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:
8: \usepackage{amsmath}
9: %\usepackage{graphicx}
10: \usepackage{epsfig}
11:
12:
13:
14:
15: \newcommand{\vp}{{\bf v}}
16: \newcommand{\vf}{{\bf u}}
17: \newcommand{\bx}{{\bf x}}
18: \newcommand{\by}{{\bf y}}
19: \newcommand{\br}{{\bf r}}
20: \newcommand{\bg}{{\bf g}}
21: \newcommand{\hdr}{{\delta{\bf r}}}
22: \newcommand{\bdr}{\delta {\bf r}}
23: \newcommand{\BE}{\begin{equation}}
24: \newcommand{\EE}{\end{equation}}
25: \newcommand{\beqn}{\begin{eqnarray}}
26: \newcommand{\eeqn}{\end{eqnarray}}
27: \newcommand{\intc}{\int_{|\bf x -\bf r| \le R}}
28: \newcommand{\etab}{\mbox{\boldmath $\eta$}}
29:
30:
31: \begin{document}
32: %\draft
33:
34: \title{Clustering transition in a system of
35: particles self-consistently driven by a shear flow }
36:
37: \author{Crist\'obal L\'opez}
38: \address{Instituto
39: Mediterr\'aneo de Estudios Avanzados IMEDEA (CSIC-UIB),
40: Campus de la Universitad de las Islas Baleares,
41: E-07122 Palma de Mallorca, Spain.
42: }
43: \date{\today}
44:
45: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
46: \begin{abstract}
47:
48: We introduce a simple model of active transport for an ensemble of particles
49: driven by an external shear flow.
50: Active refers to the fact that the flow of the particles is modified by
51: the distribution of particles itself.
52: The model consists in that the effective velocity of every particle is
53: given by the average of the external flow velocities felt by the particles
54: located at a distance less than a typical radius, $R$. Numerical analysis reveals
55: the existence of a transition to clustering depending on the parameters of the
56: external flow and on $R$. A continuum description in terms of the
57: number density of particles is derived,
58: and a linear stability analysis of the density equation is performed
59: in order to characterize the transitions observed in the model of interacting particles.
60:
61:
62:
63: \end{abstract}
64: \pacs{05.45.-a, 05.60.-k}
65: \maketitle
66:
67: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
68:
69: %\begin{twocolumns}
70:
71: \section{Introduction}
72:
73: Two different types of transport problems can be roughly distinguished:
74: passive and active. The case of passive transport occurs when the
75: transported quantity does not affect the advecting flow,
76: as exemplified by a dye immersed in a fluid, or of any
77: reacting substance like a chemical pollutant having no feedback on the carrying atmospheric
78: or oceanic flow~\cite{general}.
79: Conversely, in the active transport problem, the subject of this paper,
80: the flow itself is modified by the advected
81: substance. Sometimes this is also refered as self-consistent transport since the velocity field
82: is in general determined by the substance via a dynamical constraint~\cite{diegochaos}.
83: The temperature field,
84: an ensemble of charged particles moving in a self-generated electric field, the vorticity of a
85: fluid flow, and gravitationally interacting particles, are a few examples reflecting
86: the ubiquity and relevance of active transport processes in Nature.
87:
88: Recently, much progress has been achieved in both self-consistent and passive processes through
89: its reformulation within the Lagrangian description~\cite{falkovich, cencio}, which studies transport in terms
90: of individual particle trajectories instead of scalar fields.
91: Thus, the Lagrangian description of a non-reacting passive scalar in an external
92: velocity field, $\vp (\bx,t)$, is given by
93: \BE
94: \frac{d\bx}{dt}=\vp (\bx,t)+\sqrt{2D_0}\etab (t),
95: \label{passive}
96: \EE
97: where $D_0$ is the diffusion coefficient of the passive scalar, and
98: $\etab$ is a normalized Gaussian white noise with zero mean and
99: delta correlated in time. In eq.~(\ref{passive}) the passive character is shown in the
100: fact that there is no coupling between the equations of motion.
101: On the contrary, in active transport the interactions among particles
102: alters the trajectory of any of them, so that for an ensemble of $N$ particles
103: immersed in a fluid flow
104: one can write in general~\cite{boffetta}
105: \BE
106: \frac{d {\bf x}_{i}(t)}{d t} = {\vp}({\bx}_{1}(t), ..., {\bf x}_{N}(t)),
107: \label{active}
108: \EE
109: $i=1,...,N$.
110: This $N$-body problem is often treated in a mean-field approximation
111: where every particle is considered independently of the rest but in an average potential
112: determined self-consistently from the motion of all the particles~\cite{diegochaos, diego2}.
113: In this way, the influence of any particle on the system is just through its contribution
114: to the potential.
115:
116: In this work, we introduce a distinct type of self-consistent transport model.
117: At difference of the mean-field approach, our model
118: assumes a finite range of interaction, $R$,
119: for any particle, so that particles only interact with others surrounding them.
120: Eq.~(\ref{active}) takes the form
121: \BE
122: \frac{d {\bf x}_{i}(t)}{d t} = {\vp}({\bx}_{i}(t), {\bx}_{i+1}(t) ..., {\bf x}_{i+N_R(i)}(t)),
123: \label{activemodelo}
124: \EE
125: where $N_R (i)$ is the number of particles at distance less than $R$ of particle $i$,
126: and with $i+1,..., i+N_R(i)$ we label these particles.
127: Most importantly, the external flow is {\it given} and the invidual particles modify their
128: response to the flow according to the local density around them. Like the mean-field one,
129: our model is an intermediate case between the passive transport case and the many-body
130: self-consistent models with real interactions decaying with distance. The aim
131: of the present paper is to show that even very simple active systems can show a
132: very rich behaviour, and, in particular,
133: the formation of clusters of particles may appear. Also, due its simplicity one can present
134: a detailed analytical study of the model, and show that clustering emerges as a deterministic
135: instability of the density equation of the system.
136:
137:
138: The paper is organized as follows. In the next section we introduce the model and present numerical
139: results showing the clustering. Then, in Sec. III we derive the density equation
140: for the dynamics of the particles, and perform a linear stability analysis of this continuum
141: description. Then, we finish in Sec. IV with the summary of the work.
142:
143:
144: \section{Self-Consistent model of particles driven by an external shear flow. Numerical results}
145:
146: Let us consider $N$ particles in a two-dimensional system of size $L \times L$,
147: and
148: the presence of a stationary incompressible two-dimensional shear flow
149: $\vp (x,y)=(0, v(x))$.
150: In the model the effective velocity of particle $i$ at time $t$,
151: $\vp^{eff}_i (t)$, which is in the position
152: $\bx_i (t)$, is the
153: average velocity of the external velocities felt by the particles in its $R$-neighbourhood.
154: Mathematically
155: \begin{eqnarray}
156: \label{modelo1}
157: \vp^{eff}_i (t)&=&\frac{1}{N_R (i)}\sum_j \vp(\bx_j(t), t), \\
158: \frac{d\bx_i(t)}{dt}&=&\vp^{eff}_i (t),
159: \label{modelo2}
160: \end{eqnarray}
161: where, as indicated above, $N_R (i)$ denotes the number of particles
162: at a distance less than $R$ of particle $i$,
163: and the sum is restricted to the particles $j$ such that $|\bx_i (t)-\bx_j(t)| \le R$.
164: Periodic boundary conditions are considered, and finite-size effects of the particles, like
165: inertia and collisions,
166: are neglected.
167: Note that the self-consistent character of the model comes from the fact that at every
168: time the velocity of any particle is determined by the (local) distribution of particles
169: itself.
170: A noise term similar to eq.~(\ref{passive})
171: could be added to the r.h.s of eq.~(\ref{modelo1}), but this is not considered in this work and
172: we just suppose that advection induced by the external flow dominates on the random motion
173: of the individual particles.
174:
175: Two limits are clearly identified, $R \to 0$ is the tracer limit, i.e., every particle is
176: simply driven by the flow. In the opposite $R \to L$ all the particles move with the same velocity,
177: which is just an average of the external velocity field over all the particles in the system.
178: Physically, the model mimicks particles transported by a flow and
179: with some kind of effective non-local interaction
180: that force them to move locally with the same velocity. In the context of living organisms,
181: traffic or behaviour of human societies many different
182: models have been proposed where the density
183: of particles modify their velocity~\cite{flierl}:
184: repulsion, attraction, distribution of resources, cooperation,
185: are some of the types of interactions among the individuals that are usually studied.
186: These interactions
187: are mediated (in a biological framework) through vision, hearing, smelling
188: or other kinds of sensing, which is reflected, as in our model, by the appearence of a typical
189: interaction radius,
190: $R$. However, a crucial difference of these biologically oriented models with
191: (\ref{modelo1}-\ref{modelo2})
192: is that in those the particles are self-propelled, i.e., they have their own velocity. In our model,
193: the velocity is externally given, and it is our aim to study the properties of the system of
194: particles depending on the characteristics of the external flow. Regarding
195: a biological motivation, our model is adequate for acuatic organisms
196: moving by the water flow that
197: modify their velocity as a response to other individuals living within a certain distance.
198:
199:
200: Concerning the clustering properties of the model, which is the main focus of this work,
201: it is clear that
202: shear enhances encounters among particles, and this favours that particles
203: group together due to the averaging of velocities. On the contrary, local strength
204: of the external velocity field tends to disperse the particles, breaking clusters.
205: Combining these two effects a typical length scale is introduced:
206: \BE
207: \lambda^{-2} =\frac{<(dv(x)/dx)^2>}{<v(x)^2>}.
208: \label{escalataylor}
209: \EE
210: Here $<.>=1/L\int_0^L dx .$, and $\lambda$ is related to the Taylor microscale of turbulence,
211: though here the meaning is somewhat different since it refers to the length scale at which
212: shear is comparable to the amplitud of the velocity.
213: Therefore, one expects the formation
214: of clusters when $\lambda$ is smaller than the typical interaction diameter, $2R$.
215: In other words, the typical length scale emerging from the comparison of shear and velocity must
216: be smaller than the scale at which we average the velocity of any particle.
217: On the other
218: side, it is clear that when $R \approx L$ most of the particles of the system move with the
219: same velocity (all the particles enter in the average sum of (\ref{modelo1})), avoiding the
220: aggregation of the particles.
221: Thus our hypothesis for clustering requires that:
222: \BE
223: \lambda/2 \le R < L
224: \label{condition}
225: \EE
226:
227:
228: To be specific, in the following
229: the external shear flow is
230: given by
231: $v(x)=U_0 +V_0 \sin{(\omega x/L)}$, with $L$ the system size (which
232: we take $L=1$ so that all length-scales are measured in units of $L$), $U_0$,
233: $V_0$ positive constants, and $\omega= 2\pi n$, $n=0, 1, 2, ...$.
234: For this flow it is not difficult to calculate
235: $\lambda=\sqrt{1+2U_0^2/V_0^2}/\omega$
236: so that on
237: taking $\sqrt{1+2 U_0^2/V_0^2}=2\pi$,
238: eq.~(\ref{condition}) becomes $1/n \le 2R <1$.
239:
240: For a spatial distribution of particles
241: the quantitative characterization of clustering \cite{puglio}
242: is performed by means of an entropy-like measure
243: \BE
244: S_M=-\sum_{i=1}^M \frac{m_i}{N}\ln \frac{m_i}{N},
245: \label{entropy}
246: \EE
247: where $M$ is the number of boxes in which we divide the system, and $m_i$ is the number of particles
248: in box $i$. One has that
249: $0\le S_M \le \ln M$, such that $S_M=0$ is obtained when all the particles are
250: in just one of the boxes, and the $\ln M$ value is reached
251: when $m_i= N/M$ for all $i$ (Poisson
252: distribution of particles), i.e., $S_M$ decreases when the clustering increases. We define
253: the clustering coefficient as $C_M=\exp(<H_M>_t)/ M$, where $<.>_t$ denotes a temporal average at long
254: times, so that when there is no clustering $C_M \approx 1$.
255: In the left panel of fig. (\ref{fig:clustering1}) we fix $R=0.1$ (much smaller than
256: the system size $L=1$) and plot $C_M$ vs $n$ observing
257: that the transition to clustering is obtained for $n \approx 5$, fitting perfectly (\ref{condition}).
258: In the right panel we take $n=10$ and plot $C_M$ vs $R$ observing the two transitions indicated
259: in (\ref{condition}).
260: In figure~\ref{fig:pattern} we plot
261: the spatial distribution of particles (in the left panel we plot the initial distribution)
262: in the regime of clustering
263: at time $t=16$ (right) for $R=0.1$ and $n=10$. Here one sees that the particles
264: tend to aggrupate following the sinusoidal
265: flow.
266:
267:
268: Similar results are obtained for other shear flows. E.g., for the linear shear given by
269: $v(x)=\Gamma x$ if $x\in [0, 1/2]$ and $v(x)=\Gamma (1-x)$ for $x\in [1/2, 1]$,
270: $\lambda$
271: is $1/\sqrt{12}$ so that altering the features of the external flow
272: (the values of $\Gamma$) the aggregation properties of the system for fixed $R$
273: are not changed.
274: However, transitions between non-clustering and clustering distributions are observed by
275: varying $R$.
276:
277: In the next section we explain analytically the transition to clustering observed in the numerics.
278: This is done by deriving the density evolution equation for the system of particles.
279:
280:
281: \section{Continuum description in terms of the density of particles. Linear
282: stability analysis}
283:
284:
285: A continuum theory can give further insight on the model.
286: The process to obtain it is standard \cite{Dean}, and we just present here a sketch:
287: define the particle
288: density as $\rho (\bx, t)=\sum_{i=1}^N \rho_i (\bx, t)=\sum_{i=1}^N \delta (\bx_i (t)-\bx)$,
289: then use an arbitrary function $f(\bx)$ defined on the coordinate space, and take the
290: time derivative on both sides of the obvious
291: relation $f(\bx_i (t))=\int d\bx \ \rho_i (\bx, t) f(\bx )$. Finally, using
292: $\int_{|\bx - \bx_i (t)|\le R} d\bx \ \rho (\bx, t)= N_R (i)$ one arrives to
293: \BE
294: \partial_t \rho (\bx, t)+\nabla_{\bx} \cdot
295: \left[\frac{\rho (\bx, t) \int_{|\br -\bx|\le R} d\br\ \vp (\br, t)\ \rho(\br, t)}
296: {\int_{|\br -\bx|\le R} d\br\ \rho(\br, t)}\right]=0.
297: \label{continua}
298: \EE
299: Note that we have maintained the time dependence of the velocity field to reflect the
300: generality of the approach.
301: Eq.~(\ref{continua}) can be simply read as that the density of particles is driven by
302: the effective velocity $\vp^{eff} (\bx, t)= \int_{|\br -\bx|\le R}
303: d\br\ \vp (\br, t)\ \rho(\br, t)/\int_{|\br -\bx|\le R} d\br\ \rho(\br, t)$, whose
304: dependence on the density reveals the self-consistent character of the model. Note also
305: the two trivial limits: a) $R \to 0$ or passive limit, $\vp^{eff} (\bx, t) \to \vp (\bx, t)$, and
306: b) $R \to 1$ ($L=1$) for which $\vp^{eff} (\bx, t) \to 1/N\int d\br \rho(\br, t) \vp (\br, t) $,
307: i.e., the average velocity
308: of the system of particles, which is the same for all of them (and constant for a time-independent
309: velocity field).
310: %If a noise term is added in the r.h.s. of eq.~(\ref{modelo2}) (similarly to eq.~(\ref{passive}))
311: %a laplacian term of the density appears on the r.h.s of eq.~(\ref{continua}).
312:
313:
314: Next we make a linear stability analysis of the stationary homogenous
315: solution, $\rho_0$, of eq.~(\ref{continua}). We first write
316: $\rho (\bx, t)=\rho_0 + \epsilon \psi (\bx, t)$ where $\epsilon$ is a small parameter,
317: and $\psi (\bx, t)$ the space-time dependent perturbation, and substitute it in
318: eq.~(\ref{continua}). To first order in $\epsilon$, using incompressibility
319: of the flow and denoting $\int_B . =\int_{|\br -\bx|\le R} .$ we obtain
320: \begin{eqnarray}
321: &\partial_t \psi+\frac{1}{\pi R^2}\bg (\bx)\cdot\nabla_{\bx}\psi
322: +\frac{1}{\pi R^2}\nabla_{\bx} \cdot \int_B d\br \ \vp (\br, t) \psi (\br, t)& \nonumber \\
323: &-\frac{1}{(\pi R^2)^2}\bg (\bx)\cdot \int_B d\br \ \psi (\br, t) =0,&
324: \label{monstruo}
325: \end{eqnarray}
326: with $\bg (\bx)=\int_B d\br \ \vp (\br, t)$. Though linear, the above expression
327: is still rather complicated
328: since it is non-local in space. For the sinusoidal shear flow
329: (taking for simplicity and without lost of generality $U_0=0$),
330: we have that $\bg (\bx)=\hat \by 2\pi R/\omega \sin (\omega x) J_1 (\omega R)$
331: with $\hat \by$ a unitary vector in the $y$ direction, and $J_1$ the first order Bessel function,
332: so that
333: eq.~(\ref{monstruo}) becomes
334: \begin{eqnarray}
335: &\partial_t \psi+\frac{2 V_0 J_1(\omega R)}{\omega R}\sin (\omega x)
336: \partial_y \psi& \nonumber \\
337: &+\frac{V_0}{\pi R^2}\partial_y
338: [\int_B d\br \ \sin (\omega r_x)\psi (r_x, r_y, t)]& \nonumber \\
339: &-\frac{2 V_0 J_1 (\omega R)}{\pi \omega R^3}\sin (\omega x)\partial_y[
340: \int_B d\br \ \psi (\br, t)] =0,&
341: \label{monstruosin}
342: \end{eqnarray}
343: where $\br=(r_x, r_y)$.
344:
345:
346: We are mainly interested in the clustering transition driven by the relative values of
347: $\lambda$ and $R$, so that we next consider the limit $R << 1$. It is very important
348: to note that, to be the expansion in $R$ consistent, all length-scales of the system must
349: also be very small compared with the system size. Specifically, $\lambda=1/\omega <<1$, so that
350: in particular, one cannot expand the $\sin (\omega r_x)$ in the integrand in eq.~(\ref{monstruosin}).
351: Let us detail the calculations. The two integrals appearing in eq.~(\ref{monstruosin})
352: have the approximations (for simplicity of notation we skip the time dependence):
353: \begin{eqnarray}
354: &I_a=\int_B d\br \psi (r_x, r_y)= \int_{|\br'|\le R} d\br'
355: \psi (r_x'+x, r_y'+y) & \nonumber \\
356: &\approx \int_{|\br'|\le R} d\br'[\psi(x, y) +r_x' \partial_y \psi (x, y)
357: +r_y' \partial_y \psi (x, y)]&\nonumber \\
358: &=\pi R^2 \psi (x, y) +\Theta (R^4);&
359: \label{exp1} \\
360: &I_b=\int_B d\br \sin (\omega r_x) \psi (r_x, r_y)&\nonumber \\
361: &=\int_{|\br'|\le R} d\br'\sin (\omega r_x'+\omega x) \psi (r_x'+x, r_y'+y)& \nonumber \\
362: &\approx \int_{|\br'|\le R} \sin (\omega r_x'+\omega x)
363: [\psi(x, y) +r_x' \partial_y \psi
364: +r_y' \partial_y \psi ]& \nonumber \\
365: &= \frac{2\pi J_1 (\omega R) R}{\omega}\sin (\omega x) \psi +
366: \frac{4 \pi R^2 J_2 (\omega R)}{\omega}\cos (\omega x) \partial_x \psi +\Theta(R^4). & \nonumber \\
367: \label{exp2}
368: \end{eqnarray}
369: Here $\Theta (R^4)$ indicates terms of order $R^4$ and superior.
370: After substituting expressions (\ref{exp1})-(\ref{exp2}) in eq.~(\ref{monstruosin}) the evolution of
371: the perturbation in the small $R$ limit (or better, when the typical length scales of the problem
372: are small) is finally given by
373: \begin{equation}
374: \partial_t \psi+\frac{2 V_0 J_1(\omega R)}{\omega R}\sin (\omega x)
375: \partial_y \psi
376: +\frac{4 V_0 J_2 (\omega R)}{\omega}\cos (\omega x)\partial_{xy}^2 \psi =0,
377: \label{definitivaper}
378: \end{equation}
379: where we have neglected terms of order $R$.
380:
381:
382: %\begin{eqnarray}
383: %&\partial_t \psi+\frac{2 V_0 J_1(\omega R)}{\omega R}\sin (\omega x)
384: %\partial_y \psi& \nonumber \\
385: %&+\frac{4 V_0 J_2 (\omega R)}{\omega}\cos (\omega x)\partial_{xy}^2 \psi =0.&
386: %\label{definitivaper}
387: %\end{eqnarray}
388:
389: Two
390: fundamental features further
391: simplifies the analysis: a) the coefficients are periodic in the spatial coordinates
392: so that Floquet theory can be applied, and b) the coefficients are independent of the
393: $y$ coordinates so that plane waves are solutions on the $y$ direction. Therefore we make the
394: ansatz:
395: \BE
396: \psi (x, y, t) = e^{\Lambda t +i \hat \omega y +i K x}\sum_{m=-\infty}^{\infty} \phi_m e^{i\omega x m},
397: \label{ansatz}
398: \EE
399: where, because of periodic boundary conditions, $\hat \omega =2 \pi p$ ($p=1, 2, ...$),
400: $K=2 \pi p$ ($p=1, 2, ...$), and
401: $\phi_m$ are complex coefficients.
402: $K$ is restricted to the first Brillouin zone determined by $-\omega/2 \le K \le \omega/2$,
403: and $\hat \omega$ is not bounded.
404:
405: If any of the eigenvalues $\Lambda$ is positive then
406: the perturbation grows (the homogenous solution is unstable) and clustering emerges in the system.
407: Thus we look for the conditions to have $\Lambda >0$. Using the exponential formula for
408: the sine and cosine functions, and substituting expression (\ref{ansatz}) in
409: eq.~(\ref{definitivaper}) we obtain, after grouping the terms with the same exponential argument,
410: \begin{equation}
411: \Lambda_m \phi_m +
412: \phi_{m-1}[\alpha_1 -\beta m]
413: +\phi_{m+1} [\alpha_2 -\beta m ]=0,
414: \label{fourier}
415: \end{equation}
416: with $\alpha_1=a_1 \hat \omega /2 -a_2 \hat \omega K/2 +a_2 \omega \hat \omega/2 $,
417: $\alpha_2= -a_1 \hat \omega /2 -a_2 \hat \omega K/2-a_2 \omega \hat \omega/2$,
418: and
419: $\beta=a_2 \omega \hat \omega /2 $, with the notations $a_1=2 V_0 J_1(\omega R)/(\omega R)$
420: and $a_2=4 V_0 J_2 (\omega R)/\omega$.
421:
422:
423: For a simple theoretical analysis we just consider
424: the three Fourier modes $m=0, \pm 1$ and neglect the rest. Diagonalizing the
425: corresponding
426: $3 \times 3$ matrix of coefficients of the
427: system in eq.~(\ref{fourier}) we obtain three eigevanlues, one zero and the other two
428: given by
429: \begin{equation}
430: \Lambda_{\pm} (K) = \pm
431: \frac{V_0\hat \omega}{\omega R}
432: \sqrt{8 J_2^2 K^2 R^2 -4 \omega R J_2 J_1 -2 J_1^2},
433: \label{eigenvalue}
434: \end{equation}
435: where the Bessel functions, $J_1$ and $J_2$, are evaluated at $\omega R$.
436: The expression for $\Lambda_+$ is cuadratic in $K$ with a positive coefficient for
437: the term in $K^2$,
438: so that taking into account that $-\omega/2\le K \le \omega/2$, the
439: inestability is obtained when $\Lambda_+ (K=\omega /2)$ is positive, i.e.,
440: \begin{equation}
441: J_2(\omega R)^2 \omega^2 R^2 - 2 \omega R J_2(\omega R) J_1 (\omega R)- J_1 (\omega R)^2 \ge 0.
442: \label{condition3modos}
443: \end{equation}
444: Numerically one solves the above inequality and obtains that the condition for instability
445: is $\omega R \ge 2.5$,
446: which, despite the many approximations made to derive it, fits well with the numerical result
447: $1/\omega \le 2 R$ (eq.~(\ref{condition}) for $U_0=0$). We have checked that the above
448: result is improved by including more modes in the Floquet analysis. In particular,
449: considering $m=0, \pm 1, \pm 2$, the final condition for the maximum exponent to be positive
450: is $\omega R \ge 1.32 $. We believe that in the limit $m \to \infty$ the numerical result is approached.
451: Therefore this analysis confirms that the derived continuum description eq.~(\ref{continua}) properly
452: describes the discrete interacting particles model, and that the clustering emerges as a
453: deterministic instability of the density equation.
454:
455:
456:
457:
458: \section{Summary}
459:
460: In this work we have proposed a very simple model for an ensemble of particles self-consistently
461: driven by an external shear flow. Despite its simplicity the model shows a very interesting
462: behavior where a transition to grouping of particles are observed. An hypothesis for the
463: appearence of the clustering has been presented. It esentially says that the clustering
464: appears when the length scale that comes from the comparison of the shear flow and
465: the velocity field amplitudes is smaller than the typical interaction radius of the particles.
466: This hypothesis has been numerically checked and also a continuum description has been derived that
467: confirms it.
468:
469:
470: A more realistic interaction of the particles, for example decaying with the distance within
471: $R$, is planned
472: to be studied in the future. Also, it will be interesting
473: a detailed study of the role of a noise term in the dynamics of the particles
474: (which in the continuum description is a difussion term), and the analysis
475: when a chaotic flow is considered.
476:
477: \section{Acknowledgments}
478:
479: I have benefited from many useful conversations with Emilio Hern\'andez-Garc\'\i a.
480: I also acknowledge discussions with Dami\`a Gomila and Pere Colet.
481: Work supported
482: from MCyT of Spain under projects
483: REN2001-0802-C02-01/MAR (IMAGEN) and BFM2000-1108 (CONOCE), and from a
484: {\it Ram\'on y Cajal}
485: fellowship of the Spanish MEC.
486:
487:
488: \begin{figure}
489: \begin{center}
490: \epsfig{figure=transic1.eps,width=\columnwidth,angle=0}
491: \end{center}
492: \caption{Left: $C_M$ vs $n$ with $R=0.1$. Right: $C_M$ vs $R$ for $n=10$.
493: In both plots, $U_0=10$, $\sqrt{1+2\frac{U_0^2}{V_0^2}}=2\pi$, and the
494: time average is performed over the last $2000$ steps in a numerical simulation running
495: for $5000$ steps with $dt=0.01$.
496: }
497: \label{fig:clustering1}
498: \end{figure}
499:
500:
501: \begin{figure}
502: \begin{center}
503: \epsfig{figure=patron2.eps,width=\columnwidth,angle=0}
504: \end{center}
505: \caption{Spatial distribution (statistically stationary) of particles at time $t=0$ (left),
506: and $t=16$ (right panel).
507: Here $R=0.1$, $n=10$,
508: $U_0=1$ ($ \sqrt{1+2\frac{u_0^2}{V_0^2}}=2\pi$, and the initial number of particles $N_0=1500$.
509: }
510: \label{fig:pattern}
511: \end{figure}
512:
513:
514:
515:
516:
517:
518:
519:
520:
521:
522:
523:
524:
525:
526: \begin{thebibliography}{99}
527:
528: \bibitem{general}
529: J. M. Ottino,
530: {\it The Kinematics of Mixing: Stretching, Chaos and Transport} (Cambridge Univ. Press, Cambridge, 1989);
531: T. Bohr, M. Jensen, G. Paladin, and A. Vulpiani,
532: {\it Dynamical Systems Approach to Turbulence} (Cambridge University Press, Cambridge, 1998);
533: Focus issue on {\it Activity in Chaotic Flows} [Chaos {\bf 12}, 372 (2002)].
534:
535: \bibitem{diegochaos}
536: D. del-Castillo-Negrete, CHAOS {\bf 10}, 75 (2000).
537:
538: \bibitem{falkovich}
539: G. Falkovich, K. Gawedzki, M. Vergassola, Rev. Mod. Phys. {\bf 73}, 913 (2001).
540:
541: \bibitem{cencio}
542: A. Celani, M. Cencini, A. Mazzino, and M. Vergassola,
543: Phys. Rev. Lett. {\bf 89}, 234502 (2002).
544:
545: \bibitem{boffetta}
546: G. Boffetta, D. del-Castillo-Negrete, C. L\'opez, G. Pucacco,
547: and A. Vulpiani,
548: Phys. Rev. E {\bf 67}, 026224 (2003).
549:
550: \bibitem{diego2}
551: D. del-Castillo-Negrete, and M.C. Firpo,
552: CHAOS {\bf 12}, 496 (2002);
553: D. del-Castillo-Negrete,
554: In {\it Dynamics and Thermodynamics of Systems with Long Range Interactions}, Edited by T.
555: Dauxois et al. Lecture Notes in Physics Vol. 602, Springer, 2002.
556:
557: \bibitem{flierl}
558: G. Flierl, D. Grunbaum, s. Levin, and D. Olson,
559: J. Theor. Biol. {\bf 196}, 397 (1999);
560: T. Vicsek et al., Phys. Rev. Lett. {\bf 75}, 1226 (1995);
561: D. Helbing, Rev. Mod. Phys. {\bf 73}, 1067 (2001).
562:
563: \bibitem{puglio}
564: A. Puglisi, V. Loreto, U. Marini Bettolo Marconi, and A. Vulpiani
565: Phys. Rev. E {\bf 59}, 5582 (1999)
566:
567: \bibitem{Dean}
568: D. S. Dean, J. Phys. A {\bf 29}, L613 (1996);
569: U. M.B. Marconi and P. Tarazona, J. Chem. Phys. {\bf 110}, 8032 (1999).
570:
571:
572: \end{thebibliography}
573:
574:
575: %\end{twocolumns}
576:
577: \end{document}
578: