Datasets:

marin-community
/

ar5iv-no-problem-markdown

Dataset card Data Studio Files Files and versions Community

Split (1)

train · 366k rows

id stringlengths 30 36	source stringclasses 1 value	format stringclasses 1 value	text stringlengths 5 878k
no-problem/9907/astro-ph9907214.html	ar5iv	text	# 1 Introduction ## 1 Introduction The field of Very High Energy (VHE) gamma-ray astronomy has exploded in recent years, mainly pushed by the development of more sensitive telescopes. The emphasis has been to lower energy thresholds, improve angular and energy resolutions and most importantly hadronic cosmic ray background rejection. Considerable efforts have also been made to develop telescopes which detect VHE extensive air showers (EAS) which have secondaries that survive to ground level, such as Milagro and the Tibet Array. If reasonable sensitivity at VHE energies can be achieved with these detectors, they will offer powerful capabilities, such as full overhead sky coverage both day and night regardless of weather and skylight conditions. This would allow much higher temporal coverage of sources that are already known to be highly variable, such as Active Galactic Nuclei. The Milagro detector is progressing toward reaching the necessary VHE sensitivity. It is a large (60m x 80m x 8m) water pond instrumented with 723 8” photomultiplier tubes (pmts) in two layers. These pmts detect the Cherenkov light produced by EAS secondaries passing through the optically clear water. Its high altitude (2650m) and sensitivity to both photonic and leptonic EAS components give it an energy threshold such that for zenith traversing sources the peak primary energy will be 1 TeV. After calibrations it will have good angular resolution and hadronic cosmic ray rejection (see McCullough 1999 for more details). To improve the sensitivity of the current Milagro detector, 172 instrumented, large area ($`5m^2`$), water Cherenkov detectors (tanks) will be deployed around the pond to effectively extend its active area. As discussed below, this will improve both the energy and angular resolution of Milagro and increase its hadronic cosmic ray rejection, thus improving its overall VHE sensitivity. It can also be used to increase Milagro’s efficiency for detecting EAS below 1.0 TeV which have core positions significantly away from the Milagro pond. ## 2 Water Tank Detector & Array The criteria for selecting a detector design that will improve the performance of EAS experiments are: low cost and low maintence (a large ground area needs to be covered, typically at a remote high altitude site), high sensitivity to EAS secondary particles, and good timing and particle density resolution. Previous works (Yodh 1996) showed that water Cherenkov detectors have superior sensitivity to those of scintillation counters for detecting EAS secondaries. Thus the tank design proposed here satisfies these design criteria, although the particle density resolution is somewhat poor. On average the pmt signal is about 100 photoelectrons for a through-going vertical muon. Figure 1 displays a crossectional view of a tank showing the position of the top-mounted, downward-looking 8” pmt and the Tyvek-lined bottom, sides, and floating top. This position of the pmt gives a fairly uniform response across the full tank, although it does degrade the timing resolution somewhat compared to a bottom mounted, upward looking position. Due to its active material, water, the tank is sensitive to both the photonic and leptonic components of EAS as opposed to plastic scintillator based detectors which are mainly sensitive to the leptonic component. The Tyvek lining provides a diffusivly reflective inner surface with $`>90\%`$ reflectivity at the important wavelengths determined from convoluting the Cherenkov photon spectrum and pmt quantum efficiency (wavelengths around 350 nm). : Schematic of example water Cherenkov tank. Key features are top mounted, downward looking pmt and Tyvek lined inner surfaces (units are feet). The Milagro inspired tank array has 172 tanks placed on a square grid with a spacing of 15 m, giving a full array area of 200m x 200m centered on the Milagro pond. Monte Carlo generated data was used to determine the performance characteristics of these tanks, and the improvement of the sensitivity of the Milagro detector generated by using these tanks. Corsika was used for generating simulated EAS and the Geant package was used to simulate the tank and Milagro detector responses (see (Westerhoff 1998) for more details). ## 3 Monte Carlo Estimates of Milagro Performance Improvements The information acquired with the tanks discussed above can be used in two separate ways. First, it may improve the angular and energy reconstruction resolutions of EAS which trigger the Milagro pond detector by making additional independent shower front timing measurements and by improving the EAS core position resolution for EAS whose cores done not strike the pond. A simple multiplicity trigger condition of 50 pond pmts being hit by an EAS was used as a pond-trigger in simulations. Second, the information can be used to increase the effective area of the Milagro detector by using it in a combined pond-tank trigger. ### 3.1 Improvements in Pond Triggered Events From simulation, on average about 24 tank pmts are hit per event where a hit is the detection of 1 or more photoelectrons. The occupancy (fraction of the time a given pmt or tank is hit) for pond pmts is about 30% and for tanks is about 10%. The tanks have fewer low pulse height hits than the pond pmts (below 30 pes) but about the same number of large pulseheight hits (above 30 pes). As seen in Figure 2, our simulations predict that using the tank array in reconstructing EAS core positions can improve the position resolution tremendously for EAS whose core positions are off the pond. This improvement is crucial for EAS energy determination and is also important in EAS angle determination because the pmt hit times must be corrected for EAS shower front curvature about the core position. Current ongoing studies of Monte Carlo generated EAS show that a good core position resolution should improve the hadronic cosmic ray rejection capabilities of Milagro as well. : Plot of median core position error versus core distance from center of Milagro pond using tanks (circles) and not using tanks (crosses). Improvements in the angular reconstruction resolution of EAS is displayed in Figure 3. The improvement is maximal for low multiplicity (number of pmts in Milagro which detect light) events which are typically low primary energy EAS. It is also maximal for EAS whose cores land far from the pond. : Plots of median angle error versus number of pmt hits in top layer of Milagro pond and versus EAS core distance from center of Milagro pond. Circles are values when tanks are used and crosses are value when tanks are not used. ### 3.2 Improvements in Trigger Sensitivity Including tank acquired information within the Milagro trigger condition can increase the efficiency for seeing low energy events. This is clearly seen in Figure 4 which displays a plot of the effective area of the Milagro detector for three types of triggers. The pond-only trigger is a requirement that at least 50 pmts be hit by the EAS. The tank+pond trigger is that either the pond trigger be satisfied or that at least 5 tanks be hit by the EAS. The tank-only trigger is that at least 5 tanks be hit and that less than 50 pond pmts be hit by the EAS. The tank-only trigger is included to explicitly show the contribution of the tanks to the effective area. : Plot of effective area versus energy. Dashed is pond-only trigger, dotted is tank-only trigger, and solid is pond-tank trigger. Those events obtained by using the tanks in a trigger condition have an average pond pmt multiplicity of 20 and an angular resolution of about $`2.5^o`$. This resolution is significantly worse than the resolution of pond-triggered events ($`<1.0^o`$) but is adequate for doing coincident searches with most BATSE-detected Gamma Ray Bursts and for photon counting analyses where the event angles are not used. ## 4 Results from a Prototype Array A prototype tank array has been installed near the Milagro pond to study the response of the water tanks to typical EAS that trigger Milagro. The array consists of 11 tanks built with commercially available polyethylene storage tanks. The installed pmts are of the same type as those in the Milagro pond (Hamamatsu R5912). The tanks are at various distances from the pond which will enable us to study their response as a function of EAS core distance and particle density. The tank hit multiplicity with at pond trigger requirement of approximately 120 hit pond pmts is 2.5. Results from these prototypes will also be presented. ## 5 Summary From the above simulation results one can see the predicted large improvement to both the angular and core position resolutions of the Milagro detector using information acquired by a spare array of instrumented Cherenkov water tanks. This improvement is mainly for EAS whose cores do not fall directly on the Milagro pond. Since the sensitivity of an VHE detector is proportional to its angular precision, this improvement will have a large positive effect on Milagro’s sensitivity. The greatly improved core position resolution will increase Milagro’s sensitivity to various source spectral characteristics. This research was supported in part by the National Science Foundation, the U. S. Department of Energy Office of High Energy Physics, the U. S. Department of Energy Office of Nuclear Physics, Los Alamos National Laboratory, the University of California, the Institute of Geophysics and Planetary Physics, The Research Corporation, and CalSpace. References McCullough, J.F. 1999, HE 6.1.02, these ICRC procedings. Westerhoff et al. 1998, 19th Texas Symposium on Relativistic Astrophysics procedings. Yodh, G.B. 1996, Space Science Reviews, 75: 199-212.
no-problem/9907/astro-ph9907305.html	ar5iv	text	# 1 INTRODUCTION ## 1 INTRODUCTION In a recent paper Ciardullo et al. (1999) resolved the companions (or possible companions) of 19 planetary nebulae (PNs) nuclei. By assuming that the resolved companions are main sequence stars, they calculated the distances to these PNs. The projected orbital separations of the binary nuclei are in the range of a few hundred to several thousand astronomical units. At such large orbital separations the companions will not influence the mass loss from the PN progenitors. However, the companions can influence the structures of the descendant PNs. For orbital periods of $`P_{\mathrm{orb}}10\tau _{PN}`$, where $`\tau _{PN}`$ is the formation time for the nebula, the orbital motion may cause a departure from axisymmetrical structure (Soker 1994), while for larger separations the companion may blow a small bubble inside the nebula (Soker 1996). Such a bubble, formed by a very wide companion, might be the “vertical bridge” observed by Corradi et al. (1999) near a star located in Wray 17-1. The list of binary companions presented by Ciardullo et al. allows a qualitative comparison between the properties of the binary nuclei and the observed structures of these PNs. This is the aim of this paper: to search for a consistency between the claimed binary nuclei and the PNs morphologies, thereby to support the arguments of Ciardullo et al. (1999) for an association, or not, between the observed companions and the central stars of the PNs. As an example, let us examine the Stingray nebula, for which a visual companion was resolved by the Hubble space telescope near its central star (Bobrowsky et al. 1998). The projected orbital separation is $`2200\mathrm{AU}`$, while the outer nebular radius is $`8000\mathrm{AU}`$. This radius implies a nebular expansion age of $`\tau _{PN}4000`$ years for an expansion velocity of $`10\mathrm{km}\mathrm{s}^1`$. A companion at such a large separation cannot be responsible for the axisymmetrical structure of this PN. Its orbital separation, which implies an orbital period of $`10^5\mathrm{yrs}`$, may manifest itself in a departure from axisymmetry. During the time of $`\tau _{PN}4000`$, the mass-losing star would have moved a distance of $`0.5\left(4000/10^5\right)\times 2\pi \times 2200\mathrm{AU}275\mathrm{AU}`$ in its orbital motion (assuming the two stars have equal masses). This is $`3\%`$ of the nebular radius, and it implies a clear deviation from axisymmetry. This PN, indeed, has a large degree of deviation from axisymmetry, part of which may be attributed to the presence of this companion. However, this PN has a complicated structure, including a dense ring which is inclined to the long axis of the nebula. The overall degree of deviation from axisymmetry seems to me to be too large to be solely explained by the visual companion, and I hypothesize that a closer companion in an eccentric orbit may exist. The visually resolved companion’s wind may also have shaped the nebula in its immediate vicinity. The Stingray nebula deserves further observations. The paper is organized as follows. In $`\mathrm{\S }2`$ I find the masses of the main sequence companions, and calculate the orbital periods and velocities. I then mark each PN according to whether its structure is compatible or not with the properties of its central binary system, and in $`\mathrm{\S }3`$ I elaborate on the individual PNs. In $`\mathrm{\S }4`$ I compare the sample studied in $`\mathrm{\S }3`$ to the sample of PNs for which Ciardullo et al. found no companions. A short summary is in $`\mathrm{\S }5`$. ## 2 PROPERTIES OF BINARY NUCLEI In Table 1 the same 19 PNs that are given by Ciardullo et al. (1999) in their table 7 are listed. The columns in Table 1 are: (1) The common PN name. (2) The PN G designation (galactic coordinates). (3) The secondary mass in $`M_{}`$. It is calculated from the magnitude $`M_V`$, which is found from the colors (V-I) and reddening \[E(B-V)\] as given by Ciardullo et al. Only for K 1-27 was the companion’s mass taken to be $`0.6M_{}`$ since Ciardullo et al. argue that the companion is a WD rather than a main sequence star. (4) The projected orbital separation in AU. The values are taken from Ciardullo et al., where more details can be found. (5) The orbital period, in years, calculated by taking the progenitor of the PN (now the WD) to be of mass $`0.8M_{}`$. The real orbital separation is likely to be larger than the projected orbital separation. An effect in the opposite direction is the mass-loss process during the AGB phase. The orbital separation increases during this phase, and hence it was smaller at the beginning of the process. The projection effect is likely to dominate in most cases, hence the orbital periods are likely to be longer than those given in the table. (6) The orbital velocity of the primary around the binary center of mass, taking the projected orbital separation and assuming a circular orbit and a mass of $`0.8M_{}`$ for the WD progenitor. (7) The nebular diameter as given by Ciardullo et al. (1999), in parsecs. (8) The classification of Ciardullo et al. (1999; Table 7): P$``$Probable binary; S$``$Possible binary; D$``$Doubtful. (9) Departure from axisymmetry: Y$``$a departure from axisymmetry is clearly seen; N$``$no clear departure from axisymmetry is seen in the images. By departure from axisymmetry, I refer only to large-scale departures, and not to small-scale blobs, filaments etc. (10) My estimate of the compatibility of the PN departure from axisymmetry and the possibility that its binary companion caused this departure: ($`+`$)Compatible; $`\left(\mathrm{?}\right)`$hard to tell. For example, for A 30 no departure is expected since the orbital period is very long, and indeed no large-scale departure from axisymmetry is observed, hence the $`+`$ sign. For NGC 7008 a deviation from symmetry is expected and is indeed observed, hence a $`+`$ sign. In K 1-14 a departure is observed, but the main signature is on the outskirts of the nebula, and hence an interaction with the ISM is possible. If the companion is associated with the PN central star, as claimed by Ciardullo et al., then it can also cause a departure. Hence for this PN I put a question mark, since it is not clear if the morphology is compatible with the claimed companion, or it is solely due to an interaction with the ISM. Four main processes can cause a large-scale deviation from axisymmetry. ($`i`$) Interaction with the ISM. In this case the most prominent features are on the outskirts on the nebula (e.g., Tweedy & Kwitter 1994, 1996), with smaller, or none at all, deviations from axisymmetry in the inner regions of the nebula. It is worth noting that in large PNs the ISM may penetrate the outer regions of the nebula, and influence the inner structure as well as the outer regions (Dgani & Soker 1998). ($`ii`$) A close companion in an eccentric orbit (Soker, Rappaport, & Harpaz 1998). This occurs when the companion is close enough to influence the mass loss process from the AGB star, and the eccentricity is large. ($`iii`$) A wide companion, with an orbital period in the range of $`0.3\tau _{PN}P_{\mathrm{orb}}30\tau _{PN}`$, where $`\tau _{PN}`$ is the formation time of the relevant part of the nebula. The formation time can be $`\tau _{PN}\mathrm{several}\times 10^4\mathrm{yr}`$ for a halo, $`\tau _{PN}\mathrm{several}\times 10^3\mathrm{yrs}`$ for the dense inner shell, and $`\tau _{PN}\mathrm{several}\times 100\mathrm{yrs}`$ for possible jets. For most of the PNs listed in Table 1, the relevant time is the expansion time of the nebula to its current radius $`\tau _{PN}R_{PN}/10\mathrm{km}\mathrm{s}^110^45\times 10^4\mathrm{yrs}`$. Another requirement of the latter mechanism is that the velocity of the mass-losing star around the center of mass $`v_1`$ will be not too small. This means a companion of mass $`M_20.3M_{}`$, depending on the orbital separation (see $`\mathrm{\S }4`$). ($`iv`$) During the mass loss process itself, if one or a few long live cool (or hot) spots exist on the surface of the AGB star. This process seems to be important to massive stars (e.g., as suggested for the $`30M_{}`$ star HD 179821 by Jura & Werner 1999), but it is not clear if it can operate efficiently in AGB stars, where the strong convection may not allow such spots to live long enough. Simple estimates suggest that the deviation from axisymmetry, e.g., the dislocation of the central star from the center of the nebula, will be of the order of $`v_1/v_e`$, where $`v_e`$ is the expansion velocity of the nebula. However, interaction of the wind with previously ejected mass from the AGB star may change the density of the wind in a different sense on both sides, and because of the sensitivity of emission to the density, a larger degree of departure from axisymmetry will be observed. The observed degree can be lower than expected if the direction of deviation is not perpendicular to the line of sight; if it is along the line of sight, the departure from axisymmetry will not be observed. In most cases I expect the degree of axisymmetry to be in the range from one to several times $`v_1/v_e`$. In the next section I discuss the structure of each PN, and show that in most cases the PNs’ structures can be used to strengthen the claims of Ciardullo et al. regarding the association, or non-association, of the putative companions with the central stars of the PNs. ## 3 NOTES ON THE INDIVIDUAL NEBULAE In this section I analyze each PN separately. I will not discuss the position angle of the companion versus the direction of the departure from axisymmetry. This is because there are no details on the orbit (e.g., velocity), and the current position of the companion has little information about the orbit inclination, eccentricity, etc. In analyzing the different PNs, I use images from Balick (1987), Schwarz, Corradi, & Melnick, (1992), Acker et al. (1992), Manchado et al. (1996), and other sources as indicated below. K1-14. There is a clear deviation from axisymmetry. The central star’s distance from the center of the nebula $`\mathrm{\Delta }r`$ is $`15\%`$ of the nebular radius, i.e., $`\mathrm{\Delta }r/R_{PN}0.15`$. For an expansion velocity of 15 $`\mathrm{km}\mathrm{s}^1`$ the age of the nebula is $`2\times 10^4\mathrm{yrs}`$, just a little shorter than the orbital period. The velocity of the progenitor around the center of mass is relatively high, $`0.5\mathrm{km}\mathrm{s}^1`$, and it can explain the departure from axisymmetry. Therefore, for this nebula the departure from axisymmetry may be attributed to the presence of the companion. However, this is not the only possibility. The central star is closer to the north-east part of the nebula, where a brighter arc is seen. This structure hints at an interaction with the ISM. Hence I put a question mark for this nebula (10th column in Table 1). Future observations will have to examine in more detail the properties of the companion and its association with the central star. A 63. No clear deviation from axisymmetry is observed. This suggests that either the “companion” is not associated with the progenitor of the nebula (Ciardullo et al. marked it as a possible association), or that the real orbital separation is larger than the projected orbital separation, hence the orbital period is $`30`$ times the age of the nebula. In any case, the absence of a clear departure from axisymmetry is compatible with the properties of the orbit even if the observed companion is a member of a binary system. NGC 7008. A clear departure is observed, both in the outer and inner regions. This is expected, since the mass-losing star’s orbital velocity is large, $`1.9\mathrm{km}\mathrm{s}^1`$, and the orbital period, although short, is within an order of magnitude of the formation time of this small nebula. NGC 650-51. A clear asymmetry is seen along the waists of this bipolar PN. It cannot be attributed to an interaction with the ISM. The orbital velocity of $`0.26\mathrm{km}\mathrm{s}^1`$ is too low to explain the large departure in the equatorial plane. The only way in which this “companion” could have caused the deviation is that it has a very eccentric orbit and currently it is close to its apastron position. In this case the orbital period is shorter, and when the companion was closer the orbital velocity of the mass-losing star was much higher. Although plausible, this is an unlikely explanation. More likely, a closer companion, but still at a wide orbital separation, exists, or a close companion which directly affects the mass-loss process orbits the central star with a highly eccentric orbit (Soker et al. 1998). I marked this PN with a question mark. PuWe 1. The images of this PN is from its discoverers, Purgathofer & Weinberger (1980), and Tweedy & Kwitter (1996). This PN shows a clear departure from symmetry, but with a bright arc on the outskirts of the nebula. This clearly suggests an interaction with the ISM, as was argued by Tweedy & Kwitter (1996), rather than a wide binary companion influence. This is compatible with the claim of Ciardullo et al. that the association of the observed “companion” with the central star is doubtful. Hence I marked it with a +. NGC 2392. A small but clear departure is observed along the south-north direction, which is the long axis of the nebula, in the outer as well as the inner parts. Along the minor axis, the eastern side is longer by several percents than the western side. Such a deviation could not be caused by a companion at such a large orbital separation. However, the distance found by Ciardullo et al., $`6.41\mathrm{kpc}`$, is an upper limit. Most methods find a distance of $`0.51.5\mathrm{kpc}`$ (Acker et al. 1992). Taking a distance to NGC 2392 to be $`1.2\mathrm{kpc}`$, the orbital period is $`130,000\mathrm{yr}`$, the orbital velocity is $`v_10.5\mathrm{km}\mathrm{s}^2`$, and the companion will be less massive. These parameters for the companion and orbital motion can explain the observed departure from axisymmetry. Therefore, the morphology of this PN suggests that the companion is associated with the central star. Hence I marked it with a + sign, despite the classification of “possible association” by Ciardullo et al.. Another possibility is discussed together with NGC 1535 below. NGC 1535. A small but clear asymmetry is observed along position angle $`220^{}`$. Such a small departure is compatible with the parameters of the companion. Another possibility, for both this PN and NGC 2392, is that these are elliptical PNs with their long axis almost along the line of sight. The two sides are not identical due to an unequal collimated flows on the two sides of the long axis, which are not related to the presence of the companion. A study of the expansion velocities of these PNs along different directions is needed. A 30. This PN has a very small departure from sphericity on its outskirts, which can safely be attributed to the ISM. No large-scale deviation is observed in the inner region, only blobs and filaments, hence I marked it to have no real deviation from axisymmetry. This is compatible with the very long orbital motion and low velocity. A 7. This a large nebula, which does not show a bright arc, is located far from the galactic plane and is surrounded by a tenuous ISM, hence it does not seem to interact with the ISM (Xilouris et al. 1996; Tweedy & Kwitter 1996). The distance to this PN is $`0.2\mathrm{kpc}`$, much smaller than the upper limit given by Ciardullo et al., for which the orbital separation is $`200\mathrm{AU}`$ and the orbital period $`2,500\mathrm{yr}`$. Such a binary should cause a noticeable departure. Although the nebula has an almost circular outer boundary, its inner region is elliptical, with unidentical bright regions on both sides. However, this is an old and large PN, and it is not clear whether this small deviation from axisymmetry is related to the presence of a companion, to instabilities which had time to grow, or to a weak interaction with the ISM. Therefore, I marked it with a question mark. A 24. This PN has a point symmetric structure. The western side is a little more extended. I marked it as having a questionable large-scale deviation from point symmetric. Although the orbital period is long, this is a large PN, hence old one. Tweedy & Kwitter (1996) argue that it does not interact with the ISM. For all these, it is hard to tell whether its structure is compatible with the companion properties. A 31. The image is taken from Tweedy & Kwitter (1994), who find that this PN interacts with the ISM. The morphology of this PN clearly comes from an interaction with the ISM. Any influence by the companion will be much smaller than the effect of the ISM. Therefore, I marked this PN with a question mark. A33. A large deviation from axisymmetry, which cannot be attributed to the ISM since there is a deviation in the inner regions, and no bright arc is observed on the side closer to the central star. The distance of the star from the center of the nebula is $`\mathrm{\Delta }r/R_{PN}0.07`$. The departure of this PN is compatible with the properties of the companion, as already noted in a previous paper (Soker 1997). NGC 6210. Ciardullo et al. find a distance smaller than $`32\mathrm{kpc}`$. Most other methods give distances of $`2\mathrm{kpc}`$, i.e., a factor of 16 smaller. This means an orbital period $`64`$ times shorter, and an orbital velocity of $`0.8\mathrm{km}\mathrm{s}^1`$. Such a binary companion should cause a noticeable deviation from axisymmetry. This PN shows a small deviation from axisymmetry along it major axis. Since the deviation is not so clear, and Ciardullo et al. argue for a possible association between the companion and central star, I marked it with a question mark. NGC 3132. This PN shows a clear asymmetry along position angle $`45^{}`$. In that direction the nebula extends to a distance smaller by $`20\%`$ than the distance in the opposite direction. Such a large departure from axisymmetry is expected from a companion having the listed properties. K 1-22. Although the image in Acker et al. (1992) is very faint, a small deviation from axisymmetry is observed in the inner region, more or less along the minor axis of the PN. This is compatible with the properties of the companion given in Table 1. K 1-27. In Soker (1997) I suggested that the morphology indicates an interaction with the ISM. The finding of a companion by Ciardullo et al. suggests that it is the companion that causes the deviation from axisymmetry, or at least plays an important role in it. For this PN Ciardullo et al. argue that the companion is a WD. Hence I took its mass to be $`0.6M_{}`$. Mz 2. A clear asymmetry is observed: the two outer lobes are slightly displaced to the east relative to the inner bright shell, and a “missing” segment in the shell is observed at position angle $`220^{}`$. This departure is compatible with the finding of Ciardullo et al.. Sp 3. In a previous paper (Soker 1997, Table 4) I noticed that the morphology of this PN suggests that it interacts with the ISM or contains a wide binary companion. The findings of Ciardullo et al. show that it is the companion that causes the departure from axisymmetry. IC 4637. A departure from axisymmetry is observed, but it may come from an interaction with the ISM. I therefore marked it with a question mark. ## 4 PNs FOR WHOM NO COMPANIONS WERE FOUND The purpose of the present paper is to demonstrate how the morphology of a PN can be used to strengthen (or not) the claim for the presence of a wide companion. However, in many cases a deviation from axisymmetry will be observed, but no companion will be found. These cases are: (1) The deviation is caused by the motion of the PN through the ISM. I discussed this case in previous sections. Below a question arises in regard to A66 which has a clear asymmetry (Hua, Dopita & Mertinis 1998). However, this is a large low density nebula, and therefore the asymmetry is most probably due to the ISM. (2) The projected separation of the companion and the PN central star is too small to be resolved, e.g., in the snapshot survey of Ciardullo et al., the resolution with the HST is $`0.1`$ arcsec. This will happen if the orbital separation is small, or if the orbital separation is large but the binary plane is almost edge-on, and the companion line of sight happens to be closed to that of the central star. Note that to observe a deviation from axisymmetry the orbit must not be face on, so in many cases the orbit will be close to being edge-on. (3) The companion is a low mass main sequence star, and it is too faint to be detected. In Table 1, 18 out of 19 companions, or possible companions, have a mass of $`0.5M_{}`$. Therefore, we can safely deduce that the study of Ciardullo et al. is limited mainly to main sequence companions of masses $`0.5M_{}`$ (we return to this point below). (4) The deviation from axisymmetry is due to a close companion at an eccentric orbit (Soker et al. 1998). Such systems are likely to form Bipolar PNs. (5) A companion that enters the envelope of the progenitor of the PN may cause a deviation from axisymmetry as well. This has been suggested as a possible explanation for the displacement of the two outer rings of SN1987A (Soker 1999). Regarding mechanisms (4) and (5), a clear deviation from axisymmetry is seen in the bipolar PN NGC 2346, for which there is a very close companion with an orbital period of 16 days (Bond & Livio 1990). (6) In the survey of Ciardullo et al. PNs located in a high field star density were also classified as non-detection, since a wide binary companion could not be statistically identified in these cases. It is possible therefore, that the companions of some of the non-detection PNs which have nebular departure from axisymmetry are seen in the survey, but can not be statistically distinguished from the field stars. With these possibilities in mind, I examine the PNs from the snapshot survey of Ciardullo et al. for which no companions have been detected. Here again, I consider a PN to have a clear departure from axisymmetry when the scale of departure is larger than the typical size of blobs and filaments in the PN. The list of 67 PNs and their properties was kindly supplied to me by R. Ciardullo. In this list there are 4 PNs which are classified as Bipolar, and are likely to be formed from a close binary interaction (Soker 1997, Table 3). Considering point (4) above, I do not consider these PNs. Out of these 67 PNs, 13 PNs were claimed by Soker (1997, Table 4) to be descendants of common envelope evolution. Following point (5) above, I do not consider these PNs. Out of the rest 50 PNs, The deviations from axisymmetry of 15 PNs were noted by Soker (1997) as follows. For 9 PNs the deviation from axisymmetry seems to result from interaction with the ISM; two out of these 9 show no deviation in the inner regions, and therefore have no influence of a wide companion (A16; NGC 6894), while for the other 7 it is hard to tell (due to, e.g., an extended filamentary structure). For NGC 5979 Soker (1997) left it open whether the departure from axisymmetry is due to an interaction with the ISM or to a wide companion. IC 4593 show both signs of interaction with the ISM and a wide companion. For 4 PNs Soker (1997) attributes the departure to a wide companion (IC 5148-50;NGC 2022; NGC 3242; NGC 7662). In these PNs the departure from axisymmetry is most prominent in the inner regions, and therefore can be caused by a companion with a relatively short orbital period. Soker, Zucker, & Balcik (1992) attributed the departure from axisymmetry of NGC 3242 to a star with an orbital period of $`4,000\mathrm{yrs}`$, but a shorter period is also possible. A shorter orbital period is possible for NGC 2022, for which the deviation is of the inner “jets”. A careful examination of images in the literature reveals 3 other PNs with a clear departure from axisymmetry (NGC 6153, Gorny et al. 1999; NGC 5882, NGC 6804 Schwarz et al. 1992), and 2 with departure which I cannot tell if compatible with a wide companion (NGC 6891, Chu, Jacoby & Arendt 1987; NGC 2792, Schwarz et al. 1992). For 23 PNs out of the 50 I consider, I could not tell whether they have a departure from axisymmetry which is compatible with a wide companion (e.g., they have interaction with the ISM; they have filamentary structure with small number of large filaments; no high resolution image exists). Out of the rest 27 PNs, 8 show clear departure, and for 3 it is not clear (these 11 PNs were discussed in the previous paragraph). I conclude that $`35\pm 5\%`$ of PNs for whom no companions were found by Ciardullo et al., show departure from axisymmetry compatible with the presence of a wide stellar companion. This should be compared to the 10 PNs with probable detection, from which 8 have departure compatible with the claimed companion (both ’Y’ and ’$`+`$’ in Table 1). Considering the detection limits, and in particular the companion mass of $`0.5M_{}`$, I find the differences in detection, $`80\%`$ compared with $`35\%`$, significant, despite the small number of PNs. I return now to the question of the limit on the companions masses. I used the I-band limiting magnitude (supplied by Ciardullo) of the 8 PNs I claim to have wide companions and the 3 PNs with question marks, to estimate the upper limit on the masses of possible main sequence companions. I took distances from Cahn, Kaler & Stanghellini (1992), and the mass-luminosity relation for main sequence stars. Because of the strong dependence of luminosity on mass in that relation, the calculation is quite robust to the uncertainties in the different quantities entering the calculation. The uncertainties in the distances may be up to a factor of $`3`$, which introduce uncertainties of up to $`70\%`$ in the limiting masses. Because of these uncertainties I do not find it appropriate to list each object separately (unlike the case of the PNs presented in Table 1, for which Ciardullo et al. used the companions to determine the distances). Overall, I find the limiting masses to be around $`0.5M_{}`$ for most objects. This is compatible with the masses of detected companions (Table 1). A detailed statistical study of binary systems that cause departure from axisymmetry using a population synthesis code, is planned for a forthcoming paper. The code that will be used is the one used by Rappaport & Soker (1999), from which I take the different values used below. I crudely estimate the number of expected wide binaries as follows. For an orbital separation smaller than $`1\mathrm{AU}`$ (an orbital period of one year), a strong interaction of the binary system will form a bipolar PNs, or the system will enter a common envelope. I exclude such systems here. For the rest, the distribution of binary orbital period is uniform in $`\mathrm{log}\left(P\right)`$, up to $`P10^6\mathrm{yrs}`$. The orbital velocity of the mass-losing star, of mass $`M_1`$, around the center of mass with a companion of mass $`M_2`$, is (assuming a circular orbit) $`v_1=0.4\left({\displaystyle \frac{P}{10^4\mathrm{yr}}}\right)^{1/3}\left({\displaystyle \frac{M_1+M_2}{1M_{}}}\right)^{2/3}\left({\displaystyle \frac{M_2}{0.3M_{}}}\right)\mathrm{km}\mathrm{s}^1.`$ (1) As mentioned before, an orbital velocity of $`0.3\mathrm{km}\mathrm{s}^1`$ may cause a noticeable departure from axisymmetry. The mass distribution in binary systems is somewhat peaked toward $`M_2/M_1=1`$, hence most companions of PN progenitors will have masses of $`M_20.3M_{}`$. This means that binaries with orbital periods of up to few$`\times 10^5\mathrm{yrs}`$ can be counted, especially if eccentric orbits are considered as well. From below, we require that the orbital period will not bee too short, this gives an orbital period longer than $`500\mathrm{yrs}`$. Overall, most companions in a logarithmic interval of $`2`$, which is $`1/3`$ of all binary systems may cause deviation from axisymmetry. The question is what is the fraction of binary systems among all progenitors of PNs. This fraction is taken by different authors to be in the range of 0.6-1 (Rappaport & Soker 1999). I conclude that wide binaries will cause a clear departure from axisymmetry in $`2035\%`$ of all PNs. The fraction of PNs below the detection limit mentioned above, I crudely estimate to be half this number $`1017\%`$. Adding to these some interaction with the ISM (e.g., a dense cloud), and large scale instabilities in the mass loss process, which I interpreted as wide binaries in analyzing the PNs in this section, I find the fraction of PNs with departure from pure axisymmetry among the non-detection PNs ($`35\%`$) to be quite reasonable. ## 5 SUMMARY Stellar companions to progenitors of PNs will influence the morphology of the descendant PNs. Close companions will affect directly the mass loss process from the AGB progenitor. Such close binaries are not likely to be resolved. Very wide binaries (as well as the wide binaries discussed here) may form a small bubble inside the nebula (Soker 1996). In the intermediate range, binary systems which have orbital periods in the range of several$`\times 100`$ to a few$`\times 10^5`$ years, may cause the PN to have a large-scale departure from axisymmetrical structure. The goal of the present paper was to show that the structure of a PN can be used to strengthen an argument for an association, or a non-association, of a putative wide companion with the progenitor of the PN. In a recent work Ciardullo et al. (1999) list 19 PNs for which they argue a probable, possible, or doubtful association of stars with the progenitors of the PNs. Ciardullo et al. did not use the structure of the PNs as a tool to further support their claims. In the present paper I used their list, and demonstrated that the structures of the PNs are compatible in most cases with their claims, hence this paper strengthens them. The greatest confusion may come from a possible influence of the ISM on the PN morphology, especially for large PNs. I tried to present arguments for and against ISM interaction where possible. In the opposite direction, PNs whose central stars have companions can be used to further study the influence of wide companions on the PNs morphology. For this, a determination of the binary component velocity is needed. Very helpful but more difficult to determine will be the eccentricity and inclination of the orbital plane. From a list of 27 PNs observed with HST for which no companions were found by Ciardullo et al., I claimed that 8 PNs are likely to have companions which have caused their structures to depart from axisymmetry. In addition to 5 PNs from this list which were predicted to have wide binary companions by Soker (1997), I added here to the list three PNs: NGC 5882, NGC 6153, and NGC 6804. I estimated that the observations of Ciardullo et al. will not detect main sequence companions of masses below $`0.5M_{}`$ in these 8 PNs. A search for a wide binary companion in each of these 8 PNs is encouraged. For three PNs out of the 27 I could not tell if a companion is compatible or not with their structure. That only 8 of the 27 PNs have morphology compatible with a wide binary companion, compared with 8 out of 10 PNs that were claimed to have probable companion by Ciardullo et al., strengthen the arguments presented in the present paper. ACKNOWLEDGMENTS: I would like to thank the referee, Robin Ciardullo, for his suggestion to study the sample of PNs for whom no companions were found, and for supplying the data for these PNs, and Matthew Bobrowsky for a careful reading of the manuscript. This research was supported in part by a grant from the University of Haifa and a grant from the Israel Science Foundation.
no-problem/9907/math9907057.html	ar5iv	text	# References
no-problem/9907/hep-ph9907410.html	ar5iv	text	# Particle and Nuclear Physics with High Energy Leptons ## 1 Introduction The space-time charge resolution of the leptonic probe depends upon the following three kinematical variables: $`x_{Bj}`$ \- the invariant Bjorken variable, $`Q^2`$ \- the invariant 4-momentum of the exchanged photon ($`W`$,$`Z`$ boson) and $`s`$ \- the centre-of-mass energy. These variables can be expressed in terms of the incoming and outgoing lepton 4-momenta $`k`$ and $`k^{^{}}`$ and by the 4-momentum of the nucleon $`p`$: $`Q^2=q^2=(kk^{^{}})^2`$ $`x_{Bj}=Q^2/(2pq)`$ $`s=(p+k)^2`$ In the plane perpendicular to the collision axis the constituents of hadronic matter carrying electro-weak charges can be resolved within the distances of $`l_t1/Q`$. The corresponding space-time resolution in the longitudinal direction is determined by the value of the invariant Bjorken variable - $`x_{Bj}`$ and by the value of the reference frame-dependent Lorenz-factor of the nucleon (nucleus) - $`\gamma `$: $`l_lt1/(2\gamma Mx_{Bj})`$. In high energy collisions $`l_l`$ and $`t`$ are strongly correlated: $`(tl_l)1/s`$ simplifying the large-distance structure of hadronic matter to “frozen configurations”. Strong interactions provide natural scales for the resolution of transverse distances: $`1/\mathrm{\Lambda }_{QCD}`$ and the inverse mass of the $`\rho `$ meson, $`1/m_\rho `$. Using these yardsticks three kinematical regions can be defined: * the photo-production region (PR) ($`Q^2<\alpha _1\mathrm{\Lambda }_{QCD}^2`$ ) where the large-distance structure of hadronic matter is of importance and where the quasi-real photon interacts with the hadronic matter predominantly via the vector-meson component of its wave function. * the transition region (TR) ($`\alpha _1\mathrm{\Lambda }_{QCD}^2<Q^2<m_\rho ^2/\alpha _2`$ ) where a direct coupling of the photon ($`W`$, $`Z`$ boson) to a charged parton becomes important * the deep inelastic region (DIS) ($`Q^2>m_\rho ^2/\alpha _2`$) where the direct coupling of the photon to a charged parton dominate These kinematical regions have rather fuzzy boundaries corresponding to $`\alpha _1`$ and $`\alpha _2`$ values in the range of approximately 0.1-1.0. The natural scales of the resolution of the longitudinal distances in electron-nucleon and electron-nucleus collisions are: the size of nucleons $`R_N`$ and size of nuclei $`R_A`$ defined here in a Lorenz-invariant way as distances over which the valence quarks of the nucleon (nucleus) are localised. These values define three kinematical regions: * large $`x_{Bj}`$ region ($`x_{Bj}>1/(2MR_N)0.1`$) where the the photon (W,Z boson) interaction with the nucleon (nucleus) is localised within the longitudinal distances smaller than the nucleon size * intermediate $`x_{Bj}`$ region ($`1/(2MR_A)0.01<x_{B_j}<1/(2MR_N)0.1`$) where the photon interacts coherently within the longitudinal distances exceeding the size of the nucleon * small $`x_{Bj}`$ region ($`x_{Bj}<1/(2MR_A)0.01`$ ) where the light-cone-coherent interaction of the photon extends over the longitudinal distances exceeding the size of the nucleus The $`(x,Q^2)`$ region accessible to the HERA and earlier fixed target experiments is shown in Fig. 1. The main distinction between the fixed target and the collider kinematical domains, resulting from the different centre-of-mass energies, is that in the latter case the small $`x_{Bj}`$ region can be studied in DIS regime where the relevant hadronic degrees of motion are quarks and gluons. In addition, the HERA collider experiments extend the measured DIS region to the $`Q^2>M_W^2`$ values where the Neutral-Current and Charged-Current processes are of comparable strength. The small $`x_{Bj}`$, large coherence length, DIS processes can be viewed in two equivalent and complementary ways. In the Bjorken reference frame in which the nucleon (nucleus) moves with asymptotically large momentum the photon can be considered as colliding with delocalised partons, which are described by the light-cone wave function of the nucleon (nucleus). In this picture nuclei can be considered as sources of variable ($`A^{1/3}`$-dependent) strength of the colour fields. The quark and gluon interaction dynamics is expressed in this frame in terms of effective partonic densities in the nucleon (nucleus). In the rest frame of the nucleon (nucleus) the small $`x_{Bj}`$ DIS collisions can be viewed as coherent scattering of various quark-gluon Fock components of the virtual photon wave function. In this picture nuclei can be considered as effective ($`A^{1/3}`$-dependent) filters of various Fock projections of the virtual photon wave function. The dynamics of quark and gluon interactions is expressed in this frame in terms of the cross sections for interactions of $`q\overline{q}`$, $`q\overline{q}g`$ and other Fock states of the virtual photon with the nucleon (nucleus). ## 2 Comments on the present HERA program The experimental program carried out presently at HERA is more exciting than one might think while reading the papers and listening to the conference presentations of the results approved for public release. The relevance of the HERA electron-proton scattering program and the importance of its extension to electron-nucleus collisions for studies of strong interactions can be seen more clearly if one re-formulates research goals from a perspective which is different from the HERA orthodoxy. The change of perspective boils down to using the QED processes and the hard perturbative QCD processes, which are theoretically well understood and cross-measured at $`p\overline{p}`$ and $`e^+e^{}`$ collisions, as tools to investigate the QCD-vacuum and the medium effects in large (as compared to the confinement scale) distance propagation of point-like and extended colour objects. Within such a perspective the role of what is considered to be the “physics signal” and what is considered to be the “measurement noise” are reversed. In more practical terms, instead of getting rid of such non-perturbative effects by “re-tuning” of the Monte-Carlo generators or by absorbing them into various sets of “structure functions” studies are focused on these effects. The discovery of the rise of the proton $`F_2`$ structure function at small $`x_{Bj}`$ provided experimental evidence of striking differences in “large-friction” propagation of quarks in the ”sticky” QCD medium with respect to a “frictionless” propagation of electric charges in “almost-transparent” QED media. In the published analyses of the structure function data , these effects are absorbed into the effective gluon distribution derived in the the QCD analyses of the proton structure function within the framework of the DGLAP evolution equations. The questions if such a procedure is justified down to low $`x_{Bj}`$ and if the medium dependent effects can be neglected are open. Changing the medium in which the photon couples to partons (replacing nucleons by nuclei) provides the most straightforward way to factorise out the medium effects and to verify the applicability of the DGLAP equations to the large density partonic systems. One of the most striking medium effects observed at HERA is the difference in the fragmentation of the quarks ejected from the proton in DIS processes with respect to the fragmentation of quarks produced in $`e^+e^{}`$ annihilation processes . The Breit-frame spectrum of charged hadrons produced in fragmentation of low $`x_{Bj}`$ quarks is incompatible with the corresponding spectrum measured in the $`e^+e^{}`$ annihilation. In addition, the observed pattern of energy deposition in the direction of the current low $`x_{Bj}`$ quarks deviates significantly from that of the large $`x_{Bj}`$ quarks . These observations indicate clearly that the medium induced effects of quark-energy-loss and quark-multiple-scattering are large and can be studied quantitatively in a future electron-nucleus scattering program. The large fraction of rapidity gap events observed in low $`x_{Bj}`$ DIS scattering was unexpected and remains a mystery if one believes in the validity of the DGLAP evolution equations in this region. The conventional way of analysing these events within the Regge phenomenology boils down to absorbing rather than explaining the source of these events into the pomeron and reggeon structure functions. Studies of rapidity gap events in electron-nucleus scattering are indispensable in understanding universality of various de-excitations modes of the QCD vacuum. ## 3 Electron-nucleus scattering at high energies High-energy electron-nucleus collisions were discussed at several workshops , , . Physics highlights of the $`eA`$ experimental program include: * studies of the large density partonic systems and searches for nonlinear QCD phenomena * studies of partonic structure of the large distance colour singlet excitations of the QCD-Vacuum * filtering out soft from hard processes for perturbative QCD studies * understanding of colour transparency and colour opacity * studies of the space-time structure of strong interactions using the nucleus as a femto-vertex detector * studies of luminous photon-photon scattering In addition $`eA`$ collisions provide precision measurements of partonic distributions in nuclei in the $`x_{Bj}`$ range of $`10^4`$ \- $`10^1`$ and precise understanding of the high $`E_T`$ quark-jets which can be used as probes of nuclear medium. They may turn out to be vital in pinning down the quark-gluon plasma signals in the forthcoming analyses of the RHIC and the LHC-AA data. In an optimal scenario the atomic numbers of ions selected for the $`eA`$ collisions should cover uniformly the $`A^{1/3}`$ range (e.g $`D_2`$ $`O_{16}`$, $`Ca_{40}`$, $`Sn_{120}`$ and $`Pb_{207}`$) and the collected luminosities should satisfy the condition: $`L^{eA}\times A10pb^1`$/ion. The shopping list of measurements which can be made if the above criteria are fulfilled include: * $`F_2^A`$ and $`F_L^A`$ structure functions: inclusive and tagged by the number of wounded nucleons and evaporation fragments * A-dependence of the gluon distribution * Jet spectra and single particle inclusive spectra in the photo-production, DIS, and the transition regions * A-dependence of the vector meson production * A-dependence of the open charm and beauty production * A-dependence of the fraction of rapidity gap events * A-dependence of jet profiles and jet energy loss * Fragmentation spectra of tagged low-$`x`$ partons in nuclear medium, also in correlation with the observed number of wounded nucleons and evaporation fragments * Particle multiplicities and particle correlations in $`\gamma ^{}A`$ and $`\gamma A`$ scattering * Bose-Einstein correlations and their A-dependence Electron-nucleus collisions can be realized at DESY and/or at BNL. Preliminary machine studies presented by J. Maidement at the “Physics with HERA as an $`eA`$ collider” workshop show that with relatively modest investment (ion source, RFQ) isoscalar nuclei of the energies of $`410\times A`$ GeV can be accelerated and stored at HERA providing luminosities: * $`L_{eA}^{HERA}\times A=(0.51.0)\times L_{ep}^{HERA}100pb^1`$/year for electron - deuteron collisions * $`L_{eA}^{HERA}\times A1/6\times L_{ep}^{HERA}20pb^1`$/year for electron - oxygen collisions In order to collide electrons with heavier nuclei a purpose-built heavy ion preinjector system is needed. For electron - lead collisions the expected luminosity is: * $`L_{eA}^{HERA}\times A(1/201/50)\times L_{ep}^{HERA}37pb^1`$/year. At BNL, preliminary machine studies show that it is feasible to collide heavy ions with electrons by building a purpose-designed room-temperature electron(positron) ring in the RHIC tunnel. As an example, the expected luminosity for eAu collisions is estimated to be: * $`L_{eA}^{BNL}\times A=3.7\times A\times 10^{29}cm^2s^1L_{ep}^{HERA}`$ for collisions of 10 GeV electrons with $`100\times A`$ GeV ions. The above luminosities indicate that the electron-nucleus program requires at least 2-4 years of $`eA`$ collisions at HERA. It is worthwhile mentioning that the simultaneous storage of two or three types of isoscalar nuclei allows one to drastically reduce the systematic uncertainties of A dependent ratios of various observables to $``$ 1 % level. Such a running mode is considered possible by the machine experts at HERA. What could be the time-schedule for such a program? If the BNL and/or DESY $`eA`$ projects are approved by the year 2002, the first electron-ion collisions can be observed in 2005. For HERA, the eA collisions would follow the “high lumi” program which is expected to start in 2001. Two possible detector scenarios are being considered. The first consists of building a dedicated detector for $`eA`$ collisions. An alternative, cost-effective solution is to use the existing detector(s) upgraded for the $`eA`$ runs. The $`(x,Q^2)`$ coverage of the existing HERA detectors for collisions of 27.6 GeV electrons with ions accelerated to the energy of 410 Gev/nucleon is shown in Fig. 2. The $`Q^2`$ region which can be measured by these detectors is limited by the angular coverage of the central detectors. Note, that both HERA detectors provide measurements of the energies and angles of electrons and photons emitted collinearly with the incoming electron. Consequently, the photoproduction events can be tagged and radiative processes can be experimentally controlled within the present detector set-up at HERA. In Fig. 3 the pseudorapidity spectrum of particles produced in collisions of electrons with lead nuclei is shown. The present HERA detectors cover the pseudorapidity range down to $`\eta 3.5`$. Below this pseudorapidity value the produced particles (mostly nucleons and nuclear fragments) remain undetected. This region is particularly interesting both for $`ep`$ and $`eA`$ collisions and the upgrade of the existing detectors to detect these particles is indispensable for the future experimental program. ## 4 Conclusions Over the last 20 years we have witnessed impressive experimental and theoretical progress in understanding the point-like structure of hadrons and in understanding short-distance processes involving quarks and gluons. Understanding the role of quarks and gluons in large-distance, soft processes remains one of the most important challenges for the coming decade. Electro-weak probes of tunable space-time resolution applied to hadronic media of tunable colour force strength provide an “adiabatic approach-path” to large-distance processes. The experimental program of high energy electron-nucleus scattering, which can address these challenges is exciting, feasible and cost-effective. It can be realized with relatively modest modifications of the existing accelerators at DESY and/or BNL, but needs a joint-effort of the particle and nuclear physics communities. ## Acknowledgments I would like to thank all colleagues from the Department of Physics of the Oxford University and from Balliol College who made my sabbatical stay so ejoyable. I am indebted particularly to R. Devenish for his hospitality.
no-problem/9907/cond-mat9907489.html	ar5iv	text	# Observation of supercurrent enhancement in SNS junctions by non-equilibrium injection into supercurrent carrying bound Andreev states ## Abstract We report for the first time enhancement of the supercurrent by means of injection in a mesoscopic three terminal planar SNSNS device made of Al on GaAs. When a current is injected from one of the superconducting Al electrodes at an injection bias $`V=\mathrm{\Delta }(T)/e`$, the DC Josephson current between the other two superconducting electrodes has a maximum, giving evidence for an enhancement due to a non-equilibrium injection into bound Andreev states of the underlying semiconductor. The effect persists to temperatures where the equilibrium supercurrent has vanished. The study of non-equilibrium phenomena generated by current across SN boundaries of SNS junctions has attracted much interest in recent years. It has been realised that supercurrents in SNS junctions are transmitted by the correlated electron-hole pairs generated by Andreev reflections at the SN boundaries. In this paper, we demonstrate experimentally a new supercurrent enhancement effect, where these correlated electron hole pairs are affected by current injection across a third SN boundary. In general, two effects have been studied in three terminal SNS devices. One is the field effect, where the carrier density in the normal region is controlled . The other is injection, where the normal region is connected to one or two other reservoirs, biased with respect to the superconductor . This modifies the quasiparticle energy distribution, either by simply increasing the electron temperature of the normal conductor, or by inducing a suitable non-thermal distribution function. These approaches are related to much older work , where superconductivity was enhanced close to $`T_c`$ by redistributing quasiparticles by microwave irradiation or by tunnelling resulting in an enhancement of the energy gap and other properties derived from this. Recent theories for SNS junctions have predicted how injection can alter the Josephson coupling of the junction . The supercurrent in the junction is carried by bound Andreev states, which in multichannel diffusive systems form a continuum. The role of a non-equilibrium electron distribution can be seen if one considers this spectral distribution $`N_j`$ of supercurrent carrying states , together with the energy distribution of electrons $`f`$ in the normal conductor. From these the supercurrent as function of the phase difference between the superconductors can be written: $$I_s(\varphi )=2ev_FA_0^{\mathrm{}}𝑑\epsilon N_j(\epsilon ,\varphi )\left(f(\epsilon )f(\epsilon )\right)$$ (1) where $`A`$ is the cross-section of the conductor. The energy $`\epsilon `$ is defined with respect to the chemical potential of the superconductors. In equilibrium $`f`$ is the Fermi distribution function. The distribution $`N_j`$ depends on the sample geometry, and can for simple systems be calculated directly from quasi-classical Green’s function theory . In diffusive systems $`N_j`$ varies on the scale of the Thouless-energy $`\epsilon _L=\mathrm{}D/L^2`$, where $`D`$ is the diffusion constant and $`L`$ is the distance between the superconductors. For long junctions where $`\mathrm{\Delta }\epsilon _L`$ ($`\mathrm{\Delta }`$ is the superconductor energy gap), $`N_j`$ peaks at $`\epsilon \epsilon _L`$. In equilibrium the critical supercurrent decay with the normal conductor coherence length $`\xi _N(T)=\sqrt{\mathrm{}D/2\pi k_BT}`$. For long junctions $$I_c=I_0\frac{\mathrm{\Delta }^2(T)}{\mathrm{\Delta }^2(0)}\sqrt{\frac{T_c}{T}}\mathrm{exp}(L/\xi _N(T))$$ (2) where $`I_0=8\mathrm{\Delta }^2(0)L/(eR_N\sqrt{2\pi \mathrm{}Dk_BT_c})`$, and $`R_N`$ is the junction normal resistance. If however, the distribution function contains a sharp feature, narrower than $`k_BT`$, at the energies where the supercurrent- carrying states are present, a very different temperature dependence and even a change of sign can be expected. In the Baselmans experiment the non-equilibrium distribution function had the form of a thermally rounded step function comprised of a linear combination of two mutually bias-displaced Fermi functions of the normal reservoirs at each end of the filamentary normal conductor. The observed effect was however much weaker than the equilibrium supercurrent (without injection). A considerably stronger effect can be anticipated if the relevant non-equilibrium distribution function derived from the thermally rounded Fermi functions of the normal reservoirs is replaced by a distribution function induced between two voltage biased superconducting reservoirs. If the size of the normal conductor is smaller than the phasebreaking diffusion length, the non-equilibrium electron energy distribution in the normal conductor will contain replica of the sharp singularities in the superconducting density of states of the reservoirs. The fingerprints of these are seen as a sub-gap structure in the I-V characteristics of SNS junctions . At a bias voltage of $`V=\mathrm{\Delta }(T)/e`$ the singularities in the normal conductor distribution function match the Fermi energies of the superconductors, where the supercurrent carrying states are concentrated. The resulting injection induced supercurrent in the adjacent SNS junction is thus expected to have a maximum for an injection bias $`V=\mathrm{\Delta }(T)/e`$, and a weak temperature dependence due to the small thermal smearing of the distribution function in play. Besides these non-equilibrium effects a broad heating effect is expected, which will always suppress the critical supercurrent. In this letter we demonstrate experimentally for the first time that the superconductivity of an SNS-junction can be enhanced above the equilibrium value by injecting quasiparticles with an energy distribution which match the spectral supercurrent density $`N_J(E)`$. The effect is realised in a three terminal sample geometry, where three superconducting Al electrodes are connected to the same piece of highly doped GaAs semiconductor. One of the electrodes is used as a common ground for the current flow. Another electrode is used as a detector and the third electrode as an injector. For the particular devices studied we find that the injection-induced supercurrent exceeds the equilibrium supercurrent at temperatures above $`0.6K`$ (approx. $`\frac{1}{2}T_c`$). The samples were formed from a layered structure of GaAs and Al, grown in an MBE chamber. Here, $`200nm`$ of highly doped n-GaAs were grown on an undoped/insulating substrate. This was then capped with 150-$`200nm`$ Al. In order to reduce the Schottky-barrier between GaAs and Al, five layers of $`\delta `$-doping with $`510^{13}cm^2`$ Si was inserted in the GaAs just below the Al. The Al film was subsequently deposited without breaking the vacuum. This resulted in a contact resistance of $`810^9\mathrm{\Omega }cm^2`$. Samples with a planar geometry as shown in Figure 1 were then formed by removing Al in selected areas with conventional E-beam lithography and wet etch. Larger scale patterning to form a $`20\mu m`$ wide mesa structure was done with UV-lithography. The details of the fabrication has been published elsewhere . The measurements were performed in a pumped $`{}_{}{}^{3}He`$ cryostat with a base temperature of $`235mK`$. At low temperatures the GaAs-film had a carrier density of $`n=5.510^{18}cm^3`$ and a mean free path of $`\mathrm{}=42nm`$, corresponding to a diffusion constant $`D=127cm^2/s`$. In an independent weak localisation experiment the phase-breaking diffusion length was found to be $`\mathrm{}_\varphi 5\mu m`$ at the base temperature. The Al-film had a superconducting critical temperature of $`1.196K`$. All measured samples had nearly identical characteristics. The data presented in this paper is based on a single sample, although all samples exhibited similar effects. This particular sample is shown in Figure 1. The three superconducting Al electrodes are connected to the same piece of GaAs within mesoscopic range. The distance between two neighbouring electrodes is approximately $`400nm`$ corresponding to a Thouless-energy $`\epsilon _L=52\mu eV`$. The junctions AB and BC formed individual Josephson junctions with normal resistances $`3.3\mathrm{\Omega }`$ and $`3.2\mathrm{\Omega }`$, and critical currents respectively $`2.2\mu A`$ and $`3.4\mu A`$ at $`240mK`$, when all electrodes were kept at zero voltage. The junction AC was much weaker coupled because of the longer distance between A and C. In this experiment we define the AB junction as the detector, and the BC junction as the injector. The critical supercurrent was measured on the detector while the injector was biased at high voltages. Because the samples were symmetrical, the roles of the two junctions could be (and was) interchanged. During the experiment, a detector current was passed through electrode A, and an injection current was passed through electrode C, both from high impedance ($`470k\mathrm{\Omega }`$) DC current sources. Electrode B was connected to ground through an Ampere-meter, in order to check that no current leaks were present in the set- up. The detector (AB) and injector (BC) voltages were measured by separate leads to the superconductors. All leads to the sample were fitted with $`50cm`$ low temperature THERMOCOAX filters providing high frequency power attenuation of $`12.5\sqrt{f[GHz]}dB`$ and room temperature $`\pi `$ filters giving approximately $`20dB`$ attenuation at $`700kHz`$. Both the detector and injector junctions showed a “Fraunhofer -type of pattern with up to 10 well-developed lobes in the critical supercurrent vs. applied magnetic field, indicating a high degree of homogeneity of the junctions. Before each measurement series on the 3-terminal samples, the magnetic field was carefully zeroed to within a small fraction of a flux quantum . In Figure 2 we show the detector supercurrent at various injection currents. In the left panel, the detector critical current $`I_c`$ is plotted as function of the injection current at three different temperatures. Above the highest injection currents shown in the figure, the central electrode suddenly went normal due to a current in excess of its critical current and above this, injector bias measurements could not be made. Each data point is the result of a fit to a noise rounded resistively shunted junction (RSJ) model of individual I-V characteristics of the detector junction. In the right panel of Figure 2 we show three examples of the fits. The injection currents and fitted $`I_c`$ of these are indicated with arrows in the left panel. The fitted noise temperature varied randomly in the range $`13K`$ due to the error on the measured data . The I-V characteristics of the detector junction show current offsets. The offsets are simply a fraction of the injector current and are exclusively due to the sample geometry, where the three electrodes are connected to the same piece of GaAs. Consequently the detector current $`I_{det}`$ has to be compensated by about one fourth of the injection current for the detector to be at zero bias. In the left panel of Figure 2 we see, that the critical supercurrent exhibit two clear features as function of injection current (and voltage). The equilibrium supercurrent appears to be falling off rapidly with increasing injection current, and is seen to be strongly temperature dependent. The next maximum appearing for higher injection currents is interpreted as a non-equilibrium supercurrent. At very high injection currents, close to the critical current of the central electrode, we observe a weak enhancement of the critical current. This effect remains unexplained. The non-equilibrium critical current builds up with increasing injection current to a maximum (at $`400mK`$ indicated by C in Figure 2). We see that while the equilibrium supercurrent is dominating at $`400mK`$, at $`900mK`$ it has almost vanished. The non-equilibrium supercurrent on the other hand is almost temperature independent and by far exceeds the equilibrium supercurrent at $`900mK`$. At $`600mK`$ the two critical currents have equal magnitude. Although, in the figure we mostly plot the properties for positive injection currents, we would like to emphasise that the effect was completely symmetrical upon reversal of current direction. The effects were reproduced in several samples. In Figure 3 we show a detailed plot of the temperature dependence of both the critical current at zero injection, and the optimal non-equilibrium current. The detector $`I_c`$ without injection (squares) has a strong temperature dependence, which has been fitted to Eq. (2) with parameters $`I_0=4.9\mu A`$ and $`\xi _N(T_c)=115nm`$. From the independently measured sample parameters we find $`I_0=363\mu A`$ and $`\xi _N(T_c)=113nm`$. The large deviation of the prefactor $`I_0`$ may be attributed to the fact that the calculation of Eq. (2) does not take interface barriers into account. The other curve (triangles) shows the maxima of the injection induced non-equilibrium $`I_c`$ as exemplified for three temperatures in Figure 2. This only decreases slightly with increasing temperature. In the insert, we plot the corresponding voltage across the injector junction for each of these points. The injector voltage varied slightly during each detector I-V measurement. The plotted value was taken where the detector current $`I_{det}`$ matched the earlier mentioned current offset. The full curve in the insert shows the BCS theory gap function, where $`T_c=1.196K`$ has been measured, and $`\mathrm{\Delta }(0)=158\mu eV`$ has been adjusted from the bulk value $`175\mu eV`$ to fit the data. It is clearly seen that the injection bias giving the maximum non-equilibrium supercurrent, corresponds to the superconductor energy gap $`\mathrm{\Delta }(T)`$. Together with the weak temperature dependence of the non-equilibrium critical current, we take this as the most important experimental indication that the observed injection induced phenomena are related to a non-equilibrium population of the supercurrent carrying states. This non-equilibrium population is induced by Andreev reflections on the superconducting injection electrode, and is possessing a sharp kink at the superconducting gap energy. This kink originates from the singularities in the density of states of the superconductor. The functional dependence of the spectral supercurrent density of Eq. (1) has not been calculated for the planar geometry of the samples used in the experiment, however the form of the distribution function may allow us to probe the varying part of $`N_j`$. We believe the observed phenomena are related to the earlier observations published in Ref. , where a maximum in oscillation amplitude in a flux sensitive interferometer was observed for the bias voltages $`V=\mathrm{\Delta }/e\pm \epsilon _L`$. Here the phase dependent (coherent) change of the normal resistance was probed instead of the supercurrent. In conclusion, we have measured the injection-induced supercurrent in a three terminal device. Above $`600mK`$ the induced non-equilibrium supercurrent exceeds the equilibrium (zero injection) supercurrent. The optimal injection induced critical current is observed when the injector electrode is biased at $`V=\mathrm{\Delta }(T)/e`$. Further studies, are needed to map out the detailed properties of the non-equilibrium supercurrent, e.g. the magnetic field dependence and the d. c. Josephson effects. We acknowledge useful discussions with Vitaly Shumeiko and Jonn Lantz. We thank the III-V Nanolab of the Niels Bohr Institute for providing the processing facilities. This work has been supported by the Danish Technical Research Council, and the Velux Foundation.
no-problem/9907/hep-th9907033.html	ar5iv	text	# References EDO-EP-27 July, 1999 On Alternate Derivation of Type IIB Matrix Model Ichiro Oda <sup>1</sup><sup>1</sup>1 E-mail address: ioda@edogawa-u.ac.jp. Address till August 4, 1999: Universita Degli Studi Di Padova, Dipartimento Di Fisica ”Galileo Galilei”, Via F. Marzolo, 8, 35131 Padova, Italia, e-mail address: ICHIRO.ODA@pd.infn.it Edogawa University, 474 Komaki, Nagareyama City, Chiba 270-0198, JAPAN ## Abstract Starting with Green-Schwarz superstring action, we construct a type IIB matrix model. We fix the local $`\kappa `$ symmetry in the Killing spinor gauge and then perform the world-sheet duality transformation. A matrix model obtained from this gauge-fixed action is shown to be equivalent to the type IIB matrix model constructed by Ishibashi et al. Our construction does not make use of an analytic continuation of spinor variable. Moreover, it seems that our construction is applicable to that of more general type IIB matrix models in a curved background such as $`AdS_5\times S^5`$. Matrix models provide us non-perturbative formulations of M theory and type IIB superstring theory . It is quite surprising that such simple supersymmetric quantum mechanics theories seem to contain within them a complicated structure of supergravity. Despite much remarkable success of matrix models, we are still far from having a complete understanding of several important problems. In particular, one big mystery is the problem of background independence . The actions of the matrix models include a flat metric in the kinetic and/or the potential parts, which is not allowed as stressed in a previous article . This is because matrix models, which describe quantum gravity, should pick up its own space-time background in a dynamical manner and should not be $`a\mathrm{𝑝𝑟𝑖𝑜𝑟𝑖}`$ formulated in a specific fixed background. Thus it is desirable to construct a matrix model which is independent of background metric , but its relation to quantum gravity is not clear at present. Another interesting approach would be construction of a matrix model in an arbitrary curved background. (But we believe that the background is not completely arbitrary in that the backgorund fields would have to satisfy the proper on-shell constraints of supergravity in order that the action is $`\kappa `$-symmetric.) Recently, in an attempt of constructing a type IIB matrix model on $`AdS_5\times S^5`$ , we have utilized the Killing spinor gauge developed in as the gauge-fixing condition of the $`\kappa `$-symmetry. (See also related works .) The aim of the present article is just to apply the method developed there to the case of a flat background geometry. Indeed, following the method we will obtain the same matrix model as found by Ishibashi et al . One advantage of our construction is that we do not have to make use of an analytic continuation of spinor variable in the path integral as used by Ishibashi et al. Instead, we will use the Killing spinor gauge and the world-sheet duality symmetry of type IIB superstring theory. Another advantage is of course that our construction is extensible to construction of more complicated matrix models . First of all, let us start with the type IIB Green-Schwarz superstring action in a flat background in the Nambu-Goto form : $`S=S_1+S_2={\displaystyle d^2\sigma (L_1+L_2)},`$ (1) where $`L_1`$ $`=`$ $`\sqrt{{\displaystyle \frac{1}{2}}\mathrm{\Sigma }^2},`$ $`L_2`$ $`=`$ $`i\epsilon ^{ij}_iX^a(\overline{\theta }^1\mathrm{\Gamma }^a_j\theta ^1\overline{\theta }^2\mathrm{\Gamma }^a_j\theta ^2)+\epsilon ^{ij}\overline{\theta }^1\mathrm{\Gamma }^a_i\theta ^1\overline{\theta }^2\mathrm{\Gamma }^a_j\theta ^2`$ (2) $`=`$ $`i\epsilon ^{ij}𝒦^{IJ}\overline{\theta }^I\mathrm{\Gamma }^a_j\theta ^J(_iX^a{\displaystyle \frac{1}{2}}i\overline{\theta }^K\mathrm{\Gamma }^a_i\theta ^K),`$ with $`\mathrm{\Sigma }^2`$ $`=`$ $`\mathrm{\Sigma }_{ab}\mathrm{\Sigma }^{ab},`$ $`\mathrm{\Sigma }^{ab}`$ $`=`$ $`\mathrm{\Pi }_i^a\mathrm{\Pi }_j^b\epsilon ^{ij},`$ $`\mathrm{\Pi }_i^a`$ $`=`$ $`_iX^ai\overline{\theta }^I\mathrm{\Gamma }^a_i\theta ^I,`$ $`𝒦`$ $`=`$ $`\sigma _3=\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right).`$ (3) Here $`\theta ^I`$ are Majorana-Weyl spinors in ten dimensions and various indices take the following values: $`i,j,\mathrm{}=0,1`$; $`a,b,\mathrm{}=0,1,\mathrm{},9`$; $`I,J,\mathrm{}=1,2`$. It is straightforward to check that this action has the $`N=2`$ supersymmetry $`\delta _s\theta ^I`$ $`=`$ $`\epsilon ^I,`$ $`\delta _sX^a`$ $`=`$ $`i\overline{\epsilon }^I\mathrm{\Gamma }^a\theta ^I,`$ (4) and the local $`\kappa `$-symmetry $`\delta _\kappa \theta ^I`$ $`=`$ $`\alpha ^I,`$ $`\delta _\kappa X^a`$ $`=`$ $`i\overline{\theta }^I\mathrm{\Gamma }^a\alpha ^I,`$ (5) where $`\alpha ^1`$ $`=`$ $`(1+\stackrel{~}{\mathrm{\Gamma }})\kappa ^1,`$ $`\alpha ^2`$ $`=`$ $`(1\stackrel{~}{\mathrm{\Gamma }})\kappa ^2,`$ $`\stackrel{~}{\mathrm{\Gamma }}`$ $`=`$ $`{\displaystyle \frac{1}{2\sqrt{\frac{1}{2}\mathrm{\Sigma }^2}}}\mathrm{\Sigma }_{ab}\mathrm{\Gamma }^{ab},`$ $`\stackrel{~}{\mathrm{\Gamma }}^2`$ $`=`$ $`+1.`$ (6) Now let us fix the $`\kappa `$-symmetry in the Killing spinor gauge . The flat Minkowskian space-time has the Poincare group $`ISO(1,9)`$ as the isometry group, so it is natural to consider the following Killing spinor gauge condition for the case at hand $`\theta _{}^I𝒫_{}^{IJ}\theta ^J=0,`$ (7) where the projection operator $`𝒫_\pm ^{IJ}`$ is defined as $`𝒫_\pm ^{IJ}={\displaystyle \frac{1}{2}}(\delta ^{IJ}\pm i\mathrm{\Gamma }_{11}\epsilon ^{IJ}),`$ (8) with the chiral matrix $`\mathrm{\Gamma }_{11}=\mathrm{\Gamma }^0\mathrm{\Gamma }^1\mathrm{}\mathrm{\Gamma }^9`$. (The other simple choice of the projection operator is $`𝒫_\pm ^{IJ}=\frac{1}{2}(\delta ^{IJ}\pm i\epsilon ^{IJ})`$. It is easy to see that this choice leads to the same result as that shown in this paper.) Note that this projection operator satisfies the following identities: $`𝒫_\pm ^2`$ $`=`$ $`𝒫_\pm ,`$ $`𝒫_+𝒫_{}`$ $`=`$ $`𝒫_{}𝒫_+=0,`$ $`𝒫_\pm \mathrm{\Gamma }^a`$ $`=`$ $`\mathrm{\Gamma }^a𝒫_{}.`$ (9) As a result, we can show $`\overline{\theta }^I\mathrm{\Gamma }^a_i\theta ^I`$ $`=`$ $`2i\overline{\theta }\mathrm{\Gamma }^a_i\theta ,`$ $`𝒦^{IJ}\overline{\theta }^I\mathrm{\Gamma }^a_i\theta ^J`$ $`=`$ $`0.`$ (10) where we have defined $`\theta _+^1=\theta `$. The latter equation in Eq.(10), together with Eq.(2), leads to the fact that the Wess-Zumino term $`L_2`$ vanishes in the present gauge condition. Thus, we arrive at the $`\kappa `$-symmetry gauge-fixed action given by $`S={\displaystyle d^2\sigma \sqrt{\frac{1}{2}\mathrm{\Sigma }^2}},`$ (11) where $`\mathrm{\Sigma }^2`$ $`=`$ $`\mathrm{\Sigma }_{ab}\mathrm{\Sigma }^{ab},`$ $`\mathrm{\Sigma }^{ab}`$ $`=`$ $`\mathrm{\Pi }_i^a\mathrm{\Pi }_j^b\epsilon ^{ij},`$ $`\mathrm{\Pi }_i^a`$ $`=`$ $`_iX^a2i\overline{\theta }\mathrm{\Gamma }^a_i\theta .`$ (12) The $`\kappa `$-symmetry gauge-fixed action (11) is still quartic with respect to the fermionic variable, but it is possible to reduce it to a much simpler quadratic form by making the world-sheet duality transformation as follows . To this aim, let us first rewrite the action (11) into the Polyakov form $`S={\displaystyle \frac{1}{2}}{\displaystyle d^2\sigma \sqrt{g}g^{ij}(_iX^a2i\overline{\theta }\mathrm{\Gamma }^a_i\theta )(_jX^a2i\overline{\theta }\mathrm{\Gamma }^a_j\theta )}.`$ (13) Next, we rewrite the action (13) into the first-order form by introducing the Lagrange multiplier $`P_i^a`$ $`\stackrel{~}{S}={\displaystyle \frac{1}{2}}{\displaystyle d^2\sigma \sqrt{g}g^{ij}\left[P_i^aP_j^a+2P_i^a(_jX^a2i\overline{\theta }\mathrm{\Gamma }^a_j\theta )\right]}.`$ (14) Actually the integration over $`P_i^a`$ gives rise to the action (13). After integrating out $`X^a`$ and then solving the resulting equation in terms of the dual variable $`\stackrel{~}{X}^a`$ as $`P^{ia}={\displaystyle \frac{1}{\sqrt{g}}}\epsilon ^{ij}_j\stackrel{~}{X}^a,`$ (15) we can obtain the desired dual action $`\stackrel{~}{S}={\displaystyle \frac{1}{2}}{\displaystyle d^2\sigma (\sqrt{g}g^{ij}_i\stackrel{~}{X}^a_j\stackrel{~}{X}^a+4i\epsilon ^{ij}_i\stackrel{~}{X}^a\overline{\theta }\mathrm{\Gamma }^a_j\theta )}.`$ (16) It is quite remarkable that the $`\kappa `$-symmetry gauge-fixed action (11) is transformed to a much simpler quadratic form through the world-sheet duality transformation. The action $`\stackrel{~}{S}`$ is still invariant under the residual $`N=2`$ supersymmetry. This supersymmetry can be found by amalgamating the original supersymmetry (4) with the $`\kappa `$-symmetry (5) to preserve the gauge fixing condition (7), in other words, $`\theta ^1=i\mathrm{\Gamma }_{11}\theta ^2`$. Following a similar procedure to the reference , let us consider the mixed transformation of the $`N=2`$ supersymmetry (4) and the $`\kappa `$-symmetry (5) $`\delta \theta ^I`$ $`=`$ $`\delta _s\theta ^I+\delta _\kappa \theta ^I,`$ $`\delta X^a`$ $`=`$ $`\delta _sX^a+\delta _\kappa X^a.`$ (17) Provided that we define $`\stackrel{~}{\epsilon }^2=i\mathrm{\Gamma }_{11}\epsilon ^2`$ and $`\stackrel{~}{\kappa }^2=i\mathrm{\Gamma }_{11}\kappa ^2`$ and then choose $`\kappa ^1`$ $`=`$ $`{\displaystyle \frac{\epsilon ^1+\stackrel{~}{\epsilon }^2}{2}},`$ $`\stackrel{~}{\kappa }^2`$ $`=`$ $`{\displaystyle \frac{\epsilon ^1\stackrel{~}{\epsilon }^2}{2}},`$ (18) the gauge fixing condition is preserved as desired. With these conditions and the definition $`\xi `$ $`=`$ $`{\displaystyle \frac{\epsilon ^1+\stackrel{~}{\epsilon }^2}{2}},`$ $`\epsilon `$ $`=`$ $`{\displaystyle \frac{\epsilon ^1\stackrel{~}{\epsilon }^2}{2}},`$ (19) it is easy to derive the new $`N=2`$ supersymmetry law of $`\theta `$ whose result is given by $`\delta \theta =\xi \stackrel{~}{\mathrm{\Gamma }}\epsilon ,`$ (20) whereas it is necessary to make some work to do the transformation law of $`\stackrel{~}{X^a}`$. In a similar way to the case of $`\theta `$, we can find the new $`N=2`$ supersymmetry transformation law of the original variable $`X^a`$ $`\delta X^a=2i\overline{\theta }\mathrm{\Gamma }^a(\xi +\stackrel{~}{\mathrm{\Gamma }}\epsilon ),`$ (21) which gives us the transformation of $`P_i^a`$ $`\delta P_i^a=4i_i\overline{\theta }\mathrm{\Gamma }^a\stackrel{~}{\mathrm{\Gamma }}\epsilon .`$ (22) From this expression, we can derive the the new $`N=2`$ supersymmetry law of $`\stackrel{~}{X^a}`$ $`\delta \stackrel{~}{X}^a=4i\overline{\epsilon }\mathrm{\Gamma }^a\theta .`$ (23) Note that to derive this equation we have used the fact that $`g_{ij}\stackrel{~}{\mathrm{\Gamma }}`$ can be replaced with $`\frac{\epsilon _{ij}}{\sqrt{g}}`$ in $`\delta P_i^a`$ since both the expressions make the same contribution in the action $`\stackrel{~}{S}`$. At this stage, it is important to notice that $`\stackrel{~}{\mathrm{\Gamma }}`$ can be expressed in terms of the dual variable as $`\stackrel{~}{\mathrm{\Gamma }}={\displaystyle \frac{1}{2\sqrt{g}}}_i\stackrel{~}{X}^a_j\stackrel{~}{X}^b\epsilon ^{ij}\mathrm{\Gamma }^{ab}.`$ (24) Finally, integrating out $`g_{ij}`$, we finish with the dual action $`\stackrel{~}{S}={\displaystyle d^2\sigma (\sqrt{\frac{1}{2}\sigma ^2}+2i\epsilon ^{ij}_iX^a\overline{\theta }\mathrm{\Gamma }^a_j\theta )}.`$ (25) The $`N=2`$ supersymmetry of this action is then given by $`\delta \theta `$ $`=`$ $`{\displaystyle \frac{1}{2\sqrt{\frac{1}{2}\sigma ^2}}}\sigma _{ab}\mathrm{\Gamma }^{ab}\epsilon +\xi ,`$ $`\delta X^a`$ $`=`$ $`4i\overline{\epsilon }\mathrm{\Gamma }^a\theta .`$ (26) In the above equations, we have written $`\stackrel{~}{X}^a`$ simply as $`X^a`$ and defined $`\sigma ^{ab}`$ as $`\sigma ^{ab}=_iX^a_jX^b\epsilon ^{ij}.`$ (27) These expressions are equivalent to those in , so we can obtain the same type IIB matrix model by following the same path of thought as in . Notice that even if we have derived the same matrix model by starting with the same type IIB superstring action (1), (2) the derivation method is different. In particular, our approach does not rely on an analytic continuation of the spinor variable in the path integral which plays an important role in and seems to be extensible to more general situation . Of course, the key idea behind the approach at hand is a symmetry of the Green-Schwarz superstring action under the world-sheet duality transformation. Incidentally, we have recently examined the duality transformations of the Green-Schwarz superstring as well as supersymmetric D-brane actions in a general type II supergravity background .
no-problem/9907/hep-th9907193.html	ar5iv	text	# An Application of the Schwarzian DerivativeTalk given by B. Zhou at the CCAST Workshop on Integrable System, Beijing, May 4–7. ## 1 Notations and Conventions Let $`f`$ be an arbitrary function of $`r`$ and $`f^{}`$ be its derivative. The notation $`f(r)`$ always represents the value of the function $`f`$ at $`r`$. Similarly, $`f^{}(r)`$ is the value of the function $`f^{}`$ at $`r`$. For two functions $`f`$ and $`g`$, $`fg`$ is their product<sup>1</sup><sup>1</sup>1We denote $`ff`$ by $`f^2`$, $`fff`$ by $`f^3`$, and so on., and $`{\displaystyle \frac{f}{g}}`$ is their division, while the composition of two functions $`f`$ and $`g`$ is denoted by $`fg`$. These functions are defined by $`(fg)(r)`$ $`=`$ $`f(r)g(r),`$ (10) $`\left({\displaystyle \frac{f}{g}}\right)(r)`$ $`=`$ $`{\displaystyle \frac{f(r)}{g(r)}},`$ (11) $`(fg)(r)`$ $`=`$ $`f(g(r)).`$ (12) For product of three or more functions we assume that the composition of two functions has a higher rank of precedence of associavity than the products of functions. Nevertheless the symbol $`{\displaystyle \frac{f}{g}}`$ is considered to be a pure entity and can’t be broken. To understand these conventions we have the following examples: $`fgh=ghf=f(gh),`$ (13) $`{\displaystyle \frac{f}{g}}h=\left({\displaystyle \frac{f}{g}}\right)h={\displaystyle \frac{fh}{gh}}.`$ (14) If $`k`$ or $``$, we define a function $`l_k`$ as the multiplication of its variable with the number $`k`$, i.e., $$l_k(r)=kr.$$ For convenience we also consider $`k`$ itself as a constant function taking the value $`k`$: $$k(r)=k.$$ Other functions such as the power function $`r^s`$ with real number $`s`$ and the exponential function $`e^r`$ are denoted by $`p_s`$ and $`\mathrm{exp}`$ and other elementary functions are denoted by their standard mathematical symbols. With these conventions, we can derive the following derivative rules: $$\begin{array}{cc}l_k^{}=k,\hfill & k^{}=0,\hfill \\ p_s^{}=sp_{s1}\hfill & \mathrm{arctan}^{}=\frac{1}{1+p_2},\hfill \\ \mathrm{exp}^{}=\mathrm{exp},\hfill & \mathrm{ln}^{}=p_1,\hfill \end{array}$$ (15) and the derivative rule for composition of function is $$(fg)^{}=g^{}f^{}g$$ (16) ## 2 Some Properties of the Schwarzian Derivative Now we list some elementary properties of the Schwarzian derivative. (1) If $`f`$ and $`g`$ are two functions, we have $$S(fg)=(g^{})^2S(f)g+S(g).$$ (17) (2) If $`f={\displaystyle \frac{l_a+b}{l_c+d}}`$ for some numbers $`a`$, $`b`$, $`c`$ and $`d`$, namely, $`f(r)={\displaystyle \frac{ar+b}{cr+d}}`$, then $$S\left(\frac{l_a+b}{l_c+d}\right)=0.$$ (18) This is the well-know fact that the fractional linear transformation is a global conformal transformation of the complex sphere. By using this result we have $$S\left(\frac{l_a+b}{l_c+d}g\right)=S(g).$$ (19) That is to say, if $`g`$ is a special solution of the equation $`S(f)=R`$, the general solution will be $$f=\frac{l_a+b}{l_c+d}g=\frac{ag+b}{cg+d}$$ (20) with constants $`a`$, $`b`$, $`c`$, $`d`$ such that $`adbc=1`$. (3) For $`s`$, we have $`S(l_k)`$ $`=`$ $`0,`$ (21) $`S(p_s)`$ $`=`$ $`{\displaystyle \frac{s^21}{2p_2}},`$ (22) $`S(\mathrm{exp})`$ $`=`$ $`{\displaystyle \frac{1}{2}},`$ (23) $`S(\mathrm{ln})`$ $`=`$ $`{\displaystyle \frac{1}{2p_2}},`$ (24) $`S(\mathrm{tan})`$ $`=`$ $`2,`$ (25) $`S(\mathrm{arctan})`$ $`=`$ $`{\displaystyle \frac{2}{(1+p_2)^2}}.`$ (26) ## 3 The Complete Solution of Eq. (9) The function $`\stackrel{~}{R}`$ in Eq. (9) is $$\stackrel{~}{R}=\frac{\stackrel{~}{d}^21}{2p_2}+\frac{\stackrel{~}{\mathrm{\Lambda }}p_{2\stackrel{~}{d}2}}{2}\frac{1}{(C_2+C_1p_2)^2}p_{\stackrel{~}{d}}.$$ (27) For $`C_1C_2>0`$, we have $`\stackrel{~}{R}`$ $`=`$ $`S(p_{\stackrel{~}{d}})+(p_{\stackrel{~}{d}}^{})^2S(h_1)p_{\stackrel{~}{d}}`$ (28) $`=`$ $`S(h_1p_{\stackrel{~}{d}})`$ where $`h_1`$ is a yet-unknown function such that $`S(h_1)`$ $`=`$ $`{\displaystyle \frac{\stackrel{~}{\mathrm{\Lambda }}}{2\stackrel{~}{d}^2C_2^2}}{\displaystyle \frac{1}{(1+\frac{C_1}{C_2}p_2)^2}}`$ (29) $`=`$ $`S(l_{\sqrt{C_1/C_2}})+(l_{\sqrt{C_1/C_2}}^{})^2{\displaystyle \frac{\stackrel{~}{\mathrm{\Lambda }}}{2\stackrel{~}{d}^2C_1C_2}}{\displaystyle \frac{1}{(1+p_2)^2}}l_{\sqrt{C_1/C_2}}`$ $`=`$ $`S(h_2l_{\sqrt{C_1/C_2}}).`$ Now we want to find a function $`h_2`$ such that $`S(h_2)`$ $`=`$ $`{\displaystyle \frac{\stackrel{~}{\mathrm{\Lambda }}}{2\stackrel{~}{d}^2C_1C_2}}{\displaystyle \frac{1}{(1+p_2)^2}}`$ (30) $`=`$ $`S(\mathrm{arctan})+{\displaystyle \frac{\mathrm{\Delta }K}{2\stackrel{~}{d}^2C_1C_2}}{\displaystyle \frac{1}{(1+p_2)^2}}`$ $`=`$ $`S(\mathrm{arctan})+(\mathrm{arctan}^{})^2{\displaystyle \frac{\mathrm{\Delta }K}{2\stackrel{~}{d}^2C_1C_2}}\mathrm{arctan}`$ $`=`$ $`S(h_3\mathrm{arctan}).`$ Note that in the third line we have considered $`{\displaystyle \frac{\mathrm{\Delta }K}{2\stackrel{~}{d}^2C_1C_2}}`$ to be a constant function. The function $`h_3`$ in the above is easy to find to be $`\mathrm{tan}l_k`$ with $$k=\frac{1}{2\stackrel{~}{d}}\sqrt{\frac{\mathrm{\Delta }K}{C_1C_2}},$$ (31) because $`S(h_3)`$ $`=`$ $`{\displaystyle \frac{\mathrm{\Delta }K}{2\stackrel{~}{d}^2C_1C_2}}`$ (32) $`=`$ $`S(l_k)+(l_k^{})^2S(\mathrm{tan})l_k`$ $`=`$ $`S(\mathrm{tan}l_k).`$ Combining all the above steps, one finds that $$\stackrel{~}{R}=S(\mathrm{tan}l_k\mathrm{arctan}l_{\sqrt{C_1/C_2}}p_{\stackrel{~}{d}}).$$ (33) and a special solution of eq. (9) is found to be: $$g_0=\mathrm{tan}l_k\mathrm{arctan}l_{\sqrt{C_1/C_2}}p_{\stackrel{~}{d}},$$ (34) namely, $$g_0(r)=\mathrm{tan}\left(k\mathrm{arctan}\sqrt{\frac{C_1}{C_2}}r^{\stackrel{~}{d}}\right).$$ (35) The special case considered in corresponds to $`k=1`$. By using Mathematica, one can easily check that the above function is indeed a solution of eq. (9). The general solution of eq. (9) is obtained from the above special solution, eq. (35), by an arbitrary $`SL(2,R)`$ transformation: $$g(r)=\frac{a_0g_0(r)+b_0}{c_0g_0(r)+d_0},$$ (36) where $`a_0`$, $`b_0`$, $`c_0`$ and $`d_0`$ consist of an $`SL(2,R)`$ matrix: $$a_0d_0b_0c_0=1.$$ (37) Here we have three independent constants. This is the right number for a third order ordinary differential equation. So we obtain the complete solution to eq. (9). With this general solution in hand one can proceed to obtain the general solution of (3). For details please see . ## 4 More examples There are also other examples which can be solved by the above method. One example comes from the special case $`\stackrel{~}{d}=0`$ and the equation is as follows ($`\mathrm{\Delta }=a^2`$): $$Y^{\prime \prime }\frac{a^2}{2}(Y^{})^2+Q(r)Y^{}=R(r)$$ (38) where $`Q(r)`$ $`=`$ $`{\displaystyle \frac{1+a^2}{r}}+{\displaystyle \frac{1+aC_3}{r\mathrm{ln}\frac{r_0}{r}}},`$ (39) $`R(r)`$ $`=`$ $`{\displaystyle \frac{a^2}{2r^2}}+{\displaystyle \frac{1+aC_3}{r^2\mathrm{ln}\frac{r_0}{r}}}+{\displaystyle \frac{C_3^2}{2r^2\left(\mathrm{ln}\frac{r_0}{r}\right)^2}}+{\displaystyle \frac{(D3)C_4^24C_44(D1)}{4(D2)r^2\left(\mathrm{ln}\frac{r_0}{r}\right)^2}}.`$ (40) In terms of $$g(r)=𝑑re^{a^2Y{\scriptscriptstyle Q(r)𝑑r}},$$ (41) or, equivalently, $$Y=\frac{1}{a^2}(\mathrm{ln}(g^{})+Q(r)𝑑r),$$ (42) the equation for $`Y`$ can be written as $$S(g)=\stackrel{~}{R}(r),$$ (43) where $`S(g)`$ is the Schwarzian derivative of $`g`$ and $`\stackrel{~}{R}(r)`$ $`=`$ $`a^2R(r)Q^{}(r){\displaystyle \frac{1}{2}}Q^2(r)`$ (44) $`=`$ $`{\displaystyle \frac{1}{2r^2}}+{\displaystyle \frac{1K}{2r^2\left(\mathrm{ln}\frac{r_0}{r}\right)^2}}`$ with $$K=4(1+aC_3)+\frac{4(D1)+4C_4(D3)C_4^2}{2(D2)}a^2.$$ (45) For different choices of $`K`$ the special solution is as follows: (1) For $`K=0`$, $$g_0=\mathrm{ln}\left\|\mathrm{ln}\frac{r_0}{r}\right\|;$$ (46) (2) For $`K>0`$ ($`k=\sqrt{K}`$), $$g_0=\left\|\mathrm{ln}\frac{r_0}{r}\right\|^k;$$ (47) (3) For $`K<0`$ ($`k=\sqrt{K}`$), $$g_0=\mathrm{tan}\left(\frac{k}{2}\mathrm{ln}\left\|\mathrm{ln}\frac{r_0}{r}\right\|\right).$$ (48) For more examples and extensive discussions, we refer the readers to . ## 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. C.-J. Zhu would like to thank the Abdus Salam International Center for Theoretical Physics for hospitality at the Extended Workshop on String Theory (June 1st–July 16th, 1999, Trieste, Italy). He would also like to thank SISSA/ISAS (International School for Advanced Studies, Trieste, Italy) for hospitality where we finally find the time to finish the writing of this note.
no-problem/9907/quant-ph9907006.html	ar5iv	text	# Optical Quantum Random Number Generator ## Acknowledgments This work was partly supported by the Esprit project EQCSPOT. ## Figure Caption Figure 1: Schematic diagram of the random number generator
no-problem/9907/hep-th9907031.html	ar5iv	text	# 1 Introduction ## 1 Introduction The triviality of simple random surface models on the lattice, due to the dominance of branched polymers , and the non-scaling of the string tension in analogous Euclidean invariant models based on dynamical triangulations made it imperative to study models with an action depending on the extrinsic curvature of surfaces. For a review of work in this area up to 1997, see . One of the models with extrinsic curvature dependent action is the so-called gonihedric random surface model introduced in and further studied in . This model is unstable in the simplest cases . However, there is a lattice discretization of the model which is not plagued by stability problems and is more amenable to analytical study, see -. Another lattice model with similar properties was introduced and studied in . In a simplified version of the three-dimensional gonihedric lattice model was studied and the two largest eigenvalues of the transfer matrix were determined. In this paper we finish the calculation of the spectrum of the transfer matrix for the model and find all the eigenfunctions explicitly. ## 2 Transfer matrix for loops Let $`T_3`$ denote a sublattice of $`𝐙^3`$ of size $`M\times M\times N`$ with periodic boundary conditions. We shall think of the third coordinate direction where the lattice has extension $`N`$ as the ‘time’ direction. The configuration space of the system we wish to study is a collection of subsets of the plaquettes in $`T_3`$, denoted $``$, which we refer to as singular surfaces or simply as surfaces. A collection of plaquettes $`M`$ belongs to $``$ if and only if any link in a plaquette in $`M`$ belongs to an even number of plaquettes in $`M`$. This means that the surfaces in $``$ do not overlap themselves but they can intersect themselves at right angles and they are closed. Note that the surfaces need not be connected. It is not hard to see that the conguration space $``$ is identical to the configuration space of the three-dimensional Ising model on the periodic lattice $`T_3`$. The surfaces are in one to one correspondence with phase boundaries. We say that a link $`\mathrm{}`$ in a surface $`M`$ is an edge if $`\mathrm{}`$ is contained in exactly two plaquettes that meet at a right angle. We denote the collection of all edges in $`M`$ by $`(M)`$. The action $`S(M)`$ of the surface $`M`$ is defined as the total number of edges in $`M`$ $$S(M)=\mathrm{\#}(M).$$ (1) Note that no action is associated with those links where the surfaces crosses itself. The partition function is given by $$Z(\beta )=\underset{M}{}e^{\beta S(M)}.$$ (2) We define an intermediate plane in $`𝐑^3`$ to be a plane that lies parallel to but in between planes where one of the coordinates takes an integer value. If we slice a surface $`M`$ (regarded as a subset of $`𝐑^3`$) by an intermediate plane we obtain a collection of links which we shall call generalized loops or simply loops. The collection of all possible loops that can arise in this way will be denoted $`\mathrm{\Pi }`$. A collection of links $`P`$ in an intermediate plane belongs to $`\mathrm{\Pi }`$ exactly when each vertex in $`P`$ belongs to an even number of links in $`P`$. We note that the loops $`\mathrm{\Pi }`$ coincide with the phase boundaries that arise in the two-dimensional Ising model. It will not come as a surprise that this close connection to the Ising model will make the present model exactly soluble in a very strong sense. For $`P\mathrm{\Pi }`$ we let $`\|P\|`$ denote the number of links in $`P`$. Note that if $`P_1,P_2\mathrm{\Pi }`$, then the symmetric difference $$(P_1P_2)(P_1P_2)P_1\mathrm{}P_2\mathrm{\Pi }$$ (3) is again a loop in $`\mathrm{\Pi }`$. It is easy to check that $$\|P_1\mathrm{}P_2\|=\|P_1\|+\|P_2\|2\|P_1P_2\|$$ (4) and $`d(P_1,P_2)=\|P_1\mathrm{}P_2\|`$ is a metric on $`\mathrm{\Pi }`$. Let $`k(P)`$ denote the number of corners of $`P`$, i.e. the number of vertices of order 2 whose adjacent links are at right angles. Let us denote by $``$ the space of real-valued functions on $`\mathrm{\Pi }`$. Let $`K_\beta `$ be the operator on $``$ whose matrix (kernel) is given by $$K_\beta (P_1,P_2)=\mathrm{exp}\left(\beta (k(P_1)+k(P_2)+2\|P_1\mathrm{}P_2\|)\right).$$ (5) With this notation the partition function of the model can be expressed as $$Z_N(\beta )=\mathrm{Tr}K_\beta ^N$$ (6) by slicing surfaces in intermediate planes orthogonal to the time direction. This partition function can be evaluated if we ignore the interesection term $`\|P_1P_2\|`$ in the exponent of $`K_\beta `$ or if we replace $`k(P_1)+k(P_2)`$ by $`k(P_1\mathrm{}P_2)`$ . In this approximation the model becomes a stack of noninteracting two-dimensional models. The critical beahaviour of the model is identical to that of the two-dimensional Ising model and the free energy can be computed exactly by standard methods . Here we will make an approximation different from the one of . We drop the curvature terms in the action and study the simplified transfer matrix $$K_\beta (P_1,P_2)=\mathrm{exp}\left(2\beta (\|P_1\mathrm{}P_2\|M^2)\right),$$ (7) where we have inserted a normalization factor $`M^2`$ in the exponent for later convenience. Dropping the curvature term is the same as neglecting the action associated with edges in the time-direction. We note the mathematical analogy of the transfer matrix (7) with the transition function $`\mathrm{exp}(\beta \|xy\|)`$, $`x,y𝐑^d`$, for random walk in $`𝐑^d`$. This makes it reasonable to regard $`K_\beta `$ as describing the diffusion of loops. With abuse of notation we denote the new transfer matrix (7) by the same symbol as the old one (5). This should not cause any confusion since we stick to the notation of Eq. (7) in the sequel. ## 3 Eigenvalues of the transfer matrix In this section we solve the eigenvalue problem $$\underset{Q\mathrm{\Pi }}{}K_\beta (P,Q)\psi _i(Q)=\mathrm{\Lambda }_i(\beta )\psi _i(P),$$ (8) $`\psi _i`$. We prove that only the eigenvalues depend on $`\beta `$, not the eigenfunctions. The eigenvalues can all be expressed in terms of the partition function, the internal energy and the spin correlation functions of the two-dimensional Ising model. The eigenfunctions are explicit functions on $`\mathrm{\Pi }`$ which can be normalized to take only the values $`1`$ and $`1`$. Note first that $`{\displaystyle \underset{Q}{}}K_\beta (P,Q)`$ $`=`$ $`{\displaystyle \underset{Q}{}}e^{2\beta (\|Q\|M^2)}`$ (9) $`=`$ $`\mathrm{\Lambda }_0(\beta ),`$ where we have in the first step shifted the summation variable from $`Q`$ to $`P\mathrm{}Q`$ (permissible since the mapping $`PP\mathrm{}Q`$ is bijective) and $`\mathrm{\Lambda }_0(\beta )`$ is the partition function of a two-dimensional Ising model on a periodic $`M\times M`$ lattice. This lattice will be denoted $`T_2`$. The loops $`Q`$ are the phase boundaries of the Ising model and the Ising spin variables sit on the lattice dual to $`T_2`$. It follows from Eq. (9) that the constant function is an eigenfunction of $`K_\beta `$ and $`\mathrm{\Lambda }_0(\beta )`$ is the corresponding eigenvalue. Since all matrix elements of $`K_\beta `$ are positive and all entries in the eigenvector corresponding to $`\mathrm{\Lambda }_0`$ have the same sign, it follows from the Perron–Frobenius theorem that $`\mathrm{\Lambda }_0`$ is simple and it is the largest eigenvalue of $`K_\beta `$. We conclude that the free energy per site in the present model is the same as the free energy per site in the two-dimensional Ising model. In particular, the two models have the same critical point and the same specific heat. However, the correlations are different, and we proceed to the calculation of the next eigenvalue of $`K_\beta `$. Let us introduce the notation $$P\|Q2M^22\|P\mathrm{}Q\|,$$ (10) so $`K_\beta (P,Q)=\mathrm{exp}\left(\beta P\|Q\right)`$. We find it convenient to think of $`P\|Q`$ as the ‘inner product’ between the loops $`P`$ and $`Q`$ even though the set of loops is of course not a linear space. We shall refer to $`P\|Q`$ as the invariant product on $`\mathrm{\Pi }`$. If $`P`$ is a loop we let $`\overline{P}`$ denote its complement, i.e. the loop made up of exactly those links in $`T_2`$ that $`P`$ does not contain. We let $`0`$ denote the empty loop and $`U`$ the loop that contains all links in the lattice. Then $`\overline{0}=U`$ and $`\overline{P}=P\mathrm{}U`$. The invariant product is clearly symmetric in its two arguments and it has the following properties: $`P\|P`$ $`=`$ $`2M^2`$ (11) $`P\|0`$ $`=`$ $`2M^22\|P\|`$ (12) $`P\|\overline{Q}`$ $`=`$ $`P\|Q.`$ (13) One can easily construct a finite dimensional inner product space which contains all loops in a natural fashion such that the inner product of loops coincides with the invariant product and the loops lie on a sphere of radius $`\sqrt{2}M`$. The invariant product is invariant under a nonabelian group $`𝒢`$ of motions in loop-space generated by * the translations of loops $$T_a:PP+a,$$ (14) $`aT_2`$, * ‘generalized antipodal maps’ $$A_Q:PP\mathrm{}Q,$$ (15) $`Q\mathrm{\Pi }`$, * reflections and rotations in $`T_2`$. Invariance means that $`gP\|gQ=P\|Q`$ for any $`g𝒢`$. As explained above we can view the set of loops as a sphere and we are free to regard the empty loop as the north-pole and the complete loop $`U`$ at the south-pole. Loops increase in size as one moves from north to south. The mapping $`P\overline{P}`$ is in this picture the usual antipodal map and the mappings $`A_Q`$ all have the property $`A_Q^2=I`$ where $`I`$ is the identity map. The group $`𝒢`$ acts on the functions in $``$ in a natural way: $$g\psi (P)=\psi (g^1P).$$ (16) Let $`O_n`$ denote the linear operator on $``$ with matrix elements $$O_n(P,Q)=P\|Q^n.$$ (17) Then $`O_n`$ is invariant under the action of $`𝒢`$ on $``$, i.e. $$gO_ng^1=O_n$$ (18) and the operators $`O_n`$ commute with each other. Since they are symmetric it follows that they have common eigenvectors and the same applies of course to the transfer matrix $$K_\beta =\underset{n=0}{\overset{\mathrm{}}{}}\frac{\beta ^n}{n!}O_n.$$ (19) This proves that the eigenvectors of $`K_\beta `$ are independent of $`\beta `$. In view of the invariance of the transfer matrix under $`𝒢`$ it is natural to look for eigenvectors which depend on $`P`$ only via the invariant product. Let us define $$\psi _Q(P)=P\|Q.$$ (20) We claim that $$K_\beta \psi _Q=\mathrm{\Lambda }_1(\beta )\psi _Q$$ (21) for any $`Q\mathrm{\Pi }`$, where $$\mathrm{\Lambda }_1(\beta )=\frac{1}{2M^2}\underset{P}{}e^{\beta P\|0}P\|0,$$ (22) so the eigenvalue ratio $`\mathrm{\Lambda }_1\mathrm{\Lambda }_0^1`$ is minus the internal energy per link of the Ising model on $`T_2`$ . In order to prove Eqs. (21) and (22) it is convenient to introduce a function $`\eta _{\mathrm{}}`$ on $`\mathrm{\Pi }`$, where $`\mathrm{}`$ is a link in the lattice $`T_2`$, defined as $$\eta _{\mathrm{}}(P)=\{\begin{array}{cc}\hfill 1\text{if}lP,& \\ \hfill 1\text{if}lP.& \end{array}$$ (23) Note that $$P\|Q=\underset{\mathrm{}}{}\eta _{\mathrm{}}(P\mathrm{}Q),$$ (24) where the sum runs over all links in the lattice $`T_2`$. Consider the function $$\mathrm{\Phi }_{Q,l}:P\eta _{\mathrm{}}(P\mathrm{}Q)$$ (25) on $`\mathrm{\Pi }`$. If $`\mathrm{}Q`$ then $`\mathrm{}P\mathrm{}Q`$ if and only if $`\mathrm{}P`$. Similarly, if $`\mathrm{}Q`$ then $`\mathrm{}P\mathrm{}Q`$ if and only if $`\mathrm{}P`$. We conclude that $$\eta _{\mathrm{}}(P\mathrm{}Q)=\eta _{\mathrm{}}(P)\eta _{\mathrm{}}(Q).$$ (26) It follows that $`{\displaystyle \underset{P}{}}e^{\beta 0\|P}\eta _{\mathrm{}}(P\mathrm{}Q)`$ $`=`$ $`{\displaystyle \underset{P}{}}e^{\beta 0\|P}\eta _{\mathrm{}}(P)\eta _{\mathrm{}}(Q)`$ (27) $`=`$ $`{\displaystyle \frac{1}{2M^2}}{\displaystyle \underset{P}{}}e^{\beta 0\|P}0\|P\eta _{\mathrm{}}(Q).`$ The last equality is obtained by observing that the sum $$\underset{P}{}e^{\beta 0\|P}\eta _{\mathrm{}}(P)$$ (28) is independent of $`\mathrm{}`$ due to the translational and rotational invariance of the invariant product and using Eq. (24). This proves Eqs. (21) and (22) since it suffices to verify $$\underset{P}{}K_\beta (P^{},P)P\|Q=\mathrm{\Lambda }_1P^{}\|Q$$ (29) for $`P^{}=0`$ due to the invariance of $`K_\beta `$ under $`𝒢`$. In fact we have shown that the functions $`\mathrm{\Phi }_{Q,\mathrm{}}`$ are all eigenfunctions of $`K_\beta `$ with the eigenvalue $`\mathrm{\Lambda }_1`$. We now use the elementary functions $`\mathrm{\Phi }_{Q,\mathrm{}}`$ to construct the higher eigenfunctions of $`K_\beta `$. Let $`\mathrm{}_1`$ and $`\mathrm{}_2`$ be two distinct links. Define $$\mathrm{\Phi }_{Q_1Q_2,\mathrm{}_1\mathrm{}_2}(P)=\mathrm{\Phi }_{Q_1,\mathrm{}_1}(P)\mathrm{\Phi }_{Q_2,\mathrm{}_2}(P).$$ (30) Then, by Eq. (26), $$\left(K_\beta \mathrm{\Phi }_{Q_1Q_2,\mathrm{}_1\mathrm{}_2}\right)(0)=\mathrm{\Lambda }_2(\mathrm{}_1,\mathrm{}_2)\mathrm{\Phi }_{Q_1Q_2,\mathrm{}_1\mathrm{}_2}(0),$$ (31) i.e. $`\mathrm{\Phi }_{Q_1Q_2,\mathrm{}_1\mathrm{}_2}`$ is an eigenfunction of $`K_\beta `$ with eigenvalue $$\mathrm{\Lambda }_2(\mathrm{}_1,\mathrm{}_2)=\underset{P}{}e^{\beta 0\|P}\eta _\mathrm{}_1(P)\eta _\mathrm{}_2(P).$$ (32) Note that $`\mathrm{\Lambda }_2(\mathrm{}_1,\mathrm{}_2)`$ depends on the links $`\mathrm{}_1`$ and $`\mathrm{}_2`$ but only on their relative orientation and the distance between them. Of course the eigenvalue also depends on $`\beta `$ but we suppress this from our notation. Similarly, if the links $`\mathrm{}_1,\mathrm{}_2,\mathrm{},\mathrm{}_n`$ are all distinct, then $$\mathrm{\Phi }_{Q_1\mathrm{}Q_n,\mathrm{}_1\mathrm{}\mathrm{}_n}(P)=\underset{i=1}{\overset{n}{}}\mathrm{\Phi }_{Q_i,\mathrm{}_i}(P)$$ (33) is an eigenfunction of $`K_\beta `$ with the eigenvalue $$\mathrm{\Lambda }_n(\mathrm{}_1,\mathrm{},\mathrm{}_n)=\underset{P}{}e^{\beta 0\|P}\underset{i=1}{\overset{n}{}}\eta _\mathrm{}_i(P).$$ (34) The eigenvalue is symmetric under permutations of the $`\mathrm{}_i`$’s and simultaneous lattice rotations or translations of the links. If $`\mathrm{}`$ is a link in the lattice $`T_2`$ let $`\sigma _{\mathrm{}}^1`$ and $`\sigma _{\mathrm{}}^2`$ be the two Ising spin variables on the dual sites adjacent to the link $`\mathrm{}`$. Then $`\eta _{\mathrm{}}(P)=\sigma _{\mathrm{}}^1(P)\sigma _{\mathrm{}}^2(P)`$ where $`\sigma _{\mathrm{}}^i(P)`$ is the value taken by $`\sigma _{\mathrm{}}^i`$ in the spin configuration corresponding to the loop $`P`$. The expectation value for an Ising model on $`T_2`$ is given by $$()=\mathrm{\Lambda }_0^1\underset{P}{}e^{\beta P\|0}()$$ (35) so $$\mathrm{\Lambda }_n(\mathrm{}_1,\mathrm{},\mathrm{}_n)=\mathrm{\Lambda }_0\underset{i=1}{\overset{n}{}}\sigma _\mathrm{}_i^1\sigma _\mathrm{}_i^2.$$ (36) The correlation inequalities $$\underset{i=1}{\overset{n}{}}\sigma _\mathrm{}_i^1\sigma _\mathrm{}_i^2\underset{i=1}{\overset{k}{}}\sigma _\mathrm{}_i^1\sigma _\mathrm{}_i^2\underset{i=k+1}{\overset{n}{}}\sigma _\mathrm{}_i^1\sigma _\mathrm{}_i^2$$ (37) now imply $$\mathrm{\Lambda }_n(\mathrm{}_1,\mathrm{},\mathrm{}_n)\mathrm{\Lambda }_0\mathrm{\Lambda }_k(\mathrm{}_1,\mathrm{},\mathrm{}_k)\mathrm{\Lambda }_{nk}(\mathrm{}_{k+1},\mathrm{},\mathrm{}_n).$$ (38) We expect that $$\mathrm{\Lambda }_n(\mathrm{}_1,\mathrm{},\mathrm{}_n)<\mathrm{\Lambda }_1,$$ (39) for any $`n>1`$ but do not have a general proof of this inequality. It can be checked in special cases using the explicit form of the two spin correlation function and the cluster property of correlations. ## 4 Multiplicities As remarked above, the largest eigenvalue $`\mathrm{\Lambda }_0`$ is simple by the Perron–Frobenius theorem. Let us consider the first nontrivial eigenvalue $`\mathrm{\Lambda }_1`$ with eigenfunctions $`\mathrm{\Phi }_{Q,\mathrm{}}`$. Since $`\mathrm{\Phi }_{Q,\mathrm{}}`$ is a multiple of $`\mathrm{\Phi }_{0,\mathrm{}}`$ by the constant $`\eta _{\mathrm{}}(Q)`$ it suffices to consider the functions $`\mathrm{\Phi }_{0,\mathrm{}}=\eta _{\mathrm{}}`$. We claim that these functions are linearly independent so the multiplicity of $`\mathrm{\Lambda }_1`$ is at least $`2M^2`$. Suppose there are constants $`c_{\mathrm{}}`$ such that $$\underset{\mathrm{}}{}c_{\mathrm{}}\eta _{\mathrm{}}(P)=0$$ (40) for any $`P\mathrm{\Pi }`$. We adopt the convention that sums on $`\mathrm{}`$ run over all links in $`T_2`$ unless otherwise specified. Then $$\underset{\mathrm{}P}{}c_{\mathrm{}}+\underset{\mathrm{}P}{}c_{\mathrm{}}=0.$$ (41) By taking $`P=0`$ we also obtain $$\underset{\mathrm{}}{}c_{\mathrm{}}=0.$$ (42) It follows that $$\underset{\mathrm{}P}{}c_{\mathrm{}}=0$$ (43) for any $`P\mathrm{\Pi }`$. Let now $`x`$ and $`y`$ be two different lattice points in $`T_2`$ and suppose $`P`$ is made up of two simple nonintersecting curves $`\gamma _1`$ and $`\gamma _2`$ from $`x`$ to $`y`$. Then clearly $$\underset{\mathrm{}\gamma _1}{}c_{\mathrm{}}=\underset{\mathrm{}\gamma _2}{}c_{\mathrm{}}.$$ (44) Let $`\gamma _3`$ be one more simple curve from $`x`$ to $`y`$ which also avoids $`\gamma _1`$ and $`\gamma _2`$. Then we can join $`\gamma _3`$ with either $`\gamma _1`$ or $`\gamma _2`$ to make a closed curve (i.e. a loop) in $`\mathrm{\Pi }`$. We conclude that $$\underset{\mathrm{}\gamma _1}{}c_{\mathrm{}}=\underset{\mathrm{}\gamma _3}{}c_{\mathrm{}}=\underset{\mathrm{}\gamma _2}{}c_{\mathrm{}}$$ (45) which implies that $$\underset{\mathrm{}\gamma _i}{}c_{\mathrm{}}=0$$ (46) for $`i=1,2,3`$. Now take $`x`$ and $`y`$ to be nearest neighbours and $`\gamma _1`$ the curve that joins them by one link $`\mathrm{}_1`$. It follows that $`c_\mathrm{}_1=0`$ and hence $`c_{\mathrm{}}=0`$ for all $`\mathrm{}`$ since $`x`$ and $`y`$ are arbitrary. We could have obtained the above result more easily by noting that the functions $`\eta _{\mathrm{}}`$ are mutually orthogonal in the natural inner product on $``$ which we denote by $`(,)`$. This follows from the fact that $`\eta _\mathrm{}_1\eta _\mathrm{}_2`$ with $`\mathrm{}_1\mathrm{}_2`$ is an eigenfunction of $`K_\beta `$ with an eigenvalue $`\mathrm{\Lambda }_2(\mathrm{}_1,\mathrm{}_2)\mathrm{\Lambda }_0`$ so $`\eta _\mathrm{}_1\eta _\mathrm{}_2`$ is orthogonal to the constant function, i.e. $$(1,\eta _\mathrm{}_1\eta _\mathrm{}_2)=\underset{P}{}\eta _\mathrm{}_1(P)\eta _\mathrm{}_2(P)=0.$$ (47) Hence, $$(\eta _\mathrm{}_1,\eta _\mathrm{}_2)=\delta _{\mathrm{}_1\mathrm{}_2}\mathrm{\hspace{0.17em}2}^{M^2}$$ (48) since the total number of loops in $`\mathrm{\Pi }`$ is equal to $`2^{M^2}`$. This can also be seen by a direct calculation. If the inequality (39) is valid then the multiplicity of $`\mathrm{\Lambda }_1`$ is exactly $`2M^2`$. This follows from the fact that the functions $`\mathrm{\Phi }_{Q_1\mathrm{}Q_n,\mathrm{}_1\mathrm{}\mathrm{}_n}`$ span $``$. We now prove this spanning property. Let $`\mathrm{}_1,\mathrm{}_2,\mathrm{}_3,\mathrm{}_4`$ be four different links which contain the same vertex. We shall call such a collection of links a star. In this case $$\eta _\mathrm{}_1(P)\eta _\mathrm{}_2(P)\eta _\mathrm{}_3(P)\eta _\mathrm{}_4(P)=1$$ (49) for any loop $`P`$ since $`P`$ contains $`0`$, $`2`$ or $`4`$ of the links $`\mathrm{}_1,\mathrm{},\mathrm{}_4`$. If $`\mathrm{}_1,\mathrm{}_2,\mathrm{}_3,\mathrm{}_4`$ are a star and $`n4`$ it follows that $$\mathrm{\Phi }_{Q_1\mathrm{}Q_n,\mathrm{}_1\mathrm{}\mathrm{}_n}=\pm \mathrm{\Phi }_{Q_5\mathrm{}Q_n,\mathrm{}_5\mathrm{}\mathrm{}_n}.$$ (50) Let $`=\{\mathrm{}_1,\mathrm{},\mathrm{}_n\}`$ be a collection of links in $`T_2`$ and define $$E_{}(P)=\underset{\mathrm{}}{}\eta _{\mathrm{}}(P).$$ (51) Clearly $`E_{}=\pm \mathrm{\Phi }_{Q_1\mathrm{}Q_n,\mathrm{}_1\mathrm{}\mathrm{}_n}`$ so in order to prove the spanning property it suffices to identify $`2^{M^2}`$ mutually orthogonal functions of the form $`E_{}`$. We adopt the convention $`E_{\mathrm{}}=1`$. If $``$ and $`^{}`$ are two collections of links we say that they are equivalent if the symmetric difference $`\mathrm{}^{}`$ is a star or the symmetric difference of two or more stars. This defines an equivalence relation on the set of all collections of links. If $``$ is a collection of links let $`[]`$ denote the equivalence class of $``$. It is easy to see that the total number of elements in each equivalence class is $`2^{M^2}`$ and the number of different equivalence classes is also $`2^{M^2}`$ since the total number of collections of links from $`T_2`$ is $`4^{M^2}`$. If we choose one $`_i`$ from each equivalence class $`[_i]`$, $`i=1,\mathrm{},2^{M^2}`$, then $$(E__i,E__j)=\underset{P}{}\underset{\mathrm{}_{ij}}{}\eta _{\mathrm{}}(P),$$ (52) where $`_{ij}`$ is a maximal subset of $`_i\mathrm{}_j`$ which contains no star. Hence, for $`ij`$, $$(E__i,E__j)=(1,E_{_{ij}})=0$$ (53) since $`_{ij}`$ is nonempty for $`ij`$. This proves that $`\{E__i\}`$ is a family of mutually orthogonal functions. ## 5 Discussion Using the results derived in the previous section it is quite easy to calculate correlation function, i.e. functions of the form $$G_\beta ^{(N)}(P,Q)=\underset{M=PQ}{}e^{\beta S_1(M)},$$ (54) where the loops $`P`$ and $`Q`$ lie in two intermediate constant time planes separated by $`N`$ lattice spacings. In Eq. (54) we sum over all surfaces whose intersection with the two intermediate planes are $`P`$ and $`Q`$ and $`S_1(M)`$ is a modified action functional, counting only edges that are orthogonal to the time direction. We have $$G_\beta ^{(N)}(P,Q)=(\delta _P,K_\beta ^N\delta _Q),$$ (55) where $`\delta _P`$ and $`\delta _Q`$ are delta functions in $``$. Up to normalization we can interpret $`G_\beta ^{(N)}(P,Q)`$ as the probability of having the loop $`P`$ at time $`N`$ given that we have $`Q`$ at time $`0`$. An easy calculation gives $$G_\beta ^{(N)}(P,Q)=2^{M^2}\underset{i=1}{\overset{2^{M^2}}{}}E__i(P\mathrm{}Q)\mathrm{\Lambda }__i^N,$$ (56) where the sum runs over the orthogonal family of functions constructed in the last section and $`\mathrm{\Lambda }__i`$ is the eigenvalue of $`K_\beta `$ corresponding to $`E__i`$. Unfortunately it is not clear how to extend the analysis of the present paper to include the curvature terms in the action of the full model. The original transfer matrix (5) is not invariant under the group $`𝒢`$, only translations and rotations remain symmetries. The constant function is not an eigenfunction any more. The only result that survives is the fact that the largest eigenvalue is in this case also simple by the Perron–Frobenius theorem. However, we have managed to solve exactly a three dimensional lattice model that describes the diffusion of loops. As a surface model it is not isotropic but can be viewed as a stack of two-dimensional Ising systems with an intereaction that is sufficiently simple for us to solve the model exactly. Acknowledgement. T. J. is indebted to NRCPS “Demokritos” and the CERN Theory Division for hospitality.
no-problem/9907/cond-mat9907272.html	ar5iv	text	# Comment on “Temperature Dependence of the Josephson Current” In a recent letter , Overhauser has claimed that the accepted calculations of the temperature dependence of the Josephson current are incorrect. His contention is that broken pair states, in the BCS terminology, should not be included as initial states in calculating the contribution of tunneling to the free energy of two weakly coupled superconductors. In fact, all terms that depend on the difference of the phases of the order parameters in the two superconductors must be included and such phase dependences are present (via intermediate states) in the omitted terms. As Overhauser himself shows, including them reproduces the usual result. This is really all that needs to be said. The following remarks may, however, clarify the issue—if further clarification is needed. Consider the anomalous averages $$F^>(𝐤,t,t^{})a_𝐤(t)a_𝐤(t^{}),\mathrm{and}F^<(𝐤,t,t^{})a_𝐤(t^{})a_𝐤(t).$$ (1) The difference between the conventional ($`C`$) and Overhauser ($`O`$) way of evaluating these averages may be summarized in the following relations for their Fourier transforms with respect to $`tt^{}`$: $$F_C^>(𝐤,\omega )=B(𝐤,\omega )[1f(\omega )],F_C^<(𝐤,\omega )=B(𝐤,\omega )f(\omega ),$$ (2) as opposed to $$F_O^>(𝐤,\omega )=B(𝐤,\omega )[1f(\omega )]^2,F_O^<(𝐤,\omega )=B(𝐤,\omega )f(\omega )^2.$$ (3) Above, $`B(𝐤,\omega )`$ is an odd function of $`\omega `$ whose form is not essential for the argument and $`f(\omega )=[\mathrm{exp}(\beta \omega )+1]^1`$ is the Fermi function, with $`\beta `$ the reciprocal of the temperature. Note that the conventional calculation yields forms which agree with the general fluctuation-dissipation requirement $$F^>(\omega )=e^{\beta \omega }F^<(\omega ),$$ (4) whereas these quantitites calculated according to the Overhauser prescription do not, showing that the ad hoc suppression of states (which treats initial and intermediate states asymmetrically) is unphysical. Work supported in part by the NSF under grant DMR-9805613.
no-problem/9907/cond-mat9907068.html	ar5iv	text	# Mean-field theory for scale-free random networks ## 1 Introduction Contemporary science has been particularly successful in addressing the physical properties of systems that are composed of many identical elements interacting through mainly local interactions. For example, many successes of materials science and solid state physics are based on the fact that most solids are made of relatively few types of elements that exhibit spatial order by forming a crystal lattice. Furthermore, these elements are coupled by local, nearest neighbor interactions. However, the inability of contemporary science to describe systems composed of non-identical elements that have diverse and nonlocal interactions currently limits advances in many disciplines, ranging from molecular biology to computer science . The difficulty in describing these systems lies partly in their topology: many of them form complex networks, whose vertices are the elements of the system and edges represent the interactions between them. For example, living systems form a huge genetic network, whose vertices are proteins, the edges representing the chemical interactions between them . Similarly, a large network is formed by the nervous system, whose vertices are the nerve cells, connected by axons . But equally complex networks occur in social science, where vertices are individuals, organizations or countries, and the edges characterize the social interaction between them , in the business world, where vertices are companies and edges represent diverse business relationships, or describe the world wide web (www), whose vertices are HTML documents connected by links pointing from one page to another . Due to their large size and the complexity of the interactions, the topology of these networks is largely unknown or unexplored. Traditionally, networks of complex topology have been described using the random graph theory of Erdős and Rényi (ER) . However, while it has been much investigated in combinatorial graph theory, in the absence of data on large networks the predictions of the ER theory were rarely tested in the real world. This is changing very fast lately: driven by the computerization of data acquisition, topological information on various real world networks is increasingly available. Due to the importance of understanding the topology of some of these systems, it is likely that in the near future we will witness important advances in this direction. Furthermore, it is also possible that seemingly random networks in Nature have rather complex internal structure, that cover generic features, common to many systems. Uncovering the universal properties characterizing the formation and the topology of complex networks could bring about the much coveted revolution beyond reductionism . A major step in the direction of understanding the generic features of network development was the recent discovery of a surprising degree of self-organization characterizing the large scale properties of complex networks. Exploring several large databases describing the topology of large networks, that span as diverse fields as the www or the citation patterns in science, recently Barabási and Albert (BA) have demonstrated that independently of the nature of the system and the identity of its constituents, the probability $`P(k)`$ that a vertex in the network is connected to $`k`$ other vertices decays as a power-law, following $`P(k)k^\gamma `$. The generic feature of this observation was supported by four real world examples. In the collaboration graph of movie actors, each actor is represented by a vertex, two actors being connected if they were casted in the same movie. The probability that an actor has $`k`$ links was found to follow a power-law for large $`k`$, i.e. $`P(k)k^{\gamma _{actor}}`$, where $`\gamma _{actor}=2.3\pm 0.1.`$ A rather complex network with over $`300`$ million vertices is the www, where a vertex is a document and the edges are the links pointing from one document to another. The topology of this graph determines the web’s connectivity and, consequently, our effectiveness in locating information on the www . Information about $`P(k)`$ can be obtained using robots , indicating that the probability that $`k`$ documents point to a certain webpage follows a power-law, with $`\gamma _{www}^{in}=2.1`$ , and the probability that a certain web document contains $`k`$ outgoing links follows a similar distribution, with $`\gamma _{www}^{out}=2.45`$. A network whose topology reflects the historical patterns of urban and industrial development is the electrical powergrid of western US, the vertices representing generators, transformers and substations, the edges corresponding to the high voltage transmission lines between them . The connectivity distribution is again best approximated with a power-law with an exponent $`\gamma _{power}4`$. Finally, a rather large, complex network is formed by the citation patterns of the scientific publications, the vertices standing for papers, the edges representing links to the articles cited in a paper. Recently Redner has shown that the probability that a paper is cited $`k`$ times (representing the connectivity of a paper within the network) follows a power-law with exponent $`\gamma _{cite}=3`$. These results offered the first evidence that large networks self-organize into a scale-free state, a feature unexpected by all existing random network models. To understand the origin of this scale invariance, BA have shown that existing network models fail to incorporate two key features of real networks: First, networks continuously grow by the addition of new vertices, and second, new vertices connect preferentially to highly connected vertices. Using a model incorporating these ingredients, they demonstrated that the combination of growth and preferential attachment is ultimately responsible for the scale-free distribution and power-law scaling observed in real networks. The goal of the present paper is to investigate the properties of the scale-free model introduced by BA , aiming to identify its scaling properties and compare them with other network models intended to describe the large scale properties of random networks. We present a mean field theory that allows us to predict the dynamics of individual vertices in the system, and to calculate analytically the connectivity distribution. We apply the same method to uncover the scaling properties of two versions of the BA model, that are missing one of the ingredients needed to reproduce the power-law scaling. Finally, we discuss various extensions of the BA model, that could be useful in addressing the properties of real networks. ## 2 Earlier network models ### 2.1 The Erdős-Rényi model Probably the oldest and most investigated random network model has been introduced by Erdős and Rényi (ER) , who were the first to study the statistical aspects of random graphs by probabilistic methods. In the model we start with $`N`$ vertices and no bonds (see Fig. 1a). With probability $`p_{\mathrm{ER}}`$, we connect each pair of vertices with a line (bond or edge), generating a random network. The greatest discovery of ER was that many properties of these graphs appear quite suddenly, at a threshold value of $`p_{\mathrm{ER}}(N)`$. A property of great importance for the topology of the graph is the appearance of trees and cycles. A tree of order $`k`$ is a connected graph with $`k`$ vertices and $`k1`$ edges, while a cycle of order $`k`$ is a cyclic sequence of $`k`$ edges such that every two consecutive edges and only these have a common vertex. ER have demonstrated that if $`p_{\mathrm{ER}}c/N`$ with $`c<1`$, then almost all vertices belong to isolated trees, but there is an abrupt change at $`p_{\mathrm{ER}}1/N`$, (i.e. $`c=1`$), when cycles of all orders appear. In the physical literature the ER model is often referred to as infinite dimensional percolation, that is known to belong to the universality class of mean field percolation . In this context $`p_c1/N`$ is the percolation threshold of the system. For $`p<p_c`$ the system is broken into many small clusters, while at $`p_c`$ a large cluster forms, that in the asymptotic limit contains all vertices. To compare the ER model with other network models, we need to focus on the connectivity distribution. As Erdős and Rényi have shown in their seminal work, the probability that a vertex has $`k`$ edges follows the Poisson distribution $$P(k)=e^\lambda \lambda ^k/k!,$$ (1) where $$\lambda =\left(\begin{array}{c}N1\\ k\end{array}\right)p_{\mathrm{ER}}^k(1p_{\mathrm{ER}})^{N1k},$$ (2) its expectation value being $`(N1)p_{\mathrm{ER}}`$. For sake of comparison, in Fig. 2a we show $`P(k)`$ for different values of $`p_{\mathrm{ER}}`$. ### 2.2 The small-world model Aiming to describe the transition from a locally ordered system to a random network, recently Watts and Strogatz (WS) have introduced a new model , that is often referred to as small-world network. The topological properties of the network generated by this model have been the subject of much attention lately . The WS model begins with a one-dimensional lattice of $`N`$ vertices with bonds between the nearest and next-nearest neighbors (in general, the algoritm can include neighbors up to an order $`n`$, such that the coordination number of a vertex is $`z=2n`$) and periodic boundary conditions (see Fig. 1b). Then each bond is rewired with probability $`p_{\mathrm{WS}}`$, where rewiring in this context means shifting one end of the bond to a new vertex chosen at random from the whole system, with the constraint that no two vertices can have more than one bond, and no vertex can have a bond with itself. For $`p_{\mathrm{WS}}=0`$ the lattice is highly clustered, and the average distance between two vertices $`l`$ grows linearly with $`N`$, while for $`p_{\mathrm{WS}}=1`$ the system becomes a random graph, poorly clustered and $`l`$ grows logarithmically with $`N`$. WS found that in the interval $`0<p_{\mathrm{WS}}<0.01`$ the model exhibits small-world properties, ($`ll_{random}`$), while it remains highly clustered. The connectivity distribution of the WS model depends strongly on $`p_{\mathrm{WS}}`$: for $`p_{\mathrm{WS}}=0`$ we have $`P(k)=\delta (kz)`$, where $`z`$ is the coordination number of the lattice, while for finite $`p_{\mathrm{WS}}`$, $`P(k)`$ is still peaked around $`z`$, but it gets broader. Ultimately, as $`p_{\mathrm{WS}}1`$, the distribution $`P(k)`$ approaches the connectivity distribution of a random graph, i.e. the distribution converges to that obtained for the ER model with $`p_{\mathrm{ER}}=z/N`$ (see Fig. 2b). ## 3 The scale-free model A common feature of the models discussed in the previous section is that they both predict that the probability distribution of the vertex connectivity, $`P(k)`$, has an exponential cutoff, and has a characteristic size $`k`$, that depends on $`p`$. In contrast, as we mentioned in the Introduction, many systems in nature have the common property that $`P(k)`$ is free of scale, following a power-law distribution over many orders of magnitude. To understand the origin of this discrepancy, BA have argued that there are two generic aspects of real networks that are not incorporated in these models . First, both models assume that we start with a fixed number ($`N`$) of vertices, that are then randomly connected (ER model), or reconnected (SW model), without modifying $`N`$. In contrast, most real world networks are open, i.e. they form by the continuous addition of new vertices to the system, thus the number of vertices, $`N`$, increases throughout the lifetime of the network. For example, the actor network grows by the addition of new actors to the system, the www grows exponentially in time by the addition of new web pages, the research literature constantly grows by the publication of new papers. Consequently, a common feature of these systems is that the network continuously expands by the addition of new vertices that are connected to the vertices already present in the system. Second, the random network models assume that the probability that two vertices are connected is random and uniform. In contrast, most real networks exhibit preferential connectivity. For example, a new actor is casted most likely in a supporting role, with more established, well known actors. Similarly, a newly created webpage will more likely include links to well known, popular documents with already high connectivity, or a new manuscript is more likely to cite a well known and thus much cited paper than its less cited and consequently less known peer. These examples indicate that the probability with which a new vertex connects to the existing vertices is not uniform, but there is a higher probability to be linked to a vertex that already has a large number of connections. The scale-free model introduced by BA, incorporating only these two ingredients, naturally leads to the observed scale invariant distribution. The model is defined in two steps (see Fig. 3): (1) Growth: Starting with a small number ($`m_0`$) of vertices, at every timestep we add a new vertex with $`m`$($`m_0`$) edges (that will be connected to the vertices already present in the system). (2) Preferential attachment: When choosing the vertices to which the new vertex connects, we assume that the probability $`\mathrm{\Pi }`$ that a new vertex will be connected to vertex $`i`$ depends on the connectivity $`k_i`$ of that vertex, such that $$\mathrm{\Pi }(k_i)=k_i/\underset{j}{}k_j.$$ (3) After $`t`$ timesteps the model leads to a random network with $`N=t+m_0`$ vertices and $`mt`$ edges. As Fig. 4a shows, this network evolves into a scale-invariant state, the probability that a vertex has $`k`$ edges following a power-law with an exponent $`\gamma _{model}=2.9\pm 0.1`$. The scaling exponent is independent of $`m`$, the only parameter in the model. Since the power-law observed for real networks describes systems of rather different sizes at different stages of their development, one expects that a correct model should provide a distribution whose main features are independent of time. Indeed, as Fig. 4b demonstrates, $`P(k)`$ is independent of time (and, subsequently, independent of the system size $`N=m_0+t`$), indicating that despite its continuous growth, the system organizes itself into a scale-free stationary state. We next describe a method to calculate analytically the probability $`P(k)`$, allowing us to determine exactly the scaling exponent $`\gamma `$. The combination of growth and preferential attachment leads to an interesting dynamics of the individual vertex connectivities. The vertices that have the most connections are those that have been added at the early stages of the network development, since vertices grow proportionally to their connectedness relative to the rest of the vertices. Thus some of the oldest vertices have a very long time to acquire links, being responsible for the high-$`k`$ part of $`P(k)`$. The time dependence of the connectivity of a given vertex can be calculated analytically using a mean-field approach. We assume that $`k`$ is continuous, and thus the probability $`\mathrm{\Pi }(k_i)=k_i/_jk_j`$ can be interpreted as a continuous rate of change of $`k_i`$. Consequently, we can write for a vertex $`i`$ $$\frac{k_i}{t}=A\mathrm{\Pi }(k_i)=A\frac{k_i}{_{j=1}^{m_0+t1}k_j}.$$ (4) Taking into account that $`_jk_j=2mt`$ and the change in connectivities at a time step is $`\mathrm{\Delta }(k)=m`$, we obtain that $`A=m`$, leading to $$\frac{k_i}{t}=\frac{k_i}{2t}.$$ (5) The solution of this equation, with the initial condition that vertex $`i`$ was added to the system at time $`t_i`$ with connectivity $`k_i(t_i)=m`$, is $$k_i(t)=m\left(\frac{t}{t_i}\right)^{0.5}.$$ (6) As the inset of Fig. 4b shows, the numerical results are in good agreement with this prediction. Thus older (smaller $`t_i`$) vertices increase their connectivity at the expense of the younger (larger $`t_i`$) vertices, leading with time to some vertices that are highly connected, a “rich-gets-richer” phenomenon that can be easily detected in real networks. Furthermore, this property can be used to calculate $`\gamma `$ analytically. Using (6), the probability that a vertex has a connectivity $`k_i(t)`$ smaller than $`k`$, $`P(k_i(t)<k)`$, can be written as $$P(k_i(t)<k)=P(t_i>\frac{m^2t}{k^2}).$$ (7) Assuming that we add the vertices at equal time intervals to the system, the probability density of $`t_i`$ is $$P_i(t_i)=\frac{1}{m_0+t}.$$ (8) Substituting this into Eq. (4) we obtain that $$P(t_i>\frac{m^2t}{k^2})=1P(t_i\frac{m^2t}{k^2})=1\frac{m^2t}{k^2(t+m_0)}.$$ (9) The probability density for $`P(k)`$ can be obtained using $$P(k)=\frac{P(k_i(t)<k)}{k}=\frac{2m^2t}{m_0+t}\frac{1}{k^3},$$ (10) predicting $$\gamma =3,$$ (11) independent of $`m`$. Furthermore, Eq. (10) also predicts that the coefficient $`A`$ of the power-law distribution, $`P(k)Ak^\gamma `$, is proportional to the square of the average connectivity of the network, i.e., $`Am^2`$. In the inset of Fig. 4a we show $`P(k)/2m^2`$ vs. $`k`$. The curves obtained for different $`m`$ collapse into a single one, supporting the analytical result (10). ## 4 Limiting cases of the scale-free model ### 4.1 Model A The development of the power-law scaling in the scale-free model indicates that growth and preferential attachment play an important role in network development. To verify that both ingredients are necessary, we investigated two variants of the BA model. The first variant, that we refer to as model A, keeps the growing character of the network, but preferential attachment is eliminated. The model is defined as follows (see Fig. 3b): (1) Growth : Starting with a small number of vertices ($`m_0`$), at every time step we add a new vertex with $`m(m_0)`$ edges. (2) Uniform attachment : We assume that the new vertex connects with equal probability to the vertices already present in the system, i.e. $`\mathrm{\Pi }(k_i)=1/(m_0+t1)`$, independent of $`k_i`$. Fig. 5a shows the probability $`P(k)`$ obtained for different values of $`m`$, indicating that in contrast with the scale-free model, $`P(k)`$ has an exponential form $$P(k)=B\mathrm{exp}(\beta k)$$ (12) We can use the mean field arguments developed in the previous section to calculate analytically the expression for $`P(k)`$. The rate of change of the connectivity of vertex $`i`$ in this case is given by $$\frac{k_i}{t}=A\mathrm{\Pi }(k_i)=\frac{A}{m_0+t1}.$$ (13) At one timestep $`\mathrm{\Delta }(k)=m`$, implying that $`A=m`$. Solving the equation for $`k_i`$, and taking into account that $`k_i(t_i)=m`$, we obtain $$k_i=m\left(\mathrm{ln}(m_0+t1)\mathrm{ln}(m_0+t_i1)+1\right),$$ (14) a logarithmic increase with time, verified by the numerical simulations (see Fig. 5b). The probability that vertex $`i`$ has connectivity $`k_i(t)`$ smaller than $`k`$ is $$P(k_i(t)<k)=P\left(t_i>(m_0+t1)\mathrm{exp}(1\frac{k}{m})m_0+1\right).$$ (15) Assuming that we add the vertices uniformly to the system, we obtain that $`P\left(t_i>(m_0+t1)\mathrm{exp}(1{\displaystyle \frac{k}{m}})m_0+1\right)`$ $`=1{\displaystyle \frac{(m_0+t1)\mathrm{exp}(1\frac{k}{m})m_0+1}{m_0+t}}.`$ (16) Using Eq. (10) and assuming long times, we obtain $$P(k)=\frac{e}{m}\mathrm{exp}(\frac{k}{m}),$$ (17) indicating that in (12) the coefficients are $$B=\frac{e}{m},\beta =\frac{1}{m}.$$ (18) Consequently, the vertices in the model have the characteristic connectivity $$k^{}=\frac{1}{\beta }=m,$$ (19) which coincides with half of the average connectivities of the vertices in the system, since $`k=2m`$. As the inset of Fig. 5a demonstrates the numerical results approach asymptotically the theoretical predictions. The exponential character of the distribution for this model indicates that the absence of preferential attachment eliminates the scale-free feature of the BA model. ### 4.2 Model B This model tests the hypothesis that the growing character of the model is essential to sustain the scale-free state observed in the real systems. Model B is defined as follows (see Fig. 3c): We start with $`N`$ vertices and no edges. At each time step we randomly select a vertex and connect it with probability $`\mathrm{\Pi }(k_i)=k_i/_jk_j`$ to vertex $`i`$ in the system. Consequently, in comparison with the BA model, this variant eliminates the growth process, the numbers of vertices staying constant during the network evolution. While at early times the model exhibits power-law scaling (see Fig. 6, $`P(k)`$ is not stationary: Since $`N`$ is constant, and the number of edges increases with time, after $`TN^2`$ timesteps the system reaches a state in which all vertices are connected. The time-evolution of the individual connectivities can be calculated analytically using the mean field approximation developed for the previous models. The rate of change of the connectivity of vertex $`i`$ has two contributions: the first describes the probability that the vertex is chosen randomly as the origin of the link, $`\mathrm{\Pi }_{random}(k_i)=1/N`$ and the second is proportional to $`\mathrm{\Pi }(k_i)=k_i/_jk_j`$, describing the probability that an edge originating from a randomly selected vertex is linked to vertex $`i`$: $$\frac{k_i}{t}=A\frac{k_i}{_{j=1}^Nk_j}+\frac{1}{N}.$$ (20) Taking into account that $`_jk_j=2t`$ and that the change in connectivities during one timestep is $`\mathrm{\Delta }(k)=2`$, and excluding from the summation edges originating and terminating in the same vertex, we obtain $`A=N/(N1)`$, leading to $$\frac{k_i}{t}=\frac{N}{N1}\frac{k_i}{2t}+\frac{1}{N}.$$ (21) The solution of this equation has the form $$k_i(t)=\frac{2(N1)}{N(N2)}t+Ct^{\frac{N}{2(N1)}}.$$ (22) Since $`N>>1`$, we can approximate $`k_i`$ with $$k_i(t)=\frac{2}{N}t+Ct^{1/2}$$ (23) Since the number of vertices is constant, we do not have “introduction times” $`t_i`$ for the vertices. There exists, however, a time time analogous to $`t_i`$: the time when vertex $`i`$ was selected for the first time as the origin of an edge, and consequently its connectivity changed from $`0`$ to $`1`$. Equation (22) is valid only for $`t>t_i`$, and all vertices will follow this dynamics only after $`tN`$. The constant $`C`$ can be determined from the condition that $`_jk_j=2t`$, and has the value $$C=0,$$ (24) thus $$k_i(t)\frac{2}{N}t.$$ (25) The numerical results shown in Fig. 6b agree well with this prediction, indicating that after a transient time of duration $`tN`$ the connectivity increases linearly with time. Since the mean-field approximation used above predicts that after a transient period the connectivities of all vertices should have the same value given by Eq. (25), we expect that the connectivity distribution becomes a Gaussian around its mean value. Indeed, Fig. 6a illustrates that as time increases, the shape of $`P(k)`$ changes from the initial power-law to a Gaussian. The failure of models A and B in leading to a scale-free distribution indicates that both ingredients, namely growth and preferential attachment, are needed to reproduce the stationary power-law distribution observed in real networks. ## 5 Discussion and conclusions In the following we discuss some of the immediate extensions of the present work. (i) A major assumption in the model was the use of a linear relationship between $`\mathrm{\Pi }(k_i)`$ and $`k_i`$, given by (3). However, at this point there is nothing to guarantee us that $`\mathrm{\Pi }(k)`$ is linear, i.e. in general we could assume that $`\mathrm{\Pi }(k)k^\alpha `$, where $`\alpha 1`$. The precise form of $`\mathrm{\Pi }(k)`$ could be determined numerically by comparing the topology of real networks at not too distant times. In the absence of such data, the linear relationship seems to be the most efficient way to go. In principle, if nonlinearities are present (i.e. $`\alpha 1`$), that could affect the nature of the power-law scaling. This problem will be addressed in future work . (ii) An another quantity that could be tested explicitelly is the time evolution of connectivities in real networks. For the scale-free model we obtained that the connectivity increases as a power of time (See Eq. (6)). For model A we found logarithmic time dependence (Eq. (14)), while for model B linear (Eq. (25)). Furthermore, if we introduce $`p_{\mathrm{ER}}=at`$ in the ER model, one can easily show that $`k_{\mathrm{ER}}(t)t`$. If time resolved data on network connectivity becomes available, these predictions could be explicitelly tested for real networks, allowing us to distinguish between the different growth mechanisms. (iii) In the model we assumed that new links appear only when new vertices are added to the system. In many real systems, including the movie actor networks or the www, links are added continuously. Our model can be easily extended to incorporate the addition of new edges. Naturally, if we add too many edges, the system becomes fully connected. However, in most systems the addition of new vertices (and the growth of the system) competes with the addition of new internal links. As long as the growth rate is large enough, we believe that the system will remain in the universality class of the BA model, and will continue to display scale-free features. (iv) Naturally, in some systems we might witness the reconnection or rewiring of the existing links. Thus some links, that were added when a new vertex was added to the system, will break and reconnect with other vertices, probably still obeying preferential attachment. If reattachment dominates over growth (i.e. addition of new links by new vertices), the system will undergo a process similar to ripening: the very connected sites will acquire all links. This will destroy the power-law scaling in the system. However, similarly to case (iii) above, as long as the growth process dominates the dynamics of the system, we expect that the scale-free state will prevail. (v) The above discussion indicates that there are a number of “end-states” or absorbing states for random networks, that include the scale-free state, when power-law scaling prevails at all times, the fully connected state, which will be the absorbing state of the ER model for large $`p`$, and the ripened state, which will characterize the system described in point (iv). Note that the end state of the WS model, obtained for $`p_{\mathrm{WS}}=1`$, is the ER model for $`p_{\mathrm{ER}}=z/N`$. The precise nature of the transition between these states is still an open question, and will be the subject of future studies . (vi) Finally, the concept of universality classes has not been properly explored yet in the context of random network models. For this we have to define scaling exponents that can be measured for all random networks, whether they are generated by a model or a natural process. The clustering of these exponents for different systems might indicate that there are a few generic universality classes characterizing complex networks. Such studies have the potential to lead to a better understanding of the nature and growth of random networks in general. Growth and preferential attachment are mechanisms common to a number of complex systems, including business networks , social networks (describing individuals or organizations), transportation networks , etc. Consequently, we expect that the scale-invariant state, observed in all systems for which detailed data has been available to us, is a generic property of many complex networks, its applicability reaching far beyond the quoted examples. A better description of these systems would help in understanding other complex systems as well, for which so far less topological information is available, including such important examples as genetic or signaling networks in biological systems. Similar mechanisms could explain the origin of the social and economic disparities governing competitive systems, since the scale-free inhomogeneities are the inevitable consequence of self-organization due to the local decisions made by the individual vertices, based on information that is biased towards the more visible (richer) vertices, irrespective of the nature and the origin of this visibility. This work was supported by the NSF Career Award DMR-9710998.
no-problem/9907/nucl-th9907118.html	ar5iv	text	# Transport calculation of dilepton production at ultrarelativistic energies ## I Introduction Recently, electromagnetic radiation in form of lepton pairs has been observed at CERN in the CERES experiment. A strong enhancement above the cocktail of hadronic decays has been reported. This is interesting, because dileptons may escape nearly undisturbed from the hot and dense zone which is formed in the heavy ion collision. Dileptons could probe the intermediate stage of these interactions. On the other hand, hadrons are thought to probe the later freeze-out stages. This folklore may not be entirely correct: A number of investigations indicate that hadronic signals such as strangeness and entropy, as well as flow are developped in the early phase, while a large fraction of the observed dileptons is simply due to post-freeze-out decays (”feeding”) of isolated hadrons. The fact that also dileptons are emitted from various sources and during the entire reaction makes the interpretation of this signal not as easy as hoped for. Electromagnetic probes are definitely not a pure signal from the most interesting high density stage only! Rather, their spectrum is composed of different contributions integrated over the full interaction dynamics. Transport models need to be applied to decompose the dilepton cocktail into its various individual contributions. The most discussed feature of the data is the enhancement of intermediate mass lepton pairs $`(300\text{MeV}<M_{ee}<700\text{MeV})`$ as compared to the hadronic cocktails based on primary collisions. While the enhancement at BEVALAC energies is still not understood , first predictions for Pb+Au collisions from transport models including secondary production of dileptons (e.g. $`\pi \pi e^+e^{}`$) indicated only a small (if any) deviation from the CERES data . A large variety of attempts has been made to explain this deviation. The two most popular of them are the hypothesis of lowering vector meson masses (the Brown-Rho scaling) and the collisional broadening of the spectral functions in the hadronic environment . Here we analyse the recent CERES data on the multiplicity and $`p_t`$ of the lepton pairs. First, we give a survey of the dilepton production in the framework of the ultrarelativistic quantum molecular dynamics model, UrQMD . The present paper demonstrates that a state-of-the-art transport model based on hadronic and string degrees of freedom can predict the present data without assuming phase transitions or unconventional in-medium effects. This compares well to recent results from a simple transport model . ## II Description of the model Dilepton production in the framework of the UrQMD model has been considered in some detail in ref. . In this model dileptons are produced perturbatively, mainly in hadronic decays. Dalitz decays of neutral mesons and direct decays of vector mesons are taken into account. The incoherently summed $`pn`$ bremsstrahlung as well as the $`\mathrm{\Delta }`$(1232) Dalitz decay are of minor importance at CERES energies because the system is dominated by mesonic degrees of freedom. All dilepton decay widths have been specified in . For the vector meson decay width $`\mathrm{\Gamma }_{\rho e^+e^{}}(M)`$ we use the typical $`M^3`$ dependence which results from vector meson dominance (VMD model). There seems to be some inconsistency between the original formulation of VMD (see e.g. Sakurai ), which results in a $`M^3`$ dependence for $`\mathrm{\Gamma }_{\rho e^+e^{}}`$ and the extended VMD from Kroll, Lee and Zumino implying a dilepton decay width proportional to $`M`$ (see e.g. ref. ). The discrepancy between these two approaches seems to be as $`M^4`$! However, if one considers the channel $`\pi \pi e^+e^{}`$, the extended VMD requires also a direct coupling between the electromagnetic and the hadronic channel. If this term is included (as well as the interference), one ends up with the overall $`M^3`$ dependence again. In the extended VMD the direct term can be motivated only by a $`\pi \pi \gamma `$ coupling. Therefore it is not clear how to treat the dilepton decay of $`\rho `$ mesons which themselves come from heavy resonance decays or even string fragmentation. Similar problems arise for the $`\omega `$ and $`\varphi `$ decays. However, due to their small total widths, this effect does not manifest itself in the dilepton mass spectra and is thus not discussed separately here. In the UrQMD model the following procedure is applied: For $`\rho `$ mesons which are produced by $`\pi \pi `$-annihilations or in decays of baryonic or mesonic resonances the dilepton decay width scales like $`M^3`$. The dilepton width for $`\rho `$ mesons from string fragmentation is chosen according to the direct term of the extended VMD model which scales like $`M`$. Other approaches are possible and yield different results: see discussion below. ## III Results The undisturbed dilepton mass spectrum for proton on beryllium collisions at 450 GeV incident energy is shown in fig. Transport calculation of dilepton production at ultrarelativistic energies. This ’naked’ spectrum could be verified only with an ideal detector covering the $`4\pi `$ phase space angle and having a perfect momentum resolution (the bin width is 8 MeV). It shows all details of the model calculation. The CERES data are drawn for orientation only and should not be directly compared to the actual calculation. One sees, however, that roughly two orders of magnitude are lost due to the acceptance. At low invariant masses the overwhelming background is due to the Dalitz decays of the long-living mesons $`\pi ^0`$ and $`\eta `$. Between 500 and 750 MeV the $`\omega `$ Dalitz and the direct $`\rho `$ meson decays dominate the spectrum. At higher energies the characteristic peaks of the direct $`\omega `$ and $`\varphi `$ decays overlap the $`\rho `$ decay. The $`\eta ^{}`$ Dalitz decay or baryonic channels like $`pn`$ bremsstrahlung play only a subordinate role. The asymmetrical shape of the $`\rho `$ mass distribution is due to phase space limitations and especially due to the $`M^3`$ scaling of the decay width. At invariant masses around 1.2 GeV there is a drop in the $`\rho `$ distribution. This is due to a limitation of the model not to create resonances from string fragmentation with masses exceeding the pole mass more than three widths ($`M<M_R+3\mathrm{\Gamma }_R`$). For comparison with the experimental data from the CERES collaboration, one needs to correct the model calculations for the limited detector acceptance and momentum resolution. Because the detector covers only a small geometric region, every detected particle must have a certain angle to the beam axis. This requires the pseudo rapidity of the lepton $`\eta _{e^\pm }`$ to be $`2.1\eta _{e^\pm }2.65`$. A cut on small opening angles $`\mathrm{\Theta }_{e^+e^{}}>0.035\text{mrad}`$ of the electron-positron pair has to be matched because close pairs cannot be identified by the RICH detectors. The individual leptons which have passed the initial RICH detectors are then deflected in a magnetic field with a momentum resolution parameterised as $`{\displaystyle \frac{\mathrm{\Delta }p}{p}}=\sqrt{\alpha ^2+(\beta p)^2},`$ (1) where $`\alpha `$ and $`\beta `$ depend on the detector setup and are listed in table 1 . The momenta of the electron and positron are washed out accordingly with a Gaussian distribution function of width $`\frac{\mathrm{\Delta }p}{p}`$. Finally, these tracks have to survive a transverse momentum cut, which is typically $`p_t>50\text{MeV}`$ for the proton induced and $`p_t>200\text{MeV}`$ for the ion induced reactions. In fig. Transport calculation of dilepton production at ultrarelativistic energies we present the dilepton mass spectra for proton on beryllium collisions including the CERES acceptance and resolution. As compared to fig. Transport calculation of dilepton production at ultrarelativistic energies the momentum resolution has washed out the peak structures and the low mass pairs have been suppressed due to the acceptance. As an effect of the finite resolution one can see also contributions from the direct $`\rho ^0`$ decay below the two pion threshold. The transport calculation slightly overestimates the data around the vector meson poles but seems to be consistent with the overall data set. At invariant masses higher than 1.3 GeV the data possibly indicate a need for additional sources like direct production in meson+meson collisions (see e.g. ) which were not considered in the present calculation. The UrQMD result for the heaviest system measured by CERES can be found in fig. Transport calculation of dilepton production at ultrarelativistic energies together with the data from the ’96 run . These default calculations do not show the typical trend to underestimate the data in the region around 400 MeV. Like in a number of other free or conventional transport calculations the data point at 780 MeV, just at the $`\rho /\omega `$ mass, is slightly overestimated . This has been interpreted that ’some of the signal should be distributed away’ by medium effects. However, in view of the good agreement of the conventional calculation at lower masses one must ask, where this strength should go. Furthermore, we have observed that introducing collision widths results in a significant enhancement of vector meson production, which is not compatible to the data. Two data points are missed at 140 MeV and 180 MeV. By lowering the $`p_t`$ cut to less than 200 MeV this dip can be artificially removed on cost of overshooting the data points at 63 MeV and 88 MeV. Our interpretation of this observation is that the UrQMD $`p_t`$-distribution of pions and etas is too soft to be compatible to this data. Furthermore the spectral density of the $`\rho `$ meson in nuclear matter might extend to below $`2m_\pi `$ because the $`\rho `$ may decay into a nucleon nucleon-hole pair. This contribution is neglected in all on-shell transport models like UrQMD, but might fill in this gap. The results of the transport simulation depend on the details of the VMD model (fig. Transport calculation of dilepton production at ultrarelativistic energies). In the default calculation, the dilepton decay width of $`\rho `$ mesons scales like $`M^3`$ if they are from hadronic sources and like $`M`$ for $`\rho `$’s from strings (see the discussion above). If no VMD was considered at all, i.e. assuming constant dilepton decay widths for the $`\rho `$, a clear enhancement of the data over the model calculation can be observed (dashed line in fig. Transport calculation of dilepton production at ultrarelativistic energies). However, as pointed out above, neglecting the $`M^3`$-dependence of the $`\rho `$ meson dilepton decay width is not a valid assumption. In the hydrodynamical calculations of ref. the $`M^3`$ factor is missed, which might be the reason for their underestimation of the CERES data. On the other hand, one can assume ’total VDM’ by requiring that the vector mesons which were produced in string fragmentation scale like the ’hadronic’ $`\rho `$’s with $`M^3`$. This leads to an overestimation of the Pb+Au data (see dotted line in fig. Transport calculation of dilepton production at ultrarelativistic energies) and also the p+Be data are overestimated. From a theoretical point of view this approach has to be considered as an upper limit to the contribution of dilepton emission from $`\rho `$ mesons from string fragmentation. Contrary, the default calculation provides a lower limit for a realistic dilepton spectrum. The range between both approaches is the result of our uncertainty how to treat dilepton production from string fragmentation. The recent lead data have also been analysed for the transverse momenta of the lepton pairs (i.e. the virtual photons). The idea was to cut on pairs emitted from low-$`p_t`$ vector mesons for which the interaction with the surrounding medium should be strongest. Consequently, in-medium effects would result in a stronger enhancement of the low-$`p_t`$ data set over conventional models. The results of the corresponding UrQMD analysis are shown in fig. Transport calculation of dilepton production at ultrarelativistic energies in comparison to the CERES data . Note, first of all that the dip below $`M_{e^+e^{}}=200\text{MeV}`$ which was already observed in fig. Transport calculation of dilepton production at ultrarelativistic energies can be found only in the high $`p_t`$ region (lower frame). This indicates once more that UrQMD predicts less high-$`p_t`$ $`\pi ^0`$ and $`\eta `$ mesons. In the $`\rho `$ region a reasonable agreement between simulation and data can be found both for the low-$`p_t`$ (upper frame) and the high-$`p_t`$ sample. In particular the low-$`p_t`$ data are consistent with the present UrQMD calculation – without a need for additional medium effects. ## IV Conclusion In summary, the p+Be and the Pb+Au CERES data can be consistently explained by the UrQMD transport simulations. A reasonably good agreement was found especially in the region of $`0.3\text{GeV}<M<0.6\text{GeV}`$. It was discussed that the parameterisation of the vector meson decay width is not constrained from first principles. The variation between different approaches for $`\mathrm{\Gamma }_{\rho e^+e^{}}(M)`$ manifests itself in dilepton spectra clearly underestimating or slightly overshooting the CERES data. Further efforts in theory are required to provide a microscopic scheme for $`\mathrm{\Gamma }_{\rho e^+e^{}}(M)`$. Measurements with higher precision could help to identify the $`\omega `$ peak in the invariant mass spectra and thus pin down the shape of the $`\rho `$ meson. In addition more experimental information on the elementary channels p+p, $`\pi `$+p, especially a separate measurement of the Dalitz contribution (e. g. with TAPS), could be used to verify the different realisations of the VMD. ## Acknowledgments The authors want to thank W. Cassing, B. Friman and J. Knoll for helpful discussions. This work was supported in parts by Graduiertenkolleg Theoretische und Experimentelle Schwerionenphysik, GSI, BMBF, DFG, J. Buchmann Foundation, A. v. Humboldt Foundation and D. O. E. grant DE-FG02-96ER40945.
no-problem/9907/astro-ph9907235.html	ar5iv	text	# DISCOVERY OF THE OPTICAL TRANSIENT OF THE GAMMA RAY BURST 990308 ## 1. INTRODUCTION The old dilemma of whether Gamma Ray Bursts (GRBs) are galactic or cosmological has been solved recently with the discovery of X-ray (Costa et al. 1997), optical (van Paradijs et al. 1997), and radio transients (Frail et al. 1998). Four optical transients have measured red shifts of 0.835 (GRB970508), 0.966 (GRB980703), $`1.600`$ (GRB990123), and $`1.619`$ (GRB990510), thus proving that bursts are at cosmological distances. After the burst is over, the expanding shell emits an afterglow which eventually fades to invisibility. Early detections of the afterglow include GRB990510 at 3.52 hours (Harrison et al. 1999) and GRB990123 during the burst by ROTSE (Akerlof et al. 1999). Theoretical models of synchrotron emission from a decelerating relativistic shell shocked by collision with an external medium are in good agreement with afterglow observations from X-ray to radio and from the earliest to the latest times of available observations (Galama et al. 1998; 1999). The accurate optical/radio positions allow for very deep searches after the transient fades. For the 11 optical/radio transient positions searched to date, a very faint galaxy (typically with $`R25`$) appears within an arc-second or so for 9 of the positions. This result demonstrates that most GRBs reside inside distant host galaxies. Although the optical/radio transient positions are within the galaxies, they are significantly offset from the centers, demonstrating that the progenitors are neither associated with a giant central black hole nor are ejected from the galaxy. Bursts have been associated with active star formation (Totani 1997, Paczynski 1998), while the hosts have been shown to be significantly subluminous for normal galaxies in the majority of cases (Schaefer 1999). Three of the positionally coincident host galaxies have observed redshift values of 0.695 (GRB970228), 3.418 (GRB971214), and 1.097 (GRB980613), in reasonable agreement with the redshifts from the optical transients. The GRB explosion mechanism is still unknown, but a profitable line of study is to observe the burst afterglows over a wide range of time and frequency, with demographic arguments providing information on the progenitor and hence the explosion mechanism. Few afterglows have been observed to date, so further examples are important. This paper reports the discovery of the afterglow of GRB990308 with B, V, and R imagery starting 3.28 hours after the burst. ## 2. OBSERVATIONS GRB990308 triggered the Compton Gamma Ray Observatory Burst and Transient Source Experiment (BATSE) on 1999 March 8 05:15:07 UT. This burst has a peak flux (50-300 keV, 256 ms time bins) of $`1.6\pm 0.1phs^1cm^2`$, a fluence ($`>25`$ keV) of $`2.2\times 10^5ergcm^2`$, and a $`T_{90}`$ duration of $`106\pm 12`$ s. GRB990308 also triggered one camera of the Rossi X-Ray Timing Explorer (RXTE) All-Sky Monitor (Levine et al. 1996; Smith et al. 1999a). At 06:32 UT, a preliminary position had been determined and distributed as e-mail by the GRB Coordinate Network (GCN; Barthelmy et al. 1994). This position is a long thin error box which has since been updated (see below) to one with a $`3\sigma `$ total width of close to 6.1’. GRB990308 was also detected, very weakly, by the Ulysses GRB detector (Hurley et al. 1992). A cross correlation of the light curves between Ulysses and BATSE (separated by 2066.004 light seconds) as part of the Interplanetary Network (IPN; Hurley et al. 1999) yields an annulus with $`3\sigma `$ total-width of 35.9’. The IPN/RXTE error boxes overlap at an angle of $`80^{}`$ to form a $`3\sigma `$ error box defined by four points with J2000 coordinates of 12:22:12 +06:35:06, 12:22:23 +06:29:06, 12:24:05 +06:57:40, and 12:24:18 +06:52:05. The final error region was imaged with the LOTIS (Park et al. 1997) and Super- LOTIS (Park et al. 1998) wide-field fast response cameras located at the Lawrence Livermore National Laboratory. Both systems responded to a GCN notice of the preliminary BATSE coordinates, and the first images were taken 12.6 s after the trigger. An updated BATSE position from the GCN shifted the position, with the cameras responding quickly, such that the first LOTIS images that cover the error box are at 132 s after the BATSE trigger. This 10 s integration reached $`V=12.0`$ (S/N = 3, no filter) with no indication of an optical transient. A sum of roughly ten minutes of exposure from 132-1029 s after the BATSE trigger showed no optical transient to $`V=13.4`$. Unfortunately, the night was somewhat foggy, so the limiting magnitude was dominated by the unusually bright sky background. The Super-LOTIS camera has a $`0.8^{}`$ square field-of-view which took 30 s exposures in a spiral pattern around the center of the BATSE error region. The burst position was imaged starting 28.2 min after the burst to a depth of $`V=15.3`$ (S/N = 3, no filter) with no evidence for an optical transient. With the QUEST camera on the 1.0-m Schmidt telescope of the Observatorio Nacional de Llano del Hato near Mérida, Venezuela, we started scans of the preliminary error box approximately three hours after the burst. The QUEST camera consists of sixteen $`2048\times 2048`$ CCD detectors in a $`4\times 4`$ array operating in a drift scan mode (Snyder 1998, Sabbey, Coppi, & Oemler 1998). At any instant the QUEST camera is viewing 5.4 square degrees and in one hour will scan a $`2.3^{}\times 15^{}`$ region, which makes the camera ideal for rapid response to large preliminary GRB error boxes. Each row of detectors is covered with a different filter, so that the GRB990308 error box was imaged four independent times with standard (Bessel 1990) V, R, B, and R bands in time order. The integration time for each image was 142 s, with a total time of 11.5 min separating the centers of first and last. The limiting magnitudes were $`B=19.2,V=20.2`$, and $`R=21.6`$ to the $`S/N=3`$ level with the 69% illuminated Moon $`53^{}`$ away. The pixel size is 1.02”, the FWHM seeing was 2.6”, and the entire IPN/RXTE box was covered. The large size of the QUEST field-of-view is good for rapid imaging of large preliminary GRB positions, but it also means that rapid identification is difficult until a small error box is reported. For GRB990308, the early error box was $`8.6^{}`$ long until March 14 when the small IPN/RXTE region was produced (Smith et al. 1999b). The deep limiting magnitude of QUEST is good for catching faint optical transients, but it also means that the afterglow cannot be recognized if it is near or below the threshold of the Digital Sky Survey until deeper comparison images can be made. For GRB990308 where our combined images go to $`R=22.1`$ for $`S/N=3`$ with no obvious transient brighter than $`R=18`$, we had to await the acquisition of deep comparison images. These were made with the Yale 1-m telescope at Cerro Tololo (operated by the YALO consortium; Mendez, Depoy, and Bailyn 1998) in the R-band on 1999 May 29 as a 14 element mosaic reaching $`R=22.9`$ mag. A star-by-star comparison rapidly revealed an optical transient close to the center of the IPN/RXTE region (Fig. 1). The transient appears with a S/N of 15.0, 17.1, 4.7, and 20.8 on the V, R, B, and R images, respectively. All four images show the transient with the shape and FWHM of our point-spread-function. The astrometric position on all four images is identical to within an rms scatter of 0.4”, with the fitted motion during the 11.5 min of observation equal to $`0.03^{\prime \prime }\pm 0.28^{\prime \prime }`$. We take the appearance of four independent and significant images with good stellar shapes as proof that the images are not artifacts of any kind. For the position near opposition, the lack of motion rules out all Solar System objects, including Kuiper Belt Objects. The source must have an amplitude of $`>7.5`$ mag since the source position is empty to $`R>25.7`$ in June (see below). The only known astrophysical objects outside our Solar System with such amplitudes are dwarf novae, novae, supernovae, and GRBs. The transient cannot be a dwarf nova or nova since the $`R=18.14`$ observed magnitude (see below) would imply a distance far outside any galaxy. The transient cannot be a supernova (with $`M_R>19.5`$ for a distance modulus of $`<37.64`$) as our limit on any parent galaxy (which must have $`R>25.7`$, see later) would then be $`M_R>11.94`$. The transient is rather similar to GRBs seen previously (van Paradijs et al. 1997; Galama et al. 1998; 1999; Harrison et al. 1999). Thus we conclude that our optical transient is definitely the afterglow of GRB990308. We calibrated the B, V, and R magnitudes of nearby comparison stars with images taken on 1999 June 10 with the Yale 1-m telescope. This calibration was made by the observation of 22 standards stars (Landolt 1992) which was then applied to images of the GRB990308 field. We then performed differential photometry on the QUEST images with respect to 4 nearby comparison stars of known magnitude. We find $`V=18.32\pm 0.07`$ at 196.8 min, $`R=18.14\pm 0.06`$ at 200.6 min, $`B=18.65\pm 0.23`$ at 204.5 min, and $`R=18.22\pm 0.05`$ at 208.3 min after the burst. The times are for the middle of the integrations relative to the BATSE trigger. The position of the transient was measured for all four images with respect to stars in the USNO-A2.0 catalog (Monet et al. 1998). The combined position is J2000 $`12:23:11.44\pm 0.02`$ $`+06:44:05.10\pm 0.17`$. For comparison purposes, our star #1 ($`B=14.75\pm 0.01`$, $`V=13.94\pm 0.02`$, $`R=13.50\pm 0.03`$) is at $`12:23:11.272+06:45:38.19`$ while star #6 ($`B=18.40\pm 0.07`$, $`V=17.63\pm 0.03`$, $`R=17.16\pm 0.03`$) is at 12:23:10.824 +06:43:20.67. We observed this position with the Very Large Array telescope and found no significant radio sources to the $`3\sigma `$ level of $`258\mu Jy`$ at 8.5 GHz on 1999 June 18, to $`114\mu Jy`$ at 8.5 GHz on 1999 June 26, and to $`165\mu Jy`$ at 1.4 GHz on 1999 July 4. The transient had probably faded to invisibility by the mid-March, but late time imaging was carried out to find the GRB host galaxy. The transient position is empty on the Palomar Sky Survey and our Yale 1-m images. We took deep R-band images with the WIYN 3.5-m on Kitt Peak on 1999 June 12. Our six 15-min exposures were processed and co-added by the normal procedures. The resulting picture shows no significant source within 6” to a $`S/N=3`$ limiting magnitude of $`R=25.4`$. We also obtained deep K-band and R-band images of the optical transient position with the 10-m Keck I and II telescopes, respectively. The R-band image was obtained on 1999 June 19 as a combination of five 200 s exposures with the Low-Resolution Imaging Spectrometer (Oke et al. 1995). This reached a $`S/N=3`$ limiting magnitude of $`R=25.7`$. No star or galaxy appears to this limit within $`6.3^{\prime \prime }`$ of the optical transient position. The K-band images were obtained on 1999 June 23 and 24 as a combination of 12 second exposures totaling 71 and 72 min with the Near Infrared Camera (Matthews & Soifer 1994). The nights were photometric and we obtained standard star observations to calibrate the photometry. The fields are roughly 38.4” on a side with 0.15 pixels, reaching a $`S/N=3`$ limiting magnitude of $`K=23.3`$. The position of the optical transient was empty to this threshold. ## 3. ANALYSIS Our observations can be used to constrain the decline rate of the afterglow. We will parameterize the decline with a power law index $`\delta `$, such that the flux varies as $`t^\delta `$ with $`t`$ the time since the burst. The two QUEST R-band observations 7.7 min apart show a fading by $`0.08\pm 0.08`$ mag which suggests that $`\delta =2\pm 2`$. The QUEST V-band measurements can be combined with each of the three limits from LOTIS and Super-LOTIS to give similar constraints that $`\delta >1.3`$ for early times. The WIYN and Keck limits imply that $`\delta <1.1`$ for late times. Taken together, all of our detections and limits are consistent with $`\delta =1.2\pm 0.1`$ for the assumption of a single power law over all time. As discussed below, the spectral slope gives reason to expect a more complicated time dependance with $`\delta `$ close to zero for early times. The QUEST B, V, and R magnitudes can yield a spectral slope. The first step is to correct for the known galactic extinction of E(B-V) = 0.023 (Schlegel, Finkbeiner, & Davis 1998). The second step is to deduce the magnitudes and colors at one instant of time, which we take as the time of our first image. If $`\delta =1`$ then the colors are $`BV=0.26\pm 0.24`$ and $`VR=0.17\pm 0.08`$, while if $`\delta =0`$ then the colors are $`BV=0.30\pm 0.24`$ and $`VR=0.13\pm 0.08`$. The third step is to convert these magnitudes into flux units in Janskys, $`f_\nu `$. The fourth step is to fit the three fluxes to a presumed power law with $`f_\nu \nu ^\alpha `$. We find $`\alpha =0.38\pm 0.25`$ for $`\delta =1`$ and $`\alpha =0.49\pm 0.26`$ for $`\delta =0`$. Within the framework of afterglow models for the synchrotron emission from external shocks (Sari, Piran, and Narayan 1998), the spectral slope is either in an $`\alpha =1/3`$ regime or in regimes with $`\alpha 0.5`$. Our measured spectral slope is easily consistent with the $`\alpha =1/3`$ regime, yet is inconsistent with the $`\alpha 0.5`$ regime at the $`5\times 10^4`$ probability level. Thus we conclude that the afterglow was in the $`\nu ^{1/3}`$ regime at a time 3.28 hours after the burst. Synchrotron models require that any afterglow in the $`\nu ^{1/3}`$ regime have $`1/3\delta 1/2`$ (Sari, Piran, and Narayan 1998), which is to say that the afterglow is not fast-fading and may even be brightening. GRB970508 provides a precedent for an afterglow brightening for the first two days, during which time the spectral slope is positive and consistent with $`\nu ^{1/3}`$ (Castro-Tirado et al. 1998). A $`\delta 0`$ light curve for GRB990308 would explain the absence of a detection by LOTIS and Super-LOTIS, yet is not significantly inconsistent with our observed decline by $`0.08\pm 0.08`$ mag in 7.7 minutes. Within the standard synchrotron fireball model, we can place interesting constraints on the burst energetics. For the GRB 990308 optical light to be in the $`\nu ^{1/3}`$ regime, the characteristic times $`t_c`$ and $`t_m`$ (Sari, Piran, & Narayan 1998, eqns 15 and 16) must both be greater than 3.28 hours. For the case of adiabatic evolution, this requires $`E_{52}>0.0023ϵ_B^1ϵ_e^4`$ and $`E_{52}<0.0002ϵ_B^3n^2`$, where $`E_{52}`$ is the energy in the spherical shock in units of $`10^{52}`$ ergs, n is the number density in the external medium in units of $`cm^3`$, while $`ϵ_e`$ and $`ϵ_B`$ are the fractions of the shock energy in the electrons and magnetic field. These two constraints limit the allowed values to a small sliver of $`E_{52}ϵ_B`$ parameter space, centered roughly on the $`40ϵ_B=E_{52}^{0.5}`$ line. With $`n1`$ and the required $`ϵ_e1`$, then $`ϵ_B0.25`$ and $`E_{52}>0.01`$. Since for reasonable models and beaming factors we have $`E_{52}<100`$ and $`n>1`$, we derive $`ϵ_e>0.2`$. Similarly, for $`E_{52}<100`$ and $`ϵ_e1`$, we find that $`n<10^4`$, which might be a problem for models where the burster sits in a massive wind from a massive supernova precursor or in dense star forming regions. For the case of equipartition of energy, there is no acceptable solution other than in the case that $`n<0.1`$. For the cases of $`n=1`$ and $`n=1000`$, $`ϵ_B`$ must be less than $`24\%`$ and $`0.03\%`$ of $`ϵ_e`$ respectively with both $`t_c`$ and $`t_m`$ pushed to 3.28 hours. Similar conclusions are reached in the radiative case with the Lorentz factor of the shocked material is $`>10`$. So, with GRB990308 being in the $`\nu ^{1/3}`$ regime at 3.28 hours after the burst for standard fireball models with $`n>0.1`$, we argue that the external medium has $`n<10^4`$, that the fraction of the shock energy in electrons must be greater than 20%, that the fraction of the shock energy in the magnetic field must be less than 25%, and that the electrons not be in equipartition with the magnetic field. It is generally thought that GRBs were born in distant host galaxies, so the absence of any possible host galaxy to $`R=25.7`$ is problematic. A possible solution to the lack of a visible host galaxy is that the burster was ejected from its home galaxy long ago. The possibility of ejection is natural in some models, such as neutron star collision scenarios, where the progenitor binary system may receive large velocity kicks during neutron star formation. This solution might be implausible since all other optical/radio transients with exact positions and associated hosts are within the visible galaxies, so at least these bursters have not been ejected. For the specific case of GRB990308, the nearest galaxy is one with $`R=24.7`$ at 6.3” angular distance, which implies a transverse separation of $`50`$ kpc or more for a plausible host and luminosity. Such a transverse separation is unlikely within current models (Bloom, Sigurdsson, & Pols 1999), so we consider ejection improbable. The second possible solution to the lack of a visible host is that the galaxy is substantially subluminous. This case has already been demonstrated for the majority of the very brightest bursts irrespective of the breadth of the GRB luminosity function (Schaefer 1999), while several of the hosts associated with optical transients with redshifts are in the lower few per cent of the luminosity-weighted Schechter luminosity function (GRB970228, GRB970508, GRB990510). Indeed, if GRB990308 has the average luminosity associated with the no-evolution fits to the LogN-LogP curve ($`10^{57}phs^1`$ or $`6\times 10^{50}ergs^1`$; Fenimore et al. 1993), then the host galaxy would be at $`z=0.50`$ with an absolute magnitude fainter than $`17.2`$ which is in the faintest $`2\%`$ of galaxies. The third possible solution to the lack of a visible host is that the burst can be at a high redshift. For an average galaxy, say one in the middle of the luminosity-weighted luminosity function with an absolute R magnitude of $`20.8`$, the burst must have $`z>1.2`$ for us to not detect the host to $`R=25.7`$. Here and throughout this paper, we have adopted a Hubble Constant of $`65kms^1Mpc^1`$, $`\mathrm{\Omega }_0=0.3`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0.7`$, the galaxy K-corrections for Sb galaxies of Rocca-Volmerange & Guideroni (1988), and the GRB K-corrections of Fenimore et al. (1992) with spectral slope -1.5. With this approximate limit based on an assumption that the host is average in brightness , the deduced burst luminosity must be $`>6\times 10^{57}phs^1`$. Pushing GRB990308 to $`z>1.2`$ is plausible. This work was supported by the National Science Foundation, the Department of Energy, and the National Aeronautics and Space Administration, while the Observatorio Astronómico Nacional is operated by CIDA for the Consejo Nacional de Investigaciones Científicas y Tecnológicas.
no-problem/9907/cond-mat9907400.html	ar5iv	text	# Spontaneous spin stripe dimerization in the doped t-J model. \[ ## Abstract To investigate spin dimerization in the $`tJ`$ model we consider the extended $`tJ\delta `$ model with explicit spin dimerization introduced via parameter $`\delta `$. At zero doping the dimerized spin liquid is unstable at $`\delta \delta _c0.3`$. We demonstrate that the doping stabilizes the dimerized stripe phase: the $`\delta _c(x)`$ decreases when doping $`x`$ increases. At doping larger than critical, $`x>x_c`$, the dimerized phase is stable even without explicit dimerization, i.e. at $`\delta =0`$. \] It is widely believed that the 2D $`tJ`$ model is relevant to the low energy physics of high-temperature superconductors. This is why investigation of this model is of great interest both for theory and experiment. In spite of great efforts during more than a decade there is no full understanding of the phase diagram of the $`tJ`$ model, however some facts are well established. At zero doping the model is equivalent to the Heisenberg model on a square lattice which has long range Neel order . Doping by holes destroys the order. A simplified picture of noninteracting holes leads to the Neel state instability with respect to spirals at arbitrary small but finite doping . However more sophisticated numerical calculations which take into account renormalization of the hole Green’s function under the doping indicate that the Neel order is stable below some critical hole concentration $`x_c`$ . In the Neel phase ($`x<x_c`$), in all waves except s-wave, there is magnon mediated superconducting pairing interaction between holes. . It is also clear that at very small hopping there is phase separation in the model because separation leads to reduction of the number of destroyed antiferromagnetic links The purpose of the present work is to elucidate spin structure of the ground state in quantum disordered regime, i.e at $`x>x_c`$. The most important hint comes from experiment: indications of stripes in the high-$`T_c`$ materials . Another important hint is a remarkable stability of the spin dimerized phase in the frustrated $`J_1J_2`$ model. The idea of such state for this model was first formulated by Read and Sachdev , and was then confirmed by further work . The stability of such a configuration implies that the lattice symmetry is spontaneously broken and the ground state is four-fold degenerate. Such a route towards quantum disorder is known rigorously to take place in one dimension, where the Lieb-Schultz-Mattis (LSM) theorem guarantees that a gapped phase always breaks the translational symmetry. Some time ago Affleck suggested that the LSM theorem can be extended to higher dimensions, and the gapped states of quantum systems necessarily break the discrete symmetries of the lattice . The example of the $`J_1J_2`$ model provides further support for this idea. There have been several attempts to consider the spin-dimerized phase in a doped Heisenberg antiferromagnet. For this purpose Affleck and Marston analyzed Hubbard-Heisenberg model in the weak-coupling regime, Grilli, Castellani and G. Kotliar considered $`SU(N)`$, $`N\mathrm{}`$, $`tJ`$ model, and very recently Vojta and Sachdev considered $`Sp(2N)`$, $`N\mathrm{}`$, $`tJ`$ model with long range Coulomb interaction. These works indicated a stability of the spin-dimerized phase in some region of parameters, providing a very important guiding line. However relevance of these results to ”physical regime” of the $`tJ`$ model remained unclear. In the present work we demonstrate that the spin-dimer order is stable, the only small parameter used in the analysis is hole concentration with respect to the half filling. The Hamiltonian under consideration is $$H=t\underset{ij\sigma }{}c_{i\sigma }^{}c_{j\sigma }+\underset{ij}{}J_{ij}\left(𝐒_i𝐒_j\frac{1}{4}n_in_j\right).$$ (1) $`c_{i\sigma }^{}`$ is the creation operator of an electron with spin $`\sigma `$ $`(\sigma =,)`$ at site $`i`$ of the two-dimensional square lattice. The $`c_{i\sigma }^{}`$ operators act in the Hilbert space with no double electron occupancy. The $`ij`$ represents nearest neighbour sites. The spin operator is $`𝐒_i=\frac{1}{2}_{\alpha ,\beta }c_{i\alpha }^{}\sigma _{\alpha \beta }c_{i\beta }`$, and the number density operator is $`n_i=_\sigma c_{i\sigma }^{}c_{i\sigma }`$. Antiferromagnetic interactions $`J_{ij}>0`$ are arranged in a stripe pattern shown in Fig. 1: solid links correspond to $`J_{ij}=J_{}=J(1+\delta )`$, and dashed links correspond to $`J_{ij}=j=J(1\delta )`$. At half filling ($`n_i=1`$) the Hamiltonian (1) has already been studied: for $`\delta >\delta _c0.303`$ the ground state is a quantum disordered state with gapped spectrum, and for $`\delta <\delta _c`$ there is spontaneous Neel ordering with gapless spin waves. In order to study the stability of the dimer phase we first derive an effective Hamiltonian in terms of bosonic operators creating spin-wave triplets (magnons) $`t_{i\alpha }^{}`$, $`\alpha =x,y,z`$ and fermionic operators creating holes $`a_\sigma ^{}`$, $`\sigma =,`$ from the spin singlets shown in Fig. 1. This Hamiltonian consists of four parts: the spin-wave part $`H_t`$, the hole part $`H_h`$, the spin-wave-hole interaction $`H_{th}`$, and the hole-hole interaction $`H_{hh}`$. Let us start from $`H_t`$. Similar effective theories have been derived in Refs. and we only present the result: $`H_t=H_2+H_3+H_4+H_U`$, where $`H_2=_{𝐤,\alpha }\left\{A_𝐤t_{𝐤\alpha }^{}t_{𝐤\alpha }+\frac{B_𝐤}{2}\left(t_{𝐤\alpha }^{}t_{𝐤\alpha }^{}+\text{h.c.}\right)\right\}`$, $`H_3=_{1+2=3}\text{R}(𝐤_\mathrm{𝟏},𝐤_\mathrm{𝟐})ϵ_{\alpha \beta \gamma }t_{𝐤_\mathrm{𝟏}\alpha }^{}t_{𝐤_\mathrm{𝟐}\beta }^{}t_{𝐤_\mathrm{𝟑}\gamma }+\text{h.c.}`$, and $`H_4=_{1+2=3+4}\text{T}(𝐤_\mathrm{𝟏}𝐤_\mathrm{𝟑})(\delta _{\alpha \delta }\delta _{\beta \gamma }\delta _{\alpha \beta }\delta _{\gamma \delta })t_{𝐤_\mathrm{𝟏}\alpha }^{}t_{𝐤_\mathrm{𝟐}\beta }^{}t_{𝐤_\mathrm{𝟑}\gamma }t_{𝐤_\mathrm{𝟒}\delta }`$. We also introduce an infinite repulsion on each site, in order to enforce the kinematic constraint $`t_{i\alpha }^{}t_{i\beta }^{}=0`$. $$H_U=U\underset{i,\alpha \beta }{}t_{i\alpha }^{}t_{i\beta }^{}t_{i\beta }t_{i\alpha },U\mathrm{}$$ (2) The following definitions are used in $`H_2`$: $`B_𝐤=j(\mathrm{cos}k_y0.5\mathrm{cos}k_x)`$ and $`A_𝐤=J_{}+B_𝐤`$. The matrix elements in the quartic and cubic interaction terms are: $`\text{T}(𝐤)=j(0.25\mathrm{cos}k_x+0.5\mathrm{cos}k_y)`$ and $`\text{R}(𝐩,𝐪)=0.25j(\mathrm{sin}q_x\mathrm{sin}p_x)`$. Throughout the paper we work in the Brillouin zone of the dimerized lattice. At zero doping ($`n_i=1`$) $`H_t`$ is an exact mapping of the original Hamiltonian (1). To analyse this case it is enough to apply the technique . The result for the normal spin-wave Green’s function reads: $$G_N(𝐤,\omega )=\frac{\omega +\stackrel{~}{A}_𝐤(\omega )}{\{\omega +\stackrel{~}{A}_𝐤(\omega )\}\{\omega \stackrel{~}{A}_𝐤(\omega )\}+\stackrel{~}{B}_𝐤^2}$$ (3) where $`\stackrel{~}{A}_𝐤(\omega )=A_𝐤+\mathrm{\Sigma }_4^N(𝐤)+\mathrm{\Sigma }_{Br}^{(1)}(𝐤,\omega )`$ and $`\stackrel{~}{B}_𝐤(\omega )=B_𝐤+\mathrm{\Sigma }_4^A(𝐤)`$. Normal $`\mathrm{\Sigma }_4^N`$ and anomalous $`\mathrm{\Sigma }_4^A`$ self-energies are caused by the quartic interaction $`H_4`$ and the most important contribution $`\mathrm{\Sigma }_{Br}^{(1)}`$ comes from the Brueckner diagrams as described in . Strictly speaking there is also some contribution to the self-energy caused by the ”triple” interaction $`H_3`$. However this contribution is very small (see, e.g. Ref.) and therefore we neglect it. Expansion of the self-energy in powers of $`\omega `$ near $`\omega =0`$ gives quasiparticle residue and spin-wave spectrum $`Z_𝐤^1=1\frac{\mathrm{\Sigma }_{Br}^{(1)}}{\omega }`$, $`\omega _𝐤=Z_𝐤\sqrt{[\stackrel{~}{A_𝐤}(0)]^2\stackrel{~}{B_𝐤}^2}`$. Expressions for effective Bogoliubov parameters $`u_𝐤`$ and $`v_𝐤`$ are given in . The spin-wave gap $`\mathrm{\Delta }=\omega _{𝐤_\mathrm{𝟎}}`$, $`𝐤_\mathrm{𝟎}=(0,\pi )`$, obtained as a result of a selfconsistent solution of Dyson’s equations is plotted in Fig.2 (line at $`x=0`$). The critical value of the explicit dimerization (point where the gap vanishes) $`\delta _c=0.298`$ is in agreement with results of series expansions and quantum Monte Carlo simulations . The validity of the Brueckner approximation is justified by the smallness of the gas parameter $`n_t=_\alpha t_{i\alpha }^{}t_{i\alpha }`$. At the critical point $`n_t=0.13`$. Consider now doping by holes. On the single dimer $`\|s`$ the hole can exist in symmetric and antisymmetric states. Because of hopping between the dimers there is mixing between these states, but the mixing is very small (few per cent) and can be neglected. The symmetric state has substantially lower energy and therefore only this state is populated at doping. The corresponding hole creation operator $`a^{}`$ is defined in the following way: $`\sqrt{2}a_\sigma ^{}\|s=(c_{2,\sigma }^{}+c_{1,\sigma }^{})\|0`$, where $`1`$ and $`2`$ numerate the dimer sites. Bare hole dispersion can be found by calculating hopping matrix elements. This gives $`H_h=(ϵ_𝐩+const)a_{𝐩\sigma }^{}a_{𝐩\sigma }`$, where the $`const`$ is chosen in such a way that $$ϵ_𝐩=t(\mathrm{cos}p_y+0.5\mathrm{cos}p_x+1.5)$$ (4) vanishes at the minimum: $`ϵ_{𝐩_\mathrm{𝟎}}=0`$, $`𝐩_\mathrm{𝟎}=(\pi ,\pi )`$. The spin-wave-hole interaction $`H_{th}`$ can be easily calculated in the way similar to that for doped spin-ladder . This interaction consists of two parts. The first one is interaction of a hole and a magnon positioned at different dimers. This is a relatively weak interaction which can be neglected . The second part, which gives the main effect, comes from the constraint that a hole and a magnon can not coexist at the same dimer: $`t_{i\alpha }^{}a_{i\sigma }^{}=0`$. To deal with this constraint we introduce, similarly to (2), an infinite repulsion $$H_{U1}=U\underset{i,\alpha \sigma }{}t_{i\alpha }^{}t_{i\alpha }a_{i\sigma }^{}a_{i\sigma },U\mathrm{}.$$ (5) The exact hole-magnon scattering amplitude caused by this interaction can be found via Bethe-Salpeter equation shown in Fig.3a. It is similar to that for magnon-magnon scattering . The result is $$\mathrm{\Gamma }(E,𝐤)=\left(\underset{𝐪}{}\frac{Z_𝐪u_𝐪^2}{E\omega _𝐪ϵ_{𝐤𝐪}}\right)^1,$$ (6) where $`E`$ and $`𝐤`$ is total energy and total momentum of the incoming particles. Let us denote the hole concentration by $`x=n/N`$, where $`n`$ is number of holes, and $`N`$ is number of sites. Hence on-site electron occupation number is $`n_i=1x`$. Concentration of holes in terms of the dimerized lattice is two times larger $`n/(0.5N)=2x`$, and this is the gas parameter of the magnon-hole Brueckner approximation. According to (4) the holes are concentrated in the pocket in the vicinity of $`𝐩_\mathrm{𝟎}=(\pi ,\pi )`$. Therefore the magnon normal self-energy described by the diagram Fig. 3b is $$\mathrm{\Sigma }_{Br}^{(2)}(𝐤,\omega )=2x\mathrm{\Gamma }(\omega ,𝐤+𝐩_\mathrm{𝟎})$$ (7) It is instructive to consider first the case which allows an analytical solution: $`J_{}j`$, $`\sqrt{2}\pi x1`$. Bare magnon dispersion in this case is $`\omega _𝐤J_{}+j(\mathrm{cos}k_y0.5\mathrm{cos}k_x)`$ and hence the integrals in (6,7) can be calculated analytically with logarithmic accuracy. This gives $$\mathrm{\Sigma }_{Br}^{(2)}(𝐤,\omega )\frac{2\sqrt{2}\pi x(t+j)}{\mathrm{ln}(12.5/\mu )+i\pi \theta (\delta \omega )},$$ (8) where $`\delta \omega =\left[\omega \omega _𝐤+j(\omega _𝐤\omega _{𝐤_\mathrm{𝟎}})/(t+j)\right]/(t+j)`$, $`\mu =\mathrm{max}(\|\delta \omega \|,\sqrt{2}\pi x)`$, and $`\theta (\delta \omega )`$ is a step function. The magnon Green’s function is $`G(𝐤,\omega )=\left(\omega \omega _𝐤\mathrm{\Sigma }_{Br}^{(2)}(𝐤,\omega )\right)^1`$. For illustration the spectral function $`ImG(\omega )`$ at $`𝐤=𝐤_\mathrm{𝟎}`$, $`t/j=3`$ and different $`x`$ is plotted in Fig.4. There are several conclusions from formula (8) and Fig. 4: 1) doping pushes the spin-wave spectrum up, 2) the effect is increasing with hopping $`t`$, 3) finite width appears, 4) there is only a logarithmic dependence on the infrared cutoff. Let us stress the importance of the point (4). It means that the effect is practically independent of the long-range dynamics. Moreover, near the critical point ($`\mathrm{\Delta }=0`$) the situation is even better: the spin-wave spectrum is linear and even the logarithmic divergence disappears. Thus in the 2D case there is separation of scales which justifies Brueckner approximation. If we tried to apply the described approach to the 1D case (say a doped spin ladder) we would get into trouble: power infrared divergence appears in Brueckner diagram and hence there is no justification for gas approximation. Let us also comment on the point (3) (width). There is a ”triple” contribution to the magnon self-energy, Fig. 3c. This is a long-range contribution which is much less important than the Brueckner one, and this is why we neglect it (cf. with Ref.). However this diagram influences the width of the magnon spectral function. In the general case there are two contributions to the Brueckner self-energy: $`\mathrm{\Sigma }_{Br}^{(1)}`$, which is due to the magnon-magnon constraint, and $`\mathrm{\Sigma }_{Br}^{(2)}`$ which is due to the magnon-hole constraint. To find the spin wave spectrum one has to solve selfconsistently Dyson’s equation for Green’s function (3), as it is described in Ref. . Results for the spin-wave ”gap” $`\mathrm{\Delta }`$ as a function of explicit dimerization $`\delta `$ for different hole concentrations $`x`$ and $`t/J=3`$ are plotted in Fig. 2. Strictly speaking at $`x0`$ the $`\mathrm{\Delta }`$ is not a gap because of the large decay width. What we plot is the position of the centre of gravity of the magnon spectral function. However at $`\mathrm{\Delta }0`$ the width vanishes, and therefore the critical regime is uniquely defined. When $`\mathrm{\Delta }`$ is not small there is an interesting question of calculation of the exact shape of magnon spectral function. However to resolve this problem one needs to include long-range dynamics (diagram Fig.3c). It is clear from Fig.2 that at $`t/J=3`$ and $`x>x_c0.090`$ the ”gap” remains finite even at $`\delta =0`$. This is regime of spontaneous dimerization. For $`t/J=2`$ the critical concentration is $`x_c=0.106`$, and for $`t/J=1`$, $`x_c=0.132`$. Thus the doping stabilizes the dimerized phase. The larger the hopping $`t`$, the stronger the effect of stabilization. (The same follows from eq. (8).) This statement is true only if $`t/J10`$. At $`t/J10`$ there is a crossover to quasiparticles with higher spin (hole-magnon bound states) which indicate transition to the Nagaoka regime. The small parameter of the Brueckner approximation is concentration of holes in the dimerized lattice: $`2x`$. Therefore at $`t/J=3`$ one should expect $`20\%`$ accuracy in calculation of $`x_c`$. Note that the value of $`x_c`$ is close to that found in from the Neel state. An important effect related to the stabilisation is suppression of spin wave quantum fluctuations by doping. At $`x=0`$ and $`\delta =\delta _c=0.298`$ the density of spin fluctuations is $`n_t0.13`$. Increasing of the hole concentration to $`x=0.1`$ (at $`t/J=3`$ and $`\delta =\delta _c`$) gives $`n_t0.02`$. At the critical point $`t/J=3`$, $`\delta =0`$, $`x=x_c=0.09`$ the density is just $`n_t0.07`$ (see also ). The phase diagram of the $`tJ\delta `$ model at zero temperature is presented in Fig.5 Because of the mobile holes the dimerized spin liquid is a conducting state. Stability of this state is a very robust effect because it is due to the high energy correlations (typical energy scale $`2t`$). There are also low energy (long range) effects with typical energy scale $`2tx`$ which can lead to hole-hole pairing, small amplitude density waves etc. We do not consider these effects in the present work because they are secondary with respect to the main one: spin dimerization. It is also worth noting that one can introduce additional parameters $`t^{}`$ and $`t^{\prime \prime }`$ (next and next-next neighbour hopping) which influence single hole dispersion (4), shift position of the minimum and change shape of the ”Fermi surface”: stabilization is not sensitive to these details. Strictly speaking one can not exclude a possibility that at decreasing $`\delta `$, at some point there is a first order phase transition to an absolutely different phase. However such a transition at zero temperature always has a precursor: some specific bound state. The examples are: hole-magnon bound state for transition to Nagaoka regime in the present model at very large $`t`$, multi-magnon bound states for a frustrated spin-ladder as a precursor of spinon deconfinement , and multi-magnon bound states in 2D $`J_1J_2`$-model at the 1st order transition point $`J_2/J_10.6`$ . In the present model we do not see any hint of such precursors and therefore we conclude that the first order phase transition is highly unlikely. We have discussed in detail the transition to the Neel state at small hole concentration $`x`$. It is also clear that at large $`x`$ there is a 2nd order transition to normal Fermi liquid shown schematically in Fig. 5 by the dashed line. One can say that it happens when hole concentration in the dimerized lattice is equal to unity, $`2x1`$. However this is only a crude estimate. Unfortunately we can not describe this part of the diagram more precisely because our approach assumes that $`2x1`$. It has been shown recently that Heisenberg antiferromagnet on triangular lattice is very close to the instability with respect to spontaneous dimerization . We would like to note that it is highly likely that in this model doping also stabilises the dimerized phase. In conclusion, using the dilute gas approximation we have analyzed the phase diagram of the $`tJ\delta `$ model and the stability of spin-dimerized phase. At doping larger than the critical one, $`x>x_c0.09`$, the dimerized phase is stable even without explicit dimerization. I thank V. N. Kotov and M. Yu. Kuchiev for stimulating discussions.
no-problem/9907/cond-mat9907501.html	ar5iv	text	# Thermal vortex dynamics in a two-dimensional condensate ## I Introduction Vortices and other topological field configurations play a fundamental role in determining the properties of many fascinating materials and control the physical mechanisms underlying several applications. Examples include superfluids , superconductors , periodic solids , liquid crystals , two-dimensional magnets , propagating coherent light beams , and perhaps even the hot plasma that filled the very early Universe . Common aspects of the phenomenology of these diverse applications stem from the mathematical similarity of the underlying field theories and their classical static solutions. Models of vortices as Brownian point particles, characterized by a mass, mutual interactions, and damping, are commonly used to study the behavior of superconductors and superfluids . In this picture the equation of motion for a single overdamped vortex, not subject to any external forces, is extremely simple: $`\gamma \dot{𝐫}(t)`$ $`=`$ $`𝐟(t);`$ (1) $`f_\mu (t)f_\nu (t^{})`$ $`=`$ $`{\displaystyle \frac{2\gamma k_BT}{M}}\delta _{\mu ,\nu }\delta (tt^{}),`$ (2) where $`\gamma `$ is the damping coefficient, $`M`$ is the vortex mass, $`𝐟(t)`$ is a Gaussian thermal noise, $`\mu ,\nu \{x,y\}`$, and $`\mathrm{}`$ denotes canonical ensemble averaging. It follows that the vortex velocity distribution is also Gaussian. This effective picture is extremely appealing—primarily because of the drastic reduction of the number of degrees of freedom. However, to our knowledge, it has not been shown to be derivable from the dynamics of an underlying field theory. In this paper we study this question by solving both analytically and numerically for the motion of the vortex as an effective degree of freedom arising in a stochastic nonlinear Schrödinger equation. It is well-known that the conventional description of Brownian motion in terms of equations such as (2), that of a heavy particle interacting with light scatterers, ignores the presence of hydrodynamic fluctuations. A more complete description would allow us not only to test the validity of the Brownian motion model for the vortex but also to compute corrections to it and, possibly, to find new physical effects. Our analytic approach rests on the use of a singular perturbation expansion around a rigid vortex utilizing a low-temperature or weak-noise expansion. We are able to derive a Fokker-Planck equation for the single vortex distribution function which corresponds to that of a passive scalar advected to a background flow \[not just a simple diffusion equation as would be predicted by Eqs. (2)\]. In our numerical work we are able to extract the diffusion constant for the vortex motion which turns out to be in good agreement with the theoretical prediction. Consistent with our theoretical analysis, the vortex effective mass diverges logarithmically with the system size. We also investigate the statistics of vortex flights and demonstrate that the probability distribution function (PDF) has an exponential tail implying nontrivial correlations in the background thermal flow field. In Section I we present the stochastic nonlinear Schrödinger equation and discuss its physical relevance. In Section II we use the low-temperature perturbative expansion about the vortex collective coordinates to derive its equation of motion. Section III describes our numerical techniques and the associated results. We end with Section IV with a further discussion of our conclusions. ## II The Stochastic Nonlinear Schrödinger Equation We study the motion of an isolated vortex by considering it as a singularity of a classical stochastic nonlinear Schrödinger field $`\psi (𝐫,t)`$ in two spatial dimensions. The equation of motion of this model theory is $`(a<0)`$: $`(i\mathrm{}\gamma ){\displaystyle \frac{\psi }{t}}`$ $`=`$ $`{\displaystyle \frac{\mathrm{}^2}{2m}}^2\psi +a\psi +b\|\psi \|^2\psi +\eta (𝐫,t);`$ (3) $`\eta (𝐫,t)\eta ^{}(𝐫^{},t^{})`$ $`=`$ $`2k_BT\gamma \delta (tt^{})\delta (𝐫𝐫^{}),`$ (4) $`\eta (𝐫,t)\eta (𝐫^{},t^{})`$ $`=`$ $`0.`$ (5) The damping coefficient $`\gamma `$ and the stochastic force $`\eta (𝐫,t)`$ model the coupling of the condensate to a heat bath, satisfying the fluctuation-dissipation relation. Coefficients $`a`$, $`b`$ and $`m`$ are assumed constant and independent of temperature; thermal effects are fully described by the stochastic force. The above equations (5) describe the phenomenology of physically important systems such as atomic Bose condensates and superfluid helium II . The same equation (at $`T=0`$), amended as appropriate by the vector potential, describes the order-parameter dynamics of type-II superconductors . Although the above is known to be strictly true only for a class of type-II superconductors with magnetic impurities , it is often assumed that Eqs. (5) apply to all type-II superconductors, unless indicated otherwise. We will henceforth study these equations as an important paradigmatic field theory representing a broad range of related physical phenomena. It is convenient to make all quantities dimensionless by standard substitutions. We set $`\psi =\stackrel{~}{\psi }\sqrt{a/b}`$, $`\gamma =\stackrel{~}{\gamma }\mathrm{}`$, $`t=\stackrel{~}{t}\gamma /(a)`$, $`𝐫=\stackrel{~}{𝐫}\mathrm{}/\sqrt{2am}`$, $`\eta =\stackrel{~}{\eta }(a)\sqrt{a/b}`$, and $`k_BT=a\mathrm{}^2/(2\stackrel{~}{\beta }mb)`$. Note that $`𝐫`$ is normalized to a unit of length $`\mathrm{}/\sqrt{2am}`$, which is the Ginzburg-Landau coherence length, usually denoted $`\xi `$. With these substitutions, and after dropping the tildes, Eqs. (5) become $`(i/\gamma 1){\displaystyle \frac{}{t}}\psi `$ $`=`$ $`^2\psi +[\|\psi \|^21]\psi +ϵ\eta (𝐫,t);`$ (6) $`\eta (𝐫,t)\eta ^{}(𝐫^{},t^{})`$ $`=`$ $`{\displaystyle \frac{2}{\beta }}\delta (tt^{})\delta (𝐫𝐫^{}),`$ (7) $`\eta (𝐫,t)\eta (𝐫^{},t^{})`$ $`=`$ $`0.`$ (8) The position of an isolated vortex, $`𝐑(t)`$, is defined via a contour integral $$[\text{arg}\psi (𝐫,t)]𝑑𝐥,$$ (9) which equals $`2\pi `$ when the integration path encloses the point $`𝐑(t)`$, and vanishes otherwise. The Brownian motion of vortices as point particles has been argued to hold primarily in the overdamped limit ($`\gamma \mathrm{}`$), which is directly relevant to the superconducting case; this is also the limit studied in this paper. We envision a situation in which a single vortex exists in the ground state of the field $`\psi (𝐫,t)`$. In a thin superconducting film one can in principle achieve the same effect by placing the sample in a magnetic field of one flux quantum per sample area. In the latter case there are many other factors of practical importance, such as geometrical and point pinning, which however will not be dealt with here. The vortex is assumed to be located near the center of the spatial extent of the condensate, so that boundary effects can be neglected. The radius of the sample $`\mathrm{\Lambda }_c`$ provides a natural infrared cutoff. At sufficiently low temperature, thermally induced vortex-antivortex pairs will be exponentially suppressed and we may assume that there is only one vortex in a spatially bounded superfluid film at all times. This assumption has been verified numerically for a range of temperatures and sample sizes studied. ## III Perturbative Analysis of Vortex Transport In this Section we utilize the singular perturbation expansion of Kaup to extract the vortex position from the stochastic equation of motion (8). We first expand the field as $`\psi (𝐫,t)`$ $`=`$ $`\psi _0(𝐗)+ϵ\varphi (𝐫,t);`$ (10) $`\varphi (𝐫,t)`$ $`=`$ $`\varphi _1(𝐗,t)+ϵ\varphi _2(𝐗,t)+\mathrm{},`$ (11) where $`𝐗(X,Y)`$ is the comoving coordinate, measured relative to a moving reference point; $`𝐗(t)=𝐫𝐑^0(t)`$. The absence of explicit time dependence in $`\psi _0(𝐗)`$ in the comoving reference frame indicates that it is the static vortex field in the absence of fluctuations. Tomboulis has shown that introduction of a collective coordinate such as $`𝐑^0(t)`$, while conserving the total number of dynamical degrees of freedom, is a canonical transformation. The fluctuating field $`\varphi (𝐫,t)`$ can be thought of as a superposition of harmonic modes in the background of the rigid vortex $`\psi _0(𝐗)`$. The phonons are gapless , therefore they will be an important consideration at any finite temperature. The variable $`ϵ1`$ plays the role of a small parameter as well as that of a bookkeeping device to control the perturbation series . Another useful way to think about $`ϵ`$ emerges when one absorbs $`ϵ`$ into the definition of $`\eta (𝐫,t)`$ in Eq. (8): $`ϵ`$ then reappears as the square root of temperature in the fluctuation-dissipation relation in Eqs. (8). Therefore, the small-$`ϵ`$ expansion is actually the small temperature or weak noise expansion. The equation of motion (8) is required to hold at every order in $`ϵ`$, which results in a hierarchy of equations for $`\psi _0(𝐗)`$, $`\varphi _1(𝐗,t)`$, etc. Because of the added fluctuations in Eq. (10), the actual vortex position $`𝐑(t)`$ as the singularity of the full dynamical field $`\psi (𝐫,t)`$ does not in general coincide with the coordinate $`𝐑^0(t)`$, which by definition is the singularity of the rigidly moving static vortex field $`\psi _0(𝐗)`$. The two are related via $$𝐑(t)=\stackrel{}{\rho }[𝐑^0(t),ϵ\varphi ]=𝐑^0(t)+ϵd^2𝐫\frac{\delta \stackrel{}{\rho }[𝐑^0(t),\varphi ]}{\delta \varphi (𝐫)}\|_{\varphi =0}\varphi (𝐫,t)+\text{c.c.}+\mathrm{},$$ (12) where $`\stackrel{}{\rho }[𝐑^0(t),ϵ\varphi ]`$ is a vector functional of the fluctuation field $`ϵ\varphi (𝐫,t)`$ and of the reference point $`𝐑^0(t)`$. We define $`\dot{𝐑}^0(t)`$ $`=`$ $`0+ϵ𝐯_1^0(t)+ϵ^2𝐯_2^0(t)+\mathrm{},`$ (13) $`\dot{𝐑}(t)`$ $`=`$ $`0+ϵ𝐯_1(t)+ϵ^2𝐯_2(t)+\mathrm{},`$ (14) which makes explicit the fact that there is no vortex motion in the absence of thermal fluctuations. Differentiating Eq. (12) with respect to $`t`$, we find at order $`ϵ`$, $$𝐯_1(t)=𝐯_1^0(t)+𝐕[𝐑(t),t],$$ (15) where $`𝐕[𝐑(t),t]\frac{d}{dt}d^2𝐫\frac{\delta \stackrel{}{\rho }[𝐑(t),\varphi ]}{\delta \varphi (𝐫)}\|_{\varphi =0}\varphi (𝐫,t)+\text{c.c.}`$ is a time-dependent velocity field. In the last expression we replaced $`𝐑^0(t)`$ with $`𝐑(t)`$, as the two are equal to $`𝒪(ϵ)`$. The statistical properties of $`𝐕[𝐑(t),t]`$ are given by those of the excitation field $`\varphi (𝐫,t)`$, i.e., the phonons, and will therefore depend on temperature. Substituting Eqs. (10), (11), (13) and (14) into Eqs. (8) and collecting powers of $`ϵ`$ we obtain, at order $`ϵ^0`$, $$0=_𝐗^2\psi _0[1\|\psi _0\|^2]\psi _0.$$ (16) The properties of the static vortex solution are well known; $`\psi _0(𝐗)=f(\|𝐗\|)\mathrm{exp}[i\text{arg}(X+iY)]`$, where $`f(x)x`$ for $`x1`$ and $`f(x)1`$ for $`x1`$. At order $`ϵ`$ we find $$\frac{}{t}\left(\begin{array}{c}\varphi _1\\ \varphi _1^{}\end{array}\right)=\left(\begin{array}{c}\varphi _1\\ \varphi _1^{}\end{array}\right)+𝐯_1^0_𝐗\left(\begin{array}{c}\psi _0\\ \psi _0^{}\end{array}\right)\left(\begin{array}{c}\eta \\ \eta ^{}\end{array}\right),$$ (17) where $``$ is a Hermitean matrix $$\left(\begin{array}{cc}_𝐗^2+[12\|\psi _0\|^2]\hfill & \hfill \psi _0^2\\ (\psi _0^{})^2\hfill & \hfill _𝐗^2+[12\|\psi _0\|^2]\end{array}\right).$$ (18) The noise $`\eta (𝐗,t)`$ remains white in space and time also in the moving reference frame, just as in the original Eqs. (8). Differentiating Eq. (16) with respect to $`X,Y`$ we find $$\left(\begin{array}{c}_X\psi _0\\ _X\psi _0^{}\end{array}\right)=\left(\begin{array}{c}_Y\psi _0\\ _Y\psi _0^{}\end{array}\right)=0.$$ (19) Hence the null space of the linear operator $``$ is spanned by the eigenvectors $`(_X\psi _0,_X\psi _0^{})^T`$ and $`(_Y\psi _0,_Y\psi _0^{})^T`$. These generate uniform translations of $`\psi _0`$ in the $`xy`$ plane and correspond to the Goldstone modes. It turns out that all secular terms in the $`ϵ`$-expansion vanish when we demand that the arbitrary perturbation $`(\varphi ,\varphi ^{})^T`$ be orthogonal to the null space of the operator $``$. With this condition in place \[which also uniquely specifies the reference point $`𝐑^0(t)`$\] we now take the scalar product of Eq. (17) with the eigenvector $`(_X\varphi _0,_X\varphi _0^{})^T`$, and integrate over the sample area to obtain $$0=\text{Re}\left\{d^2𝐗(_X\psi _0^{})[𝐯_1^0(t)_𝐗\psi _0\eta (𝐗,t)]\right\}.$$ (20) The same equation is obtained if we use the other eigenvector in the scalar product, only now $`_X\psi _0^{}`$ is replaced by $`_Y\psi _0^{}`$. Solving Eq. (20) for $`𝐯_1^0(t)`$ leads to $$𝐯_1^0(t)=𝐅(t),$$ (21) where the stochastic force $`𝐅(t)`$ is Gaussian with zero mean and autocorrelator $$F_\mu (t)F_\nu (t^{})=\frac{2}{\beta M(\mathrm{\Lambda }_c)}\delta (tt^{})\delta _{\mu \nu },$$ (22) where $`\mu ,\nu \{X,Y\}`$, and $$M(\mathrm{\Lambda }_c)^{\mathrm{\Lambda }_c}d^2𝐗\|_𝐗\psi _0(𝐗)\|^22\pi \mathrm{ln}\mathrm{\Lambda }_c$$ (23) can be interpreted as the (cutoff-dependent) inertial mass of the vortex. The fully dimensional form of this equation is $`M(\mathrm{\Lambda }_c)(2\pi \gamma ^2/b)\mathrm{ln}(\mathrm{\Lambda }_c/\xi )`$. This quantity, in a different derivation, has received the same interpretation by Šimánek . The so-called vortex core mass is implicitly present in the first part of Eq. (18) and makes a constant contribution to $`M(\mathrm{\Lambda }_c)`$, whereas the asymptotic expression in the latter part of Eq. (18) contains the $`\mathrm{\Lambda }_c`$-dependence which is dominant in the limit of large $`\mathrm{\Lambda }_c`$. In charged superfluids there is yet another contribution to the vortex inertial mass, the so-called electromagnetic mass, which originates from the energy of the electric field generated by a moving vortex. This contribution is absent in Eq. (18), which strictly applies only to neutral superfluids. Substituting for $`𝐯_1^0(t)`$ in Eq. (21) from Eq. (15) and using definitions (13) and (14), we arrive at an equation of motion for the vortex, valid to $`𝒪(ϵ)`$, $$\frac{d}{dt}𝐑(t)=ϵ𝐅(t)+ϵ𝐕[𝐑(t),t].$$ (24) As the last step we set $`ϵ=1`$ in this equation. If we assume that $`𝐕[𝐑(t),t]`$ is slowly varying relative to the noise $`𝐅(t)`$, then we may treat $`𝐕[𝐑(t),t]`$ as an external field and arrive immediately at the Fokker-Planck equation for the density $`c(𝐫,t)\delta [𝐫𝐑(t)]`$: $$\frac{c(𝐫,t)}{t}+[𝐕(𝐫,t)c(𝐫,t)]=\sigma ^2c(𝐫,t),$$ (25) where the diffusion coefficient $$\sigma \frac{1}{\beta M(\mathrm{\Lambda }_c)}.$$ (26) This equation is of the same form as that for a passive scalar advected to background fluid flow $`𝐕(𝐫,t)`$ . It is well established—particularly in the best-studied case of incompressible fluid flow $`𝐕(𝐫,t)=0`$—that in contrast to pure diffusion, an advected passive scalar may display non-Gaussian (for example, exponential) tails in the probability density function (PDF) $`c(𝐫,t)`$. These tails arise due to nontrivial spacetime correlations in the advecting velocity field . Interestingly, neither diffusion nor advection when acting alone can produce non-Gaussian tails in the PDF’s. In our case, the non-Gaussian tails would appear in the PDF of vortex displacements during a fixed time interval $`\mathrm{\Delta }t`$ (vortex flights). In order to test this possibility without detailed knowledge of the velocity field $`𝐕(𝐫,t)`$, we study the thermal motion of an isolated vortex numerically as described below. ## IV Numerical Analysis and Results In this section we describe our numerical methods used to solve Eqs. (8) under the constraint that only a single vortex be present in the simulation volume. An unpaired vortex is forced into the sample by coupling the field $`\psi (𝐫,t)`$ to a static external gauge potential $`𝐀(𝐫)`$, replacing $``$ in Eq. (6) by $`+i𝐀(𝐫)`$. If we choose $`𝐀(𝐫)=(2\pi /L^2)y\widehat{x}`$, there will be on average one vortex per area $`L\times L`$ in an infinite 2D sample. The introduction of $`𝐀(𝐫)`$ is a mathematical artifice and does not represent a dynamical field. In the language of superconductivity, $`𝐀(𝐫)`$ represents an external uniform magnetic field that completely penetrates the sample without screening, and vanishes in the thermodynamic limit $`L\mathrm{}`$. This is a realistic approximation to a superconducting thin film geometry, as in this case the external magnetic flux must fully penetrate the superconductor. Screening is only manifested as an inhomogeneity of the flux density threading the plane of the superconductor. The characteristic length scale for these variations is $`\mathrm{\Lambda }=\lambda ^2/d`$, where $`\lambda `$ is the bulk magnetic penetration length of the superconducting material, and $`d`$ is the thickness of the film. We study the system in the limit where $`\mathrm{\Lambda }`$ is much larger than any other length of the problem. In this limit the flux density threading the superconductor is uniform and screening can be neglected. One might also worry that the derivation leading up to Eq. (18), which was done explicitly for the case of the neutral superfluid, no longer applies when $`𝐀(𝐫)`$ is introduced. While it is a fact that the massless Goldstone boson in a superluid is in general replaced by a massive vector boson in a superconductor, the mass of this boson vanishes in the thin film limit considered, and the gapless phonon spectrum is restored, as pointed out by Šimánek. By the same token the electromagnetic vortex mass vanishes in our thin film limit, as the moving vortex does not generate any electric field. We choose our system to be a square, $`S=0,L)\times 0,L)`$. We wish to employ a variant of periodic boundary conditions so that the entire $`xy`$ plane consists of physically equivalent copies of the system at all times; by this we mean that the local superfluid density $`\|\psi (𝐫,t)\|^2`$, and the supercurrent density $$𝐉(𝐫,t)=\frac{i}{2}(\psi \psi ^{}\psi ^{}\psi )+\|\psi \|^2𝐀$$ (27) should both be periodic functions with periodicity $`L`$ in both spatial directions. To this end we impose quasi-periodic boundary conditions on $`\psi (𝐫,t)`$ ($`𝐫S`$, $`n_y`$ and $`n_y`$ arbitrary integers): $$\psi (𝐫+n_xL\widehat{x}+n_yL\widehat{y},t)=\psi (𝐫,t)\mathrm{exp}[in_x\theta _x+in_y(2\pi x/L+\theta _y)].$$ (28) The angles $`\theta _x`$ and $`\theta _y`$ are as yet arbitrary and may be functions of time. Upon discretization of Eqs. (8) we observe that any given time dependence of $`\theta _x`$ and $`\theta _y`$ will in general affect the motion of $`\psi (𝐫,t)`$. Yet $`\theta _x`$ and $`\theta _y`$ are not dynamical degrees of freedom in the sense that their motion is not determined by Eqs. (8). Therefore, we are left with a problem of choosing the “right” time dependence for $`\theta _x`$ and $`\theta _y`$. Fortunately the physical interpretation of these variables is straightforward. In order to find it, we rewrite the stochastic field equation in (8) in the form $$(i/\gamma 1)\frac{}{t}\psi =\frac{\delta H[\psi ,\psi ^{},𝐀]}{\delta \psi ^{}}+ϵ\eta (𝐫,t),$$ (29) where $$H[\psi ,\psi ^{},𝐀]d^2𝐫\left[\frac{1}{2}\|(+i𝐀)\psi \|^2\|\psi \|^2+\frac{1}{2}\|\psi \|^4\right]$$ (30) is the Hamiltonian of the system (the condensate) disconnected from the thermal reservoir. One can show that $$\frac{H}{\theta _x}=\widehat{x}_0^L𝑑y\frac{\delta H}{\delta 𝐀(𝐫)}\|_{x=0}\widehat{x}_0^L𝑑y𝐉_s(𝐫,t)\|_{x=0}.$$ (31) The first part of this equation is an algebraic equality; in the second part we used the definition of the supercurrent, whose explicit form appears in Eq. (27). Therefore, the first derivative of the Hamiltonian with respect to $`\theta _x`$ is equal to the net supercurrent passing through the edge of the sample perpendicular to the $`x`$ axis. A completely analogous statement holds for the other spatial direction as well. If one is only interested in the thermal motion of the vortex in the absence of external forces—such as, for example, the Magnus force created by macroscopic superflow—the desired constraints on angles $`\theta _x`$ and $`\theta _y`$ are $$\frac{H}{\theta _x}=\frac{H}{\theta _y}=0.$$ (32) These constraints define a zero flow ensemble , relevant for experiments with no externally imposed driving flow. The angles $`\theta _x`$ and $`\theta _y`$, though not dynamical variables, have their values fixed by two dynamical constraints, (32), at all times. As a result of the quasi-periodic boundary conditions, the isolated vortex is part of an infinite (always perfectly square) vortex lattice, with lattice constant $`L`$. The physical cutoff $`\mathrm{\Lambda }_c`$ needed to evaluate the vortex mass is estimated as $`\mathrm{\Lambda }_cL/\sqrt{\pi }`$ (this is when a square of side $`L`$ and a circle of radius $`\mathrm{\Lambda }_c`$ have equal areas). Our numerical approach consists of simultaneously solving the lattice version of Eqs. (8) on the spatial domain $`S`$, and the two dynamical constraints (32). The numerical grid spacing $`\delta `$ must resolve both the vortex core size $`\xi `$, which is the unit of length, and the phonon correlation length $`\xi _p`$, which is the characteristic length scale of the four-point correlator $`\|\psi (𝐫,t)\|^2\|\psi (𝐫^{},t)\|^2`$, and depends on temperature. We have found it adequate to set $`\delta =0.15`$ throughout the temperature range studied. Numerical stability of the second-order stochastic Runge-Kutta method used to integrate Eqs. (8) then allows a time step of $`\tau =0.005`$. We study sample sizes $`L=6`$, 12, and 24, and temperatures $`1/\beta =0.0125`$, 0.025, 0.05, and 0.1. The vortex is clearly identifiable and no thermal vortex-antivortex pairs are generated during the numerical runs. These ranged in duration from $`6\times 10^4`$ to $`6\times 10^5`$, in dimensionless units (up to $`1.2\times 10^8`$ time steps). The characteristic equilibration time was $`500`$. The vortex position is taken to be the coordinate of the lower left-hand corner of the plaquette where it is detected with the discretized version of the integral (9). Fig. 1 shows $`[\mathrm{\Delta }𝐑(t)]^2`$ for three different $`L`$ and the same $`\beta `$. It is apparent from the data offsets that vortex motion is not purely diffusive. On the other hand, the calculated diffusion coefficient $`\sigma `$ correctly approximates the slopes of the numerical data at long times. Therefore we conclude that the vortex mass, measured through $`\sigma `$, displays a logarithmic divergence with the system size $`L`$. The issue of the value of the vortex inertial mass has been the subject of much controversy in the literature . Our results show that even in the presence of random flow the logarithmic divergence of the classical vortex mass persists. This stands in agreement with an earlier result in the presence of coherent flow. It would be interesting to investigate whether this divergence persists in the presence of multiple vortices or in the limit of low damping. An independent measure of diffusion is obtained through the concept of the mean first passage time $`\overline{t}(\mathrm{})`$ over the boundaries of the square box of size $`2\mathrm{}\times 2\mathrm{}`$. For pure diffusion one has $`\overline{t}(\mathrm{})=0.295\mathrm{}^2/\sigma `$ . Fig. 2 reiterates that the vortex motion is diffusive at long times. Deviations from pure diffusion are manifest at short times that correspond to vortex flights of $`𝒪(1)`$ and less. The evidence from Fig. 2 also shows that $`\sigma `$ depends linearly on temperature, as expected from Eq. (26). This is a useful check on the quality of our numerics—the vortex is not subject to spurious pinning due to the discreteness of the underlying lattice. The flattening-out of the data for large $`\mathrm{}`$ is due to finite sampling. Finally in Fig. 3 we present the PDF of vortex flights. This distribution has exponential tails (straight lines on a logarithmic scale as in Fig. 3), in addition to the expected Gaussian profile typical of Brownian motion. Our numerical results show that these tails persist for all system sizes and $`\beta `$ studied, and the slope decreases as the vortex mass $`M(\mathrm{\Lambda }_c)`$ is increased (which corresponds to a volume increase). The presence of exponential tails has been linked to nontrivial time correlations in the advecting velocity field . The presence of finite correlation times can be either a feature of a stochastic velocity field or can arise from deterministic, chaotic motion (e.g., intermittency). In our case, we expect the exponential tails to arise from the time correlations of overdamped phonon modes which are allowed to exist in the fluctuation spectrum of our theory. This aspect is presently under investigation. ## V Discussion and Conclusions The particular models of vortex motion of the type described in the Introduction typically ignore the presence of hydrodynamic modes. In contrast, real physical systems, in particular those described by stochastic field theories, necessarily include these effects. Consequently, as demonstrated here both analytically and numerically, the motion of a single vortex must be viewed as diffusive and randomly advected at the same time. The Fokker-Planck equation (25) makes explicit the formal analogy between single vortex fluctuations and the dynamics of a passive scalar advected to random fluid flow. The quantity $`c(𝐫,t)`$ is the ensemble-averaged probability density of finding a vortex at $`𝐫`$ at an instant $`t`$. The ensemble averaging is done over the fast degrees of freedom of the heat bath. The slow, hydrodynamic, degrees of freedom of the superfluid—the phonons—remain implicitly present in the velocity field $`𝐕(𝐫,t)`$. This equation presents a picture of the vortex as a diffusing massive particle subject to random advection. The advection is apparently due to phonons, which propagate through the vortex core randomly. As a result, the probability of large-scale vortex flights is increased. In turbulent fluid flow experiments with a dye as the passive scalar it is possible to measure the dye concentration, analogous to our $`c(𝐫,t)`$, directly, because individual dye particles do not interact among themselves (apart from contact interaction). Vortices, however, interact logarithmically, so one cannot confirm the exponential tails in $`c(𝐫,t)`$ directly by studying motion of a large aggregate of vortices in a sample. Instead, one must study the motion of a single vortex, and measure the statistics of vortex flights. Examples where thermal vortex motion plays a key role include theories of thermal depinning, flux creep, and dilute vortex lattice melting in high-$`T_c`$ superconductors. Our current understanding of these phenomena relies exclusively on the Brownian particle description. It would be interesting to re-examine these problems in the light of the advected scalar aspect of vortex motion. An exciting new possibility is a direct experimental observation of vortex motion in two-dimensional high-$`T_c`$ superconductors by fast scanning tunneling microscopy . With this method the overwhelmingly more frequent occurrence of large vortex flights in comparison with that of a simple Brownian particle could be put to direct test. ## VI Acknowledgments We thank A.R. Bishop, D.P. Arovas, T. Hwa, and G. Lythe for useful discussions. Simulations were carried out at the Advanced Computing Laboratory (ACL), Los Alamos National Laboratory, the Department of Physics, Ohio State University, and at the National Energy Research Scientific Computing Center (NERSC), Lawrence Berkeley National Laboratory.
no-problem/9907/hep-ph9907411.html	ar5iv	text	# Mass shift, width broadening and spectral density of 𝜌-mesons produced in heavy ion collisions ## The problem of how the properties of hadrons change in hadronic or nuclear matter in comparison to their free space values has attracted a lot of attention. It is clear on physical grounds that the in-medium mass shift and width broadening of a particle are only due to its interaction with the constituents of the medium, for not too dense media anyway. Thus one can use phenomenological information on this interaction to calculate the mass shift and width broadening. For meson $`a`$ scattering on hadron $`b`$ in the medium the contribution to the self-energy is: $$\mathrm{\Pi }_{ab}(E,p)=4\pi \frac{d^3k}{(2\pi )^3}n_b(\omega )\frac{\sqrt{s}}{\omega }f_{ab}^{(\mathrm{cm})}(s)$$ (1) where $`E`$ and $`p`$ are the energy and momentum of the meson, $`\omega ^2=m_b^2+k^2`$, $`n_b`$ is the occupation number, and $`f_{ab}`$ is the forward scattering amplitude. The normalization of the amplitude corresponds to the standard form of the optical theorem $`\sigma =(4\pi /q_{\mathrm{cm}})\mathrm{Im}f^{(\mathrm{cm})}(s)`$. The applicability of eq. (1) is limited to those cases where interference between sequential scatterings is negligible. In the limit that the target particles $`b`$ move nonrelativistically, $`\mathrm{\Pi }_{ab}=4\pi f_{ab}^{(b\mathrm{rest}\mathrm{frame})}\rho _b`$, where $`\rho _b`$ is the spatial density. This corresponds to the mass shift and width broadening $$\mathrm{\Delta }m_a(E)=2\pi \frac{\rho }{m_a}\mathrm{Re}f_{ab}^{(b\mathrm{rest}\mathrm{frame})}(E),\mathrm{\Delta }\mathrm{\Gamma }_a(E)=\frac{\rho }{m_a}k\sigma _{ab}(E).$$ (2) These relations hold also in the general case provided the amplitudes are averaged over momentum distributions of the constituents. We assumed that $`\rho `$-mesons are formed during the last stage of the evolution of hadronic matter created in a heavy ion collision, when the matter can be considered as a weakly interacting gas of pions and nucleons. This stage is formed when the local temperature is on the order of 100 to 150 MeV and when the local baryon density is on the order of the normal nucleon density in a nucleus. The description of nuclear matter as such a gas, of course, cannot be considered as a very good one, so it is clear that our results may be only semiquantitative. The main ingredients of our calculation are on-shell $`\rho \pi `$ and $`\rho N`$ forward scattering amplitudes and total cross sections. The scattering amplitudes were obtained by saturation of the low energy part with resonances and using a combination of vector meson dominance (VMD) and Regge theory at high energy. Using VMD we are restricted to the case of transversally polarized $`\rho `$-mesons. One can argue though, that for unpolarized $`\rho `$-mesons our results should be multiplied by a factor ranging from 2/3 to 1. Using experimental information on momentum distributions of pions and nucleons and on the $`\pi /N`$ ratio shows that $`\mathrm{\Delta }m_\rho `$ tens of MeV, while the width increases by several hundred MeV at beam energies of a few GeV$``$A and by twice that amount at about a hundred GeV$``$A. We also considered the $`\rho `$ meson dispersion relation for finite temperature and baryon density for momenta up to a GeV/c or so as this is very interesting for the production of dileptons in high energy heavy ion collisions. The dispersion relation is determined by the poles of the propagator after summing over all target species and including the vacuum contribution to the self-energy. In the narrow width approximation we have $`E_R^2(p)`$ $`=`$ $`p^2+m_\rho ^2+\mathrm{Re}\mathrm{\Pi }_{\rho \pi }(p)+\mathrm{Re}\mathrm{\Pi }_{\rho N}(p),`$ $`\gamma (p)`$ $`=`$ $`\left[\mathrm{Im}\mathrm{\Pi }_\rho ^{\mathrm{vac}}+\mathrm{Im}\mathrm{\Pi }_{\rho \pi }(p)+\mathrm{Im}\mathrm{\Pi }_{\rho N}(p)\right]/E_R(p).`$ (3) where $`E(p)=E_R(p)i\gamma (p)/2`$. We can also define a mass shift, $$\mathrm{\Delta }m_\rho (p)=\sqrt{m_\rho ^2+\mathrm{Re}\mathrm{\Pi }_{\rho \pi }(p)+\mathrm{Re}\mathrm{\Pi }_{\rho N}(p)}m_\rho $$ (4) While $`\mathrm{\Delta }m_\rho `$ in eq.(4) coincides with the one in eq.(1) for small $`\mathrm{Re}\mathrm{\Pi }/m_\rho `$, the two definitions of width correspond to the rest frame of the $`\rho `$-meson (eq.(2)) and to that of the thermal system (eq.(3)) and differ by the time dilation factor, $`\gamma =\mathrm{\Gamma }m_\rho /E_R(p)`$. We evaluate $`\mathrm{\Delta }m`$ and $`\gamma `$ for $`T`$ = 100 and 150 MeV and nucleon densities of 0, 1 and 2 times normal nuclear matter density (0.155 nucleons/fm<sup>3</sup>). This is done by utilizing a Fermi-Dirac distribution for nucleons. The nucleon chemical potentials are 745 and 820 MeV for densities of 1 and 2 times normal at $`T`$ = 100 MeV, and 540 and 645 MeV for densities of 1 and 2 times normal at $`T`$ = 150 MeV. Anti-nucleons are not included. Fig.1 shows the mass shift for different temperatures and nucleon densities (in units of normal nuclear density). The effect with pions alone is negligible (on the order of 1 MeV). The main effect comes from nucleons. The effective mass increases with nucleon density and with momentum, but is almost independent of temperature. These trends and numbers are roughly consistent with other analyses . Fig.2 shows the behavior of the $`\rho `$ meson width $`\gamma (p)`$. Once again pions have very little effect. The main effect comes from nucleons. Contrary to $`\rho `$ mesons moving in vacuum or through a pure pion gas the width remains roughly constant with momentum when nucleons are present. The width is about 240 MeV at 1 times nuclear density and about 370 MeV at 2 times nuclear density. This means that the $`\rho `$ meson becomes a rather poorly defined excitation with increasing nucleon density. The rate of dilepton production is proportional to the imaginary part of the photon self-energy which is itself proportional to the imaginary part of the $`\rho `$ meson propagator because of VDM. $$E_+E_{}\frac{dR}{d^3p_+d^3p_{}}\frac{\mathrm{Im}\mathrm{\Pi }_\rho }{[M^2m_\rho ^2\mathrm{Re}\mathrm{\Pi }_\rho ]^2+[\mathrm{Im}\mathrm{\Pi }_\rho ]^2}$$ (5) The vacuum part of $`\mathrm{\Pi }_\rho `$ can only depend on the invariant mass, $`M^2=E^2p^2`$, whereas the matter parts can depend on $`E`$ and $`p`$ separately. Since we are using the on-shell amplutudes, the matter parts only depend on $`p`$ because $`M`$ is fixed at $`m_\rho `$. The vacuum parts are obtained from the Gounaris-Sakurai formula . The imaginary part of the propagator is proportional to the spectral density. The former is plotted in Fig.3 for a pure pion gas and in Fig.4 for a gas of pions and nucleons at $`T`$ and $`n`$ characteristic of the final stages of a high energy heavy ion collision. Pions have very little effect on the spectral density even at such a high temperature. The effect of nucleons, however, is dramatic. The spectral density is greatly broadened, so much so that the very idea of a $`\rho `$ meson may lose its meaning. The above remarks on the relative importance of pions and nucleons may need to be re-examined when really applying these calculations to heavy ion collisions, where $`\pi /N`$ 6. The pions seem to be overpopulated in phase space, compared to a thermal Bose-Einstein distribution, and this could be modeled either by introducing a chemical potential for pions or simply by multiplication by an overall normalization factor. Pions would need to be enhanced by a substantial factor (5 or more) to make a noticeable contribution at a density of 0.155 nucleons per fm<sup>3</sup>. Recently data in Pb-Au collisions at 160 GeV$``$A have been presented where it was found that the $`\rho `$-peak is absent at $`k_T(e^+e^{})<400`$ MeV, but reappears at $`k_T(e^+e^{})>400`$ MeV. This seems to be just the opposite of our findings. However, our calculations refer to the $`\rho `$ momentum relative to the local rest frame of the matter and a low momentum $`\rho `$ may actually be moving faster relative to the outflowing matter than a higher momentum one. No conclusion can really be drawn without putting our results into a space-time model of the evolution of matter. In summary, we have studied the properties of the neutral $`\rho `$ meson in the gas of pions and nucleons with experimental and thermal momentum distributions. In the former case, pions give the dominant effect, in the latter case they are not important. This difference is due to completely different $`\pi /N`$ ratios. However, in both cases interaction with the gas provides a generally positive mass shift for the $`\rho `$ mesons and greatly increase their width. The $`\rho `$ meson spectral density is so broadened that the $`\rho `$ may lose its identity as a well defined particle or resonance. At sufficiently high energy density the matter can no longer be described very well as a gas of noninteracting pions and nucleons. Nevertheless the trends must be obeyed by any realistic calculations of the $`\rho `$ meson in-medium. Applications to thermal and hydrodynamic models of heavy ion collisions are under investigation. ## Acknowledgments The work reported here was supported in part by the RFBR grant 97-02-16131, CRDF grant RP2-132 and the US Department of Energy grant DE-FG02-87ER40382. V.L.E. thanks the Organizing Committee of QM99 for local support.
no-problem/9907/cond-mat9907181.html	ar5iv	text	# Theory of the 𝜋 state in 3He Josephson junctions \[ ## Abstract The flow of superfluid <sup>3</sup>He-B through a $`65\times 65`$ array of nanometer size apertures has been measured recently by Backhaus et al. They find in the current–phase relation a new branch, so-called $`\pi `$ state. We study two limiting cases which show that the $`\pi `$ state arises from coupling of the phase degree of freedom to the spin-orbit rotation. The $`\pi `$ state exists in a single large aperture, but is difficult to observe because of hysteresis. A better correspondence with experiments is obtained by assuming a thin wall, where the Josephson coupling between the two sides arises from a dense array of pin-holes. \] The flow of superfluid <sup>3</sup>He-B through a single nanometer-size aperture was studied by Avenel and Varoquaux some time ago . At temperatures near the superfluid transition temperature $`T_\mathrm{c}`$, the current–phase relation is sinusoidal $$J(\varphi )=J_\mathrm{c}\mathrm{sin}\varphi $$ (1) as expected for a Josephson junction. Also according to expectation, they find that the sine form (1) gradually becomes tilted when the temperature is lowered. More recently, Backhaus et al. studied a $`65\times 65`$ array of small apertures . They discovered a new behavior where the current–phase relation acquires a positive slope at phase differences $`\varphi \pi `$. This $`\pi `$ state develops when the temperature is lowered to approximately $`0.6T_\mathrm{c}`$. A few theoretical explanations for the $`\pi `$ state have been proposed . In this letter we present a theory, that is based on the many-component form of the order parameter in <sup>3</sup>He. It differs from the previous suggestions because it contains no unjustified assumptions and an order-of-magnitude agreement with experiments is obtained without any adjustable parameters. Unusual current-phase relations occur also in other systems. A $`\pi `$ junction, where $`J_\mathrm{c}`$ in (1) is negative, can be induced by adding magnetic impurities to a tunneling barrier between two s-wave superconductors . Similar $`\pi `$ shifts appear in nonmagnetic junctions between d-wave superconductors. In addition, current-phase relations with additional zeros ($`J(\varphi )=0`$ for $`\varphi 0`$ or $`\pi `$) can appear for special orientations of the anisotropic crystals . The $`\pi `$ state in <sup>3</sup>He differs from these in several respects, most fundamentally because it arises from interplay of two soft modes of the order parameter, the phase $`\varphi `$ and the spin-orbit rotation. We present calculations in two limiting cases. In the case of a tunneling barrier, the existence of the $`\pi `$ state can be demonstrated by analytic calculations. The parameters of the tunneling model are estimated using the quasiclassical theory. In the case of a single aperture, the $`\pi `$ state is obtained by numerical simulations using the Ginzburg-Landau theory of <sup>3</sup>He. Tunneling junction.—The simplest case to demonstrate the $`\pi `$ state is to consider a planar wall through which the <sup>3</sup>He atoms can tunnel. The energy arising from tunneling between the left ($`L`$) and right ($`R`$) sides can be written as $$F_\mathrm{J}=Re\underset{\mu }{}\left[aA_{\mu z}^LA_{\mu z}^R+b(A_{\mu x}^LA_{\mu x}^R+A_{\mu y}^LA_{\mu y}^R)\right].$$ (2) Here $`A_{\mu j}`$ is the $`3\times 3`$ matrix order parameter where the first index $`\mu `$ refers to the orientation of the Cooper pair in spin space, and the latter index $`j`$ in orbital space. The $`z`$ axis is taken perpendicular to the tunneling wall. Equation (2) is a simple generalization of $`F_\mathrm{J}=aRe(A^LA^R)`$, which describes the Josephson coupling of two s-wave superconductors with order parameters $`A^L`$ and $`A^R`$ . The mass current through the wall is given by $`J=(2m_3/\mathrm{})F_\mathrm{J}/\varphi `$, where $`\varphi =\varphi ^L\varphi ^R`$ is the phase difference and $`m_3`$ the mass of a <sup>3</sup>He atom. In the B phase of <sup>3</sup>He, the order parameter has the form $`A_{\mu j}=\mathrm{\Delta }\mathrm{exp}(\mathrm{i}\varphi )R_{\mu j}`$. Here $`\mathrm{\Delta }`$ is the amplitude, $`\mathrm{exp}(\mathrm{i}\varphi )`$ a phase factor, and $`R_{\mu j}`$ a $`3\times 3`$ rotation matrix: $`_\mu R_{\mu j}R_{\mu k}=\delta _{jk}`$. The rotation matrices can be parametrized by an axis $`\widehat{𝐧}`$ and an angle $`\theta `$. Substituting into (2) gives ($`\alpha `$, $`\beta >0`$) $$F_\mathrm{J}=\underset{\mu }{}\left[\alpha R_{\mu z}^LR_{\mu z}^R+\beta (R_{\mu x}^LR_{\mu x}^R+R_{\mu y}^LR_{\mu y}^R)\right]\mathrm{cos}\varphi .$$ (3) In deriving this expression from (2) one must pay attention to the fact that the order parameter of the p-wave superfluid is strongly suppressed near a wall. As a consequence the parameters $`\alpha `$ and $`\beta `$ in (3) are not simply related to the coefficients $`a`$ and $`b`$ in (2), but otherwise the dependence of $`F_\mathrm{J}`$ on the soft variables $`\varphi `$ and $`R_{\mu j}`$ remains the same as obtained by the simple substitution above. Let us consider the case that the rotation matrices on the left and right sides are the same. This gives rise to the “zero-state” with the critical current $`J_\mathrm{c}=(2m_3/\mathrm{})(\alpha +2\beta )>0`$. This state has lowest energy when $`\|\varphi \|<\pi /2`$ because it corresponds to the maximum of the expression in square brackets in (3). The situation changes when $`\varphi `$ exceeds $`\pi /2`$. There one has to look for a minimum of the expression in the square brackets. This corresponds to the $`\pi `$ state, which is illustrated by the solid line in Fig. 1a. The critical current $`J_\mathrm{c}`$ in (1) is negative: $`J_\mathrm{c}=(2m_3/\mathrm{})\alpha `$ if $`\alpha >\beta `$ and $`J_\mathrm{c}=(2m_3/\mathrm{})(2\beta \alpha )`$ otherwise. In order to make the tunneling model realistic, we have to consider three additional contributions to the energy. Firstly, there is the magnetic dipole-dipole energy $`F_\mathrm{d}=8g_\mathrm{d}\mathrm{\Delta }^2(\frac{1}{4}+\mathrm{cos}\theta )^2`$ . In the bulk it fixes the rotation angle $`\theta `$ equal to $`\theta _0=\mathrm{arccos}(\frac{1}{4})=104^{}`$. This remains valid also near the junction because both the Josephson energy (3) and the dipole-dipole energy can reach their minima simultaneously: the products of two rotation matrices appearing in the former are not limited by the fact that both matrices have a fixed rotation angle $`\theta _0`$. Secondly, there is a surface energy that arises from coupling of the dipole-dipole energy to the suppression of the order parameter near walls . It has the form $$F_\mathrm{s}=b_4(\widehat{𝐧}\widehat{𝐬})^4b_2(\widehat{𝐧}\widehat{𝐬})^2,$$ (4) where $`\widehat{𝐬}`$ is the surface normal. The lowest surface energy is achieved when the rotation axis $`\widehat{𝐧}`$ is perpendicular to the wall, $`\widehat{𝐧}=\pm \widehat{𝐬}`$, because $`b_2>2b_4>0`$. Thirdly, there is a gradient energy associated with spatial bending of the rotation axis $`\widehat{𝐧}`$. It arises because in practice all tunnel junctions are of finite size, and other walls in the container favor a different orientation of $`\widehat{𝐧}`$ than may be the minimum of the Josephson energy. We model the gradient energy by the simple quadratic forms $$F_\mathrm{g}^L=\gamma (\eta ^L\eta _{\mathrm{}}^L)^2,F_\mathrm{g}^R=\gamma (\eta ^R\eta _{\mathrm{}}^R)^2,$$ (5) where $`\eta `$ is the polar angle of $`\widehat{𝐧}`$, i.e., $`\mathrm{cos}\eta =\widehat{n}_z`$. $`\eta ^L`$ and $`\eta ^R`$ denote the polar angles on both sides just at the junction, and we assume that the values $`\eta _{\mathrm{}}^L`$ and $`\eta _{\mathrm{}}^R`$ further away are either $`0`$ or $`\pi `$. In the experimental case the surface energy (4) is important in fixing $`\eta _{\mathrm{}}^L`$ and $`\eta _{\mathrm{}}^R`$, but otherwise its contribution is so small that we can neglect it in the following. The current–phase relations for the tunneling model, (3) and (5), are plotted in Fig. 1. It can be seen that a large value of the gradient energy parameter $`\gamma `$ suppresses the $`\pi `$ state. Furthermore, we find two cases where the rotation axes $`\widehat{𝐧}`$ far from the junction are either parallel or antiparallel. The latter has smaller critical current \[$`J_\mathrm{c}=(2m_3/\mathrm{})(\alpha \frac{7}{4}\beta )`$\] but a relatively a more pronounced $`\pi `$ state. This “bi-stability” was theoretically discussed in Ref. and has recently been observed experimentally . Evaluation of tunneling parameters.—The experiment has a square array of apertures of diameter $`D=100`$ nm with spacing $`S=3\mu `$m in a wall of thickness $`W=50`$ nm. In order to make the tunneling model to imitate the experiment, we estimate $`\alpha `$ and $`\beta `$ by letting all the three lengths to approach zero but keeping their ratios unchanged . The calculation for such “pin-holes” is relatively simple once the self-consistent solution for the order parameter near a wall is known . We assume diffuse scattering of quasiparticles at surfaces. The tunneling form (3) is reproduced in these calculations at temperatures $`T0.5T_\mathrm{c}`$, and values of $`\alpha `$ and $`\beta `$ can be extracted. The parameter $`\gamma `$ can be estimated using the bending energy of the B phase and assuming a simple form $`\eta (𝐫)\eta _{\mathrm{}}r^1`$, where $`r`$ is the radius from the center of the aperture array, and this expression is cut off at the radius of the array. In agreement with experiments, we find that the $`\pi `$ state appears at low temperatures because $`\alpha `$ and $`\beta (1T/T_\mathrm{c})^2`$ have stronger temperature dependence than $`\gamma 1T/T_\mathrm{c}`$, see Fig. 2. Moreover, the parameters $`\alpha `$, $`\beta `$, and $`\gamma `$ agree within one order of magnitude to those that give an approximate best fit to the experiments (see caption of Fig. 2). This fit reproduces also the absolute magnitude of the critical current, and the same values of the parameters are used for cases of both parallel and antiparallel $`\widehat{𝐧}`$’s. The tunneling model can be improved trivially by extending the pin-hole calculation to the whole temperature range $`0<T<T_\mathrm{c}`$. A more ambitious project for the future would be the self-consistent calculation for aperture sizes on the order of the coherence length $`\xi _0`$. In both cases the resulting Josephson energy $`F_\mathrm{J}(\varphi ,R_{\mu j}^L,R_{\mu j}^R)`$ will no more be of the simple form (3). Single aperture.—The limit opposite to the tunneling barrier is a single large aperture. There the major task is to calculate the order parameter self-consistently. We have done this using the Ginzburg-Landau (GL) theory of <sup>3</sup>He. The differential equations were solved numerically on a grid in and around the aperture. Our calculations are more general than the previous ones because we use a full a three-dimensional grid. Vanishing $`A_{\mu j}`$ was assumed at surfaces. The order parameter of the $`\pi `$ state is shown in Fig. 3. It is plotted along the axis of a circularly symmetric aperture. For simplicity, we have normalized the order parameter to unit matrix in the bulk, $`A_{\mu j}(z=\pm \mathrm{})=\mathrm{exp}(\pm \mathrm{i}\varphi /2)\delta _{\mu j}`$ (assuming the case of parallel $`\widehat{𝐧}`$’s). This is possible because for aperture sizes on the order of the GL coherence length $`\xi _{\mathrm{GL}}`$ the dipole-dipole energy can be neglected. The characteristic property of the $`\pi `$ state is the components $`A_{yz}`$ and $`A_{zy}`$. These are the dominant components in the orifice, and they decay slowly towards the bulk. They imply broken symmetry: the symmetry group of the $`\pi `$ branch is $`m^{}m2^{}`$ compared to $`\frac{\mathrm{}}{m^{}}\frac{2^{}}{m}`$ of the zero branch. (Here prime denotes time-inversion.) This sets rather strong requirement for the calculation because the circular symmetry of the aperture cannot be used to simplify the computation. The current–phase relations are summarized in Fig. 4. We see that the occurrence of the $`\pi `$ branch depends sensitively on the diameter of the aperture, whereas the wall thickness is less important. For small apertures no $`\pi `$ branch is found. When $`D`$ exceeds approximately $`5\xi _{\mathrm{GL}}`$, the $`\pi `$ branch appears. In the region (b) the current–phase relation has negative slope. Such a state can be stabilized if the left and right sides are connected, like in a torus geometry. Increasing the diameter, the current–phase relation gets a positive slope in region (c). This state is stable also in a piston-driven flow channel. The $`\pi `$ state is also the absolute energy minimum at $`\varphi =\pi `$ in region (c). In region (d) the $`\pi `$ state continues to exist but it has higher energy than the zero branch. The calculation assumes the idealized case of flow between two infinite bulk fluids. Any additional hydrodynamic inductance shifts upwards the border between regions (c) and (d) as it increases the energy of current carrying states. Although the $`\pi `$ branch constitutes the absolute energy minimum, it may be difficult to find it experimentally in a single aperture. The reason is that whenever it is locally stable, there always exists a locally stable zero state at the same $`\varphi `$. Because the order parameters of the $`\pi `$ and zero states differ considerably, it may be that the phase slips only take place between two branches of the zero state without ever finding the way to the lower energy $`\pi `$ state. This is what we find in the numerical calculations, where the $`\pi `$ state was found only if the initial $`A_{\mu j}`$ was chosen close enough to the converged solution. We remind, though, that our calculations are not meant to simulate the correct dynamics of the phase slip. The dimensions of one aperture in the array at Berkeley are marked on figure 4. This is clearly in the region where no $`\pi `$ state is found. Although the Ginzburg-Landau calculation is accurate only at temperatures near $`T_\mathrm{c}`$, it is unlikely that the $`\pi `$ state could be stabilized in a more accurate calculation at lower temperatures. This statement is based on experience gained in previous low-temperature calculations . Thus we conclude that the appearance of the $`\pi `$ state in the Berkeley experiment essentially depends on the presence of many apertures. We have also done two-dimensional calculations that simulate the flow through a long narrow slit. The $`\pi `$ state is found and its properties are qualitatively similar to those in a circular aperture . In particular, the transitions from the zero branch to the $`\pi `$ branch seem to be absent. This is consistent with the fact that no $`\pi `$ branch was found in the experiments by Avenel and Varoquaux . The $`\pi `$ state can be interpreted so that a half-quantum vortex has crossed the orifice. There are no free half quantum vortices in superfluid <sup>3</sup>He-B, but the double-core vortex can be interpreted as a bound pair of two half-quantum vortices . Indeed, the order parameter in Fig. 3 is very similar to that in the double-core vortex on the axis going between the two cores . Conclusion.—The $`\pi `$ state was found to occur in both the limits investigated above. Its mechanism is the same in both cases: a lower coupling energy is achieved by producing a spin-orbit rotation that heals slowly in the bulk liquid. The Berkeley experiment is somewhat an intermediate case to the two limits studied, which makes it evident that the $`\pi `$ state also there arises from the same mechanism. The theory above provides several predictions that can be tested experimentally. For example, the $`\pi `$ state depends on the linear dimension $`L`$ of the aperture array because $`\alpha /\gamma `$, $`\beta /\gamma L`$. An external magnetic field fixes the surface orientation of $`\widehat{𝐧}`$, and thus can be used to suppress the $`\pi `$ state.
no-problem/9907/hep-ph9907309.html	ar5iv	text	# 1 Introduction ## 1 Introduction The successful description of the nucleon structure function by perturbative QCD, using the DGLAP (Dokshitzer-Gribov-Lipatov-Altarelli-Parisi) parton evolution equations constitutes a major success of QCD. However, at very small Bjorken-$`x`$, these equations are expected to become invalid. An alternative ansatz for the small-$`x`$ regime is the BFKL (Balitsky-Fadin-Kuraev-Lipatov) equation . At lowest order, the BFKL and DGLAP equations resum the leading logarithmic $`(\alpha _s\mathrm{ln}1/x)^n`$) or $`(\alpha _s\mathrm{ln}Q^2/Q_0^2)^n`$) contributions, respectively, with $`Q^2`$ being the virtuality of the exchanged photon in deep-inelastic neutral current $`ep`$ collisions. The leading-log DGLAP ansatz implies strong ordering ($`Q_0^2k_{T}^{}{}_{1}{}^{2}\mathrm{}k_{T}^{}{}_{i}{}^{2}\mathrm{}Q^2`$) of the transverse momenta $`(k_T)_i`$ in the parton cascade. Here, $`(k_T)_i`$ is the transverse momentum (measured relative to the incident proton direction) of the $`i`$-th emitted “final state” parton (see Fig. 1). According to BFKL kinematics, the transverse momenta are no longer strongly ordered but may be thought of as forming, roughly speaking, a random walk with $`(k_T^2)_i(k_T^2)_{i+1}`$. Measurements of the hadronic final state resulting from the initial-state parton cascade should be sensitive to the type of evolution. For example, without strong $`k_T`$-ordering, more transverse energy $`E_T`$ is to be expected from the BFKL than from the DGLAP type of evolution in a rapidity interval between the proton fragmentation region and the current fragmentation region. A similar expectation holds for the inclusive transverse momentum distribution of hadrons . So far, however, none of the studied observables (including recent measurements of forward jet and large-$`p_T`$ charged and neutral pions ) are directly probing the correlation structure of the parton cascade. To allow additional discrimination between a strictly $`k_T`$-ordered and an unordered scenario, it is necessary to go beyond single-particle observables and consider directly measures of the transverse momentum correlations in the parton cascade. ## 2 The transverse momentum transfer In this paper we consider the transverse momentum transfer across rapidity $`y`$, $`\stackrel{}{\mathrm{\Pi }}(y)`$. This quantity is defined as the vector-sum of the transverse momenta $`\stackrel{}{k_i}`$ of all particles with rapidities smaller than $`y`$ in an event with $`n`$ particles. For simplicity, we consider in the following only one of the components of the transverse momentum vectors (denoted by $`k_i`$) and define $$\mathrm{\Pi }(y)=\underset{i=1}{\overset{n}{}}k_i\theta (yy_i),$$ (1) where $`\theta (x)=0`$ for $`x<0`$ and $`\theta (x)=1`$ for $`x0`$. $`\mathrm{\Pi }(y)`$ is a random function, varying from event to event. A set of $`k`$-th order moment-functions can be constructed defined as<sup>1</sup><sup>1</sup>1See for a discussion of the theory of random functions $$\mathrm{\Pi }(y_1)\mathrm{\Pi }(y_2)\mathrm{}\mathrm{\Pi }(y_k),$$ (2) where $``$ means that the average is taken over an ensemble of events. From the moment-functions, one can further define a set of correlation functions by taking a cluster decomposition (see eg. ). The two-point correlation function $`D^2(y_1,y_2)`$ is then given by $$D^2(y_1,y_2)=\mathrm{\Pi }(y_1)\mathrm{\Pi }(y_2)\mathrm{\Pi }(y_1)\mathrm{\Pi }(y_2).$$ (3) In particular, for $`y_1=y_2=y`$ we define $`D^2(y)=D^2(y,y)`$. This quantity measures the variance of the transverse momentum transfer distribution across the rapidity “boundary” $`y`$. $`D^2(y)`$ provides a measure of the rapidity structure of the correlations between the transverse momenta as can be seen from the relation $$D^2(y)=_{\mathrm{}}^y𝑑y_1_y^{\mathrm{}}𝑑y_2d^2k_1d^2k_2\stackrel{}{k}_1\stackrel{}{k}_2\rho _2(\stackrel{}{k}_1,y_1;\stackrel{}{k}_2,y_2),$$ (4) where $`\rho _2(\stackrel{}{k}_1,y_1;\stackrel{}{k}_2,y_2)`$ is the inclusive two-particle density. With reference to the parton-level diagram of Fig. 1, it should be noted that $`\mathrm{\Pi }(y)`$ is in that case equal to the (propagator) transverse momentum exchanged between the vertices $`i`$ and $`i+1`$ if the position $`y`$ is located between particles $`i`$ and $`i+1`$ on the rapidity axis. Transverse momentum transfer correlations are, therefore, a direct measure of the correlations between the exchanges. In models of uncorrelated production, the transverse momentum of a given particle can be compensated by that of any other particle or group of particles. Consequently, the correlation length describing the transverse momentum correlations, which follow from momentum conservation alone, is of the order of half the available rapidity range and should thus increase with energy or $`W`$ in DIS. For uncorrelated transverse momenta $`D^2(y)`$ is expected to increase when $`y`$ moves from the edge of the rapidity space to $`y=0`$ since, for independent random variables, the variance of a sum increases proportionally to the number of terms included. For the same reason also $`D^2(0)`$ should increase with the number of particles produced. Going beyond the uncorrelated production assumption, one could assume that the relevant part of the diagram in Fig. 1 (neglecting any dependence on longitudinal momenta) can be schematically written as a product of nearest-neighbour correlated “links” $$\|T_n\|^2\underset{i}{}V(\stackrel{}{\mathrm{\Pi }}_i,\stackrel{}{\mathrm{\Pi }}_{i+1}),$$ (5) where the functions $`V`$ represent the (iterated) kernel of the diagram. In that case, an important role is played by the eigenvalues and eigenfunctions of the kernel, considered as a transfer integral operator . For a sufficiently long ladder, it can be shown that the function $`D^2(y_1,y_2)`$ is controlled by the largest eigenvalue $`\lambda _{01}`$ leading to a behaviour $$D^2(y_1,y_2)\mathrm{exp}\left\{\rho (1\lambda _{01})\|y_1y_2\|\right\},$$ (6) with $`\rho `$ the mean parton multiplicity per unit of rapidity. In this short-range model, the correlation length is therefore given by $`1/\rho (1\lambda _{01})`$. In models with short-range correlations, by definition, particles far away in rapidity cannot be correlated. As a result, the compensation of transverse momenta is local, i.e. it is compensated by its neighbours. The corresponding correlation length $`\lambda `$ is energy independent. These general properties are reflected in the behaviour of $`D^2(y)`$. Only particles in the rapidity interval $`\mathrm{\Delta }=(\lambda +y,y+\lambda )`$ contribute to the transverse momentum transfer at $`y`$. Therefore, $`D^2(y)`$ is expected not to increase faster with energy than the particle density in the interval $`\mathrm{\Delta }`$ (for constant $`\lambda `$) which is known to increase only logarithmically. In a $`k_T`$-ordered cascade, correlations will be long-range. The correlation length can be expected to increase with the ‘length of the cascade’ in rapidity space, i.e. proportional to $`\mathrm{ln}1/x`$ or $`\mathrm{ln}W`$ (at fixed $`Q^2`$) in DIS. For the case of the BFKL scenario, it can further be expected that the correlation length is related to the dominant eigenvalues of the BFKL kernel (cfr. Eq. 6) ## 3 A Monte Carlo study Predictions for the cases of the ordered and unordered cascades are obtained from DIS Monte Carlo models, which incorporate the QCD evolution in different approximations and utilize phenomenological models for the non-perturbative hadronization phase. The MEPS model (Matrix Element plus Parton Shower) , incorporates the QCD matrix elements up to first order, with additional soft emissions generated by adding leading log parton showers. In the colour dipole model (CDM) radiation stems from colour dipoles formed by the colour charges. Both programs use the Lund string model for hadronization. The CDM description of gluon emission is similar to that of the BFKL evolution, because the gluons emitted by the dipoles do not obey strong $`k_T`$ordering . In MEPS the partons are strongly ordered in $`k_T`$, because they are based upon leading log DGLAP parton showers. Two samples of one million events have been generated using the latest versions of the models (Lepto 6.5 for MEPS, Ariadne 4.08 for CDM) with the parton density parametrisation GRV-94 . The fraction of events undergoing a Soft Colour Interaction was set to zero. In all figures the transverse momentum transfer was calculated using final state hadrons, including neutrals so that transverse momentum is conserved for the full event. Figure 3 shows the variance of the transverse momentum transfer, $`D^2(y)`$, scaled to the average hadron transverse momentum squared in different bins of $`Q^2`$ and $`x`$. Large differences between models are observed at low $`x`$. Because the variance is scaled, this effect is not due to the larger $`p_T`$-flow produced by Ariadne. As expected, the transverse momentum transfer fluctuations are larger in the unordered scenario. The correlation function, computed as the ratio $`D^2(y_1,y_2)/D(y_1)D(y_2)`$, is shown in Fig. 2. The correlation length (shown in Fig. 4) is calculated along slices perpendicular to the diagonal $`y_1=y_2`$. Within these slices the correlation decreases exponentially with $`\|y_1y_2\|`$ and the correlation length is defined as the slope of an expontial fit. The models used in this simulation show no strong dependence of the correlation length on $`x`$ or $`Q^2`$. Again, differences are observed in the behaviour of the correlation length between the Lepto and Ariadne Monte Carlo models. In particular, Ariadne predicts a symmetric behaviour in the proton and photon fragmentation hemisphere with a double-peaked structure for the correlation length, while Lepto predicts longer correlation lengths in the photon fragmentation hemisphere. ## 4 Conclusion It has been demonstrated that the transverse momentum transfer correlation is a theoretically attractive variable to discriminate between models for QCD evolution in parton cascades. The two considered models, Lepton and Ariadne, show large differences in the behaviour of the transverse momentum transfer fluctuation and correlation length. However, further work is needed to investigate the feasibility of an experimental measurement. ## References
no-problem/9907/hep-ph9907485.html	ar5iv	text	# References Four-neutrino mixing C. Giunti INFN, Sezione di Torino, and Dipartimento di Fisica Teorica, Università di Torino, Via P. Giuria 1, I–10125 Torino, Italy, and School of Physics, Korea Institute for Advanced Study, Seoul 130-012, Korea ## Abstract It is shown that only two four-neutrino schemes are compatible with the results of all neutrino oscillation experiments. These two schemes have a mass spectrum composed of two pairs of neutrinos with close masses separated by the “LSND gap” of the order of 1 eV. Talk presented at the X<sup>th</sup> International School “PARTICLES and COSMOLOGY”, 19–25 April 1999, Baksan Valley, Kabardino-Balkaria, Russia, KIAS-P99067, hep-ph/9907485. The investigation of neutrino properties is considered today as one of the most exciting fields of research in high-energy physics. It has received a stupendous boost from the evidence in favor of neutrino oscillations discovered recently in the Super-Kamiokande experiment . The main Super-Kamiokande evidence in favor of neutrino oscillations consists in the observation of an up–down asymmetry of high-energy $`\mu `$-like events generated by atmospheric neutrinos: $$𝒜_\mu (D_\mu U_\mu )/(D_\mu +U_\mu )=0.311\pm 0.043\pm 0.01\text{[2]}.$$ (1) Here $`D_\mu `$ and $`U_\mu `$ are, respectively, the number of downward-going and upward-going events, corresponding to the zenith angle intervals $`0.2<\mathrm{cos}\theta <1`$ and $`1<\mathrm{cos}\theta <0.2`$. Since the fluxes of high-energy downward-going and upward-going atmospheric neutrinos are predicted to be equal with high accuracy on the basis of geometrical arguments (see ), the Super-Kamiokande evidence in favor of neutrino oscillations is model-independent and provides a definite confirmation of the indications in favor of oscillations of atmospheric neutrinos found in the Super-Kamiokande experiment itself and in other experiments through the measurement of the ratio of $`\mu `$-like and $`e`$-like events (Kamiokande, IMB, Soudan 2 ) and through the measurement of upward-going muons produced by neutrino interactions in the rock below the detector (MACRO ). Large $`\nu _\mu \nu _e`$ oscillations of atmospheric neutrinos are excluded by the absence of a up–down asymmetry of high-energy $`e`$-like events generated by atmospheric neutrinos and detected in the Super-Kamiokande experiment ($`𝒜_e=0.036\pm 0.067\pm 0.02`$ ) and by the negative result of the CHOOZ long-baseline $`\overline{\nu }_e`$ disappearance experiment . Therefore, the atmospheric neutrino anomaly consists in the disappearance of muon neutrinos and can be explained by $`\nu _\mu \nu _\tau `$ and/or $`\nu _\mu \nu _s`$ oscillations (here $`\nu _s`$ is a sterile neutrino that does not take part in weak interactions). 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 (see and references therein). The disappearance of solar electron neutrinos can be explained by $`\nu _e\nu _\mu `$ and/or $`\nu _e\nu _\tau `$ and/or $`\nu _e\nu _s`$ oscillations (see the recent analyses in ). The accelerator LSND experiment is the only one that claims the observation of neutrino oscillations in specific appearance channels: $`\overline{\nu }_\mu \overline{\nu }_e`$ and $`\nu _\mu \nu _e`$. Since the appearance of neutrinos with a different flavor represents the true essence of neutrino oscillations, the LSND evidence is extremely interesting and its confirmation (or disproof) by other experiments should receive high priority in future research. Four such experiments have been proposed and are under study: BooNE at Fermilab, I-216 at CERN, ORLaND at Oak Ridge and NESS at the European Spallation Source . Among these proposals only BooNE is approved and will start in 2001. Neutrino oscillations occur if neutrinos are massive and mixed particles (see ), i.e. if the left-handed components $`\nu _{\alpha L}`$ of the flavor neutrino fields are superpositions of the left-handed components $`\nu _{kL}`$ ($`k=1,\mathrm{},N`$) of neutrino fields with definite mass $`m_k`$: $$\nu _{\alpha L}=\underset{k=1}{\overset{N}{}}U_{\alpha k}\nu _{kL},$$ (2) where $`U`$ is a $`N\times N`$ unitary mixing matrix. From the measurement of the invisible decay width of the $`Z`$-boson (see ) it is known hat the number of light active neutrino flavors is three, corresponding to $`\nu _e`$, $`\nu _\mu `$ and $`\nu _\tau `$ (active neutrinos are those taking part to standard weak interactions). This implies that the number $`N`$ of massive neutrinos is bigger or equal to three. If $`N>3`$, in the flavor basis there are $`N_s=N3`$ sterile neutrinos, $`\nu _{s_1}`$, …, $`\nu _{s_{N_s}}`$, that do not take part to standard weak interactions. In this case the index $`\alpha `$ in Eq. (2) takes the values $`e,\mu ,\tau ,s_1,\mathrm{},s_{N_s}`$. The three evidences in favor of neutrino oscillations found in solar and atmospheric neutrino experiments and in the accelerator LSND experiment imply the existence of at least three independent neutrino mass-squared differences. This can be seen by considering the general expression for the probability of $`\nu _\alpha \nu _\beta `$ transitions in vacuum, that can be written as (see ) $$P_{\nu _\alpha \nu _\beta }=\left\|\underset{k=1}{\overset{N}{}}U_{\alpha k}^{}U_{\beta k}\mathrm{exp}\left(i\frac{\mathrm{\Delta }m_{kj}^2L}{2E}\right)\right\|^2,$$ (3) where $`\mathrm{\Delta }m_{kj}^2m_k^2m_j^2`$, $`j`$ is any of the mass-eigenstate indices, $`L`$ is the distance between the neutrino source and detector and $`E`$ is the neutrino energy. The range of $`L/E`$ characteristic of 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. From Eq. (3) it is clear that neutrino oscillations are observable in an experiment only if there is at least one mass-squared difference $`\mathrm{\Delta }m_{kj}^2`$ such that $$\frac{\mathrm{\Delta }m_{kj}^2L}{2E}0.1$$ (4) (the precise lower bound depends on the sensitivity of the experiment) in a significant part of the energy and source-detector distance intervals of the experiment (if the condition (4) is not satisfied, $`P_{\nu _\alpha \nu _\beta }\left\|_kU_{\alpha k}^{}U_{\beta k}\right\|^2=\delta _{\alpha \beta }`$). Since the range of $`L/E`$ probed by the LSND experiment is the smaller one, a large mass-squared difference is needed for LSND oscillations: $$\mathrm{\Delta }m_{\mathrm{LSND}}^210^1\mathrm{eV}^2.$$ (5) Furthermore, from Eq. (3) it is clear that a dependence of the oscillation probability from the neutrino energy $`E`$ and the source-detector distance $`L`$ is observable only if there is at least one mass-squared difference $`\mathrm{\Delta }m_{kj}^2`$ such that $$\frac{\mathrm{\Delta }m_{kj}^2L}{2E}1.$$ (6) Indeed, all the phases $`\mathrm{\Delta }m_{kj}^2L/2E1`$ are washed out by the average over the energy and source-detector ranges characteristic of the experiment. Since a variation of the oscillation probability as a function of neutrino energy has been observed both in solar and atmospheric neutrino experiments and the ranges of $`L/E`$ characteristic of these two types of experiments are different from each other and different from the LSND range, two more mass-squared differences with different scales are needed: $`\mathrm{\Delta }m_{\mathrm{sun}}^210^{10}\mathrm{eV}^2\text{(VO)},`$ (7) $`\mathrm{\Delta }m_{\mathrm{atm}}^210^310^2\mathrm{eV}^2.`$ (8) The condition (7) for the solar mass-squared difference $`\mathrm{\Delta }m_{\mathrm{sun}}^2`$ has been obtained under the assumption of vacuum oscillations (VO). If the disappearance of solar $`\nu _e`$’s is due to the MSW effect , the condition $$\mathrm{\Delta }m_{\mathrm{sun}}^210^4\mathrm{eV}^2\text{(MSW)}$$ (9) must be fulfilled in order to have a resonance in the interior of the sun. Hence, in the MSW case $`\mathrm{\Delta }m_{\mathrm{sun}}^2`$ must be at least one order of magnitude smaller than $`\mathrm{\Delta }m_{\mathrm{atm}}^2`$. It is possible to ask if three different scales of neutrino mass-squared differences are needed even if the results of the Homestake solar neutrino experiment is neglected, allowing an energy-independent suppression of the solar $`\nu _e`$ flux. The answer is that still the data cannot be fitted with only two neutrino mass-squared differences because an energy-independent suppression of the solar $`\nu _e`$ flux requires large $`\nu _e\nu _\mu `$ or $`\nu _e\nu _\tau `$ transitions generated by $`\mathrm{\Delta }m_{\mathrm{atm}}^2`$ or $`\mathrm{\Delta }m_{\mathrm{LSND}}^2`$. These transitions are forbidden by the results of the Bugey and CHOOZ reactor $`\overline{\nu }_e`$ disappearance experiments and by the non-observation of an up-down asymmetry of $`e`$-like events in the Super-Kamiokande atmospheric neutrino experiment . The existence of three different scales of $`\mathrm{\Delta }m^2`$ imply that at least four light massive neutrinos must exist in nature. Here we consider the schemes with four light and mixed neutrinos , which constitute the minimal possibility that allows to explain all the existing data with neutrino oscillations. In this case, in the flavor basis the three active 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 with three different scales of $`\mathrm{\Delta }m^2`$ 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. The six schemes are divided into four schemes of class 1 (I–IV) in which there is a group of three masses separated from an isolated mass by the LSND gap, and two schemes of class 2 (A, B) in which there are two couples of close masses separated by the LSND gap. In the following we will show that the schemes of class 1 (I–IV) are disfavored by the data if also the negative results of short-baseline accelerator and reactor disappearance neutrino oscillation experiments are taken into account . Let us remark that in principle one could check which schemes are allowed by doing a combined fit of all data in the framework of the most general four-neutrino mixing scheme, with three mass-squared differences, six mixing angles and three CP-violating phases as free parameters. However, at the moment it is not possible to perform such a fit because of the enormous complications due to the presence of too many parameters 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 of class 1 with the following relatively simple procedure. Let us define the quantities $`d_\alpha `$, with $`\alpha =e,\mu ,\tau ,s`$, in the schemes of class 1 as $$d_\alpha ^{(\mathrm{I})}\|U_{\alpha 4}\|^2,d_\alpha ^{(\mathrm{II})}\|U_{\alpha 4}\|^2,d_\alpha ^{(\mathrm{III})}\|U_{\alpha 1}\|^2,d_\alpha ^{(\mathrm{IV})}\|U_{\alpha 1}\|^2.$$ (10) Physically $`d_\alpha `$ quantifies the mixing of the flavor neutrino $`\nu _\alpha `$ with the isolated neutrino, whose mass is separated from the other three by the LSND gap. The probability of $`\nu _\alpha \nu _\beta `$ ($`\beta \alpha `$) and $`\nu _\alpha \nu _\alpha `$ transitions (and the corresponding probabilities for antineutrinos) in short-baseline experiments are given by $$P_{\nu _\alpha \nu _\beta }=A_{\alpha ;\beta }\mathrm{sin}^2\frac{\mathrm{\Delta }m_{41}^2L}{4E},P_{\nu _\alpha \nu _\alpha }=1B_{\alpha ;\alpha }\mathrm{sin}^2\frac{\mathrm{\Delta }m_{41}^2L}{4E},$$ (11) with the oscillation amplitudes $$A_{\alpha ;\beta }=4d_\alpha d_\beta ,B_{\alpha ;\alpha }=4d_\alpha (1d_\alpha ).$$ (12) The probabilities (11) have the same form as the corresponding probabilities in the case of two-neutrino mixing, $`P_{\nu _\alpha \nu _\beta }=\mathrm{sin}^2(2\vartheta )\mathrm{sin}^2(\mathrm{\Delta }m^2L/4E)`$ and $`P_{\nu _\alpha \nu _\alpha }=1\mathrm{sin}^2(2\vartheta )\mathrm{sin}^2(\mathrm{\Delta }m^2L/4E)`$, which have been used by all experimental collaborations for the analysis of the data in order to get information on the parameters $`\mathrm{sin}^2(2\vartheta )`$ and $`\mathrm{\Delta }m^2`$ ($`\vartheta `$ and $`\mathrm{\Delta }m^2`$ are, respectively, the mixing angle and the mass-squared difference in the case of two-neutrino mixing). Therefore, we can use the results of their analyses in order to get information on the corresponding parameters $`A_{\alpha ;\beta }`$, $`B_{\alpha ;\alpha }`$ and $`\mathrm{\Delta }m_{41}^2`$. The exclusion plots obtained in short-baseline $`\overline{\nu }_e`$ and $`\nu _\mu `$ disappearance experiments imply that $$d_\alpha a_\alpha ^0\text{or}d_\alpha 1a_\alpha ^0(\alpha =e,\mu ),$$ (13) with $$a_\alpha ^0=\frac{1}{2}\left(1\sqrt{1B_{\alpha ;\alpha }^0}\right)(\alpha =e,\mu ),$$ (14) where $`B_{e;e}^0`$ and $`B_{\mu ;\mu }^0`$ are the upper bounds, that depend on $`\mathrm{\Delta }m_{41}^2`$, of the oscillation amplitudes $`B_{e;e}`$ and $`B_{\mu ;\mu }`$ given by the exclusion plots of $`\overline{\nu }_e`$ and $`\nu _\mu `$ disappearance experiments. From the exclusion curves of the Bugey reactor $`\overline{\nu }_e`$ disappearance experiment and of the CDHS and CCFR accelerator $`\nu _\mu `$ disappearance experiments it follows that $`a_e^04\times 10^2`$ for $`\mathrm{\Delta }m_{41}^20.1\mathrm{eV}^2`$ and $`a_\mu ^00.2`$ for $`\mathrm{\Delta }m_{41}^20.4\mathrm{eV}^2`$ (see ). Therefore, the negative results of short-baseline $`\overline{\nu }_e`$ and $`\nu _\mu `$ disappearance experiments imply that $`d_e`$ and $`d_\mu `$ are either small or large (close to one). However, since the survival probability of solar $`\nu _e`$’s is bounded by $$P_{\nu _e\nu _e}^{\mathrm{sun}}d_e^2,$$ (15) only the possibility $$d_ea_e^0$$ (16) is acceptable in order to explain the observed deficit of solar $`\nu _e`$’s with neutrino oscillations. In a similar way, since the survival probability of atmospheric $`\nu _\mu `$’s and $`\overline{\nu }_\mu `$’s is bounded by $$P_{\nu _\mu \nu _\mu }^{\mathrm{atm}}d_\mu ^2,$$ (17) it is clear that large values of $`d_\mu `$ are incompatible with the asymmetry (1) observed in the Super-Kamiokande experiment. Indeed, it has been shown in that the Super-Kamiokande asymmetry (1) and the exclusion curve of the Bugey $`\overline{\nu }_e`$ disappearance experiment imply the upper bound $$d_\mu 0.55a_\mu ^{\mathrm{SK}}.$$ (18) This upper bound is depicted by the horizontal line in Fig. 1 (the vertically hatched area above the line is excluded). In Fig. 1 we have also shown the bound $`d_\mu a_\mu ^0`$ or $`d_\mu 1a_\mu ^0`$ obtained from the exclusion plot of the short-baseline CDHS $`\nu _\mu `$ disappearance experiment, which exclude the shadowed region. It is clear that the results of short-baseline disappearance experiments and the Super-Kamiokande asymmetry (1) imply that $`d_\mu 0.55`$ for $`\mathrm{\Delta }m_{41}^20.3\mathrm{eV}^2`$ and that $`d_\mu `$ is very small for $`\mathrm{\Delta }m_{41}^20.3\mathrm{eV}^2`$. However, this range of $`d_\mu `$ is disfavored by the results of the LSND experiment, that imply a lower bound $`A_{\mu ;e}^{\mathrm{min}}`$ for the amplitude $`A_{\mu ;e}=4d_ed_\mu `$ of $`\nu _\mu \nu _e`$ oscillations. Using also the bound (16), we obtain the constraint $$d_\mu A_{\mu ;e}^{\mathrm{min}}/4a_e^0.$$ (19) This bound is represented by the curve in Fig. 1 labelled LSND + Bugey (the diagonally hatched area is excluded) and one can see that it excludes the range of $`d_\mu `$ allowed by the results of short-baseline disappearance experiments and by the Super-Kamiokande asymmetry (1). From Fig. 1 one can see that in the framework of the schemes of class 1 there is no range of $`d_\mu `$ that is compatible with all the experimental data. Hence, the four-neutrino schemes of class 1 are disfavored by the data. The incompatibility of the experimental results with the schemes of class 1 is shown also in Fig. 1, 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). On the other hand, the four-neutrino schemes of class 2 (A, B) are compatible with the results of all neutrino oscillation experiments if the mixing of $`\nu _e`$ with the two mass eigenstates responsible for the oscillations of solar neutrinos ($`\nu _3`$ and $`\nu _4`$ in scheme A and $`\nu _1`$ and $`\nu _2`$ in scheme B) is large and the mixing of $`\nu _\mu `$ with the two mass eigenstates responsible for the oscillations of atmospheric neutrinos ($`\nu _1`$ and $`\nu _2`$ in scheme A and $`\nu _3`$ and $`\nu _4`$ in scheme B) is large . This fact implies that $`\nu _e`$’s do not oscillate in atmospheric and long-baseline neutrino oscillation experiments and one can obtain rather stringent upper bounds for the probability of $`\nu _e`$ transitions into any other state and for the size of CP or T violation that could be measured in long-baseline experiments in the $`\nu _\mu \nu _e`$ and $`\overline{\nu }_\mu \overline{\nu }_e`$ channels . Furthermore, it has been shown in that the upper bound $`N_\nu ^{\mathrm{BBN}}<4`$ for the effective number of neutrinos in Big-Bang Nucleosynthesis implies that the mixing of $`\nu _s`$ with the two mass eigenstates responsible for the oscillations of atmospheric neutrinos is very small. In this case atmospheric neutrinos oscillate only in the $`\nu _\mu \nu _\tau `$ channel and solar neutrino oscillate only in the $`\nu _e\nu _s`$ channel. This is very important because it implies that the two-generation analyses of solar and atmospheric neutrino data give correct information on neutrino mixing in the two four-neutrino schemes A and B. I would like to thank S.M. Bilenky, W. Grimus and C.W. Kim for friendship and stimulating collaboration. I would also like to express my gratitude to the Korea Institute for Advanced Study (KIAS) for kind hospitality during the writing of this report.
no-problem/9907/astro-ph9907296.html	ar5iv	text	# A new bursting X-ray transient: SAX J1750.8-2900 ## 1 Introduction A long term program to survey the 40$`\times `$40 degrees around the Galactic Centre started on mid 1996 with the large field of view instruments on board the BeppoSAX satellite (Wide Field Cameras, hereafter WFC). Previous surveys of the region with similar instruments were limited by the lack of the combination of sufficiently long, repeated exposures and the wide angular coverage. In the last 10 years, however, the use of the coded mask imaging technique increased the total number of known X-ray emitters in the region (Skinner et al. (1993); Vargas et al. (1997)) stimulating detailed measurements of individual sources and in turn their identification at different wavelengths. The Galactic Bulge monitoring program carried out by BeppoSAX WFC in the energy range 2-30 keV has been especially prolific in the study of X-ray burst sources, increasing substantially (by about 50% in 2.5 years) the number of objects of this type which were known originally in this region. As of January 1999 it led to the discovery of 6 new burst sources and, in addition, found burst emission from 7 already known sources (Heise et al. (1999); Ubertini et al. 1999a ; Cocchi et al. 1998a for earlier results) in a total time exposure of $``$2.5$`\times `$10<sup>6</sup>s. The new transient sources show dim X-ray outburst episodes during $``$1 to a few weeks, with peak fluxes generally below a few 10<sup>37</sup> erg/s at 10 kpc distance (Heise et al. (1999)). From one of these sources, SAX J1808.4-3658 (in ’t Zand et al. (1998)) a modulation period of 2.5 ms was discovered by RXTE during a second outburst (Wijnands & van der Klis (1998), Chakrabarty & Morgan (1998)). Here we report results of one of these previously unknown transients showing bursting behaviour, discovered by the WFC on March 18th, 1997 (Bazzano et al. 1997a , Heise et al. (1997)) in a celestial position 1.2 degrees off the Galactic Center. In particular, we analyse the spectral and temporal behaviour of the persistent emission and characterise the burst emission properties to determine the nature of the transient. ## 2 Observations and data analysis The Wide Field Cameras experiment on board the BeppoSAX satellite comprises 2 identical coded aperture multi-wire proportional counter detectors viewing opposite sky directions (Jager et al. (1997)), each one featuring a field of view of 40$`\times `$40 degrees full width to zero response (i.e., 3.7$`\%`$ of the sky) and an angular resolution of 5 arcmin. The source location accuracy depends on the signal-to noise ratio and is 0.7 arcmin at best (99$`\%`$ confidence level). The energy range is 2-30 keV on-axis and the time resolution is 0.5 ms. The field of view (FOV) is the largest of any flown X-ray imaging device with arcmin resolution, which allows for the search of short duration and/or weak transient events. The on-axis sensitivity for the Galactic Bulge field is $``$10 mCrab in 10<sup>4</sup>s observing time. Detector data contain a superposition of background and of multiple source shadowgrams, the latter resulting from the coding of the sky object image with the instrument aperture pattern. The reconstruction of the sky image for point-like sources involves an algorithm that consist of a cross correlation of the detector data with the aperture (see e.g. Caroli et al. (1987)). The position and intensity of any point source is determined by folding a sky model distribution through a point spread function (PSF), using iterative $`\chi `$ <sup>2</sup> minimisation (Jager et al. (1997)). For WFC this can be carried out in each individual energy channel. The full-width at half maximum of the PSF is smallest on axis at $``$5 arcmin. SAX J1750.8-2900 is located only 1.2 degrees off the Galactic Centre and so in the most sensitive region for this type of observations. The Galactic Bulge was observed during spring 1997 for 5 $`\times `$10<sup>5</sup> s, spread out along 28 days. Burst phenomena are systematically searched in data from both cameras using time profiles of the total detector over the entire energy range with a time resolution of 1 s. When a burst occurs a reconstructed sky image is generated for the burst duration and different sky images corresponding to longer time exposure are generated for intervals just before and after burst. This allows to resolve the point source responsible for the intensity increase revealed in detector ratemeters. In crowded fields and in some not evident case an image subtraction is necessary to facilitate identification of bursting sources in the FOV. ## 3 Transient source position and lightcurve Figure 1 shows the error region for SAX J1750.8-2900. The best fit position is R.A. = 17h 50m 24s, Dec = -29 02’ 18” (equinox 2000.0), with an error radius of 1 arcmin (99% confidence). This is a position refined from the previously published value which resulted from a quick-look analysis (Heise et al. (1997)). The deviation between both values is 0.4 arcmin. Also shown are the positions of two X-ray bursts that were observed simultaneously with the active phase of SAX J1750.8-2900, showing that they result from a position coincident with the transient. In March 1992 the ROSAT PSPC observed the region around SAX J1750.8-2900 four times (between MJD 48685.09 and MJD 48691.63 with exposure times up to 1976 s), during a raster scan of the Galactic Centre region. No source was detected during these observations within the 99% confidence error box of SAX J1750.8-2900. The source 1RXP J175029-2859.9 lies 1.5 arcmin outside the SAX J1750.8-2900 error box. This close-by source was marginally detected at $`6.2\pm 1.9`$ $`\times 10^3`$ counts s<sup>-1</sup> in the second observation (between MJD 48685.36 and MJD 48685.39). From this result we can derive an upper limit of $``$ 3 $`\times 10^{12}`$ erg cm<sup>-2</sup> s<sup>-1</sup> on the soft X-ray emission (0.5-2.0 keV) of SAX J1750.8-2900 during quiescence. The RossiXTE All Sky Monitor (Levine et al. (1996)) data retrieved from a public dataset (provided by the RXTE/ASM team, http://space.mit.edu/XTE) shows the onset of a fast rise, exponential decay outburst of SAX J1750.8-2900 peaking at $``$ 120$`\pm `$40 mCrab in the 2-10 keV range, starting close to MJD 50518 (two days before the initial WFC observation) and lasting $``$ 2 weeks. This transient behaviour is supported by WFC later detections as previously reported by in ’t Zand et al. (1997). The X-ray persistent emission of SAX J1750.8-2900 (i.e., the emission detected during time intervals excluding bursts) was measured by the WFC starting from March 13th, 1997. The source flux was initially at a level of $``$ 70 mCrab in the energy band 2-30 keV. The light curve detected in the 2-30 keV range is shown in Fig. 2, along with the burst occurrence time. The outburst profile decay is close to exponential, with intensity changes on the time scale of hours. The first two days measurements (when the source was more luminous) are affected by relatively large errors caused by the off-axis position of the source. The average luminosity in the 2-30 keV band, calculated for 10 kpc distance is $``$3$`\times `$10<sup>37</sup> erg s<sup>-1</sup> on March 13th. The flux was about $``$ 1.5 times weaker five days later, when the source was seen bursting for the first time, and dropped to less than $``$ 3 mCrab ($``$ 10<sup>36</sup> erg s<sup>-1</sup> at 10 kpc) on March 25th. The source was again visible at $``$ 10 mCrab on March 30-31, when two more bursts were detected. We fitted the emission spectra detected in the 2-30 keV band during six observing periods between MJD 50520 and MJD 50531 using a few spectral models. The results obtained for power law and thermal bremssthralung (both with low energy absorption) are shown in Table 1. The spectra can be described either by a power law shape having a photon index $`\mathrm{\Gamma }`$ $``$ 2.5 and extinction parameter $`N_H`$ $``$ 6 $`\times `$10<sup>22</sup> cm<sup>-2</sup>, or by bremssthralung emission with kT in the 3 to 10 keV range and $`N_H`$ $``$ 2.5 $`\times `$10<sup>22</sup> cm<sup>-2</sup>. The spectra cannot be fitted satisfactorily with single component blackbody emission. The fit results give indication that the source has experienced spectral softening during the outburst decay. By performing an F-test on the two spectra taken at MJD 50520 and MJD 50527 we find that the probability that there is no softening is less than 1%. If the X-ray emission mechanism is thermal (as observed in many X-ray bursters) the softening could be ascribed to a temperature variation of the electron plasma, possibly due to the decrease in the accretion flow. ## 4 The X-ray bursts A total of 9 X-ray bursts were detected from SAX J1750.8-2900 during an overall time span of 14 days in spring 1997. The first one (the faintest observed) occurred on MJD 50525.48150, with a peak flux of $``$0.4 Crab. 7 out of 9 bursts were detected during three days from March 18th (see Table 2 for burst occurrence times), having similar bolometric fluences in the range $``$2 to 3$`\times `$10<sup>-7</sup> erg cm<sup>-2</sup>. In Fig.3 burst profiles in two energy bands are plotted for two of these events. The study of the burst frequency is limited by the fact that during observation the effective exposure time is only a fraction ($``$60$`\%`$) of the total pointing time, due to earth occultations and other shorter non-coverage periods. The observed values of time intervals are in fact an upper limit to the real burst interval time. It is then possible that SAX J1750.8-2900 made bursts during the first observation period when its persistent flux was above $``$50 mCrab. In spite of this, there is evidence that the burst frequency decreased when the source flux dropped below $``$20 mCrab, i.e. in observations performed after MJD 50530 (see Fig.2). The primary question concerning the bursts is whether they are type I X-ray bursts. All burst profiles detected from SAX J1750.8-2900 show a fast rise ($``$ 1 s), exponential decay shape, with e-folding time in the range $``$ 5-10 s (see Fig.3). The decay times in the energy band 8-26 keV are systematically shorter than those observed in the band 2-8 keV. However, the spectral softening cannot be proven by examining the individual bursts, due to the large statistical error (see Table 2). In order to increase significance we summed up the profiles of the last 7 bursts in the energy bands 2-8 and 8-26 keV, with a time resolution of 0.1 s. The first two bursts were excluded because their detection was affected by the earth atmosphere. The start channel of each burst was determined as the first point in the time profile which differed more than $``$4$`\sigma `$ from the mean persistent emission. The fit of the two profiles obtained with an exponential function gives an e-folding decay time $`\tau `$ = 5.3$`\pm `$0.7 s in the low energy band and $`\tau `$ = 2.8$`\pm `$0.2 s in the high energy band, and proves that spectral softening is occurring during burst decay. Together with a consistency of the burst spectrum with that of a few keV blackbody emission (see Table 2) this identifies the bursts as type-I. Among the bursts detected, there is no clear evidence of an X-ray burst with double peaked or flat profile, which might have suggested saturation of the luminosity to near-Eddington level and resulting photospheric radius expansion (Lewin, van Paradijs, & Taam (1995)). However we can derive an upper limit on the source distance assuming that the maximum burst luminosity was below Eddington. The maximum observed peak flux (burst F, see Table 2) is consistent with a 3 $`\sigma `$ upper limit of $``$ 7 kpc. Burst spectra are rather soft and generally compatible with blackbody emission having colour temperatures between 2 and 3 keV (see Table 2). Under given assumptions (Lewin, van Paradijs, & Taam (1993)) the effective temperature $`T_{eff}`$ and the bolometric flux of a burst can determine the ratio between the blackbody radius $`R_{bb}`$ (that is, the radius of the emitting sphere) and the distance d of the neutron star. Assuming d=10 kpc and the observed colour temperatures as $`T_{eff}`$, and not correcting for gravitational redshift the measured blackbody radius is $``$ 8 km. For the above upper limit of 7 kpc, this value of $`R_{bb}`$ scales to a corresponding upper limit of $``$ 6 km. This value could be underestimated, due to the uncertainties in the relationship between colour and effective temperature. If, as suggested by Ebisuzaki (1987) the colour temperature exceeds $`T_{eff}`$ by a factor $``$1.5, then the neutron star radius should be at least two times the measured blackbody radius. These values are therefore consistent with a neutron star nature of the compact object. ## 5 Discussion ### 5.1 Burst emission properties In the simplest interpretation of the thermonuclear flash model which successfully explains type-I X-ray bursts (Lewin, van Paradijs, & Taam (1995) for review) the matter accreted onto a neutron star surface prior to an observed type-I burst is converted into nuclear fuel and the fraction of the total accreted energy available for burning depends on the actual reaction process and fuel composition. If the thermonuclear flash is isotropic and the accreted material is totally converted into fuel, the ratio between the mass and radius of the NS is given by $`M_{}`$/$`R_{10km}`$ = (0.01-0.04)\*$`\alpha `$, where 0.01 and 0.04 hold for helium and hydrogen burning respectively. Here $`M_{}`$ is the mass of the compact object in units of solar masses, $`R_{10km}`$ is the NS radius in units of 10 km and $`\alpha `$ is the ratio between the bolometric flux of the persistent emission (integrated over the burst interval) and the bolometric fluence of the burst. A 1.4 $`M_{}`$ neutron star would then result in a value of $`\alpha `$ $``$ 100 for pure helium burning. For the four shortest observed burst times intervals B-C,C-D,D-E and F-G (see Table 2) we estimated the $`\alpha `$ parameter and found values of 85$`\pm `$20, 120$`\pm `$30, 170$`\pm `$30 and 210$`\pm `$40 respectively. These intervals are monotonically increasing from 4.1 to 5.9 h on a time scale of 1.5 days, and all the related bursts show similar profiles and fluences. This suggests that perhaps no bursts were missed in between. The fast rise time of the bursts ($`<`$ 2 s) and the measured values of $`\alpha `$ seem to favour a pure helium flash respect to combined hydrogen-helium shell burning. (Lewin, van Paradijs, & Taam (1993)). ### 5.2 SAX J1750.8-2900 and the transients of the Galactic Bulge Most X-ray burst sources known so far are type-I bursters associated with low-mass X-ray binaries (LMXBs) containing old, weakly magnetized neutron stars, and show concentration in the direction of the Galactic Centre (van Paradijs, (1995) ). They can be persistent (though variable) or transient, and may have recurrence periods with nearly constant burst activity, like the recently studied GS 1826-338 (Ubertini et al. 1999b ), or conversely show only episodic burst emission, like for example SLX 1735-26 (Bazzano et al. 1997b ) and XTE J1709-267 (Cocchi et al. 1998b ). Correlation between burst frequency and persistent emission is not an uncommon feature. In general, type-I bursts are observed when the source persistent luminosity is comprised between $``$10<sup>-2</sup> and $``$0.3 of the Eddington limit. For the bursting soft X-ray transients (White, Kaluziensky, & Swank (1984); see Campana et al. (1998) for recent review) the burst activity is usually detected during the occurrence of outburst episodes, which show peak luminosity of up to $``$10<sup>38</sup> erg s<sup>-1</sup> and often recur on time scales of $``$ 1 to $``$ 10 years. SAX J1750.8-2900 shows this type of transient phenomenology. The outburst light curve has a rather clear fast rise and exponential decay shape. In spite of the incomplete sampling it evidently shows variable decay behaviour, as observed in other LMXB transients (Chen, Shrader & Livio (1997)). So there is no real evidence that the source had a secondary outburst after MJD 50535 as it could appear at a first glance. The X-ray flux decreased of a factor $``$ 20 in a period of $``$ 3 weeks in spring 1997, after which the source remained undetected (the RossiXTE ASM and BeppoSAX data do not show any other evident outburst in the period 1996 to 1999 May). We provide evidence that SAX J1750.8-2900 has been observed bursting only whenever the intensity was above $``$ 10 mCrab, and that the burst frequency was positively correlated with the persistent emission (at least when the persistent flux was in the range $``$ 10 to $``$ 50 mCrab). The spectral softening seen by analysis of burst profiles is an evidence that SAX J1750.8-2900 is a type-I burster, and hence that the compact object is a neutron star. The observed peak luminosity of bursts suggests an upper limit of 7 kpc on the source distance. Due to the lack of optical identification and/or visible X-ray modulation it is not possible to classify with certainty the binary source as a low mass system. Nevertheless, the detection of type-I bursts is sufficient to firmly set SAX J1750.8-2900 as a candidate member of the LMXB class. The current sample of known LMXB could be biased towards bright X-ray transients, due to instrument selection effects and the established occurrence of weak, short lasting transients with long recurrence time (like e.g., 2S 1803-245, Muller et al. (1998); and SAX J1748.9-2021, in ’t Zand al. (1999)). In fact, the recent observations by BeppoSAX and RXTE are significantly growing the number of weak LMXB. Among them, most are NS transients which are also burst sources and often show high energy tails. For this reason, these have been suggested as a possible new subclass of low mass binaries (Heise et al. (1999)). Indeed these systems could be NS soft X-ray transients (of the type of Cen X-4 or Aql X-1), which are harboured within the Galactic Bulge at quite large distances. This is what should be expected, as increasing the sensitivity and coverage will push the limit of observable distances up to a range in which many more sources are available, due to their concentration towards the Galactic Centre. We thank the staff of the BeppoSAX Science Operation Centre and Science Data Centre for their help in carrying out and processing the WFC Galactic Centre observations. The BeppoSAX satellite is a joint Italian and Dutch program. A.B., M.C., L.N. and P.U. thank Agenzia Spaziale Nazionale (ASI) for grant support.
no-problem/9907/hep-ph9907219.html	ar5iv	text	# Diminishing “charginos nearly degenerate with the lightest neutralino” slit using precision data ## 1 Introduction LEP II is very effective in bounding from below masses of charginos which, if this is kinematically allowed, should be produced in a pair in $`e^+e^{}\chi ^+\chi ^{}`$ annihilation. The present bounds are $`m_{\stackrel{~}{\chi }^\pm }90\text{ GeV}`$ for the higgsino-dominated case and $`m_{\stackrel{~}{\chi }^\pm }70\text{ GeV}`$ for the gaugino-dominated case if sneutrino is not too heavy . However, when the lightest chargino and neutralino (the latter being the LSP) are almost degenerate in mass, the charged decay products of the light chargino are very soft, and the above quoted bounds are no longer valid. Special search for such light charginos has been performed recently by DELPHI collaboration, and the case of $`\mathrm{\Delta }M^\pm m_{\stackrel{~}{\chi }_1^\pm }m_{\stackrel{~}{\chi }_1^0}100\text{ MeV}`$ is now excluded . However, still in the case of $`\mathrm{\Delta }M^\pm 1\text{ GeV}`$ LEP II does not provide a lower bound and charginos as light as $`45\text{ GeV}`$ are allowed (this bound comes from the measurements of $`Z`$ decays at LEP I and SLC). The case of almost degenerate chargino and neutralino can be naturally realized in SUSY and the possibilities to find such particles are discussed in literature . In this letter we investigate the radiative corrections to $`m_W`$ and to $`Z`$ boson decay parameters generated by such almost degenerate particles. When their masses are close to $`m_Z/2`$ radiative corrections are large and they spoil the perfect description of experimental data by the Standard Model. Due to the decoupling property of SUSY models, when $`m_{\stackrel{~}{\chi }^{\pm ,0}}m_Z`$ the radiative corrections are power suppressed. ## 2 Discussion In the simplest supersymmetric extensions of the Standard Model the chargino-neutralino sector is defined by the numerical values of four parameters: $`M_1`$, $`M_2`$, $`\mu `$ and $`\mathrm{tan}\beta `$. The case of nearly degenerate lightest chargino and neutralino naturally arise when: 1. $`M_2\mu `$: in this case the particles of interest form an $`SU(2)`$ doublet of Dirac fermions, whose wave functions are dominated by higgsinos; 2. $`\mu M_2`$: in this way we get an $`SU(2)`$ triplet of Majorana fermions, with the wave functions dominated by winos. In this way we get higgsino- and gaugino-dominated scenarios, correspondingly. Let us start from case (1). ### 2.1 Higgsino-dominated case The degenerate Dirac doublet produces the following corrections to the three functions $`V_i`$ which determine the values of radiative corrections (for definitions of functions $`V_i`$ and further details about electroweak radiative corrections see ): $`\delta ^{\stackrel{~}{h}}V_m`$ $`={\displaystyle \frac{16}{9}}\left[\left({\displaystyle \frac{1}{2}}s^2+s^4\right)\left(1+2\chi \right)F(\chi )\left({\displaystyle \frac{1}{2}}s^2\right)\left(1+2{\displaystyle \frac{\chi }{c^2}}\right)F\left({\displaystyle \frac{\chi }{c^2}}\right){\displaystyle \frac{s^4}{3}}\right],`$ (1) $`\delta ^{\stackrel{~}{h}}V_A`$ $`={\displaystyle \frac{16}{9}}\left({\displaystyle \frac{1}{2}}s^2+s^4\right)\left[{\displaystyle \frac{12\chi ^2F(\chi )2\chi 1}{4\chi 1}}\right],`$ (2) $`\delta ^{\stackrel{~}{h}}V_R`$ $`={\displaystyle \frac{16}{9}}c^2s^2\left[\left(1+2\chi \right)F(\chi ){\displaystyle \frac{1}{3}}\right],`$ (3) where $`\chi (m_{\stackrel{~}{\chi }^{\pm ,0}}/m_Z)^2`$, the function $`F`$ is defined in Appendix B of Ref. , and $`s^2`$ ($`c^2`$) is the sine (cosine) squared of the electroweak mixing angle $`\theta `$. Comparing the experimental data with Standard Model formulas , we obtain that the $`\chi ^2`$ for the new physics contributions to $`V_i`$ can be computed in the following way: $$\chi ^2=C_{ij}\left(\delta _{\mathrm{NP}}V_i\overline{\delta V_i}\right)\left(\delta _{\mathrm{NP}}V_j\overline{\delta V_j}\right)$$ (4) $$\left(\begin{array}{ccc}C_{mm}& C_{mA}& C_{mR}\\ C_{mA}& C_{AA}& C_{AR}\\ C_{mR}& C_{AR}& C_{RR}\end{array}\right)=\left(\begin{array}{ccc}7.28& 0& 0\\ 0& 7.24& 2.54\\ 0& 2.54& 23.03\end{array}\right);\left(\begin{array}{c}\overline{\delta V_m}\\ \overline{\delta V_A}\\ \overline{\delta V_R}\end{array}\right)=\left(\begin{array}{c}0.07\\ 0.33\\ +0.01\end{array}\right).$$ (5) In Fig. 1 the functions $`\delta ^{\stackrel{~}{h}}V_i`$ are plotted against the chargino-neutralino mass $`m_{\stackrel{~}{\chi }^{\pm ,0}}`$. Comparing this graph with formulas (4) and (5), we see that at $`95\%`$ C.L. the bound $`m_{\stackrel{~}{\chi }^{\pm ,0}}54\text{ GeV}`$ should be satisfied. Note that the main contribution to $`\chi ^2`$ comes from $`\delta ^{\stackrel{~}{h}}V_A`$, which is singular at $`m_{\stackrel{~}{\chi }^{\pm ,0}}=m_Z/2`$. This singularity is not physical and our formulas are valid only for $`2m_{\stackrel{~}{\chi }^{\pm ,0}}m_Z+\mathrm{\Gamma }_Z`$; the existence of $`\chi ^\pm `$ with a mass closer to $`m_Z/2`$ will change Z-boson Breit-Wigner curve, therefore it is also not allowed. The importance of the $`Z`$ wave function renormalization for the case of light charginos was emphasized in . ### 2.2 Wino-dominated case For case (2) (gaugino dominated states) the expressions for the $`\delta ^{\stackrel{~}{w}}V_i`$ functions are: $`\delta ^{\stackrel{~}{w}}V_m`$ $`={\displaystyle \frac{16}{9}}\left[c^4\left(1+2\chi \right)F(\chi )\left(12s^2\right)\left(1+2{\displaystyle \frac{\chi }{c^2}}\right)F\left({\displaystyle \frac{\chi }{c^2}}\right){\displaystyle \frac{s^4}{3}}\right],`$ (6) $`\delta ^{\stackrel{~}{w}}V_A`$ $`={\displaystyle \frac{16}{9}}c^4\left[{\displaystyle \frac{12\chi ^2F(\chi )2\chi 1}{4\chi 1}}\right],`$ (7) $`\delta ^{\stackrel{~}{w}}V_R`$ $`={\displaystyle \frac{16}{9}}c^2s^2\left[\left(1+2\chi \right)F(\chi ){\displaystyle \frac{1}{3}}\right].`$ (8) The values of these functions are shown in Fig. 2 and at 95% C.L. we get $`m_{\stackrel{~}{\chi }^{\pm ,0}}60\text{ GeV}`$. Let us remark that, although this and the previous bounds have been obtained respectively in the limits $`\|\mu \|\mathrm{}`$ and $`\|M_2\|\mathrm{}`$ (in which case the mass splitting $`\mathrm{\Delta }M^\pm `$ is exactly zero), we have verified numerically using equations from Ref. that they are still valid for values of $`\|M_2\|`$ and $`\|\mu \|`$ small enough to allow for $`\mathrm{\Delta }M^\pm 1\text{ GeV}`$. ## 3 Conclusions Since there are a number of new additional particles in SUSY extensions, we will briefly discuss their contributions to the functions $`V_i`$. In the considered limits the remaining charginos and neutralinos are very heavy, so they simply decouple and produce negligible contributions. The contributions of the three generations of sleptons (with masses larger than $`90\text{ GeV}`$) into $`V_A`$ are smaller than $`0.1`$, so they can be safely neglected. The contributions of squarks of the first two generations are also negligible since they should be heavier than Tevatron direct search bounds; taking $`m_{\stackrel{~}{q}}200\text{ GeV}`$, we have $`\|\delta ^{\stackrel{~}{q}}V_i\|0.1`$. Concerning the contributions of the third generation squarks, they are enhanced by the large top-bottom mass difference and are not negligible. However, being positive and almost universal , they do not affect our analysis: compensating negative contributions of chargino-neutralino into $`V_A`$ they will generate positive contributions to $`V_R`$ and $`V_m`$, and $`\chi ^2`$ will not be better. When squarks are heavy enough (for $`m_{\stackrel{~}{b}}300\text{ GeV}`$), they simply decouple and their contributions become negligible as well. The last sector of the theory to be discussed is Higgs bosons. Unlike the case of Standard Model now we have one extra charged higgs and two extra neutral higgses. Their contributions to radiative corrections were studied in detail in . According to Fig. 2 from that paper it is clear that the contributions of MSSM higgses (and $`SU(2)\times U(1)`$ gauge bosons) equal with very good accuracy those of the Standard Model with the mass of SM higgs being equal to that of the lightest neutral higgs in SUSY generalization. That is why the contributions from the gauge-Higgs sector of the theory also cannot compensate those of the light chargino-neutralino. Apart from oblique corrections (those arising from vector bosons self energies) which have been considered in this letter, there are process dependent vertex and box corrections. However, due to LEP II and Tevatron low bounds on squarks and sleptons masses they are small. ## 4 Acknowledgments We are grateful to F. L. Villante for useful discussion on $`\chi ^2`$ analysis and to P. H. Chankowski for bringing Ref. to our attention. Investigation of M. V. are supported by the grants of RFBR No. 98-07-90076 and 98-02-17372.
no-problem/9907/chao-dyn9907022.html	ar5iv	text	# Anderson Localization in a String of Microwave Cavities ## I INTRODUCTION In 1958 P.W. Anderson calculated the effects of perturbations of a periodic lattice on the eigenvalues and eigenfunctions of the Schrödinger equation . He was able to show that for a variety of disordered potentials the eigenfunctions are exponentially localized in a small region of the lattice due to the interference of waves scattered from the perturbations or impurities. Till then the localization of waves played a crucial role in almost every physical domain – for a review see e.g. . In the case of a long, disordered one-dimensional (1D) chain it could be shown rigorously, that Anderson localization occurs for a very large class of potentials . However there are classes of disordered potentials where extended states do exist . It has been conjectured that localization effects are far more general and are a generic feature of wave equations . Experiments working with electromagnetic waves are usually focusing on secondary features of localization, e.g. coherent backscattering or the investigation of the transmission and absorption coefficients . A direct search for localization in electromagnetic field distributions has only very recently been attempted in two-dimensional (2D) and three-dimensional (3D) systems composed of 1D waveguides . In the present paper we like to report on the direct observation of Anderson localization in an electromagnetic system, i.e. a microwave resonator, governed by the vectorial Helmholtz equation, with a perturbed periodicity. So far microwave resonators have been a major experimental tool for the investigation of so-called quantum billiard systems , i.e. a point-like particle caught in a potential with infinitely high walls, or recently for the study of models used in nuclear physics . In all these experiments the analogy between the scalar Helmholtz equation which describes the electromagnetic field inside a flat microwave cavity and the Schrödinger equation is used and the energy spectra of the resonators are statistically analyzed – for a review of a wide range of experiments and the statistical methods used see e.g. . Here we will discuss experiments performed with a so-called three-dimensional microwave cavity, i.e. a resonator that has to be described by the vectorial Helmholtz equation and belongs to the class of resonators investigated in . As mentioned these experiments compose to our knowledge the first and direct experimental search for Anderson localization in such systems, not to be confused with the study of higher dimensional lattices still described by the Schrödinger equation . One of the main problems in such an experimental investigation is the limited size of the system at hand. In numerical simulations of finite chains described by the Schrödinger equation the number of elementary cells is usually some hundred cells or even higher (for an early review see e.g. ), while a realistic experimental set-up with a chain of microwave resonators has to have a much smaller number of elementary cells to keep the dimension of the system at an acceptable level. The paper is organized as follows. In Sect. II we will give a description of the microwave resonator and the experimental methods used to measure the resonance frequencies and the field distributions of the cavity. After having verified that the unperturbed system is indeed periodic and that we find extended states we proceeded to investigate the influence of a single perturbation of the periodicity on the eigenstates and eigenfrequencies – the results are given in Sect. III. As a third possible configuration we tried to set-up a disordered chain and measured the field distributions discussed in Sect. IV. ## II EXPERIMENT To simulate periodic and perturbed systems, an accelerating cavity of the superconducting Darmstadt linear electron accelerator S-DALINAC has been modified. The cavity itself is manufactured from 2 mm thick Niobium sheetmetal and consists of 18 identical cells and two slightly different cells at the ends of the chain to compensate the influence of the cut-off tubes attached to the outer cells. The cylindrical symmetric section with the 20 cells has a length of 1m and the diameter of a single cell varies between 91 mm and 39 mm. A sketch of the modified accelerating cavity is given in Fig. 1. Below 3.5 GHz the two cut-off tubes cause an exponential decay of the electromagnetic field outside the 20 cell section and above 3.5 GHz two Niobium plates can be used to close the first and the last cell. In both cases only the 20 cell section is excited up to 20GHz by a HP8510B vectorial network analyzer connected to a set of 20 identical and periodically mounted capacitively coupling dipole antennas. We chose a large penetration depth of 8 mm for the antennas to ensure that even modes which are localized in just one or two cells can be observed in the transmission spectra. Care has been taken to ensure that all antennas are identical and periodically mounted so that the periodicity of the resonator is not perturbed by the antennas themselves. To perturb the periodic setup every second cell is equipped with an adjustable lead screw with a diameter of 20 mm which can penetrate into the resonator’s volume. The penetration depth of each screw $`d^i=0\mathrm{m}\mathrm{m},i=1,3,\mathrm{},19`$ can be continuously varied between 0 mm and 47 mm. The cavity therefore allows the investigation of three different set-ups, i.e. a periodic system, a system with a single perturbation or impurity with variable strength and a disordered system where the $`d^i`$’s are set to random values. Despite the fact that the measurement of the field distributions has been performed at room temperature, all materials used will become superconducting at temperatures which can easily be reached inside a LHe-bath cryostat described in to allow future high resolution measurements of the cavity’s spectra. The electromagnetic fields $`\stackrel{}{E}`$ and $`\stackrel{}{B}`$ inside the cavity are described by the vectorial Helmholtz equation $$(\mathrm{\Delta }+ϵ\mu \frac{\omega ^2}{c_0^2})\stackrel{}{E}(\stackrel{}{r})=\stackrel{}{0}$$ (1) and $$(\mathrm{\Delta }+ϵ\mu \frac{\omega ^2}{c_0^2})\stackrel{}{B}(\stackrel{}{r})=\stackrel{}{0}$$ (2) with the corresponding boundary conditions $$\stackrel{}{E}_{}(\stackrel{}{r})\|_G=\stackrel{}{0}\mathrm{and}\stackrel{}{B}_{}(\stackrel{}{r})\|_G=\stackrel{}{0}$$ (3) on the walls $`\delta G`$ which are assumed to be ideally conducting. The well known analogy between flat electromagnetic cavities and two-dimensional quantum potentials, that has stimulated a multitude of experiments is of course lost. As mentioned above the effects of Anderson localization can be found either in the eigenfunctions, in our case the electromagnetic field distributions, or the energy spectrum, in our case the set of resonance frequencies, of a system. We examined therefore both, the field distributions and the energy spectrum, for signatures of localization effects. The electromagnetic field amplitudes inside the cavity can be measured using Slater’s theorem formulated in that describes shifts of the resonance frequencies of a microwave cavity if a perturbing body is brought into the resonating volume. The electric and the magnetic field energies inside a microwave cavity which is excited in resonance are of equal magnitude if the cavity is in a stationary state. If an external body is introduced into the resonator, the field energies are shifted and the resonance frequency $`f_0`$ will be adjusted in a way that the electric and magnetic field energies become equal again and a new stationary state is reached. According to the frequency shift $`\mathrm{\Delta }f=ff_0`$ of a resonance caused by a perturbing body of volume $`\mathrm{\Delta }V`$ is $$\frac{\mathrm{\Delta }f}{f_0}=\frac{1}{4U}_{\mathrm{\Delta }V}(ϵϵ_0\stackrel{}{E}^2\mu \mu _0\stackrel{}{H}^2)𝑑V,$$ (4) where $`U`$ is the total energy stored in the electromagnetic field and $`ϵ`$ and $`\mu `$ are the perimittivity and the permeability of the perturbing body. If $`\mathrm{\Delta }V`$ is small compared to the wavelength of the mode under investigation, the perturbing body is moved on a nodal line of the magnetic field (i.e. $`\stackrel{}{H}=\stackrel{}{0}`$) and is composed of a dielectric material, Eq. (4) can be written as $$\frac{\mathrm{\Delta }f}{f_0}=\frac{ϵϵ_0\mathrm{\Delta }V}{4U}\stackrel{}{E}^2,$$ (5) which immediately leads to the proportional relation $$\mathrm{\Delta }f\stackrel{}{E}^2.$$ (6) The squared field strength $`\stackrel{}{E}^2`$ can therefore be calculated directly from measurements of the frequency shift caused by a small perturbing body. To measure the frequency shift we used the experimental set-up sketched in Fig. 2. The resonator is moved with a speed of $`1\mathrm{m}/\mathrm{min}`$ while the perturbing body is fixed on the symmetry axis of the cavity with a very thin string, to prevent oscillations of the body. The first twenty resonances of the cavity are transverse magnetic (TM) modes for which the magnetic field is zero on the axis. We further used a cylindrical Teflon bead as a perturbing body with a volume of $`\mathrm{\Delta }V=2\mathrm{m}\mathrm{m}^3`$ and a permittivity of $`ϵ2.1`$ and a permeability of $`\mu 1`$ . The special symmetry of our system and the perturbing body we are using allows us to measure $`\stackrel{}{E}^2`$ directly by using Eq. (6) unlike similar experiments , where a metallic perturbing body is used and the composed quantity $`2\stackrel{}{E}+\stackrel{}{B}`$ is measured. The network analyzer can be set to continuously sweep a range of 1 MHz across the last position of the resonance which takes a time of 184ms. In the upper part of Fig. 3 a part of the transmission spectrum with a resonance at 2.8715 GHz and the Teflon cylinder outside and inside the resonator is shown. The frequency shift $`\mathrm{\Delta }f`$ caused by the perturbation of the electromagnetic field is clearly visible. Nevertheless, additional noise blurs the position of the resonance so that it is impossible to find its position by simply looking for the frequency with the highest transmission. Transmitting the data via the IEEE-bus for an offline analysis with e.g. fitted resonance curves is also not practical since during the time needed for the data transfer ($`1\mathrm{s}`$) the cavity moves about $`16\mathrm{m}\mathrm{m}`$. Beside the measurement of the ratio between the received and the emitted wave, the HP8510B allows the analysis of the phase relation $`\varphi `$ between the received and the emitted wave which shifts by $`\pi `$ at the resonance frequency. For a resonance with a frequency $`f_0`$ and a quality factor $`Q`$ one can write the phase relation as $$\varphi =\mathrm{arctan}(\frac{ff_0}{Q(f_0^2f^2)})+\pi /2$$ (7) which can be expanded around the point $`f=f_0`$ to a linear relation $$\varphi =\frac{\pi }{2}+2\frac{Q(f_0f)}{f_0}+\mathit{0}(f^2)$$ (8) As one can see in the lower part of Fig. 3 the frequency shift $`\mathrm{\Delta }f`$ is small compared to the region where the linear series approximation of Eq. (7) is applicable. The network analyzer is therefore set to emit a wave with the frequency $`f_0`$ and measure the phase relation $`\varphi `$ between the received and the emitted wave. Since no time consuming sweep is required for measurements performed at a constant frequency, the Teflon bead is assumed to be fixed during each measurement and the received signal can be averaged up to 4000 times, therefore greatly reducing the underlying noise. The difference of the phase relations with the bead outside and inside the cavity, $`\mathrm{\Delta }\varphi ,`$ as shown in the lower part of Fig. 3, can then easily be used with Eq. (8) to compute the field amplitudes. From Eqs. (6) and (8) one gets the proportionality relation $$\mathrm{\Delta }\varphi \mathrm{\Delta }f\stackrel{}{E}^2.$$ (9) In a first step we measured the field distribution of the unperturbed cavity, i.e. the penetration depths of the screws were set to zero, and compared our measurements with finite element calculations using the MAFIA computer code. The model used for the finite element calculations is cylindrically symmetric with a resolution of $`1\mathrm{m}\mathrm{m}\times 0.25\mathrm{mm}`$ and allows the calculation of the resonance frequencies and the field distributions of the first 40 TM modes. In Fig. 4 the experimental and the calculated electric field distributions for the 20th mode are compared, showing that the effects of small deviations from the ideal geometry, either due to the additional holes that were drilled into the cavity or the cut-off tubes, are negligible. ## III A SINGLE PERTURBATION If the system does show a transition from extended to localized states if the periodicity is perturbed – despite the finite size and the fact that it has to be described by the vectorial Helmholtz equation – one expects for a single perturbation of the otherwise periodic chain only one local impurity mode in each band with a localized field distribution. We therefore investigated the development of the field distribution of the first resonance if the penetration depth of an arbitrary screw (here the one in cell # 13) is increased. The evolution of the wave function is shown in Fig. 5, where we plotted the frequency shift $`\mathrm{\Delta }f`$ against the position $`z`$ of the Teflon bead. Beside perturbing the periodicity of the cavity the screw also breaks its cylindrical symmetry so that the cavities axis is not a nodal line of the magnetic field anymore. Calculations with MAFIA have nevertheless shown that the magnetic field on the axis is still zero for a wide range of penetration depths. For penetration depths above approximately $`40\mathrm{m}\mathrm{m}`$ the actual electric field is slightly larger than the one determined by Eq. (9) since the proportionality relation has to be written as $$\mathrm{\Delta }\varphi \mathrm{\Delta }fϵϵ_0\stackrel{}{E}^2\mu \mu _0\stackrel{}{H}^2.$$ (10) For $`d^{13}=15\mathrm{m}\mathrm{m}`$ the first resonance excites the whole cavity (Fig. 5a) while for an increasing $`d^{13}`$ the field amplitude drops down to zero almost everywhere in the resonator. The field distribution of the first resonance even localizes up to a point where only the disturbed and a few neighboring cells are excited (Fig. 5d). As it is the case for systems described by the Schrödinger equation, the field distributions of the other 19 modes of the first TM band are still extended over the whole cavity. In the case of strong localization in quantum systems the envelope of the localized eigenfunction can for a wide range of perturbed potentials be approximated by an exponentially decaying function , i.e. $$\|\psi _{loc}(z)\|^2\mathrm{exp}\left(\frac{\|zz_0\|}{L_{loc}}\right),$$ (11) where $`z_0`$ is the position of the perturbation and $`L_{loc}`$ is called the localization length. Despite the fact, that even for strongly localized modes the field distribution is not symmetric with respect to $`z_0`$ – mainly because of the small dimensions of our system – Eq. (11) can be used to describe the envelope of $`\stackrel{}{E}^2`$ for a wide range of penetration depths $`d^{13}`$. In Fig. 6 one can see a nearly linear dependence between $`d^{13}`$ and $`L_{loc}`$ for penetration depths above $`35\mathrm{m}\mathrm{m}`$. For a smaller perturbation the field distribution shows a clear maximum (see e.g. Fig. 5b) around the perturbing screw, but Eq. (11) cannot be used to describe the envelope of $`\stackrel{}{E}^2`$, since the decay is not exponential, and $`L_{loc}`$ is therefore not well defined. To underline the point that the field distribution has a maximum in the disturbed cell one usually calls this form of localization weak localization . The behaviour of the field distributions of our microwave resonator can be compared to the eigenfunctions computed for numerical models with a single perturbation of a periodic chain . Beside the transition from an extended to an exponentially localized eigenfunction the solutions of the Schrödinger equation do have other prominent features in periodic potentials with a single perturbation. One of these is a shift in the eigenvalue of the local impurity mode, while the eigenvalues of the solutions in a band that are still extended over the system remain fixed. An investigation of the resonance frequency $`f_0`$ of the first eight modes of the microwave resonator while $`d^{13}`$ is increased, see Fig. 7, shows that only the local impurity mode – in this case the lowest excitation of the cavity – experiences a notable shift in its resonance frequency. Even in the regime of weak localization, i.e. around $`d^{13}30\mathrm{m}\mathrm{m},`$ where the field distributions of all modes are extended over the whole cavity, no shift of the resonance frequency of the first mode is observed. The transition from an extended or weakly localized state to a strong, i.e. exponential, localized state can therefore not only be observed by measuring the eigenfunctions but also by looking at the eigenfrequencies. To conclude this section we like to state that we found a complete correspondence between the eigenfunctions and eigenvalues of the Schrödinger equation in a periodic potential with a single perturbation and the field distributions and resonance frequencies of a periodic 3D microwave cavity with a single, sufficiently large perturbation. ## IV MULTIPLE PERTURBATIONS In contrast to the emerging of a local impurity mode at the position of a single perturbation or impurity, one expects for a large class of one-dimensional disordered systems that all the eigenfunctions of a scalar wave equation are exponentially localized . For disordered acoustic systems a coexistence between extended, weakly localized and strongly localized modes has been reported by He and Maynard . To examine the behaviour of a disordered electromagnetic system we set the penetration depths of all screws to various, non periodic alternating lengths. The positions of the screws are sketched in an inset above Fig. 8 where also several different field distributions are shown. In contrast to the predictions for one-dimensional systems in not all field distributions are exponentially localized as for example the first (Fig. 8a) or the 20th (Fig. 8b) mode with localization lengths of $`L_{loc}^145\mathrm{m}\mathrm{m}`$ and $`L_{loc}^{20}60\mathrm{m}\mathrm{m}`$. We still found states which show weak localization, e.g. the fifth mode (Fig. 8c), or that are still extended throughout the whole system (ninth mode, Fig. 8d). Dean and Bacon found a similar behaviour in an early numerical model where they studied the eigenfunctions of a disordered harmonic chain composed of 22 light atoms and 28 heavy ones and which is described by the Schrödinger equation . In their model, localization occurs always on the upper edge of a band, while we are observing localization at both, the upper and the lower edge of the first band. The extended modes we are observing are found in the middle of the band like the ninth mode shown in Fig. 8d. As mentioned above a coexistence of localized and extended modes has been already observed for an acoustic system . Our particular set-up allows the examination of another interesting system: If the penetration depths off all screws are set to the same value, the system will be periodic again – an elementary cell is now composed of an unperturbed and a perturbed cell – and should therefore show only extended wave functions. As an example we compare in Fig. 9 the field distributions of the 19th mode of the cavity with $`d^i=0\mathrm{m}\mathrm{m},i=1,3,\mathrm{},19`$ and $`d^i=40\mathrm{m}\mathrm{m},i=1,3,\mathrm{},19`$. As expected we found, that both wave functions have essentially the same envelope, despite the fact that for a randomly perturbed setting the 19th mode is strongly localized. In both cases the field distribution is not exactly symmetric – an effect that is caused by geometrical imperfections and which is also visible in the upper part of Fig. 4. ## V CONCLUSION By using an appropriately shaped 3D microwave cavity we investigated a finite, periodic or disordered system with eigenfunctions described by the vectorial Helmholtz equation. The field distributions inside the cavity were measured by analyzing the phaseshifts caused by a small dieelectric body inside the resonator. In the case of a chain of quasi-identical cells we find extended field distributions – the whole resonator volume is excited. It is well known that the eigenstates of a scalar wave equation are exponentially localized in the case of infinitely long, disordered chains and that there is always one local impurity mode per band in the case of a single perturbation or impurity of the chain. Our experiments showed that the same behaviour can be observed for a finite – in fact with just 20 cells very small – system although it is described by the vectorial Helmholtz equation. We observed the transition from an extended state to an exponentially localized state in the case of a single perturbation by looking at the field distribution and the resonance frequency of the mode. In the regime of strong localization we found a linear dependence between the localization length and the penetration depth of the screw that causes the perturbation. The current set-up does not allow the investigation of totally disordered chains but enables us to study systems where extended modes coexist with strong and weakly localized ones which can be compared to certain numerical models for scalar problems or results found in acoustic systems . It is also possible to study chains composed of a periodic array of two-cell elements in which all states are extended over the whole cavity. Our experiments showed that appropriately shaped microwave cavities exhibit all the features of extended periodic systems – despite the fact that they are finite and in our case have to be described by vectorial wave equations. To our knowledge this is also the first time that the localization of an eigenfunction of the vectorial Helmholtz equation has been observed in an electromagnetic system. The experimental verification of predictions on the statistical behavior of extended billiard chains, see e.g. , should therefore be possible and combine the results of the theory of periodic systems and so-called quantum chaotic systems . The experimental set-up itself already allows the study of various problems from the fields of solid-state physics and scattering problems in a very clean and pedagogical way. ## VI acknowledgments We would like to thank the workshop of the Institut for Nuclear Physics in Darmstadt for the excellent and precise modifications of the microwave resonator. We would also like to thank T. Dittrich, S. Fishman, Y. Imry, C. Rangacharyulu, U. Smilansky and H.-J. Stöckmann for very helpful discussions. This work has still been supported in part by the Sonderforschungsbereich 185 ”Nichtlineare Dynamik” of the Deutsche Forschungsgemeinschaft (DFG) and through the Forschergruppe with contract number DFG RI242/12-1.
no-problem/9907/hep-th9907028.html	ar5iv	text	# References Let us consider the motion of a pair of 2D-electrons in a constant and uniform magnetic field. In this problem, the motion of the electrons center of mass can be considered separately from their relative motion. The stationary Schrödinger equation for the relative motion (the Laughlin problem) has the form $$\left\{\frac{1}{m^{}}\left[\left(\widehat{p}_x\frac{eB}{4c}\widehat{y}\right)^2\left(\widehat{p}_y+\frac{eB}{4c}\widehat{x}\right)^2\right]+\frac{e^2}{\sqrt{x^2+y^2}}\right\}\mathrm{\Psi }=E\mathrm{\Psi }.$$ (1) Here $`m^{}`$ is the effective electron mass in the given heterostructure, $`e`$ is the absolute value of the electron charge, $`B`$ is the magnetic field induction. The symmetric gauge for the vector potential in the above equation has been used $$\stackrel{}{A}=\frac{B}{2}y\stackrel{}{e}_x+\frac{B}{2}x\stackrel{}{e}_y=\frac{B}{2}r\stackrel{}{e}_\phi .$$ (2) Equation (1) formally corresponds to the Schrödinger equation for a mass $`m^{}/2`$ and charge $`e/2`$ particle interacting with the uniform magnetic field (2) and with a fixed charge $`2e`$ placed at the origin. Equation (1) is conveniently transformed to dimensionless variables. To this end, we introduce the magnetic length $`l_B=\sqrt{2\mathrm{}c/\left(eB\right)}`$ and the dimensionless energy eigenvalue $`\lambda =2E/\left(\mathrm{}\omega _c\right),`$ where $`\omega _c=eB/\left(m^{}c\right)`$ is the cyclotron frequency. Upon introduction of the new dimensional variables $$x=l_B\xi ,y=l_B\eta ,\rho =\sqrt{\xi ^2+\eta ^2},$$ (3) equation (1) takes the form $$\left[\left(\widehat{p}_\xi \frac{1}{2}\widehat{\eta }\right)^2+\left(\widehat{p}_\eta +\frac{1}{2}\widehat{\xi }\right)^2+\frac{a}{\rho }\right]\mathrm{\Psi }=\lambda \mathrm{\Psi }.$$ (4) Here $`a=\sqrt{B_0/B}`$ and $`B_0=2cm^2e^3/\mathrm{}^34.7\times 10^9\left(m^{}/m\right)^2`$ is the critical magnetic field. In the polar coordinates, equation (4) reads $$\left[\left(\frac{^2}{\rho ^2}+\frac{1}{\rho }\frac{}{\rho }\right)\frac{1}{\rho ^2}\frac{^2}{\phi ^2}i\frac{}{\phi }+\frac{1}{4}\rho ^2+\frac{a}{\rho }\right]\mathrm{\Psi }=\lambda \mathrm{\Psi }.$$ (5) The wave function is looked for in the form $$\mathrm{\Psi }=\mathrm{exp}\left(il\phi \right)R\left(\rho \right).$$ (6) The radial part of the wave function is described by the equation $$\left[\left(\frac{^2}{\rho ^2}+\frac{1}{\rho }\frac{}{\rho }\right)+\frac{l^2}{\rho ^2}+l+\frac{1}{4}\rho ^2+\frac{a}{\rho }\right]R\left(\rho \right)=\lambda R\left(\rho \right).$$ (7) Here $`l=0\pm 1,\pm 2,\mathrm{}`$ are eigenvalues of the operator $`L_z=i\frac{}{\phi }`$. In the following, only the case $`l=0`$ is considered, and hence the equation takes the form $$\left[\left(\frac{^2}{\rho ^2}+\frac{1}{\rho }\frac{}{\rho }\right)+\frac{1}{4}\rho ^2+\frac{a}{\rho }\right]R\left(\rho \right)=\lambda R\left(\rho \right).$$ (8) The Hamiltonian operator in the left-hand side of (8) can be written in the form $$\widehat{H}=\widehat{p}_\rho ^2+\frac{1}{4}\rho ^2+\frac{1}{4\rho ^2}+\frac{a}{\rho }.$$ (9) Here $$\widehat{p}_\rho =\frac{1}{i}\left(\frac{}{\rho }+\frac{1}{2\rho }\right).$$ (10) is the radial momentum operator, selfadjoint in the Hilbert space $`L^2(0,\mathrm{},\rho d\rho ).`$ It should be mentioned that the term $`1/\left(4\rho ^2\right)`$ has appeared in (9). Radial eigenfunctions are looked for in the form $$R\left(\rho \right)=\mathrm{exp}\left(\rho ^2/4\right)f\left(\rho \right).$$ (11) The equation for the function $`f\left(\rho \right)`$ reads $$\frac{d^2f}{d\rho ^2}+\left(\frac{1}{\rho }\rho \right)\frac{df}{d\rho }+\left(\lambda 1\frac{a}{\rho }\right)f\left(\rho \right)=0.$$ (12) For magnetic fields of arbitrary strength (the parameter $`a`$ takes arbitrary values), the regular at the origin solution of equation (12), ensuring that $`R\left(\rho \right)`$ belongs to the Hilbert space, is given by a series with a complicated and almost unknown structure. It is only for some unique values of the magnetic field that the functions $`f\left(\rho \right)`$ are reduced to polynomials \- , so that the radial functions $`R\left(\rho \right)`$ take the form $$R_{nk}\left(\rho \right)=C_{nk}\mathrm{exp}\left(\rho ^2/4\right)Q_{nk}\left(\rho \right),$$ (13) where $$Q_{nk}\left(\rho \right)=\underset{j=0}{\overset{n}{}}b_j\rho ^j$$ (14) is the order $`n`$ polynomial with exactly $`k`$ zeros ($`n`$ and $`k`$ are the principle and the radial quantum numbers respectively) in the physical region $`\left(\rho 0\right)`$. The eigenvalues $`\lambda `$ for all the states of this unique kind are given by the simple unified formula $$\lambda =n+1,n+1,2,\mathrm{}.$$ (15) Coefficients of the polynomial $`Q_{nk}\left(\rho \right)`$ are determined by the recurrence relations $$\begin{array}{cc}b_0=\hfill & 1,\hfill \\ b_1=\hfill & 2,\hfill \\ b_j=\hfill & \left[ab_{j1}+\left(jn2\right)b_{j2}\right]j^2.\hfill \end{array}$$ (16) The unique values of the magnetic fields are determined by the relations $$\left(n+1\right)^2b_{n+1}=ab_nb_{n1}=0.$$ (17) Several leading values of the parameter follow $$\begin{array}{cc}a_{10}=1,\hfill & \\ a_{20}=\sqrt{6},\hfill & \\ a_{30}=\sqrt{10+\sqrt{73}},\hfill & a_{31}=\sqrt{10\sqrt{73}},\hfill \\ a_{40}=\sqrt{25+3\sqrt{33}},\hfill & a_{41}=\sqrt{253\sqrt{33}}.\hfill \end{array}$$ (18) It is of interest to remark that the energy levels for the unique states (13) can be obtained from the Bohr - Sommerfeld quantization rule, taking into consideration both the physical $`\left(\rho 0\right)`$ and nonphysical $`\left(\rho <0\right)`$ ranges of the variable $`\rho `$. In this case, the effective potential energy should be put equal to (see (9)) $$U_{eff}\left(\rho \right)=\frac{1}{4}\rho ^2+\frac{a}{\rho }+\frac{1}{4\rho ^2}.$$ (19) The Bohr - Sommerfeld quantization rule with allowance for the above remarks can be put into the form $$2\left(\underset{\rho _1}{\overset{\rho _2}{}}𝑑\rho \sqrt{\lambda U_{eff}\left(\rho \right)}+\underset{\rho _3}{\overset{\rho _4}{}}𝑑\rho \sqrt{\lambda U_{eff}\left(\rho \right)}\right)=2\pi \left(n+1\right).$$ (20) The appearance of unity in the right-hand side of (20) is explained by the fact that there exist four regular lower turning points, each of them contributing $`1/4.`$ It should be mentioned that, when the term $`1/\left(4\rho ^2\right)`$ is not included in expression (19), we have to assume that there is an impenetrable potential wall at the point $`\rho =0`$ with the contribution $`1/2`$, instead of $`1/4`$ . Integration in the left-hand side of equation (20) is easily performed with the help of the residue technique (the integration contour is depicted in Fig. 3b). As a result, we obtain exactly formula (13) for $`\lambda `$, irrespective of $`a`$ values. The physical meaning of the particular magnetic field values determined by relations (15) was unclear up to now. In the following, we present considerations that may elucidate the physical origin of this phenomenon. Introduce the center of orbit operators $$\widehat{X}_c=\widehat{p}_\eta +\frac{1}{2}\widehat{\xi },\widehat{Y}_c=\widehat{p}_\xi +\frac{1}{2}\widehat{\eta }.$$ (21) Unlike , the symmetric gauge has been used here, and transformation to the dimensionless form has been performed by means of the magnetic length introduced. Then the operator of the square of the distance between the center of the orbit and the origin can be written in the form $$\widehat{R}_c^2=\widehat{p}_\xi ^2+\widehat{p}_\eta ^2+\frac{1}{4}\widehat{\rho }^2+\widehat{L}_z.$$ (22) In the case of a pure magnetic field, these operator commute with the Hamiltonian, so that, in a stationary state, the value of $`R_c^2`$ is quite well defined. In our case $`R_c^2`$ can be written in the form $$\widehat{R}_c^2=\widehat{H}2\widehat{L}_z\frac{a}{\rho }.$$ (23) Due to the presence of the Coulomb force, in our case, $`R_c^2`$ is not conserved even in a stationary state, and hence, only the quantum average of this physical quantity makes sense. Introduce now the average value of the radius of orbit squared as it has been done in for the pure magnetic field case $$\widehat{R}^2=\rho ^2\widehat{R}_c^2.$$ (24) With allowance for (4), (15) and for the fact that $`l=0`$, this equation can be written in the form $$\widehat{R}^2=\rho ^2+a\frac{1}{\rho }\lambda .$$ (25) Thus, the average magnetic flux in a stationary state (ordinary units are again used) takes the form $$\mathrm{\Phi }=\pi \widehat{R}^2Bl_B^2=\widehat{R}^2\mathrm{\Phi }_0,$$ (26) where $`\mathrm{\Phi }_0=2\pi c\mathrm{}/e`$ is the magnetic flux quantum. Calculations demonstrated that, for stationary states (13), the quantity $`\mathrm{\Phi }`$ is the integral multiple of $`\mathrm{\Phi }_0`$ $$\mathrm{\Phi }=\left(n+1\right)\mathrm{\Phi }_0,n=2,3,\mathrm{}.$$ (27) In the following, the results of calculations are illustrated by the examples for the cases of states $`f_{10}`$ and $`f_{20}`$. For the state $`f_{10}`$ we have $`a=1,\lambda =2`$ $$f_{10}\left(\rho \right)=\mathrm{exp}\left(\rho ^2/4\right)\left(1+\rho \right),$$ $$\rho ^2=\left(10+3\sqrt{2\pi }\right)/\left(3+\sqrt{2\pi }\right),$$ $$\frac{1}{\rho }=\left(2+\sqrt{2\pi }\right)/\left(3+\sqrt{2\pi }\right),$$ $$\widehat{R}^2=2.$$ For the state $`f_{20}`$ we have $`a=\sqrt{6},\lambda =3`$ $$f_{20}\left(\rho \right)=\mathrm{exp}\left(\rho ^2/4\right)\left(1+\sqrt{6}\rho +\rho ^2\right),$$ $$\rho ^2=\left(114+36\sqrt{3\pi }\right)/25+8\sqrt{3\pi },$$ $$\frac{1}{\rho }=6\left(\sqrt{6}+\sqrt{2\pi }\right)/\left(25+8\sqrt{2\pi }\right),$$ $$\widehat{R}^2=3.$$ We believe that formula (26) is valid for all the unique states, though we are unable to present a general proof of this statement now. We are grateful to V.Ch. Zhukovskii and A.V. Borisov for helpful discussions. Figure captions Figure 1. The curves demonstrate qualitative behaviour of the relation $`\lambda =2E/\mathrm{}\omega `$ as a function of the parameter $`a=\sqrt{B_0/B}`$. The dots mark the value of parameters $`\lambda `$ and $`a`$ of unique states. The lower curve corresponds to the ground state, the next one is for the first exited state and so on. Fig. 2. a) The curves of normalized radial functions of unique states $`f_{nk}\left(r\right)`$ in physical $`\left(\rho 0\right)`$ and nonphysical $`\left(\rho <0\right)`$ ranges at several values of principal $`\left(n\right)`$ and radial $`\left(k\right)`$ quantum numbers. b) The curves of the normalized radial density of unique states $`D_{nk}\left(r\right)=rf_{nk}^2\left(r\right)`$ at several values of principal $`\text{}\left(n\right)`$ and radial $`\left(k\right)`$ quantum numbers. The values of a radial variable are measured in the units of the effective Bohr radius $`a_B^{}=\mathrm{}^2/\left(m^{}e^2\right)`$. Fig. 3. a) On the curve of the effective potential $`U_{eff}\left(\rho \right)`$ in physical $`\left(\rho 0\right)`$ and nonphysical $`\left(\rho <0\right)`$ ranges at $`\lambda =2`$ the turning points are shown. b) The contour of circumvention of four branching points in the complex plane.
no-problem/9907/astro-ph9907196.html	ar5iv	text	# Galaxy Morphology in the GTO-NICMOS Northern Hubble Deep Field ## 1. Introduction When coupled with distance estimates such as photometric redshifts, the study of morphology has the potential to probe the dynamical state and evolution of galaxies. However, morphological classification is only reliable to redshifts of a few tenths when hampered by ground-based seeing. The deep, high-resolution WFPC 2 imaging of the Hubble Deep Field (HDF, Williams et al. 1996) dramatically pushed the study of galaxy morphology to faint magnitudes and high redshifts, revealing that by $`I_{AB}>24`$, the traditional Hubble sequence no longer provides an adequate description of most galaxies (Abraham et al. 1996). Some of the faint peculiar galaxies are sub-luminous irregulars at modest redshifts, while others are higher-$`z`$. But are they “true peculiars” – the counterparts to local irregulars? Matters are complicated by morphological $`k`$-corrections: for single waveband selection, shorter rest wavelengths are sampled in higher-$`z`$ galaxies. The rest-UV is dominated by sites of recent star formation, and it is known that the appearance of local Hubble-sequence galaxies can be very different in the UV compared to the optical (e.g., O’Connell 1997). This change in apparent morphology, resulting from a dispersion of stellar populations, is well illustrated by some HDF spirals at moderate redshift ($`z1`$) which undergo a complete metamorphosis from the observed optical to the near-IR (Fig. 2). To address whether the apparent increased incidence of peculiars in the optical at faint magnitudes is attributable to genuine evolution in the fraction of irregular galaxies, or whether it is predominantly due to band-shifting effects, the appearance of galaxies over a variety of redshifts should be compared at the same rest wavelength (Fig. 3). ## 2. A Study of HDF-North We have analyzed galaxy morphology to faint magnitudes in HDF-North using the optical & near-IR HST images (Bunker, Spinrad & Thompson 1999). We have studied the GTO-NICMOS data set (Thompson et al. 1999), a 1 arcmin<sup>2</sup> area of the HDF imaged for 49 orbits with NIC 3 in both F110W ($`J`$-band) and F160W ($`H`$-band). Combined with the four WFPC 2 pass-bands (Fig. 3), this data set provides deep, multi-color, high-resolution imaging extending out to 1.6 $`\mu `$m – the rest-optical at $`z2`$. We use the redshifts of the galaxies to match the rest-wavelengths, determine intrinsic luminosities, and to fit stellar populations/dust reddening to the spectral energy distributions. Where available, we use the spectroscopically-measured redshifts (from Cohen et al. 1996 unless otherwise noted). Where no published spectroscopic redshift exists, we adopt the photometric redshift estimate of Fernández-Soto, Lanzetta & Yahil (1999). ### Comparative Morphology: Down to $`I_{AB}26`$ (the brightest 100 galaxies in GTO-NICMOS field): only about 1/6 of galaxies change their appearance greatly between the WFPC 2 and NICMOS images – these have large morphological $`k`$-corrections; about half of the galaxies retain the same morphology in all wavebands (above the redshifted Lyman break) and are “true irregulars”; the remaining third of galaxies are too compact for changes in morphology to be ascertained (the NIC 3 PSF has a FWHM of $`0.25`$arcsec); for most cosmologies, the higher-redshift systems are on average more compact. ### Spatially-Resolved Stellar Populations: Once we correct for different resolutions of NIC 3 and WFPC 2 (through “PSF matching”), we can use the spatially-resolved colors to study different stellar populations and/or dust-reddening within a galaxy (see Figs. 2 & 5). Some of the galaxies which have the same appearance at all wavelengths fall outside the traditional Hubble tuning-fork diagram, but instead belong to new morphological groups, such as chain galaxies (Fig. 5; Cowie, Hu & Songaila 1995), tadpoles (van den Bergh et al. 1996) and bow-shock systems (Fig. 5). ### Barred Spirals: Our data can also address the evolution of galactic bars: it has been claimed that at faint magnitudes, the fraction of barred spirals in the optical HDFs declines rapidly (van den Bergh et al. 1996, Abraham et al. 1999). If this is a truly evolutionary effect, then it has great significance for the physics of disk formation: bars are supported by disk self-gravity, so the implication would be either that at high-$`z`$ the halo mass dominates that of the disk, or there are significant random motions in the stellar orbits (Ostriker & Peebles 1973). However, when the spirals are imaged in the near-IR, many are revealed to have bars which are absent in the WFPC 2 bands (Fig. 6): the bars have similar colors to the bulges (dominated by older, cooler, redder stars). It appears that morphological $`k`$-correction effects for the higher-$`z`$ spirals cause the apparent decline in optically-selected barred spirals at fainter magnitudes. From the small-number statistics of spirals in the GTO-NICMOS field, there is no significant evolution in the incidence of galactic bars. ## 3. Conclusions Some Hubble tuning-fork galaxies only reveal their true morphology in near-IR. This is particularly so for galaxies with a large dispersion in stellar ages and spatially-distinct stellar populations, such as in spiral galaxies. However, such galaxies which undergo a morphological metamorphosis from the WFPC 2 to NIC 3 images are rare; most retain the same appearance in all wavebands, or are too compact for the structural parameters to be determined. Once the morphological $`k`$-corrections have been accounted for, it appears that the fraction of true irregulars does increase at faint magnitudes/high-$`z`$. Finally, the deep near-IR data shows that there is no significant evolution in the incidence of barred spirals with redshift: their apparent scarcity in the optical is a band-shifting effect on the older stellar population of their bars. A more detailed description of this work is given in Bunker, Spinrad & Thompson (1999). #### Acknowledgments. I wish to thank my collaborators on this program, Hyron Spinrad and Rodger Thompson. We are grateful to Ray Weymann and Lisa Storrie-Lombardi at OCIW for organizing an enjoyable and timely workshop on photometric redshifts, and thank Daniel Stern, Leonidas Moustakas and Mark Dickinson for useful discussions. A.J.B. acknowledges by a NICMOS postdoctoral research fellowship, supported in part by NASA grant NAG 5-3043. The observations were obtained with the NASA/ESA Hubble Space Telescope operated by the Space Telescope Science Institute managed by the Association of Universities for Research in Astronomy Inc. under NASA contract NAS 5-26555. ## References Abraham, R. G., Tanvir, N. R., Santiago, B. X., Ellis, R. S., Glazebrook, K., & van den Bergh, S. 1996, MNRAS, 279, 47P Abraham, R. G. 1997, in “The UV Universe at Low- and High-$`z`$”, ed. Waller, W. H., Faneli, M. N., Hollis, J. E., & Danks, A. C. AIP Press, vol. 408, p195 Abraham, R. G., Merrifield, M. R., Ellis, R. S., Tanvir, N. R., & Brinchmann, J. 1999, MNRAS, in press, astro-ph/9811476 Bruzual, G., & Charlot, S. 1993, ApJ, 405, 538 Bunker, A. J., Stern, D., Spinrad, H., Dey, A., & Steidel, C. C. 1998, BAAS, 192.7008 Bunker, A. J., Spinrad, H., & Thompson, R. I. 1999, AJ, submitted, & BAAS, 194.0403 Cohen, J. G., Cowie, L. L., Hogg, D. W., Songaila, A., Blandford, R., Hu, E. M., & Shopbell, P. 1996, ApJ, 471, 5 Cowie, L. L., Hu, E. M., & Songaila, A. 1995, AJ, 110, 1576 Fernández-Soto, A., Lanzetta, K. M., & Yahil, A. 1999, ApJ, 513, 34 O’Connell, R. W. 1997, in “The UV Universe at Low- and High-$`z`$”, ed. Waller, W. H., Faneli, M. N., Hollis, J. E., & Danks, A. C. AIP Press, vol. 408, p11 Ostriker, J. P., & Peebles, J. E. 1973, ApJ, 186, 467 Steidel, C. C., Giavalisco M., Dickinson M. E., & Adelberger K. L. 1996, AJ, 112, 352 Thompson, R. I., Storrie-Lombardi, L. J., Weymann, R. J., Rieke, M. J., Schneider, G., Stobie, E., & Lytle, D. 1999, AJ, 117, 17 van den Bergh, S., Abraham, R. G., Ellis, R. S., Tanvir, N. R., Santiago, B. X., & Glazebrook, K. G. 1996, AJ, 112, 359 Williams, R. E., Blacker, B., Dickinson, M., et al. 1996, AJ, 112, 1335
no-problem/9907/quant-ph9907049.html	ar5iv	text	# A position-momentum EPR state of distantly-separated trapped atoms ## I Introduction In 1935, Einstein, Podolsky and Rosen (EPR) proposed a now famous gedanken experiment involving a system of two particles spatially separated but correlated in position and momentum as described by the Wigner function $$W(q_1,p_1;q_2,p_2)\delta (q_1+q_2)\delta (p_1p_2),$$ (1) where $`q_1`$ and $`q_2`$ are the continuous position variables of the particles with corresponding conjugate momenta $`p_1`$ and $`p_2`$. With the assumption of local realism, but with the apparent ability to assign definite values to canonically conjugate variables of one particle from measurements of the other particle in this system, a conflict with the Heisenberg uncertainty principle seemingly follows, which led EPR to conclude that quantum mechanics is incomplete. Bohm adapted this argument to a system of discrete (dichotomic) variables, to which Bell applied his classic analysis, deriving the so-called Bell inequalities which quantify explicitly the conflict between local realism and quantum mechanics. Note that although measurements of $`(q_i,p_i)`$ do not lead to a violation of a Bell inequality for the original EPR state of Eq. (1) (Ref., p.196), the entanglement of this state guarantees that an appropriate set of variables exists for which a contradiction with local realism would be manifest . Experimental demonstrations of the conflict between quantum mechanics and local realism have concentrated almost exclusively on systems of discrete variables, such as electron spin or photon polarization . Only one experiment, by Ou et al. , following suggestions by Reid and Drummond , has in fact realized the EPR paradox as originally envisioned by EPR; that is, for canonically conjugate variables with a continuous spectrum. The EPR “source” in this experiment was a nondegenerate optical parametric amplifier (NOPA), and the relevant variables were the quadrature amplitudes of the entangled electromagnetic fields generated in the parametric process. These amplitudes are analogous to the position and momentum of a particle and can be measured very efficiently via homodyne detection . In the present work, we describe a new scheme that goes beyond an analogy and actually realizes an EPR state in position and momentum for a pair of massive particles at distinct physical locations. Our proposal for achieving stored entanglement for continuous quantum variables is based upon a set of interactions in cavity quantum electrodynamics (QED) that allows for the exchange of quantum states between the motion of trapped atoms and propagating light fields . By exploiting these interactions and the light source of , we show that it should be possible to prepare deterministically a state of the form (1) for a pair of trapped atoms located at macroscopically-separated sites. Beyond conventional $`(q,p)`$ projections as in homodyne or heterodyne measurements, the setting of atom traps and cavity QED also enables detection strategies for the explicit demonstration of the nonlocal character of the resulting EPR state. Moreover, the techniques that we describe could be important resources for the realization of quantum networks, a particular example being the creation of EPR states to enable the teleportation of the center-of-mass wave function of a massive particle . ## II Trapped atom coupled to an optical cavity mode We begin with the basic setup that facilitates the motion-light coupling fundamental to our scheme ; this setup was originally considered by Zeng and Lin . We consider a single two-level atom (or ion) confined in a harmonic trap located inside an optical cavity. The atomic transition of frequency $`\omega _\mathrm{a}`$ is coupled to a single mode of the cavity field of frequency $`\omega _\mathrm{c}`$ and is also assumed to be driven by an external (classical) laser field of frequency $`\omega _\mathrm{L}`$. The physical setup and excitation scheme are depicted in Fig. 1. The cavity is aligned along the $`x`$-axis, while the laser field is incident from a direction in the $`y`$-$`z`$ plane (i.e., perpendicular to the $`x`$-axis). The Hamiltonian describing the internal and external atomic degrees of freedom plus the atom-cavity and atom-laser couplings takes the form (in a frame rotating at the laser frequency for the internal-atomic and cavity operators) $`\widehat{H}_0=`$ $`{\displaystyle \underset{j=x,y,z}{}}\mathrm{}\nu _j(\widehat{b}_j^{}\widehat{b}_j+1/2)+\mathrm{}\delta \widehat{a}^{}\widehat{a}+\mathrm{}\mathrm{\Delta }\widehat{\sigma }_+\widehat{\sigma }_{}`$ (4) $`+\mathrm{}\left[_\mathrm{L}(\widehat{y},\widehat{z},t)\widehat{\sigma }_++_\mathrm{L}^{}(\widehat{y},\widehat{z},t)\widehat{\sigma }_{}\right]`$ $`+\mathrm{}g_0\mathrm{sin}(k\widehat{x})(\widehat{a}^{}\widehat{\sigma }_{}+\widehat{\sigma }_+\widehat{a}).`$ Here, $`\{\nu _x,\nu _y,\nu _z\}`$ are the harmonic oscillation frequencies along the principal axes of the trap, $`\widehat{b}_j`$ and $`\widehat{a}`$ are annihilation operators for the quantized atomic motion and cavity field, respectively, $`\widehat{\sigma }_{}=\|ge\|`$ is the atomic lowering operator, and $`\delta =\omega _\mathrm{c}\omega _\mathrm{L}`$ and $`\mathrm{\Delta }=\omega _\mathrm{a}\omega _\mathrm{L}`$. The quantity $`_\mathrm{L}(\widehat{y},\widehat{z},t)`$ is the (possibly time-dependent) amplitude of the laser field. The single-photon atom-cavity dipole coupling strength is given by $`g_0`$, while the sine function describes the standing wave structure of the cavity field (we assume that the centre of the trap is located at a node of the cavity field), with $`k=2\pi /\lambda `$ the wavenumber of the field and $`\widehat{x}=[\mathrm{}/(2m\nu _x)]^{1/2}(\widehat{b}_x+\widehat{b}_x^{})`$. In a number of assumptions and approximations are made in order to simplify the model. In particular: * The detunings of the light fields from the atomic transition frequency are assumed to be very large (i.e., $`\mathrm{\Delta }\|_\mathrm{L}\|,g_0,\delta ,\nu _j`$), enabling atomic spontaneous emission to be neglected and the internal atomic dynamics to be adiabatically eliminated. * Any forms of motional decoherence associated with the trap itself are ignored. * The size of the harmonic trap is assumed to be small compared to the optical wavelength (Lamb-Dicke regime), enabling the approximations $`\mathrm{sin}(k\widehat{x})\eta _x(\widehat{b}_x+\widehat{b}_x^{})`$, where $`\eta _x`$ ($`1`$) is the Lamb-Dicke parameter, and $`_\mathrm{L}(\widehat{y},\widehat{z},t)_\mathrm{L}(t)e^{i\varphi _\mathrm{L}}`$. * The cavity and laser fields are tuned so that $`\delta =\omega _\mathrm{c}\omega _\mathrm{L}=\nu _x`$. * The trap frequency $`\nu _x`$ and cavity field decay rate $`\kappa _a`$ are assumed to satisfy $`\nu _x\kappa _a\|(g_0\eta _x/\mathrm{\Delta })_\mathrm{L}(t)\|`$. The first inequality allows a rotating-wave approximation to be made with respect to the trap oscillation frequency, while the second inequality enables an adiabatic elimination of the cavity field mode. Given these conditions one can show that the motional mode dynamics in the $`x`$ direction can be described by the simple quantum Langevin equation $$\dot{\stackrel{~}{b}}_x\mathrm{\Gamma }(t)\stackrel{~}{b}_x+\sqrt{2\mathrm{\Gamma }(t)}\stackrel{~}{a}_{\mathrm{in}}(t),$$ (5) where $`\stackrel{~}{b}_x=e^{i\nu _xt}\widehat{b}_x`$, $`\mathrm{\Gamma }(t)=[g_0\eta _x_\mathrm{L}(t)/\mathrm{\Delta }]^2/\kappa _a`$, and $`\stackrel{~}{a}_{\mathrm{in}}(t)`$, which satisfies the commutation relation $`[\stackrel{~}{a}_{\mathrm{in}}(t),\stackrel{~}{a}_{\mathrm{in}}^{}(t^{})]=\delta (tt^{})`$, is the quantum noise operator describing the input to the cavity field (in a frame rotating at the cavity frequency). In this way, the statistics of the input light field can be “written onto” the state of the oscillator. In it was shown how this effect can be used to efficiently prepare a squeezed state of the motion of the trapped atom. Here we extend that work further to the generation of entanglement between the motional states of trapped atoms at separated sites. In particular, our protocol transfers entanglement from a pair of quantum-correlated light fields to a pair of trapped atoms in a process of quantum state exchange, or qusex. ## III Light source: nondegenerate parametric amplifier Our source of quantum-correlated light fields is taken to be a NOPA operating below threshold . The light fields may be nondegenerate in polarization or in frequency. We denote the annihilation operators for the two intracavity field modes, of frequencies $`\omega _1`$ and $`\omega _2`$, by $`\widehat{c}_1`$ and $`\widehat{c}_2`$, and the Hamiltonian describing the coupling between these modes takes the form (in a rotating frame) $$H_\mathrm{I}=i\mathrm{}ϵ\left(\widehat{c}_1\widehat{c}_2\widehat{c}_1^{}\widehat{c}_2^{}\right),$$ (6) where $`ϵ`$ is the coupling strength, proportional to the nonlinear susceptibility of the intracavity medium and to the strength of the coherent pump field (at frequency $`\omega _1+\omega _2`$). Assuming the cavity mode amplitudes to be damped at the same rate $`\kappa _c`$, equations of motion for the mode operators (in the rotating frame) can be derived as $$\dot{\widehat{c}}_{1,2}=\kappa _c\widehat{c}_{1,2}ϵ\widehat{c}_{2,1}^{}+\sqrt{2\kappa _c}\widehat{c}_{\mathrm{in}}^{(1,2)}(t),$$ (7) where $`\widehat{c}_{\mathrm{in}}^{(1,2)}(t)`$ are the vacuum input fields to the NOPA cavity modes (see, e.g., ). The output fields from the NOPA then follow from the boundary conditions $$\widehat{c}_{\mathrm{out}}^{(1,2)}(t)+\widehat{c}_{\mathrm{in}}^{(1,2)}(t)=\sqrt{2\kappa _c}\widehat{c}_{1,2}(t),$$ (8) The (linear) equations above are readily solved in a Fourier-transformed space defined by $`Z(\omega )=(2\pi )^{1/2}𝑑tZ(t)e^{i\omega t}`$. Defining quadrature phase amplitudes (“positions” and “momenta”) for the output fields by $`X_{\mathrm{out}}^{(1,2)}(t)`$ $`=`$ $`\widehat{c}_{\mathrm{out}}^{(1,2)}(t)+\widehat{c}_{\mathrm{out}}^{(1,2)}(t),`$ (9) $`Y_{\mathrm{out}}^{(1,2)}(t)`$ $`=`$ $`i\left\{\widehat{c}_{\mathrm{out}}^{(1,2)}(t)\widehat{c}_{\mathrm{out}}^{(1,2)}(t)\right\},`$ (10) the sum of the $`X`$ amplitudes is derived as $`X_{\mathrm{out}}^{(1)}(\omega )`$ $`+X_{\mathrm{out}}^{(2)}(\omega )`$ (13) $`={\displaystyle \frac{\kappa _cϵ+i\omega }{\kappa _c+ϵi\omega }}\left\{X_{\mathrm{in}}^{(1)}(\omega )+X_{\mathrm{in}}^{(2)}(\omega )\right\}`$ $`0\mathrm{as}ϵ\kappa _c\mathrm{and}\omega 0,`$ while the difference of the $`Y`$ amplitudes is $`Y_{\mathrm{out}}^{(1)}(\omega )`$ $`Y_{\mathrm{out}}^{(2)}(\omega )`$ (16) $`={\displaystyle \frac{\kappa _cϵ+i\omega }{\kappa _c+ϵi\omega }}\left\{Y_{\mathrm{in}}^{(1)}(\omega )Y_{\mathrm{in}}^{(2)}(\omega )\right\}`$ $`0\mathrm{as}ϵ\kappa _c\mathrm{and}\omega 0.`$ So, the two output fields are highly correlated and, close to $`\omega =0`$ and for $`ϵ\kappa _c`$, their quadrature amplitudes exhibit precisely the properties of the original EPR state, as demonstrated explicitly by the Wigner function for the state of the (1,2) fields . ## IV Light-to-motion quantum state exchange As depicted in Fig. 2, the two NOPA output fields are assumed to be incident on separate cavities, each containing a trapped atom in the configuration described earlier. Note that the output fields from the NOPA are resonant with the respective cavity mode frequencies. We assume that $`\mathrm{\Gamma }(t)=\mathrm{\Gamma }`$, a constant, and, for simplicity, that $`\mathrm{\Gamma }`$ is the same for both configurations. Denoting the motional mode operators for the two atoms along the $`x`$-axis by $`\stackrel{~}{b}_{1x}`$ and $`\stackrel{~}{b}_{2x}`$, respectively, the two systems are thus described by $`\dot{\stackrel{~}{b}}_{jx}`$ $`=`$ $`\mathrm{\Gamma }\stackrel{~}{b}_{jx}+\sqrt{2\mathrm{\Gamma }}\widehat{a}_{\mathrm{in}}^{(j)}(t)`$ (17) $`=`$ $`\mathrm{\Gamma }\stackrel{~}{b}_{jx}+\sqrt{2\mathrm{\Gamma }}\widehat{c}_{\mathrm{out}}^{(j)}(t\tau ),(j=1,2)`$ (18) where $`\tau `$ is a time delay (assumed the same for both cavities); provided the coupling between the NOPA and the cavities is unidirectional, this delay can essentially be ignored . If the bandwidths of the input light fields from the NOPA are sufficiently broad, in particular if $`\kappa _c\mathrm{\Gamma }`$ \[i.e., $`\widehat{c}_{\mathrm{out}}^{(1,2)}(t)`$ can be regarded as quantum white noise operators in (18)\], then one can perform an average over the input fields and derive a master equation (see, e.g., ) for the density operator $`\rho `$ of the motional modes alone, $`\dot{\rho }`$ $`=\mathrm{\Gamma }(N+1)(2\stackrel{~}{b}_{1x}\rho \stackrel{~}{b}_{1x}^{}\stackrel{~}{b}_{1x}^{}\stackrel{~}{b}_{1x}\rho \rho \stackrel{~}{b}_{1x}^{}\stackrel{~}{b}_{1x})`$ (24) $`+\mathrm{\Gamma }N(2\stackrel{~}{b}_{1x}^{}\rho \stackrel{~}{b}_{1x}\stackrel{~}{b}_{1x}\stackrel{~}{b}_{1x}^{}\rho \rho \stackrel{~}{b}_{1x}\stackrel{~}{b}_{1x}^{})`$ $`+\mathrm{\Gamma }(N+1)(2\stackrel{~}{b}_{2x}\rho \stackrel{~}{b}_{2x}^{}\stackrel{~}{b}_{2x}^{}\stackrel{~}{b}_{2x}\rho \rho \stackrel{~}{b}_{2x}^{}\stackrel{~}{b}_{2x})`$ $`+\mathrm{\Gamma }N(2\stackrel{~}{b}_{2x}^{}\rho \stackrel{~}{b}_{2x}\stackrel{~}{b}_{2x}\stackrel{~}{b}_{2x}^{}\rho \rho \stackrel{~}{b}_{2x}\stackrel{~}{b}_{2x}^{})`$ $`+\mathrm{\hspace{0.17em}2}\mathrm{\Gamma }M(\stackrel{~}{b}_{1x}\rho \stackrel{~}{b}_{2x}+\stackrel{~}{b}_{2x}\rho \stackrel{~}{b}_{1x}\stackrel{~}{b}_{1x}\stackrel{~}{b}_{2x}\rho \rho \stackrel{~}{b}_{1x}\stackrel{~}{b}_{2x})`$ $`+\mathrm{\hspace{0.17em}2}\mathrm{\Gamma }M(\stackrel{~}{b}_{1x}^{}\rho \stackrel{~}{b}_{2x}^{}+\stackrel{~}{b}_{2x}^{}\rho \stackrel{~}{b}_{1x}^{}\stackrel{~}{b}_{1x}^{}\stackrel{~}{b}_{2x}^{}\rho \rho \stackrel{~}{b}_{1x}^{}\stackrel{~}{b}_{2x}^{}),`$ with the parameters $`N`$ and $`M`$ given, in terms of the NOPA parameters, by $$N=\frac{4ϵ^2\kappa _c^2}{(\kappa _c^2ϵ^2)^2},M=2\kappa _cϵ\frac{\kappa _c^2+ϵ^2}{(\kappa _c^2ϵ^2)^2}.$$ (25) This master equation has a steady state solution $$\rho ^{\mathrm{ss}}=\|\psi _{12}\psi _{12}\|,$$ (26) i.e., a pure state, with $`\|\psi _{12}`$ $`=`$ $`S_{12}(r)\|0_{1x}\|0_{2x}`$ (27) $`=`$ $`\left[\mathrm{cosh}(r)\right]^1{\displaystyle \underset{m=0}{\overset{\mathrm{}}{}}}\left[\mathrm{tanh}(r)\right]^m\|m_{1x}\|m_{2x},`$ (28) where $`\|m_{1x,2x}`$ are Fock states of the motional modes and $`S_{12}(r)`$ is the two-mode squeezing operator , $$S_{12}(r)=\mathrm{exp}\left[r\left(\stackrel{~}{b}_{1x}\stackrel{~}{b}_{2x}\stackrel{~}{b}_{1x}^{}\stackrel{~}{b}_{2x}^{}\right)\right].$$ (29) This operator transforms the mode annihilation operators for the atomic motion as $`S_{12}^{}(r)\stackrel{~}{b}_{1x}S_{12}(r)`$ $`=`$ $`\mathrm{cosh}(r)\stackrel{~}{b}_{1x}\mathrm{sinh}(r)\stackrel{~}{b}_{2x}^{},`$ (30) $`S_{12}^{}(r)\stackrel{~}{b}_{2x}S_{12}(r)`$ $`=`$ $`\mathrm{cosh}(r)\stackrel{~}{b}_{2x}\mathrm{sinh}(r)\stackrel{~}{b}_{1x}^{},`$ (31) where $`\mathrm{cosh}(r)=\sqrt{N+1}`$ and $`\mathrm{sinh}(r)=\sqrt{N}`$. Defining position and momentum operators as $$Q_j=\stackrel{~}{b}_{jx}+\stackrel{~}{b}_{jx}^{},P_j=i\left(\stackrel{~}{b}_{jx}\stackrel{~}{b}_{jx}^{}\right),$$ (32) it follows that $`S_{12}^{}(r)\left(Q_1+Q_2\right)S_{12}(r)`$ $`=`$ $`e^r\left(Q_1+Q_2\right)`$ (33) $`S_{12}^{}(r)\left(P_1P_2\right)S_{12}(r)`$ $`=`$ $`e^r\left(P_1P_2\right),`$ (34) and so, in the limit $`ϵ\kappa _c`$ (i.e., $`r\mathrm{}`$), an EPR state in the positions and momenta of the two trapped atoms is established. The nature of the correlations inherent in the joint state (27) of the atomic motion is most clearly expressed through the Wigner function for this state : $`W(q_1,p_1;q_2,p_2)`$ (35) $`={\displaystyle \frac{4}{\pi ^2}}\mathrm{exp}\left\{\left[(q_1+q_2)^2+(p_1p_2)^2\right]e^{+2r}\right\}`$ (36) $`\times \mathrm{exp}\left\{\left[(q_1q_2)^2+(p_1+p_2)^2\right]e^{2r}\right\}`$ (37) $`C\delta (q_1+q_2)\delta (p_1p_2)\mathrm{as}r\mathrm{},`$ (38) with $`C`$ a constant. This entangled state is achieved in steady state over a time $`t\mathrm{\Gamma }^1`$. The coupling to the external fields from the NOPA can then be turned off by setting $`_{\mathrm{L1},2}`$ to zero. The result is a stored EPR state for the motion of two trapped atoms that would persist for a duration set by the timescale for motional decoherence. ## V Discussion Before considering some of the interesting possibilities offered by this system, we return briefly to some of the major assumptions associated with the model. Firstly, the finite effect of atomic spontaneous emission events on the motional state can be estimated as in . This effect can be neglected provided the rate of these events is much smaller than the rate $`\mathrm{\Gamma }`$ at which the motional steady state is achieved. The condition one derives by enforcing this inequality amounts to the condition of a “one-dimensional” atom in cavity QED, $`C_1=g_0^2/(\kappa _a\gamma )1`$, where $`\gamma `$ is the linewidth (FWHM) of the atomic transition . Note that the experiment of Ref. has achieved $`C_1=70`$. With regards to the trapping potential, harmonic frequencies on the order of tens of MHz have been achieved in ion traps, with corresponding Lamb-Dicke parameters on the order of $`0.1`$ and smaller . Note, however, that large values of the entanglement parameter $`r`$ imply population of large-$`m`$ number states and a broader spread of the atomic wavepacket. Given that the mean excitation number for the state (27) is $`\overline{n}=\mathrm{sinh}^2(r)`$, a more precise form of the Lamb-Dicke assumption would be $`\eta _x\sqrt{\overline{n}+1}=\eta _x\mathrm{cosh}(r)1`$. With trap frequencies such as those quoted above, the condition $`\nu _x\kappa _a`$ should be satisfied for a cavity field decay rate of a few MHz or less; such values of $`\kappa _a`$ have not been realized in current experiments as in Ref., but potentially could be with improved cavity finesse as in Ref.. Assuming that this is the case, likely magnitudes for the rate $`\mathrm{\Gamma }`$ would then be tens or hundreds of kHz. Finally, timescales for motional decoherence and heating in recent ion trap experiments are of the order of milliseconds, with further improvement likely ; given the various rates discussed above, these effects would not be expected to hamper the preparation of the entangled state. As for applications of this system, further investigation of the EPR paradox would obviously be possible, with a variety of motional state measurements able to be implemented on the trapped atoms , possibly also via the cavity field . In particular, violations of a Bell inequality for the state (27) can be obtained with measurements that project onto a basis of even and odd phonon number for each of the trapped atoms . To the extent that a “macroscopic” number of quanta may in principle be involved, such investigation could also address new viewpoints on the compatibility of quantum mechanics with local realism . On a somewhat more applied side is the possibility of using the EPR state (27) for quantum dense coding , or for the teleportation of the quantum state of a system with continuous variables , generalizing the original discrete-variable teleportation protocol of Bennett et al. . This elegant adaptation of the EPR paradox has in fact been realized with light fields, again using optical parametric amplifiers and homodyne measurements of quadrature amplitudes . The scheme outlined in this paper opens the door to teleporting an atomic center-of-mass wave function , by providing the motional state entanglement required by the continuous-variable teleportation protocol. Such a capability is also of considerable interest in the related context of quantum computation with trapped atoms and light . Here, we specifically have in mind protocols that combine quantum information processing with both discrete and continuous variables . Any implementation of a qubit (e.g., internal atomic states or photon polarization) could be linked with an external degree of freedom (e.g., atomic center-of-mass or complex amplitude of the electromagnetic field), with the complete system viewed as a composite unit (qubit plus qunat ) for protocols such as quantum communication between distant nodes of a quantum network . Further to this theme, note also that we need not restrict ourselves to a single trapped atom at each site. For example, if there are $`K`$ atoms inside each cavity, then focussing the coupling lasers ($`_{\mathrm{L1},2}`$) sequentially on atoms $`1`$, $`2`$, $`\mathrm{}`$, $`K`$ at each site (and neglecting any direct interaction between neighbouring atoms at each site) would generate a set of pair-wise EPR-entangled atoms. Alternatively, and perhaps more interestingly, one might consider the case in which the $`K`$ atoms at each site are simultaneously coupled to the cavity. Assuming for simplicity that they have identical coupling strengths, then the system would again be described by Eq. (24), but with the replacements $`\stackrel{~}{b}_{1,2x}\stackrel{~}{B}_{1,2x}K^{1/2}_{j=1}^K\stackrel{~}{b}_{1,2x}^{(j)}`$ and $`\mathrm{\Gamma }K\mathrm{\Gamma }`$. Such dynamics evidently leads to a highly entangled state of all 2K atoms, a situation of potentially great utility and, indeed, of considerable general interest. ## ACKNOWLEDGMENTS ASP gratefully acknowledges support from the Marsden Fund of the Royal Society of New Zealand. HJK is supported by the National Science Foundation, by DARPA via the QUIC Institute which is administered by ARO, and by the Office of Naval Research.
no-problem/9907/quant-ph9907063.html	ar5iv	text	# REFERENCES Nuclear Magnetic Resonance Quantum Computing Using Liquid Crystal Solvents Costantino S. Yannoni, Mark H. Sherwood, Dolores C. Miller, and Isaac L. Chuang IBM Almaden Research Center, San Jose, CA 95120 Lieven M.K. Vandersypen Solid State and Photonics Laboratory, Stanford University, Stanford, CA 94305-4075, and IBM Almaden Research Center, San Jose, CA 95120 Mark G. Kubinec College of Chemistry, D62H Hildebrand Hall, University of California Berkeley, Berkeley, CA 94720-1460 ## Abstract Liquid crystals offer several advantages as solvents for molecules used for nuclear magnetic resonance quantum computing (NMRQC). The dipolar coupling between nuclear spins manifest in the NMR spectra of molecules oriented by a liquid crystal permits a significant increase in clock frequency, while short spin-lattice relaxation times permit fast recycling of algorithms, and save time in calibration and signal-enhancement experiments. Furthermore, the use of liquid crystal solvents offers scalability in the form of an expanded library of spin-bearing molecules suitable for NMRQC. These ideas are demonstrated with the successful execution of a 2-qubit Grover search using a molecule (<sup>13</sup>C<sup>1</sup>HCl<sub>3</sub>) oriented in a liquid crystal and a clock speed eight times greater than in an isotropic solvent. Perhaps more importantly, five times as many logic operations can be executed within the coherence time using the liquid crystal solvent. Quantum computations have been done using NMR techniques to manipulate ensembles of coupled nuclear spins (qubits) in molecules in solution. Isotropic solvents are favorable for NMRQC experiments since the narrow, well-resolved NMR lines satisfy the dual requirements of individual spin addressability and long coherence time. However, the quantum computer clock is slow, since the time required to execute logic gates is approximately $`1/2J`$ seconds where $`J`$, the strength of the spin-spin scalar coupling that persists in isotropic solution, is typically less than $`300`$ Hz for organic solutes. In liquid crystal solvents, however, the solute molecules become partially oriented against a background of rapid Brownian motion, resulting in well-resolved NMR spectra characterized by dipolar coupling between nuclear spins, which can be as large as $`10`$ kHz. Thus, the NMRQC clock in liquid crystal solvents can be much faster. Several other advantages of using liquid crystals for NMRQC experiments are outlined below. The behavior of $`N`$ coupled spins in a molecule dissolved in an isotropic liquid in an external magnetic field is given by the Hamiltonian $$H_{\mathrm{iso}}/h=\underset{i}{\overset{N}{}}\nu _iI_{zi}+\underset{i<j}{\overset{N}{}}J_{ij}(I_{xi}I_{xj}+I_{yi}I_{yj}+I_{zi}I_{zj})$$ (1) where $`\nu _i`$ is the resonance frequency (Hz), $`I_{zi}`$ is the z component of the angular momentum operator, and $`J_{ij}`$ is the strength of the scalar coupling (Hz). For two spins A and B, which yield a first order NMR spectrum, i.e. $`\|\nu _A\nu _B\|>>\|J\|`$, Eq.(1) becomes ) $$H_{iso}^{}/h=\nu _AI_{zA}+\nu _BI_{zB}+JI_{zA}I_{zB}.$$ (2) The time evolution of spins under $`H_{liq}^{}`$ permits the use of relatively simple pulse sequences to do NMR quantum computing. The clock frequency, $`f_{clock}`$, for an NMR quantum computer based on a two-spin system with a scalar coupling of 200 Hz is given by 2$`\|J\|`$ $``$ 400 Hz. The Hamiltonian for $`N`$ spins in a molecule dissolved in a liquid crystal solvent is $$H_{\mathrm{lc}}/h=\underset{i}{\overset{N}{}}\nu _i^{}I_{zi}+\underset{i<j}{\overset{N}{}}J_{ij}(I_{xi}I_{xj}+I_{yi}I_{yj}+I_{zi}I_{zj})+\underset{i<j}{\overset{N}{}}D_{ij}\left\{2I_{zi}I_{zj}\frac{1}{2}(I_{xi}I_{xj}+I_{yi}I_{yj})\right\}.$$ (3) A dipolar term now appears because the molecules are partially oriented and the resonance frequency $`\nu _i^{}`$ includes the effects of molecular orientation and chemical shift anisotropy. The dipolar coupling strength $`D`$, which also depends on the orientation, is typically $`100`$ Hz to $`10`$ kHz. However, the pulse sequences used for NMRQC in isotropic liquids can not be applied because of the form of the spin operators in Eq.(3). For a 2-spin system with a first order spectrum, Eq.(3) becomes $$H_{\mathrm{lc}}^{}/h=\nu _A^{}I_{zA}+\nu _B^{}I_{zB}+(J+2D)I_{zA}I_{zB}.$$ (4) Since Eq.(4) has the same form as Eq.(2), the pulse sequences that have been used successfully for NMRQC in isotropic solution can be applied directly to liquid crystal solutions, permitting computations with $`f_{clock}=2\|(J+2D)\|`$ Hz, a frequency that can be much higher than $`2\|J\|`$. This is borne out in Table I which shows the <sup>13</sup>C - <sup>1</sup>H coupling strength, spin-lattice relaxation time ($`T_1`$) and spin-spin relaxation (decoherence) time ($`T_2`$) for <sup>13</sup>C and <sup>1</sup>H in chloroform (<sup>13</sup>C<sup>1</sup>HCl<sub>3</sub>) in both liquid crystal and isotropic solution at ambient temperature. The <sup>13</sup>C-<sup>1</sup>H coupling in the liquid crystal (ZLI-1167) is eight times larger than the scalar coupling in acetone-d<sub>6</sub>, corresponding to a computer with a clock that is eight times faster. The product of the shortest coherence time and the clock frequency $`T_2f_{clock}=2T_2J`$, which approximates the number of gates that can be executed while maintaining coherence, may be used as a figure of merit. The data in Table I show that $`T_2^{lc}f_{clock}^{lc}5T_2^{iso}f_{clock}^{iso}`$, meaning that more complex algorithms requiring five times as many logic operations can be executed using this solute/liquid-crystal-solvent system. The chloroform <sup>13</sup>C and <sup>1</sup>H spin-lattice relaxation times are about $`12`$ times shorter in ZLI-1167 than in acetone-$`d_6`$. Since all NMRQC algorithms as well as NMR experiments used to set up and calibrate the spectrometer require a polarization time of $`5T_1`$ s, an order of magnitude savings in time can be significant, and will become more so as the number of qubits increases. This advantage will also be manifest when the sensitivity must be increased by co-addition of the signal from several successive NMRQC experiments, or when procedures which require multiple experiments such as quantum state tomography are used to diagnose the operation of quantum algorithms. Another advantage of using liquid crystals as solvents for NMRQC is that they permit a wider choice of spin-bearing molecules that may be suitable for NMRQC. Dipolar coupling, which is manifest in the NMR spectra of oriented molecules, requires only proximity between the spins of interest. As a result, two spins that are separated by several bonds and which have no scalar coupling may, if spatially proximate, have dipolar coupling sufficiently large for quantum computation. The ability to control the degree of orientation of the solute molecule by varying the solvent temperature and solute concentration provides the experimentalist with means of tailoring the NMR spectrum to meet the requirements for NMRQC. In addition, magic-angle spinning and multiple pulse methods can be used to preferentially scale the dipolar splitting in the spectrum of a liquid-crystal-oriented molecule to convert it to first order. Complications do arise with the use of liquid crystal solvents: (1) the NMR lines of small molecules dissolved in liquid crystal solvents are susceptible to a broadening mechanism not found in isotropic solution, most likely due to variations in the degree of orientation caused by thermal gradients in the sample. Nonetheless, resonance line widths $`<2`$ Hz (<sup>13</sup>C) and $`<3`$ Hz (<sup>1</sup>H) were obtained for <sup>13</sup>C<sup>1</sup>HCl<sub>3</sub> in ZLI-1167; (2) the large dipolar couplings may cause unwanted evolution of the spins during the relatively long pulses required for selective excitation in homonuclear spin systems. In order to show that quantum computations can be done successfully using liquid-crystal solution NMR, we have implemented the Grover search algorithm using <sup>13</sup>C<sup>1</sup>HCl<sub>3</sub> dissolved in ZLI-1167. The goal of the search is as follows: given a function $`f(x)`$, find the unknown element $`x_0`$ among four possible elements 00,01,10,11 - represented by the four spin-product states $`\|00,\|01,\|10,\|11`$ \- which satisfies $`f(x_0)=1`$, where $`f(x)=0`$ for the other three elements. Classically, this would take an average of $`2.25`$ attempts, while one query is sufficient using the Grover algorithm. The carbon and proton spins were first prepared in an effective pure state created by temporal labeling followed by a previously used Grover protocol. The prediction is that the algorithm will put the spins in the state $`\|x_0`$. The <sup>13</sup>C and <sup>1</sup>H readout spectra for the four possible $`x_0`$ are shown in Fig. 1. As predicted for two spins in an effective pure state, the value of $`x_0`$ is clearly indicated by the amplitude and phase of the two resonance lines in the <sup>13</sup>C and <sup>1</sup>H spectra. Measurement of the deviation density matrix using quantum state tomography, confirms the output states to be as theoretically predicted; Fig. 2 shows that for $`x_0=11`$, the final state of the spins is the state $`\|11`$. An immediate advantage of using liquid crystal solvents becomes clear from the savings in experimental time. A wait-time ($`5T_1`$) between experiments of only $`8`$ s sufficed, compared with $`105`$ s in acetone-$`d_6`$. The increase in spin coupling strength more than compensated for the shorter coherence time and permitted successful completion of the algorithm. The use of liquid crystals as solvents in NMR quantum computing has been proposed and demonstrated. Due to the large <sup>13</sup>C-<sup>1</sup>H dipolar coupling, the speed of the NMR quantum computer was increased by a factor of $`8`$ while the short spin-lattice relaxation times resulted in significant time saved in setting up the spectrometer and in executing the Grover search algorithm. This demonstration of the utility of liquid crystals as NMRQC solvents expands the library of potential quantum computing molecules. The authors would like to acknowledge helpful discussions on liquid crystal solvents with B. M. Fung. The support and encouragement of Nabil Amer, James Harris and Alex Pines are also appreciated. L.V. Acknowledges a Yansouni Family Stanford Graduate Fellowship. This work was performed under the auspices of the DARPA NMRQC initiative. FIGURE CAPTIONS $$\begin{array}{cc}\text{}& \text{}\\ \text{}& \text{}\\ \text{}& \text{}\\ \text{}& \text{}\end{array}$$ Figure 1 - Costantino Yannoni, Applied Physics Letters Figure 2 - Costantino Yannoni, Applied Physics Letters
no-problem/9907/hep-ph9907563.html	ar5iv	text	# UNITU-THEP-7/1999 hep-ph/9907563 Octet and Decuplet Baryons in a Confining and Covariant Diquark-Quark Model11footnote 1Supported by the BMBF (06–TU–888) and by the DFG (We 1254/4-1). Talk given by R. Alkofer at PANIC 99. ## 1 Motivation Different types of hadronic models describe various aspects of baryon physics. Among them are nonrelativistic quark models, various sorts of bag models and approaches describing baryons by means of collective variables like topological or non-topological solitons . Most of these models are designed to work in the low energy region and generally do not match the calculations within perturbative QCD. Considering the great experimental progress in the medium energy range , there is a high demand for models describing baryon physics in this region that connects the low and high energy regimes. To make progress in this direction we investigate a covariant formulation of a diquark-quark-model of baryons. Our motivation to choose such an approach is fed from two sources. On the one hand, when starting with the fully relativistic Faddeev equation for bound states of three quarks, diquarks appear as effective degrees of freedom. These diquarks stand for correlated quark-quark pairs inside baryons. Thus they should not be confused with the notion of diquark condensates in the context of colour superconductivity. On the other hand, diquarks as constituents of baryons are naturally obtained when one starts with an NJL-type of model of colour octet flavour singlet quark currents . Although in the limit $`N_c\mathrm{}`$ baryons emerge as solitons of meson fields , it can be shown for the case of three colours that both effects, binding through quark exchange in the diquark-quark picture and through mesonic effects, contribute equally . ## 2 Solving the four-dimensional Bethe-Salpeter equation: masses and wave functions Starting from the Faddeev equation, one can approximate the two-quark irrreducible $`T`$-matrix by separable contributions that can be viewed as loosely bound diquarks. The full three-body problem reduces then to a two-body one, in which bound states appear as the solution of a homogeneous Bethe-Salpeter equation. The attractive interaction between quark and diquark is hereby provided by quark exchange. In we formalize this procedure by an effective Lagrangian containing constituent quark, scalar diquark and axialvector diquark fields. This leads to a coupled set of Bethe-Salpeter equations for octet and decuplet baryons. We avoid unphysical thresholds by an effective parameterization of confinement in the quark and diquark propagators. We then solve the complete four-dimensional Bethe-Salpeter equation in ladder approximation and obtain wave functions for the octet and decuplet baryons . The Lorentz invariance of our model has been checked explicitly by choosing different frames. The implementation of the appropriate Dirac and Lorentz representations of the quark and diquark parts of the wave functions leads to a unique decomposition in the rest frame of the baryon. Besides the well known $`s`$-wave and $`d`$-wave components of non-relativistic formulations of the baryon octet we additionally obtain non-negligible $`p`$-wave contributions which demonstrates again the need for covariantly constructed models. Table 1 summarizes the structure of the octet wave function. Each of the eight components is to be multiplied with a scalar function which is given in terms of an expansion in hyperspherical harmonics and is computed numerically. In order to obtain the mass spectra for the octet and decuplet baryons we explicitly break SU(3) flavour symmetry by a higher strange quark constituent mass. Using the nucleon and the delta mass as input our calculated mass spectra are in good agreement with the experimental ones, see Table 2. The wave functions for baryons with distinct strangeness content but same spin differ mostly due to flavour Clebsch-Gordan coefficients, the respective invariant functions being very similar. Due to its special role among the other baryons, we investigated the $`\mathrm{\Lambda }`$ hyperon in more detail and discussed its vertex amplitudes. In our approach, the $`\mathrm{\Lambda }`$ acquires a small flavour singlet admixture which is absent in $`SU(6)`$ symmetric non-relativistic quark models. ## 3 Applications: Form Factors and Strangeness Production A significant test and a first application of our model is the calculation of various form factors . The most important ingredient are the fully four-dimensonal wave functions described above. It turns out that already the electromagnetic form factors of the nucleon provide severe resctrictions for the parameters of the model. The pion-nucleon form factor is calculated in view of its possible use in a spectator model for nucleon-nucleon scattering processes. At the soft point, $`Q^2=0`$, we find good agreement with experiment. In the spacelike region our $`g_{\pi NN}`$ falls like a monopole with a large cutoff used also in One-Boson-Exchange (OBE) models. Compared with a calculation including only scalar diquarks we find a lower value for the pion-nucleon coupling at the soft point. Serving as a central ingredient for strangeness production processes the kaon-nucleon-lambda form factor $`g_{KN\mathrm{\Lambda }}`$ is an issue of special interest. Due to flavour algebra the isospin configuration of the $`\mathrm{\Lambda }`$ singles out the scalar diquark as the only diquark contributing to nucleon-lambda transitions. With a pseudoscalar kaon not coupling to the scalar diquark we find ourselves in a comfortable position to handle such transitions. Results can be found in a forthcoming paper . Going beyond the calculation of form factors we work on the application of our approach to production processes like kaon photoproduction and the associated strangeness production close to threshold. Great experimental progress in the last few years gave access to kinematical as well as some spin observables due to self-analyzing $`\mathrm{\Lambda }`$ decay. We try to reproduce these observables in our model with the aim to get further insight into the mechanisms of strangeness production inside the nucleon. In the case of kaon photoproduction, $`\gamma pK\mathrm{\Lambda }`$, preliminary results are gained by computing the left diagram of fig. 1 and the corresponding one with photon and kaon line crossed. Although the outcome of the total cross section is encouraging, the result for the polarisation asymmetries of the $`\mathrm{\Lambda }`$ falls too short. Additionally, we describe the process $`pppK\mathrm{\Lambda }`$ employing an OBE-picture for internuclear forces. From our viewpoint, the diquarks act as mere spectators, so that the exchanged bosons couple only to the valence quarks of the baryons. In a first step we investigate the contributions stemming from pseudoscalar kaon and pion interchange between these two protons. The latter process is depicted in the right diagram of fig. 1. If instead of the pion a kaon is exchanged, the upper part of this diagram changes to the form factor $`g_{KN\mathrm{\Lambda }}`$. Possible applications are not exhausted by these examples. The diquark spectator picture can be extended to, e.g., real and virtual Compton scattering. The quark structure functions (see also ), which are obtained from the imaginary part of the virtual Compton scattering amplitude, would allow a comparison with the results of perturbative calculations. Acknowledgement: R. A. thanks W. Bentz and P. Maris for interesting discussions and the organizers and conveners of PANIC 99.
no-problem/9907/hep-ph9907313.html	ar5iv	text	# Chiral Phase Transition for 𝑺⁢𝑼⁢(𝑵) Gauge Theories ## Introduction The phase structure of strongly coupled gauge field theories as a function of the number of matter fields $`N_f`$ is a problem of general interest. Much has been learned about the phases of supersymmetric theories in recent years Seiberg ; Seiberg-Witten ; IntSeiberg ; Peskin ; DiVecchia . An equally interesting problem is the phase structure of a non-supersymmetric $`SU(N)`$ theory as a function of the number of fermion fields $`N_f`$. At low enough values of $`N_f`$, the chiral symmetry $`SU(N_f)_L\times SU(N_f)_R`$ is expected to break to the diagonal subgroup. At some value of $`N_f`$ less than $`11N/2`$ (where asymptotic freedom is lost), there will be a phase transition to a chirally symmetric phase. Whether the transition takes place at a relatively small value of $`N_f`$ mawhinney or a larger value remains unknown. The larger value ($`N_f/N4`$) is suggested by studies of the renormalization group improved gap equation ATW and is associated with the existence of an infrared fixed point. A recent analysis ASe indicates that instanton effects could also trigger chiral symmetry breaking at comparably large value of $`N_f/N`$. Besides being of theoretical interest, the physics of a chiral transition could have consequences for electroweak symmetry breaking AS , since near-critical gauge theories provide a natural framework for walking technicolor theories ATW2 . If a phase transition is second order, a useful approach is to find a tractable model in the same universality class. For chiral symmetry, a natural order parameter is the $`N_f\times N_f`$ complex matrix field $`M`$ describing mesonic degrees of freedom. If the meson degrees of freedom are the only ones that develop large correlation lengths at the phase transition, then the transition may be studied using an effective Landau-Ginzburg theory. For the zero-temperature transition as a function of $`N_f`$, a similar approach might also be tried. It was suggested in Ref.ATW , however, that while the order parameter vanishes continuously as $`N_fN_f^c`$, the transition is not second order. With the gap equation dominated by an infrared fixed point of the gauge theory, the transition was argued to be continuous but infinite order. It has also been noted sekhar that because of the associated long range conformal symmetry, the masses of all the physical states, not just the scalar mesons are expected to scale to zero with the order parameter. We nevertheless suggest that an effective potential using only the low lying mesonic degrees of freedom might be employed to model at least some aspects of the zero-temperature chiral phase transition. The key ingredient is the presence of a new non-analytic potential term that emerges naturally once the anomaly structure of the theory is considered. The anomalies also provide a link between this effective potential term and the underlying gauge theory. To deduce the anomaly induced effective potential we modify an effective potential toy ; HSS ; Sannino developed for $`N_f<N`$, and apply it to the range $`N_f>N`$. We use this potential to discuss the zero-temperature phases of an $`SU(N)`$ gauge theory as a function of $`N_f`$. Assuming that the transition is governed by an infrared fixed point of the theory, we deduce that chiral symmetry is restored, together with long-range conformal symmetry, when $`\gamma <1`$, where $`\gamma `$ is the anomalous dimension of the mass operator. Finally we note that by using the perturbative expansion of $`\gamma `$, chiral symmetry is predicted to be restored above $`N_f^c4N`$, in agreement with a gap equation analysis. ## The Effective Potential In this section we construct an effective potential valid to all orders in the loop expansion and appropriate for the range $`N_f>N`$. The new ingredients are: * Using the full, rather than the one loop, beta function in the trace anomaly saturation. * Taking account of the anomalous dimension of the fermion mass operator. This anomaly-induced effective potential is based on the QCD trace and $`U_A(1)`$ anomalies (see SaSc for more details). We build the potential out of the $`N_f\times N_f`$ complex meson matrix $`M_i^j`$ transforming as the operator $`q_i\stackrel{~}{q}^j`$. So we assign naive mass dimension 3 to $`M_i^j`$. The operator $`q\stackrel{~}{q}`$ acquires an anomalous dimension $`\gamma `$ when quantum corrections are considered and the full dynamical dimension is thus $`3\gamma `$. To make our effective potential capture the low-energy quantum dynamics of the underlying theory, we take $`3\gamma `$ to be the scaling dimension of $`M_i^j`$. The anomalous dimension $`\gamma `$ is of course a function of the coupling $`g`$, which in turn depends on the relevant scale. To build the final meson potential we only use the chirally invarian term $`\mathrm{det}M`$. This is plausible (see section VII of Ref. GJJS ) and would correspond to the ”holonomic” structure which emerges if the potential is considered to arise (toy ; HSS ; Sannino ) from broken super QCD. In the same spirit we take det$`M`$ to have the scaling dimension $`(3\gamma )N_f`$. The potential term we find is SaSc $$V=C\mathrm{\Lambda }^4\left[\frac{\mathrm{\Lambda }^{3N_f}}{\mathrm{det}M}\right]^{\frac{4}{f(g)}}+\mathrm{h}.\mathrm{c}.,$$ (1) where $`C`$ is related to $`A`$ via: $$C=\frac{f(g)}{4e}\mathrm{exp}\left[\frac{4A}{f(g)}\right],$$ (2) and $$f(g)=\frac{\beta (g)}{g^3}16\pi ^2(3\gamma )N_f.$$ (3) Finally we integrate out the $`\eta ^{}`$ field, which can be isolated by setting $$\mathrm{det}M=\|\mathrm{det}M\|e^{i\varphi },$$ (4) where $`\varphi \eta ^{}`$. This is done anticipating that the $`\eta ^{}`$ will be heavy with respect to the intrinsic scale of the theory and the other mesonic degrees of freedom. Now using Eq. (4) we derive the field equation $`\varphi =0`$ which leads to the final potential $$V=2C\mathrm{\Lambda }^4\left[\frac{\mathrm{\Lambda }^{3N_f}}{\|\mathrm{det}M\|}\right]^{\frac{4}{f(g)}}.$$ (5) The shape of this potential is determined by the function $`f(g)`$ Eq. (3). Our interest here is in the range $`N<N_f<(11/2)N`$ where the chiral phase transition is expected to occur. For $`N_f`$ close to $`(11/2)N`$, a weak infrared fixed point will occur. The $`\beta `$ function will be negative and small at all scales and $`\gamma `$ will also be small. Thus $`f(g)`$ will be negative. As $`N_f`$ is reduced, the fixed point coupling increases as does $`\gamma `$. However in the range of interest $`f(g)`$ will remain negative ($`(3\gamma )N_f>(\beta (g)/g^3)16\pi ^2`$). The potential in Eq. (5) may then be written as $$V=+2\|C\|\mathrm{\Lambda }^4\left[\frac{\|\mathrm{det}M\|}{\mathrm{\Lambda }^{3N_f}}\right]^{\frac{4}{\frac{\beta (g)}{g^3}16\pi ^2+(3\gamma )N_f}}.$$ (6) It is positive definite and vanishes with the field $`\|\mathrm{det}M\|`$. ### The Chiral Phase Transition To study the chiral phase transition, we need the combined effective potential $$V_{tot}=V+V_I$$ (7) where $`V_I`$ is a generic potential term not associated with the anomalies. It is instructive, however, to investigate first the extremum properties of the anomaly term (Eq. (6)). Assuming the standard pattern for chiral symmetry breaking $`SU_R(N_f)\times SU_L(N_f)SU_V(N_f)`$, $`M_j^i`$ may be taken to be the order parameter for the transition. For purposes of this discussion, we restrict attention to the vacuum value of $`M_j^i`$, which can be rotated into the form $`M_j^i=\delta _j^i\rho `$, where $`\rho 0`$ is the modulus. Substituting the previous expression in the anomaly induced effective potential gives the following expression: $$V=+2\|C\|\mathrm{\Lambda }^4\left[\frac{\rho }{\mathrm{\Lambda }^3}\right]^{\frac{4N_f}{\frac{\beta (g)}{g^3}16\pi ^2+(3\gamma )N_f}}.$$ (8) Recall that ($`(3\gamma )N_f>(\beta (g)/g^3)16\pi ^2`$) in the range of interest. The first derivative $`V/\rho `$ vanishes at $`\rho =0`$ provided that $`\frac{4N_f}{\frac{\beta (g)}{g^3}16\pi ^2+(3\gamma )N_f}>1`$, a condition that is clearly satisfied. The second derivative, $$\frac{^2V}{\rho ^2}\rho ^{\left[\frac{4N_f}{\frac{\beta (g)}{g^3}16\pi ^2+(3\gamma )N_f}2\right]},$$ (9) also vanishes at $`\rho =0`$ if the exponent in Eq. (9) is positive. The second derivative at $`\rho =0`$ is a positive constant when the exponent vanishes, and it is $`+\mathrm{}`$ for $$\frac{4N_f}{\frac{\beta (g)}{g^3}16\pi ^2+(3\gamma )N_f}2<0.$$ (10) The curvature of $`V_{tot}`$ at the origin is given by the sum of the two terms $`\frac{^2V}{\rho ^2}`$ and $`\frac{^2V_I}{\rho ^2}`$, evaluated at $`\rho =0`$. To proceed further, we assume that the phase transition is governed by an infrared stable fixed point of the gauge theory. We thus set $`\beta (g)=0`$. The curvature of V at the origin is then $`0`$ for $`\gamma >1`$, finite and positive for $`\gamma =1`$, and $`+\mathrm{}`$ for $`\gamma <1`$. The value of $`\gamma `$ depends on the fixed point coupling, which in turn depends on $`N_f`$. As $`N_f`$ is reduced from $`(11/2)N`$, the fixed point coupling increases from $`0`$, as does $`\gamma `$. Assuming that $`\gamma `$ remains monotonic in $`N_f`$, growing to $`1`$ and beyond as $`N_f`$ decreases, there will be some critical value $`N_f^c`$ below which $`\frac{^2V}{\rho ^2}`$ vanishes at the origin. The curvature of $`V_{tot}`$ will then be dominated by the curvature of $`V_I`$ at the origin. For $`N_f=N_f^c`$, there will be a finite positive contribution to the curvature from the anomaly-induced potential. For $`N_f>N_f^c`$ ($`\gamma <1`$), $`V`$ possesses an infinite positive curvature at the origin, suggesting that chiral symmetry is necessarily restored. We will here take the condition $`\gamma =1`$ to mark the boundary between the broken and symmetric phases, and explore its consequences. This condition was suggested in Ref. CG , based on other considerations. We next investigate the behavior of the theory near the transition by combining the above behavior with a simple model of the additional, non-anomalous potential $`V_I`$. We continue to focus only on the modulus $`\rho `$ and take the potential to be a traditional Ginzburg-Landau mass term, with the squared mass changing from positive to negative as $`\gamma 1`$ goes from negative to positive: $`\left(1\gamma \right)\mathrm{\Lambda }^2\rho ^2`$. Additional, stabilizing terms, such as a $`\rho ^4`$ term, could be added but will not affect the qualitative conclusions. The full potential is then $$V_{tot}=2\|C\|\mathrm{\Lambda }^4(\frac{\rho }{\mathrm{\Lambda }^3})^{\frac{4}{3\gamma }}\left(\gamma 1\right)\mathrm{\Lambda }^2\rho ^2.$$ (11) For $`\gamma >1`$ (but $`<3`$), the first term stabilizes the potential for large $`\rho `$, and the potential is minimized at $$<\rho >=\mathrm{\Lambda }^3\left[\frac{\gamma 1}{2\|C\|}\right]^{\frac{1}{\gamma 1}},$$ (12) in the limit $`\gamma 1`$. It describes an infinite order phase transition as $`\gamma 1`$ , in qualitative agreement with the gap equation studies. This behavior would not be changed by the addition of higher power terms ($`\rho ^4,\rho ^6,\mathrm{}`$) to the potential. The curvature of the potential Eq. (11) at the minimum describes a mass associated with the field $`\rho `$. To interpret this mass physically, one should construct the kinetic energy term associated with this field (at least to determine its behavior as a function of $`\gamma 1`$ ). We hence rescale $`\rho `$ to a field $`\sigma `$ via $`\rho =\sigma ^{3\gamma }\mathrm{\Lambda }^\gamma `$ with $`\sigma `$ possessing a conventional kinetic term $`\frac{1}{2}(^\mu \sigma )^2`$. This then leads to the following result for the physical mass $`M_\sigma `$ and $`<\sigma >`$ $$<\sigma >\left[\frac{\gamma 1}{2\|C\|}\right]^{\frac{1}{2(\gamma 1)}}\mathrm{\Lambda },M_\sigma 2\sqrt{6}\|C\|\left[\frac{\gamma 1}{2\|C\|}\right]^{\frac{1}{2(\gamma 1)}}\mathrm{\Lambda }.$$ (13) Likewise, in the presence of the quark mass term we have (see SaSc for details) $$\left[<\sigma >\right]_{\gamma =1}\left[\frac{mN_f\mathrm{\Lambda }}{2\|C\|}\right]^{\frac{1}{2}},\left[M_\sigma \right]_{\gamma =1}2\left[2mN_f\mathrm{\Lambda }\right]^{\frac{1}{2}}.$$ (14) Thus the order parameter $`\sigma `$ for $`\gamma =1`$ vanishes according to the power $`1/2`$ with the quark mass in contrast with an ordinary second order phase transition where the order parameter is expected to vanish according to the power $`1/3`$. Finally we note an important distinction between our effective potential describing an infinite order transition and the Ginzburg-Landau potential describing a second order transition. The latter may be used in both the symmetric and broken phases, describing light scalar degrees of freedom as the transition is approached from either side. Our potential develops infinite curvature at the origin in the symmetric phase, indicating that no light scalar degrees of freedom are formed as the transition is approached from that side. This is in agreement with the conclusions of Ref. ARTW , indicating that as one crosses to the symmetric phase, mesons melt into quarks and gluons and hence the physics is described via only the underlying degrees of freedom. The present effective Lagrangian formalism for describing the chiral/conformal phase transition is close in spirit to the one developed in Ref.MY . By perturbatively (see SaSc ) saturating at two loops the condition $`\gamma <1`$ at the fixed point value of the coupling constant leads to the conclusion that chiral symmetry is restored for $`N_f>N_f^c3.9N`$. ## Conclusions We have explored the chiral phase transition for vector-like $`SU(N)`$ gauge theories as a function of the number of flavors $`N_f`$ via an anomaly induced effective potential. The effective potential was constructed by saturating the trace and axial anomalies. It depends on the full beta function and anomalous dimension of the quark-mass operator. We showed that the anomaly induced effective potential for $`N_f>N`$ is positive definite and vanishes with the field $`M_i^j`$. We then investigated the stability of the potential at the origin, and discovered that the second derivative is positive and divergent when the underlying $`\beta `$ function and the anomalous dimension of the quark-mass operator satisfy the relation of Eq. (10). We took this to be the signal for chiral restoration. With conformal symmetry being restored along with chiral symmetry (due to the $`\beta `$ function vanishing at an infrared fixed point), the criticality relation becomes a constraint on the anomalous dimension of the quark-mass operator: $$\gamma <1.$$ (15) To convert this inequality into a condition for a critical number of flavors, we used the perturbative expansion of the anomalous dimension evaluated at the fixed point, deducing that chiral symmetry is restored for $`N_f4N`$, in agreement with gap equation studies. The core of this talk is the proposal that the chiral/conformal phase transition, suggested by gap equation studies to be continuous and infinite order, may be described by an effective potential whose form is dictated by the trace and axial anomalies of the underlying $`SU(N)`$ gauge theory. ###### Acknowledgements. I am very happy to thank Joseph Schechter for sharing the work on which this talk is based and Thomas Appelquist for enlightening discussions. The work has been partially supported by the US DOE under contract DE-FG-02-92ER-40704.
no-problem/9907/gr-qc9907013.html	ar5iv	text	# 1 The (extended) Penrose diagram describing the evaporating black hole remaining the static wormhole at the end state. In this figure, 𝑦⁺=𝑥⁺+𝐴 and 𝑦⁻=𝑥⁻+𝐵. The gap between the horizon of the evaporating black hole and the throat of the wormhole would be caused by the thunderpop at the intersection point (𝑥⁺ᵢₙₜ,𝑥⁻ᵢₙₜ). Two throat curves of the wormhole are identified in the opposite direction. Wormhole as the end state of two-dimensional black hole evaporation Sung-Won Kim Department of Science Education, Ewha Women’s University, Seoul 120-750, Korea Hyunjoo Lee Department of Physics, Ewha Women’s University, Seoul 120-750, Korea ABSTRACT We present a specific two-dimensional dilaton gravity model in which a black hole evaporates leaving a wormhole at the end state. As the black hole formed by infalling matter in a initially static spacetime evaporates by emitting Hawking radiation, the black hole singularity that is initially hidden behind a timelike apparent horizon meets the shrinking horizon. At this intersection point, we imposed boundary conditions which require disappearance of the black hole singularity and generation of the exotic matter which is the source of the wormhole as the end state of the black hole. These, of course, preserve energy conservation and continuity of the metric. E-mail address : sungwon@mm.ewha.ac.kr E-mail address : hyunjoo@mm.ewha.ac.kr Hawking’s discovery that black holes radiate thermally was the beginning of applications of quantum mechanics in black hole physics. But this discovery has raised the information problem: according to Hawking’s calculation, when a black hole formed in collapse of a pure state evaporates, the resulting outgoing state is approximately thermal and in particular is a mixed state. This conflicts with the ordinary laws of quantum mechanics which always preserve purity. This disagreement can be avoided by introducing the final geometry resulting from black hole evaporation to preserve purity. In this paper we present a specific two-dimensional dilaton gravity model in which a black hole evaporates leaving a wormhole as the end state. Recently Hayward proposed that black holes and wormholes are interconvertible. In particular, if a wormhole’s negative-energy generator fails or the negative-energy source is overwhelmed by normal ordinary positive-energy matter, it will become a black hole, and a wormhole could be constructed from a suitable black hole by irradiating it with negative energy. The classical two-dimensional Callan-Giddings-Harvey-Strominger(CGHS) action is $`S_{\mathrm{cl}}`$ $`=`$ $`S_\mathrm{G}+S_\mathrm{M}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle d^2x\sqrt{g}\left(e^{2\varphi }\left[R^{(2)}+4(\varphi )^2+4\lambda ^2\right]\frac{1}{2}\underset{i=1}{\overset{N}{}}(f_i)^2\right)},`$ where $`\varphi `$ is a dilaton field, $`R^{(2)}`$ is the 2D Ricci scalar, $`\lambda `$ is a positive constant, $``$ is the covariant derivative, and the $`f_i`$ are $`N`$ matter (massless scalar) fields. This action admits vacuum solutions, static black hole solutions, dynamical solutions describing the formation of a black hole by collapsing matter fields and wormhole solutions as we will see later. To see one-loop quantum effects and back reaction, one can use the trace anomaly, $`<T_\mu ^\mu >=(\mathrm{}/24)R^{(2)}`$, for massless scalar fields in two dimensions and the Polyakov-Liouville action, $$S_{\mathrm{PL}}=\frac{\mathrm{}}{96\pi }d^2x\sqrt{g(x)}d^2x^{}\sqrt{g^{}(x^{})}R^{(2)}(x)G(x,x^{})R^{(2)}(x^{}),$$ (2) for which $$<T^{\mu \nu }>=\frac{2\pi }{\sqrt{}g}\frac{\delta }{\delta g^{\mu \nu }}S_{\mathrm{PL}},$$ (3) where $`G(x,x^{})`$ is a Green function for $`^2`$. Here we take the large-$`N`$ limit, in which $`\mathrm{}`$ goes to zero while $`N\mathrm{}`$ is held fixed. In that limit the quantum corrections for the gravitational and dilaton fields are negligible, and one needs to take into account only the quantum corrections for the matter fields. The one-loop effective action is then $`S_{(1)}=S_{\mathrm{cl}}+NS_{\mathrm{PL}}`$. In oder to find analytic solutions including semiclassical corrections, one can modify the action as in Refs. -. We use the model modified from the original CGHS model by Bose, Parker and Peleg. They added to the classical action (S0.Ex1) a local covariant term of one-loop order: $$S_{\mathrm{corr}}=\frac{N\mathrm{}}{24\pi }d^2x\sqrt{g}[(\varphi )^2\varphi R^{(2)}].$$ (4) Now the total modified action including the one-loop Polyakov-Louville term is $$S=S_{\mathrm{cl}}+S_{\mathrm{corr}}+NS_{\mathrm{PL}}.$$ (5) We use the (extended) null coordinates $`x^\pm =x^0\pm x^1`$ (the both coordinates $`x^+`$ and $`x^{}`$ cover the entire range($`\mathrm{},+\mathrm{}`$)) and the conformal gauge $`g_{++}=g_{}=0,`$ $`g_+=\frac{1}{2}e^{2\beta }`$. We can choose $`\varphi (x^+,x^{})=\beta (x^+,x^{})`$ in analyzing the equations of motion. In the conformal gauge the equations of motion derived from $`S`$ are the same as the classical ones $`_{x^+}_x^{}(e^{2\beta (x^+,x^{})})`$ $`=`$ $`_{x^+}_x^{}(e^{2\varphi (x^+,x^{})})=\lambda ^2,`$ $`_{x^+}_x^{}f_i(x^+,x^{})`$ $`=`$ $`0,`$ (6) while the constraints are modified by nonlocal terms $`t_\pm (x^\pm )`$ arising from the Polyakov-Liouville action. In conformal gauge, one can use the trace anomaly of $`N`$ massless scalar fields $`f_i`$ to obtain $`<T_+^f>=\kappa _{x^+}_x^{}\beta ,`$ where $`\kappa =N\mathrm{}/12`$, and integrate the equation $`^\mu <T_{\mu \nu }^f>=0`$ to get the quantum corrections to the energy-momentum tensor of the $`f_i`$ matter fields: $$<T_{\pm \pm }^f>=\kappa [_{x^\pm }^2\beta (_{x^\pm }\beta )^2t_\pm (x^\pm )],$$ (7) where $`t_\pm (x^\pm )`$ are integration functions determined by boundary conditions. And the modified constraints $`\frac{\delta S}{\delta g^{\pm \pm }}=0`$ become $$_{x^\pm }^2(e^{2\varphi (x^+,x^{})})(T_{\pm \pm }^f)_{\mathrm{cl}}+\kappa t_\pm (x^\pm )=0,$$ (8) where $`(T_{\pm \pm }^f)_{\mathrm{cl}}=\frac{1}{2}_{i=1}^N(_{x^\pm }f_i)^2`$ is the classical contribution to the energy-momentum tensor of the $`f_i`$ matter fields. For a given classical matter distribution and a given $`t_\pm (x^\pm )`$ one finds the solution for the equations of motion (S0.Ex2) with constraints (8), $`e^{2\varphi }`$ $`=`$ $`e^{2\beta }=\lambda ^2x^+x^{}{\displaystyle ^{x^+}}𝑑x_2^+{\displaystyle ^{x_2^+}}𝑑x_1^+[(T_{++}^f)_{\mathrm{cl}}\kappa t_+(x_1^+)]`$ (9) $`{\displaystyle ^x^{}}𝑑x_2^{}{\displaystyle ^{x_2^{}}}𝑑x_1^{}[(T_{}^f)_{\mathrm{cl}}\kappa t_{}(x_1^{})]+a_+x^++a_{}x^{}+b,`$ where $`a_\pm `$ and $`b`$ are constants. In the choice $`(T_{\mu \nu }^f)_{\mathrm{cl}}=0`$ and $`t_\pm (x^\pm )=a_\pm =b=0`$, it means the linear dilaton flat spacetime solution. It also has static black hole solutions corresponding to choice $`(T_{\mu \nu }^f)_{\mathrm{cl}}=0`$, $`t_\pm (x^\pm )=a_\pm =0`$ and $`b=M/\lambda `$. In order to find the solution corresponding to the Minkowski vacuum asymptotically, one can use Eq.(7) to find the solution for which $`<T_{\pm \pm }^f(\sigma ^\pm )>=0`$ in flat coordinates $`\sigma ^\pm `$. The functions $`t_\pm (x^\pm )`$ are determined by imposing appropriate boundary conditions that the metric is flat, such that $`\beta `$ and its derivatives vanish in the asymptotically flat coordinates $`\sigma ^\pm `$. Then we get $$<T_{\pm \pm }^f(\sigma ^\pm )>\|_{\mathrm{boundary}}=\kappa t_\pm (\sigma ^\pm )=0,$$ (10) and $$t_\pm (x^\pm )=\left(\frac{\sigma ^\pm }{x^\pm }\right)t_\pm (\sigma ^\pm )\frac{1}{2}D_{x^\pm }^S[\sigma ^\pm ]=\frac{1}{(2x^\pm )^2},$$ (11) where $`D_y^S[z]`$ is the Schwarzian derivative $`D_y^S[z]=_y^3z/(_yz)\frac{3}{2}(_y^2z/_yz)^2`$ and we use the fact that the Minkowski vacuum corresponds to Eq.(10). Thus we find that the asymptotically Minkowski-vacuum solution is $$e^{2\varphi }=e^{2\beta }=\lambda ^2x^+x^{}\frac{\kappa }{4}\mathrm{ln}\|\lambda ^2x^+x^{}\|+C,$$ (12) where $`C`$ is a constant. This solution has two apparent horizons in the extended coordinates. Next we turn to the dynamical scenario in which the spacetime is initially described by one of the static solutions in Eq.(12), and in which the black holes are formed by collapsing matter fields, particularly the simple shock wave of infalling matter described by $`(T_{++}^f)_{\mathrm{cl}}=(M/\lambda x_0^+)\delta (x^+\pm x_0^+)`$, where $`x_0^+>0`$, and $`(T_{}^f)_{\mathrm{cl}}=0`$. Then we find the solution $$e^{2\varphi }=e^{2\beta }=\lambda ^2x^+x^{}\frac{\kappa }{4}\mathrm{ln}\|\lambda ^2x^+x^{}\|\frac{M}{\lambda x_0^+}(x^+x_0^+)\mathrm{\Theta }(x^+\pm x_0^+)+C,$$ (13) where $`\mathrm{\Theta }(x^+\pm x_0^+)`$ $`=`$ $`0x_0^+x^+x_0^+`$ (14) $`=`$ $`1\mathrm{elsewhere}.`$ For all values of $`M`$ and $`C`$, the solution after shock wave is $$e^{2\varphi }=e^{2\beta }=\lambda ^2x^+(x^{}+\mathrm{\Delta })\frac{\kappa }{4}\mathrm{ln}\|\lambda ^2x^+x^{}\|+\frac{M}{\lambda }+C,$$ (15) where $`\mathrm{\Delta }=M/(\lambda ^3x_0^+)`$. The black hole singularity curve is $$\lambda ^2x_s^+(x_s^{}+\mathrm{\Delta })\frac{\kappa }{4}\mathrm{ln}\|\lambda ^2x_s^+x_s^{}\|+\frac{M}{\lambda }+C=0,$$ (16) and an apparent horizon, defined by $`_{x^+}e^{2\varphi }=0`$, is $$\lambda ^2x_h^+(x_h^{}+\mathrm{\Delta })=\frac{\kappa }{4}.$$ (17) When the apparent horizon is formed, the black hole starts radiating. At future null infinity $`^+`$ one can calculate Hawking radiation in the asymptotically flat coordinates $`\widehat{\sigma }^\pm `$, defined by $`\lambda \widehat{\sigma }^+=\mathrm{ln}\|\lambda x^+\|`$ and $`\lambda \widehat{\sigma }^{}=\mathrm{ln}\|\lambda (x^{}+\mathrm{\Delta })\|`$, $$<T_{}^f(\widehat{\sigma }^\pm )>\|__+=\frac{\kappa \lambda ^2}{4}\left[1\frac{1}{(1+\lambda \mathrm{\Delta }e^{\lambda \widehat{\sigma }^{}})^2}\right].$$ (18) Initially the singularity is behind apparent horizon, but as the black hole evaporates by emitting Hawking radiation the apparent horizon shrinks and eventually meets the singularity at $`(x_{\mathrm{int}}^+,x_{\mathrm{int}}^{})`$, $`x_{\mathrm{int}}^+={\displaystyle \frac{1}{\lambda ^2\mathrm{\Delta }}}\left[\pm \mathrm{exp}\left({\displaystyle \frac{4(\pm M+\lambda C)}{\kappa \lambda }}+1\right){\displaystyle \frac{\kappa }{4}}\right],`$ $`x_{\mathrm{int}}^{}=\mathrm{\Delta }\left[1{\displaystyle \frac{\kappa }{4}}\mathrm{exp}\left({\displaystyle \frac{4(\pm M+\lambda C)}{\kappa \lambda }}1\right)\right]^1,`$ (19) where the upper signs do when $`x^+x^{}<0`$ and the lower when $`x^+x^{}>0`$. The singularity become naked after the singularity and the apparent horizon have merged. The future evolution is not uniquely determined unless boundary conditions are imposed at the naked singularity. Here we consider the solution to the future of the point $`(x_{\mathrm{int}}^+,x_{\mathrm{int}}^{})`$; at the intersection point the evaporating black hole matches to a stable wormhole keeping continuous metric and energy conservation. We find the boundary conditions that match the solution (15) continuously to a static wormhole solution. Before this matching, we present wormhole solutions in the (1+1)-dimensional dilaton gravity. The equations of motion from the classical action (S0.Ex1) are $`{\displaystyle \frac{2}{\pi }}e^{2\varphi }\left[_\mu _\nu \varphi +g_{\mu \nu }\left(\left(\varphi \right)^2^2\varphi \lambda ^2\right)\right]T_{\mu \nu }`$ $`=`$ $`0`$ (20) $`e^{2\varphi }\left[R+4\lambda ^2+4^2\varphi 4(\varphi )^2\right]`$ $`=`$ $`0,`$ (21) where the first equation is derived from variation of the metric and the second is the dilaton equation of motion. To find the traversable wormhole solution of this theory we first introduce the spacetime metric $$ds^2=e^{2\mathrm{\Phi }(r)}dt^2+\left(1\frac{b(r)}{r}\right)^1dr^2,$$ (22) where $`\mathrm{\Phi }(r)`$ is redshift function which determines the structure of the wormhole and defined by $$g_{tt}=e^{2\mathrm{\Phi }},$$ (23) and $`b(r)`$ is the shape function that relates to the proper radial distance $`l(r)`$ from wormhole throat by $$dl/dr=\pm \left(1b/r\right)^{1/2}.$$ (24) In order for the spatial geometry to tend to an appropriate asymptotically flat limit, $`lim_r\mathrm{}b(r)`$ must be finite. By comparison with the Schwarzschild metric this implies that the mass of the wormhole, as seen from spatial infinity, is given by $`lim_r\mathrm{}b(r)=2GM_\mathrm{w}`$, where $`M_\mathrm{w}`$ is ADM mass of the wormhole. To produce a traversable wormhole several general constraints on $`b(r)`$ and $`\mathrm{\Phi }(r)`$ are required. The first constraint is that the spatial geometry must have wormhole shape, i.e., two flat regions and a narrow one. Thus throat is at minimum of $`r=b=b_0`$, $`1b/r0`$ throughout spacetime, and $`b/r0`$ as $`l\pm \mathrm{}`$ (asymptotically flat regions of two universes) so $`r\|l\|`$. This constraint means that the embedding surface flares outward, forcing $`(d^2r/dz^2)>0`$ so that $`r(z)`$ is a minimum at the throat in two-dimensional embedding Euclidean space with coordinates $`z`$ and $`r`$. Consequently, we have $$\frac{d^2r}{dz^2}=\frac{bb^{}r}{2b^2}>0.$$ (25) Secondly, there should be no horizons or singularities since a horizon would prevent two-way travel through the wormhole, so $`\mathrm{\Phi }`$ is everywhere finite and $`\mathrm{\Phi }0`$ as $`l\pm \mathrm{}`$. The last constraint is following: the matter and field that generate the spacetime curvature for the wormhole must have a physically reasonable stress-energy tensor with non-zero components $`T_{\widehat{t}\widehat{t}}`$ and $`T_{\widehat{r}\widehat{r}}`$. In the proper reference frame of a set of observers who remain always at rest in the coordinate system $`𝐞_{\widehat{t}}`$ $`=`$ $`e^\mathrm{\Phi }𝐞_t`$ $`𝐞_{\widehat{r}}`$ $`=`$ $`(1b/r)^{1/2}𝐞_r,`$ (26) the non-zero components of the stress-energy tensor are $`T_{\widehat{t}\widehat{t}}=\rho =`$ density of mass-energy and $`T_{\widehat{r}\widehat{r}}=\tau =`$ radial tension. In this basis the metric coefficients take on the Minkowskian forms, $$g_{\widehat{\alpha }\widehat{\beta }}=𝐞_{\widehat{\alpha }}𝐞_{\widehat{\beta }}=\eta _{\widehat{\alpha }\widehat{\beta }}\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right).$$ (27) The spacetime metric (22) can be written like followings by introducing (extended) light-cone coordinates $`y^\pm =y^0\pm y^1`$ ( $`y^\pm `$ also cover ($`\mathrm{},+\mathrm{}`$) ), and by choosing conformal gauge $`g_{\mu \nu }=e^{2\beta }\eta _{\mu \nu }`$, $`ds^2`$ $`=`$ $`e^{2\mathrm{\Phi }}[dt^2dr_{}^{}{}_{}{}^{2}]`$ (28) $`=`$ $`e^{2\beta }dy^+dy^{}`$ where $`r^{}=e^\mathrm{\Phi }\left(1\frac{b}{r}\right)^{\frac{1}{2}}𝑑r`$. As $`r`$ grows very large, $`r^{}\mathrm{}`$. On the other hand, when $`r`$ approaches $`b=b_0`$ then $`r^{}0`$ since $`\mathrm{\Phi }`$ is everywhere finite and we can set $`dr^{}\mathrm{}`$ at $`r^{}=0`$. We then have the nonzero components of energy-momentum tensor in (extended) light-cone coordinates, $`T_+`$ $`=`$ $`{\displaystyle \frac{t}{y^+}}{\displaystyle \frac{t}{y^{}}}T_{tt}+{\displaystyle \frac{r}{y^+}}{\displaystyle \frac{r}{y^{}}}T_{rr}`$ $`=`$ $`e^{2\beta }(\rho +\tau )`$ $`T_{++}`$ $`=`$ $`{\displaystyle \frac{t}{y^+}}{\displaystyle \frac{t}{y^+}}T_{tt}+{\displaystyle \frac{r}{y^+}}{\displaystyle \frac{r}{y^+}}T_{rr}`$ $`=`$ $`e^{2\beta }{\displaystyle \frac{y^{}}{y^+}}(\rho \tau )`$ $`T_{}`$ $`=`$ $`{\displaystyle \frac{t}{y^{}}}{\displaystyle \frac{t}{y^{}}}T_{tt}+{\displaystyle \frac{r}{y^{}}}{\displaystyle \frac{r}{y^{}}}T_{rr}`$ $`=`$ $`e^{2\beta }{\displaystyle \frac{y^+}{y^{}}}(\rho \tau ).`$ And the conservation law of energy-momentum tensor, $`^\mu T_{\mu \nu }=0`$, gives $$\rho e^{2\beta }=\frac{c}{\|\lambda ^2y^+y^{}\|},$$ (30) where $`c`$ is a constant. In conformally invariant case ($`T_+=0`$), the equations of motion become $`{\displaystyle \frac{2}{\pi }}e^{2(\varphi +\beta )}\left[_{y^+}_y^{}\varphi +2_{y^+}\varphi _y^{}\varphi +{\displaystyle \frac{1}{2}}\lambda ^2e^{2\beta }\right]`$ $`=`$ $`\rho +\tau =0`$ (31) $`e^{2\varphi }\left[_{y^+}_y^{}\beta 2_{y^+}_y^{}\varphi +2_{y^+}\varphi _y^{}\varphi +{\displaystyle \frac{1}{2}}\lambda ^2e^{2\beta }\right]`$ $`=`$ $`0.`$ (32) Since we have gauge fixed $`g_{++}`$ and $`g_{}`$ to zero, we should impose their equations of motion as constraints in conformally invariant case, $`{\displaystyle \frac{2}{\pi }}e^{2\varphi }(2_{y^+}\beta _{y^+}\varphi _{y^+}^2\varphi )`$ $`=`$ $`2\rho e^{2\beta }{\displaystyle \frac{y^{}}{y^+}}`$ (33) $`{\displaystyle \frac{2}{\pi }}e^{2\varphi }(2_y^{}\beta _y^{}\varphi _y^{}^2\varphi )`$ $`=`$ $`2\rho e^{2\beta }{\displaystyle \frac{y^+}{y^{}}}.`$ (34) So we get the solution determined in choice of $`\varphi =\beta `$ as $$e^{2\varphi }=e^{2\beta }=C^{}+C_0\mathrm{ln}\|\lambda ^2y^+y^{}\|+C_1y^++C_2y^{}\lambda ^2y^+y^{},$$ (35) where $`C^{},C_0,C_1`$ and $`C_2`$ are constants and we can choose $`c`$ in Eq.(30) so that $`C_0=\frac{2\pi c}{\lambda ^2}=\frac{\kappa }{4}`$. For a static wormhole the constants $`C_1`$ and $`C_2`$ should be zero, $$e^{2\varphi }=e^{2\beta }=C^{}\frac{\kappa }{4}\mathrm{ln}\|\lambda ^2y^+y^{}\|\lambda ^2y^+y^{}.$$ (36) Because the wormhole has forbidden range ($`r<b_0`$) we have to restrict the ranges of $`y^+`$ and $`y^{}`$ such as $`\|\lambda ^2y^+y^{}\|1`$. In the regions of $`\lambda ^2y^+y^{}1`$, we have to fix this solution with $`C^{}>1`$ and $`\frac{\kappa }{4}<1`$ since wormhole solution has no singularities and horizons. In the other regions, $`C^{}>1`$ is required from the same reason. In this wormhole solution, one can let $`C^{}=\frac{M_\mathrm{w}}{\lambda }`$ like the black hole case. Since the physical meaning of the wormhole solution is better in coordinates where the metric is asymptotically constant on $`_R^+`$, we set that $`\lambda y^+`$ $`=`$ $`\pm e^{\lambda \sigma ^+}`$ $`\lambda y^{}`$ $`=`$ $`\pm e^{\lambda \sigma ^{}},`$ (37) where $`\sigma ^\pm =t\pm r^{}`$. These constraints preserve the conformal gauge and gives $$e^{2\beta }=C^{}\frac{\kappa }{2}\lambda r^{}+e^{2\lambda r^{}}.$$ (38) Thus, with the proper asymptotic flatness, $`e^{2\mathrm{\Phi }(r^{})}=C^{}e^{2\lambda r^{}}{\displaystyle \frac{\kappa }{2}}\lambda r^{}e^{2\lambda r^{}}+1`$ (39) $`\rho (r^{})=\tau (r^{})=c[C^{}e^{2\lambda r^{}}{\displaystyle \frac{\kappa }{2}}\lambda r^{}e^{2\lambda r^{}}+1]`$ (40) As $`r`$ grows very large, i.e., $`r^{}\mathrm{}`$, $`\mathrm{\Phi }=0`$ and $`\rho =c`$. And when $`r`$ approaches $`b=b_0`$ then $`r^{}0`$ as we mentioned before, so at the throat $`\rho `$ $`=`$ $`c(C^{}+1)\rho _0`$ (41) $`\mathrm{\Phi }`$ $`=`$ $`{\displaystyle \frac{1}{2}}\mathrm{ln}(C^{}+1),`$ (42) where $`c<0`$ so that the constant $`\rho _0`$ is a large negative value, since the exotic matter that violates the null energy condition would be concentrated near the throat for almost traversable wormhole. ¿From a classical perspective, violation of the null energy condition is not permitted. However we know that some quantum effects lead to measurable experimentally verified violations of the null energy condition. In particular, Hawking evaporation violates the area increase theorem for classical black holes. This implies that the quantum processes underlying the Hawking evaporation process must also induce a violation of the input assumption (the null energy condition - the only input assumption that seems weak in quantum violation) used in proving the classical area increase theorem. Now we give the boundary conditions that match the evaporating black hole solution (15) to the static wormhole solution (36) continuously at the intersection points $`(x_{\mathrm{int}}^+,x_{\mathrm{int}}^{})`$. The amount of energy $`E_{\mathrm{rad}}`$ radiated by the black hole up to the curve $`x^{}=x_{\mathrm{int}}^{}`$ is calculated as $`E_{\mathrm{rad}}`$ $`=`$ $`{\displaystyle _{\mathrm{}}^{\widehat{\sigma }_{\mathrm{int}}^{}}}<T_{}^f(\widehat{\sigma }^{})>𝑑\widehat{\sigma }^{}`$ (43) $`=`$ $`M+\lambda C{\displaystyle \frac{\kappa \lambda }{4}}\left[\mathrm{ln}\left({\displaystyle \frac{\kappa }{4}}\right)1\right]{\displaystyle \frac{\kappa \lambda \mathrm{\Delta }}{4x_{\mathrm{int}}^{}}}.`$ The ADM mass of the dynamical solution (15) (relative to the reference solution with $`C=C_0`$) is $`M_{\mathrm{ADM}}=M+\lambda (CC_0)`$. Thus we see that the unradiated mass $`\delta M`$ remaining as $`x^{}x_{\mathrm{int}}^{}`$ is $`\delta M`$ $`=`$ $`M_{\mathrm{ADM}}E_{\mathrm{rad}}`$ (44) $`=`$ $`{\displaystyle \frac{\kappa \lambda }{4}}\left[\mathrm{ln}\left({\displaystyle \frac{\kappa }{4}}\right)1\right]\lambda C_0+{\displaystyle \frac{\kappa \lambda \mathrm{\Delta }}{4x_{\mathrm{int}}^{}}}.`$ ¿From the constraint equations (8) we find the thunderpop $`[T_{}^f(\widehat{\sigma }^{})]_{\mathrm{cl}}`$ $`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \underset{i=1}{\overset{N}{}}}(_{}f_i)^2`$ $`=`$ $`{\displaystyle \frac{\kappa \lambda \mathrm{\Delta }}{4x_{\mathrm{int}}^{}}}\delta (\widehat{\sigma }^{}\widehat{\sigma }_{\mathrm{int}}^{}).`$ The mass remaining after the shock wave (S0.Ex16) is $`\frac{\kappa \lambda }{4}\left[\mathrm{ln}\left(\frac{\kappa }{4}\right)1\right]`$ when $`C_0=0`$, and would be mass $`M_\mathrm{w}`$ of the wormhole. Thus we try to match at intersection points (S0.Ex5) the solution (15) to one of the static wormhole solution, $$e^{2\varphi }=e^{2\beta }=\lambda ^2(x^++A)(x^{}+B)\frac{\kappa }{4}\mathrm{ln}[\lambda ^2(x^++A)(x^{}+B)]+\frac{\kappa }{4}(\mathrm{ln}\frac{\kappa }{4}1),$$ (46) where $`A=0`$, $`B=\mathrm{\Delta }`$ and only the regions of $`\lambda ^2y^+y^{}=\lambda ^2(x^++A)(x^{}+B)0`$ is considered(see Figure 1). The solution (46) satisfies the constraints for wormhole that we mentioned below Eq.(36) since $`\frac{\kappa }{4}(\mathrm{ln}\frac{\kappa }{4}1)>1`$ when $`\frac{\kappa }{4}<1`$, and has no horizons and singularities in the wormhole spacetime, $`\lambda ^2(x^++A)(x^{}+B)1`$. One get the scalar curvature, $`R=8e^{2\beta }_{x^+}_x^{}\beta `$, which is negative in the entire region of the wormhole spacetime except the asymptotically flat region. The no-positive property of the scalar curvature is required for an appropriate shape of the wormhole. At the throat of the wormhole, $$e^{2\beta }\|_{\mathrm{at}\mathrm{throat}}=e^{2\mathrm{\Phi }}\|_{\mathrm{at}\mathrm{throat}}=1+\frac{\kappa }{4}(\mathrm{ln}\frac{\kappa }{4}1)$$ (47) and the scalar curvature is also finite and negative, $$R\|_{\mathrm{at}\mathrm{throat}}=\lambda ^2\kappa \frac{1\frac{\kappa }{4}\mathrm{ln}\frac{\kappa }{4}\frac{\kappa }{4}}{1+\frac{\kappa }{4}\mathrm{ln}\frac{\kappa }{4}\frac{\kappa }{4}}<0.$$ (48) In asymptotic region, one can get easily $`\mathrm{\Phi }=0`$ and $`R=0`$. In this work we introduced the extended light-cone coordinates and showed that there are boundary conditions in which an evaporating black hole remains a static wormhole as a candidate of end states of a evaporating black hole. These conditions preserve energy conservation and continuity of the metric. This final geometry from the black hole evaporation can avoid the information problem.
no-problem/9907/cond-mat9907103.html	ar5iv	text	# Electronic structure without exchange? ## Abstract The correlation holes for densities of equal and opposite spin around a test electron are determined from the Schrödinger equation with proper boundary conditions. The traditional ”exchange” term follows from the boundary condition which respects a spacial exclusion principle for equal spins. The resulting potential compares reasonably well with standard local density potentials and should simplify extensions towards non-local effects. The Pauli principle excludes multiple occupancy of spin orbitals. When the wave function is a Slater determinant, it gives rise to the concept of exchange, i.e. when two electron indicies in the Coulomb integral are exchanged there will be a perfect annulation of the direct Coulomb term for multiple occupancy . An exchange hole is created around a certain electron, excluding other electrons of the same spin to come too close. When applied to free electron wave functions, this exchange hole has a well-known form, and the resulting exchange potential is in the Hartree-Fock-Slater approximation (HFS) applied to atoms, molecules and solids, which in general are not free electron like. In this way it is possible to reduce the difficult N-body problem to a 1-particle like Schrödinger equation: $$^2\mathrm{\Psi }_i(r)+(V_{ext}(r)+\frac{\rho (r^{})}{rr^{}}d^3r^{}+\mu _{xc}(r))\mathrm{\Psi }_i(r)=ϵ_i\mathrm{\Psi }_i(r)$$ (1) where $`\mathrm{\Psi }_i(r)`$ is the wave function for electron state $`i`$ with energy $`ϵ_i`$, and $`\rho (r)`$ is the electron density. The external potential $`V_{ext}(r)`$ is the attractive potential due to all nuclear charges and the exchange-correlation potential is reduced in a local form, $`\mu _{xc}(r)`$. The density functional (DF) theory in the local density approximation (LDA) showed that this is correct in variational calculations with the electron density as a variable . The exchange hole still serves as a valid pair correlation function between two electrons of the same spin, but the resulting exchange potential is in the LDA Kohn-Sham (KS) potential 2/3 of that in HFS. Correlation is an additional effect that stems from the fact that all electrons repel each other, so that due to the correlated movement of the electrons there will be a reduced Coulomb term in the in the DF-type Schrödinger equation for a one-particle potential . Similarly as the exchange hole for equal spin, there will be a correlation hole also for opposite spin electron density around an electron, but it is much less profound than the exchange hole especially at high density. Whereas the exchange hole goes to zero at the position of the given electron and totally must contain one electron (meaning that Coulomb repulsion must exclude one electronic charge), the correlation hole integrates up to zero, i.e. it represents just a redistribution of charge around a given electron. Furthermore there is no condition for its amplitude at the site of the first electron. The LDA formalism is applied including the corrections due to correlation and spin-polarization, and it has proven to be very successful in electronic structure calculations . More recent corrections due to weak variations of the density, such as in the generalized gradient approximation (GGA) have improved some results of LDA without making the method more difficult . In this work we propose a different derivation of the DF-type potential for exchange and correlation than what is done in HFS and DF theories. The correlation part between electrons of opposite spin has been presented previously in ref. . Here we adapt the formalism to densities of equal spin, i.e. for the dominating part of the potential which is denoted exchange. The term ”exchange” is no longer appropriate, since as we shall see there is only correlation due to Coulomb repulsion between electrons. But there is an important boundary condition which modifies completely the correlation hole for equal or opposite spin. However, the term ”exchange” is so common that we in the following continue to use it for the interaction between electrons of equal spin. The goal is to generalize the method for non-constant densities as in real atoms and solids, but in this work we consider free-electron conditions of constant density. The electron gas parameter $`r_s`$ determines the average radius around a given electron of density $`\rho `$, $`4\pi /3r_s^3\rho =1`$. An electron at some point will on the average have another electron $`r_s`$ away from it, so the Coulomb repulsion is of the order $`1/r_s`$ when using atomic units. The density in a solid is large near to the atoms and $`r_s`$ is small and rarely larger than 2 a.u., i.e. always smaller than the atomic radius. In the near neighborhood of an electron, which is fixed at some position $`r=0`$, other electrons will feel a large repulsive Coulomb potential due to the fixed electron. The amplitude of this potential will even be larger than the mean-field crystal potential that determines the crystal wave function $`\mathrm{\Psi }_i(r)`$ in eq 1 . $`\mathrm{\Psi }_i(r)`$ will not be a good description for the wave function of other electrons near the one fixed at $`r=0`$. This is even more so for electrons of equal spin, where a second electron cannot be at the origin. Therefore, in order to determine the exchange and correlation potential one must consider that $`\mathrm{\Psi }_i(r)`$ is deformed by the Coulomb interaction and the boundary condition for equal spin, around the single electron that we temporarily fix at a local origin. This happens for all surrounding electrons, independent of which shell they belong to, when the Coulomb repulsion excedes the value of the crystal potential at the local origin. For equal spin, this is even more certain, since the boundary condition for r=0 apply to all electrons independent of their relative differences in kinetic energy. First, for electrons of opposite spin and total density $`\rho /2`$ (no spin-polarization) the redistributed charge density can be determined from a Schrödinger equation of type $$^2\varphi (r)+(g/r+\frac{\rho (r^{})}{rr^{}}d^3r^{}+\mu _{xc}(r))\varphi (r)=ϵ\varphi (r)$$ (2) with the boundary condition that the density $`\varphi (r)^2`$ should tend to $`\rho /2`$ at the limit of the correlation hole (at $`r=r_s`$ or so) . In eq. 2, $`g`$ is the effective interaction strength and the following two potential terms are due to the possible Coulomb exchange-correlation interaction among more than one electron within the correlation hole. (It is not necessary to limit the problem to two electrons, one at $`r=0`$ and the other around it). Further in eq. 2 the effective mass is 1/2. The lowest energy $`ϵ`$ is for $`\varphi (r)`$ being an $`s`$-state, i.e. $`\mathrm{}=0`$. Secondly, for electrons of the same spin as the one at the center, one can solve a similar type of equation. However, we need to recall the Pauli principle a second time: Two electrons of the same spin cannot be in the same state, and in particular at a given instant they cannot be at the same place. This puts a boundary condition on the wave function $`\phi (r)`$ for electrons of the same spin; $`\phi (0)`$=0, and it implies solutions of $`\mathrm{}1`$. Furthermore, the electron at the center is unique and cannot be contained in the surrounding cloud, $$(\phi (r)^2\rho /2)d^3r=1$$ (3) Since correlation is allowed even between electrons of the same spin, it is possible that $`\phi (r)^2`$ could be larger than $`\rho /2`$ for some points far from 0, but at the limit of the exchange hole $`\phi (r)^2`$ should tend $`\rho /2`$ as in case of correlation only. In principle, this requires a continous matching of both the amplitude and the derivative, but in this work only one continuity condition is applied. The equation for $`\phi `$ becomes; $$^2\phi (r)+(g/r+\mathrm{}(\mathrm{}+1)/r^2+\frac{\rho (r^{})}{rr^{}}d^3r^{}+\mu _{xc}(r))\phi (r)=ϵ\phi (r)$$ (4) with $`\mathrm{}`$=1 for the lowest energy. The influence of the coupling strength $`g`$ is strong on $`\varphi (r)`$, but only minor on $`\phi (r)`$. The DF value for the xc-energy is $$\epsilon =0.5(\phi _g(r)\rho /2)/rd^3r𝑑g$$ (5) and $$\epsilon =0.5(\varphi _g(r)\rho /2)/rd^3r𝑑g$$ (6) for equal and opposite spin, respectively. The integration over the coupling strength is from 0 to 1, and the potential is found from $$\mu =\frac{d}{d\rho }(\rho \epsilon )$$ (7) These equations are solved for different densities, but some technical points are to be noted. The boundary condition for $`\phi (r_c)`$ is that the derivative at $`r_c`$ is zero, where $`4\pi /3r_c^3\rho =4`$. This makes approximately $`\phi ^2(r_c)=\rho /2`$. Smaller $`r_c`$ gives a very similar $`\mu `$-potential, but the discontinuity at the hole boundary is evident. The same $`r_c`$ is used for the correlation between opposite spins with the boundary condition $`\varphi ^2(r_c)=\rho /2`$. $`\varphi (r)`$ has one maximum (larger than $`\rho `$) before $`r=r_c`$, while $`\phi (r)`$ is an ever increasing function. The results for the coupling strength $`g`$=1/2 are very close to those with the integrated values from 0 to 1. Finally, we ignore the coupling between the two spin densities since it turns out to have a quite small effect on the result (even ignoring terms 3 and 4 in eqs. 2 and 4 give almost the same result). The high density solution for equal spin $`\phi (r)`$ with $`g`$=0, gives a potential value which is about 12 % larger than the KS-value. The reason for this is probably connected with the imposed cut-off radius. The Slater function , which defines the hole in KS-theory, has Friedel-like oscillations outside its first node. A simple exercise will motivate that the present scheme should give the KS-value for interaction $`g`$=0, if the solutions for large $`r`$ are retained. For very large $`r`$ it is simpler to handle solutions going to zero than solutions which are approaching the electron density. Therefore we search the solutions $`j(r)`$ for a fictive positive charge that cancels exactly one electron, and which equals the electron density at $`r`$=0. Instead of eq. 4, we have $$^2j(r)+\mathrm{}(\mathrm{}+1)/r^2j(r)=ϵj(r)$$ (8) when no interaction within the electron cloud is considered. This is the V=0 central potential problem with the known solutions, the Bessel functions $`j_{\mathrm{}}(\sqrt{ϵ}r)`$. The solution $`j_0`$ fulfills the condition at $`r`$=0, but it is not normalizable over all space. Linear combinations with higher-$`\mathrm{}`$ functions fulfill the condition at $`r`$=0, but $`(j_0+j_1)`$ is not normalizable either. The low-$`\mathrm{}`$ combination which fulfills the two conditions is $`(j_0+j_2)`$. The square of this solution is precisely the Slater function, normalized to one electron: $$_0^{\mathrm{}}(j_0+j_2)^2d^3r=1$$ (9) At the first node of $`(j_0+j_2)`$ the normalization is about 0.8, and the potential is about 95 % of its full value, the KS-value. Verification by numerical solutions will be difficult because of the requirement of normalization over all space. But at the same time it would be incorrect to impose a cut-off at the first node, as it will localize the hole too much. Instead we can return to the solutions for the electron density ($`\phi ^2(r)`$ instead of the fictive positive density) and extend the solution beyond the first maximum. By letting the interacting density reach the non-interacting density at $`r_c`$, and normalizing to unity at a larger radius, one finds that some fraction of the hole is beyond the maximum at $`r_c`$, similar to (but still more localized than) the Slater function. This procedure is used with the coupling $`g`$, although it is no longer required that $`\phi ^2\rho /2`$ at some point, because of the Coulomb repulsion. One can summarize the final results in terms of powers of the density, $$\mu =2.1\rho ^{1/3}+0.07\rho ^{1/6}(Ryd)$$ (10) where the major part of the second term is the part from different spin. The first part is still a little larger (6 %) than the KS-value for exchange, $`1.961\rho ^{1/3}`$ $`(Ryd)`$. The effective parameter ($`\alpha `$) in front of the $`\rho ^{1/3}`$-term is shown in fig 1, together with two commonly used LDA potentials. In the case of spin polarization, it is possible to solve the equations above for two different densities. Due to the Hartree and $`\mu _{xc}`$ terms in eqns. 2 and 4, there is a coupling between the two densities, which diminish the tendency for polarization. However, this interaction will also complicate the search for self-consistent solutions, and approximate results can be obtained rather quickly by omitting the two terms completely. The preliminary results give a slightly larger tendency for polarization than standard spin-polarized versions of LDA . The resulting potential is close to, but not better than LDA. However, the results are suffiently promising to continue the search for improved solutions at large $`r`$ and for non-locality. Apart from the problem at large $`r`$ the formalism contains no free parameters or other ad-hoc assumptions. Further, the present formalism helps to understand one difference between HF and DF approaches. The exchange in HF acts selectively between orbitals, so the exchange between core and valence appears different from within valence states, for example. In LDA all electrons have the same status independent of the electronic shell, since the density is made up by all electrons, core and valence. (However, in practise HF and LDA results may come out not too different after convergence.) From the way we look at the electron interactions in the present method, it is natural that all electrons should be equal because of the boundary condition at $`r`$=0. Equal spin electrons have zero probability to be at the same place, and this is independent of the energy of their atomic orbital. For opposite spin electrons the probablity is not zero but significanly reduced, because the Coulomb repulsion is strong and will separate the electrons, independently of their kinetic energy. Selective shell dependent interactions like a self-interaction correction, is not a natural extension of the present approach. Also methods which assign a strong orbital dependent parameter like a Hubbard parameter to some of the electrons, are difficult to understand, because the electron cannot distinguish between the interaction to a special orbital from the rest of the electron density. However, all this reasoning has to be revised if the interaction length ($`r_c`$ roughly), becomes large in comparison to the size of an atom. A large part of the core density may exist as a ”bump” within the radius $`r_c`$, and corrections due to non-locality will be important. Therefore it is essential to correct the LDA as in GGA or for even stronger non-constant density corrections, especially for f-electron systems where self-interaction or Hubbard models give different results than LDA. In conclusion, the most important result is that the hole for equal spins has similar shape as the exchange hole for free electrons. As a cut-off radius is imposed, it will be slightly more localized and give a slightly stronger potential than the exchange hole which has small oscillations far away. The zero amplitude at the origin is the result of a boundary condition, and it is the rigidity of the Schrödinger equation which determines the shape of the exchange hole for equal spins. Further development of the long-range tails of the hole is necessary before applications, and generalization for non-constant density will be of great interest.
no-problem/9907/astro-ph9907056.html	ar5iv	text	# Ab initio galaxy formation ## 1. Introduction The main ideas of hierarchical galaxy formation were set out more than twenty years ago by White & Rees (1978). These authors proposed that galaxy formation is a two stage process. In the first stage, dark matter haloes form by the dissipationless accretion of smaller units and through mergers. The second stage consists of the dissipative condensation of baryons to the centre of dark matter haloes. After the first stars form from this gas, feedback effects can play an important role, regulating star formation and thus controlling the efficiency of galaxy formation in dark matter haloes of different mass. The past decade has seen the development of semi-analytic models to study the ab initio formation and evolution of galaxies, within the framework of the growth of structure in the dark matter (e.g. Kauffmann etal 1993, Cole etal 1994, Somerville & Primack 1998). The dissipationless physics in these models is well understood and has been explored extensively using N-body simulations of gravitational instability (for recent illustrative examples see Jenkins etal 1998 and Ghigna etal 1999). The dissipative processes are, however, not at all well understood. The idea that hot gas would cool radiatively to make galaxies was also proposed in the late 1970s, but numerical simulations of this process with sufficient resolution to identify “galaxies” within cosmological volumes are only now becoming possible (Pearce etal 1999). In the semi-analytic models, the processes of gas cooling, star formation, the attendant feedback and galaxy mergers are described by a set of simple, physically motivated rules. The values of the required parameters are set by comparing the model results with properties of the local galaxy population, such as the field galaxy luminosity function or the Tully-Fisher relation (see Somerville & Primack for a discussion of the various models), to produce a fully specified model with strong predictive capabilities. ## 2. Ab initio galaxy formation The semi-analytic model described by Cole etal (1999) is a development of the one described in earlier papers by the Durham group (e.g. Cole etal 1994, Baugh, Cole & Frenk 1996). In addition to the use of improved algorithms, for example in the Monte-Carlo generation of dark matter halo merger histories, a broader range of physical processes is now modelled, greatly expanding the number of galaxy properties that we can predict. The size of a galactic disk is determined by the conservation of the angular momentum of cooling gas. The gas is assumed to have the same specific angular momentum profile as the dark matter halo, whose spin originates from tidal torques that act during its formation. The size of a bulge is computed by conserving energy when two fragments merge to form the bulge and applying the virial theorem. The condensation of baryonic material at the centre of a dark matter halo alters the structure of the halo, causing a contraction. The resulting size of the disk and bulge components depends upon the self-gravity of the baryons and the gravity of the modified dark matter halo. The chemical enrichment of the baryons is also followed in the new model. Episodes of star formation result in the production of metals that can be transferred to the reservoirs of hot and cold gas within each dark matter halo. Feedback from star formation can reheat some of the cold gas, thus providing a further channel through which to transfer metals to the hot gas component. The calculation of the scale length of the galactic disk and of the metallicity of the cold gas allows us to obtain an optical depth for the disk, and hence to calculate the extinction of starlight by dust. The main property of the local galaxy population that we use to constrain our models is the field galaxy luminosity function. Figure 1 shows two recent determinations of the luminosity function in the $`b_J`$-band. These surveys are taken from the same parent photometric catalogue, but extend to different apparent magnitude limits. The solid line shows the luminosity function of the fiducial model of Cole etal (1999), which is in very good agreement with the measurement by Zucca etal (1997) over a dynamic range of $`10^4`$ in luminosity and $`10^5`$ in space density. The dotted line shows a model in which the feedback arising from star formation is assumed to be a much stronger function of circular velocity, as in the model of Cole etal (1994). Star formation in low circular velocity objects is extremely inefficient in this case, and a reasonable match is obtained to the flat faint end slope of the luminosity function found by Ratcliffe etal (1998). The dashed line shows the slope of the dark matter halo mass function; this slope, $`\alpha 1.8`$, would be obtained for the luminosity function in the absence of feedback and if galaxies merged on the same timescale as their parent dark matter haloes. ## 3. The star formation history of the universe The use of the redshifted Lyman-break spectral feature to isolate candidate high redshift galaxies has proven to be a remarkably successful and efficient way of constructing large samples of galaxies at significant lookback times (Steidel etal 1996 and references therein). The application of this photometric technique to deep images obtained from the ground and to the Hubble Deep Fields has allowed the star formation history of the universe to be constructed (e.g. Madau etal 1996, Steidel etal 1999). Figure 2 shows a subset of the currently available observational estimates of the star formation rate per unit volume, as inferred from the flux density at various wavelengths, indicated by the key in the Figure. The dotted line shows the star formation history predicted by the model of Cole etal (1994), which predates the oldest data points on the Figure by two years (see Figure 16 of Baugh etal 1998). The dashed line shows the star formation history in the fiducial model of Cole etal (1999), estimated from the rest-frame luminosity density at $`1500\AA `$. The solid line shows how this estimate is changed when obscuration by dust is taken into account; this line is the one that should be compared to the data points, which have not been corrected for the effects of dust. The star formation rate per unit volume in the model is reduced by a factor of $`2.5`$ at $`z=4`$ as a result of dust extinction of the $`1500\AA `$ light. The small differences between the dotted line (Cole etal 1994) and the solid line are mainly due to the different parameterisations of the strength of feedback as a function of circular velocity used in the two models. The particular choice of feedback model is motivated by the attempt to reproduce the faint end of the local luminosity function (see Figure 1 and the discussion in the previous Section). The weaker feedback employed by Cole etal (1999) results in a somewhat broader peak in the global star formation rate around $`z23`$, compared with the dotted line, which peaks at $`z1.52`$. However, when expressed in terms of lookback time (as shown by the scale at the top of the Figure), which is the appropriate variable for computing the integrated stellar mass, this difference is in fact quite small. ## 4. The evolution of galaxy clustering Two approaches have been used to obtain predictions for the clustering of galaxies from semi-analytic models. In the first approach, a bias parameter is computed for dark matter halos to relate the amplitude of the correlation function of the haloes to that of the underlying dark matter distribution (Mo & White 1996). The semi-analytic model is then used to populate dark matter haloes with galaxies, and a selection criterion, such as an apparent magnitude limit, is applied to the model galaxies. The bias parameter for the galaxies is obtained by summing the bias of each halo, weighted by its abundance and the number of galaxies it contains that satisfy the selection criterion (e.g. Baugh etal 1999). This approach was used by Baugh etal (1998) to predict successfully that Lyman break galaxies have a correlation length similar to that of bright galaxies at the present day, and much larger than that of the dark matter in any viable model at $`z3`$. Another method is to use the semi-analytic model to populate dark matter haloes in a N-body simulation of hierarchical clustering with galaxies (Kauffman etal 1997, 1999; Governato etal 1998, Benson etal 1999). This approach has the advantage of being able to probe galaxy clustering down to small scales and to include the distortion of the clustering pattern that arises from the peculiar motions of galaxies. Benson etal (1999) find that the clustering of galaxies in the fiducial model of Cole etal (1999) is in very good agreement with the two point correlation function measured for APM galaxies. This is a remarkable result because the measured correlation function is a power law over more than a decade in separation, whereas the correlation function of the dark matter has two inflection points over the same range of scales. Furthermore, on scales around $`1h^1`$Mpc, the amplitude of the dark matter correlation function is higher than that measured for galaxies, implying that galaxies are $`anti`$-biased or less clustered compared to the dark matter. The level of small scale clustering depends sensitively on how dark matter haloes are populated by galaxies, and thus upon the details of the galaxy formation process. The first approach described above has been used to predict the evolution of the correlation length of galaxies, measured in comoving units, with redshift, as shown in Figure 3 (Baugh etal 1999; similar results are found using the N-body technique, Kauffmann etal 1998). The clustering evolution of galaxies is markedly different to that displayed by the dark matter, and furthermore is not well described by the ‘$`ϵ`$-model’ commonly used to interpret such data. The correlation length of galaxies initially decreases to $`z1`$, and then starts to increase again at higher redshifts. The model predicts that galaxies that are bright enough to be seen at high redshift are hosted by the most massive dark matter haloes in place at these redshifts. Such haloes are biased tracers of the dark matter distribution. Traditionally, the angular clustering of galaxies imaged in one band has been quantified in terms of the correlation amplitude measured at a fixed angular scale plotted as a function of the limiting apparent magnitude of the sample. Even at very faint magnitudes, the median redshift of the sample does not increase rapidly and in addition, the clustering signal is diluted by projection effects. A powerful observational technique that is now being pursued by many groups is to take deep images in many different filters and to measure the angular clustering of galaxies selected to lie within a limited baseline in photometric redshift. This method has been applied to study the evolution of galaxy clustering to redshift $`z1`$ (Connolly etal 1998) and beyond using the Hubble Deep Field (e.g. Magliocchetti & Maddox 1998). At present, the fields studied are small and the results may be subject to sample variance. The detection of the dip in the galaxy correlation length shown in Figure 3 would provide evidence for hierarchical galaxy formation, but it may tell us more about the details of the process of galaxy formation than about the underlying cosmology. In any event, the semi-analytic models described here will have an important part to play in the interpretation of such results. ### Acknowledgments. CMB would like to thank the organisers for an enjoyable and informative meeting and for providing assistance to attend the workshop. ## References Baugh, C. M., Cole, S., Frenk, C. S., 1996, MNRAS, 283, 1361 Baugh, C. M., Cole, S., Frenk, C. S., Lacey, C.G., 1998, ApJ, 498, 504 Baugh, C. M., etal, 1999, MNRAS, 305, L21 Benson, A.J., etal, 1999, MNRAS, submitted, astro-ph/9903343 Cole, S., etal, 1994, MNRAS, 271, 781 Cole, S., Lacey, C.G., Baugh, C.M., Frenk, C.S., 1999, in preparation. Connolly, A.J., Szalay, A.S., Brunner, R.J., 1998, ApJ, 499, L125 Ghigna, S., etal, 1999, MNRAS, 300, 146. Governato, F., etal 1998, Nature, 392, 359. Jenkins, A.R., etal (The Virgo Consortium), 1998, ApJ, 499, 20 Kauffmann, G., White, S.D.M., Guiderdoni, B., 1993, MNRAS, 264, 201 Kauffmann, G., Nusser, A., Steinmetz, M., 1997, MNRAS, 286, 795 Kauffmann, G., etal, 1999, MNRAS, 303, 188 Kauffmann, G., Colberg, J.M., Diaferio, A., White, S.D.M., 1998, astro-ph/9809168 Madau, P., etal 1996, MNRAS, 283, 1388 Magliocchetti, M., Maddox, S.J., 1998, astro-ph/9811320 Mo H. J., White S. D. M., 1996, MNRAS, 282, 347 Ratcliffe A., Shanks T., Parker Q. A., Fong R., 1998, MNRAS, 293, 197 Pearce, F.R., etal (The Virgo Consortium) 1999, ApJ, in press, astro-ph/9905160 Somerville R. S., Primack J. R., 1998, astro-ph/9806228 Steidel, C.C., etal, 1996, ApJ, 462, L17 Steidel, C.C., etal, 1999, ApJ, 519, 1. White, S.D.M., Rees, M.J., 1978, MNRAS, 183, 341 Zucca E., et al., 1997, A&A, 326, 477
no-problem/9907/cond-mat9907330.html	ar5iv	text	# Phase Transitions in a Two-Component Site-Bond Percolation Model ## Abstract A method to treat a $`N`$-component percolation model as effective one component model is presented by introducing a scaled control variable $`p_+`$. In Monte Carlo simulations on $`16^3`$, $`32^3`$, $`64^3`$ and $`128^3`$ simple cubic lattices the percolation threshold in terms of $`p_+`$ is determined for $`N=2`$. Phase transitions are reported in two limits for the bond existence probabilities $`p_=`$ and $`p_{}`$. In the same limits, empirical formulas for the percolation threshold $`p_+^c`$ as function of one component-concentration, $`f_b`$, are proposed. In the limit $`p_==0`$ a new site percolation threshold, $`f_b^c0.145`$, is reported. The percolation model goes back to Flory who introduced it in the context of polymer gelation. Since then it has been used in a wide range of approaches and techniques. In standard percolation models either bond or site percolation is dealt with. Site-bond percolation combines the two formulations, dealing with randomly occupied sites (vertices) and randomly existing bonds (open edges) connecting these sites. However in this version of the model only one active component exists, the other sites are considered unoccupied. A further generalization is to consider several components, which was done for site percolation as well as bond percolation by Zallen and called polychromatic percolation. Zallen focused on the coexistence of percolating species in highly connected lattices, giving a criterion for the occurrence of a panchromatic regime where all species percolate. Site-bond percolation using two components was investigated previously by one of us and applied to the question of the nuclear liquid gas phase transition. Site-bond percolation with several species was considered in and an approximate percolation criterion was given. In this Letter, we investigate a two component site-bond percolation model on a simple cubic lattice, focusing on two specific limits which exhibit novel behavior. Let us begin by describing the approach we have taken, in the general case of $`N`$ different component flavors. No assumption concerning topological dimensions or lattice structure is made. We have $`N`$ component concentrations $`f_i`$ with $`_{i=1}^Nf_i=1`$ and different bond probabilities to connect all possible combinations of sites, resulting in $`A_{par}=(N1)+\left(\genfrac{}{}{0pt}{}{N+21}{2}\right)`$ free parameters $`a_i`$. The bonds have been assumed to be directionless, meaning that their probabilities only depend on the species of the sites they are connecting. We now want to know in which region of this $`A_{par}`$-dimensional parameter space an infinite network $`C_{\mathrm{}}`$ of connected bonds occurs, that is, where the probability for a given site to belong to the infinite network, $`p_{\mathrm{}}(\{a_i\})`$, is non-zero. The particular type of a bond shall be irrelevant in order for it to belong to the infinite network. For a system with $`N3`$ components, however, this approach is quite impractical. It would be preferable to be able to reduce the dependence of the order parameter, $`p_{\mathrm{}}`$, to one variable at fixed particle concentrations. We propose one such variable in following the definitions of , generalizing them to $`N`$ components. In analogy to the bond existence probability used in ordinary one component bond percolation models, which gives the bond density in the system under observation, we introduce the scaled control parameter $`p_+`$: $$p_+=\underset{ij=1}{\overset{N}{}}\alpha _{ij}p_{ij}.$$ (1) Here, the $`p_{ij}`$’s denote the probability for a bond to exist between two sites occupied by species $`i\text{and}j`$ and $`\alpha _{ij}=\alpha _{ji}`$ is the probability that any given nearest neighbor edge is one that connects two sites of flavors $`i\text{and}j`$: $$\alpha _{ij}=2f_if_j,$$ (2) with the constraint that $`_{ij}^N\alpha _{ij}=1`$. For our simulation we consider a two component system on $`16^3`$, $`32^3`$, $`64^3`$ and $`128^3`$ simple cubic lattices. For simplicity we shall call one species blue, the other red. We now have four free parameters to vary: The fraction of one of the components, say of the blue sites, $`f_b`$ and three bond activation probabilities: $`p_{bb}`$ for bonds connecting two blue sites, $`p_{rr}`$ for bonds connecting two red sites and $`p_{}`$ for $`b\text{-}r`$-bonds. However we shall set $`p_=p_{bb}=p_{rr}`$, introducing a symmetry in the system. This is motivated by considerations of, for example, isospin symmetry, where the $`e^+e^+`$ and $`e^{}e^{}`$ interactions are identical. Eqs. (1) and (2) now read $$p_+=\alpha _{}p_{}+\alpha _=p_=,$$ (3) with $`\alpha _==(1\alpha _{})`$, and $$\alpha _{}=2f_b(1f_b),$$ (4) respectively, where we have replaced the double indices by a more intuitive notation for only two components. Again the question is in which region of the three dimensional $`p_=\text{}p_{}\text{}f_b`$ space an infinite network of bonds appears in the lattice. We defer the question of the concentration dependence to later and for the moment set $`f_b`$ to some fixed value, which shall, for now, be $`f_b=0.5`$. In the simulation the lattice is populated at random, without correlations, according to $`f_b`$ and bonds are formed for varied values of $`p_=,p_{}[0,1]`$ using a Monte Carlo algorithm. The resulting cluster structure is analyzed using a cluster-find-algorithm described in and $`p_{\mathrm{}}(p_=,p_{})`$ is recorded. As always, a cluster is defined as a set of vertices connected by open edges. In Fig. 1 we show $`p_{\mathrm{}}`$ as a function of the two control parameters, $`p_=\text{and}p_{}`$, at a concentration of $`f_b=0.5`$. We can see that $`p_{\mathrm{}}`$ changes from 0 (front corner) to 1 (back corner), with a critical line of a second order phase transition in the $`(p_=,p_{})`$-plane. We now follow our previous consideration and analyze the same data in terms of the scaled control parameter $`p_+`$, the result of which is displayed in Fig. 2. Two distinct branches are seen, both in the shape of a second order phase transition. Figure 2 first suggests that $`p_+`$ is a good control parameter, as we only have these two universal scaled curves, but furthermore that another transition seems to take place in the system. Analysis of the data shows that the ’upper’ branch constitutes of points with both $`p_=`$ and $`p_{}`$ non-zero, whereas all points with $`(p_==0,p_{}0)`$ and $`(p_=0,p_{}=0)`$ fall on the ’lower’ branch. For other values of $`f_b`$ the same behavior is found, but then the two curves for the zero-limits in $`p_=`$ or $`p_{}`$ are not the same. Also shown in Fig. 2 are the expectations from one component bond percolation theory, $`p_{\mathrm{}}(p_+p_+^c)^\beta `$, (smooth lines). For the finite $`(p_=,p_{})`$ regime the critical value $`p_+^c`$ has the same numerical value as the bond existence probability in one component bond percolation (aside from a small difference due to finite size effects), $`p_+^c=0.251\pm 0.002`$. As long as both bond-types are active, the system under observation here and the one component model show an identical phase transition behavior, which is consistent with the findings presented in . In the zero-limits of $`p_=`$ or $`p_{}`$, however, $`p_+^c`$ is shifted to $`p_+^c=0.280\pm 0.002`$. The critical exponent $`\beta =0.41`$ from one component bond percolation theory is the same in both cases presented here, as shown in the double logarithmic plot in the inset of Fig. 2. What causes the change of $`p_+^c`$ in the limits $`p_=0`$ and $`p_{}0`$? We restrict ourselves to a discussion of $`p_{}0`$, as the same line of arguments applies in the other limit. For simplicity we first choose $`p_==1`$. Then setting $`p_{}=0`$ corresponds to a lattice in which all available $`bb`$\- and $`rr`$-edges are open, but all $`br`$-edges are closed. There will be an infinite network present, as usual identified with the biggest percolating cluster (with two components each above the site percolation threshold being present we could have two percolating clusters), with $`p_{\mathrm{}}`$ given by $$p_{\mathrm{}}(p_{}=0)=f_b\underset{s=1}{\overset{\mathrm{}1}{}}n_s^bs,$$ (5) where $`n_s^b`$ is the number of $`b`$-clusters of size $`s`$ and the upper limit in the sum is meant to indicate that the infinite cluster is excluded. Letting $`p_{}=ϵ=1/N_{edges}`$ introduces on average one $`br`$-bond in the lattice. This bond might, with some small probability, which is related to the number of perimeter edges of all finite clusters, connect two finite clusters. With a noticeably higher probability however, it will open an edge that connects the infinite cluster to the biggest $`r`$-cluster, $`C_{max}^r`$, with $`\|C_{max}^r\|\|C_{\mathrm{}}\|`$. The infinite cluster can now consist of both $`b`$\- and $`r`$-sites and $`p_{\mathrm{}}`$ reads $$p_{\mathrm{}}(p_{}=ϵ)=1\underset{s=1}{\overset{\mathrm{}1}{}}n_ss,$$ (6) which, by rewriting $`f_b`$ in Eq. 5 as $`f_b=1_{s=1}^{max}n_s^rs`$, leads to a difference in $`p_{\mathrm{}}`$: $`\delta _{\mathrm{}}(ϵ)`$ $`=`$ $`p_{\mathrm{}}(p_{}=ϵ)p_{\mathrm{}}(p_{}=0)`$ (7) $`=`$ $`\left({\displaystyle \underset{s=1}{\overset{\mathrm{}1}{}}}n_s^bs+{\displaystyle \underset{s=1}{\overset{max}{}}}n_s^rs\right){\displaystyle \underset{s=1}{\overset{\mathrm{}1}{}}}n_ss,`$ (8) with the property that $`lim_{ϵ0}\delta _{\mathrm{}}(ϵ)0`$ for all events where $`C_{\mathrm{}}`$ and $`C_{max}^r`$ are connected. Averaging over all events will yield some effective $`\delta _{\mathrm{}}^{\text{eff}}`$ with $`0<\delta _{\mathrm{}}^{\text{eff}}<\delta _{\mathrm{}}`$. This leads to the conclusion that in the limit $`p_{}0`$ a first order phase transition takes place in $`p_{\mathrm{}}`$. The simulation results support this conjecture. Figure 3 shows these results, where the average was taken over events in which the number of $`br`$-bonds actually formed in the simulation, $`n_{br}`$, was non-zero. For values of $`p_=`$ other than $`1`$ one still finds the same behavior, somewhat less pronounced due to smaller $`\|C_{\mathrm{}}\|`$ and $`\|C_{max}^r\|`$. The same holds for $`f_b0.5`$ where it is clear that the effect vanishes continuously in the limit of a one component bond percolation system: $`f_b0`$ or $`f_b1`$. Obviously the question arises what the behavior of $`p_+^c`$ is. We found that for $`p_{}0`$ the percolation threshold $`p_+^c`$ continually changes to a new, concentration dependent value. In Fig. 4(a) we show this transition, which is of second order type. We also show a fit to the curve, which is of the form $$p_+^c(p_=,p_{}0)=\frac{1}{(u+vp_{})}+t,$$ (9) where the fit parameters were found to be $`u=34.6\pm 0.3`$, $`v=1823\pm 72`$ and $`t=0.246\pm 0.001`$. It has to be noted that this formula only stands on empirical grounds, fits by exponential functions may also be useful. As stated earlier, the same line of thought is applicable to the $`p_=0`$ limit and indeed do our simulations present the same results in the system with $`f_b=0.5`$. For $`f_b0.5`$ this direct symmetry is broken, but qualitatively the results are still the same. Even for $`f_b=0.5`$ they differ however with respect to the afore-mentioned dependence of $`p_+^c`$ on the concentration $`f_b`$ in the limits $`p_==0`$ and $`p_{}=0`$. For $`p_+^c(f_b,p_{}=0)`$ we find the functional form depicted in Fig. 4(b). The results shown are fits of $`p_+^c`$ to the scaling relation, $$\|p_+^c(L)p_+^c\|L^{1/\nu },$$ (10) as given in Ref. , where we kept $`\nu `$ fixed at $`0.88`$ and lattice sizes $`L=16`$, $`L=32`$, $`L=64`$ and $`L=128`$ were taken into account. The simulation data is fitted with $$p_+^c(f_b,p_{}=0)=\frac{2(f_b\frac{1}{2})^2+\frac{1}{2}}{h+mf_b}.$$ (11) This is Eq. (3) with the purely empirical assumption of a hyperbola for $`p_=^c(f_b)`$, which is in agreement with the results of Heermann and Stauffer for a one component site-bond model . Fitting the parameters to our simulation data results in $`h=4.007\pm 0.002`$ and $`m=4.428\pm 0.005`$. A comparison with the formula given by Heermann and Stauffer yields $`h=1/p_{bond}^c(0)=4.019`$ and $`m=(1p_{bond}^c)/(p_+^c(f_{site}^c1))=4.386`$, where $`p_{bond}^c=0.2488`$ and $`f_{site}^c=0.3116`$ are the percolation thresholds for one component bond- and site- percolation on a three dimensional simple cubic lattice, respectively. However, in contrast to the one component model, due to the symmetry introduced in the system, notably that $`p_=p_{bb}=p_{rr}`$, we only have $`f_b[0.0,0.5]`$ as independent regime here, with the interval $`f_b[0.5,1.0]`$ being symmetric to the one shown here, with only the roles of blue and red sites being switched. This is a manifestation of the two components behaving like two superposed, non-interfering one component site-bond percolation systems and lets us conclude that the phase transition in $`p_+^c`$ for $`p_{}0`$ discussed above may be interpreted as an effective transition from a one component bond percolation model to a one component site-bond percolation model. For the case of the second limit discussed, $`p_=0`$, the results are shown in Fig. 5. In Fig. 5(a) we go back to the bond existence probability $`p_{}`$. Its critical value, as a function of the concentration, $`p_{}^c(f_b)`$, is well reproduced with an exponential fit: $$\phi =p_{}^c(f_b)=a\mathrm{exp}(df_b)+c$$ (12) Fitting this empirical formula to the simulation data gives $`a=2.1\pm 0.07`$, $`d=10.9\pm 0.2`$ and $`c=0.547\pm 0.002`$. We define $`f_b^c\phi ^1(p_{}=1)`$ and numerically get $`f_b^c0.14`$ from Eq. (12), whereas in an independent simulation we find $`f_b^c=0.145\pm 0.001`$, which again is a result from a fit to the scaling relation Eq. (10). The parameter $`f_b^c`$ can be regarded as a new threshold, which would correspond to the percolation threshold in a simple site percolation model in which nearest neighbors only belong to the same cluster if they are of opposite flavor, unlike the normal site percolation model which yields $`f_{site}^c=0.3116`$. In Fig. 5(b) $`f_b^c`$ determines the critical point of the phase transition line $`p_+^c(f_b,p_==0)`$. Another argument in favor of a qualitatively new behavior arises by considering the density of accessible edges at the critical concentrations in the two models. For the one considered here it is given by $`\alpha _{}(f_b^c)`$, see Eq. (4), for the usual site percolation model we might define one in an analogous manner: $`\alpha _{site}=(f_{site}^c)^2`$. By setting $`\alpha _{}(f_b^c)=\alpha _{site}`$ one would expect $`f_b^c`$ to be $`0.051`$ which stands in contradiction to our findings. In concluding, we introduced a new way to treat N-component percolation. This approach was applied to a two component site-bond percolation model and new first order phase transitions of $`p_{\mathrm{}}`$ were reported in the limits $`p_{}0`$ and $`p_=0`$. In the latter case we could furthermore establish a novel empirical formula for the percolation threshold as a function of component concentration, whereas in the first earlier findings of a one component site-bond percolation model were found to apply in the two component model too. The field for future work in this area seems vast, one might, for example, try to apply the same method to multi-component systems on lattices of higher dimensions and/or higher connectivity. This approach should also find a broad range of possible applications. One might think of special networks or gelation phenomena with several components involved, which only interact with each other, as well as wetting phenomena. Furthermore an application to stock-market simulations seems possible and is being undertaken by the authors. This work was supported by the National Science Foundation under Grant No. PHY-9605207. One of us (H.M.H.) is supported in part by the Studienstiftung des deutschen Volkes.
no-problem/9907/cond-mat9907364.html	ar5iv	text	# Effect of impact energy on the shape of granular heaps ## 1 Introduction The study of granular materials has received wide attention in the last few years , because of the challenging problems posed by this state of matter. Sand heaps are among the most familiar granular systems that have been studied experimentally and theoretically. Despite the apparent simplicity a sand heap presents a number of intriguing phenomena, such as avalanches internal texture , segregation and stratification . Several theoretical models have been employed to describe the dynamics and the static profile of a sandpile, ranging from purely numerical models, such as lattice gas and molecular dynamics simulations , to discrete lattice models and continuous equations . Granular heaps can be generated by slowly pouring grains (i.e. spherical beads, sand) from a given height over a rigid support (3D piles) or into a vertical Hele-Shaw cell (2D piles). The pile typically displays a well defined slope, corresponding to the angle of repose, with clear deviations at the top and at the tail. In particular, the tail of a 2D pile was found experimentally to follow a logarithmic dependence . This result was anticipated by the solution of a continuous model solved in a 2D silo geometry and by theoretical arguments . The angle of repose depends on the characteristics of the material such as density, humidity, packing history and boundary conditions. Here we analyze the shape of the top of the granular heap, focusing on the effect of the impact energy of the particles. To this end, we perform a set of experiments with glass and leads beads poured at very slow rate into a vertical Hele-Shaw cell and study the changes of the heap profile when grains are poured from different heights. We propose a simple continuous model to describe the steady-state properties of the grains flowing on the top of the pile, providing an explicit expression for the heap profile. The model treats explicitly the dissipation of the kinetic energy of the flowing grains due to inelastic collisions and provides a good fit to the experimental data on glass and lead beads. ## 2 Experiments Experiments are performed within a vertical Hele Shaw cell (20x30cm) with a fixed thickness of $`5mm`$ . Lead beads, with diameter $`d=0.2cm`$ are poured through a funnel placed above the center of the cell. We carefully determine the impact energy of the particles , controlling the distance $`Z_0`$ between the top of the heap and the funnel: the impact energy is thus $`e_0=mgZ_0`$. The falling heights range from $`Z_0=1cm`$ to $`Z_0=14cm`$ and are determined with a with a precision of $`\mathrm{\Delta }Z_0=0.5cm`$. Experiments are performed at very low particle flux, dropping one particle at a time through the funnel. In the range of falling heights employed , the reorganization of the pile after impact is usually very weak and the particle jumps down the slope and is eventually trapped, giving rise to a local change in the slope of the pile. The dynamics of the beads can be sometimes very complex: we observe collisions with the walls of the cell, rolling motion, ejection of other particles and occasionally big avalanches of a few layers. After a short transient stage, the heap grows steadily and the profile translates at constant velocity, proportional to the input flux. In this regime we retrieve the experimental heap profiles (see Fig. 1)from image analysis and we average them. The averaging is done superimposing the profiles obtained at different time steps in the steady state. In Fig. 2 we show the averaged profiles for different impact energies. It can be observed that the heap has a flat top and shows a well defined angle of repose, which turns out to be essentially independent on $`Z_0`$, at least for the experimental values used. When the tails of the heaps obtained with different $`Z_0`$ are superimposed, we observe that the maximum height $`h(0)`$ decreases linearly with $`Z_0`$ (inset of Fig. 2). We find that the shape of the heap is well described by the form $$h(x)=A\theta _cx\theta _c\xi e^{(x/\xi )},$$ (1) where $`\theta _c`$ is the angle of repose, $`\xi `$ is a correlation length and $`A`$ is an inessential constant, that can be eliminated through an appropriate vertical translation of the profile. From Eq. (1) we have that $`h(0)=A\theta _c\xi `$, which implies that $`\xi `$ is proportional to $`Z_0`$, which is verified by the fitting parameters (see Fig. 3) This result reflects the fact that $`Z_0`$ is, together with the particle diameter, the only characteristic length present in the system. The basic features of these results are quite robust and do not depend on the particular material used. We verified this by forming heaps with glass beads of considerably smaller diameter ($`d=0.025cm`$), poured at low flux $`W=0.32g/s`$. In this case $`Z_0`$ can be controlled very precisely by moving the outlet up at the same velocity as the growth of the top of the heap. In this experiment, grains are poured from one of the two sides of the Hele-Shaw cell. We see in Fig. 4 that also in this case $`h(0)`$ is linear in $`Z_0`$ and the decay is reasonably well fitted by an exponential, excluding from the fitting region the first 0.5cm, which correspond to the radius of the funnel. Finally in this case, the measured angle of repose displays a small dependence on $`Z_0`$, of the order of $`\mathrm{\Delta }\theta =0.05`$. Looking at Fig. 3, we note that the linear dependence of $`\xi `$ from $`Z_0`$ is steeper in the case of glass beads by roughly a factor $`2`$. The origin of this effect is probably due to the different restitution coefficients of the materials. ## 3 Model In order to understand the experimental observations, we consider a sandpile model in which energy dissipation is explicitly treated. The complete dynamical evolution of all the grains and their energy dissipation is extremely difficult to treat explicitly. The evolution of the heap is indeed a very complex phenomenon, being due to the way momentum is transfered during particle collisions, frictional properties and other forms of energy dissipation. However, we note that in our experiment the growth of the heap displays a remarkable stationary state, which we can try to describe with a simplified model taking into account essentially only energy dissipation. Alonso et al. have introduced a simple automaton model which has been used to derive the angle of repose of a pile in terms of microscopic parameters, such as the restitution coefficient of the grains. In this model, grains are poured one by one on the top of a heap with an initial kinetic energy $`e_0`$. As the grains fall down the pile, parameterized by the local height $`h(x)`$, their kinetic energy $`e`$ decreases due to inelastic collisions and increases because of gravity as $$e(x+\delta )=r(e(x)+mg(h(x)h(x+\delta ))),$$ (2) where $`m`$ is the mass of the grains, $`\delta `$ is the characteristic size of the jumps, $`g`$ is the gravity acceleration and $`r`$ is an “effective” restitution coefficient. The grains jump to the right until their kinetic energy is less than a threshold $`u`$. When a grain comes to rest at $`x`$ the height $`h`$ of the column $`x`$ is increased by one unit and a new grain is poured at the origin. This model has been simulated and, despite its simplicity, was shown to explain several experimental results . We study here a continuous version of the model in which the steady-state kinetic energy of the rolling grains follows Eq. (2) $$\delta \frac{de(x)}{dx}=(r1)e(x)+r\gamma (x),$$ (3) where $`\gamma (x)mg\delta dh/dx`$ measures the local slope. Eq. (3) describes the steady-state energy profile in terms of a given slope profile $`\gamma (x)`$. To close the problem, we need to specify how the steady-state slope profile depends on the energy profile. We assume that the system in the steady state is at the verge of stability and that the variations of the local slope compensate the difference between the kinetic energy $`e`$ and the trapping energy $`u`$: the slope will decrease for $`e>u`$ and increase for $`e<u`$. The simplest equation of this form is given by $$\delta \frac{d\gamma (x)}{dx}=\mathrm{\Gamma }(ue(x)),$$ (4) where $`\mathrm{\Gamma }`$ is a phenomenological parameter. Eq. (4) implies that for large $`x`$ the grains are at the threshold of mobility $`e(x)u`$. Finally, we impose the boundary conditions $`e(0)=e_0=mgZ_0`$ and $`\gamma (0)=0`$. We note that by dividing each term of the equation by $`mg`$ all the quantities can be expressed in terms of lengths. Eqs. (3-4) are linear and can be solved in order to obtain an explicit expression for the profile of the pile. It is convenient to transform the pair of equations into an homogeneous one using the transformation $$E(x)=e(x)u,G(x)=\gamma (x)+(1r)u/r.$$ (5) The solution of the homogeneous equations $`\delta {\displaystyle \frac{de(x)}{dx}}=(r1)E(x)+rG(x),`$ (6) $`\delta {\displaystyle \frac{dG(x)}{dx}}=\mathrm{\Gamma }E(x),`$ (7) is given by $`e(x)=u+E_+e^{\lambda _+x/\delta }+E_{}e^{\lambda _{}x/\delta }`$ (8) $`\gamma (x)=(r1)u/r+G_+e^{\lambda _+x/\delta }+G_{}e^{\lambda _{}x/\delta }`$ (9) where $`\lambda _\pm =(r1\pm \sqrt{(r1)^24r\mathrm{\Gamma }})/2`$ are the eigenvalues of the matrix $$\left(\begin{array}{cc}𝓇\mathcal{1}& 𝓇\\ \mathcal{\Gamma }& \mathcal{0}\end{array}\right)$$ (10) The coefficients $`E_\pm `$ and $`G_\pm `$ can be explicitly obtained imposing the boundary conditions. In the limit of large $`x`$ the slope of the pile is given by $`\gamma _{\mathrm{}}=(r1)u/r`$, from which we obtain the angle of repose $`\theta _c=(1r)u/(mg\delta r)`$. Since the real part of $`\lambda _\pm `$ is negative for all the values of the parameters the model has a well defined angle of repose. The relaxation towards the angle of repose is expressed by the sum of two exponentials with characteristic lengths $`\xi _\pm \delta /\lambda _\pm `$. Since $`\xi _+>\xi _{}`$, the relaxation is dominated by $`\xi _+`$ that we can identify with $`\xi `$ of Eq. 1. Finally, integrating the equation for $`\gamma (x)`$, we obtain an analytical form for the shape of the pile that can be compared with experiments $$h(x)=h(0)(1r)ux/(mg\delta r)$$ $$\delta [G_+/\lambda _+(e^{\lambda _+x/\delta }1)+G_{}\lambda _{}(e^{\lambda _{}x/\delta }1)].$$ (11) The exponential relaxation to the angle of repose is experimentally observed, although in the present form the model can not account for the impact energy dependence of the relaxation length. An original assumption of Eq. (2) was that the jump length $`\delta `$ is constant during the grain motion. The problem of a ball rolling and jumping over a rough inclined plane has received a large attention in recent years . It has been experimentally observed that as a function of the angle of the inclined plane the motion is decelerated or accelerated, with a region in between where the motion is effectively viscous. The energy relaxation was recently studied in Ref. , where the stopping length in the decelerated regime was measured as a function of the initial energy. The result clearly indicates that the stopping length is linearly dependent on the initial energy. We can show this fact by a simple argument. Consider an inelastic ball hitting a plane with energy $`e_0`$ and let $`\alpha `$ be the angle between the plane and the particle velocity before the impact. We consider for simplicity that normal and tangential restitution coefficients are both equal to $`r`$ so that at each jump $`n`$ the energy decreases by $`e_n=re_{n1}`$. A straightforward calculation shows that the jump length decreases with $`n`$ as $`\delta _n=2e_0r^n\mathrm{sin}(2\alpha )/(mg)`$, which yields $`e(x)=e_0[1mgx(1r)/(2e_0\mathrm{sin}(2\alpha ))]=e_0(1x/\xi )`$. We see here that the characteristic length for energy relaxation $`\xi `$ is linear in $`e_0`$, in agreement with the experimental result seen in Fig. 3 ## 4 Conclusions In conclusion, we have studied the effect of impact energies on the shape of the sand heap. Experiments show an exponential relaxation of the slope to the angle of repose, with a characteristic length that is linearly dependent on the impact energy of the grains. We model this process by simple relaxation equations and explain theoretically the experimental results. S. Z. is supported by EC TMR Research Network contract ERBFMRXCT960062.
no-problem/9907/astro-ph9907070.html	ar5iv	text	# Constraints on the bulk Lorentz factor in the internal shock scenario for gamma–ray bursts ## 1 introduction In the last few years, increasing evidence in favor of the fireball model (Rees & Mészáros 1992) for gamma–ray bursts has been gathered, thanks to the observations of BeppoSAX (Boella et al. 1997). The power law decay of the optical afterglow of several bursts has lasted for time–scales of a year, in striking agreement with the simplest fireball scenario (Wijers, Rees & Mészáros 1997), in which a shock wave propagates in the interstellar medium (ISM), accelerating particles which then emit by the synchrotron process (Sari, Piran & Narayan 1998). However a single shock wave cannot account for both the temporal behavior of the $`\gamma `$–ray emission (Fenimore et al. 1999a) and for the requirement of a high efficiency in converting its kinetic energy into radiation (Sari & Piran 1997). In fact the time variability structure of bursts appears to be constant during the whole of the high energy emission (Fenimore et al. 1999b), while the deceleration of a shock slowed down by interactions with the ISM would produce a time dilation between the first and the last spike of the burst. Moreover the observed variability cannot be explained as the consequence of inhomogeneities in the ISM since a radiative efficiency of less than $`1\%`$ would be expected (Sari & Piran 1997; see however Dermer, Böttcher & Chiang 1999). In the internal shock scenario, put forward by Rees & Mészáros (1994), the inner engine produces many relativistic expanding shells (or an unsteady wind) with a distribution of Lorentz factors centered on a mean value $`\mathrm{\Gamma }`$ and width $`\mathrm{\Delta }\mathrm{\Gamma }\mathrm{\Gamma }`$. The burst radiation is produced through the dissipation occurring when a faster shell catches up a slower one. No matter the physical mechanism producing the observed photons, the inner (hidden) engine is responsible for the temporal structure of the observed burst through the time history of the shell emission (Kobayashi, Piran & Sari 1997). Even if we may have information on the time–scale of the shell ejection, it is hard to estimate the bulk Lorentz factor $`\mathrm{\Gamma }`$ of the relativistic outflow. This is however a critical parameter to unveil the mechanism(s) that powers the outflow and the radiating process that produces $`\gamma `$–ray photons. In fact, the knowledge of $`\mathrm{\Gamma }`$ allows us to estimate the amount of baryon loading of the fireball and the intrinsic frequency of the emitted photons. The constraints related to the compactness problem (see e.g. Piran 1992) give a lower limit $`\mathrm{\Gamma }\stackrel{>}{}100`$, while an upper limit $`\mathrm{\Gamma }\stackrel{<}{}10^5`$ is inferred from the need of having the fireball opaque till the acceleration stage is completed. This poorly constrains the properties of the outflow, leaving a broad interval for physically possible $`\mathrm{\Gamma }`$. In this letter we analyze the kinematic evolution of an inhomogeneous fireball, in which the flow is approximated by discrete shells with different intrinsic properties (baryon load, energy and Lorentz factor) and the outflow energy is dissipated through binary shell–shell collisions. We derive general constraints on the relativistic properties of the outflow both in the simplest internal shock scenario, that requires a single interaction for each couple of shells, and in the case of hierarchically developed internal shocks, in which successive shell mergers end up in a single (or a few) more massive shells, whose interaction with the ISM produces the afterglow. The latter scenario is strongly favored by the need of dissipating a significant fraction of ordered outflow energy (see also Kobayashi et al. 1997). ## 2 Kinematic of the flow We describe the relativistic outflow as constituted of $`N_\mathrm{S}`$ shells, each with its own relativistic Lorentz factor $`\mathrm{\Gamma }_\mathrm{i}`$, mass $`M_\mathrm{i}`$ and energy $`E_\mathrm{i}=\mathrm{\Gamma }_\mathrm{i}M_\mathrm{i}c^2`$. The global properties of the flow are described by the averages $`\mathrm{\Gamma }`$, $`M`$ and $`E`$ and their dispersions, by the total duration of the ejection of shells $`T`$ (as measured in the rest frame of the inner engine) and by the time interval between the ejection of two successive shells $`\mathrm{\Delta }t`$. If the ejection time of a typical shell equals the time of quiescent phase between the production of two consecutive shells, $`\mathrm{\Delta }t=T/2N_\mathrm{S}`$ (it is thus assumed that the time necessary to eject a fast or a slow shell is the same, see however Panaitescu, Spada and Mészáros 1999). The average mass, energy and relativistic factor are related as $$M=\frac{E}{N_\mathrm{S}\mathrm{\Gamma }c^2}5.6\times 10^7E_{52}\mathrm{\Gamma }_2^1N_{\{\mathrm{S},2\}}^1M_{},$$ (1) where $`E=10^{52}E_{52}`$ erg is the total energy of the outflow <sup>1</sup><sup>1</sup>1Here and in the following we parameterize a quantity $`Q`$ as $`Q=10^\mathrm{x}Q_\mathrm{x}`$ and adopt CGS units.. The dynamics of each shell is characterized by an initial phase in which it is accelerated to its final Lorentz factor as $`\mathrm{\Gamma }R/R_0`$, where $`R_0`$ is the distance of the shell from the center of expansion at the ejection time (see e.g. Piran 1998). This phase ends when the final Lorentz factor is reached at the ‘acceleration’ radius: $$R_\mathrm{A}=\mathrm{\Gamma }_\mathrm{i}R_010^9\mathrm{\Gamma }_{\{\mathrm{i},2\}}R_{\{0,7\}}\mathrm{cm}.$$ (2) During the following evolution, the shell coasts with constant velocity until an interaction with another shell or with the ambient medium takes place. Assuming that the acceleration phase lasts for a time interval negligible with respect to the coasting phase, the interaction between two shells with Lorentz factors $`\mathrm{\Gamma }_1`$ and $`\mathrm{\Gamma }_2`$ ($`>\mathrm{\Gamma }_1`$) and initially spaced by a time interval $`\mathrm{\Delta }t`$ occurs at a radius: $$R_I=2\frac{\alpha _\mathrm{\Gamma }^2}{\alpha _\mathrm{\Gamma }^21}\mathrm{\Gamma }_1^2c\mathrm{\Delta }t=4.0\times 10^{13}\mathrm{\Gamma }_{\{1,2\}}^2T_1N_{\{\mathrm{S},2\}}^1\mathrm{cm}$$ (3) where the numerical value has been computed for $`\alpha _\mathrm{\Gamma }=\mathrm{\Gamma }_2/\mathrm{\Gamma }_1=2`$ and assuming $`\mathrm{\Delta }t=T/2N_S`$. The interaction of the shells with the ambient medium is a continuous process and does not happen at a well determined radius. However a typical scale can be estimated as the distance at which the shell Lorentz factor is half its initial value, which is reached when the shell has swept up an external mass $`m`$ equal to its rest mass divided by the initial Lorentz factor (see Mészáros & Rees 1997). Hence the deceleration radius can be approximated as $`M_i/\mathrm{\Gamma }_i=(4/3)\pi R_D^3nm_p`$, i.e.: $$R_D1.2\times 10^{16}E_{52}^{1/3}n_0^{1/3}\mathrm{\Gamma }_{\{\mathrm{i},2\}}^{2/3}N_{\{\mathrm{S},2\}}^{1/3}\mathrm{cm}$$ (4) where $`n`$ is the density (assumed uniform) of the ambient medium and shells of equal energy have been considered. Therefore, the more the shell is relativistic, the smaller the deceleration radius is (Eq. 4), while, on the contrary, the largest the radius at which internal shocks occur (see Eq. 3). This allows us to put some constraints on the average relativistic factor of the flow, since observationally the internal shocks set up before the afterglow, i.e. before the development of the external shock. To fully describe the kinematic of the flow, a fourth transition radius is important, i.e. the radius at which the shell becomes transparent (to Thomson scattering). If, in fact, the shell becomes optically thin before the acceleration phase is completed, the internal energy can escape from the shell and the acceleration process is damped. On the other hand, if two shells collide when still opaque, they give rise to a single shell that is reaccelerated to an intermediate Lorentz factor. The transparency radius corresponds to: $$R_\tau =\left(\frac{M_i\sigma _T}{4\pi m_p}\right)^{1/2}5.9\times 10^{12}E_{52}^{1/2}\mathrm{\Gamma }_{\{\mathrm{i},2\}}^{1/2}N_{\{\mathrm{S},2\}}^{1/2}\mathrm{cm}$$ (5) ## 3 Comparing the radii Fig. 1 shows a comparison of the four critical radii for a typical burst with energy $`E=10^{52}`$ erg and duration $`T=`$ 10 s. The ISM has been assumed uniform with density $`n=1`$ cm<sup>-3</sup>. Two cases have been considered: the left panel refers to a burst made by a (relatively) small number of shells ($`N_S=100`$), while for the right panel $`N_S=25000`$. The smaller value has been derived assuming that each of the burst pulses fitted by Norris et al. (1996) is produced by the interaction of a shell pair. Norris et al. (1996) find a number of pulses $`N_p\stackrel{<}{}50`$, and hence we have $`N_S=2N_P100`$. In the latter case, instead, the central engine emits shells at regular time intervals, equally spaced by the smallest variability time–scale observed in GRB light–curves ($`200\mu `$s, Schaefer & Walker 1999). In this situation each of the Norris et al. (1996) pulses is considered as a blend of sub–pulses, with an envelope corresponding to the varying efficiency of the central engine: indeed, Walker, Schaefer & Fenimore (1999) have shown that millisecond variability on top of the larger time–scale modulation is a common feature of GRB light–curves. In both cases the maximum value of $`\mathrm{\Gamma }`$ is bound by the requirement that internal shocks must happen before external ones, while the minimum value is constrained by the transparency condition. This second constraint is less severe since, as already mentioned, collisions between opaque shells would simply cause a reacceleration of the merged shell. Note however that this could play a role in preventing the presence of very slow shells outside the transparency radius $`R_\tau `$. It should be also stressed that numerically the limit given by the transparency requirement is similar to the minimum value $`\mathrm{\Gamma }\stackrel{>}{}100`$, already obtained from the compactness argument (e.g. Piran et al. 1996). The acceleration radius does not impose any significant constraint, showing that from the kinematic and radiative points of view, shells could be potentially accelerated even to very high Lorentz factors $`\mathrm{\Gamma }10^5`$. From the condition $`R_I<R_D`$ we get an upper limit: $$\mathrm{\Gamma }_{\mathrm{lim}}=800T_1^{3/8}E_{52}^{1/8}n_0^{1/8}N_{\mathrm{S},2}^{1/4}.$$ (6) This limit is quite robust with respect to $`E`$ and $`n`$ while it is more dependent of $`T`$ and $`N_S`$. A variation of three orders of magnitude of the total energy changes it by a factor $`3`$ only. The density of the ISM has been assumed uniform and $``$ one proton cm<sup>-3</sup>. This is probably a lower limit. However, even in the case of the hypernova scenario (Paczynski 1998), in which GRB would occur in a much denser environment ($`n10^4`$ cm<sup>-3</sup>), $`\mathrm{\Gamma }_{\mathrm{lim}}`$ decreases only by a factor of $`3`$. Finally, the total burst duration of 10 seconds corresponds to the mean T90 parameter of the long GRBs, and is appropriate for the majority of bursts. Its small uncertainty influences $`\mathrm{\Gamma }_{\mathrm{lim}}`$ rather weakly. ## 4 Efficiency of internal shocks The efficiency of internal shocks in converting the bulk outflow energy into internal energy can be easily estimated since each collision satisfies energy and momentum conservation (see also Kobayashi et al. 1997). Consider two shells of rest masses $`m_1`$ and $`m_2`$ and Lorenz factors $`\mathrm{\Gamma }_1`$ and $`\mathrm{\Gamma }_2`$ ($`>\mathrm{\Gamma }_1`$), respectively. Calling $`ϵ`$ the internal (random) energy of the merged shell after the interaction, we have: $`\mathrm{\Gamma }_1m_1+\mathrm{\Gamma }_2m_2`$ $`=`$ $`\mathrm{\Gamma }_f\left(m_1+m_2+ϵ/c^2\right)`$ $`\mathrm{\Gamma }_1\beta _1m_1+\mathrm{\Gamma }_2\beta _2m_2`$ $`=`$ $`\mathrm{\Gamma }_f\beta _f\left(m_1+m_2+ϵ/c^2\right),`$ (7) where the subscript $`f`$ refers to quantities after the interaction and $`\mathrm{\Gamma }=(1\beta ^2)^{1/2}`$. If we assume that all of the internal energy is converted into radiation, we obtain - independently of the emission mechanism - an upper limit for the efficiency $`\eta =ϵ/(\mathrm{\Gamma }_1m_1+\mathrm{\Gamma }_2m_2)`$. From the conservation equations an implicit solution for the final bulk Lorentz factor of the merged shells and the maximum radiative efficiency can then be derived: $`\beta _f`$ $`=`$ $`{\displaystyle \frac{\beta _1+\alpha _\mathrm{\Gamma }\alpha _m\beta _2}{1+\alpha _\mathrm{\Gamma }\alpha _m}}`$ $`\eta `$ $`=`$ $`1{\displaystyle \frac{\mathrm{\Gamma }_f(1+\alpha _m)}{\mathrm{\Gamma }_1(1+\alpha _\mathrm{\Gamma }\alpha _m)}},`$ (8) where $`\alpha _m=m_2/m_1`$. The above relations give an upper limit to the fraction of energy that can be radiated in photons. In fact, strictly speaking, $`\eta `$ is the fraction of bulk kinetic energy converted into internal energy. In the standard synchrotron shock model, this random energy is equally shared among protons, electrons and magnetic field, and only one third of this energy (the fraction going to electrons) can be radiated (see e.g. Panaitescu et al. 1999). The major features of Eq. 7 are that for a large difference in the Lorentz factors ($`\alpha _\mathrm{\Gamma }1`$) the efficiency can approach unity and that for a fixed value of $`\alpha _\mathrm{\Gamma }`$ the maximum efficiency is reached when $`m_1=m_2`$. For $`\mathrm{\Gamma }_1,\mathrm{\Gamma }_21`$, as for internal shocks, the expression for $`\eta `$ reduces to: $$\eta 1\frac{1+\alpha _m}{\sqrt{1+\alpha _m\left(\alpha _m+\alpha _\mathrm{\Gamma }+1/\alpha _\mathrm{\Gamma }\right)}}.$$ (9) An even simpler relation is found if the two shells have the same total energy (i.e. $`\alpha _\mathrm{\Gamma }=1/\alpha _m`$): $$\eta 1\frac{1+\alpha _m}{\sqrt{2+2\alpha _m^2}}.$$ (10) In this latter case the efficiency is always lower than 30%, independently of $`\alpha _\mathrm{\Gamma }`$. Fig. 2 shows the efficiency for low–intermediate values of $`\alpha _\mathrm{\Gamma }`$ in two limits: the most efficient situation ($`m_1=m_2`$, dashed line) and the equal energy case ($`\mathrm{\Gamma }_1m_1=\mathrm{\Gamma }_2m_2`$, solid line). The limit $`\mathrm{\Gamma }_1,\mathrm{\Gamma }_21`$ has been assumed. Note that, in this case, the efficiency does not depend on the value of $`\mathrm{\Gamma }_1`$. Let us now consider the global efficiency of the burst. In the ‘standard’ internal shock scenario (Rees & Mészáros 1994) the shells have a distribution of Lorentz factors with mean value $`\mathrm{\Gamma }`$ and width $`\mathrm{\Delta }\mathrm{\Gamma }\mathrm{\Gamma }`$. This in turn corresponds to a distribution of $`\alpha _\mathrm{\Gamma }`$ with $`\alpha _\mathrm{\Gamma }2`$ within a factor of order unity (which accounts for the particular shape of the distribution). This implies (see Fig. 2) $`\eta 5\%`$, independently of the mass ratio $`\alpha _m`$. A numerical simulation for a log–normal distribution of $`\mathrm{\Gamma }`$ with $`\mathrm{\Delta }\mathrm{\Gamma }=\mathrm{\Gamma }`$, gives $`\eta =6.8\%`$. These (upper limits on) efficiencies are problematic since, in the absence of extreme beaming, the total energy release of the most powerful GRBs would exceed by some orders of magnitude the maximum energy release achievable by current models, that involve a stellar mass black hole as the energy source. If, e.g., GRB 990123 had an efficiency of $`5\%`$, the required energy would be $`E_{\mathrm{iso}}5\times 10^{55}`$ erg. An extremely narrow beaming angle $`\theta 0.5^{}`$ would be then needed to reduce the energy release to the value $`E_\theta =10^{52}`$ erg. A higher efficiency can be only obtained by allowing for a broader distribution of relativistic factors $`\mathrm{\Gamma }`$ and imposing that all shells have roughly the same mass, somehow requiring a finely tuned variable engine. Moreover, a very broad distribution of Lorentz factors implies that the Compton drag effect becomes crucial, as described in the following section. ### 4.1 Compton drag Let us assume that the emission produced in the interaction between two shells is isotropic in the shell comoving frame $`K^{}`$. Consider also a third shell, faster than the previous two, with Lorentz factor $`\mathrm{\Gamma }^{}`$ in that frame. This moves in the radiation bath of total energy $`E_{\mathrm{ph}}^{}`$ produced in the interaction, and thus Compton scatters a fraction $`\tau _T`$ of the photons, increasing their energy by a factor $`\mathrm{\Gamma }^2`$ (here $`\tau _T`$ is the Thomson optical depth of the fast shell). The total energy lost by the fast shell in the frame $`K^{}`$ is hence $`\tau _T(\mathrm{\Gamma }^21)E_{\mathrm{ph}}^{}`$. If $`\tau _T(\mathrm{\Gamma }^21)`$ approaches unity, there are two major consequences: first a significant part of the energy of photons is due to the inverse Compton mechanism rather than to the internal shock dissipation process; second, the fast shell looses a significant fraction of its bulk kinetic energy to up–scatter the primary photons and this causes a braking of the fast shell (in the frame $`K^{}`$). Thus, even if the original distribution of $`\mathrm{\Gamma }`$ factors were very broad, the early interactions would give rise to an efficient inverse Compton drag on the faster shells, reducing the mean value of $`\alpha _\mathrm{\Gamma }`$ in the successive collisions and hence the global efficiency of internal shocks. Moreover, in this case the net energy produced through inverse Compton scattering would be equal or even larger than that directly due to shocks, and the primary emission mechanism would be different. We plan to investigate this issue in a forthcoming paper. ## 5 Hierarchical internal shocks The low efficiency predicted by Eq. 9 refers to a single interaction between two shells. However, each time this interaction takes place, a new shell is formed which can in turn catch up (or be caught by) another one. If the $`i^{th}`$ interaction has efficiency $`\eta _i`$, after $`N`$ of them the fraction of the initial bulk energy converted into photons will be: $$\eta _\mathrm{N}=\underset{i=1}{\overset{N}{}}\eta _i(1\eta _{i1})\mathrm{}(1\eta _1)\stackrel{\eta _i=\eta }{===}\eta \underset{i=1}{\overset{N}{}}(1\eta )^{i1}$$ (11) If the efficiency is constant ($`\eta _i=\eta `$) and $`N`$ is arbitrarily large, all the energy of the outflow can be converted into radiation, as $`\eta _{_{\mathrm{}}}=1`$ for any value of $`\eta `$. For the typical efficiency $`\eta =6.8\%`$ derived above, about 9 interactions are needed to obtain an overall $`50\%`$ efficiency. Assuming that, as sketched in Fig. 3, the collision ‘tree’ develops as a binary bifurcation, the number $`𝒩`$ of shells that has to merge into a single one after $``$9 interactions is $`𝒩=2^9=512`$. This result has two main consequences. Firstly, it is unlikely that a burst with high efficiency is produced by a small ($`<`$ 100) number of shells: a scenario in which the inner engine emits a large number of shells is thus favored. Secondly, in deriving $`\mathrm{\Gamma }_{\mathrm{lim}}`$ (Eq. 6) we assumed that a single interaction was enough to power a GRB. However, within the hierarchical scenario, a shell must interact with $`𝒩`$ other shells and hence Eq. 6 must be modified as: $$\mathrm{\Gamma }_{\mathrm{lim}}=800T_1^{3/8}E_{52}^{1/8}n_0^{1/8}N_{\mathrm{S},4}^{1/4}𝒩_2^{1/4}$$ (12) which gives $`\mathrm{\Gamma }_{\mathrm{lim}}650`$ for - say - $`N_S=25000`$ and $`𝒩=512`$. Even if this is an illustrative value only (the actual number depending on the details of the initial distribution of the Lorentz factors), we conclude that values of $`\mathrm{\Gamma }`$ larger than $`1000`$ are unlikely. ## 6 Discussion We have analyzed the kinematic efficiency of internal shocks in a fireball made of many individual shells. Since the radius at which the internal shocks set up grows with the average Lorentz factor of the flow, while the development of an external shock is favored by a higher relativistic motion, we can put limits on the average Lorentz factor $`\mathrm{\Gamma }`$ of the flow if - as inferred from observations - internal shocks develop before the external one. Moreover (see also Kobayashi et al. 1997) we conclude that a simple internal shock scenario in which each shell is caught up by a single faster shell, suffers from a very low radiative efficiency, unless extremely different Lorentz factors are involved. These would however cause a dramatic Compton drag effect, which in turn constrains the possibility of a very broad $`\mathrm{\Gamma }`$ distribution. A possible solution is to have a flow with moderately relativistic Lorentz factors $`100\stackrel{<}{}\mathrm{\Gamma }\stackrel{<}{}600`$ in which collisions develop until a single (or a few) massive shell is formed. For this to happen the shells have to be ejected at small time intervals, of the order of milliseconds. In this case each burst is made of more than a thousand spikes that merge to produce the broad peaks often observed in GRB lightcurves. These peaks would then reflect a modulation of the hidden central engine that powers the burst, rather than a single collision. It is interesting to ask whether an outflow with significantly higher Lorentz factors produces a burst or not. In a flow with very high $`\mathrm{\Gamma }`$, the internal shocks set-up at a larger radius, and hence a hierarchical internal shock cannot develop. Since the efficiency is roughly proportional to the number of shocks a shell undergoes, a higher bulk gamma is linked to a lower efficiency. Observationally, this implies that the ratio between burst and afterglow fluences is a function of $`\mathrm{\Gamma }`$, being larger for smaller $`\mathrm{\Gamma }`$. A further consequence is that very short (millisecond) bursts should be characterized by smaller efficiencies and brighter afterglows. To date, this can be only a prediction since the BeppoSAX trigger works only for long burst. As a final note, following the explosion of the bright burst GRB 990123, two values of $`\mathrm{\Gamma }`$ have been derived. Sari & Piran (1999) obtain $`\mathrm{\Gamma }200`$ from the optical flash intensity while Liang et al. (1999) estimate $`\mathrm{\Gamma }350`$ from the lightcurve properties. Despite the differences, both values are in agreement with the general limits derived in this letter. ## Acknowledgments We thanks S. Campana for stimulating comments and for carefully reading the manuscript. DL and AC thank the Cariplo foundation and the Italian MURST for financial support, respectively. This research was supported in part by the National Science Foundation under Grant No. PHY94-07194 (AC).
no-problem/9907/chao-dyn9907012.html	ar5iv	text	# Learning Driver-Response Relationships from Synchronization Patterns ## 1 Introduction The study of synchronization between chaotic systems has been a topic of increasing interest since the beginnings of the ’90s. One important step in this direction was the introduction of the concept of generalized synchronization , extending previous studies of coupled identical systems (identical synchronization ) to the study of coupled systems with different dynamics. Let us denote by $`𝐗`$ and $`𝐘`$ two dynamical systems, and by $`𝐱=\{x_1,\mathrm{},x_d\}`$ and $`𝐲=\{y_1,\mathrm{},y_r\}`$ their state vectors, obtained for example by delay embedding. We assume in the following that the dynamics is deterministic with continuous time (the case of maps is completely analogous, and will be treated in Secs. 3 and 4). We further assume the systems are unidirectionally coupled, say $`𝐗`$ is the autonomous driver and $`𝐘`$ the driven response $`\dot{𝐱}(t)`$ $`=`$ $`F(𝐱(t)),`$ $`\dot{𝐲}(t)`$ $`=`$ $`G(𝐲(t),𝐱(t))`$ (1) We speak about generalized synchronization between $`𝐗`$ and $`𝐘`$ if the following relation exists: $$𝐲(t)=\mathrm{\Psi }(𝐱(t))$$ (2) This requirement is less restrictive than the one of identical synchronization, in which $`\mathrm{\Psi }=1`$. Equation (2) implies that the state of the response system is a function only of the state of the driver. It is not to be confused with the opposite relation $`𝐱(t)=\mathrm{\Phi }(𝐲(t))`$ (considered in ), which is generically valid for sufficiently high embedding dimensions if the coupling is non-singular, in the sense of obeying $`det(G_i/x_n)0`$ everywhere. This follows from the implicit function theorem, which allows us to invert eq.(1b) to $`𝐱=\chi (𝐲,\dot{𝐲})`$, and from the fact that $`\stackrel{~}{𝐲}(t)=(𝐲(t),\dot{𝐲}(t))`$ is as good a state vector as $`𝐲(𝐭)`$. In particular, if we consider $`G(𝐲(t),𝐱(t))=H(𝐲(t))+U(𝐱(t))`$, we immediately have $`𝐱(t)=U^1(\dot{𝐲}(t)H(𝐲(t)))`$ if $`U(𝐱)`$ can be inverted. The transformation $`\mathrm{\Psi }`$ does not need to be smooth as considered in and explicitly required in . In fact, Pyragas defined as strong and week synchronizations the cases of smooth and non-smooth transformations, respectively (see also ). If one of the systems drives the other and a relationship like equation (2) exists, it is possible to predict the response from the simultaneous state of the driver. But the opposite is not true. Just knowing that a relationship like equation (2) exists and that the state of $`𝐘`$ can be predicted from that of $`𝐗`$, it is in general not possible to establish which is the driver and which is the response. This is obvious when $`\mathrm{\Psi }`$ is bijective (i.e., $`\mathrm{\Phi }=\mathrm{\Psi }^1`$ exists and is unique). The above arguments tell us that $`\mathrm{\Psi }`$ is indeed likely to be bijective in case of generalized synchronization, at least for nearly all $`𝐱`$: If a coupling is not regular in the above sense, then its singularities are typically located on a set of measure zero. One might tend to believe that $`𝐗`$ must control $`𝐘`$ (and not the opposite), if $`𝐲`$ follows the motion of $`𝐱`$ with a positive time delay. But even then one cannot be sure since there could be an internal delay loop in $`𝐘`$ which causes the emitted time series to lag behind. Also, both systems could be driven by a third system. Thus, detecting causal relationships is not easy in general, although it is of course of utmost importance in many applications. In the above we pretended that we could detect exactly whether the state of one of the systems is a function of the other. This is of course never the case in practical applications. Different observables which should enable one to detect interdependencies in realistic cases were introduced by several authors. Following an original idea of Rulkov et al. , mutual cross-predictabilities were defined and studied by Schiff et al. and Le van Quyen et al. . A quantity more closely related to that of , but optimized for robustness to noise and imperfections in the data was used in . In the latter paper also a number of other variants were discussed. Some of these variants were tested and found to be inferior, but no systematic tests were made. In contrast to our above discussion, the authors of and claimed that driver/response relationships can be deduced from such interdependencies. But their proposals, backed by numerical studies of simple model systems, were mutually contradictory. While it was argued in that the driver state $`𝐱`$ should be more dependent on the response state $`𝐲`$ (i.e., there exists a stronger functional dependency $`𝐲\stackrel{\mathrm{\Phi }}{}𝐱`$) than vice versa (which is, as we said, a bit counter intuitive), exactly the opposite was claimed in . Finally, in it was claimed that neither can be expected to be correct in realistic situations with finite noisy data, and that it is in general the system with more excited degrees of freedom (the more ‘active’ system) that is more independent, while the state of the more ‘passive’ system (with less excited degrees of freedom) depends on it. It is the purpose of the present work to settle this question by carefully studying simple toy models, including Lorenz, Roessler, and Henon systems, using two of the interdependence measures proposed in . Basic notions involved in generalized synchronization are reviewed in Sec.2. In Sec.3, we recall the operational definition of interdependence used in . Numerical results are presented in Sec.4, and our conclusion is drawn in Sec.5. ## 2 Generalized Synchronization with Exactly Known Dynamics While identical synchronization is easily visualized by plotting the difference between one of the coordinates of the driver and the corresponding coordinate of the response, no similarly simple way exists to detect generalized synchronization. Constructing the function $`\mathrm{\Psi }`$ explicitly might be possible in particularly simple cases, but since this will be never exact, it will be never clear whether deviations from eq.(2) are due to lack of synchronization or inexactness of $`\mathrm{\Psi }`$. Instead, the methods of choice in cases where the exact equations of motion are known and where arbitrary initial states can be prepared are the study of Lyapunov exponents and the identical synchronization of two identical response systems differing in their initial conditions. For the driver/response systems as in eq.(1), one has $`d+r`$ different Lyapunov exponents. Of these, $`d`$ exponents coincide with those of the (autonomous) driver denoted by $`\lambda _i^{(𝐗)},i=1\mathrm{}d`$. The other $`r`$ exponents coincide with those of the response, considered as a non-autonomous system driven by the external signal $`𝐱(t)`$ (called conditional Lyapunov exponents in ) <sup>1</sup><sup>1</sup>1 To see this, we have to recall that all Lyapunov exponents are obtained by iterating $`d+r`$ basis vectors in tangent space, and re-orthogonalize them repeatedly. Tangent vectors corresponding to $`\lambda _i^{(𝐗)}`$ span only the first $`d`$ coordinates. The remaining tangent vectors have the first $`d`$ components equal to zero, either by orthogonalization or because their last $`r`$ components increase faster than any of the first $`d`$ components.. They will be called $`\lambda _i^{(𝐘)},i=1\mathrm{}r`$. Ranking the Lyapunov exponents as usual by magnitude, we have generalized synchronization iff $`\lambda _1^{(𝐘)}<0`$. Furthermore, once the Lyapunov exponents are known, the dimension of the combined system $`𝐗+𝐘`$ can be estimated from the Kaplan-Yorke formula $$D_{𝐗+𝐘}=l+\underset{j=1}{\overset{l}{}}\frac{\lambda _j}{\|\lambda _{l+1}\|}$$ (3) (here, $`l`$ is the largest integer for which the sum over $`j`$ is non negative). Generically, we must expect this also to be the dimension of $`𝐘`$ alone <sup>2</sup><sup>2</sup>2I.e., the dimension of an attractor constructed exclusively from components of $`𝐲`$; we keep of course the fact that $`𝐘`$ is driven by $`𝐗`$.. The reason is that, as pointed out in the introduction, $`𝐗`$ will be a (single- or multivalued) function of $`𝐘`$, if the inverse of $`G(𝐱,𝐲)`$ is single- or multivalued. On the other hand, the Kaplan-Yorke dimension of $`𝐗`$ alone may be equal to $`D_{𝐗+𝐘}`$ or smaller. It is given by a formula similar to eq.(3) but with $`\lambda _i`$ replaced by $`\lambda _i^{(𝐗)}`$. We see that $$D_𝐗<D_{𝐗+𝐘}\mathrm{iff}\lambda _1^{(𝐘)}>\lambda _{l+1}^{(𝐗)},$$ (4) where $`l`$ is determined by $`_{jl}\lambda _j^{(𝐗)}0<_{jl+1}\lambda _j^{(𝐗)}`$. If this inequality holds (together with $`\lambda _1^{(𝐘)}<0`$), we have weak synchronization in the sense of Pyragas . In the opposite case, i.e. $`\lambda _1^{(𝐘)}\lambda _{l+1}^{(𝐗)}`$, one is likely to have strong synchronization, although this might not be true due to multifractality: Due to the latter, it is possible that the box-counting dimension of $`𝐗`$ is strictly smaller than that of $`𝐗+𝐘`$, although the equality holds for the Kaplan-Yorke (i.e., information) dimensions. In such a case $`\mathrm{\Psi }`$ cannot be smooth, but regions where $`\mathrm{\Psi }`$ is non-smooth might well be of measure zero . Another approach for detecting generalized synchronization is by using two identical response systems which differ only in their initial conditions. If these replicas get synchronized after some transient, their trajectories are obviously independent of the initial conditions, thus being only a function of the driver. This is most easily checked visually, e.g. by plotting the difference between two analogous components of the two replicas against time. In this way one can also check for intermittencies and long transients which can, together with finite numerical resolution, severely obscure the interpretation. ## 3 Generalized Synchronization in Real Life The above considerations depend on the availability of the exact equations of motion, and on the ability to prepare identical replicas. Neither holds for typical applications. Real signals usually consist of short segments of data contaminated by noise. Furthermore, the dynamics of the system is not known, and therefore the methods described in the previous section are not applicable. While identically synchronized systems describe the same trajectory in the phase space, the hallmark of a relationship as in eq.(2) is that any recurrence of $`𝐗`$ implies a recurrence of $`𝐘`$. If $`𝐗`$ comes exactly back to a state it had already been in before, the same must be true for $`𝐘`$. In real data, one cannot of course expect exact recurrence. We will therefore use as a criterion that whenever two states of $`𝐗`$ are similar, the contemporary states of $`𝐘`$ are also similar. In , this was implemented by making forecasts of $`𝐱_n`$ using local neighborhoods (e.g., by means of locally linear maps), and comparing the quality of these forecasts with that of forecasts based on “wrong” neighborhoods. In the latter, the nearest neighbors of $`𝐱_n`$ are replaced by the equal time partners of the nearest neighbors of $`𝐲_n`$. For reasons explained in , we prefer to use instead a measure closer to the original proposal of Rulkov et al. . But we should stress that we see no reason why our results should not be carried over to the observables used in immediately. Let us suppose we have two simultaneously measured univariate time series from which we can reconstruct $`m`$-dimensional delay vectors $`𝐱_n=(x_n,\mathrm{},x_{nm+1})`$ and $`𝐲_n=(y_n,\mathrm{},y_{nm+1})`$, $`n=1,\mathrm{}N`$. Let $`r_{n,j}`$ and $`s_{n,j}`$, $`j=1,\mathrm{},k`$, denote the time indices of the $`k`$ nearest neighbors of $`𝐱_n`$ and $`𝐲_n`$, respectively. For each $`𝐱_n`$, the squared mean Euclidean distance to its $`k`$ neighbors is defined as $$R_n^{(k)}(𝐗)=\frac{1}{k}\underset{j=1}{\overset{k}{}}\left(𝐱_n𝐱_{r_{n,j}}\right)^2$$ (5) and the Y-conditioned squared mean Euclidean distance is defined by replacing the nearest neighbors by the equal time partners of the closest neighbors of $`𝐲_n`$, $$R_n^{(k)}(𝐗\|𝐘)=\frac{1}{k}\underset{j=1}{\overset{k}{}}\left(𝐱_n𝐱_{s_{n,j}}\right)^2.$$ (6) If the point cloud $`\{𝐱_n\}`$ has average squared radius $`R(𝐗)=R^{(N1)}(𝐗)`$ and effective dimension $`D`$ (for a stochastic time series embedded in $`m`$ dimensions, $`D=m`$), then $`R_n^{(k)}(𝐗)R(𝐗)`$ for $`kN`$. More precisely, we expect $$R_n^{(k)}(𝐗)/R(𝐗)(k/N)^{2/D},$$ (7) where $`D`$ is the dimension of the probability measure from which the points $`𝐱_n`$ are drawn. Furthermore, $`R_n^{(k)}(𝐗\|𝐘)R_n^{(k)}(𝐗)R(𝐗)`$ if the systems are strongly correlated, while $`R_n^{(k)}(𝐗\|𝐘)R_n^{(k)}(𝐗)`$ if they are independent. Accordingly, we can define an interdependence measure $`S^{(k)}(𝐗\|𝐘)`$ as $$S^{(k)}(𝐗\|𝐘)=\frac{1}{N}\underset{n=1}{\overset{N}{}}\frac{R_n^{(k)}(𝐗)}{R_n^{(k)}(𝐗\|𝐘)}.$$ (8) Since $`R_n^{(k)}(𝐗\|𝐘)R_n^{(k)}(𝐗)`$ by construction, we have $$0<S^{(k)}(𝐗\|𝐘)1.$$ (9) Low values of $`S^{(k)}(𝐗\|𝐘)`$ indicate independence between $`𝐗`$ and $`𝐘`$, while high values indicate synchronization (becoming maximal when $`S^{(k)}(𝐗\|𝐘)1`$). The opposite interdependence $`S^{(k)}(𝐘\|𝐗)`$ is defined in complete analogy. It is in general not equal to $`S^{(k)}(𝐗\|𝐘)`$. If $`S^{(k)}(𝐗\|𝐘)>S^{(k)}(𝐘\|𝐗)`$, i.e. if $`𝐗`$ depends more on $`𝐘`$ than vice versa, instead of assuming a causal relationship, we just say that $`𝐘`$ is more “active” than $`𝐗`$. As was argued in , high activity is mainly due to a large effective dimension $`D`$, on the typical length scale set by the distances $`\|𝐱_n𝐱_{r_{n,k}}\|`$ and $`\|𝐱_n𝐱_{s_{n,k}}\|`$. The second interdependence measure to be used in this work was also introduced in . In eq.(8) we essentially compare the $`𝐘`$-conditioned mean squared distances to the unconditioned m.s. nearest neighbor distances. Instead of this, we could have compared the former to the m.s. distances to random points, $`R_n(𝐗)=(N1)^1_{jn}(𝐱_n𝐱_j)^2`$. Also, in ergodic theory often geometric averages are more robust and more easy to interpret than arithmetic ones. Therefore, let us use the geometrical average in the analogon of eq.(8), and define $$H^{(k)}(𝐗\|𝐘)=\frac{1}{N}\underset{n=1}{\overset{N}{}}\mathrm{log}\frac{R_n(𝐗)}{R_n^{(k)}(𝐗\|𝐘)}$$ (10) This is zero if $`𝐗`$ and $`𝐘`$ are completely independent, while it is positive if nearness in $`𝐘`$ implies also nearness in $`𝐗`$ for equal time partners. It would be negative if close pairs in $`𝐘`$ would correspond mainly to distant pairs in $`𝐗`$. This is very unlikely but not impossible. Therefore, $`H^{(k)}(𝐗\|𝐘)=0`$ suggests that $`𝐗`$ and $`𝐘`$ are independent, but does not prove it. This (and the asymmetry under the exchange $`𝐗𝐘`$) is the main difference between $`H^{(k)}(𝐗\|𝐘)`$ and mutual information. The latter is strictly positive whenever $`𝐗`$ and $`𝐘`$ are not completely independent. As a consequence, mutual information is quadratic in the correlation $`P(𝐗,𝐘)P(𝐗)P(𝐘)`$ for weak correlations ($`P`$ are here probability distributions), while $`H^{(k)}(𝐗\|𝐘)`$ is linear. Thus $`H^{(k)}(𝐗\|𝐘)`$ is more sensitive to weak dependencies which might make it useful in applications. Also, it should be easier to estimate than mutual informations which are notoriously hard to estimate reliably. ## 4 Numerical Examples The aim of this section is to see numerically whether there exists any relationship between the ‘activity’ defined in the last section, and a driver/response relationship. In principle there should exist such a relationship, since we have argued that the system with higher dimension should be more active, and usually the response does have higher dimension. This would agree with the conclusion of , and contradict . But it is well known that observed attractor dimensions can be quite different from real ones, in particular if one has only a finite amount of noisy data and weakly coupled systems . In order to obtain results which can be easily compared to those of , we study the same systems as these authors. ### 4.1 Lorenz Driven by Rössler As a first example, we studied the unidirectionally coupled systems proposed in reference . The driver is an autonomous Rössler system with equations: $`\dot{x}_1`$ $`=`$ $`\alpha \{x_2+x_3\}`$ $`\dot{x}_2`$ $`=`$ $`\alpha \{x_1+0.2x_2\}`$ (11) $`\dot{x}_3`$ $`=`$ $`\alpha \{0.2+x_3(x_15.7)\}`$ which drives a Lorenz system in which the equation for $`\dot{y}_2`$ is augmented by a driving term involving $`x_2`$, $`\dot{y}_1`$ $`=`$ $`10(y_1+y_2)`$ $`\dot{y}_2`$ $`=`$ $`28y_1y_2y_1y_3+Cx_2^2`$ (12) $`\dot{y}_3`$ $`=`$ $`y_1y_2{\displaystyle \frac{8}{3}}y_3.`$ The unidirectional coupling is introduced in the last term of the second equation, the constant $`C`$ being its strength. Notice that $`x_2`$ enters quadratically in the coupling, whence it cannot be written as a univalent function of $`y_1,y_2,y_3`$, and $`\dot{y}_2`$. Thus we do not have a strict argument telling us that $`D_𝐘D_𝐗`$, but the latter seems extremely likely. As in reference the parameter $`\alpha `$ is introduced in order to control the relative frequencies between the two systems. The differential equations were iterated, together with the equations for the tangent vectors, by using fourth and fifth order Runge-Kutta algorithms with $`\mathrm{\Delta }t=0.0030.009`$. This was checked to yield numerically stable results, while larger $`\mathrm{\Delta }t`$ and/or a third order algorithm would have given different results. In order to eliminate transients, the first $`10^6`$ iterations were discarded. From the increase of the tangent vectors during the following $`10^6`$ iterations we obtained the Lyapunov exponents. Delay vectors with delay $`\tau =0.3`$ and embedding dimensions 4 and 5 were constructed from $`x_1(t)`$ and $`y_1(t)`$. This delay corresponds to roughly 1/4 of the average period of the Lorenz equations. All time sequences had length $`N=5000`$. In order to check for stability and for very long transients, all computations were repeated several times with different initial conditions. For the parameters considered here, the Lyapunov exponents of the Rössler are $`\lambda _i0.09\alpha ,\mathrm{\hspace{0.33em}0},5\alpha `$; and the ones of the Lorenz without coupling are $`\lambda _i0.84,\mathrm{\hspace{0.33em}0},14.5`$. Figure 1 shows the maximum Lyapunov exponent of the driven Lorenz system, as a function of the coupling strength $`C`$. The continuous curve shows the result for $`\alpha =6`$, the broken one corresponds to $`\alpha =10`$. For generalized synchronization, the maximum Lyapunov exponent should be negative. When $`\alpha =10`$, this is observed for $`2.1<C<2.7`$ and for $`c>2.9`$. For all values of $`C`$ considered in fig. 1, the maximal Lyapunov exponent of the driven system is larger than the smallest one of the driver, and therefore we have only weak synchronization. Figure 2 shows the interdependencies $`S(𝐗\|𝐘)`$, $`S(𝐘\|𝐗)`$, $`H(𝐗\|𝐘)`$ and $`H(𝐘\|𝐗)`$ for $`\alpha =6`$ and $`\alpha =10`$, and for different number of nearest neighbors. The embedding dimension was $`m=5`$. In all panels, lines with crosses (lower curves) are for $`S`$ and lines with squares (upper curves) are for $`H`$; the dark lines denote $`(𝐘\|𝐗)`$ interdependencies and the grey lines are for $`(𝐗\|𝐘)`$. All interdependencies rise with the coupling strength, with only few exceptions. These exceptions (at $`C2.5`$ for $`\alpha =6`$ and at $`2.3<C<2.8`$ for $`\alpha =10`$) occur exactly at places where the maximal Lyapunov exponent of the driven system is non-monotonic. Thus the dependencies are monotonic functions of the Lyapunov exponent. The measure $`S`$ is more sensitive to the sign of the Lyapunov exponent than is $`H`$, as seen from the sharper increase of $`S`$ when the Lyapunov exponent passes through zero and the systems synchronize. In reference this system was studied only for $`\alpha =6`$ and $`C=8`$. The latter corresponds to very strong coupling. It was found that the interdependence of $`𝐘`$ from $`𝐗`$ was larger than vice versa. This was taken as a proof that in general the response depends more on the driver than vice versa, and it was proposed that this result could be used as a general method to detect driver/response relations. Our results with $`S`$ and $`H`$ agree perfectly with those of , if we keep the same values for $`\alpha `$ and $`C`$, and use $`k20`$, i.e. for large neighborhoods. In case of $`S`$ we find $`S(𝐘\|𝐗)S(𝐗\|𝐘)`$ for all $`k`$ and $`\alpha `$, provided $`C>5`$. Finally, we also find the same inequality $`S(𝐘\|𝐗)>S(𝐗\|𝐘)`$ for very small couplings (below the synchronization threshold), while the opposite inequality sometimes holds for intermediate $`C`$. A more consistent picture is seen in the behavior of $`H(𝐗\|𝐘)`$ and $`H(𝐘\|𝐗)`$. Except in the case $`\alpha =6,C>4`$, and $`k20`$, we always found $`H(𝐗\|𝐘)>H(𝐘\|𝐗)`$, in agreement with the prediction of . This inequality is most pronounced for small values of $`C`$. Our results can be understood by the following heuristic arguments: * For strong couplings the two systems are so strongly synchronized that the differences in interdependence are small, and we can predict $`𝐗`$ from $`𝐘`$ essentially as well as $`𝐘`$ from $`𝐗`$. * The theoretical predictions $`H(𝐗\|𝐘)>H(𝐘\|𝐗)`$ and $`S(𝐗\|𝐘)>S(𝐘\|𝐗)`$ are based on the limiting behavior for small neighborhoods. It is thus not too surprising that they can be violated for large values of $`k`$. * For uncoupled systems ($`C=0`$) one has $`S(𝐗\|𝐘)<S(𝐘\|𝐗)`$ if $`D_𝐗<D_𝐘`$ and vice versa. Notice that this (which is easily obtained from eq.(7)) is the opposite of what we expect if $`D_𝐗<D_𝐘`$ holds only due to the coupling. In our case, $`D_{\mathrm{Roessler}}2.018<D_{\mathrm{Lorenz}}2.058`$. This explains why $`S(𝐗\|𝐘)<S(𝐘\|𝐗)`$ for very small $`C`$ ($`C<1`$). In contrast, $`H(𝐗\|𝐘)=H(𝐘\|𝐗)=0`$ for uncoupled systems, whence no such problem exists for $`H`$. Thus we expect that the behavior of $`H(𝐗\|𝐘)`$ at small couplings is easier to interpret than the behavior of $`S`$ which should depend nontrivially on $`k`$ and $`N`$. This is precisely what we found. ### 4.2 Two Coupled Henon Maps As a second example we studied two unidirectionally coupled Henon maps similar to the ones proposed in , with equations $`x_1^{}`$ $`=`$ $`1.4x_1^2+b_1x_2`$ $`x_2^{}`$ $`=`$ $`x_1`$ (13) for the driver, and $`y_1^{}`$ $`=`$ $`1.4(Cx_1y_1+(1C)y_1^2)+b_2y_2`$ $`y_2^{}`$ $`=`$ $`y_1`$ (14) for the response. Again we discarded the first $`10^6`$ iterations, and obtained Lyapunov exponents from the next $`10^6`$. Interdependencies were then estimated from $`N=5000`$ iterations, using 3-dimensional delay vectors. As in the previous example, the stability of the results was checked by starting from different initial conditions. For calculating the interdependencies we used $`k=10`$ nearest neighbors. No significant changes were observed for other values of $`k`$. The constants $`b_1`$ and $`b_2`$ were both set to 0.3 when analyzing identical systems, and to 0.3, 0.1 when analyzing non-identical ones. Furthermore, in all cases we also studied how the results changed if white measurement noise was added either to the driver, to the response, or to both. #### 4.2.1 Identical systems We first studied the case $`b_1=b_2=0.3`$. This is the “canonical” value for the Henon map, for which $`\lambda _1=0.4192`$ and $`D=1.26`$. One easily sees that $`(y_1,y_2)=(x_1,x_2)`$ is a solution of eqs.(13,14). Thus we can have identical synchronization , but due to the asymmetry of the coupling we cannot rule out non-identical (generalized) synchronization either. Figure 3 (solid line) shows the maximum Lyapunov exponent of the response system. It becomes negative for couplings larger than 0.7, when identical synchronization between the systems takes place. But it is also slightly negative for $`0.47<C<0.52`$. Plotting differences $`x_1y_1`$, e.g., one sees that there is no identical synchronization in this window. But making a cut through the attractor of the combined system, by plotting e.g. pairs $`(x_2,y_2)`$ whenever $`\|x_1\|<10^4`$, one sees a fractal structure which clearly shows that there is no identical synchronization. On the other hand, $`\lambda _1^{(𝐘)}<0`$ which leaves only the possibility of generalized synchronization. In this window $`(0.47<C<0.52)`$, the Kaplan-Yorke formula gives $`D2.25`$, showing that this synchronization is weak. For $`C>0.7`$, the Kaplan-Yorke formula does of course not apply to the combined system and $`D=D_{\mathrm{Henon}}=1.26`$. Interdependencies are shown in fig. 4A and fig. 4D. As expected, we see that $`S`$ and $`H`$ both rise sharply at $`C0.7`$, where identical synchronization sets in. The fact that synchronization is perfect is seen from the fact that $`S(𝐘\|𝐗)=1`$ for $`C0.7`$. In contrast, we do not see any anomaly for $`C0.5`$, showing again that the synchronization at $`c0.5`$ is very weak indeed. For $`C<0.7`$ we see that $`H(𝐗\|𝐘)>H(𝐘\|𝐗)`$, in agreement with our general prejudice that the response has higher dimension and is thus more active. Although not so pronounced, this difference is also seen for $`S`$. We also analyzed how both measures changed with the inclusion of measurement noise ($`S/N`$ amplitude ratio $`25\%`$). Figs. 4B and 4E show the results when the noise is added to the driver and figs.4C and 4F when it is added to the response. In general, we expect of course a decrease of any dependences when noise is added. Formally, we have to discuss how $`R^{(k)}(𝐗\|𝐘)`$, $`R^{(k)}(𝐘\|𝐗)`$, $`R^{(k)}(𝐗)`$, and $`R(𝐗)`$ change if noise is added to $`𝐗`$. * $`R(𝐗)`$ changes very little, since $`S/N1`$; * $`R^{(k)}(𝐗)`$ increases strongly, since the dimension of the noisy time series is large; * $`R^{(k)}(𝐗\|𝐘)`$ increases little if $`𝐗`$ and $`𝐘`$ are weakly dependent (it is already large), but increases even more than $`R^{(k)}(𝐗)`$ if $`𝐗`$ and $`𝐘`$ are strongly dependent. * The last is also true for $`R^{(k)}(𝐘\|𝐗)`$. From these we see that $`H(𝐘\|𝐗)`$ should decrease roughly as much as $`H(𝐗\|𝐘)`$, and this decrease should be strongest if $`𝐗`$ and $`𝐘`$ are fully synchronized. In contrast, $`S(𝐗\|𝐘)`$ should decrease much less (or even increase) if noise is added to $`𝐗`$, and if $`𝐗`$ and $`𝐘`$ are weakly dependent. If $`𝐗`$ and $`𝐘`$ are fully synchronized, adding noise to $`𝐗`$ suppresses $`S(𝐘\|𝐗)`$ much more than $`S(𝐗\|𝐘)`$ since the strong increase of $`R^{(k)}(𝐘\|𝐗)`$ is then not compensated by any change of $`R(𝐘)`$. Adding noise to $`𝐘`$ can be discussed similarly. All these predictions are fully verified in fig.4. Notice that measurement noise can reverse the general inequality $`H(\mathrm{driver}\|\mathrm{response})>H(\mathrm{response}\|\mathrm{driver})`$, as seen e.g. in fig.4E for $`C>0.7`$. In general, $`S`$ seems to be less robust against measurement noise than $`H`$. #### 4.2.2 Non identical systems Figures 5A and 5D show dependencies for different $`b`$-parameters ($`b_1=0.3,b_2=0.1`$) where identical synchronization is impossible, and where the driver has higher dimension than the uncoupled response. In this case, the interdependencies do not increase as sharply as in the previous case, and they reach lower values. With both measures we see an increase between $`C=0.10.4`$ in agreement with the negative values of the maximum Lyapunov exponent for these coupling strengths (see fig.3). As in reference , we found $`S(𝐗\|𝐘)>S(𝐘\|𝐗)`$ and $`H(𝐗\|𝐘)>H(𝐘\|𝐗)`$. These inequalities still hold when adding measurement noise to the driver (fig.5B,E). But when the noise is added to the response (fig.5C,F), only the inequality for $`H`$ survives, while that for $`S`$ is reversed. These dependences on noise can be discussed in complete analogy with the previous case of identical systems. The situation changes slightly if the uncoupled response has higher dimension than the driver, as in the case $`b_1=0.1,b_2=0.3`$ shown in fig. 6. The panels of this figure show more structure than those of fig.5, mainly since also the Lyapunov exponent shows more structure (see fig.3). From fig. 6F we see that there is a parameter window, $`0.4C0.6`$, where $`H(𝐘\|𝐗)>H(𝐗\|𝐘)`$ after adding noise to $`𝐘`$. This is not yet understood, but all other features in this plot can be understood heuristically along the lines discussed above. ## 5 Conclusion In this work we studied the possibility of predicting driver/response relationships from asymmetries in nonlinear interdependence measures. More precisely, we studied two particular interdependence measures, introduced in , and applied them to simple asymmetrically coupled strange attractors. In contrast to previous works, we find that such predictions are not always reliable, although we agree with that they would be possible for ideal (noise-free, infinitely long) data. We agree with as far as one of their numerical examples is concerned, but we show that this was a mere coincidence, and cannot be generalized. Instead, we confirmed the conjecture of that asymmetries in interdependence measures reflect mainly the different degrees of complexity of the two systems at the level of resolution at which these measures are most sensitive. For practical applications, this is not the infinitely fine level at which theoretical arguments like those of apply. The latter predict correctly that the response is more complex, i.e. has a higher Kaplan-Yorke dimension. But this argument can become irrelevant even for the extremely simple toy models which we studied in the present paper. It should be even less relevant in realistic situations where all sorts of noise, non-stationarity, and shortness of data present additional limitations. Nevertheless, we propose that asymmetries of measured interdependencies can be very useful in understanding coupled systems. Indications for this were given in . All these papers were dealing with neurophysiology. Even if no causal relationships can be deduced from such asymmetries, it was found in that the resulting patterns are closely related to clinical observations, and could e.g. contribute to a more precise localization of epileptic foci and might be useful for predicting epileptic seizures. An unexpected result of our study is that for the simulations performed the measure $`H(𝐗\|𝐘)`$ which had not done very well in preliminary tests is actually more robust and easier to interpret than the measure $`S(𝐗\|𝐘)`$ which was mostly used in . However, this should not be directly extended to real life data. A more systematic comparative study with a large database of EEGs from epilepsy patients is under way.
no-problem/9907/astro-ph9907047.html	ar5iv	text	# On the origin of the Tully-Fisher relation ## 1. INTRODUCTION The Tully-Fisher relation (hereafter TF: Tully & Fisher 1977) is one of the most basic relations in spiral galaxies, and would be a clue to understand the origin of disk galaxies. The TF relation has been roughly considered as a product of the virial theorem and a nearly constant mass-to-light ratio since its discovery. Recently, the origin of the TF relation is discussed by, e.g., Silk (1997) and Mo, Mao & White (1998). In their semi-analytical approach, Mo, Mao & White (1998) succeeded in reproducing the TF relation, assuming a constant mass-to-light ratio and empirical profiles of disks and haloes. Heavens & Jimenez (1999) took the same approach but including an empirical star formation model, and reproduced the TF relation in four passbands simultaneously. In these semi-analytical approaches, however, observed features of galaxies, such as the exponential profile and flat rotation curve, were not constructed as the results of simulations, but assumed a priori. Steinmetz & Navarro (1999) performed direct simulations of galaxy formation within the cosmological context, and also succeeded in getting the slope and scatter of the TF relation. They treated the volume much larger than the scale of galaxies, and considered environmental effects (e.g., tidal field and infall/outflow of mass). However, due to the complicated behaviors of these effects, it still remains unknown what physics produces the TF relation. Based on these arguments, we simulate formation and evolution of galaxies with the $`N`$-body/SPH method including cooling, star formation and their feedback of energy, mass and metals to the ISM. Using the similar method, Katz (1992) and Steinmetz & Müller (1994, 1995) succeeded in constructing internal structures as observed in spiral galaxies, e.g., the exponential density profile, flat rotation curve, and distributions of stellar age and metallicity in the bulge, disk and halo. In contrast to their simulations of single galaxy formation, we consider formation of many galaxies with different masses and spin parameters. In order to investigate what physics produces the TF relation, we concentrate on formation of isolated galaxies with different initial conditions. We note that as well as the TF relation, observed features of spiral galaxies, such as the exponential light-profiles and flat rotation curve, are reproduced in our simulations, which were assumed a priori in past semi-analytical approaches. ## 2. METHODS ### 2.1. Numerical Methods We use a GRAPE-SPH code, a hybrid scheme of the smoothed particle hydrodynamics (SPH) and the $`N`$-body integration hardware GRAPE-3 (Sugimoto et al. 1990), the code which was firstly discussed by Steinmetz (1996) and developed further by us (Koda, Wada & Sofue 1999). The code is implemented to treat three-component systems: the gas, stars and dark mater, and calculate gravitational and hydrodynamical forces, which are symmetrized to satisfy the Newton’s third law. We consider the radiative and inverse Compton cooling of the gas which is optically thin and in collisional ionization equilibrium of H and He ($`X_H=0.76`$, $`X_{He}=0.24`$; see Katz et al. 1996). We also take phenomenological models of star formation and their feedback to mimic real galaxy formation. The star formation model adopted is basically the same as those discussed in Katz (1992), Navarro & White (1993) and Steinmetz & Müller (1994, 1995). Stars are formed in regions which are locally contracting and Jeans-unstable at a rate given by $`\dot{\rho _{}}=c_{}\rho _{\mathrm{gas}}/\mathrm{max}(\tau _{\mathrm{dyn}},\tau _{\mathrm{cool}})`$. Here, $`\rho _{}`$, $`\rho _{\mathrm{gas}}`$, $`\tau _{\mathrm{dyn}}`$ and $`\tau _{\mathrm{cool}}`$ are the densities of stars and the gas, local dynamical and cooling timescales, respectively. In most cases, $`\tau _{\mathrm{dyn}}`$ is larger than $`\tau _{\mathrm{cool}}`$, and the star formation timescale is typically $`20\tau _{\mathrm{dyn}}`$ under the adopted parameter $`c_{}=0.05`$. When one third of the mass in a gas particle is transformed into stars, we create a new collisionless star particle, which inherits the position and velocity of its parent gas particle. We take a simple model of feedback. Massive stars with $`8M_{}`$ are assumed to release energy, mass and metals into the surrounding gas at a constant rate through stellar wind (SW) and type II supernovae (SNII). All of the massive stars then become white dwarfs, and $`15\%`$ of them results in type Ia supernovae (SNIa, Tsujimoto et al. 1995). In Table 1, we list total released energy, mass and metal per star, and the periods of feedback (see Nomoto et al. 1997a & 1997b; Yoshii et al. 1996). We can estimate the number of stars with $`8M_{}`$, in a star particle, with the initial mass function (IMF) of Salpeter (1955). Released energy is provided into the surrounding gas as thermal energy. Since stars are formed and release energy at dense gas regions, where feedback energy is soon radiated away, feedback does not so much affect the simulations in our model. ### 2.2. Initial Conditions We simulate formation and evolution of galaxies from $`z=25`$ to $`0`$ in the CDM cosmology ($`\mathrm{\Omega }_0=1`$, $`h=0.5`$). Our initial conditions are similar to those of Katz (1992) and Steinmetz & Müller (1994, 1995), but we consider 14 isolated spheres, on which small scale CDM fluctuations are superimposed with the Zel’dovich approximation (Zel’dovich 1970). The densities of the spheres are enhanced above the background field by $`\delta \rho /\rho =0.25`$. We normalize the CDM spectrum so that the rms fluctuation in a sphere of radius $`8h^1\mathrm{Mpc}`$ becomes equal to $`\sigma _8=0.63`$ at $`z=0`$. The spheres are rigidly rotating, and following the reduced Hubble expansion (see Steinmetz & Müller 1995). Two free parameters, mass and spin parameter, are listed in Table 2. The gas and dark matter are represented by the same number of particles, and their mass ratio is set to $`1/9`$. The initial temperature of the gas is set to 70 K of the cosmic microwave background at $`z=25`$. The mass of a gas particle varies between $`2.4\times 10^6`$ and $`1.9\times 10^7M_{}`$ according to the system mass considered (Table 2). The mass of a dark matter particle varies between $`2.1\times 10^7`$ and $`1.7\times 10^8M_{}`$. Low resolution may cause an artificial heating due to two-body relaxation between the SPH and dark matter particles, however this range of particle mass is small enough to exclude the artificial heating effect (Steinmetz & White 1997). The gravitational softenings are taken to be $`1.5\mathrm{kpc}`$ for gas and star particles, and $`3\mathrm{kpc}`$ for dark matter. The total number of particles typically becomes $`4\times 10^4`$ at the end of the simulations ($`z=0`$). ### 2.3. Data Reduction In order to compare the properties of observed and simulated galaxies, we compute the observables such as luminosity and the line-width for each “spiral galaxies” at $`z=0`$. Stellar luminosity is computed with the simple stellar population (SSP) synthesis models of Kodama & Arimoto (1997). The models provide integrated spectra along stellar isochrones corresponding to appropriate age and metallicity. We take the Salpeter’s IMF. Total magnitude $`M_I`$ of simulated galaxies is computed to sum up star particles using appropriate SSP tables, which are selected in accordance with age and metallicity of the star particles. The line-width $`W_{20}`$ is derived similarly to the observable as constructing a line-profile of gas weighted by mass, and measuring the width at $`20\%`$ level of a peak flux. The catalog of all the simulated galaxies is presented in Table 2. ## 3. RESULTS ### 3.1. Evolution and Structures of Individual Galaxies Figure 1 shows snapshots of star particles at four redshifts for the case of $`M=4\times 10^{11}M_{}`$ and $`\lambda =0.06`$. These view angles provide a face-on projection at $`z=0`$. Two clumps at $`z=4`$ are merging and form a bulge-like system between $`z=4`$ and $`3`$. After the bulge formation, the surrounding gas gradually cools and falls to form a gaseous disk, and then a stellar disk is gradually formed. The panel at $`z=2`$ shows that a disk-like structure surrounds the central bulge. Since the gas density of the inner disk is higher than that of the outer part, the star formation timescale is shorter in the inner part than in the outer part. The stellar disk is gradually formed from the inside to outside in the gas disk. The panel at $`z=1`$ shows a larger disk than that of $`z=2`$. In Figure 2 we plot the total star formation rate, total magnitude and color in a rest frame against the look-back time (and also the redshift). The star formation rate peaks at $`z3`$ at a rate of $`29M_{}\mathrm{yr}^1`$, and then declines and reaches an almost constant value of $`1M_{}\mathrm{yr}^1`$ at $`z1`$, when the stellar mass has already been about $`80\%`$ of the total disk mass. At $`z=0`$, about $`10\%`$ of the total baryonic mass remains in the gas and the rest is already in stars, which is consistent with the observations of spiral galaxies. Total magnitude also peaks near the time of the maximum star formation. Then the magnitudes in all passbands are monotonically declining until $`z=0`$ because the mean age of the stellar component becomes older and massive stars gradually die. The amount of the decline between $`z=3`$ to the present is $`1.6\mathrm{mag}`$ in $`I`$-band. We show the final snapshots of star particles in Figure 4. The $`I`$-band luminosity profile and rotation-velocity profile of the final galaxy are shown in Figure 3. Observed internal structures, such as the exponential light-profile and flat rotation curve, are well reproduced as the results of the simulation. We confirm that all the 14 simulated “spiral galaxies” attain such properties. About $`30\%`$ of total gas angular momentum transfers to that of dark matter in the period $`z=52`$, when condensed gas core with dark halos merge into a larger object (Katz & Gunn 1991; Navarro, Frenk & White 1995). After the period, the baryon disk scarcely looses the angular momentum, because of no infall of lumpy clumps in our simulations. The specific angular momenta of the baryon disks lie in the range of those in observed spiral galaxies at $`z=0`$ (Fall 1983; Contardo et al 1998). In the following discussion, we investigate a statistical property of 14 simulated galaxies, all of which have internal structures similar to observed spiral galaxies as discussed above. ### 3.2. The Tully-Fisher Relation In Figure 4, we compare the observed (open squares) and simulated (closed circles) TF plot in $`I`$-band. We use the observed data presented by Han (1992) and translate them to absolute values with distances assuming $`h=0.5`$. The slopes of the solid and dashed lines are derived by fitting to observed data (Giovanelli et al. 1997a), and the zero points are fitted by eye. In Figure 4, we notice that the simulated TF relation shows the following three points: (i) The slope of TF is well reproduced. (ii) The scatter of TF is also similar to the observations. (iii) The zero point is systematically fainter. These points are consistent with the previous results which included environmental effects, e.g., tidal field and infall/outflow of mass (Steinmetz & Navarro 1999). Here, we artificially control initial conditions of isolated spheres, and examine how final structures of “spiral” galaxies depend on the initial conditions. Based on these simulations, we discuss the TF relation in detail on three points: the zero point, slope and scatter. The zero point: Our simulated galaxies are $`1.5\mathrm{mag}`$ fainter than observed ones. We could find a possible solution of this discrepancy by changing the adopted cosmological model: Changing $`h`$ ($`\mathrm{\Omega }_0`$) would cause a vertical (horizontal) shift of the plotted points in Figure 4. If larger $`h`$ ($`>0.5`$) is taken, distance estimations for the observed galaxies become smaller, and open squares (observed galaxies) vertically shift downward in Figure 4. Larger $`h`$ also provides younger age, and therefore, brighter luminosity for galaxies, because galaxies loose luminosity with their age. It results in an upward shift of the simulated galaxies. If smaller $`\mathrm{\Omega }_0`$ ($`<1`$) is taken, the baryon fraction $`\mathrm{\Omega }_\mathrm{b}/\mathrm{\Omega }_0`$ becomes larger because $`\mathrm{\Omega }_\mathrm{b}`$ should be fixed by the Big Bang Nucleosynthesis. Since $`I`$-band luminosity is almost determined by the total baryonic mass, it results in a smaller mass-to-light ratio $`M/L`$, which shifts the filled circles (simulated galaxies) leftward in Figure 4. If we take a cosmological model with larger $`h>0.5`$ and/or smaller $`\mathrm{\Omega }_0<1`$, which is suggested by recent observations (Giovanelli et al. 1997b, Perlmutter et al. 1998), the simulated TF relation would become consistent of the observed one. Note that the zero point also depends on the redshift when the whole mass is merging into a galaxy in hierarchical cosmology (Mo, Mao & White 1998). The $`\mathrm{\Omega }_0<1`$ universe will shift the zero point brighter (upward), and also alleviate the discrepancy. The slope: As seen in Figure 4, our simulations well reproduce the slope and scatter of the TF relation. The shifted zero point may not affect the mutual comparison among the simulated galaxies. In the left panel of Figure 5, the simulated galaxies with different masses are represented by different symbols on the TF plot. The slope of the solid line is derived from observed data (Giovanelli et al. 1997a). Figure 5 clearly shows that larger total mass of galaxies result in brighter luminosity and larger line-width. The resultant slope of model galaxies on the TF plot is quite similar to the observations. The TF correlation originates in the differences of total masses of galaxies. The scatter: Figure 5 (left panel) shows that $`I`$-band luminosity are different among the galaxies with the same mass. The reason of the non-constant $`M/L`$ is found in the right panel of Figure 5. The right panel is the same plot as the left panel, but different symbols are used for different initial spin parameters $`\lambda `$. It shows that galaxies with different $`\lambda `$ distribute nearly perpendicular to the TF correlation, resulting in the scatter of the TF relation. The amplitude of the scatter depends on the range of $`\lambda `$. We took $`\lambda =0.040.10`$, which is a range if angular momenta are caused by the cosmological tidal field (Barnes & Efstathiou 1987). The role of spin parameters on the scatter can be explained in two ways: (i) Low (high) $`\lambda `$ leads to a centrifugally concentrated disk, which provides large (small) line-width $`W_{20}`$. (ii) Low (high) $`\lambda `$ results in smaller (larger) disk, and high (low) surface density of the gas disk, which controls the time-scale of star formation $`\tau _{}`$. For low $`\lambda `$, $`\tau _{}`$ is relatively short due to the high surface density, and most of stars are formed at the earlier phase of galaxy formation. Therefore, the mean age of the stellar component in low $`\lambda `$ galaxies is older than that in high $`\lambda `$ galaxies at $`z=0`$. Low $`\lambda `$ makes fainter galaxies than high $`\lambda `$ through the age of their stellar component. This effect also makes the scatter of the TF relation. ## 4. SUMMARY AND DISCUSSION We have investigated the origin of the TF relation by simulating formation of 14 galaxies with the $`N`$-body/SPH direct calculations, which includes the cooling, star formation and stellar feedback. Internal structures of spiral galaxies, such as the exponential disk profile and flat rotation, are well reproduced as the results of the simulations, while these properties were assumed a priori in past semi-analytic approaches (Mo, Mao & White 1998; Heavens & Jimenez 1999). We observed the total magnitude and gas line-widths of the simulated ”spiral galaxies”, and found that they also reproduce the slope and scatter of the TF relation, except for the zero-point. We have simulated galaxy formation from many different initial conditions, and found that the slope of the TF relation is produced by the difference of total galactic masses, and the scatter is produced by the difference of initial spin parameters. We should comment upon our model of star formation and feedback. The following issues about star formation and feedback remain for future investigations with higher resolution simulations. Luminosity in red passbands is determined firstly by the stellar amount, and secondly by their age. Simulated galaxies have the stellar amount similar to the observations, typically $`8090\%`$ of the total baryonic mass at $`z=0`$. Their luminosity hence depends on stellar age distribution, i.e., star formation history. Our star formation model is based on the Schmidt’s law ($`\dot{\rho _{}}\rho _{\mathrm{gas}}^{1.5}`$), which is an optimal but robust model if the mean rate of star formation can be determined by a mean local density of the gas. This averaging treatment of star formation would be most suitable for the current resolution $`1\mathrm{kpc}`$. In the Milky Way Galaxy, star formation is observed at molecular clouds, and further discussion about different kinds of star formation model would requires to resolve at least the scale of molecular clouds $`100\mathrm{p}\mathrm{c}`$, 10 times smaller than our spatial resolution. In our simulations, the star formation timescale is mostly set to $`20\tau _{\mathrm{dyn}}`$. Since the local dynamical timescale $`\tau _{\mathrm{dyn}}`$ is much shorter than galaxy age, the gas is transformed to stars efficiently in high redshift. Taking the model which gives longer timescales of star formation would provide brighter galaxies because galaxies which consist of younger stars are brighter. Stellar feedback would control the self-regulation mechanisms in spiral galaxies as discussed by Silk (1997). Our adopted model behaved like a minimal feedback, because feedback energy released as thermal energy which could be radiated away in a short timescale. If the energy is released as kinetic form as discussed in Navarro & White (1993), the feedback would become more efficient and could change the history of star formation. It however is not clear that feedback through kinetic energy can be occurred in the scale of $`1\mathrm{k}\mathrm{p}\mathrm{c}`$, i.e., our spatial resolution. Since supernovae remnants expand to about $`10100\mathrm{p}\mathrm{c}`$ in the inter stellar medium, $`10100`$ times higher spatial resolution than the present one should be required for the detailed modeling of the local stellar feedback. These issues remain for future investigations. As discussed in Section 3.2, the discrepancy of the $`1.5\mathrm{mag}`$ fainter zero point in $`I`$-band would be alleviated in the cosmological model of $`h>0.5`$ and $`\mathrm{\Omega }_0<1`$. We should note, however, that this discrepancy cannot be solved in the $`h=0.5`$ and $`\mathrm{\Omega }_0=1`$ cosmology with different models of star formation and feedback. The $`I`$-band luminosity declines by only $`1.6\mathrm{mag}`$ from $`z=3`$, the time of the peak luminosity, to the present. In order to shift the zero point $`1.5\mathrm{mag}`$ brighter by changing the star formation history, almost all the stars must be formed at very low redshift $`z<0.1`$ (recent 2 Gyr) in all the spiral galaxies in the local universe. Our initial conditions consider collapse of isolated spheres. Under these idealized initial conditions, the infall of sub-galactic clumps at $`z2`$, which is expected in the $`\mathrm{\Omega }=1`$ CDM universe, cannot occur, though a continuous gas supply would enhance the star formation rate at a low redshift. If we consider later gas infalls, the resultant galaxies would have brighter luminosity. Note however that as discussed in Section 3.2, our derived TF relation was consistent with the previous simulations (Steinmetz & Navarro 1999), which included the environmental effects, e.g., tidal field and infall/outflow of mass. In our isolated conditions, the simulations do not suffer from an extreme transfer of angular momentum from baryon to dark matter (Navarro, Frenk & White 1995; Navarro & Steinmetz 1997) because no merging occurs at a low redshift, Numerical computations were carried out on the GRAPE-3 system at the Astronomical Data Analysis Center of the National Astronomical Observatory, Japan. J.K. thank the Hayakawa Fund for the financial support to participate the 15th IAP meeting. We would like to thank Dr. N. Arimoto for providing us with their tables of the stellar population synthesis. We thank M. Honma for reading the manuscript and giving helpful comments.
no-problem/9907/hep-th9907023.html	ar5iv	text	# 2 The Weakly-Coupled Heterotic Theory ## 2 The Weakly-Coupled Heterotic Theory The SQFT hypothesis is particularly compelling in the context of the weakly-coupled heterotic string . Both the graviton and the gauge bosons live in this case in the ten-dimensional bulk, and their leading interactions are given by the same order in string perturbation theory (i.e. the sphere diagram). This leads to the universal relation between the four-dimensional Planck mass ($`M_P`$) and the tree-level Yang-Mills couplings , $$M_P^2M_H^2/g_{\mathrm{YM}}^2,$$ (2.1) independently of the details of compactification. If we assume that $`g_{\mathrm{YM}}o(1)`$, then the heterotic string scale ($`M_H`$) is necessarily tied to the Planck scale. Furthermore, the standard Kaluza-Klein formula for the four-dimensional gauge couplings is $$1/g_{\mathrm{YM}}^2(RM_H)^6/g_H^2,$$ (2.2) with $`R`$ the typical radius of the six-dimensional compact space and $`g_H`$ the dimensionless string coupling. Pushing the Kaluza-Klein scale ($`M_{\mathrm{KK}}R^1`$) much below $`M_H`$ requires therefore a hierarchically-strong string coupling, and invalidates the semiclassical treatment of the vacuum. Of course all radii need not be equal but, at least in orbifold compactifications, T-duality allows us to take them all larger or equal to the string length, and then the above argument forbids any single radius from becoming too large. There is actually a loophole in the above reasonning. If some compact dimensions are much larger than the heterotic string length, loop corrections to the inverse squared gauge couplings will generically grow like a power of radius . <sup>2</sup><sup>2</sup>2In special models, such as orbifolds without N=2 sectors, these large threshold corrections can be made to vanish at one-loop. The evolution of gauge couplings with energy is thus unaffected by the openning of large extra dimensions . However, since $`g_H`$ must in these models be hierarchically strong, the semiclassical string vacuum cannot be trusted. <sup>3</sup><sup>3</sup>3 Power corrections to gauge couplings have been also recently invoked as a way to speed up the unification process . It is thus logically conceivable that even though the observed low-energy gauge couplings are of order one, their tree-level values are hierarchically smaller. Since it is the tree-level couplings that enter in the relation (2.1), the heterotic string scale could thus in principle be significantly lower than the four-dimensional Planck mass . The main motivation for contemplating such possibilities in the past was the search for string models with low-energy supersymmetry broken spontaneously at tree level. Existing heterotic vacua of this type employ a string variant of the Scherk-Schwarz mechanism , which breaks supersymmetry in a way reminiscent of finite-temperature effects. The scale of (primordial) breaking is proportional to an inverse radius, so that lowering it to the electroweak scale requires the openning of extra dimensions at the TeV – a feature shown to be generic in orbifold models. <sup>4</sup><sup>4</sup>4For more general compactifications, the limit of supersymmetry restoration is also known to be a singular limit , even though there is no precise relation between the scale of symmetry breaking and some Kaluza Klein threshold. Insisting on tree-level breaking is, on the other hand, only a technical requirement – there is no reason why the breaking in nature should not have a non-perturbative origin. Furthermore, Scherk-Schwarz compactification has not so far lead to any new insights on the problems of vacuum selection and stability. Thus, there seems to be little theoretical motivation at this point for abandonning the SQFT hypothesis, and its successful unification predictions, in heterotic string theory. ## 3 Brane World and Open String Theory The story is different in the theory of (unoriented) open and closed strings, in which gauge and gravitational interactions have different origins. While the graviton (a closed-string state) lives in the ten-dimensional bulk, open-string vector bosons can be localized on defects – the worldvolumes of D(irichlet)-branes . Furthermore while closed strings interact to leading order via the sphere diagram, open strings must be attached to a boundary and thus interact via the disk diagram which is of higher order in the genus expansion. The four-dimensional Planck mass and Yang-Mills couplings therefore read $$1/g_{\mathrm{YM}}^2(R_{}M_I)^{6n}/g_I,M_P^2R_{}^nR_{}^{6n}M_I^8/g_I^2,$$ (3.3) where $`R_{}`$ is the typical radius of the n compact dimensions transverse to the brane, $`R_{}`$ the typical radius of the remaining (6-n) compact longitudinal dimensions, $`M_I`$ the type-I string scale and $`g_I`$ the string coupling constant. As a result (a) there is no universal relation between $`M_P`$, $`g_{\mathrm{YM}}`$ and $`M_I`$ anymore, and (b) tree-level gauge couplings corresponding to different sets of branes have radius-dependent ratios and need not unify. A few remarks before going on. First, we are here discussing a theory of unoriented strings, because orientifolds are required in order to cancel the tension and RR charges of the (non-compact) space-filling D-branes. Second, using T-dualities we can ensure that both $`R_{}`$ and $`R_{}`$ are greater than or equal to the string scale . This may take us either to Ia or to Ib theory (also called I or I’, respectively) – I will not make a distinction between them in what follows. Finally, it should be stressed that D-branes are the only known defects which can localize non-abelian gauge interactions in a perturbative setting. Orbifold fixed points can at most ‘trap’ matter fields and abelian vector bosons (from twisted RR sectors).<sup>5</sup><sup>5</sup>5Non-perturbative symmetry enhancement is of course a possibility, as has been discussed for instance in . The great success of the perturbative Standard Model makes one, however, reluctant to start with a theory in which $`W`$ bosons, and all quarks and leptons do not correspond to perturbative quanta. Relations (3.3) tell us that type I string theory is much more flexible (and less predictive) than heterotic theory. The string scale $`M_I`$ is now a free parameter, even if one insists that both $`g_{\mathrm{YM}}`$ and $`g_I`$ be kept fixed and of $`o(1)`$. This added flexibility can be used to remove the order-of-magnitude discrepancy between the unification and string scales . A much more drastic proposal is to lower $`M_I`$ down to the experimentally-allowed limit $`o(\mathrm{TeV})`$. Keeping for instance $`g_I`$, $`g_{\mathrm{YM}}`$ and $`R_{}M_I`$ of order one, leads to the condition $$R_{}^nM_P^2/M_I^{2+n}.$$ (3.4) A TeV string scale would then require from n=2 millimetric to n=6 fermi-size dimensions transverse to our Brane World – the relative weakness of gravity being in this picture attributed to the transverse spreading of gravitational flux. What has brought this idea <sup>6</sup><sup>6</sup>6For early discussions of a Brane Universe see . into sharp focus was (a) the realization that submillimeter dimensions are not at present ruled out by mesoscopic gravity experiments,<sup>7</sup><sup>7</sup>7That such experiments do not rule out light scalar particles, such as axions, with gravitational-force couplings and Compton wavelengths of a millimeter or less, had been already appreciated in the past . The Kaluza-Klein excitations of the graviton are basically subject to the same bound. and (b) the hope that lowering $`M_I`$ to the TeV scale may lead to a new understanding of the gauge hierarchy. Needless to say that a host of constraints (astrophysical and cosmological bounds, proton decay, fermion masses etc.) will make realistic model building a very strenuous exercise indeed. Finding type I vacua with three chiral families of quarks and leptons is already a non-trivial problem by itself . None of these difficulties seems, however, a priori fatal to the Brane World idea, even in its most extreme realization . ## 4 Renormalization Group or Classical Supergravity? Although the type I string scale could lie anywhere below the four-dimensional Planck mass,<sup>8</sup><sup>8</sup>8Arguments in favour of an intermediate string scale were given in . I will now focus on the extreme case where it is close to its experimental lower limit, $`M_Io(\mathrm{TeV})`$. Besides being a natural starting point for discussing the question of the gauge hierarchy, this has also the pragmatic advantage of bringing string physics within the reach of future acceleretor experiments. This extreme choice is at first sight antipodal to the minimal SQFT hypothesis : the MSSM is a stable renormalizable field theory, and yet one proposes to shrink its range of validity to one order of magnitude at most! Nevertheless, as I will now argue, the Brane World and SQFT scenaria share many common features when the number of large transverse dimensions in the former is exactly two . The key feature of the SQFT hypothesis is that low-energy parameters receive large logarithmic corrections, which are effectively resummed by the equations of the Renormalization Group. This running with energy can account for the observed values of the three gauge couplings, and of the mass matrices of quarks and leptons, in a way that is relatively ‘robust’.<sup>9</sup><sup>9</sup>9One must of course assume initial conditions for the RG equations, typically imposed by unification and by discrete symmetries, but there is no need to know in greater detail the physics in the ultraviolet regime. Furthermore the logarithmic sensitivity of parameters generates naturally hierarchies of scales, and has been the key ingredient in all efforts to understand the origin of the $`M_Z/M_P`$ hierarchy in the past . Consider now the Brane World scenario. The parameters of the effective Brane Lagrangian are dynamical open- and closed-string moduli. These latter, denoted collectively by $`m_K`$, include the dilaton, twisted-sector massless scalars, the metric of the transverse space etc. Their vacuum expectation values are constant along the four non-compact space-time dimensions, but vary generically as a function of the transverse coordinates $`\xi `$. For weak type-I string coupling and large transverse space these variations can be described by a Lagrangian of the (schematic) form $$_{\mathrm{bulk}}+_{\mathrm{source}}d^n\xi \left[\frac{1}{g_I^2}(_\xi m_K)^2+\frac{1}{g_I}\underset{s}{}f_s(m_K)\delta (\xi \xi _s)\right].$$ (4.1) This is a supergravity Lagrangian reduced to the n large transverse dimensions, and coupling to D-branes and orientifolds which act as sources localized at transverse positions $`\xi _s`$.<sup>10</sup><sup>10</sup>10In the general case there could be also branes extending only partially into the large transverse bulk. Our discussion can be adapted easily to take those into account. The couplings $`f_s(m_K)`$ may vary from source to source – they can for instance depend on open-string moduli – and are subject to global consistency conditions. What is important, however, to us is that they are weak in the type-I limit, leading to weak variations, $$m_K(\xi )=m_K^0+g_Im_K^1(\xi )+\mathrm{},$$ (4.2) with $`m_K^0`$ a constant, $`m_K^1`$ a sum of Green’s functions etc. For $`n=2`$ dimensions the leading variation $`m_K^1`$ grows logarithmically with the size of the transverse space, $`R_{}`$. Since our Standard Model parameters will be a function of the moduli evaluated at the position of our Brane World, they will have logarithmic sensitivity on $`M_P`$ in this case, very much like the (relevant) parameters of a supersymmetric renormalizable QFT. Similar sensitivity will occur even if $`n>2`$, as long as some twisted moduli propagate in only two extra large dimensions. Let me now discuss the validity of the approximation (4.1). The bulk supergravity Lagrangian receives both $`\alpha ^{}`$ and higher-genus corrections, but these involve higher derivatives of fields and should be negligible for moduli varying logarithmically over distance scales $`\sqrt{\alpha ^{}}`$. The source functions, $`f_s(m_K)`$, are also in general modified by such corrections – our $`\delta `$-function approximation is indeed only valid to within $`\delta \xi o(\sqrt{\alpha ^{}})`$. Such source modifications can, however, be absorbed into boundary conditions for the classical field equations at the special marked points $`\xi _s`$. The situation thus looks (at least superficially) analogous to that prevailing under the SQFT hypothesis : large corrections to low-energy parameters can be in both cases resummed by differential equations with appropriate boundary conditions. There are, to be sure, also important differences : in particular, the Renormalization Group equations are first order differential equations in a single (energy) scale parameter, while the classical supergravity equations are second-order and depend on the two coordinates of the large transverse space. The analogy between energy and transverse distance is also reminiscent of the holographic idea , considered in the context of compactification in . It is, however, important to stress that our discussion here stayed pertubative (and there was no large-N limit involved). I have just tried to argue that large string-loop corrections to the parameters of a brane action can, in appropriate settings, be calculated reliably as the sum of two superficially similar effects: (a) RG running from some low energy scale up to string scale, and (b) bulk-moduli variations over a transverse two-dimensional space of size much greater than string length. The two corresponding regimes – of renormalizable QFT and of reduced classical supergravity – are a priori different and need not overlap. ## 5 The Puzzle of Unification The logarithmic sensitivity of brane parameters on $`R_{}`$ can be used to generate scale hierarchies dynamically, exactly as with renormalizable QFT. Gauge dynamics on a given brane, for example, can become strong as the transverse space expands to a hierarchically large size, thereby inducing gaugino condensation and possibly supersymmetry breaking. Rather than discussing such scenaria further, I would now like to return to the main piece of evidence in favour of the SQFT hypothesis : the apparent unification of the Standard Model gauge couplings. Can their observed low-energy values be understood in an equally robust and controlled manner, as coming from logarithmic variations in the (real) space transverse to our Brane World ?<sup>11</sup><sup>11</sup>11For another recent idea see . I dont yet know the answer to this important question, but let me at least refute the following possible objection : since the three gauge groups of the Standard Model live at the same point in transverse space (or else matter charged under two of them would have been ultraheavy) how can real-space variations split their coupling constants apart ? This objection would have been, indeed, fatal if all gauge couplings were determined by the same combination of bulk fields. This is fortunately not the case : scalar moduli from twisted sectors of orbifolds have been, for instance, shown to have non-universal couplings to gauge fields living on the same brane . The logarithmic variations of such fields could split the three Standard Model gauge couplings apart, although it is unclear why this splitting should be in the right proportion. Acknowledgements: I thank the organizers of the Göteborg, Brussels and Bad Honnef meetings for the invitations to speak, and in particular François Englert for teaching us all that ‘physics is great fun’. I also thank G. Aldazabal, C. Angelantonj, A. Dabholkar, M. Douglas, G. Ferretti, B. Pioline, A. Sen and H. Verlinde for discussions, and the ICTP in Trieste for hospitality while this talk was being written up. Research partially supported by EEC grant TMR-ERBFMRXCT96-0090.
no-problem/9907/hep-lat9907010.html	ar5iv	text	# What we do understand of Colour Confinement ## 1 Introduction Lattice formulation is a gauge invariant regulator of non abelian gauge theories. Numerical simulations on the lattice produce from first principles regulated correlators of physical quantities. Simulations can be used to compute quantities involving low energy modes, which are out of reach of perturbation theory, such as weak interaction matrix elements, masses, matrix elements of operators in the light cone expansion. The typical problems encountered in this “phenomenological” use of lattice are the removal of the cut off (renormalization), and limitations in computer power. Simulations can also be used to test theoretical ideas and to investigate the structure of the theory. An example of investigation of “theoretical” type is the study of the mechanism of confinement. The typical difficulty in this approach is to have good theoretical ideas to test numerically, and possibly to falsify from the first principles. Apart from confinement itself there are a few fundamental issues at the background of our understanding of QCD. Among them * The $`1/N_c0`$ limit. The conjecture is that the basic properties of a gauge theory, e.g. confinement, are already contained in the limit in which the number of colours $`N_c`$ goes large, with $`g^2N_c`$ fixed. Corrections $`𝒪(1/N_c)`$ can be treated as a small perturbation. A consequence of this conjecture is that also quark loops can be viewed as a small perturbation, apart from their effect on the scale of the theory. Indeed apart from the loop with two vertices which is proportional to $`g^2N_fN_f/N_c1`$, and which enters in the $`\beta `$ function, loops with more vertices have additional factors $`1/N_c`$ and are negligible. According to this conjecture also the mechanism of confinement and the corresponding order parameter should then be marginally affected by the presence of quarks. * Understanding the ground state is also important to understand why perturbation theory works at small distances. Perturbative quantization describes interaction of quarks and gluons, and the ground state is the Fock vacuum. Quarks and gluons are not observed in nature, and the Fock vacuum is certainly not the ground state. This reflects in the renormalized perturbation expansion as a lack of convergence, even in the sense of asymptotic expansion. ## 2 Confinement in Nature. Colour is confined in nature. The expected ratio of abundance of quarks $`n_q`$ to abundance of nucleons $`n_p`$ is in the standard cosmological model $$\frac{n_q}{n_p}10^{12}.$$ (1) The experimental upper limit is $$\frac{n_q}{n_p}10^{27},$$ (2) coming from Millikan like experiments on $`1g`$ of matter. The estimate (1) is conservative. If we assume no confinement and $`T`$ is the temperature at which quarks decouple their effective mass $`m_q`$ is $`T`$. The reactions $$q+\overline{q}\mathrm{mesons}q+q\overline{q}+\mathrm{baryons}$$ are esothermic: let $`\sigma `$ the corresponding cross section and $`\sigma _0lim_{v0}\sigma v`$. Then quarks will decouple when $$n_q\sigma _0=G_N^{1/2}T^2.$$ (3) Since $`n_\gamma T^3`$ this implies $$\frac{n_q}{n_\gamma }\frac{G_N^{1/2}}{T\sigma _0}=\frac{10^{18}}{m_pm_q\sigma _0},$$ (4) $`\sigma _0m_\pi ^2`$. The ratio (1) corresponds to $`m_q=T=10GeV`$. The factor $`10^{15}`$ between the observation and the expectation cannot be explained by fine tuning of a small parameter. Like the experimental limit on the resistivity of a superconductor, it can only be explained in terms of symmetry. A suggestive idea in that direction is that vacuum is a dual superconductor . The chromoelectric field between a $`q`$ $`\overline{q}`$ pair is constrained by dual Meissner effect into an Abrikosov flux tube with energy proportional to the distance. A relativistic version of the free energy of a superconductor, which is the analog of effective action in field theory, is $$G=\frac{1}{4}F_{\mu \nu }F_{\mu \nu }+(D_\mu \phi )^{}(D_\mu \phi )+V(\phi ),$$ (5) where $$D_\mu \phi =(_\mu iqA_\mu )\phi $$ (6) is the covariant derivative and $$V(\phi )=\mu ^2\phi ^{}\phi \frac{\lambda }{2}(\phi ^{}\phi )^2$$ (7) is the effective potential. $`\mu ^2`$ and $`\lambda `$ are funtions of the temperature, and $`\mu ^2(T)>0`$ in the superconducting phase, where the potential has a mexican hat shape. Putting $`\phi =fe^{iq\theta }`$, with $`f>0`$ gives $`D_\mu \phi =iq(A_\mu _\mu \theta )fe^{iq\theta }`$. Under a gauge transformation $`A_\mu A_\mu +_\mu \mathrm{\Lambda }`$ $$_\mu \theta _\mu \theta +_\mu \mathrm{\Lambda }$$ (8) and $`\stackrel{~}{A}_\mu =A_\mu _\mu \theta `$ is gauge invariant. Moreover $$\stackrel{~}{F}_{\mu \nu }=_\mu \stackrel{~}{A}_\nu _\nu \stackrel{~}{A}_\mu =F_{\mu \nu }$$ (9) and the free energy can be rewritten as $$G=\frac{1}{4}\stackrel{~}{F}_{\mu \nu }\stackrel{~}{F}_{\mu \nu }+q^2f^2\stackrel{~}{A}_\mu \stackrel{~}{A}_\mu +_\mu f_\mu f+\mu ^2f^2\frac{\lambda }{2}f^4.$$ (10) The equations of motion are $$_\mu \stackrel{~}{F}_{\mu \nu }+q^2f^2\stackrel{~}{A}_\nu =0,(^2+\mu ^2)f=2\lambda f^3$$ (11) A static solution in the gauge $`A_0=0`$ has $`\stackrel{}{E}=0`$, $`f=\stackrel{~}{f}\sqrt{\mu ^2/2\lambda }`$ and eq.(11) reads $`\stackrel{}{}\stackrel{}{H}+q^2\stackrel{~}{f}^2\stackrel{}{\stackrel{~}{A}}`$ $`=`$ $`0,`$ (12) $`^2\stackrel{}{H}q^2\stackrel{~}{f}^2\stackrel{}{H}`$ $`=`$ $`0.`$ (13) Eq. (12) means that also in the absence of electric field there is a permanent current, or that $`\sigma =\mathrm{}`$. Eq. (13) is Meissner effect: the field $`\stackrel{}{\stackrel{~}{A}}`$ penetrates by a length $`(q\stackrel{~}{f})^1`$. At large distance from the center of a flux tube $`\stackrel{}{\stackrel{~}{A}}=0`$ or $$e\stackrel{}{\stackrel{~}{A}}𝑑\stackrel{}{x}=e\mathrm{\Phi }(B)=2\pi n,$$ (14) which is the Dirac quantization condition. Abrikosov flux tubes have monopoles at their ends. Those phenomena are a consequence of symmetry: the order parameter is $`f=\|\phi \|`$, or the non vanishing v.e.v. of a charged operator. For dual superconductivity the signal of the phase should be the v.e.v. $`\mu `$ of an operator carrying magnetic charge. ## 3 Phenomenology of confinement on the lattice. Lattice produces evidence for confinement. Wilson loops, defined as parallel transport along square contours in space time, provide the static force between $`q\overline{q}`$ pairs, in the limit of large $`T`$ $$W(R,T)\underset{T\mathrm{}}{}\mathrm{exp}(\sigma V(R)T)$$ (15) The area law observed in lattice gauge theory $$W(R,T)\mathrm{exp}(\sigma RT)$$ (16) means $$V(R)=\sigma R$$ (17) or that an infinite amount of energy is needed to pull the two particles at infinite distance from each other. $`\sigma `$ is the string tension, related to the slope of Regge trajectories. Also chromoelectric flux tubes between $`q\overline{q}`$ pairs are observed, with transverse size $`0.5\mathrm{fm}`$. The colour orientation of the chromoelectric field inside them can also be studied. Finally the collective modes of the string formed by the flux tube can be analysed. All these facts support the picture of confinement as due to dual superconductivity of vacuum. A microscopic understanding is however needed. In particular monopoles which condense in the vacuum have to be identified. ## 4 Monopoles In QED, which is a $`U(1)`$ gauge theory, magnetic charges are omitted, since they are not observed in nature. As a consequence the general solution of Maxwell’s equations can be given in terms of a vector potential $`A_\mu `$. The field strength tensor $$F_{\mu \nu }=_\mu A_\nu _\nu A_\mu $$ (18) obeys the equations $$_\mu F_{\mu \nu }=j_\nu .$$ (19) The absence of magnetic charges indentically follows from eq.(18). The dual tensor $`F_{\mu \nu }^{}\frac{1}{2}\epsilon _{\mu \nu \rho \sigma }F^{\rho \sigma }`$ is, by virtue of eq. (18) identically conserved: $$_\mu F_{\mu \nu }^{}=0.$$ (20) Eq. (20) is known as Bianchi identity. The only way to have a monopole and to preserve Bianchi identity is to introduce a singularity, and consider the monopole as the end point of an infinitely thin solenoid (Dirac string), which can be made invisible if the parallel transport of any charge $`q`$ around it is trivial or if $`\mathrm{exp}\left(iq{\displaystyle \stackrel{}{A}𝑑\stackrel{}{x}}\right)=1.`$ (21) The line integral is intended on a path which encircles the string and is equal to the magnetic flux, or to $`4\pi `$ the magnetic charge of the monopole. Eq. (21) implies $`4\pi qM=2n\pi `$ or $`qM=\frac{n}{2}`$, which is the celebrated Dirac quantization condition. As a consequence the theory becomes compact. In non abelian gauge theories, in the familiar multipole expansion the monopole term obeys abelian equation of motion, has a Dirac string and a number of independent abelian magnetic charges which is $`N1`$ for the gauge group $`SU(N)`$. The ’t Hooft-Polyakov monopole of the $`SO(3)`$ Georgi-Glashow model obeys this classification. ## 5 Monopoles in QCD To understand the monopoles in QCD we shall phrase the classification of ref. in the language of ref. , or in terms of “abelian projections”. We shall refer to $`SU(2)`$ for simplicity: extension to $`SU(N)`$ is trivial. Let $`\stackrel{}{\varphi }(x)\stackrel{}{\sigma }`$ be any local operator belonging to the adjoint representation. We define $`\widehat{\varphi }(x)=\frac{\stackrel{}{\varphi }(x)}{\left\|\stackrel{}{\varphi }(x)\right\|}`$. $`\widehat{\varphi }`$ is well defined except at zeros of $`\stackrel{}{\varphi }(x)`$. Consider the field strength tensor $`F_{\mu \nu }(x)=\widehat{\varphi }\stackrel{}{G}_{\mu \nu }(x){\displaystyle \frac{1}{g}}\widehat{\varphi }(x)\left(D_\mu \widehat{\varphi }(x)D_\nu \widehat{\varphi }(x)\right)`$ (22) where $`\stackrel{}{G}_{\mu \nu }(x)=_\mu \stackrel{}{A}_\nu _\nu \stackrel{}{A}_\mu `$ and $`D_\mu =_\mu g\stackrel{}{A}`$ is the covariant derivative. The coefficient of the second term in eq. (22) is chosen in such a way that the quadratic term $`\stackrel{}{A}_\mu \stackrel{}{A}_\nu `$ cancels with the first term. Both terms are gauge invariant under regular gauge tranformations. A gauge transformation $`U(x)`$ which brings $`\widehat{\varphi }`$ along the 3 axis $`\widehat{\varphi }=(0,0,1)`$, and diagonalizes $`\stackrel{}{\varphi }(x)\stackrel{}{\sigma }`$ is called an abelian projection. After abelian projection $`F_{\mu \nu }=_\mu A_\nu ^3_\nu A_\mu ^3`$ (23) is an abelian field. This holds in all points where $`U(x)`$ is regular. Defining the dual tensor $`F_{\mu \nu }^{}`$ as $`F_{\mu \nu }^{}={\displaystyle \frac{1}{2}}ϵ_{\mu \nu \rho \sigma }F_{\rho \sigma },`$ (24) the magnetic current $`j_\mu ^M`$ is defined as $`_\mu F_{\mu \nu }^{}=j_\nu ^M`$ (25) and is identically conserved. It is zero except at the singular points of $`U(x)`$, where monopoles can appear. Thus $`_\mu j_\nu ^M=0`$ (26) defines a magnetic $`U(1)`$ symmetry of the theory. It is not a subgroup of the gauge group since both $`F_{\mu \nu }`$ and $`j_\mu ^M`$ are colour singlet. If the vacuum is not invariant under that $`U(1)`$ symmetry, monopoles condense like the Cooper pairs and there is dual superconductivity. Notice that 1. there is a magnetic $`U(1)`$ symmetry for each operator $`\stackrel{}{\varphi }(x)`$ in the adjoint representation. 2. Under the abelian projection $`U(x)`$, due to singularities, the field strength tensor $`\stackrel{}{G}_{\mu \nu }`$ acquires a singular component $`\stackrel{}{G}_{\mu \nu }U\stackrel{}{G}_{\mu \nu }U^1+\stackrel{}{G}_{\mu \nu }^{sing}.`$ (27) The regular part can have monopole sources. $`\stackrel{}{G}_{\mu \nu }^{sing}`$ describes Dirac strings starting from the monopoles. The strategy will then be to detect condensation of different monopole species by measuring with numerical simulations across the deconfining transition a “disorder” parameter, which detects dual superconductivity. The usual order parameter is the Polyakov line. Our disorder parameter will be zero in the deconfined phase, where the order parameter is non zero, and different from zero in the confined phase, where it is zero. The concept of disorder parameter is typical of systems which admit a dual description. They usually have extended structures with non trivial topology (monopoles in QCD), which condense in the disordered phase. In a dual description these structures are described by local fields, and the definition of order and disorder is interchanged. Before giving the results and a few details on how they have been obtained, we shall present the expectations. ## 6 Expectations vs. results As we have seen dual superconductivity of the vacuum is not a well defined concept. There are infinitely many choices for the operator $`\stackrel{}{\varphi }(x)`$, and for each of them ther can or can not be condensation. What is the good choice, if any? 1. There is club of practitioners of the “maximal abelian projection”, saying that their choice is better than others. In fact with this choice the abelian field “dominates” the configurations, in particular the part of it which is produced by monopoles. This can prove convenient to attempt a construction of effective lagrangeans, but in principle does not preclude any pattern of symmetry. 2. There is a conjecture that all abelian projections are equivalent. Most probably both attitudes reflect our imperfect knowledge of the symmetry of the disordered phase. Discriminating between (A) and (B) is possible on the lattice. The results that we have obtained by a systematic study of dual superconductivity show unambiguously 1. That confinement is a transition from normal to dual superconductor ground state. This is strong evidence that the mechanism of confinement is indeed dual superconductivity of the vacuum. 2. A few different abelian projections have been analyzed. All of them show the same behaviour. The scenario (B) seems to be true. This is reassuring for the validity of the mechanism itself. If only one abelian projection would show dual superconductivity only the particles with non zero charge with respect to that $`U(1)`$ could be confined: there exist states for which that charge is zero, e.g. the gluon which is parallel to $`\widehat{\varphi }`$. Moreover the colour direction of the electric field in the flux tubes observed in the lattice should also be parallel to $`\widehat{\varphi }`$, being the electric partner of the magnetic $`U(1)`$ of the monopoles. This has been shown to be not true. Both these facts are naturally explained if the scenario (B) is at work. We conclude by giving a few details on the technique used to detect dual superconductivity. The technique has been checked in many well known systems showing order disorder duality versus traditional descriptions. ## 7 The disorder parameter An operator $`\mu `$ is constructed which carries non zero magnetic charge with respect to the $`U(1)`$ under study. Finite temperature is realized on the lattice by the usual thermodynamical recipe of having euclidean time running from $`0`$ to $`1/T`$ (the temperature), with periodic boundary conditions for barions, antiperiodic for fermions. On a lattice this is done by using a size $`N_s^3\times N_t`$, with $`N_sN_t`$ and $`a(\beta )N_t=1/T`$ ($`\beta =2N/g^2`$). By renormalization group arguments $`a(\beta )\frac{1}{\mathrm{\Lambda }_L}\mathrm{exp}(b_0\beta )`$, or $`T=\frac{\mathrm{\Lambda }_L}{N_t}\mathrm{exp}(b_0\beta )`$. The technique used to construct $`\mu `$ is inspired to ref. and and is a complicated version of the simple formula for translations in elementary quantum mechanics $`e^{ipa}\|q=\|q+a.`$ (28) In the Schrödinger representation, the field $`\stackrel{}{A}(x,t)`$ plays the rôle of $`x`$, the conjugate momentum $`\stackrel{}{\mathrm{\Pi }}(\stackrel{}{x},t)`$ the rôle of $`p`$ and $`\mu \|\stackrel{}{A}(\stackrel{}{x},t)\mathrm{exp}^{\left(i{\scriptscriptstyle d^3x\stackrel{}{\mathrm{\Pi }}(\stackrel{}{x},t)\stackrel{}{\overline{A}}(\stackrel{}{x}\stackrel{}{y})}\right)}\|\stackrel{}{A}(\stackrel{}{x},t)=\|\stackrel{}{A}(\stackrel{}{x},t)+\stackrel{}{\overline{A}}(\stackrel{}{x}\stackrel{}{y}).`$ (29) If $`\stackrel{}{\overline{A}}`$ is the field of a monopole at $`\stackrel{}{y}`$, $`\mu `$ is indeed the creation operator for a monopole, at site $`\stackrel{}{y}`$ and at time $`t`$. When inserted in the Feynman integral the operator is nothing but a linear term in the conjugated momentum added to the lagrangean and hence a shift of $`\stackrel{}{\mathrm{\Pi }}`$ at time $`t`$. Care is needed to adapt the definition (29) to a compact system, in a form which does not depend on the choice of the gauge for the classical field $`\stackrel{}{\overline{A}}`$. This can be done, and the result is, as sketched above $`\mu ={\displaystyle \frac{Z[S+\mathrm{\Delta }S]}{Z[S]}},`$ (30) with $`\mathrm{\Delta }S`$ different from zero on a hyperplane at constant $`x_0`$. Being the exponential of a sum on $`N_s^3`$ sites $`\mu `$ is subject to strong fluctuations. Numerically it is better to measure the quantity $`\rho ={\displaystyle \frac{\text{d}}{\text{d}\beta }}\mathrm{log}\mu =S_SS+\mathrm{\Delta }S_{S+\mathrm{\Delta }S}`$ (31) and to reconstruct $`\mu `$ as $`\mu =\mathrm{exp}\left[{\displaystyle _0^\beta }\rho (\beta ^{})𝑑\beta ^{}\right].`$ (32) A typical shape of $`\rho `$ is shown in fig. 2, one of $`\mu `$ in fig. 2. As is well known $`\mu `$ as an analytic function of $`\beta `$, can not vanish indentically in the deconfined phase if the number of degrees of freedom is finite. An extrapolation to $`N_s\mathrm{}`$ must be done by finite size analysis. Fig. 4 shows that a few different abelian projections behave in the same way. Fig. 4 shows that two different monopole species of $`SU(3)`$ also behave in the same way. As for the extrapolation to $`N_s\mathrm{}`$ three different ranges of $`\beta `$ are considered: 1. $`\beta \beta _c`$: there perturbative theory is at work and $`\rho cN_s+d`$, with $`c`$ a positive constant. As $`N_s\mathrm{}`$ $`\rho \mathrm{}`$ and $`\mu =0`$. 2. $`\beta \beta _c`$: there $`\rho `$ has a finite limit as $`N_s\mathrm{}`$, and $`\mu `$ is finite. 3. $`\beta \beta _c`$: there we expect $`\mu \left(\beta _c\beta \right)^\delta `$. From dimensional analysis $`\mu =N_s^\delta \mathrm{\Phi }({\displaystyle \frac{a}{\xi }},{\displaystyle \frac{\xi }{N_s}},{\displaystyle \frac{N_t}{N_s}}).`$ (33) Where $`\xi a`$, and if $`N_sN_t`$ $`\mu =N_s^\delta \mathrm{\Phi }(0,{\displaystyle \frac{\xi }{N_s}},0),`$ (34) and since $`\xi \left(\beta _c\beta \right)^\nu `$, we get the scaling law $`{\displaystyle \frac{\rho }{N_s^{1/\nu }}}=f\left(N_s^{1/\nu }(\beta _c\beta )\right).`$ (35) The quality of the scaling is shown in fig. 6 for $`SU(2)`$ and in fig. 6 for $`SU(3)`$. It gives a determination of the critical indices $`\nu `$, $`\delta `$, and of $`\beta _c`$. For $`SU(2)`$ we get $`\begin{array}{c}\nu =0.63(5)\hfill \\ \beta _c=2.30(2)N_t=4\hfill \\ \delta =0.20(8)\hfill \end{array}.`$ (39) $`\nu `$ and $`\beta _c`$ are in agreement within errors with independent determinations, and $`\nu `$ indicates a second order phase transition. For $`SU(3)`$ $`\begin{array}{c}\nu =0.33(2)\hfill \\ \beta _c=5.70(3)N_t=4\hfill \\ \delta =0.54(4)\hfill \end{array},`$ (43) indicating that the transition is first order. The method used has been tested on a number of known systems and understood in its details. The result show beyond any reasonnable doubt that dual superconductivity occurs, in different abelian projections, in connection with confinement. The part of the work reported due to our group has been done in collaboration with L. Del Debbio, G. Paffuti, P. Pieri, B. Lucini, D. Martelli in the last few years. Their contribution was determinant to the results.
no-problem/9907/cond-mat9907243.html	ar5iv	text	# Comment on “Stripes and the 𝑡-𝐽 Model” \[ \] In a recent Letter Hellberg and Manousakis (HM) studied a set of 16-site $`t`$-$`J`$ clusters with two holes using exact diagonalization to determine whether the 2D $`t`$-$`J`$ model has a striped ground state at a doping of $`x=1/8`$. Based on these diagonalizations, they concluded that the ground state of the 2D $`t`$-$`J`$ model is uniform. They observed low-lying nonuniform states with stripe-like features, but argue that these could only represent ground state configurations if one applied artificial boundary conditions. In particular, (1) they concluded that there is “no physical reason for such a simplified model \[the $`t`$-$`J`$ model\] to have a ground state with a periodic array of interfaces”; and (2) the striped states found in density matrix renormalization group (DMRG) calculations are simply an “artifact” of the boundary conditions used in DMRG. We disagree with these conclusions and believe that their analysis is flawed in several important respects. First, there is a physical mechanism which favors stripe formation in the $`t`$-$`J`$ model. As noted by HM and others , for $`J/t`$ in the relevant physical range, two holes on a $`t`$-$`J`$ cluster will form a pair with $`d_{x^2y^2}`$ symmetry. Now the point is the pairs can lower their energy further by forming a domain wall of holes with local pairing correlations, across which there is a $`\pi `$-phase shift in the antiferromagnetic background . Such an arrangement reduces the frustration of the antiferromagnetic background produced by the localized hopping of the holes and lowers their transverse kinetic energy, leading to a stabilization of the domain walls. The details of the correlations involved in this process have been studied extensively using DMRG and exact diagonalizations . The second point raised by HM concerned the question of boundary conditions. In using finite size clusters to study models which may have broken-symmetry ground states, it is often convenient to introduce a symmetry-breaking field and then study the limiting behavior by first letting the size of the system go to infinity and then letting the strength of the perturbation go to zero. We view the open end boundary conditions that we have used in this way and argue that far from being artificial, they are important for understanding the physics. Unfortunately, at present we are unable to carry out a finite-size scaling analysis to obtain the infinite size limit. Note that in the case of striped structures, the domain wall spacing rather than the lattice spacing enters in setting the lattice sizes required. Nevertheless, we have compared on numerous occasions systems of different lengths, and not seen any significant reduction in the stripe amplitudes. We have also compared a $`12\times 6`$ system with the interior $`12\times 6`$ region of a $`24\times 6`$ system at the same doping where we found the energy per hole to be the same to within about $`\pm 0.01t`$. Note also that we see the stripes regardless of whether we apply a staggered field to the edges of the system. In the absence of the staggered field, it simply takes longer for the calculation to converge. Finally, we are able to observe an essentially uniform ground state even with open boundary conditions; they occur when a next-nearest neighbor hopping $`t^{}`$ is made large enough ($`t^{}0.3t`$) . The effect of this term is to destabilize the domain walls and favor a gas of pairs. We also see stripes develop (without $`t^{}`$) as our DMRG calculation progresses from a starting point with all the holes in a clump in the center of a long system; the stripes appear spontaneously long before the holes have any probability amplitude of being near the open ends of the system. Such a calculation is shown in Figs. 1 and 2. Here, a $`16\times 6`$ system with $`J/t=0.35`$ and cylindrical boundary conditions, with eight holes, is studied with DMRG. No external fields were applied. In the initial DMRG sweep, all the holes were forced onto the center two columns of sites. Subsequently, as the finite system DMRG sweeps are performed, the system moves the holes in order to decrease the energy of the wavefunction. Since hole density is locally conserved, the essentially local DMRG sweeps move the holes slowly. Néel order develops spontaneously in the $`z`$ spin direction, since we have quantized the spins in the $`z`$ basis. As the calculation converges, this spontaneously broken spin symmetry slowly disappears, corresponding to an averaging of the overall spin direction over all possible directions. (This reduction in the local spin moments is not yet visible in the sweeps shown in Fig. 1.) Substantially before the hole density has approached either end of the system, two stripes appear spontaneously. The $`\pi `$ phase shift also appears spontaneously, and is visible in the local measurements because of the broken spin symmetry. As the calculation converges to the ground state, the $`\pi `$ phase shift becomes visible only through spin-spin correlations. The two stripes repel, and continue to move slowly apart as the sweeps progress until they are roughly equidistant from each other and the open left and right ends. Turning now to the calculations reported by HM, first, it appears that by searching for the lowest energy states in a variety of 16 site clusters, they have simply found the clusters with the largest finite size effects. In this regard, it is interesting to note that the lowest-energy labeled states (a), (b), and (c) are all on the most one-dimensional clusters of the ones they studied, those with one of the primitive translation vectors being (2,2). The fourth lowest energy state (d) is on the next most one-dimensional lattice. The authors excluded clusters which were even more one dimensional than these because of finite size effects; if they had also excluded the (2,2) clusters, their conclusions would have been quite different. The important point is that the energy differences they obtain by comparing different small clusters are far too large to relate to the subtle competition between pairing and stripes. Their lowest energy of $`0.660t`$ per site translates to an energy of $`1.979t`$ per hole. In an exact diagonalization, we find that the ground state energy of the ordinary $`4\times 4`$ periodic cluster with two holes is $`0.628t`$ per site, or $`1.72t`$ per hole. From this we see that none of the states shown in Fig. 1 (which unfortunately are not labeled according to cluster) is from the $`4\times 4`$ cluster—they were omitted by HM because their energy was too high. This difference of $`0.26t`$ per hole between these two uniform ground states of different 16 site clusters is about an order of magnitude larger than the energy difference between pairs and stripes on large systems at low doping shown in Fig. 1 of Ref. . Secondly, the stripes we have found involve correlations between two or more pairs of holes, while the 16-site clusters studied by Hellberg and Manousakis have only one pair. As noted earlier by Prelovsek and Zotos one needs at least 4 holes (i.e. two pairs) on a cluster to study stripe formation. In fact, using diagonalizations of clusters containing 4 holes, Prelovsek and Zotos found evidence for domain wall formation. We believe that meaningful information about stripe stability cannot come from calculations which involve only one pair on a cluster. In fact, the “striped” states reported in the HM Letter have quite a different origin than the stripes in the many-hole systems we have found. Just as one can combine $`p`$ and $`p`$ excited states of a single particle in a periodic box to create a spatial density wave oscillation, these authors have combined degenerate excited states of one pair in a cluster to create a standing density wave. The stripes we have found arise from correlations of pairs rather than the excited state of one pair. To summarize: there is a physical mechanism for domain wall formation in the $`t`$-$`J`$ model and the boundary conditions used in our calculations are not artificial, but, in fact, provide a simple way of introducing a symmetry breaking field. We have seen the stripes disappear when we change the model by adding a nearest-neighbor hopping $`t^{}`$, showing that the model itself does contain a mechanism for stripe formation. We have seen little change in behavior in going to larger lattices, although we have not been able to carry out a finite size scaling analysis. We believe that the calculations reported by HM are misleading because of large finite size effects on their 16-site clusters and the fact that they only have one pair. In particular, the stripe-like patterns they observe are only standing wave patterns of the motion of a pair of holes rather than the stripes arising from correlations of many pairs of holes (e.g. in the $`16\times 8`$ cluster of reference there were 16 holes or 8 pairs). Finally, it is important to note that we have not argued that stripes in the 2D $`t`$-$`J`$ model are necessarily static, although they certainly are in our calculations. There may be low energy fluctuations of the stripes which restore translational or rotational symmetry, which require larger systems and higher accuracy than we currently can manage. However, what we do believe our calculations show is that a “uniform” many-hole state which has no manifestation of static or dynamic stripes is, in fact, not a low-lying state of the $`t`$-$`J`$ model near a doping $`x=0.125`$. Indeed, most of the experiments which see stripes find dynamic stripes ; they are observable in dynamical susceptibilities, regardless of whether there are broken spatial symmetries. S.R. White acknowledges support from the NSF under grant # DMR98-70930 and D.J. Scalapino acknowledges support from the NSF under grant # DMR95-27304.
no-problem/9907/cond-mat9907249.html	ar5iv	text	# Phase transitions in the antiferroelectric-ferroelectric mixed systems ## I Introduction Mixed antiferroelectric-ferroelectric systems like PbZr<sub>1-x</sub>Ti<sub>x</sub>O<sub>3</sub> (PZT), Rb<sub>1-x</sub>(NH<sub>4</sub>)<sub>x</sub>H<sub>2</sub>PO<sub>4</sub> (RADP) ($`0x1`$) attract much attention of scientists because of their unusual properties and phase diagram peculiarities. These anomalies made PZT material useful for many technical application . In particular compositions of PZT enriched by Zr ($`x<0,1`$) appeared to be useful for application in the novel branches of electronic technique (see e.g. , and ref. therein). Theoretical description of the properties and phase diagram nearby antiferroelectric phase transition is hard problem because of complexity of description of antiferroelectric phase itself. The proximity of free energies of antiferroelectric and ferroelectric state is known to be the characteristic feature of antiferroelectrics . This proximity lead to dielectric permittivity anomalies and to the possibility to induce ferroelectric order by application of small enough electric field. The latter can be the clue of the reasons of ferroelectric phase appearance even at small $`x`$ values. Small content of ferroelectric material in the mixed system can be considered as the source of electric dipoles and induced by them random electric field. Calculation of this field magnitude allowing for dipole-dipole interaction via ferroelectric and nonferroelectric soft modes had shown that mean field is induced by the interaction via ferroelectric soft mode whereas other soft modes increase mean square field . The supposition that nonzero internal mean field can destroy antiferroelectric order by the same way as external field made it possible to describe semiquantitively the observed phase diagram . In present work we performed detailed calculations of phase diagram of mixed antiferroelectric-ferroelectric system for the region of small concentration of ferroelectric material. The obtained data fitted pretty good measured phase diagram for PZT and RADP. The application of proposed method for calculation of phase diagrams in different mixed ferroelectrics is discussed. ## II General equations Let us begin with the description of antiferroelectric phase. In Ising model free energy $`F`$ can be written as : $`F={\displaystyle \frac{1}{2}}J_S\sigma ^2+{\displaystyle \frac{1}{2}}J_AS^2p^{}E\sigma `$ (1) $`{\displaystyle \frac{T}{2}}\mathrm{ln}\left[2ch2\beta (p^{}E+J_S\sigma )+2ch2\beta J_AS\right]`$ (2) where $`\beta 1/(kT)`$, effective dipole moment $`p^{}=(\epsilon _a1)\gamma _ap/3`$, $`\epsilon _a`$ and $`\gamma _a`$ are antiferroelectric component susceptibility and Lorentz factor respectively, $`p=Zb`$ ($`\pm b`$ are the positions of two-well potential minima), $`E`$ is electric field, $`\sigma `$ and $`S`$ are respectively ferroelectric and antiferroelectric transition order parameters ($`\sigma =\overline{\xi }/b`$, $`S=\eta /b`$, $`\overline{\xi }`$ and $`\eta `$ being mean homogeneous and inhomogeneous displacements respectively); $`J_A=T_A`$ and $`J_S=T_S`$ are respectively antiferroelectric transition and Curie temperatures. Note, that in this model mean dipole moment of the unit cell has the form: $$d_r=Z(\overline{\xi }\pm \eta )$$ (3) Thus we consider the simplest two-sublattice model of antiferroelectric. ### A Order parameters Eq. (2) leads to the following expressions for the order parameters: $`\sigma `$ $`=`$ $`{\displaystyle \frac{sh2\beta (p^{}E+J_S\sigma )}{ch2\beta (p^{}E+J_S\sigma )+ch(2\beta J_AS)}}`$ (4) $`S`$ $`=`$ $`{\displaystyle \frac{sh2\beta J_AS}{ch2\beta (p^{}E+J_S\sigma )+ch(2\beta J_AS)}}`$ (5) Eqs. (2)-(4) describe antiferroelectric component, i.e. mixed system at $`x=0`$. Ferroelectric component corresponds to $`x=1`$ and its order parameter $`L`$ in Ising model has the form $$L=th\beta (E_0L+d^{}E)$$ (6) where $`E_0=T_{cmf}`$ is mean field, that defined ferroelectric transition temperature, $`d^{}`$ is effective dipole moment of ferroelectric component ions. Addition of small amount of ferroelectric component to antiferroelectric one leads to appearance of random electric field $`\epsilon `$, induced by electric dipoles of this component in the mixed system. Mean field in the mixed system has to include the contribution from ferroelectric component and from ferroelectric order parameter of antiferroelectric component namely $$\overline{\epsilon }=J_S(1x)\sigma +xE_0L$$ (7) This field as the most probable one has to define the position of maximum of random field distribution function $`f(\epsilon ,\sigma ,L)`$. It is equal to $`\delta (\epsilon J_S(1x)\sigma xE_0L)`$ in mean field approximation. Beyond of this approximation one has to take into account random field, i.e. to substitute $`E+\epsilon `$ ($`E0`$) or $`\epsilon `$ ($`E=0`$) for $`E`$ in Eqs. (4), (6) and to perform averaging with random field distribution function. This yields: $`\sigma ={\displaystyle \underset{\mathrm{}}{\overset{\mathrm{}}{}}}{\displaystyle \frac{sh(2\beta p^{}(\epsilon +E))f(\epsilon ,\sigma ,L)d\epsilon }{ch(2\beta p^{}(\epsilon +E))+ch(2\beta J_A(1x)S)}}`$ $$S=\underset{\mathrm{}}{\overset{\mathrm{}}{}}\frac{sh(2\beta J_A(1x)S)f(\epsilon ,\sigma ,L)d\epsilon }{ch(2\beta p^{}(\epsilon +E))+ch(2\beta J_A(1x)S)}$$ (8) $`L={\displaystyle \underset{\mathrm{}}{\overset{\mathrm{}}{}}}th(\beta d^{}(\epsilon +E))f(\epsilon ,\sigma ,L)𝑑\epsilon `$ One can see that in mean field approximation when $`f(\epsilon ,\sigma ,L)`$ has the form of $`\delta `$-function Eqs. (8) give Eqs. (4) at $`x=0`$ and Eq. (6) at $`x=1`$. Eqs. (8) are general expressions for order parameters dependence on concentrations and characteristics of mixed system components. Note, that in the case $`S=0`$ (both components of mixed system are ferroelectrics) the equations for $`\sigma `$ and $`L`$ have the same form as it has to be expected. ### B Transition temperatures Let us proceed now to calculations of the transition temperatures $`T_a`$ (to antiferroelectric phase) and $`T_c`$ (to ferroelectric phase) dependence on concentration. For the phase transitions of the second order these temperatures can be obtained from Eqs. (8) in the limit of $`\sigma 0`$, $`S0`$, $`L0`$. In such a limit one can simplify the integrands in Eqs. (8) substituting $`ch2\beta J_AS1`$, $`sh2\beta J_AS2\beta J_AS`$ and representing the distribution function in the form: $$f(\epsilon ,\sigma ,L)=f_0(\epsilon )\left(\frac{J_S(1x)\sigma }{p^{}}+\frac{E_0xL}{d^{}}\right)\left(\frac{df}{d\epsilon }\right)_{\sigma =L=0}$$ (9) Here $`f_0(\epsilon )`$ is the distribution function in the case of mean field absence. Keeping in mind that $`f_0(\epsilon )`$ is even parity function in aforementioned limit Eqs. (8) give: $`\sigma =\left[(1x)J_S\sigma +xE_0L{\displaystyle \frac{p^{}}{d^{}}}\right]\beta I(x)`$ $$S=(1x)\beta J_ASI(x)$$ (10) $`L=\left[(1x)J_S\sigma {\displaystyle \frac{d^{}}{p^{}}}+xE_0L\right]\beta I(x)`$ $$I(x)=\underset{\mathrm{}}{\overset{\mathrm{}}{}}\frac{f_0(y)dy}{ch^2(\beta y)}$$ (11) Integral (11) has to depend on concentration because of dependence on $`x`$ of the distribution function width. The solution of the system of Eqs. (10) yields: $$T_a=(1x)J_AI_a(x),T_c=\left[(1x)J_S+xE_0\right]I_c(x)$$ (13) or in dimensionless variables $$\tau _a=\frac{T_a}{T_A};\tau _c=\frac{T_c}{T_S};\lambda =\frac{T_{cmf}}{T_S}$$ (14) Eq. (13) can be rewritten as $$\tau _a=(1x)I_a(x);\tau _c=\left[x\lambda +(1x)\right]I_c(x)$$ (15) Here $`I_a(x)`$ and $`I_c(x)`$ are integrals (11) at $`\beta =\beta _a`$ and $`\beta =\beta _c`$ respectively. One can see, that the transition temperatures are defined by the corresponding transition temperatures and concentrations of the components of mixed system and random field distribution function characteristics. Equations (10) define concentrational dependence of transition temperature, i.e. phase diagram of mixed system for small concentration of ferroelectric component. One can see, that Eqs. (10) give the correct values in mean field approximation: $`T_a=T_A`$, $`T_c=T_S`$ at $`x=0`$ and $`T_a=0`$, $`T_c=E_0=T_{cmf}`$ at $`x=1`$. ### C Critical concentrations Random field induced by electric dipoles of ferroelectric component of mixed system can destroy completely antiferroelectric phase at some concentration of the dipoles. This concentration $`x=x_{ca}`$ is known to be critical concentration at which $`T_a=0`$, so that antiferroelectric phase exists at $`x<x_{ca}`$ only. On the other hand ferroelectric phase in the mixed system arises at $`x>x_{cc}`$ where $`x_{cc}`$ is critical concentration for ferroelectric phase transition. It has to be $`T_cT_A`$ at $`xx_{cc}`$. Therefore critical concentrations $`x_{ca}`$ and $`x_{cc}`$ in the region $`x1`$ can be obtained from the equations $`T_a=0`$ ($`x_{ca}`$) and $`T_c=T_A`$ ($`x=x_{cc}`$). Let us limit ourselves by the case $`\mathrm{\Delta }=0,`$ which is of importance both for case of PZT and RADP on the ”edges” of phase diagram. To perform the calculations on the base of Eqs. (15) we have to know concentrational dependence of integrals $`I_{a,c}(x)`$ defined by the distribution function $`f_0(y)`$. For electric dipoles as random field sources it can be represented in Gaussian form (see for details): $$f_0(y)=\frac{1}{2\sqrt{\pi c}}\mathrm{exp}\left(\frac{y^2}{4c}\right)$$ (16) $$c=x\frac{16\pi }{15}\frac{d^2}{\epsilon _0^2a^3r_c^3}\xi c_0x$$ (17) Here $`\epsilon _0`$, $`a`$ and $`r_c`$ are respectively dielectric permittivity, lattice constant and correlation radius connected with soft ferroelectric mode of antiferroelectric component of mixed system. Parameter $`\xi 1`$ takes into account dipole-dipole interaction via antiferroelectric and other nonferroelectric soft modes . Performing changing of variables in integrals $`I_{a,c}`$ ($`\beta mE=u`$), we obtain that this reduces just to renormalization of parameter $`c_0`$ in (17), giving $`c_m=c_0m^2,m=p^{},d^{}`$. To map correctly the interval $`0<z<\mathrm{}`$ into interval $`0<x<1`$ we adopt following parametrization for $`z_{c,a}`$ $`z_{c,a}=\phi _{c,a}(x)\rho ^3,\rho ={\displaystyle \frac{r_c}{a}},\phi _c(x)={\displaystyle \frac{1x}{1(1x)^\mu }},`$ (18) $`\phi _a(x)={\displaystyle \frac{x}{1x^\mu }},\mu \mathrm{}.`$ (19) Let us begin with critical concentration for antiferroelectric phase transition. At $`\beta =\beta _a`$ Eqs.(13) with respect to Eqs. (16), (17) at $`\beta _a`$ $`\mathrm{}`$ yields $$I_a(x)=\frac{1}{\sqrt{x}}\frac{1}{\sqrt{\pi }}\frac{\tau _a}{q_a};q_a=\frac{c_p^{}}{T_A}$$ (20) Substitution of Eq. (20) into the expression for $`\tau _a`$ (see Eqs. (15)) gives the following equation for critical concentration $`x_{ca}`$: $$\frac{x_{ca}}{1x_{ca}^\mu }=\frac{1}{\pi q_a^2}.$$ (21) Eq. (21) has been solved numerically during the calculation of $`\tau _a.`$ Critical concentration for ferroelectric phase transition can be obtained by the same way, but in the limit of $`T_cT_A`$, i.e. at $`\tau _cT_A/T_S`$, this ratio being larger or close to unity. Since in such a limit integral $`I_c(x)`$ can be calculated only numerically we shall obtain $`x_{cc}`$ value in the next section devoted to numerical calculations with two fitting parameters $`\lambda =T_{cmf}/T_S`$ and $`q_c=c_d^{}/T_S`$. ## III Phase diagram of mixed antiferro-ferroelectric system Phase diagram. i.e. concentrational dependence of antiferroelectric and ferroelectric phase transition temeprature was calculated with help of Eqs.(10). Integrals $`I_{c,a}(x)`$ were calculated numerically. The obtained phase diagram is represented in Fig.1 in the region of small concentration of ferroelectric component for several values of dimensionless parameters $`q_a,`$ $`q_c`$ and $`\lambda `$ (see Eq.(14)). One can see from Fig.1 that antiferroelectric phase transition critical concentration $`x_{ca}`$ increases with $`q_a`$ decrease, its values being close to those calculated on the base of Eq.(21). The obtained critical concentration of ferroelectric phase transition several times smaller than that for antiferroelectric transition. This seems to be the result of proximity of free energies and may be of $`T_S`$ and $`T_A`$ which is characteristic feature of antiferroelectrics. The dimensionless transition temperature to antiferroelectric phase $`\tau _a`$ decreases with concentration increase for all the considered parameters, whereas the behaviour of that to ferroelectric phase $`\tau _c`$ depends on the choise of the parameters values. In particular $`\tau _c`$ can increase or decrease with increase of electric dipoles concentration (see Fig.1). Physically these types of behaviour depend on distribution function width and ratio ($`\lambda `$) of ferroelectric phase transition temepratures $`T_{cmf}`$ of ferroelectric component of mixed system to that of antiferroelectric component $`T_S`$: the smaller the distribution function width and the larger $`\lambda `$ value, the larger ferroelectric phase transition temperature in the mixed system. ## IV Discussion 4.1. Let us begin with comparison of obtained data with experiment. The prominent example of mixed ferro-antiferro system is known to be PbZr<sub>1-x</sub>Ti<sub>x</sub>O<sub>3</sub> ($`0x1`$), where PbZrO<sub>3</sub> is antiferroelectric with transition temeprature from cubic paraelectric phase to rhombic antiferroelectric phase at $`T_A=500`$ K and PbTiO<sub>3</sub> is ferroelectric with transition temeprature from cubic paraelectric phase to tetragonal ferroelectric phase at $`T_{cmf}=763`$ K. Phase diagram of mixed PZT system was investigated in many papers, but mainly at $`TT_{room}`$ (see e.g. and ref. therein). Therefore critical concentration $`x_{ca}`$ is known, and $`T_a=300`$ K at $`x=(56)\%`$. One can see that the curve for $`q_a=1.7`$ gives $`T_a300`$ K at $`x=0.06`$ and $`x_{ca}0.11`$. General form of calculated $`\tau _a`$ concentrational dependence looks like experimental curve at $`x<0.1`$ ( Fig.1). There is a scattering of experimental data for ferroelectric phase transition temperature at small PbTiO<sub>3</sub> content ($`x<0.05`$) because of influence of some uncontrolled defects and impurities on $`T_c`$ value. The most probable behaviour of $`T_c`$ is its slight increase (Fig.1) . Such behaviour is similar to the calculated curve with $`q_c=1.7`$ and $`\lambda =7`$. Another important example of mixed ferro-antiferro system is Rb<sub>1-x</sub>(NH<sub>4</sub>)<sub>x</sub>H<sub>2</sub>PO<sub>4</sub> ($`0x1`$) as a representative of hydrogen-bonded material. It was studied in many details (see e.g. ). RbH<sub>2</sub>PO<sub>4</sub> (RDP) undergoes a ferroelectric phase transition at $`T_{cmf}=146`$ K, NH<sub>4</sub>H<sub>2</sub>PO<sub>4</sub> (ADP) becomes antiferroelectric near $`T_A=148`$ K. In mixed RADP system antiferroelectric transition temperature $`T_a`$ decreases with increase of RDP content and $`T_a=0`$ K near $`x=0.2`$ of RDP admixture, i.e. $`x_{ca}0.2`$ . Substitution of this $`x_{ca}`$ value into Eq.(21) gives $`q_a1`$. One can see from Fig.2 that the curve, which corresponds to $`q_a=1.175`$ fits the experimental points. Contrary to PZT there is no data about existence of ferroelectric phase transition at small ($`<10\%`$) content of RDP. The reason for this difference can be the approximately equal values of $`T_{cmf}`$ and $`T_A`$ in RADP whereas they are different in PZT. Because of this theoretical parameter $`\lambda T_A/T_S1`$ in RADP and at $`\lambda =1`$ $`\tau _c=T_c/T_S`$ curve may be lower than $`\tau _a`$ curve and only antiferroelectric phase transition exists. One can see from Fig.2, that proposed theory explaines the observed behaviour of ferroelectric phase transition decrease with increase of antiferroelectric component concentration and observed critical concentration $`x_{cc}`$ ($`q_c=1.424,\lambda =1.1`$). Note that tuneling has to be taken into account for mixed systems of KDP family. Preliminary calculations have shown, that tuneling result into change of fitting parameters values. Therefore the proposed theory describes pretty good phase diagrams of PZT and RADP for small $`x`$ values. Note, that only for small $`x`$ ($`x<0.2`$) Ti and Rb ions can be considered as impurities randomly distributed in PbZrO<sub>3</sub> and NH<sub>4</sub>H<sub>2</sub>PO<sub>4</sub> matrices so that they can be the sources of random field. 4.2. The proposed theory can be applied for another types of mixed systems. For example, mixture of two ferroelectric systems $`(F_1)_{1x}(F_2)_x`$ at $`x<0,1`$ described with the help of Eqs. (8), (10) at zero antiferroelectric phase order parameter ($`S=0`$). In this case the Equations for $`\sigma `$ and $`L`$ completely the same, i.e. $`\sigma L`$. In such situation mean field $`E_0=T_{cmf}`$ is ferroelectric phase transition temperature of $`F_1`$ component whereas $`F_2`$ component defines the form and width of random field distribution function. Because the main physical idea of the proposed model is the inhibition and distruction of long range ferroelectric or antiferroelectric order by random electric field of different sources this model can be useful for description to relaxor ferroelectrics like PbMg<sub>1/3</sub>Nb<sub>2/3</sub>O<sub>3</sub> (PMN), PbSc<sub>1/2</sub>Ta<sub>1/2</sub>O<sub>3</sub> (PST), Pb<sub>1-y</sub>La<sub>y</sub>Zr<sub>0,65</sub>Ti<sub>0,35</sub>O<sub>3</sub> (PLZT, $`y=0,08`$; 0,09). But the relaxors like PMN, PST are not solid solutions of two component, and in PLZT La ions are the impurities. Thus in all these materials there is onlu one ferroelectric component known as Burns reference phase (see and ref. therein), which is PbZr<sub>0,65</sub>Ti<sub>0,35</sub>O<sub>3</sub> in the of PLZT. The transition temperature from para- to ferroelectric phase of the reference phase $`T_{cmf}E_0`$ defines the position of random field distribution maxima whereas the substitutional disorder, vacancies of lead and oxygen, impurities like La in PLZT define the distribution function width. Namely such a model has been proposed recently . It permitted to describe the observed phase diagram of the relaxors, their dynamic and static properties peculiarities (see e.g. and ref. therein). Since the model is able to describe the distraction of long range order by any random field sources it can be useful also for description of solid solutions of ferroelectric and dielectric components, e.g. Ba(Ti<sub>1-x</sub>Sn<sub>x</sub>)O<sub>3</sub> ($`x<0,1`$). The observed in this material relaxor-like behaviour as well as change of phase transition order can be the consequence of strong random field influence as it was shown recently . Note that some residual ferroelectric domains were observed in another relaxor PLZT (9/65/35) . As the matter of fact, the dipole glass and mixed ferroglass phases (which appear in the relaxors with temperature lowering) can be considered as the reentrant phases with respect to this residual ferroelectric phase existing near the Burns temperature. Therefore the generality of proposed model gives a possibility to apply it to various mixed systems which are solid solutions of ferro- and antiferroelectrics, two ferroelectrics, ferroelectric and dielectric etc. as well as to ferroelectric relaxors.
no-problem/9907/cond-mat9907324.html	ar5iv	text	# 1 Introduction ## 1 Introduction The scaling equation of state near a critical point provides fundamental information on the critical behaviour of a thermodynamic system. For bulk Ising-like systems accurate predictions have been made recently on the basis of the $`\phi ^4`$ field theory in three dimensions. Testing these predictions by Monte Carlo simulations would be of considerable interest. Such simulations, however, are necessarily made only for finite systems and thus phenomenological concepts like finite-size scaling are needed to perform extrapolations from mesoscopic lattices. In order to test the theory in a more conclusive way it is desirable to go beyond the phenomenological finite-size scaling concept and to calculate explicitly the finite-size effects on the equation of state, i.e., on the magnetization as a function of the temperature $`T`$ and external field $`h`$ in a finite geometry. Although such field-theoretic calculations are perturbative and not exact they have been found to be in good agreement with the MC simulations. So far the calculations of thermodynamic quantities were restricted to the case of zero external field $`h`$. At finite $`h`$, only the order-parameter distribution function was calculated . In the present paper we extend these calculations to the equation of state of the three-dimensional Ising model in a finite cubic geometry at finite $`h`$. For simplicity these calculations are performed at infinite cutoff and thus our results neglect lattice effects. The latter have been shown to yield (exponentially small) non-universal non-scaling contributions. Here we focus our interest on the universal scaling part of the equation of state. Our theory predicts non-monotonic effects in the $`h`$ dependence of the equation of state which we then confirm with surprisingly good agreement by standard Monte Carlo simulations. The identification of such non-monotonic effects is of great practical importance. If, for example, the critical temperature $`T_c(L)`$ of a lattice with $`L^3`$ sites in three dimensions varies asymptotically as $`T_c(L)T_c(\mathrm{})1/L^y`$ with some correlation length exponent $`y=1/\nu `$, then a plot of the numerical $`T_c(L)`$ versus $`1/L^y`$ gives the extrapolated $`T_c(\mathrm{})`$ as an intercept. If, however, higher order terms make the curve $`T_c(L)`$ non-monotonic, then such a plot for finite $`L`$, where the non-monotonicity is not yet visible, would give a wrong estimate for $`T_c(\mathrm{})`$. Similar effects would make estimates of other quantities unreliable, and there exist examples of such estimates in the literature. ## 2 Field-theoretic calculations We consider the $`\phi ^4`$ model with the standard Landau-Ginzburg-Wilson Hamiltonian $$H(h)=_V\left[\frac{1}{2}r_0\phi ^2+\frac{1}{2}(\phi )^2+u_0\phi ^4h\phi \right]$$ (1) where $`\phi (𝐱)`$ is a one-component field in a finite cube of volume $`V=L^d`$ and $`h`$ is a homogeneous external field. We assume periodic boundary conditions. Accordingly we have $$\phi (𝐱)=L^d\underset{𝐤}{}\phi _𝐤e^{i𝐤𝐱}$$ (2) where the summation $`_𝐤`$ runs over discrete $`𝐤=\frac{2\pi }{L}𝐦`$ vectors with componets $`k_j=\frac{2\pi }{L}m_j,m_j=0,\pm 1,\pm 2,\mathrm{},j=1,2,\mathrm{},d`$ in the range $`\mathrm{\Lambda }k_j<\mathrm{\Lambda }`$, i.e., with a sharp cutoff $`\mathrm{\Lambda }`$. The temperature enters through $`r_0=r_{0c}+a_0t,t=(TT_c)/T_c`$. As pointed out recently the Hamiltonian (1) for periodic boundary conditions with a sharp cutoff $`\mathrm{\Lambda }`$ does not correctly describe the exponential size dependence of physical quantities of finite lattice models in the region $`\xi L`$. Instead of (1), a modified continuum Hamiltonian with a smooth cutoff would be more appropriate. Even better would be to employ the lattice version of the $`\phi ^4`$ theory to describe the non-scaling lattice effects of finite Ising models in the region $`\xi L`$. In the present paper, however, we shall neglect such effects by taking the limit $`\mathrm{\Lambda }\mathrm{}`$ (see below). The fluctuating homogeneous part of the order-parameter of the Hamiltonian (1) is $`\mathrm{\Phi }=V^1_Vd^dx\phi (𝐱)=L^d\phi _0`$. As previously $`\phi `$ is decomposed as $$\phi (𝐱)=\mathrm{\Phi }+\sigma (𝐱)$$ (3) where $`\sigma (𝐱)`$ includes all inhomogeneous modes $$\sigma (𝐱)=L^d\underset{𝐤\mathrm{𝟎}}{}\phi _𝐤e^{i𝐤𝐱}.$$ (4) The order-parameter distribution function $`P(\mathrm{\Phi })P(\mathrm{\Phi },t,h,L)`$ is defined by functional integration over $`\sigma `$, $$P(\mathrm{\Phi },t,h,L)=Z(h)^1D\sigma e^{H(h)},$$ (5) where $$Z(h)=_{\mathrm{}}^{\mathrm{}}𝑑\mathrm{\Phi }D\sigma e^{H(h)}$$ (6) is the partition function of system. This distribution function depends also on the cutoff $`\mathrm{\Lambda }`$ which implies non-scaling finite-size effects . From the order-parameter distribution function $`P(\mathrm{\Phi })`$ we can calculate the magnetization $$M=<\|\mathrm{\Phi }\|>=_{\mathrm{}}^{\mathrm{}}𝑑\mathrm{\Phi }\|\mathrm{\Phi }\|P(\mathrm{\Phi }).$$ (7) The functional integration over $`\sigma `$ in Eqs. (5) and (6) can only be done perturbatively . Recently a novel perturbation approach was presented . Using this approach the order-parameter distribution can be written in the form $$P(\mathrm{\Phi })=e^{H^{eff}(\mathrm{\Phi })}/_{\mathrm{}}^{\mathrm{}}𝑑\mathrm{\Phi }e^{H^{eff}(\mathrm{\Phi })}$$ (8) where the (bare) effective Hamiltonian of the Ising-like system reads $`H^{eff}(\mathrm{\Phi })`$ $`=`$ $`H_0(\mathrm{\Phi },h){\displaystyle \frac{1}{2}}{\displaystyle \underset{𝐦\mathrm{𝟎}}{}}\mathrm{ln}\{{\displaystyle \frac{Z_1[y_{0𝐦}(r_{0L})]}{Z_1[y_{0𝐦}(r_0)]}}\},`$ (9) $`H_0(\mathrm{\Phi },h)`$ $`=`$ $`L^d({\displaystyle \frac{1}{2}}r_0\mathrm{\Phi }^2+u_0\mathrm{\Phi }^4h\mathrm{\Phi }),`$ (10) with $`r_{0L}=r_0+12u_0\mathrm{\Phi }^2`$ and $`y_{0𝐦}(r)=(2L^d/3u_0)^{1/2}(r+\frac{4\pi ^2}{L^2}𝐦^2)`$. The function $`Z_1[y]`$ in Eq.(9) is defined as $$Z_1[y]=_0^{\mathrm{}}𝑑ss\mathrm{exp}(\frac{1}{2}ys^2s^4).$$ (11) After minimal renormalization at fixed dimension $`d<4`$ in the limit $`\mathrm{\Lambda }\mathrm{}`$, we obtain the finite-size scaling form of the order-parameter distribution function $$P(\mathrm{\Phi },t,h,L)=L^{\beta /\nu }p(hL^{\beta \delta /\nu },tL^{1/\nu },\mathrm{\Phi }L^{\beta /\nu }),$$ (12) where $$p(x,q,z)=\frac{\mathrm{exp}[F(x,q,z)]}{_{\mathrm{}}^{\mathrm{}}𝑑z\mathrm{exp}[F(x,q,z)]},$$ (13) with $`x=hL^{\beta \delta /\nu }`$, $`q=tL^{1/\nu }`$, $`z=\mathrm{\Phi }L^{\beta /\nu }`$ and $`F(x,q,z)`$ $`=`$ $`c_2(x,\widehat{q})\widehat{z}^2+c_4(x,\widehat{q})\widehat{z}^4xz`$ (14) $`{\displaystyle \frac{1}{2}}{\displaystyle \underset{𝐦\mathrm{𝟎}}{}}\mathrm{ln}\{{\displaystyle \frac{Z_1[y_𝐦(\stackrel{~}{r}_L(x,\widehat{q},\widehat{z}))]}{Z_1[y_𝐦(\stackrel{~}{r}_L(x,\widehat{q},0))]}}\}.`$ Here $`\widehat{q}=Q^{}t(L/\xi _0)^{1/\nu }`$ and $`\widehat{z}=(2Q^{})^\beta (\mathrm{\Phi }/A_M)(L/\xi _0)^{\beta /\nu }`$ are dimensionless variables normalized by the asymptotic amplitudes $`\xi _0`$ and $`A_M`$ of the bulk correlation length $`\xi =\xi _0t^\nu `$ at $`h=0`$ above $`T_c`$ and of the bulk order-parameter $`M_{bulk}=A_M\|t\|^\beta `$ at $`h=0`$ below $`T_c`$. The bulk parameter $`Q^{}`$ is known . The coefficients $`c_2(x,\widehat{q})`$ and $`c_4(x,\widehat{q})`$ read for $`d=3`$ $`c_2(x,\widehat{q})`$ $`=`$ $`(64\pi u^{})^1\widehat{q}\stackrel{~}{\mathrm{}}(x,\widehat{q})^{3(2\beta +1)/\nu }(1+12u^{}),`$ (15) $`c_4(x,\widehat{q})`$ $`=`$ $`(256\pi u^{})^1\stackrel{~}{\mathrm{}}(x,\widehat{q})^{34\beta /\nu }(1+36u^{}),`$ (16) where $`u^{}`$ is the fixed point value of the renormalized coupling . In three dimensions we have $`y_𝐦(\stackrel{~}{r}_L(x,\widehat{q},\widehat{z}))`$ $`=`$ $`[6\pi u^{}\stackrel{~}{\mathrm{}}(x,\widehat{q})]^{1/2}[\stackrel{~}{r}_L(x,\widehat{q},\widehat{z})\stackrel{~}{\mathrm{}}(x,\widehat{q})^2+4\pi ^2𝐦^2],`$ (17) $`\stackrel{~}{r}_L(x,\widehat{q},\widehat{z})`$ $`=`$ $`\widehat{q}\stackrel{~}{\mathrm{}}(x,\widehat{q})^{1/\nu }+(3/2)\stackrel{~}{\mathrm{}}(x,\widehat{q})^{2\beta \nu }\widehat{z}^2.`$ (18) The auxiliary scaling function $`\stackrel{~}{\mathrm{}}(x,\widehat{q})`$ of the flow parameter is determined by $`\stackrel{~}{\mathrm{}}(x,\widehat{q})^{3/2}`$ $`=`$ $`(4\pi u^{})^{1/2}[\stackrel{~}{y}(x,\widehat{q})+12\vartheta _2(\stackrel{~}{y}(x,\widehat{q}),\widehat{x})],`$ (19) $`\stackrel{~}{y}(x,\widehat{q})`$ $`=`$ $`(4\pi u^{})^{1/2}\stackrel{~}{\mathrm{}}(x,\widehat{q})^{3/21/\nu }\widehat{q},`$ (20) $`\widehat{x}`$ $`=`$ $`A_M(2Q^{})^\beta \xi _0^{\beta /\nu }(4\pi u^{})^{1/4}\sqrt{8}\stackrel{~}{\mathrm{}}(x,\widehat{q})^{\beta /\nu 3/4}x,`$ (21) where $$\vartheta _2(\stackrel{~}{y},\widehat{x})=\frac{_0^{\mathrm{}}𝑑ss^2\mathrm{cosh}(\widehat{x}s)\mathrm{exp}(\frac{1}{2}\stackrel{~}{y}s^2s^4)}{_0^{\mathrm{}}𝑑s\mathrm{cosh}(\widehat{x}s)\mathrm{exp}(\frac{1}{2}\stackrel{~}{y}s^2s^4)}.$$ (22) From Eqs. (7) and (12) we obtain the scaling form $`M(h,t,L)`$ $`=`$ $`L^{\beta /\nu }f_M(hL^{\beta \delta /\nu },tL^{1/\nu }),`$ (23) $`f_M(x,q)`$ $`=`$ $`{\displaystyle \frac{_{\mathrm{}}^{\mathrm{}}𝑑z\|z\|\mathrm{exp}[F(x,q,z)]}{_{\mathrm{}}^{\mathrm{}}𝑑z\mathrm{exp}[F(x,q,z)]}}.`$ (24) Correspondingly the asymptotic equation of state for a finite Ising-like system in the limit of zero lattice spacing can be written as $$h/M^\delta =f(hL^{\beta \delta /\nu },t/h^{1/\beta \delta })$$ (25) where $$f(x,y)=x/f_M(x,yx^{1/\beta \delta })^\delta .$$ (26) In the comparison with the MC data of the simple-cubic (sc) Ising model in Section 4, the quantities $`h`$, $`M`$ and $`L`$ are used in a dimensionless form in units of the lattice constant, see Sect. 4 of Ref.. We have taken the bulk parameters $`u^{}=0.0412,Q^{}=0.945`$ from Ref.. From Ref. we have taken the bulk amplitudes $`\xi _0=0.495,A_M=1.71`$ in units of the lattice constant (of the sc Ising model) and the bulk critical exponents $`\beta =0.3305,\nu =0.6335`$. Thus our determination of the scaling function $`f(x,y)`$ does not require a new adjustment of nonuniversal parameters. Taking the limit $`hL^{\beta \delta /\nu }\mathrm{}`$ at fixed $`t/h^{1/\beta \delta }`$, we obtain the scaling form of the bulk equation of state $$h/M^\delta =f_b(t/h^{1/\beta \delta })=f(\mathrm{},t/h^{1/\beta \delta }).$$ (27) At $`T=T_c`$ we find from Eqs. (23)-(27) $$h/M^\delta D_c=f(\mathrm{},0)=0.202$$ (28) in three dimensions (in units of the lattice constant). ## 3 Monte Carlo simulation Standard heat bath techniques were used for the Glauber kinetic Ising model, with multi-spin coding (16 spins in each 64-bit computer word). Since the effect could be seen best in lattices of intermediate size $`L80`$ for $`L\times L\times L`$ spins, memory requirements were tiny and only trivial parallelization by replication, not by domain decomposition, was used. However, thousands of hours of processor time were needed since we see the non-monotonic effects clearly in our figure if $`h/M^\delta `$ is plotted. The exponent $`\delta `$ is nearly five and thus five percent accuracy in $`h/M^\delta `$ requires one percent accuracy in the directly simulated magnetization $`M`$. Earlier simulations by the same author and algorithm did not show the non-monotonic effects at $`T=T_c`$ because they were not searched for; at that time the field-theoretical predictions presented above were not yet known. But even if they had been known it is doubtful that with the Intel Paragon used in instead of the Cray-T3E now the non-monotonic trends would have been seen in about the same computer time. Because of the limited system size, errors in the magnetization of order 1 + const/$`L^{\beta /\nu }`$ 1 + const/$`\sqrt{L}`$ are expected; in light of these errors the agreement to be presented now is surprisingly good. ## 4 Results and discussion Figure 1 compares our Monte Carlo results for $`32^3`$ and $`80^3`$ spins with our theoretical predictions, Eqs. (23)-(26), and shows good quantitative agreement; below $`T_c`$ a pronounced peak is seen in the scaling function (Fig.1a), at $`T_c`$ it is somewhat weaker (Fig.1b), above $`T_c`$ it is barely visible (Fig.1c), and far above $`T_c`$ (Fig.1d) it has vanished. The simulations, which at $`T=T_c`$ have an accuracy of the order of one and five percent for $`L=32`$ and 80, respectively, agree nicely with the theoretical predictions. (Far above $`T_c`$ therefore no simulations were made.) Thus, if one tries to determine the bulk critical amplitude of $`h/M^\delta `$ at the critical isotherm, then with varying $`L`$ at a fixed field one first gets a too small value (left border of the figures), then a too high value (peak), and then a roughly correct value (plateau in the right part of the figure). These non-monotonicities do $`not`$ vanish if we take the lattices large enough. They are part of the asymptotic scaling function and thus whatever lattice size $`L`$ we take there will be a field $`h1/L^{\beta \delta /\nu }`$ where $`h/M^\delta `$ has a maximum and near which it thus varies non-monotonically with $`L`$. We also tested the prediction of that for fixed $`T<T_c`$ the leading finite-size deviation from the bulk value of $`M`$ should vanish exponentially in $`L`$, and not with a power law $`1/L^d`$ (as predicted by perturbation theory based on the separation of the zero-mode \[4-7\]). Fig. 2a shows again non-monotonic behaviour, in both three and five dimensions. But only in three dimensions these data are accurate enough to distinguish between a tail varying exponentially and one $`1/L^d`$; Fig.2b clearly supports the theoretically predicted exponential variation. (These three-dimensional data were taken with Ito’s fast algorithm .) In order to make contact with bulk properties we also used simulations at the critical isotherm with $`1292^3`$ spins . From the simulations we obtain the bulk value of $`h/M^\delta `$ as $`D_c=0.21\pm 0.02`$. Our field-theoretic result in Eq. (28) is in very good agreement with this value. It is interesting to compare our simulation result also with other bulk theories. From series expansion Zinn and Fisher obtained $`C^c=0.299`$ for the amplitude of the bulk susceptibility at the critical isotherm $`\chi =C^c\|h\|^{\gamma /\beta \delta }`$ with $`\gamma =1.2395,\nu =0.6320`$. This leads to $`D_c=0.182`$ according to the relation $`D_c=(C^c\delta )^\delta `$. $`D_c`$ is also contained in the universal combination of amplitudes $$R_\chi =\mathrm{\Gamma }D_cA_M^{\delta 1}$$ (29) where $`\mathrm{\Gamma }`$ is the amplitude of the bulk susceptibility $`\chi =\mathrm{\Gamma }t^\gamma `$ above $`T_c`$ at $`h=0`$ and $`A_M`$ is the amplitude of the spontaneous magnetization $`M_{bulk}=A_M\|t\|^\beta `$ below $`T_c`$. Using $`\phi ^4`$ field theory at $`d=3`$ dimensions Guida and Zinn-Justin have obtained $`R_\chi =1.649`$. They also used the $`ϵ=4d`$ expansion and obtained $`R_\chi =1.674`$ at $`ϵ=1`$. Using these values for $`R_\chi `$ and the high-temperature series expansion results $`\mathrm{\Gamma }=1.0928`$, $`A_M=1.71`$ and $`\delta =(d\nu +\gamma )/(d\nu \gamma )`$ with $`\gamma =1.2395,\nu =0.6335`$, we obtain from Eq. (29) $`D_c=0.205`$ ($`d=3`$ field theory) and $`D_c=0.202`$ ($`ϵ`$-expansion), respectively, in good agreement with our theoretical result, Eq. (28), and with our MC simulation. In summary, our simulations confirmed a posteriori our theoretical predictions of Sect. 2 for the asymptotic finite-size effects in the three-dimensional Ising model. The agreement in Fig.1 is remarkable in view of the fact that the non-universal parameters of the theory were adjusted only to bulk parameters of the Ising model at $`h=0`$ and not to any finite-size MC data. Acknowledgments Support by Sonderforschungsbereich 341 der Deutschen Forschungsgemeinschaft and by NASA under contract numbers 960838, 1201186, and 100G7E094 is acknowledged. X.S.C. thanks the National Science Foundation of China for support under Grant No. 19704005, D.S. thanks the German Supercomputer Center in Jülich for time on their Cray-T3E. Figure Captions Fig. 1. Scaling plot (in units of the lattice constant, see Ref.) of $`h/M^\delta `$ versus $`x=hL^{\beta \delta /\nu }`$ below $`T_c`$ with $`t/h^{1/\beta \delta }=1.0`$ in part a, $`t/h^{1/\beta \delta }=0`$ (that means $`T=T_c`$) in part b, $`t/h^{1/\beta \delta }=1.0`$ in part c, and $`t/h^{1/\beta \delta }=1.6`$ in part d. Monte Carlo data for $`L=32`$ and $`80`$. Solid line is the theoretical prediction Eqs. (23)-(26); no Monte Carlo data are shown in part d, where $`h/M^\delta `$ is a monotonic function of external field $`x=hL^{\beta \delta /\nu }`$. Fig.2. a) Monte Carlo data for the spontaneous magnetization in units of the lattice constant in three (diamonds and plusses) and five (square) dimensions at $`T/T_c=0.99`$, versus linear lattice size $`L`$. The horizontal line $`M=0.3671`$ for three dimensions is determined from $`L=2496`$. b) Selected three-dimensional data from part a) are shown as $`M(L=\mathrm{})M(L)`$ versus $`L`$ in a semilogarithmic plot. The straight line represents an exponential decay , the two curved lines are power law decays $`1/L^2`$ and $`1/L^3`$ which fail to fit the data. The five-dimensional data of part a) were not accurate enough to distinguish between an exponential and a $`1/L^5`$ decay.
no-problem/9907/astro-ph9907390.html	ar5iv	text	# Chemically Consistent Evolutionary Synthesis Models ## 1 Introduction In principle, any stellar system with star formation (SF) going on over more than the lifetime of the most massive stars plus the cooling time of the gas is composite in terms of age and metallicity. Since this is the case for any composite stellar system like a galaxy – as opposed to star clusters – the importance to account for realistic metallicity distributions in evolutionary galaxy models is evident. For quite some time, observational evidence has been a ccumulating for finite and often very large metallicity distributions, e.g. spanning a range of more than a factor of 100 in bulges from $`(0.013)\mathrm{Z}_{}`$ (e.g. Jacoby & Ciardullo 1999). While some years ago, the focus was on super-solar metallicities e.g. in (the centers of) massive ellipticals, bulges, X-halos around ellipticals, and the hot ICM, by today, it is clear that the average metallicities in all those cases are subsolar. When averaged over $`1\mathrm{R}_\mathrm{e}`$, line strength gradients in E/S0s show $`\mathrm{Z}_{}(0.51)\mathrm{Z}_{}`$ (Carollo & Danziger 1994). For stars in bulges $`\mathrm{Z}_{}(0.30.7)\mathrm{Z}_{}`$ (e.g. McWilliam & Rich 1994), for the X-gas halos of ellipticals ASCA observations give $`0.1[\mathrm{Fe}/\mathrm{H}]0.7`$ (e.g. Loewenstein 1999). Characteristic HII region abundances (i.e. measured at $`1\mathrm{R}_\mathrm{e}`$) range from $`\mathrm{Z}\stackrel{>}{}\mathrm{Z}_{}`$ for Sa spirals down to $`\frac{1}{2}\mathrm{Z}_{}`$ for Sd galaxies (e.g. Oey & Kennicutt 1993, Zaritsky et al. 1994, Ferguson et al. 1998, van Zee et al. 1998). The sun, our reference star, stands out in metallicity among solar neighborhood stars. For F, G, K dwarfs the \[Fe/H\] distributions extend from $`0.8\mathrm{to}+0.4`$ (Rocha-Pinto & Maciel 1998), while B-stars show $`[\mathrm{O}/\mathrm{H}]=0.31`$ (Kilian-Montenbruck et al. 1994). Locally already, dwarf irregulars have metallicities in the range (2 – 30)% $`\mathrm{Z}_{}`$ (e.g. Richer & McCall 1995). The first spectra of Lyman break galaxies at $`\mathrm{z}34`$ have shown that their metallicities, derived from stellar wind features, are considerably subsolar, sometimes even sub-SMC (Lowenthal et al. 1997, Trager et al. 1997). Neutral gas in damped Ly$`\alpha `$ absorbers observed to $`\mathrm{z}>4`$ shows abundances $`3\stackrel{<}{}[\mathrm{Zn}/\mathrm{H}]\stackrel{<}{}0`$ (cf. Sect.5). ## 2 Chemically Consistent Modelling Evolutionary synthesis models start from a gas cloud of mass G, initially comprising the total mass M, with primordial abundances, give a star formation rate $`\mathrm{\Psi }(\mathrm{t})`$ and an IMF, and form the 1<sup>st</sup> generation of stars with $`\mathrm{Z}=0`$. We solve a modified form of Tinsley’s equations with stellar yields for SNII, SNI, PN, and stellar mass loss for $`\mathrm{Z}=0`$ to obtain ISM abundances and abundance ratios. The next generation of stars is formed with abundances $`Z>0`$ and abundance ratios \[X<sub>i</sub>/X<sub>j</sub>\] $`0`$, and again, the modified Tinsley equations are to be solved with yields for $`Z>0`$ and \[X<sub>i</sub>/X<sub>j</sub>\] $`0`$, … . It is seen that via the SFH galaxy evolution and stellar evolution become intimately coupled. In principle, stellar evolutionary tracks and yields would be required not only for various metallicities (and He contents), but also for various abundance ratios. The SFH – if short and burst-like or mild and $`const`$ – leads to different abundance ratios between elements with different nucleosynthetic origin, as e.g. \[C/O\] or \[O, Mg, …./Fe\], where C comes from intermediate mass stars and Fe has important SNI contributions (cf. Sect.5), both leading to a delayed production with respect to the SNII products O, Mg, etc. However, no complete grid of stellar evolutionary tracks or yields for varying abundance ratios \[Mg/Fe\] or \[$`\alpha /\mathrm{Fe}`$\] is available. Our chemically consistent ($`=`$ cc) evolutionary synthesis models follow the evolution of ISM abundances together with the spectrophotometric properties of galaxies and account for the increasing initial metallicity of successive generations of stars by using various sets of input physics – stellar evolutionary tracks, model atmosphere spectra, color and absorption index calibrations, yields, lifetimes, and remnant masses – ranging in metallicity from $`\mathrm{Z}=10^4`$ up to $`\mathrm{Z}=0.05`$. The two basic parameters of our evolutionary synthesis model are the IMF, which we take from Scalo, and SFHs, which we appropriately chose for different galaxy types. For ellipticals we use $`\mathrm{\Psi }(\mathrm{t})\mathrm{e}^{\mathrm{t}/\mathrm{t}_{}}`$, for spiral types Sa … Sc $`\mathrm{\Psi }(\mathrm{t})\frac{\mathrm{G}}{\mathrm{M}}(\mathrm{t})`$, and for Sd $`\mathrm{\Psi }(\mathrm{t})=const.`$ with characteristic timescales for SF $`\mathrm{t}_{}`$ (for spirals defined via $`_0^\mathrm{t}_{}\mathrm{\Psi }\mathrm{dt}=0.63\mathrm{G}_{\mathrm{t}=0}`$) ranging from 1 Gyr (E) to 2, 3, 10, and 16 Gyr for Sa, Sb, Sc, and Sd, respectively. Our chemically consistent models simultaneously describe the spectrophotometric evolution in terms of spectra, luminosities, colors (UV – IR), emission and absorption lines as a function of time or – for any cosmological model as given by $`\mathrm{H}_0,\mathrm{\Omega }_0,\mathrm{\Lambda }_0,`$ and a redshift of galaxy formation $`\mathrm{z}_{\mathrm{form}}`$ – the redshift evolution of apparent magnitudes UBVRIJHK and colors, including evolutionary and cosmological corrections, as well as the attenuation by intervening HI, and the chemical evolution in terms of ISM abundances of individual elements <sup>12</sup>C, …, <sup>56</sup>Fe as a function of time or, again, of redshift. The SFHs have been chosen as to provide agreement, together with the IMF, of our model galaxies after a Hubble time with integrated colors, luminosities, absorption features (E/S0s), emission line strengths (spirals), typical for the respective galaxy types, as well as with template spectra (Kennicutt 1992, see Möller et al. 1999a for details), and characteristic HII region abundances. As compared to models using solar metallicity input physics only, our cc models use somewhat different SFHs, even for galaxies that by today come close to $`\mathrm{Z}_{}`$. Clearly, differences between cc and $`\mathrm{Z}_{}`$ models become increasingly important towards higher redshift for all galaxy models. We caution that our models are simple 1-zone descriptions without any dynamics or spatial resolution, meant to describe global average quantities like integrated spectra or colors, and absorption line strengths or HII region abundances around $`1\mathrm{R}_\mathrm{e}`$. ## 3 Chemically Consistent Photometric Evolution Having only available a very incomplete grid of stellar evolutionary tracks and color calibrations, Arimoto & Yoshii 1986 were the first to attempt a cc approach to photometric evolution. Einsel et al. 1995 compiled more complete data sets for 5 metallicities to describe the photometric evolution in terms of colors and absorption line indices on the basis of stellar tracks from the Geneva, color calibrations from the Yale, and index calibrations from the Lick groups, respectively. In Möller et al. 1997 we present cc spectrophotometric evolution models based on most recent input physics. We discuss the comparison between the (mass-weighted) ISM metallicities for various galaxy types and the luminosity-weighted metallicities of the stellar population seen in different wavelength bands. At late stages, stars in models with $`const.`$ SF (Sd) show a stellar metallicity distribution strongly peaked at $`\frac{1}{2}\mathrm{Z}_{}`$ at all wavelengths and close to the ISM metallicity. Elliptical models, on the other hand, show broad stellar metallicity distributions extending from $`\mathrm{Z}=10^4\mathrm{to}\mathrm{Z}=0.05`$ in all bands with the maximum luminosity contribution in U coming from stars with $`<\frac{1}{2}\mathrm{Z}_{}`$, and in K from stars with $`\mathrm{Z}_{}`$, as shown in Fig.1. The largest difference of as much as a factor 2 is obtained in our Sb model between the luminosity-weighted stellar metallicity (e.g. in V) and the ISM metallicity. The time evolution of all these metallicities for different galaxy types are presented in Möller et al. 1997 together with the evolution of colors and absorption line indices. In Kurth et al. 1999 we present single burst single metallicity models using stellar tracks from the Padova group and compare the time evolution of colors and absorption indices to globular cluster observations. We give theoretical color and index calibrations in terms of \[Fe/H\] and investigate how they evolve with age. Stellar abundance ratios, like \[Mg/Fe\], as derived from absorption line ratios, depend on the initial abundance ratio $`[\mathrm{Mg}/\mathrm{Fe}]_{}`$ of the star, on modifications through nucleosynthesis and mixing during the life of the star, and on physical parameters in the stellar atmosphere. The Fe index, e.g., is as sensitive to Fe as to global metallicity Z, which to 50% is made up by oxygen from SNII (Tripicco & Bell 1995). Abundance ratios in galaxies additionally depend on the age and metallicity distributions of the stars and, hence, on the SFH, the IMF, any possible pre-enrichment (from Pop 3, halo, …), on metal-poor infall and/or metal selective outflow, which determine the ISM abundance ratio $`[\mathrm{Mg}/\mathrm{Fe}]_{\mathrm{ISM}}`$ before the birth of the stars. In particular do solar abundance ratios, taken for reference, reflect the local SFH, IMF, and all pre-enrichment & dilution effects of the local ISM. Looking at the metallicity dependent SNII – yields (cf. Sect.5) we notice that, integrated over the IMF, the ejected mass ratio $`\mathrm{M}(\mathrm{Mg})/\mathrm{M}(\mathrm{Fe})`$ increases by large factor when going down in metallicity from $`\mathrm{Z}_{}`$ to $`10^3\mathrm{Z}_{}`$. A lower metallicity limit for SNIa, as discussed by Kobayashi et al. 1998 would further increase $`[\mathrm{Mg}/\mathrm{Fe}]_{\mathrm{ISM}}`$ at early stages. Therefore, all conclusions drawn for galaxies from a comparison of observed $`[\mathrm{Mg}/\mathrm{Fe}]_{}>0`$ with $`\mathrm{Z}_{}`$ models – concerning a top-heavy IMF, a shorter SF timescale in massive Es, a higher SFE, etc. – should be taken with extreme caution (cf. Fritze - v. Alvensleben 1998). ## 4 Chemically Consistent Spectro-Cosmological Evolution Using sets of model atmosphere spectra covering all spectral types, luminosity classes, and our 5 metallicities (Lejeune et al. 1997, 1998) we describe the cc spectral evolution of our model galaxies. The agreement of our model spectra after a Hubble time with nearby templates is shown by Möller et al. 1997 and 1999a. With any kind of cosmological model we calculate evolutionary and cosmological corrections, apparent magnitudes U, …, K, and colors as a function of redshift z. Attenuation by intervening hydrogen (Madau 1995) is included, dust extinction depending on metallicity and gas content is currently being added in collaboration with D. Calzetti (cf. Möller et al. 1999b). For the Sd galaxy model, Fig.2a shows the cc $`(\mathrm{e}+\mathrm{k})`$ corrections in B and K as compared to those for the $`\mathrm{Z}_{}`$ model. In B, cc models give $`(\mathrm{e}+\mathrm{k})`$ corrections that make Sd galaxies brighter by $`\stackrel{>}{}1`$ mag at $`\mathrm{z}0.7`$ and by $`\stackrel{>}{}2`$ mag for $`\mathrm{z}\stackrel{>}{}2`$ as compared to $`\mathrm{Z}_{}`$ models. In the K-band as well, differences are quite significant with Sd galaxies being brighter in cc models by $`\stackrel{>}{}1`$ mag at $`\mathrm{z}\stackrel{>}{}1`$. Moreover, the successive building-up of the stellar metallicity distributions tends to increase the evolutionary effects as compared to $`\mathrm{Z}_{}`$ models. In addition to the cosmological dimming ($`\mathrm{k}_{\mathrm{UBVRI}}>0`$) the Sd model is brightened by the evolutionary correction by $`\stackrel{>}{}2`$ mag at $`\mathrm{z}\stackrel{>}{}0.7`$ and by $`3.5`$ mag at $`\mathrm{z}3`$ in B (cf. Fig.2b), while in K, Sd models still have $`(\mathrm{e}+\mathrm{k})\stackrel{<}{}1.5`$ mag at all $`\mathrm{z}0.73`$. The cc spectro-cosmological models and the comparison with high redshift galaxy data will be presented in detail in Möller et al. (in prep.). ## 5 Chemically Consistent Chemo-Cosmological Evolution Timmes et al. 1995 and Portinari et al. 1998 were the first to use stellar yields for a range of metallicities in models for the chemical evolution of the Milky Way and the solar neighbourhood, respectively. For our “spectroscopically successful” models for various galaxy types we now use element yields for different metallicities for successive generations of stars to describe the time evolution of a series of individual elements from <sup>12</sup>C through <sup>56</sup>Fe. For any cosmological model the time evolution directly transforms into a redshift evolution. SNII yields for massive stars ($`>8\mathrm{M}_{}`$) are from Woosley & Weaver 1995, yields for intermediate mass stars from van den Hoek & Groenewegen 1997. SNIa contributions to Fe, C, …, are included for the carbon deflagration white dwarf binary scenario as outlined by Matteucci & Greggio 1986. SNIa yields are only available for $`\mathrm{Z}_{}`$ (Nomoto et al. 1997, model W7), however, no important metallicity dependence is expected for SNIa yields except for a possible lower metallicity limit to the explosion (Kobayashi et al. 1998). Mass loss from stellar winds as e.g. given by Portinari et al. 1998 is not included yet. We recall that metallicity dependent stellar yields are only available for solar abundance ratios and we have no idea, if and in how far non-solar abundance rations would influence the stellar shell structure and, hence, the yields. Moreover, stellar yields depend on $`\frac{\mathrm{\Delta }\mathrm{Y}}{\mathrm{\Delta }\mathrm{Z}}`$, explosion energies, remnant masses, etc. and the metallicity dependence of these factors is still poorly understood. No clear trends are seen, neither in the output of element $`\mathrm{X}_\mathrm{i}`$ from stars of given mass as a function of metallicity, nor, at const. metallicity, as a function of stellar mass. We stress that for a given SF history and IMF our models yield absolute abundances that do not require any scaling or normalisation. We compared the redshift evolution of C and Mg abundances with CIV- and MgII-QSO absorber statistics (Fritze - v. Alvensleben et al. 1989, 1991 using yields for $`\mathrm{Z}_{}`$ stars only). CIV- and MgII-absorption is caused by the moderate column density gas in extended galaxy halos ($`17.5\mathrm{log}\mathrm{N}(\mathrm{HI})[\mathrm{cm}^2]20`$). We argued that the cross section for CIV- and MgII-absorption should scale with the abundance of the respective elements and that, hence, the redshift evolution of the comoving number density of absorbers should trace the abundance evolution. The first direct abundance determinations (Stengler - Larrea 1995) a posteriori justified our assumption. We found good agreement of the observations with a standard model for halo SF ($`\mathrm{t}_{}1\mathrm{Gyr}`$), derived constraints for $`\mathrm{t}_{}`$, the IMF, and the cosmological parameters, and predicted a low number of CIV systems at low redshift, which was impressively confirmed by HST key project data (Bahcall et al. 1993). Our cc chemo-cosmological models are compared with observed abundances in damped Ly$`\alpha `$ absorbers (=DLAs) in Lindner et al. 1999. DLAs show radiation damped Ly$`\alpha `$ lines due to high column density gas ($`\mathrm{log}\mathrm{N}(\mathrm{HI})[\mathrm{cm}^2]20.3`$) and a large number of associated low ionisation lines of C, N, O, Al, Si, S, Cr, Mn, Fe, Ni, Zn, … High resolution observations (KECK and WHT) that fully resolve the complex velocity structure in the lines allow to derive precise element abundances in a large number of DLAs over the redshift range 0 … $`4`$ (Boissé et al. 1998, Lu et al. 1993, 1996, Pettini et al. 1994, 1999, Prochaska & Wolfe 1997, …). Based on similarities of their HI column densities with those of local spiral disks, of their comoving gas densities at high z with (gas $`+`$ star)-densities in local galaxies, and based on line asymmetries indicative of rotation, damped Ly$`\alpha `$ absorption is thought to arise in (proto-)galactic disks along the line of sight to a distant QSO (e.g. Wolfe 1995). Alternatively, Matteucci et al. 1997 propose starbursting dwarf galaxies on the basis of \[N/O\] ratios, while Jimenez et al. 1999 propose LSB galaxies, and Haehnelt et al. 1998 subgalactic fragments to explain DLA galaxies at low and high redshift, respectively. After referring all observed DLA abundances to one homogeneous set of oscillator strengths and solar reference values, we compare with our spiral galaxy models Sa, …, Sdearlier shown to agree with characteristic HII region abundances at $`\mathrm{z}=0`$, and with spectrophotometric properties as observed to $`\mathrm{z}\stackrel{>}{}1`$. As can be seen in Fig.3 on the example of Zn, our Sa and Sd models bracket the redshift evolution of DLA abundances from $`\mathrm{z}4.4`$ to $`\mathrm{z}0.4`$. Similar agreement is found for all 8 elements with a reasonable number of DLA data available. Since Phillipps & Edmunds 1996 and Edmunds & Phillipps 1997 have shown that the probability for an arbitrary QSO sightline to cut through an intervening gas disk and produce DLA absorption is highest around $`1\mathrm{R}_\mathrm{e}`$, our models bridge the gap from high-z DLAs to nearby spiral HII region abundances. We conclude that from the point of view of abundance evolution, DLA galaxies may well be the progenitors of normal spirals Sa – Sd, allthough we cannot exclude that some starbursting dwarf or LSB galaxies may also be among the DLA galaxy sample. The influence of the metallicity dependent yields is seen from the comparison with $`\mathrm{Z}_{}`$ models in Fig.3 and varies from element to element. Whenever a significant difference is seen, the cc models give a good representation of the data while the $`\mathrm{Z}_{}`$ models lose data points above the Sa or below the Sd curves. We also show models with SNII yields calculated under the assumption of higher explosion energies (model C of Woosley & Weaver 95), which does not make much difference. Curves for Sb and Sc models run between those for Sa and Sd and are omitted for clarity. Comparison of cc chemo-cosmological models with observed DLA abundances further shows that the weak redshift evolution of DLA abundances is a natural result of the long SF timescales for disks galaxies, and the range of SF timescales $`\mathrm{t}_{}`$ for spirals from Sa through Sd fully explains the abundance scatter among DLAs at any redshift. Somewhat surprisingly, abundances of elements which locally are known to strongly deplete onto dust grains (like Fe or Cr) are as well described by our models as are non-refractory elements like Zn. We prefer, however, not to draw conclusions about the importance of dust in DLAs in view of the uncertainties in the stellar yields and the simplicity of our closed-box 1-zone model. Fig.3 also shows that while at high redshift all spiral types seem to give rise to DLA absorption, no more data points at $`\mathrm{z}\stackrel{<}{}1.5`$ reach close to our early type spiral models. At low redshift, the gas poor early type spirals seem to drop out of DLA samples. While a deficiency of high N(HI) systems at low z has been noted before (Lanzetta et al. 1997), and attributed to their high metallicity and dust content (Steidel et al. 1997), our models indicate an additional reason: as the global gas content drops, the probability for a QSO sightline to cut through a high N(HI) part of the galaxy decreases, i.e. the cross section for damped Ly$`\alpha `$ absorption gets reduced. If this were confirmed by further low-z DLA data, it would have serious implications as to the posibility to optically identify DLA galaxies. Locally, on average, Sd galaxies are fainter by $`2`$ mag in B than Sa’s and our cc spectro-cosmological models predict that the low-z DLA galaxies should be about as faint in B, $``$, and K as the brightest members of the high-z population: $`\mathrm{B}25`$, $``$$`24.5`$, K $`22`$ mag. Luminosities of the few optically identified DLA galaxies (and candidates) to date are in good agreement with our predictions (cf. Fritze - v. Alvensleben et al. 1999a, b, c). Optically identified DLA absorbers with information about ISM abundances from the metal absorption lines and spectrophotometric properties of the stellar population allow to much better constrain the model parameters than either aspect (ISM or stars) alone (cf. Lindner et al. 1996). Since DLA galaxies are within the reach of 10m-class telescopes up to redshifts z $`>3`$ and trace the normal galaxy population to these high redshifts without any bias as to high luminosity, radio power, or the like, they can give powerful constraints on the evolutionary histories and ages, and, ultimately, even on the cosmological parameters. Accurate abundance data in very low metallicity DLAs may provide valuable clues for the nucleosynthesis at low metallicity. ## 6 Conclusions and Outlook I pointed out the importance of a chemically consistent modelling of the spectrophotometric and chemical evolution of galaxies that takes into account the evolving metallicity distribution of the stellar population. The comparison with models using solar metallicity input physics only showed substantial differences due to important contributions of subsolar metallicity stars both to the yields of various elements and to the spectrophotometric properties for all galaxy types, in particular when going to high redshift. I stressed the need for reliable low metallicity input physics (stellar tracks, mass loss, lifetimes, yields, remnant masses, spectra, and absorption index calibrations) and cautioned the use of solar metallicity models to derive conclusions from non-solar absorption line ratios in the clearly composite stellar populations of E/S0 galaxies. Observations of the stellar metallicity distributions in nearby galaxies will be necessary before we can reasonably relax the crude simplification of closed box models. With the same number of parameters (IMF and SF history) as standard chemical or spectrophotometric models, our unified chemical, spectrophotometric and cosmological model provides a powerful tool to constrain the parameters and ages of galaxies for which both kind of information is available. ###### Acknowledgements. I gratefully acknowledge partial financial support from the Organisers and from the Deutsche Forschungsgemeinschaft (Fr 916/7-1).
no-problem/9907/hep-ph9907278.html	ar5iv	text	# 1 Kinetic theory of particle creation ## 1 Kinetic theory of particle creation After years of effort and anticipation, the Relativistic Heavy Ion Collider (RHIC) at Brookhaven will soon begin producing collisions of gold nuclei in colliding beams, at an energy of 100 GeV per nucleon in each beam. Theoretical considerations, as well as data from the successful heavy-ion program at CERN’s SPS, lead us to expect that in a central Au–Au collision the incoming baryons will pass through each other with minimal deflection. The net outgoing baryon number will thus be concentrated within a couple of units of rapidity of the beam rapidities, leaving a large central region to be filled by a baryon-free plasma. Energy deposition in the central region will take place mainly via soft mechanisms, although minijet production will play a significant role. A QCD-inspired model for soft production in the central region begins with the exchange of soft gluons between the two nuclei as they pass through each other. This leaves them charged with color, sources of color electric flux. The flux fills a tube connecting the receding nuclei, which thus form a “color capacitor” if the field is coherent across the tube, or a “color rope” if it is not. The electric field strength in either case is proportional to the square root of the number of gluon exchanges per unit area, and thus $`EA^{1/3}`$, as a result of a random walk in color space. An electric field will pop pairs of charged particles (in this case, colored particles) out of the vacuum, especially if they are light on the scale of the available field energy density. Casher, Neuberger, and Nussinov derived via WKB the rate for particle creation per unit volume as a function of transverse momentum, $$\frac{dN}{dtdVd^2p_{}}=eE\mathrm{log}\left[1+\mathrm{exp}\left(\frac{\pi (m^2+p_{}^2)}{eE}\right)\right].$$ (1) (Integrating (1) over $`p_{}`$ gives the famed Schwinger formula .) If the flux tube is narrow to begin with, as in the case of $`e^+e^{}`$ annihilation into quarks or of $`pp`$ scattering, the quarks produced via (1) will break the tube and begin an inside–outside cascade. For nucleus–nucleus collisions, however, the geometry more closely resembles that of the idealized problem of a homogeneous electric field created by infinite, parallel capacitor plates. Then the created particles form a uniform current that screens the field gradually according to (Abelian) Maxwell’s equations, $$\frac{dE}{dt}=j.$$ (2) This is called back-reaction. A straightforward approach to the particle dynamics is given by kinetic theory, where the pair creation rate (1) appears as a source term for the Vlasov equation , $$\frac{f}{t}+eE\frac{f}{p_z}=eE\mathrm{log}\left[1+\mathrm{exp}\left(\frac{\pi (m^2+p_{}^2)}{eE}\right)\right]\delta (p_z).$$ (3) (The particle density $`f`$ feeds back into the Maxwell equation (2) through the current.) This gives a beautiful description of matter creation and subsequent plasma oscillations. The drawback of this approach, of course, is that the use of the classical Vlasov formalism lacks fundamental justification. ## 2 Enter quantum field theory Quantum field theory allows a first-principles approach to the problem . One begins by taking the number of flavors $`N_f`$ to infinity in order to justify a classical, mean-field approximation for $`E`$. If we begin with a scalar field in one dimension, the ingredients of the theory are then the Klein–Gordon equation for the matter field, $$[(^\mu +ieA^\mu )(_\mu +ieA_\mu )+m^2]\mathrm{\Phi }(x)=0,$$ (4) and the semiclassical Maxwell equation, $$_\mu F^{\mu \nu }=0\|j^\nu \|0.$$ (5) A Fourier expansion of $`\mathrm{\Phi }`$, $$\mathrm{\Phi }(x,t)=\frac{dk}{2\pi }\left[f_k(t)a_k+f_k^{}(t)b_k^{}\right]e^{ikx},$$ (6) inserted into (4), gives evolution equations for the Fourier amplitudes $`f_k`$, $$\frac{d^2f_k(t)}{dt^2}+\omega _k^2(t)f_k(t)=0,$$ (7) where $$\omega _k(t)^2[keA(t)]^2+m^2.$$ (8) The canonical commutation relations for $`\mathrm{\Phi }`$ imply that $`f_k`$ can be written as $$f_k(t)=\frac{1}{\sqrt{2\mathrm{\Omega }_k(t)}}e^{i^t\mathrm{\Omega }_k(t^{})𝑑t^{}}.$$ (9) Inserting this into (7) gives equations for the effective frequencies $`\mathrm{\Omega }_k(t)`$, $$\mathrm{\Omega }_k^2(t)=\frac{\ddot{\mathrm{\Omega }}_k}{2\mathrm{\Omega }_k}+\frac{3}{4}\left(\frac{\dot{\mathrm{\Omega }}_k}{\mathrm{\Omega }_k}\right)^2+\omega _k^2(t).$$ (10) We can write the current $`j`$ in terms of the amplitudes $`f_k`$, and (5) becomes the subtracted, renormalized equation $$\ddot{A}=e\frac{dk}{2\pi }(keA)\left[\frac{1}{\mathrm{\Omega }_k(t)}\frac{1}{\omega _k(t)}\right].$$ (11) Equations (10) and (11) can be integrated numerically. We applied this formalism to the creation of bosons and fermions in one and three dimensions. For illustration, I show in fig. 1 the time evolution (in scaled variables) of a system of bosons in three dimensions. Particles created at early times are accelerated into plasma oscillations, with continued particle creation whenever the field is non-zero. A comparison to the results of kinetic theory \[a slight modification of (3): dashed curves\] shows that the latter approximates the field-theoretic solution well. Space limitations prevent me from showing the beautiful results of comparing the phase space distributions of produced particles in the two approaches. ## 3 Extensions The idealized problem of infinite, static capacitor plates was of course only a first step in development of the flux tube model. With an eye towards nucleus–nucleus collisions, we extended the analysis to the problem of capacitor plates separating at the speed of light, $`z=\pm ct`$. As is natural in the context of the central rapidity region , we assumed that the initial conditions are invariant under longitudinal boosts so that the time development depnds only on the invariant, comoving time $`\tau =\sqrt{t^2z^2}`$. The phase-space distribution of the produced particles now gives the rapidity distribution $`dN/dy`$, and analysis of the energy-momentum tensor gives the energy density and equivalent temperature as functions of $`\tau `$ (though there is as yet no mechanism for equilibration). As it turns out, the results are again well approximated by kinetic theory. The geometry to this point is still that of an idealized system of infinite transverse extent. Upon solving the problem in a cylinder of finite radius (a computationally formidable problem due to nonlinear mode mixing), Judah found that the agreement between field theory and kinetic theory is at best qualitative. As mentioned, the semiclassical approximation presented above finds its justification in the limit of large $`N_f`$. The next order in $`1/N_f`$ introduces particle–particle collisions and thus a mechanism for relaxation to thermal equilibrium . Unfortunately, the resulting integro-differential equations are (so far) computationally unmanageable. Kluger, Mottola, and Eisenberg later eschewed this approach and showed, in what was to be Judah’s last paper, that equilibration can arise even in the lowest-order mean field theory shown above if one allows for dephasing of the highest-frequency quantum oscillations. The connection between reversible quantum mechanics and irreversible kinetic theory has long been an area of mystery and of hard work. In this last paper, a Vlasov equation non-local in time is actually derived for the pair-creation problem.<sup>3</sup><sup>3</sup>3The paper is a mathematical tour de force of Airy functions, uniform asymptotic expansions, etc. Anyone who knew Judah’s great love for mathematical physics of the Morse & Feshbach variety will recognize Judah walking in its pages. ## 4 The flux tube of the dual superconductor Judah, as noted, applied the field-theoretic formalism to particle production in a cylindrical flux tube of finite, fixed radius. The QCD flux tube, however, is a dynamical object, governed by field equations. ’t Hooft and Mandelstam noted long ago that flux tubes arise naturally in superconductors via the Meissner effect; just as magnetic monopoles would be confined by a magnetic flux tube in a superconductor, quarks with their electric color charge would be confined by an electric flux tube if the QCD vacuum has the structure of a dual superconductor. Melissa Lampert and I have taken the first step of studying the dynamics of classical charges moving in an electric flux tube and the reaction of the flux tube in this model . To specify the dual superconductor model, we begin with Maxwell’s equations coupled to both magnetic and electric currents, $`_\mu F^{\mu \nu }`$ $`=`$ $`j_e^\nu `$ (12) $`_\mu \stackrel{~}{F}^{\mu \nu }`$ $`=`$ $`j_g^\nu .`$ (13) Eq. (13) is no longer just a Bianchi identity; thus a vector potential can be introduced only if a new term is added to take care of the magnetic current, $$F^{\mu \nu }=^\mu A^\nu ^\nu A^\mu +ϵ^{\mu \nu \lambda \sigma }G_{\lambda \sigma },$$ (14) $$G^{\mu \nu }=n^\mu (n)^1j_g^\nu .$$ (15) This vector potential can be coupled to electric charges as usual; in order to introduce magnetic charges, one introduces a dual potential via $$\stackrel{~}{F}^{\mu \nu }=^\mu B^\nu ^\nu B^\mu +ϵ^{\mu \nu \lambda \sigma }M_{\lambda \sigma },$$ (16) $$M^{\mu \nu }=n^\mu (n)^1j_e^\nu .$$ (17) Now we can write a model for the monopoles, for which the simplest is an Abelian Higgs theory , $$D_\mu ^BD^{\mu B}\psi +\lambda (\|\psi \|^2v^2)\psi =0,$$ (18) where $$D_\mu ^B_\mu igB_\mu .$$ (19) This theory should produce the desired magnetic condensate to confine electric charge. The magnetic current appearing in (13) is $$j_g^\mu =2g\mathrm{Im}\psi ^{}D^{\mu B}\psi .$$ (20) For the electric charges, we take simple two-fluid MHD (see for details). The parameters of the theory may be adjusted to put the superconductor into the Type I or the Type II regime; the density of the charged fluid will make the plasma frequency $`\omega _p`$ larger or smaller than the mass $`m_V`$ of the gauge boson in this Higgs theory. I show in fig. 2 plasma oscillations in the flux tube for a Type I superconductor with $`\omega _p<m_V`$. The oscillations are made nonlinear by the reaction of the flux tube, which tries to close up when the field is weak. This is a first step towards the inclusion of the dynamics of the flux tube in a model of particle creation. The latter might be included, as discussed above, as a kinetic theory based on a Vlasov equation or via a full field-theoretic treatment. As argued in , however, the simplest Higgs theory (18) doesn’t contain enough dynamical confinement physics to produce good phenomenology.
no-problem/9907/cond-mat9907212.html	ar5iv	text	# Zero-Conductance Resonances due to Flux States in Nanographite Ribbon Junctions ## Abstract The electronic transport properties through junctions in nanographite ribbons are investigated using the Landauer approach. In the low-energy regime ribbons with zigzag boundary have a single conducting channel of edge states. The conductance as a function of the chemical potential shows a rich structure with sharp dips of zero conductance. Each zero-conductance resonance is connected with a resonant state which can be interpreted as the superposition of two degenerate flux states with Kekulé-like current patterns. These zero-conductance dips are connected with a pronounced negative magneto resistance. 72.10.-d, 72.80.Rj, 73.23.-b, 73.20.-r, 73.23.Ad, 73.50.-h, 73.40.Rw The discovery of fullerene molecules and carbon nanotubes has triggered intensive research on various nanometer size carbon materials. In these systems, the geometry of sp<sup>2</sup> carbon networks crucially affects the electronic states near the Fermi level. Studies with scanning tunneling microscopy and spectroscopy have confirmed the connection between the electronic states of the single wall carbon nanotubes (SWCN) and their geometry. Recently the electrical transport measurement of individual SWCN became possible and the quantized conductance of multi-wall carbon nanotubes was observed . This initiated theoretical studies devoted to effects of non-magnetic impurities, electron correlation and topological defects. Besides these closed carbon molecules, there are also systems with open boundaries which display unusual features connected with their shape. They include small scale systems based on graphite, so-called nanographites. There are two basic shapes of regular graphite edges, called armchair and zigzag edges, depending on the cutting direction of the graphite sheet. Properties originating from such edges have been studied recently using the model of graphite ribbons, one-dimensional graphite stripes of infinite length and finite width. Ribbons with zigzag edges (zigzag ribbons) possess electron states localized near the edge with energies very close to the Fermi level . Such states are absent for ribbons with armchair edges. The edge states of zigzag ribbons were analyzed in terms of nearest-neighbor tight binding models and density functional approach. It was also pointed out that the edge states play important roles in the magnetic properties in nanometer-size systems due to their relatively large contribution to the density of states at the Fermi energy . Recently, experimental evidence for edge states has been reported for nanographite systems derived from graphitized diamond nanoparticles . In this letter, we investigate the transport properties of ribbons related to these edge states. For this purpose we design three different types of junctions which connect zigzag ribbons of the same or different widths. The electronic states are described by a single-orbital nearest-neighbor tight binding model. The conductance of the junctions is evaluated using the multi-channel Landauer formula, $`G(E)={\displaystyle \frac{\mathrm{e}^2}{\pi \mathrm{}}}{\displaystyle \underset{\mu ,\nu }{}}\|t_{\mu \nu }(E)\|^2,`$ (1) where $`t_{\mu \nu }(E)`$ is a transmission coefficient from $`\mu `$-th channel to $`\nu `$-th channel at energy $`E`$, calculated by a recursive Green’s function method . Before discussing the design of the junctions and their conductance properties, we review a few facts concerning the low-energy states in two kinds of graphite ribbons, the zigzag and the “bearded” ribbon (a variant of a zigzag ribbon). Zigzag ribbon: The zigzag ribbons are metallic for arbitrary ribbon width. The most remarkable feature is the presence of a partly flat band at the Fermi level, where the electrons are strongly localized near the zigzag edge. Each edge state has a non-vanishing amplitude only on one of the two sublattices, i.e. non-bonding character. However, in a zigzag ribbon of finite width, two edge states coming from both sides, have finite overlap. Because they are located on different sublattices, they mix into a bonding and anti-bonding configuration. In this way the partly flat bands acquire a dispersion \[Fig.1(b)\]. Note that the overlap is increasing as $`k`$ deviates from $`\pi /a`$, because the penetration depth of the edge states increases and diverges at $`k=2\pi /3a`$, where $`a`$ is the lattice constant. The dispersion depends on the ribbon width $`N`$ (number of zigzag lines from one side to the other). Close to $`k=\pi /a`$, the spectrum has the approximate form $`E_k=\pm 2tND_k^{N1}(1D_k/2)`$, where $`D_k=2\mathrm{cos}\left(\frac{ka}{2}\right)`$, and $`t`$ is the hopping matrix element. Thus, although the edge states on each side separately have non-bonding character, together through their overlap they provide one conducting channel except at exactly $`E=0`$. The energy region of single-channel transport is restricted by the energy gap ($`\mathrm{\Delta }_z`$) to the next channel $`\mathrm{\Delta }_z4t\mathrm{cos}[(N1)\pi /(2N+1)]`$ . Bearded ribbon: The bearded ribbon has one zigzag edge and one edge which has additional bonds (beard) attached to the zigzag edge \[Fig.1(a) region M\]. The edge states of both sides reside on the same sublattice so that in spite of their overlap their non-bonding character is retained and a completely flat band at $`E=0`$ is resulting for any width $`N`$ \[Fig.1(b)\]. The absence of dispersion leads to the insulating behavior for the edge state channel. The gap ($`\mathrm{\Delta }_b`$) to the first conducting channel is given by $`\mathrm{\Delta }_b4t\mathrm{cos}[N\pi /(2N+2)]`$ . We now turn to the design of a junction connecting two zigzag ribbons denoted by L (left) and R (right). As a first example we consider the case depicted in Fig.1(a) where the junction region denoted by M is a bearded ribbon of length $`l`$ (number of attached bonds). This junction model represents a metal-insulator-metal junction providing an illustrative example for the peculiar transport properties of graphite ribbons regardless of the question as to whether bearded ribbons could be realized in nature. We calculate the conductance $`G(E)`$ numerically within the energy range $`\|E\|<\mathrm{\Delta }_b/2`$. The result for $`N=30`$ and different values of $`l`$ is shown in Fig.1(c) using a logarithmic scale on the energy axis. The most striking feature is the large number of zero-conductance dips at energy values which depend on $`l`$. These dips represent resonances of total reflection. In the whole energy range no transmission resonances are observed for any value of $`l`$. Note also that the change of $`N`$ does not lead to a qualitative change and only modifies the energy range of single-channel conductance. Resonances are associated with discrete quantum levels in the junction regions. In the following we would like to characterize the quantum states encountered here and discuss the origin of total reflection. It was noticed earlier that some graphite ribbons form a triangular Kekulé pattern of circular current driven by special external boundary conditions . To investigate whether the junction states responsible for the resonance display a similar structure, we introduce the current vortex amplitude defined on the dual (triangular) lattice as the clockwise circular sum of the currents $`I_{i,p}`$ on the bonds of the $`p`$-th hexagonal plaquette, $`V_p=_{i=1}^6I_{i,p}`$. An incident current $`J_{\mathrm{in}}`$ coming from lead $`L`$ yields a current flow in the junction region, whose circulating component (vorticity), we probe by the following two response functions $`\chi _V^{(1)}=\|V_p\|/J_{\mathrm{in}}`$ and $`\chi _V^{(2)}=V_p/J_{\mathrm{in}}`$, where the bracket $`A`$ denotes the average value of $`A`$ in the junction region. The first function is a measure for the presence of circular currents and the second for the overall vorticity of the system. In Fig.1(d) the energy dependence of both functions is shown for $`l=15`$ and $`N=30`$. The susceptibilities $`\chi _V^{(1,2)}`$ show sharp divergences in the sense of linear response to the driving current $`J_{\mathrm{in}}`$ when $`E`$ approaches the energy values of a zero-conductance dip. (Note that $`\chi _V^{(1,2)}`$ are not defined at zero-conductance energies, because no driving current $`J_{\mathrm{in}}`$ can flow.) This suggests that the states at these energy values form current vortex patterns. The overall vorticity measured by $`\chi _V^{(2)}J_{\mathrm{in}}`$ changes sign at each resonance. We find that the circular current form a Kekulé-like pattern. This is most easily seen in the Fourier transform $`F(𝒌)=_p\left(\frac{V_p}{J_{\mathrm{in}}}\right)\mathrm{e}^{i𝒌𝒓_p}`$ in the junction region which has a clear peak structure of the triangle correlation of the circular currents pattern (Fig.1(e)) ($`𝒓_p`$: coordinate of the ring center; $`k_x(k_y)`$: wave number along (perpendicular to) zigzag lines in the junction). This type of pattern appears in $`F(𝒌)`$ very close to every zero-conductance dip and gradually disappears when deviating from the resonance energies. Because $`F(𝒌)`$ is the correlation among the circular currents on the hexagonal plaquettes, the state driven by the incident current is in this sense a flux state. The quantum state associated with the resonance is the superposition of the two flux states with opposite circular currents, which combine to a standing wave within the junction region under the condition of time reversal symmetry. The presence of this quantum level plays an important role for the realization of the zero-conductance resonances. At each resonance the transmission is not only carried by the usual tunneling through the insulating junction region, but also through this resonant state. This decomposition into two channels yields the following form for the transmission amplitude between the two leads close to a resonance at energy $`E_0`$, $$t(E)=\stackrel{~}{t}(E)\left(1\frac{i\mathrm{\Gamma }/2}{EE_0+i\mathrm{\Gamma }/2}\right)=\frac{\stackrel{~}{t}(E)(EE_0)}{EE_0+i\mathrm{\Gamma }/2},$$ (2) where $`\stackrel{~}{t}(E)`$ is a regular complex function of $`E`$ and $`\mathrm{\Gamma }`$ the width of the resonance. The destructive interference of the two channels is a consequence of the symmetric form of the $`S`$-matrix which relates the in- and outgoing waves of the two leads with and of the junction region . This situation has been discussed in the context of the three-way splitter and a rigorous proof was given for the exact cancelling at the resonant energies . A consequence of this resonant form of the transmission amplitude (Eq.(2)) is that the phase $`\theta `$ of $`t(E)`$ exhibits a jump by $`\pi `$ at each resonance (see Fig.1f) (similar to Ref. ). The behavior of the phase is well-described by Eq.(2). The form of the $`S`$-matrix mentioned above as well as the degeneracy of the flux states of opposite chirality, combining to a standing wave state, are based on time reversal symmetry. This symmetry can be destroyed by applying a magnetic field. We observe a pronounced negative magnetoresistance at the zero-conductance dip where the conductance grows proportionally to $`B^2`$ for small magnetic fields $`B`$, as we will discuss in detail elsewhere. We would like to mention here that circular currents associated with zero-conductance dips were recently also reported for two-dimensional wave guides including junctions with stub geometry . Also in this case a negative magnetoresistance should occur. Next, we discuss two junctions with a more realistic design which do not contain bearded ribbons. They are shown in Fig.2(a) and connect zigzag ribbon leads of different width (we show $`N_L=50`$ and $`N_R=30`$ as a representative case). The M-region contains a tilted zigzag edge for junction A and an armchair edge for junction B. The conductance of the two junctions is qualitatively different. The tilted edge in A supports an edge state, similar to the bearded ribbon, on the same sublattice as the edge state on the other side. Indeed we find a very similar behavior of the conductance with large number of zero-conductance resonances \[Fig.2(b)\] which are associated with flux states too. For junction B there is no localized state near the armchair edge. The conductance is rather featureless without any resonance in the energy regime of single-channel transport. There is a zero-conductance dip above the single-channel regime of the lead on the left-hand side ($`E>\mathrm{\Delta }_z(N=50)/2=0.062t`$) for both junctions. The origin is also connected with analogous junction states which are based on flux states as the analysis of the susceptibilities $`\chi _V^{(1,2)}`$ show, although the conditions are different here, since the transmission occurs from three channels on the left hand side to a single channel on the right hand side. We omit here a detailed discussion of this more complicated case. For energies above the single channel threshold of the right-hand side lead ($`E>\mathrm{\Delta }_z(N=30)/2=0.101t`$) no zero-dip features appear in any case. From this analysis we conclude that the edge structure of the junction plays an important role in forming the resonant states in the M-region. It is well known that the electron states display chiral properties in graphite sheets, if there is an imbalance between the two sublattices e.g., by different onsite potential. The edge states on zigzag edges are a consequence of this imbalance, since the outer most sites belong to a single sublattice. The abrupt change of the sublattice on the edge as it occurs for previous junction and junction A yields the boundary condition to form the degenerate flux states in the M-region. These flux states combine to one resonant state of standing-wave nature. Further numerical analysis shows that even single non-magnetic impurities and other simple defect structures disturbing the sublattice balance can cause a zero-conductance dip associated with flux states in zigzag ribbons. In conclusion, we numerically analyzed three types of nanographite junctions. We found that the conductance of various junctions having zigzag edges shows many zero-conductance dips as a function of energy (chemical potential). These dips are identified as resonances connected with resonant states based on flux states which form circular-current Kekulé patterns. It is obvious that the topology of the edges is crucial for this phenomenon and the chirality connected with the sublattice structure plays an important role. The time reversal symmetry necessary for the existence of these resonances is violated by external magnetic fields, leading to negative magneto resistance. While the structures used in the calculation might be difficult to produce at present , our results also suggest that transport properties of defective carbon nanotubes, carpet-roll or papier-mâché structures could be rather different from the transport properties of usual closed multi-wall nanotubes or SWCN which have only weak features in the low-energy regime . The present study not only clarifies the importance of the edges and their shapes on transport properties, but also indicates the importance of theoretical studies to explicate the interplay between the transport properties and the network topology of carbon atoms. The present numerical work provides the foresight concerning the analysis based on a low-energy effective theory, which is beyond the scope of this paper and will be presented elsewhere. Such an effective theory will serve as a basis for designing carbon-based electronic devices and for further theoretical work on effects of impurities or electron correlation. The authors are grateful to T. Enoki, C. Oshima, A. Furusaki, H. Yoshioka, K. Kusakabe, K. Nakada, S. Okada, M. Igami and Y. Takagi for many helpful discussions. This work was supported partly by Grants-in-Aid for Scientific Research No. 10309003 and 10640341 from the Ministry of Education, Science and Culture, Japan. K. W. acknowledges the financial support of the Japan Society for the Promotion of Science for Young Scientist. Numerical calculations were performed in part on VPP500 in Institute for Solid State Physics, University of Tokyo and on SX5 in Institute for Molecular Science.
no-problem/9907/cond-mat9907018.html	ar5iv	text	# Enhanced fluctuations of the tunneling density of states near bottoms of Landau bands measured by a local spectrometer. ## Abstract We have found that the local density of states fluctuations (LDOSF) in a disordered metal, detected using an impurity in the barrier as a spectrometer, undergo enhanced (with respect to SdH and dHvA effects) oscillations in strong magnetic fields, $`\omega _c\tau 1`$. We attribute this to the dominant role of the states near bottoms of Landau bands which give the major contribution to the LDOSF and are most strongly affected by disorder. We also demonstrate that in intermediate fields the LDOSF increase with $`B`$ in accordance with the results obtained in the diffusion approximation. Resonant tunneling through individual impurities has been identified and studied in vertical and lateral mesoscopic structures. When an impurity level in a potential barrier passes through the Fermi level in the emitter, it manifests itself as a step in the current-voltage (IV) characteristic , with the magnitude determined by the impurity coupling to the reservoirs and the onset smeared due to the coupling or the thermal distribution of carriers in the contact. Upon increasing bias, the current onset is followed by a plateau where temperature-independent and magnetic-field-sensitive reproducible features have been observed in several experiments on small-area vertical structures . The latter were attributed to the fluctuations in the local density of single-particle states in a disordered emitter. It has then been suggested that the impurity carrying the current (spectrometer) can act as a probe of the local density of states fluctuations (LDOSF) in the bulk of metallic contacts. When shifted with a bias, the spectrometer detects a ’fingerprint’ of the LDOSF as a function of energy. In the present paper, we study the evolution with magnetic field of the LDOSF in a 3D disordered metal, a heavily doped semiconductor, and discuss the results from the point of view of the fluctuation and correlation properties of single-particle wave functions in disordered media. We have measured the fingerprint of the LDOSF, $`\delta \nu (\epsilon )`$, in the differential conductance $`G(V)=\frac{dI}{dV}(V)`$ in a broad range of magnetic fields, $`B`$, and analyzed its variance, $`\delta G^2`$, and correlation parameters. In intermediate fields, $`\omega _c\tau 1`$, we have detected an increase of the fluctuation magnitude, in agreement with the theoretically predicted behavior : $`(\delta G)^2_B/(\delta G)^2_{B=0}1+(\omega _c\tau )^2`$. At higher fields, $`\omega _c\tau 1`$, we have observed large $`1/B`$periodic oscillations in $`\delta G^2`$. We conclude that the LDOSF in strong fields are dominated by the states near the bottoms of the Landau bands which have a distinguished role relative to the rest of the spectrum. The investigated structure consists of a 50Å GaAs well imbedded between two 81Å Al<sub>0.33</sub>Ga<sub>0.67</sub>As barriers. Each Si doped GaAs contact consists of three layers: 4800Å with nominal doping 10<sup>18</sup>cm<sup>-3</sup> is followed by 4800Å with 2$`\times `$10<sup>17</sup>cm<sup>-3</sup> and the latter is separated from the barrier by undoped spacer of 300Å and 200Å, for top and bottom contact respectively. The lateral area of the nominally undoped quantum well is reduced to a 700Å diameter disk using the ion bombardment technique . This decreases the number of active impurities in the barrier, thus avoiding overlapping spectra of the LDOSF produced by individual spectrometers. A schematic band diagram of the resonant tunneling device with an impurity level $`S`$ in the quantum well is shown in Fig. 1, inset. By testing several samples, we have selected one with a distinct impurity level, which is also well separated from the states of the quantum well which lie about 10 meV above. This energy range determines the interval where the LDOS in the contact can be studied. At zero bias, the spectrometer level $`S`$ is above the Fermi level $`\mu `$ of the emitter with 3D metallic conduction. The alignment of $`S`$ and $`\mu `$ with increasing bias is registered as a step in IV. In the differential conductance $`G(V)`$ shown in Fig. 1, this current threshold corresponds to a peak at 0.05 V. At low temperatures, its height is $`G_\mathrm{\Gamma }\frac{4e^2}{h}\frac{\mathrm{\Gamma }_{\mathrm{max}}\mathrm{\Gamma }_{\mathrm{min}}}{\mathrm{\Gamma }_{\mathrm{max}}^2+\mathrm{\Gamma }_{\mathrm{min}}^2}`$ $`\frac{4e^2}{h}\frac{\mathrm{\Gamma }_{\mathrm{min}}}{\mathrm{\Gamma }_{\mathrm{max}}}`$ and its width is related to the energetic width of the spectrometer $`\mathrm{\Gamma }\mathrm{\Gamma }_{\mathrm{max}}`$ $``$ $`120`$ $`\mu `$eV determined by the tunneling coupling between the impurity and the contacts . The values of $`\mathrm{\Gamma }_{\mathrm{min},\mathrm{max}}`$ depend on the transparencies of the two barriers, so that $`\mathrm{\Gamma }_{\mathrm{max}}`$ corresponds to the lower (collector) barrier and $`\mathrm{\Gamma }_{\mathrm{min}}`$ to the higher (emitter) barrier, $`\mathrm{\Gamma }_{\mathrm{min}}5\times 10^3\mathrm{\Gamma }_{\mathrm{max}}`$ as estimated from the value of $`G_\mathrm{\Gamma }`$. The relation between bias $`V`$ and the energy scale of the spectrometer is established by the coefficient $`\beta =\frac{dE}{d(eV)}=0.24`$, found for the selected structure from the analysis of the temperature smearing of the threshold peak. Above the threshold, the current is determined by the emitter barrier transparency $`\mathrm{\Gamma }_{\mathrm{min}}`$ and the emitter density of states $`\nu `$ at the energy $`E_S`$ below the Fermi level. As the barrier height does not change significantly over a small $`V`$-range, the current becomes a measure of the LDOS in the emitter: $`I(V)\nu (E_S)`$. Fluctuations with energy of the LDOS give rise to the reproducible, temperature independent fine structure in IV. It is seen on top of a smooth decrease in the current reflecting the averaged 3D density of states. In some samples, we have detected Zeeman splitting of the spectrometer level in magnetic fields parallel and perpendicular to the current. In such cases it is seen that the upper spin level generates a replicated image of the LDOSF which is shifted along the $`V`$-scale with respect to the image produced by the lower level. We use this observation as a proof that two levels of the same spectrometers probe the same pattern in the LDOSF. The shift of the images corresponds to the difference between the spin-splitting of an impurity level in the quantum well and that of free electrons in the bulk. To avoid the overlap, a sample has been chosen with no significant splitting of the images in magnetic field. Fig. 2 shows the dependence $`G(V)`$ measured in magnetic fields $`0<B<10.5`$T applied parallel to the current and changed with a step of 20 mT. Fluctuations $`\delta G(V)`$ have a correlation voltage of $`\mathrm{\Delta }V_c0.5`$ mV, which is comparable to the width $`\mathrm{\Gamma }/e\beta `$ of the threshold conductance peak. With increasing magnetic field up to $`B`$ 4 T the fingerprint in $`G(V,B)`$ changes randomly, with a correlation field $`\mathrm{\Delta }B_c`$ 0.05T. At high fields, the fluctuations transform into a more regular pattern where individual features, assigned to Landau bands, tend to move with increasing field towards the threshold peak, similar to the observation by Schmidt et al . To interpret fluctuations $`\delta G(V)`$ as an image of LDOSF, we employ a picture based on the properties of single-electron wave functions, $`\psi _i(𝐫)`$ in a disordered metal . We want to stress that the LDOSF measured by the resonant tunneling spectroscopy reflect not only the randomness in the structure of chaotic wave functions obeying the Porter-Thomas statistics , $`\|\psi (𝐫)\|^{2n}=f(n)L^{nd}`$ ($`L`$ and $`d`$ are the sample size and dimensionality, $`f(n)=n!`$ in weak fields and $`\mathrm{}`$ means the averaging over disorder or energy), but also the existence of local correlations between energetically close eigenstates . In a phase-coherent 3D metal, the local density of states at a point $`𝐫`$ detected by spectrometer with width $`\mathrm{\Gamma }`$ can be considered as a sum of local densities, $`\|\psi _i(𝐫)\|^2`$, of all eigenstates within energy interval $`\mathrm{\Gamma }`$: $$I(V)\nu (E)\mathrm{\Gamma }^1\underset{\|E_SE_i\|<\mathrm{\Gamma }}{}\|\psi _i(𝐫)\|^2.$$ (1) The sum in Eq. (1) includes a large number of eigenstates, $`N(\mathrm{\Gamma },L)\nu _0\mathrm{\Gamma }L^d`$, each of which typically contributing as little as $`\|\psi (𝐫)\|^2L^d`$, with the mean value of the LDOS, $`\nu _0`$, independent of the sample size $`L`$. As far as fluctuations $`\delta \nu `$ are concerned, one might naively expect that these fluctuations should vanish upon enlarging the sample, since for the sum of $`N(\mathrm{\Gamma },L)`$ independently fluctuating values $`\delta \|\psi (𝐫)\|^2`$ $`L^d`$ the variance $`\delta \nu ^2`$ can be estimated as $`N(\mathrm{\Gamma },L)\left(\delta _L\|\psi (𝐫)\|^2\right)^2`$, which is equivalent to $`\mathrm{\Gamma }^2\nu _0\mathrm{\Gamma }L^dL^{2d}\nu _0\mathrm{\Gamma }^1L^d0`$ when $`L\mathrm{}`$. However, the correlations between wave functions at close energies make the LDOSF in a large sample finite and independent of its size. This statement can be explained using Thouless’s scaling picture of quantum diffusion . We construct the electron states in a large sample by representing them as linear combinations of wave functions defined in its smaller parts, one of which contains the observation point $`𝐫`$, and by gradually combining the smaller parts up to the actual size $`L`$ of the sample. For an intermediate length scale $`\xi `$ of the constituent part containing the observation point, its states (which we call ’mother’ states of generation $`\xi `$ ) are spaced by $`\mathrm{\Delta }(\xi )\left(\nu _0\xi ^d\right)^1`$. Diffusive spreading of these states into a larger part, when it is combined of several blocks, leads to their random mixing with the states from the neighboring $`\xi `$-size blocks within the Thouless energy $`\gamma hD/\xi ^2`$ , $`D`$ is the classical diffusion coefficient. Since at each stage only a finite basis is involved in the construction of the new states, some correlations exist between the new eigenstates, although at small $`\xi `$ the spread of ’mother’ states $`\gamma `$ is larger than $`\mathrm{\Gamma }`$ and these correlations are small. However, the Thouless energy $`\gamma `$ decreases with increasing $`\xi `$ , and once $`\xi `$ exceeds length $`L_\mathrm{\Gamma }=\sqrt{\mathrm{}D/\mathrm{\Gamma }}`$ , the information carried by a set of ’mother’ states from a generation $`\xi >L_\mathrm{\Gamma }`$ will not leave the energy interval covered by the spectrometer. As a result, the sum of the densities $`\|\psi _i(𝐫)\|^2`$ in Eq. (1) will only depend on the situation at the length scale $`L_\mathrm{\Gamma }`$ and not vary with further refining of the spectrum. Thus, $`L_\mathrm{\Gamma }`$ and $`N(\mathrm{\Gamma },L_\mathrm{\Gamma })`$ represent the largest length scale and number of states for which correlations between individual eigenfunctions could be neglected and the above naive estimate of $`\delta \nu ^2`$ from independent fluctuators used. Then, for the random difference between two values of $`\nu `$ in the neighboring $`\mathrm{\Gamma }`$-intervals one should take $`\delta \nu ^2N(\mathrm{\Gamma },L_\mathrm{\Gamma })\left(\delta _{L_\mathrm{\Gamma }}\|\psi (𝐫)\|^2\right)^2\nu _0\mathrm{\Gamma }^1L_\mathrm{\Gamma }^d`$. From this, we can now estimate the fluctuations in the current plateau regime of the differential conductance which is a measure of the derivative, with respect to energy, of the LDOS in Eq. (1). We normalise the variance $`\delta G^2`$ by the height of the threshold conductance peak which depends on the average LDOS and the spectrometer width $`V_\mathrm{\Gamma }=\mathrm{\Gamma }/\beta e`$, so that $`G_\mathrm{\Gamma }N(\mathrm{\Gamma },L_\mathrm{\Gamma })\|\psi (𝐫)\|^2/V_\mathrm{\Gamma }`$ . Because $`\delta G^2`$ can be taken from the above uncorrelated difference in $`\nu `$ as $`\delta G^2\delta \nu ^2/V_\mathrm{\Gamma }^2`$, we arrive at $$\delta G^2/G_\mathrm{\Gamma }^2N(\mathrm{\Gamma },L_\mathrm{\Gamma })^1\left(\mathrm{\Gamma }/hD\right)^{(d2)/2}/(\nu hD).$$ (2) We also estimate the correlation voltage of fluctuations as $`V_\mathrm{\Gamma }`$ and the correlation magnetic field as $`\mathrm{\Delta }B_c\mathrm{\Phi }_0/L_\mathrm{\Gamma }^2`$, where $`\mathrm{\Phi }_0`$ is the flux quantum. In a 3D system with an anisotropic diffusion tensor $`(D_x,D_y,D_z)`$, Eq. (2) transforms into $`\delta G^2/G_\mathrm{\Gamma }^2\left(D_xD_yD_z\right)^{1/2}`$. This relation also determines the classical effect on the variance $`\delta G^2`$ of a magnetic field $`𝐁=B𝐥_z`$. Assuming that the cyclotron motion suppresses transverse diffusion as $`D_{x,y}=D/(1+(\omega _c\tau )^2)`$ , gives : $$\delta G^2_B/\delta G^2_{B=0}1+(\omega _c\tau )^2,\omega _c\tau 1.$$ (3) Fig. 3 represents the result of our statistical analysis of conductance fluctuations in small magnetic fields. The amplitude of fluctuations is found from an individual $`G(V)`$ curve at a fixed $`B`$ in Fig. 2 as $`\delta G^2=(G(B,V)G(B,V))^2`$ ( $`\mathrm{}`$ stands for the averaging over range $`\mathrm{\Delta }V`$ $``$ 6 mV after the threshold peak ). To decrease the scatter, a further averaging over a $`B`$-range of 0.25T has been performed. The result is compared to that in Eq. (3). The increase in $`\delta G^2_B`$ agrees with the expected quadratic dependence. From Fig. 3, we find the momentum relaxation time $`\tau 0.9\times 10^{13}`$s and use it to estimate the mobility, $`\mu `$ = 0.22 m<sup>2</sup>/Vs, in the emitter. The obtained values agree with those expected for the emitter with the same nominal doping , and justify our use of the diffusion approximation since $`\tau \mathrm{\Gamma }/\mathrm{}10^2`$. We also use these values to estimate the zero-field diffusion coefficient, $`D40`$ cm<sup>2</sup>/Vs and $`\mathrm{\Delta }B_c\mathrm{\Gamma }/eD0.03`$T which is close to the experimental value. The value of $`\tau `$ confirms that the crossover from weak to strong fields, $`\omega _c\tau 1`$, takes place at $`B`$ 4T where the Landau band (LB) formation is seen in Fig. 2. In the $`\omega _c\tau 1`$ regime, the field dependence of the variance $`\delta G^2`$ has a strong oscillatory character similar to de Haas-van Alphen (dHvA) effect, with a sequence of peaks periodic in $`1/B`$ , Fig. 4a. However, the oscillations in $`\delta G^2`$ are much more pronounced than the oscillations in the threshold peak $`G_\mathrm{\Gamma }`$ reflecting the modulation of the average density of states at the Fermi level in the emitter caused by depopulation of LB’s, Fig.4b. Also, the observed oscillations look significantly enhanced when compared to the Shubnikov-de Haas (SdH) oscillations of conductance in a lateral GaAs MESFET structure with the same nominal doping as the emitter, Fig. 4a, inset. The positions of the peaks in the variance $`\delta G^2(B)`$ appear to be different from those in $`G_\mathrm{\Gamma }(B)`$ and correspond to the filling of the LB’s for the electron concentration of approximately $`3\times 10^{17}`$cm<sup>-3</sup>, which is larger than that near the barrier and suggests that electrons in the emitter further from the barrier could contribute to their origin. These enhanced dHvA-type oscillations in the fluctuation amplitude suggest that the above estimation of $`\delta G^2`$ using statistical properties of typical wave functions should be modified. This can be done by considering a special role of the states with anomalously large fluctuations of a local density, by analogy with where these states were ’prelocalized’ states. In the case of a smooth random potential with suppressed inter-LB scattering, these anomalous states are the states near the bottoms of LB’s.When in strong fields Abrikosov’s dimerization of electron motion takes place, the contribution to the LDOSF from the bottom of the highest filled LB becomes distinguished from typical LDOSF and dominates in the magnitude of the variance $`\delta G^2`$. For energies $`E_S`$ close to the bottom of the $`n`$-th LB, $`E_n`$ $`=(n+1/2)\mathrm{}\omega _c`$, not only transverse but also the longitudinal diffusion coefficient related to the highest LB, $`D_z^{(n)}u_z^2\tau \left[E_SE_n\right]`$, is suppressed due to the decrease in the kinetic energy of the quasi-one-dimensional electron motion along magnetic field. When the characteristic length scale $`L_\mathrm{\Gamma }^z=\sqrt{\mathrm{}D_z^{(n)}/\mathrm{\Gamma }}`$ becomes smaller than the inter-LB scattering length, the states from the upper LB start providing a contribution $`\delta ^{(n)}G^2`$ to the LDOSF that is enhanced compared to the typical variance $`\delta ^{(typ)}G^2`$: $$\delta ^{(n)}G^2/\delta ^{(typ)}G^2\left(\nu ^{(n)}/\nu \right)\left(D_0/D_z^{(n)}\right)^{1/2}$$ (4) The structure of Eq. (4) explains the enhancement of oscillations in the fluctuation amplitude in Fig.4 relative to oscillations of $`G_\mathrm{\Gamma }(B)`$ and SdH oscillations. The latter are the measure of the ratio of the LDOS in the highest ($`n`$th) LB and the total LDOS, i.e. they are represented by the first factor in Eq. (4). The unusual factor $`\left(D_0/D_z^{(n)}\right)^{1/2}`$, which is responsible for the enhancement of the oscillations of the fluctuation amplitude, is a specific feature of the LDOSF effect. We thank I. Lerner, R.J. Haug, T. Schmidt and A.D. Stone for discussions and EPSRC for support.
no-problem/9907/cond-mat9907031.html	ar5iv	text	# A density functional study of lithium bulk and surfaces ## I Introduction Lithium has been the subject of considerable interest over many years. Although its electronic structure is relatively simple, its structural properties still pose a significant challenge to both experiment and simulation . Lithium is very soft; the determination of its elastic constants and surface energies requires experiments of high accuracy and simulations of high numerical stability. The calculation of surface formation energies is particularly delicate as has been discussed recently . The aim of this article is to present a comprehensive and systematic study of the band structure, cohesive energy, elastic constants, phase stability, and surface energies of lithium. Very few systematic studies of the dependence of results on the computational parameters are available. This however is especially important when energy differences are required as for example in calculation of surface energies or phase stabilities. We present the results of fully converged, full potential, all-electron calculations based on both density functional theory and the Hartree-Fock approximation. We examine the use of finite temperature density functional theory as a technique for accelerating convergence with respect to reciprocal space sampling. We expand the crystalline orbitals as a linear combination of Gaussian type orbitals (LCGTO). This approach is very well established for insulators . We find, in accord with a recent study of magnesium , that this approach is also well suited to the simulation of a free-electron metal. The paper is organized as follows: in section II, we discuss the computational parameters. In sections III and IV, we discuss results for lithium bulk and surfaces, respectively, and summarize our conclusions in section V. ## II Basis set and Method All the calculations were performed with the program package crystal . The main numerical approximation in our approach is the choice of the Gaussian basis set. The difficulties of selecting basis sets for metallic systems have been explored in previous studies . In principle the quality of a calculation can be systematically improved by adding additional functions to the basis set and optimizing their exponents in a suitable reference state - usually the bulk crystal. In practice one must balance the overcompleteness of the basis set, which leads to linear dependence, with the need for additional variation freedom. For molecular systems and insulating solids these problems have largely been overcome and Gaussian basis sets are very widely used. For metallic systems and their surfaces in particular there have been very few systematic studies. In a solid the tails of the atom centered functions overlap strongly and so diffuse basis functions optimized for the description of atomic or molecular systems are not useful and indeed may give rise to linear dependence. We are thus unable to simply use basis sets from the many libraries developed for the description of molecular systems. We have therefore developed an hierarchy of basis sets of increasing quality in order to examine the convergence of computed properties. The smallest basis set used has 3 $`s`$-symmetry functions and 2 $`p`$-symmetry functions and is denoted as $`[3s2p]`$ . The $`[1s]`$ radial function was taken from Ref. . The exponents of the two additional $`sp`$-shells were optimized in local density approximation (LDA) calculations (with Dirac-Slater exchange and Perdew-Zunger’s correlation functional ) for the solid at the experimental lattice constant. The lowest energy was obtained with exponents of 0.50 and 0.08. Similar results were obtained when using the Perdew Wang gradient corrected approximation (PWGGA). However, as an exponent of 0.08 gives rise to a very diffuse basis function close to numerical instability, instead exponents of 0.50 and 0.10 were chosen. This $`[3s2p]`$ basis set is very robust and computationally efficient — it does not give rise to linear dependence even when the bulk is strongly distorted (for example, to determine the elastic constants). A $`[4s3p]`$ basis set was obtained by using three exponents (0.50, 0.20, and 0.08) - which were chosen to be ”even tempered”, i.e. the ratio between the exponents is kept fixed (2.5 in this case). This ratio is close to the lowest which can be tolerated before on-site (atomic) linear dependence is seen. It is however also known to converge the atomic energy to within less than $`10^4E_h`$ ($`E_h`$ = 27.2114 eV) of the exact Hartree-Fock ground state energy (see the analysis in Ref. ). Finally, an additional polarization function of $`d`$-symmetry was added and the exponent optimized within a PWGGA calculation to be 0.15. However, the $`d`$-function leads only to a minor change in the total energy. The energy varies only by $`5\times 10^5E_h`$ when, e.g., changing the exponent to 0.5. The basis sets developed in this manner are displayed in table I. Both at the Hartree-Fock (HF) and B3LYP (invoving a hybrid of Fock exchange and a modification of the Becke gradient corrected exchange functional, and the Vosko-Wilk-Nusair local correlation functional V and the gradient corrected correlation potential by Lee, Yang and Parr ) levels, an optimization of basis set exponents was not possible. Instead, the outermost exponent became more and more diffuse until finally the solution became unstable. This is a well known pathology of the use of Fock exchange in metallic systems (see also the discussion in Ref. ). When features of the HF solution are discussed in this article, they were obtained with the $`[3s2p]`$ basis set (outermost exponents 0.50 and 0.10). In order to compute binding energies the free atom is calculated within a spin-polarized formalism with the same $`[1s]`$ function but with additional $`s`$-exponents 0.60, 0.24, 0.096, 0.04, and 0.016 to describe the long range behaviour of the atomic wavefunction. For the LDA and PWGGA calculations we also expand the exchange and correlation potentials in an auxiliary Gaussian basis set which consists of 13 even tempered $`s`$-functions with exponents from 0.1 to 2000, 3 even tempered $`p`$-functions with exponents from 0.1 to 0.8, and 2 $`d`$-functions with exponents of 0.12 and 0.3. This is sufficient to integrate the charge density to an accuracy of 10<sup>-7</sup> $`\|e\|`$. For the free atom, we use an auxiliary basis set with 18 even tempered $`s`$-functions with exponents from 0.0037 to 4565. Reciprocal space sampling is a delicate problem especially in metals. The sampling is performed on a Pack-Monkhorst net where the density of points is determined by a shrinking factor. The Fermi energy and shape of the Fermi surface are determined by interpolation onto a ”Gilat” net. This net is simply related to the Pack-Monkhorst net by an additional subdivision factor. To further improve convergence, the finite temperature generalization of density functional theory can be used to apply Fermi surface smearing . In table II, the dependence of the total energy on the density of points in the Pack-Monkhorst net is displayed. For this purpose, PWGGA calculations on a body centered cubic (bcc) lattice at the equilibrium lattice constant of 3.44 Å were performed. At zero temperature, even with the largest net used, the energy is still slightly decreasing as more points are used. For a smearing of 0.001 $`E_h`$ (which corresponds to a temperature of $`T=\frac{0.001E_h}{k_B}=316K`$ with Boltzmann’s constant $`k_B=3.1667\times 10^6\frac{E_h}{K}`$), the energy is stable up to a few $`\mu E_h`$. A higher number of sampling points in the Gilat net leads to a systematic improvement at zero temperature. At finite temperature, the number of sampling points in the Gilat net does not influence the results when a sufficiently high number in the Pack-Monkhorst net is chosen. As shown in table II, the difference in energy for the different number of sampling points in the Gilat net is of the order of only a few $`\mu E_h`$ for a fixed shrinking factor of 24 in the Pack-Monkhorst net. Note that for the most dense net at $`k_BT=0.001E_h`$, the difference in energy between smeared and unsmeared results is less that $`10^4E_h`$. At an even higher temperature of 0.02 $`E_h`$, the energy converges to a value which is 6 $`mE_h`$ higher than at 0.001 $`E_h`$. An estimate of the zero temperature energy is possible by using the approximation $`E(0K)=\frac{1}{2}(E(T)+F(T))`$ (with $`F=ETS`$ being the free energy and exploiting the fact that the energy increases quadratically with temperature for low temperature) as suggested in Ref. . The electronic entropy $`S`$ is defined as $`S=k_B_i^{N_{states}}f_i\mathrm{ln}f_i+(1f_i)\mathrm{ln}(1f_i)`$ with $`f_i`$ being the Fermi function. This leads to a value of $`E(0)`$ extrapolated from $`k_BT=0.02E_h`$ which deviates by less than $`10^3E_h`$ from the value obtained at a temperature of 0.001 $`E_h`$. The functions $`E(T)`$, $`F(T)`$ and $`\frac{1}{2}(E(T)+F(T))`$ for a fixed value of 145 sampling points (corresponding to a shrinking factor of 16 in the Pack-Monkhorst net) are also displayed in Figure 1. Indeed, even for relatively high temperature, $`E(0)`$ is well approximated by $`\frac{1}{2}(E(T)+F(T))`$. In conclusion, as we need a high accuracy for the energy difference between the different phases of Li, we chose for the calculations on lithium bulk a shrinking factor of 16 for the Pack-Monkhorst net which gives 145 sampling points in the irreducible Brillouin zone of the non-distorted bulk and a temperature of 0.001 $`E_h`$. This ensures that convergence of the energy to at least $`10^4E_h`$ with respect to reciprocal space sampling is achieved. ## III Results for bulk Lithium ### A Band structure In figure 2 the LDA band structure for the bcc structure is displayed. The occupied bands and the lower unoccupied bands are in excellent agreement with results earlier obtained with Gaussian basis sets and $`X_\alpha `$ exchange , by the Kohn-Korringa-Rostoker (KKR) method and Slater exchange, augmented plane wave (APW) and modified APW (MAPW) calculations both using LDA, or linear muffin tin orbital (LMTO) calculations with a combination of exchange using the Langreth and Mehl functional and LDA correlation. When using PWGGA instead of LDA, the band structure does not exhibit major differences. The experimentally known conduction band width ($``$ 4 eV) is slightly lower than the result calculated here (4.6 eV). The slow decay of the density matrix in metals leads to difficulties when Fock exchange is involved: the summation of the exchange series in direct space is very long ranged and is truncated at a large but finite distance. This cutoff for large distances in direct space results in numerical instabilities for small $`\stackrel{}{k}`$. Thus, both in the Hartree-Fock and B3LYP band structures, artificial oscillations can be found around the Gamma point. As usual for Hartree-Fock calculations, we find that the bandwidth is roughly twice as large as the experimental bandwidth. ### B Cohesive properties Table III gives results for ground state properties of bcc Li (lattice constant, bulk modulus, cohesive energy and elastic constants C<sub>11</sub>, $`C_{11}C_{12}`$, and C<sub>44</sub>). The elastic constants were obtained by applying a rhombohedral distortion for $`C_{44}`$, a tetragonal distortion for $`C_{11}`$, and an orthorhombic distortion for $`C_{11}C_{12}`$ to the solid at the equilibrium lattice constant. Cohesive energy and lattice constant agree well with experiment and depend only weakly on the basis set. The bulk modulus is obtained from the energy as a function of volume and agrees within the accuracy of the fit with that obtained using the relation $`B=\frac{1}{3}(C_{11}+2C_{12})`$. Elastic constants have a strong dependence on the basis set and the deviation from experiment improves especially when going from $`[3s2p]`$ to $`[4s3p]`$; the $`d`$-function has only minor impact on the results. We find that PWGGA is closer to experiment with results similar to Ref. . ### C Relative stabilities The experimental crystal structure of lithium at zero temperature is still unclear (at room temperature a bcc structure is favoured). Both face-centered cubic (fcc) and hexagonal (hex) structures have been suggested as well as more sophisticated structures such as 9R (a nine layer sequence of close-packed planes ABCBCACAB) or mixtures of these phases. Therefore, we also investigated the relative stability of the different phases. In Figure 3, total energies of bcc, fcc and hex phase, obtained with the PWGGA functional and the best basis set ($`[4s3p1d]`$), are displayed. The c/a ratio of the hexagonal phase remains close to the ideal close packed value of 1.633 (it varies between 1.631 and 1.635 which is within the accuracy of the fit). We find that the closed-packed structures are slightly lower in energy than the bcc structure. Although it is reasonable to conclude that the close packed structures are lower in energy than bcc it is not possible to resolve the difference in energy between the fcc and hex phases. The variation of the energy with basis set is given in table IV. As the basis set is systematically improved the energy difference between bcc and fcc increases from 1 $`\times 10^5E_h`$ to 4 $`\times 10^5E_h`$, and between bcc and hex decreases from 7 $`\times 10^5E_h`$ to 4 $`\times 10^5E_h`$. We note that even the $`d`$-function influences this energy splitting. These results, which do not include the zero point energy, indicate a preference for closed-packed structures which is in agreement with most of the previous calculations except for one (see table IV). HF calculations were only possible with the smallest $`[3s2p]`$ basis set where the order of phases is different with hcp being the lowest in energy, followed by bcc and fcc being highest. The same is found in LDA with the smallest $`[3s2p]`$ basis set, but changes when the basis set is increased to $`[4s3p]`$ where both fcc and hcp are 2 $`\times 10^4E_h`$ lower in energy than bcc (again, calculations with the best $`[4s3p1d]`$ basis set were not possible because of linear dependence) — table IV. As the energy splitting between the close packed and bcc phases is rather small the zero-point energy difference cannot be neglected. The first published calculation of this, based on the harmonic approximation, gave an additional stabilization of $`9\times 10^5E_h`$ of the fcc phase compared to the bcc phase. In these calculations however the hcp phase was found to be higher in energy that the bcc phase (table IV). Very recently, in a calculation also including anharmonic effects, the stabilization was calculated to be about $`1.5\times 10^5E_h`$ both for fcc and 9R phases relative to the bcc phase . In addition, the authors computed the variation of the vibrational free energy as a function of temperature and found a phase transition from a closed packed to a bcc phase at a temperature of $`T200K`$. ## IV Results for surfaces A further test of this approach is the calculation of surface energies. We model lithium surfaces by using a slab and varying the numbers of layers (a bcc structure is assumed). Surface energies can be calculated in two ways, either by deriving a bulk energy by subtracting energies of two slabs with $`n`$ and $`m`$ layers: $$E_{surface}=\frac{1}{2}(E_{slab}(n)(E_{slab}(n)E_{slab}(nm))\frac{n}{m})$$ (1) which has the advantage of a systematic error cancellation (in particular the reciprocal space sampling is consistent between the bulk and slab energies) or by using an independent bulk energy $$E_{surface}=\frac{1}{2}(E_{slab}(n)E_{bulk}\times n)$$ (2) All the quantities $`E_{surface}`$, $`E_{slab}(n)`$, and $`E_{bulk}`$ are energies per atom. In Figure 4, results for surface energies using both equation 1 (with $`m`$=1) and 2 are displayed. Equation 1 leads to relatively strong oscillations (dotted line with pluses) and gets more stable the larger $`m`$. Numerical noise in the expression $`(E_{slab}(n)E_{slab}(nm))\frac{n}{m}`$ is reduced for larger values of $`m`$. Equation 2, however, leads at zero temperature to a slight linear decreasing of the surface energy as a function of the number of layers (thin line without additional symbol). The reason for the nonvanishing slope is that the energy difference $`E(n)E(n1)`$ is not identical to the energy of the bulk due to the systematic errors in the convergence of the total energy with respect to reciprocal lattice sampling (this was also emphasized in Refs. ). As shown in table II, at zero temperature the bulk energy is still changing on the order of $`10^4E_h`$ with increasing sampling point number. Similarly the bulk energy varies when extracted from the slab. This slight discrepancy gives rise to a variation of the surface energy with the number of layers with a slope of $`10^4E_h/atom/layer`$. The origin of the poor convergence with respect to reciprocal lattice sampling is due to the sharp cutoff imposed by the Fermi energy. One possibility to obtain the surface energy would be to use the intercept from a linear fit, but a better and simpler way to alleviate the difficulties associated with reciprocal lattice sampling is smearing the Fermi surface with a finite temperature. Already at a temperature of 0.001 $`E_h`$, the slope is virtually zero (thick line without additional symbol). This is consistent with table II as the bulk energy converges much faster at finite temperature. At a higher temperature of $`k_BT=0.02E_h`$, $`E(T)`$ clearly deviates from $`E(0)`$ (thin line with stars) and the approximation $`E(0)=\frac{1}{2}(E(T)+F(T))`$ should be applied. This works very well when comparing the corresponding results (thin line with crosses) with results calculated at $`k_BT=0.001`$ $`E_h`$ (thick line without additional symbol). The surface energy obtained this way is only slightly higher than that from a calculation at $`k_BT=0.001`$ $`E_h`$ which is consistent with figure 1. A higher smearing temperature also reduces the oscillations in the curves both when using equation 1 or 2. It even allows to substantially reduce the number of sampling points as shown in Figure 5. At low temperature, a high number of sampling points is necessary to obtain the correct result whereas at high temperature already a shrinking factor of 4 (resulting in 6 sampling points in the irreducible zone) is sufficient. It should be noted that in this case we used Equation 1 so that all the data is consistently extracted from calculations on slabs. In conclusion, calculations at zero temperature are very cumbersome when results from calculations on slabs and on the bulk have to be combined. The error cancellation can be maximized by extracting the surface energy from calculations on slabs only. The problem of error cancellation between bulk and slab is already improved at very low temperature when it is possible to fully converge the bulk energy with respect to reciprocal lattice sampling. Higher temperature also leads to a smoother behaviour of the surface energy as a function of number of layers. Finally, calculations at high temperature can be performed with a strong reduction of the number of sampling points and using $`E(0)=\frac{1}{2}(E(T)+F(T))`$ as an approximation for zero temperature results as suggested in Ref. . In table V, LDA and PWGGA results for the unrelaxed (100), (110) and (111) surface are summarized. The lattice constant was chosen as the bulk equilibrium lattice constant, a temperature of 0.001 $`E_h`$ and a higher shrinking factor of 24 resulting in 91 sampling points for the slabs and 413 sampling points for the bulk was used. As found in Ref. for the jellium model without additional long-range corrections, surface energies in PWGGA are lower than in LDA (the different lattice constants for PWGGA and LDA are not the reason, an evaluation at the LDA equilibrium lattice constant leads to a change of the PWGGA surface energy which is negligible compared to the difference between LDA and PWGGA surface energies). The energies are reduced when going from the smaller $`[3s2p]`$ to the $`[4s3p]`$ basis set, introducing a $`d`$ function leads only to minor changes. Our results agree well with the literature . ## V Conclusion In this study of lithium metal, we presented accurate results using a full potential, all electron density functional scheme. Results for cohesive properties, elastic constants, band structure, and surface energies are in full agreement with experiment and calculated values from literature. The results in best agreement with experiment were obtained with the gradient corrected functional of Perdew and Wang. Hartree-Fock calculations for lithium are very difficult as it is impossible to optimize the exponents because of the very long range of the exchange interaction; the same problems appear in functionals involving an admixture of Fock exchange. We demonstrated the convergence of the different properties with respect to the computational parameters by using an hierarchy of basis sets, different reciprocal lattice samplings and different smearing temperatures. We showed that finite temperature calculations can be used to improve convergence and still an extrapolation to zero temperature is possible and accurate. Quantities like cohesive energy and lattice constant are already stable with the smallest basis set; elastic constants and surface energies are more sensitive. The most difficult quantity to calculate is the energy splitting between the different phases where we have reached the limit of numerical accuracy. We can not make a prediction about the preferred crystal structure; the energy difference is so small that, from the computational point of view, even subtle changes such as introducing a $`d`$-function are important and zero point energies must be included . We confirm the finding of Ref. that an approach based on Gaussian type functions provides a reliable and very efficient description of metallic systems.
no-problem/9907/cond-mat9907148.html	ar5iv	text	# Mesoscopic fluctuations of the ground state spin of a small metal particle According to Hund’s rule, electrons in a partially filled shell in an atom form a many-body ground state with maximum possible spin. The maximum spin is preferred because it allows a maximally antisymmetric coordinate wavefunction in order to minimize the electrostatic repulsion between the electrons. In recent experiments, Hund’s rule was also observed in a cylindrically-shaped semiconductor quantum dot, or “artificial atom”. The close similarity with real atoms is due to the degeneracy of single-particle levels, caused by the the high degree of symmetry of the device. In generic ultrasmall systems such as small metal grains, semiconductor quantum dots, or carbon nanotubes there is no systematic degeneracy due to a spherical (or cylindrical) symmetric potential. However, even in the absence of degeneracies, a nonzero value of the ground state spin may occur, as long as the gain in electrostatic energy is larger than the loss in kinetic energy when an antisymmetric coordinate ground state wavefunction is formed. Such a ground state is most likely to be detected in ultrasmall metal and semiconductor devices, since in those systems, unlike in macroscopic samples, the spacing between single-particle energy levels and the typical interaction energies can be larger than the temperature. In fact, the possibility of such a “weakly ferromagnetic” ground state has been suggested as an explanation for some recent experiments, that could not be explained by simple noninteracting models. In addition, a nonzero ground state spin from numerical simulations of a few particles in a chaotic dot, and a theory of spin polarization in larger dots were already mentioned in the literature. The stability of the zero spin ground state in a quantum dot was analyzed for weak interactions in Ref. . In this paper, we consider small metal grains in the mesoscopic regime, in which fluctuations of wavefunctions and energy levels, caused by, e. g., disorder or an irregular shape, control the behavior of kinetic and interaction energies at the vicinity of the Fermi energy. As a result, the ground state spin becomes subject to sample-to-sample fluctuations. Then, the relevant quantity to consider is the statistical distribution of the ground state spin for an ensemble of small metal grains or chaotic quantum dots, rather than the spin of a specific sample. Our starting point is a simple toy model that captures the essential mechanisms for mesoscopic fluctuations of the ground state spin. In second-quantized form, our model Hamiltonian $``$ reads $``$ $`=`$ $`{\displaystyle \underset{n,m,\sigma }{}}c_{n,\sigma }^{}_0(n,m)c_{m,\sigma }^{}`$ (2) $`+uM{\displaystyle \underset{n}{}}c_{n,}^{}c_{n,}^{}c_{n,}^{}c_{n,}^{},`$ where $`c_{n,\sigma }^{}`$ ($`c_{n,\sigma }^{}`$) is the creation (annihilation) operator for an electron with spin $`\sigma `$ at site $`n`$. The indices $`m,n`$ are summed over $`M`$ sites. The Hamiltonian $`_0`$ contains the kinetic energy and the impurity potential. We describe the electron-electron interaction by an on-site (Hubbard) interaction, $`uM`$. While the long-range Coulomb interaction can be trivially included via a charging energy, the model (2) does not include the Coulomb interaction at intermediate distances, which leads to a Gaussian level spacing distribution at the Fermi energy. In this work, we report a calculation of the ground state spin of the Hamiltonian (2) using a restricted version of the Hartree-Fock (HF) approximation with a random-matrix assumption for the eigenvalues and eigenvectors of the self-consistent HF Hamiltonian. We first present our main result. It consists of an equation that relates the candidate ground state energies $`E_\mathrm{G}(s)`$ for different values of the total spin $`s`$ in terms of eigenvalues $`\epsilon _\mu ^0`$ of a hermitian random matrix with level spacing $`\mathrm{\Delta }`$, the interaction parameter $`\lambda =u/\mathrm{\Delta }`$, and a (nonuniversal) numerical constant $`c`$ that describes the density response to a local perturbation of the impurity potential in $`_0`$, $`E_\mathrm{G}(s)E_\mathrm{G}(s_0)`$ $`=`$ $`{\displaystyle \underset{\mu =1}{\overset{s}{}}}(\epsilon _{N+\mu +2s_0}^0\epsilon _{N+1\mu }^0)`$ (4) $`\lambda \mathrm{\Delta }\left[s^2s_0^2+{\displaystyle \frac{2(ss_0)}{\beta (1\lambda ^2c^2)}}\right].`$ The total number of electrons is $`2(N+s_0)`$, $`s_0`$ being $`0`$ or $`1/2`$. The spin of the true ground state is found by minimizing Eq. (4) with respect to $`s`$. The parameter $`\beta =1`$ ($`2`$) if time-reversal symmetry is present (absent). The effect of spin orbit coupling and Zeeman splitting is not included here. (The case $`\beta =2`$ is only relevant for semiconductor quantum dots in a weak magnetic field, that affects orbital motion, but causes no Zeeman splitting. It is not relevant for small metal grains, as laboratory magnetic fields do not affect orbital motion in this case.) Equation (4) reflects the competition between kinetic energy (first term on the r.h.s.), which favors small $`s`$ and the on-site interaction (second term), which favors finite $`s`$. The interaction term, in turn, consists of two parts: A term quadratic in $`s`$, which describes the exchange interaction, and a term linear in $`s`$, which describes the additional “dressed” Coulomb repulsion of two particles with the same spatial wavefunction. For large $`s`$, the contribution from the kinetic term is approximately $`s^2\mathrm{\Delta }`$, so that for $`u\mathrm{\Delta }`$ a finite fraction of the total number of spins will align, rather than a small number of spins as in the case $`u<\mathrm{\Delta }`$. The instability at $`u=\mathrm{\Delta }`$ is known as the Stoner instability. In Ref. , a result similar to Eq. (4) was obtained for $`s=1`$, but with a different and fluctuating interaction term. The difference is due to the absence of a self-consistent approximation scheme in Ref. . As a consequence of the additional dressed Coulomb repulsion of particles with the same wavefunction, we find that already for interaction strengths considerably below the Stoner instability $`u=\mathrm{\Delta }`$, there is an appreciable probability of nonzero ground-state spin. This is illustrated in Fig. 1, where the distribution of the ground state spin at three different values of the interaction parameter $`\lambda `$ is shown: Already at the quite modest interaction strength $`u0.4\mathrm{\Delta }`$ a ground state spin $`s=1`$ is more likely than $`s=0`$. The effect of a weak magnetic field is twofold: First, it changes the statistics of the $`\epsilon _\mu ^0`$, and, second, it suppresses the interference in the “Cooper channel”, leading to a factor of two reduction of the interaction term linear in $`s`$ \[last term in Eq. (4)\]. Both effects favor lower $`s`$ than without a magnetic field. However, even in the absence of a magnetic field we expect that, similar to 3D metals, inclusion of the electron–electron interaction beyond the HF approximation will also lead to a suppression of the interference in the Cooper channel (logarithmically in the system size $`M`$), and hence to a prefactor in that term that is smaller than 2. Let us now turn to the details of our calculation. To find the ground state of the Hamiltonian (2) we use a simplified version of the HF approximation: We assume that the ground state has the form of a Slater determinant of single-particle wave functions $`\psi _{\mu ,}`$ and $`\psi _{\mu ,}`$ of particles which have either spin up or spin down. In this case the self-consistent HF equations read $`_{\mathrm{HF},\sigma }\psi _{\mu ,\sigma }`$ $`=`$ $`\epsilon _{\mu ,\sigma }\psi _{\mu ,\sigma };`$ (5) $`_{\mathrm{HF},\sigma }(n,m)`$ $`=`$ $`_0(n,m)+uM\rho _\sigma (n)\delta (n,m),`$ (6) $`\rho _\sigma (n)`$ $`=`$ $`{\displaystyle \underset{\mu }{}}f_{\mu ,\sigma }\|\psi _\mu (n)\|^2.`$ (7) The occupation number $`f_{\mu ,\sigma }`$ is $`1`$ ($`0`$) if the level $`\mu ,\sigma `$ is occupied (unoccupied) and $`\delta (n,m)`$ is the Kronecker delta function. The ground state energy $`E_\mathrm{G}`$ is given by $$E_\mathrm{G}=\underset{\mu ,\sigma }{}f_{\mu ,\sigma }\epsilon _{\mu ,\sigma }uM\underset{n}{}\rho _{}(n)\rho _{}(n).$$ (8) Our strategy is as follows: We start from a reference state with zero spin, in which $`N`$ particles of each spin are placed in the same levels $`\epsilon _{\mu ,}=\epsilon _{\mu ,}`$ and with the same wavefunctions $`\psi _{\mu ,}=\psi _{\mu ,}`$. We assume, that for this symmetric case the eigenvectors and the eigenvalues of $`_{\mathrm{HF}}`$ are distributed like those of a random matrix, except that the energy levels below $`E_F`$ are shifted upwards, by a small constant amount, relative to the levels above $`E_F`$, see Eq. (22) below. (If we would have included long-range Coulomb interactions via a charging energy, the shift would have been much larger and in the opposite direction. Omission of the charging energy has no consequence in our case, as we compare ground states with the same number of particles.) The assumption that the single-particle eigenvalues and wavefunctions in a self-consistent potential for a mesoscopic system below and above $`E_F`$ obey random matrix statistics, even though they may be quite different from their counterparts in the noninteracting system, was checked numerically for short range interaction models somewhat similar to ours. The energy shifts in our case result from the spin-degeneracy, which was not present in these calculations. Starting from this reference state, we build other states by the subsequent addition and removal of electrons. We first discuss the addition of a single up spin in the $`(N+1)`$st level. The first question that needs to be answered is how this addition affects the self-consistent density $`\rho _\sigma (n)`$. The density change $`\delta \rho _{}(n)`$ consists of a direct and an induced contribution, while $`\delta \rho _{}(n)`$ has an induced density shift only, $`\delta \rho _{}`$ $`=`$ $`\delta \rho _{,\mathrm{dir}}+\delta \rho _{,\mathrm{ind}},\delta \rho _{,\mathrm{dir}}(n)=\|\psi _{N+1,}(n)\|^2,`$ (9) $`\delta \rho _{}`$ $`=`$ $`\delta \rho _{,\mathrm{ind}}.`$ (10) Since the density shifts change the HF Hamiltonians by an amount $`\delta _{\mathrm{HF},\sigma }(n,m)=uM\delta \rho _\sigma (n)\delta (n,m)`$, we obtain the following self-consistency equations for $`\delta \rho _\sigma `$, $`\delta \rho _{\sigma ,\mathrm{ind}}(n)`$ $`=`$ $`2uM\text{Re}{\displaystyle \underset{\mu ,\nu ,m}{}}f_{\mu ,\sigma }(1f_{\nu ,\sigma })\delta \rho _\sigma (m)`$ (12) $`\times {\displaystyle \frac{\psi _{\mu ,\sigma }^{}(n)\psi _{\mu ,\sigma }^{}(m)\psi _{\nu ,\sigma }^{}(m)\psi _{\nu ,\sigma }^{}(n)}{\epsilon _{\mu ,\sigma }\epsilon _{\nu ,\sigma }}}.`$ Both $`\delta \rho _{}`$ and $`\delta \rho _{}`$ are of order $`1/M`$. In Eq. (12) we have computed the induced density change to first order in $`\delta _{\mathrm{HF}}`$. Higher order terms do not contribute to $`\delta \rho _{\mathrm{ind},\sigma }`$ to order $`1/M`$ and are neglected. To evaluate Eq. (12), we first sum the r.h.s. over the space index $`m`$ and then over the energy levels $`\epsilon _\mu `$ and $`\epsilon _\nu `$. Because the eigenfunction has a random sign, a single term in the latter summation is of order $`\lambda M^{3/2}\mathrm{\Delta }/(\epsilon _\mu \epsilon _\nu )`$, which is not relevant in the limit $`M1`$, even if $`\epsilon _\mu `$ and $`\epsilon _\nu `$ are both close to the Fermi level $`E_F`$. For the summation over all levels we may perform an average over the wave functions (since the denominator is a slowly varying function of $`\mu `$ and $`\nu `$ away from the Fermi level). This average is done using that for general $`\mu \nu `$ and in the limit $`M1`$ one has, $$\psi _{\mu ,\sigma }^{}(n)\psi _{\mu ,\sigma }^{}(m)\psi _{\nu ,\sigma }^{}(m)\psi _{\nu ,\sigma }^{}(n)=\frac{\delta (m,n)}{M^2}\frac{1}{M^3}.$$ Putting everything together, we find the following solution of the self-consistency equations (10) and (12), $`\delta \rho _{}(n)`$ $`=`$ $`{\displaystyle \frac{1}{1(c\lambda )^2}}\left(\|\psi _{N+1,}(n)\|^2{\displaystyle \frac{1}{M}}\right)+{\displaystyle \frac{1}{M}},`$ (13) $`\delta \rho _{}(n)`$ $`=`$ $`{\displaystyle \frac{c\lambda }{1(c\lambda )^2}}\left(\|\psi _{N+1,}(n)\|^2{\displaystyle \frac{1}{M}}\right).`$ (14) where $`c`$ is a numerical constant of order unity defined by $$c=\underset{M\mathrm{}}{lim}\frac{2\mathrm{\Delta }}{M}_{\mathrm{}}^{E_F}𝑑\epsilon _1_{E_F}^{\mathrm{}}𝑑\epsilon _2\frac{\rho (\epsilon _1)\rho (\epsilon _2)}{\epsilon _2\epsilon _1}.$$ (15) In this equation, $`\rho (\epsilon )`$ is the mean density of HF energy levels. \[The mean level spacing $`\mathrm{\Delta }`$ is taken at the Fermi energy, $`\mathrm{\Delta }=1/\rho (\epsilon _F)`$.\] The constant $`c`$ gives the linear density response $`\delta \rho (n)`$ to a shift of the impurity potential $`_0(n,n)`$ at that same site, $`\delta \rho (n)=c\delta _0(n,n)/M\mathrm{\Delta }`$, which can be verified using first order perturbation theory with respect to $`\delta _0`$. Notice that $`c`$ is not a universal constant, but depends on an integration of the density of states over the entire bandwidth. For example, for the Wigner semicircular density of states we find $`c=4/3`$ if the Fermi energy $`E_F`$ is at the band center and $`c0`$ if $`E_F`$ is at a band edge. Equation (14) expresses that the interaction enhances the fluctuations of the spin density: if $`\|\psi _{N+1,}(n)\|^2`$ is larger than average, the on-site repulsion reduces $`\rho _{}(n)`$, which in turn causes an increase of $`\rho _{}(n)`$, and so on. At the same time, the interaction reduces fluctuations of the charge density $`\rho _{}+\rho _{}`$. At $`\lambda c=1`$, which may occur before the Stoner instability $`\lambda =1`$ if $`c>1`$, the density changes diverge. Although this instability signals a breakdown of our approach, it is not clear whether it will also lead to a true macroscopic ground state spin. Below, we restrict our discussion to the case $`\lambda c<1`$. Next we address the HF energy levels $`\epsilon _{\mu ,\sigma }`$ and find $`\delta \epsilon _{\mu ,}`$ $`=`$ $`\lambda \mathrm{\Delta }{\displaystyle \frac{2}{\beta }}{\displaystyle \frac{c\lambda }{1(c\lambda )^2}}\delta _{\mu ,N+1},`$ (16) $`\delta \epsilon _{\mu ,}`$ $`=`$ $`\lambda \mathrm{\Delta }\left(1+{\displaystyle \frac{2}{\beta }}{\displaystyle \frac{1}{1(c\lambda )^2}}\delta _{\mu ,N+1}\right).`$ (17) The shift of $`\epsilon _{N+1,\sigma }`$ is extra large, since for that level the interaction effects are enhanced by the spatial fluctuations of the wavefunction. Equation (17) is the result of first order perturbation theory in $`\delta \rho _\sigma `$; second order perturbation theory $`\delta \epsilon _{\mu ,\sigma }`$ gives a correction of order $`\lambda ^2\mathrm{\Delta }\mathrm{ln}M/M`$, which we may neglect in the limit $`M1`$. In the same way, one finds that the changes in each individual wave functions is not significant for $`M1`$. Finally, we consider the change in the ground state energy $`E_\mathrm{G}`$. Because the summation over $`\mu `$ in Eq. (8) extends over $`𝒪(M)`$ levels, it is important to follow the shifts in the HF levels to second order perturbation theory, although this level of accuracy was not needed for the shift of each level individually, cf. Eq. (17). Putting everything together, we find that $$\delta E_\mathrm{G}=\epsilon _{N+1,}c\lambda ^2\mathrm{\Delta }[\beta (1c^2\lambda ^2)]^1.$$ (18) (No terms proportional to $`\mathrm{log}M`$ appear here since they cancel in the summation over the energy levels.) With the help of Eq. (17), we can rewrite Eq. (18) as $`\delta E_\mathrm{G}=\epsilon _{N+1,}+\frac{1}{2}\delta \epsilon _{N+1,}`$, which is the average of the energy for the newly occupied level $`\epsilon _{N+1,}`$ before and after its occupation. This may be interpreted as a simple extension of Koopmans’ theorem to the present case, where the modification in each one-electron wavefunction is small (of relative order $`M^{1/2}`$), but the resulting contribution to $`\delta E_\mathrm{G}`$ cannot be neglected to the order we are interested in. In the usual form of Koopmans’ theorem, where one ignores any change in the one-particle wavefunctions, the HF energies of the lowest unoccupied state before addition of the electron, and of the highest occupied state after addition are identical. (The usual Koopmans’ theorem is correct for an infinite system, in general, or for a translationally invariant finite system, as the one-electron states are trivially plane waves in that case.) The simple extension of Koopmans’ theorem also works, in our model, for the addition of several electrons. We have repeated these calculations for the addition of two electrons with opposite spin in the $`(N+1)`$th level, $`\delta \rho _\sigma (n)`$ $`=`$ $`{\displaystyle \frac{1}{M}}+{\displaystyle \frac{1}{1+c\lambda }}\left(\|\psi _{N+1,}(n)\|^2{\displaystyle \frac{1}{M}}\right),`$ (19) $`\delta \epsilon _{\mu ,\sigma }`$ $`=`$ $`\lambda \mathrm{\Delta }\left(1+{\displaystyle \frac{2}{\beta }}{\displaystyle \frac{1}{1+c\lambda }}\delta _{\mu ,N+1}\right),`$ (20) $`\delta E_\mathrm{G}`$ $`=`$ $`2\epsilon _{N+1}+{\displaystyle \frac{2}{\beta }}\lambda \mathrm{\Delta }{\displaystyle \frac{1}{1+c\lambda }}.`$ (21) As in the case of the addition of a single particle, the individual wavefunctions do not change to order $`M^{1/2}`$. Equation (20) allows us to find the statistics of the HF energy levels $`\epsilon _{\mu ,\sigma }`$ in our reference system with $`N`$ electrons of each spin: The only distribution that is consistent both with the assumption that the $`\epsilon _{\mu ,\sigma }`$ obey random-matrix statistics away from the Fermi level and with the shifts of Eq. (20) is one where the $`\epsilon _{\mu ,\sigma }`$ have the form $$\epsilon _{\mu ,\sigma }=\epsilon _\mu ^0+\frac{2}{\beta }f_\mu \frac{\lambda \mathrm{\Delta }}{1+\lambda c},$$ (22) where the $`\epsilon _\mu ^0`$ have random matrix statistics and $`f_\mu =1`$ ($`0`$) if the level $`\mu `$ is (un)occupied. In other words, the distribution of the $`\epsilon _\mu `$ is the same as that of the eigenvalues of a random matrix, with all occupied levels shifted upwards by an amount $`(2/\beta )\lambda \mathrm{\Delta }/(1+c\lambda )`$. With the knowledge we have gained above, there is little work left for the calculation of our main result, Eq. (4). Some remarks about the validity of this result are appropriate. First, to make a connection between our random matrix toy model and a laboratory made quantum dot we must identify $`M=(L/\lambda _F)^3`$ as we expect the length scale for wavefunction correlations and the range of the screened Coulomb interaction to be of order of the Fermi wavelength $`\lambda _F`$. Second, while our solution is complete within the HF approximation, one must bear in mind that this approximation scheme does not include correlation effects, such as the Cooper channel renormalization, as we discussed below Eq. (4). Those correlation effects are not expected to affect our result to first order in $`\lambda `$, but it can not be excluded that they are important in the higher order terms in Eq. (4), which involve the factor $`c\lambda `$. We close this paper with a discussion of the physical consequences of a ground state with spin $`s>1/2`$ and of the experimental situations in which it can be observed. First, the temperature $`T`$ needs to be smaller than the separation of the ground states for different spin $`s`$. This separation, which is typically smaller than the single particle level spacing $`\mathrm{\Delta }`$, is a fluctuating quantity. Very small values are possible, because, unlike in noninteracting random systems, there is no level repulsion if states of different spin are involved. For sufficiently low $`T`$, the magnetization of the grain is proportional to the spin $`s`$ of the ground state. However, $`s`$ will also affect other properties which are more easily accessible in an experiment, like current–voltage characteristics: A nonzero ground state spin can serve to explain the absence of an even-odd structure in the addition spectra of Coulomb-blockaded quantum dots, or the presence of kinks in the parametric dependence of Coulomb blockade peak positions, as was noted in Ref. . Spin is also relevant for conductance measurements at a finite bias voltage, which allow for a “spectroscopy” of the quantum dot or metal grain. In the presence of a magnetic field, the ground state is split by the Zeeman energy, and the differential conductance will show two peaks, whose relative intensity differs by a factor $`2s_{N+1}+1`$ or $`2s_N+1`$, depending on whether the tunneling onto or from the grain is the faster process. Even without an external magnetic field the ground state may be split, e.g., by spin-orbit coupling or magnetic impurities, and thus give rise to a multiplet of peaks in the differential conductance. The peak separation within a multiplet is controlled by the strength of the splitting mechanism and may be much smaller than $`\mathrm{\Delta }`$. We thank D. Davidović and M. Tinkham for stimulating discussions and for sharing their experimental results with us. When this work was nearly completed, we learned of related work by H. U. Baranger, D. Ullmo, and L. I. Glazman (Ref. ), in which some similar results were obtained. We thank H. U. Baranger for discussions on these points. This work was supported in part by the NSF through the Harvard MRSEC (grant DMR 98-09363), and by grants DMR 94-16910, DMR 96-30064, DMR 97-14725.
no-problem/9907/cond-mat9907073.html	ar5iv	text	# 1. Introduction ## 1. Introduction The spin one ferromagnet has three states $`S_z=0,\pm 1`$ and can be used as a model of several interesting physical systems. For example, it can be applied to the qualitative study of the $`\lambda `$ transition and phase separation of <sup>3</sup>He-<sup>4</sup>He mixtures. In addition, it can model an Ising ferromagnet $`S_z=\pm \frac{1}{2}`$ with nonmagnetic impurities $`S_z=0`$ as well as a magnetic system with competing interactions. It can also model the behavior of a microemulsion where the interplay of the components, water, oil and amphiphile, is mimicked by the states $`S_z=0,\pm 1`$. Thus, studying the phase transitions and the phase diagram of the spin one ferromagnet has physical relevance. In this paper we analyze the properties of the spin one ferromagnet by introducing a transformation that maps the model hamiltonian onto that of the classical spherical model (CSM). Then we study the properties of the saddle points of the integrand of the partition function and look for the existence of phase transitions. The main result of this paper is the existence of a reentrant paramagnetic-ferromagnetic-paramagnetic transition. Consider a $`d`$-dimensional cubic lattice with three states $`N_\stackrel{}{R}=0,\pm 1`$ spin located at each of its nodes. Due to symmetry reasons we may assume that the states $`N_\stackrel{}{R}=\pm 1`$ are equivalent energetically, while the state $`N_\stackrel{}{R}=0`$ may have different energy. The physical motivation for these assumptions comes from the systems mentioned above. Assuming that the spins interact through the pair potential $`V(\stackrel{}{R}\stackrel{}{R}^{})`$ and with an external magnetic field $`H`$, then the general Hamiltonian that defines the model is written as $$=E\underset{\stackrel{}{R}}{}N_\stackrel{}{R}^2+\underset{\stackrel{}{R},\stackrel{}{R}^{}}{}V\left(\stackrel{}{R}\stackrel{}{R}^{}\right)N_\stackrel{}{R}N_\stackrel{}{R}^{}+H\underset{\stackrel{}{R}}{}N_\stackrel{}{R},$$ (1) where the second sum is over all pairs of spins. The first sum in $``$ has been introduced in order to have the possibility of quadrupolar ordering in addition to the magnetic ordering associated with the second sum. The competition of these two types of ordering is measured by the parameter $`E`$, the energy difference between the states with $`N_\stackrel{}{R}=\pm 1`$ and $`N_\stackrel{}{R}=0`$. With the aim at mapping the model hamiltonian onto that of the spherical model we introduce the Ising-like variables $`a_\stackrel{}{R}=\pm 1`$ and $`b_\stackrel{}{R}=\pm 1`$ that are related to the original spin variables by, $`N_\stackrel{}{R}=\frac{1}{2}\left(a_\stackrel{}{R}+b_\stackrel{}{R}\right)`$. Under this transformation the state $`N_\stackrel{}{R}=0`$ has been double weighted since it can be obtained from the states $`a_\stackrel{}{R}=b_\stackrel{}{R}=1`$ and $`a_\stackrel{}{R}=b_\stackrel{}{R}=1`$. To compensate this difference we introduce the additional entropy term $`𝒮=k_BT\mathrm{ln}2_\stackrel{}{R}N_\stackrel{}{R}^2`$, with $`k_B=1`$ the Boltzmann constant. In terms of the Ising-like variables $``$ becomes: $``$ $`=`$ $`{\displaystyle \frac{1}{4}}(ET\mathrm{ln}2){\displaystyle \underset{\stackrel{}{R}}{}}\left(a_\stackrel{}{R}+b_\stackrel{}{R}\right)^2+{\displaystyle \frac{1}{4}}{\displaystyle \underset{\stackrel{}{R},\stackrel{}{R}^{}}{}}V\left(\stackrel{}{R}\stackrel{}{R}^{}\right)\left(a_\stackrel{}{R}+b_\stackrel{}{R}\right)\left(a_\stackrel{}{R}^{}+b_\stackrel{}{R}^{}\right)`$ (2) $`+`$ $`{\displaystyle \frac{H}{2}}{\displaystyle \underset{\stackrel{}{R}}{}}\left(a_\stackrel{}{R}+b_\stackrel{}{R}\right).`$ To study analytically the properties of the model we replace the discrete Ising-variables by continuous ones $`\mathrm{}<a_\stackrel{}{R},b_\stackrel{}{R}<+\mathrm{}`$, subject to the spherical conditions: $`_\stackrel{}{R}a_\stackrel{}{R}^2=𝒩`$, and $`_\stackrel{}{R}b_\stackrel{}{R}^2=𝒩`$, with $`𝒩`$ the total number of sites (spins) in the system. Hence, the hamiltonian (2) together with the last two restrictions can be considered as the spherical version of the original hamiltonian. ## 2. Results Proceeding as in the CSM the partition function $`𝒵`$, after some straightforward calculations, can be written as $$𝒵=\pi ^𝒩\frac{ds_1ds_2}{(2\pi i)^2}\mathrm{exp}𝒩\gamma (s_1,s_2)$$ (3) where $$\gamma (s_1,s_2)=s_1+s_2\frac{1}{2𝒩}\underset{\stackrel{}{k}}{}\mathrm{ln}\left(s_1s_2+\left(s_1+s_2\right)\frac{\alpha _\stackrel{}{k}}{T}\right)+\frac{H^2}{16T^2}\frac{s_1+s_2}{\frac{\alpha (0)}{T}\left(s_1+s_2\right)+s_1s_2}$$ (4) and $`\alpha \left(\stackrel{}{k}\right)=\frac{1}{4}\left[ET\mathrm{ln}2+_\stackrel{}{R}V\left(\stackrel{}{R}\right)\mathrm{exp}(i\stackrel{}{k}\stackrel{}{R})\right]`$. The integral in Eqn. (3) can be approximated by evaluating the integrand at its saddle points: $`\frac{\gamma }{s_i}=0`$ with $`i,j=1,2`$ and $`ij`$. The solution $`s_i=s_j=s`$ is a root of the equation, $$\psi (s)=\frac{1}{𝒩}\underset{\stackrel{}{k}}{}\frac{1}{s+2\frac{\alpha (\stackrel{}{k})}{T}}+\frac{H^2}{4T^2}\frac{1}{\left(s+2\frac{\alpha (0)}{T}\right)^2}+\frac{1}{s}=4$$ (5) If $`H0`$ there is a saddle point at any temperature only if $`\mathrm{max}_\stackrel{}{k}\alpha (\stackrel{}{k})=\alpha (0)`$ which cannot be accomplished. On the other hand, for $`H=0`$ a saddle point may disappear if in the limit $`s2\alpha (0)/T`$ the sum converges to a finite value. Let us consider a system where $`V(\stackrel{}{R}\stackrel{}{R}^{})=J`$ and $`\stackrel{}{R},\stackrel{}{R}^{}`$ are nearest neighbors. Bearing in mind that for the CSM the absence (existence) of a saddle point is related to a ferromagnetic (paramagnetic) phase and assuming that $`H=0`$, Eqn. (5) can be rewritten as $$4=\psi \left(s\right)=\frac{1}{s}+\frac{T}{J}\frac{1}{(2\pi )^d}_{B.Z.}𝑑\stackrel{}{q}\frac{1}{\frac{T}{J}\left(s+\frac{E}{2T}\frac{1}{2}\mathrm{ln}2\right)\left(_{j=1}^d\mathrm{cos}q_j\right)}$$ (6) where we have replaced the sum by an integral over the Brillouin zone. The function $`\psi (s)`$ decreases monotonically and is analytic for $`s_0(J/T)<s+\mathrm{}`$ with $`s_0=\mathrm{max}(0,\frac{1}{2}\mathrm{ln}2(E2dJ)/2T)`$. Thus, a solution (a saddle point) of Eq. (6) exists if $`\psi (s_0)>4`$ (paramagnetic phase), while there is no saddle point if $`\psi (s_0)<4`$ (ferromagnetic phase). For $`d=1`$ it is easy to show that when $`ss_0`$, $`\psi (s)`$ diverges yielding no phase transition. For $`d=2`$ the evaluation of such integral is similar to that of the CSM and one can arrive to the conclusion that there is no phase transition . In $`d=3`$ for some values of the parameters $`E`$ and $`J`$, $`\psi (s_0)<4`$ and there is no saddle point in the integrand of $`𝒵`$, leading to a phase transition. A careful analysis of $`\psi [s_0(T),T]`$ indicates that there are two ranges of parameters to be considered: (1) $`E6J<0`$ in which case $`s_0(T)`$ goes from $`+\mathrm{}`$ at $`T=0`$ up to $`\frac{1}{2}\mathrm{ln}2`$ at $`T=\mathrm{}`$, and (2) $`E6J>0`$ for which $`s_0(T)`$ goes from zero at $`T=0`$ up to $`\frac{1}{2}\mathrm{ln}2`$ at $`T=\mathrm{}`$. Here following ref. we use the result $`(\frac{1}{2\pi })^3d^3q\left(3_{j=1}^3\mathrm{cos}q_j\right)^10.50541=I_{\mathrm{cr}}`$. Therefore $`\psi (s_0)`$ can be rewritten as $$\psi (s_0(T),T)=\frac{T}{J}I_{\mathrm{cr}}+\frac{2T}{T\mathrm{ln}2E+6J}=4.$$ (7) At very low temperature and for $`E6J<0`$, the parameter $`s_0`$ is quite large and $`\psi (s_0)`$ goes down to zero in which case the system settles in the ferromagnetic state ($`\psi (s_0)<4`$). Nonetheless, at high temperatures $`s_0`$ approaches to $`\frac{1}{2}\mathrm{ln}2`$ ($`\psi (s_0(T),T)>4`$) yielding the paramagnetic phase. In Fig. 1(a) we show the behavior of $`\psi (s_0(T),T)`$ as function of temperature for several values of $`E/J`$. There we see that for $`E/J<6.0`$ the function $`\psi (s_0(T),T)`$ increases monotonically as $`T`$ grows starting from $`T=0`$ up to the high temperature region. On the other hand, for $`E/J>6.0`$ and low temperatures it decreases until a minimum is reached at $`T^{}=(1/\mathrm{ln}2)\left(E6J+\sqrt{2J(E6J)/I_{\mathrm{cr}}}\right)`$, and then increases and goes asymptotically to the line $`\psi (E/J=6,T)`$ at higher temperatures. The critical line defined by $`\psi (s_0(T_c),T_c)=4`$, horizontal dotted line in Fig. 1(a), intersects the curves $`\psi (s_0(T),E/J)`$ in only one point if $`(E/J)<6`$, single critical temperature, while for $`6<(E/J)6+x_c`$ it intersects $`\psi (s_0(T),E/J)`$ in two points giving rise to two critical temperatures. Here $`x_c=(1/I_{cr})(2\sqrt{\mathrm{ln}2}\sqrt{2})^20.124`$. These curves will be denoted as reentrance lines for values of $`E/J`$ in $`[6,6+x_c]`$. In this range of parameters it is also found that the paramagnetic phase ($`\psi (s_0(T),T)>4`$) sets in at low and high temperatures while the ferromagnetic one ($`\psi (s_0(T),T)<4`$) only exists at the intermediate regime. Above the upper bound $`(E/J)_c=6+x_c`$, there will be no phase transition at all. The critical temperatures obtained as the roots of the critical line are: $$T_c^\pm =\frac{(E6J)I_{\mathrm{cr}}+4J\mathrm{ln}22J\pm 𝒟}{2I_{\mathrm{cr}}\mathrm{ln}2}$$ (8) with $`𝒟=\left[\left((E6J)I_{\mathrm{cr}}+4J\mathrm{ln}22J\right)^216(E6J)I_{\mathrm{cr}}J\mathrm{ln}2\right]^{1/2}`$ These two branches are plotted as function of $`E/J`$ in Fig. 1(b). The upper branch $`T_c^+(E/J)`$ goes from $`2.5JT_c<5.5J`$ while the lower branch $`T_c^{}(E/J)`$ extends from $`T_c=0`$ up to $`T_c=2.5J`$. In the inset of Fig. 1(b) we show the reentrance line, $`T_c(E/J)`$ for $`6\frac{E}{J}6+x_c`$. At this point it is important to note that the results presented here are stable for small changes of the entropy term $`𝒮=k_BT\mathrm{ln}2_\stackrel{}{R}N_\stackrel{}{R}^2`$ introduced as an extra term in the original $``$. ## 3. Conclusions In this paper we have shown analytically that the spherical version of the spin one ferromagnet with short range interactions and zero external magnetic field shows a reentrant phase transition to the paramagnetic state at low temperatures in $`d=3`$ for $`6<(E/J)6.124`$. However, in lower dimensions we find no phase transition as it occurs in the spherical model. We also obtained the phase diagram $`T_c`$ versus $`E/J`$ where the different phases are indicated. This mapping of the system onto its spherical version has proven to be very useful since it may help to shed light upon the physics of several interesting systems . For instance, we are currently applying this approach to the study of phase transitions in Josephson junctions arrays with disorder . An extended version of the present paper that includes more details as well as additional results will be published elsewhere . This work has been supported by CONACYT-MEXICO under contract 25298-E.
no-problem/9907/cond-mat9907493.html	ar5iv	text	# Dipoles and fractional quantum Hall masses ## Abstract We develop a microscopic formalism to study the fractional quantum Hall plateaus at filling factors $`\nu `$ away from $`1/2\beta `$ $`\beta `$ an integer. The theory is in terms of quasiparticles which carry a charge $`e^{}`$ equal to $`12\beta \nu `$ times the charge of the electron. The wave functions obtained following our approach are shown to coincide precisely with the form predicted by Jain and this holds independently of the interaction potential. Microscopically this rigidity originates from the fact that two different charges interacting attractively in their lowest Landau levels form a bound state with a universal wave function. From the expressions of the gaps we compute an effective mass which agrees well with the experiments carried at $`\nu =1/2`$ and $`1/4`$. Many fundamental aspects of the fractional quantum Hall effect (FQHE) have resulted from a better understanding of the vicinity of the $`\nu =1/2`$ filling factor. Jain was able to obtain the most prominent FQHE plateaus by reinterpreting the FQHE as an integer quantum Hall effect (IQHE) for particles which experience a reduced magnetic field . Halperin Lee and Read have predicted that the system behaves in a Fermi liquid way at the filling fraction $`\nu =1/2\beta ,\beta =1,2`$. Their theory provides a convincing explanation of the anomaly observed at $`\nu =1/2`$ (and less strongly at $`\nu =1/4`$) in the surface acoustic propagation by Willet et al. . Several experiments have seen that slightly away from $`\nu =1/2\beta `$ the electrons move in circles as if they experience a reduced magnetic field $`\mathrm{\Delta }B=BB_0`$. A phenomenological way to account for these properties is to assume that the relevant excitations, also called composite fermions (CF), carry a fractional charge varying continuously with the filling factor, $`e^{}=(12\beta \nu )e`$. Although such fractional charges have been observed in noise experiments , they have not been clearly identified with the CF. We propose here a microscopic derivation of the CF properties in terms of quasiparticles which carry a fractional charge. Each electron is replaced by a quasiparticle and the missing charge is carried by an incompressible charged vacuum. Another feature of the FQHE is that unlike in the IQHE, the gaps responsible for it are due to the interactions. Here we set mass of the electrons to zero and the CF acquire an effective cyclotron energy due to their interactions. The novelty is to predict microscopically a rigid structure for the wave functions independently of the interacting potential. When the filling fraction gets close to $`1/2\beta `$ the states connect continuously to the Fermi-liquid states at $`\nu =1/2\beta `$. The activation gaps are then given by the Landau Level spacing $`\mathrm{\Delta }=e^{}\mathrm{}B/m^{}`$ where $`m^{}`$ defines the composite-fermions effective mass. The masses we obtain are very sensitive to the parameter $`\lambda `$ which simulates the effect of the thickness of the sample. A very good agreement with experimental results is obtained when this parameter is of the order of the magnetic length . In order to understand the origin of the effective charge $`e^{}`$ and how the cyclotron energy results from the repulsive interactions one approach initiated by Read (see also ) suggests that the $`\nu =1/2`$ low energy quasiparticles are dipoles. This picture can be related to the trial wave functions if one notes that the effect of the Lowest Landau Level (LLL) projection inherent to these wave functions is to displace the electrons from their positions in a $`\nu =1/2\beta `$ bosonic Laughlin wave function. In we developed a theory where the fundamental excitations were these dipoles. We represented one excitation by two opposite charges interacting attractively in a magnetic field . They form a bound state which moves in a straight line and carries a dipole momentum perpendicular to its canonical momentum. When projected into the LLL the eigenfunctions are independent of the attractive potential. They can be expanded on the LLL basis of the particle and hole and the expansion coefficients form a matrix which is the projection of a plane wave into the LLL. To a first approximation the $`\nu =1/2\beta `$ CF can be described by non interacting dipoles and the wave function studied in is the antisymmetrized product of these projected plane waves. We follow the same approach to study the CF at filling fractions away from $`\nu =1/2\beta `$. The main modification is that the electron charge differs from that of the hole. The charge of the hole varies with the magnetic field so that a system of particles with this charge and a density equal to that of the electrons has a constant filling factor equal to $`1/2\beta `$. The reason for this choice is to construct a bosonic system which remains incompressible as the magnetic field is varied and which defines the vacuum of the CF excitations. In this description the CF is a particle hole excitation where the statistics and the charge of the particle differ from that of the hole. As a consequence the bound state made by the electron and its correlation hole acquires a charge $`e^{}=(12\nu )e`$ . In the following we use units where $`\mathrm{}=c=l=1`$, $`l=\sqrt{\mathrm{}c/eB}`$ is the magnetic length. The product of the charge of the electron and of the hole by the magnetic field $`B`$ are denoted $`q_a`$ and $`q_b`$. $`q_a=1`$ in our units and $`q_b=2\beta \nu q_a`$. Let us first consider a simple model which contains in essence the main points developed in this letter: A single electron and its correlated hole are coupled by a spring. Their coordinates are $`x_{a(b)},y_{a(b)}=\stackrel{}{x}_{a(b)}`$. In the Landau gauge their dynamics follows the Lagrangian: $`L=(q_ax_a\dot{y}_aq_bx_b\dot{y}_b)K/2((x_ax_b)^2+(y_ay_b)^2)`$ (1) where in (1) we have have taken the strong $`B`$ field limit which enables to neglect the masses of the particles to be neglected. For simplicity we consider here the case $`q_a>q_b`$. The momenta of $`y_a`$ and $`y_b`$ are $`p_a=q_ax_a`$ and $`p_b=q_bx_b`$. The Hamiltonian is: $`H=K/2((p_a/q_a+p_b/q_b)^2+(y_ay_b)^2)`$ (2) Its eigenfunctions are given by: $`\chi _{P_Y,n}(y_a,y_b)=e^{iP_Y(q_ay_aq_by_b)/(q_aq_b)}e^{(y_ay_b)^2/(q_b^1q_a^1)}H_n((y_ay_b)/(q_b^1q_a^1)^{1/2})`$ (3) where $`P_Y`$ is the momentum in the $`y`$ direction and $`H_n(x)`$ are the Hermite polynomials. The $`x`$ coordinate $`X=(q_ax_a+q_bx_b)/(q_a+q_b)`$ and the $`y`$-momentum $`P_Y=p_a+p_b`$ are related by $`P_Y=X(q_aq_b)`$. This means that the two charges move as a charge $`q^{}=q_aq_b`$ in the magnetic field. The harmonic oscillator frequency gives the effective mass of the bound state to be: $`q_aq_b/K`$. Assuming that the electrons behave as a gas of noninteracting such bound states, at the filling factors of the principal series $`\nu =n/2\beta n+1`$ the effective filling factor of the reduced charge $`q^{}`$ is equal to $`n`$, which explains the occurrence of the FQHE at these fractions. At $`\nu =1/2\beta `$, $`q^{}=0`$, and the Hamiltonian reduces to a free Hamiltonian $`H=P^2/2m^{}`$ so that we expect a Fermi sea to form. The momentum $`\stackrel{}{P}`$ and the relative coordinate are then related by $`P_X=q_a(y_ay_b),P_Y=q_a(x_ax_b)`$ so that the bound state behaves as a neutral dipole with a dipole vector perpendicular and proportional to its momentum. Note that since the strength $`K`$ of the spring enters the Hamiltonian (2) as a normalization factor the wave function (3) which describes the two charges is independent of $`K`$. If we replace the spring by a rotation invariant potential $`V(r)`$, one can refine the preceding approach to show that the wave function of the bound state is independent of the potential. To see it we need to consider the problem of the particle and the hole interacting in their respective LLL. The orbital of a charge $`q`$ particle in the $`n^{th}0`$ Landau level is denoted $`u_{n,t}^q`$ (simply $`u_t^q`$ for $`n=0`$) where $`t`$ is the momentum in the $`y`$ direction: $`u_{n,t}^q(x,y)=e^{ity}H_n(\sqrt{q}(xt/q))e^{q(xt/q)^2/2}`$ (4) The wave function for the charge and the hole can be decomposed on a basis of LLL orbitals: $`\chi _{P_Y,n}(x_a,x_b,y_a,y_b)={\displaystyle \underset{st=P_Y}{}}\widehat{\chi }_{st}^{P_Y,n}u_s^{q_a}(x_a,y_a)\overline{u}_t^{q_b}(x_b,y_b)`$ (5) and the coefficients $`\widehat{\chi }_{st}^{P_Y,n}`$ are obtained by diagonalizing the potential interaction: $`{\displaystyle \underset{s^{}t^{}}{}}st^{}\|V\|s^{}t\widehat{\chi }_{s^{}t^{}}^{P_Y,n}=ϵ_n^0\widehat{\chi }_{st}^{P_Y,n}`$ (6) The matrix elements $`st\|V\|s^{}t^{}`$ are taken between LLL orbitals of charge $`q_a`$ for $`s,s^{}`$ and $`q_b`$ for $`t,t^{}`$. Without loss of generality we restrict ourselves to the zero momentum case $`P_Y=0`$, $`(s=t,s^{}=t^{}),`$ and we consider a Gaussian potential $`V(r)=e^{r^2/2\mu ^2}`$. In the large system size limit, the momentum index $`s`$ becomes continuous and the secular problem (6) rewrites: $`{\displaystyle \frac{1}{1+1/\mu ^2q_b}}\sqrt{{\displaystyle \frac{1/q_b1/q_a}{\pi (1\tau ^2)}}}{\displaystyle _{\mathrm{}}^{\mathrm{}}}e^{(1/q_b1/q_a)((s^2+s^2)(1+\tau ^2)/22\tau ss^{})/(1\tau ^2)}\chi ^n(s^{})𝑑s^{}=ϵ_n^0\widehat{\chi }^n(s)`$ (7) $`\mathrm{with}\tau =(1+1/\mu ^2q_a)/(1+1/\mu ^2q_b)`$ (8) One recognizes the Poisson kernel for the Hermite polynomials. The eigenfunctions $`\chi ^n(s)`$ and eigenvalues $`ϵ_n^0`$ are thus given by: $`\chi ^n(s)=H_n((1/q_b1/q_a)^{1/2}s)e^{(1/q_b1/q_a)s^2/2}`$ (9) $`ϵ_n^0=\tau ^n/(1+1/\mu ^2q_b)`$ (10) The general eigenvector for $`P_Y0`$ can be put in the gauge independent form: $`\widehat{\chi }_{st}^{n,P_Y}=s\|u_{n,P_Y}^q^{}\|t`$ (11) Here as earlier the momentum $`s`$ and $`t`$ refer respectively to charge $`q_a`$ and $`q_b`$ orbitals in the LLL. This is our main result: the wave function of the bound state is universal and given by the projection of higher Landau levels in the LLL. The eigenvalues give the kinetic energy of the bound state and since the eigenvectors do not depend on $`\mu `$, they can be obtained for an arbitrary potential using a superposition principle. In the case of the potential $`V_\lambda (r)=1/\sqrt{r^2+\lambda ^2}`$ one has: $`V_\lambda (r)`$ $`=`$ $`\sqrt{{\displaystyle \frac{2}{\pi }}}{\displaystyle _0^+\mathrm{}}e^{(r^2+\lambda ^2)/2\mu ^2}{\displaystyle \frac{d\mu }{\mu ^2}}`$ (12) $`ϵ_n^0`$ $`=`$ $`\sqrt{{\displaystyle \frac{2}{\pi }}}{\displaystyle _0^+\mathrm{}}e^{\lambda ^2/2\mu ^2}{\displaystyle \frac{(1+1/\mu ^2q_a)^n}{(1+1/\mu ^2q_b)^{n+1}}}{\displaystyle \frac{d\mu }{\mu ^2}}`$ (13) If we consider the filling factors $`\nu _n=n/2\beta n+1`$ one has $`q_a=1,q_b=2\beta n/2\beta n+1`$. In the absence of interactions the gap $`\mathrm{\Delta }_{\nu _n}`$ is the energy needed to put a particle from the last occupied LL into the first empty one $`\mathrm{\Delta }_{\nu _n}=ϵ_n^0ϵ_{n1}^0`$. In the limit where $`n`$ is large the gaps become proportional to the charge $`q_n^{}=q_aq_b`$ and one can reinterpret them as an effective cyclotron energy for a particle with an effective mass $`\alpha ^0`$, $`\mathrm{\Delta }_{\nu _n}=q^{}/\alpha ^0`$. One obtains the following expression for the effective mass: $`{\displaystyle \frac{1}{\alpha ^0}}=\sqrt{{\displaystyle \frac{2}{\pi }}}{\displaystyle _0^{\mathrm{}}}e^{\frac{1}{2\beta (1+\mu ^2)}\frac{\lambda ^2}{2\mu ^2}}{\displaystyle \frac{d\mu }{(1+\mu ^2)^2}}`$ (14) We can see the main features of the inverse-mass behavior in this expression. It decreases mildly with the filling factor $`1/2\beta `$ and drastically with the thickness parameter $`\lambda `$. In the following we improve this formula to take into account the interactions between the CF. The characteristic feature of the orbitals (11) is that they contain no adjustable parameters and we expect that the wave functions obtained by antisymmetrizing them coincide with the trial wave functions which are by construction independent of the interactions. Nevertheless, in order to obtain quantitative results we need to determine the coupling between the hole and the electron in a consistent way. To reach this goal we use a second quantized formalism. The bosonic ($`b_t^+,b_t`$) and fermionic ($`a_s^+,a_s`$) creation and annihilation operators create respectively a charge $`q_b`$ boson with the y-momentum equal to $`t`$ and a charge $`q_a`$ electron with the y-momentum equal to $`s`$ in the LLL. The fields $`\mathrm{\Phi }_b(\underset{¯}{x})=_tu_t^{q_b}(\underset{¯}{x})b_t`$ ($`\mathrm{\Phi }_a(\underset{¯}{x})=_su_s^{q_a}(\underset{¯}{x})a_s`$) and its hermitian conjugate annihilate and create a boson (an electron) at position $`\underset{¯}{x}`$. The field $`\psi ^+(\underset{¯}{x})=\mathrm{\Phi }_a^+\mathrm{\Phi }_b(\underset{¯}{x})`$ removes a boson of charge $`q_b`$ at position $`\underset{¯}{x}`$ and replaces it with a fermion with charge $`q_a`$. The state $`\|\mathrm{\Omega }`$ defines the ground state of $`N`$ bosons at the filling factor $`2\beta ^1`$. If we act on $`\|\mathrm{\Omega }`$ with $`N`$ creation operators $`\psi ^+(\underset{¯}{x})`$ the resulting state contains only electrons. The idea which motivates the following construction is that the state $`\|\mathrm{\Omega }`$ is rigid against a density wave excitation so that the low energy physics results from the dynamics of the field $`\psi ^+(\underset{¯}{x})`$. When we act on $`\|\mathrm{\Omega }`$ with $`\psi ^+(\underset{¯}{x})`$ we assume that the separation of the electron and its correlation hole remains sufficiently small so that we can treat the pair as a bound state. We derive an effective Hamiltonian for these excitations which we break into a kinetic term and an interaction potential and we recover the wave functions of the bound states as the eigenstates of the kinetic term. Let $`\rho (\underset{¯}{x})`$ denote the electron density. The dynamics of the electrons is governed by the Hamiltonian: $`H=1/2{\displaystyle d^2\underset{¯}{x}d^2\underset{¯}{y}V_e(\underset{¯}{x}\underset{¯}{y})\rho (\underset{¯}{x})\rho (\underset{¯}{y})}`$ (15) where $`V_e(\underset{¯}{x})`$ is the electron-electron potential. In the operatorial approach the coordinate of CF operators acting on $`\|\mathrm{\Omega }`$ coincide with the position of the electrons and we can identify $`\rho `$ with the CF density operator $`\rho _a=\mathrm{\Phi }_a^+\mathrm{\Phi }_a`$. Since there is no boson in the physical system the bosonic density $`\rho _b=\mathrm{\Phi }_b^+\mathrm{\Phi }_b`$ is equal to zero. Therefore, if we substitute $`\rho `$ with $`\rho _a+\kappa \rho _b`$ in (15) the quantities we compute with this modified density are in principle independent of $`\kappa `$. The scheme we are using, however, does not allow us to set $`\kappa `$ to $`0`$ because the potential which binds the particle to the hole is equal to $`\kappa V_e(\underset{¯}{x}\underset{¯}{y})`$ and our approach thus requires a positive value of $`\kappa `$ to bind them. We can determine the best value for $`\kappa `$ so that the derivative of a physical quantity with respect to $`\kappa `$ is zero. We shall therefore work out the formalism using the above expression of $`\rho `$ in (15) and we shall set the value of $`\kappa `$ using this procedure when we compute the gaps. Alternatively, Shankar has suggested to determine $`\kappa `$ by letting the bosonic and fermionic particles interact proportionally to their charge and set $`\rho (\underset{¯}{x})=q_a\rho ^a(\underset{¯}{x})+q_b\rho ^b(\underset{¯}{x})`$ (in the $`\nu =1/2\beta `$ case this coincides with our proposal ). Although both procedures give comparable results at $`\nu =1/2\beta `$, the first one is more reliable away from this filling factor. To recast the dynamics in terms of CF it is useful to replace the composite operator $`a_s^+b_t`$ by a fermionic creation operator $`\psi _{st}^+`$. For this, we use the fact that the densities $`\rho _{ss^{}}^a=a_s^+a_s^{}`$ and $`\rho _{tt^{}}^b=b_t^+b_t^{}`$ entering the definition of $`\rho `$ in (15) are the generators of two commuting algebras $`U(q_a)`$ and $`U(q_b)`$ which have a natural representation in terms of matrix operators $`\psi _{st}^+`$. We define the CF creation and annihilation operators as a set of matrix fermions: $`\{\psi _{ts},\psi _{t^{}s^{}}\}=\{\psi _{st}^+,\psi _{s^{}t^{}}^+\}=0`$ (16) $`\{\psi _{st}^+,\psi _{t^{}s^{}}\}=\delta _{tt^{}}\delta _{ss^{}}`$ (17) and we represent the generators as: $`\left(\begin{array}{cc}a_s^+a_s^{}& a_s^+b_t^{}\\ b_t^+a_s^{}& b_t^+b_t^{}\end{array}\right)=\left(\begin{array}{cc}(\psi ^+\psi )_{ss^{}}& \psi _{st^{}}^+\\ (\psi \psi ^+\psi )_{ts^{}}& (\psi \psi ^+)_{tt^{}}\end{array}\right)`$ (18) It is easy to verify that the commutation relations of the matrix elements of the right and left matrix coincide and in fact, the right hand side generators can be viewed as a generalized Holstein-Primakov transformation for the left hand side generators. More familiar generators are given by the Fourier modes of the densities $`\rho ^a(\underset{¯}{x})`$ and $`\rho ^b(\underset{¯}{x})`$ which generate the two Girvin, MacDonald and Platzman algebras of the Fermionic and Bosonic densities . The effective Hamiltonian is obtained by substituting the densities $`\rho (\underset{¯}{x})`$ with their expression in terms of CF operators (17) in the Hamiltonian (15). In order to be useful this description requires the use of two approximations. The first one assumes the vacuum $`\|\mathrm{\Omega }`$ is annihilated by the CF operators $`\chi _{ts^{}}`$. It is justified if the gap a bosonic system at the filling factor $`1/2\beta `$ is considerably larger than the physical gap of the electron system and we can disregard the fluctuations of the density ($`\rho _b`$) in this state. In we considered a system of bosons at $`\nu =1`$ $`(2\beta =1)`$ where this condition is exactly realized. The second approximation consists in identifying the physical Hilbert space generated by $`a_s^+b_t`$ with the over-complete space generated by $`\psi _{st}^+`$. (It has been related to a gauge invariance by Read in ). A consequence is that in the CF description physical quantities depend on the bosonic interactions through the parameter $`\kappa `$. As a consistency check we shall verify that the value of $`\kappa `$ which minimizes the gaps is such that the kinetic energy contribution is large compared to the interactions. To obtain the physical CF creation operators and their kinetic energy we diagonalize the Hamiltonian in the one particle Hilbert space. If we denote by $`\psi _{n,P_y}^+=_{st}\widehat{\chi }_{st}^{n,P_y}\psi _{s,t}^+\|\mathrm{\Omega }`$ the eigenstates, the secular problem coincides precisely with (6). The eigenstates of the kinetic energy are thus generated upon acting on the vacuum $`\|\mathrm{\Omega }`$ with $`N`$ operators $`\psi _{n,P_y}^+`$ with $`\widehat{\chi }_{st}^{n,P_y}`$ given by (11). It is now straightforward to verify that the ground state wave function takes the form: $`\mathrm{\Psi }(\underset{¯}{x}_i)=𝒫^{(q_a)}\mathrm{\Phi }_n^{(q^{})}(\underset{¯}{x}_i)\mathrm{\Phi }_\mathrm{\Omega }^{(q_b)}(\underset{¯}{x}_i)`$ (19) where $`\mathrm{\Phi }_\mathrm{\Omega }^{(q_b)}(\underset{¯}{x}_i)`$ is the ground state wave function for $`N`$ bosons of charge $`q_b`$ at the filling factor $`1/2\beta `$ and $`\mathrm{\Phi }_n^{(q^{})}(\underset{¯}{x}_i)`$ is the Slater determinant of $`N`$ charge $`q^{}`$ orbitals filling $`p`$ Landau levels. The symbol $`𝒫^{(q_a)}`$ means that we expand the total wave function on the basis of Slater determinants of $`N`$ charge $`q_a`$ orbitals belonging to all possible LL and we project it on the subspace of determinants with all their orbitals in the LLL. It has the effect of replacing the higher orbitals $`u_{n,t}^q^{}`$ appearing in the factor $`\mathrm{\Phi }_n^{(q^{})}`$ by their matrix elements (11) between LLL orbitals of charge $`q_a`$ and $`q_b`$. This projection differs slightly from the one in which is defined to act separately on different factors of the total wave function. The CF excited states can be described in a similar way as the states of a Fermi liquid theory. We put the system in a box of area $`2\pi /q^{}`$ by keeping only the CF orbitals $`u_{n,t}^q^{}`$ with the position index $`t`$ equal to $`0`$. In presence of a magnetic field, the momenta of the CF must be replaced by the pseudo-momenta: $`\pi _x=p_x,\pi _y=p_y+q^{}x`$ which not commute and therefore cannot be diagonalized simultaneously. We can still define coherent states $`\|k`$ which diagonalize the complex pseudomomenta $`\pi _{}=\pi _xi\pi _y`$: $`\underset{¯}{x}\|k=e^{\frac{\overline{k}k}{4q^{}}}{\displaystyle \underset{m=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{1}{m!}}({\displaystyle \frac{k}{\sqrt{2q^{}}}})^mu_{m,0}^q^{}(\underset{¯}{x})`$ (20) An excited state is characterized by the occupation number $`n(k)`$ of the different coherent states $`\|k`$. The energy per unit area is obtained by taking the expectation value of the Hamiltonian (15) in this state : $`E_{\{n\}}={\displaystyle \frac{2\pi }{q^{}}}{\displaystyle \underset{k,k^{},p}{}}n(k)(1n(k^{}))\widehat{V}(\underset{¯}{p})\overline{k}\|\rho (\underset{¯}{p})\|k^{}\overline{k}^{}\|\rho (\underset{¯}{p})\|k`$ (21) (22) where the density $`\rho =\rho ^a+\kappa \rho ^b`$ and the matrix elements of $`\rho ^a`$ and $`\rho ^b`$ in the coherent states are given by: $`\overline{k}\|\rho ^b(\underset{¯}{p})\|k^{}=e^{(p\overline{p}(1+2\frac{q^{}}{q_b})+k\overline{k}+k^{}\overline{k}^{}2k\overline{k}^{}2\frac{\sqrt{q_a}}{\sqrt{q_b}}(\overline{k}pk^{}\overline{p}))/4q^{}}`$ (23) $`\overline{k}\|\rho ^a(\underset{¯}{p})\|k^{}=e^{(p\overline{p}+k\overline{k}+k^{}\overline{k}^{}2k\overline{k}^{}2\frac{\sqrt{q_b}}{\sqrt{q_a}}(\overline{k}pk^{}\overline{p}))/4q^{}}`$ (24) (25) The energy of a quasiparticle with pseudo-momentum $`k`$ ,$`ϵ(k)`$, is obtained by differentiating $`E_{\{n\}}`$ with respect to $`n(k)`$. When the filling factor $`\nu _n=n/(2n\beta +1)`$ gets closed to $`1/2\beta `$, the effective charge $`q^{}0`$ and the pseudo-momentum $`k`$ gets peaked around the level $`m`$ where $`m/n=\|\frac{k}{k_f}\|^2`$ and $`k_f=\sqrt{1/\beta }`$. In this limit, the ground state coincides with the state for which $`n(k)=1`$ for $`\|k\|<k_f`$ so that $`k_f`$ defines the Fermi momentum. We can also verify that the matrix elements (25) yield back the conservation of the momentum $`\delta (\underset{¯}{k}\underset{¯}{p}\underset{¯}{k}^{})`$ and the expression of the energy (22) converges towards the HF energy of the Fermi liquid at $`\nu =1/2\beta `$ : $`E_{\{n\}}`$ $`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \underset{k,k^{}}{}}n(\underset{¯}{k})(1n(\underset{¯}{k}^{}))\stackrel{~}{V}(\underset{¯}{k}\underset{¯}{k}^{})(1+\kappa ^22\kappa \mathrm{cos}(\underset{¯}{k}\times \underset{¯}{k}^{}))`$ (26) $`\mathrm{with}`$ $`\stackrel{~}{V}(\underset{¯}{k})=e^{k^2/2}\widehat{V}(\underset{¯}{k})`$ (27) The effective mass of the Fermi liquid is given by: $`k_f/\alpha =\frac{ϵ}{k}(k_f)`$. This gives a quadratic expression in the variational parameter $`\kappa `$ which we minimize with respect to $`\kappa `$. The inverse masses $`1/\alpha `$ are plotted in Fig.1 for the potential $`V_\lambda (r)=1/\sqrt{\lambda ^2+r^2}`$ and for the two filling factors $`\nu =1/2,1/4`$. As a consistency test of the approximation we can verify that the parameter $`\kappa `$ adjusts its value so that the effect of the interactions is small. Indeed the ratio $`\alpha /\alpha ^0`$ where $`\alpha ^0`$ is given by (14) has its minimum $`.8`$ for $`\lambda =0`$ and increases very fast to $`1`$. The Coulomb inverse mass ($`\lambda =0`$) at $`\nu =1/2`$ is slightly larger than the prediction of Jain and Kamilla : $`1/\alpha =.31`$. The gaps and the inverse mass decrease very fast with the thickness parameter $`\lambda `$ which reflects the fact that they are mainly due to the part of the potential of the order of the magnetic length. In order to compare this results with experiments we express the ratio of the effective mass to the bare mass of the electron as : $`{\displaystyle \frac{m^{}}{m_e}}={\displaystyle \frac{\mathrm{}\omega }{V_c}}`$ (28) $`\mathrm{}\omega =\mathrm{}eB/m_ec`$ is the cyclotron energy of a bare electron and $`V_c=e^2/ϵl`$ is the energy scale of the problem ($`ϵ=12.8`$ is the dielectric constant for GaAs). It gives $`\frac{m^{}}{m_e}=\mathrm{2.63\; 10}^2\alpha \sqrt{B}`$ ($`B`$ in tesla). A value of $`m^{}=0.26\sqrt{B}`$ in good agreement with the experimental observations is obtained for $`\lambda =1.2`$ (in units of $`l`$). This value has been argued by Morf to be a reasonable choice. For this value of $`\lambda `$ the masses at $`\nu =1/4`$ and $`\nu =1/2`$ differ very little, in agreement with Pan et al. . In conclusion we have proposed a simple analytical model to study the FQHE gaps in the vicinity of fillings $`\nu =1/2\beta `$ with $`\beta `$ an integer. Its main virtue is to predict microscopically wave functions with no adjustable parameter. Two main feature are predicted. The composite fermion mass depends little on the filling fraction $`1/2\beta `$ as seen in and the gaps collapse with the thickness parameter $`\lambda `$. The first feature is easy to understand: The filling fraction enters only the definition the Fermi momentum $`k_f=1/\sqrt{\beta }`$ and as long as the CF intraction is weak, the mass depends little on $`k_f`$. The second feature raises questions when we compare this theory to the scaling $`m^{}\sqrt{B}`$ seen in . The Thickness parameter $`\mathrm{\Lambda }=\lambda l`$ introduces a new length scale and we see only two ways to explain this scaling. Either $`\lambda `$ is independent of the density (in other words $`\lambda `$ scales like $`l`$) which needs to be explained or the mass depends little on $`\lambda `$ which contradicts our predictions. We note however that a strong dependence of the gaps in $`\mathrm{\Lambda }`$ may be related to the experimental observation of Shayegan et al .
no-problem/9907/cond-mat9907118.html	ar5iv	text	# High-𝑇_𝑐 Superconductors in Applied Magnetic Fields Parallel to the CuO Planes: First Order Transition with Slow Onset of Resistivity \[ ## Abstract The three dimensional uniformly frustrated XY model is used as a model of a high temperature superconductor in an applied magnetic field parallel to the CuO-planes. Through Monte Carlo simulations with anisotropy $`\eta ^2=10`$ on large lattices we find evidence for a first order transition. Earlier simulations and theoretical treatments are discussed and the experimentally found smooth onset of resistivity is suggested to be due to a large potential barrier against vortex line motion above $`T_c`$ present for perfect alignment of the applied field. \] The behavior of cuprate superconductors in applied magnetic fields has been, and continue to be, a very active area of research. The most studied geometry is with the applied field perpendicular to the CuO-planes. In that case, for clean samples, a first order transition associated with the melting of the Abrikosov lattice is by now well established both through calorimetric measurements and measurements of the jump in the magnetic induction. Also, a number of transport measurements are consistent with a first order transition. The geometry with the applied field parallel to the planes, has not been that thoroughly examined. It has, however, been found that the sharp increase in resistivity associated with the first order melting transition vanishes for perfect alignment of the field, and that the temperature-dependence of the resistivity is consistent with a continuous transition. This remarkable finding naturally rises the question of the possibility of a continuous vortex line lattice melting, and the commonly accepted theoretical picture is that the system undergoes a continuous transition analogous to the nematic to smectic transformation in liquid crystals. A more recent experiment suggests that the transition from a vortex liquid to a vortex solid phase takes place in two steps, with an intermediate ‘possible smectic’ phase. In the first Monte Carlo study of this model the authors examined a model with anisotropy $`\eta ^2=100`$ (see below), and argue that the behavior of both the helicity modulus and the specific heat suggest a continuous transition. However, a comparison between their Fig. 2 and corresponding data for a 2D XY model of the same size shows a remarkable similarity, suggesting that it is the individual layers that dominate the behavior of this three-dimensional system. This makes it difficult to draw any conclusions based on these results. In this Letter we report on a study of a similar system, but with anisotropy $`\eta ^2=10`$. The smaller value for the anisotropy is chosen with the hope to get a more clearly visible signal from the transition. The behavior of the system is clearly dependent on the system size. Only for very large systems does it become clear that the melting transition is first order instead of continuous. We also discuss problems with the suggested smectic phase and suggest a simple mechanism behind the apparently continuous onset of resistivity at the transition. The suggested scenario is directly open for experimental verification. For the study of superconductors in applied magnetic fields we make use of the uniformly frustrated anisotropic 3D XY model, with the Hamiltonian $$=\underset{i,\widehat{\mu }}{}J_\mu \mathrm{cos}(\theta _{i+\widehat{\mu }}\theta _iA_{i\mu }),$$ (1) where the sums are over the lattice points $`i`$ and $`\widehat{\mu }=\widehat{x}`$, $`\widehat{y}`$, and $`\widehat{z}`$. To include a magnetic field in the $`y`$-direction we choose $`A_{i\mu }=(2\pi /\varphi _0)_i^{i+\widehat{\mu }}𝐀𝑑𝐥`$ such that $$\frac{1}{2\pi }𝐃\times A_{i\mu }=f\widehat{y},$$ where $`𝐃`$ is the discrete difference operator. We use equal couplings in the $`x`$-$`y`$ planes, $`J_x=J_y=J_{xy}`$, but a different coupling, $`J_z`$, between the planes. The anisotropy is given by the parameter $$\eta =\sqrt{\frac{J_{xy}}{J_z}}=\frac{\lambda _z}{\lambda _{xy}}\frac{d}{\xi _{xy}},$$ (2) where the relation to the physical parameters, the penetration lengths $`\lambda _\mu `$ in the different directions, the inter-plane distance $`d`$ and the bare vortex core size in the planes $`\xi _{xy}`$, is given by the second equality. A derivation and justification of Eqs. (1) and (2) may be found in Ref. . To examine the behavior of this system we have simulated Eq. (1) with periodic boundary conditions, $`f=1/24`$, and several values for the anisotropy and system sizes. The results reported here are for $`\eta ^2=10`$ and $`L_x=L_z=48`$, and $`L_{}=48`$, 96, 192, and 384. We measure several different quantities. The superconducting coherence in the different directions as examined through the helicity moduli $`\mathrm{{\rm Y}}_x`$, $`\mathrm{{\rm Y}}_z`$, and $`\mathrm{{\rm Y}}_{}\mathrm{{\rm Y}}_y`$. To study the melting of the vortex lattice we measure the structure factor $`S(𝐤_{})`$ where $`𝐤_{}`$ is a wave vector perpendicular to the field and monitor $`\mathrm{\Delta }S=\|S(𝐊)S(R_x[𝐊])\|`$ where $`𝐊`$ is a reciprocal lattice vector of the ordered vortex lattice, and $`R_x`$ reflects $`𝐊`$ through the $`x`$ axis, and we average over the three smallest non-zero values of $`𝐊`$. Our Monte Carlo simulations consist, for the larger systems, typically of $`10^5`$ passes through the lattice for equilibration followed by $`1\times 10^6`$ through $`4\times 10^6`$ passes for calculating averages. The runs on the smaller systems were typically somewhat shorter. For the geometry with the field perpendicular to the CuO layers it has been found that the vortex lattice melting transition is best examined by starting with a disordered system and slowly lowering the temperature until the system finds an ordered configuration with a vortex lattice. However it turns out that the choice of $`L_{}`$ – the extension of the system in the field direction – is crucial in order to succeed with this cooling into an ordered state and obtain the correct transition behavior. With a too large value of $`L_{}`$ it becomes very difficult – or due to limited computer resources even impossible – to cool into a lattice. With a too small $`L_{}`$ the simulations incorrectly give at hand that there are two separate transitions with the vortex lattice melting taking place at the lower temperature followed by the vanishing of $`\mathrm{{\rm Y}}_{}`$ at a higher temperature. The proper choice of the length $`L_{}`$ depends on the parameters of the model, $`f`$ and $`\eta `$. In the first simulations with clear evidence of a first order transition, with $`\eta ^2=10`$ and $`f=1/25`$, this length was chosen as $`L_{}=40`$, but for isotropic systems, $`\eta ^2=1`$ and $`f=1/20`$, the evidence of a sharp transition was only found for $`L_{}=128`$ or $`L_{}=120`$. Guided by the experience with fields perpendicular to the CuO planes we therefore first tried slowly cooling the system, but we never succeeded cooling into a lattice. We therefore instead started the simulations with an artificially prepared initial configuration with a perfect vortex lattice and slowly increased the temperature. In this way we obtained results similar to Fig. 1 of Ref. . The vortex lattice melting as probed by $`\mathrm{\Delta }S(𝐊)`$ is shown together with $`\mathrm{{\rm Y}}_{}`$ in Fig. 1 for $`L_{}=48`$, 96, 192, and 384. This data shows that one needs at least $`L_{}192`$ for obtaining evidence that the vanishing of $`\mathrm{{\rm Y}}_{}`$ goes together with the melting of the vortex lattice. The similarity with data from Ref. for the field perpendicular to the planes strongly suggests a first order transition also for this case with the field parallel to the planes. From these simulations for $`\eta ^2=10`$ we can of course not immediately draw any conclusions regarding the character of the transition for larger $`\eta `$. Still, we will argue below that both the simulation results presented in Ref. and the experimentally found smooth onset of the resistivity are consistent with first order transitions. Since we also find problems with the continuous melting scenario, we suggest that the transition actually is first order for all finite values of the anisotropy. Turning to the simulation results for $`\eta ^2=100`$, we first note that the choice $`L_{}=40`$, or 80, in the light of Fig. 1, not is large enough for giving the correct behavior. But it also seems that their results are dominated by the behavior of the (weakly coupled) two-dimensional layers in the $`x`$-$`y`$ plane. To understand the relation to the 2D XY model we note that it is possible to choose a gauge with $`A_{ix}=A_{iy}=0`$, and the frustration included in the $`A_{iz}`$. Considered in this way, the system consists of some stacked unfrustrated 2D XY planes with a weak interplane coupling that at some points between the planes is ferromagnetic ($`A_{iz}=0`$), at some points is antiferromagnetic ($`A_{iz}=\pi `$), and otherwise is something in between. Fig. 2 of Ref. shows that the specific heat in the three dimensional model has a peak around $`T=1.05`$. The maximum value of the plotted quantity is $`24\times C36`$. The result was found to be about the same for all three different system sizes in the figure, of which the smallest was $`48\times 40\times 48`$. To compare with the specific heat for a single 2D layer, we have made simulations with $`L=44`$ which is chosen to have close to the same area as $`48\times 40`$. Recall that the peak of the specific heat in the 2D model is a non-singular feature well separated from the Kosterlitz-Thouless temperature $`T_{\mathrm{KT}}0.892`$. Our results for $`24\times C`$ for the 2D system are shown in Fig. 2. The similarity to the data of Hu and Tachiki (inset of their Fig. 2) is striking. Not only the shape of the curves is the same but also the position and the height of the maximum: at $`T1.05`$ we find the value $`36.6`$ which is the same as – or possibly slightly higher than – the data in their Fig. 2. We therefore conclude that the data for $`\eta ^2=100`$ of Ref. is dominated by 2D fluctuations. The absence of any sharp features in this data should therefore not be taken as evidence for a continuous transition in the 3D model. Why is the interplane coupling not more important? An obvious but rough way to assess the importance of the coupling between two planes is to compare the total coupling with the temperature. The size of the planes should then be large enough if $`L_x\times L_y\times J_zT`$, which for our present values becomes $`19.21.05`$. This would lead to the conclusion that the planes are large enough to make the interplane coupling very significant. The reason why this approach fails is the presence of the frustrating vector potential $`A_{iz}`$ that varies between 0 and $`2\pi `$. This means that the interplane couplings tend to cancel each other out, and we believe this to be the reason for the very 2D-like behavior found in the model. In the theory by Balents and Nelson the transition takes place in two steps with an intermediate smectic phase. The smectic phase is characterized by order in one direction only, which results in only certain layers being occupied with vortex lines and the others being empty. However, to us this picture doesn’t seem to be a convincing explanation of the experimental findings of Ref. for two reasons. First, the experimentally found smooth onset of resistivity was only found for very small angles, $`\theta <0.5^{}`$, but there seems to be no reason to expect that the transition between the smectic and the liquid phases should be very sensitive to the precise angle of the applied magnetic field. The transition between the tilted smectic phase and the liquid is also expected to be a continuous transition, which only changes character to first order for large tilts. Second, considering the thermal fluctuations of the vortex lattice, Balents and Nelson found that the average relative displacements in the $`x`$ and $`z`$ directions are unequal with $`u_x^2/a_x^2u_z^2/a_z^2`$. This was then taken to suggest that the Lindemann criterion for melting can be satisfied in the one direction only and not in the other, leading to a partial melting of the lattice. However, this line of reasoning presumes that $`u_z^2`$ not is strongly affected by the melting in the $`\widehat{x}`$ direction, but there is no a priori reason for this assumption. Figure 3 illustrates that the disorder in the $`\widehat{x}`$ direction may well give rise to forces that destroy the layered structure. This suggests problems with the stability of the smectic phase, which would seem to altogether invalidate the continuous melting scenario. So far we have both discussed the validity of the claim of a continuous transition from simulations and some problems with the theory of continuous melting. We now turn to the experimentally found smooth onset of the resistivity suggestive of a continuous transition. We will argue that this observation not is inconsistent with a first order melting transition. A current through the system perpendicular to the applied field gives rise to a force on the vortex lines perpendicular to both the current and the field. If the vortex lines move in response to this force the system dissipates energy, which is experimentally seen as a resistivity in the sample. In the geometry of Ref. with the current in the CuO planes and the applied field also parallel to these planes, dissipation is related to the motion of the vortex lines in the $`\widehat{z}`$ direction. An interesting observation is that the mechanism for motion of vortex lines in the $`\widehat{z}`$ direction becomes qualitatively different for perfect alignment of the field with the CuO layers as compared to a tilted magnetic field. This is illustrated in Fig. 4. Panel a) shows the situation in the presence of a tilted magnetic field where the upward motion of a vortex line is linked to the vortices moving to the left in the figure. Panel b) shows the qualitatively different situation for fields perfectly aligned with the CuO planes. In this case the motion of a vortex line has to start with a ‘kink’ that then grows bigger until the whole line has crossed the plane. The crucial thing to note is that such a kink gives rise to a vortex anti-vortex pair in the $`x`$-$`y`$ plane, and that this vortex pair separates as the vortex line gradually moves upward across the plane. The energy barrier to be overcome for motion of the vortex lines in Fig. 4a is the pinning of individual vortices. In contrast, the energy barrier in Fig. 4b is the kink energy barrier – related to the vortex anti-vortex interaction in the plane – which may be quite large just above $`T_c`$. If we assume that the transition implies a change of this potential barrier from infinite to quite large, the first order transition should not be clearly seen in the resistivity. In a 2D system this potential barrier diverges as $`TT_{\mathrm{KT}}`$, and as we expect $`T_cT_{\mathrm{KT}}`$ as the anisotropy increases, and that the behavior is 2D-like above $`T_c`$, we believe that this assumption of a large potential barrier right above $`T_c`$ is a reasonable one at least for fairly large anisotropies. With this mechanism for the transition the very strong dependence of the resistivity on the angle of the applied field follows naturally. Perfect alignment of the field corresponds to two-dimensional XY layers with only thermal vortex pairs and no free vortices in the low-temperature phase. But even a small non-zero angle of the field gives some additional free vortices, and the presence of them would lead to a qualitatively different behavior. The existence of a first order melting transition should be possible to verify from measurements of e.g. the specific heat. This would be a direct experimental test of our proposed picture, since in the alternative scenario with a continuous transition into a smectic phase there should be no sharp anomaly in the specific heat. However, so far the experiments are inconclusive. In Ref. the authors found features in the specific heat both for field parallel and perpendicular to the planes. The work was later extended to several different values of the angle of the field, including a parallel field, and it was found that the sizes of these features scaled with the angle in accordance with a simple scaling relation. This seems to point to the existence of a first order melting transition, but not conclusively. As the authors point out, the reason for the apparent first order character could also be that the angular resolution was not sufficient to access the true perfect alignment region $`\|\theta \|<0.5^{}`$. A carefully designed experiment should make it possible to discriminate between the two different proposed pictures. To conclude, we have found a single first order transition in the anisotropic 3D $`XY`$ model with magnetic field applied parallel to the planes. The finding that very large system sizes are needed for obtaining the true behavior make us doubt the continuous transitions reported from simulations on smaller systems. Some weaknesses with the commonly accepted theory of continuous vortex lattice melting have been discussed and we suggest that the reason for the smooth onset of resistivity is a large potential barrier right above $`T_c`$ associated with the separation of vortex anti-vortex pairs which is linked to the motion of the vortex lines in the $`\widehat{z}`$ direction. We acknowledge discussions with Professor Steve Teitel and Professor Petter Minnhagen. This work has been supported by the Swedish Natural Science Research Council through Contract No. E-EG 10376-312, and by the resources of the Swedish High Performance Computing Center North (HPC2N).
no-problem/9907/cond-mat9907220.html	ar5iv	text	# Magneto-thermodynamics of the spin-1/2 Kagomé antiferromagnet ## Abstract In this paper, we use a new hybrid method to compute the thermodynamic behavior of the spin-$`\frac{1}{2}`$ Kagomé antiferromagnet under the influence of a large external magnetic field. We find a $`T^2`$ low-temperature behavior and a very low sensitivity of the specific heat to a strong external magnetic field. We display clear evidence that this low temperature magneto-thermal effect is associated to the existence of low-lying fluctuating singlets, but also that the whole picture ($`T^2`$ behavior of $`C_v`$ and thermally activated spin susceptibility) implies contribution of both non magnetic and magnetic excitations. Comparison with experiments is made. Introduction: Spin systems with strong geometrical frustration exhibit a generic feature: the deviation of their spin susceptibility from the Curie law occurs at a much lower temperature than for non-frustrated ones and is often accompanied by very peculiar low-temperature properties . The two-dimensional spin-$`\frac{1}{2}`$ kagomé antiferromagnet (KAFM) is an example of such systems. All theoretical and numerical treatments point to a disordered Spin-Liquid ground-state , with a small spin-gap estimated to be of the order of $`1/20`$ of the coupling constant. But contrarily to what prevails in 1-D or ladders Spin-Liquids, we have shown that the first excitations of this system have a spin 0. The spectrum of these many-particle $`S_{\mathrm{tot}}=0`$ eigen-states forms a continuum adjacent to the ground-state, which extends smoothly up to energies higher than the spin-gap . In this paper we report some thermodynamic consequences of this property and focus on the magneto-thermal effects which are seemingly a unique signature of this kind of Spin-Liquids. Numerical method: The $`\frac{1}{2}`$-spins on the Kagomé lattice interact through the Hamiltonian: $$=J\underset{<ij>}{}𝐒_i.𝐒_j\gamma H\underset{i}{}S_i^z$$ (1) where the double sum is limited to first neighbors. The temperature $`T`$ and the magnetic field $`H`$ are measured in coupling constant units ($`J=1,\gamma =1`$). Up to now two kinds of methods were available to obtain thermodynamic quantities: high temperature series and exact diagonalizations. The first gives essentially exact results down to temperatures of the order of $`J/2`$, but unhappily no direct information on the range of temperature of interest in this work. The second allows the exact computation of the thermodynamics properties at all values of $`T`$, but for a set of sizes limited to $`N21`$. For such sizes and in spite of the supposed-to-be short range spin-spin correlations , their might be some uncertainties on the low $`T`$ thermodynamic properties related to the discrete nature of the finite size spectra. To deal with larger sizes and in order to obtain significant information on the effect of an applied field at low $`T`$ ( $`T\mathrm{0.05..0.2}`$), we have devised an hybrid method which takes advantage of both previous approaches. An approximate density of states in each Irreducible Representation (I. R.) of the complete symmetry group of the lattice and Hamiltonian is reconstructed via a maximum entropy procedure. It uses exact diagonalization data to fix the edges of the density of states and the six first moments of the Hamiltonian in each given subspace. In the $`N=36`$ sample there are 264 different I.R. and thus the procedure involves a very large amount of exact results. Most of the low $`T`$ physics comes from the relative position, weight and general shape of the different I.R., which are an essential input of our method. The knowledge of these quantities for each I.R. of $`SU(2)`$ allows a straightforward computation of the effect of any magnetic field. We have checked on different frustrated spin problems that this approach gives reliable information on the intermediate temperature range and on multiple-peak structures . Numerical results: High temperature range down to $`T=0.2`$: Numerical results are summarized in Fig. 1, together with the Pade approximant of the high-temperature series of Elstner and Young . In this range of temperature, the spin susceptibility $`\chi `$ shows a shoulder at $`T1`$ and the specific heat $`C_v`$ a high $`T`$ peak around $`T=2/3`$. The Pade approximants cannot be distinguished from the $`N=18`$ results down to $`T=0.4`$ and then merge smoothly with the exact $`N=24`$ low $`T`$ results. We thus infer that above $`T=0.4`$ these different data give a very good approximation of the thermodynamic limit. It should be noticed that there is only $`50\%`$ of the total entropy in the high $`T`$ peak above $`T=0.2`$, whereas the Néel ordered Heisenberg magnet on the triangular lattice develops $`50\%`$ of the total entropy above $`T=1`$, and only $`6\%`$ below $`T=0.2`$. The low $`T`$ entropy is indeed a distinctive mark of exotic spin liquids and the KAFM represents an extreme situation. From $`T=0.2`$ down to $`T0.2\mathrm{\Delta }`$: when cooling down the KAFM, one encounters a second energy scale $`\mathrm{\Delta }`$ related to the spin-gap . In the range of sizes presently studied, the gap is still decreasing and $`\mathrm{\Delta }`$ varies from $`\mathrm{\Delta }_{18}=0.104`$ to $`\mathrm{\Delta }_{36}=0.074`$ which explains the evolution of $`C_v`$ in Fig. 2, 3. From the study of spin-spin correlations and of finite size effects on the spin-gap one might expect that the $`N=36`$ sample is not very far from the thermodynamic limit. Remembering that $`50\%`$ of the total entropy has to be accounted for below $`T=0.2`$, one expects the low $`T`$ peak in $`C_v`$ to remain for $`N\mathrm{}`$. The spin-gap clearly explains the thermally activated spin susceptibility $`\chi `$. However, $`C_v`$ (in contrast to $`\chi `$) does not decrease exponentially in the gap. For the larger samples ($`N=1836`$), $`C_v`$ below the peak is well fitted by a $`T^2`$ law in the range $`T=0.30.6\mathrm{\Delta }`$ (Fig. 2). The non exponential behavior of $`C_v`$ is indeed due to the presence of singlets in the gap, but, as can be seen in Fig. 3, higher spin channels do contribute approximately to one half of $`C_v`$ in this range of temperature and the $`T^2`$ behavior appears as a cooperative effect due to different spin channels. This fact contradicts an intuitive expectation: from our knowledge of fully gapped systems one would naively expect an absence of contribution of the triplets for temperatures lower than the spin-gap. In fact $`C_v`$, at temperature $`T`$, essentially measures the energy fluctuations around the average energy $`e(T)`$. In a fully gapped system, when $`T`$ is lower than the spin-gap, the ground-state alone is populated and the energy fluctuations are thermally activated. In the present case, the thermal population at a temperature $`T`$ is dominated by the singlet states with an energy $`e(T)`$: this large number of relay levels in the spin-gap and the very large number of low lying triplets are at the origin of this surprising behavior. There has been in the past many speculations about the origin of the $`T^2`$ behavior in $`SrCrGaO`$ and in the jarosites : simple or nematic spin-waves modes have been suggested . Strictly speaking the presence of a spin-gap precludes such an explanation in the spin-$`\frac{1}{2}`$ case, but the small spin-gap is a proof that the spin-$`\frac{1}{2}`$ KAFM is relatively near a quantum critical point. An alternative explanation can also be featured by focussing on the Resonating-Valence-Bound picture of the singlets : in such a picture the modes are associated to the dimer-dimer long range orientational order (resonons of Rokhsar and Kivelson ). Such a picture gives a $`T^2`$ behavior of $`C_v`$ provided that the dimer-dimer structure factor is linear in k at small wave-vectors. This would imply an algebraic decrease of the dimer-dimer correlations. Regarding the range of sizes available, this hypothesis cannot be completely discarded (dimer-dimer correlations at distance 6 are of the order of $`10^2`$ in the spin-$`\frac{1}{2}`$ KAFM , whereas they are about $`10^4`$ in the M.S.E. ). In this last hypothesis, a global picture involving both singlet and magnetic excitations should be worked out. Temperatures lower than $`0.2\mathrm{\Delta }`$: The $`C_v`$ curves for $`N=18`$ and $`N=36`$ are numerically exact in this temperature range. The entropy of the singlet excitations in the spin-gap ($`15\%`$) is certainly a significant information in the thermodynamic limit. But the low-temperature peak (dots in Fig. 1, 2 and 3) is unphysical and a consequence of the discreteness of the spectra. Large size effects in the lowest part of the energy spectrum preclude its extrapolation for $`N\mathrm{}`$, in particular, they make it impossible to decide the exact value of the entropy at $`T=0`$ and the very low $`T`$ behavior of $`C_v`$ in the thermodynamic limit. Comparison with experimental results: The comparison with experimental results is only indicative, insofar as only pseudo-spin-1 , spin-$`\frac{3}{2}`$ and spin-$`\frac{5}{2}`$ Kagomé lattices have been realized and studied in the laboratory. Pseudo Spin-1: The organic compound $`mMPYNN.BF_4`$ studied by Wada and coworkers may behave at low enough temperatures as a spin-$`1`$ KAFM. $`\chi `$ has been measured in a large range of temperature. There are similarities with the present spin-$`\frac{1}{2}`$ results: a shoulder at $`T1`$, a thermally activated behavior with a spin-gap of the order of $`\mathrm{7.7\hspace{0.33em}10}^2`$. It might be noticed that the experimental gap is large enough compared to the spin-$`\frac{1}{2}`$ case; half integer and integer spin systems might indeed behave differently. Spin-$`3/2`$ systems: We will concentrate here on $`SrCrGaO`$ which displays the most convincing example of a non-conventional low-temperature behavior . From the magnetic point of view, it is not clear whether $`SrCrGaO`$ is a good representative of a system of Kagomé planes (with intervening dilute triangular Heisenberg planes) or whether it is better described as a stack of 2-dimensional slabs of pyrochlore. However, on the qualitative level these two theoretical models display similarities, absence of long range order, resonating valence-bond ground-state, so that a comparison of the experimental results for $`SrCrGaO`$ with our numerical results might be justified . The double peak structure in $`C_v`$ has not be seen in these compounds: this is not a surprise at the light of present results. In $`SrCrGaO`$ the Curie-Weiss temperature is of the order of $`500K`$, the coupling constant of equation (1) is thus of the order of $`100K`$ and from results shown in Fig. 3 we expect the low-$`T`$ peak to appear at $`T0.05J5K`$ or below. On the other hand the scale for the semi-classical behavior is given by $`S(S+1)J`$: we thus predict that the high $`T`$ peak in $`SrCrGaO`$ should appear around $`350K`$, but experimental data only extend to $`40K`$. We might notice that in $`SrCrGaO`$ as in the $`S=\frac{1}{2}`$ numerical data, the low-$`T`$ peak in $`C_v`$ accounts for half of the total entropy of the spin system . Endly the $`T^2`$ behavior of $`C_v`$ is consistent with the experimental data. The stronger argument in favor of a parentage between the spin-$`\frac{1}{2}`$ KAFM and $`SrCrGaO`$ is the quasi absence of sensitivity of $`C_v`$ to applied magnetic fields $`H`$ as large as twice the temperature reported by Ramirez et al. . This is a highly surprising result, which cannot be explained in any of the following hypothesis: Néel long range order, ordinary spin glass, Spin-Liquid with a full gap both for magnetic and non magnetic excitations. i) Since a Néel-ordered system spontaneously breaks the rotational invariance, the ground-state (and the first excitations) are superpositions of quasi-degenerate eigenstates with different total spin. In such a case, $`\chi `$ goes to a constant at low $`T`$. Though the magnetization per spin vanishes, the typical spin is not zero: $`S_{\mathrm{tot}}^2N\chi (T=0)`$. This gives a $`H`$ dependent $`C_v`$. ii) On the other hand in a fully gapped Spin-Liquid, as soon as the applied field is large enough to match the gap, $`C_v`$ is completely washed out, as for example in the Spin-Peierls compounds. In $`SrCrGaO`$ a field of 8 to 11 Tesla is large enough to excite magnetic excitations (as can be seen in looking at the non zero values of $`\chi `$ at equivalent temperature) and nevertheless there is no effect on $`C_v`$. We show on the spin-$`\frac{1}{2}`$ example below, that the presence of an important background of non-magnetic excitations is enough to drastically decrease the effect of the field on $`C_v`$. The magneto-thermal effect on spin-$`\frac{1}{2}`$ KAFM is displayed in Fig. 3 for a field equal to the temperature of the peak $`T_p`$ (corresponding approximately to the 6 Tesla experiments of Ramirez et al. ). i) The contribution of singlet excitations to $`C_v`$ is displayed on the same graph. In the interesting range of temperature, the singlets only account for half of $`C_v`$. ii) The higher spin channels are responsible for the missing half of $`C_v`$. These higher spin channels are indeed fully sensitive to $`H`$. If these magnetic levels were separated from the singlet ground-sate by a spin-gap void of singlet excitations the decrease of the $`C_v`$ peak would be 3 times larger than it is in the present situation. Here again, the presence of the background of singlets brings a great change in the fluctuation spectrum of this system and explains its paradoxical low sensitivity to $`H`$. iii) The effect of a field of the order of $`T_p`$ seems larger in the spin-$`\frac{1}{2}`$ KAFM than in $`SrCrGaO`$: this suggests that in this range of $`T`$ the relative weight of the singlets is larger in $`SrCrGaO`$ than it is for spin-$`\frac{1}{2}`$. We suggest the following conjecture: $`SrCrGaO`$, as a spin-$`\frac{3}{2}`$ system, is closer to the quantum critical transition than the spin-$`\frac{1}{2}`$ KAFM. Its spin-gap, if not zero is certainly smaller than that of the spin-$`\frac{1}{2}`$ KAFM, and there may be a collapse of the different low-energy scales appearing in the low $`S_{tot}`$ sectors of the spin-$`\frac{1}{2}`$ system (see Fig. 3) . In such a situation we expect a weaker sensitivity of $`C_v`$ to $`H`$. An important issue in experimental systems is the problem of dilution. We have checked the effect of quenched disorder on the spin-$`\frac{1}{2}`$ KAFM by including 1 or 3 non magnetic sites in a $`N=21`$ sample. The general qualitative properties of the KAFM (thermally activated $`\chi `$, low-$`T`$ peak in $`C_v`$ and singlets in the spin-gap) do not disappear with such dilutions! This seems an interesting indication that the thermodynamic features we have been describing in this paper are rather robust. Conclusion: In this paper we describe the results of a numerical investigation of the magneto-thermal properties of the spin-$`\frac{1}{2}`$ KAFM. We show that the specificities of its low energy spectrum induce a very low sensitivity of the specific heat to the external magnetic field. Such an effect is a very strong signature of a large weight of singlet excitations and would not be present neither in a Néel ordered system, in a standard Spin-Glass nor in a dimerized Spin-Peierls like system. We have verified that this effect is qualitatively insensitive to quenched disorder. The qualitative similarities to Ramirez et al results on $`SrCrGaO`$ are probably more than a coincidence but a new independent proof advocating for the presence of low lying fluctuating singlets in this spin-$`\frac{3}{2}`$ compound. The precise nature of the lattice (Kagomé or pyrochlore slab) might not be a central issue. We have preliminary results showing that similar Spin-Liquid behavior does appear on many other lattices. We now suspect this behavior to be a generic feature of one class of Spin-Liquids. Acknowledgments: We thanks A. Ramirez for discussion and communication of his results a long time before publication. We have benefited from very interesting discussions with J.P. Boucher, N. Elstner, M. Gingras, S. Kivelson and P. Mendels. Computations were performed on C98 and T3E at the Institut de Développement des Recherches en Informatique Scientifique of C.N.R.S. under contracts 984091-980076 and on the T3E at NIC Jülich.
no-problem/9907/nucl-th9907043.html	ar5iv	text	# Quark distributions in the nucleon based on a relativistic 3-body approach to the NJL model ## 1 Introduction Deep inelastic lepton-nucleon scattering experiments are providing detailed information on the quark and gluon distributions in the nucleon . By analyzing the distribution functions derived from the measured cross sections with the help of the $`Q^2`$ evolution based on perturbative QCD, one can extract the fraction of the nucleon’s momentum and spin carried by the quarks at some renormalization scale $`\mu `$, and obtain valuable information on the spin-flavor structure of the nucleon . For example, of great interest in this connection are the flavor dependences of the valence quark distributions and of the sea quark distributions , which reflect the roles of $`qq`$ (diquark) and $`\overline{q}q`$ (mesonic) correlations in the nucleon. Current experiments at HERA are exploring the region of small $`x`$ in order to get more information on large-distance (non-perturbative) phenomena, and future experiments at RHIC will concentrate on the spin distributions. An essential tool to analyze the data on the nucleon structure functions in the Bjorken limit are the factorization theorems , which allow us to separate the long-distance parts (parton distributions) from the short-distance parts (hard scattering cross sections). The latter ones are calculable from perturbative QCD, while the parton distributions require the knowledge of the nucleon wave function. The central point of the factorization theorems is that the infrared-divergent (long-distance) contributions to the perturbative diagrams can be absorbed into the definition of ’renormalized’ parton distributions . Due to this procedure, the parton distributions eventually depend on a factorization scale ($`\mu `$), which is usually taken to be the same as the renormalization scale. The requirement that the structure functions should not depend on this scale then leads to the famous DGLAP equations for the $`\mu `$ dependence of the distribution functions, which in turn also determines the $`Q^2`$ dependence of the structure functions. One can therefore use effective quark theories to calculate the distribution functions at some ’low energy scale’ $`\mu =Q_0`$, where a description in terms of quark degrees of freedom alone is expected to be valid, and then use the DGLAP equations to relate them to the distribution functions extracted from experimental data at $`\mu =Q`$. As an effective quark theory in the low energy region, the Nambu - Jona-Lasinio (NJL) model is a powerful tool to investigate the properties of hadrons . It exhibits the spontaneous breaking of chiral symmetry in a simple and clear way, and allows the solution of the relativistic two- and three-body equations in the ladder approximation due to the simplicity of the interaction . Concerning the ground state properties of baryons, the Faddeev approach to the NJL model has been quite successful , and since in this approach the translational invariance and covariance are preserved, it is natural to apply it also to the structure functions of the nucleon.<sup>2</sup><sup>2</sup>2There are already several investigations of the nucleon structure functions using the soliton approach to the NJL model . For a calculation using the Bethe-Salpeter equation for a quark and a structureless diquark, see ref. . Actually, the NJL model has been formulated recently on the light cone (LC) and the structure function of the pion has been calculated . For the case of the nucleon, however, there is the problem that the relativistic Faddeev equations are usually solved by performing a Wick rotation using the Euclidean sharp cut-off , and this method cannot be applied directly to the calculation of the LC momentum distributions. A possibility is first to go to the moment space and then re-construct the distribution functions by the inverse Mellin transformation. While such a calculation is now in progress , it is desirable for a first orientation to use some simple approximation which allows a direct calculation of the distributions. The ’static approximation’ to the Faddeev kernel , which amounts to neglect the momentum dependence of the propagator of the exchanged quark, allows an analytic solution of the Faddeev equation, and it has been shown in ref. that for the nucleon mass this approximation is not unreasonable. However, the overbinding of the nucleon compared to the exact Faddeev result and the ’point-like’ quark-diquark interaction lead to radii which are too small , and therefore we can expect that the resulting momentum distributions will be too stiff. However, our main purpose here is to see some general trends, to investigate the influence of the regularization scheme and to demonstrate that in the Faddeev approach the validity of the number and momentum sum rules is guaranteed from the outset, which is not the case in the soliton or bag model approaches. The calculation employing the static approximation to the Faddeev kernel resembles the quark-diquark calculations performed earlier . We will go beyond these calculations by including the structure of the diquark (and also of the pion when estimating the pionic cloud effects), and investigating the sensitivity to the regularization scheme while preserving the number and momentum sum rules. The rest of this paper is organized as follows: In sect. 2 we explain the model for the nucleon wave function, in sect. 3 we explain our method to calculate the quark distribution functions, and in sect. 4 we discuss the numerical results. A summary is presented in sect. 5. ## 2 NJL model for the nucleon wave function The NJL model is characterized by a chirally symmetric four-Fermi interaction lagrangian $`_I`$. By means of Fierz transformations, one can rewrite any $`_I`$ into a form where the interaction strength in a particular $`\overline{q}q`$ or $`qq`$ channel can be read off directly . That part which generates the constituent quark mass $`M_Q`$ and the pion as a collective $`\overline{q}q`$ bound state is given by $$_{I,\pi }=\frac{1}{2}g_\pi \left(\left(\overline{\psi }\psi \right)^2\left(\overline{\psi }(\gamma _5𝝉)\psi \right)^2\right),$$ (2.1) and that which describes the $`qq`$ interaction in the scalar diquark ($`J^\pi =0^+,T=0`$) channel is $$_{I,s}=g_s\left(\overline{\psi }\left(\gamma _5C\right)\tau _2\beta ^A\overline{\psi }^T\right)\left(\psi ^T\left(C^1\gamma _5\right)\tau _2\beta ^A\psi \right),$$ (2.2) where $`\beta ^A=\sqrt{3/2}\lambda ^A(A=2,5,7)`$ are the color $`\overline{3}`$ matrices, and $`C=i\gamma _2\gamma _0`$. The coupling constants $`g_\pi `$ and $`g_s`$ are related to the ones appearing in the original $`_I`$ by numerical factors due to the Fierz transformation, but instead of choosing a particular form of $`_I`$ we will treat $`g_\pi `$ and the ratio $`r_s=g_s/g_\pi `$ as free parameters, where the latter reflects different possible forms of $`_I`$ . The reduced t-matrices in the pionic and scalar diquark channels are obtained from the respective Bethe-Salpeter (BS) equations as $$\tau _\pi (k)=\frac{2ig_\pi }{1+2g_\pi \mathrm{\Pi }_\pi (k)},\tau _s(k)=\frac{4ig_s}{1+2g_s\mathrm{\Pi }_s(k)}$$ (2.3) with the bubble graph $$\mathrm{\Pi }_\pi (k)=\mathrm{\Pi }_s(k)=6i\frac{d^4q}{(2\pi )^4}tr_D\left[\gamma _5S(q)\gamma _5S(qk)\right],$$ (2.4) where $`S(q)={\displaystyle \frac{1}{\overline{)}qM_Q+iϵ}}`$ is the Feynman propagator and $`M_Q`$ the constituent quark mass. If the interacting two-body channels are restricted to the scalar diquark one, the relativistic Faddeev equation can be reduced to an effective BS equation for a composite scalar diquark and a quark, interacting via quark exchange . As we explained in the previous section, in this paper we will restrict ourselves to the static approximation , where the Feynman propagator in the quark exchange kernel is simply replaced by $`1/M_Q`$. Then the effective BS equation reduces to a geometric series of quark-diquark bubble graphs ($`\mathrm{\Pi }_N(p)`$), and the solution for the t-matrix in the color singlet channel is $$T(p)=\frac{3}{M_Q}\frac{1}{1+\frac{3}{M_Q}\mathrm{\Pi }_N(p)}$$ (2.5) with $$\mathrm{\Pi }_N(p)=\frac{d^4k}{(2\pi )^4}S(k)\tau _s(pk).$$ (2.6) The quark-diquark vertex function $`\mathrm{\Gamma }_N(p)`$ in the covariant normalization is then obtained from the pole behaviour $`T_s\mathrm{\Gamma }_N(ps)\overline{\mathrm{\Gamma }}_N(ps)/(p^2M_N^2+iϵ)`$, where $`M_N`$ is the nucleon mass, as <sup>3</sup><sup>3</sup>3Our conventions for LC variables are $`a^\pm =\frac{1}{\sqrt{2}}\left(a^0\pm a^3\right),a_\pm =\frac{1}{\sqrt{2}}\left(a_0\pm a_3\right),`$ and $`a_{}^i=a_i(i=1,2)`$. The Lorentz scalar product is $`ab=a_+b^++a_{}b^{}𝒂_{}𝒃_{}`$. We will frequently call $`p_{}(p_+)`$ the ’LC minus (plus) component’ of the four vector $`p`$. $`\mathrm{\Gamma }_N(p)`$ $`=`$ $`Z_Nu_N(p)`$ (2.7) $`Z_N`$ $`=`$ $`\left({\displaystyle \frac{1}{\mathrm{\Pi }_N(p)/\overline{)}p\|_{\overline{)}p=M_N}}}\right)^{\frac{1}{2}}=\left({\displaystyle \frac{p_{}/M_N}{\overline{u}_N(p)\frac{\mathrm{\Pi }_N(p)}{p_+}u_N(p)}}\right)^{\frac{1}{2}},`$ (2.8) where $`u_N(p)`$ is a free Dirac spinor with mass $`M_N`$ normalized by $`\overline{u}_N(p)u_N(p)=2M_N`$. In this normalization, the vertex function satisfies the relation $$\frac{1}{2p_{}}\overline{\mathrm{\Gamma }}_N(p)\frac{\mathrm{\Pi }_N(p)}{p_+}\mathrm{\Gamma }_N(p)=1,$$ (2.9) which leads to charge and baryon number conservation in any treatment which preserves the Ward identity $$\mathrm{\Lambda }_{q/P}^+(p,p)=\frac{\mathrm{\Pi }_N(p)}{p_+}N_{q/P}$$ (2.10) for the vertex of the quark number current of the proton at $`q=0`$ ($`N_{u/P}=2,N_{d/P}=1`$). <sup>4</sup><sup>4</sup>4This vertex is defined by $`p\|\overline{\psi }(0)\gamma ^+\frac{1\pm \tau _z}{2}\psi (0)\|p=\overline{\mathrm{\Gamma }}_N(p)\mathrm{\Lambda }_{q/P}^+(p,p)\mathrm{\Gamma }_N(p)`$ for $`q=u(d)`$. ## 3 Quark distribution functions The twist-2 quark LC momentum distribution in the proton (momentum $`p`$) is defined as $$\stackrel{~}{f}_{q/P}(x)=\frac{1}{2}\frac{dz^{}}{2\pi }e^{ip_{}xz^{}}p\|T\left(\overline{\psi }_q(0)\gamma ^+\psi _q(z^{})\right)\|p,$$ (3.1) where $`q`$ denotes the quark flavor, $`\|p`$ is the proton state, and $`x`$ is the Bjorken variable which corresponds to the fraction of the proton’s LC momentum component $`p_{}`$ carried by the quark $`q`$. As has been discussed in detail in ref. , for connected LC correlation functions the T-product is identical to the usual product, from which it follows that the distribution (3.1) is non-zero in the interval $`1<x<1`$. The physical quark and antiquark distributions which determine the structure functions $`F_1`$ and $`F_2`$ are obtained for $`0<x<1`$ as $`f_{q/P}(x)=\stackrel{~}{f}_{q/P}(x)`$ and $`f_{\overline{q}/P}(x)=\stackrel{~}{f}_{q/P}(x)`$. The valence ($`v`$) and sea ($`s`$) quark distributions are then given by $`f_{q_v/P}(x)=f_{q/P}(x)f_{\overline{q}/P}(x)`$, $`f_{q_s/P}(x)=f_{\overline{q}_s/P}(x)=f_{\overline{q}/P}(x)`$. The evaluation of the distribution (3.1) can be reduced to a straight forward Feynman diagram calculation by noting that it can be expressed as <sup>5</sup><sup>5</sup>5The original Lorentz invariant expression is recovered by $`k_{}/p_{}kq/pq`$ and $`\gamma ^+/p_{}\overline{)}q/pq`$. The expression (3.2) corresponds to the Bjorken limit in the frame where $`𝒒_{}=0,q_z<0`$, i.e; $`q_+\mathrm{},q_{}p_{}x`$. $$\stackrel{~}{f}_{q/P}(x)=\frac{i}{2p_{}}\frac{d^4k}{(2\pi )^4}\delta (x\frac{k_{}}{p_{}})Tr_D\gamma ^+M_q(p,k),$$ (3.2) with the quark 2-point function in the proton given by $$M_{q,\beta \alpha }(p,k)=id^4ze^{ikz}<p\|T\left(\overline{\psi }_{q,\alpha }(0)\psi _{q,\beta }(z)\right)\|p>.$$ (3.3) We therefore have to evaluate the Feynman diagrams for the quark propagator in the nucleon, trace it with $`\gamma ^+`$, fix the LC minus component of the quark momentum as $`k_{}=p_{}x`$ and integrate over the remaining components $`k_+`$ and $`𝒌_{}`$. Since in our model for the nucleon discussed in the previous section the quark can either appear as a spectator or as a constituent of the scalar diquark, the Feynman diagrams to be evaluated are shown in fig.1. (In the full Faddeev approach, there is also a diagram where the external operator acts on the exchanged quark, but in the present static approximation this diagram does not contribute.) To present the formulae for the diagrams fig.1, we note that the second diagram (the diquark contribution) can be expressed conveniently as a convolution integral if we insert the identity $$1=𝑑y𝑑z𝑑q_0^2\delta (y\frac{q_{}}{p_{}})\delta (z\frac{k_{}}{q_{}})\delta (q^2q_0^2),$$ (3.4) i.e; $`y`$ is the fraction of the nucleon’s momentum component $`p_{}`$ carried by the diquark, $`z`$ is the fraction of the diquark’s momentum component $`q_{}`$ carried by the quark inside the diquark ($`x=yz`$), and $`q_0^2`$ is the virtuality of the diquark. Using also the identity $$S(k)\gamma ^+S(k)=\frac{S(k)}{k_+},$$ (3.5) and performing partial integrations in the plus components of the loop momenta, which is permissible since these integrations are convergent and not restricted by the regularization schemes to be discussed later, we obtain the following expression<sup>6</sup><sup>6</sup>6To distinguish the case without the pion cloud (valence quark picture) from that including the pion cloud, we replace $`qQ`$ in the formulae corresponding to the diagrams of fig.1.: $$f_{Q/P}(x)=\delta _{Q,U}F_{Q/P}(x)+\frac{1}{2}F_{Q(D)/P}(x).$$ (3.6) Here the first term corresponds to the first diagram in fig.1 and is expressed as $$F_{Q/P}(x)=\frac{1}{2p_{}}\overline{\mathrm{\Gamma }}_N\left(\frac{}{p_+}\mathrm{\Pi }_N(x,p)\right)\mathrm{\Gamma }_N,$$ (3.7) where $`\mathrm{\Pi }_N(x,p)`$ is the quark-diquark bubble graph for fixed minus component of the quark momentum: $$\mathrm{\Pi }_N(x,p)=\frac{d^4k}{(2\pi )^4}\delta (x\frac{k_{}}{p_{}})S(k)\tau _s(pk).$$ (3.8) The second term in (3.6) corresponding to the second diagram in fig.1 is purely isoscalar and given by the convolution integral $$F_{Q(D)/P}(x)=_0^1𝑑y_0^1𝑑z\delta (xyz)_{\mathrm{}}^{\mathrm{}}𝑑q_0^2F_{Q/D}(z,q_0^2)F_{D/P}(y,q_0^2),$$ (3.9) where the distributions $`F_{Q/D}(z,q_0^2)`$ and $`F_{D/P}(y,q_0^2)`$ for fixed virtuality of the diquark (D) are expressed as $`F_{Q/D}(z,q_0^2)`$ $`=`$ $`2g^2(q_0^2){\displaystyle \frac{\mathrm{\Pi }_s(z,q_0^2)}{q_0^2}}`$ (3.10) $`F_{D/P}(y,q_0^2)`$ $`=`$ $`\overline{\mathrm{\Gamma }}_N\left({\displaystyle \frac{1}{2p_{}}}{\displaystyle \frac{}{p_+}}+y{\displaystyle \frac{}{q_0^2}}\right)\mathrm{\Pi }_N(y,q_0^2;p)\mathrm{\Gamma }_N.`$ (3.11) Here $`\mathrm{\Pi }_s(z,q_0^2)`$ is the quark-quark bubble graph for fixed minus momentum component of the quark, $`g^2(q_0^2)=1/\left({\displaystyle \frac{\mathrm{\Pi }_s(q_0^2)}{q_0^2}}\right)`$ is the quark-diquark coupling constant, and $`\mathrm{\Pi }_N(y,q_0^2;p)`$ is the quark-diquark bubble graph for fixed virtuality and minus momentum component of the diquark: $`\mathrm{\Pi }_s(z,q_0^2)`$ $`=`$ $`\left[6i{\displaystyle \frac{d^4k}{(2\pi )^4}\delta (z\frac{k_{}}{q_{}})tr_D\left(\gamma _5S(k)\gamma _5S(kq)\right)}\right]_{q^2=q_0^2}`$ (3.12) $`\mathrm{\Pi }_N(y,q_0^2;p)`$ $`=`$ $`{\displaystyle \frac{d^4q}{(2\pi )^4}\delta (y\frac{q_{}}{p_{}})\delta (q^2q_0^2)S(pq)\tau _s(q)}.`$ (3.13) If we use the dispersion representation for the diquark t-matrix $`\tau _s`$, we can perform the $`k_+`$ and $`q_+`$ integrals analytically and verify that the distribution (3.6) is non-zero only in the interval $`0<x<1`$. This fact was anticipated already in our notation ($`\stackrel{~}{f}_{Q/P}(x)=f_{Q/P}(x)`$) and in the integration limits in eq. (3.9), and corresponds to a valence quark model ($`f_{\overline{Q}/P}(x)=0`$). Using the above expressions and the normalization (2.9), it is a simple matter to confirm the validity of the number and momentum sum rules <sup>7</sup><sup>7</sup>7To verify these relations, it is only necessary to note that the relations $`{\displaystyle _0^1}𝑑zF_{Q/D}(z,q_0^2)=2`$ and $`{\displaystyle _0^1}𝑑zzF_{Q/D}(z,q_0^2)=1`$ hold for any $`q_0^2`$ (here $`F_{Q/D}(z,q_0^2)`$ is symmetric around $`z=1/2`$), and therefore the second term in eq. (3.11) gives a vanishing surface term when integrated over $`q_0^2`$. The sum rules are then obvious since the integral of $`\mathrm{\Pi }_N(y,q_0^2,p)`$ over $`q_0^2`$ reduces to $`\mathrm{\Pi }_N(1y,p)`$, and that of $`\mathrm{\Pi }_N(x,p)`$ over $`x`$ to $`\mathrm{\Pi }_N(p)`$. $`{\displaystyle _0^1}𝑑xf_{Q/P}(x)`$ $`=`$ $`N_{Q/P}`$ (3.14) $`{\displaystyle _0^1}𝑑xx\left(f_{U/P}(x)+f_{D/P}(x)\right)`$ $`=`$ $`1.`$ (3.15) In more general terms, what we have really confirmed here is the validity of the Ward identities for quark number and momentum conservation: Concerning the number conservation, if we integrate the distribution function of fig.1 over $`x`$, the restriction $`k_{}=p_{}x`$ is lifted, and the diagrams correspond to $`\mathrm{\Lambda }_{Q/P}^+(p,p)/2p_{}`$. The validity of the Ward identity (2.10) follows then from (3.5) and partial integrations in the plus components of the loop momenta. A similar argument holds for the Ward identity expressing momentum conservation. Therefore, the Ward identities and the sum rules (3.14), (3.15) hold in any regularization scheme which does not restrict the LC plus components of the loop momenta. The regularization schemes to be discussed at the end of this section satisfy this requirement. As we have noted above, the model described so far gives essentially only valence-like distributions at the low energy scale. Although sea quark distributions will be generated in the process of the $`Q^2`$ evolution, those will be flavor independent ($`f_{\overline{u}/P}=f_{\overline{d}/P}`$), which contradicts the experimentally measured violation of the Gottfried sum rule . Also, it is a general trend of valence quark models that the resulting valence quark distributions are too stiff (too strongly pronounced peak and too small support at low values of $`x`$). We therefore consider here the effects of the pion dressing of the constituent quarks, as has been done also in previous works . (For a recent investigation using the LC quantization, see ref. .) In order to take into account the pion cloud, in principle we should solve the Schwinger-Dyson equation for the quark Feynman propagator $`S(q)={\displaystyle \frac{1}{\overline{)}qM_Q\mathrm{\Sigma }_Q(p)}}`$, where $$\mathrm{\Sigma }_Q(p)=3\frac{d^4k}{(2\pi )^4}\left(\gamma _5S(k)\gamma _5\right)\stackrel{~}{\tau }_\pi (pk)\mathrm{\Pi }_Q(p)$$ (3.16) is the quark self energy due to the pion cloud, and the reduced pion t-matrix $`\stackrel{~}{\tau }_\pi \tau _\pi +2ig_\pi `$ depends also on $`S(p)`$. This propagator should then be used to calculate the diquark t-matrix $`\tau _s`$ and the nucleon vertex function $`\mathrm{\Gamma }_N`$. Using these modified propagators and vertex functions, one should, in addition to the diagrams of fig.1, also evaluate the diagrams of fig.2, where the operator insertion is made on the quark while the pion is ’in flight’, or on the quark and the antiquark in the pion. Clearly, such a calculation is very complicated, and the usually employed convolution formalism involves the following two major approximations: First, pionic effects can be renormalized into a re-definition of the constituent quark mass and the four fermi coupling constants if one approximates the quark propagator by its pole part <sup>8</sup><sup>8</sup>8Green functions and vertex functions which differ from those without pionic effects by the replacements $`M_Q\widehat{M}_Q`$, $`G_\alpha \widehat{G}_\alpha G_\alpha Z_Q^2(\alpha =\pi ,s)`$ will be denoted by a hat. Since due to the discussion following eq.(3.18) these renormalized quantities are numerically equivalent to the ones used previously, this distinction will eventually be dropped. $$S(p)=\frac{Z_Q}{\overline{)}p\widehat{M}_Q+iϵ}Z_Q\widehat{S}(p)$$ (3.17) with $$Z_Q=\left(1+\frac{\mathrm{\Pi }_Q}{\overline{)}k}\|_{\overline{)}k=\widehat{M}_Q}\right)^1.$$ (3.18) If we define ’renormalized’ coupling constants $`\widehat{G}_\alpha `$ by $`G_\alpha =\widehat{G}_\alpha /Z_Q^2(\alpha =\pi ,s)`$, it is easy to see that the Green functions and vertex functions are renormalized according to $`\tau _\alpha =\widehat{\tau }_\alpha /Z_Q^2,T=Z_Q\widehat{T},\mathrm{\Gamma }_N=\sqrt{Z_Q}\widehat{\mathrm{\Gamma }}_N`$. If we then impose the same conditions on the parameters as commonly used in the case without pion cloud effects (that is, $`f_\pi =93MeV`$, $`m_\pi =140MeV`$, $`\widehat{M}_Q=300500MeV`$, and $`M_N=940MeV`$), the cut-off and the coupling constants $`\widehat{G}_\alpha `$ take the same values as in the case without pionic cloud effects. (In terms of the lagrangian, such a renormalization procedure corresponds to writing $`G\left(\overline{\psi }\psi \right)^2=\widehat{G}\left(\overline{\widehat{\psi }}\widehat{\psi }\right)^2`$ with $`\psi =\sqrt{Z_Q}\widehat{\psi }`$ and $`G=\widehat{G}/Z_Q^2`$.) It is also easy to check that all diagrams in fig.1 and fig.2 get a factor $`Z_Q`$. If one then calculates the diagrams of fig.2 in terms of these renormalized quantities and writes the results in terms of a convolution integral, one finds that due to the Dirac structure of the insertions on the ’parent’ quark line there appear three convolutions terms . Only one of them involves the generalization of the ’bare’ quark distribution in the nucleon (eq.(3.6)) to the off shell case ($`f_{Q/P}(x)f_{Q/P}(x,k_0^2)`$, where $`k_0^2`$ is the virtuality of the parent quark), convoluted with the quark distribution within the parent quark. Each of the other two terms involve one additional factor of $`(k_0^2M_Q^2)`$ in the integrand compared to the first term, which has a sharp peak at $`k_0^2=M_Q^2`$. The second approximation commonly used is therefore to neglect these two terms, and to assume that due to the sharp peak of $`f_{Q/P}(x,k_0^2)`$ the quark distribution within the parent quark can be evaluated at $`k_0^2=M_Q^2`$ and taken outside of the $`k_0^2`$ integral . In this way one arrives at the familiar convolution form $$f_{q/P}(x)=\underset{Q=U,D}{}_0^1𝑑y_0^1𝑑z\delta (xyz)f_{q/Q}(z)f_{Q/P}(y)$$ (3.19) and a similar expression with $`q\overline{q}`$, where the parent quark distribution in the proton $`f_{Q/P}`$ is given by eq.(3.6), and $`f_{q/Q}`$ ($`f_{\overline{q}/Q}`$) is the quark (antiquark) distribution within an on-shell parent quark which is obtained by evaluating the Feynman diagrams shown in fig.3. The quark and antiquark distributions in the parent quark obtained from the diagrams of fig.3 can be expressed as $`f_{u/U}(x)`$ $`=`$ $`Z_Q\delta (x1)+{\displaystyle \frac{1}{3}}F_{q/Q}(x)+{\displaystyle \frac{5}{6}}F_{q(\pi )/Q}(x)`$ (3.20) $`f_{d/U}(x)`$ $`=`$ $`{\displaystyle \frac{2}{3}}F_{q/Q}(x)+{\displaystyle \frac{1}{6}}F_{q(\pi )/Q}(x)`$ (3.21) $`f_{\overline{u}/U}(x)`$ $`=`$ $`{\displaystyle \frac{1}{6}}F_{q(\pi )/Q}(x)`$ (3.22) $`f_{\overline{d}/U}(x)`$ $`=`$ $`{\displaystyle \frac{5}{6}}F_{q(\pi )/Q}(x).`$ (3.23) The distributions for $`Q=D`$ are also determined from these expressions due to isospin symmetry. The detailed formulae for the distributions $`F_{q/Q}(x)`$ and $$F_{q(\pi )/Q}(x)=_0^1𝑑y_0^1𝑑z\delta (xyz)_{\mathrm{}}^{\mathrm{}}𝑑q_0^2F_{q/\pi }(z,q_0^2)F_{\pi /Q}(y,q_0^2),$$ (3.24) corresponding to the second and third diagrams of fig.3, respectively, can be obtained from the previous expressions (3.7)-(3.13) as follows: $`F_{q/Q}(x),F_{q/\pi }`$ and $`F_{\pi /Q}`$ are given by expressions similar to eqs.(3.7), (3.10) and (3.11), respectively, but with the following replacements: (i) The nucleon spinor $`\mathrm{\Gamma }_N`$ is replaced by the quark spinor defined via the residue of the propagator $`S(p)`$: $`\mathrm{\Gamma }_Q(p)=Z_Qu_Q(p)`$, where $`u_Q(p)`$ is a free Dirac spinor with mass $`M_Q`$ normalized by $`\overline{u}_Q(p)u_Q(p)=2M_Q`$. From eq.(3.18) it follows that this spinor satisfies the relation $$\frac{1}{2p_{}}\overline{\mathrm{\Gamma }}_Q(p)\frac{\mathrm{\Pi }_Q(p)}{p_+}\mathrm{\Gamma }_Q(p)=1Z_Q.$$ (3.25) (ii) the polarizations $`\mathrm{\Pi }_N(x,p)`$ and $`\mathrm{\Pi }_N(y,q_0^2;p)`$ are replaced by $`\mathrm{\Pi }_Q(x,p)`$ and $`\mathrm{\Pi }_Q(y,q_0^2;p)`$. These are defined analogously to eqs. (3.8) and (3.13) by introducing the $`\delta `$ function insertions to fix the minus momentum components of the quark or the pion and the virtuality $`q_0^2`$ of the pion into $`\mathrm{\Pi }_Q`$ defined by eq. (3.16) instead of $`\mathrm{\Pi }_N`$, and (iii) $`F_{q/\pi }`$ is given by the r.h.s. of eq. (3.10), but without the overall factor 2. We note that $`f_{q/Q}`$, and therefore also the distribution $`f_{q/P}`$ of eq. (3.19), involves an overall factor $`Z_Q`$, in accordance with our discussion following eq. (3.18). ¿From these expressions and the normalization (3.25), the validity of the number and momentum sum rules $`{\displaystyle _0^1}𝑑x\left(f_{q/Q}(x)f_{\overline{q}/Q}(x)\right)=\delta _{Q,q}`$ (3.27) $`{\displaystyle _0^1}𝑑xx\left(f_{u/Q}(x)+f_{\overline{u}/Q}(x)+f_{d/Q}(x)+f_{\overline{d}/Q}(x)\right)=1`$ can be easily checked in the same way as outlined above for the parent quark distributions in the proton. (The number sum rule is a consequence of the Ward identity for the quark number current of the parent quark $`\mathrm{\Lambda }_{q/Q}^+(p,p)={\displaystyle \frac{\mathrm{\Pi }_Q(p)}{p_+}}\delta _{q,Q}`$.) The validity of the number and momentum sum rules $`{\displaystyle _0^1}𝑑xf_{q_v/P}(x){\displaystyle _0^1}𝑑x\left(f_{q/P}(x)f_{\overline{q}/P}(x)\right)`$ $`=`$ $`N_{q/P}`$ (3.28) $`{\displaystyle _0^1}𝑑xx\left(f_{u/P}(x)+f_{\overline{u}/P}(x)+f_{d/P}(x)+f_{\overline{d}/P}(x)\right)`$ $`=`$ $`1`$ (3.29) is then a consequence of eqs. (3.14), (3.15), (3.27), refbmom2 and (3.19). Of particular interest is also the Gottfried sum $`S_G`$ $`=`$ $`{\displaystyle \frac{1}{3}}{\displaystyle _0^1}𝑑x\left(f_{u/P}(x)+f_{\overline{u}/P}(x)f_{d/P}(x)f_{\overline{d}/P}(x)\right)`$ (3.30) $`=`$ $`{\displaystyle \frac{1}{3}}{\displaystyle \frac{4}{9}}{\displaystyle _0^1}𝑑xF_{q/Q}(x)={\displaystyle \frac{1}{3}}{\displaystyle \frac{4}{9}}\left(1Z_Q\right),`$ which shows that the deviation from the valence quark model result ($`S_G=\frac{1}{3}`$) is due to the decrease of the probability of the ’bare’ valence quark state ($`Z_Q<1`$) . We now discuss our regularization scheme. Since the above expressions for the quark distributions involve loop integrals with one of the LC momentum components fixed, it is clear that we need a regularization scheme which can be formulated in terms of LC momenta. Two such schemes which have been discussed extensively in ref. are the Lepage-Brodsky (LB) or invariant mass scheme , and the transverse cut-off (TR) scheme . The basic graphs which are regularized in both schemes are the $`qq`$ and $`\overline{q}q`$ bubble graphs $`\mathrm{\Pi }_s=\mathrm{\Pi }_\pi `$, the quark-diquark bubble graph $`\mathrm{\Pi }_N`$ and the quark self energy $`\mathrm{\Pi }_Q`$, either for the case that all internal momentum components are integrated out or one of the LC momentum components is fixed. Concerning the LB scheme, it has been shown in detail for the case of $`\mathrm{\Pi }_s`$ in ref. that, if all momentum components are integrated out, this scheme is equivalent to the covariant 3-momentum (or dispersion) cut-off scheme. <sup>9</sup><sup>9</sup>9By ’covariant 3-momentum cut-off scheme’ we mean the procedure where the 3-momentum cut-off is introduced in the particular Lorentz frame where the total momentum of the two-body ($`qq`$, $`\overline{q}q`$, or quark-diquark) state is zero ($`𝒑=0`$), and the result is ’boosted’ to a general frame. For the graphs $`\mathrm{\Pi }_\alpha (\alpha =s,\pi ,N,Q)`$ considered here, this ’boosting’ simply means the replacement $`p_0^2p^2`$, and for $`\mathrm{\Pi }_N`$ and $`\mathrm{\Pi }_Q`$ also $`p_0\gamma ^0\overline{)}p`$. It is known that this procedure is equivalent to the dispersion cut-off scheme . Generally, if the intermediate state involves particles with masses $`m_1`$ and $`m_2`$, the LB cut-off applied in the frame where the transverse components of the total momentum are zero ($`𝐩_{}=0`$) restricts the invariant mass of the state according to $$\frac{𝐤_{}^2+m_1^2}{x}+\frac{𝐤_{}^2+m_2^2}{1x}<\mathrm{\Lambda }_{LB}^2,$$ (3.31) where $`x`$ and $`1x`$ are the fractions of the total momentum component $`p_{}`$ carried by the two particles. The LB regulator $`\mathrm{\Lambda }_{LB}`$ is related to the 3-momentum cut-off $`\mathrm{\Lambda }_3`$ by $`\mathrm{\Lambda }_{LB}=\left(\sqrt{m_1^2+\mathrm{\Lambda }_3^2}+\sqrt{m_2^2+\mathrm{\Lambda }_3^2}\right)`$. For the case of $`\mathrm{\Pi }_s=\mathrm{\Pi }_\pi `$ we have $`m_1=m_2=M_Q`$, and the value of $`\mathrm{\Lambda }_3`$ is determined as usual by requiring that $`f_\pi =93MeV`$. In the case of the graphs $`\mathrm{\Pi }_N`$ ($`\mathrm{\Pi }_Q`$) we have $`m_1=M_Q`$, while $`m_2`$ is the mass parameter in the dispersion representation of $`\tau _s`$ ($`\tau _\pi `$). In order not to increase the number of parameters, we will take the same value of $`\mathrm{\Lambda }_3`$ for all graphs $`\mathrm{\Pi }_s=\mathrm{\Pi }_\pi `$, $`\mathrm{\Pi }_N`$ and $`\mathrm{\Pi }_Q`$. Concerning the TR cut-off scheme, it has been discussed in ref. that the use of this scheme requires a mass renormalization procedure, since the basic self energies $`\mathrm{\Pi }_\alpha (\alpha =s,\pi ,N,Q)`$ involve also logarithmically divergent longitudinal momentum $`(k_{})`$ integrals, which are not affected by the TR regularization prescription $`\|𝒌_{}\|<\mathrm{\Lambda }_{TR}`$. In this scheme one has therefore to impose the pion, the diquark and the nucleon masses as renormalization conditions rather than to relate them to the parameters $`g_\pi `$, $`g_s`$ and $`r_s`$ via the pole conditions. For example, if we impose the condition $`1+2g_s\mathrm{\Pi }_s(M_D^2)=0`$ for some fixed $`M_D`$, the t-matrix $`\tau _s`$ in eq. (2.3) can be re-written in the renormalized form $`\tau _s=2i/(\mathrm{\Pi }_s(k^2)\mathrm{\Pi }_s(M_D^2))`$, which is formally independent of $`g_s`$ and free of divergencies due to the longitudinal momentum integration. In the calculation using the TR cut-off we will impose the same value of $`M_D`$ as obtained in the calculation using the LB cut-off. ## 4 Numerical results In both the LB and TR regularization schemes we use $`M_Q=400MeV`$ for the constituent quark mass, and determine the cut-off so as to reproduce $`f_\pi =93MeV`$. This gives $`\mathrm{\Lambda }_3=593MeV`$ for the equivalent 3-momentum cut-off in the LB scheme, and $`\mathrm{\Lambda }_{TR}=407MeV`$ in the TR scheme. In the LB scheme, we then obtain $`g_\pi =6.92GeV^2`$ and $`r_s=g_s/g_\pi =0.727`$ from the requirements $`m_\pi =140MeV`$ and $`M_N=940MeV`$, respectively, and the resulting scalar diquark mass becomes $`M_D=600MeV`$. <sup>10</sup><sup>10</sup>10The current quark mass obtained from the gap equation is $`m=5.96MeV`$ in the LB scheme. As we explained earlier, in the TR scheme we use the same value $`M_D=600MeV`$, and rewrite the t-matrices $`\tau _\pi `$ and $`\tau _s`$, which are needed to calculate the distribution functions, in terms of $`m_\pi `$ and $`M_D`$ such that they become independent of $`g_\pi `$ and $`r_s`$. Our results for the valence and sea quark distributions are shown in figs. 4-7 both for the LB and the TR cut-off scheme. As we have explained in sect.1, in order to make contact to the empirical distributions extracted from the measured structure functions, we have to evolve our calculated distributions from the low energy scale $`\mu ^2=Q_0^2`$ to the value $`\mu ^2=Q^2`$ where empirical parametrizations are available. For this $`Q^2`$ evolution we use the computer code of ref. to solve the DGLAP equation in the next-to-leading order. (For the $`Q^2`$ evolution we use $`N_f=3,\mathrm{\Lambda }_{QCD}=250MeV.`$) We will compare our evolved distributions to the parametrizations of ref. for $`Q^2=4GeV^2`$. Both the calculated and the empirical distribution functions refer to the $`\overline{MS}`$ renormalization and factorization scheme. The value of $`Q_0^2`$ is treated as a free parameter which is determined so as to reproduce the overall features of the empirical valence quark distributions at $`Q^2=4GeV^2`$. In this way we obtain a value of $`Q_0^2=0.16GeV^2`$, i.e; $`Q_0`$ is equal to our constituent quark mass $`M_Q`$. Let us first discuss the valence quark distributions shown in figs. 4 and 5. Although we do not show the results of the pure valence quark model (no pions), we note that the input distributions at $`\mu ^2=Q_0^2`$ shown here are softer than in the case without pionic cloud effects, that is, the pionic effects reduce the peak heights of the valence quark distributions and increase their support at low $`x`$. The integral over the input distributions shows that at $`\mu ^2=Q_0^2`$ the valence quarks carry $`92\%(87\%)`$ of the nucleon’s LC momentum for the case of the LB (TR) cut-off. The rest is carried by the sea quarks. This reduction of the peak heights due to pionic effects has a beneficial effect on the overall behaviour of the valence quark distributions, although it is insufficient in particular for the $`d`$ quark in the LB scheme. The input distributions are still rather stiff even when pionic effects are taken into account, which necessitates the use of a low value of $`Q_0^2`$ in order to approach the empirical distributions via the $`Q^2`$ evolution. We can expect some improvement concerning this point in a full Faddeev calculation, since in the present static approximation the size of the nucleon is too small, corresponding to momentum distributions which are too stiff. In the LB cut-off scheme, the input distributions are zero for large (and also very small) values of $`x`$, and therefore the output distributions (at $`Q^2=4GeV^2`$) show a too strong variation with $`x`$ compared to the empirical ones. On the contrary, for the TR cut-off the input distributions are non-zero in the whole region of $`x`$, which leads to a smoother behaviour of the output distributions. <sup>11</sup><sup>11</sup>11For the TR cut-off, the input distributions show a sharp increase when $`x`$ becomes very close to $`1`$. Since the computer code used for the $`Q^2`$ evolution requires an input distribution which vanishes for $`x=1`$, we artificially modified it for $`x`$ very close to $`1`$ such that it goes like $`(1x)^n`$ with some power $`n`$. ($`n=10`$ was used in the actual calculation.) This is the same feature as noted in ref. for the quark distribution in the pion, and indicates that for phenomenological applications the TR cut-off seems to be superior over the LB cut-off. On the other hand, as we have explained earlier, the shortcoming of the TR cut-off scheme is that the diquark mass must be treated as a free parameter since in this scheme mass renormalizations are necessary in order to get finite results. In our calculation, the difference between the valence $`u`$ and $`d`$ quark distributions reflects the scalar diquark correlations in the proton: Since the $`d`$ quark appears inside the diquark and not as a spectator quark (see eq.(3.6)), its distribution is given by the convolution of two distributions (eq.(3.9)), which is more concentrated at low values of $`x`$ compared to the spectator quark distribution. This is in agreement with the behaviour shown by the empirical distributions, and this observation was in fact one of the motivations to introduce diquark degrees of freedom also into the bag model description of the nucleon structure functions . We now turn to the antiquark distributions shown in figs. 6 and 7. As in the case of the valence quark distributions, the TR cut-off scheme leads to an overall better agreement with the empirical distributions than the LB scheme. The enhancement of $`\overline{d}`$ over $`\overline{u}`$ is clearly seen both in the input and the output distributions. Since the numerical value of the probability of the quark state without pion cloud is $`Z_Q=0.84`$ ($`Z_Q=0.83`$) for the LB (TR) cut-off scheme, the Gottfried sum (3.30) becomes $`S_G=0.262`$ ($`S_G=0.257`$), compared to the experimental value <sup>12</sup><sup>12</sup>12It has been shown that the Gottfried sum is almost unchanged by the $`Q^2`$ evolution . reported by the NM collaboration $`S_G=0.235\pm 0.026`$. Our results for the difference $`f_{\overline{d}/P}f_{\overline{u}/P}`$ are shown in fig. 8. We see that with our value of $`Q_0^2`$ which has been chosen such as to reproduce the overall behaviour of the valence quark distributions, the calculated difference is smaller than the empirical one for intermediate values of $`x`$, but larger for small $`x`$. Concerning the ratio $`\overline{u}/\overline{d}`$, there are also data from Drell- Yan processes which give $`\overline{u}/\overline{d}=0.51\pm 0.04\pm 0.05`$ at $`x=0.18`$, compared to our calculated value of $`0.68(0.70)`$ at $`Q^2=4GeV^2`$ for the LB (TR) cut-off. This, too, shows that the observed flavor asymmetry of the Dirac sea is larger than our calculated one in this range of $`x`$. ## 5 Summary and outlook In this paper we used the framework of the relativistic Faddeev equation in the NJL model to calculate the quark LC momentum distributions in the nucleon. As a first step towards a full Faddeev calculation, we used the nucleon vertex functions obtained in the simple static approximation to the Faddeev kernel, and included pionic cloud effect approximately using the familiar convolution formalism. We can summarize our results as follows: First, we have shown a method based on a straight forward Feynman diagram evaluation, which we believe to be best suited for the calculation of the distribution functions in the Faddeev framework. Besides being simple and straight forward, the method also guarantees the validity of the number and momentum sum rules from the outset. Second, we discussed two regularization schemes which can be formulated in terms of LC coordinates, and which preserve the number and momentum sum rules. Based on our numerical results we concluded that for the description of the momentum distributions in the nucleon the transverse momentum cut-off scheme is superior over the invariant mass regularization scheme, which is similar to the situation found previously for the momentum distribution in the pion. Third, we have shown that the resulting distribution functions reproduce the overall behaviours of the empirical ones if the low energy scale for the $`Q^2`$ evolution is taken to be about the same as the constituent quark mass ($`400MeV`$ in our calculation). Such a low value is required since the input valence quark distributions calculated in our model are rather stiff, although the pionic cloud effects served to soften them compared to the pure valence quark results. In this respect, a full Faddeev calculation which gives larger and more realistic nucleon radii, as well as the inclusion of higher mass diquark channels (axial vector diquark channel) are expected to improve the situation. We have also shown that the value for the Gottfried sum obtained in this simple calculation is in basic agreement with the experimental one. The formulation and results of this work can be used as a basis for at least the following three extensions: First, one could easily use the present framework to calculate the quark spin distributions, provided that the axial vector diquark $`(J=1)`$ channel is also taken into account. Second, one should use the full Faddeev vertex functions to calculate the quark momentum distributions. As we have noted in sect.1, the most convenient way for this purpose might be first to go to the moment space and then to reconstruct the distribution functions. The third extension concerns the case of finite density: For a finite density calculation the full Faddeev framework seems to be intractable, and approximations like the static approximation used in this paper might be unavoidable. For this purpose, however, it is necessary first to construct an equation of state for nuclear matter based on the Faddeev (quark-diquark) picture of the single nucleon, similar to the Guichon equation of state which is based on the MIT bag picture of the single nucleon. The construction of such an equation of state and its applications are now under consideration . Acknowledgements The authors would like to thank M. Miyama and S. Kumano for the computer program used for the $`Q^2`$ evolution (ref. ). One of the authors (W.B.) is grateful to A. W. Thomas, A. W. Schreiber, K. Suzuki and K. Tanaka for discussions on the nucleon structure function. ## Figure captions 1. Graphical representation of the quark LC momentum distribution in the Faddeev framework. The single (double) line denotes the quark propagator (scalar diquark t-matrix), the hatched circle stands for the quark- diquark vertex function, and the operator insertion denoted by a cross stands for $`\gamma ^+\delta (k_{}p_{}x)(1\pm \tau _z)/2`$ for the $`U(D)`$ quark distribution. The second diagram stands symbolically for those 2 diagrams obtained by inserting the cross into both particle lines in the diquark. The quark-diquark vertex contains the isospin operator $`\tau _2`$. The diagram where the operator insertion is made on the exchanged quark is not shown here since it does not contribute in the static approximation. 2. Feynman diagrams which have to be evaluated in addition to those shown in fig.1 due to the presence of the pion cloud. Here the dashed line indicates the $`\overline{q}q`$ t-matrix in the pionic channel, and the other lines are as in fig.1. The second diagram stands symbolically for those 2 diagrams obtained by inserting the cross into both the $`q`$ and $`\overline{q}`$ lines in the pion. The dots indicate all remaining diagrams where the pion line is attached to a quark in the diquark. 3. Graphical representation of the quark distribution within an on-shell parent quark. The meaning of the lines is as in figs.1 and 2. The quark spinor $`\mathrm{\Gamma }_Q(p)`$ is associated with the incoming and outgoing quark lines in all diagrams. The operator insertion in the first diagram stands for $`\gamma ^+\delta (p_{}p_{}x)(1\pm \tau _z)/2`$, and in the other diagrams for $`\gamma ^+\delta (k_{}p_{}x)(1\pm \tau _z)/2`$ for the u(d) quark distributions. The third diagram stands symbolically for those 2 diagrams obtained by inserting the cross into both the $`q`$ and $`\overline{q}`$ lines in the pion. 4. LC momentum distributions of the valence $`u`$-quark in the proton, using the LB (solid lines) and TR (dashed lines) regularization scheme. The lines associated with $`Q_0^2=0.16GeV^2`$ show the NJL model results, and the lines associated with $`Q_0^2=0.16GeV^2`$ show the results obtained by the QCD evolution in next-to-leading order from $`Q_0^2=0.16GeV^2`$ to $`Q^2=4GeV^2`$, using $`\mathrm{\Lambda }_{QCD}=0.25GeV`$ and $`N_f=3`$. The dotted line shows the parametrization for $`Q^2=4GeV^2`$ obtained in ref. by analyzing the experimental data. 5. LC momentum distributions of the valence $`d`$-quark in the proton. For explanation of the lines, see the caption to fig. 4. 6. LC momentum distributions of the $`\overline{u}`$-quark in the proton. For explanation of the lines, see the caption to fig. 4. 7. LC momentum distributions of the $`\overline{d}`$-quark in the proton. For explanation of the lines, see the caption to fig. 4. 8. The difference of $`\overline{d}`$ and $`\overline{u}`$-quark momentum distributions in the proton. For explanation of the lines, see the caption to fig. 4.
no-problem/9907/cond-mat9907119.html	ar5iv	text	# Quantum dot to disordered wire crossover: A complete solution in all length scales for systems with unitary symmetry Transport in phase coherent devices has become a rather topical subject, ever since modern nanolithographic techniques have achieved sufficient precision to manufacture a great variety of such systems . These mesoscopic structures can be broadly divided into two categories: (I) disordered conductors, where static defects generates disorder scattering with an elastic mean free path, $`l`$, smaller then the device dimensions and (II) ballistic cavities, where $`l`$ exceeds the device dimensions and boundary scattering dominates. The hallmark of mesoscopic phenomena is the relevance of quantum interference effects due to multiply scattered waves. These in turn implies that the only really important feature of the microscopic scattering mechanism is the preservation of phase memory. In spite of the many similarities of mesoscopic effects in systems of type I and II, the microscopic theoretical models to describe them are quite different. In the most understood regime (the semiclassical), impurity average diagrammatic techniques are used to study disordered conductors, whilst quantum chaotic scattering theory and trace formulas are used to describe transport in ballistic cavities . There are many alternative (non-microscopic) approaches to quantum transport that substitute the detailed information on the scattering potentials by more generic statistical hypothesis on macroscopic quantities, e.g. by assuming independent randomness in the scattering matrix of a macroscopic portion of the system. These alternative theories (known generically as random-matrix models ) have proved exact in many situations and are very useful in describing non-perturbative mesoscopic effects. The most remarkable aspect of these random-matrix models is the existence of an underlying mathematical structure that captures the relevant statistical information to the description of mesoscopic phenomena, which is known as the supersymmetric non-linear $`\sigma `$-model and has been introduced in condensed matter physics through the pioneering work by Efetov . For disordered systems non-linear $`\sigma `$-models of various kinds have been rigorously derived from microscopic models and succesfully applied to the non-perturbative regime . More recently, derivations of non-linear $`\sigma `$-models for ballistic chaotic cavities have been presented . The crucial underlying concept of these recent works is the existence of an intimate connection between the irreversible classical dynamics and the statistical quantum properties of the system. The inevitable conclusion is that the non-linear $`\sigma `$-model is a key concept in the search for a unified description of systems of type I and II. Earlier atempts in this direction have been confined to the semiclassical regime or to the thick-wire limit . In this letter we provide the complete picture by incorporating all physical regimes and useful limits. Specifically, we use a supersymmetric non-linear $`\sigma `$-model to describe the continuous crossover between a quantum dot (ballistic chaotic cavity) and a quantum wire (quasi-one-dimensional disordered conductor) in the absence of time-reversal symmetry. The system is coupled ideally (no transmission barriers) to two electron reservoirs through perfectly conducting leads that support and arbitrary number of open channels. We find an exact solution of the model and obtain closed expressions for several important transport properties. We show that our solution reproduces many known results of the literature by taking particular limits. Our starting point is the Landauer-Büttiker formalism, by which transport observables can be related to scattering functions, such as the transmission matrix $`t`$ of the system. Let $`N_1`$ and $`N_2`$ denote the number of channels in lead 1 and 2 respectively. We define $`t`$ as an $`N_1\times N_2`$ random transmission matrix describing the ideal coupling of the modes in lead 1 to those in lead 2 though a disordered sample of length $`L`$. In the limit $`L0`$ the system represents a quantum dot or a ballistic chaotic cavity, in which disorder scattering is negligible and the irregular fluctuations in the transmission matrix are due to surface scattering at the boundaries of the cavity. We shall measure the sample’s size $`L`$ in units of the localization length $`\xi `$, which we take as a free parameter, so that $`s2L/\xi `$ is a dimensionless variable controling the crossover from the ballistic ($`s=0`$) to the insulating regime ($`s1`$). The mean free path, $`l`$, sets the typical size of the cavity, so that when the disordered portion has length $`L>0`$ the combined dot-wire system has length $`L_{sys}>l`$. For $`\xi l`$ (large number of scattering channels) the system exhibits a metallic (or diffusive) behaviour which interpolates between the ballistic and the insulating regimes. A conventional approach to quantum transport is to study the statistical properties of a particular observable, such as the system’s conductance. Here, we shall instead consider a generating function from which many transport properties can be derived (such as the first three moments of the conductance or the average shot-noise power). Consider thus the following function $$𝒵(\{\theta \})=\mathrm{det}\left(\frac{1\mathrm{sin}^2(\theta _0/2)tt^{}}{1+\mathrm{sinh}^2(\theta _1/2)tt^{}}\right)$$ (1) and let $`\mathrm{\Psi }_s(\{\theta \})`$ be the average of $`𝒵(\{\theta \})`$ over realizations on the ensemble of random transmission matrices. Note that the joint distribution of the eigenvalues of $`tt^{}`$ contains $`s`$ as a parameter. Using a non-linear $`\sigma `$-model to be presented later we shall prove that $`\mathrm{\Psi }_s(\{\theta \})`$ $`=`$ $`1+(\mathrm{cos}\theta _0\mathrm{cosh}\theta _1){\displaystyle \underset{n=0}{\overset{N1}{}}}{\displaystyle _0^{\mathrm{}}}𝑑\mu _{nk}c_{nk}^{(N)}c_{nk}^{(M)}`$ (3) $`\times P_n(\mathrm{cos}\theta _0)P_{1/2+ik}(\mathrm{cosh}\theta _1)e^{\epsilon _{nk}s},`$ where $`d\mu _{nk}=(\epsilon _{nk})^1(2n+1)k\mathrm{tanh}\left(k\pi \right)dk`$ is the integration measure, $`\epsilon _{nk}=k^2+1/4+n(n+1)`$, $`N=\mathrm{min}(N_1,N_2)`$, $`M=\mathrm{max}(N_1,N_2)`$, $`P_n(x)`$ is the Legendre polynomial, $`P_{1/2+ik}(x)`$ is the conical function and $`c_{nk}^{(N)}={\displaystyle \frac{\left\|\mathrm{\Gamma }(N+1/2+ik)\right\|^2}{(Nn1)!(N+n)!}}.`$Equation (3) is the central result of this letter. It constitutes the exact solution of a realistic model of quantum transport that applies to all regimes of physical interest, namely: ballistic, diffusive, insulating, semiclassical and the extreme quantum limit. Furthermore, it contains the exact solution of important models, such as the thick wire and the DMPK theory , as particular cases. In order to demonstrate the usefulness of our explicit solution, we shall now discuss some concrete applications. (i) Moments of the conductance According to the Landauer-Bütikker scattering approach, the two probe conductance of a phase coherent system is given by $`G=G_0\mathrm{tr}(tt^{})`$, where $`G_0=2e^2/h`$ is the conductance quantum. It is easy to verify from (1) that the first three moments of the conductance can be obtained from $`\mathrm{\Psi }_s(\{\theta \})`$ through the relations: $`G/G_0_s=2\left(D_{\theta _0}\mathrm{\Psi }_s(\{\theta \})\right)\|_{\{\theta \}=0}`$ $`\left(G/G_0\right)^2_s=4\left(D_{\theta _0}D_{\theta _1}\mathrm{\Psi }_s(\{\theta \})\right)\|_{\{\theta \}=0}`$ $`\left(G/G_0\right)^3_s=4\left(D_{\theta _0}D_{\theta _1}(D_{\theta _1}D_{\theta _0})\mathrm{\Psi }_s(\{\theta \})\right)\|_{\{\theta \}=0}`$where $`D_\theta /(\mathrm{cos}\theta )`$. Using Eq. (3) we obtain the compact result $$\left(G/G_0\right)^m_s=2\underset{n=0}{\overset{N1}{}}_0^{\mathrm{}}𝑑\mu _{nk}c_{nk}^{(N)}c_{nk}^{(M)}g_{nk}^{(m)}e^{\epsilon _{nk}s},$$ (4) where $`g_{nk}^{(1)}=1`$, $`g_{nk}^{(2)}=k^2+1/4+n(n+1)`$ and $`g_{nk}^{(3)}=\frac{1}{4}(k^2+1/4)(k^2+9/4)+2n(n+1)(k^2+1/4)+\frac{1}{4}(n+2)(n+1)n(n1)`$. This expression reproduces (for $`m=1,2`$) the exact solution of the DMPK equation, presented in Ref. , by letting $`M\mathrm{}`$ (note that in this case $`c_{nk}^{(M)}1`$ and Eq. (4) remains otherwise unchanged). Another known exact solution is that obtained in Ref. for the first two moments of the conductance of a thick wire. It can be recovered by taking the limits $`N,M\mathrm{}`$, which implies $`c_{nk}^{(N)}c_{nk}^{(M)}1`$ and the sum is extended to infinity. Physically, in each of these limits we are neglecting (or reducing) the effects of the contact resistances at the lead-sample interfaces. We remark that the exact expressions for the third moment of the conductance in both limits (DMPK and thick wire) are presented here for the first time. There are a number of interesting transport regimes that can be described by our explicit solution. Let us consider here some of them separately. (a) The insulating regime ($`L\xi `$) In this regime the system becomes an effective disordered 1D wire and all moments of the conductance decay exponentially with increasing length as a consequence of wave-function localization. Is this case we may evaluate Eq. (4) for $`s1`$ using the saddle-point method, we find $`G/G_0_s`$ $``$ $`4\left(G/G_0\right)^2_s{\displaystyle \frac{64}{9}}\left(G/G_0\right)^3_s`$ $``$ $`2c_{0,0}^{(N_1)}c_{0,0}^{(N_2)}\left({\displaystyle \frac{\pi }{s}}\right)^{3/2}e^{s/4}.`$ (b) The quasi-ballistic regime ($`L_{sys}l`$) In this regime the system is in the transition from boundary surface dominated to bulk disorder dominated scattering as its length, $`L_{sys}`$, increases from its ballistic value $`l`$. We get expressions in this limit by expanding $`(G/G_0)^m_s`$ in a power series about $`s=0`$. We find $`G/G_0_s=g_c\left(1{\displaystyle \frac{g_cs}{1(\alpha g_c)^2}}+\mathrm{}\right)`$and $`\mathrm{var}(G/G_0)={\displaystyle \frac{(\alpha g_c^2)^2}{1(\alpha g_c)^2}}\left(1+{\displaystyle \frac{2(14\alpha g_c^2)s}{\alpha g_c\left(1(2\alpha g_c)^2\right)}}+\mathrm{}\right),`$where $`g_c=N_1N_2/(N_1+N_2)`$ is the dimensionless contact conductance and $`\alpha (N_1N_2)^1`$. Note that the first term reproduces the results of Ref. for quantum dots. (c) The semiclassical regime ($`N_1,N_21`$) This is the most studied regime of quantum transport and corresponds to a situation where quantum interference effects can be treated as a perturbation to classical transport theory. There are several useful methods to obtain expressions for observables in this regime, such as the diagrammatic impurity average perturbation theory , the large $`N`$ expansion of the non-linear $`\sigma `$-model and the semiclassical approach using trace formulas . For the present model we obtain $`G/G_0_s{\displaystyle \frac{1}{1/g_c+s}},`$which corresponds to the expected length dependence of the classical Ohm’s law. For the variance of the conductance we get $`\mathrm{var}(G/G_0){\displaystyle \frac{1}{15}}\left(1{\displaystyle \frac{6\gamma _5\gamma _1(1+g_cs)5\gamma _3^2}{\left(1+g_cs\right)^6\gamma _1^6}}\right),`$where $`\gamma _mN_1^m+N_2^m`$. This result agrees qualitatively with that of Ref. , where a discrete version of the non-linear $`\sigma `$-model has been used. The quantitative difference is not surprising since the continuum limit smoothes the ballistic-diffusive crossover (see also Ref. , where a different continuum model has been studied). In the diffusive limit, i.e. when $`sg_c1`$, we get simply $`G/G_0_ss^1`$ and $`\mathrm{var}(G/G_0)1/15`$ in agreement with rigorous microscopic diagrammatic calculations . We remark that one recovers the result of the ballistic-diffusive crossover predicted by the DMPK theory simply by taking the limit $`N_2\mathrm{}`$ (or $`N_1\mathrm{}`$) in all formulas. (ii) The average density of transmission eigenvalues We define $$\rho (\tau ,s)=\mathrm{tr}\delta (\tau tt^{})_s$$ (5) as the average density of eigenvalues of $`tt^{}`$ (usually called transmission eigenvalues). In our formalism $`\rho (\tau ,s)`$ can be obtained from Eq (3) through the equation $$\rho (\tau ,s)=\frac{1}{\pi \tau }\mathrm{Im}\left(\mathrm{tan}(\theta _0/2)\frac{}{\theta _0}\mathrm{\Psi }_s(\{\theta \})\right),$$ (6) in which $`\theta _0`$ and $`\theta _1`$ are related to $`\tau `$ by $`\mathrm{cos}\theta _0=\mathrm{cosh}\theta _1=12(\tau +i0^+)^1`$. Evaluating Eq. (6) we get $`\rho (\tau ,s)`$ $`=`$ $`{\displaystyle \frac{2}{\pi \tau ^2}}{\displaystyle \underset{n=0}{\overset{N1}{}}}{\displaystyle _0^{\mathrm{}}}𝑑\mu _{nk}\mathrm{cosh}(k\pi )c_{nk}^{(N)}c_{nk}^{(M)}`$ (8) $`\times P_n\left(12\tau ^1\right)P_{1/2+ik}(2\tau ^11)e^{\epsilon _{nk}s}.`$ In the limit $`s0`$ we recover the result of Ref. for quantum dots $`\rho (\tau ,0)=\tau ^\mu {\displaystyle \underset{n=0}{\overset{N1}{}}}(2n+\mu +1)\left\{P_n^{(\mu ,0)}(12\tau )\right\}^2,`$where $`\mu =MN`$ and $`P_n^{(\alpha ,\beta )}(x)`$ is the Jacobi polynomial. A useful application of Eq. (8) is the calculation of the ensemble average of arbitrary linear statistics, i.e. observables that can be written in the form $`A=\mathrm{tr}\left[f_A(tt^{})\right]`$. Using Eq. (5) one can see that $$A_s=_0^1\rho (\tau ,s)f_A(\tau ).$$ (9) As a particular application of this formula we consider the shot-noise power, which is given by $`P=P_0\mathrm{tr}\left[tt^{}(1tt^{})\right]`$, where $`P_0=2e\|V\|G_0`$ ($`V`$ is the applied voltage), and thus $`f_P(\tau )=\tau (1\tau )`$. Using Eq. (9) we find $`P/P_0_s=2{\displaystyle \underset{n=0}{\overset{N1}{}}}{\displaystyle _0^{\mathrm{}}}𝑑\mu _{nk}p_{nk}c_{nk}^{(N)}c_{nk}^{(M)}e^{\epsilon _{nk}s},`$where $`p_{nk}=3/4k^2+n(n+1)`$. We shall now sketch the proof of Eq. (3). The starting point is the supersymmetric non-linear $`\sigma `$-model representation of the ensemble average of the generating function (Eq. (1)), which reads $`\mathrm{\Psi }_s(\{\theta \})={\displaystyle 𝑑Q^{}𝑑Q^{\prime \prime }V_2(Q_0,Q^{})W_s(Q^{},Q^{\prime \prime })V_1(Q^{\prime \prime },Q)}`$where $`V_i(Q,Q^{})=\mathrm{Sdet}^{N_i}(Q+Q^{})`$, $`i=1,2`$, (Sdet stands for the superdeterminant in the convention of Ref. ) and $`dQ`$ in the invariant measure of the coset space $`𝒞=U(1,1/2)/(U(1/1)U(1/1))`$. This non-linear $`\sigma `$-model representation can be rigorously derived from a number of sofisticated stochastic approaches, such as the IWZ model or the gaussian white noise potential . Here we take, however, the point of view of regarding the non-linear $`\sigma `$-model as an independent minimal field theoretical description of the system, in the sense that it is free from all the statistically irrelevant details of specific models (either microscopic or random-matrix approaches). The price one has to pay for this model independence is that the localization length becomes a free parameter and the theory merely stablishes a scaling law for the physical observables. It has been shown that points in $`𝒞`$ can be represented by supermatrices of the form $`Q=U\left(\begin{array}{cc}\mathrm{cos}\widehat{\theta }& i\mathrm{sin}\widehat{\theta }\\ i\mathrm{sin}\widehat{\theta }& \mathrm{cos}\widehat{\theta }\end{array}\right)U^1;U=\left(\begin{array}{cc}u_1& 0\\ 0& u_2\end{array}\right),`$where $`\widehat{\theta }\mathrm{diag}(i\theta _1,\theta _0)`$; $`\theta _1>0`$, $`0<\theta _0<\pi `$, and $`u_1`$, $`u_2`$ are $`2\times 2`$ unitary supermatrices. In particular the origin of $`𝒞`$ is represented by $`Q_0=\mathrm{diag}(1,1,1,1)`$. The function $`W_s(Q^{},Q^{\prime \prime })`$ is the diffusion kernel of $`𝒞`$ and can be written as a path integral given by $`W_s(Q^{},Q^{\prime \prime })={\displaystyle \underset{x=0}{\overset{s}{}}dQ(x)\mathrm{exp}\left(\frac{1}{16}_0^s𝑑x\mathrm{Str}(_xQ)^2\right)},`$with boundary conditions $`Q(0)=Q^{\prime \prime }`$ and $`Q(s)=Q^{}`$. Here Str is the supertrace as defined in Ref . As shown in Ref. , the high degree of symmetry in this approach implies that $`\mathrm{\Psi }_s(\{\theta \})`$ satisfies the following diffusion equation $$\left(_s\mathrm{\Delta }_{\widehat{\theta }}\right)\mathrm{\Psi }_s(\{\theta \})=0$$ (10) with initial condition $`\mathrm{\Psi }_0(\{\theta \})={\displaystyle 𝑑Q^{}V_2(Q_0,Q^{})V_1(Q^{},Q)},`$where $`\mathrm{\Delta }_{\widehat{\theta }}`$ is the radial part of the Laplace-Beltrami operator of $`𝒞`$. The initial condition integral has been studied in Ref. and can be expressed in the following rather convenient form $`\mathrm{\Psi }_0(\{\theta \})`$ $`=`$ $`1+(\mathrm{cos}\theta _0\mathrm{cosh}\theta _1){\displaystyle \underset{l=0}{\overset{N1}{}}}{\displaystyle \frac{(1)^{l+1}}{2}}`$ $`\times {\displaystyle \frac{\mathrm{sin}^{2l}(\theta _0/2)}{\mathrm{sinh}^{2l+1}(\theta _1/2)}}\left[f_{Nl1}^{(\mu )}(\theta _0)g_{Nl1}^{(\mu )}(\theta _1)1\right]`$ where $`f_n^{(\mu )}(\theta _0)=F(n,n\mu ;2n\mu ;\mathrm{sin}^2(\theta _0/2))`$ and $`g_n^{(\mu )}(\theta _1)=F(n+1,n+\mu +1;2n+\mu +2;\mathrm{sinh}^2(\theta _1/2))`$ are hypergeometric functions. We solve Eq. (10) by expanding $`\mathrm{\Psi }_s(\{\theta \})`$ in a complete set of eigenfunctions of $`\mathrm{\Delta }_{\widehat{\theta }}`$ (for a detailed discussion of this procedure see Ref ). The final result is Eq. (3). In summary, we have studied a unified field-theoretical description of the quantum dot - quantum wire crossover for system with broken time-reversal symmetry. Using a generating function approach we have found closed expressions for a number of important transport properties, such as the first three moments of the conductance, the average shot-noise power and the average density of transmission eigenvalues. Extensions of our results to other symmetry classes (orthogonal and symplectic) and to more elaborate correlation functions (with explicit dependence on energy and/or magnetic field) appear to be possible by means of the theory of superharmonic analysis presented in Ref. . An interesting immediate application of the techniques presented here would be the derivation of extensions of the variance and the length-correlation formulas to all physical regimes and length scales. This work was partially supported by CNPq, FACEPE and FINEP (Brazilian Agencies).
no-problem/9907/hep-ph9907352.html	ar5iv	text	# Absorption and percolation in the production of J/𝜓 in heavy ion collisions ## Acknowledgements This work was partially supported by project PRAXIS/PCEX/P/FIS/124/96.R.U. gratefully acknowledges the financial support of the Funda o Ci ncia e Tecnologia via the “Sub-Programa Ciência e Tecnologia do $`2^o`$ Quadro Comunitário de Apoio.” ## References
no-problem/9907/nucl-th9907072.html	ar5iv	text	# Intrinsic State for an extended version of the interacting boson model (IBM-4) ## Abstract An intrinsic-state formalism for IBM-4 is presented. A basis of deformed bosons is introduced which allows the construction of a general trial wave function which has Wigner’s spin–isospin SU(4) symmetry as a particular limit. Intrinsic-state calculations are compared with exact ones showing good agreement. PACS numbers: 21.60 -n, 21.60 Fw, 21.60 Ev The Interacting Boson Model (IBM) was originally proposed to describe collective low-lying states in even-even nuclei. The model building blocks are monopolar ($`s`$) and quadrupolar ($`d`$) bosons. In the original formulation of the model (IBM-1) no distinction was made between neutrons and protons . Later, connections with the nuclear shell model were investigated and a new version was proposed in terms of neutron ($`s_\nu ,d_\nu `$) and proton ($`s_\pi ,d_\pi `$) bosons, known as IBM-2 . The model has been widely applied to medium-mass and heavy nuclei, where neutrons and protons are filling different major shells. In lighter nuclei with $`NZ`$, however, neutrons and protons are in the same shell and a boson made of one neutron and one proton (known as a $`\delta `$ boson) should be included. This version of the boson model, called IBM-3 , is the simplest isospin invariant formulation of IBM. The three types of bosons ($`\nu `$, $`\pi `$, and $`\delta `$) form an isospin $`T=1`$ triplet and correspond, microscopically, to spatially symmetric nucleon pairs with $`S=0`$. In particular, the $`\delta `$ boson corresponds to a spatially symmetric $`S=0`$ neutron-proton pair. A further extension of the IBM introduces the neutron-proton boson with $`T=0`$ or $`\sigma `$ boson, corresponding to a spatially symmetric nucleon pair with $`S=1`$. This version is known as IBM-4 and gives a proper description of even-even as well as odd-odd $`NZ`$ nuclei. The IBM-3 and IBM-4 are appropriate models for $`NZ`$ nuclei approaching the proton drip line. Such nuclei are studied intensively at the moment in particular with radioactive nuclear beams. Also, the IBM-4 is a reasonably simple, yet detailed model to study the competition between $`T=0`$ and $`T=1`$ pairing, one of the hot topics in current-day nuclear structure physics. All versions of IBM are algebraic in nature and do not have a direct geometrical interpretation. Such interpretation can be achieved, however, by introducing an intrinsic state which provides a connection to geometric models such as that of Bohr and Mottelson . Intrinsic states have been proposed for IBM-1 , for IBM-2 , and for IBM-3 . Their primary use is to provide a geometric visualization of the model. In addition, a considerable reduction is achieved in the complexity of calculations, which leaves room for the inclusion of extra degrees of freedom. The purpose of this letter is to propose an intrinsic state for IBM-4. In the limit of strong isovector pairing it reduces to the intrinsic state for IBM-3; in general, it can be used for studying the competition between $`T=0`$ and $`T=1`$ pairing in $`NZ`$ nuclei. First, the mean-field formalism for IBM-4 is presented. This formalism is subsequently checked against the results of an exact calculation. The ensemble of bosons in the IBM-4 consists of isovector $`T=1`$ and isoscalar $`T=0`$ bosons which have intrinsic spin $`S=0`$ and $`S=1`$, respectively, to ensure spatial symmetry. The allowed spin-isospin combinations are thus $`(T,S)=(1,0)`$ and $`(T,S)=(0,1)`$. These, together with the orbital angular momenta $`\mathrm{}=0,2`$, give rise to 36 different bosons. The corresponding boson creation and annihilation operators are $`\gamma _{\mathrm{}m,T\tau ,S\sigma }^{}`$ and $`\gamma _{\mathrm{}m,T\tau ,S\sigma }`$ where $`\mathrm{}`$ is the orbital angular momentum, $`m`$ is its projection, $`T`$ is the isospin, $`\tau `$ is its projection, $`S`$ is the spin, and $`\sigma `$ is its projection. The operators $`\stackrel{~}{\gamma }_{\mathrm{}m,T\tau ,S\sigma }=(1)^{\mathrm{}+T+Sm\tau \sigma }\gamma _{\mathrm{}m,T\tau ,S\sigma }`$ are introduced for having appropriate tensor transformation properties. The construction of an intrinsic state requires two ingredients. First, it needs a basis of deformed bosons and secondly, it requires a trial wave function. The deformed bosons are defined in terms of the spherical ones through a unitary Hartree-Bose transformation, $$\mathrm{\Omega }_{p,T\tau ,S\sigma }^{}=\underset{\mathrm{}m}{}\lambda _\mathrm{}m^{pT\tau S\sigma }\gamma _{\mathrm{}m,T\tau ,S\sigma }^{},\gamma _{\mathrm{}m,T\tau ,S\sigma }^{}=\underset{p}{}\lambda _\mathrm{}m^{pT\tau S\sigma }\mathrm{\Omega }_{p,T\tau ,S\sigma }^{},$$ (1) and their hermitian conjugates. The deformation parameters $`\lambda `$ in these equations verify the following orthonormalization relations: $$\underset{\mathrm{}m}{}\lambda _\mathrm{}m^{p^{}T\tau S\sigma }\lambda _\mathrm{}m^{pT\tau S\sigma }=\delta _{pp^{}},\underset{p}{}\lambda _\mathrm{}m^{pT\tau S\sigma }\lambda _\mathrm{}^{}m^{}^{pT\tau S\sigma }=\delta _{\mathrm{}\mathrm{}^{}}\delta _{mm^{}}.$$ (2) For convenience, a global label $`\xi `$ is used for spin, isospin, and their projections, $`\xi (T\tau S\sigma )`$. This new index plays the same role in IBM-4 as the isospin projection $`\tau `$ does in the IBM-3 intrinsic-state formalism . The index $`p`$ labels the different deformed bosons. The fundamental deformed boson has $`p=0`$ while excited ones have $`p=1,2,\mathrm{},_{\mathrm{}}(2\mathrm{}+1)1`$. If only $`s`$ and $`d`$ bosons are included, the maximum value of $`p`$ is 5. Only the ground-state properties are considered here; so the superscript $`p`$ is always zero and can be omitted henceforth. The definition of the ground-state trial wave function follows Ref. ; it is different depending on whether the system is even-even and odd-odd. For even-even nuclei with proton excess (the case of neutrons excess is obtained by interchanging $`N_p`$ and $`N_n`$) the proposed trial wave function for the ground state has the form $$\|\varphi (\delta ,\alpha _T,\alpha _S)_{ee}=\mathrm{\Delta }_{}^{}{}_{}{}^{N_n}(\delta ,\alpha _T,\alpha _S)\mathrm{\Omega }_{T=1\tau =1}^{N_pN_n}\|0,$$ (3) where $`N_n`$ ($`N_p`$) is half the number of valence neutrons (protons) and $`\mathrm{\Delta }^{}(\delta ,\alpha _T,\alpha _S)`$ $`=`$ $`(\mathrm{\Omega }_{T=1\tau =1}^{}\mathrm{\Omega }_{T=1\tau =1}^{}+\alpha _T\mathrm{\Omega }_{T=1\tau =0}^{}\mathrm{\Omega }_{T=1\tau =0}^{})`$ (4) $`+`$ $`\delta (\mathrm{\Omega }_{S=1\sigma =1}^{}\mathrm{\Omega }_{S=1\sigma =1}^{}+\alpha _S\mathrm{\Omega }_{S=1\sigma =0}^{}\mathrm{\Omega }_{S=1\sigma =0}^{}).`$ (5) The description of odd-odd nuclei is complicated even in the ground state, since in general its spin-isospin values are not known a priori. In $`N=Z`$ nuclei, which is the case of most interest, those values are known, being either $`(T,S)=(1,0)`$ or $`(T,S)=(0,1)`$. Correspondingly, two trial wave functions are proposed $$\|\varphi (\delta ,\alpha _T,\alpha _S)_{oo1}=\mathrm{\Delta }_{}^{}{}_{}{}^{N_n\frac{1}{2}}(\delta ,\alpha _T,\alpha _S)\mathrm{\Omega }_{T=1\tau =1}^{}\|0,$$ (6) and $$\|\varphi (\delta ,\alpha _T,\alpha _S)_{oo2,\sigma }=\mathrm{\Delta }_{}^{}{}_{}{}^{N_n\frac{1}{2}}(\delta ,\alpha _T,\alpha _S)\mathrm{\Omega }_{S=1\sigma }^{}\|0.$$ (7) Which of these two states is lower in energy depends on a delicate balance of the different terms in the Hamiltonian which in turn follow from the competition between $`T=0`$ and $`T=1`$ pairing. In addition to the deformation parameters, three variational parameters, $`\alpha _T`$, $`\alpha _S`$, and $`\delta `$, appear in the trial wave functions. The first two are related to isospin and spin symmetry breaking in the trial wave function. For deformation parameters independent of $`\xi `$ and for $`\alpha _T=\alpha _S=1/2`$, the operator $`\mathrm{\Delta }^{}(\delta ,\alpha _T,\alpha _S)`$ corresponds to a bosonic pair scalar in spin and isospin. The parameter $`\delta `$ measures the relative importance of isovector and isoscalar bosons in the ground state. In the limit of $`\delta =0`$, the number of isoscalar bosons in the ground state is zero and the IBM-3 trial state is recovered . Another interesting limit is $`\delta =\pm 1`$ which is obtained if the IBM-4 Hamiltonian has Wigner’s SU(4) symmetry . In this case $`T=0`$ and $`T=1`$ bosons are treated on equal footing. Given a general IBM-4 Hamiltonian, $`\widehat{H}`$, the ground-state equilibrium parameters are obtained by minimizing the expectation value of the Hamiltonian in the appropriate trial wave function (3,6,7). A general expression for this expectation value can be written as $`E(\lambda ,\delta ,\alpha _T,\alpha _S)`$ $`=`$ $`{\displaystyle \underset{\xi _1\xi _2}{}}ϵ_{\xi _1\xi _2}\stackrel{~}{f_1}(\delta ,\alpha _T,\alpha _S,\xi _1\xi _2)`$ (8) $`+`$ $`{\displaystyle \underset{\xi _1\xi _2\xi _3\xi _4}{}}V_{\xi _1,\xi _2,\xi _3,\xi _4}^c\stackrel{~}{f_2}(\delta ,\alpha _T,\alpha _S,\xi _1\xi _2\xi _3\xi _4),`$ (9) where, $$ϵ_{\xi _1,\xi _2}=\underset{\mathrm{}_1m_1\mathrm{}_2m_2}{}\stackrel{~}{\epsilon }_{\mathrm{}_1m_1\xi _1\mathrm{}_2m_2\xi _2}\lambda _{\mathrm{}_1m_1}^{\xi _2}\lambda _{\mathrm{}_2m_2}^{\xi _2},$$ (10) $$\begin{array}{c}V_{\xi _1,\xi _2,\xi _3,\xi _4}^c=\underset{\mathrm{}_1m_1\mathrm{}_2m_2\mathrm{}_3m_3\mathrm{}_4m_4}{}V_{\mathrm{}_1m_1\xi _1,\mathrm{}_2m_2\xi _2,\mathrm{}_3m_3\xi _3,\mathrm{}_4m_4\xi _4}\lambda _{\mathrm{}_1m_1}^{\xi _1}\lambda _{\mathrm{}_2m_2}^{\xi _2}\lambda _{\mathrm{}_3m_3}^{\xi _3}\lambda _{\mathrm{}_4m_4}^{\xi _4},\end{array}$$ (11) $$\stackrel{~}{f_1}(\delta ,\alpha _T,\alpha _S,\xi _1\xi _2)=\delta _{\xi _1\xi _2}\frac{\varphi (\delta ,\alpha _T,\alpha _S)\|\mathrm{\Omega }_{\xi _1}^{}\mathrm{\Omega }_{\xi _2}\|\varphi (\delta ,\alpha _T,\alpha _S)}{\varphi (\delta ,\alpha _T,\alpha _S)\varphi (\delta ,\alpha _T,\alpha _S)},$$ (12) and $$\stackrel{~}{f_2}(\delta ,\alpha _T,\alpha _S,\xi _1\xi _2\xi _3\xi _4)=\frac{\varphi (\delta ,\alpha _T,\alpha _S)\|\mathrm{\Omega }_{\xi _1}^{}\mathrm{\Omega }_{\xi _2}^{}\mathrm{\Omega }_{\xi _3}\mathrm{\Omega }_{\xi _4}\|\varphi (\delta ,\alpha _T,\alpha _S)}{\varphi (\delta ,\alpha _T,\alpha _S)\varphi (\delta ,\alpha _T,\alpha _S)}.$$ (13) The isospin matrix elements (12,13) are calculated straightforwardly from a binomial expansion of the trial wave function. Furthermore, the parameters $`\stackrel{~}{\epsilon }_{\mathrm{}_1m_1\xi _1\mathrm{}_2m_2\xi _2}`$ and $`V_{\mathrm{}_1m_1\xi _1,\mathrm{}_2m_2\xi _2,\mathrm{}_3m_3\xi _3,\mathrm{}_4m_4\xi _4}`$ in equations (10,11) are defined as $$\stackrel{~}{\epsilon }_{\mathrm{}_1m_1\xi _1\mathrm{}_2m_2\xi _2}\mathrm{}_1m_1\xi _1\|\widehat{H}\|\mathrm{}_2m_2\xi _2,$$ (14) $`V_{\mathrm{}_1m_1\xi _1,\mathrm{}_2m_2\xi _2,\mathrm{}_3m_3\xi _3,\mathrm{}_4m_4\xi _4}`$ $``$ $`{\displaystyle \frac{1}{4}}\mathrm{}_1m_1\xi _1,\mathrm{}_2m_2\xi _2\left\|\widehat{V}\right\|\mathrm{}_3m_3\xi _3,\mathrm{}_4m_4\xi _4`$ (15) $`\times `$ $`\sqrt{1+\delta _{\mathrm{}_1\mathrm{}_2}\delta _{m_1m_2}\delta _{\xi _1\xi _2}}\sqrt{1+\delta _{\mathrm{}_3\mathrm{}_4}\delta _{m_3m_4}\delta _{\xi _3\xi _4}},`$ (16) where $`\widehat{V}`$ stands for the two-body terms in the Hamiltonian $`\widehat{H}`$. It is worth noting that $`\stackrel{~}{\epsilon }`$ is not diagonal in $`\mathrm{}`$. The energy (8) depends explicitly on the deformation parameters $`\lambda `$ and implicitly on $`\alpha _T`$, $`\alpha _S`$, and $`\delta `$ through $`\stackrel{~}{f}_1`$ and $`\stackrel{~}{f}_2`$. The deformation parameters $`\lambda `$ are obtained by minimizing the energy with the constraint of conserving the transformation norm, $$\delta [E(\lambda ,\delta ,\alpha _T,\alpha _S)\underset{\xi \mathrm{}m}{}E_\xi \lambda _\mathrm{}m^\xi \lambda _\mathrm{}m^\xi ]=0.$$ (17) The Hartree-Bose equations for IBM-4 are obtained by deriving with respect to $`\lambda _\mathrm{}m^\xi `$, $$\underset{\mathrm{}_2m_2}{}h_{\mathrm{}_1m_1,\mathrm{}_2m_2}^\xi \lambda _{\mathrm{}_2m_2}^\xi =E_\xi \lambda _{\mathrm{}_1m_1}^\xi ,$$ (18) where the Hartree-Bose matrix, $`h^\xi `$, is $`h_{\mathrm{}_1m_1,\mathrm{}_2m_2}^\xi =\stackrel{~}{\epsilon }_{\mathrm{}_1m_1\xi \mathrm{}_2m_2\xi }\stackrel{~}{f_1}(\delta ,\alpha _T,\alpha _S,\xi \xi )\delta _{m_1m_2}`$ (19) $`+`$ $`2{\displaystyle \underset{\mathrm{}_3m_3\mathrm{}_4m_4\xi _2\xi _3\xi _4}{}}V_{\mathrm{}_1m_1\xi ,\mathrm{}_3m_3\xi _3,\mathrm{}_4m_4\xi _4,\mathrm{}_2m_2\xi _2}{\displaystyle \frac{\lambda _{\mathrm{}_3m_3}^{\xi _3}\lambda _{\mathrm{}_4m_4}^{\xi _4}\lambda _{\mathrm{}_2m_2}^{\xi _2}}{\lambda _{\mathrm{}_2m_2}^\xi }}\stackrel{~}{f_2}(\delta ,\alpha _T,\alpha _S,\xi \xi _3\xi _4\xi _2).`$ (20) There are six coupled equations of this form, which are solved for fixed values of $`\alpha _T`$, $`\alpha _S`$, and $`\delta `$ in a self-consistent way. This procedure yields the equilibrium deformation parameters $`\lambda `$. The equilibrium values for the parameters $`\alpha _T`$, $`\alpha _S`$, and $`\delta `$ are obtained by an additional minimization. In fact, if the deformation parameters are independent of $`\xi `$, the parameters $`\alpha _T`$ and $`\alpha _S`$ can be fixed to the value $`1/2`$ which corresponds, as mentioned above, to a state with well-defined spin and isospin. As was shown for IBM-3 this is a good approximation for $`N=Z`$ nuclei. To test the present formalism, comparisons with exact calculations are carried out. Numerical calculations in the framework of IBM-4 are now possible but still difficult. Also, only few dynamical limits have been studied. Here the following schematic Hamiltonian is considered $$\widehat{H}=a\widehat{C}_2[SU_{TS}(4)]+b\widehat{C}_2[SU_S(3)]+c\widehat{C}_2[SU_L(3)],$$ (22) where $`\widehat{C}_2[G]`$ stands for the quadratic Casimir operator of the algebra $`G`$. The first operator is an invariant of the algebra $`SU_{TS}(4)`$ which is the boson equivalent of Wigner’s supermultiplet algebra . It is worth noting here that, as mentioned in , there are two alternative $`SU_{TS}(4)`$ limits with the same eigenspectrum but different phases in the wave function. The results presented below are obtained within one them, which is associated to the election of the operator $`\widehat{Y}_{\mu \nu }^+`$ (using the notation of Ref. ). The use of the alternative limit, using $`\widehat{Y}_{\mu \nu }^{}`$, gives identical results but changing sign to $`\delta `$. The second operator in (22) is an invariant of the $`SU_S(3)`$ algebra associated with the $`\sigma `$ ($`S=1,T=0`$) boson. (Its definition is analogous to that of $`SU_S(3)`$ considered in the $`L=0`$ IBM-4 of Ref. .) The last invariant in (22) is an orbital deformation term associated with an $`SU(3)`$ algebra which is scalar in spin and isospin. With this Hamiltonian three situations are studied. The first case corresponds to $`a0`$ and $`c=0`$. In Ref. the competition of $`T=0`$ and $`T=1`$ pairing was discussed using this Hamiltonian. This is a relevant test for the formalism presented here since it explores the spin-isospin degrees of freedom which makes the main difference of IBM-4 with respect to previous versions of IBM. In this case the mean-field and exact calculations are almost identical although the exact calculation is always slightly lower in energy. This can be appreciated in table I where exact and mean-field ground-state energies (in units of $`a`$) are given for $`N=5`$ and $`N=15`$ bosons for selected values of $`b/a`$. The value $`b=0`$ yields a Hamiltonian with the $`SU(4)`$ symmetry, and degenerate lowest $`T=0`$ and $`T=1`$ states. Negative values of the ratio $`b/a`$ favor $`T=0`$ while positive values favor $`T=1`$ pairing. The expectation value of the schematic Hamiltonian (22) with $`c=0`$ is independent of the deformation parameters. The minimum of the energy occurs for $`\xi `$ independent parameters and $`\alpha _T=\alpha _S=\frac{1}{2}`$. This is so because it has no orbital dependence. It is worth noting that the variational wave functions (3-7) contain for special values of $`\delta `$ the lowest eigenfunctions of the $`SU_T(3)SU_S(3)`$ limit ($`\delta =\mathrm{}`$) and of the $`SU_{TS}(4)`$ limit ($`\delta =1`$). The second case corresponds to $`a0`$ and $`b=0`$ and includes the deformation term $`\widehat{C}_2[SU_L(3)]`$. The exact ground-state energy is known analytically: $$E=a\lambda (\lambda +4)+c2N(2N+3).$$ (23) This is so because, for sufficiently large negative $`c`$, the ground state belongs to the $`SU_{TS}(4)`$ representation $`(0,\lambda ,0)`$ and the $`SU_L(3)`$ representation $`(2N,0)`$, where $`N`$ is the boson number and $`\lambda =T`$ for even-even nuclei and $`\lambda =1`$ for odd-odd $`N=Z`$ nuclei . The corresponding calculation is also performed with the mean-field formalism presented here and the exact results are reproduced. The calculation gives an intrinsic state with $`\beta _\xi \lambda _{20}^\xi /\lambda _{00}^\xi =\sqrt{2}`$, $`\alpha _T=\alpha _S=\frac{1}{2}`$ and $`\delta =1`$ which is an eigenstate of the Hamiltonian (22) with $`a0`$ and $`b=0`$. This result is similar to that obtained for the intrinsic state of IBM-1 . The last case considered is the general one with $`a,b,c`$ different from zero. The ground state still belongs to the $`SU_L(3)`$ representation $`(2N,0)`$ and hence the contribution to the ground-state energy coming from $`c\widehat{C}_2[SU_L(3)]`$ is diagonal. The other two terms in the Hamiltonian can be diagonalized as in Ref. . Thus the exact energies are those calculated in table I (under ‘exact’) plus $`c2N(2N+3)`$. This calculation is repeated with the mean-field formalism and produces an intrinsic state with the same $`SU_L(3)`$ symmetry as the exact one, $`(2N,0)`$. As in the preceding case $`\beta _\xi \lambda _{20}^\xi /\lambda _{00}^\xi =\sqrt{2}`$, $`\alpha _T=\alpha _S=\frac{1}{2}`$, but now $`\delta 1`$. The mean-field energies are those given in table I (under ‘mean field’) plus $`c2N(2N+3)`$. These results demonstrate that the present mean-field formalism has the correct ingredients for reproducing the full IBM-4 calculation. In addition, this formalism allows calculations for an arbitrary number of bosons and a general Hamiltonian, not necessarily corresponding to a dynamical symmetry limit of the model. To summarize, a Hartree-Bose mean-field approximation for IBM-4 has been presented, along with trial wave functions valid for even-even and odd-odd nuclei with $`N=Z`$. The trial wave functions include boson correlations in the spin and isospin spaces. Comparisons with exact calculations show good agreement from which can be inferred that the present formalism gives a good approximation to the full diagonalization. The aim is now to consider more realistic IBM-4 Hamiltonians that include both types of pairing ($`T=0`$ and $`T=1`$), and a spin-orbit coupling as well as more general quadrupole deformation terms. This will enable the study of the interplay between single-particle, spin-isospin, and orbital degrees of freedom. This work is currently in progress. We wish to thank S. Pittel and D.D. Warner for useful comments. This work has been supported by the Spanish DGICYT under contract Nos. PB98-1111 and PB95-0123 and a DGICYT-IN2P3 agreement.
no-problem/9907/astro-ph9907244.html	ar5iv	text	# 1 Introduction ## 1 Introduction The universe is observed through electromagnetic waves. From the stars the waves are visible, from hot plasmas they are x-rays, from the hyperfine transition in hydrogen they are radio waves, and they are microwaves from the cosmic background radiation. Not all matter in the universe, however, emits detectable radiation. Examples of this are blackholes and zero-mass neutrinos. The difference between the detectable mass and the total mass that should be according to the laws of gravity is ascribed to the so-called dark matter, whose existence is inferred only from its gravitational interaction. The visible parts of the galaxies are composed mainly of stars which do not satisfy Newton’s mechanics and thus are hypothized to be surrounded by extended halos of dark matter which may be a factor of 30 or more in both mass and size. The existence of the planet Neptune was predicted from the unexpected residuals in the motion of Uranus. A negative example, on the other hand, is the precession of the planet Murcury’s perihelion. A hypothetical planet or a ring of matter, inside Murcury’s orbit, was hypothized to exist in order to explain the anomaly. No planet or material ring was observed. As is well known, the anomaly was resolved by Einstein’s general relativity theory . This is a reminder that much of the assumed missing matter might be explained by new theories. In this paper we use a space-velocity version of general relativity theory \[2-4\] in combination with Einstein’s standard general relativity theory to show that much of the unexplained observation can be satisfactorily described. More precisely, we simply prove that the Tully-Fisher law is included in the equations of motion obtained. ## 2 The Tully-Fisher Law Astronomical observations show that for disk galaxies the fourth power of the circular speed of stars moving around the core of the galaxy, $`v_c^4`$, is proportional to the total luminosity $`L`$ of the galaxy to an accuracy of more than two orders of magnitude in $`L`$, namely $`v_c^4L`$. Since $`L`$ is proportional to the mass $`M`$ of the galaxy, one obtains $`v_c^4M`$. This is known as the Tully-Fisher law . No dependence on the distance of the star from the center of the galaxy as Newton’s law $`v_c^2=GM/r`$ requires for circular motion. In order to rectify this deviation from Newton’s laws astronomers assume the existence of halos around the galaxy which are filled with dark matter and arranged in such ways so as to satisfy the Tully-Fisher law for each particular situation. It is well known that Newton’s second law also follows from Einstein’s general relativity theory in the lowest approximation in $`v/c`$, where $`v`$ is a characteristic velocity and $`c`$ the speed of light. For this reason we exclude the possibility of modifying Newton’s second law of motion by such as adding to it a term which takes care of the anomaly . Any arbitrary modification of Newton’s second law is therefore spurious even if it yields results that fit observations quite well. ## 3 Constraints on Motion of Stars A star moving around the galaxy experiences the expansion of the universe. This is a constraint on the dynamical system that should be taken into account, and without which the theory is invalid. The expansion of the universe causes an increase to the distance between the star and the center of the galaxy. But when this distance increases, the circular velocity changes accordingly. This constraint on the dynamical system should be taken into account along with the centrifugal formula $`v_c^2=GM/r`$. But a space-velocity version of general relativity is exactly the right theory to give this extra relation between the mass of the galaxy, the circular velocity and the distance of the star to the center of the galaxy. In this paper we derive the extra relation from the space-velocity version of general relativity theory and show that its combination with Einstein’s general relativity yield a term of the form $`v_c^4M`$. ## 4 Equations of Motion In Einstein’s general relativity theory the equations of motion follow from the vanishing of the covariant divergence of the energy-momentum tensor. This is a result of the restricted Bianchi identities. The equations obtained are usually the geodesic equation. By means of a successive approximation in $`v/c`$, one obtains the Newtonian equations of motion and its generalization to a higher accuracy \[8-25\]. In the space-velocity version of general relativity the situation is the same. Accordingly one has $$\frac{d^2x^\mu }{ds^2}+\mathrm{\Gamma }_{\alpha \beta }^\mu \frac{dx^\alpha }{ds}\frac{dx^\beta }{ds}=0.$$ $`(1)`$ We now find the lowest approximation of Eq. (1) in terms of $`t/\tau `$, where $`t`$ is a characteristic cosmic time and $`\tau `$ is the Hubble time, using the Einstein-Infeld-Hoffmann method . To this end we change variables in Eq. (1) from $`s`$ to $`v`$, where $`v`$ is related to the velocity-like coordinate $`x^0`$ by $`x^0=\tau v`$. A simple calculation gives $$\frac{d^2x^k}{dv^2}+\left(\mathrm{\Gamma }_{\alpha \beta }^k\frac{1}{\tau }\frac{dx^k}{dv}\mathrm{\Gamma }_{\alpha \beta }^0\right)\frac{dx^\alpha }{dv}\frac{dx^\beta }{dv}=0,$$ $`(1a)`$ with $`k=1,2,3`$. As seen, one can neglect the second term in the paranthesis since it is one order smaller than the first, and thus $$\frac{d^2x^k}{dv^2}+\mathrm{\Gamma }_{\alpha \beta }^k\frac{dx^\alpha }{dv}\frac{dx^\beta }{dv}0.$$ The second term is equal to $$\mathrm{\Gamma }_{00}^k\left(\frac{dx^0}{dv}\right)^2+2\mathrm{\Gamma }_{0b}^k\frac{dx^0}{dv}\frac{dx^b}{dv}+\mathrm{\Gamma }_{ab}^k\frac{dx^a}{dv}\frac{dx^b}{dv}.$$ But $`x^0=\tau v`$, thus the 2nd and the 3rd terms may be neglected with respect to the 1st, and we obtain $$\frac{d^2x^k}{dv^2}+\tau ^2\mathrm{\Gamma }_{00}^k0.$$ The Christoffel symbol can be calculated also, $$\mathrm{\Gamma }_{00}^k=\frac{1}{2}g^{k\rho }\left(2_0g_{\rho 0}_\rho g_{00}\right).$$ Again a $`x^0`$-derivative $`_0=\tau ^1_v`$ which is of higher order in $`t/\tau `$, thus $$\mathrm{\Gamma }_{00}^k\frac{1}{2}g^{k\rho }_\rho g_{00}\frac{1}{2}g^{ks}_sg_{00}.$$ Since $`g^{ks}\eta ^{ks}=\delta ^{ks}`$, we obtain $$\mathrm{\Gamma }_{00}^k\frac{1}{2}\frac{g_{00}}{x^k},$$ $`(2)`$ and thus the geodesic equation yields $$\frac{d^2x^k}{dv^2}+\frac{\tau ^2}{2}\frac{g_{00}}{x^k}0.$$ $`(3)`$ Writing now $`g_{00}=1+2\varphi /\tau ^2`$, we then obtain $$\frac{d^2x^k}{dv^2}=\frac{\varphi }{x^k},$$ $`(4)`$ for the equations of motion in the lowest approximation. It remains to find out the function $`\varphi \left(x\right)`$. ## 5 Field Equations To find out the function $`\varphi `$ we have to solve the gravitational field equations. Again this should be in the lowest approximation as explained before. We have \[2-4\] $$R_{\mu \nu }=\kappa \left(T_{\mu \nu }\frac{1}{2}g_{\mu \nu }T\right),$$ $`(5)`$ where $`T=T_{\rho \sigma }g^{\rho \sigma }`$ and $$\kappa =\frac{8\pi k}{\tau ^4},$$ $`(6)`$ with $`k=G\left(\tau ^2/c^2\right)`$. We then have $$T=T_{\mu \nu }g^{\mu \nu }T_{\mu \nu }\eta ^{\mu \nu }T_{00}\eta ^{00}=T_{00}.$$ $`(7)`$ Thus we obtain $$R_{00}=\kappa \left(T_{00}\frac{1}{2}g_{00}T\right)\frac{1}{2}\kappa T_{00}=\frac{1}{2}\kappa \tau ^2\rho \left(x\right),$$ $`(8)`$ where $`\rho \left(x\right)`$ is the mass density. The approximate value of $`R_{00}`$ is $$R_{00}=\frac{\mathrm{\Gamma }_{00}^\rho }{x^\rho }\frac{\mathrm{\Gamma }_{0\rho }^\rho }{x^0}+\mathrm{\Gamma }_{00}^\rho \mathrm{\Gamma }_{\rho \sigma }^\sigma \mathrm{\Gamma }_{0\rho }^\sigma \mathrm{\Gamma }_{0\sigma }^\rho \frac{\mathrm{\Gamma }_{00}^\rho }{x^\rho }\frac{\mathrm{\Gamma }_{00}^s}{x^s}.$$ $`(9)`$ Using now Eq. (2) for the value of the Christoffel symbol, we obtain $$R_{00}\frac{\mathrm{\Gamma }_{00}^s}{x^s}=\frac{1}{2}\frac{^2g_{00}}{x^sx^s}=\frac{1}{2}^2g_{00}\frac{1}{\tau ^2}^2\varphi ,$$ $`(10)`$ where $`^2`$ is the ordinary Laplace operator. Equations (8) and (10) then give $$^2\varphi =\frac{1}{2}\kappa \tau ^4\rho ,$$ $`(11)`$ or, using Eq. (6), $$^2\varphi \left(x\right)=4\pi k\rho \left(x\right).$$ $`(12)`$ This equation is exactly the Newtonian equation for gravity but with $`k=G\tau ^2/c^2`$ replacing the Newtonian constant $`G`$ \[2-4\]. ## 6 Integration of the Equations of Motion The integration of the equation of motion (4) is identical to that familiar in classical Newtonian mechanics. But there is an essential difference which should be emphasized. In Newtonian equations of motion one deals with a path of motion in the 3-space. In our theory we do not have that situation. Rather, the paths here indicate locations of particles in the sense of the Hubble distribution which now takes a different physical meaning. With that in mind we proceed as follows. Equation (4) yields the first integral $$\left(\frac{ds}{dv}\right)^2=\frac{kM}{r},$$ $`(13)`$ where $`v`$ is the velocity of the particles, in analogy to the Newtonian $$\left(\frac{ds}{dt}\right)^2=\frac{GM}{r}.$$ $`(14)`$ In these equations $`s`$ is the length parameter along the path of accumulation of the particles. Comparing Eqs. (13) and (14), and remembering that $`k=G\tau ^2/c^2`$, we obtain $$\frac{ds}{dv}=\frac{\tau }{c}\frac{ds}{dt},$$ $`(15)`$ thus $$\frac{dv}{dt}=\frac{c}{\tau }.$$ $`(16)`$ Accordingly we see that the particle experiences an acceleration $`a_0=c/\tau =cH_0`$ directed outward when the motion is circular. This extra term is not included or derivable from Einstein’s general relativity theory and it appears only in ours. ## 7 Effective Potential The motion of a particle in a central field is best described in terms of an “effective potential”, $`V_{eff}`$. In Newtonian mechanics this is given by $$V_{eff}=\frac{GM}{r}+\frac{L^2}{2r^2},$$ $`(17)`$ where $`L`$ is the angular momentum per mass unit. In our case the effective potential is $$V_{eff}\left(r\right)=\frac{GM}{r}+\frac{L^2}{2r^2}+a_0r.$$ $`(18)`$ The circular motion is obtained at the minimal value of (18), i.e. $$\frac{dV_{eff}}{dr}=\frac{GM}{r^2}\frac{L^2}{r^3}+a_0=0,$$ $`(19)`$ with $`L=v_cr`$, and $`v_c`$ is the rotational velocity. This gives $$v_c^2=\frac{GM}{r}+a_0r,$$ $`(20)`$ thus $$v_c^4=\left(\frac{GM}{r}\right)^2+2GMa_0+a_0^2r^2,$$ $`(21)`$ where $`a_0=c/\tau =cH_0`$. ## 8 Concluding Remarks The first term on the right-hand side of Eq. (21) is purely Newtonian which cannot be evaded by any reasonable theory. The second one is the Tully-Fisher term. The third term is extremely small at the range of distances of stars around a galaxy. It has been shown that a term of the form $`GMa_0=GMcH_0`$ can explain most of the observations of the dynamics of stars around the galaxies. The “modified Newtonian law of motion” proposed in Refs. 6 and 7 was by adding arbitrarily an attractive force term in the very far distances. We have seen, however, that our theory predicts a repulsive force term rather than an attractive one. In conclusion it appears to us that one should have great doubts about the necessity and existence of halo dark matter around galaxies. Rather, this is a property of spacetime . ## References 1. See, for example, M. Carmeli, Classical Fields: General Relativity and Gauge Theory (Wiley, New York, 1982). 2. M. Carmeli, Commun. Theor. Phys. 5, 159 (1996). 3. M. Carmeli, Commun. Theor. Phys. 6, 45 (1997). 4. M. Carmeli, Commun. Math. Theor. Phys. 1, 50 (1998). 5. B.C. Whitemore, Rotation curves of spiral galaxies in clusters, in: Galactic Models, Eds. J.R. Buchler, S.T. Gottesman and J.H. Hunter, Jr. (Ann. New York Academy Sciences, Vol. 596, New York, 1990). 6. M. Milgrom, Astrophys. J 270, 365, 371, 384 (1983). 7. R.H. Sanders, Astron. Astrophys. Rev. 2, 1 (1990). 8. A. Einstein and J. Grommer, Preuss. Akad. Wiss., Phys.-Math. Klasse 1, 2 (1927). 9. A. Einstein, L. Infeld and B. Hoffmann, Annals of Mathematics 39, 65 (1938). 10. A. Einstein and L. Infeld, Can. J. Math. 1, 209 (1949). 11. L. Infeld and A. Schild, Revs. Mod. Phys. 21, 408 (1949). 12. L. Infeld, Revs. Mod. Phys. 29, 398 (1957). 13. V. Fock, Revs. Mod. Phys. 29, 325 (1957). 14. V. Fock, The Theory if Space, Time and Gravitation (Pergamen Press, Oxford, 1959). 15. L. Infeld and J. Plebanski, Motion and Relativity (Pergamen Press, Oxford, 1960). 16. B. Bertotti and J. Plebanski, Ann. Phys. (N.Y.) 11, 169 (1960). 17. M. Carmeli, Phys. Lett. 9, 132 (1964). 18. M. Carmeli, Phys. Lett. 11, 24 (1964). 19. M. Carmeli, Ann. Phys. (N.Y.) 30, 168 (1964). 20. M. Carmeli, Phys. Rev. 138, B1003 (1965). 21. M. Carmeli, Nuovo Cimento 37, 842 (1965). 22. M. Carmeli, Ann. Phys. (N.Y.) 34, 465 (1965). 23. M. Carmeli, Ann. Phys. (N.Y.) 35, 250 (1965). 24. M. Carmeli, Phys. Rev. 140, B1441 (1965). 25. T. Damour, Gravitational radiation and the motion of compact bodies, in Gravitational Radiation, N. Deruelle and T. Piran, Eds. (North-Holland, Amsterdam 1983), pp. 59-144. 26. H. Goldstein, Classical Mechanics (Addison-Wesley, Reading, Mass., 1980). 27. M. Carmeli, Inter. J. Theor. Phys. 37, 2621 (1998). 28. M. Carmeli, Commun. Math. Theor. Phys. 1, 54 (1998).
no-problem/9907/gr-qc9907091.html	ar5iv	text	# 1 Introduction ## 1 Introduction Static solutions to the Einstein Yang-Mills (EYM) equations differ considerably depending on the value of the cosmological constant. The solutions can be separated into two families; $`\mathrm{\Lambda }0`$, and $`\mathrm{\Lambda }<0`$. The solutions where $`\mathrm{\Lambda }=0`$ were discovered by Bartnik and McKinnon (BK) and their asymptotically de Sitter (dS) analogs ($`\mathrm{\Lambda }>0`$) were discovered independently by Volkov et. al. and Torii et. al. . The BK solutions and the cosmological extensions to them all share similar behavior and have been studied in detail (see Ref. for a review). Recently, asymptotically anti-de Sitter ($`\mathrm{\Lambda }<0`$) black hole solutions and soliton solutions were found which are strikingly different than the BK type solutions. In particular, there exist asymptotically anti-de Sitter (AdS) solitons with no nodes in the field strength which are stable. Furthermore, in asymptotically AdS space dyon solutions are allowed. ## 2 General formalism Given the spherically symmetric metric in the Schwarzschild gauge $$ds^2=\frac{Hdt^2}{p^2}+\frac{dr^2}{H}+r^2(d\theta ^2+\mathrm{sin}^2\theta d\varphi ^2),$$ (1) where $`H`$, $`p`$ are functions of $`t`$ and $`r`$, the coupled static EYM equations of motion are $`\left({\displaystyle \frac{H}{p}}w^{}\right)^{}`$ $`=`$ $`{\displaystyle \frac{p}{H}}u^2w{\displaystyle \frac{w}{p}}{\displaystyle \frac{(1w^2)}{r^2}}`$ (2) $`\left(r^2pu^{}\right)^{}`$ $`=`$ $`{\displaystyle \frac{2p}{H}}w^2u`$ (3) $`p^{}`$ $`=`$ $`{\displaystyle \frac{2v}{r}}p\left[(w^{})^2+{\displaystyle \frac{u^2w^2p^2}{H^2}}\right]`$ (4) $`m^{}`$ $`=`$ $`v\left[{\displaystyle \frac{(w^21)^2}{2r^2}}+{\displaystyle \frac{1}{2}}r^2p^2(u^{})^2+H(w^{})^2+{\displaystyle \frac{u^2w^2p^2}{H}}\right],`$ (5) where $`w(r)`$ and $`u(r)`$ are the magnetic and electric components of the $`SU(2)`$ Yang-Mills fields , $`H(r)=12m(r)/r\mathrm{\Lambda }r^2/3`$, $`v=G/4\pi e^2`$ and $`\mathrm{\Lambda }`$ is the cosmological constant. We require that solutions to Eq.’s (2) to (5) are regular everywhere and have finite ADM mass $`M=m(\mathrm{})`$. The electric and magnetic charges of solutions are given by $$\left(\begin{array}{c}Q_E\\ Q_M\end{array}\right)=\frac{e}{4\pi }𝑑S_k\sqrt{g}\left(\begin{array}{c}F^{k0}\\ \stackrel{~}{F}^{k0}\end{array}\right)=\left(\begin{array}{c}u_1p_0\\ 1w_0^2\end{array}\right)\frac{\tau _3}{2}$$ (6) where $`w=w_0+w_1/r+\mathrm{O}(r^2)`$ etc. For solutions in asymptotically flat or dS space, $`w(r)`$ has at least one node . The situation is quite different in asymptotically AdS space ($`\mathrm{\Lambda }<0`$). $`H(r)`$ is positive everywhere and there are solutions where $`w`$ has no node. ## 3 Monopole solutions The solutions to Eq.’s (2) to (5), for $`\mathrm{\Lambda }<0`$, were evaluated numerically using the shooting method. In the shooting method one solves Eqs. (2) to (5) at $`r=0`$ in terms of two parameters, $`a`$ and $`b`$, and ‘shoots’ for solutions with the desired asymptotic behavior. $`a`$ and $`b`$ are adjustable parameters which together specify the boundary conditions at the origin for $`w`$, $`u`$, $`H`$, and $`p`$: $`u(r)=ar+\mathrm{}`$ and $`w(r)=1br^2+\mathrm{}`$ near $`r=0`$. Purely magnetic solutions (monopoles) are found by setting $`a=0`$, corresponding to $`u(r)=0`$. A continuum of monopole solutions were obtained by varying the parameter $`b`$. The solutions are similar to the black hole solutions found by Ref. , but are also regular for all $`r`$. The number of times $`w`$ crosses the axis depends on the value of the adjustable shooting parameter $`b`$. As shown in fig. 1, the behavior of $`m`$ and $`p`$ is similar to that of the asymptotically dS solutions previously considered . In contrast, there exist solutions where $`w`$ has no nodes, which are not seen in the asymptotically dS or Minkowski cases. ## 4 Dyon solutions Dyon solutions to the EYM equations for a given negative $`\mathrm{\Lambda }`$ are found if we choose the adjustable shooting parameter $`a`$ to be non-zero. Fig. 1 shows how the electric component, $`u`$, of the EYM equations starts at zero and monotonically increases to some finite value. The behavior of $`w`$, $`m`$ , $`H`$, and $`p`$ is similar to the monopole solutions. Again we find a continuum of solutions for a continuous set of parameters $`a`$ and $`b`$, where $`w`$ crosses the axis an arbitrary number of times depending these parameters. Also similar to the monopole solutions is the existence of solutions where $`w`$ does not cross the axis. This is in sharp contrast to the $`\mathrm{\Lambda }0`$ case where the dyon solutions are forbidden. ## 5 Stability of the monopole solutions The BK solutions and the dS-EYM solutions are unstable . In contrast, the AdS black hole solutions with $`u=0`$ and the monopole solutions without nodes are stable against spherically symmetric linear fluctuations. In order to derive the time dependent EYM equations, we use the most general expression for the spherically symmetric SU(2) gauge fields in the singular gauge: $$A=\frac{1}{2e}\left\{u\tau _3dt+\nu \tau _3dr+(w\tau _1+\stackrel{~}{w}\tau _2)d\theta +(\mathrm{cot}\theta \tau _3+w\tau _2\stackrel{~}{w}\tau _1)\mathrm{sin}\theta d\varphi \right\},$$ (7) where $`\tau _i`$ ($`i=1`$,$`2`$,$`3`$) are the usual Pauli matrices and $`u`$, $`\nu `$, $`w`$ and $`\stackrel{~}{w}`$ depend on $`r`$ and $`t`$. The boundary conditions $`u=\nu =0`$ and $`w^2+\stackrel{~}{w}^2=1`$ at $`r=0`$ ensure regularity at the origin. Linearized equations when $`u(t,r)=0`$, for the gauge fields $`\delta w(t,r)`$, $`\delta \stackrel{~}{w}(t,r)`$, $`\delta \nu (t,r)`$, $`\delta p(t,r)`$, and $`\delta H(t,r)`$ have been derived in the literature . Fluctuations decouple in terms of $`\delta w(t,r)`$, $`\delta p(t,r)`$, and $`\delta H(t,r)`$ which form even-parity perturbations, and in terms of $`\delta \stackrel{~}{w}(t,r)`$ and $`\delta \nu (t,r)`$ which form odd-parity perturbations. The equation for parity-odd perturbations in $`\beta =r^2p\delta \nu /w`$, where $`\beta (t,r)=e^{i\omega t}\beta (r)`$, is $$\left\{\frac{d^2}{d\rho ^2}+U_\beta (\rho )\right\}\beta =\omega ^2\beta ,U_\beta =\frac{H}{r^2p^2}(1+w^2)+\frac{2}{w^2}\left(\frac{dw}{d\rho }\right)^2$$ (8) where $`d\rho /dr=p/H`$. Volkov et al. showed that for the BK solutions there appear exactly $`n`$ negative eigenmodes ($`\omega ^2<0`$) if $`w`$ has $`n`$ nodes . Their argument applies to the asymptotically AdS case without modification. One concludes that the solutions with nodes in $`w`$ are unstable against parity-odd perturbations. For parity-even perturbations, where $`\delta w(t,r)=e^{i\omega t}\delta w(r)`$, we find the equation $$\left\{\frac{d^2}{d\rho ^2}+U_w(\rho )\right\}\delta w=\omega ^2\delta w,U_w=\frac{H(3w^21)}{p^2r^2}+4v\frac{d}{d\rho }\left(\frac{Hw^2}{pr}\right).$$ (9) Although the potential $`U_w(\rho )`$ is not positive definite, it is regular in the entire range $`0\rho \rho _{\mathrm{max}}`$. The first term in $`U_w`$ becomes negative for $`w^2<1/3`$. The Schrödinger equation (9) was solved numerically . The potential for the solutions with $`(a,b)=(0,0.001)`$, which has no node in $`w`$, has the lowest eigenvalue $`\omega ^2`$ of 0.028 and for $`(a,b)=(0,0.005)`$, which has one node, has the lowest eigenvalue of 0.023. Therefore, these solutions are stable against parity-even perturbations, even if $`w`$ has one node, and differ from the solutions where $`\mathrm{\Lambda }0`$ where the parity-even perturbations are always unstable. ## 6 Conclusion A continuum of new monopole and dyon solutions to the asymptotically AdS EYM equations have been found. In fig. 2 the spectrum of monopole solutions is plotted. The monopole solutions in which the magnetic component $`w(r)`$ of the gauge fields never vanishes, corresponding to the portion with $`n=0`$ in fig. 2, were shown to be stable against linear perturbations. Solutions with $`b0.7`$ develop an event horizon, becoming black holes. The end point of the $`n=3`$ portion of the curve in fig. 2 shows where the solutions become black holes. Acknowledgments This work was supported in part by the U.S. Department of Energy under contracts DE-FG02-94ER-40823 and DE-FG02-87ER40328.
no-problem/9907/chao-dyn9907028.html	ar5iv	text	# REFERENCES Comment on ”The dead zone for string players” In a recent Letter Broomfield and Leask experimentally found that ”It is not possible to produce sound from a string when it is bowed at its midpoint.” This finding is trivial if one knows the following mechanism of bowing: It is well known that if one puts a horse-tail bow hair under a microscope, there are many ’small nails’ on its surface. It is these ’nails’ which continuously pluck the string as the bow moves that produce sound. This is different from what happens when one plucks a string with one’s finger. In the latter case the string can vibrate freely once the finger has left the string, while in the former case the fingers constantly intervenes the vibration. The nail spacing is so close, perhaps it is better to describe these ’nails’ as ’brushes’ or ’combs’. This kind of vibration is therefore better described as a ’forced’ or ’kicked’ vibration. Once one nail has passed over the string, the string can only vibrate freely for a short period before the next nail arrives. It is therefore better for the amplitude of vibration at the bowing point to be about the same as the spacing of the nails so that the intervention of the nails is minimized. If the amplitude is larger than the spacing of the nails, the string does not manage to vibrate much before the next nail arrives. There are already mass-and-spring models for bowing of string instruments in the literature. Generally, they consider bowing as a macroscopic frictional effect. These friction-like models will have difficulty to explain the silence of bowing in the middle of the string. However, with the microscopic mechanism proposed here the phenomenon is clear. This mechanism applies to every part of the string, but the influence of the next nail becomes most significant as one bowes at the midpoint. Furthermore, the mechanism has nothing to do with the resin used. Resin increases the friction, or the length and the number of the nails. A clean bow without any resin can still produce sound. More refined models can be developed based on the reasoning given above. It will be interesting to find out experimentally what kind of modes are excited at different bowing conditions. Remember that the string is not a rigid body and the nails do not always pluck the string in a direction normal to it. Therefore, the generation of the vibration and the interference of the next nail is always imperfect. In the field of nonlinear dynamics, phenomena of ’kicked oscillators’, which display interesting chaotic behaviour, have been studied. It will also be interesting to see how they apply to the bowing of violins. Julian Juhi-Lian Ting jlting@multimania.com Date:
no-problem/9907/hep-ph9907530.html	ar5iv	text	# 1 Introduction ## 1 Introduction The Minimal Supersymmetric Standard Model (MSSM) is one of the promising extension of Standard Model. If the nature picks up the low energy supersymmetry(SUSY), MSSM will be proven for sure, as superpartners will be copiously produced at future colliders such as Large Hadron Collider (LHC) at CERN or TeV scale $`e^+e^{}`$ linear colliders (LC) proposed by DESY, KEK, and SLAC. The symmetry also offers natural solution of the hierarchy problem, amazing gauge coupling unification, and dark matter candidates. On the other hand, the MSSM suffers sever flavor changing neutral current (FCNC) constraints if no mass relation is imposed on sfermion mass parameters. Various proposals have been made of the mechanism to incorporate the SUSY breaking to “our sector”, trying to offer the natural explanation of such mass relations. In short, it would be very surprising if sparticles are found in any future collider— The discovery is not the goal, but it is the beginning of a new quest of “the mechanism” of SUSY breaking. Measurements of soft breaking masses would be an important aspects of the SUSY study at future colliders, because different SUSY breaking mechanism predict different sparticle mass patterns. In this talk, I will review attempts to measure soft breaking parameters at LHC and LC’s. In section 2 and 3, I will concentrate on the process that $`\stackrel{~}{g}`$ and $`\stackrel{~}{q}`$ are produced and decay, involving the leptonic second lightest neutralino decay $`\stackrel{~}{\chi }_2^0`$ $`\chi _1^0l^+l^{}`$. The decay either proceeds through virtual exchanges of $`Z^0`$ and $`\stackrel{~}{l}`$ or direct two body decays such as $`\stackrel{~}{\chi }_2^0\stackrel{~}{l}l`$ and $`\stackrel{~}{l}l\stackrel{~}{\chi }_1^0`$. Events near the end point of the $`m_{ll}`$ distribution of the three body decay play a key role to reconstruct the kinematics of $`\stackrel{~}{g}`$ and $`\stackrel{~}{q}`$ cascade decay chain, and minimal supergravity parameters is determined precisely. In section 3.1, I point out the three body decay distribution depends strongly on the decay matrix element. This dependence may reduce the $`m_{ll}`$ end point resolution, while the whole shape of $`m_{ll}`$ distribution could provide information on slepton masses and ino mixings. I also show a new analysis for the case where $`\stackrel{~}{\chi }_2^0`$ decays dominantly into $`\stackrel{~}{l}l`$. We point out that the peak position of $`p_T`$ asymmetry of the same flavor opposite sign (OS) lepton pairs in $`m_{ll}m_{ll}^{max}`$ region would be independent of $`\stackrel{~}{\chi }_2^0`$ momentum distribution, therefore may be used to constraint $`m_{\stackrel{~}{\chi }_2^0}`$, $`m_{\stackrel{~}{\chi }_1^0}`$ and $`m_{\stackrel{~}{l}}`$ directly such as the end point of the $`m_{ll}`$ distribution. In section 4, I discuss precision study at future LC’s. Thanks to low backgrounds at polarized $`e^+e^{}`$ collider, the machine is perfect to discover and study the superparticles if they are in kinematical reach. Furthermore it offers clean tests of relations of soft mass parameters and couplings. I discuss the radiative correction to the SUSY coupling relations which can be probed precisely at LC. The measurement of the deviation from the SUSY tree level relation offers a way to determine the squark mass scale in the “decoupling scenario”. ## 2 Supersymmetry and LHC Squarks ($`\stackrel{~}{q}`$) and gluinos ($`\stackrel{~}{g}`$) will be copiously produced at LHC, and they subsequently decay into charginos ($`\stackrel{~}{\chi }_i^\pm `$) or neutralinos ($`\stackrel{~}{\chi }_i^0`$). They could further decay into sleptons ($`\stackrel{~}{l}`$). The signal of the sparticle production will be leptons and/or jets with missing $`p_T`$ if LSP is stable<sup></sup><sup></sup>*I will concentrate on the MSUGRA motivated scenario where $`\stackrel{~}{\chi }_1^0`$ is the LSP.. Various study indicates that LHC will find the excess of the sparticle signal if $`m_{\stackrel{~}{q}},m_{\stackrel{~}{g}}<2`$ TeV in MSUGRA scenarios. The question is then if we could understand the nature of sparticles in detail. MSSM contains many parameters, on the other hand, the observed signal distributions are sum of products of production cross sections, branching ratios, and acceptances. The substantial complexities may prevent simple and model independent interpretations. However some kinematical quantity can be extracted model independently by investigating some characteristic decay distributions. One of impressive examples is the case studied for Snowmass ’96, so called “LHC point 3”. It is a case that the production of gluino followed by $`\stackrel{~}{g}\stackrel{~}{b}b`$, $`\stackrel{~}{b}b\stackrel{~}{\chi }_2^0`$ occurs with substantial branching fraction. The leptonic decay of the second lightest neutralino $`\stackrel{~}{\chi }_2^0\stackrel{~}{\chi }_1^0l^+l^{}`$ occurs with branching fraction of 16%. The number of $`b\overline{b}b\overline{b}l^+l^{}+2`$ jet events then would be around 2.3 M for one year low luminosity run with S/N ratio about 10:1; This is substantially larger production ratio compared to typical s-channel sparticle production at LC. The end point of $`m_{ll}`$ distribution of OS dileptons would be identified as $`m_{\stackrel{~}{\chi }_2^0}`$ $`m_{\stackrel{~}{\chi }_1^0}`$. The end point could be measured within 50 MeV error. For point 3, $`O(10^5)`$ events near the $`m_{ll}`$ end point could be selected for further analysis. In the limit where $`m_{ll}m_{ll}^{\mathrm{max}}`$, $`\stackrel{~}{\chi }_1^0`$ is stopped in the rest frame of $`\chi _2^0`$, therefore $`\stackrel{}{\beta }_{\stackrel{~}{\chi }_2^0}\stackrel{}{\beta }_{\stackrel{~}{\chi }_1^0}`$ in the laboratory frame. Assuming further an approximate MSUGRA relation $`m_{\stackrel{~}{\chi }_2^0}`$ $`=2m_{\stackrel{~}{\chi }_2^0}`$, one would reconstruct $`m_{\stackrel{~}{b}}`$ and $`m_{\stackrel{~}{g}}`$ through the invariant mass distribution of all possible combination of bottom jets and $`\stackrel{~}{\chi }_2^0`$ momentum. This leads to the resolution of MSUGRA parameters $`m_0=200_8^{+13}`$ GeV and $`M=100\pm 0.7`$ GeV. The above analysis showxs that the event distribution (which in principle depends on hundreds of parameters of MSSM model) could be factrized into a few distributions which sensitively reflects a few parameters of the model. The rest of the distributions will be understood better with the constraints. To this end, we may be able to provide enough cross checks between events and theoretical calculations (or MC simulations), so that we would be able to use event rates and whole distributions to determine model parameters precisely, or even reject some SUSY breaking scenarios. ## 3 Neutralino decay into leptons In the previous section, we find the invariant mass distribution of OS lepton pairs from $`\stackrel{~}{\chi }_2^0`$ decay is the important part of the analysis. In this section we concentrate on some new aspects on the nature of the decay distribution and discuss the constraints to MSSM parameters that would be obtained from the distribution measurement. ### 3.1 Three body decay into $`\stackrel{~}{\chi }_1^0l^{}l^+`$ and the decay matrix element Three body decays of $`\stackrel{~}{\chi }_2^0`$ are dominant as long as two body decays such as $`\stackrel{~}{\chi }_2^0\stackrel{~}{\chi }_1^0Z^0`$, $`\stackrel{~}{\chi }_1^0H`$, $`\stackrel{~}{\chi }_2^0\stackrel{~}{l}l`$ are not open. The branching ratio of the three body leptonic decay of the second lightest neutralino, $`\stackrel{~}{\chi }_2^0\stackrel{~}{\chi }_1^0l^+l^{}`$, is known to be very sensitive to the values of the underlying MSSM parameters. The dependence is enhanced by the negative interference between the decay amplitude from $`Z^0`$ exchange and that from slepton exchange. In this subsection we show that the effect of the interference appears not only in the branching ratios, but also in the decay distributions, such as the distribution of the invariant mass $`m_{ll}`$ of the lepton pairs. The partial decay width is given by $$\frac{d\mathrm{\Gamma }}{dxdy}(\stackrel{~}{\chi }_A^0\stackrel{~}{\chi }_B^0\overline{f}f)=\frac{N_C}{256\pi ^3}m_{\stackrel{~}{\chi }_A^0}\|\|^2(x,y,z=1+r_{\stackrel{~}{\chi }_B}^2xy).$$ (1) The range of ($`x`$, $`y`$) is given by the conditions $`z(xyr_{\stackrel{~}{\chi }_B}^2)0,`$ $`r_{\stackrel{~}{\chi }_B}^2x1,`$ $`r_{\stackrel{~}{\chi }_B}^2y1,`$ $`x+y+z=1+r_{\stackrel{~}{\chi }_B}^2,`$ (2) when $`m_f=0`$. In the phase space of the decay $`\stackrel{~}{\chi }_2^0\stackrel{~}{\chi }_1^0l^+l^{}`$, the $`Z^0`$ exchange amplitude and $`\stackrel{~}{l}`$ exchange amplitude behave differently. When the $`Z^0`$ contribution dominates, distributions are enhanced in the region of large $`m_{ll}^2=zm_{\stackrel{~}{\chi }_2^0}^2`$. In contrast, when the $`\stackrel{~}{l}`$ exchange contribution dominates, distributions are enhanced in regions with large $`x`$ and/or large $`y`$, therefore in small $`m_{ll}`$ and large $`\|E_l^{}^{\mathrm{rest}}E_{l^+}^{\mathrm{rest}}\|`$ region. Note that $`E_{l^+}^{\mathrm{rest}}=`$ $`(1x)m_{\stackrel{~}{\chi }_2^0}/2`$ and $`E_l^{}^{\mathrm{rest}}=`$ $`(1y)m_{\stackrel{~}{\chi }_2^0}/2`$ are lepton energies in the $`\stackrel{~}{\chi }_2^0`$ rest frame. We consider the case where $`2M_1M_2\|\mu \|`$, a typical case in MSUGRA model. In this case, $`\stackrel{~}{\chi }_2^0`$ is Wino-like and $`\stackrel{~}{\chi }_1^0`$ is Bino-like. An interesting property in this case is that the $`Z^0`$ and $`\stackrel{~}{l}`$ amplitudes could be of comparable size in some region of phase space. Furthermore, their interference is generally destructive for leptonic decays. These effects cause complicated situations, which we discuss below. For illustration, we use two sets of parameters for the neutralino sector, (A) and (B), shown in Table 1. These values are fixed to give the same masses for three inos, ($`m_{\stackrel{~}{\chi }_1^0}`$, $`m_{\stackrel{~}{\chi }_2^0}`$, $`m_{\stackrel{~}{\chi }_2^+}`$) $`=`$ (71.4, 140.1, 320.6) GeV. For calculating the branching ratios, we take generation-independent slepton masses and a universal soft SUSY breaking squark mass $`m_{\stackrel{~}{Q}}=500`$ GeV. In Fig. 1a, we show the $`m_{ll}`$ distribution of the decay $`\stackrel{~}{\chi }_2^0l^+l^{}\stackrel{~}{\chi }_1^0`$ for parameter set (A) and varying $`m_{\stackrel{~}{l}}`$ from 170 GeV to 500 GeV. Because $`m_{\stackrel{~}{\chi }_2^0}`$ and $`m_{\stackrel{~}{\chi }_1^0}`$ are fixed, the end points of the distributions $`m_{ll}^{\mathrm{max}}=68.7`$ GeV are same for each curve, while the shape of the distribution changes drastically with slepton mass. For a slepton mass of 170 GeV (thick solid line), the decay proceeds dominantly through slepton exchanges, therefore the $`m_{ll}`$ distribution is suppressed near $`m_{ll}^{\mathrm{max}}`$. On the other hand, once slepton exchange is suppressed by its mass, $`Z^0`$ exchange dominates and the distribution peaks sharply near $`m_{ll}^{\mathrm{max}}`$. In Fig. 1b, we show an example for $`\mu >0`$, parameter set (B). The dependence on the slepton mass is different from the previous case. As $`m_{\stackrel{~}{l}}`$ increases from 170 GeV, $`m_{ll}`$ distribution becomes softer. For $`m_{\stackrel{~}{l}}>250`$ GeV, a second peak appears due to strong cancellation of $`Z^0`$ exchange and slepton exchange contributions for a certain value of $`m_{ll}`$. At the same time, the branching ratio reaches its minimum at $`m_{\stackrel{~}{l}}300`$ GeV, much less than 1%. Notably, one can find slepton masses where a complete cancellation occurs very close to the end point $`m_{ll}^{\mathrm{max}}`$ of the $`m_{ll}`$ distribution. The thick dashed line shows distribution for $`m_{\stackrel{~}{l}_L}=245`$ GeV and $`m_{\stackrel{~}{l}_R}=204`$ GeV. Events near the end point ($`m_{ll}^{\mathrm{max}}m_{ll}<4`$ GeV) becomes too few, and it is very hard to observe the real end point for this case. As it has discussed already, the lepton invariant mass distribution is an important tool for studying MSSM at hadron colliders. In previous studies, the end point of $`m_{ll}`$ distribution is treated as ambiguous measurement of $`m_{\stackrel{~}{\chi }_2^0}m_{\stackrel{~}{\chi }_1^0}`$, and the end point samples are used for further analysis. The slepton mass dependence of $`\stackrel{~}{\chi }_2^0`$ decay distribution shown in Fig. 1 suggests that not only the end point of the distributions but also the distributions themselves contain information about the underlying parameters such as $`m_{\stackrel{~}{l}}`$. The negative side of this is that the fitted end point may depend on the assumed values of these parameters, introducing additional systematic errors to the fit. For an extreme case shown in Fig. 2, the observed end point of the lepton invariant mass distribution does not coincide with $`m_{\stackrel{~}{\chi }_2^0}m_{\stackrel{~}{\chi }_1^0}`$. Note that realistic simulations including the parameter dependence of the decay distribution were not available for hadron colliders until recently. The most recent ISAJET release ($`>`$ISAJET 7.43) allows to simulate the effect of exact matrix elements for all three body decay distributions. The $`m_{ll}`$ distribution may be used to extract the underlying MSSM parameters. The distribution depends strongly on $`m_{\stackrel{~}{l}}`$, and also on $`\stackrel{~}{\chi }_2^0l\stackrel{~}{l}`$ and $`\stackrel{~}{\chi }_2^0\stackrel{~}{\chi }_1^0Z^0`$ couplings. The $`Z^0`$ coupling is proportional to the Higgsino components of $`\stackrel{~}{\chi }_2^0`$ and $`\stackrel{~}{\chi }_1^0`$. Because we take these neutralinos to be gaugino-like, their Higgsino components depends on $`\mathrm{tan}\beta `$ and the Higgsino mass parameter $`\mu `$, and $`\|\mu \|`$ is roughly equal to $`m_{\stackrel{~}{\chi }_2^+}`$. Therefore the $`m_{ll}`$ distribution gives at least one constraint on $`\mu `$, $`\mathrm{tan}\beta `$ and $`m_{\stackrel{~}{l}}`$ in addition to the well known constraint on $`m_{\stackrel{~}{\chi }_2^0}m_{\stackrel{~}{\chi }_1^0}`$. We estimate the sensitivity, assuming that backgrounds can be neglected or subtracted, and dependence of acceptance on $`m_{ll}`$ can be corrected. We define the sensitivity function $`𝒮`$ as follows: $$𝒮=\sqrt{\underset{i}{}\left(n_i^{\mathrm{fit}}n_i^{\mathrm{input}}\right)^2/n_i^{\mathrm{input}}}.$$ (3) Here $`n_i^{\mathrm{fit}}`$ ($`n_i^{\mathrm{input}}`$) is the number of events in the $`i`$-th bin of the $`m_{ll}`$ distribution for the MSSM parameters $`(M_1,M_2,\mu ,\mathrm{tan}\beta ,m_{\stackrel{~}{l}_{L,R}})\|_{\mathrm{fit}(\mathrm{input})}`$. We normalize $`_in_i^{\mathrm{fit}}`$ and $`_in_i^{\mathrm{input}}`$ to some number $`N`$. $`𝒮`$ gives the deviation of the input distribution $`n_i^{\mathrm{input}}`$ from the distribution for the fit ($`n_i^{\mathrm{fit}}`$) in units of standard deviations. We take $`N=2500`$ and an $`m_{ll}`$ bin size of 2 GeV. In Fig. 2, we show contours of constant $`𝒮=1,2,3,4`$ (corresponding to 1$`\sigma `$, 2$`\sigma `$, 3$`\sigma `$, 4$`\sigma `$ for N=2500) in the ($`\mathrm{tan}\beta ^{\mathrm{fit}}`$, $`m_{\stackrel{~}{l}}^{\mathrm{fit}}`$) plane. For the solid lines, we take parameter set (A) and $`m_{\stackrel{~}{l}}=`$ 250 GeV as input parameters, while for fitting parameters we vary $`\mathrm{tan}\beta `$ and $`m_{\stackrel{~}{l}_L}=m_{\stackrel{~}{l}_R}`$, fixing ($`M_1^{\mathrm{fit}}`$, $`M_2^{\mathrm{fit}}`$, $`\mu ^{\mathrm{fit}}`$) to reproduce the input values of ($`m_{\stackrel{~}{\chi }_1^0}`$, $`m_{\stackrel{~}{\chi }_2^0}`$, $`m_{\stackrel{~}{\chi }_2^+}`$). The resulting contours (solid lines) correspond to the sensitivity of the $`m_{ll}`$ distribution to $`m_{\stackrel{~}{l}}`$ and $`\mathrm{tan}\beta `$ when the three ino masses are known. In the figure, a strong upper bound on the slepton masses emerges, $`m_{\stackrel{~}{l}}<260`$ GeV is obtained if $`𝒮<1`$ is required. This is consistent with the large change of the distribution between $`m_{\stackrel{~}{l}}=270`$ GeV and $`m_{\stackrel{~}{l}}=500`$ GeV found in Fig. 1. The $`m_{ll}`$ distribution also constrains $`\mathrm{tan}\beta `$ mildly. The constraint is not very strong due to our choice of parameters $`\|\mu \|M_2`$; gaugino-Higgsino mixing is suppressed in this case. ### 3.2 The two body decay into $`\stackrel{~}{l}l`$ and lepton $`p_T`$ asymmetry We now discuss the case where $`m_{\stackrel{~}{\chi }_2^0}>m_{\stackrel{~}{l}}`$. The decay could proceed through two body decays $`\stackrel{~}{\chi }_2^0`$ $``$ $`\stackrel{~}{l}^+l^{},\stackrel{~}{l}^{}l^+,`$ (4) $`\stackrel{~}{l}`$ $``$ $`\stackrel{~}{\chi }_1^0l.`$ (5) Because of the phase space factor, the decay could dominate over the three body decay $`\stackrel{~}{\chi }_2^0\stackrel{~}{\chi }_1^0f\overline{f}`$. Note that the decay distribution is now completely fixed by the two body kinematics. The $`m_{ll}`$ distribution of the two leptons from the $`\stackrel{~}{\chi }_2^0`$ cascade decay is $$\frac{1}{\mathrm{\Gamma }}\frac{d\mathrm{\Gamma }}{dm_{ll}}=\frac{2m_{ll}}{(m_{ll}^{\mathrm{max}})^2}.$$ (6) Here the end point of the $`m_{ll}`$ distribution, $`m_{ll}^{\mathrm{max}}`$, is expressed as $$m_{ll}^{\mathrm{max}}=\frac{\sqrt{(m_{\stackrel{~}{\chi }_2^0}^2m_{\stackrel{~}{l}}^2)(m_{\stackrel{~}{l}}^2m_{\stackrel{~}{\chi }_1^0}^2)}}{m_{\stackrel{~}{l}}}.$$ (7) In addition to that, it has been known that the the lepton $`p_T`$ asymmetry $`A_T(p_{T2}/p_{T1})`$ distribution (where $`p_{T1}>p_{T2}`$) is sensitive to slepton masses. The $`A_T`$ can distribute off from 1 when it originates from $`\stackrel{~}{\chi }_2^0`$ and sometimes strongly peaks. (On the other hand, $`A_T`$ distribution of the three body decay peaks at 1.) The asymmetry comes from the monochromatic nature of lepton energy from the $`\stackrel{~}{\chi }_2^0`$ decay in the rest frame. For example, when the mass difference between $`\stackrel{~}{\chi }_2^0`$ and $`\stackrel{~}{l}`$ is small, the lepton energy from the $`\stackrel{~}{\chi }_2^0`$ decay is substantially smaller than that from $`\stackrel{~}{l}`$ decay. The nature of the lepton $`p_T`$ asymmetry is then qualitatively understand, because the lepton with high (low) energy in the $`\stackrel{~}{\chi }_2^0`$ rest frame has better chance to get high (low) $`p_T`$ in the laboratory frame. The purpose of this subsection is to improve this qualitative nature to quantitative one. We note that two leptons go exactly the same direction if $`m_{ll}=0`$. In the limit, the ratio of the energies of the lepton antilepton pair is unchanged even if $`\stackrel{~}{\chi }_2^0`$ is boosted; $`A_T`$ at $`m_{ll}=0`$, $`A_T^0`$, would be estimated by the lepton energy ratio $`A_E(E_{l2}/E_{l1})`$ at the $`\stackrel{~}{\chi }_2^0`$ rest frame at $`m_{ll}=0`$, $`A_E^0`$, as, $`A_T^0A_E^0E_{l2}/E_{l1}\|_{m_{ll}0}`$ $`=`$ $`{\displaystyle \frac{m_{\stackrel{~}{l}}^2m_{\stackrel{~}{\chi }_1^0}^2}{m_{\stackrel{~}{\chi }_2^0}^2m_{\stackrel{~}{l}}^2}},(\mathrm{for}m_{\stackrel{~}{l}}^2m_{\stackrel{~}{\chi }_1^0}^2<m_{\stackrel{~}{\chi }_2^0}^2m_{\stackrel{~}{l}}^2)`$ (8) $`\mathrm{or}`$ $`{\displaystyle \frac{m_{\stackrel{~}{\chi }_2^0}^2m_{\stackrel{~}{l}}^2}{m_{\stackrel{~}{l}}^2m_{\stackrel{~}{\chi }_1^0}^2}}(\mathrm{for}m_{\stackrel{~}{l}}^2m_{\stackrel{~}{\chi }_1^0}^2>m_{\stackrel{~}{\chi }_2^0}^2m_{\stackrel{~}{l}}^2).`$ (9) Even though $`m_{ll}0`$, we see that distribution peaks at the same value; $`A_T^{\mathrm{peak}}=A_E^0`$ holds approximately. In Fig 3, we show $`A_T`$ distributions with various invariant mass cuts. Here we take the universal scaler mass $`m=100`$ GeV, the universal gaugino mass $`M=150`$ GeV, $`\mathrm{tan}\beta =2`$, $`A=0`$, and $`\mu <0`$, when $`m_{\stackrel{~}{e}_R}=120.68`$ GeV, $`m_{\stackrel{~}{\chi }_1^0}=65.15`$ GeV and $`m_{\stackrel{~}{\chi }_2^0}=135.49`$ GeV. The peak structure is quite prominent even without $`m_{ll}`$ cuts. The edge of the $`m_{ll}`$ decay distribution is 52 GeV.The MSSM parameters and cuts are same to that is taken by Iashvili and Khar-chilava. We use ISAJET7.44 and ATLFAST2.21 for simulations. We can see that the peak solely comes from $`m_{ll}m_{ll}^{max}`$ events. When we compare the distribution of the events with $`m_{ll}^2<100`$ (GeV)<sup>2</sup> (Fig.3 a) and $`m_{ll}^2<400`$ (GeV)<sup>2</sup> (Fig.3 b), we found that the event distribution is more sharply peaked at the value close to $`A_E^0=0.368`$ as $`m_{ll}`$ cut decreases. The result of non-symmetric Gaussian fit to the events near the peak for $`𝑑t=2.4fb^1`$ is summarized in the table 1. The sample for $`m_{ll}<10`$ GeV is in perfect agreement with the expected value $`A_E`$. With sufficient statistics one may be able to measure the $`m_{ll}`$ dependence of $`A_T`$ in $`m_{ll}m_{ll}^{max}`$ region. The lepton energy ratio for generic $`m_{ll}`$ would be given as $`{\displaystyle \frac{E_{l2}}{E_{l1}}}`$ $`=`$ $`{\displaystyle \frac{m_{\stackrel{~}{\chi }_2^0}^2m_{\stackrel{~}{l}}^2}{m_{\stackrel{~}{l}}^2m_{\stackrel{~}{\chi }_1^0}^2+2m_{ll}^2}}(\mathrm{for}m_{\stackrel{~}{l}}^2m_{\stackrel{~}{\chi }_1^0}^2+m_{ll}^2>m_{\stackrel{~}{\chi }_2^0}^2m_{\stackrel{~}{l}}^2)`$ (11) $`\mathrm{or}`$ $`{\displaystyle \frac{m_{\stackrel{~}{l}}^2m_{\stackrel{~}{\chi }_1^0}^2+m_{ll}^2}{m_{\stackrel{~}{\chi }_2^0}^2m_{\stackrel{~}{l}}^2}}(\mathrm{for}m_{\stackrel{~}{l}}^2m_{\stackrel{~}{\chi }_1^0}^2+m_{ll}^2<m_{\stackrel{~}{\chi }_2^0}^2m_{\stackrel{~}{l}}^2).`$ (12) According to Eq.(11), the measurement of $`m_{ll}`$ dependence of the peak position could correspond to the measurement of $`m_{\stackrel{~}{l}}^2m_{\stackrel{~}{\chi }_1^0}^2`$. In Table 2, the fitted peak value reduces as $`m_{ll}`$ cut increases. This could be due to the reduction of average $`E_{l1}/E_{l2}`$ in the $`\stackrel{~}{\chi }_2^0`$ rest frame. $`A_E=0.354`$ at $`m_{ll}=20`$ GeV for the parameters we take. The average $`A_E`$ between $`m_{ll}=0`$ and 20 GeV agrees with the peak value of $`A_T`$ distribution, though the statistics of this simulation is not sufficient to claim the deviation of the peaks between $`m_{ll}^2<100`$ (GeV)<sup>2</sup> and $`m_{ll}^2<400`$ (GeV)<sup>2</sup> samples. On the other hand, $`A_E=0.2915`$ at $`m_{ll}=m_{ll}^{\mathrm{max}}=52`$ GeV. The average $`A_E`$ between $`m_{ll}=0`$ to $`m_{ll}=m_{ll}^{\mathrm{max}}`$ is too small compared to the $`A_T^{\mathrm{peak}}`$ without $`m_{ll}`$ cut. This is consistent with the fact that no peak structure is observed for $`m_{ll}^2>1000(\mathrm{GeV})^2`$ (Fig. 3c). To see the importance of $`A_T`$ measurement, we first show the expected constraint on $`m_{\stackrel{~}{l}}`$ and $`m_{\stackrel{~}{\chi }_1^0}`$ when $`m_{\stackrel{~}{\chi }_2^0}`$ is fixed, provided that $`A_T`$ and $`m_{ll}^{\mathrm{max}}`$ is measured with the error of 0.07 and 0.5 GeV respectively (Fig.4). The error on $`m_{\stackrel{~}{l}}`$ and $`m_{\stackrel{~}{\chi }_1^0}`$ could be of the order of 1%, consistent with the previous fits. Note they did not identify the origin of the peak structure and used whole $`A_T`$ distribution for the fit. The used distribution may depend on parent squark and gluino masses, while our fit relays solely on the peak position, or only on $`m_{\stackrel{~}{\chi }_1^0}`$, $`m_{\stackrel{~}{\chi }_2^0}`$ and $`m_{\stackrel{~}{l}}`$. One may also observe the end point of $`m_{ll}`$ distribution of the three body decay in addition to the edge of the $`m_{ll}`$ distribution originated from two body cascade decay through slepton. This is because right handed slepton coupling to wino and higgsino is zero. The measurements of $`m_{ll}^{\mathrm{max}}`$ (2 body), $`A_T^0`$, and $`m_{ll}^{\mathrm{max}}`$ (3 body) $``$ $`m_{\stackrel{~}{\chi }_2^0}m_{\stackrel{~}{\chi }_1^0}`$ are potentially sufficient to determine all sparticle masses involved in the $`\stackrel{~}{\chi }_2^0`$ cascade decay. Assuming a rather optimistic error on $`m_{ll}^{\mathrm{max}}(3\mathrm{body})`$, $`\delta m_{ll}^{\mathrm{max}}(3\mathrm{body})=1`$ GeV, $`m_{\stackrel{~}{\chi }_1^0}`$, $`m_{\stackrel{~}{\chi }_2^0}`$, and $`m_{\stackrel{~}{l}}`$ are constrained within $`\pm 8`$ GeV, without assuming any relation between ino and slepton masses. The errors are substantially larger than those shown in Fig.4, due to the correlations between the constraints. On the other hand, $`m_{\stackrel{~}{e}}/m_{\stackrel{~}{\mu }}`$ ratio would be constrained strongly. Assuming $`\delta A_T<0.07`$, $`\delta m_{ee,\mu \mu }<0.5`$ GeV, $`\delta m_{ll}^{\mathrm{max}}(3\mathrm{b}\mathrm{o}\mathrm{d}\mathrm{y})=4`$ GeV, we obtain $`\delta (m_{\stackrel{~}{e}}/m_{\stackrel{~}{\mu }})=2.5`$ % for $`\mathrm{\Delta }\chi ^2<1`$, and 7% for $`\mathrm{\Delta }\chi ^2<9`$. Note that the background in $`m_{ll}m_{ll}^{max}`$ region must be studied to claim the above measurement is possible. Backgrounds from $`t\overline{t}l\overline{l}`$ could be important in low $`m_{ll}`$ region. Note that the full amplitude level study of $`W\gamma ^{}`$ production has been done for the background process of $`\stackrel{~}{\chi }_2^0`$ $`\stackrel{~}{\chi }_1^+3l`$, and large background is found in $`m_{ll}<10`$ GeV region. However it is unlikely that the background distribution has peak in $`0A_T`$ region. The peak of the signal distribution may be observed precisely on the top of such backgrounds, especially when signal rate is high enough to allow precision studies. ## 4 Precision study at LC A TeV scale linear colliders could be a powerful discovery machine. In the context of MSUGRA model, 1 TeV LC roughly corresponds to LHC in its discovery potential. This is because that in the model with universal scalar and gaugino mass at very high scale, relations $`m_{\stackrel{~}{Q}}m_{\stackrel{~}{l}}`$ and $`m_{\stackrel{~}{g}}m_{\stackrel{~}{W}}`$, $`m_{\stackrel{~}{B}}`$ are predicted naturally. In future LC’s, search modes are the production and the decay of sleptons, charginos, and neutralinos. We should also note that LC experiments cover the case where superparticles takes nasty patterns of mass spectrum. Because of the available high beam polarization, backgrounds from $`W^\pm `$ boson pair production can be highly suppressed. Notice also that its effective $`\sqrt{s}`$ is monochromatic for $`e^+e^{}`$ colliders, therefore $`\stackrel{~}{l}`$ and $`\stackrel{~}{\chi }_1^+`$ will be produced subsequently from lighter to heavier, and we can measure production cross sections and the decay distributions systematically. Systematical studies of physics potential at LC when slepton, chargino and neutralino are produced have been done in detail by several authors, and it has been shown that gaugino mass relations, slepton mass relation, and coupling relations can be confirmed with errors of $`O(1\%)`$. In this note, we concentrate on measurements of the coupling relations imposed by supersymmetry, $`g_{\stackrel{~}{B}\stackrel{~}{e}_Re_R}`$ $`=`$ $`\sqrt{2}g_B,`$ (13) $`g_{\stackrel{~}{W}\stackrel{~}{e}_Le_L}`$ $`=`$ $`\sqrt{2}g_W.`$ (14) It has been argued that this coupling relation could be measured within $`O(1\%)`$ accuracy or better by measuring sparticle production cross sections, angular distributions, and sparticle masses involved in the production process . The measurement of the couplings is important because the equivalence of the gauge coupling and gaugino coupling is ultimate probe of the supersymmetry at low energy, although the partial discovery of sparticle of course suggests the existence. Another way to say, the existence of (large) SUSY breaking sector couples to (observed) sparticles will appear as the deviation of the sparticle coupling from those predicted by the tree level symmetry. Such corrections might come from the existence of squarks which is much heavier than sleptons, charginos or neutralino. Such scenarios, with relatively light third generation squarks are occasionally quoted as “decoupling” scenarios, and attractive because they are free from large flavor changing neutral currents. When such mass spectrum is realized, SUSY coupling relations do not hold in the effective theory below $`m_{\stackrel{~}{q}}`$, and the corrections to the couplings from the tree level predictions are expressed as follows; $`\delta \left({\displaystyle \frac{g_{\stackrel{~}{B}\stackrel{~}{e}e}}{g_{\stackrel{~}{B}}^{SM}}}\right)`$ $`=`$ $`0.7\%\mathrm{log}_{10}\left({\displaystyle \frac{m_{\stackrel{~}{q}}}{m_{\stackrel{~}{l}}}}\right),`$ (15) $`\delta \left({\displaystyle \frac{g_{\stackrel{~}{W}\stackrel{~}{e}e}}{g_{\stackrel{~}{W}}^{SM}}}\right)`$ $`=`$ $`2\%\mathrm{log}_{10}\left({\displaystyle \frac{m_{\stackrel{~}{q}}}{m_{\stackrel{~}{l}}}}\right).`$ (16) If $`t`$-channel exchange of sparticles dominates over $`s`$-channel exchange of gauge bosons, the cross section can be proportional to the 4th power of the coupling, and the correction to the cross section could be around 8% when $`m_{\stackrel{~}{q}}=10m_{\stackrel{~}{l}}`$ for sneutrino and wino productions. A specific example is considered for $`\stackrel{~}{\nu }\stackrel{~}{\nu }^{}`$ production with charginos lighter than $`\stackrel{~}{\nu }`$. The production is dominated by $`t`$ channel exchange of charginos, and involve the wino-sneutrino-electron coupling. For integrated luminosity around $`100fb^1`$, the accepted number of events consisted with $`e^+e^{}`$ and some other jets or leptons activity exceeds more than $`10^4`$ events. With suitable constraint to $`\mathrm{tan}\beta `$ and heavier ino masses, $`\mathrm{log}_{10}(m_{\stackrel{~}{q}}/m_{\stackrel{~}{l}})`$ would be constrained within 0.09 (statistics) $`\pm `$ 0.08 (sneutrino mass error). Here the error from sneutrino mass uncertainty is relatively large due to $`\beta _{\stackrel{~}{\nu }}^3`$ behavior of the cross section near the sneutrino production threshold. Chargino threshold production is proportional to $`\beta _{\stackrel{~}{\chi }}`$ and the error due to the mass uncertainty might be controlled better. It would not be very surprising if $`𝒪(10^6)`$ sparticle events is accumulated in future, with sufficient understanding on underlying parameter of MSSM models. Note that different sparticles produced simultaneously, and the proposed TESLA integrated luminosity is as large as 1 ab<sup>-1</sup>= 1000fb<sup>-1</sup>. Then Does this mean that the production cross sections are measured within O(0.1%) accuracy; a few % measurement of the first generation squark masses without producing them? Apparently, measuring the number of signal events is not equivalent to the measurement of the production cross section. Measured production cross sections suffer various uncertainties, which is schematically given as $`{\displaystyle \frac{\delta \sigma }{\sigma }}`$ $`=`$ $`{\displaystyle \frac{1}{\sqrt{N_{\mathrm{accept}}}}}{\displaystyle \frac{\delta \sigma }{\delta M_i}}\delta M_i{\displaystyle \frac{\delta \sigma }{\delta m_i}}\delta m_i\mathrm{}`$ (18) $`\mathrm{luminosity}\mathrm{error}\mathrm{energy}\mathrm{resolution}\mathrm{QED},\mathrm{QCD}\mathrm{corrections}\mathrm{}`$ where $`N_{\mathrm{accept}}`$ may be expressed as $`N_{\mathrm{accept}}`$ $`=`$ $`Br(\mathrm{sparticle}\mathrm{visible}\mathrm{or}\mathrm{clean}\mathrm{mode})`$ (21) $`\times (\mathrm{acceptance})\times ({\displaystyle 𝑑t\sigma }).`$ The branching ratios may be around $`(50\%)^2`$ and the acceptance could be as high as 50 %. In the right hand side of equation (18), the first line contains errors of underlying MSSM parameters that could be negligible in the limit of infinite statistics. The second line contains machine dependent errors and potentially large and uncontrolled QED and QCD corrections. They must be very small if we want to extract $`<1\%`$ deviation of cross section, and would require the huge efforts. ## 5 Conclusion In this talk, I discussed a “precision” study of supersymmetry in future colliders, LHC and LC. The motivation of the precision study is to explore the origin of supersymmetry breaking and mechanisms to bring it to our sector. The signature must appear on the sparticle mass patterns, and would be studied in detail in LHC and LC. For LHC, charginos and neutralinos are produced as decay products of $`\stackrel{~}{g}`$ and $`\stackrel{~}{q}`$, and the nature of weak interacting sparticles will be studied. In this talk, I discussed the decay of the second lightest neutralino. The leptonic decays of the second lightest neutralino $`\stackrel{~}{\chi }_2^0`$ could be studied even though one does not know the parent neutralino momentum. The $`m_{ll}`$ distribution of the three body decay is sensitive to neutralino mixing and slepton masses. If systematical errors can be controlled, one may be able to constrain slepton masses. When the two body decay $`\stackrel{~}{\chi }_2^0\stackrel{~}{l}`$ is open, one would measure the peak of lepton $`p_T`$ asymmetry in $`m_{ll}m_{ll}^{\mathrm{max}}`$. in addition to the edge of $`m_{ll}`$ distribution of the two body decay and occasionally the $`m_{ll}`$ end point of the three body . The information constrain the parent and daughter neutralino and slepton masses rather stringently in model independent manner. In LC, not only masses of sparticles, but production cross sections and sparticle decay distributions will be measured precisely. Underlying MSSM parameters, such as sparticle soft mass parameters, $`\mathrm{tan}\beta `$, gaugino-sfermion-fermion coupling would be measured within precision of $`O(1\%)`$ or less. In this talk, we discuss the determination of squark mass scale in the “decoupling scenario” where $`\stackrel{~}{q}`$ is much heavier than $`\stackrel{~}{W},\stackrel{~}{B}`$, and $`\stackrel{~}{l}`$. The squark mass maybe constrained within O(10%) through the measurement of the deviation of $`\stackrel{~}{W}\stackrel{~}{l}l`$ coupling from its tree level value, $`g_2^{SM}`$.
no-problem/9907/cond-mat9907167.html	ar5iv	text	# Two-terminal conductance fluctuations in the integer quantum Hall regime ## I Introduction Mesoscopic sample-to-sample fluctuations in the conductance, $`G`$, are universal in the sense that the deviation from its average, $`(\delta G)^2=(GG)^2`$, is always about the order of $`(e^2/h)^2`$ at very low temperature . In zero or weak external magnetic fields, this behaviour can be understood in terms of the quantum phase coherence of diffusive electrons . This picture, however, is not expected to work in the quantum Hall regime since the trajectories of electrons are drastically deformed in strong magnetic fields. In this situation, the existence of extended states at sample edges makes the transport properties rather different from the weak-field case . Indeed, magnetoconductance measurements on mesoscopic multiterminal devices seem to agree with this expectation. To be consistent with these observations, explicit inclusion of the edge-channel effects in theories seem to be necessary . The appearance of the bulk delocalised states when the Fermi level is between two quantised Hall plateaus makes a mesoscopic system in strong magnetic field even more interesting . It is only recently, however, that experimental data on the statistical behaviour of the conductance in this regime has become available . Specifically, conductance fluctuations has been studied in mesoscopic $`n`$-channel Silicon Metal-Oxide-Semiconductor Field-Effect Transistors(Si MOSFETs) with two terminals. In these systems, conductance is quantised (in units of $`e^2/h`$) at some values of the gate voltage, $`V_g`$, and the complication due to the Shubnikov-de Haas oscillation in multiterminal devices is absent in the two-terminal setting. It is found that the distribution of conductance below the first conductance plateau is almost uniform between $`G=0`$ and $`e^2/h`$. This is in clear contrast to the Gaussian distribution in the weak-field case . Theoretical studies have been carried out on models based on single-particle interference effect, e.g. tight-binding Hamiltonian , Chalker-Coddington network model , and so on numerically and also a renormalisation group analysis . Some capture this feature of the conductance distribution. More recently, Cobden and collaborators demonstrate further that there is a great difference between conductance fluctuations in the lowest few Landau levels and the weak field case . Using the same sets of Si MOSFET devices as described, they observed in strong fields strong correlations of fluctuation peaks and dips at different magnetic fields. This is achieved by studying explicitly the evolution of fluctuation peaks and dips with magnetic fields in different field regimes. Using a greyscale plot of $`dG/dV_g`$ as a function of $`B`$ and $`V_g`$ , each bright or dark line can be viewed as the ‘history’ line of a peak or a dip of the conductance at different combinations of $`V_g`$ and $`B`$. The contrast between weak and strong magnetic fields is very clear (See Fig. 2 in Ref. 22). For weak magnetic fields, no pattern can be seen whatsoever. At $`B`$10T, we can see peaks and dips of conductance fluctuations in a transition region between two successive conductance plateaus, which are wide and grey regions. The fluctuations ‘move’ in gate voltage, in straight lines as the magnetic field is varied. This pattern means that there exist strong correlations between fluctuation peaks and dips at different gate voltages at high magnetic fields. A closer look reveals that there are two types of lines, with different slopes, in a transition region. The slope of one set of lines is parallel to that associated with the centre of the plateau below, while the slope of the other set of lines is parallel to that associated with the centre of the plateau above. The strong contrast between the fluctuations on the $`V_g`$-$`B`$ plots in weak and strong magnetic fields means that what occurs in two different field regimes is dramatically different. The straight lines appearing in the strong-field results makes it no longer appropriate to treat the phenomenon as a single-particle one. Thus those models we mentioned before are obviously not compatible with strong correlations of conductance fluctuations at different fields. In this paper, we consider a model taking into account the electron-electron interactions. The conductance plateaus are understood in terms of the edge-state picture and the absence of backscattering between channels at opposite edges of the sample. In the transition region between plateaus, we assume that the paths followed by electron currents coming from two contacts (or terminals) percolate into the bulk as the chemical potential is increased and eventually become connected by tunneling through a single saddle point. In addition to the electrons in these edge states, we also take account of electrons in localised states. Conductance fluctuations arising from Coulomb interactions between both sets of electrons are treated in a simple way. As electrons are added to the system, the ones already in the localised states will rearrange due to the interactions. Therefore, the occupation of a given localised state can fluctuate several times between $`0`$ and $`1`$ when the chemical potential, $`\mu `$, is changed from $`\mathrm{}`$ to $`+\mathrm{}`$. The total electrostatic potential between the localised states and the saddle point then also fluctuates as $`\mu `$ is varied. Because of the fluctuations of the saddle point potential, the tunneling of percolation paths fluctuates as a function of the chemical potential. In our model, fluctuations of the conductance are purely due to these interaction effects. Numerical study of this model shows that the conductance between plateaus as a function of the chemical potential does indeed fluctuate significantly within the range between $`G=0`$ and $`e^2/h`$. In the next section, we first develop our model, based on the transmission of edge-state channels through a single saddle point in the transition region. Fluctuations of a single potential saddle point are related to the Coulomb interaction between localised states in the bulk of the disordered 2DES. These localised states are treated using the Efros-Shklovskii Coulomb glass model. Models at finite temperature $`T`$ with localised states arranged on regular lattice points or random sites are then studied numerically by Monte Carlo methods and exact enumeration. Results for $`G`$ versus $`V_g`$ at different disorder configurations and temperatures are obtained. After that, comparisons of our numerical results and experiments are discussed. We conclude with a summary. ## II Theoretical model Before introducing the model, we note that there are different electron states in a disordered two dimensional electron system (2DES) in a high magnetic field. Edge state exists on an equi-energy line along the boundary of the sample connecting two contacts. The width of each edge-state channel is about the order of magnetic length, $`l_B1/\sqrt{B}`$. Therefore, in a high magnetic field the backscattering between edge states at opposite boundaries can be neglected. Since each of the edge-state channels is now one-dimensional, we can apply the two-terminal Landauer-Büttiker formula to calculate the conductance. The current injected into each of the edge states is proportional to the difference of chemical potentials between two contacts and also each edge-channel pair. In the context of the measurements on Si MOSFETs , the bias voltage between two contacts is fixed, but the gate voltage of the metallic gate on top of the 2DES, therefore the Fermi energy, is varied. As the number of electrons is increased such that the Fermi level is between Landau levels $`E_n<E_F<E_{n+1}`$, there are $`n`$ edge channels at each boundary. In a high magnetic field with no inter-edge backscattering, the conductance of the system is thus quantised and given by $`G=ne^2/h`$. In the transition region between two quantised plateaus, it is clear that the Fermi level is within a disorder-broadened Landau level. For clarity, let us consider what happens when the conductance is between the first and the second plateaus (see Fig. 1). At zero temperature, there are localised states in the bulk associated with closed local equipotential lines , shown as dashed and dotted closed lines in Fig. 1. Apart from them, we can imagine that, in the bulk of the system, there are also directed extended states at the boundaries of the regions which are occupied by electrons and connected to one or the other contact. At some value of the chemical potential, these extended states undergo large excursions into the bulk. As $`\mu `$ is increased further, the directed extended states approach each other more closely and tunneling between them occurs (shown as the dotted line in Fig. 1). Eventually, a second pair of edge states is formed in this way at the inner side near the system boundaries. Let us consider the simplest tunneling process in which the two growing regions of electrons form a single potential saddle point assumed to be located at about the centre of the 2DES. The transmission probability $`T_{ij}=\|t_{ij}\|^2`$ for the incoming channel $`i`$ and outgoing channel $`j`$ is already known, for the saddle-point potential of the form $`V(x,y)=V_0(1/2)m\omega _x^2x^2+(1/2)m\omega _y^2y^2`$ with $`\omega _x`$ and $`\omega _y`$ characterising the shape of the saddle-point potential . It is given by $$T_{ij}=\delta _{ij}\frac{1}{1+\mathrm{exp}(\pi ϵ_n)},$$ (1) where $`ϵ_n=[EE_2(n+1/2)V_{sp}]/E_1`$ with $`E`$ denoting the total energy of the electron and $`V_{sp}=V_0`$ being the bare potential strength. The two-terminal conductance is then given by $`G=(e^2/h)_iT_{ii}`$. In Eq. (1), $`E_1`$ and $`E_2`$ are in general complicated functions of $`\omega _c`$ (cyclotron frequency), $`\omega _x`$, and $`\omega _y`$. At high magnetic fields, when $`\omega _c`$ is much bigger than $`\omega _x`$ and $`\omega _y`$, $`ϵ_n`$ in Eq. (1) is simply a dimensionless measure of the energy of the electron’s guiding-centre motion relative to $`V_0`$. For the edge state of the percolating region with a guiding centre energy $`E_G=E_F\mathrm{}\omega _c(n+1/2)`$, the transmission probability is one if $`E_GV_0`$. In the other limit that $`E_GV_0`$, the edge state is completely reflected. The two-terminal conductance as a function of Fermi energy at various magnetic-field strengths have been obtained numerically by Büttiker . Interestingly, although these results are for a 2DES with a point-contact constriction or split-gates on top, the same structure is observed in a ‘macroscopic’ Si MOSFET . From this, it seems that considering only one potential saddle-point in the mesoscopic Si MOSFET where measurements were carried out may be reasonable. Within the framework of the edge-state picture described above, fluctuations of the transmission of percolating states lead to the conductance fluctuations. Here the transmission of electron currents depends solely on the potential energy at the saddle point. In general, this potential energy should have a contribution from the interactions. It is our purpose to relate the interactions to the conductance fluctuations. Electrons in our model are separated into categories: those in edge states and those in localised states. In principle, a full treatment of two interactions between electrons is very complicated . For simplicity, we neglect interactions amongst electrons in edge states, and just consider interactions between different electrons in localised states and their effect on the saddle-point potential. Equally, for simplicity, we neglect tunneling processes that mix the states we term localised with each other or with those we term edge states. This is correct in a finite sample in the semiclassical limit, when for a given value of the Fermi energy tunneling is important only at one saddle-point of the potential. More generally, one expects states at the plateau transition to be delocalised by the tunneling processes we omit, at least within a single-particle description. Note that, although interactions between extended electrons are left completely untreated in the model, it turns out, as we shall present later, that this way of including the Coulomb interaction does indeed produce dramatic fluctuations of the saddle-point potential as the chemical potential of the system is varied. The model Hamiltonian describing a system of localised states interacting with the Coulomb interaction is essentially the one which was studied first by Efros and Shklovskii . The same Hamiltonian produces a gap in the single-electron density of states at the Fermi level due to the Coulomb interaction. In the ‘Coulomb-gap’ system, electrons are strongly localised on a discrete set of sites due to the impurities. These quantum particles can thus be thought of as being in the regime where they behave classically. The Hamiltonian for localised states on $`N`$ sites at positions denoted by $`i`$ is $`H`$ $`=`$ $`{\displaystyle \underset{i=1}{\overset{N}{}}}ϵ_i(n_i{\displaystyle \frac{1}{2}})+{\displaystyle \frac{1}{2}}{\displaystyle \underset{ij}{\overset{N}{}}}{\displaystyle \frac{e^2}{4\pi \epsilon \epsilon _0}}{\displaystyle \frac{(n_i\frac{1}{2})(n_j\frac{1}{2})}{r_{ij}}}`$ (3) $`\mu {\displaystyle \underset{i=1}{\overset{N}{}}}(n_i{\displaystyle \frac{1}{2}}),`$ where $`ϵ_i`$ is the random site-energy, $`n_i`$ is the occupation number which can be either 0 or 1, and $`r_{ij}`$ is the distance between site $`i`$ and $`j`$. The subtraction of $`1/2`$ from $`n_i`$ represents neutralising background charge. $`\epsilon `$ and $`\epsilon _0`$ are dielectric constants for the Silicon and the vacuum, respectively. The last term in the Hamiltonian is crucial to model the experimental variation in the chemical potential in the system as the gate voltage is varied. More explicitly, as the voltage of the gate electrode on top of the 2DES is varied, there will be electrons brought in or pulled out of the 2DES in the inversion layer in the Silicon. The occupancy in each of the localised states will then fluctuate as a function of $`\mu `$, due to the Coulomb interaction. Accordingly, the electrostatic potentials between the $`N`$ localised states and the saddle-point potential will also fluctuate. Hence we have a fluctuating total saddle-point potential $$V_{sp}=\underset{i}{\overset{N}{}}\frac{e^2}{4\pi \epsilon \epsilon _0}\frac{(n_i\frac{1}{2})}{\sqrt{r_{i,sp}^2+h^2}}V_0$$ (4) with $`r_{i,sp}`$ representing the distance between site $`i`$ and the saddle point, where $`h`$ is the perpendicular distance between the inversion layer containing localised states and the centre of the saddle point. Again, the occupation at site $`i`$ is given by $`(n_i1/2)`$ to take into account background charges. Note that at finite temperature $`T`$, in Eq. (4) we need the thermal average of the occupation number, $`n_i`$, instead of $`n_i`$ at each site $`i`$. As a result, the transmission probability given by Eq. (1) fluctuates as a function of the gate voltage. ## III Numerical results and comparisons with experiments In this section, we present the results obtained numerically at finite temperature. The 2DES is chosen to have a rectangular shape with length $`L`$ and width $`W`$ in units of lattice spacing. The localised states are chosen to be fixed at sites $`i`$ arranged on a regular lattice with positions given by $`(x,y)`$. At fixed temperature $`T`$, we compute $`n(x,y)`$, at each site $`i`$ with coordinates ($`x,y`$) for each chemical potential $`\mu `$. Note that, here and in all the following computation, we have set the electric charge, $`4\pi \epsilon \epsilon _0`$, and the Boltzmann constant $`k_B`$ to be one. For each configuration of impurities, a random energy is attached to each localised state. The random site-energy $`ϵ_i`$ at site $`i`$ is some value in the interval $`[𝒲/2,𝒲/2]`$. After $`n(x,y)`$ is obtained, Eq. (4) determines the value of $`V_{sp}`$ at this $`\mu `$ and temperature $`T`$. The corresponding transmission probability is then given by Eq. (1), rewritten here as $`T_{ii}=1/1+\mathrm{exp}[(V_{sp}+c)/E_0]`$, where $`c`$ is a constant representing the energy of the electron’s guiding centre (or the equipotential line) ; $`c`$, however, is always set to be equal to $`V_0`$ in our numerical study. $`E_0`$ here is an energy scale, chosen arbitrarily, which in principle can depend on the magnetic field and the characteristics of the saddle-point. The conductances at different chemical potentials are calculated in this way by employing the same set of random energies at $`N`$ sites. For different sets of random site-energies which essentially represent different disorder realisations, sample-to-sample fluctuations can then be compared. Numerical studies have been carried out using different thermal-averaging methods: the Metropolis Monte Carlo algorithm and exact enumeration. Due to the probabilistic nature of the Metropolis Monte Carlo algorithm, it is important to ensure that the system has reached equilibrium, and that the Monte Carlo average indeed gives the thermal average. Due to the long-range interaction term, we expect that a large system will need an extremely long time to reach equilibrium. At higher temperature, stronger disorder (i.e. larger $`𝒲`$) and larger $`\|\mu \|`$, this difficulty may be avoided since the contribution of the Coulomb interaction is then small. It turns out that at low temperature fluctuations due to lack of equilibration are rather large up to the biggest(6$`\times `$6) system we have reached. In order to study the fluctuations due to interactions in more detail, we turn to calculate the thermal average using exact enumeration. In doing so, what occurs in lower temperature can be investigated more explicitly. Exact enumeration means that the thermal averages are achieved by finding explicitly all possible state configurations and calculating their Boltzmann factors. From the comparison of results using two different methods, it is clear that those small-amplitude fluctuations obtained in using the Monte Carlo algorithm are due to the non-equilibration, instead of the interaction. With reasonable computing time, we obtain results for systems of sizes up to 5$`\times `$4. Here, only the results for 5$`\times `$4 systems are presented. Fig. 2 clearly exhibits significant conductance fluctuations ranging from 0 to $`e^2/h`$ as the chemical potential is varied. The amplitudes of the fluctuations, however, depend on the value of the energy scale $`E_0`$. We can choose different sequences of random numbers in the program to change the random site-energy at each site on the lattice. For the same disorder strength (i.e. $`𝒲`$), the sample-to-sample fluctuations due to different realisations of impurities can then be studied. In the two cases shown in Fig. 2, there are large amplitude fluctuations and the details are sample dependent. To avoid the confusions due to the arbitrary $`E_0`$, only plots of $`V_{sp}`$ versus $`\mu `$, instead of $`G`$ versus $`\mu `$, are shown in Fig. 3 where $`h`$ is the only free parameter to obtain $`V_{sp}`$. We then compare results at different disorder strengths. Fluctuations due to the Coulomb interaction should be suppressed as the strength of the disorder is increased. This is more or less consistent with what is demonstrated in Fig. 3. Finite temperature is another source in our model which can smear out the amplitudes of the fluctuations. This is because, with a fixed chemical potential, different thermal energies give different averaged occupation numbers on the same lattice site. At higher temperature, each electron gains more thermal energy on average and thus the effect of Coulomb interaction is again suppressed. The temperature dependence of the solid and dashed lines shown in Fig. 3 with the same disorder strength clearly shows this behaviour. Finally, we discuss the results on a system which contains randomly distributed localised states. This seems to be a more realistic representation of real devices. We use the exact enumeration for the thermal averaging. The positions of the localised states are chosen independently with a uniform distribution over the system. For the same number of localised states as that in the lattice system, we conclude that finite-size effects are stronger than on the lattice. Although not shown here, results for different disorder strengths show no significant fluctuations. Comparing with the results for the lattice system, fluctuations are much smaller. In order to obtain prominent interaction-driven fluctuations off lattice, it seems possible that we need to study very large random-site systems. We now turn to describe the comparison between the numerical results from our model on a lattice and the experiments. First, let us focus on the conductance as a function of the gate voltage at some fixed magnetic field. The conductance fluctuations produced from our model are sample dependent, and varying the saddle-point potential with chemical potential can produce conductance fluctuations between $`G`$=0 and $`e^2/h`$. Both of these features are consistent with what has been observed in experiments . However, the fluctuation patterns are rather different for our model and the experiments. It is observed in experiments that the conductance fluctuates in sharp peaks and dips in the transition region between two plateaus . By contrast, at least up to the system sizes for which we have done the computation, it is observed that, from Fig. 2 and 3, sharp spikes of the fluctuations are obviously absent in our results. Instead, steps with rounded or flat tops or bottoms occur as we vary $`\mu `$. This feature can be understood in terms of the total electrostatic potential energy changing slowly or even remaining constant for some finite interval of the chemical potential. It is also true that the systems we have studied show a relatively small number of conductance fluctuations as $`\mu `$ varies, compared with experiments. However, we should expect more fluctuations in a bigger system. Our model seems to show this behaviour as the system size is increased. It is therefore not appropriate to analyse the distribution of conductance from our results which only contain small numbers of independent conductance values. Another important comparison is the temperature dependence of the conductance fluctuations. Finite temperature reduces the amplitudes of the fluctuations because the fluctuation of the occupation number in each localised state is smoothed out by the finite thermal energy. In our model, it can be observed that the width of each fluctuation ‘steps’ does not vary with temperature (see Fig. 3). In experiments, data demonstrate indeed that the amplitudes of peaks and dips are enhanced as the temperature is lowered. No obvious shrinking of their widths can be seen there. We now come to the behaviour of fluctuation peaks and dips at different magnetic fields. Although there is no explicit magnetic-field dependence in our model, we argue here, by noting that straight lines on the $`V_g`$-$`B`$ plane are obtained provided the system preserves the filling factor along each line, the occurrence of straight lines is consistent with the spirit of our model. As mentioned in the introduction section, the slopes of two sets of lines are parallel to those associated with the centres of the neighbouring plateaus. More explicitly, it is actually observed that each of these straight lines follows the equation $`V_g=CB+D`$ with $`C`$ and $`D`$ being constants. This behaviour can be understood in terms of the physics of the Si MOSFET , as has been discussed in Ref. 23. With the filling factor given by the relation $`\nu =(V_gV)\epsilon \epsilon _0h/de^2B`$ as a perpendicular magnetic field $`B`$ is applied, we have $$V_g=\frac{e^2d}{\epsilon \epsilon _0h}\nu B+V.$$ (5) Here $`V`$ is some constant and represents a threshold voltage, and $`d`$ is a distance of the order of the thickness of the $`SiO_2`$ layer. To describe the $`i`$th straight line on the $`V_g`$-$`B`$ plot, another constant $`V_i`$, for example, is needed. For different parallel lines on the $`V_g`$-$`B`$ plane, we have different $`V_i`$’s. Along each line, the filling factor and $`V_i`$ are constants. This means that, as $`V_g`$ and $`B`$ are both varied along the line, the fluctuation peaks and dips evolve in such a way that the filling factor of the system is unchanged for a given fluctuation. More explicitly, the value of $`\nu `$ is observed to be either $`i`$ or $`i+1`$ for the $`(i+1)`$th transition region, depending on which plateau region the lines belong to. In our model, conductance fluctuations are associated with the occupancy of localised states in the 2DES. For the occupancy to fluctuate, the state must have energy near the chemical potential. Hence, along each straight line, which connects peaks or dips for different $`(V_g,B)`$, the localised states are at the chemical potential. The fact that $`\nu `$ along the straight line is the same as $`i`$, for example, associated with the $`i`$th plateau centre means that the localised states in both cases must have the same total (kinetic plus electrostatic) energy as the states in the $`i`$th Landau level. This is indeed possible if the localised states belong to the Landau levels of the the 2DES. Following this argument, as the chemical potential crosses the centre of a disordered-broadened Landau level in the transition region, localised states at two tails of Landau level then give the two slopes of straight lines corresponding to two different energies of the states at two plateau centres. ## IV Conclusions In this paper, we have constructed a simple semiclassical model which produces conductance fluctuations in strong magnetic fields due to the Coulomb interaction between electrons. The inclusion of the long-range Coulomb interaction causes the occupation number in each localised state in the bulk to fluctuate as the chemical potential is varied. By taking into account the influence of the electrostatic potential between localised states and the saddle point, the energy of the saddle point in the potential seen by mobile electrons then also varies with the chemical potential. Through this saddle point, the conductance due to the transmission of edge states from one contact to the other thus fluctuates with the chemical potential. We study the model at finite temperature by numerical simulation using the Monte Carlo methods and exact enumeration. At low temperature, the Monte Carlo results suffer slow equilibration. Strong fluctuations due to the non-equilibration of the system in this case make it difficult to extract interaction-driven fluctuations. By contrast, results obtained using exact enumeration clearly exhibit significant fluctuations as a function of the chemical potential. In comparing these results with the experiments, our model shows qualitatively consistent behaviours with the experiments as the gate voltage and the magnetic field are both varied. There are, however, some different features existed between our results and experiments. In particular, although our simulations indeed exhibit fluctuations depending on realisations of disorder they give fluctuations which are like steps instead of the sharp peaks and dips observed in experiments. These discrepancies could arise because we have neglected interactions between bulk extended electrons and many specific details in the Si MOSFET. ## acknowledgements The author is greatly indebted to Dr. John Chalker for numerous discussions. Special thanks to Dr. David Cobden for discussions and providing his experimental results before publishing, Dr. Derek Lee for the help of the programming, and Dr. Chi-Te Liang for the consultation of the general experimental details. This work was supported in part by the ORS Award from the CVCP in United Kingdom.
no-problem/9907/quant-ph9907083.html	ar5iv	text	# 1 Introduction ## 1 Introduction It is well known that phase-insensitive amplifiers introduce at least 3 dB extra noise in the output, whereas phase-sensitive amplifiers may preserve the input signal-to-noise ratio, and in this sense are “noiseless” . A spectral analysis of some cavity-based noiseless amplifiers has been given in Ref., but these investigations are carried out exclusively in the time domain, while spatial aspects are neglected by introducing the plane wave approximation, i.e. by considering only one spatial mode. However, the spatial domain is relevant for the subject of noiseless amplification. Indeed, many areas of physics would benefit from having a possibility of noiseless amplification of faint optical images. Here, based on this motivation, we analyse the parametric image amplification in two different cavity geometries . ## 2 The Optical Scheme A possible realization of a parametric image amplifier is shown in Fig.1. A faint image which is to be amplified is located in the object plane $`O`$. This image is projected by a lenses system in the input plane of a ring-cavity degenerate optical parametric amplifier. The amplified image from the output plane is then projected in the image plane $`I`$ by another lenses system. Each lens has a focal length $`f`$. The presence of the dashed part in the scheme depends on the geometry one wants to consider, as we shall see. Instead, the presence of a pupil $`𝒫`$ of finite area $`S_p`$, is necessary for evaluation of the noise properties. Let us now intruduce the two-component transverse wave vector $`\stackrel{}{q}(k_x,k_y)`$ and position vector $`\stackrel{}{\rho }(x,y)`$. Moreover, let $`a(\stackrel{}{\rho },t)`$, $`a^{}(\stackrel{}{\rho },t)`$, and $`e(\stackrel{}{\rho },t)`$, $`e^{}(\stackrel{}{\rho },t)`$ the photon annihiliation and creation operators of the field in the object plane and image plane respectively. We shall assume that the field in the object plane is in a coherent state with real amplitude $`s(\stackrel{}{\rho })`$ modulated (and even) in space. The observable, in the image plane, is the surface photocurrent density $`i(\stackrel{}{\rho },t)=\eta e^{}(\stackrel{}{\rho },t)e(\stackrel{}{\rho },t)`$, with $`\eta `$ the photodetection efficiency. However, the quantity of interest for us is the number of photodetections $`N_I(\stackrel{}{\rho },t)`$ registered by the pixel, of area $`S_d`$, centered at the point $`\stackrel{}{\rho }`$ in the image plane in the time window $`[tT_d/2,t+T_d/2]`$, i.e. $$N_I(\stackrel{}{\rho },t)=_{S_d}𝑑\stackrel{}{\rho }^{}_{T_d}𝑑t^{}i(\stackrel{}{\rho }^{},t^{}).$$ (1) We shall consider the mean number $`N_I(\stackrel{}{\rho },t)`$ of registered electrons as the amplified signal of our scheme. Its variance characterizes the noise properties of the image $$\mathrm{\Delta }N_I^2(\stackrel{}{\rho },t)=_{S_d}𝑑\stackrel{}{\rho }^{}_{T_d}𝑑t^{}_{S_d}𝑑\stackrel{}{\rho }^{\prime \prime }_{T_d}𝑑t^{\prime \prime }\frac{1}{2}\{\delta i(\stackrel{}{\rho }^{},t^{}),\delta i(\stackrel{}{\rho }^{\prime \prime },t^{\prime \prime })\}_+.$$ (2) The power signal-to-noise ratio (SNR) of the amplified image is given by $$R_I=N_I(\stackrel{}{\rho },t)^2/\mathrm{\Delta }N_I^2(\stackrel{}{\rho },t).$$ (3) Equivalent quantities, $`N_O(\stackrel{}{\rho },t)`$, $`\mathrm{\Delta }N_O^2(\stackrel{}{\rho },t)`$ and $`R_O`$, can be built up in the object plane. Finally, by definition, the noise figure is $$F=R_O/R_I,$$ (4) and we refer to the situation of $`F1`$ as the noiseless amplification. In order to investigate this possibility we have to express the image field in terms of object field. ## 3 Plane Cavity We first consider the case of planar cavity . Hence, the dashed part in Fig.1 has to be neglected. The behavior of the slowly varying part of the field operator $`b(\stackrel{}{\xi },t)`$ inside the cavity is described by the following Langevin equation $$_tb(\stackrel{}{\xi },t)ic_{}^2/2k=\gamma \left[(1+i\mathrm{\Delta })b(\stackrel{}{\xi },t)𝒜_pb^{}(\stackrel{}{\xi },t)\right]+\sqrt{2\gamma }b_{in}(\stackrel{}{\xi },t),$$ (5) where $`k=2\pi /\lambda `$ is the wave number, $`\gamma `$ is the cavity decay rate, and $`\mathrm{\Delta }`$ is the cavity detuning. $`𝒜_p`$ is the constant of parametric interaction proportional to the amplitude of the pump field taken as a classical quantity. The symbol $`_{}^2`$ denotes the transverse laplacian. The above equation can be solved by means of spatio-temporal Fourier transformations; furthermore, with the aid of transformations at lenses $$b_{in}(\stackrel{}{\xi },t)=\frac{d\stackrel{}{\rho }}{\lambda f}a(\stackrel{}{\rho },\mathrm{\Omega })e^{i\frac{2\pi }{\lambda f}\stackrel{}{\rho }\stackrel{}{\xi }},e(\stackrel{}{\rho },t)=\frac{d\stackrel{}{\xi }}{\lambda f}𝒫(\stackrel{}{\xi })b_{out}(\stackrel{}{\xi },t)e^{i\frac{2\pi }{\lambda f}\stackrel{}{\rho }\stackrel{}{\xi }}$$ (6) we arrive at $$e(\stackrel{}{\rho },\mathrm{\Omega })=\frac{1}{\lambda f}𝑑\stackrel{}{\rho }^{}\mathrm{}(\stackrel{}{\rho }\stackrel{}{\rho }^{})\left[U(\stackrel{}{\rho }^{},\mathrm{\Omega })a(\stackrel{}{\rho }^{},\mathrm{\Omega })+V(\stackrel{}{\rho }^{},\mathrm{\Omega })a^{}(\stackrel{}{\rho }^{},\mathrm{\Omega })\right]$$ (7) where $`\mathrm{}`$ is the impulse response function, i.e. the Fourier transform of the pupil function $`𝒫`$ (for an infinitely large pupil $`\mathrm{}(\stackrel{}{\rho })=\lambda f\delta (\stackrel{}{\rho })`$). The coefficients $`U,V`$ are $`U(\stackrel{}{\rho },\mathrm{\Omega })`$ $`=`$ $`{\displaystyle \frac{[1i\delta (\stackrel{}{\rho },\mathrm{\Omega })][1i\delta (\stackrel{}{\rho },\mathrm{\Omega })]+𝒜_p^2}{[1+i\delta (\stackrel{}{\rho },\mathrm{\Omega })][1i\delta (\stackrel{}{\rho },\mathrm{\Omega })]𝒜_p^2}},`$ (8) $`V(\stackrel{}{\rho },\mathrm{\Omega })`$ $`=`$ $`{\displaystyle \frac{2𝒜_p}{[1+i\delta (\stackrel{}{\rho },\mathrm{\Omega })][1i\delta (\stackrel{}{\rho },\mathrm{\Omega })]𝒜_p^2}}.`$ (9) Here, $`\delta (\stackrel{}{\rho },\mathrm{\Omega })=\mathrm{\Delta }\mathrm{\Omega }+(\rho /\rho _0)^2`$ is the local mismatch function, with $`\rho _0=f\sqrt{\lambda \gamma /\pi c}`$ a characteristic transverse length. Now, the quantities of interest can be calculated by using Eq.(7), and some simplifying assumptions, i.e.: $`\lambda f/S_p^{1/2}`$ much smaller than the typical scale of change of $`U`$, $`V`$ and $`s`$, so we can take the latter functions out of integral when they enter as a product with $`\mathrm{}`$; the size of each pixel is the smallest of all the spatial scales, so that we can substitute integration over the pixel area by multiplication by $`S_d`$; observation time $`T_d`$ long compared with the inverse cavity bandwidth $`\gamma ^1`$. All that leads to $$N_I(\stackrel{}{\rho },t)=\eta S_dT_ds^2(\stackrel{}{\rho })G(\stackrel{}{\rho })+\mathrm{noise},$$ (10) with $`G`$ the gain factor $$G(\stackrel{}{\rho })=\{[(1+𝒜_p)^2\delta (\stackrel{}{\rho },0)]^2+4\delta ^2(\stackrel{}{\rho },0)\}/\{[1+\delta ^2(\stackrel{}{\rho },0)𝒜_p^2]^2\}.$$ (11) For the variance we get $`\mathrm{\Delta }N_I^2(\stackrel{}{\rho },t)`$ $`=`$ $`\eta S_dT_ds^2(\stackrel{}{\rho })G(\stackrel{}{\rho })\left\{1\eta +\eta \left[\mathrm{cos}^2\theta (\stackrel{}{\rho })e^{2R(\stackrel{}{\rho })}+\mathrm{sin}^2\theta (\stackrel{}{\rho })e^{2R(\stackrel{}{\rho })}\right]\right\}`$ (12) $`+`$ $`\mathrm{self}\mathrm{interference}\mathrm{of}\mathrm{the}\mathrm{noise},`$ where we have introduced the squeezing parameter $`\mathrm{exp}\left[\pm R(\stackrel{}{\rho })\right]=\|U(\stackrel{}{\rho },0)\|\pm \|V(\stackrel{}{\rho },0)\|`$, and the orientation angle $`2\theta (\stackrel{}{\rho })=\mathrm{arg}\left[U(\stackrel{}{\rho },0)+V(\stackrel{}{\rho },0)\right]\mathrm{arg}\left[U(\stackrel{}{\rho },0)\right]\mathrm{arg}\left[V(\stackrel{}{\rho },0)\right]`$. The condition to neglect the (unspecified) noise terms in Eqs.(10), (12) reads $$s^2(\stackrel{}{\rho })(\lambda ^2f^2/S_p)(2\pi /\gamma )1.$$ (13) It also fixes the resolution of the scheme. Once it is satisfied, the noise figure becomes $$F=\{1\eta +\eta [\mathrm{cos}^2\theta (\stackrel{}{\rho })e^{2R(\stackrel{}{\rho })}+\mathrm{sin}^2\theta (\stackrel{}{\rho })e^{2R(\stackrel{}{\rho })}]\}/\{\eta G(\stackrel{}{\rho })\}.$$ (14) The optimum condition for noiseless amplification, coming from Eqs.(11) and (14), is $`\delta (\stackrel{}{\rho },0)=0`$. Hence, depending on the value of the detuning, the noiseless amplification takes places in a small region around the optical axis or in a anular region. Therefore, for reconstruction of the whole image one has to use scanning. ## 4 Confocal cavity Let us now go to the configuration with confocal cavity . In this case the dashed part of the optical scheme in Fig.1 has to be considered. In particular the confocality is guaranteed by means of a specific relation between the focal length of the intracavity lens and the cavity round trip length. The intracavity field $`b`$ can be expanded on the basis of the Gauss-Laguerre modes $`\{f_{p,l,i}(\stackrel{}{\rho })\}`$ as $$b(\stackrel{}{\rho },t)=\underset{p,l,i}{}f_{p,l,i}(\stackrel{}{\rho })b_{p,l,i}(t).$$ (15) All the modes with $`l`$ even have the same frequency; the same is true for the modes with $`l`$ odd; the frequency separation between the two groups of modes is equal to one half the free spectral range. Due to the frequency degeneracies, the field can be split into even and odd part $`b(\stackrel{}{\rho },t)=b_+(\stackrel{}{\rho },t)+b_{}(\stackrel{}{\rho },t)`$. We consider the modes with $`l`$ even quasi-resonant with the signal. Since the detuning is equal for all these modes, it is possible to get the following Langevin equation for the intracavity field $$_tb_+(\stackrel{}{\rho },t)=\gamma \left[(1+i\mathrm{\Delta }_+)b_+(\stackrel{}{\rho },t)𝒜_pb_+^{}(\stackrel{}{\rho },t)\right]+\sqrt{2\gamma }b_+^{in}(\stackrel{}{\rho },t).$$ (16) This equation can be solved in the frequency domain; furthermore, with the aid of transformations at lenses $$b_{in}(\stackrel{}{\rho },t)a(\stackrel{}{\rho },t),e(\stackrel{}{\rho },t)=\frac{d\stackrel{}{\xi }}{\lambda f}𝒫(\stackrel{}{\xi })\frac{d\stackrel{}{\rho }^{}}{\lambda f}b_{out}(\stackrel{}{\rho }^{},t)e^{i\frac{2\pi }{\lambda f}(\stackrel{}{\rho }^{}\stackrel{}{\rho })\stackrel{}{\xi }},$$ (17) we can write $$e(\stackrel{}{\rho },\mathrm{\Omega })=\frac{1}{\lambda f}𝑑\stackrel{}{\rho }^{}\mathrm{}(\stackrel{}{\rho }\stackrel{}{\rho }^{})\left[U(\mathrm{\Omega })a_+(\stackrel{}{\rho }^{},\mathrm{\Omega })+V(\mathrm{\Omega })a_+^{}(\stackrel{}{\rho }^{},\mathrm{\Omega })\right],$$ (18) where the coefficients $`U(\mathrm{\Omega })`$ and $`V(\mathrm{\Omega })`$ are the same of Eqs.(8) and (9) with now the mismatch function no longer dependent from the position vector, i.e. $`\delta (\mathrm{\Omega })=\mathrm{\Delta }_+\mathrm{\Omega }`$. Repeating the steps of previous Section, we easily obtain the mean number of photoelectrons $$N_I(\stackrel{}{\rho },t)=\eta S_dT_ds^2(\stackrel{}{\rho })G+\mathrm{noise},$$ (19) where now the gain factor $`G`$ takes a simpler form $$G=[(1+𝒜_p)/(1𝒜_p)]^2,$$ (20) considering the situation of perfect resonance, i.e. $`\mathrm{\Delta }_+=0`$. For the variance we have $$\mathrm{\Delta }N_I^2(\stackrel{}{\rho },t)=\eta S_dT_ds^2(\stackrel{}{\rho })G\left\{1\eta +\eta G\right\}+\mathrm{self}\mathrm{interference}\mathrm{of}\mathrm{the}\mathrm{noise}.$$ (21) To neglect the noise terms in Eqs.(19) and (21), we again use the condition (13). Then, the noise figure becomes $$F=\{1\eta +\eta G\}/\{\eta G\}.$$ (22) As can be seen from Eqs.(20) and (22), the noiseless amplification occours uniformely over the transverse plane. Hence, this scheme offer the possibility of amplification of the whole image at once. Moreover, a high gain is required to only compensate the effect of non efficient detection. FIGURE CAPTIONS Fig.1 A possible realization of the parametric image amplifier.
no-problem/9907/cond-mat9907334.html	ar5iv	text	# Vicinal Surfaces, Fractional Statistics and Universality ## Abstract We propose that the phases of all vicinal surfaces can be characterized by four fixed lines, in the renormalization group sense, in a three-dimensional space of coupling constants. The observed configurations of several Si surfaces are consistent with this picture. One of these fixed lines also describes one-dimensional quantum particles with fractional exclusion statistics. The featureless steps of a vicinal surface can therefore be thought of as a realization of fractional-statistics particles, possibly with additional short-range interactions. A crystal surface cut at a small angle to a symmetry direction is called a vicinal surface. Such a miscut surface consists of terraces of the symmetry plane, separated by monatomic steps running across the sample in a preferred direction dictated by the cut. The density of the steps (number of steps per unit transverse length), $`\rho `$, is related to the miscut angle $`\theta `$ $`(\rho \mathrm{tan}\theta )`$. In thermal equilibrium, the configuration of the surface is determined by the steps and their interactions. Vicinal surfaces constitute a special class of objects in low dimensional statistical mechanics and are useful as substrates in many technological and experimental situations. An understanding of the phases and phase transitions of vicinal surfaces is therefore important. Various types of behaviors of vicinal surfaces are known. For example, it is known that vicinal Si(111) surfaces have single steps or triplets but never pairs, while pairs of steps are seen on Si(001) . An unusual phenomenon occurs in the case of Si(113) where a uniform-step-density phase, on cooling, phase-separates into a flat surface and a phase with a large angle (i.e., a high density of steps) . In other words, a small-angle vicinal Si(113) surface becomes thermodynamically unstable. The coexistence curve, on the temperature-versus-miscut-angle orientational phase diagram, ends at a tricritical point with the shape given by an exponent $`\beta =0.42`$-$`0.54`$. A variant of the phase diagram has also been reported in Ref. \[4b\]. Is there a natural way of characterizing this zoo of vicinal surfaces? In this paper our aim is to develop a long-distance, universal behavior for vicinal surfaces, where microscopic details like the specific material, the lattice structure, surface reconstruction if any, etc., do not play a direct role. This is best done in a continuum approach. The steps are treated as fluctuating elastic strings (directed polymers), and all the effects of the surface go into the elastic energies of the steps and their effective interactions. Therefore, our approach is valid for any vicinal surface that can be characterized by featureless, wandering steps. A major interaction of the steps is known from theory and experiments to be $`𝐫_i(z)𝐫_j(z)^2`$ where $`z`$ is the special direction of the steps and $`𝐫_i`$ is the transverse $`d`$-dimensional coordinate of a point at $`z`$ along the length of the $`i`$th step. (Our interest is, of course, at $`d=1`$.) This long-range interaction is generally repulsive, originating from the elastic effects of the terraces, although, for metals, dipoles or quenched impurities on the steps can produce an attractive $`r^2`$ potential. A theoretical explanation of the observed tricriticality requires an additional attractive short-range interaction. The continuum description of Ref. is justified a posteriori by the existence of a renormalization group fixed point with a diverging length-scale (see also Refs. ). In an effective Hamiltonian for the steps, the interaction need not be restricted to just pairwise-additive potentials, and, in a renormalization-group (RG) approach, one should include allowed marginal operators. We keep the three-body short-range interaction since it is marginal at $`d=1`$ (see below) . All $`m`$-step interactions are irrelevant for $`m>3`$ at $`d=1`$. The Hamiltonian is $`H`$ $`=`$ $`{\displaystyle \underset{i}{}}{\displaystyle _0^N}𝑑z{\displaystyle \frac{\kappa _i}{2}}\dot{𝐫}_i^2(z)+{\displaystyle \underset{i>j}{}}{\displaystyle _0^N}𝑑z\left[v_2\delta _\mathrm{\Lambda }(𝐫_{ij}(z))+{\displaystyle \frac{g}{r_{ij}^2(z)}}\right]`$ (2) $`+v_3{\displaystyle \underset{i>j>k}{}}{\displaystyle _0^N}𝑑z\delta _\mathrm{\Lambda }(𝐫_{ij}(z))\delta _\mathrm{\Lambda }(𝐫_{jk}(z)).`$ where $`\dot{𝐫}_i(z)=𝐫_i/z`$, $`\kappa _i`$ is the elastic constant of the $`i^{\mathrm{th}}`$ step, each of length $`N(\mathrm{})`$, and $`g,v_2,v_3`$ are respectively the long-range two-body, the short-range two-body and the short-range three-body coupling constants. The short-range interactions are taken as contact interactions, and power counting shows that this is sufficient . There is a short-distance cut-off (reminiscent of the lattice) so that, in Fourier modes, $`\delta _\mathrm{\Lambda }(𝐪)=1,`$ for $`q<\mathrm{\Lambda },`$ and $`0,`$ otherwise. For the tricritical point, one takes $`v_2=v_{20}(TT_t)`$ but, for generality, we consider both positive and negative values of all the three parameters $`(g,v_2,v_3)`$. The partition function is then given by the summation of the Boltzmann factor over all possible configurations with free end-points. Two special cases of Eq. (2) $`(v_3=0,\mathrm{and}\mathrm{g}=0)`$ were considered in Ref. in connection with the tricritical point seen in Si(113). The physical picture used is that of phase separating polymers with the critical point of the coexistence curve coinciding with the binding-unbinding critical point of two steps (zero density). It was shown in Ref. that $`\beta =0.5`$ occurs for $`g=3/4`$. The resemblance of Eq. (2) with the Calogero-Sutherland model is apparent and we shall discuss some issues related to this. The Hamiltonian $`(\mathrm{}=1)`$ describes a set of interacting quantum particles which can be chosen as bosons. The thermodynamic properties of the quantum system can be obtained from Eq. (2) with $`N`$ as the inverse temperature. In the limit of infinite lengths of the steps, the properties of the vicinal surface are given by the ground state of the corresponding quantum problem. The main contents of the paper are the following: (a) We show that for, the $`v_3=0`$ case, at $`g=3/4`$ the continuous transition goes over to a first-order transition. This requires a study of reunion exponents for the steps with the long-range interaction. The change in the order of the transition is then connected to the observed behavior of Si(113). (b) We obtain a fixed-point (or, rather, a fixed-line) description for the above Hamiltonian, and, from the nature of the fixed points, we argue that the observed features of various Si vicinal surfaces are described by these fixed points. We then conjecture that all vicinal surfaces are described by these fixed lines (and the flows) in a three-dimensional parameter space. (c) It is quite common to use a quantum description where the steps are treated as fermions, but we conclude, from the equivalence with the Calogero-Sutherland model on the $`v_3=0`$ fixed line, that the steps should rather be treated as one-dimensional fractional-exclusion-statistics particles with only short-range interactions. So far as the RG flows of the various parameters are concerned, one may just restrict oneself to two or three steps (“vertex functions”). For a many-step system, one also needs the chemical potential $`\mu `$, but its flow is determined by dimensional analysis. In the momentum-shell renormalization-group approach, short-wavelength fluctuations are integrated out in a thin shell $`(\mathrm{\Lambda }/b,\mathrm{\Lambda })`$ in Fourier space, and the effect is absorbed by redefining the parameters. The system is then rescaled to its original state, whereby the cutoff goes back to $`\mathrm{\Lambda }`$. (We choose $`\mathrm{\Lambda }=1`$.) The change of the parameters with distance in the long-distance limit is then recovered from the RG flow-equations. The spatial rescaling factor is anisotropic, $`rbr`$, $`zb^\zeta z`$, and the choice $`\zeta =2`$ keeps the elastic constant invariant. For simplicity, we choose $`\kappa _i=1,`$ for all $`i`$. Since the RG transformation is analytic, the long-range term does not get renormalized. That there is no renormalization of $`g`$ turns out to be rather natural when we discuss the quantum problem. We just quote the well-known recursion relations in terms of the dimensionless variable $`u_2=L^{2d}K_dv_2+g/(d2),`$ where $`K_d=2\pi ^{d/2}/[\mathrm{\Gamma }(d/2)(2\pi )^d]`$ and $`L`$ is an arbitrary length-scale in the transverse direction $`(b=1+\delta L/L)`$. Restricting ourselves to $`d=1`$, the flow equations are $$L\frac{du_2}{dL}=u_2u_2^2+g,L\frac{dg}{dL}=0,\mathrm{and}L\frac{d\mu }{dL}=2\mu .$$ (3) The flow equation for $`u_2`$ has two fixed points $`u_{\mathrm{s},\mathrm{u}}^{}=\{1\pm [1+4g]^{1/2}\}/2`$, s,u denoting respectively the stable and unstable fixed points (Fig. 1a). The unstable fixed point describes the binding-unbinding transition of two steps. The critical point has a length-scale exponent $`\nu _{}=1/\sqrt{1+4g}`$. For a many-step system this unstable fixed point corresponds to the tricritical point mentioned earlier, provided the fixed point describes a critical behavior. Since the density $`\rho `$ vanishes at this tricritical point, the assumption of a single length scale then tells us that $`\rho ^1`$ should diverge as the length scale. Therefore, $`\beta =\nu _{}`$. For $`\beta =1/2`$, as for Si(113), one would require $`g=3/4`$ as obtained in Ref. . We now show that $`g=3/4`$ is indeed a special point. Let us consider the free energy of two steps. Configurations of the steps are of the type shown in Fig. 2. The bubble-like contributions (region B in Fig. 2a) represent the steps in the high temperature phase described by the stable fixed point. It is shown in Ref. (we, therefore, skip the details) that the nature of the singularity of the free energy is determined by the decay of the partition function of the bubble of length $`N`$, $`Z_R(N)N^\psi `$, where $`\psi `$ is the reunion exponent. If $`1<\psi 2`$, then the transition is continuous, while for $`2<\psi `$ it is first order, but with a weak singularity and a diverging length scale. The reunion exponent $`\psi `$ determines the behavior of the partition function of two steps starting at the origin and reuniting anywhere in space. We need the exponent in the high-temperature phase characterized by the stable fixed point. The zeroth- and the first-order diagrams for the reunion partition function $`Z_R`$ are shown in Figs. 2b,c. These contributions require renormalization of the partition function itself, over and above the renormalization of the parameters. This extra renormalization yields a nontrivial $`\psi `$ different from the Gaussian value $`\psi =d/2`$ in $`d`$ dimensions. In the limit $`N\mathrm{}`$, the loop contributions lead to the recursion relation for $`Z_R(N)`$ as $`LdZ_R/dL=2(d/2+u_2)Z_R,`$ so that at $`d=1`$ at the stable fixed point $`u_2^{}=u_\mathrm{s}`$, we get $`\psi =1+(1+4g)^{1/2}/2`$. This is an exact result from which we recover the vicious walker exponent $`\psi =3/2`$ for $`g=0`$. (The details and generalizations to arbitrary number of steps will be discussed elsewhere.) We see that $`\psi =2`$ for $`g=3/4`$. This then establishes that the transition is first order for $`g>3/4`$. In the case of a first-order two-step phase-transition, this may not be the critical point of the phase-coexistence line of a many-step system. In such a case, a phase diagram of the type shown in Fig. 1b.2 is plausible, where a first-order line continues from the zero-density transition point, ending on the coexistence curve (hidden region in Fig. 1b.2, see figure caption). The first-order line represents a tranition from a uniform phase of single steps to a uniform phase of pairs, and, at the tranition, there will be diverging fluctuations in the mean spacing between the members of the pairs (for $`3>\psi >2`$). In recent experiments on Si(113) such a phase diagram has been observed for certain azimuthal angles of the miscut direction, though the low-density line is yet to be detected. Let us now consider the effect of the three-body interaction. Even though $`v_3`$ is marginal (dimensionless at $`d=1`$), there are contributions in $`v_3`$ renormalization, from both $`v_2`$ and $`g`$ (i.e. $`u_2`$) via the reunion exponent of three steps. Note that $`v_3`$ does not affect $`u_2`$ or $`g`$. Taking into account the first nontrivial contribution (Figs. 2d, e), the recursion relation for $`u_3=L^{1d}v_3`$ is ($`d=1`$) $$L\frac{du_3}{dL}=3u_2u_3Au_3^2,$$ (4) where the numerical value of the constant $`A>0`$ is not crucial. The fixed points (Fig. 1c) are $`u_3^{}=0`$ and $`u_3^{}=3u_2/A`$. The stability of the $`u_3^{}=0`$ fixed-point depends on the sign of $`u_2^{}`$. At $`u_2^{}=u_s^{}`$ (fixed point A), a small $`u_3`$ is irrelevant, but the new fixed-point C, $`u_3^{}=3u_s/A<0`$, is unstable. In contrast, at $`u_2^{}=u_u^{}`$ (fixed point B), with $`g>0`$, $`u_3`$ is relevant, and $`u_3^{}=(3\|u_u\|)/A>0`$ is stable. The flows for any $`g>0`$ are shown in Fig. 1c. Figure 1c shows that for certain combinations of $`g>0,u_2`$, and $`u_3`$, the steps in the high-temperature phase behave like the $`u_3=0`$ case (region $`\mathrm{P}_\mathrm{f}`$) but for sufficiently large $`u_3<0`$ there will be triplets of steps but never pairs (Phase $`\mathrm{P}_3`$). The transition will be first-order but with singularities determined by the $`g`$-dependent fixed point C of Fig. 1c. We like to associate Si(111) vicinal surfaces with this fixed point and the associated flow, because of similar behavior. For fixed points B and D, there could be pairings of steps, induced by a variation in $`u_2`$. Such pairings are seen, e.g., in Si(001). A many-step system will show a tricritical behavior at these fixed points, as e.g. in Si(113). These two fixed points B and D of Figs. 1c,d have the same thermodynamic exponents but they differ in the three-step correlation functions. Moreover, if $`u_2`$ is a temperature-like axis, then the temperature deviation is not the proper scaling variable at D. Taking this as a reason for the large range in the observed exponent $`\beta `$, we may associate Si(113) with the fixed point D with $`g`$ around $`3/4`$ and Si(001) with fixed point B. There is, in fact, no requirement that the long-range interaction should be exactly at $`g=3/4`$ for Si(113). Unfortunately no systematic studies are available for the dependence of $`g`$ on the various miscut parameters. In case $`g`$ depends on the azimuthal angle even slightly, it is possible that for certain azimuthal angles $`g>3/4`$ and for these cases, one would see a phase diagram of type Fig. 1b.2, while, for cases with $`g<3/4`$, one would see type Fig. 1b.1. We now make a conjecture that all vicinal surfaces are described by the RG flows and the fixed lines we obtained in the three-dimensional parameter space. Other higher-order short-range interactions are irrelevant. All universal properties are determined by the fixed lines, and it is the long-range parameter $`g`$ that determines the universality class. Let us now go back to Eq. (2) as a quantum-mechanical problem. With $`v_3=0`$, the connection of this Hamiltonian with the Calogero-Sutherland model has been noted in the past. The quantum problem is described by the stable fixed line and the negative-$`g`$-part of the unstable line in Fig. 1a. Furthermore, the Calogero-Sutherland model describes a gas of noninteracting particles obeying Haldane-Wu fractional exclusion statistics. The long-range $`1/r^2`$ potential is best thought of as a statistical interaction, with $`g`$ determining the statistics of the particles. We have established this directly in our approach by a computation of the quantum second virial coefficient along the fixed lines of Fig. 1a, since the second virial coefficient was shown, from general considerations, by Murthy and Shankar to determine the statistics uniquely. We wish to discuss this technical issue elsewhere. Suffice it to say here that the non-renormalization of $`g`$ in our RG approach is a direct manifestation of the fact that the statistics of the particles is independent of length scale and hence a renormalization-group invariant. This correspondence therefore tells us that the steps on vicinal surfaces are an analog realization of one-dimensional fractional-statistics particles. Though fermions are extensively used to study the equilibrium properties of steps, the latter are better represented by fractional-statistics particles with only short-range interactions. Our main results have already been summarized. It would be interesting to study the case of unequal elastic constants of the steps, and the regions not describable by any fixed point, where other details of the surface, like lattice periodicity, may be important. We end with the suggestions that (i) attempts be made to determine accurately the long-range (or statistics) parameter $`g`$ and its dependence on the miscut features, as e.g. azimuthal angle, for various surfaces and materials, (ii) measurements be done for three-step correlations, and (iii) the thermodynamic behavior of a finite number of steps be studied even if in finite geometries (by putting barriers). Most interesting would be the special case of $`g=1/4`$ which, in the surface context, is the point of coalescence of the stable uniform-density-phase fixed point and the unstable critical point, and, in the quantum context, is the semion, exactly half way between boson and fermion. Can one find such a surface of a metal?
no-problem/9907/astro-ph9907021.html	ar5iv	text	# A NICMOS Survey of Early-Type Galaxy Centers1: The Relation Between Core Properties, Gas and Dust Content and Environment ## 1 Introduction Optical imaging surveys carried out with The Hubble Space Telescope (HST) find that almost all early-type galaxies harbor dust (van den Bosch et al. (1994)). Even though projection effects cause a bias against detecting dust features in most of these galaxies (e.g., van Dokkum & Franx (1995)) small 100 pc scale dust features are observed in a large fraction of galaxies with a variety of morphologies (spiral, warped or irregular). Peletier et al. (1990) find that asymmetries in the isophote shapes are particularly sensitive to the presence of dust. Because the dust is found near the galaxy nuclei, a simple dust screen model can underestimate the extinction or dust column depth. This implies that a significant fraction of light in the nuclei could be absorbed by dust. Despite this problem, visible WFPC and WFPC2 images from HST have been part of a major effort to classify the nuclear stellar profiles in these galaxies, resulting in the classification of light profiles into two categories, galaxies with shallow inner cusps and galaxies with ‘power law’ light profiles (Lauer et al. (1995)). We note that this classification is dependent on the angular resolution of the image, and so on the distance of the galaxy. Near-IR images, which are less sensitive to extinction from dust, provide superior light profiles for dynamical studies of early-type galaxy cores. Although almost all early-type galaxies harbor dust, the samples used to classify light profiles are biased against the presence of dust precisely because of the sensitivity of the optical images to extinction. By using near-IR light profiles we can probe the central light profiles for a class of ellipticals with a higher dust content. Since the dust itself could be related to the formation process and subsequent evolution of the galaxy (related to merger induced accretion, or caused by dissipation in the gas, Dubinski (1994)) comparison of IR-observed light profiles between galaxies with varying amounts of dust may test models of core formation. In this paper we present the results of a survey at 1.6 microns of early type galaxies. We present high angular resolution images observed with NICMOS Camera 2 on board HST. Surface brightness profiles and isophotal parameters are measured from the images and compared with the most comprehensive summary of the visible band based studies (Faber et al. (1997)). We investigate the possibility of correlations between core properties and cold gas content (traced by emission from dust in the far infrared), hot gas content (traced by X-ray luminosity) and cluster environment. ## 2 Observations ### 2.1 The Sample Twenty-seven early-type galaxies, listed in Table 1, were observed in F160W ($`1.6\mu `$m) on camera 2 of NICMOS primarily as part of a snap shot program. Galaxies were chosen preferentially to have existing HST visible band images so that color maps could be made. This included a wide variety of early type galaxies including some galaxies with kinematically distinct cores (those studied in Carollo et al. (1997)). We then added galaxies from the RSA listed in Roberts et al. (1991) as having higher dust contents so as to insure that our sample was not grossly biased against dusty galaxies. Our sample was initially chosen to be representative of the distribution of ellipticals in terms of cold ISM content, however only a third of our 80 target galaxies were observed. In Fig. 1 we show a histogram of dust masses estimated from IRAS far infrared emission in our sample compared to the RSA (compiled by Roberts et al. (1991)) and sample of visible HST images compiled by van Dokkum & Franx (1995). We see that our sample is not as biased towards dust free galaxies as was the sample observed with WFPC and WFPC2 (compiled and discussed in Faber et al. (1997)). ### 2.2 The NICMOS images The NICMOS observation sequences were MULTIACCUM with 13 samples of step32 with a sequence exposure time of 192s. The galaxies were observed in the F160W (1.6 micron) filter with this sequence at 4 different positions on the sky separated by $`0.56^{\prime \prime }`$ resulting in a total exposure time of 12.8 minutes per galaxy. Images were reduced with the nicred data reduction software (McLeod (1997)) with on-orbit flats and darks taken near the time the data were observed. Then each set of 4 images in a given filter were combined according to the position observed. The pixel size for the NICMOS camera 2 is $`0.076^{\prime \prime }`$. Flux calibration for the NICMOS images was performed using the conversion factors based on measurements of the standard stars P330-E and P172-D during the Servicing Mission Observatory Verification program (M. Rieke, private communication). The NICMOS images coupled with visible broad band images observed with WFPC2 on board HST when available are shown in Fig. 2. Galaxies lacking WFPC2 images are shown in Fig. 3. These images show the improved ability of the 1.6 micron images to measure stellar surface brightness profiles in the presence of moderate amounts of dust. For example in NGC 524, NGC 1400, NGC 1553, NGC 3056, NGC 4261, NGC 4278, NGC 4374, NGC 4589 and NGC 7626 we can more accurately measure profiles from the NICMOS images than possible with the WFPC2 images. However we do find galaxies such as NGC 7052 and NGC 4150 which show evidence for large extinctions from dust even at 1.6 microns. Though we can make better measurements than possible from visible wavelength images, even in the near-IR the presence of the dust in these galaxies hampers our ability to accurately measure the stellar surface brightness profiles. ### 2.3 Isophote fitting To measure properties of the stellar cores, we fit ellipses to the isophotes using the ellipse routine in the stsdas package of iraf which uses an iterative method described by Jedrzejewski (1987). We also deconvolved our images with 20 iterations of the Lucy-Richardson method (from the stsdas package of iraf) to take into account the point spread function of the telescope. For deconvolution we used a model point spread function created by the program Tinytim (Krist & Hook (1997)) at the position of the nucleus of each galaxy on the camera. We then reran our ellipse fitting algorithm on the deconvolved images. The results of the ellipse fitting are shown in Fig. 4. The “Nuker Law” (e.g., Faber et al. (1997)) was then fit to the surface brightness profile derived from the deconvolved images. This profile is described by $$I(r)=I_b2^{(\beta \gamma )/\alpha }\left(\frac{r_b}{r}\right)^\gamma \left[1.0+\left(\frac{r}{r_b}\right)^\alpha \right]^{(\gamma \beta )/\alpha }$$ Here $`I_b`$ is the surface brightness at $`r_b`$ and the logarithmic slope inside the break radius $`r_b`$ is $`\gamma `$ and that outside is $`\beta `$. The parameters resulting from this fit are given in Table 2. Surface brightness profiles are classified as either ‘power law’ galaxies or ‘core’ galaxies. To identify a galaxy as having a core we require that the absolute value of inner logarithmic slope be $`\gamma <0.3`$ (e.g., Faber et al. (1997)). We note that this classification is dependent on the angular resolution of the image, and so on the distance of the galaxy. This implies that distant galaxies are more likely to be classified as power law galaxies. The fitted values of $`r_b`$ are meaningful only for the core profiles where they represent a physical change in the nature of the stellar distribution (see for example the discussion in Faber et al. (1997)). We note that our range of radii fit ($`10^{\prime \prime }`$ from the nucleus) is comparable to that of the visible band studies. These studies restricted the range of their fits because outside a radius of $`10^{\prime \prime }`$ a deVaucouleurs law might be a better description of the profile. The Nuker Law was used to fit Planetary Camera images and is intended precisely to fit over a short range of radius. However we note that the parameters resulting from such a fit are necessarily highly correlated. The quality of our fits is identical to that of Byun et al. (1996). #### 2.3.1 Comparison with previous visible band fits We have galaxies 14 in common with previous HST based visible band studies (Faber et al. (1997), Carollo et al. (1997) and Ferrarese et al. (1994)). We compare the results of our fits (see Fig. 5) to those found from these previous studies. For the most part our break radii and exponents agree with those found from the previous studies. (NGC 2636 was excluded from this comparison since no break in surface brightness profile was listed in Faber et al. (1997)). Outliers (or those galaxies with fitting parameters that grossly disagree with those found previously) mostly correspond to power law galaxies where the break radii and inner exponent value are not particularly meaningful (NGC 1172 and NGC 1331). We note that we do find a small bias towards measuring smaller break radii from the well resolved core galaxies in the NICMOS images compared to previous visible band based studies. We consider these cases: NGC 1600, NGC 4374, NGC 4261 and NGC 5845. NGC 4374, NGC 4261 and NGC 5845 contain prominent dust features within their break radii. The visible band images suffer from absorption in their central regions. This would result in measurement of a smaller break radius or a shallower central surface brightness profile (however the discrepancy would depend on the dust distribution). We do not feel that the discrepancy between our measured break radius for NGC 1600 is significant given that the parameters of the Nuker fit are highly correlated and that the NGC 1600 visible image used to measure surface brightness profile properties was not a WFPC2 image and so may be suffering from point spread function artifacts. We find that dust features in the optical images are the main source of discrepancies of core classification and fitting parameters between our work and previous visible band studies. Figure 6 shows correlations from bulk elliptical galaxy properties (dispersion and effective radius) and a comparison between break radius and surface brightness at this radius with galaxy luminosity. As found in previous works (e.g., Faber et al. (1997) and references therein) the galaxies display a dichotomy: those with measurable core break radii and low break surface brightnesses correspond mainly to higher visual luminosity galaxies, and those with steep central profiles and high central surface brightness correspond to lower visual luminosity galaxies. We do see a correlation between break radius and surface brightness at this radius (as proposed in Faber et al. (1997) and shown here in Fig. 7). #### 2.3.2 Classification between core and power law galaxies Our classifications (between core and power law) agree in all cases with those previous found (listed in Faber et al. (1997), Ferrarese et al. (1994) and Carollo et al. (1997)) but NGC 7262, NGC 1400 and Abell 2052. In NGC 7626 our fit to the F160W surface brightness profile results in $`\gamma =0.46`$, $`r_b=0^{\prime \prime }.5`$. We therefore classify this galaxy as a power law galaxy. However Carollo et al. (1997) classified it as a core galaxy based on WFPC2 F814W and F555W images. These images display a small warped dusty disk at $`r<0^{\prime \prime }.45`$ and consequently these authors note that their fit (with $`\gamma =0`$, $`r_b=0^{\prime \prime }.32`$) is uncertain within $`r<0^{\prime \prime }.45`$. Since extinction from dust artificially lowers the central surface brightness it is likely that the NICMOS images are a better tracer of the stellar surface brightness. This galaxy is likely to have the steep surface brightness profile typical of a power law galaxy. In NGC 1400, our fit to the F160W surface brightness profile results in $`\gamma =0.35`$, $`r_b=0^{\prime \prime }.85`$. This value of $`\gamma `$ is only slightly above $`0.3`$ that divides power law from core galaxies. Faber et al. (1997) classifies it as a core galaxy (with $`\gamma =0`$ and $`r_b=0^{\prime \prime }.33`$). However we note that the WFPC2 image does not clearly show a core. In this particular case, the classification is uncertain. Abell 2052 is quite distant and luminous so we expect it to be a core galaxy. Though we do not resolve its break radius from our images, a shallow central profile was resolved in the WFPC2 images (Byun et al. (1996)) which have slightly higher angular resolution than the NICMOS images. We therefore list it as a core galaxy. ### 2.4 Ellipticity and boxiness reduction near the break radius In Fig. 8 we show trends observed in ellipticity and boxiness near the break radius for the core galaxies. Core galaxies tend to exhibit a reduction in both boxiness (when boxiness is exhibited) and ellipticity between about two times the break radius and the break radius. No galaxy is observed to be boxy within in its break radius. Many of the cores are well resolved so this is not caused by smoothing due to the point spread function. There are some galaxies which stand out from this pattern, however. NGC 7052 clearly contains a gas disk which is probably affecting measurement of the B4 component describing boxiness or diskiness. NGC 1600 gains ellipticity and has disky (positive B4) isophotes within its core. We suspect that this galaxy might contain a weak stellar disk. Two processes are predicted to reduce ellipticity and boxiness: 1) the stochasticity caused by the central black hole (proposed and explored by Merritt & Valuri (1996)) and 2) scattering from a binary black hole. We consider whether the morphology change in the stellar isophotes near the break radius is consistent with the black-hole scattering model for the formation of the cuspy or shallow core. Boxiness is a symptom of an uneven distribution function of stellar orbits in phase space (Binney & Petro (1985)). In other words this function is peaked around orbits with a narrow range of shapes. A scattering process would be likely to smooth the distribution function in phase space and so reduce boxiness. After scattering, the angular momentum of a particle will have changed, so we expect the distribution of scattered stars to be closer to spherical. A population of scattered stars should therefore reduce both boxiness and ellipticity. A binary black hole will scatter stars at a particular mean escape velocity (e.g. Quinlan (1996)) which depends on the binary semi-major axis and the binary mass ratio. This particular velocity should then manifest as a particular length scale over which a change in the isophote shapes is observed. This would naturally result in a particular range of radius over which we see a reduction in ellipticity and boxiness. This would correspond to the region of order a few times the break radius over which we observe the change in boxiness and ellipticity in these two galaxies. We now discuss an alternate possibility whereby the isophote change is instead caused by the the stochasticity induced by the black hole (Merritt & Valuri (1996); Merritt & Quinlan (1998)). This mechanism reduces triaxiality (which might be consistent with the observed boxiness) but would not likely account for such an association with the break radius or a large change in ellipticity as well as boxiness. We would expect this mechanism to result in a smooth isophotal shape variation with radius since the diffusion timescale is primarily dependent on the local dynamical time (Ryden (1998)). Since we observe a shape change over a small region (a few times the break radius), we find that scattering from a black hole binary gives a more natural explanation for the isophotal shape changes observed near the break radius. However stochasticity induced by the central black hole would still be a natural explanation for smooth shape changes observed over larger scales (Ryden (1998); Bender & Mehlert (1998)). ### 2.5 Core properties vs dust content The ability to observe these galaxies at 1.6 microns allows us to study galaxies which were excluded from visible band studies precisely because of their dust content (e.g., NGC 4150 was excluded from Faber et al. (1997) precisely for this reason). This allows us to investigate the possibility that core properties are related to dust or cold gas content. If dust is long lived in elliptical galaxies then a high dust content may indicate a previous gas rich merger which might have resulted in a binary black hole with more extreme binary black hole mass ratio and a smaller core or break radius. Alternatively if black hole coalescence requires Gyrs to take place (Begelman, Blandford & Rees (1980)) then we might expect recently formed elliptical galaxies to have growing cores, again suggesting that dusty galaxies should have smaller core or break radii. However, if an elliptical galaxy gains sufficient gas then star formation can occur within its core (e.g., NGC 7052), a process which might increase the central density and reduce the core size. However as we show in Fig. 9 we see no correlation between dust content (as measured from IRAS far-infrared emission) and core size. This suggests that the dust or cold gas we now see in elliptical galaxy is acquired subsequent to core formation and does not strongly affect the surface brightness profile. This is consistent with the lack of correlation between dust quantity and galaxy properties such as luminosity (Goodfrooij & de Jong (1996)), and the lack of alignment of dust features with respect to galaxy axes which suggests that dust must be acquired in elliptical galaxies on relatively short timescales (van Dokkum & Franx (1995)). Since we see no strong color gradients in our color maps we find no evidence for a diffuse dust component that is distributed in a different way than the light density. If such a component exists it must be extremely diffuse and similarly distributed as the star light (Goodfrooij & de Jong (1996)). ### 2.6 Core properties vs X-ray luminosity We also investigate the possibility that X-ray luminosity may be related to elliptical core properties (see Fig. 10). X-ray and optical luminosities of early-type galaxies are correlated with $`L_XL_B^{2.0\pm 0.2}`$ (Eskridge, Fabbiano & Kim (1995)). However the scatter about this relation is enormous with $`L_X/L_B`$ varying by factors of 500. There are two approaches towards accounting for this large scatter: 1) evolution models for production of hot gas via supernovae (e.g., Ciotti et al. (1991)) and 2) consideration of environmental affects such as ram pressure stripping (e.g., white). In either scenario the formation and evolution of the core might be related, so we might expect a correlation between core properties and X-ray luminosity. However, we fail to see any strong correlation between core properties and X-ray luminosity or $`L_X/L_B`$ (see Fig. 7). In our sample we find a core galaxy in a poor group, NGC 524, which moderate X-ray flux as well as X-ray bright cluster ellipticals. The lack of correlation suggests that the process of core formation is subsequently unaffected by whatever processes determine the X-ray luminosities of galaxies. ### 2.7 Core properties vs environment Our sample combined with that of Faber et al. (1997) contain galaxies which span a range of environments from poor groups to moderately sized clusters such as the Virgo and Fornax Clusters. We find poor groups with brightest members with cores (NGC 524) and poor groups with brightest members without cores (NGC 1553, NGC 2907, NGC 5198, NGC 821, NGC 1172, NGC 1400, NGC 7626). (We have identified cluster membership based on Garcia (1993), Faber et al. (1989) and references therein.) Almost all the brighter galaxies ($`M_V<20.5`$) in Virgo and Fornax contain cores (see the histograms presented in Fig. 11). However galaxies of this luminosity or greater without cores appear to be common in poorer environments (e.g., NGC 821, NGC 1172, NGC 1400, NGC 1553, NGC 1700, NGC 3115, NGC 4594, NGC 4697, NGC 5198). As we see in Table 3 and Fig. 7 these are not primarily distant galaxies which would have cores that are unresolved (except possibly in the case of NGC 5198). Since the scatter from the fundamental plane for these galaxies (Faber et al. (1997)) is much smaller than a magnitude, distance errors are unlikely to account for the difference in the histograms between the Virgo and Fornax cluster galaxies and the whole sample (Figs. 11a and 11b). This suggests that poor galaxy groups can harbor more luminous power law galaxies than clusters. This is an interesting possibility that should be investigated further with bigger samples. ## 3 Summary and Discussion In this paper we have presented a NICMOS imaging study of early-type galaxies. In moderately dusty galaxies these images allow us to measure stellar surface brightness profiles more accurately than possible with visible band images which are more strongly affected by extinction from dust. Discrepancies between core classification and measurement of core properties between our NICMOS images and previous visible band based HST studies are primarily due to dust features in the nuclear regions of the galaxies. We observe a trend in boxiness and ellipticity in the core galaxies. Both boxiness and ellipticity are reduced near and within the break radius. No galaxy is observed to be boxy within its break radius. This is consistent with a core formation mechanism that involves scattering of stars (such as scattering from a binary black hole). We failed to find correlations between core break radius and dust content or X-ray luminosity. This suggests that the current cold or hot gas content of an elliptical galaxy is unrelated to the process of core formation. The gas content is then more likely to be determined by processes, such as minor mergers and cooling flows, that would occur after the formation of the galaxy or core. By combining our sample with that of Faber et al. (1997) we find that galaxies from the Virgo and Fornax clusters (together) show a dichotomy of core types which is strongly dependent on luminosity. However, the dependence of core type on luminosity may be weaker in the complete sample including galaxies outside of the Virgo and Fornax Clusters. In particular higher luminosity power law galaxies may be more common in poorer environments. Since both core classification and cluster identification tequniques are strongly dependent on the galaxy distance, care must be taken to ensure that the more luminous power law galaxies are not the most distant. A larger sample of galaxies (such as is now available in the HST archive) could more thoroughly probe the relation between core properties and environment. Support for this work was provided by NASA through grant number GO-07886.01-96A from the Space Telescope Institute, which is operated by the Association of Universities for Research in Astronomy, Incorporated, under NASA contract NAS5-26555. We also acknowledge support from NASA project NAG-53359. We acknowledge helpful discussions and correspondence with A. Alonso-Herrero, M. Elvis, M. Rieke, G. Rieke, S. Stolovy, N. Caldwell and J. Bechtold. We thank M. S. Roberts for providing catalogues to us in electronic form.
no-problem/9907/astro-ph9907445.html	ar5iv	text	# A Numerical Experiment of a Triple Merger of Spirals ## 1. Introduction Several studies have addressed the topic of spiral interactions in pairs or in a small group (Barnes 1998). However there is no specific study, to our knowledge, of interacting triplets of spirals. Here we present a preliminary report of an ongoing research on the dynamics of triplets that includes galaxies of different morphology. In particular we follow a triplet of equal-mass spirals up to a merger remnant that resembles an elliptical-like galaxy has been formed. It is known that the HI component of spirals is a good tracer of interactions. Indeed, several triplets (e.g. in M81 group) that do not appear to be perturbed in optical images do so in HI maps. Although a self-consistent treatment of gas requires the inclusion of e.g. pressure forces, we use here a test-particle approach to study it during a triple interaction of spirals; this can yield information about the large-scale kinematical features of gaseous tails. ## 2. The Numerical Experiment We use Galaxy-like spirals with total mass $`M5.8\times 10^{11}`$M, $`R_{\mathrm{halo}}135`$ kpc, and disc mass $`M_d4.4\times 10^{10}`$ M (Kuijken & Dubinski 1995, Model B). For computational economy, the number of particles used for each spiral were: $`N_{\mathrm{bulge}}=1000`$, $`N_{\mathrm{disc}}=5000`$ and $`N_{\mathrm{halo}}=12000`$. In isolation the spiral was relatively stable for a few orbital periods. Discs’ orientations were obtained from three random sets of Euler angles (some animated gifs of the simulations are at http://www.iaa.es/ae/triplets.html). The initial positions and bulk velocities of galaxies were obtained from one cold-collapse simulation of spherical galaxy triplets, in which a relatively rapid triple merger was found (Aceves 1999). We started the simulation at $`t_02`$ Gyr from turn-around, when two of the galaxies G1 & G2 had their halos just overlapping. These galaxies were approaching at a relative velocity of $`140`$ km/s; the third galaxy’s (G3) halo had not yet ‘touch’ the binary’s common halo and was infalling at $`40`$ km/s. In the following, times quoted will be the elapsed time since this $`t_0`$. The simulation lasted for another $`12`$ Gyr. The ‘gas’ particles were distributed uniformly in the plane of the discs extending twice the stellar component. Forces on them were calculated using the mass interior to their position, with respect to each galaxy centre. The distribution of mass in galaxies was assumed to remain the same throughout the simulation and that it had a radial dependence. We consider this a somewhat better approximation than assuming that all the mass of a galaxy resides at its centre; no important extra computational effort was required in this approach. ## 3. Results In Fig. 1 we show a projection of galaxies G1 & G2 discs interacting at $`4`$ Gyr; galaxy G3 has not yet come close enough for its halo to be perturbed. The discs show large damages and a long stellar tidal tail has developed in G2 ($`150`$ kpc, in projection). The gaseous tails generated reach $`200`$ kpc in projection. Some stars and gas particles are returning to the ‘binary’ galaxy due to conservation of angular momentum and because they do not have enough kinetic energy to leave the local potential well of the binary. At $`10`$ Gyr galaxies G1 & G2 have already merged with G3; see Fig. 2. Galaxy G3 shows a very large projected stellar tidal tail of $`350`$ kpc, which is about the same size as its gaseous one. At this time the tidal tail of G2 has almost disappear, although important traces of it exist as a more diffuse component. The configuration at the end of the simulation is shown in Fig. 3. Enormous gaseous tails $`1`$ Mpc are present and the stellar tidal tail of G3 seen in Fig. 2 is still present. Since the gas thermodynamics has not been considered, it is probable that some ‘HI gas’ has been ionized by shocks developed during the interaction and, hence, this extension can be an upper limit; also, the initial size of the gaseous discs and the spiral matter profile determine this extent. After removing all unbound particles the shape of the merger was found to be prolate. No disc or bulge particles were formally found to be unbound, but galaxies G1 & G2 had lost $`3`$% of their halo mass and G3 $`10`$%. The merger remnant shows a luminous profile that resembles that of an elliptical, although with a rather isothermal core, its phase-space has retained more information about its spiral past; see Fig. 4. The physical size of the stellar tails reach $`500`$ kpc! Figure 5 shows velocity profiles for the gas, which would mimic the result of placing a narrow slit along the $`X`$-axis, both at $`4`$ Gyr (top) and 12 Gyr (bottom). Velocities along the l.o.s. of $`100`$ km/s in the outer parts with a width of $`(50,100)`$ km/s are obtained, respectively, from G1 & G2 at $`4`$ Gyr. At the end of the simulation G1 shows a broad dispersion in the velocities, while galaxies G2 & G3 show a more well defined width of $`50`$ km/s in the $`XV_z`$-plane. Tidal tails in G3 developed very late in the merging process. ## 4. Final Comments Some particular results suggested by the present simulation, with its specific cold-collapse initial conditions, tend to show that: * Encounters of spirals can effectively destroy stellar discs in $`<t_{\mathrm{Hubble}}`$. Large projected stellar tidal tails $`200`$ kpc can develop during a binary formation, and up to $`350`$ kpc in the 3rd. when this joins in. However tidal tails are not eternal, they can fall-back to the remnant of their ‘parent’ galaxies in $`<t_{\mathrm{Hubble}}`$, although important traces of them remain for $`>t_{\mathrm{Hubble}}`$. * Diffuse light, due to stars stripped from the discs, is to be expected in deep imaging of spiral interactions and mergers. However if this stripping occurred more than a stellar evolution time-scale ago one may not detected it, but stellar remnants may be in the field. * Gaseous tidal tails may extend to $`1`$ Mpc in radius, similar to the extent of the common dark halo formed, in a triple interaction in $`t_{\mathrm{Hubble}}`$. This result suggests that gaseous halos may trace the size of a common dark halo developed during galaxy interactions. * Although the stellar merger remnant shows an elliptical-like profile, it is more physically justifiable to look for merger signatures in velocity-space; albeit the observational difficulties. ### Acknowledgments. The author thanks the Spanish Government (MUTIS Program) and J. Perea (DGICYT Project PB96-0921) for financial support. ## References Aceves, H. 1999, these proceedings (astro-ph/9907226) Barnes, J.E. 1998, in Interactions and Induced Star Formation: Saas-Fee Advanced Course 26, eds. D. Friedli, et al. (Berlin: Springer-Verlag) Kuijken, K. & Dubinski, J. MNRAS, 277, 1341
no-problem/9907/astro-ph9907238.html	ar5iv	text	# The supercluster-void network V. ## 1 Introduction The basic assumption in the standard Friedman-Robertson-Walker cosmology is that the Universe is homogeneous and isotropic on large scales (Peebles p80 (1980)). Clusters of galaxies and galaxy filaments form superclusters (Abell abell (1958)); examples are the Local supercluster (de Vaucouleurs dv (1956), Einasto et al. ekss84 (1984)) and the Perseus-Pisces supercluster (Jõeveer et al. jet78 (1978), Giovanelli g93 (1993)). Superclusters and voids between them form a continuous network of high- and low-density regions which extends over the entire part of the Universe studied in sufficient detail up to a redshift of $`z0.13`$ (Einasto et al. me94 (1994), 1997d , hereafter Paper I). According to the conventional paradigm the location of superclusters in this network is random, as density perturbations on all scales are believed to be randomly distributed (Feldman et al. fkp (1994)). There exists growing evidence that the supercluster-void network has some regularity. Broadhurst et al. (beks (1990)) have measured redshifts of galaxies in a narrow beam along the direction of the northern and southern Galactic poles and found that the distribution is periodic: high-density regions alternate with low-density ones with a surprisingly constant interval of $`128`$ $`h^1`$ Mpc (here $`h`$ is the Hubble constant in units of 100 km s<sup>-1</sup> Mpc<sup>-1</sup>). The three-dimensional distribution of clusters shows clear signs of regularity (Paper I). One method to characterise this regularity is a correlation analysis. Kopylov et al. (kkf1 (1984), kkf2 (1988)), Fetisova et al. (fkl (1993)), Mo et al. (mo (1992)), and Einasto & Gramann (eg93 (1993)) have found evidence for the presence of a secondary peak of the correlation function of clusters of galaxies at 125 $`h^1`$ Mpc. The secondary peak has been interpreted as the correlation of clusters in superclusters located on opposite sides of large voids. Recent studies show that the correlation function oscillates with a period equal to that of the periodicity of the supercluster-void network (Einasto et al. 1997b , hereafter Paper II). An oscillation with very low amplitude is seen also in the correlation function of LCRS galaxies (Tucker et al. to97 (1997)). An oscillating correlation function corresponds to a peaked power spectrum (Einasto et al. 1997c , hereafter Paper III). Peacock (p97 (1997)) and Gaztañaga & Baugh (gb97 (1998)) determined the three-dimensional power spectrum from the projected distribution of APM galaxies; the Abell cluster power spectrum was derived by Einasto et al. (1997a ), and Retzlaff et al. (r98 (1998)); Tadros et al. (tadros (1998)) calculated the power spectrum for APM clusters. All these recent studies find the peak or turnover of the power spectrum near a wavelength of $`\lambda _0=120`$ $`h^1`$ Mpc or wavenumber $`k_0=2\pi /\lambda _0=0.05`$ $`h`$ Mpc<sup>-1</sup>. In addition, in the line-of-sight correlation function of Ly $`\alpha `$-break galaxies with redshifts $`z3`$ secondary peaks were found at a redshift separation of $`\mathrm{\Delta }z0.22\pm 0.02`$ (Broadhurst & Jaffe bj99 (1999)). This separation corresponds to the comoving scale $`120(1+z)^1`$ Mpc if a flat $`\mathrm{\Lambda }`$CDM isotropic cosmological model with $`\mathrm{\Omega }_m=1\mathrm{\Omega }_\mathrm{\Lambda }0.4\pm 0.1`$ is assumed. A low-density model dominated by a cosmological term fits nicely to many other recent data, in particular, those from high-$`z`$ supernovae (Perlmutter et al. perl98 (1998)). It should be noted that no such feature is seen in the power spectra of some galaxy catalogs, in particular, in the recent IRAS PSC redshift survey (Sutherland et al. s99 (1999)). However, we suppose that this may be the result of IRAS galaxies avoiding overdense regions like rich clusters (see more detailed explanation of this point by Einasto et al. e99 (1999)). Quantitative methods used so far to describe the regularity of the matter distribution are insensitive to the directional and phase information, and thus characterise the regularity only indirectly. In this paper we shall use a new geometric method to investigate the distribution of clusters of galaxies, sensitive to the regularity and the isotropy of the distribution. ## 2 Observational data As in previous papers of this series we use the 1995 version of the compilation by Andernach & Tago (at98 (1998)) of all published galaxy redshifts towards galaxy clusters in the catalogue of rich clusters of galaxies by Abell (abell (1958)) and Abell et al. (aco (1989)) (hereafter Abell clusters). Individual galaxies were associated to a given Abell cluster if they lay within a projected distance of $`1.5`$ $`h^1`$ Mpc (1 Abell radius) and within a factor of two of the redshift estimated from the brightness of the clusters 10-th brightest galaxy, using the photometric estimate of Peacock & West (pw92 (1992)). For the present analysis we used a sample of all rich clusters (richness class R$`0`$ and excluding clusters from ACO’s supplementary list of S-clusters) in this compilation with redshifts up to $`z=0.12`$. The sample contains 1304 clusters, 869 of which have measured redshifts for at least two galaxies. Distances of clusters without measured redshifts and of clusters with only 1 galaxy measured have been estimated on the basis of the apparent magnitude of the 10-th brightest galaxy of the cluster. As representatives of high-density regions in the Universe we used rich superclusters from the list of superclusters presented in Paper I. Superclusters were identified using the friend-of-friends algorithm by Zeldovich et al. (zes82 (1982)) with a neighbourhood radius of 24 $`h^1`$ Mpc. In this way all clusters of a supercluster have at least one neighbour at a distance not exceeding the neighborhood radius. To illustrate the distribution of clusters in high-density regions we plot in Fig. 1 only clusters in rich superclusters, while in the quantitative analysis below we shall use all clusters. The sheets plotted are 300 $`h^1`$ Mpc thick, thus some superclusters overlap in projection. The sheet in the left panel of Fig. 1 crosses the majority of cells of the supercluster-void network present in our cluster survey; the sheet in the right panel has the full depth of the sample in the southern Galactic hemisphere. Fig. 1 shows clearly the quasi-regular network of superclusters interspersed with voids. The three-dimensional distribution of all Abell and APM clusters in rich superclusters in the whole space within a limiting radius 350 $`h^1`$ Mpc around us is shown in the home page of Tartu Observatory (http://www.aai.ee). For comparison we plot in Fig. 1 also the distribution of APM clusters of galaxies in rich superclusters. We see that the APM cluster sample covers a much smaller volume in space which makes it difficult to investigate the regularity of the distribution of high-density regions on large scales. The APM cluster sample is defined only in the southern Galactic hemisphere, and even here the APM sample containing clusters with measured redshifts covers only three high-density regions defined by very rich superclusters with at least 8 member-clusters. These are the Sculptor (SC9), Pisces-Cetus (SC10), and Horologium-Reticulum (SC48) superclusters of the catalogue in Paper I. To investigate the regularity of the supercluster-void network, the sample volume must exceed the period of the network at least several times. For this reason we used in the following analysis only Abell clusters. ## 3 Method To investigate the regularity of the distribution of clusters of galaxies we shall use a novel geometric method. The idea of the method is taken from the periodicity analysis of time series of variable stars. To derive the period of a variable star the time series of brightness measurements is folded using a certain trial period, i.e. individual measurements are stacked to yield a combined light curve for a single period. If the trial period is wrong then the distribution of data-points in the combined curve is almost random. In contrast, if the trial period is correct and the star under study is a periodic variable, then one has a clear mean light curve of the star. This method has been generalised to the 3-dimensional case by Toomet (ott (1997)). The whole space under study is divided into trial cells of side length $`d`$. Then all objects in the individual cubical cells are stacked to a single combined cell, preserving their phases in the original trial cells. We then vary the side length of the trial cube to search for the periodicity of the cluster distribution. The method can be used for any form of tightly packed regular trial cells. We use cubical trial cells which have the simplest possible form, and because this form gives a satisfactory explanation of the distribution of clusters. The next step is to calculate the goodness of regularity for this side length $`d`$ of the trial cell. To do this we divide the combined stacked cell into small elementary sub-cells and count, for a particular data point (i.e. cluster of galaxies) within this sub-cell, the number of other data points which come from an original trial cell different from the one in which the given data point is located. The goodness of regularity is calculated as follows: $$k(d)=\frac{2}{N(N+1)}\frac{V(d)}{V_\epsilon }\underset{i=1}{\overset{N}{}}\underset{j=i+1}{\overset{N}{}}v(f_i,f_j).$$ (1) Here $`N`$ is the total number of data points, i.e. individual Abell clusters; $`Vd^3`$ is the volume of the trial cell of side length $`d`$; $`f_i`$ is the phase-vector of the cluster $`i`$ in the combined cell (a spatial remainder-vector, the components of which are remainders, which arise if the original coordinates are divided by $`d`$); $`V_\epsilon `$ is the elementary volume inside of the combined cell. We use an elementary cube (sub-cell) of side length $`2d_\epsilon `$. The counter $`v(f_i,f_j)`$ is defined as follows: $`v_{ij}=1`$, if the phase vector of the cluster $`j`$ lies in the elementary volume $`V_\epsilon `$ (the centre of which coincides with $`f_i`$), and if clusters $`i`$ and $`j`$ are located farther away from each other than $`d_\epsilon `$. Otherwise we set $`v_{ij}=0`$. Thus $`v_{ij}`$ is unity only if the phases of clusters are close enough in the combined cube, but the clusters themselves are not members of the same supercluster (they lie in different trial cubes of the original space). The sensitivity of the method and the rms error of the variable $`k(d)`$ have been determined using mock samples of randomly and quasi-regularly located points. We use mock samples with two populations of clusters, in one population all clusters are randomly distributed, in the second population they are located in superclusters which form a regular rectangular network with a step 130 $`h^1`$ Mpc. Supercluster centre positions have random shifts $`\pm 20`$ $`h^1`$ Mpc around corners of the regular network. Mock samples are described in more detail in Paper III. Here we are interested in the cosmic error due to variance of $`k(d)`$ in different realizations of the sample. This relative error can be expressed as $`\sigma _kb(d2d_\epsilon )/N`$, where $`N`$ is the number of clusters in the sample, and $`b0.5`$ is a constant determined from the analysis of mock samples. The dependence of the error on $`d`$ comes from the fact, that, if $`d=2d_\epsilon `$, then trial cells have the same size as elementary sub-cells. Thus all clusters of the trial cell are located in the elementary sub-cell and there is no variance of $`k(d)`$ between different realizations; for larger trial cell size $`d`$ the error increases. The presence of a periodicity in the sample can be determined with confidence if the amplitude of the goodness near the maximum deviates from the Poisson value $`k=1`$ by more than $`2\sigma `$. Results of the analysis of mock samples are illustrated in Figure 3. We see in the right panel that the goodness curve has well defined maxima at integer multiples of $`r_0=130`$. In the mock sample shown in the left panel the fraction of clusters located in regular network is smaller and the amplitude of the maximum lies within the $`2\sigma `$ error corridor. This example shows that the period of a regular network of points can be determined if the population of regularly distributed points is sufficiently large (at least 10 % of the total number of points) and if the total number of particles is at least several hundred. Obviously, the method is sensitive to the direction of the axes of the trial cubes. If clusters form a quasi-rectangular cellular network, and the search cube is oriented along the main axis of the network, then the period is found to be equal to the side length of the cell. If the search cube is oriented at some non-zero angle $`\theta `$ with respect to the major axis of the network, then the presence of a periodicity and the period depend on the angle. If the angle $`\theta 45^{}`$, then the period is equal to the length of the diagonal of the cell. If the angle differs considerably from $`\theta =0^{}`$ and $`45^{}`$, the periodicity is weak or absent. This property is also illustrated in Figures 3 and 4 for mock and real cluster samples, respectively. Results of the period analysis of our cluster sample are shown in Figure 4. The axes of the search cubes were oriented at various angles with the coordinate axes of the supergalactic coordinate system. As seen from these Figures, the period really depends on the orientation. The smallest value of the period, $`P=130`$ $`h^1`$ Mpc, occurs when the cubes are oriented along supergalactic coordinates. If the trial cube is oriented at $`\theta =45^{}`$ with respect to supergalactic coordinates, then the period is $`P=190`$ $`h^1`$ Mpc, as expected for a rectangular network of cells. In both cases there exist a second maximum of the curve of goodness of regularity at $`d=250280`$ $`h^1`$ Mpc which is also well pronounced. If the trial cube is oriented at $`22^{}`$, then the periodicity is much weaker, the first maximum at $`d=120`$ $`h^1`$ Mpc has a lower amplitude, and the second is absent. In principle there are more possibilities, for instance when the trial cells are oriented along the diagonal of the cubical lattice. However, in most additional cases the periodicity is weaker. Probably this can be due to the fact the the regularity is not very strong and there exist deviations from it. Much larger samples are needed to find limits of the regularity. ## 4 Discussion and conclusions The periodicity analysis confirms the analysis made in Papers I and II using other methods. The cluster sample has a considerable fraction of clusters located in rich superclusters which form a rectangular lattice with a period of about $`120130`$ $`h^1`$ Mpc. Our analysis shows that the supercluster-void network is oriented approximately along supergalactic coordinates. This confirms earlier results on the presence of a high concentration of clusters and superclusters towards both the Supergalactic Plane (Tully et al. tu3 (1992)), and towards the Dominant Supercluster Plane, which are at right angles with respect to each other (Paper I). Supergalactic $`Y`$ axis is very close to the direction of Galactic poles, thus it is natural to expect a well-defined periodicity along Galactic poles as indeed observed by Broadhurst et al. (beks (1990)). The rectangular character of the distribution of rich clusters was also noticed by Tully et al. (tu3 (1992)). Recently Battaner (b98 (1998)) has found that many known superclusters can be identified with the vertices of a octahedron network of the superstructure. One principal result of our study is the direction dependence of the periodicity. This conclusion was not possible with the power spectrum or correlation function approach since they are not sensitive to directional information. A clear periodicity is observed only along supergalactic coordinates. This is in good agreement with results by Broadhurst et al. (beks (1990)) where a periodicity was observed in the direction of Galactic poles, i.e. almost exactly along the supergalactic $`Y`$ axis (supergalactic $`X`$ and $`Z`$ axes lie close to the galactic plane where clusters are invisible). In other directions the regularity is less pronounced (Guzzo et al. gcnl (1992), Willmer et al. wksk (1994), Ettori et al. egt (1997)). As noted already by Bahcall (b91 (1991)), the nearest peaks of the Broadhurst et al. survey coincide in position and redshift with nearby rich superclusters. Thin deep slices, such as slices of the LCRS and the Century Survey, also show a weak periodicity signal (Landy et al. ls96 (1996), Geller et al. g97 (1997)). The scale length found in these studies is of the same order as derived in the present paper. The regularity found in this paper concerns high-density regions marked by clusters in rich superclusters. When we compare the distribution of galaxies in various regions of the supercluster-void network we see considerable individual differences. On small scales the distribution of galaxies and clusters is fractal, as shown by many studies (see Wu et al. wu99 (1999)); and is well reproduced by models with randomly distributed density perturbations (see Feldman et al. fkp (1994)). On very large scales covering the whole volume of the Abell cluster sample we see in the distribution of clusters in extremely rich superclusters with $`N>30`$ members no regularity. We come to the conclusion that a clear regularity is observed only near the wavelength $`\lambda _0120`$ $`h^1`$ Mpc of the peak of the power spectrum. This result can be expressed as follows: density perturbations on very small ($`kk_0`$) and very large wavelengths ($`kk_0`$) are uncorrelated and have random phases; however, near the wavelength $`\lambda _0=2\pi /k_0`$ they are correlated and have similar phases. The importance of the use of phase information was stressed by Szalay (szalay (1998)). The regularity of the large-scale structure of the Universe has been studied also using the distribution of centres of superclusters (Kalinkov et al. kal (1998), Kerscher ker (1998)). Kalinkov et al. searched for the presence of a high-order clustering of superclusters using the correlation analysis, and found that superclusters are not clustered. Our analysis in Papers I and II and presented here suggests that the regularity is completely different: it consists of the presence of the supercluster-void network, not the clustering of superclusters (like galaxies concentrate to clusters). The supercluster-void regularity is expressed by the distribution of clusters themselves, not supercluster centres. Kerscher used a combination of the nearest neighbour distribution and the void probability function to measure the regularity of the structure. This method works well for small separations between superclusters ($`r60`$ $`h^1`$ Mpc), and yields results in good agreement with Paper I. However, the method does not characterise the supercluster-void network on larger scales. It should be emphasized that this inhomogeneity in the Universe does not contradict the observed degree of isotropy of the cosmic microwave background (CMB) radiation. As follows from general expressions for rms values of multipoles $`C_l`$ of CMB angular temperature anisotropies (see, e.g., Starobinsky st88 (1988)), any additional localized excess of power in the Fourier spectrum of density perturbations introduced at a scale $`k_01/R_H`$ affects only multipoles $`l<l_0=k_0R_H`$ where $`R_H`$ is the present horizon radius. This is valid for both Gaussian and non-Gaussian additional perturbations (assuming only that there is no correlation between perturbations with $`kk_0`$ and perturbations from the other, “regular” part of the spectrum). Moreover, multipoles with $`l=(0.70.9)l_0`$ are mainly amplified. For the pure CDM cosmological model with $`\mathrm{\Omega }_m=1`$, $`R_H=2c/H_0`$ and $`l_0=300`$ mainly the first acoustic peak is enhanced (Atrio-Barandela et al. a97 (1997); Eisenstein et al. Eis98 (1998); Broadhurst & Jaffe bj99 (1999)). However, in the $`\mathrm{\Lambda }`$CDM model (which is strongly supported by numerous recent observational data) this enhancement shifts to larger values of $`l`$. In particular, if $`\mathrm{\Omega }_m=1\mathrm{\Omega }_\mathrm{\Lambda }=0.3`$, then $`R_H=3.305c/H_0`$, and $`l_0500`$. For this model, the effect discussed in this paper mainly results in an increase of $`C_l`$ in the valley between the first and second acoustic peaks of the CMB temperature anisotropy spectrum. In particular, the amplitude $`(\mathrm{\Delta }T/T)_l=\sqrt{C_ll(l+1)/2\pi }`$ for $`l=400`$ will be $`2\times 10^5`$, in agreement with recent CAT2 data (Baker et al. b99 (1999)), as well as with previous CAT1 results (Scott et al. s96 (1996)). A more detailed treatment of resulting CMB anisotropies in the presence of a cosmological constant will be presented in a separate paper. The main conclusion we can draw from our study is that the Universe is not homogeneous and fully isotropic on scales $`120`$ $`h^1`$ Mpc. High-density regions in the Universe form a quasi-rectangular lattice. The distribution of high-density regions depends on the direction: along the main axis (coinciding with the supergalactic $`Y`$ axis) of the cellular system the regularity is well pronounced and has a period of $`120130`$ $`h^1`$ Mpc, while in other directions the distribution is less regular. ###### Acknowledgements. We thank Jaan Pelt for discussion and suggestions. The present study was supported by Estonian Science Foundation grant 2625. A.S. was partially supported by the grant of the Russian Foundation for Basic Research No. 99-02-16224. This paper was finished during his stay at the Institute of Theoretical Physics, ETH, Zurich.
no-problem/9907/physics9907049.html	ar5iv	text	# Error propagation in the hypercycle ## I Introduction The limitation of the length of a genome by the replication accuracy per nucleotide ($`q`$) has led to a deadlock in the theories of the origin of life based on the evolution of competing self-replicating polynucleotides. According to Eigen’s quasispecies model , which may serve as a paradigm here, polynucleotides have to replicate with high accuracy in order to reach a certain length, a requirement that is impossible to fulfill without the aid of specialized catalysts. However, to build those catalysts a blueprint is necessary that amounts to a large genome (the nucleotide sequence), which itself cannot be maintained without the catalysts. In particular, for polynucleotides of fixed length $`L`$, the quasispecies model predicts the existence of a minimal replication accuracy per genome $`Q_c=q_c^L`$, below which the genetic information is irreversibly lost. This information crisis has been termed error threshold transition. Above $`Q_c`$ the population is composed of a master copy together with a cloud of structurally similar mutants (quasispecies) . Equally important is the finding that, except in a trivially degenerate case, two or more quasispecies cannot coexist , thus precluding the coexistence of templates (i.e. polynucleotides) sufficiently different from each other to code for any useful set of catalysts. Although it has been claimed that the information crisis is not really a fundamental issue, since the error threshold transition appears only in some pathological, discontinuous replication landscapes , the coexistence problem seems to be more pervasive, as it is associated to the form of the growth functions in the chemical kinetics equations . In order to circumvent the aforementioned limitations of the quasispecies model, Eigen and Schuster proposed the hypercycle , that is, a catalytic feedback network whereby each template helps in the replication of the next one, in a regulatory cycle closing on itself. This model has gain plausibility when the ability of polynucleotides to help propagate each other was established experimentally through the study of the catalytic activity of the RNA (ribozymes) . Interestingly, though the error threshold phenomenon has traditionally been considered the main motivation for the proposal of the hypercycle (see , for instance), most of the seminal works in this field have dealt with the coexistence issue only, as they assume perfect replication accuracy for the hypercycle elements . In this case an arbitrary number of templates permanently coexist in a dynamical equilibrium state; if $`n>4`$, however, the template concentrations vary with time , periodically decreasing to very small values. In practice, large hypercycles are therefore susceptible to extinction via fluctuations, see e.g. , hence the information gain due to the coexistence of different templates in the hypercycle may not be very impressive after all. Furthermore, we will argue in this paper that coexistence in the absence of a stable equilibrium can also be achieved by a simpler arrangement, namely, the free chains, in which the cyclic order of the catalysts is interrupted. The effect of error-prone replication (mutation) in the hypercyclic organization was investigated by introducing a mutation field as a perturbation of the error-free kinetic equations . This approach, however, is not very appropriate to study the error threshold phenomenon, since the results obtained cannot be easily compared with those of the quasispecies model. In this sense, a better approach is to assume the existence of a special class of templates with no catalytic activity, so-called error-tail, that appear as a consequence of the replication errors of the hypercycle elements . However, the particular catalytic network investigated extensively within that framework was not the hypercycle, except for a short discussion in , but the fully connected network in which each element helps the replication of all other elements of the network . (Clearly, in the case of $`n=2`$ elements, these two networks become identical .) Such a network is more robust than the hypercycle since the mal-functioning or extinction of one of its elements does not compromise the whole network. Nevertheless, besides its aesthetic appeal, the cyclic coupling of the hypercycle seems to be more realistic . The goal of this paper is to investigate analytically the steady-states of a deterministic system comprised of two parts, namely, a hypercycle made up of $`n`$ self-replicating templates $`I_1,I_2,\mathrm{},I_n`$ and its error tail $`I_e`$. These parts are coupled such that any erroneous copy of the hypercycle elements will belong to the error tail. The focus of the present analysis is on the location in the parameters space of the model (i.e. replication accuracy per template, non-catalyzed and catalyzed productivity values, and hypercycle size) of the regions of stability of the diverse possibilities of coexistence between the templates composing the hypercycle. In particular, we give emphasis to the characterization of the critical parameters at which the hypercycle becomes unstable against the error tail. The remainder of the paper is organized as follows. In Sec. II we present the chemical kinetics equations that govern the time evolution of the system and motivate the specific choice of the parameters used throughout the paper. The fixed-points of the kinetic equations are obtained analytically in Sec. III and their stability discussed in Sec. IV. The phase-diagrams showing the regions of stability of the diverse coexistence states are presented and analyzed in Sec. V. Finally, some concluding remarks are presented in Sec. VI. ## II The model We consider a system composed of a hypercycle made up of $`n`$ elements $`I_1,\mathrm{},I_n`$ and its error tail $`I_e`$, as illustrated in Fig. 1. In contrast to the so-called elementary hypercycle , we assume that the templates are capable of self-replication with productivity values $`A_i\left(i=1,\mathrm{},n\right)`$ and $`A_e`$. Moreover, as usual, the growth promotion of template $`I_i`$ as a result of the catalysis from template $`I_{i1}`$ is measured by the kinetic constants $`K_i`$. The key ingredient in the modeling is that in both processes of growth of template $`I_i`$ the probability of success is given by the parameter $`Q[0,1]`$, so that an erroneous copy, which will then belong to the error tail, is produced with probability $`1Q`$. Hence the concentrations $`x_i\left(i=1,\mathrm{},n\right)`$ of the hypercycle elements and the concentration $`x_e`$ of the error-tail evolve in time according to the kinetic equations $$\dot{x}_i=x_i\left(A_iQ+K_ix_{i1}Q\mathrm{\Phi }\right)i=1,\mathrm{},n$$ (1) and $$\dot{x}_e=x_e\left(A_e\mathrm{\Phi }\right)+\left(1Q\right)\underset{i=1}{\overset{n}{}}x_i\left(A_i+K_ix_{i1}\right)$$ (2) where $`x_0x_n`$ and $$\mathrm{\Phi }=\underset{i=1}{\overset{n}{}}x_i\left(A_i+K_ix_{i1}\right)+A_ex_e$$ (3) is a dilution flux that keeps the total concentration constant, i.e., $`_{i=1}^n\dot{x}_i+\dot{x}_e=0`$. As usual, the dot denotes a time derivative. Henceforth we will assume that $$\underset{i=1}{\overset{n}{}}x_i+x_e=1.$$ (4) Clearly, this formulation is equivalent to considering polynucleotides of length $`L\mathrm{}`$ whose replication accuracy per nucleotide $`q`$ goes to 1 such that the replication accuracy per genome is finite, i.e. $`q^LQ`$. In this limit, the back-mutations from the error-tail elements to the templates that compose the hypercycle, as well as the mutations between those templates, can be safely neglected. Hence, mutations can only increase the concentration of the elements in the error-tail. The advantage of working in this limit is that the error threshold transition can be precisely located by determining the value of $`Q`$ at which the concentration of a relevant template vanishes. For finite $`L`$, as well as for finite population sizes, the characterization of this transition is more involved, being achieved through the use of finite-size scaling techniques . In this work we consider the single-sharp-peak replication landscape , in which we ascribe the productivity value $`A_1=a>1`$ to the so-called master template $`I_1`$ and $`A_i=A_e=1`$ to the $`n1`$ other elements of the hypercycle as well as to the error-tail. Also, for the sake of simplicity we set $`K_i=K`$ for all $`i`$. The motivation for this particular choice of parameters is the observation that the emergence of the hypercycle requires both spatial and temporal coexistence of the templates forming the network, and this can be achieved by a quasispecies distribution, which guarantees the coexistence of the master template and its close mutants, despite the purely competitive character of the quasispecies model . Once the coexistence is established, the appearance of catalytic couplings between the templates is not a very unlike event. Of course, as soon as those cooperative couplings become sufficiently strong to balance the competition imposed by the constant concentration constraint, the mutants will certainly depart from the master template due to the relentless pressure of the mutations, so that no trace will remain of the original quasispecies distribution. ## III Fixed points Let us distinguish between surviving templates $`x_i>0`$ and extinct templates $`x_i=0`$. A survivor $`I_j`$ is said isolated if $`x_{j1}=x_{j+1}=0`$. Hence, $$\dot{x}_j=x_j\left(Q\mathrm{\Phi }\right)j>1$$ (5) and $$\dot{x}_e=x_e(1\mathrm{\Phi })+(1Q)\left[x_1\left(a1\right)+1+K\underset{ij,j+1}{}x_ix_{i1}\right].$$ (6) In the steady-state regime, Eq. (5) yields $`\mathrm{\Phi }=Q`$ which, for $`Q<1`$, is incompatible with Eq. (6), since the term within brackets in this equation is positive. Therefore, all isolated survivors with the exception of the master template are unstable against the error tail. Next consider the following chain of surviving templates: $$\dot{x}_i=x_i(Q\mathrm{\Phi })$$ (7) $$\dot{x}_{i+1}=x_{i+1}(Q+KQx_i\mathrm{\Phi })$$ (8) $$\dot{x}_{i+2}=x_{i+2}(Q+KQx_{i+1}\mathrm{\Phi })$$ (9) $$\mathrm{}$$ (10) $$\dot{x}_k=x_k(Q+KQx_{k1}\mathrm{\Phi })$$ (11) which does not contain $`x_1`$. Again, in the steady-state regime the first equation yields $`\mathrm{\Phi }=Q`$, implying $`KQx_i=0`$, i.e., $`x_i=0`$. So, there is no fixed point corresponding to such a chain. Any chain of survivors therefore must start with template $`n`$ or $`1`$. In the first case we get $`\mathrm{\Phi }=Q`$ from $`\dot{x}_n=0`$ and $`aQ+KQx_n=Q`$ from $`\dot{x}_1=0`$, yielding $`x_n=(1a)/K<0`$, which rules out this possibility. The equilibria of interest for our study thus are either the interior equilibrium in which all templates survive, or a fixed point that corresponds to a chain of survivors beginning with $`I_1`$. Accordingly, we define a $`m`$-coexistence state as the $`n`$-component template vector $`𝐱=(x_1,x_2,\mathrm{},x_n)`$ in which the first $`m`$ components are strictly positive and the rest are equal to zero. Clearly, given the template vector $`𝐱`$ the concentration of the error tail $`x_e`$ is determined by the constraint (4). In the following we solve analytically the kinetic equations in the steady-state regime $`\dot{x}_i=0i`$. The simplest fixed point is the zero-coexistence state ($`m=0`$) which corresponds to the solution $`x_1=\mathrm{}=x_n=0`$ and $`x_e=1`$, existing for the complete range of parameter values. In the case of chains, i.e. $`0<m<n`$, the steady-state solutions of Eqns. (1) and (2) are straightforward. In fact, since $`x_n=0`$ by definition, we get $`\mathrm{\Phi }=aQ`$ from $`\dot{x}_1=0`$ which then yields $$x_1=x_2=\mathrm{}=x_{m1}=\frac{a1}{K}.$$ (12) Next we insert this result in Eq. (3) to obtain $$x_m=\frac{Qa1}{a1}\left(m1\right)\frac{a1}{K}.$$ (13) However, since $`x_i(0,1)i`$, this solution is physically meaningful in the region $`K>a1`$ and $`Q>Q_m`$ where $$Q_m=\frac{1}{a}+\frac{1}{Ka}\left(m1\right)\left(a1\right)^2.$$ (14) We note that the $`1`$-coexistence state (quasispecies) is obtained by setting $`m=1`$ in Eq. (13) and its region of existence is simply $`Q>1/a`$, since the other condition, namely $`K>a1`$, is derived by considering the other templates in the chain. In fact, this very simple result quantifies nicely the notion that the cooperative couplings must reach a certain minimum strength so as to balance the competition between templates. The analysis of the hypercycle, i.e. $`m=n`$, is a little more involved. ¿From $`\dot{x}_2=0`$ we get $`\mathrm{\Phi }Q=KQx_1`$ which, inserted in the equations $`\dot{x}_3=\mathrm{}=\dot{x}_n=0`$, yields $`x_1=x_2=\mathrm{}=x_{n1}`$ and $$x_n=x_1\frac{a1}{K}.$$ (15) Finally, using these results in Eq. (3) we find that $`x_1`$ is given by the roots of the quadratic equation $$nKx_1^2\left(KQ+a1\right)x_1+1Q=0.$$ (16) For $`K<\left(a1\right)^2/4n`$, this equation has real roots for all $`Q0`$, otherwise it has real roots for $`QQ_h`$ where $`Q_h`$ is the unique positive root of the equation $$K^2Q_h^2+2K\left(a1+2n\right)Q_h+\left(a1\right)^24nK=0.$$ (17) In particular, for large $`K`$ we find $`Q_h2\sqrt{n/K}`$. Furthermore, it can be easily seen from Eqns. (15) and (16) that $`x_n`$ vanishes at $`Q=Q_n`$ with $`Q_n`$ given in Eq. (14). To understand the role of $`Q_h`$ and $`Q_n`$ (we note that $`Q_nQ_h`$) in delimiting the region of existence of the $`n`$-coexistence state we must look at the behavior of the two real roots of Eq. (16). Let us denote them by $`x_1^+`$ and $`x_1^{}`$ with $`x_1^+x_1^{}`$, which, according to Eq. (15), correspond to $`x_n^+`$ and $`x_n^{}`$, respectively. Of course, these roots become identical at $`Q=Q_h`$ and so the two solutions for $`x_n`$ will vanish simultaneously only at the value of $`K=K_h`$ at which $`Q_h`$ equals $`Q_n`$. Explicitly, we obtain $$K_h=\left(a1\right)\left[n\left(a+1\right)1\right].$$ (18) by inserting Eq. (14) into Eq. (17). Although both roots $`x_1^+`$ and $`x_1^{}`$ are in the simplex $`(0,1)`$, this is not so for $`x_n^+`$ and $`x_n^{}`$. In particular, for $`K<K_h`$ both concentrations are negative within the range $`Q_hQ<Q_n`$. However, while $`x_n^+`$ becomes positive for $`Q>Q_n`$ (it vanishes at $`Q_n`$), $`x_n^{}`$ remains always negative. Since $`K_h>\left(a1\right)^2/4n`$ the same conclusion holds in the range $`K<\left(a1\right)^2/4n`$ as well, provided we define $`Q_h=0`$ in this region. The situation is reversed for $`K>K_h`$: both concentrations are positive within the range $`Q_hQ<Q_n`$, but now it is $`x_n^{}`$ that vanishes at $`Q_n`$ and becomes negative for $`Q>Q_n`$ while $`x_n^+`$ remains always positive. Despite the small region in the parameters space where the root $`x_1^{}`$ yields concentrations inside the simplex, the linear stability analysis discussed in the sequel indicates that this solution is always unstable, so we only need to consider the root $`x_1^+`$. Thus the range of existence of the hypercycle fixed point $`m=n`$ is $`QQ_n`$ if $`KK_h`$ and $`QQ_h`$ if $`K>K_h`$. In models without error-tail, i.e., pure replicator equations a much stronger statement on coexistence is possible. The “time average theorem” states that if there is a trajectory along which a certain subset of templates $`J`$ survives, then there is a fixed point with exactly the $`J`$-coordinates non-zero. While we have not been able to prove the “time average theorem” in full generality for Eqns. (1) and (2), it is easily verified for free chains. Hence, if there is no $`m`$-coexistence equilibrium, then there is no trajectory at all along which the templates $`I_1`$ through $`I_m`$ survive. Hitherto we have determined the ranges of the parameter $`Q`$ where the $`m`$-coexistence states are physically meaningful, in the sense that $`x_i(0,1)i`$. The next step is to find the regions where these states are locally stable. ## IV Stability analysis In order to perform a standard linear stability analysis of the fixed points obtained in the previous section, it is convenient to rewrite the kinetic equations (1) and (2) as follows $$\dot{x}_i=x_iF_i\left(𝐱\right)i=1,\mathrm{},n$$ (19) where $$F_i(𝐱)=A_iQ+KQx_{i1}A_e\underset{j}{}x_j\left(A_jA_e+Kx_{j1}\right)$$ (20) and we have used the constraint (4) to eliminate $`x_e`$. The stability of a fixed point is ensured provided that the real parts of all eigenvalues of the $`n\times n`$ Jacobian $`𝒥`$ are negative. In our case the elements of the Jacobian are given by $$J_{ij}=\delta _{ij}F_i+x_i\frac{F_i}{x_j}i,j=1,\mathrm{},n.$$ (21) The evaluation of the eigenvalues is simple only for the zero-coexistence state, since in this case the Jacobian is diagonal with elements $`J_{11}=aQ1`$ and $`J_{ii}=Q1,i>1`$. Therefore this steady state becomes unstable for $`Q>1/a`$, which coincides with the lowest replication accuracy required for the existence of the $`1`$-coexistence state. However, for a general $`m`$-coexistence state we have to resort to a numerical evaluation of the Jacobian eigenvalues. Fortunately, in the case of chains $`0<m<n`$ there is an alternative way to look at the stability of the fixed points, as hinted by the stability analysis of the zero-coexistence state, which becomes unstable due to the emergence of the $`1`$-coexistence state. In fact, it can be easily seen that any perturbation of the $`m`$-coexistence fixed point which makes the concentration $`x_{m+1}`$ non-zero will be amplified if $`A_{m+1}Q+KQx_m\mathrm{\Phi }`$ is positive. For $`m>0`$ we use $`\mathrm{\Phi }=aQ`$ and $`A_{m+1}=1`$ together with the value of $`x_m`$ given in Eq. (13) to obtain the following (necessary) condition for the stability of the $`m`$-coexistence state, $$Q<Q_{m+1}m>0,$$ (22) with $`Q_m`$ given in Eq. (14). Hence the maximum value of $`Q`$ allowed for the stability of the $`m`$-coexistence state coincides with the minimum $`Q`$ required for the existence of the $`(m+1)`$-coexistence state. Interestingly, though for $`m=0`$ we have $`\mathrm{\Phi }=1`$, $`A_1=a`$ and $`x_0=0`$, condition (22) holds true in this case too. At this point two caveats are in order. First, the entire argument leading to the stability condition (22) is flawed if the $`(m+1)`$-coexistence state happens to be unstable. Therefore, we must guarantee via the numerical evaluation of the Jacobian eigenvalues that the $`l`$-coexistence state is stable before using that condition to study the stability of chains with $`m<l`$. In particular, we have carried out the numerical analysis for the hypercycle solution $`l=n`$ and found the following general results: * For $`n4`$, it is always stable; * for $`n=5`$, it is stable in a very small range of $`Q`$ above $`Q_5`$; and * for $`n6`$, it is always unstable. Second, the derivation of the stability condition (22) is based on the analysis of a single eigenvalue of the Jacobian and so it does not yield a sufficient condition for the stability of the fixed points. Nevertheless, we have verified through the numerical evaluation of all $`n`$ eigenvalues that, provided the $`(m+1)`$-coexistence state is stable, the eigenvalue associated to fluctuations leading to an increase of the chain length is the first one to become positive. ## V Discussion Combining the existence and the stability results derived in the previous sections we can draw the phase-diagrams in the plane $`(K,Q)`$ for fixed $`a`$ and $`n`$. In particular, for $`n4`$ the $`m`$-coexistence state is stable within the interval $$Q_m<Q<Q_{m+1}m<n$$ (23) with $`Q_m`$ given in Eq. (14), provided that $`K>a1`$. Interestingly, for fixed $`Q`$, Eq. (23) shows that the increment $`\delta K`$ in the catalytic coupling needed to incorporate a new template into the chain is $$\delta K=\frac{\left(a1\right)^2}{aQ1},$$ (24) regardless the number of elements in the chain. The case $`K<a1`$, for which no chains with $`m>1`$ are allowed, does not require any special consideration. In fact, we find that the only stable states are the zero-coexistence ($`Q<1/a`$) and the $`1`$-coexistence ($`1/aQ1`$) states. However, since $`Q_21`$ only for $`Ka1`$, this result is consistent with Eq. (23). The $`n`$-coexistence state (i.e., the hypercycle solution) is stable for $`Q>Q_n`$ if $`KK_h`$ and for $`Q>Q_h`$, otherwise, where $`K_h`$ and $`Q_h`$ are given by Eqns. (18) and (17), respectively. We define the error threshold of the hypercycle as the value of the replication accuracy $`Q`$ that delimits the region of stability of the $`n`$-coexistence state. The phase diagram for $`a=10`$ and $`n=4`$ shown in Fig. 2 illustrates the major role played by $`K_h`$ in the hypercyclic organization: only for $`K>K_h`$ the hypercycle becomes more stable than a chain of the same length. Another important quantity is the value of $`K`$, denoted by $`K_c`$, at which $`Q_h`$ equals $`1/a`$, the minimal replication accuracy of the quasispecies. It is given by $$K_c=a\left(a1\right)\left[2n1+\sqrt{n\left(n1\right)}\right].$$ (25) Beyond this value the error threshold of the hypercycle $`Q_h`$ is smaller than that of the quasispecies. Moreover, as mentioned before, for large $`K`$, it vanishes like $`1/\sqrt{K}`$. A rather frustrating aspect of $`K_h`$ and $`K_c`$ is that both are of order $`a^2`$, indicating then that the productivity of catalytically assisted self-replication is much larger than that of non-catalyzed self-replication. While this is obviously true for biochemical catalysis, it is difficult to argue for the existence of such efficient catalysts in prebiotic conditions. On the other hand, we can take a different, more optimistic viewpoint and argue that modern biochemical catalysts (enzymes) are so efficient because their precursors had to satisfy the stringent conditions imposed by surpassing $`K_h`$. In Fig. 3 we present the phase-diagram for $`n=5`$. The main difference from the previous figure is that the $`5`$-coexistence state is stable only within the thin region between $`Q_5`$ and the dashed curve, obtained through the numerical evaluation of the Jacobian eigenvalues. As these curves intersect at some $`KK_h`$, the $`5`$-membered hypercycle is not very interesting, since it has the same characteristics of a chain of length $`m=5`$. To confirm this result we have carried out the numerical integration of the kinetic equations using the ninetieth-order Runge-Kutta method. The results are shown in Fig. 4, which illustrates the time evolution of the concentrations $`x_i\left(i=1,\mathrm{},5\right)`$ inside and outside the region of stability. Although the behavior pattern in the region of instability seems periodic, we have not explored completely the space of parameters to discard the existence of chaotic behavior. Hence we use the term complex dynamics to label this region in Fig. 3. We note that the phase-diagram shown in this figure describes also the regions of stability of hypercycles and chains of size $`n5`$, since $`m`$-coexistence states with $`m>5`$ are always unstable. An interesting limiting case which deserves special attention is the symmetric hypercycle ($`a=1`$). According to the argument put forward in the beginning of Sec. III, the only fixed points in this case are the zero-coexistence state and the hypercycle, i.e., chains are not allowed. Moreover, Eq. (15) yields $`x_1=x_2=\mathrm{}=x_n`$ where $`x_1`$ is given by Eq. (16) with $`a`$ replaced by $`1`$. The analysis of the roots of that quadratic equation and the numerical evaluation of the Jacobian eigenvalues yield that the symmetric hypercycle is stable for $$K>\frac{4n}{Q^2}\left(1Q\right),$$ (26) provided that $`n4`$. The region of stability observed in Fig. 3 for the $`5`$-coexistence state does not appear in the symmetric case $`a=1`$, so it must be a consequence of the asymmetry in the productivity values of the non-catalyzed self-replication reaction. We note that, differently from the asymmetric case ($`a>1`$), the zero-coexistence state is always stable. For the sake of completeness, we present some results on the elementary hypercycle ($`A_i=0,i=1,\mathrm{},n`$) coupled to an error tail ($`A_e=1`$) via the imperfect catalytically assisted self-replication. Inserting these parameters into Eq. (20) and setting $`F_i=0i`$ yields $`x_1=\mathrm{}=x_n`$ with $`x_1`$ given by the larger root of the quadratic equation $$Knx_1^2\left(n+KQ\right)x_1+1=0,$$ (27) since we have verified that the smaller root is always unstable. As in the symmetric case discussed above, for $`n4`$ the stability condition coincides with the condition for real $`x_1`$, namely, $$Q2\sqrt{\frac{n}{K}}\frac{n}{K}.$$ (28) Thus the term in the right hand side of this inequality yields the error threshold of the elementary hypercycle. Before concluding this section, we must note that the chains of size $`m`$ considered hitherto are bonded in a hypercycle of size $`n>m`$. We could study free chains of size $`n`$ as well by simply setting $`x_0=0`$ in the kinetic equations (1) and (2). Not surprisingly, the results are essentially the same as for bonded chains, with $`Q_m`$ playing a similar fundamental role in delimiting the regions of stability of the shorter chains ($`m<n`$). Although a full discussion of the stability of the complete chain ($`m=n`$) is beyond the scope of this paper, we point out that free chains with $`n>4`$ are always unstable. Moreover, as illustrated in Fig. 5, the oscillatory behavior pattern of the template concentrations, which ensues a dynamic coexistence among all templates in the chain, is similar to that observed in the hypercycle (compare with Fig. 4). In this sense, the free chains seem as good as the hypercycle to attain that kind of coexistence. However, we must emphasize that, for sufficiently large $`K`$ (i.e. $`K>K_h`$), the fixed points describing the coexistence of $`n4`$ templates in a hypercycle are much more robust against replication errors than their counterparts in the free chains. ## VI Conclusion Our study of the steady states of an asymmetric hypercycle composed of $`n`$ error-prone self-replicating templates indicates that, for $`n4`$, the error threshold of the hypercycle ($`Q_h`$) becomes smaller than that of the quasispecies ($`Q_1=1/a`$) for catalytic couplings ($`K`$) of order of $`a^2`$, where $`a`$ is the productivity value of the master template. In particular, $`Q_h`$ vanishes like $`\sqrt{n/K}`$ for large $`K`$. Perhaps, even more important is our finding that the asymmetry in the non-catalyzed self-replication reaction ($`a>1`$) entails the existence of chains of size $`n5`$. We note that these chains are unstable in the symmetric hypercycle as well as in the fully connected network . Adding to the scenario for the emergence of the hypercycle described in Sec. II, which starts with an isolated quasispecies, our results indicate that the chains may well be the next step in this complex evolutionary process. In fact, according to Eq. (24) the strengths of the catalytic couplings needed to form a chain are of order $`a`$, while the hypercycle only acquires its desirable stability characteristics for strengths of order $`a^2`$ (see Eq. (18)). Although we realize that an evolutionary step leading from chains to hypercycles is still a major one, it is certainly much more plausible than a direct transition from quasispecies to hypercycle. In any event, we think that the emergence of the hypercycle can be explained as a series of plausible smooth transitions, without need to postulating the hypercycle as an unique event in prebiotic evolution. In this vein, this work represents a modest first step to tackle this fundamental problem within a firm basis. To conclude we must mention an alternative resolution for the information crisis in prebiotic evolution which has received some attention recently , namely, the stochastic corrector model . This model builds on ideas of the classical group selection theory for the evolution of altruism , since it considers replicative templates competing inside replicative compartments, whose selective values depend on their template composition. However, the chemical kinetics equations governing the dynamics of the templates inside the compartments display a non-physical (non-integer exponents) dependence of growth on template concentrations. It seems to us that this basic assumption of the stochastic corrector model must be relaxed, or at least justified, before it can be considered an important alternative to the more traditional approach based on the hypercycle and its variants. P.R.A.C. thanks Prof. P. Schuster and Prof. P. F. Stadler for their kind hospitality at the Institut für Theorestische Chemie, where part of his work was done, and Prof. P. E. Phillipson for illuminating discussions. The work of J. F. F. was supported in part by Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq). The work of P.F.S. was supported in part by the Austrian Fonds zur Förderung der Wissenschaftlichen Forschung, Proj. No. 13093-GEN. P.R.A.C. is supported by FAPESP.
no-problem/9907/hep-ex9907034.html	ar5iv	text	# Estimates of Vertex Tagging Efficiencies at a Muon Collider Higgs Factory ## Introduction Measurement of the branching ratios (BR) of the Higgs boson would be an important goal of a future s-channel muon collider Higgs factory. This paper derives quick, heuristic estimates of the expected tagging efficiencies and purities for the 3 main decay modes of a light, Standard Model Higgs: $`\mathrm{H}\mathrm{b}\overline{\mathrm{b}}`$, $`\mathrm{H}\tau \overline{\tau }`$ and $`\mathrm{H}\mathrm{c}\overline{\mathrm{c}}`$. These estimates should be useful as input to theoretical assessments of the physics capabilities of Higgs factory muon colliders. The vertex tagging methods are similar for $`\mathrm{b}\overline{\mathrm{b}}`$ events and $`\mathrm{c}\overline{\mathrm{c}}`$ events and these two modes are treated together in the next section. The $`\tau \overline{\tau }`$ mode is then discussed in a separate section, before ending with a short conclusion section. ## I Charm and Beauty Tagging ### I.1 Tagging Signatures Higgs decays to c or b quark-antiquark pairs produce back-to-back 2-jet events with a displaced vertex in each jet from the decay of the c or b hadron. The goal of a vertex tagging algorithm is to find the displaced vertices and also to distinguish the $`\mathrm{b}\overline{\mathrm{b}}`$ events from $`\mathrm{c}\overline{\mathrm{c}}`$ events. The figures of merit for the tagging algorithm are the overall tagging efficiency for each of the 2 event types and the rejection factor against the other, wrong event type. In principle, one should also consider the rejection factor for hadronic events that are neither $`\mathrm{b}\overline{\mathrm{b}}`$ nor $`\mathrm{c}\overline{\mathrm{c}}`$ but, in practice, this is a much simpler task than distinguishing the 2 types with displaced vertices. For b jets, the charged tracks in the displaced vertex will have an invariant mass up to the mass of b hadrons – about 5 GeV. There will usually be several tracks. B hadrons almost always decay into a charm hadron plus additional hadrons, so some of the charged tracks will likely form a tertiary vertex downstream from the B decay vertex. If a “topological vertexing” algorithm such as ZVTOP zvtop is used then this tertiary vertex may be reconstructed. The characteristic displacement length of the vertex is $`\beta \gamma .\mathrm{c}\tau `$, where $`\mathrm{c}\tau `$ is 450 microns. $`\beta \gamma `$ would typically be about 7 for a 100 GeV Higgs decay at rest – corresponding to about 70% of the quark energy or 35% of the Higgs mass, as is seen in Z decays to b quarks – so the characteristic displacement length is about 3 mm. The displaced vertices from charm typically have a lower multiplicity and, more significantly, a lower invariant mass than B decays since the mass of all weakly decaying charm hadrons is less than 2 GeV. In analogy to Z decays, the charm hadrons should typically carry about 50% of the c quark energy. Combining this with $`\mathrm{c}\tau =317`$ (124) microns for charged (neutral) D mesons gives characteristic decay lengths of 4 mm (1.7 mm). The lower multiplicity and lower invariant mass of charm decays makes them more difficult to tag than B decays, even for the longer lived charged D mesons. ### I.2 Jet Tagging Efficiencies An estimate for jet tagging efficiencies and purities was made based on studies done for the proposed DESY 500 GeV center-of-mass electron-positron linear and presented in the Conceptual Design Report (CDR) desycdr . This study was performed for the following scenario: 1. a 3 layer barrel vertex detector with its innermost layer a cylinder at either 1.0 cm or 2.2 cm from the interaction point (ip) 2. pixel tracking elements. Studies were performed for both charge coupled devices (CCD’s) and active pixel sensors (APS’s) 3. a 50–50 mix of b and c jets 4. vertex reconstruction using the SLD topological vertexing algorithm ZVTOP zvtop 5. the central region of the detector was considered. A muon collider Higgs factory will have worse backgrounds than a linear electron-positron collider and two adjustments were made to allow for this. Firstly, APS’s were assumed to be the tracking technology because CCD’s are presumably too susceptible to radiation damage for use in the muon collider tracking environment and, secondly, the radius of the innermost tracking layer was assumed to be 5 cm, consistent with background studies for muon colliders iuliu . This requires an extrapolation of the DESY studies. With these adjustments, it is assumed that the Higgs factory jet tagging performance can be estimated using the DESY studies. This implicitly assumes that 2 further differences are not important in the study. Firstly, it is assumed that the effect of the higher uncorrelated “fake hit” density in a muon collider can be minimized in a well designed vertex detector using the redundancy afforded by the several vertexing layers. Secondly, the Higgs factory muon collider will have a beam spot size in the hundreds of microns which, in contrast to the smaller spot at a linear collider, will give little useful vertex constraint to assist the vertexing algorithm. The lack of a vertex constraint is important today’s collider detectors that use 2-D vertexing with silicon microstrips and only 2 or 3 layers of vertexing e.g. the LEP and Tevatron detectors. For these geometries the additional constraint would be very helpful. However, it is probably reasonable to assume that the loss would be much less in a multi-layer 3-D pixel vertex detector such as one would expect in a future muon collider. The purity vs. efficiency curves from the DESY study are shown in figure 2.2.2 of the DESY CDR desycdr . All of the curves have a fairly flat purity out to a certain efficiency then dive quite steeply, so little is lost in either purity or efficiency by choosing the efficiency and purity values at this lip of the curve. These values for 1.0 and 2.2 cm inner radii and b– and c–tagging are given in table 1. The table also gives the assumed extrapolation to the 5.0 cm radius at a muon collider. The extrapolated values were estimated by noting that the extrapolation ratio 2.2:5.0 cm is similar to the ratio between the 2 DESY values, 1.0:2.2 cm. The extrapolation to the muon collider scenario is rather modest since the observed changes between 1.0 cm and 2.2 cm are relatively small. Thus, it is considered that this specific choice of extrapolation introduces little uncertainty and should be entirely adequate for estimates of the physics potential of the Higgs factory muon collider – certainly until a detailed detector design and simulation is arrived at. The purities in table 1 refer to an equal mixture of c and b jets. For dealing with an arbitrary mixture of b and c jets it is useful to convert these purities into efficiencies for tagging the wrong type of jet. For an equal sample of b’s and c’s and subscript $`i`$ ($`j`$) denoting the intended (wrong) tag it is clear that: $$p_i=\frac{e_i}{e_i+e_j},$$ (1) so solving for $`e_j`$ gives: $$e_j=e_i\times (\frac{1}{p_i}1)$$ (2) This gives the jet tagging efficiencies of table 2 for the central region of the detector. ### I.3 Event Tagging Efficiencies The jet tagging efficiency must be converted to an event tagging efficiency. Each Higgs decay to $`\mathrm{b}\overline{\mathrm{b}}`$ or $`\mathrm{c}\overline{\mathrm{c}}`$ will produce an event with two approximately back-to-back jets with the same quark flavor. In this subsection, the event tagging efficiency and rejection factor against wrong flavor events is calculated from the assumed jet tagging efficiencies using a simple algorithm for combining the jet tagging information from the two jets. Several simplifying approximations are used. Since the jets are fairly back-to-back it will be assumed that either both jets are central or neither jet is. Therefore, the calculation is done assuming central events and then the efficiency is multiplied by an overall geometrical acceptance factor of $`\mathrm{cos}(\pi /4)=1/\sqrt{2}`$. Since the Higgs decay is isotropic this is roughly equivalent to assuming that jets more than 45 degrees from the beam direction can be considered central enough for reliable vertex tagging. The simple event tagging strategy is to require either one or both of the 2 jets to be tagged correctly and also to require that neither was tagged as the incorrect flavor. From table 3, it is seen that the probability for the first condition is $`1(1e_i)^2`$ and for the second condition is $`(1e_j)^2`$. (As before, $`i`$ denotes the correct jet tag and $`j`$ the incorrect tag.) If the simplifying assumption is made that the two probabilities are independent and the $`1/\sqrt{2}`$ geometrical acceptance is included then the efficiency, $`E_{ii}`$, for correctly tagging the event is: $$E_{ii}=\frac{1}{2}\times [1(1e_i)^2]\times (1e_j)^2.$$ (3) Clearly, the probability, $`E_{jj}`$ for tagging the wrong event type is obtained by simply swapping the $`i`$’s and $`j`$’s: $$E_{jj}=\frac{1}{2}\times [1(1e_j)^2]\times (1e_i)^2,$$ (4) and the rejection factor $`R`$ against the wrong flavor event is defined naturally to be $$R=\frac{E_{ii}}{E_{jj}}.$$ (5) The $`\mathrm{b}\overline{\mathrm{b}}`$ and $`\mathrm{c}\overline{\mathrm{c}}`$ tagging efficiencies and rejection factor against the wrong flavor are given in table 4. ## II Tau event tagging The tagging of $`h\tau \overline{\tau }`$ events appears to be much easier than the tagging of quark jets. An s-channel Higgs factory will produce Higgs at rest and not in association with other particles, so the geometry should be identical to the $`Z\tau \overline{\tau }`$ events seen in the LEP and SLD detectors operating at the Z pole energy. These distinctive events consist of almost back-to-back high energy tracks emanating from the ip. Each side is usually a single track – i.e. a “1-prong”, with an 85.5 % probability per side – with almost all of the remainder being tightly collimated “3-prong” jets. The tau lifetime is long enough ($`c\tau =88`$ microns) that the slight offset of the prongs from the ip should be observable in a precise vertex detector, but vertexing information should not generally be required to identify this event sample. Therefore, the purity of the sample should be close to 100 % and the efficiency should be dominated by the geometrical acceptance of the central tracker. For all practical purposes, physics studies could reasonably assume a purity of 100% and a conservative efficiency of $`\mathrm{cos}(\pi /4)=0.71`$. ## III Conclusions Heuristic estimates have been made of tagging efficiencies and purities in an s-channel muon collider Higgs factory for the 3 main decay modes of a light Higgs boson. The estimates for the $`\mathrm{b}\overline{\mathrm{b}}`$ and $`\mathrm{c}\overline{\mathrm{c}}`$ modes are given in table 4, while the $`\tau \overline{\tau }`$ mode is assumed to have approximately a 71 percent efficiency with essentially 100 percent purity.
no-problem/9907/hep-th9907038.html	ar5iv	text	# Untitled Document hep-th/9907038 LBNL-43634, SLAC-PUB-8187, SU-ITP-99/34 Orientifolds, RG Flows, and Closed String Tachyons Shamit Kachru,<sup>1</sup> Jason Kumar,<sup>2</sup> and Eva Silverstein<sup>3</sup> <sup>1</sup>Department of Physics University of California at Berkeley Berkeley, CA 94720 and Ernest Orlando Lawrence Berkeley National Laboratory Mail Stop 50A-5101, Berkeley, CA 94720 <sup>2</sup> Department of Physics <sup>3</sup> SLAC Stanford University Stanford, CA 94309 We discuss the fate of certain tachyonic closed string theories from two perspectives. In both cases our approach involves studying directly configurations with finite negative tree-level cosmological constant. Closed string analogues of orientifolds, which carry negative tension, are argued to represent the minima of the tachyon potential in some cases. In other cases, we make use of the fact, noted in the early string theory literature, that strings can propagate on spaces of subcritical dimension at the expense of introducing a tree-level cosmological constant. The form of the tachyon vertex operator in these cases makes it clear that a subcritical-dimension theory results from tachyon condensation. Using results of Kutasov, we argue that in some Scherk-Schwarz models, for finely-tuned tachyon condensates, a minimal model CFT times a subcritical dimension theory results. In some instances, these two sets of ideas may be related by duality. July 1999 1. Introduction Many closed string theories contain tachyons. In addition to the bosonic string, numerous orbifold models have tachyons appearing in a twisted sector. Although there are also many string backgrounds which do not contain tachyons in the perturbative spectrum, it is potentially of interest to understand the fate of those which do. Some work that was done earlier involved attempting to study the tachyon potential directly through (off-shell) calculations of the tachyon effective potential . In this paper we take a somewhat different approach to this question (using a combination of orientifold physics and the observations in ) and discuss some examples which realize it. A basic issue that arises is the following. A perturbative closed string compactification has vanishing tree-level vacuum energy (formally all one-point functions, including those of the dilaton and graviton, vanish in the closed-string sphere diagram). Therefore the tachyon potential, at fixed dilaton vacuum expectation value, looks like figure 1 in the regime accessible to the perturbative closed string description; in particular the tachyonic maximum of the potential is at $`V(0)=0`$. Then tachyon condensation leads to configurations with negative energy. Our main question is whether (1) there is a minimum configuration with finite negative energy into which the system can roll (and then what the eventual solution of the dilaton equation of motion is) or whether (2) the tachyon rolls to negative infinity. Figure 1: Tachyon on top of potential hill We will here consider two different ways of obtaining configurations with finite negative tree-level cosmological constant. The first involves the physics of orientifolds (and analogous configurations in closed string theories); the second involves strings propagating on spaces of subcritical dimension. As we will mention below, these two methods may in fact be related to each other by dualities. In the context of open string theories orientifold backgrounds \[3,,4,,5\] have the feature one would need for possibility (1), namely a negative contribution to the tree-level vacuum energy, which is finite for fixed nonzero string coupling. In §2 we will set up an open string theory which involves orientifolds, anti-orientifolds, branes, and anti-branes in which the potential looks like figure 1. This model is similar to one discussed earlier in the context of string-string duality by Bergman and Gaberdiel (and formulated earlier by Bianchi and Sagnotti ). Then using locally the mechanism of \[8,,9,,10\], we will see that the tachyons condense, leaving the system in a finite-negative-energy minimum involving only orientifolds and anti-orientifolds. We find that in some closed string theories, non-perturbative configurations, analogous to orientifolds in the open string case, exist and can play a similar role. These configurations include the S-duals of orientifold planes \[11,,12\] and non-level-matched but anomaly free orbifold backgrounds such as those of . We will discuss these configurations, which look locally like figure 2, in §3. In this approach, problem is then to relate configurations of the type in figure 1 with those of the type in figure 2. Figure 2: Tachyon at minimum of its potential The simplest such examples are given by the S-dual of the open string theory of §2 (and various heterotic cousins of this S-dual model). We discuss these theories in §4. They involve orbifold fixed points whose degrees of freedom microscopically represent the collective coordinates of NS five-branes sitting at the singularities , as well as other orbifold fixed points, breaking a complementary half of the supersymmetries from the first set, where NS anti-five-branes sit. Globally this system is unstable, and can undergo NS brane-anti-brane annihilation. Condensation of the unstable mode therefore leaves a vacuum with S-dual orientifolds and anti-orientifolds; the negative (S-dual) orientifold tensions account for the negative vacuum energy after tachyon condensation. Our second approach, based on , is the following. An a priori different way of getting backgrounds with finite negative tree-level cosmological constant was explained in , in which corresponding solutions to the dilaton and graviton equations of motion (linear dilaton solutions) were obtained. (The latter step in particular eliminates the dilaton and graviton tadpoles which we ignored in the preceeding discussion.) These solutions involve strings propagating on spaces of sub-critical dimension, in which the worldsheet beta functions are cancelled by the contributions of dilaton gradients. In §5 we will argue that a large class of tachyonic theories can be seen to flow to such backgrounds upon tachyon condensation. This analysis uses simple features of the closed string tachyon vertex operators in this class of examples. In particular generically these vertex operators constitute (or generate) mass terms lifting some of the degrees of freedom on the worldsheet, and one is left after renormalization-group flow with a theory of sub-critical dimension of the type considered in . For instance, we will see in §5 that by tuning the tachyon condensate in some Scherk-Schwarz models one can obtain minimal model CFTs times a sigma model on a subcritical dimension target space. A connection between non-critical string theory and tachyon condensation (in the context of the Hagedorn transition) was earlier conjectured in , where the form of the tachyon potential in heterotic Scherk-Schwarz compactifications was explored. We further find that this point of view is potentially related to the above point of view involving orientifolds if the type of duality conjectured in is realized in string theory. Before proceeding we should remark on the dilaton direction in the potential. Any Poincare-invariant configuration with negative tree-level vacuum energy is not a solution to the equations of motion. In string frame the vacuum energy is $$\mathrm{\Lambda }=\frac{1}{g_s^2}\mathrm{\Lambda }_0+\mathrm{}$$ So in order to solve the equations of motion the dilaton $`\varphi `$ must vary over spacetime. Similar remarks apply to the spacetime metric $`G_{\mu \nu }`$. As we will discuss in §5, the full solution may be obtained through the conformal field theory techniques of . However at some points in our analysis, particularly involving our first approach to the problem (in §2-§4), we will attempt to separate the dependence of the spacetime potential on the originally tachyonic mode from its dependence on the dilaton and metric by holding $`\varphi `$ and $`G_{\mu \nu }=\eta _{\mu \nu }`$ fixed artificially. The true dynamics of the system will of course involve all of these fields, as discussed more thoroughly in §5. Many interesting papers which discuss tachyons on branes in open string theories (“field theoretic” tachyons) have recently appeared; see e.g. . There have also been interesting recent papers on other types of non-supersymmetric string backgrounds . 2. An illustrative open-string theory Consider the type IIB string theory compactified on $`T^d/I_d\mathrm{\Omega }`$, where $`I_d`$ is a reflection on $`d`$ coordinates and $`\mathrm{\Omega }`$ is a reversal of orientation on the IIB worldsheet. The fixed planes of the $`I_d\mathrm{\Omega }`$ action, the orientifold $`(9d)`$-planes, have effectively negative tension ; that is, their graviton and dilaton one-point functions have the opposite sign to those of $`D(9d)`$-branes. There being no degrees of freedom tied to the orientifold planes, this leads to a negative contribution to the tree-level vacuum energy without leading to any unphysical negative kinetic energy. These orientifold planes are also charged oppositely to the $`D(9d)`$-branes under the $`(10d)`$-form RR potential. Therefore in the compact $`T^d`$, in order for RR flux lines to end consistently, we must introduce 16 $`D(9d)`$-branes. In addition to cancelling the RR charge, the positive tensions of the branes cancel the negative tree-level vacuum energy of the orientifolds. This example is very well known, as a T-dual of the type I string theory on $`T^d`$. Now consider instead a $`Z_2\times Z_2`$ orientifold of type IIB (studied in ). on $`T^d`$ generated by $$g_1=I_d\mathrm{\Omega }$$ $$g_2=I_d\mathrm{\Omega }(1)^F\delta $$ where $`\delta `$ is a translation by halfway around one of the circles, with say coordinate $`x_1`$, of the $`T^d`$. The element $`g_2`$ preserves the opposite half of the supersymmetry from the half preserved by $`g_1`$; it introduces anti-orientifolds at positions halfway around the $`x_1`$ circle from the positions of the orientifold planes introduced by $`g_1`$. The element $`g_2`$ alone would require the introduction of 16 anti-D$`(9d)`$ branes. Given both the orientifolds and the anti-orientifolds, we now have many options. We could: A) Add the 16 D$`(9d)`$-branes and 16 anti-D$`(9d)`$-branes corresponding to each orientifold action alone, and then project the resulting open string spectrum onto invariant states. This leads to a theory with RR charge conserved, zero tree-level vacuum energy, and gauge group $`O(16)\times O(16)`$ or a rank-16 subgroup depending on the positions of the branes. There is also a tachyon multiplet in the $`(\mathrm{𝟏𝟔},\mathrm{𝟏𝟔})`$ of $`O(16)\times O(16)`$, for sufficiently small radius or sufficiently small separation of branes and anti-branes, and a closed string tachyon for sufficiently small radius. Alternatively we could: B) Add $`k<16`$ D$`(9d)`$-branes and $`k`$ anti-D$`(9d)`$-branes. Again the RR charge is conserved, but the tree-level vacuum energy is negative (as there are too few branes to cancel the negative tension from the orientifolds). The gauge group is now $`O(k)\times O(k)`$ and there is a scalar in the bifundamental which becomes tachyonic in appropriate regimes of the classical moduli space where the branes and anti-branes are close enough to each other. It is clear in this example that we can annihilate $`16k`$ of the branes against $`16k`$ of the antibranes, which proceeds via condensation of $`(16k)\times (16k)`$ components of the bifundamental tachyon along with condensation of the appropriate magnetic tachyon . In particular when all the open string tachyons of A) condense, we are left with only orientifolds and antiorientifolds, no gauge group, and finite negative vacuum energy (for fixed nonzero dilaton). So in this open-string context, the zero-energy tachyonic theory A) rolls to a finite negative energy configuration B) (plus a gas of excitations, since energy is conserved in this process). 3. Negative-tension defects in closed string theory Now, we turn to a discussion of closed string theories with tachyons. It is natural to ask if there are negative-tension configurations in closed string theory, analogous to the orientifolds of §2, to which such theories could roll. 3.1. S-duality and Orientifolds One obvious place to look for such a configuration is from the S-dual of an orientifold plane in type IIB theory \[11,,19\]. Consider for example an orientifold 5-plane. In the IIB theory on $`\mathrm{IR}^4/I_4\mathrm{\Omega }`$, cancelling the RR charge locally gives a D5-brane at an orientifold 5-plane. The S-dual of this orientifold is type IIB on the orbifold $`\mathrm{IR}^4/I_4(1)^{F_L}`$. The twisted sector of this orbifold constitutes precisely the worldvolume theory on an NS 5-brane at a $`Z_2`$ singularity. Giving vacuum expectation values to the twisted sector scalars corresponds to moving the NS 5-brane off of the fixed point. The $`Z_2`$ fixed point which is left (which has no degrees of freedom associated with it perturbatively – all of the “twisted sector moduli” are accounted for by the translation modes of the NS 5-brane) is the S-dual of the orientifold. The full closed-string orbifold theory has zero vacuum energy, so since the NS 5-branes have positive tension (measured by an integral on a sphere surrounding the brane), the S-dual of the orientifold contributes negatively to the vacuum energy. So one expects, for instance, that a dilaton gradient should emerge as one turns on the marginal perturbations in the twisted sector of the orbifold (moving the NS brane away from the S-dual orientifold), in analogy with . It would be very interesting to understand how this works out in detail. 3.2. More general examples from non-perturbative orbifolds The S-dual orientifold is an example of a larger class of negative-tension backgrounds. In perturbative string theory, modular invariance imposes conditions which sometimes go beyond the physical condition of anomaly cancellation. Modular invariance in the context of orbifold models requires the inclusion of twisted sectors. In the S-dual orientifold example, after removing the NS 5-branes we are left with an orbifold fixed point with no twisted states. As we discussed above this object exists in the theory and has a negative contribution to $`\mathrm{\Lambda }_{tree}`$. More generally we can consider orbifold models in which the level-matching conditions, which ensure that physical perturbative string states exist in the twisted sectors, are violated . For example, let us consider the heterotic theory compactified on a K3 surface realized as an orbifold $`T^4/\text{ZZ}_2`$ (our comments generalize immediately to the $`T^4/\text{ZZ}_k`$ orbifold realizations of K3 with $`k2`$ as well). Green-Schwarz anomaly cancellation requires only that $$dH=trRRtrFFn_5\delta _5$$ where $`\delta _5`$ denotes delta functions localized at the positions of the $`n_5`$ fivebranes. On the compact K3, $`𝑑H=0`$, so $`trFF+n_5=trRR`$. This means that the number of instantons in the gauge bundle, $`n_{inst}`$, and the number of 5-branes $`n_5`$ are constrained by $$n_{inst}+n_5=24.$$ The perturbative string orbifold level-matching conditions require the orbifold group to act nontrivially, according to one of a few discrete possibilities, on the gauge degrees of freedom of the heterotic string. This enforces the introduction of 24 instantons. Non-perturbative orbifold backgrounds with fewer instantons and a compensating number of 5-branes involve non-modular-invariant choices of the action on the gauge degrees of freedom. These backgrounds fit into an intricate web of six-dimensional string dualities . Again in these cases, the fact that the 5-branes have positive tension, and therefore contribute positively to $`\mathrm{\Lambda }_{tree}`$ in these backgrounds, means that the non-level-matched orbifold fixed points that are left over contribute negatively to $`\mathrm{\Lambda }_{tree}`$. In fact, even without invoking non-perturbative physics we can find configurations with effectively negative tension. Suppose for example that we start with an orbifold action with the standard embedding in the gauge degrees of freedom of the heterotic string. In this case, $`dH`$ cancels locally. Now consider moving some large instantons off the orbifold fixed points. These instantons constitute gauge fivebranes of the heterotic string, and have positive tension. Since the overall vacuum energy of the string vacuum is zero, the contribution of the remaining orbifold fixed points is negative. Clearly there is a whole zoo of possibilities of this sort. Furthermore as long as the instantons are large, the whole configuration is accessible perturbatively.<sup>1</sup> We thank P. Aspinwall and R. Plesser for discussions on these points. 3.3. S-dual O9-planes? Hull and collaborators have proposed formulating the $`SO(32)`$ heterotic string as an orbifold of type IIB by $`(1)^{F_L}`$. In perturbative string theory, where we must add the twisted sectors dictated by modular invariance, orbifolding by $`(1)^{F_L}`$ produces IIA from IIB and vice versa. On the other hand, there are other solutions to the anomaly cancellation conditions. In particular, one could add an $`SO(32)`$ or $`E_8\times E_8`$ gauge group, obtaining the heterotic string theory. In the $`SO(32)`$ case, it is argued in that this background can be deconstructed into an S-dual O9-plane plus (in our way of counting) 16 NS 9-branes. In particular, by S-dualizing the Born-Infeld like action for the ordinary O9-plane plus 16 D9-branes, one obtains this formulation of the heterotic string. In this point of view, the S-dual O-9-plane has negative tension, $`\mathrm{\Lambda }_{SO9}<0`$, which is cancelled by the tensions of the 16 NS 9-branes to give a vacuum with $`\mathrm{\Lambda }_{tree}=0`$. These branes, if they exist, may provide further possibilities for the endpoint of tachyon condensation in various backgrounds. 4. Examples of tachyon condensation In the last section we saw that there exists a plethora of possible configurations of negative vacuum energy (a la figure 2), realized microscopically as S-duals of orientifolds and their generalizations. In this section we consider the problem, on which we will only make a start here, of understanding which tachyonic models (fig 1) roll to to which (if any) of these negative-energy configurations. In §5 we will consider another type of negative-energy configuration which can plausibly be the endpoint of tachyon condensation for a different class of theories. We have mostly been confined here to an analysis of orbifold backgrounds. One can hope that once more generic non-supersymmetric compactification geometries are understood, there will be more general examples of such backgrounds. Similarly many of the tachyonic theories (fig 1) that we are interested in explaining are perturbative string orbifolds. In general, as twisted sector tachyon condensation will break the quantum symmetry of the orbifold, we would expect orbifold models to roll to more generic geometrical backgrounds. So we will here consider a non-supersymmetric theory whose tachyons arise at a more generic point on the CFT moduli point (away from an orbifold point). In particular, let us consider the non-supersymmetric model which is S-dual to the $`d=4`$ ($`d`$ is the number of compact dimensions) case of the orientifold model of §2. We begin with type IIB theory on $`T^4`$ and mod out by a $`\text{ZZ}_2\times \text{ZZ}_2`$ symmetry generated by $$g_1=I_4(1)^{F_L},g_2=I_4(1)^{F_R}\delta _1.$$ Here $`\delta _1`$ is say a shift halfway around one of the circles of the torus. This theory has 16 orbifold fixed points (8 fixed points each for the $`g_1`$ and $`g_2`$ actions). Based on the analysis of , we find that there are NS fivebranes sitting at the $`g_1`$ fixed points and NS anti-five-branes sitting at the $`g_2`$ fixed points. The vacuum energy is zero, and for large enough radius there is no tachyon in the $`(1)^F\delta _1`$ twisted sector. At genus zero, this theory is consistently described by nonsingular orbifold conformal field theory. Globally the set of fivebranes and anti-five-branes is topologically indistinguishable from the vacuum. This being true at the level of the classical field configuration, we expect to find an instability in this system at some order in conformal perturbation theory about the orbifold limit. The leading possibility is that moving some combination of fivebranes off the fixed point(s)–an exactly marginal perturbation–induces a tachyonic mass term for the collective coordinate describing the motion of the anti-five-branes off the orbifold points. This corresponds to a force on the anti-five-branes linear in their displacement from the orbifold points. There is such a force in some regimes at the level of the low-energy gravity theory, valid when all distances are much larger than string scale. There is globally therefore a manifest instability in the system, though $`\alpha ^{}`$ corrections may come in at substringy distances and lift the tachyonic mode from the point where the anti-fivebranes are at the orbifold points. In any case it is natural to postulate that the fivebranes and anti-fivebranes annihilate, leaving behind the S-dual orientifolds which are hidden at the orbifold fixed points. Given this, this example gives us one case where the result of condensation of a negative mode, most probably a tachyon, is a configuration of finite negative tension of the type discussed in §3 (at fixed dilaton) . Many other similar models can be obtained by considering the heterotic theory on $`T^4`$ modded out by $`I_4`$ and $`I_4(1)^F`$, with a level-matching action on the gauge bundle that introduces 24 instantons and 24 anti-instantons. These then can move off the fixed points and enlarge; there is a global instability in the moduli space which we would expect to see perturbatively. The global minimum of the potential for the non-dilatonic moduli is again a configuration of finite negative tension of the type discussed in §3. We leave detailed analysis of this class of examples for future work. 5. Tachyons and RG Flow Another context in which backgrounds with finite negative tree-level cosmological constant arise is the following . Consider bosonic strings propagating in a spacetime of dimension $`D<D_{crit}`$.<sup>2</sup> Our discussion applies with minor modifications to other strings as well. The Weyl anomaly cancellation conditions, which include $$0\beta ^\mathrm{\Phi }=\frac{DD_{crit}}{6}\frac{\alpha ^{}}{2}^2\mathrm{\Phi }+\alpha ^{}_\mu \mathrm{\Phi }^\mu \mathrm{\Phi }\frac{\alpha ^{}}{24}H^2+O(\alpha _{}^{}{}_{}{}^{2})$$ can still be solved in such a situation by allowing the dilaton to vary over spacetime. For example one exact CFT solution is the linear dilaton background \[2,,22\]. The Weyl anomaly conditions, regarded as equations of motion for the spacetime theory, follow from a Lagrangian $$Sd^Dx\sqrt{G}e^{2\mathrm{\Phi }}\left(\frac{2(DD_{crit})}{3\alpha ^{}}+R\frac{1}{12}H^2+4_\mu \mathrm{\Phi }^\mu \mathrm{\Phi }+O(\alpha ^{})\right)$$ The first term is a finite negative cosmological constant proportional to $`DD_{crit}`$. It was noted in that such sub-critical-dimension theories could arise naturally as the result of tachyon condensation. Tachyon vertex operators are relevant operators of the internal worldsheet CFT (that is the part of the worldsheet CFT not involving the noncompact Poincare-invariant spacetime). Let us consider a specific case. Take the bosonic string theory compactified on the $`SO(32)`$ lattice. In a fermionic description the $`SO(32)^2`$ current algebra arises from 32 real left-moving fermions $`\lambda ^I,I=1,\mathrm{},32`$ and 32 real right-moving fermions $`\stackrel{~}{\lambda }^{\stackrel{~}{I}},\stackrel{~}{I}=1,\mathrm{},32`$. There are in addition 10 scalar fields $`X^\mu ,\mu =0,\mathrm{},10`$ making up the rest of the 26 units of central charge. This theory has in addition to the singlet tachyon (the universal tachyon of the bosonic string) a $`(\mathrm{𝟑𝟐},\mathrm{𝟑𝟐})`$ tachyon. The vertex operator for the latter is $$V_{(32,32)}=\lambda ^I\stackrel{~}{\lambda }^{\stackrel{~}{J}}e^{ikX}$$ To describe tachyon condensation, at leading order we wish to add the integrated tachyon vertex operator to the worldsheet action. At zero momentum $`k`$ this operator is relevant, and the covariant vertex operator which we would naively add, $$d^2\sigma \sqrt{g}\lambda \stackrel{~}{\lambda }$$ has nontrivial dependence on the conformal factor $`\omega `$ in the worldsheet metric. (Here we use diffeomorphisms to fix $`g=\eta e^{2\omega }`$ where $`\eta `$ is the flat metric on the worldsheet.) Heuristically, since (5.1) is a mass term for $`\lambda `$ and $`\stackrel{~}{\lambda }`$, we expect this relevant deformation to lift these degrees of freedom from the worldsheet theory, giving us a string theory propagating effectively in $`D=26\frac{N}{2}`$ where $`N`$ is the number of worldsheet fermions that pair up and become massive. However since this procedure involves going off shell, as formulated here it breaks the Weyl symmetry of the worldsheet theory classically. When we have a well-defined on-shell string theory, before condensing the tachyon, the formulation of the worldsheet theory involves dividing the worldsheet path integral out by the volume of the Weyl group, thereby eliminating the conformal factor $`\omega `$ from the theory. It is difficult to see how this degree of freedom could be introduced in a physically continuous deformation from that point. Let us therefore consider adding instead<sup>3</sup> We thank S. Shenker for a discussion on this approach. $$d^2\sigma \sqrt{g}\lambda \stackrel{~}{\lambda }e^{ikX}$$ with $`k^2=m^2<0`$. This operator is dimension (1,1) and the expression is Weyl invariant (due to the metric-dependence in the regulated operators ). In particular let us take $`k=(\pm i\kappa ,0,\mathrm{},0)`$ with $`\kappa `$ positive. This gives us a dimension (1,1) operator describing a time-dependent tachyon condensate. Then the terms we can add are of the form $$d^2\sigma \sqrt{g}\lambda \stackrel{~}{\lambda }(a_+e^{\kappa X_0}+a_{}e^{\kappa X_0})$$ for some parameters $`a_\pm `$. The addition (5.1) describes the initial time-dependence of the tachyon as it begins to roll down its potential hill. Let us take $`a_{}=0`$ so that the condensate vanishes at very early times $`X_0\mathrm{}`$. Though as the condensation proceeds we expect higher order corrections to (5.1), we do expect that the tachyon will continue to condense (rather than returning to its unstable maximum) and the coefficient of the $`\lambda \stackrel{~}{\lambda }`$ mass term in (5.1) will be nonzero for $`X_0>\mathrm{}`$. Therefore it seems quite plausible that the effect of tachyon condensation is to lift the degrees of freedom $`\lambda ,\stackrel{~}{\lambda }`$ and correspondingly introduce a nonzero $`D26`$ term in the beta function equations (spacetime equations of motion). In order to preserve conformal invariance, therefore, the spacetime dilaton needs to begin varying so as to maintain $`\beta ^\mathrm{\Phi }=0`$. (In other words we are adding a marginal perturbation (5.1) and therefore must maintain that the total central charge is constant.) One example of a solution to the modified equations of motion would, for instance, be the linear dilaton solution of \[2,,22\], though it is not clear that the dynamics favors this solution. 5.1. Relation to orientifold picture? This lifting of the $`\lambda ,\stackrel{~}{\lambda }`$ degrees of freedom agrees with the picture one would obtain from the relation of this theory to a certain type I orientifold conjectured by Bergman and Gaberdiel . The orientifold in question is simply an orientifold of type IIB by a $`\text{ZZ}_2\times \text{ZZ}_2`$ action generated by $`(1)^F`$ and $`\mathrm{\Omega }`$. This theory (which is extremely similar to the one we discussed in §4, differing by a shift and some duality transformations) has an orientifold 9-plane, an anti-orientifold 9-plane, 16 D9-branes and 16 anti-D9-branes. This gives a gauge group $`SO(32)^2`$ with tachyons in the $`(\mathrm{𝟑𝟐},\mathrm{𝟑𝟐})`$ representation. This theory contains in its spectrum of D1-branes one which has the worldvolume degrees of freedom of the bosonic string compactified on the $`SO(32)`$ lattice. These degrees of freedom consist of 10 bosonic collective coordinates from the 1-1 sector of open strings, and 32 left-moving fermions from the 1-9 sector and 32 right-moving fermions from the $`1\overline{9}`$ sector. In this theory it is relatively clear what happens when the $`(\mathrm{𝟑𝟐},\mathrm{𝟑𝟐})`$ tachyon condenses: the D9-branes annihilate with the anti-D9-branes. This leaves the spacetime theory with the O9 and anti-O9 planes, which contribute a finite negative amount to the tree-level cosmological constant as discussed in very similar circumstances above. Now let us consider what happens on the D1-brane probe in the orientifold theory. Once the D9-branes and anti-D9-branes have annihilated, the contributions from the 1-9 and $`1\overline{9}`$ sectors disappear. One is left with a low-energy worldvolume theory with 10 scalar collective coordinates $`X^\mu ,\mu =1,\mathrm{},10`$. This gives the same result as we found for the bosonic string in the RG flow analysis (in particular the absence of both the fermions $`\lambda `$ and $`\stackrel{~}{\lambda }`$ and the decoupling of the Liouville mode classically on the worldsheet). Therefore if the duality between the two theories conjectured in turns out to hold<sup>4</sup> The meaning of this duality proposal is not entirely clear, since the dilaton potential is not flat in the proposed duals. there is a relation between the two approaches we are considering in this paper. 5.2. Other Models There is a large set of models amenable to the analysis involving RG flow. For example, let us consider a (T-dual version of a) Scherk-Schwarz compactification. This consists of an orbifold of the toroidally compactified type II or heterotic theory by $`(1)^F\delta `$ where $`\delta `$ is a level-matched shift symmetry of the Narain lattice. We choose $`\delta `$ so that the Scherk-Schwarz tachyon(s) are momentum modes which becomes tachyonic at sufficiently large radius. The integrated vertex operators for these tachyons take the form (in the type II case for specificity) $$d^2zd^2\theta \underset{p}{}\lambda _pcos[(p+\delta ^{}+k)𝐗]+\rho _psin[(p+\delta ^{}+k)𝐗]$$ Here we take $`\delta ^{}`$ to be half a momentum lattice vector, and we sum over those $`p`$ in the momentum lattice such that the internal piece of the added operator is relevant–i.e. we only include tachyons. $`k`$ is spacetime momentum in the noncompact directions. We work in $`(1,1)`$ superspace with $`𝐗=X+\theta \psi +\overline{\theta }\stackrel{~}{\psi }+F`$ a scalar superfield. In the $`(0,0)`$ picture, in components, the vertex operator is $$V^{(0,0)}=(\delta _L^{}+k)\psi (\delta _R^{}+k)\stackrel{~}{\psi }e^{i\delta ^{}X}e^{ikX}$$ The renormalization group flow for this type of potential has been studied by Kutasov . He finds generically a mass gap for the degrees of freedom in $`𝐗\delta ^{}`$. With some fine-tuning of the coefficients $`\lambda `$ and $`\rho `$, one can obtain minima in the potential which locally take the form of $`𝐗^n`$ potentials for various $`n>2`$. Thus minimal models arise as the result of tachyon condensation in these special directions. In such cases, although one does not lift all the central charge of the $`𝐗`$ multiplet, it decreases in the flow to the value of the appropriate minimal model, and the orthogonal spacetime part of the theory is still of sub-critical dimension. A similar analysis can be done for heterotic compactifications of the form $`T^d/(1)^F\delta `$. More generally, this set of ideas might be helpful in any situation where the closed string tachyon is charged under some spacetime gauge symmetry. Such theories with charged tachyons are natural candidates for admitting dual descriptions in terms of branes (as in ), where the tachyon condensation corresponds to some kind of brane annihilation process. It would be interesting to see if these approaches to understanding closed string tachyons can teach one about string theory at the Hagedorn transition. A different approach to this problem was recently discussed in . Acknowledgements We would like to thank O. Aharony, P. Aspinwall, M. Berkooz, J. Harvey, I. Klebanov, E. Martinec, R. Plesser, and S. Shenker for discussions. We are also grateful to TASI-99 and the 1999 Amsterdam Summer Workshop on String Theory for hospitality during the course of this work. The research of S.K. is supported by NSF grant PHY-95-14797, by DOE contract DE-AC03-76SF00098, by an A.P. Sloan Foundation Fellowship and by a DOE OJI Award. The research of J.K. is supported by the Department of Defense NDSEG Fellowship Program and by NSF grant PHY-9870115. The research of E.S. is supported by the DOE under contract DE-AC03-76SF00515 and by an A.P. Sloan Foundation Fellowship. References relax See for instance: V.A. Kostelecky and S. Samuel, “The Tachyon Potential In String Theory,” Presented at 1988 Mtg. of Div. of Particle and Fields of the APS, Storrs, CT, Aug 15-18, 1988; T. Banks, “The Tachyon Potential in String Theory,” Nucl. Phys. B361 (1991) 166; A. Belopolsky and B. Zwiebach, “Off Shell Closed String Amplitudes: Towards a Computation of the Tachyon Potential,” Nucl. Phys. B442 (1995) 494, hep-th/9409015. relax S.P. de Alwis, J. Polchinski, and R. Schimmrigk, “Heterotic Strings with Tree Level Cosmological Constant,” Phys. Lett. B218 (1989) 449. relax A. Sagnotti, “Closed Strings And Their Open String Descendants,” Phys. Rept. 184, 167 (1989). relax P. Horava, “Strings on World Sheet Orbifolds,” Nucl. Phys. B327 (1989) 461. relax J. Dai, R.G. Leigh and J. Polchinski, “New Connections Between String Theories,” Mod. Phys. Lett. A4 (1989) 2073. relax O.Bergman and M. Gaberdiel, “A Nonsupersymmetric Open String Theory and S-Duality,” Nucl. Phys. B499 (1997) 183, hep-th/9701137. relax M. Bianchi and A. Sagnotti, “On The Systematics Of Open String Theories,” Phys. Lett. B247, 517 (1990). relax T. Banks and L. Susskind, “Brane - Anti-Brane Forces,” hep-th/9511194. relax A. Sen, “Tachyon Condensation on the Brane Anti-Brane System,” JHEP 9808 (1998) 012, hep-th/9805170. relax P. Yi, “Membranes from Five-Branes and Fundamental Strings from D-p Branes,” hep-th/9901159. relax D. Kutasov, “Orbifolds and Solitons,” Phys. Lett. B383 (1996) 48, hep-th/9512145. relax C. Hull, “The Nonperturbative SO(32) Heterotic String,” hep-th/9812210; E. Bergshoeff, E. Eyras, R. Halbersma, J.P. van der Schaar, C. Hull and Y. Lozano, “Space-Time Filling Branes and Strings with Sixteen Supercharges,” hep-th/9812224. relax G. Aldazabal, A. Font, L.E. Ibanez, A.M. Uranaga and G. Violero, “Non-Perturbative Heterotic D=6, D=4, N=1 Orbifold Vacua,” Nucl. Phys. B519 (1998) 239, hep-th/9706158. relax I. Antoniadis and C. Kounnas, “Superstring phase transition at high temperature,” Phys. Lett. B261, 369 (1991). relax For a recent review with references, see: A. Sen, “Non-BPS States and Branes in String Theory,” hep-th/9904207. relax See for instance: B. Kors, “D-brane Spectra of Nonsupersymmetric, Asymmetric Orbifolds and Nonperturbative Contributions to the Cosmological Constant,” hep-th/9907007; R. Blumenhagen and A. Kumar, “A Note on Orientifolds and Dualities of Type 0B String Theory,” hep-th/9906234; I. Klebanov, “Tachyon Stabilization in the AdS/CFT Correspondence,” hep-th/9906220; C. Angelatonj, I. Antoniadis and K. Forger, “Nonsupersymmetric Type I Strings with Zero Vacuum Energy,” hep-th/9904092; R. Blumenhagen, A. Font and D. Lust, “Tachyon Free Orientifolds of Type 0B Strings in Various Dimensions,” hep-th/9904069; R. Blumenhagen and L. Gorlich, “Orientifolds of Nonsupersymmetric Asymmetric Orbifolds,” hep-th/9812158; I. Antoniadis, G. D’Appolonio, E. Dudas, and A. Sagnotti, “Partial breaking of supersymmetry, open strings and M-theory”, hep-th/9812118; I.R. Klebanov and A.A. Tseytlin, “D-branes and Dual Gauge Theories in Type 0 Strings,” Nucl. Phys. B546 (1999) 155, hep-th/9811035; C. Angelantonj, “Nontachyonic open descendants of the 0B string theory,” Phys. Lett. B444, 309 (1998) hep-th/9810214; G. Shiu and S.H. Tye, “Bose-Fermi Degeneracy and Duality in Nonsupersymmetric Strings,” Nucl. Phys. B542 (1999) 45, hep-th/9808095; S. Kachru and E. Silverstein, “Self-Dual Nonsupersymmetric Type II String Compactifications,” JHEP 9811 (1998) 001, hep-th/9808056; J. Harvey, “String Duality and Nonsupersymmetric Strings,” Phys. Rev. D59 (1999) 026002, hep-th/9807213; M. Berkooz and S. Rey, “Nonsupersymmetric Stable Vacua of M Theory,” JHEP 9901 (1999) 014, hep-th/9807200; S. Kachru, J. Kumar and E. Silverstein, “Vacuum Energy Cancellation in a Nonsupersymmetric String,” Phys. Rev. D59 (1999) 106004, hep-th/9807076 ; I. Antoniadis, E. Dudas,and A. Sagnotti, “Supersymmetry breaking, open strings and M-theory”, Nucl.Phys. B544 (1999) 469-502, hep-th/9807011; S. Kachru and E. Silverstein, “4-D Conformal Field Theories and Strings on Orbifolds,” Phys. Rev. Lett. 80 (1998) 4855, hep-th/9802183; J.D. Blum and K.R. Dienes, “Duality Without Supersymmetry: The Case of the $`SO(16)\times SO(16)`$ String,” Phys. Lett. B414 (1997) 260, hep-th/9707148. relax J. Polchinski, “TASI Lectures on D-Branes,” hep-th/9611050. relax I. Antoniadis, E. Dudas,and A. Sagnotti, “Supersymmetry breaking, open strings and M-theory”, Nucl.Phys. B544 (1999) 469-502, hep-th/9807011. relax A. Sen, “Duality and Orbifolds,” Nucl. Phys. B474 (1996) 361, hep-th/9604070. relax J. Polchinski and E. Witten, “Evidence for Heterotic-Type I String Duality,” Nucl. Phys. B460 (1996) 525, hep-th/9510169. relax C. Callan, J. Harvey and A. Strominger, “World Sheet Approach to Heterotic Instantons and Solitons,” Nucl. Phys. B359 (1991) 611. relax R. Myers, “New Dimensions for Old Strings,” Phys. Lett. 199B (1987) 371; I. Antoniadis, C. Bachas, J. Ellis and D. Nanopoulos, “Cosmological String Theories and Discrete Inflation,” Phys. Lett. 211B (1988) 393. relax J. Polchinski, String Theory, Vol. I, Cambridge University Press, 1998. relax D. Kutasov, “Irreversibility of the Renormalization Group Flow in Two-Dimensional Quantum Gravity,” Mod. Phys. Lett. A7 (1992) 2943, hep-th/9207064; E. Hsu and D. Kutasov, “The Gravitational Sine-Gordon Model,” Nucl. Phys. B396 (1993) 693, hep-th/9212023. relax I. Antoniadis, J. Derendinger and C. Kounnas, “Nonperturbative Temperature Instabilities in N=4 Strings,” Nucl. Phys. B551 (1999) 41, hep-th/9902032.
no-problem/9907/astro-ph9907411.html	ar5iv	text	# Dark matter substructure within galactic halos ## 1. Introduction The growth of structure in the universe by hierarchical accretion and merging of dark matter halos is an attractive and well motivated cosmological model (White & Rees 1978, Davis et al. 1985). The gravitational clustering process is governed by the dark matter component and the baryons only play a minor role. The idea that galaxies are defined as those objects where gas can quickly cool predates the current hierarchical model (Hoyle 1953), and has been invoked to set the scale for survival versus disruption (Rees & Ostriker 1977, White & Rees 1978). It has proved difficult to compare the predictions of this model with non-linear structures, such as the internal properties of galaxy clusters. Numerical simulations had ubiquitously failed to find surviving substructure or “halos orbiting within halos” (e.g. Katz & White 1993, Summers et al. 1995, Frenk et al. 1996). It was generally thought that the so called “over-merging” problem could be overcome by the inclusion of baryonic component to increase the potential depth of galactic halos. Analytic work suggested that over-merging was due entirely to poor spatial and mass resolution (Moore, Katz & Lake 1996). This has been verified by higher resolution simulations of clusters in which galactic halos survive without any inclusion of gas dynamics (Moore et al. 1998, Ghigna et al. 1998, Klypin et al. 1998). When a galaxy and its dark matter halo enter a larger structure, the outer regions are stripped away by the global tides and mutual interactions. The central region survives intact so that a galaxy may continue to be observed as a distinct structure within a cluster, with its own truncated dark matter halo (Natarajan et al. 1998). In a hierarchical universe, galaxies form by a similar merging and accretion process as clusters (Klypin et al. 1999). Over-merging on galactic scales is a necessary requirement otherwise previous generations of the hierarchy would preclude the formation of disks. Observations ssuggest that over-merging has been nearly complete on galactic scales. The Milky Way contains just 11 satellites within its virial radius with $`\sigma _{satellite}/\sigma _{halo}>0.07`$, i.e. that is equivalent to $`\sigma _{satellite}=10\mathrm{km}\mathrm{s}^1`$ (c.f. Mateo 1998 and references within). The same velocity dispersion ratio in a cluster corresponds to counting galaxies more massive than the Large Magellanic Clouds $`\sigma _{LMC}50\mathrm{km}\mathrm{s}^1`$; there are 500-1000 such systems in a rich cluster (Binggeli et al. 1985, Driver et al. 1999). The same discrepancy exists at higher masses. The Coma cluster contains $`>30`$ galaxies brighter than the characteristic break in the luminosity function, $`L_{}\sigma >200\mathrm{km}\mathrm{s}^1`$ (Lucey et al. 1991). Scaling this limit to a galaxy halo we find just 2 satellites in the Milky Way or 3 near Andromeda. Why should substructure be destroyed in galactic halos but not in clusters? Analytic calculations suggested that galaxies should contain more satellites than observed (Kauffmann et al. 1993). The shape of the power spectrum varies over these scales such that galaxies form several billion years before clusters and the mass function of their progenitor clumps may differ. Furthermore, as the power spectrum asymptotically approaches a slope of -3, clumps of all masses will be collapsing simultaneously and the timescale between collapse and subsequent merging becomes shorter. These effects may conspire to preferentially smooth out the mass distribution within galactic halos. In this letter we use numerical simulations to study the formation of galactic halos with sufficient force and mass resolution that can resolve satellites as small as Draco. This allows us to make a comparative study with observations and simulations of larger mass halos. ## 2. Substructure within galaxies and clusters We simulate the hierarchical formation of dark matter halos in the correct cosmological context using a high resolution parallel treecode pkdgrav. An object is chosen from a simulation of an appropriate cosmological volume. The small scale waves of the power spectrum are realised within the volume that collapses to this object with progressively lower resolution at increasing distances from the object. The simulation is then re-run to the present epoch with the higher mass and force resolution. We have applied this technique to several halos identified from a $`10^6`$ Mpc<sup>3</sup> volume, including a cluster similar to the nearby Virgo cluster (Ghigna et al. 1998) and a galaxy with a circular velocity and isolation similar to the Milky Way. The cosmology that we investigate is a universe dominated with a critical density of cold dark matter, normalised to reproduce the local abundance of galaxy clusters. The important numerical parameters to remember are that each halo contains more than one million particles within the final virial radius $`r_{vir}`$, and we use a force resolution $`0.1\%r_{vir}`$. Further details of computational techniques and simulation parameters can be found in Ghigna et al. (1998) and Moore et al. (1999). Here we focus our attention directly on a comparison with observations. Figure 1 shows the mass distribution at a redshift $`z=0`$ within the virial radii of our simulated cluster and galaxy. It is virtually impossible to distinguish the two dark matter halos, even though the cluster halo is nearly a thousand times more massive and forms 5 Gyrs later than the galaxy halo. Both objects contain many dark matter substructure halos. We apply a group finding algorithm to extract the sub-clumps from the simulation data and use the bound particles to directly measure their kinematical properties; mass, circular velocity, radii, orbital parameters (c.f. Ghigna et al. 1998). Although our simulations do not include a baryonic tracer component, we can compare the properties of these systems with observations using the Tully-Fisher relation (Tully & Fisher 1977). This provides a simple benchmark for future studies that incorporate additional physics such as cooling gas and star-formation. Figure 2 shows the observed mass (circular velocity) function of substructure within the Virgo cluster of galaxies compared with our simulation results. The circular velocities of substructure halos are measured directly from the simulation, whilst for the Virgo cluster we invert the Binggeli et al. luminosity function data using the Tully-Fisher relation. There are no free parameters to this fit. The overall normalization of the simulation was fixed by large scale clustering properties and we then picked a cluster from a low resolution run that had a dispersion similar to Virgo. We count as a remarkable success that this model reproduces both the shape and amplitude of the galaxy mass function within a cluster. Also plotted in Figure 2 is the cumulative distribution of the 11 observed satellites that lie within 300 kpc of the Milky Way. Where necessary we have converted 1d velocity dispersions to circular velocities assuming isotropic velocity distributions. The model over-predicts the total number of satellites larger than the dSph’s by about a factor of 50. The distribution of circular velocities for the model galaxy and cluster can be fitted with a power law $`n(v/V_{vir})(v/V_{vir})^4`$, similar to that found by Klypin et al. (1999) for satellites in the proximity to galactic halos. The mass function within these systems can be approximated by a power law with $`n(m/M_{vir})(m/M_{vir})^2`$. The tidally limited substructure halos have profiles close to isothermal spheres with core radii equal to our resolution length - increasing the resolution only makes the halos denser and more robust to disruption (Moore et al. 1998). ## 3. Discussion Either the hierarchical model is fundamentally wrong, or the substructure lumps are present in the galactic halo and contain too few baryons to be observed. The deficiency of satellites in galactic halos is similar to a deficiency of dwarf galaxies in the field (e.g. Kauffmann et al. 1993). One possibility is that some of the missing satellites may be linked to the high velocity clouds (Blitz et al. 1999). Numerous studies have invoked feedback from star formation or an ionizing background to darken dwarfs by expelling gas and inhibiting star formation in low mass halos (Dekel & Silk 1986; Quinn, Katz & Efstathiou 1996). The case for feedback has always been weak. Galaxies outside of clusters are primarily rotationally supported disks, their final structure has clearly been set by their angular momentum rather than a struggle between gravity and winds. The strongest star-bursts seen in nearby dwarf galaxies lift the gas out of their disks, but the energy input is insufficient to expel the gas and reshape the galaxy (Martin 1998). While there might be little consequence to darkening dwarfs in the field, spiral disks will neither form nor survive in the presence of large amounts of substructure. The strongly fluctuating potential of clumpy collapses inhibits disk formation and has been shown to be an effective formation mechanism for creating elliptical galaxies (Lake & Carlberg 1988; Katz & Gunn 1991; Steinmetz & Muller 1995). Figure 3 shows the proto-galactic mass distribution at a redshift of 10, just a billion years after the big-bang. The smallest collapsed halos that we can resolve have a mass of $`10^7M_{}`$, not much larger than globular clusters. The problem of baryonic trapping by star formation in lumps arises before the first QSO’s could ionize the intergalactic material (however, see Haiman, Abel & Rees 1999). The second problem is that the lumps do not dissolve by $`z=1`$ or even by the present day, as we have shown. Even if we make the most optimistic assumptions about the fate of gas, the movements of this small tracer component will not lead to the destruction of the dark matter substructure. The most obvious observational constraints are the existence of old thin disks (Wielen 1972) and cold stellar streams (Shang et al. 1998). Just as gravitational perturbations from encounters transform disk galaxies to spheroidals in clusters (Moore et al. 1996), the passage of these lumps will heat any disk within the halo. Stellar disks extend to $`10\%`$ of the virial radius of the dark matter halo although HI can be observed to much larger distances (25% of $`r_{vir}`$ for some LSB galaxies). We find that the orbits of satellites within our simulated halos have a median apocentric to pericentric distance of 6:1, therefore over the past 10 billion years disks will suffer many thousands of impulsive shocks and resonant heating. The single accretion of a satellite as large as the Large Magellanic Cloud, has a devastating effect on the disk of the Milky Way (Toth & Ostriker 1992; Ibata et al. 1998, Weinberg 1998). While recent work has noted that disks embedded within live halos may precess in response to a single satellite and avoid strong vertical heating (Huang & Carlberg 1997; Velazquez & White 1998), there are far too many clumps in our simulations for this mechanism to be effective. An estimate of the heating can be obtained using the impulse approximation. Each dark halo that passes nearby or through the disk, will increase the stellar velocities across a region comparable to the size of the perturber by an amount $`\delta vGm_b/r_bV`$ where $`V`$ is the impact velocity. We measure $`m_b`$, $`r_b`$ and $`V`$ for each clump that orbits through the stellar disk. Summing the $`\delta v`$’s in quadrature over 10 Gyrs, we find that the energy input from encounters is a significant fraction of the binding energy of the stellar disk $`M_dv_c^2`$, where $`M_d`$ and $`v_c`$ are the disk mass and rotation velocity respectively. The heating is more than sufficient to explain the age-temperature relation for disk stars (Wielen 1974), although the validity of the impulse approximation needs to be examined using numerical simulations. We note that the existence of old thin disk components, or galaxies such as NGC 4244 that does not have a thick disk (Fry et al. 1999), presents a severe problem for hierarchical models. Substructure can be probed by gravitational lensing even if stars are not visible in the potential wells (e.g. Hogan 1999). Multiply imaged quasars are particularly sensitive to the foreground mass distribution; the quadruple images QSO 1422+231 cannot be modeled with a single smooth potential (Mao & Schneider 1998) and requires distortions of $``$ 1% of the critical surface density within the Einstein radius. Dark matter substructure located in projection near to the primary source would create such distortions. If we extrapolate our mass function to smaller masses, we expect $`10^5`$ clumps with $`v_c/V_{200}>0.01`$ ($`m_b10^6M_{}`$). This may cause many gravitationally lensed quasars to show signs of substructure within the lensing potentials. Cold dark matter candidates, such as axions and neutralinos, can be detected directly in the laboratory. Many proposed and ongoing experiments will be highly sensitive to the phase space distribution of particles at our position within the galaxy, yet calculations of experimental rates still assume that CDM particles passing through minute detectors have a smooth phase space distribution. We have shown that CDM halos are far from smooth, furthermore, the particle velocities in a single resolution element have a discrete component that results from the coherent streams of particles tidally stripped from individual dark matter halos. We may also an expect enhanced gamma-ray flux from neutralino annihilation within substructure cores (Lake 1990, Bergstrom et al. 1998). ## 4. Summary In a hierarchical universe, galaxies are scaled versions of galaxy clusters, with similar numbers and properties of dark matter satellites orbiting within their virial radii. The amplitude and tilt of the power spectrum, or varying the cosmological parameters $`\mathrm{\Omega }`$ and $`\mathrm{\Lambda }`$ will have little effect on the abundance of substructure. These only slightly alter the merger history and formation timescales. Any difference in merger history will be less than what we have already explored by comparing the cluster to the galaxy. Furthermore, we have shown that the properties of the substructure do not change over a 4 Gyr period, therefore an earlier formation epoch will not change these results. If we appeal to gas physics and feedback to hide $`95\%`$ of the Milky Way’s satellites then we must answer the question why just $`5\%`$ formed stars with relatively normal stellar populations and reasonably large baryon fractions. If this problem can be overcome, then the substructure has several observational signatures, namely disk heating, gravitational lensing and direct/indirect particle dark matter detection experiments. Unfortunately, the existence of old thin disks with no thick/halo components may force us to seek a mechanism to suppress small scale power e.g. free streaming by a neutrino of mass $`1`$ keV (Schaeffer & Silk 1988).
no-problem/9907/astro-ph9907391.html	ar5iv	text	# Two-Dimensional Line Strength Maps in Three Well-studied Early-Type Galaxies ## 1 Introduction The study of the stellar populations of local galaxies plays a major role in understanding galaxy formation and evolution, not in the least because we can only understand high-redshift galaxies by comparing them to the local universe. In recent years significant progress has been made in understanding the stellar populations of nearby galaxies, for ellipticals as well as spirals. Elliptical galaxies are found to fit well single-age, single-metallicity (SSP) models (e.g. Bruzual & Charlot 1993, Worthey 1994, Vazdekis et al. 1996), suggesting that most stars that we currently see were formed on a small time scale. Large ellipticals generally have non-solar Mg/Fe ratios (Peletier 1989, Worthey et al. 1992), while it appears that there is a large spread in age among them (González 1993), although this is not generally accepted (see Kuntschner & Davies 1998). Spiral galaxies have been less well studied, because of the presence of extinction and star formation makes stellar population studies much more difficult. Recent reviews about the stellar populations in bulges, and their differences with ellipticals, are, e.g., Worthey (1998) and Peletier (1999). This paper is part of a series of papers, with the aim of studying galactic bulges in the same detail as ellipticals. To do this, we have first developed a new spectrophotometric stellar population model, that can be used to interpret observed colours and absorption lines of galaxies (Vazdekis et al. 1996, hereinafter Paper I). To test this model, we have presented in a subsequent paper high-quality long-slit observations of three representative early-type galaxies, in many lines and colours, and applied the model to it. This allowed us to obtain a detailed understanding of ages and abundance distributions of these well-studied, ’standard’, galaxies (Vazdekis et al. 1997, hereinafter Paper II). In the current paper, we again present the same three galaxies (NGC 3379, NGC 4472 and NGC 4594) but now using two-dimensional data, derived using Integral Field Spectroscopy (IFS). This is the first time a detailed study has been performed of galaxy absorption line strengths using a multi-fiber instrument. A similar, but much more limited study has been performed by Emsellem et al. (1996), who used the multi-lenses IFU system TIGER on the CFHT to obtain 2-dimensional kinematic and line strength maps of the Sombrero galaxy. Although the number of absorption lines studied by them is much smaller, their paper offers us an excellent opportunity to carefully check our IFU results. Two-dimensional spectroscopy of early-type galaxies has also been obtained by Sil’chenko and collaborators (e.g. Sil’chenko et al. 1997, 1999), but in their papers only two-dimensional kinematics has been presented. There are several advantages in using IFS as opposed to multiple long slit spectra. Once the problems of data reduction have been overcome, the larger spatial coverage will be a great advantage in studying is much the galaxy kinematics. For example, one can study the triaxiality of a system much better than using long slits (see e.g. Statler 1991). Studying e.g. whether photometric and kinematic centres coincide is now much easier. With regard to stellar populations: one can find two-dimensional stellar population structures, like young disks and remnants of mergers, purely by analyzing line strength maps. One can detect small amounts of star formation everywhere in the area covered. And, compared to long-slit spectroscopy, the sensitivity at a certain radius is much larger, since more area is covered. We will discuss in this paper results obtained using a fiber bundle of 125 fibers. In the near future however, thousands of fibers will be used, which undoubtedly will make IFS much more popular and important than it is at present. In this paper, several tests will be presented to establish our IFS reduction techniques. But also, we will present some entirely new data, for the H$`\gamma `$ line, and for the Ca II IR triplet. H$`\gamma `$ is a good age-indicator, when used together with e.g. Mg<sub>2</sub> (Worthey & Ottaviani 1997), and can be used even when H$`\beta `$ is filled in by emission. The Ca II IR triplet in the near-infrared is one of the few lines depending strongly on stellar surface gravity (see e.g. Carter et al. 1987), allowing us to possibly measure or constrain the IMF slope. In this paper we present the Ca triplet (which we will address as Ca T in the rest of this paper) calibrated using stars of Díaz et al. (1989). We will present very little kinematics here, only a comparison with Emsellem et al. (1996), to show the reliability of our velocity dispersion corrections. Kinematics, including line profile analysis, will be presented in a future paper (Prada et al. 1999, in preparation). Here we present stellar population analysis of the 3 well-studied galaxies. At the same time, we also observed 2 other galaxies: the dwarf elliptical M 32 and the Sb NGC 2841. The analysis of these two galaxies will also be presented in future papers. The organization of our paper is as follows. In Section 2 the observations are described. In Section 3 we described the data reduction. In Section 4 a comparison is made with Paper II and Emsellem et al. (1996). In Section 5 we present our results, amongst other the population synthesis. In the last Section our conclusions are given. ## 2 Observations ### 2.1 The Instrument The data analyzed in this paper were obtained on Feb. 15-17, 1997, at the Observatorio del Roque de los Muchachos, on the island of La Palma. We used the 2D-FIS (Two-dimensional Fiber ISIS System), which is placed between the f/11 Cassegrain focus of the 4.2 m William Herschel Telescope (WHT) and the ISIS double spectrograph. A detailed technical description is provided by García et al. (1994); here we will only recall its main characteristics. The core of the system consists of a 2.5 m long bundle formed by 125 optical fibers, each 200 $`\mu `$m (0.9” on the sky) in diameter, arranged in two groups in the focal plane. One has 95 fibers, forming an array 9.4” $`\times `$ 12.2” on the sky, while the other, a ring of 38” in radius with 30 fibers, is intended to collect the background light. The relative positions of the fibers at the telescope’s focal plane are known very accurately, the distance between two adjacent fibers being about 1.1”. Figure 1 shows the actual distribution of the fibers in the central rectangle. At the slit the fibers have been arranged linearly (with a distance of 425 $`\mu `$m between adjacent fibers). With this type of arrangement, a set of spectra - corresponding to 125 zones in the circumnuclear region - is recorded in each exposure. The advantage of the system as opposed to, say, a Fabry-Perot is that all the spectral and spatial information is recorded simultaneously and that the spectral coverage is determined by the spectrograph alone. Details on the technique itself may be found in Arribas, Mediavilla, & Rasilla (1991), and references therein. ### 2.2 Our Observations The observations were made using simultaneously the blue and red arms of the ISIS double spectrograph. In the blue arm, the R300 grating was used, giving us a pixel scale of 96 km/s (1.55 $`\mathrm{\AA }`$), a range between 4100 and 5650 $`\mathrm{\AA }`$, covering most of the lines of the Lick system, including H$`\gamma `$ (Worthey & Ottaviani 1997), and a resolution of 2 pixels. In the red, we centered our spectrum on the Ca T. Here we had a pixel size of 28 km/s (resolution 2 pixels) and a spectral range from 8200 to 9000 $`\mathrm{\AA }`$. On both arms 1024 $`\times `$ 1024 thinned TEK CCDs were used. A dicroic, centered at 6100 $`\mathrm{\AA }`$, was used, and in the red arm a GG495 order sorting filter was placed as well. Weather conditions during the first two nights, during which the three galaxies were observed, were photometric, with a seeing of $``$ 1”. A log of the observations is presented in Table 1. Each night, we also observed some out-of-focus standard stars, using the same setup. In total 20 stars were observed. ## 3 Data Reduction Since observations of this type are not very common, and some steps of the data reduction process are significantly different from reducing e.g. long-slit data, we will cover the reduction process in quite some detail, and describe a number of tests showing that our results are reliable. To summarize, we first extract the light in the individual galaxy fibers after removing the stray light between the fibers. We then correct each fiber for its relative transmission using a twilight sky image, and correct for pixel to pixel sensitivity variations using tungsten lamp flatfields. After this we calibrate the spectra in wavelength using CuAr arc lamp exposures, remove cosmic rays, and subtract the sky background. We then extract for each fiber absorption line indices as well as kinematic information, and make maps of these. All of this work was done in IRAF, with some programs written by ourselves. Here we describe these steps in more detail. ### 3.1 Stray Light The first reduction step was to remove the stray light. This light has a diluting effect on the spectra, making it more difficult to measure line indices of the object. For 2D-FIS the individual fibers are well separated along the spatial direction of the CCD detector: the distance between adjacent fibers is 6 pixels, and the aperture width is 4 pixels. In Fig. 2 a cut along the spatial direction shows the relative separations of the fibers. For this set up, the optical cross-talk (contamination on a fiber due to the wings of the nearer fibers) is negligible if the difference in intensity between adjacent fibers is not too large. This is the case for most objects, except in focus standard stars. The object images are not very much affected by cross-talk because the fibers are conveniently ordered along the spectrograph slit. We determined the stray light to be subtracted by fitting legendre polynomia through the spatial regions between the fibers, and subtracted it. Subtracting stray light however barely affects the final line indices. This is mostly due to the fact that the galaxy light across the frame doesn’t vary very much, so that an addition of a few percent of light from elsewhere in the galaxy will not affect the integrated line index. Experimentally we found that for the Sombrero in the blue the average values of the line indices did not change by more than 0.2 $`\mathrm{\AA }`$. The RMS fiber difference between correcting and not correcting for stray light was about 0.3 $`\mathrm{\AA }`$ per index (0.005 for Mg<sub>1</sub> and Mg<sub>2</sub>), of the same order as the numbers in Table 3, indicating that by correcting for stray light we might improve the data. For that reason the correction has been applied, although it remains uncertain whether it was really necessary. ### 3.2 Flatfielding and Extracting the Fibers Flatfielding was done to remove sensitivity variations from pixel to pixel of the system, and fiber to fiber transmission variations. The TEK CCDs that we used suffer considerably from fringing in the red (Peletier 1996), but also in the blue, all the way up to 5000 $`\mathrm{\AA }`$. We measure fringing amplitudes of about 10% in our whole red region and between 4200 and 4400 $`\mathrm{\AA }`$, and about 5% between 4400 and 5000 $`\mathrm{\AA }`$. As explained in the technical note, this fringing depends only on the position of the telescope, and can be corrected for very well. For measuring maps of very tiny absorption lines however, our final accuracy might be determined by how well we can correct for this fringing. At the position of each object, star and galaxy, we also measured a Tungsten flatfield, and used this to correct for the fringing. The flatfield image was corrected for stray light, after which the fibers spectra were extracted. To the extracted spectra polynomials (10th order in the blue and 3rd order in the red) were fitted, in order to correct for the spectral response of the Tungsten lamp, but not for the fringes. Following this the object fibers (galaxy or star) were extracted using the flatfield image as a tracer. This is justified because flat and object (and arc) images are observed at the same position of telescope and then the shifts of each fiber are negligible. They were then divided by the normalized flatfield, and following that each fiber was divided by its relative transmission, determined from twilight sky frames taken at the beginning and end of each night. ### 3.3 Wavelength Calibration, Cosmic Ray and Background Removal Wavelength calibration was done using CuAr arc lamp exposures taken before or after each object frame. For each fiber separately the corresponding arc spectrum was extracted, and low order polynomials were fitted to pixel positions of the arc lines as a function of wavelength. On the average the accuracy of the fit for an arc line was 0.08 $`\mathrm{\AA }`$ (4.8 km/s) in the blue and 0.05 $`\mathrm{\AA }`$ (1.7 km/s) in the red. Using these polynomials the spectra were rebinned in log(lambda). Since for all of our galaxy we had observed at least 4 data frames of 1800s each, we could combine them and remove the cosmic rays with a simple median filtering technique. For the stars, for which only one exposure was taken, very few cosmic rays were present because of their short exposure times, and these were removed using a sigma-clipping algorithm. We then proceeded to subtract the sky background. Two methods were tested. We took separate sky exposures before and after the galaxy exposures. These were scaled to the galaxy integration time and subtracted. The second method was to take all the fibers in the outer ring together to determine an average sky spectrum, and to subtract that from the galaxy fiber spectra. The first method suffers from the fact that the sky varies with time, so that the sky spectra have to be scaled in some way. Using the second method one has the problem that the outer ring fibers still have some galaxy contribution in them. In our worst case, in NGC 4472, at 38” the galaxy is about 15 $`\times `$ fainter than at 10” (Peletier et al. 1990), so some contamination of the spectra in the outer regions can be expected. However in these three objects colour and line strength gradients are so small (Peletier et al. 1990, Paper II, González 1993) that such a small contamination of max. 6% will not be noticed. In the blue the sky background in the continuum is so low compared to the galaxies that no trace of it was left after subtracting. In the red the sky subtraction also was very precise, because the continuum flux of the outer parts of the galaxies was always at least 8 times larger than the background. In the region of the Ca II IR triplet there are no sharp emission features in the sky background, only a broad emission line from O2 (Nelson & Whittle 1995). ### 3.4 Making Maps and Datacubes On the now reduced spectra we continued to measure absorption line strengths. This was done by first determining the stellar radial velocity and velocity dispersion of each fiber, deredshifting it, measuring equivalent widths as explained e.g. in Paper II and Worthey (1994), calibrating these equivalent widths to put them onto the Lick system in the blue (Faber et al. 1985, Gorgas et al. 1993, Worthey et al. 1994) and the system of Díaz et al. (1989) in the red, and making two-dimensional maps of each of them. Velocities and velocity dispersions were determined using Bender’s (1990) Fourier Quotient method. A comparison with the literature for the Sombrero galaxy is given in the following section. After deredshifting the spectra using these velocities, indices were determined, which then had to be corrected for velocity broadening. To calibrate this velocity dispersion correction we used our 20 stars (for each star the average of all fibers was taken), and determined for each index a velocity dispersion correction curve (as described in Paper II). The same was done in the blue and in the red. In Table 2 our index definitions are given. We found that the Ca T indices were affected very much by the fact that the two continuum bands were very far from the feature, so that the line strengths of the features are severely affected by small variations in the continuum. For that reason we also tried a few different definitions for the 3 lines of the triplet, from studies of galactic globular clusters. Three other definitions were adopted, from Armandroff & Zinn (1988), Armandroff & da Costa (1991) and Rutledge et al. (1997) (see Table 2). Due to the fact that neither the stars on the Lick and Díaz systems are flux calibrated, nor our own spectra, we had to put our observations onto these systems using standard stars. To do this, we took maps of out-of-focus stars,measured their average line indices, and determined for each index a linear relation between the literature index and our measurements. In the red only a single offset was sufficient. In the blue our sensitivity curve is considerably different from that of the Lick IDS, used to define the Lick system, and therefore the conversions are somewhat more complicated. The calibration relations are given in Table 3. The RMS scatter now indicates how accurate our indices are in absolute terms. As one can see from Table 3, the RMS scatter is similar to the scatter for the Lick system itself (see Worthey et al. 1994, Table 1), which implies that the absolute values of our line indices are probably better than those of the Lick system. Having established the accuracy of our average fiber measurement, we then investigated systematic fiber-to-fiber differences using out-of-focus stars. In principle differences in spectral response can occur because each of the fibers sees a different area of the primary mirror, but moving the out-of-focus stars slightly in the focal plane showed that this effect was negligible. The out of focus stars were reduced, and for each fiber absorption line indices were determined. Taking all the fibers for which the integrated flux was larger than 30% of the flux of the brightest fiber we determined the RMS scatter in the line indices for each star, after applying a 3$`\sigma `$ rejection algorithm. These values are given in the last column in Table 3. The RMS scatter that we found only slightly depends on the signal-to-noise ratio in the stars, showing that these fiber-to-fiber differences are due to some systematic effect. Careful inspection of the data shows that fringing is important in the data, especially in the red, but also somewhat in the blue, and that incorrect flatfielding can affect the values of the indices severely. We therefore think that residual fringing (which is dependent on the position of the telescope) might be responsible for our small residual variations, although it is not excluded that variations in fiber stresses leading to small variations in the continuum could also be responsible. Structures smaller than this level in our maps should probably not be believed. The RMS fiber to fiber differences are on the order of 0.3 $`\mathrm{\AA }`$ and 0.010 for the indices expressed in magnitudes. This means that very small gradients are hard to determine, but in general this accuracy is good enough to obtain results of the quality currently produced in the literature (see the comparison with Paper II in the next section). As a last step we determine maps for each absorption line index, as well as for the integrated continuum, radial velocity, velocity dispersion and higher order velocity moments. The kinematics will be described in a following article (Prada et al. 1999, in preparation). Maps were reconstructed using triangular interpolation and regridded to frames with a pixelsize of 0.48”. Since the data are undersampled the maps will contain some small amount of aliasing. However, the comparison of the reconstructed intensity maps with HST data (see Section 4.1) shows that this is minor and hard to detect. We see however that in the three galaxies the position of the galaxy center in the red is somewhat different from in the blue. This is due to differential refraction (see for an example Arribas et al. 1997). The differential refraction within the red or blue band however was so small compared to the spatial resolution that we didn’t correct for it. To give an idea about the final obtained resolution Arribas et al. (1997) looked at reconstructed images of in focus stars, and found them to have FWHM values of 1.2”. For the data in this paper the analysis of Section 4.1 shows that similar values have been obtained here. In Fig. 3 we show all line index maps of the three galaxies. To make the interesting features visible we have for each map subtracted its median value, and then divided all pixels by the standard deviation of the map. Many maps only show noise, indicating that the line index everywhere in this central region of the galaxy is the same. However, also some prominent features are seen. These will be discussed in Sections 4 and 5. ## 4 Comparison with the Literature ### 4.1 Continuum images The first test we applied was to reconstruct the continuum images in red and blue, fit ellipses to these images, and compare them with recent HST archive images. The spectra were collapsed in the red between 8427 and 8761 $`\mathrm{\AA }`$, and in the blue between 4236 and 5643 $`\mathrm{\AA }`$. On the reconstructed images (see Fig. 3) ellipses were fit using Galphot (see Jørgensen et al. 1992). In Fig. 4 we show the blue surface brightness profiles, normalized in the center, of the three galaxies, and the profiles of the ellipticity (1-$`\frac{b}{a}`$), and the major axis position angle. Also shown are profiles determined on recent HST-WFPC2 images in V (F555W), convolved to a seeing of 1.2” (FWHM). Plotted on the abscissa is $`\sqrt{\mathrm{major}\times \mathrm{minor}\mathrm{axis}\mathrm{radius}}`$, while the surface brightness profiles are scaled by an arbitrary number. The reconstructed images, together with the convolved HST frames, are shown in Figure 5. The agreement between the continuum images and the convolved HST V-band images is good. Their slope can be seen to be slightly shallower, which is what one expect from colour gradients, since the effective wavelength of our images is somewhat bluer than 5550 $`\mathrm{\AA }`$. The fact that the ellipticity profiles beyond 3” start diverging is due to the fact that here not the whole ellipse in included in our 2D-FIS images. Agreement in the red is not as good as in the blue, but still good. We conclude that the good agreement between our photometry and HST shows that stray light is not a problem in the analysis of our spectra. ### 4.2 Comparison with long-slit spectra One of the main reasons why we are studying these 3 galaxies in such detail, is that accurate line strength profiles are available for most of the Lick indices from Paper II. In this paper the major axis of NGC 3379 and NGC 4472, and the minor axis of NGC 4594 were observed with the same spectrograph, ISIS, on the WHT, but using its regular long-slit mode. The seeing of those observations was not as good as for this paper - their effective seeing was about 2.5” - 3”. The spectra of Paper II were calibrated onto the Lick system and corrected for velocity dispersion in the same way as was done for the data in the current paper. For the comparison we turned the images of NGC 4472, NGC 3379 and NGC4594 by resp. -10, 75 and 0<sup>o</sup>, and then extracted the line strength profiles in a column of width 1” through the center. The comparison is given in Fig. 6. In general the comparison is good, although some comments have to be made. The errorbars in the upper right corner take into the account the fiber-to-fiber errors and the error made by converting to the Lick system. One can see that most of the discrepancies can be explained by this error. One also can see that the seeing for our observations was much better - this is clear from e.g. H$`\beta `$, Mg<sub>2</sub> and Mg$`b`$ for the Sombrero galaxy, where the inner disk is now much better resolved. H<sub>β</sub> and Fe 5015 in the center of the Sombrero galaxy are lower than elsewhere, because of contamination by emission; in case of the Fe 5015 line by the \[OIII\] line at 5007$`\mathrm{\AA }`$. Agreement for well-studied indices, like $`<`$Fe$`>`$ and Mg<sub>2</sub> is excellent. Apart from these differences there appear to be some small global offsets, in Ca 4227, and maybe somewhat in the G-band. Our Ca 4227 is now in better agreement with the models of e.g. Worthey et al. (1994) than the data of Paper II, but the line strength is still not as large as one would expect (see Paper II). Sometimes small ’jumps’ are seen in our index-profiles, like e.g. at -5” for the Sombrero galaxy, due to the problems described in Chapter 3. Apart from these, our index profiles are as good as most currently available profiles in the literature. ### 4.3 Comparison with TIGER The only observations of a similar kind available in the literature are those taken with TIGER, an instrument built in Lyon (Bacon et al. 1995). Tiger uses multilenslets, and compared to our current observations has a better sampling, but a much smaller wavelength range. Two detailed two-dimensional studies of the stars have been published (Emsellem et al. (1996) & Bacon et al. (1994)), of which the first deals with the Sombrero galaxy. The agreement with Emsellem’s figures 8 and 9 is good. In our study the same kinematic features are found. Although we will make a more detailed comparison in a next paper (Prada et al., in preparation), we show here that the radial velocitiy and the velocity dispersion along the major axis agree within 10 km/s (Fig. 7). In Emsellem et al. (1996) a comparison is made with Kormendy et al. (1988), which also is satisfactory. To compare our Mg$`b`$ image with Emsellem et al. (1996)’s image, we first corrected it for emission lines. As explained in Goudfrooij & Emsellem (1996) one of the wings of this line includes an emission line of \[NI\] at 5199 $`\mathrm{\AA }`$, which has the effect of increasing the continuum and so also the feature’s strength. This emission is relatively strong in the Sombrero (Ho et al. 1997). We used the fact that this feature is very sharp, as compared to the wide stellar features, and removed it by interpolating across it, as one does for bad pixels. The results of this operation can be seen for Mg b and Mg<sub>2</sub> (which is also slightly affected) in Fig. 6. The agreement between our Mg$`b`$ map and Fig. 18 of Emsellem et al. is good, reproducing in both cases the inner disk, which is strong in Mg$`b`$, and the peak in the center. Since the strength of the emission line drops off quickly away from the center, it is possible to obtain an emission line spectrum by taking an average spectrum at about 2-3<sup>′′</sup> from the center, scaling the continuum, adjusting the velocity dispersion, and subtracting it from the central spectrum. Some emission was also found in the center of NGC 3379, but much fainter, and not strong enough to detect the \[NI\] 5199 $`\mathrm{\AA }`$ line, and less extended. ## 5 Results ### 5.1 Radially Averaged Profiles In this section we present radially averaged profiles of the 3 galaxies. They have been averaged on the ellipses calculated on the continuum images. We see in Fig. 8 that the profiles of the three galaxies generally lie on top of each other, showing that the three galaxies have similar stellar populations (see also Paper II). An exception is the inner region of the Sombrero galaxy, not only because of the emission lines, but also because we see large colour gradients in lines like e.g. H$`\gamma `$ and Mg<sub>2</sub>. In agreement with long slit studies most indices have non-zero radial gradients. These gradients are given in Table 4. In the red the Ca T indices seem to increase somewhat going outward, consistent between the three of them. ### 5.2 Two-Dimensional Features in the line strength maps One of the main advantages of IFS is that one can rather easily make absorption line maps, and use them to derive conclusions about galaxy formation. In this paper we have found that the inner disk of the Sombrero is very well visible in the Mg maps (Mg<sub>1</sub>, Mg<sub>2</sub> and Mgb), but is almost invisible in maps of Fe and Ca indices. This is mainly due to the fact the relative noise in the Fe and Ca maps is much larger than in the Mg features. To show this, we show in Fig. 9 a diagram of Mg<sub>2</sub> vs. $`<`$Fe$`>`$ that includes the nuclei of some well-observed ellipticals and spirals, some models of Vazdekis et al. (1996), and the radial profile of the Sombrero galaxy (see Section 5.1). The figure has been taken from Peletier (1999). One can see that the slope of the Sombrero is very similar to that of the models, indicating that in the Sombrero Mg/Fe remains constant, although \[Mg/Fe\] is larger than solar. This results appears surprising at first view. A much used explanation of overabundant Mg/Fe ratios is that the enrichment of Mg is mainly due to massive stars producing SN type II, while Fe is mainly produced in SN type Ia (see e.g. Worthey et al 1992). Since the latter originate in binaries, there is a time delay of a few times 10<sup>8</sup> year between star formation and release of the Fe-peak elements in those SNe type Ia. Elliptical galaxies originally were thought to form their stars on very short timescales, as opposed to disks, and for that reason ellipticals would have enhanced Mg/Fe, while Mg/Fe in disks would be around solar. In an excellent review paper Worthey (1998), summarizing the current observational status of Mg/Fe in external galaxies, shows that this explanation is probably incorrect. Although giant ellipticals clearly have supersolar Mg/Fe ratios, this does not seem to be the case for fainter ellipticals (e.g. Davies et al. 1993), whose Mg/Fe are close to what is predicted by models with solar abundance ratios. Also, for spirals the situation is not so clear any more. Jablonka et al. (1996) find that their most luminous bulges have enhanced \[Mg/Fe\], but fainter ones do not. Sil’chenko (1993) finds that \[Mg/Fe\] is solar in most bulges of other spiral galaxies, although her sample contains very few massive galaxies. Based on these and other results Worthey concludes that the data seems to indicate that \[Mg/Fe\] is near zero in most galaxies of all types with velocity dispersion less than about 225 km/s, and that galaxies with velocity dispersions above that values are progressively more Mg-enhanced. Peletier (1999) slightly refines this for spirals, for which the potential is not determined only by random motions, replacing the words ’velocity dispersion’ by kinetic energy, or escape velocity. For the Sombrero galaxy the potential well is determined by both random and ordered, rotational motion ((v/$`\sigma `$)<sub>cen</sub> $``$ 1, Emsellem et al. 1996). Both $`v`$ and $`\sigma _{cen}`$ are larger than 225 km/s, implying, as is indeed observed, that Mg/Fe has to be larger than solar. Although the central region of the Sombrero is clearly dominated by a fast rotating disk, its Mg/Fe is much more similar to the centres of other giant early-type galaxies (for which Mg/Fe $`>`$ 0) than to large disks, like the disk of our Galaxy. If one believes that Mg/Fe only depends on the timescales of enrichment by SN type Ia and II, the fact that the range in observed central Mg/Fe amongst giant early-type galaxies is small would then imply that the timescale on which stars formed there would be independent on whether they are in a disk or not. Although we cannot measure these timescales at present, this could sound rather implausible, and it might be much more natural to assume that Mg/Fe depends on the IMF, which in turn would be dependent on the environment through the stellar potential. At present we don’t have any direct evidence that the IMF varies in this way, but it is much easier to explain the observations in this way, as opposed to invoking differences in timescales for enrichment by SN Ia and II. The inner disk in the Sombrero is the only two-dimensional feature that we have found in these three galaxies. This shows that the stellar populations in the central regions of NGC 3379 and NGC 4472 are to a good approximation homogeneous. ### 5.3 Population Synthesis In this section we compare the measured line-strengths with the stellar population synthesis model of Paper I. This comparison is very similar to the one performed in Paper II, apart from the fact that here we have a few extra lines (H$`\gamma `$ and the Ca T), and that our spatial resolution here is higher. In Paper II we showed that very good fits for the three galaxies could be obtained with single-age, single-metallicity models, SSP’s. Full evolutionary models provide solutions that are very similar, with small ranges in age and metallicity. The best fits yield good agreement for all broad band optical and near infrared colors as well as for many important line indices. Problems however are encountered when fitting Fe-dominated lines and the Ca4227 features. To be able to use the higher spatial resolution of these observations, and to avoid too much duplication with Paper II, we have analyzed the light only in an inner aperture with a diameter of 1.7$`\mathrm{}`$ centered on the galaxy center. Best-fitting model solutions were found by minimizing a merit function M of the type described in Paper II. To calculate M, for each color or line index the square of the difference between the observed and synthetic color or index is divided by the estimated observational error and then multiplied by a certain assigned weight. To assign the weight to each index we have chosen to group them by their main contributing element, and given the same global weight to each group. These groups have been defined are as follows: Balmer indices (H$`\beta `$, H$`\gamma _A`$, H$`\gamma _F`$), Magnesium-dominated features (Mg<sub>1</sub>, Mg<sub>2</sub>, Mg<sub>b</sub>), Iron-dominated features (Fe4383, Fe4531, Fe5015, Fe5270, Fe5335, Fe5406), Calcium-dominated features (Ca4227 and the three lines of the Ca T) and other features (G, Ca4455, C<sub>2</sub>4668). The reason why Ca4455 is not included in the Ca group is because it is mainly a blend (Worthey et al. 1994). For that reason we have always given a low weight to this line. To survey the solution space we have defined 4 weighting functions: in the first one the same weight is assigned to each of the groups as a whole. To assess the influence of each group individually three other weighting functions were defined where the weight for one group was set to 0, while the other weights remained the same. All these weight functions are summarized in Table 5. To fit the data, we have selected the models for three metallicities: Z=0.008, Z=0.02 (solar) andZ=0.05, and interpolated the output linearly to obtain results at Z=0.014 and 0.035. The age-range of the selected models was 1 to 17 Gyr. We preferred to use the bimodal IMF defined in Paper I, which is similar to the Salpeter IMF, but has fewer stars with mass below 0.6M, since it was shown in Paper II that this IMF provides slightly better fits to precisely the three galaxies to be analyzed here. Four IMF slopes were considered: $`\mu `$ = 1.0, 1.3 (Salpeter), 1.7 and 2.3. We will now discuss the fits for the galaxies individually. The best fits were obtained for NGC 4472, giving a metallicity of Z=0.05, $`\mu `$=2.3 (steeper than Salpeter), and a most likely age of 8 Gyr. The same solution is found for all weight-functions, except for the case in which the Mg group is neglected, in which case an optimal age of 5 Gyr is found, with the same $`\mu `$ and metallicity. One should realize that we are not claiming here that with the Salpeter IMF no reasonable solutions can be found for this galaxy. Using however the bimodal IMF, which appears to be more realistic, at least for our Galaxy (Kroupa et al. 1993) it appears that we need more stars around the mass where the IMF changes slope. The fits show clearly that we cannot obtain at the same time good fits for the Mg group on one hand and the Fe+Ca groups on the other hand. This is the well known problem that giant ellipticals are overabundant in Mg (see Worthey 1998 and references therein for more details). Since the current theoretical models do not allow us to change the \[$`\alpha `$/Fe\] ratio, we will only continue with the uniform weight function for which no group has been neglected. This fit is summarized in Table 6. One sees that most of the features of the Ca group are much less strong than predicted by the model, and that these differences are clearly larger than the expected observational errors. This is the case for the lines of the Ca T, as well as for the Ca 4227 line, for which we already found in Paper II that it was much lower than expected. We would like to warn the reader that the Ca T predictions are based on an stellar library (Díaz et al. 1989) which does not properly cover all the theoretical atmospheric parameters. It is reassuring that the present observational values for the two main features of the Ca T are in agreement with the observational results of Terlevich et al. (1990): while we measure a total EW of 6.5Å they measured 6.7Å. The results for NGC 3379 are very similar as those for NGC 4472. Independent of the weight function the best solution is found for (Z=0.05, $`\mu `$=2.3, age=8 Gyr), except in the case in which the Mg group is ignored, which gives the best solution for Z=0.035, $`\mu `$=2.3, and age=8 Gyr. The best fit with uniform weight function is also tabulated in Table 6. For this galaxy, the Ca group features are lower than predicted, similar to NGC 4472. In general the fact that the residuals in general are larger shows that the quality of the obtained global fit is considerably worse than the fit for NGC 4472. For the central region of the Sombrero galaxy we always obtain Z=0.05, $`\mu `$=2.3 and age=13 Gyr for all our weight functions except for the one where we neglect the Mg group. Here the best fit is obtained for the same IMF slope, Z=0.035 and a slightly larger age (15 Gyr). It seems that possibly better fits could be obtained with even higher metallicities. This can however not be investigated because of the lack of reliable theoretical models with Z $`>`$ 0.05. From the tabulated values we see that most of the features of the Ca and Fe groups show residuals much larger than the observed values and that the Mg/Fe overabundance is very pronounced. Apart from this, we also see that while the two H$`\gamma `$ indices are fairly well matched, this is not the case for H$`\beta `$ which is much lower than any model prediction. From the spectra it can be seen that this is the result of emission lines ’filling in’ the absorption line, and thereby reducing its equivalent width. Overall this galaxy yielded fits that were qualitatively the worst of the three galaxies. To summarize our main, new results from this stellar population analysis: The strengths of the lines of the Ca T are lower than expected, confirming the result of Paper II purely on the basis of the Ca 4227 line. We find that \[Ca/Fe\] = 0 or slightly negative. This is peculiar, since Ca is thought to come primarily from explosive and normal oxygen burning, and is expected to follow Mg in these bright galaxies. Although our models seem to indicate that Ca is underabundant with respect to Fe (a result also confirmed by García-Vargas et al. 1998), this is not certain, as explained in Idiart et al. (1997). Because of the fact that the models of Paper I, like most currently available models, assume that \[Ca/Fe\] and \[Ca/Mg\] = 0 for the input stars that are used to calculate the fitting functions, the calculated integrated Ca T index will be too large if the stars in the input library that are used are overabundant in Ca, and vice-versa. Idiart et al. have determined Ca abundances, as well as Fe and Mg abundances, for a medium-sized input library of about 100 stars, and using individual Ca/Fe ratios for those stars they calculate models with much smaller values for the equivalent widths of the Ca T than Paper I or García-Vargas et al. (1998). Using models with solar abundance ratios we find that \[Ca/Fe\] for our three galaxies is about solar (see Fig. 10). Therefore, in spite of the fact that the Ca is an $`\alpha `$-element, the Ca-dominated features do not track Mg, but the Fe-dominated group. Combining the fits obtained here with those of Paper II obtained at larger radii we can get an idea of the gradients in age, metallicity and IMF slope in these galaxies. To do this we summarize in Table 7 the fits at the 3 different positions. The table shows that the metallicity is the main parameter dominating the observed gradients. Although in the central regions of these galaxies a metallicity of more than 2 times solar is required, the metallicity decreases rapidly outward (being below solar at 15$`\mathrm{}`$), except for NGC 4472, for which the gradient is shallower. In Paper II we attributed this to the fact that this galaxy has a considerably larger effective radius than the other two. The age on the other hand does not seem to vary in an appreciable way with galactocentric radius. Only for NGC 3379 we see a slight age gradient, being somewhat younger in its center. For our understanding of these gradients it would be very useful if good observations could be obtained further away from the center. ## 6 Conclusions We have observed the well-studied galaxies the Sombrero, NGC 3379 and NGC 4594. We show that excellent two-dimensional absorption line strengths and stellar kinematics can be obtained with this instrument. We have extensively compared our results with the current literature. A detailed comparison with 2D multi-lenslet TIGER-spectroscopy of the Sombrero galaxy by Emsellem et al. (1996) shows that we can measure velocities and velocity dispersions to an accuracy of 10 km/s with our current setup, or about 0.1 pixel. We reproduce well the absorption line maps presented in that paper. An in-depth comparison with long-slit spectroscopy of Paper II shows in general good agreement between absorption line strength obtained using IFS and long-slit spectroscopy. A problem in our absorption line maps is that they contain peaks at random position which are not due to photon noise but probably to continuum variations from fiber to fiber, possibly induced by residual fringing or by variations in stress. We show that Mg/Fe is enhanced in the whole inner disk of the Sombrero galaxy. \[Mg/Fe\] there is larger than in the rest of the bulge. The large values of Mg/Fe in the central disk are consistent with the centres of other early-type galaxies, and not with large disks, like the disk of our Galaxy, where \[Mg/Fe\] $``$ 0. We confirm with this observation a recent result of Worthey (1998) that Mg/Fe is determined by the central kinetic energy, or escape velocity, of the stars, only, and not by the formation time scale of the stars. In galaxies with regions with large escape velocities the IMF would have to be skewed more towards high-mass stars, which then will favor SNe type II as compared to SN type Ia, so that the Mg/Fe ratio will be enhanced. We have obtained new observations of the Ca II IR triplet in the three galaxies, and compared them with our model predictions. We find that the observations of the Ca II IR triplet are lower than expected from the models. This result is in agreement with Paper II, where we found that the Ca 4227 line strength was lower than expected. The fact that it has been assumed in the Vazdekis models (and most others) that \[Ca/Fe\] and \[Ca/Mg\] for the stars in the stellar input libraries is solar, makes it difficult for our models to determine the \[Ca/Fe\] ratio in our galaxies. Using however the models of Idiart et al. (1997) who have taken these abundance ratios into account for individual stars, we find that Ca approximately tracks Fe for our three galaxies. We find that H$`\gamma `$ agrees well with predictions based on other lines, including H$`\beta `$. Because of the fact that H$`\beta `$ is often severely contaminated by emission lines we confirm statements by e.g. Kuntschner & Davies (1998) that H$`\gamma `$ is often a very good alternative to H$`\beta `$ when measuring e.g. ages of galaxies. Combining the data presented in this paper with the results of Paper II we find that the radial line strength gradients in the three galaxies are predominantly gradients in metallicity, in agreement with earlier work (e.g. Davies et al. 1993). ## Acknowledgements We are grateful to Adolfo Garc ía for his work developing 2D-FIS. This work has been partially supported by the Dirección General de Investigación Científica y Técnica (PB93-0658). We thank Martin Vogelaar for his programming help, and the referee for useful comments that improved the paper. The 4.2 m William Herschel Telescope is operated by the Royal Greenwich Observatory, at the Spanish Observatorio del Roque de los Muchachos of the Instituto de Astrofísica de Canarias. We thank all the staff at the observatory for their kind support.
no-problem/9907/astro-ph9907259.html	ar5iv	text	# Observations and Implications of the Star Formation History of the LMC 1footnote 11footnote 1Based on observations with the NASA/ESA Hubble Space Telescope, obtained at the Space Telescope Science Institute, operated by AURA Inc under contract to NASA ## 1 INTRODUCTION Recent improvements in data quality and analysis tools have opened up the possibility of deriving detailed star formation histories for Local Group galaxies based on observed colors and brightnesses of individual stars. Results have been somewhat surprising, indicating a wide diversity of star formation histories among galaxies of the Local Group, even for galaxies of a given morphological type (e.g., the dwarf spheroidals); some recent reviews have been presented by Mateo (1998), Grebel (1998), and Da Costa (1998). The Large Magellanic Cloud occupies a special role in these studies. As our nearest neighbor (apart from the Sagittarius dwarf), it allows observations of faint stars which include essentially unevolved stars that are fainter than the turnoff corresponding to the age of the universe. Such stars contain information about the initial mass function and also about the earliest stages of the star formation history of a galaxy. In addition, we can obtain accurate photometry of stars down to the oldest main sequence turnoff. It is critical that information derived from these stars agrees with star formation histories derived from brighter, more evolved stars if we are to believe the results on star formation histories derived for more distant galaxies, where only the brighter stars are observable. This is especially true given that uncertainties in our understanding of stellar evolution are generally larger for stars in their later stages of evolution, when they are brighter. In addition, the LMC provides a unique opportunity to compare the star formation history of its field population with that of its star clusters, since the LMC has a rich population of the latter. This has implications for understanding whether clusters form in a different mode of star formation than field stars, and is important to understand the degree to which one can trace the global star formation history of a galaxy from its constituent star clusters. Several studies have suggested star formation histories for the field population in the LMC. Early studies by Butcher (1977) and Stryker (1984) suggested that the LMC might be composed primarily of younger stars. Using ground-based observations which did not quite reach to the oldest main sequence turnoff, Bertelli et al. (1992) and Vallenari et al. (1996a,b) suggested that the LMC field population was composed primarily of younger stars with ages less than a typical burst age of 4 Gyr, with some indication that the burst age varied across the galaxy. Deeper observations with the HST have not confirmed this picture, instead suggesting that star formation has been more continuous over the lifetime of the LMC, although almost certainly with an increase in star formation rate in the past several Gyr (Holtzman et al. 1997, Geha et al. 1998). All of these studies were for regions outside the LMC bar. Inside the bar, Olsen (1999) suggests that the star formation rate has also been more continuous, possibly extending back for a longer period at a roughly constant rate than the outer fields; this differs from the conclusions of Elson et al (1997), who suggest that the bar formed relatively recently and has an age of $``$ 1 Gyr. In this paper, we present derivation of the star formation history for a field in the LMC bar observed with the HST/WFPC2, as well as for several previously published outer fields, using a more detailed analysis of the color-magnitude diagram. Along with this, we discuss some of the problems and limitations of the techniques (including ours) which are being used to extract star formation histories. We attempt to present a summary of some of the main implications of recent results on the star formation history of the LMC: the relation between the field and cluster star formation history, differences between the outer regions and the bar, the relationship of the star formation history to the dynamical history of the LMC, and the relation between the star formation history derived from studies of color-magnitude diagrams with that derived from chemical evolution studies. ## 2 OBSERVATIONS All of the data discussed in this paper were obtained with the Wide Field/Planetary Camera 2 (WFPC2) on the Hubble Space Telescope (HST) as part of Guaranteed Time Observations granted to the WFPC2 Investigation Definition Team. Three separate fields were observed, with two located several degrees from the LMC center, and one located within the LMC bar; details are given in Table 1 (for a map of the locations of the outer fields, see Geha et al. 1998) In all fields, observations were made through the F555W and F814W filters. Standard reduction procedures were applied to all of the frames, as discussed by Holtzman et al (1995a). The two outer fields are relatively uncrowded. In contrast, the bar field is fairly crowded, so stars cannot be seen as faint as in the outer field. This problem is exacerbated by the fact the the PSF in the bar exposures is significantly broader than in the other fields. This presumably occurred because of a large focus excursion (the so-called “breathing” of the HST secondary); to our knowledge, these frames represent one of the largest examples of this behavior. These exposures serve as a distinct warning to those who assume that the HST PSF is temporally stable. ## 3 ANALYSIS ### 3.1 Photometry Photometry in each of the fields was done using profile-fitting photometry as described by Holtzman et al. (1997). To summarize, we performed the photometry simultaneously on the entire stack of frames taken in each field, solving for the brightnesses in the two colors, the relative positions of the stars, and the frame-to-frame positional shifts (and scale changes between the filters), using an individual custom model PSF for each frame. The model PSFs used a separate focus for each frame as derived using phase retrieval of a few stars in the frame; the models also incorporate the field dependence of the pupil function as specified by the WFPC2 optical prescription, and the field dependence of aberrations as derived from phase retrieval from some other stellar fields. Instrumental magnitudes were placed on the WFPC2 standard system using the calibration of Holtzman et al. (1995b). No correction was made for the possible effect of charge transfer efficiency problems, since these are expected to be relatively small for the background levels in our frames, especially for the relatively bright stars on which most of our analysis is based. The software for all of the PSF modelling and photometry was implemented in the XVISTA image processing package. Figure 1 shows the derived color-magnitude diagrams for the three fields. We investigated errors in the photometry and its completeness using a series of artificial star tests, in which artificial stars were placed into each image at a range of brightnesses and the photometry was redone. The individual errors for all of the artificial stars (observed vs. input brightness) were tabulated for use in the construction of artificial color-magnitude diagrams, as discussed below. ### 3.2 Derivation of star formation histories Various groups in the past several years have published descriptions of techniques used to infer star formation histories based on the distribution of stars in a color-magnitude diagram, or, more generally, for observations of stars in multiple colors (e.g., Tolstoy and Saha 1996; Dolphin 1997; Hurley-Keller, Mateo, & Nemec 1998; Ng 1998; Olsen 1999; Gallart et al. 1999; Han, in prep.). These are all similar in concept, in that a set of observations is fit using some combination of individual simple stellar populations in an effort to derive the relative importance of different simple stellar populations and thus a star formation history. The techniques differ in detail, using different metrics against which one measures how well a given model matches the observations (e.g., maximum likelihood, minimum $`\chi ^2`$ for different bins in color-magnitude space, etc.), different techniques with which the best fit is sought (e.g., linear least squares, genetic algorithm, trial and error), and different input stellar models. For several years, we have also been doing fits for star formation histories. Our technique bins stars in a color-magnitude diagram, and searches for a best fit by minimizing the $`\chi ^2`$ between the number of observed and model stars in the different bins. The search for the best fit is automated, using nonlinear least squares to solve for the the relative amplitudes of each simple stellar population, and optionally for the distance, reddening, and metallicity; the fit is nonlinear because the problem is formulated to insure that only positive amplitudes for each simple stellar population are allowed. For our input stellar models, we use the isochrones from the Padova group (Bertelli et al. 1994), and recently have also experimented with a preliminary version of the newest isochrones from this group (Girardi, private communication). We find that the isochrones from Girardi provide a significantly better match to the observed giant branches, as they predict hotter temperatures for the giants; around the main sequence, where most of our stars are located, the Girardi isochrones are similar to the original Padova set. In all of our subsequent analysis, we use the Girardi isochrones which were available to us (Z=0.001, 0.004, and 0.008) in conjunction with Padova isochrones for other metallicities (Z=0.0004, 0.02, 0.05). The isochrones are used in conjunction with Kurucz (1993) model atmospheres to derive colors and brightnesses in the WFPC2 photometric system. We allow for arbitrary ages and metallicities by interpolating within the isochrones using a set of equivalent evolutionary points to preserve the correct isochrone shape. Our basis simple stellar populations are either discrete bursts or are constructed assuming continuous star formation within specified epochs. Typically, we use epochs with equal widths in the logarithm of the age, to account for the fact that isochrones at a fixed age difference become more similar as a population ages. For the current work, we have assumed a Salpeter initial mass function ($`dN/dMM^\alpha `$, with $`\alpha =2.35`$) for most of our models, although we have tried some other IMFs as well, as discussed briefly below. Uncertainty about the IMF is probably responsible for one of the largest sources of potential errors in our results. Given model predictions for the number of stars as a function of color and magnitude for any given star formation history, along with a distance and extinction, we account for observational errors and incompleteness by smearing the model results with the errors derived from our artificial star tests. We use the exact tabulations of measured errors for our artificial stars to do this smearing, and thus make no assumption that the errors are distributed normally (which they are often not, in particular, because of errors from crowding which are correlated for each color). The fraction of detected artificial stars is included in the smearing, so incompleteness is automatically taken into account. Previous usages of this software (Holtzman et al. 1997, Geha et al. 1998) have used it in a mode where the bins are very wide in color, effectively making this perform a fit to the luminosity function. Fits to the luminosity function are less sensitive to possible errors in reddening, photometric calibration, and color errors in the stellar models. Of course, throwing away color information lowers the ability to discriminate among different models. Fits to the full color-magnitude diagram using narrower color bins, along with a discussion of possible problems with interpreting these, are presented next. ## 4 DISCUSSION ### 4.1 Derivation of star formation histories from model fits We derived two separate star formation histories, one for the “outer” fields and one for the bar. For the outer fields, we combined the data from the two observed fields because there is no strong evidence that the star formation history of these fields differs (Geha et al. 1998) and because the relatively smaller number of stars in these fields limits the accuracy of the derived star formation histories. It is possible to attempt to determine the distance modulus and/or the extinction by allowing these to be free parameters in the fit or by comparing the residuals for various choices of these parameters. We do this below in an effort to estimate some of uncertainties in our results. However, we believe that it is important, when possible, to use additional constraints on these quantities rather than simply allowing them to be free parameters. There are many methods for estimate distances and/or extinctions which do not depend on interpretation of a color-magnitude diagram, for example, the use of variable stars or HI column densities. As we will argue below, it seems prudent to use independent information when possible to constrain the star formation histories. For the LMC, however, there is significant debate about the distance and some uncertainty about the extinction. For our initial fits, we decided to fix the distance to the LMC at a (extinction-free) distance modulus of 18.5 and initially chose an extinction of E(B-V)=0.10 based on the maps of Schwering and Israel (1991) which suggest a foreground reddening of E(B-V)=0.07; we included an additional 0.03 mag as an estimate for internal reddening. However, with these choices we found that the model main sequences were too red, which, as discussed below, affected the derived star formation histories in systematic ways. The only way we could match the color of the main sequence was to adopt lower extinctions; we adopted E(B-V)=0.04 for the outer fields and E(B-V)=0.07 for the bar field. Interestingly, ground-based studies of the LMC have been yielding relatively low reddenings for fields away from young associations (Zaritsky, private communication), although we do not have any direct estimates from these studies for our fields at this time. A possible alternative to errors in our assumed reddening is errors in our photometric zeropoints; in fact, a small color change of $``$ 0.02 mag in F555W-F814W relative to our adopted calibration is suggested by Stetson (1998). In any case, we find that our derived results are relatively insensitive to the source of the problem, and have chosen to use the empirical lower reddenings. In a subsequent section, we will present star formation histories derived from a variety of choices of distance and extinction, in an effort to understand the uncertainties in our results. Here, we note that our adopted choices for these quantities actually are those for which the best-fitting star formation histories give relatively low $`\chi ^2`$ values; there is some variation in which distance and extinction give the absolutely lowest $`\chi ^2`$ depending the field being fit and assumptions about the age-metallicity relation. For our initial fits, we constrained the basis stellar populations to have the age-metallicity relation presented by Pagel and Tautvaisiene (1998), who derived it from a chemical evolution model designed to to fit a variety of observations of LMC clusters and field stars. For our initial models, we assumed that the age-metallicity relation is dispersion-free; the cluster data (e.g., Olszewski 1993), however, suggest that there may be a range of metallicities at any given age. To obtain isochrones at the appropriate ages and metallicities, we interpolated among the Padova isochrones. We used age bins with a width of 0.1 in $`\mathrm{log}(age)`$, assuming constant star formation within each age bin. We binned the data in the color-magnitude diagrams with bin widths of 0.04 mag in color and 0.06 mag in brightness; these were compromise values given the observed number of stars. Figures 2 and 3 show a summary of the derived star formation history information for the two fields with the constrained age-metallicity relation. The upper left plot shows the derived star formation rate as a function of time (abscissa is linear in lookback time), the lower left shows the cumulative number of stars formed as a function of log(lookback time), the upper right plot shows the differential and cumulative metallicity distribution functions, and the lower right shows the observed and model F555W luminosity functions (corrected for reddening). The 3D diagram at the lower left shows the “population box” (Hodge 1989) which gives the star formation rate as a function of log(lookback time) and metallicity; older ages are at the left side of the plot, and lower metallicities are at the back. The greyscale at the lower right shows the difference between the model and the observed Hess diagrams divided by the square root of the data. Thus, the greyscale diagram gives the deviation between the data and the model in units of the deviation expected from counting statistics. The greyscale is in the sense that bright areas are regions where the observed number of stars is larger than the model; the full range from black to white is $`3\sigma `$ to $`+3\sigma `$. The quality of the fits is estimated by a reduced $`\chi ^2`$ statistic which is shown in the greyscale diagram, and also by the probability (given in the luminosity function panel) that the observed and the model luminosity functions are drawn from the same population, as inferred from a Kolmogorov-Smirnov test. These model fits suggest that star formation has been ongoing in the LMC over its entire history, with fluctuations of a factor of a few in star formation rate and a higher star formation rate in the past few Gyr. In the outer fields, there is evidence for an increasing star formation rate over the past few Gyr, whereas in the bar, the star formation rate seems to have been more constant recently. More generally, these fits suggest, as did previous studies (Holtzman et al. 1997, Geha et al. 1998, Olsen 1999), that a significant fraction ($``$ 50%) of the stars in the LMC are relatively old, i.e. older than $``$ 4 Gyr. However, if one inspects the residual greyscale Hess diagrams, one one can clearly see some systematic problems with the model fits. In particular, the model main sequences are much narrower than the observed main sequences. A similar effect, although at a reduced level, can be seen in the residuals shown in Olsen (1999); the apparently smaller effect in those data is plausibly explained by the fact that their exposure times were shorter (by a factor of $``$ 4), leading to larger photometric errors and hence broader observed and model sequences. There are several possible explanations: 1. The LMC has stars with a range of metallicities at any given age. 2. A significant number of stars in the LMC are unresolved binary stars. 3. There is a spread in distances and/or extinctions. This is unlikely given the inclination of the LMC and the relatively low total extinction towards our fields. 4. Our observational errors have been significantly underestimated; we believe this is unlikely given our careful analysis of photometric errors. We discuss the first two possibilities in the next sections. #### 4.1.1 Metallicity dispersion We feel that the mostly likely source of the broad main sequence is that the LMC has stars with a range of metallicities at any given age. Certainly, the Milky Way has a very significant dispersion in its age-metallicity relation. To test whether a spread in metallicities can account for the observed broad main sequence, we performed the fits allowing for multiple combinations of age and metallicity. We used the same age bins as before, but at each age, allowed populations with discrete metallicities of Z=0.0004, 0.001, 0.004, and 0.008 , 0.02, and 0.05. The choice of these six metallicities was motivated by the fact that these were available without any interpolation in metallicity. We allowed for all combinations of age with these metallicities. Figures 4 and 5 show the results using these models. Similar star formation histories are derived, but the resulting residuals show substantially smaller systematic deviations. Remarkably, the model fits qualitatively recover the mean age-metallicity relation observed in LMC clusters, despite the fact that no assumptions were made about this relation at all. This is demonstrated by comparing the “population boxes” for these models with those using the constrained age-metallicity relation. Although the derived relation is not especially quantitative since we only included six discrete metallicities, Figure 6 shows a representation of the derived age-metallicity relation for the bar field; squares give the metallicity of the population with the highest star formation rate at each age bin, while crosses give the mean star-formation weighted metallicity from the six metallicity bins. The solid line shows the relationship from Pagel and Tautvaisiene (1998), which does a reasonable job of matching observations of LMC clusters (e.g., Olszewski 1993). Our derived relation is similar to the model which was designed to fit the cluster observations; the width of our relation at fixed age is also qualitatively similar to that seen in the cluster distribution. While the fits are significantly better using a range of metallicities at a given age, the derived star formation histories are qualitatively similar to those derived using the constrained dispersion-free age-metallicity relation. As with the constrained age-metallicity fits, we find evidence that the bar field contains a larger relative number of older stars than the outer fields. We find it encouraging that the results on the star formation appear to be reasonably robust against assumptions made about the metallicity distribution. Since our models allow for multiple combinations of age and metallicity, the results actually make a crude prediction for the metallicity distribution of LMC stars; the prediction is crude since we are only using six discrete metallicities in our models rather than a continuous distribution. Figures 4 and 5 suggest that the LMC has a relatively broad metallicity distribution. Low metallicity stars (\[Fe/H\] $`\stackrel{<}{}`$ -1) comprise 15 and 30% of the stars in the outer and bar fields, respectively. However, one needs to beware of directly comparing these numbers to observations of giant star metallicities. Our models give the relative numbers of stars of all stellar masses at different metallicities, while giants sample only a small range of stellar masses. Because older, more metal-poor, populations feed stars to the giant branch slower than young populations, the relative number of lower metallicity giants will be lower than the true relative fraction of lower metallicity stars which is predicted by the models. This effect can be substantial; for the bar model, we estimate that that low metallicity giants will be undersampled by nearly a factor of two in a pure giant sample as compared with the true metallicity distribution. Thus relatively few metal-poor giants are predicted by these models. Our models do have a reasonable component of relatively metal-rich (solar or greater) stars, which are included to fit the reddest sections of the main sequence. It is possible that the contribution of these stars is overestimated by our models because of some contribution from unresolved binaries, as discussed next. #### 4.1.2 Unresolved binaries Unresolved binary stars can significantly affect the distribution of objects in a color-magnitude diagram. Although a variety of evidence suggests that a significant fraction of all stars are in binary systems, it is less clear whether the masses of stars in such systems are correlated or are drawn from the same initial mass function. If stellar masses of binaries are uncorrelated and the mass function rises towards lower masses, then the effect of unresolved binaries is mainly significant only for rather low mass stars at the bottom of the main sequence; for more massive stars, a binary companion is much more likely to be significantly fainter and thus have little influence on the total system luminosity and color. As a result, the only way to get a broadening of the main sequence for stars similar to those observed by us in the LMC ($``$ 1$`M_{}`$) is to require that the masses of the components of binary systems are correlated. Such a scenario has been suggested by Gallart et al. (1999) to explain the color-magnitude diagram of the Leo I dwarf spheroidal; they find significantly better fits using a large fraction of binary stars which are constrained to have mass ratios greater than 0.6. However, we find that models using such a scenario still do not accurately reproduce our observed color-magnitude diagram using a dispersion-free age-metallicity relation. The problem arises because the width of the observed sequence is broadest compared to the models at intermediate luminosities. If there is any range of mass ratios in binaries at all, the effect of binaries grows with decreasing luminosity. Thus, any model which matches the width of the main sequence at intermediate luminosities using a binary component predicts too broad of a sequence at lower luminosities. To further check the binary star hypotheses, we performed fits with multiple combinations of age and metallicity, but using an assumed binary fraction of 0.5; binary masses were drawn from the same initial mass function as the parent population but with binary mass ratios constrained to be larger than 0.5. These fits are significantly worse than fits without binaries. However, as mentioned above, if all mass ratios were allowed, then the models would allow for a significant component of binaries since their effect is small for the stars in our observations. Although we find that metallicity spread is a more likely explanation than unresolved binaries for the observed width of the main sequence, it is likely that both effects play some role. The existence of some unresolved binaries would probably reduce some of the spread in our derived age-metallicity relation; in particular, we expect it would reduce the contribution of stars at the highest metallicities at any given age. ### 4.2 Accuracy of derived star formation histories from model fits Before one reads too much into these derived star formation histories, however, one should consider some of the limitations and problems of fitting star formation histories. There are numerous assumptions that are made in the models: * The stellar models accurately predict the observed properties of stars as a function of age and metallicity, * A unique initial mass function exists which is independent of age and metallicity, and is represented (in this case) by a power law with $`dN/dMM^{2.35}`$, * All stars are found at a common distance and extinction, * The observational data are calibrated to the same system as the models, and the observational errors can be accurately measured, * The basis stellar populations used in the fits represent all populations present in the galaxy. All of these assumptions are likely to be in error at some level. Consequently, the question is the degree to which deviations from the assumptions affect the derived star formation histories. Unfortunately, this is very difficult to assess given the unknown nature of the possible errors in the assumptions. As a result of these problems, it is likely that no solution will actually matches the observed distribution of stars within the errors expected from Poisson statistics alone. This is certainly the case for the best models here; given the number of independent regions in the color-magnitude diagrams being fit, one would expect a reduced $`\chi ^2`$ much closer to unity than the values we obtain. Sometimes the fits produce model luminosity functions which are consistent with the data, but other times they are formally ruled out with a KS test. Review of the various papers which derive star formation histories for various systems suggest the same quality of matches is obtained for most other derived star formation histories. Given the known problems with the assumptions, this is usually not considered to be a major problem; instead, one makes the assumption that the “best-fitting” model represents the closest approximation to the truth, even if it is statistically inconsistent with the data. We make the same assumption, but feel the need to explicitly state it; one could certainly imagine situations in which this assumption might not be true. Because one is just choosing the best-fitting model averaged over the entire color-magnitude diagram, our method inherently weights areas where there are more stars and where the photometric errors are low. As a result the model does not give extra weight to regions which carry more unique information about stellar ages. For example, if upper main sequence stars exist, there must be a young population, but if these stars are vastly outnumbered by older stars, the model will do its best to fit the older stars even if it means sacrificing a good match to the younger stars. One could certainly devise a scheme where certain regions of the color-magnitude diagram carry extra weight, and perhaps this is the direction we should take in the future. In addition, different assumptions can lead to systematic errors in the derived star formation histories. For example, we found that changing the assumed reddening led to systematic differences in the derived star formation history. At a higher reddening, more metal-poor stars are required to fit the observed data. If the age-metallicity relation is constrained, then this in turn leads to a higher derived number of older stars. Errors in the assumed initial mass function can lead to similar systematic effects. To demonstrate the effect of varying reddenings, distances, and IMF slopes, we ran a set of solutions allowing for a range in reddening from $`0.04<E(BV)<0.10`$, a range of distance from $`18.2<mM<18.7`$, and two different IMF slopes with $`\alpha =2.35`$ and $`\alpha =2.95`$; each of these was varied independently with the other two quantities at our preferred values. For each different parameter, we derived a star formation history along with a $`\chi ^2`$ for the each fit. Figure 7 shows the derived values of $`\chi ^2`$ for different choices of reddening and distance modulus; in each panel, results are shown both for the constrained age-metallicity relation as well as for multiple combinations of age and metallicity. If the age-metallicity relation is constrained, the quality of the fit changes significantly for different choices of reddening and distance modulus; minimum $`\chi ^2`$ are reached around our preferred values of $`mM=18.5`$ and $`E(BV)=0.07`$ and 0.04 for the bar and outer fields. However, if multiple combinations of age and metallicity are allowed, the quality of the fits are relatively insensitive to the choice of reddening and distance modulus, indicating there is some degeneracy in the sensitivity to different parameters. This supports our assertion that it is better to use additional independent observational constraints on parameters relevant to the star formation fit rather than to include these parameters in the fitting process. For many systems, information about distance and reddening are readily available. We suggest that perhaps the greatest improvement in the confidence in our derived star formation histories will come from the observations of observed metallicity distributions against which one could compare the derived star formation histories. The derivation of metallicities is feasible in nearby stellar systems given current multi-object spectroscopic capabilities and/or using multi-band photometry, and, in fact, such studies are underway by several groups (e.g., Olszewski, Suntzeff, & Mateo 1996, Smecker-Hane et al., private communication). However, we reiterate that caution must be used to compare observations with model predictions; one must take into account the metallicity distribution biases which are introduced by the observational selection of stars used for metallicity determinations. Given the limited external constraints we have about the distance modulus to the LMC, the extinction, the IMF, and the metallicity distributions, we must consider how possible uncertainties in these affect our derived star formation histories. Figure 8 shows derived star formation histories for the ensemble of models comprising $`18.2<(mM)<18.7`$, $`0.04<E(BV)<0.10`$, and IMF slopes of -2.35 and -2.95; the bold line shows the results from previous figures for our preferred quantities. One can see that the star formation history is qualitatively similar independent of the parameters, but quantitatively, the star formation rate at any given time can be in error by a factor of a few. The largest qualitative difference comes for different choices of IMF slope; as the IMF becomes steeper (dashed line in Figure 8), the observations require a larger relative number of younger stars, exactly as expected (Holtzman et al. 1997). To the extent to which our parameter choices span the full range of values expected for the LMC, Figure 8 can be used to give a reasonable estimate of the uncertainties in our results, although these results do not consider the possible effect of errors in the stellar models. In addition to the systematic errors, there are random errors in our results because of the limited number of stars observed. These random errors are larger for the outer fields than for the bar field because they have fewer observed stars. However, simulations of color-magnitude diagrams suggest that the magnitude of the random errors, even for the more sparse outer fields, are smaller than those which arise because of potential systematic errors. ### 4.3 Differential comparison of color-magnitude diagrams To avoid potential problems with fitting star formation histories, it is possible to derive differences in the star formation history from one field to another by a direct comparison of the observed color-magnitude diagrams. Such differences lead to systematic residuals when comparing the Hess diagrams of different fields. Differential comparisons of fields with similar metallicities are relatively straightforward to interpret in terms of age differences, although a quantitative association of a difference with an age requires the use of stellar models. Differential comparisons may be more problematic for fields with significantly different metallicities. As an application, we consider the differences in star formation history between the LMC outer fields and the bar field. Our model-dependent derived star formation history suggested that there has been a greater relative contribution of the youngest stars in the outer fields than in the bar, in agreement with the results derived by Olsen (1999) based on similar fits, but in qualitative disagreement with the results of Elson et al. (1997) which were based on a visual inspection of the color magnitude diagram in a bar field. Figure 9 shows the difference in the Hess diagrams between the two fields, where white areas represent locations where there are more outer field stars, and darker areas regions where there are more bar stars. The Hess diagrams were normalized to have the same total number of stars between $`M_V44.5`$ where the photometry in both fields is reasonably complete and where stellar evolution effects are minimal; the difference between the Hess diagrams was smoothed to suppress the noise from counting statistics. One can clearly see a darker band in the lower parts of the residual Hess diagram which suggests that the bar contains a relatively larger number of intermediate age stars than the outer field; the difference is made up by a relatively larger number of upper main sequence (younger) stars in the outer fields. The bar field also has a relatively larger number of red clump stars which represent stars of intermediate age. Consequently, the differential, model-independent comparison supports the results derived by fitting stellar models, namely that the bar, although it contains a significant population of young stars, is relatively older than the outer fields. Although the bar field shows an apparently significant sequence which one might associate with a several Gyr old burst (as seen in Figure 1 and shown in cross-section plots in Elson et al. 1997), which shows up in contrast to the outer field color-magnitude diagrams, such a feature turns out to be a generic feature of models even with a continuous star formation rate. This arises because upper main sequence stars (those with convective cores) evolve to cooler temperatures and higher luminosities over most of their main-sequence lifetimes, but then retreat to higher temperatures, creating a jag in the evolutionary path in a color-magnitude diagram. Since the star spends proportionally more time at the coolest effective temperature, a secondary sequence which is offset from the main sequence exists for a continuous star formation history. Although this is true even for a population of fixed metallicity, an age-metallicity relation makes the secondary sequence even more pronounced. The effect is demonstrated in Figure 10, where we show a synthetic Hess diagram of a population with a constant star formation rate over the past 12 Gyr, using our adopted age-metallicity relation for the LMC. One sees a clear sequence which might be confused with an increase in the star formation rate at some time in the past, despite the fact that it is a color-magnitude diagram for a constant star formation rate. This clearly shows the peril of interpreting color-magnitude diagrams purely visually; an apparent concentration of points does not necessarily imply a burst or even a significant enhancement in the star formation rate. This observation leads to an understanding of the difference between the interpretation of Elson et al. (1997) and the results of this paper and Olsen (1999) regarding the relative age of the LMC bar. Elson et al. (1997) suggest that the LMC bar is younger than the rest of the LMC because they observe a bimodal distribution of color in the upper main sequence of their LMC bar field. They associate the blue peak with the formation of the LMC bar ($``$ 1 Gyr ago) and the red peak with the formation of the bulk of LMC field stars ($``$ 4 Gyr ago). Instead, we find that the red peak is a generic prediction of the models even for roughly constant star formation rate, and the blue peak represents a recent increase in the star formation rate which is seen in both the outer fields and the bar (in fact, it is stronger in the outer fields than in the bar). ### 4.4 Field vs. cluster star formation history Perhaps the most notable conclusion which can be drawn from our derived star formation histories is that the field star formation history in both the bar and the outer fields appears to differ from the star formation as suggested by the age distribution of LMC clusters. LMC clusters show a significant age gap between lookback times of 4 and 12 Gyr (e.g., van den Bergh 1991, Girardi et al. 1995), with 14 old clusters (which have ages comparable to those of the Galactic globulars, see Olsen et al. 1998), numerous young clusters, and only one cluster, ESO 121-SC03, at an intermediate age. In contrast, the derived star formation histories of Figures 2-5 suggest that star formation has been more continuous in the field of the LMC. Here we consider the robustness of that conclusion. Geha et al. (1998) showed that the observed luminosity function in the outer fields was strongly inconsistent with a star formation history which corresponds to the current number distribution of clusters as a function of lookback time. However, this comparison is perhaps unfair, as the older clusters are generally more massive than the younger ones, so weighting by mass would allow for a larger older component. In addition, one might consider that some fraction of clusters which were formed at an early epoch might be disrupted during the subsequent evolution of the LMC; although many of the young clusters are massive and appear tightly bound and unlikely to disrupt anytime soon, many others have lower masses and larger sizes and might plausibly disrupt. As a result, we consider the more general question of whether any star formation history with a gap in star formation between 4 and 10 Gyr is capable of reproducing the observed properties of the LMC field stars. To address this, we performed fits for the star formation history again without allowing for any component stellar populations with ages between 4 and 10 Gyr. Figures 11 and 12 show the results for the bar field for the constrained age-metallicity relation and arbitrary combinations of age and metallicity. The fit with the constrained age-metallicity relation is notably worse than allowing for intermediate age stars. This is easily explained; the existence of a broad band of subgiants around the oldest turnoff suggests that multiple ages are present. However, it is possible to get such a continuous band with different combinations of age and metallicity, since older, lower metallicity stars can blend smoothly into younger, higher metallicity stars without necessarily leaving a gap in the color-magnitude diagram. This is confirmed by Figure 12, which shows that a moderately good match to the observed Hess diagram can be made even with an age gap in the star formation history. However, one can see that the model produces too many subgiants at $`M(F555W)2.5`$; this can be seen in both the residual Hess diagrams as well as in the luminosity function. The $`\chi ^2`$ for the star formation history is only slightly worse with an age gap than without it, but the probability that the luminosity function is consistent with that of the data can be ruled out at a much higher confidence level than for models without a gap. In addition, the existence of a gap would require a relatively larger population of older, metal-poor stars; with a gap, we find that approximately 40% of all of the LMC field stars would have to be older than 10 Gyr and more metal-poor than \[Fe/H\] $``$ -1. This may not be consistent with observations of metallicity distributions (e.g. Olszewski 1993) and the lack of a strong horizontal branch in the color-magnitude diagrams; however, the possibility exists that the LMC has an extended, low density, older stellar halo which becomes more dominant over the young and intermediate age population as one moves farther from the center of the LMC. Consequently, we feel that it is unlikely that the field star formation history, as sampled by the location of our fields, has a gap in star formation between 4 and 10 Gyr ago. ### 4.5 Burstiness of star formation Another outstanding question is the degree to which star formation is “bursty” in the LMC. The degree to which we can distinguish between bursty and continuous star formation depends on several factors. For older populations, the distribution of stars in the color-magnitude diagrams changes very slowly with age, so it is difficult to get much age resolution. For younger populations, the separation between ages is larger, but the observed number of stars is smaller, so sensitivity to different distributions of star formation is limited by counting statistics. Our fits for star formation history have been performed assuming constant star formation within epochs spaced by 0.1 in $`\mathrm{log}(age)`$. This value for the width of each epoch was determined by finding the narrowest age bin which gave statistically distinguishable fits, as measured by $`\chi ^2`$. To measure the sensitivity of the technique to burstiness in the star formation rate, we performed the star formation fits in which single age bins were given a duration of $`\mathrm{\Delta }(\mathrm{log}(age))=0.01`$, while preserving the 0.1 spacing in $`\mathrm{log}(age)`$. We did this for each age bin in turn for lookback times from 1 to 4 Gyr. We found that we obtained nearly identical quality fits using the 0.01 width epochs as we did with the 0.1 width epochs, although the fits would have been worse if we had required more than one bin to be ”bursty” at a time. The basic reason we could not discriminate the duration of a star formation epoch is small number statistics in the number of stars observed on the upper main sequence; without the counting statistics, the models are straightforward to distinguish. This was true even for the bar field, which is the densest field one could observe in the LMC. We estimate that increasing the number of stars by a factor of 5-10 would allow burstiness to be distinguished, suggesting that a program with multiple pointings with WFPC2 (e.g., Smecker-Hane et al., in progress) and/or the Advanced Camera would be useful. At ages older than 4 Gyr, burstiness becomes extremely difficult to measure without exquisitely accurate photometry. ### 4.6 Star formation and the interaction history of the LMC It has been suggested that star formation in the LMC is triggered by tidal interactions with the Milky Way and the SMC. As a result, it is of interest to see whether there is any evidence for an enhanced star formation rate around the time of the last closest passage. Since the full orbits of the Magellanic Clouds are still unknown, the time of last close passage is somewhat uncertain, but the latest models place it around 2.5 Gyr ago (Zhao, private communication). Inspection of our derived star formation histories (for example, Figure 8) show a general tendency for the star formation to increase by a mild amount around this time, but no dramatic effect is seen. However, as discussed in the last section, we cannot constrain the burstiness of the star formation rate very accurately from the current data, so some correlation of star formation with orbit is not necessarily ruled out. One might expect that triggered star formation would not occur in all regions of the galaxy at the same time. If star formation were triggered in different regions at different times, subsequent mixing arising from the stellar velocity dispersion and differential rotation would smooth the bursty nature of the triggered star formation on a timescale given by the mixing. Given an approximate velocity dispersion of 50 km/s, it would take only $``$ 500 Myr for stars at a radius of 4 kpc to mix azimuthally. ### 4.7 Chemical evolution and star formation history of the LMC Pagel and Tautvaisiene (1998) have recently published a model for the chemical evolution of the LMC and compared it with previous models. Such models attempt to match the observed abundance distributions of different elements as a function of metallicity. In general, these models allow to star formation rate to be a free parameter. Models differ in the adopted yields, IMF, the presence of inflow and/or outflow, and the degree to which outflow is selectively enhanced in heavy elements. The best fitting model of Pagel & Tautvaisiene for the LMC favors a star formation history which, although they call it a “bursty” model, has an underlying constant star formation rate over the history of the LMC, with an enhancement in the star formation rate 3 Gyr ago. However, they also have a model with a smoothly varying star formation rate which also provides a reasonable match with observational data. We suspect that using a star formation history derived from our color-magnitude diagrams would be able to provide a reasonable match to the abundance data as well, as it is intermediate between the two models presented in Pagel and Tautvaisiene. We suggest that the next logical step in modeling star formation histories is to couple the derived star formation rates with chemical evolution models and simultaneously attempt to match both color-magnitude diagram data and abundance data. In principle, this might allow one to more uniquely determine the importance of mass inflow/outflow. Unfortunately, such attempts will be complicated if the star formation history is a strong function of location in the galaxy. However, from the few fields considered to date, it appears that the variation with position may not be so large as to make such an attempt futile; once data on more fields, particularly at large radii become available, we suspect a simple model with a small number of radial zones might be sufficiently accurate. ## 5 CONCLUSIONS We have derived star formation histories from the distribution of stars in deep color-magnitude diagrams obtained using HST. These data suggest that there is a significant component of stars older than 4 Gyr in both outer fields and the bar of the LMC. Models in which there is a dispersion-free age-metallicity relation cannot reproduce the width of the main sequence in our high accuracy photometric data. As a result, we have fit models which allow for multiple combinations of age and metallicity and find we can obtain accurate matches to the observed data. These fits allow us to fully construct “population boxes” from our data which are derived solely from color-magnitude diagrams. Such diagrams qualitatively reproduce the mean age-metallicity relation observed in LMC clusters as well as the spread around this relation. These derived models produce crude predictions for the abundance distribution the LMC; new observations which provide such distributions will be extremely useful in constraining the star formation histories and confirming the validity of the models. Both the model fits as well as a differential comparison between the observed color-magnitude diagrams suggest that the bar of the LMC contains a relatively larger number of older stars than the outer fields. This is consistent with the conclusions of Olsen (1999) but different from those of Elson et al. (1998), although we have presented a plausible explanation for why the latter study reached their conclusions. One main implication of the derived star formation histories is that the field star formation history appears to differ from that suggested by the LMC clusters, in that there does not appear to be an age gap in the field star age distribution. However, we note that it is actually rather difficult to constrain the star formation history for lookback times greater than 4 Gyr given the age-metallicity degeneracy in the location of isochrones; observations of larger samples of subgiants, ideally with metallicity determinations, would be desirable to confirm that field stars fill the cluster age gap. We find that it is quite difficult to constrain the degree to which star formation is bursty in the LMC on time scales less than about 25% in age with the observed number of stars even in the WFPC2 bar field. Larger samples will be required to address this issue. However, sequential star formation across the LMC followed by mixing may erase the signatures of bursty star formation even if it occurs. Future progress will be made with larger samples of stars; with accurate photometry down to the oldest main sequence turnoff, one can further constrain burstiness and the star formation history. In addition, we suggest that metallicity determinations for a large number of stars will be crucial in constraining and testing derivations of star formation histories. Coupled with chemical evolution models, we may be able to get constraints on the importance of inflow/outflow in the LMC, and begin to fully understand the nature of the star formation history in one of our nearest neighbors. This work was supported in part by NASA under contract NAS7-918 to JPL. \| Table 1. \| \| \| \| \| \| --- \| --- \| --- \| --- \| --- \| \| Summary of Observations \| \| \| \| \| \| Field \| $`\alpha _{2000}`$ \| $`\delta _{2000}`$ \| Exposure time \| STScI filenames \| \| LMC-OUTER \| $`5^h14^m44^s`$ \| $`65^{}17^{}43^{\prime \prime }`$ \| 4000s \| u2c5010\[1-8\]t \| \| HODGE-10 \| 5 58 21 \| -68 21 19 \| 2500s \| u2o9020\[1-6\]t \| \| HS-275 (BAR) \| 5 24 21 \| -69 46 27 \| 3700s \| u2o9010\[1-8\]t \|
no-problem/9907/astro-ph9907422.html	ar5iv	text	# A DETAILED ANALYSIS OF THE NEUTRON TO PROTON RATIO AT THE ONSET OF PRIMORDIAL NUCLEOSYNTHESIS 11footnote 1Talk given by S. Esposito at Second Meeting on New Worlds in Astroparticle Physics, September 1998, Faro, Portugal. ## 1 Introduction The more recent data on $`{}_{}{}^{4}He`$ mass fraction and D abundance produced during Big Bang Nucleosynthesis (BBN) are still controversial, since there are two different sets of results mutually incompatible $`Y_p`$ $`=`$ $`0.234\pm 0.0054,D/H=(1.9\pm 0.4)10^4;`$ $`Y_p`$ $`=`$ $`0.243\pm 0.003,D/H=(3.40\pm 0.25)10^5;`$ (1.1) It is possible that forthcoming measurements from high-redshift, low-metallicity QSO, or better understanding of the present data will clarify the situation. The fact which however is emerging from the above results is that Helium data are now reaching a precision of one per mille, requiring a similar effort in reducing the theoretical uncertainties. The $`{}_{}{}^{4}He`$ mass fraction $`Y_p`$ crucially depends on the ratio $`n/p`$ at the weak interaction freeze–out. To reach a precision of the order of $`10^4`$ it is demanding to analyze with an accuracy of one percent the rates of processes converting $`np`$, i.e. $`\nu _ene^{}p`$, $`\overline{\nu }_e\text{ }pe^+n`$ and $`ne^{}\overline{\nu }_e\text{ }p`$. This means that the simple Born approximation rate $`\omega _B`$, i.e. the tree level $`VA`$ interaction in the infinite mass limit for nucleons, should be corrected via the inclusion of several radiative and non radiative effects. We will report on the results of a recent evaluation of all the above corrections . For an extensive discussion, as well as a detailed analysis of the calculation, see reference 2. ## 2 Corrections to Born rates and $`{}_{}{}^{4}He`$ mass fraction Leading corrections to the weak rates of neutron decay and electron and neutrino capture processes can be classified as follows: * order $`\alpha `$ radiative corrections $`\mathrm{\Delta }\omega _R`$. These effects have been extensively studied in literature and can be classified in outer factors, involving the nucleon as a whole, and inner ones, which instead depend on the details of nucleon internal structure. Other small effects are actually expected at higher order in $`\alpha `$, since, for example, the estimated value of the neutron lifetime is compatible with the experimental value $`\tau _n^{exp}`$ at 4-$`\sigma `$ level only. These additional contributions are usually taken into account by eliminating the coupling in front of the reaction rates in favour of $`\tau _n^{exp}`$. In Figure 1 we report the result of our calculation for $`\mathrm{\Delta }\omega _R`$ as a function of the ratio $`T/m_e`$. * All Born amplitudes should be also corrected for nucleon finite mass effects $`\mathrm{\Delta }\omega _K`$. They affect the allowed phase space as well as the weak amplitudes, which should now include the contribution of nucleon weak magnetism. Initial nucleons with finite mass will also have a thermal distribution in the comoving frame, over which it is necessary to average the transition probabilities. All these effects are of the order of $`T/M_N`$ or $`m_e/M_N`$, with $`M_N`$ the nucleon mass. The percent corrections $`\mathrm{\Delta }\omega _K/\omega _B`$ are reported in Figure 2. * Since all reactions take place in a thermal bath of electron, positron, neutrinos, antineutrinos and photons, finite density and temperature radiative effects $`\mathrm{\Delta }\omega _T`$ should be also included; these corrections, of the order of $`\alpha (T/m_e)`$, account for the electromagnetic interactions of the in/out particles, involved in the microscopic weak processes $`np`$, with the surrounding medium. They can be evaluated in the real time formalism for finite temperature field theory . As can be seen in Figure 3, these corrections only represents $`0.2÷0.3\%`$ of the corresponding Born rates at the freeze out temperature $`T1MeV`$. The corrections to the $`np`$ rates may produce a sensible effect on the $`{}_{}{}^{4}He`$ mass fraction $`Y_p`$, which is strongly dependent on the neutron fraction $`X_n=n_n/(n_n+n_p)`$ at the nucleon freeze-out. The expected variation of the surviving neutron fraction $`\delta X_n`$ induced by the whole effects $`\mathrm{\Delta }\omega `$, which we report in Figure 4, can be evaluated by solving the corresponding transport equation. We find for the asymptotic abundance $`\delta X_n0.0024`$, with a relative change, in percent, $`\delta X_n/X_n=1.6\%`$. These results allows for a simple estimation of the corrections to $`{}_{}{}^{4}He`$ mass fraction $`\delta Y_p=2\delta X_n\mathrm{exp}\left(t_{ns}/\tau _n^{exp}\right)`$, where $`t_{ns}180sec`$ corresponds to the onset of nucleosynthesis and the exponential factor accounts for the depletion of relic neutrons at freeze-out due to $`\beta `$-decay. Using the results for $`\delta X_n`$ we find $`\delta Y_p0.004`$, or, in percent $`\delta Y_p/Y_p=1.6\%`$. ## References
no-problem/9907/cond-mat9907040.html	ar5iv	text	# Transport of Torsional Stress in DNA ## 1 Introduction and Summary DNA can be regarded as a linear repository of sequence information, or as a chemical compound subject to various modifications (e.g. methylation), and each of these viewpoints is important for understanding some aspects of gene function and regulation. However, many other important processes require an appreciation of DNA as a physical elastic object in a viscous environment. For example, the action-at-a-distance between eukaryotic promoters and their enhancers involves an effective concentration of bound enhancer units depending on both torsional and bend rigidity of DNA. While the equilibrium statistical mechanics of stiff macromolecules such as DNA is a classical topic (see e.g. ), still the nonequilibrium transport properties of such molecules remains incomplete, in part due to the experimental difficulty of probing those properties. In particular, Liu and Wang proposed that the transport of torsional stress (torque) along DNA during transcription could play a role in gene regulation (the “twin supercoiled domain model”) . Transcription causes axial rotation of the transcribed DNA relative to the transcribing polymerase. If free rotation is hindered in some way, a resulting torsional stress will propagate down the DNA, destabilizing (or overstabilizing) the double helix structure at some distant point. The resulting “topological coupling” between nearby genes has been observed in several experiments (see Sect. 2 below). Liu and Wang assumed a simple mechanism for the transport of torsional stress, following Levinthal and Crane<sup>1</sup><sup>1</sup>1Levinthal and Crane’s “speedometer-cable” motion will be called “plumber’s snake” motion, or “spinning” motion, in the present paper. . In a viscous medium a straight infinite rod meets a frictional resistance to axial rotation given by $$\tau =\mu _{\mathrm{spin}}\omega L$$ (1) Here the torque $`\tau `$ (with dimensions of energy) depends on the rotation rate $`\omega `$ (radians/sec) and length $`L`$ via a friction constant $`\mu _{\mathrm{spin}}`$. A simple calculation gives $`\mu _{\mathrm{spin}}=4\pi \eta R^21.310^{15}`$dyn$``$sec, where $`R1`$nm is the rod radius and $`\eta =0.01\mathrm{erg}\mathrm{sec}\mathrm{cm}^3`$ is the viscosity of water. Other authors give slightly different prefactors . Liu and Wang pointed out that the torsional friction constant $`\mu _{\mathrm{spin}}`$ appearing in eqn. (1) is extremely small due to the factor of $`R^2`$, and so they concluded that no significant torsional stress was possible in DNA of reasonable length without some additional physical anchoring. Absent such anchoring, both linear (open) and circular (plasmid) DNA would spin in place, like a plumber’s snake . For concreteness we will consider below the example of a linear DNA of length 3.5 kbp (1200 nm), rotated at its end with angular frequency $`\omega =60`$radians/sec; a related case is a 7 kbp construct, linear or circular, rotated near its center. In either case formula (1) gives a maximum torsional stress $`\tau 910^{18}\mathrm{dyn}\mathrm{cm}`$. Since the torque needed to denature DNA locally is several thousand times greater (see below), Liu and Wang’s conclusion seems to be safe. The analysis of this paper was motivated by several experimental observations which defy the familiar analysis just summarized (Sect. 2 below). A variety of assays, both in living cells and in vitro, have found significant torsional stress following transcription at a single promoter on unanchored DNA constructs. All these experiments are sensitive to topoisomerase, pointing to the role of torsional stress. The estimates given above imply that such large stresses are impossible. To resolve this paradox, the analysis in Sects. 34 below will show that the classical formula (1) can be very misleading: it vastly underestimates the torsional stress on the DNA duplex near the transcribing polymerase. The discussion rests on the observation that DNA is a heteropolymer, i.e. it is naturally bent on length scales longer than its persistence length of about 50 nm. For a curved molecule to spin in place without dragging sideways through the surrounding medium, as assumed in formula (1), requires constant flexing. The natural bends resist this flexing, forcing the molecule to translate through the fluid and greatly increasing the viscous drag through the surrounding water. (Fig. 1d below summarizes the model.) This enhanced drag indeed explains the large observed torsional stress near the point of transcription. ## 2 Experiments ### 2.1 General This section briefly reviews a few of the relevant experimental results, focusing on in vitro assays. Sect. 3 below describes our physical model. RNA polymerases are efficient motors: for example, E. Coli RNAP can generate forces of up to 20 pN against an opposing load . When the same mechanical energy is expended against a torsional load, it corresponds to a torque of 20 pN$``$0.34 nm/step divided by 2$`\pi `$ radians for every 10.5 steps, or $`10^{13}`$dyn cm, more than enough to induce structural transitions in DNA. The speed of transcription ranges from 50 nt/sec in eukaryotes to twice as great for T7 . The corresponding rotational driving rates are then $`\omega =30`$ and 60 radians/sec, respectively. The actual torsional stress during transcription need not, however, attain the maximal value just given. Liu and Wang’s twin supercoiled domain model rests on the observation that torsional stress will only build up if a) the polymerase itself is prevented from counterrotating about the DNA template, and b) a suitable torsional load opposes the rotation of the DNA at a point sufficiently close to the cranking polymerase. The present paper is concerned mainly with point (b), but for completeness we first digress to discuss (a). #### Anchoring at the transcribing polymerase A number of effects can prevent counterrotation of the polymerase. For example, in eukaryotes the polymerase may be physically attached to the nuclear matrix. Even without a rigid attachment, the eukaryotic polymerase holoenzyme is physically quite large and thus offers a large hydrodynamic drag to rotation. Similarly in prokaryotes, the nascent RNA transcript can begin translation before it is fully transcribed, leading effectively to a large complex consisting of polymerase, transcript and ribosome. Liu and Wang proposed a particularly attractive possibility: if the emerging protein is membrane-bound (for example, the tetracycline-resistance tet gene product), it can anchor its ribosome to the cell membrane . Many experiments have shown that translation of tet greatly increases twin-supercoil domain effects (see \[8, and references therein\]). The above mechanisms operate only in vivo. Remarkably, twin-supercoiled domain effects have also been observed in a number of in vitro assays, where no cellular machinery exists (see Sect. 2.2 below). At least three mechanisms can nevertheless create significant drag opposing counterrotation of the polymerase: i) polymerase has been found to create a tight loop in the DNA, greatly increasing its effective hydrodynamic radius and hence the drag for counterrotation ; ii) the nascent RNA transcript itself will create some hydrodynamic drag to rotation ; iii) under the conditions of most experiments (e.g. ) polymerase is present at concentrations leading to batteries of simultaneously-transcribing complexes. To relieve torsional stress, all active complexes would have to counterrotate simultaneously, with a drag proportional to their total number. #### Anchoring elsewhere Thus, even in vitro, transcription can effectively lead to the cranking of DNA by a nearly immobilized polymerase. As mentioned in point (b) above, however, cranking at one point still does not suffice to create torsional stress: DNA rotation must be effectively hindered somewhere else as well, since otherwise both linear and circular DNA would simply spin freely in place at the driving rate $`\omega `$. As in point (a), many mechanisms can anchor DNA in the crowded cellular environment. For example, in eukaryotes a DNA-binding protein could tie the DNA onto some part of the nuclear matrix. Another possibility, envisioned by Liu and Wang and implemented in several experiments, is to bind a second polymerase to the DNA and rely on its resistance to rotation as in (a) above. The second polymerase can either be stalled or actively transcribing in the opposite (divergent) sense from the first. Once again, however, the clearest results come from the in vitro assays mentioned earlier, in which only a single promoter is active on a circular or even linear (L. B. Rothman-Denes, unpublished results; D. Levens, unpublished results) template. In these experiments the only known hindrance to free spinning motion is the torsional hydrodynamic drag. If DNA were effectively a simple, straight, rod of diameter 2 nm, then the estimate in equation (1) would apply, and we could confidently predict that transcription would generate negligible torsional stress. Since the experiments contradict this expectation, we must modify the naïve physical picture of the transport of torsional stress in DNA. ### 2.2 Experimental results #### In vitro Tsao et al. made a circular plasmid with only one promoter actively transcribing . They assayed transient torsional stress in the wake of polymerase by allowing topoisomerase I to selectively eliminate negative supercoils, then measuring the remaining degree of positive supercoiling via 2d electorphoresis. They found that transcription induces a degree of supercoiling “much bigger than expected” and concluded that “it is possible that the degree of supercoiling generated by transcription is underestimated in the theoretical calculation” of . Dröge and Nordheim assayed torsional stress in a 3 kbp circular plasmid using the B-Z structural transition . They concluded that “Interestingly our results suggest that diffusion rate of transcription-induced superhelical twists must be relatively slow compared to their generation, and that under in vitro conditions localized transient supercoiling can reach unexpectedly high levels.” Similarly Dröge later found that transcription can induce site-specific recombination in vitro . Here the conclusion is that transcription created local torsional stress, in turn driving local writhing and bringing recombination sites into synapsis. Wang and Dröge later extended these experiments and called attention to the fact that torsional strain remains localized in a gradient region close to the polymerase, instead of spreading rapidly around the plasmid and cancelling at the antipodal point . Drolet, Bi, and Liu studied the reciprocal effects of topoisomerase I and gyrase , assaying with 1d electrophoresis. The result of interest to the present paper is that they found that membrane anchoring via the nascent TetA protein was not necessary for transcription-induced supercoiling, in contrast to earlier in vivo studies. Finally, Rothman-Denes et al. and Levens et al. (unpublished results) have used linear (open) 2300 nm templates including a T7 RNA polymerase promoter near the center. Transcription from this promoter by T7 RNA polymerase generates torsional stress. Rothman-Denes et al. used the activity of a bacteriophage N4 early promoter as a stress reporter. This promoter is inactive in its unstressed state and activated through cruciform extrusion at a superhelical density $`\sigma _{\mathrm{crit}}=0.03`$ , corresponding to a torsional stress of $`\tau _{\mathrm{crit}}710^{14}\mathrm{dyn}\mathrm{cm}`$, consistent with the estimate given above<sup>2</sup><sup>2</sup>2We estimate that about 30% of the superhelical density goes into twisting the double helix (and the rest into the mean writhe) . Multiplying $`0.3\sigma _{\mathrm{crit}}`$ by the microscopic twist stiffness $`Ck_\mathrm{B}T4.510^{19}`$erg cm and the relaxed Link density $`2\pi /(10.5\mathrm{bp}0.34\mathrm{nm}/\mathrm{bp})`$ gives the above estimate for $`\tau _{\mathrm{crit}}`$. Direct physical manipulation on stretched DNA gives similar results .. Levens et al. instead used an element of the human c-myc gene, which interacts with single-stranded DNA binding proteins, and measured unwinding using potassium permanganate, which reacts with single-stranded tracts. The results of both sets of experiments suggest that structural transitions are induced by T7 RNA polymerase transcription. Thus it again appears that transcription of linear DNA can create torsional stress several thousand times greater than that predicted by the classical formula (1). #### In vivo As mentioned above, in vivo experiments are harder to interpret, but nevertheless we mention a few illustrative results to show the very general character of the frictional-drag paradox. Rahmouni and Wells used a circular 6.3 kb plasmid, reporting its torsional stress via the B-Z structural transition . They concluded that “the diffusion of supercoils must be slower than was originally predicted \[in \]”. Lilley and collaborators have carried out an extensive series of experiments reviewed in . Their conclusion that an “as yet unidentified topological barrier should exist” may point to the same surprisingly large rotational drag argued for in the in vitro experiments above. In later work they also found that the transcribing polymerase need not be physically anchored, reinforcing the argument in point (a) of section 2.1 above . Turning finally to experiments in eukaryotes, we mention only two experiments of Dunaway and coworkers. Dunaway and Ostrander sought to eliminate any anchoring of their DNA template by injecting linear DNA with no subsequences known to associate with the nuclear architecture into Xenopus oocytes . They injected an exogenous (bacterial) polymerase into their oocytes and ensured that its promoter was the only spontaneously-transcribing promoter on their template. They also used linear templates, reducing the likelihood of any entanglement effects. Using 3.6–4.5 kb templates with a ribosomal RNA promoter to report torsional stress, they concluded that “localized, transient domains of supercoiling” could occur in open DNA, trapping significant torsional stress. Similarly, later work by Krebs and Dunaway concluded that “The viscous drag against a large DNA molecule is apparently sufficient to prevent transcription-generated supercoils from diffusing rapidly off the end of the DNA, so DNA length creates a topological domain” . Once again this conclusion is remarkable, in that it contravenes the estimates in Sect. 1 above. ## 3 Physical picture As described in Sect. 1, the surprising physical aspect of the experiment is the buildup of torsional stress in the DNA, when nothing seems to prevent the molecule from spinning almost freely in place. Apparently the simple physical model of a uniform elastic rod in a viscous fluid has left out some crucial effect. One may at this point be tempted to abandon simple physical models altogether, pointing to the many specific biochemical features of real DNA which they omit. But the elastic rod model successfully describes many detailed features of DNA stretching and fluorescence-depolarization experiments, including effects of torsional stress (e.g. ). Moreover, the surprising observed behavior is generic and robust, not specific to a particular situation, suggesting that the model needs only some simple new ingredient in order to capture the observed behavior. In this section we argue that augmenting the elastic rod model by including the natural bends in the DNA duplex dramatically changes the transport of torsional stress. The strength of these bends has been independently measured; it is not a new free parameter. Their effect on the equilibrium properties of DNA coils has long been recognized. In this section and the next we instead study their effects far from equilibrium. ### 3.1 Need for spin-locking Imagine a given segment of an elastic rod (modeling a twist-storing polymer such as DNA) as contained in a black box with only the two ends of the rod accessible. Cranking one end about its axis amounts to injecting a conserved quantity, “linking number” (or Lk), into the rod.<sup>3</sup><sup>3</sup>3Strictly speaking Lk is well defined only for a closed loop. Nevertheless, the change in Lk in an open segment with fixed end is well defined, and must vanish, whatever happens inside the black box. Rotating one end about its axis thus injects a conserved quantity. We can schematically think of linking number as taking one of five pathways away from the cranking site: 1. Lk can be elastically stored as twist in the rod: the rod segment can rotate about its axis by an amount which depends on position along the rod; 2. Lk can be elastically stored as writhe: the rod can begin to supercoil; 3. Lk can be transported by spinning (plumber’s-snake) motion, emerging at the far end with no net change in the rod state; 4. Lk can be transported by rigid rotation (crankshaft motion) of the whole segment about some axis; 5. Lk can be lost via the action of topoisomerase. We are interested in steady-state transport, in the absence of topoisomerase, and so we consider only the competition between pathways #3 and #4. This picture allows a more precise summary of the paradox reviewed in Sects. 12 above. The steady transport of injected Lk will meet with resistance in the form of effective frictional constants $`\mu _{\mathrm{spin}}`$ for spinning and $`\mu _{\mathrm{rigid}}`$ for rigid rotation, and hence a total frictional constant $`\mu _{\mathrm{tot}}=(\mu _{\mathrm{spin}}^{}{}_{}{}^{1}+\mu _{\mathrm{rigid}}^{}{}_{}{}^{1})^1`$. But we have seen that experimentally $`\mu _{\mathrm{tot}}`$ is much larger than the theoretically expected value of $`\mu _{\mathrm{spin}}`$. No matter how large $`\mu _{\mathrm{rigid}}`$ may be, it cannot resolve this paradox. In particular the well-known coupling between torsional stress and writhing motion (see e.g. and references therein) is of no help, since the problem is precisely that there is little torsional stress. What is needed is a way to shut down pathway #3, i.e. to lock the spin degree of freedom, at least partially. The fact that a uniform rod is never actually straight on length scales beyond its bend-persistence length $`A`$ does not help, either.<sup>4</sup><sup>4</sup>4Even in the absence of thermal motion, a naturally-straight rod will bend when cranked fast enough, executing a hybrid of rigid rotation and spinning . Wolgemuth et al. found, however, that for the parameters of interest to us here the Lk transport is dominated by spinning, exactly as argued above. The ease of spinning relative to translation of the rod through the fluid can alternately be understood from the point of view of Brownian fluctuations: a thin, axially-symmetric object receives random thermal kicks from the surrounding fluid, but these deliver very little torque due to the small rod radius $`R`$. By the general relation between diffusion and friction , we again obtain a rotational friction constant suppressed by powers of $`R`$, consistent with formula (1). Spinning creates no long-range hydrodynamic interaction, since the fluid velocity field falls off on the scale of the rod diameter $`R=1`$nm . Since $`A`$ is much larger than $`R`$, the straight rod approximation is adequate . Certainly the spinning in place of a thermally-bent but naturally-straight rod requires continuous flexing of the rod, as the direction of curvature rotates in the material frame of the rod, but the elastic cost of a bend in a cylindrical rod depends only on the magnitude, not the direction, of the curvature, and this does not change: such a rod has no energetic barrier to spinning. To summarize, the naïve equation (1) will be accurate, and torsional stresses will be small, unless some sort of locking mechanism inhibits free spinning of linear DNA in solution. To find such a mechanism, we must now introduce some new element of realism into our description of DNA. ### 3.2 Natural bends As mentioned in Sect. 1, the key ingredient missing so far from our model is the natural curvature of the DNA duplex. Immense effort has been focused on predicting the precise conformation of a DNA tract given its basepair sequence, using molecular modeling, oligomer crystallography, and NMR, among other techniques. Fortunately, for our problem it suffices to characterize the average effect of curvature over hundreds of basepairs. For such purposes a very simple phenomenological approach suffices. Natural DNA is a stack of similar but nonidentical subunits, arranged in an order which is fixed but random for our purposes. It is crucial that even though these bends are random, their effects do not average to zero on length scales much longer than one base-pair. Instead, the minimum-energy conformation of such a stack may be regarded as a distorted helix whose backbone follows a random walk, with a structural persistence length $`P`$. Note that $`P`$ is a purely geometrical parameter, having nothing to do with the mechanical bend stiffness $`\kappa _{\mathrm{bend}}`$ of DNA nor the thermal energy $`k_\mathrm{B}T`$. Instead $`P`$ reflects the information content in a piece of DNA. Just as in the straight case, bent (natural) DNA can also be deformed away from its minimum-energy state at some enthalpic cost characterized by a bend stiffness $`\kappa _{\mathrm{bend}}`$, with units energy$``$length. Since fluctuations are controlled by the thermal energy $`k_\mathrm{B}T`$, we define the bend length $`A=\kappa _{\mathrm{bend}}/k_\mathrm{B}T`$. The combined effect of thermal and natural bends then makes DNA a random coil with total persistence length<sup>5</sup><sup>5</sup>5Some authors call $`P`$ the “static persistence length” and $`A`$ the “dynamic persistence length”. Schellman and Harvey verified Trifonov et al.’s heuristic derivation of this formula within a number of detailed models . Since $`P^1<A^1`$ we can regard the bend disorder as smaller than the thermal disorder. In this case Trifonov’s formula also gives the effective persistence length measured by fitting DNA stretching experiments to the naïve worm-like chain model . $`A_{\mathrm{tot}}=\left(A^1+P^1\right)^1`$ . Under physiological conditions $`A_{\mathrm{tot}}`$ has the familiar value of 50 nm. Experiments on artificial, naturally-straight DNA make it possible to determine $`A`$ and $`P`$ separately, yielding $`A80`$nm and $`P130`$nm .<sup>6</sup><sup>6</sup>6 Though Bednar et al. did not estimate the uncertainty in their determination of $`P`$ , it may well be large. They note, however, that their direct experimental determination agrees with the model-dependent prediction of Bolshoy et al. . ### 3.3 Hybrid motion We wish to explore the consequences of the natural bends introduced in the previous subsection for the transport of torsional stress in DNA. Before doing any calculations, it is worthwhile to formulate some intuitive expectations, based on four increasingly realistic cartoons for the steady-state motion of a cranked DNA segment of contour length $`\mathrm{}`$ (Fig. 1a–d). As noted in Sect. 1, a straight, rigid segment (Fig. 1a) would encounter a torsional drag per unit length $`\mu _{\mathrm{spin}}\omega `$, or a net drop in torsional stress between the ends of $`\mu _{\mathrm{spin}}\omega \mathrm{}`$, with friction constant $`\mu _{\mathrm{spin}}`$ given below eqn (1). We argued in Sect. 3.1 that the case of a naturally-straight but semiflexible segment is similar (Fig. 1b). Matters change considerably when we introduce natural bends. If the rod were perfectly rigid (Fig. 1c), it would have to execute crankshaft motion; individual rod elements would then drag sideways through the fluid. We will see below that as $`\mathrm{}`$ increases, the corresponding drag per unit length would increase without bound. On long enough scales, then, we may expect that any realistic molecule cannot be regarded as infinitely stiff. At the other extreme, we could imagine the naturally-bent rod spinning in place. This however would mean that every joint periodically bends oppositely to its preferred conformation. The corresponding elastic energy cost creates a barrier to this motion. We will argue that in fact a real, semiflexible heteropolymer chooses a compromise between these extremes of motion, selecting a crossover scale $`L_C`$ and executing a hybrid motion (Fig. 1d). On length scales shorter than $`L_C`$ this motion is nearly rigid, since as just argued an activation barrier resists flexing. On longer length scales the motion must cross over to spinning, since as just argued rigid (crankshaft) motion meets a large viscous drag on long scales. We must now justify these intuitive ideas and obtain a numerical estimate for the crucial crossover scale $`L_C`$. Since $`L_C`$ will turn out to be significantly longer than the basepair step size, we will conclude that the spinning (plumber’s-snake) motion is effectively locked, as we argued was necessary in Sect. 3.1. ## 4 Scaling Analysis We must now justify and quantify the expectations sketched in Sect. 3. ### 4.1 Spin-locking Consider first the hypothetical case of a perfectly rigid, naturally-bent rod (Fig. 1c). The viscous force per length $`f`$ on a straight rod much longer than its radius $`R`$, dragged sideways through a viscous medium, is $$f\mu _{\mathrm{drag}}v=\frac{4\pi \eta }{0.8+\mathrm{ln}(X/2R)}v$$ (2) where $`v`$ is the speed and $`X`$ is the rod length. Our polymer is of course not straight on length scales beyond its structural persistence length $`P`$, so we substitute $`P`$ for the long-scale cutoff $`X`$ in eqn. (2). Since the dependence on $`X`$ is weak this is a reasonable approximation.<sup>7</sup><sup>7</sup>7A rod pulled at some angle other than $`90^{}`$ to its tangent will have a drag given by eqn. (2) with a slightly different prefactor; we will neglect this difference and use eqn. (2) in all cases. Taking $`P=130`$nm and $`R=1`$nm then gives $`\mu _{\mathrm{drag}}2.510^2\mathrm{erg}\mathrm{sec}\mathrm{cm}^3`$. Suppose we crank a rod segment of arc length $`\mathrm{}`$, which then rotates rigidly about an axis. Each element of the rod then moves through fluid at a speed $`v=r_{}\omega `$, where $`r_{}`$ is the distance from the rod element to the rotation axis (Fig. 1c). Multiplying the moment arm $`r_{}`$ times the drag force, eqn. (2), and integrating over the curve yields the torque drop $`\mathrm{\Delta }\tau =\mu _{\mathrm{drag}}\omega \mathrm{}r_{}^{2}`$ across the segment. Here $`r_{}^{}{}_{}{}^{2}`$ is the average of $`r_{}^{}{}_{}{}^{2}`$ along the rod segment. Each rod segment of course has a different sequence, and hence a different preferred shape. Each segment will therefore have a different value of $`r_{}^{}{}_{}{}^{2}`$. Fortunately, we are interested in the sum of the torque drops across many segments, each with a different, random, sequence. Thus we may replace $`r_{}^{}{}_{}{}^{2}`$ by its ensemble average over sequences, which we will call $`r_{}^{}{}_{}{}^{2}`$. This average has a simple form: eqn. 7.31 of Ref. gives $`𝐫_{}^{}{}_{}{}^{2}=\mathrm{}P/9`$, and hence $$\mathrm{\Delta }\tau =\mu _{\mathrm{drag}}\omega \mathrm{}^2P/9$$ (3) In the language of Sect. 3.1, we have just estimated the drag torque $`\mu _{\mathrm{rigid}}\omega \mathrm{}`$, finding $`\mu _{\mathrm{rigid}}\mu _{\mathrm{drag}}\mathrm{}P/9`$. Indeed we see that the drag per unit length grows with $`\mathrm{}`$, as suggested in Sect. 3.3 above. Formula (3) is valid when the segment length $`\mathrm{}`$ is longer than $`P`$, an assumption whose self-consistency we will check below. We can now relax the artificial assumption of a perfectly rigid rod and thus pass from Fig. 1c to the more realistic Fig. 1d. Suppose that a long polymer has been subdivided into segments of length $`\mathrm{}`$, each approximately executing rigid rotation about a different axis. The axes will all be different, since we are assuming that $`\mathrm{}`$ is longer than the structural persistence length $`P`$. To join these segments smoothly as they rotate, each segment therefore needs to flex. On average, each segment must periodically bend one end relative to the other by about $`90^{}`$. The least costly conformational change which accomplishes this is to spread the bending strain uniformly along the entire segment length $`\mathrm{}`$; we can then estimate the elastic bending-energy cost as <sup>8</sup><sup>8</sup>8Natural DNA can have localized regions of reduced bend stiffness. These flexible tracts will not significantly affect this estimate unless they are spaced more closely than the length scale $`L_C`$ found below. $$\frac{\mathrm{}\kappa _{\mathrm{bend}}}{2}\left(\frac{\pi }{2\mathrm{}}\right)^2$$ (4) This energy barrier becomes small for large $`\mathrm{}`$, just the opposite trend to that of eqn. (3). The physical reason for this behavior is that we do not insist on ironing out every small kink in the rotating rod’s shape; the rod segment can satisfy the imposed conditions on its ends by deforming only a fraction of its many intrinsic bends. The bending energy needed to crank the segment through an angle $`\theta `$ is roughly eqn. (4) times $`\frac{1}{2}(1\mathrm{cos}\theta )`$; the torque needed to increase $`\theta `$ is then the derivative of this formula, $`\frac{1}{2}\mathrm{sin}\theta `$. Thus the driving torque needed to overcome the bending-energy barrier turn through a complete revolution is just one half of eqn. (4). The crossover length $`L_C`$ is then the value of $`\mathrm{}`$ at which the viscous torque drop, eqn. (3), just balances this critical value: $$\mu _{\mathrm{drag}}\omega L_{C}^{}{}_{}{}^{2}P/9=\frac{k_\mathrm{B}T}{2}\frac{A}{L_C}\frac{\pi ^2}{8}$$ (5) Substituting the numerical values we find $`L_C450`$nm for T7 RNAP, and slightly larger for other, slower, polymerases. Our crossover length has indeed proven to be longer than the structural persistence length $`P`$, so the assumption $`\mathrm{}>P`$ made above is self-consistent. Indeed, $`L_C`$ has proven to be about 1.4 kbp. In our illustrative example of a 7 kbp DNA construct cranked at the midpoint, we see that intrinsic bends shut down spinning motion almost completely: the naïve model of Sect. 1 does not describe the true motion at all. We must now see what this implies for the overall torsional stress on the construct. ### 4.2 Hydrodynamic interactions In contrast to spinning in place, dragging a thin rod sideways sets up a long-range flow field. Now that we know that spinning is effectively forbidden, we must therefore study the possibility of long-range hydrodynamic interactions between rod segments. The theory of polymer dynamics tells us that a short random coil dragged through fluid can be viewed as a set of thin-rod elements moving independently in a motionless background (the “free-draining” case), but a long coil instead moves as a solid spherical object, due to hydrodynamic interactions . The crossover between these two regimes is controlled by the dimensionless parameter $`Q\sqrt{\frac{L}{A_{\mathrm{tot}}}}\frac{\mu _{\mathrm{drag}}}{\eta }`$. Free draining corresponds to the case $`Q1`$. For our illustrative example of a coil of length $`L=2300`$nm and total persistence length $`A_{\mathrm{tot}}=50`$nm, we get $`Q=17`$, interactions are important, and the coil moves as a solid sphere. The viscous drag torque on such a coil is $`\tau =\mu _{\mathrm{coil}}\omega L`$, where $`\mu _{\mathrm{coil}}=\frac{4}{9}\sqrt{3\pi ^3LA^3}1.26\eta `$ (see §31 of ). Dividing this torque equally between the upstream and downstream halves of the construct, we find the estimated torsional stress on either side of the cranking point to be $`\omega 1.010^{15}`$dyn cm sec. Taking $`\omega =60`$radian/sec then gives a torsional stress of $`610^{14}\mathrm{dyn}\mathrm{cm}`$, comparable to the value quoted in Sect. 2 as necessary to induce structural transitions and about seven thousand times greater than the naïve estimate given below eqn. (1). ### 4.3 Relation to prior theoretical work The viewpoint taken in this paper can be regarded as a synthesis of two established threads. #### Fluid-mechanics work One of these threads studies the deterministic dynamics of externally driven (i.e. far from equilibrium) rods in a viscous environment. For example, Garcia de la Torre and Bloomfield studied the effects of a single, permanent, large-angle bend on viscous drag , obtaining precise versions of some of the formulæ given above. Individual large-angle bends caused by DNA-binding factors may well be present in vivo, but our point here is that a statistical distribution of small, finite-stiffness bends still leads to dramatic effects. Several authors have studied the interplay between shape and twist in the dynamics of naturally straight, flexible rods in a viscous medium , again obtaining precise formulæ for situations simpler than that studied here. It would be very interesting to incorporate intrinsic bends into their formalism. Finally, Marko has proposed that the impulsive (jumpy) action of RNA polymerase can lead to transient torsional stresses greater than predicted by the naïve formula, eqn. (1) . The range of this enhancement, however, depends on the time scale of each step and may be too short to explain the observed phenomena. Experimental measurement of this time scale will be needed to assess this proposed mechanism. #### Simulation work A second thread is the extensively studied problem of the equilibrium fluctuations of a polymer, particularly the diffusive torsional motion of DNA as measured in fluorescence experiments. Most of this work used Monte Carlo or Brownian dynamics numerical simulation techniques; most did not introduce long-range hydrodynamic interactions as we did in Sect. 4.2 above. Fujimoto and Schurr noted that fitting experimental fluorescence polarization anisotropy data to a model of intrinsically-straight DNA yielded an effective hydrodynamic radius which increased with increasing segment length . They suggested the possibility that this effect could be caused by permanent or long-lived bends in DNA. Collini et al. took up the same problem , explicitly introducing intrinsic bends. Their physical model, however, was the crankshaft motion of a perfectly rigid, zig-zag shape. The zig-zag shape introduces structure on one length scale. A major point of the scaling analysis in Sect. 4 above, however, was that the minimum-energy conformation of natural DNA is actually a random coil, and random walks have structure on all length scales. A second key point of our analysis was that DNA is not infinitely stiff, leading to the crossover phenomenon found in Sect. 4.1. Schurr et al. distinguished between “phase-locked bends”, equivalent to the natural bends in the present work, and “non-phase-locked bends” including the thermal bends of the present work.<sup>9</sup><sup>9</sup>9Another example of a non-phase-locked bend could be a universal joint: a bend maintaining fixed polar angle but free to swivel in the azimuthal direction. Schurr et al. also distinguish between slowly- and rapidly-relaxing bends. The present work assumes that the large external applied torsional stress (absent in the equilibrium situation studied in ) suffices to overcome any kinetic barriers to elastic deformation of the DNA duplex. They verified using Monte Carlo simulation that in the absence of natural bends, the torsional drag on a thermally-bent rod is the same as that for a straight rod, as argued physically in Sect. 3.1 above. Schurr et al. went on to anticipate the hybrid motion studied in the present work, proposing that “beyond some length the degree of global phase locking should decrease, as the motion approaches that of a wobbly eccentric speedometer cable, and the effective hydrodynamic radius should reach a plateau value, which is possibly 1.2 nm. The available evidence indicates that this radius is independent of length for $`L>60`$nm” . The authors did not, however, present a model incorporating random natural-bend disorder. The present work predicts instead that the response of DNA to external cranking is controlled by an effective drag constant that does not saturate until $`L>L_C`$. The crossover scale $`L_C`$ depends on the transcription rate via eqn. (5) and is typically hundreds of nanometers; the saturation value of the effective hydrodynamic radius is then much greater than 1.2 nm. The driven situation of interest here is not, however, the same as the equlibrium situation studied in . Finally, A. Maggs has independently shown that in a naturally-straight, thermally-bent rod twist relaxation follows the same diffusive law as in a rigid straight rod, out to extremely long scales (over 2 kbp) (A. Maggs, unpublished results). Beyond this scale Maggs found that pathway #2 in Sect. 3.1 above begins to affect twist relaxation, leading to an interesting new scaling relation. ## 5 Conclusion The analysis of this paper rests upon a surprising fact from slender-body viscous hydrodynamics. The drag torque for spinning a thin rod behaves reasonably as one decreases the rod radius $`R`$: it is proportional to $`R^2`$. In surprising contrast, the drag force for pulling such a rod sideways is practically independent of $`R`$ (eqn. (2) above). The only length scale available to set the rotational drag for rigid crankshaft motion is then the radius of curvature of the rod. But a randomly-bent rod has structure on every length scale, and so the drag torque per length increases without bound for longer segments until the crossover condition, eqn. (5), is met. Since the crossover scale $`L_C`$ proves to be long, cranked DNA is effectively spin-locked on scales shorter than at least 1 kbp. This observation explains why the naïve formula, eqn. (1), is inapplicable, eliminnating the paradox described in Sect. 1. The transport of torsional stress may enter in many cell processes. While this paper has stressed its possible role in gene regulation, torsional stress has recently been assigned a role in the disassembly of nucleosomes in front of an advancing polymerase complex (e.g. ), in chromatin remodeling (e.g. ), and in the action of enzymes on DNA (e.g. ). The ideas of this paper may be relevant to these problems too, though of course in eukaryotes the phenomenon described here may be preempted by the effects of higher-order chromatin structure. Direct manipulation of single DNA molecules sometimes involves cranking as well (e.g. ). The simple scaling analysis used in this paper makes some testable predictions. The key claim has been that intrinsic bends can have a huge effect on the transport of torsional stress along DNA. For example, synthetic DNA engineered to be less bent than natural sequences will have longer crossover scale $`L_C`$ (eqn. (5)), and hence should support less torsional stress for a given length. Shortening a linear template below $`L_C`$ should also sharply reduce the overall drag coefficient. More generally, none of the experimental papers cited earlier made quantitative estimates of the effective torsional friction constant needed to explain their results. One could imagine an in vitro experiment using local stress reporters (e.g. the B-Z structural transition), inserted at various positions, to get the full torsional stress profile, in space and time, as function of transcription rate. Even a limited subset of this quantitative information would yield insight into the mechanisms of torsional stress transport. ## Acknowledgments I wish to thank S. Block, N. Dan, P. Dröge, M. Dunaway, R. E. Goldstein, J. Marko, T. R. Powers, J. M. Schurr, and C. Wiggins, for valuable discussions, and particularly D. Levens, A. Maggs, and L. B. Rothman-Denes for describing their unpublished work. This work was supported in part by NSF grant DMR98-07156.
no-problem/9907/quant-ph9907028.html	ar5iv	text	# Testing the symmetrization postulate of quantum mechanics and the spin-statistics connection ## I Introduction Several experiments have been reported that assert to test the symmetrization postulate of quantum mechanics and/or the spin-statistics connection. The symmetrization postulate and the spin-statistics connection are indeed amongst the fundamental tenets of quantum mechanics. The symmetrization postulate establishes that in a system containing identical particles the only possible states are either all symmetrical or all antisymmetrical with respect to permutations of the particles. In the first case, the particles are called bosons and follow Bose-Einstein statistics; in the second case they are called fermions and follow Fermi-Dirac statistics. Experiments indicate that particles with integer values of spin are bosons, while particles with odd-half-integer spin are fermions. The reason why only symmetric and antisymmetric states seem to occurr in nature and the connection with the spin of the particles has been a puzzle since the early days of quantum mechanics . The spin-statistics theorem, proved by W. Pauli from the basic principles of quantum field theory and special relativity such as the requirement of local commutativity of observables, states that, given the choice between Bose and Fermi statistics, integer-spin particles must obey Bose statistics and odd-half-integer-spin particles must obey Fermi statistics. Proofs of the spin-statistics theorem are reviewed in . Quantum mechanics would nevertheless allow also symmetries different from those imposed by the symmetrization postulate, and theories have been developed allowing for small deviations from conventional statistics which might have been masked in the experiments performed so far. It is worth noting that no theory so far predicts the possibility of observing a violation in a particular system or in some specific condition. Consistent theories can be formulated, however, which would lead to different symmetry properties. Experiments are needed then to discriminate between these theories imposing constraints which are the more stringent the higher is the experimental precision. In this paper, after a brief account of the theoretical background, a discussion of experimental tests is presented, trying to clarify in a consistent manner their meaning. It is shown that most of the recent experiments should be considered as testing the spin-statistics connection rather than the validity of the symmetrization postulate. An experiment to search for violations of the symmetrization postulate for nuclei is proposed. ## II Theoretical background It is not the purpose of this paper to discuss in detail the theories which lead to symmetry properties different from the ones which are peculiar to bosons and fermions. A general survey can be found in . It is of interest, however, to put in evidence the possibility of theories allowing for small deviations from the usual symmetry relations, whose search is the subject of the experiments discussed in this paper. Such deviations can be expressed as a different symmetry of the state under particle exchange or, in Fock-space representation, as a deformation of the algebra of the creation and annihilation operators. A statistics intermediate between Bose and Fermi cases was first proposed in considering the possibility that at most $`n`$ identical particles could occupy the same quantum state. This idea led to a generalized field theory , called parastatistics, in which the field operators obey trilinear commutation relations instead of the usual bilinear relations. These theories predict, however, gross violations of statistics which are immediately excluded by experimental evidence. The possibility of a continuous interpolation between bosonic and fermionic behaviours is given by ”quons” . The commutation and anticommutation relations are replaced by generalized bilinear commutation relations depending on a parameter $`q`$ ($`q`$-mutators): $$a_ka_l^+qa_l^+a_k=\delta _{kl},1q1$$ (1) with the vacuum condition $`a_k\|0=0`$. As $`q`$ varies between -1 and 1, the symmetry changes continuously from the completely antisymmetric case (fermions) to the completely symmetric case (bosons). It has been shown that this interpolation preserves positivity of norms and the non-relativistic form of locality . Other aspects are still doubtful such as the possibility of accounting for local observables in a relativistic theory or for the existence of antiparticles. Statistics other than Fermi and Bose have also been investigated for one- and two- dimensional systems and in connection with anyon high-temperature superconducting systems . Two important points must be noted to avoid confusion in the interpretation of experimental tests. The first is that in a system including two identical particles, only two ”entangled” states are possible, namely a symmetric and an antisymmetric state: $`\mathrm{\Psi }_S(1,2)={\displaystyle \frac{1}{\sqrt{2}}}[\psi _a(1)\psi _b(2)+\psi _b(1)\psi _a(2)]`$ (2) $`\mathrm{\Psi }_A(1,2)={\displaystyle \frac{1}{\sqrt{2}}}[\psi _a(1)\psi _b(2)\psi _b(1)\psi _a(2)]`$ (3) In this case, the symmetrization postulate does not play any role, the question being only the connection between the spin of the particles and the symmetry of the two-particle state. This is not the case for systems including more than two identical particles, where mixed-symmetry states are possible. The second important remark is that, as was pointed out in , in the framework of ordinary quantum mechanics transitions between states of different permutation symmetry cannot take place. This ”superselection rule” is a rigorous selection rule holding also in the presence of perturbations such as collisions or external fields. It is not possible then to consider states given by a coherent superposition of states of different permutation symmetry. The system must be described as an incoherent mixture which is represented by a density matrix. In the case of two particles, for example, the two-particles density matrix for small violations of Bose statistics is: $$\rho _2=(1\frac{1}{2}\beta ^2)\rho _s+\frac{1}{2}\beta ^2\rho _a.$$ (4) where $`\rho _{s(a)}`$ is the symmetric (antisymmetric) two-particle density matrix. Neglecting these important points led to a wrong interpretation of some of the experiments. Using the notation adopted in the literature, in the following $`\beta ^2/2`$ indicates the ”symmetry-violation” parameter. Its real meaning needs to be specified, however, for the particular physical system and the theoretical model considered. In , for example, it was shown that in the frame of a $`q`$-mutator theory, the value of $`\beta ^2/2`$ can be related to the value of the $`q`$ parameter. ## III Experiments In this section, experiments performed to search for violations of the symmetrization postulate and/or of the spin-statistics connection are discussed. The focus is on recent experiments and, in particular, on experiments on integer-spin nuclei in molecules. For a review of the work before 1989, the reader is referred to . As already mentioned, some of the initial experiments suffered from a misunderstanding of the constraints imposed by the ”superselection rule”. It is shown here that also the interpretation given for some recent experiments is not correct. In fact, although the papers published on this subject usually present their ”null results” as a confirmation of the validity of the symmetrization postulate, most of them should be considered as tests of the spin-statistics connection. The argument is simple: as discussed above, in a two-particle system, only symmetric or antisymmetric states are possible. The symmetrization postulate does not play any role in this case. Experiments involving only two identical particles should then be considered as tests of the spin-statistics connection rather than tests of the symmetrization postulate. An experiment that would allow a genuine test of the symmetrization postulate is proposed in the following. ### A Experiments on half-integer-spin particles. A few experiments have been performed to test the validity of the Pauli principle for half-integer-spin particles. In particular, a high precision test on electrons was performed in by running a current through a copper bar and searching for the X-rays that would be emitted if some of the electrons introduced in the sample were captured by a copper atom and cascaded down to the 1S state, which is already filled with two electrons. No signal was found and this was interpreted as giving a limit $`\beta ^2/21.7\times 10^{26}`$ to the probability that a new electron introduced into copper would form a mixed-symmetry state with respect to the other electrons already present in the copper sample. The large number of electrons in the system was important to reach such a high sensitivity but, on the other hand, makes the interpretation of the result more complicated. Conclusions may change depending on whether we consider the symmetry of the system composed by the injected electron plus the electrons already present in the copper bar, or we consider a model in which the electron collides with a copper atom and is captured. A simpler two-electron system was investigated in . A spectroscopic test was performed on helium atoms, searching for a transition involving the permutation symmetric $`1s2s{}_{}{}^{1}S_{0}^{}`$ state. An upper bound $`\beta ^2/25\times 10^6`$ was set to a violation of the Pauli principle. In spite of the lower sensitivity, the interpretation of this result is simpler. Since only two identical particles are involved, this must considered as a test of the spin-statistics connection. Doubts can be raised, however, about what would be the chemical stability of ”paronic” atoms in ordinary samples. In , this was taken into account by having the atoms ionize and recombine in a discharge before entering the detection region. ### B Experiments on integer-spin particles. No accurate test for integer-spin particles had been reported until recently. This is due to the fact that while there are several systems in which a violation of the Pauli exclusion principle would be detected as a signal on a zero background, the effect of a small violation for particles following Bose-Einstein statistics would usually manifest itself as a small change in the properties of a many-particle system. This obviously limits the achievable accuracy. In , a bound to a possible violation of the generalized Bose statistics for pions was inferred considering the $`K_L\pi ^+\pi ^{}`$ decay, which is usually considered as due to CP violation. A limit of $`\beta ^2/210^6`$ was obtained. Photons. Several papers have been published recently reporting or proposing experiments to set a limit to possible violations of Bose statistics for photons . The fundamental nature of the photon and its peculiar properties makes it very interesting to investigate this particle in this context. It is hard, however, to find an experiment that would give a direct evidence of a violation with a significant sensitivity. This is one case, in fact, in which a small deviation from normal statistics would usually produce only a small signal over a large background. An attempt to set a limit to a possible violation of Bose statistics was made in , based on light intensities attainable in laser systems. In , a possible dependence of the frequency of light on its intensity was searched for. This effect is expected if a $`q`$-nonlinearity is introduced in the description of the electromagnetic field. Since nonlinearities in the commutation relations give rise to mixed-symmetry states, this experiment could also be reinterpreted as a search for a violation of the symmetrization postulate for photons. The connection is not straightforward though and was not pursued in the paper. In , the experimental upper limit on the two-gamma decay of $`Z`$-boson, $`Z\gamma \gamma `$, was used to establish an upper bound $`\beta ^2/210^2`$ on a possible small violation of the exchange-symmetry for a system of two photons. The same idea, based on what is called Landau-Yang theorem, was exploited in to improve the limit by searching for the forbidden $`J=0J^{}=1`$ transitions in atoms excited by two photons of the same energy. A limit of $`\beta ^2/27\times 10^8`$ was set on the probability that two photons are in an exchange-antisymmetric state. A different approach was followed in . A very tight bound to a violation of statistics for photons was inferred considering photons and electrons as coupled ”quons” and relating the bound for photons to that obtained in for electrons. Although this argument is indirect and model-dependent, it is very interesting and it could also be extended to other particles. Nuclei in molecules. In 1931, Ehrenfest and Oppenheimer showed that a composite system of fermions is a boson or a fermion depending on whether it is made of an even or an odd number of fermions . Considering the total angular momentum resulting from the constituents angular momenta, an extension of the spin-statistics theorem to composite systems is obtained. For this argument to be valid, it is necessary that the interaction between the composite particles is negligible compared to the internal excitation energy so that the internal structure can be neglected and the system can be considered as a single particle. This has a striking demonstration in recent experiments on Bose-Einstein condensation of atoms : the phase transition is observed only for the isotopes of the atoms for which the sum of the number of protons, neutrons and electrons gives an even number, that is when the number of neutrons in the nucleus is even. The same argument applies to the symmetry of systems including identical nuclei. In , it was proposed that a high-sensitivity investigation of the spectra of molecules containing identical spin-0 nuclei would allow to search for violations of Bose statistics for the nuclei. The basic idea of these experiments is analogous to the one underlying the experiment on electrons in the helium atom . This represents indeed a rare case in which a violation of the spin-statistics connection for integer-spin particles would be detected on a virtually zero background with a sensitivity comparable to the one achieved in experiments on fermions. Let us consider a molecule containing two identical spin-0 nuclei as, for example, <sup>16</sup>O<sub>2</sub>. According to the Born-Oppenheimer approximation and neglecting the coupling of the nuclear spin with the rest of the molecule (which is not important for these experiments since the spin of the nuclei is zero), the total wave function $`\psi _t`$ can be written in the form $$\psi _t=\psi _e\psi _v\psi _r\psi _n$$ (5) where $`\psi _e`$, $`\psi _v`$ and $`\psi _r`$ are the electronic, vibrational and rotational functions, respectively, and $`\psi _n`$ is the nuclear spin function. For integer-spin nuclei, the total wavefunction $`\psi _t`$ must be symmetric in the exchange of two nuclei. Since the nuclear spin is zero, $`\psi _n`$ is obviously symmetric. The vibrational wave function $`\psi _v`$ is also unaltered in the exchange of the nuclei because it depends only on the magnitude of the internuclear distance. Since the total wavefunction $`\psi _t`$ must be symmetric, only the states corresponding to even (odd) rotational quantum numbers are allowed if $`\psi _e`$ is symmetric (antisymmetric) . In a first series of experiments , the spectrum of the <sup>16</sup>O<sub>2</sub> molecule was investigated searching for transitions between states which are antisymmetric under the exchange of the two nuclei. An upper bound $`\beta ^2/2`$ 5$`\times 10^7`$ was set in , and a similar, slightly less accurate, result was obtained in . Experiments on <sup>16</sup>O<sub>2</sub> were later reported in . In , a limit $`\beta ^2/2`$ 5$`\times 10^8`$ was obtained. In a new accurate test on <sup>16</sup>O nuclei was performed by investigating the vibrational spectrum of the <sup>12</sup>C<sup>16</sup>O<sub>2</sub> molecule. The CO<sub>2</sub> molecule has the same symmetry properties of O<sub>2</sub> but, since it is triatomic, it has strong active vibrational bands in the infrared which are lacking in O<sub>2</sub> spectra. In particular, the intensity of the combination band 12<sup>0</sup>1 - 00<sup>0</sup>0 around 2 $`\mu `$m investigated in is more than two orders of magnitude larger than the electronic transitions of oxygen previously investigated, and therefore the absorbance of a given population is correspondingly larger. This results in an increased detection sensitivity. Since the ground electronic wavefunction of the CO<sub>2</sub> molecule is symmetric in the exchange of the two <sup>16</sup>O nuclei, the rotational wavefunction in the ground vibrational state must be symmetric and only even values for the rotational quantum number J are allowed. The R-branch investigated in should then be composed only of R(2J) transitions. A bound of $`\beta ^2/2`$(2.1$`\pm `$0.7)$`\times `$ 10<sup>-9</sup> to the relative population of the forbidden states was deduced in this work. This experiment gives at present the most stringent test of the spin-statistics connection for <sup>16</sup>O nuclei. Further improvements are possible: the sensitivity could be increased by reducing the noise to the quantum level, by increasing the absorption pathlength, or by selecting stronger transitions. The fundamental vibrational band of CO<sub>2</sub> around 4.3 $`\mu `$m, which is at least a factor of thousand stronger than the one at 2 $`\mu `$m, could be investigated. A dramatic increase in the detection sensitivity could be obtained by a resonant-ionization-spectroscopy detection scheme. A similar test can be performed on other spin-0 nuclei; a straightforward extension is the investigation of the spectrum of <sup>18</sup>O<sub>2</sub> . An interesting prospect is the investigation of spectra of molecules containing more than two identical nuclei. As shown above, most of the experiments performed so far involve only two identical particles so that they provide a test of the spin-statistics connection. In order to search for possible violations of the symmetrization postulate of quantum mechanics, it is now clear that systems including more than two identical particles should be considered. In this case, indeed, more complex symmetries are possible and the Young diagrams do not reduce to the trivial completely symmetric and completely antisymmetric cases. On the other hand, systems composed of very large numbers of identical particles can complicate the interpretation of experimental results. Molecules offer the possibility of investigating more and more complex structures and, as for the experiments performed on O<sub>2</sub> and CO<sub>2</sub>, high detection sensitivity can be achieved by laser spectroscopy methods. A good candidate for this experiment is OsO<sub>4</sub>, a highly symmetric molecule with four identical spin-0 nuclei. In spite of the higher complexity of the spectrum with respect to the simpler molecules investigated previously, high resolution and high sensitivity spectroscopy schemes have been developed, especially in the region around 10 $`\mu m`$ which is of metrological interest. The good knowledge of molecular parameters makes it simpler to find the position of the relevant transitions and to separate them from spurious signals. In particular, transition frequencies can be singled out that would represent a signature of a violation of the symmetrization postulate. The expected detection sensitivity is expected to be comparable to that achieved in previous tests on molecules. ## IV Conclusions Several experiments confirm the validity of the spin-statistics connection for various types of particles to a high level of accuracy. This provides a proof of the general formalism of quantum mechanics and can be extended also to different particles. In a recent paper , the results obtained in for nuclei was used to set a bound on possible violations of the Pauli exclusion principle for nucleons and for quarks. On the other hand, not many experiments have really tested the validity of the symmetrization postulate of quantum mechanics with high precision. A possible experiment on nuclei in molecules was proposed in this paper which should allow to test the symmetrization postulate with an accuracy comparable to that achieved in previous tests of the spin-statistics connection. The author acknowledges useful discussions with G. Modugno and thanks R.C. Hilborn for providing preprints of his recent papers.
no-problem/9907/astro-ph9907034.html	ar5iv	text	# ROSAT X-ray observations of 3CRR radio sources ## 1 Introduction Soft X-ray emission from radio galaxies and radio-loud quasars probes a range of physical conditions. On the largest scales, the X-rays are the thermal emission of hot intracluster gas. In general we do not expect a strong quantitative relationship between this emission and the radio properties of the galaxy or quasar, although there is direct evidence for the influence of one on the other in a few well-studied objects \[e.g. NGC1275, Böhringer et al. (1993); Cygnus A, Carilli, Perley & Harris (1994); 3C 449, Hardcastle, Worrall & Birkinshaw (1998c)\]. On smaller scales, X-ray emission may come directly from components of the radio source, as synchrotron, inverse-Compton, or synchrotron self-Compton (SSC) radiation. There is evidence for X-ray synchrotron emission in the jets of M87 (Biretta, Stern & Harris 1991) and Centaurus A (Turner et al. 1997) and in a hotspot of 3C 390.3 (Prieto 1997), and for a synchrotron self-Compton process in the hotspots of Cygnus A (Harris, Carilli & Perley 1994). Moreover, inverse-Compton scattering of CMB or active nucleus photons may be responsible for extended emission seen in Fornax A (Feigelson et al. 1995) and 3C 219 (Brunetti et al. 1998). However, in general such kpc-scale X-ray emission is likely to be weak compared to that due to hot gas. Inverse-Compton processes are more important on parsec scales, where electron and photon densities are much higher. Such emission is unresolved by present-day X-ray observatories. Synchrotron emission, along with thermal emission from the central engine and accretion disc, may also contribute to an unresolved nuclear component. Obscuration has a strong effect on what is observed. Unified models of radio sources (e.g. Urry & Padovani 1995) require an obscuring ‘torus’ of dusty material around the central engine, so that nuclear optical continuum and broad lines are not seen in objects classed as radio galaxies. This torus should also obscure soft X-ray emission originating close to the central engine, leading to suggestions (e.g. Crawford & Fabian 1996a) that soft X-ray emission from powerful radio galaxies should in general be dominated by the extended thermal component. In hard X-rays, a strongly absorbed nuclear component has been found in a few radio galaxies, consistent with this picture (e.g. Ueno et al. 1994). However, the radio-related component of the nuclear emission may originate on scales larger than those of the torus. Although not affected by obscuration, it would be rendered anisotropic by bulk relativistic beaming. There is substantial evidence that unabsorbed radio-related non-thermal X-rays are seen in radio galaxies. Recent work has strengthened the conclusion of Fabbiano et al. (1984) that the correlation between nuclear soft X-ray and radio-core luminosity implies a jet-related origin for the X-ray emission, as high-resolution X-ray observations have allowed point-like and extended components to be separated (Worrall & Birkinshaw 1994; Edge & Röttgering 1995; Worrall 1997). A relativistically beamed radio-related component almost certainly dominates the soft X-ray emission in core-dominated quasars and BL Lac objects. A large, unbiased sample of objects is most effective for investigating the origins of X-ray emission in extragalactic radio sources. The 3CRR sample (Laing, Riley & Longair 1983; Laing & Riley, in preparation) is a flux-limited sample of the brightest radio galaxies and quasars in the Northern sky. It includes all objects with 178-MHz flux density greater than 10.9 Jy (on the scale of Baars et al. 1977) having $`\delta >10^{}`$ and $`\|b\|>10^{}`$. Excluding the starburst galaxy 3C 231 (M82), which will not be considered further in this paper, it contains 172 objects, of which, at the time of writing, 89 have been observed (either intentionally or serendipitously) in ROSAT pointed observations at off-axis angles less than 30 arcmin (and usually on-axis). Most sources in the sample are powerful, distant FRII (Fanaroff & Riley 1974) radio galaxies and quasars, and almost all are well-studied in the radio and optical, making 3CRR an ideal sample for a comparison of radio and X-ray properties. Although such comparisons have been carried out previously (e.g. Fabbiano et al. 1984; Crawford & Fabian 1995, 1996a; Prieto 1996), none has made full use of ROSAT’s high spatial resolution to try to distinguish between different physical origins for the X-ray emitting components, and in particular to test the suggested relationship between X-ray and radio core emission. In this paper we tabulate the X-ray count rate from each of the 89 ROSAT-observed 3CRR sources and present our best estimates of the count rate associated with a central unresolved component. We use radio data from the literature to discuss whether a coherent picture of the physics can be derived from the X-ray and radio observations. Complex sources are discussed in detail and we provide references to existing studies. Some objects are the subject of more detailed work which will appear elsewhere. Throughout the paper we use a cosmology in which $`H_0=50`$ km s<sup>-1</sup> Mpc<sup>-1</sup> and $`q_0=0`$. Spectral index is defined in the sense that flux density is proportional to $`\nu ^\alpha `$. ## 2 Data and analysis Table 1 lists the 89 ROSAT-observed sources. \[The ROSAT Master observation catalogue lists a further three 3CRR objects (3C 34, 3C 336 and 3C 438) for which data were not public in time for our study. With the demise of ROSAT, further data will not be forthcoming.\] Optical and radio classifications and redshifts come largely from the updated 3CRR tables of Laing & Riley (in prep.), updated with more recent spectroscopic observations where available (e.g. Laing et al. 1994; Laing, private communication; Jackson & Rawlings 1997). The classification of objects by whether or not they have strong high-excitation nuclear emission lines dates back to Hine & Longair (1979); we use the definition of Laing et al. (1994) who define high-excitation objects as having \[OIII\]/H$`\alpha >0.2`$ and equivalent widths of \[OIII\] $`>3`$Å. Low-excitation objects are excluded from many radio galaxy-quasar/broad line radio galaxy unified models (e.g. Barthel 1994) because broad lines in their spectra would be very hard to detect, so that there is no optical evidence for their orientation to the line of sight. They also appear to have rather different radio properties from high-excitation radio galaxies (Jackson & Rawlings 1997, Hardcastle et al. 1998b). For each source we give the ROSAT observation request (ROR) number of the primary data set we have used and the source’s angular offset from the pointing direction where this is significantly different from zero. Because we are most interested in the sources’ spatial properties in this paper, we have in general preferred HRI observations over PSPC where both exist (the 50 per cent encircled energy radii are $`13`$ and $`3`$ arcsec for the PSPC at 1 keV and the HRI, respectively), though we have checked all HRI results against any existing PSPC data for consistency. Where an X-ray source is reasonably bright, and there is no obvious evidence for significant source extension, we have used an on-source circle of radius $`1`$ arcmin and taken background from an annulus of radii 1 and 2 arcmin about the source (for HRI data) or $`2`$ arcmin and 2 and 3 arcmin (for PSPC data), taking care in all cases to exclude unrelated contaminating sources. Where a smaller source circle was used, we have corrected the derived count rate to that expected from the standard source and background regions, for consistency (this affects only a few sources). Counts and background-subtracted count rates for the few sources observed significantly off-axis are corrected for the effects of vignetting (calculated at 1 keV) and any additional obscuration. Upper limits on the total (resolved plus unresolved) counts, quoted where a source was undetected, are $`3\sigma `$ limits, derived by applying Poisson statistics to a detection cell of a size appropriate to the off-axis angle of the source \[for on-axis sources, the sizes used are $`5\times 5`$ arcsec (HRI) and $`30\times 30`$ arcsec (PSPC)\]. The well-known ‘aspect smearing’ problem of the HRI (e.g. Morse 1994) means that we cannot reliably distinguish unresolved from resolved components in HRI data simply by fitting the nominal HRI point response function (PRF), and determining the number of counts associated with a possible central point source necessarily involves a certain amount of interpretation. Our procedure, justified by detailed work on sources known to be point-like, is that where the X-ray emission from a source is obviously compact, we treat it all as point-like if its radial profile is consistent with a nominal HRI PRF convolved with a Gaussian of FWHM a few arcseconds. If the statistics are too poor to make this judgement, we use the total measured counts as an upper limit on the counts in a possible point-like component. Where there is emission which is unambiguously extended we attempt to separate out a compact, possibly AGN-associated, component. Our best approach, appropriate where the data are roughly radially symmetrical, is to fit models convolved with the instrumental PRF to the radial profile of the emission. $`\beta `$-models (e.g. Sarazin 1986) are physically appropriate for extended emission originating from hot gas in hydrostatic equilibrium, and provide a useful way of characterising extended emission even when the origins of the emission are not known. In all such fits the parameter $`\beta `$ is allowed to take the values 0.35, 0.5, 0.667 and 0.9, while core radius is varied over a wide range, and the central normalisations of both the $`\beta `$ model and an unresolved central source are allowed to vary, so that the counts associated with the best-fitting components need not sum exactly to the total counts obtained directly from the data. We calculate counts in the best-fitting $`\beta `$ model from the integral of the model over the source and background regions. In radial-profile fitting we are hampered, in the case of HRI data, by our inadequate knowledge of the PRF for a particular observation because of aspect-smearing effects. Some techniques exist to circumvent this problem. Where a bright nearby pointlike source is available in the image, it is possible to use it as a template to determine the PRF applicable to the observation (e.g. Hardcastle, Lawrence & Worrall 1998d), but such sources are rare. Another possible technique is ‘dewobbling’ the source (Harris et al. 1998b). The aspect errors in HRI observations are thought to arise from pixel-to-pixel gain variations in the aspect camera CCD. The spacecraft wobble exacerbates this problem because it causes the aspect stars to move across the CCD on short time-scales. The ‘dewobble’ technique relies on binning the data according to the phase of the wobble, and then recentroiding to correct for aspect errors; so long as the satellite roll angle and the properties of the CCD are constant, the aspect error should be a function of the wobble phase only. Harris et al. provide iraf/PROS scripts to perform this task. The main disadvantage of the method is that it is applicable only to sources with a relatively bright unresolved component (of order 0.1 counts s<sup>-1</sup>); this rules out its application to many of the sources in the sample. In general we expect sources that have been dewobbled to show some deviation from the nominal PRF on small scales; the procedure of stacking by centroid should tend to produce a profile that is more sharply peaked than the true PRF, as can easily be understood by considering the limit of a single photon in each bin, but on the other hand errors in centroiding and restacking will broaden the profile. Where HRI observations are clearly dominated by a point-like component (e.g. quasars) but may also contain some extended emission, we sometimes use the central regions of the (dewobbled) radial profile of the source itself to determine a broadened PRF for model fitting; the procedure here is to fit the central part of the radial profile with the nominal HRI PRF convolved with a Gaussian, assuming that the central source is pointlike. All cases where dewobbling or a broadened PRF have been applied are discussed individually in section 3. Where neither method is applicable we assume that the nominal PRF (David et al. 1997) applies. An error in the automatic (SASS) processing of HRI observations has recently been discovered (Harris 1999). This introduces an additional aspect uncertainty into HRI data taken before 1997 January 17. We have applied the SAO-provided scripts which approximately correct this problem to all affected HRI observations in which we have carried out radial profile fits. In general the scripts make a small difference to the radial profiles of our target sources, most noticeable in those sources, mainly quasars, which are dominated by a central point-source component. In no case does the correction cause a previously existing extended component to disappear. The PRF of the PSPC is sufficiently broad that it is not expected to be significantly affected by aspect problems, and PSPC data are not affected by the SASS processing error. When fitting models to data taken with this instrument, we determine an energy-dependent PRF from the data in the energy band 0.2–1.9 keV (where the PRF is well known). For consistency of presentation, the results of model fitting are scaled to the full 0.1–2.4 keV passband of ROSAT using an appropriate spectral model. ## 3 Notes on individual sources In this section we comment on any features of particular interest in the X-ray observations, and discuss our analysis of complicated objects. Where we have separated a source into extended and compact components, values for the best-fitting parameters of the $`\beta `$-model and point source are listed in Table 3. As well as the core radius and $`\beta `$ for the best-fitting $`\beta `$-model, we tabulate the central surface brightness $`b_0`$. This quantity may be converted to an estimate of the central proton density of the X-ray emitting gas, $`n_{p0}`$. From Birkinshaw & Worrall (1993), eq. 10, to within a factor of 2 $$n_{p0}^22\times 10^9\frac{(1+z)^6b_0f}{D_L\theta _c}$$ where $`b_0`$ is in counts s<sup>-1</sup> arcsec<sup>-2</sup>, $`\theta _c`$ is the core radius in arcsec and $`D_L`$ is the luminosity distance to the source in Gpc. $`f`$, the factor relating count rate to distance-normalised emission measure. $`f`$ has a relatively weak dependence on spectral shape, since the passband of ROSAT is broad. Within a factor of about 1.5, for temperatures between 1 and 5 keV, $`f`$ has the value $`4.8\times 10^{22}`$ m<sup>-5</sup> s count<sup>-1</sup> for HRI data and $`1.5\times 10^{22}`$ m<sup>-5</sup> s count<sup>-1</sup> for the PSPC. The central density can be used to calculate the central cooling time, using the relation given by Sarazin (1986): $$t_{\mathrm{cool}}3\times 10^{13}n_{p0}^1\sqrt{kT}$$ where $`kT`$ is in keV and $`t_{\mathrm{cool}}`$ in years. Where we quote a central density or cooling time in the present paper, we have calculated the value of $`f`$ appropriate for the source’s redshift, $`N_H`$ and estimated temperature. ### 3.1 3C 28 This low-excitation radio galaxy lies in the double X-ray cluster Abell 115 (Feretti et al. 1984; Worrall, Birkinshaw & Cameron 1995). It has a relaxed double radio structure with no detected radio core, although thin radio filaments emerging from the host galaxy may suggest that it is not an inactive object (Leahy, Bridle & Strom 1998). 3C 28 lies in the northern component of the X-ray cluster, with its lobes straddling (within the ROSAT pointing accuracy) the peak in the X-ray surface brightness. Radial profile fitting shows the source to be significantly extended (Table 3). A point-like component is not significant on an F-test. It seems likely that most of the emission from the central peak originates in the central regions of a cooling flow, as suggested by Feretti et al. (1984). We therefore take the best-fit point-like component as an upper limit on the AGN contribution. ### 3.2 3C 31 Our results for this FRI radio galaxy (NGC 383) are consistent with Trussoni et al. (1997). PSPC images show extended emission, but the HRI detects only point-like emission, with a component coincident with 3C 31’s nucleus. Emission coincident with NGC 380 and NGC 379, other members of 3C 31’s group, is also detected. ### 3.3 3C 33 Results for this low-redshift FRII radio galaxy are taken from Hardcastle, Birkinshaw & Worrall (1998a). The source is unresolved with the HRI. ### 3.4 3C 48 The HRI data for this compact steep-spectrum quasar seem to show a strong E-W elongation, but this is mainly contributed by a single observation interval (OBI). The observations of 3C 48 consist of 7 OBIs taken over the course of three days at three distinct nominal satellite roll angles. We first split the data into these three ‘stable roll angle intervals’ and then applied the dewobbling procedure. The result was a clear reduction in the east-west elongation of the data. The radial profile of the source was still not well fit with a nominal HRI PRF; there were large contributions to the fitting statistic both from the inner bins, on scales of a few arcseconds, and from the bins on scales of 20–50 arcsec. We therefore used the central part of the radial profile as a template for the PRF. This left an excess component of extended X-ray emission, on scales of tens of arcsec, which we interpret as emission from hot gas; results are given in Table 3. PSPC data are adequately modelled as a point source but not inconsistent with the HRI-derived model. For any reasonable temperature assumption, the bright, compact extended X-ray emission implies rapid cooling with mass deposition rates of several hundreds of solar masses per year. This would tend to support the cooling-flow model of Fabian et al. (1987) for the optical emission-line regions. The X-ray emission from 3C 48 is discussed in more detail elsewhere (Worrall et al., in preparation). ### 3.5 3C 61.1 A faint point source is detected close to the pointing centre of the HRI observation, but it is $`30`$ arcsec from the core of the FRII radio galaxy (Laing & Riley, in prep.), a larger offset than would be expected if the sources were associated (see Fig. 1). The X-ray source appears to be almost coincident with a point-like optical object in the field, about 30 arcsec to the east of 3C 61.1’s host galaxy (see the plates of Gunn et al. 1981), which is presumably a background active nucleus. We therefore derive an upper limit on the X-ray counts associated with the radio position. ### 3.6 3C 84 This source (Perseus A, NGC 1275) lies in the centre of the Perseus cluster, and in a strong cooling flow. Remarkable shell-like structure is apparent in the HRI image, which Böhringer et al. (1993) relate to displacement of thermally emitting gas by the radio lobes. Consequently it is hard to measure a good count rate for the central component; radial fitting is not appropriate. The counts we associate with the unresolved source are measured from a 15-arcsec radius source circle and a 15–25-arcsec background annulus in the centre, rather than by fitting to the data. For consistency with the other objects, the counts we tabulate are corrected to our standard 1-arcmin source circle and 1–2-arcmin background region. The total counts listed are for a 7-arcmin source circle and 7-8 arcmin background region. ### 3.7 3C 98 This nearby FRII galaxy shows some extended X-ray emission on scales of tens of arcseconds, in addition to a compact central source (Table 3). ### 3.8 3C 123 This peculiar FRII radio galaxy is dominated by an X-ray extended component. Radial profiling does not require a core component, but the X-ray distribution is complex. Since radial profiling may not be appropriate, we derive an upper limit on the core contribution from a 15-arcsec source circle around the brightest X-ray peak. The counts we tabulate are corrected to our standard 1-arcmin source circle and 1–2-arcmin background region. We will discuss this object in more detail elsewhere (Hardcastle et al., in preparation). ### 3.9 3C 212 This quasar is unresolved with the PSPC. Our count rate is consistent with that measured from the same data by Elvis et al. (1994). ### 3.10 3C 215 This object is a well-studied lobe-dominated quasar. Both the co-added HRI image used in our analysis and, to a lesser extent, the individual datasets from which we generated it show elongation of the X-rays in a NNW-SSE direction; the weak nearby point sources in the field also show elongation roughly in this direction, so we attribute the elongation to aspect-smearing problems. The long observation contains many different OBIs and satellite roll angles, and to apply dewobbling it was necessary to drop some of the shorter intervals with insufficient counts for good centroiding, leaving 65 ks, or 75 per cent of the available data. The restacked image is indeed narrower and without significant elongation. Radial profile fits, using a PRF based on the inner regions of the source, show the dewobbled source to have a significant extended component (Table 3). Galaxy counts around 3C 215 also suggest a rich cluster environment (Ellingson, Yee & Green 1991) and some extended line emission is seen (Crawford & Fabian 1989) so it seems likely that we are seeing emission from hot cluster gas, perhaps influencing the distorted structure of the radio source (Bridle et al. 1994). The core radius corresponds to 140 kpc and the cooling time of the gas will be less than $`10^{10}`$ years for $`kT4`$ keV. ### 3.11 3C 219 This broad-line radio galaxy (Laing et al. 1994) shows largely point-like emission in the short archival PSPC observation, but there is some evidence of excess counts in the radial profile on scales of 40–100 arcsec. This may be related to the poor optical cluster known to be associated with 3C 219 (Macdonald, Kenderdine & Neville 1968), though Brunetti et al. (1998), using HRI observations not yet in the public archive, suggest instead that the extension is inverse-Compton emission associated with the radio lobes, while not ruling out a thermal origin for some or all of it. A point-source fit to the PSPC radial profile has $`\chi ^2=17.3`$ with 9 degrees of freedom, a barely acceptable fit, while the addition of the $`\beta `$-model (Table 3) improves the fit significantly. The point-like component dominates, contributing $`75\pm 7`$ per cent of the counts in the 2-arcmin source circle. ### 3.12 3C 220.1 Our results for this FRII radio galaxy are taken from Hardcastle, Lawrence & Worrall (1998d). The source is decomposed into point-like and extended components (Table 3). The counts allocated to the point source here neglect any possible contribution from a cooling flow. ### 3.13 3C 220.3 Our results for this undetected FRII radio galaxy are taken from Worrall et al. (1994). ### 3.14 4C 73.08 This low-excitation FRII source is in a crowded X-ray field. The optical position is coincident with a weak extended X-ray source, but there are too few counts for adequate radial profile fitting. We take the counts in a 2-arcmin source circle as an upper limit to point-like emission. ### 3.15 3C 254 The radial profile of the PSPC observation of this quasar is poorly fit by a point-source model, with positive residuals on scales between 30 and 60 arcsec. The HRI observation detects a weak secondary source 30 arcsec to the E, unidentified on Digitized Sky Survey (DSS)<sup>1</sup><sup>1</sup>1Available on the World Wide Web at URL: $`<`$http://stdatu.stsci.edu/dss/$`>`$ plates, and too close to the quasar to be resolved by the PSPC. The secondary source is not responsible for all of the extension seen in either the HRI or PSPC data; it may, in any case, be a clump in extended thermal emission. There is optical evidence that 3C 254 lies in a cluster of galaxies (Bremer 1997), and there is no evidence in the HRI images for X-ray emission from the foreground galaxy to the NE of the quasar reported by Bremer (1997). We fit to the radial profile of the PSPC observations, where there are more counts and so better statistics; dewobbling has no effect on the results. The results are tabulated in Table 3. The counts in the extended component indicate a luminous cluster; no cooling flow is required unless $`kT2.5`$ keV, but we cannot rule out a contribution to the point-source component from unresolved cooling gas. If we assume $`kT=7.9`$ keV, then the rest-frame 0.3–3.0 keV luminosity of the extended component is $`10^{38}`$ W, consistent with the estimate of Crawford & Vanderreist (1997) who examine the spectrum of the same PSPC data. Fitting to the radial profile of the dewobbled HRI data gives a smaller and somewhat less luminous extended component, but the results are consistent within the uncertainties on the fitted parameters. ### 3.16 3C 263 After dewobbling the long HRI observation of this quasar, known to lie in a cluster of Abell richness $`1`$, we find no significant evidence of extended emission. This is consistent with the results of an earlier study using the same data by Hall et al. (1995). ### 3.17 3C 263.1 This FRII radio galaxy was not detected by Crawford & Fabian (1996a) in a far-off-axis PSPC observation. However, an HRI observation detects a source within 6 arcsec of the optical position of 3C 263.1, with a count rate consistent with the upper limit of Crawford & Fabian. There is no obvious background source on the optical plates of Laing et al. (1978) that might be responsible for the X-ray emission. The HRI emission is compact and adequately fitted by a broadened PRF. The source is surprisingly bright if it is due only to nuclear X-ray emission, given the weak flux of the radio core (Liu, Pooley & Riley 1992). Up to $`60`$ per cent of the X-ray flux may be contributed by a cluster of comparable size to that seen in 3C 220.1, but the data do not require this, so for consistency we treat the source as pointlike. ### 3.18 3C 264 This narrow-angle-tail radio galaxy lies $`7`$ arcmin away from the X-ray centroid of the cluster Abell 1367, and 8.4 arcmin off-axis in the PSPC observation. We use the parametrised off-axis PSPC PSF (Hasinger et al. 1995) to perform a radial-profile fit in a standard 2-arcmin source region, which excludes most of the emission from the cluster. The results (Table 3) suggest that there is some extended emission associated with the radio source, in addition to a strong point-like component. Our result is consistent with the count rate quoted by Edge & Röttgering (1995). ### 3.19 3C 270.1 50 arcsec to the NW of this high-redshift quasar there is a second compact source, associated with a faint compact source on DSS plates and so probably another AGN. Our counts are taken from a 30-arcsec source circle around the quasar, corrected to our standard 2-arcmin source circle and 2–3-arcmin background annulus. We are unable to say whether the apparent extended emission around 3C 270.1 is related to our target source or the other object. ### 3.20 3C 272.1 This nearby FRI source (M84) is extended in the X-ray. Our results are taken from a fit of point source plus $`\beta `$ model (Table 3). ### 3.21 Abell 1552 The cluster is detected in the short PSPC observation, but no point-like excess is found at the optical position of the radio galaxy. ### 3.22 3C 274 (M87) Harris, Biretta & Junor (1997) present evidence that the X-ray core of M87 varies by $`20`$ per cent on time-scales of years. We use their core count rate for June 1995. The total counts tabulated are for the source region they define, a 276-arcsec radius circle with background taken from a 280–300 arcsec annulus, both centred around the X-ray core, and our results are consistent with theirs. ### 3.23 3C 275.1 Both HRI and PSPC data are available for this quasar. The PSPC data are confused by weak extended emission from the nearby spiral galaxy NGC 4651; when this is removed, the data show a point-like source. The longer HRI observation shows a profile slightly broader than the nominal PRF, characteristic of a source affected by aspect smearing. When the profile is fitted with a broadened PRF, there is some indication of additional extension on scales of 20–60 arcsec. Using the broadened PSF, we fit an additional $`\beta `$-model component, with results tabulated in Table 3. As in 3C 215, the galaxy counts of Ellingson et al. (1991) imply a rich cluster environment for this source, so it is likely that the extended emission we detect is real and due to hot intracluster gas. Hinten & Romanishin (1986) suggest that the extended emission-line region around the source is due to a cooling flow, but the best-fitting $`\beta `$-model does not require cooling. ### 3.24 3C 277.2 Our count rate for this faint FRII radio galaxy (taken from a 30-arcsec source circle and corrected to our standard 1-arcmin source region) is consistent with that of Crawford & Fabian (1995). ### 3.25 3C 280 Worrall et al. (1994) describe the separation of the PSPC emission from this high-redshift FRII radio galaxy into compact and extended components. The archival HRI data are consistent with such a decomposition (cf. Dickinson et al. 1998), and so we use the Worrall et al. results here (Table 3). ### 3.26 3C 288 Because this radio galaxy is 21 arcmin off-axis in the PSPC observation, uncertainties in the PSPC PRF at this radius mean that we have not been able to separate extended and compact emission, although the source is at least partially extended. 3C 288 appears in the optical to be the dominant member of a cluster (Wyndham 1966) and so it is not surprising that extended X-ray structure is seen. Interactions with the cluster gas may be the reason for its unusual structure in the radio (Bridle et al. 1989); it is one of a small class of reasonably powerful, distorted double sources with strong jets but weak hotspots, which often have a low-excitation optical spectrum and an association with a cluster \[other good examples are 3C 401 and 3C 438, Hardcastle et al. (1997); the class is similar in radio properties to the ‘jetted doubles’ of Law-Green et al. (1995)\]. The available X-ray data are unfortunately not good enough to test whether 3C 288 lies in a cooling flow, a suggested origin for this type of source. ### 3.27 3C 294 This high-redshift radio galaxy was detected in PSPC observations by Crawford & Fabian (1996a). The counts we measure from a long archival HRI observation are consistent with the result of Crawford & Fabian within the errors. The source appears extended, but the low count rate prevents us from extracting a reliable radial profile. Dickinson et al. (1998) reach similar conclusions. ### 3.28 3C 295 This radio galaxy is known to lie in an optical cluster (e.g. Yates, Miller & Peacock 1989) and to have extended X-ray emission (Henry & Hendriksen 1986). Our best-fitting characterisation of the HRI emission with a $`\beta `$-model and point source is tabulated in Table 3. Acceptable fits are found with no point-source contribution. Our values for core radius and $`\beta `$ are consistent with the findings of Henry & Hendriksen. The point-source contribution is much lower than their adopted value (which would correspond to approximately 115 counts in the HRI observation, on the assumption of a power-law spectrum and galactic absorption) but agrees within the errors. Our best-fitting $`\beta `$ model is also consistent with Neumann (1999). The best-fitting model implies rapid cooling, with a mass deposition rate of some hundreds of solar masses per year, consistent with the picture of Henry & Hendriksen (1986), and the cooling time at the core radius is less than the Hubble time, so that the isothermal $`\beta `$-model is not physically consistent. When we fit the data with the most recent versions of the cooling-flow models used by Hardcastle et al. (1998d), we find that the dependence of pressure on radius must be weaker than the $`pr^1`$ relation used in our earlier work, because the HRI emission is not strongly centrally peaked; for example, acceptable fits are obtained for a model with $`pr^{0.5}`$, $`\rho r^{0.75}`$, $`Tr^{0.25}`$, with the cooling time at the core radius being $`2\times 10^9`$ yr. This fit is not as good as the simple $`\beta `$ model together with a weak point source, and the cooling time is rather short, but it fits the data without the need for a central radio-related contribution. For consistency with our other analyses, we use the point-source component derived from $`\beta `$-model fitting as our best estimate of the unresolved contribution to the X-ray emission. ### 3.29 3C 303 This broad-line radio galaxy is a strong point source in the X-ray. In the HRI image a faint point-like component, containing $`25\pm 6`$ counts, is approximately coincident (after aligning the radio core and X-ray centroid) with the western hotspot of 3C 303. We may therefore be seeing an X-ray counterpart to the radio and optical hotspots already known in this object (e.g. Meisenheimer, Yates & Röser 1997). If so, this would be one of only a very few X-ray hotspots known. \[The other examples are the hotspots of Cygnus A (Harris et al. 1994) where the emission mechanism is likely to be synchrotron self-Compton scattering, the western hotspot of Pictor A (Röser & Meisenheimer 1987, and unpublished ROSAT data) where the emission mechanism is not clear, and the northern hot spot of 3C 390.3 (Prieto 1997, and see below) which is interpreted as a synchrotron X-ray source.\] However, the position of this X-ray component, given the low count rate, is at least equally consistent with that of a $`z=1.57`$ quasar 4 arcsec to the SW of the optical source identified with the hot spot (Kronberg 1976; Kronberg et al. 1977), and it seems most likely that the X-ray emission comes from this quasar (see Fig. 2). If the X-ray emission is related to the hot spot, then its 1-keV flux of 7.6 nJy means that it is much too bright \[using the code of Hardcastle, Birkinshaw & Worrall (1998a)\] to be synchrotron self-Compton emission unless the source is very far from equipartition. It could be synchrotron emission if the electron energy distribution is a steep power law (electron energy index of $`2.8`$, corresponding to the observed radio and radio-optical spectral indices of $`0.9`$) and extends to very high energies, with a high-energy cutoff near the region corresponding to X-ray emission. ### 3.30 3C 309.1 In spite of the suggestion by Forbes et al. (1990), based on the presence of a luminous extended emission-line region, that this quasar lies in a massive cooling flow, its X-ray emission is well fitted by a PSPC PRF. As for 3C 254, we cannot rule out the possibility of a cooling-flow contribution to the point-source flux, which would need to be present on smaller scales than the PSPC resolution (i.e. about 100 kpc at the redshift of 3C 309.1). ### 3.31 3C 310 This relaxed double radio galaxy (van Breugel & Fomalont 1984) shows clear extended X-ray emission, first detected with Einstein by Burns, Gregory & Holman (1981). The HRI image shows the X-ray emission to be centred on the radio galaxy, with some evidence for east-west extension in the very central regions; because some other sources in the field seem to be extended in a similar direction, we are inclined to treat this as an aspect-induced artefact. When we characterise the emission with a $`\beta `$ model and point source (Table 3), $`\beta `$ is not well constrained but large core radii (between 40 arcsec for $`\beta =0.35`$ and 140 arcsec for $`\beta =0.9`$) provide the best fit to the data. No model without a point-source contribution is a good fit. ### 3.32 3C 324 We find a very weak extended source close to the catalogued position in the HRI observation of this high-redshift radio galaxy. Our count rates for this source are consistent with those of Crawford & Fabian (1996a) and Dickinson et al. (1998). ### 3.33 NGC 6109 (4C 35.40) Our results for this FRI radio galaxy are consistent with those of Feretti et al. (1995), who find a compact source. Radial-profile fitting suggests that some of the emission is extended (Table 3). ### 3.34 3C 334 This quasar appears point-like in a PSPC observation. A long HRI observation shows some extension in a NW-SE direction but, as this is in the same direction as a ‘hot line’ at the edge of the image, it seems plausible that much of the extension is due to aspect smearing. However, when the source is dewobbled, radial profiling reveals some residual extension on scales of tens of arcseconds. The best-fitting $`\beta `$ model (Table 3) has significant extended emission. Such a model is consistent with the PSPC radial profile, although the total count rate in the HRI observation is considerably higher than would be inferred from the PSPC data, suggesting variability. There is no evidence in the galaxy counts of Yee & Green (1987) that 3C 334 lies in a particularly rich environment, but Crawford & Fabian (1989) show it to have some extended optical line emission and Hintzen (1984) presents an optical plate with a number of nearby galaxies, including an apparent close companion. The compact nature of the extended emission means that a cooling flow is required unless the temperature is very high. ### 3.35 3C 338 This radio galaxy, located in the cluster A2199, shows a complex and asymmetrical X-ray structure (Owen & Eilek 1998) and is also unusual in the radio. Radial profiling is clearly not particularly appropriate, but a simple application to the inner regions of the source gives the decomposition tabulated here, which represents our best estimate of the core contribution. ### 3.36 NGC 6251 This FRI radio galaxy shows both extended and compact emission in the PSPC observation (Birkinshaw & Worrall 1993; Worrall & Birkinshaw 1994). The HRI observation detects mainly the compact core, but there is also a weak but significant detection of a component $`25`$ arcsec to the NW of the core, coincident with a bright knot in the base of the jet (Perley, Bridle & Willis 1984). \[This component is not to be confused with the possible jet-related X-ray emission claimed by Mack, Kerp & Klein (1997), which is at much larger distances from the nucleus and is not visible on the HRI image.\] The knot contributes only $`10\pm 5`$ counts to the total. The number of counts in the unresolved component is consistent with the results of the spatial decomposition of Worrall & Birkinshaw (1994), a result insensitive to the spectrum of the source. We cannot rule out a thermal contribution to the unresolved component, but it must be compact. ### 3.37 3C 346 This narrow-line radio galaxy is clearly detected with the PSPC, showing both extended and compact X-ray emission; fitting a PSPC PRF alone to the radial profile gives unacceptable results. In the radio, 3C 346 is another ‘jetted double’ source with a strong one-sided jet and a weak hot spot (compare 3C 288, section 3.26). It has an unusually bright core (Spencer et al. 1991) and an optical jet (e.g. de Koff et al. 1996) and exhibits a strong Laing-Garrington effect<sup>2</sup><sup>2</sup>2Laing (1988); Garrington et al. (1988). (Akujor & Garrington 1995) which suggests a cluster environment for the source. $`\beta `$-model fits are tabulated in Table 3. Most of the counts (70 per cent) are contributed by the point-source component; no model without a point-source contribution can adequately fit the data. A cooling flow is not required. ### 3.38 3C 345 This superluminal quasar is known to be variable in both the radio and X-ray (e.g. Unwin et al. 1997). We use a single PSPC observation, taken between 1993 Mar 06 and 1993 Mar 11, as representative (since the available HRI observations are all short). The data are acceptably fit with a point-source model. ### 3.39 3C 356 This high-redshift radio galaxy was observed with the ROSAT PSPC and HRI by Crawford & Fabian (1993, 1996b). We confirm the existence of a very weak source in the HRI observation, and our measurements are consistent with theirs within the errors. ### 3.40 4C16.49 This steep-spectrum quasar is only barely detected. It appears elongated in the HRI observation, but in the same direction as other objects in the field, and so is likely to be aspect-smeared. We treat it as unresolved. ### 3.41 3C 368 Our results for this high-redshift radio galaxy are consistent with those of Crawford & Fabian (1995). ### 3.42 3C 388 The X-ray images of this low-excitation FRII radio galaxy show a significant amount of diffuse emission. The source is known to lie in a poor cluster, with a number of nearby galaxies (Prestage & Peacock 1988). In the radio the object has an unusual structure. Its one-sided bright jet and diffuse hot spots might put it in the class of jetted doubles discussed earlier (section 3.26), but the presence of two distinct spectral regimes in its lobes suggests that it may be a restarting source (Roettiger et al. 1994). $`\beta `$-model fits are tabulated in Table 3 and show the X-ray emission to be dominated by an extended component. No model without a point source is an acceptable fit to the data, however. Gas of the density implied by the fitted $`\beta `$ model should be rapidly cooling, with a mass deposition rate of $`30M_{}`$ yr<sup>-1</sup>. When we fit with cooling-flow models, using a temperature $`T3`$ keV outside the cooling radius, the lack of a strong peak in the central regions (as with 3C 295) means that no model with $`pr^1`$ is a good fit; we find acceptable fits for models with a weaker dependence of $`p`$ upon $`r`$, for example with $`pr^{0.5}`$, $`\rho r^{0.75}`$, $`Tr^{0.25}`$, $`\beta =0.5`$, $`r_{\mathrm{core}}=13`$ arcsec. The cooling radius for this model is 9 arcsec and the cooling time at this radius is about $`5\times 10^9`$ yr. A point-like nuclear component is still required, containing $`66\pm 19`$ counts. ### 3.43 3C 390.3 Leighly et al. (1997) discuss 90 separate HRI observations of this broad-line radio galaxy in the context of a variability study; the source is shown to be highly variable on time-scales of weeks. We adopt their average value as our count rate. As discussed by Harris, Leighly & Leahy (1998a), the radial profile of the stacked HRI observations is broader than would be expected from the nominal HRI PRF, but its width is within the range observed to be a result of aspect uncertainty, and so we treat the source as point-like \[with the exception of the X-ray hot spot discussed by Prieto (1997) and Harris et al.\]. ### 3.44 3C 442A This peculiar radio galaxy (Comins & Owen 1991) is associated with NGC 7237 in the merging galaxy pair NGC 7236/7237, which lies in a poor cluster. The small-scale structure in the HRI image seems to suggest a double X-ray core with separation roughly 20 arcsec in an east-west direction. Such structure is unreliable because of aspect problems, but the large extension makes an aspect-related explanation unlikely. Other weak sources nearby are not doubled up in the same way, including one apparently coincident with the small associated galaxy to the SE. No compact source is found associated with NGC 7236. On larger scales, the map smoothed with a Gaussian of $`\sigma =16`$ arcsec seems to show a bar of X-ray emission, oriented roughly NNW-SSE (see Fig. 3); this lies in approximately the same position angle as the common optical envelope of the galaxy pair (e.g. Borne & Hoessel 1988) and approximately perpendicular to the position angle of the radio source, contrary to the prediction of Comins & Owen (1991) who expected the hot gas to be aligned with the radio source. An X-ray counterpart to one of the nearby radio point sources (‘E’ in the notation of Comins & Owen) is also detected, but source ‘A’ is not seen. Because of the clear absence of radial symmetry in this source, radial profiling is not suitable. An upper limit is found for the point-like counts by taking a 10-arcsec source circle about the X-ray component coincident with the steep-spectrum radio core, using a background region between 40 and 60 arcsec away; we correct the result to our standard 1-arcmin source region. Total counts are derived from a 2.5-arcmin source circle. ### 3.45 3C 449 The PSPC observations of this FRI radio galaxy are reported in Hardcastle et al. (1998c). The structure of the extended emission is poorly constrained by a more recent 19-ks HRI observation, but no point sources are detected which might have conspired to produce the ring-like structure seen around the southern radio lobe in the lower resolution PSPC observations, reinforcing the conclusion that this is caused by a displacement of the hot cluster gas by the radio-emitting plasma. The HRI and PSPC radial profile fits are consistent, and so we tabulate the PSPC results in Table 3, as the statistics are slightly better. ### 3.46 3C 454.3 This is a well-known highly variable core-dominated quasar. We report results from PSPC observations since the HRI exposures are all short. The count rates in a 2-arcmin source circle are significantly different in three separate PSPC observations, and so we have used the longest (from 1992 Dec 16 – 1992 Dec 18), which gives a count rate consistent with that of Sambruna (1997). We find no evidence for intraday variability or extended X-ray emission. ### 3.47 3C 465 Sakelliou & Merrifield (1998) present HRI data which show a point-like X-ray component coincident with the host galaxy of this wide-angle tail radio galaxy. After excluding X-ray emission from the northern companion galaxy of the AGN host and realigning the HRI data from the two observation epochs as suggested by Sakelliou & Merrifield, we find the radial profile of the central source is moderately well fit by a broadened PRF; we take a 1-arcmin source circle and background between 1.7 and 4 arcmin, excluding confusing sources. Further out, there is clear evidence for extended emission on arcminute scales. ## 4 Results ### 4.1 Demographics Table 4 shows the numbers of 3CRR sources observed and detected in ROSAT pointed observations, broken down by source class. Just over half of the 3CRR sample has been observed. FRII radio galaxies, which make up 60 per cent of the sample, have only 46 per cent of the observations but are consistent with the overall detection fraction of 79 per cent in giving 37 per cent of the detections. The observed subsample is biased with respect to the original (178-MHz flux density) selection criterion of 3CRR – the sources that have been observed with ROSAT tend to be brighter than those that have not, largely because the sources with 178-MHz fluxes much larger than the flux limit of the sample tend to be low-redshift objects and so more suitable for observation with ROSAT. However, dividing the observed subsample by source class, we find that the bias is only present in the FRII radio galaxies. The subsamples of FRI radio galaxies and of quasars are unbiased, in the sense of having 178-MHz radio fluxes that are consistent with having been drawn at random from the objects of their class in the 3CRR sample. We can treat at least these subsamples as representative of the 3CRR sample as a whole. ### 4.2 The radio core – X-ray core correlation In Table 5 we list for each source the total counts found in the source region and the number (or upper limit) that we associate with an unresolved central source. The point-like count rate is converted into a 1-keV flux density and rest-frame spectral luminosity density<sup>3</sup><sup>3</sup>3The X-ray luminosity density of a typical galaxy due to intragalactic gas, LMXB and stars, which will be unresolved in most of our objects, corresponds to only a few $`\times 10^{15}`$ W Hz<sup>-1</sup> sr<sup>-1</sup> at 1 keV (see e.g. Irwin & Sarazin 1998), and so most of the luminosity of almost all our sources must arise elsewhere; similarly, all our sources (except M84) have nuclear luminosities significantly higher than those of the compact nuclear sources in nearby normal galaxies studied by Colbert & Mushotzky (1998). by assuming a power-law spectrum (with $`\alpha _X=0.8`$) and galactic absorption. We also tabulate 5-GHz radio core flux density and rest-frame spectral luminosity density, assuming $`\alpha _R=0`$. For consistency we have used the arcsecond-scale rather than milliarcsecond-scale core fluxes even where both are available; in the objects (largely quasars) where good information is available the milliarcsecond- to arcsecond-scale core flux density ratio is not very different from unity. Fig. 4 shows histograms of the rest-frame radio to core-X-ray two-point spectral indices of sources with detections in the radio. A strong correlation is apparent, in both flux-flux and luminosity-luminosity plots (Figs 5 and 6) between the X-ray and radio core strengths. Analysis with the survival analysis package asurv \[Rev. 1.1; LaValley, Isobe & Feigelson (1992)\] taking account of the censoring of the data, using a modified Kendall’s $`\tau `$ algorithm, shows that the flux-flux and luminosity-luminosity correlations are significant, both for the sample as a whole and for the different classes of source (Tables 7 and 8, columns 3 and 4). The fact that the flux-flux correlation is significant gives us confidence that we are not simply seeing a redshift-induced artefact in the stronger luminosity-luminosity correlation. Akritas & Siebert (1995) present a procedure which allows partial correlation analysis in the presence of censoring, and we apply it to the distributions of radio luminosity, X-ray luminosity and redshift to make this conclusion quantitative; the correlation for the sample as a whole, and for the radio galaxies and quasars separately, is significant at better than the 95 per cent confidence level even given the effects of redshift. These results imply a physical relationship between the X-ray and radio cores of the radio sources, confirming work done with other samples and instruments on low-luminosity radio galaxies (Fabbiano et al. 1984; Edge & Röttgering 1995; Siebert et al. 1996; Canosa et al. 1999), high-luminosity radio galaxies (Worrall et al. 1994; Hardcastle et al. 1998d) and quasars (Tananbaum et al. 1983; Kembhavi, Feigelson & Singh 1986; Worrall et al. 1987; Browne & Murphy 1987; Baker, Hunstead & Brinkmann 1995; Siebert et al. 1996). What does this imply for the origin of the X-ray emission? There is much evidence for relativistic motion in the cores of all radio-loud AGN, with Lorentz factors $`\gamma 5`$ inferred from superluminal motion and unification arguments. The observed fluxes and luminosities of radio cores are therefore strongly influenced by Doppler beaming. For example, 20 of the 22 FRI radio galaxies in our sample, which have broadly similar total radio luminosities, have a distribution of core prominences (defined in this case as the ratio of 5-GHz core flux density to 178-MHz total flux density) spanning two to three orders of magnitude and consistent with having been drawn from a randomly oriented population of objects with a single intrinsic core prominence and $`\gamma =5`$. (The outliers are the peculiar objects 3C 28 and 3C 442A, where nuclear activity may have ceased). If the X-ray luminosity ($`L_X`$) were simply related to the AGN power, and therefore independent of beaming and orientation, we would not expect the core radio and X-ray emission to be strongly correlated. At best there would be two to three orders of magnitude of scatter in the correlation, and in practice for 20 objects of similar unbeamed luminosity we would simply not see the correlation at all. The very existence of a correlation between $`L_X`$ and $`L_R`$ in these radio galaxies forces us to the conclusion that the soft-X-ray emission originates in a Doppler-boosted region with a Lorentz factor similar to that of the jet. The simplest model is one in which the X-ray emission originates in the jet itself. To make this argument more quantitative, and to extend it to the higher-luminosity radio galaxies where there are fewer clear detections of nuclear X-ray emission, it is useful to characterize the X-ray/radio correlation by linear regression. Linear regression is difficult in this dataset and its subsets because not all objects have detected cores either in the radio or the X-ray wavebands. Most available algorithms for linear regression in the presence of censored data work only with unidirectional censoring. Two well-known algorithms exist for performing linear regression on doubly censored data: that of Schmitt (1985), implemented in asurv, and the Theil-Sen estimator (Akritas, Murphy & LaValley 1995). Their application to datasets such as ours is discussed in Appendix A. Our conclusion is that the Theil-Sen estimator is slightly better, but we tabulate the results of both in Tables 7 and 8 to emphasise that similar results are obtained. In a few of the sub-samples the censoring is only unidirectional (e.g. in quasars where we have complete radio core information) and we are able to use other algorithms, such as the Buckley-James method implemented in asurv, which also give similar answers. The results are most doubtful in subsamples where upper limits dominate (e.g. the FRII radio galaxies). Table 8 shows that for the subsamples of radio galaxies the slopes of the correlation are close to unity. Only when broad- and narrow-line radio galaxies are considered together is the slope clearly different from unity, and (as can be seen from Fig. 6) this is because the broad-line objects are systematically brighter in X-rays than the narrow-line objects of similar radio core luminosity, suggesting that these objects have an additional X-ray component not seen in the narrow-line and low-excitation objects. The quasars have an X-ray-radio slope significantly flatter than unity, and generally lie about the regression line for the whole sample. This result has been seen before and can be explained in terms of a model in which both beamed radio-related X-ray emission and anisotropic unrelated X-ray emission are seen from quasar nuclei (Worrall et al. 1987; Browne & Murphy 1987; Worrall et al. 1994; Baker et al. 1995). Of particular interest are the high-power, narrow-line FRII radio galaxies, of which Cygnus A is the prototype. Obscuring tori are required by unified models in these sources. Because they are distant and faint, we have detected only 17 out of the 30 observed sources in this class, and convincingly detected a central unresolved component in only 7. However, the survival-analysis slope of the $`L_X/L_R`$ correlations in these FRIIs (Table 8) is entirely consistent with the results for the much better detected FRI radio galaxies. If we normalize the $`L_X/L_R`$ relation at a 5-GHz radio spectral luminosity density of $`10^{23}`$ W Hz<sup>-1</sup> sr<sup>-1</sup>, then the expected X-ray luminosity density at 1 keV (with 90 per cent confidence range derived from simulation) is $`1.8\times 10^{16}`$ ($`6.5\times 10^{15}`$$`4.6\times 10^{16}`$) W Hz<sup>-1</sup> sr<sup>-1</sup> for the FRI radio galaxies and $`9.5\times 10^{15}`$ ($`1.4\times 10^{15}`$$`4.6\times 10^{16}`$) W Hz<sup>-1</sup> sr<sup>-1</sup> for the FRIIs, so the normalizations of the $`L_X/L_R`$ relations for the two classes of source are consistent. This is as expected if the nuclear soft X-ray emission originates from the jet in both classes of source, since there is little difference on parsec scales between the radio properties of FRI and FRII jets (e.g. Pearson 1996). There are no objects which have a bright radio core and a weak or undetected nuclear X-ray component; therefore nothing explicitly contradicts the suggestion that there is a radio-related soft X-ray component in the centre of all radio galaxies. The slope of the radio-X-ray correlations can in principle provide information on the mechanism responsible for the radio-related X-ray emission. Possible mechanisms include synchrotron emission from the same population of electrons responsible for the radio emission, and inverse-Compton scattering by those electrons of a population of low-energy photons to X-ray energies, either the radio synchrotron photons themselves (synchrotron self-Compton emission) or an external photon population (‘external Compton’). These different mechanisms would give rise to different relationships between the intrinsic radio and X-ray luminosities. However, simulation shows that the effects of Doppler beaming will tend to obscure any relationship that there may be between the rest-frame luminosities in the two wavebands. In addition, redshift-induced effects will tend to push the slope towards unity. It is conventional to compensate for these problems by considering not the luminosity but the prominence of a beamed component, where the prominence is defined as the ratio of the component’s luminosity to an unbeamed quantity such as the low-frequency radio luminosity of the source. The slope of a plot of X-ray core prominence against radio core prominence, determined using the Theil-Sen regression, is consistent with unity for the FRI radio galaxies and the narrow-line FRII radio galaxies. This suggests that the bulk Lorentz factors of the radio- and X-ray emitting material are similar. ### 4.3 Cluster emission in high-power radio galaxies and quasars We have detected extended emission, presumably hot gas associated with a cluster of galaxies, in several quasars with redshifts between $`0.4`$ and $`0.7`$ (3C 48, 3C 215, 3C 254, 3C 275.1, 3C 334). In all these objects optical observations have previously suggested a cluster environment for the radio source. Using the low-redshift cluster temperature-luminosity relation of David et al. (1993), which where tested has been reliable out to high redshift (Mushotzky & Scharf 1997; Donahue et al. 1998), we use PROS and the observed count rates to estimate<sup>4</sup><sup>4</sup>4The procedure used is to guess a temperature, calculate a luminosity from this temperature using a redshifted Raymond-Smith model with 0.5 cosmic abundance, use the temperature-luminosity relation to calculate a new temperature from this luminosity and iterate until the temperature and luminosity stabilise. self-consistent temperatures and 2-10 keV luminosities (see Table 9). These luminosities are consistent with moderately rich cluster environments for the quasars, and also match those of radio galaxies with unambiguously extended X-ray emission at similar redshift inferred by Worrall et al. (1994), Crawford & Fabian (1996a), and Hardcastle et al. (1998d), as would be expected in unified models. Deriving useful upper limits on extended luminosity for those objects (mostly high-redshift quasars) which appear pointlike in radial profile analysis is difficult; the answer depends on the radial profile and the temperature of the undetected extended X-ray emission as well as on the source and the background count rates. If we assume that the undetected emission is similar to that in detected objects, the task is better constrained. X-ray emission from the detected $`z>0.3`$ clusters in our sample can be represented on average by a $`\beta `$ model with $`\beta =0.9`$, $`r_{\mathrm{core}}150`$ kpc, and $`kT5`$ keV. We then use Monte Carlo simulations to estimate how strong such a component would have to be to be detected at the 95 per cent significance level on an F-test in 95 per cent of the trials. For each of the 21 point-like objects with $`z>0.3`$ we determine an upper limit on extended counts in this way using a simulation matched to the on-source and background counts of the source. Where the simulations show that all the observed counts could be attributed to cluster emission, as happens in a few cases with poor statistics at high redshift, we use the total counts as an upper limit instead. From these limits on count rate we compute a limit on 2-10 keV extended luminosity assuming $`kT=5`$ keV (the results are insensitive to choice of temperature). Limits are plotted with detections in Fig. 7. The upper limits on cluster luminosity are a substantial fraction of the total source luminosities for the majority of these objects, since most were observed with the PSPC which has poor spatial resolution compared with the expected angular size of clusters. Nevertheless, limits and detections populate similar ranges of luminosity for high-redshift radio galaxies and quasars. At lower redshifts, it is known that FRI radio galaxies tend to lie in group- or cluster-scale hot gas (e.g. Worrall & Birkinshaw 1994) and results here agree with this. We cannot rule out extended X-ray emission in any of the FRIs in the sample, although objects observed only with the HRI are poorly constrained. The picture is less clear for FRII objects. Of the low-excitation objects, all seem to lie in environments that are at least comparable to those of the FRIs, and some (e.g. 3C 388) are in considerably richer environments. Of the high-excitation objects, which are more typical of their class in their radio structure, at least one (3C 219) appears to lie in an environment comparable to 3C 388’s, and some (e.g. 3C 223, 3C 284) probably have luminosities comparable to those of the cluster around the FRI 3C 449, but others (e.g. 3C 326, 4C 73.08, 3C 98) are constrained by the data to lie in environments no more luminous than those of the least luminous FRIs. It seems, therefore, that FRIIs at low redshift inhabit a range of environments, and while we cannot rule out the possibility that the environments of low-$`z`$ 3CRR FRIIs are identical to those of FRIs (although there is optical evidence against this; Prestage & Peacock, 1988), our finding that luminous X-ray environments for low-$`z`$ FRIIs are not common agrees with earlier work (e.g. Miller et al. 1985). Our data provide some weak support for the common belief (based on optical work; e.g. Yates, Miller & Peacock 1989) that there is an evolution in the environments of the most powerful radio sources between $`z=0`$ and $`z0.5`$, in the sense that they are more likely to lie in rich clusters at high redshift. On the basis of Einstein observations Miller et al. (1985) suggested that the hot-gas environments of low-$`z`$ FRIIs do not confine their radio lobes. Our limits on extended emission in the environments of FRIIs, however, suggest that lobe confinement is possible; for example, the very poor environment that we detect around 3C 98, with an assumed temperature of 1 keV, provides a pressure greater than or equal to the minimum pressure derived for 3C 98 by Miller et al. out to $`2`$ arcmin (i.e. close to the hot spot regions where ram pressure confinement is believed to be important). Part of the argument of Miller et al. was based on the observation that sources with a wide range of linear sizes have approximately the same axial ratio (defined as the ratio of lobe width to length); this would not happen if the lobes were confined while the sources continued to expand linearly. But more recent studies of radio-galaxies at low redshift do show a weak relationship between source length and axial ratio, removing some of the motivation for believing that the lobes of radio galaxies are unconfined; it seems more likely that they are unconfined initially and reach equilibrium at a later stage in a source’s lifetime. We can compare the clusters detected by X-ray observation with those inferred from observations of extended emission-line gas. Emission-line regions imply that there is cool ($`10^4`$ K) gas in the environment of the quasar or radio galaxy, and this must be in pressure equilibrium with the hot, X-ray emitting gas if the emission-line region is to last more than a sound-crossing time. Pressures can be inferred from emission-line ratios if it is assumed that the lines are excited by photoionization from the central quasar. The assumption that the emission-line regions are not overpressured with respect to the X-ray emitting gas leads to the deduction in many objects (e.g. Crawford & Fabian 1989) that the X-ray gas must be so dense that its local cooling time is less than the Hubble time. There are luminous extended emission-line regions in all the quasars with detected extended X-ray emission, although the X-ray observations do not directly require cooling in all cases. Pressures deduced from line diagnostics (\[OIII\]/\[OII\] ratios) for four of these sources are tabulated in Table 9, together with pressures at similar radii derived from the $`\beta `$-model fits to the extended X-ray emission. The pressures are consistent to within a factor of a few in all cases where data are available. After submission of the first draft of this paper, Crawford et al. (1999) completed an independent radial-profile analysis of all five of the quasars in our sample with detected extended emission. They reach broadly similar conclusions. ## 5 Conclusions We have used the spatial resolution of ROSAT to find compact central X-ray components in a large fraction ($`60`$ per cent) of the observed radio galaxies and quasars in the 3CRR sample, and we report limits in the remaining sources. It appears that the model, discussed in section 1, in which all powerful radio sources have radio-related soft X-ray emission is consistent with these observations. As argued in section 4.2, the strong correlation between the core radio and nuclear X-ray emission, given that the radio core is believed to be strongly affected by relativistic beaming, implies that the soft X-ray emission originates in the jet, rather than being related to the radio emission less directly (e.g. by common correlation with the bolometric luminosity of the AGN). A jet origin for the soft X-ray emission also explains the fact that it is seen in radio galaxies in spite of the high column densities of absorbing material inferred from hard X-ray observations of some sources. A second nuclear X-ray component, which originates much closer to the nucleus and is affected by the absorbing material so that it is not seen in radio galaxies, is necessary to explain the otherwise anomalously bright emission from lobe-dominated quasars (and from radio-quiet quasars) as discussed by Worrall et al. (1994); such a component is also likely to be present in broad-line radio galaxies. The slope of the $`L_X/L_R`$ correlation for radio galaxies, qualitatively similar to that seen in core-dominated quasars, supports unified models for radio galaxies and quasars. Extended X-ray emission from cluster- or group-scale gas is common in FRI objects in the sample. Several low-redshift FRIIs also show extended X-ray emission, but others must inhabit relatively poor environments. At higher redshifts, we have found evidence for luminous X-ray emission around several radio galaxies and lobe-dominated quasars, implying rich cluster environments for these sources. The resolution of ROSAT limits our ability to detect more quasar clusters, but it is very likely that more will be detected with the new generation of X-ray instruments. The inferred environments of radio galaxies and quasars detected at $`z0.3`$ are similar, supporting the predictions of unified models. ## Acknowledgements We thank Mark Birkinshaw for writing and supporting the software used in radial profile analysis, and for helpful comments on the paper. Fortran code implementing the Theil-Sen estimator was written by Michael LaValley and kindly provided by Michael Akritas; we thank him and Eric Feigelson for helpful discussion of the linear regression problems discussed in section 4.2 and Appendix A. Fortran code implementing the partial Kendall’s $`\tau `$ procedure of Akritas & Seibert was also provided by Michael Akritas, director of the Statistical Consulting Center for Astronomy operated at the Department of Statistics, Penn State University (URL: $`<`$http://www.stat.psu.edu/scca/homepage.html$`>`$). We are grateful to Dan Harris for a thorough and wide-ranging referee’s report on the first version of this paper. This research has made use of data obtained through the High Energy Astrophysics Science Archive Research Center Online Service, provided by the NASA/Goddard Space Flight Center. The Digitized Sky Surveys were produced at the Space Telescope Science Institute under U.S. Government grant NAG W-2166. The National Radio Astronomy Observatory is operated by Associated Universities Inc., under co-operative agreement with the National Science Foundation. This work was supported by PPARC grant GR/K98582 and NASA grant NAG5-1882. ## Appendix A Linear regression in doubly censored data Linear regression when the data are censored in both directions is a common problem in astronomy, but there are relatively few algorithms for dealing with it; most of the survival-analysis techniques used in astronomy are inherited from the biological sciences where only one variable (survival time) is likely to be censored. Schmitt (1985) presents one algorithm, which is implemented in the survival-analysis package asurv. Some of the problems with Schmitt’s algorithm are discussed in Sadler, Jenkins & Kotanyi (1989). One particular practical difficulty is that the data must be binned, and the correct procedure for choosing the number of bins is not obvious. Akritas et al. (1995) present a second algorithm, a generalised version of the Theil-Sen estimator \[which determines the slope $`\beta `$ which makes Kendall’s $`\tau `$ between the residuals ($`y_i\beta x_i`$) and $`x_i`$ approximately zero\]. Here we discuss Monte Carlo simulations intended to demonstrate which of the two procedures is best suited to our data. Our aim is to find the unknown physical relationship between two quantities, both of which probably have intrinsic scatter; this differs from the more usual case where the error in one of the quantities (the independent variable) can be neglected. The measurement errors in our data are probably small compared to the intrinsic scatter, and are in any case not well known in the case of the radio core flux densities. In the case of uncensored data of this kind, where ordinary least-squares regression is applicable, Isobe et al. (1990) have shown that the best estimators of the slope are those that treat the two variables symmetrically, so that one quantity is not privileged by being treated as the independent variable; they recommend the line that bisects the two lines obtained by ordinary least squares regression of each variable on the other. Schmitt’s algorithm is essentially a weighted least-squares estimator, so it seems likely that the same problems obtain and that similar solutions are applicable; the situation is less clear in the case of the Theil-Sen estimator. Our simulated data consisted of a hundred data points. The $`x_i`$ were drawn from a uniform distribution between 21.0 and 28.0; the corresponding points $`y_i`$ were generated as $`y_i=\alpha +\beta x_i`$, where $`\alpha =10`$. These values were chosen so as to match roughly with the ranges seen in radio and X-ray luminosity. Intrinsic scatter was modelled with Gaussians having standard deviations of $`\sigma _x`$ and $`\sigma _y`$; random numbers drawn from these distributions were added to $`x_i`$ and $`y_i`$. Each point had independent probabilities $`p_x`$, $`p_y`$ of being censored in the $`x`$ and $`y`$ directions, which were of the order of tens of per cent to match our dataset. Censored data were replaced with an upper limit which was higher by the modulus of a value drawn from a Gaussian distribution than the value after intrinsic scatter had been added. We ran a number of simulations (typically several hundred) and measured the median and 90 per cent confidence levels of the estimates of the slope. For the Schmitt algorithm the data were binned in a $`5\times 5`$ grid with the bin boundaries being chosen automatically by asurv. Simulation showed that the results, and confidence ranges, were not strongly affected by other choices of number of bins around this value. Neither estimator was strongly affected by choices of other distributions for the initial values of $`x_i`$. Our modelling shows that the Schmitt procedure results in a biased estimator of the true slope, $`\beta `$, of such datasets even if no scatter is added to the $`x`$-coordinate, taken to be the independent variable; it consistently underestimates $`\beta `$. Only in the limit of negligible scatter and negligible censoring is it unbiased. By contrast, the Theil-Sen procedure gives a good, unbiased estimator of $`\beta `$ when no scatter is added to the $`x_i`$, as described by Akritas et al. Both, however, are biased (in the sense of underestimating $`\beta `$) when intrinsic scatter is added to the dependent variable. Treating the variables symmetrically by finding the bisector of the two regression lines has better results. Both the Schmitt and the Theil-Sen procedures provide essentially unbiased estimators of $`\beta `$ provided that there is similar noise and censoring in the two co-ordinates; ‘similar’ here implies that $`\sigma _y=\beta \sigma _x`$. The Theil-Sen procedure deals slightly better with cases where $`\beta 1`$, though both tend to produce results that are biased towards an estimated slope of unity; we emphasise that this bias is small compared with the breadth of the distribution of the results in small datasets like ours. The 90 per cent confidence ranges are slighly smaller when the Theil-Sen estimator is used. Neither test appears to be biased by allowing the censoring probability to depend on the value of $`x_i`$ or $`y_i`$, so long as this is done symmetrically. If the censoring probabilities or the degrees of intrinsic scatter are different in the two co-ordinates then there is some bias in the results of either algorithm, in the senses tabulated in Table A10. Bias is particularly serious when the degrees of intrinsic scatter are significantly different; for example, when $`\sigma _x=2.0`$ and $`\sigma _y=1.0`$ the median estimate of the slope is biased by $`20`$ per cent, even when censoring is negligible. We conclude that * The Theil-Sen estimator is undoubtedly better than the Schmitt estimator when there is negligible intrinsic scatter in one co-ordinate but non-zero scatter in the other. * When the bisector of the two regression lines is used and when the variables are genuinely symmetrical, both procedures provide good estimates of the true slope for $`\beta 1`$. The Theil-Sen estimator gives a smaller confidence range, is less badly affected by choices of $`\beta `$ significantly different from 1, and is less biased by asymmetry in the censoring of the two variables. * The bisector of the two regression lines is a significantly biased estimator of the true slope, using either estimator, when the noise on the two variables is significantly differently distributed.
no-problem/9907/astro-ph9907210.html	ar5iv	text	# A Self-consistent Method for Estimating Photometric Redshifts ## 1 Introduction The subject of photometric redshift measurement (i.e. estimating redshifts to galaxies from multi-waveband photometric observations) has recently been revitalised, due to a rapid increase in the number of existing multi-waveband galaxy surveys. This provides a fast way for estimating redshifts to a large number of galaxies, with some trade-off in their accuracy. Moreover, in the case of faint galaxies ($`I>25`$ mag.), where spectroscopic redshifts are more difficult to measure even with the largest telescopes, photometric redshift technique is the only way to secure their redshifts. For a given galaxy, the photometric redshift can be estimated by comparing its observed SED with a set of template SEDs, corresponding to different galaxy types and shifted to different redshifts, allowing for the galaxy evolution with look-back time. The redshift and spectral type associated with the template SED which is the closest to the observed SED will then be assigned to that galaxy. Therefore, the crucial problem is the choice of the template SEDs and their behaviour at high redshifts. There are two general ways for adopting the template SEDs, as discussed below; a). Empirical templates: in this case one uses the observed SEDs for different types of galaxies. The problem here is that there are not enough information about the SEDs for different classes of galaxies at different redshifts (particularly at high redshifts). Therefore, incorporating the spectral evolution of galaxies to the model SEDs of different types is difficult and uncertain. b). Synthetic templates: uses model SEDs for different spectral types of galaxies, shifted in redshift space, assuming evolutionary population synthesis (EPS) models. The main problem here is to construct realistic model SEDs at different redshifts by constraining the evolutionary models. In the present study, we adopt a combined approach to self-consistently model the template SEDs at different redshifts, in agreement with observations. In particular, the effect of dust and its evolution with redshift is included in the SEDs, consistent with the EPS models, and optimised using observations at different look-back times. ## 2 The Photometric Redshift Technique- A New Approach A large set of chemo-photometric Evolutionary Population Synthesis (EPS) models with different input parameters are developed, each accounting for both the local properties and evolutionary behavior of different types of galaxies<sup>1,2</sup>. For a given spectral type of galaxies, the model parameters are normalised to those at $`z=0`$, by fitting them to the local observed SED of their respective type. The evolutionary behaviour (i.e. input parameters) of the EPS models (which also include contribution from dust) are then constrained by estimating the photometric redshifts ($`z_{phot}`$) to a calibrating sample of 73 galaxies with available spectroscopic redshifts ($`z_{spec}`$) and minimising the rms scatter between them. The calibrating sample is adopted from the Hubble Deep Field (HDF) so that the galaxies will have detection in, at least, four passbands, including UV which is crucial for any photometric redshift determination. The procedure is summarised in the flow chart in Figure 1. The parameters in the EPS models, constrained by observations, consist of the IMF, formation redshift ($`z_{form}`$), local star formation rate ($`\psi _0`$) and the time evolution of the star-formation rate, as parametrised by $`n`$ (see Figure 1). Once these parameters are constrained, the evolution of the optical depth ($`\tau `$) is determined directly from the gas metallicity, $`Z_{gas}`$, and the fractional mass of gas which takes part in star formation ($`f_g`$), assuming that the gas-to-dust ratio is proportional to the gas metallicity<sup>1,2</sup>. For galaxies with $`z>2`$ in the calibrating sample, the effect of Lyman continuum and Lyman forest absorption is estimated<sup>3</sup> The fluxes for these objects are subsequently corrected for inter-galactic absorption before they are used to constrain the EPS models. The template SEDs cover a range in wavelength from 0.05 $`\mu m`$ to 1 mm. The sensitivity of the result to the EPS model parameters are shown in Table 1. The optimum rms scatter obtained on the $`z_{phot}`$ vs. $`z_{spec}`$ plane (Fig. 2) is 0.11 (model 4 in Table 1), which is found using template SEDs for four different types of galaxies (elliptical, spiral, starburst and irregular) and only four passbands (UVRI). This relation for the calibrating sample is presented in Figure 2, with its rms scatter taken as the error in our photometric redshifts. ## 3 Degeneracy in Photometric Redshifts A source of uncertainty in estimating the photometric redshifts and spectral types of galaxies is the possibility that the models might be degenerate (i.e. different template SEDs producing the same result). Furthermore, the photometric errors in the observed SEDs are likely to affect the final estimate of both the photometric redshift and the spectral type of their respective galaxy. To investigate the uniqueness of the photometric redshift solutions in this study, a Monte Carlo simulation is performed. A simulated catalogue is generated to resemble the observed, UV selected HDF survey, with UBVI magnitudes, known redshifts ($`z_{input}`$) and spectral types (a UV selected catalogue is used here to ensure the availability of UV data which are crucial for any photometric redshift measurement). The galaxies in this catalogue are randomly selected to have SEDs similar to the synthetic SEDs for any of the four types of galaxies (E, Sp, Sb, Irr) considered here, shifted in redshift space. Random Gaussian noise, resembling photometric errors are then added to the simulated SEDs. The simulated catalogue contains 273 galaxies (the same as the complete UV-selected HDF catalogue), has a magnitude limit of U=27 mag., an apparent UV magnitude distribution similar to the observed catalogue and a redshift limit of $`z=2.5`$ (maximum redshift allowed in a UV selected survey). The photometric redshift code was then used to predict the redshifts ($`z_{output}`$) and spectral types of individual galaxies in the simulated catalogue and to compare them with their input values. The $`log(z_{output}/z_{input})`$ distribution (Fig. 3) shows a distinct peak at zero, indicating that the redshifts for the simulated galaxies are well re-produced within $`\mathrm{\Delta }z0.11`$. As far as the predicted spectral types are concerned, we recover the type classification for ellipticals in the input catalogue by $`100\%`$ (i.e. no mis-identification of ellipticals). Regarding the spirals, irregulars and starbursts, we can recover, respectively, $`79\%`$, $`85\%`$ and $`71\%`$ of the spectral types of galaxies in the input catalogue. ## 4 Consistency Test In order to explore the consistency of the above procedure, we apply it on a complete sample of UV selected galaxies from the HDF, to estimate their photometric redshifts. Using this redshift catalogue, we then predict the physical quantities (i.e. the luminosity density) in redshift intervals and compare them with similar measurements from other independent studies. A UV selected survey is particularly useful for such comparison because the UV light is heavily affected by dust in galaxies<sup>4</sup>, allowing a test of the estimated extinction corrections. Moreover, such surveys do not contain objects with $`z>2.5`$ and hence, are less affected (then deep surveys in other bands) by uncertainties in correction for Lyman continuum absorption. The UV luminosity densities are estimated in redshift intervals, using the photometric redshift catalogue. The results are compared with similar measurements from other, independent, studies (Fig. 4) and show excellent agreement. This indicates that the statistical results, based on the present photometric redshift technique, are consistent with other studies. ## 5 Summary and Conclusion A new method is presented to estimate photometric redshifts of galaxies. The main improvement in this study is to develope more realistic template SEDs over a large range in redshift and in agreement with the observational data. The SEDs cover a range in wavelength from 0.05 $`\mu m`$ to 1 mm, with their main characteristics as follows: a). they are optimised to produce the observed colours of galaxies at $`z0`$; b). incorporate the chemo-photometric evolution of galaxies of different types (i.e. stellar luminosities for different metallicities), in agreement with observations; c). allow a self-consistent treatment of dust contribution and its evolution with redshift; d). include absorption and re-emission of radiation by dust and hence, realistic estimates of the far-infrared radiation; e). include correction for inter-galactic absorption by Lyman continuum and Lyman forest. Using a calibration sample of 73 galaxies with available spectroscopic redshifts, the evolutionary model parameters are optimised to get the minimum scatter between the photometric and spectroscopic redshifts in the calibrating sample. Using four passbands (UVRI), the photometric redshifts are estimated to an accuracy of $`\mathrm{\Delta }z=0.11`$. The method is applied to a UV selected sample of HDF galaxies. The UV luminosity densities, estimated in redshift intervals, are in excellent agreement with similar measurements from other, independent, surveys.
no-problem/9907/astro-ph9907120.html	ar5iv	text	# The Youngest Lobe-Dominated Radio Sources ## 1 Introduction There exists a class of enigmatic sources in which high luminosity radio emission regions are located ‘symmetrically’ on both sides of the central engine on linear scales of less than 1 kpc; the so-called Compact Symmetric Objects (Wilkinson et al.,, 1994). Several theories have been suggested to explain CSOs. It has been suggested that they are a young phase in the development of classical doubles (e.g., Phillips & Mutel,, 1982; Fanti et al.,, 1995; Readhead et al., 1996b, ) or that they are fairly old sources in which a dense environment inhibits their growth (e.g., van Breugel et al.,, 1984). Finally it has been proposed that they are a separate class of short lived objects which ‘fizzle out’ after about $`10^4`$ years (Readhead et al.,, 1994). An obvious way to try to distinguish between these models is to investigate the growth in overall sizes of CSOs and hence set limits to their ages directly. ## 2 Data analysis for 0710+439 The radio source 0710+439 ($`z=0.518`$, Lawrence et al.,, 1996) has a high radio luminosity ($`L_{5GHz}=5\times 10^{26}h^2`$ W Hz<sup>-1</sup>, Readhead et al., 1996a, ). The radio flux density is very weakly polarised and it is not time variable (Aller et al.,, 1992). 0710+439 has been observed at 5 GHz with a global VLBI array at 6 epochs spread over a period of 17 years. The first five epochs were analysed by Owsianik & Conway, (1998). Here we add new data from multi-snapshot 14 station global observations made at 18th of Sep 1997. The highest dynamic range image (see Fig. 1a) was obtained using DIFMAP (Shepherd et al.,, 1994) and was used as a starting point in re-mapping the other epochs following the procedure describe by Owsianik & Conway, (1998). This image shows clearly the overall triple structure of the source. The northern and the southern components show compact, bright subcomponents which are associated with hotspots surrounded by the faint extended lobe emission. The middle component is associated with the base of the northern jet, and the centre of activity lies at the southern end of this component (Taylor et al.,, 1996). The image in Fig. 1a shows clearly for the first time at this frequency a bridge of emission between northern hotspot and the middle component. In addition the emission previously detected at 1.6 GHz (Xu,, 1994) between the middle and southern components is for the first time now also detected at 5GHz. The hotspot components are well separated on CLEAN images which allows us to use the AIPS task JMFIT to fit the position of hotspot at each epoch. The fitted linear regression line to this data gives an estimated separation rate of $`13.614\pm 0.988`$ $`\mu `$as/yr (Fig. 1b). From the obtained correlation coefficient of 0.989 we can reject the null hypothesis of no motion at a better than $`0.1\%`$ confidence level (for a fuller discussion on errors see Owsianik & Conway,, 1998). Dividing the distance between hotspots of 87.07 $`h^1`$ pc by their observed separation velocity of $`0.243\pm 0.018h^1`$c we estimate that the source 0710+439 is $`1100\pm 100`$ years old. Assuming that hotspots are close to their equipartition pressure and assuming that the source is orientated not too far from the sky plane (as supported by the source having an arm-length ratio close to one) then ram pressure arguments imply an external density of $`2h^{18/7}`$ cm<sup>-3</sup>. From the velocity of advance of 0710+439 we can calculate that the rate of work done advancing the two hotspots is $`5\times 10^{43}h^{17/7}`$ erg s<sup>-1</sup>. Following the arguments of Readhead et al., 1996a we can estimate an upper limit on the total jet power of $`4\times 10^{44}h^{10/7}`$ erg s<sup>-1</sup>. A lower limit on the jet power equals the sum of work of advance and radio luminosity ($`7\times 10^{43}h^2`$ erg s<sup>-1</sup>). Given these numbers we can estimate that the efficiency ($`ϵ`$) of conversion of jet mechanical energy into radio emission in 0710+439 is $`23\%<ϵ<53\%`$ (for $`h=0.6`$). ## 3 Data analysis for 2352+495 The radio source 2352+495 ($`z=0.238`$, Lawrence et al.,, 1996) has a high radio luminosity ($`L_{5GHz}=1\times 10^{26}h^2`$ W Hz<sup>-1</sup>, Readhead et al., 1996a, ). The radio flux density is very weakly polarised and it does not exhibit a significant time variation (Aller et al.,, 1992). Conway et al., (1992) revealed for the first time the triple radio morphology of 2352+495 and from analysis of multi-epoch data it was argued that the center of activity was associated with the middle component (Conway et al.,, 1992). Later 15 GHz observations located a compact core component at the southern end of this middle component (Taylor et al.,, 1996). 2352+495 has been observed at 5 GHz with a global VLBI array at five epochs spread over a period of 14 years. The first two epochs included in our analysis were made at epoch 1983.93 with a global array of 6 telescopes and then at epoch 1986.89 using multiple snapshots with 9 telescopes (see Conway et al.,, 1992). Here we reanalyse these epochs and add data from three additional epochs; a multi snapshot 12 station global VLBI observations made at epoch 1993.44, a 10 station VLBA observations made at epoch 1995.67 (donated by G.B. Taylor and R.C. Vermuelen, gratefully acknowledged by the authors) and finally a 14 station global VLBI observations made at epoch 1997.71. The telescopes used included those from the European VLBI Network (EVN), the Very Long Baseline Array (VLBA), the Very Large Array (VLA) and Haystack Observatory. The analysis of the data followed the procedure described in Owsianik & Conway, (1998). Fig. 2a shows the best image of 2352+495 obtained for the last epoch data using DIFMAP (Shepherd et al.,, 1994), this map was used as a starting point in analysis of the earlier epochs. This 5th epoch map clearly shows the overall triple structure of the source. The northern and the southern components show compact, bright subcomponents which are associated with hotspots surrounded by fainter extended emission from the lobes. The middle very bright component appears to be associated with the base of the northern jet (Taylor et al.,, 1996). The image also shows an almost continuous jet connecting the middle component with the southern hotspot. For the first time this image also reveals portions of a possible jet-like component on the other side of the middle component/core feature. The well separated hotspot components in the CLEAN images allowed us to use the AIPS task JMFIT to fit the position of hotspots at each epoch. The linear regression fits to the observed changes in gaussian component separation gave us an estimate of the relative angular separation rate of $`21.062\pm 2.704\mu `$as/yr corresponding to a separation velocity of $`0.202\pm 0.026h^1`$c (Fig. 2b). The linear regression fit gives a correlation coefficient of 0.976 which allows us to reject the null hypothesis of no motion at the better than $`1\%`$ confidence level. Dividing the distance between hotspots of 117.03 $`h^1`$ pc by their observed separation rate we estimate that 2352+495 is $`1900\pm 250`$ years old. Assuming that hotspots are close to their equipartition pressure and assuming that the source is orientated not too far from the sky plane (consistent with an arm-length ratio very close to 1) then ram pressure arguments imply an external density of $`1h^{18/7}`$ cm<sup>-3</sup>, which is consistent with the NLR intercloud medium. Given our estimate of an age and the jet thrust we can compare the power of advance required to drive the hotspot forward, with the radio luminosity and the jet power (Readhead et al., 1996a, ). For an age of 2352+495 the rate of work done in advancing the two hotspots is $`1\times 10^{43}h^{17/7}`$ erg s<sup>-1</sup>. The upper limit on the total power supplied by the jet is $`8\times 10^{43}h^{10/7}`$ erg s<sup>-1</sup> (Readhead et al., 1996a, ). The lower limit on the total power of the jet is the sum of the power of advance and the radio luminosity ($`3\times 10^{42}h^2`$ erg s<sup>-1</sup>). Given these numbers we can therefore estimate that the efficiency of conversion of the jet mechanical energy into the radio emission is $`5\%<ϵ<19\%`$ (for $`h=0.6`$). ## 4 Evolution of CSOs In addition to the two sources presented in this paper there are now reliable velocities of expansion for three other CSOs; namely 0108+388 (Owsianik et al.,, 1998), 2021+614 (Conway et al., (1994) and Tschager et al., this volume) and 1943+456 (Polatidis, this volume). All of the expansion velocities are of order $`0.2h^1`$c; giving hotspot advance speeds of an order $`0.1h^1`$c and ages of a few thousand years. It has therefore become clear that the majority of CSOs are very young sources, which grow fairly rapidly (Owsianik & Conway,, 1998; Owsianik et al.,, 1998). What is still uncertain is their subsequent evolution. The simplest scenario is that CSOs evolve via Medium-size Symmetric Objects into Large-size Symmetric Objects (Fanti et al.,, 1995; Readhead et al., 1996b, ; Owsianik & Conway,, 1998). Given the rapid expansion rate of CSOs the sources will spend only a short time in the CSO phase, and it would be expected that only a small fraction of sources would be CSOs. In fact CSOs comprise about $`7.5\%`$ of sources in the flux limited surveys at 5GHz (Polatidis, this volume). However, this large fractional population can be explained if, as is theoretically expected, there is a strong negative luminosity evolution with increasing size. In such a model CSOs will evolve into more numerous, but much weaker sources (e.g., Begelman,, 1996; Readhead et al., 1996b, ; Owsianik & Conway,, 1998). As noted by Readhead et al., 1996a for 2352+495, and as we find here the limits on radiative efficiency for CSOs compared to ‘classical’ large lobe-dominated sources (in which $`ϵ`$ is of order of few per cent Owsianik & Conway,, 1998) empirically demonstrates that the expected luminosity evolution does in fact occur and with a size consistent with that expected by theoretical models (Begelman,, 1996; Readhead et al., 1996a, ).
no-problem/9907/astro-ph9907225.html	ar5iv	text	# Fast rotation of strange stars ## 1 Introduction A deconfined, beta stable quark matter is composed of the u, d, and s quarks, and in contrast to “ordinary” baryon matter has the strangeness per unit baryon number $`1`$ (therefore the name ‘strange quark matter’). An intriguing possibility that such strange matter could be the absolute ground state of matter at zero pressure and temparature was pointed out in an influential paper of Witten (1984) \[such a possibility has been contemplated before by Bodmer (1971)\]. This possibility is not excluded by what we know from laboratory nuclear physics. If the hypothesis of strange matter is true, then some of neutron stars could be actually strange stars, built entirely of strange matter (Haensel et al. 1986, Alcock et al. 1986; arguments against the existence of strange stars have been presented by Alpar (1987), and Caldwell & Friedman (1991) ). Detailed models of nonrotating strange stars were constructed in (Haensel et al. 1986, Alcock et al. 1986). Calculations concerning physics and astrophysics of strange stars, published up to 1991, were reviewed in (Madsen & Haensel 1991). References to more recent work on strange stars can be found in review articles by Weber et al. (1995) and Madsen (1999), and in the monograph of Glendenning (1997). Most recently, strange stars have been invoked in models of X-ray bursters (Bombaci 1997, Cheng et al. 1998) and of gamma-ray bursters (Dai & Lu 1998). Rapid rotation of strange stars has become a topic of interest in 1989, after a sensational (but subsequently withdrawn) claim of detection of a $`0.5`$ ms pulsar in SN 1987A. In several papers, strange star has been advanced as an appropriate model for a half-millisecond pulsar (Glendenning 1989a,b, Frieman & Olinto 1989). However, several authors pointed out that the experimental constraints on the strange matter model imply that strange stars could not rotate so fast (Haensel and Zdunik 1989, Lattimer et al. 1990, Zdunik & Haensel 1990, Prakash et al. 1990). Subsequent studies of rapid rotation of strange stars focused on the effect of rotation on the structure of the normal crust (Glendenning & Weber 1992). Recently, it has been shown that fast rotation of hot, young neutron stars is severely limited by the emission of gravitational radiation, due to r-mode instability \[for a review, see e.g. Stergioulas (1998)\]. However, in contrast to newly born neutron stars, hot strange stars are not subject to the r-mode instability (Madsen 1998). Therefore, strange stars formed in collapse of rotating stellar cores can rotate very fast. This gives additional motivation for studying rapid rotation of strange stars. It should be stressed, that nearly all calculations of uniformly rotating models of strange stars were done within the slow rotation approximation (see, e.g., Colpi & Miller 1992, Glendenning & Weber 1992). Actually, in the calculations of Glendenning & Weber (1992) the slow rotation scheme, pioneered by Hartle (1967, 1973) and Hartle & Thorne (1968), was supplemented by a “self-consistency condition” at the Keplerian frequency (Weber & Glendenning 1991, 1992). However, the slow rotation approximation is not valid near the shedding (Keplerian) limit. As discussed by Salgado et al. (1994a, b), the improved, “self-consistent” version of this approximation, overestimates by more than 10% the maximum rotation frequency of neutron star models. So far, the only exact<sup>1</sup><sup>1</sup>1In this article, the term “exact” is relative to the treatment of rotation and is used to distinguish from the slow rotation approximation calculations of the rapidly rotating models of strange stars were done by Friedman (quoted in Glendenning 1989a,b) and Lattimer et al. (1990). However, the precision of their numerical results, calculated for one specific equation of state of strange matter, was not checked using an independent exact calculation. Also, the fact that their 2-D code was based on the Butterworth & Ipser (1976) scheme, rises some suspicion concerning its precision when applied to the equation of state of strange matter, characterized by huge density discontinuity at the stellar surface. Indeed Butterworth & Ipser (1976) considered the case of a density discontinuity at the stellar surface – they computed configurations of rotating homogeneous bodies, but as we discuss below (Sec. 4.3), they treated properly the discontinuity only in the radial direction, which is not sufficient for objects which deviate substancially from the spherical symmetry. In the present paper we perform exact calculations of rapidly rotating strange stars, composed entirely of strange quark matter. Our calculations are based on a multi-domain spectral method recently developed by Bonazzola et al. (1998b). Such a multi-domain technique enables us to treat exactly the density discontinuity at the surface of strange stars. The calculations are performed for a family of equations of state of strange quark matter. On one hand, precision of our calculation is checked using internal error estimators. On the other hand, the validity of scaling relations for the rotating configurations, displayed by our numerical results, yields an additional test of precision of our calculations. Finally, we address the problem of triaxial instabilities of rapidly rotating strange stars, and point out differences with respect to normal neutron stars. In Sect. 2 we discuss the equation of state of strange matter, based on the MIT bag model of quark matter. The set of equations to be solved and the numerical procedure is presented in Sect. 3. Stationary, uniformly rotating configuration of strange stars are studied in Sect. 4, where we also derive exact scaling relations for the parameters of the rotating strange star models. Parameters of the maximally rotating configurations, stable with respect to the axially-symmetric perturbations, are calculated in Sect. 5, and an exact formula for the maximum rotation frequency of rotating strange stars is derived. We derive also exact formulae for mass and radius of the maximum mass configuration of rotating strange stars. In Sect. 6, we study the problem of triaxial instabilities of rapidly rotating strange stars, and show, that rapidly rotating strange stars might be more susceptible to these instabilities than ordinary neutron stars. Finally, Sect. 7 contains a discussion of our results and the conclusion. ## 2 Equation of state and static models of bare strange stars Equation of state of strange matter will be based on the MIT bag model. Baryon number density of strange matter is $`n=\frac{1}{3}(n_\mathrm{u}+n_\mathrm{d}+n_\mathrm{s})`$, where $`n_\mathrm{u}`$ is the number density of the u-quarks etc. In what follows, we will use the simplest model of self-bound strange quark matter, ignoring the strange quark mass and neglecting the quark interactions except for the confinement effects described by the bag constant. We consider thus massless, noninteracting u, d, s quarks, confined to the bag volume. In the case of a bare strange star, the boundary of the bag coincides with stellar surface. Using the model described above, one can express the energy density, $`ϵ`$, and the pressure, $`P`$, of strange quark matter, as functions of the baryon number density, $`n`$, in the following form $`ϵ`$ $`=`$ $`an^{\frac{4}{3}}+B,`$ $`P`$ $`=`$ $`{\displaystyle \frac{1}{3}}an^{\frac{4}{3}}B,`$ (1) where $`B`$ is the MIT bag constant, and the parameter $`a`$ is given by $$a=\frac{9}{4}\pi ^{\frac{2}{3}}\mathrm{}c=952.371\mathrm{MeV}\mathrm{fm}.$$ (2) The bag constant $`B`$ describes the difference in the energy density of the true (real) vacuum and that of the QCD vacuum. This parameter plays a crucial role in the MIT bag model, being actually responsible for the quark confinement. Let us mention, that in the case of a bare strange star of $`\mathrm{M}_{}`$ we are dealing with a bag of a radius of $`10\mathrm{km}`$, containing $`10^{57}`$ quarks. The constant $`B`$ represents a natural unit for both the energy density and the pressure. Defining dimensionless energy density and pressure of strange matter, $`\stackrel{~}{ϵ}ϵ/B`$, $`\stackrel{~}{P}=P/B`$, we can write down a dimensionless form of the EOS of strange matter, $$\stackrel{~}{P}=\frac{1}{3}(\stackrel{~}{ϵ}4).$$ (3) The above dimensionless form of the EOS of strange matter constitutes a basis for the derivation of the scaling laws relating the families of static models of strange stars, calculated for different values of $`B`$ (Witten 1984, Haensel et al. 1986). For the (gravitational) mass, baryon mass, and radius of static configuration with maximum allowable mass we get (assuming $`G=\mathrm{6.6726\; 10}^{11}\mathrm{m}^3\mathrm{kg}^1\mathrm{s}^2`$, $`\mathrm{M}_{}=\mathrm{1.989\; 10}^{30}\mathrm{kg}`$ and $`c=\mathrm{2.99792458\; 10}^8\mathrm{m}\mathrm{s}^1`$): $`M_{\mathrm{max}}^{\mathrm{stat}}`$ $`=`$ $`{\displaystyle \frac{1.9638}{\sqrt{B_{60}}}}\mathrm{M}_{},`$ $`M_{\mathrm{B},\mathrm{max}}^{\mathrm{stat}}`$ $`=`$ $`{\displaystyle \frac{2.6252}{\sqrt{B_{60}}}}\mathrm{M}_{},`$ $`R_{M_{\mathrm{max}}}^{\mathrm{stat}}`$ $`=`$ $`{\displaystyle \frac{10.712}{\sqrt{B_{60}}}}\mathrm{km}`$ (4) where $`B_{60}B/(60\mathrm{MeV}\mathrm{fm}^3)`$. The baryon mass is defined as $`M_\mathrm{B}\mathrm{1.66\; 10}^{27}\mathrm{kg}N_\mathrm{B}`$, where $`N_\mathrm{B}`$ is the total baryon number of stellar configuration. Consider the case of strange matter at zero pressure (such a situation corresponds to the surface of a “bare strange star”). For massless, noninteracting quarks, we obtain the following set of parameters, characterizing the properties of strange matter at zero pressure, $`n_0`$ $`=`$ $`0.28665\left(B_{60}\right)^{3/4}\mathrm{fm}^3,`$ $`\rho _0`$ $``$ $`ϵ_0/c^2=4.2785\times 10^{17}B_{60}\mathrm{kg}\mathrm{m}^3.`$ (5) Strange matter would be the real ground state of matter at zero pressure, only if the energy per unit baryon number, at zero pressure, $`E_0ϵ_0/n_0=837.26(B_{60})^{1/4}\mathrm{MeV}`$, was below the energy per baryon in the maximally bound terrestrial nucleus, $`{}_{}{}^{56}\mathrm{Fe}`$, equal $`930.4\mathrm{MeV}`$. This implies $`B_{60}<1.525`$. On the other hand, a consistent model should not lead to a spontaneous fusion of neutrons into droplets of u, d quarks (which would eventually transform into droplets of strange matter - “strangelets”). In view of the fact, that $`a_{\mathrm{ud}}/a=(1+2^{4/3})/3`$, we get $`E_0^{(\mathrm{ud})}=1.127E_0^{(\mathrm{uds})}`$. Therefore, the condition of the stability of neutrons with respect to a spontaneous fusion into strangelets $`E_0^{(\mathrm{ud})}=988.5(B_{60})^{1/4}\mathrm{MeV}>939.6\mathrm{MeV}`$, which implies $`B_{60}>0.9821`$. Summarizing, within the simplest model of strange matter, the bag constant is constrained by $`0.9821<B_{60}<1.525`$. An important quantity, relevant for the pulsations and stability of strange stars is the adiabatic index of strange matter, defined as $`\gamma (n/P)\mathrm{d}P/\mathrm{d}n`$. Dependence of $`\gamma `$ on baryon density of strange matter is shown in Fig. 1; it is qualitatively different from $`\gamma (n)`$ for ordinary neutron star matter. The values of $`\gamma `$ in the outer layers of strange stars are very large. Even at the highest densities, allowed for strange stars, the value of $`\gamma `$ is significantly higher, than the ultrarelativistic Fermi gas value $`4/3`$, predicted by the asymptotic freedom of QCD at $`n\mathrm{}`$. ## 3 Formulation of the problem and numerical code ### 3.1 Equations of stationary motion We refer to Bonazzola et al. (1993) for a complete description of the set of equations to be solved to get general relativistic models of stationary rotating bodies. Let us simply recall here that, under the hypothesis of stationarity, axisymmetry and purely azimuthal motion (no convection), a coordinate system $`(t,r,\theta ,\phi )`$ can be chosen so that the spacetime metric takes the form $`\mathrm{d}s^2`$ $`=`$ $`N^2\mathrm{d}t^2+B^2r^2\mathrm{sin}^2\theta (\mathrm{d}\phi N^\phi \mathrm{d}t)^2`$ (6) $`+A^2(\mathrm{d}r^2+r^2\mathrm{d}\theta ^2),`$ where $`N`$, $`N^\phi `$, $`A`$ and $`B`$ are four functions of $`(r,\theta )`$. The Einstein equations result in a set of four elliptic equations for these metric potentials<sup>2</sup><sup>2</sup>2in this section $`G=c=1`$.: $`\mathrm{\Delta }_3\nu =\mathrm{\hspace{0.33em}4}\pi A^2(E+3P+(E+P)U^2)`$ $`+{\displaystyle \frac{B^2r^2\mathrm{sin}^2\theta }{2N^2}}(N^\phi )^2\nu (\nu +\beta )`$ (7) $`\stackrel{~}{\mathrm{\Delta }}_3(N^\phi r\mathrm{sin}\theta )=16\pi {\displaystyle \frac{NA^2}{B}}(E+P)U`$ $`r\mathrm{sin}\theta N^\phi (3\beta \nu )`$ (8) $`\mathrm{\Delta }_2[(NB1)r\mathrm{sin}\theta ]=\mathrm{\hspace{0.33em}16}\pi NA^2BPr\mathrm{sin}\theta `$ (9) $`\mathrm{\Delta }_2(\nu +\alpha )=\mathrm{\hspace{0.33em}8}\pi A^2[P+(E+P)U^2]`$ $`+{\displaystyle \frac{3B^2r^2\mathrm{sin}^2\theta }{4N^2}}(N^\phi )^2(\nu )^2`$ (10) where the following abreviations have been introduced: $$\nu :=\mathrm{ln}N,\alpha :=\mathrm{ln}A,\beta :=\mathrm{ln}B,$$ (11) $`\mathrm{\Delta }_2`$ $`:=`$ $`{\displaystyle \frac{^2}{r^2}}+{\displaystyle \frac{1}{r}}{\displaystyle \frac{}{r}}+{\displaystyle \frac{1}{r^2}}{\displaystyle \frac{^2}{\theta ^2}}`$ (12) $`\mathrm{\Delta }_3`$ $`:=`$ $`{\displaystyle \frac{^2}{r^2}}+{\displaystyle \frac{2}{r}}{\displaystyle \frac{}{r}}+{\displaystyle \frac{1}{r^2}}{\displaystyle \frac{^2}{\theta ^2}}+{\displaystyle \frac{1}{r^2\mathrm{tan}\theta }}{\displaystyle \frac{}{\theta }}`$ (13) $`\stackrel{~}{\mathrm{\Delta }}_3`$ $`:=`$ $`\mathrm{\Delta }_3{\displaystyle \frac{1}{r^2\mathrm{sin}^2\theta }}`$ (14) $`ab`$ $`:=`$ $`{\displaystyle \frac{a}{r}}{\displaystyle \frac{b}{r}}+{\displaystyle \frac{1}{r^2}}{\displaystyle \frac{a}{\theta }}{\displaystyle \frac{b}{\theta }}.`$ (15) In Eqs. (7)-(10), $`U`$ and $`E`$ are respectively the fluid 3-velocity and energy density, both as measured by the locally non-rotating observer: $`U=Br\mathrm{sin}\theta (\mathrm{\Omega }N^\phi )/N`$, $`E=\mathrm{\Gamma }^2(ϵ+P)P`$, $`\mathrm{\Gamma }=(1U^2)^{1/2}`$. These equations are supplemented by the first integral of motion $$H+\nu \mathrm{ln}\mathrm{\Gamma }=\mathrm{const}.,$$ (16) where $`H`$ is pseudo-enthalpy (or log-enthalpy) defined by $$H=\mathrm{ln}\left(\frac{ϵ+P}{nE_0}\right).$$ (17) The dimensionless EOS of strange matter can be parametrized in terms of the $`H`$ variable as $`\stackrel{~}{ϵ}`$ $`=`$ $`3\mathrm{e}^{4H}+1,`$ $`\stackrel{~}{P}`$ $`=`$ $`\mathrm{e}^{4H}1.`$ (18) Consequently, the baryon number density can expressed in terms of $`H`$ by $$n=n_0\mathrm{e}^{3H},$$ (19) where for our model the baryon number density of strange matter at zero pressure $`n_0=(3B/a)^{3/4}`$. The equations of stationary motion involve the quantities with dimensions of energy density and powers of length, and universal constants $`G`$ and $`c`$. These equations can be rewritten in a dimensionless form, if we introduce the dimensionless quantities $`\stackrel{~}{ϵ}`$, $`\stackrel{~}{P}`$, $`\stackrel{~}{r}`$, $`\stackrel{~}{\mathrm{\Omega }}`$, via following relations, $`ϵ`$ $`=`$ $`B\stackrel{~}{ϵ},`$ $`P`$ $`=`$ $`B\stackrel{~}{P},`$ $`r`$ $`=`$ $`{\displaystyle \frac{c^2}{\sqrt{GB}}}\stackrel{~}{r},`$ $`\mathrm{\Omega }`$ $`=`$ $`{\displaystyle \frac{\sqrt{GB}}{c}}\stackrel{~}{\mathrm{\Omega }}.`$ (20) Notice, that the solutions of dimensionless equations of stationary motion will not depend explicitly on $`B`$. In order to recover the solution in conventional units, for a specific value of $`B`$, one has just to use relations (20). ### 3.2 Numerical procedure The non-linear elliptic equations (7)-(10) are solved iteratively by means of the multi-domain spectral method developed recently by Bonazzola et al. (1998b). In this method, the whole space is divided into three domains: D1: the interior of the star, D2: an intermediate domain whose inner boundary is the surface of the star and outer boundary a sphere located at $`r:=R_{23}2r_{\mathrm{eq}}`$ (where $`r_{\mathrm{eq}}`$ is the equatorial coordinate-radius of the star), and D3: the external domain whose inner boundary is the outer boundary of D2 and which extends up to infinity, thanks to the compactification $`u=1/r`$. A mapping $`(\xi ,\theta ^{})(r,\theta )`$ satisfying $`\theta =\theta ^{}`$ is introduced in each domain so that the domain boundaries lie at a constant value of the coordinate $`\xi `$ (typically, $`\xi =1`$, $`0`$ or $`1`$). Explicitly, this mapping reads $$\text{in D1:}r=\alpha _1\left[\xi +(3\xi ^42\xi ^6)F_1(\theta ^{})\right],\xi [0,1],$$ (21) $`\text{in D2:}r=\alpha _2\left[\xi 1+{\displaystyle \frac{\xi ^33\xi +2}{4}}F_2(\theta ^{})\right]+R_{23},`$ $`\xi [1,1],`$ (22) $$\text{in D3:}u:=\frac{1}{r}=\frac{1}{2R_{23}}(1\xi ),\xi [1,1],$$ (23) where the functions $`F_1(\theta ^{})`$ and $`F_2(\theta ^{})`$ are related to the equation $`r=S(\theta )`$ of the stellar surface by $`S(\theta )=\alpha _1[1+F_1(\theta )]=\alpha _2[2+F_2(\theta )]+R_{23}`$. The mapping (21)-(23) is a specialization to the axisymmetric case of the 3-D mapping introduced in Bonazzola et al. (1998b), to which the reader is referred for more details. A spectral expansion of each relevant field is then performed with respect to the coordinates $`(\xi ,\theta ^{})`$. We use Chebyshev polynomials in $`\xi `$ and trigonometric polynomials or Legendre polynomials in $`\theta ^{}`$ (see Bonazzola et al. 1999 for the use of spectral methods in relativistic astrophysics). Since the discontinuities in the physical fields (strong discontinuity in the density, cf. Fig. 3 below, discontinuity in the second derivative for the metric potentials) are located at the boundary between two domains (namely D1 and D2), the applied spectral method is free from any Gibbs phenomenon and leads to a very high precision. This is illustrated in Fig. 5 of Bonazzola et al. (1998b) which concerns the case of a rotating constant-density body (hence with a strong discontinuity of the density field at the surface) in the Newtonian theory. An analytical solution is available in this case (MacLaurin ellipsoid) and can be used to evaluate the accuracy of the code. The code gives a rapidly rotating MacLaurin ellipsoid with a relative error of the order $`10^{12}`$ with 49 (resp. 25) degrees of freedom in $`r`$ (resp. $`\theta `$). Figure 5 of Bonazzola et al. (1998b) also demonstrates that the error is evanescent, i.e. that it decreases exponentially with the number of degrees of freedom (i.e. grid points). This represents a major improvement with respect to the spectral method developed previously by Bonazzola et al. (1993) to compute stationary configurations of rotating bodies in general relativity. This method has been used to compute rapidly rotating neutron star models by Salgado et al. (1994a,b), Haensel et al. (1995), Goussard et al. (1997, 1998) and Nozawa et al. (1998) but it could not have been employed as such to compute models of rotating strange stars: the Gibbs phenomenon induced by the density discontinuity at the stellar surface would have been too large. ### 3.3 Description of a typical model A rapidly rotating strange star model obtained from the code is shown in Figs. 2 and 3. Assuming $`B=60\mathrm{MeV}\mathrm{fm}^3`$, this model has a baryon mass $`M_\mathrm{B}=2.00\mathrm{M}_{}`$, a gravitational mass $`M=1.60\mathrm{M}_{}`$, and a rotation period $`P=0.87\mathrm{ms}`$. The central baryon density is $`n_\mathrm{c}=0.42\mathrm{fm}^3`$, the central values of the metric potentials $`N`$, $`A`$, $`B`$ and $`N^\phi `$ \[cf. Eq. (6)\] are respectively 0.65, 1.47, 1.47 and $`0.46\mathrm{\Omega }`$. The boundary between the computational domains D1 and D2 introduced in Sect. 3.2 coincides with the stellar surface (thick solid line in Fig. 2). The boundary between the computational domains D2 and D3 is out of the scope of the Figure. The energy density profile corresponding to this model is shown in Fig. 3. Note the strong discontinuity at the stellar surface: the jump in density is $`2/3`$ of the central density. ### 3.4 Treatment of configurations at the Keplerian limit The maximum rotation rate that a star can sustain is reached when the velocity at the equator equals the velocity of an orbiting particle (a higher velocity would result in a centrifugal break-up of the star). This limit is called the Keplerian limit. It is reached when the derivative $`H/r`$ vanishes at the stellar equator. The surface of the star is then no longer smooth but exhibits a cusp along the equator (see Fig. 4). Such a non-differentiable surface cannot be decribed by the mapping (21)-(23) because the functions $`F_1(\theta )`$ and $`F_2(\theta )`$ are assumed to be expandable in $`\mathrm{cos}(k\theta )`$ series (cf. Bonazzola et al. (1998b)), which implies that they are smooth functions of $`\theta `$. The solution to this problem consists in freezing the adaptation of the mapping to the stellar surface when the ratio $`(H/r)_{\mathrm{eq}}/(H/r)_{\mathrm{pole}}`$ passes below a certain threshold during the iteration process. For instance, this threshold was chosen to be $`0.28`$ in the computation shown in Fig. 4. Consequently, in the final result, the density discontinuity at the stellar surface no longer coincides with the boundary between the computational domains D1 and D2, except at the equator. In this case, a Gibbs phenomenon is present. The accuracy of the calculation is then lower than when the mapping is adapted to the surface of the star. However, since the stellar interior covers most of domain D1 (cf. Fig. 4), the Gibbs phenomenon is rather limited. It is notably less severe than if D1 would have been a sphere of radius $`r_{\mathrm{eq}}`$, as in our previous numerical method (Bonazzola et al. 1993). ### 3.5 Tests of the numerical code Various tests have been passed by the code. First of all, we recover previous results, e.g. those presented in Nozawa et al. (1998), when using an EOS for neutron star matter instead of the strange quark matter EOS presented in Sect. 2. Regarding the treatment of the density discontinuity at the stellar surface, the code has been tested on the computation of MacLaurin ellipsoids (homogeneous rotating bodies in the Newtonian regime), and gives excellent results, as recalled in Sect. 3.2: the relative accuracy reaches $`10^{12}`$ ! Another type of test is the evaluation of the virial identities GRV2 (Bonazzola 1973, Bonazzola & Gourgoulhon 1994) and GRV3 (Gourgoulhon & Bonazzola 1994), this latter being a relativistic generalization of the classical virial theorem. GRV2 and GRV3 are integral identities which must be satisfied by any solution of the Einstein equations (7)-(10) and which are not imposed during the numerical procedure (cf. Nozawa et al. 1998 for details on the computation of GRV2 and GRV3). When presenting numerical results, we will systematically give the accuracy by which the numerical solution satisfies these virial identities. As discussed in Sect. 3.4, the computation of configurations rotating at the Keplerian limit is not free of Gibbs phenomenon. In order to gauge the resulting numerical error, we performed various computations of the same Keplerian configuration, by varying the numbers of coefficients in the spectral method and by varying the threshold on $`(H/r)_{\mathrm{eq}}/(H/r)_{\mathrm{pole}}`$ for freezing the mapping. We systematically used three sets of numbers of coefficients (= number of collocation points): $`(N_r,N_\theta )`$ = $`(33,17)`$, $`(37,19)`$ and $`(41,21)`$ in each of the three domains. The threshold on $`(H/r)_{\mathrm{eq}}/(H/r)_{\mathrm{pole}}`$ was varied from $`0.28`$ down to $`0.20`$. The variation of all these parameters lead to a relative change of the numerical solution of the order of or below $`1\%`$. This gives an estimation of the error of our method in computing strange stars at the Keplerian limit. Note the GRV2 and GRV3 errors for Keplerian configurations revealed to be better than this. For instance, for the configuration shown in Fig. 4, the GRV2 (resp. GRV3) error is $`\mathrm{3\hspace{0.17em}10}^5`$ (resp. $`\mathrm{1.3\hspace{0.17em}10}^3`$). ## 4 Numerical results ### 4.1 General properties For rigid rotation, the equilibrium configurations span a two-dimensional domain. This domain is shown in Fig. 5 in the central density - gravitational mass plane. Selected configurations are listed in Table 1. The symbols have the following meaning: $`M`$ is the gravitational mass; $`M_\mathrm{B}`$ is the baryon mass; $`\mathrm{\Omega }`$ is the angular velocity, $`P`$ the corresponding rotation period; $`n_\mathrm{c}`$ is the central baryon density; $`\rho _\mathrm{c}`$ is the central proper energy density divided by $`c^2`$; $`H_\mathrm{c}`$ is the central log-enthalpy \[Eq. (17)\]; $`R_{\mathrm{eq}}`$ is the circumferential radius, i.e. the length of the equator \[as given by the metric (6)\] divided by $`2\pi `$; $`r_{\mathrm{eq}}`$ is the equator $`r`$ coordinate; $`r_{\mathrm{pole}}/r_{\mathrm{eq}}`$ is the coordinate oblateness of the star; $`I`$ is the moment of inertia (defined as $`J/\mathrm{\Omega }`$); $`J`$ is the star angular momentum; $`Q`$ is the quadrupole moment defined as in Salgado et al. (1994a) or Laarakkers & Poisson (1999); $`T/W`$ is the “kinetic to gravitational energy” ratio (see Sect. 6); $`U_{\mathrm{eq}}`$ is the rotation velocity at the equator as measured by a locally non-rotating observer (those observer whose 4-velocity is the normal to the $`t=\mathrm{const}`$ hypersurfaces); $`z_{\mathrm{eq}}^\mathrm{f}`$ is the redshift for an emission at the equator and in the direction of rotation, $`z_{\mathrm{eq}}^\mathrm{b}`$ the redshift for an emission at the equator and in the direction opposite to rotation, $`z_{\mathrm{pole}}`$ the redshift at the stellar pole; $`N_\mathrm{c}`$, $`N_\mathrm{c}^\phi `$, $`A_\mathrm{c}`$ and $`B_\mathrm{c}`$ are the metric potentials \[Eq. (6)\] at the stellar center. Finally GRV2 and GRV3 are the two virial error indicators discussed in Sect. 3.5. ### 4.2 Evolutionary sequences A strange star which slowly loses energy and angular momentum via electromagnetic or gravitational radiation keeps its total baryon number constant. Therefore, we can compute evolutionary sequences of strange stars as sequences at fixed baryon mass $`M_\mathrm{B}`$. Similarly to neutron stars, two categories of strange stars can be distinguished: normal stars, which have a baryon mass lower than the maximum baryon mass of static configurations $`M_{\mathrm{B},\mathrm{max}}^{\mathrm{stat}}=2.625B_{60}^{1/2}\mathrm{M}_{}`$ (cf. Table 1), and supramassive stars, which have a baryon mass greater than $`M_{\mathrm{B},\mathrm{max}}^{\mathrm{stat}}`$. Any normal star belongs to an evolutionary sequence which terminates at a static configuration. On the contrary, supramassive stars exist only by virtue of rotation. The two families clearly appear in Fig. 6, which is a plot of the angular momentum as a function of the rotation frequency for evolutionary sequences. The supramassive sequences are not connected with the static limit $`\mathrm{\Omega }=0`$. At both ends, they terminate by Keplerian configuration. The limiting case $`M_\mathrm{B}=M_{\mathrm{B},\mathrm{max}}^{\mathrm{stat}}`$ is denoted by the thick line in Fig. 6. A characteristic feature of some supramassive sequences also appears in Fig. 6: they can be spun-up by angular momentum loss: $`d\mathrm{\Omega }/dJ<0`$. This effect is well known for Newtonian stars with soft EOS (adiabatic index close to $`4/3`$) (Shapiro et al. 1990), as well as for (relativistic) neutron stars (Cook et al. 1994, Salgado et al 1994a). As can be seen from Fig. 6, this effect is very prononced for strange stars. Beside the fact that they can represent evolution path of rotating strange stars, another motivation for computing constant baryon number configurations is that this permits a stability analysis, as discussed in Sect. 5.1 below. ### 4.3 Comparison with previous works First of all, in the non-rotating case, we recover the results presented in Haensel et al. (1986) for massless strange quarks \[compare the $`M_{\mathrm{max}}^{\mathrm{stat}}`$ model in Table 1 with Eq. (28) of Haensel et al. (1986)\]. The first rapidly rotating model of a strange star has been computed by Friedman (unpublished, quoted in Glendenning (1989a,b)). The obtained maximum angular velocity is $`6\%`$ lower than ours (last column in Table 1)<sup>3</sup><sup>3</sup>3Friedman’s result is presented for $`\stackrel{~}{B}^{1/4}:=(\mathrm{}c)^{3/4}B^{1/4}=170\mathrm{MeV}`$, which corresponds to $`B=108.7\mathrm{MeV}\mathrm{fm}^3`$. It must be rescaled to $`B=60\mathrm{MeV}\mathrm{fm}^3`$, according to the law (20), in order to be compared with results listed in Table 1.. The corresponding gravitational mass is $`7\%`$ lower than ours, and the central energy density is $`7\%`$ greater than ours. The only other rapidly rotating model of strange star published in the literature is that of Lattimer et al. (1990). The maximum angular velocity obtained by these authors is $`2\%`$ lower than ours (last column in Table 1) <sup>4</sup><sup>4</sup>4Lattimer et al. (1990) results (their Table 7) is presented for $`B=1.634\times 56=91.5\mathrm{MeV}\mathrm{fm}^3`$. It must be rescaled to $`B=60\mathrm{MeV}\mathrm{fm}^3`$, according to the law (20), in order to be compared with results listed in Table 1.. The corresponding mass is $`12\%`$ lower than ours, the circumferential radius $`R_{\mathrm{eq}}`$ $`14\%`$ lower, the ratio $`T/W`$ $`10\%`$ lower, and the polar redshift (linked to the injection energy $`\beta `$ given by Lattimer et al. (1990) by $`z_{\mathrm{pole}}=\beta ^{1/2}1`$) $`15\%`$ lower. The largest discrepancy arises for the baryon mass $`M_\mathrm{B}`$: the value reported by Lattimer et al. (1990) is $`21\%`$ lower than ours. However we suspect some error in the presentation of the results by Lattimer et al. (1990), because (i) their ratio $`M_\mathrm{B}/M`$ is only $`1.18`$, whereas it is as large as $`1.34`$ for the non-rotating maximum mass model (for comparison, our value of $`M_\mathrm{B}/M`$ for the $`\mathrm{\Omega }_{\mathrm{max}}`$ model is $`1.35`$), (ii) in the neutron star case, large discrepancies with baryon masses presented in Lattimer et al. (1990) have been noticed previously by Salgado et al. (1994a) \[see Sect. 5.1.2 of Salgado et al. (1994a)\]. Note that the numerical code employed by Salgado et al. (1994a) has been successfully compared with codes from two other groups in Nozawa et al. (1998) and did not show any trouble with the baryon mass. Unfortunately, the baryon mass of Friedman’s model is not reported in Glendenning (1989a,b). Apart from the baryon mass problem discussed above, the discrepancy with Friedman and Lattimer et al. (1990) results remains rather large. It may be explained as follows. Both Friedman’s code \[presented in Sect. II.a of Friedman et al. (1986)\] and Lattimer et al. (1990) code employ the Butterworth & Ipser (1976) finite difference technique for numerically solving the partial differential equations resulting from Einstein’s equations. Butterworth & Ipser (1976) considered explicitly the case of a strong discontinuity in the density field at the stellar surface \[their Sect. III.c)ii)\]. However they treat it only along the radial direction (by using a modified Lagrangian-polynomial fit in $`r`$) and keep a Legendre expansion for the angular variable $`\mu :=\mathrm{cos}\theta `$. This is correct as long as the stellar surface is exactly spherical, i.e. the star is static. But when the surface deviates from spherical symmetry (as is the case for rapidly rotating stars, cf. Figs. 2 and 4), there is also a discontinuity in the $`\theta `$ direction. This discontinuity, when described by means of Legendre polynomials, which are smooth functions, inevitably generates spurious oscillations — the so-called Gibbs phenomenon. We believe that it is this Gibbs phenomenon which explains the discrepancy between Friedman and Lattimer et al. (1990) results and ours. ## 5 Maximally rotating configuration and scaling with $`B`$ ### 5.1 Maximally rotating stable configuration A natural bound on $`\mathrm{\Omega }`$ results from the mass shedding condition, corresponding to the Keplerian angular velocity, $`\mathrm{\Omega }_\mathrm{K}`$ discussed in Sect. 3.4: stationary configurations of uniformly rotating strange stars with $`\mathrm{\Omega }>\mathrm{\Omega }_\mathrm{K}`$ do not exist. We will also require, that the stationary, uniformly rotating configuration be stable with respect to axisymmetric perturbations (see e.g. Sect. 2.5.1 of Friedman 1998). We assume that the strange star is sufficiently hot, so that it is not subject to the r-mode instabilities (Madsen 1998). The maximum value of $`\mathrm{\Omega }`$ for stationary, uniformly rotating models of strange stars which satisfy these two conditions, will be denoted by $`\mathrm{\Omega }_{\mathrm{max}}`$, and the corresponding configuration will be called a maximally rotating one. The stability with respect to axisymmetric perturbations can be investigated by means of the turning point theorem established by Friedman et al. (1988). We use the following variant of it \[also used by Baumgarte et al. (1998)\]: along an evolutionary sequence (i.e. a sequence at constant baryon number, Sect. 4.2), the change in stability is reached when the gravitational mass reaches a minimum. Using this criterion, we have determined the boundary separating stable and unstable configurations by linking the points of mimimum $`M`$ along the constant baryon mass sequences in Fig. 5: the configurations located to the left (resp. to the right) of the dashed line in Fig. 5 are stable (resp. unstable) with respect to axisymmetric perturbations. Moreover, the intersection of this line with the $`\mathrm{\Omega }=\mathrm{\Omega }_\mathrm{K}`$ line in Fig. 5 marks the maximum angular velocity allowable for stable rotating strange stars, i.e. the $`\mathrm{\Omega }_{\mathrm{max}}`$ configuration. It is denoted by a square and is listed in the last column of Table 1. As noted by Cook et al. (1994) and Stergioulas & Friedman (1995), the $`\mathrm{\Omega }_{\mathrm{max}}`$ configuration and the configuration with maximum gravitational mass, $`M_{\mathrm{max}}^{\mathrm{rot}}`$, do not coincide, although they are close to each other for typical neutron star models. We also found this for rotating strange stars, and the effect is more pronounced than in the case of neutron stars: it can be seen easily in Fig. 5. The parameters of the $`M_{\mathrm{max}}^{\mathrm{rot}}`$ and $`\mathrm{\Omega }_{\mathrm{max}}`$ configurations are listed in the last two columns in Table 1. $`\mathrm{\Omega }_{\mathrm{max}}`$ is $`4\%`$ higher than $`\mathrm{\Omega }(M_{\mathrm{max}}^{\mathrm{rot}})`$. Note that the $`M_{\mathrm{max}}^{\mathrm{rot}}`$ configuration is on the stable side, so that the situation is similar to that presented in Fig. 9 of Stergioulas & Friedman (1995), or Fig. 8 of Koranda et al. (1997). ### 5.2 Formulæ for $`\mathrm{\Omega }_{\mathrm{max}}`$ The scaling properties of the equation of the stationary motion imply that $`\mathrm{\Omega }_{\mathrm{max}}B^{1/2}`$. The calculation of $`\mathrm{\Omega }_{\mathrm{max}}`$ involves locating the threshold for the instability with respect to the axisymmetric perturbations, and requires therefore a higly precise numerical code for the 2-D calculations of stationary configurations. Therefore, the precision with which the scaling formula for $`\mathrm{\Omega }_{\mathrm{max}}`$ is fulfilled reflects the overall precision of our numerical calculations. Our numerical results for $`\mathrm{\Omega }_{\mathrm{max}}`$ can summarized in exact formulae $$\mathrm{\Omega }_{\mathrm{max}}=\mathrm{9.92\hspace{0.33em}10}^3\sqrt{B_{60}}\mathrm{rad}\mathrm{s}^1,P_{\mathrm{min}}=\frac{0.634}{\sqrt{B_{60}}}\mathrm{ms}.$$ (24) We checked that the scaling $`\mathrm{\Omega }_{\mathrm{max}}\sqrt{B}`$ holds with a very high precision. Within our simple model of EOS of strange matter, the constraints on the bag constant $`B`$, Sect. 2, imply $$0.513\mathrm{ms}<P_{\mathrm{min}}<0.640\mathrm{ms}.$$ (25) In the case of dense baryon matter, exact results for the maximum rotation frequency of uniformly rotating neutron star models, for a very broad set of realistic causal EOS, show an interesting correlation with values of the mass and radius of the static configuration with maximum allowable mass, $`M_{\mathrm{max}}^{\mathrm{stat}}`$, $`R_{M_{\mathrm{max}}}^{\mathrm{stat}}`$ (Haensel & Zdunik 1989, Shapiro et al. 1989, Friedman et al. 1989, Friedman 1989, Haensel et al. 1995). The most recent form of such an “empirical formula” for $`\mathrm{\Omega }_{\mathrm{max}}`$, derived in (Haensel et al. 1995), reads $`\mathrm{\Omega }_{\mathrm{max}}`$ $``$ $`C_{\mathrm{NS}}\left({\displaystyle \frac{M_{\mathrm{max}}^{\mathrm{stat}}}{\mathrm{M}_{}}}\right)^{\frac{1}{2}}\left({\displaystyle \frac{R_{M_{\mathrm{max}}}^{\mathrm{stat}}}{10\mathrm{km}}}\right)^{\frac{3}{2}},`$ $`C_{\mathrm{NS}}`$ $`=`$ $`7730\mathrm{rad}\mathrm{s}^1,`$ (26) where the prefactor $`C_{\mathrm{NS}}`$, which does not depend on the EOS of dense baryon matter, has been obtained via fitting “empirical formula” to exact numerical results for realistic, causal EOS of baryon matter. The “empirical formula” reproduces exact results with a surprisingly high precision, the relative deviations from Eq. (26) not exceeding 5%. In the case of strange stars, built of strange matter of massless quarks, the formula of the type (26) is exact, albeit with a different numerical prefactor. Using our numerical results, we get $`C_{\mathrm{SS}}=\mathrm{7.84\; 10}^3\frac{\mathrm{rad}}{\mathrm{s}}`$, so that $`C_{\mathrm{SS}}C_{\mathrm{NS}}`$ within better than 2%. Therefore, “empirical formula”, derived originally for (ordinary) neutron stars, holds also, with very good precision, for strange stars. Our conclusion agrees with unpublished result of Friedman (quoted in Glendenning 1989a), and with Lattimer et al. (1990). ### 5.3 Formulæ for maximum mass configurations Rotation increases maximum allowable mass of strange stars, and the equatorial radius of the maximum mass configuration. Our results for the maximum mass of rotating configurations, $`M_{\mathrm{max}}^{\mathrm{rot}}`$, and its equatorial radius, $`R_{\mathrm{eq},M_{\mathrm{max}}^{\mathrm{rot}}}`$, can be summarized in two exact formulae, $`M_{\mathrm{max}}^{\mathrm{rot}}`$ $`=`$ $`{\displaystyle \frac{2.831}{\sqrt{B_{60}}}}\mathrm{M}_{},`$ $`R_{\mathrm{eq},M_{\mathrm{max}}}^{\mathrm{rot}}`$ $`=`$ $`{\displaystyle \frac{16.54}{\sqrt{B_{60}}}}\mathrm{km}`$ (27) The ratio of the maximum mass of rotating configurations to that of static ones, and the ratio of the corresponding equatorial radii, are thus independent of $`B`$, $$\frac{M_{\mathrm{max}}^{\mathrm{rot}}}{M_{\mathrm{max}}^{\mathrm{stat}}}=1.44,\frac{R_{\mathrm{eq},M_{\mathrm{max}}}^{\mathrm{rot}}}{R_{M_{\mathrm{max}}}^{\mathrm{stat}}}=1.54.$$ (28) It has been pointed out by Lasota et al. (1996), that for realistic causal baryonic EOS the ratios $`M_{\mathrm{max}}^{\mathrm{rot}}/M_{\mathrm{max}}^{\mathrm{stat}}`$ and $`R_{\mathrm{eq},M_{\mathrm{max}}}^{\mathrm{rot}}/R_{M_{\mathrm{max}}}^{\mathrm{stat}}`$ are very weakly dependent on the EOS, and can be very well approximated (within a few percent) by constants $`C_M=1.18`$ and $`C_R=1.34`$, respectively. In the case of strange stars the values of $`C_M`$ and $`C_R`$ are significantly higher; rotation increases the value of $`M_{\mathrm{max}}`$ by 44% compared to 18% for neutron stars, while the corresponding increase in the equatorial radius is 54% compared to 34% for neutron stars. We conclude that“empirical formula” for $`\mathrm{\Omega }_{\mathrm{max}}`$ has more universal character than the “empirical formulae” for $`M_{\mathrm{max}}^{\mathrm{rot}}`$, $`R_{\mathrm{eq},M_{\mathrm{max}}}^{\mathrm{rot}}`$, proposed in Lasota et al. (1996); the former applies both for neutron and strange stars, while the latter describe neutron stars only. ## 6 About triaxial instabilities of rapidly rotating strange stars ### 6.1 Known results on the threshold in $`T/W`$ for triaxial instabilities Rapidly rotating neutron stars can be subject to a spontaneous breaking of their symmetry about the rotation axis, resulting from some triaxial instability, when their rotation rate exceeds a certain threshold (see e.g. Bonazzola & Gourgoulhon (1997), Friedman (1998) or Stergioulas (1998) for a review). In order for the symmetry breaking to take place, a (secular) dissipative mechanism has to be working. Basically two such mechanisms can be contemplated: (i) viscosity and (ii) coupling with gravitational radiation (Chandrasekhar-Friedman-Schutz (CFS) instability). For compact stars that are accelerated by accretion in X-ray binary systems, these triaxial instabilities may be an important source of gravitational waves, in the frequency band of the LIGO/VIRGO interferometric detectors currently under construction. Another situation where these triaxial instabilities might develop is in compact stars newly formed after a stellar core gravitational collapse. In the present paper, we have restricted ourselves to stationary and axisymmetric rotating strange star models. Therefore, we could not compute the triaxial instability threshold of rotating strange stars. However, an indicator of the stability of a rotating self-gravitating body is its kinetic to gravitational potential energy ratio $`T/W`$, which can be computed for any equilibrium configuration. For instance, it is well known that a homogeneous Newtonian rotating body (MacLaurin spheroid) becomes secularly unstable with respect to triaxial $`\mathrm{}=m=2`$ perturbations (bar mode) if $`T/W>0.1375`$ (Jacobi/Dedekind bifurcation point in the MacLaurin sequence). For compressible bodies (still in the Newtonian regime), this threshold can be lowered to $`0.1275`$ (Bonazzola et al. 1996). The threshold of the CFS instability decreases with the mode number $`m`$ for polar modes: for a Newtonian polytrope of abiabatic index $`\gamma =2`$, one has $`T/W>0.079`$ for a $`m=3`$ polar mode, $`T/W>0.058`$ for $`m=4`$ and $`T/W>0.046`$ for $`m=5`$ (Stergioulas & Friedman 1998). Higher order modes are likely to be damped out by viscosity. For axial r-modes, the CFS instability is generic, i.e. it occurs at any rotation rate (Andersson 1998). Madsen (1998) has however argued that the r-mode instability is suppressed by the quark matter bulk viscosity in new born strange stars. In the relativistic case, a ratio $`T/W`$ can be defined according to Friedman et al. (1986) prescription. This quantity reduces to the usual kinetic to gravitational energy ratio at the Newtonian limit, whereas its physical interpretation in the relativistic case is not so clear (in particular the value of $`T/W`$ is coordinate dependent). However it can be used to measure the importance of rotational effects. For the viscosity-driven instability, general relativistic terms increase the threshold on $`T/W`$: it becomes as high as $`T/W>0.26`$ for a compactification parameter $`M/R=0.2`$, according to the post-Newtonian analysis of Shapiro & Zane (1998). This stabilizing effect of general relativity onto the viscosity-driven bar instability has also been found by Bonazzola et al. (1998a). On the contrary, general relativity decreases the threshold of the CFS instability, down to $`T/W>0.065`$ for $`m=2`$ polar mode (this mode being always stable in the Newtonian regime for $`\gamma =2`$), $`T/W>0.046`$ for $`m=3`$, $`T/W>0.035`$ for $`m=4`$, and $`T/W>0.029`$ for $`m=5`$, for a $`\gamma =2`$ polytrope \[Stergioulas & Friedman 1998, see also Yoshida & Eriguchi (1997)\]. ### 6.2 $`T/W`$ ratio of rotating strange stars Having the above results in mind, we have computed the $`T/W`$ ratio according to the prescription of Friedman et al. (1986), for each of our rotating stange star models. In particular, Fig. 7 shows the value of $`T/W`$ as a function of the rotation frequency along evolutionary sequences. It appears clearly that the values of $`T/W`$ for strange stars are much greater than for neutron stars (compare e.g. with Fig. 4 of Cook et al. (1994)). In particular, for the $`\mathrm{\Omega }_{\mathrm{max}}`$ configuration, $$\left(\frac{T}{W}\right)_{\mathrm{\Omega }_{\mathrm{max}}}=0.201,$$ (29) (it does not depend on $`B`$) whereas it ranges from $`0.105`$ to $`0.139`$ for the set of neutron star EOS examined in Nozawa et al. (1998). The fact that $`T/W`$ for maximally rotating strange stars is significantly larger than that for neutron stars, has been noted by Lattimer et al. (1990) and Colpi & Miller (1992). Moreover, contrary to neutron stars, the value of $`T/W`$ remains very high, of the order of $`0.2`$, for rapidly rotating strange stars of moderate mass: the value $`T/W=0.2`$ is reached at a rotation period $`P=0.86\mathrm{ms}`$ (resp. $`P=0.85\mathrm{ms}`$) for a baryon mass of $`M_\mathrm{B}=1.60\mathrm{M}_{}`$ (resp. $`M_\mathrm{B}=1.80\mathrm{M}_{}`$), which corresponds to a gravitational mass $`M=1.31\mathrm{M}_{}`$ (resp. $`M=1.47\mathrm{M}_{}`$) (notice that in the static case $`M_\mathrm{B}=1.6,1.8\mathrm{M}_{}`$ for normal sequences in Fig. 7 correspond to gravitational masses $`M=1.27,1.42\mathrm{M}_{}`$, respectively). All this seems to indicate that triaxial instabilities could develop more easily in rotating strange stars. However, it has to be stressed that no actual stability analysis has been performed yet for strange stars which, in contrast to ordinary neutron stars, are bound not only by gravity, but also by confinement forces. These confinement forces are represented represented by the bag pressure, $`B`$, acting on the strange star surface. The density profiles, Fig. 3, and adiabatic index $`\gamma `$, Fig. 1, are qualitatively different from those characteristic of rotating neutron star models, for which triaxial instabilities have been studied. ## 7 Discussion and conclusions The numerical method used in the present paper is particularly suitable for exact calculations of models of rapidly rotating strange stars. In contrast to numerical approaches used in previous calculations, the multi-domain method enabled us to treat exactly the huge density discontinuity at the surface of strange stars. We used three-domain grid in both $`r`$ and $`\theta `$ variables, adapted (adjusted) via a self-consistent procedure to the stationary shape of rotating strange star surface. The precision of our calculations was checked using several tests. We found significant differences between our results, and those obtained by other authors in previous exact calculations of the models of maximally rotating strange star models. As noted in the early papers on the properties of strange stars, models of static massive ($`M\mathrm{M}_{}`$) strange stars are characterized by global stellar parameters (mass, radius, moment of inertia, surface redshift) which are rather similar to those of ordinary neutron stars (Haensel et al. 1986, Alcock et al. 1986). In contrast to the static case, rapidly rotating strange stars exhibit qualitative differences with respect to rapidly rotating neutron stars. Strange stars rotating close to the break up (Keplerian) frequency have kinetic to gravitational energy ratio $`T/W0.2`$, which is nearly twice larger than corresponding values for neutron stars. Moreover, this unusually high $`T/W0.2`$ is characteristic not exclusively for supermassive rotating models, but is typical also for normal rotating configurations with baryon mass lower than the maximum allowed mass for static models. In particular, $`T/W0.2`$ is reached for the $`1.4\mathrm{M}_{}`$ models rotating at $`P0.85`$ms. Large value of $`T/W`$ stems from a flat density profile combined with strong equatorial flattening of rapidly rotating strange stars. These particular features of rapidly rotating strange stars are consequences of a special character of EOS of strange quark matter in the strange star interior, reflected by a specific density dependence of the adiabatic index. Strange matter inside massive ($`M>\mathrm{M}_{}`$) strange stars is very stiff in the outer (surface) layers and soft in the central core. Large value of $`T/W`$ may signal a relative softness (susceptibility) of rapidly rotating strange stars to triaxial instabilities. For simple strange matter EOS, used in the present paper, several exact scaling relations hold for extremal (maximum frequency, maximum mass) configurations of rotating strange stars. Maximum rotation frequency as well as maximum mass of rotating configurations exhibit an exact scaling with respect to the value of the bag constant. Empirical formula for $`\mathrm{\Omega }_{\mathrm{max}}`$ for neutron stars, which relates this quantity to the mass and radius of static configuration with maximum allowable mass, holds with a high precision also for strange stars. Minimum period of rotating strange stars, stable with respect to axisymmetric perturbations, obtained for an acceptable range of the bag constant, is $`0.5`$ms, and therefore similar to that characteristic of realistic models of neutron stars. However, the effect of rapid rotation on the maximum allowable mass is significantly larger for strange stars than for neutron stars. Uniform rotation can increase the massimum allowable mass of strange stars by more than $`40\%`$, to be compared with only 20% increase for realistic models of neutron stars. It should be stressed, that our numerical results for rapidly rotating strange stars, summarized in Eqs.(24,25), and Eqs.(27,28,29), have been obtained for a schematic EOS, Eq.(3). This particular EOS has been selected for the sake of simplicity and, last but not the least, for numerical convenience. The most general - generic - feature of an EOS of strange matter is the existence of a self-bound state at zero pressure, with $`E_0<930.4`$ MeV. However, the actual EOS of strange matter - should it exist - is expected to be different from that given by Eq.(3). Therefore, results of the present paper neither represent the full range of possibilities for “theoretical” strange stars, nor they are expected to correspond to precise actual values of parameters of rapidly rotating strange stars - should they exist in Nature. The equation of state of strange matter, used in the present paper, has been derived neglecting strange quark mass and representing all effects of QCD interactions by one parameter - the bag constant $`B`$. However, this simplest (minimal) model exhibits generic properties of the EOS of a self-bound quark matter. Generally, inclusion of perturbative corrections, expressed in terms of the QCD coupling constant $`\alpha _\mathrm{c}`$, and of the strange quark mass $`m_s`$, imply shifting down and simultaneous narrowing of the window of the values of $`B`$ compatible with strange matter hypothesis (Farhi & Jaffe 1984). Typical values of $`m_s`$ are $`200\mathrm{MeV}/c^2`$; the effect of $`\alpha _\mathrm{c}0.3`$ on the EOS is then much smaller than that of $`m_s`$. Scaling relations between extremal strange star configurations, corresponding to different values of $`B`$ (at fixed values of $`\alpha _\mathrm{c}`$, $`m_s`$) are than no longer exact. In the case of static strange star models, they are still very precise (after the prefactors in these relations have been recalculated to include the effect of $`\alpha _\mathrm{c}`$ and $`m_s`$) and therefore useful (Haensel et al. 1986); similar situation is expected to take place in the case of rotating strange star models. In the static case, the effect of non-zero $`m_s`$ and $`\alpha _\mathrm{c}`$ consisted in making the maximum mass configurations less massive but more compact, and we expect similar effect for the extremal configurations of rotating strange stars. In this paper we restricted ourselves to the case of strange stars with superdense quark surface, built exclusively of strange quark matter (bare strange stars). In principle, one should consider also models of strange stars, covered by an envelope (crust) consisting of nuclei immersed in electron gas (Alcock et al. 1986). The nuclei, forming a crystal lattice of the crust, are separated from strange matter by a repulsive coulomb barrier. Both coulomb barrier, and characteristic spatial gap between nuclei and strange matter, result from an electric dipole layer on the quark surface, due to the nonuniform density distribution of electrons near the quark surface. The presence of electrons in strange matter, necessary for equilibrium of quark core core with crust, results from nonzero strange quark mass. The maximum mass of the crust on a $`M>\mathrm{M}_{}`$ strange star was originally estimated as $`10^5\mathrm{M}_{}`$ (Alcock et al. 1986). Further studies of the crust–strange matter coexistence conditions led to an even lower value of the maximum mass of the crust on a $`M>\mathrm{M}_{}`$ strange star, of the order of $`10^6\mathrm{M}_{}`$ (Huang & Lu 1997). In the case of the static strange star models, the effect of the presence of the crust on the maximum allowable mass configuration is very small. Rapid rotation will increase the maximum mass of a strange star crust (Glendenning & Weber 1992), but still its effect on the maximally rotating and maximum mass configurations may be expected to be small: both $`\mathrm{\Omega }_{\mathrm{max}}`$ and $`M_{\mathrm{max}}^{\mathrm{rot}}`$ will be slightly reduced, as compared with the case of bare strange stars. It should be stressed, that the only existing calculation of rapid rotation of strange stars with crust (Glendenning & Weber 1992) was performed within a version of slow rotation approximation. The multi-domain method with adaptable grid, used in the present paper, is also suitable for an exact treatment of rapid rotation of strange stars with crust, in which matter density exhibits a huge discontinuity at the quark matter–crust interface, with outward density drop by more than three orders of magnitude. This problem is now being investigated. ###### Acknowledgements. During his stay at DARC, Observatoire de Paris, P. Haensel was supported by the PAST professorship of French MENRT. This research was partially supported by the KBN grant No. 2P03D.014.13. The numerical calculations have been performed on computers purchased thanks to a special grant from the SPM and SDU departments of CNRS.
no-problem/9907/cond-mat9907381.html	ar5iv	text	# Waves on Icicles ## Abstract Icicles with wave patterns on their surfaces can sometimes be seen hanging from roofs of buildings. Surprisingly, most of these wave patterns are at intervals of about 1 cm. The reason for this uniformity of interval has not been clarified. Here we show a formula to explain this remarkable phenomenon by introducing a new instability theory. This theory is given by thermal diffusion in thin water layer streams flowing along the icicles. The streams change the temperature distribution and control waves of short wavelengths. The specific wavelength (about 1 cm) can be determined by Laplace instability of the heat field in the atmosphere and by the thermal diffusion effect in thin-layer streams. Icicles are often seen hanging from roofs of buildings in winter season in snowy regions. They are almost columnar in shape and their radii (or thickness) become smaller in the gravitational direction. The precise inner morphology of icicle is much more complicated and interesting like having hollows, which is shown by Maeno et.al. 1 . But in this article we discuss only the surface instability of icicle as the periodically modulated thickness of an icicle along its longitudinal direction. The intervals of wave (wavelength) appearing on icicles are almost uniform (about 1 cm), and do not depend on the length or thickness of the icicle or on the external temperature. Furthermore, such waves appear only on the surface of icicles that are enveloped by a layer of water flow of about 0.1 mm in thickness. The surface of a growing icicle changes from flat to wavy in a few hours matsu . Formation of such waves is also discussed qualitatively by Maeno et.al. 1 as the conjecture. But its scenario can not explain both the value of wavelength and its universality. The mechanism to construct the wave here we show, is different from above one. This kind of phenomenon is usually called surface instability. In this short article, we will discuss both the instability of the surface of an icicle and factors that determine the wavelength appearing on the surface of an icicle. In 1997 Matsuda has given the experiment to construct the wavy ice matsu in which the physical situation is the same as icicle. He has made a gutter inclined at $`\theta `$ degrees and kept the static water flow on it. He put the apparatus in cold room with temperature -8 degrees centigrade. Then the wavy ice has appeared under the water flow and its wave length is about 8 mm at $`\theta =\pi /2`$, which agrees with the case of icicle matsu . This means the waves appearing on icicle is induced by the interaction between thin water flow and thermal diffusion. In the study of crystal growth, surface instability is usually explained by Mullins-Sekerka’s theory 2 , which is based on two observations: Laplace instability and Gibbs-Thomson effect. However, the mechanism underlying the instability that occurs on icicles cannot be explained by this theory. The reason for this is that the Gibbs-Thomson effect only applies to wavelength of about 10 micrometers; i.e., wavelength much shorter than those on icicles (about 1 cm). Furthermore, Laplace instability does not occur in a thin liquid layer; it only occurs when the thickness of the layer is wider than the wavelength. Thus, Laplace instability could only occur on icicles if the thickness of liquid layer were larger than the wavelength of 1 cm, which is never the case. Let us then consider the characteristics of a fluid. There are various instability theories in fluid mechanics. Benney’s liquid film has often been used to study the characteristics of a thin water layer 3 . This theory explains why waves appear on the surface of thin water flow on a sloping road on a rainy day. This phenomenon is caused by gravity, viscosity and surface tension. However,these are not static waves but moving waves. In the case of icicles, waves move down an icicle two-times faster than the average velocity of a fluid. Thus, the formation of waves on an icicle can not be explained only on the basis of characteristics of such hydrodynamical waves. We therefore consider not only the fluid characteristics which flows along an icicle but also the thermal diffusion effect to express the static wave on the surface of icicle. Our program is the following. First, we solve the hydrodynamics of thin-layer fluid, which is not used in the usual Mullins-Sekerka theory. Next, we consider the thermal diffusion in thin-layer fluid and in the air surrounding an icicle. The thermal diffusion in air is very important since Laplace instability cannot occur in a water layer. As a boundary condition, the heat flow should be conserved on the boundary between air and the water. The thermal properties in water fluid and in air give all the necessary information on the formation of surface of icicles. Let us start with the case of an icicle having a boundary surface (ice-water) with a small perturbation. We add a perturbation to the radius $`R`$ of an infinitesimally long columnar ice (model of icicle), $$RR+\delta (t)\mathrm{sin}kx,$$ (1) (x-axis is along the length of the icicle) and we consider the time development of $`\delta (t)`$. We define the y-axis as perpendicular to the ice surface, and use the variable $`y=rR`$ with $`y<<R`$. The thin water layer on this surface has mean thickness $`h_00.1mm`$. For considering such a film fluid, it will be easier to have a fluid dynamical model as thin water layer on a inclined gutter just like Matsuda’s experiment matsu . (x-axis is taken along slope, and y-axis is taken normal to slope surface.) This simplification is well established since $`h_0R`$. This situation for fluid mechanics is usually discussed for flat gutter, and shown the specific surface velocity $$U=\frac{gh_0^2}{2\nu },$$ (2) with parabolic velocity distribution, $$v_x=U(2\frac{y}{h_0}(\frac{y}{h_0})^2),$$ (3) where $`\nu `$ is the viscosity ($`1.8\times 10^6m^2/s`$) and $`g`$ is the gravitational acceleration. (For this calculation, see for e.g. the exercise in Landau’s text book 3 .) If we observe mean surface velocity $`U`$, we obtain $`h_0`$ and flow quantity $`𝐐=2\pi R\overline{U}h_0`$, where $`\overline{U}=\frac{1}{h_0}_0^{h_0}v_x𝑑y=2U/3`$ is the mean velocity. In Matsuda’s experiment matsu , he used the flow $`𝐐=160ml/hr`$ and width of gutter $`W=3cm`$ ($`W`$ corresponds to $`2\pi R`$), where the wave appears most clear. This gives $`\overline{U}h_0=1.48\times 10^6m^2/s`$. From $`\overline{U}=2U/3`$ and $`U=\frac{gh_0^2}{2\nu }`$, we get $`U=2.4\times 10^2m/s`$ with $`h_0=0.93\times 10^4m`$. On the other hand, his direct observation of surface mean velocity by using test fine grain gives $`U=4\times 10^2m/s`$. So we use $`U=(2.44)\times 10^2m/s`$ with $`h_0=(0.931.21)\times 10^4m`$ as the experimental surface velocity and thickness of water layer. Our problem with small modulation $`\delta \mathrm{sin}kx`$ on boundary can be expressed as perturbation to above solution. To do this we make nondimensional coordinate $`x_{}`$ and $`y_{}`$ as $$x_{}=kx,y_{}=y/h_0,$$ and we put off * in a while. In this notation solid surface is given as $`y=\eta \mathrm{sin}x`$, where $`\eta =\delta /h_0`$. The fluid in such a thin-layer has boundary conditions as follows. The fluid velocity equals to zero at the ice-water boundary due to viscosity, and the gradient of fluid velocity equals to zero on the air-water surface due to the vanishing of shearing stress. By solving static Navier Stokes equation 4 with above mentioned boundary conditions and with long wavelength approximation $`\mu kh_010^2`$, we obtain the hight of air-water boundary surface in nondimensional form given as, $$y=1+\eta \mathrm{sin}x,$$ (4) which means the thickness of water layer is uniform at lowest order of long wavelength approximation. (In first order approximation, we have modulation in its thickness but it is not thinner at protrusions. The thickness modulation is like $`\mathrm{cos}kx`$ against the icicle modulation: $`\mathrm{sin}kx`$. But these effects are irrelevant. 4 ) The stream function of water is obtained as $$\psi _{}=\frac{1}{3}(y\eta \mathrm{sin}x)^3+(y\eta \mathrm{sin}x)^2,$$ (5) where $`\psi _{}`$ is also the non-dimensional stream function as, $$\psi =h_0U\psi _{},$$ and we put off * again. The form of such stream function is not surprising, and it means just the parabolic form of velocity fields, where the velocity vanishes at modulated solid-water boundary. The next task is to solve the thermal diffusion equation with a flow background as, $$\mathrm{}T\frac{\stackrel{}{v}}{D}\stackrel{}{}T=\frac{1}{D}\frac{T}{t},$$ (6) where $`D\frac{\kappa }{\rho c}1.3\times 10^7m^2/s`$ is the thermal diffusion constant and we used non-compressive condition. We rewrite above form by using stream function $`\psi `$, neglecting higher order in parameter $`\mu kh_0`$ as long wavelength approximation, and under the static thermal distribution condition. (All variables are written in dimensionless form except temperature.) $$\frac{^2T}{y^2}=\alpha (\frac{T}{x}\frac{\psi }{y}\frac{T}{y}\frac{\psi }{x}).$$ (7) Note that the dimensionless parameter $`\alpha \mu Uh_0/D`$ is proportional to wave number k. The value of $`\alpha `$ is about 2 experimentally. By using boundary condition on solid-water surface ($`y=\eta \mathrm{sin}x`$)as $`T(SW)`$ $`=`$ $`T_M=const.,`$ (8) $`Q(SW)`$ $`=`$ $`\kappa {\displaystyle \frac{T}{y}}_{y=\eta \mathrm{sin}x}\kappa a(x),`$ (9) we obtain the quantities on air-water surface ($`y=1+\eta \mathrm{sin}x`$) by solving differential equation. $`T(AW)`$ $`=`$ $`T_M+[1+{\displaystyle \frac{7}{60}}\widehat{D}+{\displaystyle \frac{13}{3360}}\widehat{D}^2]a(x),`$ (10) $`Q(AW)`$ $`=`$ $`\kappa [1+{\displaystyle \frac{5}{12}}\widehat{D}+{\displaystyle \frac{239}{10080}}\widehat{D}^2]a(x),`$ (11) where $`T_M`$ is the melting point, $`Q`$’s are heat flows, $`a(x)`$ is an unknown gradient of temperature on solid surface, and $`\widehat{D}`$ means $`\alpha d/dx`$.The higher order terms of $`\alpha `$ are negligible since their coefficients are effectively small. It should be noted that the temperature of the air-water surface is different from the one in the environment. The temperature in atmosphere changes greatly near the surface of an icicle. Thus, we must also consider the thermal diffusion in air. Since we consider the static diffusion, our equation becomes Laplace equation. We write it in cylindrical coordinate as $$[\frac{^2}{r^2}+\frac{1}{r}\frac{}{r}+\frac{1}{r^2}\frac{^2}{\theta ^2}+\frac{^2}{x^2}]T(r,\theta ,x)=0.$$ (12) Note that our previous coordinate $`y`$ is related to $`r`$ as $$r=R+y.$$ We suppose the existence of axial symmetry, and we take $`T/\theta =0`$. Therefore we work with $$[\frac{^2}{r^2}+\frac{1}{r}\frac{}{r}+\frac{^2}{x^2}]T(r,x)=0.$$ (13) Since we have oscillating surface in x-direction, it is natural to construct the solution in the form, $$T(r,x)=f(r)+g(r)\mathrm{sin}(kx+\varphi ),$$ (14) where $`f(r)`$ satisfies $$[\frac{d^2}{dr^2}+\frac{1}{r}\frac{d}{dr}]f(r)=0,$$ (15) and $`g(r)`$ satisfies $$[\frac{d^2}{dr^2}+\frac{1}{r}\frac{d}{dr}k^2]g(r)=0.$$ (16) Then the solution is given by using dimensional variables $`x,\stackrel{~}{y}yh_0,\stackrel{~}{R}R+h_0`$ as $`T(x,y)`$ $`=`$ $`A+B\mathrm{log}(1+\stackrel{~}{y}/\stackrel{~}{R})`$ (17) $`+CK_0(k(\stackrel{~}{R}+\stackrel{~}{y}))\mathrm{sin}(kx+\varphi ),`$ where A,B, and C are the coefficients and $`K_0`$ is the modified Bessel function. Last term is induced from boundary condition at water surface. The reader might think the appearance of logarithmic term strange, since it diverges at large $`r`$. But the appearance of such a term is natural for infinitesimally long axially symmetric source. In our system the length of icicle is finite. Therefore this solution is valid just around the icicle, and far from icicle source we must treat the source just like point and we need to connect two solutions at $`rL`$ (L is the order of length of icicle). In this way the coefficients A,B and C will be determined. But we have another physical observable, such as, mean growth rate of icicle which determine coefficient B. The information of temperature at infinity is included into this observable. At $`\stackrel{~}{y}=\delta \mathrm{sin}kx`$ (AW surface), temperature and thermal flow are given by above solution by using taylor expansion. $`T(AW)`$ $`=`$ $`A+[CK_0(k\stackrel{~}{R})\mathrm{cos}\varphi {\displaystyle \frac{LV}{\kappa _0}}\delta ]\mathrm{sin}kx`$ (18) $`+CK_0(k\stackrel{~}{R})\mathrm{sin}\varphi \mathrm{cos}kx,`$ $`Q(AW)`$ $`=`$ $`LV[\kappa _0CkK_0^{}(k\stackrel{~}{R})\mathrm{cos}\varphi +{\displaystyle \frac{LV\delta }{\stackrel{~}{R}}}]\mathrm{sin}kx`$ (19) $`\kappa _0CkK_0^{}(k\stackrel{~}{R})\mathrm{sin}\varphi \mathrm{cos}kx,`$ $`V`$ is the mean growth rate of icicle defined by $$V<Q>/L,$$ with $`L`$ the latent heat, and $`<>`$ means the average along x-direction. This value determines coefficient B as, $$B=\frac{L\stackrel{~}{R}V}{\kappa _0}.$$ $`\kappa _0`$ is the thermal conductivity of air. We treat $`C`$ as order $`\delta `$ and we took up to first order in $`\delta `$. Note that the thermal-diffusion inside an icicle is not considered. Since the surface of icicle is the same temperature as melting point, the temperature inside an icicle can be considered as uniform. We now give continuity condition of temperature and heat flow conservation on the water-air boundary surface. By replacing (10) and (11) in dimensional form, and compare with (18) and (19), all the coefficients are determined as well as the function $`a(x)`$. The growth rate of solid is determined from this function as $$v(x)=\kappa a(x)/L=V+b\mathrm{sin}kx+c\mathrm{cos}kx,$$ (20) where b and c are obtained some coefficients. We can obtain the amplification rate of fluctuation by $`\dot{\delta }/\delta =b/\delta `$. We have two mixed effects for the growth rate of fluctuation. One is the diffusion property of temperature in the atmosphere. The other is the effect of fluid for thermal diffusion in thin-layer. In total, we obtain the amplification rate of fluctuation under the condition $`kR>>1/2`$. $$\frac{\dot{\delta }}{\delta }=Vk\times \frac{1\frac{239}{10080}\alpha ^2}{1+\frac{1272}{10080}\alpha ^2+(\frac{239}{10080})^2\alpha ^4},$$ (21) where $`\alpha k`$ , and $`V`$ is the mean growth rate. This equation is interpreted as R.H.S = (Laplace instability) $`\times `$ (fluid effect for thermal diffusion). Therefore, thermal diffusion in atmosphere enhances fluctuation of shorter wavelength, but fluid effect for thermal diffusion tries to suppress such fluctuation of shorter wavelength. From these two effects, amplification rate has a peak value. The maximum amplification factor is given at definite wave number. $$\alpha _{max}2.2.$$ The corresponding wavelength is obtained by using the definition of $`\alpha `$ as, $$\lambda =\frac{2\pi Uh_0^2}{D\alpha }.$$ The experimental data: $$U=(2.44)\times 10^2m/s,h_0=(0.931.21)\times 10^4m$$ with $`\alpha _{max}2.2.`$ gives $`\lambda =5mm13mm`$, which well agrees with observed wave length 8mm. Let us discuss our obtained result by comparing with the discussion given by Maeno.et.al.1 . Our obtained result comes from two important effects. One is the thermal diffusion in air and its Laplace instability, another one is the effect of fluid for thermal diffusion in thin water. The fluid effect shuffles the static temperature distribution and so the higher frequency wave is suppressed. On the other hand Maeno et.al. 1 have given the conjecture on the formation of ribs on icicle. This is from three points as follows. 1: If we had an protrusion accidentally, the water flow on it may be thinner and cooled much faster. 2: The protrusions are more exposed in the surrounding cold air and giving rapid growth of ribs. 3: Since the water flow have larger viscosity, the water tends to be stagnant just upstream of ribs and giving slower growth of icicle. The first and third one are irrelevant in long wavelength approximation which we used, and the thickness of water layer is almost uniform in our explicit fluid mechanical calculation. This will be well understood by the difference between thickness 0.1mm and wavelength 10mm. The second one may correspond to the Laplace instability in air which we used. But this is not enough to express the wave properties. Really all of above three remarks enhance shorter wavelength fluctuations, and do not determine specific wavelength. So we have used not only the property of water thickness, but also the velocity distribution of water flow which affects thermal diffusion. This effect breaks the Laplace instability at shorter wavelength, and we obtain peak growth rate at definite wavelength as the quantitatively good result. The wavelength depends on the fluid flow $`𝐐2\pi R\overline{U}h_0`$ on icicle surface: $$\lambda _{\mathrm{max}}=\frac{2\pi Uh_0^2}{D\alpha _{max}}=\frac{1}{D\alpha _{max}}(\frac{\nu }{g\pi })^{1/3}(\frac{3𝐐}{2R})^{4/3},$$ (22) where $`R`$ is the mean radius of the icicle. The temperature of the environment changes the mean growth rate of solidification, but wavelength is independent of it. The wavelength depends on the ratio of $`𝐐`$ to $`R`$ but not only on $`𝐐`$. We then have a universal wavelength by assuming that $`𝐐`$ is proportional to $`R`$. The fluid effect that suppresses short wavelength fluctuation is somewhat similar to the Gibbs-Thomson effect. But their origins are quite different, and they work in different length scales. Further experimental test will be necessary to check the conclude validity of this theory by using (22). But here we have shown how the wave can be presented on icicle as instability theory. Our proposed instability theory is quite new and may be applied to a wide range of phenomena. For example, the similar phenomenon occurring on stalactites can be explained in the same way by changing thermal diffusion to solution diffusion. ###### Acknowledgements. The authors are grateful to Prof. R. Takaki for his valuable discussion. One of the authors N.O. thanks Prof. K. Fujii for his encouragement during this study. The authors are also grateful to Prof. E. Yokoyama, Dr. Nishimura, and Prof. R. Kobayashi for their helpful discussions and encouragement and to Ms. K. Norisue for her help in English corrections.
no-problem/9907/cond-mat9907500.html	ar5iv	text	# The Physical Limits of Communication — or — Why any sufficiently advanced technology is indistinguishable from noise ## Abstract It has been well-known since the pioneering work of Claude Shannon in the 1940s that a message transmitted with optimal efficiency over a channel of limited bandwidth is indistinguishable from random noise to a receiver who is unfamiliar with the language in which the message is written. In this letter we demonstrate an equivalent result about electromagnetic transmissions. We show that when electromagnetic radiation is used as the transmission medium, the most information-efficient format for a given message is indistinguishable from black-body radiation to a receiver who is unfamiliar with that format. The characteristic temperature of the radiation is set by the amount of energy used to make the transmission. If information is not encoded in the direction of the radiation, but only its timing, energy or polarization, then the most efficient format has the form of a one-dimensional black-body spectrum which is easily distinguished from the three-dimensional case. Shannon’s information theory considers the set $`\{x_i\}`$ of all possible messages $`x_i`$ which can be transmitted across a “channel” linking a sender to a receiver. In the simplest case the channel is considered to be free of noise, so that each message is received just as it was sent. Shannon demonstrated that the only consistent definition of the amount of information carried by such a channel is $$S=\underset{i}{}p_i\mathrm{log}p_i,$$ (1) where $`p_i`$ is the probability of transmission of the message $`x_i`$. (In evaluating this expression we will take natural logarithms, as is common in statistical mechanics. In information theory logarithms base 2 are the norm, but the difference is only one of a constant factor in $`S`$. Conventionally in statistical mechanics there is also a leading constant $`k`$ in the definition of $`S`$, known as the Boltzmann constant. Here we have set $`k=1`$, with the result that temperatures are measured in units of energy.) If there are no other constraints, then $`S`$ is maximized when all the $`p_i`$ are equal, i.e., when all messages are transmitted with equal probability. If we transmit a constant stream of information in this way by sending many messages one after another, then the resulting flow of data will appear completely random to anyone who does not know the language that it is written in. We now consider the corresponding problem for a message sent using electromagnetic radiation. In outline our argument is as follows. We assume that the sender of a radio message has a certain amount of energy available and we ask what is the maximum amount of information that can be sent with that energy. Generically, this is the problem of maximizing the Shannon information, Eq. (1), for an ensemble of bosons (photons in this case) subject to the constraint of fixed average energy. The solution is familiar from statistical physics, since the formula for the Shannon information is identical to that for the thermodynamic entropy of an ensemble. The maximization of entropy for an ensemble of bosons gives rise to the Bose–Einstein statistics and, in the case of electromagnetic waves, to black-body radiation. By an exact analogy we now show that the most information-rich electromagnetic transmission has a spectrum indistinguishable from black-body radiation. In order to apply Shannon’s theory to electromagnetic radiation we must pose the problem in a form resembling the transmission of information over a channel. To do that, consider the following thought experiment. Suppose we have a closed container or “cavity” with perfectly reflecting walls, which, for reasons which will shortly become clear, we take to be a long tube of constant cross-sectional area $`A_t`$ and length $`\mathrm{}`$. Suppose also that we have the technical wherewithal to set up any electromagnetic “microstate” within this cavity. (The microstates need not be single modes of the cavity; any set of states will work.) Since the walls are perfectly reflecting, this microstate will remain in place indefinitely. We can use the cavity as a communication device. Each possible message $`x_i`$ that we might wish to send is agreed to correspond to one microstate, and we simply pass the entire cavity to the receiver of the message. The information transferred by the electromagnetic radiation stored within is then given by Shannon’s formula with $`p_i`$ being the probability that we put the cavity in microstate $`i`$. (This communication is not the same as a sending a radio message but does have the same information content, as we will demonstrate shortly.) Now suppose that, as in the case of the simple channel, we wish to transmit a steady stream of information by exchanging a series of such cavities. We assume that the power available to make the transmission is limited—i.e., there is a fixed average energy “budget” $`E`$ which we can spend per message sent—and we ask what is the greatest amount of information that can be transmitted per message. This is given by maximizing Eq. (1) subject to the constraint of fixed average energy, but allowing any number of photons to be present in the cavity. The solution of this maximization problem is well-known and gives the grand canonical ensemble in which $$p_i=\frac{\mathrm{exp}(\beta E_i\gamma N_i)}{Z},$$ (2) where $`E_i`$ is the energy in microstate $`i`$, $`N_i`$ is the number of photons, $`Z`$ is the grand canonical partition function, and $`\beta `$ and $`\gamma `$ are Lagrange multipliers, usually referred to as the (inverse) temperature and chemical potential respectively. If we now denote the microstates by the numbers of photons $`\{n_k\}`$ in each (non-interacting) single-particle state $`k`$ of the cavity, then it is straightforward to show that the average of any $`n_k`$ over many messages follows the normal Bose–Einstein distribution $$n_k=\frac{1}{\mathrm{e}^{\beta \epsilon _k}1},$$ (3) where $`\epsilon _k`$ is the energy of a photon in state $`k`$ and we have set $`\gamma =0`$ since there is no chemical potential for photons in a vacuum. Suppose now that, instead of handing over the entire cavity, we remove one of its ends and allow the photons to escape in the form of a message. One should envisage the end of the cavity as being the transmitter of the message, not the cavity itself. The cross-section $`A_t`$ of the cavity is also the area of the transmitter, hence the subscript $`t`$. Usually we will want to direct the message towards a specific receiver. For concreteness we will assume that the receiver is a distance $`d`$ from the end of the cavity and presents some finite area $`A_r`$ to the cavity as depicted in Fig. The Physical Limits of Communication — or — Why any sufficiently advanced technology is indistinguishable from noise. Only those photons in the cavity which have their momentum within the correct interval of solid angle will strike the receiver, possibly after reflecting one or more times off the walls of the container. (We use only the forward interval of solid angle, disallowing reflections off the rear wall of the cavity. If we allow such reflections the transmission will convey twice as much information but last twice as long, making the rate of information transfer identical.) Given that our cavity has volume $`V=\mathrm{}A_t`$, the density of single-particle states satisfying this criterion, taking into account both polarizations of the photons, is $`\rho (\epsilon )=2\mathrm{}A_tA_r\epsilon ^2/d^2h^3c^3`$, where $`h`$ is Planck’s constant and $`c`$ is the speed of light. The average energy contained in our message per unit interval of energy is thus $$I(\epsilon )=\frac{2\mathrm{}A_tA_r}{d^2h^3c^3}\frac{\epsilon ^3}{\mathrm{e}^{\beta \epsilon }1}.$$ (4) A spectrum with this functional form for the intensity $`I(\epsilon )`$ is usually referred to as a black-body spectrum. A black-body spectrum is produced by a perfect thermal radiator in equilibrium at temperature $`T=\beta ^1`$, and, to a good approximation, by most astronomical bodies. In fact there is a small correction to Eq. (4) because photons may be delayed slightly if they have to take a path which reflects off the walls of the cavity. This correction goes inversely as the cosine of the angle subtended by the receiver at the transmitter. Here we have assumed that the distance between transmitter and receiver is large enough that this cosine is well approximated by 1. Our transmission necessarily contains all the information needed to reconstruct the original microstate of the cavity, and therefore contains the same amount of information as that microstate. Furthermore, no ensemble can exist of messages which originate within a volume less than or equal to that of our cavity and which convey more information per message for the same average energy. If such an ensemble existed, reversing time would allow us to trap messages drawn from that ensemble within a cavity of the same size and create an ensemble of microstates which also had greater information content. This conclusion is forbidden, since we have already shown that the black-body distribution maximizes information content. Now we send a stream of information in the form of many such messages, each one resulting from one microstate of the cavity. Because of the tubular shape we have chosen for our cavity each message will last a time $`\mathrm{}/c`$ and the transmission will have constant average intensity. The apparent temperature $`T`$ of the transmission is fixed by the size of the energy budget. Calculating the average energy $`E`$ per message from an integral over Eq. (4) and dividing by $`\mathrm{}/c`$, we find that for a transmission with energy budget $`ϵ`$ per unit time the apparent temperature is given by $$T^4=\frac{15h^3c^2}{2\pi ^4}\frac{d^2}{A_tA_r}ϵ.$$ (5) The information transmitted per unit time $`\mathrm{d}S/\mathrm{d}t`$ can be calculated using the formula $`S=\mathrm{log}Z\beta \mathrm{log}Z/\beta `$, noting that $`\mathrm{log}Z/\beta =E`$. The result is $$\frac{\mathrm{d}S}{\mathrm{d}t}=\frac{8\pi ^4}{45h^3c^2}\frac{A_tA_r}{d^2}T^3=\left[\frac{512\pi ^4}{1215h^3c^2}\frac{A_tA_r}{d^2}ϵ^3\right]^{\frac{1}{4}}.$$ (6) This is the greatest possible rate at which information can be transferred by any electromagnetic transmission for a given average power. It depends only on fundamental constants, the areas of the transmitter and receiver, the distance between them, and the average power or equivalently the apparent temperature. For a transmitter and receiver of one square metre each, a metre apart, with an energy budget of $`1\mathrm{Js}^1`$, the information rate is $`1.61\times 10^{21}`$ bits per second. Note that the information rate increases slightly slower than linearly with the power $`ϵ`$ of the transmitter, as $`ϵ^{3/4}`$. It also increases with the area of the transmitter and receiver, so that the best information rate for a given energy budget is achieved for large antennae with low apparent temperatures. In order to reconstruct the microstate of the original cavity and hence extract the information from a transmission of this kind, we need to have complete information about the photons arriving at our receiver. This means that in addition to energy and polarization information, we need to know the transverse momentum (or transverse position) of each one. In practice, we can at best only detect the transverse components with finite resolution, which places an upper limit on the energy range over which Eq. (4) is valid. The finite size of the receiver places a lower limit on the range. Thus, for a given receiver, the most information-efficient transmission will have a black-body spectrum only between these limits. In fact, most receivers are not capable of measuring the transverse components at all. Usually we have information about only the energy and timing of arriving photons, and possibly their polarization. In this case Eq. (3) still holds, but Eq. (4) becomes $$I(\epsilon )=\frac{2\mathrm{}}{hc}\frac{\epsilon }{\mathrm{e}^{\beta \epsilon }1}.$$ (7) This is the form which the spectrum of a black body would take in a one-dimensional world, and we refer to it as a “one-dimensional black-body spectrum”. The information transmitted per unit time is $$\frac{\mathrm{d}S}{\mathrm{d}t}=\frac{2\pi ^2}{3h}T=\left[\frac{4\pi ^2}{3h}ϵ\right]^{\frac{1}{2}}.$$ (8) For an energy budget of $`1\mathrm{Js}^1`$, this gives a maximum information rate of $`2.03\times 10^{17}`$ bits per second. A related issue is the calculation of the most information-efficient broadcast transmission, where a broadcast in this context means a transmission for which we do not know the location or size of the receiver and therefore cannot guarantee that it will receive all, or even most, of the photons transmitted. In this case we again find that the most efficient transmission is one which has the same spectrum as a one-dimensional black body. To summarize, we have shown that optimizing the information-efficiency of a transmission which employs electromagnetic radiation as the information carrier, with a fixed energy budget per unit time, gives rise to a spectrum of intensities identical to that of black-body radiation. In fact, to an observer who is not familiar with the encoding scheme used, an optimally efficient message of this type will be entirely indistinguishable from naturally occurring black-body radiation. One might be tempted to say that perhaps one could tell the two apart by spotting patterns within particular frequency bands of the message, or by performing some other decomposition of the signal. This however is ruled out by the maximization of the Shannon information; any regularities which would allow one to draw such a distinction are necessarily the result of less than optimal encoding. For the case where the transverse momentum of photons is not used to encode information, or for broadcast transmissions, the spectrum is indistinguishable from that of a one-dimensional black body. We have also shown that the characteristic temperature of the message is simply related to the energy used to send it, and we have derived an upper limit on the rate at which information can be transmitted in both cases. We end with some discussion and speculations. First, we should point out that the ideas outlined here constitute only a thought experiment. By a sequence of deductions we have placed an upper limit on the information efficiency of electromagnetic transmissions, but we have not shown how to achieve that upper limit. Second, we speculate that results similar to those presented here may apply to transmissions using other radiative media as well. Since many natural processes maximize the Gibbs–Boltzmann entropy, they should give rise to spectra indistinguishable from optimally efficient transmissions. For instance, if we had a transmitter that can emit any type of particle (rather than just photons), it seems plausible that the optimal spectrum of particle types and energies would be that of Hawking black-hole radiation .
no-problem/9907/hep-ex9907062.html	ar5iv	text	# Observation of the Cabibbo–suppressed decay Ξ_𝑐⁺→𝑝⁢𝐾⁻⁢𝜋⁺ ## Abstract We report the first observation of the Cabibbo–suppressed charm baryon decay $`\mathrm{\Xi }_c^+pK^{}\pi ^+`$. We observe $`150\pm 22`$ events for the signal. The data were accumulated using the SELEX spectrometer during the 1996–1997 fixed target run at Fermilab, chiefly from a 600 GeV/$`c`$ $`\mathrm{\Sigma }^{}`$ beam. The branching fractions of the decay relative to the Cabibbo–favored $`\mathrm{\Xi }_c^+\mathrm{\Sigma }^+K^{}\pi ^+`$ and $`\mathrm{\Xi }_c^+\mathrm{\Xi }^{}\pi ^+\pi ^+`$ are measured to be $`B(\mathrm{\Xi }_c^+pK^{}\pi ^+)/B(\mathrm{\Xi }_c^+\mathrm{\Sigma }^+K^{}\pi ^+)=0.22\pm 0.06\pm 0.03`$ and $`B(\mathrm{\Xi }_c^+pK^{}\pi ^+)/B(\mathrm{\Xi }_c^+\mathrm{\Xi }^{}\pi ^+\pi ^+)=0.20\pm 0.04\pm 0.02`$, respectively. The study of Cabibbo–suppressed (CS) charm decays can provide useful insights into the weak interaction mechanism for non–leptonic decays . The observed final state may arise either from direct quark emission at the decay stage or, in some cases, from quark rearrangement due to final–state scattering. By comparing the strengths of CS decays to their Cabibbo–favored (CF) analogs, one can, in a systematic way, assess the contributions of the various mechanisms. Modern methods for calculating non–leptonic decay rates of the charm hadrons employ heavy quark effective theory and the factorization approximation . Nonetheless, the three–body decays of charm baryons are prohibitively difficult to calculate due to the complexity of associated final state interactions. Measurements of the relative branching fractions of charm baryon states, both CF and CS, give additional information about the structure of the decay amplitude and the validity of the factorization approximation. Many CS hadronic decays of charm mesons have been measured by both $`e^+e^{}`$ collider and fixed target experiments. Until now the only CS charm baryon decay reported with significant statistics is $`\mathrm{\Lambda }_c^+pK^{}K^+`$ and its resonant state, $`\mathrm{\Lambda }_c^+p\varphi `$ . This transition requires internal W–emission and may be inhibited by color–suppression effects. In this paper, we present the first observation of the CS decay $`\mathrm{\Xi }_c^+pK^{}\pi ^+`$ and determine the branching fractions of this decay relative to the CF $`\mathrm{\Xi }_c^+\mathrm{\Sigma }^+K^{}\pi ^+`$ and $`\mathrm{\Xi }_c^+\mathrm{\Xi }^{}\pi ^+\pi ^+`$ modes. All three of these modes involve external W–emission. Figure 1 shows a spectator diagram with external $`W`$–emission for $`\mathrm{\Xi }_c^+`$ decaying into a CF and a CS mode. (The other CF $`\mathrm{\Xi }^{}`$ mode interchanges $`s`$ and $`u`$ quark lines and produces a $`d\overline{d}`$ pair from the vacuum instead of a $`u\overline{u}`$ pair.) The decay process is similar in the two modes except for the flavor change in the weak decay Cabibbo–Kobayashi–Maskawa matrix element ($`V_{cs}`$ vs. $`V_{cd}`$). Based on a simplified dimensional analysis, we would expect $`B(\mathrm{\Xi }_c^+pK^{}\pi ^+)/B(\mathrm{\Xi }_c^+\mathrm{\Sigma }^+K^{}\pi ^+)=2.1\times \alpha \times \mathrm{tan}^2\theta _c`$ where $`\theta _c`$ is the Cabibbo angle and $`\alpha `$, of order unity, measures the ratio of the hadronic contributions to the weak decay matrix element. The factor 2.1 reflects the relative phase space of the two modes. By comparing the $`\alpha `$ factors for the entire family of CS non–leptonic charm baryon decays one may develop a picture of the dominant features of the amplitudes. The SELEX(E781) experiment at Fermilab is particularly well–suited for studying CS charm baryon decays because of its excellent particle identification capabilities. SELEX is a high energy hadroproduction experiment using a 3–stage spectrometer designed for high acceptance for forward interactions ($`x_F=2p_{}^{CM}/\sqrt{s}>0.1`$). The experiment emphasizes understanding charm production in the forward hemisphere and the study of charm baryon decays. Using both a negative beam ($`50`$% $`\mathrm{\Sigma }^{}`$, $`50`$% $`\pi ^{}`$) and positive beam ($`92`$% $`p`$, $`8`$% $`\pi ^+`$), SELEX recorded 15.2 billion interaction events during the 1996–1997 fixed target run. The majority particles ($`\mathrm{\Sigma }^{}`$, $`\pi ^{}`$) in the 600 GeV/$`c`$ negative beam and the protons in the 540 GeV/$`c`$ positive beam were tagged by a beam transition radiation detector. The data were accumulated from a segmented target (5 foils: 2 Cu, 3 C, each separated by 1.5 cm) whose total thickness was 5% of an interaction length for protons. The spectrometer had silicon strip detectors to measure the beam and outgoing tracks, giving precision primary and secondary vertex reconstruction. Transverse position resolution ($`\sigma `$) was 4 $`\mu \mathrm{m}`$ for the 600 GeV/$`c`$ beam tracks. The average longitudinal vertex position resolution was 270 $`\mu \mathrm{m}`$ for primary and 560 $`\mu \mathrm{m}`$ for secondaries, respectively. Momenta of particles deflected by the analyzing magnets were measured by a system of proportional wire chambers (PWCs), drift chambers and silicon strip detectors. Momentum resolution for a typical 100 GeV/$`c`$ track was $`\sigma _p/p0.5`$%. The absolute momentum scale was calibrated using the $`K_S^0`$ mass. For $`D^0K^{}\pi ^+`$ decays the average mass resolution was 9 MeV/$`c^2`$ independent of $`D^0`$ momenta from 100–450 GeV/$`c`$. Charged particle identification was done with a Ring Imaging $`\stackrel{ˇ}{\mathrm{C}}`$erenkov detector (RICH) , which distinguished $`K^\pm `$ from $`\pi ^\pm `$ up to 165 GeV/$`c`$. The proton identification efficiency was greater than 95% above proton threshold ($`90`$ GeV/$`c`$). For pions the total mis–identification probability due to all sources of confusion was less than 4%. Interactions were selected by a scintillator trigger. The trigger for charm required at least 4 charged tracks after the targets as indicated by an interaction counter and at least 2 hits in a scintillator hodoscope after the second analyzing magnet. It accepted about 1/3 of all inelastic interactions. Triggered events were further tested in an on–line computational filter based on downstream tracking and particle identification information. The on–line filter selected events that had evidence of a secondary vertex from tracks completely reconstructed using the forward PWC spectrometer and the vertex silicon. This filter reduced the data size by a factor of nearly 8 at a cost of about a factor of 2 in charm written to tape (as determined from a study of unfiltered $`K_S^0`$, $`\mathrm{\Lambda }`$ and $`D^0K^{}\pi ^++c.c.`$ decays). Most of the charm loss came from selection cuts that were independent of charm species or kinematic variables and which improved the signal/noise in the final sample. Results presented here come from the first processing pass through all data. In this analysis, secondary vertex reconstruction was attempted when the $`\chi ^2`$ per degree of freedom ($`\chi ^2`$/dof) for the fit of the ensemble of tracks to a single primary vertex exceeded 5. All combinations of tracks were formed for secondary vertices with as many as 5 prongs. Secondary vertices were tested against a reconstruction table that specified selection criteria for each charm decay mode. Secondary vertices which occurred inside the volume of a target or after the first plane of the silicon vertex detector were rejected. The $`pK^{}\pi ^+`$ sample was selected by the following requirements: (i) primary and secondary vertex fits each have $`\chi ^2/\mathrm{dof}<5`$; (ii) vertex separation significance $`L/\sigma 8`$, where $`L`$ is the longitudinal separation between primary and secondary vertices and $`\sigma `$ is the error on $`L`$; (iii) the reconstructed momentum vector from the secondary vertex points back to the primary vertex with $`\chi ^2/\mathrm{dof}<4`$; (iv) the momentum of the $`\pi ^+`$ $``$ 5 GeV/$`c`$; (v) $``$($`K`$)/$`(\pi )>1`$ for $`K`$ identification and $``$($`p`$)/$`(\pi )>1`$ for $`p`$ identification, where $``$ is the mass–selection likelihood function for the RICH; and (vi) the sum of the squared transverse momenta of the three daughter particles with respect to the parent direction, $`\mathrm{\Sigma }p_T^2>0.3`$ $`(\mathrm{GeV}/c)^2`$. Figure 2 shows the first observation of the CS $`\mathrm{\Xi }_c^+pK^{}\pi ^+`$ decay mode. The inset of the figure shows the invariant mass distribution from the entire selected mass range of reconstructed $`pK^{}\pi ^+`$ candidates. The large peak in the inset is the decay $`\mathrm{\Lambda }_c^+pK^{}\pi ^+`$. The bump at the right is the CS $`\mathrm{\Xi }_c^+pK^{}\pi ^+`$ decay. The distribution has an artificial high–mass cutoff because of a maximum mass cut imposed in this first–pass analysis. Monte Carlo simulations show that the observed number of signal events is systematically reduced only by ($`2\pm 3`$)% due to this cutoff. The background characteristics are very similar both in the number per mass bin and lifetime through the entire mass region, so that the background subtraction under the $`\mathrm{\Xi }_c^+`$ peak is stable. The lifetime both of background and signal events for the $`\mathrm{\Xi }_c^+`$ region has been investigated. The background apparent lifetime is quite short, while the $`\mathrm{\Xi }_c^+`$ lifetime is consistent with the Particle Data Group average . A detailed lifetime analysis is underway and will be reported elsewhere. The mass is determined using a Gaussian of variable centroid, width and area. The number of signal events is determined by a sideband subtraction method. The signal window is determined from the $`\mathrm{\Lambda }_c^+`$ data and is set at 45 MeV/$`c^2`$. The background is a second order polynomial fit to the entire mass region excluding a 65 MeV/$`c^2`$ wide window centered at each mass peak. The $`\mathrm{\Xi }_c^+`$ yield is the difference between the summed events in the 45 MeV/$`c^2`$ signal window centered at 2.467 GeV/$`c^2`$ and the extrapolated background. Variations in the exclusion window size change the extrapolated background by fewer than 2 events. There are $`150\pm 22`$ events for $`\mathrm{\Xi }_c^+pK^{}\pi ^+`$ at a $`\mathrm{\Xi }_c^+`$ mass of $`2467.4\pm 1.3`$ MeV/$`c^2`$. The statistical significance for the signal, $`S/\sqrt{S+B}`$, is $`6.9\pm 0.6`$ in which $`S`$ is the number of signal events and $`B`$ is the number of background events in the signal region. Background contributions from charm baryon ($`\mathrm{\Xi }_c^0`$ and $`\mathrm{\Omega }_c^0`$) decay to the CS decay are expected to be negligible; they produce an even number of secondary vertex tracks and their decay lengths are short compared to the minimum $`\mathrm{\Xi }_c^+`$ flight path cut. Excellent proton identification strongly suppresses background from charm meson decays (three–body $`D^+`$ and $`D_s^+`$). One normalizes the CS mode by comparing it to CF modes that have a hyperon in the final state. In the first reconstruction pass, hyperons ($`\mathrm{\Sigma }^\pm ,\mathrm{\Xi }^{}`$) were identified only inclusively in a limited decay interval (5 – 12 m downstream from the target): candidates had $`p>40`$ GeV/$`c`$ and no hits assigned along the trajectory in the 14 chambers after the second analyzing magnet. This category of tracks gives unique $`\mathrm{\Sigma }^+`$ identification but is ambiguous between $`\mathrm{\Sigma }^{}`$ and $`\mathrm{\Xi }^{}`$. Figure 3 shows two CF $`\mathrm{\Xi }_c^+`$ modes decaying to $`\mathrm{\Xi }^{}\pi ^+\pi ^+`$ and $`\mathrm{\Sigma }^+K^{}\pi ^+`$, respectively. The shaded areas in Fig. 3(a) and 3(b) are the estimated reflection from $`\mathrm{\Lambda }_c^+\mathrm{\Sigma }^{}\pi ^+\pi ^+`$ and $`\mathrm{\Lambda }_c^+\mathrm{\Sigma }^+\pi ^{}\pi ^+`$, respectively. The shapes are determined by a Monte Carlo simulation and the areas are normalized to the observed number of signal events in the $`\mathrm{\Lambda }_c^+`$ data. For this part of the analysis, additional cuts were applied to all data. The pion momentum threshold was raised from 5 to 10 GeV/$`c`$ and the transverse component of the $`\mathrm{\Xi }_c^+`$ momentum with respect to its line of flight was required to be less than 0.3 GeV/$`c`$. The total acceptance (geometrical acceptance and reconstruction efficiency) for decay modes of interest was estimated by embedding Monte Carlo charm decay tracks into data events. Momentum and energy were not conserved in the process, but studies indicate this has little effect on the single–charm acceptance calculation. Events were generated with an average transverse momentum $`p_T=1.0`$ GeV/$`c`$ and longitudinal momentum distributions as observed for the data. Detector hits, including resolution and multiple Coulomb scattering smearing effects, produced by these embedded tracks were folded into the hit banks of the underlying data event. The new ensemble of hits was passed through the SELEX off–line software. The acceptance is the ratio of the number of events in the signal to the number of embedded events in a particular mode. The method was checked by measuring two well–determined relative branching fractions, $`B(D^0K^{}\pi ^+\pi ^+\pi ^{})/B(D^0K^{}\pi ^+)=1.94\pm 0.07\pm _{0.11}^{0.09}`$ and $`B(D^+K^{}K^+\pi ^+)/B(D^+K^{}\pi ^+\pi ^+)=0.093\pm 0.010\pm _{0.006}^{0.008}`$ where the first error is statistical and the second is systematic. Systematic errors include the difference of the ratio for charge–conjugate states, and uncertainties of signal yields and acceptance estimations. The results agree well with the world average . For the $`D^+`$ decays, the vertex significance cut ($`L/\sigma `$) was increased to 20 to suppress background from $`D_s^+K^{}K^+\pi ^+`$. This costs little signal, since the lifetime of $`D^+`$ is 2.4 times longer than that of $`D_s^+`$. Using this procedure, we measure the relative branching fraction for $`\mathrm{\Xi }_c^+\mathrm{\Sigma }^+K^{}\pi ^+`$ and $`\mathrm{\Xi }_c^+\mathrm{\Xi }^{}\pi ^+\pi ^+`$ to be $`(0.92\pm 0.20\pm 0.07)`$. The systematic error is due to uncertainties in the background subtraction and relative acceptance estimation. This result is comparable to the CLEO measurement, $`(1.18\pm 0.26\pm 0.17)`$ . The number of events and estimated acceptance for the three observed $`\mathrm{\Xi }_c^+`$ modes with the common set of cuts are summarized in table I. The branching fraction of the CS decay $`\mathrm{\Xi }_c^+pK^{}\pi ^+`$ relative to the CF $`\mathrm{\Xi }_c^+\mathrm{\Sigma }^+K^{}\pi ^+`$ is measured to be $`0.22\pm 0.06\pm 0.03`$ which corresponds to $`(4.2\pm 1.1\pm 0.5)\times \mathrm{tan}^2\theta _c`$. The $`\alpha `$ factor is $`2.0\pm 0.5\pm 0.2`$. To put this in some perspective, the corresponding $`\alpha `$ factor for the CS decay $`\mathrm{\Lambda }_c^+pK^{}K^+`$ relative to the CF mode $`\mathrm{\Lambda }_c^+pK^{}\pi ^+`$ is $`2.5\pm 0.6`$, using the world average for the branching fraction and correcting for phase space. Thus, the factors for these two different baryons are similar. In contrast, the CS ratios of $`D`$ mesons depend strongly on final state multiplicity, suggestive of sizable final state interactions. For the $`\mathrm{\Xi }^{}\pi ^+\pi ^+`$ mode, B($`\mathrm{\Xi }_c^+pK^{}\pi ^+`$)/ B($`\mathrm{\Xi }_c^+\mathrm{\Xi }^{}\pi ^+\pi ^+`$) is measured to be $`0.20\pm 0.04\pm 0.02`$. Systematic errors include uncertainties in (i) the relative acceptance estimation, (ii) background subtractions due to the reflection from other hyperon modes, and (iii) the $`pK^{}\pi ^+`$ yield due to the imposed mass cutoff. In summary, we observe the CS decay $`\mathrm{\Xi }_c^+pK^{}\pi ^+`$ at a mass of $`2467.4\pm 1.3`$ MeV/$`c^2`$ with $`150\pm 22`$ signal events. The relative branching fractions of the decay $`\mathrm{\Xi }_c^+pK^{}\pi ^+`$ to the CF $`\mathrm{\Xi }_c^+\mathrm{\Sigma }^+K^{}\pi ^+`$ and $`\mathrm{\Xi }_c^+\mathrm{\Xi }^{}\pi ^+\pi ^+`$ are measured to be $`B(\mathrm{\Xi }_c^+pK^{}\pi ^+)/B(\mathrm{\Xi }_c^+\mathrm{\Sigma }^+K^{}\pi ^+)=0.22\pm 0.06\pm 0.03`$ and $`B(\mathrm{\Xi }_c^+pK^{}\pi ^+)/B(\mathrm{\Xi }_c^+\mathrm{\Xi }^{}\pi ^+\pi ^+)=0.20\pm 0.04\pm 0.02`$, respectively. The authors are indebted to the staffs of Fermi National Accelerator Laboratory, the Max–Planck–Institut für Kernphysik, Carnegie Mellon University, and Petersburg Nuclear Physics Institute for invaluable technical support. This project was supported in part by Bundesministerium für Bildung, Wissenschaft, Forschung und Technologie, Consejo Nacional de Ciencia y Tecnología (CONACyT), Conselho Nacional de Desenvolvimento Científico e Tecnológico, Fondo de Apoyo a la Investigación (UASLP), Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP), the Israel Science Foundation founded by the Israel Academy of Sciences and Humanities, Istituto Nazionale de Fisica Nucleare (INFN), the International Science Foundation (ISF), the National Science Foundation (Phy #9602178), NATO (grant CR6.941058-1360/94), the Russian Academy of Science, the Russian Ministry of Science and Technology, the Turkish Scientific and Technological Research Board (TÜBİTAK), the U.S. Department of Energy (DOE grant DE-FG02-91ER40664 and DOE contract number DE-AC02-76CHO3000), and the U.S.-Israel Binational Science Foundation (BSF). We also acknowledge a useful conversation with Austin Napier.
no-problem/9907/cond-mat9907252.html	ar5iv	text	# 1 Introduction ## 1 Introduction Surface growth is a problem which may be found in mathematics, physics, biology or even social phenomena, if we define the surface as a boundary between different phases in $`d`$-dimensional space . Theoretical studies of the properties of the surfaces of the grown crystal/thin films based on Monte Carlo (MC) simulations, initially using the solid-on-solid (SOS) models, then extended to more realistic kinetic MC methods, became particularly popular in early 80’s (see for review). These studies resulted in at least a semi-quantitative understanding of growth modes and step dynamics. Then rapid progress in observational methods (including diffractional methods — RHEED, LEED and direct imaging — STM, AFM) allowed for better understanding mechanism and physics of crystal/films growth. However, MC simulations may be still applied as a complementary method which is sometimes more efficient while analytical treatment fails and surely is more cheaper than performing experiments. In this work with the SOS model we would like to check how substrate temperatures influence qualitatively the surface morphology before the percolation limit, when only a small fraction of the first monolayer (ML) is deposited. ## 2 Model The present model on a square lattice is based on random deposition (RD) enriched by a relaxation process of incoming at the fixed rate particles. After a random choice of the place of the particle’s initial contact to the surface, particles migrate to one of the four nearest neighbor (NN) sites or stick on the place of deposition. The migration procedure is repeated $`L_{\text{dif}}`$ times, thus $`L_{\text{dif}}`$ may be considered a range of diffusion . In the local relaxation process, the particle tends to maximize the number of particle-particle lateral bonds (PPLB) similarly to the Wolf–Villain model . The probability of choosing one of the five accessible sites is given by the Boltzmann factor $`\mathrm{exp}(E/k_BT)`$, where the particle’s total energy $`E=n_xJ_x+n_yJ_y`$ in all five positions depends on the number $`n_x`$ ($`n_y`$) of the virtually created bonds in $`x`$ ($`y`$) direction. $`J_x`$ and $`J_y`$ reflect strengths of the PPLB. $`k_B`$ is the Boltzmann constant and $`T`$ denotes the substrate absolute temperature. Additionally, the system tendency to increase the number of PPLB is slowed down by the barrier for diffusion $`V`$: the probability of movement (choosing NN sites) is reduced by a factor $`\mathrm{exp}(V_x/k_BT)`$ or $`\mathrm{exp}(V_y/k_BT)`$, depending on the considered direction of diffusion. The diffusion barrier $`V`$ must be positive, while the negative $`E`$ is compatible with the assumed tendency of the system to maximize the number of PPLB. The model control parameters were taken after Ref. (where kinetic MC simulations were performed) to mimic the realistic ratio of $`V_x/V_y`$ and $`J_x/J_y`$. We assume that activation energy for diffusion and activation energy for breaking PPLB (in terminology of the cited paper) may correspond to the model control parameters $`V`$ and $`E`$, respectively. $`x`$-direction is parallel to dimer rows, while $`y`$ shows the direction of cross-channel diffusion on a bcc(110) plane: $$\begin{array}{c}\hfill \text{Ag: }V_x/V_y=0.736J_x/J_y=9.000\\ \hfill \text{Cu: }V_x/V_y=0.793J_x/J_y=6.857\end{array}$$ ## 3 Results of simulations The simulations were carried out on a $`256\times 256`$ lattice with periodic boundary conditions. Diffusion was limited to $`L_{\text{dif}}=10^3`$ steps. We measure the surface anisotropy in terms of the dimensionless parameter $`\epsilon \left[G(1,0)G(0,1)\right]/G(0,0)`$, where $`G(𝐬)h(𝐫+𝐬)h(𝐫)h(𝐫)^2`$ is height-height correlation function . The average $`\mathrm{}`$ runs over all substrate sites $`𝐫`$ and the dispersion $`G(0,0)=\sigma ^2=\left[h(𝐫)h(𝐫)\right]^2`$ is defined as a square of the surface width $`\sigma `$, which usually describes the surface roughness. ### 3.1 Influence of the substrate temperature on the surface anisotropy Fig. 1 shows changes of surface morphology with increasing temperature for submonolayer coverage $`\theta =0.1`$ ML. For the limiting case $`T0`$ ($`\mathrm{exp}(V/k_BT)\mathrm{}`$), the large diffusion barriers stop any migration. Thus the probability to stay in the place of deposition $`P_0`$ goes to unity, while the probability of the migration to the NN sites $`P_{\text{NN}}`$ goes to zero, and a film surface grown on a very cold substrate contains randomly, isotropically distributed single atoms. The anisotropy parameter $`\epsilon `$ tends to zero. Increase of the temperature changes the ratio between Boltzmann factors drastically. For low temperatures isotropically distributed small aggregates on the substrate appear (see the upper part of Fig. 1). In such a case although particles are allowed to make $`L_{\text{dif}}=10^3`$ steps, they stick after making only a few of them, and the surface morphology still corresponds roughly to that generated with RD rules. Then in medium temperatures range — see the medium part of Fig. 1 — migration in $`y`$-direction is still stopped by diffusion barriers, while particles are able to diffuse in $`x`$-direction and long 1-D stripes on the substrate grow. The anisotropy parameter $`\epsilon `$ becomes maximal. Finally, in the bottom of Fig. 1 nucleation of larger islands is observed with tendency of their orientation along $`x`$-direction when the temperature is relatively high. For infinitely large temperature $`T\mathrm{}`$ ($`\mathrm{exp}(V/k_BT)1`$), the barrier for diffusion does not influence the particle migration. The probabilities of choosing any of the accessible sites become equal ($`P_0=P_{\text{NN}}=1/5`$). Particles migration is stopped only by limited to $`L_{\text{dif}}`$ steps diffusion. Thus, for very hot substrates the surface morphology is similar to that generated with RD and again $`\epsilon =0`$. The changes of the surface morphology may also be observed by the temperature dependence of the height-height correlation function $`G(𝐬)`$. In Fig. 2, as discussed earlier, the influence of the substrate temperature $`T`$ on the surface morphology anisotropy parameter $`\epsilon `$ is presented. ### 3.2 Influence of the substrate temperature on the surface roughness For submonolayer coverage, the film contains $`\theta L^2`$ occupied substrate sites (with $`h=1`$) and $`(1\theta )L^2`$ empty ones ($`h=0`$). Neither infinitely large diffusion $`L_{\text{dif}}\mathrm{}`$ nor varying substrate temperature $`T`$ can change the distribution of film heights and/or influence the single-site characteristic $`\sigma `$ — the surface roughness (they influence, however, two-sites characteristics such as $`\epsilon `$ — the surface anisotropy). Thus roughness $`\sigma `$ grows like $`\sqrt{\theta }`$ as it is predicted by Poisson process for RD case. For example, for surfaces which exhibit self-affinity and, as a results of that, their roughness dynamics may be described by Family–Vicsek law , before completing first ML — or more precisely, before substrate coverage reach the percolation threshold and the notion of film thickness can be applied — surface roughness dynamics corresponds to simple RD. Such a situation was already observed for many SOS models, and explained as an artifact of an initially flat substrate . For instance, in the Family model , particles after RD perform a local search for the column with minimal height. Thus, starting with perfectly flat substrate, and for submonolayer coverage, all sites are energetically equivalent and diffusion is unimportant. Note, that for submonolayer MBE also other law to describe dynamic scaling was proposed . ## 4 Conclusions In summary, we found from computer simulations that with increasing substrate temperature a transition in surface morphology takes place. In contrast, varying the substrate temperature does not influence the film roughness for submonolayer coverage. The changes in surface morphology are due to changes of the relative ratio between Boltzmann factors for diffusion barriers and strengths of attraction (bonds energies) in two directions perpendicular to each other. Qualitative agreement with kinetic MC results of simulation with fully reversible aggregation SOS model for Ag(110) and Cu(110) and experimental observation for Cu submonolayers was achieved. Also STM images and hot STM movies of homoepitaxial submonolayer growth on Si(001) substrate shows similar morphology transition. For low deposition rates, 1-D islands grow well within a 50 K window around 530 K. At lower temperatures, mobility is decreased, and atoms deposited on the surface nucleate rather new islands than attaching to existing islands. At higher temperatures islands merge and become 2-D but remain still anisotropic . At low coverage and at room temperature III group elements (Ga, In) and IV group elements (Si, Ge, Sn, Pb) all form 1-D chains of ad-dimers . Shen and co-workers — as it was earlier stated also in the work of Mo et al. and confirmed by STM studies — conclude that, for such a systems, not only the anisotropy in the diffusion but also the anisotropy in the sticking must be involved in modeling film growth. In presented here model first assumption is realized by different barriers for diffusion ($`V_xV_y`$) while anisotropy in bonds ($`J_xJ_y`$) mimics anisotropic sticking probability. ## Acknowledgments I am grateful to A.Z. Maksymowicz for scientific guidance and fruitful discussion. The simulations were carried out in ACC-CYFRONET-AGH. This work and machine time in ACC-CYFRONET-AGH are financed by Polish Committee for Scientific Research (KBN) with grants No. 8 T11F 02616 and No. KBN/S2000/AGH/069/1998, respectively.
no-problem/9907/cond-mat9907100.html	ar5iv	text	# Mesoscopic Analysis of Structure and Strength of Dislocation Junctions in FCC Metals ## Abstract We develop a finite element based dislocation dynamics model to simulate the structure and strength of dislocation junctions in FCC crystals. The model is based on anisotropic elasticity theory supplemented by the explicit inclusion of the separation of perfect dislocations into partial dislocations bounding a stacking fault. We demonstrate that the model reproduces in precise detail the structure of the Lomer-Cottrell lock already obtained from atomistic simulations. In light of this success, we also examine the strength of junctions culminating in a stress-strength diagram which is the locus of points in stress space corresponding to dissolution of the junction. In FCC metals, a key mechanism limiting the movement of the dislocations is the “forest intersection” mechanism, where segments of dislocations on a glide plane are rendered immobile as a result of intersection with dislocations on other glide planes. Such intersections can lead to complex dislocation junction structures since the cores of the dislocations in these metals are dissociated into partial dislocations separated by a stacking fault. Furthermore, the structure of the junction also depends strongly on the geometric disposition (such as the line directions and Burgers vector) of the participating dislocations. A core level analysis of all the possible junction configurations is therefore important from a number of perspectives. Such core level calculations in conjunction with statistical averaging procedures can provide key parameters that are used in models that predict the mechanical behavior of metallic crystals on a macro-scale. These models include single crystal plasticity models and the computational models that simulate the dynamics of a large collection of dislocations. In this letter we develop a mesoscopic dislocation dynamics model that can be used to simulate the structures and strength of dislocation junctions in FCC metals and may obviate the need for direct atomistic simulations of these junctions. For simple dislocation intersection geometries, the structure of dislocation junctions has been studied extensively in a series of classic papers by Hirth and coworkers using the theory of linear elasticity. Analytical insights into more complicated intersection geometries have been gained by using the line tension approximation for the dislocation lines. While this approach provides a great deal of physical insight into the junction structure and strength, it ignores the extended core structure of the dislocations as well as the long range interaction between the dislocation segments. With rapid advances in computational power in recent years, it has become possible to perform atomistic simulations of dislocation intersections. Typically these simulations, performed using cells of the dimensions of 300-400$`\AA `$ containing over a million atoms, are computationally very demanding and raise serious questions concerning the role that boundary conditions play in dictating the results. In the present study, we develop a mesoscopic method to study the structure and strength of dislocation junctions that includes the dissociation of the dislocation core into partial dislocations. The interaction between the dislocations are treated using the theory of anisotropic elasticity. We find that the our method reproduces, in precise detail, all the features of the dislocation junction structure obtained from a full atomistic treatment of the dislocation core. Our results demonstrate that the junction structure is almost entirely determined by elastic interaction between the partial dislocations and the stacking fault energy. Furthermore, the adaptive procedures that we use allow us to simulate dislocation junctions that have lengths of the order of microns. An atomistic simulation of these large junctions would require simulations involving an excess of a billion atoms. In order to clearly demonstrate the role of the stacking fault in determining the junction structure, we will consider dislocation junctions in two metals, namely, Al, with a high stacking fault energy (0.104 $`J/m^2`$) and Ag, with a low stacking fault energy(0.016 $`J/m^2`$). We limit our discussions to the Lomer-Cottrell lock; a complete investigation of other junctions will be reported elsewhere. The simulations are carried out through an adaptive finite element based dislocation dynamics algorithm that is described in detail in Ref.. We start our simulations with straight dislocation lines that are pinned at their end points. Simulations are carried out until the dislocations glide to their equilibrium configuration. Each dislocation line is allowed to split into partial dislocations by accounting for the energy cost of the stacking fault created in the process. In the starting configuration, the partial dislocations are discretised into straight segments of equal length. A time step of the dynamics process consists of moving the node connecting the segments with a velocity that is proportional to the nodal driving force. The computation of the nodal force requires the knowledge of the force per unit length at certain quadrature points on the dislocation segments that are attached to the node. The force per unit length at any point on a segment consists of a component arising from the stresses due to all the dislocation segments in the system including the segment itself. The Brown regularization procedure is adopted to guarantee that the self-stress contribution is well behaved. When the interaction between the segments belonging to different dislocations is attractive, they approach each other in the process of forming a junction. The stress acting on one of the segments due to the other one diverges as their separation vanishes. In our calculations, the stresses are constant for distances less than a critical separation distance, $`r_c`$, by the value of stress computed at $`r_c`$. As a result, the segments are locked once they are closer than this critical distance; a junction has formed. From an elastic perspective, at distances larger than $`r_c`$, the stresses produced by the junction segments correspond to those produced by a dislocation whose Burgers vector is the sum of the Burgers vectors of the two segments that make up the junction. Once the junction is formed, it can unzip if the external stresses cause the segments that form the junction to move away from each other. The calculations described here were carried out with $`r_c=b=a/\sqrt{2}`$, although we have also considered the cases in which the cutoff was $`b/2`$ and $`2b`$, without noticeable change to the resulting junction structures. In addition to the stress from the dislocation segments, the force per unit length consists of a component arising from the stacking fault. This component is normal to the dislocation segment and has a magnitude that equals the stacking fault energy and acts in a direction tending to shrink the stacking fault. In the simulations that we describe, we have ignored the frictional stress since we have found that its inclusion is unimportant. A key feature of our simulations is the adaptive positioning of the nodes. As the simulation progress, the nodes are redistributed, so that the regions with large curvature have more nodes per unit length. We demonstrate our results by first considering the equilibrium configuration of a Lomer-Cottrell lock as shown in Fig. 1. This configuration has been chosen so as to make a direct comparison of our results with the atomistic simulations for Al reported in Ref.. The pinning points are arranged such that in the starting configuration the dislocation line directions make an angle of $`60^0`$ with the $`[\overline{1},1,0]`$ direction. This direction coincides with the line of intersection of the slip planes of the dislocations that form the junction. The initial separation between the dislocations was chosen to be 6$`\AA `$. The line directions of each of the partial dislocations and their slip plane normals are given in the figure. We follow the notation described in Ref. to label the Burgers vectors of the partial dislocations by referring to the Thompson tetrahedron. For example, the $`a/2[0,1,\overline{1}](111)`$ dislocation splits into partial dislocations $`A\delta `$ and $`\delta C`$ with Burgers vectors $`a/6[\overline{1},2,\overline{1}]`$ and $`a/6[1,1,\overline{2}]`$ respectively. As is evident from the figure, the junction segment in the case of Al, has split into separate parts. A stair-rod segment with Burgers vector $`\gamma \delta `$ of the type $`a/6110`$ forms an extended node on the left side of the junction. The remaining part of the junction is a sessile Lomer dislocation segment. The length of the stair-rod segment is 38$`\AA `$, while the Lomer is 42$`\AA `$, which is in excellent agreement with the atomistic results. The dislocation dynamics model also agrees with the atomistic results in predicting the structure of the right hand node, which is point-like and is the meeting point of constricted dislocation segments. As an extension of earlier results and to highlight the dependence of the junction structure on the stacking fault energy, we have also computed the geometry of the Lomer-Cottrell junction in Ag. The dislocation junction in Ag, for this configuration, has an entirely different structure. The junction segment is entirely composed of a stair-rod dislocation of length $`180`$Å. The smaller stacking fault energy keeps the segments $`\delta C`$ and $`D\gamma `$ from participating in the junction formation process. Before proceeding to discuss the effect on junction structure of the length and the orientation of the participating dislocations, we pause to illustrate the adaptive nodal repositioning process used in the simulations. It can be seen from Fig.1b that the nodes on the red dislocation line have evolved from a case where they are spaced at equal intervals to a case where there are more nodes in the regions with large curvature. In order to consider the effect of varying the distance between pinning points and the orientation of the dislocations on the junction structure, we consider the highly symmetric geometry shown in Fig. 1, where the participating dislocations have the same orientation and identical distances between the pinning points. In the case of the dislocation junction in Al, we find that the length of the Lomer segment increases with increasing distance between the pinning points, while the length of the stair-rod segment is not altered. By way of contrast for the junction in Ag, the length of the stair-rod keeps increasing on increasing the distance between the pinning points up to about 1000$`\AA `$. Beyond this length, the Lomer segment appears and increases in length, while the left hand stair-rod node attains a constant length of about 600$`\AA `$. We therefore conclude that the length of the extended node is determined largely by the stacking fault energy of the crystal. This calculation also illustrates the ability of our method to handle large junction lengths which are clearly beyond the reach of the atomistic simulations. We now consider the effect of altering the junction angle on the structure of the dislocation junction. We use $`\varphi `$ to denote the angle between the dislocation line direction in the starting configuration and the $`[\overline{1},1,0]`$ direction, $`l_J`$ to represent the junction length and $`2l`$ for the distance between the pinning points. In Fig. 2, we plot the junction length and structure as the orientation of the pinning points is altered. The dislocation junction ceases to form at angles smaller than -50<sup>0</sup> or angles greater than 80<sup>0</sup>. For comparison the we also show the results of the line tension model tension model of Saada. Within this model the junction length vanishes for angles that are larger than 60<sup>0</sup> or smaller than -60<sup>0</sup>. The difference can be understood by noting the difference in the node structures in junctions with negative angles in Fig. 2. The left node for the -30<sup>0</sup> configuration is made up of a stair-rod of the type $`\gamma \delta `$, while the right node is made up of a new stair-rod segment $`\delta D/C\gamma `$, which is a $`a/3110`$ type dislocation. The existence of the $`a/3110`$ in the negative angle junctions is counterintuitive, since a simple line tension estimate shows that this type of stair rod has 4 times higher energy per unit length than the $`a/6110`$ stair-rod. The dislocation interactions that are ignored in this simple line tension argument conspire to make this stair-rod segment stable. The asymmetry in the junction formation angles can be attributed to this difference in junction structures of the positive and negative angle junctions. Far more interesting than the structure of junctions is their behavior under stress. Line tension models predict that the arms of the dislocations bow out under the influence of the external stress, while the junction translates along the line of intersection of the slip planes. As the junction translates, its length decreases as a result of the applied stress via an “unzipping” mechanism. Fig. 3 shows the evolution of the dislocation junction in Fig.1a under the influence of an externally applied stress. The resolved shear stress on the dislocations in the $`(1,1,1)`$ and $`(\overline{1},1,1)`$ slip planes are labeled $`\sigma _1`$ and $`\sigma _2`$ respectively. We have chosen the orientation of the applied stress for this series of pictures such that $`\sigma _1=\sigma _2`$. On increasing the stress, the length of the Lomer segment initially increases on going from no applied stress to a stress of 0.006$`\mu `$ ($`\mu `$ is the shear modulus). This is consistent with the atomistic simulations, but is at variance with the line tension models, which always predict a decreasing junction length with increasing stress. On increasing the stress further, the junction translates and undergoes an unzipping mechanism, whereby the length of the Lomer segment decreases. This behavior is evident for the stress level of .009$`\mu `$. On increasing the stress further, the junction breaks at a stress level of about 0.012$`\mu `$ after which they continue to bow out. The configuration shown at stress of 0.012$`\mu `$, is not in equilibrium, but a snap shot after the junction has been destroyed. We have carried out similar calculations for several values of resolved shear stress acting on the two dislocations. A “yield surface” for dislocation destruction is shown in Fig. 4. This surface is symmetric about the line $`\sigma _1=\sigma _2`$, but depends on the sign of $`\sigma _1`$ and $`\sigma _2`$. This symmetry breaking is readily understood by looking at the configurations of the dislocations under action of positive and negative $`\sigma _1`$(denoted in Fig. 4 by points A and B respectively) with $`\sigma _2=0`$. The presence of the stair-rod node makes it more difficult for the horizontal segment to bow out in case (A) compared to case (B) and requires a larger value of shear stress to break the junction. It is interesting to compare the predictions of the line tension model and the atomistic simulations with the results obtained from our simulations. An important prediction of the line tension model is the scaling of the breaking stress of symmetric junctions with the distance between the pinning points, written as $`2l`$. The line tension model predicts the average critical resolved shear stress to break the symmetric junctions to be $`0.5\mu b/l`$. We have simulated the breaking of symmetric junctions with lengths ranging from 300$`\AA `$ to 1$`\mu m`$ by applying the external stress in different orientations. We have found that the critical breaking stress scales as $`\sigma _c\mu b/l`$. For the cases $`\sigma _1=\sigma _2`$ and $`\sigma _1=\sigma _2`$ the critical resolved shear stress behaves like $`\sigma _c=0.64\mu b/l`$ and $`\sigma _c=1.24\mu b/l`$, respectively. For the $`300\AA `$ junction, for the case when $`\sigma _1=1.3\sigma _2`$, the atomistic simulations gave $`\sigma _c=.017\mu =0.8\mu b/l`$, which is in good agreement with our results. In the atomistic simulations, performing the scaling analysis is very computationally demanding and perhaps not even realizable. The scaling of the breaking stress in un-symmetric junctions with different arm length of the participating dislocations was found to be more complicated and will not be discussed here. Before presenting our concluding remarks, we point out the features that are missed in our model and the effect they may have on the results discussed thus far. In our simulations, the dislocations do not acquire a jog as they go past each other. While this does not affect the junction structure, it will alter the breaking stress, since the external stress should supply the energy required to create the jog. However, the jog contribution to the breaking stress is only a small fraction of the stress required to unzip the junctions. In order to examine the effect of the stress cut-off distance, $`r_c`$, we have also carried out all the above simulations by choosing $`r_c`$ to equal $`b/2`$ and $`2b`$. The junction structure showed very little difference in the two calculations. Also, the breaking stress in all the cases was within 10$`\%`$ of the values reported here. In conclusion, we have developed an efficient method to study dislocation interactions in FCC metals. We have illustrated that the method can provide “rules” like critical angle for junction formation and breaking stress criteria, which can be used in 3D dislocation dynamics models. The results from our simulations for junctions in different configurations, when appropriately averaged, can provide parameters related to junction strength in models for single crystal plasticity. Fits to these parameters from tension tests on single crystals have revealed a hierarchy of junction strengths in FCC crystals. Work is in progress to verify the observed hierarchy on the basis of our simulations. In the future we also hope to study other problems like the strength of dislocation junctions in alloys and interactions of extended dislocations with defects. We are grateful to M. Fivel, A. Needleman, M. Ortiz and D. Rodney for illuminating discussions. We also appreciate the support of the Caltech ASCI program and the Brown University MRSEC.
no-problem/9907/physics9907016.html	ar5iv	text	# References As is well known, magnetic reconnection is a very important physical phenomena in many astrophysical objects. It is believed that, magnetic reconnection can serve as such a mechanism to transfer the magnetic energy into heat and kinetic energy of the plasma . Generally, due to the nonlinearity of the MHD equations, it is very difficult to find its analytic solutions. We have to turn to the numerical simulations. But, unfortunately, numerical simulations are usually limited by artificial resistivity and the implementation of sensible boundary conditions . Anyway, the exact analytic solutions are needed and helpful. On one hand, they can provide the important insights into the problem concerned, and on the other hand, they can be used to test the relevant numerical methods. Just based on the beautiful exact solutions, Parker found that, diffusion is an essential part of the reconnection of lines of force at a neutral point . Since then, a lot of excellent analytic solutions have been found and applied to analyze some concrete models . Recently, Watson, Craig and Henton etc. have made a great progress in searching for the exact solutions . They found that, exact families of reconnection solutions can be constructed in both two and three dimensions, and, these solutions display the essential characteristics of fast reconnection. Furthermore, they have shown that, the reconnection solutions can be obtained by superposing the velocity and magnetic fields of simple magnetic annihilation model. So, annihilation model can provide a solid basis for our understanding the reconnection process. Up to now, there are few works on the time-dependent case . Most of the previous works in searching for the exact solutions concerntrate on the case of the steady MHD. But, it is desired to remove the restrictions of stationarity and incompressiblety etc. for the realistic application. In this paper, we present two time-dependent analytic solutions of the incompressible MHD. We begin by introducing the basic equations in section 1. The solutions are obtained in section 2. And, the relevant physics indicated by the solutions is analysized in section 3. Finally, the conclusions are summarized in section 4. 1. The Basic Equations The incompressible MHD equations can be written as the following nondimensional form : $`{\displaystyle \frac{\stackrel{}{v}}{t}}`$ $`=`$ $`(\stackrel{}{v})\stackrel{}{v}P+(\times \stackrel{}{B})\times \stackrel{}{B}`$ (1) $`{\displaystyle \frac{\stackrel{}{B}}{t}}`$ $`=`$ $`\times (\stackrel{}{v}\times \stackrel{}{B})+\eta ^2\stackrel{}{B}`$ (2) $`\stackrel{}{B}`$ $`=`$ $`0`$ (3) $`\stackrel{}{v}`$ $`=`$ $`0`$ (4) where the length, magnetic field $`\stackrel{}{B}`$, velocity $`\stackrel{}{v}`$, time $`t`$, gas pressure $`P`$ and the resistivity $`\eta `$ have been non-dimensionalized by the typical coronal parameters - length-scale $`L_c`$, background field strength $`B_c`$, Alfven speed $`v_A=B_c/(8\pi \rho )^{1/2}`$, Alfven travel time $`\tau _A=L_c/v_A`$, magnetic pressure $`B_c^2/(8\pi )`$ and $`v_AL_c`$ respectively. By taking the curl of Eq.(1), we can remove $`P`$ in above equations and get $$\frac{(\times \stackrel{}{v})}{t}=\times [(\stackrel{}{v})\stackrel{}{v}]+\times [(\times \stackrel{}{B})\times \stackrel{}{B}]$$ (5) Now, we need to deal with Eqs.(2, 3, 4, 5). Once we solve them and get $`\stackrel{}{v}`$ and $`\stackrel{}{B}`$, we can calculate the pressure $`P`$ according to Eq.(1). 2. Magnetic Annihilation Solution In order to solve Eqs.(2, 3, 4, 5), suppose that $`\stackrel{}{B},\stackrel{}{v}`$ have the following forms $`\stackrel{}{B}`$ $`=`$ $`\{\mathrm{\hspace{0.33em}0},Y(x,t),\mathrm{\hspace{0.33em}\hspace{0.33em}0}\}`$ (6) $`\stackrel{}{v}`$ $`=`$ $`\{U(x,t),yV(x,t),\mathrm{\hspace{0.33em}\hspace{0.33em}0}\}`$ (7) for the annihilation solution. Here, $`U,V,Y`$ are the functions of $`x`$ and $`t`$. And thus, Eq.(3) is satisfied automatically. Now, Eqs.(4, 5) and Eq.(2) give $`{\displaystyle \frac{}{x}}U(x,t)+V(x,t)`$ $`=`$ $`0`$ (8) $`{\displaystyle \frac{}{t}}V(x,t)+U(x,t){\displaystyle \frac{}{x}}V(x,t)+V(x,t)^2+f(t)`$ $`=`$ $`0`$ (9) $`{\displaystyle \frac{}{t}}Y(x,t)+{\displaystyle \frac{}{x}}(U(x,t)Y(x,t))\eta {\displaystyle \frac{^2}{x^2}}Y(x,t)`$ $`=`$ $`0`$ (10) where, $`f(t)`$ is a function of $`t`$ to be determined later by the boundary condition. If we further suppose that $$U(x,t)=U(x)g(t)$$ from Eq.(8), it is evident that $$V(x,t)=V(x)g(t).$$ Substitute the above two equations into Eqs.(8, 9), we achieve $`{\displaystyle \frac{}{t}}g(t)+c_1g(t)^2`$ $`=`$ $`0`$ (11) $`f(t)+c_2g(t)^2`$ $`=`$ $`0`$ (12) $`{\displaystyle \frac{}{x}}U(x)+V(x)`$ $`=`$ $`0`$ (13) $`U(x){\displaystyle \frac{}{x}}V(x)+V(x)^2c_1V(x)c_2`$ $`=`$ $`0`$ (14) with $`c_i(i=1,2,3,\mathrm{})`$ being the (integral) constants. Now, we are in position to state how to solve $`g(t),U(x)`$ and $`V(x)`$ from the above equations. Eq.(11) tells us that $$g(t)=\frac{1}{c_1t+c_3}.$$ (15) Starting out from Eq.(13, 14), we can obtain two solutions. A. Take $`U(x)`$ as the linear function of $`x`$, then we have $`U(x)`$ $`=`$ $`c_4x+c_5`$ (16) $`V(x)`$ $`=`$ $`c_4`$ (17) $`c_2`$ $`=`$ $`c_1c_4+c_4^2.`$ (18) B. Let $`U(x)=P(x)+c_4`$, and substitute it into Eqs.(13, 14), we arrive at $`{\displaystyle \frac{}{x}}P(x)+V(x)`$ $`=`$ $`0`$ (19) $`P(x){\displaystyle \frac{}{x}}V(x)+V(x)^2+c_4{\displaystyle \frac{}{x}}V(x)c_1V(x)c_2`$ $`=`$ $`0.`$ (20) if we further have $`P(x)=c_5V(x)`$ and $`\frac{}{x}V(x)=c_6V(x)`$, Eqs.(19, 20) can be satisfied easily. Then, we get $`V(x)`$ $`=`$ $`c_7e^{c_6x}`$ (21) $`U(x)`$ $`=`$ $`c_5c_7e^{c_6x}+c_4`$ (22) $`c_6c_4`$ $`=`$ $`c_1`$ (23) $`c_5c_6`$ $`=`$ $`1`$ (24) $`c_2`$ $`=`$ $`0.`$ (25) In the next step, inserting these solutions into Eq.(10), we can finally solve the magnetic field. To give a feeling on the form of the magnetic field $`\stackrel{}{B}`$, let us consider the asymptotic solution of $`Y(x,t)`$. When $`\eta 0`$, Eq.(10) is simplified as $$\frac{}{t}Y(x,t)+\frac{}{x}(U(x,t)Y(x,t))=0.$$ (26) Take Eqs.(15, 16) and Eq.(22) into account, we get $$Y(x,t)=c_9\frac{(c_1t+c_3)^{\frac{c_8}{c_1}}}{(c_4x+c_5)^{1+\frac{c_8}{c_4}}}$$ (27) for solution A, and $$Y(x,t)=c_9e^{c_6x}(c_4e^{c_6x}+c_5c_7)^{\frac{c_8}{c_4c_6}1}(c_1t+c_3)^{\frac{c_8}{c_1}}$$ (28) for solution B. With a suitable choice of the relevant constants, we can obtain $`Y(x,t)`$ $`=`$ $`{\displaystyle \frac{1}{x^3(t+1)^2}}`$ (29) $`Y(x,t)`$ $`=`$ $`{\displaystyle \frac{e^x}{(e^x1)^3(t+1)^2}}`$ (30) corresponding to A and B. Here, the singularity exists evidently at the neutral interface $`x=0`$ for both A and B. The presence of the singularity is naturally unavoidable when diffusion vanishes, as discussed by Parker . Eq.(10) tells us taht, $`Y(x,t)`$ can not be simply decomposed as the form of $`X(x)T(t)`$, after the similar supposition for $`U(x,t)`$ is made. To solve the magnetic field $`\stackrel{}{B}`$ analytically, a certain transformation is needed. For solution A, $`U(x,t)=\frac{c_4x+c_5}{c_1t+c_3}`$. Suppose $`Y(x,t)=Y(s)`$, and $$s=(c_4x+c_5)(c_1t+c_3)^\alpha $$ (31) with $`\alpha =\frac{1}{2}`$, Eq.(10) can be transformed into $$\eta \frac{^2}{s^2}Y(s)+\frac{c_12c_4}{2c_4^2}s\frac{}{s}Y(s)\frac{1}{c_4}Y(s)=0.$$ (32) and for solution B, $`U(x,t)=\frac{c_5c_7e^{c_6x}+c_4}{c_1t+c_3}`$, similarly, let $$s=c_5c_7e^{c_6x}(c_1t+c_3)^\alpha $$ (33) with $`\alpha =1`$, Eq.(10) can be read as $$c_6\eta s\frac{^2}{s^2}Y(s)+(\eta c_6s)\frac{}{s}Y(s)Y(s)=0.$$ (34) Now, the partial differential equation Eq.(10) has been transformed into the ordinary differential equations, whose solutions can be expressed as special functions. The solutions of Eq.(32) and Eq.(34) are $$Y(s)=C_1{}_{\mathrm{𝟏}}{}^{}𝐅_{\mathrm{𝟏}}^{}[\frac{c_4}{c_12c_4},\frac{1}{2},\frac{(c_1+2c_4)s^2}{4c_4^2\eta }]+sC_2{}_{\mathrm{𝟏}}{}^{}𝐅_{\mathrm{𝟏}}^{}[\frac{1}{2}\frac{c_4}{c_12c_4},\frac{3}{2},\frac{(c_1+2c_4)s^2}{4c_4^2\eta }]$$ (35) and $$Y(s)=C_1e^{\frac{s}{c_6\eta }}+C_2e^{\frac{s}{c_6\eta }}𝚪[0,\frac{s}{c_6\eta }]$$ (36) respectively. Here, $`C_1,C_2`$ are the integral constants, $`{}_{\mathrm{𝟏}}{}^{}𝐅_{\mathrm{𝟏}}^{}`$ is the Kummer confluent hypergeometric function while $`\mathrm{\Gamma }`$ the incomplete gamma function. Up to now, we have obtained two classes of the exact analytic solutions for both velocity field and magnetic field. We would like to give a short comment here. It is deserved to note that, Eqs.(31, 33) are actually the scale transformation with time $`t`$ as the scaling factor. So, our solutions will perform certain scaling characteristic, i.e., when the velocity fields are taken as Eqs.(16, 22, 15), the evolution of the magnetic field with time decribed by the solution of Eq.(10), can be described by the solution of the ordinary differential equation Eqs.(32, 34) dilated simply by time $`t`$ according to $`(c_1t+c_3)^\alpha `$. 3. Physical Characteristics of the Solution In this section, we discuss the physics implied by the above solutions. In doing so, we have to fix the free parameters. The relevant functions are taken as $$g(t)=\frac{1}{t+1}$$ (37) together with $`U(x)`$ $`=`$ $`x`$ (38) $`V(x)`$ $`=`$ $`1`$ (39) for solution A, and $`U(x)`$ $`=`$ $`e^x1`$ (40) $`V(x)`$ $`=`$ $`e^x`$ (41) for solution B. In fact, there is another possible choice such as $`U(x)=1e^x,V(x)=e^x`$ with g(t) given by Eq.(37) for the flow field of the annihilation solution, but it is very similiar to the one given by Eq.(40,41), so we will not discuss it here. The velocity field for solutions A and B are $$\stackrel{}{v}=\{\frac{x}{t+1},\frac{y}{t+1},\mathrm{\hspace{0.33em}\hspace{0.33em}0}\}$$ (42) and $$\stackrel{}{v}=\{\frac{e^x1}{t+1},\frac{ye^x}{t+1},\mathrm{\hspace{0.33em}\hspace{0.33em}0}\}$$ (43) respectively. They are symmetric (for A) or asymmetric (for B) stagnation-point flow weakening with time. For asymmetric case - solution B, the flow field is shown in Fig.1. With the given velocity fields $`U(x,t)`$, the equation of magnetic field $`B(x,t)`$ can be read as $$\eta \frac{^2}{s^2}Y(s)+\frac{3}{2}s\frac{}{s}Y(s)+Y(s)=0$$ (44) with the solution $$Y(s)=Cs{}_{\mathrm{𝟏}}{}^{}𝐅_{\mathrm{𝟏}}^{}[\frac{5}{6},\frac{3}{2},\frac{3s^2}{4\eta }](s=\frac{x}{\sqrt{t+1}})$$ (45) and $$\eta s\frac{^2}{s^2}Y(s)+(s+\eta )\frac{}{s}Y(s)+Y(s)=0$$ (46) with the solution $$Y(s)=Ce^{\frac{s}{\eta }}(𝚪[0,\frac{s}{\eta }]𝚪[0,\frac{1}{\eta }])(s=\frac{e^x}{t+1})$$ (47) for A and B correspondingly. To determine the relevant constants, $`Y\|_{x=0,t=0}=0`$ and $`\frac{Y}{x}\|_{x=0,t=0}=C`$ are required here. With the known magnetic fields, we can now solve the current and the energy dissipation rate by Joule heat. Based on Eqs.(45, 47), we achieve $$j(x,t)=C\{\frac{1}{\sqrt{t+1}}{}_{\mathrm{𝟏}}{}^{}𝐅_{\mathrm{𝟏}}^{}[\frac{5}{6},\frac{3}{2},\frac{3x^2}{4\eta (t+1)}]\frac{5x^2}{6\eta (t+1)^{3/2}}{}_{\mathrm{𝟏}}{}^{}𝐅_{\mathrm{𝟏}}^{}[\frac{11}{6},\frac{5}{2},\frac{3x^2}{4\eta (t+1)}]\}$$ (48) for solution A, and $$j(x,t)=C\{1+\frac{1}{\eta (t+1)}e^{\frac{e^x}{\eta (t+1)}x}(𝚪[0,\frac{e^x}{\eta (t+1)}]𝚪[0,\frac{1}{\eta }])\}$$ (49) for solution B. Finally, according to $$𝒲_\eta =<\eta j^2>=\eta j(x,t)^2𝑑V$$ (50) we can calculate the power provided by Ohmic dissipation. The dependence of $`B(x,t),j(x,t)`$ on $`(x,t)`$ and $`𝒲_\eta `$ on time $`t`$ are shown in Figs.(2-7). It can be seen from the figures that, for solution A, the maximum of the magnetic field departs from the initial magnetic neutral interface with time. So, the dissipation area will get gradually wide. Following it, the current becomes smaller and smaller. The biggest rate of the diffuse of the magnetic energy due to the Joule heat takes place at the beginning. The solution B is interesting, its current is evidently a soliton. Hence, when the wave-packet of the current is located in the domain that we consider, the dissipation power keeps as a constant. But, once it passes through the domain, the dissipation power will descend with time. 4. Conclusions and Discussions In conclusion, we have obtained two classes of full analytic time-dependent solutions for magnetic annihilation by directly solving the MHD equations. The solutions which we get here have such a scaling property that the evolution of the relevant physical observables are described by the solutions of certain ordinary differential equations, with the variables being some functions of $`x`$ dilated by time $`t`$. Besides, in the case of asymmetric inflow, a soliton-like current can be formed. Finally, the solutions are used to reveal the relevant physical characteristics in the process of the magnetic annihilation. However, the time-dependent solutions obtained here are the relative simple ones. Based on the method provided by Watson, Craig and Henton etc. , we expect to construct the time-dependent reconnection solutions. The work along this direction is under way. Acknowledgment: One of the authors (Y. Liu) would like to thank Dr. J. Q. Li for helpful discussions and Prof. P. G. Watson for great valuable comments. This work is partly supported by Chinese Postdoctoral Foundation.
no-problem/9907/astro-ph9907211.html	ar5iv	text	# A New Spin on the Problem of HB Gaps: Stellar Rotation along the Blue Horizontal Branch of Globular Cluster M13Based on observations obtained at the W.M. Keck Observatory, which is operated jointly by the California Institute of Technology and the University of California. Based on observations with the NASA/ESA Hubble Space Telescope, obtained at the Space Telescope Science Institute, which is operated by AURA, Inc., under NASA contract NAS 5-26555. ## 1 Introduction A number of unresolved issues in post-main-sequence stellar evolution revolve around the nature of stars on the horizontal branch (HB). The HB stars in globular clusters are readily identified by their position in a cluster’s color-magnitude diagram, and are assumed to have the same age and initial composition as the other stars in the cluster. The photometric properties of the HB are very sensitive to stellar composition and structure, so detailed study of the HBs of many different clusters allows us to test models of post-main-sequence stellar evolution. The distribution of stars along the horizontal branch differs from cluster to cluster. This variety in the HB color morphology is primarily a function of cluster metallicity (Sandage & Wallerstein 1960; Sandage & Wildey 1967), with metal-rich clusters tending to have short red HBs, and metal-poor clusters exhibiting predominantly blue HBs. Some other parameter (or set of parameters) is evidently also at work, however, as clusters with nearly identical metallicities can show very different HB color distributions (van den Bergh 1967; Sandage & Wildey 1967). This ‘second parameter’ was initially thought to be cluster age (Searle & Zinn 1978; Lee, Demarque, & Zinn 1994; Stetson, Vandenberg, & Bolte 1996), but several alternative or additional second-parameter candidates have since been suggested, including helium abundance and mixing (Sweigart 1997), CNO abundance (Rood & Seitzer 1981), central concentration of the cluster (Fusi Pecci et al. 1993, Buonanno et al. 1997), and distribution of stellar rotation rates (Peterson et al. 1995). Most of these proposed mechanisms influence the HB morphology by regulating either the core mass, or the amount of mass loss that a star experiences on the upper red giant branch (RGB). Greater mass loss results in a smaller hydrogen envelope surrounding the $`0.5M_{}`$ helium core, and the star appears on the zero-age horizontal branch (ZAHB) with a higher photospheric temperature and higher surface gravity. There has been some success in synthesizing the observed HB distributions by evolving stellar models up the RGB and on to the HB with a mean mass loss $`\mathrm{\Delta }M0.15M_{}`$ and a dispersion in mass loss $`\sigma 0.010.03M_{}`$ (Lee, Demarque, & Zinn 1990), but this approach fails to reproduce details of the complex color morphology seen in many clusters. Particularly intriguing are the gaps which appear in the blue HB sequence of many globular clusters, including M15 (Buonanno et al. 1983), M13 and M80 (Ferraro et al. 1997, Ferraro et al. 1998), NGC 2808 (Sosin et al. 1997), NGC 6229 (Catelan et al. 1998), and NGC 6273 (Piotto et al. 1999), as well as the metal-poor halo (Newell 1973). High-precision photometry reveals narrow regions of the HB color axis which are underpopulated at a statistically significant level, and which seem to occur at similar locations in different clusters, strengthening the claim that they are real features, and not just statistical artifacts (Ferraro et al. 1997, Piotto et al. 1999). The presence of such gaps poses a major challenge to HB theory, as it is difficult to envision a single mass-loss mechanism which implements such narrow ‘forbidden zones’ in the envelope mass distribution. Alternatively, the gaps may mark the boundaries between separate, discrete populations of HB stars, which differ in their origin or evolution. We then might expect to see some additional qualitative difference between stars on either side of the gap. In order to evaluate how various stellar characteristics change as a function of position along the HB and from cluster to cluster, we have undertaken a program to measure rotation rates and chemical abundances of blue horizontal-branch (BHB) stars in M3, M13, M15, M92, and M68 via high-resolution echelle spectroscopy. M13 (NGC 6205) is one of the closest and best-studied globulars, with $`(mM)=14.35`$ mag (Peterson 1993) and a metallicity \[Fe/H\] $`=1.51`$ dex measured from red giant abundances (Kraft et al. 1992). Its BHB extends from the blue edge of the RR Lyrae gap to the helium-burning sequence (a ‘long blue tail’), interrupted by one or more gaps (Ferraro et al. 1997). We have observed thirteen of M13’s BHB stars in the range $`7000\mathrm{K}<T_{\mathrm{eff}}<20000\mathrm{K}`$, specifically in order to span the most obvious gap, which is located at $`T_{\mathrm{eff}}11000\mathrm{K}`$ and labelled ‘G1’ by Ferraro et al. In this paper, we describe the measurement of projected stellar rotation ($`v\mathrm{sin}i`$) for this sample; the abundance anomalies also exhibited by these stars have been reported previously (Behr et al. 1999). Our measurements build upon prior work by Peterson, Rood, and Crocker (1995, henceforth P95), who undertook a spectroscopic survey of BHB stars in M13, using the O I triplet at 7775 Å to evaluate \[O/Fe\] and $`v\mathrm{sin}i`$. In the 29 BHB stars for which $`v\mathrm{sin}i`$ was determined, they found rotations as fast as $`40\mathrm{km}\mathrm{s}^1`$, although the small number of these higher $`v\mathrm{sin}i`$ values strongly suggests that the true distribution of rotational $`v`$ is bimodal, with roughly a third of M13’s BHB stars being ‘fast rotators’ ($`40\mathrm{km}\mathrm{s}^1`$) and the other two-thirds being ‘slow rotators’ ($`15\mathrm{km}\mathrm{s}^1`$). The existence of such fast rotation is difficult to explain. The progenitors of these HB stars were solar type or later, and are expected to have shed most of their angular momentum via magnetically-coupled winds early in their main-sequence lifetimes, reaching the $`v<2\mathrm{km}\mathrm{s}^1`$ observed in the Sun and other similar Population I dwarfs. Assuming solid-body rotation and a homogenous distribution of angular momentum per unit mass, Cohen & McCarthy (1997) estimate that the HB stars should be rotating no faster than $`10\mathrm{km}\mathrm{s}^1`$, but this prediction clearly disagrees with P95, as well as prior measurements in other clusters (Peterson 1983, 1985a, 1985b) and Cohen & McCarthy’s own $`v\mathrm{sin}i`$ findings in globular cluster M92. Pinsonneault, Deliyannis, & Demarque (1991) explain the anomalously fast rotation via core-envelope decoupling—the stellar core retains much of its original angular momentum, while only the envelope is braked, and the rapidly rotating core is then revealed after much of the envelope is lost on the RGB. Such core behavior would explain the fast rotators, although the model does not explain why only a third of the HB stars would show such an effect, nor does it agree with the slow core rotations inferred for young stars (Bouvier 1997, Queloz et al. 1998) and the solar interior (Corbard et al. 1997, Charbonneau et al. 1998). The existence of this quickly-rotating population is suggestive when M13 is compared to other clusters. The other two globular clusters in the P95 study, M3 and NGC 288, were found to have significantly lower maximum rotation among their BHB stars, $`v\mathrm{sin}i<20\mathrm{km}\mathrm{s}^1`$ in nearly all cases. The HB distributions of these two clusters are also somewhat less blue than M13—neither show any HB population at $`T_{\mathrm{eff}}`$ above 14500 K—so the obvious inference is that the fast stellar rotation is somehow related to the presence of a blue tail. Mengel & Gross (1976) point out a possible mechanism by which this might occur—fast core rotation in pre-HB stars nearing the tip of the red giant branch would forestall the helium flash, such that more envelope mass is lost to stellar winds, and the star ends up further to the blue when it reaches the HB. This mechanism would function on a star-by-star basis, so one would thus expect to see higher rotation in the hotter stars, as the quickly-rotating core is revealed by progressively greater amounts of RGB mass loss. No such variation in $`v\mathrm{sin}i`$ as a function of $`T_{\mathrm{eff}}`$ was seen in the P95 results, although the rapid disappearance of the oxygen lines above 10000 K limited the temperature coverage of their measurements. Since our sample extends to $`T_{\mathrm{eff}}19000\mathrm{K}`$, we are able to test this hypothesis more thoroughly. ## 2 Observations and Reduction Our spectra were collected using the HIRES spectrograph (Vogt et al. 1994) on the Keck I telescope during four observing runs on 1998 June 27, 1998 August 20–21, 1998 August 26–27, and 1999 March 09–11. A 0.86-arcsec slit width yielded a nominal $`R=45000`$ ($`v=6.7\mathrm{km}\mathrm{s}^1`$) per 3-pixel resolution element. Spectral coverage ran from $`39405440`$ Å ($`m=9066`$) for the June observations, and $`38906280`$ Å ($`m=9157`$) for the August and March observations, with slight gaps above 5130 Å where the free spectral range of the orders overfilled the detector. We limited frame exposure times to 1200 seconds, to minimize susceptibility to cosmic ray accumulation, and then coadded three frames per star. $`S/N`$ ratios were on the order of $`5090`$ per resolution element, permitting us to accurately measure even weak lines in the spectra. Nine of the thirteen stars in our sample were selected from photometry of archival HST WFPC-2 images of the center of M13, specifically reduced for this project. The program stars were selected to be as isolated as possible; the HST images showed no apparent neighbors within $`5`$ arcseconds. The seeing during the HIRES observations was sufficiently good ($`0.81.0`$ arcsec) to avoid any risk of spectral contamination from other stars. The other four M13 HB program stars were taken from the sample of P95, to permit a comparison to published values. All of the stars from P95 are located in the cluster outskirts, where crowding is not so problematic. Positions, finding charts, photometry, and observational details for the entire target list will be provided in a later paper (Behr 1999b). We used a suite of routines developed by J. K. McCarthy (1988) for the FIGARO data analysis package (Shortridge 1993) to reduce the HIRES echellograms to 1-dimensional spectra. Frames were bias-subtracted, flat-fielded against exposures of the internal quartz incandescent lamp (thereby removing much of the blaze profile from each order), cleaned of cosmic ray hits, and coadded. Thorium-argon arc spectra provided wavelength calibration, and arc line profiles were used to determine the instrumental broadening profile (see below). Sky background was negligible, and 1-D spectra were extracted via simple pixel summation. A 10th-order polynomial fit to line-free continuum regions completed the normalization of the spectrum to unity. ## 3 Analysis and Results The resulting stellar spectra show ten to over two hundred metal absorption lines each. Even to the eye, the hotter stars in our sample are clearly slow rotators, with exceedingly sharp, narrow lines. Measurement of $`v\mathrm{sin}i`$ broadening traditionally entails cross-correlation of the target spectrum with a rotational reference star of similar spectral type, but this approach assumes that the template star is truly at $`v\mathrm{sin}i=0`$, which is rare. Furthermore, given the abundance peculiarities that many of these stars exhibit (Behr et al. 1999), it is difficult to find an appropriate spectral analog. Since we are able to resolve the line profiles of our stars, we instead chose to measure $`v\mathrm{sin}i`$ by fitting the profiles directly, taking into account other non-rotational broadening mechanisms. First, an instrumental broadening profile was constructed from bright but unsaturated arc lines for each night of observation. Those line profiles which deviated most from the mean were successively discarded, until the rms deviation reached 1% of peak intensity or less. This technique proved very effective at eliminating blended arc lines, residual cosmic ray hits, and otherwise non-representative line profiles from the composite profile. The variation in instrumental broadening over the detector area proved negligible, as the 100 to 200 lines which made up each composite profile were distributed evenly across the chip. The resulting instrumental profile was almost perfectly Gaussian, with a FWHM of $`6.3\mathrm{km}\mathrm{s}^1`$, slightly narrower than that expected for $`R=45000`$. We estimate a thermal Doppler broadening of $`3\mathrm{km}\mathrm{s}^1`$ FWHM, and a microturbulent broadening $`\xi =2\mathrm{km}\mathrm{s}^1`$ from the previous abundance analysis of these stars, so we convolved the instrumental profile with a Gaussian of $`3.6\mathrm{km}\mathrm{s}^1`$ FWHM to account for these effects. This profile was convolved with a hemicircular rotation profile for the specified $`v\mathrm{sin}i`$ to create the final theoretical line profile. No attempt was made to include limb darkening, as this should have only minimal impact on the final result. Each observed line in a spectrum was fit to the theoretical profile using an iterative least-squares algorithm. Four free parameters were used for fitting: $`v\mathrm{sin}i`$, line center $`\lambda _{\mathrm{ctr}}`$, line depth, and continuum level. Data and best-fit profiles for three representative lines in each of two stars are depicted in Figure 1. The values for $`v\mathrm{sin}i`$ from different spectral lines generally agree quite well, as shown in Figure 2. We removed helium lines, blended lines such as the Mg II 4481 triplet, and extreme outliers (those more than $`3\sigma `$ from the mean) from each line list, and calculated the mean and rms error in the mean in standard statistical fashion. Rotation results for the 13 stars observed in this study are given in Table 1. To quantify a star’s position on the HB, we use $`T_{\mathrm{eff}}`$, which is derived from photographic $`BV`$ photometry for the P95 stars, and from $`UV`$ photometry for the nine HST-selected stars, as detailed in Behr et al. 1999. (The photographic photometry temperatures will be refined by CCD photometry of the cluster in the near future, but are adequate for our present purpose.) In Figure 3, we plot both our results and the results from P95 as a function of temperature, along with photometric data illustrating the temperature distribution of the HB star population. The $`V`$ vs. $`T_{\mathrm{eff}}`$ color-magnitude diagram in the middle panel, and the accompanying histogram in the lower panel, are based on our reduction of the archival HST images. ## 4 Discussion For the four stars in common, our $`v\mathrm{sin}i`$ agree quite well with those of P95. The existence of the quickly-rotating BHB population, then, seems secure. However, our data clearly show that the fast rotators are constrained to the cooler end of the BHB, as all of the stars between 11000 K and 19000 K appear at $`v\mathrm{sin}i<10\mathrm{km}\mathrm{s}^1`$. This abrupt change in the rotation distribution at 11000 K (the location of the gap) was hinted at in the prior measurements, but appears much more clearly as the observations are extended to hotter stars. The measurement of $`v\mathrm{sin}i`$ by line broadening is inherently statistical, since a small value of $`v\mathrm{sin}i`$ in a single star might be due to its rotation axis lying close to the line of sight. Given an isotropic distribution in axis orientation, however, a large $`\mathrm{sin}i`$ is much more likely than a small one—the probability that $`\mathrm{sin}i<0.25`$ is about 3%, for example—and with several stars in the sample, the likelihood that the small observed $`v\mathrm{sin}i`$ are due to chance polar orientation becomes exceedingly slim. We are confident, then, that the hot stars in our sample are truly rotating no faster than $`10\mathrm{km}\mathrm{s}^1`$, and that we are seeing a real difference in stellar rotation characteristics between the cooler and hotter stars on the BHB. The nature of this difference, however, seems to argue against the hypothesis proposed by P95. If the blue end of M13’s HB is populated by stars which have undergone extreme RGB mass loss due to fast core rotation, then that angular momentum should become evident at the surface and appear as higher $`v\mathrm{sin}i`$. Instead, the hotter stars, which have suffered greater mass loss, are observed to have slower rotational velocities. The mechanisms that regulate mass loss are still poorly understood, so it is certainly possible that faster rotation might somehow limit the amount of mass lost, and keep those stars at the red end of the BHB. A more likely possibility, also mentioned in P95, is that enhanced mass loss, due to some other second parameter, is very efficient at removing angular momentum from the convective envelopes of the RGB stars. Those stars which lose more mass, ending up further towards the blue end of the BHB, would thus be rotating more slowly. In this case, the range of rotation rates would be better described as a result, not a cause, of the second-parameter phenomenon. Such a mechanism would explain why the hotter stars in our sample are all slow rotators, but does not account for the presence of the slow rotating population at lower BHB temperatures. An alternative hypothesis is that the slow rotators are the ‘normal’ HB stars, and that the rapid rotators redward of the gap represent the progeny of merged stars, perhaps a subpopulation of blue straggers (BS), similar to the suggestion by Fusi Pecci et al. (1992) that BS progeny populate the red HB. Such stellar merger products would likely retain excess angular momentum even through the HB stage, although the issue is complicated by other possible effects of mergers, including mass loss. A more comprehensive assessment of such anomalously high HB rotation in M92 (Cohen & McCarthy 1997) and other clusters, and its correlation with blue straggler populations, should be undertaken. Of even greater significance is the abruptness of the change in rotational signature, and its location along the HB axis, which closely coincides with the location of the gap G1, as illustrated in Figure 3. Below 11000 K, the quickly-rotating stars account for roughly a third of the HB population, but they abruptly disappear at $`T_{\mathrm{eff}}>11000\mathrm{K}`$. In addition, the slowly-rotating segment of the cooler population does appear to have a higher peak $`v\mathrm{sin}i14\mathrm{km}\mathrm{s}^1`$ than the hotter stars with $`v\mathrm{sin}i<10\mathrm{km}\mathrm{s}^1`$. This discontinuous change in the distribution of $`v\mathrm{sin}i`$ that we observe on either side of the gap supports the hypothesis that the gap separates two different populations of HB stars, and may provide an important hint towards eventual identification of the mechanism or mechanisms which differentiate the two populations. The low $`v\mathrm{sin}i`$ exhibited by the hotter stars is also relevant to the issue of the anomalous photospheric abundances reported by Behr et al. (1999)—helium underabundance by as much as a factor of 300, and enhancement of metals to solar and supersolar levels, probably due to diffusion mechanisms in the radiative atmospheres of the hotter, higher-gravity stars. Slow stellar rotation, and smaller resulting meridional circulation currents, are likely to be prerequisites for the diffusion to be effective (Michaud et al. 1983). If the observed HB color morphologies are due to changes in atmospheric structure brought on by these metal enhancements, as suggested by Caloi (1999) and Grundahl et al. (1999), then the change in maximum $`v\mathrm{sin}i`$ along the HB could prove closely linked to the appearance of the gaps. Further theoretical consideration of the influence of stellar rotation on the onset of the diffusion effects and the impact of diffusion enhancements on photometric colors will be necessary to evaluate this possibility. Although we are still far from explaining the underlying cause of this gap, or the origin of the anomalously high rotation rates of the cooler BHB stars, these observations will help to constrain theoretical models, and perhaps stimulate new ones. Stellar rotation rate clearly has some relation to the second parameter problem, although the nature of this relation is yet to be determined. Further observations, both in M13 and in other clusters which show similar gaps, will show whether this rotational discontinuity is a ubiquitous feature of globular cluster BHBs, and how it is correlated with their morphology. These observations would not have been feasible without the HIRES spectrograph and the Keck I telescope. We are indebted to Jerry Nelson, Gerry Smith, Steve Vogt, and many others for making such marvelous machines, to the W. M. Keck Foundation for making it happen, and to a bevy of Keck observing assistants for making them work. SGD was supported, in part, by the Bressler Foundation, and STScI grant GO-7470. PC gratefully acknowledges support provided by the Sherman M. Fairchild Foundation. This research has made use of the SIMBAD database, operated at CDS, Strasbourg, France.
no-problem/9907/physics9907042.html	ar5iv	text	# Some Exact Results of Hopfield Neural Networks and Applications # Hong-Liang Lu, Corresponding author, Ph.D student major to Radio Physics in Physics Department, Science College. Address: Physics Department, Science College, 20 Chengzhong Road, Jiading, Shanghai 201800, China. Email: xjqiu@srcap.stc.sh.cn Xi-Jun Qiu, Professor and Ph.D supervisor of Theoretical Physics, Physics Department, Shanghai University. (Physics Department, Science College, Shanghai University) ## Abstract A set of fixed points of the Hopfield type neural network was under investigation. Its connection matrix is constructed with regard to the Hebb rule from a highly symmetric set of the memorized patterns. Depending on the external parameter the analytic description of the fixed points set had been obtained. And as a conclusion, some exact results of Hopfield neural networks were gained. PACS Number(s): 87.10.+e, 05.45.+b Key Words: Neural Networks, Hebb rule, Fixed points Over the past decade, there has been an explosion of interest in so-called artificial neural network(ANN) technology. ANNs are a model of computing inspired by the brain-, consisting of a collection of model ”neurons” connected by model ”synapses.” Computation in the network is performed in a distributed fashion, by propagating excitatory and inhibitory activations across the synapses, and computing the neuronal outputs as a nonlinear (typically sigmoidal) function of total synaptic input. These networks also have a capacity to ”learn” to perform a given computation by adjusting real-valued synaptic connection strengths (weight values) between units. ANNs are of considerable interest both for biological modeling of information processing in the nervous system, and for solving many classes of complex real-world applications. J.J.Hopfield boosted neural network research at the beginning of the 1980s with the publication of a famous paper on artificial neural networks, which he used for pattern completion and to solve optimization problems. These networks consist of one layer of neurons that are completely connected with each other. Hopfield analysed the behavior of networks belonging to that type, and could prove mathematically that stable behavior may be achieved under certain conditions. It can be shown that the dynamic behavior of Hopfield type neural networks is described by an energe surface. Each network state corresponds to a certain position on that surface. Through external clamping, neurons may be forced to certain states of activity, and thus the whole network may be forced to move to a well defined point on the energy surface. If the network is released, i.e. external clamping is removed, it will change its state in such a way that it moves on the energy surface towards new states of lower energy. Finally, neuron states will stop changing if a local minimum in the energy surface is reached. Through careful selection of weights, ocillations will be avoided. A set of fixed points of the Hopfield type neural networkis under investigation. Its connection matrix is constructed with regard to the Hebb rule from a $`(p\times n)`$-matrix $`𝐒`$ of memorized patterns: $$𝐒=\left(\begin{array}{ccccccc}1x& 1& \mathrm{}& 1& 1& \mathrm{}& 1\\ 1& 1x& \mathrm{}& 1& 1& \mathrm{}& 1\\ \mathrm{}& \mathrm{}& \mathrm{}& \mathrm{}& \mathrm{}& \mathrm{}& \mathrm{}\\ 1& 1& \mathrm{}& 1x& 1& \mathrm{}& 1\end{array}\right).$$ Here $`n`$ is the number of neurons, $`p`$ is the number of memorized patterns $`\stackrel{}{s}^{(l)}`$, which are the rows of the matrix $`𝐒`$, and $`x`$ is an arbitrary real number. Depending on $`x`$ the memorized patterns $`\stackrel{}{s}^{(l)}`$ are interpreted as $`p`$ distorted vectors of the standard $$\stackrel{}{\epsilon }(n)=(\underset{n}{\underset{}{1,1,\mathrm{},1}}).$$ $`(1)`$ We denote by $`\stackrel{}{\epsilon }(k)`$ the configuration vector which is collinear to the bisectrix of the principle orthant standard-vector. Next, $`n`$ is the number of the spin variables, $`p`$ is the number of the memorized patterns and $`q=np`$ is the number of the nondistorted coordinates of the standard-vector. Configuration vectors are denoted by small Greek letters. We use small Latin letters to denote vectors whose coordinates are real. The problem is as follows: the network has to be learned by $`p`$-times showing of the standard (1), but a distortion has slipped in the learning process. How does the fixed points set depends on the value of this distortion $`x`$? Depending on the distortion parameter $`x`$ the analytic description of the fixed points set has been obtained. It turns out to be very important that the memorized patterns $`\stackrel{}{s}^{(l)}`$ form a highly symmetric group of vectors: all of them correlate one with another in the same way: $$(\stackrel{}{s}^{(l)},\stackrel{}{s}^{(l^{})})=r(x),$$ $`(2)`$ where $`r(x)`$ is independent of $`l,l^{}=1,2,\mathrm{},p.`$ Namely this was the reason to use the words ”highly symmetric” in the title. It is known , that the fixed points of a network of our kind have to be of the form: $$\stackrel{}{\sigma }^{}=(\sigma _1,\sigma _2,\mathrm{},\sigma _p,1,\mathrm{},1),\sigma _i=\{\pm 1\},i=1,2,\mathrm{},p.$$ $`(3)`$ Let’s join into one class $`\mathrm{\Sigma }^{(k)}`$ all the configuration vectors $`\stackrel{}{\sigma }^{}`$ given by Eq.(3), which have $`k`$ coordinates equal to ”–1” among the first $`p`$ coordinates. The class $`\mathrm{\Sigma }^{(k)}`$ consists of $`C_p^k`$ configuration vectors of the form (3), and there are $`p+1`$ different classes $`(k=0,1,\mathrm{},p)`$. Our main result can be formulated as a Theorem. Theorem. As $`x`$ varies from $`\mathrm{}`$ to $`\mathrm{}`$ the fixed points set is exhausted in consecutive order by the classes of the vectors $$\mathrm{\Sigma }^{(0)},\mathrm{\Sigma }^{(1)},\mathrm{},\mathrm{\Sigma }^{(K)},$$ and the transformation of the fixed points set from the class $`\mathrm{\Sigma }^{(k1)}`$ into the class $`\mathrm{\Sigma }^{(k)}`$ occurs when $`x=x_k`$: $$x_k=p\frac{n(2k1)}{n+p2(2k1)},k=1,2,\mathrm{},K.$$ If $`\frac{p1}{n1}<\frac{1}{3}`$, according this scheme all the $`p`$ transformations of the fixed points set are realized one after another and $`K=p`$. If $`\frac{p1}{n1}>\frac{1}{3}`$, the transformation related to $$K=\left[\frac{n+p+2}{4}\right]$$ is the last. The network has no other fixed points. The Theorem makes it possible to solve a number of practical problems. We would like to add that the Theorem can be generalized onto the case of arbitrary vector $$\stackrel{}{u}=(u_1,u_2,\mathrm{},u_p,1,\mathrm{},1),\underset{i=1}{\overset{p}{}}u_i^2=p$$ being a standard instead the standard (1). Here memorized patterns $`\stackrel{}{s}^{(l)}`$ are obtained by the distortion of the first $`p`$ coordinates of the vector $`\stackrel{}{u}`$ with regard to the fulfillment of Eqs.(2). The obtained results can be interpreted in terms of neural networks, Ising model and factor analysis. Acknowledgement The authors acknowledge the benefit of extended interaction with Miss Jie-yan Bai of Shanghai Research and Development Center for Fiber Optic Technology, Shanghai 803 Research Institute, who has helped us develop or clarify several ideas and issues that appear in this paper. We also extend our thanks to Prof. Yu-Long Mo of the School of Information and Communication Engineering, Shanghai University.
no-problem/9907/cond-mat9907174.html	ar5iv	text	# Spectral Statistics and Dynamical Localization: sharp transition in a generalized Sinai billiard ## Abstract We consider a Sinai billiard where the usual hard disk scatterer is replaced by a repulsive potential with $`V(r)\lambda r^\alpha `$ close to the origin. Using periodic orbit theory and numerical evidence we show that its spectral statistics tends to Poisson statistics for large energies when $`\alpha <2`$ and to Wigner-Dyson statistics when $`\alpha >2`$, while for $`\alpha =2`$ it is independent of energy, but depends on $`\lambda `$. We apply the approach of Altshuler and Levitov \[Phys. Rep. 288, 487 (1997)\] to show that the transition in the spectral statistics is accompanied by a dynamical localization-delocalization transition. This behaviour is reminiscent of a metal-insulator transition in disordered electronic systems. The statistical distribution of quantum mechanical energy eigenvalues is of fundamental interest in diverse areas of physics such as condensed matter, atomic, and nuclear physics . In the strict quasiclassical limit, where the de Broglie wavelength is much smaller than all other lengthscales, the theory of quantum chaos relates the spectral statistics of a quantum system to the dynamics of its classical counterpart. While chaotic classical dynamics leads to Wigner-Dyson random matrix statistics , integrable dynamics generically gives rise to Poisson statistics (uncorrelated eigenvalues) . There is overwhelming evidence, both experimental and numerical, for these results . The spectral statistics of systems with mixed classical dynamics is expected to be described by superposing level sequences with Poisson and random matrix statistics, where the respective mean level spacing is determined from the size of the corresponding phase space volume with regular or chaotic dynamics . However, when the de Broglie wavelength is not the smallest lengthscale in the system, the spectral statistics is not solely determined by the classical dynamics. For example, the classical motion of an electron in a (three-dimensional) disordered system, such as a metal with substantial impurity scattering, can be regarded as completely chaotic. Nevertheless, upon variation of the disorder strength at a fixed energy (Fermi energy), the spectral statistics undergoes a sharp transition from Wigner-Dyson to Poisson statistics . Note that at the critical point the de Broglie wavelength is of the same order of magnitude as the elastic mean free path, a classical lengthscale. The transition in the spectral statistics is accompanied by a transition from extended to localized eigenstates (Anderson metal-insulator transition) . In the case of disordered systems one can therefore attribute the deviation of the spectral statistics from what one would expect on the basis of the classical dynamics to the quantum phenomenon of localization. A similar effect on the spectral statistics may be caused by dynamical localization . Dynamical localization occurs for example in a circular billiard with a rough boundary, where in a certain range of parameters the quantum eigenstates are localized in angular momentum space despite that the classically chaotic dynamics leads to diffusive spreading of an initial angular momentum distribution . As a function of boundary roughness there is a crossover between localized and extended eigenstates, which takes place when the de Broglie wavelength $`\lambda _{\mathrm{dB}}`$ is of the same order of magnitude as a classical lengthscale set by the roughness, $`(dR/d\theta )^2/R_0`$, where $`R(\theta )`$ defines the rough circle in polar coordinates, $`R_0=R`$, and the average is over the angle . At the same time the level statistics changes smoothly from Poisson to Wigner-Dyson statistics . It is the purpose of the present article to show that a dynamical (non-random) system may also display a sharp transition in the spectral statistics. This is accompanied by a dynamical localization-delocalization transition of the eigenstates and is reminiscent of a metal-insulator transition in disordered systems. We thereby extend the list of analogies between dynamical and disordered systems . We consider a generalization of the well-known Sinai billiard (SB). Sinai proved that the free motion of a classical particle being specularly reflected from a disk of radius $`R`$ inside a square with periodic boundary conditions is completely chaotic . We modify the SB by replacing the disk with the scattering potential $$V(r)=\{\begin{array}{cc}\lambda \left(\left(\frac{R}{r}\right)^\alpha 1\right),\hfill & r<R\hfill \\ 0,\hfill & r>R\hfill \end{array},$$ (1) where $`\alpha `$ and $`\lambda `$ are positive parameters (see also Fig. 1). Note that in the limit $`\alpha \mathrm{}`$ the SB is recovered. In a chaotic billiard, the replacement of a hard wall by a soft potential barrier generically leads to the formation of stable islands , however in the present case these are hardly visible in a Poincaré surface of section and are not relevant for the effect studied here. A similar system was considered by Altshuler and Levitov . They focused on the properties of the eigenstates and proved the occurence of a dynamical localization-delocaliztion transition. In contrast, the emphasis in the present work is on a transition in the level statistics and on its semiclassical origin. The transition occurs as a function of $`\alpha `$ at $`\alpha =2`$, irrespective of the value of $`\lambda `$. It is a sharp transition in the limit of large energy. Using periodic orbit theory we find that it is caused by a competition between regular (‘bouncing ball’-type ) and chaotic orbits. In the second part of this article we show that the approach of Altshuler and Levitov may be applied also to the present case. We thereby demonstrate that the transition in the spectral statistics is accompanied by a dynamical delocalization transition of the eigenfunctions. For large energies, the qualitative quantum dynamics of the generalized Sinai billiard (GSB) is determined by the interplay of three lengthscales, (i) the de Broglie wavelength $`\lambda _{\mathrm{dB}}`$, (ii) the radius $`r_c=(E/\lambda +1)^{1/\alpha }R`$ of the classically forbidden area, and (iii) a typical length $`\stackrel{~}{l}`$ of an orbit before its direction is randomized by scattering from the potential. When $`\alpha >2`$ the potential (1) effectively acts as a hard wall, since the radial wave function then vanishes like $`\mathrm{exp}(cr^{1\alpha /2})`$ with $`c=2\sqrt{\lambda R^\alpha }/(\alpha 2)`$ near the origin, i.e. faster than any power. To estimate the behaviour of the spectral statistics, we approximate the potential (1) by a hard disk with the energy dependent radius $`r_c`$. According to the semiclassical theory , the structure in the spectrum on the scale $`\delta E`$ is determined by periodic orbits of length $`l=hv/\delta E`$, where $`h`$ denotes Planck’s constant and $`v`$ the velocity of the particle. For $`l>\stackrel{~}{l}`$ the chaotic orbits dominate and the spectral statistics will be random-matrix-like, while for $`l<\stackrel{~}{l}`$ the regular orbits dominate, leading to Poisson-like statistics. With $`\epsilon =\delta E/\mathrm{\Delta }`$ measuring the energy on the scale of the mean level spacing $`\mathrm{\Delta }`$, we expect random-matrix-like behaviour on energy scales $`\epsilon <\stackrel{~}{\epsilon }`$ and Poisson-like behaviour for $`\epsilon >\stackrel{~}{\epsilon }`$, where $`\stackrel{~}{\epsilon }A/\lambda _{\mathrm{dB}}\stackrel{~}{l}`$ and $`A`$ denotes the area of the billiard. Now $`\stackrel{~}{l}`$ can be estimated by $`\stackrel{~}{l}A/r_c`$, so that we find $`\stackrel{~}{\epsilon }r_c/\lambda _{\mathrm{dB}}`$. This ratio scales with energy as $`r_c/\lambda _{\mathrm{dB}}E^{(\alpha 2)/2\alpha }`$. Consequently, for $`\alpha <2`$ and increasing energy, $`\stackrel{~}{\epsilon }`$ tends to zero, while $`\stackrel{~}{\epsilon }\mathrm{}`$ for $`\alpha >2`$. On this basis we expect a sharp transition between Poisson and random matrix statistics at $`\alpha =2`$ in the limit of large energy $`E`$. Intuitively, the classically forbidden area becomes invisible to quantum mechanics for $`\alpha <2`$, while for $`\alpha >2`$ it becomes more and more sizeable. To see this explicitly, we apply Berry’s periodic orbit theory of bilinear spectral statistics to the SB with energy dependent disk radius $`r_c`$. The SB has two types of periodic orbits, those that never strike the disk and those that do. The former are marginally stable, occur in one-parameter families, and will be referred to as regular orbits, while the latter are unstable, isolated, and will be referred to as chaotic orbits. In the following we determine the contribution of the regular orbits to the spectral form factor $`K(\tau )`$. Denoting the unfolded density of states by $`d(\epsilon )`$, the form factor is related to the two-point correlation function $$R(s)=d\left(\epsilon +\frac{s}{2}\right)d\left(\epsilon \frac{s}{2}\right)_\epsilon 1$$ (2) by Fourier transformation, $`K(\tau )=_{\mathrm{}}^{\mathrm{}}dse^{2\pi is\tau }R(s)`$. One way to determine the contribution of the regular orbits to the density of states is to suitably modify the trace formula for the empty billiard. To avoid degeneracies in the spectrum, we use a rectangular instead of a quadratic billiard (sidelengths $`a`$ and $`b`$) and quasi-periodic boundary conditions for the wavefunction, i.e. $`\psi (x+a,y)=e^{i\varphi _x}\psi (x,y)`$ and analogously for the $`y`$-direction . The eigenvalues of the rectangle without the disk are $`E_{jk}=(2\pi j+\varphi _x)^2/a^2+(2\pi k+\varphi _y)^2/b^2`$ (here and below $`\mathrm{}^2/2m=1`$). Applying the Poisson summation formula to the density of states of the empty rectangle and replacing the resulting Bessel function by its asymptotic form leads to a representation as a sum over periodic orbit families. Each family is specified by two integers, $`m`$ and $`n`$ (positive or negative), denoting the number of traversals across the billiard in $`x`$\- and $`y`$-direction, respectively. Now including the disk obstructs the path of the orbits that violate the condition $`2l_{mn}r_c(E)<A`$, where $`A=ab`$ and $`l_{mn}=[(ma)^2+(nb)^2]^{1/2}`$ denotes the length of the orbits. The remaining families have to be weighted by the area they cover. The oscillatory part of the regular contribution to the unfolded density of states is then $$\stackrel{~}{d}_{\mathrm{reg}}(\epsilon )=\sqrt{\frac{2}{\pi }}\underset{(m,n)^{}}{}f_{mn}(E)\underset{j=1}{\overset{\mathrm{}}{}}\frac{\mathrm{cos}(jS_{mn}\frac{\pi }{4})}{(jl_{mn})^{1/2}E^{1/4}},$$ (3) where the first sum is over all primitive orbits and the second over repetitions. In eq. (3) we have introduced the actions $`S_{mn}=l_{mn}\sqrt{E}+m\varphi _x+n\varphi _y`$ and the orbit selection function $`f_{mn}(E)=(12r_c(E)l_{mn}/A)\theta (A2r_c(E)l_{mn})`$, and set the mean level density to its large energy limit, so that $`\epsilon =AE/4\pi `$. This formula is a simple generalization of the expression for the usual SB given in Ref. . Substituting Eq. (3) into (2) and retaining only the diagonal terms in the double sum over periodic orbits (the diagonal approximation is justified by the energy average and known to be exact for regular systems ) yields the contribution of the regular orbits to the form factor, $$K_{\mathrm{reg}}(\tau )=\underset{(m,n)^{}}{}f_{mn}^2(E)\underset{j=1}{\overset{\mathrm{}}{}}\frac{\delta (\tau jl_{mn}/A\sqrt{E})}{2\pi jl_{mn}\sqrt{E}}.$$ (4) This expression may be evaluated approximately by replacing the sums over primitive orbits and repetitions by a single sum over all orbits $`(m,n)`$. In the large $`E`$ limit, $$K_{\mathrm{reg}}(\tau )\left(1cE^{\frac{\alpha 2}{2\alpha }}\tau \right)^2\theta \left(1cE^{\frac{\alpha 2}{2\alpha }}\tau \right),$$ (5) with $`c=2R\lambda ^{1/\alpha }`$. Eq. (5) shows that the contribution of the regular orbits to the form factor tends to zero for $`\alpha >2`$, while for $`\alpha <2`$ the Poisson limit $`K(\tau )=1`$ is attained. The contribution of the chaotic orbits assures that for $`\alpha >2`$ GOE statistics of random matrix theory is reached. To see this explicitly would require to go beyond the diagonal approximation. Next, we verify the above theoretical prediction of a transition in the spectral statistics of the GSB numerically. For $`\alpha <2`$, the eigenvalues of the GSB may be calculated by diagonalizing the Hamiltonian in the eigenbasis of the empty billiard. However, for $`\alpha 2`$ the matrix elements diverge and it is natural to work in the eigenbasis of the $`1/r^\alpha `$-potential . In this regime we chose to apply the Korringa-Kohn-Rostoker method, as described in Ref. for the case of the ordinary SB. The difference with respect to Ref. is that in the present case the scattering phase shifts of the potential have to be determined by solving the radial Schrödinger equation numerically, while they can be expressed in terms of Bessel functions for the usual SB. This makes the present case computationally more demanding. Fig. 2 shows the cumulative spacing distribution $`N(s)=_0^sdxP(x)`$, with $`P(s)`$ denoting the nearest-neighbour spacing distribution, for $`\alpha =1`$ and 4 (we used $`\lambda =8`$ for the former, $`\lambda =0.01`$ for the latter, and in both cases $`R=1`$ with a billiard of area $`4\pi `$). Each curve is displayed with two reference curves, $`N(s)=1e^s`$ for Poisson statistics and the Wigner surmise $`N(s)=1\mathrm{exp}(\frac{\pi }{4}s^2)`$ for the GOE. We observe a clear movement of $`N(s)`$ towards the Poisson curve with increasing energy for $`\alpha =1`$ and towards the Wigner surmise for $`\alpha =4`$. Considering the relatively weak energy dependence of the spectral statistics, see Eq. (5), the limiting distributions are not expected to be reached within the available spectra. We verified that for $`\alpha =2`$ the spectral statistics is independent of energy in accordance with Eq. (5). We turn to the properties of the eigenfunctions of the GSB. To keep the discussion general, we consider the GSB in $`d2`$ dimensions. Following Ref. we map the Schrödinger equation for the GSB on a localization problem. Applying Levitov’s criterion of a diverging number of resonances we then show that $`\alpha =2`$ corresponds to a critical point associated with a delocalization transition. Using periodic boundary conditions and a billiard of unit volume, the Schrödinger equation in momentum representation takes the form $$E_𝐤c_𝐤+\underset{𝐤^{}𝐤}{}V_{𝐤𝐤^{}}c_𝐤^{}=Ec_𝐤,$$ (6) where $`𝐤`$ denotes a site in the reciprocal lattice, $`c_𝐤`$ the Fourier coefficients of the wavefunction, $`V_𝐤`$ those of the potential, and $`E_𝐤=𝐤^2+V_\mathrm{𝟎}`$. Eq. (6) may be interpreted as the Schrödinger equation of a particle on a lattice with on-site energies $`E_𝐤`$ and hopping amplitudes $`V_𝐤`$. Considering an eigenfunction with energy $`E`$, one finds that its Fourier coefficients are non-vanishing only in an energy shell $`E\delta E<E_𝐤<E+\delta E`$, where $`\delta E`$ is proportional to $`\lambda `$, which is assumed to be much smaller than $`E`$. Within this quasi-($`d`$$``$$`1`$)-dimensional shell, the on-site energies $`E_𝐤`$ are uniformly distributed quasi-random numbers. Since $`V_𝐤1/k^{d\alpha }`$ for large $`k`$, one may interpret Eq. (6) as an Anderson model with long-range power-law hopping within the energy shell . A delocalization transition occurs when the mean number of resonances per site diverges . A resonance is defined as a pair of sites $`𝐤`$,$`𝐤^{}`$ that fulfil the condition $`\|V_{𝐤𝐤^{}}\|>\|E_𝐤E_𝐤^{}\|`$. If two sites are in resonance, the eigenstates of the corresponding two by two eigenvalue problem have amplitudes $`c_𝐤`$,$`c_𝐤^{}`$ that are comparable in magnitude. Informally, they can then be considered as ‘linked’ and an eigenstate can spread along this link. If the mean number of resonances per site is infinite, there can be no localization, irrespective of the hopping strength. Noting that for $`V_{𝐤𝐤^{}}\delta E`$ the probability that the sites $`𝐤,𝐤^{}`$ are in resonance is simply $`\|V_{𝐤𝐤^{}}\|/\delta E`$, the mean number of resonances $`𝒩`$ with a fixed site $`𝐤`$ can be estimated by summing the probability over the energy shell, $$𝒩=\frac{1}{\delta E}\underset{𝐤^{}\delta E}{}\|V_{𝐤𝐤^{}}\|\frac{1}{\sqrt{E}}\underset{0}{\overset{\sqrt{E}}{}}dkk^{d2}V_k.$$ (7) The sum is estimated by an integration over a flat region of radius $`\sqrt{E}`$ (we used $`\mathrm{\Delta }k=\delta E/2\sqrt{E}`$ for the width of the energy shell and dropped an energy independent factor). Substituting the asymptotic form of $`V_k`$ for large $`k`$ into Eq. (7) we find that the mean number of resonances per site diverges in the limit of large energy when $`\alpha >2`$, independent of the dimension. In the case that was considered by Altshuler and Levitov the same consideration leads to a critical $`\alpha `$ of one. For $`\alpha `$ below the critical value, the eigenstates are localized with a power-law tail due to direct hopping, independent of $`\lambda `$ when $`d3`$ (because the effective lattice, i.e. the energy shell, is two-dimensional for $`d=3`$), and also for small $`\lambda `$ when $`d>3`$. In summary, we have shown that the GSB displays a sharp transition in the spectral statistics, which is caused by a competition between chaotic and regular orbits. It is accompanied by a dynamical delocalization transition of the eigenstates. The investigation helps to clarify the relation between disordered systems and quantum chaos. It is a pleasure to acknowledge useful discussions with E. B. Bogomolny, O. Bohigas, A. Mirlin, F. v. Oppen, C. Schmit, U. Smilansky and H. A. Weidenmüller.
no-problem/9907/hep-lat9907014.html	ar5iv	text	# Random Matrix Theory, Chiral Perturbation Theory, and Lattice Data (20 July 1999) ## Abstract Recently, the chiral logarithms predicted by quenched chiral perturbation theory have been extracted from lattice calculations of hadron masses. We argue that the deviations of lattice results from random matrix theory starting around the so-called Thouless energy can be understood in terms of chiral perturbation theory as well. Comparison of lattice data with chiral perturbation theory formulae allows us to compute the pion decay constant. We present results from a calculation for quenched SU(2) with Kogut–Susskind fermions at $`\beta =2.0`$ and $`2.2`$. PACS: 11.30.Rd; 11.15.Ha; 12.38.Gc; 05.45.Pq Keywords: chiral perturbation theory; random matrix theory; lattice gauge calculations; SU(2) gauge theory , , , , , , and For many observables, quenched chiral perturbation theory predicts contributions which are logarithmic in the quark mass . Their identification in lattice gauge results is a long standing problem. It seems that the latest numerical results on hadron masses in quenched lattice simulations allow for an approximate determination of these $`\mathrm{ln}(m)`$ contributions. (For an earlier attempt see .) The determination of these logarithms is an important test of chiral perturbation theory (chPT) which in turn plays a central role for the connection of low-energy hadron theory on one side and perturbative and lattice QCD on the other. In a completely independent development, it has been shown by several authors that chiral random matrix theory (chRMT) is able to reproduce quantitatively microscopic spectral properties of the Dirac operator obtained from QCD lattice data (see the reviews and Refs. ). Moreover, the limit up to which the microscopic spectral correlations can be described by random matrix theory (the analogue of the Thouless energy of statistical physics) was analyzed theoretically in and identified for quenched SU(2) (SU(3)) lattice calculations in (). It has also been shown that the results obtained in chRMT can be derived directly from field theory, providing a firm theoretical basis for the RMT approach. In this letter we want to study the Dirac spectrum beyond the Thouless energy using chPT (for a first investigation of this issue see ). Lattice Monte Carlo calculations inevitably involve a finite volume, so we have to consider chPT in a finite volume, too. Chiral RMT should be valid for masses $`m`$ up to the Thouless energy $`1/L^2`$ ($`L`$ = linear extent of the lattice), whereas the chPT formulae are supposed to work if $`m`$ is larger than a certain lower limit scaling like $`1/L^4`$. Thus we expect a domain of common applicability for sufficiently large lattices. The data we are going to analyze were obtained for the gauge group SU(2) in the quenched approximation with staggered fermions. They consist of complete spectra of the lattice Dirac operator. As the bare coupling in these data is relatively strong, we shall set up our (quenched) chPT in such a way that only those symmetries are taken into account which are exactly realized on the lattice. We will find that this eliminates terms $`\mathrm{ln}(m)`$ from the quenched chiral condensate, which are present if one starts from the continuum symmetries. In the following we shall use the chiral condensate and the scalar susceptibilities, so we first give their definitions. For a finite lattice and a non-vanishing valence-quark mass, the chiral condensate is given by $$\sigma _{\mathrm{lattice}}(m)=\frac{1}{N}\underset{k=1}{\overset{N}{}}\frac{1}{\mathrm{i}\lambda _k+m}.$$ (1) In the quenched theory, the connected susceptibility is given simply by $$\chi _{\mathrm{lattice}}^{\mathrm{conn}}(m)=\frac{}{m}\sigma (m)=\frac{1}{N}\underset{k=1}{\overset{N}{}}\frac{1}{(\mathrm{i}\lambda _k+m)^2}.$$ (2) The disconnected susceptibility is defined on the lattice by $$\chi _{\mathrm{lattice}}^{\mathrm{disc}}=\frac{1}{N}\underset{k,l=1}{\overset{N}{}}\frac{1}{(\mathrm{i}\lambda _k+m)(\mathrm{i}\lambda _l+m)}\frac{1}{N}\underset{k=1}{\overset{N}{}}\frac{1}{\mathrm{i}\lambda _k+m}^2.$$ (3) Here, $`N=L^4`$ denotes the number of lattice points, and the $`\lambda _k`$ are the Dirac eigenvalues. Note that each of the doubly degenerate eigenvalues (for gauge group SU(2)) counts only once. With the abbreviation $`F(x)=_0^1tI_0(tx)`$ the chRMT result for the chiral condensate reads $$\frac{\sigma _{\mathrm{RMT}}}{\mathrm{\Sigma }}=2u\left[I_0(2u)K_0(2u)+I_1(2u)K_1(2u)K_0(2u)F(2u)\right],$$ (4) where $`\mathrm{\Sigma }`$ denotes the absolute value of the chiral condensate for infinite volume and vanishing mass, and the rescaled mass parameter $`u`$ is given by $`u=m\mathrm{\Sigma }L^4`$. The functions $`I_\nu `$, $`K_\nu `$ are modified Bessel functions. For the connected susceptibility chRMT predicts $$\frac{\chi _{\mathrm{RMT}}^{\mathrm{conn}}}{N\mathrm{\Sigma }^2}=2K_1(2u)\left[2uF(2u)I_1(2u)\right].$$ (5) The chRMT result for the disconnected susceptibility is slightly more complicated , but can be simplified to read: $`{\displaystyle \frac{\chi _{\mathrm{RMT}}^{\mathrm{disc}}}{N\mathrm{\Sigma }^2}}`$ $`=`$ $`1K_0(2u)I_0(2u)+\left[K_0(2u)2uK_1(2u)\right]F(2u)`$ (6) $`\left\{2uK_0(2u)F(2u)2u\left[K_0(2u)I_0(2u)+K_1(2u)I_1(2u)\right]\right\}^2.`$ In Ref. it was demonstrated for $`\chi ^{\mathrm{disc}}`$ that chRMT describes the Monte Carlo data perfectly up to values of $`m`$ which scale like $`1/L^2`$ (the analogue of the Thouless energy). This is also true for $`\sigma `$ and $`\chi ^{\mathrm{conn}}`$. We want to understand the behavior of the data beyond the Thouless energy using (quenched) chPT. Chiral perturbation theory uses effective actions for the Goldstone bosons to describe the mass and volume dependence of, e.g., the chiral condensate and the susceptibilities. The effective action depends only on the symmetries and the corresponding breaking pattern. The predictions of chRMT are equivalent to the leading-order results from the so-called $`ϵ`$-expansion in chPT, where $`m=O(ϵ^4)`$ and $`1/L=O(ϵ)`$. The coefficients in this expansion are nontrivial functions of $`u=m\mathrm{\Sigma }L^4`$. On the other hand, standard chiral perturbation theory (the so-called $`p`$-expansion) sets $`m=O(p^2)`$, $`1/L=O(p)`$ and expands in powers of $`p`$ (see, e.g., ). Applying chPT in the present context we have to deal with two technical problems. Since the data that we want to analyze are taken in the quenched approximation we ought to work with the quenched version of chiral perturbation theory. Secondly, we work at rather strong coupling where we cannot expect the continuum symmetries to be already effectively restored. So we should consider only symmetries which are exact on the lattice (and the corresponding Goldstone bosons), a situation usually not dealt with in the literature. Below the Thouless energy the first problem becomes trivial when we use the chRMT results, because these depend explicitly on the number of flavors which we can set equal to zero. Furthermore they also depend on the topological charge $`\nu `$ (i.e. the number of zero modes of the Dirac operator), which suggests a solution to the second problem: At strong coupling, the Dirac operator of staggered fermions has no exact topological zero modes due to lattice artifacts, hence the lattice results should be compared with the case $`\nu =0`$ . Indeed, one finds very good agreement. Above the Thouless energy, in the regime of the $`p`$-expansion, we should use (partially) quenched chPT taking into account the pattern according to which chiral symmetry is spontaneously broken in the case of staggered fermions with gauge group SU(2). To the best of our knowledge, an analysis of this particular case has not been done before. In the following, we present our own, somewhat unorthodox, approach to this problem in order to avoid the intricacies of quenched chiral perturbation theory. Starting point is a partition function of the form: $$\begin{array}{c}Z(m_v,m_s,L)\hfill \\ \mathrm{exp}\left[VT(m_v,m_s)\frac{1}{2}\underset{Q}{}\left\{K_Q\underset{p}{}\mathrm{ln}\left(\widehat{p}^2+m_Q^2(m_v,m_s)\right)\right\}\right].\hfill \end{array}$$ (7) $`T(m_v,m_s)`$ is the classical (or saddle-point) contribution to the free energy, which we will assume is a smooth function of the quark masses, and independent of lattice volume. The double sum represents the 1-loop contribution coming from light composite bosons. The sum runs over the allowed momenta $`p`$ ($`p_\mu =2\pi n_\mu /L`$ with integer $`n_\mu `$) and over particle type $`Q`$. $`K_Q`$ is the multiplicity of the particles of type $`Q`$. On the lattice we will use the expression $$\widehat{p}^22\underset{\mu }{}(1\mathrm{cos}p_\mu )$$ (8) for the boson kinetic term, where $`p_\mu `$ is expressed in lattice units. We have introduced two quark masses, valence quarks with mass $`m_v`$, and sea quarks with mass $`m_s`$. We have $`N_v`$ ‘generations’ of the quarks with mass $`m_v`$, and $`N_s`$ generations of the quarks with mass $`m_s`$. In the continuum limit, each staggered ‘generation’ yields four fermion flavors. We will take the limit $`N_v0`$, so that the valence quarks are quenched. To use Eq. (7) we need the masses and multiplicities of all the light bosons. For mesons made of two different quark flavors $`\overline{q}_iq_j`$ we use the expression $$m^2=A(m_i+m_j)/2.$$ (9) This applies to $`(N_s+N_v)(N_s+N_v1)`$ bound states. For the remaining $`N_s+N_v`$ ‘flavor-diagonal’ mesons we should also consider annihilation (see Fig. 1). However, with staggered fermions there is an anomaly-free U(1) symmetry, which would cause the amplitude for $`\overline{q}_iq_i\overline{q}_jq_j`$ to vanish if either $`m_i=0`$ or $`m_j=0`$. Therefore we expect the annihilation amplitude to be proportional to $`m_im_j`$ for small quark masses. In the final formulae this leads to contributions which possess only a mild $`m`$–dependence (there is at least one more power of $`m`$ than in the leading contribution). Hence these are hard to distinguish from the smooth background and cannot be fitted reliably. To leading order, it is consistent to neglect the annihilation terms, and we end up with $`N_v`$ ($`N_s`$) states with $`m^2=Am_v`$ ($`m^2=Am_s`$) in the ‘flavor-diagonal’ sector. (Note that for the gauge group SU(3) there are special cases where the annihilation terms contribute to leading order and can therefore not be neglected.) For the SU(2) gauge group the symmetry when all quarks are massless is $`\mathrm{U}(2N_v+2N_s)`$, which spontaneously breaks to $`\mathrm{O}(2N_v+2N_s)`$ . This is further broken to $`\mathrm{O}(2N_v)\times \mathrm{O}(2N_s)`$ if the valence and sea quarks are given different bare masses. The extra symmetry present when the gauge group is SU(2) transforms mesons into baryons ($`q_iq_j`$ and $`\overline{q}_i\overline{q}_j`$ states), which thus have the same mass as the mesons. These states will have mass-squared given by the same formula as the different-flavor mesons, i.e., $`m^2=A(m_i+m_j)/2`$. Therefore we have to include more light bosons in our partition function leading to the spectrum given in Table 1. In SU(2) we just sum over the non-degenerate eigenvalues (according to the RMT conventions), so our definition of the chiral condensate $`\sigma `$ is $$\sigma (m_v,m_s,L)=\underset{N_v0}{lim}\left(\frac{1}{V}\frac{1}{2N_v}\frac{1}{Z}\frac{Z}{m_v}\right).$$ (10) If we substitute the multiplicities of Table 1 into Eq. (7) we get $`\sigma (m_v,m_s,L)`$ $`=`$ $`S+C_cm_v+C_d2N_sm_s{\displaystyle \frac{A}{4L^4}}{\displaystyle \underset{p}{}}{\displaystyle \frac{1}{\widehat{p}^2+Am_v}}`$ (11) $`{\displaystyle \frac{AN_s}{L^4}}{\displaystyle \underset{p}{}}{\displaystyle \frac{1}{2\widehat{p}^2+A(m_v+m_s)}},`$ where the first three terms parametrize the classical background contribution. Differentiating $`\sigma `$ to give the susceptibilities we find for identical valence and sea quark masses, $`m_v=m_s=m`$, $`\chi ^{\mathrm{conn}}(m,m,L)`$ $`=`$ $`C_c+{\displaystyle \frac{A^2(1+N_s)}{4L^4}}{\displaystyle \underset{p}{}}{\displaystyle \frac{1}{(\widehat{p}^2+Am)^2}},`$ (12) $`\chi ^{\mathrm{disc}}(m,m,L)`$ $`=`$ $`C_d+{\displaystyle \frac{A^2}{8L^4}}{\displaystyle \underset{p}{}}{\displaystyle \frac{1}{(\widehat{p}^2+Am)^2}}.`$ (13) Sending finally also $`N_s`$ to zero we arrive at formulae appropriate for comparison with our quenched Monte Carlo data. Note that the leading behavior of these chPT formulae for $`m0`$ coincides with the leading term for $`m\mathrm{}`$ of the chRMT formulae. This is in accordance with the existence of a mass range where both theories apply. In the thermodynamic limit we obtain from (11), (12), and (13) chiral logarithms: $`\sigma (m)`$ $`=`$ $`\left(S{\displaystyle \frac{1}{4}}Aa_0\right){\displaystyle \frac{A^2}{64\pi ^2}}m\mathrm{ln}(Am)+\left(C_c{\displaystyle \frac{1}{4}}A^2a_1\right)m`$ (14) $`+{\displaystyle \frac{A^3}{512\pi ^2}}m^2\mathrm{ln}(Am){\displaystyle \frac{1}{4}}A^3a_2m^2+\mathrm{},`$ $`\chi ^{\mathrm{conn}}(m)`$ $`=`$ $`{\displaystyle \frac{A^2}{64\pi ^2}}\mathrm{ln}(Am)+\left(C_c{\displaystyle \frac{1}{4}}A^2a_1{\displaystyle \frac{A^2}{64\pi ^2}}\right)`$ (15) $`+{\displaystyle \frac{A^3}{256\pi ^2}}m\mathrm{ln}(Am)+A^3\left({\displaystyle \frac{1}{512\pi ^2}}{\displaystyle \frac{1}{2}}a_2\right)m+\mathrm{},`$ $`\chi ^{\mathrm{disc}}(m)`$ $`=`$ $`{\displaystyle \frac{A^2}{128\pi ^2}}\mathrm{ln}(Am)+\left(C_d{\displaystyle \frac{1}{8}}A^2a_1{\displaystyle \frac{A^2}{128\pi ^2}}\right)`$ (16) $`+{\displaystyle \frac{A^3}{512\pi ^2}}m\mathrm{ln}(Am)+A^3\left({\displaystyle \frac{1}{1024\pi ^2}}{\displaystyle \frac{1}{4}}a_2\right)m+\mathrm{}`$ for $`N_s=0`$ and $`m_v=m_s=m`$. The numerical constants $`a_0`$, $`a_1`$, and $`a_2`$ take the values $`a_0=0.1549`$, $`a_1=0.03035`$, $`a_2=0.002776`$. To highlight the special properties of the SU(2) gauge group with staggered fermions, we will now compare with the conventional continuum calculation with gauge group SU(3). In this case there are no light ‘Goldstone baryons’, so we only need to consider meson states. The other important difference is that in the continuum the chiral U(1) has an anomaly. This means that the annihilation term in the meson mass matrix does not need to vanish in the chiral limit. This leads us to the following ansatz for the mass-squared matrix, $`M^c`$, for the flavor-diagonal mesons: $$M^c=A\mathrm{diag}(m_v,\mathrm{},m_v,m_s,\mathrm{},m_s)+m_0^2\left(\begin{array}{ccc}1& \mathrm{}& 1\\ \mathrm{}& \mathrm{}& \mathrm{}\\ 1& \mathrm{}& 1\end{array}\right)$$ (17) with $`N_v`$ ($`N_s`$) entries $`Am_v`$ ($`Am_s`$) on the diagonal of the first part. Here $`m_0`$ is a constant with the dimensions of mass, which does not vanish in the chiral limit. Besides $`N_v1`$ eigenvalues equal to $`Am_v`$ and $`N_s1`$ eigenvalues equal to $`Am_s`$ this mass matrix has two non-trivial eigenvalues. The important thing to note is that one of them does not vanish in the chiral limit but goes to the value $`(N_v+N_s)m_0^2`$. This is the state corresponding to the continuum $`\eta ^{}`$. Using the continuum multiplicities and masses in our formula for $`Z`$ and omitting the smooth background for simplicity leads to $$\sigma ^c(m,m,L)=\frac{A}{2L^4}\underset{p}{}\frac{m_0^2(1N_s^2)N_s(\widehat{p}^2+Am)}{(\widehat{p}^2+Am)(\widehat{p}^2+N_sm_0^2+Am)}$$ (18) (the case of equal valence and sea quark masses is sufficient to illustrate the main differences between staggered and continuum fermions). For $`N_s0`$ the $`L\mathrm{}`$ limit of this $`\sigma `$ is $`\sigma ^c(m,m,\mathrm{})`$ $`=`$ $`{\displaystyle \frac{A}{2N_s}}{\displaystyle \frac{Am}{16\pi ^2}}\mathrm{ln}\left[{\displaystyle \frac{Am}{N_sm_0^2+Am}}\right]`$ (19) $`{\displaystyle \frac{AN_s}{2}}\left(a_0+{\displaystyle \frac{Am}{16\pi ^2}}\mathrm{ln}(Am)+a_1Am\right)`$ $`{\displaystyle \frac{A}{2}}\left({\displaystyle \frac{m_0^2}{16\pi ^2}}\mathrm{ln}(N_sm_0^2+Am)+a_1m_0^2\right)+\mathrm{}.`$ At small $`m`$ the most severe singularity is of the form $`m\mathrm{ln}m`$. However in the quenched limit the behaviour is different. Then Eq. (18) reduces to $$\sigma ^c(m,m,L)=\frac{A}{2L^4}\underset{p}{}\frac{m_0^2}{(\widehat{p}^2+Am)^2},$$ (20) which is more singular at small $`m`$. The thermodynamic limit of this expression is $$\sigma ^c(m,m,\mathrm{})=\frac{Am_0^2}{2}\left[\frac{1}{16\pi ^2}\left(\mathrm{ln}(Am)+1\right)a_1+\mathrm{}\right].$$ (21) Now we have a more severe singularity ($`\mathrm{ln}m`$) in the small mass limit. This is the well-known quenched chiral logarithm. We do not get this logarithm in our regime because the relevant U(1) symmetry is not anomalous. We have confronted our chPT formulae with lattice Monte Carlo data. We used staggered fermions in the quenched approximation and two values of the coupling strength $`\beta `$, $`\beta =2.0`$ and $`\beta =2.2`$. The lattice sizes and numbers of configurations are given in Table 2. The chPT formulae are able to describe the deviations from the RMT results beyond the Thouless energy very well. Examples of joint fits of $`\chi ^{\mathrm{conn}}`$ and $`\chi ^{\mathrm{disc}}`$ with (12) and (13), respectively, are presented in Fig. 2. It seems that for the connected susceptibility our lattices are not large enough to exhibit clearly a domain where both chRMT and chPT (to the order considered) are applicable, although there is a trend towards the formation of such a window with increasing $`L`$. Joint fits of $`\chi ^{\mathrm{conn}}`$ and $`\chi ^{\mathrm{disc}}`$ lead to the values for $`A`$ given in Table 2. The error has been estimated from the variation of the results under changes of the fit interval. This is a somewhat subjective procedure of determining the error and goes beyond the purely statistical contribution. However we consider it to be more realistic than using the MINUIT errors. The smaller errors on the larger lattices thus reflect the increasing reliability of our numbers. Due to the Gell-Mann–Oakes–Renner relation the parameter $`A`$ is related to the chiral condensate $`\mathrm{\Sigma }`$ (for infinite volume and vanishing mass) and the pion decay constant $`f_\pi `$ by $$A=\frac{2\mathrm{\Sigma }}{f_\pi ^2}.$$ (22) The chiral condensate $`\mathrm{\Sigma }`$ can be determined independently, e.g. from the mean value of the smallest eigenvalue according to the RMT formula $$\mathrm{\Sigma }=\frac{2.06636\mathrm{}}{L^4\lambda _{\mathrm{min}}}.$$ (23) The corresponding numbers together with the resulting values for $`f_\pi `$ are also given in Table 2. Note that except for the smallest lattices the results are nearly independent of $`L`$ demonstrating the reliability of our procedure. In conclusion, we have verified predictions of chiral random matrix theory and (quenched) chiral perturbation theory by means of lattice Monte Carlo data. In particular, we have investigated finite-volume effects and found a domain of common validity of both theories. It was crucial to base the analysis on the lattice symmetries and not on the continuum symmetries. We were also able to determine the pion decay constant. The extension of the present analysis to the case of gauge group SU(3) will be the subject of an upcoming publication. It is a pleasure to thank I. Zahed for an early discussion and M. Golterman, N. Kaiser, and J.J.M. Verbaarschot for helpful comments. This work was supported in part by DFG. SM, AS, and TW thank the MPI für Kernphysik, Heidelberg, for hospitality and support. The numerical simulations were performed on a CRAY T90 at the Forschungszentrum Jülich, on a CRAY T3E at the HLRS Stuttgart, and on a Cray T90 at the Leibniz-Rechenzentrum München.
no-problem/9907/hep-ph9907436.html	ar5iv	text	# 1 The results of the 125×10⁶ statistics samples (weighted events) (except for 𝐵⁢𝑜⁢𝑟⁢𝑛, where the sample is 540×10⁶ of such events) from YFSWW3-1.12 for the average value of 𝑀_𝑊 as computed with the levels of radiative corrections as indicated for both bare and calorimetric treatments of the final lepton. See the text for more details. The role of the final-state radiative (FSR) effects in the processes $`e^+e^{}W^+W^{}+n(\gamma )4f+n(\gamma )`$ at and beyond LEP2 energies is of considerable interest for the LEP2 and NLC physics programs . In this paper, we evaluate for the first time the possible interplay between the exact $`𝒪(\alpha )`$ electroweak (EW) corrections and the leading-logarithm (LL) final-state radiative effects for these processes when the $`n(\gamma )`$ radiation is realized according to the amplitude-based Monte Carlo event generator techniques described in Refs. , wherein infrared singularities are cancelled to all orders in $`\alpha `$ by using the extension to spin 1 charged particles of the theory of Yennie, Frautschi and Suura for QED . The final-state radiative effects are realized in the LL approximation using the calculation of the program PHOTOS (Ref. ) in which a non-radiative final-state process is used to generate up to two photons in the corresponding radiative process by iterating the structure function evolution equation<sup>1</sup><sup>1</sup>1To be precise an ansatz is provided, which reproduces the LL terms. It includes transverse degrees of freedom for the photon 4-momentum, assures coverage of the full phase space and rules of energy-momentum conservation. The photons’ angular distribution is chosen to reproduce exactly the one of the soft photon limit. See Ref. for more details. for QED . The exact $`𝒪(\alpha )`$ YFS exponentiated final-state $`W`$ decay radiative effects will be published elsewhere . In this connection, we note that we expect the non-leading $`𝒪(\alpha )`$ and higher order ($`𝒪(\alpha ^n),n2`$) final-state radiative effects to be small, $`1\%`$ in the peak reduction effect for example, even for a “bare trigger” acceptance for the outgoing final charged particles. This has been found by the authors of Ref. , who analysed the effects of final-state radiation in $`Z`$ decay in the naive exponentiated (exact and LL) $`𝒪(\alpha )`$ approximation and who estimated the corresponding size of the analogous effects in $`W`$ decay, such as $`14\%`$ for the total peak reduction effect. Indeed, more recently, the authors of Ref. have made an independent cross check on their estimates of the FSR line-shape effects for $`e^+e^{}W^+W^{}4f`$ in Ref. . There they present an exact $`𝒪(\alpha )`$ calculation of the process in the double-pole approximation (DPA), wherein they retain in the pole expansion of the complete $`e^+e^{}4f`$ amplitude, only the terms containing the double pole in the S-matrix at the complex mass squared, $`M^2=M_W^2iM_W\mathrm{\Gamma }_W`$, where $`M_W,\mathrm{\Gamma }_W`$ are the respective mass and width of the $`W`$ boson, and where in the residues of the respective double poles they project the respective $`𝒪(\alpha )`$ corrections to an appropriate on-shell point. Henceforth, we refer to the on-shell residue projected DPA as to the leading pole approximation (LPA), with more general applications in mind: for example, in a triply resonant process, the LPA would correspond to the triple pole terms in the respective S-matrix element with the residues projected to an appropriate on-shell point. In this gauge-invariant calculation, these authors find that the FSR peak reduction effect is $`14.4\%`$ for $`W^{+()}e^{+()}\nu _e(\overline{\nu }_e)`$, to be compared with their estimate of $`14\%`$ in Ref. . We will compare our results with those in Refs. . We emphasise that our work differs from the work of Ref. in that we include the exact EW $`𝒪(\alpha )`$ corrections with YFS exponentiation in the production process and we actually calculate the effects of the FSR in the $`W`$-pair production and decay process at LEP2/NLC energies; in Ref. , only the process $`\nu _\mu \overline{\nu }_\mu ZZe^+e^{}+\nu _\tau \overline{\nu }_\tau `$ is actually calculated, and a heuristic argument is used to estimate the corresponding results for final-state $`W`$-decay radiation. Thus, our calculations will also be a comment on the accuracy of these heuristic arguments in the presence of the YFS exponentiated exact $`𝒪(\alpha )`$ corrections to the $`W`$-pair production process. Our work differs from that in Ref. in that we include the YFS exponentiation of the exact $`𝒪(\alpha )`$ production process in the $`W`$-pair intermediate state and the $`𝒪(\alpha )^2`$ LL FSR whereas, in Ref. , the exact $`𝒪(\alpha )`$ correction to the production and decay processes for the $`W`$-pair in the leading pole approximation is calculated without exponentiation. The leading pole approximation treatment of the attendant non-factorizable corrections in Refs. is also retained. The latter non-factorizable corrections have been shown to be small and, as we explain in Ref. , when one works up to but not including $`𝒪(\frac{\alpha }{\pi }\frac{\mathrm{\Gamma }_W}{M_W})`$ as we do, such effects may be dropped; which we do. Thus, although we start our calculation in Ref. in the fermion-loop scheme , when we focus on the $`𝒪(\alpha )`$ EW correction, we go to the leading pole part of the respective production amplitude. We also make the approximation of using on-shell residues for this double pole part, with which we then approximate the corresponding $`𝒪(\alpha )`$ EW correction. In our Monte Carlo event generator approach, we stress that the full off-shell phase space is always retained here. We improve our result by using the complete on-shell residues for EW corrections rather than their on-shell fermion-loop scheme representatives. Indeed, for the QED bremsstrahlung correction we stress that, since the real photon has $`k^2=0`$, the corresponding running charge is the usual one. It can thus be shown that, in $`𝒪(\alpha )`$, bremsstrahlung residues of the on-shell fermion-loop scheme are equivalent to those in the LPA; in both cases all infrared singularities are properly cancelled and not only the QED gauge invariance is preserved but also the full $`SU_{2L}\times U_1`$ gauge invariance . For this reason, in order $`𝒪(\alpha )`$, in our final result, any reference to the fermion-loop scheme is purely pedagogical. What we arrive at is precisely the LPA, with full on-shell residues for the respective double pole approximation. Indeed, as the YFS expansion is not generally familiar, if one looks at Eqs. (1) and (2) in Ref. , which give the on-shell $`𝒪(\alpha )`$ contributions to the YFS residuals $`\overline{\beta }_0`$ and $`\overline{\beta }_1`$, respectively, for the production process in the LPA, one may think that the lowest-order contribution to $`\overline{\beta }_0`$, $`\overline{\beta }_0^{(0)}`$ in the notation of Ref. , is not required either to be evaluated at the corresponding on-shell point as well. However, the right-hand side of Eq. (2) in Ref. , for example, involves the subtraction, from the corresponding on-shell $`𝒪(\alpha )`$ bremsstrahlung cross section, of the product of the YFS real emission infrared function $`\stackrel{~}{S}`$ by the on-shell lowest order Born cross section; we need to stress that the YFS theory then forces the contribution to $`\overline{\beta }_0`$ corresponding to this respective lowest-order Born cross section, $`\overline{\beta }_0^{(0)}`$, to be evaluated at the on-shell point as well. Thus, according to the YFS theory, Eqs. (1) and (2) in Ref. are entirely equivalent to results in Refs. for the production process, for the contributions up to and including terms $`𝒪(\alpha )`$. As can also be seen from the results in Refs. , these approximations are valid up to but not including $`𝒪(\frac{\alpha }{\pi }\frac{\mathrm{\Gamma }_W}{M_W})`$. We then apply the YFS Monte Carlo methods of two of us (S.J. and B.F.L.W.) , as extended to spin 1 particles in Ref. , to arrive at the respective exact $`𝒪(\alpha )_{prod}`$ YFS exponentiated results realized in YFSWW3-1.11. Hence, we stress that, as far as the $`𝒪(\alpha )`$ correction to the production process under study is concerned, the results in Refs. should be equivalent, in view of the many cross checks carried out by the authors of Refs. on the two corresponding electroweak on-shell calculations used therein. More precisely, starting from the calculations in the program YFSWW3-1.11 in Ref. , which feature the exact $`𝒪(\alpha )_{prod}`$ YFS exponentiated results for the process $`e^+e^{}W^+W^{}+n(\gamma )4f+n(\gamma )`$, we have interfaced the outgoing final state to the program PHOTOS . The latter uses the structure function evolution equation for QED to generate up to two final-state decay photons for each $`W`$ according to the respective LL probabilities to radiate; here the corresponding angular distributions of the decay photons are all generated in accordance with this LL approximation as described in Ref. . The net probability of the respective event is unchanged, i.e. the normalization of YFSWW3-1.11 is unaffected by this interface, which will be described in more detail elsewhere . We refer to the version of YFSWW3 that contains this final-state radiative interface to PHOTOS as YFSWW3-1.12 and it is available from the authors . In what follows, we present some sample Monte Carlo data from YFSWW3-1.12 to look into the possible role of FSR in the presence of the $`𝒪(\alpha )`$ EW corrections. For definiteness, we focus here on the current LEP2 CMS energy of $`190`$ GeV and on the SM couplings. The complete discussion of both LEP2 and NLC energies with the illustration of anomalous couplings will appear elsewhere . Specifically, in Figs. 1–8, we show the results obtained with YFSWW3-1.12 on the processes $`e^+e^{}W^+W^{}+n(\gamma )\overline{c}s+\mathrm{}\overline{\nu }_{\mathrm{}}`$, $`\mathrm{}=e,\mu `$, for the cosine of the $`W`$ production angle distribution in the CM (LAB) system, for the $`W`$ mass distribution, with both “bare” and “calorimetric” definitions of that mass, for the CMS lepton final energy distribution, for both calorimetric and bare definitions of that energy, and for the corresponding distributions of the cosine of the lepton decay angle in the $`W`$ rest frame. We note the following properties of these results. First, concerning the $`W`$ mass distributions in Figs. 1 and 5, we see that the respective average values of $`M_W`$ are as given in Table 1. There, EW-ex denotes the exact $`𝒪(\alpha )_{prod}`$ calculation of EW corrections ; EW-ap denotes the approximate treatment of these EW corrections as given in Ref. ; No EW denotes that the EW corrections other than the ones coming from LL ($`𝒪(\alpha ^2)`$) initial-state radiation are turned off. The calorimetric results are all closer to their respective NO FSR analogues than are the bare results, as expected. The effects of the FSR for the muon case are all respectively smaller than the corresponding results for the electron case, again as expected because of the smaller radiation probability for the muon. The size of the shift of $`M_W`$ is generally consistent with the discussion in Ref. , which deals with primarily the line shape (peak position and height); in detail we see that, in the presence of the FSR, at the level of our statistical errors, for an average quantity such as $`M_W`$, all three calculations in the table are sufficient, as expected. With regard to the guesstimates made in Ref. concerning the peak reduction and the peak position shift, we see from the $`BARE`$ curves in Fig. 1 that our result of $`13.5\%`$ for the peak reduction in the $`e^{}`$ case (comparing the EW-ex curves with and without FSR) is in good agreement with the $`14\%`$ guesstimate of Ref. and with the $`14.4\%`$ found in the recent $`𝒪(\alpha )`$ on-shell LPA results in Ref. . The $`57`$ MeV guesstimated in Ref. for the corresponding peak position shift in the $`e^{}`$ case was recently updated to $`77`$ MeV in Ref. ; for the $`\mu `$ case, the updated expectation from Ref. for the peak position shift is $`39`$ MeV. For completeness, we note that the size of the peak reduction effect in the $`\mu `$ case has been found to be $`8\%`$ in Ref. whereas in Fig. 5 we find $`7.6\%`$, again showing good agreement between our results and those in Ref. . Indeed, to compare our results for the peak position shift with those just cited from Ref. , we have performed Breit–Wigner fits to our line shapes in Figs. 1 and 5 with the values, both fixed and floating, of the $`W`$ width. The results of our fits are shown in Table 2. For comparison, the fits are done for two different mass intervals, from $`78`$ GeV to $`82`$ GeV, and from $`76`$ GeV to $`84`$ GeV, to illustrate the role of the wings of the resonance in the fits. From these results we find that the BARE peak position shifts are estimated using the narrow fit range as $`80.16880.240=72`$ MeV and $`80.19980.241=42`$ MeV for the $`e`$ and $`\mu `$ cases, respectively. We also computed the shift in the average invariant mass $`M_W`$ of the $`W`$ in the narrow range from $`78`$ GeV to $`82`$ GeV as another estimate of the peak position shift for the BARE trigger and we found $`81.5\pm 1.4`$ MeV and $`43.9\pm 0.9`$ MeV for the $`e`$ and $`\mu `$ cases, respectively. Thus, both sets of estimators of the peak position shifts are in reasonable agreement with the results given in Ref. <sup>2</sup><sup>2</sup>2 The fit mass shift and the peak position shift approach one another as the fit range approaches a zero size interval around the peak; a similar remark applies to the shift in the average mass relative to the range over which it is taken around the peak.; in this connection, we recall the slight difference in beam energy between our studies ($`95`$ GeV) and those in Ref. ($`92`$ GeV). Moreover, we see in Table 2 the same pattern of results as we see in Table 1: the FSR effects for the $`e`$ case are more pronounced than those for the $`\mu `$ case; the calorimetric acceptance reduces the size of the FSR effects; the results are not very sensitive to the EW correction to the production process. If we compare the predictions with and without FSR for the EW-ex and no EW cases we get a measure of the modulation of the FSR on the EW correction. From the curves in our Figs. 1 and 5 and the respective plots of the $`\delta _{RAD}`$ as defined in the figures we see that this modulation is as expected. Concerning the cosine of the $`W`$ production angle distributions, we see the interplay of the exact EW corrections on the one hand and the FSR on the other. Further, we see that the approximate EW corrections of Ref. , while a definite improvement over the no EW corrections at all, are not sufficient to describe this interplay at the level of $`0.5`$$`1.0\%`$. Similar remarks hold for the lepton energy distribution in the CM system, although the corresponding insufficiency is reduced to the level of $`0.3\%`$ for the $`BARE`$ case, for example for electrons. Concerning the distributions of the cosine of the lepton decay angle in the $`W`$ rest frame, we again see the importance of including both the EW corrections and the FSR in Figs. 4 and 8, for the electron and the muon, respectively. In all cases, the results for the muon, particularly the $`BARE`$ results, are less affected by the FSR than are the corresponding results for the electron, as expected. We stress that our results in Figs. 1–8 are generally consistent with those in Ref. keeping in mind that we treat the $`𝒪(\alpha ^2)`$ LL FSR and the YFS exponentiated on-shell exact $`𝒪(\alpha )_{prod}`$ production process, whereas Ref. treats only $`𝒪(\alpha )`$ corrections in our LPA, in which only on-shell residues are used. Indeed, in addition to the agreements already cited, we call the reader’s attention to the normalization correction in Fig. 9 of Ref. : at the CMS energy of $`\sqrt{s}=190`$ GeV, it is $`11\%`$, in very good agreement with our result in Ref. , which is $`(1+\delta _{prod})(\rho _w)^2111.1\%`$, for the latter result, we have used Table 2 in Ref. for the relative correction $`\delta _{prod}=9.9\%`$ to the production process, and the result in Ref. for the $`𝒪(\alpha )`$ correction to the leptonic partial width $`\rho _w10.686\%`$. In addition, we can note that, for the case of the $`\tau \overline{\nu }_\tau `$ decay channel, our results are also consistent with those in Ref. for the peak position shift and peak reduction effects. In view of our higher-order corrections, we find quite reasonable all the agreements noted here. A more detailed discussion of such comparisons will appear . We stress that we have arrived at our results through a MC event generator realization of our calculation, in which realistic, finite $`p_T`$, $`n(\gamma )`$ radiation is incorporated in the production process on an event-by-event basis, whereas the results in Ref. are all semi-analytical. This enhances the significance of the general agreement of our results where they do overlap. The issue of whether the calorimetric results are more realistic than the bare ones appears to depend on whether one is talking about the muon or the electron<sup>3</sup><sup>3</sup>3T. Kawamoto, private communication, 1998.. For the electron, it is very difficult to separate the soft photons with energy $`\mathrm{\Gamma }_W`$ that are responsible for the FSR effects of the $`W`$ line shape as discussed already in Refs. ; they are just a part of the electromagnetic calorimeter response in general, which is used to measure the electron energy. For the muon, the energy is usually measured by a muon chamber in which, in general, these soft photons are not present. Thus, for the electron, our calorimetric results are more realistic; for the muon, it is the other way around. In either case, we see that precision $`W`$-pair production and decay studies need to take the interplay between the FSR and the EW corrections into account so as to obtain the most precise tests of the SM; our calculations in YFSWW3-1.12 offer an avenue to achieve that goal. Acknowledgements Two of us (S.J. and B.F.L.W.) acknowledge the kind hospitality of Prof. A. De Rújula and the CERN Theory Division while this work was being completed. Two of us (B.F.L.W. and W.P.) acknowledge the support of Prof. D. Schlatter and the ALEPH Collaboration in the final stages of this work. One of us (Z.W.) acknowledges the support of the L3 group of ETH Zurich during the time this work was performed.
no-problem/9907/cond-mat9907478.html	ar5iv	text	# Scaling and universality of AC conduction in disordered solids ## Abstract Recent scaling results for the AC conductivity of ionic glasses by Roling et al. \[Phys. Rev. Lett. 78, 2160 (1997)\] and Sidebottom \[Phys. Rev. Lett. 82, 3653 (1999)\] are discussed. It is shown that Sidebottom’s version of scaling is completely general. A new analytical approximation to the universal AC conductivity of hopping in the extreme disorder limit, the “diffusion cluster approximation,” is presented and compared to simulations and experiments. Disordered solids have AC electrical properties remarkably in common. These solids, in fact, have so similar frequency-dependent conductivity $`\sigma (\omega )`$ that ionic conduction cannot be distinguished from electronic. Even the temperature-dependence of $`\sigma (\omega )`$ is “quasi-universal.” The class of disordered solids with quasi-universal AC behavior is large, including polycrystalline and amorphous semiconductors, organic semiconductors, ionic conductive glasses, ionic melts, non-stoichiometric crystals, ionic or electronically conducting polymers, metal cluster compounds, transition metal oxides, and macroscopic mixtures of differently conducting phases like organic-inorganic composites, etc. Each class contains hundreds of different solids and there is a huge literature on their AC conductivities. It is possible for almost all disordered solids to scale AC data at different temperatures into one single curve. This so-called master curve gives the dimensionless AC conductivity $`\stackrel{~}{\sigma }\sigma (\omega )/\sigma (0)`$ as function of a scaled dimensionless frequency $`\stackrel{~}{\omega }`$. The existence of a master curve is sometimes referred to as the “time-temperature superposition principle” (TTSP). The master curves of different solids - while not identical - are surprisingly similar. This quasi-universality was recognized gradually in the 1970’s . The common AC features of disordered solids are the following : At low frequencies conductivity is frequency-independent. Around the dielectric loss peak frequency $`\omega _m`$ AC conduction sets in, and for $`\omega \omega _m`$ $`\sigma (\omega )`$ is close to a frequency power-law. The exponent is between 0.7 and 1.0 . As temperature is lowered the exponent goes to 1.0. In a log-log plot the AC conductivity is much less temperature-dependent than the DC conductivity. A final ubiquitous observation is the Barton-Nakajima-Namikawa (BNN) relation connecting dielectric loss strength $`\mathrm{\Delta }ϵ`$, dielectric loss peak frequency, and DC conductivity - $`\sigma (0)=p\mathrm{\Delta }ϵϵ_0\omega _m`$ \- where $`p`$ is a numerical constant close to one. To construct the master-curve frequency must be scaled by $`\omega _m`$. Because the dielectric loss strength is only weakly temperature-dependent while $`\sigma (0)`$ and $`\omega _m`$ are both Arrhenius, the BNN-relation implies $`\omega _m\sigma (0)`$. Thus, the existence of a master curve is conveniently summarized into $$\stackrel{~}{\sigma }=F\left(\frac{C}{\sigma (0)}\omega \right)F(\stackrel{~}{\omega }),$$ (1) where $`C`$ may depend on variables like charge carrier concentration $`n`$, temperature $`T`$, high frequency dielectric constant, etc. Recently, there has been renewed interest in scaling and universality of AC data for ionic glasses (excellent reviews of glass ionic conduction have been given by Nowick and coworkers and Ngai ). In 1997 Roling, Happe, Funke, and Ingram showed that $`Cn/T`$ for sodium borate glasses . This year, however, Sidebottom showed that in general scaling is not achieved by $`Cn/T`$ \- instead three quite different ionic conductive systems all obey the following scaling relation , $$\stackrel{~}{\sigma }=F\left(\frac{ϵ_0\mathrm{\Delta }ϵ}{\sigma (0)}\omega \right)F(\stackrel{~}{\omega }).$$ (2) The purpose of this paper is three-fold. First, we put the recent scaling results into a historical perspective. Second, we show that Eq. (2) is generally valid. Finally, Sidebottom’s experimental master curves are compared to two models, the symmetric hopping model and the macroscopic model. In this connection a new analytical approximation to the universal AC hopping conductivity is presented. First, the historical perspective. In a series of papers towards the end of the 1950’s Taylor analyzed the dielectric properties of ionic glasses in accordance with the Debye equation with a spread of relaxation times . He showed that the dielectric loss for all glasses fell on a single plot against scaled frequency. In 1962 Isard relabeled Taylor’s axis by plotting dielectric loss against log of the product of frequency and resistivity and thus essentially arrived at AC scaling in the form given in Eq. (1) . Since then Eq. (1), which we shall refer to as “Taylor-Isard scaling,” has been used in several different contexts. For instance, Taylor-Isard scaling was used by Scher and Lax in 1973 in their famous papers introducing the continuous time random walk approximation , by Summerfield and coworkers in 1985 for amorphous semiconductors , by van Staveren and coworkers in 1991 for metal-cluster compounds , and by Kahnt the same year for ionic glasses . We now proceed to comment on the scaling principle of Roling and coworkers, $`Cn/T`$ in Eq. (1), and Sidebottom’s recent interpretation of its occasional violations. If $`q`$ is charge, $`f_H`$ jump rate, and $`d`$ jump length, Sidebottom bases his arguments on the expressions $`\sigma (0)nq^2d^2f_H/k_BT`$ and $`\mathrm{\Delta }ϵnq^2d^2/k_BT`$ . Clearly, to obtain $`\stackrel{~}{\omega }`$ frequency should be scaled by $`f_H`$, thus leading to $`Cn/T`$ if all other parameters are constant. But since it is reasonable to assume the jump length would increase as concentration decreases, Sidebottom argues that one cannot expect $`Cn`$; on the other hand the $`n`$-dependence of $`d`$ is taken care of by Eq. (2) in which the unknown jump length is eliminated by scaling with the measured $`\mathrm{\Delta }ϵ`$. In our opinion Sidebottom’s arguments are largely correct, but Eq. (2) is much more general than it appears from his Letter. Consider hopping of completely non-interacting charge carriers on a cubic lattice. In this model conductivity is $`n`$ times charge carrier mobility and for given lattice jump frequencies the mobility scales with $`d^2`$. This shows that $`Cnd^2`$ as Sidebottom has it. However, if the lattice is homogeneous with just one jump frequency conductivity is frequency-independent; in order to have strongly frequency-dependent conductivity lattice jump frequencies must vary many decades. In such more realistic cases the expressions used by Sidebottom for $`\sigma (0)`$ and $`\mathrm{\Delta }ϵ`$ do not apply - there simply is no unique $`f_H`$. Still, we find in our simulations that these more realistic hopping models always obey Eq. (2). This leads to the question: When is Eq. (2) obeyed? Equation (2) apparently expresses two informations: 1) TTSP is obeyed, and 2) the scaled frequency is given by $`\stackrel{~}{\omega }=[ϵ_0\mathrm{\Delta }ϵ/\sigma (0)]\omega `$. We now show that 1) mathematically implies 2): TTSP implies the existence of some function $`\stackrel{~}{\sigma }(\stackrel{~}{\omega })`$ where $`\stackrel{~}{\omega }`$ is the scaled frequency. Expanding to first order in $`\stackrel{~}{\omega }`$ leads to $`\stackrel{~}{\sigma }=1+iA\stackrel{~}{\omega }`$ (where $`A`$ is real because $`\sigma ^{}(\omega )=\sigma (\omega )`$). Since $`\sigma =\stackrel{~}{\sigma }\sigma (0)`$ we have $`\sigma =\sigma (0)+iA\stackrel{~}{\omega }\sigma (0)`$. On the other hand, from the definition of $`\mathrm{\Delta }ϵ`$ one has $`\sigma =\sigma (0)+i\omega \mathrm{\Delta }ϵϵ_0`$ for $`\omega 0`$. Equating the two expressions for $`\sigma `$ leads to $`\stackrel{~}{\omega }=A^1[\mathrm{\Delta }ϵϵ_0/\sigma (0)]\omega `$, i.e., Eq. (2) (the numerical value of $`A`$ is irrelevant). Thus, Eq. (2) follows from TTSP alone. We next compare the data discussed by Sidebottom to two models, the macroscopic model and the symmetric hopping model. The macroscopic model considers AC conduction in a random mixture of phases with different (frequency-independent) conductivities . The AC conductivity is obtained by solving Maxwell’s equations. Computer simulations have shown that in the extreme disorder limit (where the local conductivities vary over many decades) the AC conductivity is accurately described by the effective medium approximation (EMA) universality equation , $$\stackrel{~}{\sigma }\mathrm{ln}\stackrel{~}{\sigma }=i\stackrel{~}{\omega }.$$ (3) The symmetric hopping model , which models AC conduction via random walks of non-interacting particles on a cubic lattice with random symmetric Arrhenius nearest-neighbor jump rates ($`\gamma _0e^{\beta E}`$, where $`E`$ is the random energy barrier), also exhibits universality in the extreme disorder limit. Even for this model does the EMA lead to Eq. (3) . For hopping, however, Eq. (3) does not give an accurate representation of the universal $`\stackrel{~}{\sigma }(\stackrel{~}{\omega })`$ . Before comparing model predictions to experimental data we present results from new simulations of the symmetric hopping model. The simulations were done on 3-dimensional samples with periodic boundary conditions, using a method based on exactly solving the master equation . Figure 1a shows $`\stackrel{~}{\sigma }^{}`$ \- the real part of $`\stackrel{~}{\sigma }`$ \- for the Box distribution of energy barriers. The frequency axis is scaled according to Eq. (2). Clearly, as $`\beta 1/k_BT`$ increases $`\stackrel{~}{\sigma }^{}`$ converges to a single curve. As seen in Fig. 1b showing the apparent exponent $`\mu `$ ($`\stackrel{~}{\sigma }^{}\stackrel{~}{\omega }^\mu `$) as function of $`\stackrel{~}{\sigma }^{}`$, the universal curve is the same for different energy barrier distributions. Figure 1b compares the simulations to three analytical approximations, the EMA universality equation Eq. (3), the percolation path approximation (PPA) , and a new “diffusion cluster approximation” (DCA), which we describe now: The reason for AC universality in hopping is the fact that in the extreme disorder limit conduction takes place on the “percolation cluster” formed by the links with largest jump rates until percolation ; the PPA-idea is to regard the conducting paths on the percolation cluster as strictly one-dimensional. The physical idea behind EMA is to replace the inhomogeneous lattice by an effective homogeneous medium determined self-consistently . Our hopping simulations show that the truth is somewhere between PPA and EMA , somewhere between one-dimensionality and homogeneity. Most likely, this is because both approximations ignore the fact that conduction takes place on some complex subset of the percolation cluster. This “diffusion cluster” must be smaller than the backbone (defined by removing dead-ends of the percolation cluster, fractal dimension=1.7 ) and larger than the set of red bonds (those that, when cut, stops the current, fractal dimension=1.1 ). At present this is all we know about the diffusion cluster and its dimension $`d_0`$ is regarded below as a fitting parameter. To derive the DCA equation we use EMA in $`d_0`$ dimensions. In the extreme disorder limit EMA implies $`\mathrm{ln}\stackrel{~}{\sigma }s\stackrel{~}{G}`$ where, if $`p(𝐤)=\frac{1}{d_0}_{i=1}^{d_0}\mathrm{cos}(k_i)`$, $`s\stackrel{~}{G}`$ is the following integral $`s\stackrel{~}{G}=_{\pi <k_i<\pi }\frac{d𝐤}{(2\pi )^d}\frac{i\omega }{i\omega +2d_0\sigma [1p(𝐤)]}`$. Whenever $`1<d_0<2`$ one finds $`s\stackrel{~}{G}(i\omega /\sigma )^{d_0/2}`$ at relevant frequencies. Thus, after rescaling frequency we arrive at the DCA equation, $$\mathrm{ln}\stackrel{~}{\sigma }=\left(\frac{i\stackrel{~}{\omega }}{\stackrel{~}{\sigma }}\right)^{d_0/2}.$$ (4) As is clear from Fig. 1b, the solution of this equation for $`d_0=1.35`$ gives an excellent fit to the universal AC hopping conductivity. In Fig. 2 the data discussed by Sidebottom are compared to the macroscopic model represented by the EMA universality equation (3) (dashed line), to the hopping model simulations in the extreme disorder limit (filled squares), and to the DCA (full line, Eq. (4)). Figure 2a shows Sidebottom’s own data for $`(\mathrm{Na}_2\mathrm{O})_\mathrm{x}(\mathrm{GeO}_2)_{1\mathrm{x}}`$ glasses, Fig. 2b shows data for the ionic melt $`0.4\mathrm{Ca}(\mathrm{NO}_3)_20.6\mathrm{KNO}_3`$ measured by Howell and coworkers , and Fig. 2c shows Patel’s data for the thioborate glass system $`(\mathrm{K}_2\mathrm{S})_\mathrm{x}(\mathrm{B}_2\mathrm{S}_3)_{1\mathrm{x}}`$ . In all three systems we find a frequency range where experimental data agree well with the symmetric hopping model. The symmetric hopping model is an extremely simple model which contains no fitting parameters but still gives a good fit to experimental data as shown above. The model does not include Coulomb interactions and it allows an arbitrary number of charge carriers at each site. We note, however, that the model is mathematically equivalent to that obtained by linearizing (with respect to the electric field) a hopping model with energy disorder and Fermi-statistics . The macroscopic model does not work very well for the ionic systems in Fig. 2. This is interesting because the macroscopic model via Gauss’ law includes all effects of Coulomb interactions. Apparently, the continuum description behind this model does not appropriately reflect the actual microscopic disorder. To summarize, we have briefly outlined the history of AC scaling in disordered solids and showed that the scaling version Eq. (2) must be obeyed whenever conductivity scaling is possible at all (TTSP obeyed). We have presented a new analytical approximation to the universal AC conductivity of hopping in the extreme disorder limit, the diffusion cluster approximation (DCA), and shown that it gives an excellent fit to simulations . Also, DCA (and thereby the extreme disorder limit of the symmetric hopping model) agrees well with the ionic data discussed by Sidebottom , in a frequency range depending on the system. ###### Acknowledgements. This work was supported in part by the Danish Natural Science Research Council.
no-problem/9907/nucl-th9907101.html	ar5iv	text	# Charge symmetry breaking in mirror nuclei from quarks ADP-99-28/T367 The binding energy differences of the valence proton and neutron of the mirror nuclei, <sup>15</sup>O – <sup>15</sup>N, <sup>17</sup>F – <sup>17</sup>O, <sup>39</sup>Ca – <sup>39</sup>K and <sup>41</sup>Sc – <sup>41</sup>Ca, are calculated using the quark-meson coupling (QMC) model. The calculation involves nuclear structure and shell effects explicitly. It is shown that binding energy differences of a few hundred keV arise from the strong interaction, even after subtracting all electromagnetic corrections. The origin of these differences may be ascribed to the charge symmetry breaking effects set in the strong interaction through the u and d current quark mass difference. PACS: 24.85.+p, 24.80.+y, 11.30.Hv, 12.39.Ba Keywords: Charge symmetry breaking, Quark-meson coupling model, Mirror nuclei, Effective mass The discrepancy between the calculated binding energy differences of mirror nuclei and those measured is a long-standing problem in nuclear physics. It is known as the Okamoto-Nolen-Schiffer (ONS) anomaly . Although it was first thought that electromagnetic effects could almost account for the observed binding energy differences, it is now believed that the ONS anomaly has its origin in charge symmetry breaking (CSB) in the strong interaction . In addition to calculations based on charge symmetry violating meson exchange potentials , a number of quark-based calculations have been performed in an attempt to resolve this anomaly. To some extent, these have been stimulated by the discovery of the nuclear dependence of the nucleon structure function measured in deep inelastic lepton-nucleus scattering (the nuclear EMC effect ). In such calculations, CSB enters through the up (u) and down (d) current quark mass difference in QCD. Despite these efforts, the difficulty of producing a realistic description of nuclear structure on the basis of explicit quark degrees of freedom has hindered the direct calculation of the binding energy differences. (One such quark-based, nuclear calculation exists , but it involved a shell model calculation for the iso-vector mass shifts of iso-spin multiplets in $`1s0d`$-shell nuclei, and the role of quarks entered through a model for the short-range CSB force.) In this study we report the results for the binding energy differences of the valence (excess) proton and neutron of the mirror nuclei, <sup>15</sup>O – <sup>15</sup>N, <sup>17</sup>F – <sup>17</sup>O, <sup>39</sup>Ca – <sup>39</sup>K and <sup>41</sup>Sc – <sup>41</sup>Ca, calculated using a quark-based model involving explicit nuclear structure and shell effects – the quark-meson coupling (QMC) model . This model has been successfully applied not only to traditional nuclear problems but also to other new areas as well . Although some exploratory QMC results on the ONS anomaly have already been reported , an early version of the model was used there, and it was applied to finite nuclei only through local density approximation, rather than a consistent shell model calculation. A detailed description of the Lagrangian density and the mean-field equations of motion needed to describe a finite nucleus is given in Refs. . A major difference in the present work compared with Refs. is that here charge symmetry is explicitly broken at the quark level through their masses. We use different values for the u and d current quark masses, and the proton and neutron (effective) masses. Thus, the saturation properties of symmetric nuclear matter needed to be recalculated to fix the relevant quark-meson coupling constants. At position $`\stackrel{}{r}`$ in a nucleus (the coordinate origin is taken at the center of the nucleus), the Dirac equations for the quarks in the proton or neutron bag are given by: $$\left[i\gamma _x\left(\left(\begin{array}{c}m_u\\ m_d\end{array}\right)V_\sigma ^q(\stackrel{}{r})\right)\gamma ^0\left(V_\omega ^q(\stackrel{}{r})\pm \frac{1}{2}V_\rho ^q(\stackrel{}{r})\right)\right]\left(\begin{array}{c}\psi _u(x)\\ \psi _d(x)\end{array}\right)=0,$$ $$(\|\stackrel{}{x}\stackrel{}{r}\|\mathrm{bag}\mathrm{radius}).$$ (1) The mean-field potentials for a bag centered at position $`\stackrel{}{r}`$ in the nucleus are defined by $`V_\sigma ^q(\stackrel{}{r})=g_\sigma ^q\sigma (\stackrel{}{r}),V_\omega ^q(\stackrel{}{r})=g_\omega ^q\omega (\stackrel{}{r})`$ and $`V_\rho ^q(\stackrel{}{r})=g_\rho ^qb(\stackrel{}{r})`$, with $`g_\sigma ^q,g_\omega ^q`$ and $`g_\rho ^q`$ the corresponding quark and meson-field coupling constants. (Note that we have neglected a possible, very slight variation of the scalar and vector mean-fields inside the nucleon bag due to its finite size .) The mean meson fields are calculated self-consistently by solving Eqs. (23) – (30) of Ref. with the proper modifications caused by the different proton and neutron (u and d quark) masses, namely, by solving a set of coupled, non-linear, differential equations for static, spherically symmetric nuclei, resulting from the variation of the effective Lagrangian density involving the quark degrees of freedom and the scalar, vector and Coulomb fields in mean field approximation. Thus, the present calculation is free from the sort of double counting questioned by Auerbach , namely that one should not add the effective mass difference reduction between the proton and neutron in medium on the top of the Coulomb displacement energies. Furthermore, the calculation also includes the shell effects which were discussed by Cohen et al. . Before discussing the results obtained, we need to specify the parameters and inputs used in the calculation. They are summarized in TABLE I. The bag constant, $`B`$, is determined by the bare proton mass in free space after allowing for the electromagnetic self-energy correction, $`+`$0.63 MeV. The parameter $`z`$ represents the sum of the center-of-mass and gluon fluctuation corrections, included in the standard MIT bag mass formula as $`z/R`$ and assumed independent of the density . $`B`$ and $`z`$ are determined by setting the bag radius in free space to be $`R=0.8`$ fm, and imposing the mass stability condition, $`\frac{m_p}{R}=0`$ . (See Ref. for details.) For the neutron, the procedure is the same as that for the proton, allowing for the electromagnetic self-energy correction, $``$0.13 MeV, but using the values of $`B`$ and $`z`$ determined above and calculating the d current quark mass and the bag radius for the neutron by the mass stability condition. Thus, the u current quark mass, $`m_u`$, is the basic input parameter used to fix the model parameters so as to reproduce the proton and neutron masses in free space after allowing for the electromagnetic self-energy corrections. The coupling constants, $`g_\sigma ^q`$ and $`g_\omega ^q`$ are determined so as to fit the saturation properties of symmetric nuclear matter – i.e., a binding energy of 15.7 MeV at the saturation density, $`\rho _0=0.15`$ fm<sup>-3</sup>. The binding energy is calculated by subtracting the average nucleon mass, $`(m_p+m_n)/2`$, and using the different scalar densities for protons and neutrons. In TABLE I, SU(2) stands for the parameters and inputs obtained and used for the calculation when SU(2) symmetry for the quarks and nucleons is assumed – i.e., using the same values for the u and d quark masses, and also for the proton and neutron masses. We should notice that the coupling constant, $`g_\sigma ^q`$, is also scaled for the present calculation of finite nuclei, by the same amount as found necessary in Ref. to fit the r.m.s. charge radius of <sup>40</sup>Ca (for the bag radius 0.8 fm) – keeping the ratio $`(g_\sigma /m_\sigma )`$ fixed, because the fixed ratio of $`(g_\sigma /m_\sigma )`$ has no effect on the properties of infinite nuclear matter. The quark-$`\rho `$ meson coupling constant, $`g_\rho ^q`$, needs some explanation. Within QMC, $`g_\rho ^q`$ is determined so as to reproduce the symmetry energy of 35 MeV. However, because the present model does not contain the $`\rho `$-nucleon tensor coupling and we work only in Hartree approximation , this gives an unrealistically large value for the coupling constant, $`g_\rho ^q=9.321`$. To make a realistic estimate, taking into account the $`\rho `$-meson central and spin-orbit potentials for the valence proton and neutron binding energies, we use the phenomenological value, $`g_\rho ^q=4.595`$ ($`g_{\rho NN}^2/4\pi =4\times 0.42`$), the value at zero three-momentum transfer corresponding to Hartree approximation, from TABLE 4.1 of Ref. . We also estimate the contributions of the $`\rho `$-potentials using the naive QMC value, $`g_\rho ^q=9.321`$, in order to test the sensitivity. In FIG. 1 we show the proton and neutron effective mass difference calculated in symmetric nuclear matter, including the electromagnetic self-energy corrections for the proton ($`+`$0.63 MeV) and neutron ($``$0.13 MeV). One notices that the proton and neutron mass difference becomes smaller as the baryon density increases – a result which was also found in Refs. . This seems to work in the right direction to resolve the ONS anomaly, but it is still not quantitative. (Recall the discussion of Auerbach .) Next, we show the calculated single-particle energies for <sup>17</sup>F and <sup>17</sup>O in TABLE II, as an example. These mirror nuclei have a common core nucleus, <sup>16</sup>O, and have an extra valence proton for <sup>17</sup>F and neutron for <sup>17</sup>O. In order to focus on the strong interaction effect for the valence proton and neutron, the Dirac equations are solved without the Coulomb and $`\rho `$-meson potentials, or the electromagnetic self-energy corrections, and keeping only the charge symmetric $`\sigma `$ and $`\omega `$ mean field potentials. Consistently, the valence nucleon contributions are not included in the Coulomb (proton) and $`\rho `$-mean field (iso-vector) source densities in the core nucleus. However, for the nucleons in the core nucleus, electromagnetic self-energy corrections and the Coulomb potential as well as the $`\rho `$ mean field potential are included in addition to the $`\sigma `$ and $`\omega `$ mean field potentials in solving the Dirac equations. Results will be shown for three cases for <sup>17</sup>F and <sup>17</sup>O: 1. Calculation performed imposing charge symmetry breaking through the u and d quark masses and the proton and neutron masses using the phenomenological $`\rho `$-quark coupling constant, $`g_\rho ^q=4.595`$ (at zero three-momentum transfer, $`\stackrel{}{q}=0`$, corresponding to Hartree approximation ) (denoted CSB). 2. Calculation performed assuming SU(2) symmetry for the u and d quark masses and the proton and neutron masses using the phenomenological $`\rho `$-quark coupling constant, $`g_\rho ^q=4.595`$ (denoted SU(2)). (See also the explanation of CSB.) 3. Calculation performed imposing charge symmetry breaking through the different u and d current quark masses and the proton and neutron masses using the $`\rho `$-quark coupling constant, $`g_\rho ^q=9.321`$ (denoted Case 3). The SU(2) results for <sup>17</sup>F and <sup>17</sup>O agree perfectly with each other as they should. Single-particle energies in the cores of <sup>17</sup>F and <sup>17</sup>O are slightly different for both CSB and Case 3. This difference is induced by the different (effective) masses for the valence proton and neutron, arising from the charge and density dependence of their coupling to the self-consistent scalar mean field. This also causes a second order effect on the Coulomb and $`\rho `$-meson potentials through the self-consistency procedure. The single-particle energies of the valence proton and neutron are practically equal for both CSB and Case 3. It is interesting to compare the binding energy differences between the valence proton in <sup>17</sup>F and neutron in <sup>17</sup>O. Both CSB and Case 3 results give, $`E(p)(1d_{5/2})E(n)(1d_{5/2})0.18`$ MeV, while the SU(2) case is zero as it should be. This amount already shows a magnitude similar to that of the observed binding energy differences, where the origin may be ascribed to the effect of the proton-neutron effective mass difference reduction and simultaneous effect of the core nucleus potentials. Note that in QMC the quark scalar charge for the d quark, which is defined by the integral of the quark scalar density over the nucleon volume, is slightly greater than that for the u quark, because the u quark mass is smaller than the d quark mass. The lower component of the u quark wave function is enhanced more than that of the d quark. This is a simple consequence of relativistic quantum mechanics. Nevertheless, as a result, the in-medium proton-$`\sigma `$ and neutron-$`\sigma `$ coupling constants, $`g_\sigma ^p(\sigma )`$ and $`g_\sigma ^n(\sigma )`$, differ from their values in free space and the proton and neutron effective mass difference is reduced . This leads to a reduction in the binding energy differences below the amount one naively expects from the proton and neutron mass difference of about 2 MeV in free space (without the electromagnetic self-energy corrections) – see also TABLE I and FIG. 1. The density dependence of the effective $`pn`$ mass difference, which we have just described, is the major source of charge symmetry violation discussed here. On the other hand, the fact that there is a small $`\rho ^0`$ mean field also affects the systematics as we vary $`A`$ and we now examine this contribution. In FIG. 2 we show the $`\rho `$-meson mean field potential generated by the core in <sup>17</sup>F and <sup>17</sup>O, for CSB and SU(2). There is no distinguishable difference between <sup>17</sup>F and <sup>17</sup>O for CSB. We will evaluate the $`\rho `$-meson central and spin-orbit potential contributions to the single-particle energies of the valence proton and neutron perturbatively. We should note that QMC gives the correct expression for the spin-orbit potentials, including the finite size of the nucleon : $`V^{s.o.}(r)\stackrel{}{l}\stackrel{}{s}`$ $`=`$ $`{\displaystyle \frac{1}{2m_N^2(r)r}}\left[\mathrm{\Delta }_\sigma +3(12\mu _s\eta (r))\mathrm{\Delta }_\omega +{\displaystyle \frac{1}{2}}\tau _3^N(12\mu _v\eta (r))\mathrm{\Delta }_\rho \right]\stackrel{}{l}\stackrel{}{s}.`$ (2) We are interested in the last term in Eq. (2), the $`\rho `$-meson spin-orbit potential, which gives opposite contributions for the valence proton and the valence neutron. Contributions from the other mesons to Eq. (2) have the same sign for protons and neutrons and their contributions to the binding energy differences are therefore expected to be tiny. Furthermore, the contribution from the effective mass difference of the proton and neutron is even higher order. Thus, we use the SU(2) value for $`m_N^2`$ in Eq. (2) to evaluate the $`\rho `$-meson spin-orbit potential. Using the calculated wave functions for the valence proton and neutron, obtained by solving the Dirac equations without the Coulomb and $`\rho `$-meson potentials or the electromagnetic self-energy corrections, we evaluate the $`\rho `$-meson contributions perturbatively: $`\delta E_\rho `$ $`=`$ $`{\displaystyle d^3r\psi _{valence}^{}(\stackrel{}{r})[\frac{1}{2}\tau _3^NV_\rho (r)]\psi _{valence}(\stackrel{}{r})},`$ (3) $`\delta E_\rho ^{s.o.}`$ $`=`$ $`{\displaystyle d^3r\psi _{valence}^{}(\stackrel{}{r})[\frac{1}{2}\tau _3^NV_\rho ^{s.o.}(r)](\stackrel{}{l}\stackrel{}{s})\psi _{valence}(\stackrel{}{r})},`$ (4) where $`1/2V_\rho (r)`$ is shown in FIG. 2, and $`V_\rho ^{s.o.}(r)=\frac{1}{2m_N^2(r)r}(12\mu _v\eta (r))\mathrm{\Delta }_\rho `$, the third term in Eq. (2). In QMC the iso-vector magnetic moment, $`\mu _v`$, is calculated to be $`\mu _v^{QMC}=2.558`$ for the bag radius $`R=0.8`$ fm. This is somewhat smaller than the empirical value, $`\mu _v^{emp.}=4.7051`$. Thus, for the CSB and SU(2) calculations, we use the empirical value, $`\mu _v^{emp.}=4.7051`$, together with the phenomenological coupling constant, $`g_\rho ^q=4.595`$, in order to make a more realistic estimate. In TABLE III we summarize the calculated single-particle energies for the valence proton and neutron of the mirror nuclei for two cases, CSB and SU(2). We expect that the results for CSB are the more realistic. Comparing the $`\rho `$-potential contributions for the hole states with core plus valence states, one notices the shell effects due to the $`\rho `$-potentials. The $`\rho `$-potential contributions for the discrepancies of the <sup>15</sup>O – <sup>15</sup>N and <sup>17</sup>F – <sup>17</sup>O binding energy differences are about $``$0.11 MeV and $``$0.011 MeV, respectively, while for the <sup>39</sup>Ca – <sup>39</sup>K and <sup>41</sup>Sc – <sup>41</sup>Ca cases, they are about $``$0.17 MeV and $``$0.013 MeV, respectively. These results reflect the difference in the shell structure, namely hole states tend to have larger $`\rho `$-potential contributions than the core plus valence nucleon states. This can be understood because, in the hole states, the excess proton or neutron sits in the region where the iso-vector density distribution is larger. For the SU(2) case, the valence state binding energy differences of mirror nuclei, $`\delta E`$, come entirely from the $`\rho `$-meson potentials. We see that $`\delta E`$ obtained in CSB is always larger than that for SU(2). Typical values for the binding energy differences are a few hundred keV. The larger binding energy differences for the valence proton and neutron obtained in CSB indicate that the prime CSB effects originate in the u-d current quark mass difference. The resulting contribution to the binding energy differences is of the order of about a few hundred keV. This is precisely the order of magnitude which is observed as the ONS anomaly . Furthermore, as we see from TABLE III, the systematic dependence on $`A`$ is also reasonably well described, except for the <sup>39</sup>Ca – <sup>39</sup>K case. It is a fascinating challenge for the future to compare this result with the traditional mechanism involving $`\rho \omega `$ mixing . This will involve the issue of the possible momentum dependence of the $`\rho \omega `$ mixing amplitude . In addition, one would need to examine whether there is any deeper connection between these apparently quite different sources of charge symmetry violation. We would like to stress that the present contribution to the ONS anomaly is based on a very simple but novel idea, namely the slight difference between the quark scalar densities of the u and d quarks in a bound nucleon, which stems from the u and d quark mass difference . Our results were obtained within an explicit shell model calculation, based on quark degrees of freedom. They show that if charge symmetry breaking is set through the u and d current quark mass difference so as to reproduce the proton and neutron masses in free space (without any electromagnetic interaction effects), it produces binding energy differences for the valence (excess) proton and neutron of mirror nuclei of a few hundred keV. The origin of this effect within relativistic quantum mechanics is so simple that it is natural to conclude that a sizable fraction of the charge symmetry breaking in mirror nuclei arises from the density dependence of the u and d quark scalar densities in a bound nucleon. We would like to thank A. G. Williams for helpful comments on the manuscript. This work was supported by the Australian Research Council and the Japan Society for the Promotion of Science.
no-problem/9907/nucl-th9907075.html	ar5iv	text	# Dynamics and Sizes of the Fireball at Freeze-outWork supported by DAAD, BMBF, GSI, and DFG. ## 1 MOTIVATION In relativistic heavy-ion collisions the momenta of the escaping hadrons are fixed at the point of “thermal freeze-out”. They carry information about the final state of the fireball which one can use to reconstruct this state and thus obtain boundary conditions for back-extrapolation into earlier stages of its evolution. In this way one can examine the question whether or not a quark-gluon plasma was created. Assuming thermalization and collective transverse flow at freeze-out, these two features can be extracted separately by combining single-particle $`m_{}`$-spectra and two-particle Bose-Einstein correlations . Particles of non-zero transverse momentum are emitted from a fireball region which moves outwards transversely; in the local comoving frame their transverse momentum is lower, resulting in an enhanced Boltzmann factor. Transverse flow thus flattens the $`m_{}`$-spectrum. A given spectral slope can, however, be reproduced by different combinations of flow and temperature, because larger flow can be compensated for by lower temperature and vice versa. On the other hand, the sizes of the regions emitting particles of given momentum (homogeneity regions, measured by correlation radii) are controlled by the expansion velocity gradients. The natural velocity measure is given by the chaotic thermal motion; increasing $`T`$ and the transverse flow (velocity gradients) together at the proper rate will thus not change the resulting correlation radii. In summary, $`T`$ and $`v_{}`$ are anticorrelated by a given spectral slope, but correlated by the measured homogeneity sizes. Exploiting both types of correlations their values can be determined unambiguously. This method was used for analyzing the data from central Pb+Pb collisions at 158 $`A`$GeV/$`c`$ taken by the NA49 collaboration . ## 2 THE MODEL It is impossible to extract from Bose-Einstein correlations the global fireball features in a model-independent way. We therefore parameterize the fireball by an emission function, i.e. by a model for the (Wigner) phase-space density of the particles at freeze-out : $`S(x,K)d^4x`$ $`=`$ $`{\displaystyle \frac{m_{}\mathrm{cosh}(y\eta )}{(2\pi )^3}}\mathrm{exp}\left({\displaystyle \frac{Ku(x)\mu }{T}}\right)\mathrm{exp}\left({\displaystyle \frac{(\eta \eta _0)^2}{2(\mathrm{\Delta }\eta )^2}}\right)`$ (1) $`\times G(r){\displaystyle \frac{d\tau }{\sqrt{2\pi (\mathrm{\Delta }\tau )^2}}}\mathrm{exp}\left({\displaystyle \frac{(\tau \tau _0)^2}{2(\mathrm{\Delta }\tau )^2}}\right)\tau d\eta rdrd\phi .`$ The main assumptions here are the following: (i) Approximate thermal equilibrium at temperature $`T`$. (ii) Boost-invariant longitudinal flow combined with a tunable transverse expansion profile with a transverse flow rapidity given by $`\eta _t(r)=\eta _f(r/r_{\mathrm{rms}})`$; this expansion is encoded in the velocity field $`u(x)`$. (iii) For the transverse density profile $`G(r)`$ two different forms were studied: $$\text{box-like:}G(r)=\theta (R_Br),\text{Gaussian:}G(r)=\mathrm{exp}\left(\frac{r^2}{2R_G^2}\right).$$ (2) If the parameters $`T`$, $`\eta _f`$, and $`R_B`$ or $`R_G`$ are independent of space-time rapidity $`\eta `$, the model cannot reproduce the rapidity dependence of the data. We therefore concentrate here on a single rapidity window ($`3.9<Y<4.4`$). The time parameters $`\tau _0`$, $`\mathrm{\Delta }\tau `$ were fixed together with the other parameters from the correlation radii, whereas the longitudinal width $`\mathrm{\Delta }\eta =1.3`$ of the density profile was fixed by adjusting the width of $`dN/dy`$. The correlation radii were calculated numerically from the “model-independent expressions” . When calculating the single-particle spectrum resonances were taken into account as in ; since we fitted the $`h^{}`$ spectrum we also added negative kaons and antiprotons. For the resonances baryonic and strangeness chemical potentials, $`\mu _\mathrm{B}`$ and $`\mu _\mathrm{S}`$, were taken into account as required by baryon number and strangeness conservation . ## 3 THE FIT Fig. 1 shows the $`\chi ^2`$ contours of our fit to the data. The valley extending from the upper left to the lower right corner shows the ambiguity of the extraction of temperature and transverse flow from the spectrum. Moreover, the left and right figure show also a model ambiguity: at fixed temperature different density profiles lead to slightly different transverse expansion velocities. The results of the fit to the correlation radii (width parameters in a Gaussian parametrization of the correlation function) indicate that a box-like density fits the data better than a Gaussian one, although the latter cannot be excluded. Strictly speaking, only models with too weak transverse flow seem to be completely wrong. With the box-model we find $`T80110\text{MeV}`$ and $`\overline{v}_{}0.47c0.62c`$. The radius of the transverse density distribution $`R_B=12.1\text{fm}`$, corresponding to $`r_{\mathrm{rms}}=8.6`$ fm, is about twice as large as the initial transverse overlap region of the two Pb nuclei, consistent with the strong transverse flow. Both models fit the $`m_{}`$-spectrum well but, as shown in Fig. 2, they differ in the quality of the fit to the correlation radii. The $`M_{}`$-dependence of transverse correlation radii $`R_o`$ and $`R_s`$ is better captured by the box-model. This can be traced back to the fact that the box-model has a rigid radial density cutoff; for increasing $`K_{}`$ the homogeneity region is thus squeezed towards that edge, making it narrower in the direction of $`K_{}`$. As a result $`R_o`$ decreases more strongly with $`K_{}`$ than for the Gaussian model where the homogeneity region expands into the tail of the density distribution (see Fig. 2b). We stress that it is the generic feature shown in the upper half of Fig. 2b (and not the box-like transverse density distribution itself) which appears to be required by the data. ## 4 MULTIPLICITY From the fitted models we can calculate the total pion multiplicity, assuming chemical equilibrium at thermal freeze-out with $`\mu _\mathrm{B}360\text{MeV}`$, corresponding to an entropy per baryon of about $`S/A`$=38 , and $`\mu _\mathrm{S}`$ fixed by strangeness neutrality. This implies $`\mu _\pi `$=0 and results in a predicted multiplicity which is more than a factor 3 below the measured value. By investigating the pion phase-space density we inferred a chemical potential for the direct pions of $`\mu _\pi 60\text{MeV}`$ at $`T100\text{MeV}`$. This implies that at thermal freeze-out the pions are out of chemical equilibrium. This is expected if chemical freeze-out happens in equilibrium near the confinement phase transition at $`T_{\mathrm{cr}}170`$ MeV . ## 5 CONCLUSIONS A combined analysis of single-particle spectra and two-particle correlations avoids the ambiguities and strongly reduces the model-dependencies resulting from an analysis of the $`m_{}`$-spectrum of only a single particle species . It leads to the unavoidable conclusion that thermal freeze-out of the fireball happens at a very low temperature ($`T_{\mathrm{therm}}80110\text{MeV}`$) and is accompanied by very strong transverse flow ($`\overline{v}_{}0.47c0.62c`$). The $`M_{}`$-dependence of the homogeneity lengths can be somewhat better reproduced by a box-like than by a Gaussian transverse density profile at freeze-out. At this low temperature the fireball is no longer in chemical equilibrium. The pion abundance appears to be fixed by chemical freeze-out already at $`T_{\mathrm{chem}}T_{\mathrm{cr}}170\text{MeV}`$ . We find that at $`T_{\mathrm{therm}}`$=100 MeV this leads to a pion chemical potential of $`\mu _\pi 60`$ MeV .
no-problem/9907/cond-mat9907098.html	ar5iv	text	# Untitled Document VISIBILITY DIAGRAMS AND EXPERIMENTAL STRIPE STRUCTURES IN THE QUANTUM HALL EFFECT Yvon Georgelin<sup>a</sup>, Thierry Masson<sup>b</sup> and Jean-Christophe Wallet<sup>a</sup> <sup>a</sup>Groupe de Physique Théorique, Institut de Physique Nucléaire F-91406 ORSAY Cedex, France <sup>b</sup>Laboratoire de Physique Théorique (U.M.R. 8627), Université de Paris-Sud Bât. 211, F-91405 ORSAY Cedex, France Abstract: We analyze various properties of the visibility diagrams that can be used in the context of modular symmetries and confront them to some recent experimental developments in the Quantum Hall Effect. We show that a suitable physical interpretation of the visibility diagrams which permits one to describe successfully the observed architecture of the Quantum Hall states gives rise naturally to a stripe structure reproducing some of the experimental features that have been observed in the study of the quantum fluctuations of the Hall conductance. Furthermore, we exhibit new properties of the visibility diagrams stemming from the structure of subgroups of the full modular group. (June 1999) LPT-99/27 1. INTRODUCTION The Quantum Hall Effect (QHE) is a remarquable phenomenon occuring in a two-dimensional electron gas in a strong magnetic field at low temperature . Since the discovery of the quantized integer and fractional Hall conductivity, the QHE has been an intensive area of theoretical and experimental investigations. The pioneering theoretical contributions analyzing the basic features of the hierarchy of the Hall plateaus have triggered numerous works aiming to provide a better understanding of the underlying properties governing the complicated phase diagram associated with the quantum Hall regime together with the precise nature of the various observed transitions between plateaus and/or focusing on a characterization of a suitable theory. It has been realized for some time that modular symmetries may well be of interest to understand more deeply salient features of the QHE. For instance, it has been suggested , \[6a-b\] that some properties of the phase diagram may be explained in terms of modular group transformations. At the present time, a fully satisfactory microscopic or effective theory for the QHE, from which the relevant modular symmetry (if any) would come out, is still lacking. This has motivated further studies focalized on the derivation of general constraints on the phase diagram coming for the full modular group or some of its subgroups \[6a-b\]. Some time ago, we have shown that a special subgroup of the full modular group, namely the group $`\mathrm{\Gamma }(2)`$, can be used to derive a model for a classification of integer as well as fractional Hall states \[7a-b\]. We have further shown that the constraints stemming from $`\mathrm{\Gamma }(2)`$ on physically admissible $`\beta `$-functions give rise to a global phase diagram as well as crossover in the various observed transitions which are in good agreement with the present experimental observations. Here, it is worth recalling that the classification based on the $`\mathrm{\Gamma }(2)`$ symmetry \[7a-b\], which refines the Jain and Haldane ones and can be somehow viewed as a kind of generalization of the law of the corresponding states proposed in , seems to reproduce successfully the observed hierarchical structure of the Hall states. This construction involves two important building blocks (in addition to the action of the group $`\mathrm{\Gamma }(2)`$ itself) called the visibility diagrams. Basically, these diagrams, inherited from theoretical studies in arithmetics and rigidely linked to the structure of $`\mathrm{\Gamma }(2)`$, have been shown to encode a great amount of information on the experimentaly observed global organization of the quantum Hall states. Recently, some new experiments on the mesoscopic conductance in the quantum Hall regime in silicon MOSFETs have been performed . The main result is that the extrema of the conductance fluctuations spread on linear trajectories in the gate voltage $`V_g`$-magnetic field $`B`$ plane. These lines are parallels to lines of constant filling factor $`\nu =p`$ (with $`p=0,1,2,\mathrm{}`$). Moreover, a clear stripe structure can be observed on this plane. The purpose of this note is to analyze more deeply various properties of the visibility diagrams (therefore extending our previous investigations \[7b\])as well as to confront them to the present experimental situation for the QHE. We show that a suitable physical interpretation of the visibility diagrams that permits one to describe successfully the observed architecture of the Quantum Hall states gives rise naturally to a stripe structure that reproduces some of the experimental features of the stripe structure of the conductance fluctuations in the $`V_g`$-$`B`$ plane. This is presented in section 2 where the interesting role played by the visibility diagrams in the description of some of the physics of the Quantum Hall regime is emphasized. Furthermore, we exhibit in section 3 new properties of these diagrams stemming from the structure of subgroups of the full modular group. Finally, we summarize the results and conclude. 2. VISIBILITY DIAGRAMS AND STRIPE STRUCTURES Let us first recall the essential features of $`\mathrm{\Gamma }(2)`$ that will be needed in the following analyzis. The group $`\mathrm{\Gamma }(2)`$ is the set of transformations $`G`$ acting on the upper-half complex plane $`𝒫`$ which can be written as $$G(z)=\frac{(2s+1)z+2n}{2rz+(2k+1)},k,n,r,sZ$$ $`(1),`$ where $$(2s+1)(2k+1)4rn=1\text{unimodularity condition}$$ $`(2),`$ and $`z𝒫`$ (Im$`z>0`$). The two generators of $`\mathrm{\Gamma }(2)`$ are defined by $$T^2(z)=z+2,\mathrm{\Sigma }(z)=\frac{z}{2z+1}$$ $`(3a,b).`$ Now to construct from $`\mathrm{\Gamma }(2)`$ a model for the classification of the quantum Hall states identify first the complex coordinate $`z`$ with the filling factor $`\nu =p/q`$. Then, as we have shown in \[7a\], for a given (fixed) even denominator metallic state $`\lambda =\frac{2s+1}{2r}`$, the hierarchy of the (liquid) odd denominator states surrounding this metallic state is obtained from the images $`G_{n,k}^\lambda (0)`$ and $`G_{n,k}^\lambda (1)`$ of $`0`$ and $`1`$ by the family of transformations $`G_{n,k}^\lambda \mathrm{\Gamma }(2)`$ where $`\lambda =\frac{2s+1}{2r}`$ holds and $`n,k`$ are constrained by the unimodularity condition (2). The action of $`\mathrm{\Gamma }(2)`$ on the filling factor $`\nu =p/q`$ can be visualized with the help of graphical representations called visibility diagrams whose construction is now summarized (for a detailed construction see \[7b\]). Consider a two-dimensional square lattice whose vertices are indexed by a couple of positive (or zero) relatively prime integers $`(q,p)`$. Since $`\mathrm{\Gamma }(2)`$ preserves the parity of the denominator of any rational fraction, there are actually two ways for organizing the vertices pertaining to this lattice, depending whether the denominator is even or odd. Consider first the case where it is even and choose therefore a given $`\lambda =\frac{2s+1}{2r}`$ as a starting vertex. Then, it is not difficult to realize that $`G_{n,k}^\lambda (0)`$ and $`G_{n,k}^\lambda (1)`$ which in the present framework label the Hall plateaus surrounding the even denominator (metallic) Hall state corresponding to $`\lambda `$ are all located on two parallel straigh lines forming an unbounded left-ended stripe surrounding the vertex $`\lambda `$. Finally, the application of a similar process to all even denominator fractions gives rise to a collection of non-overlapping stripes as depicted on fig.1, each stripe corresponding therefore to a vertex with even denominator. This visibility diagram, hereafter called even diagram, involves naturally the Jain hierarchy which corresponds to the stripe associated with $`\lambda =1/2`$ as it can be easily realized by computing the successive values for $`G_{n,k}^{1/2}(0)`$ and $`G_{n,k}^{1/2}(0)`$ using (1) and (2). It is worth recalling some experimental results performed in \[11-13\] on metal-insulator transitions. It appears that the corresponding phase diagrams (see e.g. fig.2 of ref. and fig.2 of ref.; see also a recent result reported in ) exhibit a stripe structure which bears some similarity with the stripe structure occuring on the even diagram when, anticipating what will be done in a while, the $`(q,p)`$ plane is identified with the magnetic field-charge carrier density planeat least for not too small charge carrier density. This can be easily illustrated by selecting from fig.1 the relevant stripes which corresponds here to $`\lambda =(2s+1)/2`$ as shown on fig.2 to be compared to the fig.3 of ref.. We close this part of the discussion by noting that an experimental phase diagram of the integer QHE (charge carrier density ($`V_g`$) versus magnetic field) has been explicitely obtained very recently in suggesting that the vertical leftmost half stripe that is involved in the even diagram might well be associated with some insulator state. In fact, provided the above interpretation of the $`(q,p)`$ plane is actually correct, this vertical stripe (indexed by $`\mathrm{"}1/0\mathrm{"}`$) might correspond to the type I insulator indicated in , whose properties are different from the type II insulator. The second diagram that can be constructed, hereafter called the odd diagram, can be readily obtained by using a well-known theorem in arithmetics which states that for any relatively prime integers $`q`$ and $`p`$ there exist (necessarily) prime integers $`a`$ and $`b`$ such that $$qbpa=\pm 1$$ $`(4).`$ Then, for any $`(q,p)`$ vertex of the lattice with $`q`$ odd, associate the set of points $`(a,b)`$ satisfying this relation. It is easy to realize that these points are located on two parallel straight lines forming a stripe surrounding the $`(q,p)`$ vertex as depicted on fig.3. We then obtain stripes for any odd denominator fraction $`q/p`$. Notice that the stripes can overlap, contrary to what happens for the even diagram <sup></sup><sup></sup>It is not dififcult to see that the Haldane hierarchy is involved in the odd visibility diagram.. We are now in position to show that this latter diagram encodes interesting information concerning some recent experimental results related to the quantum Hall fluctuations of the conductance that have been reported in . The main result of this study on the mesoscopic conductance in the Quantum Hall regime in Silicon MOSFET is that the extrema for the conductance fluctuations spread on linear trajectories in the $`V_g`$-$`B`$ plane parallel to constant (integer) filling factor. Namely, the following relation $$\frac{C}{e}\frac{V_g}{B}=n\frac{e}{h},n\text{integer}$$ $`(5)`$ in which $`\frac{C}{e}=\frac{\rho }{V_g}`$ is assumed to hold ($`\rho `$ denotes the electron density and the constant $`\frac{C}{e}`$$`=`$$`\mathrm{8.6\; 10}^{11}`$cm<sup>-2</sup>V<sup>-1</sup> in ) has been found to be verified within a few percent of accuracy. Moreover, a stripe structure in the $`V_gB`$ plane has been observed, as shown on fig.4 (taken from the fig.2 of ). In order to confront some of these experimental results to the present $`\mathrm{\Gamma }(2)`$ framework, we have first to exhibit a possible relation between the visibility diagram, the gate voltage and the applied magnetic field. This proceeds as follows. First, recall that the filling factor $`\nu =\frac{p}{q}`$ can be expressed from its very definition as $$\nu =\frac{N_c}{N_\mathrm{\Phi }}$$ $`(5)`$ in which $`N_c`$ is the number of charge carriers (the electrons in the present case) and $`N_\mathrm{\Phi }=`$$`BS\frac{h}{e^2}`$ ($`S`$ is the device area) is the number of unit flux. Then observe that the action of the operator $`T^2`$ (3.a), a Landau shift type operator, on any vertex $`(q,p)`$ of the visibility diagram gives rise to a vertical shift ($`T^2:(q,p)(q,p+2q)`$) whereas the action on any vertex of $`\mathrm{\Sigma }`$ (3.b) which is the flux attachement operator produces a horizontal shift ($`\mathrm{\Sigma }:(q,p)(q+2p,p)`$). Next, observe that the charge carrier density happens to be proportional to the gate voltage $`V_g`$ in the experiments reported in and that the flux attachement is obtained by magnetic field variation. Bringing this all together, this suggests to identify naturally (up to dimensionful factors) the horizontal (resp. vertical) axis on the visibility diagram with the $`B`$-axis (resp. $`V_g`$-axis). Owing to the above identification, it is now possible to perform a comparison between the stripe structure occuring in the $`V_gB`$ plane observed in (represented by the grey areas associated with the plateaus) and the one stemming from the odd diagram. These structures are depicted respectively on fig.4 and fig.5. For the sake of clarity, we have only represented on fig.5 the stripes corresponding to the integer filling factors considered in . We observe a good qualitative agreement between both structures. Each grey area on fig.4 corresponding to a given (integer) plateau appears to be bounded by two parallel straight lines (at least in the considered range for $`V_g`$ and $`B`$). Taking into account the method for constructing the odd diagram, it seems natural to identify each grey area associated with a given integer filling factor with the corresponding stripe on fig.5. If this latter identification is physically correct, it is possible to obtain from the odd diagram further information on the Hall conductance as a function of $`V_g`$ (uppermost onset of fig.4) by adapting to the present situation, for which $`B`$ is fixed while $`V_g`$ ($`N_c`$) varies, the argument that we used in \[7b\] to obtain a resistivity plot (Hall resistivity versus $`B`$) fitting well with the experimental plot. This is straighforwardly achieved by simply assuming that the vertical width of any stripe on fig.5, defined by the intercept of any vertical line with that stripe, is proportional to the width of the corresponding plateau. The subsequent analysis is then very similar to the one that we described in \[7b\]. We have found that the resulting conductivity-$`V_g`$ plot agrees qualitatively with the one depicted on fig.4. At this point, the analysis suggests that the proposed physical interpretation of the odd diagram is consistent with the experimental observations corresponding to the integer QHE. Proving that this diagram encodes some relevant properties of both integer and fractional QHE (giving therefore some global information on the phase diagrams) would require further confrontation with experiments exploring the fractional quantum Hall regime (and/or for higher magnetic field). In this regime, it is obvious from the very construction of the odd diagram that one (experimentaly testable) prediction of the present scheme is the occurence of branching tree-like structures \[7b\] among the stripes in the $`V_gB`$ plane similar to the one appearing on fig.3. Let us now consider the quantum Hall fluctuations. Recall that these fluctuations are observed in the transition regions between plateaus which narrow as the temperature decreases while the fluctuations grow and sharpen. Obviously, a full confrontation of the experimentaly verified relation (5) to the present $`\mathrm{\Gamma }(2)`$ framework would require to have at hand a detailled model for the quantum Hall fluctuations Note that relation (5) seems to contradict the predictions stemming from the non-interacting models. providing a satisfactory explanation for the origin of the observed behaviour. Nevertheless, keeping in mind the above analysis of the stripe structures, it seems plausible to conjecture that, for a given transition $`\nu =n_1\nu =n_2,n_{1,2}N`$ (extending to $`n_{1,2}Q`$ if the present scheme applies to the fractional QHE), the directions of the two coexisting families of straight lines involving the extrema of the fluctuations (which are observed in for the transitions $`01`$, $`12`$, $`23`$, $`34`$) are given by the directions defined by the corresponding stripes involved in that transition. 3. MORE ON VISIBILITY DIAGRAMS AND DISCUSSION From the above analysis, it appears that the visibility diagrams may well encode some physically relevant features of the QHE which motivates a deeper investigation of mathematical properties underlying their structure. This is what we consider now. First, we point out that the odd diagram is in fact the superposition of two ”more elementary” visibility diagrams. To see that, consider separately odd and even numerator filling factors (with odd denominator) and apply the method for constructing the odd visibility diagram that has been described in section 2. Doing this, one obtains the two new diagrams represented on fig.6 and fig.7 corresponding respectively to odd and even numerator filling factors (hereafter called respectively odd/odd and even/odd diagrams). Then, it can be easily realized that the superposition of these two latter diagrams gives rise to the odd visibility diagram. Note that the stripes appearing in these two diagrams do not overlap as it is the case for the even diagram. Furthermore, observe that this latter diagram is related to the even/odd diagram through a symmetry around the $`p=q`$ axis which corresponds to the action of an operation belonging to the full modular group $`\mathrm{\Gamma }(1)`$ but not to $`\mathrm{\Gamma }(2)`$. Let us study more closely the action of modular transformations pertaining to $`\mathrm{\Gamma }(1)`$ on the even, even/odd and odd/odd diagrams. Some remarks are in order. On one hand, it can be easily seen that the action of any $`G\mathrm{\Gamma }(1)`$ preserves the arithmetic relation (4) ruling the whole construction of these diagrams. As a consequence, the action of $`\mathrm{\Gamma }(1)`$ maps the stripe structure of each of the diagrams into another one or possibly a sub-structure. On the other hand, any $`G\mathrm{\Gamma }(2)`$ maps each of these three diagrams into itself simply because $`\mathrm{\Gamma }(2)`$ preserves the even or odd character of both numerator and denominator involved in the filling factor. In other words, the whole structure of each diagram remains invariant under $`\mathrm{\Gamma }(2)`$. In fact, it appears that $`\mathrm{\Gamma }(2)`$ is the largest subgroup of $`\mathrm{\Gamma }(1)`$ leaving invariant each of the 3 diagrams. To see that, consider the action of the coset group $`\mathrm{\Gamma }(1)/\mathrm{\Gamma }(2)`$ on these diagrams. It is known in the mathematical litterature that this coset group involves 6 elements whose corresponding representatives in $`\mathrm{\Gamma }(1)`$ can be choosen as: $$I=\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right),U=\left(\begin{array}{cc}1& 1\\ 0& 1\end{array}\right),V=\left(\begin{array}{cc}0& 1\\ 1& 0\end{array}\right)$$ $`(6a,b,c)`$ $$W=\left(\begin{array}{cc}1& 0\\ 1& 1\end{array}\right),P=\left(\begin{array}{cc}0& 1\\ 1& 1\end{array}\right),P^2=\left(\begin{array}{cc}1& 1\\ 1& 0\end{array}\right)$$ $`(6d,e,f),`$ from which it can be verified that any of the 6 coset group elements maps a visibility diagram into a sub-diagram included in another diagram. Finally, using (6) together with the definition of $`\mathrm{\Gamma }(2)`$, it is easy to prove that the even and odd diagrams (depicted respectively on fig.1 and fig.2) are invariant under the action of another subgroup of $`\mathrm{\Gamma }(1)`$ generated by $`U`$ given by (6b) and $`\mathrm{\Gamma }(2)`$. This subgroup is nothing but $`\mathrm{\Gamma }_0(2)`$, which has been proposed \[6b\] (see also second of ref. \[6a\]) as another candidate for a discrete symmetry group underlying the physics of the QHE. Note that the constraints from $`\mathrm{\Gamma }_0(2)`$ on the renormalization group flow in a two-parameter scaling framework have been examined in \[6b\]. The resulting flow diagram (phase diagram) has been shown to exhibit a specific feature. In fact, consistency with the present experimental observations requires the occurence for the $`01`$ transition Recall that, as usual, the whole flow diagram is obtained by applying successive $`\mathrm{\Gamma }_0(2)`$ transformations to the $`01`$ ”template” transition of a critical point at $`\sigma _{xy}=\sigma _{xx}=1/2`$ which appears as a pole of the corresponding $`\beta `$-function. This stemms from the existence of a fixed point of $`\mathrm{\Gamma }_0(2)`$ in its fundamental domain at $`z_0=(1+i)/2`$ (recall that in this framework one has $`z=\sigma _{xy}+i\sigma _{xx}`$ which parametrizes the conductivity plane). The situation is different in the $`\mathrm{\Gamma }(2)`$ case, as shown recently in : There is no critical point showing up as a pole (at finite distance in the conductivity plane) of the corresponding $`\beta `$-function but consistency with the two-parameter scaling hypothesis seems to require the occurence in each allowed transition of a temperature-independant point that might be identified with the crossing point appearing in the crossover of the observed transitions . Let us summarize the results involved in this paper. We have analyzed various properties of the visibility diagrams which are related to the modular symmetries. In particular, we have shown that the observed structures occuring in the (gate voltage-magnetic field) data, at least in the integer Quantum Hall regime, may well be encoded in one visibility diagram, namely the odd one. We have also indicated an experimental way to test its possible relevance to the fractional QHE. Furthermore, we have conjectured that for a given transition $`\nu =n_1\nu =n_2,n_{1,2}Q`$ the directions of the two coexisting families of straight lines involving the extrema of the fluctuations ( observed in for the transitions $`01`$, $`12`$, $`23`$, $`34`$) are given by the directions defined by the corresponding stripes of the odd diagram involved in that transition. REFERENCES 1) The Quantum Hall Effect, 2nd ed., R.E. Prange and S.M. Girvin eds. (Springer-Verlag, New-York) 1990. See also Perspectives in Quantum Hall Effect, S.D. Sorma and A. Pinczuk eds. (Wiley, New-York) 1997. 2) K. von Klitzing, G. Dorda and M. Pepper, Phys. Rev. Lett. 45 (1980) 494. 3) D.C. Tsui, H.L. Störmer and A.C. Gossard, Phys. Rev. Lett. 48 (1982) 1559. 4) R.B. Laughlin, Phys. Rev. Lett. 50 (1983) 1385; F.D. Haldane, Phys. Rev. Lett. 51 (1983) 605; B.I. Halperin, Phys. Rev. Lett. 52 (1984) 1583. 5) C.A. Lütken and G.G. Ross, Phys. Rev. B45 (1992) 11837, Phys. Rev. B48 (1993) 2500. 6a) C.A. Lütken, Nucl. Phys. B396 (1993) 670; C.P. Burgess and C.A. Lütken, On the implication of discrete symmetries for the $`\beta `$-function of Quantum Hall System, cond-mat/9812396 and references therein. See also E. Fradkin and S. Kivelson, Nucl. Phys. B474 (1996) 543. 6b) B.P. Dolan, J. Phys. A: Math. Gen. 32 (1999) L243; B.P. Dolan, Modular invariance, universality and crossover in the QHE, cond-mat/9809294. 7a) Y. Georgelin and J.C. Wallet, Phys. Lett. A224 (1997) 303. 7b) Y. Georgelin, T. Masson and J.C. Wallet, J. Phys. A:Math. Gen. 30 (1997) 5065. 8) Y. Georgelin, T. Masson and J.C. Wallet, $`\mathrm{\Gamma }(2)`$ modular symmetry, renormalization group flow and the Quantum Hall Effect, cond-mat/9906193. 9) S. Kivelson, D.H. Lee and S.C. Zhang, Phys. Rev. B46 (1992) 2223. 10) D.H. Cobden, C.H. Barnes and C.J.B. Ford, Quantum Hall Fluctuations and evidence for charging in the Quantum Hall Effect, cond-mat/9902154. 11) S.V. Kravchenko, W. Mason and J.E. Furneaux, Phys. Rev. Lett. 75 (1995) 910. 12) S.V. Kravchenko et al., Phys. Rev. B51 (1995) 7038. 13) M. Hilke et al., experimental phase diagram of the integer quantized Hall effect, cond-mat/9906212. 14) D.N. Sheng and Z.Y. Weng, phase diagram of the integer Quantum Hall Effect, cond-mat/9906261. 15) See in D. Mumford, Tata Lectures on Theta functions, vol. I-III, (Birkhauser, Boston, Basel, Stuttgart) 1983.
no-problem/9907/math9907072.html	ar5iv	text	# Untitled Document BRUNNIAN BRAIDS AND SOME OF THEIR GENERALIZATIONS Theodore Stanford † Research supported in part by the Naval Academy Research Council Mathematics Department United States Naval Academy 572C Holloway Road Annapolis, MD 21402 stanford@nadn.navy.mil Abstract. We use a variation on the commutator collection process to characterize those pure braids which become trivial when any one strand is deleted, or, more generally, those pure braids which become trivial when all the strands in any one of a list of sets of strands is deleted. 0. INTRODUCTION A Brunnian link is a tame link of $`n`$ closed curves in $`R^3`$ such that deleting any one of the curves results in a trivial link of $`n1`$ components. (See Rolfsen .) By analogy, we shall call a braid of $`n`$ strands a Brunnian braid if deleting any one of the strands produces a trivial braid of $`n1`$ strands. A Brunnian braid on more than two strands must clearly be a pure braid, so we confine our attention to the pure braid group $`P_nB_n`$. Though the idea of “deleting a strand” is topological, we take here a purely algebraic approach. Brunnian braids were considered by Levinson under the name “decomposable braids”. More generally, he considers “$`k`$-decomposable braids”, which become trivial when any $`k`$ strands are deleted. In he gives a geometric characterization of such braids, and in he gives algebraic characterizations in the cases $`(n,k)=(3,1),(4,1),(4,2)`$. We shall generalize these results below. It is not hard to see that the set of $`n`$-strand Brunnian braids is a free normal subgroup of $`P_n`$. In the Kourovka notebook , this was called the subgroup of “smooth” braids, and the problem was posed to give a set of free generators for smooth braids with a given number of strands. Johnson used the Hall commutator collection process to give a set of generators modulo any group of the lower central series. There has appeared in a conference proceedings an abstract of a solution to the Kourovka problem, after which the problem was taken from the notebook. I have been unable to locate, however, a paper which follows through on the abstract. In this paper, we give sets of generators for the Brunnian subgroups, the $`k`$-decomposable subgroups, and for a somewhat more general class of subgroups. Our generating sets are not minimal, so we do not address the Kourovka question. Our method is a finite variation on the Hall commutator collection process. A similar method was used in , in order to show that an element of an arbitrary group $`G`$ is “$`n`$-trivial” if and only if it is in the $`n`$th group of the lower central series of $`G`$. We give our characterization of the Brunnian subgroup of $`P_n`$ (Corollary 2.3) in terms of monic commutators (Definition 1.3). We set $`[x,y]=x^1y^1xy`$ for any group elements $`x,y`$. Here are some monic commutators: $`[p_{1,3},p_{2,4}]P_4`$, $`[[p_{1,2},p_{1,3}],p_{1,4}]P_4`$, and $`[[[p_{1,2},p_{2,3}],p_{3,4}],[p_{4,5},p_{5,6}]]P_6`$. It is not hard to see that these particular examples are all Brunnian braids, since deleting any one strand trivializes at least one entry in the iterated commutator, which of course trivializes the whole commutator. We define (Definition 1.4 and Proposition 1.6) the support of a monic commutator to be the strands whose indices appear somewhere in the commutator. Thus the support of $`[p_{1,2},p_{1,4}]P_4`$ is $`\{1,2,4\}`$. Deleting strands $`1`$,$`2`$, or $`4`$ from this commutator trivializes it, but deleting strand $`3`$ does nothing. We characterize the Brunnian subgroup of $`P_n`$ as being generated by all monic commutators whose support is the whole set of strands $`\{1,2,3\mathrm{}n\}`$. As noted below, the Brunnian subgroup is not finitely-generated (for $`n>2`$), so its list of generators necessarily includes commutators of with an arbitrary number of brackets. More generally, consider any finite collection of subsets $`S_1,S_2,\mathrm{}S_m\{1,2,\mathrm{}n\}`$. Our main theorem states that the subgroup of braids which become trivial when the strands in any one $`S_i`$ are cut is generated by the set of monic commutators whose support intersects each $`S_i`$ nontrivially. When $`S_i=\{i\}`$ we get the characterization of Brunnian braids, and when the $`S_i`$ consist of all subsets with $`k`$ elements, we get a characterization of $`k`$-decomposable braids. Appreciation. I would like to than Joan Birman, Tony Gaglione, and Mark Meyerson for helpful remarks and conversations. 1. BASIC IDEAS A standard reference on braids is Birman . Recall that the pure braid group $`P_n`$ is generated by $`p_{a,b}`$ for $`1a<bn`$, where $`p_{a,b}`$ is the braid which links strand $`a`$ and strand $`b`$ in front of the other strands. Artin’s semidirect product decomposition may be used to find a finite presentation for $`P_n`$. Here is one version: A. $`p_{a,b}p_{a,c}p_{b,c}=p_{a,c}p_{b,c}p_{a,b}=p_{b,c}p_{a,b}p_{a,c}`$ for all $`1a<b<cn`$ B. $`p_{a,b}p_{c,d}=p_{c,d}p_{a,b}`$ and $`p_{a,d}p_{b,c}=p_{b,c}p_{a,d}`$ for all $`1a<b<c<dn`$ C. $`p_{a,c}p_{b,c}^1p_{b,d}p_{b,c}=p_{b,c}^1p_{b,d}p_{b,c}p_{a,c}`$ for all $`1a<b<c<dn`$ Each of the above relations corresponds to a “geometrically obvious” commutation relation. For the (B) relations this is immediate, and examples for the (A) and (C) relations are shown in the figure below. $$p_{1,2}(p_{2,3}p_{1,3})=(p_{2,3}p_{1,3})p_{1,2}$$ $$p_{1,3}(p_{2,3}^1p_{2,4}p_{2,3})=(p_{2,3}^1p_{2,4}p_{2,3})p_{1,3}$$ Fix the positive integer $`n`$, and let $`N=\{1,2,\mathrm{}n\}`$. Suppose $`Sn`$. Denote by $`\overline{S}`$ the complement of $`S`$ in $`N`$. Let $`P_S`$ be subgroup of $`P_N=P_n`$ generated by $`p_{a,b}`$ with $`aS`$ and $`bS`$, and let $`Q_S`$ be the subgroup generated by $`p_{a,b}`$ such that either $`aS`$ or $`bS`$. The standard semidirect product decomposition states that there is a retraction homomorphism from $`P_N`$ to $`P_N`$ whose image is $`P_{N1}`$ and whose kernel is $`Q_{\{n\}}`$. This homomorphism is accomplished geometrically by cutting or trivializing the $`n`$th string of a pure braid. If instead we cut all the strings in $`SN`$, then we still get a retraction map and a semidirect product decomposition. Proposition 1.1. For each $`SN`$, there exists a retraction homomorphism $`\varphi _S:P_nP_n`$ whose image is $`P_{\overline{S}}`$ and whose kernel is $`Q_S`$. Moreover, $`\varphi _{S_1}\varphi _{S_2}=\varphi _{S_1S_2}`$ for any $`S_1,S_2N`$. Proof: Define $`\varphi _S(p_{a,b})=p_{a,b}`$ if $`aS`$ and $`bS`$, and $`\varphi _S(p_{a,b})=1`$ otherwise. Then check using the presentation above that this defines a homomorphism with the required properties. Now, given a set of subsets $`S_1,S_2,S_3,\mathrm{}S_mn`$, we are interested in characterizing $`_{i=1}^mQ_{S_i}`$. First note it follows from Proposition 1.1 that $`P_{S_1}P_{S_2}=P_{S_1S_2}`$. Inductively, we have Proposition 1.2. Let $`S_1,S_2,\mathrm{}S_mN`$. Then $`_{i=1}^mP_{S_i}=P_{_{i=1}^mS_i}`$. By way of contrast, the subgroup $`Q=_{i=1}^mQ_{S_i}`$ is not in general finitely-generated. For example, consider $`Q_{\{2\}}`$ and $`Q_{\{3\}}`$ in $`P_3`$. Both of these subgroups are free on two generators, so $`Q_{\{2\}}Q_{\{3\}}`$ is also free. The retraction $`\varphi _{\{3\}}:P_3P_3`$ restricts to a retraction $`Q_{\{2\}}Q_{\{2\}}`$, whose kernel is $`Q_{\{2\}}Q_{\{3\}}`$ and whose image is the infinite cyclic group $`p_{1,2}`$. Thus $`Q_{\{2\}}Q_{\{3\}}`$ has infinite index in $`Q_{\{2\}}`$, and therefore $`Q_{\{2\}}Q_{\{3\}}`$ is not finitely-generated. We shall show that $`_{i=1}^mQ_{S_i}`$ is generated by a subset of the monic commutators of $`P_n`$. Definition 1.3. A monic commutator is an element of $`P_n`$ defined recursively as follows A. $`p_{a,b}`$ and $`p_{a,b}^1`$ are monic commutators for all $`1a<bn`$. B. If $`x`$ and $`y`$ are monic commutators, and $`[x,y]1`$, then $`[x,y]`$ is a monic commutator. Definition 1.4. If $`xP_n`$, then the support $`\sigma (x)`$ is the intersection of all $`Sn`$ such that $`xP_S`$. We have $`\sigma (p_{a,b}^{\pm 1})=\{a,b\}`$. Two other things are also immediate. First, $`\sigma ([x,y])\sigma (x)\sigma (y)`$. Second, if $`\sigma (x)S=\mathrm{}`$, then $`\varphi _S(x)=x`$. For monic commutators, we can say more: Proposition 1.5. If $`x`$ is a monic commutator in $`P_n`$, then $`\varphi _S(x)=1`$ if and only $`\sigma (x)S\mathrm{}`$. Proof: If $`\sigma (x)S=\mathrm{}`$, then $`\varphi _S(x)=x1`$ by definition. For the converse, it suffices by Proposition 1.1 to show that for a monic commutator $`x`$, $`i\sigma (x)`$ implies that $`\varphi _{\{i\}}(x)=1`$. This is certainly true when $`x=p_{a,b}^{\pm 1}`$. Suppose that it is true for two monic commutators $`x`$ and $`y`$. If $`i\sigma ([x,y])`$, then $`i\sigma (x)`$ or $`i\sigma (y)`$. Then $`\varphi _{\{i\}}(x)=1`$ or $`\varphi _{\{i\}}(y)=1`$, and in either case $`\varphi _{\{i\}}([x,y])=1`$. Proposition 1.6. If $`x`$, $`y`$, and $`[x,y]`$ are monic commutators, then $`\sigma ([x,y])=\sigma (x)\sigma (y)`$. Proof: If $`i\sigma ([x,y])`$ then $`\varphi _{\{i\}}([x,y])1`$, and therefore $`i\sigma (x)\sigma (y)`$. 2. THE THEOREM Theorem 2.1. For $`1im`$, let $`S_iN`$. Then $`_{i=1}^mQ_{S_i}`$ is generated by the set of monic commutators $`x`$ such that $`\sigma (x)S_i\mathrm{}`$ for all $`1im`$. We obtain as corollaries characterizations of $`k`$-decomposable braids and of Brunnian braids, generalizing results of Levinson . Corollary 2.2. The normal subgroup of $`k`$-decomposable $`n`$-strand braids is generated in $`P_n`$ by all monic commutators whose support has cardinality at least $`nk+1`$. Corollary 2.3. The normal subgroup of all Brunnian braids in $`P_n`$ is generated by all monic commutators whose support is $`N=\{1,2,\mathrm{}n\}`$. Corollary 2.4. The subgroup of $`(n2)`$-decomposable braids is the commutator subgroup of $`P_n`$. Corollary 2.4 may also be proved directly by observing that if $`\|S\|=n2`$ then the image of $`\varphi _S`$ is a two-strand pure braid group, isomorphic to $`Z`$. More specifically, if $`S=N\{i,j\}`$, then $`\varphi _S(x)`$ can be taken to be the integer which measures the linking number of strands $`i`$ and $`j`$ in the braid $`x`$. Then $`x`$ is in the commutator subgroup of $`P_n`$ if and only if all these linking numbers vanish. Proof of Theorem 2.1: If $`x`$ is a monic commutator which intersects each $`S_i`$ nontrivially, then $`\varphi _{S_i}(x)=1`$ for all $`i`$, and therefore $`x_{i=1}^mQ_{S_i}`$. We need to show that any element in $`_{i=1}^mQ_{S_i}`$ can be written as a product of such commutators. First we will describe a type of commutator collection process, where any element of $`P_n`$ may be written as a product of monic commutators such that the monic commutators with common support are grouped together. When we apply this to an element of $`_{i=1}^mQ_{S_i}`$, we will find that all the monic commutators whose support misses one of the $`S_i`$ will drop out. Let $`qP_n`$ be given as a word in the $`p_{i,j}^{\pm 1}`$. Fix a total ordering $`T_1,T_2,\mathrm{}T_{2^n}`$ of the subsets of $`N`$, arbitrary except that we require that if $`T_iT_j`$ then $`i<j`$. For notational convenience we shall write $`T_i<T_j`$ if $`i<j`$. We claim that there exist elements $`q_1,q_2,\mathrm{}q_{2^k}P_n`$ such that $`q=q_1q_2\mathrm{}q_{2^k}`$, and such that each $`q_i`$ a product of monic commutators of support $`T_i`$, where an empty product is taken to be $`1P_n`$. Suppose inductively that we have written $`q=q_1q_2\mathrm{}q_rs`$, where $`q_i`$ is a product of monic commutators of support $`T_i`$ for $`1ir`$, and $`s`$ is a product of monic commutators, each of support $`T_r`$. We want to find all the monic commutators in $`s`$ of support $`T_r`$, and move them back into $`q_r`$. We may do this by inserting commutators, each of which has support $`>T_r`$. More precisely, let $`s=x_1x_2\mathrm{}x_tyz`$, where each $`x_i`$ is a monic commutator of support $`>T_r`$, $`y`$ is monic commutator of of support $`T_r`$, and $`z`$ is a product of monic commutators of support $`T_r`$. We may then write $`s=yx_1[y,x_1]x_2[y,x_2]x_3\mathrm{}x_t[y,x_t]`$, and then move $`y`$ into $`q_r`$. Even though this increases the number of monic commutators in $`s`$, it decreases by one the number of them that have support $`T_r`$. Thus it is possible to continue until all the monic commutators with support $`T_r`$ are contained in $`q_r`$, and inductively we can continue until $`q=q_1q_2\mathrm{}q_{2^k}`$. Now let $`q_{i=1}^mQ_{S_i}`$ and $`q=q_1q_2\mathrm{}q_{2^k}`$ as above. Fix an arbitrary $`i`$ between $`1`$ and $`n`$. We need to show that if $`T_jS_i=\mathrm{}`$, then $`q_j=1`$. This is done by induction on $`\|T_j\|`$. Suppose that $`q_j^{}=1`$ for all $`T_j^{}T_j`$. By Proposition 1.1, $`\varphi _{\overline{T}_j}`$ factors through $`\varphi _{S_i}`$, so $$1=\varphi _{\overline{T}_j}(q)=\varphi _{\overline{T}_j}(q_1)\varphi _{\overline{T}_j}(q_2)\mathrm{}\varphi _{\overline{T}_j}(q_{2^n})$$ and the only possible nontrivial element of this product is $`\varphi _{\overline{T}_j}(q_j)`$, which must then be trivial as well. References. E. Artin. Theory of braids. Annals of Mathematics (2) 48 (1947), 101–126. J.S. Birman. “Braids, Links and Mapping Class Groups.” Annals of Mathematics Studies 82. Princeton University Press, 1975. G.G. Gurzo. The group of smooth braids. 16th All-Union Algebra Conference, Abstract II, 39–40, Leningrad 1981. Magnus, Karrass, and Solitar. “Combinatorial Group Theory.” Dover Publications, New York, 1976. D.L. Johnson. Towards a characterization of smooth braids. Mathematical Proceedings of the Cambridge Philosophical Society 92 (1982) 425–427. “The Kourovka notebook of unsolved problems in group theory.” 7th edition, Novosibirsk, 1980. H. W. Levinson. Decomposable braids and linkages. Transactions of the American Mathematical Society 178 (1973), 111–126. H. W. Levinson. Decomposable braids as subgroups of braid groups. Transactions of the American Mathematical Society 202 (1975), 51–55. Magnus, Karrass, and Solitar. “Combinatorial group theory. Presentations of groups in terms of generators and relations.” Second revised edition. Dover Publications, New York, 1976. K. Y. Ng and T. Stanford. On Gusarov’s groups of knots. Mathematical Proceedings of the Cambridge Philosophical Society 126 (1999), 63–76. D. Rolfsen. “Knots and Links” Mathematics Lecture Series 7. Publish or Perish, Wilmington, DE, 1976.
no-problem/9907/astro-ph9907092.html	ar5iv	text	# 1 Introduction ## 1 Introduction Low-ionization nuclear emission line regions, LINERs, are characterized by their optical emission line spectrum which shows a lower degree of ionization than active galactic nuclei (AGN). Their major power source and line excitation mechanism have been a subject of lively debate ever since their discovery. If powered by accretion, they probably represent the low-luminosity end of the quasar phenomenon and their presence has relevance to, e.g., the evolution of quasars, the faint end of the Seyfert luminosity function, the soft X-ray background, and the presence of SMBHs in nearby galaxies. A detailed study of the LINER phenomenon is thus very important. X-rays are a powerful tool to investigate the presence of an active nucleus within the LINER galaxies via X-ray variability and luminosity, and to explore the physical properties of LINERs in general. An analysis of ASCA spectra of several low-luminosity AGN including 5 LINERs (Ptak et al. 1999) revealed the presence of a hard powerlaw–like emission component plus soft thermal emission. Here, we present a study of 13 LINERs observed in the soft X-ray spectral region (0.1–2.4 keV) with the instruments on-board ROSAT. All of them were targets during the ROSAT all-sky survey, some of them detected for the first time in X-rays, and for 8 of them additional pointed PSPC and/or HRI observations are available. Luminosities given below are based on $`H_0`$=75 km/s/Mpc; distances of very nearby galaxies were taken from Tully’s (1988) catalog. ## 2 Results of the X-ray analysis The results can be summarized as follows (see Komossa et al. 1999 for the full analysis): * X-ray luminosities range between $`\mathrm{log}L_\mathrm{x}=37.7`$ (NGC 404) and 40.8 (NGC 4450). The ratios $`L_\mathrm{x}`$/$`L_\mathrm{B}`$, when compared to early-type galaxies, are located in the intermediate region of the $`L_\mathrm{x}L_\mathrm{B}`$ diagram (Fig. 1; see Komossa et al. 1999 for an explicit comparison with elliptical galaxies) and similar emission mechanisms may contribute to the observed X-ray luminosities. * Whereas the bulk of the X-ray emission from the LINERs is consistent with arising from a point source there is some extended emission seen at weak emission levels in some sources. * We do not detect any short timescale (hours to weeks) temporal variability. This is consistent with the earlier suggestion that LINERs may accrete in the advection-dominated mode. Only one source, NGC 2768, is slightly variable on a timescale of months. * The X-ray spectra of the LINER galaxies show some spectral variety. Half of them are best described by a powerlaw of photon index $`\mathrm{\Gamma }_\mathrm{x}2`$ whereas the other ones are best matched by a Raymond-Smith model with $`kT`$ 0.6 keV and heavily depleted metal abundances. Below, we discuss possible explanations for the unusually low inferred abundances. ## 3 X-ray emission mechanisms Some sources are best described by a single AGN-like powerlaw with X-ray luminosity above that expected from discrete stellar sources. These spectra most likely indicate the presence of low-luminosity AGNs in the centers of these LINER galaxies. We concentrate here on possible explanations for the very low metal abundances (1/10 – 1/20 $`\times `$ solar) formally derived for the other sources when a Raymond-Smith model is fit. Presence of a second hard component. One way to avoid heavily depleted abundances is to invoke a second hard X-ray spectral component (see the discussion in Buote & Fabian 1998, Komossa & Schulz 1998, and Komossa et al. 1998 for details). The second component could be of powerlaw shape and originate from a low-luminosity AGN (LLAGN), or it could be hot thermal emission from an extended component. The two possibilities cannot be distinguished by spectral fits to ROSAT data. Alternatively, the emitting gas might show a range of temperatures. Strickland & Stevens (1998) carefully discussed this possibility by simulating an expanding superbubble under the assumptions of axial symmetry and collisional ionization equilibrium (CIE). They find that by fitting a single-temperature model to ROSAT data instead of a multi-temperature one, the true metal abundances are heavily underestimated. However, nature may be even more complicated. If the plasma is not in CIE, the density and temperature structure of the emitting region cannot be reproduced by a finite one-parameter family of curves with merely differing temperatures, since each ionization stage of each ion would have its ‘own temperature’. Non-equilibrium effects in a hot extended plasma component. When applying Raymond-Smith or similar codes, the assumption of CIE has to be kept in mind. In optically thin plasmas such an assumption can never be strictly fulfilled, since collisional ionization and radiative recombination are not statistically inverse processes, and therefore can only hold in an approximate sense. If in addition the plasma is in motion, e.g. due to its origin in a superbubble or galactic wind, the adiabatic cooling timescale can be shorter than any timescale associated with atomic processes. The resulting spectrum may then be dominated by a strong contribution from delayed recombination of highly ionized species (Breitschwerdt & Schmutzler 1994, 1999). The kinetic temperature of the electrons in the X-ray emitting plasma is typically much lower than in a CIE plasma, and collisional ionization contributes mainly to very soft X-ray or UV emission. Thus a two- or multitemperature plasma can be mimicked, as far as the spectral resolution of current X-ray detectors is concerned. Sub-solar ‘Raymond-Smith abundances’ can also be obtained with non-CIE spectra from a galactic outflow, folded through the ROSAT detector response matrix, as we have verified by preliminary calculations of non-CIE models. In conclusion, both effects (the presence of a LLAGN and non-equilibrium plasma) are likely to play a role to some extent. The importance of the latter also depends on the origin of the gas. The galaxies of the present sample are mostly type S0 and SA. Whereas in ellipticals the extended gas is ‘old’ and the standard equilibrium assumptions may be justified, in spirals a contribution from a young starburst that drives superwinds is well possible. In that case the gas is likely to be far from equilibrium (see the discussion in Breitschwerdt et al. 1999 and these proceedings). Given the spectral variety of LINERs with contributions from several emission components, future studies of both, spectra of individual objects as well as larger samples will certainly give further insight into the LINER phenomenon which provides an important link between active and ‘normal’ galaxies.
no-problem/9907/astro-ph9907356.html	ar5iv	text	# An investigation of the large-scale variability of the apparently single Wolf-Rayet star WR 1 ## 1 Introduction It is now fairly well-established that apparently single Wolf-Rayet (WR) stars may display two distinct (but probably non mutually exclusive) spectroscopic patterns of variability: (a) small-scale emission features systematically moving from the line center to the line wings on an hourly timescale (e.g., Lépine (1998)); (b) dramatically larger line-profile deformations operating on a much longer basis ($``$ days, e.g., Smith & Willis (1994)). Although the first phenomenon, observed in most (if not all) WR stars, is believed to be the consequence of the fragmented, possibly turbulent nature of the outflow, the origin of the latter type of variability is still very much elusive. Remarkable in this respect, is the existence of a well-established (although strongly epoch-dependent) large-scale, cyclical pattern of variability in the two apparently single WR stars WR 6 ($`𝒫`$ = 3.763 $`\pm `$ 0.002 d; Firmani et al. (1980)) and WR 134 ($`𝒫`$ = 2.27 $`\pm `$ 0.04 d; McCandliss et al. (1994); Morel et al. (1999)). A major observational effort has been directed on establishing the true nature of these peculiar objects, i.e., whether this cyclical variability is induced by an orbiting unseen (collapsed?) companion or by the rotational modulation of large-scale wind structures (e.g., Vreux et al. (1992); Morel (1998); Harries et al. (1999), and references therein). Although the exact nature of these stars has yet to be unambiguously settled, these studies reveal that rotational modulation constitutes an attractive alternative to the binary hypothesis, especially considering the recent recognition that some O stars (the progenitors of WR stars) might possess such azimuthally structured outflows (Fullerton et al. (1997); Kaper et al. (1997)). A prime target for further investigations is the seldom-studied, apparently single WN 5 star WR 1 (HD 4004) that has recently been shown to present a spectral pattern of variability very similar to that of WR 6 (Niedzielski (1995), 1996a, b; Wessolowski & Niedzielski 1996; Niedzielski (1998)). Strong line-profile variability was observed, as well as apparently cyclical (according to $`𝒫`$ $``$ 2.667 days) variations in the EWs of He ii $`\lambda `$4686 and He ii $`\lambda `$5412 (Niedzielski 1996a ). The first claim of periodic variability in WR 1 with $`𝒫`$ $``$ 7.7 days was made by Lamontagne (1983) from an analysis of the radial velocity variations of He ii $`\lambda `$4686. The first photometric monitoring of this object has shown WR 1 to be variable, with an indication of a 6.1 day period (Moffat & Shara 1986). Recently, Marchenko et al. (1998a) discussed Hipparcos broadband photometric data which revealed that WR 1 also displays relatively long-term photometric variations, with a marginal evidence for a $`𝒫`$ = 11.68 $`\pm `$ 0.14 day periodicity. As can be seen, controversy persits in the literature concerning the possible cyclical nature of the variations in WR 1. We present in this paper the results of spectroscopic and photometric monitoring of WR 1 carried out in 1995 and 1996 aiming at shedding some light on this issue. ## 2 Observations and Reduction Procedure ### 2.1 Photometry The photometric variability of WR 1 has been investigated during the interval 1996 September 18–October 5 by use of the single channel photometer Cuentapulsos on the 0.84 m telescope of the Observatorio Astronómico Nacional at San Pedro Mártir (Mexico). Two additional objects were monitored during this observing run, namely, WR 3 and WR 153. The nights were generally clear. WR 1 was observed through a narrowband $`v`$ filter centered on 5140 Å (FWHM = 90 Å). This filter samples a continuum-dominated region of the WR spectrum. We applied the following sequence of 60 s integration through a 25″ diaphragm: sky, C2, C1, WR, C1, WR, C1, C2, sky. The same nearby comparison stars as used by Moffat & Shara (1986) have been chosen. These comparison stars are similar in terms of their magnitude and colour to WR 1: $`\mathrm{\Delta }B`$ (WR – C1) = – 0.17, $`\mathrm{\Delta }`$\[$`B`$$`V`$\] (WR – C1) = – 0.17, $`\mathrm{\Delta }B`$ (WR – C2) = – 0.26, $`\mathrm{\Delta }`$\[$`B`$$`V`$\] (WR – C2) = – 0.19. An extinction coefficient $`k_v`$ = 0.20 was used throughout the data reduction. The scatter in the (C2 $``$ C1) data for the whole dataset amounts to $`\sigma `$ = 4.7 mmag. The differential magnitudes quoted in Table 1 are averaged over two consecutive cycles typically separated by about 20 minutes. ### 2.2 Spectroscopy #### 2.2.1 Long-slit and Reticon Spectra Long-slit spectra of WR 1 have been obtained during various campaigns at the Observatoire du Mont Mégantic and Dominion Astrophysical Observatory (Canada) in 1995 October and 1996 September. Reticon spectra were also obtained at DAO in 1996 November. The 1996 campaign at the Observatoire du Mont Mégantic was coordinated to support the photometric campaign described above. Table 2 lists the mode of observation, the dates of the spectroscopic observations, the interval of the observations in heliocentric Julian dates, the observatory name, the number of CCD spectra obtained, the selected spectral domain, the reciprocal dispersion of the spectra, and the typical signal-to-noise ratio (S/N) in the continuum. The spectra were reduced using the IRAF<sup>1</sup><sup>1</sup>1IRAF is distributed by the National Optical Astronomy Observatories, operated by the Association of Universities for Research in Astronomy, Inc., under cooperative agreement with the National Science Foundation. data reduction packages. The bias and sky subtraction, flat-field division, removal of cosmic ray events, extraction of the spectra, and wavelength calibration were carried out in the usual way. Spectra of calibration lamps were taken immediately before and after the stellar exposure. The stellar spectra were subsequently continuum normalized by fitting a low-order Legendre polynomial to carefully selected line-free regions. In order to minimize the spurious velocity shifts induced by an inevitably imperfect wavelength calibration, the spectra were coaligned in velocity space by using the interstellar doublet Na i $`\lambda `$$`\lambda `$5890, 5896 as fiducial marks. When not available, the doublet Ca ii $`\lambda `$$`\lambda `$3934, 3968 or the diffuse interstellar band at 4501 Å were used. A trend for a systematic shift of the zero point of the wavelength scale has been corrected by redshifting most of the spectra by an average value of 35 km s<sup>-1</sup>. #### 2.2.2 Echelle Spectra Echelle spectra have been obtained during the period 1996 September 16–19 with the Echelle spectrograph Espresso (Levine & Chakrabarty (1995)) on the 2.1 m telescope of the San Pedro Mártir Observatory. The UCL camera and a 1024 $`\times `$ 1024 coated CCD-Tek chip have been used. The selected grating (300 lines mm<sup>-1</sup>) yields a reciprocal dispersion of 0.16 and 0.23 Å pixel<sup>-1</sup> at H$`\gamma `$ and H$`\alpha `$, respectively. The spectra cover 27 orders and span the spectral range 3720-6900 Å. The reduction procedure (bias subtraction, division by a normalized flat field, removing of scattered light, extraction of the orders) was carried out using the reduction tasks in the IRAF package echelle. Comparison spectra of Th–Ar lamps have been used for the wavelength calibration. The typical accuracy of the wavelength calibration can be judged by the dispersion in the heliocentric radial velocities of the interstellar line Na i $`\lambda `$5890: $`\sigma `$ $``$ 2 km s<sup>-1</sup>. The instrumental response has been removed by fitting a low-order Legendre polynomial to the continuum sections of each order of an O-star spectrum, and then dividing the corresponding order of the WR 1 spectra by this polynomial. The resultant, almost flat orders were then combined using the IRAF task scomb. This procedure proved to be satisfactory, except when the overlapping regions fall in a spectral domain with very steep intensity gradients, as in the blue wing of He ii $`\lambda `$4686 (see Fig.2). The combined spectra were then subdivided into spectral regions roughly corresponding to those of the long-slit spectra discussed above (§2.2.1). For consistency purposes, these spectra have been continuum normalized by fitting a low-order Legendre polynomial to the same continuum sections that selected for the long-slit spectra. ## 3 Results ### 3.1 Photometric Variations The light curve of WR 1 is plotted as a function of the heliocentric Julian date of observation in Figure 1. The main feature of this light curve is the gradual increase of the stellar continuum flux amounting to $`\mathrm{\Delta }v`$ $``$ 0.09 mag beginning at HJD 2,450,346, followed by its decline on about the same timescale after HJD 2,450,350. During the last 11 nights (after HJD 2,450,353), only marginal evidence for variability is found, with $`\sigma `$(WR $``$ C)/$`\sigma `$(C2 $``$ C1) $``$ 1.3. This can be compared with a value of 1.9 derived by Moffat & Shara (1986) on the basis of 14 nights of broadband $`B`$ observations in 1984. Noticeable is the striking lack of recurrence in the continuum flux data. A periodicity search in the (WR $``$ C1) and (WR $``$ C2) data acquired after HJD 2,450,353 yields no evidence for any periodic signals in the range 6–8 days that would possibly account for the variations observed before this date. The highest peaks in the power spectra of these two datasets (after correction by the CLEAN algorithm; see Roberts et al. (1987)) appear (with considerable uncertainty) for periods of about 4 days. However, the significance of these periodic signals is very low and they must thus be regarded as spurious. The periods proposed by Niedzielski (1996a; 2.667 days) and Marchenko et al. (1998a; 11.68 days) are inconsistent with the global light-curve morphology. On the other hand, the periods claimed by Lamontagne (1983; 7.7 days) and Moffat & Shara (1986; 6.1 days) are only consistent with part of the data (before HJD 2,450,353). We will come back to this point later (§4). ### 3.2 An Overview of the Line-profile Variations In the following, we will generally illustrate the line-profile variations of WR 1 by means of the strong He ii $`\lambda `$4686 line as its variations are qualitatively similar to those affecting the other He ii features (see §3.5; Niedzielski 1996a ). One of the most outstanding features of the line-profile variability is the relatively low level of variability observed within one night.<sup>2</sup><sup>2</sup>2Note that although probably present in WR 1, the small-scale emission-excess features travelling on an hourly timescale on the top of the line profiles (probably induced in WR stars by outwardly moving shocked and/or turbulent material; Lépine & Moffat (1999)) cannot be studied here, owing to the insufficient S/N achieved for the Echelle spectra. Such a gradual pattern of variability has been observed whenever several spectra have been obtained during a single night of observation. This phenomenon is illustrated in Figure 2 where we show a superposition of the 21 He ii $`\lambda `$4686 line profiles obtained on the night of 1996 September 18 (around HJD 2,450,344.873). In sharp contrast, however, noticeable night-to-night variations are observed. This is illustrated in Figure 3 which shows a montage of the nightly mean spectra obtained in 1995 October (left panel) and 1996 September (right panel) for the spectral region encompassing He ii $`\lambda `$4686 (it has to be kept in mind that because of the paucity of the photometric data, the strength of the spectral lines in these continuum-normalized spectra has not been corrected for the varying continuum flux level). ### 3.3 Temporal Variance Spectrum Analysis The Temporal Variance Spectrum (TVS; Fullerton et al. (1996)) has been used to assess the level of spectral variability as a function of wavelength. The (square root of the) TVS, giving the typical “size” of the deviations from a template-weighted mean spectrum (expressed in percentage of the continuum flux) is shown, along with the 99.0 % confidence level for variability, in Figure 4. As can be seen, highly significant variability affects the emission lines, with the notable exception of N v $`\lambda `$ 4945 for which only little evidence for variability is found. Many peaks (i.e., locations of “preferential” variability) can be distinguished in the TVS. The locations of these peaks (in terms of the projected velocity referred to the laboratory rest wavelength of the spectral feature in question) are quoted in Table 3. This investigation shows that: (a) the variability often extends to velocities comparable to the wind terminal velocity ($`v_{\mathrm{}}`$ $``$ 2135 km s<sup>-1</sup>; Rochowicz & Niedzielski (1995)), e.g., $`v`$ $``$ – 1790 km s<sup>-1</sup> for the relatively unblended line He ii $`\lambda `$4859; (b) the TVS structure — made up of several subpeaks — presents some similarities for the various He ii features. However, the velocities quoted in Table 3 are certainly entached of considerable uncertainties (e.g., because of blending of TVS subpeaks) making difficult a clear statement as to whether the variability takes place at the same characteristic velocities in different lines. ### 3.4 The P Cygni Profile Variability The TVS analysis also reveals substantial variability at the location of the P Cygni absorption component of N v $`\lambda `$4604 (Fig.4). This mainly results from the transition of the N v $`\lambda `$4604 feature from a pure emission line-profile in 1995 October to a P Cygni line-profile in 1996 September (Fig.5). No clear variations in the strength of this absorption trough on a daily timescale are observed. An enhanced peak in the TVS can also be found at the location where the carbon triplet C iv $`\lambda `$5806 and He i $`\lambda `$5876 merge (Fig.4). The large daily changes affecting C iv $`\lambda `$5806 and/or He i $`\lambda `$5876 in October 1995 and September 1996 are illustrated in Figure 6 (this phenomenon is not observed in 1996 November). The high level of variability observed at this particular location is likely due to the superposition of two distinct type of variability: (a) red-wing variability of C iv $`\lambda `$5806 as observed in other spectral features (e.g., He ii $`\lambda `$4686; Fig.4); (b) variations in the strength of the P Cygni absorption component of He i $`\lambda `$5876.<sup>3</sup><sup>3</sup>3Such P Cygni absorption troughs of He i $`\lambda `$5876 are commonly (and more clearly) observed in other WN stars (e.g., Robert (1992)). The latter interpretation is supported by the fact that the projected velocity of the peak in the TVS ($`v`$ $``$ – 1980 km s<sup>-1</sup>, referred to the He i rest laboratory wavelength) matches fairly well the wind terminal velocity ($`v_{\mathrm{}}`$ $``$ 2135 km s<sup>-1</sup>). Considering the relatively modest level of variability affecting the red-wing part of the line profiles in WR 1 (Fig.4), it is likely that changes in the strength of the P Cygni absorption component of the He i feature mostly contribute to the observed changes. ### 3.5 Correlated Line-profile Variations in Different Lines? A possible correlated pattern of variability in two different spectral features has been investigated by calculating the Spearman rank-order correlation matrices (see, e.g., Johns & Basri (1995)), whose elements $`r(i,j)`$ yield the degree of correlation between the line intensity variations at pixels $`i`$ and $`j`$ in each line profile, respectively. In the case of perfectly, positively correlated variations, a matrix unity is obtained. The correlation matrices of He ii $`\lambda `$5412 with He ii $`\lambda `$4686, He ii $`\lambda `$4859, and N v $`\lambda `$4945 are shown in Figure 7 in the form of contour plots, where the lowest contour indicates a significant positive (or negative) correlation at the 99.5 % confidence level (the other spectral lines were not covered enough to be included in the analysis). These matrices are displayed in the projected velocity frame (referred to the line laboratory rest wavelength). A significant positive correlation is generally found between the pattern of variability of He ii $`\lambda `$5412 and He ii $`\lambda `$4686. The same conclusion, however restricted to the velocity range (– 1000, + 1000) km s<sup>-1</sup>, holds for the variations affecting He ii $`\lambda `$5412 and He ii $`\lambda `$4859. In contrast, the (weak) variations of N v $`\lambda `$4945 are apparently not linked to those of He ii $`\lambda `$5412 (the same is true for C iv $`\lambda `$5806 and He i $`\lambda `$5876). Various factors are susceptible to mask a potentially correlated pattern (e.g., noise, blends). Also, due to the stratified nature of WR winds, it is conceivable that (a) two given features present similar, yet time-delayed patterns of variability, and (b) the variability is mainly restricted to the line formation regions of the He ii ions and does not extend where the bulk of the N v $`\lambda `$4945 emission originates. This would also result in an apparent lack of correlation. Overall, the results of this analysis tend to show that the He ii features vary in a fairly similar fashion. The small number of spectral features included in the analysis prevents, however, to draw at this stage more general conclusions. ### 3.6 Centroid, FWHM, EW, and Skewness Variations We have investigated the time-dependent changes in the global line-profile properties by calculating the centroid, FWHM, EW, and skewness of He ii $`\lambda `$4686, the strongest line. The centroid and skewness were calculated as the first moment, and the ratio of the third and the (3/2 power of the) second central moments of the portion of the profile above two in units of the continuum (in order to avoid the contribution of blends). The FWHM was determined by a Gaussian fit to the entire line profile. The EWs were calculated by integrating the line flux in the interval 4643–4780 Å. One can note that the mean EW of He ii $`\lambda `$4686 increased from 313 to 343 Å from 1995 October to 1996 September. This 10 % increase may be explained by an intrinsic EW variability and/or by long-term changes in the stellar continuum flux. As can be seen in Figure 8 (right panel), the data obtained in 1996 September generally show coherent time-dependent variations. This is especially clear for the skewness time series, with the timescale of the variations being of the same order of magnitude as the variations of the continuum flux shown in the uppermost part of Figure 8, namely, in the range 5–7 days (note that, unlike the EW, the skewness is insensitive to changes in the continuum flux level). On the other hand, however, and although this may be induced by the paucity of the data, no clear time-dependent behavior is noticeable in the 1995 October dataset (left panel). ## 4 Comparison with Previous Studies Niedzielski (1996a) reported on a spectroscopic pattern of variability of WR 1 very much reminiscent of the one presented by the apparently single WN 5 star WR 6 that displays phase-locked (although strongly epoch-dependent) spectral variations according to $`𝒫`$ $``$ 3.77 days (Morel et al. (1998), and references therein). The results of our investigation of the spectral variability of WR 1 are broadly consistent with this suggestion. In particular, as general features of the spectroscopic pattern of variability, one can note in both objects the substantial variations of the absorption component of the optical P Cygni profiles (readily observable in N v $`\lambda `$$`\lambda `$4604, 4620 where the absorption trough occasionally disappears; Fig.5) or the coherent time-dependent changes in the global line-profile properties (e.g., skewness; Fig.8). Similarities can also be found with WR 134, another very rare example of single-line WR star displaying cyclical variations (Morel et al. (1999)). Evidence was also presented by Niedzielski (1996a) for cyclical (according to $`𝒫`$ $``$ 2.667 days), correlated changes in the EWs of He ii $`\lambda `$4686 and He ii $`\lambda `$5412. The data presented in the present paper do not, however, support a claim of such periodicity. In particular, a period search in our EW data (but also in our centroid, FWHM, and skewness data; Fig.8) yields no significant signal at the expected frequency. Also, this 2.667 day period can hardly account for our global light-curve morphology (Fig.1). In a speculative vein, this apparent disagreement between our results and those presented by Niedzielski (1996a) might point to a multiperiodic nature of the variability in WR 1, with the 2.667 day period being present in 1994-1995 but being dominated by other longer (non)cyclical processes during our observations. It has to be noted that based on a more comprehensive analysis of the data, Niedzielski (private communication) recently questioned the strictly periodic nature of the spectral variations. As already noted by Niedzielski (1995, 1996b) and Wessolowski & Niedzielski (1996), our data provide evidence for a more gradual spectral pattern of variability than for WR 6; a fact which may imply that the period (if any) is longer than 3.77 days. In support of this view, the centroid, EW, and skewness data of He ii $`\lambda `$4686 show coherent time-dependent patterns of variability (Fig.8), with timescales in the range 5–6 days after a formal periodicity search. However, we caution the reader that this period should not be taken too literally because of the limited time sampling of these data. In particular, a cyclical pattern in this range, as proposed by Lamontagne (1983) and Moffat & Shara (1986), is clearly inconsistent with the lack of continuum flux variations observed after HJD 2,450,353 (Fig.1) unless one invokes a sudden period of “quiescence” after this date, with the small amplitude of the variations masking any periodic patterns. ## 5 On the Presence of a Companion Although a cyclical pattern of variability has yet to be unambiguously established in WR 1, general considerations are made below regarding the model assuming an orbiting (unseen) companion as the origin of the variability. ### 5.1 A Non-degenerate Companion? Some constraints can be set on the mass, $`M_{}`$, of this putative companion on the basis of the centroid measurements of He ii $`\lambda `$4686 presented in Figure 8. Assuming that these variations are entirely attributable to orbital motion (i.e., $`K_{WR}`$ $``$ 70 km s<sup>-1</sup>), one can explore the allowed values of ($`𝒫`$, $`M_{}`$). The solid lines in Figure 9 show the result of this investigation for three illustrative values of the orbital inclination ($`i`$ = 30<sup>°</sup>, 60<sup>°</sup>, and 90<sup>°</sup>). In these calculations, we assume a circular orbit and a mass for WR 1 of 9.1 M (Hamann et al. (1995)). For a wide range in orbital inclination ($`i`$ $``$ 30<sup>°</sup>) and in orbital period ($`𝒫`$ $``$ 20 days), an upper limit for the companion’s mass of 15 M is derived. Assuming the companion to be a main sequence star, this constrains the spectral type to be later than B1. In this case, the companion’s wind is too weak (Grigsby & Morrison (1995)) to induce wind-wind collision effects that may induce the large spectral changes observed in WR 1. For a system observed nearly face-on ($`i`$ $``$ 30<sup>°</sup>), larger masses are evidently consistent with the $`K_{WR}`$ value adopted above. However, no direct evidences (i.e., photospheric lines in the integrated spectrum or dilution of the WR continuum) support the presence of a luminous, early-type companion. In the presence of a non-degenerate companion, one would also expect a flat or eclipsing-type light curve. Yet, the opposite behavior is observed in Figure 1. These arguments strongly argue against the presence of a non-degenerate star orbiting WR 1. ### 5.2 A Collapsed Companion? For a canonical mass of the companion as a neutron star, $`M_{}`$ $``$ 1.4 M, improbably small periods below one day are consistent with the adopted $`K_{WR}`$ value (Fig.9). However, it has to be kept in mind that the centroid measurements of He ii $`\lambda `$4686 included the highly variable uppermost part of the profile which is unlikely to purely reflect orbital motion. Therefore, this $`K_{WR}`$ value is probably grossly overestimated. If one considers lower $`K_{WR}`$ values (as Lamontagne 1983: $`K_{WR}`$ = 22 $`\pm `$ 5 km s<sup>-1</sup>; the allowed values of \[$`𝒫`$, $`M_{}`$\] are shown in this case by short-dashed lines in Fig.9), much larger periods (as in fact suggested by our study) are allowed. Thus, these considerations are not sufficient by themselves to rule out the presence of a neutron star companion (a black hole companion, on the other hand, would require long periods and/or small orbital inclinations). <sup>4</sup><sup>4</sup>4Considering a lower $`K_{WR}`$ value in §5.1 does not qualitatively modify our conclusions. In the presence of a collapsed companion, one may expect fairly strong, accretion-type X-ray emission. Earliest observations by the HEAO A–1 experiment gave an upper limit $`L_X`$ $``$ 4.4 $`\times `$ 10<sup>33</sup> erg s<sup>-1</sup> on the emission in the 0.5–20 keV range (Helfand (1980)). A value $`L_X`$ = 7.07 $`\pm `$ 2.85 $`\times `$ 10<sup>32</sup> erg s<sup>-1</sup> in the 0.2–2.4 keV range has been reported for WR 1 during the ROSAT all-sky survey (Pollock et al. (1995)). WR 1 is a fairly strong X-ray emitter compared to other (apparently) single WN stars. However, its emission is by no means unusual when only considering the WNE-s subclass (Wessolowski (1996)). Two subsequent pointed ROSAT PSPC observations showed that a satisfactory fit to the X-ray spectrum can be achieved, either with a Raymond-Smith thermal plasma of about 1 keV or by the model developed by Baum et al. (1992), assuming a mixture of “cool” (in radiative equilibrium) and “hot” (shocked) material (Wessolowski et al. (1995); Wessolowski (1996)). This picture is consistent with our current understanding of the X-ray production in bona fide single WR stars, as being due to radiatively-induced instabilities (e.g., Willis & Stevens (1996)). This “normal” level of X-ray emission from WR 1 does not constitute, however, a decisive argument against the presence of a collapsed companion, as the accretion of the wind material onto the neutron star is known to be inhibited in some X-ray binaries (e.g., Zhang et al. (1998)). Because of the spiral-in process that massive close binaries are believed to experience in the course of their evolution, one is led to expect periods of some hours for systems made up of a WR star and a compact companion, not days (e.g., De Donder et al. (1997)). A period of 4.8 hr is observed in Cygnus X–3, the prime candidate for a WR + compact companion system (van Kerkwijk et al. (1996)). ## 6 Concluding Remarks The single-star hypothesis appears appealing when considering the similarities between the optical spectral pattern of variability of WR 1 and the ones of the two peculiar stars WR 6 and WR 134. In this respect, although the light-curve morphology of WR 1 (a “bump” followed by a plateau) has, to our knowledge, no example among the WR population (Moffat & Shara 1986; Lamontagne & Moffat (1987); Antokhin et al. (1995); Marchenko et al. 1998b ), such a well-defined light-curve pattern is reminiscent of what is observed in WR 6 (Robert et al. (1992)).<sup>5</sup><sup>5</sup>5Curiously, a very similar light-curve morphology (both in terms of the timescales involved and of the amplitude of the variations) has been noticed in the Be star FV CMa (Balona et al. (1992)). Such a repeatable pattern would, however, imply an unlikely large value for the period in the context of this single-star hypothesis ($`𝒫`$ $``$ 18 days). Since the substantial depolarization of the emission lines observed in WR 6 and WR 134 is generally taken as evidence for an equatorially-enhanced outflow (Schulte-Ladbeck et al. (1991), 1992), revealing the same phenomenon in WR 1 may lead to the interesting suggestion that the occurence of large-scale line-profile and photometric variations (and thus possibly of azimuthally structured outflows) in single WR stars is somehow linked to the existence of a wind-compressed zone (Ignace et al. (1996)). ###### Acknowledgements. We acknowledge an anonymous referee and Alex W. Fullerton, whose comments have stimulated a substantial improvement of this manuscript. T. M., Y. G., and N. S.-L. wish to thank the Natural Sciences and Engineering Research Council (NSERC) of Canada and the Fonds pour la Formation de Chercheurs et l’Aide à la Recherche (FCAR) of Québec for financial support; T. E. is grateful for full financial aid from the Evangelisches Studienwerk/Germany, which is supported by the German Government.
no-problem/9907/hep-ex9907036.html	ar5iv	text	# Issues on NLO pQCD Programs ## 1 Features of NLO pQCD Programs During the last few years four multi-purpose pQCD calculation programs for the HERA DIS processes have become available, Mepjet, Disent, Disaster++, and JETVIP. Their main features are given in table 1. The most important difference is the method used to handle cancelations of singularities in real and virtual corrections. On the one hand, the phase space slicing method integrates analytically in regions of invariant masses lower than an extremly small cut-off parameter $`s_{\mathrm{min}}`$. On the other hand, the subtraction method uses the “plus” prescription to calculate a counter term, that is subtracted from the divergent distributions. For several features, e.g. mass treatment and the contribution of resolved and electroweak processes, only one program is available; therefore cross checks of results, where these corrections get important, are impossible. ## 2 Comment on DiJet cuts Using dijet measurements several QCD tests and parameter extractions have been performed by the HERA experiments. NLO pQCD programs are important tools needed for this task, but the measurements nowadays tend to enter regions where pQCD alone is not able to describe the data, e.g. in the transition region from photoproduction to DIS, where resolved processes become important. In those regions it is all the more important to have reliable NLO predictions. An important restriction for cuts on dijet photoproduction calculations was pointed out some time ago by Frixione and Ridolfi. The same arguments also hold for DIS processes. In figure 1a) the NLO prediction of the double differential cross section in transverse momenta of both jets in the Breit frame is shown. The jets were found using the longitudinal boost invariant $`k_t`$ algorithm. For the LO contributions the jet $`p_t`$ are balanced. The same is true for the virtual corrections and this leads to large corrections on the diagonal, which are known to be negative. The real corrections, canceling the virtual divergences, introduce differences in the jet $`p_t.`$ After imposing a cut of $`p_{t,\text{lower}}>5\text{GeV}`$ for both jets, figure 1b) shows the integrated cross section as a function of the highest jet $`p_t.`$ This plot clearly shows, that the NLO prediction breaks down at a $`p_{t,\text{higher}}`$ cut of approximately $`6\text{GeV},`$ since lowering the $`p_t`$ cuts and thus enlarging the allowed phase space leads to a reduced NLO prediction. A possible solution is to introduce an asymmetric cut, where the difference of the cut values ensures, that the NLO prediction is on the falling edge of the distribution of figure 1b). A different Ansatz is to make a symmetric cut and an additional cut on the sum of the jet $`p_t`$ values, e.g. $`_{1,2}p_{t,i}>17\text{GeV}.`$ This cut removes the main negative contribution in the lower left corner of figure 1a) and ensures, that the cancelation of singularities takes place. ## 3 Conclusions An overview of features currently available in NLO pQCD programs for HERA deep inelastic scattering processes has been given. In addition a special issue on jet cuts for dijet production has been pointed out demonstrating, that asking for a symmetric minimal $`p_t`$ on both jets leads to unreliable NLO predictions. Two alternative scenarios are given, that produce reasonable results. I would like to thank R. Devenish, W.J. Stirling, and M.R. Whalley for organising this interesting and fruitful workshop and J. Butterworth, M.H. Seymour, and G. Thompson for the kind invitation. Special thanks goes to Ch. Berger and S.J. Maxfield for proofreading and commenting on this contribution. ## References
no-problem/9907/astro-ph9907193.html	ar5iv	text	# THE PROPERTIES OF THE RELATIVISTIC IRON K-LINE IN NGC 3516 ## 1 INTRODUCTION A long ASCA exposure of the Seyfert 1 galaxy MCG-6-30-15 revealed a broad, redshifted profile to the iron K$`\alpha `$ line (Tanaka et al. 1995). The line is thought to arise from the inner regions of an accretion disk close to the central black hole, the profile being the result of extreme Doppler and gravitational shifts (e.g., Fabian et al. 1995). “Snapshot” observations of other Seyferts have shown these broad lines to be common and defined the mean properties (Nandra et al. 1997b, hereafter N97b; Reynolds 1997). Furthermore, Iwasawa et al. (1996, hereafter I96) have presented evidence that the profile of MCG-6-30-15 is variable, being extremely broad and redshifted in the “deep minimum” state (see also Reynolds & Begelman 1997; Young, Ross & Fabian 1998; Iwasawa et al. 1999). It is of great interest that the line profile is observed to vary in this way, and indeed at all. The lines typically arise around $`20R_\mathrm{g}`$ (Tanaka et al. 1995; N97b) and should therefore be highly responsive to continuum variations: the expected lag due to reverberation is of order $`1000`$s. In the time taken to accumulate a typical ASCA spectrum ($`2040`$ ks) we would therefore expect the line profile to relax to its mean state and appear similar at all times. I96 suggested that, rather than due to reverberation, the profile changes might be due to a changing pattern of illumination of the disk (see also Weaver & Yaqoob 1998; McKernan & Yaqoob 1999; Reynolds et al. 1999). Here we report on a long observation of the Seyfert 1 galaxy NGC 3516 (z=0.009), which is known to be variable and have a broad, redshifted iron K$`\alpha `$ line (Kriss et al. 1996; N97 a,b,c), offering the opportunity for performing analysis similar to that of MCG-6-30-15. ## 2 OBSERVATIONS The ASCA observation began on 1998-Apr-12 at 21:44:50 and lasted for $`4.5`$ days. We used data from both Solid–state Imaging Spectrometers (SIS0/SIS1), and the two Gas Imaging Spectrometers (GIS2/GIS3). The SIS data were collected in 1-CCD readout mode using FAINT telemetry mode. The data were analyzed according to the methods described in Nandra et al. (1997a,b) and references therein. The total exposure times after screening were 152 ks for the SIS detectors and 186 ks for the GIS, making this one of the longer ASCA exposures of a single object. The mean count rates for the observation were $`0.978\pm 0.003`$, $`0.806\pm 0.002`$, $`0.587\pm 0.002`$ and $`0.724\pm 0.002`$ for SIS0, SIS1, GIS2 and GIS3 (SIS: 0.5-10 keV; GIS: 1-10 keV). The average 2-10 keV flux, determined using a power-law model, was $`4.7\times 10^{11}`$ erg cm<sup>-2</sup> s<sup>-1</sup>, which lies between the flux states described by N97c. ## 3 MEAN LINE PROPERTIES We first analyzed the integrated spectrum. The X-ray spectrum of NGC 3516 is absorbed by a column of partially ionized gas (e.g., Kriss et al. 1996) and we restricted our initial analysis to the 3-10 keV band to minimize its effects. We found no evidence for an iron K-edge from the ionized gas. Fig. 1 shows the line profile of NGC 3516 derived from the SIS data only, derived by adopting a local power law fit in the 3-4 and 7-10 keV range to the SIS+GIS data. It is clearly broad and red-shifted, and shows a very similar shape to MCG-6-30-15 (Tanaka et al. 1995), the mean for Seyfert 1 galaxies (N97b) and previous observations of NGC 3516 (N97c). The profile can be modeled adequately with two gaussians, a narrow “core” and broad, redshifted “wing”, with parameters detailed in Table 1. The fit gave a rather flat photon index compared with previous ASCA observations (N97c), with $`\mathrm{\Gamma }=1.56\pm 0.04`$ and an acceptable $`\chi ^2`$ of 1703.0/1736 d.o.f. We next fitted the data with the Schwarzschild disk line model of Fabian et al. (1989) . Such disk line models are rather complex, with considerable degeneracy which makes the $`\chi ^2`$-fitting process susceptible to traps in local minima. The solutions we show in Table 1 are the best we have found after searching parameter space extensively. Determination of accurate error bars in these circumstances is severely complicated by the model degeneracies, however, and these should be treated with caution. Fits to the Schwarzschild model provided an excellent fit to the data ($`\chi ^2`$=1696.7/1736 d.o.f) - improving on the double-gaussian model by $`\mathrm{\Delta }\chi ^2`$=6.3 for the same number of free parameters. Both the energy of the line and the inner radius pegged at their minimum values (6.34 keV and 6.0 $`R_\mathrm{g}=R_{\mathrm{ms}}`$) in this fit. To improve the stability of the fitting, the inner radius was therefore fixed at $`R_{\mathrm{ms}}`$ when the parameters in Table 1 were determined. This low inner radius and the extraordinarily steep emissivity index of $`q=8.0`$ indicate that the line emission is very strongly concentrated in the central regions, where the most extreme relativistic effects operate. Another way of enhancing these relativistic effects is to allow the disk to extend closer to the black hole, which is possible when the hole is rotating. Fits to the Kerr model of Laor (1991) are also given in Table 1. In this case the outer radius cannot be constrained and was fixed at the maximum value allowed by the implementation of the model ($`400R_\mathrm{g}`$). The $`\chi ^2`$of 1703.4/1736 dof for is slightly worse than the Schwarzschild case but the parameters, in particular the emissivity law, are more in line with our prejudices about the source geometry. For a disk centrally-illuminated by a point source we expect q$`03`$, for example. The Kerr model is plotted as the dotted line in Fig. 1. Weaker lines from iron K$`\beta `$ and Ni K$`\alpha `$ are also expected from the accretion disk (e.g., George & Fabian 1991), and indeed Fig. 1 shows some evidence for excess flux around the expected energies of 7-7.5 keV. We have added additional disk lines to the Kerr and Schwarzschild models to represent emission from Fe K$`\beta `$ and Ni K$`\alpha `$. The energies were fixed at the neutral values of 7.06 keV and 7.47 keV (rest frame) respectively, the fluxes were left free, but all other disk line parameters were tied to those of Fe K$`\alpha `$. For the Kerr model, we find a decrease in fit-statistic of only $`\mathrm{\Delta }\chi ^2`$=2.2 for Fe K$`\beta `$, but a significant improvement ($`\mathrm{\Delta }\chi ^2`$=5.5) for Ni K$`\alpha `$. The EW of these lines are unconstrained, but the best-fit values are 60 eV and 80 eV respectively (c.f. 560 eV for Fe K$`\alpha `$). For the Schwarzschild model both lines are highly significant, with K$`\beta `$ having $`\mathrm{\Delta }\chi ^2`$=6.3 and Ni K$`\alpha `$ $`\mathrm{\Delta }\chi ^2`$=14.1, and EW=100 eV and 150 eV. Using an optimistic error prescription of $`\mathrm{\Delta }\chi ^2`$=2.3 we constrain the ratio of Fe K$`\beta `$/Fe K$`\alpha `$= $`0.14\pm 0.09`$, consistent with the theoretical expectation for neutral iron of 0.11 (Kikoin 1976). The Ni K$`\alpha `$/Fe K$`\alpha `$ ratio is $`0.19\pm 0.07`$, a little higher than than the $`0.07`$ expected based on the fluorescence yields, and solar abundances. Although the disk line clearly accounts for the bulk of the flux and provides a good fit, it is possible that there is a contribution from more distant material, such as the BLR or molecular torus (e.g. Ghisselini, Haardt & Matt 1993; Krolik, Madau & Zycki 1993). Indeed adding a narrow line with fixed energy at 6.4 keV does improve the fit in both cases with $`\chi ^2`$=1684.0/1735 dof for the Schwarzschild model ($`\mathrm{\Delta }\chi ^2`$=12.7; equivalent width, EW= $`50_{30}^{+30}`$ eV) and $`\chi ^2`$=1682.8/1735 dof for the Kerr ($`\mathrm{\Delta }\chi ^2`$=20.6; EW=$`90_{50}^{+60}`$). Aside from a reduction in the flux (and EW) the disk line parameters remain consistent within the errors. This model is shown as the dashed line in Fig. 1. Another interesting possibility is that there is an absorption component present in the red part of the line. Adding such an absorption line to the Schwarzschild model gives $`\chi ^2`$=1690.2/1734 dof ($`\mathrm{\Delta }\chi ^2`$=6.5). For the Kerr model we obtain $`\chi ^2`$=1675.1/1734 dof ($`\mathrm{\Delta }\chi ^2`$=28.3) for a narrow absorption line and a further improvement to $`\chi ^2`$=1670.4/1733 when the absorption line is slightly broadened ($`\sigma `$=0.2 keV). This last model is the best of all attempted against the integrated profile and is shown as the bold line in Fig. 1. Further evidence for the absorption feature is afforded by the time-resolved profiles, which we explore next. ## 4 LINE VARIABILITY The continuum of NGC 3516 varied by a factor of $`50`$ per cent during the observation (Fig. 2), and we have searched for variations in the emission line too. We split the dataset into 8 segments, designated P1-P8, with equal durations (46.4 ks). These are marked on the light curve in Fig. 2. These segments have a typical exposure of $`20`$ ks, which is just sufficient for meaningful determination of the line parameters. We used the same, standard binning for all spectra. The bins in each spectrum satisfy our usual criterion of $`>20`$ ct/bin, but the standard binning method ensures that when comparing line profiles, no spurious differences appear due to different binnings. We employed several different methods of modeling the continuum to deconvolve the line: a) a power law fit in the 3-4 and 7-10 keV bands (as above), b) a two-gaussian fit to the 3-10 keV data, with energies and widths fixed at those for the mean profile then the gaussians removed, and c) a photoionization fit to the full band data, excluding the 4-7 keV band. In practice these all gave similar results. We show the data/model ratios derived from a) in Fig. 3. The peak seems to occur close 6.4 keV (shown by the vertical, dotted line) in most cases. Although individual profiles are noisy there does appear to be some variation in the shape of the line. We have quantified the profile variations by dividing the line into three parts resolved in energy: the “red wing” (4-6 keV), “core” (6.0-6.4 keV) and “blue wing” (6.4-6.8 keV). We took the data model ratios in these bins and converted them into line fluxes using the fitted continuum. The resulting light curves are shown in Fig. 2. $`\chi ^2`$-tests against a constant show variability at $``$ 80, 95 and 99 per cent confidences for the red wing, core and blue wing. The red and blue wing fluxes appear uncorrelated with the continuum variations. A strictly proportional relationship between the blue wing and continuum is ruled out at $`99`$ per cent confidence, based on a $`\chi ^2`$ test. Interestingly, however, the wings appear to be correlated with each other. A linear (Pearson) correlation shows a coefficient of $`r=0.69`$, significant at $`95`$ per cent confidence. The core is evidently variable with a poor $`\chi ^2`$of $`13.1`$ for a constant intensity. Assuming a 1:1 linear relationship with the continuum gives a much better, and acceptable $`\chi ^2`$of $`10.0/8`$ d.o.f. despite having 1 fewer degree of freedom. This offers some evidence that the continuum and line core are correlated and vary in strict proportion, although the Pearson coefficient of $`r=0.49`$ is not significant in this case. We also fitted disk line models to these individual spectra, but were not able to identify a single parameter which accounts for the changes in profile. Notwithstanding the complexity in determining the emission properties, the absorption feature at $`5.9`$ keV found in the mean spectrum does appear to be variable. We have fitted the SIS spectra only, which are much more sensitive to the feature, with the “template” Kerr model from the mean spectrum. The continuum spectral index and line flux were allowed to vary, but all other line parameters were fixed at the mean. We then added an absorption line at 5.89 keV, derived from the mean spectrum, to the model. We found significant features (at $`>99.5`$ and $`>95`$ per cent confidence) in the P4 and P5 spectra ($`\mathrm{\Delta }\chi ^2`$=9.6 and $`\mathrm{\Delta }\chi ^2`$=4.7: see inset to Fig. 1), but insignificant improvements ($`\mathrm{\Delta }\chi ^2`$$`<2.2`$) in all other spectra. Examining the line profiles one could speculate that the feature was present in some of the other spectra, but at different energies. For example, P1 gives $`\mathrm{\Delta }\chi ^2`$=6.0 for a feature at 5.8 keV, significant at 95 per cent confidence. At this stage it is unclear whether the feature is variable in strength, energy or both. A possible alternative explanation for the shape of the P4/P5 line is that we are seeing two separate components, a core around 6.4 keV and a strongly-redshifted and broad component at $`5.5`$ keV. The latter is reminiscent of the profile occasionally seen in MCG-6-30-15 (I96). The absorption interpretation seems preferable, however, as an examination of, e.g., the P2 and P7 profiles is more suggestive of a single line. ## 5 DISCUSSION We have presented a broad, iron-line profile of NGC 3516 with unprecedented signal-to-noise ratio. Disk models give an excellent fit to the data, and indicate a strong concentration of the emission in the close to the central black hole. We cannot distinguish between rotating and non-rotating black hole models using these data alone, although the disk-line parameters seem more plausible for a Kerr geometry. There is also some evidence in the mean profile for a narrower component which arises from somewhere other than the inner accretion disk, and/or an absorption feature at around 5.9 keV. There is also tentative evidence for the Fe K$`\beta `$ and/or Ni K$`\alpha `$ lines in the integrated spectrum. Consideration of the line variability suggests that the profile is variable, meaning that the fits to the integrated spectrum offer only a partial picture of the central regions. The profile changes are not easily interpretable, but energy-resolved light curves indicate that the core of the line responds to the continuum, but that the extended wings do not. These fluxes are also variable, however, and are correlated with each other. The clearest interpretation of this is that the red and blue wings form a single, broad component with a common origin in the inner disk, but that a large fraction of the core comes from somewhat larger radii. The profile changes may be considered surprising given the relatively weak continuum variability, and the inference that the bulk of the flux in the line probably arises from within $`100R_\mathrm{g}`$ (Tanaka et al. 1995; N97b). The light-travel time is only 5000 $`R_{100}M_7`$s (where $`R_{100}`$ is the radius in units of $`100R_\mathrm{g}`$ and $`M_7`$ the mass in units of $`10^7`$ $`\mathrm{M}_{}`$, which we consider a reasonable estimate; see, e.g., Edelson & Nandra 1999). The line should therefore respond fully to any continuum changes in the integrations presented in Fig. 3 and they should simply exhibit the mean profile. If due to reverberation, the profile variations would imply that we have seriously underestimated either the size of the emission regions (e.g. Hua, Kazanas & Cui 1999) and/or the black hole mass. Also, the variation of the “blue wing” is in excess of a factor $`2`$, which is more than the continuum (Fig. 2). Reverberation cannot produce such an over-response without changes in ionization state, for which we have no evidence. More likely the profile variations are dominated not by reverberation, but some other process which acts in a different manner or on a longer time scale. For example the variation in the blue wing could be due to an enhancement in the illuminating flux of the inner disk portion moving towards us, due to a local flare (c.f. I96, Iwasawa et al. 1999). Strong gravitational effects in transverse or receding portions at the same radius would account for the correlation between the red wing and blue core. A more “standard” reverberation picture may be relevant for the core, which comes from further out where the X-ray source appears point-like. The apparent response of the core to the continuum on time scales of $`50`$ ks still places that flux within $`1000R_\mathrm{g}`$, however. Another physical process which can produce variations in the line profile which are not necessarily related to flux and which can occur on a longer time scale is absorption. Complete absorption (i.e. occultation) has already been suggested as a possible mechanism for profile variability (Weaver & Yaqoob 1998). We interpret our absorption feature at $`5.9`$ keV, which is evident in the mean spectrum but particularly strong in the P4 and P5 data, as being due to resonance scattering by iron. Resonance absorption features are expected, but unless there is a large velocity gradient in the absorbing material they would typically not be observable in the ASCA spectra as they would be very narrow and therefore weak. As the feature is redshifted with respect to the rest energy of iron K$`\alpha `$ emission (6.4-6.9 keV), we speculate that the material is infalling and/or suffering gravitational redshift close to the central hole. Evidence for infalling material in AGN is relatively scarce and, if confirmed, such resonance absorption features could provide rare, hard evidence for material actually accreting onto the black hole. Differential (i.e. tidal) changes in the gravitational field would naturally provide a large, apparent velocity gradient in the material, broadening and strengthening the resonance features. Further support for an origin close to the central hole is provided by the variability of the feature, which will be particularly interesting to follow up with the high-resolution XRS calorimeter aboard ASTRO-E. ## ACKNOWLEDGMENTS We thank the ASCA team for their operation of the satellite, and the ASCA GOF at NASA/GSFC for helpful discussions. We acknowledge the support of the Universities Space Research Association (KN, IMG, TY). KN is supported through NASA ADP grant NAG5-7067. This research has made use of data obtained through the HEASARC on-line service, provided by NASA/GSFC.
no-problem/9907/nucl-th9907019.html	ar5iv	text	# Lambda-proton correlations in relativistic heavy ion collisions ## Abstract The prospect of using $`\mathrm{\Lambda }p`$ correlations to extract source sizes in relativistic heavy ion collisions is investigated. It is found that the strong interaction induces a large peak in the correlation function that provides more sensitive source size measurements than $`pp`$ correlations under some circumstances. The prospect of using $`\mathrm{\Lambda }p`$ correlations to measure the time lag between lambda and proton emissions is also studied. Two-particle correlations have proven to be a powerful tool for determining source sizes and lifetimes in heavy ion collisions. At low energies, correlations of protons, neutrons and intermediate mass fragments have provided information on the space-time extent of the collision systems . At relativistic energies, pion, kaon and proton correlations have greatly enhanced our understanding of the dynamics of heavy-ion collisions ; these correlations provide different but complementary information. For instance, heavier particles are more affected by collective flow, thus making the mass-dependence of source sizes a test of our picture of explosive flow in heavy ion collisions; freeze-out conditions may be different for pions, kaons and protons, thus comparing parameters inferred from their correlations allows one to test the conjecture of sequential freeze-out. In this letter, we explore lambda-proton ($`\mathrm{\Lambda }p`$) correlations as a candidate of interferometric study. We find that an enhancement to the correlation function at low relative momentum allows one to infer the size of the emitting source. The inferred lambda source parameters may provide valuable information because lambdas are strangeness carrying baryons. Unlike two-proton ($`pp`$) system, the $`\mathrm{\Lambda }p`$ system has no repulsive Coulomb interaction. Thus the enhancement from the strong interaction better survives when source sizes become large. We illustrate the sensitivity of $`\mathrm{\Lambda }p`$ correlations and show that for large sources, they might be more sensitive than $`pp`$ correlations, but not as sensitive as coalescence measurements. We also study the possibility to determine whether lambdas and protons are emitted simultaneously by comparing the correlations for positive and negative values of the projected outward relative momentum. The correlation of two particles from a chaotic source may be estimated by assuming that they interact only with each other after they are emitted from space time points $`x_a`$ and $`x_b`$ , $`C(𝐩_a,𝐩_b)`$ $`=`$ $`{\displaystyle \frac{𝒫(𝐩_a,𝐩_b)}{𝒫(𝐩_a)𝒫(𝐩_b)}}`$ (2) $`{\displaystyle \frac{d^4x_ad^4x_bS_a(p_a,x_a)S_b(p_b,x_b)\|\varphi _{\mathrm{rel}}(𝐩_b𝐩_a)\|^2}{d^4x_ad^4x_bS_a(𝐩_a,x_a)S_b(𝐩_b,x_b)}}`$ In principle, the correlation depends on the size and shape of the source described by function $`S(𝐩,x)`$, which provides the differential probability of emitting particles of momentum p at a space-time point $`x`$. However, for the purposes of our study we will ignore the momentum dependence of the source functions and assume a Gaussian form for $`S`$. $$S(x_a)=\delta (t)\mathrm{exp}\left(\frac{x^2+y^2+z^2}{2R_g^2}\right),$$ (3) where $`R_g`$ is the Gaussian size of the source. The correlation function’s sensitivity to the source size depends on the form of the relative wave function, $`\varphi _{\mathrm{rel}}`$. The relative wave function is determined by the relative strong interaction, the relative Coulomb interaction, and in the case of identical particles, symmeterization constraints. Since the $`\mathrm{\Lambda }p`$ system involves non-identical particles, and interacts mostly through the $`s`$-wave channel at low relative momentum, it is largely insensitive to details about the shape of the emitting source. However, since the strong interaction is short range, the enhancement of the correlation function at low relative momentum is highly sensitive to the size. Furthermore, the lack of a Coulomb repulsion, which dominates $`pp`$ correlations at low relative momentum, allows the $`\mathrm{\Lambda }p`$ correlation function to remain sensitive to the volume for fairly large sources. An Urbana-type potential which was motivated from low energy $`\mathrm{\Lambda }p`$ scattering and hypernuclei bind energy data is used to generate the relative wave functions. $$V_{\mathrm{\Lambda }p}=V_C\left(\overline{V}\frac{1}{4}V_\sigma 𝝈_\mathrm{\Lambda }𝝈_p\right)T_\pi ^2,$$ (4) where $`V_C`$ is a Woods-Saxon repulsive core. $$V_C=W_C\left[1+\mathrm{exp}\left(\frac{rR}{d}\right)\right]^1,$$ (5) with $`W_C`$=2137 MeV, $`R`$=0.5 fm, $`d`$=0.2 fm. The modified one-pion exchange tensor potential $$T_\pi =\left(1+\frac{3}{x}+\frac{3}{x^2}\right)\frac{e^x}{x}\left(1e^{cr^2}\right)^2,$$ (6) where $`x`$=0.7$`r`$ and $`c`$=2 fm<sup>-2</sup>. The spin-independent part of the attractive potential is characterized by $`\overline{V}=6.2\pm 0.05`$ MeV, while the spin-dependent part is small, $`V_\sigma =0.25\pm 0.25`$ MeV, and not well determined. The correlation functions for $`R_g=4,6`$ and 10 fm sources are illustrated in Fig. 1 as a function of $`k`$, which is one half of the relative momentum, $`k=\|𝐩_\mathrm{\Lambda }𝐩_p\|/2`$, as measured in the pair center-of-mass frame. Also shown are $`pp`$ correlations for the same source sizes. Clearly, $`\mathrm{\Lambda }p`$ correlations are more sensitive than $`pp`$ correlations for determining larger source sizes. Despite the fact that the $`pp`$ scattering length is considerably longer, Coulomb effects obscure the sensitivity of $`pp`$ correlation for sources larger than approximately 6 fm. Thus, even though statistics for $`\mathrm{\Lambda }p`$ correlations are reduced compared to $`pp`$ statistics, they may provide a more accurate determination of the source size under some circumstances. Correlations involving neutrons, such as $`nn`$ or $`pn`$, are also free of a relative Coulomb interaction, and given the large $`nn`$ and $`pn`$ scattering lengths, provide sensitive correlations. However, neutrons are notoriously difficult to measure, whereas lambdas can be measured with a charged-particle detector through its $`\pi ^{}p`$ decay channel. Another candidate for source size measurement is deuteron coalescence . Statistically, coalescence should provide the most accurate determination of source parameters of all baryonic probes as coalescence does not suffer from Coulomb effects, and employs the entire strength of a bound state rather than the fraction represented by a rise in the scattering phase shift. For illustration, we have used the same source size for both lambdas and protons, and assumed thermal momentum distributions. To eliminate the Lorentz factor effect and also for computational reasons, we have used thermal temperature $`T`$=3 MeV. However, the Lorentz factor effect is small due to the large lambda and proton masses. For instance, the difference in the correlation functions between $`T`$=3 and 300 MeV is less than 5%. For the interest of experimental feasibility of $`\mathrm{\Lambda }p`$ correlation measurements, for $`T`$=300 MeV, a total of 20 million pairs results in, from pure phase space population, 10 pairs in 0$`<`$$`k`$$`<`$5 MeV/$`c`$ (hence 30% statistical uncertainty), 70 pairs in 5$`<`$$`k`$$`<`$10 MeV/$`c`$ (12%), 800 pairs in 25$`<`$$`k`$$`<`$30 MeV/$`c`$ (4%), and 7000 pairs in 0$`<`$$`k`$$`<`$50 MeV/$`c`$. An experimental correlation function often resorts to a mixed-event technique for uncorrelated pairs. Since both the height and the width of a $`\mathrm{\Lambda }p`$ correlation function are sensitive to the source size, it is important to normalize the correlation function properly. Since lambda and proton exhibit no correlation at large $`k`$, an experimental $`\mathrm{\Lambda }p`$ correlation function may be normalized to unity at large $`k`$. The $`\mathrm{\Lambda }p`$ correlation function is one fourth spin singlet ($`S`$=0) and three fourths spin triplet ($`S`$=1). To illustrate the spin dependence, correlation functions are presented for a $`R_g`$=4 fm source in the upper panel of Fig. 2, separately for $`S`$=0 and $`S`$=1 pairs. If $`V_\sigma `$ in Eq. (4) were zero, the two contributions would be identical. As $`V_\sigma `$ is not well understood, $`0<V_\sigma <0.5`$ MeV , measuring the spin dependence of the correlation function could in principle determine $`V_\sigma `$. The correlation function is largely determined by the scattering length and effective range of the potential . The scattering length and effective range corresponding to the potential (4) ($`\overline{V}`$=6.2 MeV, $`V_\sigma `$=0.25 MeV) are $`2.88`$ fm and $`2.92`$ fm for the spin singlet, and $`1.66`$ fm and $`3.78`$ fm for the spin triplet, respectively. They are in reasonable agreement with those from Refs. . Using these values and an analytical approximation similar for neutron-proton correlations , we obtain $`\mathrm{\Lambda }p`$ correlation functions that are consistent with the results in the upper panel of Fig. 2. The sensitivity of the spin-averaged correlation function to the parameters $`\overline{V}`$ and $`V_\sigma `$ is illustrated in the lower panel of Fig. 2. Using the stated uncertainties , $`\overline{V}=6.2\pm 0.05`$ MeV and $`V_\sigma =0.25\pm 0.25`$ MeV, the correlation function for a $`R_g`$=4 fm source is shown for a range of parameters. The results suggest that the uncertainties in the potential parameters translate into an approximately $`\pm 0.5`$ fm uncertainty in the source size extracted from the correlation function at $`k`$$`<`$25 MeV/$`c`$, while the large $`k`$ tail has better constrain on the source size. Recently, Lednicky et al. have shown that correlations of non-identical particles can provide information revealing whether the particles are emitted simultaneously. If the lambdas are emitted before the protons in such a way that the probability cloud describing the protons lags that for the lambdas of the same velocity, the correlation function then depends on the sign of the relative momentum in the direction defined by that of the displacement of the lambda and proton clouds. An example is illustrated in Fig. 3, where both sources are assumed to be characterized by a size of 4 fm, but are separated by $`\mathrm{\Delta }z`$=1.5 fm or 3.0 fm. In Fig. 3 the difference of the correlation functions, $`C_+(k)C_{}(k)`$, is plotted against $`k`$ where the transverse component of $`k`$ is required to be less than 10 MeV/$`c`$. Here, $`C_+`$ refers to the correlation function constructed with the requirement that $`k_z>0`$, while $`C_{}`$ is constructed with the opposite constraint. A displacement $`\mathrm{\Delta }z`$ can result from a displacement in time, $`\mathrm{\Delta }\tau =\mathrm{\Delta }z/v`$, where $`v`$ is the velocity of the $`\mathrm{\Lambda }p`$ pair. The displacement direction, $`\widehat{z}`$, is then defined by the direction of $`v`$, or that of the pair momentum relative to the source. Thus, if $`\mathrm{\Lambda }p`$ correlations can be measured with an accuracy of a few percent at $`k`$$``$30 MeV/$`c`$, the conjecture that strange and non-strange baryons are emitted simultaneously can be addressed quantitatively. One should remember that the residual interaction of the proton with the Coulomb field of the nuclear sources might distort the result. Unlike $`pp`$ correlations, where both particles feel the identical force, only the proton experiences the Coulomb field. This issue has been previously considered in the context of $`pn`$ correlations where the distortion was shown to be small for fast-moving pairs . This effect should be smaller at RHIC where the excess charge at midrapidity is expected to be smaller than observed at SPS and AGS energies. Finally, we make a brief note regarding $`\mathrm{\Lambda }\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}p`$ correlations. In the view of our $`\mathrm{\Lambda }p`$ results, we expect that the $`\mathrm{\Lambda }\mathrm{\Lambda }`$ strong interaction would also give sizeable correlations. These correlations are of great interest because of the predicted existence of a $`\mathrm{\Lambda }\mathrm{\Lambda }`$-like particle . In conjunction with $`\mathrm{\Lambda }p`$, $`\overline{\mathrm{\Lambda }}p`$ correlations may reveal valuable information on low energy $`\overline{\mathrm{\Lambda }}p`$ annihilation cross sections which are presently unknown but indispensable in modeling certain aspects of heavy-ion collisions . In summary, $`\mathrm{\Lambda }p`$ correlations may provide a useful characterization of the space-time structure of relativistic heavy ion collisions should one be able to gather sufficient statistics. The lack of a relative Coulomb interaction allows the strong interaction to produce a large peak in the correlation function even for large sources, to which $`pp`$ correlation loses its sensitivity. Furthermore, by binning according to the sign of the projected relative momentum, one might address the question of whether lambdas and protons are emitted simultaneously. However, the interpretation of the correlation function could benefit from a more precise parameterization of the $`\mathrm{\Lambda }p`$ interaction. ###### Acknowledgements. F.W. acknowledges Drs. V. Koch, R. Lednicky, J.C. Peng, A.M. Poskanzer and N. Xu for useful discussions. This work was supported by the U.S. Department of Energy under contract DE-AC03-76SF00098 and the National Science Foundation under grant PHY-96-05207.
no-problem/9907/math9907212.html	ar5iv	text	# References Journal of Nonlinear Mathematical Physics 1999, V.6, N 3, 1References. Letter Copyright © 1999 by B.A. Kupershmidt Remarks on Quantization of Classical $`r`$-Matrices Boris A. KUPERSHMIDT Department of Mathematics, University of Tennessee Space Institute, Tullahoma, TN 37388, USA E-mail: bkupersh@utsi.edu Received May 6, 1999; Revised May 26, 1999; Accepted June 1, 1999 ## Abstract If a classical $`r`$-matrix $`r`$ is skewsymmetric, its quantization $`R`$ can lose the skewsymmetry property. Even when $`R`$ is skewsymmetric, it may not be unique. Let $`r`$ be a classical $`r`$-matrix. In general, it means that we have a family of vector spaces $`\{V_\alpha \}`$, $`\alpha 𝒜`$, and a collection of linear operators $$r(\alpha ,\beta ):V_\alpha V_\beta V_\beta V_\alpha ,\alpha \beta 𝒜,$$ (1) satisfying the misnamed “Classical Yang-Baxter” equation (CYB) $$\begin{array}{c}[c(r)]_{ijk}^{\phi \psi \xi }(\alpha ,\beta ,\gamma ):=\left(r(\alpha ,\beta )_{ij}^{s\phi }r(\beta ,\gamma )_{sk}^{\xi \psi }+r(\alpha ,\beta )_{ij}^{\psi s}r(\alpha ,\gamma )_{sk}^{\xi \phi }\right)\hfill \\ +c.p.(i,j,k;\phi ,\psi ,\xi ;\alpha ,\beta ,\gamma )=0\hfill \end{array}$$ (2) where “$`c.p`$” stands for the sum on cyclically permuted triples of indices indicated, and $`r(\alpha ,\beta )_{ij}^{uv}`$ are the matrix elements of the operators $`r(\alpha ,\beta )`$ (1) in a collection of fixed basises: $$r(\alpha ,\beta )\left(e_i^\alpha e_j^\beta \right)=r(\alpha ,\beta )_{ij}^\mathrm{}ke_{\mathrm{}}^\beta e_k^\alpha ;$$ (3) the convention of summation over repeated upper-lower indices is in force. In most applications, all the vector spaces $`V_\alpha `$ are isomorphic to each other, $`V_\alpha V`$; in addition, often, – but not always, – the operator $`r:VVVV`$ is skewsymmetric: $$PrP=r,r_{ij}^k\mathrm{}=r_{ji}^\mathrm{}k,$$ (4) where $`P`$ is the permutation operator, $$P(xy)=yx.$$ (5) We shall consider this particular framework from now on. To quantize a given $`r`$-matrix $`r`$ is to find an operator family $$R=R(h):VVVV,$$ (6) depending upon a parameter $`h`$, such that $$R(h)=P+hr+O(h^2),$$ (7) and $`R`$ satisfies the Artin braid relation (also misnamed as the “Quantum Yang-Baxter” equation, QYB): $$R^{12}R^{23}R^{12}=R^{23}R^{12}R^{23},$$ (8) where this equality of operators acting on $`VVV`$ employs the standard notation $$R^{12}(xyz)=R(xy)z,R^{23}(xyz)=xR(yz).$$ (9) How does the skewsymmetry condition on $`r`$, (4), translate into $`R=R(h)`$? There are at least two possible, logically independent, answers, only one of which is correct. The first one is what is commonly accepted in the literature under the name of “unitarity”: $$R(h)^1=R(h),$$ (10a) or $$R(q)^1=R(q),$$ (10b) in the multiplicative notation $`q=e^h`$. The second one I shall call, for want of a better term, the mirror symmetry: $$R^{}(h)=R(h).$$ (11) Here $`R^{}`$ is the operator acting as the mirror image of $`R`$. If $$R\left(e_ie_j^{}\right)=R_{ij}^k\mathrm{}e_k^{}e_{\mathrm{}},$$ (12) then $$R^{}\left(R_{ij}^k\mathrm{}e_{\mathrm{}}e_\kappa ^{}\right)=e_j^{}e_i.$$ (13) This definition, useful as it is, is not connected to skewsymmetry of $`r`$. The classical $`r`$-matrix $`r`$ appears as the $`h^1`$-term in the $`h`$-expansion of the Quantum $`R`$-matrix $`R(h)`$ around $`h=0`$. The terms in $`h`$ of orders higher than 1 recede away in the quasiclassical passage. The examples that follow demonstrate that these higher-order terms can have distinctly anti-Prussian character and break out strict orders and symmetries. (In Drinfel’d proved that every skewsymmetric classical $`r`$-matrix $`r`$ represents $`h^1`$-part of some skewsymmetric Quantum $`R`$matrix $`R`$. The question of additional parameters in $`R`$ was not addressed there, or elsewhere.) In the $`1^{st}`$ example, $`\text{dim}(V)=2`$ and the $`R`$-matrix $`R=R(h;\theta )`$ acts on $`VV`$ (in a chosen basis) as $$R\left(e_0e_0^{}\right)=e_0^{}e_0,$$ (14) $$R\left(e_0e_1^{}\right)=\left(e_1^{}+he_0^{}\right)e_0,$$ (15) $$R\left(e_1e_0^{}\right)=e_0^{}(e_1he_0),$$ (16) $$R\left(e_1e_1^{}\right)=e_1^{}e_1+\theta h^2e_0^{}e_0.$$ (17) Here $`\theta `$ is an arbitrary constant. The Artin relation (8) is easily verified. The $`h^1`$-terms comprise the $`r`$-matrix $$r_{ij}^k\mathrm{}=\delta _0^k\delta _0^{\mathrm{}}\left(\delta _{ij}^{01}\delta _{ij}^{10}\right)$$ (18) which is obviously skewsymmetric. The $`R`$-matrix $`R(h;\theta )`$ is, however, not unitary unless $`\theta =0`$. Also, it’s easy to see that $$R^{}(h;\theta )=R(h;\theta ),$$ (19) so that this $`R`$-matrix is not mirror-symmetric either, again unless $`\theta =0`$. Our 2<sup>nd</sup> example is a little bit more elaborate, with $`\text{dim}(V)=3`$. Here the $`R`$-matrix is both skewsymmetric and mirror-symmetric, but it depends upon one extra parameter, in addition to the quantization parameter $`h`$, thus exhibiting clearly nonuniqueness of quantization of classical $`r`$-matrices. Fixing a basis $`(e_0,e_1,e_2)`$ in $`V`$, we set $$R\left(e_0e_0^{}\right)=e_0^{}e_0,$$ (20.1) $$R\left(e_0e_1^{}\right)=\left(e_1^{}+he_0^{}\right)e_0,$$ (20.2) $$R\left(e_1e_0^{}\right)=e_0^{}(e_1he_0),$$ (20.3) $$R\left(e_1e_1^{}\right)=e_1^{}e_1;$$ (20.4) $$R\left(e_0e_2^{}\right)=\left(e_2^{}+he_1^{}+\frac{h^2}{2}e_0^{}\right)e_0,$$ (21.1) $$R\left(e_2e_0^{}\right)=e_0^{}\left(e_2he_1+\frac{h^2}{2}e_0\right),$$ (21.2) $$R\left(e_1e_2^{}\right)=e_2^{}(e_1+he_0)+h^2\left(\frac{1}{2}e_1^{}+\lambda he_0^{}\right)e_0,$$ (21.3) $$R\left(e_2e_1^{}\right)=\left(e_1^{}he_0^{}\right)e_2+h^2e_0^{}\left(\frac{1}{2}e_1\lambda he_0\right),$$ (21.4) $$\begin{array}{c}R\left(e_2e_2^{}\right)=e_2^{}\left(e_2+he_1+\frac{h^2}{2}e_0\right)\hfill \\ he_1^{}\left(e_2+\stackrel{~}{\lambda }h^2e_0\right)+h^2e_0^{}\left(\frac{1}{2}e_2+\stackrel{~}{\lambda }he_1\right).\hfill \end{array}$$ (21.5) Here $`\lambda `$ is the new free parameter, and $$\stackrel{~}{\lambda }=\lambda \frac{1}{4}.$$ (22) From formulae (20) we see that the previous example (14)–(17) is embedded into this one, with $`\theta =0`$. It’s immediate to check that $$R(h;\lambda )^2=\mathrm{𝟏},$$ (23) $$R^{}(h;\lambda )=R(h;\lambda ),$$ (24) so that our $`R`$-matrix is both skewsymmetric and mirror-symmetric. Also, the $`h^1`$-part of $`R(h;\lambda )`$ is given by the flag-type formula $$r_{ij}^k\mathrm{}=(ic)\delta _i^{\mathrm{}}\delta _{j1}^k(jc)\delta _{i1}^{\mathrm{}}\delta _j^k,0i,j,k,\mathrm{}\text{dim}(V)1,$$ (25) where $`c`$ is an arbitrary constant. \[In our case $`c=1`$, but this constant can be adjusted to any desired value by an appropriate nonlinear transformation; in particular, we can make $$c=\frac{\text{dim}(V)1}{2}$$ (26) to have the determinant in $`GL(V)`$ being central in the induced Lie-Poisson structure . In this language, the $`R`$-matrix (20)–(21) defines the Quantum Group $`\text{Mat}_{h;\lambda }(3)`$, a 3-dimensional analog of the 2-dimensional Quantum Group $`\text{Mat}_h`$(2).\] The checking of the Artin relation for the $`R`$-matrix (20)–(21) is easy but tedious; the mirror property (24) cuts the verification procedure by $`1/3`$; there are still more symmetries present in this $`R`$-matrix which will allow another $`1/3`$ of the checking labor to be avoided. How many additional constants should one expect when quantizing a skewsymmetric classical $`r`$-matrix and requiring the Quantum $`R`$-matrix to be skewsymmetric and mirror-symmetric? For the case of the $`r`$-matrix (25), I expect the total number of additional parameters (the $`\lambda `$’s) to be $$\text{dim}(V)2,$$ (27) and in general it probably could never be larger no matter what $`r`$ is; dropping off the mirror-symmetry condition increases the number of possible parameters by 1.

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.