1: %\documentclass[aps,twocolumn,amsmath,amssymb]{revtex4}

2: \documentclass[aps,amsmath,amssymb]{revtex4}

3: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}

4: 

5: % DO NOT ERASE OR MOVE THIS LINE. IT MUST BE WITHIN FIRST 8 LINES IN A FILE.

6: \newcommand{\LastChange}{Time-stamp: "2006-06-14 14:45:16 aga SelfSimilarDraft7.tex"}

7: 

8: 

9: \usepackage[rflt]{floatflt}

10: \usepackage{hyperref}

11: \usepackage{graphicx}% Include figure files

12: %\usepackage{dcolumn}% Align table columns on decimal point

13: %\usepackage{bm}% bold math

14: %\usepackage[showrefs]{refcheck}

15: %\usepackage{epsf}

16: %\nofiles

17: \begin{document}

18: 

19: \newcommand{\Det}{\,{\rm{Det}\,}}

20: \newcommand{\Tr}{\,{\rm{Tr}\,}}

21: \newcommand{\tr}{\,{\rm{tr}\,}}

22: 

23: \newcommand{\bomega}{\bar{\Omega}}

24: \newcommand{\wsf}{\omega_{\rm sf}}

25: \newcommand{\sign}{{\rm sign}}

26: 

27: \renewcommand{\Re}{{\rm Re}}

28: \renewcommand{\Im}{{\rm Im}}

29: 

30: \newcommand{\COMMENT}{\large\bf}

31: \newcommand{\FigWidth}{0.9\columnwidth}

32: 

33: 

34: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

35: %newfont{\Bbb}{msbm10 scaled 1\@ptsize00}

36: \newcommand{\CC}{\mbox{\Bbb C}}

37: \newcommand{\NN}{\mbox{\Bbb N}}

38: \newcommand{\PP}{\mbox{\Bbb P}}        %%% "OPEN" LETTERS %%%%

39: \newcommand{\QQ}{\mbox{\Bbb Q}}

40: %\newcommand{\SS}{\mbox{\Bbb S}}

41: \newcommand{\RR}{\mbox{\Bbb R}}

42: \newcommand{\TT}{\mbox{\Bbb T}}

43: \newcommand{\XX}{\mbox{\Bbb X}}

44: \newcommand{\ZZ}{\mbox{$\mathbb Z$}}

45: \newcommand{\z}{\raise-1pt\hbox{$\mbox{\Bbbb Z}$}}

46: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

47: 

48: \def\beq{\begin{equation}}

49: \def\eeq{\end{equation}}

50: \def\p{\partial}

51: \def\G{\Gamma}

52: 

53: \def\MM{{\sf M}}

54: \def\DD{{\sf D}}

55: \def\Dc{\CC \setminus {\sf D}}

56: \def\Uc{{\sf U}_{{\rm ext}}}

57: \def\UU{{\sf U}}

58: \def\BB{{\sf B}}

59: 

60: 

61: 

62: 

63: \def\unitary{{\mathscr U}}

64: \def\hermitian{{\mathscr H}}

65: \def\normal{{\mathscr N}}

66: \def\complex{{\mathscr C}}

67: 

68: \def\kernel{K}

69: \def\lbracket{\left <}

70: \def\rbracket{\right >}

71: \def\vphcl{\varphi _{0}}

72: \def\rhocl{\rho _{0}}

73: 

74: \def\square{\hfill

75: {\vrule height6pt width6pt depth1pt} \break \vspace{.01cm}}

76: 

77: 

78: 

79: 

80: 

81: \title{Self-similarity in Laplacian Growth}

82: %\small draft}

83: 

84: \author{Ar.~Abanov}

85: %\email[]{aga@lanl.gov}

86: %\homepage[]{}

87: \affiliation{

88:             Department of Physics, MS 4242,

89:             Texas A\&M University,

90:             College Station, TX 77843-4242, USA

91: }

92: \author{M.~Mineev-Weinstein}

93: \affiliation{Los Alamos National Laboratory, MS-P365,

94: Los Alamos, NM 87545, USA

95: }

96: \author{A.~Zabrodin}

97: \affiliation{Institute of Biochemical Physics, Kosygina str. 4,

98: 119991 Moscow, Russia; also at ITEP, Bol. Cheremushkinskaya str. 25,

99: 117259 Moscow, Russia}

100: 

101: \date{\today}

102: %\date{Last Change \LastChange}

103: 

104: \begin{abstract}

105: We consider Laplacian Growth of self-similar domains in different

106: geometries. Self-similarity determines the analytic structure of the

107: Schwarz function of the moving boundary. The knowledge of this

108: analytic structure allows us to derive the integral equation for the

109: conformal map. It is shown that solutions to the integral equation

110: obey also a second order differential equation which is the one

111: dimensional Schroedinger equation with the

112: $\mbox{sinh}^{-2}$-potential. The solutions, which are expressed

113: through the Gauss hypergeometric function, characterize the geometry

114: of self-similar patterns in a wedge. We also find the potential for

115: the Coulomb gas representation of the self-similar Laplacian growth

116: in a wedge and calculate the corresponding free energy.

117: \end{abstract}

118: 

119: %\pacs{PACS numbers:75.10.Nr, 05.50.+q, 75.10.Jm}

120: 

121: 

122: \maketitle

123: \tableofcontents

124: 

125: \section{Introduction}

126: 

127: The Laplacian Growth (LG) phenomena has been studied for a long time

128: in various physical systems ranging from solidification to bacterial

129: colony growth \cite{Howison}. Displacement of viscous fluid by an

130: invicsid one (to be referred as {\it oil} and {\it air}) between two

131: closely spaced horizontal plates (the Hele-Shaw cell) is the

132: simplest process of this kind. The Laplacian Growth is defined here

133: as the dynamics of the oil/air interface (Fig. \ref{fig:LG}a).

134: For a review, see, e.g.,

135: \cite{RMP,book}.

136: 

137: The process is unstable and produces in a long time an asymptotical

138: shape which is either a finger \cite{ST,TRHC} (in channel and wedge

139: geometries) or a fractal with the same universal characteristics

140: \cite{HLS-DLA} as in diffusion-limited aggregation

141: \cite{WittenSanderPRL,PhTod} (in all geometries).

142: 

143: It has been recently discovered that the Laplacian Growth is

144: \begin{itemize}

145: \item

146: described by the 2D Toda integrable hierarchy \cite{MWZ},

147: \item

148: equivalent to the evolution of the support of eigenvalues in

149: certain ensembles of random matrices \cite{MWZ,KKMWZ},

150: \item

151: relevant to the 2D fermions in the magnetic field

152: \cite{ABWZ}, and

153: \item

154: equivalent to Whitham equations of the soliton theory \cite{KMWZ}.

155: \end{itemize}

156: 

157: \begin{figure}

158: %\includegraphics[scale=1.2]{WZNW}

159: %\includegraphics[width=2in,height=2in]{figures/HeleShaw}

160: \includegraphics[width=\FigWidth]{figures/HeleShaw}

161: 

162: 

163: \caption{\label{fig:LG} Laplacian Growth in a Hele-Shaw cell for the

164: radial {\it a}), channel {\it b}), and wedge {\it c}) geometries. }

165: \end{figure}

166: 

167: The most well studied geometry is a rectangular Hele-Shaw cell

168: (Fig. \ref{fig:LG}b) where air enters the channel from the left

169: side. The boundary dynamics in this system was considered first by

170: Saffman and Taylor in 1958 \cite{ST}.  They observed a relatively

171: stable finger (known today as the Saffman-Taylor finger) formed

172: after a competition of many fingers generated by an interfacial

173: instability. They also found a continuous family of traveling wave

174: solutions (in the absence of surface tension) and posed a problem

175: of selection of a proper member of the family to describe the

176: observable finger. (See the development of the selection problem

177: in \cite{Asympt}-\cite{M98}). In the radial geometry (Fig.

178: \ref{fig:LG}a) an asymptotic shape is a fractal with nontrivial

179: universal self-similar features \cite{HLS-DLA}, which are still

180: not explained. Another geometry which has been studied both

181: theoretically and experimentally is the wedge geometry (Fig. \ref{fig:LG}c).

182: In this setting air enters the wedge filled with oil through the

183: wedge corner and pushes oil outward. A number of exact solutions

184: were found in this geometry by different methods

185: \cite{TRHC},\cite{Benamar}-\cite{MV}.

186: Some of them are

187: analogs of ``traveling wave'' solutions. They are also called

188: fingers.

189: 

190: The latter experimental setting is of the great interest for several

191: reasons. One of them is that traveling wave solutions in a wedge

192: geometry are the only known explicit solutions of LG described by

193: conformal maps (from the unit circle), which derivatives contain

194: non-integer singularities.  Moreover, they have a very important

195: property of being self-similar.

196: 

197: It is the latter property that we are going to study in detail in

198: this paper. We prove that the only smooth

199: self-similar solution in the

200: radial geometry is an ellipse. The family of self-similar

201: solutions with a singularity at the origin is much more

202: interesting. We show that the exact solutions in the wedge geometry

203: first obtained in \cite{Benamar} belong to this class.

204: 

205: To study self-similar solutions of general type,

206: we develop a new method which allows us to represent

207: the results for the wedge geometry in a more explicit

208: and transparent form. We will show that the

209: self-similarity completely determines the analytical structure of

210: the Schwarz function of the moving boundary.  Using the knowledge of

211: the analytical structure we will derive an integral equation for the

212: conformal map from the exterior of the unit circle to the domain

213: filled by oil.  We will establish then the equivalence of this

214: integral equation to the Schrodinger equation with a well-studied

215: (Poeschle-Teller) potential \cite{Flugge} which was derived for

216: LG in wedge earlier \cite{Benamar,Tu}.  We will also calculate the

217: free energy of the 2D Coulomb gas representation of LG for the case

218: of self-similar solutions in a wedge.

219: 

220: The paper is organized as follows. In Section \ref{sec:SF}, we

221: introduce the notion of the Schwarz function of a curve. In Section

222: \ref{sec:LGSF}, we formulate the Laplacian Growth in terms of the

223: Schwarz function of the moving boundary.  Section \ref{sec:SSSGR}

224: treats the property of self-similarity in terms of the

225: Schwarz-function. Sections \ref{sec:W} and \ref{sec:IndDiff} are

226: devoted to the self-similar solutions in the wedge. They contain

227: geometric interpretation, derivation of both integral and

228: differential equations for these solutions, proof of their

229: equivalence, and the analytic continuation of the solutions.  In

230: section \ref{sec:Coulomb}, we present the 2D Coulomb gas

231: representation of the self-similar solution and, finally, in section

232: \ref{sec:SLC}  we investigate different limits and special cases

233: which are of particular interest for physics.

234: 

235: \section{Schwarz Function}\label{sec:SF}

236: 

237: The Schwarz function \cite{Davis} plays a very important role in

238: what follows, so we devote some space here to introduce this

239: object. Let a curve ${\cal C}$ in the plane be given by an

240: equation

241: \begin{equation}

242: F(x,y)=0,

243: \end{equation}

244: where $x$ and $y$ are cartesian coordinates in the plane. Replacing

245: them by complex coordinates $z=x+iy$, $\bar{z}=x-iy$, we rewrite the

246: equation for the curve in the form

247: \begin{equation}

248: G(z,\bar{z})=F\left(\frac{z+\bar{z}}{2},\frac{z-\bar{z}}{2i}\right)=0

249: \end{equation}

250: Solving this equation with respect to $\bar{z}$ (always possible

251: locally for analytic curves) we find

252: \begin{equation}\label{eq:Schwarz}

253: \bar{z}=S(z)

254: \end{equation}

255: The so defined function $S(z)$ is called the Schwarz function of the

256: curve ${\cal C}$.  The equation (\ref{eq:Schwarz}) is complex and

257: thus contains two real equations, which are mutually dependent,

258: since each of them bears the same information as the Eq. (1).  The

259: constraint on $S$ is clear after taking the conjugate of equation

260: (\ref{eq:Schwarz}):

261: \begin{equation}\label{eq:unitarity}

262: z=\bar{S}(S(z)).

263: \end{equation}

264: This property of the Schwarz function is called unitarity \cite{MWZ}.

265: The Schwarz function is holomorphic and has a great advantage over a

266: conformal map description of curves, as it does not imply any

267: particular parametrization on the curve.

268: 

269: The geometrical meaning of the Schwarz function becomes clear if one

270: imagines that the curve ${\cal C}$ is a mirror.  Then for the object

271: at a point $z$ close enough to the mirror the image is at the

272: point $\overline{S(z)}$.  For instance, the Schwarz function

273: for a circle centered at the

274: origin is $S(z)=R^2/z$, where $R$ is the radius

275: of the circle.

276: 

277: Given the Schwarz function $T(w)$ of a closed curve ${\cal T}$ in

278: the (``mathematical") plane $w$ and the conformal map $f(w)$ from

279: the outside/inside of that curve onto the outside/inside of a curve

280: ${\cal C}$ on the (``physical") plane $z$, one can easily find the

281: Schwarz function of the curve ${\cal C}$:

282: $S(z)=\bar{f}(T(f^{-1}(z)))$, where $f^{-1}$ denotes the inverse

283: function, and we adopt the standard notation $\bar f(z)= \overline{f

284: (\bar{z})}$. In particular, if the curve ${\cal T}$ in the

285: mathematical plane is the unit circle, we find that the Schwarz

286: function is given by

287: \begin{equation}\label{eq:Scircle}

288: S(z)=\bar{f}(1/w), \quad \quad z=f(w)

289: \end{equation}

290: 

291: 

292: \section{Laplacian Growth and the Schwarz function}\label{sec:LGSF}

293: 

294: In this section we formulate the Laplacian Growth in terms of the

295: Schwarz function of the boundary following \cite{Howison91}.

296: 

297: For a quasistatic motion of a viscous fluid between two closely

298: spaced parallel surfaces, the Navier-Stokes equation is reduced

299: to the Darcy's Law \cite{Lamb} which states that the 2D velocity

300: $\bf{v}$ of the fluid is proportional to the gradient of pressure

301: $p$.  In properly rescaled variables it has the form

302: \begin{eqnarray}

303: &&{\bf v}=-{\bf \nabla} p\label{eq:darcy}\\

304: &&\Delta p =0 \label{eq:continuity}

305: \end{eqnarray}

306: where the last equation (\ref{eq:continuity}) follows from the

307: incompressibility condition $\mbox{div}\, \bf{v}=0$. In the air

308: (zero viscosity) domain pressure is constant, $p=\mbox{const}$.

309: Neglecting surface tension, one concludes that $p(x,y)$ is a

310: harmonic function in the oil domain constant along the boundary. In

311: this paper we consider only a single sink placed at infinity. In

312: this case pressure $p$ behaves at infinity as

313: \begin{equation}

314: p(x,y) \simeq -\frac{Q}{2\pi}\log R, \quad R\to \infty ,

315: \end{equation}

316: where $R^2=x^2+y^2$ and $Q$ is the pumping rate (that is the

317: decrease of the oil area per unit time).

318: 

319: 

320: Introducing the holomorphic complex potential $W(z)=-p+i\psi$ and

321: the complex velocity $v=v_x+iv_y$, one can rewrite equation

322: (\ref{eq:darcy}) in the form $v=\overline{\partial_z W}$. In the

323: case of a single air bubble in the cell the complex potential is

324: given by

325: \begin{equation}\label{eq:Ww}

326: W(z,t) = \frac{Q}{2\pi}\log w(z,t)

327: \end{equation}

328: Here $w(z,t)$ is (time dependent) conformal map from the oil domain

329: onto the exterior of the unit circle normalized in such a way that

330: $w(z, t)\simeq z/r(t)$ as $|z|\to \infty$ with a real positive

331: coefficient $r(t)$. This coefficient is called the (external)

332: conformal radius

333: of the air domain. In what follows we set $Q=\pi$. In these units,

334: area of the growing domain is equal to $\pi t$.

335: 

336: Let us describe the oil/air interface by its time-dependent Schwarz

337: function $S(z,t)$.  If at time $t$ a point $z$ is at the boundary,

338: $\bar{z}=S(z)$, then at time $t+dt$ the point $z+vdt$ belongs to the

339: new boundary, i.e., $\overline{z+vdt}=S(z+vdt,t+dt)$. Therefore, on

340: the boundary it holds

341: \begin{equation}\label{eq:infinitezimal}

342: \bar{v}=\dot{S}+S'v,

343: \end{equation}

344: where dot and prime denote partial derivatives with respect to $t$ and $z$.

345: 

346: The unit tangent vector to the boundary equals to $dz/dl$, where $l$

347: is the arclength $dl =\sqrt{dzd\bar{z}}=dz \sqrt{S'}$, so

348: \begin{equation}\label{eq:tangent}

349: dz/dl =\frac{1}{\sqrt{S'}}

350: \end{equation}

351: 

352: 

353: In the case of zero surface tension pressure is constant along the boundary,

354: and so its gradient is orthogonal to it.

355: In the complex notation we have

356: \begin{equation}

357: \mbox{Re}\frac{\bar{v}}{\sqrt{S'}}=0.

358: \end{equation}

359: From unitarity (\ref{eq:unitarity}) we find $S'=1/\bar{S'}$.

360: Together with the previous equation this gives

361: \begin{equation}

362: \bar{v}+vS'=0.

363: \end{equation}

364: Using equation (\ref{eq:infinitezimal}) and (\ref{eq:Ww}) (with

365: $Q=\pi$) we then find

366: \begin{equation}\label{eq:LGS}

367: \dot{S}=2\bar{v}=2\partial_z W = \partial_z \log w

368: \end{equation}

369: After analytical continuation from the boundary, this equation

370: is valid everywhere in the oil domain.

371: Equation (\ref{eq:LGS}) connects the change on the Schwarz function

372: of the boundary with the solution to the Dirichlet

373: boundary value problem for the

374: oil domain with given sources/sinks.

375: 

376: One can represent the Schwarz function in the form

377: \begin{equation}\label{eq:S+-}

378: S(z)=S_{-}(z)+S_{+}(z)

379: \end{equation}

380: where $S_{-}$ is regular in the oil domain, and $S_{+}$

381: is regular in the air domain. The function

382: $S_{+}(z)$ is given by the Cauchy integral

383: \beq\label{basic3}

384: S_{+}(z,t)=\frac{1}{2\pi i} \oint_{\p \DD (t)}

385: \frac{\bar z' dz'}{z' - z}

386: \eeq

387: where $\DD$ is the air domain and the point $z$ is

388: assumed to be inside it. (A similar Cauchy integral, with the point

389: $z$ outside $\DD$, can be written for $S_{-}$.)

390: As the function $\partial_z W$ has no singularities in the oil domain, we find

391: \begin{equation}\label{eq:S+dot}

392: \dot{S}_{+}(z)=0

393: \end{equation}

394: From this we conclude that all harmonic moments of the oil domain

395: \begin{equation}\label{eq:harmMoments}

396: t_k=

397: \frac{1}{2\pi i k}\oint_{\p \DD} \bar z z^{-k}dz

398: %\frac{1}{2i}\int_{D_{-}}z^{-k-1}dzd\bar{z}=

399: \frac{1}{2\pi i k}\oint_{\p \DD} S_{+}(z)z^{-k}dz, \quad

400: k\geq 1,

401: \end{equation}

402: are integrals of motion.  This fact was observed by S.Richardson

403: \cite{Richardson}.

404: 

405: 

406: \section{Self-similar solutions. General relations}\label{sec:SSSGR}

407: 

408: Self-similarity expresses the fact that the evolution of the domain

409: is just a dilatation of its shape.  It means that if initially the

410: Schwarz function of the boundary is $S(z)$, then at any other time

411: $\bar{z}/r(t)=S(z/r(t))$ for the points on the boundary, where the

412: real dilatation parameter, $r(t)$, initially equals to $1$. In other

413: words,

414: \beq\label{SS2} S(z,t)=r(t)S(z/r(t)).

415: \eeq

416: Let us take $r(t)$

417: to be the conformal radius of the growing (air) domain. It is defined as

418: the derivative at infinity of the conformal mapping from the

419: exterior of the unit circle onto the oil domain normalized by the

420: conditions that infinity is sent to infinity and $r(t)>0$. Then

421: $S(z)$ is the Schwarz function of the oil domain scaled so that its

422: conformal radius is $1$.

423: 

424: The self-similar reduction of eq.(\ref{eq:LGS}) reads

425: \beq

426: \label{SS3} S(z)-zS'(z)=2A \, \p_z \log w(z)

427: \eeq

428: where $r^{2}(t)= t/A$ and $w(z)$ is the conformal map of the domain

429: with conformal radius $1$. Taking the contour integral of equation

430: (\ref{SS3}) over the boundary we find that the constant $A$ is the

431: area (divided by $\pi$) of the domain with conformal radius $1$.

432: 

433: Since the r.h.s. of (\ref{SS3}) is an analytic function in the

434: exterior of $\DD$, this equation implies the relation

435: $$

436: S_{+}(z)-zS_{+}'(z)=0

437: $$

438: with the general solution \beq\label{SS4} S_{+}(z)=sz \eeq where $s$

439: is a constant. This solution corresponds to an ellipse with the axes

440: ratio $(1+s)/(1-s)$. We thus see that {\it regular} self-similar

441: solutions are exhausted by ellipses.

442: 

443: This result is not surprising and can be seen right from equations

444: (\ref{eq:S+dot}) and (\ref{eq:harmMoments}). Under the dilatations

445: $z\rightarrow rz$ the $k$th harmonic moment changes

446: as $t_{k}\rightarrow t_{k}r^{2-k}$. On the other hand, the harmonic

447: moments are integrals of motion and as such cannot change. This

448: leaves as with the only possibility to have all the moments equal to

449: $0$ except $t_{2}$, which is exactly the statement (\ref{SS4}).

450: 

451: %\begin{floatingfigure}[l]{2.1in}

452: \begin{figure}

453: \includegraphics[width=2in,height=2in]{figures/SelfCrossing}

454: \caption{\label{fig:SelfCrossing} A singular self-crossing domain.

455: The thick lines are the branch cuts of the function $S_{+}$.

456: Spacings between two adjacent fingers are called fjords.}

457: \end{figure}

458: %\end{floatingfigure}

459: 

460: However, there is a family of {\it singular} self-similar solutions

461: of equation (\ref{SS3}) for which the interface may cross itself at

462: the origin, see Fig. \ref{fig:SelfCrossing} (to be prepared

463: for the next section, the domain

464: looks like $\ZZ _N$-symmetric one

465: but in general no additional symmetry is implied).

466: For such solutions the

467: constant $s$ in (\ref{SS4}) is different in different sectors of the

468: plane and, therefore, $S_{+}$ has jumps across some rays with

469: $\mbox{arg}\, z = \mbox{const}$ starting at $0$ and coming to

470: $\infty$.  We stress that these rays are singularities of $S_{+}$,

471: so they do not depend on $t$. Let $\mit\Gamma$ be such a ray. Take

472: the jump of the both sides of eq. (\ref{SS3}) on this ray. Since the

473: r.h.s. is regular, we get

474: $$

475: [S(z)]_{\mit\Gamma}-z[S'(z)]_{\mit\Gamma}=0\,,

476: \quad z\in \mit\Gamma

477: $$

478: (the jump is defined as $[S(z)]_{\mit\Gamma}=

479: S(ze^{+i0})-S(ze^{-i0})$, $z\in \mit\Gamma$), hence

480: $$

481: [S(z)]_{\mit\Gamma}=c(\mit\Gamma )\, z\,,

482: \quad z\in \mit\Gamma

483: $$

484: 

485: The jump of $S(z)$ on the branch cut is related to the geometry of

486: the interface around the origin.  Let us choose coordinates so that

487: the branch cut is along the positive real semi-axis. Consider a

488: vicinity of the point $z=0$. The function $S'(z)$ has a constant

489: jump on the positive part of the real axis equal to $c=c(\mit\Gamma )$.

490: On the other hand, we know that the unit tangent vector to the

491: boundary curve is (\ref{eq:tangent}) $l (z)=1/\sqrt{S'(z)}$, so

492: using the notation in Fig. \ref{fig:jump}, we write

493: $l_{1}^{-2}=e^{-2i\theta_1}= S'(0_{+})$,

494: $l_{2}^{-2}=e^{-2i(\pi -\theta_2 )}= S'(0_{-})$ (here $0_{\pm}$

495: means $\epsilon e^{\pm i0}$ as $\epsilon \to 0$) and so the jump is

496: $$

497: c=S'(0_+)-S'(0_-)=e^{-2i\theta_1}-e^{2i\theta_2}

498: $$

499: Introducing $\theta = \theta_1 + \theta_2$ (the

500: fjord \footnote{Spacings between two adjacent fingers are called

501: fjords.} angle) and $\phi = \theta_1 -\theta_2$, we obtain the

502: desired relation \beq\label{SS5} c=-2ie^{-i\phi}\sin \, \theta \eeq

503: We see that if $c$ is purely imaginary, then the fjord is symmetric

504: ($\phi =0$) with respect to the real axis.

505: \begin{figure}

506: \includegraphics[width=2in,height=2in]{figures/jump}

507: \caption{\label{fig:jump}

508: The jump of the Schwarz function is related to the fjord angle.}

509: \end{figure}

510: 

511: The connection between the jump of $S_{+}$ and the fjord angle can

512: be understood purely geometrically. Let us consider a symmetric

513: fjord, i.e., the ray $\mit\Gamma $ is the symmetry axis.  Very close to

514: the origin the boundaries are almost straight lines.  Let us

515: consider a point $ze^{+i0}$, where $z$ is on $\mit\Gamma$. Its

516: reflection in the boundary lies at $ze^{i\theta}=\overline{S(ze^{+i0})}$.

517: For the point $ze^{-i0}$ we have

518: $ze^{-i\theta}=\overline{S(ze^{-i0})}$, so

519: $$

520: [S(z)]=S(ze^{+i0})-S(ze^{-i0})=(e^{-i\theta}-e^{i\theta })z=cz,

521: $$

522: from which equation (\ref{SS5}) follows.

523: 

524: Using the representation of the Schwarz function via conformal map

525: (\ref{eq:Scircle}), equation (\ref{SS3}) can be rewritten in the

526: familiar form \beq\label{separation} \bar f (w^{-1}) \p_w f(w)-

527:  f(w) \p_w \bar f (w^{-1}) =\frac{2A}{w}.

528: \eeq

529: The constant $A$ can now be expressed as

530: \beq\label{A}

531: A=\frac{1}{2\pi i} \oint_{|w|=1} \bar f (w^{-1}) d f(w).

532: \eeq

533: 

534: We note that equation (\ref{separation}) looks as the

535: Wronskian of $f(w)$ and $\bar f(w^{-1})$. This fact suggests that

536: they are two linearly independent solutions of a second order

537: differential equation of the form

538: $$

539: w^2 f''(w)+wf'(w)+V(w)f(w)=0

540: $$

541: with a real ``potential" $V$, such that $V(w)=\bar V(w^{-1})$. Below

542: we shall see that this is indeed the case.

543: 

544: 

545: \section{Symmetric self-similar solutions and

546: solutions in the wedge}\label{sec:W}

547: 

548: In this section, we study a family of self-similar solutions with the

549: $\ZZ _N$-symmetry (Fig. \ref{fig:SelfCrossing}).

550: Equivalently, these are self-similar solutions in

551: the wedge with the angle $\theta_0 = 2\pi/N$.

552: 

553: Let $\omega =e^{2\pi i/N}$ be the primitive root

554: of unity of degree $N$. The $\ZZ _N$-symmetry means that

555: if $z \in \DD$, then $\omega z \in \DD$.

556: For the conformal map this mean that

557: $f(\omega w)=\omega f(w)$ and for the Schwarz function

558: that $S(\omega z)=\omega^{-1}S(z)$. Moreover, from the

559: integral representation (\ref{basic3}) we see that the

560: latter property holds for $S_{+}$ and $S_{-}$ separately.

561: 

562: We need to introduce the following notation: Let $\mit\Sigma_k$ be

563: the sector $\frac{2\pi}{N}(k-1) \leq \mbox{arg}\, z \leq

564: \frac{2\pi}{N}k$ and let $\mit\Gamma_k = \mit\Sigma_{k+1} \cap

565: \mit\Sigma_k$ be the ray $\mbox{arg}\, z = \frac{2\pi}{N}k$, see

566: Fig. \ref{fig:SectorsAndCuts}.

567: 

568: 

569: 

570: \begin{figure}

571: \includegraphics[width=2in]{figures/SectorsAndCuts}

572: \caption{\label{fig:SectorsAndCuts}

573: Thick lines $\Gamma_{k}$ are the branch cuts in $S_{+}$.

574: Enumeration of sectors

575: $\Sigma_{k}$ is also shown.}

576: \end{figure}

577: 

578: \subsection{The integral equation}

579: 

580: Combining self-similarity ($S_{+}(z)

581: =s_k z$ if $z\in \mit\Sigma_k$) with $\ZZ _{N}$-symmetry

582: ($S_{+}(\omega z)=\omega^{-1} S_{+}(z)$) we find that

583: $s_{k+1}=\omega^{-2}s_k$ and so

584: \beq\label{SZ1}

585: s_k =\omega^{-2k}s_0

586: \eeq

587: Therefore,

588: \beq\label{SZ2}

589: [S(z)]_{\mit\Gamma_k} =c\omega^{-2k} z\,,

590: \quad z\in \mit\Gamma_k

591: \eeq

592: where we have set

593: $c=(\omega^{-2}-1)s_0$. These jumps are singularities

594: of $S_{+}$\footnote{One can

595: restore $S_{+}$ from its jumps by means of the

596: Cauchy type integral:

597: $$

598: S_{+}(z)=\frac{1}{2\pi i} \sum_{k=0}^{N-1}

599: \int_{\mit\Gamma_k} \frac{[S_{+}(z')]_{\mit\Gamma_k}

600: dz'}{z'-z}

601: =\frac{c}{2\pi i}\int_{0}^{\infty} \sum_{k=0}^{N-1}

602: \frac{xdx}{\omega^k x-z}

603: =\,\, -\, \frac{c}{2\pi i}\sum_{k=0}^{N-1}

604: \omega^{2k} z \, \log(-\omega^{k}z)

605: $$

606: that is equal to $\omega^{-2k}s_0z$ in the sector

607: $\mit\Sigma_k$ as it must be.}.

608: 

609: The definition of the Schwarz function written in the form

610: \beq\label{SZ3} \bar f(w)=S(f(w^{-1})) \eeq implies that

611: singularities of $\bar f(w)$ inside the unit circle exactly

612: correspond to singularities of $S(z)$ outside the domain $\DD$,

613: i.e., to singularities of $S_{+}(z)$. In our case $\bar f (w)=f(w)$

614: due to the reflection symmetry with respect to the real axis, and

615: singularities of $f(w)$ are jumps on $N$ intervals $\mbox{arg}\, w

616: =\frac{2\pi}{N}k$, $0<|w|<1$ inside the unit circle, which we denote

617: by $\gamma_k$. Therefore, we can write $f(w)$ as the Cauchy type

618: integral \beq\label{SZ4} f(w)=w+\frac{1}{2\pi i} \sum_{k=0}^{N-1}

619: \int_{0}^{\omega^k} \frac{[f(v)]_{\gamma_k}dv}{v-w} \eeq (the

620: integral is over $\gamma_k$). From (\ref{SZ3}) we conclude that

621: \beq\label{SZ5}

622: \left. \phantom{A/B} [f(w)]_{\gamma_k}=-\,

623: [S_{+}(z)]_{\mit\Gamma_{N-k}} \right |_{z=f(w^{-1})}=

624: -c\omega^{-2k}f(w^{-1})

625: \eeq

626: (The minus sign is due

627: to the interchange of the two sides of the branch cut when one

628: inverts $w$ \footnote{It comes from the fact that $1/(x+i0)=1/x-i0$,

629: so the inversion interchanges the branches.}.)

630: Plugging this into (\ref{SZ4}), one

631: obtains, after some transformations:

632: $$

633: f(w)=w -\frac{c}{2\pi i}\sum_{k=0}^{N-1}

634: \omega^{-2k} \int_{1}^{\infty} \frac{f(\omega^k x)\, dx}{(1-

635: \omega^k xw)x}

636: $$

637: Since $f(\omega^k x)=\omega^k f(x)$,

638: the sum is $\sum_k \omega^{-k}/(1-\omega^k xw)$ which is

639: easily calculated to be

640: $Nxw/(1-x^N w^N)$, and so we obtain the integral equation

641: for the conformal map:

642: \beq\label{SZ6}

643: f(w)=w -\frac{cN}{2\pi i}\int_{1}^{\infty}

644: \frac{wf(x)\, dx}{1-w^N x^N}

645: \eeq

646: with the ``boundary condition" $f(1)=0$.

647: Let us denote the fjord angle by $\theta = \pi \beta$, then

648: $c=-2i \sin \, \pi \beta$ (from eq. (\ref{SS5})) and

649: the integral equation acquires the form

650: \beq\label{integral}

651: f(w) =w\, + \,  \frac{N\sin \, \pi \beta}{\pi}

652: \int_{1}^{\infty} \frac{ wf(v)\, dv}{1-w^N v^N}\,,

653: \quad \quad f(1)=0

654: \eeq

655: Without using eq. (\ref{SS5}), one can find

656: the coefficient in front of the integral by noting

657: that the fjord angle $\pi \beta$ implies that

658: $f(w)\simeq (w-1)^{\beta}$ near the point $w=1$.

659: Expanding eq. (\ref{SZ6}) around this point, one

660: obtains $c=-2i \sin \, \pi \beta$.

661: 

662: 

663: \subsection{Geometrical interpretation}\label{sec:geom}

664: 

665: There is another way to obtain the Equation (\ref{SZ6}), namely,

666: from the geometrical arguments applied to multi-logarithmic

667: solutions of 2D LG \cite{MS}.

668: 

669: We start from a simple observation that each logarithm in the

670: standard multi-logarithmic solution (anzats) for the conformal map

671: of the LG process asymptotically describes a single ``U'' shaped

672: fjord, a fjord with parallel walls. Let us consider such an anzats:

673: \begin{equation}\label{eq:anz}

674: f(w)=rw+\sum_{k}\alpha_{k}\log \left(1-a_{k}/w \right)

675: \end{equation}

676: where $\alpha_k$'s are constants of motion, while all $a_{k}$

677: ($|a_{k}|<1$) and $r$ are functions of time. Asymptotically at large

678: time $r\rightarrow \infty$ and $|a_{k}|\rightarrow 1$. Integrals of

679: motion for such a solution are

680: \begin{equation}\label{eq:beta}

681: \beta_{k}=f(1/\bar{a}_{k})=r/\bar{a}_{k}+

682: \sum_{k'}\alpha_{k'}\log \left(1-a_{k'}\bar{a}_{k} \right)

683: \end{equation}

684: At large time, when all $a_{k}$s are close to $1$, we can neglect all

685: logarithms except one. Imagine for simplicity that $a_{0}$ is real

686: and $a_{0}\rightarrow 1$, then we can write

687: \begin{eqnarray}

688: &&\beta_{0}\approx r/a_{0}+\alpha_{0}\log (1-a_{0}^{2})\approx

689: r+\alpha_{0}\log (1-a_{0})+\alpha_{0}\log 2,      \label{eq:beta0}\\

690: &&f(1)\approx r+\alpha_{0}\log (1-a_{0})=\beta_{0}-\alpha_{0}\log 2.

691: \label{eq:f1}

692: \end{eqnarray}

693: This means that asymptotically the bottom of the fjord is at the point

694: $\beta_{0}-\alpha_{0}\log 2$.

695: 

696: Let us now take a number $\delta $ such that $\delta \ll 1$ and

697: $\delta \gg 1-a_{0}$, then using $e^{i\delta }\approx 1+i\delta $

698: we can calculate

699: \begin{eqnarray}

700: &&f(1+i\delta )\approx r+\alpha_{0}\log (i\delta ),\nonumber\\

701: &&f(1+i\delta )-f(1-i\delta ) = i\pi \alpha_{0}. \nonumber

702: \end{eqnarray}

703: This shows that the width of the fjord is constant and

704: equals $\pi |\alpha_{0}|$. The direction of the fjord is given by

705: the phase of $\alpha_{0}$.

706: 

707: Using this simple correspondence between the logarithmic

708: singularities and the ``U'' shape fjords we can construct a

709: so-called ``V'' shape fjord, i.e., a fjord which walls form a

710: constant angle between them.

711: 

712: It is clear that for this construction we need to put many identical

713: fjords in such a way that they have equal distance between their

714: vertices).  All fjord vertices must be on a ray which direction is

715: the same as the direction of the fjord. For the $\ZZ _{N}$-symmetric

716: case we then write:

717: \begin{eqnarray}

718: &&f(w)=rw+\sum_{l=0}^{\infty }\sum_{k=0}^{N-1}

719: \alpha \omega^{k}\log (1-a_{l}\omega^{k}/w), \label{eq:fDiscret}\\

720: &&\beta_{l}=l\Delta =f(1/a_{l}),\label{eq:betaDiscret}

721: \end{eqnarray}

722: where $a_{l}$, $\alpha $, $\Delta $ are real and $\Delta $ is the distance

723: between the nearest $\beta $s. The numbering of $a_{l}$ starts at the largest

724: $a$, which is $1$, so

725: \begin{equation}\label{eq:l0}

726: a_{0}=1.

727: \end{equation}

728: We now want to take the limit of $\Delta, \alpha \rightarrow 0 $. At this

729: limit the spacing between $a$'s becomes small, and we can think of $a_{l}$

730: as a smooth function $a(l)$, or alternatively $l(a)$. We then can go from

731: summation to integration and write

732: \begin{eqnarray}

733: &&f(w)=rw+\alpha \sum_{k=0}^{N-1}

734: \omega^{k}\int_{0}^{\infty } \log (1-a\omega^{k}/w)dl

735: =rw+\alpha \sum_{k=0}^{N-1}

736: \omega^{k}\int_{0}^{1}\frac{l(a)}{a-w\omega^{-k}}da,

737: \label{eq:fCont}\\

738: &&l(a)=\frac{1}{\Delta }f(1/a),\label{eq:betaCont}

739: \end{eqnarray}

740: where we integrate by parts and use (\ref{eq:l0}).

741: Substituting eq. (\ref{eq:betaCont}) into (\ref{eq:fCont}), we get

742: \begin{equation}\label{eq:fGeom}

743: f(w)=rw+\frac{\alpha }{\Delta }\sum_{k=0}^{N-1}\omega^{k}\int_{0}^{1}

744: \frac{f(1/a)}{a-w\omega^{-k}}

745: \end{equation}

746: from which (\ref{SZ6}) follows.

747: We note that the relation $\beta_{l}=l\Delta $ (see (\ref{eq:betaDiscret}))

748: is crucial for the derivation and it is this relation that leads to the

749: linear integral equation.

750: 

751: 

752: \section{Integral and differential equations for

753: the conformal map}\label{sec:IndDiff}

754: 

755: \subsection{The integral equation. Direct solution}

756: 

757: In the previous section we have derived the integral equation for the

758: conformal map $f(w)$ (see (\ref{integral})). It is an inhomogeneous

759: equation. Let us transform it into a homogeneous one. Putting $w=1$

760: in (\ref{integral}), we get

761: $$

762: \frac{N\sin \, \pi \beta}{\pi}

763: \int_{1}^{\infty} \frac{ f(v)\, dv}{ v^N -1}\, =\, 1

764: $$

765: that allows us to rewrite (\ref{integral}) as

766: $$

767: f(w)=\frac{N\sin \, \pi \beta}{\pi}

768: \int_{1}^{\infty}\frac{v^N (w^N -1)w}{(v^N -1)

769: (w^N v^N -1)} \, f(v) \, dv.

770: $$

771: After the substitution

772: $$

773: f(w)= w(1-w^{-N})^{\beta} F(w^{-N})\,,

774: \quad w^{-N}=x\,, \quad \kappa =2/N

775: $$

776: the equation acquires the form \beq\label{DS2} (1-x)^{\beta -1} F(x)

777: = \frac{\sin \, \pi \beta}{\pi} \int_{0}^{1}

778: \frac{t^{-\kappa}(1-t)^{\beta -1}}{1-xt} \, F(t) \, dt \eeq which

779: makes sense for arbitrary $\kappa <1$. The structure of the r.h.s.

780: suggests to look for solutions in hypergeometric functions.

781: Specifically, we use the identity

782: $$

783: \int_{0}^{1}\frac{t^{c-1}(1-t)^{a-1}}{1-xt}

784: \phantom{a}_{2}F_{1} \left (

785: \begin{array}{c}a, \,\, b\\c \end{array}

786: ; t \right ) dt

787: $$

788: $$

789: =\, (1-x)^{-1} \, \frac{\Gamma (a) \Gamma (c-b)}{\Gamma (c-b+a)}

790: \phantom{a}_{3}F_{2} \left (

791: \begin{array}{c}a, \,\,\, c\! -\! b, \,\,\, 1 \\

792: c, \, c \! - \! b\! + \! a

793: \end{array}; \frac{x}{x-1} \right )

794: $$

795: (taken from \cite{GR}, section 7.512).

796: Notice that at $c-b+a=1$ the

797: function $_{3}F_{2}$ reduces to

798: $_{2}F_{1}$ and the r.h.s. becomes

799: $$

800: (1-x)^{-1} \, \Gamma (a) \Gamma (1-a)

801: \phantom{a}_{2}F_{1} \left (

802: \begin{array}{c}a, \, 1\! -\! a\\c

803: \end{array}; \frac{x}{x-1} \right ).

804: $$

805: Using the identities

806: $$

807: \Gamma (a) \Gamma (1-a)=\frac{\pi}{\sin \, \pi a}

808: $$

809: and

810: $$

811: \phantom{a}_{2}F_{1} \left (

812: \begin{array}{c}a, \, b\\c

813: \end{array}; z \right )

814: =(1-z)^{-a}

815: \phantom{a}_{2}F_{1} \left (

816: \begin{array}{c}a, \,\, c\! -\! b\\c

817: \end{array}; \frac{z}{z-1} \right )

818: $$

819: we see that the function

820: $$

821: F(x)=

822: \phantom{a}_{2}F_{1} \left (

823: \begin{array}{c}\beta , \,\, \beta \! -\! \kappa \\

824: 1-\kappa \end{array}; x \right )

825: $$

826: solves eq. (\ref{DS2}).

827: 

828: Summing up, we have obtained the explicit

829: formula for the conformal map:

830: \beq\label{confmap}

831: f(w)=w(1-w^{-2/\kappa})^{\beta}

832: \phantom{a}_{2}F_{1} \left (

833: \begin{array}{c}\beta , \,\, \beta \! -\! \kappa \\

834: 1-\kappa \end{array}; w^{-2/\kappa} \right ).

835: \eeq

836: For $\ZZ _N$-symmetric patterns $\kappa =2/N$

837: and $\pi \beta$ is the fjord angle. Equivalently,

838: this solution describes the self-similar growth

839: in the wedge with angle $\theta_0 =2\pi/N$.

840: Being restricted to the sector $0\leq \mbox{arg}\, w

841: \leq \pi \kappa$, the r.h.s. of (\ref{confmap})

842: makes sense for all $0< \kappa <1$ and provides

843: the solution for the self-similar growth in the wedge

844: with angle $\theta_0 =\pi \kappa$.

845: 

846: We note some simple properties of the map (\ref{confmap}). First,

847: there is the symmetry $\beta \rightarrow 1-\beta $. Second, at $\beta =0$

848: the hypergeometric function is identically $1$ and the function

849: (\ref{confmap}) is simply $f(w)=w$ that corresponds to the self-similar

850: growth of a disk or sector. In the

851: opposite limit, the function (\ref{confmap})

852: gives the conformal map to the complement of a needle-like

853: domain (a degenerate finger).

854: Let us consider two cases separately: i) $0<\kappa <1/2$ and

855: ii) $1/2<\kappa <1$ (the case $\kappa >1$

856: is discussed in Appendix A). In case i), at $\beta =\kappa$

857: the hypergeometric function is again identically $1$ and

858: $f(w)=(w^{1/\kappa }-w^{-1/\kappa })^{\kappa }$

859: is the map to the complement of a needle growing from the origin.

860: In case ii), at $\beta =1-\kappa $ we use the symmetry

861: $\beta \rightarrow 1-\beta $ and again get the same mapping.

862: Note that in case ii)

863: for values of $\beta$ in the range $1-\kappa < \beta < 1/2$ a critical

864: point of the function (\ref{confmap}) (a value of $w$ at which $f'(w)=0$)

865: is outside the unit circle, so the map (\ref{confmap}) is no longer

866: conformal. Such values of $\beta$ are thus non-physical.

867: 

868: To summarize, the physically meaningful range of the parameters

869: is as follows.

870: If $\kappa$ does not exceed $\frac{1}{2}$, then $\beta$ can be any

871: number between $0$ and $\kappa$, and $\pi (\kappa - \beta )$ is the

872: angle of the finger (the angle between the tangent lines to the

873: finger boundary at the origin). If $\frac{1}{2} < \kappa < 1$, then

874: $\beta$ can not exceed $1-\kappa$.

875: As $\beta$ approaches $1-\kappa$, the finger becomes a very thin

876: needle. Formally, in this limit the angle of the finger tends to

877: $\pi (\kappa - \beta )$ (not to zero!)

878: but the neighborhood of the origin where

879: this angle properly describes the form of the finger vanishes

880: as $\beta$ tends to $1-\kappa$.

881: 

882: With the conformal map (\ref{confmap}) known explicitly, one is able

883: to find some geometric characteristics of the growing self-similar

884: fingers. In particular, the distance from the finger tip to the

885: origin is

886: $$

887: R= |f(e^{i\pi \kappa /2)}| =2^{\beta}\!\!

888: \phantom{a}_{2}F_{1} \left (

889: \begin{array}{c}\beta , \,\, \beta \! -\! \kappa \\

890: 1-\kappa \end{array}; -1\right )

891: $$

892: (the conformal radius is $1$). Using some identities for the

893: hypergeometric function, one can represent it in a more explicit

894: form:

895: \beq\label{DS3}

896: R=2^{\beta -1} \frac{\Gamma \left (

897: \frac{\beta -\kappa}{2} \right ) \Gamma (1-\kappa )}{\Gamma (\beta -

898: \kappa ) \Gamma \left (1- \frac{\beta +\kappa}{2} \right )}.

899: \eeq

900: The area of the finger with conformal radius $1$ will be found

901: below. The result is $\frac{1}{2}\pi \kappa A$, where the constant

902: $A$ (see (\ref{A})) is given by

903: \beq\label{A2}

904: A=\frac{\Gamma

905: (1-\kappa )\Gamma (-\kappa)}{\Gamma (\beta - \kappa ) \Gamma

906: (1-\beta -\kappa )}.

907: \eeq

908: For physical values of the parameters $\kappa$ and $\beta$

909: this constant is always positive.

910: For non-physical values ($1/2<\kappa <1$ and

911: $1-\kappa <\beta <1/2$) formula (\ref{A2}) gives a negative

912: area.

913: 

914: 

915: 

916: 

917: 

918: \subsection{Differential equations for the conformal map}

919: 

920: It turns out that the

921: solution to the integral equation (\ref{integral})

922: obeys a differential equation of second order

923: (that is not surprising because the solution is expressed

924: through the hypergeometric function).

925: For the first time it was written in \cite{Tu}.

926: Specifically, the equation reads

927: \beq\label{differ1}

928: w\p_w (w \p_w f)+ N^2\, \frac{\beta

929: (1-\beta )w^N}{(w^N -1)^2}\, f = f.

930: \eeq

931: It has two linearly independent solutions:

932: \beq\label{f}

933: f(w)=w(1-w^{-N})^{\beta}

934: \phantom{a}_{2}F_{1} \left (

935: \begin{array}{c}\beta , \,\, \beta \! -\! \frac{2}{N} \\

936: 1-\frac{2}{N} \end{array}; w^{-N} \right ),

937: \eeq

938: \beq\label{g}

939: g(w)=w^{-1}(1-w^{-N})^{\beta}

940: \phantom{a}_{2}F_{1} \left (

941: \begin{array}{c}\beta , \,\, \beta \! +\! \frac{2}{N} \\

942: 1+\frac{2}{N} \end{array}; w^{-N} \right ).

943: \eeq

944: The solution $f(w)$ (growing at infinity as $w$)

945: is our conformal map (\ref{confmap}).

946: The meaning of the second solution (tending to $0$

947: at infinity) is clarified below.

948: 

949: It is useful to present the integral and differential equations in

950: another form which may be more suggestive for some purposes. Set

951: $w=e^{2x/N}=e^{\kappa x}$ and $f(e^{\kappa x})=\psi (x)$. In this

952: notation, the equations are:

953: \begin{itemize}

954: \item

955: The integral equation:

956: \beq\label{integral2}

957: \psi (x)=e^{\kappa x}+\frac{2\sin \, \pi \beta}{\pi}

958: \int_{0}^{\infty} \frac{e^{\kappa (x+x')}

959: \psi (x')dx'}{1-e^{2(x+x')}},

960: \eeq

961: \item

962: The differential equation:

963: \beq\label{differ2}

964: \psi '' (x) +\frac{\beta (1-\beta )}{\mbox{sinh}^2 x}

965: \, \psi (x) =\kappa ^2 \psi (x).

966: \eeq

967: \end{itemize}

968: The latter is the Schroedinger equation in the

969: singular potential $V(x)=-\,

970: \frac{\beta (1-\beta )}{\mbox{sinh}^2 x}$. We need

971: solutions on the positive real axis such that $\psi (0)=0$.

972: The equation has two linearly independent solutions

973: $\psi_{\pm \kappa}(x)$, where

974: $$

975: \psi_{\kappa}(x)=e^{\kappa x}

976: (1-e^{-2x})^{\beta}

977: \phantom{a}_{2}F_{1} \left (

978: \begin{array}{c}\beta , \,\, \beta \! -\! \kappa \\

979: 1-\kappa \end{array}; e^{-2x} \right ).

980: $$

981: The solution that grows at infinity, $\psi_{+\kappa}(x)$,

982: corresponds to the conformal map.

983: 

984: In order to show that $\psi_{\kappa}$ obeys both

985: integral and differential equations, we rewrite them

986: as the eigenvalue equations

987: $\hat I \psi = \frac{\pi}{\sin \, \pi \beta}\psi$,

988: $\hat D \psi = \beta (\beta -1)\psi$ for the operators

989: $$

990: \hat I \psi (x)=\int_{0}^{\infty}

991: \frac{e^{\kappa (x+x')}\, \mbox{sinh}\, x}{\mbox{sinh}\,

992: (x+x')\, \mbox{sinh}\, x'}\, \psi(x')\, dx' ,

993: $$

994: $$

995: \hat D \psi (x)=\mbox{sinh}^2 x (\p_{x}^{2}-\kappa^2)

996: \psi (x).

997: $$

998: A direct calculation shows that the operators $\hat I$, $\hat D$

999: commute on the space of functions vanishing at the origin, and thus

1000: have a family of common eigenfunctions. The correspondence between

1001: the eigenvalues is established by their local behavior around the

1002: origin.

1003: 

1004: A change of the variables leads to another form

1005: of the differential equation, where the parameters

1006: $\kappa$ and $\frac{1}{2}-\beta$ get interchanged.

1007: Namely, introduce the new variable $\xi$ such that

1008: \beq\label{xi}

1009: \mbox{cosh}\, \xi = \mbox{coth}\, x

1010: \eeq

1011: (equivalent forms: $\mbox{sinh}\, \xi =-(\mbox{sinh}\, x)^{-1}$,

1012: $\xi = \log \, \mbox{tanh}\, (x/2)$) and set

1013: $$

1014: \tilde \psi (\xi )=(\mbox{sinh} \, \xi)^{1/2}

1015: \psi (x(\xi )).

1016: $$

1017: Then eq. (\ref{differ2}) becomes

1018: \beq\label{differ3}

1019: \tilde \psi '' -

1020: \frac{\kappa^2 -\frac{1}{4}}{\mbox{sinh}^2 \xi}

1021: \, \tilde \psi

1022: =\left (\beta -\frac{1}{2}\right )^2 \tilde \psi.

1023: \eeq

1024: 

1025: \subsection{Analytic continuation of the conformal map}

1026: 

1027: An important information is encoded in the analytic

1028: continuation of the function $f(w)$ to the interior

1029: of the unit circle. It can be achieved by means of the

1030: identity

1031: $$

1032: \phantom{a}_{2}F_{1} \left (

1033: \begin{array}{c}a , \, b \\

1034: c \end{array}; z \right )

1035: =\frac{\Gamma (c)\Gamma (b-a)}{\Gamma (b)\Gamma (c-a)}

1036: \, (-z)^{-a}

1037: \phantom{a}_{2}F_{1} \left (

1038: \begin{array}{c}a , \,\, a-c+1 \\

1039: a-b+1 \end{array}; z^{-1} \right )

1040: $$

1041: $$

1042: \phantom{aaaaa}

1043: +\,\,\, \frac{\Gamma (c)\Gamma (a-b)}{\Gamma (a)\Gamma (c-b)}

1044: \, (-z)^{-b}

1045: \phantom{a}_{2}F_{1} \left (

1046: \begin{array}{c}b , \,\, b-c+1 \\

1047: b-a+1 \end{array}; z^{-1} \right ).

1048: $$

1049: In our case this identity gives:

1050: $$

1051: f(w)\, =\,

1052: A w(1-w^{-N})^{\beta} (-w^{-N})^{-\beta}

1053: \phantom{a}_{2}F_{1} \left (

1054: \begin{array}{c}\beta , \,\, \beta +\frac{2}{N} \\

1055: 1+\frac{2}{N} \end{array}; w^N \right )

1056: $$

1057: $$

1058: \phantom{aaaaa}

1059: +\,\, \frac{\sin \, \pi \beta}{\sin \frac{2\pi}{N}}

1060: \,  w(1-w^{-N})^{\beta} (-w^{-N})^{-\beta +\frac{2}{N}}

1061: \phantom{a}_{2}F_{1} \left (

1062: \begin{array}{c}\beta , \,\, \beta -\frac{2}{N} \\

1063: 1-\frac{2}{N} \end{array}; w^N \right ),

1064: $$

1065: where \beq\label{A1} A=\frac{\Gamma (1-\kappa )\Gamma (-\kappa

1066: )}{\Gamma (\beta - \kappa ) \Gamma (1-\beta -\kappa )}\,, \quad

1067: \kappa = 2/N . \eeq Taking appropriate branches of the multi-valued

1068: functions, we have

1069: $$

1070: (1-w^{-N})^{\beta} (-w^{-N})^{-\beta}=

1071: (1-w^N)^{\beta},

1072: $$

1073: $$

1074: (-w^{-N})^{2/N}= \omega^{2k-1}w^{-2} \quad \mbox{if

1075: $w\in \mit\Sigma_k$}.

1076: $$

1077: Therefore, we can compactly write the analytically

1078: continued function in the form

1079: \beq\label{cont}

1080: f(w)=B_k f(w^{-1})+

1081: Ag(w^{-1})\,,

1082: \quad w\in \mit\Sigma_k \,,

1083: \eeq

1084: where

1085: \beq\label{Bk}

1086: B_k =\omega^{2k-1}\,

1087: \frac{\sin \, \pi \beta}{\sin \frac{2\pi}{N}}\,,

1088: \eeq

1089: $A$ is given by (\ref{A1}),

1090: and $g(w)$ is the second solution (\ref{g}) of the

1091: differential equation (\ref{differ1}).

1092: (Note that $f(w^{-1})$ and $g(w^{-1})$ are also

1093: solutions of this equation.)

1094: 

1095: In fact (\ref{cont})

1096: is nothing else than the decomposition of the

1097: Schwarz function $S=S_{+}+S_{-}$. To see this,

1098: we rewrite (\ref{cont}) in the equivalent form

1099: $$

1100: f(w^{-1})=B_{1-k} f(w)+

1101: Ag(w)\,,

1102: \quad w\in \mit\Sigma_k \,,

1103: $$

1104: which makes it clear (\ref{eq:Scircle}) that the l.h.s. is

1105: $S(z)$ while the first term in the r.h.s. is

1106: $S_{+}(z)$ (cf. (\ref{SZ2}), (\ref{SZ5})).

1107: Therefore, the second term is just $S_{-}(z)=g(f^{-1}(z))$.

1108: This observation allows us to compute the area

1109: of the domain with conformal radius $1$ and to

1110: conclude that it is equal to $\pi A$, in agreement

1111: with the previously introduced notation (\ref{A}).

1112: Indeed, we write:

1113: $$

1114: \mbox {Area}\, (\DD ) =\frac{1}{2i}\oint_{\p \DD}

1115: S(z)dz =\frac{1}{2i}\oint_{\p \DD}

1116: S_{+}(z)dz + \frac{1}{2i}\oint_{\p \DD}

1117: S_{-}(z)dz\,.

1118: $$

1119: The first integral vanishes since $S_+$ is analytic

1120: inside $\DD$ while the second one is

1121: $$

1122: \frac{A}{2i}\oint g(w) df(w) =\pi A

1123: $$

1124: (by taking the residue at infinity).

1125: 

1126: 

1127: 

1128: 

1129: \section{The 2D Coulomb gas representation

1130: for the self-similar solutions}\label{sec:Coulomb}

1131: 

1132: As it has been shown in \cite{MWZ,KKMWZ,TBAZW}, the LG

1133: dynamics (with zero surface tension) can be simulated by the 2D

1134: Coulomb gas in an external field in the limit ${\cal N}\to \infty$,

1135: where ${\cal N}$ is the number of charges:

1136: \beq\label{C1}

1137: Z_{{\cal N}} (t) = \int \prod_{i<j}^{{\cal N}} |z_i - z_j |^2

1138: \prod_{l=1}^{{\cal N}} e^{\frac{{\cal N}}{t}W(z_l , \bar z_l )} d^2

1139: z_l \,,

1140: \eeq

1141: where the potential $W$ is determined by the ``positive"

1142: part of the Schwarz function:

1143: \beq\label{C2} W(z, \bar z)=-|z|^2 + 2

1144: {\cal R}e \, \int_{0}^{z}S_{+}(z')\, dz' \,.

1145: \eeq

1146: Specifically, it can

1147: be shown that in the limit ${\cal N}\to \infty$ the charges

1148: uniformly fill a domain $\DD (t)$ in the plane and the shape of this

1149: domain depends on $t$ according to the Laplacian growth equation.

1150: For the self-similar $\ZZ _{N}$-symmetric solutions a simple

1151: calculation gives

1152: \beq\label{C3} W(z, \bar z)=-|z|^2 + \frac{\sin \,

1153: \pi \beta}{2\sin \, \frac{2\pi}{N}} \left ( e^{-2\pi i(2k-1)/N}z^2 +

1154: e^{2\pi i (2k-1)/N}\bar z^2 \right ) \, , \quad z\in \mit\Sigma_k\,.

1155: \eeq

1156: This potential is continuous across the boundary between the

1157: sectors and enjoys the $\ZZ _{N}$-symmetry.

1158: 

1159: The free energy of the Coulomb gas (\ref{C1}), normalized

1160: in a special way,

1161: \beq\label{C4}

1162: {\cal F}(t)=t^2  \lim_{{\cal N}\to \infty}

1163: \frac{\log Z_{{\cal N}}}{{\cal N}^2}\,,

1164: \eeq

1165: plays an important role. (As is shown in \cite{KKMWZ},

1166: it is related to tau-function of certain integrable hierarchies.)

1167: To find it for the self-similar solutions, we use

1168: the integral representation

1169: \beq\label{C5}

1170: 2\pi {\cal F}(t)=\int_{\DD (t)} W(z, \bar z)d^2 z +

1171: t \int_{\DD (t)}\log |z|^2 \, d^2 z

1172: \eeq

1173: and the relation

1174: \beq\label{C6}

1175: \frac{\p^2 {\cal F}(t)}{\p \, t^2} = 2\log \, r(t)

1176: \eeq

1177: ($r(t)$ is the external conformal radius

1178: of the domain $\DD (t)$) obtained in \cite{KKMWZ}.

1179: Since the potential (\ref{C3}) is homogeneous,

1180: the self-similar ansatz plugged into (\ref{C5})

1181: yields

1182: $$

1183: 2\pi {\cal F}(t)=r^4 (t) \int_{\DD (A)} W\, d^2z

1184: +Ar^4(t)\int_{\DD (A)}\log (r^2 (t)|z|^2 ) d^2 z\,,

1185: $$

1186: where the domain $\DD (A)$ has conformal radius $1$

1187: and $r^2 (t)=t/A$. Therefore, we can write

1188: $$

1189: {\cal F}(t)\, =

1190: \, r^4(t){\cal F}(A)+\frac{A^2}{2} r^4(t)\log r^2(t) \, =\,

1191: \frac{t^2}{A^2}{\cal F}(A) +\frac{1}{2} t^2 \log \frac{t}{A}

1192: $$

1193: and calculate the second $t$-derivative:

1194: $$

1195: \frac{\p^2 {\cal F}}{\p \, t^2} =

1196: \frac{2}{A^2}{\cal F}(A) + \log \frac{t}{A} +\frac{3}{2}\,.

1197: $$

1198: The r.h.s. must be equal to $2\log r(t) = \log(t/A)$, hence

1199: ${\cal F}(A)=-\frac{3}{4}A^2$ and eventually we obtain

1200: \beq\label{C7}

1201: {\cal F}(t)=\frac{1}{2}t^2 \log \frac{t}{A} \, -\, \frac{3}{4}\, t^2\,,

1202: \eeq

1203: where $A$ is given in (\ref{A1}).

1204: 

1205: \section{Special and limiting cases}\label{sec:SLC}

1206: 

1207: The trivial special cases are $\beta =0$ (no fjords,

1208: the domain is the disk, $f(w)=w$) and

1209: $\beta =2/N$ (the domain is a collection of $N$

1210: symmetric needles growing from $0$, $f(w)=w(1-w^{-N})^{2/N})$.

1211: A non-trivial special case, where the solution is available

1212: in elementary functions is $N=4$ (wedge with the right

1213: angle). At $N=4$, the conformal map is \cite{TRHC}

1214: $$

1215: f(w)=\frac{w}{2}\left [

1216: (1-w^{-2})^{\beta}

1217: (1+w^{-2})^{1-\beta} +

1218: (1+w^{-2})^{\beta}

1219: (1-w^{-2})^{1-\beta}

1220: \right ]

1221: $$

1222: and $A=(1-2\beta )\cos  \pi \beta$.

1223: A direct substitution $N=1$ leads to a singularity. To resolve it,

1224: we set $N=1-\epsilon$, $\beta =4\epsilon t_2$ and tend $\epsilon \to

1225: 0$. The general formulas then give: $S_{+}(z)=2t_2 z$, $f(w)=w

1226: +2t_2/w$, $A=1-4t_{2}^{2}$, i.e., the domain is an ellipse with the

1227: second harmonic moment $t_2$ (in the normalization used in

1228: \cite{MWZ}). The case $N=2$ gives, in a similar way, the same family

1229: of ellipses (restricted to the half-plane). See also Appendix A.

1230: 

1231: The most interesting limit is $N\to \infty$. In this case

1232: we expect the solutions to reproduce, after a proper

1233: rescaling, the Saffman-Taylor

1234: fingers in the channel geometry.

1235: In the rest of this section we show

1236: that this is indeed the case.

1237: 

1238: First of all, let us change $w^N \to w$ and consider the

1239: conformal map

1240: $$

1241: f_1 (w)=w^{\kappa /2} (1-w^{-1})^{\beta}

1242: \phantom{a}_{2}F_{1} \left (

1243: \begin{array}{c}\beta , \,\, \beta -\kappa \\

1244: 1-\kappa \end{array}; w^{-1} \right )

1245: $$

1246: in the limit $\kappa = 2/N \to 0$ and $\beta \to 0$ so that

1247: $\beta/\kappa$ remains a finite parameter.

1248: We set

1249: $$

1250: \beta = \kappa (1-\lambda )\,,

1251: $$

1252: so that $\pi \lambda \kappa$ is the angle of the

1253: finger (tending to zero in the limit).

1254: Second, passing to the channel of a finite

1255: width (say $2\pi$) requires the rescaling

1256: of the physical plane $z \to Nz$. At last,

1257: in order to be closer to the finger tip

1258: one should shift the origin. Taking this into account,

1259: it is not difficult to see that the correct

1260: limit to the channel geometry is

1261: \beq\label{ch1}

1262: f_{{\rm ch}} (w)= \lim_{N\to \infty}

1263:  \left [ Nf_1 (w) -N\right ]

1264: \eeq Performing this limit with the $f_1(w)$ given above, one

1265: obtains \beq\label{ch2} f_{{\rm ch}} (w) = \log w +2(1-\lambda )

1266: \log (1-w^{-1}) \eeq which is the standard form of the

1267: Saffman-Taylor finger with the relative width $\lambda$. Note that

1268: the hypergeometric function behaves as $1+ O(\kappa^2)$ and so does

1269: not contribute to $f_{{\rm ch}}(w)$.

1270: 

1271: It is also instructive to derive the limiting form

1272: of the time-dependent conformal map and to follow in detail

1273: how self-similarity in the wedge

1274: turns into translational invariance in the channel.

1275: At nonzero $\kappa$ the time evolution is just a similarity

1276: transformation, so we can write, taking into account the shift

1277: of the origin:

1278: $$

1279: \tilde f_{{\rm ch}}(w,t)=\lim _{N \to \infty}

1280: \left (\frac{r(T_0)+t}{r(T_0)}\, N\, f_1 (w)\, -

1281: \, N \right )

1282: $$

1283: where $T_0 =AN^2$, i.e., $r(T_0)=N$.

1284: Expanding

1285: $$

1286: \frac{r(T_0)+t}{r(T_0)} =

1287: \left (1+\frac{t}{T_0}\right )^{1/2}=1+ \frac{t}{2T_0} +\ldots

1288: $$

1289: we get

1290: $$

1291: \tilde f_{{\rm ch}}(w,t)=f_{{\rm ch}}(w)+

1292: \frac{tN}{2T_0}\,.

1293: $$

1294: It remains to rescale time as

1295: $t=2NT$, where $T$ is kept finite

1296: (note that still $t\ll T_0$),

1297: then

1298: the last term becomes $T/A$. The limiting

1299: value of $A$ is $1-\beta \kappa^{-1} = \lambda$,

1300: so we obtain the time-dependent Saffman-Taylor

1301: finger in the standard form

1302: $$

1303: f_{{\rm ch}}(w,T)=\tilde f_{{\rm ch}}(w,t)=

1304: \frac{T}{\lambda}+

1305: \log w +2(1-\lambda ) \log (1-w^{-1})\,.

1306: $$

1307: 

1308: \acknowledgments

1309: 

1310: Ar.A. was supported by the Oppenheimer fellowship at LANL.

1311: M.M. gratefully acknoweldges generous support

1312: of the LDRD project 20030037DR at LANL during

1313: completion of this work.

1314: The work of A.Z. was supported in part by

1315: RFBR grant 06-02-17383 and by grant NSh-8004.2006.2

1316: for support of scientific schools.

1317: 

1318: 

1319: \appendix

1320: 

1321: \section{Shapes of the domains}\label{sec:forms}

1322: 

1323: In this Appendix we discuss the

1324: range of parameters of the conformal map (\ref{confmap})

1325: and the corresponding

1326: shapes of the air domains. It is convenient to adopt

1327: the function (\ref{confmap}) to the case of a single sector in the physical

1328: plane:

1329: \begin{equation}\label{eq:confmapSingle}

1330: f(w)=w^{\kappa /2}(1-1/w)^{\beta}

1331: \phantom{a}_{2}F_{1} \left (

1332: \begin{array}{c}\beta , \,\, \beta \! -\! \kappa \\

1333: 1-\kappa \end{array}; 1/w \right ).

1334: \end{equation}

1335: (we remind that in our notation the wedge angle

1336: is $\Theta_{w}=\pi \kappa $ and the fjord angle

1337: is $\Theta_{fjord}=\beta \pi $).

1338: 

1339: First, we notice that at $\beta =0$ the hypergeometric function is

1340: identically $1$, so the conformal map is just $f(w)=w^{\kappa /2}$ --

1341: the map from the exterior of the unit circle to the

1342: exterior of the sector.

1343: 

1344: Further investigation should be done separately in the different regions:

1345: i) $0<\kappa <1/2$, ii) $1/2<\kappa <1$, iii) $1<\kappa <3/2$,

1346: and iv) $3/2<\kappa <2$. Two special points $\kappa =1$ and $\kappa =2$

1347: require separate considerations.

1348: 

1349: In the region i) the allowed values of $\beta $ are $0<\beta <\kappa $.

1350: At $\beta =\kappa $ the hypergeometric function is again identically $1$.

1351: The map then is $f(w)=(w^{1/2}-w^{-1/2})^{\kappa }$ and

1352: the air domain is a needle. The conformal map has a critical point

1353: (a point where $f'(w)=0$) $w=-1$ right on the unit circle.

1354: If we formally consider

1355: the case $\beta >\kappa $ then we see that the critical point

1356: is outside the

1357: unit circle. The map (\ref{eq:confmapSingle}) is then no longer conformal.

1358: 

1359: In the region ii) the allowed values of $\beta $ are $0<\beta <1-\kappa $.

1360: At $\beta =1-\kappa $ we use the symmetry

1361: $\beta \rightarrow 1-\beta $ to see

1362: that the map is again given by

1363: $f(w)=(w^{1/2}-w^{-1/2})^{\kappa }$ and

1364: the air domain is again a needle. The

1365: critical point is at $w=-1$. If we increase the value of $\beta$, then the

1366: critical point moves outside the unit circle and the map

1367: is no longer conformal.

1368: 

1369: 

1370: In the region iii) the allowed values of $\beta $ are $0<\beta <\kappa-1$.

1371: At $\beta =\kappa-1$ the hypergeometric function is just $1+1/w$.

1372: The function (\ref{eq:confmapSingle}) is then

1373: $f(w)=(w^{1/2}-w^{-1/2})^{\kappa -1}(w^{1/2}+w^{-1/2})$. This gives

1374: a needle-like domain but the needle is

1375: ``two-sided" and rotated by $\pi /2$ (see the figure).

1376: The map has two

1377: critical points on the unit circle at the

1378: points $w=\exp (\pm i\phi_{0})$, where

1379: $\tan \phi_{0}=\sqrt{\kappa -1}$. This corresponds to the two tips of the

1380: needle. If we increase the value of $\beta$, then both

1381: critical points move outside the unit circle and the map

1382: is no longer conformal.

1383: 

1384: 

1385: 

1386: In the region iv) the allowed values of $\beta $ are $0<\beta <2-\kappa$.

1387: At $\beta =2-\kappa $ we use the symmetry

1388: $\beta \rightarrow 1-\beta $ and again see that the

1389: situation is identical to the one in region iii).

1390: There are two critical points right on the unit

1391: circle. Both of them go outside should we increase the

1392: value of $\beta $ even further.

1393: 

1394: 

1395: All the four cases are illustrated in the figure.

1396: 

1397: 

1398: 

1399: \begin{tabular}{cccc}

1400: \begin{minipage}{0.2\columnwidth}

1401: \includegraphics[width=\columnwidth]{figures/kappa1}

1402: \end{minipage}

1403: &

1404: \begin{minipage}{0.22\columnwidth}

1405: \includegraphics[width=\columnwidth]{figures/kappa2}

1406: \end{minipage}

1407: &

1408: \begin{minipage}{0.28\columnwidth}

1409: \includegraphics[width=\columnwidth]{figures/kappa3}

1410: \end{minipage}

1411: &

1412: \begin{minipage}{0.3\columnwidth}

1413: \includegraphics[width=\columnwidth]{figures/kappa4}

1414: \end{minipage}

1415: \end{tabular}

1416: \vspace{.1in}

1417: 

1418: The special value $\kappa =1$ requires careful consideration. We set

1419: $\kappa =1-\delta $ and take the limit $\delta  \rightarrow 0$,

1420: $\beta \rightarrow 0$ with the arbitrary ratio $a=\beta /\delta <1$.

1421: The hypergeometric function is then $1-a/w$ and the conformal map

1422: (\ref{eq:confmapSingle}) becomes $f(w)=w^{1/2}-aw^{-1/2}$ which gives

1423: half of an ellipse with axes $1-a$ and $1+a$.

1424: 

1425: 

1426: At the special  value $\kappa =2$,  we set

1427: $\kappa =2-\delta $ and again take the limit $\delta  \rightarrow 0$,

1428: $\beta \rightarrow 0$ with an arbitrary ratio $a=\beta /\delta <1$.

1429: The hypergeometric function becomes $1-a/w^{2}$ and the conformal map

1430: $f(w)=w-a/w$ gives an ellipse with axes $1-a$ and $1+a$.

1431: 

1432: 

1433: 

1434: \begin{thebibliography}{99}

1435: 

1436: \bibitem{Howison}

1437: K.A.Gillow and S.D.Howison, {\it A bibliography

1438: of free and moving boundary problems for Hele-Shaw and Stokes

1439: flow} (1998), http://www.maths.ox.ac.uk/~howison/Hele-Shaw/

1440: 

1441: \bibitem{RMP} D.Bensimon, L.P.Kadanoff, S.Liang, B.I.Shraiman

1442: and C.Tang, Rev. Mod. Phys. {\bf 58} (1986) 977-999

1443: 

1444: \bibitem{book} B.Gustafsson and A.Vasil'ev, {\it Conformal and

1445: potential analysis in Hele-Shaw cells}, Birkh\"auser Verlag, 2006

1446: 

1447: \bibitem{ST} P.G.Saffman and G.I.Taylor, Proc. R. Soc.

1448: {\bf A 245} (1958) 312

1449: 

1450: \bibitem{TRHC} H.Thome, M.Rabaud, V.Hakim and Y.Couder,

1451: Phys. Fluids {\bf A1} (1989) 224

1452: 

1453: \bibitem{HLS-DLA} H.L.Swinney and O.Praud, Phys. Rev. E {\bf 72}

1454: (2003) 011406

1455: 

1456: \bibitem{WittenSanderPRL} T.A.Witten and L.Sander, Phys. Rev. Lett.

1457: {\bf 47} (1981) 1400

1458: 

1459: \bibitem{PhTod} T.C.Halsey, Physics Today, {\bf 53} (2000) 36-41

1460: 

1461: \bibitem{MWZ}

1462: M.Mineev-Weinstein, P.Wiegmann and A.Zabrodin,

1463: Phys. Rev. Lett. {\bf 84} (2000) 5106-5109, e-print

1464: archive: nlin.SI/0001007

1465: 

1466: \bibitem{KKMWZ} I.Kostov, I.Krichever, M.Mineev-Weinstein,

1467: P.Wiegmann and A.Zabrodin, {\it $\tau$-function for analytic

1468: curves}, in: Random Matrices and Their Applications

1469: (MSRI publications, vol. 40), ed. P.Bleher and A.Its (Cambridge:

1470: Canbridge Academic Press), 285-299,

1471: e-print archive: hep-th/0005259

1472: 

1473: \bibitem{ABWZ} O.Agam, E.Bettelheim, P.Wiegmann, and A.Zabrodin,

1474: Phys. Rev. Lett. {\bf 88} (2002) 236801, e-print archive:

1475: cond-mat/0111333

1476: 

1477: \bibitem{KMWZ} I.Krichever, M.Mineev-Weinstein, P.Wiegmann and A.Zabrodin,

1478: Physica {\bf D 198} (2004) 1-28, e-print archive: nlin.SI/0311005

1479: 

1480: \bibitem{Asympt} ``Asymptotics beyond All Orders'', series NATO ed. by Couder,

1481: Levine, and Tanveer (1991)

1482: 

1483: \bibitem{Aldushin} Aldushin, B.Matkowsky, Physics of Fluids, {\bf 11},

1484: (1999) 1287-1296

1485: 

1486: \bibitem{M98} M.Mineev-Weinstein, Phys. Rev. Lett. {\bf 80}

1487: (1998) 2113-2116, e-print archive: patt-sol/9705004

1488: 

1489: \bibitem{Benamar} M.Ben Amar, Phys. Rev. {\bf A43}

1490: (1991) 5724-5727; Phys. Rev. {\bf A44} (1991) 3673-3685

1491: 

1492: \bibitem{Tu} Y.Tu, Phys. Rev. {\bf A44} (1991) 1203-1210

1493: 

1494: \bibitem{Combescot} R.Combescot, Phys. Rev. {\bf A45}

1495: (1992) 873-884

1496: 

1497: \bibitem{Cummings} L.Cummings, Euro. J. Appl. Math. {\bf 10}

1498: (1999) 547-560

1499: 

1500: \bibitem{Richwedge} S.Richardson, Euro. J. Appl. Math. {\bf 12}

1501: (2001) 665-676

1502: 

1503: \bibitem{MV} I.Markina and A.Vasil'ev, Scientia {\bf 9} (2003)

1504: 33-43; Euro. J. Appl. Math. {\bf 15} (2004) 781-789

1505: 

1506: \bibitem{Flugge} S. Flugge ``Practical Quantum Mechanics'',

1507: Springer-Verlag, Berlin, Heidelberg, New York, 1974.

1508: 

1509: \bibitem{Davis}

1510: P.J.Davis, ``The Schwarz

1511: function and its applications", The

1512: Carus Math. Monographs, No. 17,

1513: The Math. Assotiation  of America, Buffalo,

1514: N.Y., 1974

1515: 

1516: \bibitem{Howison91} S.D.Howison, Euro. J. of Appl. Math,

1517: {\bf 3} (1992) 209-224

1518: 

1519: \bibitem{Lamb} H.Lamb ``Hydrodynamics'', Dover Publication,

1520: 1945

1521: 

1522: \bibitem{Richardson} S.Richardson, J. Fluid Mech. {\bf 56} (1972) 609-618

1523: 

1524: \bibitem{MS} M.Mineev-Weinstein and S.Ponce Dawson, Phys. Rev E,

1525: {\bf 50} (1994) R24-R27; S.Ponce Dawson and M Mineev-Weinstein,

1526: Physica D, {\bf 73} (1994) 373 - 387

1527: 

1528: \bibitem{TBAZW}

1529: R.Teodorescu, E.Bettelheim, O.Agam, A.Zabrodin and P.Wiegmann,

1530: Nucl. Phys. {\bf B704} (2005) 407-444, e-print archive:

1531: hep-th/0401165;

1532: Nucl. Phys. {\bf B700} (2004) 521-532, e-print archive: hep-th/0407017

1533: 

1534: \bibitem{GR}

1535: I.Gradshtein and I.Ryzhik,

1536: Tables of Integrals, Series and Products,

1537: Academic, NY, 1965

1538: 

1539: 

1540: 

1541: 

1542: 

1543: \end{thebibliography}

1544: 

1545: \end{document}

1546: