nlin0005049/spot.tex
1: \tolerance=10000
2: %\documentstyle[preprint,aps]{revtex}
3: \documentstyle[prl,psfig,twocolumn,aps]{revtex}
4: \begin{document}
5: 
6: %\twocolumn[\hsize\textwidth\columnwidth\hsize\csname@twocolumnfalse\endcsname
7: \title{Nonlocal Boundary Dynamics of Traveling Spots in a Reaction-Diffusion 
8: System}
9: \author{L.M. Pismen \\
10: {\it Department of Chemical Engineering and 
11: Minerva Center for Nonlinear Physics of Complex Systems,\\
12: Technion -- Israel Institute of Technology, 32000 Haifa, Israel}} 
13: \date{\today}
14: \maketitle
15: 
16: \begin{abstract}
17: The boundary integral method is extended to derive closed integro-differential 
18: equations applicable to computation of the shape and propagation speed of a 
19: steadily moving spot and to the analysis of dynamic instabilities in the sharp 
20: boundary limit. Expansion of the boundary integral near the locus of traveling 
21: instability in a standard reaction-diffusion model proves that the bifurcation 
22: is supercritical whenever the spot is stable to splitting, so that propagating 
23: spots can be stabilized without introducing additional long-range variables. 
24: \end{abstract}
25: 
26: \pacs{82.40.Bj, 82.20.Mj, 05.65.+b}
27: 
28: Localized structures in non-equlibrium systems (dissipative solitons) have been 
29: studied both in experiments and computations in various applications, including 
30: chemical patterns in solutions \cite{swin} and on surfaces \cite{imb}, gas 
31: discharges \cite{pur99} and nonlinear optics \cite{firth}. 
32: The interest to dynamic solitary 
33: structures, in particular, in optical \cite{firth} and gas discharge systems  
34: \cite{purw} has been recently driven by their possible role in information 
35: transmission and processing. 
36: 
37: A variety of observed phenomena can be reproduced qualitatively with the help of 
38: simple reaction-diffusion models with separated scales 
39: \cite{ohta,keros,meron,pi94,osmur}. Extended models of this type included 
40: nonlocal interactions due to gas transport \cite{pi94,mikh}, Marangoni flow 
41: \cite{pi97} or optical feedback \cite{firth,piop}. A great advantage of scale 
42: separation is a possibility to construct analytically strongly nonlinear 
43: structures in the sharp interface limit. An alternative approach based on 
44: Ginzburg--Landau models supplemented by quintic and/or fourth-order differential 
45: (Swift-Hohenberg) terms \cite{rab} have to rely on numerics in more than one 
46: dimension. 
47: 
48: Dynamical solitary structures are most interesting from the point of view of 
49: both theory and potential applications. Existence of traveling spots in 
50: sharp-interface models is indicated by translational 
51: instability of a stationary spot 
52: \cite{mikh}. This instability is a manifestation of a general phenomenon of parity 
53: breaking (Ising--Bloch) bifurcation \cite{coul,hagmeron} which takes a single 
54: stable front into a pair of counter-propagating fronts forming the front and the 
55: back of a traveling pulse. Numerical simulations, however, failed to produce 
56: stable traveling spots in the basic activator-inhibitor model, and the tendency 
57: of moving spots to spread out laterally had to be suppressed either by global 
58: interaction in a finite region \cite{mikh} or by adding an extra inhibitor with 
59: specially designed properties \cite{bode}.
60: 
61: The dynamical problem is difficult for theoretical study, since a moving spot 
62: loses its circular shape, and a free-boundary problem is formidable even for 
63: simplest kinetic models. Numerical simulation is also problematic, due to the 
64: need to use fine grid to catch sharp gradients of the activator; therefore 
65: actual computations were carried out for moderate scale ratios.  
66: A large amount of numerical data, such as the inhibitor field far from the spot 
67: contour, is superfluous. This could be overcome if it was possible 
68: to reduce the PDE solution to 
69: local dynamics of a sharp boundary. Unfortunately, a purely local equation of 
70: front motion \cite{hagmeron} is applicable only when the curvature far exceeds 
71: the diffusion scale of the long-range variable, whereas a spot typically  
72: suffers splitting instability \cite{ohta} before growing so large. On the other 
73: hand, the nonlocal boundary integral method \cite{gope} is applicable only when 
74: the inhibitor dynamics is fast compared to the characteristic propagation scale 
75: of a front motion, i.e. under conditions when no dynamic instabilities arise and
76: traveling spots do not exist.
77: 
78: It is the aim of this Letter, to extend the nonlocal boundary integral method  
79: to dynamical problems, and to find out with its help conditions of supercritical 
80: bifurcation for steadily moving spots. We consider the standard FitzHugh--Nagumo 
81: model including two variables -- a short-range activator $u$ and a long-range 
82: inhibitor $v$:
83:   \begin{eqnarray}
84: \epsilon^2 \tau u_t & = & \epsilon^2 \nabla^2 u + V'(u) - \epsilon v, 
85:       \label{sueq}   \\
86:  v_t & = & \nabla^2 v - v - \nu  + \mu u.
87:     \label{sveq}   \end{eqnarray}
88: Here $V(u)$ is a symmetric double-well potential with minima at $u=\pm 1$; 
89: $\epsilon \ll 1$ is a scale ratio, and other parameters are scaled in such a way 
90: that the effects of bias and  curvature on the motion of the front separating 
91: the up- and down states of the short-range variable are of the same order of 
92: magnitude. The local {\it normal} velocity of the front is 
93: \begin{equation}
94: c_n = \tau^{-1}(b v -\kappa) + O(\epsilon), 
95: \label{eqmot} \end{equation}
96: where $\kappa$ is curvature and 
97: $b$ is a numerical factor dependent on the form of $V(u)$; for example, 
98: $b=3/\sqrt{2}$ for the quartic potential $V(u)= -\frac{1}{4}(1-u^2)^2$. By 
99: definition, the velocity is positive when the down-state $u<0$ advances.
100: 
101: In the sharp boundary approximation valid at $\epsilon \ll 1$, a closed equation 
102: of motion for a solitary spot propagating with a constant speed can be written by 
103: expressing the local curvature in Eq.~(\ref{eqmot}) with the help of a suitable 
104: parametrization of the spot boundary, and resolving Eq.~(\ref{sveq}) rewritten 
105: in a coordinate frame propagating with a speed $c$ (as yet unknown). It is 
106: convenient to shift the long-range variable $v= w- \nu +\mu$, so that 
107: $w(\infty)=0$ when the up-state $u=1-O(\epsilon)$ prevails at infinity. The 
108: stationary equation of $w$ in the coordinate frame translating with the speed 
109: {\bf c} is 
110: \begin{equation}
111:   {\bf c} \cdot \nabla w + \nabla^2 w - w = 2\mu H,
112: \label{sweq}  \end{equation}
113: where, neglecting $O(\epsilon)$ corrections, $H=1$ inside and $H=0$ outside the 
114: spot. The solution can be presented in the form of an integral over the spot 
115: area $\cal S$:
116: \begin{equation}
117: w({\bf x} ) =  -\frac{\mu}{\pi}  \int_{\cal S} {\cal G}({\bf x}-\mbox{\boldmath 
118: $\xi$}) d^2 \mbox{\boldmath $\xi$},
119: \label{wint}  \end{equation}
120: where the kernel ${\cal G}$ contains a modified Bessel function $K_0$:
121: \begin{equation}
122: {\cal G}({\bf r})  = \frac{1}{2\pi} e^{-\frac{1}{2}{\bf c \cdot r}} 
123:   K_0\left( |{\bf r}|\sqrt{1+\mbox{$\frac{1}{4}c^2$}}\right) .
124: \label{wker}  \end{equation}
125: This integral can be transformed into a contour integral with the help of the 
126: Gauss theorem. To avoid divergent expressions, the contour should exclude the 
127: point ${\bf x}=\mbox{\boldmath $\xi$}$. Clearly, excluding an infinitesimal 
128: circle around this point does not affect the integral (\ref{wint}), since the 
129: kernel (\ref{wker}) is only logarithmically divergent. Replacing ${\cal G}({\bf 
130: r}) =\nabla^2 {\cal G}({\bf r}) + {\bf c} \cdot \nabla {\cal G}({\bf r})$ $({\bf r} 
131: \neq 0)$, we transform the integral in Eq.~(\ref{wint}) as
132: \begin{eqnarray}
133: -\int_{\cal S} {\cal G}({\bf x}-\mbox{\boldmath $\xi$}) 
134: d^2 \mbox{\boldmath $\xi$} =
135: \int_{\cal S} \nabla_{\mbox{\boldmath $\xi$}} \cdot 
136: {\bf H} ({\bf x}-\mbox{\boldmath $\xi$}) d^2 \mbox{\boldmath $\xi$} \cr
137:  = \oint_{\Gamma'} {\bf n}(s) \cdot {\bf H}
138: ({\bf x}-\mbox{\boldmath $\xi$}(s)) ds, 
139: \label{gauss}  \end{eqnarray}
140: where ${\bf H}({\bf r}) = \nabla{\cal G}({\bf r}) + {\bf c} {\cal G}({\bf r})$ 
141: and {\bf n} is the normal to the contour $\Gamma'$. The vector Green's function  
142: {\bf H} corresponding to the kernel in Eq.~(\ref{wint}) is computed as
143: \begin{eqnarray}
144:  {\bf H}({\bf r}) &=& e^{-\frac{1}{2}{\bf c}\cdot {\bf r}} 
145: \left[\mbox{$\frac{1}{2}$}{\bf c}
146: K_0\left( |{\bf r}| \sqrt{1+\mbox{$\frac{1}{4}$} c^2}\right) \right. \cr
147: &-& \left. \sqrt{1+\mbox{$\frac{1}{4}$} c^2} \frac{{\bf r}}{|{\bf r}|}
148: K_1\left( |{\bf r}| \sqrt{1+\mbox{$\frac{1}{4}$} c^2}\right) \right].
149: \label{gauss2}  \end{eqnarray}
150: When ${\bf x}$ is a boundary point, $\Gamma'$ consists of the spot boundary 
151: $\Gamma$ cut at this point and closed by an infinitesimally small semicircle 
152: about ${\bf x}$. The integral over the semicircle equals to $\pi$. Defining the 
153: external normal to $\Gamma$ as the tangent ${\bf t}={\bf x}'(s)$ rotated 
154: clockwise by $\pi/2$, the required value of the long-range variable on the spot 
155: boundary (parametrized by the arc length $s$ or $\sigma$) is expressed,
156: using the 2D cross product $\times$, as
157: \begin{equation}
158:  v (s) = -\nu + \frac{\mu}{\pi}  \oint_{\Gamma} 
159:    {\bf H} ({\bf x}(s)- {\bf x} (\sigma)) \times {\bf x}' (\sigma) d\sigma .
160: \label{woint}  \end{equation}
161: 
162: To obtain a closed integral equation of a steadily moving spot, it remains to 
163: define a shift of parametrization accompanying shape-preserving translation. 
164: Recall that Eq.~(\ref{eqmot}) determines the propagation velocity $c_n$ along 
165: the {\it normal} to the boundary. In addition, one can introduce arbitrary {\it 
166: tangential} velocity $c_t$ which has no physical meaning but might be necessary 
167: to account for the fact that each ``material point'' on a translated contour is, 
168: generally, mapped onto a point with a different parametrization even when the 
169: shape remains unchanged. The tangential velocity can be defined by requiring 
170: that each material point be translated strictly parallel to the direction of 
171: motion, i.e. $c_n{\bf n}+c_t{\bf t}={\bf c}$. Taking the cross product with 
172: ${\bf c}$ yields $c_t= c_n ({\bf c} \times {\bf t})$ $/ ({\bf c} \cdot {\bf 
173: t}).$ Then eliminating $c_t$ gives the normal velocity $c_n = {\bf c} \times 
174: {\bf t}$ necessary for translating the contour along the $x$ axis with the 
175: velocity $c$. Using this in Eq.~(\ref{eqmot}) yields the condition of stationary 
176: propagation
177: \begin{equation}
178: {\bf c} \times {\bf x}'(s) = \tau^{-1} [b v(s)- \kappa(s)] .
179:   \label{shape}  \end{equation}
180: 
181: The form and the propagation speed of a slowly moving and weakly distorted 
182: circular contour can be obtained by expanding Eq.~(\ref{shape}) in $c=|{\bf c}|$ 
183: near the point of traveling bifurcation $\tau=\tau_0$, which is also determined 
184: in the course of the expansion. For a circular contour with a radius $a$, 
185: Eq.~(\ref{woint}) takes the form 
186: \begin{eqnarray}
187:  v(\phi) &=& -\nu + \frac{\mu a}{\pi}  \int_{0}^{2\pi} 
188:    e^{-\frac{1}{2} c a(\cos \phi- \cos \varphi)} \; \times \cr
189: && \left[ \mbox{$\frac{1}{2}$} c \cos \varphi \,
190:  K_0\left((2a\sqrt{1+\mbox{$\frac{1}{4}c^2$}}
191: \sin \mbox{$\frac{1}{2}$} |\phi-\varphi| \right) \right. \cr
192: & & + \sin \mbox{$\frac{1}{2}$}|\phi-\varphi|
193: \sqrt{1+\mbox{$\frac{1}{4}c^2$}}\, \; \times \cr 
194: && \left. K_1\left(2a\sqrt{1+\mbox{$\frac{1}{4}c^2$}}
195: \sin \mbox{$\frac{1}{2}$} |\phi-\varphi| \right)  \right] d \varphi,
196: \label{woint1}  \end{eqnarray}
197: where $\phi$ or $\varphi$ is the polar angle counted from the direction of 
198: motion. The angular integrals that appear in the successive terms of the 
199: expansion are evaluated iteratively, starting from $\Phi_0(a)= \pi I_0(a) 
200: K_0(a)$ and using the relations
201: \begin{eqnarray*}
202: \Psi_k(a) &=& \int_{0}^{\pi} \sin^{2k+1} \frac{\phi}{2}\,
203:   K_1\left( 2 a\sin \frac{\phi}{2}\right) d \phi = 
204: - \frac{1}{2}\frac{d \Phi_k}{da}, \cr 
205: \Phi_k (a) &=& \int_{0}^{\pi}\!\! \sin^{2k} \frac{\phi}{2}\,
206:   K_0 \!\left( 2 a\sin \frac{\phi}{2} \right) \! d\phi = 
207: -\frac{1}{2a}\, \frac{d(a\Psi_{k-1})}{da}. 
208: \end{eqnarray*}
209: 
210: 
211: Effect of small boundary distortions on $v$ can be computed directly with the 
212: help of Eq.~(\ref{wint}), where the integration should be carried out only over 
213: a small area swept by the displaced spot boundary. This approach is most useful 
214: for stability analysis with respect to small perturbations of a known static 
215: shape, and is easier than using the expansion of Eq.~(\ref{woint}) with a 
216: perturbed boundary. For a circular spot, we expand the perturbations of both $v$ 
217: and $\rho$ in the Fourier series 
218: \begin{eqnarray}
219: \widetilde \rho(\phi,t) &=& \rho(\phi)-a =
220:  \sum_{n \geq 2} c^n a_n e^{\lambda_n t}\cos n\phi , \cr
221:  \widetilde v(\phi,t) &=& 
222: \sum_{n \geq 2} \widehat v_n e^{\lambda_n t}\cos n\phi .
223: \label{expand}  \end{eqnarray}
224:  The curvature is expressed as
225: \begin{eqnarray}
226: \kappa(\phi) &=& \frac{\rho^2 -2 \rho^2_\phi - 
227:           \rho\rho_{\phi\phi}}{(\rho^2+\rho^2_\phi)^{3/2}} \cr
228: &=& a^{-1}+ 3(c/a)^2 a_2 e^{\lambda_2 t}\cos 2\phi + O(c^3). 
229: \label{curv}  \end{eqnarray}
230: 
231: Since the displaced point should remain on the boundary, the distortion 
232: $\widetilde \rho(\varphi)$ should be compensated by rigid displacement of the 
233: spot by an increment $\widetilde \rho(\phi)$ when $\widetilde v(\phi)$ is 
234: computed (see the inset in Fig.~\ref{f1}). The resulting equation for 
235: eigenvalues $\lambda_n$ following from Eq.~(\ref{eqmot}) is
236: \begin{eqnarray}
237: && \tau \lambda_n = \frac{ n^2-1}{a^2} - \frac{4ab\mu}{\pi^2} 
238: \int_0^{\pi} \cos n\phi \, d \phi \; \times \cr
239: && \int_0^\pi  [\widetilde \rho(\varphi)- \widetilde \rho(\phi)
240: \cos (\varphi-\phi)] 
241: e^{-\frac{1}{2} c a(\cos \phi- \cos \varphi)}\; \times \cr
242: && K_0\left(2a\sqrt{1+ \lambda_n +\mbox{$\frac{1}{4}c^2$}}
243: \sin \mbox{$\frac{1}{2}$} |\phi-\varphi| \right) d \varphi.
244: \label{wint1}  \end{eqnarray}
245: 
246: 
247: \begin{figure}[b]
248: \centerline{\hspace{.0cm} \psfig{figure=f1.eps,width=8.5cm}}
249: \caption{
250: The bifurcation diagram for stationary spots at $\tau=1$. C -- existence 
251: boundary, S -- locus of splitting instability. The stability region is bounded by 
252: the locus of breathing instability B, branching off at the point of double zero 
253: eigenvalue D, and the locus of traveling instability T. Inset: a circular spot 
254: distorted by second and third harmonics with amplitudes proportional to $c^n$. 
255: The shape is characteristic to a spot propagating to the right, and the 
256: amplitudes are chosen in such a way that the curvature on the back side 
257: vanishes. The center of the gray circle 
258: is shifted from the black to the gray spot to compensate 
259: the distortion at $\phi=0$, so that the integral   
260: is taken over the area between the black contour and the gray circle when the 
261: effect of small distortions on the $v$ field at this point is computed. 
262: \label{f1}} 
263: \end{figure}
264: 
265: Using the constant zero-order term in the expansion of Eq.~(\ref{woint1}) 
266: together with $\kappa= a^{-1}$ in Eq.(\ref{eqmot}) yields the stationarity 
267: condition 
268: \begin{equation}
269:  \nu = -(ba)^{-1} + \mu a \left[ K_1(a)\,I_0(a) - K_0(a)\,I_1(a) \right] .
270: \label{cstat} \end{equation}
271: A stationary solution stable against collapse or uniform swelling exists in the 
272: region in the parametric plane $\mu,\nu$ (Fig.~\ref{f1}) bounded by the cusped 
273: curve C and the axis $\nu=0,\, \mu>2/b$. This curve is drawn as a parametric 
274: plot with $\nu(a)$ given by Eq.~(\ref{cstat}) and $\mu(a)$ by Eq.~(\ref{wint1}) 
275: with $c, \, n$ and $\lambda_0$ set to zero (or, equivalently, by the condition 
276: $F_0'(a)=0$, where $F_0(a)$ is the right-hand side of Eq.~(\ref{cstat}).
277: 
278: The first-order term in the expansion of Eq.~(\ref{woint1}) is proportional to 
279: $\cos \phi$, and should compensate at the traveling bifurcation point the 
280: left-hand side of Eq.~(\ref{shape}). This yields the bifurcation condition 
281: \begin{equation}
282: \tau_0  =  b\mu a [a(I_1(a)K_0(a) - I_0(a)K_1(a)) + 2I_1(a)K_1(a)] ,
283:   \label{disp1d} \end{equation}
284: which coincides with the known result obtained by other means \cite{mikh}.
285: The curve T in Fig.~\ref{f1} shows the traveling instability threshold for 
286: $\tau_0  =1$. The static spot is unstable below this curve; the locus shifts up 
287: (to smaller radii) as $\tau$ decreases, and exits the existence domain 
288: at $\tau<1/4$. At $\tau>1$, the dominant instability at large radii is a static 
289: splitting instability. Its locus, determined by Eq.~(\ref{wint1}) with $n=2$ and 
290: $c=\lambda_2=0$, is the curve S in Fig.~\ref{f1}.
291:  
292: Another possible dynamic instability is breathing instability 
293: \cite{ohta,haim,pur99}. Its locus is given by Eq.~(\ref{wint1}) with $c=n=0$ and 
294: $\lambda_0=i\omega$. The frequency $\omega$ as a function of the spot radius $a$ 
295: is computed by solving the equation $\tau \omega =a^{-2}$Im $F(a,\omega)/$Re  
296: $F(a,\omega)$, where $F(a,\omega)$ is the right-hand side of Eq.~(\ref{wint1}) 
297: computed as
298: \begin{eqnarray}
299: F(a,\omega)  &=& 2 \mu a \left[I_1\left(a\sqrt{1+i\omega}\right)
300: K_1\left(a\sqrt{1+i\omega}\right) \right. \cr
301: &-& \left. I_0\left(a\sqrt{1+i\omega}\right)
302: K_0\left(a\sqrt{1+i\omega}\right) \right] . 
303:   \label{dis00} \end{eqnarray}
304: The curve B in Fig.~\ref{f1} shows the bifurcation locus at $\tau=1$. The 
305: instability region retreats to small radii (large $\nu$) at large $\tau$ and 
306: spreads downwards as $\tau$ decreases. The balloon of stable solutions disappears 
307: altogether at $\tau<0.5$ after the tips of both dynamic loci meet on the 
308: existence boundary.
309: 
310: In the second order, Eq.~(\ref{woint1}) yields a constant term 
311: \begin{equation}
312: v^{(2,0)}  = - \mu a^2 [a(I_1(a)K_0(a) - I_0(a)K_1(a)) + I_1(a)K_1(a)]  
313:   \label{dis20} \end{equation}
314: and a dipole term $ v^{(2,2)} = q^{(2,2)} \cos 2\phi$, where
315: \begin{eqnarray}
316: q^{(2,2)} = \mbox{$\frac{1}{4}$} \mu a^2 
317:  [a(I_0(a)K_1(a) - I_1(a)K_0(a)) \cr
318: - 3I_1(a)K_1(a) +2 I_2(a)K_2(a)] .
319:   \label{dis22} \end{eqnarray}
320: The constant term is positive and causes contraction of the average radius of 
321: the moving spot by an increment $\widetilde a = - a^2 c^2 bv^{(2,0)}$. 
322: 
323: The second-order dipolar term in the right-hand side of Eq.~(\ref{shape}), 
324: $\widetilde v^{(2,2)} = \widetilde q^{(2,2)} a_2 \cos 2\phi$, as well as the 
325: third-order first harmonic term, $\widetilde v ^{(3,1)} = \widetilde q^{(3,1)} 
326: a_2 \cos \phi$, needed for the solvability condition to follow, are read from 
327: Eq.~(\ref{wint1})  with $n=2$ and $\lambda_2=0$, respectively, in zero and first 
328: order in $c$:
329: \begin{eqnarray}
330: \widetilde q^{(2,2)} &=&  - 3a^{-2} + 2b\mu [I_1(a)K_1(a)- I_2(a)K_2(a)],
331:   \label{disp22} \\
332: \widetilde q^{(3,1)} &=& b\mu a^2 I_1(a)K_1(a) .
333:   \label{disp31} \end{eqnarray}
334: The coefficient $\widetilde q^{(2,2)}$ vanishes at the splitting instability 
335: threshold (curve S in Fig.~\ref{f1}), and must be negative when the circular 
336: spot is stable. Consequently, the distortion amplitude is $a_2 = - q^{(2,2)}/ 
337: \widetilde q^{(2,2)} <0$, so that the dipole term causes contraction of the 
338: moving spot in the direction of motion and expansion in the normal direction. 
339: 
340: Continuing the expansion to the third order, we compute the first harmonic term 
341: contributing to the solvability condition. The latter has the form $ \widetilde 
342: \tau c = k c^3$, where $ \widetilde \tau = \tau-\tau_0$ and the coefficient $k$
343: determining the character of the bifurcation is computed as
344: \begin{equation}
345: k = b\mu \left( q^{(3,1)} - \tau_0'(a) a^2 v^{(2,0)} 
346: - \widetilde q ^{(3,1)} q ^{(2,2)}/ \widetilde q ^{(2,2)}\right).
347:   \label{disp3} \end{equation}
348: The first term is the coefficient at the first harmonic in the third order of 
349: the expansion of Eq.~(\ref{woint1}). The second term takes into account the 
350: second-order radius correction to the first-order first harmonic term. The last 
351: term gives the effect of dipolar shape distortion; it becomes dominant when the 
352: locus of splitting instability is approached. Stable traveling solution should 
353: be observed beyond the traveling instability threshold, i.e.\ at $\widetilde 
354: \tau <0$; hence, the condition of supercritical bifurcation is $k<0$. The 
355: numerical check of the symbolically computed expression shows that the traveling 
356: bifurcation is always supercritical when the spot is stable to splitting. The
357: traveling solution bifurcating supercritically must be stable, at least close to the
358: bifurcation point where it inherits stability of the stationary spot to other kinds
359: of perturbations.
360: 
361: The third harmonic term that appears in the third order of the expansion 
362: delineates, together with the second-order dipolar term, the characteristic 
363: shape of a translating spot, pointed in the direction of motion and spread 
364: sidewise, as in the inset in Fig.~\ref{f1}, 
365: which has been also observed in numerical 
366: simulations \cite{bode}. Beyond the range of the bifurcation expansion, the 
367: shape, as well as the propagation speed can be determined by solving numerically 
368: Eq.~(\ref{shape}) with $v(s)$ given by Eq.~(\ref{gauss2}) and curvature computed
369: using  the 
370: fully nonlinear expression in Eq.~(\ref{curv}). Although the boundary integral 
371: method reduces a PDE to a 1D integro-differential equation, the equation is 
372: rather difficult. Iterative numerical solution \cite{dima} tends to break down 
373: rather close to the bifurcation point, as soon as the shape distortion becomes 
374: strong enough to flatten the spot at the back side. Since the boundary integral 
375: equation is non-evolutionary, there is no way to distinguish between a purely 
376: numerical failure of convergence and a physical instability  that would lead to 
377: lateral spreading observed in PDE simulations \cite{mikh}. 
378: 
379: The above bifurcation expansion proves that a stable traveling solution does 
380: exist in the basic model (\ref{sueq}), (\ref{sveq}) in the sharp boundary limit. 
381: The result is applicable at $1 \gg c\gg \sqrt{\epsilon}$. It can be extended 
382: straightforwardly to models with more than one long-range variable, provided all 
383: long-range equations are linear. Stable traveling spot solutions should be, 
384: indeed, more robust in an extended model where they have been obtained in PDE 
385: simulations \cite{bode}, whereas in the basic model they require fine parametric 
386: tuning aided by the analytical theory.
387: 
388: \paragraph*{Acknowledgement.} 
389: This work has been supported by the German--Israeli Science
390: Foundation. 
391: 
392: \begin{thebibliography}{9}
393: \bibitem{swin} G.~Li, Q.~Ouyang, and H.L.~Swinney, J. Chem. Phys. {\bf 105}, 
394: 10830 (1996).
395: \bibitem{imb} G.~Haas , M.~B\"ar, I.G.~Kevrekidis, P.B.~Rasmussen, 
396: H.-H.~Rotermund , and G.~Ertl, Phys. Rev. Lett. {\bf 75}, 3560 (1995).
397: \bibitem{pur99} I. M\"uller, E.~Annelt and H.-G.~Purwins, Phys. Rev. Lett. {\bf 
398: 82}, 3428 (1999).
399: \bibitem{firth} W.J.~Firth and A.J.~Scroggie, \prl {\bf 76}, 1623 (1996).
400: \bibitem{purw} L.M.~Portsel, Yu.A.~Astrov, I.~Reimann, E.~Annelt and 
401: H.-G.~Purwins, J. Appl. Phys. {\bf 85}, 3960 (1999).
402: \bibitem{ohta} T.~Ohta, M.~Mimura, and R.~Kobayashi, Physica (Amsterdam) {\bf D 
403: 34} 115 (1989).
404: \bibitem{keros} B.S.~Kerner and V.V.~Osipov, Usp. Fiz. Nauk. {\bf 157}, 201 
405: (1989) [Sov. Phys. Usp. {\bf 32}, 101 (1989)].
406: \bibitem{meron} E.~Meron, Phys. Rep. {\bf 218}, 1 (1992).
407: \bibitem{pi94} L.M.~Pismen, J. Chem. Phys. {\bf 101} 3135 (1994).
408: \bibitem{osmur} C.B.~Muratov and V.V.~Osipov, Phys. Rev. {\bf E 53}, 3101 
409: (1996). 
410: \bibitem{mikh} K.~Krischer and A.~Mikhailov, Phys. Rev. Lett. {\bf 73}, 3165 
411: (1994) 
412: \bibitem{pi97} L.M.~Pismen, Phys. Rev. Lett. {\bf 78}, 382 (1997).
413: \bibitem{piop} L.M.~Pismen, Phys. Rev. {\bf 75}, 228 (1995).
414: \bibitem{rab} I.S.~Aranson, K.A.~Gorshkov, A.S.~Lomov and  M.I.~Rabinovich, 
415: Physica (Amsterdam) {\bf D 42}, 435 (1990);
416: W.~van Saarloos and P.C.~Hohenberg, Physica (Amsterdam) {\bf D 56}, 303 (1992);
417: H.~Sakaguchi and H.R.~Brand, Physica {\bf D 97}, 274 (1996);
418: K.~Ouchi and H.~Fujisaka, Phys. Rev. {\bf E 54}, 3895 (1996). 
419: \bibitem{coul} P.~Coullet , J.~Lega, B.~Houchmanzadeh, and J.~Lajzerowicz, Phys. 
420: Rev. Lett. {\bf 65}, 1352 (1990). 
421: \bibitem{hagmeron} A.~Hagberg and E.~Meron, Nonlinearity {\bf 7}, 805 (1994).
422: \bibitem{bode} C.P.~Schenk, M.~Or-Guil, M.~Bode, and H.-G.~Purwins, Phys. Rev. 
423: Lett. {\bf 78}, 3781 (1997).
424: \bibitem{gope}
425: D.M.~Petrich and R.E.~Goldstein, Phys. Rev. Lett {\bf E 72}, 1120 (1994); 
426: R.E.~Goldstein, D.J.~Muraki, and D.M.~Petrich, Phys. Rev. {\bf E 53}, 3933 
427: (1996). 
428: \bibitem{haim}
429: D.~Haim, G.~Li, Q.~Ouyang, W.D.~McCormick, H.L.~Swinney, A.~Hagberg, and 
430: E.~Meron, Phys. Rev. Lett. {\bf 77}, 190 (1996).
431: \bibitem{dima} L.M.~Pismen and D.~Kazhdan, unpublished.
432: \end{thebibliography}
433: 
434: \end{document}
435: 
436: 
437: