1: \documentclass[aps,pre,twocolumn,superscriptaddress,showpacs]{revtex4}
2: %\documentclass[aps,prl,preprint,superscriptaddress,showpacs]{revtex4}
3: \usepackage{amsfonts,amssymb,amsmath,latexsym,epsfig,times}
4: \baselineskip 25pt
5:
6:
7: \begin{document}
8:
9: [Phys. Rev. E {\bf 68}, 056307 (2003)]
10:
11: \title{Reactive dynamics of inertial particles in nonhyperbolic chaotic flows}
12:
13: \author{Adilson E. Motter}
14: \email{motter@mpipks-dresden.mpg.de}
15: \affiliation{Max Planck Institute for the Physics of Complex Systems,
16: N\"othnitzer Strasse 38, 01187 Dresden, Germany}
17:
18: \author{Ying-Cheng Lai}
19: \affiliation{Departments of Mathematics, Electrical Engineering, and Physics,
20: Arizona State University, Tempe, Arizona 85287, USA}
21:
22: \author{Celso Grebogi}
23: \affiliation{Instituto de F\'{\i}sica, Universidade de S\~{a}o Paulo,
24: Caixa Postal 66318, 05315-970 S\~{a}o Paulo, Brazil}
25:
26: \date{\today}
27:
28: \begin{abstract}
29:
30: Anomalous kinetics of infective (e.g., autocatalytic) reactions in open,
31: nonhyperbolic chaotic flows are important for many applications in biological,
32: chemical, and environmental sciences. We present a scaling theory for the
33: singular enhancement of the production caused by the universal, underlying
34: fractal patterns. The key dynamical invariant quantities are the {\it effective
35: fractal dimension} and {\it effective escape rate}, which are primarily
36: determined by the hyperbolic components of the underlying dynamical invariant
37: sets. The theory is general as it includes
38: all previously studied hyperbolic reactive dynamics as a special case.
39: We introduce a class of dissipative embedding maps for numerical verification.
40:
41:
42: \end{abstract}
43:
44: \pacs{47.70.Fw, 47.53.+n, 47.52.+j, 05.45.-a}
45:
46: \maketitle
47:
48: Many chemical and biological processes in fluids are characterized by a
49: filamental distribution of active particles along fractal invariant sets of the
50: advection chaotic dynamics. These fractal structures act as dynamical catalysts
51: for the reaction, which is relevant for a variety of environmental processes in
52: open flows, such as ozone depletion in the atmosphere \cite{SLLE:1996} and
53: population dynamics of plankton in the oceans \cite{A:1998}.
54:
55: The study of active processes in open chaotic flows has attracted a great deal of
56: interest from the dynamical system community \cite{TKPTG:1998,open_flows}. Most
57: of the studies have been performed for time-dependent two-dimensional
58: incompressible flows, in the limit of weak diffusion. The flow is nonturbulent,
59: but the particle dynamics is considered to be chaotic (Lagrangian chaos) and the
60: active particles to interact with one another without modifying the flow. The
61: advection dynamics of such particles can be cast in the context of chaotic
62: scattering, where incoming tracers spend some time in a mixing (scattering)
63: region before being scattered along the unstable manifold of the chaotic saddle.
64: As a result, the products of the reaction concentrate along a fattened-up copy
65: of the unstable manifold, giving rise to the observed fractal patterns.
66:
67: Although filamental patterns have been observed in nature,
68: a clear relation between the observed value of the fractal dimension and the
69: underlying advection dynamics has been lacking. For example, in the ``flow past a
70: cylinder" system previously considered \cite{TKPTG:1998}, the dimension of the
71: unstable manifold is known to be 2 but the relevant dimension governing
72: infective and collisional reactions is about 1.6. This lack of relation,
73: while not reducing the importance of the previous phenomenological
74: characterization of filamental distributions of active particles, has led to some skepticism
75: about the merit of the dynamical system approach to the
76: problem. In general, the advection dynamics can be characterized as either
77: hyperbolic or nonhyperbolic. In hyperbolic chaotic scattering, all the periodic
78: orbits are unstable and there are no Kolmogorov-Arnold-Moser (KAM) tori in the
79: phase space, while the nonhyperbolic counterpart is frequently characterized as
80: having both chaotic and marginally stable periodic orbits. Fundamental
81: assumptions in such works are: (1) the active particles are massless point-like
82: tracers and (2) the advection dynamics of these particles is hyperbolic
83: \cite{inertial}. However, in realistic situations, the Lagrangian dynamics is
84: typically nonhyperbolic and the active particles have finite size and inertia.
85: Indeed, fully hyperbolic systems are quite rare and represent very idealized
86: situations as the advection dynamics of tracers in fluids is usually constrained
87: to have a nonhyperbolic character because of no-slip boundary conditions at the
88: surface of obstacles. Obstacles are at the same time the origin of Lagrangian
89: chaos and the origin of nonhyperbolicity. Even away from obstacles and
90: boundaries, chaotic motions of tracers typically coexist with regular motions.
91: In addition, the individual active particles are often too large to be regarded
92: as noninertial, as is the case for many species of zooplankton in the sea.
93: Therefore, nonhyperbolic and inertial effects are prevalent in nature
94: and expected to play an important role in most environmental processes. A
95: question of physical importance is then: what happens to the reactive
96: dynamics when assumptions (1) and (2) are dropped?
97:
98:
99: In this article, we present a scaling theory for the reactive
100: dynamics of inertial particles in nonhyperbolic chaotic flows. The key
101: concepts in our framework are the {\it effective fractal dimension} and {\it
102: effective escape rate}, which are respectively defined as
103: \begin{eqnarray}
104: \label{d_eff}
105: D_{eff}(\varepsilon )&=&-\frac{d\ln N(\varepsilon)}{d \ln
106: \varepsilon},\\ \kappa_{eff}(\varepsilon )&=& -\frac{d\ln R(n)}{d
107: n},
108: \label{e_eff}
109: \end{eqnarray}
110: where $N(\varepsilon)$ is the number of $\varepsilon$-squares needed
111: to cover the relevant fractal set, and $R(n)$ is the fraction of
112: particles that takes more than $n=n(\varepsilon )$ steps to escape from
113: the mixing region (see below). As a representative application of these
114: concepts, we show, for autocatalytic reactions of the form $A + B
115: \rightarrow 2B$, that the area covered by $B$-particles in the steady state
116: obeys the following scaling law:
117: \begin{equation} \label{scaling}
118: {\cal A}_B\sim\left[\frac{\sigma}{e^{\mu_{eff}/(2-D_{eff})}-1}
119: \right]^{2-D_{eff}},
120: \end{equation}
121: where $\mu_{eff}\equiv (\kappa_{eff} + \tilde{\kappa})\tau$, $\tau$
122: is the time interval between successive reactions (time lag), $\sigma$
123: is the reaction range, and $\tilde{\kappa}$ is the contraction rate
124: due to dissipation. For nonhyperbolic flows in two dimensions, $D_{eff}<2$ and
125: $\kappa_{eff}>0$ are non-trivial functions of the scale $\varepsilon$,
126: which can often be regarded as constants over a wide interval, even
127: though $D=\lim_{\varepsilon\rightarrow 0}D_{eff}(\varepsilon)=2$ and
128: $\kappa=\lim_{\varepsilon\rightarrow 0}\kappa_{eff}(\varepsilon)=0$.
129: We find, surprisingly, that
130: $D_{eff}$ and $\kappa_{eff}$ are significantly different from $D$ and
131: $\kappa$, respectively, not only for noninertial but also for inertial
132: particles, even though the advection dynamics of the latter is hyperbolic,
133: meaning that {\it scars} of the nonhyperbolic conservative dynamics
134: are observable in the hyperbolic dynamics of slightly dissipative
135: systems ($\tilde{\kappa}\ll 1$). The previous relations for
136: noninertial particles in hyperbolic fluids appear as a particular
137: case of our results.
138:
139: The nature of the chaotic scattering arising in the context of particle
140: advection in incompressible fluids may change
141: fundamentally as the mass and size of the particles are increased from
142: zero. Physically, this happens because of the detachment of the
143: particle motion from the local fluid motion. For spherical
144: particles of finite size, the particle velocity ${\bf v}\equiv d{\bf
145: x}/dt$ is typically different from the (time dependent) fluid
146: velocity ${\bf u}={\bf u}({\bf x},t)$ and, in first order, is governed
147: by the equation \cite{MR:1984}
148: \begin{equation}
149: \frac{d{\bf v}}{dt} - \alpha \frac{d{\bf u}}{dt} = - a\;({\bf v} - {\bf u}).
150: \label{fluid}
151: \end{equation}
152: The parameters are
153: $\alpha = 3\rho_f/(\rho_f+2\rho_p)$
154: and
155: $a=\frac{2}{3}\alpha/St$,
156: where $\rho_f$ and $\rho_p$ are the densities of fluid and particle,
157: respectively, and $St$ is the Stokes number, which goes to zero in the
158: limit of point-like particles \cite{approx}. For neutrally buoyant particles, the mass
159: ratio parameter is $\alpha = 1$, while for aerosols and bubbles, we have
160: $\alpha < 1$ and $\alpha > 1$, respectively. The inertia
161: parameter $a$ determines the contraction rate or dissipation in the
162: {\it phase space} $({\bf x},{\bf v})$, which for incompressible flows
163: can be shown to be $-2a$. In the limit $a\rightarrow \infty$, the
164: dynamics is projected on a surface defined by ${\bf v} = {\bf u}$,
165: which corresponds to the advection dynamics of point particles. The
166: {\it configuration-space} projection of the particle motion is
167: strongly influenced by $\alpha$ \cite{BTT:2002}. For small inertia
168: (large $a$), $\boldsymbol{\nabla}\cdot {\bf v}\approx
169: a^{-1}(\alpha-1)\boldsymbol{\nabla}\cdot \boldsymbol{(}({\bf
170: v}\cdot\boldsymbol{\nabla}){\bf u}\boldsymbol{)}
171: =a^{-1}(\alpha-1)(s^2-\omega^2)$, where $s$ and $\omega$ are
172: proportional to the strain rate and vorticity of the fluid
173: \cite{M:1987}, respectively. The behaviors of bubbles and aerosols are
174: then qualitatively different. For instance, along a closed orbit,
175: aerosols are pushed outward, while bubbles are pushed inward.
176: We first consider bubbles, whose configuration-space dynamics
177: is dissipative when the vorticity overcomes the strain rate.
178:
179: Dynamically, the inertial effects are effectively those due to
180: dissipation, so that the transition to finite inertia is equivalent to
181: a transition from open Hamiltonian to dissipative
182: dynamics. It has been recently shown \cite{ML:2002} that,
183: while hyperbolic dynamics is robust, nonhyperbolic
184: chaotic scattering typically undergoes a metamorphosis in the presence of
185: arbitrarily small amount of dissipation. For nonhyperbolic scattering in
186: open Hamiltonian systems, particles can spend a long time in the neighborhood
187: of KAM tori, resulting in an algebraic decay for the survival
188: probability of particles in the scattering region. As a consequence, the fractal
189: dimension of the invariant manifolds is the phase-space dimension
190: \cite{LFO:1991}. This should be contrasted with the
191: hyperbolic case, whose decay is exponential and fractal dimension is
192: typically smaller. The dissipation, however, may convert
193: marginally stable periodic orbits of the KAM islands into
194: attractors. The survival probability then becomes exponential, the
195: dimension of the chaotic saddle becomes fractional, and the overall
196: dynamics of the scattering process becomes hyperbolic.
197:
198: To understand the meanings of $D_{eff}$ and $\kappa_{eff}$ for fractal sets
199: arising in the transition from Hamiltonian nonhyperbolic to weakly dissipative
200: chaotic scattering, we consider a Cantor set, which is
201: constructed in the interval $[0,1]$ according to the rule that
202: in the $n$th time step, a fraction $\Delta_n= \gamma/(\beta +n) +\delta$
203: is removed from the middle of each one of the $N=2^{n-1}$ remaining
204: subintervals, where $\beta$, $\gamma$, and $\delta$ are constants.
205: The conservative case corresponds to $\delta=0$, which is
206: characterized by an algebraic decay with $n$ of the total length
207: remaining, given by $R(n)\sim n^{-\gamma}$ for $n\gg \beta$, and by a
208: unity fractal dimension for the invariant set, $D=1$. The removed
209: fraction $\Delta_n$ decreases at each time step and, as a result, a
210: systematic change of scales is induced, resulting in a non-self-similar
211: invariant set that becomes denser as we go to smaller scales.
212: The relevant consequence is that the box-counting
213: dimension converges slowly to 1, leading to a scale-dependent
214: effective fractal dimension, $D_{eff}\approx
215: 1-\gamma/\ln\varepsilon^{-1}$ for small $\varepsilon$. Similarly, the
216: effective escape rate behaves as $\kappa_{eff}\approx \gamma/n \approx
217: \gamma\ln 2/\ln \varepsilon^{-1}$, where $n=n(\varepsilon)$ is defined
218: as the number of iterations needed to make the length of each
219: remaining subinterval smaller than $\varepsilon$.
220:
221: The limiting dynamics changes drastically and acquires properties of
222: hyperbolic dynamics when a small amount of dissipation is allowed,
223: which is modeled by $0<\delta \ll \gamma/\beta$. In particular, the
224: total length of the remaining intervals decays exponentially,
225: $R(n)\sim (1- \delta)^n$ for $n\gg \gamma/\delta -\beta$, and the
226: dimension of the invariant set becomes smaller than one, namely,
227: $D=\ln 2\boldsymbol{/}\ln{[2/(1-\delta)]}$. At finite time, however, the transition
228: from the conservative to the dissipative case is much smoother. For
229: $\ln\varepsilon^{-1}\gg\beta$, the effective fractal dimension and the
230: effective escape rate are $D_{eff}(\varepsilon )\approx
231: \ln{2}\boldsymbol{/}\ln{[2/(1-\delta)]} - \gamma'/\ln{\varepsilon^{-1}}$ and
232: $\kappa_{eff}(\varepsilon )\approx \ln (1-\delta)^{-1} + \gamma'
233: \ln{[2/(1-\delta)]}\boldsymbol{/}\ln{\varepsilon^{-1}}$, respectively,
234: where $\gamma'\equiv \gamma/(1-\delta)$. The key feature is that
235: unrealistically small scales are required to resolve the limiting
236: values of the fractal dimension and the escape rate, rendering them
237: physically irrelevant. For instance, to obtain $D_{eff} > 0.95$, scales
238: $\varepsilon < 10^{-20}$ may be required. Thus, the physically important
239: characteristics of the fractal set are the effective dimension and
240: escape rate.
241:
242:
243: We now present a physical theory for the scaling law (\ref{scaling}), valid
244: for autocatalytic reactions in two-dimensional time-periodic flows.
245: Consider the area covered by $B$-particles in the open part of the
246: flow (i.e., region where particles eventually escape to infinity) and,
247: to be specific, that the time lag $\tau$ is integer multiple of the
248: flow's period. After a sufficiently long time from the onset of the
249: reaction, the reagent $B$ is distributed along stripes of
250: approximately uniform width, mimicking the unstable manifold. The
251: average width $\epsilon$ of the stripes changes aperiodically over
252: time until the steady state is reached, when it undertakes the
253: periodicity $\tau$ of the reaction. We assume that the reaction is
254: sufficiently close to the steady state so that $D_{eff}(\epsilon)$ and
255: $\kappa_{eff}(\epsilon)$ can be considered constant over time. This
256: condition is not so restrictive because for many systems $D_{eff}$ and
257: $\kappa_{eff}$ are essentially constant over several decades (see
258: below). Therefore, for scales larger than $\epsilon$, the area covered
259: by $B$-particles can be regarded as a fractal characterized by
260: dimension $D_{eff}(\epsilon)$ and escape rate
261: $\kappa_{eff}(\epsilon)$.
262:
263: Let $\epsilon^{(n-1)}(\tau)$ and $\epsilon^{(n)}(0)$ denote the
264: average widths of the stripes right before and right after the $n$th
265: reaction \cite{TKPTG:1998}, respectively. Between successive reactions,
266: the stripes shrink due to escape and dissipation as follows:
267: $\epsilon^{(n)}(\tau)=\epsilon^{(n)}(0) e^{-h_{eff}\;\tau}$,
268: where $h_{eff}=(\kappa_{eff}+\tilde{\kappa})/(2-D_{eff})$ plays the
269: role of an {\it effective} (contracting) Lyapunov exponent, while $\tilde{\kappa}$
270: accounts for the nonconservative contribution. When the reaction occurs, the
271: widening due to the reaction is proportional to the reaction range:
272: $\epsilon^{(n+1)}(0)-\epsilon^{(n)}(\tau)\propto\sigma$.
273: The area covered by $B$-particles right before the $(n+1)$th reaction
274: ${\cal A}_B^{(n)}$ satisfies
275: ${\cal A}_B^{(n)}\propto\boldsymbol{(}\epsilon^{(n)}(\tau)\boldsymbol{)}^{2-D_{eff}}$.
276: These relations can then be combined to
277: yield the following recursive relation for the area:
278: ${\cal A}_B^{(n+1)}=e^{-\mu_{eff}}\left[ ({\cal
279: A}_B^{(n)})^{1/(2-D_{eff})}+c\;\sigma \right]^{2-D_{eff}}$,
280: where $\mu_{eff} = (\kappa_{eff} +\tilde{\kappa})\tau$ and $c$ is a
281: constant geometric factor. From the condition ${\cal
282: A}_B^{(n+1)}={\cal A}_B^{(n)}$ follows our main scaling (\ref{scaling})
283: for the area ${\cal A}_B$ in the steady state \cite{cont}.
284: This scaling holds for both noninertial and inertial
285: particles, regardless of whether the flow is hyperbolic or nonhyperbolic.
286: The hyperbolic case with inertial particles is studied in
287: Ref. \cite{TNMGT:PNAS}. The scaling (\ref{scaling}) represents a
288: further step toward generality
289: since it is also valid for nonhyperbolic flows.
290:
291:
292: To make possible a numerical verification of the scaling law (\ref{scaling}),
293: it is necessary at present to use discrete-time maps. To construct a class of
294: maps that captures all essential features of continuous-time chaotic flows, we
295: note: (1) the fluid dynamics, determined by $d{\bf
296: x}/dt={\bf u}$, is {\it embedded} in the particle's advection equation
297: and is recovered in the limit $a\rightarrow \infty$; (2) the
298: phase-space contraction is determined by $a$ (irrespective of
299: $\alpha$); (3) for small inertia, the configuration-space contraction
300: is proportional to $a^{-1}(\alpha-1)$.
301: For an area preserving map ${\bf x}_{n+1}={\bf M}({\bf x}_n)$,
302: representing the dynamics of a time-periodic incompressible fluid, a
303: possible choice for the corresponding {\it embedding map} representing
304: the {\it inertial} particle dynamics is
305: ${\bf x}_{n+2} - {\bf M}({\bf x}_{n+1}) =
306: e^{-a}\boldsymbol{(}\alpha{\bf x}_{n+1}-{\bf M}({\bf x}_n)\boldsymbol{)}$,
307: where the factors involving $a$ and $\alpha$ are naturally imposed by
308: the particle dynamics \cite{c10}. This can be written as
309: \begin{eqnarray}
310: \label{embed2x}
311: {\bf x}_{n+1} &=& {\bf M}({\bf x}_{n}) + \boldsymbol{\delta}_n,\\
312: \boldsymbol{\delta}_{n+1}&=&e^{-a}\boldsymbol{ (}\alpha{\bf x}_{n+1}-{\bf M}({\bf x}_n)\boldsymbol{ )},
313: \label{embed2d}
314: \end{eqnarray}
315: where ${\bf x}$ and $\boldsymbol{\delta}$ can be interpreted as the
316: configuration-space coordinates and the detachment from the fluid
317: velocity, respectively, so that (${\bf x}$, $\boldsymbol{\delta}$)
318: represents the phase-space coordinates.
319: This class of embedding maps
320: can be a paradigm to address many problems in inertial advection dynamics
321: as it captures the essential properties of
322: Eq.~(\ref{fluid}). In particular, it is uniformly dissipative, with
323: phase-space contraction rate equal to $e^{-2a}$; the noninertial
324: dynamics ${\bf x}_{n+1} = {\bf M}({\bf x}_{n})$ is recovered in the
325: limit $a\rightarrow \infty$; and the configuration-space contraction rate
326: is proportional to $e^{-a}(\alpha-1)$ for $e^{-a}(\alpha-1)\ll 1$, in
327: agreement with the distinct behavior expected for aerosols and
328: bubbles. Therefore, for finite $a$, a rich higher dimensional dynamics
329: with $\alpha$-dependent ${\bf x}$-space projection is expected. Next
330: we consider such a dynamics for both $\alpha > 1$ and $\alpha < 1$.
331:
332:
333:
334: To simulate the flow, we consider a two-dimensional area-preserving map that has
335: a pronounced nonhyperbolic character \cite{LFO:1991}: $(x,y)\rightarrow
336: [\lambda(x-w^2/4),\lambda^{-1}(y+w^2)]$, where $w\equiv x+y/4$ and $\lambda>1$
337: is the bifurcation parameter. The dynamics is nonhyperbolic for
338: $\lambda\lesssim 6.5$. For $\lambda=4$, for example, there is a major KAM
339: island in the $xy$-space, as shown in Fig.~\ref{fig1}(a). Also, from
340: Fig.~\ref{fig1}(a), one can see tangencies between the stable and unstable
341: manifolds in the neighborhood of the KAM island, which is a signature of
342: nonhyperbolicity. It is well established that, within the nonhyperbolic region,
343: the dimension of the invariant manifolds is $D=2$ and the escape rate is
344: $\kappa=0$ \cite{LFO:1991,ML:2002}. When this map is embedded in
345: Eqs.~(\ref{embed2x}-\ref{embed2d}), for $\alpha > 1$, the $xy$-projection of the
346: resulting 4-dimensional map is dissipative in the mixing region (KAM islands and
347: neighborhood). In this regime, the dissipation stabilizes marginally stable
348: periodic orbits in the KAM islands of the conservative map, converting the KAM
349: islands and neighborhood into the corresponding basin of attraction of the newly
350: created attractors, as shown in Fig.~\ref{fig1}(b). The basin itself extends
351: around the mixing region mimicking the stable manifold of the conservative
352: dynamics. As a result, the tangencies between the invariant manifolds
353: apparently disappear, suggesting that the advection dynamics of bubble particles
354: is {\it hyperbolic}. For $\alpha < 1$, on the other hand, the
355: configuration-space projection expands in the mixing region and almost all the
356: orbits eventually escape to infinity. However, for small inertia and $\alpha$
357: close to 1, particles in the regions corresponding to KAM islands of the
358: conservative dynamics and neighborhood are {\it almost trapped} in the sense
359: that the time it takes to escape is much larger in these
360: regions than outside
361: them. These regions are neglected in our analysis of the open part of the flow,
362: as shown in Fig.~\ref{fig1}(c), because filamental structures cannot be resolved
363: inside them.
364:
365:
366: Numerical simulation of the autocatalytic reaction is performed by dividing the
367: mixing region with a grid where the size of the cells represents the reaction
368: range $\sigma$. Particles are placed in the center of the cells. When a
369: reaction takes place in a cell occupied by a $B$-particle, all the cells
370: adjacent to it are {\it infected} with $B$-particles \cite{c4}. We assume that
371: $A$ is the background material and that the reaction takes place simultaneously
372: for all the particles at time intervals $\tau$ \cite{c3}. If we start with a small seed
373: of $B$-particles near the stable manifold, after a transient time a steady state
374: is reached where $B$-particles are accumulated along a fattened-up copy of the
375: unstable manifold, as shown in Fig.~\ref{fig1}(d) for massless point particles,
376: in Fig.~\ref{fig1}(e) for bubbles, and in Fig.~\ref{fig1}(f) for aerosols. In
377: the computation, particles $A$ and $B$ are set to have the same mass ratio and
378: inertia parameters. To compute the effective fractal dimension $D_{eff}$ of
379: the unstable manifold, we use the uncertainty algorithm \cite{MGOY:1985} applied
380: to the first order approximation of the inverse map. The effective dimension
381: turns out to be constant over many orders of magnitude of variations in
382: $\varepsilon\;$ and it is approximately the same for both noninertial and
383: slightly inertial bubble particles ($D_{eff}= 1.73$ for $\varepsilon >
384: 10^{-15}$), while it is somewhat smaller for slightly inertial aerosol particles
385: ($D_{eff}= 1.68$ for $\varepsilon > 10^{-15}$), as shown in Fig.~\ref{fig2}(a).
386: Strong evidence of the scaling law (\ref{scaling}) is presented in
387: Fig.~\ref{fig2}(b) for two different values of the time lag $\tau$,
388: where the scaling exponent is consistent with
389: $D_{eff}= 1.73$ for noninertial and bubble particles, and with $D_{eff}= 1.68$ for
390: aerosol particles. We see that even though the area ${\cal A}_B$ changes with
391: the inertial properties of the particles, the {\it scaling} of ${\cal A}_B$
392: remains essentially the same for bubbles, as expected from our
393: Cantor-set model.
394:
395:
396: It is instructive to compare this result with the reaction-free dynamics. The
397: dimension of nonhyerbolic invariant sets can be argued to be integer by mean of
398: a zoom-in technique, where a fast numerical convergence is achieved by focusing
399: on the densest parts of the fractal \cite{LFO:1991}. The reaction, however, has
400: a {\it global} character, as it takes place along the unstable manifold around
401: the entire mixing region. This makes the convergence of the relevant effective
402: dimension extremely slow, and that is why the effective dimension is apparently
403: constant.
404:
405: In summary, we have shown that the dynamical system approach to the reactive
406: dynamics in imperfectly mixed flows also applies to realistic situations where
407: nonhyperbolic and inertial effects are relevant. The rate equations of reactive
408: processes are primarily governed by finite-time dynamics and as such change
409: smoothly in the {\it noninertial} $\rightarrow$ {\it inertial} transition, which
410: is in sharp contrast with the metamorphosis undergone by the long-term and
411: asymptotic dynamics of reaction-free particles. We have focused on
412: autocatalytic reactions, but our results are expected to hold whenever the
413: reaction front mimics the underlying unstable manifold and the activity takes
414: place along the boundary of a fattened-up fractal. Examples of this kind of
415: process include infective reactions (e.g., combustion \cite{KMSSKN:2003}) and
416: collisional reactions in general (e.g., $A+B\rightarrow 2C$, where an unlimited
417: amount of material $A$ is present and material $B$ is continuously injected in
418: the vicinity of the stable manifold \cite{TKPTG:1998}). Finally, we observe
419: that our analysis does not rely on the existence of a well-defined fractal set
420: in the advection dynamics. The results remain valid as long as effective values
421: for the fractal dimension and escape rate can be properly defined and are
422: approximately constant over the relevant interval of observation. This is
423: important for environmental processes, whose underlying dynamics is only
424: partially understood \cite{TNMGT:PNAS}.
425:
426: AEM and YCL were supported by AFOSR under Grant No. F49620-03-1-0290. CG was supported by Fapesp and CNPq.
427: AEM thanks Tamas T\'el for illuminating discussions.
428:
429: \begin{references}
430:
431: \bibitem{SLLE:1996}
432: S. Edouard, B. Legras, F. Lefevre, and R. Eymard, Nature (London) {\bf 384}, 444 (1996).
433:
434: \bibitem{A:1998}
435: E. R. Abraham, Nature (London) {\bf 391}, 577 (1998).
436:
437: \bibitem{TKPTG:1998}
438: Z. Toroczkai, G. K\'arolyi, \'A. P\'entek, T. T\'el, and
439: C. Grebogi, Phys. Rev. Lett. {\bf 80}, 500 (1998).
440:
441: \bibitem{open_flows}
442: See, for example, Focus issue: {\it Active chaotic flow}, Chaos {\bf 12}(2), 372 (2002);
443: references therein.
444:
445: \bibitem{inertial}
446: Inertial effects on reative particles have been
447: previouly considered for turbulent flows [R. Reigada, F. Sagu\'es, and
448: J. M. Sancho, Phys. Rev. E {\bf 64}, 026307 (2001)] and for cellular
449: flows [T. Nishikawa, Z. Toroczkai, C. Grebogi, and T. T\'el,
450: Phys. Rev. E {\bf 65}, 026216 (2002); Z. H. Liu, Y.-C. Lai, and
451: J. M. Lopez, Chaos {\bf 12}, 417 (2002)].
452:
453: \bibitem{MR:1984}
454: M. R. Maxey and J. J. Riley, Phys. Fluids {\bf 26}, 883 (1983).
455:
456: \bibitem{approx}
457: Eq. (\ref{fluid}) is an approximation of the Maxey-Riley equation
458: [A. Babiano, J. H. E. Cartwright,
459: O. Piro, and A. Provenzale, Phys. Rev. Lett. {\bf 84}, 5764 (2000)].
460:
461: \bibitem{BTT:2002}
462: I. J. Benczik, Z. Toroczkai, and T. T\'el, Phys. Rev. Lett. {\bf 89}, 164501 (2002).
463:
464: \bibitem{M:1987}
465: M. R. Maxey, Phys. Fluids {\bf 30}, 1915 (1987);
466: E. Balkovsky, G. Falkovich, and A. Fouxon, Phys. Rev. Lett. {\bf 86}, 2790 (2001).
467:
468: \bibitem{ML:2002}
469: A. E. Motter and Y.-C. Lai, Phys. Rev. E {\bf 65}, 015205 (2002).
470:
471: \bibitem{LFO:1991}
472: Y.-T. Lau, J. M. Finn, and E. Ott, Phys. Rev. Lett. {\bf 66}, 978 (1991).
473:
474:
475: \bibitem{cont}
476: A similar scaling law can be derived
477: in the continuous-time limit $\tau\rightarrow 0$.
478:
479: \bibitem{TNMGT:PNAS}
480: T. T\'el, T. Nishikawa, A. E. Motter, C. Grebogi, and Z. Toroczkai, Chaos {\bf 14}, 72 (2004).
481:
482: \bibitem{c10}
483: For $\alpha =1$, this equation reduces to
484: the first-order approximation of the bailout embedding map introduced in
485: [J. H. E. Cartwright, M. O. Magnasco, and O. Piro, Phys. Rev. E {\bf 65}, 045203 (2002)]
486: to study the advection dynamics of neutrally buoyant particles.
487:
488: \bibitem{c4} For numerical convenience, the autocatalytic reaction is
489: followed by coalescence so that when more than one $B$-particle is
490: found in the same cell, they are all replaced by a single $B$-particle
491: in the center of the cell. Without coalescence, the number of $B$-particles
492: diverges because of the shrinking due to dissipation, but the area
493: ${\cal A}_B$ still reaches a steady state. Since we focus on the {\it area}
494: covered by $B$-particles, our results do not depend on the
495: coalescence.
496:
497: \bibitem{c3}
498: The time lag $\tau$ is assumed to be on
499: the order of or smaller than the relevant time scale of the advection
500: dynamics $\tau_f=1/\kappa_{eff}(\sigma)$. For $\tau\gg \tau_f$, the reaction
501: undergoes an emptying transition and the Cantori structures near
502: the KAM tori cannot be neglected. In this regime,
503: almost all the particles along the unstable manifold escape in between
504: successive reactions and the hypothesis that reagent $B$ is distributed along
505: approximately uniform stripes breaks down outside the Cantori. However, since
506: Cantori are obstacles to the transport of particles, for $\sigma$ small enough,
507: high concentration of particles can still be found inside the hierarchy of
508: Cantori structures. For details we refer to [A. P. S. de Moura and C. Grebogi,
509: preprint (2003)].
510:
511: \bibitem{MGOY:1985}
512: C. Grebogi, S. W. McDonald, E. Ott, and J. A. Yorke, Phys. Lett. A {\bf 99}, 415 (1983).
513:
514: \bibitem{KMSSKN:2003}
515: I. Z. Kiss, J. H. Merkin, S. K. Scott, P. L. Simon, S. Kalliadasis, and Z. Neufeld,
516: Physica D {\bf 176}, 67 (2003).
517:
518: \end{references}
519:
520:
521:
522:
523: % FIGURE 1
524: \begin{figure}[pt]
525: \begin{center}
526: %\epsfig{figure=xfig1.gif,width=13.0cm}
527: %\epsfig{figure=fig1.eps,width=\linewidth}
528: \caption
529: { For $\lambda = 4$, (a) KAM island (light grey), and stable (grey) and unstable (black)
530: manifolds for $a=\infty$. (b) Fixed point attractor (black
531: dot), basin of attraction (grey), and unstable manifold (black), for
532: $\alpha=1.05$ and $a=1$. (c) Stable (grey) and unstable (black) manifolds
533: for
534: $\alpha=0.95$ and $a=1$, outside the region covered by the almost trapped orbits (light grey). Particles
535: are launched with initial velocity matching the fluid velocity
536: ($\boldsymbol{\delta}_{0}={\bf 0}$).
537: (d-f) Corresponding area covered by $B$-particles in the ``open" part of the flow right before
538: the reaction, for $\tau=5$ and $\sigma=5.10^{-3}$. }
539: \label{fig1}
540: \end{center}
541: \end{figure}
542:
543:
544:
545: % FIGURE 2
546: \begin{figure}[ph]
547: \begin{center}
548: %\epsfig{figure=xfig2.gif,width=10cm}
549: %\epsfig{figure=fig2a.eps,width=4.0cm}
550: %\epsfig{figure=fig2b.eps,width=4.0cm}
551: \caption
552: {For $\lambda = 4$, (a) Effective dimension of the unstable
553: manifold as computed from the uncertainty method,
554: where $f(\varepsilon)$ is the fraction of $\varepsilon$-uncertain points
555: in the line $x=0$, $0<y<0.1$ of the time reversed dynamics.
556: (b) Scaling of the relative area ${\cal A}_B$ covered by $B$-particles
557: [in the region shown in Figs.~\ref{fig1}(d)-(f),
558: right before the reaction] as a function of the
559: reaction range $\sigma$ for two choices of the time lag $\tau$.
560: In both plots, stars correspond to
561: noninertial particles ($a=\infty$), circles to
562: bubble particles with $\alpha=1.05$ and $a=1$, and plus signs to aerosol
563: particles with $\alpha=0.95$ and $a=1$. The aerosol data in (a) are
564: shifted vertically downward for clarity.}
565: \label{fig2}
566: \end{center}
567: \end{figure}
568:
569: \end{document}
570:
571:
572:
573:
574:
575:
576:
577:
578: