1: \documentclass[aps,prl,twocolumn]{revtex4}

2: %\documentclass[letterpaper,fullpage,aps,prl,twocolumn]{revtex4}

3: \usepackage{graphicx,amsmath}

4: %\topmargin +0.0in

5: 

6: \newcommand{\Pe}{\mathrm{P\hspace{-1pt}e}}

7: \newcommand{\Peo}{\mathrm{P\hspace{-1pt}e}\hspace{-0.5pt}_o}

8: \newcommand{\re}{\mbox{Re\hspace{1pt}}}

9: 

10: 

11: \begin{document}

12: \title{ Interfacial dynamics in transport-limited dissolution }

13: 

14: \author{Martin Z. Bazant}

15: \affiliation{

16: Department of Mathematics,

17:   Massachusetts Institute of Technology,

18:   Cambridge, MA 02139}

19: 

20: \date{\today}

21: 

22: \begin{abstract}

23:   Various model problems of ``transport-limited dissolution'' in two

24:   dimensions are analyzed using time-dependent conformal maps.  For

25:   diffusion-limited dissolution (reverse Laplacian growth), several

26:   exact solutions are discussed for the smoothing of corrugated

27:   surfaces, including the continuous analogs of ``internal

28:   diffusion-limited aggregation'' and ``diffusion-limited erosion''. A

29:   class of non-Laplacian, transport-limited dissolution processes are

30:   also considered, which raise the general question of when and where

31:   a finite solid will disappear. In a case of dissolution by

32:   advection-diffusion, a tilted ellipse maintains its shape during

33:   collapse, as its center of mass drifts obliquely away from the

34:   background fluid flow, but other initial shapes have more

35:   complicated dynamics.

36: \end{abstract}

37: 

38: \maketitle

39: 

40: The analysis of interfacial dynamics using conformal maps (``Loewner

41: chains'') is a classical subject, which is finding unexpected 

42: applications in physics~\cite{gruzberg04,bauer04,handbook05}. For

43: dynamics controlled by Laplacian fields, there is a vast literature on

44: continuous models of viscous fingering~\cite{bensimon86,howison92},

45: and stochastic models of diffusion-limited aggregation

46: (DLA)~\cite{hastings98,handbook05} and fractal curves in critical

47: phenomena~\cite{gruzberg04,bauer04}.  Conformal-map dynamics has also

48: been formulated for a class of non-Laplacian growth phenomena of both

49: types~\cite{bazant03}, driven by conformally invariant transport

50: processes~\cite{bazant04}. For growth limited by advection-diffusion

51: in a potential flow, the connection between continuous and stochastic

52: growth patterns has been elucidated~\cite{david05}, and the continuous

53: dynamics has also been studied in cases of freezing in flowing

54: liquids~\cite{maksimov76,kornev88,kornev94,alimov98}.

55: 

56: In all of these examples, the moving interface separates a ``solid''

57: region, where singularities in the map reside, a ``fluid'' region,

58: where the driving transport processes occur and the map is

59: univalent. (In viscous fingering, these are the inviscid and viscous

60: fluid regions, respectively.) Most attention has been paid to problems

61: of ``transport-limited growth'', where the solid region grows into the

62: fluid region, since the dynamics is unstable and typically leads to

63: cusp singularities in finite time~\cite{meyer82,shraiman84,howison86}

64: (without surface tension~\cite{bensimon86}). Here, we consider various

65: time-reversed problems of ``transport-limited dissolution'' (TLD),

66: where the solid recedes from the fluid region, e.g. driven by

67: advection-diffusion in a potential flow. These are stable processes,

68: so we focus on continuous dynamics, without surface tension.

69: 

70: Stochastic diffusion-limited dissolution (DLD), sometimes called

71: ``diffusion-limited erosion'' (DLE) or ``anti-DLA'', has been

72: simulated by allowing random walkers in the fluid to annihilate

73: particles of the solid upon contact

74: ~\cite{paternson84,tang85,meakin86,krug91}, rather than aggregating as

75: in DLA~\cite{witten81}.  Outward radial DLE on a lattice, or

76: ``internal DLA'' (IDLA), where the random walkers start at the origin

77: and cause a fluid cavity to grow in an infinite solid, has also been

78: studied by mathematicians, who proved that the asymptotic shape is a

79: sphere in any dimension~\cite{lawlor92}. 

80: 

81: We begin our analysis by summarizing some exact solutions for

82: continuous DLD, which could help to understand fluctuations and the

83: long-time limits of DLE and IDLA. Although these solutions are known

84: in somewhat different forms for Laplacian

85: growth~\cite{meyer82,shraiman84,howison86}, it is instructive to

86: summarize them prior to considering problems of TLD by

87: advection-diffusion, to highlight the effects of fluid flow.

88: 

89: {\it Dissolution of surface corrugation. --} Let $G(w,t)$ be a

90: conformal map from the left half plane to the fluid region, where

91: steady (Laplacian) diffusion with uniform flux at $-\infty$ drives

92: dissolution of the solid. The interfacial dynamics is given by the

93: (dimensionless) Polubarinova-Galin (PG)

94: equation~\cite{handbook05,howison92},

95: \begin{equation}

96: \re\left(\overline{G^\prime}{G_t}\right) = 1, \ \ \mbox{ on } \re w = 0.

97: \end{equation}

98: An exact solution has the form,

99: \begin{equation}

100:   G(w,t) = w + h(t) + \log(1+a(t)e^w)  \label{eq:fingers}

101: \end{equation}

102: with $|a(t)|<1/2$ (real) where

103: \begin{eqnarray}

104: a &=& \frac{Ce^{-t}}{(1-a^2)^{3/2}} \sim Ce^{-t} + \frac{3}{2}C^3e^{-3t} + \ldots\\

105: \frac{dh}{dt} &=& \frac{1+a-a^2}{1-a^2}, \ h \sim t + C(e^{-t}-1) +

106: \ldots  \label{eq:ah}

107: \end{eqnarray}

108: and $C = a(0)(1-a(0)^2)^{3/2}$. This solution, shown in

109: Fig.~\ref{fig:fingers}, describes the decay of surface corrugations in

110: electropolishing~\cite{meakin86,krug91,wagner54,edwards53} or the

111: displacement of an inviscid fluid by an immiscible viscous fluid in a

112: Hele-Shaw cell~\cite{paterson84,tang85,bensimon86}. 

113: % Upon time-reversal, the interface forms $2/3$-power cusps as $a\to 1/2$

114: %~\cite{bensimon86,meyer82,shraiman84,howison86}.)

115: and closely ressembles simulations of ``mean-field DLE'' (Fig. 2 of

116: ~\cite{tang85}).

117: 

118: The linear stability of a flat interface in DLD is contained in the

119: long-time limit of Eqs.~(\ref{eq:fingers})-(\ref{eq:ah}). With the

120: dimensions restored ($w \mapsto wk$, $G \mapsto Gk$, and $t \mapsto

121: kvt$), the interface shape becomes sinusoidal,

122: \begin{equation}

123: x(y) \sim vt + C e^{-kvt} \cos(ky),

124: \end{equation}

125: for a Fourier mode of wavenumber $k$ and mean interfacial velocity

126: $v$. We thus recover the classical

127: result~\cite{tang85,krug91,wagner54} that $kv$ is the exponential

128: decay rate of mode $k$, as observed in experiments~\cite{edwards53}.

129: 

130: \begin{figure}

131: \includegraphics[width=.9\linewidth]{fingers}

132: \caption{ \label{fig:fingers} Diffusion-limited dissolution of a

133:   corrugated surface from left to right, for $h(0)=0$, $a(0)=.3$,

134:   $t=0,.25,.50,\ldots,2.50$ in Eq.~(\ref{eq:fingers}). }

135: \end{figure}

136: 

137: 

138: {\it Outward dissolution of clover-like shapes. -- } Next we consider

139: the continuous analog of IDLA: DLD in a radial geometry driven by a

140: constant diffusive flux from the origin. This could model quasi-steady

141: melting of an infinite solid around a point source of heat, or the

142: injection of a viscous fluid into a Hele-Shaw cell, displacing an

143: inviscid fluid. The radial PG equation is~\cite{handbook05,howison92},

144: \begin{equation}

145: \re\left(\overline{w g^\prime}

146: g_t \right) = 1, \ \ \mbox{ on }

147: |w|=1, \label{eq:pgr} 

148: \end{equation}

149: where $ z=g(w,t)$ is a univalent mapping of the {\it interior} of

150: the unit disk.

151: % Substituting the Taylor expansion,

152: % \begin{equation}

153: % g(w,t) = \sum_{n=0}^\infty a_n(t) w^n \ \ \mbox{for } |w| \leq 1.

154: %\label{eq:taylor}

155: %\end{equation}

156: %yields a system of nonlinear ordinary differential equations for the

157: %coefficients. 

158: A tractable case is the clover-like $(N-1)$-fold perturbation of a

159: circle,

160: \begin{equation}

161: g(w,t) = a_1(t) w + a_N(t) w^N   \label{eq:gout}

162: \end{equation}

163: first analyzed by Meyer for the (time-reversed) Hele-Shaw

164: problem~\cite{meyer82}.  For $a_1(0)=1$ and $0 < c = a_N(0) < 1/N$

165: (real), the solution has the implicit form,

166: \begin{equation}

167: a_1^2 + N a_N^2 = 1 + Nc^2+2t \ \mbox{ and } \  a_1^Na_N=c,   \label{eq:aNout}

168: \end{equation}

169: Since $a_1(t)$ increases to $\infty$ from $a_1(0)=1$ and $a_N(t)$

170: decreases to $0$ from $a_N(0)=c<1/N < 1$, the following recursion

171: converges very quickly,

172: \begin{equation}

173: a_1 = \sqrt{2t + 1 + Nc^2(1-a_1^{-2N})}, \label{eq:rec}

174: \end{equation}

175: and yields asymptotic approximation upon recursive substitution.

176: %

177: % \begin{equation}

178: % a_1 \sim \sqrt{2t + 1+Nc^2(1-(2t+1)^{-N})}   \label{eq:a1outsim}

179: % \end{equation}

180: An example shown in Figure \ref{fig:out4}, illustrates the rapid

181: smoothing of a four-leafed clover shape.

182: 

183: From Eqs.~(\ref{eq:aNout})-(\ref{eq:rec}) we see that the $(N-1)$-fold

184: radial perturbation decays as a power-law, $a_N \propto t^{-N/2}$, in

185: contrast to the exponential decay of perturbations of a flat

186: interface. We conjecture that the stochastic interface in IDLA is

187: asymptotic to the continuous DLD solution above, up to small

188: logarithmic fluctuations~\cite{lawlor95}, where $w^N$ is the smallest

189: perturbation in the initial cluster shape. In general, the continuous

190: dynamics of DLD may also accurately approximate the ensemble-averaged

191: stochastic dynamics of IDLA, which is not the case for DLA and other

192: unstable aggregation processes~\cite{david05}.

193: 

194: \begin{figure}

195: \includegraphics[width=1.2\linewidth]{out4}

196: \caption{ \label{fig:out4} Outward radial diffusion-limited dissolution

197:   driven by a point sink at the origin for an initial four-lobed

198:   perturbation of a circle. This exact solution is given by

199:   Eqs.~(\ref{eq:out})-(\ref{eq:outeqns}) with $N=5$, $c=0.15$, $t=0,

200:   0.2, 0.4,\ldots,1.0$.  }

201: \end{figure}

202: 

203: An (apparently new) explicit solution is possible for $N=2$. In that

204: case, Equation (\ref{eq:aNout}) reduces to a depressed

205: cubic~\cite{nick93}, $a_2^3 + 3p a_2 = 2 q$, solved by the

206: formula of Cardano and dal Ferro,

207: \begin{equation}

208: a_1 = \sqrt[3]{q + \sqrt{q^2-p^3}} + \sqrt[3]{q-\sqrt{q^2-p^3}}.

209: \end{equation}

210: where $p=-(t+c^2+1/2)/4$ and $q=-c/4$. 

211: 

212: {\it Inward dissolution of star-like shapes. -- } Next we briefly

213: consider inward DLD with a constant diffusive flux at infinity, which

214: simply is the time-reverse of Laplacian growth in a radial

215: geometry~\cite{howison92}. Now the map in Eq.~(\ref{eq:pgr}) must be

216: univalent {\it outside} the unit disk, with a Laurent expansion,

217: \begin{equation}

218: g(w,t) = \sum_{n=-\infty}^1 a_n(t) w^n, \ \ |w|\geq 1   \label{eq:ls}

219: \end{equation}

220: There are well-known solutions for $(N+1)$-fold

221: perturbations~\cite{shraiman84,howison86},

222: \begin{equation}

223: g(w,t) = a_1(t) w + a_{-N}(t) w^{-N} \  \ \mbox{ for } \ |w|\geq 1, \label{eq:in}

224: \end{equation}

225: where, without loss of generality, $a_1(0)=1$ and $a_{-N}(0) = c<1/N$

226: is real. Again, the Laurent coeficients satisfy a pair of nonlinear

227: equations, which is most easily solved as a fixed-point iteration,

228: \begin{equation}

229: a_1 = \sqrt{1-2t+Nc^2(a_1^{2N}-1)} \ \mbox{ and } \ a_{-N} = c a_1^N. \label{eq:ineqs}

230: \end{equation}

231: The only qualitative difference with outward DLD is that the solid

232: collapses to a point in a finite time, $t_c = (1-Nc^2)/2$. For $N=1$,

233: an ellipse ($0<c<1$) or circle ($c=0$) maintains its shape during

234: collapse,

235: \begin{equation}

236: g(w,t) = \sqrt{1 - \frac{t}{t_c}} \left(w + \frac{c}{w}\right)

237: \end{equation}

238: For $N\geq 1$, the shape approaches a circle prior to collapse, 

239: according to the asymptotic formula,

240: \begin{equation}

241: a_1(t) \sim \sqrt{2(t_c-t) + Nc^2(2(t_c-t))^N} 

242: \end{equation}

243: %a_{-N}(t) &\sim& c\left(2(t_c-t) + Nc^2(2(t_c-t))^N\right)^{N/2} 

244: %\end{eqnarray}

245: with $a_{-N}(t)=c a_1(t)^N$.  The collapse of a four-pointed 

246: shape ($N=3$) is shown in Figure~\ref{fig:in4}.

247: 

248: \begin{figure}

249: \includegraphics[width=1.2\linewidth]{in4}

250: \caption{ \label{fig:in4} Inward radial diffusion-limited dissolution

251:   driven by a sink at $\infty$ for a four-pointed shape.  This exact

252:   solution is given by Eqs.~(\ref{eq:in})-(\ref{eq:ineqs}) with $N=3$,

253:   $c=0.3$, $t=0, 0.04, 0.08,\ldots,0.36, 0.3649$; the collapse occurs

254:   at $t_c = 0.365$.  }

255: \end{figure}

256: 

257: 

258: {\it Advection-diffusion-limited dissolution. --} The dynamics of

259: dissolution become more interesting when driven by non-Laplacian (but

260: conformally invariant~\cite{bazant04}) transport

261: processes~\cite{bazant03}, where right-hand side of the PG equation

262: (\ref{eq:pgr}) is replaced by the nonuniform, time-dependent flux to

263: the interface, $\sigma(w,t)$.  An important example is dissolution by

264: advection-diffusion in a potential flow, e.g. the erosion of rock by

265: flowing water, the evaporation of a fiber coating in a flowing gas, or

266: the melting of a solid column in a flowing liquid. The time-reversed

267: growth problem has been studied extensively in the contexts of

268: freezing~\cite{maksimov76,kornev88,kornev94,alimov98} and

269: advection-diffusion-limited aggregation~\cite{bazant03,david05}, but

270: it seems that dissolution -- which leads to collapse in finite time --

271: has not been analyzed. For a given initial shape and background flow,

272: {\it when and where will collapse occur?}

273: 

274: Consider a finite solid of constant concentration and arbitrary shape

275: in a two dimensional potential flow of zero concentration and uniform

276: velocity far away. The relative importance of advection to diffusion

277: is measured by the P\'eclet number, $Pe_0 = UL/D$, for a background

278: fluid velocity $U$, diffusivity $D$ and length $L$. The time-dependent

279: P\'eclet number, $Pe(t) = Pe_0 a_1(t)$, is defined by the conformal

280: radius, $a_1(t)$, in Eq.~(\ref{eq:ls}). When solid dissolves

281: ($a_1(t)\to 0$), diffusion eventually dominates, and it is natural to

282: focus on the low-$Pe$ limit.

283: 

284: The flux profile $\sigma(\theta,\Pe)$ on the absorber has been studied

285: extensively, and very accurate asymptotic approximations are

286: available~\cite{choi05}. (A numerical code in matlab is also at

287: http://advection-diffusion.net.) From the low-$\Pe$ approximation,

288: \begin{equation}

289: \sigma \sim \frac{I_0(Pe)}{K_0(Pe/2)}e^{Pe\,\cos\theta} -

290: \Pe\left(\cos\theta + \int_0^{Pe} dt e^{t\cos\theta} \frac{I_1(t)}{t}\right)  \label{eq:sig}

291: \end{equation}

292: which is uniformly accurate in angle $\theta$ up to $Pe=10^{-1}$, let

293: us keep only the leading terms,

294: \begin{equation}

295:   \sigma \sim

296:   \frac{1+Pe\cos\theta}{-\log(Pe/4)-\gamma}-Pe\cos\theta \label{eq:sigsim} 

297: \end{equation}

298: where $\gamma = 0.577215\ldots$ is Euler's constant. 

299: 

300: In the final stage of collapse where $-\log Pe \gg \gamma$, the interface

301: is asymptotically circular, $g(w,t) \sim a_1(t) w$. From

302: (\ref{eq:sigsim}), the radius in this regime satisfies, $a_1 da_1/dt

303: \sim -1/\log a_1$, and thus has the asymptotic form,

304: \begin{equation}

305: a_1 \sim \sqrt{\frac{4(t-t_c)}{\log\frac{1}{t-t_c}-\log\log

306:     \frac{1}{t-t_c} + \ldots }}

307: \end{equation}

308: 

309: To describe the dynamics starting at small $Pe$ and ending just prior

310: to collapse, it is reasonable to further set $\log Pe\approx$

311: constant.  In this intermediate regime, the interfacial dynamics is

312: given by

313: \begin{equation}

314:   \re(\overline{w g^\prime} g_t ) = -1 + b\, a_1(t) \cos\theta \ \

315:   \mbox{ for } \ w=e^{i\theta}   \label{eq:pgb}

316: \end{equation}

317: (with a suitable choice of units). The $\cos\theta$ term has the

318: effect of exciting new modes in the shape of the dissolving

319: solid, which are not

320: present in the initial condition. We will see that a circle, $g(w,0) =

321: w$, translates away from the flow, $g(w,t) = w + a_0(t)$, as its

322: upstream side dissolves more quickly prior to collapse. It can also be

323: shown that an $(N+1)-$fold perturbation of a circle (\ref{eq:in}) will

324: translate and develop an $N-$fold perturbation, $g(w,t) =

325: a_1(t)w+a_0(t) + a_{-N+1}/w^{N-1} + a_N(t)/w^N$. Higher Fourier modes

326: in $\sigma(w,t)$ excite additional terms in the Laurent

327: expansion of the shape, $g(w,t)$.

328: 

329: \begin{figure}

330: \includegraphics[width=1.2\linewidth]{inflow}

331: \caption{ \label{fig:inflow} Inward advection-diffusion-limited

332:   dissolution of a tilted elliptical solid in a uniform background

333:   flow from left to right ($c=0.5i$, $b = 0.8$, $t=0, 0.04,

334:   0.08,\ldots,0.36, 0.3749$).  The major axis of the ellipse remains

335:   oriented at $\pi/4$, while its center of mass moves at an angle of

336:   $\pi/6$ relative to the background flow.  At time $t_c = 0.375$, the

337:   solid disappears at the point $a_0(t_c) = 0.4+0.2i$.  }

338: \end{figure}

339: 

340: {\it Collapse of a tilted ellipse in a uniform flow. --} To illustrate

341: these principles, let us consider a solid ellipse, $g(w,0) = w + c/w$

342: ($c<1$), which is tilted at an angle $\phi=(\mbox{arg} c)/2$ with

343: respect to a background flow in the positive $x$ direction. During

344: dissolution, the shape remains an ellipse, but its center of mass,

345: $a_0(t)$, move. An example is shown in Fig.~\ref{fig:inflow}.

346: 

347: The Laurent expansion is given by

348: %\begin{equation}

349: $g(w,t) = a_1(t)w + a_0(t) + a_{-1}(t)/w$,  

350: %\label{eq:inflow}

351: %\end{equation}

352: where $a_1(t)$ is real, but $a_0(t)$ and $a_{-1}(t)$, are

353: complex. Substituting into (\ref{eq:pgb}) and integrating shows that

354: the conformal has a square-root singularity,

355: \begin{equation}

356: a_1(t) = \sqrt{1 - \frac{t}{t_c}}, \ \ t_c = \frac{1-|c|^2}{2},

357: \end{equation}

358: since the area decreases linearly to zero (in the approximation of

359: constant total flux): $A(t) = A(0)-2\pi t$, where $A(0) =

360: \pi(1-|c|^2)$. The collapse time depends on the initial shape through

361: $|c|$. The slowest collapse, $t_c=1/2$, occurs for a circle, $c=0$,

362: while the collapse time tends to zero for a very elongated ellipse,

363: $|c| \to 1$, regardless of orientation.  

364: 

365: For a constant total flux, the collapse time $t_c$ does not depend on

366: the bias introduced by the flow velocity (through $b$), although the

367: flow affects the time scale through the initial P\'eclet number,

368: $Pe_0$. Mathematically, this is a general consequence of the conformal

369: invariance of the transport process~\cite{bazant03}, which causes the

370: total flux (Nusselt number) to depend only on the conformal radius and

371: not the asymmetric shape of the particle~\cite{choi05}. Physically,

372: the enhancement of dissolution on the upstream side of the solid is

373: cancelled by the reduction in dissolution on the downstream side.

374: 

375: During dissolution, the ellipse keeps its shape and its orientation

376: with respect to the flow direction since $a_{-1}(t) = c

377: a_1(t)$. However, the center of mass moves away from the flow, but

378: also away from the end of the ellipse which protrudes upstream. The

379: velocity of the center of mass is constant and in the $1+c$ direction,

380: \begin{equation}

381: a_0(t) = \frac{b(1+c)t}{1-|c|^2}.

382: \end{equation}

383: Note that the center of mass does not move along the major axis of the

384: ellipse at angle $\phi$, but instead at an oblique angle $\theta$

385: given by $\sin \theta = |c| \sin(2\phi-\theta)$. The final collapse

386: occurs at the point, $a_0(t_c) = b(1+c)/2$. In the simplest case,

387: $c=0$, a circle maintains its shape while its center of mass

388: translates away from the flow at (dimensionless) velocity $b$ until

389: collapse occurs at $x=b/2$ at time $t_c=b/2$. 

390: 

391: For any initial shape, the center of mass will drift away from the

392: flow, as well as away from any protrusions in the direction of the

393: flow. For non-elliptical shapes, it is nontrivial to predict the exact

394: time and place of collapse, even for the simplified dynamics of

395: Eq.~(\ref{eq:pgb}). For more general transport-limited

396: dynamics~\cite{bazant03}, predicting the collapse seems like an

397: interesting open problem.

398: 

399: 

400: 

401: 

402: The author thanks J. Propp for an introduction to IDLA.

403: 

404: \begin{thebibliography}{99}

405: 

406: \bibitem{gruzberg04} I. Gruzberg and L. P. Kadanoff,

407:   J. Stat. Phys. {\bf 114}, 1183 (2004).

408: 

409: \bibitem{bauer04} M. Bauer and D. Bernard, cond-mat/0412372.

410: 

411: \bibitem{handbook05} M. Z. Bazant and D. Crowdy, in {\it Handbook of

412:     Materials Modeling}, Vol. I, ed. by S. Yip, Art. 4.10 (Springer,

413:   2005).

414: 

415: \bibitem{bensimon86} D. Bensimon, L. P. Kadanoff, S. Liang,

416:   B. I. Shraimain, and C. Tang, Rev. Mod. Phys. {\bf 58}, 977 (1986).

417: 

418: \bibitem{howison92} S. D. Howsion, Euro. J. Appl. Math. {\bf 3}, 209 (1992).

419: 

420: \bibitem{hastings98} M. Hastings and L. Levitov, Physica D {\bf 116},

421:   244 (1998).

422: 

423: \bibitem{bazant03} M. Z. Bazant, J. Choi, and B. Davidovitch,

424:   Phys. Rev. Lett. {\bf 91}, 04503 (2003).

425: 

426: \bibitem{bazant04} M. Z. Bazant, Proc. Roy. Soc. Lond. A {\bf 460},

427:   1433 (2004).

428: 

429: \bibitem{david05} B. Davidovitch, J. Choi, and M. Z. Bazant,

430:   Phys. Rev. Lett. {\bf 95}, 075504 (2005).

431: 

432: \bibitem{maksimov76} V. A. Maksimov, Prikl. Mat. Mekh. {\bf 40}, 264 (1976).

433: 

434: \bibitem{kornev88} K. G. Kornev and V. A. Chugunov,

435:   Prikl. Mat. Mekh. {\bf 52}, 773 (1988).

436: 

437: \bibitem{kornev94} K. Kornev and G. Mukhamadullina,

438:   Proc. Roy. Soc. Lond. A {\bf 447}, 281 (1994).

439: 

440: \bibitem{alimov98}  M. Alimov, K. Kornev, and G. Mukhamadullina, SIAM

441:   J. Appl. Math. {\bf 59}, 387 (1998).

442: 

443: \bibitem{meyer82} G. H. Meyer, in {\it Numerical Treatment of Free

444:     Boundary Value Problems}, p. 202 (Birkh\"auser, 1982).

445: 

446: \bibitem{shraiman84} B. I. Shraiman and D. Bensimon, Phys. Rev. A {\bf 30},

447:   2840 (1984).

448: 

449: \bibitem{howison86} S. D. Howison, SIAM J. Appl. Math. {\bf 46}, 20

450:   (1986). 

451: 

452: \bibitem{paternson84} L. Paterson, Phys. Rev. Lett. {\bf 52}, 1621

453:   (1984).

454: 

455: \bibitem{witten81} T. Witten and L. M. Sander, Phys. Rev. Lett. {\bf

456:   47}, 1400 (1981).

457: 

458: \bibitem{tang85} C. Tang, Phys. Rev. A {\bf 31}, 1977 (1985).

459: 

460: \bibitem{meakin86} P. Meakin and J. M. Deutch, J. Chem Phys. {\bf 85}, 2320

461:   (1986).

462: 

463: \bibitem{krug91} J. Krug and P. Meakin, Phys. Rev. Lett. {\bf 66}, 703 (1991).

464: 

465: \bibitem{lawlor92} G. Lawlor, M. Bramson, and D. Griffeath,

466:   Ann. Prob. {\bf 20}, 2117 (1992).

467: 

468: \bibitem{wagner54} C. Wagner, J. Electrochem. Soc. {\bf 101}, 225

469:   (1954).

470: 

471: \bibitem{edwards53} J. Edwards, J. Electrochem. Soc. {\bf 100}, 189C,

472:   223C (1953).

473: 

474: \bibitem{lawlor95} G. F. Lawlor, Ann. Probab. {\bf 23}, 71 (1995).

475: 

476: \bibitem{nick93} R. W. D. Nickalls, Math. Gazette {\bf 77}, 354 (1993).

477: 

478: \bibitem{choi05} J. Choi, D. Margetis, T. M. Squires, and

479:   M. Z. Bazant, J. Fluid Mech. {\bf 536}, 155 (2005).

480: 

481: \end{thebibliography}

482: 

483: \end{document}

484: 

485: %

486: 

487: 

488: {\it Conclusions. --}

489: 

490: reverse laplacian growth: exact solutions to compare to DLE, IDLA

491: analysis and deterministic models; by comparing exact continuous

492: solution to stochastic growth simulations, can check mean-field

493: approximation, as in ~\cite{david05} [and Sander for DLA in a

494: channel]. due to the stable growth, it may be that the naive mean

495: field approx works fine...

496: 

497: advection-diffusion-limited dissolution. new set of basic questions:

498: for given tranport process, when and where does collapse of a finite

499: solid occur?

500: 

501: 

502: ------

503: 

504: 

505: 

506: 

507: Following a suggestion of Jim Propp, in these notes we derive some

508: exact solutions for general problems of ``reverse Laplacian growth'',

509: where an interface advances {\it away} from a region of quasi-steady

510: heat or mass transport which drives the dynamics, rather than toward

511: it as in Laplacian growth, DLA, viscous fingering, etc. As a result,

512: the well-known finite-time singularities of Laplacian growth are

513: replaced by stable dynamics leading to simple, smooth shapes.

514: 

515: These continuous models might describe a variety of physical

516: situations, such as the electro-dissolution of an electrode, corrosion

517: of a solid limited by the diffusion of a catalyst, the slow melting of

518: a solid limited by heat diffusion, and the displacement of a less

519: viscous fluid by a more viscous fluid in a Hele-Shaw cell or porous

520: medium.  Surface tension can be neglected since the dynamics is stable

521: and tends to minimize interfacial curvature, as we shall see below.

522: 

523: In two dimensions, conformal-map dynamics (a reverse ``Loewner

524: chain'', which expands the region of interest) can be formulated for

525: interfacial dynamics limited by any conformally invariant transport

526: process, such as advection-diffusion in a potential flow or

527: electrochemical conduction.  We suggest the term ``transport-limited

528: dissolution'' to describe all such cases, when the dynamics is

529: continuous, although here we will only consider the simplest case of

530: steady linear diffusion (described by a harmonic concentration field).

531: 

532: In each problem, the interfacial dynamics can be either discrete and

533: stochastic, or continuous and deterministic. The former case, on a

534: lattice, corresponds to what probabilists call ``internal DLA'' for

535: outward radial dynamics or ``diffusion-limited erosion'' for inward

536: radial dynamics. Here, we will present some simple exact solutions to

537: the continuous analogs of these two discrete models, although these

538: naive mean-field approximations may not precisely describe the average

539: shape of the interface in the discrete, stochastic models, as in DLA

540: and ADLA.

541: -----

542: 

543:  Paterson~\cite{paterson84} introduced the model of ``anti-DLA''

544: where random walkers in the fluid region annihiliate particles of the

545: solid region upon contact, rather than aggregating as in

546: DLA~\cite{witten81}. Tang~\cite{tang85} simulated a ``mean-field''

547: generalization of anti-DLA (where many walkers must strike the same

548: solid site before annihilation) and compared to numerical solutions of

549: Laplacian dissolution. The smoothing and residual roughness of

550: anti-DLA surfaces was studied by Meakin and Deutch~\cite{meakin86}, as

551: model of electrochemical polishing. Krug and Meakin~\cite{krug91}

552: introduced the term ``diffusion-limited erosion'' (DLE) for anti-DLA

553: and analyzed the roughness of a flat front. Bram

554: 

555: