1:
2: \documentclass[12pt]{iopart}
3: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4: %TCIDATA{TCIstyle=article/art4.lat,jart,sw20jart}
5:
6: %TCIDATA{Created=Sat Jul 28 14:09:51 2001}
7: %TCIDATA{LastRevised=Thu Feb 07 09:24:49 2002}
8:
9: %\input{tcilatex}
10: \begin{document}
11:
12: \title{Phase diagram of a driven interacting three-state lattice gas}
13: \author{E Lyman and B Schmittmann}
14: \address{Center for Stochastic Processes in Science and Engineering, and Physics Department, Virginia Tech, Blacksburg, VA 24061-0435, USA}
15:
16: \date{today}
17: \vspace{-0.5cm}
18: \begin{abstract} We present Monte Carlo simulations of a three-state lattice
19: gas, half-filled with two types of particles which attract one another,
20: irrespective of their identities. A bias drives the two particle species in
21: opposite directions, establishing and maintaining a non-equilibrium steady
22: state. We map out the phase diagram at fixed bias, as a function of
23: temperature and fraction of the second species. As the temperature is
24: lowered, a continuous transition occurs, from a disordered homogeneous into
25: two distinct strip-like ordered phases. Which of the latter is selected
26: depends on the admixture of the second species. A first order line separates
27: the two ordered states at lower temperatures, emerging from the continuous
28: line at a non-equilibrium bicritical point. For intermediate fraction of the
29: second species, all three phases can be observed.
30: \end{abstract}
31: \vspace{-0.7cm}
32: \pacs{05.10.Ln, 05.50+q, 64.60.Cn}
33:
34:
35: %\\Key Words: Driven lattice gases, phase transitions, non-equilibrium.
36: %\\Contact: B. Schmittmann, e-mail: SCHMITTM@vt.edu; +1 540 231-6518,
37: % FAX: +1 540 231-7511.\
38:
39: %\submitted
40: %\maketitle
41:
42:
43: {\em Introduction. }For systems in thermal equilibrium, the theoretical
44: framework is firmly laid, resting on the work of Boltzmann and Gibbs over a
45: hundred years ago. In particular, the study of simple equilibrium models has
46: a long and illustrious history, as reduction in complexity facilitates the
47: development of theoretical techniques and intuition. In contrast, for
48: systems far from equilibrium there exists no general theoretical framework,
49: and the field remains in an undeveloped state. The strategy of investigating
50: simple models motivates our Monte Carlo study of a driven diffusive system
51: far from equilibrium. A modification of the Ising model, our system departs
52: from well-travelled ground in equilibrium statistical mechanics. Our goal is
53: to develop some intuition about systems far from equilibrium while extending
54: earlier work in the field \cite{SZ-rev}.
55:
56: Almost twenty years ago, Katz, Lebowitz, and Spohn (KLS) \cite{KLS}
57: introduced a generalization of the Ising lattice gas \cite{Ising}, motivated
58: by the physics of fast ionic conductors \cite{FIC}: A bias $E$
59: is applied along a specified lattice axis, driving the particles much like
60: an electric field would drive positive charges. With conserved density and
61: periodic boundary
62: conditions, the system settles into a non-equilibrium steady state,
63: characterized by a uniform particle current. Similar to the equilibrium
64: Ising model, the KLS phase space consists of a high temperature disordered
65: phase and a low temperature phase-separated phase, characterized by a
66: particle-rich strip {\em parallel} to the field direction. At half-filling,
67: the transition remains continuous, but shifts to a higher temperature $%
68: T_{c}(E)$. Remarkably, the transition belongs to a novel universality
69: class \cite{FT,MC-KLS,AFSS}, distinct from the Ising class. One of its key
70: signals is strong anisotropy: wave vectors scale as $k_{\Vert }\sim k_{\bot
71: }^{1+\Delta }$, with $k_{\Vert }$ in the bias direction and $\Delta =2$ in
72: dimension $d=2$. Away from $T_{c}(E)$, a conserved order parameter, coupled
73: with the lifting of the detailed balance constraint, generates power law
74: decays of correlations at {\em all} $T>T_{c}$ \cite{2-pt}.
75:
76: A natural generalization \cite{SHZ} of the KLS model introduces a second,
77: negatively ``charged'' particle species, driven in the opposite direction by
78: the bias. In the high $T$, high $E$ limit where interparticle interactions
79: can be neglected, the system has
80: a line of phase transitions, separating a disordered phase from an
81: inhomogeneous ordered phase. In the ordered phase, which prevails at high
82: density, particles of opposite charge block each other's progress,
83: forming a charge-segregated strip{\em \ transverse} to the field. Both first
84: and second order transitions can occur \cite{VZS,KSZ}. The blocking
85: transition persists in systems carrying {\em nonzero} charge, giving rise to
86: slowly drifting strips \cite{LZ,BZ}. Slow and fast cars, observed in a
87: co-moving frame, offer a good analogy: a blocking transition (traffic jam)
88: occurs when vehicles are sufficiently dense \cite{traffic}. We refer to this
89: noninteracting system as the `two-species model' for short.
90:
91: It is natural to wonder what will happen if we lift the high field, high
92: temperature constraint. Now the particles should ``feel'' the Ising
93: interaction over some range of temperatures, and all three phases (disorder,
94: parallel and transverse strips) may exist in phase space. Several questions
95: emerge immediately. First, we can explore the stability of the KLS
96: universality class by replacing a few positive particles by negative ones.
97: Eventually, however, a blocking transition will occur when a critical charge
98: is exceeded. Similarly, the two-species limit can be probed by taking the
99: strength of interparticle interactions to zero. In this letter, we limit
100: ourselves to establishing the {\em presence of all three phases}: the
101: disordered phase, the transverse strip associated with the blocking
102: transition, and the parallel strip associated with KLS order. Details will
103: be deferred to a future publication \cite{LS2}. In the next section, we will
104: introduce the model specifications and our choice of order parameters, to set
105: the stage for our Monte Carlo results. We conclude with some open questions.
106:
107: {\em The microscopic model and order parameters.} The configurations of
108: our model are specified by a set of occupation variables, $\left\{ s\left(
109: {\bf r}\right) \right\} $, where ${\bf r}\equiv \left( x,y\right) $ labels a
110: site on a fully periodic square lattice of dimensions $L_{x}\times L_{y}$,
111: and each $s\left( {\bf r}\right) $ can take the values $+1$, $-1$, or $0$
112: for a positive particle, negative particle, or hole. The drive
113: $E$ points in the positive $y$-direction. We also introduce the variable
114: $n\left( {\bf r}\right) \equiv \left| s\left( {\bf r}\right) \right| $ to
115: distinguish particles (of either species) from holes. For later reference,
116: we define the mass density $m=\frac{1}{L_{x}L_{y}}\sum_{{\bf r}}n\left( {\bf %
117: r}\right) $ and the charge density $q=\frac{1}{L_{x}L_{y}}\sum_{{\bf r}%
118: }s\left( {\bf r}\right) $. To ensure access to the KLS critical
119: point, we study systems with $m=0.5$, i.e., half-filled lattices.
120: The particles are endowed with attractive nearest-neighbor interactions of
121: strength $J>0$, which are {\em independent }of charge and controlled by the
122: usual Ising Hamiltonian
123: \begin{equation}
124: H=-4J\sum\limits_{\left\langle {\bf r,r}^{\prime }\right\rangle }n\left(
125: {\bf r}\right) n\left( {\bf r}^{\prime }\right) \label{Ising-H}
126: \end{equation}
127: We may set $J=1$ without losing any interesting physics. A given
128: configuration evolves in time as follows. A nearest-neighbor bond is chosen
129: at random, and, if occupied by a particle-hole pair, its contents are
130: exchanged according to the Metropolis \cite{MRRTT} rate $\min \left\{ 1,\exp
131: \left[ -\left( \Delta H-\delta yEs\left( {\bf r}\right) \right) /T\right]
132: \right\} $. Here, the second term models the effect of
133: the drive: if the particle, of charge $s$, is initially located at ${\bf r}$%
134: , $\delta y$ is the change in its $y$-coordinate due to the jump. Thus,
135: positive (negative) charges jump preferentially along (against) the field
136: direction. The parameter $T$ (``temperature'') models the coupling to a
137: thermal bath. Particle-particle (i.e., charge) exchanges are not allowed.
138:
139: We note, first, that this dynamics is {\em diffusive}, i.e., it conserves
140: particle and charge densities. Second, even though the drive mimics an
141: electrostatic potential, the boundary conditions prohibit the existence of a
142: {\em global} Hamiltonian. As a consequence, the system settles into a
143: generic {\em nonequilibrium} steady state. Third, we briefly review the
144: different limits of this model: For $q=m=0.5$, we obtain the KLS model,
145: while $J/T\rightarrow 0$ at finite $E/T$ is the two species case. Of course,
146: the equilibrium Ising model is recovered for $E=0$, $q=0.5$.
147: Thus, the natural control parameters for our study are temperature $T$
148: (measured in units of the Onsager value), the drive $E$ (measured in units
149: of $J$) and the system charge $q$.
150:
151: Due to the conservation laws, the ordered phases are spatially
152: inhomogeneous. Anticipating strip-like ordered domains, we select an
153: order parameter sensitive to such structures, i.e., the equal-time
154: structure factor associated with the particle distribution,
155: \begin{equation}
156: \left\langle \Phi (m_{x},m_{y})\right\rangle \equiv \left\langle \left|
157: \frac{\pi}{L_{x}L_{y}}\sum_{x,y}n(x,y)e^{2\pi i(m_{x}x/L_{x}+m_{y}y/L_{y})}%
158: \right| ^{2}\right\rangle \label{SF}
159: \end{equation}
160: Here, $\left\langle \cdot \right\rangle $ denotes a configurational average,
161: and the integers $m_{x},m_{y}$ index the wave vector.
162: Strips transverse and parallel to the drive are
163: easily identified by considering the smallest nonzero wavevectors in the $x$%
164: - and $y$-directions, respectively. Specifically, a perfect strip along the $%
165: y$-direction corresponds to $\left\langle \Phi (1,0)\right\rangle =1$ while
166: a random configuration gives $\left\langle \Phi \right\rangle =O(\frac{1}{%
167: L_{x}L_{y}})$. By disregarding the phase, the fluctuations in the strip
168: position do not interfere with the averaging procedure. While other choices of order parameter are of
169: course possible, we prefer $\left\langle \Phi (m_{x},m_{y})\right\rangle $
170: since both its high- and low-temperature limits are exactly known.
171: All simulations are run on $40\times 40$ lattices, starting
172: from random initial configurations except where noted.
173: One Monte Carlo step (MCS) is defined as
174: $2L_{x}L_{y}$ update attempts. When averaging, the first $2\times 10^{5}$ MCS
175: are discarded
176: to let the system reach the steady state, and measurements are taken every $%
177: 200$ MCS for the next $8\times 10^{5}$ MCS.
178:
179: {\em Monte Carlo results. }The parameter space for our model is spanned by $T
180: $, $E$, and $q$. In order to establish the presence of all three phases, $E$
181: must be chosen judiciously. To date, the driven Ising model has mostly been
182: studied at infinite drive, where jumps against $E$ are completely
183: suppressed, in order to maximize its nonequilibrium characteristics. This
184: choice, however, renders our two-species system non-ergodic: any
185: configuration in which the minority species forms a blockage, even if it is
186: just a single row spanning the system in the transverse direction, will
187: never break up, regardless of whether such a configuration is stable,
188: metastable, or just a random fluctuation. Thus, a much smaller $E$ must be
189: selected if we wish to observe transitions from transverse to parallel
190: strips, with reasonable decay times. Exploratory runs show $E=2$ to be a
191: good choice. The remaining parameter space is now two-dimensional with axes $%
192: (T,q)$, and we map out the phase diagram in this plane. We first consider $%
193: q=0.5$, corresponding to {\em zero} negative charges, at finite $E$. On this
194: line, we find a single continuous transition, at $T_{c}(2)\simeq 1.15$, to
195: the KLS ordered state, i.e., a single strip aligned with $E$. A detailed
196: anisotropic finite-size scaling \cite{AFSS} study, to be published elsewhere \cite{LS2},
197: indicates that this transition is in the usual KLS class, consistent with
198: field-theoretic predictions \cite{FT}. Next, we turn to a smaller charge, $%
199: q=0.425$, which corresponds to exactly $60$ negative particles, i.e., $1.5$
200: rows, on a half-filled $40\times 40$ lattice. Fig.~1 shows time traces of $%
201: \Phi \left( 1,0\right) $ and $\Phi \left( 0,1\right) $ at $4$ different
202: temperatures, all starting from a random initial configuration.
203: At high temperatures, the system is disordered, with both modes essentially
204: zero. As $T$ is lowered to $T\simeq 1.80$, the {\em blocking
205: transition} occurs first, evidenced by $\Phi \left( 0,1\right) $ emerging
206: from the noise. At $T=1.00$,
207: the system settles quickly into a well-developed transverse strip
208: with a strong nonzero signal
209: in $\Phi \left( 0,1\right) $. Lowering $T$ further to $T=0.90$, we
210: observe signatures of metastability: the corresponding time trace
211: initially develops a large $\Phi \left( 1,0 \right) $ which decays after
212: about $2.5\times 10^{5}$ MCS and reorganizes itself into a transverse strip,
213: signalled by $\Phi \left( 0,1 \right) $. Finally, a
214: well-developed parallel strip is observed at $T=0.80$.
215: Thus, we establish a sequence of {\em two
216: transitions} at $q=0.425$, with the blocking transition occuring {\em first}
217: as the temperature is lowered.
218:
219: %===================================
220: \begin{figure}[tbp]
221: \input{epsf}
222: %\par
223: \begin{center}
224: \vspace{-2.cm}
225: \begin{minipage}{0.6\textwidth}
226: \epsfxsize = \textwidth \epsfysize = 0.8\textwidth \hfill
227: \epsfbox{./LS_fig1.eps}
228: \vspace{-1.cm}
229: \end{minipage}
230: \end{center}
231: \caption{Time traces for $\Phi \left( 0,1\right) $ ($ \diamond $) and
232: $\Phi \left( 1,0\right) $ ($\bullet $) vs time (in units of 400 MCS),
233: at four different temperatures for $q=0.425$.}
234: \vspace{-0.5cm}
235: %\label{p1b}
236: \end{figure}
237: %==============================================================
238:
239:
240: The results of our simulation study are summarized in Fig.~2 which shows the
241: phase diagram in the $q$-$T$ plane, for a $40\times 40$ system. As $q$ {\em %
242: decreases} from $0.5$ to $0$, the number of negatively charged particles
243: {\em increases} from $0$ to $400$. To locate and distinguish continuous and
244: first order transitions, we monitor time traces of the order parameters and
245: extract their fluctuations, i.e., $\left\langle \Phi ^{2}\left( 1,0\right)
246: \right\rangle -\left\langle \Phi \left( 1,0\right) \right\rangle ^{2}$ and $%
247: \left\langle \Phi ^{2}\left( 0,1\right) \right\rangle -\left\langle \Phi
248: \left( 0,1\right) \right\rangle ^{2}$. Crossing a continuous transition, the
249: appropriate order parameter rises smoothly from the noise, accompanied by a
250: peak in its fluctuations. For example, at $q=0.425$, the fluctuations of
251: the $\left( 0,1\right) $ mode peak at $T=1.80\pm 0.05$, whence we use this
252: value to (approximately) locate the transition. Proceeding in this manner,
253: we find two lines of continuous transitions, separating the disordered phase
254: from two different ordered phases: For $0.50\geq q\geq $ $0.46$, the
255: disordered (D) phase becomes unstable with respect to a {\em parallel}
256: strip (PS)
257: as in the KLS model, while for $q\leq $ $0.45$ the system orders into
258: the {\em transverse} strip (TS) associated with the blocking transition.
259: However, the parallel strip reemerges, as the true low-temperature
260: configuration: A line of first order transitions begins at $q\simeq 0.46$, $%
261: T\simeq 1.1$, extending to smaller $q$'s and $T$'s. This line separates two
262: {\em ordered} phases: transverse strips which persist at higher
263: temperatures, and parallel strips at lower $T$'s. Near this line, time
264: traces of the order parameters show metastability and hysteresis. To
265: separate stable from metastable configurations (to the accuracy of our
266: simulations), we analyzed long runs up to $2.4\times 10^{6}$ MCS,
267: starting from different initial configurations. For $%
268: q=0.425$, a sharp transition is easily located at $T=0.84\pm 0.01$. This
269: becomes more difficult as $q$ increases since the continuous and first order
270: lines approach one another and the first order character of the lower
271: transition weakens. For smaller $q$, the first order transition shifts to
272: such low temperatures that metastable states are effectively frozen on the
273: time scales of our simulations.
274:
275:
276: %===================================
277: \begin{figure}[tbp]
278: \input{epsf}
279: %\par
280: \begin{center}
281: \vspace{-1.5cm}
282: \begin{minipage}{0.4\textwidth}
283: \epsfxsize = \textwidth \epsfysize = 0.8\textwidth \hfill
284: \epsfbox{./LS_fig2.eps}
285: \vspace{-1.2cm}
286: \end{minipage}
287: \end{center}
288: \caption{Phase diagram for a $ 40 \times 40 $ system in the $q,T$ plane,
289: for $E=2 $.
290: The D-PS ($ \diamond $) and D-TS ($ \triangle $) transition lines
291: are second order. The PS and TS phases
292: are separated by first order transitions ($ \circ $). The junction of the three
293: lines marks the bicritical point.}
294: \vspace{-0.5cm}
295: %\label{p1b}
296: \end{figure}
297: %==============================================================
298:
299: Provided our qualitative picture is confirmed by further tests, the junction
300: of the first and second order lines, at $q\simeq 0.46$, $T\simeq 1.1,$ is a
301: {\em non-equilibrium bicritical} point. In fact, the system shows markedly
302: different behavior at $q=0.45$ and $q=0.46$. At $q=0.46$ the order parameter
303: signalling transverse strips, $\left\langle \Phi \left( 0,1\right)
304: \right\rangle $, reaches a maximum value of just $0.09$ at $T=1.2$, in stark
305: contrast to a maximum value of $0.33$ at $T=1.1$ if $q=0.45$.
306: Moreover,
307: the $q=0.45$ system clearly shows a lower transition
308: with signs of metastability,
309: while none
310: is observed at $q=0.46$. While it is quite remarkable that changing the sign
311: of just {\em eight} particles makes such a difference, we emphasize that $%
312: q=0.45$ corresponds to precisely {\em one full row} of negative particles in
313: a $40\times 40$ system while $q=0.46$ results only in a partially filled row
314: ($32$ negative charges).
315:
316:
317:
318: {\em Conclusions. }We have simulated a lattice gas, consisting of two
319: oppositely charged particle species and holes subject to
320: an ``electric'' field. The particles attract one
321: another, independent of charge. This model interpolates between two
322: well-studied limits: the KLS model \cite{KLS} which has just a single
323: species, and the high-field, high-temperature version of this model \cite
324: {SHZ} where the interactions are irrelevant. Both limits exhibit
325: order-disorder transitions, characterized however, by different ordered
326: phases: a density-segregated strip {\em parallel }to the drive in the KLS
327: limit, and a density- and charge-segregated strip {\em transverse} to $E$ for the
328: two-species limit. Here, we have mapped out the phase diagram for an
329: intermediate value of the drive, where both ordered phases are observed in
330: different regions of parameter space. Lines of first and second order
331: transitions, joined at a bicritical point, demarkate their stability domains.
332:
333: We conclude with some remarks on work in progress \cite{LS2} and open
334: questions. Clearly, we have explored only a limited portion of the huge
335: parameter space. Moreover, a systematic finite-size scaling analysis is
336: needed to obtain a better estimate of the transition lines and to extract
337: the critical properties of the continuous transitions. Preliminary studies
338: show that the mean features of our phase diagram are independent of system
339: size. Analytic work, ranging from mean-field to full-fledged renormalized
340: field theory, should provide further insights. A particularly intriguing
341: question is how $q$ and $E$ should scale with the lattice dimensions. If
342: both are held fixed when performing the standard finite size analysis for
343: the KLS model, the blocking transition will eventually supersede the KLS
344: transition: a fixed charge density corresponding to a single row in a square
345: system becomes several rows thick in a ``long skinny'' system. Clearly, much
346: remains to be explored before this rich system is understood in detail.
347:
348: {\em Acknowledgements.} We thank R.K.P.~Zia, R.J.~Astalos and U.C.~T\"{a}uber for helpful discussions. Partial support from the National Science Foundation through DMR-0088451 is gratefully acknowledged.
349:
350: { \bf References}
351: \begin{thebibliography}{99}
352:
353: \bibitem{SZ-rev} Schmittmann B and Zia R K P 1995 {\em Phase Transitions and
354: Critical Phenomena} vol 17, ed C Domb and J L Lebowitz
355: (London: Academic)
356:
357: \bibitem{KLS} Katz S, Lebowitz J L and Spohn H 1983 {\em Phys.~Rev}.~{\bf B28} 1655 and 1984 {\em J.~Stat.~Phys.}~\ {\bf 34} 497
358:
359: \bibitem{Ising} Ising E 1925 {\em Z.~Physik} {\bf 31} 253; McCoy B M and
360: Wu T T 1973 {\em The Two-dimensional Ising Model} (Cambridge, Mass: Harvard Univ.~Press)
361:
362: \bibitem{FIC} See, e.g., Chandra S 1981 {\em Superionic Solids. Principles
363: and Applications} (Amsterdam: North Holland)
364:
365: \bibitem{FT} Janssen H K and Schmittmann B 1986 {\em Z.~Phys} {\bf B64} 503;
366: Leung K-t and Cardy J L 1986 {\em J.~Stat.~Phys.}~{\bf 44} 567
367: and {\bf 45} 1087 (erratum)
368:
369: \bibitem{MC-KLS} Wang J-S 1996 {\em J.~Stat.~Phys} {\bf 82} 1409; Leung K -t
370: and Wang J -S 1999 {\em Int.~J.~Mod.~Phys.}~{\bf C10} 853; Caracciolo S, Gambassi A,
371: Gubinelli M, Pelissetto A cond-mat/0106221.
372:
373: \bibitem{AFSS} Leung K-t, 1991 {\em Phys.~Rev.~Lett.}~{\bf 66} 453 and 1992 {\em Int.~
374: J.~Mod.~Phys.}~{\bf C3} 367
375:
376: \bibitem{2-pt} Zhang M Q, Wang J-S, Lebowitz J L and Vall\`{e}s J L
377: 1988 {\em J.~Stat.~Phys.}~{\bf 52} 1461; \ Garrido P L, Lebowitz J L,
378: Maes C and Spohn H 1990 {\em Phys.~Rev.}~{\bf A42} 1954; Grinstein G 1991 J.~Appl.~
379: Phys.~{\bf 69} 5441; Schmittmann B and Zia R K P 1998 {\em J.~Stat.~Phys.}~
380: {\bf 91} 525
381:
382: \bibitem{SHZ} Schmittmann B, Hwang K and Zia R K P 1992 {\em Europhys.~Lett.}~{\bf 19} 19
383:
384: \bibitem{VZS} Vilfan I, Zia R K P and Schmittmann B 1994
385: {\em Phys.~Rev.~Lett.}~{\bf 73} 2071
386: \bibitem{KSZ} Korniss G, Schmittmann B and Zia R K P 1995 {\em Europhys.~Lett.}~{\bf 32},
387: 49 and 1996 {\em J.~Stat.~Phys.}~{\bf 86} 721
388:
389: \bibitem{LZ} Leung K-t and Zia R K P 1997 {\em Phys.~Rev.}~{\bf E56} 308
390:
391: \bibitem{BZ} Brookings T and Zia R K P, private communication
392:
393: \bibitem{traffic} For a review, see Chowdhury D, Santen L, and Schadschneider A 2000
394: {\em Phys.~Rep.}~{\bf 329} 199
395:
396: \bibitem{MRRTT} Metropolis N, Rosenbluth A W, Rosenbluth M M, Teller A H
397: and Teller E 1953 {\em J.~Chem.~Phys.}~{\bf 21} 1087
398:
399: \bibitem{LS2} Lyman E and Schmittmann B, unpublished
400:
401: \end{thebibliography}
402:
403: \end{document}
404: