1: %\documentstyle[twocolumn,aps,prl,times,epsfig]{revtex}
2: %\documentclass[aps,pre,floats,twocolumn,aps,superscriptaddress,showpacs]{revtex4}
3: %\documentstyle[aps,prl,twocolumn,times,epsfig]{revtex}
4: %\documentclass[pre,floats,twocolumn,aps,superscriptaddress,showpacs]{revtex4}
5: %\documentclass[prl,floats,aps,superscriptaddress,showpacs]{revtex4}
6: %\documentclass[preprint,floats,aps,superscriptaddress,showpacs]{revtex4}
7: \documentclass[prl,twocolumn,floats,aps,superscriptaddress,showpacs]{revtex4}
8:
9: \usepackage{amsmath}
10: %\usepackage{graphicx}
11: \usepackage{epsfig}
12:
13:
14: \newcommand{\bv}{{\bf v}}
15: \newcommand{\bu}{{\bf u}}
16: \newcommand{\bx}{{\bf x}}
17: \newcommand{\by}{{\bf y}}
18: \newcommand{\br}{{\bf r}}
19: \newcommand{\bg}{{\bf g}}
20: \newcommand{\bk}{{\bf k}}
21: \newcommand{\hdr}{{\delta{\bf r}}}
22: \newcommand{\bdr}{\delta {\bf r}}
23: \newcommand{\BE}{\begin{equation}}
24: \newcommand{\EE}{\end{equation}}
25: \newcommand{\BA}{\begin{array}}
26: \newcommand{\EA}{\end{array}}
27: \newcommand{\beqn}{\begin{eqnarray}}
28: \newcommand{\eeqn}{\end{eqnarray}}
29: \newcommand{\intc}{\int_{|\bf x -\bf r| \le R}}
30: \newcommand{\etab}{\mbox{\boldmath $\eta$}}
31: \newcommand{\zetab}{\mbox{\boldmath $\zeta$}}
32: \newcommand{\nablab}{\mbox{\boldmath $\nabla$}}
33:
34: \begin{document}
35:
36: \title{Macroscopic description of particle systems with
37: non-local density-dependent diffusivity.
38: }
39:
40: \author{Crist\'obal L\'opez}
41: \affiliation{Instituto
42: Mediterr\'aneo de Estudios Avanzados IMEDEA (CSIC-UIB),
43: Campus de la Universidad de las Islas Baleares,
44: E-07122 Palma de Mallorca, Spain.
45: }
46:
47: \date{\today}
48:
49: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
50: \begin{abstract}
51:
52: In this paper we study macroscopic
53: density equations
54: in which the diffusion coefficient
55: depends on a weighted spatial average of the density itself.
56: We show that large differences
57: (not present in the local density-dependence case)
58: appear between
59: the density equations that are derived from
60: different representations
61: of the Langevin
62: equation describing a system of interacting
63: Brownian particles.
64: Linear stability analysis
65: demonstrates that under some circumstances the density
66: equation interpreted like Ito
67: has pattern solutions, which
68: never appear for the
69: H\"anggi-Klimontovich interpretation, which is the other one typically
70: appearing in the context of nonlinear diffusion processes.
71: We also introduce a discrete-time microscopic model of
72: particles that
73: confirms the results
74: obtained at the macroscopic density level.
75:
76: \end{abstract}
77: \pacs{87.23.Cc, 05.10.Gg, 05.40.-a}
78: \maketitle
79:
80: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
81:
82: %\begin{twocolumns}
83: %\section{Introduction}
84:
85: In the mathematical description of
86: many ecological systems the mobility of the species
87: depends on the local density or concentration of the individuals \cite{okubo,murray}.
88: A standard example of this is given by models for animal or insect
89: dispersal where it is usually considered that there is an
90: increase of the diffusion coefficient due to population pressure.
91: These models are written down as
92: an extension of the diffusion equation
93: which includes
94: a diffusivity depending on the population density.
95: Under the generic name of nonlinear diffusion equations
96: similar equations arise in other contexts like in
97: the flow through a porous medium, bacterial dynamics,
98: and transport in
99: plasmas~\cite{murray,aronson}.
100:
101:
102: Generally, the nonlinear diffusion
103: equations for the density
104: correspond to
105: a coarse-grained description of the more fundamental
106: particle stochastic dynamics. Similarly to the case of diffusion
107: in nonhomogeneous media where the diffusion coefficient
108: varies in space, a heuristic derivation
109: of the nonlinear diffusion density equations
110: has to deal with the Ito-Stratonovich
111: dilemma \cite{vankampen, sancho}. (To be precise
112: one should speak about the interpretation of
113: the Langevin equation since,
114: as we will see later, the {\it Stratonovich}
115: realization does not appear in our study
116: and one has instead the H\"anggi-Klimontovich
117: interpretation of the Langevin equation \cite{hanggi}.)
118: In the context of noise-induced
119: phase transitions this is of a major importance
120: \cite{raul},
121: and this work pretends to show that it
122: is also specially relevant when one considers
123: nonlocal interactions between the particles.
124:
125:
126: In a biological setting,
127: nonlocal effects have been widely considered
128: since they account for the interaction between
129: individuals that are separated in space
130: \cite{flierl}.
131: More recently, there has appeared interest in models
132: where the mobility depends on an
133: averaged density rather than on its
134: value at a point \cite{holm}.
135: The spatial non-locality is there
136: introduced in order
137: to take into account the
138: finite-size of the particles or
139: the senses (visual, hearing, etc..) of the organisms, that
140: somehow determine their mobility.
141: The idea behind is that the diffusivity of any
142: particle may depend on the configuration
143: of other particles in its vicinity.
144:
145:
146: Thus,
147: in this work we study a nonlocal version of a general
148: density-dependent diffusion equation for
149: the number density of individuals.
150: We focus on its pattern and clustering formation
151: properties.
152: Making linear stability analysis we first show that,
153: contrarily to what happens in the local case, completely
154: different results can be obtained for the
155: Ito and H\"anggi-Klimontovich
156: descriptions.
157: In fact, spatial patterns can emerge for the first but never for
158: the second.
159: Then, we introduce two microscopic
160: interacting particles systems
161: whose continuum evolutions
162: are given by the
163: Ito and H\"anggi-Klimontovich density equations.
164: Numerical simulations of these systems
165: confirm the calculations made
166: with the continuum density equations.
167:
168:
169:
170: Let us consider a population dynamics
171: model of relevance for insect dispersal.
172: The diffusion coefficient depends on
173: the local density of particles, i.e., $D=D(\rho)$, where $\rho (\bx ,t)$ is
174: the population density (we assume that $D$ is not an explicit function of the
175: spatiotemporal coordinates).
176: We only study
177: positive density-dependence, which
178: is the most widespread hypothesis and indicates that
179: competition increases dispersing because individuals have
180: better survival rates by leaving high-density areas~\cite{ecography}.
181: Mathematically, $D'(\rho) \equiv dD/d\rho >0$.
182: The mean-field density descriptions in Ito
183: and H\"anggi-Klimontovich interpretations
184: are~\cite{vankampen,raul,hanggi}:
185: \begin{eqnarray}
186: \frac{\partial \rho (\bx, t)}{\partial t}&=&\nablab^2 \left( D(\rho) \rho \right),
187: \label{IDESDE}
188: \\
189: \frac{\partial \rho (\bx, t)}{\partial t}&=&\nablab \cdot \left( D(\rho) \nablab
190: \rho \right).
191: \label{SDESDE}
192: \end{eqnarray}
193: The first one corresponds to Ito and the second to H\"anggi-Klimontovich, and in the
194: following we will denote them, respectively, as IE and HKE.
195: The HKE is the one that is
196: usually considered when modeling at the macroscopic level~\cite{okubo,murray},
197: since one writes down the density evolution as
198: a flux equation $\partial_t \rho =- \nablab \cdot {\bf J} $
199: with the density-dependent flux ${\bf J}=-D(\rho) \nablab \rho$.
200: But this is
201: not a definitive argument and, as it is
202: discussed in \cite{horstemke}, the IE
203: should be the correct one for a continuum description
204: of a discrete population dynamics model of non-overlapping generations.
205: A proper derivation of density equations from the fundamental
206: stochastic particle dynamics, once a right interpretation
207: for this is
208: assumed, can be consulted in \cite{dean}. Note that they are stochastic,
209: but in our study we neglect the fluctuations by considering a
210: kind of mean-field approximation \cite{umberto}.
211: Let us
212: assume an initial
213: constant density $\rho (\bx, 0) =\rho_0$. We now perform a linear
214: stability analysis of the stationary homogeneous (identical to $\rho_0$
215: because particle number is conserved) solution for
216: both IE and HKE. Writing $\rho (\bx, t) =\rho_0 + \epsilon \psi (\bx, t)$,
217: where $\epsilon$ is a small parameter, and $\psi$ the space-time dependent
218: perturbation, we obtain (we label the equations with IE or HKE):
219: \begin{eqnarray}
220: (IE) \ \ \partial_t \psi (\bx, t)& =& \left( D\left(\rho_0\right)+ \rho_0 D' \left(\rho_0\right)\right)
221: \nablab^2 \psi \\
222: (HKE) \ \ \partial_t \psi (\bx, t) &=& D(\rho_0)
223: \nablab^2 \psi.
224: \label{locales}
225: \end{eqnarray}
226: Note that in both the perturbation follows a simple diffusion equation and there
227: is, therefore, no instability in the density equation that could give rise
228: to the formation of patterns or aggregates of particles. From this point of view,
229: the Ito and H\"anggi-Klimontovich
230: frameworks (Eqs.~(\ref{IDESDE}) and (\ref{SDESDE})) are equivalent.
231:
232: Now let us consider that the diffusivity depends on an
233: averaged density over
234: the whole system, ${\bar \rho} (\bx, t)=\int_{\Re^d}
235: d\br G(\bx -\br) \rho (\br, t)$,
236: where the kernel, $G(\bx -\br)$,
237: represents the effect of population density at $\br$ on the density at $\bx$.
238: It embodies the forces of attraction and repulsion of
239: neighbors~\cite{kenkre}.
240: We assume that $G$ is normalized so that $\int_{\Re^ d} d\br G(\br )=1$.
241: Then let us consider $D=D({\bar \rho})$ in
242: Eqs.~(\ref{IDESDE}-\ref{SDESDE})
243: and make again a
244: linear stability analysis around $\rho_0$
245: for the now nonlocal HKE and IE,
246: \begin{eqnarray}
247: (HKE) \nonumber \\ &\epsilon \partial_t \psi (\bx, t) =
248: \nablab \left[ D(\rho_0 +\epsilon \int d\br G(\bx -\br) \psi (\br, t)) \right. & \nonumber \\
249: &\left. \nablab (\rho_0 + \epsilon \psi) \right],&
250: \label{sdenolocal}
251: \end{eqnarray}
252: so that to first order in $\epsilon $ the perturbation simply diffuses,
253: $\partial_t \psi = D(\rho_0)\nablab^2 \psi$, and no instability
254: happens in HKE.
255: However, when we consider the Ito description
256: we have:
257: \begin{eqnarray}
258: &(IE)\ \ \ \ \ \ \ \ \ \epsilon \partial_t \psi (\bx, t) \ \ \ \ & \nonumber \\
259: &=\nablab^2 \left[ D\left(\rho_0 +
260: \epsilon \int d\br G(\bx -\br) \psi (\br, t)\right)
261: \left( \rho_0 + \epsilon \psi (\bx, t)\right) \right]& \nonumber \\
262: &= \epsilon \nablab^2 \left[ D(\rho_0) \psi + \rho_0 D(\rho_0)\int d\br
263: G(\bx-\br) \psi(\br, t)\right].&
264: \label{idenolocal}
265: \end{eqnarray}
266: Considering a harmonic perturbation
267: $\psi (\bx, t)= \exp(\lambda t+ i \bk \cdot \bx)$
268: it is not difficult to obtain the following dispersion
269: relation
270: \begin{equation}
271: \lambda (k)= -k^2 \left( D(\rho_0)+
272: \rho_0 D'(\rho_0) {\hat G} (\bk) \right),
273: \label{dispersion}
274: \end{equation}
275: where $k=|\bk|$ and
276: ${\hat G}(\bk)=\int d\br G(\br) e^{-i\bk\cdot \br}$ is the Fourier
277: transform of the kernel.
278: The important fact with Eq.~(\ref{dispersion}) is that
279: depending on the kernel function (note that the
280: form of the diffusivity functional is almost irrelevant
281: once we have assumed that it is an increasing function
282: of the density, $D'(\rho)>0$),
283: the perturbation growth
284: rate can be positive, giving rise to the aggregation
285: of particles or formation of spatial patterns~\cite{flierl,kenkre}.
286: As we have
287: just seen, this is absolutely different to the result
288: obtained in the H\"anggi-Klimontovich
289: framework, Eq.~(\ref{sdenolocal}) and below.
290:
291: Next we introduce a model, choosing
292: the density and the kernel functions,
293: with a clearly oriented
294: ecological application and where $\lambda$ can take positive values.
295: We study the distribution of organisms over a spatial
296: area and thus we
297: restrict to two spatial dimensions.
298: A typical form for
299: $D(\rho)$ in
300: animal dispersal models is $D_0 (\rho/\rho_0)^p$ where $D_0$ and $p$ are
301: positive real numbers~\cite{okubo,murray}, and $\rho_0$ is a reference density
302: which we take equals to the initial one without loss of
303: generality. Because of the normalization of the
304: kernel function the averaged density, ${\bar \rho}$, has
305: the same dimensions as $\rho$.
306: Concerning the kernel, we
307: take the usual top-hat function \cite{kenkre}:
308: $G(\br)=1/(\pi R^2)$ if $|\br | \leq R$ and
309: $G(\br)=0$ otherwise, which introduces a typical
310: interaction length, $R$, in the system.
311: One obtains
312: ${\hat G}(k)=2 J_1(k R)/(k R)$ so that the dispersion relation
313: takes the form
314: \begin{equation}
315: \lambda (k)= - D_0 k^2 \left( 1 + \frac{2 p J_1 (kR)}{kR} \right).
316: \label{dispersion2}
317: \end{equation}
318: $J_1$ is the first order Bessel function, and it is clear from
319: Eq.~(\ref{dispersion2}) that
320: the value of $D_0$ only sets the time scale of the system.
321: The onset of pattern formation is $\lambda >0$ which numerically
322: is approached for $p$ larger than $p_c \approx 7.6$.
323: In Fig.~\ref{fig:patternscont}
324: we show a long-time density solution for the IE (left) and
325: HKE (right) with this kernel and $D$ functions, and $p=9$.
326: Note the hexagonal pattern formed in IE and the homogenous solution for
327: HKE, i.e.,
328: the numerical results
329: confirm the different spatial structures obtained for the
330: two descriptions. We have checked that this is so for
331: all $p$, as can be seen below in Fig.~\ref{fig:structure}.
332:
333:
334:
335:
336: \begin{figure}
337: \begin{center}
338: \epsfig{figure=figpatternscont1.eps,clip=true,width=\columnwidth,angle=0}
339: \end{center}
340: \caption{Steady number density of particles with nonlocal density-dependence
341: diffusivity.
342: Left is Ito and right H\"anggi-Klimontovich.
343: The diffusivity functional is $D(\rho)=D_0(\rho/\rho_0)^p$, and
344: we use the top-hat kernel with $R=0.1$. The values of the
345: other
346: parameters, in both plots,
347: are $D_0=0.0001$, $p=9$, $\rho_0=2000$, and system sizes
348: in $x$ and $y$ are equal to $1$.
349: }
350: \label{fig:patternscont}
351: \end{figure}
352:
353:
354: Up to now we have shown that, for
355: nonlinear nonlocal diffusion phenomena,
356: the different
357: interpretations, in terms of density equations,
358: of the same system dynamics lead to
359: very distinct results.
360: One may ask if this
361: also happens at the level of the discrete
362: interacting particle dynamics.
363: Since the IE and HKE
364: are the mean-field density
365: descriptions
366: derived from
367: the Langevin equation
368: of a system
369: of Brownian interacting particles,
370: we expect that
371: the same differences encountered before
372: between both descriptions
373: appear.
374: We will show that this is in
375: fact the case for two time-discrete microscopic
376: systems
377: which, in the continuous time limit,
378: are the Ito and H\"anggi-Klimontovich
379: Langevin equations corresponding to the IE and HKE
380: with the top-hat kernel and the typical diffusivity
381: functional for insect dispersal.
382:
383: Therefore, the
384: problem is to find a discrete dynamics from
385: which deriving HKE and IE.
386: Following the discussion
387: in \cite{horstemke} (Section 5.4.2)
388: we first take a population dynamics
389: system of non-overlapping generations whose
390: Langevin equation, in the time continuum limit,
391: should
392: be interpreted in Ito. The system consists in
393: $N$ particles with positions
394: $\br_i (t)=(x_i (t), y_i (t))$, $i=1,..., N$, in a two-dimensional periodic box of
395: size $L=1$
396: evolving as follows
397: \begin{equation}
398: \br_i(t+\Delta t)= \br_i(t) + \sqrt{2 D_0 (N_R (\br_i(t))/N)^p \Delta t}
399: \ \etab^{(i)} (t),
400: \label{itodiscreta}
401: \end{equation}
402: where $D_0$ and $p$ are real positive numbers, $\Delta t$ is the time step,
403: $N_R (\br_i(t))$ is the number
404: of particles at a distance less than $R$ of particle $i$ at time $t$,
405: and
406: $\etab^{(i)}(t)=(\eta^{(i)}_x, \eta^{(i)}_y)$
407: is a Gaussian White noise with
408: correlations $\left<\eta^{(i)}_a (t) \eta^{(j)}_b (s) \right>=
409: \delta_{i j} \delta_{a b} \delta_{t s}$.
410: As already mentioned, in the
411: limit $\Delta t \to 0$
412: one obtains an Ito Langevin equation,
413: \begin{equation}
414: \frac{d\br_i}{dt}=\sqrt{2 D_0 (N_R (\br_i(t))/N)^p} \etab^{(i)} (t),
415: \label{itole}
416: \end{equation}
417: and one can calculate~\cite{dean,umberto}
418: that
419: the corresponding density equation of this Langevin is
420: the nonlocal IE with the top-hat kernel
421: and the density functional of insect dispersal.
422: For the H\"anggi-Klimontovich case, the first step is to interpret
423: eq.~(\ref{itole}) in this sense so that it is equivalent to the
424: following Ito Langevin equation \cite{nota}:
425: \begin{equation}
426: \frac{d\br_i}{dt}=g(\br_i(t))\nabla_{\br_i} g(\br_i(t))+
427: \sqrt{2 D_0 (N_R (\br_i(t))/N)^p} \etab^{(i)} (t),
428: \label{hkle}
429: \end{equation}
430: where $g(\br_i(t))=\sqrt{2 D_0 (N_R (\br_i(t))/N)^p}$.
431:
432: Thus, our discrete particle systems equivalent to the IE and HKE
433: are given by Eq.~(\ref{itodiscreta}) and the discrete-time version
434: up to order $\Delta t$ of Eq.~(\ref{hkle}), respectively.
435: Note that, since our simulations are off-lattice, i.e.,
436: particles move in the continuum space,
437: we have to consider an auxilary
438: grid to compute the gradients of $g(\br_i(t))$
439: for the H\"anggi-Klimontovich case.
440: In Fig.~(\ref{fig:patternsdis}) we show
441: the long-time spatial distribution of particles
442: for the same parameter values used in Fig.~(\ref{fig:patternscont}).
443: Left corresponds to Ito
444: and right to
445: H\"anggi-Klimontovich.
446: The particles following the time-discrete-dynamics
447: which
448: give rise to the Ito Langevin equation arrange
449: in an hexagonal pattern of clusters, while in the
450: H\"anggi-Klimontovich case the particles are randomly
451: distributed in the space. This is completely in accord
452: with what it happens in the macroscopic description.
453:
454:
455: \begin{figure}
456: \begin{center}
457: \epsfig{figure=patternsdis2.eps,clip=true,width=\columnwidth,angle=0}
458: \end{center}
459: \caption{Spatial distribution of particles at large time.
460: Left, is for Ito and right for H\"anggi-Klimontovich.
461: We use the same parameters as in Fig.~(\ref{fig:patternscont}),
462: i.e., $D_0=0.0001$, $p=9$, and $R=0.1$. The number of particles is $N=2000$,
463: for
464: both plots, and $dt=0.01$.
465: For the right panel we take a square grid (to calculate the gradients
466: of the $g$ function) with lattice spacing $dx=0.005$.
467: }
468: \label{fig:patternsdis}
469: \end{figure}
470:
471:
472: A more quantitative comparison between
473: the macroscopic and microscopic dynamics follows.
474: The characterization of the pattern
475: can be performed via
476: the structure function, whose
477: maxima identify periodic structures. In the case of
478: the continuum density, the structure function in the steady state, $S_c$,
479: is computed as the modulus of the Fourier
480: transform of $\rho (\bx, t)$ averaged spherically
481: and in time. In the case of the particle system
482: the structure function, which is related but not
483: identical to $S_c$, is calculated as
484: $S_d(k)=\left< \left| \sum_j e^{i \bk \cdot \bx_j (t)}\right|^2 /N
485: \right>$, where
486: $\bx_j (t)$ is the position vector of particle $j$,
487: $\bk$ is a two-dimensional wave vector with modulus $k$,
488: and the average indicates a spherical average over the wave vectors
489: with modulus $k$, a temporal average in
490: the steady state, and an average
491: over many different realizations.
492: In Fig.~\ref{fig:structure}
493: we plot the maximum (for $k>0$) value of $S_c$ (left panel) and of $S_d$ (right)
494: versus the parameter $p$. With circles we plot the computed ones
495: for IE and with squares for HKE.
496: It is clear from
497: the plot that the value of $p$ at which the transition to pattern,
498: in Ito,
499: occurs is the same for both, and coincident with
500: the numerical value obtained from stability analysis.
501: Moreover, for any value of $p$ no peak (i.e., no pattern)
502: is ever observed in the H\"anggi-Klimontovich prescription.
503:
504:
505:
506: \begin{figure}
507: \begin{center}
508: \epsfig{figure=funes3.eps,clip=true,width=\columnwidth,angle=0}
509: \end{center}
510: \caption{Maxima of the structure functions for different values of
511: the parameter $p$. In the left panel we plot the maximum value of $S_c$,
512: i.e., the corresponding to the continuum density equation. It is normalized
513: with the number of points of the grid used to make the computation, $10201$.
514: Circles are for the Ito prescription and squares for the H\"anggi-Klimontovich one.
515: On the right, we plot, also vs $p$,
516: the maximum value of $S_d$, i.e, of the structure function
517: for a distribution of particles. Also circles are for Ito and squares
518: for H\"anggi-Klimontovich. The rest of the parameters for both plots are $D_0=0.0001$
519: and
520: $R=0.1$. For the H\"anggi-Klimontovich particle dynamics we use
521: a lattice with $dx=0.005$.
522: }
523: \label{fig:structure}
524: \end{figure}
525:
526:
527: In summary,
528: density evolution equations with density-dependent diffusion
529: coefficient
530: are standards in the macroscopic
531: modeling of biological, physical and
532: chemical processes.
533: Typically, they are postulated
534: as $\partial_t \rho (\bx, t)= - \nablab \cdot {\bf J}$
535: with ${\bf J}$ the population flux given by
536: ${\bf J}= -D(\rho) \nablab \rho$. In this way one
537: obtains a macroscopic density equation in the
538: so-called
539: H\"anggi-Klimontovich framework, which is different to the Ito
540: one, Eq.~(\ref{IDESDE}).
541: Though this last expression
542: rarely appears in the literature,
543: this is no reason to exclude it~\cite{vankampen},
544: and one has to resort
545: to the real microscopic dynamics to infer the proper density
546: equation describing the system dynamics.
547: In fact, a proper macroscopic equation for population
548: dynamics models of non-overlapping generations goes through Ito
549: \cite{horstemke}.
550: In this work we have
551: shown that, in the context of biological population
552: dynamics but easily extended to others,
553: no relevant differences, at the level of clustering or
554: pattern formation, arise between
555: both equations when the functional $D(\rho)$ is local.
556: However, when $D$ is nonlocal
557: the differences
558: are rather important, and non-homogenous
559: spatial structures can appear in Ito but never
560: in H\"anggi-Klimontovich.
561: Furthermore, we have discussed the discrete
562: dynamics of non-locally
563: interacting particle systems
564: from which the Ito and H\"anggi-Klimontovich density
565: equations are derived, confirming also at this level
566: the large differences encountered in the macroscopic
567: description.
568: It is thus clear that the main message of this work
569: is that it is specially relevant in the case
570: of systems with
571: nonlocal density-dependent diffusivity that
572: the macroscopic description should be properly derived
573: from the discrete particle model.
574: The inclusion of reaction terms and further extensions
575: of the model will be pursued
576: in the future.
577:
578:
579:
580:
581: I acknowledge useful conversations with
582: Emilio Hern\'andez-Garc\'\i a and Umberto
583: Marini Bettolo Marconi.
584: Financial support from MEC (Spain) and FEDER through project
585: CONOCE2 (FIS2004-00953) is greatly acknowledged.
586: C.L. is a {\sl
587: Ram{\'o}n y Cajal} research fellow of the Spanish MEC.
588:
589:
590:
591: \begin{thebibliography}{99}
592:
593:
594:
595: \bibitem{okubo}
596: A. Okubo and S. Levin,
597: \emph{Diffusion and Ecological Problems: Modern Perspectives},
598: Springer Verlag, New York (2001).
599:
600:
601: \bibitem{murray}
602: J.D. Murray,
603: \emph{Mathematical biology},
604: Springer-Verlag, Berlin (2002).
605:
606:
607:
608:
609: \bibitem{aronson}
610: D. G. Aronson, in \emph{Nonlinear Diffusion Problems},
611: A. Fasano and M. Primicerio (eds.),
612: Lecture Notes in Mathematics 1224, Springer Verlag, Berlin (1985);
613:
614:
615:
616: \bibitem{vankampen}
617: N.G. van Kampen, \emph{Stochastic processes in physics and chemistry},
618: 2nd ed., North-Holland, Amsterdam (1992);
619: N. Van Kampen, J. Phys. Chem. Solids {\bf 49}, 673 (1988);
620: C.W. Gardiner, \emph{Handbook of Stochastic Methods}, 2nd ed., Springer, Berlin (1985).
621:
622:
623: \bibitem{sancho}
624: J. M. Sancho, M. San Miguel and D. D\"urr,
625: J. Stat. Phys.
626: {\bf 28}, 291 (1982).
627:
628:
629: \bibitem{hanggi}
630: P. H\"anggi and H. Thomas, Phys. Rep. {\bf 88}, 207 (1982);
631: Y. L. Klimontovich, Physica A {\bf 163}, 515 (1990).
632:
633:
634:
635: \bibitem{raul}
636: C. Van den Broeck, J. M. R. Parrondo, and R. Toral,
637: Phys. Rev. Lett. {\bf 73}, 3395 (1994);
638: C. Van den Broeck, J. M. R. Parrondo, R. Toral,
639: and R. Kawai,
640: Phys. Rev. E {\bf 55}, 4084 (1997).
641:
642:
643: \bibitem{flierl}
644: G. Flierl, D. Grunbaum, S. Levin, and D. Olson,
645: J. Theor. Biol. {\bf 196}, 397 (1999);
646: A. Mogiler and L. Edelstein-Keshet,
647: J. Math. Biol. {\bf 38}, 534 (1999);
648: F. Cecconi, M. Marsili, J. R. Banavar, and A. Maritan,
649: Phys. Rev. Lett. {\bf 89}, 088102 (2002);
650: C. Topaz and A. Bertozzi,
651: SIAM J. Appl. Math. {\bf 65}, 152 (2004).
652: Y. E. Maruvka and N. M. Shnerb,
653: Phys. Rev. E {\bf 73}, 011903 (2006).
654:
655:
656: \bibitem{holm}
657: D. D. Holm and V. Putkaradze,
658: Phys. Rev. Lett. {\bf 95}, 226106 (2005);
659: D. D. Holm and V. Putkaradze,
660: nlin.PS/0506020;
661: C. L\'opez,
662: Physical Review E {\bf 72}, 061109 (2005).
663:
664:
665: \bibitem{ecography}
666: E. Matthysen,
667: Ecography {\bf 28}, 403 (2005).
668:
669:
670: \bibitem{horstemke}
671: W. Horsthemke and R. Lefever,
672: \emph{Noise-induced transitions}, (Springer, Berlin, 1984).
673:
674: \bibitem{dean}
675: D. S. Dean, J. Phys. A {\bf 29}, L613 (1996).
676:
677: \bibitem{umberto}
678: U.M.B. Marconi and P. Tarazona, J. Chem. Phys. {\bf 110}, 8032 (1999);
679: A.J. Archer and M. Rauscher, J. Phys. A: Math. Gen. {\bf 37}, 9325 (2004).
680:
681:
682: \bibitem{kenkre}
683: M. A. Fuentes, M. N. Kuperman, and V. M. Kenkre, Phys. Rev. Lett. {\bf 91}, 158104 (2003);
684: E. Hern\'andez-Garc\'\i a and C. L\'opez, Phys. Rev. E
685: {\bf 70}, 016216 (2004);
686: C. L\'opez and E. Hern\'andez-Garc\'\i a,
687: Physica D {\bf 199}, 223 (2004);
688: M. G. Clerc, D. Escaff, and V. M. Kenkre,
689: Phys. Rev. E {\bf 72}, 056217 (2005).
690:
691: \bibitem{nota}
692: According to the H\"anggi-Klimontovich interpretation,
693: the discrete-time version of the Langevin equation
694: Eq.~(\ref{itole}) goes through the
695: post-point discretization. However, in order to have
696: a dynamics of non-overlapping generations we
697: write down first its equivalent
698: Langevin equation in Ito, Eq.~(\ref{hkle}),
699: and then consider its time-discrete form to
700: order $\Delta t$.
701:
702: %\bibitem{sanmiguel}
703: %M. San Miguel, R. Toral, in: E. Tirapegui, J. Mart\'\i n,
704: % R. Tiemann (Eds.),
705: %\emph{Instabilities and Nonequilibrium Structures VI},
706: %(Kluwer Academic Publishers, 1999).
707:
708: \end{thebibliography}
709:
710:
711: %\end{twocolumns}
712:
713: \end{document}
714: