1: \documentclass[prl,preprint,aps,epsf]{revtex4}
2: \usepackage{graphics}
3: \usepackage{epsfig}
4: \begin{document}
5: % \draft command makes pacs numbers print
6: \draft
7: \title{Non-Reversible Evolution of Quantum Chaotic System. Kinetic Description.}
8: \author{L. Chotorlishvili, V. Skrinnikov }
9: \affiliation{Department of Physics,Tbilisi State University, Chavchavadze av. 3, 0128 Tbilisi, Georgia \\
10: E-mail: lchotor33@yahoo.com}
11: \date{\today}
12:
13: \begin{abstract}
14: It is well known that the appearance of non-reversibility in classical chaotic systems is connected with a local instability of phase trajectories relatively to a small change of initial conditions and parameters of the system. Classical chaotic systems reveal an exponential sensitivity to these changes. This leads to an exponential growth of the initial error with time, and as the result after the statistical averaging over this error, the dynamics of the system becomes non-reversible.
15:
16: In spite of this, the question about the origin of non-reversibility in quantum case remains actual. The point is that the classical notion of instability of phase trajectories loses its sense during quantum consideration.
17:
18: The current work is dedicated to the clarification of the origin of non-reversibility in quantum chaotic systems.
19:
20: For this purpose we study a non-stationary dynamics of the chaotic quantum system. By analogy with classical chaos, we consider an influence of a small unavoidable error of the parameter of the system on the non-reversibility of the dynamics.
21:
22: It is shown in the paper that due to the peculiarity of chaotic quantum systems, the statistical averaging over the small unavoidable error leads to the non-reversible transition from the pure state into the mixed one.
23:
24: The second part of the paper is dedicated to the kinematic description of the chaotic quantum-mechanical system. Using the formalism of superoperators, a muster kinematic equation for chaotic quantum system was obtained from Liouville equation under a strict mathematical consideration.
25: \end{abstract}
26:
27: \pacs{PACS number: 03.65.Sq, 05.45.Mt}
28:
29: \maketitle
30:
31: %\begin{center}
32: %INTRODUCTION
33: %\end{center}
34: \section*{INTRODUCTION}
35:
36: Study of the quantum reversibility and motional stability is of
37: great interest [1-4]. This interest is due to not only the
38: fundamental problem of irreversibility in quantum dynamics, but
39: also to particular application. In particular, it reveals itself in relarion to the
40: field of quantum computation. A quantity of central importance
41: which has been on the focus of many studies [5-9] is the so-called
42: fidelity $f(t)$, which measures the accuracy to which a quantum
43: state can be recovered by inverting, at time $t$, the dynamics
44: with a perturbed Hamiltonian
45: \begin{equation}
46: f(t)=|\langle \psi|e^{i\hat{H}t}e^{-i\hat{H}_ot}|\psi\rangle|^2.
47: \end{equation}
48:
49: Here $\psi$ is the initial state which evolves in time $t$ with
50: the Hamiltonian $\hat{H}_o$, while $\hat{H}=\hat{H}_o+\hat{V}$ is
51: the perturbed Hamiltonian. The analysis of this quantity has shown
52: that under some restrictions, the series taken from $f(t)$ is exponential
53: with a rate given by the classical Lyapunov exponent [5-6]. But
54: here a question appears. The point is that the origin of the dynamic
55: stochasticity, which is the reason of irreversibility in
56: classical case, is directly related to the nonlinearity of
57: equation of motion. For classical chaotic system this nonlinearity
58: leads to the repulsion of phase trajectories at a sufficiently
59: quick rate [10-13]. In case of quantum consideration, the
60: dynamics of a system is described by a wave function that obeys a
61: linear equation and the notion of a trajectory is not used at all.
62: Hence, at first sight it seems problematic to find out the quantum
63: properties of systems, whose classical consideration reveals their
64: dynamic stochasticity. For this reason, according to the widely
65: accepted opinion the notion of "Quantum Chaos" includes phenomena,
66: related to the quantum-mechanical description of systems chaotic
67: in the classical limit [14-17].
68:
69: Unfortunately, our understanding of quantum dynamics of chaotic
70: systems is still quite limited. The majority of the existing
71: quantum chaos literature concentrates on understanding the
72: properties of eigenfunctions and eigenvalues.
73:
74: In this paper we consider the issue of irreversibility for
75: quantized chaotic system. For simplicity we shall discuss a
76: periodic driving.
77:
78: In given work we shall try to show that in essentially
79: chaotic domain, quantum chaotic dynamics is characterized by the
80: transition from the pure quantum-mechanical state into the mixed
81: one. In this case, irreversibility in the system appears as a
82: result of loss of informaintion about the phase factor of the wave
83: function. To prove this phenomenon, we shall consider model
84: Hamiltonian studied in [1,18]. We shall try to give our
85: explanation of the mixed state formation in a chaotic
86: quantum-mechanical system. The main purpose of given work is the
87: deriving of kinetic equation for chaotic quantum-mechanical
88: system. Interesting method of how to derive a kinetic equation for
89: semi-classical chaotic system was offered in [16]. In this work we
90: shall try to do the same in exceptionally quantum domain, without
91: application to the semi-classical methods. The paper is organized
92: as follows. In the first part formation of a mixed state is
93: considered. The second part is dedicated to the kinetic
94: description of quantum chaos.
95:
96:
97: \section{Formation of mixed state.}
98:
99:
100: During the discussion of nonreversible evolution of the
101: quantum-mechanical system naturally the question emerges. How does a non-reversibility originate in a
102: quantum system? If the quantum system
103: evolves according to the Schrodinger equation, how can a pure
104: quantum-mechanical state become a mixed one? The question is that,
105: in contrast to the classical chaos, quantum-mechanically irregular
106: motion cannot be characterized by extreme sensitivity to tiny
107: changes of the initial data. Due to the unitarity of the quantum
108: dynamics, the overlap of two wave functions remains
109: time-independent
110: $|\langle\Theta(t)|\xi(t)\rangle|^2=|\langle\Theta(0)|\xi(0)\rangle|^2$,
111: provided time-dependence of $\Theta(t)$ and $\xi(t)$ is generated
112: by the same Hamiltonian. However, an alternative characterization
113: of classical chaos, extreme sensitivity to slight changes of the
114: dynamics does carry over into quantum mechanics.
115:
116: As it was mentioned in the introduction, the model Hamiltonian,
117: studied in [1,18] is taken as a model for our considerations
118:
119: \begin{equation}
120: \hat{H}(Q,P,x)=\frac{1}{2}(P^2_1+P^2_2+Q^2_1+Q^2_2)+xQ^2_1Q^2_2,
121: \label{2}
122: \end{equation}
123: where $P_1$, $P_2$, and $Q_1$, $Q_2$ are canonical momenta and
124: coordinates. Here and further we shall consider dimensionless
125: quantities.
126:
127: It follows from works [1,18] that for values of parameters
128: $x=x_o\approx 1$ and $E\approx 3$ (where $E$ is energy of system),
129: dynamics related to Hamiltonian (2) is chaotic and is
130: characterized by correlation function, the width of which is equal to
131: $\tau_c=1$ [1]. Our goal is to study mechanisms of irreversibility and mixed state formation in non-stationary quantum dynamics. For this let present Hamiltonian (2) in the following
132: form $$\hat{H}(t)=\hat{H}_o+V(t),$$
133: \begin{equation}
134: \hat{H}_o=\frac{1}{2}(P^2_1+P^2_2+Q^2_1+Q^2_2)+x_oQ^2_1Q^2_2,
135: \end{equation}
136: $$V(t)=V_o(Q)f_o(t),~V_o=\Delta x_o \cdot Q^2_1Q^2_2.$$
137:
138: As one can see from (3), the time dependence of the Hamiltonian
139: $\hat{H}(t)$ is provide by amplitude modulation of the parameter
140: $x=x_o+\Delta x_of_o(t)$, where $f_o(t)$ is a periodic function
141: of time with $T_o$ period. Hereinafter we shall work in a
142: domain of chaotic motion $x_o\sim 1,~ \Delta x_o<x_o$.
143:
144: After taking (3) into account, the Schrodinger equation for the
145: wave function takes the form
146: \begin{equation}
147: i\frac{\partial |\psi(Q,x(t))\rangle}{\partial
148: t}=\hat{H}(Q,x(t))|\psi(Q,x(t))\rangle. \label{4}
149: \end{equation}
150:
151: The solution of the time-dependent Schrodinger equation (4)
152: can be written formally with the help a time-dependent
153: exponential [14]
154: $$U(t)=exp[-i\int_{o}^{t}dt'\hat{H}(t')]_t,$$
155: where the positive time ordering requires $$[A(t)B(t')]_t=
156: \left\{
157: \begin{array}{l}
158: A(t)B(t')\hskip 0.7cm{\mbox if}\hskip 0.2cm t>t' \\
159: B(t')A(t)\hskip 0.7cm {\mbox if} \hskip 0.2cm t<t'
160: \end {array} .
161: \right. $$
162:
163: In our case $H(t+T_ok)=H(t)$, $k=1,2...$, the evolution operator
164: referring to one period $T_o$ , the so-called Flouqet operator
165: $U(T_o)\equiv \hat{F}$ [14], is worthy of consideration, since it
166: yields a stroboscope view of the dynamics
167: $$|\psi(kT_o)\rangle=(\hat{F})^n |\psi(0)\rangle.$$
168:
169: The Flouqet operator being unitary has unimodular eigenvalues.
170: Suppose we can find eigenvectors $|\varphi _{\chi}\rangle$ of the
171: Flouqet operator $$\hat{F}|\varphi _{\chi}\rangle =
172: e^{-i\varphi_{\chi}}|\varphi _{\chi}\rangle,$$
173: \begin{equation}
174: \langle \varphi _{\chi}|\varphi _{\Theta}\rangle =\delta _
175: {\chi\Theta}. \label{5}
176: \end{equation}
177: Then, with the eigenvalue problem solved, the stroboscopic
178: dynamics may be written explicitly [14]
179: \begin{equation}
180: |\psi_{n}(kT_o)\rangle = \sum_{\chi} e^{-ik \varphi_{\chi}}\langle
181: \varphi _{\chi}|\psi_{n}(0)\rangle|\varphi _{\chi}\rangle. \label{6}
182: \end{equation}
183:
184: Here $|\psi_{n}(0)\rangle$ is the eigenfunction of the Hamiltonian
185: $H$ (3), corresponding to the value of parameter $x(0)=x_0+\Delta
186: x_0f(0)$.
187:
188: As it was mentioned above, our aim is to prove that one of the signs of the emergence
189: of quantum chaos is a formation of the mixed state. Being
190: initially in a pure quantum-mechanical state, described by the
191: wave function $|\psi_n\rangle$, the system during the evolution
192: makes an irreversible transition to the mixed state.
193:
194: The information about whether the system is in the mixed state or in the pure one, may be obtained from the form of the density matrix [19-21]. Introducing the definition $C_{nm}(k)=e^{-ik\phi_m}\langle\phi_m | \psi_n(0)\rangle$, let us rewrite (6) in more convenient way for further usage:
195: \begin{equation}
196: | \psi_{n}(kT_0)\rangle = \sum_{m}C_{nm}(k)| \phi_m\rangle.
197: \end{equation}
198: Next, according to the commonly accepted rules [22], the density matrix of the system may be expressed via the expansion coeficients $C_{nm}(k)$
199: \begin{equation}
200: \rho_{nm}(k)=\sum_{p}C_{np}(k)C_{mp}^*(k).
201: \end{equation}
202: After the substitution of the explicit form of $C_{nm}(k)$ coefficients into (8), we get:
203: \begin{equation}
204: \rho_{nm}(k)=A_{mn}e^{ik(\phi_m-\phi_n)},
205: \end{equation}
206: where
207: \begin{equation}
208: A_{mn}=\sum_p \langle\phi_m | \psi_p(0)\rangle\langle\psi_p(0) | \phi_n\rangle.
209: \end{equation}
210:
211: As seen from (9), non-diagonal matrix elements of the density matrix contain fast oscillating in time exponential phase factors $e^{ik(\phi_m-\phi_n)}$, and the diagonal matrix elements
212: \begin{equation}
213: \rho_{nn}=A_{nn}=\sum_p |\langle\phi_n | \psi_p(0)\rangle |^2.
214: \end{equation}
215:
216: Exponential phase factors of the non-diagonal matrix elements $\rho_{nm}(k)$ express the principle of quantum coherence [23] and correspond to the complete quantum-mechanical description of the system in pure quantum-mechanical state. While they are not equal to zero, the system is in the pure state, and zeroing of these elements is the sign of transition into mixed state [20-22].
217:
218: But here a question appears: if the system was initially in the pure quantum-mechanical state, how can the transition into mixed state take place? Before answering on this question, let us recall, that existence of non-complete quantum information about the state of the system corresponds to the mixed quantum-mechanical state [19-23]. I.e. for transition into the mixed state, a partial loss of information about the state of the system must occur. According to the commonly accepted notions [20], this corresponds to the loss of information about the phases of the system and to the zeroing of non-diagonal elements of the density matrix.
219:
220: So, to prove the formation of the mixed state one has to show
221: the zeroing of non-diagonal elements of density matrix, the
222: equality of which to zero is a sign of a mixed state [19-21].
223:
224:
225: Let us recollect that value $\phi_n$ is the
226: eigenvalue of the Floquet operator. Owing to the non-integrability
227: of the system, eigenvalues can be obtained only by way of
228: numerical diagonalization of the Hamiltonian $H_0$ (3). According
229: to the main hypothesis of the random matrix theory [14,17],
230: the elements of this matrix in the chaotic domain are random numbers.
231: So, their eigenvalues can also be considered as random values.
232:
233: This statement is valid only in the chaotic domain [24,25].
234: According to this, the phase
235: \begin{equation}
236: f(n,m)=\phi_m-\phi_m,
237: \end{equation}
238: in exponential factors of the non-diagonal matrix elements in (9)
239: is random quantity with all its properties. Such an interpretation
240: of the eigenvalues of chaotic quantum-mechanical system is stated
241: in recently published works [24,25]. In these papers the analogy
242: between spectral characteristics of a Hamilton system and random
243: time series is discussed. Further we shall treat $f(n,m)$ phase as
244: a mathematical random variable putting aside its physical meaning.
245: According to the preceding, it is clear that the values of matrix
246: elements of density matrix of the chaotic quantum-mechanical
247: system $\rho_{nm}(k)$ are random values too. Taking a statistical
248: average of expression (9), we have:
249: \begin{equation}
250: \langle \rho_{nm}(k)\rangle = \langle A_{mn}e^{ik(\phi_m -
251: \phi_n)}\rangle ,
252: \end{equation}
253: where $\langle ... \rangle$ denotes the averaging over the statistical ensemble.
254:
255: Before proceeding the discussion, we should define what one implies under quantum-statistical ensemble in our case more precisely.
256:
257: Recall, that while studying classical chaos, one usually examines the stability of the system relatively to a small change of initial conditions and system parameters. A small initial error of these parameters always exists and remains unavoidable (one may measure the parameters of the system at a very high precision, but even in this case there is still a small error, the removal of which, i.e. the measurement at absolute precision, is impossible [10]). So that, not the existence of the unavoidable error is fundamental, but what kind of influence it brings over the system dynamics. It is well known that in case of regular systems such an influence is negligible, but if the system is chaotic, the effect of it increases exponentially [12].
258:
259: Making a complete analogy to classical chaos, let us assume, that the parameter, characterizing interaction between oscillators $x_0$, has some small unavoidable error $\delta x_0$. Then the set of chaotic Hamiltonians (3), with the values of interaction parameter taken from interval $[x_0-\delta x_0, x_0+\delta x_0]$, may be considered as quantum-mechanical ensemble. According to the random matrix theory [14,17,27], there is a random distribution of distances between levels that corresponds to ensemble of chaotic Hamiltonians. In particular for chaotic Hamiltonians with real matrix elements the distribution of the distances between levels obeys normal Gaussian distribution [14,17].
260:
261:
262: The further study of (13) can be performed using more rigorous mathematical
263: substantiation. For this let us recall some details from the
264: probability theory [26].
265:
266: a) In general case, under the characteristic function of the
267: random variable $X$, mathematical expectation of the following
268: exponent is meant
269: \begin{equation}
270: F(t)=M(exp(itX)),
271: \end{equation}
272: where $t$ is a real parameter.
273:
274: b) Mathematical expectation itself is defined as a first initial
275: moment $\mu_1$ of the random variable $X$
276: \begin{equation}
277: M(X)=\langle X\rangle =\mu_1=\sum_k x_kP_k,
278: \end{equation}
279: where $X$ is the discrete random variable which takes possible
280: values $x_1,x_2,...$ with appropriate probabilities
281: $P_k=P(X=x_k), <...>$ means average.
282:
283: Taking (14), (15) into account and considering $f(n,m)$ as a
284: random value, we get:
285: \begin{equation}
286: \langle\rho_{nm}(k)\rangle=A_{mn}F(k),
287: \end{equation}
288: where
289: \begin{equation}
290: F(k)=\langle e^{ikf(n,m)}\rangle,
291: \end{equation}
292: is the mathematical expectation of the characteristic function of
293: the random phase $f(n,m)$, and
294: \begin{equation}
295: A_{mn}=\langle\phi_m|\psi(0)\rangle\langle\psi(0)|\phi_n\rangle.
296: \end{equation}
297:
298: Let us assume that random phase $f(n,m)$ has a normal dispersion
299: [26]. This assumption is based on the fact that matrix elements of Hamiltonian (3) are real values. Then from (17) one can obtain
300: \begin{equation}
301: F(k)=e^{iak}e^{-\frac{\sigma^2k^2}{2}},
302: \end{equation}
303: where
304: \begin{equation}
305: a=M(f(n,m))
306: \end{equation}
307: is the mathematical expectation of the random phase,
308: \begin{equation}
309: \sigma^2=M(f^2(n,m))-(M(f(n,m)))^2.
310: \end{equation}
311:
312: On basis of obtained expressions (16), (19) one may say the following. For small $k$, the system is still in a pure state (non-diagonal matrix elements differ from zero). But after some time the system passes to the mixed state as the non-diagonal matrix elements of the density matrix decrease with exponential law
313: \begin{equation}
314: \rho_{nm}(k)\sim e^{-\frac{\sigma^2k^2}{2}}.
315: \end{equation}
316: This phenomenon is connected with the "phase incursion" [12]. Uncertainty of phase in (9) is accumulated little by little with time. Finally, when $k>\sqrt{2}/\sigma$, the uncertainty of phase is of order $~2\pi$, and the phase is totally chaotized (see [3,4]). As the result the system passes into mixed state.
317:
318: But here a question emerges. How does the uniqueness of the nonlinear chaotic system show itself? If we assume the existence of initial dispersion of parameter for linear system, shall we obtain the same result? Before answering on this question let us consider a peculiarity of already obtained result (22):
319:
320: When approaching to (22), the fact of randomicity of the distribution of distances between levels of energy terms was fundamental. We should notice, that this randomicity was made conditional on the randomicity of the distribution of energy termes themselves and not due to the widening of the line $\delta\phi_n$ of the energy terms $\phi_n$ [14,17].
321:
322: Now let us consider a linear system and see whether this condition is carried out or not.
323:
324: Consider ensemble of linear oscillators
325: \begin{equation}
326: H_L=\frac{p^2}{2m}+\frac{\beta x^2}{2},
327: \end{equation}
328: that are characterized by a small dispersion of parameter $\delta\beta\ll\beta$.
329:
330: The spectrum of linear oscillator is well known and has a regular character [20]
331: \begin{equation}
332: E_n(\omega)=\omega(n+\frac{1}{2}),
333: \end{equation}
334: where $\omega=\sqrt{\beta/m}$ is the distance between the levels of the system.
335:
336: Easy to see, that if uncertainty of parameter $\beta$ is small $\delta\beta\ll\beta$, then this may lead to the widening of the levels, but not to the mixing of them. As a result, the value of such widening in $\omega$ units is:
337: \begin{equation}
338: \delta E_n\approx\delta\omega(n+\frac{1}{2}),
339: \end{equation}
340: where $\delta\omega\Rightarrow\frac{\delta\omega}{\omega}=\frac{1}{2}\delta\beta$.
341:
342: It is already clear, that the condition is not fulfilled. In this case the damping of non-diagonal matrix elements of density matrix may be caused only by the widening of energy terms and not by their random distribution. One should take into account that the widening of term depends linearly on the dispersion $\delta\beta$, and each of the Hamiltonians makes an equal contribution to the average value. This happens because in the case of linear oscillator, each of the Hamiltonians of the ensemble has the same spectrum as others have, and the averaging is performed over the widening of energy terms. In case of chaotic ensemble, each of the Hamiltonians of the system has its own spectrum (due to exponential repulsion of the energy terms of the chaotic ensemble [14,17] they are exponentially sensitive to a small dispersion) and the averaging is performed over these spectrum characteristics, and not over the widening of separate line. When taking such average, the contribution from separate Hamiltonians of the chaotic ensemble is not the same and is described by Gaussian esemble [14,17]. This is the fundamental and not the formal difference between these two cases.
343:
344: Taking all this into account, we obtain for ensemble average matrix elements of linear oscillator:
345: \begin{equation}
346: \langle\rho_{mn}^L(t)\rangle=\frac{1}{2\delta\beta}\int_{-\delta\beta}^{\delta\beta}\rho_{mn}^L(t,\delta\beta')d(\delta\beta').
347: \end{equation}
348:
349: Considering (24), (25) and (26) we get in the end:
350: \begin{equation}
351: \langle\rho_{mn}^L(t)\rangle_{\delta\beta}=\frac{2}{\delta\beta t|n-m|}e^{i(m-n)\omega t}\cos{\left( \frac{\delta\beta}{2}(n-m)\omega t\right)}.
352: \end{equation}
353:
354: Let us start the analysis of the obtained result. Firstly it is clearly seen, that time dependence is not of Gaussian type, but is inverse proportional function of time $1/t$. This as minimum decreases the rate of formation of mixed state in comparison to the case of chaotical ensemble (see Eq.(22)). More than this, the time of formation of the mixed state is completely defined by the value of dispersion $\delta\beta$:
355: \begin{equation}T_c=\frac{2}{|n-m|}\frac{1}{\delta\beta}.
356: \end{equation}
357:
358: Actually this means the following. Observing the evolution of the system on an arbitrary large interval of time $t\in[0,T]$, one can always select such an accuracy while measuring the system parameter $\delta\beta_0$ to fulfill the condition $T<T_c$. This means that the system will be in a pure state for given observation interval $t\in[0,T]$ and given accuracy of measurement $\delta\beta_0$.
359:
360: Hence in some sense we have an analogy with classical chaos. A small initial dispersion of values has an exponential effect, which reveals itself in formation of mixed state. In case of regular system, one may neglect this dispersion with an accurace up to the widening of the energy terms.
361:
362: Finally let us discuss one more question. Can one formally obtain the result (22) in a case of linear oscillator? From a formal mathematical point of view it is possible, but this case will not have a physical sense. For this one has to choose an ensemble of linear oscillators with predetermined random frequences that obey Gaussian distribution. Next one does the averaging over this ensemble, taken over the random frequences and not over the widening of lines. But this case will not have anything common with a real physical situation. The thing is that the Gaussian ensemble of Hamiltonians in case of chaotic system as well as the ensemble of harmonic oscillators in a case of regular system is directly connected with an unavoidable error of measurement of physical parameters of the system, and the averaging over them is not a pure formality.
363:
364: Zeroing of the non-diagonal elements of density matrix is the sign
365: of quantum chaos beginnings and formation of mixed state. From
366: that moment the quantum dynamics is non-reversible, since the
367: information about the wave functions phase is lost.
368:
369: After the formation of a mixed state the quantum-mechanical
370: description loses its sense and we have to use a
371: quantum-statistical interpretation. Along with this, the notion of
372: fidelity loses its meaning, i.e. by the time of the beginning of
373: reverse evolution the system is no longer in the pure state
374: described by the wave-function. The reverse transition from the
375: mixed state to the pure one and reconstruction of the
376: wave-function are impossible because of the irreversibility of the
377: process.
378:
379: Given above reasoning can be checked by numerical diagonalization of the
380: Hamiltonian (2), defining eigenvalues $\varphi_{\chi}$ and by
381: estimation of the averaged non-diagonal elements of the density
382: matrix according to (9).
383:
384: The result of numerical calculations is represented on Fig.1.
385:
386: \begin{figure}[tbp]
387: \includegraphics[scale=0.5]{Image1.eps}
388: \caption{
389: \label{Fig.1} The graph of the dependence $\rho_{nm}$ (9) on time
390: $t=kT_o$ . The graph is obtained using (9) and by way of numerical
391: diagonalization of the Hamiltonian (3) $\hat{H}(P,Q,x(0))$. As it
392: is easily seen from the graph that when $t=kT_o=\sqrt{2}/\sigma>\tau_c,~
393: T_o=\frac{2\pi}{\Omega}=0.1\tau_c$, $\tau_c=1$ [1], zeroing of
394: $\rho_{nm}$ happens. In the numerical calculations we have
395: used definite initial state $\psi(0)$, where $\psi(0)$ is the
396: eigenvector of the Hamiltonian $\hat{H}(P,Q,x(0)), x(o)=x_o+\Delta
397: x_o$, and definite eigenfunctions $|\phi_n\rangle$,
398: $|\phi_m\rangle$, $n=6$, $m=80$ of the Hamiltonian
399: $\hat{H}(P,Q,x_0)$. The dimension of the diagonalized matrix was
400: $1000 \times 1000,\Delta x_o=0.1$. For the statistical averaging
401: of the numerical data we have used ensemble of 50 Hamiltonians,
402: corresponding to the small dispersion of parameter $\delta
403: x_0=0.01$.
404: }
405:
406: \end{figure}
407:
408:
409: \section{Kinetic description of the chaotic quantum-mechanical
410: system.}
411:
412: In previous section we studied mechanism of mixed state
413: formation. After formation of mixed state quantum-mechanical
414: consideration loses meaning and there is a need to use a kinetic
415: description. Kinetic equation for the chaotic quantum-mechanical
416: system for the first time was obtained in [16]. But this study was done in
417: the semi-classical domain. Namely for example, zeroing of
418: non-diagonal part of density matrix was proved by use of
419: semi-classical approximation. The main purpose of given work is
420: the deriving of kinetic equation for chaotic quantum-mechanical
421: system in exceptionally quantum domain without application to the
422: semi-classical methods. For obtaining of muster equation for the
423: density matrix we shall use a method of projection operator [28].
424:
425: Let us split the density matrix operator $\hat{\rho}$ on a slow
426: $\hat{\rho}_{R}$ and a fast varying $\hat{\rho}_{NR}$ operators:
427: $\hat{\rho}=\hat{\rho}_{R}+\hat{\rho}_{NR}$. Relevant part
428: $\hat{\rho}_{R}$ in the basis $|\varphi_n \rangle$ contains only
429: diagonal elements, whereas non-relevant part $\hat{\rho}_{NR}$
430: contains only non-diagonal elements. These elements, as it was shown in previuos section, contain fast oscillating exponents and when taking average over the ensemble, the zeroing of them takes place. Elimination of the diagonal
431: part from the density matrix is a linear operation, which
432: satisfies the property of projection operator $\hat{D}^2=\hat{D}$
433: [28]
434: \begin{equation}
435: \hat{\rho}_{R}=\hat{D}\hat{\rho},~~
436: \hat{\rho}_{NR}=(1-\hat{D})\hat{\rho}.
437: \end{equation}
438:
439: Let us note that this reflection is non-reversible. Due to the
440: zeroing of non-diagonal part of the density matrix, part of
441: information is lost.
442:
443: Inasmuch as relevant statistical operator $\hat{\rho}_{R}(t)$ is
444: different from the total operator $\hat{\rho}(t)$, generally
445: speaking it does not suit the Liouville-Fon Neumann equation [21]
446:
447: \begin{equation}
448: \frac{\partial \hat{\rho}}{\partial
449: t}+i\hat{L}\hat{\rho}=0,
450: \end{equation}
451: where $\hat{L}$ is Liouville operator [19]. After acting on the
452: equation (30) by operator $\hat{D}$, we get
453: $$\frac{\partial
454: \hat{\rho} _{R}}{\partial
455: t}+i\hat{D}\hat{L}(\hat{\rho}_{R}+\hat{\rho}_{NR})=0,$$
456:
457: \begin{equation}
458: \frac{\partial \hat{\rho} _{NR}}{\partial
459: t}+i(1-\hat{D})\hat{L}(\hat{\rho}_{R}+\hat{\rho}_{NR})=0.
460: \end{equation}
461:
462: For the purpose to obtain closed equation for $\hat{\rho} _{R}$ we
463: exclude from the first equation (31) $\hat{\rho} _{NR}$ . As a
464: result we get $$\frac{\partial \hat{\rho} _{R}(t)}{\partial
465: t}+i\hat{D}\hat{L}\hat{\rho}_{R}(t)+\int\limits_{t_o}^{t}
466: K(t-t_1)\hat{\rho}_{R}(t_1)dt_1=-i\hat{D}\hat{L}exp[-i(t-t_o)(1-\hat{D})\hat{L}]\hat{\rho}_{NR}(t_o),$$
467: \begin{equation}
468: K(t-t_1)=\hat{D}\hat{L}exp[-i(t-t_1)(1-\hat{D})\hat{L}](1-\hat{D})\hat{L}.
469: \end{equation}
470: This equation is valid for $t=\sqrt{2}/\sigma>\tau_c$. Evidently
471: $\hat{\rho}_{R}(t)$ is expressed by way of values of
472: $\hat{\rho}_{R}(t_1)$ taken for the time interval $t_o<t_1\leq t$,
473: and additionally trough the value of $\hat{\rho}_{NR}(t_o)$. If in
474: the initial moment of time $t=t_o$ the system is in a pure
475: quantum-mechanical state then $\hat{\rho}_{NR}(0)\not=0$.
476:
477: In this case solving the equation (32) is problematically. However
478: let us recollect that for $t_o>\sqrt{2}/\sigma$ system is already in a
479: mixed state. Therefore equation (32) takes more simple form
480: $(\hat{\rho}_{NR}(t_o)=0)$
481: \begin{equation}
482: \frac{\partial \hat{\rho} _{R}(t)}{\partial
483: t}+i\hat{D}\hat{L}\hat{\rho}_{R}(t)+\int\limits_{t_o}^{t}
484: K(t-t_1)\hat{\rho}_{R}(t_1)dt_1=0.
485: \end{equation}
486:
487: For solving equation (33) we shall use a method of super-operators
488: [29]. But before performing this, we should notice, that we have obtained a closed equation for relevant part of statistical operator. We were able to come to this only because when $t>t_0=\sqrt{2}/\sigma$ the system is in the mixed state and all non-diagonal matrix elements of the density matrix are equal to zero.
489:
490: Further when studying the evolution of the system we shall consider as the origin of time the moment of the formation of the mixed state in the system. This corresponds to a formal transition to limit $t_0\rightarrow -\infty$. Further for simplification of (33) we shall use Abel theorem [26]
491: \begin{equation}
492: \lim_{T\rightarrow \infty}\frac{1}{T}\int_T^0 f(t)dt=f(0)-\lim_{\epsilon\rightarrow +0}\int_{-\infty}^{0}e^{\epsilon(t')}\frac{d}{dt'}f'(t')dt' .
493: \end{equation}
494: Taking (34) into account, Eq.(33) will have the following form:
495: \begin{equation}
496: \frac{\partial\hat{\rho}_R(t)}{\partial t}+iDL\hat{\rho}_R(t)=\lim_{\epsilon\rightarrow +0}\int_{-\infty}^{t}e^{\epsilon (t' -t)}K(t-t')\hat{\rho}_R(t')dt' .
497: \end{equation}
498: According to the method of superoperators [29], the correspondence of one operator to another may be considered as representation. The operator in this case will be represented by a matrix element with two indices, while the linear product of operators is a matrix with four indices, i.e. a superoperator [29]. A concrete example is the projection of $\hat{D}$ operator on diagonal elements. We should notice, that in our case, the projection operator is some definite physical procedure of averaging matrix elements of the density matrix over Gaussian chaotic ensemble.
499:
500: From the relation $\hat{D}\rho_{nm}(t)=\rho_{nm}(t)\delta_{mn}$ we come to the following representation of $\hat{D}$ superoperator $\hat{D}_{nmn'm'}=\delta_{nn'}\delta_{mm'}\delta_{nm}$, so that
501: \begin{equation}
502: \sum_{m'n'}\hat{D}_{mnm'n'}\rho_{m'n'}=\rho_{nm}\delta_{nm}.
503: \end{equation}
504: Taking (36) into account, Eq.(35) is
505: \begin{equation}
506: \frac{d\rho_{nn}(t)}{dt}+i\hat{D}(\hat{L}\hat{\rho}_R)_{nn}=-\int_{-\infty}^{0}dt'[K(-t')\hat{\rho}_R(t+t')]_{nn}e^{\epsilon t'}.
507: \end{equation}
508: In expression (37) and further for short we shall omit $R$ index for diagonal matrix elements of the operator $\hat{\rho}$. Considering the relation
509: \begin{equation}
510: [\hat{L},\hat{\rho}_R]_{nn} = \sum_a L_{nnaa}\rho_{aa}=0,
511: \end{equation}
512: and representing Liouville operator as $L=L_0+L'$ in compliance with (3) we get:
513: \begin{equation}
514: \frac{d\rho_{nn}(t)}{dt}=-\int_{-\infty}^{t}dt_1 e^{\epsilon(t-t_1)}\sum_{n}K_{nnmm}(t-t_1)\rho_{mn}(t_1),
515: \end{equation}
516: and
517: \begin{equation}
518: K_{nnmm}(t)=[L' e^{-it(1-D)L}(1-D)L']_{nnmm}.
519: \end{equation}
520: From (40), (3) and from representation of Liouville operator in form $L=L_0+L'$, one can see, that kernel $K_{nnmm}(t)$ is at least of second order by $\Delta x_0$. Next it is easy to check the corectness of the expression:
521: \begin{equation}
522: \sum_{m}L_{abmm}'=\sum_m(V_{am}\delta_{bm}-V_{mb}\delta_{am})=V_{ab}-V_{ab}=0,
523: \end{equation}
524: for Liouville superoperator
525: \begin{equation}
526: L_{mnm'n'}=(H_{mm'}\delta_{nn'}-H_{nn'}\delta_{mm'}).
527: \end{equation}
528: The relation (41) in its turn leads to the rule of sums $\sum_{m}K_{nnmm}=0$. Taking the symmetries $L_{abcd}=L_{cdab}$, $D_{abcd}=D_{cdab}$ into account, we get from (39):
529: \begin{equation}
530: \frac{d\rho_{nn}(t)}{dt}=-\int_{\infty}^{t}dt_1e^{\epsilon(t-t_1)}\sum_{m\neq n}[K_{nnmm}(t-t_1)\rho_{mm}(t_1)-K_{mmnn}(t-t_1)\rho_{nn}(t_1)].
531: \end{equation}
532: Next we shall make the following approximations. With accuracy up to the value of $(\Delta x_0)^2$ order in the exponent in expression (40) we shall replace the complete Liouville operator $\hat{L}=\hat{L}_0+\hat{L}'$ with $\hat{L}_0$. With the same precision we may set $\rho_{nn}(t-t')=\rho_{nn}(t)$. As a result, from (43) we arrive at
533: \begin{equation}
534: \frac{d\rho_{nn}(t)}{dt}=\sum_{m\neq n}[W_{nm}\rho_{mm}(t)-W_{mn}\rho_{nn}(t)],
535: \end{equation}
536: where
537: \begin{equation}
538: W_{nm}=-\int_{-\infty}^{0}dt' e^{\epsilon t'}[\hat{L}' e^{it(1-\hat{D})\hat{L}_0}(1-\hat{D})\hat{L}']_{nnmm}.
539: \end{equation}
540: Since
541: \begin{equation}
542: (\hat{D}\hat{L}_0)_{abcd}=\delta_{ab}[\phi_a\delta_{ac}\delta_{bd}-\phi_b\delta_{bd}\delta_{ac}]=0,
543: \end{equation}
544: in expression (45) the $\hat{D}$ operator in the argument of exponential function may be omitted.
545:
546: Taking into account time dependence of the operator
547: $x(t)=x_o+\Delta
548: x_of(t),~f(t)=\sum\limits_{\nu=-\infty}^{\infty}e^{i\nu\Omega
549: t},~\Omega=2\pi/T_o$ for
550: non-diagonal matrix elements we get
551: \begin{equation}
552: \frac{d\rho _{nn}(t)}{dt}=\sum_{m\not
553: =n}(W_{nm}\rho_{mm}(t)-W_{mn}\rho_{nn}(t)),
554: \end{equation}
555: where $W_{nm}=\frac{\pi}{2}
556: |V_{nm}|^2\sum\limits_{\nu=-\infty}^{\infty} \delta
557: (E_{nm}-\nu\Omega)$ is the transition amplitude between the
558: eigenstates of the Hamiltonian $\hat{H}_o~$ (3), $E_{nm}=\phi_n-\phi_m$
559: , $V_{nm}$ is the matrix element of the operator $\hat{V}_o=\Delta
560: x_oQ^2_1Q^2_2$ in the basis of eigenfunctions of the Hamiltonian
561: $\hat{H}_o,~ V_{nm}=\langle\psi_n|\hat{V}_o|\psi_m\rangle.$
562:
563: Equation (47) describes non-reversible evolution of the system
564: from non-stationary state to the stationary state defined by the
565: principle of detail equilibrium [20]. To prove irreversibility of
566: the process let us consider time dependence of non-equilibrium
567: entropy [30]
568:
569: \begin{equation}
570: S(t)=-K_B\sum_n\rho_{nn}(t)\ln(\rho_{nn}(t)),\label{25}
571: \end{equation}
572: where $K_B$ is the Boltzmann constant. Taking into account
573: $\sum_n\rho_{nn}(t)=1$ , from (48) we get
574: $$
575: \frac{dS(t)}{dt}=-K_B\sum_n\sum_mW_{nm}[\rho_{mm}(t)-\rho_{nn}(t)]\ln(\rho_{nn}(t))-
576: $$
577: \begin{equation}
578: -K_B\sum_n\frac{\partial \rho_{nn}(t)}{\partial
579: t}=\frac{1}{2}K_B\sum_n\sum_mW_{nm}[\rho_{nn}(t)-\rho_{mm}(t)][\ln(\rho_{nn}(t))-\ln(\rho_{mm}(t))].
580: \end{equation}
581:
582: Due to the property of logarithmic function
583: $$(\rho_{nn}(t)-\rho_{mm}(t))(\ln(\rho_{nn}(t))-\ln(\rho_{mm}(t)))\geq
584: 0$$ we see that $\frac{dS}{dt}\geq 0$. This testifies the growth
585: of entropy during the evolution process.
586:
587: More exact estimation of entropy growth may be obtained from the
588: principle of detail equilibrium [20]. It is evident from the
589: expression of transition probability $W_{nm}=\frac{\pi}{2}
590: |V_{nm}|^2\sum\limits_{\nu=-\infty}^{\infty} \delta
591: (E_{nm}-\nu\Omega)$ that only resonant states $E_{nm}=\nu \Omega$
592: are included in the process. Transitions between states lead to
593: the redistribution of the initial probabilities given by
594: $\rho_{nn}(t_o)$. As a result after the time being we have
595: stationary distribution of the probabilities defined by the
596: principle of detail equilibrium.
597:
598: According to this principle number of transitions from the state
599: $m$ into the state $n$ is equal to the quantity of reverse
600: transitions
601: \begin{equation}
602: W_{nm}\rho_{mm}=W_{mn}\rho_{nn}.
603: \end{equation}
604:
605: Taking into account that in our case $W_{nm}=W_{mn}$ , for the
606: entropy growth we get
607: \begin{equation}
608: \Delta S=S(t\gg \sqrt{2}/\sigma)-S(t=0)=K_B\ln N,
609: \end{equation}
610: where the number of levels $N$, completely included into the process
611: is defined by the spectral characteristics of the
612: Hamiltonian (3) and perturbation (according to (47) ).
613:
614: The authors express their gratitude to Professor A.Ugulava for
615: valuable suggestions and useful discussions.
616:
617: \newpage
618: \begin{references}
619:
620: \bibitem{1} M.Hiller, T.Kottos, D.Cohen, T.Geisel,
621: Phys.Rev.Lett. 92,010402 (2004).
622:
623: \bibitem{2} A.Ugulava, L.Chotorlishvili, and K.Nickoladze,
624: Phys.Rev.E68,026216 (2003).
625:
626: \bibitem{3} A.Ugulava, L.Chotorlishvili, and K.Nickoladze, Phys.Rev.E70,
627: 026219 (2004).
628:
629: \bibitem{4} A.Ugulava, L.Chotorlishvili, and K.Nickoladze, Phys.Rev.E71, 056211
630: (2005).
631:
632: \bibitem{5} F.Haug, M.Bienert, W.Schleich, T.Seligman, M.Raizen, Phys.Rev.
633: A71,043803 (2005).
634:
635: \bibitem{6} T.Prosen, M.Znidaric, J.Phys. A35, 1455 (2002).
636:
637: \bibitem{7} G.Benenti, G.Casati, Phys.Rev.E65, 066205 (2002).
638:
639: \bibitem{8} R.Jalbert, H.Pastawski, Phys.Rev.Lett. 86,2490 (2001).
640:
641: \bibitem{9} V.Flambaum, F.Izrailev, Phys.Rev. E64, 036220 (2001).
642:
643: \bibitem{10} A.J.Lichtenberg, M.A.Lieberman, Regular and Stochastic Motion,
644: (Springer Verlag , New York, Heidelberg, Berlin,1983).
645:
646: \bibitem{11} M.C.Gutzwiller, Chaos in Classical and Quantum Mechanics
647: (Springer, New York,1990).
648:
649: \bibitem{12} R.Z.Sagdeev, G.M.Zaslavsky, Introduction to the Nonlinear
650: Physics (in Russian), (Nauka, Moscow, 1988).
651:
652: \bibitem{13} K.T.Alligood, T.D.Sauer, J.York, Chaos and Introduction to
653: Dynamical Systems (Springer, New York,1996).
654:
655: \bibitem{14} F.Haake, Quantum Signatures of Chaos (Springer, Berlin,2001).
656:
657: \bibitem{15} P.Gaspard, Chaos, Scattering and Statistical Mechanics
658: (Cambridge Univ.Press, Cambridge, 1998).
659:
660: \bibitem{16} G.M.Zaslavsky, Chaos in Dynamical Systems (in Russian), (Nauka,
661: Moscow, 1984).
662:
663: \bibitem{17} H.J.Stockman, Quantum Chaos, An Introduction (Cambridge
664: Univ.Press, Cambridge, 1993).
665:
666: \bibitem{18} D.Cohen, T.Kottos, Phys.Rev.E63, 036203 (2001).
667:
668: \bibitem{19} L.D.Landau and E.M.Lifschitz, Statistical Mechanics,
669: v.5, (in Russian) (Nauka Moscow 1976)
670:
671: \bibitem{20} L.D.Landau and E.M.Lifshitz , Quantum Mechanics. Nonrelativistic
672: Theory (Pergamon, Oxford, 1977).
673:
674: \bibitem{21} R.P.Feynman, Statistical Mechanics (W.A.Benjamin,
675: Inc.Massachusetts,1972).
676:
677: \bibitem{22} K.Blum, Density Matrix Theory and Applications (Plenum Press, New York and London, 1981)
678:
679: \bibitem{23} W.P.Schleich, Quantum Optics in Phase Space (Wiley, Berlin, Weinheim, NY, Toronto, 2001)
680:
681: \bibitem{24} A.Relano, J.M.G.Gomer, R.A.Molina, J.Retamosa and
682: E.Faleiro, Phys.Rev.Lett.89,244102 (2002).
683:
684: \bibitem{25} A.Relano, J.Retamosa, E.Faleiro and J.M.G.Gomer, Phys.Rev.E,066219 (2005).
685:
686: \bibitem{26} W.Feller, An Introduction to Probability Theory and
687: Its Applications, John Wiley @ Sons, Inc. New York, London, Sidney
688: v1(1958), v2 (1966).
689:
690: \bibitem{27} M.Hiller, D.Cohen, T. Geisel, T.Kottos,
691: arXiv:cond-mat/0506300 v1 (2005)
692:
693: O.Bohigas, M.Giannoni, C.Schmidt, Phys.Rev.Lett. 52, 1 (1984)
694:
695: \bibitem{28} G.Ropke, Statistische Mechanik fur das Nichtgleichgewicht ( VEB
696: Deutscher Verlag der Wissenschaften, Berlin, 1987).
697:
698: \bibitem{29} S.Nakajima, Progr.Theor.Phys. 20,948 (1958).
699:
700: \bibitem{30} S.Fujita, Introduction to Non-Equilibrium Quantum Statistical Mechanics (W.B.Saunders Company, Philadelphia-London,
701: 1966).
702:
703: \end{references}
704:
705:
706:
707: \end{document}
708: