1: \documentclass[nobm,figures,cite]{epl}
2:
3: \title{Rocking bistable systems: use and abuse of Linear Response
4: Theory} \shorttitle{Rocking bistable systems: use and ...}
5: \author{J.~Casado-Pascual\inst{1,2} \and J.~G\'omez-Ord\'o\~nez\inst{1}
6: \and M.~Morillo\inst{1} \and P.~H\"anggi\inst{2}} \institute{ \inst{1}
7: F\'{\i}sica Te\'orica, Universidad de Sevilla - Apartado de Correos
8: 1065, Sevilla 41080, Spain\\ \inst{2} Institut f\"ur Physik,
9: Universit\"at Augsburg - Universit\"atsstra\ss e 1, D-86135 Augsburg,
10: Germany } \shortauthor{J.~Casado-Pascual \etal}
11: \pacs{05.40.-a}{Fluctuation phenomena, random processes, noise, and
12: Brownian motion} \pacs{05.20.-y} {Classical statistical mechanics}
13: \pacs{02.50.Ey}{Stochastic processes}
14:
15: \begin{document}
16:
17: \maketitle
18:
19: \begin{abstract}
20: The response of a nonlinear stochastic system driven by an external
21: sinusoidal time dependent force is studied by a variety of numerical and
22: analytical approximations. The validity of linear response theory is put
23: to a critical test by comparing its predictions with numerical
24: solutions over an extended parameter regime of driving
25: amplitudes and frequencies. The relevance of the driving frequency
26: for the applicability of linear response theory is explored.
27: \end{abstract}
28:
29: %\section{}
30:
31:
32:
33: % body of paper here
34: %\begin{multicols}{2}
35: The response of dissipative physical systems to small amplitude external
36: perturbations is usually described with the powerful tools of linear
37: response theory (LRT)\cite{kubo57}, as it is generally accepted that the
38: effect of the perturbation can be described in terms of small deviations
39: from the behavior of the unperturbed system. In particular, for long times
40: and for systems which in the absence of driving reach an equilibrium
41: distribution, LRT provides an approximate expression for the probability
42: distribution obtained by keeping just the linear terms in a series
43: expansion in the external
44: amplitude. The purpose of the present letter is to point out the
45: relevance of parameters other than the amplitude of the driving
46: force, for the validity of LRT. We will show that
47: for a periodic external force, the validity of LRT depends not just on the
48: amplitude of the driving term but also crucially on its frequency.
49:
50: Let us consider a system characterized by a single degree of freedom,
51: $x$, whose time evolution is governed by the nonlinear Langevin equation
52: (in dimensionless form),
53: \begin{equation}
54: \dot{x}(t)=x(t)-x^{3}(t)+A\cos \Omega t+\eta(t),
55: \end{equation}
56: where $A\cos\Omega t$ represents an external signal and
57: $\eta(t)$ is a Gaussian white noise with zero average and
58: $\langle \eta(t)\eta(s)\rangle = 2D\delta(t-s)$. The corresponding linear
59: Fokker-Planck equation (FPE) for the probability density $P(x,t)$ reads
60: \begin{equation}
61: \label{lfpe}
62: \frac{\partial P}{\partial t}=\frac{\partial}{\partial
63: x}\big\{(-x+x^{3}- A\cos\Omega t)P\big\}+D\frac{\partial^{2}P}{\partial
64: x^{2}}.
65: \end{equation}
66: The unperturbed system has an equilibrium distribution of the form
67: \begin{equation}
68: \label{peq}
69: P_{eq}(x)=N \exp \left (-\frac {U_0(x)}{D}\right ),
70: \end{equation}
71: where $N$ is a normalization constant and $U_0(x)$ is the unperturbed
72: potential
73: \begin{equation}
74: U_0(x)=-\frac {x^2}2+\frac {x^4}4.
75: \end{equation}
76: This potential has two minima located at $x_m=\pm 1$ and a maximum at
77: $x_M=0$, with a barrier height of $0.25$. The potential $U_0(x)-Ax\cos
78: \Omega t$ loses its bistable character for $A \ge A_T=\sqrt{4/27}$.
79:
80: The analysis of the dynamics is simplified by making use of two
81: important theorems: the H-theorem, which ensures the existence of a
82: unique long time distribution function $P_{\infty}(x,t)$
83: \cite{Lebowitz,Risken} and the Floquet theorem, which guarantees that
84: $P_{\infty}(x,t)$ is periodic in time with the same period as the
85: external force \cite{junhan91}. For the system at hand, the symmetry of
86: $U_0(x)$ implies the following properties for the long time unique
87: solution of the FPE: $P_\infty(-x,t;-A)=P_\infty(x,t;A)$ and
88: $P_\infty(-x,t;A)=P_\infty(x,t+T/2;A)$, where $T=2\pi/\Omega$ and we
89: have indicated explicitely the dependence of $P_{\infty}$ on $A$. Using
90: the Fourier expansion,
91: \begin{equation}
92: \label{fourier}
93: P_{\infty}(x,t;A)=\sum_{m=-\infty}^\infty H_m(x;A) e^{im\Omega t}
94: \end{equation}
95: the first property leads to $H_m(x;A)=H_m(-x;-A)$, while the second one
96: implies that $H_m(x;A)=(-1)^m H_m(-x;A)$. From both of them, we obtain
97: $H_m(x;-A)=(-1)^m H_m(x;A)$. It then follows immediately that the odd
98: moments of the distribution, $\langle x^n(t)
99: \rangle_\infty,\;n=1,3,\ldots$ can be written as Fourier series
100: containing only odd harmonics as the even harmonics vanish due to the
101: symmetries above. Analogously, even moments $\langle
102: x^p(t)\rangle_\infty,\;p=0,2,\ldots$ contain just even harmonics in
103: their Fourier series expansions\cite{Haenggi93}. Inserting the Fourier
104: expansion, eq.~(\ref{fourier}), into the FPE, an infinite set of
105: equations for the coefficients $H_m(x;A)$ is obtained. Inspection of the
106: set indicates that if $H_m(x,A)$ is expanded in powers of $A$, it can
107: not contain powers smaller than $A^{|m|}$. From the above general
108: considerations, we have, in particular, for the first two moments,
109: \begin{eqnarray}
110: \langle x(t)\rangle_\infty &=& \sum_{n\,odd} M_n(A) e^{in\Omega t}
111: =2\sum_{n>0,odd}^\infty |M_n(A)| \cos (n \Omega t- \phi_n)\nonumber \\
112: &=&2\sum_{n>0,odd}^\infty
113: |M_n(A)| \left ( \cos \phi_n \cos n \Omega t + \sin \phi_n \sin n \Omega
114: t \right )
115: \end{eqnarray}
116: with $M_n(A)=c_n^{(0)}A^{|n|}+c_n^{(2)}A^{|n+2|}+\ldots$, and
117: \begin{equation}
118: \langle x(t)^2\rangle_\infty = \sum_{p\;even} L_p(A) e^{ip\Omega t}
119: \end{equation}
120: with $L_p(A)=b_p^{(0)}A^{|p|}+b_p^{(2)}A^{|p+2|}+\ldots$.
121:
122: The exact analytical expression for $P_{\infty}(x,t)$ is unknown. LRT
123: amounts to write
124: \begin{equation}
125: \label{appr}
126: P_{\infty}(x,t)=P_{eq}(x) + A P_1^{(1)}(x,t),
127: \end{equation}
128: with $P_1(x,t)$ obtained from a first order perturbation analysis of the
129: FPE; see \cite{junhan91} and the Appendix of \cite{gamhan98} for details
130: \footnote{Note that the plus signs in (A.23) of Ref. \cite{gamhan98}
131: should read minus. This in turn yields a minus sign on the
132: right hand side in (A.28).}
133:
134: \begin{equation}
135: \label{p1}
136: A P_1^{(1)}(x,t)=-\sum_{n=1}^\infty \frac A{\lambda_n^2+\Omega^2} \left [
137: \lambda_n \cos \Omega t + \Omega \sin \Omega t \right ] d_n \varphi_n(x),
138: \end{equation}
139: where $\varphi_n(x)$ are the right eigenstates of the unperturbed FP
140: operator and $\lambda_n$ the corresponding eigenvalues. The coefficients
141: $d_n$ are $\langle \varphi_n|\partial /\,\partial
142: x|\varphi_0\rangle$. It follows from eqs.~(\ref {appr}) and (\ref{p1})
143: that the average value $ \langle x(t) \rangle_\infty^{LRT}$ is given by
144: \begin{equation}
145: \label{avg}
146: \langle x(t)\rangle^{LRT}_\infty = a_1 \cos (\Omega t - \phi_1^{LRT}).
147: \end{equation}
148: The explicit calculation of the amplitude, $a_1$, and phase lag,
149: $\phi_1^{LRT}$, requires the knowledge of the spectrum of the unperturbed
150: system. For the bistable system at hand, no exact analytical expressions
151: for the eigenfunctions and eigenvalues exist, although useful
152: approximate expressions are known
153: \cite{hanggi82,dykman93}. Alternatively, the amplitude and phase lag
154: \cite{gamhan98,junhan93,gommor94} can
155: be obtained from the response function. Using the two-mode
156: approximation of Jung and H\"anggi \cite{junhan93}, we write
157: \begin{equation}
158: \langle x(t)\rangle^{LRT}_\infty = b_1 \cos(\Omega t- \beta_1) +b_2
159: \cos(\Omega t- \beta_2)
160: \end{equation}
161: where the first term on the right hand side is due to the interwell
162: hops, while the second one describes the influence of intrawell
163: dynamics. It is convenient to cast the expression for $\langle
164: x(t)\rangle^{LRT}_\infty$ as in eq.~(\ref{avg}), and within the two-mode
165: approximation, we get for the amplitude
166: \begin{equation}
167: \label{amp}
168: a_1= \frac AD \left [ \frac {g_1^2 \lambda_1^2}{\lambda_1^2+\Omega^2}
169: +\frac {g_2^2 \alpha^2}{\alpha^2+\Omega^2} +\frac{2g_1g_2 \lambda_1
170: \alpha (
171: \lambda_1\alpha+\Omega^2)}{(\lambda_1^2+\Omega^2)(\alpha^2+\Omega^2)}
172: \right ]^{\frac 12}
173: \end{equation}
174: while the phase lag of the response with respect to the input signal,
175: $0 \le \phi_1^{LRT} \le \pi/2$, is given by
176: \begin{equation}
177: \label{fase}
178: \phi_1^{LRT}=\arctan \frac{\frac
179: {g_1\lambda_1\Omega}{\lambda_1^2+\Omega^2}+\frac{g_2\alpha\Omega}{\alpha^2+\Omega^2}}{\frac{g_1\lambda_1^2}{\lambda_1^2+\Omega^
180: 2}+\frac{g_2\alpha^2}{\alpha^2+\Omega^2}}.
181: \end{equation}
182: In the above formulas, $\lambda_1$ is given by \cite{hantal90}
183: \begin{equation}
184: \label{lambda1}
185: \lambda_1 \approx \frac {\sqrt 2}\pi\, (1-\frac 32 D)\,\exp(-1/4D),
186: \end{equation}
187: and $\alpha=2$. The weights, $g_1$ and $g_2$ can be obtained from the
188: expressions
189: \begin{equation}
190: \label{weight2}
191: g_2= \frac {\lambda_1 \langle x^2\rangle_{eq}}{\lambda_1 -\alpha} +
192: \frac { \langle x^2\rangle_{eq} - \langle x^4\rangle_{eq}}{\lambda_1 -\alpha}
193: \end{equation}
194: \begin{equation}
195: \label{weight1}
196: g_1=\langle x^2\rangle_{eq} -g_2
197: \end{equation}
198: To leading order in $D$, we can replace $\lambda_1$ by $\lambda_K=\sqrt
199: 2/\pi \exp(-1/4D)$, $g_1 \approx 1$ and $g_2 \approx D/ \alpha$. This is
200: the limit considered in \cite{gang92}.
201:
202: Linear response theory leads to the following predictions: the first
203: moment $\langle x(t)\rangle_\infty$ should contain a single harmonics
204: with the frequency of the driving force, the output amplitude should
205: behave linearly with $A$. Certainly, for finite values of $D$, if the
206: amplitude of the driving force is infinitesimally small, the expansion
207: procedure in $A$ is valid and LRT applies. The point that we want to
208: address here is that for finite small amplitudes, $A < A_T$, the value
209: of $\Omega$ has to be taken into account when applying LRT. The upper
210: limit for the values of $A$ for which LRT remains valid\footnote{ In the
211: linear response regime, the dimensionless ratio $A/D$ is assumed to obey
212: $A < D$. In the opposite singular limit, $A \gg D$, the dynamics assumes
213: universal weak noise spectral properties \cite{gamhan98,shnjun94}.}
214: depends as well on the driving frequency.
215:
216: The adiabatic approximation gives a description of the dynamics when
217: $\Omega$ is small compared to any other characteristic frequency of the
218: system. In this approach \cite{junhan91}, the probability density is
219: assumed to be given by
220: \begin{equation}
221: \label{adiab}
222: P_{ad}(x,t)= N(t) \exp \left (-\frac {U_0(x)-Ax\cos(\Omega t)}{D}\right),
223: \end{equation}
224: where $N(t)$ is the normalization constant. An analysis of the
225: corrections to the bare adiabatic approximation has been recently
226: presented by Talkner \cite{talkne99}.
227:
228: Even in the absence of driving, no exact explicit time dependent
229: analytical solution of the FPE for the model system at hand is known. We
230: have resorted to numerical solutions of eq.~(\ref{lfpe}). We
231: follow a technique based on the use of the split propagator method of
232: Feit et al. \cite{feifle82} and detailed in \cite{gommor92}. From the
233: numerical solution of the FPE we can easily obtain the time dependence
234: of $\langle x(t)\rangle_\infty$. As this is a periodic function of time,
235: its Fourier components can be obtained by numerical quadrature.
236:
237: \begin{figure}
238: \onefigure[scale=0.5]{Fig1.eps}
239: \caption{Amplitudes of the Fourier components of $\langle
240: x(t)\rangle_\infty$ for noise strength $D=0.1$ and input amplitude
241: $A=0.2$ and frequencies $\Omega=10^{-4}$ (upper panel) and $\Omega=10^{-1}$
242: (lower panel). }
243: \label{f.1}
244: \end{figure}
245:
246: In Fig.~\ref{f.1}, we show the amplitudes of the relevant Fourier
247: components of the output signal for $D=0.1$, $A=0.2$ and two very
248: different driving frequencies, $\Omega=10^{-1}$ and $\Omega=10^{-4}$.
249: In this figure, as well as in the subsequent ones, we have taken
250: $D=0.1$. This is a typical value and it is adequate for the validity of the
251: two-mode approximation leading to eqs.~(\ref{amp}, \ref{fase}).
252: On the horizontal axis we indicate the order of
253: the harmonics. It is clear that even for this driving amplitude,
254: relatively large in relation to its threshold value, the response of the
255: system at the larger frequency contains essentially the first
256: harmonics. On the other hand, for the small driving frequency, higher
257: order harmonics are generated. This is an indication of the failure of
258: LRT to describe the dynamics at these low frequencies, while LRT might
259: still be a good description for higher frequencies.
260:
261: \begin{figure}
262: \onefigure[scale=0.5]{Fig2.eps}
263: \caption{Time evolution of $\langle x(t)\rangle$ for $D=0.1$, $A=0.04$
264: and $\Omega=10^{-4}$ as obtained from the numerical solution of the FPE
265: (solid line), the adiabatic approximation (dotted line), the two-mode
266: LRT (dashed line) and the two-mode LRT to leading order in $D$
267: (dot-dashed line).}
268: \label{f.2}
269: \end{figure}
270:
271: In Fig.~\ref{f.2}, we depict the time evolution of $\langle x(t)\rangle$
272: obtained from the numerical solution of the FPE for $D=0.1,\, A=0.04$
273: and $\Omega=10^{-4}$. We also show the behaviors obtained using the
274: adiabatic ansatz, eq.~(\ref{adiab}), LRT within the two-mode
275: approximation, eqs.~(\ref{amp}, \ref{lambda1}, \ref{weight2},
276: \ref{weight1}) and LRT to leading order in $D$. The input signal is
277: largely amplified at this small frequency. The adiabatic result deviates
278: slightly from the numerical one at the peaks. The deviations from the
279: numerical results are larger with the LRT description. Nonetheless, the
280: two-mode LRT and the adiabatic approximations yield an acceptable
281: description of the dynamics. This is expected within the linear response
282: regime, where $ A \ll D$.
283:
284: \begin{figure}
285: \onefigure[scale=0.5]{Fig3.eps}
286: \caption{The same as in Fig.~\ref{f.2} but with $\Omega=10^{-1}$.}
287: \label{f.3}
288: \end{figure}
289:
290: In Fig.~\ref{f.3}, we show the behavior for $\Omega=10^{-1}$. It is
291: clear that the adiabatic approach yields a signal with a very large
292: amplitude and a large phase shift compared with the numerics. The
293: two-mode LRT still yields a very acceptable behavior. The same
294: qualitative features are observed in Fig.~\ref{f.4} where
295: $\Omega=1$. For this large frequency, the deviations of the two-mode LRT
296: from the numerical result are very small and they can not be noticed in
297: the plot. In Fig.~\ref{f.3}, we show the full time evolution including
298: the short transient. In Fig.~\ref{f.4}, as we consider a
299: large frequency value, we only show a few oscillations in the asymptotic
300: regime, so that the details of a cycle can be distinguished. These last three
301: figures show that for very small $A$, LRT gives a satisfactory
302: description of the system response, with deviations from the numerics
303: more pronounced as the external frequency assumes smaller values.
304:
305: \begin{figure}
306: \onefigure[scale=0.5]{Fig4.eps}
307: \caption{The same as in Fig.~\ref{f.2} but with $\Omega=1.0$. Notice
308: that due to the large value of the driving frequency, we only plot a few
309: cycles of the output in the asymptotic regime.}
310: \label{f.4}
311: \end{figure}
312: To test the validity of LRT as the input amplitude is increased, we have
313: carried out an extensive numerical analysis of the system response to
314: input signals of increasing amplitudes and different frequencies. We
315: evaluate the relative error $e_{ampl} = |A_{out}-a_1|/ A_{out}$, between
316: the output amplitude, $A_{out}$, provided by the numerics and the one
317: obtained within LRT with the two-mode approximation , $a_1$ in
318: eq.~(\ref{amp}), as a function of the input amplitude $A$. Our findings
319: are shown in Fig.~\ref{f.5}. The upper panel shows the dependence of
320: $e_{ampl}$ on $A$ for several frequencies, when the LRT is evaluated to
321: leading order in $D$, while in the lower panel, the full expressions,
322: eqs.~(\ref{amp}, \ref{lambda1}, \ref{weight2}, \ref{weight1}) have been
323: used. For relatively high frequencies, $\Omega=1.0$ (circles), and
324: $\Omega=10^{-1}$ (plus signs), the error remains small and is
325: practically constant, even for input amplitudes which are rather large
326: compared to its threshold value. On the other hand, for small values of
327: $\Omega$, $\Omega=10^{-3}$ (x) and $\Omega=10^{-4}$ (triangles), the
328: error increases drastically with the input amplitude. In particular, the
329: explicit relative errors at $D=0.1$, i.e. ($e_1$, $e_2$, $e_3$, $e_4$),
330: corresponding to the driving frequencies ($\Omega_1=10^{-4}$,
331: $\Omega_2=10^{-3}$, $\Omega_3=10^{-1}$, $\Omega_4=1.0$) respectively,
332: read for $A=0.01$: (0.028, 0.028, 0.056, 0.063); for $A=0.1$: (0.249,
333: 0.249, 0.072, 0.0659); and for $A=0.2$: (2.539, 0.797, 0.090, 0.074).
334: Thus, the output amplitude predicted by LRT at these small external
335: frequencies is very much in error, even though, for the same external
336: amplitudes and moderate-to-large frequencies, LRT predictions are still
337: adequate.
338:
339: \begin{figure}
340: \onefigure[scale=0.5]{Fig5.eps}
341: \caption{Plots of the relative error of the output amplitude, $e_{ampl}
342: = |A_{out}-a_1|/ A_{out}$, vs. input amplitude $A$ and several values of
343: the driving frequency. In the upper panel, $a_1$ is evaluated using the
344: LRT two-mode expressions to leading order in $D$, while the full
345: two-mode formulas are used in the lower panel; see eqs.~(\ref{amp},
346: \ref{lambda1}, \ref{weight2}, \ref{weight1}) in the main text. The noise
347: strength is $D=0.1$ and the frequencies are: $\Omega=1.0$ (circles),
348: $\Omega=10^{-1}$ (plus signs), $\Omega=10^{-3}$ (x) and
349: $\Omega=10^{-4}$ (triangles).}
350: \label{f.5}
351: \end{figure}
352:
353: \begin{figure}
354: \onefigure[scale=0.5]{Fig6.eps}
355: \caption{Plot of the phase lag between the average output and the driving
356: force versus the angular frequency
357: $\Omega$. With the solid line we depict $\phi_1^{LRT}$, evaluated using
358: the LRT two-mode expressions to leading order in $D$ (see
359: eq.~(\ref{fase}) in the main text). The symbols denote the numerically
360: determined values of the phase lag $\Psi$ (see text), for $A=0.01$
361: (circles), $A=0.05$ (triangles), $A=0.1$ (diamonds), and $A=0.2$ (x).
362: The noise strength is set at $D=0.1$. }
363: \label{f.6}
364: \end{figure}
365:
366: The average output lags behind the input signal with a phase shift
367: between $0$ and $\pi/2$. The value of the phase shift predicted by LRT,
368: $\phi_1^{LRT}$ given by eqs.~(\ref{avg}, \ref{fase}), is independent of
369: the driving amplitude, but depends on $D$ and $\Omega$. It starts at $0$
370: for very small frequencies, then reaches a local maximum, and tends to
371: its limiting value $\pi/2$ for very large frequencies. Its behavior for
372: the small-to-moderate frequencies considered here ($\Omega < 1 $) is
373: depicted with the solid line in Fig.~\ref{f.6} for $D=0.1$. On the other
374: hand, the phase lag of the numerical result, $\Psi$, depends on $D$, $A$
375: and $\Omega$. The $\Psi$ values plotted have been calculated from the
376: difference between the instant of times within a period, at which the
377: driving signal and the periodic output, $\langle x(t) \rangle_\infty$,
378: cross signs, i.e., the corresponding phase delay in crossing zero. In
379: Fig.~\ref{f.6}, we plot the values of $\Psi$ for several values of the
380: driving amplitude and frequency. For very small frequencies, the output
381: is almost in phase with the input for all the amplitudes considered. As
382: the frequency increases, deviations between the numerical predictions
383: and LRT results are manifested, being larger for larger driving
384: amplitudes.
385:
386: In conclusion, our analysis clearly indicates the influence of the
387: driving frequency $\Omega$ on the validity of the LRT predictions for
388: the amplitude, phase and number of higher harmonics of the response of
389: the system to subthreshold input signals. As the driving frequency
390: assumes sufficiently small values, the output amplitude {\it significantly}
391: deviates from its linear behavior predicted by the two-mode
392: approximation LRT, even though the driving amplitude might still be
393: quite small in order to preserve the bistable character of the
394: unperturbed potential. Even for subthreshold inputs, higher order
395: harmonics might contribute to the system response for small driving
396: frequencies, contrary to the predictions of LRT. Although the
397: global behavior of the phase lag indicated by LRT is qualitatively
398: correct, as expected, its {\it quantitative} predictions are not reliable as
399: the input amplitude increases.
400:
401: \acknowledgments Support by the Direcci\'on General de Ense\~nanza
402: Superior of Spain (Project No. PB98-1120), the Junta de Andaluc\'\i a
403: (J.C.-P., J.G.-O., M.M.) and the Deutsche Forschungsgemeinschaft
404: HA1517/13-4 (P.H.) is gratefully acknowledged.
405:
406:
407: \begin{thebibliography}{0}
408:
409: \bibitem{kubo57}
410: \Name{Kubo R.}
411: \REVIEW{J. Phys. Soc. Japan}{12}{1957}{570}.
412:
413: \bibitem{Lebowitz}
414: \Name{Lebowitz J. L. \and Bergmann P. G.}
415: \REVIEW{Ann. Physik}{1}{1957}{1}.
416:
417: \bibitem{Risken}
418: \Name{Risken H.}
419: \Book{The Fokker-Planck Equation}
420: \Publ{Springer-Verlag, Berlin}
421: \Year{1984}
422: \Page{135}.
423:
424: \bibitem{junhan91}
425: \Name{Jung P. \and H\"anggi P.}
426: \REVIEW{Phys. Rev. A}{44}{1991}{8032}.
427:
428: \bibitem{Haenggi93}
429: \Name{H\"anggi P., Jung P., Zerbe C. \and Moss F.}
430: \REVIEW{J. Stat. Phys.}{70}{1993}{25}.
431:
432: \bibitem{gamhan98}
433: \Name{Gammaitoni L., H\"anggi P., Jung P. \and
434: Marchesoni F.} \REVIEW{Rev. Mod. Phys.}{70}{1998}{223}.
435:
436: \bibitem{hanggi82}
437: \Name{H\"anggi P. \and Thomas H.}
438: \REVIEW {Phys. Rep.}{88}{1982}{207} .
439:
440: \bibitem{dykman93}
441: \Name{Dykman M.I., Haken H., Gang Hu, Luchinsky D.G.,
442: Mannella R., McClintock P.V.E., Ning C.Z., Stein N.D. \and Stocks N.G.}
443: \REVIEW{Phys. Lett. A}{180}{1993} {332}.
444:
445: \bibitem{junhan93}
446: \Name{Jung P. \and H\"anggi P.}
447: \REVIEW{Z. f\"ur Physik B}{90}{1993}{255}.
448:
449:
450: \bibitem{gommor94}
451: \Name{G\'omez-Ord\'o\~nez J. \and Morillo M.}
452: \REVIEW{Phys. Rev. E}{49}{1994}{4919}.
453:
454: \bibitem{hantal90}
455: \Name{H\"anggi P., Talkner P. \and Borkovec M.}
456: \REVIEW{Rev. Mod. Phys.}{62}{1990}{251}.
457:
458: \bibitem{gang92}
459: \Name{Gang Hu, Haken H. \and Ning C.Z.}
460: \REVIEW{Phys. Lett. A}{172}{1992}{21}.
461:
462: \bibitem{shnjun94}
463: \Name{Shneidman V.A., Jung P. \and H\"anggi P.}
464: \REVIEW{Europhys. Lett.}{26}{1994}{5710}.
465:
466: \bibitem{talkne99}
467: \Name{Talkner P.} \REVIEW{New J. Phys.}{1}{1999}{4.1}.
468:
469: \bibitem{feifle82}
470: \Name{Feit M.D., Fleck J.A. Jr., \and Steiger A.}
471: \REVIEW{J. Comput. Phys.}{47}{1982}{412}.
472:
473: \bibitem{gommor92}
474: \Name{G\'omez-Ord\'o\~nez J. \and Morillo M.} \REVIEW{Physica
475: A}{183}{1992}{490}.
476:
477: \end{thebibliography}
478:
479:
480: \end{document}
481:
482:
483:
484:
485:
486:
487:
488: