quant-ph0308112/eco.tex
1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: %%%  (December 2003)
3: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4: 
5: \documentclass[aps,prl,twocolumn,floats]{revtex4}
6: \usepackage{epsfig}
7: \usepackage{bm}
8: \usepackage{latexsym}
9: 
10: \begin{document}
11: 
12: \newcommand{\hide}[1]{}
13: \newcommand{\tbox}[1]{\mbox{\tiny #1}}
14: \newcommand{\half}{\mbox{\small $\frac{1}{2}$}}
15: \newcommand{\sinc}{\mbox{sinc}}
16: \newcommand{\const}{\mbox{const}}
17: \newcommand{\trc}{\mbox{trace}}
18: \newcommand{\intt}{\int\!\!\!\!\int }
19: \newcommand{\ointt}{\int\!\!\!\!\int\!\!\!\!\!\circ\ }
20: \newcommand{\eexp}{\mbox{e}^}
21: \newcommand{\bra}{\left\langle}
22: \newcommand{\ket}{\right\rangle}
23: \newcommand{\EPS} {\mbox{\LARGE $\epsilon$}}
24: \newcommand{\ar}{\mathsf r}
25: \newcommand{\im}{\mbox{Im}}
26: \newcommand{\re}{\mbox{Re}}
27: \newcommand{\bmsf}[1]{\bm{\mathsf{#1}}} 
28: 
29: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
30: 
31: \title{Quantum Reversibility: Is there an Echo?}
32: 
33: \author{
34: Moritz Hiller$^{1}$, 
35: Tsampikos Kottos$^{1}$, 
36: Doron Cohen$^{2}$ 
37: and Theo Geisel$^{1}$
38: }
39: 
40: \affiliation{
41: $^{1}$Max-Planck-Institut f\"ur Str\"omungsforschung und Fakult\"at Physik
42: der Universit\"at G\"ottingen, Bunsenstra\ss e 10, D-37073 G\"ottingen, Germany \\
43: $^{2}$Department of Physics, Ben-Gurion University, 
44: Beer-Sheva 84105, Israel
45: }
46: 
47: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
48: 
49: \begin{abstract}
50: We study the possibility to undo the quantum mechanical evolution in a time reversal 
51: experiment. The naive expectation, as reflected in the common terminology (``Loschmidt 
52: echo"), is that maximum compensation results if the reversed dynamics extends to the 
53: same time as the forward evolution. We challenge this belief, and demonstrate that the 
54: time $t_r$ for maximum return probability is in general shorter. We find that $t_r$ 
55: depends on $\lambda = \varepsilon_{\rm evol}/\varepsilon_{\rm prep}$, 
56: being the ratio of the error in setting the parameters (fields) for the time reversed 
57: evolution to the perturbation which is involved in the preparation process.
58: Our results should be observable in spin-echo experiments where the dynamical 
59: irreversibility of quantum phases is measured.
60: \end{abstract}
61: 
62: \maketitle
63: 
64: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
65: 
66: In this Letter we study the probability of return $P(t_1,t_2)$ for a generalized wavepacket 
67: dynamics scenario. The system is prepared in some initial state $\Psi_{\tbox{prep}}$, which 
68: can be regarded as the outcome of a preparation procedure which is governed by a Hamiltonian 
69: ${\cal H}_{\tbox{prep}}$. We assume that the quantum mechanical evolution is generated by
70: Hamiltonians with classically chaotic limit: The state is propagated for a time $t_1$ 
71: using a Hamiltonian ${\cal H}_1$, and then the evolution is time-reversed for a time $t_2$  
72: using a perturbed Hamiltonian ${\cal H}_2$. The corresponding evolution operators are $U_1$ 
73: and $U_2$. The probability of return to the initial state is 
74: \begin{eqnarray}
75: \label{fkernel1}
76: P(t_1,t_2)= 
77: |\langle \Psi_{\tbox{prep}} |U_2(t_2)^{-1} U_1(t_1) | \Psi_{\tbox{prep}} \rangle|^2
78: \end{eqnarray}
79: There are two special cases that have been extensively studied in the literature. The 
80: traditional wavepacket dynamics scenario \cite{H91} is obtained if we set $t_2=0$. In 
81: this context the ``survival probability" is defined as 
82: \begin{eqnarray}
83: \label{fkernel2}
84: P_{\tbox{SR}}(t) \ = \ P(t,0) 
85: \end{eqnarray}
86: The ``Loschmidt echo" (LE) scenario is obtained if we set $t_1=t_2=t$. In this context 
87: the ``fidelity" is defined as 
88: \begin{eqnarray}
89: \label{fkernel3}
90: P_{\tbox{LE}}(t) \ = \ P(t,t) 
91: \end{eqnarray}
92: The theory of the fidelity was the subject of intensive studies during the last 3~years 
93: \cite{P84,JP01,JSB01,CT02,PS02,BC02,WC02,PLU95,VH03}. It has been adopted as a standard 
94: measure for quantum reversibility following \cite{P84} and its study was further motivated 
95: by the realization that it is related to the analysis of dephasing in mesoscopic systems 
96: \cite{Z91}.
97: 
98: 
99: In the present Letter we consider the full scenario 
100: of a time reversal experiment. The probability to find 
101: the system in its original state is $P(t)=P(t,0)$ 
102: before the time reversal ($t<T/2$), and 
103: %
104: \begin{eqnarray}
105: \label{fkernel4}
106: P(t) \ = \ P(T/2,t-T/2) 
107: \end{eqnarray}
108: %
109: after the time reversal ($t>T/2$).
110: The period $T$ is the total time of the experiment.
111: %
112: The naive expectation, which is also reflected in the term ``Loschmidt echo", 
113: is to have a maximum for $P(t)$ at the time $t=T$. 
114: We are going to show that this expectation is wrong. 
115: We find that the maximum return probability 
116: is obtained at a time $t_r$ which
117: in general is shorter than that. Namely,
118: % 
119: \begin{equation}
120: \label{tr}
121: T/2 \le t_r \le T. 
122: \end{equation}
123: %
124: If we have $t_r=T/2$ we say that there is no reversibility. 
125: If we have $t_r \sim T$ we say that we have a nearly perfect echo. 
126: We show that $t_r/T$ is a function of a dimensionless 
127: parameter $0< \lambda < \infty$. Namely,   
128: %
129: \begin{eqnarray}
130: \label{sf}
131: {t_r\over T} = f(\lambda)\quad ;\quad \lambda = \varepsilon_{\tbox{evol}}/\varepsilon_{\tbox{prep}}
132: \end{eqnarray}
133: %
134: where $\varepsilon_{\tbox{prep}}$ quantifies the difference 
135: between the evolution Hamiltonian ${\cal H}$ 
136: and the preparation Hamiltonian ${\cal H}_{\tbox{prep}}$, 
137: while $\varepsilon_{\tbox{evol}}$ quantifies the difference 
138: between the two instances ${\cal H}_1$ and ${\cal H}_2$ 
139: of the evolution Hamiltonian, which are used for the forward 
140: and for the time-reversed evolution respectively. 
141: %
142: The idea is that there is no way to have a complete control over 
143: the parameters (fields) of the systems. Therefore there is 
144: an unavoidable difference ($\varepsilon_{\tbox{evol}}$) 
145: between these two instances of ${\cal H}$, which by the setup of 
146: the experiment are regarded as identical. 
147: %
148: The scaling function (\ref{sf}) takes the limiting value $f(0)=1$ (echo) 
149: while for $\lambda > \lambda^*$ we get $f(\lambda) = 0.5$ (no reversibility). 
150: Here $\lambda^*$ is some system-specific constant of order unity.
151: 
152: The most popular preparation which is considered in the literature, either in the context of 
153: wavepacket dynamics or fidelity (LE) studies, is a Gaussian wavepacket. Obviously the choice 
154: of such preparation is motivated mainly by the wishful thinking of theoreticians. However, in 
155: many applications, one is not so much interested in evolving an initial Gaussian wavepacket. 
156: This is certainly the situation in quantum information processing \cite{NC00} and in spin-echo 
157: experiments \cite{PLU95} where one starts with a {\it random} initial state. Formally a 
158: Gaussian wavepacket can be regarded as the ground state of a phase-space shifted Harmonic 
159: oscillator. Therefore it is characterized by a very large $\varepsilon_{\tbox{prep}}$, 
160: leading to $\lambda \ll 1$. In this Letter we do not assume $\lambda \ll 1$, but rather 
161: consider the general case. 
162: 
163: In order to develop a general theory we need a model in which we have control over both 
164: $\varepsilon_{\tbox{prep}}$ and $\varepsilon_{\tbox{evol}}$. We consider a quantized system
165: whose classical analog has positive Lyapunov exponent. Its Hamiltonian ${\cal H} = {\cal 
166: H}(Q,P;x)$ depends on a parameter (field) variable~$x$  
167: which is determined by the experimental setup. For example it can be either a gate voltage or 
168: a magnetic flux. The dynamics takes place within a classically small (but quantum mechanically 
169: large) energy window. The classical dynamics is assumed to have a well defined finite correlation 
170: time $\tau_{\rm cl}$. We consider classically small (but possibly quantum mechanically large)  
171: perturbations ($x \mapsto x+\delta x$). Accordingly, the Hamiltonian can be linearized as follows: 
172: \begin{eqnarray}
173: \label{WBM}
174: {\cal H} = {\cal E} + \delta x {\cal B}
175: \end{eqnarray}
176: We define ${\cal H}_1$ and ${\cal H}_2$ by setting $\delta x =  \pm\varepsilon_{\tbox{evol}}$. 
177: The requirement of having {\em classically small} $\delta x$ means that 
178: the phase space structure of ${\cal H}_1$ and ${\cal H}_2$ is similar, 
179: and that any (small) difference in the chaoticity can be neglected. 
180: (we have verified that this smallness condition is satisfied for the example below).  
181: 
182:    
183: 
184: The preparation issue requires further discussion. The traditional possibility is to prepare a 
185: Gaussian wavepacket (also known as a coherent state preparation). We regard this possibility as 
186: {\em uncontrolled} because the value of  $\varepsilon_{\tbox{prep}}$ is ill defined. To have a 
187: physically meaningful definition of $\varepsilon_{\tbox{prep}}$ the natural procedure is as 
188: follows: We define a preparation Hamiltonian ${\cal H}_{\tbox{prep}}$ by setting $\delta x =  
189: \varepsilon_{\tbox{prep}}$. Then we  start with an initial eigenfunction of ${\cal E}$, and 
190: evolve it with ${\cal H}_{\tbox{prep}}$ until we get an ergodic-like steady state within an 
191: energy shell (note \cite{rmrk}). 
192: The width of this energy shell is proportional to $\varepsilon_{\tbox{prep}}$. 
193: The resulting wavepacket is used as an initial state for the time reversal experiment. 
194: For the purpose of comparison we shall consider also a Gaussian wavepacket preparation. 
195: For such preparation $\lambda \ll 1$ irrespective of its energy width.
196: The reason is that this preparation does not occupy ergodically its energy shell. 
197: Formally one can say that for a Gaussian wavepacket ${\cal H}_{\tbox{prep}}$ differs  
198: enormously from the evolution Hamiltonian. There is no point in quantifying this 
199: difference. This is the reason why we use our controlled preparation procedure, where 
200: ${\cal H}_{\tbox{prep}}$ differs from the evolution Hamiltonian in a well defined manner. 
201: In any case we shall verify that a Gaussian wavepacket is indeed 
202: like taking a preparation with $\lambda \ll 1$. 
203: 
204: 
205: In our numerical investigation we use the model Hamiltonian
206: \begin{eqnarray} 
207: \label{2DW}
208: {\cal H}(Q,P;x) = \half(P_1^2{+}P_2^2 + Q_1^2{+}Q_2^2) + x Q_1^2 Q_2^2
209: \end{eqnarray}
210: with $x=1+\delta x$. It describes the motion of a particle in a 2D well (2DW). The physical 
211: units are chosen so as to have dimensionless variables. Therefore upon quantization the Planck 
212: constant $\hbar$ is a dimensionless  quantity. Our numerical study is focused on an energy 
213: window around $E \approx 3$ where the motion is mainly chaotic with characteristic correlation 
214: time $\tau_{\rm cl} \approx 1$ \cite{CK01}. The quantization is done with $\hbar=0.012$. We write 
215: the Hamiltonian matrix as in Eq.~(\ref{WBM}), using a basis such that ${\cal E}$ is diagonal. 
216: The mean level spacing is $\Delta \approx 4.3 \times \hbar^2$. As expected, on the basis of a 
217: general ``quantum chaos" argumentation \cite{FP86}, the matrix ${\cal B}$ is a banded matrix. 
218: More details regarding the band profile can be found in Ref.~\cite{CK01}. The only additional 
219: piece of information that is needed for the following analysis is the parametric scale $\delta 
220: x_c$. This is defined as the $\delta x$ which is needed in order to mix neighboring levels. 
221: It is given by the ratio $\Delta/\sigma$, where $\sigma$  is the root mean square value of 
222: the near diagonal matrix elements of the ${\cal B}$ matrix. For the above model $\delta x_c 
223: \approx 3.8*\hbar^{3/2}$.  
224: 
225: Fig.~1 (left panel) displays representative results of simulated time reversal experiments. One 
226: experiment is done with a coherent state preparation, and we indeed see behavior that looks like 
227: an echo ($t_r\sim T$). Qualitatively the same behavior is observed for a random preparation that 
228: has $\lambda \ll 1$. Once we take a preparation with a larger value of $\lambda$, we realize that 
229: the compensation time is in general $t_r < T$. In particular with the $\lambda > 1$ preparation 
230: we do not observe any quantum reversibility ($t_r=T/2$). 
231: 
232: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
233: \begin{figure}
234: \epsfig{clip,figure=eco_fig1,scale=0.35,angle=0}
235: \caption
236: {
237: Left panels:  
238: The probability $P(t)$ to find the system in its original state, for the 2DW model. 
239: The two panels, from top to bottom, are for the representative values 
240: $\varepsilon_{\tbox{evol}}=0.00315$ and $0.26$. 
241: We use various preparations: a coherent 
242: state preparation (solid line), a random superposition with the same energy width 
243: ($\circ$), a random superposition (dot-dashed line) with a larger~$\lambda$ where 
244: $\epsilon_{\tbox{prep}}= 0.4$, and a $\lambda > 1$ preparation (dashed line) which 
245: was obtained by setting $\epsilon_{\tbox{prep}}= 0$. The arrows indicate the time $t_r$ 
246: for the case $\epsilon_{\tbox{prep}}= 0.4$. Right panel: The simulations are now done 
247: with the corresponding RMT model. 
248: }
249: \end{figure}
250: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
251: 
252: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
253: \begin{figure}[b]
254: \epsfig{clip,figure=eco_fig2,height=0.7\hsize,angle=0}
255: \caption
256: {
257: The time $t_r$ is calculated from the simulation of $P(t)$,  
258: and the ratio $t_r/T$ is presented for various values of the period $T$, 
259: and for various values (see legend) of $\varepsilon_{\tbox{evol}}$. 
260: An average over a number of initial wavepackets 
261: with the same $\varepsilon_{\tbox{prep}}=0.4$ was performed. We clearly see that $t_r$ 
262: is smaller for larger values of $\varepsilon_{\tbox{evol}}$. The left panel corresponds 
263: to the 2DW simulations, while the right panel to the effective RMT model. 
264: }
265: \end{figure}
266: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
267: 
268: In Fig.~2 we present some results for $t_r/T$. We clearly see that $t_r$ is smaller for 
269: larger values of $\varepsilon_{\tbox{evol}}$. Much more illuminating is Fig.~3 where we 
270: present our numerical results for $t_r$ and various $T$. The points corresponding to the 
271: same $\lambda$ but different values of $\varepsilon_{\tbox{evol}}$ and $\varepsilon_{\tbox{prep}}$ 
272: fall onto the same smooth curve with a good accuracy, confirming the scaling hypothesis 
273: (\ref{sf}). Moreover, we see that for $\lambda>\lambda^*$, with $\lambda^* \approx 1.3$, 
274: we have $t_r=T/2$ (no echo) irrespective of the actual value of $\varepsilon_{\tbox{prep}}$
275: and $\varepsilon_{\tbox{evol}}$. Thus, if we want a $\lambda > \lambda^*$ type of preparation, 
276: we simply can set \mbox{$\varepsilon_{\tbox{prep}}=0$}, which means an eigenstate of ${\cal E}$. 
277: From Fig.~3 we also see that in the other limiting case of $\lambda\ll 1$ we have (nearly) 
278: an echo. This is the same as in the case of a Gaussian wavepacket preparation. 
279: 
280: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
281: \begin{figure}[b]
282: \epsfig{clip,figure=eco_fig3,height=0.7\hsize,angle=0}
283: \caption
284: {
285: The ratio $t_r/ T$ for different initial random wavepackets which are determined by 
286: $\varepsilon_{\tbox{prep}}$,  and for various $\varepsilon_{\tbox{evol}}$. 
287: The horizontal axis is the scaled variable $\lambda$. 
288: The data scale nicely in accordance with Eq.~(\ref{sf}). 
289: The dashed line is a simple polynomial fit. The left panel 
290: corresponds to the 2DW simulations while the right one to the corresponding RMT model. 
291: }
292: \end{figure}
293: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
294: 
295: In order to explain the observed results for $t_r$ 
296: we look on Fig.~4. There are two axes along which 
297: the theory of $P(t_1,t_2)$ is quite well known. 
298: One is the traditional wavepacket dynamics axis 
299: $t_2=0$ along which $P_{\tbox{SR}}(t)$ is defined, 
300: while the other is the ``LE axis" ($t_2=t_1$) 
301: along which $P_{\tbox{LE}}(t)$ is defined. 
302: In the figure we also indicate the course 
303: of a time reversal experiment.  
304: It is clear that in order to have $t_r<T$ 
305: the contor line $P(t_1,t_2)=P_{\tbox{SR}}(T/2)$ 
306: should meet the $t_1$ axis in a sharp angle. 
307: Furthermore, if we want to have some remnant of 
308: an echo at the end of the period ($t=T$)  
309: we have to cross the contour line  
310: $P(t_1,t_2)=P_{\tbox{LE}}(T/2)$. 
311: The condition for that is 
312: %
313: \begin{eqnarray}
314: P_{\tbox{SR}}(T/2) < P_{\tbox{LE}}(T/2)
315: \end{eqnarray}
316: %
317: In the following discussion we would like to assume that both $\varepsilon_{\tbox{prep}}$ 
318: and $\varepsilon_{\tbox{evol}}$ are larger than $\delta x_c$, which means that the 
319: perturbations are strong enough to mix levels. In such case the decay of $P_{\tbox{SR}}(t)$ 
320: or $P_{\tbox{LE}}(t)$ is approximately exponential: 
321: \begin{eqnarray}
322: \label{expon}
323: P_{\tbox{SR}}(t), \ P_{\tbox{LE}}(t) \ = \ \exp(-\gamma t)
324: \end{eqnarray}
325: Moreover in both cases $\gamma$ is given by the expression 
326: \begin{eqnarray}
327: \gamma = \min( \gamma_{\tbox{PT}}, \gamma_{\tbox{SC}} )
328: \end{eqnarray}
329: where $\gamma_{\tbox{PT}}$ is the value which is determined by perturbation theory, while  
330: $\gamma_{\tbox{SC}}$ is determined by semiclassical considerations.   
331: 
332: Once details are concerned the theory behind the exponential approximation Eq.~(\ref{expon}) 
333: becomes quite different in the two respective cases [$P_{\tbox{SR}}(t)$, $P_{\tbox{LE}}(t)$]. 
334: The theory of the survival probability is related to the parametric theory of the LDOS 
335: \cite{CH00,CK01}. Namely, for relatively small perturbations $\gamma = \gamma_{\tbox{PT}} 
336: \propto (\varepsilon_{\tbox{prep}} / \delta x_c)^2$ is essentially the width of Wigner's Lorentzian, 
337: while for large perturbations $\gamma = \gamma_{\tbox{SC}} \propto \varepsilon_{\tbox{prep}}$ 
338: is the width of the energy shell (in the latter case the exponential approximation is at best 
339: a good fit). In contrast to that, the theory of the fidelity $P_{\tbox{LE}}(t)$ is related to 
340: a theory of dynamical correlations, and cannot be reduced to the LDOS analysis \cite{WC02}. 
341: The best theory to date is semiclassical (see ~\cite{VH03} and references therein). As in the 
342: case of the survival probability we have 
343: $\gamma = \gamma_{\tbox{PT}} \propto (\varepsilon_{\tbox{evol}}/ \delta x_c)^2$ for small perturbations, while for large perturbations $\gamma = 
344: \gamma_{\tbox{SC}}$, in some typical cases, is related to the Lyapunov exponent. We would like 
345: to point out that the study of the general conditions for having a fingerprint of the Lyapunov 
346: exponent in time reversal experiments is still an open issue for future study \cite{HKC03}. 
347: The existing semiclassical theory for the ``echo" phenomena assumes Gaussian wavepackets.  
348:   
349: By definition $P_{\tbox{SR}}(T/2)$ depends only on $\varepsilon_{\tbox{prep}}$, while 
350: $P_{\tbox{LE}}(T/2)$ is mainly sensitive to $\varepsilon_{\tbox{evol}}$. Therefore it 
351: is evident that a small $\lambda$ is a condition for having $t_r<T$. If we have $\lambda
352: \ll 1$ then the contour line $P(t_1,t_2)=P_{\tbox{LE}}(T/2)$ can be very close to the 
353: LE-axis. This implies that for $\lambda\ll 1$ we can get nearly 
354: a perfect echo behavior ($t_r\sim T$).
355: This picture, as we have seen before, is supported by our numerical findings.
356:  
357: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
358: \begin{figure}
359: \epsfig{clip,figure=echo,scale=0.6,angle=0}
360: \caption
361: {
362: The contour line of $P(t_1,t_2)$ 
363: that goes through the point $(T/2,T/2)$. 
364: One solid line (``no echo") is for the case $\lambda\geq 1$. 
365: The other solid line is for the case of a relatively 
366: small~$\lambda$. The dashed line illustrates   
367: the course of a time reversal experiment, 
368: while the ``LE axis" is the line along 
369: which $P_{\tbox{LE}}$ is defined.
370: }
371: \end{figure}
372: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
373: 
374: As we have seen above, the applicability of semiclassical considerations is not essential 
375: for having an ``echo". The general picture that we have outlined should be valid also in 
376: the absence of a semiclassical limit. This is in contrast to the impression that one might 
377: get from the recent literature. 
378: In order to establish this provocative statement, in a way that leaves no doubts, 
379: we use a simple random matrix theory (RMT) procedure. We take the 
380: resulting banded matrix ${\cal B}$ of the 2DW model~(\ref{2DW}), and randomize the signs 
381: of the off-diagonal terms. In this way we get an effective RMT (ERMT) model of the type 
382: that had been introduced by Wigner 50~years ago \cite{wigner}. The model is characterized 
383: by the same mean level spacing, and by the same band-profile as the physical 2DW model. 
384: Consequently the generated dynamics is characterized by the same correlation time (the 
385: latter is determined by the bandwidth). But unlike the 2DW model, the ERMT model is 
386: lacking a semiclassical limit. In the right panels of figures 1-3 we demonstrate the results 
387: of simulations that were done with the ERMT model. We clearly see that we get similar 
388: results (in Fig.3 the RMT drop is slightly sharper).
389: 
390: 
391: In summary, we have shed a new light on the physics of quantum reversibility, and in 
392: particular we have introduced the concept of compensation time~$t_r$, which replaces 
393: the misleading terminology of ``echo". Our predictions should be tested in wave field 
394: evolution experiments such as spin polarization echoes in nuclear magnetic resonances
395: \cite{PLU95,prv}. In particular we have considered the realistic case of a general 
396: preparation, and clarified the role of semiclassical considerations in the theory. 
397: 
398: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
399: \ \\
400: It is our pleasure to thank Horacio Pastawski and Bilha Segev (BGU) for useful discussions. 
401: This research was supported by the Israel Science Foundation (grant No.11/02), and by a grant 
402: from the GIF, the German-Israeli Foundation for Scientific Research and Development.
403: 
404: 
405: 
406: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
407: 
408: \begin{thebibliography}{99}
409: 
410: \bibitem{H91} E. J. Heller, {\it Chaos and Quantum Systems}, edt. M.-J. Giannoni et al. 
411: (Elsevier, Amsterdam), (1991)
412: 
413: \bibitem{P84} A. Peres, Phys. Rev. A {\bf 30}, 1610 (1984).
414: 
415: \bibitem{JP01} R.A. Jalabert and H.M. Pastawski, Phys. Rev. Lett. {\bf 86}, 2490 (2001);
416: F.M. Cucchietti, H.M. Pastawski, R. Jalabert Physica A, {\bf 283}, 285 (2000); F.M. 
417: Cucchietti, H.M. Pastawski and D.A. Wisniacki Phys. Rev. E {\bf 65}, 046209 (2002); 
418: 
419: \bibitem{JSB01} Ph. Jacquod, I. Adagdeli and C.W.J. Beenakker, Phys. Rev. Lett. {\bf 89}, 
420: 154103 (2002); Ph. Jacquod, I. Adagdeli and C.W.J. Beenakker, Europhys. Lett. {\bf 61},
421: 729 (2003); Ph. Jacquod, P.G. Silvestrov and C.W.J. Beenakker, Phys. Rev. E {\bf 64}, 
422: 055203(R) (2001). 
423: 
424: \bibitem{CT02} N.R. Cerruti and S. Tomsovic, Phys. Rev. Lett. {\bf 88}, 054103 (2002);
425: J. Phys. A: Math. Gen. {\bf 36}, 3451 (2003). 
426: 
427: \bibitem{PS02} T. Prosen, Phys. Rev. E {\bf 65}, 036208 (2002); T. Prosen and M. 
428: Znidaric, J. Phys. A: Math. Gen {\bf 35}, 1455 (2002); ibid {\bf 34}, L681 (2001); T. 
429: Prosen and T. H. Seligman, ibid {\bf 35}, 4707 (2002); T. Kottos and D. Cohen, Europhys. 
430: Lett. {\bf 61}, 431 (2003).
431: 
432: \bibitem{BC02} G. Benenti and G. Casati, Phys. Rev. E {\bf 66}, 066205 (2002); W. Wang
433: and B. Li, ibid. {\bf 66}, 056208 (2002).
434: 
435: \bibitem{WC02}D. A. Wisniacki and D. Cohen, ibid. {\bf 66}, 046209 (2002).
436: 
437: \bibitem{VH03} Jiri Vanicek and Eric J. Heller, quant-ph/0302192.
438: 
439: \bibitem{PLU95} H. M. Pastawski, P. R. Levstein and G. Usaj, Phys. Rev. Lett. {\bf 75}, 4310 (1995);
440: G. Usaj, H. M. Pastawski and P. Levstein, Molecular Physics {\bf 95}, 1229 (1998).
441: 
442: \bibitem{Z91} W. H. Zurek, Phys. Today {\bf 44}, 36 (1991); D. Cohen and T. Kottos 
443: cond-mat/0302319; D. Cohen, Phys. Rev. E {\bf 65}, 026218 (2002).
444: 
445: \bibitem{NC00} M. A. Nielsen and I. L. Chuang, {\it Quantum Computation and Quantum Information}
446: (Cambridge University Press, 2000).
447: 
448: \bibitem{CK01} Doron Cohen, and Tsampikos Kottos, Phys. Rev. E, {\bf 63} 36203, (2001).
449: 
450: \bibitem{FP86} M. Feingold and A. Peres, Phys. Rev. A {\bf 34} 591, (1986); M.
451: Feingold, D. Leitner, M. Wilkinson, Phys. Rev. Lett. {\bf 66}, 986 (1991); M.
452: Feingold, A. Gioletta, F. M. Izrailev, L. Molinari, ibid. {\bf 70}, 2936 (1993);
453: M. Wilkinson, M. Feingold, D. Leitner, J. Phys. A {\bf 24}, 175 (1991).
454: 
455: \bibitem{CH00} Doron Cohen, and E. J. Heller, Phys. Rev. Lett. {\bf 84} 2841, (2000).
456: 
457: \bibitem{HKC03} M. Hiller, T. Kottos and D. Cohen, in preparation (2003).
458: 
459: \bibitem{wigner} E. Wigner, Ann. Math {\bf 62} 548 (1955); {\bf 65} 203 (1957);
460: V.V. Flambaum, A.A. Gribakina, G.F. Gribakin and M.G. Kozlov, Phys. Rev. A {\bf 50}
461: 267 (1994); G. Casati, B.V. Chirikov, I. Guarneri, F.M. Izrailev, Phys. Rev. E
462: {\bf 48}, R1613 (1993); \ Phys. Lett. A {\bf 223}, 430 (1996); Ph. Jacquod and D.
463: L. Shepelyansky, Phys. Rev. Lett. {\bf 75}, 3501 (1995); Y. V. Fyodorov, O. A.
464: Chubykalo, F. M. Izrailev, and G. Casati, ibid. {\bf 76}, 1603 (1996).
465: 
466: \bibitem{prv} H. M. Pastawski, private communication.
467: 
468: \bibitem{rmrk} To ensure phase space (semiclassical) ergodicity 
469: of the wavefunction within the energy shell we randomize the signs 
470: of the state vector elements in the ${\cal H}_0$ basis.    
471: 
472: \end{thebibliography}
473: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
474: \end{document}
475: