1: \documentclass[aps,pre,twocolumn,showpacs,floatfix]{revtex4}
2: %\documentclass[aps,pre,preprint,showpacs,floatfix]{revtex4}
3: \usepackage{graphicx}
4: \usepackage{bm}
5: \usepackage{amsmath}
6: \usepackage[centerlast]{subfigure}
7: \newif\iffigs
8: \figstrue % uncomment if figures are present
9: %\figsfalse % uncomment if there are no figures available
10: \iffigs
11: \fi
12: \def\drawing #1 #2 #3 {
13: \begin{center}
14: \setlength{\unitlength}{1mm}
15: \begin{picture}(#1,#2)(0,0)
16: \put(0,0){\framebox(#1,#2){#3}}
17: \end{picture}
18: \end{center}}
19:
20:
21:
22: \begin{document}
23: \title{Thin front propagation in random shear flows}
24:
25: \author{M. Chinappi$^{1}$, M. Cencini$^{2,3}$ and A. Vulpiani$^{3,4}$
26: } \affiliation{ $^1$ Dipartimento di Meccanica e Aeronautica,
27: Universit\`a di Roma "la Sapienza", Via Eudossiana 18, I-00184 Roma,
28: Italy.\\ $^2$ ISC-CNR via dei Taurini, 19 I-00185 Roma, Italy\\ $^3$
29: SMC-INFM c/o Dipartimento di Fisica, Universit\`a di Roma ``La
30: Sapienza'', p.zzle Aldo Moro, 2 I-00185 Roma, Italy\\ $^4$Dipartimento
31: di Fisica and INFN, Universit\`a di Roma ``La Sapienza'', p.zzle Aldo
32: Moro, 2 I-00185 Roma, Italy} \date{\today}
33:
34: \begin{abstract}
35: Front propagation in time dependent laminar flows is investigated in
36: the limit of very fast reaction and very thin fronts, i.e. the
37: so-called geometrical optics limit. In particular, we consider fronts
38: evolving in time correlated random shear flows, modeled in terms of
39: Ornstein-Uhlembeck processes. We show that the ratio between the time
40: correlation of the flow and an intrinsic time scale of the reaction
41: dynamics (the wrinkling time $t_w$) is crucial in determining both the
42: front propagation speed and the front spatial patterns. The relevance
43: of time correlation in realistic flows is briefly discussed in the
44: light of the bending phenomenon, i.e. the decrease of propagation
45: speed observed at high flow intensities.
46: \end{abstract}
47:
48: \pacs{47.70.Fw,82.40.Ck}
49: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
50: %47.70.Fw Chemically reactive flows
51: %82.40.Ck Pattern formation in reactions with diffusion,
52: % flow and heat transfer
53: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
54: \maketitle
55:
56:
57: Front propagation in fluid flows is a problem relevant to many areas
58: of science and technology ranging from combustion technology
59: \cite{W85} to chemistry \cite{chem} and marine ecology \cite{bio}. In
60: the last years several theoretical
61: \cite{ABP00,Const,KBM01,ACVV01,ACVV02,NX03,CTVV03} and experimental
62: \cite{LMRS03,PH04,SS05,LMRS05} works studied chemically reactive
63: substances stirred by laminar flows. This problem, though considerably
64: simpler than the case of turbulent flows \cite{W85}, is non trivial
65: and displays a very rich and interesting phenomenology. We mention
66: here the front speed locking phenomenon in time dependent cellular
67: flows, which was numerically and theoretically found in
68: Ref.~\cite{CTVV03} and then experimentally observed in
69: Ref.~\cite{SS05}. Another interesting example is represented by the
70: theoretical studies on time periodic shear flows \cite{KBM01,NX03} and
71: the recent experimental work \cite{LMRS05} which study aqueous
72: reactions in periodically modulated Hele-Shaw flows.\\\indent Laminar
73: flows are interesting also because they constitute a theoretical
74: laboratory to study some problems which can be encountered in more
75: complex (turbulent) flows. For instance, this is the case of time
76: correlations \cite{D99,A00} that are believed to be very important in
77: determining the bending of turbulent premixed flame velocity when the
78: intensity of turbulence is increased (see \cite{Ronney} for a
79: discussion about this problem). Actually for bending several
80: mechanisms have been proposed like reaction quenching \cite{quench},
81: dynamics of pockets of material which did not react left behind
82: \cite{pockets} and finally time correlations \cite{D99,A00}.\\\indent
83: The aim of this paper is to investigate the role of time correlations
84: in the propagation of reactions in random shear flows. In particular,
85: we shall consider the problem in the so-called geometrical optics, or
86: Huygens regime \cite{Peter} that is realized in the case of very fast
87: reactions, taking place in very thin regions. As in
88: Refs.\cite{D99,A00}, we neglect possible back-effects of the
89: transported reacting scalar on the velocity field, i.e. we treat the
90: problem in the context of passive reactive transport. The latter
91: assumption is justified for dilute aqueous auto-catalytic reactions
92: and more in general for chemical reaction with low heat release. It
93: should be also remarked that in the chosen framework pockets cannot be
94: generated due to shear geometry and quenching of reaction cannot
95: happen due to the choice of working in the geometrical optics
96: limit. Therefore, the case under consideration allows us to focus on
97: the effects due to time correlations solely.\\\indent Let us start to
98: introduce our problem by shortly recalling the main equations. Since
99: we consider premixed reactive species, the simplest model consists in
100: studying the dynamics of a scalar field $\theta(\bm x,t)$ representing
101: the fractional concentration of the reaction's products ($\theta=1$
102: inert material, $\theta=0$ fresh one and $0<\theta<1$ coexistence of
103: fresh material and products). The evolution of $\theta$ is ruled by
104: the advection-reaction-diffusion equation:
105: \begin{equation}
106: \partial_t \bm u+\bm u\cdot\bm\nabla \theta=D\Delta\theta+
107: \frac{f(\theta)}{\tau_r}\,,
108: \label{eq:ard}
109: \end{equation}
110: where $\bm u$ is a given velocity field (incompressible $\bm
111: \nabla\cdot\bm u= 0$ through this paper). The $f(\theta)$ (that is
112: typically a nonlinear function with one unstable $\theta=0$ and one
113: stable $\theta=1$ state) models the production process occurring on a
114: time-scale $\tau_r$.\\\indent Eq.~(\ref{eq:ard}) may be studied for
115: different geometries and boundary conditions. In this work we consider
116: an infinite two-dimensional stripe along the x-direction with a
117: reservoir of fresh material on the right, inert products on the left
118: and periodic boundary conditions in the transverse direction (which
119: has size $L$). In particular, we shall be concerned with the
120: concentration initialized as a step, i.e. $\theta(x,y,0)=1$ for $x\leq
121: 0$, and zero otherwise.
122: With this geometry a front of inert material (stable phase) propagates
123: from left to right with an instantaneous velocity which can be defined
124: as:
125: \begin{equation}
126: v_f(t) = \frac{1}{L} \int_\mathcal{D} d \bm x \;\partial_t \theta(\bm x,t)\,,
127: \label{eq:burningrate}
128: \end{equation}
129: more precisely this is the bulk burning rate~\cite{Const} (integration
130: is over the entire domain $\mathcal{D}$). Most of the theoretical
131: studies aim to predict the dependence of the average front speed
132: $V_f=\langle v_f \rangle$ on the details of the velocity field. Very
133: important are of course also the propagation speed fluctuations; one
134: would like to predict how these are related to the fluid velocity
135: fluctuations. These are in general very difficult issues, but very
136: important in technological applications, where one has to project the
137: reactor geometry and flow characteristics. Definite answers about the
138: reaction propagation exists only in particular conditions, e.g. when
139: the flow is motionless ($\bm u =0$) and under rather general
140: hypothesis on the production function $f(\theta)$ it is possible to
141: show that the reaction asymptotically propagates with a velocity $v_0$
142: within the bounds (see Ref.~\cite{saarloos} for an exhaustive review):
143: \begin{equation}
144: 2\sqrt{\frac{Df'(0)}{\tau_r} } \leq v_0
145: \leq 2\sqrt{\frac{D}{\tau_r} \sup_{\theta} \left\{ \frac{f(\theta)}{\theta}
146: \right\}}\,,
147: \label{eq:aw}
148: \end{equation}
149: where $f'$ indicates the derivative, and the thickness of the reaction
150: zone varies as $\xi \propto \sqrt{D\tau_r}$. For a wide class of
151: reaction terms $f$, such as the autocatalytic reaction dynamics,
152: $f(\theta)=\theta(1-\theta)$, and more in general for convex functions
153: ($f^{''}(\theta)<0$) one can prove that $v_0=2\sqrt{Df'(0)/\tau_r}$
154: exactly. In the presence of a velocity field $\bm u$, generally one
155: has that the speed $V_f$ is larger than the bare velocity
156: $v_0$. Specifically here we consider the limit in which the reaction
157: is much faster than the other time-scales of the problem, formally
158: this regime is reached when $\tau_r \to 0$ and $D\to 0$ but with
159: $D/\tau_r =const$ such that the bare propagation velocity $v_0=
160: 2\sqrt{D/\tau_r}$ is finite and well defined, while the reaction zone
161: thickness shrinks to zero ($\xi \to 0$) \cite{Peter}, where for the
162: sake of notation simplicity we posed $f'(0)=1$.. It should be noted
163: that this regime, also called geometrical optics limit is commonly
164: encountered in many applications \cite{Peter}. In this limit, being
165: sharp ($\xi\to 0$), the front dynamics can be described in terms of
166: the evolution of the surface (line in $2d$) which divides inert
167: ($\theta=1$) and fresh ($\theta=0$) material. The effect of the flow
168: is thus to wrinkle the front increasing (in two dimensions) its length
169: $\mathcal{L}_f$ and, as a consequence of the relation \cite{W85,Peter}
170: \begin{equation}
171: v_f = \frac{v_0 \mathcal{L}_f}{L}\,,
172: \label{eq:relation}
173: \end{equation}
174: (where $L$ is the length of a flat front in the absence of fluid
175: motion) its propagation velocity, i.e. $v_f>v_0$. Quantifying such an
176: enhancement is one of the main goals of, e.g., the community
177: interested in combustion propagation \cite{W85}. It should be also
178: remarked that the presence of complicated flow has also an important
179: role in the generation of patterns, i.e., front spatial
180: structures.\\\indent From a formal point of view the evolution of
181: $\theta$ can be recast in terms of the evolution of a scalar field
182: $G({\bf r},t)$, where the iso-line (in 2d) $G({\bf r},t)=0$ represents
183: the front, i.e. the boundary between inert ($G>0$) and fresh ($G<0$)
184: material. $G$ evolves according to the so-called $G$-equation
185: \cite{Peter,EMS95}
186: \begin{equation}
187: \partial_t G +\bm u\cdot \bm \nabla G=v_0 |\bm \nabla G|\,.
188: \label{eq:Geq}
189: \end{equation}
190: The analytical treatment of this equation is not trivial, and even in
191: relatively simple cases (e.g., shear flows) numerical analysis is
192: needed. Also on the numerical side solving (\ref{eq:Geq}) is a non
193: trivial issue, indeed the presence of strong gradients usually requires
194: the regularization of (\ref{eq:Geq}) through the introduction of a
195: diffusive term (see e.g. \cite{Ald}). Here, following
196: Ref.~\cite{CTVV03}, we adopt a Lagrangian integration scheme the basic
197: idea of which is now briefly sketched.\\\indent
198: \begin{figure}[b!]
199: \iffigs \includegraphics[scale=.45]{fig1}
200: \else \drawing 70 90 {Fig1.eps} \fi
201: \caption{\label{fig:1} Typical front patterns for a stationary shear
202: flow (a), time correlated shear flow with $\tau_f=200$ (b) and with
203: $\tau_f=2$ (c). In the stationary case $U=1/\sqrt{2}$, while in the time
204: correlated case we set $U_{rms}=\sqrt{2}$. In all cases the bare
205: velocity is $v_0=0.2$. For (a) and (b) we used $N_y=800$ grid
206: points and $N_y=3000$ for (c). Here and in the following figures $L=2\pi$.}
207: \end{figure}
208: First of all let us introduce the type of flow we are interested in,
209: we consider shear flows that can be written as
210: \begin{equation}
211: \bm u=(U(t)g(y),0)\,,
212: \label{eq:shear}
213: \end{equation}
214: being $g(y)$ the functional shape of the flow (here $g(y)=\sin (2\pi
215: y/L)$) and $U(t)$ its intensity. The domain of integration is chosen
216: as a stripe $[0:N\,L]\times[0:L]$, where $N$ (typically in the range
217: $5-20$) is the maximum number of cells of size $L$ in the
218: $x$-direction that are used in the integration (the number should be
219: fixed according to the front width). The number of cells $N$ is
220: dynamically adjusted. In particular, after the propagation sets in a
221: statistically stationary regime, while the front propagates the cells
222: on the left that are completely inert with $\theta=1$ are eliminated
223: by the integration domain. On the right side we retain only a finite
224: number (which depends on the maximum allowed speed) of cells with
225: fresh material $\theta=0$. The domain is discretized and the value of
226: $\theta$ in each point of the lattice is updated with the following
227: rule. At each time step, each grid point $\bm r_{n,m}=(x_n,y_m)$ is
228: backward integrated in time according to the advection by the flow
229: $d\bm r/dt=-\bm u$. Once the point $\bm r'$ that will arrive in $\bm
230: r_{n,m}$ at time $t$ is known, $\theta(\bm r_{n,m},t)$ is set to $1$
231: if in a circle centered in $\bm r'$ and having radius $v_0dt$ there is
232: at least one grid point with $\theta=1$. This is a straightforward way
233: to implement the Huygens dynamics.
234: The algorithm works as soon as $v_0dt$ is sufficiently larger than the
235: spatial discretization $dx=dy=L/N_y$ (where $N_y$ is the number of
236: grid points in the $y$-direction, and $N_x=N\,N_y$). For a detailed
237: description of the algorithm see the Appendix in Ref.~\cite{CTVV03}.
238: \begin{figure}[t!]
239: \iffigs \includegraphics[scale=.55]{fig2}
240: \else \drawing 70 40 {Fig2.eps} \fi
241: \caption{\label{fig:2} Normalized velocity $V_n=(v_f(t)-v_0)/U$ vs
242: $t/t_w$ for: $v_0=1$, $U=1$ (circles, green online), $v_0=0.2$, $U=1$
243: (squares, red online) and $v_0=0.5$, $U=1$ (triangles, blue
244: online). The corresponding $t_w$'s were numerically computed as
245: $W_f^\ast/2U$ obtaining $2.6$, $9.48$ and $18.96$, respectively
246: ($W_f^\ast$ is estimated by counting the number of pixel in the
247: border between inert and fresh material). The inset shows the unscaled
248: results. The resolution used is $N_y=800$.}
249: \end{figure}
250: For a stationary shear flow, i.e. $U(t)=U$, by means of
251: simple geometrical reasonings one can show that at long times the
252: front evolves with velocity \cite{ABP00}:
253: \begin{equation}
254: V_f=v_0+U\sup_{y} \{g(y)\}\,,
255: \label{eq:vel}
256: \end{equation}
257: which, with the choice of the sin flow, means $V_f=v_0+U$. Similarly
258: one can predict the asymptotic shape of the front, which is shown in
259: Fig.~\ref{fig:1}a. The important features are the presence of a
260: stationary (maximum) point in correspondence of the point where $g(y)$
261: has its maximum, and a cusp in its minimum. The asymptotic speed
262: (\ref{eq:vel}) is reached only after the transient time $t_w$
263: necessary to the front shape to reach its maximum length
264: (corrugation). Following \cite{KBM01} we call $t_w$ as the {\it
265: wrinkling } time, that can be defined as the time the front width
266: $\mathcal{W}_f$ (i.e. the distance between the leftmost point in which
267: $\theta=0$ and the rightmost in which $\theta=1$) employs to pass from
268: the initial zero-value (indeed at the beginning the front is flat) to
269: the asymptotic one $\mathcal{W}_f^\ast$. For $U \gg v_0$ this time can
270: be estimated as
271: \begin{equation}
272: t_w\propto L/v_0\,.
273: \label{eq:tw}
274: \end{equation}
275: This comes from the fact that starting from the flat profile the front
276: width $\mathcal{W}_f(t)$ grows in time as $2Ut$ up to the moment in
277: which the cusp (see Fig.~\ref{fig:1}a) is formed (see also
278: \cite{KBM01}). Then the growth slows down up to the stationary value
279: $\mathcal{W}_f^\ast$. Assuming the linear growth up to the end one may
280: estimate $t_w=\mathcal{W}_f^\ast/(2U)$. Further, since in the shear
281: flow case (where the formation of pockets of inert material is not
282: possible), for $U \gg v_0$, the width $\mathcal{W}_f^\ast$ is
283: proportional to the stationary front length $\mathcal{L}_f$, which is
284: linked to the asymptotic velocity by Eq.~(\ref{eq:relation}). Finally
285: since the latter given by (\ref{eq:vel}) one ends up with
286: $t_w=(L/U)(1+U/v_0)$ which reduces to (\ref{eq:tw}) for $U\gg v_0$. In
287: Fig.~\ref{fig:2} we show $v_f(t)$ as a function of $t/t_w$, as one can
288: see with this rescaling the asymptotic speed is reached at the same
289: instant for systems which have different $U$ and $v_0$, as the nice
290: collapse of the different curves indicates (compare with the
291: inset). We noticed that as soon as $U/v_0 \geq 4$ $t_w \propto L/v_0$
292: as predicted by Eq.~(\ref{eq:tw}). \\\indent The wrinkling time is an
293: inner time scale of the reaction dynamics, which is very important
294: when considering time-dependent flows. In particular, here we study
295: the example of random shear flows (\ref{eq:shear}) with
296: $g(y)=\sin(2\pi y/L)$ (as in the stationary case) and random
297: amplitudes $U(t)$ which are chosen according to an Ornstein-Uhlembeck
298: process. Therefore, $U$ evolves according to the Langevin dynamics
299: \begin{equation}
300: \frac{dU}{dt}=-\frac{U}{\tau_f}+\sqrt{\frac{2 U_{rms}^2}{\tau_f}}\: \eta
301: \end{equation}
302: where $\eta$ is a zero-mean Gaussian white
303: noise and $\tau_f$ defines the flow correlation time so that $\langle
304: U(t)U(t')\rangle= U^2_{rms} \exp(-|t-t'|/\tau_f)$. Clearly one has to
305: distinguish two limiting cases: i) when the flow fluctuations are
306: slower than the wrinkling time: $\tau_f\gg t_w$; ii) when they are
307: much faster: $\tau_f\ll t_w$.\\
308: \begin{figure}[t!]
309: \iffigs \includegraphics[scale=.55]{fig3}
310: \else \drawing 70 90 {Fig3.eps} \fi
311: \caption{\label{fig:3} Measured front velocity $v_f(t)$ (thick black
312: line) and the adiabatic prediction $v_0 + |U(t)|$ (solid line, red
313: online) versus $t/\tau_f$ for a correlated flow with $U_{rms}=1$,
314: $v_0=0.2$ and: (top) $\tau_f=200 (\gg t_w=31.4$), (bottom) $\tau_f (2 \ll
315: t_w=31.4)$. The resolution used was $N_y=800$ in the first case and
316: $N_y=3000$ in the second one.}
317: \end{figure}
318: $i)$ In this condition the front has enough time for adiabatically
319: adjust itself on the instantaneous flow velocity. Thus by generalizing
320: (\ref{eq:vel}) it is natural to expect that $v_f(t)=v_0+|U(t)|$ (as
321: confirmed in Fig.~\ref{fig:3}a) so that $V_f=v_0+\langle |U(t)|
322: \rangle$. In other words if the velocity fluctuations are slower than
323: the wrinkling time the front can be efficiently corrugated close to
324: the maximal wrinkled shape allowed by the flow and so by
325: (\ref{eq:relation}) can reach maximal speed.\\ $ii)$ On the other hand
326: in opposite limit $\tau_f\ll t_w$ the front has not time to be
327: maximally corrugated by the flow, and so its speed cannot reach the
328: maximal amplification allowed by the fluid. In this case it is not
329: anymore true that $v_f(t)=v_0+|U(t)|$ (see
330: Fig.~\ref{fig:3}b). \\\indent These effects on the propagation speed
331: have a counterpart in the patterns of the front. This is evident by
332: looking at the front shape (compare Fig.~\ref{fig:1}b and c with
333: a). Indeed while in the case $\tau_f \gg t_w$ at any instant the shape
334: of front closely resembles that obtained in the stationary case, when
335: $\tau_f \ll t_w$ one notices that the front length is strongly reduced
336: and the spatial structure complicated by the presence of more than one
337: cuspid.\\\indent Looking at Fig.~\ref{fig:3}b it is clear that the
338: reactive dynamics acts as a sort of filtering of the fluid velocity so
339: that not only the front speed is not enhanced at the maximal allowed
340: value but also its fluctuations are much decreased. In
341: Fig.~\ref{fig:4} we show the normalized front speed
342: $V_n=(V_f-v_0)/\langle|U(t)|\rangle$ and the normalized variance
343: $\sigma_n=\sigma_f\/(\sqrt{\langle \vert U(t) \vert^2\rangle-\langle
344: \vert U(t)\vert\rangle^2}$ (i.e. the standard deviation of the
345: $v_f(t)$ normalized by that expected on the basis of the adiabatic
346: process $|U(t)|$), by fixing the flow intensity $U_{rms}$ and varying
347: the correlation time. Note that in the limit of very long correlation
348: times $V_n\approx 1$ and $\sigma_n\approx 1$. As one can see a fast
349: drop of the front speed and average fluctuations with respect to its
350: maximum value is observed when $\tau_f/t_w<1$, confirming the above
351: picture.\\\indent
352: \begin{figure}[t!]
353: \iffigs \includegraphics[scale=.6]{fig4}
354: \else \drawing 70 40 {Fig4.eps} \fi
355: \caption{\label{fig:4} Normalized velocity $V_n=(V_f-v_0)/\langle
356: \vert U \vert \rangle$ as a function of $\tau_f/t_w$ for $v_0=0.2$ and
357: $U_{rms}=2\sqrt{2}$ (circles, green online), $U_{rms}=\sqrt{2}$
358: (squares, red online) and $U_{rms}=1/\sqrt{2} $ (triangles, blue
359: online). The inset displays the normalized variance $\sigma_n$ in the
360: same cases. The resolution used goes from $N_y=3000$ (for the lowest
361: value of $\tau_f$) up to 800 (for the highest one).}
362: \end{figure}
363: These results along with those of Refs.~\cite{D99,A00,KBM01} confirm
364: the importance of time correlations in the flow in determining the
365: front speed. This may be relevant to more realistic flows in the light
366: of the above cited bending phenomenon. Indeed in turbulent flows one
367: has that at increasing the turbulence intensity fluctuations on
368: smaller and smaller scales appear. These are characterized by faster
369: and faster characteristic time scales. In this respect, as suggested
370: in by the results of this work, one may expect that in the corrugation by
371: these scales become less and less important, so that the average front
372: speed may increase less than expected.\\
373: \indent We conclude by noticing that it would be very interesting to test the
374: effects of time-correlations on the front propagation also in other
375: kind of laminar flows. In particular this could be performed in
376: experimental studies in settings similar to those of
377: Refs.~\cite{LMRS05} where flows of the form $\bm u(\bm r,t)=U(t)\bm
378: v(\bm r)$ can be easily generated with a good control of the
379: time dependence of the amplitude.\\
380: \indent We are grateful to C. Casciola for useful discussions. MC and AV
381: acknowledge partial support by MIUR Cofin2003 ``Sistemi Complessi e
382: Sistemi a Molti Corpi'', and EU under the contract HPRN-CT-2002-00300.
383:
384:
385: \begin{thebibliography}{99}
386:
387: \bibitem{W85} F.~A.~Williams, {\it Combustion Theory}
388: (Benjamin-Cummings, Menlo Park 1985).
389:
390: \bibitem{chem} J.~Ross, S.~C.~M\"uller, and C.~Vidal, Science
391: \textbf{240}, 460 (1988); I.~R.~Epstein, Nature \textbf{374}, 231
392: (1995).
393:
394: \bibitem{bio}E.~R.~Abraham, Nature \textbf{391}, 577 (1998);
395: E.~R.~Abraham et al., Nature \textbf{407}, 727 (2000).
396:
397:
398: \bibitem{ABP00} B.~Audoly, H.~Beresytcki and Y.~Pomeau,
399: C. R. Acad. Sci. \textbf{328}, S\'erie II b, 255 (2000).
400:
401:
402: \bibitem{Const} P.~Constantin, A.~Kiselev, A.~Oberman and L.~Ryzhik,
403: Arch. Rational Mechanics \textbf{154}, 53 (2000).
404:
405:
406:
407: \bibitem{KBM01}B.~Khouider, A.~Bourlioux and A.J.~Majda,
408: Comb. Th. Model. \textbf{5} 295 (2001).
409:
410: \bibitem{ACVV01} M.~Abel, A.~Celani, D.~Vergni and A.~Vulpiani,
411: Phys. Rev. E \textbf{64}, 046307 (2001).
412:
413: \bibitem{ACVV02} M.~Abel, M.~Cencini, D.~Vergni and A.~Vulpiani
414: Chaos \textbf{12}, 481 (2002).
415:
416:
417: \bibitem{CTVV03} M.~Cencini, A.~Torcini, D.~Vergni and A.~Vulpiani
418: Phys. Fluids \textbf{15}, 679 (2003).
419:
420:
421: \bibitem{NX03}J.~Nolen and J.~Xin, SIAM J. Multiscale Modeling and Simulations
422: \textbf{1}, 554 (2003).
423:
424: \bibitem{LMRS03} M. Leconte, J. Martin, N. Rakotomalala, and D. Salin,
425: Phys. Rev. Lett., \textbf{90} 128302 (2003).
426:
427:
428: \bibitem{LMRS05} M.~Leconte, J.~Martin, N.~Rakotomalala and D.~Salin,
429: ``Autocatalitic reaction front in a pulsating periodic flow'', preprint
430: arXiv:physics/0504012.
431:
432: \bibitem{SS05} M. S. Paoletti and T. H. Solomon, Europhys. Lett.,
433: \textbf{69}, 819 (2005).
434:
435: \bibitem{PH04}A. Pocheau and F. Harambat,
436: in proceedings of the 21th international congress of theoretical an
437: applied mechanics, ICTAM 2004.
438:
439:
440: \bibitem{D99}B.~Denet, Comb. Th. Model. \textbf{3}, 585 (1999).
441:
442: \bibitem{A00} W.T.~Ashurst, Comb. Th. Model \textbf{4}, 99 (2000).
443:
444: \bibitem{Ronney} P.~D.~Ronney,
445: in {\it Modeling in Combustion Science}, pp. 3-22, Eds. J. Buckmaster
446: and T. Takeno (Springer-Verlag Lecture Notes in Physics, 1994).
447:
448: \bibitem{quench} L.~Kagan and G.~Sivashinshy, Combust, Flame {\bf 120}, 222
449: (2000).
450: \bibitem{pockets} J.~Zhu and P.D.Ronney, Combus. Sci. Technol. {\bf 100}, 183
451: (1994).
452:
453:
454: \bibitem{saarloos} W.~van Saarloos, Phys. Rep., \textbf{386}, 29 (2003).
455:
456:
457: \bibitem{Peter} N.~Peters, {\it Turbulent combustion} (Cambridge
458: University Press, 2000).
459:
460:
461:
462: \bibitem{EMS95} P.~F.~Embid, A.~J.~Majda and P.~E.~Souganidis,
463: Phys. Fluids \textbf{7} (8), 2052 (1995).
464:
465:
466: \bibitem{Ald} R.~C.~Aldredge, Comb. and Flame {\bf 106}, 29 (1996).
467:
468: \end{thebibliography}
469:
470: \end{document}
471:
472: