1: %\documentclass[aps,prl,twocolumn]{revtex4}
2: %\documentstyle[aps,prl,multicol,epsf]{revtex4}
3: %\documentstyle[aps,prl,multicol,epsf,graphics]{revtex}
4: %\documentstyle[aps,prl,twocolumn,epsf,graphics]{revtex}
5: %\documentstyle[pra,aps,eqsecnum,epsf]{revtex}
6: %\usepackage{graphicx}
7:
8: \documentclass[aps,prb,twocolumn,superscriptaddress,floatfix]{revtex4}
9: \usepackage{psfig}
10: \usepackage{graphicx}
11: \usepackage{epsfig}
12: \bibstyle{apsrev.bib}
13:
14:
15: \begin{document}
16: \input epsf.sty
17:
18: \title{The partially asymmetric zero range process with quenched disorder}
19:
20:
21: \author{R\'obert Juh\'asz}
22: \email{juhasz@lusi.uni-sb.de}
23: \affiliation{Fachrichtung Theoretische Physik, Universit\"at des
24: Saarlandes, D-66041 Saarbr\"ucken, Germany}
25: \author{Ludger Santen}
26: \email{santen@lusi.uni-sb.de}
27: \affiliation{Fachrichtung Theoretische Physik, Universit\"at des
28: Saarlandes, D-66041 Saarbr\"ucken, Germany}
29: \author{Ferenc Igl\'oi} \email{igloi@szfki.hu}
30: \affiliation{ Research Institute for Solid
31: State Physics and Optics, H-1525 Budapest, P.O.Box 49, Hungary}
32: \affiliation{ Institute of Theoretical Physics, Szeged University,
33: H-6720 Szeged, Hungary}
34:
35: \date{\today}
36:
37: \begin{abstract}
38: We consider the one-dimensional partially asymmetric zero range
39: process where the hopping rates as well as the easy direction of
40: hopping are random variables. For this type of disorder there is a
41: condensation phenomena in the thermodynamic limit: the particles
42: typically occupy one single site and the fraction of particles
43: outside the condensate is vanishing. We use extreme value statistics
44: and an asymptotically exact strong disorder renormalization group
45: method to explore the properties of the steady state. In a finite
46: system of $L$ sites the current vanishes as $J \sim L^{-z}$, where
47: the dynamical exponent, $z$, is exactly calculated. For $0<z<1$ the
48: transport is realized by $N_a \sim L^{1-z}$ active particles, which
49: move with a constant velocity, whereas for $z>1$ the transport is
50: due to the anomalous diffusion of a single Brownian particle.
51: Inactive particles are localized at a second special site and their
52: number in rare realizations is macroscopic. The average density
53: profile of inactive particles has a width of, $\xi \sim
54: \delta^{-2}$, in terms of the asymmetry parameter, $\delta$. In
55: addition to this, we have investigated the approach to the steady
56: state of the system through a coarsening process and found that the
57: size of the condensate grows as $n_L \sim t^{1/(1+z)}$ for large
58: times. For the unbiased model $z$ is formally infinite and the
59: coarsening is logarithmically slow.
60: \end{abstract}
61:
62: \maketitle
63:
64: \newcommand{\bc}{\begin{center}}
65: \newcommand{\ec}{\end{center}}
66: \newcommand{\be}{\begin{equation}}
67: \newcommand{\ee}{\end{equation}}
68: \newcommand{\beqn}{\begin{eqnarray}}
69: \newcommand{\eeqn}{\end{eqnarray}}
70:
71:
72: \vskip 2cm
73: \section{Introduction}
74:
75: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
76: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
77:
78: The properties of interacting many particle systems can be strongly
79: affected by the presence of quenched disorder. This in particular
80: true for systems of self-driven particles\cite{krug,im,barma}, where even pointlike
81: defects\cite{janowsky} are able to change the macroscopic properties of the
82: system. Often these defects cause phase separated states, a
83: phenomenon, which is known from jam formation at bottlenecks\cite{krug,barma}.
84:
85: Particular interesting features arise, if not only the amplitude
86: of the hopping rates are quenched random variables, but the
87: directional bias as well. Then the dynamics of the particles is
88: governed by a complex landscape of energy barriers. As the escape
89: time growth exponentially with the heights of the barriers, the
90: largest barriers in the system determine the velocity of the
91: particles. This property of the many particle system is in agreement
92: with more classical problem of single particle diffusion in a
93: disordered environment\cite{sinai}, which is rather well understood and
94: serves for numerous applications, e.g. polymer translocation through a
95: narrow pore\cite{kln04} or the motion of molecular motors
96: on heterogeneous tracks\cite{kln05}.
97:
98: Since there is no general framework of studying nonequilibrium
99: disordered systems it is of interest to investigate specific simple
100: models. Here we consider the zero range process\cite{evansreview} (ZRP) with quenched
101: disorder. The ZRP is particularly well suited for theoretical analysis
102: because the stationary weights of a given configuration factorize and
103: can be exactly calculated\cite{spitzer}. But the ZRP is not only conceptually
104: interesting, there are a number of important applications as well: it
105: has been used in order to describe e.g. the formation of traffic jams\cite{chowd},
106: the coalescence in granular systems, and the gelation in networks.
107:
108: The effect of quenched disorder on the properties of the ZRP have been
109: investigated in one dimension in the totally asymmetric version\cite{jainbarma}, i.e. when the
110: particles hop in one direction with position dependent rates. In this
111: case a dynamical phase transition takes place from a low density
112: phase, where one observes the condensation of holes, to a homogeneous
113: high density state\cite{kf,evans}. At low densities the average speed of the
114: particles is related to the lower cut-off of the effective hopping
115: rates. The properties of the phase transition are determined by the
116: asymptotics of the effective hopping rate distribution at the lower
117: cut-off and not by its mean or variance. If the ZRP is partially
118: asymmetric and the easy direction of hopping is random as well, one
119: observes a strong dependence of the average speed of the particles on
120: the system size rather than on the density as in case of the totally
121: asymmetric model\cite{jsi}. Already this result illustrates the
122: qualitative difference
123: between the two different realization of the disorder.
124:
125: The ZRP can be exactly mapped
126: onto a one-dimensional asymmetric simple
127: exclusion process\cite{liggett,hinrichsen,schutzreview} (ASEP) if sites are considered as particles and
128: masses as hole clusters. Disorder in the ZRP is transformed into
129: particle dependent hopping rates in the ASEP, which is generally
130: referred to as particle-wise disorder. In the ASEP one can also realize disorder
131: which depends on sites. The ASEP both with particle-wise\cite{kf,evans} and
132: site-wise disorder\cite{barma,derrida,stinchcombe,jsi} have been extensively studied, in particular for systems
133: which are totally asymmetric.
134:
135: The partially asymmetric ASEP with particle-wise disorder has been recently
136: studied by a strong disorder
137: renormalization group (RG) method\cite{jsi} which provides asymptotically
138: exact results for large sizes. This RG
139: method has originally been introduced in order to
140: study random quantum spin chains\cite{MDH}, but afterwards it has been applied
141: for a large variety of quantum\cite{DF,fisherxx} as well as classical\cite{RGsinai} systems, both in
142: and out of equilibrium\cite{hiv}, for a review see\cite{im}.
143:
144: In this paper we use this RG method to study the
145: partially asymmetric ZRP with quenched disorder. With this type of
146: disorder there is a condensation phenomenon: almost all particles
147: occupy one single site, whereas the fraction of particles outside the
148: condensate converges to zero in the thermodynamic limit. Here we address
149: questions regarding the size dependence of different quantities, such
150: as the stationary current, the density profile, the mean density in
151: the bulk and the number of particles outside the condensate. We also
152: investigate the approach to the steady state through a coarsening
153: process and the time dependence of the mass of the condensate.
154:
155: The paper is organized as follows. In the next section we will
156: introduce the disordered ZRP and discuss its relation to the ASEP with
157: particle disorder. In Sec.\ref{sec_RG} we introduce the
158: RG method which is used to calculate the steady
159: state current in the system. The density profile as well as
160: finite-size behavior of the bulk density is calculated in
161: Sec.\ref{sec_profile}, whereas the coarsening behavior is analyzed
162: in Sec.\ref{sec_coarsening}. We summarize and discuss our results in
163: the final section. Some details of the calculations are given in the
164: Appendices.
165:
166:
167: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
168: \section{Definitions and preliminaries}
169: \label{model}
170: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
171:
172: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
173: \begin{figure}
174: \includegraphics[width=0.8\linewidth]{fig1.eps}
175: \caption{\label{fig_zrp} The zero range process on a one-dimensional
176: lattice. The top most particle at site $i$ hops with rate $p_i$ to
177: site $i+1$ and with rate $q_{i-1}$ to site $i-1$.}
178: \end{figure}
179: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
180:
181: In the ZRP particles hop from site to site on a lattice and the hop
182: rates depend on the departure site and on the number of particles
183: (mass) at that site. With these conditions the stationary weight of a
184: configuration is given in a factorized form, which offers an
185: opportunity to analyze the steady state properties exactly\cite{evansreview}.
186: The type of ZRP we consider in the present work
187: is defined on a one-dimensional periodic lattice with
188: $l=1,2, \dots,L$ sites and $N$ particles.
189: Particles are allowed to hop to nearest neighbor sites.
190: The hopping rates are quenched random variables,
191: where we denote
192: the forward hopping rate from site $l$ to $l+1$ by $p_l$ and the
193: backward hopping rate from site $l+1$ to $l$ by $q_l$ (see Fig.~\ref{fig_zrp}).
194: Compared to the most general ZRP, we do not consider mass-dependent
195: hopping rates, i.e. $p_l$ and $q_l$ are valid for all occupation numbers
196: $n_l\geq 1$.
197: A configuration of the system is characterized by the distribution
198: of masses (number of particles), $n_l$, where $\sum_{l=1}^L n_l=N$.
199:
200: The one-dimensional ZRP defined above is equivalent to an ASEP where
201: $l=1,2, \dots, L$ particles are placed on a ring with $L+N$
202: sites. The $l$-th particle hops to empty neighboring sites with a
203: forward (backward) rate of $q_l$ ($p_l$) and a configuration is
204: defined by the number of empty sites, $n_l$, behind the particle (i.e.
205: in front of the particle there are $n_{l+1}$ holes).
206:
207: The stationary weight of a configuration of the ZRP is
208: given in a product form:
209: %
210: \be
211: P(\{n_l\}) = Z^{-1}_{L,N} \prod_{l=1}^L f_l(n_l)
212: \label{product}
213: \ee
214: %
215: where
216: %
217: \be
218: Z_{L,N} = \sum_{\{n_l\}} \prod_{l=1}^L f_l(n_l)\delta \left ( \sum_{l=1}^{L} n_l - N \right )\;
219: \label{ZLN}
220: \ee
221: %
222: is the canonical partition sum. The factors are given by:
223: %
224: \be
225: f_l(n_l)=g_l^{n_l}\;
226: \ee
227: %
228: and the $g_l$ satisfy the equations:
229: %
230: \be
231: g_l(p_l+q_{l-1})=g_{l-1} p_{l-1} + g_{l+1} q_l\;.
232: \label{geq}
233: \ee
234: %
235: These equations are identical to the stationary weights of a random
236: walker with random hop rates: $p_l$ ($l \to l+1$) and $q_{l-1}$ ($l
237: \to l-1$). The solution of Eq.~(\ref{geq}) is given by:
238: %
239: \be
240: g_l=\frac{C}{p_l} \left[ 1 + \sum_{i=1}^{L-1} \prod_{j=1}^i \frac{q_{l+j-1}}{p_{l+j}} \right]\;,
241: \label{g}
242: \ee
243: %
244: where $C$ is a constant. From Eq.~(\ref{geq}) one obtains:
245: %
246: \be
247: g_{l-1} p_{l-1} - g_l q_{l-1}=g_l p_l - g_{l+1} q_l=const\;,
248: \label{const}
249: \ee
250: %
251: which is a constant of motion. Taking the constant in Eq.~(\ref{const})
252: to be one we get:
253: %
254: \be
255: C=\left[ 1 - \prod_{l=1}^L \frac{q_l}{p_l} \right]^{-1}\;.
256: \ee
257: %
258: One can obtain a number of useful results for the stationary
259: state\cite{evansreview}.
260: E.g. the occupation probability, $p_l(n_l)$, that the $l$-th
261: site contains $n_l$ particles is given by:
262: %
263: \be
264: p_l(n_l)=f_l(n_l) \frac{Z_{L-1,N-n_l}} {Z_{L,N}} \;,
265: \label{p_n}
266: \ee
267: %
268: and the particle current reads as:
269: %
270: \be
271: J_{L,N}=\langle p - q \rangle
272: =\frac{Z_{L,N-1}}{Z_{L,N}}\;.
273: \label{J}
274: \ee
275: %
276:
277: These expressions are simplified if the number of particles goes to
278: infinity, $N \to \infty$, whereas there is no restriction on the value
279: of $L$. In this limit we obtain for the canonical partition sum:
280: %
281: \be
282: Z_{L,N}=g_L^N \prod_{l=1}^{L-1} \frac{1}{1-g_l/g_L},\quad N \to \infty\;,
283: \label{Z_inf}
284: \ee
285: %
286: where we label the sites in a way that $g_L=\max(\{g_l\})$. Then the
287: stationary current reads as:
288: %
289: \be
290: J_{L}=1/g_L,\quad N \to \infty\;,
291: \label{JL}
292: \ee
293: %
294: and the occupation probability
295: follows a geometrical distribution:
296: %
297: \be
298: p(n_l)=(1-\alpha_l)\alpha_l^{n_l}, \quad \alpha_l=g_l/g_L, \quad N \to \infty\;,
299: \label{geom}
300: \ee
301: %
302: so that
303: \be
304: \langle n_l \rangle =\alpha_l/(1-\alpha_l), \quad N \to \infty.
305: \label{nav}
306: \ee
307:
308: For large, but finite $N$ and $L$ the sum of the occupation numbers
309: should be equal to the total number of particles\cite{evansreview}:
310: %
311: \be \sum_{l=1}^{L}\langle n_l
312: \rangle=\sum_{l=1}^{L}{1\over 1/(g_lJ_{L,N})-1} =N\;.
313: \label{finite}
314: \ee
315: %
316:
317: In this work the hop rates are independent and identically
318: distributed random variables taken from the distributions, $\rho
319: (p)dp$ and $\pi (q)dq$, respectively which will be specified later.
320: Generally we allow for the existence of links with $p_l>q_l$ as well
321: as links with $p_l<q_l$ with finite probability, i.e. the easy
322: direction of hopping is a random variable, too.
323:
324: We introduce a control-parameter, $\delta$,
325: which characterizes the average asymmetry between forward and backward rates:
326: %
327: \be \delta=\frac{[\ln p]_{\rm av} - [\ln q]_{\rm av}}{{\rm var}[\ln
328: p]+{\rm var}[\ln q]}\;,
329: \label{delta}
330: \ee
331: %
332: such that for $\delta>0$ ($\delta<0$) the particles move on average to
333: the right (left). Here, and in the following $[\dots]_{\rm av}$
334: denotes average over quenched disorder, whereas ${\rm var}(x)$ stands
335: for the variance of $x$.
336:
337: One can show, that for any non-zero density, $\rho=N/L>0$, a Bose
338: condensation occurs, in the sense
339: that a finite fraction of the particles,
340: $\langle n_L \rangle/N >0$, are condensed at site $L$, and $\langle
341: n_L \rangle/N$ tends to one in typical samples for $N \to \infty$.
342: The condensation of particles can be understood by analyzing
343: the distribution of $g_i$'s, as we will show in the next section
344: (see for comparison Refs.[\onlinecite{evans,jsi}]).
345: We shall also show that the stationary
346: current vanishes in the thermodynamic limit, i.e. $\lim_{L\to\infty}J_L=0$.
347: These results are in complete agreement with a recently introduced
348: criterion \cite{kafri}, which predicts strong phase separation for
349: vanishing stationary current.
350:
351: The numerical calculations are carried out using two kinds of disorder
352: distributions. First, a bimodal distribution, where
353: $p_iq_i=r$ holds for all $i$ and \be \rho (p)=c\delta (p-1)+(1-c)\delta
354: (p-r),
355: \label{bimodal}
356: \ee
357: with $r>1$ and $0<c\le 1/2$.
358: Second, we use a uniform distribution defined by:
359: \beqn
360: \rho (p)&=&p_0^{-1}\Theta (p)\Theta (p_0-p), \nonumber \\
361: \pi (q) &=& \Theta (q)\Theta (1-q),
362: \label{uniform}
363: \eeqn
364: where $p_0>0$ and $\Theta (x)$ is the Heaviside function.
365:
366: In the following we use the terminology, that the
367: model is asymmetric or biased, for $\delta>0$, and it is unbiased for
368: $\delta=0$. This latter model is realized if the distribution of the
369: hopping rates is symmetric. For random quantum spin chains, which show analogous
370: low-energy properties, $\delta=0$, corresponds to the critical point and
371: (a part of) the $\delta>0$ region is the so called Griffiths phase\cite{griffiths}.
372: Since the same type of mechanism takes place in the two systems we use the
373: terminology of Griffiths phase for the random ZRP with $\delta>0$, too.
374:
375: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
376: \section{Analysis of the current}
377: \label{sec_current}
378:
379: In section~\ref{model} we pointed out that the stationary
380: weights of the disordered ZRP are
381: related to the statistical weights, $g_l$, of a random walker in a
382: random environment.
383: For the further analysis it is important to notice that the
384: random variables $g$ defined by Eq.(\ref{g}),
385: are so-called Kesten variables\cite{kesten}, because there exist a number of rigorous
386: mathematical results concerning their distribution $P_L(g)$. In the
387: Griffiths phase, i.e. for
388: $\delta>0$ and in the thermodynamic limit, $L \to \infty$, the limit
389: distribution presents an algebraic tail:
390: %
391: \be
392: P_{L}(g) \sim g^{-1-1/z}, \quad L \to \infty\;,
393: \label{kesten}
394: \ee
395: %
396: where $z$ is the positive root of the equation
397: %
398: \be
399: \left[\left({q\over p}\right)^{1/z} \right]_{av}=1.
400: \label{z}
401: \ee
402: %
403: In particular we obtain for small, $\delta$, that:
404: %
405: \be
406: z \approx \frac{1}{2 \delta}, \quad \delta \ll 1\;,
407: \label{z_small}
408: \ee
409: %
410: which is divergent and independent of the actual form of the
411: distributions, $\rho(p)$ and $\pi(q)$.
412:
413: If we now apply the results for Kesten variables given above
414: to the disordered ZRP two remarks are in order.
415: First, the algebraic tail in Eq.~(\ref{kesten}) should be present
416: in the large (but finite) $L$ limit. Second, the $g_l$-s in
417: Eq.~(\ref{g}) are not
418: strictly independent. However, as will be shown
419: later $g_l$ and $g_k$ can be considered to be uncorrelated
420: if $|l-k|>\xi$, where $\xi$ is
421: the finite correlation length of the problem. Therefore
422: $g_L$ has to be treated
423: as the largest event of a distribution in
424: Eq.~(\ref{kesten}) among $\sim L/\xi$ terms. According to extreme value
425: statistics\cite{galambos} the typical value of $g_L$ follows from the relation:
426: $g_L^{-1/z}L=O(1)$, such that:
427: %
428: \be
429: g_L \sim L^{z}\;.
430: \label{g_L}
431: \ee
432: %
433: Another result, which we obtain from extreme value statistics,
434: is the typical value of the $i$-th largest occupation probability, $g(i)/g_L$:
435: %
436: \be
437: \frac{g(i)}{g_L} \approx i^{-z}\;.
438: \label{g_i}
439: \ee
440: %
441: From these results and from Eq.~(\ref{J}) follows that the typical
442: value of the current is
443: %
444: \be
445: |J_L|\sim L^{-z},\quad \delta>0\;,
446: \label{jgriffiths}
447: \ee
448: %
449: thus vanishes algebraically in the thermodynamic limit. Our second
450: result concerns the typical value of particles outside the condensate,
451: $N_{out}$, which follows from Eqs.(\ref{nav}) and (\ref{g_i}) as:
452: %
453: \be
454: N_{out} = \sum_{l=1}^{L-1} \langle n_l \rangle \sim \sum_{i=1}^{L-1} i^{-z}
455: \sim \left\{ \begin{array}{l} L^{1-z},\quad z<1 \\
456: O(1), \quad z>1 \end{array} \right. \;.
457: \label{N_out}
458: \ee
459: %
460: Therefore, for any $z>0$ the fraction of
461: particles in the condensate goes to one in the limit of large
462: system sizes $L\to\infty$ and constant density, $\rho=L/N$.
463: The typical number of particles outside the condensate, however,
464: behaves differently for $z<1$, when
465: it is divergent, and for $z>1$, when it tends to a finite value. We
466: shall discuss this issue in more detail in Sec.\ref{sec_profile}.
467:
468: Next, we turn to analyze the behavior of the $g_l$ weights in the
469: weakly asymmetric limit, $\delta \ll 1$ and estimate the correlation
470: length, $\xi$.
471: For our analysis it is convenient to use a correspondence between the
472: sequences of the hop rates and random walk paths. (A similar reasoning has been introduced for
473: the random transverse-field Ising chain in Ref.\cite{ir}.)
474: We consider the situation where the walker starts
475: at $i=0$, $x_0=0$ and takes in its
476: $l$-th step an oriented distance of $\delta x_l=\ln q_{l-1} - \ln
477: p_l$, such that its position, $x_l$, is given by: $x_l=\sum_{i_1}^l
478: \delta x_i$, see Fig. \ref{rw1}. For $\delta > 0$ the
479: motion of the walker is biased (due to the average slope of the
480: energy landscape), but
481: decorated by fluctuations, i.e. one observes local deviations
482: from the average behavior. The typical time during which the walker makes a large
483: excursion, $\xi$, against the bias follows from the Gaussian nature of the fluctuations
484: and given by\cite{ir}:
485: %
486: \be
487: \xi \sim \delta^{-2}\;.
488: \label{xi}
489: \ee
490: %
491: More precisely the probability of an excursion time, $i$, reads as
492: $P(i) \sim \exp(-i/\xi)$. Similarly, from the Gaussian form of the
493: fluctuations follows the typical transversal size of excursions of
494: the walker, which scales as, $\xi_{\perp} \sim \xi^{1/2} \sim \delta^{-1}$.
495: Consequently the probability of a transversal size of excursions,
496: $\Delta$, is given by $P(\Delta) \sim \exp(-\Delta/\xi_{\perp})$.
497:
498: $\xi$ as defined in Eq.~(\ref{xi}) is the correlation length of the
499: random walk and at the same time it is the correlation length of the
500: ZRP as discussed in the first part of this section. Indeed, the
501: stationary weight, $g_i$, is connected to the part of the landscape
502: which starts at position, $i$, and its value is dominated by the height of the
503: largest excursion, $x_i^{max}-x_i$, see Fig.\ref{rw1}. Since the
504: landscape becomes uncorrelated for distances (times), which are larger
505: than $\xi$, the corresponding $g_l$ weights are
506: uncorrelated, too. The value of $g_L$, i.e. the largest weight, is
507: related to the largest possible transverse fluctuation, $\Delta_L$. In
508: a chain of length, $L$, this extremal position can be chosen out of
509: $\sim L$ positions, therefore the typical value of $\Delta_L$ follows
510: from extreme value statistics as: $\exp(-\Delta_L/\xi_{\perp}) L
511: =O(1)$, thus $\Delta_L \sim \ln L \delta^{-1}$. Keeping in mind the
512: definition of $\delta x_i$ and putting this result into Eq.~(\ref{g})
513: we obtain: $\ln g_L \sim \delta^{-1} \ln L$, which is compatible with
514: the exact result in Eq.~(\ref{g_L}) and the small $\delta$
515: expansion of $z$ in Eq.~(\ref{z_small}).
516:
517: In the random unbiased ZRP $\delta=0$ and the correlation length is
518: divergent. In this case the transverse fluctuations of the walker in
519: the thermodynamical limit are unbounded. In a finite system
520: they typically behave as: $\Delta_L \sim L^{1/2}$.
521: From this follows that
522: the fluctuations in the particle current in a typical sample are of
523: the form:
524: %
525: \be
526: |\ln |J_L|| \sim L^{1/2},\quad \delta=0\;.
527: \label{jcrit}
528: \ee
529: %
530:
531: We can thus conclude that in the random walk picture the largest local
532: fluctuation of the energy landscape is responsible for the small value of the
533: particle current. The particles are accumulated in front of this large
534: barrier and built the condensate at a given site of the system.
535: At other subleading barriers there are only a few, $O(1)$, particles
536: the sum of which gives typically $N_{out}$, as estimated in
537: Eq.~(\ref{N_out}). Later in Sec.\ref{sec_cloud} and \ref{sec_density}
538: we shall use this random walk mapping to obtain the size of the cloud
539: of particles around the condensate and the density profile.
540:
541: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
542: \begin{figure}
543: \includegraphics[width=0.8\linewidth]{fig2.eps}
544: \caption{\label{rw1} Mapping the configurations to random walk paths.}
545: \end{figure}
546: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
547:
548:
549: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
550: \section{Renormalization and the particle flow}
551: \label{sec_RG}
552: The results about the particle current presented in the previous
553: section can be obtained, together with another results, by the
554: application of an RG method. Here we first
555: illustrate how single sites of the lattice can be decimated,
556: afterwards the method is adopted to the random system and finally the
557: properties of the fixed points, both for the unbiased and asymmetric (biased)
558: models are discussed.
559: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
560: \begin{figure}
561: \includegraphics[width=\linewidth]{fig3.eps}
562: \caption{\label{fig_elim} Renormalization scheme for the zero range
563: process: Site $i$ is eliminated and
564: hopping from site $i-1$ to $i+1$ and vice versa occurs with
565: rates $\tilde p$ and $\tilde q$, respectively.}
566: \end{figure}
567: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
568:
569: \subsection{Exact decimation of a single site}
570:
571: The factorized form of the stationary distribution in
572: Eq.~(\ref{product}) allows to explicitly integrate out a
573: typical site, say $i \ne L$. To do this we start with Eq.~(\ref{const})
574: and eliminate $g_i$: \be {p_{i-1}p_{i}\over
575: p_i+q_{i-1}}g_{i-1}-{q_{i-1}q_{i}\over p_i+q_{i-1}}g_{i+1} = 1, \ee
576: so that the effective (two site) hop rates are identified as
577: %
578: \be
579: \tilde p = {p_{i-1}p_{i}\over p_i+q_{i-1}}, \qquad
580: \tilde q = {q_{i-1}q_{i}\over p_i+q_{i-1}}.
581: \label{effective}
582: \ee
583: %
584: At the same time the particle current is transformed as:
585: %
586: \be
587: \tilde{J}=\sum_{n_i=0}^{\infty} \frac{Z_{L-1,N-n_l-1}}{Z_{L-1,N-n_l}} p(n_l)\;,
588: \label{Jtilde}
589: \ee
590: %
591: what can be obtained along the lines of Eq.~(\ref{J}). From a practical
592: point of view it is of importance that the particle current remains
593: invariant, which happens both, i) for $N \to \infty$ (even for finite $L$)
594: and ii) if $\langle n_i \rangle \to 0$. For a random system case ii)
595: is realized for a typical site, if $L \to \infty$, even if
596: $N$ is finite. Indeed,
597: in this case $q_i=O(1)$ whereas
598: $g_L=J_{L,N}^{-1} \to \infty$. Since a site outside the condensate
599: generally contains finite number of particles these are all
600: ``typical'' so that one expects to be able to repeat the
601: transformation up to site, $L$.
602:
603: \subsection{Renormalization of the random ZRP}
604:
605: In the following we shall apply the RG transformation in case ii),
606: i.e. for a random system in the infinite (or very large) lattice
607: limit. The corresponding single particle problem of a random
608: walker in a random potential is thoroughly studied in the literature
609: and many exact results have been obtained by using the strong disorder RG
610: method\cite{RGsinai,im}. The principles of the RG procedure
611: apply as well to the
612: the ZRP with many particles.
613: There are, however, questions which
614: are specific in the large $N$ limit, such as the properties of the
615: condensate, the profile of the particles and the coarsening process
616: towards the stationary state, which will be studied in the following
617: sections.
618:
619: In the following we adopt the RG rules in Eq.~(\ref{effective}) for the
620: random system. In the traditional application of the strong disorder
621: RG method for a Brownian particle one renormalizes the energy
622: landscape\cite{RGsinai} in Fig.~\ref{rw1}. Here we apply a somewhat different
623: approach and also refer to a mapping to random quantum spin chains.
624: The first step is to select the site (bond) to be decimated out.
625: We choose the fastest rate, $\Omega= \max(\{p_i\},\{q_i\})$, which
626: means that processes which are faster than the time-scale,
627: $\tau=1/\Omega$, are integrated out. If the distribution of disorder
628: is sufficiently broad, then the rate $\Omega$ is much larger than
629: the neighboring rates and the expressions for the renormalized rates
630: in Eq.~(\ref{effective}) can be simplified. For example if $\Omega=p_i$
631: we have:
632: %
633: \be
634: \tilde p \approx p_{i-1}, \qquad
635: \tilde q \approx {q_{i-1}q_{i}\over \Omega },
636: \label{back}
637: \ee
638: %
639: and similarly for $\Omega=q_{i-1}$, by replacing $q_i$ and $p_i$. It
640: is evident from Eq.~(\ref{back}) that the generated new rate is smaller
641: than the eliminated rate $\Omega$. Repeating the decimation
642: procedure the energy scale is gradually lowered and we monitor the
643: distribution of the hop rates, $P(p,\Omega)$ and $R(q,\Omega)$.
644: respectively. In particular we are interested in the scaling
645: properties of the transformation at the fixed point, which is located
646: at $\Omega^*=0$. This latter statement is in accordance with the fact,
647: that the stationary current vanishes in the system.
648:
649: Using the approximate decimation rules in Eq.~(\ref{back}) the RG
650: equations can be formulated in the continuum limit as a set of
651: integral-differential equations, that can be exactly solved at the
652: fixed point, both at $\delta =0$ \cite{DF} and for $\delta\neq0$
653: \cite{i02}. These calculations are identical to that of the random
654: transverse-field Ising chain (RTFIC) and we borrow the results
655: obtained for the latter model. Details about the mapping between the
656: two problems as well as the fixed-point solutions are presented in
657: Appendix A.
658:
659: \subsection{The unbiased ZRP}
660:
661: The unbiased ZRP with $\delta =0$ corresponds to the critical point
662: of the RTFIC. At this point distribution of forward and backward
663: rates in Eq.~(\ref{sol}) are identical having the same exponents:
664: $r_0=p_0\sim 1/\ln(\Omega/\Omega_0)$. Here, $\Omega_0$ is a reference
665: energy scale. The appropriate scaling variable is given by $\eta=-(\ln
666: \Omega - \ln p)/\ln \Omega=-(\ln \Omega - \ln q)/\ln \Omega$, having a
667: distribution $\rho(\eta)=\exp(-\eta) {\rm d} \eta$, $\eta > 0$. The
668: length scale, $L_{\Omega}$, which is the size of the effective
669: cluster, and the energy scale are related as:
670: %
671: \be
672: L_{\Omega} \sim \left[ \ln \frac{\Omega_0}{\Omega} \right]^2,\quad \delta=0\;,
673: \label{scale_crit}
674: \ee
675: %
676: which shows unusual, activated scaling. In the ZRP in the course
677: of renormalization
678: huge particle clusters are created and the
679: distance, $X$, they travel is the accumulated distance covered by the
680: original particles that form the cluster, $X=\sum_{k=1}
681: x_k$. Finally after eliminating all but the last site, we
682: obtain the accumulated distance traveled by all the particles during
683: time, $t$. In the unbiased case $[\langle X \rangle ]_{\rm av}=0$
684: holds and the average mean-square of the accumulated displacement is given by:
685: %
686: \be
687: [\langle X^2 \rangle ]_{\rm av} \sim \ln^4 t\;,
688: \label{sinai}
689: \ee
690: %
691: in agreement with the diffusion of a Sinai walker\cite{sinai1}.
692: Furthermore an
693: appropriate scaling combination between current (measured in a
694: specific sample) and size is obtained as: $|\ln
695: |J_L|| L^{-1/2}$, which is compatible with the relation in
696: Eq.~(\ref{jcrit}).
697:
698: \subsection{The asymmetric ZRP}
699: \label{sec_asym}
700:
701: The ZRP with a global bias, $\delta >0$ (or equivalently $\delta <0$),
702: corresponds to the so called disordered Griffiths phase
703: \cite{griffiths} of the RTFIC. In this case a time scale $\tau
704: \sim\Omega_{\xi}$ exists, which separates two characteristic
705: areas of the renormalization process.
706: In the initial part of
707: the renormalization, for $\Omega>\Omega_{\xi}$, both backward and
708: forward rates are decimated out, until effective clusters of typical
709: size, $\xi$, are created, where the correlation length $\xi$ is defined in
710: Eq.~(\ref{xi}), through Eq.~(\ref{scale_crit}).
711: In the random walk picture in Fig. \ref{rw1}
712: this corresponds to eliminate the fluctuations of the landscape and
713: obtain a monotonically decreasing curve. Now continuing the decimations
714: for $\Omega<\Omega_{\xi}$, almost exclusively forward rates are
715: eliminated and the ratio of the typical backward and forward rates
716: tends to zero. Consequently the system is renormalized to a totally
717: asymmetric ZRP in which the distribution of the forward rates can be
718: calculated (see Eq.~(\ref{sol})), and follows a pure power-law (see
719: Eq.~(\ref{sol})):
720: %
721: \be
722: P_0(p,\Omega) \approx \frac{1}{z \Omega}\left(\frac{\Omega}{p}\right)^{1-1/z},
723: \quad \Omega < \Omega_{\xi}\;,
724: \label{Pp}
725: \ee
726: %
727: where $z$ is the dynamical exponent as defined in Eq.~(\ref{z}).
728: For comparison with Eq.~(\ref{kesten}) one should note that for the
729: totally asymmetric model the weights in Eq.~(\ref{g}) are $\sim 1/p_l$,
730: consequently Eqs.~(\ref{Pp}) and (\ref{kesten}) are of identical form.
731: Furthermore the effective particles become indeed uncorrelated in a
732: length-scale of $\xi$.
733:
734: If effective clusters of size $L$ are created the corresponding energy
735: scale is lowered as:
736: %
737: \be
738: L \sim \left(\frac{\Omega}{\Omega_0} \right)^{-1/z}, \quad \delta > 0\;.
739: \label{scale_gr}
740: \ee
741: %
742: Consequently the smallest effective forward hop rate is given by:
743: $\tilde{p}_L \sim L^{-z}$. This implies the same relation for the
744: stationary current as given in Eq.~(\ref{jgriffiths}).
745:
746: The results presented in the previous subsections are expected to be
747: asymptotically exact. Indeed during renormalization the distributions
748: of the hop rates broaden without limits both at the critical point and
749: in the Griffiths phase consequently the decimation rule in
750: (\ref{back}) becomes exact in the fixed point.
751:
752: The scaling forms of the current in Eqs.(\ref{jgriffiths}) and
753: (\ref{jcrit}) are tested by numerical calculations in Ref.\cite{jsi}
754: for the partially asymmetric ASEP with bimodal, i.e. discrete disorder
755: in Eq.~(\ref{bimodal}). Here we performed calculations on the ZRP using
756: the uniform, i.e. continuous disorder in Eq.~(\ref{uniform}). The
757: agreement is indeed satisfactory, both for the unbiased,
758: (Fig.~\ref{du1}) and for the biased (asymmetric) (Fig.~\ref{du2}) cases.
759:
760: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
761: \begin{figure}
762: \includegraphics[width=1.0\linewidth]{fig4.eps}
763: \caption{\label{du1} Distribution of the steady state current in the random
764: unbiased ZRP calculated from Eqs.(\ref{g}) and (\ref{JL}). a) Note
765: that the distribution broadens with the size, which is a clear
766: indication of an infinite disorder fixed point. b) An appropriate
767: scaling collapse is obtained using the scaling combination in
768: Eq.~(\ref{jcrit}).}
769: \end{figure}
770: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
771:
772: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
773: \begin{figure}
774: \includegraphics[width=1.0\linewidth]{fig5.eps}
775: \caption{\label{du2} Distribution of the steady state current in the random
776: asymmetric ZRP, i.e. in the Griffiths phase calculated from
777: Eqs.(\ref{g}) and (\ref{JL}). a) In a log-log plot the distribution
778: is shifted to the left with increasing sizes, the asymptotic slope
779: of the curves is $1/z$ and can be used to measure the dynamical
780: exponent. b) An appropriate scaling collapse is obtained using the
781: scaling combination in Eq.~(\ref{jgriffiths}) and the exact result
782: for $z$ in Eq.~(\ref{z}). }
783: \end{figure}
784: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
785:
786: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
787: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
788: \section{Distribution of particles}
789: \label{sec_profile}
790:
791: In order to characterize the microscopic states of the disordered
792: ZRP it is instructive to distinguish between the
793: $\langle n_L \rangle$ particles in the condensate and
794: particles outside the condensate, which are either considered
795: to be {\it active}, i.e. they behave as a single random walker
796: and carry the current of the system, or {\it inactive} particles
797: that are localized outside the condensate.
798:
799: \subsection{Active particles}
800:
801: In a system with large finite number of sites, $L$, the number of
802: active particles is expected to scale as, $N_{a} \sim L^a$, with an
803: exponent $0 \le a <1$. The value of $a$ can be estimated by using the
804: relation: $J_{L}=v N_a/L$, where $v$ is the average velocity of a
805: Brownian particle in that landscape. In a finite system $v$ scales as:
806: $v \sim L^{-\omega}$, with $\omega=\max(z-1,0)$. Indeed a single
807: Brownian walker moves with a constant speed for $z<1$ and has
808: anomalous diffusion for $z>1$. Thus the particle current in a finite
809: system behaves as: $J_{L} \sim L^{a-1-\omega}$. This result should be
810: compared to our analysis in Eq.~(\ref{jgriffiths}), where we
811: distinguish between the $z > 1$ and $z < 1$, respectively.
812:
813:
814: \subsubsection{Single particle transport $z > 1$}
815:
816: For $z>1$ we have $\omega=z-1$ and the current is so small that it is
817: produced by a finite number of active particles, i.e. $N_{a}=O(1)$. In
818: this case the accumulated distance traveled by all particles is simply
819: given by:
820: %
821: \be
822: X \sim t^{1/z}, \quad z>1\;.
823: \label{X_>}
824: \ee
825: %
826:
827: \subsubsection{Many particle transport $z < 1$}
828:
829: For $z < 1$, however, $\omega=0$ and $a=1-z$, thus the current in the
830: ZRP is produced by $N_a \sim L^{1-z}$ active particles.
831: Now the active
832: particles have a constant velocity, thus during time, $t$ they travel
833: a distance of $\sim t$.
834:
835:
836: Note that the number of active particles, $N_a$, is of the same order
837: of magnitude as the typical value of particles outside the condensate,
838: $N_{out}$, see Eq.~(\ref{N_out}). This is due to the fact that in a
839: typical sample there are only finite number of inactive particles,
840: this issue is discussed in details in the following subsection.
841:
842: \subsection{Inactive particles}
843:
844: Inactive particles in the random ZRP are also of two kinds. The first
845: kind of inactive particles is found in a ``cloud'', which is localized
846: next to the condensate. The formation of
847: this cloud is due to an attractive property of the
848: energy landscape: particles that left the condensate
849: and did not travel farther than a
850: distance $\xi$ will typically turn back.
851: The second sort of inactive particles are found to be localized at the
852: subleading extrema of the energy landscape, i.e. they are at a site
853: $\tilde{l}$, where $g_{\tilde{l}}$ in Eq.~(\ref{g}) takes the
854: next-to-leading value. As shown in Eq.~(\ref{g_i}) the typical value of
855: $\langle n_{\tilde{l}} \rangle$ is of the order of one. Its average
856: value, however, is divergent and will be determined below.
857:
858: \subsubsection{Typical behavior of the cloud}
859: \label{sec_cloud}
860:
861: Here we consider a typical sample in which the cloud is due to the
862: attraction of the condensate and well separated from
863: the subleading extrema of the landscape. In this situation the density
864: of inactive particles is expected to decay exponentially, $n_l \sim
865: \exp(-l/l_w)$, where $l$ measures the distance from the condensate and
866: $l_w$ is the typical width of the cloud. We estimate $l_w$ by the
867: following consideration. The correlated cluster at the condensate has
868: the largest size among the clusters and its value, $\xi_L$, follows
869: from extreme value statistics. Using the distribution function of
870: cluster sizes below Eq.~(\ref{xi}) we obtain for its typical value:
871: $\xi_L \sim \xi \ln L $. At this distance, $l=\xi_L$, the typical
872: value of the weight is $g_l=O(1)$, and with Eqs.~(\ref{nav}),
873: (\ref{geom}), and (\ref{g_L}) we obtain $n_l\sim L^{-z} \sim
874: \exp(-\xi_L/l_w)$. Consequently
875: %
876: \be
877: l_w \sim \frac{\xi}{z}=\xi_{typ} \sim \delta^{-1}\;,
878: \label{l_w}
879: \ee
880: %
881: where we have made use of the scaling relation\cite{fisher99,im}:
882: $\xi/\xi_{typ}=z$, where $\xi_{typ}$ is the typical
883: correlation length. The small $\delta$ behavior in the last equation
884: follows directly from Eqs.(\ref{xi}) and (\ref{z_small}). We have
885: obtained thus the result that the typical width of the cloud of
886: inactive particles is measured by the typical correlation length of
887: the biased Sinai walker.
888:
889: The typical density profile of inactive particles has the same scaling
890: behavior as the typical and the average value of the occupation
891: number, $\alpha_i$, as defined in Eq.~(\ref{geom}). For this latter
892: quantity we have checked numerically the scaling form in
893: Eq.~(\ref{l_w}) for different values of $\delta$ and satisfactory
894: agreement is found, see Fig.~\ref{pfg}.
895:
896:
897: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
898: \begin{figure}
899: \includegraphics[width=0.8\linewidth]{fig6.eps}
900: \caption{\label{pfg} Average of the occupation probability, $\alpha_i$, which corresponds to
901: the typical density of the cloud. The data were obtained by solving
902: Eq.(\ref{finite}) numerically for $L=N=2048$ and the disorder average was
903: performed over 500000 samples.
904: We used the uniform distribution in Eq.(\ref{uniform}) for different values of the
905: asymmetry parameter, $\delta$, and thus for the dynamical exponent, $z>1$.
906: The data are rescaled according to (\ref{alpha1}) and using the length scale
907: in (\ref{l_w}) with (\ref{z}).
908: }
909: \end{figure}
910: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
911:
912: \subsubsection{Average number of inactive particles}
913: \label{sec_inactive}
914:
915: The typical value of inactive particles is finite, however in rare
916: realizations it is possible to find even a macroscopic number of inactive
917: particles. This is due to the fact that the leading value of the
918: weights in Eq.~(\ref{g}), $g_L$, and the subleading value,
919: $g_{\tilde{l}} \equiv g$
920: can be arbitrarily close to each other.
921: The average value of $[\langle n_{\tilde{l}} \rangle ]_{\rm av}\approx
922: N_{ia}$ is dominated by such rare realizations, in which both $\langle n_L \rangle$ and
923: $\langle n_{\tilde{l}} \rangle$ are macroscopic, i.e. they are of $\mathcal{O}(N)$.
924: However, as we checked numerically, the contribution to the average value of such samples in which more than
925: two sites have macroscopic occupation is negligible. Therefore in the subset of rare
926: realizations we should consider only two active sites, $L$ and $\tilde{l}$,
927: and the unnormalized weight that the subleading site contains $n_{\tilde{l}}=n$
928: particles is in leading order
929: proportional to $g^{n}g_L^{N-n}= \alpha^{n}g_L^N$.
930: The distribution of $n$ is thus approximately
931: $P_{\alpha}(n)=(1-\alpha)/(1-\alpha^{N+1})\alpha^{n}$ for $\alpha<1$ and $P_{\alpha}(n)=1/(N+1)$,
932: for $\alpha=1$\cite{current}. Thus we obtain for the expectation value:
933: \be
934: \langle n \rangle (\alpha)={\alpha\over
935: 1-\alpha}-(N+1){ \alpha^{N+1}\over 1-\alpha^{N+1}},\quad \alpha<1\;,
936: \label{corr}
937: \ee
938: and $\langle n \rangle=N/2$ for $\alpha=1$. Using the distribution function, $\rho(\alpha)$ we
939: can average over the rare realizations:
940: $[\langle n_{\tilde{l}} \rangle ]_{\rm av}=
941: \int_0^1 \langle n \rangle (\alpha)\rho(\alpha) {\rm d}\alpha$, which is dominated by the
942: contribution as $\alpha \to 1$. Keeping in mind that the maximal value of $\langle n \rangle$ is
943: $N/2$, we can write
944: %
945: \be
946: N_{ia}\approx [\langle n_{\tilde{l}} \rangle ]_{\rm av} \approx
947: \int_0^{1-2/N} \frac{\alpha}{1-\alpha}\rho(\alpha) {\rm d}\alpha \sim \rho(1)
948: \ln N.
949: \ee
950: %
951: provided that the distribution $\rho(\alpha)$ has a finite limiting
952: value at $\alpha \to 1$. As we have
953: checked numerically this is indeed the case, both for
954: the unbiased and the biased (asymmetric) models. Consequently
955: the average number of inactive particles is logarithmically divergent, although its typical value
956: is finite. Furthermore the ratio between the average numbers of the active and inactive particles,
957: $N_a/N_{ia}$, tends to zero for $z>1$ and tends to infinity, for $z<1$.
958:
959: \subsubsection{Average density profile of inactive particles}
960: \label{sec_density}
961:
962: The average density profile of inactive particles, $[\langle n_l
963: \rangle]_{\rm av}$, is proportional to $P(l)$, which is the probability
964: density, that the subleading, almost degenerate $g \approx g_L$ is located at
965: $l=\tilde{l}$. In the asymmetric model $\tilde{l}$ can be either at the
966: renormalized cluster of the condensate or at any other cluster.
967: In the second case $P(l)$ is a constant, which happens for $l > \xi$.
968: This behavior is illustrated in the left part of Fig.~\ref{pf}.
969:
970: In the unbiased model, as can be seen from Eq.~(\ref{xi}) the width
971: of the cluster diverges. Therefore $\tilde{l}$ and $L$ are always
972: in the same cluster, and as a consequence the probability distribution is
973: scale-free and a function of $l/L$.
974: The form of the probability distribution for $l \ll L$
975: can be obtained from the random walk picture in Fig.~\ref{rw1}.
976: The rare event for this process is represented by a landscape which is
977: drawn by an unbiased ($\delta=0$) random walker, which starts at one
978: minimum and after $l$ steps arrives to a degenerate second minimum. This
979: means that the random walker is surviving (since it does not cross its
980: starting position) and returns to its origin. The fraction of such
981: random walks is given by\cite{ir}: $P(l) \sim l^{-3/2}$, $l \ll L$,
982: consequently $P(l)=L^{-3/2} \tilde{P}(l/L)$, where $\tilde{P}(x)$ is a
983: smooth scaling function, which behaves for small $x$ as $\tilde{P}(x)
984: \sim x^{-3/2}$. With this prerequisite we obtain for the average density
985: profile of inactive particles for a given density $\rho=N/L>0$:
986: %
987: \be
988: [\langle n_l\rangle ]_{av}(L)=\ln (L)L^{-3/2}\tilde{P}(l/L).
989: \label{scaling}
990: \ee
991: %
992: Results of numerical calculations are presented in the right part of
993: Fig.~\ref{pf}, which are in excellent agreement with the theoretical
994: predictions.
995:
996: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
997: \begin{figure}
998: \includegraphics[width=0.45\linewidth]{fig7a.eps}
999: \includegraphics[width=0.45\linewidth]{fig7b.eps}
1000: \caption{\label{pf} Average density profiles of inactive particles
1001: for different system sizes, obtained by solving
1002: Eq.(\ref{finite}) numerically. The density was $\rho =1$ and average
1003: was performed
1004: over 500000 samples.
1005: Left: asymmetric model with the uniform distribution, $p_0=3$. The profile is constant outside
1006: the renormalized cluster of the condensate, the size of which is finite.
1007: Right: unbiased model, $p_0=1$, rescaled according to (\ref{scaling}).}
1008: \end{figure}
1009: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1010:
1011:
1012: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1013: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1014:
1015: \section{Coarsening}
1016: \label{sec_coarsening}
1017: In the stationary state of the process typically almost all particles
1018: occupy single site. If initially the particles are uniformly
1019: distributed on the lattice, the system undergoes a coarsening process,
1020: meaning that the number of particles in the condensate and the typical
1021: size of empty regions is growing as time elapses. To be specific we
1022: consider finite $L$ and $N$ and define the length scale as\cite{kf} $l(t)\equiv
1023: \sigma^2(t)=[{1\over L}\sum_{i=1}^L\langle n_i^2\rangle (t)]_{av}$
1024: , which characterizes the typical number of particles at
1025: sites with a non-microscopic $n_i$, or equivalently the typical
1026: distance between
1027: these sites- In the RG procedure
1028: the latter corresponds to the typical cluster sizes (defined
1029: as the sum of the original links) at the energy scale $\Omega$.
1030: The growth rate of the typical distance, ${\rm d} l /{\rm d} t$,
1031: is proportional to the typical current at that energy scale.
1032: This leads to the differential equation:
1033: %
1034: \be
1035: \frac{{\rm d} l}{{\rm d} t} \sim J_l\;.
1036: \label{l_t}
1037: \ee
1038: %
1039: At the Griffiths phase we have in (\ref{jgriffiths}), $J_l\sim l^{-z}$,
1040: thus the solution of Eq.~(\ref{l_t}) is given by:
1041: %
1042: \be
1043: l \sim t^{1/\zeta},\quad \zeta=z+1\;.
1044: \label{dyn}
1045: \ee
1046: %
1047:
1048: Here $\zeta$ is the dynamical coarsening exponent. Note that $\zeta$ is exactly
1049: known through Eq.(\ref{z}) and it is a continuous function of
1050: the parameters appearing in the distributions of hop rates.
1051: Close to the critical point $\zeta$ becomes universal,
1052: depending only on the control parameter:
1053: $\zeta\approx 1/(2\delta )$, see Eq.(\ref{z_small}).
1054: We note that the same result can be obtained by referring to the result of the
1055: RG procedure. As shown in subsection \ref{sec_asym} for $\Omega <
1056: \Omega_{\xi}$ the ZRP is transformed to a totally asymmetric ASEP with a rate
1057: distribution in Eq.~(\ref{Pp}). The coarsening of the
1058: totally asymmetric ASEP
1059: has been analyzed in Refs.~[\onlinecite{kf,jainbarma,bennaim}]
1060: with the result given in Eq.~(\ref{dyn}).
1061:
1062: At the critical point $\zeta$ diverges and the coarsening is ultra-slow
1063: and strictly at $\delta = 0$ the length scale grows
1064: anomalously (logarithmically) with $t$.
1065: This type of growth is the so called {\it anomalous coarsening}
1066: \cite{evanscoars},
1067: which can be observed e.g. in spin glasses.
1068:
1069: The asymptotic time dependence of $l$ can be obtained by inserting
1070: the scaling form of the typical current through an empty region
1071: of size $l$ into the differential equation in Eq.~(\ref{l_t}). The
1072: scaling form is according (\ref{jcrit}) given by $J_l\sim e^{-c l^{1/2}}$.
1073: Eq.~(\ref{jcrit}) has the asymptotic (large $t$) solution $l^{1/2}e^{{\rm c}
1074: l^{1/2}}\sim t$,
1075: thus the growth of the length scale is logarithmically slow
1076: %
1077: \be
1078: l \sim \left[\ln\left({t\over \ln t}\right) \right]^2.
1079: \label{log}
1080: \ee
1081: %
1082: Results of numerical simulations are presented in Fig. \ref{fig20} for
1083: the Griffiths phase (biased model) and in Fig. \ref{fig21} for the unbiased model. The
1084: agreement with the theoretical results is satisfactory.
1085:
1086:
1087:
1088: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1089: \begin{figure}
1090: \includegraphics[width=0.45\linewidth]{fig8a.eps}
1091: \includegraphics[width=0.45\linewidth]{fig8b.eps}
1092: \caption{\label{fig20}
1093: Time dependence of the coarsening length scale, $l$, in the
1094: Griffiths phase in a log-log plot. We used the bimodal
1095: distribution (\ref{bimodal}) with c=0.2, r=0.5 where $z=0.5$ (left)
1096: and c=0.3, r=0.5 where $z\approx 0.818$ (right), having a density
1097: $\rho =3$, and disorder average was
1098: performed over a few hundred samples. The slope of the straight line
1099: is the theoretical result in Eq.(\ref{dyn}): $1/(1+z)$.}
1100: \end{figure}
1101: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1102:
1103:
1104: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1105: \begin{figure}
1106: \includegraphics[width=0.45\linewidth]{fig9a.eps}
1107: \includegraphics[width=0.45\linewidth]{fig9b.eps}
1108: \caption{\label{fig21}
1109: Time dependence of the coarsening length scale, $l$, for the unbiased
1110: model using the theoretical combination in Eq.~(\ref{log}). We used the bimodal
1111: distribution (\ref{bimodal})) with
1112: c=0.5, r=0.1 (left) and c=0.5, r=0.2 (right), having a density
1113: $\rho =3$, and disorder average was
1114: performed over a few hundred samples.}
1115: \end{figure}
1116: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1117:
1118: At the boundary of the Griffiths phase, where $z=0$,
1119: the coarsening is controlled
1120: by the behavior of the forward rate distribution at $p\to 0$.
1121: To see this, let us consider an arbitrary forward rate distribution for which
1122: $\rho (p)\sim p^{-1+1/z_0}$~~($z_0>0$) as $p\to 0$, and an arbitrary backward rate distribution,
1123: %e.g. $q_i=q_0$ for all $i$.
1124: such that $\delta >0$.
1125: Analyzing (\ref{z}) it is easy to show that $\zeta > z_0+1$ in the
1126: whole Griffiths phase $\delta >0$, and at the boundary
1127: \be
1128: \lim_{\delta\to\infty}\zeta (\delta )= z_0+1,
1129: \ee
1130: which is the coarsening exponent of a totally
1131: asymmetric process obtained from the original one by setting
1132: the backward rates to zero.
1133: Thus, as expected, the presence of nonzero backward rates slows down
1134: the coarsening in so far as the dynamical exponent is larger in the
1135: whole Griffiths phase than that of
1136: the totally asymmetric process with zero backward rates, and the latter
1137: value is recovered at the boundary of the Griffiths phase.
1138:
1139: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1140: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1141: \section{Conclusion}
1142:
1143: In this paper the prototype of a diffusive many particle system, the
1144: zero range process, is studied in one dimension, in the presence of
1145: quenched disorder. This model system is of large importance since its
1146: stationary state is in a product form therefore can be constructed
1147: analytically. At the same time there exists a wide range of
1148: different realisations of the ZRP, and therefore it can be applied
1149: to various problems of stochastic transport.
1150: We have analyzed the properties of the steady state of
1151: the disordered system and obtained many exact results. These results
1152: are expected to be generic for random, driven non-equilibrium systems
1153: and therefore can be useful to analyze more complicated,
1154: non-integrable processes, too. It also important to notice
1155: that our results are directly related to another archetypical model
1156: of non-equilibrium transport, i.e. the
1157: ASEP with particle-wise disorder.
1158:
1159: \subsection{Types of the transport}
1160:
1161: The transport in the system is related to the form of the
1162: hop rate distribution and characterized by the value of the dynamical
1163: exponent $z$ defined in Eq.~(\ref{z}). We can distinguish four main
1164: types of transport in the random ZRP.
1165:
1166: \subsubsection{$z=0$}
1167:
1168: In this case the maximal value of the backward hop rates, $q_{max}$,
1169: is smaller than the minimal value of the forward hop rates, $p_{min} >
1170: q_{max}$. This situation is equivalent to the totally asymmetric ZRP,
1171: which has been previously analyzed in the literature. In this case
1172: there is a finite stationary current in the system, such that the number
1173: of active particles (which carry the current) is $N_a \sim L$ and the
1174: particles have a finite velocity, $v=O(1)$. Condensation of
1175: particles at a particular site is not generic, but appears only
1176: for hop rate distributions that are vanishing fast enough at the
1177: lower cut-off $c>0$ (see \cite{krug} for details).
1178:
1179: \subsubsection{$0<z<1$}
1180:
1181: In this case $q_{max}>q_{min}$ and the current is vanishing in the
1182: thermodynamic limit as $J \sim L^{-z}$. The transport in the system is
1183: effected by $N_a \sim L^{1-z}$ active particles, which move with a finite
1184: velocity, $v=O(1)$. Note that in the system there are infinitely many
1185: active particles, but their density is zero.
1186:
1187: \subsubsection{$1<z<\infty$}
1188:
1189: The current in the system is vanishing according to $J \sim L^{-z}$,
1190: and the
1191: transport is effected by $N_a=O(1)$ active particles, which have a
1192: vanishing velocity. Thus the transport is realized by the anomalous
1193: diffusion of a few biased random walker.
1194:
1195: \subsubsection{Unbiased ZRP: $z \to \infty$}
1196:
1197: The average current is zero and the fluctuations of the accumulated
1198: displacement are due to a few Sinai walkers and given in
1199: Eq.~(\ref{sinai}).
1200:
1201: \subsection{Problem with self averaging}
1202:
1203: In the partially asymmetric random ZRP with $z>0$ the number of active
1204: particles, $N_a$, and thus the transport is self averaging, i.e. the
1205: typical and the average values have the same scaling behavior. The
1206: inactive particles, however, which are outside the condensate but do
1207: not contribute to the transport, have different properties. Their
1208: typical number is $O(1)$ whereas their average diverges as $\ln
1209: N$. This latter value is dominated by rare realizations in which there
1210: is a second site with macroscopic number of particles.
1211:
1212: \subsection{Relation with the totally asymmetric ZRP}
1213:
1214: The RG framework used in this paper has revealed a
1215: relation between the partially asymmetric ZRP with arbitrary type of
1216: initial disorder and the totally asymmetric ZRP with a hop
1217: rate distribution,
1218: which has a vanishing power-law tail. Indeed, during renormalization
1219: if the typical size of the renormalized new clusters becomes larger
1220: than the correlation length the new effective particles perform a totally
1221: asymmetric motion with a power-law distribution of the hop rates given
1222: in Eq.~(\ref{Pp}). The exponent of this distribution is related to the
1223: form of the distribution in the original (i.e. partially asymmetric)
1224: model, see in Eq.~(\ref{z}). In this way the problem which has been
1225: throughly studied for the random totally asymmetric ZRP appears
1226: naturally as the fixed point problem of the partially asymmetric ZRP.
1227: The power-law exponent then appears in a self-organized fashion. In
1228: the terminology of random systems both problems have the same type of
1229: {\it strong disorder fixed point}. We note that a similar relation is
1230: encountered between the biased Sinai walk and the directed trap model\cite{im}.
1231: On the other hand for the unbiased ZRP the
1232: singular behavior is governed by a so called {\it infinite disorder fixed point}.
1233:
1234: \subsection{Condensation transition for limited number of particles}
1235:
1236: Throughout the analysis of
1237: the partially asymmetric model we considered exclusively the
1238: case of a finite density, $\rho=N/L>0$. In this case there is
1239: always a condensate present, where one typically finds
1240: a finite fraction of the particles. However, if the number
1241: of particles scales as $N \sim
1242: L^{\omega}$, $0<\omega<1$, one expects that the condensate will
1243: disappear for sufficiently small values of $\omega$. Indeed, the
1244: transport in the system can involve $N_a \sim L^{1-z}$ particles, for $0<z<1$,
1245: consequently for $\omega<1-z$ all the particles carry current and the
1246: condensate is absent in the system.
1247: At the borderline case, with $N=A L^{1-z}$ one expects
1248: that the fraction of particles in the condensate varies with $A$ and
1249: that there is a condensation transition possibly at a finite value of
1250: $A=A_c=O(1)$. First numerical results support this scenario of the condensation
1251: transition, but a more detailed analysis of the phase transition,
1252: which is due to the slow dynamics of the process quite involved,
1253: is not yet completed. We are going to clarify this phenomena in
1254: a separate work.
1255:
1256: \subsection{Higher dimensions}
1257:
1258: The ZRP process has the same type of factorized steady state for any
1259: type of lattices, thus also in higher dimensions. Also the stationary
1260: weights involve site dependent quantities, $g_i$, which are the
1261: stationary weights of the random walk on the given lattice. The
1262: renormalization described in Sec.\ref{sec_RG} can be implemented
1263: numerically in this case. Similar calculations have already been
1264: performed for random quantum systems\cite{im}. During the RG procedure one
1265: obtains large effective clusters having a complicated topology. One
1266: might ask the question whether the strong and infinite disorder nature of
1267: the fixed points as obtained in 1d remains also in higher dimensions.
1268: The answer to this question is negative. We know that the underlying
1269: random walk process has an upper critical dimension, $d_u=2$, so that
1270: the Gaussian nature of the random walk remains\cite{luck} (in $d=2$ with
1271: logarithmic corrections) even in the presence of quenched disorder.
1272: The same type of irrelevance of disorder is expected to hold for the
1273: ZRP, too.
1274:
1275:
1276: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1277: \section{Acknowledgments}
1278:
1279: R.J. is indebted to H. Rieger for stimulating discussions.
1280: R.J. and L.S. acknowledge support by the Deutsche
1281: Forschungsgemeinschaft under grant No. SA864/2-1.
1282: This work has been
1283: supported by a German-Hungarian exchange program (DAAD-M\"OB), by the
1284: Hungarian National Research Fund under grant No OTKA TO34138, TO37323,
1285: TO48721, MO45596 and M36803.
1286:
1287: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1288: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1289:
1290:
1291: \appendix
1292:
1293: \section{Mapping to the RTFIC, the RG equations and their fixed point solution}
1294:
1295: Renormalization of the one-dimensional ZRP with quenched disorder
1296: as described in Sec.~\ref{sec_RG} is equivalent to the same procedure of a
1297: random transverse-field Ising chain (RTFIC) described by the Hamiltonian:
1298: %
1299: \be
1300: H=-\sum_{l=1}^{L} J_i
1301: \sigma^x_{l}\sigma^x_{l+1}-\sum_{l=1}^{L}h_l\sigma^z_{l},
1302: \label{Ising}
1303: \ee
1304: %
1305: where $\sigma^{x,z}_l$ are Pauli spin operators at site $l$, and the
1306: couplings, $J_l$, and the transverse fields, $h_l$, are
1307: quenched random variables. During renormalization the largest term in
1308: the Hamiltonian, $\Omega=\max(\{J_i\},\{h_i\})$ is successively eliminated.
1309: For example, if the largest term is a coupling, $\Omega=J_i$, the connected two sites, $(i,i+1)$,
1310: flip coherently and form an effective two-site cluster. The renormalized value
1311: of the transverse field acting on the spin cluster is given by a second-order perturbation
1312: calculation as:
1313: %
1314: \be
1315: \tilde{h} \approx \frac{h_i h_{i+1}}{\Omega}\;.
1316: \label{htilde}
1317: \ee
1318: %
1319: Decimating a strong transverse field, $\Omega=h_i$, will result in decimating out site, $i$,
1320: and generating a new coupling, $\tilde{J}$, between the remaining sites, $i-1$ and $i+1$. The
1321: value of $\tilde{J}$ follows from self-duality, in Eq.~(\ref{htilde}) one should replace $h_i$
1322: by $J_i$. Comparing the decimation rules for the ZRP in Eq.~(\ref{back}) to that for
1323: the RTFIC in Eq.~(\ref{htilde}) one observes a complete agreement with the correspondences:
1324: $p_i \leftrightarrow J_i$ and $q_i \leftrightarrow h_i$. Now properties of the RG flow
1325: for the ZRP can be obtained from the equivalent expressions for the RTFIC.
1326:
1327: For example the probability distribution of the backward, $R(q,\Omega)$, and forward,
1328: $P(p,\Omega)$, hop rates satisfy the set of integral-differential equations,
1329: %
1330: \beqn
1331: {{\rm d} R(q,\Omega) \over {\rm d}
1332: \Omega} &=& R(q,\Omega)[P(\Omega,\Omega)-R(\Omega,\Omega)]\\
1333: & - & P(\Omega,\Omega) \int_{q}^{\Omega} {\rm d} q' R(q',\Omega) R({q \Omega \over q'},\Omega)
1334: {\Omega \over q'} \nonumber \\
1335: {{\rm d} P(p,\Omega) \over {\rm d}
1336: \Omega}&=&P(p,\Omega)[R(\Omega,\Omega)-P(\Omega,\Omega)]\\
1337: & - &R(\Omega,\Omega) \int_{p}^{\Omega} {\rm d} p' P(p',\Omega) P({p \Omega \over p'},\Omega)
1338: {\Omega \over q'}, \nonumber
1339: \eeqn
1340: %
1341: having a solution at the fixed point:
1342: \beqn
1343: P_0(p,\Omega)=\frac{p_0(\Omega)}{\Omega}\left(\frac{\Omega}{p}\right)^{1-p_0(\Omega )}, \nonumber \\
1344: R_0(q,\Omega)=\frac{r_0(\Omega)}{\Omega}\left(\frac{\Omega}{q}\right)^{1-r_0(\Omega )},
1345: \label{sol}
1346: \eeqn
1347: with $0<p,q\le \Omega$.
1348: The value of the exponents, $p_0(\Omega)$ and $r_0(\Omega)$, depend on original distributions
1349: and therefore on the value of the control parameter, $\delta$. The specific values are given in
1350: Sec.\ref{sec_RG}.
1351:
1352: \section{Average of the occupation probability}
1353:
1354: Here we start with the asymptotic distribution of the (uncorrelated) $g_i$ weights in
1355: Eq.~(\ref{kesten}) with the condition that the largest value is fixed, $g_L=G$,
1356: so that
1357: %
1358: \be
1359: \rho(g|G)dg={1/zg^{-1/z-1}\over 1-G^{-1/z}}\Theta(G-g).
1360: \ee
1361: %
1362: From this we obtain for the distribution of the occupation probabilities, $\alpha_i=g_i/g_L$, for large $L$ as
1363: \be
1364: \rho(\alpha)d\alpha\approx {1\over z}{1\over \alpha}\left[{1\over L\alpha^{1/z}}-
1365: e^{- L\alpha^{1/z}}\left(1+{1\over L\alpha^{1/z}} \right) \right]d\alpha.
1366: \ee
1367: %
1368: The average of $\alpha$ is therefore given by:
1369: %
1370: \beqn
1371: [\alpha]_{av}&=&\int_0^1\alpha\rho(\alpha)d\alpha \nonumber \\
1372: &\approx &
1373: L^{-z}\int_0^L\left[{1\over y}-e^{-y}\left(1+{1\over y}\right) \right]
1374: y^{-1+z}dy \ .
1375: \eeqn
1376: %
1377: In leading order of $L$ the second term in the bracket can be neglected and we arrive at:
1378: %
1379: \beqn
1380: [\alpha]_{\rm av} &\approx& L^{-z}\left({\rm const}+{L^{z-1}\over z-1}\right)\sim
1381: L^{-z},\quad z<1, \nonumber \\
1382: \left[\alpha\right]_{\rm av} &\approx& L^{-1}\left({\rm const}+\ln L\right)\sim
1383: L^{-1}\ln L \quad z=1,\nonumber \\
1384: \left[\alpha\right]_{\rm av}&\approx& L^{-z}\left({\rm const}+{L^{z-1}\over z-1}\right)\sim
1385: \frac{L^{-1}}{z-1} \quad z>1. \label{alpha1}
1386: \eeqn
1387: %
1388: Thus, as far as $z>1$ the occupation probability is
1389: non-self-averaging, and the decay exponent is becoming
1390: one independent of $\delta$.
1391:
1392:
1393: \begin{thebibliography}{}
1394:
1395: \bibitem{krug} J. Krug, Braz. J. Phys. {\bf 30}, 97 (2000).
1396:
1397: \bibitem{im} F. Igl\'oi and C. Monthus, Physics Reports {\bf 412}, 277, (2005),
1398: preprint cond-mat/0502448.
1399:
1400: \bibitem{barma} G. Tripathy and M. Barma, Phys. Rev. Lett. {\bf 78}, 3039 (1997); Phys. Rev. E {\bf 58}, 1911 (1998.)
1401:
1402: \bibitem{janowsky} S.\ A.\ Janowsky and J.L. Lebowitz, Phys. Rev A {\bf 45}, 618 (1992); J. Stat. Phys. {\bf 77}, 35 (1994).
1403:
1404: \bibitem{sinai} For a review see: J.P. Bouchaud and A. Georges, Phys. Rep. {\bf 195}, 127 (1990).
1405:
1406: \bibitem{kln04} Y.~Kafri, D.K.~Lubensky, and D.R. Nelson,
1407: Biophys. J. {\bf 86}, 3373 (2004).
1408:
1409: \bibitem{kln05} Y.~Kafri, D.K.~Lubensky, and D.R. Nelson,
1410: Phys.~Rev. E {\bf 71}, 041906 (2005).
1411:
1412: \bibitem{evansreview} For a review see: M.R. Evans and T. Hanney,
1413: preprint cond-mat/0501338.
1414:
1415: \bibitem{spitzer} F. Spitzer (1970) Advances in Math. {\bf 5} 246.
1416:
1417: \bibitem{chowd} D. Chowdury, L.~Santen, A.~Schadschneider, Phys. Rep. {\bf 329}, 199 (2000).
1418:
1419: \bibitem{jainbarma} K. Jain and M. Barma, Phys. Rev. Lett. {\bf 91}, 135701 (2003).
1420:
1421: \bibitem{liggett} T.M. Liggett {\it Stochastic interacting systems:
1422: contact, voter, and exclusion processes}, (Berlin, Springer, 1999).
1423:
1424: \bibitem{hinrichsen} H. Hinrichsen, Adv.~Phys.~{\bf 49}, 815 (2000).
1425:
1426: \bibitem{schutzreview} G.M. Sch\"utz, in {\it Phase Transitions and
1427: Critical Phenomena}, vol. 19, Eds. C. Domb and J.L. Lebowitz (Academic Press, San Diego, 2001).
1428:
1429: \bibitem{kf} J. Krug and P. A. Ferrari, J. Phys. A {\bf 29}, L465 (1996).
1430:
1431: \bibitem{evans} M. R. Evans, Europhys. Lett. {\bf 36}, 13 (1996).
1432:
1433: \bibitem{jsi} R. Juh\'asz, L. Santen and F. Igl\'oi,
1434: Phys. Rev. Lett. {\bf 94}, 010601 (2005).
1435:
1436: \bibitem{derrida} C. Enaud and B. Derrida, Europhys. Lett. {\bf 66}, 83 (2004).
1437:
1438: \bibitem{stinchcombe} R.J. Harris and R.B. Stinchcombe, Phys. Rev. E {\bf 70} 016108(E) (2004).
1439:
1440: \bibitem{MDH} S.K. Ma, C. Dasgupta, and C.-K. Hu, \prl {\bf 43}, 1434 (1979);
1441: C. Dasgupta and S.K. Ma, Phys. Rev. B {\bf 22}, 1305 (1980).
1442:
1443: \bibitem{DF} D.S. Fisher, Phys. Rev. Lett. {\bf 69}, 534 (1992); Phys. Rev. B {\bf 51}, 6411 (1995).
1444:
1445: \bibitem{fisherxx} D.S. Fisher, Phys. Rev. B {\bf 50}, 3799 (1995).
1446:
1447: \bibitem{RGsinai} D.S. Fisher, P. Le Doussal, and C. Monthus, Phys. Rev. E{\bf 59} 4795 (1999).
1448:
1449: \bibitem{hiv} J. Hooyberghs, F. Igl\'{o}i, and C. Vanderzande, Phys. Rev. Lett. {\bf 90}, 100601 (2003);
1450: Phys. Rev. E {\bf 69}, 066140 (2004).
1451:
1452: \bibitem{kafri} Y. Kafri, E. Levine, D. Mukamel, G.M. Sch\"utz and
1453: J. T\"or\"ok, Phys. Rev. Lett. {\bf 89}, 035702 (2002).
1454:
1455: \bibitem{griffiths} R.B. Griffiths, Phys. Rev. Lett. {\bf 23}, 17 (1969).
1456:
1457: \bibitem{kesten} H. Kesten, Acta Math. {\bf 131}, 298 (1973);
1458: B. Derrida and H. Hilhorst, J. Phys. A{\bf 16}, 2641 (1983); C. de
1459: Calan, J. M. Luck, T. M. Nieuwenhuizen and D. Petritis, J. Phys. A {\bf 18}, 501 (1985).
1460:
1461: \bibitem{galambos} J. Galambos, {\it The Asymptotic Theory of Extreme
1462: Order Statistics} (John Wiley and Sons, New York, 1978).
1463:
1464: \bibitem{ir} F. Igl\'oi and H. Rieger, Phys. Rev. B {\bf 57}, 11404 (1998).
1465:
1466: \bibitem{i02} F. Igl\'oi, Phys. Rev. B{\bf 65}, 064416 (2002).
1467:
1468: \bibitem{sinai1} Y.A.G. Sinai, Theor. Prob. Appl. {\bf 27}, 256 (1982).
1469:
1470: \bibitem{fisher99}
1471: D.S. Fisher, Physica A {\bf 263}, 222 (1999).
1472:
1473: \bibitem{current} In the framework of this simplified picture we
1474: obtain for the finite $N$ correction to the current:
1475: $J_L/J_{L,N}\approx 1+P_{\alpha}(N)$.
1476:
1477: \bibitem{bennaim} E. Ben-Naim, P.L. Krapivsky, and S. Redner,
1478: Phys. Rev. E {\bf 50}, 822 (1994).
1479:
1480: \bibitem{evanscoars} M.R. Evans, J. Phys: Condens. Matter {\bf 14},
1481: 1397 (2002).
1482:
1483: \bibitem{luck}
1484: J.M. Luck, Nucl. Phys. B {\bf 225}, 169 (1983); D.S. Fisher, Phys. Rev. A {\bf 30}, 960 (1984).
1485:
1486: \end{thebibliography}
1487:
1488: \end{document}
1489:
1490: