1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: %%% Jiri Vanicek and Doron Cohen (March 2003) (July 2003)
3: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4:
5:
6: \documentclass{iopart}
7: \usepackage{epsfig}
8: \usepackage{latexsym}
9:
10: \begin{document}
11:
12: \newcommand{\hide}[1]{}
13: \newcommand{\tbox}[1]{\mbox{\tiny #1}}
14: \newcommand{\mbf}[1]{{\mathbf #1}}
15: \newcommand{\half}{\mbox{\small $\frac{1}{2}$}}
16: \newcommand{\sinc}{\mbox{sinc}}
17: \newcommand{\const}{\mbox{const}}
18: \newcommand{\trc}{\mbox{trace}}
19: \newcommand{\ointt}{\int\!\!\!\!\int\!\!\!\!\!\circ\ }
20: \newcommand{\eexp}{\mbox{e}^}
21: \newcommand{\bra}{\left\langle}
22: \newcommand{\ket}{\right\rangle}
23:
24:
25: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
26:
27: \title
28: [Survival probability and LDOS for 1D systems]
29: {Survival probability and local density of states for one-dimensional Hamiltonian systems}
30:
31: \author{Ji\v{r}\'{\i} Van\'{\i}\v{c}ek$^{1,2}$ and Doron Cohen$^3$}
32:
33: \date{March 2003}
34:
35: \address{
36: $^1$ \mbox{Department of Physics, Harvard University, Cambridge, MA 02138} \\
37: $^2$ \mbox{Mathematical Sciences Research Institute, Berkeley, CA 94720} \\
38: $^3$ \mbox{Department of Physics, Ben-Gurion University, Beer-Sheva 84105, Israel}
39: }
40:
41:
42: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
43:
44:
45: \begin{abstract}
46: For chaotic systems there is a theory for the decay of
47: the survival probability, and for the parametric dependence of
48: the local density of states. This theory leads to the distinction
49: between ``perturbative" and ``non-perturbative" regimes,
50: and to the observation that semiclassical tools are useful
51: in the latter case. We discuss what is ``left" from this
52: theory in the case of one-dimensional systems.
53: We demonstrate that the remarkably accurate {\em uniform} semiclassical
54: approximation captures the physics of {\em all} the different regimes,
55: though it cannot take into account the effect of strong localization.
56: \end{abstract}
57:
58:
59:
60: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
61: \section{\label{introduction}Introduction}
62:
63:
64: The quantum mechanical state of a particle, or of a system,
65: is represented by the probability matrix $\rho$.
66: This object corresponds to a classical distribution $\rho^{cl}(Q,P)$
67: in phase space, where $Q$ and $P$ are the canonical coordinates
68: of the system. One way to represent the probability matrix is
69: by using the Wigner function $\rho(Q,P)$.
70: Many calculations in quantum mechanics reduce eventually
71: to calculation of a trace over a pair of probability matrices.
72: This includes in particular calculations of the local density of states (LDOS) \cite{wls},
73: and calculations of the survival probability ${\cal P}(t)$ \cite{heller}.
74: Strongly related are calculations of Franck-Condon factors for
75: non-adiabatic transitions between Born-Oppenheimer surfaces \cite{bilha}.
76: %
77: Lately \cite{fidelity,jacq,fdl,fdl2} there is a special interest in calculation of ``fidelity"
78: (also known as ``Loschmidt echo") in the context of quantum computation.
79: This ``fidelity" is in fact a synonym for a ``survival probability" ${\cal P}(t)$
80: which is calculated for a particular dynamical scenario (namely, the forward evolution
81: is followed by a reversed evolution with a perturbed Hamiltonian).
82:
83:
84: The calculation of a trace over a pair of probability matrices
85: has a clear classical limit. To be specific let us consider
86: the survival probability, which is defined as
87: %
88: \begin{eqnarray} \label{e1}
89: {\cal P}(t) = | \langle \psi_0 | \psi_t \rangle |^2
90: = \trc (\rho_t \rho_0).
91: \end{eqnarray}
92: %
93: where $\rho_t$ and $\psi_t$ are the probability matrix
94: and the associated wavefunction of an evolving
95: quantum mechanical pure state.
96: In the Wigner representation the trace operation
97: means $dQdP/(2\pi\hbar)^d$ integral over phase space,
98: where $d$ is the number of freedoms.
99: It should be emphasized that the Wigner representation
100: is quantum-mechanically exact. The classical limit/approximation
101: is obtained by treating $\rho(P,Q)$ as a classical distribution,
102: whose evolution is governed by classical equations of motion.
103:
104:
105:
106: To take the classical limit as a leading order
107: approximation for ${\cal P}(t)$ is one possibility.
108: Another possibility is to use perturbation theory,
109: in particular ``Fermi golden rule" (FGR).
110: It should be obvious that the results that are
111: obtained by using these methods are typically
112: very different from each other.
113:
114:
115:
116: A major objective of recent studies \cite{ovrv}
117: is to understand the limitations of perturbation theory on the one hand,
118: and to explore the capabilities of the semiclassical tools, on the other.
119: A specific question is whether the decay of ${\cal P}(t)$ is
120: of perturbative nature (e.g. FGR type
121: \footnote{
122: In section~5 we explain that perturbation theory can
123: lead to either no decay, or Gaussian decay, or Wigner type decay.
124: The latter can be regarded as the outcome of FGR transitions.
125: }),
126: or else whether it is of classical nature ("semiclassical" type).
127: %
128: The main approach towards this question is to allow
129: the specification of a control parameter that represents
130: the ``strength" of a perturbation.
131: Depending on the value of of this control parameter
132: one may have either perturbative or semiclassical behavior.
133: %
134: The first publications \cite{crs,wls} that have taken this approach,
135: led to the distinction between ``perturbative" and ``non-perturbative"
136: regimes, and to the realization that the ``semiclassical" behavior is
137: contained in the"non-perturbative" regime.
138: Later \cite{jacq} the idea was adopted into the context of ``fidelity" studies.
139:
140:
141:
142: The above mentioned studies have mainly concentrated on
143: quantized {\em chaotic} systems. The chaos assumption allows
144: simplification of certain calculations. In particular one
145: can invoke the random matrix theory (RMT) conjecture,
146: in order to obtain some ``generic" results.
147: The natural question that arises is whether some
148: of the general theory regarding the LDOS and ${\cal P}(t)$,
149: applies also to the world of one dimensional ($d=1$) systems.
150:
151:
152:
153: A central observation in the theory of quantized chaotic
154: systems is the existence of two distinct energy scales.
155: One is the mean level spacing ($\Delta\propto\hbar^d$),
156: while the other is the bandwidth ($\Delta_b \propto \hbar$).
157: The latter is semiclassically related to the correlation time
158: of the classical motion. The dimensionless bandwidth
159: is defined as $b=\Delta_b/\Delta$. The classical limit
160: ($\hbar\rightarrow0$) of quantized chaotic system ($d>1$)
161: is characterized by $\Delta\ll\Delta_b$ and hence $b\gg1$.
162: But in one dimensional systems ($d=1$) we do
163: not necessarily have this separation of energy scales.
164: For some typical perturbations $b={\cal O}(1)$.
165: We shall explain that in such case there is no FGR regime
166: in the theory. A necessary, but not sufficient condition
167: for having FGR regime in one-dimensional systems is to have $b\gg 1$.
168: This means that ``small features" should characterize
169: the phase space manifolds that support the perturbed
170: (or evolving) quantum mechanical states.
171:
172:
173: The analysis of one dimensional ($d=1$) systems is highly
174: non-universal, but typically allows analytical calculations
175: that go well beyond the leading semiclassical approximation.
176: In particular we demonstrate the capabilities and the limitations
177: of the uniform approximation \cite{jiri,jiri2}.
178:
179:
180: The outline of this paper is as follows:
181: In Section~\ref{models} we define some simple prototype models.
182: In Section~\ref{definitions} we define the main objects of the
183: studies which are the LDOS and the survival probability~${\cal P}(t)$.
184: In Section~\ref{qcc} we discuss the issue of quantum-classical
185: correspondence (QCC) and make the distinction
186: between ``semiclassical" and ``perturbative" approximations.
187: In Sections~\ref{regimes1} and~\ref{regimes2}
188: we discuss the notion of ``regimes"
189: in the context of LDOS studies. In Sections~\ref{large_bandwidth}
190: and~\ref{uniform_regimes} we discuss the existence of LDOS regimes
191: in 1D within the framework of a uniform approximation.
192: In Section~\ref{strong_localization} we discuss the implication of the strong
193: localization effect. In Section~\ref{survival} we extend the general
194: consideration into the context of survival probability studies.
195: Conclusion and final remarks are summarized in Section~\ref{conclusions}.
196: The appendices are an integral part of the Paper.
197: They contain details of derivations and have not been
198: integrated into the main text in order to simplify the
199: presentation of the physical picture.
200:
201:
202: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
203: \section{\label{models}Definition of the models}
204:
205: In this paper we are going to consider one-dimensional model systems.
206: The simplest is the deformable harmonic oscillator:
207: %
208: \begin{eqnarray} \label{e2}
209: {\cal H}(Q, P; x) = \half (x P)^2 + \half (Q / x)^2.
210: \end{eqnarray}
211: %
212: where $(Q,P)$ are the canonical coordinates,
213: and $x$ is a constant parameter. The energy surfaces
214: ${\cal H}(Q, P; x) =E$ are ellipses in phase space.
215: We shall assume that $E\gg1$.
216: Without loss of generality we shall regard $x=x_0=1$
217: as the ``unperturbed" value of the parameter $x$.
218: Later we shall assume that $\delta x \equiv (x-x_0)$
219: is small in a {\em classical} sense. Namely $\delta x \ll 1$.
220: Then it is possible to write the Hamiltonian as
221: ${\cal H}={\cal H}_0+\delta {\cal H}$, where
222: %
223: \begin{eqnarray} \label{e3}
224: {\cal H}_0 \ &= \ {\cal H}(Q,P;x_0{=}1) \ = \ \frac{1}{2} (P^2+Q^2),
225: \\ \label{e4}
226: \delta {\cal H} \ &= \
227: \delta x \left.\frac{\partial {\cal H}}{\partial x}\right|_{x=1} \ = \ \delta x \ (P^2-Q^2)
228: \end{eqnarray}
229: %
230: Note that the perturbation $\delta x$ is allowed to be
231: large in a quantum mechanical sense: many levels
232: can be ``mixed" by the perturbation $\delta {\cal H} $.
233:
234:
235:
236:
237: A more complicated case is to consider a particle in a ring
238: ($=$ one dimensional box with periodic boundary conditions).
239: The Hamiltonian is of the general form
240: %
241: \begin{eqnarray} \label{e5}
242: {\cal H}(Q, P; x) = \frac{1}{2{\mathsf m}} P^2 + x \ V(Q)
243: \end{eqnarray}
244: %
245: Without loss of generality we can take ${\mathsf m}=1$
246: as the mass of the particle, and $L=2\pi$
247: as the perimeter (length) of the ring.
248: The parameter $x$ controls the ``height" of the
249: potential landscape. Naturally we shall regard $x_0=0$
250: as the unperturbed value of $x$.
251: We shall consider the case of a single ``bump" where
252: %
253: \begin{eqnarray} \label{e6}
254: V(Q) = V_0\exp \left[-\left(Q-Q_0 \right)^2 / 2 \ell^2 \right]
255: \end{eqnarray}
256: %
257: It is implicitly assumed that $\ell \ll L$.
258: The Fourier components of this potential are
259: $|\tilde{V}(k)| = V_0\ell \exp\left[-(k\ell)^2 /2 \right]$.
260: Hence the non-vanishing ($|k\ell| < 1$) Fourier components
261: satisfy $|\tilde{V}(k)| \approx V_0\ell$.
262: If we have many bumps, then we have a ``disordered" potential
263: %
264: \begin{eqnarray} \label{e7}
265: V(Q) \ = \ \sum_{\alpha} (\pm\mbox{random}) V_0 \exp \left[-\left(Q-Q_{\alpha} \right)^2 / 2 \ell^2 \right]
266: \end{eqnarray}
267: %
268: Here we implicitly assume that the bumps are non-overlapping
269: and randomly distributed along the ring, such that the correlation
270: function $\langle V(Q+r)V(Q) \rangle$ is characterized by
271: the correlation length $\ell$. Consequently the non-vanishing
272: ($|k\ell| < 1$) Fourier components of the potential
273: satisfy $|\tilde{V}(k)| \approx V_0\ell \times \sqrt{L/\ell}$.
274: Note that the phases of the Fourier components
275: look ``random", unlike the case of a single bump.
276:
277: All models that we have introduced are
278: of the generic form ${\cal H}= {\cal H}_0 +\delta {\cal H}$.
279: The representation in the basis that is determined by the
280: unperturbed Hamiltonian is
281: %
282: \begin{eqnarray} \label{e8}
283: {\cal H} \mapsto \mbf{E} + \delta x \mbf{B}
284: \end{eqnarray}
285: %
286: where $\mbf{E}=\{E_n\}$ is a diagonal matrix
287: consisting of the eigenenergies of the unperturbed Hamiltonian.
288: With scaled parameters, such that $\hbar=1$,
289: the mean level spacing of the eigenenergies
290: is $\Delta=1$ for Hamiltonian (\ref{e2})
291: and $\Delta = \sqrt{2E}$ for Hamiltonian (\ref{e5}).
292: %
293: The matrix $\mbf{B}$ corresponds to the perturbation.
294: It is a banded matrix. The bandwidth is $b$. It is defined such that
295: $2b$ is the number of the levels which are coupled by the
296: perturbation. For the Hamiltonian (\ref{e2}) we have $b=1$,
297: because only neighboring levels of the de-symmetrized Hamiltonian ($|n-m|=2$)
298: are coupled by the perturbation $\delta {\cal H}$.
299: For the Hamiltonian (\ref{e5}) the bandwidth is $b=L/\ell$.
300: See Section~\ref{large_bandwidth} for details. Thus for the latter model we may have $b\gg1$.
301:
302:
303: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
304: \section{\label{definitions}Definitions of $P(n|m)$ and ${\cal P}(t)$ }
305:
306: Let $|n(x)\rangle$ and $E_n(x)$ be the eigenstates and the corresponding
307: eigenvalues of the Hamiltonian ${\cal H}(Q,P;x)$.
308: We define the ``parametric" kernel
309: %
310: \begin{eqnarray}
311: P(n|m) \ = \ | \langle n(x) | m(x_0) \rangle |^2 = \trc (\rho_n \rho_m).
312: \label{ldos}
313: \end{eqnarray}
314: %
315: Note that for $\delta x=0$ we have $P(n|m)=\delta_{nm}$.
316: Given a reference state $|m(x_0)\rangle$ this kernel
317: can be regarded as a probability distribution with respect to $n$.
318: The LDOS is just a scaled version of this kernel, namely
319: %
320: \begin{eqnarray} \label{e10}
321: P(\omega) \ = \ \sum_n P(n|m) \ 2 \pi \delta\Big( \omega-\left[E_n(x){-}E_m(x_0)\right] \Big)
322: \end{eqnarray}
323: %
324: One important measure that characterizes the LDOS
325: is its variance:
326: %
327: \begin{eqnarray} \label{e11}
328: \delta E^2 \ = \ \sum_n P(n|m) \ (E_n-E_m)^2
329: \end{eqnarray}
330: %
331: In the next section we shall explain that the variance has
332: a special role in the theory of the LDOS.
333:
334:
335:
336: The survival probability of the state $|m(x_0)\rangle$ for evolution
337: which is generated by the perturbed Hamiltonian ${\cal H}(Q,P;x)$
338: can be written as ${\cal P}(t)=|F(t)|^2$, where
339: %
340: \begin{eqnarray} \label{e12}
341: F(t) \ = \ \langle m(x_0) | \eexp{-it({\cal H}-E_m)} | m(x_0) \rangle
342: \end{eqnarray}
343: %
344: In the above definition we have taken $E=E_m$ as
345: the natural reference for the energy. With this definition we
346: see that $F(t)$ is just the Fourier transform of the LDOS.
347: We note that in more complicated scenarios
348: a simple Fourier transform relation between $F(t)$
349: and $P(\omega)$ does not exist \cite{fdl}.
350:
351:
352:
353: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
354: \section{\label{qcc}Quantal-classical correspondence (QCC)}
355:
356: The classical limit/approximation for the LDOS kernel $P(n|m)$ is obtained
357: by taking for the Wigner function $\rho_n(Q,P) \approx \rho^{cl}_n(Q,P)$,
358: where \mbox{$\rho^{cl}_n(Q,P) \propto \delta\left[{\cal H}(Q,P;x)-E_n(x)\right] $}
359: is the corresponding classical microcanonical distribution.
360: For the deformable harmonic oscillator one obtains (\ref{appA}):
361: %
362: \begin{eqnarray} \label{e13}
363: P^{cl}(n|m) \ \approx \ \frac{1}{\pi}
364: \frac{1}{\sqrt{ 4( \delta x / x )^2E^2 -(E_n-E_m)^2 }}
365: \end{eqnarray}
366: %
367: The dispersion (square root of the variance) is
368: %
369: \begin{eqnarray} \label{e14}
370: \delta E^{cl} \ = \ \sqrt{2} \left(\frac{\delta x}{x}\right) E
371: \end{eqnarray}
372:
373:
374:
375: What about the quantum mechanical LDOS?
376: The simplest limit that can be considered
377: is first order perturbation theory (FOPT).
378: One obtains (\ref{appB}):
379: %
380: \begin{eqnarray} \label{e15}
381: P^{\tbox{prt}}(n|m) \ \approx \ \delta_{nm} \ + \
382: \frac{1}{4} \left(\frac{\delta x}{x}E\right)^2 \ \delta_{|n-m|,2}
383: \ \ \ \ \ \ \ \ \ \mbox{[FOPT]}
384: \end{eqnarray}
385: %
386: This result is very different from
387: the classical result. Therefore we say
388: that there is no ``detailed QCC". On the other
389: hand one easily calculates the variance.
390: One obtains $\delta E = \delta E^{cl}$.
391: We call the latter equality ``restricted QCC".
392:
393: It is possible to write an exact analytical expression for
394: the quantum mechanical LDOS (see \ref{appA}).
395: But this expression is not very useful since its complexity makes
396: it virtually impossible to extract the simple limits in various regimes.
397: It is more illuminating
398: to obtain a uniform approximation (\ref{appC})
399: %
400: \begin{eqnarray} \label{e16}
401: P^{\rm uniform}(n|m) \ \approx \ \left[ J_{(m-n)/2} \left( \frac{\delta x}{x}E \right) \right]^2
402: \end{eqnarray}
403: %
404: In figure~\ref{fig_ldos} we demonstrate that the uniform approximation
405: is almost indistinguishable from the exact result for any value of~$\delta x$.
406: One can verify that this approximation
407: reduces to the FOPT result (\ref{e15})
408: in the parametric regime $\delta x \ll (x/E)$.
409: Disregarding an oscillatory component it
410: reduces to the semiclassical result (\ref{e13})
411: in the parametric regime $\delta x \gg (x/E)$.
412: The two regimes are separated by the quantum mechanical
413: parametric scale $\delta x_{\tbox{prt}} = (\hbar\omega_{osc}/E) x $.
414: For the convenience of the reader we have
415: reverted here to non-scaled units (with our scaling
416: $\omega_{osc}=1$ and $\hbar=1$.)
417:
418:
419:
420: \begin{figure}
421: \centerline{\epsfig{figure=pert.eps,width=0.7\hsize}} \ \\
422: \centerline{\epsfig{figure=inter.eps,width=0.7\hsize}} \ \\
423: \centerline{\epsfig{figure=semicl.eps,width=0.7\hsize}} \ \\
424:
425: \caption{\label{fig_ldos}
426: The local density of states for E=100.
427: (a) For $(\delta x / x) E= 0.5$ (perturbative regime).
428: (b) For $(\delta x / x) E= 2$ (intermediate regime).
429: (c) For $(\delta x / x) E= 20$ (semiclassical regime).
430: In each panel the thick dashed line is the exact result,
431: the solid line is the uniform result,
432: the dashed-dotted line is the classical result,
433: and the solid circles are the perturbative result.
434: The classical result is divided by 2 for a reason
435: which is explained at the end of \ref{appA}.
436: }
437: \end{figure}
438:
439:
440:
441:
442: The example above demonstrates some general
443: observations regarding the LDOS. On the one
444: hand we have restricted QCC, which means
445: $\delta E = \delta E^{cl}$. It can be proved that
446: this type of QCC holds in general \cite{lds}.
447: The proportionality $\delta E \propto \delta x$
448: is guaranteed by the classical linearization condition,
449: which is a fixed assumption in this paper
450: (in the present example it means $\delta x \ll x$).
451: On the other hand we do not have in general detailed QCC,
452: which means that the approximation $P(n|m) \approx P^{cl}(n|m)$
453: holds only in a specific parametric regime.
454: In the above example we have only two parametric regimes: \\
455: %
456: \begin{minipage}{\hsize}
457: \vspace*{0.1cm}
458: \begin{itemize}
459: \setlength{\itemsep}{0cm}
460: \item
461: The ``perturbative regime" ($\delta x \ll \delta x_{\tbox{prt}}$) \\
462: in which (in this example) FOPT can be used.
463: \item
464: The ``non-perturbative regime" ($\delta x > \delta x_{\tbox{prt}}$) \\
465: in which (in this example) the semiclassical approximation can be used.
466: \end{itemize}
467: \vspace*{0.0cm}
468: \end{minipage}
469: %
470: Hence upon quantization there is only one parametric scale
471: in the theory \mbox{($\delta x_{\tbox{prt}} \propto \hbar$)}.
472: This parametric scale marks a border between the
473: perturbative and the non-perturbative regimes.
474:
475:
476:
477: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
478: \section{\label{regimes1}Regimes in case of chaotic systems}
479:
480: In case of chaotic systems the generic case is having
481: three regimes. The ``perturbative regime" is subdivided
482: into a ``FOPT regime" and
483: a Wigner/Lorentzian/core-tail/FGR regime.
484: In the latter case the various names mean the same,
485: and we shall use from now on the term ``Wigner regime".
486: The FOPT regime is defined as
487: the parametric regime where we can use first order
488: perturbation theory (FOPT):
489: %
490: \begin{eqnarray} \label{e17}
491: P^{\tbox{prt}}(n|m) \ \approx \ \delta_{nm} \ + \
492: \delta x^2 \frac{\ |\mbf{B}_{nm}|^2}{(E_n{-}E_m)^2}
493: \ \ \ \ \ \ \ \ \ \ \mbox{[FOPT]}
494: \end{eqnarray}
495: %
496: The condition for the validity of this approximation
497: is $\delta x \ll \delta x_c$, where
498: $\delta x_c = \Delta/\sigma$.
499: Here $\Delta$ is the mean level spacing,
500: and $\sigma$ is defined as the RMS value
501: of the ``in-band" off-diagonal
502: elements of the $\mbf{B}$ matrix.
503:
504:
505:
506:
507: Outside of the ``FOPT regime" we can still use
508: first order perturbation theory for the {\em tails} of the LDOS \cite{vrn}.
509: This leads to a generalized Wigner's Lorentzian
510: (a more meaningful name is ``core-tail structure"):
511: %
512: \begin{eqnarray} \label{e18}
513: P^{\tbox{prt}}(n|m) \ \approx \ \delta x^2
514: \frac{|\mbf{B}_{nm}|^2}{\left[\Gamma(\delta x)/2\right]^2 + (E_n{-}E_m)^2}
515: \ \ \ \ \ \ \ \ \ \ \mbox{[Wigner]}
516: \end{eqnarray}
517: %
518: The width parameter $\Gamma(\delta x)$ is determined
519: so as to have proper normalization \mbox{($\sum_n P(n|m)=1$)}.
520: For strongly chaotic systems,
521: for which the band profile is quite flat,
522: it follows that the ``core" width is
523: $\Gamma\approx (\delta x/\delta x_c)^2\Delta$.
524: %
525: The ``core-tail" approximation makes sense
526: as long as we have separation of energy scales
527: $\Gamma(\delta x) \ll \Delta_b$,
528: where $\Delta_b=b\Delta$ is the bandwidth in units of energy.
529: This translates into the condition
530: $\delta x \ll \delta x_{\tbox{prt}}$ where
531: %
532: \begin{eqnarray} \label{e18b}
533: \delta x_{\tbox{prt}} \ \ = \ \ \sqrt{b} \ \delta x_c.
534: \end{eqnarray}
535:
536:
537:
538: The ``core-tail" approximation (\ref{e18})
539: can be regarded as the outcome of
540: perturbative summation of diagrams
541: to infinite order: The FOPT diagrams are
542: re-iterated, while the interference
543: terms are neglected. In time domain
544: analysis this corresponds to a {\em Markovian approximation}
545: for the survival probability ${\cal P}(t)$,
546: leading to an exponential decay.
547: %
548: Hence we say that there is a {\em perturbative regime}
549: that includes the FOPT sub-regime, and the Wigner sub-regime.
550: In the Wigner sub-regime we need {\em all orders}
551: of perturbation theory leading to FGR transitions and Wigner decay.
552: %
553: On the other hand in the FOPT sub-regime there is {\em no decay}.
554: This can be easily deduced by Fourier transforming
555: the LDOS which is associated with Eq.(\ref{e17}):
556: Up to negligible first order correction, the survival amplitude $F(t)$
557: comes out as a pure phase factor, whose absolute value squared is ${\cal P}(t)\approx1$.
558: %
559: It is appropriate at this point to make a connection with
560: recent fidelity studies \cite{fidelity}. There, it is customary to consider
561: the dynamics of a wavepacket that is a superposition of many eigenstates.
562: Consequently, in fidelity studies, there is an effective
563: averaging over the survival amplitude $F(t)$,
564: leading (in the FOPT regime) to a slow Gaussian decay which is trivially related
565: to the statistics of the first order correction to the eigenenergies.
566:
567:
568: We can summarize this section by stating that in case
569: of a generic quantized chaotic system we have three
570: regimes as follows: \\
571: %
572: \begin{minipage}{\hsize}
573: \vspace*{0.1cm}
574: \begin{itemize}
575: \setlength{\itemsep}{0cm}
576: \item
577: The FOPT regime ($\delta x \ll \delta x_c$) \\
578: where Eq.(\ref{e17}) can be trusted.
579: \item
580: The Wigner (perturbative) regime ($\delta x_c \ll \delta x \ll \delta x_{\tbox{prt}}$), \\
581: where Eq.(\ref{e18}) can be trusted.
582: \item
583: The non-perturbative regime ($\delta x > \delta x_{\tbox{prt}}$), \\
584: in which (typically) the semiclassical approximation can be used.
585: \end{itemize}
586: \vspace*{0.0cm}
587: \end{minipage}
588: %
589: Hence upon quantization there are two parametric scales
590: in the theory, which are $\delta x_c$
591: and \mbox{$\delta x_{\tbox{prt}} \propto \hbar$}.
592: These parametric scale mark the borders between the regimes.
593:
594:
595:
596:
597: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
598: \section{\label{regimes2}Regimes in case of 1D systems}
599:
600: By Weyl law we know that the mean level spacing
601: for $d$-dimensional system is $\Delta\propto\hbar^d$.
602: On the other hand the bandwidth $\Delta_b\propto\hbar$
603: is inversely proportional to the period, or to the
604: correlation time of the dynamics.
605: For chaotic systems (with $d>1$)
606: we generally have $b = \Delta_b/\Delta \gg1$.
607: But for $d=1$ systems we can have $b={\cal O}(1)$,
608: as in the example of equation~(\ref{e2}).
609: Therefore in the latter case $\delta x_{\tbox{prt}} \sim \delta x_c$,
610: and we do not have a ``Wigner regime".
611: This means that the LDOS cannot have a Lorentzian-like structure,
612: and consequently the survival probability ${\cal P}(t)=|F(t)|^2$
613: cannot have an exponential decay.
614: While this latter statement strictly holds for the
615: specific scenario which has been defined by equation~(\ref{e12}),
616: it typically holds also in more complicated circumstances (``fidelity" studies).
617: Thus we conclude that the FGR picture cannot be applicable
618: to the analysis of the dynamics unless $b\gg1$.
619:
620:
621: At this stage of the discussion we can say that in order
622: to have three regimes (FOPT, Wigner, semiclassical) in the
623: theory of one-dimensional systems, we have to consider
624: models where the perturbation is characterized by a large bandwidth ($b\gg1$).
625: In the next sections we would like to further discuss the following: \\
626: %
627: \begin{minipage}{\hsize}
628: \vspace*{0.2cm}
629: \begin{itemize}
630: %\setlength{\itemsep}{0cm}
631: \item
632: A uniform approximation can be applied in
633: order to address both the perturbative
634: and the non-perturbative regimes.
635: \item
636: We can have $b\gg1$ by considering
637: potentials that have small (or sharp) features.
638: (For example ``bump" or ``disorder").
639: \item
640: It is possible to get a Lorentzian-like LDOS
641: from the uniform approximation,
642: but $b\gg1$ is not a sufficient condition.
643: \item
644: The uniform approximation does not take
645: into account the effect of strong localization.
646: \item
647: Three LDOS regimes can be observed in case of disordered potential
648: in spite of the strong localization effect.
649: \end{itemize}
650: \vspace*{0.0cm}
651: \end{minipage}
652:
653:
654:
655:
656: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
657: \section{\label{large_bandwidth}Perturbations with large bandwidth}
658:
659: For any one-dimensional system we can find the action-angle variables,
660: in which the problem becomes essentially equivalent to the
661: problem (\ref{e5}) of a particle in a ring.
662: Consider for example the deformable harmonic oscillator (\ref{e2}):
663: We already saw that all the essential physics of the perturbed
664: eigenstates can be obtained via a uniform approximation
665: which is based on action-angle variables description of the system.
666: %
667: The Hamiltonian (\ref{eC3}) of a deformable harmonic oscillator
668: in action-angle variables,
669: and the Hamiltonian (\ref{e5}) of a particle in a ring,
670: are similar as far as the semiclassical
671: treatment is concerned. Having a quadratic rather
672: than a linear dispersion relation is not an essential difference.
673:
674:
675: More generally, we may encounter circumstances where
676: the perturbation creates small phase space structures.
677: This is the new ingredient (compared with the deformable
678: harmonic oscillator) that we are going to consider in
679: the following sections. The problem (\ref{e5}) of a particle in a ring
680: constitutes a prototype example of having such small phase
681: space structures. The perturbation $V(Q)$ is characterized
682: by a different, much smaller scale ($\ell$), compared with
683: the size of the system ($L$). This means that we have
684: large bandwidth ($b\gg1$) rather than $b={\cal O}(1)$.
685:
686:
687: In the absence of a perturbation ($x=x_0=0$),
688: the eigenenergies of a particle in a ring are
689: $E_n=p_n^2/ 2 {\mathsf m}$,
690: where $n$ is an integer index, and $p_n = (2\pi\hbar/L) n$.
691: The perturbation matrix is related to the
692: Fourier components of the potential:
693: %
694: \begin{eqnarray} \label{e19_a}
695: \mbf{B}_{nm} \ = \ \frac{1}{L} \tilde{V}(p_n-p_m)
696: \end{eqnarray}
697: %
698: The expressions for $\tilde{V}(k)$ in case of either
699: the ``bump" (\ref{e6})
700: or the ``disorder" (\ref{e7}),
701: were given in Section~\ref{models}.
702: Substitution in (\ref{e19_a})
703: allows to determine the bandwidth:
704: %
705: \begin{eqnarray} \label{e19}
706: \Delta \ &=& \ 2\pi\hbar v_E/L
707: \\ \label{e20}
708: \Delta_b \ &=& \ 2\pi\hbar v_E/\ell
709: \\ \label{e21}
710: b \ &=& \ L/\ell
711: \end{eqnarray}
712: %
713: where $v_E=\sqrt{2 E / {\mathsf m}}$.
714: One difference between the two models is related
715: to the coupling parameter $\sigma$, which has been
716: defined in the beginning of Section~\ref{regimes1}
717: as the RMS value of the in-band off-diagonal elements:
718: %
719: \begin{eqnarray} \label{e22}
720: \sigma \ =& \ (\ell/L) V_0 & \ \ \ \ \mbox{for the bump}
721: \\ \label{e23}
722: \sigma \ =& \ (\ell/L)^{1/2} V_0 & \ \ \ \ \mbox{for the disorder}
723: \end{eqnarray}
724: %
725: This leads to
726: %
727: \begin{eqnarray} \label{e24}
728: \delta x_c^{\rm bump} \ &= \ \Delta/\sigma \ = \ (\Delta_b/V_0)
729: \\ \label{e25}
730: \delta x_c^{\rm disorder} \ &= \ \Delta/\sigma \ = \ (\Delta_b/V_0) / \sqrt{b}
731: \end{eqnarray}
732: %
733: However, the difference in $\sigma$, and hence in $\delta x_c$
734: is not the significant difference between the ``bump"
735: potential and the ``disorder" potential. The significant
736: difference is related to the {\em statistical properties} of the
737: $\mbf{B}$ matrix. In the case of ``disorder" the
738: matrix elements look ``random", which is not
739: the case for a single bump.
740:
741:
742: In both cases, of having either single bump or disorder,
743: the FOPT regime is $\delta x \ll \delta x_c$.
744: What do we have beyond FOPT?
745: More specifically, the question is whether we have,
746: as in the theory of chaotic systems, a distinct parametric scale
747: $\delta x_{prt} = \sqrt{b} \delta x_c$ that distinguishes
748: between a ``Wigner regime" and a ``semiclassical regime".
749: In the next section we shall try to address this question
750: within the framework of the uniform approximation.
751:
752:
753:
754: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
755: \section{\label{uniform_regimes}Regimes within the framework of the uniform approximation}
756:
757: There is a general semiclassical procedure that associates
758: wavefunctions of integrable systems with phase space manifolds.
759: The traditional implementation of this procedure, in case of
760: one dimensional systems, is known as ``the WKB method".
761: The WKB method is problematic near turning points.
762: This problem can be solved by ``uniformization" of the solution.
763: The simplest point of view regarding this ``uniformization"
764: is obtained by using action-angle variables. This leads
765: to a Hamiltonian that looks like that of a particle in a ring.
766: In general the dispersion relation can be different
767: (not quadratic as in (\ref{e5})), but we shall see that
768: this is not an important difference for the issues under study.
769:
770: The eigenstates of Hamiltonian (\ref{e5}) are supported by the
771: manifolds $H(Q,P;x)=E_n$. An alternate (explicit) way
772: to describe a given manifold is $P=p_n(Q;x)$, where
773: %
774: \begin{eqnarray} \label{e26}
775: p_n(Q; x) \ = \
776: \sqrt{2{\mathsf m}\left[ E_n{-}\delta xV(Q) \right]}
777: \ \approx \
778: p_n - \ \delta x \ \hbar \ k(Q)
779: \end{eqnarray}
780: %
781: and $k(Q)=V(Q)/\hbar v_E$. In the following we
782: use units such that $\hbar=1$.
783: The semiclassical formula for the corresponding eigenfunction is
784: %
785: \begin{eqnarray} \label{e27}
786: \langle Q|n(x)\rangle \ = \ \frac{1}{\sqrt{L}}
787: \exp\left(i \int_0^Q p_n(Q';x)dQ' \right)
788: \end{eqnarray}
789: %
790: This can be regarded as a special case of (\ref{eC5}).
791: Above we approximate the classical pre-exponential prefactor
792: by ${1}/{\sqrt{L}}$. This is legitimate because a fixed assumption of
793: this Paper is that we are considering high lying eigenstates,
794: and assume {\em classically small} perturbations
795: %
796: \footnote{
797: It can also be regarded as an example of a more general semiclassical
798: perturbation approximation \cite{miller2,fdl2}.}.
799: %
800: We already saw, in the context of the deformable harmonic oscillator
801: (see the end of \ref{appC}), that the numerical error which is associated
802: with this approximation is insignificant. The essential physics of
803: having various ``regimes" is not related to this approximation.
804:
805:
806: The overlap of the semiclassical wavefunctions is given
807: by the integral
808: %
809: \begin{eqnarray} \label{e28}
810: \hspace*{-2cm}
811: \langle n(x) | m(x_0) \rangle = \frac{1}{L} \int_0^L dQ
812: \ \exp\left[- i\left( (p_n - p_m)Q - \delta x \int_0^Q k(Q')dQ' \right)\right]
813: \end{eqnarray}
814: %
815: In the following subsections we discuss the consequences
816: of this expression in case of either bump or disorder.
817:
818:
819: %%%%%%%%%%%%%%%%%%%%%%%%%%
820: \subsection{bump case}
821:
822: In case of the bump, $k(Q)$ has an amplitude
823: $k_0=V_0/\hbar v_E$ over a spatial scale $\ell$.
824: The total phase variation in (\ref{e28}) is
825: %
826: \begin{eqnarray} \label{e29}
827: \delta\phi_{\rm bump} \ = \ \delta x \times k_0 \ell
828: \end{eqnarray}
829: %
830: This phase variation is in fact the
831: phase space area of the bump.
832: So we have two possibilities:
833: Either $\delta\phi_{\rm bump} \ll 2\pi$
834: or $\delta\phi_{\rm bump} \gg 2\pi$.
835: This can be shown to be equivalent
836: to either $\delta x \ll \delta x_c^{\rm bump}$ or
837: $\delta x \gg \delta x_c^{\rm bump}$ respectively.
838: In the former case it is easy to recover
839: the FOPT result (\ref{e17}).
840: One should simply put the perturbation
841: off the exponent, and then do integration
842: by parts. For $n\ne m$ it leads to
843: %
844: \begin{eqnarray} \label{e30}
845: \hspace*{-1cm}
846: P(n|m) = \left|
847: \frac{1}{L}\int_0^L dQ
848: \frac{\eexp{- i (p_n - p_m)Q} }{p_n-p_m}
849: \ \delta x k(Q)
850: \right|^2 \ = \
851: \delta x^2 \left|\frac{\mbf{B}_{nm}}{E_n-E_m} \right|^2
852: \end{eqnarray}
853: %
854: The other possibility ($\delta\phi_{\rm bump} \gg 2\pi$)
855: guarantees the validity of the standard
856: semiclassical approximation. In such case
857: we cannot put the perturbation off the exponent,
858: but instead we can make a stationary phase
859: approximation. We shall not dwell further
860: on details because it is a standard textbook procedure.
861:
862:
863: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
864: \subsection{disorder case}
865:
866: We see that for a simple bump we have either
867: FOPT or semiclassical approximation.
868: So we have just two regimes, in spite of the fact
869: that $b\gg 1$.
870: This means that large bandwidth is not a sufficient
871: condition for having three parametric regimes.
872: So let us try to make things more complicated by
873: considering a disordered potential with many bumps.
874: Equation~(\ref{e29}) still describes the phase variation over a single bump.
875: This means that $\delta\phi_{\rm bump} \gg 2\pi$ is still
876: the relevant condition for a semiclassical approximation!
877: What about first order perturbation theory?
878: The total phase variation in (\ref{e28}) for many bumps is
879: %
880: \begin{eqnarray} \label{e31}
881: \delta\phi_{\rm disorder} = \delta x \times \sqrt{b} \times k_0 \ell
882: \end{eqnarray}
883: %
884: (note that $b$ is essentially the number of bumps involved).
885: Consequently the validity condition for first order perturbation theory
886: is $\delta\phi_{\rm disorder} \ll 2\pi$, which can be easily converted
887: into $\delta x \ll \delta x_c^{\rm disorder}$.
888:
889:
890: The considerations of the previous paragraph imply
891: that for {\em disordered} potential we have three regimes.
892: In the intermediate regime
893: $\delta x_c \ll \delta x \ll \delta x_{\tbox{prt}}$,
894: we have $\delta \phi_c^{\rm disorder} \gg 2\pi$
895: while $\delta \phi_c^{\rm bump} \ll 2\pi$.
896: Consequently we can use neither FOPT, nor the semiclassical approximation.
897: This is the Wigner regime where we expect to find a Lorentzian-like LDOS.
898: Let us demonstrate that indeed a Lorentzian-like LDOS
899: can be obtained from the uniform approximation (\ref{e28}).
900: For this purpose we average $P(n|m)$ over realizations of the disorder:
901: %
902: \begin{eqnarray}
903: \hspace*{-2.5cm}
904: P(n|m) = \frac{1}{L^2}\int_0^L\int_0^L dQ_1dQ_2
905: \ \eexp{-i(p_n-p_m)(Q_2-Q_1)}
906: \left\langle \exp\left[i\delta x\int_{Q_1}^{Q_2}k(Q')dQ'\right] \right\rangle_{\rm disorder}
907: \nonumber \\ \hspace*{-1.3cm}
908: %
909: = \frac{1}{L^2}\int_0^L\int_0^L dQ_1dQ_2
910: \ \eexp{-i(p_n-p_m)(Q_2-Q_1)}
911: \exp\left[-\frac{1}{2} \left(\frac{\delta x }{\hbar v_E}\right)^2
912: \int_{Q_1}^{Q_2} \int_{Q_1}^{Q_2}\langle V(Q')V(Q'')\rangle dQ'dQ'' \right]
913: %
914: \nonumber \\ \hspace*{-1.3cm}
915: \approx \frac{1}{L}\int_{-\infty}^{\infty} dr \ \eexp{-i(p_n-p_m)r}
916: \exp\left[-\delta x^2 \left(\frac{V_0}{\hbar v_E}\right)^2 \ell |r| \right]
917: %
918: \label{e32}
919: \end{eqnarray}
920: %
921: The last integral is the Fourier transform of an exponential,
922: leading to the Lorentzian LDOS as defined in (\ref{e18}).
923:
924:
925:
926: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
927: \section{\label{strong_localization}Strong localization effect}
928:
929: Still we have to address the question whether we can trust
930: the uniform approximation, which is based on the WKB wavefunction (\ref{e27}).
931: The answer is known to be negative in case of {\em disordered}
932: potential. The WKB approximation does not take into account
933: backscattering, which is responsible for the strong localization
934: effect in 1D disordered potential. It is well known \cite{richter} that
935: the localization length is equal (in 1D) to twice the mean
936: free path. Up to a numerical prefactor the Born approximation estimate is
937: %
938: \begin{eqnarray} \label{e33}
939: \hspace*{-1cm}
940: \frac{1}{L_{\tbox{loc}}}
941: \ &\approx & \
942: \frac{1}{\hbar v_E} \frac{\sigma^2}{\Delta}
943: \times \delta x^2
944: \ = \ \left( \frac{\delta x}{\delta x_c}\right)^2 \frac{1}{L}
945: \ = \ \left( \frac{\delta x}{\delta x_{\tbox{prt}}}\right)^2 \frac{1}{\ell}
946: \end{eqnarray}
947: %
948: The condition for {\em not} being affected by the strong localization
949: effect is $L_{\tbox{loc}} \gg L$, which leads to $\delta x \ll \delta x_c$.
950: Thus it follows that only in the FOPT regime we can ignore
951: the strong localization effect. The strong localization effect cannot be
952: ignored neither in the Wigner regime nor in the semiclassical regime.
953:
954:
955: The above Born approximation assumes $L_{\tbox{loc}} \gg \ell$.
956: This condition breaks down if $\delta x > \delta x_{\tbox{prt}}$.
957: Recall that we also assume that $\delta x$ is small in the
958: classical sense ($\delta x V_0 \ll E$). The two inequalities are consistent
959: if and only if the de Broglie wavelength of the particle is much smaller
960: compared with $\ell$. In this regime we can analyze the localization
961: using the well known transfer matrix approach: Each bump has
962: some transfer matrix, and the random distance between the bumps
963: provides the phase randomization which is assumed in ``combining"
964: adjacent transfer matrices. Denoting the
965: average transmission of a ``bump" by $g$, one gets
966: %
967: \begin{eqnarray} \label{e33a}
968: \frac{1}{L_{\tbox{loc}}} \ \approx \ \ln(1/g) \times \frac{1}{\ell}
969: \end{eqnarray}
970: %
971: Thus even in the $\delta x > \delta x_{\tbox{prt}}$ regime
972: we can have a very long localization length ($L_{\tbox{loc}} \gg \ell$),
973: which is in fact consistent with the naive expectation.
974:
975:
976: The implications of the above discussion are, that in spite of
977: the strong localization effect, it is still meaningful to distinguish
978: between three parametric regimes (FOPT, Wigner, semiclassical).
979: We just have to remember that the wavefunction is not ergodic
980: in real space, so the role of $L$ is taken by $L_{\tbox{loc}}$.
981:
982:
983:
984: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
985: \section{\label{survival}The survival probability ${\cal P}(t)$ }
986:
987: We turn to discuss the calculation
988: of the survival probability (\ref{e1})
989: for the specific ``wavepacket dynamics" scenario
990: that has been defined in Section~\ref{definitions}.
991: We have the following five strategies of calculation: \\
992: %
993: \begin{minipage}{\hsize}
994: \vspace*{0.1cm}
995: \begin{itemize}
996: \setlength{\itemsep}{0cm}
997: \item
998: Uniform approximation (which is essentially exact)
999: \item
1000: Time domain classical approximation
1001: \item
1002: Energy domain classical approximation ($\leadsto$ LDOS $\leadsto$ Fourier transform)
1003: \item
1004: Time domain perturbation theory
1005: \item
1006: Energy domain perturbation theory ($\leadsto$ LDOS $\leadsto$ Fourier transform)
1007: \end{itemize}
1008: \vspace*{0.0cm}
1009: \end{minipage}
1010:
1011:
1012:
1013: It can be shown that in typical circumstances
1014: the two versions of perturbation theory give in leading order consistent results.
1015: This means that we can write a perturbative (essentially first order) result that can be trusted
1016: for sufficiently short times (for any perturbation~$\delta x$)
1017: or for sufficiently weak perturbation (for any time~$t$).
1018: %
1019: As an example we consider the deformable harmonic oscillator.
1020: Taking equation~(\ref{e15}) with appropriate second order
1021: compensation of normalization, we get after Fourier transform,
1022: %
1023: \begin{eqnarray} \label{e34}
1024: {\cal P}^{\tbox{prt}}(t) \ \approx \ 1 - 2\left(\frac{\delta x}{x}\right)^2 E^2 \sin^2 t
1025: \end{eqnarray}
1026: %
1027: In a strict time domain FOPT we get only $t^2$ time
1028: dependence, while in a strict energy domain FOPT
1029: we do not get the correct normalization (${\cal P}^{\tbox{prt}}(0) =1$).
1030: Still we are able to get one consistent result.
1031: %
1032: The situation is different with the classical
1033: approximation. Here time domain and energy domain
1034: calculations do not give the same result.
1035: As an example we consider again the deformable
1036: harmonic oscillator. The calculation of overlap
1037: between $\rho^{cl}_t$ and $\rho^{cl}_0$ is
1038: simpler in action-angle variables, but otherwise
1039: it is similar to the calculation in \ref{appA}, leading to
1040: %
1041: \begin{eqnarray} \label{e35}
1042: {\cal P}^{cl}(t) \ = \ \left|2\pi\frac{\delta x}{x}E\sin t \right|^{-1}
1043: \end{eqnarray}
1044: %
1045: The energy domain classical approximation is
1046: obtained by squaring the Fourier transform of
1047: equation~(\ref{e10}) using the classical approximation (\ref{e13}) of $P^{cl}(n|m)$, leading to
1048: %
1049: \begin{eqnarray} \label{e36}
1050: {\cal P}^{cl,E}(t) \ = \ \left|J_0\left(2\frac{\delta x}{x} Et\right)\right|^2
1051: \end{eqnarray}
1052: %
1053: Although there is no simple exact expression for ${\cal P}(t)$,
1054: there again exists a very simple uniform approximation which is remarkably accurate
1055: in all regimes and therefore we can regard it as ``exact,'' namely
1056: %
1057: \begin{eqnarray} \label{e37}
1058: {\cal P^{\rm uniform}}(t) \ = \ \left|J_0\left(2\frac{\delta x}{x} E \sin t \right)\right|^2
1059: \end{eqnarray}
1060: %
1061: It is derived in \ref{appD} by using semiclassical expressions for the initial and evolved states,
1062: while calculating the overlap exactly rather than by the stationary phase approximation.
1063:
1064:
1065: Results (\ref{e34})-(\ref{e37}) are graphically displayed in figure~\ref{fig_survival}.
1066: Approximations (\ref{e34})-(\ref{e36}) can be regarded as various limits of the uniform approximation (\ref{e37}).
1067: It is easily seen that (\ref{e37}) reduces to the perturbative result (\ref{e34})
1068: whenever $2 (\delta x / x ) E\sin(t) \ll 1$.
1069: So as expected the perturbative result can be trusted for either small time $t$
1070: or for small perturbation $\delta x / x$.
1071: %
1072: It is also easy to see that the uniform result (\ref{e37}) reduces
1073: to the energy domain classical result (\ref{e36}) for $t \ll 1$.
1074: %
1075: The relation of (\ref{e37}) to the time domain classical result (\ref{e35})
1076: is more subtle: For large perturbation the two expressions agree in an asymptotic sense,
1077: and either time smoothing or energy averaging is required in order
1078: to demonstrate this agreement (see figure~\ref{fig_survival}c).
1079:
1080:
1081:
1082:
1083:
1084: \begin{figure}
1085: \centerline{\epsfig{figure=sp_pert.eps,width=0.45\hsize}} \ \\
1086: \centerline{\epsfig{figure=sp_inter.eps,width=0.45\hsize}} \ \\
1087: \centerline{\epsfig{figure=sp_semicl.eps,width=0.45\hsize}} \ \\
1088: \centerline{\epsfig{figure=sp_semiclZ.eps,width=0.45\hsize}} \ \\
1089:
1090: \caption{\label{fig_survival}
1091: The survival probability, using the same parameters as in figure~\ref{fig_ldos}.
1092: (a) Perturbative regime.
1093: (b) Intermediate regime.
1094: (c) Semiclassical regime.
1095: In each panel the solid line is the uniform result (\ref{e37}),
1096: the dashed line is the perturbative result (\ref{e34}),
1097: the dashed-dotted line is the time domain classical result (\ref{e35}),
1098: and the dotted line is the energy domain classical result (\ref{e36}).
1099: Panel (d) gives a short-time detail of panel (c).
1100: }
1101:
1102: \end{figure}
1103:
1104:
1105:
1106:
1107: An important message of this section is that the discussion of regimes
1108: in the context of ${\cal P}(t)$ is in one to one correspondence
1109: with the discussion of LDOS regimes.
1110:
1111:
1112:
1113: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1114: \section{\label{conclusions}Conclusions and final remarks}
1115:
1116: In general, for more complicated ``wavepacket dynamics" scenarios,
1117: which may involve time dependent Hamiltonian,
1118: an explicit reduction of the survival probability problem
1119: to LDOS study is not possible \cite{fdl}.
1120: The simplest way to make the Hamiltonian ``time dependent"
1121: is by changing it either once (as in ``fidelity" or ``Loschmidt echo" studies)
1122: or repeatedly (as in kicked systems).
1123: The latter possibility opens the way for ``chaos"
1124: in the classical dynamics of one-dimensional (1D) systems.
1125: A prototype system for 1D chaos studies is the kicked rotator:
1126: This is a particle in a ring, which is periodically kicked by a cosine potential.
1127: As in the standard formulation of ``wavepacket dynamics" it is common
1128: to assume an initial preparation which is supported by a manifold $P=p_m$.
1129: As a result of the kicks the initial manifold becomes extremely convoluted.
1130: If we want to calculate ${\cal P}(t)$ we have to trace
1131: the evolving manifold with the initial one. The calculation
1132: of this overlap can be carried out using the uniform approximation
1133: that we have discussed previously~\cite{jiri,jiri2,fdl2}.
1134:
1135: There is one major difference that makes the ``chaotic" scenario
1136: that we have described above different from the LDOS calculation:
1137: The issue of strong localization, which has been
1138: discussed in Section~\ref{strong_localization} is no longer relevant.
1139: The evolving state is ``distributed" over the whole manifold
1140: %
1141: \footnote{
1142: The only reservation for this statement is the possibility
1143: of witnessing a ``dynamical localization effect" after an extremely
1144: long time (known as the ``breaktime").
1145: For small $\hbar$ the existence of a ``breaktime" will have
1146: a negligible effect on the behavior of ${\cal P}(t)$.
1147: In fact we should remember that also in non-kicked systems
1148: we have a ``breaktime", which is just the Heisenberg time.
1149: We also note that this type of localization is apparently
1150: captured by semiclassical methods \cite{kaplan}}.
1151:
1152: {\em In view of the above, the calculations that we have presented
1153: in this paper, using the uniform approximation, are in fact generic}.
1154: In particular we have explained how a Fermi-golden-rule behavior
1155: arises within this framework. A refinement
1156: of the uniform approximation using the
1157: ``replacement manifolds" approach has been introduced
1158: in references \cite{jiri,jiri2}.
1159:
1160: The reader can be tempted to reach the conclusion that
1161: the discussion of ``regimes" within the framework
1162: of the ``uniform approximation" can be trivially generalized
1163: from $d=1$ chaotic systems to $d>1$ chaotic systems.
1164: This is in fact not quite correct.
1165: In order to explain the difficulty let us consider
1166: again LDOS calculation. In one-dimensional systems
1167: what we have to do is to calculate the overlap of
1168: two one-dimensional manifolds. We can do this
1169: calculation semiclassically if we can trust the
1170: stationary phase approximation. This leads to the
1171: "$\hbar$ area" condition: the stationary phase approximation
1172: is accurate if the phase space area delimited by the two manifolds between
1173: two stationary phase points is larger than $\hbar$.
1174: What is the generalization of
1175: this condition in the $d>1$ case? Now the manifolds
1176: are ``surfaces" and they intersect along ``lines".
1177: So the concept of ``stationary points" becomes
1178: inapplicable, and also the ``$\hbar$ area" condition
1179: becomes meaningless. The way out of this difficulty \cite{vrn}
1180: is to use the Wigner function point of view. Then one realizes
1181: that the proper way to formulate the semiclassical condition
1182: is to say that the separation between the surfaces should be much
1183: larger compared with the ``thickness" of Wigner function.
1184: This ``thickness" is just the ``bandwidth" that we have
1185: discussed in this Paper.
1186:
1187: Conventionally, semiclassical methods are applied
1188: in the ``semiclassical" regime (large enough perturbation).
1189: In the perturbative regime people use perturbation theory.
1190: A major motivation for the present line of study is
1191: to extend the applicability of semiclassical methods into
1192: the perturbative regime. Most of the calculations that we have
1193: presented under the heading ``uniform approximation" can be
1194: reformulated using the Wigner function language.
1195: This opens the way towards a unified semiclassical
1196: understanding of ``regimes" in case of $d>1$ systems.
1197:
1198:
1199: \ack
1200:
1201: It is our pleasure to thank Bilha Segev (BGU) and Eric Heller (Harvard)
1202: for useful discussions. The research was supported by
1203: the Institute for Theoretical Atomic and Molecular Physics at Harvard University,
1204: by the Mathematical Sciences Research Institute at Berkeley, and
1205: by the Israel Science Foundation (grant No.11/02).
1206:
1207:
1208: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1209: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1210: \appendix
1211:
1212:
1213: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1214: \section{\label{appA}Calculation of $P^{cl}(n|m)$ and $P^{\rm exact}(n|m)$}
1215:
1216: We shall demonstrate in this appendix the calculation of $P^{cl}(n|m)$
1217: for the deformable harmonic oscillator:
1218: %
1219: \begin{eqnarray} \label{eA1}
1220: P^{cl}(n|m) &
1221: \ = \ \int\frac{dQdP}{2\pi} \delta\left[H(Q,P;x_2)-E_n\right]\delta\left[H(Q,P;x_1)-E_m\right]
1222: \nonumber \\ &
1223: \ = \ \int\frac{dQ}{2\pi} \sum_{P = \pm \sqrt{2 E_{m} - (Q / x_1)^2}/ x_1}
1224: \frac{\delta [ {\cal H}(Q, P; x) - E_n ]}{x_1^2 |P|}
1225: \nonumber \\ &
1226: \ = \ \frac{1}{\pi} \int dQ
1227: \frac{\delta [f(Q)]}{\sqrt{2 x_1^2E_m - Q^2}}
1228: \nonumber \\ &
1229: \ = \ \frac{1}{\pi} \sum_{\pm} \left ( 2 x_1^2 E_m - Q_{\pm}^{2} \right)^{-1/2} \left| f'(Q_{\pm}) \right|^{-1}
1230: \nonumber \\ &
1231: \ = \ \frac{1}{\pi} \frac{x_1 x_2}
1232: {\sqrt{\left(x_{2}^{2} E_n - x_{1}^{2} E_m \right) \left(x_{2}^{2} E_m - x_{1}^{2} E_n \right)}}
1233: \end{eqnarray}
1234: %
1235: where
1236: %
1237: \begin{eqnarray} \label{eA2}
1238: f(Q) \ \equiv \ \frac{x_{2}^{4} - x_{1}^{4}}{2 x_{1}^{4} x_{2}^{2}} Q^2 + \left
1239: ( \frac{x_2}{x_1} \right)^2 E_m - E_n.
1240: \end{eqnarray}
1241: %
1242: and $Q_{\pm}$ are the roots of the equation $f(Q)=0$,
1243: %
1244: \begin{eqnarray} \label{eA3}
1245: Q_{\pm} \ = \ \pm x_{1} x_{2} \sqrt{\frac{2 (x_{2}^{2} E_m - x_{1}^{2} E_n)}{x_{2}^{4} - x_{1}^{4}}}
1246: \end{eqnarray}
1247: %
1248: for which
1249: %
1250: \begin{eqnarray}
1251: f'(Q_{\pm}) \ = \ \pm x_1^{-3} x_2^{-1} \sqrt{2 (x_{2}^{2} E_m - x_{1}^{2} E_n) (x_{2}^{4} - x_{1}^{4})}
1252: \end{eqnarray}
1253: %
1254: Equation~(\ref{e13}) is a simplified version of this result, assuming that $\delta x \ll x$.
1255:
1256:
1257: It is also possible to obtain an exact result in the quantum mechanical case.
1258: The explicit expression for the eigenfunction using Hermite polynomials is
1259: %
1260: \begin{eqnarray}
1261: \bra Q | n(x) \ket \ = \ (\pi x^2)^{-1/4} (2^n n!)^{-1/2} H_n(Q/x)
1262: \eexp{-(Q/x)^2 / 2}
1263: \end{eqnarray}
1264: %
1265: This leads to
1266: %
1267: \begin{eqnarray} \nonumber
1268: & \langle n(x_2)|m(x_1) \rangle = (\pi x_1 x_2)^{-1/2}
1269: (2^{n+m}n!m!)^{-1/2} \nonumber
1270: \\ \label{eA4}
1271: & \ \ \times \int_{-\infty}^{\infty} dQ H_n(Q / x_2) H_m(Q / x_1)
1272: \eexp{ -\half (x_1^{-2} + x_2^{-2} ) Q^2}
1273: \end{eqnarray}
1274: %
1275: Upon squaring one obtains $P(n|m)$. The integral in (\ref{eA4}) becomes highly oscillatory
1276: for high-lying eigenstates in which we are interested, and numerical calculation is tricky.
1277: To our surprise, this intimidating integral can be evaluated analytically even for $x_1 \neq x_2$,
1278: resulting in a finite sum of terms, which is not very elegant, but very simple to evaluate on a computer.
1279: For the exact benchmark in the numerical results presented in this paper,
1280: we therefore used this analytical expression instead of numerically evaluating the integral (\ref{eA4}).
1281:
1282:
1283: Note that both classically and quantum mechanically
1284: the overlap $P(n|m)$ depends only on the ratio $x_2/x_1$.
1285: In the classical case $n$ is a real index ($E_n$ can have
1286: any real positive value). In the quantum mechanical
1287: case $n$ is an integer index ($E_n= n + \half$).
1288: Due to the reflection symmetry of the Hamiltonian
1289: there is an overlap only between states with the same
1290: parity. The overlap for $|n-m|=\mbox{odd}$ vanishes.
1291: Whenever we say that $P(n|m) \approx P^{cl}(n|m)$,
1292: it should be interpreted in a coarse grain sense.
1293: Therefore, for sake of graphical presentation we have
1294: plotted $P(n|m)$ versus $P^{cl}(n|m)/2$.
1295: One may say that the plotted $P(n|m)$ and $P^{cl}(n|m)/2$
1296: correspond to a de-symmetrized oscillator.
1297:
1298:
1299: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1300: \section{\label{appB}Calculation of $P^{\tbox{prt}}(n|m)$}
1301:
1302: We shall demonstrate in this appendix the calculation of $P^{prt}(n|m)$
1303: for the deformable harmonic oscillator. From first order perturbation theory
1304: we know that for $m \ne n$
1305: %
1306: \begin{eqnarray}
1307: \langle m(x_0) | n(x_0+\delta x) \rangle \approx
1308: \frac{ \langle m |\delta {\cal H} | n \rangle }{E_n-E_m }
1309: \end{eqnarray}
1310: %
1311: Using
1312: %
1313: \begin{eqnarray} \nonumber
1314: \hspace*{-2.5cm}
1315: & \bra m \vert Q^2 \vert n \ket ,\ \bra m \vert P^2 \vert n \ket
1316: \ = \ (\half + n) \delta_{mn}
1317: & \pm \frac{1}{2}
1318: \left[\sqrt{(n+1)(n+2)} \delta_{m, n+2} + \sqrt{n (n-1)} \delta_{m,n-2} \right]
1319: \end{eqnarray}
1320: %
1321: one obtains
1322: %
1323: \begin{eqnarray}
1324: P^{\tbox{prt}}(n|m) \ \approx \
1325: \delta_{n,m} + \frac{1}{4}\left(\frac{\delta x}{x}\right)^2 \left(E^2-\frac{1}{4}\right) \delta_{|n-m|,2}
1326: \end{eqnarray}
1327: %
1328: where $E=(E_n+E_m)/2$. In the text we have presented
1329: a simplified version that assumes $E \gg 1$.
1330:
1331:
1332:
1333: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1334: \section{\label{appC}Calculation of $P^{\rm uniform}(n|m)$}
1335:
1336: The following canonical transformation is used in order
1337: to transform the Hamiltonian of the harmonic oscillator
1338: to action-angle variables:
1339: %
1340: \begin{eqnarray}
1341: Q \ &=& \ \sqrt{2I} \cos(\phi) \\
1342: P \ &=& \ \sqrt{2I} \sin(\phi)
1343: \end{eqnarray}
1344: %
1345: In the vicinity of $x_0=1$ it leads to
1346: %
1347: \begin{eqnarray} \nonumber
1348: {\cal H}(\phi,I; x) \ &=& \
1349: I \left[(1/x)\cos^2\phi+x\sin^2\phi\right]
1350: \\ \label{eC3}
1351: \ & \approx & \
1352: I - \delta x I \cos(2\phi)) + {\cal O}(\delta x^2)
1353: \end{eqnarray}
1354: %
1355: The canonical transformation from $(\phi,I)$ to the
1356: action angle variables $(\phi',I')$ of the perturbed
1357: Hamiltonian is derived from a generating function $S(\phi,I')$.
1358: The manifold $I'=\const$ is determined from
1359: the equation ${\cal H}(\phi,I; x)=\const$,
1360: leading to the relation $I = \left[ 1+\delta x \cos(2\phi) \right] I'$.
1361: The generating function should satisfy
1362: $I={\partial S}/{\partial \phi}$. Therefore one deduces that
1363: %
1364: \begin{eqnarray}
1365: S(\phi,I') \ = \ \left[ \phi+\delta x \sin(2\phi) \right] I'
1366: \end{eqnarray}
1367: %
1368: The semiclassical expression for the wavefunction is
1369: %
1370: \begin{eqnarray} \label{eC5}
1371: \bra \phi \vert I' \ket_{\rm sc} \ = \ (2 \pi)^{-1/2} \sqrt{\frac{\partial^2 S}{\partial I' \partial \phi}} \ \eexp{iS(\phi,I')}.
1372: \end{eqnarray}
1373: %
1374: The $n$th semiclassical eigenstate corresponds
1375: to the substitution $I'=E_n=n + \half$.
1376: For technical simplicity it is convenient to
1377: calculate the overlap between two perturbed
1378: wavefunctions ($x=x_0 \pm \delta x/2$) .
1379: %
1380: \begin{eqnarray} \nonumber
1381: \hspace*{-2cm}
1382: \langle n(x_0 + \delta x / 2 ) | m(x_0 - \delta x / 2 ) \rangle
1383: \ = \ \int_0^{2\pi} d\phi \langle \phi | I''{=}E_n \rangle^{*} \langle \phi | I'{=}E_m \rangle
1384: \\ \nonumber
1385: = \frac{1}{2\pi} \int_0^{2\pi} d\phi \sqrt{1 - \delta x^2 \cos^2 2\phi}
1386: \times \exp \left[ i (E_m - E_n) \phi - i \delta x E \sin 2 \phi \right]
1387: \\
1388: \approx J_{(m-n)/2}(\delta x E)
1389: \end{eqnarray}
1390: %
1391: Above $I'$ and $I''$ correspond to the perturbations $\pm \delta x/2$,
1392: and $E=(E_n+E_m)/2$. In the main text we have reverted to
1393: a more general version of this expression that does not assume $x_0=1$.
1394: It is important to realize that because of the assumption $\delta x \ll x$
1395: the pre-exponential factor (which is in fact a ``classical" factor)
1396: can be neglected in leading order. Figure~\ref{fig_ldos} confirms the remarkable
1397: agreement of the uniform approximation with the exact result.
1398:
1399:
1400:
1401:
1402: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1403: \section{\label{appD}Calculation of ${\cal P}^{\rm uniform}(t)$}
1404:
1405: For the purpose of calculation it is more convenient
1406: to regard the prepared state as a ``perturbed state"
1407: and the evolution Hamiltonian as the ``unperturbed
1408: Hamiltonian". This is of course equivalent to the presentation
1409: in the text upon the replacement $\delta x\mapsto -\delta x$.
1410:
1411: The initial state $|m(x_0+\delta x)\rangle$
1412: is characterized by the action $I'=E_m\equiv E$.
1413: It is represented as in \ref{appC} by
1414: %
1415: \begin{eqnarray}
1416: \langle \phi | I'{=}E \rangle \approx \frac{1}{2\pi} \exp[i(\phi+\delta x\sin 2\phi) E]
1417: \end{eqnarray}
1418: %
1419: The evolving state is represented by
1420: %
1421: \begin{eqnarray}
1422: \langle \phi |\eexp{-it{\cal H}_0} | I'{=}E \rangle = \langle \phi-t | I'{=}E \rangle
1423: \end{eqnarray}
1424: %
1425: The overlap between the evolving and the initial state is
1426: %
1427: \begin{eqnarray} \nonumber
1428: \frac{1}{2\pi} \int_0^{2\pi} d\phi \exp \Big[ i \Big( -t + \delta x \left[ \sin 2 (\phi-t) - \sin 2 \phi \right] \Big) E \Big]
1429: \end{eqnarray}
1430: %
1431: For $F(t)$ as defined in (\ref{e12}) we obtain
1432: (after the required replacement $\delta x\mapsto -\delta x$):
1433: %
1434: \begin{eqnarray}
1435: \hspace*{-2cm}
1436: F(t) \ = \ \frac{1}{2\pi} \int_0^{2\pi} d\phi \exp \left[ i \delta x E \sin t \cos(2 \phi-t) \right]
1437: \ = \ J_0\left(2 \delta x E \sin t\right)
1438: \end{eqnarray}
1439: %
1440: The survival probability is obtained by squaring this result.
1441:
1442:
1443: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1444: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1445:
1446: \Bibliography{99}
1447:
1448: \bibitem{wls}
1449: Cohen D and Heller E J 2000 {\it Phys. Rev. Lett.} {\bf 84} 2841
1450:
1451: \bibitem{heller}
1452: Heller E J 1991 {\it Chaos and Quantum Systems},
1453: ed M-J~Giannoni {\it et al} (Amsterdam: Elsevier)
1454:
1455: \bibitem{bilha}
1456: Kallush S, Segev B, Sergeev A V and Heller E J 2002
1457: % ``Surface Jumping: Franck Condon factors and Condon points in phase space ``,
1458: {\it J. Phys. Chem. A} {\bf 106} 6006
1459:
1460: \par\item[] Sergeev A V and Segev B 2002
1461: % ``Most probable path on phase space for a radiationless transition in a molecule",
1462: {\it J. Phys. A: Math. Gen.} {\bf 35} 1769
1463:
1464: \bibitem{fidelity}
1465: Jalabert R A and Pastawski H M 2001
1466: {\it Phys. Rev. Lett.} {\bf 86} 2490
1467: %
1468: \par\item[] Cucchietti F M, Pastawski H M, Jalabert R 2000
1469: {\it Physica A} {\bf 283} 285
1470: %
1471: \par\item[] J.V. Emerson, Phys. Rev. Lett. {\bf 89}, 284102 (2002).
1472: %
1473: \par\item[] Cucchietti F M, Pastawski H M and Wisniacki D A 2002
1474: {\it Phys. Rev. E} {\bf 65} 045206
1475: %
1476: \par\item[] Cucchietti F M, Lewenkopf H, Mucciolo E R,
1477: Pastawski H M and Vallejos R O 2002
1478: {\it Phys. Rev. E} {\bf 65} 046209
1479: %
1480: \par\item[] Benenti G and Casati G 2002 {\it Phys. Rev. E} {\bf 65} 066205
1481: %
1482: \par\item[] Cerruti N R and Tomsovic S
1483: {\it Phys. Rev. Lett.} {\bf 88} 054103
1484: %
1485: \par\item[] Prosen T 2001 {\it Preprint} quant-ph/0106149
1486: %
1487: \par\item[] Prosen T and Znidaric M 2001 {\it J. Phys. A} {\bf 34} L681
1488: %
1489: \par\item[] Eckhardt B 2003 {\it J.Phys. A} {\bf 36} 371
1490: %
1491: \par\item[] Silvestrov P G, Tworzydlo J and Beenakker C W J 2002 {\it Preprint} nlin.CD/0207002
1492:
1493: \bibitem{jacq}
1494: Jacquod P, Silvestrov P G and Beenakker C W J 2001
1495: {\it Phys. Rev. E} {\bf 64} 055203
1496:
1497: \bibitem{fdl}
1498: Wisniacki D A and Cohen D 2002 {\it Phys. Rev. E} {\bf 66} 046209
1499:
1500: \bibitem{fdl2}
1501: Van\'{\i}\v{c}ek J and Heller E J 2003 {\it Preprint} quant-ph/0302192
1502:
1503:
1504:
1505: \bibitem{ovrv}
1506: For a pedagogical presentation, including references,
1507: see references \cite{dsp} and \cite{vrn}, which can be
1508: downloaded from http://www.bgu.ac.il/$\sim$dcohen.
1509:
1510:
1511: \bibitem{dsp}
1512: Cohen D 2002
1513: "Driven chaotic mesoscopic systems, dissipation and decoherence"
1514: in
1515: {\it Dynamics of Dissipation: Proceedings of the 38th Karpacz Winter School of
1516: Theoretical Physics}
1517: ed P Garbaczewski and R Olkiewicz (Springer)
1518:
1519: \bibitem{vrn}
1520: Cohen D 2000
1521: "Chaos, dissipation and quantal Brownian motion" in
1522: {\it New Directions in Quantum Chaos: Proceedings of the International School of Physics Enrico Fermi Course CXLIII}
1523: ed G Casati, I Guarneri and U Smilansky (Amsterdam: IOS Press)
1524:
1525: \bibitem{crs}
1526: Cohen D 1999 {\it Phys. Rev. Lett.} {\bf 82} 4951
1527:
1528: \bibitem{jiri}
1529: Van\'{\i}\v{c}ek J and Heller E J 2001 {\it Phys. Rev. E} {\bf 64} 026215
1530:
1531: \bibitem{jiri2}
1532: Van\'{\i}\v{c}ek J and Heller E J 2003 {\it Phys. Rev. E} {\bf 67} 016211
1533:
1534: \bibitem{lds}
1535: Cohen D and Kottos T 2001 {\it Phys. Rev. E} {\bf 63} 36203
1536:
1537: \bibitem{miller}
1538: Miller W H 1974 {\it Adv. Chem. Phys.} {\bf 25} 69
1539:
1540: \bibitem{miller2}
1541: Miller W H and Smith F T 1978 {\it Phys. Rev. A} {\bf 17} 939
1542:
1543: \bibitem{richter}
1544: Richter K, Ullmo D and Jalabert R A 1996 {\it Phys. Rev. B} {\bf 54} R5219
1545:
1546: \bibitem{kaplan}
1547: L. Kaplan, Phys. Rev. Lett. {\bf 81}, 3371 (1998).
1548:
1549: \end{thebibliography}
1550:
1551: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1552: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1553: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1554: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1555:
1556:
1557:
1558: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1559: \end{document}
1560: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1561: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1562:
1563:
1564:
1565:
1566:
1567:
1568:
1569:
1570:
1571:
1572:
1573:
1574:
1575: