1: \documentstyle[aps,multicol,epsfig]{revtex}
2: %\documentclass[pre,twocolumn,showpacs]{revtex4}
3:
4: %\usepackage{epsfig}
5:
6: \begin{document}
7:
8: \title{
9: Intrinsic noise-induced phase transitions: beyond the noise interpretation.}
10:
11: \author{O. Carrillo$^1$, M. Iba\~nes$^1$, J. Garc\'{\i}a-Ojalvo$^{2,3}$, J.
12: Casademunt$^1$, J.M. Sancho$^1$}
13:
14: \address
15: {$^1$Departament d'Estructura i Constituents de la Mat\`eria,
16: Universitat de Barcelona, Diagonal 647, E--08028 Barcelona, Spain\\
17: $^2$Departament de F\'{\i}sica i Enginyeria Nuclear,
18: Universitat Polit\`ecnica de Catalunya,
19: Colom 11, E--08222 Terrassa, Spain
20: $^3$Center for Applied Mathematics, Cornell University, 657 Rhodes
21: Hall, Ithaca, New York 14853}
22:
23: \date{\today}
24:
25: \maketitle
26:
27: \begin{abstract}
28: We discuss intrinsic noise effects in stochastic multiplicative-noise partial
29: differential equations, which are qualitatively independent of the noise
30: interpretation (It\^o vs. Stratonovich),
31: in particular in the context of noise-induced ordering phase transitions. We
32: study a model which, contrary to all cases known so far, exhibits such ordering
33: transitions when the noise is interpreted not only according to Stratonovich,
34: but also to It\^o. The main feature of this model is the absence of a linear
35: instability at the transition point. The dynamical properties of the resulting
36: noise-induced growth processes are studied and compared in the two
37: interpretations and with a reference Ginzburg-Landau type model. A detailed
38: discussion of new numerical algorithms used in both interpretations is also
39: presented.
40: \end{abstract}
41:
42: \pacs{05.40.-a, 02.50.Ey, 64.60.Cn}
43:
44: %\maketitle
45:
46: \begin{multicols}{2}
47:
48: \section{Introduction}
49:
50: An important feature of nonlinear systems is their ability to
51: sustain an organized behavior even in the presence of a substantial
52: amount of randomly fluctuating influences. Even more strikingly,
53: systems which in the absence of fluctuations exhibit a disordered
54: behavior can experience, under certain conditions, the emergence of
55: spatiotemporal order upon addition of a suitable amount of noise
56: \cite{nises}. The most basic manifestation of this fact is the
57: existence of ordering phase transitions induced by noise in dynamical
58: systems with spatial degrees of freedom \cite{chris,nio}. These transitions
59: bring the system from a disordered to an ordered phase as the intensity
60: of the noise increases, contrary to naive intuition. By disordered
61: (ordered) phase we mean for example the homogeneous zero (non-zero) state
62: corresponding to the coarse graining of a spin field with random
63: (uniform) orientation.
64:
65: Ordering phase transitions are usually
66: driven by multiplicative noise terms, which depend on the system's
67: variables \cite{notapie}. But the stochastic integrals associated with
68: stochastic
69: differential equations with multiplicative noise are not uniquely
70: defined \cite{vankampen}. Among the many interpretations that can be
71: given to these integrals, two are frequently used: the Stratonovich
72: interpretation which follows the standard rules of calculus,
73: but gives rise to non-intuitive statistical properties of
74: the noise terms, and the It\^o interpretation which avoids these
75: problems, but at the expense of requiring new rules of calculus.
76:
77: Beyond the technical mathematical definitions, the physical implications of
78: both noise prescriptions boil down to an important fact. The Stratonovich
79: prescription for white noise yields the result one would get for a
80: time-correlated noise in the limit of vanishing correlation time. The key point
81: is that, as soon as the noise is slightly correlated, the stochastic variables
82: defined by the corresponding Langevin equation build up correlations with the
83: noise variable at equal time. This immediately implies that the multiplicative
84: noise terms in the equation have a nonzero mean, even with a zero-mean noise.
85: The result is the so-called Stratonovich drift, a net force induced by noise
86: which is at the heart of most noise-induced phenomena, in particular concerning
87: noise-induced ordering transitions.
88:
89: As for a given stochastic differential equation with multiplicative
90: noise the results do depend on the
91: interpretation, a preliminary analysis of the physical problem has to be
92: performed to make a judicious choice. Our experience indicates that there
93: are a minimum of three possible situations:
94: \begin{itemize}
95: \item If we start with a well established deterministic differential equation
96: and some controlled parameter is allowed to fluctuate (experimental or
97: realistic external noise) one would always expect the noise to have a
98: high--frequency cut-off and as a consequence the Stratonovich
99: interpretation is usually argued to be the reasonable choice.
100: \item
101: If the starting scheme is a master equation which is approximated by a
102: Fokker-Planck equation, then one can write a stochastic differential
103: equation with multiplicative noise in the It\^o interpretation.
104: This happens, for instance, in front
105: propagation problems
106: on a lattice \cite{doering1}.
107: \item
108: Moreover quite often our initial scheme is a set of stochastic differential
109: equations, and we would like to simplify the problem eliminating the
110: most irrelevant fast variables (those with a very short time scale). The
111: interpretation of the final stochastic differential equation will depend on
112: the order in which this procedure is performed with respect to the white
113: noise limit. This is indeed a nontrivial task.
114: \end{itemize}
115:
116: Since the Stratonovich drift
117: can drastically modify the behavior of systems, and since it may not always be
118: obvious what the appropriate noise prescription is in a given problem, it is
119: particularly important to distinguish which noise effects are {\em intrinsic},
120: in the sense of occurring regardless the noise
121: interpretation, and which
122: ones are strictly
123: associated to the Stratonovich drift. In other words, it is important to
124: elucidate when the noise interpretation may only affect the quantitative
125: behavior, and when it may indeed change the problem at a qualitative level.
126:
127: For the case of noise-induced phase-transitions, the noise prescription used so
128: far in the literature is that of Stratonovich. Nevertheless, it could be argued
129: that if the noise has an internal origin, one should in principle expect It\^o
130: noise too, so it would be good to establish whether, in the latter case,
131: noise-induced transitions can occur. We will see that this is indeed the case
132: for a recently discovered class of noise--induced phase
133: transitions. From a theoretical
134: point of view, it is also important to deal with It\^o noise since then the
135: continuum white-noise limit is either well defined or less singular than in the
136: Stratonovich case \cite{nualart}. This has important consequences in order
137: to
138: establish when the macroscopic observables will carry out a nontrivial,
139: singular dependence on the spatial cut-off of the noise (Stratonovich case)
140: and
141: when such residual dependence will be weak (It\^o case) \cite{rocco}.
142:
143: Few contributions have appeared in the physics literature on It\^o calculus in
144: extended systems. A comparative discussion about the mathematical problems
145: involved in the two interpretations appeared in \cite{doering}. The role of the
146: multiplicative noise in the It\^o interpretation has been analyzed in the
147: context of spatiotemporal intermittency \cite{palma} and front propagation
148: \cite{anxo}. Dynamical renormalization group calculations were presented in
149: \cite{grinstein}. However, noise-induced ordering phase transitions had been
150: reported so far only in the framework of the Stratonovich interpretation
151: \cite{nises,chris,nio}. In that case, the mechanism underneath these
152: transitions is that the multiplicative noise term has a non-zero average value,
153: which produces a short-time instability of the disordered phase and induces the
154: ordered phase to arise \cite{chris,nio,kramers}. The instability can be linear
155: \cite{chris,marta2} or nonlinear \cite{nio,oliver}, but is in any case induced
156: by the so-called Stratonovich shift. Due to the absence of such a drift, the
157: It\^o interpretation does not present this type of noise-induced ordered phase,
158: or any other spatially ordered state \cite{mamunoz}.
159:
160: Recently, however, a new type of noise-induced phase transition has been found
161: which does not occur via an instability of the disordered phase
162: \cite{prlmarta}. Here, the ordered phase arises due to the balance between the
163: relaxing deterministic forces pushing the system toward the disordered state,
164: and the activating multiplicative fluctuations pulling the field away from that
165: state, in a type of entropy-driven phase transition (EDPT). This behavior is
166: the spatio-temporal extension of noise-induced transitions in purely temporal,
167: zero-dimensional systems, where the probability distribution of the
168: time-dependent variable exhibits a change in the number and type of its extrema
169: as noise intensity varies \cite{horsthemke}. A key idea in the model here
170: studied is that the bimodality in the stationary probability density is not
171: associated to a potential barrier, but has a dynamical origin. In fact, the
172: dependence of the multiplicative noise term on the field is such that, for
173: sufficiently large noise strength, the system escapes more easily from the
174: central region than from the sides, despite the fact that the deterministic
175: force always drives the system towards the center. As a result the peaks of the
176: probability density are off-center. An important difference with the usual
177: bimodality associated to a potential barrier is that in our case the
178: characteristic relaxation time scales for the zero-dimensional model are of
179: order one (${\cal{O}}(\varepsilon^0)$) as opposed to
180: ${\cal{O}}(\exp(1/\varepsilon)$) which is
181: characteristic of activation processes, $\varepsilon$ being a
182: generic measure of the noise strength.
183: In the spatially extended case, the spatial (diffusive) coupling of
184: the field introduces an additional crucial ingredient, namely it freezes the
185: domains impeding the fast relaxation process of the zero-dimensional case. This
186: gives rise to a well-defined, stable interface which then drives the much
187: slower domain-growth dynamics.
188: Since no Stratonovich drift is required to induce this effect (as opposed for
189: instance to the case of Ref.\cite{nises} where it takes the form of an
190: effective
191: barrier) it is to be expected that the corresponding class of model exhibiting
192: this behavior should also display noise-induced ordering in the It\^o
193: interpretation. In this paper we show that this is indeed the case, by
194: comparing the behavior of the model introduced in \cite{prlmarta} for both the
195: It\^o and Stratonovich interpretations with that of a standard Ginzburg-Landau
196: model with multiplicative noise (section II). We also analyze in detail the
197: dynamical properties of the growth processes arising from the noise-induced
198: ordering transitions in the two cases (section III), which will be shown to
199: share universal characteristics (i.e. growth exponents) but differ in
200: non-universal features (such as power-law prefactors). Finally, algorithms that
201: have been specially developed for generating the results presented in this
202: paper, for both the Stratonovich and It\^o interpretations, are described in
203: detail in the Appendix.
204:
205:
206: \section{Theoretical analysis}
207:
208: We will use a model of a class of systems for which the steady-state
209: probability distribution can be obtained {\it exactly}. As a consequence,
210: the existence of a phase transition in this kind of systems can be studied
211: without any dynamical reference.
212:
213: Our model corresponds to a relaxational flow in a free--energy potential
214: ${\cal F}(\{\phi \})$,
215: with a field-dependent kinetic coefficient $\Gamma(\phi)$ and
216: a fluctuating term fulfilling a fluctuation-dissipation relation
217: \cite{prlmarta}. The model is defined by the following
218: stochastic partial differential equation:
219: \begin{equation}
220: \frac{\partial \phi (\vec x,t)}{\partial t} = -
221: \Gamma(\phi(\vec x,t)) \frac{\delta
222: \cal{F}}{\delta \phi(\vec x,t)} + \Gamma(\phi(\vec x,t))^{1/2}
223: \xi(\vec x,t)
224: \end{equation}
225: We suppose that the noise $\xi(\vec x,t)$ is Gaussian, with zero mean and
226: correlation
227: \begin{equation}
228: \langle \xi(\vec x,t) \xi(\vec x',t') \rangle = 2 \sigma^{2}
229: \delta(\vec x - \vec x') \delta(t - t')\,,
230: \end{equation}
231: where $\sigma^{2}$ is the noise intensity.
232: Moreover, we choose the following form for the free-energy potential
233: $\cal{F}$,
234: \begin{equation}
235: {\cal{F}} = \int d^{d}\vec x \left\{ V_{0}(\phi(\vec x,t)) +
236: \frac{D}{4d}
237: \left[
238: \vec{\nabla} \phi (\vec x,t) \right]^{2}
239: \right\}\,.
240: \end{equation}
241:
242: Since we are dealing with spatially uncorrelated
243: noise, we perform the analysis in a discrete space in order to avoid
244: singularities \cite{doering}. In a d--dimensional
245: square lattice of mesh size $\Delta x$ and $N=L^{d}$ cells, our model
246: reads
247: \begin{equation}
248: \frac{d\, \phi_i}{d\,t} = -\Gamma_i \frac{\partial {F}}{\partial \phi_i} +
249: \Gamma_i^{1/2}\xi_i(t)\,,
250: \label{eq:motion}
251: \end{equation}
252: where only one index is used to label the cells, $\phi_i\equiv\phi(\vec x_i)$,
253: $\Gamma_i \equiv \Gamma(\phi_i)$, and the noise satisfies the correlation:
254: \begin{equation}
255: \langle \xi_i(t) \xi_j(t') \rangle = 2 \sigma^{2} \frac{\delta_{ij}}
256: {\Delta x^{d}}\delta(t-t')\,.
257: \label{eq:noiseNdis}
258: \end{equation}
259: In discrete space, the free energy has the form
260: \begin{equation}
261: {F}(\{\phi\}) = \sum_{i=1}^{N} \left [V_0(\phi_i) + \frac{D}{4d\Delta x^{2}}
262: \sum_{j\in nn^{+}(i)}(\phi_j-\phi_i)^{2} \right]\,,
263: \label{freenergyNd}
264: \end{equation}
265: where the gradient term is approximated by the sum over nearest neighbors
266: on the lattice in a standard way,
267: $|\vec \nabla\phi|^{2}\to\sum_{j \in
268: nn^{+}(i)}\frac{(\phi_j-\phi_i)^{2}}{\Delta
269: x^{2}}$
270: , and $nn^{+}(i)$ stands for the $d$--nearest neighbors of $i$ in the
271: positive direction of each axis.
272: For simplicity, we choose a monostable local potential,
273: \begin{equation}
274: V_{0}(\phi) = \frac{a}{2} \phi^{2}\,,
275: \end{equation}
276: where $a>0$.
277: Finally, the kinetic coefficient $\Gamma(\phi)$ is taken to depend on
278: the field in the following way \cite{prlmarta}
279: \begin{equation}
280: \Gamma( \phi ) = \frac{1}{1 + c \phi^{2}}\,.
281: \label{gamma}
282: \end{equation}
283: This functional dependence of the kinetic coefficient favors diffusion
284: due to fluctuations in the disordered state.
285:
286: Our objective now is to study the equation (\ref{eq:motion}) in the
287: Stratonovich and It\^o stochastic interpretations.
288: The corresponding Fokker--Planck equation for the probability
289: density of the field $P(\{\phi\},t)$ can be written in a unified notation
290: for both interpretations \cite{vankampen},
291: \begin{eqnarray}
292: \frac{\partial P}{\partial t} = \sum_i \frac{\partial}{\partial \phi_i}
293: \left[\Gamma_i \frac{\partial {F}}{\partial \phi_i} P +
294: \frac{B \sigma^{2}}{\Delta x^{d}}\Gamma_i^{1/2}
295: \frac{\partial \Gamma_i}{\partial \phi_i} P\right.
296: \nonumber
297: \\
298: \left.+ \frac{\sigma^2}{\Delta x^d}
299: \frac{\partial}{\partial {\phi_i}} \Gamma_i P \right]\,,
300: \label{probd}
301: \end{eqnarray}
302: where $B=1$ for the Stratonovich interpretation and $B=2$ in the It\^o case.
303:
304: If no probability flux is present, the stationary solution $P_{\rm st}$
305: of (\ref{probd}) satisfies
306: \begin{equation}
307: \left(\frac{\partial {F}}{\partial \phi_i}+ \frac{B\sigma^2}{2\Delta x^d}
308: \frac{\partial \ln \Gamma_i}{\partial \phi_i} \right) P_{\rm st} +
309: \frac{\sigma^2}{\Delta x^d} \frac{\partial P_{\rm st}}{\partial \phi_i}=0\,.
310: \label{Stratoprobd3}
311: \end{equation}
312: The solution of this equation is
313: \begin{equation}
314: \label{pstNd}
315: P_{\rm st}(\{\phi\}) \sim {\rm e}^{-{F}_{\rm eff}\Delta x^d/\sigma^2}\,,
316: \end{equation}
317: where we have introduced the effective free energy
318: \begin{equation}
319: {F}_{\rm eff}(\{\phi\}) \equiv {F}(\{\phi\}) + \frac{B\sigma^2}{2\Delta
320: x^d} \sum_{i=1}^N \ln \Gamma_i\,.
321: \label{FeffNd}
322: \end{equation}
323:
324: The above expressions can be written in continuum space as
325: \begin{equation}
326: P_{\rm st}(\{\phi\}) \sim {\rm e}^{-{\cal F}_{\rm eff}/\sigma^2}\,,
327: \qquad\qquad
328: \label{pstNc}
329: \end{equation}
330: \begin{equation}
331: {\cal F^S}_{\rm eff}(\{\phi\}) \equiv {\cal F}(\{\phi\}) +
332: \frac{B{\sigma_0}^2}{2}
333: \int \, d^d x \ln \Gamma(\phi(\vec x))\,,
334: \label{FeffNc}
335: \end{equation}
336: where ${\sigma_0}^2\equiv \sigma^2/\Delta x^d$ stands for the effective
337: noise intensity of a spatially white noise in a discrete space.
338:
339: We have thus seen that the stationary multivariate probability
340: distribution can be obtained exactly in both the It\^o and Stratonovich
341: interpretations
342: for the spatially extended EDPT model, and that both lead to very
343: similar qualitative results. The only difference is an extra factor $2$ in
344: the new term of the effective potential in the It\^o interpretation. As
345: is already known \cite{prlmarta}, the EDPT model presents a continuous
346: ordering noise--induced phase transition in the Stratonovich interpretation.
347: But according to the results shown above, and as will be shown in the
348: following Section,
349: this model also exhibits an {\it ordering} transition in
350: the {\it It\^o} interpretation, although the location of the critical point
351: will be different. We should remark here that, as in the case
352: of the Stratonovich interpretation \cite{prlmarta}, this phase
353: transition is not due to a short-time instability of the homogeneous
354: null phase. Indeed, the linear equation for the first statistical moment
355: $\langle \phi \rangle$ can be computed to be \cite{nises}
356: \begin{equation}
357: \frac{\partial \langle \phi \rangle}{\partial t} = -\left[a +
358: (2-B)\sigma_{0}^2 c\right]
359: \langle \phi \rangle + \frac{D}{2d} \nabla^2 \langle \phi \rangle\,.
360: \label{mean}
361: \end{equation}
362: For $a>0$, the homogeneous null solution of this equation is stable for all
363: noise intensities, both for $B=1$ and $B=2$.
364: Therefore, the mechanism of this phase transition must be different from
365: the standard one.
366:
367: \section{Steady-state behavior}
368:
369: A standard way of determining the existence of a noise-induced phase
370: transition is by applying a mean-field approximation to the Langevin or
371: Fokker-Planck equations of the system \cite{nises,chris}.
372: In the present case, however, since we have obtained the exact multivariate
373: probability distribution in both interpretations,
374: we will implement that approximation directly on the effective potential
375: derived from (\ref{FeffNd}).
376:
377: The mean-field approximation consists in replacing the exact value of
378: the neighbor field in the Langevin or Fokker-Planck equation
379: by a common mean-field value $\langle\phi\rangle$.
380: In the present case, we make such an identification in the neighboring
381: values of the gradient term appearing in the effective free
382: energy [see Eqs.~(\ref{freenergyNd}) and (\ref{FeffNd})]:
383: \begin{equation}
384: \label{mfappgrad}
385: \frac{1}{\Delta x^2}\sum_{j\in nn^+(i)}(\phi_j-\phi_i)^2\approx
386: \frac{2 d}{\Delta x^2}\,(\langle\phi\rangle-\phi_i)^2\,.
387: \end{equation}
388: In this way, the effective free energy becomes
389: \begin{eqnarray}
390: F_{\rm eff}(\left\{\phi\right\},\langle \phi \rangle) =
391: \sum_{i=1}^{N}
392: %\Delta x^{d}
393: \left\{ V_0(\phi_{i}) +
394: \frac{{B\sigma_{0}}^{2}}{2}
395: \ln \Gamma(\phi_{i})
396: \right. \nonumber \\
397: \left. \frac{D}{2 \Delta x^{2}} (\phi_{i} - \langle\phi\rangle)^{2} \right\}
398: \equiv \sum_{i=1}^{N}
399: %\Delta x^{d}
400: V_{\rm eff} (\phi_{i},
401: \langle\phi\rangle).\quad
402: \end{eqnarray}
403: The unknown mean--field value $\langle\phi\rangle$ is obtained
404: self--consistently, according to
405: \begin{equation}
406: \langle\phi\rangle=\int_{-\infty}^{\infty} \phi\,
407: P_{\rm st}(\phi,\langle\phi\rangle)
408: \label{self-consNIT}
409: \end{equation}
410: where the one-site probability distribution ($P_{\rm st}(\{\phi\})
411: =\prod_{i=1}^{N}\, P_{\rm st}(\phi_i)$) is given by
412: \begin{equation}
413: P_{\rm st}(\phi)\sim {\rm e}^{-V_{\rm eff}/{\sigma_0}^{2}}\,.
414: \end{equation}
415: The mean-field predictions for $\langle\phi\rangle$ in the
416: two-dimensional case are plotted in Fig.~\ref{fig:mfentropy2},
417: where lines separating the situations where $\langle\phi\rangle=0$
418: (disorder) and $\langle\phi\rangle\neq 0$ (order) are plotted for
419: both the It\^o and Stratonovich interpretations in the
420: space of parameters $D$ and $\sigma^2$. The figure shows
421: that both interpretations predict a continuous noise--induced
422: ordering phase transition, which occurs earlier (i.e., for lower noise
423: intensities) in the It\^o case. In particular, in the large coupling
424: limit ($D\to\infty$) the transition in the It\^o interpretation takes
425: place at a critical noise intensity ($\sigma^2_c = a/Bc$, for $\Delta x =
426: 1$),
427: that is half the critical value in the Stratonovich case, both of which
428: coincide with the transition point in zero dimensional systems (with
429: noise intensity $\sigma^2/\Delta x^2$) \cite{horsthemke}.
430:
431: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
432: \begin{figure}[htb]
433: \narrowtext
434: \epsfig{file=fig1.eps,width=6.9cm}
435: \caption{
436: Phase diagram of the EDPT model, obtained from a mean-field analysis, in
437: the It\^o (continuous line) and Stratonovich (dashed line)
438: interpretations. The horizontal dotted line corresponds to the value
439: of $D$ used in Fig.~\protect\ref{fig:mfentropy1}.
440: The parameter values are $a=1$, $c=0.5$, and $\Delta x=1$.}
441: \label{fig:mfentropy2}
442: \end{figure}
443: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
444:
445: Note that, in contrast with the usual noise-induced transitions
446: (which exhibit reentrant phenomena), the transition lines of
447: Fig. \ref{fig:mfentropy2}
448: decay monotonously with $\sigma^2$. This implies therefore that no minimum
449: coupling strength is required in these models for a phase transition to
450: occur.
451:
452: In order to validate the results obtained from the mean-field
453: approximation, we have performed extensive numerical simulations
454: of model (\ref{eq:motion})-(\ref{gamma}) in both the It\^o and
455: Stratonovich interpretations. To that end, we have
456: developed a new type of numerical algorithm suitable for the
457: implementation of both stochastic interpretations of the multiplicative
458: noise. The derivation of this algorithm and a comparison with the
459: well-known Heun algorithm (derived only for the Stratonovich interpretation)
460: is presented in the Appendix. The simulations have been performed
461: on a square lattice of $256\times256$ cells of mesh size $\Delta x=1$,
462: with a time step $\Delta t=0.01$ and periodic boundary conditions
463: (except when explicitly indicated). Where necessary, we have averaged
464: over 10 realizations of the noise and the initial random
465: conditions, corresponding to Gaussian or uniform distributions.
466: In order to compute the mean field, we first evaluate the spatial
467: average of the system:
468: \begin{equation}
469: \langle\phi(t)\rangle = \frac {1}{N} \left| \sum_{i=1}^N \phi_{i}(t)
470: \right| \,,
471: \label{avnum}
472: \end{equation}
473: where $N$ is the number of lattice cells, and $\phi_{i}(t)$ is the field
474: value at the $i$ cell. Once the spatial average reaches a stationary state,
475: the temporal average is evaluated as
476: \begin{equation}
477: \langle\phi\rangle = \frac{1}{T_{M} - T_{m}} \sum^{T_{M}}_{t=T_{m}}
478: \langle\phi(t)\rangle\,,
479: \end{equation}
480: where $T_{M}$ and $T_{m}$ delimit the time interval within the steady-state
481: regime in which the temporal average is calculated. Afterwards, the
482: realization average can be computed.
483: %We make all
484: %these computations for a given value of the effective multiplicative noise
485: %intensity, and for a given initial configuration
486: %of our system.
487:
488:
489:
490: The numerical simulation results for the two interpretations are shown in
491: Fig.~\ref{fig:mfentropy1}, where they are also compared with the predictions
492: coming from the mean-field approximation. Due to the value of $D$ chosen,
493: the agreement between the mean-field estimate and the simulations is better
494: for the It\^o interpretation. In any case, the model exhibits a noise-induced
495: ordering phase transition for both interpretations, as predicted by
496: the mean-field approach.
497:
498: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
499: \begin{figure}[htb]
500: \narrowtext
501: \epsfig{file=fig2.eps,width=7cm}
502: \caption{
503: Mean-field and numerical simulation results for the EDPT model in the
504: It\^o (continuous line) and Stratonovich (dashed line) interpretations.
505: Simulations have been performed for different system sizes:
506: $L=16$ (circles), $L=24$ (squares) and $L=32$ (triangles) for It\^o,
507: and $L=64$ (triangles), $L=48$ (diamonds), $L=32$ (squares) and
508: $L=16$ (circles) for Stratonovich. $D=4$, and the rest of
509: parameter values are those of the previous figure.}
510: \label{fig:mfentropy1}
511: \end{figure}
512: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
513:
514: We stress at this point that standard models exhibiting noise-induced
515: phase transitions caused by short-term instabilities of the disordered
516: phase do so only in the case of the Stratonovich interpretation. In
517: order to illustrate this point, we present here for comparison what
518: happens in the well-known case of the Ginzburg-Landau model with external
519: multiplicative fluctuations \cite{nio}:
520: \begin{equation}
521: \frac{d\phi}{dt} = a\phi - b\phi^{3} + D\nabla^{2}\phi +
522: \phi\,\xi(\vec x,t) + \eta(\vec x,t)
523: \label{GLeq}
524: \end{equation}
525: where $\eta(\vec x,t)$ and $\xi(\vec x,t)$ are Gaussian and white noises.
526: This system presents a noise-induced phase transition if we interpret
527: the noise in the Stratonovich sense, but not if one uses the
528: It\^o interpretation. This can be seen in Fig.~\ref{4}, where
529: the two simulations share the same conditions and parameters.
530: In the It\^o interpretation the ordered parameter $\langle\phi\rangle$
531: remains always in the
532: disordered state, due to the fact that the noise-dependent drift
533: that causes the short-time instability is only present in the
534: the Stratonovich prescription \cite{mamunoz}.
535:
536: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
537: \begin{figure}[htb]
538: \narrowtext
539: \centerline{\psfig{figure=fig3.eps,width=8cm,height=7cm}}
540: \caption{
541: Bifurcation diagram of the Ginzburg-Landau model in the
542: It\^o and Stratonovich interpretations. Both mean-field and 2-d simulation
543: results are shown. The parameters used are $L=30$, $a=-0.2$, $b=1$, $D=4$
544: and the additive noise intensity $0.5$.
545: }
546: \label{4}
547: \end{figure}
548: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
549:
550:
551: \section{Domain growth dynamics}
552:
553: We have seen that the EDPT model in the presence of external fluctuations
554: can reach a stationary ordered state described by a non-zero order
555: parameter $\langle\phi\rangle$, for both the It\^o and Stratonovich
556: interpretations. This means that, if the system is initially in a
557: disordered steady state $\langle\phi\rangle=0$
558: corresponding to a small noise intensity,
559: %($\sigma^2<a/Bc$)
560: as the intensity of external fluctuations is increased above its critical
561: value the system develops
562: domains of the two new symmetric stationary ordered phases, that grow
563: with time as shown in Fig.~\ref{3}. The figure shows that the system
564: behaves differently in the two stochastic interpretations for the same
565: noise intensity, the It\^o case being much more contrasted due to the
566: fact that the order parameter is larger than in the Stratonovich case,
567: which is very noisy.
568:
569: In this section we are concerned
570: with the growth of these noise-induced domains.
571: Although the mechanism that induces the phase transition is different from
572: those that have been reported before, we can expect that, once the domains
573: have appeared, their dynamics has the same characteristics as those of the
574: domain growth following the quench of a system below its order-disorder
575: transition temperature, as happens in the Ginzburg--Landau model
576: \cite{marta3}.
577:
578: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
579: \begin{figure}[htb]
580: \narrowtext
581: \begin{center}
582: {\psfig{figure=fig4a.eps,width=4cm,height=4cm}}
583: \hskip0.001cm
584: {\psfig{figure=fig4b.eps,width=4cm,height=4cm}}
585: \\
586: {\psfig{figure=fig4c.eps,width=4cm,height=4cm}}
587: \hskip0.001cm
588: {\psfig{figure=fig4d.eps,width=4cm,height=4cm}}
589: \end{center}
590: \caption{
591: Snapshots of evolving noise-induced domains for the EDPT model
592: at $t=750$ (left figures) and $t=1750$ (right figures) in the It\^o
593: (top) and Stratonovich (bottom) interpretations. Parameters are: $a=1$,
594: $c=3$, $\sigma^{2}=3.5$ and $L=256$.}
595: \label{3}
596: \end{figure}
597: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
598:
599: For non-conserved order parameter models, one of the domains
600: grows until it fills the whole system. The mechanism underlying domain
601: growth in this case is the motion of the interface between domains
602: caused by the interface structure. The translational velocity of the
603: domain boundary has been found to be proportional to the mean curvature
604: of the boundary, and independent of the free energy of the interface.
605: This can be quantified by the equation of
606: motion obeyed by the characteristic length (i.e. the average radius) of the
607: domains of equilibrium phases, $R(t)$, \cite{maxi}
608: \begin{equation}
609: \frac{d R}{d t} = A \frac{\Gamma}{R}\,,
610: \label{CA}
611: \end{equation}
612: where $A$ is a model-dependent constant and $\Gamma$ is the
613: kinetic coefficient multiplying the diffusion term. This expression
614: leads in a straightforward way to the Allen--Cahn law of domain growth:
615: \begin{equation}
616: R(t)\propto \sqrt{2A\Gamma} \,\,t^{1/2}
617: \label{A-Claw}
618: \end{equation}
619: In the time regime where this law is verified, $R(t)$ is the only
620: characteristic length of the system, and a scaling behavior for its
621: spatial structure at different times is found. All these results are
622: known to apply also in the case of standard noise-induced phase
623: transitions caused by linear instabilities of the homogeneous disordered
624: phase \cite{marta2}. We want to find out whether the same thing happens
625: in the EDPT described in this paper.
626:
627: In order to characterize the dynamics of model (\ref{eq:motion}), we
628: let our system evolve from an initial disordered state, and compute
629: the isotropic correlation $G(r,t)$ function
630: at different times. We use the following normalization
631: \begin{equation}
632: g(r,t)= \frac{G(r,t)}{G(0,t)}.
633: \label{normals}
634: \end{equation}
635: Let us consider a time regime in which there is only one characteristic
636: length $R(t)$ in the system, which is related to the average size of
637: the domains. There are several possible definitions for $R(t)$, but
638: all of them should lead to the same results. We have chosen $R(t)$ as
639: the distance at which $g(r,t)$ has half its maximum value. In this time
640: regime, we can apply the scaling hypothesis for a d-dimensional system
641: \begin{equation}
642: g(r,t)=g(r/R(t)),
643: \label{scaling}
644: \end{equation}
645: with no other explicit time dependence. When these relations hold,
646: the spatial structure of the system at different times is statistically
647: equivalent, except for a scale factor. Since the domain growth is more
648: clearly observed far from the critical point, we have taken new parameter
649: values accordingly. The numerical results in the Stratonovich interpretation
650: for the scaled pair correlation function
651: are represented in Fig.~\ref{fig:scalNIT}. As it was shown in
652: \cite{marta3}, the pair correlation function exhibits a discontinuity in
653: its first derivative in the presence of noise sources.
654: We have eliminated this discontinuity by fitting a parabolic function in the
655: origin ($r=0$, $\Delta x$). The same study has been made in the case
656: of It\^o interpretation under the same conditions and parameters.
657:
658: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
659: \begin{figure}[htb]
660: \narrowtext
661: \centerline{
662: \psfig{file=fig5.eps,width=7cm}
663: }
664: \caption{
665: Scaled pair correlation function
666: for the EDPT model in the Stratonovich interpretation
667: for t=1300 (circles) and t=2000 (squares), and
668: in the Ito interpretation for
669: $t=1200$ (triangles) and $t=1800$ (diamonds).
670: The parameter values are $a=1$, $c=3$, $D=4$, $\sigma^2=3.5$ and
671: $\Delta x=1$.
672: }
673: \label{fig:scalNIT}
674: \end{figure}
675: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
676: We now compare the temporal evolution of the characteristic length of the
677: system $R(t)$ for the two stochastic interpretations. Fig. \ref{ept2}
678: presents this comparison for equal values of the noise intensity.
679: From these numerical results we can conclude that the Allen-Cahn law
680: is satisfied for the two interpretations, and that there is a time regime
681: in which the system is self-similar. One interesting fact is that
682: domain evolution in It\^o is slower than in Stratonovich, and in both cases
683: much slower than the Ginzburg--Landau model. This fact can be
684: explained looking at the constant prefactor $\sqrt{2A\Gamma}$ of the
685: Allen-Cahn law (\ref{A-Claw}). In the Ginzburg-Landau model $\Gamma=1$,
686: but in the EDPT model this quantity is field dependent (\ref{gamma}),
687: and can be approximated by
688: \begin{equation}
689: \Gamma \approx \frac{1}{1 + c \langle\phi^2\rangle} \approx \frac{1}{1 +
690: c \langle\phi\rangle^2}.
691: \label{approx}
692: \end{equation}
693: According to this expression, and since for a fixed $\sigma^2$ we have
694: that
695: $\langle\phi\rangle_I>\langle\phi\rangle_S$, as a consequence we
696: should expect the slowest growth for the It\^o EDPT case, and the
697: fastest one for the Ginzburg-Landau model. This is what we can
698: see in Fig.~\ref{ept2}.
699: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
700: \begin{figure}[htb]
701: \narrowtext
702: \centerline{\psfig{figure=fig6.eps,width=7cm}}
703: \caption{
704: Allen-Cahn law for the GL model (same parameter values as in Fig.\ref{4}
705: with
706: $\epsilon=10^{-4}$ and $\sigma^{2}=0.6$)
707: and the EDPT model for equal
708: noise intensities (and different mean fields). The latter is
709: computed in both the It\^o and Stratonovich interpretations.
710: The parameter values for the EDPT model are $a = 1$, $c = 3$, $D = 4$,
711: $\sigma^{2} = 3.5$ and $\Delta x = 1$.
712: }
713: \label{ept2}
714: \end{figure}
715: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
716:
717: In order to eliminate the influence of the stationary mean-field value
718: $\langle\phi\rangle$
719: on the growth rate, we have compared the evolution of
720: the system under the two interpretations using in each case a different
721: noise intensity, so that the mean field has the same value in the two
722: cases. The results are shown in Fig \ref{fig:r}, where we have fixed
723: $\langle\phi\rangle=3.15$ for the two interpretations, for which we need
724: $\sigma^{2} = 16$ in the Stratonovich interpretation and
725: $\sigma^{2} = 6$ in the It\^o interpretation.
726: As can be seen, in both interpretations the system seems to evolve at the
727: same rhythm, although the slope in It\^o is slightly higher than in
728: Stratonovich. We think that this small difference is due to the fact
729: that although $\langle\phi\rangle$ is the same for both interpretations,
730: the Stratonovich case is much more fluctuating, and hence we have to expect
731: a larger $\langle\phi^ 2\rangle$ and accordingly a lower slope.
732:
733: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
734: \begin{figure}[htb]
735: \narrowtext
736: \centerline{\psfig{file=fig7.eps,width=7cm}}
737: \caption{
738: Allen-Cahn law for the EDPT model for equal mean fields
739: (and different noise intensities), under both the It\^o and
740: Stratonovich interpretations.
741: For Stratonovich $\sigma^{2} = 16$, for It\^o
742: $\sigma^{2} = 6$.
743: Other parameter values are: $a=1$, $c=0.5$.}
744: \label{fig:r}
745: \end{figure}
746: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
747:
748: \section{Comments and conclusions}
749:
750: It is worth commenting here that an effective model can be developed
751: which has the same stationary solutions as the EDPT model, but different
752: dynamics. The dynamical equation for this effective model with $\Delta x =
753: 1$ is
754: \begin{equation}
755: \frac{\partial \phi}{\partial t} = -a \phi + \frac{B \sigma^{2} c \phi}{1 +
756: c \phi^{2}} + \frac{D}{2d} \nabla^{2} \phi + \xi(\vec x, t)\,,
757: \end{equation}
758: where, as before, $B$ is a parameter whose value indicates
759: the interpretation that we are mimicking ($B=1$ for Stratonovich
760: and $B=2$ for It\^o). The correlation of noise is given by Eq. (2).
761: The equation of motion of the mean value of the field in the linear
762: approximation is
763: \begin{equation}
764: \frac{d \langle\phi \rangle}{d t} = (B\sigma^2 c -a)\langle\phi \rangle +
765: \frac{D}{2d}
766: \nabla^{2} \langle\phi\rangle\,,
767: \end{equation}
768: which tells us that, for $\sigma^2>a/Bc$,
769: the homogeneous phase $\langle\phi\rangle=0$ is unstable.
770: This instability does not appear in the EDPT model (see Eq. \ref{mean}).
771: According to this result, we have to expect an initial transient faster
772: in this model (as in the Ginzburg-Landau model) than in the EDPT cases.
773: This fact has been checked numerically and it can be seen in Fig.
774: \ref{fig:g0}. We can clearly appreciate that the effective model has a
775: much faster initial transient that the EDPT model, for the same values
776: of the parameters which is a signature of the different character of the
777: instability of the initial state. While this observation applies also to
778: the zero-dimensional version of the model, the crucial ingredient in our
779: EDPT model is the role of the spatial coupling, which prevents the fast
780: transition between the two probability peaks. The fact that this
781: transition occurs in a deterministic time scale in the zero-dimensional case
782: is what distinguishes the problem from the usual, barrier-crossing
783: bistability. In the spatially extended case, however, it is precisely
784: the spatial coupling what generates an effective barrier allowing for
785: the formation of stable domains, with an interface-driven dynamics.
786:
787:
788: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
789: \begin{figure}[htb]
790: \narrowtext
791: \centerline{\psfig{file=fig8.eps,width=7cm}}
792: \caption{
793: Transient evolution of the quantity $m_2 = <\phi^2(t)>/<\phi>^2_{st}$
794: which
795: measures the emergence of order from homogeneous initial condition
796: $\phi=0$. The letter $I$ means the It\^o case and the value of the intensity
797: of
798: the noise is inside the parenthesis.
799: }
800: \label{fig:g0}
801: \end{figure}
802: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
803:
804: In conclusion, we have presented a nonequilibrium field model for which one
805: can compute exactly the stationary probability distribution, and which
806: exhibits an intrinsic noise-induced ordering phase transition
807: irrespective of the
808: stochastic interpretation of the multiplicative noise term. In particular,
809: the phase transition is found in the It\^o interpretation, where so far,
810: noise had only been seen to have a disordering effect. The same model can
811: be studied changing the diffusive term by the spatial coupling of the
812: Swift-Hohenberg model in which case a noise-induced pattern transition is
813: found \cite{buceta}. These type of models do constitute a
814: generalization of the Horsthemke-Lefever noise induced transitions to
815: genuine noise induced {\it phase} transitions in extended systems.
816:
817: \acknowledgments This research was supported by the Direcci\'on General de
818: Ense\~nanza Superior (Spain) under projects BFM2000-0624, BFM2001-2159,
819: BXX2000-0638-C02-02 and from the Generalitat de Catalunya (project
820: 2001SGR00223).
821:
822: \appendix
823: \section{Stochastic algorithms}
824:
825: Here we will derive an alternative algorithm which is an extension of the
826: well-known Heun algorithm, valid for both the It\^o and Stratonovich
827: interpretations of stochastic differential equations with multiplicative
828: noise. Our aim is to simulate numerically the following stochastic
829: differential equation on a d-dimensional lattice,
830: \begin{equation}
831: \frac{\partial\phi(\vec x,t)}{\partial t}=f(\phi(\vec x,t),\nabla)
832: +g(\phi(\vec x,t)) \xi(\vec x,t).
833: \end{equation}
834: First of all, we write this equation in a discrete space as follows,
835: \begin{equation}
836: \frac{d \phi_{i}(t)}{dt} = f_{i} (\phi(t)) + g_{i}(\phi
837: (t)) \xi_{i} (t)\,,
838: \label{sde}
839: \end{equation}
840: where $i$ stands for the position inside the lattice, and the noise
841: correlation is given by Eq. (\ref{eq:noiseNdis}).
842:
843: The first step in the derivation of the algorithm is to
844: integrate formally Eq. (\ref{sde}) to get
845: \begin{eqnarray}
846: \phi_{i} (t + \Delta t) = \phi_{i} (t) &+&
847: \int_{t}^{t + \Delta t} f_i(\phi(t')) dt'
848: \nonumber \\
849: &+& \int_{t}^{t + \Delta t} g(\phi(t'))
850: \xi(t') dt'
851: \label{eq1}
852: \end{eqnarray}
853: The first integral in (\ref{eq1}) is evaluated according to a second-order
854: predictor-corrector algorithm.
855: \begin{eqnarray}
856: \phi_{i} (t + \Delta t) = \phi_{i} (t) &+& \frac{f_{i}(\phi(t)) +
857: f_{i}(\tilde{\phi}(t))}{2} \Delta t
858: \nonumber \\
859: &+& \int_{t}^{t + \Delta t} g(\phi(t'))
860: \xi(t') dt'
861: \end{eqnarray}
862: where $\tilde{\phi}(t)$ is the predictor term defined as the first-order
863: solution of (\ref{eq1}),
864: \begin{equation}
865: \tilde{\phi}_{i} (t) = \phi_{i} (t) + f_{i}(\phi_{i} (t)) \Delta t + g_{i}
866: (\phi (t)) X_{i}
867: \label{eqpredictor}
868: \end{equation}
869: This expression defines the first equation of the algorithm, which is
870: independent of the stochastic interpretation. $X_{i}(t)$ is the
871: Wiener process, defined as
872: \begin{equation}
873: X_{i}(t) = \int_{t}^{t + \Delta t} \xi_{i} (t') dt'
874: \end{equation}
875: and whose numerical implementation is
876: \begin{equation}
877: X_{i}(t)= \sqrt{ \frac{2 \sigma^2 \Delta t}{\Delta x^2}} \alpha_i\,,
878: \end{equation}
879: where $\alpha_i$ are independent Gaussian random numbers of zero mean and
880: unity variance, and they are implemented using \cite{raul}.
881:
882: The second integral in Eq. (\ref{eq1}) is not well defined, and
883: one needs to make a prescription for its evaluation, at least up to first
884: order in $\Delta t$.
885:
886: The standard Heun algorithm works for the Stratonovich interpretation,
887: and makes the following assumption
888: \begin{equation}
889: \int_{t}^{t+\Delta t} g(\phi(t')) \xi(t') dt'=\left(\frac{g(\phi_{i}(t))
890: +g(\tilde{\phi}_{i}(t))}{2}\right) X_{i}(t)
891: \end{equation}
892: Accordingly, the second equation of this algorithm is
893: \begin{eqnarray}
894: \phi_{i}(t + \Delta t) = \phi_{i}(t) +
895: \frac{f_{i}(\phi (t)) + f_{i}(\tilde{\phi} (t))}{2} \Delta t
896: \nonumber \\
897: + \left(\frac{
898: g(\phi_{i}(t)) + g(\tilde{\phi}_{i}(t))}{2}\right) X_{i}(t)
899: \label{eqHeun}
900: \end{eqnarray}
901: On the other hand, in the
902: Stratonovich calculus this integral is interpreted as \cite{vankampen}
903: \begin{equation}
904: \int_{t}^{t + \Delta t} g (\phi (t')) \xi(t') dt' = g \left(\frac{
905: \phi_{i}(t) + \tilde{\phi}_{i}(t)}{2}\right) X_{i}(t)\,,
906: \end{equation}
907: so that the second equation of the algorithm is
908: \begin{eqnarray}
909: \phi_{i} (t + \Delta t) = \phi_{i}(t) +
910: \frac{f_{i}(\phi (t)) + f_{i}(\tilde{\phi} (t))}{2} \Delta t
911: \nonumber \\
912: + g \left(\frac{
913: \phi_{i}(t) + \tilde{\phi}_{i}(t)}{2}\right) X_{i} (t),
914: \label{eqS}
915: \end{eqnarray}
916: which is not exactly the standard Heun algorithm. This is the algorithm
917: that has been used in this paper for the Stratonovich interpretation
918: results. Its advantage is that, in contrast to Heun algorithm, our
919: method has an analogue in the Ito interpretation, for which the same
920: integral is defined as \cite{vankampen}
921: \begin{equation}
922: \int_{t}^{t + \Delta t} g_{i}(\phi (t')) \xi_{i} (t') dt' = g_{i}( \phi
923: (t) ) X_{i}(t).
924: \end{equation}
925: Therefore, the second equation of the algorithm in Ito interpretation
926: reads
927: \begin{eqnarray}
928: \phi_{i}(t + \Delta t) = \phi_{i}(t) + \frac{f_{i}(\phi(t)) +
929: f_{i}(\tilde{\phi}(t))}{2} \Delta t
930: \nonumber \\
931: + g_{i}( \phi(t) ) X_{i}(t).
932: \label{eqI}
933: \end{eqnarray}
934: Given these results, the algorithm proceeds by evaluating first the
935: predictor contribution (\ref{eqpredictor}) and, using this value,
936: computing the corrector term (\ref{eqHeun}), (\ref{eqS}) or
937: (\ref{eqI}), corresponding to the Heun, Stratonovich or Ito algorithms
938: respectively. All these three different algorithms are approximations
939: up to the same order (second order in the deterministic part but first
940: order in the stochastic one), when properly expanded in powers of
941: $\Delta t$. One can check that there are no differences,
942: up to these orders, between the Heun and Stratonovich algorithms, as it
943: should be. Nevertheless, the Stratonovich prescription has an extra term,
944: $1/2 g(\phi_i)g'(\phi_i)X_i(t)^2$, with respect the It\^o one, which is of
945: order $\Delta t$. Our Ito algorithm also agrees with the one presented in
946: \cite{rao} up to order $\Delta t^{2}$.
947:
948: \begin{thebibliography}{99}
949:
950: \bibitem{nises} J. Garc\'{\i}a-Ojalvo and J.M. Sancho, {\em Noise in
951: Spatially Extended Systems} (Springer, New York, 1999).
952:
953: \bibitem{chris} C. Van den Broeck, J.M.R. Parrondo, and R. Toral,
954: Phys. Rev. Lett. {\bf 73}, 3395 (1994);
955: C. Van den Broeck, J.M.R. Parrondo, R. Toral, and
956: R. Kawai, Phys. Rev. E {\bf 55}, 4084 (1997).
957:
958: \bibitem{nio} J.M. Sancho and J. Garc\'{\i}a-Ojalvo,
959: in {\em Lecture Notes in Physics}, edited by J.
960: Freund, T. P\"{o}schel, vol. 557 (Springer, Berlin 2000).
961:
962: \bibitem{notapie} Additive noise has been seen to lead to order as well,
963: provided multiplicative noise is also present; see for instance A.
964: Zaikin and L. Schimansky-Geier, Phys. Rev. E {\bf 58}, 4355 (1998).
965:
966: \bibitem{vankampen} N.G. van Kampen, {\em Stochastic Processes in Physics
967: and Chemistry} (North-Holland, 1983)
968:
969: \bibitem{doering1} Ch. R. Doering, C. Mueller and P. Smereka, {\em
970: Interacting particles, the stochastic
971: Fisher--Kolgomorov--Petrousky--Piscunov equation, and duality}, Physica A
972: (in press, 2003).
973:
974: \bibitem{nualart} E. Al\'os, J. A. Le\'on, and D. Nualart, Prob. Theory
975: Relat. Fields {\bf 115}, 41 (1999).
976:
977: \bibitem{rocco} A. Rocco, L. Ram\'{\i}rez--Piscina, and J. Casademunt,
978: Phys. Rev. E{\bf 65}, 056116 (2002).
979:
980: \bibitem{doering} C. Doering, Phys. Lett. A {\bf 122}, 133 (1987).
981:
982: \bibitem{palma} M.G. Zimmermann, R. Toral, O. Piro, and M. San Miguel,
983: Phys. Rev. Lett. {\bf 85}, 3612 (2000).
984:
985: \bibitem{anxo} J.M. Sancho and A. S\'anchez, Eur. Phys. J. B {\bf 16},
986: 127 (2000).
987:
988: \bibitem{grinstein} Y. Tu, G. Grinstein, and M.A. Mu\~noz,
989: Phys. Rev. Lett. {\bf 78}, 274 (1997).
990:
991: \bibitem{kramers} A. Becker and L. Kramer, Phys. Rev. Lett. {\bf 73},
992: 955 (1994).
993:
994: \bibitem{marta2}M. Iba\~{n}es, J. Garc\'{\i}a-Ojalvo, R.
995: Toral, and J.M. Sancho, in {\em Lecture Notes in Physics}, edited by J.
996: Freund, T. P\"{o}schel, vol. 557 (Springer, Berlin 2000), p. 247.
997:
998: \bibitem{oliver} O. Carrillo, M. Iba\~nes, and J.M. Sancho,
999: Fluct. Noise Lett. {\bf 2}, L1 (2002).
1000:
1001: \bibitem{mamunoz} W. Genovese and M.A. Mu\~noz, Phys. Rev. E {\bf 60}, 69
1002: (1999).
1003:
1004: \bibitem{prlmarta} M. Iba\~{n}es, J. Garc\'{\i}a-Ojalvo, R. Toral, and J.M.
1005: Sancho, Phys. Rev. Lett. {\bf 87}, 020601 (2001).
1006:
1007: \bibitem{horsthemke} W. Horsthemke and R. Lefever,
1008: {\em Noise-induced transitions} (Springer, Berlin, 1984).
1009:
1010: \bibitem{marta3} M. Iba\~{n}es, J. Garc\'{\i}a-Ojalvo, R.
1011: Toral, and J.M. Sancho, {\em Eur. Phys. J.} B {\bf 18}, 663 (2000).
1012:
1013: \bibitem{maxi} J.D. Gunton, M. San Miguel, and P.S. Sahni,
1014: in {\em Phase transitions and critical phenomena}, Vol. 8, edited
1015: by C. Domb and J.L. Lebowitz (Academic Press, New York, 1983).
1016:
1017: \bibitem{buceta} J. Buceta, M. Iba\~nes, J.M. Sancho, and K. Lindenberg,
1018: {\em Noise--Driven Mechanism for Pattern Formation}, Phys. Rev. E (in
1019: press) (2003).
1020:
1021: \bibitem{raul} R. Toral and A. Chakrabarti,
1022: Comp. Phys. Commun. {\bf 74} (1993) 4084.
1023:
1024: \bibitem{rao} N.J. Rao, J.D. Borwankar, and D. Ramkrishna,
1025: SIAM J. Control {\bf 12}, 124 (1974).
1026:
1027: \end{thebibliography}
1028:
1029: \end{multicols}
1030: \end{document}
1031:
1032:
1033: