cond-mat0501327/sdf.tex
1: \documentclass[preprint,twocolumn,tightenlines,aps,prb,10pt,floats,floatfix]{revtex4}
2: \usepackage{psfig}
3: \usepackage{epsfig}
4: \usepackage{graphicx}
5: \usepackage{dcolumn}
6: \usepackage{longtable}
7: 
8: 
9: \def\erf{\mbox{\rm{erf}}}
10: 
11: \begin{document}
12: 
13: % \title{Langevin dynamics with space-dependent friction arising from 
14: % a secondary stochastic force}
15: \title{Dissipating the Langevin equation in the 
16: presence of an external stochastic potential}
17: \author{Jeremy M. Moix}
18: \author{Rigoberto Hernandez}
19: \thanks{Author to whom correspondence should be addressed} 
20: \email{hernandez@chemistry.gatech.edu.}
21: \affiliation{Center for Computational and Molecular Science and Technology, \\
22: School of Chemistry and Biochemistry, \\
23: Georgia Institute of Technology, \\
24: Atlanta, GA  30332-0400}
25: \date{\today}
26: %
27: %
28: \begin{abstract}%
29: In the Langevin formalism, 
30: the delicate balance maintained between the fluctuations
31: in the system and their corresponding dissipation may be 
32: upset by the presence of a secondary, space-dependent stochastic force, 
33: particularly in the low friction regime. 
34: In prior work, the latter was dissipated self-consistently
35: through an additional uniform (mean-field) friction
36: [Shepherd and Hernandez, {\it J.~Chem.~Phys.},
37: {\bf 115}, 2430-2438 (2001).]
38: An alternative approach to ensure that equipartition is satisfied 
39: relies on the use of a space-dependent friction while ignoring nonlocal
40: correlations.
41: The approach is evaluated with respect to its ability to
42: maintain constant temperature for two simple one-dimensional, 
43: stochastic potentials of mean force wherein the friction can
44: be evaluated explicitly when there is no memory in the barriers.
45: %in the memory-less limit.
46: The use of a space-dependent friction is capable of 
47: providing qualitatively similar results to those obtained previously, 
48: but in extreme cases, deviations from equipartition may be observed
49: due to the neglect of the memory effects present in the stochastic potentials.
50: \end{abstract}
51: 
52: \maketitle
53: 
54: \section{Introduction}
55: 
56: In the theory of diffusion processes over fixed barriers, 
57: numerous studies have shown that the dissipative
58: term in the Langevin equation is rarely constant along the reaction 
59: coordinate.\cite{carm83,sbb88,zhu88,krishnan92,bere93,straus93b,neria96,schwartz99}
60: %\marginpar{citations?}
61: A general rate theory when the friction is both
62: space- and time-dependent has been developed to account for this
63: phenomenon over the entire friction regime.\cite{haynes94,haynes93,voth92b}
64: One might na\"{i}vely expect that a space-dependent 
65: component must be included in the friction kernel to capture the essential
66: dynamics of a given system.  
67: However, this is not always the case.  
68: Several groups have shown that the average dynamical properties
69: may still be adequately described by a 
70: generalized Langevin equation with space-independent friction
71: even when the reaction 
72: coordinate has a strong spatial dependence.\cite{straus93b,sbb88,hynes87,krishnan92}
73: %\marginpar{CITATIONS?}
74: An analysis by Haynes and Voth 
75: concluded that the key factor is not whether
76: the friction is space-dependent, since it generally will be, 
77: but rather how the friction varies 
78: along the reaction coordinate.\cite{voth95}
79: In particular, they suggest that the symmetry of the space-dependent friction
80: with respect to the barrier can be used as a metric for evaluating the 
81: role of the friction in the dynamics. 
82: Similar product and reactant states will give 
83: rise to similar (symmetric) friction
84: components about the transition state.
85: Perhaps surprisingly, 
86: an antisymmetric friction does not have a significant impact on the 
87: dynamics,
88: while a symmetric friction can result in large deviations from the 
89: predictions of standard rate theories for processes with space-independent 
90: friction.\cite{voth95,straus93b,zhu88,sbb87,krishnan92,schwartz99}
91: Thus, the Langevin model with a uniform effective friction can often 
92: approximate the dynamics of projected variables even if the formal 
93: projection would have required a space-dependent model.
94: 
95: The central question explored in this work is whether a single
96: uniform effective friction suffices even when the Langevin system is
97: subjected to an external space-dependent stochastic potential.
98: %
99: %In the present work, 
100: %the main concern is the description of the space-dependence in the 
101: %friction that may arise from fluctuations in an external
102: %(but space-dependent) stochastic potential that is symmetric
103: %about a given barrier,
104: %rather than from the space-dependent coupling to the thermal bath.
105: %With periodic potentials of mean force, 
106: %the friction is perfectly symmetric with respect to a given barrier, 
107: %and may give rise to interesting dynamical effects according to
108: %previous studies.
109: %
110: The behavior of a Brownian particle diffusing across 
111: various subsets of this class of potentials has been
112: the subject of intense research.\cite{DG92,march98,pollak93a,
113: pollak99,astumian98,VDB93,Reim95,Hang95b,lind94}
114: This activity has 
115: largely been motivated by the discovery of resonant activation
116: in which the rate of transport over a stochastic barrier
117: exhibits a maximum as a function of the correlation time in
118: the fluctuations of the barrier height.\cite{DG92}
119: However,
120: until recently, simulations of these systems have not been performed in
121: the low friction regime, where deviations from
122: equipartition may occur,
123: due to an inability to adequately
124: describe the friction in the presence of an additional
125: stochastic force.\cite{hern00c,hern02c}
126: In previous work, 
127: the dissipation of this excess energy 
128: was achieved through a self-consistent approach in which
129: the friction constant is renormalized iteratively 
130: until equipartition is satisfied.\cite{hern00c}
131: This renormalization is approximate because it does not explicitly
132: account for the correlations between the external stochastic forces 
133: across space and time, but rather uses a single mean friction
134: to dissipate theses forces at times longer than their correlation times.
135: %
136: % However, 
137: % an analytic form for this contribution to the friction 
138: % has not yet been found. 
139: % Here, 
140: A possible improvement to the self-consistent approach can be obtained 
141: by allowing the friction to be space-dependent while 
142: explicitly ignoring the memory in the stochastic potential,
143: In the special case that the stochastic potential has no memory, then
144: this treatment is exact.
145: However, this approximation is often not justified when modeling real 
146: systems and therefore, the model potentials employed are chosen to have an
147: exponentially decaying memory of their past states.  
148: In the most extreme cases, these correlations can result in deviations from 
149: equipartition during the course of the simulation, although the space-dependent 
150: friction dissipates such fluctuations correctly in most situations.
151: The general conclusion appears to be that the more detailed space-dependent
152: approach is in qualitative agreement with the self-consistent approach
153: and hence, as in the fixed barrier case, Langevin systems with 
154: stochastic forces may be dissipated by a single (though renormalized)
155: uniform friction.
156: 
157: The conclusions of this work are supported by a study of two different
158: classes of one-dimensional problems in which the particle diffuses
159: across a periodic array of coherent or incoherent barriers.
160: These two cases can be specified by sinusoidal or 
161: merged-harmonic-oscillator potentials, respectively.
162: For such simple forms of the stochastic potential, analytic expressions
163: for the friction as a function of the spatial coordinate can 
164: readily be obtained and are presented in Sec.~\ref{sec:formalism}.
165: The resulting Langevin dynamics across these potentials 
166: dissipated either uniformly or through the space-dependent
167: friction are illustrated in Sec.~\ref{sec:results}.
168: 
169: \section{Langevin Model with Stochastic Potentials}\label{sec:formalism}
170: 
171: An equation of motion describing the diffusion of a particle  
172: influenced by a stochastic potential of
173: mean force can be adequately described by a
174: phenomenological Langevin equation of the form,
175: \begin{equation}
176: \dot{v}=-\gamma(t) v
177:         +\xi(t)
178:         +F(x;t)
179:         \;, \label{eq:eqm}
180: \end{equation}
181: where $F(x;t) \equiv -\nabla_{x}U(x;t)$ is an external stochastic force,
182: and $\gamma(t)$ is the friction required to dissipate
183: both the thermal forces and those due to the external stochastic potential.
184: The thermal bath is described by $\xi(t)$, 
185: which is a Gaussian white noise source
186: with time correlation given by the 
187: fluctuation-dissipation relation (FDR),
188: \begin{equation}
189: \langle\xi(t)\xi(t^{\prime})\rangle=2k_{\rm B}T\gamma_{\rm th}
190: \delta(t-t^{\prime})
191: \;.\label{eq:xi_xi}
192: \end{equation} 
193: In the limit that $F(x;t)=F(x;0)$ for all $t$, %(what if F(x;t)=F(x))
194: these equations reduce to the 
195: Langevin equation with $\gamma(t)=\gamma_{\rm th}$.
196: Otherwise, the question remains as to what is the appropriate form
197: of $\gamma(t)$.
198: Two approaches for addressing this question are presented 
199: in Sections~\ref{sec:uniform} and~\ref{sec:sdf},
200: after first describing the explicit forms of the stochastic
201: potentials.
202: 
203: \subsection{Stochastic Potential Representation}\label{sec:pot}
204: 
205: The space-dependent friction (SDF) that arises from the 
206: fluctuations in $F(x;t)$ can readily be evaluated analytically for
207: two different classes of one-dimensional
208: stochastic potentials.
209: The first of these is a sinusoidal potential taking the 
210: general form,
211: \begin{equation}
212: U(x;t)=\left(E_{\rm b}+\frac{1}{2}\eta(t)\right)
213:        \left(\sin\left(\frac{\pi x}{2}\right)+1\right)
214: \;,\label{eq:sin}
215: \end{equation}
216: in which the barriers fluctuate {\it coherently} with each other.
217: The second is constructed using a series of merged harmonic oscillators (MHOs) 
218: in which each barrier is allowed to fluctuate independently
219: ({\it incoherently}) of one another, and is specified by
220: \begin{equation}
221: U(x,t)= \left\{\begin{array}{ll}
222:         \frac{1}{2}k_0(x-x_m^0)^2   &\mbox{for $x_m^0 < x \le x_m^-$}\\
223:         V_{m}^{\ddagger} + 
224:         \frac{1}{2}k_{m}^{\ddagger}(x-x_m^{\ddagger})^2 
225:                                     &\mbox{for $x_m^- < x \le x_m^+$}\\
226:         \frac{1}{2}k_0(x-x_{m+1}^0)^2  &\mbox{for $x_m^+ < x \le x_{m+1}^0$}
227: 
228:       \end{array}\right.
229: \;,\label{eq:mho}
230: \end{equation}
231: where the $m^{th}$ well and adjacent barrier
232: are centered at $x_m^0=-\lambda/2+m\lambda$ and 
233: $x_m^\ddagger=m\lambda$, respectively. 
234: The connection points are chosen to ensure
235: continuity in the potential and its first derivative such that
236: $x_{m}^{\pm}=\pm{k_0\lambda}/{(2k_{0}-2k_{m}^\ddagger)}+m\lambda$. 
237: As opposed to the sinusoidal potential, the width of the MHO barriers 
238: varies stochastically in time according to the relation 
239: $k_{m}^{\ddagger}=-(k_{0}+\eta(m,t))$, 
240: which, in turn, defines the barrier height
241: $V_{m}^{\ddagger}=-k_{0}k_{m}^{\ddagger}\lambda^2/(8k_{0}-8k_{m}^{\ddagger})$.
242: The remaining parameters in the  
243: potentials are chosen such that the lattice spacing is 4 and the
244: thermal energy of the particle is $1/6$ of the average value of 
245: the barrier heights.
246: %average barrier height for each is $6 k_{\rm B} T$ 
247: 
248: 
249: The stochastic term,
250: $\eta(t)$, is defined as an Ornstein-Uhlenbeck process governed by 
251: the following differential equation,
252: \begin{equation}
253: \dot{\eta}(t)=-\frac{\eta(t)}{\tau_{\rm{c}}}+
254:               \sqrt{\frac{2\sigma^2}{\tau_{\rm{c}}}}\zeta(t)
255: \;,
256: \end{equation}
257: with the probability distribution,
258: \begin{equation}
259: P\left(\eta(t)\right)=\frac{1}{\sqrt{2\pi\sigma^2}}\exp\left(-\frac{\eta(t)^2}
260:         {2\sigma^2}\right)
261: \;,\label{eq:prob}
262: \end{equation}
263: and time correlation,
264: \begin{equation}
265: \langle\eta(t)\eta(t^{\prime})\rangle=\sigma^2\exp\left(-\frac
266:                                       {|t-t^{\prime}|}{\tau_{\rm{c}}}\right)
267: \;.
268: \end{equation}
269: The variance of the distribution is given by $\sigma^2$, $\tau_{\rm{c}}$ is
270: the correlation time, and $\zeta(t)$ is an additional
271: white noise source.
272: The distribution of barriers heights for the sinusoidal potential
273: is given directly by the distribution of $\eta(t)$,
274: but due to the nature of the expression for the barrier heights of the MHOs, 
275: the resulting distribution for this potential takes on a more complex
276: form that is sharper and slightly skewed compared with Eq.~\ref{eq:prob}.
277: As a result, a much smaller range of fluctuations is allowed for the MHO
278: than the sinusoidal potential to ensure that the distribution 
279: does not become significantly non-Gaussian.  
280: More details on the exact behavior of the
281: MHO barrier heights are provided in Ref.~\onlinecite{hern00c}.
282: 
283: \subsection{Uniform Dissipation}\label{sec:uniform}
284: 
285: In previous work,\cite{hern00c} 
286: a self-consistent procedure was developed to ensure that the evolution
287: of the system using Eq.~\ref{eq:eqm} remains in thermal equilibrium.
288: This was accomplished through an iterative
289: procedure in which the friction, given by the sum of the two 
290: contributions from the thermal bath and the stochastic potential, $\it{i.e.}$
291: $\gamma \equiv  \gamma_{\rm th}+\gamma_{\rm F}$, 
292: is renormalized according to the relation,
293: \begin{equation}
294: \gamma^{(n+1)}=\gamma^{(n)}\left(\frac{\langle v^2(t) 
295:                                        \rangle_n}{k_{\rm b}T}\right)
296: \label{eq:SCF}
297: \;.
298: \end{equation}
299: The friction for the next iteration is determined from the value of the 
300: friction at the current step scaled by the magnitude of the 
301: deviation from equipartition seen in the dynamics
302: until convergence is reached to within a desired accuracy.
303: The main criticism to this approach lies in the approximation 
304: made in developing Eq.~\ref{eq:SCF} in which the stochastic potential
305: is treated as a local noise source,
306: $\gamma_{\rm F}$, obeying a fluctuation-dissipation relation
307: equivalent to Eq.~\ref{eq:xi_xi}.
308: However, the stochastic potentials have memory and are therefore 
309: nonlocal in nature leading to non-vanishing cumulants at 
310: third and higher orders.
311: These effects are included, but only in an average manner, to second order
312: in this approach.
313: 
314: \subsection{Space-Dependent Dissipation}\label{sec:sdf}
315:  
316: % An alternative approach that explicitly includes the spatial 
317: % dependence of the friction will be described.  
318: 
319: An alternative approach to dissipating the external stochastic force
320: relies on replacing the space- and time-dependent friction, $\gamma(x,t)$, 
321: by a space-dependent friction, $\gamma(x(t))$, satisfying a local 
322: FDR.
323: Given that the size of the fluctuations in $F(x;t)$
324: depend on $x$ at a given $t$,
325: a Brownian particle moving quickly across the surface will experience
326: a series of forces whose relative magnitudes depend on the particle's velocity.
327: However when the the Brownian particle moves slowly, 
328: the particle will sample only the local fluctuations of the
329: stochastic potential in the vicinity of its local position $x$.
330: In this regime, the particle arrives at a local quasi-equilibrium  
331: which must necessarily satisfy the FDR locally. 
332: This suggests that the dissipation should not be uniform, but rather
333: should depend on position, and therefore indirectly on time.
334: It should be noted that while the mean-field approach 
335: described in the previous subsection
336: is capable of including the average of the 
337: correlations between the fluctuations, 
338: the approximation made here does not account for any of the memory effects. 
339: However, in the limit that there is no 
340: memory in the external stochastic potential,
341: %\tau_{\rm{c}}=0$, 
342: the following results are exact.
343: 
344: The question now arises of how to explicitly describe the friction constant in 
345: the presence of an additional fluctuating force resulting from the
346: potentials of mean force given in Eqns.~\ref{eq:sin} and ~\ref{eq:mho}.  
347: The friction constant must dissipate the excess energy 
348: that arises from the fluctuating forces through a local 
349: space-dependent FDR,
350: \begin{equation}
351: 2k_{\rm B}T\gamma_{\rm c}(x;t)=\langle\delta F_{\rm c}(x;t)^2\rangle
352: \;,\label{eq:FDT}
353: \end{equation}
354: where the cumulative force is simply the sum of the thermal Gaussian noise and
355: the stochastic force arising from the external potential, 
356: $F_{\rm c}=F_{\rm th}+F_{\rm U}$. 
357: Assuming the respective 
358: fluctuations in the bath and the potential are uncorrelated, {\it i.e.} 
359: $\langle\delta F_{\rm th} \delta F_{\rm U} \rangle=0$,
360: then Eq.~(\ref{eq:FDT}), reduces to 
361: \begin{equation}
362: 2 k_{\rm B}T\gamma_{\rm c}(x;t)=\langle\xi(t)\xi(t^\prime)\rangle+
363:                                 \langle \delta F_{\rm U}(x;t)^2 \rangle
364: \;,
365: \label{eq:fric}
366: \end{equation} 
367: %where the time dependent fluctuations in 
368: %the potential are implicitly contained in $\eta$.  (redundant)
369: The thermal fluctuations 
370: are ohmic as given in Eq.~\ref{eq:xi_xi}, 
371: and the relationship for the fluctuations in the force is
372: $\delta F_{\rm U}(x;t) \equiv 
373:  F_{\rm U}(x;t)-\langle F_{\rm U}(x;t)\rangle_{\eta}$, 
374: where the average is taken with respect to the auxiliary stochastic
375: variable, $\eta$.
376: The average value of the force can be determined according 
377: to the usual integrals, 
378: \begin{equation}
379: \langle F_{\rm U}(x;t)\rangle=
380:                              \frac{-\int_{-\infty}^{\infty}\!d\eta\, P(\eta)
381:                              \nabla_x U(x;t) }
382:                              {\int_{-\infty}^{\infty}\!d\eta\, P(\eta)}
383: \;,
384: \end{equation}
385: where the fluctuations in the force are governed by the stochastic 
386: Ornstein-Uhlenbeck process, $\eta$, whose probability distribution is 
387: given by Eq.~\ref{eq:prob}. 
388: 
389: The remaining steps of the derivation rely
390: upon the specific form of the potential.
391: As an illustration, the SDF is evaluated explicitly below
392: for the simpler sinusoidal (coherent) stochastic potential.
393: (The results for the incoherent MHO potential can be found in the Appendix.)
394: The derivation begins by direct evaluation of Eq.~\ref{eq:fric}
395: for the specific class of potentials.  
396: As remarked above, the first term reproduces 
397: the FDR, Eq.~\ref{eq:xi_xi}, for the thermal forces.
398: Ignoring the correlation in the forces at different times, the second 
399: reduces to:
400: \begin{eqnarray}
401: \langle \delta F_{\rm U}(x;t)^2 \rangle & = &
402:                     \frac{\pi^2}{4}\cos^2\left(\frac{\pi x}{2};t\right)
403:                     \times \nonumber \\
404:              &   &  \int_{-\infty}^\infty\!d\eta\,
405:                     \left(E_{\rm b}+\frac{1}{2}\eta\right)^2P(\eta)- \nonumber\\
406:              &   &  \left[\frac{\pi}{2}\cos\left(\frac{\pi x}{2};t\right)
407:                     \right.\times \nonumber \\
408:              &   &  \left. \int_{-\infty}^{\infty}\!d\eta\, 
409:                     \left(E_{\rm b}+\frac{1}{2}\eta\right)P(\eta)\right]^2
410: \;.
411: \end{eqnarray}
412: The Gaussian integrals are readily evaluated to yield:
413: \begin{equation}
414: \langle \delta F_{\rm U}(x;t)^2 \rangle=\frac{\sigma^2\pi^2}{16}
415:                                         \cos^2(\frac{\pi x}{2};t)
416: \;.
417: \end{equation}
418: Upon substitution into Eq.~\ref{eq:FDT},
419: the explicit form of the SDF is
420: \begin{equation}
421: \gamma_{\rm c}(x;t)=\gamma_{\rm th}\delta(t-t^\prime)+
422:                    \frac{\sigma^2\pi^2}{32k_{\rm B}T}\cos^2(\frac{\pi x}{2};t)
423: \label{eq:sdfsin}
424: \;.
425: \end{equation}
426: This is the simplest possible form for this result,
427: and is due to the separability of the potential
428: into a sum of deterministic and linear stochastic terms.  
429: In fact, it is easily shown that for any separable potential 
430: of the form,
431: \begin{equation}
432: U(x;t)=\bar{U}(x)+\eta(t)W(x)
433: \;,\label{eq:separable}
434: \end{equation}
435: where $\bar{U}(x)$ is the deterministic 
436: component of the potential of mean force,
437: then the additional friction due
438: to the stochastic potential is given by
439: \begin{equation}
440: \langle \delta F_{\rm U}(x;t)^2 \rangle=\left(\nabla_x W(x)\right)^2
441:                                         \int\!d\eta\, (\eta^2-\eta) P(\eta)
442: \;,
443: \label{eq:MomentGen}
444: \end{equation} 
445: provided the distribution is normalized.
446: The MHO does not satisfy the condition of Eq.~\ref{eq:separable}
447: and hence its friction correction can not be
448: obtained by Eq.~\ref{eq:MomentGen}.
449: The form of the friction correction for the MHO consequently
450: contains more terms, but the requisite approximation (that the
451: forces are uncorrelated at different times) enters the
452: derivation in a conceptually equivalent way.
453: 
454: \subsection{Mean First-Passage Times}
455: 
456: The dynamics of the system were characterized by the mean first passage
457: time (MFPT) of a particle to escape its initial minima and establish
458: a quasi-equilibrium within another well.  With periodic, stochastic 
459: potentials, this may be accomplished by defining a region
460: of the phase space of the particle bounded by an 
461: energetic constraint.\cite{hern02b}
462: The MFPT
463: is simply the average of a sufficient number of first passage processes into
464: this region,
465: with the corresponding rate given by the inverse of the MFPT.
466: While the incorporation of a space-dependent friction in the algorithm for
467: the numerical integration of the equations of motion would seemingly result
468: in a dramatic increase in computational expense, 
469: the actual effort is comparable to the previous mean-field approach because the 
470: preliminary convergence procedure for the friction constant is now unnecessary.
471: 
472: \section{Results and Discussion}\label{sec:results}
473: 
474: The analytic and numerical space-dependent components of the friction
475: over one period of the MHO and sinusoidal potentials can be seen in the bottom
476: panel of Figs.~\ref{fig:sdfmho} and 
477: ~\ref{fig:sdfsin}, respectively, with the numerical results
478: averaged over 500 representative trajectories.
479: 
480: \begin{figure}[t]
481: \begin{center}
482: \includegraphics*[width=7.5cm]{./mhofluc.eps}
483: \end{center}
484: \caption{
485: Representative fluctuations over one period of the MHO
486: potential and force (top panel),
487: and the resulting space-dependent friction (bottom panel).
488: The numerical component in the bottom panel
489: is displayed as the solid black line, with
490: the analytic result, given in the Appendix, as the dotted white line.
491: The temperature is 2/3, the variance
492: is 0.22, and the thermal friction is 0.08.
493: }
494: \label{fig:sdfmho}
495: \end{figure}
496: %%%%%%%%%%%%%%%%%%%%%%
497: %\newpage
498:                                                                                 
499: %%%%%%%%%%%%%%%%%%%%%%%
500: \begin{figure}[t]
501: \begin{center}
502: \includegraphics*[width=7.5cm]{./sinfluc.eps}
503: \end{center}
504: \caption{
505: Representative fluctuations over one period of the sinusoidal
506: potential and force (top panel),
507: and the resulting space-dependent friction (bottom panel).
508: The numerical result is displayed as the solid black line,
509: with the analytic result, given by Eq.~\ref{eq:sdfsin}, shown as
510: the dotted white line.
511: The parameters used are the same as in Fig.~\ref{fig:sdfmho}.
512: }
513: \label{fig:sdfsin}
514: \end{figure}
515: %%%%%%%%%%%%%%%%%%%%%%%
516: %/newpage
517:  
518: The top panels display the fluctuations in the potential and the resulting
519: forces that give rise to the space-dependent friction.
520: The analytic forms of the SDF, displayed as the dotted white line, 
521: agree with the corresponding numerical results, and exact agreement
522: is obtained upon further averaging.  
523: The fluctuations in the forces reach
524: a maximum at approximately the midpoint between the minima and maxima, 
525: where deviations from the average force take on the largest values.  
526: The fluctuations in the potential are largest at the barriers, 
527: while the forces are zero at these locations.
528: This leads to a vanishing contribution to the total friction 
529: from the space-dependent component at these points.  
530: In the well region, the behavior of the SDF 
531: for the sinusoidal and MHO potentials is inherently different.  
532: The SDF for the MHO is zero outside of the barrier 
533: region since the wells do not fluctuate by construction.
534: However, the sinusoidal potential fluctuates continuously throughout
535: leading to a friction correction along the entire reaction coordinate.  
536: Consequently, the magnitude of the friction correction 
537: in simulations employing the sinusoidal potential are
538: slightly larger than that in those employing the MHO.
539: But, as illustrated below,
540: this effect does not have a dramatic effect on the resulting dynamics.  
541: 
542: Values of the friction corrections 
543: calculated from the iterative and space-dependent
544: approaches for the MHO and sinusoidal potentials are displayed
545: in Table~\ref{tb:mho} with  
546: the values of the thermal friction listed in the left-most column.
547: \begin{table*}[ht]
548: \begin{center}
549: \begin{tabular}{ccccccccccccc}
550: \hline \\[-9truept]
551: \hline \\[-8truept]
552: \multicolumn{1}{c}{} &
553: \multicolumn{3}{c}{
554:     \underline{$
555:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
556:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
557:                  \rm{MHO}
558:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
559:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
560:                $}
561: }&
562: \multicolumn{3}{c}{
563:    \underline{$
564:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
565:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
566:                  \rm{Sin}
567:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
568:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
569:    $}
570: }\\
571: %\multicolumn{3}{c}{
572: %   \underline{$
573: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
574: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
575: %                 \tau_{\rm c}=10^1
576: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
577: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
578: %   $}
579: %}\\
580:                                                                                 
581: $\gamma_{\rm th}$ &
582:                                                                                 
583: $ \mbox{ } \mbox{ }  \langle \gamma_{\rm{F}} \rangle_{\rm{0}}
584:   \mbox{ } \mbox{ }$ &
585: $ \mbox{ } \mbox{ }  \langle \gamma_{\rm{F}}\rangle_{\rm{sdf}}
586:   \mbox{ } \mbox{ }$ &
587: $ \mbox{ } \mbox{ }  \langle v^2 \rangle_{\rm{sdf}}
588:   \mbox{ } \mbox{ }$ &
589:                                                                                 
590: $ \mbox{ } \mbox{ }  \langle \gamma_{\rm{F}}\rangle_{\rm{0}}
591:   \mbox{ } \mbox{ }$ &
592: $ \mbox{ } \mbox{ }  \langle \gamma_{\rm{F}}\rangle_{\rm{sdf}}
593:   \mbox{ } \mbox{ }$ &
594: $ \mbox{ } \mbox{ }  \langle v^2 \rangle_{\rm{sdf}}
595:   \mbox{ } \mbox{ }$ \\
596:                                                                                 
597: %$ \mbox{ } \mbox{ }  \langle \gamma_{\rm{F}}\rangle_{\rm{0}}
598: %  \mbox{ } \mbox{ }$ &
599: %$ \mbox{ } \mbox{ }  \langle \gamma_{\rm{F}}\rangle_{\rm{sdf}}
600: %  \mbox{ } \mbox{ }$ &
601: %$ \mbox{ } \mbox{ }  \langle v^2 \rangle_{\rm{sdf}}
602: %  \mbox{ } \mbox{ }$ \\
603: \hline\\[-7truept]
604:                                                                                 
605: $0.08$ & $0.00$ & $0.01$ & $0.67$ &
606:          $0.00$ & $0.03$ & $0.69$ \\
607: %         $0.00$ & $0.01$ & $0.66$  \\
608:                                                                                 
609: $0.2$ &  $0.00$ & $0.01$ & $0.67$ &
610:          $0.00$ & $0.03$ & $0.68$ \\
611: %         $0.00$ & $0.01$ & $0.67$  \\
612:                                                                                 
613: $0.4$ &  $0.00$ & $0.01$ & $0.67$ &
614:          $0.01$ & $0.03$ & $0.67$ \\
615: %         $0.00$ & $0.01$ & $0.67$  \\
616:                                                                                 
617: \hline\\[-9truept]
618: \hline\\[-8truept]
619: \end{tabular}
620: \end{center}
621: \caption{The average of the friction corrections, $\gamma_{\rm F}$,
622: calculated by the iterative
623: self-consistent (0) and space-dependent (sdf)
624: approaches for the MHO and sinusoidal potentials.
625: The resulting temperatures are also included for
626: the space-dependent friction.
627: In all cases the temperature is 2/3
628: (in units of a standard temperature, $k_{\rm b} T_0$),
629: the variance, $\sigma^2=0.22$, and the correlation
630: time, $\tau_{\rm c}=1$.
631: }
632: \label{tb:mho}
633: \end{table*}
634: 
635: The variance and correlation time
636: for both potentials is 0.22 and 1, respectively.
637: The resulting
638: temperatures, ($k_{\rm B}T\equiv \langle v^2 \rangle$), are also 
639: listed for the space-dependent approach.  
640: The friction correction in the self-consistent method
641: ensures equipartition by definition, and therefore, is not listed.
642: The magnitude of the SDF
643: for all values of $\tau_{\rm c}$ follow accordingly;
644: however this is the only value with respect to the given variance
645: for which any deviation from equipartition is observed.
646: As can be seen, 
647: both the self-consistent and space-dependent components of the total
648: friction for each potential provide negligible contributions for this
649: variance since 
650: the magnitude of the fluctuations
651: in the barrier height are relatively small.
652: Therefore the total friction is a sum of 
653: a large thermal component, and a space-dependent contribution.
654: The slight differences in the magnitudes of the SDF for the 
655: two potentials can be attributed to the piecewise nature of the MHO potential.
656: The particles spend most of the simulation time in the wells which
657: do not fluctuate.  
658: A contribution to the total friction from the space-dependent term is 
659: included only when the energetically-limited particle 
660: accumulates enough energy to
661: explore the upper portion of the MHO potential.
662: 
663: To further explore the accuracy of
664: the space-dependent approach, the sinusoidal
665: potential has been studied with a ten-fold increase in the variance from
666: 0.22 to 2.2.  
667: %This corresponds to fluctuations in the barrier heights of
668: %approximately 25 \% around the average.
669: The values of the friction correction
670: from these simulations are listed in Table~\ref{tb:sin}.
671: %\begin{longtable}{cc}
672: \begin{table*}[t]
673: \begin{center}
674: \begin{tabular}{ccccccccccccc}
675: %\begin{longtable}{ccccccccccccc}
676: \hline \\[-9truept]
677: \hline \\[-8truept]
678: \multicolumn{1}{c}{} &
679: \multicolumn{3}{c}{
680:     \underline{$
681:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
682:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
683:                  \tau_{\rm c}=10^{-1}
684:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
685:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
686:    $} }&
687: \multicolumn{3}{c}{
688:    \underline{$
689:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
690:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
691:                  \tau_{\rm c}=10^0
692:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
693:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
694:    $} }&
695: \multicolumn{3}{c}{
696:    \underline{$
697:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
698:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
699:                  \tau_{\rm c}=10^1
700:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
701:                  \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
702:    $} }\\
703: $\gamma_{\rm th}$ &
704: $ \mbox{ }   \mbox{ }
705:   \langle \gamma_{\rm{F}} \rangle_{\rm{0}}
706:   \mbox{ } \mbox{ } $ &
707: $ \mbox{ }  \mbox{ }
708:   \langle \gamma_{\rm{F}}\rangle_{\rm{sdf}}
709:   \mbox{ }  \mbox{ }$ &
710: $ \mbox{ }   \mbox{ }
711:   \langle v^2 \rangle_{\rm{sdf}}
712:   \mbox{ }  \mbox{ }$ &
713: $ \mbox{ }   \mbox{ }
714:   \langle \gamma_{\rm{F}}\rangle_{\rm{0}}
715:   \mbox{ } \mbox{ } $ &
716: $  \mbox{ } \mbox{ }
717:   \langle \gamma_{\rm{F}}\rangle_{\rm{sdf}}
718:   \mbox{ } \mbox{ }$ &
719: $ \mbox{ }  \mbox{ }
720:   \langle v^2 \rangle_{\rm{sdf}}
721:   \mbox{ }  \mbox{ }$ &
722: $  \mbox{ } \mbox{ }
723:   \langle \gamma_{\rm{F}}\rangle_{\rm{0}}
724:   \mbox{ }   \mbox{ }$ &
725: $   \mbox{ } \mbox{ }
726: \langle \gamma_{\rm{F}}\rangle_{\rm{sdf}}
727:   \mbox{ } \mbox{ } $ &
728: $ \mbox{ }  \mbox{ }
729:   \langle v^2 \rangle_{\rm{sdf}}
730:   \mbox{ }   \mbox{ }$ \\
731: \hline\\[-7truept]
732: $0.08$ & $0.04$ & $0.28$ & $0.72$ &
733:          $0.05$ & $0.29$ & $0.74$ &
734:          $0.01$ & $0.28$ & $0.68$  \\
735: $0.2$ & $0.04$ & $0.28$ & $0.71$ &
736:         $0.05$ & $0.29$ & $0.72$ &
737:         $0.01$ & $0.28$ & $0.68$  \\
738: $0.4$ & $0.04$ & $0.28$ & $0.70$ &
739:         $0.06$ & $0.29$ & $0.71$ &
740:         $0.01$ & $0.28$ & $0.67$  \\
741: \hline\\[-9truept]
742: \hline\\[-8truept]
743: \end{tabular}
744: %end{longtable}
745: \end{center}
746: \caption{The average of the friction corrections, $\gamma_{\rm F}$,
747: calculated by the iterative self-consistent (0)
748: and space-dependent approaches (sdf) for the sinusoidal potential.
749: The resulting temperatures are also included for
750: the space-dependent friction method.
751: The temperature is 2/3 in all cases and the variance, $\sigma^2=2.2$.
752: }
753: \label{tb:sin}
754: \end{table*}
755: %\end{longtable}
756: 
757: The displayed correlation times, $\tau_{\rm c}$, are those that
758: exhibit the largest resonant activation.  
759: Consequently, if memory effects in the barrier heights are 
760: important in determining the friction constant, 
761: it should be manifested here.  
762: Although not shown for brevity, 
763: outside this region of the correlation time,
764: the magnitude of the deviations from equipartition 
765: %for the space-dependent friction 
766: decrease rapidly, but the size of the space-dependent components
767: remains roughly constant.
768: Similarly, the corresponding corrections arising in the
769: self-consistent method also approach zero.
770: As can be seen from Table~\ref{tb:sin}, the space-dependent approach 
771: results in a correction that is roughly constant for all values of 
772: the correlation time, while the iterative approach does exhibit some 
773: variation with $\tau_{\rm c}$.  
774: This is the expected result since the space-dependent
775: friction assumes the fluctuations in the potential are local
776: and therefore, ignores any correlation in
777: the barrier heights. 
778: The iterative approach, however, is capable of
779: incorporating the memory of the potential into the friction correction, 
780: but only in an average manner.  
781: As a consequence, significant deviations from equipartition may be 
782: observed when simulations are performed with a 
783: space-dependent friction that ignores the correlation effects, as
784: illustrated by this extreme example.
785: 
786: Figs.~\ref{fig:mfpt0.05} and ~\ref{fig:mfpt0.22} display the 
787: MFPTs obtained for the MHO potential 
788: with the results from the space-dependent and self-consistent 
789: approaches in the top and bottom panels, respectively. 
790: 
791: \begin{figure}[ht]
792: \begin{center}
793: \includegraphics*[width=7.5cm]{./0.05mfpt.eps}
794: \end{center}
795: \caption{The mean first-passage times (MFPT)
796: for a particle diffusing across the MHO stochastic potential
797: are displayed for two possible scenarios of the
798: dissipative mechanism.
799: The top panel uses space-dependent friction, and the bottom
800: displays the uniform friction determined by the self-consistent method.
801: The variance for both is 0.05, and
802: the three lines correspond to values of the thermal
803: friction of 0.08 (solid curve with x symbols), 0.2 (dashed curve
804: with triangles), and 0.4 (dot-dashed curve with squares).
805: The symbols on the broken axis represent the
806: numerically calculated MFPTs at the limits of the correlation time.
807: }
808: \label{fig:mfpt0.05}
809: \end{figure}
810: %%%%%%%%%%%%%%%%%%%%%%%
811: %\newpage
812:                                                                                 
813: %%%%%%%%%%%%%%%%%%%%%%%
814: \begin{figure}[ht]
815: \begin{center}
816: \includegraphics*[width=7.5cm]{./0.22mfpt.eps}
817: \end{center}
818: \caption{The mean first-passage times (MFPT)
819: for a particle diffusing across the MHO stochastic potential
820: are displayed for two possible scenarios of the
821: dissipative mechanism.
822: The parameters are the same as in Fig.~\ref{fig:mfpt0.05}, except the
823: variance is 0.22.
824: }
825: \label{fig:mfpt0.22}
826: \end{figure}
827: %%%%%%%%%%%%%%%%%%%%%%%
828: %\newpage
829:  
830: The results in Fig.~\ref{fig:mfpt0.05} have been calculated using a variance
831: of $\sigma^2=0.05$, while those in Fig.~\ref{fig:mfpt0.22} use
832: $\sigma^2=0.22$.
833: The corresponding results for the sinusoidal potential using 
834: a variance of 0.22 can be seen in Fig.~\ref{fig:mfptsin}.
835: %%%%%%%%%%%%%%%%%%%%%%%
836: \begin{figure}[t]
837: \begin{center}
838: \includegraphics*[width=7.5cm]{./sin.eps}
839: \end{center}
840: \caption{The mean first-passage times (MFPT)
841: for a particle diffusing across the sinusoidal stochastic potential
842: are displayed for two possible scenarios of the
843: dissipative mechanism.
844: Other than for the change from the MHO to the sinusoidal potential,
845: the parameters are the same as in Fig.~\ref{fig:mfpt0.22}.
846: }
847: \label{fig:mfptsin}
848: \end{figure}
849: %%%%%%%%%%%%%%%%%%%%%%%
850: The values on the broken axis represent the numerically calculated
851: MFPTs in the limits of correlation time, $\tau_{\rm c}$.
852: In the zero-correlation time limit, 
853: the fluctuations in the potential are so rapid that the 
854: particle effectively experiences the average, stationary potential, 
855: from which the dynamics were calculated.  
856: In the limit of infinite correlations, 
857: fluctuations in the potential are nonexistent, 
858: and therefore the particle experiences a single realization 
859: of the potential with constant barrier heights determined by the 
860: initial value sampled from the distribution. 
861: The MFPTs displayed in Fig.~\ref{fig:mfpt0.22}
862: obtained with a larger variance alters
863: the magnitude of the resonant activation, but 
864: influences the results for the two approaches equally.
865: The results from the simulations with a space-dependent friction 
866: are systematically shifted to lower MFPTs as seen in all three figures.  
867: This trend is most readily explained through by the 
868: trends in Table.~\ref{tb:mho}.  
869: In the low friction regime, an increase in the friction
870: increases the corresponding rate of transport.
871: The average space-dependent contribution is always larger than
872: its respective mean field counterpart, 
873: and is expected to have the largest effect
874: on the results with the smallest thermal friction.
875: The fluctuations present along the entire reaction coordinate
876: of the sinusoidal potential do not appear to have a dramatic effect
877: on the dynamics.  
878: The results in Fig.~\ref{fig:mfptsin} for the sinusoidal potential follow
879: the same trend as those in Figs.~\ref{fig:mfpt0.05} and \ref{fig:mfpt0.22}
880: for the MHO potential indicating that the SDF approach is capable of 
881: adequately describing the fluctuations in the system.
882: Aside from the shift, the general behavior of the MFPT is adequately
883: reproduced by both methods, particularly at larger values
884: of the thermal friction when the space-dependent component becomes
885: less significant.  
886: At this level of description,
887: each of the two approaches for constructing the friction
888: are capable of capturing the essential dynamics of the system.
889: However, 
890: some advantage is gained by using the self-consistent
891: method because it ensures the system is kept at constant temperature 
892: for all values of the correlation time throughout the simulation,
893: while the space-dependent approach may lead to deviations in extreme cases.
894: The most significant difference between the two methods can be
895: seen at intermediate correlation times, in which the resonant
896: activation observed from the iterative approach is slightly 
897: more pronounced. 
898: This can particularly be seen in the MFPTs
899: when the friction case takes on the smallest value of $\gamma_{\rm{th}}=0.08$.
900: Since the resonant activation arises from correlations in the barrier heights, 
901: it is not surprising that simulations incorporating a friction
902: capable of accounting for this phenomenon can have a noticeable
903: impact on the dynamics, 
904: even if it does so only in an average manner.
905: 
906: \section{Conclusions}\label{sec:conclusion}
907: 
908: The space-dependent friction arising from the 
909: presence of a secondary (external) stochastic potential in 
910: the Langevin equation has been explicitly derived for two simple classes of
911: the stochastic potentials.
912: The numerical results are
913: in excellent agreement with analytic expressions 
914: describing the space-dependent friction. 
915: The resulting dynamics have been compared to those obtained using
916: an alternate approach in which a uniform correction is calculated 
917: self-consistently.
918: Although the latter approach does effectively include the 
919: time correlation between the barrier fluctuations at long times,
920: the former does not in any sense.
921: This neglect may result in deviations from equipartition in 
922: some extreme cases.  
923: However, both approaches are capable of capturing
924: the essential dynamics of the system and lead
925: to the now-expected resonant activation phenomenon.
926: Consequently, the central result of this paper is that the Langevin
927: dynamics of a particle under external stochastic potentials
928: can be properly dissipated by a single uniform renormalized friction
929: without loss of qualitative (and often quantitative) accuracy.
930: 
931: The role of the memory time in
932: an external stochastic potential acting on a particle 
933: described by a generalized Langevin equation of motion
934: is still an open question.
935: In this limit, there would presumably be an interplay between the
936: memory time of the thermal friction and that of the stochastic potential.
937: When the latter is small compared to the former, the quasi-equilibrium
938: condition central to this work would no longer be satisfied by the particle,
939: and hence it is expected that a non-uniform (and time-dependent) friction
940: correction would then be needed.
941: 
942: \section{Acknowledgments}%
943: 
944: RH gratefully acknowledges Abraham Nitzan for
945: an insightful question whose answer became this paper.
946: This work has been partially supported by a National Science Foundation Grant,
947: No.~NSF 02-123320.  
948: The Center for Computational Science and Technology
949: is supported through a Shared University Research (SUR)
950: grant from IBM and Georgia Tech.
951: Additionally, RH is the Goizueta Foundation Junior Professor.
952: 
953: \onecolumngrid
954: \section{Appendix}%
955: 
956: The piecewise nature of the MHO potential results in a
957: piecewise form for the associated SDF.
958: Although incoherent, every barrier gives rise to the same averages,
959: and hence the procedure needs to be carried out only over a 
960: small region defined by the closed interval, $[x_m^0,x_m^\ddagger]$.  
961: The limits of integration over this region can be 
962: determined from the expression for the connection points
963: \begin{equation}
964: x_m^-=-\frac{k_0\lambda}{2k_0-2k_m^\ddagger}+m\lambda
965: \;,
966: \end{equation}
967: where $k_m^{\ddagger}=-(k_0+\eta(t))$. 
968: This can equivalently be expressed as
969: \begin{equation}
970: \eta(t)=-\frac{k_0\lambda}{2(x_m^--m\lambda)}-2k_0
971: \;.
972: \end{equation}
973: At the top of the barrier, when $x_m^-=x_m^\ddagger,~
974: \mbox{$\eta(t)=\infty$}.$
975: In the intermediate region for arbitrary $x$,
976: \begin{eqnarray}
977: \eta(t) & = & -\frac{k_0\lambda}{2(x-m\lambda)}-2k_0 \nonumber \\
978:         & \equiv & \eta^*
979: \;.
980: \end{eqnarray}
981: Otherwise, at the minimum when $x_m^-=x_m^0,~\mbox{$\eta(t)=-k_0$} $.
982: 
983: Although it is apparent from the expression for the barrier height
984: that the corresponding distribution is non-Gaussian, the 
985: resulting forces are Gaussian with the probability given by Eq.~\ref{eq:prob}. 
986: The average force for a given $x$ is simply the weighted average of 
987: the forces when $x$ is in the respective regions,
988: $(x_m^0,x_m^-)$ and $(x_m^-,x_m^{\ddagger})$, 
989: which correspond to $\eta$ regions of
990: $(-k_0,\eta^*)$ and $(\eta^*,\infty)$. 
991: The resulting integral for the average value of $F(x;t)$ is now:
992: \begin{equation}
993: \langle F_{\rm U}(x;t) \rangle= \frac{\int_{-k_0}^{\eta^*}\!d\eta\, F(x)P(\eta)
994:                                 +\int_{\eta^*}^{\infty}\!d\eta\,F(x)P(\eta)}
995:                                 {\int_{-k_0}^{\infty}\!d\eta\,P(\eta)}
996: \;.
997: \end{equation}
998: Here, one must be careful 
999: in determining which portion of the force to use in the above
1000: equation.  
1001: For example, when $\eta<\eta^*$, 
1002: the majority of the force is due to the barrier portion of the potential, 
1003: not the well component.  
1004: The average can thus be expressed as
1005: \begin{equation}
1006: \langle F_{\rm U}(x;t) \rangle= -\int_{\eta^*}^{\infty}\!d\eta\, 
1007:                                k_0(x-x_m^0)P^\prime(\eta)
1008:                                +\int_{-k_0}^{\eta^*}\!d\eta\,
1009:                                (k_0+\eta)(x-x_m^{\ddagger})P^\prime(\eta)
1010: \;,
1011: \end{equation}
1012: where $P^\prime(\eta)$ is defined through the 
1013: normalization condition, $\it{i.e.}$,
1014: \begin{equation}
1015: \int_{-k_0}^{\eta^*}\!d\eta\,P^\prime(\eta)+
1016: \int_{\eta^*}^{\infty}\!d\eta\,P^\prime(\eta)\equiv 1
1017: \;,
1018: \end{equation}
1019: which leads to the probability distribution
1020: \begin{equation}
1021: P^\prime(\eta)=\frac{2}{\sqrt{2\pi\sigma^2}}
1022:                \frac{\exp\left({-\frac{\eta^2}{2\sigma^2}}\right)}
1023:                {1+\erf\left(\frac{k_0}{\sqrt{2\sigma^2}}\right)}
1024: \;,
1025: \end{equation}
1026: where $\erf(x)$ is the standard error function.
1027: Use of the normalization condition reduces the
1028: average force to
1029: \begin{equation}
1030: \langle F_{\rm U}(x;t) \rangle  =  k_0(x-x_m^\ddagger)-
1031:                              [k_0(x-x_m^0)+k_0(x-x_m^{\ddagger})]
1032:                              \int_{\eta^*}^{\infty}\!d\eta\,
1033:                              P^\prime(\eta)  \nonumber 
1034:                              \mbox{} + (x-x_m^{\ddagger})\int_{-k_0}^{\eta^*}
1035:                              \!d\eta\, \eta P^\prime(\eta)
1036: \;.
1037: \end{equation}
1038: The remaining integrals are readily computed; the explicit form of 
1039: the average force is
1040: \begin{eqnarray}
1041: \langle F_{\rm U}(x;t) \rangle & = & k_0(x-x_m^\ddagger)-
1042:                                    \left( k_0 (2x-x_m^0-x_m^\ddagger)\right)
1043:                                    \left(\frac{1-\erf\left(\frac{\eta^*}
1044:                                    {\sqrt{2\sigma^2}}\right)}{1
1045:                                     +\erf\left(\frac{k_0}{\sqrt{2\sigma^2}}
1046:                                     \right)}\right) \nonumber \\
1047:                                &   & \mbox{}+(x-x_m^\ddagger)
1048:                                      \sqrt{\frac{2\sigma^2}{\pi}}
1049:                                      \left[\frac{\exp\left(-
1050:                                      \frac{k_0^2}{2\sigma^2}\right)-
1051:                                      \exp\left(-
1052:                                      \frac{(\eta^*)^2}{2\sigma^2}\right)}
1053:                                      {1+\erf\left(
1054:                                     \frac{k_0}{\sqrt{2\sigma^2}}\right)}\right]\;.               
1055: \end{eqnarray}
1056: 
1057: The second quantity to be computed is the average of the 
1058: square of the force, and the derivation follows that (above) of
1059: the average force.
1060: The limits of integration are the same 
1061: and the Gaussian integrals can be calculated in the same manner.
1062: Again using the normalization requirement, 
1063: the first integral is eliminated such that
1064: \begin{eqnarray}
1065: \langle F_{\rm U}(x;t)^2 \rangle & = & k_0^2(x-x_m^\ddagger)^2 +
1066:                                        [k_0^2(x-x_m^0)^2-
1067:                                k_0^2(x-x_m^\ddagger)^2]
1068:                                \int_{\eta^*}^{\infty}\!d\eta\, 
1069:                                P^\prime(\eta)
1070:                                \nonumber   \\
1071:                          &   & \mbox{}+2k_0(x-x_m^\ddagger)^2
1072:                                \int_{-k_0}^{\eta^*}
1073:                                \!d\eta\,\eta P^\prime(\eta)+(x-x_m^\ddagger)^2
1074:                                \int_{-k_0}^{\eta^*}\!d\eta\,\eta^2 
1075:                                P^\prime(\eta)   
1076: \;.
1077: \end{eqnarray}
1078: The first two integrals are the same as before, 
1079: and the third can be obtained with little effort.  
1080: The resulting mean squared force is
1081: \begin{eqnarray}
1082: \langle F_{\rm U}(x;t)^2 \rangle & = & k_0^2(x-x_m^\ddagger)^2+\left(k_0^2
1083:                                (x-x_m^0)^2-k_0^2(x-x_m^\ddagger)^2\right)
1084:                                \left(\frac{1-\erf\left(
1085:                                \frac{\eta^*}{\sqrt{2\sigma^2}}
1086:                                \right)}{1+\erf\left(\frac{k_0}
1087:                                {\sqrt{2\sigma^2}}\right)}\right)
1088:                                \nonumber   \\
1089:                          &   & \mbox{} +\frac{4k_0\sigma^2}{\sqrt{2
1090:                                \pi\sigma^2}}(x-x_m^{\ddagger})^2
1091:                                \left[\frac{ \exp\left(-
1092:                                \frac{k_0^2}{2\sigma^2}\right)-
1093:                                \exp\left(-\frac{(\eta^*)^2}{2\sigma^2}\right)}
1094:                                {\left(1+\erf\left(\frac{k_0}
1095:                                {\sqrt{2\sigma^2}}\right)\right)}\right]
1096:                                \nonumber   \\
1097:                          &   & \mbox{} +\sigma^2(x-x_m^{\ddagger})^2
1098:                                \left[ \frac{
1099:                                \erf\left(\frac{\eta^*}{\sqrt{2
1100:                                \sigma^2}}\right)
1101:                                +\erf\left( \frac{k_0}{\sqrt{2
1102:                                \sigma^2}} \right)}
1103:                                {1+\erf\left(\frac{k_0}{\sqrt{2
1104:                                \sigma^2}}\right)}\right]\nonumber\\
1105:                          &   & \mbox{} -\sqrt{\frac{2
1106:                                \sigma^2}{\pi}}(x-x_m^\ddagger)^2
1107:                                \left[ \frac{\eta^*\exp\left(-
1108:                                \frac{(\eta^*)^2}{2\sigma^2}\right) 
1109:                                +k_0\exp\left(-\frac{k_0^2}{2\sigma^2}\right)}
1110:                                {1+\erf\left(\frac{k_0}{2
1111:                                \sigma^2}\right)}\right]
1112: \;.
1113: \end{eqnarray}
1114: The SDF for the MHO potential is then obtained by 
1115: appropriate substitutions into Eq.~\ref{eq:FDT}.
1116: 
1117: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1118: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5%
1119:                %  FIGURES  %
1120: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1121: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1122: 
1123: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1124: %%%%%%%% begin mho friction table %%%%%%%%%
1125: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1126: %\newpage
1127: %\begin{table}[h]
1128: %\begin{center}
1129: %\begin{tabular}{ccccccccccccc} 
1130: %\hline \\[-13truept] 
1131: %\hline \\[-12truept]
1132: %\multicolumn{1}{c}{} &
1133: %\multicolumn{3}{c}{ 
1134: %    \underline{$
1135: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1136: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1137: %                 \rm{MHO} 
1138: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1139: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1140: %               $} 
1141: %}&
1142: %\multicolumn{3}{c}{
1143: %   \underline{$ 
1144: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1145: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1146: %                 \rm{Sin}
1147: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1148: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1149: %   $} 
1150: %}\\
1151: %\multicolumn{3}{c}{
1152: %   \underline{$
1153: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1154: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1155: %                 \tau_{\rm c}=10^1
1156: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1157: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1158: %   $} 
1159: %}\\
1160: 
1161: %$\gamma_{\rm th}$ &
1162: 
1163: %$ \mbox{ } \mbox{ }  \langle \gamma_{\rm{F}} \rangle_{\rm{0}}  
1164: %  \mbox{ } \mbox{ }$ &
1165: %$ \mbox{ } \mbox{ }  \langle \gamma_{\rm{F}}\rangle_{\rm{sdf}}  
1166: %  \mbox{ } \mbox{ }$ &
1167: %$ \mbox{ } \mbox{ }  \langle v^2 \rangle_{\rm{sdf}}             
1168: %  \mbox{ } \mbox{ }$ &
1169: 
1170: %$ \mbox{ } \mbox{ }  \langle \gamma_{\rm{F}}\rangle_{\rm{0}}   
1171: %  \mbox{ } \mbox{ }$ &
1172: %$ \mbox{ } \mbox{ }  \langle \gamma_{\rm{F}}\rangle_{\rm{sdf}}  
1173: %  \mbox{ } \mbox{ }$ &
1174: %$ \mbox{ } \mbox{ }  \langle v^2 \rangle_{\rm{sdf}}             
1175: %  \mbox{ } \mbox{ }$ \\
1176: 
1177: %$ \mbox{ } \mbox{ }  \langle \gamma_{\rm{F}}\rangle_{\rm{0}}   
1178: %  \mbox{ } \mbox{ }$ &
1179: %$ \mbox{ } \mbox{ }  \langle \gamma_{\rm{F}}\rangle_{\rm{sdf}}  
1180: %  \mbox{ } \mbox{ }$ &
1181: %$ \mbox{ } \mbox{ }  \langle v^2 \rangle_{\rm{sdf}}               
1182: %  \mbox{ } \mbox{ }$ \\ 
1183: 
1184: %\hline\\[-12truept]
1185: 
1186: %$0.08$ & $0.00$ & $0.01$ & $0.67$ & 
1187: %         $0.00$ & $0.03$ & $0.69$ \\ 
1188: %         $0.00$ & $0.01$ & $0.66$  \\
1189: 
1190: %$0.2$ &  $0.00$ & $0.01$ & $0.67$ &
1191: %         $0.00$ & $0.03$ & $0.68$ \\ 
1192: %         $0.00$ & $0.01$ & $0.67$  \\
1193: 
1194: %$0.4$ &  $0.00$ & $0.01$ & $0.67$ &
1195: %         $0.01$ & $0.03$ & $0.67$ \\ 
1196: %         $0.00$ & $0.01$ & $0.67$  \\
1197: 
1198: %\hline\\[-13truept]
1199: %\hline\\[-12truept]
1200: %\end{tabular}
1201: %\end{center}
1202: %\caption{The average of the friction corrections, $\gamma_{\rm F}$,
1203: %calculated by the iterative 
1204: %self-consistent (0) and space-dependent (sdf)
1205: %approaches for the MHO and sinusoidal potentials.  
1206: %The resulting temperatures are also included for
1207: %the space-dependent friction. 
1208: %In all cases the temperature is 2/3 
1209: %(in units of a standard temperature, $k_{\rm b} T_0$),
1210: %the variance, $\sigma^2=0.22$, and the correlation
1211: %time, $\tau_{\rm c}=1$.
1212: %}
1213: %\label{tb:mho}
1214: %\end{table}
1215: %%%%%%%% end mho friction table %%%%%%%%%%%%%
1216: %\newpage
1217: 
1218: %%%%%%%% begin sin friction table %%%%%%%%%%%%
1219: %\begin{table}[h]
1220: %\begin{center}
1221: %\begin{tabular}{ccccccccccccc}
1222: %\hline \\[-13truept]
1223: %\hline \\[-12truept]
1224: %\multicolumn{1}{c}{} &
1225: %\multicolumn{3}{c}{
1226: %    \underline{$
1227: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1228: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1229: %                 \tau_{\rm c}=10^{-1}
1230: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1231: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1232: %   $} }&
1233: %\multicolumn{3}{c}{
1234: %   \underline{$
1235: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1236: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1237: %                 \tau_{\rm c}=10^0
1238: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1239: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1240: %   $} }&
1241: %\multicolumn{3}{c}{
1242: %   \underline{$
1243: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1244: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1245: %                 \tau_{\rm c}=10^1
1246: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1247: %                 \mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }\mbox{ }
1248: %   $} }\\
1249: %$\gamma_{\rm th}$ &
1250: %$ \mbox{ }   \mbox{ }
1251: %  \langle \gamma_{\rm{F}} \rangle_{\rm{0}}
1252: %  \mbox{ } \mbox{ } $ &
1253: %$ \mbox{ }  \mbox{ }
1254: %  \langle \gamma_{\rm{F}}\rangle_{\rm{sdf}}
1255: %  \mbox{ }  \mbox{ }$ &
1256: %$ \mbox{ }   \mbox{ }
1257: %  \langle v^2 \rangle_{\rm{sdf}}
1258: %  \mbox{ }  \mbox{ }$ &
1259: %$ \mbox{ }   \mbox{ }
1260: %  \langle \gamma_{\rm{F}}\rangle_{\rm{0}}
1261: %  \mbox{ } \mbox{ } $ &
1262: %$  \mbox{ } \mbox{ }
1263: %  \langle \gamma_{\rm{F}}\rangle_{\rm{sdf}}
1264: %  \mbox{ } \mbox{ }$ &
1265: %$ \mbox{ }  \mbox{ }
1266: %  \langle v^2 \rangle_{\rm{sdf}}
1267: %  \mbox{ }  \mbox{ }$ &
1268: %$  \mbox{ } \mbox{ }
1269: %  \langle \gamma_{\rm{F}}\rangle_{\rm{0}}
1270: %  \mbox{ }   \mbox{ }$ &
1271: %$   \mbox{ } \mbox{ }
1272: %\langle \gamma_{\rm{F}}\rangle_{\rm{sdf}}
1273: %  \mbox{ } \mbox{ } $ &
1274: %$ \mbox{ }  \mbox{ }
1275: %  \langle v^2 \rangle_{\rm{sdf}}
1276: %  \mbox{ }   \mbox{ }$ \\
1277: %\hline\\[-12truept]
1278: %$0.08$ & $0.04$ & $0.28$ & $0.72$ & 
1279: %         $0.05$ & $0.29$ & $0.74$ &
1280: %         $0.01$ & $0.28$ & $0.68$  \\
1281: %$0.2$ & $0.04$ & $0.28$ & $0.71$ & 
1282: %        $0.05$ & $0.29$ & $0.72$ &
1283: %        $0.01$ & $0.28$ & $0.68$  \\
1284: %$0.4$ & $0.04$ & $0.28$ & $0.70$ & 
1285: %        $0.06$ & $0.29$ & $0.71$ &
1286: %        $0.01$ & $0.28$ & $0.67$  \\
1287: %\hline\\[-13truept]
1288: %\hline\\[-12truept]
1289: %\end{tabular}
1290: %\end{center}
1291: %\caption{The average of the friction corrections, $\gamma_{\rm F}$,
1292: %calculated by the iterative self-consistent (0) 
1293: %and space-dependent approaches (sdf) for the sinusoidal potential.
1294: %The resulting temperatures are also included for
1295: %the space-dependent friction method.
1296: %The temperature is 2/3 in all cases and the variance, $\sigma^2=2.2$.
1297: %}
1298: %\label{tb:sin}
1299: %\end{table}
1300: %%%%%%%% end sin friction table %%%%%%%
1301: %\newpage
1302: 
1303: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1304: %\begin{figure}[ht]
1305: %\begin{center}
1306: %\includegraphics*[width=15.0cm]{./figs/mhofluc.eps}
1307: %\end{center}
1308: %\caption{
1309: %Representative fluctuations over one period of the MHO 
1310: %potential and force (top panel),
1311: %and the resulting space-dependent friction (bottom panel).
1312: %The numerical component in the bottom panel 
1313: %is displayed as the solid black line, with
1314: %the analytic result, given in the Appendix, as the dotted white line. 
1315: %The temperature is 2/3, the variance
1316: %is 0.22, and the thermal friction is 0.08.
1317: %}
1318: %\label{fig:sdfmho}
1319: %\end{figure}
1320: %%%%%%%%%%%%%%%%%%%%%%
1321: %\newpage
1322: 
1323: %%%%%%%%%%%%%%%%%%%%%%%
1324: %\begin{figure}[ht]
1325: %\begin{center}
1326: %\includegraphics*[width=15.0cm]{./figs/sinfluc.eps}
1327: %\end{center}
1328: %\caption{
1329: %Representative fluctuations over one period of the sinusoidal
1330: %potential and force (top panel),
1331: %and the resulting space-dependent friction (bottom panel).
1332: %The numerical result is displayed as the solid black line,
1333: %with the analytic result, given by Eq.~\ref{eq:sdfsin}, shown as 
1334: %the dotted white line.
1335: %The parameters used are the same as in Fig.~\ref{fig:sdfmho}.
1336: %}
1337: %\label{fig:sdfsin}
1338: %\end{figure}
1339: %%%%%%%%%%%%%%%%%%%%%%%
1340: %\newpage
1341: 
1342: %%%%%%%%%%%%%%%%%%%%%%%
1343: %\begin{figure}[ht]
1344: %\begin{center}
1345: %\includegraphics*[width=10.0cm]{./figs/0.05mfpt.eps}
1346: %\end{center}
1347: %\caption{The mean first-passage times (MFPT)
1348: %for a particle diffusing across the MHO stochastic potential
1349: %are displayed for two possible scenarios of the
1350: %dissipative mechanism.
1351: %The top panel uses space-dependent friction, and the bottom 
1352: %displays the uniform friction determined by the self-consistent method.
1353: %The variance for both is 0.05, and 
1354: %the three lines correspond to values of the thermal
1355: %friction of 0.08 (solid curve with x symbols), 0.2 (dashed curve
1356: %with triangles), and 0.4 (dot-dashed curve with squares).  
1357: %The symbols on the broken axis represent the 
1358: %numerically calculated MFPTs at the limits of the correlation time.
1359: %}
1360: %\label{fig:mfpt0.05}
1361: %\end{figure}
1362: %%%%%%%%%%%%%%%%%%%%%%
1363: %\newpage
1364: 
1365: %%%%%%%%%%%%%%%%%%%%%%
1366: %\begin{figure}[ht]
1367: %\begin{center}
1368: %\includegraphics*[width=10.0cm]{./figs/0.22mfpt.eps}
1369: %\end{center}
1370: %\caption{The mean first-passage times (MFPT)
1371: %for a particle diffusing across the MHO stochastic potential
1372: %are displayed for two possible scenarios of the
1373: %dissipative mechanism.
1374: %The parameters are the same as in Fig.~\ref{fig:mfpt0.05}, except the
1375: %variance is 0.22.
1376: %}
1377: %\label{fig:mfpt0.22}
1378: %\end{figure}
1379: %%%%%%%%%%%%%%%%%%%%%%
1380: %\newpage
1381: 
1382: %%%%%%%%%%%%%%%%%%%%%%%
1383: %\begin{figure}[ht]
1384: %\begin{center}
1385: %\includegraphics*[width=10.0cm]{./figs/sin.eps}
1386: %\end{center}
1387: %\caption{The mean first-passage times (MFPT)
1388: %for a particle diffusing across the sinusoidal stochastic potential
1389: %are displayed for two possible scenarios of the
1390: %dissipative mechanism.
1391: %Other than for the change from the MHO to the sinusoidal potential,
1392: %the parameters are the same as in Fig.~\ref{fig:mfpt0.22}.
1393: %}
1394: %\label{fig:mfptsin}
1395: %\end{figure}
1396: %%%%%%%%%%%%%%%%%%%%%%%
1397: \twocolumngrid
1398: \bibliography{j,surf,tst,miller2,hern,voth,flucbar,mfpt,liquid,sdf}
1399: 
1400: \end{document}
1401: