1: \documentstyle[aps,pre,manuscript]{revtex}
2:
3:
4: \begin{document}
5:
6: \draft
7:
8: \title{The noise properties of stochastic processes and entropy production}
9:
10: \author{Bidhan Chandra Bag, Suman Kumar Banik and
11: Deb Shankar Ray{\footnote{e-mail : pcdsr@mahendra.iacs.res.in} }
12: }
13:
14: \address{Indian Association for the Cultivation of Science,
15: Jadavpur, Calcutta 700 032, India.}
16:
17: \date{\today}
18:
19: \maketitle
20:
21: \begin{abstract}
22: Based on a Fokker-Planck description of external Ornstein-Uhlenbeck noise
23: and cross-correlated noise processes driving a dynamical system we examine the
24: interplay of the properties of noise processes and the dissipative characteristic
25: of the dynamical system in the steady state entropy production and flux. Our
26: analysis is illustrated with appropriate examples.
27: \end{abstract}
28:
29: \pacs{PACS number(s) : 05.45.-a, 05.70.Ln, 05.20.-y}
30:
31:
32: \section{Introduction}
33: A dynamical system in contact with a reservoir has been a subject of wide
34: attention in dissipative dynamics and irreversible thermodynamics.
35: The focal theme lies on the possible link between the rate of phase space volume
36: contraction and the thermodynamically inspired quantities like entropy
37: production, entropy flux, Onsager coefficients etc.
38: \cite{hol,leb,ruelle,jou,gaspard,bb3,cohen,pat,bb1,bb2,sc1}.
39: While on the other
40: hand it has been argued that the entropy production is related to the intrinsic
41: properties of phase space structure of the dynamical systems through the
42: Lyapunov exponents \cite{bb3,cohen,pat,bb1,bb2,sc1}, the traditional wisdom asserts that entropy production
43: in a class of thermostatted Hamiltonian system
44: is defined \cite{ruelle} as the work per unit time (in the leading order) done on the system
45: by an external constraint under nonequilibrium steady state condition. Recently
46: based on a Markovian description of a stochastic process
47: Daems and Nicolis \cite{nicolis} have
48: critically analysed the two aspects from the consideration
49: of an information entropy balance equation.
50:
51: The object of the present paper is to extend the treatment to colour \cite{hr}
52: and cross-correlated noise processes \cite{jz,jz3} and to search for an appropriate
53: signature of an intrinsic interplay between the noise properties
54: of these processes and the dissipative characteristics of the dynamical system
55: in the steady state entropy production and flux. We specifically consider
56: the overall system to be open, i. e. , the noises
57: are of external origin such that they do not, in general, satisfy
58: fluctuation-dissipation (F-D) relations. Whenever possible we allow
59: ourselves to make a fair comparison with the standard results for closed
60: systems.
61:
62: The organisation of the paper is as follows: In Sec. II we
63: consider two types of external, stationary and Gaussian noise processes
64: namely, the Ornstein-Uhlenbeck and cross-correlated noise processes in terms
65: of a Fokker-Planck description and set up an entropy balance equation to
66: identify the drift term which reveales that in addition to dissipation
67: constant it contains the
68: essential properties of noise processes. Sec. III is devoted to explicit
69: examples to calculate the entropy production. The paper is concluded
70: in Sec. IV.
71:
72:
73: \section{The noise processes and entropy production}
74: \subsection{Fokker-Planck description}
75:
76:
77: \subsubsection{External Ornstein-Uhlenbeck noise processes}
78:
79: The Langevin equations of motion in phase space for an N-degree-of-freedom
80: system which is driven by the
81: external colour noise process $\eta_i$ can be written as
82:
83: \begin{eqnarray}
84: \dot {q}_i & = & \frac{\partial H}{\partial p_i} = p_i \; \;, \nonumber \\
85: \dot {p}_i & = &
86: -\frac{\partial H}{\partial q_i} -\gamma_i p_i +\eta_i, \; \;
87: \; \; \; \; i = 1 \cdots N.
88: \label{e1}
89: \end{eqnarray}
90:
91: \noindent
92: $N$ is the number of degrees of freedom of the system. $\gamma_i$ is the
93: damping constant for i-th degree of freedom.
94: $q_i, p_i$ are the corresponding co-ordinate and the
95: momentum, respectively.
96: While the presence of $\gamma_i$ imparts a dissipative character in the
97: dynamics, the stochastic forcing $\eta_i$ ensures a canonical
98: distribution at equilibrium when the fluctuation-dissipation relation
99: is satisfied. $H$ is Hamiltonian
100: of an initially the conservative system and is given by
101: \begin{equation}
102: H = \sum_{i=1}^N \frac{p_i^2}{2 } +V(\{q_i\}, t) \; \;.
103: \end{equation}
104:
105: \noindent
106: The masses of all the degrees of freedom have been set to unity.
107: $V(\{q_i\}, t)$ is the potential of the Hamiltonian system.
108:
109: The term $\eta_i$ in Eq. (1) refers to an external, Gaussian colour noise
110: for the i-th degree of freedom and follows the two time correlation function
111:
112: \begin{equation}
113: \langle \eta_i(t) \eta_i(t') \rangle = \frac{D_i^0}{\tau_i}
114: e^{-\frac{|t-t'|}{\tau_i}}
115: \end{equation}
116:
117: \noindent
118: where $\tau_i$ is the correlation time and $D_i^0$ is the noise strength.
119: The time evolution of $\eta_i$ can be conveniently expressed in terms of
120: the white noise process $\zeta_i (t)$-for the i-th component
121:
122: \begin{eqnarray*}
123: \dot{\eta_i} & = &-\frac{\eta_i}{\tau_i} + \frac{\sqrt{D_i^0}}{\tau_i} \zeta_i
124: \end{eqnarray*}
125: \begin{eqnarray}
126: \langle \zeta_i(t) \zeta_i(t') \rangle & = & 2 \delta (t-t') \nonumber\\
127: \langle \zeta_i\rangle & =& 0 \; \;.
128: \end{eqnarray}
129:
130: In case there exists no fluctuation-dissipation relation between $\gamma_i$ and $\eta_i$
131: the system described by the Eq. (1) is sometimes termed as
132: thermodynamically open \cite{lin}.
133:
134: Eq.(4) implies that $\eta_i$ can be treated as a phase space variable on the
135: same footing as $q_i$, $p_i$. Thus the original $2N$ dimensional stochastic
136: system (1, 3) now becomes a $3N$ dimensional Markovian process where Eq. (1)
137: and (4) are written in a compact form
138:
139: \begin{equation}
140: \dot{X_i} = F_i(X) + \zeta_i
141: \end{equation}
142: where
143: \begin{eqnarray*}
144: X_i \left\{
145: \begin{array}{l}
146: = q_i \; \; \; {\rm for} \; i = 1, \cdots , N \\
147: = p_i \; \; \; {\rm for} \; i = N+1, \cdots , 2N \\
148: = \eta_i \; \; \; {\rm for} \; i = 2N+1, \cdots , 3N \\
149: \end{array}\right.
150: \end{eqnarray*}
151: \begin{eqnarray*}
152: F_i & = & X_{i+N} \; \; \; {\rm for} \; i = 1, \cdots , N \nonumber\\
153: F_i & = & - \frac{\partial V(X_1, \cdots, X_N)}{\partial X_i} -\gamma_i X_i
154: + X_{i+N} \; \; \; {\rm for} \; i = N+1, \cdots , 2N \nonumber\\
155: F_i & = & -\frac{X_i}{\tau_i} + \frac{\sqrt {D_i^0}}{\tau_i} \zeta_i
156: \; \; \; {\rm for} \; i = 2N+1, \cdots , 3N
157: \end{eqnarray*}
158:
159: \noindent
160: and
161:
162: \begin{eqnarray}
163: \langle \zeta_i(t) \zeta_i(t') \rangle = 0 \; \; \; {\rm for} \; i = 1, \cdots , 2N \nonumber\\
164: \langle \zeta_i(t) \zeta_i(t') \rangle = 2 \delta (t-t') \; \; \; {\rm for} \; i = 2N+1, \cdots , 3N
165: \end{eqnarray}
166:
167: The Fokker-Planck equation \cite{hr} corresponding to Langevin Eq.(5) can be written as
168:
169: \begin{equation}
170: \frac{\partial P(X, t)}{\partial t} = - \sum_{i=1}^{3N} \frac{\partial}{\partial X_i}(F_i P)
171: + \sum_{i=2N+1}^{3N} D_i
172: \frac{\partial^2 P}{\partial X_i^2 } \; \; \;,
173: \label{e7}
174: \end{equation}
175:
176: \noindent
177: where $D_i = \frac{D_i^0}{\tau_i^2}$.
178:
179: $P(X, t)$ is the extended phase space probability distribution function. The
180: extension is due to the inclusion of $N$ noise variables due to
181: the external agency as phase variables. We conclude by pointing
182: out that the above formulation contains the thermodynamically closed
183: system as a special case where the internal noise strength $D_i^0$ is related to
184: dissipation $\gamma_i$
185: through the relation $D_i^0 = \gamma_i kT$, where $T$ refers to the equilibrium
186: temperature of the reservoir.
187:
188: \subsubsection{Cross-correlated noise processes}
189:
190: Next we consider a dynamical system driven by both additive and
191: multiplicative noise processes $\eta_i$ and $\zeta_i$,
192: respectively. The Langevin equation for this process, in general, can be written as
193:
194: \begin{equation}
195: \dot{X_i} = L_i(\{X_i\}, t) + g_i(X_i) \zeta_i +\eta_i \; \; \; i=1, \cdots ,N
196: \end{equation}
197:
198: \noindent
199: $L_i$ contains the dissipative term as well as the external applied
200: deterministic force, if any. $g_i(X_i)$ is the
201: coupling between the system and the multiplicative process.
202: $\zeta_i$ and $\eta_i$ are white, Gaussian
203: noise processes with the following correlation between them;
204:
205: \begin{eqnarray}
206: \langle \zeta_i(t) \zeta_j(t') \rangle
207: & = & 2 D_{ij}' \delta (t-t') \delta_{ij} \nonumber\\
208: \langle \eta_i(t) \eta_j(t') \rangle & = &
209: 2 \alpha_{ij} \delta (t-t') \delta_{ij} \nonumber\\
210: \langle \zeta_i(t) \eta_j(t') \rangle & = &
211: \langle \zeta_i(t') \eta_j(t) \rangle =
212: 2 \lambda_{ij} \sqrt{D_{ij}'\alpha_{ij}} \delta (t-t') \delta_{ij}
213: \end{eqnarray}
214:
215: \noindent
216: $D_{ij}'$ and $\alpha_{ij}$ correspond to the strength of multiplicative
217: and additive noises, respectively. $\lambda$ represents the cross correlation
218: between them with the limit $0\le \lambda \le 1$. The cross correlation between
219: these noise processes is known to cause symmetry breaking leading to
220: non-equilibrium phase transitions \cite{jz} in spatially extended systems and generate
221: interesting ratchet motion \cite{jz3} in systems with symmetric potential under
222: isothermal condition.
223:
224: The Fokker-Planck equation corresponding to Langevin Eq.(8) can be written as
225:
226: \begin{equation}
227: \frac{\partial P(X)}{\partial t} = - \sum_{i=1}^{N} \frac{\partial}{\partial X_i}(F_i P)
228: + \sum_{i=1}^{N} D_i
229: \frac{\partial^2 P}{\partial X_i^2 } \; \; \;.
230: \label{e10}
231: \end{equation}
232:
233: \noindent
234: where the drift for the $i$-th component $F_i$ is
235:
236: \begin{equation}
237: F_i = L_i(\{X_i\}, t) + \nu \left[ D_{ii}' \frac{\partial g_i(X_i)}{\partial X_i}
238: g_i(X_i) + \lambda_{ii} \sqrt{\alpha_{ii} D_{ii}'}\right] +
239: D_{ii}' \frac{\partial g_i^2(X_i)}{\partial X_i} +
240: 2 \lambda_{ii} \sqrt{\alpha_{ii} D_{ii}'} \frac{\partial g_i(X_i)}{\partial X_i}
241: \label{e11}
242: \end{equation}
243:
244: \noindent
245: $\nu =1$ stands for the Stratonovich and $\nu=0$ for the Ito convention, respectively.
246: Diffusion coefficient $D_i$ within small noise approximation can be written as
247:
248: \begin{equation}
249: D_i = \alpha_{ii} + D_{ii}' g_i^2(X_{ie}) + 2 \lambda_{ii} \sqrt{\alpha_{ii} D_{ii}'} g_i(X_{ie})
250: \label{e12}
251: \end{equation}
252:
253: \noindent
254: where $e$ in $X_{ie}$ refers to the steady state value of $X_i$, i. e., $X_{ie}$
255: is a solution of
256:
257: \begin{equation}
258: F_i(\{X_i\}) = 0 \; \; \; i = 1, \cdots , N
259: \label{e13}
260: \end{equation}
261:
262: The choice of specific forms of nonlinearity in $L_i(\{X_i\}, t)$ results
263: in typical features of nonequilibrium phase transitions in model systems.
264: For the present purpose, however, we retain a general structure for
265: the rest of the treatment.
266:
267: \subsection{Information Entropy production}
268:
269: Information entropy $S$ is formally defined in terms of the phase space
270: distribution function $P(X, t)$ through the well-known relation
271:
272: \begin{equation}
273: S = -\int dX P(X,t) \ln P(X,t) \; \;.
274: \label{e14}
275: \end{equation}
276:
277: The above definition allows us to have an evolution equation for entropy.
278: To this end we observe from Eqs.(\ref{e10}) (or (7)) and (\ref{e14}) that
279:
280: \begin{equation}
281: \frac{dS}{dt} = -\int dX \left[ - \sum_i \frac{\partial}{\partial X_i}(F_i P)
282: + \sum_i D_i
283: \frac{\partial^2 P}{\partial X_i^2}
284: \right] \ln P \; \; \;.
285: \label{e15}
286: \end{equation}
287:
288: Performing partial integration of the right hand side of the above Eq.(15)
289: and then dropping boundary terms (since the probability density tends
290: to zero as $|X|\rightarrow\infty$), one obtains the following form of
291: information entropy balance:
292:
293: \begin{equation}
294: \frac{dS}{dt} = \int dX P \nabla_X \cdot F
295: + \sum_i D_i \int \frac{1}{P}
296: \left(\frac{\partial P}{\partial X_i}\right)^2 \; \; \;.
297: \label{e16}
298: \end{equation}
299:
300: The first term in (\ref{e16}) has no definite sign while
301: the second term is positive
302: definite because of positive definiteness of $D_i$. Therefore the
303: second one can be identified as the entropy production ($\dot{S}_0$) \cite{nicolis}
304: \begin{equation}
305: \dot{S}_0 = \sum_i D_i \int \frac{1}{P_s}
306: \left(\frac{\partial P_s}{\partial X_i}\right)^2 dX
307: \label{e17}
308: \end{equation}
309:
310: \noindent
311: in the steady state. The subscript $s$
312: of $P_s$ refers to steady state. It is therefore evident from Eq.(\ref{e17}) that
313:
314: \begin{eqnarray}
315: \dot{S}_{flux} & = & \int dX \; P_s(X) \; \nabla_X \cdot F \; \;
316: = {\overline{\nabla_X \cdot F} } \nonumber\\
317: \dot{S}_0 & = & -\dot{S}_{flux} \; \;.
318: \label{e18}
319: \end{eqnarray}
320:
321: \noindent
322: Note that since we consider the system to be dissipative
323: $\overline{\nabla_X . F}$ is negative and therefore $\dot{S}_0$ turns out to
324: be positive.
325:
326:
327: \subsection{Influence of external perturbation}
328:
329: It is now interesting to examine the entropy production when the dissipative
330: system is thrown away from the steady state due to an
331: additional weak applied force. To this end we consider the drift $F_1$ due to
332: external force
333: so that the total drift $F$ has now two contributions:
334:
335: \begin{equation}
336: F(X) = F_0(X) + h F_1(X) \; \;.
337: \label{e19}
338: \end{equation}
339:
340: When $h = 0$, $P = P_s$. The deviation of $P$ from $P_s$
341: in presence of nonzero
342: small $h$ can be explicitly taken into account once we make use of the identity
343: for the diffusion term \cite{nicolis}
344:
345: \begin{equation}
346: \frac{\partial^2 P}{\partial X_i^2} = \frac{\partial}{\partial X_i}
347: \left[ P \frac{\partial \ln P_s}{\partial X_i}\right] +
348: \frac{\partial}{\partial X_i}\left[ P_s \frac{\partial}{\partial X_i}\frac{P}{P_s}\right]
349: \; \; .
350: \label{e20}
351: \end{equation}
352:
353: When $P = P_s$ the second term in
354: (\ref{e20}) vanishes. In presence of additional
355: forcing the Eq.(\ref{e10}) becomes,
356: \begin{equation}
357: \frac{\partial P}{\partial t} = - \nabla_X . \psi P - h \nabla_X . F_1 P
358: + \sum_i D_i \frac{\partial}{\partial X_i} \left( P_s \frac
359: {\partial}{\partial X_i}\frac{P}{P_s} \right)
360: \label{e21}
361: \end{equation}
362: where $\psi$ is defined as
363: \begin{equation}
364: \psi = F_{0} - \sum_i D_i \frac{\partial \ln P_s}{\partial X_i}
365: \; \; .
366: \label{e22}
367: \end{equation}
368:
369: Here we have assumed for simplicity that $D_i$ is not
370: affected by the additional forcing. The leading order influence
371: is taken into account through the additional drift term in Eq.(\ref{e21}).
372:
373: Under steady state condition $(P = P_s)$ and $h = 0$, the second and
374: the third
375: terms in (21) vanish yielding
376:
377: \begin{equation}
378: \nabla_X . \psi P_s = 0 \; \;.
379: \label{e23}
380: \end{equation}
381:
382: It is immediately apparent that $\psi P_s$ refers to a current ${\cal J}$
383: where ${\cal J} = \psi P_s$. The steady state condition therefore reduces to
384: an equilibrium condition (${\cal J}=0$) if
385: \begin{equation}
386: \psi = 0 \; \; .
387: \label{e24}
388: \end{equation}
389: (In the next section we shall consider two explicit examples to show that $\psi=0$).
390: This suggests a formal relation between $F_0$ and $D_i$ as
391: \begin{equation}
392: F_0 = \sum_i D_i \frac{\partial \ln P_s}{\partial X_i}
393: \label{e25}
394: \end{equation}
395: where $P_s$ may now be referred to as the {\it equilibrium} density function
396: in phase space.
397: $F_0$ contains dissipation constant $\gamma$. Depending on the problem it
398: also depends on the correlation time $\tau_i$ of the colour noise or on the cross
399: correlation $\lambda_{ii}$ between the noise processes.
400:
401: To consider the information entropy
402: balance equation in presence of external forcing
403: we first differentiate
404: Eq.(\ref{e14}) with respect to time and use Eq.(\ref{e21}).
405: Following Ref. \cite{nicolis}
406: one can show that in the new steady state
407: (in presence of $h \ne 0$), the entropy production ($\dot{S}_h$)
408: and the flux ($\dot{\Delta S}_{flux}$) like terms balance each other as follows:
409: \begin{equation}
410: \dot{S}_h = - \dot{\Delta S}_{flux}
411: \label{e26}
412: \end{equation}
413: with
414: \begin{equation}
415: \dot{S}_h = \sum_{i,j} {\cal D}_{ij} \int dX P\left(
416: \frac{\partial}{\partial X_i}
417: \ln \frac{P}{P_S}\right)^2
418: \label{e27}
419: \end{equation}
420: and
421: \begin{equation}
422: \dot{\Delta S}_{flux} = h^2 \int dX \delta P \nabla_X . F_1 +h^2
423: \int dX \left( \sum_i F_{1i} \frac{\partial \ln P_S}{\partial X_i}\right)
424: \delta P \; \; \;.
425: \label{e30}
426: \end{equation}
427:
428: \noindent
429: where we have put $h\delta P = P-P_s$.
430:
431: In the following section we shall work out the specific cases to provide
432: explicit expressions for the entropy production
433: and some related quantities due to external forcing for different kinds of open
434: systems mentioned in the last section.
435:
436: \section{Applications}
437:
438: \subsection{Entropy production in a system driven by an external colour noise}
439:
440: To illustrate the theory we now consider a damped
441: harmonic oscillator driven by an external, Gaussian Ornstein-Uhlenbeck
442: noise, $\eta_1$. The noise correlation of $\eta_1$ is given by Eq.(30).
443:
444: \begin{eqnarray}
445: \dot{q_1} &=& p_1 \nonumber\\
446: \dot{p_1} &=& -\omega_0^2 q_1 -\gamma + \eta_1
447: \end{eqnarray}
448:
449: \noindent
450:
451: \begin{equation}
452: \langle \eta_1 (t) \eta_1 (t') \rangle = \frac{D^0}{\tau}
453: e^{-\frac{|t-t'|}{\tau}}
454: \end{equation}
455:
456: \noindent
457: where $\omega_0$ is the frequency of the oscillator.
458:
459: To make notation consistent with Eq.(5) we would like to let $X_1$, $X_2$
460: and $X_3$ correspond to $q_1$, $p_1$ and $\eta_1$ respectively.
461:
462: The relevant equations of motion are therefore as follows
463:
464: \begin{eqnarray}
465: \dot{X_1} &=& F_1 = X_2 \; \; \; , \nonumber\\
466: \dot{X_2} &=& F_2 = - \omega_0^2 X_1 -\gamma X_2 + X_3 \; \; \; , \nonumber\\
467: \dot{X_3} &=& F_3 = - \frac{X_3}{\tau} + \frac{\sqrt{D^0}}{\tau} \zeta_3 \; \;,
468: \end{eqnarray}
469:
470: \noindent
471: where $\zeta_3$ is a $\delta$-correlated noise
472:
473: \begin{eqnarray*}
474: \langle \zeta_3 (t) \zeta_3 (t') \rangle = 2 \delta (t-t') \; \; \;.
475: \end{eqnarray*}
476:
477: Therefore for the Langevin Eq. (31) the Fokker-Planck Eq.(7) becomes
478:
479: \begin{eqnarray}
480: \frac{\partial P}{\partial t} & = & -X_2 \frac{\partial P}{\partial X_1}
481: + \frac{\partial}{\partial X_2} (\omega_0^2 X_1 + \gamma X_2 -X_3) P +
482: +\frac{1}{\tau} \frac{\partial}{\partial X_3} (X_3 P) +
483: \frac{D^0}{\tau^2} \frac{\partial^2 P}{\partial X_3^2}
484: \end{eqnarray}
485:
486: We now use the following transformation
487: \begin{equation}
488: U=a X_1+ b X_2 +X_3 \; \; \;,
489: \end{equation}
490: where $a$ and $b$ are constants to be determined.
491:
492: Then under steady state condition Eq.(32) reduces to the following form:
493: \begin{equation}
494: \frac{\partial}{\partial U}(\Gamma U)P_s + D_s
495: \frac{\partial^2 P_s}{\partial U^2} = 0 \; \; ,
496: \end{equation}
497:
498: where
499: \begin{eqnarray}
500: D_s = \frac{D^0}{\tau},
501: \end{eqnarray}
502: and
503: \begin{eqnarray}
504: \Gamma U = -a X_2 + b \omega^2 X_1+ b \gamma X_2 -b X_3 + \frac{X_3}{\tau} \; \; .
505: \end{eqnarray}
506:
507: Here $\Gamma$ is again a constant to be determined. Putting (33) in
508: Eq. (36) and comparing the coefficients of $X_1$, $X_2$ and $X_3$ we find
509:
510: \begin{eqnarray}
511: \Gamma a = -\omega_0^2 b \; \; \; \; , \nonumber
512: \; \; \; \; \Gamma b = -a + b \gamma \; \;.
513: \end{eqnarray}
514: \noindent
515: and
516:
517: \begin{equation}
518: \Gamma = -b + \frac{1}{\tau}
519: \end{equation}
520:
521: The physically allowed solutions for $a$, $b$ and $\Gamma$ are as follows ;
522:
523: \begin{eqnarray*}
524: a = \frac{1}{2}(-\frac{\gamma}{2}+\frac{1}{\tau} -\frac{1}{2}
525: \sqrt{\gamma^2 -4 \omega_0^2})(-\gamma-\sqrt{\gamma^2-4\omega_0^2})
526: \end{eqnarray*}
527:
528: \begin{eqnarray*}
529: b = -\frac{\gamma}{2}+\frac{1}{\tau} -\frac{1}{2}
530: \sqrt{\gamma^2 -4 \omega_0^2}
531: \end{eqnarray*}
532:
533: \noindent
534: and
535:
536: \begin{equation}
537: \Gamma = -\frac{\gamma}{2}+\frac{1}{2}
538: \sqrt{\gamma^2 -4 \omega_0^2}
539: \end{equation}
540:
541: The stationary solution of (34) $P_s$ is then given by
542: \begin{equation}
543: P_s = N_s e^{-\frac{\Gamma U^2}{2 D_s}} \; \;.
544: \end{equation}
545:
546: Here $N_s$ is the normalization constant. By virtue of (39) $\psi$
547: corresponding to Eq.(22) is therefore
548:
549: \begin{equation}
550: \psi = \Gamma U - D_s \frac{\partial \ln P_s}{\partial U} = 0 \; \;.
551: \end{equation}
552:
553: Since $\psi P_s$ defines a current, $P_s$
554: defines a zero current situation or an equilibrium condition.
555: The equilibrium solution $P_s$ from (39) can now be used to calculate the
556: steady state entropy production as given by Eq.(17). We thus have
557:
558: \begin{equation}
559: \dot{S}_0={\cal D}_s\int_{-\infty}^\infty \frac{1}{P_s}
560: \left(\frac{\partial P_s}{\partial U}\right)^2 dU \; \; .
561: \end{equation}
562:
563: Explicit evaluation shows
564: \begin{equation}
565: \dot{S}_0 = \Gamma \; \; ,
566: \end{equation}
567:
568: \noindent
569: where $\Gamma$ is given by Eq.(38). Thus at equilibrium the entropy production
570: is inversely proportional to relaxation time of the process.
571:
572: We now introduce an additional weak forcing in the dynamics. This may
573: achieved by adding a constant external force field $f_c$ in the dynamics.
574: Eq. (31) then becomes
575:
576: \begin{eqnarray}
577: \dot{X_1} &=& X_2 \nonumber\\
578: \dot{X_2} &=& - \omega_0^2 X_1 -\gamma X_2 + f_c + X_3 \nonumber\\
579: \dot{X_3} &=& - \frac{X_3}{\tau} + \frac{\sqrt{D^0}}{\tau} \zeta_3
580: \end{eqnarray}
581:
582: Then the non-equilibrium situation (due to additional forcing, $F_{12}=f_c$)
583: corresponding to Eq.(43) is governed by
584:
585: \begin{eqnarray}
586: \frac{\partial P}{\partial t} & = & -X_2 \frac{\partial P}{\partial X_1}
587: + \frac{\partial}{\partial X_2} (\omega_0^2 X_1 + \gamma X_2 -X_3) P +
588: -\frac{\partial}{\partial X_2} (f_c P)+\frac{1}{\tau} \frac{\partial}{\partial X_3} (X_3 P) +
589: \frac{D^0}{\tau^2} \frac{\partial^2 P}{\partial X_3^2}
590: \end{eqnarray}
591:
592:
593: Using the transformation (33) again in Eq.(44) we have
594:
595: \begin{equation}
596: \frac{\partial}{\partial U}(\Gamma U)P_s -\frac{\partial}{\partial U} F_u P+ D_s
597: \frac{\partial^2 P_s}{\partial U^2} = 0 \; \; ,
598: \end{equation}
599:
600: \noindent
601: $\Gamma$ and other constants are given by the Eq. (38). Here $F_u$ is
602:
603: \begin{equation}
604: F_u = b f_c
605: \end{equation}
606:
607:
608: The now stationary solution of (45) in presence of external forcing
609: is now given by,
610:
611: \begin{equation}
612: P'_s=N'e^{-\frac{\Gamma}{2D_s}[ U^2-\frac{2F_u U}{\Gamma}]} \; \; ,
613: \end{equation}
614: where $N'$ is the normalization constant.
615:
616: We are now in a position to calculate the steady state entropy flux
617: ($\dot{\Delta S}_{flux}$) due to external forcing ($h \ne 0$) from Eq.(28)
618:
619: \begin{equation}
620: \dot{\Delta S}_{flux}=\int dX\delta P \nabla . F_1+ \int dX
621: \left(\sum_i F_{1i}\frac{d\ln P_s}{dX_i}\right) \delta P \; \;,
622: \end{equation}
623:
624: \noindent
625: putting $h =1$.
626:
627: The components of $F_1$ in $U$-space can be identified as
628:
629: \begin{equation}
630: F_{11} = F_u \; \; \; {\rm and} \; \; \nabla_U \cdot F_1 = 0
631: \end{equation}
632:
633: $ \delta P=P'_s-P_s$ denotes the deviation from the initial equilibrium
634: state due to external forcing. For normalized probability functions $P'_s$
635: and $P_s$ the first integral in (48) vanishes.
636: Thus the entropy production at steady state due to weak forcing is given by
637:
638: \begin{eqnarray*}
639: \dot{S}_h & = & -\dot{\Delta S}_{flux} \nonumber \\
640: & = & \frac{\Gamma}{D_s} \int F_u U \delta P dU \nonumber\\
641: \end{eqnarray*}
642:
643: Making use of the definition of $\delta P$ and integrating
644: explicitly we obtain
645:
646: \begin{equation}
647: \dot{S}_h = \frac{b^2 f_c^2}{D_s}
648: \end{equation}
649:
650: Putting $D_s$ from Eq.(35) and $b$ from (38) we obtain
651:
652: \begin{equation}
653: \dot{S}_h = \frac{\left[4 -4 \gamma \tau + \gamma^2 \tau^2
654: + \tau^2(\gamma^2-4 \omega_0^2) -2 \tau \sqrt{\gamma^2 -4 \omega_0^2}(
655: 2 - \gamma \tau)\right] f_c^2}{4 D^0}
656: \end{equation}
657:
658: We now examine specifically the following two limits:
659:
660:
661: \noindent
662: (i) In the Markovian limit $\tau \rightarrow 0 $ the above expression
663: reduces to the following form:
664:
665: \begin{equation}
666: \dot{S}_h = \frac{f_c^2}{D^0}
667: \end{equation}
668:
669: For the closed thermodynamic system $D^0 = \gamma kT$ which reduces the above
670: expression to the standard result for entropy production of irreversible
671: thermodynamics for Brownian oscillator.
672:
673: \noindent
674: (ii) Next we consider an interesting limiting case $\omega_0 \rightarrow 0$,
675: which implies that for a free Brownian particle we have
676:
677: \begin{eqnarray*}
678: a = 0 \; \; \;, b = \frac{1-\gamma \tau}{\tau} \; \; \; {\rm and} \; \; \; \Gamma = \gamma
679: \end{eqnarray*}
680:
681: \begin{equation}
682: \dot{S}_h = \frac{(1- \gamma \tau)^2 f_c^2}{D^0}
683: \end{equation}
684:
685: The above expression depicts an interplay of the dissipation constant $\gamma$
686: of the system and the correlation time $\tau$ of the noise in determining the
687: entropy production. Two different cases are noteworthy;
688:
689: \noindent
690: (a) $\gamma \tau < 1$ or $\tau < \frac{1}{\gamma}$:
691:
692: When relaxation time of the system greater than correlation time of external
693: noise the entropy production $\dot{S}_h$ decreases with increase of
694: $\tau$ until $\tau \ge \frac{1}{\gamma}$.
695:
696: \noindent
697: (b) $\gamma \tau > 1$ or $\tau > \frac{1}{\gamma}$:
698:
699: The entropy production $\dot{S}_h$ increases with increase of $\tau$
700: until $\tau > \frac{1}{\gamma}$. It is interesting to note that
701: in the limit $\gamma \tau = 1$ entropy production is zero. A plot of entropy
702: production in the steady state vs correlation time therefore exhibits a minimum
703: (See Fig. 1). It is thus apparent that in presence in presence of the nonequilibrium
704: constraint the properties of noise processes as well as the dynamic characteristic
705: of the system are important for entropy production.
706:
707: \subsection{Entropy production in a cross-correlated noise driven system}
708:
709: We now turn to the second case where a simple dissipative system is
710: driven by both additive and multiplicative noises.
711:
712: \begin{equation}
713: \dot{X_1} = - \gamma X_1 -\zeta_1 X_1 +\eta_1
714: \end{equation}
715:
716: Here $L_1$ in Eq.(8) corresponds to $-\gamma X_1$. The correlation between
717: the noise processes are given by,
718:
719:
720: \begin{eqnarray}
721: \langle \zeta_1(t) \zeta_1(t') \rangle &=& 2 D_{11}' \delta (t-t') \nonumber\\
722: \langle \eta_1(t) \eta_1(t') \rangle &=& 2 \alpha_{11} \delta (t-t') \nonumber\\
723: \langle \zeta_1(t) \eta_1(t') \rangle &=& \langle \zeta_1(t') \eta_1(t) \rangle =
724: 2 \lambda_{11} \sqrt{D_{11}'\alpha_{11}} \delta (t-t') \; \;, 0\le \lambda_{11} \le 1
725: \end{eqnarray}
726:
727: \noindent
728: $\lambda_{11}$ denotes the cross-correlation between the two noise processes.
729:
730: Eq.(10) for this system reduces to
731:
732: \begin{equation}
733: \frac{\partial P(X_1)}{\partial t} = -\frac{\partial (F_1 P)}{\partial X_1}
734: + D_1 \frac{\partial^2 P}{\partial X_1^2}
735: \end{equation}
736:
737: \noindent
738: where the drift term is
739:
740: \begin{equation}
741: F_1 = -(\gamma + 2 D_{11}' -\nu) X_1 + (2-\nu) \lambda_{11} \sqrt{D_{11}' \alpha_{11}}
742: \end{equation}
743:
744: \noindent
745: and
746:
747: \begin{equation}
748: D_1 = D_{11}' X_{1e}^2 -2 \lambda_{11} \sqrt{D_{11}' \alpha_{11}} X_{1e} + \alpha_{11}
749: \end{equation}
750:
751: \noindent
752: where,
753:
754: \begin{equation}
755: X_{1e} = \frac{(2-\nu) \lambda_{11} \sqrt{D_{11}' \alpha_{11}}}{\gamma + 2 D_{11}' -\nu}
756: \end{equation}
757:
758: Making use of steady state value of $X_1$, i. e. $X_{1e}$ in Eq.(58) we obtain the
759: following constant diffusion coefficient in the weak noise limit
760:
761: \begin{equation}
762: D_1 = \frac{\left[\alpha_{11} \gamma^2 + (2-\nu) D_{11}' \alpha_{11} \{(2-\nu)D_{11}'
763: +2\gamma -2 \gamma \lambda_{11}^2 -\lambda_{11}^2 (2-\nu) D_{11}' \} \right]}
764: {\Gamma'^2}
765: \end{equation}
766:
767: \noindent
768: where
769:
770: \begin{equation}
771: \Gamma' = \gamma + 2 D_{11}' -\nu
772: \end{equation}
773:
774: Now the stationary solution of Eq.(56) is given by
775:
776: \begin{equation}
777: P_s=N_1 e^{-\frac{\Gamma'}{2D_1}[ X_1^2-2\frac{2lX_1}{\Gamma'}]} \; \; ,
778: \end{equation}
779:
780: \noindent
781: where $N_1$ is the normalization constant.
782:
783: $l$ is given by
784:
785: \begin{equation}
786: l=(2-\nu)\lambda_{11} \sqrt{D_{11}' \alpha_{11}}
787: \end{equation}
788:
789: Putting (62) in Eq.(22) one may show as before that
790:
791: \begin{equation}
792: \psi = 0
793: \end{equation}
794:
795: Thus $P_s$ is an equilibrium probability distribution function.
796:
797: Using Eq.(62) in Eq.(17) we obtain the standard expression for entropy production
798: at equilibrium
799:
800: \begin{equation}
801: \dot{S}_0 =\Gamma'
802: \end{equation}
803:
804: As before $\Gamma'$ is a negative divergence of the drift term in Eq.(61).
805: Eq.(65) carries same message as in Eq. (42) but for a different
806: system. It is apparent that the cross correlated diffusion coefficient
807: $D_{11}'$ between the noise processes is as important as the dissipation factor
808: $\gamma$ that determines the steady state entropy production.
809:
810: To study the effect of additional weak forcing on the stationary system we again
811: add a constant field of force $f_e$ in the Eq.(54). Due to the additional
812: forcing ($F_{11} = f_e)$ in Eq. (54) we have
813:
814: \begin{equation}
815: \dot{X_1} =\gamma X_1 -\zeta_1 +\eta_1 +f_e
816: \end{equation}
817:
818: Then the non-equilibrium situation corresponding to Eq.(66) is given by
819:
820: \begin{equation}
821: \frac{\partial P}{\partial t} = \frac{\partial \Gamma' X_1}{\partial X_1}
822: -F_c' \frac{\partial P}{\partial X_1} +D_1 \frac{\partial^2 P}{\partial X_1^2}
823: \end{equation}
824:
825: \noindent
826: where,
827:
828: \begin{equation}
829: F_c' = f_e +l
830: \end{equation}
831:
832: Using the stationary solution of Eq.(67) in Eq.(28) as in the previous section we
833: obtain the expression for entropy production in the steady state
834:
835: \begin{equation}
836: \dot{S}_h = \frac{\left[\gamma^2 +(2-\nu)^2 D_{11}'^2 + 2\gamma (2-\nu) D_{11}'\right] f_e^2}
837: {\left[\alpha_{11} \gamma^2 +(2-\nu) D_{11}' \alpha_{11} \{(2-\nu)D_1 +2\gamma
838: -2\gamma \lambda_{11}^2 -\lambda_{11}^2(2-\nu) D_{11}'\}\right]}
839: \end{equation}
840:
841: One may recover the standard results for a closed system by switching off the
842: multiplicative noise ($D_{11}' = 0$) and implementing fluctuation-dissipation
843: relation $\alpha_{11} = \gamma kT$ in Eq.(69). We then obtain
844:
845: \begin{equation}
846: \dot{S}_h = \frac{f_e^2}{\alpha_{11}} \; \; \; =\frac{f_e^2}{\gamma k T}
847: \end{equation}
848:
849: Eq. (69) implies that for finite $D_{11}'$ entropy production is an increasing
850: function of the cross correlation (i.e $\lambda_{11}$) between the two noise
851: processes.
852:
853: \section{Conclusions}
854: In this paper we have examined the role of noise properties of stochastic
855: processes in entropy production under a steady state condition. As specific
856: cases we have considered Ornstein-Uhlenbeck noise with finite correlation
857: time and cross-correlated noises driving the dynamical system. Based on
858: an information entropy balance equation we have shown that the entropy production
859: and flux like terms not only depend on the dissipative characteristics of the dynamics
860: of the phase space of the dynamical system, particularly, the rate of phase space volume contraction,
861: but also on the correlation time and strength of cross correlation of the noises.
862: Since the steady state entropy production is identified as a drift term in the Fokker-Planck
863: description in the present formalism and the correlation time or the strength
864: of cross-correlated noises make their presence felt in this term, it is not
865: difficult to trace the origin of the role of interplay of dissipation and
866: the properties of the noise processes.
867: In view of the fact that the Ornstein-Uhlenbeck noise processes or the
868: cross-correlated
869: noise processes are commonly occurring situations in condensed matter physics
870: and chemistry, we hope that the present analysis will be useful in
871: irreversible thermodynamics in relation to dynamical systems, in general.
872:
873:
874: \acknowledgements
875: S. K. Banik is indebted to the Council of Scientific and Industrial
876: Research for partial financial support.
877:
878: \begin{thebibliography}{99}
879:
880: \bibitem{hol}
881: B. L. Holian, W. Hoover, and H. Posch,
882: Phys. Rev. Lett. {\bf 59}, 10 (1987).
883:
884: \bibitem{leb}
885: N. I. Chernov, G. L. Eyink, J. L. Lebowitz, and Ya. G. Sinai,
886: Commun. Math. Phys. {\bf 154}, 569 (1993).
887:
888: \bibitem{ruelle} D. Ruelle, J. Stat. Phys. {\bf 85}, 1 (1996).
889:
890:
891: \bibitem{jou} A. Compte and D. Jou, J. Phys. A {\bf 29}, 4321 (1996).
892:
893: \bibitem{gaspard} P. Gaspard, J. Stat. Phys. {\bf 88}, 1215 (1997).
894:
895: \bibitem{bb3} B. C. Bag, J. Ray Chaudhuri, and D. S. Ray,
896: J. Phys. A {\bf 33}, 8331 (2000).
897:
898:
899: \bibitem{cohen} D. J. Evans, E. G. D. Cohen, and G. P. Morriss,
900: Phys. Rev. A {\bf 42}, 5990 (1990).
901:
902: \bibitem{pat} A. K. Pattanayak, Phys. Rev. Lett. {\bf 83}, 4526 (1999).
903:
904: \bibitem{bb1} B. C. Bag, S. Chaudhuri, J. Ray Chaudhuri, and D. S. Ray,
905: Physica D {\bf 125}, 47 (1999).
906:
907: \bibitem{bb2} B. C. Bag and D. S. Ray, J. Stat. Phys. {\bf 96}, 271 (1999).
908:
909: \bibitem{sc1} S. Chaudhuri, G. Gangopadhyay, and D. S. Ray,
910: Phys. Rev. E {\bf 47}, 311 (1993).
911:
912: \bibitem{nicolis} D. Daems and G. Nicolis, Phys. Rev. E {\bf 59}, 4000 (1999).
913:
914: \bibitem{hr}
915: See, for example, H. Risken, {\it Fokker-Plank Equation} (Springer-Verlag,
916: Berlin, 1989).
917:
918: \bibitem{jz}
919: J. H. Li and Z. Q. Huang, Phys. Rev. E {\bf 53}, 3315 (1996).
920:
921: \bibitem{jz3}
922: J. H. Li and Z. Q. Huang, Phys. Rev. E {\bf 57}, 3917 (1998).
923:
924: \bibitem{lin}
925: See, for example, K. Lindenberg and B. J. West, {\it The Nonequlibrium
926: Statistical Mechanics of Open and Closed Systems} (VCH, New York, 1990).
927:
928: \end{thebibliography}
929:
930:
931: \begin{figure}
932: \caption{
933: Plot of entropy production $\dot{S}_h$ vs correlation time $\tau$
934: for the Eq.(53) using $\gamma =1.0$, $f_c = 1.0$ and $D^0 = 1$ (Units are arbitrary).
935: }
936: \end{figure}
937:
938: \end{document}
939: