1: \documentclass[12pt]{article}

2: \usepackage{latexsym,amsmath,amssymb}

3: %\headheight=10mm \headsep=-10mm \oddsidemargin=-1mm \evensidemargin=-1mm

4: %\textheight=210mm \textwidth=150mm \normalbaselineskip=12pt

5: %\normalbaselines %\parskip 6pt plus 1pt

6: 

7:      \addtolength{\textwidth}{2cm} \addtolength{\hoffset}{-1cm}

8:      \addtolength{\textheight}{3cm} \addtolength{\voffset}{-1.5cm}

9: \newif\ifdviwin

10: 

11: \usepackage[latin1]{inputenc}

12: \usepackage[english]{babel}

13: \usepackage{indentfirst}

14: \usepackage[mathscr]{eucal}

15: \usepackage{amssymb,amsmath,amsfonts}

16: \usepackage{fancybox,fancyhdr}

17: \usepackage{graphicx}

18: 

19: \newif\ifdviwin

20: 

21: \dviwintrue 

22: 

23: \def\cA{\mathcal{A}}

24: \def\cP{\mathcal{P}}

25: \def\cS{\mathcal{S}}

26: \def\cC{\mathcal{C}}

27: \def\cR{\mathcal{R}}

28: \def\cB{\mathcal{B}}

29: \def\cD{\mathcal{D}}

30: \def\cL{\mathcal{L}}

31: \def\cM{\mathcal{M}}

32: \def\cG{\mathcal{G}}

33: \def\cF{\mathcal{F}}

34: \def\cN{\mathcal{N}}

35: \def\cT{\mathcal{T}}

36: \def\osm{(\Omega,\Sigma,\mu)}

37: 

38: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

39: %%%% NUEVOS COMANDOS PARA LATEX

40: \let\8=\infty \let\0=\emptyset  \let\sii=\iff

41: \let\hat=\widehat \let\dd=\displaystyle \let\por=\times \let\pertenece=\in

42: \let\paratodo=\forall \let\existe=\exists \let\contenido=\subset

43: \let\contiene=\supset \let\union=\cup \let\interseccion=\cap

44: \let\barra =\backslash

45: \let\tilde=\widetilde

46: \let\derecha=\flushright \let\endderecha=\endflushright

47: \let\izquierda=\flushleft \let\endizquierda=\endflushleft

48: \let\landa=\lambda

49: \let\alfa=\alpha

50: \let\ro=\rho

51: \let\ex=\wedge

52: \let\dobleflecha=\rightleftharpoons

53: \let\parc=\partial

54: 

55: \def\ep{\varepsilon}

56: \def\vart{\vartheta}

57: \def\landa{\lambda}

58: \def\Landa{\Lambda}

59: \def\lap{\Delta}

60: \def\flecha{\rightarrow}

61: \def\esiz{\langle}

62: \def\esde{\rangle}

63: 

64: \def\cte.{\mathop{\rm cte.}\nolimits}

65: \def\nor{\mathop{\rm nor}\nolimits}

66: \def\disc{\mathop{\rm disc (\xi)}\nolimits}

67: \def\det{\mathop{\rm det}\nolimits}

68: \def\tr{\mathop{\rm tr}\nolimits}

69: \def\diag{\mathop{\rm diag}\nolimits}

70: \def\div{\mathop{\rm div }\nolimits}

71: \def\Re{\mathop{\rm Re }\nolimits} \def\Im{\mathop{\rm Im }\nolimits}

72: \def\ran{\mathop{\rm ran }\nolimits}

73: \def\sen{\mathop{\rm sen }\nolimits} \def\arcsen{\mathop{\rm arcsen }\nolimits}

74: \def\tg{\mathop{\rm tg }\nolimits} \def\arctg{\mathop{\rm arctg }\nolimits}

75: \def\cotg{\mathop{\rm cotg }\nolimits}

76: \def\cosec{\mathop{\rm cosec }\nolimits}

77: \def\senh{\mathop{\rm senh }\nolimits}

78: \def\cosh{\mathop{\rm cosh }\nolimits}

79: \def\tanh{\mathop{\rm tanh }\nolimits}

80: \def\arcsinh{\mathop{\rm arcsinh }\nolimits}

81: \def\arccosh{\mathop{\rm arccosh }\nolimits}

82: \def\Arg{\mathop{\rm Arg }\nolimits}

83: \def\co{\mathop{\overline{\rm co}}\nolimits}

84: \def\lin{\mathop{\rm lin }\nolimits} \def\dim{\mathop{\rm dim }\nolimits}

85: \def\card{\mathop{\rm card }\nolimits} \def\ext{\mathop{\rm ext }\nolimits}

86: \def\graf{\mathop{\rm Graf }\nolimits}

87: \def\spand{\mathop{\rm span }\nolimits}

88: \def\Dem{\par\noindent{\sc Dem:}}

89: \def\almargen#1{\marginpar{\footnotesize\sf #1}}

90: \def\N{\mathbb{N}}

91: \def\L{\mathbb{L}}

92: \def\E{\mathbb{E}}

93: \def\B{\mathbb{B}}

94: \def\Q{\mathbb{Q}}

95: \def\R{\mathbb{R}}

96: \def\Z{\mathbb{Z}}

97: \def\C{\mathbb{C}}

98: \def\D{\mathbb{D}}

99: \def\CP{\mathbb{C}\mathbb{P}}

100: \def\H{\mathbb{H}}

101: \def\S{\mathbb{S}}

102: \def\X{\mathfrak{X}}

103: \def\SL{\mathbf{SL}(2,\mathbb{C})}

104: \def\He{\rm{Herm}(2)}

105: 

106: 

107: %%%%%%%%%%%%%%%%%%

108: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

109: %%%% Grupos tipo teorema

110:  \newtheorem{defi}{Definition}

111:  \newtheorem{teo}[defi]{Theorem}

112:  \newtheorem{pro}[defi]{Proposition}

113:  \newtheorem{cor}[defi]{Corollary}

114:  \newtheorem{lem}[defi]{Lemma}

115:  \newtheorem{eje}[defi]{Example}

116:  \newtheorem{remark}[defi]{Remark}

117:  \newtheorem{note}[defi]{Note}

118:  \newtheorem{obs}{Observación}

119: 

120:  \newenvironment{proof}{\rm \trivlist \item[\hskip \labelsep{\it

121:       Proof}:]}{\par\nopagebreak \hfill $\Box$ \endtrivlist}

122: 

123:  \newenvironment{proof1}{\rm \trivlist \item[\hskip \labelsep{\it

124:       Proof of Theorem \ref{main}}:]}{\par\nopagebreak \hfill $\Box$ \endtrivlist}

125: 

126:  \newenvironment{proof2}{\rm \trivlist \item[\hskip \labelsep{\it

127:       Proof of Theorem \ref{truco}}:]}{\par\nopagebreak \hfill $\Box$ \endtrivlist}

128: 

129: 

130: 

131: %\newenvironment{proof}[1]{\par\noindent{\sc Proof:}#1}

132: %{\hfill$\square$\par\bigskip\par} %\newenvironment{dem}[1]{}{}

133: %\newenvironment{sol}[1]{\par\noindent{\sc Solución:}

134: %#1}{\hfill$\square$\par\bigskip\par} %%%%%%%%%%%%%%%%%%%%%%%%%%%%%

135: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

136: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

137: %\pagestyle{myheadings}

138: 

139: \numberwithin{equation}{section} 

140: 

141: \begin{document}

142: \mbox{}\vspace{0.4cm}\mbox{} 

143: 

144: \begin{center}

145: \rule{15.2cm}{1.5pt}\vspace{0.5cm} 

146: 

147: {\Large \bf The space of solutions to the Hessian one equation 

148: \\[0.3cm] in the finitely punctured plane}\\ \vspace{0.5cm} {\large José A.

149: Gálvez$\mbox{}^a$, Antonio Martínez$\mbox{}^b$ and Pablo 

150: Mira$\mbox{}^c$}\\ \vspace{0.3cm} \rule{15.2cm}{1.5pt} 

151: \end{center}

152:   \vspace{1cm}

153: $\mbox{}^a$, $\mbox{}^b$ Departamento de Geometría y Topología, Universidad de Granada,

154: E-18071 Granada, Spain. \\ e-mail: jagalvez@ugr.es ; amartine@ugr.es 

155: \vspace{0.2cm} 

156: 

157: \noindent $\mbox{}^c$ Departamento de Matemática Aplicada y Estadística, 

158: Universidad Politécnica de Cartagena, E-30203 Cartagena, Murcia, Spain. \\ 

159: e-mail: pablo.mira@upct.es \vspace{0.2cm} 

160: 

161: \noindent Date: September 10, 2004 \\ Keywords: Monge-Ampère equation, 

162: parabolic affine spheres, Jörgens theorem 

163: 

164: \vspace{0.3cm} 

165: 

166:  \begin{abstract}

167: We construct the space of solutions to the elliptic Monge-Ampère equation 

168: ${\rm det} (D^2 \phi)=1$ in the plane $\R^2$ with $n$ points removed.

169: We show that, modulo equiaffine transformations and for $n>1$, this space 

170: can be seen as an open subset of $\R^{3n-4}$, where the coordinates are 

171: described by the conformal equivalence classes of once punctured bounded 

172: domains in $\C$ of connectivity $n-1$. This approach actually provides a 

173: constructive procedure that recovers all such solutions to the 

174: Monge-Ampère equation, and generalizes a theorem by K. Jörgens. 

175:  \end{abstract}

176: 

177: %%\vspace{0.5cm} 

178: %%\begin{center}

179: %%{\bf Résumé} 

180: %%\end{center}

181: %%\begin{quotation}

182: %%{\small Nous construisons l'espace des solutions de l'équation elliptique 

183: %%de Monge-Ampère ${\rm det} (D^2 \phi)=1$ définies dans le plan $\R^2$ sauf 

184: %%en $n$ points. Nous montrons que, pour $n >1$ et modulo les tranformations 

185: %%équiaffines, cet espace peut être identifié à un ouvert de $\R^{3n-4}$, 

186: %%dont les coordonnées de l'espace sont décrites par des classes 

187: %%d'équivalences conformes de domaines bornés de $\C$ perforés en un seul 

188: %%point et de connectivité $n-1$. Ce point de vue permet de construire et de 

189: %%retrouver toutes ces solutions de l'équation de Monge-Ampère et généralise 

190: %%un théorème de K. Jörgens.} 

191: %%\end{quotation}

192: 

193: %\vspace{0.3cm}

194: 

195: 

196: \section{Introduction}

197: A celebrated result by K. Jörgens \cite{Jor1} states that all solutions to 

198: the elliptic Monge-Ampère equation 

199: \begin{equation}\label{monge}

200: \phi_{xx} \phi_{yy} - \phi_{xy}^2 =1 

201: \end{equation}

202: which are globally defined on $\R^2$ are quadratic polynomials. This 

203: theorem has motivated a large amount of works, which essentially follow 

204: two research lines. One of them deals with the extension of the result to 

205: more general classes of Monge-Ampère equations (see for instance 

206: \cite{BCGJ,Cal,CaLi2,ChYa,ChWa,Pog,TrWa1,TrWa2}). The other one was 

207: initiated by K. Jörgens in 1955, and concerns the validity of the theorem 

208: in large proper domains of $\R^2$, and the study of the new solutions that 

209: may arise. Some results of this type can be found in 

210: \cite{Jor2,CaLi1,FMM1,FMM2,TrWa1}. In particular, in \cite{Jor2} all the 

211: solutions to \eqref{monge} globally defined in $\R^2 \setminus \{(0,0)\} $ 

212: were obtained. 

213: 

214: In the present work we follow this last direction, and plan to determine 

215: all the solutions to \eqref{monge} in $\R^2$ with $n$ points removed. In 

216: this way, our result generalizes the two mentioned theorems by Jörgens, 

217: and suggests a new line of inquiry into the theory of Monge-Ampère 

218: equations. 

219: 

220: For stating such a general description, some remarks have to be made. 

221: First, note that if $\phi$ is a solution to \eqref{monge} in a finitely 

222: punctured plane, the punctures correspond then to isolated singularities 

223: of $\phi$. We shall assume without loss of generality that isolated 

224: singularities are \emph{non-removable}, i.e. $\phi$ cannot be 

225: $\mathcal{C}^1$-extended across the singularity. A well-known criterion to determine

226: whether an isolated singularity of a solution to \eqref{monge} is 

227: removable is found in \cite{Jor2}. We shall provide a different one in 

228: Section 3, in terms of the underlying conformal structure of the solution 

229: $\phi$.

230: 

231: Another remark concerns non-unicity of the solutions to \eqref{monge}. Let 

232: $\mathcal{E} $ be the group of \emph{equiaffine transformations} of $\R^3$

233: whose differentials fix the $x_3$-direction, i.e. the set of maps $\Phi 

234: :\R^3 \flecha \R^3$ of the form 

235:  \begin{equation}\label{equi}

236:  \Phi \left(\def\arraystretch{1.2}\begin{array}{c} x_{1} \\

237: x_{2}\\ x_{3} 

238: \end{array} \right) =

239: \left(\def\arraystretch{1.2}\begin{array}{ccc} a_{11} & a_{12} & 0 \\ 

240: a_{21} & a_{22} & 0 \\ a_{31} & a_{32} & 1 

241: \end{array} \right) \left(\def\arraystretch{1.2}\begin{array}{c} x_{1} \\

242: x_{2}\\ x_{3} 

243: \end{array} \right) + \left(\def\arraystretch{1.2}\begin{array}{c} b_{1} \\

244: b_{2}\\ b_{3} 

245: \end{array} \right), \hspace{0.5cm} a_{11} a_{22} - a_{21} a_{12} =1.

246:  \end{equation}

247: Then, if $\phi(x,y)$ solves \eqref{monge} and $\Phi \in \mathcal{E}$, the 

248: function $\phi' (x',y')$ given by $\Phi (x,y,\phi (x,y))= 

249: (x',y',\phi'(x',y'))$ is a new solution to \eqref{monge} in terms of the 

250: variables $x'$, $y'$. Therefore, we shall be interested in obtaining the 

251: classification result for solutions to \eqref{monge} \emph{modulo 

252: equiaffine transformations}. In other words, two solutions to 

253: \eqref{monge} differing only by an element of $\mathcal{E}$ will be 

254: considered to be equivalent. 

255: 

256: Keeping these facts in mind, the classification by Jörgens of the 

257: solutions to \eqref{monge} in the once punctured plane is formulated as 

258: follows: \emph{the unique solutions to \eqref{monge} in 

259: $\R^2$ and in $\R^2 \setminus \{(0,0)\}$ are respectively, up to equiaffine transformations, the

260: quadratic polynomial $\phi(x,y)= \left(x^2 +y^2 \right)/2$ and the 

261: rotational example} 

262: $$\phi(x,y)= \frac{1}{2}\left(\sqrt{x^2 + y^2} \sqrt{1 + x^2 + y^2} + \sinh^{-1} \left(\sqrt{x^2 + y^2} \right)

263: \right).$$ So, the following question arises naturally: given $n>1$, is 

264: there a unique (up to equiaffine transformations) solution to 

265: \eqref{monge} defined in an $n$-times punctured plane, and so that the 

266: punctures are non-removable isolated singularities? In Section 3 we shall 

267: analyze to what extent the answer to this question is \emph{yes}. 

268: 

269: The paper is organized as follows. In Section 2 we review the theory of 

270: solutions to \eqref{monge}, and its equivalence to the theory of locally 

271: convex embedded parabolic affine spheres in the affine $3$-space 

272: $\R^3$. In particular, we shall recall a conformal representation

273: available for these surfaces, and indicate some basic properties of 

274: parabolic affine spheres in terms of the holomorphic data. The main result 

275: in this Section is the construction of solutions to \eqref{monge} in a 

276: punctured plane $\R^2 \setminus \{p_1, \dots, p_n\}$ so that the $p_j$'s 

277: are non-removable isolated singularities of such solutions. The 

278: construction relies on the conformal representation of parabolic affine 

279: spheres, and the existence of two classical holomorphic functions with the 

280: fundamental property of mapping bijectively a once punctured bounded 

281: domain in $\C$ with $n$ boundary components onto 

282: $\C$ with $n$ vertical (resp. horizontal) slits.

283: 

284: Section 3 contains the main theorem of this work, which characterizes the 

285: examples constructed in Section 2 as the only solutions (up to equiaffine 

286: transformations) to \eqref{monge} that are globally defined over a 

287: finitely punctured plane. This theorem provides an explicit one-to-one 

288: correspondence between the conformal equivalence classes of $\C$ with $n$ 

289: disks removed, and the space of solutions to \eqref{monge} in a finitely 

290: punctured plane with exactly $n$ non-removable isolated singularities, 

291: modulo equiaffine transformations. Particularly, for $n>1$, this quotient 

292: space can be naturally identified with an open subset of $\R^{3n-4}$. 

293: 

294: Finally, in Section 4 we show that our main theorem for the particular 

295: choice $n=1$ recovers the Jörgens one, and we determine explicitly in 

296: terms of theta functions all the solutions to \eqref{monge} in the twice 

297: punctured plane. This indicates that all the solutions we have constructed 

298: are highly non-trivial. 

299: 

300: \section{Construction of the canonical examples}

301: 

302: Let $\phi(x,y):D\subseteq \R^2\flecha \R$ be a solution to \eqref{monge} 

303: on a planar domain $D$. We shall assume without loss of generality that 

304: $\phi_{xx}$ is always positive. Then the subset

305: $\mathcal{S}_{\phi} =\{(x,y,\phi(x,y)): (x,y)\in D\} \subset \R^3$

306: describes an embedded locally convex \emph{parabolic affine sphere} in the 

307: affine $3$-space $\R^3$ with affine normal vector $\xi =(0,0,1)$ and 

308: affine metric $ds^2 = \phi_{xx}\, dx^2 + 2 \phi_{xy}\, dx dy + \phi_{yy}\, 

309: dy^2$ (see for instance \cite{FMM1,FMM2} for more information). 

310: Conversely, any embedded locally convex parabolic affine sphere with 

311: affine normal vector $\xi=(0,0,1)$ is locally the graph over a domain in 

312: the $x,y$-plane of a solution to \eqref{monge}. Along this paper, by a 

313: parabolic affine sphere we will always mean a locally convex one. 

314: 

315: The affine metric $ds^2$ of $\mathcal{S}_{\phi}$ induces a Riemann surface 

316: structure on $\mathcal{S}_{\phi}$. We shall call this Riemann surface 

317: structure the \emph{underlying conformal structure} of 

318: $\phi(x,y)$.

319: 

320: The solutions to \eqref{monge} can be explicitly recovered in terms of 

321: holomorphic data with respect to this underlying conformal structure. We 

322: expose here the global version of this fact in \cite{FMM2}, for the 

323: general case of immersed parabolic affine spheres. 

324: 

325: \begin{teo}\label{reptem}

326: Let $\Sigma$ be a simply connected Riemann surface, and let 

327: $F,G:\Sigma\flecha \C$ be two holomorphic functions for which $dG\neq 0$ and $|dF|<|dG|$ hold

328: everywhere. Then the map $\psi:\Sigma\flecha \R^3$ given by 

329: \begin{equation}\label{rep}

330: \psi = \frac{1}{2} \left( G +\overline{F}, \frac{1}{4} \left\{ |G|^2 - 

331: |F|^2 + 2\Re (G F) \right\} - \Re \int F dG \right) 

332: \end{equation}

333: is a parabolic affine sphere with affine normal vector $\xi=(0,0,1)$ and 

334: affine metric $ds^2 =\frac{1}{4} \left( |dG|^2 - |dF|^2 \right)$. 

335: 

336: Conversely, any parabolic affine sphere can be recovered in this way with 

337: respect to the conformal structure induced by its affine metric. 

338: \end{teo}

339: 

340: %%%It comes clear from this conformal representation that there are, at a

341: %%%first sight, two kinds of singularities for the solutions to

342: %%%\eqref{monge}: the \emph{branch points}, i.e. isolated points in $\Sigma$

343: %%%at which $dF=dG=0$, and the singular curves in $\Sigma$ along which $|dF|=

344: %%%|dG|$. However, branch points cannot occur in solutions to \eqref{monge}

345: %%%(this follows directly from Lemma \ref{remo} in Section 3). Thus, all the

346: %%%singularities of solutions to \eqref{monge} are characterized by the

347: %%%condition $|dF/dG|=1$.

348: 

349: In order to construct examples of solutions to \eqref{monge} in the 

350: $n$-times punctured plane, we need to recall first a

351: classical description of two special holomorphic functions, which we will 

352: denote by $p(z), q(z)$. 

353: 

354: Let $\Omega\subset \C$ be a bounded domain whose boundary is made up by 

355: $n\geq 1$ analytic Jordan curves $C_1, \dots, C_n$, where we will let $C_n$ be the exterior

356: component, and take $z_0 \in \Omega$. Let $\cG (z,z_0)$ be the Green 

357: function of 

358: $\Omega$ with respect to $z_0$, given by $\cG(z,z_0)= \mathcal{D} (z,z_0)

359: - \log |z-z_0|$, where $\cD$ is the solution to the Dirichlet problem in 

360: $\Omega$ with boundary conditions

361: $\cD (\zeta) = \log |\zeta - z_0| $, $\zeta \in C_k$. Then the map

362:  \begin{equation}\label{aru}

363:  u (z)=  \frac{\parc \cG (z,z_0)}{\parc z_0} + \frac{\parc \cG (z,z_0)}{\parc \overline{z_0}}

364:  \end{equation}

365: is harmonic in $\Omega \setminus \{z_0\}$, and vanishes on $\parc \Omega$. 

366: Now let $\omega_j$ denote the \emph{harmonic measure} of $C_j$ with 

367: respect to 

368: $\Omega$, characterized by the conditions

369:  \[ \left\{\def\arraystretch{1.2} \begin{array}{lll} \Delta \omega_j & = & 0 \hspace{0.5cm} \text{ on } \Omega, \\ \omega_j (C_j) &

370:  = & 1, \\ \omega_j (C_k) & = & 0 \hspace{0.5cm} \text{ if } k \neq j.

371:   \end{array} \right. \] We shall denote by $\alfa_{j,k}$ the period of the

372: conjugate harmonic function of $\omega_j$ along $C_k$: \[ \alfa_{j,k} = 

373: \int_{C_k} \frac{\parc \omega_j}{\parc {\bf n}} \ ds,\] where ${\bf n}$ is 

374: the unit normal to the curve $C_k$. Let finally $A_k \in \R$ denote the 

375: period of the conjugate harmonic function of $u(z)$ along $C_k$. Then, if 

376: $\landa_1, \dots, \landa_{n-1}$ are real numbers, the harmonic function

377:  \begin{equation}\label{truqui}

378:   u + \landa_1 \omega_1 +\dots + \landa_{n-1} \, \omega_{n-1} :

379: \Omega\setminus \{z_0 \} \flecha \R 

380:  \end{equation}

381: has a well defined conjugate harmonic function on $\Omega \setminus 

382: \{z_0\}$ if and only if the 

383: $\landa_i$'s solve the following linear system:

384: 

385: \[ \left\{\def\arraystretch{1.2} \begin{array}{ccc}

386: \landa_1 \, \alfa_{1,1} + \cdots + \landa_{n-1}\, \alfa_{n-1,1} & = & -A_1 

387: \\ \vdots & & \vdots \\ \landa_1 \, \alfa_{1,n-1} + \cdots + \landa_{n-1}

388: \, \alfa_{n-1,n-1} & = & -A_{n-1} 

389:   \end{array} \right. \]

390: 

391: This system is classically known to have a unique solution (see \cite{Ahl} 

392: for instance), and thus there is a holomorphic function $p(z): 

393: \Omega\setminus \{z_0\} \flecha \C$ whose real part is the above harmonic 

394: function, i.e. 

395: $\Re p(z)= u(z) + \sum_{i=1}^{n-1} \landa_i \, \omega_i (z)$. We must 

396: remark that $p(z)$ is well defined up to pure imaginary additive 

397: constants. 

398: 

399: In addition, by the construction of $u(z)$ it becomes clear that $p(z)$ 

400: has a pole of order one at $z_0$, with residue $1$. Moreover, $\Re p(z) = 

401: \landa_k \in \R$ along $C_k$ for every $k\in \{1, \dots, n\} $. 

402: 

403: Actually, $p(z)$ is characterized by these conditions: it is the unique 

404: (up to pure imaginary additive constants) holomorphic map in $\Omega 

405: \setminus \{z_0\}$ with a simple pole of residue $1$ at $z_0$ whose real 

406: part is constant along each boundary component in $\parc \Omega$. Finally, 

407: $p(z)$ is a conformal equivalence between $\Omega \setminus \{z_0\}$ and 

408: its image, which is a vertical slit domain in $\C$ \cite{Ahl}. 

409: 

410: The analogous process to the above one, but this time starting with 

411: \begin{equation}\label{ardo}

412:  v (z)= \frac{\parc \cG (z,z_0)}{\parc

413:  z_0} - \frac{\parc \cG (z,z_0)}{\parc \overline{z_0}},

414:  \end{equation}

415: produces a holomorphic function $q(z):\Omega \setminus \{z_0\} \flecha \C$ 

416: with a simple pole of residue $1$ at $z_0$, and such that its imaginary 

417: part $\Im q(z)$ is constant along each boundary curve $C_k$. Again, $q(z)$ 

418: is defined up to real additive constants, and it is the unique holomorphic 

419: map in $\Omega\setminus \{z_0\}$ satisfying these properties. Further, it 

420: maps $\Omega\setminus \{z_0\} $ bijectively onto a horizontal slit domain 

421: in $\C$. 

422: 

423: \begin{teo}\label{ejes}

424: For any $n\in \N$, there exists a regular solution $\phi(x,y)$ to 

425: \eqref{monge} globally defined in $\R^2$ minus $n$ points, and such that 

426: the punctures are non-removable isolated singularities of $\phi(x,y)$. 

427: \end{teo}

428: \begin{proof}

429: We start with a complex domain $\Omega$ as above, with a distinguished 

430: point $z_0\in \Omega$. Let then $p(z),q(z)$ be the holomorphic functions 

431: constructed following the previous procedure. Now, define 

432: \begin{equation}\label{fg}

433: G(z)= p(z) + q(z), \hspace{1cm} F(z)= p(z) -q(z). 

434: \end{equation}

435: It then follows readily that $(G + \overline{F})|_{C_k}$ is constant for 

436: all $k\in \{1,\dots, n\} $. Differentiating now this expression we infer 

437: that $|dF / dG| =1$ at every point in $\parc \Omega$. On the other hand, 

438: since 

439: $p(z),q(z)$ have simple poles of equal residues at $z_0$, we obtain that

440: $|dF /dG| (z_0)=0$. Hence, $|dF / dG |<1$ in all

441: $\Omega$. In addition, as $p,q$ are holomorphic bijections between $\Omega\setminus{\{z_0\}}$

442: and a certain slit domain in $\C$, we get that $dF,dG$ cannot vanish 

443: simultaneously. Particularly, $dG$ never vanishes by $|dF /dG|<1$. 

444: 

445: Therefore, by Theorem \ref{reptem} the holomorphic functions $F,G$ define 

446: a parabolic affine sphere $$\psi : \widetilde{\Omega \setminus \{z_0\} } 

447: \flecha \R^3,$$ where $\widetilde{\Omega \setminus \{z_0\}} $ is the 

448: conformal universal covering of $\Omega \setminus \{z_0\}$. The affine 

449: sphere $\psi$ is actually well defined on $\Omega \setminus \{z_0\}$. To 

450: see this we have to check that 

451:   \begin{equation}\label{perio}

452: \Re \int_{\Gamma} F dG =0 

453:   \end{equation}

454: along any loop $\Gamma$ in $\Omega \setminus \{z_0\}$. First, observe that 

455: as $p(z)$ and $q(z)$ map each $C_k$ into a piece of a straight line, the 

456: functions $F,G$ can be holomorphically extended across each $C_k$ by 

457: Schwarzian reflection. In this way, we may extend 

458: $\Omega \setminus \{z_0\}$ slightly so that the curves $C_k$ are interior to

459: the extended domain, and 

460: $F,G$ are holomorphic there. But in this larger domain, the curves $C_1, \dots, C_n$

461: provide a basis of its first homology group. So, we just need to check 

462: condition \eqref{perio} for $\Gamma = C_1, \dots, C_n$. To do so, we first 

463: observe that by \eqref{fg} it holds for all $k\in \{1,\dots, n\}$ 

464: $$\int_{C_k} F dG = \int_{C_k} \left( p\,  dq - q \, dp \right).$$ In addition, $$0 = \Re

465: \int_{C_k} d \left( p \bar{q} \right) = - \Re \int_{C_k} \left( p\, dq - q 

466: \, dp \right) +2 \int_{C_k} \left( \Im q \, \Im p' + \Re p \, \Re 

467: q'\right) dz.$$ As both $\Re p(z)$ and $\Im q(z)$ are constant along 

468: $C_k$, we obtain that $\Re \int_{C_k} F dG =0 $ for

469: all the $C_k$'s, as we wished. 

470: 

471: Finally, we show that $\psi=(\psi_1,\psi_2,\psi_3):\Omega \setminus 

472: \{z_0\} \flecha \R^3 

473: $ is actually an embedding. For this, observe first of all that $F$ is

474: holomorphic at $z_0$, and $G$ has a simple pole at $z_0$. Thus, 

475: $(\psi_1,\psi_2)$ is a homeomorphism from a small enough punctured disk about $z_0$

476: onto the exterior of a Jordan curve in $\R^2$. 

477: 

478: Next, observe that $\cF := (\psi_1, \psi_2):\Omega \setminus 

479: \{z_0\}\flecha \R^2$ is well defined and continuous on the once punctured 

480: topological sphere $E$ which results when each curve $C_k$ in 

481: $\overline{\Omega} \setminus \{z_0\}$ is identified with a single point.

482: This happens because by \eqref{rep} we have that $\mathcal{F} (C_j)=p_j 

483: \in \R^2$ for every $j\in \{1,\dots, n\} $ and for some points 

484: $p_1,\dots, p_n \in \R^2$.

485: 

486: By the behaviour of $\cF$ about $z_0$, there is a topological punctured 

487: disk $\mathcal{D} \subset E$ with puncture at $z_0$ such that 

488: $\cF $ maps $\cD$ bijectively onto the exterior of a closed disk $B$ in $\R^2$.

489: Thus, $\cF (E\setminus \cD)$ must fill the whole $\overline{B}$. As a 

490: consequence, $(\psi_1,\psi_2):\overline{\Omega}\setminus \{z_0\} \flecha 

491: \R^2$ is onto. 

492: 

493: Now, assume the existence of two points $z_1,z_2\in \Omega \setminus 

494: \{z_0\}$ with $\cF (z_1)=\cF(z_2) =a\in \R^2$, and let $C$ denote a 

495: divergent curve in $\R^2\setminus \{p_1,\dots, p_n\}$ with endpoint $a$ 

496: (that may belong to $\{p_1,\dots, p_n\}$). As $\cF$ is a local 

497: homeomorphism, there exist two paths $\gamma_1, \gamma_2$ in 

498: $\Omega$ with endpoints $z_1,z_0$ and $z_2,z_0$ respectively, such that

499: $\cF (\gamma_i)=C$. But $\cF$ is a homeomorphism around $z_0$, and this

500: implies that $\gamma_1, \gamma_2$ must coincide from a \emph{first point} 

501: to the endpoint $z_0$. This contradicts that $\cF$ is a local 

502: homeomorphism, unless $z_1=z_2$. Hence, 

503: $\psi :\Omega \setminus \{z_0\} \flecha \R^3$ is one-to-one.

504: 

505: Finally, we need to check that $\cF (\Omega\setminus \{z_0\}) \subseteq 

506: \R^2 \setminus \{p_1,\dots, p_n\} $. If $\cF (\widetilde{z})= p_j$ for 

507: some 

508: $\widetilde{z}\in \Omega$, we may take disjoints neighbourhoods $A_1,A_2$ of $\widetilde{z}$ and $C_j$

509: in $\overline{\Omega} \setminus \{z_0\}$, respectively. There exist then 

510: points $\zeta_i\in A_i$ with $\cF(\zeta_1)=\cF(\zeta_2) \in \R^2 \setminus 

511: \{p_1, \dots, p_n\} $. Thus both points lie in $\Omega$, which is 

512: impossible. 

513: 

514: Summing up, we have an embedded parabolic affine sphere $\mathcal{S}=\psi 

515: (\Omega \setminus \{z_0\}) \subset \R^3$ such that $(\psi_1, \psi_2 

516: ):\Omega \setminus \{z_0\} \flecha \R^2 \setminus \{p_1, \dots, p_n\} $ is 

517: a global homeomorphism. This produces a regular solution to \eqref{monge} 

518: globally defined in $\R^2 \setminus \{p_1, \dots, p_n\} $ and with 

519: isolated singularities exactly at the points $p_1, \dots, p_n$. All these 

520: isolated singularities are non-removable by Lemma \ref{remo}. 

521: 

522: \end{proof}

523: 

524: \section{The classification theorem}

525: 

526: This Section is devoted to show that any solution to \eqref{monge} in a 

527: finitely punctured plane must be equiaffinely equivalent to one of the 

528: examples of Theorem \ref{ejes}. In order to do so, it is convenient to 

529: introduce the following terminology. By definition, a \emph{global 

530: solution} to \eqref{monge} in $\R^2 \setminus \{p_1,\dots, p_n\} $ will be 

531: a $\mathcal{C}^2$ function $\phi (x,y): \R^2 \setminus \{p_1, \dots, p_n\} 

532: \flecha \R$ verifying \eqref{monge} and that is not $\mathcal{C}^1$ at the 

533: $p_j$'s. We remark (see \cite{Jor2}) that if a solution to \eqref{monge}

534: extends as a $\mathcal{C}^1$ function across an isolated singularity, then 

535: it actually extends analytically across it. 

536: 

537: With this, we shall prove 

538: 

539: \begin{teo}\label{main}

540: Any global solution $\phi (x,y):\R^2 \setminus \{p_1, \dots, p_n\} \flecha 

541: \R$ to \eqref{monge} is one of the examples in Theorem \ref{ejes} up to 

542: equiaffine transformations. 

543: 

544: In particular, there exists an explicit bijective correspondence between 

545: the conformal equivalence classes of once punctured bounded domains in 

546: $\C$ of connectivity $n-1$, and the space of global solutions to \eqref{monge}

547: with $n$ punctures, modulo equiaffine transformations. 

548: \end{teo}

549: 

550: Let $\cM_n$ be the quotient space of global solutions to \eqref{monge} 

551: with $n>1$ punctures, modulo equiaffine transformations. As an interesting 

552: consequence of Theorem \ref{main} we can endow $\cM_n$ with a finite 

553: dimensional analytic manifold structure: 

554: 

555: \begin{cor}\label{space}

556: The space $\cM_n$ can be naturally identified with an open subset of 

557: $\R^{3n-4}$.

558: \end{cor}

559: \begin{proof}

560: Let $\Omega \subset \C$ be a bounded domain in $\C$ of connectivity 

561: $n-1$, and take $z_0 \in \Omega$. There exists then a Möbius

562: transformation of the Riemann sphere $\C \cup \{\8\} $ taking $\Omega 

563: \setminus \{z_0\}$ to a new complex domain $\Landa\subset \C$ consisting 

564: of 

565: $\C$ with the interior of $n$ Jordan curves removed. Thus, by Koebe's

566: uniformization theorem, $\Landa$ is conformally equivalent to $\C$ with 

567: $n$ disjoint disks removed. Moreover, these disks $D_1,\dots, D_n$ are

568: uniquely determined if we assume the following two restrictions: $D_1 =\D$ 

569: is the unit disk, and 

570: $D_2$ has its center in the positive real line.

571: 

572: Let now $\mathcal{B}_n$ denote the space of conformal equivalence classes 

573: of once punctured bounded domains in $\C$ of connectivity $n-1$. The above 

574: comments show that $\mathcal{B}_n $ can be canonically identified with the 

575: open subset in $(1,+\8)\times \C^{n-2} \times (\R^+)^{n-1} \subset 

576: \R^{3n-4}$ consisting of those points $a = (c_1,\dots, c_{n-1}, r_1, 

577: \dots, r_{n-1})$ for which the open disks of centers $c_j$ and associated 

578: radii $r_j$ (with $c_1 \in (1,+\8)$) together with the unit disk $\D$ are 

579: pairwise disjoint. Once here, the proof follows immediately from Theorem 

580: \ref{main}. 

581: \end{proof}

582: 

583: Before coming to the proof of Theorem \ref{main}, we shall give a 

584: criterion to determine when an isolated singularity of a solution to 

585: \eqref{monge} can be removed. Our perspective here is completely different 

586: from the usual results on removing isolated singularities of Monge-Ampère 

587: equations that can be found in the literature. However this description is 

588: to some extent implicit in the result by Jörgens \cite{Jor2} on removing 

589: isolated singularities of \eqref{monge}. 

590: 

591: \begin{lem}\label{remo}

592: An isolated singularity of a solution $\phi(x,y)$ to \eqref{monge} is 

593: removable if and only if the underlying conformal structure of $\phi(x,y)$ 

594: around the singularity is that of a punctured disk. 

595: \end{lem}

596: \begin{proof}

597: Let $\phi(x,y)$ be a solution to \eqref{monge} in a once punctured 

598: topological disk $\mathcal{U}\setminus \{(x_0,y_0)\}\subset \R^2$. Let 

599: $\zeta :\Omega \subset \C\flecha \mathcal{U}$ be a global holomorphic

600: parameter with respect to the underlying conformal structure of 

601: $\phi(x,y)$. Then, shrinking $\mathcal{U}$ if necessary,

602: $\Omega$ is biholomorphic to a punctured disk or an annulus

603: 

604: Suppose that $(x_0,y_0)$ is a removable isolated singularity of 

605: $\phi$, i.e. $\phi$ extends as a $\mathcal{C}^1$ function across

606: $(x_0,y_0)$. As the holomorphic data verify (see \cite{FMM1} for instance)

607: $G+F = 2x +2 i \phi_y$, $G-F= 2 \phi_x + 2 i y $, both $F,G$ have a well defined value at the

608: singularity. This implies that we must have the punctured disk conformal 

609: type (otherwise, $F,G$ would be constant). 

610: 

611: 

612: Now, suppose that the underlying conformal structure is that of a 

613: punctured disk, say $\Omega =\D^*=\{z :0<|z|<1\}$, and let $F,G$ be the 

614: holomorphic data of the conformal representation. As was proved in 

615: \cite{FMM2}, it must hold on $\D^*$ that $dx^2 + dy^2 \leq |dG|^2$. Hence 

616: $G$ has at most a pole at $0$, and $dG(0) \neq 0$. In addition, as $|dF

617: /dG|<1$ on $\D^*$, we conclude that $F$ has at most a pole at $0$. But 

618: now, if $F$ or $G$ had a pole at $0$, the same would happen to $G+F$ or 

619: $G-F$. This is not possible, by \eqref{rep}. So $F,G$ extend

620: holomorphically to $0$. Finally 

621: $|dF /dG|(0) \neq 1$, because $|dF /dG|<1$ on $\D^*$.

622: Therefore, the conformal representation assures that the parabolic affine 

623: sphere extends regularly across $0$, i.e. $(x_0,y_0)$ is a removable 

624: singularity of $\phi(x,y)$. 

625: \end{proof}

626: 

627: \begin{remark}

628: Lemma \ref{remo} indicates that if $\phi(x,y)$ is a solution to 

629: \eqref{monge} in a punctured neighbourhood of a non-removable isolated 

630: singularity $(x_0,y_0)$, it has the underlying conformal structure of an 

631: annulus $A_r = \{z: 1<|z|<r\} $. Moreover, it is deduced from the 

632: conformal representation that $(G + \overline{F} )|_{\S^1} = x_0 +i y_0$, 

633: where $\S^1 =\{z: |z|=1\} $. So, both $G$ and $F$ can be analytically 

634: continued across $\S^1$ by Schwarzian reflection. In particular, 

635: $\phi(x,y)$ extends continuously to $(x_0,y_0)$, although it is not

636: $\mathcal{C}^1$ at this point.

637: \end{remark}

638: 

639: \begin{proof1}

640: Let $\phi(x,y):\R^2\setminus\{p_1,\dots, p_n\} \flecha \R$ be a global 

641: solution to \eqref{monge}, and let $\Sigma$ denote its underlying Riemann 

642: surface structure. By Lemma \ref{remo}, the underlying conformal type of 

643: $\phi(x,y)$ about any of the isolated singularities is that of an annulus.

644: Moreover, it was proved in \cite{FMM2} that the conformal type of 

645: $\phi(x,y)$ in the exterior of a sufficiently large disk $x^2 +y^2 >R^2$

646: is that of a punctured disk. Putting together these conditions and 

647: applying Koebe's uniformization theorem we deduce that $\Sigma$ is 

648: conformally equivalent to $\C$ with $n$ disjoint disks removed. We shall 

649: denote by $\zeta$ the complex parameter of this domain. At last, by 

650: reflecting this domain with respect to the circle $C$ that bounds one of 

651: these disks, $\Sigma$ can be assumed to be a once punctured bounded domain 

652: $\Omega \setminus \{z_0\}\subset \C$ whose boundary is made up by

653: $n$ disjoint circles, the exterior one being precisely $C$.

654: 

655: Once here, by the representation formula we obtain that the parabolic 

656: affine sphere $\cS_{\phi}=(x,y,\phi(x,y))$ can be conformally parametrized 

657: as $\psi : \Omega\setminus \{z_0\} \flecha \R^3$, where $\psi$ is given by 

658: \eqref{rep} in terms of two holomorphic functions $F,G:\Omega\setminus 

659: \{z_0\} \flecha \C$. Moreover, the map $G +\overline{F} :\Omega\setminus 

660: \{z_0\} \flecha \R^2 \setminus \{p_1, \dots , p_n\} $ is a global 

661: diffeomorphism, $G +\overline{F}$ is constant along each boundary circle 

662: in $\overline{\Omega}$, and $G$ has a pole of order one at $z_0$ 

663: \cite{FMM2}. 

664: 

665: We recall at this point another result in \cite{FMM2}, which describes the 

666: asymptotic behaviour of parabolic affine spheres at infinity. This result 

667: tells in our situation that the solution $\phi(x,y)$ to \eqref{monge} we 

668: started with has the decomposition 

669:   \begin{equation}\label{elipse}

670: \phi (x,y)= E_{\phi} (x,y) + \alfa \log |\zeta|^2 + O(1). 

671:   \end{equation}

672: Here $\zeta$ is the conformal parameter of the above mentioned complex 

673: domain consisting of $\C$ with $n$ disks removed, $\alfa\in \R$, $O(1)$ 

674: stands for a term bounded in absolute value by a constant, and $E_{\phi} 

675: (x,y)$ is a quadratic polynomial. When 

676: $k>0$ is a large positive number, the ellipse $E_{\phi} (x,y)=k$ describes the shape of

677: $\phi (x,y)$ at infinity, and is called the \emph{ellipse at infinity} of

678: $\phi(x,y)$.

679: 

680: Now let $\Phi \in \mathcal{E}$ be an equiaffine transformation as in 

681: \eqref{equi}, and let $\phi'(x',y')$ be the solution to \eqref{monge} 

682: associated to the embedded parabolic affine sphere $\mathcal{S}' = \Phi 

683: \circ \cS_{\phi} $. Then $\phi' (x',y'):\R^2 \setminus \{q_1, \dots, q_n\} 

684: \flecha \R$ is a global solution to \eqref{monge}, where the isolated 

685: singularities $q_1, \dots, q_n$ are in general different from $p_1,\dots, 

686: p_n$. Thus, $\phi'(x,',y')$ has all the properties established for 

687: $\phi(x,y)$, and it becomes clear from the decomposition \eqref{elipse}

688: that its ellipse at infinity differs from the one of $\phi$ only by an 

689: affine transformation of $\R^2$. With this, it is easy to choose an 

690: adequate equiaffine transformation $\Phi$ so that the ellipse at infinity 

691: of $\phi'(x',y')$ is actually a circle. 

692: 

693: Consider next the holomorphic data of the embedded parabolic affine sphere 

694: $\mathcal{S}'$, denoted by $G^*, F^*$, and defined on

695: $\Omega \setminus \{z_0\} $. Then $G^*$ has a simple pole at $z_0$ and, as

696: the ellipse at infinity of $\mathcal{S}'$ is a circle, $F^*$ extends 

697: holomorphically about $z_0 $ (see \cite{FMM2}). 

698: 

699: Furthermore, by performing an adequate dilatation and rotation of 

700: $\Omega$ if necessary, we may assume that the residue of $G^*$ at $z_0$ is

701: $2$.

702: 

703: With all of this we get that $$\psi_1 = \frac{1}{2} \Re \left( G^*+ F^* 

704: \right) :\Omega \setminus \{z_0\} \flecha \R$$ is harmonic, constant on 

705: each boundary component of $\Omega$, and $(G^*+F^*)/2$ has a simple pole 

706: of residue one at $z_0$. Thus, by uniqueness of the holomorphic map $p(z)$ 

707: defined in Section 2 we infer that $G^*(z) +F^*(z) =2 p(z) + c_1$, where 

708: $c_1 \in \C$. Analogously, $G^*(z) -F^*(z) = 2 q(z) + c_2$ for $c_2 \in \C$.

709: Finally we conclude that $$G^*(z)= p(z) +q(z) + d_1, \hspace{1cm} F^*(z)= 

710: p(z) -q(z) +d_2,$$ for constants $d_1,d_2\in \C$. This lets us conclude 

711: that the global solution $\phi'(x',y')$, and thus the original global 

712: solution $\phi(x,y)$, differs from one of the examples in Theorem 

713: \ref{ejes} only by an equiaffine transformation. This proves the first 

714: claim. 

715: 

716: The second claim is essentially direct. We assign to a specific once 

717: punctured complex domain in $\C$ the example constructed in Theorem 

718: \ref{ejes}. As changing the domain within its conformal equivalence class 

719: does not change the surface, we have a well defined mapping from conformal 

720: equivalence classes of once punctured domains with $n$ boundary curves 

721: into global solutions to \eqref{monge} in the $n$-times punctured plane, 

722: modulo equiaffine transformations. This mapping is surjective by the first 

723: part of the theorem, and injective because solutions to \eqref{monge} 

724: differing just by an equiaffine transformation have the same underlying 

725: conformal structure. This ends up the proof. 

726: \end{proof1}

727: 

728: \begin{remark}

729: The two classical Jörgens theorems indicate that there is exactly one (up 

730: to equiaffine transformations) global solution to \eqref{monge} in the 

731: whole plane $\R^2$, and in the once punctured plane. This does not hold 

732: anymore when we consider the $n>1$ times punctured plane, as shown by 

733: Corollary \ref{space}. In any case, given $n>1$ and a conformal 

734: equivalence class of 

735: $\C$ with $n$ disjoint disks removed, there exists a unique (up to equiaffine

736: transformations) global solution to \eqref{monge} in an $n$-times 

737: punctured plane whose underlying conformal structure is the previously 

738: given one. So, unicity in the $n=1$ case comes from the fact that all 

739: domains of the form $\C \setminus D$, where $D$ is a disk in $\C$, are 

740: conformally equivalent. 

741: \end{remark}

742: 

743: Theorem \ref{main} can also be stated in the geometric context of 

744: parabolic affine spheres. Indeed, if a parabolic affine sphere with a 

745: finite number of singularities is the boundary of a convex body then it 

746: must be regular at infinity (see \cite{FMM2}), and reasoning as in Theorem 

747: \ref{ejes} we see that it must be an entire graph. So, we have: 

748: 

749: \begin{cor}\label{Pcv}

750: All embedded parabolic affine spheres in $\R^3$ that have a finite number 

751: of isolated singularities and are the boundary of a convex body of 

752: $\R^3$ must be equiaffinely equivalent to one of the examples in Theorem

753: \ref{ejes}. 

754: \end{cor}

755: Actually, the condition of being the boundary of a convex body can be 

756: substituted by the property of being closed in $\R^3$. Without going into 

757: detail, we just indicate that this fact can be proved as follows: given a 

758: closed parabolic affine sphere $S$, its Legendre transform surface must 

759: lie on the boundary of a convex body, and so $S$ must be globally convex. 

760: Thus, Corollary \ref{Pcv} can be applied. 

761: 

762: \section{Explicit construction for one and two punctures}

763: 

764: To begin this last Section, we show that Theorem \ref{main} for $n=1$ 

765: recovers the classical description by Jörgens of the global solutions to 

766: \eqref{monge} in the once punctured plane. To do so, we just need to 

767: analyze the example with $n=1$ in Theorem \ref{ejes}. Obviously, we may 

768: assume there that $\Omega = \D$, the unit circle, and $z_0 =0$. 

769: 

770: Let $z_0 \in \D$. The Green function in $\D$ with respect to $z_0$ is then 

771: $$\cG (z,z_0)= \log \left|\frac{1- z \overline{z_0}}{z -

772: z_0}\right| .$$ Deriving this expression with respect to $z_0$ as in 

773: \eqref{aru}, \eqref{ardo} we get the harmonic functions $$ u(z,z_0)=\Re 

774: \left( \frac{1}{z-z_0} - \frac{z}{1 - z\overline{z_0}}\right), 

775: \hspace{0.6cm} v(z,z_0)=\Im \left( \frac{1}{z-z_0} + \frac{z}{1 - 

776: z\overline{z_0}}\right) .$$ Therefore, letting $z_0=0$ we see that the 

777: canonical functions $p(z),q(z)$ in 

778: $\D^*$ are $$p(z)= \frac{1}{z} -z, \hspace{1cm}

779: q(z)= \frac{1}{z} +z.$$ Finally, by \eqref{fg} the holomorphic data of the 

780: solution to \eqref{monge} are $G(z)=2 /z$ and $F(z)=-2z$. These are known 

781: to be the holomorphic data (up to the conformal change $w=2/z$) of the 

782: rotational example in the introduction (see \cite{FMM1}). Also observe 

783: that $(G + \overline{F})|_{\S^1} =0$. Hence, Jörgens' theorem is 

784: recovered. 

785: 

786: \begin{figure}[h]\label{rev}

787:   \begin{center}

788: \includegraphics[clip,width=5.5cm]{figura7.eps}\vspace{-2cm}

789: \caption{The unique solution to the Monge-Ampère equation in the once 

790: punctured plane, modulo equiaffine transformations.} 

791:  \end{center}

792: \end{figure}

793: 

794: It is also possible to write down explicitly in terms of theta functions 

795: all global solutions to \eqref{monge} in a twice punctured plane 

796: $\R^2 \setminus \{p_1,p_2\} $, as we show next.

797: 

798: 

799:  \begin{figure}[h]

800:   \begin{center}

801:     \begin{tabular}{cc}

802: \includegraphics[clip,width=7cm]{figura6.eps} &

803: \includegraphics[clip,width=6.5cm]{figura5.eps}\vspace{-.5cm}

804: \end{tabular}

805: \caption{Two different views of a solution to the Monge-Ampère equation in 

806: a twice punctured plane.} 

807:  \end{center}

808: \end{figure}

809: 

810: 

811: Let $\theta_0 (w)$ be the Jacobi theta function (see \cite{CoHi} for 

812: instance) 

813: $$\theta_0(w) =C \prod_{k=1}^{\8} \left(1 - r^{2k-1} e^{2 \pi i w}\right)

814: \left(1 - r^{2k-1} e^{-2 \pi i w}\right),$$ where $C=\prod_{k=1}^{\8} 

815: (1-r^{2k})$ and $0<r<1$. Consider also $\mathbb{A}_r$ to be the annulus 

816: given by $\mathbb{A}_r = \{z : r<|z| <1\}$, and let $\vart_1 (z)$ denote 

817: the \emph{annular} Jacobi theta function given by $\vart_1 (e^{2\pi i w})= 

818: \theta_0 (r w)$. In other words, 

819:  \begin{equation}\label{tet}

820:  \vart_1 (z)= C \left( 1- \frac{1}{z}\right)

821: \prod_{k=1}^{\8} \left( 1- r^{2k} z \right) \left( 1- r^{2k}/ z \right). 

822:  \end{equation}

823: It comes then clear from this expression that 

824:  \begin{equation}\label{tetrel}

825:  \vart_1 (z)= \overline{\vart_1(\bar{z})} = -r^2 z \, \vart_1 (r^2 z),

826:  \hspace{0.7cm} \vart_1 (z /r^2)= -z \vart_1 (z),

827:  \end{equation}

828: and thus that 

829:  \begin{equation}\label{tetre2}

830:  \vart_1 '(z) = - r^2 \vart_1 (r^2 z) - r^4 z \vart_1'(r^2 z),

831:  \hspace{0.5cm} \vart_1'(z/r^2)= -r^2 \vart_1(z) - r^2 z \vart_1 '(z).

832:  \end{equation}

833: 

834: 

835: A formula for the Green function of an annulus of the type 

836: $\{r^{1/2} < |z|< r^{-1/2}\}$ for $r \in (0,1)$ can be found in

837: \cite[pg. 386-387]{CoHi}. By the conformal invariance property of the 

838: Green function, we obtain from this formula that the Green function 

839: $\cG(z,z_0)$ of $\mathbb{A}_r$ with respect to a point $z_0\in \mathbb{A}_r$ can be

840: obtained in terms of 

841: $\vart_1$ as $$\cG (z,z_0)= \log |z| \left( 1 +\frac{\log

842: |z_0|}{\log r}\right) -\log \left|\frac{\vart_1 ( z_0 /z)}{\vart_1 (z 

843: \overline{z_0})} \right| . 

844: $$ In other words, $$\cG(z,z_0)= \Re h(z) = \Re \left\{\log

845: \left( \frac{z^{1 +\log |z_0| / \log r}}{\vart_1 (z_0/z) / \vart_1 

846: (\overline{z_0} z)}\right) \right\} .$$ At this point we can find the 

847: harmonic function in \eqref{aru} by differentiation. We obtain $$u (z)= 

848: \Re \left\{ \frac{\log z}{\log r} \left(\frac{1}{z_0} + 

849: \frac{1}{\overline{z_0}}\right)\right\} -\Re \left \{ \frac{\vart_1'(z_0 

850: /z)}{z \vart_1 (z_0 /z)} \frac{\vart_1'(\overline{z_0} z) z}{\vart_1 

851: (\overline{z_0}z)} \right\} .$$ Noting that $$ \frac{\vart_1'(z_0 /z)}{z 

852: \vart_1 (z_0 /z)} -\frac{\vart_1'(\overline{z_0} z) z}{\vart_1 

853: (\overline{z_0}z)}$$ is a well defined meromorphic function on 

854: $\mathbb{A}_r$, the way we defined the holomorphic function $p(z)$ in

855: \eqref{truqui} suggests that 

856:  \begin{equation}\label{p2}

857:  p(z)=\frac{\vart_1'(\overline{z_0}

858: z) z}{\vart_1 (\overline{z_0}z)} - \frac{\vart_1'(z_0 /z)}{z \vart_1 (z_0 

859: /z)} . 

860:  \end{equation}

861: The validity of this assertion is easily checked: $p(z)$ is holomorphic on 

862: $\mathbb{A}_r \setminus \{z_0\} $, it has a simple pole of residue $1$ at

863: $z_0$, and using \eqref{tetrel} and \eqref{tetre2} we see that

864:  \begin{equation}\label{cambi}

865: p (1/\bar{z}) + \overline{p(z)} =0, \hspace{1cm} p (r^2 /\bar{z}) + 

866: \overline{p(z)} = -\frac{1}{z_0} - \frac{1}{\overline{z_0}}. 

867:  \end{equation}

868: So, the real part of $p$ is constant on the boundary curves $|z|=r,1$. 

869: 

870: The corresponding process starting this time with the harmonic function 

871: defined via \eqref{ardo} lets us recover the other canonical holomorphic 

872: function $q(z):\mathbb{A}_r \setminus \{z_0\} \flecha \C$ as 

873:   \begin{equation}\label{p3}

874:  q(z)=-\frac{\vart_1'(z_0 /z)}{z \vart_1 (z_0 /z)} -\frac{\vart_1'(\overline{z_0}

875: z) z}{\vart_1 (\overline{z_0}z)}. 

876:  \end{equation}

877: Putting together \eqref{p2} and \eqref{p3}, and recalling the formula 

878: \eqref{fg}, we obtain the holomorphic data on $\mathbb{A}_r \setminus 

879: \{z_0\} $ 

880: 

881:  \begin{equation}\label{data2}

882:  G(z)= -\frac{2 \vart_1'(z_0 /z)}{z \vart_1 (z_0 /z)}, \hspace{0.6cm} F(z)= \frac{2\vart_1'(\overline{z_0}

883: z) z}{\vart_1 (\overline{z_0}z)}. 

884:  \end{equation}

885: 

886: Finally, by Theorem \ref{main}, all solutions to \eqref{monge} in a twice 

887: punctured plane are recovered, up to equiaffine transformations, by the 

888: holomorphic data in \eqref{data2}. The 2-parametric family of such 

889: solutions described by Corollary \ref{space} is given by the variation of 

890: $r\in (0,1)$ and $|z_0| \in (r,1)$. In the solution to \eqref{monge} we have just

891: constructed, the singularities are located at the points $(0,0)$ and $(\Re 

892: (-2 /\overline{z_0}), \Im (-2/\overline{z_0}))$ of $\R^2$. This follows 

893: from \eqref{rep} and the relation \eqref{cambi}. 

894: 

895: 

896: 

897: 

898: 

899: 

900: \def\refname{References}

901: \begin{thebibliography}{9}

902: 

903: \bibitem[Ahl]{Ahl}  L.V. Ahlfors, {\it Complex analysis. An introduction to the theory of 

904: analytic functions of one complex variable}. McGraw-Hill, New 

905: York-Toronto-London, 1953. 

906: 

907: \bibitem[BCGJ]{BCGJ} J. Bao, J. Chen, B. Guan, M. Jin, Liouville property

908: and regularity of a Hessian quotient equation, {\it Amer. J. Math.} {\bf 

909: 125} (2003), 301--316. 

910: 

911: \bibitem[CaLi1]{CaLi1} L. Caffarelli, Y. Li, An extension to a theorem of Jörgens,

912: Calabi, and Pogorelov, {\it Comm. Pure Appl. Math.} {\bf 56} (2003), 

913: 549--583. 

914: 

915: \bibitem[CaLi2]{CaLi2} L. Caffarelli, Y. Li, A Liouville theorem for solutions of the

916: Monge-Ampère equation with periodic data, {\it Ann. Inst. H. Poincaré 

917: Anal. Non Linéaire} {\bf 21} (2004), 97--120 

918: 

919: \bibitem[Cal]{Cal} E. Calabi, Improper affine hyperspheres of convex type

920: and a generalization of a theorem by K. Jörgens, {\it Michigan Math. J.} 

921: {\bf 5} (1958), 105--126. 

922: 

923: \bibitem[ChYa]{ChYa} S.Y. Cheng, S.T. Yau, Complete affine hypersurfaces.

924: Part I. The completeness of affine metrics, {\it Comm. Pure Appl. Math.} 

925: {\bf 39} (1986), 839--866. 

926: 

927: \bibitem[ChWa]{ChWa} K-S. Chou, X-J. Wang, Entire solutions of the

928: Monge-Ampère equation, {\it Comm. Pure Appl. Math.} {\bf 49} (1996), 

929: 529--539. 

930: 

931: \bibitem[CoHi]{CoHi} R. Courant, D. Hilbert, {\it Methods of Mathematical

932: Physics}, vol. I. John Wiley and Sons, 1953. 

933: 

934: \bibitem[FMM1]{FMM1} L. Ferrer, A. Martínez, F. Milán, Symmetry and uniqueness of parabolic

935: affine spheres, {\it Math. Ann.} {\bf 305} (1996), 311--327. 

936: 

937: \bibitem[FMM2]{FMM2} L. Ferrer, A. Martínez, F. Milán, An extension of a theorem by K.

938: Jörgens and a maximum principle at infinity for parabolic affine spheres, 

939: {\it Math. Z.} {\bf 230} (1999), 471--486. 

940: 

941: %%\bibitem[FMM3]{FMM3} L. Ferrer, A. Martínez, F. Milán, The space of parabolic affine

942: %%spheres with fixed compact boundary, {\it Monatsh. Math.} {\bf 130}

943: %%(2000), 19--27.

944: 

945: \bibitem[Jor1]{Jor1} K. Jörgens,  Über die Lösungen der Differentialgleichung $rt-s\sp 2=1$. {\it Math.

946: Ann.} {\bf 127} (1954), 130--134. 

947: 

948: \bibitem[Jor2]{Jor2} K. Jörgens, Harmonische Abbildungen und die Differentialgleichung

949: $rt-s\sp 2=1$. {\it Math. Ann.} {\bf 129} (1955), 330--344.

950: 

951: \bibitem[Pog]{Pog} A.V. Pogorelov, On the improper convex affine hyperspheres,

952: {\it Geom. Dedicata} {\bf 1} (1972), 33--46. 

953: 

954: \bibitem[TrWa1]{TrWa1} N.S. Trudinger,  X-J. Wang, The Bernstein problem for affine

955: maximal hypersurfaces. {\it Invent. Math.} {\bf 140} (2000), 399--422. 

956: 

957: \bibitem[TrWa2]{TrWa2} N.S. Trudinger, X-J. Wang, Bernstein-Jörgens theorem

958: for a fourth order partial differential equation, {\it J. Partial 

959: Differential Equations} {\bf 15} (2002), 78--88. 

960: 

961: 

962: \end{thebibliography}

963: \end{document}

964: 

965: 

966: \end{document}

967: