cond-mat0604170/ver4.tex
1: \documentclass[doublecol]{epl2} 
2: % or \documentclass[page-classic]{epl2} for one column style
3: 
4: 
5: 
6: 
7: 
8: \title{  Equilibration  problem for the generalized Langevin equation}
9: %\shorttitle{Title}
10: %Insert here a short version of the title if it exceeds 70 characters
11: 
12: \author{Abhishek Dhar\inst{1} \and Kshitij Wagh\inst{2} }
13: %\shortauthor{F. Author \etal}
14: 
15: \institute{                    
16:   \inst{1} Raman Research Institute, Bangalore 560080 \\
17:   \inst{2} Department of Physics and Astronomy, Rutgers University, Piscataway, New Jersey 08854-8019 
18: }
19: \pacs{05.40.-a} {Fluctuation phenomena, random processes, noise, and
20:   Brownian motion} 
21: \pacs{05.10.Gg} {Stochastic analysis methods}
22: \pacs{05.70.Ln} {Nonequilibrium and irreversible thermodynamics}
23: 
24: 
25: \abstract {We consider the problem of equilibration of a single
26: oscillator system with dynamics  given by the generalized classical 
27: Langevin equation.  It is well-known that this dynamics can be obtained if one
28: considers a model where the single oscillator is coupled to an
29: infinite bath of harmonic oscillators which are initially in equilibrium. 
30: Using this equivalence we first determine the conditions  necessary for
31: equilibration for the case when the system potential is harmonic.
32: We then give an example with a particular bath where we show that,
33: even for parameter values where the harmonic case always equilibrates,
34: with any  finite amount of nonlinearity the system does not 
35: equilibrate for arbitrary initial conditions. We understand this as a
36: consequence of the formation of nonlinear localized excitations similar
37: to the discrete breather modes in nonlinear lattices.   
38: }
39: 
40: 
41: \begin{document}
42: 
43: \maketitle
44: \def\bea{\begin{eqnarray}}
45: \def\eea{\end{eqnarray}}
46: \def\a{\alpha}
47: \def\d{\delta}
48: \def\D{\Delta}
49: \def\p{\partial} 
50: \def\nn{\nonumber}
51: \def\r{\rho}
52: \def\rv{\bar{r}}
53: \def\la{\langle}
54: \def\ra{\rangle}
55: \def\e{\epsilon}
56: \def\om{\omega}
57: \def\Om{\Omega}
58: \def\n{\eta}
59: \def\g{\gamma}
60: \def\bFi{{\Phi}}
61: \def\bM{{ M}}
62: \def\bY{{\cal{Y}}}
63: \def\bG{{\Gamma}}
64: \def\l{\lambda}
65: \def\break#1{\pagebreak \vspace*{#1}}
66: \def\f{\frac}
67: \def\dg{\dagger}
68: \def\zh{\hat{Z}}
69: \def\s{\sigma}
70: \def\bd{{\bf d}}
71: \def\ba{{\bf a}}
72: \def\be{{\bf e}}
73: \def\bb{{\bf b}}
74: \def\l{\lambda}
75: \def\tg{\tilde{\gamma}}
76: \def\tx{\tilde{x}}
77: \def\teta{\tilde{\eta}}
78: 
79: 
80: \section{Introduction}
81: One of the simplest phenomenological ways of
82: modeling the interactions of a system 
83: with a heat bath is through the Langevin equation which, for a single
84: particle, of unit mass and  moving in a one dimensional potential $V(x)$,
85: is given by 
86: \bea
87: \ddot{x} = -d V(x)/d x-\g \dot{x}+\eta(t)~, 
88: \label{leq}
89: \eea 
90: where  $\eta (t)$ is a Gaussian white
91: noise which satisfies the fluctuation-dissipation (FD) relation 
92: $\la \eta(t) \eta(t') \ra = 2 k_B T \g \delta (t-t') $ . Here $T$ is 
93: the temperature of the heat bath. The dynamics in eq.~(\ref{leq})
94: ensures that at long times the system reaches thermal
95: equilibrium. Thus the phase space 
96: density $P(x,p,t)$, where $p=\dot{x}$, converges in the 
97: limit $t\to \infty$ to the Boltzmann distribution $e^{-\beta H_s}/Z$ where
98: $H_s=p^2/2 +V(x)$ and $Z$ is the corresponding partition function.
99: The proof for this uses the  correspondence between the Langevin
100: equation and the Fokker-Planck equation \cite{risken}. 
101: 
102:      
103: However the $\delta$-correlated nature of the noise is 
104: unphysical and this has led to the study of the generalized Langevin
105: equation \cite{mori65,kubo66} 
106: \bea
107: \ddot{x} = -d V(x)/d x-\int_{-\infty}^t dt' \g (t-t')  \dot{x}(t')
108: +\eta(t)~, 
109: \label{gleq}
110: \eea
111: where the noise is correlated and the  the dissipative term involves a memory kernel.
112: The noise is again Gaussian and is related to the dissipative
113: term  through the generalized FD relation 
114: \bea
115: \la \eta(t) \eta(t') \ra = k_B T \g(t-t')~.
116: \label{gFD}
117: \eea
118:  A standard
119: method of microscopically modeling a heat bath is to consider an
120: infinite collection of harmonic oscillators, with a distribution of
121: frequencies, coupled linearly to the system. In that case it can
122: be shown \cite{zwanzig,weiss99}  that the effective equation of motion of the system is
123: precisely given by eq.~(\ref{gleq}) where the dissipation kernel
124: $\g(t)$ depends on the bath oscillator frequencies and the coupling
125: constants. 
126: Unlike the case with $\d$-correlated noise there exists no general
127: proof that, for a general potential $V(x)$, the system will
128: reach thermal equilibrium at long times.  The reason for this is that
129: in this case the construction of a Fokker-Planck description is
130: difficult and is  known in few cases ({\emph{e.g.}} harmonic
131: oscillator case treated in Ref.~\cite{adelman}).  For the special case of a
132: harmonic potential and with certain restrictions on the form of
133: $\g(t)$ one can prove equilibration  by a direct solution
134: of the equations of motion \cite{kubo}. 
135: In the quantum mechanical case the oscillator bath model has been
136:  widely used to model the effects of noise, dissipation and decoherence in
137:  quantum systems \cite{ford,caldeira}. In this
138: case the approach to equilibrium has been proved only for a
139: special class of  potentials for the cases where the system-reservoir
140: coupling is weak \cite{kac}.
141: 
142: 
143: A number of   papers \cite{sroko00,wang,bao} have attempted to understand various aspects 
144: of the  generalized Langevin equation such as anomalous diffusion,
145: nonstationarity and ergodicity.  
146: In this paper we address the question of approach to equilibrium for
147: the generalized Langevin equation. For a particle in a harmonic
148: potential $V(x)=\om^2_0 x^2/2$ and 
149: coupled to a heat bath with a  finite band-width (and hence long-time
150: memory) we show that  
151: equilibration is not always ensured and depends on the oscillator
152: frequency $\om_0$. We find the necessary
153: conditions for equilibration which is related to the existence of
154: bound states (localized modes) of the coupled system-plus-bath. 
155: We note that Ref.~\cite{bao} looks at the
156: question of ergodicity which is similar to the question addressed
157: here. One of their results is that the motion of a  harmonic
158: oscillator coupled to a general heat bath is ergodic.
159: This is based on the fact  that the coefficients in the 
160: Fokker-Planck equation  asymptotically approach constant values. Our
161: work shows that this condition does {\emph{not}} ensure ergodicity.
162: Next we consider a 
163: particle in a nonlinear potential $V(x)=\om_0^2 x^2/2 + u x^4/4$ and
164: coupled to a special bath, the so-called Rubin model \cite{rubin60,weiss99}.
165: The usual expectation
166: would be that nonlinearity should help in equilibration. However
167: surprisingly we find the contrary to be true. For values of $\om_0$ for
168: which the linear system (with $u=0$) does equilibrate, we show that
169: switching on  the nonlinearity can lead to loss of equilibration.
170: We relate this to the formation of nonlinear localized  modes which
171: are similar to the discrete breather modes found in nonlinear lattices
172: \cite{sievers,flach}. 
173: 
174: \section{Definition of model}
175: We consider the following Hamiltonian
176: of the usual model of a system coupled to a bath of $N$ oscillators:
177: \bea
178: H=H_s+ \sum_{\a=1}^N \f{P_\a^2}{2}+\f{\om_\a^2}{2} 
179: \left( X_\a-\f{c_\a  x}{\om^2_\a} \right)^2 ~, \label{ham}
180: \eea
181: where $\{ X_\a,~P_\a \}$ denotes  degrees of
182: freedom of the $\a^{\rm th}$ bath oscillator ($\a=1,2,...N$), $\om_\a $ is its
183: frequency,  and $c_\a$ is the strength of the coupling to the
184: system. For dissipation it is necessary that the frequencies $\om_\a$
185: have a continuous spectrum in the limit $N \to \infty$.  
186: We assume that at time $t=t_0$ the system and bath are decoupled
187: ($\{c_\a\}=0$ ).
188: The bath is in thermal equilibrium and described by the canonical
189: distribution $e^{-\beta H_b}$ where $H_b=\sum_\a [~P^2_\a/2+\om_\a^2
190:   X_\a^2/2~]$ and the system is in an arbitrary initial state
191: $\{x(t_0),~p(t_0)\}$. The coupling is switched on at time $t=t_0$. It
192: can 
193: then be shown that eliminating the bath degrees of freedom 
194: leads to the following equation of motion for the system (for $t > t_0$):
195: \bea
196:  \ddot{x} = -\f{d V(x)}{d x}-\g(0) x +\int_{t_0}^t dt' \f{d\g (t-t')}{dt'}  {x}(t')
197: +\eta(t)~, 
198: \label{redeq}
199: \eea
200: with the dissipation kernel given by
201: $ \g(t)=\sum_\a ({c_\a^2}/{\om_\a^2}) \cos {(\om_\a t)}$. 
202: In the limit $t_0 \to -\infty$, using $\g(t \to \infty)=0$, we get eq.~(\ref{gleq}). 
203: The noise correlations satisfy the generalized FD relation in eq.~(\ref{gFD})
204: where the noise average $\la ...\ra$ is obtained by averaging 
205: over the initial conditions of the bath.
206: 
207: \section{Equilibration in a harmonic potential}
208: We first  
209: consider the case with $V(x)=\om^2_0 x^2/2$. For this case
210: eq.~(\ref{redeq}) is a linear inhomogeneous equation whose solution is:
211: \bea
212: x(t)&=& H(t-t_0) x(t_0) + G(t-t_0) p(t_0) \nn \\  
213: &&+ \int_{t_0}^t dt' G(t-t') \eta(t')~, \label{gensol}
214: \eea
215: where $H(t)$ and $G(t)$ are the solutions of the
216: homogeneous part of eq.~(\ref{redeq}) (with $t_0=0$) for initial conditions 
217: $H(0)=1,~\dot{H}(0)=0$ and  $G(0)=0,~\dot{G}(0)=1$ respectively. 
218: For equilibration we require that, in the long time limit, the
219: solution in eq.~(\ref{gensol}) should not
220: depend on initial conditions. Thus a necessary condition is that
221: $G(t \to \infty) \to
222: 0$. At large times it can be shown that 
223: $H(t) \to \dot{G}(t)$ and hence $H(t \to \infty) \to 0$ also.
224: Let us define the Green's function $G^+(t)= G(t) \theta
225: (t)$. It is easy to see that $G^+$ satisfies the equation of
226: motion:
227: \bea
228:  \ddot{G^+} + \om_0^2  G^+ +\int_{-\infty}^\infty dt' \g^+ (t-t')
229: \dot{G}^+(t')= \delta(t)~, \nn 
230: \eea
231: where $\g^+(t)=\g(t) \theta(t)$.
232: Defining the Fourier transforms $G^+(\om)= \int_{-\infty}^\infty dt
233: e^{i \om t} G^+(t)$ and $\g^+(\om) =\int_{-\infty}^\infty dt e^{i \om t}
234: \g^+(t)$ we find that $G^+(\om)$ is given by 
235: \bea
236: G^+(\om)&=&\f{1}{- (\om^2-\om_0^2)-i \om \g^+(\om) }~  \label{gom}
237: \eea
238: with $\g^+(\om)=\g^+_R(\om)+ i \g^+_I (\om)$ where the real and imaginary
239: parts are given by:
240: $\g^+_R(\om)= \sum_\a ({\pi c_\a^2}/{2 \om_\a^2})
241:   [\d(\om+\om_\a)+\d(\om-\om_\a)],~~ 
242: \g^+_I(\om) =  \sum_\a {c_\a^2 \om}/{[ \om_\a^2 (\om^2-\om_\a^2)]}$.  
243: Now from eq.~(\ref{gom}) it follows that  $G^+(t \to \infty)=
244: \lim_{t \to \infty} (1/2\pi) \int d\om  e^{-i \om t }
245: G^+(\om)$ will vanish (this follows from the Riemann-Lebesgue lemma
246: \cite{bender})  {\emph{unless}} $G^+(\om)$ blows up at some real
247: value of $\om$.
248: Thus the condition under which $G^+(t\to \infty)$ does not vanish  is
249: that the equation     
250: \bea
251: - (\om^2-\om_0^2)-i \om \g^+(\om)=0 \label{bseq}
252: \eea
253: has a real $\om$ solution, which we will denote by $\Om_b$. It follows from
254: eq.~(\ref{bseq}) that $\g_R(\Om_b)=0$ which means that $\Om_b$
255: necessarily lies outside the bath band-width. 
256: This solution corresponds to a bound state. To see this we now solve 
257: the equations of motion of the coupled system-plus-bath 
258: by a resolution into its normal modes. Let us denote the normal mode
259: frequencies by $\Om_Q$ and the corresponding eigenfunction by
260: ${\bf{U}}_Q=\{ U_{0,Q},U_{\a=1,Q},...U_{\a=N,Q}\}$ where $U_{0Q}$
261: corresponds to the system variable $x$. They satisfy the
262: equations:
263: \bea
264: - \Om_Q^2 U_{0Q}&=& - \om_0^2 U_{0Q} +\sum_\a c_\a \left( U_{\a Q} -\f{c_\a
265: U_{0Q}}{\om_\a^2} \right)\nn \\
266: -\Om_Q^2 U_{\a Q} &=& -\om_\a^2 \left(U_{\a Q}-\f{c_\a
267:   U_{0Q}}{\om_\a^2} \right)~ ~~\a=1...N~ \label{nmodeeq}
268: \eea   
269: A bound state \cite{ziman} occurs if there is a mode such that its
270: frequency $\Om_b$ 
271: lies outside the bandwidth of the isolated bath. The corresponding
272: eigenfunction ${\bf U}_b$ will have a finite weight at the system
273: point (\emph{i.e.} $U_{0b}= O(1)$). Solving, for $U_{\a b}$, the second
274:   equation in eq.~(\ref{nmodeeq}) and substituting into the first
275:   equation, we find that the condition for a bound 
276: state is given precisely by the solution (if it exists) of eq.~(\ref{bseq}).
277: The corresponding eigenfunction is given by:
278: $U_{\a b}={c_\a U_{0b}}/({-\Om_b^2+\om_\a^2}),~~\a=1...N$ and 
279: $U_{0 b} =  [ 1+ \sum_\a {c^2_\a}/{(-\Om_b^2+\om_\a^2)^2}]^{-1/2}$.
280: The general solution of the full equations of motion in terms of
281: normal modes is 
282: \bea
283: Y=U\cos{\Om (t-t_0)} U^{-1} Y(t_0) + U \f{\sin{
284: \Om(t-t_0)}}{ \Om} U^{-1} \dot{Y}(t_0) ~, \nn
285: \eea
286: where
287: $Y=\{x,X_1,...X_\a...X_N\}^T$. From this one can identify
288: the part of $x(t)$ involving 
289: $p(t_0)$. Comparing with eq.~(\ref{gensol}) we get 
290: $G(t)=\sum_Q U^2_{0Q} {\sin{ (\Om_Q t)}}/{\Om_Q}$.   
291: If there are no bound states then the frequencies $\Om_Q$ form a
292: continuous spectrum, the sum can be converted into an
293: integral, and in the limit $t \to \infty$ the infinite oscillations
294: lead to the integral vanishing. In the presence of a bound state we
295: get a non-vanishing contribution given by 
296: \bea
297: G(t \to \infty) = U_{0b}^2 \f{\sin{( \Om_b t )}}{\Om_b}~. \label{Ginf}
298: \eea 
299: Thus we again arrive at the conclusion that in the presence of bound
300: states, thermal equilibration is not achieved and the steady state
301: properties depend on the initial conditions of the system.
302: The long time form in eq.~(\ref{Ginf})  can also be obtained directly
303: from eq.~(\ref{gom}) by integrating 
304: around the singularity given by eq.~(\ref{bseq}).  
305: In the absence of bound states, eq.~\ref{gensol} gives the  unique
306: steady state solution $x(t)=\int_{-\infty}^\infty dt' G^+(t-t')
307: \n(t')$, and one can verify that:
308: \bea
309: \la \f{\om_0^2 x^2}{2} \ra = \f{\om_0^2 k_B T}{2 \pi} 
310: \int_{-\infty}^\infty d \om |G^+(\om)|^2 Re[\g^+ (\om)] \nn &=&\f{k_B T}{2}\\ 
311: \la \f{p^2}{2} \ra = \f{ k_B T}{2 \pi} 
312: \int_{-\infty}^\infty d \om~ \om^2 |G^+(\om)|^2 Re[\g^+ (\om)] 
313: &=&\f{k_B T}{2} ~,\nn
314: \eea
315: where the integrals are easy to evaluate if one chooses appropriate
316: contours \cite{kubo}. The inset in fig.~(\ref{equilib}) gives a
317: numerical demonstration of the nonequilibration problem (see following
318: section for details of numerics). 
319: 
320: \section{ Equilibration in a nonlinear potential}
321: Usually one
322: expects that introducing nonlinearity should help in equilibration and
323: we will now test this expectation. We will consider a particular model
324: of a heat bath for which it will be  possible to analyze things in
325: detail. This is the Rubin model in which the bath  is a one-dimensional
326: chain of coupled oscillators. Thus the full Hamiltonian is:
327: $H=H_s+\sum_{i=1}^N [ {p_i^2}/{2} +{(x_{i+1}-x_{i})^2}/{2}
328:   ] + {(x_1-x)^2}/{2}$,  
329: with $x_{N+1}=0$ and where $x=x_0$ refers to the system. 
330: Transforming to normal mode coordinates of the bath we recover the
331: form eq.~(\ref{ham}) with $\om_\a=4 \sin^2(q_\a/2)$ and $c_\a = 
332: [2/(N+1)]^{1/2} \sin (q_\a)$ with $q_\a=\a \pi/(N+1)$ and $\a=1,2...N$. 
333: From this we can find the explicit form of $\g^+(\om)$ which is:
334: \bea
335: && i\om \g^+(\om)=-(1-e^{iq}) ~~~{\rm for}~|\om|=2| \sin{(q/2)}| ~<~2 \nn
336: \\
337: &&=-(1+e^{-\s}) ~~~{\rm for}~|\om|=2 \cosh{(\s/2)} ~>~2.   \label{alpeq} 
338: \eea
339: For a harmonic potential $V(x)=\om_0^2 x^2/2$ we see from
340: eq.~(\ref{bseq})  that the condition
341: for getting bound states is $ \om_0^2 > 2 $. The bound-state mode
342: is given by $x_i^b=A  (-1)^i e^{-\s i}\cos(\Om_b t)$.
343: where 
344: $\Om_b=\om_0^2/\sqrt{\om_0^2-1}$ is the bound state frequency,
345: $\s=\ln{(\om^2_0-1)}$ and $A$ is an arbitrary 
346: constant which can be fixed by normalization.    
347: 
348: We now introduce the nonlinear term $u x^4/4$ and study its effect
349: numerically. Let us choose $\om_0^2=1$, in
350: which case, in the absence of the nonlinear term, there are no bound
351: states and we expect equilibration. In our simulation we consider a
352: heat bath with $N=1000$ particles whose initial positions and momenta
353: are chosen from the  Boltzmann distribution $e^{- H_b/T}$ with
354: $T=1$.  
355: The system has the initial state $\{p(0)=0, x(0)\}$. We
356: solve the equations of motion numerically with this initial state with
357: the system-bath coupling turned on at time $t=0$. We compute the expectation
358: values $K_e(t)=\la p^2 \ra$ and $P_e(t)=\la x ~\p H/\p x \ra$ where
359: $\la....\ra$ denotes an average over the bath initial conditions. In
360: our simulations we averaged over $10^4$ initial conditions. We have
361: checked that results do not change on  increasing the bath size or
362: number of aveages. For
363: equilibration we expect the steady state values $K_e=P_e=T$.    
364: In fig.~\ref{equilib} we see that for the initial condition $x(0)=7.0$
365: the linear problem equilibrates as expected. On the other hand for a
366: small value of the nonlinearity parameter $u=0.1$ the system \emph{fails to 
367:   equilibrate}. For a different initial condition $x(0)=5.0$ the system
368: equilibrates. 
369: The loss of equilibration infact occurs whenever $x(0)$ is greater than
370: a critical value $x_c\approx 5.6$. 
371: \begin{figure}
372: \onefigure[width=3.3in]{fig1p.eps}
373: \caption{Plot of average kinetic energy $K_e(t)$ as a function of time
374:   for different initial conditions. The purely linear case always
375:   equilibrates (for $\om_o=1$) while in the nonlinear case, equilibration  depends on
376:   initial conditions. The behaviour of $P_e(t)$ is similar. The inset shows the
377: nonequilibration problem in the linear case. The parameter
378: value $\om_0^2=3.0$ gives a bound state with $\Om_b=3/2^{1/2}$ and we
379: plot $K_e(t)$ at large times for $x(0)=5.0$ and
380: $T=1$. The solid line gives the analytic prediction $K_e(t)=\ddot
381: {G}(t)^2 x(0)^2 +T$ with $G(t)$ given by eq.~(\ref{Ginf}).   
382: }
383: \label{equilib}
384: \end{figure}
385: 
386: The dependence of equilibration on initial
387: conditions can be traced to the formation of localized states that can 
388: be generated once we have nonlinearity.  
389: These {\emph{nonlinear localized modes}} \cite{sievers,flach} are stable time-dependent
390: solutions of the equations of motion that are localized and largely
391: monochromatic and have the same form as the impurity bound states
392: namely  $x^b_i(t)=A (-1)^i e^{-\s i}
393: \cos(\Om_b t)$. If we plug this into the equations of motion and
394: neglect terms containing higher harmonics ($\sim 3 \Om_b$) 
395: we then get the following
396: conditions relating $\Om_b, \s$ and $A$ (for the parameter choice $\om_0=1$):
397: \bea
398: A^2=\f{4 e^\s}{3 u},~~~~
399: \Om^2_b=2+e^{\s}+e^{-\s}~. 
400: \label{bscond}
401: \eea 
402: Note that, of the three parameters needed to describe a bound state
403: namely $\Om_b, \s, A$ only two get fixed and we thus have a continuum
404: set of possible bound states which can be described by a single
405: parameter.  
406: It is clear that for any non-zero $u$ we can always choose 
407: $\s >0$ and determine $A,~\Om_b$ from eq.~(\ref{bscond}). Hence bound
408: states exist for \emph{any} value of $u$. 
409: A clean  way to numerically observe these
410: localized modes is by initially preparing the bath at $T=0$ and
411: the system with initial conditions $\{p(0)=0, x(0)\}$. Then at very
412: long times we find that this initial condition relaxes to a localized
413: mode  if $x(0)$ is sufficiently large. 
414: In fig.~\ref{breath} we plot the coordinates $\{x(t),x_i(t)\}$ for
415: $i=1,2,3,4$ as functions of time (for parameter values $\om_0=1$
416: and $u=0.1$) . For the initial 
417: condition $x(0)=7.0$  it is clear that the long time solution is
418: quite accurately described by the localized mode solution. 
419: In the fits with the localized mode we fix $A$ from the numerically
420: obtained amplitude of $x(t)$  and obtain  $\s$ and $\Om_b$ from
421: eq.~(\ref{bscond}). 
422: For the  initial condition $x(0)=5.0$ we find that the initial energy
423: quickly dissipates into the bath and at long times is distributed uniformly.
424: A more accurate form for the localized mode can be obtained by
425: including higher harmonics. At the next order, the solution is given by 
426: $x_i^b=A^{(1)}(-1)^ie^{-\s^{(1)} i} \cos{\Om_b t}+
427: A^{(2)}(-1)^ie^{-\s^{(2)} i} \cos{3\Om_b t}$, where again there is only
428: one unknown constant. In our example we find that $A^{(2)} << A^{(1)}$
429: (For all values of $u$ we get $A^{(2)}/A^{(1)} < 1/21$). 
430: \begin{figure}
431: \onefigure[width=3.3in]{fig2.eps}
432: \caption{Plots of the positions of various particles $x(t)$ and
433:   $x_i(t)$ for $i=1,2,3,4$ as functions of time plotted after the
434:   system has reached a steady state. Initial condition was $x(0)=7.0$
435:   and all other positions and momenta set to zero. The solid lines
436:   correspond to the analytical prediction for  the breather mode with 
437:   $A=6.44$, $\s=1.135$ and $\Om_b=2.331$. }
438: \label{breath} 
439: \end{figure}
440: 
441: Thus we have shown that the finite-temperature non-equilibration
442: problem is related to the formation
443: of nonlinear localized modes which, unlike in the harmonic case, 
444: depends on initial conditions.  
445: 
446: \section{Discussion}
447: In summary we have examined  the  conditions
448: necessary for thermal equilibration of a  particle in a  
449: potential well and evolving through a generalized Langevin equation.
450: For the case of a harmonic potential we show that 
451: for finite band-width baths the system will not always equilibrate. 
452: Using the microscopic model of a bath as a collection of oscillators
453: we show that non-equilibration arises because of the formation of
454: bound states in the coupled system-plus-bath. 
455: Surprisingly we find that making the system nonlinear does not
456: restore equilibration. On the contrary nonlinearity assists in the
457: formation of localized modes and hence causes loss of equilibration for
458: arbitrary  initial conditions. These nonlinear localized modes
459: are usually studied in the context of periodic nonlinear lattices,
460: while here they arise in a situation where only one spring is nonlinear. 
461: In this paper we have considered a special one-dimensional
462: bath. However localized modes also occur in higher dimensional
463: nonlinear lattices with finite band-widhts \cite{flach} and hence we
464: expect that the problem of equilibration is quite general.
465: 
466: In Ref.~\cite{dhar}  electronic systems described by the 
467: tight binding noninteracting Hamiltonian were discussed and it was 
468: shown that the formation of bound states with finite band width
469: reservoirs leads to a similar problem of equilibration. One example
470: which was studied was that of a single site attached to a
471: one-dimensional free electron reservoir. 
472: An interesting question that the present study raises is whether
473: the introduction of an onsite interaction, say of the  Hubbard type,  
474: would lead to similar effects as the nonlinear term in the case of the
475: oscillator.  
476: 
477: \begin{thebibliography}{0}
478: \bibitem{risken} 
479:   \Name{Risken H.} 
480:   \Book{The Fokker Planck Equation}
481:   \Publ{Springer-Verlag, Berlin} 
482:   \Year{1984}
483:   \Page{134}. 
484: \bibitem{mori65} 
485:   \Name{Mori H.} 
486:   \Review{Prog. Theor. Phys.} {\bf{33}} {(1965)} {423}. 
487: \bibitem{kubo66} 
488:   \Name{Kubo R.} 
489:   \Review{Rep. Prog. Theor. Phys.} {\bf{29}} {(1966)} {255}.
490: \bibitem{zwanzig} 
491:   \Name{Zwanzig R.} 
492:   \Book{Nonequilibrium Statistical Mechanics} 
493:   \Publ{Oxford University Press, Oxford} 
494:   \Year{2001}
495: \bibitem{weiss99} 
496:   For the quantum case see: 
497:   \Name{Weiss U.} 
498:   \Book{Quantum Dissipative Systems (Second Edition)}  
499:   \Publ{World Scientific, London} 
500:   \Year{1999};
501:   \Name{Hanggi P.} 
502:   \Review{Lect. Notes Phys.} {\bf 484} {(1997)} {15}.  
503: \bibitem{adelman} 
504:   \Name{Adelman S. A.}
505:   \Review{J. Chem. Phys.} {\bf   64} {(1976)} {124}. 
506: \bibitem{kubo} 
507:   \Name{Kubo R., Toda M. \and Hashitsume N.} 
508:   \Book{Statistical Physics II} 
509:   \Publ{Springer-Verlag, Berlin} 
510:   \Year{1985}.  
511: \bibitem{ford} 
512:   \Name{Ford G. W., Kac M. \and Mazur P.}
513:   \Review{J. Math. Phys.} {\bf 6} {(1965)} 504.
514: \bibitem{caldeira} 
515:   \Name{Caldeira A. O. \and Leggett A. J.} 
516:   \Review{Physica A} {\bf 121} {(1983)} {587};
517:   \Name{Hakim V. \and Ambegaokar V.} 
518:   \Review{Phys. Rev. A} {\bf 32} {(1985)} {423};
519:   \Name{Grabert H., Schramm P. \and Ingold G. L.}   
520:   \Review{Phys. Rep.} {\bf 168} {(1988)} {115}.
521: \bibitem{kac} 
522:   \Name{Benguria R. \and Kac M.} 
523:   \Review{Phys. Rev. Lett.} {\bf 46} {(1981)} {1}.
524: \bibitem{sroko00}  
525:   \Name{Srokowski T.} 
526:   \Review{Phys. Rev. Lett.} {\bf 85} {(2000)} {2232};  
527:   \Name{Morgado R. {\it et al}},  
528:   \Review{Phys. Rev. Lett.} {\bf 89} {(2002)} {100601};
529:   \Name{Mokshin A. V., Yulmetyev R. M. \and Hanggi P.}  
530:   \Review{Phys. Rev. Lett.} {\bf 95} {(2005)} {200601}; 
531:   \Name{Vinales A. D. \and Desposito M. A.} 
532:   \Review{Phys. Rev. E} {\bf 73} {(2006)} {016111}; 
533:   \Name{Plyukhin A. V. \and Schofield J.}
534:   \Review{Phys. Rev. E} {\bf 64} {(2001)} {041103}.  
535: \bibitem{wang} K. G. Wang, Phys.  Rev. A {\bf 45}, 833 (1992); K. G. Wang 
536: and M. Tokuyama, Physica A {\bf 265}, 341 (1999). 
537: \bibitem{bao} 
538:   \Name{Bao J. D., Hanggi P. \and Zhuo Y. Z.} 
539:   \Review{Phys. Rev. E} {\bf 72} {2005} {061107}. 
540: \bibitem{rubin60} 
541:   \Name{Rubin R. J.} 
542:   \Review{J. Math. Phys.} {\bf 1} {(1960)} {309}.
543: \bibitem{sievers} 
544:   \Name{Sievers A. J. \and Takeno S.} 
545:   \Review{Phys. Rev. Lett.} {\bf 61} {(1988)} {970}. 
546: \bibitem{flach} 
547:   \Name{Flach S. \and Willis C. R.} 
548:   \Review{Phys. Rep.} {\bf 295} {(1998)} {181}. 
549: \bibitem{bender} 
550:   \Name{Bender C. M. \and Orszag S. A.} 
551:   \Book{Advanced Mathematical Methods for Scientists and Engineers} 
552:   \Publ{McGraw-Hill, New York} 
553:   \Year{1978}. 
554: \bibitem{ziman} 
555:   \Name{Ziman J. M.} 
556:   \Book{Principles of the theory of solids (Second Edition)}
557:   \Publ{Cambridge University Press, London} 
558:   \Year{1972}
559:   \Page{71}.
560: \bibitem{dhar} 
561:   \Name{Dhar A. and Sen D.} 
562:   \Review{Phys. Rev. B} {\bf 73} {2006} {085119}.
563: \end{thebibliography}
564: 
565: \end{document}
566: 
567: 
568: \bibitem{b.a}
569:   \Name{Author F., Author S. \and Author T.}
570:   \REVIEW{Some Rev. A}{69}{1969}{9691}.
571: 
572: \bibitem{b.b}
573:   \Name{Author F. \and Author S.}
574:   \Book{Some Book of Interest}
575:   \Editor{A. Editor}
576:   \Vol{9}
577:   \Publ{Publishing house, City}
578:   \Year{1939}
579:   \Page{666}.
580: 
581: \bibitem{b.c}
582:   \Editor{Editor A.}
583:   \Book{Some Book of Interest}
584:   \Vol{9}
585:   \Publ{Publishing house, City}
586:   \Year{1939}
587:   \Section{A}.
588: 
589: \end{thebibliography}
590: 
591: 
592: 
593: