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,floatfix]{revtex4}
9: \documentclass[twocolumn,aps,amsmath,amssymb,pre]{revtex4}
10: %\usepackage{psfig}
11: \usepackage{graphicx}
12: \usepackage{epsfig}
13: \bibstyle{apsrev.bib}
14:
15:
16: \begin{document}
17: \input epsf.sty
18:
19: \title{Weakly coupled, antiparallel, totally asymmetric simple exclusion processes}
20:
21:
22: \author{R\'obert Juh\'asz}
23: \email{juhasz@szfki.hu}
24: \affiliation{Research Institute for Solid
25: State Physics and Optics, H-1525 Budapest, P.O.Box 49, Hungary}
26:
27: \date{\today}
28:
29: \begin{abstract}
30: We study a system composed of two parallel totally asymmetric
31: simple exclusion processes with open boundaries,
32: where the particles move in the two
33: lanes in opposite directions and are allowed to jump to the other lane
34: with rates
35: inversely proportional to the length of the system.
36: Stationary density profiles are determined and the phase diagram of
37: the model is constructed in the hydrodynamic limit, by
38: solving the differential equations describing the steady state of the system,
39: analytically for vanishing total current and
40: numerically for nonzero total current.
41: The system possesses phases with a localized shock in the density
42: profile in
43: one of the lanes, similarly to exclusion processes endowed with
44: nonconserving kinetics in the bulk. Besides, the system undergoes a
45: discontinuous phase transition,
46: where coherently moving delocalized shocks emerge in both lanes
47: and the fluctuation of the global density is
48: described by an unbiased random walk. This phenomenon
49: is analogous to the phase coexistence observed at the coexistence line of the totally asymmetric simple
50: exclusion process, however, as a consequence of
51: the interaction between lanes, the density profiles are deformed and
52: in the case of asymmetric lane change, the motion of the shocks is
53: confined to a limited domain.
54:
55: \end{abstract}
56:
57:
58: \maketitle
59:
60: \newcommand{\bc}{\begin{center}}
61: \newcommand{\ec}{\end{center}}
62: \newcommand{\be}{\begin{equation}}
63: \newcommand{\ee}{\end{equation}}
64: \newcommand{\beqn}{\begin{eqnarray}}
65: \newcommand{\eeqn}{\end{eqnarray}}
66:
67:
68:
69: \vskip 2cm
70: \section{Introduction}
71:
72: The investigation of interacting stochastic driven diffusive systems
73: plays an important role in the understanding of nonequilibrium
74: steady states \cite{liggett,zia}.
75: As opposed to equilibrium statistical mechanics,
76: phase transitions may occur in these systems even in one spatial
77: dimension \cite{evans}.
78: The paradigmatic model of driven lattice gases is
79: the one-dimensional totally asymmetric
80: simple exclusion process (TASEP) \cite{mcdonald,schutzreview},
81: which exhibits boundary induced
82: phase transitions \cite{krug} and the steady state of which is exactly
83: known \cite{derrida,schutzdomany}.
84: Beside theoretical interest, this model and its numerous variants have
85: found a wide range of applications,
86: such as the description of vehicular traffic \cite{santen} or modeling
87: of transport processes in biological systems \cite{schad}.
88: Inspired by the traffic of cytoskeletal motors \cite{howard}, such models were
89: introduced where a totally asymmetric exclusion
90: process is coupled to a finite compartment where the motion of
91: particles is diffusive \cite{nieuwenhuizen,klumpp,muller}.
92: Recently, the attention has turned to exclusion processes endowed with
93: various types of reactions which violate the conservation
94: of particles in the bulk
95: \cite{challet,frey,schutz03,ejs,rakos1,js,klumpp2,levine,ehk,rakos2,pierobon,mobilia,hinsch,oshanin}.
96: The simplest one among these models is the TASEP with ``Langmuir
97: kinetics'', where particles are created and annihilated also
98: at the bulk sites of the
99: system \cite{frey}---a process, which may serve as a simplified model
100: for the cooperative motion of molecular motors along a filament from
101: which motors can detach and attach to it again.
102: For these types of systems, the time scale of nonconserving processes compared
103: to that of directed motion and the processes at the boundaries is crucial.
104: If the nonconserving reactions occur with rates of larger order than
105: the inverse of the system size $L$, then in the large $L$ limit,
106: they dominate the stationary state.
107: On the contrary, when they are of smaller order than
108: $\mathcal{O}(1/L)$,
109: they are irrelevant and the stationary state is identical to that of
110: the underlying driven diffusive system.
111: However, in the marginal case when the rates of nonconserving processes
112: are of order
113: $\mathcal{O}(1/L)$, the
114: interplay between them and the boundary processes may result in
115: intriguing phenomena, such as ergodicity breaking \cite{rakos1,rakos2} or
116: the appearance of a localized shock in the density
117: profile \cite{frey}, which is in contrast to the delocalized shock
118: dynamics at the coexistence line of the TASEP
119: \cite{schutzdomany,schutz98}. The formation of domain walls
120: can be observed also experimentally in the transport of kinesin
121: motors in accordance with theoretical predictions \cite{konzack,nishinari,greulich}.
122:
123:
124: Other systems which have an intermediate complexity compared to exclusion
125: processes with bulk reactions and those coupled to a compartment
126: are the two-channel or multichannel systems.
127: In these models, particles are either conserved by the
128: dynamics in each lane and interaction is realized by the dependence
129: of the hop rates on the configuration of the parallel lanes
130: \cite{popkov,peschel,lee,salerno,popkov1,narrow},
131: or particles can jump between lanes \cite{pk,harris,mitsudo,reichenbach}.
132: We study in this work a two-lane exclusion process where particles
133: move in the two lanes in opposite directions.
134: Particles are allowed to change lanes and we restrict ourselves to the
135: case of weak lane change rates, i.e. they are inversely proportional
136: to the system size. This means that the probability that a
137: marked particle changes lanes during the time it resides in the system
138: is $\mathcal{O}(1)$.
139: If particles in one of the lanes are regarded as holes, and vice versa,
140: this model can also be interpreted as a two-channel driven system where
141: particles move in the same direction in the channels and are
142: created and annihilated in pairs.
143: In the hydrodynamic limit of the model, we shall construct the
144: steady-state phase diagram by means of analyzing the differential
145: equations describing the model on the macroscopic scale.
146: At the coexistence line, where coherently moving delocalized shocks
147: develop in both lanes, which is reminiscent of the delocalized
148: shock dynamics at the coexistence line of the TASEP, the density
149: profiles are studied in the framework of a phenomenological domain
150: wall picture based on the hydrodynamic description.
151: Recently, a two-lane exclusion process has been investigated with
152: weak, symmetric lane change, where particles move in the lanes
153: in the same direction
154: \cite{reichenbach}. In this model, the formation of delocalized
155: shocks in both lanes has been found, as well.
156: In our model, even the case of asymmetric lane change can be
157: treated analytically in the hydrodynamic limit if
158: the total current is zero, which holds also at the coexistence line.
159:
160: The paper is organized as follows. In Sec. \ref{description}, the model
161: is introduced and the hydrodynamic description is discussed.
162: In Sec. \ref{slc}, the case of symmetric lane change is investigated, while
163: Sec. \ref{alc} is devoted to the asymmetric case. The results are
164: discussed in Sec. \ref{discussion} and some of the calculations
165: are presented in two Appendixes.
166:
167: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
168: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
169:
170: \section{Description of the model}
171: \label{description}
172:
173: The model we focus on consists of two parallel one-dimensional
174: lattices with $L$ sites, denoted by $A$ and $B$, the sites of which
175: are either empty or occupied
176: by a particle.
177: The state of the system is specified by the set of
178: occupation numbers $n^{A,B}_i$ which are zero (one) for empty
179: (occupied) sites.
180: We consider in this system a continuous-time stochastic process where
181: the occupations of pairs of adjacent sites change independently
182: and randomly after an exponentially distributed waiting time.
183: The possible transitions and the
184: corresponding rates, i.e. the inverses of the mean waiting times,
185: are the following (Fig. \ref{model}).
186: On chain $A$, particles attempt to jump to the adjacent
187: site on their right-hand side, whereas on chain $B$ to the adjacent
188: site on their left-hand side, with a rate which is set to unity, and attempts are
189: successful when the target site is empty.
190: On the first site of chain $A$ and on the $L$th site of chain $B$
191: particles are injected with rate $\alpha$, provided these sites are
192: empty, whereas on the $L$th site
193: of chain $A$ and on the first site of chain $B$ they are removed with
194: rate $\beta$.
195: So the system may be regarded to be in contact with
196: virtual particle reservoirs with densities $\alpha$ and
197: $1-\beta$ at the entrance- and exit sites, respectively.
198: The process described so far is composed of two independent
199: totally asymmetric simple exclusion processes.
200: The interaction between them is realized by allowing
201: a particle residing at site $i$ of chain $A$($B$) to hop to site $i$ of chain
202: $B$($A$) with rate $\omega_{A}$ ($\omega_{B}$), provided the target site
203: is empty.
204:
205: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
206: \begin{figure}[h]
207: \includegraphics[width=0.8\linewidth]{fig1.eps}
208: \caption{\label{model} Transitions and the corresponding
209: rates in the model under study.}
210: \end{figure}
211: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
212:
213: As in the case of the TASEP with Langmuir kinetics,
214: one must distinguish here between three cases, concerning the order
215: of magnitude of the lane change rates in the large $L$ limit.
216: If the rates $\omega_{A}$ and $\omega_{B}$ are of larger order than
217: $\mathcal{O}(1/L)$, then in the limit $L\to\infty$,
218: the interchain processes are dominant compared to the effects of the
219: boundary reservoirs and the horizontal motion of the particles.
220: The densities $\rho$ and $\pi$ in lane $A$ and $B$, respectively,
221: are expected to be constant far from
222: the boundaries and to fulfill the relation
223: $\omega_{A}\rho (1-\pi)= \omega_{B}\pi (1-\rho)$,
224: which is forced by the lane change kinetics.
225: When the interchain hop rates are smaller than
226: $\mathcal{O}(1/L)$, then (apart from some possible special parameter combinations\cite{js})
227: they are irrelevant in the $L\to \infty$ limit and the stationary state
228: is that of two independent exclusion processes.
229: An interesting situation arises
230: if the rates $\omega_{A}$ and $\omega_{B}$ are proportional to $1/L$.
231: In this case the effects of boundary reservoirs and those of lane
232: change kinetics are comparable and the
233: competition between them results in nontrivial density profiles.
234: We focus on this case in the present work, and
235: parametrize the lane change rates as $\omega_{A}=\Omega_{A}/L$ and
236: $\omega_{B}=\Omega_{B}/L$, with the constants $\Omega_{A}$ and
237: $\Omega_{B}$.
238: Setting the lattice constant $a$ to $a=1/L$ and
239: rescaling the time $t$ as $\tau=t/L$,
240: we are interested in the properties of the system in the (continuum) limit
241: $L\to\infty$, where the state of the system is
242: characterized by the local densities
243: $\rho(x,\tau)$ and $\pi(x,\tau)$ on
244: chain $A$ and $B$, respectively, which are functions of the continuous
245: space variable $x\in [0,1]$ and time $\tau$.
246: Turning our attention to the subsystem containing lane
247: $A$($B$) alone, we see that the
248: interchain hoppings can be interpreted as
249: bulk nonconserving processes for the TASEP in lane $A$($B$).
250: The bulk reservoir which the TASEP is connected to is, however,
251: not homogeneous but
252: it is characterized by the position and time-dependent density $\pi(x,\tau)$($\rho(x,\tau)$).
253: Generally, driven diffusive systems which
254: are combined with a weak (i.e. $\mathcal{O}(1/L)$)
255: bulk nonconserving process
256: are described on the macroscopic scale specified above by the partial
257: differential equation
258: \be
259: \frac{\partial \rho (x,\tau)}{\partial \tau}+\frac{\partial J(\rho (x,\tau))}{\partial x}=S(\rho(x,\tau)),
260: \label{hgen}
261: \ee
262: where $S(\rho(x,t))$ is the
263: source term related to the nonconserving process and $J(\rho)$
264: is the current as a function of the density
265: in the steady state of the corresponding translation invariant
266: infinite system
267: without nonconserving processes
268: (i.e. $S(\rho(x,\tau))\equiv 0$) \cite{schutz03}.
269: Under these circumstances, the TASEP has a product measure stationary state
270: and the current-density relationship is simply $J(\rho)=\rho(1-\rho)$,
271: hence
272: the currents in the two lanes are
273: given as $J_A(\rho(x,\tau))=\rho(x,\tau)[1-\rho(x,\tau)]$ and $J_B(\pi(x,\tau))=-\pi(x,\tau)[1-\pi(x,\tau)]$ in terms
274: of the local densities.
275: For the source terms in the two lanes, we may write
276: $S_A(\rho(x,\tau),\pi(x,\tau))=-S_B(\rho(x,\tau),\pi(x,\tau))=\Omega_{B}[1-\rho(x,\tau)]\pi(x,\tau)-\Omega_{A}\rho(x,\tau)[1-\pi(x,\tau)]$
277: since the lane change events at a given site are infinitely rare in the limit
278: $L\to\infty$.
279: Setting these expressions into eq. (\ref{hgen}) we obtain that
280: in the steady state, where
281: $\partial_{\tau}\rho(x,\tau)=\partial_{\tau}\pi(x,\tau)=0$,
282: the density profiles $\rho(x)$ and $\pi(x)$
283: satisfy the coupled differential equations
284: \beqn
285: (2\rho-1)\partial_x\rho +
286: \Omega_{B}(1-\rho)\pi-\Omega_{A}\rho(1-\pi)=0, \nonumber \\
287: (2\pi-1)\partial_x\pi +
288: \Omega_{B}(1-\rho)\pi-\Omega_{A}\rho(1-\pi)=0.
289: \label{diff1}
290: \eeqn
291: We mention that one arrives at the same differential equations when in
292: the master equation of the process the expectation values
293: of pairs of occupation numbers $\langle n_in_{j}\rangle$
294: are replaced by the products $\langle n_i\rangle \langle n_{j}\rangle$, and
295: afterwards it is turned to a continuum description with retaining only
296: the first derivatives of the densities and neglecting the higher derivatives
297: which are at most of the order $\mathcal{O}(1/L)$ almost everywhere.
298:
299: For the stationary density profiles the boundary conditions
300: $\rho(0)=\pi(1)=\alpha$ and $\rho(1)=\pi(0)=1-\beta$
301: are imposed.
302: In fact, we shall keep these boundary conditions only for
303: $\alpha,\beta\le 1/2$; otherwise, we modify them
304: for practical purposes at the level of the hydrodynamic
305: description. The reason for this is the following.
306: In the domain $\alpha,\beta>1/2$ of the TASEP, the so-called maximum
307: current phase, the current is limited by the maximal
308: carrying capacity in the bulk, $J=1/4$, which is realized at
309: the bulk density $\rho=1/2$ \cite{derrida,schutzdomany}.
310: In this phase, boundary layers form in the stationary density profile at both
311: ends, where the density drops to the bulk value $\rho=1/2$.
312: Similarly, in the case of the TASEP with Langmuir kinetics,
313: if the entrance rate $\alpha$ exceeds the value $1/2$,
314: then in the density profile dictated by the reservoir at the entrance
315: site, a boundary layer develops, where the density drops to $1/2$.
316: The width of the boundary layer
317: is growing sublinearly with $L$ \cite{js},
318: such that in the hydrodynamic limit, it shrinks to $x=0$
319: and $\lim_{x\to 0}\rho(x)=1/2$ holds, independently of
320: $\alpha$, which influences only the shape of the microscopic
321: boundary layer.
322: These considerations apply also to the present model at both ends and for both
323: lanes.
324: Therefore, in order to simplify the treatment of the problem
325: at the level of the {\it hydrodynamic} description, we use the
326: effective boundary conditions
327: \be
328: \rho(0)=\pi(1)=a, \qquad \rho(1)=\pi(0)=1-b,
329: \label{bc}
330: \ee
331: where $a\equiv \min\{\alpha,1/2\}$ and $b\equiv \min\{\beta,1/2\}$.
332: However, we stress that, although, the profile propagating from e.g. the
333: left-hand boundary$\rho_l(x)$ is continuous at $x=0$ according to the
334: effective boundary conditions (\ref{bc}) for $\alpha>0$, a boundary
335: layer forms on the {\it microscopic} scale.
336:
337: In addition to the boundary layers related to the maximal
338: carrying capacity in the bulk, the stationary density profiles may
339: in general contain another type of boundary layer of finite width
340: or a localized shock in the bulk, where the density has a
341: finite variation within a region the width of
342: which is growing sublinearly with $L$ \cite{frey,ejs}.
343: This leads to the appearance of discontinuities in
344: $\rho(x)$ and $\pi(x)$ in the hydrodynamic limit, either
345: in the bulk $0<x<1$ in the case of a shock or
346: at $x=0,1$ in the case of a boundary layer.
347: This is in accordance with the fact that, in general, there does not
348: exist a continuous solution to the two first order differential
349: equations, which fulfills all four boundary
350: conditions.
351: Apart from some special parameter combinations, there is
352: one discontinuity in each lane, which is either in the bulk (a shock)
353: or at $x=0,1$ (a boundary layer).
354: The location of the discontinuity is determined by the requirement
355: that the currents in both lanes $J_A(\rho(x))$ and $J_B(\rho(x))$ must be
356: continuous functions of $x$ in the bulk $0<x<1$ \cite{schutz03,ejs}.
357: This follows from that the width of the shock region is proportional
358: to $\sqrt{L}$,
359: thus the rate of a lane change event is vanishing there in the limit
360: $L\to\infty$. This condition permits only such a shock
361: which separates complementary densities on its two sides, i.e.
362: $\rho$ and $1-\rho$ in lane $A$ or $\pi$ and $1-\pi$ in lane $B$.
363: The position of the shock $x_s$ e.g. in lane A is thus given implicitly
364: by the equation $\rho_l(x_s)=1-\rho_r(x_s)$, where $\rho_l(x)$ and
365: $\rho_r(x)$ are the solutions on the two sides of the shock.
366: For the detailed rules on the stability of the discontinuity at
367: $x=0,1$ see Ref. \cite{schutz03}.
368:
369:
370: Subtracting the two differential equations yields the obvious result that the total
371: current
372: \be
373: J\equiv \rho(x)[1-\rho(x)]-\pi(x)[1-\pi(x)]
374: \label{J}
375: \ee
376: is a (position independent) constant.
377: This relation makes it possible to eliminate one of the
378: functions, say, $\pi(x)$ and to reduce the problem to the
379: integration of a single differential equation
380: \be
381: \frac{d\rho}{dx}=\Omega_{A}\frac{\rho-[\frac{1}{2} \pm \sqrt{(\rho -\frac{1}{2})^2+J}][K(1-\rho)+\rho]}{2\rho -1},
382: \label{diffgen}
383: \ee
384: where we have introduced the ratio of lane change rates
385: $K\equiv \Omega_B/\Omega_A$
386: and the signs in front of the square root are related to the
387: two solutions $\pi_+(x)>1/2$ and $\pi_-(x)<1/2$ of the quadratic
388: equation (\ref{J}).
389: Disregarding the simple case $K=1$,
390: there are two difficulties about this equation.
391: First, the solution depends on the current $J$ as a parameter, which itself
392: depends on the density profiles and is {\it a priori} not known.
393: Fortunately, apart from two phases in the phase diagram, $\rho(x)$ and
394: $\pi(x)$ simultaneously fit to the boundary conditions either at
395: $x=0$ or $x=1$, consequently, the current is exclusively determined
396: by the entrance- and exit rates as $J=a(1-a)-b(1-b)$.
397: In the remaining two phases, the functions $\rho(x)$ and $\pi(x)$
398: meet the boundary
399: conditions at the opposite ends of the system. Here, one may solve
400: eq. (\ref{diffgen}) iteratively until self-consistency is attained.
401: Second, even in the case when $J$ is known, eq. (\ref{diffgen}) cannot be
402: analytically integrated in general, except for the case when the current
403: is zero.
404: This is realized in three cases, two of which are related to the
405: symmetries of the system. We discuss these possibilities in the rest
406: of the section.
407:
408: As the two entrance- and exit rates were chosen to be identical,
409: the obvious relation holds when the rates $\Omega_{A}$ and
410: $\Omega_{B}$ are interchanged:
411: \be
412: \rho(x;\alpha,\beta,\Omega_{A},\Omega_{B})=
413: \pi(1-x;\alpha,\beta,\Omega_{B},\Omega_{A}),
414: \label{sym1}
415: \ee
416: where the dependence of the profiles on the four parameters
417: $\alpha$,$\beta$,$\Omega_{A}$ and $\Omega_{B}$ is explicitly indicated.
418: This relation, together with eq. (\ref{J}) implies
419: that the current changes sign if $\Omega_{A}$ and $\Omega_{B}$ are
420: interchanged.
421: Thus $\Omega_{A}=\Omega_{B}$ implies $J=0$, that holds apparently
422: since none of the chains is singled out in this case.
423:
424: As a consequence of the particle-hole symmetry of the model,
425: we have the relation
426: \be
427: \rho(x;\alpha,\beta,\Omega_{A},\Omega_{B})=
428: 1-\pi(x;\beta,\alpha,\Omega_{A},\Omega_{B}).
429: \label{sym2}
430: \ee
431: Using eq. (\ref{J}), it follows that the current
432: changes sign when $\alpha$ and $\beta$ are interchanged,
433: so it must be zero for $\alpha=\beta$.
434: Alternatively, this can be seen by interchanging particles and holes
435: in one of the chains, which results in a two-channel system where
436: particles move in the channels in the same direction
437: and particles are created
438: and annihilated in pairs at neighboring sites of the two chains with
439: rates $\Omega_{A}$ and $\Omega_{B}$, respectively. Particles are injected and
440: removed in the channels with the same rate, hence the channels are
441: equivalent. Since the currents of particles and holes are equal,
442: the total current must be zero.
443:
444:
445: The third parameter regime where the current is zero is the domain
446: $\alpha,\beta \ge1/2$.
447: Here, as aforesaid, the density profiles and the current are
448: independent of $\alpha$ and $\beta$ in the hydrodynamic limit.
449: Since the current is zero for $\alpha=\beta$ it follows that
450: $J=0$ in the whole domain $\alpha,\beta \ge 1/2$.
451:
452:
453: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
454: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
455: \section{Symmetric lane change}
456: \label{slc}
457: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
458: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
459:
460: We start the investigation of the model with the simple case
461: of equal lane change rates ($\Omega_{A}=\Omega_{B}\equiv\Omega$),
462: where the solutions to the hydrodynamic
463: equations are analytically found and some general features of the model
464: can be understood.
465: Since the total current is zero, either $\rho(x)=\pi(x)$
466: or $\rho(x)=1-\pi(x)$ must hold.
467: Substituting the former relation into eq. (\ref{diff1}) yields
468: \be
469: \rho_e(x)=\pi_e(x)={\rm const},
470: \label{ek1}
471: \ee
472: whereas the latter gives
473: \be
474: \rho_c(x)=1-\pi_c(x)=\Omega x+{\rm const}.
475: \label{ck1}
476: \ee
477: Thus, the profiles
478: are piecewise linear and consist of constant segments with equal
479: densities in the two lanes and segments of slope $\Omega$
480: ($-\Omega$) in lane $A$($B$) with complementary densities.
481: Switching off the interchain particle exchange ($\Omega=0$),
482: we get two identical TASEPs,
483: which have, apart from the coexistence line
484: $\alpha=\beta <1/2$, constant density profiles in the bulk.
485: In the high-density phase ($\beta<\min\{\alpha,1/2\}$),
486: the density, being $1-\beta$, is controlled by the exit rate
487: and the profile is discontinuous at $x=0$.
488: In the low-density phase ($\alpha<\min\{\beta,1/2\}$),
489: the density is
490: $\alpha$ and a discontinuity appears at $x=1$ \cite{derrida,schutzdomany}.
491: In the maximum current phase ($\alpha,\beta> 1/2$), as we have already
492: mentioned, the bulk density is $1/2$ and boundary layers appear at
493: both ends.
494: On the other hand, the effect of symmetric lane change processes
495: is to diminish the difference between the local densities in the two lanes.
496: Since the densities are already equal without the interaction,
497: this situation is obviously not altered when switching on the vertical
498: hopping processes.
499: Consequently, the density profiles in the bulk are identical to that
500: of the TASEP in these phases.
501:
502: This is, however not the case at the coexistence line $\alpha=\beta
503: <1/2$. In the TASEP, a sharp domain wall emerges here in the density
504: profile, which separates a low- and a high-density phase with
505: constant densities far from the domain wall $\alpha$ and $1-\alpha$,
506: respectively.
507: The stochastic motion of the domain wall is
508: described by an unbiased random walk with reflective boundaries
509: \cite{schutzdomany,schutz98}, such that the average stationary
510: density profile connects linearly the boundary
511: densities $\alpha$ and $1-\alpha$.
512: Returning to our model, we consider first the closed system,
513: i.e. $\alpha=\beta=0$. The
514: profiles which fulfill the requirement about the continuity of the
515: currents are depicted in Fig. \ref{k1fig1} for various global
516: particle densities $\varrho\equiv\lim_{L\to\infty}\frac{N}{2L}$, where
517: $N$ is the number of particles in the system.
518: Here, the density profiles consist of three segments in general. In
519: the middle part of the system an equal-density segment is found
520: (Fig. \ref{k1fig1}a,b,d). This region is connected with the boundaries
521: by complementary-density segments on its left-hand side and on its
522: right-hand side, which
523: are continuous at $x=0$ and $x=1$, respectively.
524: In both lanes, the density profile is continuous
525: at one end of the equal-density segment and a shock is located
526: at the other one, such that the two shocks are at opposite ends.
527: The density in the equal-density region (and at the same time the
528: location of the shocks) depend on the global particle density.
529: At $\varrho =1/2$, the equal-density segment is lacking if $\Omega<1$
530: (Fig. \ref{k1fig1}c) and the profiles are linear if $\Omega =1$.
531: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
532: \begin{figure}[h]
533: \includegraphics[width=1.0\linewidth]{fig2.ps}
534: \caption{\label{k1fig1} Density profiles of the closed system
535: ($\alpha=\beta=0$) with $\Omega=0.5$ for different
536: global densities: (a) $\varrho=0.26$, (b) $\varrho=0.44$, (c)
537: $\varrho=0.5$ and (d) $\varrho=0.74$. The thin solid and dashed
538: lines represent the flow field of the differential equation
539: (\ref{diffgen}) corresponding to the complementary-density and
540: equal-density solutions, respectively. The thick solid and dashed lines are
541: the density profiles $\rho(x)$ and $\pi(x)$, respectively.}
542: \end{figure}
543: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
544:
545: If particles are allowed to enter and exit from the system at the
546: boundaries, i.e. $0<\alpha=\beta<1/2$, the total number of particles
547: is no longer conserved.
548: Nevertheless, we expect that
549: the stationary density profiles averaged over
550: configurations with a fixed global density $\varrho$, $\rho_{\varrho}(x)$ and $\pi_{\varrho}(x)$,
551: can still be constructed from the solutions (\ref{ek1}) and (\ref{ck1})
552: of the hydrodynamic equations. These profiles
553: are similar to those of the closed system and the only
554: difference is that the complementary-density segments fit to the
555: altered boundary conditions
556: $\rho_{\varrho}(0)=\pi_{\varrho}(1)=\alpha$ and $\rho_{\varrho}(1)=\pi_{\varrho}(0)=1-\alpha$.
557: This is indeed the case in the limit $\alpha\to 0$. Here,
558: the injections and removals of particles at the boundary sites, which
559: modify the global density, are infinitely rare, such that
560: the system has always enough time to relax, i.e. to adjust the density
561: profiles to the slightly altered global density.
562: As long as the shocks are not in the vicinity of the boundaries,
563: the densities at the boundary sites are independent from the
564: global density, which influences only the
565: position of the equal density segment.
566: Therefore, the stochastic variation of the global density $\varrho
567: (t)$ is described by a homogeneous, symmetric random walk in the
568: interval $[0,1]$ with reflective boundary conditions.
569: For finite $\alpha$, we can give only a heuristic argument why we
570: expect that the fluctuations of the global density are quasistationary
571: in the above sense.
572: In the stationary state, the center of the mass of a small instantaneous local
573: perturbation propagates with a
574: velocity $v(x)=1-2\rho(x)$ \cite{lighthill,schutz98}, which changes sign at
575: $\rho=1/2$.
576: In the complementary-density segments, the perturbations in the
577: density, which come from the fluctuations of the boundary reservoirs,
578: are thus driven toward the equal-density segment with a finite velocity.
579: The characteristic traveling time of the perturbation, as well as the
580: time scale related to the lane change processes in a finite system of size $L$
581: is $\mathcal{O}(L)$. The relaxation time
582: of the perturbation is
583: thus expected to be $\mathcal{O}(L)$. However, the random walk
584: dynamics of the global density implies that the time scale of a finite
585: change in the global density is $\mathcal{O}(L^2)$, which is large
586: compared to the relaxation time, thus the density profile
587: has enough time to follow the instantaneous global density.
588: The fluctuating global density $\varrho(t)$ is thus expected
589: to be a symmetric random walk with reflective boundaries at $\alpha$ and
590: $1-\alpha$.
591: In the stationary state, the global density is therefore homogeneously
592: distributed in the interval $[\alpha,1-\alpha]$.
593:
594: On the other hand, one can easily calculate that
595: if the position of the shock in lane $A$ is $x_s$,
596: the global density of particles in the system is
597: $\varrho(x_s)=(1-x_s)[\Delta(x_s)+\Omega]+\alpha$ for $x_s\ge 1/2$, where
598: we have introduced the (position dependent) height of the shock:
599: $\Delta(x_s)=2\Omega(x_s-1)+1-2\alpha$. Note that,
600: as opposed to the single lane
601: TASEP, this relation is no longer linear, therefore the
602: probability distribution of the position of the shock
603: is not uniform in the steady state.
604:
605: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
606: \begin{figure}[h]
607: \includegraphics[width=0.9\linewidth]{fig3.ps}
608: \caption{\label{k1num4}
609: Density profiles in lane $A$ at the coexistence
610: line at $\alpha=\beta=0.1$, obtained by
611: numerical simulation for different system sizes
612: and for different values of $\Omega$:
613: a) $\Omega=0.4$, b) $\Omega=0.8$, and c) $\Omega=1.2$.
614: In the case $\Omega=0.8$, the height of the shock at $x_s=1/2$ is zero.
615: The solid curves are the analytical predictions in the hydrodynamic limit.
616: The thick solid lines represent the profile
617: $\rho_{\varrho}(x)$ at global density $\varrho=1/2$.}
618: \end{figure}
619: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
620: With these prerequisites,
621: the steady-state density profile in lane A can be easily calculated
622: by averaging $\rho_{\varrho}(x)$ over the
623: steady-state distribution of the global density:
624: $\rho(x)=\frac{1}{1-2\alpha}\int_{\alpha}^{1-\alpha}\rho_{\varrho}(x)d\varrho$.
625: Skipping the details of the straightforward calculations,
626: we shall give the profile $\rho(x)$ in the interval
627: $\frac{1}{2}\le x \le 1$, whereas for $0 \le x\le \frac{1}{2}$, it
628: is obtained by the help of the relation
629: $\rho(x)=1-\rho(1-x)$, that follows from eq. (\ref{sym1}) and
630: (\ref{sym2}). The density profile in lane $B$ can be calculated by
631: making use of eq. (\ref{sym2}), which implies $\pi(x)=1-\rho(x)$.
632: The cases corresponding to the different signs of $\Delta(\frac{1}{2})$ must
633: be treated separately.
634: For $\Delta(\frac{1}{2})\ge 0$, we obtain
635: \be
636: \rho(x)=\frac{1}{2}+\frac{(x-\frac{1}{2})\Delta^2(x)}{1-2\alpha} \qquad
637: \Delta(\frac{1}{2}) \ge 0,
638: \label{positive}
639: \ee
640: which is a third-degree polynomial of $x$.
641: If $\Delta(\frac{1}{2})>0$, the second
642: derivative of $\rho(x)$ is discontinuous at $x=\frac{1}{2}$.
643: In the limit $\Omega\to 0$, the linear profile of the TASEP at the
644: coexistence line is recovered.
645: If $\Delta(\frac{1}{2})=0$, eq. (\ref{positive}) simplifies to
646: \be
647: \rho(x)=\frac{1}{2}+4\Omega (x-\frac{1}{2})^3 \qquad \Delta(\frac{1}{2})=0,
648: \ee
649: which is everywhere analytic.
650: For $\Delta(\frac{1}{2})<0$, the profile is constant in the
651: interval $\frac{1}{2} \le x
652: \le|\Delta(\frac{1}{2})|/(2\Omega)+\frac{1}{2}$,
653: where $\rho(x)=1/2$, while it is given by
654: eq. (\ref{positive}) in the interval
655: $|\Delta(\frac{1}{2})|/(2\Omega)+\frac{1}{2}\le x\le 1$.
656: These curves, as well as results of Monte Carlo simulations for finite
657: systems of size $L=64,128$ and $256$ are shown in Fig \ref{k1num4}.
658: In the numerical simulations, after waiting a period of $10^6$ Monte Carlo
659: steps in order to reach the steady state, we have measured the local
660: occupancies every $10$ Monte Carlo steps during a period of
661: $5\cdot 10^9$ steps.
662: For increasing $L$, the properly scaled profiles seem
663: to tend to the analytical curves expected to be valid
664: in the continuum limit.
665:
666: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
667: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
668: \section{Asymmetric lane change}
669: \label{alc}
670: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
671: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
672:
673: In this section, the stationary properties of the model are
674: investigated in the case $K\neq 1$.
675: Due to the symmetries of the system, we may restrict ourselves to the
676: investigation of the part $K<1$,
677: $\beta\le \alpha$ of the parameter space, which is related to the remaining part through eqs. (\ref{sym1}) and (\ref{sym2}).
678: Analyzing the solutions to the hydrodynamic equation (\ref{diffgen}),
679: one can construct the phase diagram in the four-dimensional parameter
680: space spanned by $\alpha,\beta,\Omega_A$ and $\Omega_B$.
681: Two representative two-dimensional cross sections
682: of the parameter space
683: at fixed lane change rates are shown in
684: Fig. \ref{pd5} and \ref{pd2}. As can be seen, the phase boundaries are
685: symmetric to the diagonal, which is a consequence of eq. (\ref{sym2}).
686: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
687: \begin{figure}[h]
688: \includegraphics[width=1\linewidth]{fig4.ps}
689: \caption{\label{pd5} Phase diagram at $\Omega_A=2$,
690: $\Omega_B=0.4$. Phase boundaries are indicated by solid lines.
691: Letters L,H and S refer to low-density,
692: high-density and localized shock phase, respectively; the first(second)
693: letter refers to lane $A$($B$).
694: The thick solid line indicates the coexistence line.
695: At the dashed lines, the function $J(\alpha,\beta)$ is nonanalytic.}
696: \end{figure}
697: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
698: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
699: \begin{figure}[h]
700: \includegraphics[width=1\linewidth]{fig5.ps}
701: \caption{\label{pd2} Phase diagram at $\Omega_A=2$,
702: $\Omega_B=1$. The dotted curves are the discontinuity lines, which
703: terminate at the points indicated by the full circles.}
704: \end{figure}
705: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
706: On the other hand, the phase diagrams are richer compared
707: to that of the symmetric model:
708: Besides the phases where the profiles are continuous in the interior
709: of the system, the asymmetry in the lane change kinetics leads
710: to the appearance of phases where one of the lanes contains a
711: localized shock in the bulk. This is reminiscent of the
712: shock phase of the single lane TASEP with Langmuir
713: kinetics. As a new feature, the position of the shock may vary
714: discontinuously with the boundary rates here, when the so-called
715: discontinuity line is crossed. The coexistence line, where
716: coherently moving delocalized shocks emerge in both lanes, is still present,
717: however, it is shorter than in the symmetric case and the shocks
718: walk only a shrunken domain.
719: The subsequent part of the section is devoted to the detailed analysis of
720: these findings.
721:
722: \subsection{Density profiles}
723:
724: We start the presentation of the results with the description of the
725: density profiles in the phases below the diagonal $\alpha=\beta$ of
726: the two-dimensional phase diagrams.
727:
728: If the exit rate is small enough, the densities in the bulk
729: exceed the value $1/2$ in both lanes (Fig. \ref{4prof}a).
730: Both profiles are continuous in the bulk and at the
731: exits, i.e. $\lim_{x\to 1}\rho(x)=\lim_{x\to 0}\pi(x)=1-b$,
732: but they are discontinuous at the
733: entrances, i.e. $\lim_{x\to 0}\rho(x)\neq a$ and $\lim_{x\to
734: 1}\pi(x)\neq a$, which signals the appearance of boundary
735: layers on the microscopic scale.
736: The profiles $\rho(x)$ and $\pi(x)$, as well as the current, depend
737: exclusively on $\beta$, while $\alpha$
738: influences only the boundary layers at the entrances. This
739: situation is observed also in the high-density phase of the TASEP with
740: Langmuir kinetics \cite{frey,ejs},
741: therefore we call this phase H-H phase, referring to
742: the high density in both lanes.
743: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
744: \begin{figure}[h]
745: \includegraphics[width=1\linewidth]{fig6.ps}
746: \caption{\label{4prof} Density profiles in lane $A$ (lower line) and lane
747: $B$ (upper line) for the parameters $\Omega_A=2$, $\Omega_B=0.4$,
748: $\alpha=0.45$ and for various exit rates: a) $\beta=0.15$, b) $\beta=0.25$,
749: c) $\beta=0.35$ and d) $\beta=0.45$. The profiles in panels a)-c)
750: were obtained by numerically solving eq. (\ref{diffgen}), whereas those
751: in panel d) are the analytical curves in eq. (\ref{comp}). Results of
752: numerical simulations (dotted lines) obtained for system size
753: $L=10000$ by averaging the occupancies in a period of $10^7$ Monte Carlo steps
754: in the steady state are
755: hardly distinguishable from these curves.}
756: \end{figure}
757: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
758:
759: In the phase denoted by S-H in the phase diagram,
760: $\rho(x)$ and $\pi(x)$ are continuous at the exits, similarly to the
761: H-H phase, however, the discontinuity in $\rho(x)$ is no longer
762: at $x=0$ but it is shifted to the interior of the system (Fig. \ref{4prof}b).
763: Thus $\lim_{x\to 0}\rho(x)=a$ holds and a shock is located in
764: lane $A$ at some $x_s$ ($0<x_s<1)$.
765: Therefore this phase is termed S-H phase, where the letter S refers to the shock
766: in lane $A$ and letter H refers to the high density ($\pi(x)>1/2$)
767: in lane $B$. The function $\pi(x)$
768: is discontinuous at $x=1$, i.e. $\lim_{x\to 1}\pi(x)\neq a$
769: and it is not differentiable (although continuous) at $x_s$.
770: Since both $\rho(x)$ and $\pi(x)$ are continuous at $x=0$, the
771: total current is given by
772: \be
773: J=a(1-a)-b(1-b)
774: \label{Jexact}
775: \ee
776: in this phase.
777: The profile $\rho(x)$ at fixed $\alpha$ and $\beta$
778: can be computed by substituting the current calculated from
779: eq. (\ref{Jexact}) into the differential equation (\ref{diffgen}).
780: Then, the solutions propagating from the left-hand and the right-hand
781: boundary, i.e. the solutions $\rho_l(x)$ and $\rho_r(x)$
782: fulfilling the boundary conditions $\rho_l(0)=a$ and
783: $\rho_r(1)=1-b$, respectively, are calculated numerically.
784: Finally, the position of the shock $x_s$ is obtained from
785: the relation $\rho_l(x_s)=1-\rho_r(x_s)$, which is
786: implied by the continuity of the current in lane $A$.
787: Once $\rho(x)$ is at our disposal, $\pi(x)$ can be calculated from eq. (\ref{J}).
788:
789:
790: Apart from the discontinuity line to be discussed in the next section,
791: the position of the shock $x_s$ varies continuously with the boundary rates
792: in the S-H phase.
793: Fixing $\alpha$ and reducing $\beta$, $x_s$ is
794: decreasing and at a certain value of $\beta$, $\beta=\beta_H(\alpha)$,
795: the shock reaches the left-hand boundary at $x=0$.
796: At this point, the right-hand solution $\rho_r(x)$ extends entirely to
797: the left-hand boundary and a further increase in $\beta$ drives the system to
798: the H-H phase.
799: The phase boundary $\beta_H(\alpha)$ between the S-H and the H-H
800: phase is thus determined from the condition $x_s=0$ or, equivalently,
801: $\rho_r(1)=1-a$.
802: When $\beta$ is increased along a vertical path in the phase
803: diagram at a fixed $\alpha$, $x_s$ increases and
804: for $\alpha<\rho_1$, where the constant $\rho_1$ will be determined
805: later, the path hits the coexistence line before the shock would reach
806: the right-hand boundary (see Sec. III. D).
807: Increasing $\beta$ along a path at some $\alpha>\rho_1$,
808: the shock reaches the right-hand boundary at $x=1$ for a certain value
809: of $\beta$, $\beta =\beta_L(\alpha)$ and the path leaves the S-H phase.
810: At the phase boundary,
811: the left-hand solution $\rho_l(x)$ extends to the whole system
812: and $\rho_l(1)=b$ must hold.
813:
814: Crossing the phase boundary $\beta_L(\alpha)$, the L-H phase is
815: entered, where letter L refers to the low density in lane $A$ since
816: here, $\rho(x)<1/2$ and $\pi(x)>1/2$ hold in the bulk (Fig. \ref{4prof}c).
817: In this phase, $\rho(x)$ and $\pi(x)$ are discontinuous at $x=1$, whereas
818: they are continuous at $x=0$ hence the
819: current is given by eq. (\ref{Jexact}).
820: In the part of the L-H phase where the current is zero, i.e. if
821: $\alpha=\beta$ or $\alpha,\beta\ge 1/2$, the profiles
822: can be calculated analytically.
823: The equal-density solutions of eq. (\ref{diffgen})
824: are
825: \beqn
826: \ln [\rho_e(x)(1-\rho_e(x))]&=&\Omega_{A}(K-1)x+{\rm const},
827: \nonumber \\
828: \pi_e(x)&=&\rho_e(x),
829: \label{equal}
830: \eeqn
831: whereas the complementary-density solutions read as
832: \beqn
833: \frac{\rho_1}{\rho_2}\ln |\rho_c(x)-\rho_1|-\frac{\rho_2}{\rho_1}\ln
834: |\rho_c(x)-\rho_2|&=&2\Omega_A\sqrt{K}x+{\rm const}, \nonumber \\
835: \pi_c(x)&=&1-\rho_c(x),
836: \label{comp}
837: \eeqn
838: where the constants $\rho_1\equiv 1/(1+K^{-1/2})$ and
839: $\rho_2\equiv 1/(1-K^{-1/2})$
840: are the roots of the equation $S_A(\rho,1-\rho)=0$.
841: There is, furthermore, a special complementary-density solution with
842: constant densities:
843: \beqn
844: \rho_c(x)&=&\rho_1, \nonumber \\
845: \pi_c(x)&=&1-\rho_1.
846: \label{compsp}
847: \eeqn
848: In the part of the L-H phase where $J=0$, the profiles are given by
849: the complementary-density solution which fulfills
850: the boundary conditions $\rho_c(0)=a$ and $\pi_c(0)=1-a$ (Fig. \ref{4prof}d).
851:
852: \subsection{Phase boundaries and the discontinuity line}
853:
854: In the S-H phase (and the L-H phase), the profiles
855: $\rho(x)$ and $\pi(x)$, as well as the
856: current are independent of $\alpha$ if $\alpha\ge 1/2$. Here, $\lim_{x\to 0}\rho(x)=1/2$
857: and $\alpha$ influences only the microscopic boundary layer, as we
858: argued in Sec. II. As a
859: consequence, the phase boundaries $\beta_H(\alpha)$ and
860: $\beta_L(\alpha)$ are horizontal lines in the domain
861: $\alpha\ge 1/2$ (see Fig. \ref{pd5} and \ref{pd2}) and
862: we may restrict the investigation of the phase boundaries
863: to the domain $\alpha\le 1/2$.
864:
865: Although we cannot give an analytical expression for the density profiles in
866: general, some
867: information can be gained on the phase boundaries of the S-H phase
868: by investigating the constant solutions of the hydrodynamic equations.
869: A constant solution $\rho(x)=r$, $\pi(x)=p$ must obey
870: $S_A(r,p)=0$, otherwise the spatial derivatives
871: $\partial_x\rho(x)$ and $\partial_x\pi(x)$
872: would not
873: vanish in eq. (\ref{diff1}). On the other hand, the constants satisfy
874: the equation $J=r(1-r)-p(1-p)$, where $J$ is
875: determined by the boundary rates via eq. (\ref{Jexact}).
876: Eliminating $p$ yields that $r$
877: is given by the implicit equation:
878: \be
879: r(r-1)\left[1-\frac{K}{(K+(1-K)r)^{2}}\right]=J(\alpha,\beta).
880: \label{Jrho}
881: \ee
882: In the S-H phase, this equation has two roots $r_1(J(\alpha,\beta),K)$ and
883: $r_2(J(\alpha,\beta),K)$ (shortly $r_1$ and $r_2$)
884: in the interval $[0,1]$. One can check that for the
885: larger root $r_2$, the relation $1/2<1-\beta<r_2$ holds, whereas
886: the smaller one, $r_1$, may be larger or smaller than $1/2$.
887:
888: First, we examine the phase boundary separating the L-H phase and the
889: S-H phase. One can check that at this boundary line, $r_1<1/2$
890: holds. Moreover, it follows from eq. (\ref{diff1}) that
891: $\frac{d\rho(x)}{dx}> 0$ if $0\le\rho(x)<r_1$, and
892: $\frac{d\rho(x)}{dx}< 0$ if $r_1<\rho(x)<1/2$.
893: Thus, the line $\rho(x)=r_1$ behaves as an attractor for the
894: solutions $\rho_l(x)$ propagating from the left-hand boundary
895: $x=0$ if $\rho_l(0)=\alpha\le 1/2$,
896: meaning that $\rho_l(x)$ approaches monotonously to $r_1$ as $x$ increases and $\lim_{x\to\infty}\rho_l(x)=r_1$.
897: Since $\rho_l(x)$ is monotonous and $\rho_l(0)=\alpha$, as well as $\rho_l(1)=\beta$ hold at the phase boundary, at the common point of
898: the boundary line and the diagonal $\alpha=\beta$, $\rho_l(x)$ must
899: be a constant function $\rho_l(x)=\alpha$. This, however, implies that $\alpha$
900: must coincide with $r_1$. The endpoint of the boundary line
901: $\beta_L(\alpha)$ is therefore at
902: $\alpha=\beta=r_1(J=0,K)=1/(1+K^{-1/2})$, which depends only on $K$.
903:
904: As opposed to this point, the whole function $\beta_L(\alpha)$ depends both on
905: $\Omega_A$ and $\Omega_B$.
906: Nevertheless, we can find an analytical expression for $\beta_L(\alpha)$
907: in the limit $K={\rm const}$, $\Omega_A\to\infty$.
908: We can see from eq. (\ref{diffgen}) that the
909: derivative $\frac{d\rho(x)}{dx}$ is proportional to $\Omega_A$ for
910: a fixed $K$. As a consequence, the larger $\Omega_A$ is the more rapidly
911: $\rho_l(x)$ tends to $r_1$. Thus, in the limit specified above,
912: $\lim_{\Omega_A\to\infty}(\rho_l(1)-r_1)=0$, which leads to $\beta=r_1$.
913: Substituting this into eq. (\ref{Jrho}), we
914: obtain for the inverse of the boundary curve $\beta_{L\infty}(\alpha)$
915: in the limit $K={\rm const}$, $\Omega_A\to\infty$:
916: \beqn
917: [\beta_{L\infty}]^{-1}(\beta)= \qquad \qquad \qquad \qquad \qquad
918: \qquad \qquad \qquad& & \nonumber \\
919: \frac{1}{2}\left(1-\sqrt{1-4\beta(1-\beta)
920: \left[2-\frac{K}{\left[K+(1-K)\beta\right]^2}\right]}\right).& &
921: \label{pb2}
922: \eeqn
923: This curve is plotted in Fig. \ref{phase5}.
924: The phase boundaries obtained by integrating eq. (\ref{diffgen})
925: numerically for
926: finite lane change rates tend rapidly to this limiting curve.
927: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
928: \begin{figure}[h]
929: \includegraphics[width=0.9\linewidth]{fig7.ps}
930: \caption{\label{phase5} Phase boundaries of the S-H
931: phase obtained by integrating eq. (\ref{diffgen}) numerically for $K=1/5$ and for various $\Omega_A$.
932: The solid lines are the limiting
933: curves: $\beta_{L\infty}(\alpha)$, given by eq. (\ref{pb2}) and
934: $\beta_{H\infty}(\alpha)$, given solely by eq. (\ref{pb1}) since $K<K^*$.}
935: \end{figure}
936: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
937:
938: Next, we turn to examine the boundary curve between the S-H and
939: the H-H phase, $\beta_H(\alpha)$. Along this line,
940: $\rho_r(0)=1-\alpha$ and $\rho_r(1)=1-\beta$ hold,
941: and the roots of eq. (\ref{Jrho}) are
942: arranged as $r_1<1-\alpha<1-\beta<r_2$.
943: One can show that
944: the line $\rho(x)=r_2$ is an attractor for the
945: solutions $\rho_l(x)$ which start at $x=0$ if
946: $\rho_l(0)> \max\{1/2,r_1\}$. Moreover, if $r_1>1/2$, the line
947: $\rho(x)=r_1$ repels the solutions $\rho_l(x)$ starting from $x=0$ if
948: $r_1<\rho_l(0)<r_2$, or, in other words, the solutions $\rho_r(x)$
949: propagating from the right-hand boundary, for which $r_1<\rho_r(1)<r_2$,
950: are attracted by the line $\rho(x)=r_1$.
951: When the diagonal is approached along the phase
952: boundary $\beta_L(\alpha)$, the profile $\rho_r(x)$, being monotonous,
953: must tend to the constant function $\rho_r(x)=1-\alpha$, as well as
954: $\pi(x)$ since the current is zero at the diagonal.
955: However, equal densities in the two lanes are possible for $K\neq 1$
956: only if the density is one (or zero),
957: therefore the boundary line must approach the
958: diagonal at $\alpha=0$. This is in accordance with the fact that
959: $r_2=1$ if $J=0$.
960: Thus, we obtain $\lim_{\alpha\to 0}\beta_L(\alpha)=0$, independently
961: of the lane change rates.
962:
963:
964: Similarly to $\beta_L(\alpha)$, the location of
965: the whole boundary line $\beta_H(\alpha)$
966: depends on both lane change rates (see Fig. \ref{phase5} and \ref{phase2})
967: and we can give an analytical expression for $\beta_H(\alpha)$
968: again only in the limit $K={\rm const}$, $\Omega_A\to\infty$.
969: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
970: \begin{figure}[h]
971: \includegraphics[width=0.9\linewidth]{fig8.ps}
972: \caption{\label{phase2} Phase boundaries
973: between the S-H phase and the H-H phase obtained by numerical
974: integration of eq. (\ref{diffgen})
975: for $K=1/2>K^*$ and for several values of $\Omega_A$. The solid lines are
976: the limiting curves given in eq. (\ref{pb1}) and eq. (\ref{pb3}).}
977: \end{figure}
978: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
979: As aforesaid, the profile given by $\rho_r(x)$ lies between two
980: attractors, $\rho(x)=r_1$ and $\rho(x)=r_2$, to which
981: the solutions tend in the limit $x\to-\infty$ and $x\to\infty$, respectively.
982: When $\Omega_A$ is increased (such that $K$ is fixed), then
983: $\beta_L(\alpha)$ at a fixed $\alpha$ decreases.
984: Thus, the current is increasing and the two roots of
985: eq. (\ref{Jrho}), $r_1$ and $r_2$ are coming closer.
986: Keeping in mind that $r_1<1-\alpha<1-\beta<r_2$,
987: there are now two possible cases. Depending on the value of $\alpha$,
988: either the gap between $1-\beta$ and $r_2$ or the gap between
989: $r_1$ and $1-\alpha$ vanishes first.
990: In other words, in the former case, the profile is attracted to the
991: line $\rho(x)=r_2$ at $x=1$ in the limit $\Omega_A\to\infty$,
992: i.e. $\lim_{\Omega_A\to\infty}(\rho_r(1)-r_2)=0$, whereas in the
993: latter case it is attracted to the line $\rho(x)=r_1$ at $x=0$
994: (provided that $r_1>1/2$) , i.e.
995: $\lim_{\Omega_A\to\infty}(\rho_r(0)-r_1)=0$.
996: In the first case, substituting $r=1-\beta$ into eq. (\ref{Jrho}),
997: we obtain for the inverse of the limiting curve $\beta_{H\infty}^b(\alpha)$
998: in terms of $\beta_{L\infty}(\alpha)$
999: \be
1000: [\beta_{H\infty}^b]^{-1}(\beta)=[\beta_{L\infty}]^{-1}(1-\beta),
1001: \label{pb1}
1002: \ee
1003: whereas in the second case, $r=1-\alpha$ yields
1004: \beqn
1005: \beta_{H\infty}^{a}(\alpha)= \qquad \qquad \qquad \qquad \qquad \qquad \qquad \nonumber \\
1006: \frac{1}{2}\left(1-\sqrt{1-\frac{4\alpha(1-\alpha)K}
1007: {\left[K+(1-K)(1-\alpha)\right]^2}}\right).
1008: \label{pb3}
1009: \eeqn
1010: These curves are plotted in Fig. \ref{phase2}.
1011: The phase boundary in the limit $K={\rm const}$, $\Omega_A\to\infty$ is
1012: given by
1013: \be
1014: \beta_{H\infty}(\alpha)=\max\{\beta_{H\infty}^a(\alpha),\beta_{H\infty}^b(\alpha)\}.
1015: \label{env}
1016: \ee
1017: The value $\alpha^*$ at which the functions $\beta_H^a(\alpha)$ and
1018: $\beta_H^b(\alpha)$ intersect varies with $K$. If $K\to 1$,
1019: $\alpha^*$ tends to $\frac{\sqrt{3}-1}{2\sqrt{3}}=0.21132\dots$,
1020: while $\alpha^*=1/2$ if $K$ is equal to
1021: \be
1022: K^*=1+\sqrt{2}-\sqrt{2(1+\sqrt{2})}=0.21684\dots
1023: \label{kstar}
1024: \ee
1025: Thus, for $K\le K^*$, the limiting curve of $\beta_H(\alpha)$ is given
1026: by eq. (\ref{pb1}) alone,
1027: otherwise it is composed of eq. (\ref{pb1}) and eq. (\ref{pb3}) as
1028: given by eq. (\ref{env}).
1029:
1030: Although the curve $\beta_{H\infty}(\alpha)$ gives the phase boundary line
1031: only in the limit $K={\rm const}$, $\Omega_A\to\infty$,
1032: we show that for
1033: $K>K^*$ and for large enough $\Omega_A$, $\Omega_A\ge \Omega_A^*(K)$,
1034: the phase transition point at $\alpha=1/2$ is given exactly by
1035: eq. (\ref{pb3}), i.e. $\beta_H(1/2)=\beta^a_{H\infty}(1/2)=\frac{K}{1+K}$.
1036:
1037: In order to see this, we discuss first the possible
1038: appearance of a discontinuity line in the S-H phase, at which
1039: the position of the shock $x_s$ changes discontinuously.
1040: If $r_1=1/2$, one can see from eq. (\ref{diff1})
1041: that for the spatial derivative of the profile,
1042: $\lim_{\rho\to 1/2}\frac{d\rho(x)}{dx}>0$ holds and the left-hand and
1043: right-hand solutions $\rho_l(x)$ and $\rho_r(x)$ may propagate as far as the line
1044: $\rho(x)=1/2$. If $\rho_l(x_1)=1/2$ and $\rho_r(x_2)=1/2$ hold for some
1045: $x_1$ and $x_2$, such that $0<x_1<x_2<1$, then the left-hand and right-hand solutions are connected by a
1046: constant segment $\rho(x)=1/2$ in the interval $[x_1,x_2]$ and the
1047: profile is continuous (Fig. \ref{crit}c).
1048: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1049: \begin{figure}[h]
1050: \includegraphics[width=1.0\linewidth]{fig9.ps}
1051: \caption{\label{crit} Density profiles for parameters $\Omega_A=2$,
1052: $\Omega_B=1$ and a) $\alpha=0.4$, $\beta=0.29$, where $r_1$ is slightly
1053: above $1/2$; b) $\alpha=0.4$, $\beta=0.32$, where $r_1$ is slightly
1054: below $1/2$; c) $\alpha=0.4$, $\beta\approx 0.305$, where
1055: $r_1=1/2$.
1056: d) The endpoint of the discontinuity line at $\alpha\approx 0.297$, $\beta \approx 0.237$.}
1057: \end{figure}
1058: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1059: Substituting $r=1/2$ into eq. (\ref{Jrho}) we obtain the equation
1060: of the discontinuity line:
1061: $\alpha(1-\alpha)-\beta(1-\beta)=(\frac{1-K}{2(1+K)})^2$,
1062: along which the current is constant. Solving this equation for
1063: $\beta$, we obtain
1064: \beqn
1065: \beta_{d}(\alpha)= \qquad \qquad \qquad \qquad \qquad \qquad \qquad \nonumber \\
1066: \frac{1}{2}\left(1-\sqrt{1-4
1067: \alpha(1-\alpha)+4\left(\frac{1-K}{2(1+K)}\right)^2}\right).
1068: \label{disc}
1069: \eeqn
1070:
1071: This curve is shown in Fig. \ref{pd2}.
1072: When at an arbitrary point of this line, $\beta$ is infinitesimally
1073: decreased, then $r_1$ exceeds the value $1/2$ and a shock
1074: appears at $x_1$ with an infinitesimal height (Fig. \ref{crit}a).
1075: Conversely, an
1076: infinitesimal increase in $\beta$ decreases $r_1$ below $1/2$
1077: and an infinitesimal shock appears at $x_2$ (Fig. \ref{crit}b).
1078: Thus, when this line is crossed, the position of the shock jumps from
1079: $x_1$ to $x_2$.
1080: At the point of the discontinuity line at $\alpha=1/2$, $x_1=0$ holds and
1081: an infinitesimal increase (decrease) in $\beta$ drives the system
1082: to the H-H (S-H) phase. Therefore the point of this curve
1083: at $\alpha=1/2$
1084: coincides with the phase boundary, i.e. $\beta_d(1/2)=\beta_H(1/2)$.
1085: On the other hand, comparing eq. (\ref{pb3}) and eq. (\ref{disc}), we
1086: see that $\beta_d(1/2)=\beta_{H\infty}^a(1/2)=\frac{K}{1+K}$.
1087: The function $\beta_{H\infty}^a(\alpha)$ thus gives the phase transition point
1088: at $\alpha=1/2$ exactly, provided the curve $\beta_d(\alpha)$
1089: is located in the S-H phase.
1090: That means if $K>K^*$ and $\Omega_A\ge \Omega_A^*(K)$, where
1091: $\Omega_A^*(K)$ is the value of $\Omega_A$ at which $\beta_H(1/2)$ first
1092: reaches the value $\frac{K}{1+K}$, when $\Omega_A$ is
1093: increased from zero. For example, for $K=1/2$, we have found $\Omega_A^*(1/2)\approx
1094: 0.719$ (Fig. \ref{phase2}).
1095: We emphasize, however, that the discontinuity line is lacking
1096: if $K<K^*$ or $K>K^*$ but $\Omega_A<\Omega_A^*(K)$.
1097:
1098:
1099: If $K>K^*$ and $\Omega_A> \Omega_A^*(K)$, then, at the transition point at
1100: $\alpha=1/2$, we have $x_1=0$; $\Omega_A$ influences only $x_2$
1101: and $x_2\to0$ if $\Omega_A\to\Omega_A^*(K)$.
1102: Moving away from the point at $\alpha=1/2$ along the discontinuity line,
1103: the length of the constant segment
1104: $x_2-x_1$ is decreasing and vanishes at a certain point.
1105: Here, the density profile becomes analytical and the
1106: discontinuity line terminates (see Fig. \ref{crit}d and Fig. \ref{pd2}).
1107: The position of this endpoint depends on $\Omega_A$ and it is moving
1108: toward smaller values of $\alpha$ for increasing $\Omega_A$;
1109: in the large $\Omega_A$ limit, it tends to the
1110: point of intersection of $\beta_d(\alpha)$ and $\beta_{H\infty}^b$, whereas
1111: in the limit $\Omega_A\to\Omega_A^*(K)$, the abscissa of the end point
1112: tends to $1/2$.
1113:
1114: Finally, we mention that another special curve in the S-H phase
1115: is defined by the equation $S_A(\alpha,1-\beta)=0$.
1116: At this curve the left-hand solution $\rho_l(x)$ is constant,
1117: therefore both $\rho(x)$ and $\pi(x)$ are constant in the interval $[0,x_s]$.
1118:
1119:
1120: \subsection{Current}
1121:
1122: We have seen that, apart from the H-H (and the L-L phase), the current
1123: is given by eq. (\ref{Jexact}). It is zero at the line $\alpha=\beta$ and
1124: in the domain $\alpha,\beta\ge 1/2$ and it is
1125: independent of $\alpha$ ($\beta$) in the H-H (L-L) phase.
1126: It is a continuous function of the boundary rates everywhere, although
1127: nonanalytic at the boundaries of the H-H phase and the L-L phase,
1128: as well as at the lines $\alpha=1/2$ and $\beta=1/2$
1129: outside the H-H phase and the L-L phase,
1130: where the effective boundary rates $a$ and $b$ saturate at $1/2$.
1131:
1132: In the following, we concentrate on the current in the H-H phase.
1133: Since it is continuous at the boundary line
1134: $\beta_H(\alpha)$, it can be expressed in the H-H phase
1135: in terms of $\beta_H(\alpha)$ as
1136: $J(\beta)=\beta_H^{-1}(\beta)(1-\beta_H^{-1}(\beta))-\beta(1-\beta)$.
1137: According to numerical results (see Fig. \ref{current}), the current $J(\beta)$ has a maximum in
1138: the H-H phase, and for large $\Omega_A$, the location of the maximum tends to $\beta=\beta_H(1/2)$
1139: for $K\le K^*$, while for $K>K^*$, it tends to $\beta^*$ where the curves
1140: $\beta_{H\infty}^a(\alpha)$ and $\beta_{H\infty}^b(\alpha)$ intersect.
1141: Contrary to this, the current is a monotonously decreasing function of $\beta$ in
1142: the S-H phase.
1143: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1144: \begin{figure}[h]
1145: \includegraphics[width=1.0\linewidth]{fig10.ps}
1146: \caption{\label{current} The current as a function of $\beta$ along
1147: the line $\alpha=1/2$ for $K=1/5$ (a) and $K=1/2$ (b). The solid
1148: line is the current in the limit $\Omega_A\to\infty$, the other
1149: curves from bottom to top correspond to $\Omega_A=0.25,0.5,1,2,4$, respectively.}
1150: \end{figure}
1151: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1152: We now turn to the question, at which parameter combination the total current
1153: is maximal in the steady state.
1154: For fixed $K$, the maximal current is realized in the limit
1155: $\Omega_A\to\infty$ at $\beta_{H\infty}(1/2)$ for $K<K^*$ and at $\beta^*$
1156: for $K>K^*$.
1157: Since the current is growing faster in the S-H phase for decreasing
1158: $\beta$ than in the H-H phase above $\beta^*$ (if $K>K^*$),
1159: we conclude that the parameter
1160: combination that maximizes the current is found in the domain $K<K^*$.
1161: For $K<K^*$, the maximal current is thus
1162: $J_{\rm max}(K)=1/4-\beta_{H\infty}(1/2)(1-\beta_{H\infty}(1/2))$.
1163: Since $\beta_{H\infty}(1/2)$ decreases monotonously with decreasing $K$, the
1164: current is maximal at $K=0$, where
1165: $\beta_{H\infty}(1/2)=\frac{2-\sqrt{2}}{4}$ and $J_{\rm max}(0)=1/8$.
1166: The value $1/8$ is thus an upper bound for
1167: the total current and $J\to 1/8$ in the limit $\Omega_A\to \infty$ if
1168: $\alpha\ge 1/2$, $\beta=\frac{2-\sqrt{2}}{4}$ and $\Omega_B=0$.
1169:
1170: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1171: \begin{figure}[h]
1172: \includegraphics[width=1.0\linewidth]{fig11.ps}
1173: \caption{\label{approx} The current as a function of $\beta$
1174: in the H-H phase obtained by numerical integration for $K=1/2$ and for
1175: different values of $\Omega_A$. The analytical approximation
1176: given in eq. (\ref{smallJ}) is indicated by solid lines.}
1177: \end{figure}
1178: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1179: We close this section with the discussion of the current in the H-H
1180: phase when the lane change rates are small, i.e. $\Omega_A,\Omega_B \ll 1$.
1181: If the vertical hopping processes are switched off, the current is zero, hence
1182: we expect that for small lane change rates the current is small, as well.
1183: Assuming that $J\ll (\frac{1}{2}-\beta)^2$ and expanding the right-hand side
1184: of eq. (\ref{diffgen}) in a Taylor series up to first order in $J$, then
1185: solving the resulting differential equation yields finally
1186: \beqn
1187: J=\frac{1-K}{1+K}\beta(1-\beta)(1-2\beta)\times \nonumber \\
1188: \left[\sqrt{1+2(1+K)\frac{\Omega_A}{1-2\beta}}-1\right]+
1189: \mathcal{O}(\Omega_A^3).
1190: \label{smallJ}
1191: \eeqn
1192: The details of the calculation are presented in Appendix A.
1193: This expression is compared to the current calculated by integrating
1194: eq. (\ref{diffgen}) numerically in
1195: Fig. \ref{approx}.
1196: Expanding this expression for small $\Omega_A$, we obtain
1197: \be
1198: J=\beta(1-\beta)(1-K)\Omega_A\left[1-\frac{1+K}{2}\frac{\Omega_A}{1-2\beta}\right]
1199: + \mathcal{O}(\Omega_A^3).
1200: \label{smallJ2}
1201: \ee
1202: The current is thus in leading order proportional to $\Omega_A$.
1203: Examining the higher-order terms in the series expansion of the
1204: right-hand side of eq. (\ref{diffgen}), one
1205: can show that for arbitrary $\Omega_A$, the current vanishes
1206: as $J\sim 1-K$ when $K\to 1$.
1207:
1208:
1209: \subsection{Coexistence line}
1210:
1211: Let us assume that a given point of the section $\alpha=\beta<\rho_1$
1212: is approached along a path in the S-H phase.
1213: In this case, the position of the shock in lane $A$, $x_s$, tends to some
1214: $x_{\rm min}=x_{\rm min}(\alpha,\Omega_A,\Omega_B)$, which is somewhere
1215: in the bulk,
1216: i.e. $0<x_{\rm min}<1$.
1217: When the same point is approached
1218: from the L-S phase, the position of the shock in lane $B$ tends
1219: to the same $x_{\rm min}$ according to eq. (\ref{sym2}).
1220: Thus, when the boundary line $\alpha=\beta<\rho_1$ is
1221: passed from the S-H phase, the shock in
1222: lane $A$ jumps from $x_{\rm min}$ to the right-hand boundary at $x=1$, where a discontinuity appears, while
1223: the discontinuity at $x=1$ in lane $B$, which
1224: can be regarded as a shock localized there, jumps to $x_{\rm min}$.
1225: So, the density profile changes discontinuously.
1226: Strictly at $\alpha=\beta$, the shocks in both lanes are delocalized
1227: and perform a stochastic motion in the domain $[x_{\rm min},1]$,
1228: similarly to the symmetric model with $K=1$ at the line $\alpha=\beta<1/2$.
1229:
1230: Now, this phenomenon is investigated in detail in the case of asymmetric
1231: lane change.
1232: Since the current is zero if $\alpha=\beta$, the solutions of
1233: the hydrodynamic equations are those given in eq. (\ref{equal}) and
1234: eq. (\ref{comp}) (see Fig. \ref{ff1}).
1235: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1236: \begin{figure}[h]
1237: \includegraphics[width=1.0\linewidth]{fig12.ps}
1238: \caption{\label{ff1} The flow field of the differential equation
1239: (\ref{diffgen}) for $\Omega_A=2$, $\Omega_B=0.5$ and $J=0$. The solid
1240: curves represent the complementary-density solutions given in
1241: eq. (\ref{comp}), whereas dashed curves represent the
1242: equal-density solutions given in eq. (\ref{equal}).}
1243: \end{figure}
1244: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1245: The argumentation about the quasistationarity of the fluctuations of the
1246: global density presented in the case $K=1$ apply to the case $K\neq
1247: 1$, as well.
1248: Thus, in the open system, the density profiles averaged over
1249: configurations with a fixed global density, $\rho_{\varrho}(x)$
1250: and $\pi_{\varrho}(x)$, can be constructed from the solutions
1251: (\ref{equal}) and (\ref{comp}).
1252: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1253: \begin{figure}[h]
1254: \includegraphics[width=1.0\linewidth]{fig13.ps}
1255: \caption{\label{ff2} Density profiles in lane $A$ (thick solid lines)
1256: and in lane $B$ (thick dashed lines) which
1257: belong to a fixed global
1258: particle density $\varrho$, at four different values of $\varrho$.
1259: The flow field of eq. (\ref{diffgen}) is indicated by thin lines.
1260: The rates are $\Omega_A=0.5$, $\Omega_B=0.25$ and $\alpha=\beta=0.1$.}
1261: \end{figure}
1262: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1263: The structure of these profiles is identical to those obtained in the
1264: case $K=1$ (see Fig. \ref{ff2}).
1265: Generally, they consist of three segments (see Fig. \ref{ff2}b):
1266: An equal-density segment is located in the middle part of the system
1267: in the domain $[x_0,\tilde x_0]$. This region is connected with the
1268: left-hand boundary by a complementary-density segment, which is continuous at
1269: $x=0$, i.e. $\rho_{c,l}(0)=\alpha$, $\pi_{c,l}(0)=1-\alpha$, and with
1270: the right-hand boundary by another complementary-density segment,
1271: which is continuous at $x=1$, i.e.
1272: $\rho_{c,r}(1)=1-\alpha$, $\pi_{c,r}(1)=\alpha$.
1273: Each lane contains a shock, which are at the opposite ends of the
1274: equal-density segment (one at $x_0$, the other one at $\tilde x_0$).
1275: The location of the equal-density segment, as well as $x_0$
1276: and $\tilde x_0$ are determined by the actual global density.
1277: If $x_0=\tilde x_0\equiv \overline x$, the equal-density segment is lacking and
1278: $\rho_{c,l}(x)$ and $\rho_{c,r}(x)$ are directly connected by a shock
1279: at $\overline x$ (Fig. \ref{ff2}c).
1280: Since $x_0$ and $\tilde x_0$ are not independent, the shocks move in a
1281: synchronized way, and their motion is confined to the range $[x_{\rm min},1]$.
1282: If one of the shocks is at $x=1$, the other one is at $x_{\rm min}$, thus,
1283: the lower bound $x_{\rm min}$ is determined by the
1284: equation $\rho_{c,l}(x_{\rm min})=1-\rho_{e}(x_{\rm min})$, where
1285: $\rho_{e}(x)$ is the equal-density solution which fulfills
1286: $\rho_{e}(1)=1-\alpha$ (see Fig. \ref{ff2}a).
1287: The lower bound $x_{\rm min}$ is an increasing
1288: function of $\alpha$ (see Fig \ref{ff3}).
1289: In the limit $\alpha\to 0$, $x_{\rm min}$ tends to zero and if
1290: $\alpha\to\rho_1$, $x_{\rm min}$ tends to 1, thus, at $\alpha=\rho_1$,
1291: the shock becomes localized at $x=1$ and the system enters the L-H phase.
1292: At this point, the density
1293: profile is given by the special complementary-density solution:
1294: $\rho_c(x)=\alpha$, $\pi_c(x)=1-\alpha$.
1295: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1296: \begin{figure}[h]
1297: \includegraphics[width=0.9\linewidth]{fig14.ps}
1298: \caption{\label{ff3} The shock in lane $A$ at the left most possible
1299: position for $\Omega_A=0.5$, $\Omega_B=0.25$ and for different rates $\alpha$($=\beta$).}
1300: \end{figure}
1301: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1302:
1303: Similarly to the case $K=1$, the stochastic variation of the global density
1304: $\varrho(t)$ is described by a bounded symmetric random
1305: walk.
1306: The stationary density profile $\rho(x)$ can be obtained from
1307: the profile $\rho_{\varrho}(x)$ at a fixed global
1308: density by averaging it over $\varrho$.
1309: For $K\neq 1$, we could not carry out the averaging analytically,
1310: nevertheless, we can gain some information on the stationary density
1311: profiles at $\overline x$ without the knowledge of the entire profiles.
1312: At the point $\overline x$, the relation
1313: $\rho_{\varrho}(\overline x)=\pi_{\varrho}(\overline x)$
1314: holds for all $\varrho$. This, together with the relation $\rho(x)=1-\pi(x)$
1315: following from eq. (\ref{sym2}) implies that
1316: $\rho(\overline x)=\pi(\overline x)=1/2$.
1317: As it is shown in Appendix B,
1318: the ratio of the first derivatives of the stationary density profile
1319: in lane $A$ on the
1320: two sides of the point $\overline x$ is
1321: \be
1322: \frac{d\rho(\overline x-)}{dx}/\frac{d\rho(\overline x+)}{dx}=
1323: \frac{K+(1-K)\rho_0}{1-(1-K)\rho_0},
1324: \label{rat}
1325: \ee
1326: where $\rho_0\equiv \rho_{c,\alpha}(\overline x)$.
1327: In the case $K<1$, $\rho_0<\rho_1<1/2$ always holds,
1328: hence this ratio is smaller
1329: than 1 and the first derivative of the density profile is
1330: discontinuous at $\overline x$.
1331: Furthermore, it is clear that the stationary density profile
1332: is identical to the complementary-density solution in the
1333: interval $[0,x_{\rm min}]$, since this domain is forbidden for the shocks.
1334: We have performed numerical simulations for finite systems
1335: of size $L=64,128$ and $256$
1336: and measured the density profiles in the same way as in the symmetric
1337: case at the coexistence line.
1338: Results are shown in Fig. \ref{coexnum}.
1339:
1340: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1341: \begin{figure}[h]
1342: \includegraphics[width=0.8\linewidth]{fig15.ps}
1343: \caption{\label{coexnum} Density profiles in lane $A$ obtained by numerical
1344: simulations with rates $\alpha=\beta=0.1$, $\Omega_A=0.5$,
1345: $\Omega_B=0.25$ and for different system sizes (thin lines). The
1346: thick line represents the density profile $\rho_{\varrho}(x)$
1347: at global density $\varrho=1/2$ in the continuum limit.}
1348: \end{figure}
1349: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1350:
1351:
1352: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1353: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1354: \section{Discussion}
1355: \label{discussion}
1356: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1357:
1358: In this work, a two-lane exclusion process was studied,
1359: where the particles are conserved in the bulk and
1360: each lane can be thought of as a totally asymmetric simple
1361: exclusion process with nonconserving kinetics in the bulk.
1362: As a consequence, the model
1363: unifies the features of the particle conserving and bulk
1364: nonconserving exclusion processes, as far as the dynamics of the
1365: shock is concerned. Namely, it exhibits
1366: phases with a localized shock in one lane, while the other one acts as a
1367: nonhomogeneous bulk reservoir, the position-dependent density
1368: of which is determined by the dynamics itself in a self-organized manner.
1369: On the other hand, the model undergoes a discontinuous phase
1370: transition at the coexistence line, where
1371: delocalized shocks form in both lanes and move in a synchronized way.
1372: Here, the global density of particles behaves as an
1373: unbiased random walk, similarly to the TASEP at the
1374: coexistence line, however, the density profiles in the coexisting
1375: phases are not constant here.
1376:
1377: Although we considered throughout this work exchange rates
1378: proportional to $1/L$, one may imagine other types of scaling.
1379: For the TASEP with Langmuir kinetics,
1380: shock localization is observed at the coexistence line when
1381: the creation and annihilation rates vanish proportionally to $1/L^a$
1382: with $1\le a<2$ \cite{js}. It might be worth examining
1383: whether the synchronization of
1384: shocks in the present model persists for lane change rates
1385: vanishing faster than $1/L$.
1386:
1387: In the limit of large lane change rates, $K={\rm const}$,
1388: $\Omega_A\to\infty$, the density profiles and
1389: the boundaries of the phases exhibiting
1390: a localized shock are related to the zeros of the source term in
1391: the hydrodynamic equation, which is generally valid for systems with
1392: weak bulk nonconserving kinetics \cite{frey,pierobon,reichenbach}.
1393: However, different behavior is observed in general when the lane change rates
1394: $\omega_A$ and $\omega_B$ are finite in the limit $L\to\infty$.
1395: In this case, the particle current from one lane
1396: to the other may be finite at the boundary layers.
1397: These currents add to the incoming and/or outgoing currents at the boundaries,
1398: which may lead to nontrivial bulk densities even in the case $K=1$,
1399: where the profiles are constant.
1400:
1401: Possible extensions of the present model are obtained when
1402: different exit and entrance rates or different hop rates
1403: in the two lanes are taken into account. Nevertheless, these generalized
1404: versions are more difficult
1405: to treat because of the reduced symmetry compared to that of the
1406: present model.
1407:
1408: \appendix
1409: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1410: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1411: \section{}
1412: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1413: We derive here an approximative expression for the current
1414: in the H-H phase in the limit of small lane change rates.
1415: Assuming that $J\ll (\frac{1}{2}-\beta)^2$,
1416: we may expand the right-hand side of
1417: the differential equation (\ref{diffgen}) in a Taylor series up to
1418: first order in $J$.
1419: Integrating the differential equation obtained in this way yields
1420: \beqn
1421: F(\rho)\equiv \frac{1}{K-1}\ln(\rho-\rho^2)+
1422: J\left\{\frac{K}{(1-K)^2}\left[\ln\frac{\rho}{1-\rho}-\frac{1}{\rho}\right]+\right.
1423: \nonumber \\
1424: \left.\frac{1}{(1-K)^2}\left[\ln\frac{\rho}{1-\rho}+\frac{1}{1-\rho}\right]
1425: \right\}=\Omega_Ax+{\rm const}, \qquad
1426: \label{first}
1427: \eeqn
1428: where the first term on the left-hand side is just the equal-density
1429: solution for $J=0$.
1430: The solution which obeys the boundary condition $\rho(1)=1-\beta$
1431: is $F(\rho)-F(1-\beta)=\Omega_A(x-1)$.
1432: From this equation, we get for the density at the left-hand boundary,
1433: $\rho_0\equiv\lim_{x\to 0}\rho(x)$, the implicit equation:
1434: \be
1435: F(\rho_0)-F(1-\beta)=-\Omega_A.
1436: \label{e1}
1437: \ee
1438: On the other hand, we have another relation between $J$ and $\rho_0$:
1439: \be
1440: J=\rho_0(1-\rho_0)-\beta(1-\beta),
1441: \label{e2}
1442: \ee
1443: thus, we have closed equations for current.
1444: Note that the leading order term on the left-hand side of
1445: eq. (\ref{e1}), which comes
1446: from the difference of the leading terms of $F(\rho)$ evaluated at
1447: $\rho_0$ and $1-\beta$, is $\mathcal{O}(J)$, while the next-to-leading
1448: contribution in eq. (\ref{e1}) is $\mathcal{O}(J^2)$.
1449: Expressing $\rho_0$ from eq. (\ref{e2}) and expanding it for small
1450: $J$, we get
1451: $\rho_0=1-\beta-\frac{J}{1-2\beta}-\frac{J^2}{(1-2\beta)^3}+\mathcal{O}(J^3)$.
1452: Substituting this expression into eq. (\ref{e1}) gives an implicit
1453: equation for $J$. Assuming that $J\ll \beta$ and expanding the terms
1454: containing $J$ in this equation in Taylor series, we obtain finally
1455: \be
1456: \Theta+\frac{1+K}{2}\Theta^2+\mathcal{O}(\Theta^3)=\frac{\Omega_A}{1-2\beta},
1457: \label{quadr}
1458: \ee
1459: where $\Theta\equiv\frac{J}{(1-2\beta)\beta(1-\beta)(1-K)}$.
1460: For small $\Theta$, which amounts to $\Omega_A\ll 1-2\beta$, we get a
1461: good approximation for the current by solving this quadratic equation
1462: and arrive at eq. (\ref{smallJ}).
1463:
1464: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1465: \section{}
1466: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1467:
1468:
1469: As we have seen, the positions of the shocks in lane $A$ and lane $B$ are not
1470: independent, thus, we may define thereby a function $\tilde x_0(x_0)$,
1471: which is given implicitly by the equation
1472: $\rho_{e}(\tilde x_0)=1-\rho_{c,r}(\tilde
1473: x_0)$, where $\rho_e(x)$ is the equal-density solution which satisfies
1474: the condition $\rho_{c,l}(x_0)=\rho_e(x_0)$.
1475: In the following, we shall denote the density in lane $A$ at $x_0$, if
1476: the shock is located at $x_s$, by $\rho_{x_s}(x_0)$.
1477: First, we notice that at the reference point $x_0$
1478: ($x_0<\overline x$), $\rho_{x_s}(x_0)$=$\rho_{c,l}(x_0)$ holds
1479: whenever the shock in lane $A$
1480: resides between $x_0$ and $\tilde x_0(x_0)$, i.e. $x_0<x_s<\tilde
1481: x_0(x_0)$. Similarly, $\rho_{x_s}(\tilde x_0(x_0))$=$\rho_{c,r}(\tilde
1482: x_0)$ if $x_0<x_s<\tilde x_0(x_0)$.
1483: When the shock is outside this interval
1484: ($x_s\notin [x_0,\tilde x_0]$),
1485: then $\rho_{x_s}(x_0)$=$\rho_{e}(x_0)$, where $\rho_{e}(x)$ is
1486: some equal-density solution determined by the global density
1487: and $\rho_{x_s}(x_0)>1/2$ or $\rho_{x_s}(x_0)<1/2$ if
1488: $x_s<x_0$ or $x_s>\tilde x_0(x_0)$, respectively.
1489: As a consequence of the particle-hole symmetry,
1490: the relation $\rho_{x_s}(x_0)=1-\rho_{\tilde
1491: x_s(x_s)}(x_0)$ holds if $x_s\notin [x_0,\tilde x_0]$.
1492: Furthermore, in the stationary state, the probability that $x_s<x_0$ is
1493: equal to the probability that $x_s>\tilde x_0(x_0)$
1494: for any $x_0\le \overline x$.
1495: Therefore the contribution to
1496: the average profile is $1/2$ when $x_s\notin [x_0,\tilde x_0]$.
1497: We can thus write for the average densities at $x_0<\overline x$ and
1498: $\tilde x_0(x_0)>\overline x$
1499: \beqn
1500: \rho(x_0)=p(x_0)\rho_{c,l}(x_0)+[1-p(x_0)]\frac{1}{2}, \nonumber \\
1501: \rho(\tilde x_0)=p(x_0(\tilde x_0))\rho_{c,r}(\tilde
1502: x_0)+[1-p(x_0(\tilde x_0))]\frac{1}{2},
1503: \label{average}
1504: \eeqn
1505: respectively, where $p(x_0)$ is the probability that the shock in lane
1506: $A$ resides in the interval $[x_0,\tilde x_0(x_0)]$, and
1507: $x_0(\tilde x_0)$ is the inverse function of $\tilde x_0(x_0)$.
1508: For the spatial derivatives of the densities at $x_0$ and $\tilde
1509: x_0(x_0)$, we get
1510: \beqn
1511: \rho'(x_0)=p'(x_0)\left(\rho_{c,l}(x_0)-\frac{1}{2}\right)+p(x_0)\rho_{c,l}'(x_0),
1512: \nonumber \\
1513: \rho'(\tilde x_0)=p'(x_0)x_0'(\tilde x_0) \left(\rho_{c,l}(\tilde
1514: x_0)-\frac{1}{2}\right)+p(x_0(\tilde x_0))\rho_{c,r}'(\tilde x_0), \nonumber
1515: \\
1516: \eeqn
1517: where the prime denotes derivation.
1518: Using that $p(\overline x)=0$ and $\rho_{c,l}(\overline
1519: x)=1-\rho_{c,r}(\overline x)$, we obtain for the ratio of the left-
1520: and right-hand side
1521: derivatives at $\overline x$:
1522: \be
1523: \rho'(\overline x-)/\rho'(\overline x+)=|\tilde x_0'(\overline x)|.
1524: \ee
1525: Expanding the functions $\rho_{c,l}(x)$, $\rho_{c,r}(x)$ and
1526: $\rho_{e}(x)$ in Taylor series up to first order in $x$
1527: around $x=\overline x$, we obtain
1528: \be
1529: |\tilde x_0'(\overline x)|=
1530: \frac{\rho'_{c,l}(\overline x)-\rho'_{e}(\overline x)}
1531: {\rho'_{e}(\overline x)-\rho'_{c,r}(\overline x)}.
1532: \ee
1533: Using eq. (\ref{diff1}), these derivatives can be given
1534: in terms of $\rho_0\equiv \rho_{c,\alpha}(\overline x)$
1535: and we arrive at eq. (\ref{rat}).
1536:
1537:
1538:
1539: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1540: \acknowledgments
1541: The author thanks L. Santen and F. Igl\'oi for useful discussions.
1542: This work has been supported by the National Office of Research and
1543: Technology under grant No. ASEP1111 and by the Deutsche
1544: Forschungsgemeinschaft under grant No. SA864/2-2.
1545:
1546:
1547: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1548:
1549:
1550: \begin{thebibliography}{}
1551:
1552: \bibitem{liggett} T.M. Liggett {\it Stochastic interacting systems:
1553: contact, voter, and exclusion processes}, (Berlin, Springer, 1999).
1554:
1555: \bibitem{zia}
1556: B. Schmittmann and R. Zia, in {\it Phase Transitions and Critical Phenomena},
1557: vol. 17, edited by C. Domb and J.L. Lebowitz
1558: (Academic, London, 1995), Vol. 17.
1559:
1560: \bibitem{evans}
1561: M.R. Evans, Braz. J. Phys. {\bf 30}, 42 (2000).
1562:
1563: \bibitem{mcdonald} C. MacDonald, J. Gibbs, and A. Pipkin, Biopolymers
1564: {\bf 6}, 1 (1968).
1565:
1566: \bibitem{schutzreview} For a review, see G.M. Sch\"utz, in {\it Phase Transitions and Critical Phenomena}, vol. 19, edited by C. Domb and J.L. Lebowitz (Academic, San Diego, 2001), Vol. 19.
1567:
1568: \bibitem{krug} J. Krug, Phys. Rev. Lett. {\bf 67}, 1882 (1991).
1569:
1570: \bibitem{derrida} B. Derrida, M.R. Evans, V. Hakim and V. Pasquier,
1571: J. Phys. A {\bf 26}, 1493 (1993).
1572:
1573: \bibitem{schutzdomany} G. Sch\"utz and E. Domany, J. Stat. Phys. {\bf
1574: 72}, 277 (1993).
1575:
1576: \bibitem{santen} D. Chowdhury, L.~Santen, and A.~Schadschneider,
1577: Phys. Rep. {\bf 329}, 199 (2000).
1578:
1579: \bibitem{schad} D. Chowdhury, A. Schadschneider, and K. Nishinari,
1580: Physics of Life Reviews (Elsevier, New York, 2005), Vol. 2, p. 318.
1581:
1582: \bibitem{howard}
1583: J. Howard, {\it Mechanics of Motor Proteins and the Cytoskeleton}
1584: (Sinauer, Sunderland, 2001).
1585:
1586: \bibitem{nieuwenhuizen}
1587: R. Lipowsky, S. Klumpp and T.M. Nieuwenhuizen,
1588: Phys. Rev. Lett. {\bf 87}, 108101 (2001).
1589:
1590: \bibitem{klumpp}
1591: S. Klumpp and R. Lipowsky, J. Stat. Phys. {\bf 113}, 233 (2003).
1592:
1593: \bibitem{muller} M.J.I. M\"uller, S. Klumpp and R. Lipowsky,
1594: J. Phys.: Condens. Matter {\bf 17}, 3839 (2005).
1595:
1596: \bibitem{challet} R.D. Willmann, G.M. Sch\"utz and D. Challet, Physica
1597: A {\bf 316}, 430 (2002).
1598:
1599: \bibitem{frey} A. Parmeggiani, T. Franosch and E. Frey, Phys. Rev. Lett. {\bf 90}, 086601 (2003); Phys. Rev. E {\bf 70}, 046101 (2004).
1600:
1601: \bibitem{schutz03} V. Popkov, A. R\'akos, R.D. Willmann,
1602: A.B. Kolomeisky and G.M. Sch\"utz, Phys. Rev. E {\bf 67}, 066117 (2003).
1603:
1604: \bibitem{ejs} M.R. Evans, R. Juh\'asz and L. Santen, Phys. Rev. E {\bf
1605: 68}, 026117 (2003).
1606:
1607: \bibitem{rakos1} A. R\'akos, M. Paessens and G.M. Sch\"utz,
1608: Phys. Rev. Lett. {\bf 91}, 0238302 (2003).
1609:
1610: \bibitem{js}
1611: R. Juh\'asz and L. Santen, J. Phys. A {\bf 37}, 3933 (2004).
1612:
1613: \bibitem{klumpp2}
1614: S. Klumpp and R. Lipowsky, Europhys. Lett. {\bf 66}, 90 (2004).
1615:
1616: \bibitem{levine}
1617: E. Levine and R.D. Willmann, J. Phys. A: Math. Gen. {\bf 37}, 3333 (2004).
1618:
1619: \bibitem{ehk}
1620: M.R. Evans, T. Hanney and Y. Kafri, Phys. Rev. E {\bf 70}, 066124 (2004).
1621:
1622: \bibitem{rakos2}
1623: A. R\'akos and M. Paessens, J. Phys. A: Math. Gen. {\bf 39} 3231 (2006).
1624:
1625: \bibitem{pierobon}
1626: P. Pierobon, T. Franosch and E. Frey, Phys. Rev. E {\bf 74}, 031920 (2006).
1627:
1628: \bibitem{mobilia}
1629: P. Pierobon, M. Mobilia, R. Kouyos and E. Frey, Phys. Rev. E {\bf 74},
1630: 031906 (2006).
1631:
1632: \bibitem{hinsch}
1633: H. Hinsch, R. Kouyos and E. Frey, in {\it Traffic and Granular Flow
1634: '05}, edited by A. Schadschneider, T. P\"oschel, R. K\"uhne,
1635: M. Schreckenberg and D.E. Wolf (Springer, New York, 2006).
1636:
1637: \bibitem{oshanin}
1638: O. B\'enichou, A.M. Cazabat, A. Lemarchand, M. Moreau and G. Oshanin,
1639: J. Stat. Phys. {\bf 97}, 351 (1999).
1640:
1641: \bibitem{schutz98} A.B. Kolomeisky, G.M. Sch\"utz, E.B. Kolomeisky and
1642: J.P. Straley, J. Phys. A: Math. Gen. {\bf 31}, 6911 (1998).
1643:
1644: \bibitem{konzack}
1645: S. Konzack, P.E. Rischitor, C. Enke and R. Fischer,
1646: Mol. Biol. Cell {\bf 16}, 497 (2005).
1647:
1648: \bibitem{nishinari} K. Nishinari, Y. Okada, A. Schadschneider and
1649: D. Chowdhury, Phys. Rev. Lett. {\bf 95}, 118101 (2005).
1650:
1651: \bibitem{greulich}
1652: P. Greulich, A. Garai, K. Nishinari, A. Schadschneider and
1653: D. Chowdhury, Phys. Rev. E {\bf 75}, 041905 (2007).
1654:
1655: \bibitem{lee}
1656: H.-W. Lee, V. Popkov and D. Kim, J. Phys. A {\bf 30}, 8497 (1997).
1657:
1658: \bibitem{peschel}
1659: V. Popkov and I. Peschel, Phys. Rev. E {\bf 64}, 026126 (2001).
1660:
1661: \bibitem{popkov} V. Popkov and G.M. Sch\"utz, J. Stat. Phys. {\bf
1662: 112}, 523 (2003).
1663:
1664: \bibitem{salerno}
1665: V. Popkov and M. Salerno, Phys. Rev. E, {\bf 69}, 046103 (2004).
1666:
1667: \bibitem{popkov1}
1668: V. Popkov, J. Phys. A: Math. Gen. {\bf 37}, 1545 (2004).
1669:
1670: \bibitem{narrow}
1671: E. Pronina and A.B. Kolomeisky, J. Phys. A: Math. Theor. {\bf 40}, 2275 (2007).
1672:
1673: \bibitem{pk}
1674: E. Pronina and A.B. Kolomeisky, J. Phys. A: Math. Gen. {\bf 37}, 9907
1675: (2004); Physica A {\bf 372}, 12 (2006).
1676:
1677: \bibitem{harris}
1678: R.J. Harris and R.B. Stinchcombe, Physica A {\bf 354}, 582 (2005).
1679:
1680: \bibitem{mitsudo}
1681: T. Mitsudo and H. Hayakawa, J. Phys. A: Math. Gen. {\bf 38}, 3087 (2005).
1682:
1683: \bibitem{reichenbach}
1684: T. Reichenbach, T. Franosch and E. Frey,
1685: Phys. Rev. Lett. {\bf 97}, 050603 (2006); T. Reichenbach, E. Frey, and
1686: T. Franosch, New. J. Phys. {\bf 9}, 159 (2007).
1687:
1688: \bibitem{lighthill}
1689: M.J. Lighthill and G.B. Whitham, Proc. R. Soc. London, Ser. A {\bf 229}, 281 (1955).
1690:
1691:
1692: \end{thebibliography}
1693: \end{document}
1694:
1695: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1696: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1697: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1698: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1699: % End of the document
1700: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1701: