1: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
2: \usepackage{graphicx}% Include figure files
3: \usepackage{dcolumn}% Align table columns on decimal point
4: \usepackage{bm}% bold math
5:
6: %\documentclass[prl,twocolumn,showpacs]{revtex}
7: \usepackage{epsfig}
8: %\pagestyle{empty}
9: %
10: %-- Define Page size
11: %
12: %\setlength{\topmargin}{0cm}
13: %\setlength{\headsep}{0in}
14: %\setlength{\textheight}{23cm}
15: %\setlength{\oddsidemargin}{0.4cm}
16: %\setlength{\marginparwidth}{.25in}
17: %\setlength{\textwidth}{16.0cm}
18: %\setlength{\parindent}{36pt}
19: %\baselineskip 12 pt
20: %
21: \begin{document}
22: \title{Theory of Single File Diffusion in a Force Field
23: }
24: \author{E. Barkai$^{1,2}$}
25: \author{R. Silbey$^1$}
26: \affiliation{
27: $^1$Department of Chemistry, Massachusetts Institute of Technology,
28: Cambridge, Massachusetts 02139, USA \\
29: $^2$ Department of Physics,
30: Bar Ilan University,
31: Ramat-Gan 52900 Israel}
32: %\email{barkaie@mail.biu.ac.il}
33: \begin{abstract}
34: {
35: The dynamics of hard-core interacting
36: Brownian particles in an external
37: potential field is studied in one dimension.
38: Using the Jepsen line we find a very general and
39: simple formula relating the motion
40: of the tagged center particle,
41: with the classical, time dependent single particle
42: reflection ${\cal R}$ and transmission ${\cal T}$
43: coefficients.
44: Our
45: formula describes rich physical behaviors both in equilibrium
46: and the approach to equilibrium of this many body problem.
47: }
48: \end{abstract}
49:
50: % O2.50.Ey Stochastic Processes
51: % 05.40.Fb Random walks and Levy flights
52: % 82.39.-k chemical kinetics in biological systems
53: % 05.40.Jc Brownian Motion
54: % 82.70.-y Dispersive systems, complex fluids
55: % 05.20.-y Classical Statistical Mechanics
56: \pacs{05.20.-y,05.40.Jc,02.50.Ey}
57: \maketitle
58:
59: Single file diffusion of a tagged Brownian particle,
60: interacting
61: with other Brownian particles, is a model for motion of
62: a molecule or particle in a crowded one-dimensional environment, such as
63: a biological pore or channel
64: \cite{Hanggi,Eok} and for experimentally studied physical systems
65: such as zeolites \cite{Hahn} and confined colloid particles \cite{Lutz,Haim}.
66: The confinement of the tagged particle by the other particles is strong
67: and severely restricts the motion of the particle. The description of
68: this single file motion has been of much theoretical interest
69: \cite{Hanggi,Harris,Levitt,Kollmann,Henk,Roden,Taloni,Tobias,Aslangul,Flomenbom}.
70: For an infinite system and uniform initial particle density,
71: Harris \cite{Harris} and Levitt \cite{Levitt} first showed that a
72: tagged particle exhibits anomalous diffusion
73: $\langle (x_T)^2 \rangle \sim t^{1/2}$,
74: rather then normal diffusion $\langle (x_T)^2 \rangle \sim t$,
75: due to the strong many body confinement effect. In recent years,
76: two important research direction have begun to emerge. First,
77: the effect of an external force field acting on the particles is important
78: since, in many cases, pores induce entropic barriers \cite{Hanggi}
79: and are generally
80: inhomogeneous; hence, single file motion in a periodic potential
81: \cite{Taloni} and a confining box \cite{Tobias} were investigated.
82: Secondly, initial conditions have a profound effect on single file
83: motion \cite{Aslangul,Flomenbom}: for example, particles with initial
84: delta function distribution in space (rather than a uniform
85: distribution as assumed in \cite{Harris,Levitt})
86: yield normal diffusion \cite{Aslangul}.
87: This is important since if a potential field is acting on the particles,
88: thermal initial conditions will have a Boltzmann weighting, leading to generally
89: non-uniform initial conditions.
90:
91:
92: Here we provide a general and surprisingly simple
93: theory of single file diffusion
94: valid in the presence (or absence) of
95: a potential field, $V(x)$, as well as for thermal and
96: non-thermal
97: initial conditions. In addition, we briefly discuss
98: why our main results are valid even for
99: the case where the underlying dynamics is anomalous
100: \cite{remark1}.
101: Our general result reproduces those previously obtained
102: as well as many new ones,
103: by mapping the many particle problem onto a solvable single particle model.
104:
105: In our model,
106: $2 N +1$ identical
107: particles with hard core particle-particle interactions
108: are undergoing Brownian
109: motion in one dimension, so particles cannot pass one another.
110: An external potential field $V(x)$ is acting on the particles.
111: We tag the central particle, which has $N$ other particles to its left,
112: and $N$ to its right.
113: Initially the tagged particle is at $x=0$.
114: The motion of a single particle, in the absence of
115: interactions with other particles,
116: is over-damped Brownian motion
117: so that the single non interacting particle Green function
118: $g(x,x_0,t)$, with the initial condition $g(x,x_0,0) = \delta(x-x_0)$,
119: is obtained from the Fokker-Planck equation \cite{Risken}
120: %
121: \begin{equation}
122: {\partial g(x,x_0,t) \over \partial t} =
123: D \left[ {\partial^2 \over \partial x^2} - { 1 \over k_b T} {\partial \over \partial x} F\left(x \right) \right] g(x,x_0,t),
124: \label{eq000}
125: \end{equation}
126: %
127: and $F(x) = - V'(x)$ is the force field.
128:
129:
130:
131: In Fig. \ref{fig1} a schematic diagram of the problem is presented.
132: The straight line is called the Jepsen line, which starts from the
133: origin $x=0$
134: and follows the rule $x(t)=v t$, where $v$ is a test velocity
135: \cite{Levitt}.
136: In the interacting system we label particles according to their
137: initial position increasing to the right (see Fig. 1).
138: As noticed in \cite{Jepsen,Levitt},
139: since the Brownian particles are impermeable, every time a particle
140: crosses the Jepsen line from the right (or left),
141: the particle number immediately
142: to the left of the line will be raised (or lowered) by one, respectively.
143: Hence, the particle number immediately to the left of the Jepsen line
144: defines a stochastic process
145: decreasing and increasing its value $+1$ or $-1$
146: or zero randomly.
147:
148: Following Levitt \cite{Levitt} we now consider a non interacting system,
149: equivalent to
150: the interacting one in the large $N$ limit.
151: We let particles pass through each other,
152: but switch labels upon collision,
153: and introduce the
154: counter $\alpha(t)$ which increases by $+1$ if a particle
155: crosses the Jepsen line from left, and decreases by $-1$ when
156: a particle crosses this line from right.
157: The event when the counter $\alpha$
158: has the value zero, is equivalent to finding the tagged particle
159: to the left of the Jepsen line in the interacting system.
160: This is the case since then
161: the total number of crossings from left to right is equal
162: the number of crossings from right to left.
163: So the probability
164: of the tagged particle
165: being in the vicinity of $x_T=v t$ is given by the probability that
166: $\alpha=0$, i.e. by the statistics of the number of transitions
167: of the Jepsen line \cite{Levitt}. In what follows we depart from
168: the approach in \cite{Levitt,remark3}.
169:
170: \begin{figure}[!htb]
171: \begin{center}
172: \includegraphics[width=5.cm,height=8cm,angle=-90]{fig1.ps}%
173: \caption{\label{fig1} Schematic motion of Brownian particles in a
174: harmonic potential (the parabola) where particles cannot
175: penetrate through each other.
176: The straight line is called the Jepsen line, as explained in the text.
177: The center tagged particle is labeled $0$.
178: In an equivalent non-interacting picture,
179: we allow particles to pass through each other,
180: and we follow the trajectory of the particle which is at the center.
181: %At a given time
182: %$t$ this particle may be either particle $0$ or $-1$ or $1$.
183: }
184: \end{center}
185: \end{figure}
186:
187:
188: Our aim is to calculate the probability of the random variable
189: $\alpha$, $P_N(\alpha)$, and then switch $v t \to x_T$ to find the
190: probability density function (PDF) of the tagged particle.
191: For that we designate
192: $P_{LL}(x_0 ^{-j})$ as the probability that a non tagged particle $j$
193: starting to the left of the Jepsen line $x_0 ^{-j}<0$,
194: is found also at time $t$ on the left
195: of this line.
196: $P_{LR}$ is the probability of a particle starting to
197: the left of the Jepsen line to
198: end on the right, and similarly for $P_{RR}$ and $P_{RL}$.
199: Consider first
200: $N=1$, that is, one particle which starts at $x_0 ^{1}>0$ and
201: a second particle which starts at $x_0 ^{-1}<0$.
202: Then clearly we have
203: either $\alpha=\pm 1$ or $\alpha=0$. The probabilities of these
204: events are easily calculated, for example
205: %
206: $P_{N=1} \left( \alpha=1 \right) =
207: P_{LR} (x_0 ^{-1}) P_{RR} (x_0 ^{+1})$
208: is the probability that one particle crossed from $L$ to $R$
209: and the other remained in domain $R$.
210: Similarly
211: $P_{N=1} \left( \alpha=0 \right)= P_{LL} (x_0 ^{-1}) P_{RR} (x_0 ^{+1})
212: + P_{LR} (x_0 ^{-1}) P_{RL} (x_0 ^{+1}) $ and
213: $P_{N=1} (\alpha=-1)= P_{LL} (x_0 ^{-1}) P_{RL} (x_0 ^{+1})$.
214: Since in the non-interacting picture, the motion of the particles
215: are independent, we can use random walk theory and
216: Fourier analysis \cite{Montroll} to find
217: the behavior for any $N$
218: %
219: \begin{equation}
220: P_N \left(\alpha\right) = { 1 \over 2 \pi} \int_{-\pi} ^\pi {\rm d} \phi \Pi_{j=1} ^N
221: \lambda\left( \phi, x_0 ^{-j} , x_0 ^{j} \right) e^{ i \alpha \phi}.
222: \label{eq02}
223: \end{equation}
224: %
225: Where the structure function is
226: %
227: $$ \lambda\left(\phi,x_0 ^{-j}, x_0 ^{j} \right) =
228: e^{i \phi} P_{LR} (x_0 ^{-j}) P_{RR} (x_0 ^{j}) +
229: P_{LL}(x_0 ^{-j}) P_{RR} (x_0 ^{j}) $$
230: \begin{equation}
231: +
232: P_{LR}(x_0 ^{-j}) P_{RL} (x_0 ^{j}) + e^{-i \phi}
233: P_{LL}(x_0 ^{-j}) P_{RL} (x_0 ^{j}).
234: \label{eq03}
235: \end{equation}
236: %
237: We average Eq. (\ref{eq02}) with respect to the initial conditions
238: $x_0^{j}$ and $x_0 ^{-j}$, which are assumed to be independent identically
239: distributed
240: random variables and we find
241: %
242: \begin{equation}
243: \langle P_N \left( \alpha \right) \rangle=
244: { 1 \over 2 \pi} \int_{-\pi} ^\pi {\rm d} \phi \langle \lambda\left(\phi\right) \rangle^N e^{ i \alpha \phi}
245: \label{eq04}
246: \end{equation}
247: %
248: where
249: from Eq. (\ref{eq03})
250: the averaged structure function is
251: %
252: $$ \langle \lambda\left(\phi\right) \rangle = $$
253: \begin{equation}
254: e^{i \phi} \langle P_{LR} \rangle \langle P_{RR} \rangle +
255: \langle P_{LL}\rangle \langle P_{RR} \rangle +
256: \langle P_{LR} \rangle \langle P_{RL} \rangle + e^{-i \phi}
257: \langle P_{LL} \rangle \langle P_{RL} \rangle.
258: \label{eq05}
259: \end{equation}
260: %
261: The averages in Eq. (\ref{eq05})
262: are easy to find in principle, in terms of the Green function of
263: the non-interacting particle and the initial density of particles,
264: for example
265: %
266: \begin{equation}
267: \langle P_{LR} \rangle =
268: \int_{-\overline{L}} ^0 f_{L}(x_0) \int_{x_T} ^{\overline{L}}
269: g(x,x_0,t) {\rm d} x {\rm d} x_0
270: \label{eq09a}
271: \end{equation}
272: %
273: where the mentioned replacement $v t \to x_T$ was made.
274: Here $f_L(x_0)$ is the PDF
275: of initial positions of the particles which initially are at
276: $x_0<0$, similarly $f_R(x_0)$ describes the initial conditions of the
277: right particles.
278: In Eq. (\ref{eq09a})
279: $\overline{L}$
280: is the system size,
281: which can be taken to
282: infinity in the usual way.
283:
284:
285:
286: Eq. (\ref{eq04}) describes a random walk where the number of particles
287: $N$ serves as an operational time. We may use the Gaussian central limit
288: theorem (CLT) to analyze this walk, when $N \to \infty$. In that limit
289: the first two moments of the structure function, $\mu_1$ and $\mu_2$,
290: found
291: in the small $\phi$ expansion
292: %
293: \begin{equation}
294: \langle \lambda(\phi) \rangle = 1 + i \mu_1 \phi - {1 \over 2 } \mu_2 \phi^2 + O(\phi^3)
295: \label{eq06}
296: \end{equation}
297: %
298: are the only two parameters needed to determine the behavior of $P_N (\alpha)$.
299: Defining the variance
300: $\sigma^2 = \mu_2 - (\mu_1)^2$, using Eq. (\ref{eq05})
301: and the normalization condition e.g. $\langle P_{LR} \rangle + \langle P_{LL}
302: \rangle = 1$ we find the expected result
303: %
304: \begin{equation}
305: \mu_1 = \langle P_{LR} \rangle - \langle P_{RL} \rangle
306: \label{eq07}
307: \end{equation}
308: %
309: \begin{equation}
310: \sigma^2 =
311: \langle P_{RR} \rangle \langle P_{RL} \rangle
312: + \langle P_{LL} \rangle \langle P_{LR} \rangle.
313: \label{eq08}
314: \end{equation}
315: %
316: Using the CLT we have the probability of zero crossing, namely
317: $\alpha =0$ in the $N \to \infty$ limit
318: %
319: \begin{equation}
320: P_N\left( \alpha=0 \right) \sim { \exp\left( - {N (\mu_1)^2 \over 2 \sigma^2} \right) \over \sqrt{ 2 \pi N } \sigma}.
321: \label{eq09}
322: \end{equation}
323: %
324: This is our first general result, valid for a large
325: class of
326: Green functions
327: and initial conditions and thus suited for
328: the investigation of a wide range of problems.
329:
330:
331:
332: Symmetric potential fields
333: $V(x)= V(-x)$, and symmetric initial conditions are now investigated.
334: The latter simply means that the
335: density of the initial positions of the left
336: particles, i.e. those residing initially in $x_0<0$,
337: is the same as that of the right particles,
338: $f_R(x_0)= f_L (-x_0)$.
339: In this case the subscript $R$ and $L$ is redundant
340: and we use $f(x_0)=f_R(x_0)=f_L(-x_0)$ to describe the initial conditions \cite{remark4}.
341: From symmetry it is clear that
342: the tagged particle is unbiased, namely $\langle x_T \rangle=0$.
343: Further, since $N$ is large we may expand expressions in the $\exp$ in
344: Eq. (\ref{eq09}) in
345: $x_T$, to obtain leading terms
346: %
347: \begin{equation}
348: \mu_1 = \Delta J x_T + O(x_T)^2
349: \label{eq10}
350: \end{equation}
351: %
352: where
353: we used the symmetry of the problem which
354: implies
355: $(\langle P_{LR}\rangle - \langle P_{RL}\rangle)_{x_T =0} = 0$,
356: and by definition
357: %
358: \begin{equation}
359: \Delta J = {\partial \over \partial x_T}
360: \left[ \langle P_{LR}(x_T) \rangle - \langle P_{RL}\left( x_T\right) \rangle \right]|_{x_T=0}.
361: \label{eqx}
362: \end{equation}
363: %
364: Similarly
365: %
366: \begin{equation}
367: \sigma^2 =
368: 2\langle P_{RR} (x_T) \rangle
369: \left[ 1 - \langle P_{RR} \left(x_T\right) \rangle \right]|_{x_T = 0}+O(x_T).
370: \label{eq12}
371: \end{equation}
372: %
373: We designate $\langle P_{RR}(x_T) \rangle|_{x_T=0} ={\cal R}$ as
374: a reflection coefficient,
375: since it is the probability that a particle starting at $x_0<0$
376: is found at $x<0$ at time $t$ when an average over all initial
377: conditions is made
378: %
379: \begin{equation}
380: {\cal R}= \int_0 ^{\overline{L}} f(x_0) \int_0 ^{\overline{L}} g(x,x_0,t) {\rm d} x {\rm d} x_0.
381: \label{eq13}
382: \end{equation}
383: %
384: As usual the transmission coefficient ${\cal T} =1- {\cal R}$
385: is defined through ${\cal T} = \langle P_{RL}(x_T) \rangle|_{x_T=0}$.
386: Notice that these reflection and transmission coefficients are
387: time dependent single particle
388: quantities which give useful information for the
389: many body problem.
390: Also from symmetry we have
391: %
392: $ {\partial \over \partial x_T} \langle P_{RL}(x_T) \rangle |_{x_T=0}
393: =-{\partial \over \partial x_T} \langle P_{LR}( x_T) \rangle |_{x_T=0}$
394: %
395: in Eq. (\ref{eqx}).
396: Hence we define
397: $j=-\partial \langle P_{LR} (x_T) \rangle / \partial x_T | _{x_T =0}$ where
398: from its definition Eq.
399: (\ref{eq09a})
400: %
401: \begin{equation}
402: j= \int_0 ^{\overline{L}} f(x_0) g(0,x_0,t) {\rm d} x_0.
403: \label{eq14}
404: \end{equation}
405: %
406: So $j$ is the
407: density of non-interacting particles at $x=0$ for an initial density $f(x_0)$.
408: Using Eqs.
409: (\ref{eq09}-\ref{eq12})
410: we find our main result the PDF of the tagged particle
411: %
412: \begin{equation}
413: P(x_T) \sim {1 \over \sqrt{ 2 \pi \langle \left( x_T \right)^2 \rangle}}
414: \exp\left[ - { \left( x_T \right)^2 \over 2 \langle \left( x_T \right)^2 \rangle} \right]
415: \label{eq15}
416: \end{equation}
417: %
418: where
419: $\langle \left( x_T \right)^2 \rangle =
420: {\cal R} {\cal T} / ( 2 N j^2)$
421: is the mean square displacement (MSD).
422:
423:
424: {\em Gaussian packet}. As our first example, we consider particles
425: without external forces $V(x)=0$,
426: in an infinite system with symmetric
427: Gaussian
428: initial conditions with a width $\xi$:
429: $f(x_0) = \sqrt{2} \exp( - x_0 ^2 / (2 \xi^2) ))/ \sqrt{\pi \xi^2}$
430: $(x_0>0)$.
431: The free particle Green function is
432: %
433: \begin{equation}
434: g(x,x_0,t)={\exp\left( - {(x - x_0)^2 \over 4 D t } \right) \over \sqrt{ 4 \pi D t}}.
435: \label{eq15dd}
436: \end{equation}
437: %
438: Using Eqs. (\ref{eq13}-\ref{eq15dd}) we find
439: the MSD of the tagged particle \cite{remark}
440: %
441: \begin{equation}
442: \langle (x_T)^2\rangle \sim \xi^2 {\pi \over N} \left( 1 + { 2 D t \over \xi^2} \right) \left[ { 1 \over 4} - { 1 \over \pi^2} \mbox{arccot}^2 \left( \sqrt{ {2 D t \over \xi^2} } \right) \right].
443: \label{eq15cc}
444: \end{equation}
445: %
446: This solution exhibits a transition from anomalous sub-diffusion
447: to normal diffusion. At short times $2 D t/\xi^2 \ll 1$
448: %
449: %\begin{equation}
450: $\langle (x_T)^2 \rangle \sim \xi {\sqrt{ 2 D t} \over N}$
451: %\end{equation}
452: %
453: while at long times
454: %\begin{equation}
455: $\langle (x_T)^2 \rangle \sim {\pi D t \over 2 N}$.
456: %\end{equation}
457: %
458: For short times the particles do not have time to disperse, hence
459: the motion of the tagged particle is slower than normal, increasing
460: as $t^{1/2}$ since it is confined by the other particles in
461: the system.
462:
463:
464: {\em Particles in a box}. We now consider the example of
465: particles in a box extending
466: from $- \overline{L}$ to $\overline{L}$,
467: which was recently investigated using the Bethe--ansatz and
468: numerical simulations \cite{Tobias}.
469: The tagged particle initially on $x=0$ has $N$ particles
470: uniformly distributed to its left and similarly to its right. In this case
471: $f(x_0) = 1/ \overline{L}$ for $0<x_0<\overline{L}$.
472: The green function $g(x,x_0,t)$
473: of a single particle in a box with
474: reflecting walls is found using an eigenvalue approach
475: \cite{Risken}, similar to the method of solution
476: of the
477: undergraduate quantum mechanical problem of a particle in a box.
478: We find
479: %
480: \begin{equation}
481: {\cal R} = {1\over 2} + { 4 \over \pi^2} \sum_{n=1, \mbox{Odd}} ^\infty
482: { \exp \left[ - D {\left( \pi^2 n^2\right) \over 4 \overline{L}^2} t \right] \over
483: n^2}
484: \label{eq16}
485: \end{equation}
486: %
487: so at $t=0$, ${\cal R} = 1$ since all particles initially say in
488: $(0,\overline{L})$ did
489: not have time to move to the other side of box, and
490: $\lim_{t \to \infty} {\cal R} =1/2$
491: since in the long time limit there is equal probability for
492: a non-interacting particle to occupy half of the box.
493: The eigenvalues of the non-interacting particle
494: control the {\em exponential}
495: decay of ${\cal R}$ Eq. (\ref{eq16}) which in turn
496: determines the dynamics of the interacting tagged particle.
497: We also find $j=1/2 \overline{L}$ using Eq. (\ref{eq14}).
498: For short times $D \pi^2 t / 4 \overline{L}^2 \ll 1$
499: we can replace the
500: summation in Eq. (\ref{eq16}) with integration and then using Eq. (\ref{eq15})
501: and $\rho = N / \overline{L}$
502: %
503: \begin{equation}
504: P\left(x_T \right) \sim {\sqrt{\rho} \over 2 \left( \pi D t \right)^{1/4}}
505: \exp\left[ - {\rho \left( x_T \right)^2 \sqrt{\pi} \over 4 \sqrt{ D t } } \right].
506: \label{eq17}
507: \end{equation}
508: %
509: Thus the tagged particle undergoes a single file
510: sub-diffusive process \cite{Harris,Levitt}
511: $\langle (x_T)^2 \rangle \sim 2 (D t)^{1/2}/ (\rho \sqrt{\pi})$,
512: since for short times the particles do not interact with walls.
513: In the long time limit, the tagged particle reaches an
514: equilibrium easily found using Eqs. (\ref{eq15},\ref{eq16})
515: %
516: \begin{equation}
517: p\left( x_T \right) \sim { \sqrt{ N } \over \sqrt{ \pi} \overline{L} } e^{ - { N (x_T)^2 \over \overline{L}^2} }
518: \label{eq18}
519: \end{equation}
520: %
521: and $\lim_{t \to \infty} \langle (x_t)^2 \rangle = \overline{L}^2/ (2 N)$.
522: More generally from Eq.
523: (\ref{eq15}),
524: $\langle(x_T)^2 \rangle=2{\cal R}\left( 1 - {\cal R} \right)\overline{L}^2/N$
525: which nicely matches and simplifies considerably,
526: the Bethe-ansatz solution \cite{Tobias}
527: already for $N=70$ \cite{remark5}.
528:
529: {\em Thermal initial conditions}. If we assume that initially the particles
530: are in thermal equilibrium, our simple formulas simplify even more.
531: If $f(x_0) = 2 \exp( - V(x_0)/ k_b T)/ Z$ where
532: the normalizing partition function is
533: %
534: \begin{equation}
535: Z = \int_{-\infty} ^\infty \exp \left[ - {V\left(x\right) \over k_b T} \right]
536: {\rm d} x
537: \label{eqZ}
538: \end{equation}
539: %
540: then $j=1/Z$.
541: To see this, note that since we have a symmetric case $V(x) = V(-x)$ then
542: $g(0,-x_0,t) = g( 0, x_0,t)$ and
543: %
544: $j =
545: 2 \int_0 ^\infty \exp\left[ - V(x_0) / k_b T \right] g(0,x_0,t) {\rm d} x_0/Z$
546: gives
547: %
548: \begin{equation}
549: j= \int_{-\infty} ^\infty {\exp\left[ -{ V\left( x_0 \right) \over k_b T } \right] \over Z} g(0,x_0, t) {\rm d} x_0.
550: \label{eqjj}
551: \end{equation}
552: %
553: Where we assume that the potential is binding so a stationary solution of
554: the Fokker-Planck equation is reached i.e. free particle Eq.
555: (\ref{eq15dd}) is excluded.
556: Therefore $j$ Eq. (\ref{eqjj})
557: is the probability of finding non-interacting particles on the origin,
558: with thermal equilibrium initial conditions.
559: Since the latter is the
560: stationary solution of the Fokker-Planck operator,
561: $j$ is time independent and
562: equal to $j= \exp[ - V(0)/ k_b T]/Z$.
563: We can always choose $V(0)= 0$ and then $j = 1/Z$.
564: Using Eq.
565: (\ref{eq15})
566: we find the MSD of the tagged particle
567: %
568: \begin{equation}
569: \langle (x_T)^2 \rangle ={ {\cal R} {\cal T} Z^2\over 2 N}.
570: \label{eqsimple}
571: \end{equation}
572: %
573:
574: {\em Single file motion in Harmonic Potential.}
575: We now consider particles
576: in an Harmonic potential field
577: $V(x) = m \omega^2 x^2 / 2$
578: where $\omega$ is the Harmonic frequency and
579: initially the particles
580: are in thermal equilibrium.
581: The corresponding single particle green function $g(x,x_0,t)$
582: describes the Ornstein--Uhlenbeck process, and is well
583: known \cite{Risken}.
584: Defining the thermal length scale
585: $\xi_{{\rm th}} = \sqrt{ k_b T / m \omega^2}$
586: and using thermal initial conditions
587: $f(x_0)=\sqrt{2 / \pi \xi_{{\rm th}} ^2}\exp \left[-(x_0)^2/ 2\xi_{{\rm th}} ^2 \right]$,
588: we find using
589: Eqs. (\ref{eq13},\ref{eqZ},\ref{eqsimple})
590: %
591: \begin{equation}
592: \langle \left( x_T \right)^2 \rangle = { \pi \over N} \left( \xi_{{\rm th}}\right)^2 \left\{ { 1 \over 4} - { 1 \over \pi^2} \mbox{arccot}^2 \left[ \sqrt{ \exp\left( 2 \tilde{t} \right) - 1} \right] \right\}
593: \label{eqHO}
594: \end{equation}
595: %
596: where $\tilde{t} = D t / (\xi_{{\rm th}})^2$ is dimensionless time.
597: For short times $\tilde{t} \ll 1$ we obtain sub-diffusive behavior
598: $\langle (x_T)^2 \rangle \sim \xi_{{\rm th}} \sqrt{ 2 D t} / N$ since
599: then the effects of the binding field are negligible, while the tagged particle
600: motion is restricted by all others leading to sub-diffusive behavior.
601: For long times the tagged particle reaches an equilibrium
602: $ \langle (x_{T})^2 \rangle \sim \pi (\xi_{{\rm th}} )^2 / 4 N$.
603:
604: {\em Equilibrium of tagged particle.} In the long time limit,
605: and for binding potential fields we find again a simple limiting behavior.
606: First, note that $\lim_{t \to \infty} {\cal R} = 1/2$
607: as is easily obtained from Eq. (\ref{eq13}) and physically
608: obvious for the symmetric system under investigation.
609: Secondly, for any initial condition $g(0,x_0,t)= 1/Z$ when $t \to \infty$,
610: namely the Green function reached an equilibrium
611: and hence we find from Eq. (\ref{eq14})
612: $\lim_{t \to \infty} j = 1 / Z$. Therefore in the long time
613: limit
614: %
615: \begin{equation}
616: \lim_{t \to \infty} \langle \left( x_T \right)^2 \rangle \sim
617: {Z^2 \over 8 N}.
618: \label{eqZN}
619: \end{equation}
620: %
621: Consistently, this result can be derived directly from the canonical ensemble,
622: using the many body Hamiltonian of the system,
623: in the large $N$ limit, without resorting to dynamics as we did here.
624: The long time limit of our examples
625: of particles in a box
626: Eq. (\ref{eq18})
627: and in a harmonic field
628: Eq. (\ref{eqHO})
629: are
630: easily recovered from the more general Eq. (\ref{eqZN}).
631:
632:
633: In the examples analyzed so far,
634: we considered symmetric external fields, and initial
635: conditions, which yield no drift $\langle x_T \rangle=0$.
636: Our general results Eqs. (\ref{eq09},\ref{eq15}) can be extended to
637: cases where the drift is non-zero,
638: then expansions Eqs. (\ref{eq10},\ref{eqx},\ref{eq12})
639: include additional terms which make
640: the final expression slightly more cumbersome and
641: hence suited for a longer publication.
642: Another interesting extension of this work is the
643: case where the particle dynamics,
644: in the {\em absence} of interactions with each other,
645: is anomalous as was considered in \cite{Flomenbom,Band} for the
646: case of sub-diffusing particles \cite{remark1}.
647: These results will be addressed in
648: the future.
649:
650: To conclude, we have mapped the many body problem of interacting
651: hard core Brownian particles to a single particle problem
652: where calculation of the reflection coefficient ${\cal R}$ and
653: $j$ yield the motion of the tagged particle. In the large $N$ limit,
654: information
655: on the motion of the tagged particle is contained in the single particle
656: Green function
657: $g(x,x_0,t)$ which can be calculated with known methods.
658: Rich physical behaviors emerge, which depend on the
659: initial distribution of the particles and the force field.
660: For particles in a confining field, a transition from sub-diffusion
661: behavior to an equilibrium value is found to be controlled by
662: the eigen modes of the single particle green function, thus
663: the approach to equilibrium is fast in this sense, even though
664: the diffusion is slower than normal. In contrast, for Gaussian
665: initial conditions a transition between sub-diffusive to normal
666: diffusion is found, which is the expected transition for a wide variety
667: of well behaved initial conditions in the absence of a force.
668: The most rewarding part of this work
669: are the simple equations we find for the single file problem.
670:
671:
672: {\bf Acknowledgments} EB is supported by the Israel Science foundation.
673: We thank
674: Tobias Ambj\"ornsson
675: for
676: helpful discussions.
677:
678: \begin{thebibliography}{99}
679:
680: %Diffusion in confined geometries
681: \bibitem{Hanggi} P.S. Burada, P. H\"anggi, F. Marchesoni
682: G. Schmid, and P. Talkner
683: arXiv:0808.2345v1 [cond-mat.stat-mech] (2008).
684:
685: %Wall-mediated self diffusion in slit and cylindrical pores
686: \bibitem{Eok} H. Kim, C. Kim, E. K. Lee, P. Talkner, and P. H\"anggi
687: {\em Phys. Rev. E} {\bf 77}, 031202 (2008).
688:
689: % Ionic diffusion through confined geometries: from Langevin equations
690: % to partial differential equations.
691: %\bibitem{Nadler} B. Nadler, Z. Schuss, A. Singer, and R. S. Eisenberg
692: %{\em J. Phys.: Condens. Matter} {\bf 16} S2153-S2165 (2004).
693:
694: % Single-File Diffusion Observation
695: \bibitem{Hahn} K. Hahn, J. K\"argeri, and V. Kukla {\em Phys. Rev. Lett.}
696: {\bf 76} 2762 (1996).
697:
698: % Single File Diffusion of Colloids in One Dimensional Channels
699: % Experiment
700: \bibitem{Lutz} C. Lutz, M. Kollmann, and C. Bechinger
701: {\em Phys. Rev. Lett.} {\bf 93} 026001 (2004).
702:
703: % From Random Walk to Single File Diffusion
704: % Experiments colloid suspensions
705: \bibitem{Haim} B. Lin, M. Meron, B. Cui, S. A. Rice, H. Diamant
706: {\em Phys. Rev. Lett.} {\bf 94} 216001 (2005).
707:
708: % Diffusion with Collisions Between Particles
709: \bibitem{Harris} T. E. Harris {\em J. Appl. Prob.} {\bf 2}, 323 (1965).
710:
711: % Dynamics of a Single File Pore: Non Fickian Behavior
712: \bibitem{Levitt} D. G. Levitt {\em Phys. Rev. A} {\bf 8} 3050 (1973).
713:
714:
715: % Single File Diffusion of Atomic and Colloidal Systems: Asymptotic Laws
716: \bibitem{Kollmann} M. Kollmann {\em Phys. Rev. Lett.} {\bf 90} 180602 (2003).
717:
718: % Diffusion in concentrated lattice gases III Tracer diffusion on a one dimensional lattice.
719: \bibitem{Henk} H. van Beijeren, K. W. Kehr, and R. Kutner {\em Phys. Rev. B} {\bf 28} 5711 (1983).
720:
721: \bibitem{Roden} C. R\"odenbeck, J. K\"arger, and K. Hahn {\em Phys. Rev. E}
722: {\bf 57} 4382 (1998).
723:
724:
725: % Single File Diffusion on a Periodic Surface
726: \bibitem{Taloni} A. Taloni, and F. Marchesoni {\em Phys. Rev. Lett.} {\bf 96}, 020601 (2006).
727:
728: % Single-File Diffusion in a Box
729: \bibitem{Tobias} L. Lizana, T. Ambj\"ornsson
730: {\em Phys. Rev. Lett.} {\bf 100}, 200601 (2008).
731:
732: % Classical diffusion of N interacting particles in one dimension:
733: %General results and asymptotic laws
734: \bibitem{Aslangul} Cl. Aslangul {\em Europhysics Letters} {\bf 44} 284 (1998).
735:
736: % On Single File and Less Dense Systems
737: \bibitem{Flomenbom} O. Flomenbom, and A. Taloni
738: {\em Europhysics Letters} {\bf 83} 20004 (2008).
739:
740: \bibitem{remark1} Physically
741: this might be important for dynamics in slit pores,
742: where it was shown that non-interacting particles, exhibit super-diffusion, due to forward scattering from
743: the walls of pore \cite{Eok}.
744:
745: \bibitem{Risken} H. Risken {\em The Fokker-Planck Equation} Springer-Verlag
746: (Berlin) (1984).
747:
748: % Dynamics of a Simple Many-Body System of Hard Rods
749: \bibitem{Jepsen} D. W. Jepsen {\em J. Math. Phys.} (N.Y) {\bf 6} 405 (1965).
750:
751: \bibitem{remark3} Levitt \cite{Levitt}
752: uses a Poissonian approximation to calculate
753: the motion of a tagged particle in the absence of a potential field.
754: We find that for our
755: aim, where an external force is acting on the particles,
756: the Poissonian approximation is wrong. This can be shown by an exact
757: calculation of the equilibrium distribution of the tagged particle,
758: using Boltzmann-Gibbs canonical ensemble.
759:
760: % On the enriched collection of stochastic processes
761: \bibitem{Montroll} E. W. Montroll and B. J. West in {\em Fluctuation Phenomena}
762: E. W. Montroll and J. L. Lebowitz Editors, Studies in Statistical Mechanics
763: Vol. VII North Holland, Amsterdam (1979).
764:
765: \bibitem{remark4} $f(x_0)=0$ if $x_0<0$, $f(x_0)\ge 0$ and
766: $\int_0 ^\infty f(x_0) {\rm d} x_0 = 1$.
767:
768:
769:
770: \bibitem{remark} To obtain Eq. (\ref{eq15cc}) we need the following integral
771: $ \int_0 ^\infty \exp( - \eta^2 y^2 / 2) \mbox{Erf} \left( y / \sqrt{2} \right) {\rm d} y =
772: \sqrt{2/ \pi} \mbox{arccot}( \eta ) / \eta$.
773:
774: \bibitem{remark5} We compared our analytical result with
775: simulations of \cite{Tobias} in Fig. 4.
776: %There finite size particles
777: %where considered, while we consider point particles. Simple rescaling
778: %according to $\rho \to N / (\overline{L} - N \Delta)$ where $\Delta$ is the
779: %diameter of the spheres makes a direct comparison possible.
780: %Deviations
781: %between results in \cite{Tobias} and our work are found at very short
782: %times when the particles did not have time to collide.
783:
784:
785: % Single File Diffusion of sub-diffusive particles
786: \bibitem{Band}
787: T. Bandyopadhyay {\em Europhysics Letters} {\bf 81} 16003 (2008).
788:
789: \end{thebibliography}
790: \end{document}
791:
792: