1: \documentstyle[multicol,overcite,pre,aps,epsf]{revtex}
2:
3: \begin{document}
4: \tightenlines
5:
6:
7: \title{Emergence of Quantum-Classical Dynamics in an Open Quantum Environment}
8: \author{Kazutomu Shiokawa and Raymond Kapral
9: \\ Chemical Physics Theory Group, Department of Chemistry,\\
10: University of Toronto, Toronto, ON M5S 3H6, Canada}
11: \maketitle
12:
13: \begin{abstract}
14: The conditions under which an open quantum mechanical system may
15: be described by mixed quantum-classical dynamics are investigated.
16: Decoherence is studied using influence functional methods in a model
17: composite quantum system comprising two coupled systems, $A$ and $C$,
18: interacting with a harmonic bath with Ohmic and
19: super-Ohmic spectral densities. Subsystem $A$ is directly coupled
20: to subsystem $C$, while $C$ is coupled directly to the bath.
21: Calculations are presented for a model where subsystem $A$ is
22: taken to be a two-level system which is bilinearly coupled to a
23: single harmonic oscillator $C$ subsystem. The loss of quantum
24: coherence in each subsystem is discussed in the extreme
25: non-adiabatic regime where the dynamics of subsystem $A$ is
26: essentially frozen.
27: Subsystem $C$ is shown to lose its coherence rapidly, while
28: subsystem $A$ maintains coherence for longer time periods since
29: $C$ modulates the influence of the bath on $A$. Thus,
30: one may identify situations where the coupled $AC$ system
31: evolution effectively obeys mixed quantum-classical dynamics.
32: \end{abstract}
33: \pacs{xx}
34: \begin{multicols}{2}
35: \narrowtext
36:
37: \section{Introduction}
38:
39: Since it is often impossible to isolate a quantum system
40: completely from its environment, the study of open quantum
41: dynamical systems is necessary in many circumstances. Techniques
42: used to treat open quantum systems have been developed extensively
43: and used in numerous applications. \cite{davis,Mukamel95,Weiss99}
44: Projection operator techniques \cite{Nakajima58,Zwanzig61} and
45: influence functional methods \cite{FeyVer63} have been
46: actively used in chemical physics in studies of electron, proton,
47: and exciton transfer processes in the condensed phase and in
48: biological systems. \cite{QDOS,makri,EgoEveSki99,MayKuhn00}
49:
50: In this paper, we investigate some of the circumstances under
51: which an open quantum system can be described using mixed
52: quantum-classical dynamics. Quantum-classical dynamics is
53: used to study condensed phase many-body systems
54: \cite{tully0,herman}, especially in the context of methods for
55: treating non-adiabatic dynamics
56: \cite{tully1,coker,rossky,martinez,prezhdo,martens,kapral,schofield}.
57: In the studies presented here, we consider a composite quantum
58: mechanical system $AC$ comprising two coupled subsystems $A$ and
59: $C$; subsystem $C$ is assumed to be in direct contact with a
60: thermal quantum mechanical bath $B$. Systems with this structure
61: can arise in condensed phase dynamics where certain quantum
62: degrees of freedom ($A$), for example, those associated with
63: protons or electrons, may interact directly with neighboring
64: solvent molecules ($C$), which in turn interact with the rest of
65: the solvent ($B$). It is interesting to determine if
66: the composite system may be treated as
67: mixed quantum-classical system where the dynamics of subsystem
68: $C$, which is in direct contact with the heat bath, is classical
69: in character while subsystem $A$ retains its quantum nature. Some
70: aspects of the dynamics of such quantum-classical composite systems
71: have been investigated. \cite{kapral2}
72:
73: In order to investigate this problem we study a simple model
74: system where subsystem $A$ depends on spin degrees of freedom and
75: subsystem $C$ is a single harmonic oscillator bilinearly coupled
76: both to subsystem $A$ and the bath. The bath is a collection of
77: independent harmonic oscillators. While this is a highly
78: simplified model of the realistic systems discussed above, it does
79: capture some essential features of real coupled systems and is
80: amenable to detailed analysis. Due to its interaction with the
81: bath, the dynamics of subsystem $C$ is dissipative and executes
82: brownian motion. The brownian motion of a quantum particle
83: governed by different potential functions and immersed in a
84: thermal harmonic oscillator bath has been studied
85: extensively by influence functional methods\cite{CaldeiraLeggett84,
86: HakimAmb85,GSI88,UnruhZurek89,QBM1,QBM2}.
87: The character of its dynamics is
88: determined by the system-bath coupling and the spectral density
89: and temperature of the bath. In the composite system we study, the
90: dynamics of the $C$ subsystem oscillator is also influenced by the
91: quantum dynamics of subsystem $A$. The dynamics of subsystem $A$
92: is more complicated. It is also dissipative but its energy must be
93: transmitted through subsystem $C$ to the bath. Our results on the
94: applicability of mixed quantum-classical dynamics are based on the
95: nature of decoherence\cite{EID,Dec96} in the coupled system: when one subsystem
96: behaves quantum mechanically and the other classically, there must
97: be a mechanism making the former decohere slowly and the latter
98: quickly.
99:
100: The outline of this paper is as follows: In Sec.~\ref{sec:IF} we
101: specify the model in detail and outline the application of the
102: influence functional formalism to it. Although we
103: use influence functional methods, similar results can be obtained
104: by other methods. Section~\ref{sec:QBM} considers the
105: equilibration of the $CB$ system in the absence of $A$ and
106: establishes that this composite system can be described by an
107: effective spectral density. The emergence of mixed
108: quantum-classical dynamics is investigated in
109: Sec.~\ref{sec:emergence}. In this section we use the results
110: developed in the previous sections, specialized to a two-level system
111: in the strongly non-adiabatic regime where the populations change
112: very slowly. We study the decoherence of the off-diagonal elements
113: of the subsystem $A$ reduced density matrix in this regime and
114: show under what conditions one may observe quantum-classical
115: dynamics. The conclusions of the study are presented in
116: Sec.~\ref{sec:conc}. The Appendices contain additional details of
117: the calculations for Ohmic and super-Ohmic spectral densities.
118:
119: \section{General formulation} \label{sec:IF}
120:
121: \subsection{Composite system in a thermal bath}
122:
123: The composite $ACB$ quantum system consists of three coupled subsystems:
124: subsystem $A$ is coupled directly to subsystem $C$, while subsystem $C$
125: is in direct contact with a thermal bath $B$. The $ACB$ system hamiltonian
126: is,
127: \begin{eqnarray}
128: \hat{H} = \hat{H}_A + \hat{V}_{AC} + \hat{H}_C + \hat{V}_{CB}
129: +\Delta \hat{H}_C + \hat{H}_B\;.
130: \label{H}
131: \end{eqnarray}
132: The thermal bath is taken to be a set of $N$ independent harmonic
133: oscillators with frequencies $\omega_{n}$, masses $m_{n}$, and coordinates
134: and conjugate momenta
135: $(\hat{q},\hat{p})\equiv (\hat{q}_1,...,\hat{q}_N,\hat{p}_1,...,\hat{p}_N)$
136: so that the hamiltonian is
137: \begin{eqnarray}
138: \hat{H}_{B} = \sum_{n=1}^{N}( \frac{\hat{p}_{n}^{2}}{2 m_{n}}
139: + \frac{1}{2}m_{n} \omega_{n}^{2}\hat{q}_{n}^{2}).
140: \label{HB}
141: \end{eqnarray}
142: Subsystem $C$ consists of a single harmonic oscillator with renormalized
143: mass $M$ and frequency $\omega_r$ with coordinate and momentum operators
144: $(\hat{x}, \hat{p})$,
145: \begin{eqnarray}
146: \hat{H}_C = \frac{\hat{p}^2}{2M} + \frac{M}{2} \omega_r^2 \hat{x}^2\;.
147: \label{HC}
148: \end{eqnarray}
149: The interaction potential between $C$ and the bath is assumed to be bilinear,
150: \begin{eqnarray}
151: \hat{V}_{CB} = \hat{x} \sum_{n=1}^{N} c_{n} \hat{q}_{n},
152: \label{HI}
153: \end{eqnarray}
154: where $c_n$ is the coupling constant to the $n$th oscillator. The coupling
155: constants are related to the spectral density of the bath by,
156: \begin{eqnarray}
157: J_B(\omega)
158: \equiv \pi \sum_{n}\frac{c_{n}^2}{2 m_n \omega_n} \delta(\omega - \omega_n).
159: \label{SpectralDensity}
160: \end{eqnarray}
161: We suppose the spectral density has the form
162: \begin{equation}
163: J_{B}(\omega) = \eta \omega^{\nu} e^{-\omega/\Lambda},
164: \end{equation}
165: and consider both Ohmic ($\nu=1$) and super-Ohmic ($\nu=3$) cases.
166: A counter term $\Delta \hat{H}_C$
167: has been added to the hamiltonian. It depends on $c_n$, $m_n$,
168: $\omega_n$, $\hat{p}$ and $\hat{x}$ and is given by
169: \begin{eqnarray}
170: \Delta \hat{H}_c &=&
171: \left\{
172: \begin{array}{lll}
173: \eta \Lambda \hat{x}^2 / \pi & \hspace{1cm} &(\nu=1) \\
174: \eta \Lambda \hat{p}^2 / M^2 \pi + \eta \Lambda^3 \hat{x}^2 / \pi
175: &\hspace{1cm} &(\nu=3).
176: \end{array}
177: \right.
178: \label{counterterm}
179: \end{eqnarray}
180: This term is introduced to
181: cancel the shift of the mass and frequency of the $C$ subsystem
182: oscillator due to the interaction with the bath which will
183: become divergent when the frequency cutoff $\Lambda$ goes to infinity.
184: As is customary, we consider the
185: renormalized quantities after including a counter term
186: as physical observables with specified values in our analysis \cite{IZ}.
187: The hamiltonian of subsystem $A$, $\hat{H}_A(\sigma)$, depends on
188: spin variables $\sigma$ and we assume subsystem $A$ is bilinearly coupled
189: to subsystem $C$, $\hat{V}_{AC}=\lambda \sigma \hat{x}$.
190:
191: The dynamical properties of interest can be computed from the
192: density matrix of the system at time $t$. In the coordinate
193: representation $\{x,q\}$ of the $C$ subsystem and bath $B$, the
194: density matrix takes the form, $\hat{\rho}(x,x',q,q',t) \equiv
195: \langle x ~q | \hat{\rho}(t) | x' ~q' \rangle$, which is still an
196: operator in the $A$ subsystem degrees of freedom. The density
197: matrix may be written more explicitly in integral form in terms of
198: the propagator
199: \begin{eqnarray}
200: \hat{K}(x,q;t \mid x_0, q_{0};0)
201: \equiv \langle x ~q | e^{- i \hat{H} t/ \hbar} | x_0 ~q_{0} \rangle \;,
202: \label{KernelK}
203: \end{eqnarray}
204: as
205: \begin{eqnarray}
206: \hat{\rho}(x,x',q,q',t) &=&
207: \int d x_{0} d x'_{0} d q_{0} d q'_{0}
208: \hat{K}(x,q; t \mid x_0, q_{0};0) \nonumber \\
209: &\times& \hat{\rho}(x_0, x'_0, q_{0}, q'_{0}, 0)
210: \hat{K}^{*}(x',q'; t \mid x'_0, q'_{0};0) \;. \nonumber \\
211: \label{DMintegral}
212: \end{eqnarray}
213:
214: We are primarily interested in the dynamics of the composite $AC$
215: subsystem under the influence of the thermal bath. In such a
216: circumstance the reduced density matrix $\hat{\rho}_r$, obtained by
217: integrating over the bath degrees of freedom, is the relevant
218: quantity. Such a reduction is especially appropriate if the
219: characteristic time scale for the bath is much shorter than those
220: for the $A$ and $C$ subsystems. We assume a factorized initial
221: condition between the $AC$ subsystem and the bath,
222: \begin{eqnarray}
223: \hat{\rho}(x_0, x'_0, q_{0}, q'_{0}, 0) =
224: \hat{\rho}_{AC}(x_0, x'_0, 0)
225: \otimes
226: \rho_{B}(q_{0}, q'_{0},0)\;,
227: \label{IDMAC}
228: \end{eqnarray}
229: and the bath is always taken to be initially in thermal equilibrium.
230: Under these conditions the integral form of the reduced density matrix
231: at time $t$ is
232: \begin{eqnarray}
233: \hat{\rho}_r(x,x',t) &=&
234: \int d x_{0} d x'_{0}
235: \hat{J}_r(x, x';t \mid x_0, x'_0; 0)
236: \hat{\rho}_{AC}(x_0, x'_0, 0), \nonumber \\
237: \label{DMredintegral}
238: \end{eqnarray}
239: where the time evolution kernel $J_r$ is given by
240: \begin{eqnarray}
241: \hat{J}_r(x, x';t \mid x_0, x'_0; 0) &=&
242: \int d q d q_{0} d q'_{0}
243: \hat{K}(x,q;t \mid x_0, q_{0};0) \nonumber \\
244: &\times& \rho_{B}(q_{0}, q'_{0} ,0)
245: \hat{K}^{*}(x',q; t \mid x'_0, q'_{0};0). \nonumber \\
246: \label{Jr}
247: \end{eqnarray}
248: Given the initial conditions discussed above, this exact
249: expression for the reduced density matrix specifies a
250: non-Markovian time evolution since the solution at time $t$
251: depends on its past history. Approximate Markovian evolution equations
252: may miss essential features of the quantum/classical correspondence and
253: tend to underestimate the loss of quantum coherence. In addition, a Markovian
254: approximation is not generally valid for a harmonic oscillator
255: model, except for systems with Ohmic type dissipation in the high
256: temperature regime.
257: Below we use
258: influence functional methods where the exact solution is expressed
259: as a path integral in order to deal with the nonlocal time
260: evolution.
261:
262: \subsection{Influence functional method}
263:
264: The formulation of the reduced density matrix in terms of an
265: influence functional for the composite $ACB$ system parallels
266: that for spin-boson models often discussed in the literature.
267: \cite{LCDFGZ87,Weiss99} Consequently the presentation of the
268: generalization to our system will be brief.
269: In order to write the reduced density matrix in terms of an
270: influence functional, we first introduce a basis of spin
271: functions with labels $\sigma$ and $\sigma'$ to represent the states of
272: the $A$ subsystem. In this basis we may write eq.~(\ref{DMredintegral}) as,
273: \begin{eqnarray}
274: && \rho_r(\sigma, \sigma', x, x', t)= \int d \sigma_0 d \sigma'_0
275: \int d x_{0} d x'_{0} \nonumber \\
276: && \times J_r(\sigma, \sigma', x, x';t
277: \mid \sigma_0, \sigma_0', x_0, x'_0; 0)
278: \rho_{r}( \sigma_0, \sigma_0', x_0, x'_0, 0). \nonumber \\
279: \label{DMACf}
280: \end{eqnarray}
281: For the factorized initial condition (\ref{IDMAC}) the kernel
282: $J_r$ can be written in the path integral form as
283: \begin{eqnarray}
284: && J_r(\sigma, \sigma', x, x';t \mid \sigma_0, \sigma'_0, x_0, x'_0; 0) = \nonumber \\
285: && \int^{(\sigma \sigma' )}_{(\sigma_0 \sigma'_0 )} {\mathcal D}
286: \sigma {\mathcal D} \sigma' K[\sigma] K^{*}[\sigma']
287: J_{C}^{\sigma \sigma'}(x, x';t \mid x_0, x'_0; 0),
288: \label{Jrpathint}
289: \end{eqnarray}
290: where $K[\sigma]$ is a probability amplitude for subsystem $A$
291: following the path $\sigma$ in the absence of the subsystem $C$
292: and the environment. The evolution kernel $J_{C}^{\sigma \sigma'}$ for
293: subsystem $C$ in the presence of external sources $(\sigma,
294: \sigma')$ is
295: \begin{eqnarray}
296: &&J_{C}^{\sigma \sigma'}(x, x';t \mid x_0, x'_0; 0)
297: \equiv \int^{(x x')}_{(x_0 x'_0)} {\mathcal D} x {\mathcal D} x'
298: e^{i {\cal S}[\sigma,\sigma',x,x']/\hbar},
299: \nonumber \\
300: \label{JF}
301: \end{eqnarray}
302: The action ${\cal S}[\sigma,\sigma',x,x']$ consists of several
303: contributions and can be decomposed as follows:
304: \begin{eqnarray}
305: {\cal S}[\sigma,\sigma',x,x']&=& {\cal S}_C[x,x'] +
306: {\cal S}_{AC}[\sigma,\sigma',x,x'] \nonumber \\
307: &&+\Delta {\cal S}_C[x,x']
308: +{\cal S}_{IF}[x,x'].
309: \end{eqnarray}
310: The actions for subsystem $C$, plus its counter action, and the
311: interaction between subsystems $A$ and $C$ are
312: \begin{eqnarray}
313: ({\cal S}_C+\Delta {\cal S}_C)[x,x']&=& \int_{0}^{t} ds \Big(
314: \frac{M_{0}}{2} \dot{x}^2(s) -
315: \frac{M_{0}}{2}\omega_{0}^2 x^2(s) \nonumber \\
316: &&-\frac{M_{0}}{2} \dot{x'}^2(s) +
317: \frac{M_{0}}{2}\omega_{0}^2 x^{\prime 2}(s) \Big), \nonumber \\
318: {\cal S}_{AC}[\sigma,\sigma',x,x']&= &
319: \lambda \int_{0}^{t} ds \left( x(s) \sigma(s) - x'(s) \sigma'(s)
320: \right)\;, \nonumber \\
321: \label{SAC}
322: \end{eqnarray}
323: where, for notational convenience, we used the bare mass $M_{0}$ and bare frequency $\omega_{0}$ as
324: $M_{0}=M$ and $M_{0}\omega_{0}^2=M\omega_r^2+
325: 2\eta \Lambda/\pi$ for $\nu=1$ while $M_{0}=M+2\eta \Lambda/\pi$
326: and $M_{0}\omega_{0}^2=M\omega_r^2+ 2\eta \Lambda^3/(3\pi)$
327: for $\nu=3$.
328:
329: The influence action ${\cal S}_{IF}[x,x']$ accounts for the effect
330: of the bath on $C$ and is given by
331: \begin{eqnarray}
332: {\cal S}_{IF}[x,x'] &=&
333: i \int_{0}^{t} ds \int_{0}^{s} ds'
334: [ x(s) - x'(s) ] \times \nonumber \\
335: &&[\alpha(s-s') x(s') - \alpha^{*}(s-s') x'(s')],
336: \label{IFaction}
337: \end{eqnarray}
338: where $\alpha(t)$ is a complex kernel whose real $\alpha^R(t)$
339: and imaginary $\alpha^I(t)$ parts, respectively, are given by
340: \begin{eqnarray}
341: \alpha^R(t)&=& \frac{1}{\pi} \int_{0}^{\infty} d \omega J_B(\omega)
342: \coth \frac{\beta \hbar \omega}{2} \cos \omega t, \\
343: \alpha^I(t)&=&-\frac{1}{\pi} \int_{0}^{\infty} d \omega J_B(\omega)
344: \sin \omega t.
345: \label{noisekernel}
346: \end{eqnarray}
347: We define new variables as $R \equiv (x+x')/2$,
348: $r \equiv x-x'$, $\sigma_{\pm} \equiv \sigma \pm \sigma'$, and
349: write the actions in these new variables as
350: \begin{eqnarray}
351: &&{\cal S}_{AC}[\sigma_{\pm},R,r]= \frac{\lambda}{2}
352: \int_{0}^{t} ds \left( \sigma_{+}(s) r(s) + 2 \sigma_{-}(s) R(s)
353: \right), \nonumber \\
354: &&({\cal S}_{C}+\Delta {\cal S}_C)[R,r]= \int_{0}^{t} ds \{ M_{0}
355: \dot{R}(s)\dot{r}(s) - M_{0}\omega_{0}^2 R(s)r(s) \}, \nonumber \\
356: &&{\cal S}_{IF}[R,r] = i \int_{0}^{t} ds \int_{0}^{s} ds' r(s)
357: \alpha^R(s-s') r(s') \nonumber \\
358: &&\qquad \qquad - 2 \int_{0}^{t} ds \int_{0}^{s} ds' r(s)
359: \alpha^I(s-s') R(s').
360: \label{SAC2}
361: \end{eqnarray}
362:
363: \subsection{Euler-Lagrange equations and solution} \label{EL}
364:
365: >From eqs.~(\ref{SAC2}) the Euler-Lagrange equations for $R$ and $r$ are
366: \begin{eqnarray}
367: M_{0} \ddot{R}_c(s)
368: + M_{0} \omega_{0}^2 R_c(s)
369: &+& 2 \int_{0}^{s} ds'
370: \alpha^I(s-s') R_c(s') \nonumber \\
371: &=& \frac{1}{2}\lambda \sigma_{+}(s),
372: \label{EL1}
373: \end{eqnarray}
374: \begin{eqnarray}
375: M_{0} \ddot{r}_c(s) +
376: M_{0} \omega_{0}^2 r_c(s)
377: &-& 2 \int_{s}^{t} ds' \alpha^I(s-s') r_c(s') \nonumber \\
378: &=&\lambda \sigma_{-}(s). \label{EL2}
379: \end{eqnarray}
380: The initial and final conditions for eq.~(\ref{EL1})
381: (eq.~(\ref{EL2})) are $R_0$ and $R_t$ ($r_0$ and $r_t$).
382: If we let the two independent solutions of the homogeneous part of
383: eq.~(\ref{EL1}) (eq.~(\ref{EL2})) be $u_i(s)$($v_i(s)$), $i=1,2$,
384: with boundary conditions
385: $u_1(0)=1,u_1(t)=0$, $u_2(0)=0,u_2(t)=1$
386: ($v_1(0)=1,v_1(t)=0$, $v_2(0)=0,v_2(t)=1$),
387: the solutions of these uncoupled equations can be written as
388: \begin{eqnarray}
389: R_c(s)
390: &=& R_0 u_1(s) + R_t u_2(s)
391: +\frac{\mu}{2} \sigma_{+}(g(s)),
392: \nonumber \\
393: r_c(s)
394: &=& r_0 v_1(s) + r_t v_2(s)
395: + \mu \sigma_{-}(g(s)),
396: \label{SolutionRr}
397: \end{eqnarray}
398: where we have used the notation,
399: \begin{eqnarray}
400: \sigma_{+}(g(s)) &\equiv& \int_{0}^{s} ds' g_{+}(s-s')
401: \sigma_{+}(s') \nonumber \\
402: &-& u_2(s) \int_{0}^{t} ds' g_{+}(t-s') \sigma_{+}(s'),
403: \nonumber \\
404: \sigma_{-}(g(s)) &\equiv& \int_{0}^{s} ds' g_{-}(s-s')
405: \sigma_{-}(s') \nonumber \\
406: &-& v_2(s) \int_{0}^{t} ds' g_{-}(t-s') \sigma_{-}(s').
407: \label{g+g-}
408: \end{eqnarray}
409: and defined $\mu \equiv \lambda / M$.
410: The solutions $v_1$ and $v_2$ satisfy the homogeneous part of
411: the backward time equation (\ref{EL2})
412: and are related to $u_1$ and $u_2$ by $v_1(s)=u_2(t-s)$ and
413: $v_2(s)=u_1(t-s)$. The function
414: $g_{+}(s)$ ($g_{-}(s)$) also satisfies the homogeneous part of
415: eq.~(\ref{EL1}) (eq.~(\ref{EL2}))
416: with boundary conditions $g_{\pm}(0)=0,\dot{g}_{\pm}(0)=1$.
417:
418: The solutions for $g_{\pm}$ are given in Appendix A for Ohmic
419: and super-Ohmic spectral densities. From these solutions $u_{1,2}$
420: and $v_{1,2}$ can be determined.
421:
422: \subsection{Reduced density matrix solution}
423:
424: Since the potentials in our model are harmonic, an exact evaluation of the path integral
425: can be carried out. It is dominated by the classical solution given
426: in eq.~(\ref{SolutionRr}). The $\sigma$ dependence in eq.~(\ref{SolutionRr})
427: arises from the back reaction of the $A$ subsystem on the $C$
428: subsystem. To separate out the contribution from this back reaction,
429: we expand ${\cal S}$ in powers of $\sigma$ and write,
430: \begin{eqnarray}
431: {\cal S}[\sigma_{\pm},R_c,r_c]&=&
432: {\cal S}^{(0)}[R_c,r_c] +
433: {\cal S}^{(1)}[\sigma_{\pm},R_c,r_c] +
434: {\cal S}^{(2)}[\sigma_{\pm}]. \nonumber \\
435: \label{S-SSS}
436: \end{eqnarray}
437: The zeroth order term in $\sigma$ takes the form
438: \begin{eqnarray}
439: {\cal S}^{(0)}[R_c,r_c]
440: &=& \Big( M \dot{u}_1(t) R_0 + M \dot{u}_2(t) R_t\Big) r_t \nonumber \\
441: &-& \Big( M \dot{u}_1(0) R_0 + M \dot{u}_2(0) R_t\Big)r_0 \nonumber \\
442: &+& i \Big( a_{11}(t) r_0^2 + ( a_{12}(t) + a_{21}(t)) r_0 r_t
443: \nonumber \\
444: &&+ a_{22}(t) r_t^2\Big)\;,
445: \label{SQBM}
446: \end{eqnarray}
447: and is an influence action for a damped harmonic
448: oscillator\cite{GSI88,QBM1}. Here
449: \begin{eqnarray}
450: a_{kl}(t) &=& \frac{1}{2} \int_{0}^{t} ds \int_{0}^{t} ds'
451: v_k(s) \alpha^R(s-s') v_l(s'),
452: \label{aij}
453: \end{eqnarray}
454: for $(k,l=1,2)$ contains the effects of fluctuations due to the bath on
455: subsystem $C$. The term linear in $\sigma$ is
456: \begin{eqnarray}
457: &&{\cal S}^{(1)}[\sigma_{\pm},R_c,r_c] = \lambda \int_{0}^{t} ds
458: \left( u_1(s) R_0 + u_2(s) R_t \right) \sigma_{-}(s) \nonumber \\
459: &+& i \mu r_0 \int_{0}^{t} ds \int_{0}^{t} ds'
460: \{ v_1(s) \alpha^R(s-s') \sigma_{-}(g(s')) \} \nonumber \\
461: &+& i \mu r_t \int_{0}^{t} ds \int_{0}^{t} ds'
462: \{ v_2(s) \alpha^R(s-s') \sigma_{-}(g(s')) \} \nonumber \\
463: &+& \frac{\lambda r_0}{2} \int_{0}^{t} ds \;
464: v_1(s) \sigma_{+}(s) + \frac{\lambda r_t}{2}\int_{0}^{t} ds
465: v_2(s) \sigma_{+}(s)\;.
466: \label{SABC}
467: \end{eqnarray}
468: This term arises from the interaction between subsystem $A$ and
469: the bath, modulated by the trajectory of subsystem $C$.
470: The quantum back reaction of subsystem $A$ on $C$ induces
471: self-coupling in $A$, which is contained in the last term,
472: \begin{eqnarray}
473: {\cal S}^{(2)}[\sigma_{\pm}]
474: &=& \frac{\lambda \mu}{2}
475: \int_{0}^{t} ds \; \sigma_{-}(s) \sigma_{+}(g(s)) \nonumber \\
476: &+&i \mu^2 \int_{0}^{t} ds \int_{0}^{s} ds' \;
477: \sigma_{-}(g(s)) \alpha^R(s-s') \sigma_{-}(g(s')). \nonumber \\
478: \label{SAB}
479: \end{eqnarray}
480:
481: Using the results above, $J_C^{\sigma \sigma'}$ in eq.~(\ref{JF})
482: can be written in the compact form,
483: \begin{eqnarray}
484: J_{C}^{\sigma \sigma'}(R_t, r_t;t \mid R_0, r_0; 0) =
485: N(t) \exp\Big\{\frac{i}{\hbar} {\mathcal L} \Big\},
486: \label{Jrsimple}
487: \end{eqnarray}
488: where ${\mathcal L}= \vec{R}^{T} {\bf u} \vec{r} +i \vec{r}^{T}{\bf a} \vec{r}
489: + \vec{\sigma}_{u}^{T} \vec{R} + \vec{\sigma}_{v}^{T} \vec{r}
490: + {\cal S}^{(2)}[\sigma_{\pm}] $ and $N(t)$ is a normalization factor
491: independent of dynamical variables. In writing this equation we have
492: introduced the notation and $({\bf a})_{ij} = a_{ij}$,
493: $\vec{R}^T=(R_0, R_t)$ and $\vec{r}^T=(r_0, r_t)$, where $T$
494: stands for the transpose. The matrix ${\bf u}$ is defined as
495: \begin{eqnarray}
496: {\bf u} =
497: \left( \begin{array}{cc}
498: u_{11} & u_{12} \\
499: u_{21} & u_{22}
500: \end{array} \right)
501: \equiv
502: M \left( \begin{array}{cc}
503: - \dot{u}_{1}(0) & \dot{u}_{1}(t) \\
504: - \dot{u}_{2}(0) & \dot{u}_{2}(t)
505: \end{array} \right)\;,
506: \label{U}
507: \end{eqnarray}
508: while the vectors $\vec{\sigma}_{u}$ and $\vec{\sigma}_{v}$ are given by
509: \begin{eqnarray}
510: \vec{\sigma}_{u} =
511: \lambda \left( \begin{array}{c}
512: \sigma_{-}(u_1) \\
513: \sigma_{-}(u_2)
514: \end{array} \right), \quad
515: \vec{\sigma}_{v} =
516: \lambda \left( \begin{array}{c}
517: \sigma(v_1) \\
518: \sigma(v_2)
519: \end{array} \right)\;.
520: \label{Sigmauv}
521: \end{eqnarray}
522: In eqs.~(\ref{Sigmauv}) we have used the symbolic notation
523: \begin{eqnarray}
524: \sigma_{-}(u_i) &\equiv& \int_{0}^{t} ds u_i(s) \sigma_{-}(s) \;, \nonumber \\
525: \sigma(v_i) &\equiv& \int_{0}^{t} ds v_i(s) [ \sigma_{+}(s)/2 \nonumber \\
526: &&+ \int_{0}^{t} ds' i \alpha^R(s-s') \sigma_{-}(g(s'))/M]\;,
527: \end{eqnarray}
528: for $i=1,2$.
529:
530: Having derived the expression for $J_C^{\sigma \sigma'}$ in
531: eq.~(\ref{Jrsimple}), we may use it to obtain the partial Wigner
532: transform of the reduced density matrix. The density matrix in the
533: partial Wigner representation \cite{Wigner} over the $C$ subsystem
534: degrees of freedom is defined by
535: \begin{eqnarray}
536: \rho_{rW}(\sigma,\sigma',R,P,t)&=& \frac{1}{2\pi \hbar} \int dr
537: e^{-iPr/\hbar} \nonumber \\
538: && \times \rho_r(\sigma,\sigma',R+r/2,R-r/2,t).
539: \end{eqnarray}
540: Wigner transforming eq.~(\ref{DMACf}) and using this definition we
541: obtain
542: \begin{eqnarray}
543: \rho_{rW}(\sigma,\sigma',R,P,t)&=&\int d\sigma_0 d\sigma_0' \;
544: \int^{(\sigma \sigma' )}_{(\sigma_0 \sigma'_0 )} {\mathcal D}
545: \sigma {\mathcal D} \sigma' K[\sigma] K^{*}[\sigma']\nonumber \\
546: && \int dR_0 dP_0 \; K_{C}^{\sigma \sigma'}(R, P;t \mid R_0, P_0;
547: 0)\nonumber \\
548: &&\times \rho_{rW(}\sigma_0,\sigma'_0,R_0,P_0,0) \;,
549: \label{Wrdensity}
550: \end{eqnarray}
551: where the kernel $K_{C}^{\sigma \sigma'}(R, P;t \mid R_0,P_0;0)$
552: is defined by
553: \begin{eqnarray}
554: K_{C}^{\sigma \sigma'}(R, P;t \mid R_0,P_0;0)&=& \frac{1}{2 \pi
555: \hbar} \int dr dr_0\; e^{-i(Pr-P_0r_0)/\hbar} \nonumber \\
556: && \times J_{C}^{\sigma \sigma'}(R, r;t \mid R_0, r_0; 0)\;.
557: \end{eqnarray}
558:
559: Inserting the expression for $J_{C}^{\sigma \sigma'}$ in eq.~(\ref{Jrsimple}), the
560: time evolution kernel $K_{C}^{\sigma \sigma'}$ for the Wigner function is given by
561: \begin{eqnarray}
562: K_{C}^{\sigma \sigma'}(R_t, P_t; t &\mid& R_0, P_0; 0) \nonumber \\
563: &=& \frac{N(t)}{2 \pi \hbar} \int dr dr_0 \;
564: e^{i (-P_t r_t + P_0 r_0+ {\mathcal L})/\hbar} \nonumber \\
565: &=& N_W(t) \exp\Big[ - \delta \vec{X}_{W}^{T}
566: {\bf Q}^{-1} \delta \vec{X}_{W} \nonumber \\
567: &&+ \frac{i}{\hbar} \vec{\sigma}_{u}^{T} \vec{R}
568: + \frac{i}{\hbar}{\cal S}^{(2)}[\sigma_{\pm}] \Big],
569: \label{Ksigma}
570: \end{eqnarray}
571: where $N_W(t)=N(t)/2 \hbar \sqrt{|{\bf a}|}$ and $|{\bf a}|$ is the
572: determinant of ${\bf a}$. The vector $\delta \vec{X}_W=
573: \vec{X}_W -\langle \vec{X} \rangle$, with
574: \begin{eqnarray}
575: \vec{X}_{W} =
576: \left( \begin{array}{c}
577: R_t \\
578: P_t - \lambda \sigma(v_2)
579: \end{array} \right) \;,
580: \label{RWvec}
581: \end{eqnarray}
582: and
583: \begin{eqnarray}
584: \langle \vec{X} \rangle=
585: \left( \begin{array}{c}
586: \langle R \rangle \\
587: \langle P \rangle
588: \end{array} \right)
589: =\frac{-1}{u_{21}}
590: \left( \begin{array}{cc}
591: u_{11} & 1 \\
592: |{\bf u}| & u_{22}
593: \end{array} \right)
594: \left( \begin{array}{c}
595: R_0 \\
596: P_0 + \lambda \sigma(v_1)
597: \end{array} \right).
598: \label{RC}
599: \end{eqnarray}
600: The matrix ${\bf Q}$ is given by
601: \begin{eqnarray}
602: &&{\bf Q} = \frac{4\hbar}{u_{21}^2} \times \nonumber \\
603: &&
604: \left( \begin{array}{cc}
605: a_{11} & a_{12} u_{21} - a_{11} u_{22} \\
606: a_{12} u_{21} - a_{11} u_{22} &
607: a_{11} u_{22}^2 - 2 a_{12} u_{21} u_{22} + a_{22} u_{21}^2
608: \end{array} \right). \nonumber \\
609: \label{QW2}
610: \end{eqnarray}
611: In the next section we shall need $Q_{11}$ and the
612: explicit expression for this matrix element is
613: \begin{eqnarray}
614: Q_{11}(t)&=& \frac{2 \hbar}{\pi M^2} \int_{0}^{\infty} d \omega
615: \coth \left( \frac{\beta \hbar \omega}{2} \right)
616: J_B(\omega) \Delta(\omega,t), \nonumber \\
617: \label{Q11}
618: \end{eqnarray}
619: where $\Delta(\omega,t)$ is given by
620: \begin{eqnarray}
621: \Delta(\omega,t) &\equiv& \int_{0}^{t} ds
622: \int_{0}^{t} ds' g_{+}(t-s) \cos \omega (s-s') g_{+}(t-s').
623: \nonumber \\
624: \label{Deltaomega}
625: \end{eqnarray}
626: These results provide the tools needed to analyze the
627: conditions under which passage to the quantum-classical limit
628: may be carried out. In the next sections we present explicit
629: results for both the Ohmic and super-Ohmic bath spectral
630: densities.
631:
632: \section{Thermalization of quantum brownian motion} \label{sec:QBM}
633:
634: Whether a linear or nonlinear system interacting with a bath will eventually reach thermal equilibrium or not is one of the fundamental issues
635: in statistical mechanics. For a linear quantum brownian motion,
636: from a general class of initial conditions, the state of C is
637: considered to reach asymptotically the gaussian
638: form\cite{Tegmark}.
639: In this section, first we will show that the
640: C subsystem evolution starting from an arbitrary initial condition
641: possesses an asymptotic limit. We then study the influence of such
642: a thermalized quantum system on A. For our analysis, we shall need
643: the results for a system composed of the particle $C$ and the
644: thermal bath $B$ in the absence of $A$. In this case
645: eq.~(\ref{Wrdensity}) takes the form,
646: \begin{eqnarray}
647: \rho_{rW}(R_t,P_t,t)&=&
648: \int dR_0 dP_0 \; K_{C}(R_t, P_t;t \mid R_0, P_0; 0)\nonumber \\
649: &&\times \rho_{rW}(R_0,P_0,0) \;,
650: \label{WrdensityC}
651: \end{eqnarray}
652: In the absence of subsystem $A$, from eq.~(\ref{Ksigma}) the time
653: evolution kernel for the Wigner function is
654: \begin{eqnarray}
655: K_{C}(R_t, P_t; t \mid R_0, P_0; 0) &=& N_W(t)
656: \exp[ - \delta \vec{X}_{W}^{T} {\bf Q}^{-1} \delta
657: \vec{X}_{W}], \nonumber \\
658: \label{Knosigma}
659: \end{eqnarray}
660: where $\vec{X}_{W}^T = (R, P)$ and
661: \begin{eqnarray}
662: \langle \vec{X} \rangle =
663: \left( \begin{array}{c}
664: \langle R \rangle \\
665: \langle P \rangle
666: \end{array} \right)
667: =\frac{-1}{u_{21}}
668: \left( \begin{array}{cc}
669: u_{11} & 1 \\
670: |{\bf u}| & u_{22}
671: \end{array} \right)
672: \left( \begin{array}{c}
673: R_0 \\
674: P_0
675: \end{array} \right).
676: \label{RC2}
677: \end{eqnarray}
678:
679: For a harmonic potential, as a result of the Ehrenfest theorem,
680: the center phase space coordinate $\langle \vec{X} \rangle$ of
681: this distribution function follows a classical trajectory and will
682: decay because of energy dissipation into the bath.
683: Using the exact expressions for the matrix elements of ${\bf u}$,
684: obtained from the formulas given in Appendix A,
685: one may show that $\langle \vec{X} \rangle$ vanishes in the long
686: time limit for both Ohmic and super-Ohmic baths.
687: As a result, in both cases, for times long compared to the
688: characteristic relaxation time for this decay, the time evolution
689: kernel reduces to
690: \begin{eqnarray}
691: K_{C}(R_t, P_t; t \mid R_0, P_0; 0) &\to & N_W(t)
692: \exp[ - \vec{X}_{W}^{T} {\bf Q}^{-1} \vec{X}_{W} ],
693: \nonumber \\
694: &\equiv& K_{C}(R_t, P_t; t)
695: \label{Knosigmaasymptotic}
696: \end{eqnarray}
697: which is independent of $R_0$ and $P_0$. Consequently,
698: eq.~(\ref{WrdensityC}) reduces to
699: \begin{eqnarray}
700: \rho_{rW}(R_t,P_t,t)= K_{C}(R_t, P_t;t);
701: \label{WrdensityD}
702: \end{eqnarray}
703: thus, for an arbitrary initial condition, the Wigner transformed
704: reduced density matrix is uniquely determined by the time evolution
705: kernel $K_C$. Furthermore, the off-diagonal elements of ${\bf
706: Q}$ vanish in this limit and we obtain a Gaussian form
707: for the density matrix whose widths are uniquely specified and
708: given by $\langle (\Delta R)^2 \rangle=Q_{11}/2$ and
709: $ \langle (\Delta P)^2 \rangle=Q_{22}/2$.
710: The width $\langle (\Delta R)^2 (\infty) \rangle$ can be obtained from
711: eq.~(\ref{Q11}) as
712: \begin{eqnarray}
713: \langle (\Delta R)^2 (\infty)\rangle&=& \frac{ \hbar}{\pi M^2}
714: \int_{0}^{\infty} d \omega
715: \coth \left( \frac{\beta \hbar \omega}{2} \right)
716: J_B(\omega) \Delta(\omega,\infty), \nonumber \\
717: \end{eqnarray}
718: We show in Appendix A for the Ohmic case and for the weak
719: coupling limit of the super-Ohmic case, that
720: $J_B(\omega) \Delta(\omega,\infty)=M^2 \chi''_C (\omega)$, where
721: \begin{eqnarray}
722: \chi''_C (\omega) = \frac{1}{M} \frac{\omega \gamma'(\omega)}
723: { (\omega_{r}^2 - \omega^2 + \omega \gamma''(\omega))^2
724: + \omega^2 \gamma'^{2}(\omega) },
725: \label{chi}
726: \end{eqnarray}
727: is the dynamical susceptibility for $C$.
728: For the spectral density defined in eq.~(\ref{SpectralDensity}),
729: the explicit form of the frequency dependent dissipation coefficient
730: $\gamma(\omega)$ is given by
731: \begin{eqnarray}
732: M \gamma(\omega) &\rightarrow&
733: \left\{
734: \begin{array}{llll}
735: \eta
736: \hspace{1cm} &(\nu=1)&\\
737: \eta \omega^2 \hspace{1cm}
738: &(\nu=3)&
739: \end{array}
740: \right.
741: \label{gammaexact2}
742: \end{eqnarray}
743: in the limit of a large cutoff parameter $\Lambda$, where
744: $\gamma''(\omega)$ was absorbed in the mass renormalization for the
745: $\nu=3$ case. Consequently, we may write
746: $\langle (\Delta R)^2 (\infty)\rangle$ as
747: \begin{eqnarray}
748: \langle \Delta^2 R(\infty) \rangle
749: = \frac{\hbar}{\pi}
750: \int_{0}^{\infty} d \omega \coth \left( \frac{\beta \hbar \omega}{2} \right)
751: \chi''_C (\omega).
752: \label{FDR}
753: \end{eqnarray}
754: Equation~(\ref{FDR}) is a statement of the fluctuation-dissipation
755: relation \cite{Kubo57}. The form in eq.~(\ref{FDR}) suggests that
756: the combined system $CB$ is now in thermal equilibrium with
757: spectrum specified by $\chi''_C (\omega)$ but the system $C$
758: itself is not in equilibrium. The fluctuation-dissipation relation
759: is violated at finite $t$ while the system is still far from
760: equilibrium and holds only asymptotically.
761: This result generalizes the result previously
762: obtained by Caldeira and Leggett
763: for an Ohmic bath\cite{CaldeiraLeggett84}.
764:
765: When the $CB$ system is in thermal equilibrium, we show in
766: Appendix~\ref{app:LRT} using linear response theory that
767: $J_{CB}(\omega) =
768: \lambda^2 \chi''_C (\omega)$: the effective spectral density is
769: proportional to the dynamical susceptibility of the $C$ system.
770: If we use the relation
771: \begin{eqnarray}
772: J_{B}(\omega) = M \omega \gamma'(\omega),
773: \label{ICBtogamma}
774: \end{eqnarray}
775: for the bath spectral density assumed in eq.~(\ref{SpectralDensity}), we see
776: that the effective spectral density in the combined system $CB$ can
777: be written as
778: \begin{eqnarray}
779: J_{CB}(\omega) = \frac{\lambda^2 J_{B}(\omega) }
780: { M^2 (\omega_{r}^2 - \omega^2 + \omega \gamma''(\omega))^2
781: + J_{B}^2(\omega) }.
782: \label{EffectiveSpectralDensity}
783: \end{eqnarray}
784: This relation can also be obtained by solving the classical
785: equations of motion as suggested earlier \cite{FeyVer63,Leggett84,GOA85}
786: owing to the linear coupling assumed between the $C$ system and the bath.
787:
788: \section{Emergence of quantum-classical dynamics via decoherence}
789: \label{sec:emergence}
790:
791: The investigation of the emergence of quantum-classical dynamics
792: presented here will be restricted to cases where the dynamics of subsystem
793: $A$ occurs on time scales which are very long compared to those
794: that characterize subsystem $C$. We suppose that subsystem $A$ is a
795: two-level system with $\sigma=\pm1$ and focus on the extreme
796: non-adiabatic regime where the population dynamics of subsystem $A$ is
797: essentially frozen on the time scales of interest so that
798: $\rho_{A}(-1,-1,t)=\rho_{A}(-1,-1,0),\rho_{A}(1,1,t)=\rho_{A}(1,1,0)$.
799: We further assume that subsystems $A$ and $C$ are decoupled initially
800: and use the factorized initial condition,
801: \begin{eqnarray}
802: \hat{\rho}_{AC}(x_0, x'_0, 0) = \hat{\rho}_A(0) \otimes \rho_{C}(x_0, x'_0, 0).
803: \label{DMACinitial}
804: \end{eqnarray}
805: This is a reasonable assumption if they are weakly coupled.
806:
807: \subsection{Decoherence in $A$ interacting with an equilibrium $CB$ bath}
808: If the interaction between the $A$ system and the combined system
809: $CB$ is turned on after an equilibrium state is reached, the
810: effect of $CB$ on $A$ is
811: given simply by the harmonic bath directly coupled to the $A$
812: subsystem but with its spectrum characterized by the effective spectral
813: density. Under these conditions, the off-diagonal part of the density matrix
814: can be obtained from eq.~(\ref{DMACf}) by dropping the dependence on
815: $(x,x')$, omitting the integrals over spin variables and by making the
816: variable replacements $x \rightarrow \sigma$ and $J_B \rightarrow J_{CB}$
817: in eq.~(\ref{IFaction}) for the influence functional. We find,
818: \begin{eqnarray}
819: \rho_A(-1,1, t)&=& e^{-D_{A}^{(0)}(t)} \rho_A(-1,1, 0),
820: \label{ODDM}
821: \end{eqnarray}
822: where $D_{A}^{(0)}(t)$ is a decoherence factor for A and is defined as
823: \begin{eqnarray}
824: D_{A}^{(0)}(t) &=& \frac{4}{\pi \hbar}
825: \int_{0}^{\infty} d \omega J_{CB}(\omega)
826: \coth \left( \frac{\beta \hbar \omega}{2} \right)
827: \nonumber \\ &&\times \left( \frac{1 - \cos(\omega t)}{\omega^2} \right)\;.
828: \label{ODDM2}
829: \end{eqnarray}
830: Thus, as mentioned above, the dynamics of the composite system is equivalent
831: to that of a two-level spin-boson system coupled to a harmonic bath,
832: with an effective spectral density.
833:
834: \begin{figure}[htbp]
835: \epsfxsize=.45\textwidth \epsfbox{Fig1a.eps}
836: \epsfxsize=.45\textwidth \epsfbox{Fig1b.eps}
837: \caption{ The temporal behavior of the off-diagonal element of the
838: reduced density matrix $\rho_A(-1,1,t)$ for subsystem $A$ is plotted
839: in panel (a) versus time. The initial value $\rho_A(-1,1,0)=1$. Parameters are:
840: $\beta=0.66$, $\Lambda=500$, $\omega_{r}=1$, $\gamma=0.3$,
841: $M=1$ and $\hbar=1$. The same quantity is shown in panel (b) using a logarithmic
842: time scale. In both figures the solid line
843: corresponds to $A$ under the
844: influence of $CB$
845: with the effective Ohmic spectral density,
846: while the dashed line shows the influence of Ohmic bath $B$ without $C$.
847: (1) $\lambda^2=0.1$ and (2) $\lambda^2=0.3$.}
848: \label{fig1}
849: \end{figure}
850: Using these results we may examine the coherence in $A$ under the influence
851: of the effective bath $CB$ and the coherence of $A$ under the influence of
852: the bath $B$ separately. In Figs.~1(a) and (b), the off-diagonal part
853: of the density matrix of subsystem $A$ is plotted for an Ohmic environment.
854:
855: The solid line corresponds to our system while the dashed line
856: is the ordinary spin boson model without $C$.
857: The time evolution of $\rho_A(-1,1, t)$ for the ordinary ordinary spin
858: boson model is characterized by (1) an initial
859: period where subsystem $A$ system has not yet felt the
860: existence of the environment and its coherence is maintained; (2)
861: a quantum regime for $t> 1/\Lambda$ where the system begins to interact
862: with the vacuum fluctuations of the environment and, finally, (3) a thermal
863: regime for $t> \hbar \beta$ where the effects of thermal fluctuations
864: on subsystem $A$ have set in. For $1/\Lambda < t < \hbar \beta$,
865: only vacuum fluctuations interact with the system.
866:
867: By contrast, the existence of the intermediate subsystem $C$ changes
868: the evolution of $\rho_A(-1,1, t)$ significantly. Subsystem $C$ is
869: characterized by its harmonic oscillator frequency and mass. Fluctuations
870: originating in the bath $B$ are modulated by $C$ through these parameters.
871: For a large mass $M$, the effective modes of the combined $CB$ system are
872: concentrated in the neighborhood of $\omega_{r}$ and the decoherence
873: behavior of subsystem $A$ is governed by these modes.
874: Therefore, after the initial period, decoherence will begin to occur at
875: $t \sim 1/ \omega_r$.
876:
877: Although some systems are approximately characterized by an Ohmic
878: spectral density, generic systems will have non-Ohmic spectral densities.
879: The density of states typically varies as $\omega^{\nu}$ with
880: $\nu$ depending on the spatial dimension $D$. We consider the 3D
881: super-Ohmic case with $\nu=3$. Such an environment is
882: relevant for the study of polarons\cite{LCDFGZ87},
883: macroscopic magnetization tunneling in crystals\cite{GargKim89},
884: and radiation damping of atoms
885: \cite{BaroneCaldeira91}. A super-Ohmic environment
886: affects subsystem $A$ evolution on short time scales more significantly
887: than an Ohmic environment. A sub-Ohmic environment induces non-trivial
888: long time behavior. Results for the super-Ohmic case with $\nu=3$ are
889: shown in Figs.~\ref{fig2} (a) and (b).
890:
891: For the super-Ohmic high dimensional environment, the vacuum fluctuations
892: play a more significant role than for an Ohmic environment, reflecting the
893: fact that there are larger number of high frequency modes in the higher
894: dimensional environment. The same effect is also responsible
895: for the large difference in the behavior of
896: correlation functions of the system in quantum and classical baths
897: when the bath is super-Ohmic\cite{EgoEveSki99}.
898: The dynamics of $A$ in the presence of the intermediate subsystem
899: $C$ is worth noting. For the present choice of parameters,
900: $\rho_A(-1,1, t)$ asymptotically reaches a non-zero value indicating
901: that quantum coherence in the $A$ system will never be lost completely.
902: The origin of this behavior can be understood
903: in terms of the time-dependence of the diffusion coefficient
904: (See Ref.~\cite{QBM1}).
905: System $C$ initially executes brownian motion with
906: a time-dependent diffusion coefficient
907: as a result of the non-Markovian time evolution of the reduced density matrix.
908: This time-dependent diffusion coefficient of $C$, which also determines
909: the decoherence rate,
910: exhibits a rapid increase
911: at early times $t > 1/\Lambda$ and then asymptotically vanishes.
912: This initial increase can be large enough to wash away the
913: quantum coherence of $C$ directly coupled to the bath.
914: Owing to the modulation effect from $C$, these high frequency modes are
915: filtered out and do not directly affect $A$ if $\omega_{r}<<\Lambda$.
916: The late time value of the diffusion coefficient is too small to eliminate
917: the quantum coherence from subsystem $A$ completely.
918: Hence, the non-Markonvian nature of the density matrix evolution
919: for the super-Ohmic case is responsible for the significant difference
920: in the dynamics of coherence between the $A$ and $C$ subsystems.
921:
922:
923: \begin{figure}[htbp]
924: \epsfxsize=.45\textwidth \epsfbox{Fig2a.eps}
925: \epsfxsize=.45\textwidth \epsfbox{Fig2b.eps}
926: \caption{ Plot of $\rho_A(-1,1,t)$ versus time (a) and versus $\log(t)$ (b)
927: for a super-Ohmic environment with $\nu=3$. Parameters and labeling
928: are the same as in \protect Fig.~\ref{fig1}. The dashed line in panel
929: (a) for super-Ohmic bath $B$ without $C$ overlaps the vertical axis.
930: The behavior is visible in panel (b) using the logarithmic time scale. }
931: \label{fig2}
932: \end{figure}
933:
934: \subsection{Decoherence in $A$ for arbitrary $C$ initial conditions}
935:
936: Thus far we have considered only the asymptotic limit where $CB$ has
937: reached equilibrium and the fluctuation-dissipation relation holds.
938: It is interesting to consider the coherence of $A$ under the
939: influence of intrinsic dynamics of $C$ starting from arbitrary $C$
940: initial conditions. As an example of this situation, we assume that $C$
941: is initially in a gaussian coherent state given by the wave function,
942: \begin{eqnarray}
943: \psi_{\bar{x} \bar{p}}(x) &=& \frac{1}{(\pi \epsilon^2)^{1/4}} \exp \left[
944: -\frac{(x-\bar{x})^2}{2 \epsilon^2} + i \frac{\bar{p} x}{\hbar} \right],
945: \label{WF0}
946: \end{eqnarray}
947: with width $\epsilon=1/\sqrt{\omega_r}$ where $\bar{x}$ and $\bar{p}$
948: are parameters. The $C$ subsystem density matrix at the initial time is then
949: given by
950: \begin{eqnarray}
951: \rho_{C}(x_0, x'_0, 0) &=&\frac{1}{\sqrt{\pi \epsilon^2}}
952: \exp \left[ -\frac{(R_0-\bar{x})^2}{\epsilon^2} -\frac{r_0^2}{4 \epsilon^2}
953: +i \frac{\bar{p} r_0}{\hbar} \right].
954: \nonumber \\
955: \label{2IDMx}
956: \end{eqnarray}
957: We may now use this expression for $\rho_{C}(x_0, x'_0, 0)$ in the
958: factorized initial condition for $AC$ given in eq.~(\ref{DMACinitial}).
959: The off-diagonal element of reduced density matrix for subsystem $A$
960: in the non-adiabatic limit can then be determined from
961: eqs.~(\ref{DMACf}) and (\ref{Jrpathint}) for
962: $\rho_r(\sigma, \sigma', x, x', t)$ for $AC$ by integrating over the
963: $C$ variables. The result is,
964: \begin{eqnarray}
965: \rho_A(-1,1,t)&=&\int dx dx'\; \Big[\int dx_0 dx'_0\; \nonumber \\
966: && \times J^{-1,1}_C(x,x',t|x_0,x'_0,0)\rho_{C}(x_0, x'_0, 0)\Big] \nonumber \\
967: && \times \rho_A(-1,1,0)\;.
968: \end{eqnarray}
969: Changing integration variables $(x,x') \to (R,r)$
970: and $(x_0,x_0') \to (R_0,r_0)$ and using eq.~(\ref{Jrsimple}) for
971: $J^{-1,1}_C(R,r,t|R_0,r_0,0)$ we may write this as
972: \begin{eqnarray}
973: \rho_A(-1,1,t)&=&\int dR dr\; \Big[\int dR_0 dr_0\; \nonumber \\
974: && \times N(t) \exp\Big\{\frac{i}{\hbar} {\mathcal L} \Big\}
975: \rho_{C}(R_0, r_0, 0)\Big] \nonumber \\
976: && \times \rho_A(-1,1,0)\;.
977: \end{eqnarray}
978: Making use of the expression for ${\mathcal L}$ evaluated for
979: $\sigma=-1$ and $\sigma'=1$ and carrying out the integrations we find,
980: \begin{eqnarray}
981: \rho_A(-1,1, t)&=& \rho_A(-1,1, 0) e^{-D_A(t)} \nonumber \\
982: && \times
983: \exp \left[\frac{i}{\hbar} ( f_{R}(t) \bar{x} + f_{P}(t) \bar{p} ) \right]
984: \nonumber \\
985: && \times
986: \exp \left[-\frac{1}{\hbar} ( f_{R}^2(t)/ M \omega_r + f_{P}^2(t) M \omega_r )
987: \right]\:, \nonumber \\
988: \label{DMAnonadiabatic_offdiag}
989: \end{eqnarray}
990: %
991: where
992: \begin{eqnarray}
993: f_{R}(t) = \lambda \int_{0}^{t} ds \frac{u_{11}(s)}{u_{21}(s)},
994: \quad f_{P}(t) = \lambda \int_{0}^{t} ds \frac{1}{u_{21}(s)}
995: \label{AB}
996: \end{eqnarray}
997: and
998: \begin{eqnarray}
999: D_A(t) = \frac{2 \lambda^2}{\pi \hbar M^2}
1000: \int_{0}^{\infty} d \omega J_{B}(\omega)
1001: \coth \left( \frac{\beta \hbar \omega}{2} \right) \Delta_A(\omega,t).
1002: \nonumber \\
1003: \label{Decfunctionfull}
1004: \end{eqnarray}
1005: The explicit expression for $\Delta_A(\omega,t)$ can be computed
1006: from the solutions presented in Appendix~A by a lengthy but
1007: straightforward calculation. We find,
1008: \begin{eqnarray}
1009: &&\Delta_A(\omega,t) = \frac{1}{(\Omega_{r}^2 - \omega^2)^2 + 4
1010: \Gamma^2 \omega^2 } \left[ 2 \frac{1 - \cos(\omega t)}{\omega^2}
1011: \right. \nonumber \\
1012: && \quad +\frac{2}{\Omega \Omega_{r}^2} \frac{1 -
1013: \cos(\omega t)}{\omega}\nonumber \\
1014: & &\quad \times
1015: \left( \omega \Omega (1-e^{-\Gamma t} \cos(\Omega t) )
1016: - \omega \Gamma e^{-\Gamma t} \sin(\Omega t)\right)\nonumber \\
1017: &&\quad+ \frac{2}{\Omega \Omega_{r}^2} \frac{\sin(\omega
1018: t)}{\omega}
1019: \nonumber \\
1020: & &\quad \times \left( 2 \Gamma \Omega (1-e^{-\Gamma t}
1021: \cos(\Omega t) ) + (\Omega^2 - \Gamma^2) e^{-\Gamma t} \sin(\Omega
1022: t) \right)
1023: \nonumber \\
1024: && \quad +\frac{1}{\Omega^2 \Omega_{r}^{4}} \left\{ \left(
1025: \omega^2 + 4 \Gamma^2 \right) \Omega^2 ( 1- \cos(\Omega t) e^{-
1026: \Gamma t} )^2
1027: \right. \nonumber \\
1028: && \quad -
1029: 2 \Gamma \Omega \left( \omega^2 + 2 \Gamma^2 - 2 \Omega^2 \right)
1030: \nonumber \\
1031: && \quad \times ( 1 - \cos(\Omega t) e^{- \Gamma t} )\sin(\Omega
1032: t)e^{- \Gamma
1033: t}\nonumber \\
1034: && \quad +\left. \left.
1035: \left( \omega^2 \Gamma^2 + (\Omega^2 - \Gamma^2)^2 \right)
1036: \sin^2(\Omega t) e^{- 2 \Gamma t}\right\}\right].
1037: \label{DeltaAfullOhmic}
1038: \end{eqnarray}
1039: Equations (\ref{DMAnonadiabatic_offdiag})-(\ref{DeltaAfullOhmic}) are
1040: our main results. The first term in
1041: eq.~(\ref{DMAnonadiabatic_offdiag}) is the initial condition for
1042: $A$, the second term accounts for the decoherence arising from
1043: thermal fluctuations of the bath mediated by $C$. The third term
1044: arises from the initial condition for $C$ and the last term is
1045: responsible for the decoherence of $A$ due to the averaged damped
1046: oscillatory motion of $C$.
1047:
1048:
1049: The modulus of the off-diagonal reduced density matrix element for $A$
1050: is shown in Figs.~\ref{fig3}(a) and (b) as a function of time.
1051: The solid line shows the second term eq.~(\ref{DMAnonadiabatic_offdiag}),
1052: $S \equiv e^{-D_A(t)}$,
1053: which gives the decoherence due to the quantum-back-reaction-induced
1054: self-interaction of subsystem $A$.
1055: The dashed line shows the last term,
1056: $L \equiv \exp \left[-\frac{1}{\hbar} ( f_{R}^2(t)/ M \omega_r + f_{P}^2(t) M \omega_r) \right]$,
1057: the decoherence
1058: due to the motion of $C$. The latter effect
1059: is not strong enough to eliminate coherence of $A$ completely.
1060: Since, initially, the CB is far from equilibrium,
1061: those modes that affect A are not necessarily
1062: near resonant modes around $\omega_r$.
1063: Consequently, the decoherence of A is more rapid
1064: in this case than in the case of evolution
1065: from the thermal equilibrium initial condition
1066: studied in Sec. IV. A.
1067: This tendency is more evident in a super-Ohmic bath than in an Ohmic bath.
1068: With generic initial conditions for $C$,
1069: A is under the influence of the nonequilibrium bath $CB$,
1070: which is no longer equivalent to the effective thermal bath with the
1071: spectral density $J_{CB}$.
1072: \begin{figure}[pp]
1073: \epsfxsize=.45\textwidth \epsfbox{Fig3a.eps}
1074: \epsfxsize=.45\textwidth \epsfbox{Fig3b.eps}
1075: \caption{
1076: Plot of contribution to the modulus of $\rho_A(-1,1,t)$ versus time for an Ohmic bath
1077: (Fig.~\ref{fig3}(a)) and for a super Ohmic bath (Fig.~\ref{fig3}(b)).
1078: Subsystem $C$ is in a coherent state initially.
1079: In both figures the solid line corresponds to the second term S,
1080: while the dashed line corresponds to the last term L
1081: in the right hand side in eq.(\ref{DMAnonadiabatic_offdiag}).
1082: $\lambda^2=0.1$ for S1,L1 and $\lambda^2=0.3$ for S2,L2.
1083: Parameters values are the same as in \protect Fig.~\ref{fig1}.
1084: }
1085: \label{fig3}
1086: \end{figure}
1087:
1088: Now let us consider the situation where the decoherence of $C$ is
1089: fast and that of $A$ is slow. This may occur at high temperatures
1090: and for weak coupling between $A$ and $C$. In such a case, the
1091: back reaction from $A$ on $C$ may be neglected. In this
1092: circumstance, we may consider the decoherence of $C$ in the
1093: absence of $A$. Quantum brownian motion and decoherence of a
1094: damped harmonic oscillator has been studied previously
1095: \cite{CaldeiraLeggett84,GSI88,UnruhZurek89,QBM1,QBM2}. Using the
1096: result in eq.~(3.9) of Ref.~\cite{QBM1}, the reduced density matrix may
1097: be approximated by,
1098: \begin{eqnarray}
1099: \rho_{r}(\sigma,\sigma',x,x',t) \sim e^{-(x-x')^2 D_C(t)}
1100: \rho_{r}(\sigma,\sigma',x,x',0),
1101: \label{MasterEquationDecC}
1102: \end{eqnarray}
1103: where $D_C(t) =\int_{0}^{t} ds \int_{0}^{s} ds'\alpha^R(s') \cos
1104: \omega_r s'$ is the decoherence factor for $C$.
1105: \begin{figure}[hh]
1106: \begin{center}
1107: \epsfxsize=.45\textwidth \epsfbox{Fig4a.eps}
1108: \epsfxsize=.45\textwidth \epsfbox{Fig4b.eps}
1109: \end{center}
1110: \caption{
1111: Plot of the decoherence factor versus time. The solid line corresponds
1112: to the logarithm of the modulus of
1113: the density matrix $\rho_A$ in eq.(\ref{DMAnonadiabatic_offdiag})
1114: with the inverse sign,
1115: while the dashed line is $D_C$.
1116: System parameters are:
1117: $x-x'=1$, $\omega_{r}=1$, $\gamma=0.3$, $M=1$, $\hbar=1$.
1118: Panel (a) is for the Ohmic bath with $\beta =0.002$, $\Lambda=500$.
1119: Panel (b) is for the $\nu=3$ super-Ohmic bath with $\beta =0.02$,
1120: $\Lambda=20$.}
1121: \label{fig4}
1122: \end{figure}
1123: In Fig.~\ref{fig4}, the
1124: decoherence factors for $A$ and $C$ are plotted.
1125: For an Ohmic bath (Fig.~\ref{fig4}(a)), initially,
1126: the quantum coherence is
1127: lost faster $C$ than in $A$; hence, the $A$ subsystem behaves more
1128: quantum mechanically than the $C$ subsystem during this initial
1129: period. Figure~\ref{fig4}(a) shows that there is a crossover and
1130: for longer times, $t>1/\omega_r$, subsystem $A$ experiences
1131: stronger decoherence than $C$. The decoherence factor for $C$
1132: varies linearly with $t$ indicating that the dynamics of $C$
1133: asymptotically approaches Markovian evolution in the high temperature
1134: Ohmic bath.
1135: For a super-Ohmic bath (Fig.~\ref{fig4}(b)), a substantial portion
1136: of the quantum coherence in $C$ is lost during the initial period
1137: due to large fluctuations coming from the bath, while $A$ retains
1138: quantum coherence for a long time. The dynamics of the
1139: coupled system can be approximated by mixed quantum-classical
1140: evolution for long times after the short transient period.
1141:
1142:
1143: \section{Conclusion} \label{sec:conc}
1144:
1145: An appreciation of the conditions under which a quantum mechanical
1146: system may be approximated as a quantum-classical system is
1147: essential in many applications to the dynamics of many-body
1148: systems. Adopting an open quantum system point of view has
1149: provided a natural way to explore this issue based on the
1150: decoherence of the superposition of quantum states into a
1151: statistical mixture under the influence of the environment
1152: \cite{EID,Dec96}. Indeed, considerations of decoherence have played an
1153: important role in discussions of schemes for the simulation of
1154: condensed phase systems. \cite{decoh}
1155:
1156: Using influence functional methods, we have shown for our simple
1157: model system that the decoherence time scales that characterize
1158: the $A$ and $C$ subsystems can differ significantly. In
1159: particular, in the limit of nonadiabatic dynamics we have
1160: identified the following three regimes: (1) the full quantum
1161: regime where both the $A$ and $C$ subsystems behave quantum
1162: mechanically; (2) the quantum-classical regime where
1163: subsystem $A$ maintains coherence owing to its indirect coupling
1164: to the bath, while $C$ has lost its coherence and behaves
1165: effectively classically; (3) the classical regime where the
1166: quantum coherence of both the $A$ and $C$ subsystems has been lost
1167: and the composite $AC$ subsystem exhibits effectively classical
1168: dynamics.
1169: The different roles that the environment plays
1170: in its effect on two subsystems is responsible for the separation
1171: of decoherence time scales.
1172:
1173:
1174: We also saw that a different choice of initial conditions will
1175: modify above picture. When the initial condition for $C$ is chosen
1176: to be a pure gaussian coherent state, $CB$ can no longer be
1177: considered to be a thermal bath and the coherence of $A$ will be
1178: lost in a shorter time scale
1179: than times given by the inverse characteristic
1180: frequency of $C$.
1181:
1182:
1183:
1184: Quantum/classical correspondence in nonlinear systems, in
1185: particular, in chaotic systems, has many nontrivial
1186: features\cite{Chi}. The quantum open system approach adopted in this
1187: paper can also shed light on this problem\cite{ShiokawaHu95,PazZurek00}.
1188: While the model system studied here is very simple and the extreme
1189: non-adiabatic regime is only one limiting case to consider, our
1190: results nevertheless have provided some insight into the emergence
1191: of quantum-classical dynamics and should be useful in the study of
1192: the quantum dynamics of more complex systems.
1193:
1194: \section*{Acknowledgments}
1195: This work was supported in part by a grant from the Natural
1196: Sciences and Engineering Research Council of Canada.
1197: Acknowledgment is made to the donors of The Petroleum Research
1198: Fund, administered by the ACS, for partial support of this
1199: research.
1200:
1201: \appendix
1202: \renewcommand{\theequation}{\thesection\arabic{equation}}
1203: \setcounter{equation}{0}
1204:
1205: \section{Solutions of equations of motion} \label{app:Formulas}
1206:
1207: In this Appendix we give some details of the solutions of the
1208: equations of motion discussed in Sec.~\ref{EL} which are used in
1209: the calculations presented in Sec.~\ref{sec:emergence}. We require
1210: the functions $u_{1,2}(s)$ (and $v_{1,2}(s)$) which are the
1211: solutions of the homogeneous parts of the Euler-Lagrange equations
1212: (\ref{EL1}) and (\ref{EL2}). These functions can be found if
1213: $g_{\pm}(s)$, which are also solutions of the homogeneous parts
1214: eqs.~(\ref{EL1}) and (\ref{EL2}), are known, since
1215: \begin{eqnarray}
1216: u_{1}(s) = \dot{g}_{+}(s) - \frac{g_{+}(s)}{g_{+}(t)} \dot{g}_{+}(t),
1217: \qquad u_{2}(s) = \frac{g_{+}(s)}{g_{+}(t)},
1218: \label{u1u2}
1219: \end{eqnarray}
1220: in order to satisfy the boundary conditions on these functions. Using
1221: the spectral density for an Ohmic bath, the equation that $g_{\pm}(s)$
1222: satisfies is
1223: \begin{eqnarray}
1224: \ddot{g}_{\pm}(s) \pm 2 \Gamma \dot{g}_{\pm}(s) + \omega_{r}^2
1225: g_{\pm}(s)=0\;. \label{EOMg}
1226: \end{eqnarray}
1227: The solution of eq.~(\ref{EOMg}) is
1228: \begin{eqnarray}
1229: g_{\pm}(s) &=& \frac{\sin \Omega s}{\Omega } e^{\mp \Gamma s}\;,
1230: \label{g+-exp}
1231: \end{eqnarray}
1232: where $\Gamma=\gamma$ and $\Omega^2 \equiv \omega_{r}^2
1233: -\gamma^2$.
1234:
1235: Using the spectral density for a super-Ohmic bath with
1236: $\nu=3$, the equations of motion for $g_{\pm}(s)$ can be written
1237: as
1238: \begin{equation}
1239: \ddot{g}_{\pm}(s) + \omega_{r}^2 g_{\pm}(s) \mp 2\gamma
1240: \stackrel{...}{{g}}_{\pm}(s)=0\;. \label{AL}
1241: \end{equation}
1242: Thus, the equation of motion has the form of the Abraham-Lorentz
1243: equation \cite{Jackson75}. This equation is an approximation to
1244: the full equation of motion with non-local time dependence. As in
1245: the electromagnetic field case, only the physically relevant roots
1246: of the characteristic equation must be retained. These are $\Gamma
1247: \pm i\Omega$ with
1248: \begin{eqnarray}
1249: \Gamma \equiv \frac{1}{6 \gamma} \left(\frac{D}{4}+\frac{1}{D} -1
1250: \right), \qquad \Omega \equiv \frac{\sqrt{3}}{6 \gamma}
1251: \left(\frac{D}{4}-\frac{1}{D} \right) \label{LambdaGammaOmega}
1252: \end{eqnarray}
1253: where
1254: \begin{eqnarray}
1255: D = 2 \left( 1+ 54 (\gamma \omega_{r})^{2}+ \sqrt{108}\gamma
1256: \omega_{r} \sqrt{27 (\gamma \omega_{r})^{2} +1} \right)^{1/3}.
1257: \label{Ddef}
1258: \end{eqnarray}
1259: Retaining the physically relevant roots of the characteristic
1260: equation, the solution for $g_{\pm}(s)$ for the super-Ohmic case
1261: has the same form as eq.~(\ref{g+-exp}) but with the above values
1262: of $\Gamma$ and $\Omega$. In the weak coupling limit that we
1263: consider in our calculations, $\Gamma \to \gamma \omega_r^2$ and
1264: $\Omega \to \omega_r^2$.
1265:
1266:
1267: In our model, eq.~(\ref{AL}) can be reduced to a second order ordinary
1268: differential equation by differentiating the homogeneous part of the
1269: Euler-Lagrange equations with respect to time as
1270: \begin{equation}
1271: \stackrel{...}{{g}}_{\pm}(s)=- \omega^2_r \dot{g}_{\pm}(s)
1272: +{\mathcal O}(\gamma)\;,
1273: \end{equation}
1274: which may be substituted into eq.~(\ref{AL}) to give (\ref{EOMg}) with
1275: $\Gamma=\gamma \omega^2_r$ and
1276: $\Omega=\omega_r$ in the leading order approximation as indicated above.
1277: Although one can improve the
1278: approximation leading to eq.~(\ref{EOMg}) for the super-Ohmic case
1279: to arbitrary higher order in $\gamma$, we restrict our study to the leading
1280: order in this parameter.
1281:
1282: Since the $u_{1,2}$ and $v_{1,2}$ solutions are now known, we have
1283: all the information needed to compute the various quantities
1284: necessary to obtain the numerical results. In particular, the
1285: expression for the matrix elements of ${\bf a}$ in eq.~(\ref{aij})
1286: for a general environment can be found.
1287: (See Refs.~\cite{CaldeiraLeggett84,QBM1,QBM2} for an Ohmic bath.) In our
1288: computations, we need $a_{11}(t)$ which has the form,
1289: \begin{eqnarray}
1290: a_{11}(t) &=& \frac{1}{2 \pi g_{+}^2(t)} \int_{0}^{\infty} d
1291: \omega J_B(\omega) \coth \frac{\beta \hbar \omega}{2}
1292: \Delta(\omega,t)\;,
1293: \label{a11general}
1294: \end{eqnarray}
1295: where $\Delta(\omega,t)$ was defined in eq.~(\ref{Deltaomega}).
1296: The evaluation of $\Delta(\omega,t)$ is straightforward and leads to,
1297: \begin{eqnarray}
1298: &&\Delta(\omega,t)= \frac{1}{2 \Omega^2}
1299: \frac{1}{(\Omega_{r}^2 - \omega^2)^2 + 4 \Gamma^2 \omega^2 }
1300: \left[ 2 \Omega^2 + e^{-\Gamma t} \right.\nonumber \\
1301: && \quad \times \left.
1302: \left[ 2 \cos(\omega t) \{ \left( \Gamma^2 + \omega_{+} \omega_{-} \right)
1303: \cos(\Omega t) - 2 \Gamma \Omega \sin(\Omega t) \} \right. \right.
1304: \nonumber \\
1305: &&\quad -
1306: \left. \left.
1307: \left( \Gamma^2 + \omega_{-}^2 \right) \cos(\omega_{+} t)
1308: -\left( \Gamma^2 + \omega_{+}^2 \right) \cos(\omega_{-} t)
1309: \right]\right. \nonumber \\
1310: &&\quad + \left. e^{-2 \Gamma t} \{ \Omega_{r}^2 + \omega^2
1311: +2 \Gamma \Omega \sin(2 \Omega t) \right. \nonumber \\
1312: &&\quad \left.
1313: -\left( \Gamma^2 + \omega_{+} \omega_{-} \right) \cos(2 \Omega t) \}
1314: \right]
1315: \label{G2Ohmic}
1316: \end{eqnarray}
1317: where $\omega_{\pm} \equiv \omega \pm \Omega$ and $\Omega_{r}^2
1318: \equiv \Omega^2 + \Gamma^2$. Using this result, in the asymptotic
1319: limit we obtain,
1320: \begin{eqnarray}
1321: J_B(\omega) \Delta(\omega,\infty)&=&
1322: \left\{
1323: \begin{array}{llll}
1324: \frac{2M \gamma \omega}{(\Omega_{r}^2 - \omega^2)^2 + 4 \gamma^2 \omega^2 }
1325: &(\nu=1)& \nonumber
1326: \\
1327: \frac{2M \gamma \omega^3}{(\Omega_{r}^2 - \omega^2)^2 + 4 \gamma^2 \omega^6 }
1328: &(\nu=3)&.
1329: \end{array}
1330: \right.
1331: \label{G2longtime}
1332: \end{eqnarray}
1333: In writing this equation, we used the fact that the distribution
1334: $J_B(\omega) \Delta(\omega,\infty)$ is highly peaked around
1335: $\omega_r$ for small $\gamma$.
1336:
1337: Similarly, for long times we may write the elements of ${\bf a}$
1338: as,
1339: \begin{eqnarray}
1340: a_{11}(t)
1341: &\rightarrow&
1342: \frac{\Omega^2 e^{2 \Gamma t}}{2 \pi \sin^2 \Omega t}
1343: \int_{0}^{\infty} d \omega J_B(\omega)
1344: \coth \frac{\beta \hbar \omega}{2} \Delta(\omega,\infty)
1345: %
1346: \nonumber \\
1347: a_{12}(t) &=& a_{21}(t) \rightarrow
1348: \frac{\Omega e^{\Gamma t}}
1349: {2 \pi \sin \Omega t}
1350: \int_{0}^{\infty} d \omega J_B(\omega)
1351: \coth \frac{\beta \hbar \omega}{2}
1352: \nonumber \\
1353: (&\Gamma& -\Omega \cot (\Omega s) )
1354: \Delta(\omega,\infty)
1355: %
1356: \nonumber \\
1357: a_{22}(t) &\rightarrow& \frac{1}{2 \pi} \int_{0}^{\infty} d
1358: \omega J_B(\omega) \coth \frac{\beta \hbar \omega}{2} \nonumber \\
1359: \left\{
1360: \right. &\omega^2& \left. + (\Gamma -\Omega \cot (\Omega s) )^2
1361: \right\} \Delta(\omega,\infty).
1362: %
1363: \label{aklinfty}
1364: \end{eqnarray}
1365: Using these results we can compute the asymptotic properties of
1366: ${\bf Q}$ needed in the calculations presented in the text.
1367:
1368: \section{Effective spectral density}\label{app:LRT}
1369:
1370: In this Appendix we derive the relation between the effective
1371: spectral density for the $CB$ system and the dynamical
1372: susceptibility of subsystem $C$ using an argument that by-passes
1373: the actual diagonalization procedure. A similar argument is given
1374: in Refs.~\cite{Leggett84,GOA85}. From the Hamiltonian given in
1375: eq.~(\ref{H}), we can write the equation of motion for $C$ with a
1376: harmonic potential in the form,
1377: \begin{eqnarray}
1378: x(\omega) = \chi_{C}(\omega) F_C(\omega)
1379: \label{xchif}
1380: \end{eqnarray}
1381: in the complex Fourier representation with a susceptibility
1382: function $\chi_{C}(\omega)$. The force $F_C(\omega)=- \partial
1383: V_{AC}(\omega)/\partial x= - \lambda \sigma(\omega)$ is the
1384: external force from subsystem $A$ acting on $C$. On the other
1385: hand, the equation of motion for subsystem $A$ gives
1386: \begin{eqnarray}
1387: - m_A \omega^2 \sigma(\omega) - \lambda x(\omega) = F_A(\omega)
1388: \label{EOMforsigma}
1389: \end{eqnarray}
1390: with $F_A(\omega)=- \partial V_{A}(\sigma) / \partial \sigma$.
1391: Here we assumed $H_{A}$ has the form $p_A^2/2m_A + V_A(\sigma)$
1392: for simplicity. Our argument does not depend on the form of
1393: $H_{A}$ as clear from the context. Combining with
1394: eq.~(\ref{xchif}), we have
1395: \begin{eqnarray}
1396: \left[ - m_A \omega^2 - \lambda^2 \chi_{C}(\omega)
1397: \right]\sigma(\omega) = F_A(\omega).
1398: \label{chisigmaf}
1399: \end{eqnarray}
1400:
1401: Now, suppose we have already diagonalized last three terms of
1402: eq.~(\ref{H}) and replaced them with $N+1$ - harmonic oscillators,
1403: \begin{eqnarray}
1404: H_C + V_{CB} + H_B \rightarrow H'_{B} =
1405: \sum_{n=0}^{N}( \frac{p_{n}^{'2}}{2 m'_{n}}
1406: + \frac{m'_{n} \omega_{n}^{'2} q_{n}^{'2}}{2}).
1407: \label{HB'}
1408: \end{eqnarray}
1409: We write the interaction term $V_{AC}$ in terms of new coordinates
1410: $q'_{n}$
1411: \begin{eqnarray}
1412: V_{AC} \rightarrow V'_{AB} = \sigma \sum_{n=0}^{N} c'_{n} q'_{n}.
1413: \label{ACnew}
1414: \end{eqnarray}
1415: Assuming that the system described by the Hamiltonian in
1416: eq.~(\ref{H}) reaches thermal equilibrium in the asymptotic limit,
1417: we can replace the effect of the combined system $CB$ by the
1418: equivalent thermal bath $H'_{B}$ at the same temperature. Although
1419: this diagonalization procedure is straightforward, for our
1420: purpose, we only need the form of the equation of motion for
1421: $\sigma$ expressed by the dissipation coefficient
1422: $\gamma_{CB}(\omega)$ for the effective bath eq.~(\ref{HB'}),
1423: \begin{eqnarray}
1424: \left[ - m_A \omega^2 - i m_A \omega \gamma_{CB}(\omega)
1425: \right]\sigma(\omega) = F_A(\omega).
1426: \label{FAomega}
1427: \end{eqnarray}
1428: Comparing the above with eq.~(\ref{chisigmaf}) we have
1429: \begin{eqnarray}
1430: \lambda^2 \chi_{C}(\omega)
1431: = i m_A \omega \gamma_{CB}(\omega).
1432: \label{chifromgammaCB}
1433: \end{eqnarray}
1434: Recall that the spectral density is related to the dissipation
1435: coefficient by
1436: \begin{eqnarray}
1437: J_{CB}(\omega) = m_A \omega \gamma'_{CB}(\omega),
1438: \label{JCBtogamma}
1439: \end{eqnarray}
1440: for real $\omega$. From this the desired relation follows:
1441: \begin{eqnarray}
1442: J_{CB}(\omega) = \lambda^2 \chi''_{C}(\omega).
1443: \label{JCBtoChiC}
1444: \end{eqnarray}
1445:
1446: \begin{thebibliography}{99}
1447:
1448: \bibitem{davis} See, e.g.,
1449: E.~B. Davis, {\em Quantum Theory of Open Systems}, (Academic Press,
1450: London, 1976);
1451: B. J. Lindenberg and B. J. West,
1452: {\em The Nonequilibrium Statistical Mechanics of Open
1453: and Closed Systems}, (VCH Press, New York, 1990);
1454: T. Dittrich, P. H\"anggi G.-L. Ingold, B. Kramer G. Sch\"on and W. Zwerger,
1455: {\em Quantum Transport and Dissipation}, (Wiley, New York, 1998).
1456:
1457: \bibitem{Mukamel95} S. Mukamel, {\it Principles of Nonlinear
1458: Optical Spectroscopy}, (Oxford University Press, New York, 1995).
1459:
1460: \bibitem{Weiss99}
1461: U. Weiss, {\it Quantum Dissipative Systems}, (World Scientific,
1462: Sigapore, 1999).
1463:
1464: \bibitem{Nakajima58} S. Nakajima, Prog. Theor. Phys. {\bf 20}, 948 (1958).
1465:
1466: \bibitem{Zwanzig61} R. Zwanzig, Lect. Theor. Phys. {\bf 3}, 106 (1961).
1467:
1468: \bibitem{FeyVer63}
1469: R. P. Feynman and F. L. Vernon, Ann. Phys. {\bf 24}, 118 (1963).
1470:
1471: \bibitem {QDOS} P. Pechukas and U. Weiss, eds., Special Issue,
1472: {\it Quantum Dynamics of Open Systems}, Chem. Phys, {\bf 268}, (2001).
1473:
1474: \bibitem{makri} D. E. Makarov and N. Makri, Chem. Phys. Lett.
1475: {\bf 221}, 482 (1994); N. Makri and D. E. Makarov, J. Chem. Phys.
1476: {\bf 102}, 4600, 4611 (1995); N. Makri, J. Math. Phys. {\bf 36},
1477: 2430 (1995); E. Sim and N. Makri, Comp. Phys. Commun. {\bf 99},
1478: 335 (1997): N. Makri, J. Phys. Chem. {\bf 102}, 4414 (1998).
1479:
1480: \bibitem{EgoEveSki99} S. A. Egorov, K. F. Everitt, and J. L. Skinner,
1481: J. Phys. Chem. A {\bf 103}, 9494 (1999).
1482:
1483: \bibitem{MayKuhn00}
1484: V. May and O. K\"{u}hn,
1485: {\it Charge and Energy Transfer Dynamics in Molecular Systems}
1486: (Wiley, New York, 2000).
1487:
1488: \bibitem{tully0} J.~C. Tully in {\em Modern Methods for
1489: Multidimensional Dynamics Computations in Chemistry}, ed. D.~L.
1490: Thompson, (World Scientific, NY, 1998), p. 34.
1491:
1492: \bibitem{herman} M.~F. Herman, Annu. Rev. Phys. Chem. {\bf 45},
1493: 83 (1994); M.~F. Herman, J. Chem. Phys. {\bf 87}, 4779 (1987).
1494:
1495: \bibitem{tully1} J.~C.~Tully, J. Chem. Phys. {\bf 93,} 1061 (1990);
1496: J.~C.~Tully, Int. J. Quantum Chem. {\bf 25}, 299 (1991); D.~S.
1497: Sholl and J.~C. Tully, J. Chem. Phys. {\bf 109}, 7702 (1998).
1498:
1499: \bibitem{coker} L. Xiao and D.~F. Coker, J. Chem. Phys. {\bf 100},
1500: 8646 (1994); D.~F. Coker and L. Xiao, J. Chem. Phys. {\bf 102},
1501: 496 (1995); H.~S. Mei and D.F. Coker, J. Chem. Phys. {\bf 104},
1502: 4755 (1996).
1503:
1504: \bibitem{rossky} F. Webster, P.~J. Rossky, and P.~A. Friesner,
1505: Comp. Phys. Comm. {\bf 63}, 494 (1991); F. Webster, E.~T. Wang,
1506: P.~J. Rossky, and P.~A. Friesner, J. Chem. Phys. {\bf 100}, 483
1507: (1994).
1508:
1509: \bibitem{martinez} T.~J. Martinez, M. Ben-Nun, and R.~D. Levine,
1510: J. Phys. Chem. A {\bf 101}, 6389 (1997).
1511:
1512: \bibitem{prezhdo} O.~V. Prezhdo and V.~V. Kisil, Phys. Rev. A {\bf 56},
1513: 162 (1997).
1514:
1515: \bibitem{martens} C.~C. Martens and J.-Y. Fang, J. Chem. Phys.
1516: {\bf 106}, 4918 (1996); A. Donoso and C.~C. Martens, J. Phys.
1517: Chem. {\bf 102}, 4291 (1998).
1518:
1519: \bibitem{kapral} R. Kapral and G. Ciccotti, J. Chem. Phys.
1520: {\bf 110}, 8919 (1999); S. Nielsen, R. Kapral and G. Ciccotti, J.
1521: Chem. Phys. {\bf 112}, 6543 (2000); S. Nielsen, R. Kapral and G.
1522: Ciccotti, J. Stat. Phys. {\bf 101}, 225 (2000); S. Nielsen, R.
1523: Kapral and G. Ciccotti, J. Chem. Phys. {\bf 115}, 5805 (2001); D.
1524: Mac Kernan, G. Ciccotti and R. Kapral, J. Chem. Phys. {\bf 116},
1525: 2346 (2002).
1526:
1527: \bibitem{schofield} C. Wan and J. Schofield, J. Chem.
1528: Phys. {\bf 113}, 7047 (2000).
1529:
1530: \bibitem{kapral2} R. Kapral, J. Phys. Chem. {\bf 105}, 2885
1531: (2001).
1532:
1533: \bibitem{CaldeiraLeggett84}
1534: A. O. Caldeira and A. J. Leggett,Physica {\bf A121}, 587 (1983);
1535:
1536: \bibitem{HakimAmb85} V. Hakim and V. Ambegaokar, Phys. Rev. A {\bf 32},
1537: 423 (1985).
1538:
1539: \bibitem {GSI88}
1540: H. Grabert, P.Schramn and G. L. Ingold, Phys. Rep. {\bf 168}, 115
1541: (1988).
1542:
1543: \bibitem{UnruhZurek89}
1544: W. G. Unruh and W. H. Zurek, Phys. Rev. D40, 1071 (1989);
1545:
1546: \bibitem{QBM1} B. L. Hu, J. P. Paz and Y. Zhang, Phys. Rev. D {\bf 45},
1547: 2843 (1992).
1548:
1549: \bibitem{QBM2}
1550: W. H. Zurek, S. Habib, and J. P. Paz, Phys. Rev. Lett. {\bf 70}, 1187 (1993);
1551: B. L. Hu and Y. Zhang, Int.\ J.\ Mod.\ Phys.\ {\bf A10} (1995) 4537;
1552: C. Anastopoulos and J. J. Halliwell, Phys. Rev. D {\bf 51}, 6870 (1995).
1553:
1554: \bibitem{EID}
1555: W. H. Zurek, Phys. Rev. D24, 1516 (1981); D26, 1862 (1982);
1556: E. Joos and H. D. Zeh, Z. Phys. B59, 223 (1985).
1557:
1558: \bibitem{Dec96} {\it Decoherence and the Apperance of the
1559: Classical World in Quantum Theory}, eds. D. Giulini, {\it et al.},
1560: (Springer, Berlin, 1996).
1561:
1562:
1563: \bibitem{IZ}
1564: C. Itzykson and J.-B. Zuber, {\it Quantum Field Theory},
1565: (McGraw Hill, New York, 1980).
1566:
1567: \bibitem{LCDFGZ87}
1568: A.~J. Leggett, S. Chakravarty, A.~T. Dorsey, M.~P. A. Fisher, A.
1569: Garg, and W. Zwerger, Rev. Mod. Phys. {\bf 59}, 1 (1987).
1570:
1571: \bibitem{Wigner} E.~Wigner, Phys. Rev. {\bf 40}, 749 (1932);
1572: K.~Imre, E.~\"{O}zizmir, M.~Rosenbaum and P.~F.~Zwiefel, J. Math.
1573: Phys. {\bf 5}, 1097 (1967).
1574:
1575: \bibitem{Tegmark}
1576: M. A. Huerta and H. S. Robertson, Jour. Stat. Phys. {\bf 1}, 393 (1969);
1577: M. Tegmark and H. S. Shapiro, Phys. Rev. E {\bf 50}, 2538 (1994).
1578:
1579: \bibitem{Kubo57} M. Toda, N. Saito, R. Kubo, and H. Hashizume
1580: {\it Statistical Physics I,II}
1581: (Springer, Berlin, 1991).
1582:
1583: \bibitem{Leggett84} A.~J. Leggett, Phys. Rev. B{\bf 30}, 1208 (1984).
1584:
1585: \bibitem{GOA85} A. Garg, J. N. Onuchic, and V. Ambegaokar,
1586: J. Chem. Phys. {\bf 83}, 4491 (1985).
1587:
1588: \bibitem{GargKim89} A. Garg and G. H. Kim, Phys. Rev. Lett. {\bf 63},
1589: 2512 (1989).
1590:
1591: \bibitem{BaroneCaldeira91} P. M. V. B. Barone and A. O. Caldeira, Phys. Rev.
1592: A {\bf 43}, 57 (1991).
1593:
1594: \bibitem{decoh} E.~R. Bittner and P.~J. Rossky, J. Chem. Phys.
1595: {\bf 103}, 8130 (1995); O.~V. Prezhdo and P.~J. Rossky, Phys. Rev.
1596: Lett. {\bf 81}, 5294 (1998).
1597:
1598: \bibitem{Jackson75} J. D. Jackson, {\it Classical Electrodynamics},
1599: (Wiley, New York, 1975).
1600:
1601: \bibitem {Chi}
1602: B. V. Chirikov, in {\it Chaos and Quantum Physics} Les Houches
1603: Lectures Session LII, eds M. J. Giannoni, A. Voros, and J.
1604: Zinn-Justin (North Holland 1991).
1605:
1606: \bibitem{ShiokawaHu95} K. Shiokawa and B. L. Hu, Phys. Rev. E{\bf 52}, 2497
1607: (1995); S. Habib, K. Shizume, and W. H. Zurek, Phys. Rev. Lett.
1608: {\bf 80}, 4361 (1998).
1609:
1610: \bibitem{PazZurek00}
1611: J. P. Paz and W. H. Zurek, in
1612: {\it Coherent Matter Waves}, Les Houches Lectures Session LII,
1613: (North Holland, Amsterdam, 1999).
1614:
1615:
1616:
1617:
1618:
1619: \end{thebibliography}
1620:
1621:
1622: \end{multicols}
1623: \end{document}
1624: