1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: %%% (December 2003)
3: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4:
5: \documentclass[aps,pre,twocolumn,floats]{revtex4}
6: \usepackage{epsfig}
7: \usepackage{bm}
8:
9: \begin{document}
10:
11:
12: % basic commands
13: \newcommand{\hide}[1]{}
14: \newcommand{\tbox}[1]{\mbox{\tiny #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{\intt}{\int\!\!\!\!\int }
20: \newcommand{\ointt}{\int\!\!\!\!\int\!\!\!\!\!\circ\ }
21: \newcommand{\eexp}{\mbox{e}^}
22: \newcommand{\bra}{\left\langle}
23: \newcommand{\ket}{\right\rangle}
24: \newcommand{\EPS} {\mbox{\LARGE $\epsilon$}}
25: \newcommand{\ar}{\mathsf r}
26: \newcommand{\im}{\mbox{Im}}
27: \newcommand{\re}{\mbox{Re}}
28: \newcommand{\bmsf}[1]{\bm{\mathsf{#1}}}
29:
30:
31: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
32:
33: \title{Quantum Dissipation due to the Interaction with Chaos}
34:
35: \author{Doron Cohen$^{1}$ and Tsampikos Kottos$^{2}$}
36:
37: \affiliation{
38: $^{1}$ Department of Physics, Ben-Gurion University, Beer-Sheva 84105, Israel \\
39: $^{2}$Max-Planck-Institut f\"ur Str\"omungsforschung, Bunsenstra\ss e 10,
40: D-37073 G\"ottingen, Germany
41: }
42:
43: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
44:
45: \begin{abstract}
46: We discuss the possibility of having ``quantum dissipation" due to the interaction
47: with chaotic degrees of freedom. We define the conditions that should be satisfied
48: in order to have a dissipative effect similar to the one due to an interaction
49: with a (many body) bath. We also compare with the case where the environment
50: is modeled by a random matrix model. In case of interaction with ``chaos" we observe
51: a regime where the relaxation process is non-universal, and reflects the underlaying
52: semiclassical dynamics. As an example we consider a two level system (spin)
53: that interacts with a two dimensional anharmonic oscillator.
54: \end{abstract}
55:
56: \maketitle
57:
58: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
59:
60:
61: The interaction of a system with its environment is
62: a central theme in classical and quantum mechanics.
63: The main effects that are associated
64: with this interaction are ``dissipation" (irreversible loss of energy)
65: and ``noise". The latter is due to ``fluctuations"
66: of the environmental degrees of freedom.
67: On short time scales the main effect is
68: the ``decoherence" due to the noise.
69: On long time scales the interplay of dissipation and noise
70: leads to a state of thermal equilibrium.
71:
72:
73: The common modeling of the system-environment
74: interaction is provided by a Hamiltonian
75: ${\cal H}_{total} = {\cal H}_0 + {\cal H}(Q,P; x)$,
76: where ${\cal H}_0$ is the system Hamiltonian,
77: $x$ is a system observable,
78: and ${\cal H}(Q,P; x)$ describes the
79: environment including the interaction
80: with the system. The simplest (and most
81: popular) modeling of the environment is
82: as a large collection of {\em harmonic oscillators}.
83: This is known as the Caldeira-Leggett approach \cite{weiss,leggett}.
84: Another approach is to use {\em random matrix} modeling
85: of the environment \cite{mello,pastur,esposito}.
86: %
87: %
88: However, in this paper we are interested
89: in another possibility, where the interaction
90: is with (few) {\em chaotic} degrees of freedom.
91:
92:
93:
94: In what follows {\bf ``interaction with chaos"} means
95: that the environment is the quantized version
96: of a few degrees of freedom chaotic system.
97: This should be contrasted with {\bf ``interaction with bath"}
98: where the environment is modeled as a large
99: collection of quantized harmonic oscillators.
100: %
101: %
102: Let us regard the environment as a ``black box"
103: (one does not know what is there). The questions
104: that we would like to address are:
105: {\bf \ (1)} How to characterize the bath in a way
106: which does not assume a specific model.
107: {\bf \ (2)} Is it possible to distinguish ``interaction with chaos"
108: from ``interaction with bath".
109: As we explain below the second question
110: is related to the notions of ``thermodynamic limit"
111: and ``universality".
112:
113:
114: Common models for dissipation assume
115: a ``thermodynamic limit", which means
116: interaction with infinitely many degrees
117: of freedom. In this paper we inquire
118: whether {\em few} chaotic degrees
119: of freedom may have the same effect.
120: This is a question of great practical importance.
121: Future ``quantum electronics" may consist
122: of several interacting ``quantum dots".
123: One wonders how a coherent process
124: in one part of the ``circuit" is affected by the
125: "noise" which is induced by the quantized chaotic
126: motion of an electron in a nearby quantum dot.
127: In other words: one would like to know whether
128: it is possible to use in the nano-scale reality
129: notions such as ``dissipation" and ``dephasing",
130: that are traditionally associated with having an
131: interaction with many degrees of freedom.
132:
133:
134: As a specific example we consider a two level system (spin)
135: that interacts with a two-dimensional anharmonic oscillator.
136: This would be the well known ``Spin-Boson" model \cite{weiss,leggett}
137: if the interaction were with a bath of harmonic oscillators.
138: The motivation to deal with this model
139: is well documented in the cited literature.
140:
141: We also compare with the case where the interaction
142: is with a {\em random-matrix modeled environment}.
143: In the ``quantum chaos" literature, and in mesoscopic
144: physics, random matrix theory (RMT) is regarded
145: as the ``reference" case. Any deviation from RMT is called
146: "non-universality", and has to do with the underlying
147: semiclassical dynamics. In this paper we show
148: that the notion of (non) universality can be extended
149: into the studies of quantum dissipation.
150:
151: The basic parameters that characterize {\em any}
152: system-environment modeling are listed in Table~1.
153: We shall define these parameters in a way which
154: is independent of the theoretical modeling of the bath.
155: A common assumption is going to be that the mean level spacing
156: is very small. For completeness of presentation,
157: and for the purpose of defining what does it mean ``very small",
158: we keep $\Delta$ as an explicit free parameter (note \cite{rmrk}).
159:
160:
161: %%%%%%%%%%%%%%%%%%%%%%%%%%%
162: %%%%%%%%%%%%%%%%%%%%%%%%%%%
163:
164: \vspace*{0.2cm}
165:
166: \begin{tabular}{|l|l|}
167: \hline
168: {\small \em parameter} & {\small \em \ \ \ \ significance} \\
169: \hline
170: $\Delta\propto\hbar^d$ & environment mean level spacing \\
171: $\Delta_b\propto\hbar$ & environment bandwidth \\
172: $T$ & environment temperature \\
173: $d_{\tbox{T}}$ & environment heat capacity \\
174: \hline
175: $\EPS$ & system energy \\
176: $A$ & system amplitude of motion \\
177: $V$ & system rate of motion \\
178: \hline
179: $\Gamma$ & strength of coupling \\
180: \hline
181: \end{tabular}
182:
183: {\footnotesize \ \\ Table 1: The various parameters that
184: characterize a generic quantum dissipation problem.
185: (See text for details).}
186:
187: \vspace*{0.0cm}
188: \pagebreak
189:
190: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
191: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
192:
193:
194: %%%%% leading example
195:
196: As a leading example we consider a two level system.
197: For the system Hamiltonian we write
198: ${\cal H}_0 = (\hbar\Omega/2) \bm{\sigma}_1$
199: where $\Omega$ is the Bloch frequency,
200: and $\bm{\sigma}_{i=1,2,3}$ are the Pauli matrices.
201: One can think of this Hamiltonian as
202: describing a particle in a double well potential \cite{weiss,leggett}.
203: Then it is natural to define its position as $x = v \bm{\sigma}_3$,
204: where $v$ is a constant. We assume that the interaction with the
205: environment is via this ``position" coordinate:
206: %
207: ${\cal H}(Q,P;x) =
208: \half(P_1^2{+}P_2^2 + Q_1^2{+}Q_2^2) + (1{+}x) Q_1^2 Q_2^2$.
209: %
210: This environment can be interpreted as a particle moving
211: in a 2-dimensional anharmonic well (2DW).
212: In the representation $|\nu,n \rangle$,
213: which is determined by ${\cal H}(Q,P;0)$
214: and $\bm{\sigma}_3$, the Hamiltonian matrix takes the form
215: %
216: \begin{eqnarray} \label{e7}
217: H_{total} \ = \
218: \left[
219: \matrix{\bm{E}+v\bm{B} & \hbar\Omega/2 \cr
220: \hbar\Omega/2 & \bm{E}-v\bm{B}}
221: \right]
222: \end{eqnarray}
223: %
224: where $\bm{E}=\mbox{diag}\{E_n\}$ is a diagonal
225: matrix that contains the energy levels of the environment,
226: and $\bm{B}$ is a banded matrix (see \cite{lds}).
227: % The bandprofile of $\bm{B}$ is plotted in Fig.~2 of \cite{lds}.
228: The initial state of the total Hamiltonian $\Psi(t=0)$
229: is assumed to be factorized as $\varphi \otimes \psi$,
230: where $\varphi$ is the initial state
231: of the spin, and $\psi=|n_0\rangle$
232: is the initial state of the environment.
233: It is implicit that we average over states
234: with $E_{n_0} \sim E$ corresponding to
235: a microcanonical preparation.
236: As for the spin, we would like to consider
237: the standard scenario where the initial
238: state is a coherent superposition
239: $|\varphi\rangle = ( |\uparrow\rangle + |\downarrow\rangle )/\sqrt{2}$.
240: The reduced probability matrix after time $t$ is
241: %
242: \begin{eqnarray} \label{e8}
243: \rho_{\nu,\nu'}(t) =
244: \sum_{n} \Psi_{\nu,n}(t)^{*}\Psi_{\nu',n}(t)
245: \equiv \half (1+ \vec{M} {\cdot} \vec{\bm{\sigma}})_{\nu,\nu'}
246: \end{eqnarray}
247: %
248: where $\vec{M}=(M_1,M_2,M_3)$ is the
249: polarization of the spin. It is most convenient
250: to describe the state of the spin using $\vec{M}$.
251: In particular we define
252: $S(t) =\vec{M}\cdot\vec{M}$ as a
253: measure for the purity of the spin state.
254:
255:
256:
257: %%%% general formulation
258:
259: We turn to formulate the general case.
260: The interaction of the system with the
261: environment is assumed to be of the general form
262: ${\cal H}_{int}=-x{\cal F}$ where
263: $x$ and ${\cal F}$ are system and environmental
264: observables respectively. (In the above example
265: $x = v \bm{\sigma}_3$ and ${\cal F}=-Q_1^2 Q_2^2$).
266: In the absence of system-environment coupling
267: we can characterize the fluctuations of the
268: observable ${\cal F}(t)$ by a correlation
269: function $C(\tau)$.
270: (We are using here Heisenberg picture language).
271: %
272: Its Fourier transform $\tilde{C}_{\tbox{E}}(\omega)$
273: is known as the power spectrum of the fluctuations.
274: The observable ${\cal F}$ has a matrix representation
275: $\langle n | {\cal F} | m \rangle = -\bm{B}_{nm}$
276: where we use the basis which is determined by ${\cal H}$.
277: The fluctuations are related to the {\em bandprofile}
278: of this matrix:
279: %
280: \begin{eqnarray} \label{e1}
281: \tilde{C}_{\tbox{E}}(\omega) &=&
282: \left[\sum_{m}
283: \left|\bm{B}_{nm}\right|^2
284: \ 2\pi\delta\left(\omega-\frac{E_m{-}E_n}{\hbar}\right)\right]_{E_n {\sim} E}
285: \\ \label{e2}
286: &\equiv& 2\pi\sigma^2\delta(\omega)+
287: \frac{2\pi\hbar\sigma^2}{\Delta} R\left(\frac{\hbar\omega}{\Delta}\right)
288: G\left(\frac{\hbar\omega}{\Delta_b}\right)
289: \end{eqnarray}
290: %
291: In the first expression there is an implicit microcanonical averaging
292: over the states $E_n \sim E$. In the second expression $\sigma^2$
293: is the average value $\overline{|\bm{B}_{nm}|^2}$,
294: taken over the near-diagonal matrix elements.
295: %
296: The lower cutoff function $R()$
297: depends on the level spacing statistics \cite{ophir}:
298: It is $R() \approx 1$ for $\omega > \Delta/\hbar$.
299: The mean level spacing $\Delta$ is proportional to
300: $\hbar^d$, where $d$ is the the number of environmental
301: degrees of freedom.
302: %
303: %
304: Any {\em chaotic motion} is characterized by
305: a finite correlation time $\tau_{c}$. Therefore
306: $\tilde{C}_{\tbox{E}}(\omega)$ has a a cutoff frequency
307: $\omega_{c} = 2\pi/\tau_{c}$. This implies
308: (via Eq.(\ref{e1})) that $\bm{B}_{nm}$ is a banded
309: matrix with a bandwidth $\Delta_b=\hbar\omega_c$.
310: The envelope function $G()$, with $G(0)\equiv1$,
311: describes the bandprofile.
312:
313:
314:
315: For sake of comparison we refer to the {\em Spin-Boson}
316: model. The distribution of the bath oscillators
317: is described \cite{leggett,weiss} by a spectral function $J(\omega)$,
318: leading to $\tilde{C}_{\tbox{E}}(\omega){=}2\hbar J(\omega)/(1{-}\eexp{-\beta\hbar\omega})$,
319: where $\beta$ is the reciprocal temperature of the bath.
320: The ``ohmic" assumption of having ``white noise"
321: ($C(\tau)$ with short correlation time $\tau_c$)
322: for high temperatures is imposed {\em by construction},
323: by setting $J(\omega)\propto\omega G(\omega/\omega_c)$.
324: This corresponds to a {\em strong chaos assumption}.
325: However, it should be realized that even if we ``cook"
326: a harmonic bath that has the same $\tilde{C}_{\tbox{E}}(\omega)$
327: as that of a chaotic environment, still there is
328: a major difference: In spite of having the same
329: bandprofile, the $\bm{B}_{nm}$ matrix of a harmonic
330: bath is very {\em sparse}: only states that differ
331: by ``one photon" excitation of a {\em single}
332: oscillator are coupled (hence the differences $E_m{-}E_n$
333: are the frequencies of the harmonic oscillators).
334:
335:
336:
337:
338: In what follows we would like to define the notion
339: of temperature ($T$) without assuming a specific modeling.
340: There are three features of the environment that has to do
341: with this notion: {\bf (i)} The growing density of states;
342: {\bf (ii)} The growing fluctuations intensity;
343: {\bf (iii)} The asymmetry of $\tilde{C}_{\tbox{E}}(\omega)$ with
344: respect to $\omega$.
345: %
346: Let us regard the system as a ``thermometer".
347: The equilibrium is determined by the microcanonical
348: temperature $T=(\partial_E \ln(1/\Delta))^{-1}$.
349: For $d{=}2$ environment with constant density of states
350: we would get $T{=}\infty$. [We consider this hypothetical
351: case for argumentation purpose. This would be indeed
352: the case if the environment were modeled as a billiard.
353: In a later paragraph we discuss the model
354: of Eq.(\ref{e7}) for which $T$ is finite].
355: In case of a two level system, having $T=\infty$
356: implies an equal probability for the two energy states.
357: %
358: Now we can take (instead of a two level system)
359: a particle with one degree of freedom
360: as a ``thermometer". Since we already know that the
361: ``temperature" is (say) $T=\infty$, we may deduce that
362: the friction coefficient, as determined by
363: the fluctuation-dissipation theorem is $\eta = \nu/(2T) = 0$,
364: where $\nu=\tilde{C}_{\tbox{E}}(\omega{\sim}0)$ is the
365: intensity of the fluctuations.
366: This conclusion is {\em wrong}.
367: In fact the friction coefficient
368: is $\eta = \half\Delta {\times} \partial_E(\nu/\Delta)$,
369: as discussed in \cite{frc} (and references therein).
370: Thus we can define an effective temperature,
371: which is related to the thermalization process:
372: %
373: \begin{eqnarray} \label{e3}
374: T_{\tbox{eff}} \ \equiv \ \frac{\nu}{2\eta} \ = \
375: \left(\frac{\partial}{\partial E}
376: \ln\left(\frac{1}{\Delta} \tilde{C}_{\tbox{E}}(\omega{\sim}0) \right)
377: \right)^{-1}
378: \end{eqnarray}
379: %
380: In generic circumstances the distinction between
381: $T_{\tbox{eff}}$ and $T$ is not so dramatic.
382: In fact, some further inspection into Eq.(\ref{e1})
383: reveals that generically $T_{\tbox{eff}} = T/2$.
384: %
385: The above subtlety does not apply to a thermal preparation.
386: Assuming a reciprocal temperature~$\beta$,
387: the friction coefficient (as obtained by canonical averaging
388: over the above cited microcanonical result) is $\eta = \half \beta \nu$.
389: Therefore we get $T_{\tbox{eff}}{=}1/\beta$ as expected.
390:
391:
392:
393: Having defined $T$, we can make a quantum mechanical
394: distinction between high and low temperature regimes.
395: This is related to the asymmetry of $\tilde{C}_{\tbox{E}}(\omega)$
396: with respect to $\omega$.
397: Some inspection into Eq.(\ref{e1}) reveals that
398: a finite temperature implies that the band profile acquires
399: a factor $\exp(\omega/(2T_{\tbox{eff}}))$,
400: which is consistent with the Spin-Boson modeling
401: (see a previous paragraph).
402: {\em ``High temperature"} means that for
403: the physically relevant frequencies $\hbar\omega/T\ll 1$.
404: A~sufficient condition is $T\gg\Delta_b$.
405: In the latter case $\tilde{C}_{\tbox{E}}(\omega)$ can be treated as
406: a symmetric function with respect to $\omega\mapsto-\omega$,
407: and therefore it can be interpreted as the power
408: spectrum of a {\em classical noise}.
409:
410:
411:
412: The issue of {\em thermodynamic limit} is related
413: to the heat capacity of the environment. This is
414: $d_{\tbox{T}} = ({\partial T}/{\partial E})^{-1} \sim d$,
415: where $d$ is its number of {\em environmental} degrees of freedom.
416: The energy that the system can exchange with the environment
417: is denoted by $\EPS$.
418: %
419: If we want to assume a stable temperature $T$,
420: the heat capacity of the environment should be large enough,
421: so that energy exchange between the system
422: and the environment does not have a big effect.
423: This leads to the condition $\EPS \ll d_{\tbox{T}} \times T$.
424: For a generic few degrees of freedom system
425: $\EPS \sim d_0 \times T$,
426: where $d_0$ is the number of {\em system} degrees of freedom.
427: This leads to $d_0 \ll d_{\tbox{T}}$.
428: In case of a two level system the condition is much easier.
429: Namely, we have $\EPS \sim \hbar\Omega$, and therefore
430: we get the ``easy" condition $\hbar\Omega \ll d_{\tbox{T}} \times T$.
431:
432:
433:
434:
435: Assuming the typical circumstances of
436: an oscillating system, the time variation of the system
437: observable $x(t)$ is characterized by an amplitude $|x| \sim A$,
438: and by a rate of change $|\dot{x}| \sim V$.
439: This specification of the system dynamics is essential
440: in order to define a dimensionless parameter
441: that characterizes the system-bath interaction:
442: %
443: \begin{eqnarray} \label{e5}
444: & \frac{\Gamma}{\Delta} \ \ = \ \ \mbox{minimum}\left(
445: \frac{2\pi\sigma^2}{\Delta^2}A^2, \ \left(\frac{\hbar\sigma}{\Delta^2}V\right)^{2/3}\right)
446: \end{eqnarray}
447: %
448: Loosely speaking this parameter indicates how many
449: environmental energy levels are mixed non-perturbatively
450: due to the interaction with the system. The $A$ dependence
451: of $\Gamma$ is the consequence of the well known theory by Wigner:
452: It is the number of levels which are mixed by the perturbation
453: in ${\cal H} \mapsto E_n\delta_{nm} + x \bm{B}_{nm}$ with $|x|\sim A$.
454: If the perturbation is slow (small $\dot{x}$) this $A$ based
455: estimate becomes non-relevant. For a proper analysis \cite{frc}
456: one should switch to the adiabatic ($x$ dependent) basis,
457: leading to ${\cal H} \mapsto E_n\delta_{nm} + \dot{x} (i\hbar\bm{B}_{nm}/(E_n{-}E_m))$
458: with $|\dot{x}|\sim V$. The $V$ based estimate for $\Gamma$
459: follows from the latter representation. Thus one realizes that
460: for slow rate of $x$ variation there is a crossover form
461: a $A$-determined to a $V$-determined mixing.
462:
463:
464: %%%%%%%%%%%
465: %%%%%%%%%%%
466: % numerics
467:
468: The rest of this paper is aimed in clarifying the significance
469: of the parameter $\Gamma$. Some results of the simulations
470: with the 2DW model Eq.(\ref{e7}) are presented in Fig.1.
471: The energy of the environment was in the range
472: $2.8<E<3.2$ where the classical dynamics is predominantly chaotic.
473: The classical correlation time is $\tau_{c}\sim1$.
474: The simulations are done with $\hbar=0.03$.
475: This means that the bandwidth is $\Delta_b \sim 0.2$.
476: This should be contrasted with the mean level spacing $\Delta\sim 0.004$.
477: The temperature in the specified energy range is $T\sim 1.3$,
478: and the heat capacity is $d_{\tbox{T}}\sim2.4$.
479: The amplitude of the motion is $A=v$,
480: while the rate of the motion is formally $V=\infty$.
481: The latter should be understood in
482: the path-integral context (Eq.(\ref{e9})),
483: where $x(t)$ makes ``jumps" between the two sites ($x=\pm v$).
484: Hence $\Gamma\propto(v/\hbar)^2$.
485: One can easily verify that the Kondo parameter \cite{leggett}
486: is $\alpha = (1/16\pi)\Gamma/T$. In our simulations $\alpha \ll 1$.
487:
488:
489: In the lower panels of Fig.1 we present
490: the corresponding results of simulations
491: with a RMT model where the induced fluctuations
492: have {\em exactly} the same power spectrum
493: as in the 2DW model. The RMT model has been
494: obtained by taking the Hamiltonian Eq.(\ref{e7})
495: with a randomized $\bm{B}$.
496: We simply randomized the signs of the off-diagonal
497: elements. This procedure destroys all the
498: correlations between the elements, but does
499: not affect the bandprofile (which is implied by Eq.(\ref{e1})).
500: Two observations should be made
501: immediately: {\bf (i)} A few degrees of freedom chaos
502: indeed provides a dissipative effect,
503: as in the case of a many degrees of freedom bath.
504: {\bf (ii)} The effect of interaction with ``chaos"
505: can be distinguished from the case of
506: a random-matrix modeled environment.
507: The latter claim is based on the observation that
508: in the regime $v>0.16$ there is a two orders
509: of magnitude difference between the
510: corresponding curves. Also the scaling of curves
511: with respect to $v$ is ``broken" (upper panel):
512: The sensitivity to $v$ is much smaller than implied
513: by the overcompensating scaling of the time axis.
514:
515:
516:
517: %%% Semiclassical considerations
518:
519:
520: As explained in the introduction the deviation from RMT
521: is regarded as ``non-universality" and its understanding
522: requires a proper definition of the {\bf classical limit}.
523: This should not be mistaken as a synonym for
524: "high temperatures". The confusing, and possibly meaningless
525: procedure to define this limit is by taking $\hbar\rightarrow0$.
526: We argue below that {\em the meaningful definition
527: of a ``semiclassical regime" is by considering the
528: nature of the dynamics of the environment}.
529: This leads to the condition $\Gamma \gg \Delta_b$.
530: In order to explain this condition we adopt
531: the Feynman-Vernon picture of the dissipation process.
532: %
533: Within this picture the propagator of the
534: reduced probability matrix
535: is written as a path integral:
536: %
537: \begin{eqnarray} \label{e9}
538: K(\nu,\nu'|\nu_0,\nu_0') =
539: \sum_{x_A,x_B} F[x_A, x_B]
540: \ \eexp{i({\cal A}[x_A]-{\cal A}[x_B])}
541: \end{eqnarray}
542: %
543: The summation is over pairs of system
544: trajectories (with weight factors absorbed into the
545: definition of the integration measure).
546: The action ${\cal A}[x]$ is defined as the phase
547: which is accumulated along a given trajectory.
548: We note that for a spin the trajectory is piecewise
549: constant ($x=\pm v$).
550: The influence functional is defined as
551: %
552: %\begin{eqnarray} \label{e10}
553: $F[x_A, x_B] =
554: \langle \psi |
555: U[x_B]^{-1} U[x_A]
556: |\psi \rangle$
557: %\end{eqnarray}
558: %
559: where the expectation value is taken for the
560: initial preparation of the environment
561: (it is typically a mixture of many states
562: implying that an appropriate average should
563: be taken over $\psi$).
564: The environmental evolution operator in case
565: of the models that we consider is
566: %
567: $U[x] = {\cal T}\exp
568: \left( -(i/\hbar)\int_0^t (\bm{E}+x(t')\bm{B})dt'\right)$.
569: %
570: Thus, in a semiclassical framework the problem
571: of ``quantum dissipation" reduces to that
572: of analyzing ``{\em driven degrees of freedom}".
573: It has been realized \cite{crs} that the
574: driven dynamics becomes non-perturbative
575: if $\Gamma>\Delta_b$. Below we discuss the implication
576: of this observation.
577:
578:
579: The purity $S(t)$ is related to the dephasing factor $|F[x_A, x_B]|$.
580: [For $\Omega{=}0$, the Hamiltonian Eq.(\ref{e7}) becomes block diagonal,
581: leading to $S(t) {=} |F[x_A, x_B]|^2$, with $x_A{=}v$ and $x_B{=}{-}v$].
582: If the fluctuations are regarded as ``noise",
583: one obtains the {\em standard} expression
584: %
585: %\begin{eqnarray} \label{e11}
586: $|F[x_A,x_B]| = \eexp{-\frac{1}{2\hbar^2}\int_0^t\int_0^t
587: C(t'{-}t'')(x_B(t')-x_A(t''))^2 dt'dt''}$.
588: %\end{eqnarray}
589: %
590: This expression implies a short time
591: Gaussian decay $S(t)=\exp(-4 C(0) (vt/\hbar)^2)$,
592: which evolves into a long time Gaussian decay
593: $S(t)=\exp(-(\sigma v t / \hbar)^2)$
594: in the regime $\Gamma < \Delta$.
595: If the envelope $G()$ of the bandprofile
596: were smooth, then we would expect,
597: in the regime $\Delta < \Gamma < \Delta_b$,
598: an intermediate stage of exponential decay
599: $S(t)=\exp(-2\gamma t)$ with
600: $\gamma=2 (v/\hbar)^2 \tilde{C}_{\tbox{E}}(\Omega)\sim \Gamma/\hbar$.
601: In case of our numerical example the bandprofile has a structure \cite{lds},
602: and hence $C(\tau)$ has oscillations that show up in the simulations.
603:
604:
605: Can we trust the standard expression for $|F[x_A,x_B]|$
606: if we have a dynamical environment
607: rather than a noise source [with the same $C(\tau)$]?
608: It is not difficult to observe that the standard expression
609: can be derived from Fermi-golden-rule (FGR) considerations.
610: In case of Harmonic bath the FGR treatment is valid,
611: and this expression, with $C(\tau)$ replaced by its
612: symmetrized version, is {\em exact}.
613: But for interaction with "chaos" we have claimed above
614: that non-universality should show up
615: in the non-perturbative regime ($\Gamma > \Delta_b$).
616: Our numerics confirm this prediction:
617: %
618: In the non-perturbative regime ($\Gamma > \Delta_b$)
619: we find a premature ($t < \tau_c$) crossover
620: from the expected short time Gaussian decay,
621: to a non-universal behavior which is not captured
622: by the standard formula. In the 2DW case the system has a
623: classical limit, and therefore the nonperturbative decay of $S(t)$
624: is slowed down (compared with RMT) because it is limited by the classical dynamics.
625: As observed (note the inset) the decay becomes much less sensitive to $v$.
626:
627:
628: The above discussed non-universality can be regarded
629: as the manifestation of ``semiclassical" correlations
630: between the off-diagonal matrix elements of $\bm{B}_{nm}$.
631: These are {\em not} reflected in $\tilde{C}_{\tbox{E}}(\omega)$.
632: %
633: In case of the {\em Spin-Boson} model the $\bm{B}_{nm}$ matrix
634: is sparse, which implies (in the limit of infinite bath)
635: that off-diagonal correlations can be neglected.
636: %
637: In case of a {\em random matrix} modeled environment,
638: absence of correlations is guaranteed by construction
639: for $x=0$, which implies (assuming no sparsity)
640: lack of ``invariance" with respect to $x$ \cite{kuzna}.
641: %
642: Accordingly we have three classes of models that
643: become distinct in the non-perturbative regime.
644:
645:
646:
647: %%%%%%%%%%%%%%
648: %%%%%%%%%%%%%%
649: % conclusions
650:
651: In conclusion, we have discussed the consequences
652: of having four energy scales $(\Delta, \Delta_b, T, \Gamma)$
653: in any generic problem of ``quantum dissipation".
654: The strength of the interaction is characterized by $\Gamma$.
655: A universal "quantum dissipation" behavior
656: requires a separation of energy scales
657: $\Delta \ll \Gamma \ll \Delta_b$.
658: Non-perturbative breakdown of this universality
659: due to the underlaying semiclassical dynamics
660: is found if $\Gamma > \Delta_b$.
661:
662:
663: We thank M. Esposito, P. Gaspard, L. Pastur, and V.~Falco for stimulating discussions.
664: This work was supported by a Grant from the GIF,
665: the German-Israeli Foundation for Scientific Research and Development,
666: and by the Israel Science Foundation (grant No.11/02).
667:
668: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
669: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
670:
671: \begin{thebibliography}{99}
672:
673: \bibitem{weiss}
674: {\em Quantum Dissipative Systems}, U. Weiss, World Scientific, Singapore (1999).
675:
676: \bibitem{leggett}
677: A.J. Leggett et al, Rev. Mod. Phys. {\bf 59}, 1 (1987).
678:
679: \bibitem{pastur} J. Lebowitz and L. Pastur,
680: % ``on random matrix model of relaxation",
681: preprint (2002).
682:
683: \bibitem{esposito}
684: M. Esposito and P. Gaspard,
685: % spin relaxation in complex environment,
686: Phys. Rev. E {\bf 68}, 066113 (2003);
687: % Quantum master equation for a system influencing its environment
688: Phys. Rev. E {\bf 68}, 066112 (2003).
689:
690: \bibitem{mello}
691: P. Pereyra, J. Stat. Phys. {\bf 65}, 773 (1991).
692: P.A. Mello, P. Pereyra and N. Kumar, J. Stat. Phys. {\bf 51}, 77 (1988).
693: % P.A. Mello et al, J. Stat. Phys. {\bf 51}, 77 (1988).
694: % P. A. Bulgac, G.D. Dang and D. Kusnezov, Phys. Rev. E {\bf 58}, 196 (1998).
695:
696: \bibitem{rmrk}
697: From here on we ignore the effect
698: of recurrences that happen after an extremely large time.
699:
700: \bibitem{ophir}
701: O.M. Auslaender and S. Fishman,
702: J. Phys. A {\bf 33}, 1957 (2000);
703: Phys. Rev. Lett. {\bf 84}, 1886 (2000).
704:
705: \bibitem{frc} D. Cohen, Annals of Physics 283, 175 (2000).
706:
707: \bibitem{lds}
708: D. Cohen and T. Kottos, Phys. Rev. E 63, 36203 (2001).
709:
710: \bibitem{crs}
711: D. Cohen, Phys. Rev. Lett. {\bf 82}, 4951 (1999).
712: D. Cohen and T. Kottos, Phys. Rev. Lett. {\bf 85}, 4839 (2000).
713: % T. Kottos and D. Cohen, Phys. Rev. E {\bf 64}, R-065202 (2001).
714:
715: % \bibitem{wls}
716: % D. Cohen and E.J. Heller, PRL 84, 2841 (2000). \\
717: % R.A. Jalabert and H.M. Pastawski, PRL {\bf 86}, 2490 (2001).
718:
719: \bibitem{kuzna}
720: Attempts to have an $x$ invariant random matrix model, [such as in
721: P. A. Bulgac et al, PRE {\bf 58}, 196 (1998)],
722: make the model ``perturbative by construction" \cite{crs}.
723:
724: \end{thebibliography}
725:
726: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
727: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
728:
729: \clearpage
730: \onecolumngrid
731:
732: \ \\
733:
734: \epsfig{figure=dfs_fig, width=0.85\hsize}
735:
736: \ \\
737:
738: {\footnotesize\noindent {\bf FIG.1:}
739: The decay of $S(t)$ in case of interaction with
740: chaos (upper panels) and in case of interaction with a ``randomized"
741: environment that has exactly the same fluctuations (lower panels).
742: The selected values of the coupling parameter are in the range \mbox{$10^{-4}<v<0.3$}.
743: The left panels are for the $\Gamma<\Delta$ regime \mbox{($v<0.02$)} and for the
744: $\Gamma>\Delta_b$ nonperturbative regime \mbox{($v>0.16$)}.
745: The inset is for some of the \mbox{$v>0.16$} curves without
746: scaling. The arrows point on the largest $v$ value.
747: The right panels are for the $\Delta<\Gamma<\Delta_b$ regime \mbox{($0.02<v<0.16$)}.
748: The scaling methods of the time axis are implied by the analysis of $S(t)$,
749: which should hold if there is a universal behavior which
750: is determined by $C(\tau)$ alone (see text).}
751:
752:
753: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
754: \end{document}
755: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
756: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
757:
758:
759:
760:
761:
762:
763: