1: \documentclass[aps,preprint,showpacs]{revtex4}
2:
3: \usepackage{graphicx}
4: \usepackage{bm}
5:
6: \begin{document}
7:
8: \title{Microscopic spectral density in random matrix models for chiral
9: and diquark condensation}
10:
11: \author{Beno\^\i t Vanderheyden} \affiliation{Departement
12: of Electrical Engineering and Computer Science, \\
13: B-28, Universit{\'e} de Li\`ege, Sart-Tilman, \\
14: B-4000 Li\`ege, Belgium}
15:
16: \author{A. D. Jackson}
17: \affiliation{
18: The Niels Bohr Institute, Blegdamsvej 17, DK-2100 Copenhagen \O, Denmark}
19:
20: \date{\today}
21:
22: \begin{abstract}
23:
24: We examine random matrix models of QCD which are capable of
25: supporting both chiral and diquark condensation. A numerical study
26: of the spectral densities near zero virtuality shows that the
27: introduction of color in the interactions does not alter the
28: one-body results imposed by chiral symmetry. A model with three
29: colors has the spectral density predicted for the chiral ensemble
30: with a Dyson index $\beta = 2$; a pseudoreal model with two colors
31: exhibits the spectral density of the chiral ensemble with $\beta =
32: 1$.
33:
34: \end{abstract}
35:
36: \pacs{11.30. Fs, 11.30. Qc, 11.30. Rd, 12.38. Aw}
37:
38: \maketitle
39:
40: Chiral random matrix theory ($\chi$RMT) is based on the observation
41: that many of the low-energy properties of QCD are dominated by its
42: global symmetries~\cite{reviewchiRMM}. Random matrix
43: models~\cite{chiRMM} thus attempt to capture the basic mechanisms for
44: chiral condensation by reducing the QCD interactions to their
45: essential structure. These models introduce low-lying modes which
46: respect a basic left-right symmetry but which interact via random
47: matrix elements. The consequences of chiral symmetry have been
48: investigated at two levels. The microscopic level deals with the
49: statistical properties of the eigenvalues of the Dirac operator and
50: their correlations. Here, $\chi$RMT
51: studies~\cite{VerZah93,Ver94,Ver94-2} have helped in understanding how
52: these properties are determined by the spontaneous breaking of chiral
53: symmetry as indicated, for example, by the universality of the
54: spectral density near zero virtuality and associated sum
55: rules~\cite{reviewchiRMM}. At the macroscopic level, it is possible
56: to consider the consequences of chiral symmetry for the global state
57: of the random matrix system. With the aid of the mean field
58: approximation and additional prescriptions for including the effects
59: of temperature and chemical potential, it is straightforward to
60: construct the partition function for $\chi$RMT models and thus examine
61: the patterns of chiral symmetry breaking as a function of $T$, $\mu$,
62: and the quark mass~\cite{JacVer96,HalJac98}. Many of the properties
63: of the resulting phase diagram are direct consequences of chiral
64: symmetry and largely independent of the detailed form of the
65: interactions. They are thus expected to provide guidance in our
66: understanding of the QCD phase diagram particularly in cases (e.g.,
67: $N_c = 3$ and $\mu \ne 0$) where the Dirac operator is non-Hermitean.
68: These are cases where lattice simulations cannot rely on importance
69: sampling and are very difficult to
70: perform~\cite{AlfKap99,ChaWie99,FodKat02,ForPhi02}.
71:
72: There has been considerable interest in the possibility that both
73: chiral and diquark condensates can develop and compete
74: thermodynamically in QCD. (For a review of diquark condensation in
75: QCD, see~\cite{early,RapSch98,AlfRaj98,review}.) Thus, we recently
76: proposed a random matrix model for QCD which goes beyond $\chi$RMT and
77: has interactions which implement coexisting chiral and color
78: symmetries~\cite{VanJac99,VanJac01}. We have examined this model at
79: the macroscopic level by identifying the allowed topologies for the
80: $(T,\mu)$ phase diagram. The phase structure is determined by a
81: single parameter $\alpha$, defined as a ratio between coupling
82: constants in the chiral and diquark condensation channels. Physically,
83: this ratio measures the relative strengths of the chiral and color
84: symmetries in the random matrix interactions. In the cases relevant
85: for QCD [either ${\rm SU}(2)$ or ${\rm SU}(3)$], $\alpha$ has a fixed
86: value and is associated with a given phase structure. The topology of
87: this phase structure is robust with respect to moderate variations in
88: the detailed form of the interactions (i.e., variations in $\alpha$).
89:
90: The addition of color structure to the interactions implies additional
91: constraints on the random matrix ensemble considered. As we shall see
92: below, the color generators modify the statistical weights of the
93: interaction matrix elements. The question then arises whether these
94: constraints are capable of altering the various results associated
95: with chiral symmetry alone. Our previous considerations suggest that
96: this is not the case at the macroscopic level. In the case of an
97: interaction which is completely dominated by chiral symmetry ($\alpha
98: \to 0$), our models precisely reproduce the phase diagram of $\chi$RMT
99: (Ref.~\cite{JacVer96}). In cases relevant for QCD with fundamental
100: fermions~\footnote{In this work, we consider only fundamental
101: fermions. For models with adjoint fermions,
102: see~\cite{reviewchiRMM,KogSteTou00,KimSon01,KimVer01}} and three
103: colors ($N_c = 3$ and $\alpha = 0.75$), the diquark phase develops in
104: regions where it is thermodynamically advantageous, but color does not
105: otherwise modify the phase structure. Thus, for $N_c = 3$, color does
106: not seem to weaken the chiral correlations at the macroscopic level.
107:
108: For $N_c = 2$, the situation is more subtle: the gauge interaction is
109: pseudoreal, and baryons and mesons belong to the same multiplets. For
110: $m =0$, $\mu = 0$, and $N_f$ flavors, the quark Lagrangian has an
111: extended ${\rm SU}(2 N_f)$ invariance which relates chiral and diquark
112: fields. This flavor symmetry is explicitly broken for either $m >0$ or
113: $\mu >0$. Studies of the symmetry breaking patterns showed that a
114: diquark condensed phase becomes favorable for $\mu \sim m_\pi \propto
115: m^{1/2}$ and $T <
116: T_c$~\cite{KogSteTou99,KogSteTou00,KimSon01,KimVer01}. Our model
117: exhibits the ${\rm SU}(2 N_f)$ symmetry and produces results which agree
118: with the general predictions of Ref.~\cite{KogSteTou99} as well as
119: with those of chiral perturbation
120: theory~\cite{KogSteTou00,KimVer01}. Thus, for $N_c = 2$, the
121: introduction of color in the random matrix interactions does not lead
122: to unexpected topologies of the phase diagram (i.e. unexpected
123: macroscopic properties).
124:
125: The purpose of this paper is to investigate the consequences of the
126: additional color correlations at the microscopic level. To this end,
127: we examine the distributions of the eigenvalues of the Dirac operator
128: near zero virtuality and compare them with the analytic forms
129: predicted by $\chi$RMT. The rest of the paper is organized as follows:
130: We present the models in Sec.~\ref{s:models}, discuss aspects of their
131: macroscopic and microscopic spectral densities in
132: Secs.~\ref{s:macrosd} and \ref{s:microsd}, comment on other properties
133: in Sec.~\ref{s:others}, and conclude in Sec.~\ref{s:concl}.
134:
135: \section{The random matrix models}
136: \label{s:models}
137:
138: To understand the form of the correlations induced by each symmetry,
139: it is useful to provide a brief description of the models of
140: Refs.~\cite{VanJac99}, which we will refer to as I. In the present
141: paper we will restrict our attention to a theory with $N_f$ light
142: flavors, zero temperature, zero chemical potential, and zero quark
143: mass. We start by recalling the basic form of chiral random matrix
144: theory, in which only chirality is introduced, and then extend the
145: model to include color.
146:
147: We consider first a chiral random matrix model and work in the sector
148: of zero topological charge for simplicity. The partition function has
149: the form
150: \begin{eqnarray}
151: Z = \int\,{D} W \,\,\prod_{i = 1}^{N_f}\,\, {D}\psi_i^*\,{D}
152: \psi_i^{\phantom{*}} \,
153: \exp\left[i \sum_{i= 1}^{N_f} \, \psi^*_i \, {\cal D} \, \psi_i
154: \right]\,\exp\Big(- \frac{N \beta \Sigma^2}{2}\,{\rm Tr}[W W^\dagger]\Big),
155: \label{Z}
156: \end{eqnarray}
157: where $\psi_i^{\phantom *}$ and $\psi^*_i$ are independent Grassmann
158: variables representing the quark fields and where the matrix ${\cal
159: D}$ represents the Dirac operator. Its block structure reflects chiral
160: symmetry. Working in a suitable basis of left and right states $(1 \pm
161: \gamma_5) \phi_n$, ${\cal D}$ has the form~\cite{chiRMM}
162: \begin{eqnarray}
163: {\cal D} & = & \left( \begin{array}{cc} 0 & W \\ W^\dagger & 0
164: \end{array} \right),
165: \label{D}
166: \end{eqnarray}
167: where $W$ is an $N\times N$ block matrix. The integral in
168: Eq.~(\ref{Z}) is over the matrix elements of $W$, $DW$ is a Haar
169: measure, and Tr denotes a trace over the $N$ matrix indices. The model
170: is thus a theory of $2 N$ low-lying modes which respect chiral
171: symmetry and whose interaction matrix elements $W_{ij}$ are drawn on a
172: Gaussian distribution. The number of modes scales with the volume of
173: the system; the thermodynamic limit is taken as $N \to \infty$.
174:
175: For later comparisons, it is useful to understand how the gauge group
176: is taken into account. In a model which implements the global
177: symmetries of QCD with $N_c = 3$, the matrix elements of $W$ are
178: complex. For the analysis below, it is worth noting that their real
179: and imaginary parts satisfy no particular relationship and are thus
180: drawn independently. This case corresponds to a Dyson index $\beta =
181: 2$ and is described by the chiral unitary ensemble ($\chi$GUE). For
182: $N_c = 2$, the gauge group is pseudoreal as mentioned above. The
183: Dirac operator then contains an additional antiunitary
184: symmetry~\cite{chiRMM}, which allows one to choose a basis in which
185: the matrix elements of $W$ are real. This case leads to the chiral
186: orthogonal ensemble ($\chi$GOE). In short, the value of $N_c$ enters
187: $\chi$RMT only through the reality of the matrix elements of $W$.
188:
189: In model I, we included color directly in the interactions in a way
190: which mimics single-gluon exchange. For two flavors, the partition
191: function takes the form
192: \begin{eqnarray}
193: Z = \int\,{D} H \,\,{D}\psi_1^\dagger\,{D}\psi_1^{\phantom{*}} \,
194: {D}\psi_2^*\,{D}\psi_2^T \,
195: \exp\left[i
196: \left(
197: \begin{array}{c}
198: \psi_1^\dagger \\
199: \psi_2^T \\
200: \end{array}
201: \right)^T
202: \left(
203: \begin{array}{cc}
204: {\cal D}_c & 0 \\
205: 0 & -{\cal D}_c^T \\
206: \end{array}
207: \right)
208: \left(
209: \begin{array}{c}
210: \psi_1 \\
211: \psi_2^*
212: \end{array}
213: \right)
214: \right],
215: \label{Zcolor}
216: \end{eqnarray}
217: where $\psi_1$ and $\psi_2^T$ denote the quark fields for flavor $1$
218: and flavor $2$, respectively, $D H$ is a measure to be defined below,
219: and ${\cal D}_c$ is the single-quark Dirac operator. Note that the
220: subblock associated with flavor $2$ has been transposed in order to
221: exhibit the possibility of forming $\langle \psi_2^T \psi_1^{\phantom
222: T} \rangle$ condensates (see I).
223:
224: The diagonal block ${\cal D}_c$ now reflects both chiral and color
225: symmetries. In order to be able to define an order parameter which is
226: antisymmetric under the permutation of an odd number of quantum
227: numbers, spin has to be introduced together with
228: color~\cite{VanJac99}. The Dirac operator ${\cal D}_c$ then has the
229: chiral structure of Eq.~(\ref{D}) where $W$ is exploded into a $2 N_c
230: \times 2 N_c$ matrix of embedded spin and color subblocks,
231: \begin{eqnarray}
232: W = \sum_{\mu = 0}^3 \, \sum_{a = 1}^{N_c^2 - 1}
233: (\sigma_\mu \otimes \lambda_a)\, A_{\mu a}.
234: \label{W}
235: \end{eqnarray}
236: Here $\sigma_\mu = (1, i \vec{\sigma})$ are $2 \times 2$ spin
237: matrices, $\lambda_a$ are the $N_c \times N_c$ matrices of ${\rm
238: SU}(N_c)$ and $A_{\mu a}$ are $n \times n$ real matrices representing
239: the gluon fields. Taking into account all substructures, each matrix
240: $W$ is thus $N \times N$ with $N = 2 N_c n$. The matrix elements of
241: $A_{\mu a}$ in Eq.~(\ref{W}) are distributed according to the measure
242: ${D}H$, Eq.~(\ref{Zcolor}), which takes the form
243: \begin{eqnarray}
244: {D}H = \left\{ \prod_{\mu a}\,{D}A_{\mu a} \right\}
245: \,\exp\left(- N \Sigma_0^2 \sum_{\mu a} {\rm Tr}[A_{\mu a} \,
246: (A_{\mu a})^T]\right),
247: \label{DH}
248: \end{eqnarray}
249: where $D A_{\mu a}$ are Haar measures, $\Sigma_0$ is a constant, and the
250: superscript $T$ denotes a transposition. Again, the thermodynamic limit
251: corresponds to $N \to \infty$.
252:
253: Evidently, these spin and color substructures impose strong
254: constraints between the real and imaginary parts of the matrix
255: elements of $W$. Consider, for example, the contributions to ${\cal
256: D}_c$, Eqs.~(\ref{Zcolor}) and (\ref{W}), of the random matrices
257: $A_{01}$ ($\mu =0$ and $a = 1$) and $A_{02}$. Their matrix elements
258: are real and are drawn independently. When combined to form $W$ as
259: prescribed in Eq.~(\ref{W}), the matrix elements of $A_{01}$ multiply
260: $\sigma_0$, a real diagonal spin matrix, and $\lambda_1$, a real color
261: matrix. Hence, $A_{01}$ contributes to the real part of $W$
262: only. Similarly, the matrix elements of $A_{02}$ combine with
263: $\sigma_0$ (real) and $\lambda_2$, an imaginary color matrix. They
264: thus contribute to the imaginary part of $W$. Hence, in contrast to
265: ordinary $\chi$RMT, the matrix elements of $W$ are complex for both
266: $N_c =3$ and $N_c = 2$. The real and imaginary parts of $W$ arise from
267: well-defined combinations of the matrix elements of $A_{\mu a}$. Their
268: statistical distributions are then dictated by the content of the spin
269: and color block matrices, which thus introduce well-defined
270: correlations.
271:
272: Do these additional correlations preserve those imposed in
273: Eq.~(\ref{D}) by chiral symmetry? The $W$ matrix elements are complex
274: for all $N_c$; their real and imaginary parts are no longer
275: independent random variables. One might thus be concerned that the
276: statistical properties of the eigenvalues of the Dirac operator would
277: differ from those of the $\chi$GUE and $\chi$GOE. We now consider a
278: number of spectral properties to indicate that this is {\em not\/} the
279: case and that the additional color symmetries do not alter the
280: statistical properties due to chiral symmetry.
281:
282: \section{The macroscopic spectral density}
283: \label{s:macrosd}
284:
285: As an initial measure of the statistical properties, we consider the
286: spectral density $\rho(\lambda)$, defined as
287: \begin{eqnarray}
288: \rho(\lambda) & \equiv & {1 \over 2 N} \, \left\langle
289: \sum_{i = 1}^{2 N} \delta(\lambda - \lambda_i)\right\rangle,
290: \end{eqnarray}
291: where $\lambda_i$ are the $2 N$ eigenvalues of the Dirac operator and
292: $\langle \rangle$ denotes an ensemble average. While $\rho(\lambda)$
293: is not universal in the usual random matrix sense, we will focus on
294: those symmetry properties which are expected to be protected.
295: Consider first the chiral random matrix model defined in Eq.~(\ref{Z}).
296: Because of the block structure of ${\cal D}$, Eq.~(\ref{D}), the eigenvalues
297: $\lambda_i$ occur in pairs of opposite signs. Moreover, Eq.~(\ref{Z})
298: shows that the spectrum is $N_f$-fold degenerate. Consider next model I,
299: Eq.~(\ref{Zcolor}), which may seem different at first glance
300: because of its more elaborate structure. In fact, each flavor subblock has
301: the same chiral substructure as in Eq.~(\ref{D}), and the eigenvalue
302: spectrum remains symmetric about $\lambda = 0$. Further, the eigenvalues
303: of $- {\cal D}_c^T$, Eq. (\ref{Zcolor}), are degenerate with those of
304: ${\cal D}_c$ so that the spectrum is again twofold degenerate ($N_f = 2$).
305: Hence, the same basic chiral and flavor symmetries prevail in the two models.
306:
307: In order to facilitate numerical evaluation, we now consider the
308: quenched limit $N_f \to 0$. This limit is free from contributions from
309: vacuum graphs (through powers of the determinant of the Dirac operator
310: in the partition function; see~\cite{reviewchiRMM}). The ensemble
311: average of a given quantity then amounts to a mere counting of the
312: contributions from the individual eigenvalues of ${\cal D}$ (or ${\cal
313: D}_c$), distributed according to the normal laws of Eq.~(\ref{Z}) or
314: Eq.~(\ref{DH}), as appropriate.
315:
316: For large matrices, we find that the model I is numerically consistent
317: with the (non universal) semicircle law familiar from the $\chi$GUE
318: and $\chi$GOE,
319: \begin{eqnarray}
320: \lim_{N \to\infty} \rho(\lambda) & = & \left\{ \begin{array}{cc}
321: \frac{\Sigma}{\pi}\, \sqrt{1
322: - \left(\frac{\Sigma \lambda}{2}\right)^2} &
323: \textrm{if}~|\lambda| \le 2/\Sigma, \\
324: 0 & \textrm{otherwise}.
325: \end{array}
326: \right.
327: \label{rho}
328: \end{eqnarray}
329: In the case of chiral random matrix models, $\Sigma$ is the variance
330: of the distribution in Eq.~(\ref{Z}). In model I, $\Sigma$ is
331: proportional to $\Sigma_0$, Eq.~(\ref{DH}), in a manner that will be
332: discussed shortly. Note that thanks to the Banks-Casher
333: relationship~\cite{BanCas80}, $\Sigma$ is in all cases to be
334: identified with the chiral order parameter:
335: \begin{eqnarray}
336: \langle \bar\psi \psi\rangle = \lim_{\lambda \to 0}\,\, \lim_{N \to \infty}
337: \pi \rho(\lambda) = \Sigma.
338: \end{eqnarray}
339: Having obtained the semicircle distribution in model I is a natural result.
340: The random matrix interactions of Eq.~(\ref{W}) mix a set of $4 \times (N_c^2
341: - 1)$ independent real matrices, $A_{\mu a}$, in a democratic way. This
342: ensemble naturally leads to the semicircle law familiar from most elementary
343: random ensembles, including the chiral ones.
344:
345: Further remarks can be made about the dependence of $\Sigma$ with
346: respect to $N_c$. For $\chi$RMT, $\Sigma$ does not depend on $\beta$,
347: and hence does not depend on $N_c$. In model I, however, $\Sigma$ is
348: a function of $N_c$ and $\Sigma_0$ which can be easily determined by
349: noting the following relationship~\footnote{This can be established by
350: considering the resolvent operator $G(z) = \langle {\rm Tr} (z -
351: D)^{-1}\rangle$, and by matching its asymptotic expansion at large $z$
352: to the form it assumes for the spectral density of Eq.~(\ref{rho}).}
353: between the radius of the semicircle $2/\Sigma$ and the variance
354: $\langle {\rm Tr}[W W^\dagger]\rangle$:
355: \begin{eqnarray}
356: \langle {\rm Tr}[W W^\dagger]\rangle =
357: \frac{N}{\Sigma^2}.
358: \label{varChi}
359: \end{eqnarray}
360: From the definition of $W$, Eq.~(\ref{W}), and the distribution in
361: Eq.~(\ref{DH}), we have
362: \begin{eqnarray}
363: \langle {\rm Tr}[W W^\dagger]\rangle = 4 \sum_{\mu a}
364: \langle {\rm Tr} [A_{\mu a} (A_{\mu a})^T] \rangle =
365: \frac{2 (N_c^2 - 1)}{N_c^2}\,\frac{N}{\Sigma_0^2},
366: \label{SigmaSquared}
367: \end{eqnarray}
368: which gives
369: \begin{eqnarray}
370: \Sigma & = & \frac{N_c}{\sqrt{2 \,(N_c^2 - 1)}} \,\,\Sigma_0 =
371: \left\{\begin{array}{cc}
372: 0.75\,\Sigma_0 & \textrm{if}~N_c = 3, \\
373: \\
374: 0.8165\,\Sigma_0 & \textrm{if}~N_c = 2.
375: \end{array}
376: \right.
377: \label{SigmaColor}
378: \end{eqnarray}
379: The $N_c$ dependence of the chiral condensate reflects the fact that
380: the strength of the interactions in the chiral channel varies with
381: $N_c$. In fact, the $N_c$-dependent prefactor in
382: Eq.~(\ref{SigmaSquared}) is directly proportional to the Fierz
383: coefficient which appears when the random matrix interactions are
384: projected onto the chiral condensation channel (Ref.~\cite{VanJac99}).
385:
386: \section{The microscopic spectral density}
387: \label{s:microsd}
388:
389: The structures that are characteristic of chiral symmetry are most
390: clearly seen in the distributions of the eigenvalues near zero
391: virtuality. It is thus worth considering the microscopic spectral
392: density, defined at the level of single eigenvalues as
393: \begin{eqnarray}
394: \rho_S(z) = \lim_{N \to \infty} {1 \over 2 N \Sigma}\,
395: \rho\left({z \over 2 N \Sigma}\right),
396: \end{eqnarray}
397: where $\Sigma$ is the chiral order parameter of the theory at hand. The
398: microscopic spectral densities of the various chiral ensembles are strikingly
399: different. In the quenched limit, $\chi$GUE gives
400: \begin{eqnarray}
401: \rho_S(z) & = & {z \over 2}\,\left(J_0^2(z) + J_1^2(z)\right)
402: ~~~~(\textrm{$\chi$GUE}, N_f = 0, z >0),
403: \label{rhocGUE}
404: \end{eqnarray}
405: while $\chi$GOE gives~\cite{reviewchiRMM,ForNag99,KleVer00}
406: \begin{eqnarray}
407: \rho_S(z) & = & {z \over 2}\, \left( J_1^2(z) - J_0(z) \,J_2(z) \right)
408: - {1 \over 2} \, J_0(z) \, \left(\int_0^z du \,\,J_2(u) - 1 \right)
409: \nonumber \\
410: && (\textrm{$\chi$GOE}, N_f = 0, z >0).
411: \label{rhocGOE}
412: \end{eqnarray}
413:
414: Figures~1 and 2 show the histograms obtained in the models of
415: Refs.~\cite{VanJac99}, plotted against Eq.~(\ref{rhocGUE}) for the
416: model with $N_c = 3$ and against Eq.~(\ref{rhocGOE}) for that with
417: $N_c = 2$. The data were obtained from an ensemble of $2 \times 10^5$
418: matrices with blocks $W$ of size $120 \times 120$ (i.e., $2 n N_c =
419: 120$). As in $\chi$RMT, the behavior of the small eigenvalues depends
420: strongly on whether $N_c = 3$ or $N_c = 2$. In fact, the data agree
421: with the theoretical curves for the corresponding chiral ensembles to
422: expected statistical accuracy. The microscopic spectral density thus
423: appears to preserve the form dictated by chiral symmetry in spite of
424: the presence of additional color symmetries. In particular, we
425: observe the familiar depletion near $z = 0$ for $N_c =3$.
426:
427: The spectrum in Fig.~2 is flatter than that for $N_c = 3$. This is a
428: consequence of the greater motion of the individual eigenvalues in the
429: model with $N_c = 2$. This difference is familiar in chiral ensembles,
430: where it is interpreted as a consequence of the fact that the spectrum
431: of a real matrix ($\chi$GOE) is less rigid than that of a complex
432: matrix ($\chi$GUE). It is interesting to see that the level motion in
433: our model with $N_c = 2$ agrees with that expected for $\chi$GOE even
434: though the matrix elements of $W$ are no longer real. Again, the
435: correlations induced by chiral symmetry are maintained in spite of the
436: additional constraints introduced by the color structure of $W$.
437:
438: Of course, the preservation of the $N_c = 2$ microscopic spectral
439: density in spite of the complexity of $W$ should not come as a surprise.
440: The interactions in Eq.~(\ref{Zcolor}) are, in fact, pseudoreal when
441: $N_c = 2$~\cite{VanJac01}. Thus, we could have used an appropriate basis
442: of states such that all matrix elements have vanishing imaginary parts.
443: The model would then have displayed a greater similarity to a $\chi$RMT
444: model with $\beta = 1$, and it would thus be natural to find a spectrum
445: which reproduces that of $\chi$GOE.
446:
447: \section{Other spectral properties}
448: \label{s:others}
449:
450: We have also studied short range eigenvalue correlations by examining
451: the distribution of level spacings, $p(s)$. Here, $s$ is the nearest
452: neighbor spacing measured in units of the local average spacing.
453: (Hence, the average value of $s$ is $1$ by definition.) Note that
454: $p(s)$ is a bulk observable which is sensitive to chiral symmetry only
455: for small $s$. Figure 3 shows the relative difference between the
456: results obtained for the model of I with $N_c = 3$ and those of the
457: $\chi$GUE. The data were obtained from a set of nine independent
458: series of diagonalizations of 10000 matrices of size $N \times N = 120
459: \times 120$. To avoid side effects due to the finite range of the
460: random matrix support, we only kept a central portion of the spectrum
461: (30th eigenvalue to 90th). The level spacing distribution for a given
462: model is taken as a bin-by-bin average over the nine series of runs,
463: with error bars corresponding to the variance in each bin. The plot
464: shows $\delta p(s) = [p_{\chi{\rm GUE}}(s) - p_{N_c =
465: 3}(s)]/p_{\chi{\rm GUE}}(s)$, where $p_{\chi{\rm GUE}}(s)$ is the
466: level spacing distribution for $\chi {\rm GUE}$ and $p_{N_c = 3}(s)$
467: is that for model I with $N_c = 3$. [Note that $p_{\chi{\rm GUE}}(s)
468: = p_{{\rm GUE}}(s)$.] The level spacing difference is everywhere
469: consistent with zero. Similar results were obtained for a comparison
470: with $\chi$GOE. Thus, the level spacing distribution reveals no
471: statistically significant differences between the models of
472: Refs.~\cite{VanJac99} and their corresponding chiral ensembles.
473:
474: It is also useful to construct the variance-covariance matrix $D$,
475: defined as
476: \begin{equation}
477: D_{ij} = \langle (\lambda_i - {\bar \lambda_i})
478: (\lambda_j - {\bar \lambda_j}) \rangle,
479: \end{equation}
480: where ${\bar \lambda_i}$ is the mean value of eigenvalue $i$.
481: Numerical studies of the models of Refs.~\cite{VanJac99} indicate that
482: all eigenvalues of $D$ are greater than zero. Thus, the eigenvalues
483: of the Dirac operator do not satisfy any linear constraints. The
484: eigenvalues of $D$ are related to the statistically independent
485: fluctuations of the eigenvalues of the Dirac operator about their mean
486: values. (This relation would be exact if the joint probability
487: distribution for the eigenvalues of the Dirac operator were strictly
488: Gaussian.) Numerical studies reveal good agreement with the spectrum
489: of normal modes known analytically for the $\chi$GOE and
490: $\chi$GUE~\cite{OswSpl01}. This provides an additional indication
491: that the inclusion of color symmetry has not altered the statistical
492: properties of the models of Refs.~\cite{VanJac99} from those of the
493: corresponding chiral ensembles.
494:
495: \section{Conclusions}
496: \label{s:concl}
497:
498: We have investigated a number of properties of the eigenvalue spectrum
499: of random matrix models which include both chiral and color
500: symmetries. We find no deviation from the analytic results of
501: $\chi$RMT for either the microscopic spectral density or the level
502: spacing distribution. Given the relatively elaborate block structure
503: of these models, such studies are most easily performed numerically.
504: While more complicated spectral correlators have not been
505: investigated, we find no grounds to doubt that they will also reflect
506: the underlying chiral structure of the problem. (It should be
507: emphasized that this extended block structure does not lead to
508: complicated forms for the partition function and that studies of the
509: macroscopic properties of these models are straightforward for all $T$
510: and $\mu$.) The spin and color block structures of the interactions
511: do not seem to upset spectral features associated with chiral
512: symmetry. Rather, these additional correlations appear to act in a
513: channel which is ``orthogonal'' to the chiral channel. We do not
514: regard this result as surprising. One may think of the models of
515: Refs.~\cite{VanJac99} as schematic lattice calculations which
516: implement single-gluon exchange between discrete quark fields. This
517: ``lattice calculation'' does not possess any other length scale than
518: that introduced by $\Sigma$ and is, hence, free of a Thouless
519: energy. Because chiral symmetry is respected by the interactions, the
520: results of $\chi$RMT should be anticipated at all small energies.
521: This is indeed what our numerical analysis reveals.
522:
523: The inclusion of color does, of course, introduce additional
524: correlations in the spectrum of the Dirac operator. It would be of
525: interest to construct new spectral measures which could probe these
526: correlations. In particular, it would be valuable to find a color
527: analogue of the Banks-Casher relation to facilitate the determination
528: of the strength of the diquark condensate.
529:
530: \section*{Acknowledgments}
531:
532: We thank J. J. M. Verbaarschot for provocative questions which
533: initiated this work and D. Toublan and K. Splittorff for useful
534: discussions.
535:
536: \begin{thebibliography}{99}
537:
538: \bibitem{reviewchiRMM} For a review of $\chi$RMT, see
539: J. J. M. Verbaarschot and T. Wettig, Annu. Rev. Nucl. Part. Sci. {\bf
540: 50}, 343 (2000).
541:
542: \bibitem{chiRMM} E. V. Shuryak and J. J. M. Verbaarschot,
543: Nucl. Phys. {\bf A560}, 306 (1993); J. J. M. Verbaarschot,
544: Phys. Rev. Lett. {\bf 72}, 2531 (1994); Phys. Lett. B {\bf 329}, 351
545: (1994).
546:
547: \bibitem{VerZah93}
548: J. J. M. Verbaarschot and I. Zahed, Phys. Rev. Lett. {\bf 70}, 3852 (1993).
549:
550: \bibitem{Ver94}
551: J. J. M. Verbaarschot, Nucl. Phys. {\bf B427}, 534 (1994).
552:
553: \bibitem{Ver94-2}
554: J. J. M. Verbaarschot, Nucl. Phys. {\bf B426[FS]}, 559 (1994).
555:
556: \bibitem{JacVer96}
557: A. D. Jackson and J. J. M. Verbaarschot, Phys. Rev. D {\bf 53}, 7223 (1996).
558:
559: \bibitem{HalJac98}
560: M. A. Halasz, A. D. Jackson, R. E. Schrock, M. A. Stephanov, and
561: J. J. M. Verbaarschot, Phys. Rev. D {\bf 58}, 096007 (1998).
562:
563: \bibitem{AlfKap99} M. G. Alford, A. Kapustin, and F. Wilczek,
564: Phys. Rev. D {\bf 59}, 054502 (1999).
565:
566: \bibitem{ChaWie99} S. Chandrasekharan and U. J. Wiese,
567: Phys. Rev. Lett. {\bf 83}, 3116 (1999).
568:
569: \bibitem{FodKat02}
570: Z. Fodor and S. D. Katz, Phys. Lett. B {\bf 534}, 87 (2002);
571: J. High Energy Phys. {\bf 03}, 014 (2002).
572:
573: \bibitem{ForPhi02}
574: P. de Forcrand and O. Philipsen, Nucl. Phys. {\bf B642}, 290 (2002).
575:
576: \bibitem{early} B. Barrois, Nucl. Phys. {\bf B129}, 390 (1977); D.
577: Bailin and A. Love, Phys. Rep. {\bf 107}, 325 (1984).
578:
579: \bibitem{RapSch98}
580: R. Rapp, T. Sch{\"a}fer, E. V. Shuryak, and M. Velkovsky,
581: Phys.\ Rev.\ Lett.\ {\bf 81}, 53 (1998); Ann. Phys. (N.Y.) {\bf 280},
582: 35 (2000).
583:
584: \bibitem{AlfRaj98}
585: M. Alford, K. Rajagopal, and F. Wilczek, Phys.\ Lett.\ B {\bf 422}, 247 (1998).
586:
587: \bibitem{review} For a review on color superconductivity, see
588: K. Rajagopal and F. Wilczek, {\it B. L. Ioffe Festschrift, At the frontier
589: of particle physics/Handbook of QCD}, edited by M. Shifman (World
590: Scientific, 2001), Vol. 3, p. 2061 hep-ph/0011333.
591:
592: \bibitem{VanJac99}
593: B. Vanderheyden and A. D. Jackson, Phys.~Rev. D~{\bf 61}, 076004 (2000);
594: {\bf 62}, 094010 (2000).
595:
596: \bibitem{VanJac01}
597: B. Vanderheyden and A. D. Jackson,
598: Phys.~Rev. D{\bf 64}, 074016 (2001).
599:
600: \bibitem{KogSteTou99} J. B. Kogut, M. A. Stephanov, and D. Toublan,
601: Phys. Lett. B {\bf 464}, 183 (1999).
602:
603: \bibitem{KogSteTou00} J. B. Kogut, M. A. Stephanov, D. Toublan,
604: J. J. M. Verbaarschot, and A. Zhitnitsky, Nucl. Phys. {\bf B582}, 477 (2000).
605:
606: \bibitem{KimSon01}
607: K. Splittorff, D. T. Son, and M. A. Stephanov, Phys. Rev. D {\bf 64},
608: 016003 (2001).
609:
610: \bibitem{KimVer01}
611: K. Splittorff, D. Toublan, and J. J. M. Verbaarschot,
612: Nucl. Phys. {\bf B620}, 290 (2002).
613:
614: \bibitem{BanCas80}
615: T. Banks and A. Casher, Nucl. Phys. {\bf B169}, 103 (1980).
616:
617: \bibitem{ForNag99}
618: P. J. Forrester, P. Nagao, and G. Honner, Nucl. Phys. {\bf B533}, 601 (1999).
619:
620: \bibitem{KleVer00}
621: B. Klein and J. J. M. Verbaarschot, Nucl. Phys. {\bf B588}, 483 (2000).
622:
623: \bibitem{OswSpl01}
624: A. D. Jackson, C. B. Lang, M. Oswald, and K. Splittorff,
625: Nucl. Phys. {\bf B616}, 233 (2001).
626:
627:
628: \end{thebibliography}
629:
630:
631:
632: \newpage
633:
634: \begin{center}
635: \begin{figure}
636: \includegraphics{sd_fig1.eps}
637: \caption{Histogram: microscopic spectral density of the Dirac operator for a
638: random matrix model with a color subblock and $N_c = 3$. Solid line:
639: spectral density predicted for $\chi$GUE ($N_f \to 0$).}
640: \end{figure}
641: \end{center}
642:
643: \newpage
644:
645: \begin{center}
646: \begin{figure}
647: \includegraphics{sd_fig2.eps}
648: \caption{Histogram: microscopic spectral density of the Dirac operator for a
649: random matrix model with a color subblock and $N_c = 2$. Solid line:
650: spectral density predicted for $\chi$GOE ($N_f \to 0$).}
651: \end{figure}
652: \end{center}
653:
654: \newpage
655:
656: \begin{center}
657: \begin{figure}
658: \includegraphics{sd_fig3.eps}
659: \caption{Relative difference between the mean level spacing distribution
660: for $\chi$GUE and that for a model with a color subblock and $N_c = 3$.
661: The errorbars are estimated from a series of nine independent
662: runs of diagonalizations.}
663: \end{figure}
664: \end{center}
665:
666: \end{document}
667:
668:
669:
670:
671:
672:
673:
674:
675:
676:
677: