1: %\documentstyle[prl,aps,preprint,psfig]
2: %\tightenlines
3: %\documentstyle[galley,prl,aps,psfig]{revtex}
4: %\documentstyle[preprint,eqsecnum,aps,epsfig]{revtex}
5: %\newlength{\figwidth}
6: %\setcounter{equation}{0}
7: %\setlength{\figwidth}{3.4in}
8:
9: %\documentstyle[preprint,eqsecnum,osa,epsfig]{revtex}
10: \documentstyle[12pt,epsf]{article}
11: \renewcommand\baselinestretch{1.2}
12: \renewcommand{\theequation}{\arabic{section}.\arabic{equation}}
13: \renewcommand{\thefigure}{\arabic{figure}}
14: \parindent 30pt
15: \textheight 8.5in
16: \textwidth 6in
17:
18: \title{Quantum thermodynamic fluctuations of a chaotic Fermi--gas model}
19:
20: \author{P. Leboeuf$~ ^1$ \ and A. G. Monastra$^2$}
21:
22: \newcommand{\tauH}{\tau_{\scriptscriptstyle H}}
23: \newcommand{\kb}{k_{\scriptscriptstyle B}}
24: \newcommand{\ef}{E_{\scriptscriptstyle F}}
25: \newcommand{\leqa}{\stackrel{<}{\scriptstyle \sim}}
26: \newcommand{\geqa}{\stackrel{>}{\scriptstyle \sim}}
27: \newcommand{\tmin}{\tau_{\rm min}}
28: \newcommand{\Tc}{T_{\rm c}}
29: \newcommand{\Ec}{E_{\rm c}}
30:
31: \begin{document}
32: \maketitle
33:
34: \begin{center}
35: \small{1. \ Laboratoire de Physique Th\'eorique et Mod\`eles
36: Statistiques,\footnote[3]{Unit\'e de Recherche de l'Universit\'e Paris XI
37: associ\'ee au CNRS} \\ B\^at. 100, Universit\'e de Paris-Sud, 91405
38: Orsay Cedex, France}
39: \end{center}
40:
41: \begin{center}
42: \small{2. \ Department of Physics of Complex Systems, The Weizmann
43: Institute of Science,\\ 76100 Rehovot, Israel}
44: \end{center}
45:
46: \baselineskip 0.25in
47:
48: \hspace{1.5in}
49:
50: \begin{abstract}
51: We investigate the thermodynamics of a Fermi gas whose single--particle energy
52: levels are given by the complex zeros of the Riemann zeta function. This is a
53: model for a gas, and in particular for an atomic nucleus, with an underlying
54: fully chaotic classical dynamics. The probability distributions of the quantum
55: fluctuations of the grand potential and entropy of the gas are computed as a
56: function of temperature and compared, with good agreement, with general
57: predictions obtained from random matrix theory and periodic orbit theory
58: (based on prime numbers). In each case the universal and non--universal
59: regimes are identified.
60: \end{abstract}
61:
62: \vspace{2cm}
63:
64: \noindent PACS numbers: 21.60.-n, 24.60.Lz, 05.30.Fk, 05.45.Mt \\
65: \noindent Keywords: Fermi gas, statistical fluctuations.
66:
67: \pagebreak
68:
69: \section{Introduction}
70: \label{sec:Intro}
71: \setcounter{equation}{0}
72:
73: Fermi gases confined to small volumes present distinct finite size effects in
74: their thermodynamic properties. Among the different corrections to the bulk
75: behavior, some are smooth functions of the parameters of the system. They
76: typically depend on geometric aspects of the volume occupied by the gas (e.g.
77: surface and curvature terms). In contrast, other corrections are fluctuating
78: functions. These quantum (or mesoscopic) fluctuations are related to the
79: discreteness of the single--particle spectrum, and are sensitive to the nature
80: of the dynamics of the particles in the gas. In many cases they are small
81: compared to the bulk, but may play nevertheless an important role (like, for
82: example, in nuclear physics where shell effects are crucial in the
83: determination of the shape of atomic nuclei \cite{bm,sm}). In other (more
84: spectacular) situations, they provide the dominant contribution. This occurs
85: for example in the electronic orbital magnetism in quantum dots -- where they
86: can be larger than the Landau susceptibility \cite{shapiro,voppen,ruj} --, or
87: in the persistent currents in mesoscopic rings \cite{agi,sch} -- where the
88: bulk contribution vanishes. In all cases the quantum fluctuations disappear
89: when the temperature is raised.
90:
91: Instead of a detailed computation and description of the quantum fluctuations
92: for a particular system, the aim here is to study their statistical
93: properties. The interest of such an analysis is well known: using a minimum
94: amount of information, a statistical approach allows to establish a
95: classification scheme among the fluctuations of different physical systems. It
96: also allows to distinguish the generic from the specific, and provides a
97: powerful predictive tool in complex systems. A general description of the
98: statistical properties of the quantum fluctuations of thermodynamic functions
99: of integrable and chaotic ballistic Fermi gases, and of their temperature
100: dependence, was developed in Ref.\cite{lm3}.
101:
102: From a classical point of view, there are two extreme cases of
103: single--particle motion, namely integrable and fully chaotic. The case of a
104: mixed phase space (e.g., coexistence of regular and chaotic motion) being the
105: generic one. In all cases, in a Fermi gas model the quantum thermodynamic
106: fluctuations can be directly related to the fluctuations of the
107: single--particle, discrete, energy levels. Schematically, there are two
108: distinct types of single--particle fluctuations. On the one hand there are the
109: local fluctuations, i.e. those occurring on scales of the order of the
110: single--particle mean level spacing $\delta$. For the two extreme dynamics
111: mentioned above, these fluctuations are known to be universal \cite{bg},
112: namely Poisson for integrable motion, random matrix in full chaos. The second
113: type of fluctuations generically present in any single--particle spectrum are
114: long range correlations, that occur on a scale $E_c$, where $\Ec$ is the
115: energy associated with the shortest classical periodic orbit of the mean field
116: (inverse time of flight across the system). In contrast to the local
117: fluctuations, the long range modulations depend on the short--time specific
118: properties of the dynamics and are therefore non--universal, either for
119: integrable, mixed, or fully chaotic dynamics (thought their importance varies
120: in each case).
121:
122: Properties related to both types of single--particle fluctuations have been
123: identified in the nuclear behavior. For example, shell effects in the nuclear
124: masses are a clear manifestation of the long range correlations, whereas
125: nearest--neighbor spacing statistics and the distribution of widths of neutron
126: resonances illustrate universal local fluctuations. However, a global and
127: comprehensive picture of the quantum fluctuations has not yet emerged. An
128: important additional ingredient are certainly the variations in the nucleus of
129: the nature of the dynamics as a function of the excitation energy. But even in
130: the case of a scaling mean--field motion where the classical dynamics is
131: energy independent (like when putting a Fermi gas inside a hard wall box or
132: billiard) the physical behavior of the gas can be quite complex and rich, and
133: both types of fluctuations (local and long range) can in fact manifest in
134: different quantities related to the system, or in the same quantity at
135: different temperatures \cite{lm3}.
136:
137: One might think that a fully chaotic single--particle dynamics implies
138: universal quantum fluctuations of the thermodynamics of the gas, well
139: described by some random matrix model (Wigner, two--body random, etc). This
140: belief is not true in general. The reason is that the thermodynamic
141: fluctuations are universal only when they are controlled by the local
142: single--particle fluctuations. And this is not always the case. For example,
143: it has been shown, in particular, that the statistical properties of the
144: ground--state energy of a many--body system are controlled by the short
145: periodic orbits (at least in systems with a ground state well described by a
146: mean--field approximation, like the atomic nucleus) \cite{lm3}. The
147: corresponding distribution is therefore non--universal. As a consequence, it
148: makes no sense to model the ground--state fluctuations by some random matrix
149: model, that possess no information on the specific mean--field short time
150: properties of the dynamics.
151:
152: Our purpose here is to illustrate the richness of the thermodynamic quantum
153: fluctuations and to explicitly check some of the predictions by considering a
154: particular example of a Fermi gas with a fully chaotic classical dynamics. The
155: system considered is a mathematical model. It is achieved by considering a
156: spectrum of single--particle energy levels constructed from the imaginary part
157: of the complex zeros of the Riemann zeta function. The Fermi gas is obtained
158: by filling the single--particle energy levels with an average occupation
159: number determined by the Fermi--Dirac distribution. This model of a chaotic
160: gas was introduced in Ref.\cite{blm1}, where general motivations as well as
161: dynamical analogies were given. If this Fermi--gas model is viewed as an
162: element of the periodic table, following nuclear physics terminology we
163: referred to it as the {\it Riemannium} \cite{blm1}. Aside its mathematical
164: interest, physically the Riemannium is an important model because it possesses
165: all the typical dynamical properties and characteristics of a realistic
166: chaotic gas, while it greatly facilitates the numerical and analytical
167: computations (e.g., all the ingredients required in a semiclassical
168: description of its thermodynamic properties are known). This is quite unusual
169: in fully chaotic systems, and therefore offers a rather unique model to test
170: general results and ideas.
171:
172: The basic quantities to be computed are the probability distributions, at a
173: given temperature, of the quantum fluctuations of the grand potential and of
174: the entropy of the gas. We consider a degenerate gas in the grand canonical
175: ensemble; other ensembles (e.g., canonical) may be considered as well. In
176: particular, it can be shown that to leading order in an expansion of large
177: chemical potential, the fluctuations of the energy of the gas at a fixed
178: number of particles coincide with those of the grand potential. Our results
179: therefore apply to the total energy. In a semiclassical approximation, the
180: fluctuations are described by the periodic orbits of the classical dynamics
181: associated with the single--particle motion \cite{bb,gutz,sm,brack}. In the
182: Riemannium, the role of the periodic orbits is played by the prime numbers.
183: The basic equations and the relevant energy scales of the Riemannium are
184: introduced in \S \ref{sec:Maineq}. In \S \ref{sec:Omega} we analyze the
185: fluctuations of the grand potential. The corresponding distribution is
186: non--universal at all temperatures (including $T=0$, the case considered in
187: \cite{blm1}). The asymptotic moments of the distribution, obtained for a large
188: chemical potential, are well reproduced by some convergent sums dominated by
189: the smaller prime numbers (i.e., short periodic orbits). Odd moments are
190: non--zero, implying an asymmetric distribution at all temperatures. At a
191: finite chemical potential, large prime numbers introduce corrections to the
192: asymptotic moments. These corrections to the non--universal leading--order
193: behavior are, in contrast, universal, and described by random matrix theory.
194:
195: The entropy fluctuations are studied in \S \ref{sec:Entropy}. Their character
196: strongly depends on temperature. For a chaotic dynamics without time reversal
197: symmetry, which corresponds to the Riemannium, these fluctuations present
198: several remarkable features, predicted in \cite{lm3}. For temperatures $T \ll
199: \delta$, the variance of the entropy fluctuations increases linearly. In a
200: second regime, when $\delta \ll T \ll \Ec$, the variance saturates to a
201: constant value. This initial behavior of the variance (i.e. linear growth
202: followed by a saturation), and more generally of the probability distribution
203: of the entropy fluctuations, was predicted to be universal, namely identical
204: for any chaotic system without time reversal symmetry (with a given
205: single--particle mean level spacing at Fermi energy). It is well described by
206: the fluctuations obtained from a GUE random matrix single--particle spectrum.
207: The shape of the probability distribution strongly depends on temperature, and
208: converges towards a Gaussian as temperature increases. However, this universal
209: behavior disappears for temperatures of the order $\Ec$ or higher, where the
210: statistical properties of the gas do not coincide any more with random matrix
211: theory. The probability distribution of the fluctuations is now described by
212: the short periodic orbits, which are system dependent. In this high
213: temperature regime the typical size of the fluctuations shows an overall
214: exponential decay. All these predictions are worked out explicitly for the
215: Riemannium and compared with numerical simulations. We have found very good
216: agreement between theory and numerical experiments in all the regimes.
217:
218: Although only two particular thermodynamic functions are discussed, they
219: illustrate all the rich possibilities that may be encountered in other
220: functions. Similar methods apply to them.
221:
222: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
223:
224: \section{Thermodynamic framework}
225: \label{sec:Maineq}
226: \setcounter{equation}{0}
227:
228: In the grand canonical ensemble the equilibrium properties of a Fermi gas are
229: defined by the grand potential
230: \begin{equation}
231: \Omega (\mu,T) = - T \int dE ~ \log [1+e^{(\mu-E)/T}] ~ \rho(E) \ ,
232: \label{omegint}
233: \end{equation}
234: where $T$ is the temperature (we set Boltzmann's constant $k_B =1$), $\mu$ the
235: chemical potential, and
236: \begin{equation}
237: \rho (E) = \sum_j \delta (E - E_j)
238: \end{equation}
239: is the single--particle density of states. In the Riemannium, the
240: single--particle energies are given by the imaginary part of the upper--half
241: complex zeros $s=1/2 + i E_j$ of the Riemann zeta function $\zeta (s)$, $ E_j
242: = 14.1347, 21.0220, 25.0109, \ldots$ (the values used in all our numerical
243: simulations were obtained from \cite{odlyzko}). They are expressed in
244: arbitrary units, which are also those of temperature (since $k_B =1$). From
245: the grand potential, other thermodynamic functions can be computed. We are
246: particularly interested in the behavior of the entropy of the gas,
247: \begin{equation} \label{Sdef}
248: S(\mu,T) = -\left( \frac{\partial \Omega}{\partial T} \right)_{\mu} \ .
249: \end{equation}
250:
251: The number of complex zeros with imaginary part less than $E$ is
252: \cite{edwards},
253: \begin{equation}
254: n(E) = 1 - \frac{1}{2\pi} E \log \pi - \frac{1}{\pi} Im \log \Gamma \left(
255: \frac{1}{4} - i \frac{E}{2} \right) - \frac{1}{\pi} Im \log \zeta \left(
256: \frac{1}{2} - i ~ E \right) \ . \label{nzexact}
257: \end{equation}
258: Expanding the $\Gamma$ function and using the Euler product representation of
259: $\zeta$ in the last term, this expression splits naturally into a smooth plus
260: an oscillatory part, as in the semiclassical approximation in quantum systems.
261: The density of zeros $\rho (E) = d n / d E$ is written $ \rho (E) =
262: \overline{\rho} (E) + \widetilde{\rho} (E) $, with
263: \begin{eqnarray} \label{rhobar}
264: \overline{\rho} (E) &=& \frac{1}{2\pi} \log \left( \frac{E}{2\pi}
265: \right) + {\cal O} (E^{-2}) \ , \\
266: \widetilde{\rho} (E) &=& - \frac{1}{\pi} \sum_p \sum_{r=1}^{\infty}
267: \frac{\log p}{p^{r/2}} \cos( E ~ r ~ \log p) \ . \label{rhoosc}
268: \end{eqnarray}
269: In the Riemannium, the first term corresponds therefore to the asymptotic
270: average density of the single--particle energy levels. In the fluctuating part
271: $\widetilde{\rho}$, the sum is made over the prime numbers $p$. The comparison
272: of this equation with the semiclassical Gutzwiller trace formula for the
273: spectral density of a dynamical system \cite{gutz}, $\widetilde{ \rho } (E) =
274: \sum_{{\rm p.o.}} A_{{\rm p.o.}} \cos (S_{{\rm p.o.}}(E)/\hbar)$ (the sum is
275: over the classical periodic orbits), shows that each prime number may be
276: interpreted as the label of an unstable primitive periodic orbit $p$ of a
277: fully chaotic system without time reversal symmetry, of action $S_p = E \,
278: \tau_p$, period $\tau_p = \log p$, Lyapounov stability $\lambda_p = 1$,
279: repetitions labeled by $r$, and $\hbar=1$ \cite{berry2,foot}. Because of this
280: mapping, in the following we often refer to the prime numbers as periodic
281: orbits. Putting aside the issue of the very existence of a classical system
282: whose quantum eigenvalues coincide with the imaginary part of the Riemann
283: zeros, it is the formal analogy with the semiclassical theory of dynamical
284: systems that makes the Riemann zeros in general, and the Riemannium in
285: particular, interesting on a physical basis as a model to test general
286: theories of quantum chaotic systems. Additional support for a dynamical
287: interpretation comes from numerical simulations, that clearly indicate
288: \cite{odlyzko} that the sequence of the imaginary part of the Riemann zeros
289: obeys the Bohigas--Giannoni--Schmit conjecture for chaotic systems without
290: time--reversal symmetry \cite{bgs,bg}: asymptotically, their statistics
291: coincide with those of eigenvalues of the GUE ensemble of random matrices
292: \cite{mont}.
293:
294: Replacing the decomposition (\ref{rhobar})--(\ref{rhoosc}) of the density of
295: states into (\ref{omegint}) leads to a corresponding decomposition of the
296: grand potential. A similar decomposition for other thermodynamic functions is
297: obtained by derivation. We are interested in the statistical properties of the
298: fluctuating part of those functions, and in their temperature dependence. In
299: the forthcoming analysis, the different relevant scales are the following.
300: \begin{itemize}
301: \item[i)] Largest scale: chemical potential $\mu$
302: \item[ii)] Intermediate scale: shortest periodic orbit
303: \begin{eqnarray}\label{ec}
304: &~& \hspace*{-2cm} {\rm Energy:} \ \Ec = h / \tmin = 2 \pi / \log 2
305: \approx 9.06472 \ , \nonumber \\ &~& \hspace*{-2cm} {\rm Time:} \ \tmin =
306: \log 2 \ , \\ &~& \hspace*{-2cm} {\rm Temperature:} \ \Tc = \Ec / 2 \pi^2
307: \approx 0.459224 \nonumber \ .
308: \end{eqnarray}
309: \item[iii)] Smallest scale: single--particle mean level spacing
310: \begin{eqnarray}\label{delta}
311: &~& \hspace*{-2cm} {\rm Energy:} \ \delta = 1 / \overline{\rho} = 2 \pi / \log
312: (\mu/2\pi) \ , \nonumber \\
313: &~& \hspace*{-2cm} {\rm Time:} \ \tauH = h/\delta = \log (\mu/ 2\pi) \ , \\
314: &~& \hspace*{-2cm} {\rm Temperature:} \ T_{\delta} = \delta / 2 \pi^2 =
315: 1 / \pi \log (\mu/2\pi) \nonumber \ .
316: \end{eqnarray}
317: \end{itemize}
318: The intermediate scale is related to the shortest periodic orbit (or time of
319: flight across the system at energy $\mu$) which, according to the previous
320: analogy, corresponds here to the period $\tmin = \log 2$ associated with the
321: prime number 2. $\Ec$ defines the outer energy scale of oscillations in the
322: density of states (cf Eq.(\ref{rhoosc})). Due to the independence of the
323: periods with energy in the Riemann dynamics, this energy scale is a fixed
324: number. In the definition of the mean level spacing $\delta$, we have used the
325: asymptotic leading order approximation of the average density of states.
326: $\tauH$ is the so--called Heisenberg time. The $2 \pi^2$ factor (approximately
327: 20) that relates energies and temperatures is dictated by semiclassical
328: arguments \cite{lm3}, and it has to be taken into account in numerical
329: comparisons. A final important parameter is the ratio of $\Ec$ to $\delta$,
330: which measures the number of single--particle states contained in a scale
331: $\Ec$,
332: $$
333: g=\Ec /\delta = \log (\mu/2\pi)/\log 2 \ .
334: $$
335: In the semiclassical limit $\mu \rightarrow \infty$, the three energy scales
336: are well separated, $ \delta \ll \Ec \ll \mu$, and $g \rightarrow \infty$.
337:
338: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
339:
340: \section{The grand potential}
341: \label{sec:Omega}
342: \setcounter{equation}{0}
343:
344: Replacing $\rho (E)$ by $\overline{\rho}$ in Eq.(\ref{omegint}), integrating
345: by parts, and applying Sommerfeld's approximation (valid when $T \ll \mu$), we
346: obtain the usual thermodynamic relation
347: \begin{equation}
348: \overline{\Omega} ( \mu , T ) = \overline{\Omega}_0 (\mu) -
349: \frac{\pi^2}{6} ~ \overline{\rho} (\mu) T^2 \ , \label{omegliso}
350: \end{equation}
351: where $\overline{\Omega}_0$ is the grand potential at zero temperature. For
352: the Riemannium
353: $$
354: \overline{\Omega}_0 (E) = -\frac{E^2}{4 \pi} \log \left( \frac{E}{2\pi} \right)
355: + \frac{3}{8 \pi} E^2 - \frac{7}{8} E - \frac{\log E}{48 \pi} + c + {\cal O}
356: (E^{-2}) \ .
357: $$
358: Terms of order up to $E^{-2}$ in $\bar{\rho}$ have been included in the
359: integration. The determination of the constant $c$ is of numerical importance
360: when computing the fluctuations. From the work of Selberg on the function
361: $S_1(t)$ \cite{selberg} (which, up to an overall sign, coincides with the
362: fluctuating part of the grand potential at zero temperature), we can extract
363: its value
364: $$
365: c = \frac{1}{\pi} \int_{1/2}^{\infty} \log \left| \zeta (x) \right| dx
366: - \lim_{E \rightarrow \infty} \left\{ \int_{0}^{E} \overline{n} (x) \ dx
367: - \frac{E^2}{4 \pi} \log \left( \frac{E}{2\pi} \right) + \frac{3}{8
368: \pi} E^2 - \frac{7}{8} E - \frac{\log E}{48 \pi} \right\} ,
369: $$
370: where $\overline{n} (x)$ is the exact smooth part of the counting function,
371: computed from equation (\ref{nzexact}). Evaluating numerically the integrals,
372: the value of the constant is found to be $c \approx 0.75575$.
373:
374: Replacing $\rho (E)$ by (\ref{rhoosc}) in Eq.(\ref{omegint}), the oscillating
375: part of the grand potential is
376: \begin{equation}
377: \widetilde{\Omega} ( \mu , T ) = - \frac{1}{\pi} \sum_p
378: \sum_{r=1}^{\infty} \frac{\kappa ( \pi \ T \ r \log p) }{ r^2 \ p^{r/2}
379: \log p} \cos( \mu \ r \log p) \ . \\ \label{Oosc}
380: \end{equation}
381: The function
382: $$
383: \kappa (x) = x / \sinh x
384: $$
385: takes into account temperature effects \cite{sm}, and cuts exponentially the
386: contribution of large prime numbers. If $T > \Tc$, the contribution of each
387: orbit, including the shortest one, is exponentially small.
388:
389: Equation (\ref{Oosc}) is an oscillatory function of $\mu$ that describes the
390: quantum fluctuations of the grand potential. To study their statistical
391: properties we define the average
392: \begin{equation}
393: \langle f \rangle \equiv \frac{1}{\Delta \mu} \int_{\mu - \Delta \mu /2}^{
394: \mu + \Delta \mu / 2 } f(\mu') ~ d \mu' \ , \label{mean}
395: \end{equation}
396: for any thermodynamic function $f(\mu)$ associated to the Riemannium. The size
397: of the average window $\Delta \mu$ has to be much smaller than $\mu$ (in order
398: to extract local properties), but large enough to contain a sufficient number
399: of oscillations to make the statistical approach meaningful, i.e. $ \Ec \ll
400: \Delta \mu \ll \mu $.
401:
402: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
403:
404: \subsection{Variance}\label{secvaro}
405:
406: The most basic feature of the probability distribution of the fluctuations is
407: the variance. The average of the square of $\widetilde{\Omega}$ can be written
408: as an integral \cite{lm3}
409: \begin{equation}
410: \langle \widetilde{\Omega}^2 \rangle (\mu , T ) = \frac{1}{2 \pi^2}
411: \int_{0}^{\infty} d\tau ~ \frac{ \kappa^2 ( \pi T \tau ) }{ \tau^4 } ~
412: K (\tau) \ . \label{varint}
413: \end{equation}
414: The function
415: \begin{equation} \label{ff}
416: K (\tau) = \left\langle \sum_{i,j} A_i A_j \cos[ \mu (r_i \log p_i - r_j
417: \log p_j) ] \delta ( \tau - \bar{\tau} ) \right\rangle \ ,
418: \end{equation}
419: is the form factor, i.e. the Fourier transform of the two--point correlation
420: function of the single--particle spectrum; $A_i = \log p_i/ p_i^{r_i / 2} $,
421: and $ \bar{\tau} = ( \tau_{p_i}+ \tau_{p_j} ) / 2 $. This function is central
422: in our analysis. In the semiclassical limit $g \rightarrow \infty$ it is
423: conjectured to coincide with the GUE form factor
424: \begin{equation}\label{kgue}
425: K_{GUE} (\tau) = {\rm min} \{ \tau, \tauH \} \ .
426: \end{equation}
427: If we naively replace in (\ref{varint}) $K (\tau)$ by $K_{GUE} (\tau)$, the
428: integral diverges at $\tau \rightarrow 0$. This is because at finite $g$
429: random matrix theory provides a wrong description of the short--time behavior
430: of the form factor. For low values of $\tau$ the diagonal terms in
431: Eq.(\ref{ff}) should be used \cite{berry1},
432: \begin{equation}\label{kd}
433: K_D (\tau) = \sum_{i} A_i^2 \ \delta ( \tau - \tau_i ) \ .
434: \end{equation}
435: This shows that the form factor is non--universal at short times, and that
436: $K(\tau) = 0$ for $\tau < \tmin$. The form factor behaves like $K_{GUE}$ only
437: for times larger than some $\tau_*$. The intermediate time $ \tau_* $
438: satisfies $({\rm Re} \gamma)^{-1} < \tau_* < \tauH $, where $\gamma$ is the
439: classical resonance (eigenvalue of the Perron--Frobenius operator) with the
440: smallest real part \cite{agam}.
441:
442: Since short times dominate the fluctuations of the grand potential, in the
443: limit $\tau \rightarrow 0$ we use in (\ref{varint}) the diagonal approximation
444: of the form factor, and neglect for the moment the contributions of longer
445: times,
446: \begin{equation}
447: \langle \widetilde{\Omega}^2 \rangle_{{\rm D}} = \frac{1}{2 \pi^2}
448: \sum_{ p , r} \frac{ \kappa^2 ( \pi T \ r \log p ) }{ r^4 ~ p^r ~
449: \log^2 p } \ . \label{varodiag}
450: \end{equation}
451: This is a convergent sum for all temperatures, and is independent of the
452: chemical potential. At $T = 0$ it gives the value $\langle
453: \widetilde{\Omega}^2 \rangle_{{\rm D}} \approx 0.079290$ \cite{blm1}. Since
454: the temperature factor is a monotonically decreasing function of $\tau$, the
455: variance exponentially decreases with increasing temperature, and the role of
456: the short orbits (small primes) in the fluctuations is further enhanced. Fig.1
457: compares Eq.(\ref{varodiag}) evaluated at different temperatures to numerical
458: results obtained for the Riemannium (for reference, in the chemical potential
459: window used to compute the fluctuations we have $T_\delta \approx 0.013$).
460:
461: The agreement with the leading order description of the variance is excellent.
462: However, as the chemical potential is lowered deviations from the asymptotic
463: behavior are observed. The corrections may be obtained from Eq.(\ref{varint}),
464: $\langle \widetilde{\Omega}^2 \rangle = \langle \widetilde{\Omega}^2
465: \rangle_{{\rm D}} + \langle \widetilde{\Omega}^2 \rangle_{{\rm off}}$, by
466: using the GUE form factor (\ref{kgue}) for times $\tau > \tau_*$. This gives,
467: \begin{equation}
468: \langle \widetilde{\Omega}^2 \rangle_{{\rm off}} = \frac{1}{2 \pi^2}
469: \int_{\tauH}^{\infty} d\tau ~ \frac{ (\tauH - \tau ) ~ \kappa^2 ( \pi
470: T \tau ) }{ \tau^4 } \ .
471: \end{equation}
472: As temperature goes to zero, the temperature factor $\kappa^2$ tends to 1, and
473: the integral is easily performed giving $ \langle \widetilde{\Omega}^2
474: \rangle_{{\rm off}} = - 1 / (12 \pi^2 \tauH^2) $. This provides a
475: finite--$\mu$ correction to the variance, which decreases as $(\log
476: \mu)^{-2}$. For increasing temperatures the correction is smaller. The
477: off-diagonal terms thus provide a universal correction to the leading non
478: universal diagonal term (namely, the only specific information on the system
479: that enters the correction is the average density of states at Fermi energy,
480: through $\tauH = h \overline{\rho}$).
481:
482: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
483:
484: \subsection{Higher moments and distribution}
485: \label{sec:moments}
486:
487: We already showed that to calculate the variance the diagonal approximation is
488: accurate because the short orbits dominate the fluctuations. A generalization
489: of the diagonal approximation allows to evaluate all the moments of the
490: probability distribution of the grand potential by a method developed in
491: \cite{lm3} and \cite{blm1}. The mechanism is an interference process between
492: repetitions of primitive periodic orbits.
493:
494: Defining the amplitude
495: \begin{equation}\label{apr}
496: {\cal A}_{p,r} (T) = - \frac{1}{2 \pi} \frac{\kappa ( \pi \ T \ r \log
497: p) }{ r^2 \ p^{r/2} \log p} \ ,
498: \end{equation}
499: the third and fourth moments of the distribution of $\widetilde{\Omega}$ are
500: found to be, in the diagonal approximation,
501: \begin{eqnarray}
502: \langle \widetilde{\Omega}^3 \rangle &=& 6 ~ \sum_p \sum_{r_1, r_2 =
503: 1}^{\infty} {\cal A}_{p,r_1} ~ {\cal A}_{p,r_2} ~ {\cal A}_{p , r_1 +
504: r_2} \ , \label{m3} \\
505: \langle \widetilde{\Omega}^4 \rangle &=& 2 ~ \sum_p \Big[ ~ 4
506: \sum_{r_1, r_2, r_3 = 1}^{\infty} {\cal A}_{p, r_1} ~ {\cal A}_{p,
507: r_2} ~ {\cal A}_{p, r_3} ~ {\cal A}_{p, r_1 + r_2 + r_3} - ~ 6
508: \sum_{r_1, r_2=1}^{\infty} {\cal A}^2_{p, r_1} ~ {\cal A}^2_{p, r_2}
509: \nonumber \\
510: & ~ & + ~ 3 \sum_{r_1, r_2 = 1}^{\infty} \sum_{r_3 = 1}^{r_1 + r_2 -
511: 1} {\cal A}_{p, r_1} ~ {\cal A}_{p, r_2} ~ {\cal A}_{p, r_3} ~ {\cal
512: A}_{p, r_1 + r_2 - r_3} \Big] + 3 ~ \langle \widetilde{\Omega}^2
513: \rangle^2 . \label{m4}
514: \end{eqnarray}
515: All the sums are convergent. We see explicitly that the third moment and the
516: excess of the fourth moment (the term $3 ~ \langle \widetilde{\Omega}^2
517: \rangle^2$ being the fourth moment of a Gaussian distribution) are different
518: from zero. The distribution of the fluctuations of the grand potential is
519: therefore a non universal asymmetric function that strongly depends on the
520: small prime numbers. It is displayed, for different temperatures, in Fig.2
521: \cite{foot2}.
522:
523: Higher moments $k \geq 5$ may also be computed, with increasing complexity of
524: the result. Corrections to the diagonal terms coming from off--diagonal
525: contributions exist, but as already shown they are negligible in the limit
526: $\mu \rightarrow \infty$. We thus ignore them.
527:
528: These results are confirmed to high accuracy by the numerical data from the
529: Riemann zeros (cf Fig.1 and Table 1).
530:
531: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
532:
533: \section{The entropy}
534: \label{sec:Entropy}
535: \setcounter{equation}{0}
536:
537: From Eq.(\ref{Sdef}) the entropy of the gas is expressed as,
538: \begin{equation}
539: S(\mu,T) = \int s \left( \frac{\mu - E}{T} \right) \rho (E) ~ dE =
540: \sum_j s \left( \frac{\mu - E_j}{T} \right) \ ,
541: \label{Sint}
542: \end{equation}
543: where
544: \begin{equation}
545: s (x) = \log ( 1 + e^x ) - \frac{x}{1+e^{-x}} \ .
546: \end{equation}
547: The function $s(x)$ is localized around $x=0$, and decays exponentially for
548: $|x| \gg 1$. At very low temperatures ($T \ll T_\delta$), the function
549: $S(\mu,T)$ has peaks as a function of $\mu$ centered at each single particle
550: energy level (Riemann's zero) of width $\ll \delta$. At these temperatures the
551: entropy is therefore zero for $\mu \neq E_j$, and takes the value $\log 2$
552: when $\mu = E_j$. As the temperature increases, the width increases (while the
553: height remains constant) and the peaks start to overlap. At temperatures $T
554: \sim T_\delta$ and at a fixed $\mu$, only the energy levels distant by a few
555: mean spacings from the chemical potential contribute to the entropy. The
556: fluctuations are governed by the local statistics of the eigenvalues, which
557: are universal (e.g., GUE). In this regime universality in the entropy
558: fluctuations is expected. In contrast, at higher temperatures $T \geqa \Tc$,
559: when peaks separated from $\mu$ by a distance of order $\Ec$ start to
560: contribute to the entropy, the universality will be lost because information
561: on the scale of the shortest periodic orbit enters. We now substantiate this
562: analysis by an explicit quantitative calculation.
563:
564: Deriving with respect to temperature Eqs. (\ref{omegliso}) and (\ref{Oosc}),
565: the smooth and fluctuating part of the entropy are given, respectively, by
566: \begin{eqnarray}
567: \overline{S} ( \mu , T ) &=& \frac{\pi^2}{3} \ \overline{\rho} ( \mu )
568: \ T \ , \\ \label{Slis}
569: \widetilde{S} ( \mu , T ) &=& \sum_p \sum_{r=1}^{\infty} \frac{
570: \kappa' ( \pi \ T \ r \log p) }{ r \ p^{r/2} } \cos( \mu \ r \log p) \
571: . \label{Sosc}
572: \end{eqnarray}
573: The function $\kappa' (x) = d \kappa / d x $ decreases also exponentially for
574: $x \gg 1$, producing a cut-off for prime numbers satisfying $\log p \gg 1 / (
575: \pi T)$. When $x \rightarrow 0$, $\kappa' (x)$ vanishes linearly. The maximum
576: of the function is located at $x \approx 1.6061$. Prime numbers that satisfy
577: $r \log p \sim 1 / ( \pi T)$ give therefore the main contribution to the
578: oscillations. At low temperatures large primes (long orbits) are selected.
579: Smaller primes (e.g., short orbits) contribute as the temperature is raised.
580:
581: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
582:
583: \subsection{Variance}
584:
585: Analogously to the grand potential, the variance of the entropy fluctuations
586: may also be written in terms of the form factor,
587: \begin{equation}
588: \langle \widetilde{S}^2 \rangle = \frac{1}{2} \int_{0}^{\infty} d\tau
589: ~ \frac{ \kappa'^2 ( \pi \ T \ \tau ) }{ \tau^2 } ~ K (\tau) \ .
590: \label{Svar}
591: \end{equation}
592: We briefly review here the general results obtained in Ref.\cite{lm3}, and
593: check them with the Riemannium.
594:
595: The two different regimes of the form factor described in \S \ref{secvaro}
596: split the integral (\ref{Svar}) into three different parts,
597: \begin{equation}
598: \langle \widetilde{S}^2 \rangle = \frac{1}{2} \left\{ \sum_{r \log p < \tau_*
599: } \frac{ \kappa'^2 ( \pi T \ r \log p ) }{ r^2 ~ p^r } + \int_{\tau_*}^{\tauH}
600: d\tau ~ \frac{ \kappa'^2 ( \pi T \tau ) }{ \tau } + \tauH \int_{ \tauH }^{
601: \infty } d\tau ~ \frac{ \kappa'^2 ( \pi T \tau ) }{ \tau^2 } \right\} \ .
602: \label{S2}
603: \end{equation}
604: The sum corresponds to the non universal short--time regime, whereas
605: the two integrals correspond to the long--time random matrix behavior.
606: These different terms dominate the integral at different temperatures.
607:
608: \vspace{0.2cm}
609: \noindent
610: * \underline{Low temperatures}: $T \ll T_\delta$. In this regime the maximum
611: of $\kappa'$ is centered at times much larger than the Heisenberg time
612: $\tauH$. The dominant term in Eq.(\ref{S2}) is the last one. We can extend the
613: integral down to zero with a negligible error. The variance of the entropy is
614: \begin{equation} \label{s2l}
615: \langle \widetilde{S}^2 \rangle \approx I_2 \ \overline{\rho} (\mu) \ T =
616: I_2 \ T/\delta \ ,
617: \end{equation}
618: where
619: \begin{equation}
620: I_2 = \pi^2 \int_{0}^{\infty} dx ~ \frac{\kappa'^2 ( x ) }{ x^2 }
621: \approx 1.51836 \ .
622: \end{equation}
623: In this initial regime the growth is linear, with a slope proportional to the
624: density of states. No additional specific information on the Riemann zeta is
625: present in this formula. Eq.(\ref{s2l}) reflects the discreteness of the
626: single--particle spectrum, and is thus a very general result valid for
627: arbitrary systems, independently of their dynamics \cite{lm3}. The smooth part
628: of the entropy grows also linearly with temperature, and is proportional to
629: the density of states \cite{landau}. Compared to the mean value
630: $\overline{S}$, the typical size of the fluctuations $\sqrt{\langle
631: \widetilde{S}^2 \rangle}$ are large (of order $(T/\delta)^{-1/2})$. In this
632: regime the quantum fluctuations dominate.
633:
634: \vspace{0.2cm}
635: \noindent
636: * \underline{Intermediate temperatures}: $T_\delta \ll T \ll \Tc$. Now the
637: maximum of $\kappa'$ is centered at times in between $\tmin$ and $\tauH$.
638: The dominant term in expression (\ref{S2}) is the second one, where we have
639: introduced the GUE diagonal approximation of the form factor $K_D (\tau) =
640: \tau$. Neglecting the contributions of short and long times we extrapolate the
641: integral from 0 to $\infty$,
642: \begin{equation} \label{sat}
643: \langle \widetilde{S}^2 \rangle \approx I_1 = \frac{1}{2}
644: \int_{0}^{\infty} dx ~ \frac{ \kappa'^2 ( x ) }{ x } \approx
645: 0.0709159 \ .
646: \end{equation}
647: In this regime, the size of the entropy fluctuations are therefore insensitive
648: to temperature variations, they saturate to a universal constant.
649:
650: \vspace{0.2cm}
651: \noindent
652: * \underline{High temperatures}: $T \sim \Tc $ and higher. In this regime the
653: diagonal approximation of the form factor is still accurate, but the short
654: time (non universal) structure is now apparent. At this temperatures the
655: entropy fluctuations are also sensitive to the fact that below $\tmin = \log
656: 2$ the form factor is strictly zero. The dominant term in Eq.(\ref{S2}) is the
657: first sum. The sum can be extended to all prime numbers and repetitions with
658: negligible error, and gives a variance
659: \begin{equation}
660: \langle \widetilde{S}^2 \rangle \approx \frac{1}{2} \sum_p \sum_{r =
661: 1}^{\infty} \frac{\kappa'^2 ( \pi \ T \ r \log p)}{ r^2 \ p^{r}
662: } \ .
663: \end{equation}
664: For $T \gg \Tc$, all the terms of this sum are exponentially small,
665: and the fluctuations vanish accordingly.
666:
667: To have a global description of $\langle \widetilde{S}^2 \rangle$ that
668: interpolates between the different regimes described above a numerical
669: evaluation of Eq.(\ref{S2}) is necessary. The result depends on $\mu$ through
670: the Heisenberg time.
671:
672: We have checked these predictions by a direct comparison of Eq.(\ref{S2}) to a
673: numerical computation of the entropy variance as a function of temperature.
674: The result is displayed in Fig.3 (for reference, $T_\delta \approx 0.013$).
675: There is an excellent agreement with theory for all temperatures. The initial
676: linear growth and the saturation to a plateau are amplified in the inset. The
677: size of the fluctuations almost reaches the theoretical prediction (\ref{sat})
678: at $T \approx 2.5 \ T_{\delta} \approx 0.03$. The expected intermediate
679: plateau is however short, because the temperatures $T_\delta$ and $\Tc$ are
680: not sufficiently well separated (due to the slow logarithmic decrease of
681: $\delta$, even at these large values of $\mu$ we are not sufficiently
682: asymptotic). The exponential decay is also well described.
683:
684: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
685:
686: \subsection{Distribution}
687:
688: In the regime $T \ll \Tc$ the previous results confirm that the behavior of
689: the statistical properties of the entropy fluctuations are universal. They
690: depend only on the structure of the GUE form factor and not on any specific
691: property of Riemann's zeros. The universality is not expected to be valid only
692: for the variance, but more generally for the full probability distribution. To
693: check this, we compare the probability distribution of the entropy
694: fluctuations obtained from Eq.(\ref{Sint}) using two different
695: single--particle spectra $\{ E_j \}$:
696: \begin{itemize}
697: \item[a)] the zeros of the Riemann zeta function,
698: \item[b)] the eigenvalues of a GUE ensemble of random matrices.
699: \end{itemize}
700: In the first case the probability distribution is computed by varying the
701: chemical potential in a small window, whereas in the second by averaging over
702: the Gaussian ensemble. Both probability distributions, computed at different
703: temperatures, are plotted in Fig.4. At low temperatures both distributions are
704: almost indistinguishable. Notice the strong sensitivity of the distribution to
705: temperature variations. When temperatures of order $\Tc$ are reached (cf part
706: d)), the universality is lost. For $T \sim \Tc$ the fluctuations are dominated
707: by short orbits, and are system specific. The moments of the distribution can
708: be computed by the same techniques used in \S \ref{sec:moments} for the grand
709: potential replacing ${\cal A}_{p,r}$ in Eq.(\ref{apr}) by $ {\cal A}_{p,r} (T)
710: =\kappa' ( \pi \ T \ r \log p) / (2 r \ p^{r/2} )$.
711:
712: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
713:
714: \section{Concluding remarks}
715: \label{sec:Conclu}
716: \setcounter{equation}{0}
717:
718: The use of the Riemannium as a test model in quantum mechanics is justified by
719: two main reasons. First, by its genericity: the Riemann spectrum possesses all
720: the generic features of a classically chaotic quantum system with no time
721: reversal symmetry. Second, by its practical advantages, namely all the
722: necessary quantum and semiclassical information required to work out
723: accurate computations and comparisons is available. Hence, though this "number
724: theoretic" model may seem somewhat remote from a realistic system, it provides
725: an excellent arena to verify the non trivial quantum mechanical properties of
726: chaotic systems.
727:
728: Based on semiclassical techniques and random matrix theory, several aspects of
729: the thermodynamics of a chaotic Fermi gas have been verified. An accurate
730: description of the probability distribution of the quantum fluctuations of the
731: grand potential (or energy) and of the entropy of the Riemannium were
732: obtained. In particular, the universal linear growth of the entropy variance
733: followed by a saturation, with a further non universal exponential decay, were
734: confirmed. The size of the saturation plateau, predicted in the regime
735: $T_\delta \ll T \ll \Tc$, was relatively small (see the inset in Fig.3). This
736: is due to the slow logarithmic asymptotic convergence properties of the
737: Riemannium (in spite of the large chemical potential used in the numerical
738: simulations, we are not very deep in the semiclassical limit. In fact, for the
739: window analyzed in the figures the number of particles is around $10^{12}$,
740: with $g \approx 35$. For an atomic nucleus or for electrons in a metallic
741: grain, this value of $g$ corresponds to approximately $40\sim 50$ particles.
742: In the Riemannium, in order to have $T_\delta$ and $\Tc$ separated by a factor
743: of, say, 100, $\mu$ need to be of the order of $10^{30}$).
744:
745: The high accuracy of the results obtained for all the quantities studied
746: confirm the validity of the different approximations employed. The present
747: theory therefore provides a solid ground to go beyond and test realistic
748: systems. Of particular interest is the interplay between mean--field
749: approximations, residual interactions and dynamics. Some encouraging results
750: in this direction were already obtained in the study of nuclear masses
751: \cite{mass}.
752:
753: This work has been supported by the European Commission under the Research
754: Training Network MAQC (HPRN-CT-2000-00103) of the IHP Programme.
755:
756: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
757:
758: \begin{thebibliography}{99}
759: \bibitem{bm} A. Bohr and B. R. Mottelson, {\it Nuclear Structure},
760: Benjamin, Reading, MS, 1969, Vol.I.
761:
762: \bibitem{sm} V. M. Strutinsky and A. G. Magner, Sov. J. Part. Nucl.
763: {\bf 7} (1976) 138.
764:
765: \bibitem{shapiro} B. Shapiro, Waves Random Media {\bf 9} (1999) 271.
766:
767: \bibitem{voppen} F. von Oppen, Phys. Rev. B {\bf 50} (1994) 17151.
768:
769: \bibitem{ruj} K. Richter, D. Ullmo and R. Jalabert, Phys. Rep. {\bf
770: 276} (1996) 1.
771:
772: \bibitem{agi} B. L. Altshuler, Y. Gefen, Y. Imry, Phys. Rev.
773: Lett. {\bf 66} (1991) 88.
774:
775: \bibitem{sch} A. Schmid, Phys. Rev. Lett. {\bf 66} (1991) 80.
776:
777: \bibitem{lm3} P. Leboeuf and A. G. Monastra, {\sl Ann. Phys.} {\bf
778: 297} (2002) 127.
779:
780: \bibitem{bg} O. Bohigas and M.-J. Giannoni, in {\sl Lecture Notes in
781: Physics} {\bf 209}, (Springer, Berlin, 1984) p.1.
782:
783: \bibitem{blm1} P. Leboeuf, A. G. Monastra and O. Bohigas, {\sl Reg.
784: Chaot. Dyn.} {\bf 6} (2001) 205.
785:
786: \bibitem{bb} R. Balian and C. Bloch, Ann. Phys. (N.Y.) {\bf 69} (1972)
787: 76.
788:
789: \bibitem{gutz} M. C. Gutzwiller, J. Math. Phys. {\bf 12}, 343 (1971);
790: {\sl Chaos in Classical and Quantum Mechanics} (Springer--Verlag, New
791: York, 1990).
792:
793: \bibitem{brack} M. Brack, Rev. Mod. Phys. {\bf 65} (1993) 677.
794:
795: \bibitem{odlyzko} A. M. Odlyzko, {``The $10^{20}$-th zero of the
796: Riemann zeta function and $175$ million of its neighbors''}, AT \& T
797: Report, 1992 (unpublished); see also http://www.dtc.umn.edu/\verb+~odlyzko+/.
798:
799: \bibitem{edwards} H. M. Edwards, {\sl Riemann's Zeta Function}, New York,
800: Academic Press, 1974.
801:
802: \bibitem{berry2} M. V. Berry, in {\sl Quantum Chaos and Statistical
803: Nuclear Physics} edited by T. H. Seligman and H. Nishioka, {\it
804: Lectures Notes in Physics} {\bf 263}, (Springer Verlag, Berlin, 1986)
805: p.1.
806:
807: \bibitem{foot} Three unusual features of the Riemann dynamics are the
808: following: a) the independence of the periods with energy. This can be
809: circumvented by interpreting the logarithm of the prime numbers as lengths,
810: instead of periods; b) contrary to semiclassical approximations, the
811: oscillating part is here exact (no correction terms are present) and, c) the
812: negative sign in front of Eq.(\ref{rhoosc}).
813:
814: \bibitem{bgs} O. Bohigas, M.-J. Giannoni, and C. Schmit, Phys. Rev.
815: Lett. {\bf 52} (1984) 1.
816:
817: \bibitem{mont} for the zeros of the Riemann zeta function, this conjecture is
818: originally due to H. L. Montgomery, {\sl Proc. Symp. Pure Math.} {\bf 24}, 181
819: (1973); D. Goldston and H. L. Montgomery, {\sl Proceeding Conf. at Oklahoma
820: State Univ. 1984}, edited by A. C. Adolphson et al, p.183.
821:
822: \bibitem{selberg} A. Selberg, {\sl Collected Papers}, Vol. {\bf I},
823: (Springer Verlag, Berlin, 1989) p.214.
824:
825: \bibitem{berry1} M. V. Berry, Proc. Roy. Soc. Lond. A {\bf 400} (1985)
826: 229.
827:
828: \bibitem{agam} O. Agam, J. Phys. I France {\bf 4} (1994) 697.
829:
830: \bibitem{foot2} A further development of the technique allows to compute the
831: tail of the distribution function, P. Bleher, O. Bohigas, P. Leboeuf and A.
832: Monastra, to be published.
833:
834: \bibitem{landau} L. Landau and E. M. Lifchitz, {\it Physique
835: Statistique}, \'Editions Mir, Moscou, 1988.
836:
837: \bibitem{mass} O. Bohigas and P. Leboeuf, Phys. Rev. Lett. {\bf 88}
838: (2002) 092502; Ibid 129903.
839:
840: %\end{references}
841: \end{thebibliography}
842:
843: \pagebreak
844:
845: %--------------
846:
847: \begin{figure}
848: \begin{center}
849: \leavevmode
850: \epsfysize=3.0in
851: \epsfbox{RII_varO.epsf}
852: \vspace{0.5cm}
853:
854: \caption{{\small Variance of the grand potential fluctuations of the
855: Riemannium as a function of temperature. Dots: numerical results computed in
856: the chemical potential window $\mu = (267653402147 \pm 6000)$ containing the
857: zeros $(10^{12}+1940)$ to $(10^{12}+48684)$. Full line: theoretical prediction
858: (\ref{varodiag}).}}
859:
860: \end{center}
861: \label{VarOmega}
862: \end{figure}
863:
864: %-------------
865:
866: \begin{figure}
867: \begin{center}
868: \leavevmode
869: \epsfysize=5.0in
870: %\epsfbox{/home/leboeuf/figures/RII_distO.epsf}
871: \epsfbox{RII_distO.epsf}
872: \vspace{0.5cm}
873:
874: \caption{{\small Normalized numerical histograms of the probability
875: distribution of $\widetilde{\Omega}$ at different temperatures: (a) $T = 0$,
876: (b) $T = 0.1$, (c) $T = 0.3$, (d) $T = 0.5$, (e) $T = 0.8$, (f) $T = 1.2$
877: (same chemical potential window as in Fig.1).}}
878:
879: \end{center}
880: \label{DistOmega}
881: \end{figure}
882:
883: %-------------
884:
885: \begin{figure}
886: \begin{center}
887: \leavevmode
888: \epsfysize=3.0in
889: %\epsfbox{/home/leboeuf/figures/RII_varS.epsf}
890: \epsfbox{RII_varS.epsf}
891: \vspace{0.5cm}
892:
893: \caption{{\small Variance of the entropy fluctuations of the Riemannium as a
894: function of temperature. Dots: numerical results computed in the same chemical
895: potential window as Fig.1. Full line: theoretical prediction (\ref{S2}).
896: Dashed line (inset): low temperature approximation (\ref{s2l}). Dot-dashed
897: line (inset): saturation value (\ref{sat}).}}
898:
899: \end{center}
900: \label{VarS}
901: \end{figure}
902:
903: %-------------
904:
905: \begin{figure}
906: \begin{center}
907: \leavevmode
908: \epsfysize=5.0in
909: %\epsfbox{/home/leboeuf/figures/RII_distS.epsf}
910: \epsfbox{RII_distS.epsf}
911: \vspace{0.5cm}
912:
913: \caption{{\small Normalized probability distribution of $\widetilde{S}$ at
914: different temperatures. Full line: numerical results for the Riemannium
915: computed in the same chemical potential window as Fig.1. Dash line: numerical
916: results from a GUE single--particle spectrum. (a) $T = T_{\delta} \approx
917: 0.013$, (b) $T = 4 \ T_{\delta} \approx 0.052$, (c) $T = 0.1$, (d) $T = 0.5$,
918: (e) $T = 0.8$, (f) $T = 1.2$.}}
919:
920: \end{center}
921: \label{DistS}
922: \end{figure}
923:
924: \
925:
926: %-------------------
927:
928: \pagebreak
929:
930: \begin{center}
931:
932: \begin{tabular}{|r||r|r|r|}
933: \hline
934: ~ & T = 0 ~~~~~~~~~~~ & T = 0.3 ~~~~~~~~~~~ & T = 0.5 ~~~~~~~~~~~ \\
935: \hline
936: $\langle \widetilde{\Omega}^2 \rangle$ & $(7.928 \pm 0.002) \times
937: 10^{-2}$ & $(5.9885 \pm 0.0025) \times 10^{-2}$ & $(4.2953 \pm 0.0015)
938: \times 10^{-2}$ \\
939: ~ & $7.9290 \times 10^{-2}$ & $5.9886 \times 10^{-2}$ & $4.2953 \times
940: 10^{-2}$ \\
941: \hline
942: $\langle \widetilde{\Omega}^3 \rangle$ & $-(5.78 \pm 0.02) \times
943: 10^{-3}$ & $-(3.44 \pm 0.02) \times 10^{-3}$ & $-(1.65 \pm 0.01)
944: \times 10^{-3}$ \\
945: ~ & $-5.7822 \times 10^{-3}$ & $-3.4377 \times 10^{-3}$ & $-1.6508
946: \times 10^{-3}$ \\
947: \hline
948: $\langle \widetilde{\Omega}^4 \rangle$ & $(1.480 \pm 0.002) \times
949: 10^{-2}$ & $(7.625 \pm 0.005) \times 10^{-3}$ & $(3.586 \pm 0.003)
950: \times 10^{-3}$ \\
951: ~ & $1.4814 \times 10^{-2}$ & $7.6273 \times 10^{-3}$ & $3.5869 \times
952: 10^{-3}$ \\
953: \hline
954:
955: \end{tabular}
956:
957: \end{center}
958:
959: {\small Table 1: Moments of $\widetilde{\Omega}$ at different temperatures.
960: The upper values (with the errors) were obtained from the numerical
961: distributions of figure (2.a), (2.c) and (2.d). The lower values are the
962: semiclassical results (\ref{varodiag}), (\ref{m3}) and (\ref{m4}),
963: respectively.}\\
964:
965: %-------------------
966:
967: \end{document}
968:
969:
970: