math0510071/uqc.tex
1: \documentclass[11pt,twoside]{article}
2: \include{epsf}
3: \usepackage{latexsym, amssymb, amsmath, theorem}
4: \usepackage{calc, epsfig}
5: \pagestyle{myheadings} 
6: \markboth{A.~Glutsyuk}{Simple proofs of uniformization theorems} 
7: 
8: \setlength{\textwidth}{15.5cm}
9: \setlength{\textheight}{22cm}
10: \setlength{\oddsidemargin}{1cm}
11: \setlength{\evensidemargin}{1cm}
12: \setlength{\headheight}{12pt}
13: \setlength{\headsep}{20pt}
14: \setlength{\topmargin}{0cm}
15: 
16: \numberwithin{equation}{section}
17: \renewcommand{\labelenumi}{(\arabic{enumi})}
18: 
19: \theoremstyle{change}
20: \theorembodyfont{\itshape}
21: \newtheorem{theorem}{Theorem}[section]
22: \newtheorem{proposition}[theorem]{Proposition}
23: \newtheorem{lemma}[theorem]{Lemma}
24: \newtheorem{corollary}[theorem]{Corollary}
25: 
26: \theorembodyfont{\rmfamily}
27: \newtheorem{definition}[theorem]{Definition}
28: \newtheorem{claim}[theorem]{Claim}
29: \newtheorem{example}[theorem]{Example}
30: \newtheorem{remark}[theorem]{Remark}
31: \newenvironment{proof}{{\noindent \textbf{Proof}\,\,}}{\hspace*{\fill}$\Box$\medskip}
32: \def\rd{\mathbb R^2}
33: \def\cc{\mathbb C}
34: \def\bc{\overline{\cc}}
35: \def\ci{C^{\infty}}
36: 
37: \title{\textbf 
38: {Simple proofs of uniformization theorems}}
39: \author{A.A.Glutsyuk\\[5pt]
40: \textit{CNRS, Unit\'e de Math\'ematiques Pures et Appliqu\'ees, M.R.,} \\
41: \textit{\'Ecole Normale Sup\'erieure de Lyon} \\
42: \textit{46 all\'ee d'Italie, 69364 Lyon Cedex 07 France}}
43: %\\
44: %\textit{Email: aglutsyu$\@$umpa.ens-lyon.fr}}
45: \begin{document}
46: \maketitle
47: \def\td{\mathbb T^2}
48: \begin{abstract} 
49: The measurable Riemann mapping theorem proved 
50: by Morrey and in some particular cases by Ahlfors, Lavrentiev 
51: and Vekua, says that any measurable almost complex 
52: structure on $\rd$ ($S^2$) with bounded dilatation is integrable: there is a 
53: quasiconformal homeomorphism of $\rd$ ($S^2$) onto $\cc$ 
54: ($\bc$) transforming the given almost complex structure to the standard one. 
55: We give an elementary proof of this theorem that is done as follows. 
56: Firstly we prove its double-periodic version: each $\ci$ almost complex 
57: structures on the two-torus can be transformed by a diffeomorphism 
58:  to the standard complex structure on appropriate complex torus. The proof is 
59:  based on the homotopy method for the Beltrami equation on $\td$ 
60: with parameter. (As a by-product, 
61: we present a simple proof of the Poincar\'e-K\"obe theorem saying that each 
62: simply-connected Riemann surface is conformally equivalent to either 
63: $\overline{\cc}$, or $\cc$, or the unit disc.) Afterwards the 
64: general case is treated by $\ci$ double-periodic approximation  and simple
65:  normality arguments (involving Gr\"otzsch inequality) following 
66:  the classical scheme.  
67: \end{abstract}
68: 
69: \tableofcontents
70: \vspace{.5cm}
71: \hrule
72: 
73: \section{Introduction, the plan of the paper and history}
74: \subsection{Uniformization theorems. The plan of the paper}
75: A linear complex structure on $\rd$ is a structure of a linear space 
76: over $\mathbb C$ (we fix an orientation and consider it to be 
77: compatible with the complex structure). 
78: The {\it (almost) complex structure} on a real two-dimensional surface is 
79: a family of linear complex structures on the tangent planes at its points.  
80: A linear complex structure on $\rd$ defines an ellipse 
81: in $\rd$ centered at 0, which is an orbit under the $S^1$- action 
82: by multiplication by complex numbers with unit module. (The ellipse 
83: corresponding to the standard complex structure on $\cc$ is a circle.) 
84: The {\it dilatation} of a nonstandard linear complex structure on $\cc$ 
85: (with respect to the standard complex structure) is the excentricity 
86: of the corresponding ellipse (i.e., the ratio of the largest radius over 
87: the smallest one). An almost complex 
88: structure defines an ellipse field in tangent planes, and vice versa: 
89: the ellipse field determines the almost complex structure in a unique way. 
90: 
91: If our surface is a Riemann surface (with a fixed complex structure), 
92: then any  (nonstandard) almost complex 
93: structure has a well-defined dilatation at each point of the surface. 
94: In this case an almost complex structure is said to be {\it bounded}, 
95: if its dilatation is bounded. 
96: The  {\it (total) dilatation} of a bounded almost complex structure is 
97: the supremum of its dilatations (more precisely, the minimal supremum 
98: of dilatations after possible correction of the almost complex 
99: structure over a measure zero set). 
100: 
101: Each real linear isomorphism $\cc\to\cc$ acts on the space of the ellipses centered 
102: at 0, and hence, on the space of linear complex structures. Its {\it 
103: dilatation} is defined to be the dilatation of the image of the 
104: standard complex structure (which is equal to the excentricity of the 
105: image of a circle centered at 0). 
106: The action of a differentiable homeomorphism of domains in $\mathbb C$ 
107: on the almost complex structures 
108: and its dilatation (at a point) are defined to be those of its derivative. 
109: Its {\it (total)} dilatation is the supremum of the dilatations through all 
110: the points. 
111: 
112: It appears that {\it any $\ci$ (and even measurable) bounded almost 
113: complex structure is integrable}, that is, can be transformed to a true  
114: complex structure by a $\ci$ (respectively, quasiconformal) homeomorphism, see 
115: the following Definition and Theorem. 
116: 
117: \begin{definition} (see, e.g., [Ah2]). Let $K>0$. 
118: A homeomorphism of domains in $\cc$ is said to be $K$- 
119: {\it quasiconformal} (or $K$- homeomorphism), 
120: if it  has local $L_2$ derivatives and its dilatation 
121: (at the differentiability points with nonzero derivative) is 
122:  no greater than $K$. 
123: A homeomorphism is said to be quasiconformal if it is 
124: $K$- quasiconformal for some $K>0$. 
125: \end{definition}
126: 
127: \begin{remark} \label{grr} The dilatations of a 
128: differentiable homeomorphism and its inverse are equal.  In particular, the inverse to a $K$- diffeomorphism is also a 
129: $K$- diffeomorphism. The composition of two $K$- diffeomorphisms is a 
130: $K^2$- diffeomorphism. This follows from definition. 
131: \end{remark}
132: 
133: \begin{proposition} \label{group}  (see [Ah2])   
134: The quasiconformal homeomorphisms of a Riemann surface form a group.
135: \end{proposition}
136: 
137: \begin{proposition} \label{measure0} The image of a zero measure 
138: set under a quasiconformal homeomorphism has also zero measure.  
139: \end{proposition}
140: 
141: \begin{corollary} \label{cormeasure} For any quasiconformal homeomorphism 
142: the set of its differentiability points with zero derivative has measure 
143: zero. 
144: \end{corollary} 
145: 
146: \begin{proof} The image of the set from the Corollary has zero measure 
147: by definition. Therefore, the set itself has zero measure (Proposition 
148: \ref{measure0} applied to the inverse mapping, which is quasiconformal 
149: by Proposition \ref{group}).
150: \end{proof}  
151:  
152: Both Propositions are proved in Subsection 3.4 and neither them, nor the 
153: Corollary  will be used in the paper. 
154: 
155: \begin{definition}
156: A homeomorphism $\cc\to\cc$ is said to be {\it normalized}, if it fixes 0 and 1. 
157: \end{definition}
158: 
159: \begin{theorem} \label{c} ([AhB], [M]). For any measurable 
160: bounded almost complex structure $\sigma$ on $\mathbb C$ there exists a unique 
161: normalized quasiconformal homeomorphism 
162: $\mathbb C\to\mathbb C$ that transforms $\sigma$ 
163: to the standard complex structure 
164: (at the differentiability points with nonzero derivative). 
165: If $\sigma$ is $\ci$ in some domain, then 
166: the homeomorphism is a $\ci$ diffeomorphism while restricted to this domain.  
167: \end{theorem}
168: 
169: {\bf Addendum [AhB].} If a bounded almost
170: complex structure on $\cc$ varies analytically in a complex parameter, then 
171: so does the corresponding homeomorphism from Theorem \ref{c}.  
172: \medskip
173: 
174: \begin{remark} A quasiconformal homeomorphism of a once punctured domain 
175: extends quasiconformally to the puncture 
176: (in particular, the homeomorphism from Theorem \ref{c} is quasiconformal 
177: at infinity). This follows easily 
178: from the local uniqueness of the quasiconformal homeomorphism up to 
179: composition with conformal mapping (Proposition \ref{uniq}, see 
180: Subsection 3.4) and the theorem on erasing isolated singularities 
181: of bounded holomorphic functions. 
182: \end{remark}
183: 
184: In the present paper we give proofs of Theorem \ref{c} (Sections 2, 3) 
185: and the Addendum (Subsection 3.5) 
186: that seem to be simpler than the known proofs and easier to explain. 
187: A historical overview will be given in Subsection 1.4. 
188: 
189: \begin{remark} The proof of the local integrability of an analytic almost 
190: complex structure is elementary: it is done immediately 
191: by analyzing the complexification of the corresponding $\cc$- linear 1- form 
192: (this proof is due to Gauss). But it is already nontrivial in the $\ci$ case. 
193: \end{remark}
194: 
195: The measurable versions of the Theorem and the Addendum have many 
196: very important 
197: applications in various domains of mathematics, especially in holomorphic 
198: dynamics and the 
199: Kleinian group theory (quasiconformal surgery, where one deals with 
200: invariant almost complex structures that are discontinuous...), see, e.g.,   
201: [CG]. 
202: 
203: For the proof of Theorem \ref{c} we firstly prove (in Section 2) its 
204: version for $\ci$ almost 
205: complex structures on the two-torus: the proof uses only elementary 
206: Fourier analysis. 
207: 
208: \begin{theorem} \label{td} ([Ab]) For any $\ci$ 
209: bounded almost complex structure $\sigma$ on $\td$ there exists a 
210: $\ci$ diffeomorphism of $\td$ onto appropriate 
211: complex torus (the latter torus depends on $\sigma$) 
212: that transforms $\sigma$ to the standard complex structure.
213: \end{theorem} 
214: 
215: 
216: Then in Section 3 we deduce Theorem \ref{c} from 
217: Theorem \ref{td} by using double-periodic approximations of a given  
218: almost complex structure on $\cc$ and simple normality arguments  involving a 
219: Gr\"otzsch inequality for annuli diffeomorphisms. This deduction 
220: follows the classical scheme [Ah2]. 
221: 
222: The proof of  Theorem \ref{td} presented below is implicitly contained in the 
223: previous paper [Gl] by the author, where the same method was used to 
224: prove a foliated version of Theorem \ref{c}. We prove the existence of a global 
225: nowhere vanishing $\sigma$- holomorphic differential. To do this, 
226: we use the homotopy method for the Beltrami equation with parameter, which   
227: reduces the proof to solving a linear 
228: ordinary differential equation in $L_2(\td)$. We prove
229: regularity of its solution by 
230: showing that the equation is bounded in any Sobolev space $H^s(\td)$.
231: 
232: In Subsection 1.3 we give a proof of  
233: the classical Poincar\'e-K\"obe uniformization theorem using 
234: Theorem \ref{td}: 
235: \begin{theorem} \label{pk} [Ko1], [Ko2], [P]. Each simply-connected Riemann surface is 
236: conformally equivalent to either unit disc, or $\cc$, or the Riemann sphere.
237: \end{theorem}
238: 
239: In the proofs of the previously mentioned Theorems we use  the 
240: well-known notations (recalled in the next Subsection) concerning 
241: almost complex structures. 
242: \def\omu{\omega_{\mu}}
243: 
244: \subsection{Complex structures and uniformizing differentials. Basic notations}
245: 
246:  To a (nonstandard) almost complex structure (denoted $\sigma$) 
247:  on a subset $D\subset \mathbb C$ 
248: we put into correspondence a $\cc$- valued 1- form that is 
249: $\cc$- linear with respect 
250: to $\sigma$. The latter form can be normalized to have the type 
251: \begin{equation}
252: \omu=dz+\mu(z) d\bar z, \ |\mu|<1.\label{1.1}
253: \end{equation}
254:  The function $\mu$ is uniquely defined by $\sigma$. Vice versa, for arbitrary 
255: complex-valued function $\mu$, $|\mu|<1$, the 1- form (\ref{1.1}) defines the  
256: unique complex structure for which it is $\cc$- linear. We denote by 
257: $\sigma_{\mu}$ the almost complex structure thus defined (whenever the contrary 
258: is not specified). Then $\sigma_{\mu}$ is bounded, if and only if $\sup|\mu|<1$. 
259: 
260: \begin{remark} The ellipse associated to $\sigma_{\mu}$ 
261: on the tangent plane at a point $z$ is given 
262: by the equation $|dz+\mu(z)d\bar z|=1$; the dilatation (excentricity) is 
263: equal to $\frac{1+|\mu(z)|}{1-|\mu(z)|}$. 
264: \end{remark}
265: We will be looking for a 
266: differentiable homeomorphism $\Phi(z)$ that is holomorphic, i.e., 
267: that transforms $\sigma_{\mu}$ to the standard complex structure. This is equivalent 
268: to say that the differential of $\Phi$ (which is a closed form) 
269: is a $\cc$- linear form, i.e., has the type $f(z)(dz+\mu d\bar z)$:  
270: $$\frac{\partial\Phi}{\partial\bar z}=\mu\frac{\partial\Phi}{\partial z}.$$
271: 
272: \begin{remark} Conversely, let $\mu$ be $\ci$, $|\mu|<1$. 
273: Then any $\ci$ closed 
274: 1- form $f(z)(dz+\mu d\bar z)$ is $\sigma_{\mu}$-  holomorphic, 
275: i.e., is a differential of a complex-valued $\ci$ function $\Phi$ transforming 
276: $\sigma_{\mu}$ to the standard complex structure. 
277: A form $f(z)(dz+\mu d\bar z)$ is closed if and only if 
278: \begin{equation}\partial_{\bar z}f=\partial_z(\mu f).\label{dz}\end{equation}
279: \end{remark}
280: \begin{definition} A Riemann surface is said to be {\it parabolic}, 
281: if its universal covering is conformally equivalent to $\mathbb C$ 
282: (i.e, the surface is either $\mathbb C$, or $\mathbb C^*$, or a complex torus). 
283: \end{definition}
284: 
285: \begin{definition} \label{unifdif} The {\it uniformizing differential} on 
286: $\mathbb C$ (or on a complex torus) 
287: with the affine coordinate $z$ is the 1- form $dz$ or its nonzero constant 
288: multiple. More generally, a holomorphic 1- form on a parabolic Riemann surface 
289: is said to be {\it a uniformizing differential}, if the primitive of 
290: its lifting to the universal cover is a conformal isomorphism onto $\mathbb C$. 
291: \end{definition}
292: 
293: \def\smu{\sigma_{\mu}}
294: 
295: \begin{remark} \label{umetric} The uniformizing differential is well-defined up 
296: to multiplication by constant. It coincides with the unique (up to constant) 
297: nowhere vanishing holomorphic differential whose squared module is 
298: a complete metric. 
299: \end{remark}
300: \begin{proposition} \label{uniftd} Let $\mu:\td\to\cc$ be a $\ci$ function, 
301: $|\mu|<1$.  Suppose there is a $\ci$ 
302: nowhere vanishing function $f:\td\to\cc\setminus0$ satisfying (\ref{dz}). 
303: Then the corresponding almost complex structure $\sigma_{\mu}$ is 
304: integrable and the form $f\omu$ 
305:  is a uniformizing differential of $(\td,\smu)$. 
306: \end{proposition}
307: 
308: The Proposition follows from compactness and the two previous Remarks.
309: 
310: \subsection{Proof of the uniformization Theorem \ref{pk} modulo 
311: Theorem \ref{td}}
312: 
313: Let $S$ be a simply-connected Riemann surface. Then it is either contractible, 
314: then homeomorphic $\rd$, or is homeomorphic to the two-sphere. 
315: We prove the statement of Theorem \ref{pk} only in 
316: the case, when $S$ is contractible: we show that 
317: $S$ is conformally equivalent to 
318: either $\cc$ or disc. Then if $S$ is sphere, it follows that 
319: $S$ is conformally-equivalent to $\overline{\cc}$ (by the previous 
320: statement applied to once punctured $S$ and the theorem on 
321: erasing isolated singularities of bounded holomorphic functions). 
322: In the proof of Theorem \ref{pk} we use the following Corollary 
323: of Theorem \ref{td} and 
324: the Riemann mapping theorem (saying that any simply-connected domain in $\cc$ 
325: distinct from $\cc$ is conformally equivalent to unit disc: the proof 
326:  is elementary and is contained in standard courses of complex analysis.)
327: 
328: \begin{corollary} For any bounded $\ci$ almost complex structure $\sigma$ 
329: on the closed unit disc $\overline D$ there exists a $\ci$ diffeomorphism of the open 
330: disc $D$ onto itself transforming $\sigma$ to the standard complex structure. 
331: \end{corollary}
332: 
333: \begin{proof} Let us extend $\sigma$ 
334: to $\rd$ up to a double-periodic bounded $\ci$ almost complex structure 
335: (say, with periods $4$ and $4i$) and consider the quotient torus equipped 
336: with the induced almost complex structure.  
337: Then the corresponding tori diffeomorphism from 
338: Theorem \ref{td} transforms the latter structure to the standard one. Its  
339: lifting to the universal covers 
340: transforms $D$ to a simply-connected domain in $\cc$ and 
341: sends $\sigma$ to the standard complex structure. Now applying 
342: the Riemann mapping theorem to the image of $D$ proves the Corollary. 
343: \end{proof}
344: 
345: We assume that the Riemann surface $S$ is contractible, hence, 
346: admits a $\ci$ 1-to-1 parametrization by $\rd$. 
347: Its complex structure induces a $\ci$ almost complex structure 
348: (denote it $\sigma$) on $\rd$  
349: (not necessarily bounded). Take a growing sequence of discs  
350: $S_1\Subset S_2\Subset\dots S$ exhausting $S$ centered at 0. 
351: On each $S_n$ the almost complex structure $\sigma$ is bounded. By the 
352: Corollary, for any $n$ 
353: there is a diffeomorphism $\phi_n:S_n\to D$ conformal with respect 
354: to the complex structure of $S$, $\phi_n(0)=0$. 
355: Let $w$ be a local holomorphic chart on $S$ near 0, $w(0)=0$. 
356: Let us change $\phi_n$ to its constant multiple 
357: $\Phi_n=\lambda_n\phi_n$ having unit derivative in $w$ at 0. The family 
358: $\Phi_n$ is normal: each subsequence contains a subsequence converging 
359: uniformly on compact sets in $S$. Indeed, fix a $k\in\mathbb N$ 
360: and consider the $\ci$ injections 
361: $\Phi_n\circ\phi_k^{-1}:D\to\Phi_n(S_n)$, $n\geq k$. By construction, the 
362: latters are holomorphic and univalent, they 
363: send 0 to 0 and have one and the same derivative at 0. Therefore, they form a 
364: normal family, see [CG], hence, so do the $\Phi_n$'s. By construction, the 
365: limit of a converging subsequence of the $\Phi_n$' s is a conformal 
366: diffeomorphism of $S$ onto either a disc, or $\cc$. Theorem \ref{pk} is proved. 
367: 
368: \def\onu{\omega_{\nu}}
369: 
370: \subsection{Historical overview} The local integrability of a 
371: $\ci$ (and even H\"older) almost complex structure was proved by 
372: Korn [Korn] and Lichtenstein [Licht]; a simpler proof was obtained 
373: by Chern [Chern] and Bers [Be]. The local integrability together with the 
374: Poincar\'e-K\"obe uniformization Theorem \ref{pk} imply the  
375: global integrability  statement of Theorem \ref{c}. Lavrentiev [La] 
376: gave a direct proof of Theorem \ref{c} for continuous almost complex 
377: structures. Later Ahlfors [Ah1] and Vekua [Vek] gave another direct proofs  
378: under the previous (stronger) H\"older condition. 
379: 
380: In the general measurable case Theorem \ref{c} was proved by Morrey [M]. 
381: Later new proofs were obtained by Ahlfors and Bers [AhB], Bers and 
382: Nirenberg [BeN] and Boyarskii [Bo].  
383: (In fact, Lavrentiev and Morrey stated 
384: their theorems for almost complex structures on a disc, but their versions 
385: on $\rd$ follow immediately, e.g.,  by the arguments from the previous 
386: Subsection.) A new simpler proof of Theorem \ref{c}  
387: using $L_2$ analysis and Fourier transformation on $\rd$ 
388: was recently obtained by A.Douady and X.Buff [DB].  
389: 
390: \section{Smooth complex structures on $\td$. Proof of Theorem \ref{td}}
391: 
392: \subsection{Homotopy method. The sketch of the proof of Theorem \ref{td}}
393: 
394: Let $\mu:\td\to\cc$ be a $\ci$ complex-valued function, $|\mu|<1$, 
395: $\smu$ be the corresponding almost complex structure, see (\ref{1.1}). 
396: Theorem \ref{td} says that there exists 
397: a diffeomorphism transforming $(\td,\smu)$ into a complex torus equipped 
398: with the standard complex structure. To prove this statement, 
399: it suffices to construct 
400: a uniformizing differential, more precisely, a $\ci$ nowhere vanishing 
401: function $f:\td\to\cc\setminus0$ such that the form $f\omu$ is closed 
402: (see Proposition \ref{uniftd}), i.e., to solve partial differential 
403: equation (\ref{dz}) in a $\ci$ nowhere vanishing function $f$. 
404: 
405: 
406: To solve (\ref{dz}), we use the homotopy method. 
407: Namely, we include $\smu$ into the one-parametric 
408: family of complex structures (denoted by $\sigma_{\nu}$) defined by  their 
409: $\mathbb C$- linear  1- forms 
410: $$\onu=dz+\nu(z,t)d\bar z,\ \nu(z,t)=t\mu(z),\ t\in[0,1].$$
411: The complex structure corresponding to the parameter value $t=0$ is the 
412: standard one, the given structure $\sigma_{\mu}$ corresponds to 
413: $t=1$. 
414: We will find a $C^{\infty}$ family $f(z,t):\td\times[0,1]\to\cc\setminus0$ of 
415: complex-valued nowhere vanishing $C^{\infty}$ functions on $\td$ depending on 
416: the same parameter $t$, $f(z,0)\equiv1$, such that the differential forms 
417: $f(z,t)\onu$ are closed, i.e., 
418: \begin{equation} \partial_{\bar z}f=\partial_z(f\nu).\label{dzt}\end{equation}
419: Then the function $f=f(z,1)$ is the one we are looking for. 
420: 
421: To construct the previous family of functions, we will find firstly 
422: a family $f(z,t)$ of 
423: {\it nonidentically-vanishing} (not necessarily nowhere vanishing) 
424: functions satisfying (\ref{dzt}): 
425: 
426: \begin{lemma} \label{homot} Let $\nu(z,t):\td\times[0,1]\to\cc$ be a $\ci$ family of 
427: $C^{\infty}$ functions on $\td$, $|\nu|<1$, $\nu(z,0)\equiv0$, $z$ be 
428: the complex coordinate on $\td$. There exists a $\ci$ family 
429: $f(z,t):\td\times[0,1]\to\cc$ of $\ci$ functions on $\td$ that are solutions of 
430: (\ref{dzt}) with the initial condition 
431: $f(z,0)\equiv1$ such that for any fixed $t\in[0,1]$ $f(z,t)\not\equiv0$ in $z$.
432: \end{lemma}
433: 
434: The Lemma will be proved in the next Subsection.
435:  
436: Below we show that in fact, the functions $f(z,t)$ from the Lemma vanish 
437: nowhere. To do this (and only in this place) we use the 
438: local integrability of  a $\ci$ complex structure:  
439: 
440: \begin{proposition} \label{loc} ([Korn], [Licht], [La], [Chern], [Be]). 
441: Let $D\subset\cc$ be a disc centered at 0, 
442: $\mu:D\to\cc$, $\mu\in\ci$, $|\mu|<1$, $\smu$ be the corresponding 
443: almost complex structure, see (\ref{1.1}). There exists a 
444: local $\sigma_{\mu}$- holomorphic univalent coordinate near 0.
445: \end{proposition} 
446: 
447: The Proposition will be proved in Subsection 2.3. 
448: 
449: \begin{proof} {\bf of Theorem \ref{td} modulo Lemma \ref{homot} 
450: and Proposition \ref{loc}.} Let $f(z,t)$ be a family of functions 
451: from the previous Lemma. By the previous discussion, it suffices 
452: to show that $f(z,t)\neq0$. This inequality holds for $t=0$, where $f=1$. 
453: 
454:  Let us prove that $f(z,t)\neq0$ by contradiction. Suppose the contrary. 
455: Then the set of the parameter values $t$ corresponding to the functions $f(z,t)$ 
456: having zeroes is nonempty (denote this set by $M$). Its complement 
457: $[0,1]\setminus M$ is open by definition. Let us show that the set $M$ is open 
458: as well. This will imply that the parameter segment is a union of two disjoint 
459: open sets, which will bring us to contradiction. It suffices to show that 
460: the (local) presense of a zero of a function $f$ persists under perturbation. 
461: 
462: Suppose $f(z_0,t)=0$ for some $z_0$, $t$ (let us fix them). It suffices to 
463: show that for $t'$ close to $t$ the function $f(z,t')$ has a zero 
464: near $z_0$. Let $w$ 
465: be the local holomorphic coordinate on $\td$ near $z_0$ from the previous 
466: Proposition corresponding to $\mu=\nu(z,t)$, $w(z_0)=0$. 
467: Suppose that 
468: the function $f(z,t)$ does not vanish identically on $\td$ 
469:  locally near $z_0$: one can achieve this by changing  
470: $z_0$, since $f$ does not vanish identically. Recall that $f\onu$ is 
471: a closed $\cc$- linear 1-form with  respect to the variable complex structure 
472: $\sigma_{\nu}$, hence, it is holomorphic in the 
473: coordinate $w$. Therefore, $f\onu=(w^k+\text{higher terms})dw$, $k\geq1$. 
474: Now by the index argument, the local presense of zero of $f$ on 
475: $\td$ persists under 
476: perturbation. This together with the previous discussion proves the inequality 
477:  $f(z,t)\neq0$ and Theorem \ref{td}.
478:  \end{proof}
479:  
480:  \subsection{Variable holomorphic differential: proof of Lemma \ref{homot}}
481:  
482:  Differentiating (\ref{dzt}) in $t$ yields (we denote 
483:  $\dot f$ the partial derivative in $t$ of a function $f$)
484:  \begin{equation}
485: \partial_{\bar z}\dot f-(\partial_z\circ\nu)\dot f=(\partial_z\circ\dot\nu)f.\label{2.2}
486: \end{equation}
487: where $\partial_z\circ\nu$ ($\partial_z\circ\dot\nu$) is the composition of the operator 
488: of the multiplication by the function $\nu$ (respectively, $\dot\nu$) and the 
489: operator $\partial_z$. 
490: Any solution $f$ of equation (\ref{2.2}) with the initial condition 
491: $f(z,0)\equiv1$ 
492: that vanishes identically on the torus for no value of $t$ is a one we 
493: are looking for. Let us show that (\ref{2.2}) is implied by a bounded 
494: linear differential equation in $L_2(\td)$. To do this, 
495: we use the following properties of the operators $\partial_z$ and $\partial_{\bar z}$. 
496:  
497:  \begin{remark} Denote $z=x_1+ix_2$, $x=(x_1,x_2)\in\rd$. 
498:  The operators $\partial_z$, $\partial_{\bar z}$ on $\td$ have common eigenfunctions 
499:  $e_n(x)=e^{i(n,x)}$, $n=(n_1,n_2)\in\mathbb Z^2$. The corresponding eigenvalues 
500:  (denote them $\lambda_n$ and $\lambda_n'$ respectively) have equal modules, 
501:  more precisely, 
502:  \begin{equation}\lambda_n'=-\overline{\lambda_n}.\label{2.3}\end{equation}
503:  This is implied by the fact that the operator $\partial_{\bar z}$ is conjugated to 
504:  $-\partial_z$ in the $L_2$ scalar product, which follows from definition. In fact, 
505:  $$\lambda_n=\frac i2(n_1-in_2),\ \lambda_n'=\frac i2(n_1+in_2).$$
506:  \end{remark}
507:  
508:  \begin{corollary} There exists a unique unitary operator $U:L_2(\td)\to 
509:  L_2(\td)$ preserving averages such that "$U=\partial_{\bar z}^{-1}\circ\partial_z$" 
510:  (more precisely, $U\circ\partial_{\bar z}=\partial_{\bar z}\circ U=\partial_z$). The operator 
511:  $U$ commutes with partial differentiations and extends up to a unitary operator 
512:  to any Hilbert Sobolev space of functions on $\td$. In particular, it preserves 
513:  the space of $\ci$ functions.
514:  \end{corollary}
515:  \begin{proof} 
516:  The operator $U$ from the Corollary is defined to have the previous 
517:  eigenfunctions $e_n$ with the eigenvalues $\frac{\lambda_n}{\lambda_n'}=
518:  \frac{n_1-in_2}{n_1+in_2}$. Its uniqueness follows immediately from the 
519:  previous operator equation on $U$ applied to the functions $e_n$. 
520:  The rest of the statements of the Corollary follow immediately 
521:  from definition and Sobolev embedding theorem (see [Ch], p.411). 
522:  \end{proof}
523:  
524:  Let us write down equation (\ref{2.2}) in terms of the new operator $U$. Applying 
525: the "operator" $\partial_{\bar z}^{-1}$ to (\ref{2.2}) and substituting 
526: $U=\partial_{\bar z}^{-1}\circ \partial_z$ yields  
527: $$(Id-U\circ\nu)\dot f=(U\circ\dot\nu)f.$$ 
528: This equation implies (\ref{2.2}). For any $t\in[0,1]$ the operator 
529: $Id-U\circ\nu$ in the left-hand side is invertible in $L_2(\td)$ and 
530: the norm of the 
531: inverse operator is bounded uniformly in $t$, since $U$ is unitary and the 
532: module  
533: $|\nu|$ is less than 1 and bounded away from 1 by compactness. 
534: Thus, the last equation can be rewritten as 
535: \begin{equation}
536: \dot f=(Id-U\circ\nu)^{-1}(U\circ\dot\nu)f,\label{2.6}
537: \end{equation}
538:  which is an ordinary differential equation in $f\in L_2(\td)$ with a 
539:  uniformly $L_2$- bounded operator in the right-hand side. 
540:   As it is shown below 
541: (in Proposition \ref{2.70}), the inverse $(Id-U\circ\nu)^{-1}$ is also uniformly 
542: bounded in each Hilbert Sobolev space $H^j(\td)$.  
543:  Therefore, equation (\ref{2.6}) written in arbitrary Hilbert Sobolev space has a 
544:  unique solution 
545:  with a given initial condition, in particular, with $f(z,0)\equiv1$ (the 
546:  theorem on existence and uniqueness of solution of ordinary 
547:  differential equation in Banach space with the right-hand side having 
548:  uniformly bounded derivative [Ch]). For any $t\in[0,1]$ this solution does 
549:  not vanish identically on $\td$ (uniqueness of solution) and belongs to 
550:  all the spaces $H^j(\td)$; hence, it is $C^{\infty}(\td)$ by Sobolev embedding 
551:  theorem (see [Ch], p.411). Thus, Lemma \ref{homot} is implied by the following
552:  
553:  \begin{proposition} \label{2.70} Let $x=(x_1,x_2)$ be affine coordinates on 
554:  $\mathbb R^2$, 
555:  $\td=\mathbb R^2\slash2\pi\mathbb Z^2$. 
556:  Let $s\geq0$, $s\in\mathbb Z$, $U$ be a linear operator 
557:  in the space of $C^{\infty}$ functions on $\td$ that commutes with the operators 
558:  $\frac{\partial}{\partial x_i}$, $i=1,2$, and extends to any Sobolev space 
559:  $H^j=H^j(\td)$, $0\leq j\leq s$, up to a unitary operator. Let $0<\delta<1$, 
560:  $\nu\in C^s(\td)$ be a complex-valued function, 
561:  $|\nu|\leq\delta$. The operator $Id-U\circ\nu$ is invertible and the 
562:  inverse operator 
563:  is bounded in all the spaces $H^j$, $0\leq j\leq s$. 
564: For any $0<\delta<1$, $j\leq s$, there 
565:  exists a constant $C>0$ (depending only on $\delta$ and $s$) such that for any 
566:  complex-valued  function $\nu\in C^s(\td)$ with $|\nu|\leq\delta$  
567:  $$||(Id-U\circ\nu)^{-1}||_{H^j}\leq C(1+\sum_{k\leq j}\max
568:  |\frac{\partial^k\nu}{\partial x_{i_1}, \dots,\partial x_{i_k}}|^j).$$
569:  \end{proposition}
570:  
571:  \begin{proof} Let us prove the Proposition for $s=1$. For higher $s$ its proof is 
572:  analogous. 
573:  \begin{equation}\text{By definition,}\ \ 
574: ||U\circ\nu||_{L_2}\leq\delta<1.\label{2.7}
575: \end{equation}
576:  %2.7
577:  Hence, the operator 
578:  $Id-U\circ\nu$ is invertible in $L_2=H^0$ and 
579: \begin{equation}
580: (Id-U\circ\nu)^{-1}=Id+\sum_{k=1}^{\infty}(U\circ\nu)^k:\label{2.8}
581: \end{equation}
582:  the sum of the $L_2$ operator norms of the sum entries in 
583:   (\ref{2.8}) is  finite by (\ref{2.7}). 
584:  Let us show that the operator in the right-hand side of (\ref{2.8}) is well-defined 
585:  and bounded in $H^1$. 
586:  To do this, it suffices to show that the sum of the operator $H^1$- norms of 
587:  the same entries is finite. 
588:  
589:  Let $f\in H^1(\td)$. Let us estimate 
590:  $||(U\circ\nu)^kf||_{H^1}$. We show that for any $k\in\mathbb N$ 
591: \begin{equation}
592: ||\frac{\partial}{\partial x_r}((U\circ\nu)^kf)||_{L_2}<ck\delta^{k-1}
593: ||f||_{H^1}, \ \ c=\delta+\max|\frac{\partial\nu}{\partial x_r}|,\ \ 
594: r=1,2.\label{2.9}
595: \end{equation}
596: 
597:  This will imply the finiteness of the operator $H^1$- norm of the sum in the 
598:  right-hand 
599:  side of (\ref{2.8}) and Proposition \ref{2.70} 
600:  (with $C=4\sum_{k\in\mathbb N}k\delta^{k-1}=
601:  \frac4{(1-\delta)^2}$).  
602: 
603: Let us prove (\ref{2.9}), e.g., for $r=1$. The derivative in the 
604: left-hand side of (\ref{2.9}) equals  
605: $$(U\circ\nu)^k\frac{\partial f}{\partial
606: x_1}+\sum_{i=1}^k(U\circ\nu)^{k-i}
607: \circ(U\circ\frac{\partial\nu}{\partial x_1})\circ(U\circ\nu)^{i-1}f$$
608: (since $U$ commutes with the partial differentiation by the condition of 
609: Proposition \ref{2.70}). The $L_2$- norm of the first term in the previous formula 
610: is no greater than 
611: $\delta^k||f||_{H^1}$ by (\ref{2.7}).  Each term in its sum has $L_2$- norm 
612: no greater 
613: than $\delta^{k-1}\max|\frac{\partial\nu}{\partial x_1}|||f||_{L_2}$ by (\ref{2.7}). 
614: This proves (\ref{2.9}). The Proposition is proved. Lemma \ref{homot} 
615: is proved. 
616: \end{proof}
617: \begin{remark} The solution of equation (\ref{2.6}) with the initial condition 
618:  $f|_{t=0}\equiv1$ admits the following formula: 
619:  \begin{equation}
620:  f(x,t)=(Id-U\circ\nu)^{-1}(1)=
621:  1+U(\nu)+(U\circ\nu\circ U)(\dot\nu)+\dots\label{2.10}
622:  \end{equation}
623:  Indeed, its right-hand side is a well defined $\ci$  
624:  family of $C^{\infty}$ functions on $\td$, which follows from the 
625:  uniform boundedness of the operators $(Id-U\circ\nu)^{-1}$ in any given 
626:  Hilbert Sobolev space. By definition, it satisfies  the unit initial 
627:  condition. Differentiating (\ref{2.10}) in $t$ yields
628:  $$(Id-U\circ\nu)^{-1}\circ (U\circ\dot\nu)\circ(Id-U\circ\nu)^{-1}(1)=
629: (Id-U\circ\nu)^{-1}\circ(U\circ\dot\nu)f(x,t).$$
630: Hence, the function (\ref{2.10}) satisfies (\ref{2.6}).  
631:  \end{remark}
632:  
633:  \def\wt#1{\widetilde#1}
634:  \def\tdv{\mathbb T^2}
635: 
636: \subsection{Zero of holomorphic differential. Proof of Proposition \ref{loc}} 
637:  Let us prove the existence of local holomorphic coordinate. Without loss of 
638:  generality we assume that $\mu(0)=0$ (applying a linear change of variables).  
639:  One can achieve also that $\mu$ is arbitrarily 
640:  small with derivatives of orders up to 3 applying a homothety and taking 
641:  the restriction to a smaller disc centered at 0. 
642:  We consider that the disc where $\mu$ is defined 
643: is embedded into $\td$ and extend the 
644:  function $\mu$ smoothly to $\td$. We assume that the extended function 
645:  satisfies the inequality $||\mu||_{C^3(\td)}<\delta$; one can make $\delta$ 
646:  arbitrarily small. 
647:  
648:  Let $\nu(x,t)=t\mu$, $f(x,t)$ be the corresponding function 
649:  family from Lemma \ref{homot} constructed as the solution of 
650: differential equation (\ref{2.6}) with unit initial condition,  $f(x)=f(x,1)$. 
651: We show in the next paragraph 
652:  that $f(0)\neq0$, if the previous constant $\delta$ is small enough. 
653:  Then the local coordinate we are looking for is the function 
654:  $$w(z)=\int_0^zf(dz+\mu d\bar z).$$
655:  Indeed, it is well-defined and holomorphic by definition. Its local univalence 
656:  follows from the nondegeneracy of its differential 
657:  $f(0)(dz+\mu d\bar z)$ at 0 (the inequalities $|\mu|<1$, $f(0)\neq0$). 
658:  
659:  Recall that by (\ref{2.10}),    
660: $$f(x,t)=(Id-tU\circ\mu)^{-1}(1), \ \text{where}\ 
661: U=(\partial_{\bar z})^{-1}
662:  \partial_z.$$ The functions $f(x,t)$ are equal to 1, if 
663:  $\mu=0$. Let us show that they are $C^0$- close to 1 (and hence, 
664: $f(0,1)\neq0$), whenever $\mu$ is small enough with 
665:  derivatives up to order 3. Consider the operator functional 
666:  $\mathcal A(\mu)=(Id-tU\circ\mu)^{-1}$: its value being an operator 
667:  acting in $H^3(\td)$ (it is well-defined, see Proposition \ref{2.70}). 
668:  As it will be shown in the next paragraph, it  
669:  depends continuously on small functional parameter $\mu\in C^3(\td)$, 
670:  $max|\mu|<1$, in the 
671:  $H^3(\td)$ operator norm, and moreover, it has a bounded derivative in $\mu$. 
672: Therefore, if $||\mu||_{C^3}$ is small enough, then  each function $f(x,t)$ 
673:  is close to 1 in $H^3$ (thus, in $C^0$, by Sobolev embedding theorem). 
674: 
675:  Now for the proof of Proposition \ref{loc} it suffices to prove 
676:  the boundedness of the previous derivative $\mathcal A'(\mu)$. 
677:  For any $0<\delta'<1$ the $\mathcal A(\mu)$ is uniformly bounded in all 
678:  $\mu$ with $||\mu||_{C^3}<\delta'$ 
679:  (Proposition \ref{2.70}), so, we can apply the usual formula for 
680:  the derivative of the inverse operator: the derivative of $\mathcal A(\mu)$ 
681:  along a vector $h\in C^3(\td)$ is equal to 
682:  $$\nabla_h\mathcal A(\mu)=\mathcal A(\mu)\circ U\circ h\circ\mathcal A(\mu).$$
683:  To prove the boundedness of the derivative, we have to show that 
684:  the $H^3$- norm of the operator in the right-hand side of the previous formula 
685:  is no greater than some constant (depending on $\mu$) times $||h||_{C^3}$. 
686:  Indeed, the previous $H^3$ operator norm is no greater 
687:  than $||\mathcal A(\mu)||_{H^3}^2$ times the $H^3$- norm of the operator 
688:  of multiplication by the function $h$, the latter is no greater than 
689:  $||h||_{C^3}$ times some universal constant. This proves 
690:  the boundedness of the derivative. 
691:  Proposition \ref{loc} is proved. The proof of Theorem \ref{td} is completed.
692:  
693:  
694:  \section{Quasiconformal mappings. Proof of Theorem \ref{c}}
695:  
696:  \subsection{The plan of the proof of Theorem \ref{c}}
697:  
698:  We have already proved the statement of Theorem \ref{c} for a 
699:  $\ci$ double-periodic almost complex structure on $\cc$  
700:  (i.e., a lifting to the universal cover $\cc$ 
701:  of a $\ci$ complex structure on $\td$). In this case the diffeomorphism 
702:  $\cc\to\cc$ from the Theorem is the lifting to the universal covers 
703:  of the diffeomorphism of the tori given by Theorem \ref{td}.
704:  To prove Theorem \ref{c} in the general case (let $\sigma$ be a given 
705:  (may be measurable) bounded complex structure on $\cc$) we 
706:  consider a sequence $\sigma_n$ of $\ci$ double-periodic complex 
707:  structures on $\cc$ with growing periods and 
708:  uniformly bounded dilatations (say less than a fixed $K>0$)  
709:  that converge to $\sigma$ almost everywhere. 
710:  For each $\sigma_n$ there is a normalized quasiconformal 
711:  diffeomorphism $\Phi_n:\cc\to\cc$ 
712:  transforming $\sigma_n$ to the standard complex structure. 
713:  We show that the diffeomorphisms $\Phi_n$ converge 
714:  (uniformly on $\overline{\cc}$) to a homeomorphism (denoted $\Phi$). 
715:  We will prove that $\Phi$ is a quasiconformal homeomorphism sending 
716:  $\sigma$ to the standard complex structure 
717:  (see the end of the Subsection). The uniqueness of a latter homeomorphism 
718:  and its diffeomorphic property  
719:  on a smoothness domain of $\sigma$ will be proved in 3.4. 
720:  Its analytic dependence on 
721:  parameter (the Addendum to Theorem \ref{c}) will be proved in 3.5.  
722:  
723:  We prove the convergence of $\Phi_n$ by equicontinuity  
724:  of the normalized $K$- homeomorphisms: 
725:  
726:  \begin{lemma} \label{norm} [Ah2]. For any $K>0$ the normalized 
727:  $K$- homeomorphisms $\cc\to\cc$ (see Definition 1.1) 
728:  are equicontinuous with their inverses as mappings of the Riemann sphere. 
729:  \end{lemma}
730:  Lemma \ref{norm} (proved in  3.2) together with Arzela-Ascoli 
731:  theorem imply the following 
732:  \begin{corollary} \label{normc} For any $K>0$ each sequence of 
733:  normalized $K$- homeomorphisms $\cc\to\cc$ 
734:  contains a subsequence converging to a homeomorphism 
735:  $\overline{\cc}\to\overline{\cc}$ uniformly on $\overline{\cc}$. 
736: \end{corollary}
737: 
738: \begin{lemma} \label{limqc} [Ah2]. Let $K>0$, $U\subset\cc$ be a domain 
739: (that may be the whole $\cc$) $\Phi_n:U\to\Phi_n(U)\subset\cc$ be a sequence 
740:  of $K$- homeomorphisms converging uniformly on compact subsets to a 
741:  homeomorphism (denote $\Phi$ the limit). 
742:  Let $\sigma_n$ be the almost complex structures sent to the standard one by 
743:  $\Phi_n$. Let $\sigma_n$ converge almost everywhere (denote $\sigma$ 
744:  their limit). Then $\Phi$ is a $K$- homeomorphism  
745:  sending $\sigma$ to the standard complex structure. 
746: \end{lemma}
747: 
748: Lemma \ref{limqc} will be proved in Subsection 3.3 (using Lemma \ref{norm} and 
749: Corollary \ref{normc}).
750: 
751: 
752: \begin{proof} {\bf of existence in Theorem \ref{c} modulo Lemmas 
753: \ref{norm} and \ref{limqc}.} Let $\sigma_n$, $\sigma$, $K$, $\Phi_n$ be as at 
754: the beginning of the Section. Then $\Phi_n$ are $K$- 
755: diffeomorphisms. Passing to a subsequence, one can achieve that $\Phi_n$ 
756: converge to a homeomorphism (Corollary \ref{normc}, 
757: denote $\Phi$ the limit homeomorphism).  By Lemma \ref{limqc}, $\Phi$ is a $K$- 
758: homeomorphism 
759: transforming $\sigma$ to the standard complex structure. Theorem \ref{c} is proved. 
760: \end{proof}
761: 
762: \begin{remark} In the proof of the existence  in Theorem \ref{c} 
763: we had used only the statements of the previous 
764: Lemmas for $\ci$ diffeomorphisms. Their statements  
765: for general quasiconformal homeomorphisms will be used in the proof of 
766: the uniqueness in Theorem \ref{c} (Subsection 3.4). 
767: \end{remark} 
768: 
769: 
770: \subsection{Normality. Proof of Lemma \ref{norm}}
771: The proof of Lemma \ref{norm} is based on the Gr\"otzsch inequality (the 
772: next Lemma) comparing moduli of $K$- homeomorphic complex annuli. To state it, 
773: let us firstly recall the following
774: 
775: \begin{definition} see [Ah2]. The {\it modulus} of an annulus 
776: $A=\{ r<|z|<1\}$ is $m(A)=-\frac1{2\pi}\ln r$. 
777: \end{definition}
778: \begin{remark} Consider the cylinder $\mathbb R\times S^1$ with the 
779: coordinates $(x,\phi)$, $S^1=\mathbb R\slash2\pi\mathbb Z$, and the standard 
780: complex structure, which is induced by the Euclidean metric 
781: $dx^2+d\phi^2$.  
782: \begin{equation}\text{For any}\ R>0 \ \text{put}\ A(R)=\{ 0<x<R\};\ \text{then}\ 
783: m(A(R))=\frac R{2\pi}.\label{ann}\end{equation}
784: The modulus of an annulus is invariant under conformal mappings [Ah2]. 
785: \end{remark} 
786: \begin{lemma} \label{grl} (Gr\"otzsch, see [Ah2]). Let $K>0$, 
787: $f:A_1\to A_2$ be a $K$- homeomorphism of complex annuli. Then 
788: \begin{equation} m(A_2)\geq K^{-1}m(A_1).\label{gri}\end{equation}
789: \end{lemma}
790: \begin{proof} For completeness of presentation, we give the classical proof 
791: of the Lemma. Firstly we prove the Lemma for a $K$- diffeomorphism; the general 
792: case is treated analogously (see the end of the 
793: proof). Let us consider that the annuli are drawn on the previous cylinder, 
794: say, $A_1=A(R_1)$, $A_2=A(R_2)$, then $m(A_i)=\frac{R_i}{2\pi}$, $i=1,2$, 
795: see (\ref{ann}).  Thus, it suffices to show that $R_2\geq K^{-1}R_1$. To do this, 
796: consider 
797:  the pullback (denoted $g$) to $A_1$ under $f$ of the Euclidean metric of $A_2$ 
798:  (denote $|\ |_g$ ($Area_g$) the corresponding norm of vector 
799:  fields on $A_1$ (respectively, the area), $Area$ being the Euclidean area). 
800:  One has $Area(A_i)=2\pi R_i$, $Area(A_2)=Area_g(A_1)$. We show that
801:  \begin{equation} Area_g(A_1)\geq K^{-1} Area(A_1).
802:  \label{area}\end{equation}
803:  This together with the previous formulas for the areas will prove the Lemma. 
804:  For the proof of (\ref{area}) we consider the family 
805:  $A(r)=\{ 0<x<r\}\subset A_1$ of subannuli in $A_1$, $r\leq R_1$, and prove 
806:  a lower bound of the derivative $(Area_g(A(r))'_r$. To do this, consider 
807:  the vector field $\frac{\partial}{\partial x}$ as the sum of its 
808:  component tangent to the circles $x=const$ and the $g$- orthogonal component 
809:  
810:  \begin{tabular}{ll}
811: \begin{minipage}{10em}
812: \begin{center}
813: \hskip-0.5cm
814: \epsfbox{figuq1.eps}
815: \end{center}
816: \end{minipage}
817: \begin{minipage}{27em} 
818: (denote the latter component normal to the circles by $n$, see Fig.1). 
819: The vector field $n$ has the 
820:  same projection to the $x$- axis, as $\frac{\partial}{\partial x}$ and 
821:  its flow leaves invariant the fibration by circles $x=const$: its time $t$ 
822:  flow map transforms $A(r)$ to $A(r+t)$. Therefore, 
823: \end{minipage}
824: \end{tabular}
825:  \def\dpf{\frac{\partial}{\partial\phi}}
826:  \def\dpx{\frac{\partial}{\partial x}}
827:  \begin{equation}(Area_g(A(r)))'_r=\int_{x=r,\phi\in[0,2\pi]} 
828:  |\dpf|_g|n|_gd\phi.\label{arder}\end{equation}
829:   One has $|n|_g\geq K^{-1}|\dpf|_g$. Indeed, the $g$- norm $|\ |_g$ of a 
830:  vector tangent to $A_1$ 
831:  is equal to the standard Euclidean norm $|\ |$ of its image under $f$: 
832:  $|n|_g=|f_*n|$, $|\dpf|_g=|f_*\dpf|$. By definition,  
833:  $|\dpf|=1$, $|n|\geq|\frac{\partial}{\partial x}|=1=|\dpf|$. Therefore, 
834:  by the $K$- quasiconformality of $f$ (see, Definition 1.1), 
835:  $|n|_g=|f_*n|\geq K^{-1}|f_*\dpf|=K^{-1}|\dpf|_g$. Hence, 
836: the previous derivative is no less than 
837: $$K^{-1}\int_{x=r,\phi\in[0,2\pi]}|\dpf|_g^2d\phi\geq K^{-1}(2\pi)^{-1}
838: (\int_{\phi\in[0,2\pi]}|\dpf|_gd\phi)^2$$
839: (Cauchy-Bouniakovskii-Schwarz inequality). The latter integral is no less than 
840: $2\pi$. Indeed, it is equal to 
841: the length in the metric $g$ of the circle $x=r$, or in other terms, 
842:  the Euclidean length of its image under $f$, which 
843: is a closed curve in $A_2$ isotopic to a circle $x=const$. Therefore, 
844: $(Area_g(A(r)))'_r\geq2\pi K^{-1}$, thus,  
845: $Area_g(A_1)\geq 2\pi K^{-1}R_1=K^{-1}Area(A_1)$. 
846: This proves (\ref{area}) and the Lemma for a $K$- diffeomorphism $f$. 
847: In the case, when $f$ is a $K$- homeomorphism, thus just having  
848: local $L_2$ derivatives, the previous discussion remains valid: the 
849: previous integrals are well-defined for almost all $r$, since the subintegral 
850: expression $|\dpf|_g|n|_g$ in (\ref{arder}) is bounded from above by 
851: $||df(r,\phi)||^2$ times a constant depending on $K$. This follows 
852: from definition and the uniform boundedness of the Euclidean norm $|n|$:  
853: by definition,  $n$ is projected to the vector field 
854: $\frac{\partial}{\partial x}$ with unit norm; the  
855: angle between $n$ and a circle $x=const$ is bounded from below by a 
856: constant depending on $K$ (quasiconformality). Lemma \ref{grl} is proved. 
857:  \end{proof}
858:  
859:  \def\var{\varepsilon}
860: 
861: To prove Lemma \ref{norm}, we need to show that close points 
862: cannot be mapped to distant points under a normalized K- homeomorphism or its 
863: inverse. This is proved by comparing moduli of appropriate annuli with 
864: those of their images (using Lemma \ref{grl}). 
865: 
866: For the proof of Lemma \ref{norm} we recall the notion of the Poincar\'e metric 
867: [CG]. The Poincar\'e metric of the unit disc $|z|<1$ is 
868: $\frac{4|dz|^2}{(1-|z|^2)^2}$ (it is invariant under its conformal 
869: automorphisms). A Riemann surface is {\it hyperbolic}, 
870: if its universal covering is conformally equivalent to the unit disc 
871: (see Theorem \ref{pk}), e.g., any domain in $\cc$ whose complement contains 
872: more than one point. 
873: The Poincar\'e metric of a hyperbolic Riemann surface is the pushforward 
874: of the Poincar\'e metric of the unit disc under the universal covering. 
875: 
876: \begin{remark} \label{poim} (see [CG]). The Poincar\'e metric is well-defined, complete 
877: and decreasing: the Poincar\'e metric of 
878: a subdomain of a hyperbolic Riemann surface is greater than that of the ambient 
879: surface. The Poincar\'e metric of $\cc\setminus\{0,1\}$ is greater than 
880: its standard spherical metric times a constant. 
881: \end{remark}
882: 
883: In the proof of Lemma \ref{norm} we use the following relation of modulus 
884: of an annulus and its Poincar\'e metric, whose proof is a straightforward 
885: calculation.
886: 
887: \begin{proposition} \label{geod} see, e.g., [DH]. The modulus of an annulus is equal to 
888:  $\pi$ times the inverse of the length of its closed geodesic. 
889: \end{proposition}
890: 
891: Let us prove the equicontinuity of normalized $K$- homeomorphisms  
892: by contradiction. Suppose the contrary, i.e., 
893: there exist an $\var>0$, a sequence of normalized 
894: $K$- homeomorphisms $\Phi_n:\cc\to\cc$ and a sequence of pairs 
895: $x_n,y_n\in\cc$, $|x_n-y_n|\to0$, $|\Phi_n(x_n)-\Phi_n(y_n)|>\var$ 
896: (in the spherical metric of $\overline{\cc}$). Without loss of 
897: generality we assume that the sequence $x_n$ (and hence, $y_n$) converges 
898: (one can achieve this by passing to a subsequence, denote $x$ the limit). 
899: Then there is a sequence $A_n$ of annuli in 
900: $\overline{\mathbb C}\setminus\{0,1,\infty\}$ bounded by circles 
901: centered at $x$ and surrounding the pairs $x_n$, $y_n$: one of the circles 
902: is fixed, the other one contracts to $x$, as $n\to\infty$ see Fig.2a. 
903:  By definition, the annuli $A_n$ 
904: tend to once punctured disc, hence, 
905: $m(A_n)\to\infty$. The point $x$ may coincide with some of the 
906: three points 0, 1, $\infty$. Let us take two of the latters that 
907: are distinct from $x$ (say, let them be 0, 1). Then each annulus $A_n$ 
908: separates the pairs $(x_n,y_n)$ and $(0,1)$.  
909: By Lemma \ref{grl}, $m(\Phi_n(A_n))\to\infty$ as well. Hence, by  Proposition 
910: \ref{geod}, the 
911: lengths of the geodesics (denoted by $\gamma_n$) of the annuli $\Phi_n(A_n)$ 
912: in their Poincar\'e metrics tend to zero.
913:  But the latter lengths are greater than the lengths of $\gamma_n$ 
914: taken in the Poincar\'e metric of $\cc\setminus\{0,1\}$, and hence, 
915: also greater than their lengths  in the spherical metric times a constant independent 
916: from $n$ (by the previous Remark). Thus, each $\gamma_n$ separates the pairs 
917: $(\Phi_n(x_n), \Phi_n(y_n))$ and $(0,1)$ and is a closed curve with spherical 
918: length tending to 0. Hence, the spherical distance between 
919: $\Phi_n(x_n)$ and $\Phi_n(y_n)$ tends to 0 - a contradiction. 
920: \medskip
921: 
922: \begin{tabular}{ll}
923: \begin{minipage}{15em}
924: \begin{center}
925: \epsfbox{figuq2.eps}
926: \end{center}
927: \end{minipage} 
928: %\begin{minipage}{20em}
929: %\end{minipage}
930: \end{tabular}
931: \medskip
932: 
933: Now let us prove that the inverses to the normalized $K$- homeomorphisms 
934: are also equicontinuous by contradiction, analogously to the previous 
935: discussion. Suppose the contrary:  
936: there exist an $\var>0$, a sequence of normalized 
937: $K$- homeomorphisms $\Phi_n:\cc\to\cc$ and a sequence of pairs 
938: $x_n,y_n\in\cc$, $|x_n-y_n|\to0$, $|\Phi_n^{-1}(x_n)-\Phi_n^{-1}(y_n)|>\var$ 
939: (in the spherical metric of $\overline{\cc}$). Without loss 
940: of generality we assume that all the 
941: sequences $x_n$, $y_n$, $\Phi_n^{-1}(x_n)$, $\Phi_n^{-1}(y_n)$ converge 
942: (one can achieve this by passing to a subsequence); denote their 
943: limits by $x$, $y$, $\tilde x$, $\tilde y$ respectively. 
944: By definition, $x=y$, $\tilde x\neq\tilde y$. 
945: Firstly consider the case, when $x\neq0,1,\infty$. Then $\tilde x,\tilde y
946: \neq0,1,\infty$ 
947: as well: otherwise $\Phi_n^{-1}(x_n)$, $\Phi_n^{-1}(y_n)$ would accumulate 
948: to $\{0,1,\infty\}$, while their $\Phi_n$- images $x_n$, $y_n$ would not - 
949: a contradiction to the equicontinuity of the $\Phi_n$'s (already proved). 
950: Fix an annulus $A$ separating the pair $0,\tilde x$ and the triple 
951:  $1,\tilde y,\infty$ (we assume that its closure is disjoint from 
952: $\Phi_n^{-1}(x_n)$, $\Phi_n^{-1}(y_n)$ for any $n$). 
953: Its images $\Phi_n(A)$ are disjoint from $0,1,x_n,y_n$ and 
954: have moduli bounded away from zero (Lemma \ref{grl}), 
955: and hence, closed geodesics (denoted $\gamma_n$, see Fig.2b) of 
956: uniformly bounded lengths. 
957: Thus, the lengths of $\gamma_n$  in the Poincar\'e metric of 
958: $\cc\setminus\{0,1,x_n,y_n\}$ are also uniformly bounded. On the other hand, 
959: $\gamma_n$ separates the pair $(0, x_n)$ and the triple $(1,y_n,\infty)$ 
960: for any $n$, see Fig.2b. The points 0, $x_n$ in the first pair are distant 
961: ($x=\lim x_n\neq0$), thus, 
962: the spherical length of $\gamma_n$ is bounded from below. 
963: The points $x_n$, $y_n$, which are separated by $\gamma_n$, 
964: collide towards $x$, so, $\gamma_n$ comes arbitrarily close to $x_n$, as 
965: $n\to\infty$. This implies that $\gamma_n$ has length tending to infinity 
966: in the Poincar\'e metric of $\cc\setminus\{0,x_n\}$, and hence, in 
967: the Poincar\'e metric of $\cc\setminus\{0,1,x_n,y_n\}$. If   
968: $x_n$ does not move while $n$ changes, this  follows from the 
969: completeness of the Poincar\'e metric (Remark \ref{poim}). The case, when 
970: $x_n\not\equiv const$, is reduced to the previous one by 
971: applying the variable change $w=\frac z{x_n}$. This contradicts to the 
972: previous statement saying that the latter Poincar\'e length of $\gamma_n$ 
973: is uniformly bounded. 
974: 
975: Now let $x\in\{0,1,\infty\}$, say, $x=1$. 
976: Then $\tilde x,\tilde y\neq0,\infty$, as before, and at least one of 
977: $\tilde x\neq\tilde y$ (say, $\tilde x$) is distinct from 1. In these 
978: notations we repeat the previous argument. Lemma \ref{norm} is proved. 
979: 
980: \subsection{Quasiconformality and weak convergence. Proof of Lemma \ref{limqc}}
981: Let $\Phi_n$, $\sigma_n$, $\Phi$, $\sigma$ be as in Lemma \ref{limqc}. 
982: Recall that 
983: the dilatations of the $\sigma_n$' s are no greater than $K$, as are 
984: those of the $\Phi_n$' s, hence, the same is true for $\sigma$. Let us show 
985: that $\Phi$ is quasiconformal, more precisely: 1) has local $L_2$ derivatives 
986: that are weak $L_2$ limits of those of $\Phi_n$; 2) transforms $\sigma$ to the 
987: standard complex structure (and hence, is $K$- quasiconformal). 
988: This will prove Lemma \ref{limqc}.
989: 
990: For the proof of statement 1) we use the fact that the norms of 
991: the  differentials $d\Phi_n$ (in the spherical metric of $\overline{\cc}$) 
992: are uniformly bounded in each space $L_2(D)$, $D\Subset\cc$. 
993: Indeed, on each disc $D\Subset\cc$ 
994: $||d\Phi_n||^2_{L_2(D)}\leq K (Area(\Phi_n(D))$, which 
995: follows from definition and $K$- quasiconformality (the areas are taken  
996: in the spherical metric). The latter areas  
997: converge to $Area(\Phi(D))$, hence, they are uniformly bounded, and so are 
998: the previous $L_2$- norms. 
999: 
1000: Thus, the derivatives are locally $L_2$- bounded, hence, passing to a 
1001: subsequence one can achieve that they converge $L_2$- weakly. On the other 
1002: hand, they converge to the 
1003: derivative of $\Phi$ in sense of distributions. Therefore,  the latter is 
1004: also $L_2$ locally and the convergence is $L_2$- weak. Statement 1) is proved. 
1005: 
1006: Let $\mu_n$, $\mu$ be the functions 
1007: from (\ref{1.1}) defining the complex structures $\sigma_n$ and $\sigma$ 
1008: respectively, thus, $d\Phi_n=f_n(dz+\mu_nd\bar z)$. By assumption, 
1009: $|\mu_n|<1$, $\mu_n\to\mu$ almost everywhere. We have to show that 
1010: $\frac{\partial \Phi}{\partial\bar z}=\mu\frac{\partial\Phi}{\partial z}$. 
1011:  Indeed,  $f_n\to f=\frac{\partial\Phi}{\partial z}$, 
1012: $f_n\mu_n\to\frac{\partial\Phi}{\partial\bar z}$  (both $L_2$ weakly), 
1013: as $n\to\infty$. Since, $f_n$ are uniformly bounded in a local space $L_2$ 
1014: and weakly converge, $\mu_n$ are uniformly bounded and converge 
1015: almost everywhere, the weak limit of their product is the product 
1016: $f\mu$ of their limits. This proves the previous partial differential 
1017: equation on $\Phi$ together with statement 2) and Lemma \ref{limqc}. 
1018: \subsection{Uniqueness, smoothness and group property}
1019: 
1020: Here we prove the uniqueness of the normalized 
1021: quasiconformal homeomorphism from Theorem \ref{c} and the group 
1022: and measure  
1023: properties of quasiconformal mappings (Propositions \ref{group} 
1024: and \ref{measure0}).  
1025: The uniqueness follows from the local uniqueness up to composition with 
1026: a conformal mapping and from normalizedness. The local uniqueness 
1027: (together with the diffeomorphic property on a smoothness domain of 
1028: the complex structure) are implied by the following 
1029: 
1030: \begin{proposition} \label{uniq} Let $D\subset\cc$ be a simply-connected 
1031: domain, $\sigma$ be a bounded measurable  almost complex structure on $D$, 
1032: $\Phi:D\to\Phi(D)\subset\cc$ be a quasiconformal homeomorphism transforming 
1033: $\sigma$ to the standard complex structure. Then $\Phi$ is unique up to 
1034: left composition with a conformal mapping. It is a $\ci$ diffeomorphism, 
1035: if $\sigma$ is $\ci$. 
1036: \end{proposition}
1037: 
1038: \begin{proof}  
1039: Let $\mu:D\to\cc$ be the function defining the almost 
1040: complex structure $\sigma$. 
1041: 
1042: {\bf Case $\mu\equiv0$.} Then $\frac{\partial \Phi}{\partial\bar z}=0$ and 
1043: $\Phi$ has local $L_2$ derivatives. Let us show that $\Phi$ is conformal. 
1044: Fix a $z_0\in D$ and put 
1045: $U(z)=\int_{z_0}^z\Phi(\zeta)d\zeta$. We show that the function $U(z)$   
1046: is well-defined (independent on the choice of path connecting $z_0$ 
1047: to $z$). Then it is holomorphic by definition, hence, so is $\Phi(z)=
1048: \frac{\partial U}{\partial z}$. It suffices to show that the integral 
1049: of the form $\Phi dz$ along any Jordan curve is zero. Since the derivatives 
1050: of $\Phi$ are locally $L_2$, we can apply the Stokes formula:  
1051: the previous integral is equal to the integral of 
1052: the differential $d(\Phi dz)$ over the domain bounded 
1053: by the curve. But $d(\Phi dz)=\frac{\partial\Phi}{\partial\bar z}d\bar z dz=0$, 
1054: so, it is zero.  
1055: 
1056: {\bf Case $\mu\in\ci$.} There exists at least one $\ci$ quasiconformal 
1057: diffeomorphism $\Psi$ transforming $\sigma$ to the standard complex 
1058: structure (Theorem \ref{td}, see also the discussion in 
1059: Section 1.2). The composition $\Phi\circ\Psi^{-1}$ preserves 
1060: the standard complex structure by definition and is quasiconformal: 
1061: it has local $L_2$ derivatives, since so does $\Phi$ and $\Psi^{-1}$ is $\ci$.  
1062: Therefore, it is conformal, as is proved above, hence, $\Phi$ is a $\ci$ 
1063: diffeomorphism. 
1064: 
1065: {\bf Case $\mu$ is measurable.} 
1066: Let $0<\delta<1$, $|\mu|<\delta$, $\mu_n$ be a sequence of $\ci$ functions, 
1067: $|\mu_n|<\delta$, $\mu_n\to\mu$ almost everywhere (we extend 
1068: $\mu_n$, $\mu$ to $\cc$ with the latter 
1069: inequality and convergence). Consider the corresponding 
1070: almost complex structures $\sigma_{\mu_n}$, see (\ref{1.1}), and 
1071: the quasiconformal diffeomorphisms (denoted $\Phi_n$) from 
1072: Theorem \ref{c} (the latters exist as is proved above). 
1073: Passing to subsequence, one can assume that they converge uniformly 
1074: on $\overline{\cc}$ (by Lemma \ref{norm}). Denote $\Psi$ their limit, 
1075: which is a quasiconformal homeomorphism transforming the  extended 
1076: complex structure $\sigma$ to the standard one (Lemma \ref{limqc}). 
1077: It suffices to show that 
1078: $\Phi\circ\Psi^{-1}:\Psi(D)\to\Phi(D)$ is a conformal homeomorphism. 
1079: It is a homeomorphism, since so are $\Phi$ and $\Psi$, and preserves 
1080: the standard complex structure, thus, if we show that it is 
1081: quasiconformal, this will imply conformality 
1082: (as is proved in the previous case $\mu\equiv0$). 
1083:  To do this, 
1084: consider the homeomorphisms $h_n=\Phi\circ\Phi_n^{-1}:\Phi_n(D)\to\Phi(D)$. 
1085: They are quasiconformal homeomorphisms 
1086: with uniformly bounded dilatations, as in the previous paragraph. They converge 
1087: to $\Phi\circ\Psi^{-1}$ uniformly on compact subsets of $\Psi(D)$. The 
1088: corresponding pullbacks of the standard complex structure converge 
1089: to the latter almost everywhere, which follows from definition and 
1090: convergence $\mu_n\to\mu$. Hence, by Lemma \ref{limqc}, the limit is 
1091: quasiconformal. Proposition \ref{uniq} is proved. Theorem \ref{c} is proved.
1092: \end{proof}
1093: 
1094: \begin{proof} {\bf of Proposition \ref{group}.} The statement of the 
1095: Proposition is local: it suffices to show that compositions 
1096: (inverses) of local $K$- homeomorphisms are $K^2$- (respectively, 
1097: $K$-) quasiconformal. We prove this statement for composition 
1098: (for inverse the proof is analogous): given domains $U,V,W\subset\cc$ 
1099: and $K$- homeomorphisms $\Psi:U\to V$, $\Phi:V\to W$, let us show that 
1100: $\Phi\circ\Psi$ is a $K^2$- homeomorphism. 
1101: By Remark \ref{grr} the previous statement holds 
1102: for diffeomorphisms, and in the general case the dilatation of the 
1103: composition is no greater than $K^2$, thus, to prove the quasiconformality 
1104: means to show that the composition has local $L_2$ derivatives. To do this, 
1105: consider the pullback $\sigma(\Phi)$ of the standard complex 
1106: structure under $\Phi$. Let us extend it to $\cc$ without increasing the 
1107: dilatation and construct a sequence $\sigma(\Phi_n)$ of $\ci$ almost complex 
1108: structures on $\cc$ converging to $\sigma(\Phi)$ 
1109: almost everywhere with dilatations 
1110: no greater than $K$. Let $\Phi_n:\cc\to\cc$ be the corresponding 
1111: normalized quasiconformal homeomorphisms (which are $K$- homeomorphisms) from 
1112: Theorem \ref{c}. They are $\ci$ diffeomorphisms as is proved above. 
1113: By Lemma \ref{limqc}, they converge uniformly on $overline{\cc}$ to a 
1114: $K$- homeomorphism $\wt{\Phi}:\cc\to\cc$ transforming 
1115:  $\sigma(\Phi)$ to the standard complex structure. 
1116:  By the previous Proposition, 
1117:  $\wt{\Phi}=\Phi$ up to left composition with a conformal mapping. 
1118:  Now the compositions $\Phi_n\circ\Psi$ are $K^2$- homeomorphisms 
1119:  (since $\Phi_n$ is $\ci$) converging to $\wt{\Phi}\circ\Psi$, and the 
1120:  corresponding pullbacks of the standard complex structure converge also. Hence, 
1121:  by Lemma \ref{limqc}, the limit is quasiconformal. Since the limit coincides with 
1122:  $\Phi\circ\Psi$ up to composition with a conformal mapping, the latter is 
1123:  quasiconformal too. Proposition \ref{group} is proved.
1124: \end{proof} 
1125: 
1126: \begin{proof} {\bf of Proposition \ref{measure0}.} The statement 
1127: of Proposition \ref{measure0} is local and is reduced to the 
1128: case of quasiconformal homeomorphisms $\overline{\cc}\to\overline{\cc}$, as 
1129: Proposition \ref{group} proved above. We prove it 
1130: by contradiction. Suppose the contrary: some quasiconformal 
1131: homeomorphism $h$ of the Riemann sphere sends a zero measure set $S$ 
1132: to a posivite measure set $h(S)$ (without loss of generality  we assume 
1133: that $h$ fixes 0, 1 and $\infty$). 
1134: Let $\sigma$ be the pull-back under $h$ 
1135: of the standard complex structure  (it is 
1136: well-defined almost everywhere). Then $h$ is the unique normalized 
1137: quasiconformal homeomorphism transforming $\sigma$ to the standard 
1138: complex structure. Let us change the standard structure in the image as 
1139: follows: on $h(S)$ we change it 
1140: to some constant nonstandard almost complex structure; on the rest we 
1141: keep it standard. Denote  $\sigma'$ the almost 
1142: complex structure thus obtained on the Riemann sphere in the image. 
1143: By Theorem \ref{c}, there exists a unique normalized 
1144: quasiconformal homeomorphism 
1145: $H$ transforming $\sigma'$ to the standard complex  structure. 
1146: One has $H\not\equiv Id$, since the set $h(S)$ has a positive measure. 
1147: By definition and Proposition \ref{group}, the composition 
1148: $H\circ h$, which is different from $h$, 
1149: is a normalized quasiconformal homeomorphism transforming 
1150: $\sigma$ to the standard complex structure. This contradicts the 
1151: uniqueness of $h$. Proposition \ref{measure0} is proved. 
1152: \end{proof}
1153: 
1154: 
1155: \subsection{Analytic dependence on parameter. Proof of the Addendum}
1156: 
1157: {\bf Double-periodic case.} Consider a family of double-periodic $\ci$ 
1158: almost complex 
1159: structures $\sigma(t)$ on $\cc$ depending holomorphically 
1160: on a complex parameter $t$ 
1161: (this means that the corresponding function $\mu=\mu(z,t)$ from (\ref{1.1}) is 
1162: holomorphic in $t$). We assume that the periods are fixed, thus, $\sigma(t)$ 
1163: are the lifting to the universal cover $\cc$ of an analytic family of 
1164: almost complex structures on the two-torus. 
1165: Then the corresponding quasiconformal diffeomorphisms (denoted 
1166: $\Phi_t$) from Theorem \ref{c} are holomorphic in $t$ as well. Indeed, 
1167: their differentials are uniformizing differentials. Hence, for any $t$, 
1168: $d\Phi_t=f_t(dz+\mu(z,t)d\bar z)$ up to multiplication by complex constant 
1169: depending on $t$, where $f_t$ is given by formula (\ref{2.10}). The right-hand 
1170: side of (\ref{2.10}) is analytic in the functional parameter $\mu$, hence, 
1171: $f_t$ is holomorphic in $t$ and $z\mapsto\int_0^zf_t(dz+\mu(z,t)d\bar z)$ 
1172: is a holomorphic family of diffeomorphisms of $\cc$. The family $\Phi_t$ is 
1173: obtained from the latter by multiplication by a function in $t$ that 
1174: makes the previous diffeomorphisms normalized (fixing 1), hence, the 
1175: multiplier function (and thus, $\Phi_t$ as well) are also holomorphic in $t$. 
1176: The Addendum is proved in the double-periodic case. 
1177: 
1178: {\bf General case.} 
1179: Now consider arbitrary analytic family $\sigma(t)$ of bounded 
1180: almost complex structures 
1181: on $\cc$ depending on a complex parameter $t$ (we suppose that $t$ runs through 
1182: the unit disc $D$). Let $\mu(z,t)$ be the corresponding functions, 
1183: see (\ref{1.1}), which are holomorphic in $t$. Then there exists a 
1184: $0<\delta<1$ such that $|\mu(z,0)|<\delta$ for any $z$. The  
1185: corresponding mapping $M_z:t\mapsto\mu(z,t)$ is a holomorphic mapping 
1186: $D\to D$ depending on $z$ in a measurable way such that $|M_z(0)|<\delta$. 
1187:  (Recall that for a given $\delta<1$ the space of holomorphic mappings 
1188:  $M:D\to D$ with $|M(0)|<\delta$ is compact, see [CG].) 
1189:  Vice versa, for any $0<\delta<1$ each measurable collection of holomorphic 
1190:  mappings $M_z:D\to D$ with $|M_z(0)|<\delta$ 
1191: defines an analytic family of bounded almost complex structures; they 
1192: are uniformly bounded when restricted to a smaller  
1193: parameter disc $D_r=\{|t|<r\}$, $r<1$. Indeed, in the case, when $M_z(0)\equiv0$, 
1194: $|M_z|_{D_r}<r$ (Schwarz Lemma); the general case is easily reduced to the 
1195: previous one.  
1196:  
1197: Denote $\Phi_t$ the corresponding 
1198: normalized quasiconformal homeomorphisms from Theorem \ref{c}.  
1199: To prove the analyticity of $\Phi_t$ in $t$, we approximate $\sigma(t)$ 
1200: (in the sense of convergence almost everywhere) 
1201: by  analytic families $\sigma_n(t)$ of $\ci$ double-periodic almost 
1202: complex structures depending holomorphically on the same parameter $t$ 
1203: with growing periods $2n$, $2in$, 
1204: $\sigma_n\to\sigma$. (For example, consider the restriction of $\sigma$ 
1205: to the period square centered at 0 and take $\sigma_n$ to be its  
1206: double-periodic  extension. Then approximate the new double-periodic  
1207: family $M_z$ by a $\ci$ family of holomorphic mappings $D\to D$.) One can 
1208: do this in such a way that $\sigma_n(t)|_{t\in D_r}$ be uniformly bounded.  
1209: Denote $\Phi_{n,t}$ the normalized 
1210: quasiconformal homeomorphisms transforming $\sigma_n(t)$ to the standard 
1211: complex structure. They depend analytically on $t$, as is proved above. 
1212: By Lemma \ref{limqc}, for any $t$, $z$, 
1213:  $\Phi_{n,t}(z)\to\Phi_t(z)$, as $n\to\infty$. 
1214: Thus, $\Phi_t(z)$ is a function in $t$ that 
1215: is a limit of pointwise converging sequence of holomorphic functions. 
1216: Let us prove that for any fixed $z$ the functions $\Phi_{n,t}(z)$ in $t\in D_r$ 
1217: are bounded uniformly in $n$: then their limit $\Phi_t(z)$ is holomorphic. 
1218: Indeed,  the almost complex structures $\sigma_n(t)|_{t\in D_r}$ 
1219: are uniformly bounded. Therefore, the family $\Phi_{n,t}$ (depending 
1220: on the two parameters $n$ and $t\in D_r$) together with their inverses 
1221: is equicontinuous (Lemma \ref{norm}). Hence, the previous functions are 
1222: uniformly bounded, so, their limit is holomorphic. The Addendum is proved. 
1223: 
1224: \section{Acknowledgements} 
1225: In late 1990-ths \'E.Ghys stated a question concerning a foliated version of 
1226: Theorem \ref{c} for linear foliations of tori. 
1227: The proof of Theorem \ref{td} presented in the paper was obtained 
1228: as a by-product of the author's solution to his question, and I wish to 
1229: thank him.  
1230: I wish also to thank him and  J.-P.Otal, \'E.Giroux for helpful discussions.
1231: The research was supported by part 
1232: by CRDF grant  RM1-2358-MO-02 and by 
1233: Russian Foundation for Basic Research (RFFI) grant 02-02-00482. 
1234: 
1235: \section{References}
1236: 
1237: [Ab] Abikoff, W. Real analytic theory of Teichm\"uller space, - Lect. Notes in  
1238: Math., 820, Springer-Verlag (1980). 
1239: 
1240: [Ah1] Ahlfors, L. Conformality with respect to Riemannian metrics. - 
1241: Ann. Acad. Sci. Fenn. Ser. A. I. 1955 (1955), no. 206, 22 pp.
1242:  
1243: %2. 
1244: [Ah2] Ahlfors, L. Lectures on quasiconformal mappings, - Wadsworth (1987).  
1245: 
1246: %3. 
1247: [AhB] Ahlfors, L.; Bers, L. Riemann's mapping theorem for variable metrics, - 
1248: Ann. of Math. (2) 72 (1960), 385-404. 
1249: 
1250: [Be] Bers, L. Riemann surfaces (mimeographed lecture notes), New York 
1251: University (1957-58). 
1252: 
1253: [BeN] Bers, L.; Nirenberg, L.   
1254: On a representation theorem for linear elliptic systems with discontinuous 
1255: coefficients and its applications. 
1256: Convegno Internazionale sulle Equazioni Lineari alle Derivate Parziali, Trieste, 1954, pp. 111--140. 
1257: Edizioni Cremonese, Roma, 1955. 
1258: 
1259: [Bo] Boyarski\u\i, B. V. 
1260: Generalized solutions of a system of differential equations of first order and of elliptic type with
1261: discontinuous coefficients. (Russian) 
1262: Mat. Sb. N.S. 43(85) 1957 451--503.
1263: 
1264: [CG] Carleson, L., Gamelin, Th.W. Complex Dynamics, - Springer-Verlag 1993. 
1265: 
1266: [Ch] Choquet-Bruhat, Y., \ de Witt-Morette, C., \ Dillard-Bleick, M. Analysis, 
1267: Manifolds and Physics, - North-Holland, 1977.
1268: 
1269: [Chern] Chern, S.-S., 
1270: An elementary proof of the existence of isothermal parameters on a surface. 
1271: Proc. Amer. Math. Soc. 6 (1955), 771--782.
1272: 
1273: [DB] Douady, A.; Buff, X. Le th\'eor\`eme d'int\'egrabilit\'e 
1274: des structures pr\`esque complexes. (French) [Integrability theorem
1275:    for almost complex structures] The Mandelbrot set, theme and variations, 307--324, London Math. Soc. Lecture Note Ser., 274,
1276:    Cambridge Univ. Press, Cambridge, 2000.
1277: 
1278: [DH] Douady, A.; Hubbard, J. 
1279: A proof of Thurston's topological characterization of rational functions. 
1280: Acta Math. 171 (1993), no. 2, 263--297. 
1281: 
1282: [Gl] Glutsyuk, A.; Simultaneous metric uniformization of foliations by 
1283: Riemann surfaces. - To appear in Commentarii Mahtematici Helvetici. 
1284: 
1285: [Ko1] K\"obe, P. \"Uber die Uniformisierung beliebiger analytischer Kurven I. 
1286: - Nachr. Acad. Wiss. G\"ottingen (1907), 177-190.
1287: 
1288: [Ko2] K\"obe, P. \"Uber die Uniformisierung beliebiger analytischer Kurven II. 
1289: - Ibid. (1907), 633-669. 
1290: 
1291: [Korn] Korn, A., Zwei Anwendungen der Methode der sukzessiven Ann\"aherungen, - 
1292: Schwarz Festschrift, Berlin (1919), pp. 215-229. 
1293: 
1294: [La] Lavrentiev, M.A., Sur une classe des repr\'esentations continues. - Mat. Sb., 
1295: 42 (1935), 407-434. 
1296: 
1297: [Licht] Lichtenstein, L., Zur Theorie der konformen Abbildungen; Konforme 
1298: Abbildungen nicht-analytischer singularit\"atenfreier Fl\"achenst\"ucke auf 
1299: ebene Gebiete, - Bull. Acad. Sci. Cracovie, (1916), 192-217. 
1300: 
1301: [M] Morrey, C. B., Jr. 
1302: On the solutions of quasi-linear elliptic partial differential equations. - 
1303: Trans. Amer. Math. Soc. 43 (1938), no. 1, 126--166.
1304: 
1305: [P] Poincar\'e, H. Sur l'uniformisation des fonctions analytiques, - 
1306: Acta Math., 31 (1907), 1-64. 
1307: 
1308: [Vek]  Vekua, I. N. The problem of reduction to canonical form of differential forms of elliptic type and the generalized
1309:    Cauchy-Riemann system. - (Russian) Dokl. Akad. Nauk SSSR (N.S.) 100, 
1310:    (1955), 197--200.
1311: 
1312: \end{document}
1313: 
1314:  \begin{figure}[ht]
1315:   \begin{center}
1316:    \epsfig{file=uqc4.eps}
1317:     \caption{}
1318:     \label{fig:4}
1319:   \end{center}
1320: \end{figure}
1321: 
1322:    
1323: 
1324: 
1325: