1: \documentclass[aps,prl,twocolumn,groupedaddress,showpacs,amsmath,

2: amssymb,amsfonts]{revtex4}

3: \usepackage{amsmath}

4: \usepackage{amsfonts}

5: \usepackage{amssymb}

6: \usepackage{bm}

7: \usepackage{graphicx}%

8: \def\eps{\varepsilon}

9: \def\goesto{\rightarrow}

10: \def\dirac{{\mbox{\boldmath$\delta$}}}

11: \begin{document}

12: \title{Spot deformation and replication in the two-dimensional

13: Belousov-Zhabotinski reaction in water-in-oil microemulsion}

14: \author{Theodore Kolokolnikov}

15: \author{Mustapha Tlidi}

16: \email[]{tkolokol@mathstat.dal.ca}

17: \email[]{mtlidi@ulb.ac.be}

18: \affiliation{Department of Mathematics and Statistics, Dalhousie

19: University, Halifax, Canada}

20: \affiliation{Optique Nonlin\'eaire Th\'eorique, Universit\'e Libre de Bruxelles, Campus Plaine CP 231, 

21: 1050 Bruxelles, Belgium}

22: 

23: \date{\today}

24: \begin{abstract}

25: In the limit of large diffusivity ratio, 

26: spot-like solutions in the two-dimensional Belousov-Zhabotinski 

27: reaction in water-in-oil microemulsion

28: are studied. It is

29: shown analytically that

30: such spots undergo an instability as the diffusivity ratio is

31: decreased. An instability threshold is derived. For spots of 

32: small radius, it is shown that this instability leads to a spot splitting

33: into precisely two spots. For larger spots, it leads to deformation,

34: fingering patterns and space-filling curves.

35: Numerical simulations are shown to be in close agreement with the

36: analytical predictions.

37: \end{abstract}

38: 

39: \pacs{45.70.Qj, 82.40.Ck, 02.30.Jr, 02.60.Lj}

40: 

41: \maketitle

42: Localized patterns such as spots belong to the class of dissipative

43: structures found far from equilibrium  \cite{DStruc}.  In recent years,

44: considerable progress has been made in the understanding of these

45: systems. 

46: The question of stability of such patterns is central and the source of

47: instabilities must be carefully scrutinized.  In particular, the

48: occurrence of instability can lead to the deformation of spots followed

49: by spot multiplication (also called self-replication) or fingering.  This

50: intriguing phenomenon has been the subject of research since the

51: pioneering work of Pearson  \cite{pearson}.  Shortly after, thanks to the

52: development of open spatial chemical reactors, self-replication was

53: observed in various experiments such as ferrocyanide-iodate-sulphite

54: reaction \cite{lmps}, the Belousov-Zhabotinsky reaction \cite{mpm, kve,

55: kve2},

56: and chloride dioxide-malonic-acid reaction \cite{DBDK}. By now this

57: phenomenon is believed to be universal \cite{mo, hayase1}. It is not

58: restricted to chemical reactions and occurs in many 

59: systems in biology \cite{meron}, material science \cite{RW, ns} and

60: nonlinear optics \cite{laser}. 

61: 

62: Analytically, spot replication is relatively well understood in

63: one dimensional setting. Nishiura and Ueyama \cite{N5} proposed that

64: self-replication of spikes occurs when the spike solution dissapears due

65: to the presence of a fold point.  Similar explanation has been reported

66: for the box-like patterns \cite{OK}.  In two dimensions, 

67: a mechanism for spot instability has been

68: proposed in \cite{mo} for general reaction-diffusion systems; related

69: analysis was performed earlier in the framework of  a piecewise-linear

70: approximation in \cite{omk}, \cite{km} and more recently in the context

71: of diblock copolymer systems \cite{RW}, \cite{ns}. 

72: See also \cite{Goldstein}, \cite{Mmotion} and \cite{gcom} and for a related 

73: approach from the point of view of interface motion.

74: 

75: In this letter we perform an analytical and numerical investigation of

76: the two-dimensional localized spots that were recently reported for the

77: Belousov-Zhabotinski (BZ) reaction in water-in-oil microemulsion

78: \cite{kve}, \cite{kve2}.

79: In the classical BZ reaction, spiral waves are observed \cite{KT} but no

80: localized spot solutions are possible.  Indeed, localized structures

81: develop only when the ratio of diffusion coefficients is sufficiently

82: large, which occurs in the microemulsion system but not in the

83: classical BZ reaction.

84: 

85: We consider the

86: water-in-oil microemulsion model of the BZ reaction as described in 

87: \cite{kve} and \cite{kve2}: 

88: \begin{subequations}

89: \label{bz}

90: \begin{align}

91: &\begin{aligned}

92: \varepsilon_{0}v_{t}  = 

93: \varepsilon_{0}D_{v}\Delta v+\left[  f_{0}

94: z+i_{0}\left(  1-mz\right)  \right]  \frac{v-q_{0}}{v+q_{0}}\\

95: +\left[

96: \frac{1-mz}{1-mz+\varepsilon_{1}}\right]  v-v^{2}

97: \end{aligned}\\

98: &z_{t}    =D_{z}\Delta z-z+v\left[  \frac{1-mz}{1-mz+\varepsilon_{1}}\right]

99: \end{align}

100: \end{subequations}

101: where $v, z$ are dimensionless concentrations of activator HBrO$_2$ and

102: oxidized catalyst $[Ru(bpy)_3]^{3+}$ respectively; $D_v$ and $D_z$ are

103: dimensionless diffusion coefficients of activator and catalyst; $f, \eps$

104: and $q$ are parameters of the standard Keener-Tyson model \cite{KT}; $i_0$

105: represents the photoinduced production of inhibitor, and $m$ represents

106: the strength of oxidized state of the catalyst with $0 <mz < 1$.

107: This reaction was shown experimentally and

108: numerically to admit localized spot patterns that persist for long time

109: \cite{kve}, \cite{kve2}. 

110: 

111: We rescale the variables as

112: $

113: z=1/m-m^{-3/2}w\varepsilon_{1},~~v=m^{-1/2}\hat{v},~~ t=\varepsilon

114: _{0}m^{1/2}\hat{t}.

115: $

116: In the new variables, after dropping the hats, we obtain

117: \begin{subequations}

118: \begin{align}

119: v_{t} =\varepsilon^{2}\Delta v+f(v,w);\ \ \tau w_{t}=D\Delta

120: w+g(v,w) \label{vw}

121: \end{align}

122: where

123: \begin{align}

124: f(v,z)   &=-\left[  f_{0}+f_{1}w\right]  \frac{v-q}{v+q}+\left[

125: \frac{w}{1+\alpha w}\right]  v-v^{2};\\

126: g(v,w)&=1-\left[  \frac

127: {w}{1+\alpha w}\right]  v

128: \end{align}

129: \end{subequations}

130: and with the nondimensional constants given by

131: \begin{align}

132: \begin{aligned}

133: &\alpha   =m^{-1/2},~~ f_{1}=\varepsilon_{1}m^{1/2}\left(  i_{0}%

134: -\frac{f_{0}}{m}\right),~~

135: q  =q_{0}m^{1/2},\\

136: &\varepsilon^2=\varepsilon_{0}D_{v}m^{1/2}

137: ,~~ D=D_{z}\varepsilon_{1}m^{-1/2},~~ \tau=\frac{1}{m}\frac{\varepsilon_{1}

138: }{\varepsilon_{0}}.

139: \end{aligned}

140: \end{align}

141: In the limit $m\rightarrow\infty,\ f_{1}\rightarrow0$ and $\tau\rightarrow

142: 0\ \ $we obtain the reduced system,%

143: \begin{equation}

144: \left\{

145: \begin{aligned}

146: v_{t}&=\varepsilon^{2}\Delta v-f_{0}\frac{v-q}{v+q}+wv-v^{2}\\

147: 0&=D\Delta w+1-vw.

148: \end{aligned}

149: \right.  \label{simple}%

150: \end{equation}

151: In particular, parameter values used in Fig.~14 of \cite{kve2} are 

152: $f_0=2.18, i_0=0,

153: m=10, \eps_1 = 0.01, \eps_0=0.1$ and $D_z/D_v=100$ which gives $\alpha=0.3,

154: f_1 = -0.007, \tau=0.01$, 

155: so that the simplification (\ref{simple}) is appropriate.

156: 

157: 

158: { \begin{figure}[b]

159: { \begin{minipage}[t]{0.15\textwidth}

160: \begin{center}{\small

161: \setlength{\unitlength}{1\textwidth} \begin{picture}(1,0.42)(0,0)

162: \put(-0.2,0.0){\includegraphics[width=1.2\textwidth,

163: height=1.1\textwidth]{steady3d.eps}}

164: \end{picture}

165: }(a)\end{center}

166: \end{minipage}

167: \begin{minipage}[t]{0.30\textwidth}

168: \begin{center}{

169: \setlength{\unitlength}{1\textwidth} \begin{picture}(1.0,0.42)(0,0)

170: \put(-0.05,0.03){\includegraphics[height=0.42\textwidth,

171: ]{ss1.eps}}

172: \put(+0.21,0.28){\includegraphics[height=0.12\textwidth,

173: ]{ss1b.eps}}

174: \put(0.5,0.03){\includegraphics[height=0.42\textwidth,

175: ]{ss2.eps}}

176: \end{picture}

177: }

178: (b) \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \  (c)

179: \end{center}

180: \end{minipage}

181: }\caption{(a) Radially symmetric, stable spot solution of the reduced BZ

182: system (\ref{simple})

183: in the BZ system on a unit disk.

184: (b) Radial profile of $v$ 

185: (solid curve) and its asymptotic approximation (dashed 

186: curves) given by (\ref{ss}). Insert: a blowup showing the profile of the

187: interface.

188: (c) Radial profile of $w$ (solid curve) and its two-term

189: asympotic approximations

190: (dashed curves) given by (\ref{w}).

191: Parameter values are $D = 20, f_0 = 2.0, \eps=0.02, q=0.005.$

192: }

193: \label{fig:steady}

194: \end{figure}} 

195: 

196: Experimental and numerical evidence in \cite{kve} and \cite{kve2} 

197: suggest that (\ref{bz}) admits localized spot soltuions,

198: such as shown in Fig.~\ref{fig:steady}.

199: Such spots occur in the regime where $\eps \ll 1.$

200: Our goal is to describe analytically the radius and profile of such a spot

201: and then study its

202: stability using singular perturbation techniques 

203: similar to those described in \cite{mo}.

204: As we will demonstrate, the instability thresholds appear in

205: the regime where  $D \gg 1$.  For sufficiently large values of $D$ a

206: spot pattern is stable. However as $D$ is decreased, instabilities of the

207: form $\exp(\lambda t) \cos(m\theta) \phi(r)$ may develop, where $\theta$

208: and $r$ are the angular and radial coordinates, respectively.

209: We provide the analytic description of the profile of the spot

210: (\ref{ss}, \ref{w}) and the dispersion relation (\ref{main-result}) 

211: between $m$ and $\lambda$. This relation leads directly to the estimate

212: for the instability threshold.  

213: 

214: We begin by constructing a stationary (time-independent)\ spot-type solution

215: on a two dimensional unit disk $\left\{  x:\left\vert x\right\vert <1\right\}

216: $. We assume that $D\gg1$. Then to leading order, $w\sim w_{0}$

217: is a constant to be determined, and the solution for $v$ consists of an

218: interface located at $\left\vert x\right\vert \sim l$ which connects two

219: nearly spatially homogeneous layers. To find the profile of such an interface

220: and its location $l$, let us rescale near $l$ as $v\left(  x\right)  =V\left(

221: y\right)  $ with $y=\left(  r-l\right)  /\varepsilon$ and $r=\left\vert

222: x\right\vert .$ For the steady state we then obtain, to leading order,

223: $V^{\prime\prime}\left(  y\right)  +f\left(  V;w_{0}\right)  =0.$ The

224: interface solution corresponds to a heteroclinic orbit of this ODE. The

225: existence of such an orbit is only possible whenever the equations

226: \begin{equation}

227: f\left(  V_{-};z_{0}\right)  =0=f\left(  V_{+};w_{0}\right)  ,\ \ \ \ \int

228: _{V_{-}}^{V_{+}}f\left(  s;w_{0}\right)  ds=0\label{Vpm}%

229: \end{equation}

230: are simultaneously satisfied for some values $V_{-}$ $<V_{+}.$ Since $q\ll1$,

231: we have $V_{-}\sim 0$ and (\ref{Vpm}) can be written

232: as

233: \begin{equation}

234: f_{0}+f_{1}w_{0}\sim\frac{3}{16}\left[  \frac{w_{0}}{1+\alpha w_{0}}\right]

235: ^{2};\ \ \ V_{+}\sim\frac{3}{4}\left[  \frac{w_{0}}{1+\alpha w_{0}}\right]

236: .\label{Vpw0}%

237: \end{equation}

238: Next, we ignore the $O(q)$ terms and integrate $V''(y)+f(V,w_0)=0$ 

239: to obtain

240: \begin{equation}

241: \begin{aligned}

242: v&\sim\left\{

243: \begin{aligned}

244: &V_{+}\tanh^{2}\left(  \sqrt{\frac{V_{+}}{6}}\left(  \dfrac{r-l}{\varepsilon

245: }\right)  \right)  ,~~& r<l\\

246: &0,& r>0

247: \end{aligned}

248: \right.  

249: \end{aligned}

250: \label{ss}%

251: \end{equation}

252: with $V_{+}$ and $w_{0}$ given by (\ref{Vpw0}) and where $r=\left\vert

253: x\right\vert $. This formula describes the profile of the interface in

254: the limit $q \to 0$. Its

255: thickness is of $O(\eps V_+^{-1/2})  .$ To determine its location

256: $l$, we integrate the second equation in (\ref{vw}). Zero-flux conditions then

257: yield $\int g=0$ so that

258: \begin{equation}

259: g\left(  V_{+},w_{0}\right)  l^{2}+\left(  1-l^{2}\right)  g\left(

260: 0,w_{0}\right)  \sim0\label{gio}%

261: \end{equation}

262: In the limit of the reduced system (\ref{simple}),\ we obtain%

263: \begin{align}

264: \label{ss2}

265: l\sim\frac{1}{2\sqrt{f_{0}}},\ \ w_{0}\sim\frac{4}{\sqrt{3}}\sqrt{f_{0}%

266: },\ \ \ V_{+}\sim\sqrt{3f_{0}}.

267: \end{align}

268: To determine the correction to $w,$ we write $w=w_{0}+D^{-1}%

269: w_{1}+O\left(  D^{-2}\right)  $. to obtain $\Delta w_{1}+g\left(  w_{0}%

270: ,v_{0}\right)  =0.$ Imposing continuity at the interface $r=l$, the

271: solvability condition $w_{1}(l)=0$, and using (\ref{gio})\ and $g\left(

272: 0,w_{0}\right)  =1,$ we then obtain

273: \begin{align}

274: w\sim w_{0}+\frac{1}{4D}\left\{

275: \begin{aligned}

276: &-\dfrac{(1-l^2)(l^2-r^2)}{l^2}, &r<l\\

277: &

278: 2\ln(\dfrac{r}{l})+l^2-r^2, &r>l

279: \end{aligned}

280: \right. .

281: \label{w}

282: \end{align}

283: An example of a localized spot and its radial profile are shown in Fig 

284: \ref{fig:steady}.

285: 

286: When decreasing the diffusion coefficient $D$ of the recovery variable,

287: numerical simulations show that the spot becomes unstable. To compute the

288: threshold associated with this instability, we linearlize around the

289: localized spot solution (\ref{ss}) as

290: \begin{align*}

291: v(x,t)  &  =v(x)+\exp\left(  \lambda t\right)  \cos\left(  m\theta\right)

292: \phi\left(  r\right) \\

293: w(x,t)  &  =w(x)+\exp\left(  \lambda t\right)  \cos\left(  m\theta\right)

294: \psi\left(  r\right)

295: \end{align*}

296: where $v$ and $w$ are given to leading order in (\ref{ss} and \ref{ss2}) spot-type solution, $m$ is an integer, $\phi,\psi\ll1$

297: and $(r,\theta)$ are the polar coordinates of $x.$ Substituting into

298: (\ref{bz}) we then obtain

299: {

300: \begin{subequations}

301: \begin{align}\label{phi}

302: \lambda\phi &  =\varepsilon^{2}\left(  \phi_{rr}+\frac{1}{r}\phi_{r}%

303: -\frac{m^{2}}{r^{2}}\phi\right)  +f_{v}  \phi+f_{w}  \psi\\ 

304: \label{psi}

305: \tau\lambda\psi &  =D\left(  \psi_{rr}+\frac{1}{r}\psi_{r}-\frac{m^{2}}{r^{2}%

306: }\psi\right)  +g_{v}  \phi+g_{w}  \psi.

307: \end{align}

308: \label{phipsi}

309: \end{subequations} }

310: Note that $v_r$ satsifies

311: that $\varepsilon^{2}(v_{rrr}+1/rv_{rr}-1/r^2v_r)+f_{v}v_{r}+w_{r}f_{w}=0.$ 

312: Therefore multiplying (\ref{psi}) by $v_r$ and integrating by parts, we obtain

313: \begin{align}\label{2:56}

314: \left(  \lambda+\frac{(m^{2}-1)\varepsilon^{2}}{l^{2}}\right)  \int v_{r}^{2}%

315: \sim\int v_{r}f_{w}\left(  \psi-w_{r}\right)  .

316: \end{align}

317: where we have assumed that to leading order, 

318: $\phi\sim v_{r}, \lambda \ll 1, \psi \ll \phi$.

319: Since $v_{r}$ is exponentially small outside the interface, we simplify $\int

320: v_{r}f_{w}\left(  \psi-w_{r}\right)  \sim-\left(  \psi\left(  l\right)

321: -w_{r}\left(  l\right)  \right)  \int_{0}^{V_{+\ }}f_{w}\left(  v,w_{0}%

322: \right)  dv$. We estimate

323: $w_r(l)  \sim -1/(2D)g\left(  V_{+},w_{0}\right)$ and

324: to determine $\psi\left(  l\right),$ we integrate (\ref{psi})

325: over the interface $l$ to obtain

326: $

327: D\psi_{r}|_{l^{-}}^{l^{+}}

328: \sim-g\left(  0,w_{0}\right)  /l^{2}

329: $

330: where we have used $\phi \sim v_r$ and (\ref{gio}). Keeping only

331: leading order terms in $D$ we then obtain the following problem for

332: $\psi$:

333: \begin{subequations}

334: \label{reducedpsi}

335: \begin{align}\label{djump}

336: &0=

337: \psi_{rr}+\frac{1}{r}\psi_{r}-\frac{m^{2}}{r^{2}}\psi=0,

338: ~~~~ r\ne l

339: \\

340: &  \psi^{\prime}(  0)=0=\psi^{\prime}(  1)

341: ,~~

342:   \psi^{\prime}\left(  l^{+}\right)  -\psi^{\prime}\left(  l^{-}\right)

343: =-\frac{1}{Dl^{2}}g\left(  0,w_{0}\right)  . 

344: \end{align}

345: \end{subequations}

346: Using the continuity of $\psi$ at $r=l$ we get

347: \[

348: \psi\left(  l\right)  \sim\frac{1}{2Dml}g\left(  0,w_{0}\right)  \left(  1+l^{2m}\right)

349: \]

350: and substituting into (\ref{2:56}) we obtain

351: 

352: \begin{align}

353: \lambda\frac{l^{2}}{\varepsilon^{2}}\sim 1-m^{2}+A\left[  1-l^{2}-\frac{1}%

354: {m}\left(  1+l^{2m}\right)  \right]

355: \label{main-result}

356: \end{align}

357: where

358: \begin{align}

359: A=\frac

360: {l}{2\varepsilon D}\frac{g\left(  0,w_{0}\right)  \left(  \int_{0}^{V_{+\ }%

361: }f_{w}\left(  v,w_{0}\right)  dv\right)  }{\varepsilon\int v_{r}^{2}}.

362: \end{align}

363: For the reduced model (\ref{simple}) we obtain an explicit result

364: \begin{align}

365: \label{simple-A}

366: A \sim \frac{3^{5/4}2^{1/2}5}{64}\frac{1}{f_0^{3/4}\eps D}

367: =

368: 0.43622\frac{1}{f_{0}^{3/4}\varepsilon D}.

369: \end{align}

370: In Fig.~\ref{fig:evals}, the analytical prediction given by

371: (\ref{main-result}) is found to be in good agreement with the numerical

372: computations of problem (\ref{phipsi}).  Full numerical simulations of

373: (\ref{simple}) also agree with this prediction.  For example, when taking

374: $\eps=0.03, f_0=1.3$ and $q=0.01$, slight spot deformation corresponding

375: to mode $m=2$ is observed when $D = 1.0$ 

376: but not when $D=1.3.$ This agrees well with $D\sim1.1$, 

377: the threshold predicted by (\ref{main-result}).

378: \begin{figure}[tb]

379: \setlength{\unitlength}{0.5\textwidth} \begin{picture}(1,0.30)(0,0)

380: \put(0.0,0.17){\footnotesize $\dfrac{\lambda}{\eps^2}$}

381: \put(0.04,0.09){\includegraphics[width=0.12\textwidth,

382: height=0.08\textheight]{eval1.eps}}

383: \put(0.18,0.056){\footnotesize $m$}

384: \put(0.18,0.00){(a)}

385: \put(0.38,0.09){\includegraphics[width=0.12\textwidth,

386: height=0.08\textheight]{eval2.eps}}

387: \put(0.33,0.17){\footnotesize $\dfrac{\lambda}{\eps^2}$}

388: \put(0.51,0.056){\footnotesize $D$}

389: \put(0.51,0.00){(b)}

390: \put(0.71,0.09){\includegraphics[width=0.12\textwidth,

391: height=0.08\textheight]{Dl.eps}}

392: \put(0.66,0.17){\footnotesize $D$}

393: \put(0.79,0.063){\footnotesize $l=\frac{1}{4f_0^2}$}

394: \put(0.84,0.00){(c)}

395: \end{picture}

396: \caption{(a), (b) Comparison of numerical computations of $\lambda$ given by

397: (\ref{phipsi}) (diamonds) with

398: the analytical result (\ref{main-result}, \ref{simple-A}) (dashed line)

399: for the reduced model

400: (\ref{simple}) on a unit disk. Parameter values are $q=0.01$, $\eps=0.01$ and 

401: (a) $D=3, f_0=1.3$; (b) $m=2, f_0=1.3$.

402: (c) Simultaneous solution of $\lambda=0=d\lambda/dm$, showing the first

403: value of $D$ for which instability occurs. The system is stable above the

404: curve and unstable below it.

405: To compute $\lambda$ numerically, 

406: (\ref{phipsi}) was reformulated as a boundary value problem by adjoining

407: the equation $d\lambda/dr=0$ along with fixing $\psi(1).$

408: Maple's numerical 

409: boundary value problem solver was then used with initial guesses 

410: $\phi=v_r, \lambda=0$ and $\psi =$ the solution of

411: (\ref{reducedpsi}).

412: All computations are

413: correct to four significant digits. }

414: \label{fig:evals}

415: \end{figure}

416: 

417: From (\ref{main-result}) and (\ref{simple-A}), it is clear that 

418: the mode $m=1$ is always stable and that

419: for large enough $D$, all modes $m\geq1$ are stable. As $D$ is decreased,

420: instability sets in when $D=O(\eps^{-1})$.

421: The threshold

422: value is found by simultaneously solving $\lambda=d\lambda/dm=0$. 

423: The resulting graph is shown on Fig.~\ref{fig:evals}c.

424: In particular, note that the system is stable if $D\eps > 0.038$, independent

425: of the value of $f_0.$

426: In the limit

427: of small radius $l\rightarrow0$, the first unstable integer mode is $m=2$

428: corresponding to $A=6,$ so that

429: the spot of small radius becomes unstable whenever 

430: $\eps D f^{3/4} \le 0.0727.$

431: More generally, by eliminating $A$ 

432: we find

433: that there exist constants $l_{1}<l_{2}<\ldots<1$ such the first unstable mode

434: is $m$ provided that $l_{m-1}<l<l_{m},$ where $l_1=0$, $l_2=0.491$, $l_3=0.667$,

435: $l_4=0.753$ and

436: $l_{m}\sim 1-0.937/m$ as $m\rightarrow\infty;$ where 0.937 is the root of

437: $e^{-2z}\left(  3+2z\right)  +3-4z=0.$

438: 

439: 

440: \begin{figure*}

441: \begin{center}

442: \includegraphics[width=0.45\textwidth]{bif.eps}

443: \includegraphics[width=0.45\textwidth]{bif-ex.eps}\\

444: \hfill (a)\hfill$~$\hfill  (b) \hfill$~$\\

445: \end{center}

446: \caption{ (a) Bifurcation diagram for (\ref{simple}) in $D$ and $f_0$ with 

447: $\eps=0.05, q=0.01.$ Solid dots represent deformations of a spot without

448: topological change. Points marked by "2" represent spot-replication into

449: two spots. An empty circle represents spot-to-ring bifurcation 

450: and empty circle

451: with a number inside represents spot-to-ring-to-spots bifurcation.

452: A solid line represents the boundary of spot-to-ring replication which

453: occurs at the fold point of the radially symmetric steady state. 

454: The bifurcation diagram was obtained by solving the full two-dimensional

455: system (\ref{simple}) 

456: using the finite element package {\tt FlexPDE} \cite{FlexPDE} with zero-flux

457: boundary conditions and 800 elements on a quarter-disk. For initial

458: conditions, (\ref{ss}, \ref{ss2}, \ref{w}) was used, but with $r$

459: replaced by $r(1+0.05 \cos2\theta)$.

460: The solid

461: line was obtained by solving for the fold point of the radially symmetric

462: steady state using Maple's boundary value problem solver. (b) Full

463: numerical simulations starting from a spot-state (\ref{ss}, \ref{w}) as

464: initial conditions. 

465: First row: spot deformation, $f_0=1.3 , D=0.35$.

466: Note that the final

467: stedy state of the system is a deformed blob shown here at $t=4000$.

468: Second row: self-replication, $f_0=2.2, D=0.24$. 

469: Third row: spot-to-ring

470: bifurcation, $f_0=1.3, D=0.1$. Fourth row: spot-to-ring-to-spots

471: bifurcation, $f_0=2.6, D=0.1$.}

472: \label{fig:bif}

473: \end{figure*}

474: 

475: 

476: Numerical computations indicate that self-replication is more prevalent

477: for spots of small radius (see Fig.~\ref{fig:bif}).

478: For larger spots, a deformation usually leads to 

479: to the so-called "finger growth'' and space filling curves. 

480: Others  studies have shown the occurrence of

481: fingering instabilities leading to labyrinthine patterns \cite{Fingers},

482: \cite{Goldstein}, \cite{mo}. The

483: question of whether the self-replication or fingering instability 

484: occurs first is still open. 

485: 

486: 

487: For smaller, $O(1)$ values of $D$, there is also a different instability

488: mechanism that can lead to splitting of a spot into a ring as illustrated

489: in Fig.~\ref{fig:bif}.  Unlike spot multiplication, this instability

490: is radially symmetric, and is caused by the {\em dissapearence} of the

491: steady state solution -- whereby the steady state solution ceases to

492: exist due to the presence of a saddle-node bifurcation -- 

493: rather than by its lateral instability. Numerical simulations suggest

494: that that spot replication occurs only for spots of smaller radius,

495: whereas spot-to-ring instability is dominant for larger spots.

496: 

497: To conclude, 

498: we have estimated analytically for which

499: value of diffusion $D$ spot deformation first occurs. 

500: As is $D$ is decreased further, self-replication and/or spot-to-ring

501: instability is observed. As evidenced by numerical simulations, we

502: conjecture that spot deformation is the precursor to this phenomenon.

503: 

504: 

505: \begin{thebibliography}{}

506: The authors are grateful to Vladimir Vanag and Irving Epstein 

507: for introducing us to this subject and for illuminating discussions. 

508: We are also grateful to the anonymous referees whose

509: suggestions have improved the paper significantly.

510: T.K.~is supported by NSERC Canada, and the 

511: Dalhousie University startup grant.  M.T.~is supported by the

512: FNRS Belgium.

513: 

514: 

515: 

516: \bibitem{DStruc}P.~Glansdorff and I.~Prigogine, {\it {Thermodynamic Theory of Structures, Stability and Fluctuations}} (Wiley, New York, 1971); E. Meron, Phys. Rep. {\bf{218}}, 1 (1992);

517: J.~Chanu and R.~Lefever, {\it Inhomogeneous phases and pattern

518: formation}, Physica A {\bf 213}, 1-276 (1995);

519: M.C.~Cross and P.C.~Hohenberg, Mod. Phys. Rep. {\bf{65}}, 851 (1993); 

520: K.~Kapral and K.~Showalter, {\it{Chemical Waves and Patterns}} 

521: (Kluwer Academic press, Dordercht, 1995).

522: 

523: \bibitem{pearson}J.E.~Pearson, 

524: Science, {\bf {261}}, 189 (1993).

525: 

526: \bibitem{lmps}K.~Lee, W.D.~McCormick, J.E.~Pearson, and H.L.~Swinney,

527: Nature, {\bf{396}}, 215 (1994).

528: 

529: \bibitem {mpm}A.P. Mu\~nuzuri, V. P\'erez-Villar, and M.~Markus,

530: Phys. Rev. Lett. {\bf{79}}, 1940 (1997).

531: 

532: \bibitem{kve}A.~Kaminaga, V.K.~Vanag, and I.R.~Epstein,

533: Angewandte Chemie, {\bf{45}}, 3087 (2006).

534: 

535: \bibitem{kve2}A.~Kaminaga, V.K.~Vanag, and I.R.~Epstein,

536: J. Chem. Phys. {\bf 122}, 174706 (2005).

537: 

538: \bibitem{DBDK}P.W.~Davis, P.~Blanchedeau, E.~Dulos, and P.~De~Kepper,

539: J.~Phys.~Chem.~A {\bf{102}} 8236 (1998).

540: 

541: \bibitem{mo} C.~Muratov and V.V.~Osipov, Phys. Rev. E {\bf 53}, 3101

542: (1996); Phys. Rev. E {\bf 54}, 4860 (1996);

543: C.~Muratov, Phys. Rev. E {\bf 66}, 066108 (2002).

544: 

545: 

546: \bibitem{hayase1}Y. Hayase and T. Ohta, Phys. Rev. Lett. {\bf {81}}, 1726 (1998); 

547: Y. Hayase, Phys. Rev. E {\bf {62}}, 5998 (2000).

548: 

549: \bibitem{meron} 

550: E.~Meron, E.~Gilad J.~von~Hardenberg, and M.~Shachak,

551: Chaos, Solitons and Fractals, {\bf{19}}, 367 (2004).

552: 

553: 

554: \bibitem{RW}X.\ Ren and J.\ Wei,  SIAM J. Math. Anal. {\bf {35}}, 1 (2003).

555: 

556: \bibitem{ns} Y.Nishiura and H.Suzuki, SIAM Math.Anal. {\bf 36}(3), 916

557: (2004).

558: 

559: \bibitem{laser} M. Tlidi A.G. Vladimirov, and P. Mandel, Phys. Rev.

560: Lett.

561: {\bf {89}}, 233901 (2002).

562: 

563: 

564: \bibitem{N5}Y. Nishiura and D. Ueyama, Physica D {\bf {130}}, 73 (1999).

565: 

566: \bibitem{OK}B.S. Kerner and V.V. Osipov, {\it{Autosolitons: A new Approach to Problems of Self-Organization and Turbulence}} (Kluwer, Dordrecht, 1994).

567: 

568: 

569: \bibitem{omk}T.~Ohta, M.~Mimura, and R.~Kobayashi, Physica D {\bf {34}}, 115 (1989).

570: 

571: \bibitem{km} S.Kawaguchi and M.Mimura, SIAM J.Appl,Math {\bf 59}, 920(1999)

572: 

573: \bibitem{KT}J.J.~Tyson and P. Fife, J. Chem. Phys. {\bf {73}}, 2224 (1980); 

574: J.P.~Keener and J.J.~Tyson, Physica D {\bf {21}}, 307 (1986).

575: 

576: \bibitem{Goldstein}

577: R.E.~Golstein, D.J. Muraki, and D.M. Petrich, Phys. Rev. E {\bf {53}}, 3939 (1996). 

578: 

579: \bibitem{Mmotion}

580: C.B.~Muratov, Phys. Rev. E {\bf 54}, 3369 (1996) 

581: 

582: 

583: \bibitem{gcom}

584: D.~Gomila, P.~Colet, G.L.~Oppo and M.~San Miguel,

585: Phys.~Rev.~Lett.~{\bf 87}, 194101 (2001).

586: 

587: 

588: \bibitem{Fingers}

589: K. Lee and H. Swinney, Phys. Rev. E {\bf {51}}, 1899 (1995).

590: 

591: \bibitem{FlexPDE} See FlexPDE website, {\tt www.pdesolutions.com}.

592: 

593: \end{thebibliography}

594: \end{document}

595: