1: \documentclass[12pt]{iopart}

2: \usepackage{graphicx}

3: \eqnobysec

4: 

5: \newcommand{\beq}{\begin{equation}}

6: \newcommand{\beqa}{\begin{eqnarray}}

7: \newcommand{\eeq}{\end{equation}}

8: \newcommand{\eeqa}{\end{eqnarray}}

9: \newcommand{\al}{\alpha}

10: \newcommand{\be}{\beta}

11: \newcommand{\bin}[2]{{#1\choose#2}}

12: \renewcommand{\d}{{\rm d}}

13: \newcommand{\demi}{{\textstyle\frac12}}

14: \newcommand{\ea}{{\rm e}^{2J_A}}

15: \newcommand{\eb}{{\rm e}^{2J_B}}

16: \newcommand{\ec}{{\rm e}^{2J_C}}

17: \newcommand{\eab}{{\rm e}^{2J_A+2J_B}}

18: \newcommand{\eac}{{\rm e}^{2J_A+2J_C}}

19: \newcommand{\ebc}{{\rm e}^{2J_B+2J_C}}

20: \newcommand{\eabc}{{\rm e}^{2J_A+2J_B+4J_C}}

21: \newcommand{\eps}{\varepsilon}

22: \newcommand{\frad}[2]{\displaystyle{\displaystyle#1\over\displaystyle#2}}

23: \newcommand{\haut}{\rule[-14pt]{0pt}{36pt}}

24: \renewcommand{\in}{{\rm in}}

25: \newcommand{\lef}{{\rm left}}

26: \newcommand{\mean}[1]{\left\langle#1\right\rangle}

27: \newcommand{\out}{{\rm out}}

28: \newcommand{\rig}{{\rm right}}

29: \newcommand{\s}{{\rm st}}

30: \renewcommand{\u}{u}

31: \newcommand{\w}{w}

32: \newcommand{\x}{x}

33: \newcommand{\y}{y}

34: \newcommand{\C}{{\cal C}}

35: \newcommand{\D}{\Delta}

36: \renewcommand{\H}{{\cal H}}

37: \newcommand{\K}{{\rm K}}

38: \renewcommand{\L}{{\rm L}}

39: \newcommand{\M}{{\rm M}}

40: \newcommand{\N}{{\rm N}}

41: 

42: \begin{document}

43: 

44: \title[Nonequilibrium stationary states with Gibbs measure

45: for two or three species]

46: {Nonequilibrium stationary states with Gibbs measure

47: for two or three species of interacting particles}

48: 

49: \author{J M Luck\dag\ and C Godr\`{e}che\ddag}

50: 

51: \address{\dag\ Service de Physique Th\'eorique\footnote{URA 2306 of CNRS},

52: CEA Saclay, 91191 Gif-sur-Yvette cedex, France}

53: 

54: \address{\ddag\ Isaac Newton Institute for Mathematical Sciences,

55: 20 Clarkson Road, Cambridge\break CB3 0EH, UK, and

56: Service de Physique de l'\'Etat Condens\'e\footnote{URA 2464 of CNRS},

57: CEA Saclay, 91191 Gif-sur-Yvette cedex, France}

58: 

59: \begin{abstract}

60: We construct explicit examples of one-dimensional driven diffusive sys\-tems

61: for two and three species of interacting particles,

62: defined by asymmetric dynamical rules which do not obey detailed balance,

63: but whose nonequilibrium stationary-state measure coincides

64: with a prescribed equilibrium Gibbs measure.

65: For simplicity, the measures considered in this construction

66: only involve nearest-neighbor interactions.

67: For two species,

68: the dynamics thus obtained generically has five free parameters,

69: and does not obey pairwise balance in general.

70: The latter property is satisfied only by the totally asymmetric dynamics

71: and the partially asymmetric dynamics with uniform bias,

72: i.e., the cases originally considered by Katz, Lebowitz, and Spohn.

73: For three species of interacting particles,

74: with nearest-neighbor interactions between particles of the same species,

75: the totally asymmetric dynamics thus obtained has two free parameters,

76: and obeys pairwise balance.

77: These models are put in perspective with other examples of driven

78: diffusive systems.

79: The emerging picture is that asymmetric (nonequilibrium) stochastic dynamics

80: leading to a given stationary-state measure

81: are far more constrained (in terms of numbers of free parameters)

82: than the corresponding symmetric (equilibrium) dynamics.

83: \end{abstract}

84: 

85: \pacs{05.70.Ln, 05.40.--a, 02.50.Ey}

86: 

87: \eads{

88: \mailto{luck@dsm-mail.saclay.cea.fr},

89: \mailto{godreche@dsm-mail.saclay.cea.fr}}

90: 

91: \maketitle

92: 

93: \section{Introduction}

94: 

95: Driven diffusive systems~\cite{s,sz,l,sch}

96: are defined by stochastic dynamical rules

97: that incorporate the effect of an external drive,

98: and therefore do not obey detailed balance, which makes

99: their nonequilibrium stationary states difficult to study in general.

100: The first step of evaluating the stationary-state measure,

101: i.e., the probability $P_\s(\C)$

102: of any given configuration~$\C$, is already a non-trivial task.

103: In view of the lack of a general theory of nonequilibrium stationary states,

104: one has to rely on the investigation of specific models,

105: for which the stationary-state measure is analytically tractable.

106: Examples of such models are the simple exclusion processes,

107: the zero-range process, and the one-dimensional Katz-Lebowitz-Spohn (KLS)

108: model.

109: 

110: For the simple exclusion process,

111: particles obeying an exclusion constraint perform random walks on a lattice,

112: with either symmetric or biased moves~\cite{l,sch,der}.

113: On a ring, i.e., with periodic boundary conditions,

114: the stationary-state measure is uniform, irrespectively of the bias:

115: all the configurations are equally probable.

116: The same model with open boundaries has a stationary-state measure

117: which can be described in terms of a matrix product Ansatz.

118: The zero-range process (ZRP)~\cite{spitzer}, in the presence of a bias,

119: belongs to the class of driven stochastic

120: processes with multiple occupancies.

121: Its stationary state has a product measure, again irrespectively of the bias.

122: The occupation numbers of the sites are independent random

123: quantities with a common distribution,

124: up to the conservation of the total number of particles.

125: The existence of unbounded occupancies however opens up the possibility

126: of having a condensation transition,

127: irrespective of the geometry of the underlying lattice,

128: and therefore also in one dimension~\cite{ev,lux}.

129: 

130: The KLS model is a lattice gas model of interacting charged particles

131: subjected to an external electric field~\cite{kls}.

132: It is representative of the class of models with non-equilibrium stationary

133: state measures incorporating physical interactions between particles.

134: The stationary-state measure of the KLS model is non trivial

135: in two dimensions and above, where e.g.~the critical temperature

136: depends continuously on the applied field~\cite{sz}.

137: The situation however simplifies in the one-dimensional geometry,

138: where the model is equivalent to a chain of classical Ising spins $s_n=\pm1$.

139: In this case, there exists a class of biased stochastic dynamics,

140: for which the nonequilibrium stationary-state measure

141: is the canonical finite-temperature Gibbs measure,

142: where the probability $P_\s(\C)$ of the configuration $\C=\{s_1,\dots,s_N\}$

143: is given by the Boltzmann formula (with $k_BT=1$)

144: \beq

145: P_\s(\C)=\frac{1}{Z}\,\exp(-\H(\C))

146: \label{boltz}

147: \eeq

148: associated with the Ising Hamiltonian with nearest-neighbor interactions

149: \beq

150: \H=-J\sum_ns_ns_{n+1}.

151: \label{ham2}

152: \eeq

153: This very stochastic model with antiferromagnetic interactions ($J<0$)

154: was subsequently

155: rediscovered in the context of polymer crystallization~\cite{gw}.

156: 

157: At this point it is natural to question the generality of the result

158: found in~\cite{kls} for the case of the Ising chain.

159: In the present work we show that for systems with three species of particles

160: there also exist asymmetric stochastic dynamics

161: which do not obey detailed balance,

162: but whose stationary-state measure is the Gibbs measure

163: corresponding to a simple Hamiltonian.

164: In particular this stationary-state measure is independent of the bias.

165: In our construction we restrict the choice of measures

166: to those involving only nearest-neighbor interactions.

167: We first revisit in Section~2 the case of two species of interacting particles,

168: thus generalizing the study of~\cite{kls} to a wider class of dynamics.

169: The emphasis is put on the role of various symmetries,

170: and especially on the number of free parameters left by imposing them.

171: We then consider, in Section~3, the entirely novel situation

172: of three species of interacting particles.

173: Section~4 contains a discussion, where our results are put in perspective

174: with yet other examples.

175: 

176: Let us finally give a brief reminder of the concepts

177: of detailed balance~\cite{kampen} and pairwise balance~\cite{pairw},

178: which will be used throughout this work.

179: Consider a finite set of configurations $\C$,

180: and a Markovian dynamics in continuous time,

181: defined by the transition rates $W(\C\to\C')$.

182: The master equation for the time-dependent probability $P(\C,t)$ reads

183: \beq

184: \frac{\d P(\C,t)}{\d t}

185: =\sum_{\C'}\Bigl(W(\C'\to\C)P(\C',t)-W(\C\to\C')P(\C,t)\Bigr).

186: \eeq

187: The stationary probability $P_\s(\C)$ therefore obeys the equation

188: \beq

189: \sum_{\C'}\Bigl(W(\C'\to\C)P_\s(\C')-W(\C\to\C')P_\s(\C)\Bigr)=0.

190: \label{bal}

191: \eeq

192: 

193: In the particular case where the Markov process is reversible,

194: the dynamics brings the system to an equilibrium state.

195: Reversibility requires the {\it detailed balance} property~\cite{kampen},

196: that is the absence of probability flux

197: between any pair of configurations~$\C$ and $\C'$ at stationarity:

198: \beq

199: W(\C\to\C')P_\s(\C)=W(\C'\to\C)P_\s(\C').

200: \label{detbal}

201: \eeq

202: Equation~(\ref{detbal}) clearly implies~(\ref{bal}).

203: 

204: A weaker property, referred to as {\it pairwise balance}~\cite{pairw},

205: is adapted to the situation of driven diffusive systems,

206: where the presence of a preferred direction of motion, i.e., a bias,

207: precludes the property of detailed balance.

208: Pairwise balance is defined as follows:

209: for every pair of configurations $\C$ and $\C'$ such that $W(\C\to\C')\neq0$,

210: there exists a third configuration $\C''$ such that

211: \beq

212: W(\C\to\C')P_\s(\C)=W(\C''\to\C)P_\s(\C'').

213: \label{pairbal}

214: \eeq

215: The moves $\C\to\C'$ and $\C''\to\C$ are said to be conjugate to each other.

216: Equation~(\ref{pairbal}) also implies~(\ref{bal}),

217: since terms corresponding to pairs of conjugate moves cancel each other.

218: Several of the examples of dynamics constructed in this paper

219: obey P-related pairwise balance,

220: where pairs of conjugate moves are related to each other by parity.

221: The first explicit example is given below~(\ref{eq3}).

222: An important general consequence of P-related pairwise balance

223: is shown at the end of Section~3.

224: If the totally asymmetric dynamics obeys P-related pairwise balance,

225: the partially asymmetric dynamics

226: with uniform bias~$p$ also obeys pairwise balance,

227: and it has the same stationary-state measure as the totally asymmetric one,

228: irrespective of the value of the bias.

229: 

230: \section{Two species}

231: 

232: Consider a ring made of $N$ sites.

233: Each site is occupied by a particle,

234: which can be either of type $A$ (positively charged)

235: or of type $B$ (negatively charged).

236: We represent the species of particle at site $n$

237: by an Ising spin $s_n=\pm1$ equal to the charge of the particle.

238: Table~\ref{2s} also gives the corresponding indicator variables.

239: For instance, $(1+s_n)/2$ is equal to 1 if the particle at site $n$

240: is of type $A$, and to 0 else.

241: 

242: \begin{table}[ht]

243: \caption{Spin (charge) and indicator variables

244: associated with each particle species in the case of two species.}

245: \label{2s}

246: \begin{center}

247: \begin{tabular}{|c|c|c|}

248: \hline

249: Species at site $n$&Spin (charge)&Indicator variable\\

250: \hline

251: $A$&$s_n=+1$&$(1+s_n)/2$\\

252: $B$&$s_n=-1$&$(1-s_n)/2$\\

253: \hline

254: \end{tabular}

255: \end{center}

256: \end{table}

257: 

258: Our goal is to construct nonequilibrium dynamics

259: such that the stationary-state measure is the Gibbs measure

260: given by~(\ref{boltz}) associated with

261: the nearest-neighbor Hamiltonian~(\ref{ham2}).

262: We consider the asymmetric exchange (Kawasaki) dynamics.

263: Consistently with the form of the Hamiltonian~(\ref{ham2}),

264: the rates depend on the two neighbors

265: of the pair to be exchanged, according to Table~\ref{2d}.

266: The dynamics so defined conserves the numbers $N_A$ and $N_B$

267: of particles of each species, with $N_A+N_B=N$.

268: We first determine the most general dynamics of this form

269: in Section~2.1, and then discuss the interplay between various possible

270: symmetries in Section~2.2.

271: 

272: \begin{table}[ht]

273: \caption{List of moves in the general exchange dynamics

274: for two species of interacting particles,

275: notation for the corresponding exchange rates,

276: and energy difference $\D\H$ involved in the moves,

277: where the Hamiltonian $\H$ is defined in~(\ref{ham2}).}

278: \label{2d}

279: \begin{center}

280: \begin{tabular}{|c|c|c||c|c|c|}

281: \hline

282: Move&Rate&$\D\H$&Move&Rate&$\D\H$\\

283: \hline

284: $AABA\to ABAA$&$\w_{AA}$&$0$&

285: $ABAA\to AABA$&$\x_{AA}$&$0$\\

286: $AABB\to ABAB$&$\w_{AB}$&$4J$&

287: $ABAB\to AABB$&$\x_{AB}$&$-4J$\\

288: $BABA\to BBAA$&$\w_{BA}$&$-4J$&

289: $BBAA\to BABA$&$\x_{BA}$&$4J$\\

290: $BABB\to BBAB$&$\w_{BB}$&$0$&

291: $BBAB\to BABB$&$\x_{BB}$&$0$\\

292: \hline

293: \end{tabular}

294: \end{center}

295: \end{table}

296: 

297: \subsection{The general case}

298: 

299: Consider the numbers $N_{AA},\dots,N_{BB}$ of oriented pairs

300: of neighbors of each species.

301: These numbers obey the sum rules

302: \beq

303: \fl N_A=N_{AA}+N_{AB}=N_{AA}+N_{BA},\qquad N_B=N_{BA}+N_{BB}=N_{AB}+N_{BB}.

304: \label{paireqs}

305: \eeq

306: The sum of the two equations gives twice the same equation,

307: so that the four pair numbers obey three independent equations,

308: leaving one single free quantity.

309: It is convenient to take the latter as being the Hamiltonian $\H$

310: of~(\ref{ham2}).

311: The pair numbers can indeed be expressed

312: as linear combinations of $\H$ and of the particle numbers $N_A$ and~$N_B$:

313: \beq

314: \matrix{

315: N_{AA}=\frad{1}{4}\left(3N_A-N_B-\H/J\right),\quad\hfill

316: &N_{BB}=\frad{1}{4}\left(3N_B-N_A-\H/J\right),\cr

317: N_{AB}=N_{BA}=\frad{1}{4}\left(N+\H/J\right).\hfill&

318: }

319: \label{pairex}

320: \eeq

321: Incidentally, this proves that the Hamiltonian~(\ref{ham2})

322: is the most general form of a pair Hamiltonian for two species of particles

323: with nearest-neighbor interactions.

324: 

325: Throughout this paper, we make use of an alternative and convenient way

326: of automatically taking into account sum rules such as~(\ref{paireqs}).

327: This consists in expressing all the quantities

328: in terms of the spin variables $s_n$ introduced in Table~\ref{2s}.

329: For instance

330: \beq

331: N_A=\frac{1}{2}\sum_n(1+s_n)=\frac{N}{2}\,(1+\mean{s_1}).

332: \label{cordef}

333: \eeq

334: Here and in the following, the brackets $\mean{\dots}$ denote

335: a uniform spatial average {\it for a fixed generic configuration~$\C$.}

336: Recall that all the sites are equivalent, because of translational invariance.

337: The pair numbers and the Hamiltonian read

338: \beq

339: \matrix{

340: \haut N_{AA}=\frad{N}{4}\,(1+2\mean{s_1}+\mean{s_1s_2}),\quad\hfill

341: &N_{BB}=\frad{N}{4}\,(1-2\mean{s_1}+\mean{s_1s_2}),\hfill\cr

342: \haut N_{AB}=N_{BA}=\frad{N}{4}\,(1-\mean{s_1s_2}),\quad\hfill

343: &\H=-NJ\mean{s_1s_2}.\hfill

344: }

345: \eeq

346: Equations such as~(\ref{paireqs}) and~(\ref{pairex})

347: are then automatically satisfied.

348: 

349: Consider now the fate of a generic configuration $\C$.

350: The total exit rate $W_\out(\C)$ from~$\C$ to any other configuration $\C'$

351: per unit time can be read off from Table~\ref{2d}:

352: \beqa

353: &&W_\out(\C)

354: =\w_{AA}N_{AABA}+\w_{AB}N_{AABB}+\w_{BA}N_{BABA}+\w_{BB}N_{BABB}\nonumber\\

355: &&{\hskip 14.3truemm}

356: +\x_{AA}N_{ABAA}+\x_{AB}N_{ABAB}+\x_{BA}N_{BBAA}+\x_{BB}N_{BBAB}.

357: \label{wout}

358: \eeqa

359: An analogous expression can be derived for the total entrance rate

360: $W_\in(\C)$ from any other configuration $\C'$ to $\C$.

361: Using again Table~\ref{2d}, as well as~(\ref{boltz}) to express

362: the stationary-state weight $P_\s(\C')$ as

363: \beq

364: P_\s(\C')=P_\s(\C)\exp(\D\H),

365: \label{delham}

366: \eeq

367: in terms of $P_\s(\C)$ and of the energy difference

368: \beq

369: \D\H=\H(\C)-\H(\C')

370: \eeq

371: involved in the move, we obtain

372: \beqa

373: &&\fl W_\in(\C)=

374: \w_{AA}N_{ABAA}+{\rm e}^{4J}\w_{AB}N_{ABAB}+{\rm e}^{-4J}\w_{BA}N_{BBAA}+\w_{BB}N_{BBAB}

375: \nonumber\\

376: &&\fl{\hskip 12.7truemm}

377: +\x_{AA}N_{AABA}+{\rm e}^{-4J}\x_{AB}N_{AABB}+{\rm e}^{4J}\x_{BA}N_{BABA}+\x_{BB}N_{BABB}.

378: \label{win}

379: \eeqa

380: In the stationary state we have

381: \beq

382: W_\out(\C)-W_\in(\C)=0

383: \label{winwout}

384: \eeq

385: for every configuration~$\C$.

386: In order to determine the number of independent conditions on the rates

387: imposed by this equation,

388: it is convenient to rewrite~(\ref{wout}) and~(\ref{win})

389: in terms of spin correlations, i.e., spatial averages of products of spin

390: variables, denoted as~$\mean{\dots}$, along the lines of~(\ref{cordef}).

391: With these notations, we obtain

392: \beqa

393: &&\fl W_\out(\C)-W_\in(\C)=\frac{N}{16}\,\bigg\{

394: ({\rm e}^{-4J}-1)(\mean{s_1s_2s_3s_4}+1)\,R_1

395: +\mean{s_1(s_3-s_2)s_4}\,R_2\nonumber\\

396: &&\fl{\hskip 42truemm}

397: +\Big[\mean{s_1(3s_2-2s_3+s_4)}+{\rm e}^{-4J}\mean{s_1(s_2-2s_3-s_4)}\Big]

398: \,R_1\bigg\},

399: \label{dw}

400: \eeqa

401: where $R_1$ and $R_2$ stand for the following linear combinations of the rates:

402: \beqa

403: &&R_1={\rm e}^{4J}(\w_{AB}+\x_{BA})-\w_{BA}-\x_{AB},\nonumber\\

404: &&R_2=({\rm e}^{4J}+1)(\w_{AB}-\x_{BA})+({\rm e}^{-4J}+1)(\w_{BA}-\x_{AB})

405: \nonumber\\

406: &&{\hskip 5truemm}-2(\w_{AA}+\w_{BB}-\x_{AA}-\x_{BB}).

407: \label{r1r2}

408: \eeqa

409: The condition~(\ref{winwout}) therefore gives two linear relations,

410: \beq

411: R_1=R_2=0,

412: \label{eq}

413: \eeq

414: between the eight exchange rates defining the general asymmetric dynamics.

415: Let us choose the time unit by setting

416: \beq

417: \w_{AA}+\w_{BB}+\x_{AA}+\x_{BB}=1.

418: \label{2nor}

419: \eeq

420: 

421: The most general asymmetric dynamics for two species of interacting particles

422: such that the stationary-state measure is given by~(\ref{boltz}),~(\ref{ham2})

423: therefore has five free parameters.

424: An explicit parametrization of the rates is given in Table~\ref{2gene}.

425: The dynamics thus obtained does not obey pairwise balance in general.

426: 

427: The parametrization of the solutions to~(\ref{eq}) and~(\ref{2nor})

428: given in Table~\ref{2gene} has been carefully chosen

429: in such a way that the various symmetries

430: to be described below correspond to the simple

431: constraints~(\ref{cstr1}), (\ref{cstr2}), (\ref{cstr3}), (\ref{cstr4})

432: in terms of the parameters~$\delta$ and $\eps_1,\dots,\eps_4$.

433: The parameters~$\delta$ and $\eps_1,\dots,\eps_4$

434: all lie in the range $[-1,+1]$, and are such that the combination

435: \beq

436: \lambda=\frac{(1+\delta)\eps_1+(1-\delta)\eps_2}{\eps_3+\eps_4}

437: \label{lamdef}

438: \eeq

439: is positive.

440: 

441: \begin{table}[ht]

442: \caption{Explicit parametrization of the rates

443: of the most general asymmetric dynamics for two species

444: with Gibbs stationary-state measure~(\ref{boltz}),~(\ref{ham2}).

445: The notation $\lambda$ is defined in~(\ref{lamdef}).}

446: \label{2gene}

447: \begin{center}

448: \begin{tabular}{|c|c||c|c|}

449: \hline

450: Rate&expression&Rate&expression\\

451: \hline

452: \haut$\w_{AA}$&$\frad{(1+\eps_1)(1+\delta)}{4}$&

453: \haut$\x_{AA}$&$\frad{(1-\eps_1)(1+\delta)}{4}$\\

454: \haut$\w_{AB}$&$\frad{(1+\eps_3)\lambda}{2({\rm e}^{4J}+1)}$&

455: \haut$\x_{AB}$&$\frad{(1-\eps_4)\lambda\,{\rm e}^{4J}}{2({\rm e}^{4J}+1)}$\\

456: \haut$\w_{BA}$&$\frad{(1+\eps_4)\lambda\,{\rm e}^{4J}}{2({\rm e}^{4J}+1)}$&

457: \haut$\x_{BA}$&$\frad{(1-\eps_3)\lambda}{2({\rm e}^{4J}+1)}$\\

458: \haut$\w_{BB}$&$\frad{(1+\eps_2)(1-\delta)}{4}$&

459: \haut$\x_{BB}$&$\frad{(1-\eps_2)(1-\delta)}{4}$\\

460: \hline

461: \end{tabular}

462: \end{center}

463: \end{table}

464: 

465: \subsection{The interplay between various symmetries}

466: 

467: The number of free parameters of the dynamics thus obtained

468: is decreased if various kinds of symmetries are imposed onto the dynamics.

469: 

470: \medskip\noindent$\bullet$

471: {\it Symmetric ({\rm P}-invariant) dynamics.}

472: Consider a symmetric dynamics, invariant under the spatial parity P

473: which reverses the orientation of the ring (i.e., interchanges left and right).

474: This symmetry property reads

475: \beq

476: \x_{IJ}=\w_{JI}

477: \label{eq1}

478: \eeq

479: for all values of the indices $I,J=A,B$.

480: The stationary state thus obtained is an equilibrium state.

481: The first equation of~(\ref{eq}),

482: \beq

483: \w_{BA}={\rm e}^{4J}\w_{AB},

484: \label{deb}

485: \eeq

486: expresses detailed balance.

487: Equation~(\ref{eq1}) amounts to setting

488: \beq

489: \eps_i=0\quad(i=1,\dots,4)

490: \label{cstr1}

491: \eeq

492: in Table~\ref{2gene}.

493: The symmetric (equilibrium) dynamics therefore has two free parameters:

494: $\delta$ and $\lambda$, in the ranges $-1<\delta<1$ and $\lambda>0$.

495: The expression~(\ref{lamdef}) for $\lambda$ indeed becomes indeterminate

496: in the limit where all the $\eps_i$ go simultaneously to zero.

497: The rates read

498: \beq

499: \matrix{

500: \haut\w_{AA}=\x_{AA}=\frad{1+\delta}{4},\quad\hfill

501: &\w_{AB}=\x_{BA}=\frad{\lambda}{2({\rm e}^{4J}+1)},\cr\hfill

502: \haut\w_{BA}=\x_{AB}=\frad{\lambda\,{\rm e}^{4J}}{2({\rm e}^{4J}+1)},\quad\hfill

503: &\w_{BB}=\x_{BB}=\frad{1-\delta}{4}.\hfill

504: }

505: \label{ress}

506: \eeq

507: 

508: \medskip\noindent$\bullet$

509: {\it Totally asymmetric dynamics.}

510: Consider a dynamics driven by an infinitely strong electric field,

511: so that the positively (resp.~negatively) charged $A$ particles

512: (resp.~$B$ particles) hop exclusively to the right (resp.~to the left).

513: Therefore

514: \beq

515: \x_{IJ}=0

516: \label{eq2}

517: \eeq

518: for all values of the indices $I,J=A,B$.

519: Equation~(\ref{eq}) becomes

520: \beq

521: \w_{BA}={\rm e}^{4J}\w_{AB}.

522: \label{eq3}

523: \eeq

524: This equation coincides with~(\ref{deb}).

525: It expresses {\it {\rm P}-related pairwise balance:}

526: conjugate moves are related to each other by parity P,

527: i.e., the first and the fourth move of the left column of Table~\ref{2d}

528: are their own conjugates,

529: whereas the second and the third moves are conjugate to each other.

530: Equation~(\ref{eq2}) amounts to setting

531: \beq

532: \eps_i=1\quad(i=1,\dots,4)

533: \label{cstr2}

534: \eeq

535: in Table~\ref{2gene}, so that $\lambda=1$.

536: The totally asymmetric dynamics

537: with stationary-state measure~(\ref{boltz}),~(\ref{ham2})

538: therefore has one free parameter:

539: $\delta$, in the range $-1<\delta<1$.

540: The rates read

541: \beq

542: \matrix{

543: \haut\w_{AA}=\frad{1+\delta}{2},\quad\hfill

544: &\w_{AB}=\frad{1}{{\rm e}^{4J}+1},\hfill\cr

545: \w_{BA}=\frad{{\rm e}^{4J}}{{\rm e}^{4J}+1},\quad\hfill

546: &\w_{BB}=\frad{1-\delta}{2}.\hfill

547: }

548: \label{resas}

549: \eeq

550: 

551: \medskip\noindent$\bullet$

552: {\it Partially asymmetric dynamics with a uniform bias.}

553: This is the most general case originally considered by KLS~\cite{kls}.

554: Consider a dynamics driven by a finite electric field,

555: so that the positively (resp.~negatively) charged $A$ particles

556: (resp.~$B$ particles) hop preferentially to the right (resp.~to the left).

557: Let

558: \beq

559: p=\frad{1+\eps}{2},\qquad q=\frad{1-\eps}{2}

560: \eeq

561: be the a priori probabilities of respectively hopping

562: to the right and to the left,

563: where $0\le\eps\le1$ provides a measure of the applied electric field.

564: This translates into the following uniform bias condition:

565: \beq

566: \frac{\x_{IJ}}{\w_{JI}}=\frac{1-\eps}{1+\eps}

567: \label{eqpq}

568: \eeq

569: for all values of $I,J=A,B$.

570: This situation interpolates between the symmetric case ($p=1/2$, $\eps=0$)

571: and the totally asymmetric one ($p=1$, $\eps=1$).

572: Equation~(\ref{eqpq}) amounts to setting

573: \beq

574: \eps_i=\eps\quad(i=1,\dots,4)

575: \label{cstr3}

576: \eeq

577: in Table~\ref{2gene}, so that again $\lambda=1$.

578: As a consequence, there is a two-parameter family of dynamics

579: with uniform bias and stationary-state

580: measure~(\ref{boltz}),~(\ref{ham2}), parametrized by~$\eps$ and $\delta$.

581: We thus recover the original KLS model~\cite{kls}.

582: The fact that the stationary-state weights are independent of the bias

583: is actually a general property of dynamics obeying P-related pairwise balance

584: (see Section~3.2).

585: 

586: \medskip\noindent$\bullet$

587: {\it {\rm CP}-invariance.}

588: The CP operation is the product of C and P,

589: where the charge conjugation C changes the charge of the particles

590: to its opposite (i.e., interchanges $A$ and~$B$ particles),

591: whereas the spatial parity P

592: changes the orientation of the ring (i.e., interchanges left and right).

593: In physical terms, in the stationary state of a CP-invariant dynamics,

594: the current due to a positively charged particle

595: and to a negatively charged particle are equal.

596: Requiring CP-invariance yields the two conditions

597: \beq

598: \w_{AA}=\w_{BB},\qquad\x_{AA}=\x_{BB},

599: \label{eqcp}

600: \eeq

601: which amount to setting

602: \beq

603: \delta=0,\qquad\eps_1=\eps_2=\eps

604: \label{cstr4}

605: \eeq

606: in Table~\ref{2gene}.

607: The most general CP-invariant dynamics

608: has therefore three free pa\-ra\-me\-ters:~$\eps$, $\eps_3$, and $\eps_4$.

609: It does not obey pairwise balance in general.

610: 

611: CP-invariance can be combined with any of the above symmetries:

612: 

613: \smallskip\noindent $\star$

614: The CP-invariant symmetric dynamics corresponds to $\delta=\eps_i=0$.

615: It has a single free parameter: $\lambda$.

616: The rates read

617: \beq

618: \matrix{

619: \haut\w_{AA}=\x_{AA}=\frad{1}{4},\hfill\quad

620: &\w_{AB}=\x_{BA}=\frad{\lambda}{2({\rm e}^{4J}+1)},\hfill\cr

621: \haut\w_{BA}=\x_{AB}=\frad{\lambda\,{\rm e}^{4J}}{2({\rm e}^{4J}+1)},\hfill\quad

622: &\w_{BB}=\x_{BB}=\frad{1}{4}.\hfill

623: }

624: \eeq

625: 

626: \smallskip\noindent $\star$

627: The partially asymmetric CP-invariant dynamics with uniform bias

628: has one single free parameter: $\eps$.

629: 

630: \smallskip\noindent $\star$

631: The totally asymmetric CP-invariant dynamics

632: is the most constrained of all the dynamics: it has no free parameter at all.

633: The rates

634: \beq

635: \w_{AA}=\w_{BB}=\frac{1}{2},\qquad\w_{AB}=\frac{1}{{\rm e}^{4J}+1},\qquad

636: \w_{BA}=\frac{{\rm e}^{4J}}{{\rm e}^{4J}+1}

637: \label{resascp}

638: \eeq

639: only depend on the energy difference $\D\H$ involved in the exchange moves.

640: They coincide with those of the heat-bath rule~\cite{binder,krauth}:

641: \beq

642: w(\D\H)=\frac{1}{\exp(\D\H)+1}

643: =\frac{1}{2}\left(1-\tanh\frac{\D\H}{2}\right).

644: \label{hb}

645: \eeq

646: 

647: The above discussion is summarized in Table~\ref{2summary},

648: giving the number of free parameters

649: for every symmetry class of dynamics, both without and with imposing

650: CP-invariance.

651: 

652: \begin{table}[ht]

653: \caption{

654: List of the symmetry classes of dynamics for two species of particles

655: with Gibbs stationary-state measure~(\ref{boltz}),~(\ref{ham2}),

656: with balance property: detailed balance (DB) or pairwise balance (PB),

657: and number of free parameters, both without and with CP-invariance.}

658: \label{2summary}

659: \begin{center}

660: \begin{tabular}{|l|c|c|c|}

661: \hline

662: Class of dynamics&Balance property&Without CP&With CP\\

663: \hline

664: General&none&5&3\\

665: Symmetric (equilibrium)&DB&2&1\\

666: Totally asymmetric&PB&1&0\\

667: Partially asymmetric (uniform bias)&PB&2&1\\

668: \hline

669: \end{tabular}

670: \end{center}

671: \end{table}

672: 

673: \section{Three species}

674: 

675: Consider again a finite ring of $N$ sites.

676: Each site is now occupied by a particle which can be

677: either of type $A$ (positively charged), of type $B$ (negatively charged),

678: or of type~$C$ (neutral, i.e., with no charge).

679: We again represent the species of particle at site $n$

680: by a spin $S_n=0,\pm1$ equal to the charge of the particle,

681: as shown in Table~\ref{3s}.

682: 

683: \begin{table}[ht]

684: \caption{Spin (charge) and indicator variables

685: associated with each particle species in the case of three species.}

686: \label{3s}

687: \begin{center}

688: \begin{tabular}{|c|c|c|}

689: \hline

690: Species at site $n$&Spin (charge)&Indicator variable\\

691: \hline

692: $A$&$S_n=+1$&$S_n(S_n+1)/2$\\

693: $B$&$S_n=-1$&$S_n(S_n-1)/2$\\

694: $C$&$S_n=0$&$1-S_n^2$\\

695: \hline

696: \end{tabular}

697: \end{center}

698: \end{table}

699: 

700: We consider Gibbs measures corresponding to the most general

701: (ferromagnetic or antiferromagnetic) Hamiltonian involving pairs of identical

702: nearest neighbors:

703: \beq

704: \H=-2(J_AN_{AA}+J_BN_{BB}+J_CN_{CC}),

705: \label{ham3}

706: \eeq

707: where the coupling constants $J_A$, $J_B$, and $J_C$ can take both signs.

708: The factor 2 is introduced for consistency with the case of two species.

709: Using the spin variables $S_n=0,\pm1$ defined in Table~\ref{3s},

710: the Hamiltonian~(\ref{ham3}) can be rewritten as

711: \beq

712: \fl\H=E_0-\case{1}{2}\sum_n

713: \Bigl[(J_A+J_B+4J_C)S_nS_{n+1}+(J_A-J_B)(S_n+S_{n+1})+J_A+J_B\Bigr]

714: S_nS_{n+1},\;

715: \label{hambeg}

716: \eeq

717: where $E_0=2(N-2N_C)J_C$ is a constant.

718: This is a generalized Blume-Emery-Griffiths spin-1 Hamiltonian~\cite{beg}.

719: 

720: We again address the question of the existence of nonequilibrium stochastic

721: dynamics whose stationary-state measure is the measure~(\ref{boltz})

722: associated with the Hamiltonian~(\ref{ham3}).

723: The results obtained in Section~2 for two species of particles suggest

724: that the case of totally asymmetric exchange dynamics is already of interest.

725: We therefore restrict our investigation to this limiting situation

726: for the time being.

727: The positively (resp.~negatively) charged~$A$ particles

728: (resp.~$B$ particles) only hop to the right (resp.~to the left),

729: whereas the neutral~$C$ particles can hop in both directions.

730: The exchange rates depend on the two neighbors

731: of the pair to be exchanged, according to Table~\ref{3d}.

732: The dynamics so defined conserves the numbers $N_A$, $N_B$, and $N_C$

733: of particles of each species, with $N_A+N_B+N_C=N$.

734: 

735: \begin{table}[ht]

736: \caption{List of moves in the totally asymmetric dynamics

737: for three species of interacting particles,

738: notation for the corresponding exchange rates,

739: and energy difference $\D\H$ involved in the moves,

740: where the Hamiltonian $\H$ is defined in~(\ref{ham3}).}

741: \label{3d}

742: \begin{center}

743: \begin{tabular}{|l|c|c||c|c|c|}

744: \hline

745: Move&Rate&$\D\H$&Move&Rate&$\D\H$\\

746: \hline

747: $AABA\to ABAA$&$\w_{AA}$&$0$&

748: $BCBC\to BBCC$&$\x_{BC}$&$-2J_B-2J_C$\\

749: $AABB\to ABAB$&$\w_{AB}$&$2J_A+2J_B$&

750: $CCBA\to CBCA$&$\x_{CA}$&$2J_C$\\

751: $AABC\to ABAC$&$\w_{AC}$&$2J_A$&

752: $CCBB\to CBCB$&$\x_{CB}$&$2J_B+2J_C$\\

753: $BABA\to BBAA$&$\w_{BA}$&$-2J_A-2J_B$&

754: $CCBC\to CBCC$&$\x_{CC}$&$0$\\

755: $BABB\to BBAB$&$\w_{BB}$&$0$&

756: $AACA\to ACAA$&$\y_{AA}$&$0$\\

757: $BABC\to BBAC$&$\w_{BC}$&$-2J_B$&

758: $AACB\to ACAB$&$\y_{AB}$&$2J_A$\\

759: $CABA\to CBAA$&$\w_{CA}$&$-2J_A$&

760: $AACC\to ACAC$&$\y_{AC}$&$2J_A+2J_C$\\

761: $CABB\to CBAB$&$\w_{CB}$&$2J_B$&

762: $BACA\to BCAA$&$\y_{BA}$&$-2J_A$\\

763: $CABC\to CBAC$&$\w_{CC}$&$0$&

764: $BACB\to BCAB$&$\y_{BB}$&$0$\\

765: $ACBA\to ABCA$&$\x_{AA}$&$0$&

766: $BACC\to BCAC$&$\y_{BC}$&$2J_C$\\

767: $ACBB\to ABCB$&$\x_{AB}$&$2J_B$&

768: $CACA\to CCAA$&$\y_{CA}$&$-2J_A-2J_C$\\

769: $ACBC\to ABCC$&$\x_{AC}$&$-2J_C$&

770: $CACB\to CCAB$&$\y_{CB}$&$-2J_C$\\

771: $BCBA\to BBCA$&$\x_{BA}$&$-2J_B$&

772: $CACC\to CCAC$&$\y_{CC}$&$0$\\

773: $BCBB\to BBCB$&$\x_{BB}$&$0$&&&\\

774: \hline

775: \end{tabular}

776: \end{center}

777: \end{table}

778: 

779: \subsection{The CP-invariant case}

780: 

781: Motivated by the form of the results of Section~2 on two species,

782: we first consider the CP-invariant case,

783: which can be anticipated to be simpler than the generic one.

784: 

785: As far as statics is concerned, C-invariance implies

786: \beq

787: J_A=J_B=J,\qquad J_C=J_0.

788: \label{jcp}

789: \eeq

790: The Hamiltonian~(\ref{hambeg})

791: becomes the usual Blume-Emery-Griffiths Hamiltonian~\cite{beg}

792: \beq

793: \H=E_0-\sum_n\Bigl[(J+2J_0)S_nS_{n+1}+J\Bigr]S_nS_{n+1}.

794: \label{hambegc}

795: \eeq

796: As far as dynamics is concerned, CP-invariance yields

797: 12 equalities among the 27 exchange rates:

798: \beq

799: \matrix{

800: \w_{AA}=\w_{BB},\quad\hfill

801: &\w_{AC}=\w_{CB},\quad\hfill

802: &\w_{BC}=\w_{CA},\quad\hfill

803: &\x_{AA}=\y_{BB},\hfill\cr

804: \x_{AB}=\y_{AB},\hfill

805: &\x_{AC}=\y_{CB},\hfill

806: &\x_{BA}=\y_{BA},\hfill

807: &\x_{BB}=\y_{AA},\hfill\cr

808: \x_{BC}=\y_{CA},\hfill

809: &\x_{CA}=\y_{BC},\hfill

810: &\x_{CB}=\y_{AC},\hfill

811: &\x_{CC}=\y_{CC}.\hfill\cr

812: }

813: \label{3cpcd}

814: \eeq

815: 

816: The analysis follows the lines of Section~2.

817: The algebra is however far more cum\-ber\-some,

818: so that intermediate expressions are too lengthy to be reported here.

819: Calculations have been worked out with the help of the software MACSYMA.

820: We start from the expressions for the total rates

821: $W_\out(\C)$ and $W_\in(\C)$ for a generic configuration $\C$,

822: similar to~(\ref{wout}) and~(\ref{win}),

823: which can be read off from Table~\ref{3d}.

824: The difference $W_\out(\C)-W_\in(\C)$

825: is then recast in terms of products of the spin variables~$S_n$.

826: We thus obtain an expression similar to~(\ref{dw}),

827: involving 42 different correlations of two to eight spin variables.

828: One example of a correlation of two variables is $\mean{S_1S_2}$,

829: whereas there is a unique correlation of eight variables:

830: $\mean{S_1^2S_2^2S_3^2S_4^2}$.

831: Requiring that the coefficients of all these correlations vanish,

832: we thus obtain 42 linear (but not independent) relations of the form

833: \beq

834: R_1=\dots=R_{42}=0,

835: \label{r42}

836: \eeq

837: where the $R_i$ are linear combinations of the exchange rates,

838: similar to $R_1$ and $R_2$ given in~(\ref{r1r2}).

839: In the CP-invariant situation under study,

840: equations~(\ref{3cpcd}) and~(\ref{r42}) together

841: yield 26 independent linear relations among the 27 exchange rates,

842: and therefore leave a single free parameter,

843: which can be fixed by choosing the time unit.

844: For consistency with the case of two species, we set

845: \beq

846: \w_{AA}+\w_{BB}=1.

847: \label{3nor}

848: \eeq

849: This normalization condition uniquely determines all the exchange rates.

850: 

851: We have therefore shown that there is a single totally asymmetric

852: CP-in\-va\-riant dynamics for three species of interacting particles

853: with stationary-state meas\-ure~(\ref{boltz}),~(\ref{ham3}).

854: This uniquely determined dynamics can be viewed

855: as a non-trivial extension to three species of the result~(\ref{resascp}).

856: The explicit expressions of the exchange rates are given in Table~\ref{3cp}.

857: This dynamics obeys pairwise balance

858: (see equation~(\ref{pb3}) below for the general case).

859: However, at variance with the case of two species,

860: the rates are not of the heat-bath form~(\ref{hb}).

861: 

862: \begin{table}[ht]

863: \caption{Expressions of the rates of

864: the totally asymmetric CP-invariant dynamics for three species.

865: The label $I$ stands for any particle species ($I=A$, $B$,~$C$).}

866: \label{3cp}

867: \begin{center}

868: \begin{tabular}{|c|c||c|c||c|c|}

869: \hline

870: Rate&expression&Rate&expression&Rate&expression\\

871: \hline

872: \haut$\w_{AA}=\w_{BB}$&$\frad{1}{2}$&

873: $\w_{BA}$&$\frad{{\rm e}^{4J}}{{\rm e}^{4J}+1}$&

874: $\x_{IA}=\y_{BI}$&$\frad{{\rm e}^{2J}}{2({\rm e}^{4J}+1)}$\\

875: \haut$\w_{AB}$&$\frad{1}{{\rm e}^{4J}+1}$&

876: $\w_{BC}=\w_{CA}$&$\frad{{\rm e}^{2J}({\rm e}^{2J}+1)}{2({\rm e}^{4J}+1)}$&

877: $\x_{IB}=\y_{AI}$&$\frad{1}{2({\rm e}^{4J}+1)}$\\

878: \haut$\w_{AC}=\w_{CB}$&$\frad{{\rm e}^{2J}+1}{2({\rm e}^{4J}+1)}$&

879: $\w_{CC}$&$\frad{{\rm e}^{2J}}{{\rm e}^{4J}+1}$&

880: $\x_{IC}=\y_{CI}$&$\frad{{\rm e}^{2J+2J_0}}{2({\rm e}^{4J}+1)}$\\

881: \hline

882: \end{tabular}

883: \end{center}

884: \end{table}

885: 

886: \subsection{The general case}

887: 

888: We now turn to the general totally asymmetric dynamics.

889: We view $J_A$, $J_B$, and $J_C$ as three independent coupling constants,

890: and consider the 27 exchange rates entering Table~\ref{3d}

891: as being a priori all different from each other.

892: 

893: Following the above procedure,

894: and choosing time units according to~(\ref{3nor}),

895: we are left after some lengthy algebra with a two-parameter family of dynamics

896: with stationary-state measure~(\ref{boltz}),~(\ref{ham3}).

897: An explicit parametrization of the rates is given in Table~\ref{3gene},

898: where the parameters $\alpha$ and $\beta$ enter linearly, and with the notation

899: \beq

900: f=\frac{1}{2({\rm e}^{2J_A+2J_B}+1)}.

901: \label{fdef}

902: \eeq

903: The form of the CP-invariant rates given in Table~\ref{3cp} has been helpful

904: in working out this parametrization of the general case.

905: 

906: \begin{table}[ht]

907: \caption{Explicit parametrization of rates

908: of the most general totally asymmetric dynamics for three species.

909: The notation $f$ is defined in~(\ref{fdef}).}

910: \label{3gene}

911: \begin{center}

912: \begin{tabular}{|c|c||c|c|}

913: \hline

914: Rate&expression&Rate&expression\\

915: \hline

916: $\w_{AA}$&$(1-\al+(1+\al)\eab)f$&

917: $\x_{BC}$&$(1+\al+\be)\ebc f$\\

918: $\w_{AB}$&$2f$&

919: $\x_{CA}$&$((1+\al)\eb+\be)f$\\

920: $\w_{AC}$&$(1-\al+(1+\al)\eb)f$&

921: $\x_{CB}$&$(1+\al+\be)f$\\

922: $\w_{BA}$&$2\eab f$&

923: $\x_{CC}$&$((1+\al)\ebc+\be)f$\\

924: $\w_{BB}$&$(1+\al+(1-\al)\eab)f$&

925: $\y_{AA}$&$(1-\al-\be\eac)f$\\

926: $\w_{BC}$&$(1+\al+(1-\al)\ea)\eb f$&

927: $\y_{AB}$&$(1-\al-\be\ec)f$\\

928: $\w_{CA}$&$(1-\al+(1+\al)\eb)\ea f$&

929: $\y_{AC}$&$(1-\al-\be)f$\\

930: $\w_{CB}$&$(1+\al+(1-\al)\ea)f$&

931: $\y_{BA}$&$(1-\al-\be\ec)\ea f$\\

932: $\w_{CC}$&$((1-\al)\ea+(1+\al)\eb)f$&

933: $\y_{BB}$&$((1-\al)\ea-\be\ec)f$\\

934: $\x_{AA}$&$((1+\al)\eb+\be\ec)f$&

935: $\y_{BC}$&$((1-\al)\ea-\be)f$\\

936: $\x_{AB}$&$(1+\al+\be\ec)f$&

937: $\y_{CA}$&$(1-\al-\be)\eac f$\\

938: $\x_{AC}$&$((1+\al)\eb+\be)\ec f$&

939: $\y_{CB}$&$((1-\al)\ea-\be)\ec f$\\

940: $\x_{BA}$&$(1+\al+\be\ec)\eb f$&

941: $\y_{CC}$&$((1-\al)\eac-\be)f$\\

942: $\x_{BB}$&$(1+\al+\be\ebc)f$&&\\

943: \hline

944: \end{tabular}

945: \end{center}

946: \end{table}

947: 

948: The two parameters $\al$ and $\be$ run over some domain~$D$,

949: such that all the rates of Table~\ref{3gene} are positive.

950: It can be checked that~$D$ is an asymmetric quadrilateral,

951: shown schematically in Figure~\ref{figquadra}.

952: The co-ordinates of its vertices read

953: \beq

954: \matrix{

955: \haut\al_\K=-1,\quad\hfill&

956: \al_\L=-\frad{L_A+S_B}{L_A-S_B},\quad\hfill&

957: \al_\M=1,\quad\hfill&

958: \al_\N=\frad{L_B+S_A}{L_B-S_A},\hfill\cr

959: \haut\be_\K=0,\hfill&

960: \be_\L=\frad{2}{L_A-S_B},\hfill&

961: \be_\M=0,\hfill&

962: \be_\N=-\frad{2}{L_B-S_A},\hfill

963: }

964: \eeq

965: where $L_A$ (resp.~$S_A$) is the largest (resp.~the smallest)

966: of the three quantities

967: $\exp(2J_A+2J_C)$, $\exp(-2J_A+2J_C)$, and $\exp(-2J_A-2J_C)$,

968: and $L_B$ (resp.~$S_B$) is the largest (resp.~the smallest)

969: of the three quantities

970: $\exp(2J_B+2J_C)$, $\exp(-2J_B+2J_C)$, and $\exp(-2J_B-2J_C)$.

971: 

972: \begin{figure}

973: \begin{center}

974: \includegraphics[angle=90,width=.5\linewidth]{figquadra.eps}

975: \caption{\small

976: Typical shape of the quadrilateral domain $D$ of the $\al$-$\be$ plane

977: such that all the rates of Table~\ref{3gene} are positive (see text).}

978: \label{figquadra}

979: \end{center}

980: \end{figure}

981: 

982: The general dynamics of Table~\ref{3gene} contains as special cases

983: several of the situations considered so far.

984: The CP-invariant situation corresponds to $\al=\be=0$.

985: The rates of Table~\ref{3cp} are thus recovered,

986: with the notations~(\ref{jcp}).

987: The case of two species is also recovered.

988: The rates~(\ref{resas}) are reproduced,

989: again with the notations~(\ref{jcp}), and with the identification

990: \beq

991: \delta=\frad{{\rm e}^{4J}-1}{{\rm e}^{4J}+1}\,\al.

992: \eeq

993: 

994: The family of dynamics described in Table~\ref{3gene}

995: obeys P-related pairwise balance.

996: This property is expressed by the nine relations

997: \beq

998: {\hskip -16pt}\matrix{

999: \w_{BA}=\eab\w_{AB},\quad\hfill

1000: &\w_{CA}=\ea\w_{AC},\hfill

1001: &\w_{BC}=\eb\w_{CB},\hfill\cr

1002: \x_{BA}=\eb\x_{AB},\hfill

1003: &\x_{AC}=\ec\x_{CA},\hfill

1004: &\x_{BC}=\ebc\x_{CB},\hfill\cr

1005: \y_{BA}=\ea\y_{AB},\hfill

1006: &\y_{CA}=\eac\y_{AC},\quad\hfill

1007: &\y_{CB}=\ec\y_{BC},\hfill\cr

1008: }

1009: \label{pb3}

1010: \eeq

1011: which are identically fulfilled by the rates of Table~\ref{3gene},

1012: for any values of the parameters~$\al$ and $\be$.

1013: In other words, the relations~(\ref{pb3})

1014: are built in as a subset of~(\ref{r42}).

1015: 

1016: Pairwise balance has the following consequence.

1017: Consider the partially asymmetric dynamics with uniform bias $p$,

1018: where all the `right' moves, i.e, those of Table~\ref{3d},

1019: take place with rates equal to $p$ times those given in Table~\ref{3d},

1020: whereas the P-related `left' moves take place

1021: with rates equal to $q$ times those given in Table~\ref{3d}, with $q=1-p$.

1022: For example:

1023: \beqa

1024: &&AABC\to ABAC\quad\hbox{with rate}\ \ pw_{AC},\nonumber\\

1025: &&CBAA\to CABA\quad\hbox{with rate}\ \ qw_{AC}.

1026: \eeqa

1027: For this uniformly biased dynamics, the total entrance and exit rates

1028: for a given configuration read

1029: \beqa

1030: &&W_\in(\C)=pW^\rig_\in(\C)+qW^\lef_\in(\C),\nonumber\\

1031: &&W_\out(\C)=pW^\rig_\out(\C)+qW^\lef_\out(\C),

1032: \eeqa

1033: with self-explanatory notations.

1034: The stationarity condition for the totally asymmetric dynamics $(p=1)$

1035: reads $W^\rig_\in(\C)=W^\rig_\out(\C)$.

1036: On the other hand, P-related pairwise balance implies

1037: $W^\rig_\in(\C)=W^\lef_\out(\C)$ and $W^\rig_\out(\C)=W^\lef_\in(\C)$.

1038: We have therefore

1039: \beq

1040: W_\in(\C)=W^\rig_\in(\C)=W^\lef_\in(\C)

1041: =W_\out(\C)=W^\rig_\out(\C)=W^\lef_\out(\C).

1042: \eeq

1043: These equations show that the partially asymmetric dynamics

1044: with uniform bias $p$ has the same stationary-state measure

1045: as the totally asymmetric one.

1046: This dynamics interpolates between the symmetric (equilibrium) case ($p=1/2$)

1047: and the totally asymmetric one ($p=1$).

1048: The fact that the stationary-state measure is independent of the bias $p$

1049: thus appears as a general consequence of P-related pairwise balance.

1050: 

1051: \section{Discussion}

1052: 

1053: In this paper we explicitly constructed classes of nonequilibrium dynamics

1054: for two and three species of interacting particles,

1055: i.e., asymmetric stochastic dynamics which do not obey detailed balance,

1056: but whose nonequilibrium stationary-state measure is a prescribed measure.

1057: We have chosen to work with finite-temperature canonical Gibbs measures

1058: associated with spin Hamiltonians with nearest-neighbor interactions.

1059: The stationary current, as well as many other observables in the stationary

1060: state, can therefore be evaluated, at least in principle,

1061: by means of the transfer-matrix formalism.

1062: We have emphasized the role of the various symmetries

1063: which can be imposed onto the dynamics.

1064: 

1065: For two species of interacting particles,

1066: a situation first considered by KLS~\cite{kls},

1067: stationary-state measures are associated with

1068: the usual (anti)ferromagnetic Hamiltonian on the spin-1/2 Ising chain.

1069: Our result for the most general dynamics is given by Table~\ref{2gene}.

1070: This dynamics has five free parameters,

1071: and does not obey pairwise balance in general.

1072: Only the cases considered by KLS, namely the totally asymmetric dynamics

1073: and the partially asymmetric dynamics with a uniform bias,

1074: obey P-related pairwise balance,

1075: where pairs of conjugate moves are related by parity.

1076: 

1077: We then turned to the novel situation of three species of interacting

1078: particles.

1079: Stationary-state measures are given by the most general Hamiltonian

1080: involving pairs of neighboring particles of the same species.

1081: This translates into a Blume-Emery-Griffiths spin-1 Hamiltonian.

1082: We first restricted the search of dynamics

1083: to the totally asymmetric case.

1084: The most general situation is described by Table~\ref{3gene}.

1085: This dynamics obeys pairwise balance.

1086: It can therefore be extended to a partially asymmetric dynamics

1087: with uniform bias $p$.

1088: The three-parameter family of dynamics thus obtained

1089: interpolates between the symmetric (equilibrium) case ($p=1/2$)

1090: and the totally asymmetric one ($p=1$).

1091: 

1092: The most constrained class of stochastic dynamics

1093: we have investigated is the CP-invariant totally asymmetric one.

1094: For a prescribed stationary-state measure,

1095: there is indeed a uniquely defined such dynamics,

1096: with no free parameter, both for two species (see equation~(\ref{resascp}))

1097: and for three species (see Table~\ref{3cp}) of interacting particles.

1098: 

1099: Throughout this work we have put a strong emphasis on the numbers

1100: of free parameters in symmetric and asymmetric dynamics

1101: leading to a given stationary-state measure.

1102: Our results suggest the following rule:

1103: asymmetric stochastic dynamics leading to a given nonequilibrium

1104: stationary-state measure are far more constrained

1105: than symmetric dynamics leading to the same measure as an equilibrium measure.

1106: 

1107: To close up, let us demonstrate that the above empirical rule

1108: holds in a much broader class of stochastic models.

1109: To do so, we have chosen to put the results of this work

1110: in perspective with the following two characteristic examples,

1111: which also belong to the realm of driven diffusive systems.

1112: 

1113: \medskip\noindent {\it Example~1.}

1114: 

1115: Our first example is much in the spirit of the present paper.

1116: Consider a driven diffusive system

1117: consisting of $K$ species of non-interacting particles on a ring,

1118: denoted by $I=A,B,\dots$, where $K\ge2$ is arbitrary.

1119: 

1120: The most general exchange dynamics is defined by the $K(K-1)$ rates~$\u_{IJ}$

1121: corresponding to the moves $IJ\to JI$ for $I\neq J$.

1122: We look for dynamics such that the stationary-state measure is uniform,

1123: i.e., all the configurations with given particle numbers~$N_I$ of each species

1124: are equally probable.

1125: This is indeed the right concept for a Gibbs measure

1126: in the absence of interactions,

1127: or, equivalently, in the limit of an infinite temperature.

1128: 

1129: The condition for having a uniform stationary-state measure reads

1130: \beq

1131: W_\out(\C)-W_\in(\C)=\sum_{IJ}\u_{IJ}(N_{IJ}-N_{JI})=0

1132: \label{pdif}

1133: \eeq

1134: for every configuration $\C$.

1135: The number of independent conditions on the rates imposed

1136: by this equation can be evaluated as follows.

1137: There are $K^2$ numbers of oriented pairs~$N_{IJ}$, which obey the sum rules

1138: \beq

1139: \sum_JN_{IJ}=\sum_JN_{JI}=N_I,\qquad\sum_IN_I=N.

1140: \eeq

1141: Only $(K-1)^2$ pair numbers are therefore linearly independent.

1142: Equation~(\ref{pdif}) shown that each of the $(K-1)(K-2)/2$

1143: antisymmetric combinations of these independent pair numbers

1144: yields one condition.

1145: The $K(K-1)$ rates therefore obey $(K-1)(K-2)/2$ conditions,

1146: so that the general asymmetric exchange dynamics for $K$ species

1147: of non-interacting particles depends on

1148: \beq

1149: A_K=\case{1}{2}(K-1)(K+2)

1150: \eeq

1151: dimensionful parameters.

1152: On the other hand, for the symmetric exchange dynamics

1153: obeying the detailed balance property $\u_{IJ}=\u_{JI}$,

1154: the $K(K-1)/2$ rates are not constrained at all.

1155: Indeed~(\ref{pdif}) vanishes identically.

1156: The general symmetric exchange dynamics therefore depends on

1157: \beq

1158: S_K=\case{1}{2}K(K-1)

1159: \eeq

1160: parameters.

1161: One has

1162: \beq

1163: A_K=S_K+K-1.

1164: \label{dif}

1165: \eeq

1166: 

1167: For two species ($K=2$), we have $A_2=2$ and $S_2=1$.

1168: There is no condition on the exchange rates,

1169: because there exists no antisymmetric combination of pair numbers.

1170: Equation~(\ref{pairex}) indeed implies $N_{AB}=N_{BA}$.

1171: As a consequence, the stationary-state measure

1172: is uniform for any value of the rates $u_{AB}$ and $u_{BA}$.

1173: Interpreting $A$ particles as particles and $B$ particles as holes,

1174: we thus recover a known property of the ASEP~\cite{sch,der},

1175: namely that its stationary-state measure is uniform, irrespective of the bias.

1176: 

1177: For three species ($K=3$), we have $A_3=5$ and $S_3=3$.

1178: There is indeed one single antisymmetric combination of pair numbers:

1179: \beq

1180: Q=N_{AB}-N_{BA}=N_{BC}-N_{CB}=N_{CA}-N_{AC}.

1181: \eeq

1182: There is accordingly a single condition on the six exchange rates

1183: for having a uniform stationary-state measure:

1184: \beq

1185: u_{AB}+u_{BC}+u_{CA}=u_{BA}+u_{CB}+u_{AC}.

1186: \label{skew}

1187: \eeq

1188: This condition is known in the context of the matrix-product

1189: formalism~\cite{ahr}.

1190: It can be checked that~(\ref{skew}) is fulfilled by the rates

1191: of Table~\ref{3gene} in the absence of interactions ($J_A=J_B=J_C=0$).

1192: The only non-zero rates indeed read $u_{AB}=w_{IJ}=1/2$,

1193: $u_{CB}=x_{IJ}=(1+\al+\be)/4$, and $u_{AC}=y_{IJ}=(1-\al-\be)/4$,

1194: irrespective of $I,J$.

1195: In the CP-invariant case, one has $u_{AB}=1/2$ and $u_{CB}=u_{AC}=1/4$.

1196: 

1197: Finally, for a large number of species ($K\gg1)$, asymmetric (driven,

1198: nonequilibrium)

1199: dynamics are far more constrained that symmetric (equilibrium) ones.

1200: Indeed the condition of having a uniform stationary-state measure

1201: roughly cuts off

1202: half the parameters, reducing their number from $K(K-1)$ to $A_K\approx K^2/2$,

1203: whereas for symmetric (equilibrium)

1204: dynamics the $K(K-1)/2$ rates are not constrained.

1205: The expression~(\ref{dif}) shows that the difference $A_K-S_K\approx K\ll S_K$

1206: is relatively negligible for a large number of species.

1207: In other words, for the uniform stationary-state measure,

1208: the full space of nonequilibrium dynamics

1209: is hardly larger than the subspace of equilibrium dynamics.

1210: 

1211: \medskip\noindent {\it Example~2.}

1212: 

1213: Our second example still belongs to the realm of driven diffusive systems,

1214: albeit with multiple occupancies.

1215: The results below strengthen our conclusion

1216: and broaden its range of applicability.

1217: 

1218: Consider the class of dynamical urn models defined as follows.

1219: $N$ particles are distributed among $M$ sites around a ring,

1220: with multiple occupancies.

1221: Let $N_m$ be the number of particles at site $m=1,\dots,M$.

1222: The system is subjected to the following stochastic dynamics.

1223: 

1224: \noindent (i) a departure site $d$ is chosen uniformly at random.

1225: 

1226: \noindent (ii) a neighboring arrival site $a$ is chosen as

1227: the right neighboring site ($a=d+1$) with probability~$p$,

1228: or the left neighboring site ($a=d-1$) with probability $q=1-p$.

1229: 

1230: \noindent (iii) a particle is transferred from site $d$ to site $a$ at a rate

1231: $W_{kl}$ which only depends on the occupancies $k=N_d$ and $l=N_a$

1232: of the two sites.

1233: 

1234: The relevant question in the present context is the following one.

1235: Under which conditions on the rates $W_{kl}$

1236: is the stationary-state measure a product measure of the form

1237: \beq

1238: P(\C)=P(N_1,\dots,N_M)=\frac{1}{Z_{M,N}}\ p_{N_1}\dots p_{N_M}

1239: \ \delta(N_1+\cdots+N_M,N)?

1240: \label{fss}

1241: \eeq

1242: The answer to this question is known~\cite{cocozza} (see~\cite{lux}

1243: for a simple presentation).

1244: Consider first the case of an asymmetric dynamics ($p\ne1/2$).

1245: The stationary-state measure is given by~(\ref{fss})

1246: if and only if the rates $W_{kl}$ obey the two conditions

1247: \beqa

1248: &&p_{k+1}p_lW_{k+1,l}=p_kp_{l+1}W_{l+1,k},\label{m1}\\

1249: &&W_{kl}-W_{k0}=W_{lk}-W_{l0}.\label{m2}

1250: \eeqa

1251: The first condition~(\ref{m1})

1252: relates the rates $W_{kl}$ and the one-site factors $p_k$

1253: of the stationary-state measure distribution.

1254: The meaning of this relation is clear:

1255: it just expresses P-related pairwise balance.

1256: The second condition~(\ref{m2}), which does not involve the $p_k$,

1257: is therefore more `kinematic' than `dynamical' in essence.

1258: 

1259: The zero-range process (ZRP) corresponds to the particular case

1260: where the rates $W_{kl}=u_k$ only depend on the occupation of the departure

1261: site.

1262: The condition~(\ref{m2}) is then automatically satisfied,

1263: whereas~(\ref{m1}) yields the following relation between the rates $u_k$

1264: and the factors $p_k$:

1265: \beq

1266: u_k=\omega\;\frac{p_{k-1}}{p_k}

1267: \eeq

1268: for $k\ge1$, where the constant $\omega$ fixes the time unit.

1269: 

1270: The most general dynamical urn model with stationary-state product measure

1271: is hardly more general than the ZRP.

1272: Let us state the following result, skipping the proof.

1273: For a given product measure of the form~(\ref{fss}),

1274: with prescribed factors $p_k$,

1275: the general solution of~(\ref{m1}) and~(\ref{m2})

1276: is entirely determined by the one-dimensional array of rates $\al_k=W_{k0}$.

1277: One has indeed (with $\al_0=0$)

1278: \beq

1279: W_{kl}=\frac{1}{p_kp_l}\sum_{m=0}^lp_{k+m}p_{l-m}(\al_{k+m}-\al_{l-m}).

1280: \eeq

1281: The rates $\al_k$ are the rates at which an empty site ($l=0$) is refilled,

1282: by receiving one particle from a non-empty neighboring site

1283: containing $k\ge1$ particles.

1284: 

1285: In the case of a symmetric dynamics ($p=1/2$),

1286: only the first condition~(\ref{m1}) is requested~\cite{cocozza,lux}.

1287: This relation expresses detailed balance.

1288: The resulting stationary state is therefore an equilibrium state.

1289: The condition~(\ref{m1}) determines the rates $W_{kl}$ for $k>l$

1290: in terms of those for $k\le l$.

1291: 

1292: For a general dynamical urn model, the stationary product measure

1293: thus depends on the one-dimensional array of rates $\al_k$

1294: in the asymmetric case,

1295: and on the two-dimensional array of rates $W_{kl}$ for $1\le k\le l$

1296: in the symmetric case.

1297: 

1298: To sum up, the two above examples of driven diffusive systems

1299: corroborate the picture which emerges from the results of the present work.

1300: Asymmetric (driven, nonequilibrium) stochastic dynamics producing a given

1301: stationary-state measure are far more constrained

1302: (in terms of numbers of free parameters)

1303: than symmetric dynamics producing the same measure as an equilibrium measure.

1304: 

1305: \section*{References}

1306: 

1307: \begin{thebibliography}{99}

1308: 

1309: \bibitem{s}

1310: Spohn H 1991 {\it Large Scale Dynamics of Interacting Particles} (Berlin:

1311: Springer)

1312: 

1313: \bibitem{sz}

1314: Schmittmann B and Zia R K P 1995 {\it Statistical Mechanics of Driven Diffusive

1315: Systems} in {\it Phase Transitions and Critical Phenomena} Vol {\bf 17}

1316: edited by C Domb and J Lebowitz (London: Academic)

1317: 

1318: \bibitem{l}

1319: Liggett T M 1999 {\it Stochastic Interacting Systems: Contact Voter

1320: and Exclusion Processes} (New York: Springer)

1321: 

1322: \bibitem{sch}

1323: Sch\"utz G M 2001 {\it Exactly Solvable Models for Many-body Systems far

1324: from Equilibrium} in {\it Phase Transitions and Critical Phenomena} Vol

1325: {\bf 19} edited by C Domb and J Lebowitz (London: Academic)

1326: 

1327: \bibitem{der}

1328: Derrida B 1998 Phys. Rep. {\bf 301} 65

1329: 

1330: \bibitem{spitzer}

1331: Spitzer F 1970 Adv. Math. {\bf 5} 246

1332: 

1333: \bibitem{ev}

1334: Evans M R and Hanney T 2005 J. Phys. A {\bf 38} R195

1335: 

1336: \bibitem{lux}

1337: Godr\`eche C 2006 in {\it Ageing and the Glass Transition} edited by M Henkel

1338: M Pleimling and R Sanctuary

1339: Lecture Notes in Physics (Berlin: Springer) to be published

1340: (preprint cond-mat/0604276)

1341: 

1342: \bibitem{kls}

1343: Katz S Lebowitz J L and Spohn H 1983 Phys. Rev. B {\bf 28} 1655

1344: \nonum

1345: Katz S Lebowitz J L and Spohn H 1984 J. Stat. Phys. {\bf 34} 497

1346: 

1347: \bibitem{gw}

1348: Gates D J and Westcott M 1988 Proc. R. Soc. London A {\bf 416} 443;~463

1349: 

1350: \bibitem{kampen}

1351: van Kampen N G 1992 {\it Stochastic Processes in Physics and Chemistry}

1352: (Amsterdam: North-Holland)

1353: 

1354: \bibitem{pairw}

1355: Sch\"utz G M Ramaswamy R and Barma M 1996 J. Phys. A {\bf 29} 837

1356: 

1357: \bibitem{binder}

1358: Binder K and Heermann D W 1997 {\it Monte Carlo Simulation in Statistical

1359: Physics. An Introduction} (New York: Springer)

1360: 

1361: \bibitem{krauth}

1362: Krauth W 2006 {\it Statistical Mechanics: Algorithms and Computations}

1363: (Oxford: Oxford University Press)

1364: 

1365: \bibitem{beg}

1366: Blume M Emery V J and Griffiths R B 1971 Phys. Rev. A {\bf 4} 1071

1367: 

1368: \bibitem{ahr}

1369: Arndt P F Heinzel T and Rittenberg V 1998 J. Phys. A {\bf 31} 833

1370: 

1371: \bibitem{cocozza}

1372: Cocozza-Thivent C 1985 Z. Wahr. {\bf 70} 509

1373: 

1374: \end{thebibliography}

1375: \end{document}

1376: