1: \documentclass[prl,aps,amsfonts,amssymb,floats,
2: twocolumn,showpacs]{revtex4}
3: \usepackage{epsf}
4: \usepackage{graphicx}
5:
6: %\newcommand{\be}{\begin{equation}}
7: %\newcommand{\ee}{\end{equation}}
8:
9: \begin{document}
10:
11: \title{Quantum correlations from Brownian diffusion of chaotic level-spacings}
12:
13: \author{S.N. Evangelou\footnote{e-mail: sevagel@cc.uoi.gr}
14: and D.E. Katsanos}
15: %\address
16: \affiliation{Department of Physics, University of Ioannina,
17: Ioannina 45110, Greece}
18:
19:
20: \begin{abstract}
21: Quantum chaos is linked to Brownian diffusion
22: of the underlying quantum energy level-spacing sequences.
23: The level-spacings viewed as functions of their order
24: execute random walks which imply uncorrelated
25: random increments of the level-spacings
26: while the integrability to chaos transition becomes
27: a change from Poisson to Gauss statistics for the
28: level-spacing increments.
29: This universal nature of quantum chaotic spectral
30: correlations is numerically demonstrated for eigenvalues
31: from random tight binding lattices and for zeros of
32: the Riemann zeta function.
33: \end{abstract}
34:
35:
36: \pacs{ 05.45.Mt, 05.40.-a, 03.67.-a, 05.45.Pq}
37:
38: \maketitle
39:
40: \par
41: \medskip
42: Over the last couple of decades a general consensus has emerged
43: concerning remnants of classical chaos in quantum physics. Since
44: the smooth and reversible quantum evolution allows no chaos what
45: is nowadays called quantum chaos had to be traced elsewhere,
46: usually in the stationary eigenvalues of a quantum system
47: \cite{1}. A key quantity which was introduced to distinguish
48: between quantum chaos and quantum integrability is the
49: distribution function of level-spacings between nearest
50: eigenvalues \cite{1,2}. The quantum chaotic behavior is then
51: defined as the resemblance of the level-spacing statistics to
52: those obtained from Gaussian random matrices having highly
53: correlated levels, which are described by random matrix theory
54: (RMT) \cite{2,3,4,5}. The major characteristic of quantum chaos is
55: the so-called ``level-repulsion'' between energy levels which is
56: due to the dramatic lowering of symmetries in chaotic systems,
57: while the corresponding level-spacing distribution obeys the
58: Wigner surmise. On the contrary, quantum systems with integrable
59: classical analogues are described by Poisson level-spacing
60: distribution function which implies uncorrelated random
61: level-spacings with possible ``degeneracies'' related to the
62: presence of symmetries. Quantum chaos has been experimentally
63: observed for the irregular energy spectra in atom-optics billiards
64: \cite{6,7}.
65:
66: \par
67: \medskip
68: The present era of quantum computation, where the exploitation of
69: quantum information resources came to the forefront, requires a
70: better understanding of spectral noise arising from quantum
71: chaotic systems. This noise could be related to quantum
72: correlations, such as entanglement which is measured via von
73: Neumann entropy and usually violates Bell inequalities \cite{8,9}.
74: For such a purpose the energy level-spacing statistical
75: distribution function is insufficient to describe fully a quantum
76: system. In order to reveal more about the nature of quantum
77: spectral noise one must examine level-to-level correlations. This
78: was done in \cite{10} where they conjecture that quantum chaotic
79: energy spectra are characterized by $1/f$ noise, which is also
80: verified by numerical data \cite{10}. In this paper we do not look
81: only correlations between the enrgy eigenvalues but also view the
82: nearest level-spacings as functions of their order, so that they
83: can imitate discrete time series with their order corresponding to
84: time. We provide evidence that in quantum chaos the classically
85: chaotic temporal evolution is replaced by diffusion of the
86: level-spacing dynamics. In other words, we can interpret the
87: evolution of level-spacings (not the energies themselves) as that
88: of fictitious particles executing random walks in one-dimensional
89: space. Moreover, we arrive at a more precise definition of quantum
90: chaos which includes quantum correlations in addition to
91: statistical behavior.
92:
93:
94: \par
95: \medskip
96: The energy level series of a quantum system make up an irregular
97: continuous but non-differentiable self-affine fractal curve
98: described by a fractal dimension $D$. For quantum chaotic levels
99: the noisy curve is space filling with $D=2$ while for energy
100: levels from quantum integrable systems the absence of correlations
101: implies less irregular curves with fractal dimension $D=1.5$. The
102: corresponding power-spectra $P(f)\propto f^{-\alpha}$ are
103: characterized by a power-law exponent $\alpha$ which obeys the
104: dimensional relation $D=(5-\alpha)/2$ \cite{11}. Therefore,
105: quantum chaos implies $1/f$ spectral noise with $\alpha=1$ while
106: quantum integrability gives $1/f^{2}$ noise with $\alpha=2$,
107: respectively \cite{10}. In order to provide a better understanding
108: for the nature of correlations in quantum chaotic energy spectra
109: we ask the question: ``What is the nature of correlations when
110: dealing with the level-spacings, instead?". Our main finding is
111: that the correlations in the level-spacing sequences of quantum
112: chaotic spectra are Brownian. Nevertheless, these are quantum
113: correlations which occur in addition to the unpredictability
114: associated to the classical chaotic behavior. Therefore, a link
115: can be established between quantum chaos and quantum correlations
116: such as entanglement \cite{8}. These facts are illustrated for two
117: basic examples of quantum chaotic behavior, the weakly disordered
118: lattice and the Riemann zeros. Furthermore, quantum chaos is
119: related to a Gaussian distribution for the $\it
120: {level}$-$\it{spacing}$ $\it{increments}$ while quantum
121: integrability is tight up to a Poisson distribution, respectively.
122: This allows a simple description of the integrability to chaos
123: transition as a change of the level-spacing increments from the
124: discrete rare event Poisson distribution to the continuous Gauss
125: distribution.
126:
127:
128: \par
129: \medskip
130: \begin{figure}
131: \includegraphics[width=8cm]{fig_1.eps}
132: \caption{(Color online) Energy-level sequences $ {\varepsilon}_{n}$ with
133: subtracted mean spacings for chaotic, integrable and
134: quasi-integrable systems obtained from disordered and
135: quasi-periodic lattices. The decreasing irregularity of the curves
136: can be seen with the space filling chaotic curve approaching
137: $D=2$, the integrable curve approaching $D=1.5$ and $D$ close to
138: $1$ for the chosen quasi-integrable model. The dimensions shown in
139: the figure differ from the expected values since they were
140: obtained by box-counting algorithms which are usually not very
141: accurate for such self-affine curves. Quantum correlations exist
142: in (a) for the maximally noisy quantum chaotic curves. The quantum
143: correlations die out in (b),(c).}
144: \end{figure}
145:
146:
147:
148: \par
149: \medskip
150: In Fig. 1 energy level signals from electrons in lattices as a
151: function of the level-index $n$ are displayed. The $n$th level is
152: expressed as a sum of level-spacings, beginning from the left end
153: of the spectrum up to the $n$th level, via
154: \begin{eqnarray}
155: {\varepsilon}_{n}=\sum_{i=1,n} \left(S_{i}-\langle S \rangle
156: \right).\nonumber
157: \end{eqnarray}
158: From each level-spacing $S_{i}$ we have subtracted a local mean
159: level-spacing $\langle S\rangle$ by following the usual unfolding
160: procedure for the density of states to become constant on average
161: everywhere in the spectrum. The chaotic, quasi-integrable and
162: integrable energy-level series can be seen to display very
163: different curve roughness. Very rough curves with $D$ approaching
164: $2$ are obtained for chaotic levels, the roughness is smaller with
165: $D$ close to $1.5$ for integrable ones while it becomes much
166: lower, closer to $D=1$, for the studied quasi-integrable system
167: where the corresponding curve approaches a line. In Fig. 2 the
168: corresponding Fourier analysis of these energy-level signals is
169: shown via the power-spectrum
170: \begin{eqnarray}
171: P(f) & = & {\frac {1}{N}} \left|\sum_{n=1}^{N}
172: {\varepsilon}_{n} \exp\left (-i{\frac {2\pi f n}{N}}\right)
173: \right|^{2}, \nonumber
174: \end{eqnarray}
175: averaged over many configurations. We show the energy levels from
176: weakly disordered solids and Riemann zeroes which give $1/f$
177: noise. Our findings for the presence of $1/f$ power-spectra for
178: quantum chaotic levels agree with the conjecture of Ref.
179: \cite{10}. We emphasize that quantum correlations exist only for
180: the maximally noisy spectrum of quantum chaotic levels of Fig.
181: 1(a). The integrable system of Fig. 1(c) does not show any
182: correlations described by a less rough curve with Brownian $1/f
183: ^{2}$noise. The quasi-integrable system of Fig. 1(b) does not
184: display any quantum correlations either. Below we shall
185: demonstrate that Brownian noise describes quantum correlations for
186: chaotic systems not for the energy levels but for the
187: level-spacing sequences, instead.
188:
189: \begin{figure}
190: \includegraphics[width=8.0cm] {fig_2.eps}
191: \caption {(Color online) The averaged power-spectra of the energy levels for
192: weakly disordered three-dimensional lattice systems, for a
193: quasi-integrable critical model and for the spectrum of zeros of
194: the Riemann zeta function. }
195: \end{figure}
196:
197:
198: \par
199: \medskip
200: The $1/f$ noise found to describe the quantum chaotic levels is a
201: common law for many classical complex phenomena where very rare
202: intense effects require the cooperation of various favorable
203: cases. It can be also viewed as a scaling law since a change of
204: the intensity combined with contraction in time leaves its form
205: unchanged. The $1/f$ law is very old and appears in many cases
206: ranging from fluid flows, stock market price indices, earthquakes,
207: etc. \cite{12}. However, despite of its generality its origin
208: remains a mystery without any widely accepted general mechanism
209: for its interpretation. The most common explanation of $1/f$ noise
210: relies on the extension of Lorentzian correlation function
211: $\gamma/(\omega^{2}+\gamma^{2})$ from usual random telegraph noise
212: of a single fluctuator with exponential relaxation $\gamma$, to
213: several fluctuators with a broad distribution of relaxation rates.
214: Although hard to justify physically if $log(\gamma)$ is uniformly
215: distributed the average over Lorentzians with the chosen broad
216: distribution of $\gamma$'s gives $1/f$ noise \cite{13}. Another
217: mechanism of $1/f$ noise seems more appropriate for quantum
218: chaotic levels. It was proposed (see for example Ref. \cite{14})
219: that a possible origin of $1/f$ noise can be Brownian correlated
220: ${\it interevents}$ between narrow pulses. Therefore, in order to
221: obtain $1/f$ noise for the chaotic levels $\varepsilon_{n}$ the
222: corresponding level-spacings $S_{n}$ which appear in Eq. (1)
223: should follow a Brownian random walk and the $\it {level}$-$\it
224: {spacing}$ $\it {increments}$ (spacings of spacings)
225: $S_{n}-S_{n-1}$ to be random uncorrelated having white type of
226: noise. This can be understood rather easily since Brownian
227: $1/f^{2}$ noise of the levels arises from uncorrelated $f^{0}$
228: white noise of the level-spacings, similarly $1/f$ noise observed
229: for chaotic levels can arise from uncorrelated white noise
230: level-spacing increments.
231:
232:
233:
234: \par
235: \medskip
236: \begin{figure}
237: \includegraphics[width=8cm]{fig_3.eps}
238: \caption { (Color online) Level-spacing sequences $S_{n}$ with unit mean-level
239: spacing $\langle S\rangle=1$ for chaotic, quasi-integrable and
240: integrable systems. On the left the corresponding normalized
241: level-spacing distribution function $P(S)$ can be seen.}
242: \end{figure}
243:
244: \par
245: \medskip
246: Our main results for the level-spacing series
247: $S_{n}=\varepsilon_{n}-\varepsilon_{n-1}$ are displayed in Fig. 3
248: together with the corresponding level-spacing distributions $P(S)$
249: on the left of the figure. It is well-known that from the chaotic
250: series the obtained $P(S)$ obeys the Wigner surmise, the less
251: irregular integrable series is described by a Poisson
252: distribution, while the quasi-integrable ones follow an
253: intermediate semi-Poisson distribution \cite{15}. The difference
254: between the amount of noise displayed in the various cases is
255: immediately obvious. In order to show that quantum correlations
256: are included in the chaotic level-spacings which follow Brownian
257: motion we have looked at the statistical distribution of the
258: level-spacing differences $S_{n}-S_{n-1}$. For chaotic levels it
259: should be Gaussian and for integrable levels of Poisson type which
260: correspond to correlated and uncorrelated levels, respectively.
261: The two behaviors are clearly distinguished in Fig. 4. We have
262: verified this picture by computing all the levels, the
263: level-spacings and the increments of level-spacings in lattice
264: models and the Riemann zeros. In Fig. 4 we show the statistics of
265: the level-spacing increments $S_{n}-S_{n-1}$. For weakly
266: disordered lattices where $\beta=1$ and $30000$ Riemann zeroes
267: \cite{16} where $\beta=2$ the data can be described by simple
268: Gauss distributions for both cases. In the integrable case the
269: statistics is Poisson while a similar result of different Poisson
270: type is obtained for the quasi-integrable statistics in the inset.
271:
272:
273: \par
274: \medskip
275: %\section{Quantum chaos transition}
276:
277:
278: \par
279: \medskip
280: %\subsection{ Disordered tight-binding lattice}
281:
282: \begin{figure}
283: \includegraphics[width=8.0cm] {fig_4.eps}
284: \caption {(Color online) The distribution of the level-spacing increments.
285: The transition from integrability to quantum chaos becomes a
286: transition from Poisson to Gauss. For weak disorder the obtained
287: variance of the Gaussian is $\sigma^{2}\approx 0.71$ and for the
288: Riemann zeros is $\sigma^{2}\approx 0.48$. The quasi-integrable
289: system shown in the inset is described by the simple Poisson
290: $e^{-2x}$.}
291: \end{figure}
292:
293:
294: \par
295: \medskip
296: We summarize our numerical computations of eigenvalues responsible
297: for the level-spacings. First, we consider the weakly disordered
298: Anderson model studied on $L^{3}$ sites of a 3D lattice. The
299: diagonal matrix elements of the corresponding are random taken
300: from a uniform probability distribution of zero mean and variance
301: $W^{2}/12$ and the off-diagonal matrix elements connecting nearest
302: neighbors are unity. The short ranged and sparse real symmetric
303: matrices are diagonalized to obtain long enough signals of raw
304: eigenvalues $E_{n}$. For such purpose the Lanczos algorithm was
305: employed and the scaling behavior is achieved by considering
306: linear lattice size up to $L=25$ for many random configurations.
307: In the unfolding procedure one replaces each level-spacing
308: $E_{n+1}-E_{n}$ by
309: $\varepsilon_{n+1}-\varepsilon_{n}={\cal{N}}(E_{n+1})-{\cal{N}}(E_{n})$,
310: where $\varepsilon_{n}={\cal{N}}(E_{n})$ are the unfolded levels,
311: with ${\cal{N}}(E)$ the averaged integrated density of states. In
312: this procedure the mean level-spacing becomes $\langle S \rangle
313: =1$ throughout the spectrum. We have taken linear size $L=25$ for
314: the chaotic levels disorder $W=10$ for $57$ random configurations
315: total of $147000$ eigenvalues, for the critical levels disorder
316: $W_{c}=16.5$ for $80$ random configurations total of $140000$
317: eigenvalues and the integrable case disorder $W=30$ total of
318: $10763$ random configurations total of $102800$ eigenvalues. The
319: averaged power spectra which are shown in Fig. 2 for an energy
320: window $[0,2]$ which include approximately $2000$ eigenvalues. We
321: have estimated the power spectrum laws $\alpha=2.0\pm 0.07$ for
322: the chaotic case, $\alpha=1.52\pm 0.04$ for the integrable case
323: while a law close to $\alpha=1.62\pm 0.08$ was found for the
324: critical case (not shown).
325:
326: \par
327: \medskip
328: We have also diagonalized a simpler critical model replacing the
329: critical case of 3D disordered systems to consider the behavior
330: at the borderline from integrability to chaos. This is the
331: one-dimensional quasi-integrable Harper model \cite{15} where the
332: site potential is $V_{n}=2\cos(2\pi\sigma n)$ with irrational
333: $\sigma =(\sqrt{5}-1)/2$. This model displays critical behavior
334: with multifractal spectra and wave functions. We find that the
335: averaged power spectra gave power-law estimates close to
336: $\alpha=3.3$ to $3.8$ as the system size increased.
337: %\par
338: %\medskip
339: %\subsection{Zeroes of the Riemann zeta function}
340: %\par
341: %\medskip
342: The level-spacings of adjacent Riemann complex zeroes \cite{16}
343: also exhibit fluctuations similar to the ones of the energy
344: spectra of electrons in disordered lattices. The Riemann
345: hypothesis states that all non-trivial zeros of the Riemann
346: $\zeta(s)=\sum_{s}n_{s}$ function have real part equal to $1/2$
347: with $\zeta_{n}=1/2+\varepsilon_{n}$. Although numerical
348: computations agree with Riemann hypothesis a full theoretical
349: proof is not known. However, the RMT is believed to be valid in
350: this case so that the spectrum of Riemann zeros represents a
351: hypothetical quantum chaotic system in the unitary universality
352: class described by GUE, which corresponds to broken time-reversal
353: invariance via a magnetic field. The $P(S)$ statistics for small
354: $S$ in this case is $\propto S^{\beta}$ with $\beta=2$, compared
355: to $\beta=1$ for the orthogonal case of disordered systems. In
356: RMT language we have a crossover from GOE to GUE statistics. In
357: Fig. 2 we demonstrate $1/f$ noise also for sequences of the
358: Riemann zeroes. We summarize that our findings for fractal
359: fluctuations and $1/f$ noise for the chaotic spectra are linked to
360: quantum fluctuations.
361:
362:
363: \par
364: \medskip
365: It is reasonable to expect that fingerprints of chaos present in
366: the classical time evolution can be also found in the stationary
367: energy levels of the corresponding quantum system. In addition to
368: classical chaos the chaotic spectra should also contain some form
369: of quantum correlations. Therefore, it is worth searching for the
370: two ingredients of quantum chaos: ``correlations" for the quantum
371: part and ``randomness" for the chaotic part. We find that both
372: appear in the self-affine fractal noise of the spectral curves
373: which is exemplified via the random walk behavior found for the
374: level-spacings. Although quantum evolution is very different from
375: that of classical point particles in phase space Brownian
376: diffusion also occurs in the quantum case. In Hilbert space
377: level-spacing evolution is important, which seems to play the same
378: role for quantum chaos as the time evolution in phase space for
379: classical chaos. Moreover, quantum chaos implies random
380: uncorrelated level-spacing increments which obey a Gaussian
381: distribution. In other words, quantum memory effects in the
382: evolution, such as dependence on the initial wave packet
383: superpositions, eventually die out for the level-spacing
384: increments where quantum chaos meets classical chaos. The quantum
385: chaotic level-spacing increments obey Gauss distribution
386: distinguished from integrable ones which obey Poisson
387: distribution.
388:
389: \par
390: \medskip
391: We have resolved numerically important issues concerning the
392: nature of spectral correlations and noise in quantum chaos. Common
393: signatures of quantum chaos such as ``level-repulsion" and
394: ``spectral rigidity" arise from quantum ``correlations" which
395: appear in chaotic level-spacings. If combined with ``randomness",
396: which might have a classical chaotic origin, make up what we call
397: quantum chaos. On the other hand, integrability usually concerns
398: uncorrelated random levels where the level-spacings obeys the
399: Poisson distribution. We should add that integrability occurs
400: either for pure ballistic systems where quantum correlations exist
401: without randomness or strongly disordered systems where due to
402: localization in random positions randomness exists without
403: correlations since localized wave functions do not communicate.
404: Quantum chaos requires both ingredients, ``quantum correlations"
405: and ``randomness", simultaneously, for example both of them are
406: present in RMT spectra.
407:
408: \par
409: \medskip
410: In summary, starting from the identification of proposed $1/f$
411: spectral fluctuations in quantum chaos numerical evidence was
412: given in favor of random walk behavior for the level-spacings. We
413: also argued that the definition of quantum chaos requires both
414: ``quantum correlations" and ``randomness". Our numerical study
415: suggests a connection between chaotic motion of classical
416: particles in phase space and motion of the level-spacings between
417: eigenvalues in Hilbert space. In the quantum case Brownian
418: correlations are shown to contribute to quantum effects. Although
419: the level-spacing distribution is widely accepted as the best
420: diagnostic tool of quantum chaos there is no direct theory
421: attached to it, as it exists for example for other level
422: correlation functions via RMT (see however the recent study of
423: \cite{17}). This might explain why the simple random walk behavior
424: of level-spacings in quantum chaotic systems, where quantum
425: correlations and unpredictable behavior are combined was not paid
426: much attention before.
427:
428: \begin{thebibliography}{99}
429:
430: \bibitem{1} M.V. Berry and M. Tabor, Proc. R. Soc. London A
431: {\bf 356}, 375 (1977); M.V. Berry, Ann. Phys. (NY) {\bf 131}, 163
432: (1981).
433:
434: \bibitem{2} O. Bohigas, M.J. Giannoni and C. Schmit,
435: Phys. Rev. Lett. {\bf 52}, 1 (1984).
436:
437: \bibitem{3} T. Guhr, A. M${\ddot u}$ller - Groeling ~ and ~
438: H.A.~ Weidenm${\ddot u}$ller, Phys. Rep. {\bf 299}, 189 (1998).
439:
440: \bibitem{4} H.J. St${\ddot{o}}$ckmann,
441: {\it ``Quantum Chaos"}, Cambridge~ University Press, Cambridge
442: (1999).
443:
444: \bibitem{5}F. Haake, {\it ``Quantum Signatures of Chaos"},
445: Springer-Verlag, Berlin (2001).
446:
447: \bibitem{6}V. Milner, J.L. Hanssen, W.C. Campbell, M.G. Raizen,
448: Phys. Rev. Lett. {\bf 86}, 1514 (2001).
449:
450: \bibitem{7}N. Friedman, A. Kaplan, D. Carasso and N. Davidson,
451: Phys. Rev. Lett. {\bf 86}, 1518 (2001).
452:
453: \bibitem{8} For the interplay between entanglement and quantum
454: chaos see J.N. Bandyopadhyay and A. Lakshminarayan, Phys. Rev.
455: Lett. {\bf 89}, 060402 (2002) and references therein.
456:
457: \bibitem{9} M.A. Nielsen and I.L. Chuang,
458: {\it ``Quantum Computation and Quantum Information"}, Cambridge
459: University Press, Cambridge (2000).
460:
461: \bibitem{10} A. Rela$\tilde{n}$o, J.M.G. G$\acute{o}$mez,
462: R.A. Molina, J. Retamosa and E. Faleiro, Phys. Rev. Lett. {\bf 89}, 244102
463: (2002).
464:
465: \bibitem{11} M.V. Berry, J. Phys. A: Math. and Gen., {\bf 29},
466: 6617 (1996).
467:
468: \bibitem{12} B.B. Mandelbrot, {\it ``Multifractals and 1/f noise"},
469: Springer, New York (1999)
470:
471: \bibitem{13} B.B. Mandelbrot and J.W. Van Ness, SIAM Rev {\bf 10},
472: 422 (1968).
473:
474: \bibitem{14} B. Kaulakys, cond-mat/0305067 and references therein.
475:
476: \bibitem{15} S.N. Evangelou and J.-L. Pichard,
477: Phys. Rev. Lett. {\bf 84}, 1643 (2000).
478:
479: \bibitem{16} Non-trivial Riemann zeroes are taken from tables in
480: http://www.dtc.umn.edu/~odlyzko/doc/zeta.html, see
481: A.M. Odlysko. Math. Comp. {\bf 48}, 273 (1987).
482:
483: \bibitem{17} O. Bohigas, P. Leboeuf and M.J. S${\acute{a}}$nchez,
484: CD/0012049.
485:
486:
487:
488: \end{thebibliography}
489:
490: \end{document}
491: