1: % > 538 000 bytes ?? YES : 

2: % Already : 714 000 bytes

3: \documentclass[10pt,twoside,leqno]{amsart}

4: \usepackage{amssymb,amsbsy,amsmath,amsfonts,amssymb,amscd,epsfig,times,graphics,color}

5: \usepackage[latin1]{inputenc}

6: \sloppy

7: \tolerance = 1000

8: 

9: %Format des cadres pour les figures sous xfig : 1-13 cm par 1-6 cm

10: 

11: \newcommand{\B}                 {\mathbb{B}}

12: \newcommand{\C}                 {\mathbb{C}}

13: \newcommand{\I}                 {\mathbb{I}}

14: \newcommand{\K}                 {\mathbb{K}}

15: \newcommand{\LL}                {\mathbb{L}}

16: \newcommand{\R}                 {\mathbb{R}}

17: \newcommand{\N}                 {\mathbb{N}}

18: \renewcommand{\P}               {\mathbb{P}}

19: \newcommand{\V}                 {\mathbb{V}}

20: \def\dl{{[\![}}

21: \def\dr{{]\!]}}

22: 

23: \newtheorem{theorem}{Theorem}

24: \newtheorem{lemma}{Lemma}

25: \newtheorem{proposition}{Proposition}

26: \newtheorem{corollary}{Corollary}

27: \newtheorem{definition}{Definition}

28: \newtheorem{remark}{Remark}

29: \newtheorem{example}{Example}

30: 

31: \setlength{\textwidth}{12.5cm}

32: % Impression :         12.5cm

33: % Impression :         15.5cm

34: % Ecriture   :         13.5cm

35: \setlength{\textheight}{20.5cm}

36: % Impression :          20.5cm

37: % Impression :          24.5cm

38: % Ecriture   :          21.5cm

39: \voffset                  0cm

40: % Impression :            0cm

41: % Impression :            -1.5cm

42: % Ecriture   :            -2.5cm

43: \hoffset                  -0.5cm

44: % Impression :            -0.5cm

45: % Impression :            -2cm

46: % Ecriture   :            -3.5cm

47: 

48: \begin{document}

49: 

50: %\large

51: 

52: \title[Characteristic foliations and envelopes of

53: holomorphy]{Characteristic foliations on maximally real submanifolds

54: of $\C^n$ and envelopes of holomorphy}

55: 

56: \author{Jo\"el Merker and Egmont Porten}

57: 

58: \address{

59: CNRS, Universit\'e de Provence, LATP, UMR 6632, CMI, 

60: 39 rue Joliot-Curie, F-13453 Marseille Cedex 13, France}

61: 

62: \email{merker@cmi.univ-mrs.fr} 

63: 

64: \address{Humboldt-Universit\"at zu Berlin, 

65: Mathematisch-Naturwissenschaftenschaft\-liche Fa\-kult\"at II, 

66: Institut f\"ur Mathematik, Rudower Chaussee 25, 

67: D-12489 Berlin, Germany}

68: 

69: \email{egmont@mathematik.hu-berlin.de}

70: 

71: \subjclass[2000]{Primary: 32D20. 

72: Secondary: 32A20, 32D10, 32V10, 32V25, 32V35}

73: 

74: \date{\number\year-\number\month-\number\day}

75: 

76: \begin{abstract}

77: Let $S$ be an arbitrary real $2$-surface, with or without boundary,

78: contained in a hypersurface $M$ of $\C^2$, with $S$ and $M$ of class

79: $\mathcal{ C}^{ 2, \alpha}$, where $0< \alpha <1$. If $S$ is totally

80: real except at finitely many complex tangencies which are hyperbolic

81: in the sense of E.~Bishop and if the union of separatrices is a tree

82: of curves without cycles, we show that every compact $K$ of $S$ is

83: CR-, $\mathcal{ W}$- and $L^{\sf p}$-removable (Theorem~1.3). Our

84: purely local techniques enable us to formulate substantial

85: generalizations of this statement, for the removability of closed sets

86: in totally real $1$-codimensional submanifolds contained in generic

87: submanifolds of CR dimension $1$.

88: \end{abstract}

89: 

90: \maketitle

91: 

92: \begin{center}

93: \begin{minipage}[t]{11cm}

94: \baselineskip =0.35cm

95: {\scriptsize

96: 

97: \centerline{\bf Table of contents}

98: 

99: \smallskip

100: 

101: {\bf Part~I \dotfill 1.}

102: 

103: \ \

104: {\bf 1.~Introduction \dotfill 1.}

105: 

106: \ \

107: {\bf 2.~Description of the proof of Theorem~1.2 and organization 

108: of the paper \dotfill 8.}

109: 

110: \ \

111: {\bf 3.~Strategy per absurdum for the proofs of

112: Theorems~1.2' and~1.4 \dotfill 13.} 

113: 

114: \ \

115: {\bf 4.~Construction of

116: a semi-local half-wedge \dotfill 19.}

117: 

118: \ \

119: {\bf 5.~Choice of a special point of $C_{\rm nr}$ to be removed 

120: locally \dotfill 32.}

121: 

122: {\bf Part~II \dotfill 45.}

123: 

124: \ \

125: {\bf 6.~Three preparatory lemmas on H\"older spaces \dotfill 45.}

126: 

127: \ \

128: {\bf 7.~Families of analytic discs half-attached

129: to maximally real submanifolds \dotfill 47.}

130: 

131: \ \

132: {\bf 8.~Geometric properties of families of half-attached analytic

133: discs \dotfill 55.}

134: 

135: \ \

136: {\bf 9.~End of proof of Theorem~1.2': application of 

137: the continuity principle \dotfill 63.}

138: 

139: \ \

140: {\bf 10.~Three proofs of Theorem~1.4 \dotfill 69.}

141: 

142: \ \

143: {\bf 11.~$\mathcal{ W}$-removability implies

144: $L^{\sf p}$-removability \dotfill 73.}

145: 

146: \ \

147: {\bf 12.~Proofs of Theorem~1.1 and of Theorem~1.3

148: \dotfill 78.}

149: 

150: \ \

151: {\bf 13.~Applications to the edge of the wedge theorem \dotfill 87.}

152: 

153: \ \

154: {\bf 14.~An example of a nonremovable three-dimensional torus \dotfill 

155: 91.}

156: 

157: \ \

158: {\bf 15.~References \dotfill 96.}

159: 

160: \smallskip

161: 

162: {\footnotesize\tt [With 24 figures]}

163: 

164: }\end{minipage}

165: \end{center}

166: 

167: \bigskip

168: 

169: \section*{\S1.~Introduction}

170: 

171: In the past fifteen years, remarkable progress has been made towards

172: the understanding of the holomorphic extendability properties of CR

173: functions. At the origin of this development, the most fundamental

174: achievement was the deep discovery, due to the effort of numerous

175: mathematicians, that the so-called {\sl CR orbits} are the adequate

176: underlying objects for the semi-local CR analysis on a general

177: embedded CR manifold. As an independent and now established theory in

178: several complex variables, one may find a precise correspondence

179: between such orbits and progressively attached analytic discs covering

180: a thick part of the envelope of holomorphy of CR manifolds, {\it cf.}

181: \cite{ b}, \cite{ trv}, \cite{ tr1}, \cite{ tu1}, \cite{ ber}, \cite{

182: tu2}, \cite{ m1}, \cite{ j2} and \cite{ p3} for a recent synthesis.

183: 

184: Within this framework, it became mathematically accessible to

185: endeavour the general study of removable singularities on embedded CR

186: manifolds $M\subset \C^n$ of arbitrary CR dimension and of arbitrary

187: codimension, not necessarily being the boundaries of (strictly)

188: pseudoconvex domains. With respect to their size or ``mass'', the

189: interesting singularities can be essentially ordered by

190: their codimension in $M$. For instance, provided it does not perturb

191: the fact that $M$ consists of a single CR orbit, an arbitrary closed

192: subset $C\subset M$ which is of vanishing 2-codimensional Hausdorff

193: content is {\it always}\, removable, as is shown in~\cite{ cs} in the

194: hypersurface case and in \cite{ mp3}, Theorem~1.1, in arbitrary

195: codimension. Hence one is left to study the removability of

196: singularities of codimension at most two. Since the general problem

197: of characterizing removability seems at the moment to be out of reach

198: (even for $M$ being a hypersurface), it is advisible to focus on

199: geometrically accessible singularities, namely singularities contained

200: in a CR submanifold of $M$. A complete study of the automatic

201: removability of two-codimensional singularities may be found in

202: Theorem~4 of~\cite{ mp1}. Having in mind the classical Painlev\'e

203: problem, we will mainly consider in this paper singularities which are

204: closed sets $C$ contained in a {\it codimension one}\, 

205: submanifold $M^1$ of $M$ which is {\it generic}\, in $\C^n$.

206: 

207: The known results on singularities of codimension one can be

208: subdivided into two strongly different groups according to the {\it CR

209: dimension} of $M$. If ${\rm CRdim}\, M \geq 2$, then a generic

210: hypersurface $M^1\subset M$ is itself a CR manifold of positive CR

211: dimension, and singularities $C\subset M^1$ can be understood on the

212: basis of the interplay between $C$ and the CR orbits of $M^1$. Deep

213: results in this direction were established when $M$ is a hypersurface

214: of $\C^n$ in~\cite{ j4}, \cite{ j5} and then generalized to CR

215: manifolds of arbitrary codimension in~\cite{ p1}: the 

216: geometric condition insuring automatic removability is simply

217: that $C$ does not contain any CR orbit of $M^1$.

218: 

219: On the other hand, if ${\rm CRdim}\, M=1$ the geometric situation

220: becomes highly different, as a generic hypersurface $M^1\subset M$ is

221: now (maximal) {\it totally real}. Fortunately, as a substitute for the

222: CR orbits of $M^1$, one can consider the so-called {\sl characteristic

223: foliation}\, of $M^1$, obtained by integrating the characteristic line

224: field $T^c M|_{M^1} \cap TM^1$. But removability theorems exploiting

225: this concept were only known for hypersurfaces in $\C^2$ and, until

226: very recently, only in the strictly pseudoconvex case. Furthermore,

227: a geometric condition insuring automatic removablitity

228: has not yet been clearly delineated.

229: 

230: Hence, with respect to the current state of the art, there was a

231: two-fold gap about codimension one removable singularities contained

232: in generic submanifolds $M$ of CR dimension one: firstly, to establish

233: a satisfying theory for non-pseudoconvex hypersurfaces in $\C^2$ and

234: secondly, to understand the situation in higher codimension. This

235: second main task was formulated as the first open problem in a list

236: p.~432 of \cite{j5} ({\it see}\, also the comments pp.~431--432 about

237: the relative geometric simplicity of the case ${\rm CRdim}\, M \geq

238: 2$). {\it A priori}, it is not clear at all whether the two directions

239: of research are related somehow, but in the present work, we shall

240: fill in this two-fold gap by devising a new semi-local approach which

241: applies uniformly with respect to codimension.

242: 

243: For the detailed discussion of our result we have to introduce some

244: terminology which will be used throughout the article. Let $M$ be a

245: generic submanifold of $\C^n$ and let $C$ be a closed subset of $M$.

246: Recall from~\cite{mp3} that a {\sl wedgelike domain} attached to a

247: generic submanifold $M'\subset \C^n$ is a domain containing a local

248: wedge of edge $M'$ at every point of $M'$. Our wedgelike domains will

249: always be {\it nonempty}. Let us define three basic notions of

250: removability. Firstly, we say that $C$ is {\sl CR-removable} if there

251: exists a wedgelike domain $\mathcal{ W}$ attached to $M$ to which

252: every continuous CR function $f\in\mathcal{ C}_{CR}^0(M\backslash C)$

253: extends holomorphically. Secondly, as in~\cite{mp3}, p.~486, we say

254: that $C$ is {\sl $\mathcal{ W}$-removable} if for every wedgelike

255: domain $\mathcal{ W}_1$ attached to $M\backslash C$, there is a

256: wedgelike domain $\mathcal{ W}_2$ attached to $M$ and a wedgelike

257: domain $\mathcal{ W}_3 \subset \mathcal{ W}_1 \cap \mathcal{ W}_2$

258: attached to $M\backslash C$ such that for every holomorphic function

259: $f\in\mathcal{ O}(\mathcal{ W}_1)$, there exists a holomorphic

260: function $F\in\mathcal{ O}(\mathcal{ W}_2)$ which coincides with $f$

261: in $\mathcal{ W}_3$. Thirdly, with ${\sf p} \in\R\cup \{+\infty\}$

262: satisfying ${\sf p} \geq 1$, we say that $C$ is {\it $L^{\sf

263: p}$-removable} if every locally integrable function $f\in L^{\sf

264: p}_{loc}(M)$ which is CR in the distributional sense on $M\backslash

265: C$ is in fact CR on all of $M$.

266: 

267: The first notion of removability is a generalization of the kind of

268: removability considered in most of the pioneering papers \cite{cs},

269: \cite{d}, \cite{fs}, \cite{j1}, \cite{kr}, \cite{l}, \cite{lu},

270: \cite{st} about removable singularities in boundaries of domains $D

271: \subset \subset \C^n$. We observe that a wedgelike open set attached

272: to a hypersurface $M$ is just a (global) one-sided neighborhood of

273: $M$, namely a domain $\omega$ with $\overline{\omega}\supset M$ such

274: that for every point $p\in M$, the domain $\omega$ contains the

275: intersection of a neighborhood of $p$ in $\C^n$ with one side of

276: $M$. If now a closed set $C$ contained in a $\mathcal{ C}^1$-smooth

277: bounded boundary $\partial D$ is CR-removable, then an application of

278: the Hartogs-Bochner theorem shows that CR functions on $\partial

279: D\backslash C$ can be holomorphically extended to $D$. The second

280: notion of removability is a way to isolate the part of the question

281: related to envelopes of holomorphy. The third notion of removability

282: has the advantage of being completely intrinsic with respect to $M$

283: and may be relevant in the study of non-embeddable CR manifolds.

284: 

285: To avoid confusion, we state precisely our submanifold notion: $Y$ is

286: a {\sl submanifold of $X$} if $Y$ and $X$ are equipped with a manifold

287: structure, if there exists an immersion $i$ of $Y$ into $X$ and if the

288: manifold topology of $Y$ and the topology of $i(Y)$ inherited from the

289: topology of $X$ coincide, so that one may identify the submanifold $Y$

290: with the subset $i(Y)\subset X$. Furthermore, our submanifolds will

291: always be connected.

292: 

293: Let us now enter the discussion of the case $n=2$. Here we shall

294: denote the submanifold $M^1\subset M$, which is a surface in $\C^2$,

295: by $S$. In~\cite{b}, E.~Bishop showed that a two-dimensional surface

296: in $\C^2$ of class at least $\mathcal{ C}^2$ having an isolated

297: complex tangency at one of its points $p$ may be represented by a

298: complex equation of the form $w=z\bar z+\lambda(z^2+\bar z^2)+{\rm

299: o}(\vert z\vert^2)$, in terms of local holomorphic coordinates $(z,w)$

300: vanishing at $p$, where the real parameter $\lambda\in [0,\infty)$ is

301: a biholomorphic invariant of $S$. The point $p$ is said to be {\sl

302: elliptic} if $\lambda\in [0,\frac{1}{2})$, {\sl parabolic} if $\lambda

303: = \frac{1}{2}$ and {\sl hyperbolic} if $\lambda \in (\frac{1}{2},

304: \infty)$. Recall that $M$ is called {\sl globally minimal} if it

305: consists of a single CR orbit ({\it cf.} \cite{tr1}, \cite{tr2};

306: \cite{mp1}, pp.~814--815; and~\cite{j4}, pp.~266--269). Throughout

307: this paper, we shall work in the $\mathcal{ C}^{ 2, \alpha}$-smooth

308: category, where $0 < \alpha < 1$. Our first main new result is as

309: follows.

310: 

311: \def\thetheorem{{\bf 1.1}}\begin{theorem}

312: Let $M$ be a globally minimal $\mathcal{ C}^{ 2, \alpha}$-smooth

313: hypersurface in $\C^2$ and let $D \subset M$ be a $\mathcal{ C}^{ 2,

314: \alpha}$-smooth surface which is

315: \begin{itemize}

316: \item[{\bf (a)}]

317: $\mathcal{C}^{2,\alpha}$-diffeomorphic to the unit

318: $2$-disc of $\R^2$ and 

319: \item[{\bf (b)}]

320: totally real outside a discrete subset of isolated complex tangencies

321: which are hyperbolic in the sense of E.~Bishop.

322: \end{itemize}

323: Then every compact subset $K$ of $D$ is CR-, $\mathcal{ W}$- and

324: $L^{\sf p}$-removable.

325: \end{theorem}

326: 

327: As a corollary, one obtains a corresponding result about holomorphic

328: extension from $\partial\Omega\backslash K$ for the case that $M$ is

329: the boundary of a relatively compact domain $\Omega\subset\C^2$. Note

330: that $\partial \Omega$ is automatically globally minimal

331: (\cite{ j4}, Section~2). We will first recall the historical

332: background of Theorem 1.1 and explain afterwards on this basis the

333: main ideas and techniques necessary for the proof. 

334: 

335: In 1988, applying a global

336: version of the Kontinuit\"atssatz, B.~J\"oricke~\cite{j1} established

337: a remarkable theorem: every compact subset of a totally real

338: $\mathcal{ C}^2$-smooth $2$-disc lying on the boundary of the unit

339: ball in $S^3=\partial \B_2\subset \C^2$ is CR-removable. This

340: discovery motivated the work~\cite{fs} by

341: F.~Forstneri$\check{\text{\rm c}}$ and E.L.~Stout, where it is shown

342: that every $\mathcal{ C}^2$-smooth compact $2$-disc contained in a

343: strictly pseudoconvex $\mathcal{ C}^2$-smooth boundary $\partial

344: \Omega$ contained in a $2$-dimensional Stein manifold $\mathcal{ M}$

345: which is totally real except at a finite number of hyperbolic complex

346: tangencies is removable; the proof mainly relies on a previous work by

347: E.~Bedford and W.~Klingenberg about the hulls of $2$-spheres contained

348: in such strictly pseudoconvex boundaries $\Omega \subset \mathcal{

349: M}$, which may be filled by Levi-flat $3$-spheres after a generic

350: small perturbation (\cite{bk}, Theorem~1). Indirectly, it followed

351: from~\cite{j1} and~\cite{fs} that such compact totally real $2$-discs

352: $D\subset \partial \Omega$ (possibly having finitely many hyperbolic

353: complex tangencies) are $\mathcal{ O}(\overline{\Omega})$-convex and

354: in particular polynomially convex if $D=\B_2$ and $\mathcal{ M}=\C^2$,

355: thanks to a previous work~\cite{st} by E.L.~Sout, where it is shown

356: (Theorem~II.10) that a compact subset $K$ of a $\mathcal{ C}^2$-smooth

357: strictly pseudoconvex boundary $\partial \Omega$ in a Stein manifold

358: is removable {\it if and only if}\, $K$ is $\mathcal{

359: O}(\overline{\Omega})$-convex. It is also established in~\cite{fs}

360: that a neighborhood of an isolated hyperbolic complex tangency in

361: $\C^2$ is polynomially convex. These papers have been followed by the

362: work~\cite{d}, where the question of $\mathcal{

363: O}(\overline{\Omega})$-convexity of {\it arbitrary compact surfaces}\,

364: $S$ (with or without boundary, not necessarily diffeomorphic to a

365: $2$-disc) contained in a $\mathcal{ C}^2$-smooth strictly pseudoconvex

366: domain $\Omega\subset \C^2$ is dealt with directly. Using K.~Oka's

367: characterization of the envelope of a compact, J.~Duval shows that the

368: essential hull $\widehat{ K}_{\rm ess}:= \overline{ \widehat{ K}_{

369: \mathcal{ O }( \overline{ \Omega} )} \backslash K}$ must cross

370: every leaf of the characteristic foliation on the totally real part of

371: $S$ and he deduces that a compact $2$-disc having only hyperbolic

372: complex tangencies is $\mathcal{ O }(\overline{ D })$-convex.

373: 

374: All the above proofs heavily rely on strong pseudoconvexity, in

375: contrast to the experience, familiar at least in the case ${\rm

376: CRdim}\, M \geq 2$, that removability should depend rather on the

377: structure of CR orbits than on Levi curvature. The first theorem for

378: the non-pseudoconvex situation was established by the second author in

379: \cite{p2}. He proved that every compact subset of a totally real disc

380: embedded in a globally minimal $\mathcal{ C}^\infty$-smooth

381: hypersurface in $\C^2$ is always CR-removable. We would like to point

382: out that, seeking theorems without any assumption of pseudoconvexity

383: leads to substantial open problems, because one loses almost all of

384: the strong interweavings between function-theoretic tools and

385: geometric arguments which are valid in the pseudoconvex realm, for

386: instance: Hopf Lemma, plurisubharmonic exhaustions, envelopes of

387: function spaces, local maximum modulus principle, Stein neighborhood

388: basis, {\it etc.}

389: 

390: To discuss the main elements of our approach, let us briefly explain

391: the geometric setup of the proof of Theorem 1.1. The characteristic

392: foliation has isolated singularities at the hyperbolic points, where it

393: looks like the phase diagram of a saddle point. In particular there

394: are four local separatrices accumulating orthogonally at each

395: hyperbolic point. Hence we can decompose the $2$-disc $D$ as a union

396: $D=T_D \cup D_o$, where $T_D$ consists of the union of the hyperbolic

397: points of $D$ together with the separatrices issuing from them, and

398: where $D_o:= D \backslash T_D$ is the remaining open submanifold of

399: $D$, contained in the totally real part of $D$. By

400: H.~Poincar\'e and I.~Bendixson's theory, $T_D$ is a tree of $\mathcal{

401: C}^{2,\alpha}$-smooth curves which contains no subset homeomorphic to

402: the unit circle, {\it cf.}~\cite{ d}. Accordingly, we decompose $K:=

403: K_{ T_D} \cup C_o$, where $K_{T_D}:= K \cap T_D$ is a proper closed

404: subset of the tree $T_D$ and where $C_o := K \cap D_o$ is a relatively

405: closed subset of $D_o$.

406: 

407: The hard part of the proof, which was actually the starting point of

408: the whole paper, will consist in removing the closed subset $C_o$ of

409: the $2$-dimensional surface $S:=D_o$ lying in $M \backslash

410: T_D$. Thereafter the removal of the remaining part $K_{T_D}$ will

411: be done by means of an investigation of the behaviour of

412: the CR orbits near $T_D$, close in spirit to our previous

413: methods in ~\cite{mp1} ({\it see}\, Section~12 below for the details).

414: 

415: Let us formulate the first crucial part of the above argument as an

416: independent theorem about the removal of closed subsets contained in a

417: totally real surface $S$. We point out that now $S$ may have {\it

418: arbitrary topology}.

419: 

420: \def\thetheorem{{\bf 1.2}}\begin{theorem}

421: Let $M$ be a $\mathcal{ C}^{ 2, \alpha }$-smooth globally minimal

422: hypersurface in $\C^2$, let $S \subset M$ be a $\mathcal{ C}^{2,

423: \alpha}$-smooth surface, open or closed, with or without boundary,

424: which is totally real at every point. Let $C$ be a proper closed

425: subset of $S$ and assume that the following topological condition

426: holds{\rm :}

427: 

428: \smallskip

429: \begin{itemize}

430: \item[$\mathcal{ F }_S^c \{C\}:$] For every closed subset $C' \subset

431: C$, there exists a simple $\mathcal{ C}^{2,\alpha}$-smooth curve

432: $\gamma: [-1, 1] \to S$, whose range is contained in a single leaf of

433: the characteristic foliation $\mathcal{ F }_S^c$ {\rm (}obtained

434: by integrating the characteristic line field $T^cM \vert_S \cap 

435: TS${\rm )}, with $\gamma (-1)

436: \not \in C'$, $\gamma (0) \in C'$ and $\gamma (1) \not \in C'$, such

437: that $C'$ lies completely in one closed side of $\gamma [-1, 1]$

438: with respect to the topology of $S$ in a neighborhood of

439: $\gamma[-1,1]$.

440: \end{itemize}

441: 

442: \smallskip

443: \noindent

444: Then $C$ is CR-, $\mathcal{ W}$- and $L^{\sf p}$-removable.

445: \end{theorem}

446: 

447: The condition $\mathcal{ F}_S^c \{C \}$ is a {\it common condition}\,

448: on $C$ and on the characteristic foliation $\mathcal{ F}_S^c$, namely

449: on the relative disposition of $\mathcal{ F}_S^c$ with respect to $C$,

450: not only on $S$; an illustration may be found in {\sc Figure~2} below.

451: In the strictly pseudoconvex context, this condition appeared

452: implicitly during the course of the proofs given in \cite{d}. Note

453: that the relevance of the characteristic foliation had earlier been

454: discovered in contact geometry, {\it cf.} \cite{be}, \cite{e}. It is

455: interesting to notice that it re-appears in the situation of Theorem

456: 1.2, where the underlying distribution $T^c M$ is allowed to be very

457: far from contact.

458: 

459: As is known, it follows from a subcase of H.~Poincar\'e and

460: I.~Bendixson's theory that if $S$ is diffeomorphic to a real $2$-disc

461: or if $S=D_o$ as above, then $\mathcal{ F}_S^c \{C\}$ is automatically

462: satisfied for an arbitry compact subset $C$ of $S$. On the contrary, it

463: may be not satisfied when for instance $S$ is an annulus equipped with

464: a radial foliation together with $C$ containing a continuous closed

465: curve around the hole of $S$. Crucially, it is elementary to construct

466: an example of such an annulus which is truly nonremovable. Indeed, the

467: small closed curve $C$ which consists of the transversal intersection

468: of a strictly convex boundary $\partial D$ with a complex line close

469: to a boundary point may be enlarged as a thin maximally real strip $S

470: \subset \partial D$ which is diffeomorphic to an annulus; in this

471: setting, $C$ is obviously nonremovable and {\it the characteristic

472: foliation is everywhere transversal to $C$}. Consequently, the

473: geometric condition $\mathcal{ F}_S^c\{C\}$ is the optimal one

474: insuring automatic removability for all choices of $M$, $S$ and

475: $C$. Further examples of closed subsets in surfaces with arbitrary

476: genus equipped with such foliations may be exhibited.

477: 

478: In the proof of Theorem~1.2, after some contraction $C'$ of $C$, we

479: may assume that no point of $C'$ is locally removable ({\it see}\,

480: Sections~2 and~3 below). Then the very assumption $\mathcal{

481: F}_S^c\{C\}$ yields the existence of a characteristic segment

482: $\gamma[-1,1]$, such that $C'$ lies on one side of

483: $\gamma[-1,1]$. Reasoning by contradiction, our aim is to show that

484: there exists at least one special point $p\in C'\cap \gamma(-1,1)$,

485: which is {\it locally}\, CR-, $\mathcal{ W}$- and $L^{\sf

486: p}$-removable. The choice of such a point $p$, achieved in Section~5

487: below, will be nontrivial.

488: 

489: The strategy for the local removal of $p$ is to construct an analytic

490: disc $A$ such that a segment of its boundary $\partial A$ is attached

491: to $S$ and touches $C'$ in only one point $p$. Several geometrical

492: assumptions have to be met to ensure that a sufficiently rich family

493: of deformations of $A$ have boundaries disjoint from $C'$, that

494: analytic extension along these discs is possible (i.e.~appropriate

495: moment conditions are satisfied), and that the union of these good

496: discs is large enough to give analytic extension to a one-sided

497: neighborhood of $M$: this is where the (semi)localization and the

498: choice of the special point $p\in C'$ will be key ingredients.

499: Let us explain why localization is crucial.

500: 

501: Working globallly, the second author produced in~\cite{ p2} a

502: convenient disc by applying the powerful E.~Bedford and W.~Klingenberg

503: theorem to an appropriate 2-sphere containing a neighborhood of the

504: {\it entire}\, singularity $C'$. This method requires global

505: properties of $S$ like $S$ being a totally real $2$-disc, which

506: ensures the existence of a nice Stein neighborhood basis of

507: $C'$. Already for real discs with isolated hyperbolic points, it is

508: not clear whether this argument can be generalized (however, we would

509: like to mention that recent results of M.~Slapar in~\cite{ sl}

510: indicate that this could be possible at least if the geometry near the

511: hyperbolic points satisfies some additional assumptions). In the case

512: where $M$ is an arbitrary globally minimal hypersurface, where $S$ has

513: arbitrary topology and has complex tangencies, the reduction to

514: E.~Bedford and W.~Klingenberg's theorem seems impossible, {\it cf.}

515: the example of an unknotted nonfillable $2$-sphere in $\C^2$

516: constructed by J.E.~Forn{\ae}ss and D.~Ma in~\cite{fm}. Also, to the

517: authors' knowledge, the possibility of filling by Levi-flat

518: $3$-spheres the (not necessarily generic) $2$-spheres lying on a {\it

519: nonpseudoconvex}\, hypersurface is a delicate open problem. In

520: addition, for the higher codimensional generalization of Theorem~1.2,

521: the idea of global filling seems to be irrelevant at present times,

522: because no analog of the E.~Bedford and W.~Klingenberg theorem is

523: known in dimension $n\geq 3$. As we aim to deal with surfaces $S$

524: having arbitrary topology and to generalize these results in arbitrary

525: codimension, we shall endeavour to firmly {\it localize the

526: removability arguments}, using only small analytic discs.

527: 

528: Thus, our way to overcome these obstacles is to consider {\it local}

529: discs $A$ which are only partially attached to $S$. The delicate point

530: is that we have at the same time (i) to control the geometry of

531: $\partial A$ near $p\in C'$ and (ii) to guarantee that the rest of the

532: boundary stays in the region where holomorphic extension is already

533: known. In fact, (ii) will be incorporated in our very special and

534: tricky choice of $p\in C'$. For (i), we have to sharpen known existence

535: theorems about partially attached analytic discs and to combine it

536: with a careful study of the complex/real geometry of the pair

537: $(M,S)$. Importantly, our construction of such analytic discs is

538: achieved elementarily in a self-contained way. A precise description

539: of the proof in the hypersurface case (only) may be found in Section~2

540: below. With some substantial extra work, we shall generalize this

541: purely local strategy of proof to higher codimension.

542: 

543: To conclude with the removal of surfaces, let us formulate a more

544: general version of Theorem~1.1, whitout the restricted topological

545: assumption that $S$ be diffeomorphic to a real disc. Applying

546: Theorem~1.2 for the removal of $K\cap (S\backslash T_S)$ and a slight

547: generalization of Theorem~4 (ii) in~\cite{mp1} for the removal of

548: $K\cap T_S$ (more precisions will be given in \S13 below), we shall

549: obtain the following statement, implying Theorem~1.1 as a 

550: direct corollary.

551: 

552: \def\thetheorem{1.3}\begin{theorem}

553: Let $M$ be a $\mathcal{ C}^{ 2, \alpha }$-smooth globally minimal

554: hypersurface in $\C^2$, let $S \subset M$ be a $\mathcal{ C}^{2,

555: \alpha}$-smooth totally real surface, open or closed, with or without

556: boundary, which is totally real outside a discrete subset of isolated

557: complex tangencies which are hyperbolic in the sense of E.~Bishop.

558: Let $T_S$ be the union of hyperbolic points of $S$ together with all

559: separatrices issued from hyperbolic points and assume that $T_S$ does

560: not contain any subset which is homeomorphich to the unit circle. Let

561: $K$ be a proper compact subset of $S$ and assume that $\mathcal{ F

562: }_{S\backslash T_S}^c \{K \cap (S \backslash T_S)\}$ holds.

563: 

564: Then $K$ is CR-, $\mathcal{ W}$- and $L^{\sf p}$-removable.

565: \end{theorem}

566: 

567: As was already emphasized, our main motivation for this work was to

568: {\it devise a local strategy of proof for Theorems~1.1, 1.2 and~1.3 in

569: order to generalize them to higher codimension}. In fact, we will

570: realize the program sketched above for generic submanifolds of CR

571: dimension $1$ and of arbitrary codimension. Thus, let $M$ be a

572: $\mathcal{C}^{2,\alpha}$-smooth globally minimal

573: generic submanifold of codimension $(n-1)$ in $\C^n$, hence of CR

574: dimension $1$, where $n\geq 2$. Let $M^1$ be a maximally real

575: $\mathcal{C}^{2,\alpha}$-smooth one-codimensional submanifold of $M$

576: which is generic in $\C^n$. As in the surface case, $M^1$ carries a

577: {\sl characteristic foliation} $\mathcal{ F}_{M^1}^c$, whose leaves

578: are the integral curves of the line distribution $TM^1\cap

579: T^cM\vert_{M^1}$. Of course the assumption that the singularity lies

580: on one side of some characteristic segment is no longer reasonable.

581: We will generalize it as a condition requiring (approximatively

582: speaking) that there be always a characteristic segment which is

583: accessible from the complement of $C$ in $M^1$ along one direction

584: normal to the characteristic segment.

585: 

586: The generalization of Theorem~1.2, which is our

587: principal result in this paper, is as follows.

588: 

589: \def\thetheorem{{\bf 1.2'}}\begin{theorem}

590: Let $M$, $M^1$, $\mathcal{F}_{M^1}^c$ be as above and let $C$ be a

591: proper closed subset of $M$. Assume that the following topological

592: condition, meaning that $C$ is not transversal to the characteristic

593: foliation, holds{\rm :}

594: \smallskip

595: \begin{itemize}

596: \item[$\mathcal{ F}_{ M^1 }^c \{C\}:$] For every closed subset $C'

597: \subset C$, there exists a simple $\mathcal{ C}^{2, \alpha}$-smooth

598: curve $\gamma: [-1,1] \to M^1$ whose range $\gamma[-1,1]$ is

599: contained in a single leaf of the characteristic foliation

600: $\mathcal{F}_{ M^1}^c$ with $\gamma(-1)\not\in C'$, $\gamma (0) \in

601: C'$ and $\gamma (1) \not \in C'$, there exists a local $(n-

602: 1)$-dimensional transversal $R^1 \subset M^1$ to $\gamma$ passing

603: through $\gamma(0)$ and there exists a thin elongated

604: open neighborhood $V_1$ of

605: $\gamma [-1,1]$ in $M^1$ such that if $\pi_{ \mathcal{F}_{ M^1}^c}:

606: V_1 \to R^1$ denotes the semi-local projection parallel to the leaves

607: of the characteristic foliation $\mathcal{F}_{ M^1}^c$, then

608: $\gamma(0)$ lies on the boundary, relatively to the topology of $R^1$,

609: of $\pi_{\mathcal{F}_{M^1}^c}(C' \cap V_1)$.

610: \end{itemize}

611: \smallskip

612: \noindent 

613: Then $C$ is CR-, $\mathcal{ W}$- and $L^{\sf p}$-removable.

614: \end{theorem}

615: 

616: The condition $\mathcal{ F}_{ M^1}^c \{C \}$, which is of course

617: independent of the choice of the transversal $R^1$ and of the thin

618: neighborhood $V_1$, is illustrated in {\sc Figure~8} of \S5.1 below;

619: clearly, in the case $n =2$, it means that $C' \cap V_1$ lies

620: completely in one side of $\gamma [-1, 1]$, with respect to the

621: topology of $M^1$, as written in the statement of

622: Theorem~1.2. Applying some of our previous results in this direction

623: (\cite{mp1}, \cite{mp3}), we shall provide in the end of Section~13

624: below some formulations of applications of Theorem~1.2', close to

625: being analogs of Theorem~1.3 in higher codimension.

626: 

627: Importantly, in order to let the geometric condition $\mathcal{

628: F}_{M^1}^c\{C\}$ appear less mysterious and to argue that it provides

629: the adequate generalization of Theorem~1.2 to higher codimension, in

630: the last Section~14 below, we shall describe an example of $M$, $M^1$

631: and $C$ in $\C^3$ violating the condition $\mathcal{ F}_{M^1}^c\{C\}$,

632: such that $C$ is transversal to the characteristic foliation and is

633: truly nonremovable. This example will be analogous in some sense to

634: the example of a nonremovable annulus discussed after the statement of

635: Theorem~1.2. Since there is no H.~Poincar\'e and I.~Bendixson 

636: theorem for

637: foliations of $3$-dimensional balls by curves, it will be even

638: possible to insure that $M$ and $M^1$ are diffeomorphic to real balls

639: of dimension $4$ and $3$ respectively. We may therefore conclude that

640: Theorem~1.2' provides the desirable answer to the (already cited {\em

641: supra}) Problem~2.1 raised by B.~J\"oricke in~\cite{ j5}, p.~432.

642: 

643: To pursue the presentation of our results, let us comment the

644: assumption that $M$ be of codimension $(n-1)$. Geometrically speaking,

645: the study of closed singularities $C$ lying in a one-codimensional

646: generic submanifold $M^1$ of a generic submanifold $M\subset \C^n$

647: which is of CR dimension $m\geq 2$ is more simple. Indeed, thanks to

648: the fact that $M^1$ is of CR dimension $m-1\geq 1$, there exist local

649: Bishop discs {\it completely attached to $M^1$}, and this helps much in

650: describing the envelope of holomorphy of a wedge attached to

651: $M\backslash C$. On the contrary, in the case where $M$ is of CR

652: dimension $1$, small analytic discs attached to a maximally real $M^1$

653: are (trivially) inexistent. This is why, in the proof of our main

654: Theorems~1.2 and~1.2', we shall deal only with small analytic discs

655: whose boundary is in part (only) contained in $M^1$. Such discs are

656: known to exist; we would like to mention that historically speaking,

657: the first construction of discs partially attached to maximally real

658: submanifolds was exhibited by S.~Pinchuk in~\cite{p}, who developed

659: the ideas of E.~Bishop~\cite{b}.

660: 

661: Finally we will test our main Theorem~1.2' in applications. First of

662: all, we clarify its relation to known removability results in CR

663: dimension greater than one. Here the motivation is simply that most

664: questions of CR geometry should be reducible to CR dimension $1$ by

665: slicing. It turns out that the main known theorems about removable

666: singularities, due to E.~Chirka, E.~L.~Stout, and B.~J\"oricke for

667: hypersurfaces, and by the authors in higher codimension (\cite{p1},

668: Theorem~1 about $L^{\sf p}$-removability; \cite{m2}, Theorem~3 about

669: CR- and $\mathcal{ W}$- removability) are all a rather direct

670: consequence of Theorem 1.2'. Since these results have not yet been

671: published in complete form, we take the occasion of including them in

672: the present paper, {\it as a corollary of Theorem~1.2'}, yet devising

673: a new geometric approach.

674: 

675: \def\thetheorem{{\bf 1.4}}\begin{theorem} 

676: Let $M$ be a $\mathcal{ C}^{2, \alpha}$-smooth globally minimal

677: generic submanifold of $\C^n$ of CR dimension $m\geq 2$ and of

678: codimension $d=n- m\geq 1$, let $M^1 \subset M$ be a $\mathcal{ C}^{2,

679: \alpha}$-smooth one-codimensional submanifold which is generic in

680: $\C^n$ and let $C \subset M^1$ be a proper closed subset of

681: $M$. Assume that the following

682: condition holds{\rm :}

683: \begin{itemize}

684: \item[$\mathcal{ O}_{M^1}^{CR}\{C\}:$] The closed subset

685: $C$ does not contain any CR orbit of $M^1$.

686: \end{itemize}

687: Then $C$ is CR-, $\mathcal{ W}$- and $L^{\sf p}$-removable.

688: \end{theorem}

689: 

690: Notice the difference with the case $m=1$, where the analog of CR

691: orbits would consist of characteristic curves: the condition

692: $\mathcal{ F}_{M^1}^c\{C\}$ does {\it not}\, say that $C$ should not

693: contain any maximal characteristic curve. In fact, we observe that

694: there cannot exist a uniform removability statement covering both the

695: case $m=1$ and the case $m\geq 2$, whence Theorem~1.2' is stronger

696: than Theorem~1.4. Indeed, the elementary example of a nonremovable

697: circle in an annulus contained in the boundary of a strictly convex

698: domain of $\C^2$ shows that $C$ may be truly nonremovable whereas it

699: does not contain any characteristic curve. In the strictly

700: pseudoconvex hypersurface setting, it is well known that Hopf's Lemma

701: implies that boundaries of Riemann surfaces contained in $C$ (and also

702: the track on $C$ of its essential hull, {\it cf.}~\cite{d}) should be

703: everywhere transversal to the characteristic foliation. Of course, this

704: implies conversely that $C$ cannot contain such boundaries (unless

705: they are empty) if $\mathcal{ F}_{M^1}^c\{C\}$ is satisfied. The

706: reason why $\mathcal{ F}_{M^1}^c\{C\}$ implies that $C$ is removable

707: also in the nonpseudoconvex setting and in arbitrary codimension will

708: be appearant later. Finally, we mention that the $L^{\sf

709: p}$-removability of $C$ in Theorem~1.4 holds more generally with no

710: assumption of global minimality on $M$, as already noticed in~\cite{

711: j5}, \cite{ p1}, \cite{mp1}. However, since the case where $M$ is not

712: globally minimal essentially reduces to the consideration of its CR

713: orbits, which are globally minimal by definition, we shall only

714: deal with globally minimal generic submanifolds $M$ throughout this

715: paper. 

716: 

717: As a final comment, we point out that, {\it because the previously known

718: proofs of Theorem~1.4 were of local type, it is satisfactory to bring

719: in this paper a purely local framework for the treatment of

720: Theorems~1.1, 1.2, 1.3 and~1.2'}.

721: 

722: Our second group of applications concerns the classical

723: edge of the wedge theorem. Typically one considers a maximally real

724: edge $E$ to which an open double wedge $(\mathcal{ W}_1, \mathcal{

725: W}_2)$ is attached from opposite directions. One may interprete this

726: configuration as a partial thickening of a generic CR manifold

727: $M\subset E\cup \mathcal{ W}_1\cup \mathcal{ W}_2$ containing $E$ as a

728: generic hypersurface. The classical edge of the wedge theorem states

729: that functions which are continuous on $ \mathcal{ W}_1 \cup \mathcal{

730: W}_2 \cup E$ and holomorphic in $\mathcal{ W}_1 \cup \mathcal{ W}_2$

731: extend holomorphically to a neighborhood of $E$. Theorem 1.2' implies

732: that it suffices to assume continuity outside a removable

733: singularities of $E$. This allows us to derive an edge of the wedge

734: theorem for meromorphic extension (Section~13 below).

735: 

736: This paper is divided in two parts: Part I contains the strategy per

737: absurdum for the proof of Theorem~1.2', the construction of what we

738: call a semi-local half-wedge and the choice of a special point to be

739: removed locally. Part~II contains the explicit construction of

740: families of half-attached analytic discs, the end of proof of

741: Theorem~1.2' and the proofs of the various applications. The reader

742: will find a more detailed description of the content of the paper in

743: \S2.16 below.

744: 

745: \subsection*{1.5.~Acknowledgements}

746: The authors would like to thank B.~J\"oricke for several valuable

747: scientific exchanges. They acknowledge generous support from the

748: European TMR research network ERBFMRXCT 98063 and they also thank the

749: universities of Berlin (Humboldt), of G\"oteborg (Chalmers), of

750: Marseille (Provence) and of Uppsala for invitations which provided

751: opportunities for fruitful mathematical research.

752: 

753: \section*{\S2.~Description of the proof of Theorem~1.2

754: and organization of the paper}

755: 

756: The main part of this paper is devoted to the proof of Theorem~1.2',

757: which will occupy Sections~3, 4, 5, 6, 7, 8 and~9 below. In this

758: preliminary section, we shall summarize the hypersurface version of

759: Theorem~1.2', namely Theorem~1.2. Our goal is to provide a

760: conceptional description of the basic geometric constructions, which

761: should be helpful to read the whole paper. Because precise, complete

762: and rigorous formulations will be developed in the next sections, we

763: allow here the use of a slightly informal language.

764: 

765: \subsection*{2.1.~Strategy per absurdum} Let $M$, $S$, and $C$ 

766: be as in Theorem~1.2. It is essentially known that both the CR- and

767: the $L^{\sf p}$-removability of $C$ are a (relatively mild)

768: consequence of the $\mathcal{ W}$-removability of $C$ ({\it see}\,

769: \S3.14 and Section~11 below). Thus, we shall describe in this section

770: only the $\mathcal{ W}$-removability of $C$. 

771: 

772: First of all, as $M$ is globally minimal, it may be proved that for

773: every closed subset $C' \subset C$, the complement $M\backslash C'$ is

774: also globally minimal ({\it see}\, Lemma~3.5 below). As $M$ is of

775: codimension one in $\C^2$, a wedge attached to $M\backslash C$ is

776: simply a connected one-sided neighborhood of $M\backslash C$ in

777: $\C^2$. Let us denote such a one-sided neighborhood by $\omega_1$. The

778: goal is to prove that there exists a one-sided neighborhood $\omega$

779: attached to $M$ to which holomorphic functions in $\omega_1$ extend

780: holomorphically. By the definition of $\mathcal{ W}$-removability,

781: this will show that $C$ is $\mathcal{ W}$-removable.

782: 

783: Reasoning by contradiction, we shall denote by $C_{\rm nr}$ the

784: smallest {\sl nonremovable} subpart of $C$. By this we mean that

785: holomorphic functions in $\omega_1$ extend holomorphically to a

786: one-sided neighborhood $\omega_2$ of $M\backslash C_{\rm nr}$ in

787: $\C^2$ and that $C_{\rm nr}$ is the smallest subset of $C$ such that

788: this extension property holds. If $C_{\rm nr}$ is empty, the

789: conclusion of Theorem~1.2 holds, gratuitously: nothing has to be

790: proved. If $C_{ \rm nr}$ is nonempty, to come to an absurd, it

791: suffices to show that at least one point of $C_{\rm nr}$ is {\sl

792: locally removable}. By this, we mean that there exists a local

793: one-sided neighborhood $\omega_3$ of at least one point of $C_{\rm

794: nr}$ such that holomorphic functions in $\omega_2$ extend

795: holomorphically to $\omega_3$. In fact, the choice of such a point

796: will be the most delicate and the most tricky part of the proof.

797: 

798: In order to be in position to apply the continuity principle, we now

799: deform slightly $M$ inside the one-sided neighborhood $\omega_2$,

800: keeping $C_{\rm nr}$ fixed, getting a hypersurface $M^d$ (with $d$

801: like ``{\sl d}eformed'') satisfying $M^d\backslash C_{\rm nr} \subset

802: \omega_2$. We notice that a local one-sided neighborhood of $M^d$ at

803: one point $p$ of $C_{\rm nr}$ always contains a local one-sided

804: neighborhood of $M$ at $p$ (the reader may draw a figure), so we may

805: well work on $M^d$ instead of working on $M$ (however, the analogous

806: property about wedges over deformed generic submanifolds is untrue in

807: codimension $\geq 2$, {\it see}\, \S3.16 below, where supplementary

808: arguments are needed).

809: 

810: Replacing the notation $C_{\rm nr}$ by the notation $C$, the notation

811: $M^d$ by the notation $M$ and the notation $\omega_2$ by the notation

812: $\Omega$, we see that Theorem~1.2 is reduced to the following main

813: proposition, whose formulation is essentially analogous to that

814: of Theorem~1.2, except that it

815: suffices to remove at least one special point.

816: 

817: \def\theproposition{2.2}\begin{proposition}

818: Let $M$ be a $\mathcal{ C}^{ 2, \alpha}$-smooth globally minimal

819: hypersurface in $\C^2$, let $S \subset M$ be a $\mathcal{ C}^{ 2,

820: \alpha}$-smooth surface which is totally real at every point. Let $C$

821: be a {\rm nonempty} proper closed subset of $S$ and assume that the

822: nontransversality condition $\mathcal{ F }_S^c \{C\}$ formulated in

823: Theorem~1.2 holds. Let $\Omega$ be an arbitrary neighborhood of $M

824: \backslash C$ in $\C^n$. Then there exists a special point $p_{\rm sp}

825: \in C$ and there exists a local one-sided neighborhood $\omega_{ p_{

826: \rm sp}}$ of $M$ in $\C^2$ 

827: at $p_{\rm sp}$ such that holomorphic functions in

828: $\Omega$ extend holomorphically to $\omega_{ p_{ \rm sp}}$.

829: \end{proposition}

830: 

831: \subsection*{2.3.~Holomorphic extension to a half-one-sided 

832: neighborhood of $M$} The choice of the special point $p_{\rm sp}$ will

833: be achieved in two main steps. According to the nontransversality

834: assumption $\mathcal{ F}_S^c\{C\}$, there exists a characteristic

835: segment $\gamma: [-1, 1] \to S$ with $\gamma(-1)\not\in C$, with

836: $\gamma(0)\in C$ and with $\gamma(1)\not \in C$ such that $C$ lies in

837: one (closed, semi-local) side of $\gamma$ in $S$. As $\gamma$ is a

838: Jordan arc, we may orient $S$ in $M$ along $\gamma$, hence we may

839: choose a semi-local open side $(S_\gamma)^+$ of $S$ in $M$ along

840: $\gamma$. In the first main step (to be conducted in Section~4 in the

841: context of the general codimensional case Theorem~1.2'), we shall

842: construct what we call a {\sl semi-local half-wedge $\mathcal{

843: HW}_\gamma^+$ attached to $(S_\gamma)^+$ along $\gamma$}. By this, we

844: mean the ``half part'' of a wedge attached to a neighborhood of the

845: characteristic segment $\gamma$ in $M$, which yields a wedge attached

846: to the semi-local one-sided neighborhood $(S_\gamma)^+$. For an

847: illustration, {\it see}\, {\sc Figure~6} below, in which one should

848: replace the notation $M^1$ by the notation $S$. Such a half-wedge may

849: be interpreted as a wedge attached to a neighborhood of $\gamma$ in

850: $S$ which is not arbitrary, but should satisfy a further property:

851: locally in a neighborhood of every point of $\gamma$, either the

852: half-wedge contains $(S_\gamma)^+$ or one of its two ribs contains

853: $(S_\gamma)^+$, as illustrated in {\sc Figure~6} below. Importantly

854: also, the cones of this attached half-wedge should vary continuously

855: as we move along $\gamma$, {\it cf.} again {\sc Figure~6}.

856: 

857: The way how we will construct this half-wedge $\mathcal{ HW

858: }_\gamma^+$ is as follows. As illustrated in {\sc Figure~1} just

859: below, we shall first construct a string of analytic discs $Z_{ r :s }

860: ( \zeta)$, where $r$ is the approximate radius of $Z_{r: s}( \partial

861: \Delta)$, whose boundaries are contained in $(S_\gamma )^+ \subset M$

862: and which touch the curve $\gamma$ only at the point $\gamma (s)$, for

863: every $s \in [-1, 1]$, namely $Z_{r: s} (1) = \gamma (s)$ and $Z_{r

864: :s}\left(\partial \Delta \backslash \{ 1\} \right) \subset (S_\gamma

865: )^+$.

866: 

867: \bigskip

868: \begin{center}

869: \input disc-string.pstex_t

870: \end{center}

871: 

872: From now on, we fix a small radius $r_0$. By deforming the discs

873: $Z_{r_0:s}(\zeta)$ in $\Omega$ near their opposite points

874: $Z_{r_0:s}(-1)$, which lie at a positive distance from the singularity

875: $C$, we shall construct in Section~4 a family of analytic disc

876: $Z_{r_0,t:s}(\zeta)$, where $t\in \R$ is a small parameter, so that

877: the disc boundaries $Z_{r_0,t:s}(\partial \Delta)$ are pivoting

878: tangentially to $S$ at the point $\gamma(s)\equiv Z_{r_0,t:s}(1)$,

879: which remains fixed as $t$ varies. Precisely, we mean that

880: $\frac{\partial Z_{r_0,t:s}}{ \partial \theta}(1)\in T_{\gamma(s)}S$

881: and that the mapping $t\longmapsto \frac{\partial Z_{r_0,t:s}}{

882: \partial \theta}(1)$ is of rank $1$ at $t=0$. This construction and

883: the next ones will be achieved thanks to the solvability Bishop's

884: equation. Furthermore, we may add a small translation parameter $\chi

885: \in \R$, getting a family $Z_{r_0, t,\chi :s} (\zeta)$ with the

886: property that the mapping $(\chi, s) \longmapsto Z_{ r_0,t, \chi

887: :s}(1)\in S$ is a diffeomorphism onto a neighborhood of $\gamma( [-1,

888: 1])$ in $S$, still with the property that the point $Z_{r_0,t,\chi

889: :s}(1)$ is fixed equal to the point $Z_{r_0, 0,\chi:s}(1)$ as $t$

890: varies. Finally, we may add a small translation parameter $\nu\in \R$

891: with $\nu >0$, getting a family $Z_{r_0,t,\chi,\nu:s}(\zeta)$ with

892: $Z_{r_0,t,\chi,0:s}(\zeta)\equiv Z_{r_0,t,\chi:s}(\zeta)$, such that

893: the mapping $(\chi,\nu,s)\longmapsto Z_{r_0,t,\chi,\nu:s}(1)$ is a

894: diffeomorphism onto the semi-local one-sided neighborhood

895: $(S_\gamma)^+$ of $S$ along $\gamma$ in $M$, provided $\nu >0$. Then

896: the semi-local attached half-wedge may be defined as

897: \def\theequation{2.4}\begin{equation}

898: \mathcal{ HW}_\gamma^+:=\left\{

899: Z_{r_0,t,\chi,\nu:s}(\rho): \

900: \vert t \vert < \varepsilon, \

901: \vert \chi \vert < \varepsilon, \ 

902: 0< \nu < \varepsilon, \

903: 1-\varepsilon < \rho < 1, \

904: -1\leq s \leq 1

905: \right\},

906: \end{equation}

907: for some small $\varepsilon >0$. In the first main technical step (to

908: be conducted in Section~4 below in the context of Theorem~1.2'), we

909: shall show that every holomorphic function $f \in \mathcal{ O} (

910: \Omega)$ extends holomorphically to $\mathcal{ HW }_\gamma^+$. To

911: prove Proposition~2.2, we shall find a special point $p_{\rm sp} \in

912: C$ such that there exists a local one-sided neighborhood $\omega_{

913: p_{\rm sp}}$ at $p_{\rm sp}$ such that holomorphic functions in

914: $\Omega \cup \mathcal{ HW }_\gamma^+$ extend holomorphically to

915: $\omega_{ p_{\rm sp}}$.

916: 

917: \subsection*{2.5.~Field of cones on $S$}

918: We continue the description of the proof of Theorem~1.2 with the full

919: family of analytic discs $Z_{ r_0, t, \chi, \nu:s} (\zeta)$. Thanks to

920: a technical application of the implicit function theorem, we can

921: arrange from the beginning that the vectors $\frac{ \partial Z_{r_0,

922: t, \chi,0: s} }{ \partial \theta } (1)$ are tangent to $S$ at the

923: point $Z_{ r_0, 0,\chi,0:s} (1) \in S$ when $t$ varies, for all fixed

924: $s$. Then by construction, the disc boundaries

925: $Z_{r_0, t, \chi, 0:s}( \partial \Delta)$ are pivoting tangentially to

926: $S$ at the point $Z_{ r_0, t, \chi, 0 :s} (1) \equiv Z_{r_0, 0, \chi,

927: 0:s} (1)$. It follows that when $t$ varies, the oriented half-lines

928: $\R^+ \cdot \frac{ \partial Z_{ r_0,t,\chi,0 : s} }{ \partial \theta}

929: (1)$ describe an open infinite oriented cone in the tangent space to

930: $S$ at the point $Z_{r_0, 0, \chi, 0:s}(1)$. Consequently, we may

931: define a {\sl field of cones} $p \mapsto {\sf C}_p$ as

932: \def\theequation{2.6}\begin{equation}

933: {\sf C}_p:= \left\{ \R^+ \cdot 

934: \frac{ \partial Z_{r_0,t,\chi,0: s} }{

935: \partial \theta} (1) : \ \vert

936: t \vert < \varepsilon \right\},

937: \end{equation}

938: at every point $p=Z_{ r_0, 0, \chi, 0: s} (1) \in S$ of a neighborhood

939: of $\gamma$ in $S$. The following figure provides an intuitive

940: illustration. One should think that the small cones are generated when

941: the small discs boundaries of {\sc Figure~1} pivote tangentially to

942: $S$.

943: 

944: \bigskip

945: \begin{center}

946: \input field-cones-hypersurface.pstex_t

947: \end{center}

948: 

949: After having defined this field of cones, we shall {\sl fill} the

950: cones as follows. Remind that a neighborhood of $\gamma$ in $S$ is

951: foliated by characteristic segments, which are approximatively

952: parallel to $\gamma$. In {\sc Figure~2} above, one should think that

953: the characteristic foliation is horizontal. So there exists a nowhere

954: vanishing vector field $p \mapsto X_p$ defined in a neighborhood of

955: $\gamma$ whose integral curves are characteristic segments. We define

956: the {\sl filled cone} ${\sf FC}_p$ by

957: \def\theequation{2.7}\begin{equation}

958: {\sf FC}_p:=\left\{

959: \lambda \cdot X_p + (1-\lambda) \cdot v_p : \ 

960: 0 \leq \lambda < 1, \

961: v_p\in {\sf C}_p

962: \right\}.

963: \end{equation}

964: Geometrically, we rotate every half-line $\R^+ \cdot v_p$ towards the

965: characteristic half-line $\R^+ \cdot X_p$ and we call the result the

966: {\sl filling} of ${\sf C}_p$. In {\sc Figure~2} above, all the cones

967: ${\sf C}_p$ coincide with their fillings. Thus we have constructed a

968: field of filled cones $p \longmapsto {\sf FC}_p$ over a neighborhood

969: of $\gamma$ in $S$.

970: 

971: \subsection*{2.8.~Small analytic discs half-attached to $S$}

972: The next main observation is that small analytic discs which are

973: half-attached to $S$ are essentially contained in the half-wedge

974: $\mathcal{ HW }_\gamma^+$, provided that they are approximatively

975: directed by the filled cone ${\sf FC}_p$. Let us be more precise. Let

976: $\partial^+ \Delta:= \{\zeta \in \partial \Delta: \ {\rm Re} \, \zeta

977: \geq 0 \}$ denote the {\sl positive half part} of the unit circle. We

978: say that an analytic disc $A: \overline{ \Delta } \to \C^2$ is {\sl

979: half-attached} to $S$ if $A( \partial^+ \Delta)$ is contained in

980: $S$. Here, $A$ is at least of class $\mathcal{ C}^1$ over

981: $\overline{\Delta}$ and holomorphic in $\Delta$. In addition, we shall

982: always assume that our discs $A$ are embeddings of $\overline{\Delta}$

983: into $\C^2$. We shall say that $A$ is {\sl approximatively straight}

984: (in an informal sense) if $A(\Delta)$ is close in $\mathcal{

985: C}^1$-norm to an open subset of the complex line generated by the

986: complex vector $\frac{\partial A}{ \partial \zeta}(1)$. Finally, we

987: say that $A$ is {\sl approximatively directed by the filled cone ${\sf

988: FC}_p$ at $p=A(1)$}, if the vector $\frac{\partial A}{\partial

989: \theta}(1)\in T_pS$ belongs to ${\sf FC}_p$. Although this

990: terminology will not be re-employed in the next sections, we may

991: formulate a crucial geometric observation as follows.

992: 

993: \def\thelemma{2.9}\begin{lemma}

994: A sufficiently small approximatively straight analytic disc $A:

995: \overline{ \Delta} \to \C^2$ of class at least $\mathcal{ C}^1$ which

996: is half-attached to $S$ and which is approximatively directed by the

997: filled cone ${\sf FC}_p$ at $p=A(1)\in S$, necessarily satisfies

998: \def\theequation{2.10}\begin{equation}

999: A\left(\overline{ \Delta } \backslash \partial^+ \Delta\right) 

1000: \subset \mathcal{ HW}_\gamma^+.

1001: \end{equation}

1002: \end{lemma}

1003: 

1004: In the context of the general Theorem~1.2',

1005: this property (with more precisions) will be checked in 

1006: Section~8 below. 

1007: 

1008: \subsection*{2.11.~Choice of a special point}

1009: In the second main step of the proof (to be conducted in Section~5 in

1010: the context of Theorem~1.2'), we shall choose the desired special

1011: point $p_{ \rm sp}$ of Proposition~2.2 to be removed locally as

1012: follows. Since we shall remove $p_{ \rm sp}$ by means of half-attached

1013: analytic discs (applying the continuity principle), we want to find a

1014: special point $p_{ \rm sp}\in C$ so that the following two conditions

1015: hold true:

1016: \begin{itemize}

1017: \item[{\bf (i)}]

1018: There exists a small approximatively straight analytic disc $A:

1019: \overline{\Delta} \to \C^2$ with $A(1)= p_{\rm sp}$ which is

1020: half-attached to $S$ such that $A$ is approximatively directed by the

1021: filled cone ${\sf FC}_{ p_{ \rm sp}}$ (so that the conclusion of

1022: Lemma~2.9 above holds true).

1023: \item[{\bf (ii)}]

1024: The same disc satisfies $A \left( \partial^+ \Delta \backslash \{1\}

1025: \right) \subset S \backslash C$.

1026: \end{itemize}

1027: 

1028: In particular, since $M \backslash C$ is contained in $\Omega$, it

1029: follows from these two conditions that the blunt disc boundary

1030: $A\left( \partial \Delta \backslash \{ 1 \} \right)$ is contained in

1031: the open subset $\Omega \cup \mathcal{ HW }_\gamma^+$, a property that

1032: will be appropriate for the application of the continuity principle,

1033: as we shall explain in Section~9 below.

1034: 

1035: To fulfill conditions {\bf (i)} and {\bf (ii)} above, we first

1036: construct a supporting real segment at a special point of the nonempty

1037: closed subset $C\subset S$.

1038: 

1039: \def\thelemma{2.12}\begin{lemma}

1040: There exists at least one special point $p_{ \rm sp } \in C$

1041: arbitrarily close to $\gamma$ in a neighborhood of which the following

1042: two properties hold true{\rm :}

1043: \begin{itemize}

1044: \item[{\bf (i')}]

1045: There exists a small $\mathcal{ C}^{2, \alpha}$-smooth open segment

1046: $H_{p_{\rm sp}}\subset S$ passing through $p_{ \rm sp}$ such that an

1047: oriented tangent half-line to $H_{p_{\rm sp}}$ at $p_{\rm sp}$ is

1048: contained in the filled cone ${\sf FC}_{p_{\rm sp}}$, as illustrated

1049: in {\sc Figure~3} below.

1050: \item[{\bf (ii')}]

1051: The same segment is a supporting segment in the following

1052: sense{\rm :} locally in a

1053: neighborhood of $p_{\rm sp}$, the set $C \backslash \{p_{\rm sp}\}$ is

1054: contained in one open side $(H_{ p_{\rm sp}})^-$ if $H_{p_{\rm sp}}$

1055: in $S$, as illustrated in {\sc Figure~3} just below.

1056: \end{itemize}

1057: \end{lemma}

1058: 

1059: \bigskip

1060: \begin{center}

1061: \input one-touch.pstex_t

1062: \end{center}

1063: 

1064: The way how we prove Lemma~2.12 is illustrated intuitively in {\sc

1065: Figure~2} above. For $\lambda\in \R$ with $0 \leq \lambda < 1$ very

1066: close to $1$, the vector field $p \longmapsto v_p^\lambda := \lambda

1067: \cdot X_p + (1-\lambda) \cdot v_p$ is very close to the characteristic

1068: vector field $p\mapsto X_p$. By construction, this vector field runs

1069: into the filled field of cones $p\mapsto {\sf FC}_p$. In {\sc

1070: Figure~2}, the integral curves of $p\mapsto v_p^\lambda$ are drawn as

1071: dotted lines, which are almost horizontal if $\lambda$ is very close

1072: to $1$. If we choose the first dotted integral curve from the lower

1073: part of {\sc Figure~2} which touches $C$ at one special point $p_{\rm

1074: sp}\in C$ and if we choose for $H_{p_{\rm sp}}$ a small segment of

1075: this first dotted integral curve, we may check that properties {\bf

1076: (i)} and {\bf (ii)} are satisfied, modulo some mild technicalities. A

1077: rigorous complete proof of Lemma~2.12 will be provided in Section~5

1078: below.

1079: 

1080: \subsection*{2.13.~Construction of analytic discs

1081: half-attached to $S$} Small analytic discs which are half-attached to

1082: a $\mathcal{ C}^{2,\alpha}$-smooth maximally real submanifold $M^1$ of

1083: $\C^n$ and which are approximatively straight will be constructed in

1084: Section~7 below. For this, we shall use the solution of Bishop's

1085: equation with parameters in H\"older spaces, obtained by A.~Tumanov

1086: in~\cite{ tu3} with an optimal loss of regularity. In Section~8, we

1087: shall check that it follows from the general constructions of

1088: Section~7 that there exists a small analytic disc $A$ half-attached to

1089: $S$ with $A(1)= p_{\rm sp}$ whose half boundary is tangent to

1090: $H_{p_{\rm sp}}$ at $p_{\rm sp}$ and which satisfies property {\bf

1091: (ii)} above, as drawn in {\sc Figure~3} above. Thus, the two geometric

1092: properties {\bf (i')} and {\bf (ii')} satisfied by the real segment

1093: $H_{p_{\rm sp}}$ may be realized by the half-boundary of a

1094: half-attached analytic disc.

1095: 

1096: \subsection*{2.14.~Translation of half-attached and continuity 

1097: principle} By means of the results of Section~7, we shall see that we

1098: may include the disc $A (\zeta)$ is a parametrized family $A_{ x,v}

1099: (\zeta)$ of analytic discs half-attached to $S$, where $x\in \R^2$ and

1100: $v \in \R$ are small, so that the mapping $x \mapsto A_{ x,0} (1) \in

1101: S$ is a local diffeomorphism onto a neighborhood of $p_{ \rm sp}$ in

1102: $S$ and so that the mapping $v \mapsto \frac{ \partial A_{ 0,v }}{

1103: \partial \theta} (1)$ is of rank $1$ at $v=0$. Furthermore, we

1104: introduce a new parameter $u \in \R$ in order to ``translate'' the

1105: totally real surface $S$ in $M$ by means of a family of $S_u \subset

1106: M$ with $S_0=S$ and $S_u \subset ( S_\gamma )^+$ for $u>0$. Thanks to

1107: the tools developed in Section~7, we deduce that there exists a

1108: deformed family of analytic discs $A_{ x, v, u} (\zeta)$ which are

1109: half-attached to $S_u$ and which satisfies $A_{x, v,0} (\zeta) \equiv

1110: A_{ x,v} (\zeta)$. In particular, this family covers a local

1111: one-sided neighborhood $\omega_{p_{\rm sp}}$ of $M$ at $p_{\rm sp}$

1112: defined by

1113: \def\theequation{2.15}\begin{equation}

1114: \omega_{p_{\rm sp}}:= \left\{

1115: A_{x,v,u}(\rho): \ 

1116: \vert x \vert < \varepsilon, \ 

1117: \vert v \vert < \varepsilon, \

1118: \vert u \vert < \varepsilon, \ 

1119: 1-\varepsilon < \rho < 1

1120: \right\},

1121: \end{equation}

1122: for some $\varepsilon>0$.

1123: 

1124: In the third and last main step of the proof (to be conducted in

1125: Section~9 below), we shall prove that every disc $A_{x, v,u} (\zeta)$

1126: with $u \neq 0$ is analytically isotopic to a point with the boundary

1127: of every disc of the isotopy being contained in $\Omega \cup \mathcal{

1128: HW}_\gamma^+$. Thanks to the continuity principle, we shall deduce

1129: that every holomorphic function $f \in \mathcal{ O}\left( \Omega \cup

1130: \mathcal{ HW}_\gamma^+ \right)$ extends holomorphically to $\omega_{

1131: p_{\rm sp}}$ minus a certain thin closed subset $\mathcal{ C}_{p_{\rm

1132: sp}}$ of $\omega_{ p_{\rm sp}}$. Finally, we shall conclude both the

1133: proof of Proposition~2.2 and the proof of Theorem~1.2 by checking that

1134: the thin closed set $\mathcal{ C}_{p_{\rm sp}}$ is in fact removable

1135: for holomorphic functions defined in $\omega_{ p_{\rm sp }} \backslash

1136: \mathcal{ C}_{ p_{ \rm sp}}$.

1137: 

1138: \subsection*{2.16.~Organization of the paper} 

1139: As was already announced, Sections~3, 4, 5, 6, 7, 8 and~9 below will be

1140: entirely devoted to the proof of Theorem~1.2', which will be

1141: endeavoured directly in arbitrary codimension, without any further

1142: reference to the hypersurface version. Only in Section~3 shall we also

1143: consider the beginning of the proof of Theorem~1.4. During the

1144: development of the proof of Theorem~1.2', in comparison to the quick

1145: description of the proof of Theorem~1.2 achieved just above, we shall

1146: unavoidably encounter some supplementary technical complications caused

1147: by the codimension being 

1148: $\geq 2$, namely technicalities which are absent in

1149: codimension $1$. We would like to mention that the crucial geometric

1150: argument which enables us to choose the desired special point will be

1151: conducted in the central Section~5 below.

1152: 

1153: Then Section~10 is devoted to summarize three geometrically distinct

1154: proofs of Theorem~1.4. In Section~11, we check that both the CR- and

1155: the $L^{\sf p}$-removability of $C$ are a consequence of the

1156: $\mathcal{ W}$-removability of $C$. In Section~12, we

1157: provide the proof of Theorem~1.1, of Theorem~1.3 and of further

1158: applications. This Section~12 may be read before entering the proof of

1159: Theorem~1.2'. Finally, in Section~13, we provide some applications of

1160: our removability results to the edge of the wedge theorem for

1161: meromorphic functions.

1162: 

1163: \section*{\S3.~Strategy per absurdum

1164: for the proofs of Theorems~1.2' and~1.4} 

1165: 

1166: \subsection*{3.1.~Preliminary}

1167: For the proof of Theorems~1.2', as in \cite{cs}, \cite{m2},

1168: \cite{mp1}, \cite{mp3}, \cite{p1}, we shall proceed by contradiction.

1169: This strategy possesses a considerable advantage: it will not be

1170: necessary to control the size of the local subsets of $C$ that are

1171: progressively removed, which simplifies substantially the presentation

1172: and the understandability of the reasonings. We shall explain how to

1173: reduce CR- and $L^{\sf p}$-removability of $C$ to its $\mathcal{

1174: W}$-removability. Also, it may be argued that the $\mathcal{

1175: W}$-removability of $C$ can be reduced to the simpler case where the

1176: functions which we have to extend are even holomorphic in a

1177: neighborhood of $M\backslash C$ in $\C^n$. Whereas such a strategy is

1178: essentially carried out in detail in previous references (with some

1179: variations), we shall for completeness recall the complete reasonings

1180: briefly here, in \S3.2 and in \S3.16 below.

1181: 

1182: \subsection*{3.2.~Global minimality of some complements} 

1183: For background notions about CR orbits in CR manifolds, we refer the

1184: reader to~\cite{su}, \cite{j2}, \cite{mp1}, \cite{j4}. We just recall

1185: a few standard facts: if $p$ belongs to a generic submanifold of

1186: $\C^n$ of class at least $\mathcal{ C}^2$, a point $q\in M$ belongs to

1187: the CR orbit $\mathcal{ O}_{CR}(M,p)$ if and only if there exists a

1188: piecewise smooth curve $\lambda: [0,1] \to M$ with $\lambda(0)=p$,

1189: $\lambda(1)=q$ such that $d\lambda(s)/ds\in T_{\gamma(s)}^cM

1190: \backslash \{0\}$ at every $s\in [0,1]$ at which $\lambda$ is

1191: differentiable; CR orbits make a partition of $M$; CR orbits are

1192: immersed $\mathcal{ C}^{1,\alpha}$-smooth submanifolds of $M$,

1193: according to H.J.~Sussmann's Theorem~4.1 in~\cite{su} specialized in

1194: the CR category; Every maximal $T^cM$-integral immersed submanifold of

1195: $M$ must contain the CR orbit of each of its points; and finally, a

1196: trivial, but often useful fact: if $N$ is a $T^cM$-integral

1197: submanifold of $M$, namely $T_pN \subset T_p^cM$ for every point $p\in

1198: N$, then the local flow of every $T^cM$-tangent vector field on $M$

1199: stabilizes locally $N$.

1200: 

1201: In the two geometric situations of Theorems~1.2' and~1.4, we shall

1202: apply the following two ~Lemmas~3.3 and~3.5 about the CR structure of

1203: the complement $M\backslash C'$, where $C'\subset C\subset M^1$ is an

1204: arbitrary proper closed subset of $C$.

1205: 

1206: \def\thelemma{3.3}\begin{lemma} Let $M$ be a $\mathcal{ C}^{2,

1207: \alpha}$-smooth generic submanifold of $\C^n$ $(n\geq 2)$ of

1208: codimension $d\geq 1$ and of CR dimension $m:=n-d\geq 1$, let $M^1$ be

1209: a $\mathcal{ C}^{2,\alpha}$-smooth 

1210: one-codimensional submanifold of $M$ which is generic in $\C^n$ and

1211: let $C'$ be an arbitrary proper closed subset of $M^1$. If either

1212: \begin{itemize}

1213: \item[{\bf (1)}]

1214: $M$ is of CR dimension $m=1$ and the condition 

1215: $\mathcal{ F}_{M^1}^c\{C'\}$ of Theorem~1.2' holds; or if

1216: \item[{\bf (2)}]

1217: $M$ is of CR dimension $m\geq 2$ and the condition 

1218: $\mathcal{ O}_{M^1 }^{CR }\{C'\}$ of Theorem~1.4 holds,

1219: \end{itemize}

1220: then for each point $p'\in C'$, there exists a piecewise $\mathcal{

1221: C}^{2, \alpha}$-smooth curve $\gamma: [0,1]\to M^1$ satisfying $d

1222: \gamma(s)/ds\in T_{\gamma(s) }M^1 \cap T_{\gamma(s)}^c M\backslash

1223: \{0\}$ at every $s\in [0,1]$ at which $\lambda$ is differentiable,

1224: such that $\gamma(0)=p'$ and $\gamma(1)$ does not belong to $C'$.

1225: \end{lemma}

1226: 

1227: \proof

1228: In the case $m=1$, we proceed by contradiction and we suppose that

1229: there exists a point $p'\in C'$ such that all piecewise $\mathcal{

1230: C}^{2,\alpha}$-smooth curves $\gamma: [0,1]\to M^1$ with $d\gamma(s)

1231: /ds\in T_{\gamma(s)}M^1\cap T_{\gamma(s)}^cM\backslash \{0\}$ at every

1232: $s\in [0,1]$ at which $\gamma$ is differentiable, which have origin

1233: $p'$ are entirely contained in $C'$. Notice that since the bundle

1234: $TM^1\cap T^cM\vert_{M^1}$ is of real dimension one and of class

1235: $\mathcal{ C}^{1,\alpha}$, such curves $\gamma$ are in fact $\mathcal{

1236: C}^{2,\alpha}$-smooth at every point. It follows immediately there

1237: cannot exist a curve $\gamma: [-1,1] \to M^1$ contained in a single

1238: leaf of the characteristic foliation $\mathcal{ F}_{M^1}^c$ with

1239: $\gamma(0)\in C'$ and $\gamma(-1), \gamma(1)\not \in C'$, which

1240: contradicts the condition $\mathcal{ F}_{M^1}^c\{C'\}$.

1241: 

1242: In the case $m\geq 2$, by genericity of $M^1$, the complex tangent

1243: bundle $T^cM^1$, which is of real dimension $(2m-2)$, is a

1244: one-codimensional subbundle of the $(2m-1)$-dimensional bundle

1245: $TM^1\cap T^cM\vert_{M^1}$, namely $T^cM^1\subset TM^1 \cap

1246: T^cM\vert_{M^1}$. Let $p\in C'$ be an arbitrary point. By the

1247: assumption $\mathcal{ O}_{M^1}^{CR}\{C'\}$, the CR orbit of $p'$ is

1248: not contained in $C'$. Equivalently, there exists a piecewise

1249: $\mathcal{ C}^{2,\alpha}$-smooth curve $\gamma: [0,1] \to M^1$

1250: satisfying

1251: \def\theequation{3.4}\begin{equation}

1252: d\gamma(s)/ds\in T_{\gamma(s)}^cM^1\backslash \{0\}\subset

1253: T_{\gamma(s)}M^1 \cap T_{\gamma(s)}^cM\backslash \{0\} 

1254: \end{equation} 

1255: at each $s\in [0,1]$ at which $\gamma$ is

1256: differentiable,

1257: such that $\gamma(0)=p'$ and $\gamma(1)$ does not belong to $C'$.

1258: Hence the conclusion of Lemma~3.3 is immediately satisfied.

1259: This completes the proof.

1260: \endproof

1261: 

1262: As an application, we deduce that under the respective assumptions

1263: $\mathcal{ F}_{M^1}^c\{C\}$ and $\mathcal{ O}_{M^1}^{CR}\{C\}$ of

1264: Theorems~1.2' and of Theorem~1.4, the complement $M \backslash C'$ is

1265: globally minimal, for every closed subset $C'\subset C$.

1266: 

1267: \def\thelemma{3.5}\begin{lemma} 

1268: Let $M$ be a $\mathcal{ C}^{2, \alpha}$-smooth generic submanifold of

1269: $\C^n$ $(n \geq 2)$ of codimension $d\geq 1$ and of CR dimension

1270: $m:=n-d \geq 1$, let $M^1$ be a $\mathcal{ C}^{2,\alpha}$-smooth

1271: one-codimensional submanifold of $M$ which is generic in $\C^n$ and

1272: let $C'$ be an arbitrary {\rm nonempty} proper closed subset of

1273: $M^1$. Assume that for each point $q'\in C'$, there exists a piecewise

1274: $\mathcal{ C}^{2,\alpha}$-smooth curve $\gamma: [0,1] \to M^1$ with

1275: $d\gamma(s) / ds\in T_{ \gamma(s)}M^1\cap T_{ \gamma(s)}^cM\backslash

1276: \{0\}$ at every $s\in [0,1]$ at which 

1277: $\gamma$ is differentiable, such that $\gamma(0)=q'$ and $\gamma(1)$

1278: does not belong to $C'$. Then the CR orbit in $M\backslash C'$ of

1279: every point $p\in M\backslash C'$ coincides with its CR orbit in $M$

1280: minus $C'$, namely

1281: \def\theequation{3.6}\begin{equation}

1282: \mathcal{ O}_{CR}(M\backslash C', p)= 

1283: \mathcal{ O}_{CR}(M,p)\backslash C'.

1284: \end{equation}

1285: In particular, as a corollary, if $M$ is globally minimal, then

1286: $M\backslash C'$ is also globally minimal.

1287: \end{lemma}

1288: 

1289: \proof

1290: Let us first explain the last sentence, which applies

1291: to the situations considered in both Theorems~1.2' and~1.4: 

1292: if $\mathcal{ O}_{CR}(M,p)=M$, then by~\thetag{3.6}, 

1293: $\mathcal{ O}_{CR}(M\backslash C',p)\equiv

1294: M\backslash C'$, which proves that

1295: $M\backslash C'$ is globally minimal.

1296: 

1297: To establish~\thetag{3.6}, we shall need the following crucial lemma,

1298: deserving an illustration: {\sc Figure~4} below.

1299: 

1300: \def\thelemma{3.7}\begin{lemma}

1301: Under the assumptions of Lemma~3.5, for every point $q'\in C' \subset

1302: M^1$, there exists a $\mathcal{ C}^{1,\alpha}$-smooth locally embedded

1303: submanifold $\Omega_{q'}$ of $M$ passing through $q'$ which is

1304: transverse to $M^1$ at $q'$ in $M$, namely which satisfies

1305: $T_{q'}\Omega_{q'}+T_{q'}M^1= T_{q'}M$, such that

1306: \begin{itemize}

1307: \item[{\bf (1)}]

1308: $\Omega_{q'}$ is a $T^cM$-integral submanifold, namely 

1309: $T_p^cM\subset T_p\Omega_{q'}$, for every point 

1310: $p\in \Omega_{q'}$.

1311: \item[{\bf (2)}]

1312: $\Omega_{q'}\backslash C'$ is contained in 

1313: a single CR orbit of $M$.

1314: \item[{\bf (3)}] $\Omega_{q'}$ is

1315: also contained in a single CR orbit of $M\backslash C'$.

1316: \end{itemize}

1317: \end{lemma}

1318: 

1319: \proof

1320: So, let $q'\in C'\subset M^1$. Since $M^1$ is generic in $\C^n$, there

1321: exists a $\mathcal{ C}^{1,\alpha}$-smooth vector field $Y$ defined in

1322: a neighborhood of $q'$ which is complex tangent to $M$ but locally

1323: transversal to $M^1$, {\it cf.} {\sc Figure~4} just below (for easier

1324: readability, we have erased the hatching of $C'$ in a neighborhood of

1325: $q'$).

1326: 

1327: \bigskip

1328: \begin{center}

1329: \input orbit-complement.pstex_t

1330: \end{center}

1331: 

1332: \noindent

1333: Following the integral curve of $Y$ issued from $q'$, we can define a

1334: point $q_\epsilon'$ in an $\epsilon$-neighborhood of $q'$ which does

1335: not belong to $M^1$. By assumption, there exists a piecewise smooth

1336: curve $\gamma: [0,1]\to M^1$ with $d\gamma(s) /ds \in T_{\gamma(s)}M^1

1337: \cap T_{\gamma(s)}^cM\backslash \{0\}$ at every $s\in [0,1]$ at which

1338: $\gamma$ is differentiable, such that $\gamma(0)=q'$ and $\gamma(1)$

1339: does not belong to $C'$. For simplicity, we shall assume that

1340: $\gamma$ consists of a single smooth piece, the case where $\gamma$

1341: consists of finitely many smooth pieces being treated in a completely

1342: similar way. With this assumption (which will simplify slightly the

1343: technicalities), it follows that there exists a vector field $X$

1344: defined in a neighborhood of the curve $\gamma([0,1])$ in $M$ which is

1345: complex tangent to $M$ and whose restriction to $M^1$ is a semi-local

1346: section of $TM^1\cap T^cM\vert_{M^1}$, such that $\gamma$ is an

1347: integral curve of $X$ and such that $\gamma(1)=\exp(X)(q')\in

1348: M^1\backslash C'$. In addition, we can assume that the vector field

1349: $Y$ is defined in the same neighborhood of $\gamma([0,1])$ in $M$ and

1350: everywhere transversal to $M^1$. If $\epsilon$ is sufficiently small,

1351: {\it i.e.} if $q_\epsilon'$ is sufficiently close to $q'$, the point

1352: $r_\epsilon':= \exp(X)(q_\epsilon')$ is still very close to

1353: $M^1$. Thus, we can define a new point $r'\in M^1$ to be the unique

1354: intersection of the integral of $Y$ issued from $r_\epsilon'$ with

1355: $M^1$. By choosing $\epsilon$ small enough, the point $r_\epsilon'$

1356: will be arbitrarily close to $\gamma(1)\not \in C'$, and consequently,

1357: we can assume that $r'$ also does not belong to $C'$, as drawn in {\sc

1358: Figure~4} above. Notice that the integral curve of $X$ from

1359: $q_\epsilon'$ to $r_\epsilon'$ is contained in $M\backslash M^1$,

1360: since the flow of $X$ stabilizes $M^1$, whence the two points

1361: $r_\epsilon'$ and $r'$ belong to the CR orbit $\mathcal{

1362: O}_{CR}(M\backslash C',q_\epsilon')$.

1363: 

1364: Let $\Omega_{r'}$ denote a small piece of the immersed submanifold

1365: $\mathcal{ O}_{CR}(M\backslash C',r')$ passing through $r'$. By the

1366: standard properties of CR orbits, we can assume that $\Omega_{r'}$ is

1367: an embedded $\mathcal{ C}^{1,\alpha}$-smooth submanifold of

1368: $M\backslash C'$ of the same CR dimension as $M\backslash C'$ and we

1369: can in addition assume that $\Omega_{r'}$ contains $r_\epsilon'$,

1370: provided $\epsilon$ is small enough. Since $Y$ is a vector field

1371: complex tangent to $M$, the submanifold $\Omega_{r'}$ is necessarily

1372: stretched along the flow lines of $Y$, hence it is transversal to

1373: $M^1$.

1374: 

1375: We then define the submanifold $\exp(-X)(\Omega_{r'})$, close to the

1376: point $q'$ (we shall argue in a while that it passes in fact through

1377: $q'$). Since the flow of $X$ stabilizes $M^1$, it follows that

1378: $\exp(-X)(\Omega_{r'})$ is transversal to $M^1$ and that

1379: $\exp(-X)(\Omega_{r'})$ is divided in two sides by its

1380: one-codimensional $\mathcal{ C}^{1,\alpha}$-smooth submanifold

1381: $M^1\cap \exp(-X)(\Omega_{r'})$. Furthermore, the flow of $X$

1382: stabilizes the two sides of $M^1$ in $M$, semi-locally in a

1383: neighborhood of $\gamma([0,1])$, {\it see} again {\sc Figure~4} above.

1384: Consequently, every integral curve of $X$ issued from every point in

1385: $\Omega_{r'}\backslash M^1$ stays in $M\backslash M^1$, hence in

1386: $M\backslash C'$ and it follows that the submanifold

1387: \def\theequation{3.8}\begin{equation}

1388: \exp(-X)(\Omega_{r'})\backslash M^1,

1389: \end{equation}

1390: consisting of two connected pieces, is contained in the single CR

1391: orbit $\mathcal{ O}_{CR}(M\backslash C', p')$. By the characteristic

1392: property of a CR orbit, this means that the two connected pieces of

1393: $\exp(-X)(\Omega_{r'})\backslash M^1$ are CR submanifolds of

1394: $M\backslash C'$ of the same CR dimension as $M\backslash C'$.

1395: Furthermore, since the intersection $M^1\cap \exp(-X)(\Omega_{r'})$ is

1396: one-codimensional, it follows by continuity that {\it the $\mathcal{

1397: C}^{1,\alpha}$-smooth submanifold $\exp(-X) (\Omega_{r'})$ is in fact a

1398: CR submanifold of $M$ of the same CR dimension as $M$}. This 

1399: proves property {\bf (1)}. 

1400: 

1401: Since $q_\epsilon'$ belongs to $\exp (-X) (\Omega_{r'})$ and since the

1402: flow of the complex tangent vector field $Y$ necessarily stabilizes

1403: the $T^cM$-integral submanifold $\exp( -X) (\Omega_{ r'})$, the point

1404: $q'$ which belongs to an integral curve of $Y$ issued from

1405: $q_\epsilon'$, must belong to the submanifold $\exp(-X)( \Omega_{

1406: r'})$, which we can now denote by $\Omega_{ q'}$, as in {\sc

1407: Figure~4} above.

1408: 

1409: Observe that locally in a neighborhood of $q'$, the integral curves of

1410: $Y$ are transversal to $M^1$ and meet $M^1$ only at one point.

1411: Shrinking if necessary $\Omega_{q'}$ a little bit and using positively

1412: or negatively oriented integral curves of $Y$ with origin all points

1413: in $\Omega_{q'} \backslash M^1$ lying in both sides, we deduce that

1414: $\Omega_{q'}\backslash C'$ is contained in the single CR orbit

1415: $\mathcal{ O}_{CR}(M\backslash C', r')$, which proves property {\bf

1416: (3)}. Using again $Y$ to join points of $C'\cap \Omega_{q'}$, we

1417: deduce also that $\Omega_{q'}$ is contained in the single CR orbit

1418: $\mathcal{ O}_{CR}(M,r')$, which proves property {\bf (2)}.

1419: 

1420: The proof of Lemma~3.7 is complete.

1421: \endproof

1422: 

1423: We can now prove Lemma~3.5. It suffices to establish that for every two

1424: points $p \in M \backslash C'$ and $q \in \mathcal{ O}_{ CR} (M, p)$

1425: with $q\not\in C'$, the point $q$ belongs in fact to the CR orbit of

1426: $p$ in $M\backslash C'$, namely $q$ belongs to 

1427: $\mathcal{ O}_{CR}(M\backslash C', p)$.

1428: 

1429: Since $q$ belongs to the CR orbit of $p$ in $M$, there exists a

1430: piecewise $\mathcal{ C}^{2,\alpha}$-smooth curve $\lambda: [0,1] \to

1431: M$ with $\lambda(0)=p$, $\lambda(1)=q$ and $d\lambda(s)/ds\in

1432: T_{\lambda(s)}^cM\backslash \{0\}$ at every $s\in [0,1]$ at which

1433: $\lambda$ is differentiable. For every $s$ with $0\leq s \leq 1$, we

1434: define a local $\mathcal{ C}^{1,\alpha}$-smooth submanifold

1435: $\Omega_{\lambda(s)}$ of $M$ passing through $\lambda(s)$ as follows:

1436: \begin{itemize}

1437: \item[{\bf (1)}]

1438: If $\lambda(s)$ does not belong to $C'$, choose for $\Omega_{

1439: \lambda(s)}$

1440: a piece of the CR orbit of $\lambda(s)$ in $M\backslash C'$.

1441: \item[{\bf (2)}]

1442: If $\lambda(s)$ belongs to $C'$, choose for $\Omega_{\lambda(s)}$

1443: the submanifold constructed in Lemma~3.7 above.

1444: \end{itemize}

1445: By Lemma~3.7, for each $s$, the complement

1446: $\Omega_{\lambda(s)}\backslash C'$ is contained in a single CR orbit

1447: of $M\backslash C'$. Since each $\Omega_{\lambda(s)}$ is a

1448: $T^cM$-integral submanifold, for each $s\in [0,1]$, a neighborhood of

1449: $\lambda(s)$ in the arc $\lambda([0,1])$ is necessarily contained

1450: $\Omega_{\gamma(s)}$. By the Borel-Lebesgue covering lemma, we can

1451: therefore find an integer $k\geq 1$ and real numbers

1452: \def\theequation{3.9}\begin{equation}

1453: 0=s_1<r_1<t_1<s_2<r_2<t_2<\cdots\cdots <

1454: s_{k-1} < r_{k-1} < t_{k-1} < s_k=1,

1455: \end{equation}

1456: such that

1457: $\lambda([0,1])$ is covered by $\Omega_{\lambda(0)} \cup

1458: \Omega_{\lambda(s_2)} \cup \cdots \cup \Omega_{\lambda(s_{k-1})}\cup

1459: \Omega_{\lambda(1)}$ and such that in addition, 

1460: $\lambda([r_j,t_j])\subset \Omega_{\lambda(s_j)}\cap \,

1461: \Omega_{\lambda(s_{j+1})}$ for $j=1,\dots,k-1$.

1462: 

1463: \def\thelemma{3.10}\begin{lemma}

1464: The following union minus $C'$

1465: \def\theequation{3.11}\begin{equation}

1466: \left(

1467: \Omega_{\lambda(0)} \cup \Omega_{\lambda(s_2)}\cup \cdots\cdots

1468: \cup \Omega_{\lambda(s_{k-1})}\cup

1469: \Omega_{\lambda(1)}\right)\backslash C'

1470: \end{equation}

1471: is contained in a single CR orbit of $M\backslash C'$.

1472: \end{lemma}

1473: 

1474: \proof

1475: It suffices to prove that for every $j=1,\dots,k-1$, the union $\left(

1476: \Omega_{\lambda(s_j)}\cup \Omega_{\lambda(s_{j+1})}\right) \backslash

1477: C'$ minus $C'$ is contained in a single CR orbit of $M\backslash C'$.

1478: 

1479: Two cases are to be considered. Firstly, assume that

1480: $\lambda([r_j,t_j])$ is not contained in $C'$, namely there exists

1481: $u_j$ with $r_j\leq u_j\leq t_j$ such that

1482: \def\theequation{3.12}\begin{equation}

1483: \gamma(u_j)\in \left(\Omega_{\lambda(s_j)}\cap 

1484: \Omega_{\lambda(s_{j+1})}\right)\backslash C'.

1485: \end{equation}

1486: Because $\Omega_{\lambda(s_j)}\backslash C'$ and

1487: $\Omega_{\lambda(s_{j+1})} \backslash C'$ are both contained in a

1488: single CR orbit of $M\backslash C'$, it follows from~\thetag{3.12}

1489: that they are contained in the same CR orbit of $M\backslash C'$, as

1490: desired.

1491: 

1492: Secondly, assume that $\lambda([r_j,t_j])$ is contained in

1493: $C'$. Choose $u_j$ arbitrary with $r_j\leq u_j \leq t_j$. By

1494: construction, $\lambda(u_j)$ belongs to $\Omega_{\lambda(s_j)}\cap

1495: \Omega_{\lambda(s_{j+1})}$ and both $\Omega_{\lambda(s_j)}$ and

1496: $\Omega_{\lambda(s_{j+1})}$ are $T^cM$-integral submanifolds of $M$

1497: passing through the point $\lambda(u_j)$. Let $Y$ be a local section

1498: of $T^cM$ defined in a neighborhood of $\lambda(u_j)$ which is not

1499: tangent to $M^1$ at $\lambda(u_j)$. On the integral curve of $Y$

1500: issued from $\lambda(u_j)$, we can choose a point

1501: $\lambda(u_j)_\epsilon$ arbitrarily close to $\lambda(u_j)$ which does

1502: not belong to $C'$. Since $Y$ is a section of $T^cM$, it is tangent to

1503: both $\Omega_{\lambda(s_j)}$ and $\Omega_{\lambda(s_{j+1})}$, hence

1504: we deduce that

1505: \def\theequation{3.13}\begin{equation}

1506: \gamma(u_j)_\epsilon\in \left(\Omega_{\lambda(s_j)}\cap 

1507: \Omega_{\lambda(s_{j+1})}\right)\backslash C'.

1508: \end{equation}

1509: Consequently, as in the first case, it follows that

1510: $\Omega_{\lambda(s_j)}\backslash C'$ and $\Omega_{\lambda(s_{j+1})}

1511: \backslash C'$ are both contained in the same CR orbit of $M\backslash

1512: C'$, as desired. This completes the proof of Lemma~3.10.

1513: \endproof

1514: 

1515: Since $p$ and $q$ belong to the set~\thetag{3.11}, we

1516: deduce that the points $p=\lambda(0)\in M\backslash C'$ and

1517: the point $q=\lambda(1)\in \mathcal{ O}_{CR}(M,p)\backslash C'$

1518: belong to the same CR orbit of $M\backslash C'$, as desired. 

1519: This completes the proof of Lemma~3.5.

1520: \endproof

1521: 

1522: \subsection*{3.14.~Reduction of CR- and of $L^{\sf p}$-removability 

1523: to $\mathcal{ W}$-removability} First of all, we remind that it

1524: follows from successive efforts of numerous mathematicians ({\it

1525: cf.}~\cite{a}, \cite{j2}, \cite{m1}, \cite{tr2}, \cite{tu1},

1526: \cite{tu2}, \cite{tu3}) that for every $\mathcal{

1527: C}^{2,\alpha}$-smooth globally minimal submanifold $M'$ of $\C^n$,

1528: there exists a wedge $\mathcal{ W}'$ attached to $M$ to which all

1529: continuous CR functions on $M$ extend holomorphically. It follows that

1530: the CR-removability of the closed subset $C\subset M^1$ claimed in

1531: Theorems~1.2' and~1.4 is an immediate consequence of its $\mathcal{

1532: W}$-removability. Based on the construction of analytic discs

1533: half-attached to $M^1$ which will be achieved in Section~7 below, we

1534: shall also be able to settle the reduction of $L^{\sf p}$-removability

1535: in the end of the paper, and we formulate a convenient lemma, whose

1536: proof is postponed to \S11 below.

1537: 

1538: \def\thelemma{3.15}\begin{lemma}

1539: Under the assumptions of Theorem~1.2' and of Theorem~1.4, if the closed subset

1540: $C\subset M^1$ is $\mathcal{ W}$-removable, then it is

1541: $L^{\sf p}$-removable, for all ${\sf p}$ with 

1542: $1\leq {\sf p}\leq \infty$.

1543: \end{lemma}

1544: 

1545: \subsection*{3.16.~Strategy per absurdum: removal of a single point 

1546: of the residual non-removable subset} Thus, it suffices to demonstrate

1547: that the closed subsets $C$ of Theorems~1.2' and~1.4 are $\mathcal{

1548: W}$-removable, {\it cf.} the definition given in Section~1. Let us fix

1549: a wedgelike domain $\mathcal{ W}_1$ attached to $M\backslash C$ and

1550: remind that all our wedgelike domains are assumed to be nonempty. Our

1551: precise goal is to establish that there exists a wedgelike domain

1552: $\mathcal{ W}_2$ attached to $M$ (including $C$) and a wedgelike

1553: domain $\mathcal{ W}_3\subset \mathcal{ W}_1 \cap \mathcal{ W}_2$

1554: attached to $M\backslash C$ such that for every holomorphic function

1555: $f\in \mathcal{ O}(\mathcal{ W}_1)$, there exists a holomorphic

1556: function $F\in \mathcal{ O}(\mathcal{ W}_2)$ which coincides with $f$

1557: in $\mathcal{ W}_3$. At first, we need some more definition.

1558: 

1559: Let $C'$ be an arbitrary closed subset of $C$. We shall say that

1560: $M\backslash C'$ {\sl enjoys the wedge extension property} if there

1561: exist a wedgelike domain $\mathcal{ W}_2'$ attached to $M\backslash

1562: C'$ and a wedgelike subdomain $\mathcal{ W}_3 \subset \mathcal{

1563: W}_1\cap \mathcal{ W}_2'$ attached to $M\backslash C$ such that, for

1564: every function $f\in\mathcal{ O}(\mathcal{ W}_1)$, there exists a

1565: function $F'\in \mathcal{ O}(\mathcal{ W}_2')$ which coincides with

1566: $f$ in $\mathcal{ W}_3$.

1567: 

1568: The notion of wedge removability can be localized as follows. Let

1569: again $C'\subset C$ be arbitrary. We shall say that a point $p' \in

1570: C'$ is {\sl locally $\mathcal{ W}$-removable with respect to $C'$} if

1571: for every wedgelike domain $\mathcal{ W}_1'$ attached to $M\backslash

1572: C'$, there exists a neighborhood $U'$ of $p'$ in $M$, there exists a

1573: wedgelike domain $\mathcal{ W}_2'$ attached to $(M \backslash C') \cup

1574: U'$ and there exists a wedgelike subdomain $\mathcal{ W}_3' \subset

1575: \mathcal{ W}_1' \cap \mathcal{ W}_2'$ attached to $M\backslash C'$

1576: such that for every holomorphic function $f \in \mathcal{ O}

1577: (\mathcal{ W}_1')$, there exists a holomorphic function $F'\in

1578: \mathcal{ O}(\mathcal{ W}_2')$ which coincides with $f$ in $\mathcal{

1579: W}_3'$.

1580: 

1581: Supppose now that $M\backslash C_1'$ and $M\backslash C_2'$ enjoy the

1582: wedge extension property, for some two closed subsets $C_1', C_2'

1583: \subset C$. Using Ayrapetian's version of the edge of the wedge

1584: theorem (\cite{a}, \cite{tu1}, \cite{tu2}), the two wedgelike domains

1585: attached to $M\backslash C_1'$ and to $M\backslash C_2'$ can be glued

1586: (after appropriate shrinking) to a wedgelike domain $\mathcal{ W}_1$

1587: attached to $M\backslash (C_1'\cap C_2')$ in such a way that

1588: $M\backslash (C_1'\cap C_2')$ enjoys the $\mathcal{ W}$-extension

1589: property. Also, if $M\backslash C'$ enjoys the wedge extension

1590: property and if $p'\in C'$ is locally $\mathcal{ W}$-removable with

1591: respect to $C'$, then again by means of the edge of the wedge theorem,

1592: it follows that there exists a neighborhood $U'$ of $p'$ in $M$ such

1593: that $(M\backslash C') \cup U'$ enjoys the wedge extension property.

1594: 

1595: Based on these preliminary remarks, we may define the following set of

1596: closed subsets of $C$:

1597: \def\theequation{3.17}\begin{equation}

1598: \mathcal{ C}:=\{C'\subset C \ \hbox{closed} \ ;

1599: \ M\backslash C' \ \text{\rm enjoys the 

1600: $\mathcal{ W}$-extension property} \}.

1601: \end{equation}

1602: Then the residual set 

1603: \def\theequation{3.18}\begin{equation}

1604: C_{\text{\rm nr}}:= \bigcap_{C' \in\mathcal{ C}} C'

1605: \end{equation}

1606: is a closed subset of $M^1$ contained in $C$. It follows from the

1607: above (abstract nonsense) considerations that $M\backslash C_{\rm nr}$

1608: enjoys the wedge extension property and that no point of $C_{\rm nr}$

1609: is locally $\mathcal{ W}$-removable with respect to $C_{\rm nr}$.

1610: Here, we may think that the letters ``nr'' abbreviate ``{\sl n}on-{\sl

1611: r}emova\-ble'', because by the very definition of $C_{\rm nr}$, none

1612: of its points should be locally $\mathcal{ W}$-removable. Notice also

1613: that $M\backslash C_{\rm nr}$ is globally minimal, thanks to

1614: Lemma~3.5.

1615: 

1616: Clearly, to establish Theorem~1.1, it is enough to show that

1617: $C_{\text{\rm nr}}=\emptyset$.

1618: 

1619: We shall argue indirectly (by contradiction) and assume that

1620: $C_{\text{\rm nr}}\neq \emptyset$. In order to derive a

1621: contradiction, it clearly suffices to show that there exists at least

1622: one point $p\in C_{\rm nr}$ which is in fact locally $\mathcal{

1623: W}$-removable with respect to $C_{\rm nr}$.

1624: 

1625: At this point, we notice that the main assumptions $\mathcal{

1626: F}_{M^1}^c\{C\}$ and $\mathcal{ O}_{M^1}^{CR}\{C\}$ of Theorem~1.2'

1627: and of Theorem~1.4 imply trivially that for every closed subset $C'$

1628: of $C$, then the condition $\mathcal{ F}_{M^1}^c\{C'\}$ and the

1629: condition $\mathcal{ O}_{M^1}^{CR}\{C'\}$ also hold true: these two

1630: assumptions are obviously stable by passing to closed subsets. In

1631: particular, $\mathcal{ F}_{M^1}^c\{C_{\rm nr}\}$ and $\mathcal{

1632: O}_{M^1}^{CR}\{C_{\rm nr}\}$ hold true. Consequently, by following a

1633: {\it per absurdum} strategy, we are led to prove two statements wich

1634: are totally similar to Theorem~1.2' and to Theorem~1.4, except that we

1635: now have only to establish that {\it a single point of $C_{\rm nr}$}

1636: is locally $\mathcal{ W}$-removable with respect to $C_{\rm nr}$. This

1637: preliminary logical consideration will simplify substantially the

1638: whole architecture of the two proofs. Another important advantage of

1639: this strategy which will not be appearant until the very end of the

1640: two proofs in Sections~9 and~10 below is that we are even allowed to

1641: select a special point $p_{\rm sp}$ of $C_{\rm nr}$ by requiring

1642: some nice geometric disposition of $C_{\rm nr}$ in a neighborhood of

1643: $p_{\rm sp}$ before removing it. Sections~4 and~5 below are devoted to

1644: such a selection.

1645: 

1646: So we are led to show that for every wedgelike domain $\mathcal{ W}_1$

1647: attached to $M\backslash C_{\rm nr}$, there exists a special point

1648: $p_{\rm sp}\in C_{\rm nr}$, there exists a neighborhood $U_{p_{\rm

1649: sp}}$ of $p_{\rm sp}$ in $M$, there exists a wedgelike domain

1650: $\mathcal{ W}_2$ attached to $\left(M\backslash C_{\rm nr}\right)\cup

1651: U_{p_{\rm sp}}$ and there exists a wedgelike domain $\mathcal{

1652: W}_3\subset \mathcal{ W}_1 \cap \mathcal{ W}_2$ attached to

1653: $M\backslash C_{\rm nr}$ such that for every holomorphic function

1654: $f\in \mathcal{ O}(\mathcal{ W}_1)$, there exists a function

1655: $F\in\mathcal{ O}(\mathcal{ W}_2)$ which coincides with $f$ in

1656: $\mathcal{ W}_3$.

1657: 

1658: A further convenient simplification of the task may be achieved by

1659: deforming slightly $M$ inside the wedge $\mathcal{ W}_1$ attached to

1660: $M\backslash C_{\rm nr}$. Indeed, by means of a partition of unity, we

1661: may perform arbitrarily small $\mathcal{ C}^{2,\alpha}$-smooth

1662: deformations $M^d$ of $M$ leaving $C_{\text{\rm nr}}$ fixed and moving

1663: $M\backslash C_{\text{\rm nr}}$ inside the wedgelike domain $\mathcal{

1664: W}_1$. Furthermore, we can make $M^d$ to depend on a single small

1665: real parameter $d\geq 0$ with $M^0=M$ and $M^d\backslash C_{\rm

1666: nr}\subset \mathcal{ W}_1$ for all $d>0$. Now, {\it the wedgelike

1667: domain $\mathcal{ W}_1$ becomes a neighborhood of $M^d$ in

1668: $\C^n$}. Let us denote by $\Omega$ this neighborhood. After some

1669: substantial technical work has been achieved, at the end of the proofs

1670: of Theorem~1.2' and~1.4 to be conducted in Sections~9 and~10 below, we

1671: shall construct a local wedge $\mathcal{ W}_{p_{\rm sp}}^d$ of edge

1672: $M^d$ at $p_{\rm sp}$ by means of small Bishop analytic discs glued to

1673: $M^d$, to $\Omega$ and to another subset (which we will call a {\sl

1674: half-wedge}, {\it see} Section~4 below) such that every holomorphic

1675: function $f\in \mathcal{ O}(\Omega)$ extends holomorphically to

1676: $\mathcal{ W}_{p_{\rm sp}}^d$. Using the stability of Bishop's

1677: equation under perturbation, we shall argue in \S9.27

1678: below that our constructions are stable under such small deformations,

1679: whence in the limit $d\rightarrow 0$, the wedges $\mathcal{ W}_{p_{\rm

1680: sp}}^d$ tend smoothly to a local wedge $\mathcal{ W}_{p_{\rm sp}}:=

1681: \mathcal{ W}_{p_{\rm sp}}^0$ of edge a neighborhood $U_{p_{\rm sp}}$

1682: of $p_{\rm sp}$ in $M^0 \equiv M$ (notice that in codimension $\geq

1683: 2$, a wedge of edge a deformation $M^d$ of $M$ does not contain in

1684: general a wedge of edge $M$, hence such an argument is needed). In

1685: addition, we shall derive univalent holomorphic extension to

1686: $\mathcal{ W}_{p_{\rm sp}}$. Finally, using again the

1687: edge of the wedge theorem to fill the space between $\mathcal{ W}_1$

1688: and $\mathcal{W}_{p_{\rm sp}}$, possibly after appropriate

1689: contractions of these two wedgelike domains, we may construct a

1690: wedgelike domain $\mathcal{ W}_2$ attached to $\left(M\backslash

1691: C\right) \cup U_{p_{\rm sp}}$ and a wedgelike domain $\mathcal{

1692: W}_3\subset \mathcal{ W}_1 \cap \mathcal{ W}_{p_{\rm sp}}$ attached to

1693: $M\backslash C$ such that for every holomorphic function $f\in

1694: \mathcal{ O}(\mathcal{ W}_1)$, there exists a function $F\in\mathcal{

1695: O}(\mathcal{ W}_2)$ which coincides with $f$ in $\mathcal{ W}_3$. In

1696: conclusion, we thus reach the desired contradiction to the definition

1697: of $C_{\text{\rm nr}}$.

1698: 

1699: As a summary of the above discussion, we have essentially shown that

1700: it suffices to prove Theorems~1.2' and~1.4 with the following two

1701: extra simplifying assumptions:

1702: \begin{itemize}

1703: \item[{\bf 1)}] 

1704: Instead of functions which are holomorphic in a

1705: wedgelike domain attached to $M\backslash C_{\rm nr}$, we consider

1706: functions which are holomorphic in a neighborhood

1707: $\Omega$ of $M\backslash

1708: C_{\rm nr}$ in $\C^{m+n}$. 

1709: \item[{\bf 2)}]

1710: Proceeding by contradiction, we

1711: have argued that it suffices to remove at least one point of $C_{\rm

1712: nr}$.

1713: \end{itemize}

1714:  

1715: Consequently, we may formulate the local statement that remains to

1716: prove: after replacing $C_{\rm nr}$ by $C$ and $M^d$ by $M$, we are

1717: led to establish the following main assertion, to which Theorems~1.2'

1718: and~1.4 are essentially reduced.

1719: 

1720: \def\thetheorem{3.19}\begin{theorem}

1721: Let $M$ be a $\mathcal{ C}^{ 2,\alpha}$-smooth globally minimal

1722: generic submanifold of $\C^n$ of CR dimension $m\geq 1$ and of

1723: codimension $d=n- m\geq 1$, let $M^1 \subset M$ be a $\mathcal{ C}^{2,

1724: \alpha}$-smooth one-codimensional submanifold which is generic in

1725: $\C^n$, and let $C$ be a {\rm nonempty} proper closed subset of $M^1$.

1726: \begin{itemize}

1727: \item[{\bf (i)}]

1728: If $m=1$, assume that the condition $\mathcal{ F}_{M^1}^c\{C\}$

1729: holds.

1730: \item[{\bf (ii)}]

1731: If $m\geq 2$, assume that the condition 

1732: $\mathcal{ O}_{M^1}^{CR}\{C\}$ holds.

1733: \end{itemize}

1734: Let $\Omega$ be an arbitrary neighborhood of $M \backslash C$ in

1735: $\C^n$. Then there exist a special point $p_{ \rm sp} \in C$, there

1736: exists a local wedge $\mathcal{ W}_{ p_{ \rm sp}}$ of edge $M$ at

1737: $p_{\rm sp}$ and there exists a subneighborhood $\Omega' \subset

1738: \Omega$ of $M \backslash C$ in $\C^n$ with $\mathcal{ W}_{p_{ \rm sp

1739: }} \cap \Omega'$ connected such that for every holomorphic function $f

1740: \in \mathcal{ O}( \Omega)$, there exists a holomorphic function $F\in

1741: \mathcal{ O} \left( \mathcal{ W}_{p_{ \rm sp}}\cup \Omega' \right)$

1742: which coincides with $f$ in $\mathcal{ W}_{ p_{\rm sp }}\cap \Omega'$.

1743: \end{theorem}

1744: 

1745: However, we remind the necessity of some supplementary arguments about

1746: the stability of our constructions under deformation. The proof of our

1747: main Theorem~3.19 will occupy Sections~4, 5, 6, 7, 8, 9 and~10 below

1748: and the deformation arguments will appear lastly in \S9.27.

1749: From now on, the main question is: {\it How to choose

1750: the special point $p_{\rm sp}$ to be removed locally}~?

1751: 

1752: \subsection*{3.20.~Choice of a special point $p\in C$ in the CR 

1753: dimension $m\geq 2$ case} In the case of CR dimension $m\geq 2$,

1754: essentially all points of $C$ can play the role of the special point

1755: $p_{\rm sp}$. However, since we want to devise a new proof of

1756: Theorem~1.4 which differs from the two proofs given in~\cite{ j4},

1757: ~\cite{p1} and in \cite{m2}, it will be convenient to choose a special

1758: point $p_1 \in M^1$ which has the property that locally in a

1759: neighborhood of $p_1$, the singularity $C\subset M^1$ lies behind a

1760: generic ``wall'' $H^1$ contained in $M^1$ and of codimension one in

1761: $M^1$. Notice that as $m\geq 2$, the dimension of a two-codimensional

1762: submanifold $H^1$ of $M$ is equal to $2m+d-2\geq n$, whence $M^1$ may

1763: perfectly be generic in $\C^n$.

1764: 

1765: \def\thelemma{3.21}\begin{lemma}

1766: Let $M$ be a $\mathcal{ C}^{2, \alpha}$-smooth globally minimal

1767: generic submanifold of $\C^n$ of CR dimension $m \geq 2$ and of

1768: codimension $d = n-m \geq 1$, let $M^1\subset M$ be a $\mathcal{

1769: C}^{2, \alpha}$-smooth one-codimensional submanifold which is generic

1770: in $\C^n$ and let $C\subset M^1$ be a {\rm nonempty} proper closed

1771: subset which does not contain any CR orbit of $M^1$. Then there

1772: exists a point $p_1\in C$ and a $\mathcal{ C}^{2,\alpha}$-smooth

1773: one-codimensional submanifold $H^1\subset M^1$ passing through $p_1$

1774: which is {\rm generic} in $\C^n$ such that $C\backslash \{p_1\}$ lies,

1775: in a neighborhood of $p_1$, in one open side $(H^1)^-$ of $H^1$ in

1776: $M^1$.

1777: \end{lemma}

1778: 

1779: \proof

1780: The proof is completely similar to the proof of

1781: Lemma~2.1 in~\cite{mp3}, {\it see} especially 

1782: {\sc Figure~1}, p.~490. 

1783: \endproof

1784: 

1785: With Lemma~3.21 at hand, we can now state a more precise version 

1786: of case {\bf (ii)} of Theorem~3.19, which will be the

1787: main removability proposition. 

1788: 

1789: \def\theproposition{3.22}\begin{proposition}

1790: Let $M$ be a $\mathcal{ C}^{2, \alpha}$-smooth globally minimal

1791: generic submanifold of $\C^n$ of CR dimension $m \geq 2$ and of

1792: codimension $d = n-m \geq 1$, let $M^1\subset M$ be a $\mathcal{

1793: C}^{2, \alpha}$-smooth one-codimensional submanifold which is generic

1794: in $\C^n$, let $p_1 \in M^1$, let $H^1\subset M^1$ be a $\mathcal{

1795: C}^{2, \alpha}$-smooth one-codimensional submanifold of $M^1$ passing

1796: through $p_1$ which is also generic in $\C^n$ {\rm (}this is possible,

1797: thanks to the assumption $m \geq 2${\rm )} and let $(H^1 )^-$ denote

1798: an open local one-sided neighborhood of $H^1$ in $M^1$. Suppose that

1799: $C \subset M^1$ is a {\rm nonempty} proper closed subset of $M^1$ with

1800: $p_1 \in C$ which is situated, locally in a neighborhood of $p_1$,

1801: only in one side of $H^1$, namely $C \subset (H^1)^- \cup \{

1802: p_1\}$. Let $\Omega$ be a neighborhood of $M \backslash C$ in

1803: $\C^n$. Then there exists a local wedge $\mathcal{ W}_{p_1}$ of edge

1804: $M$ at $p_1$ with $\Omega \cap \mathcal{ W}_{p_1}$ connected

1805: (shrinking $\Omega$ if necessary) such that for every holomorphic

1806: function $f\in\mathcal{ O} (\Omega)$, there exists a holomorphic

1807: function $F\in \mathcal{ O} \left( \mathcal{ W}_{p_1} \cup \Omega

1808: \right)$ with $F \vert_\Omega =f$.

1809: \end{proposition}

1810: 

1811: In the CR dimension $m=1$ case, the choice of a special point $p\in C$

1812: is much more delicate and will be done in the next two Sections~4

1813: and~5 below, where the analog of Proposition~3.22 appears as the main

1814: removability proposition~5.12. In the case $m=1$, the submanifold

1815: $M^1$ is of real dimension equal to $n$ and it is not difficult to

1816: generalize Lemma~3.21, obtaining a submanifold $H^1$ which is of real

1817: dimension $(n-1)$ and totally real (but not generic) in

1818: $\C^n$. However, in general, such a point $p_1\in C\cap H^1$ is {\it not

1819: locally $\mathcal{ W}$-removable}\, in general. For instance, in the

1820: hypersurface case $n=2$, locally in a neighborhood of $p_1$, the

1821: closed set $C\subset (H^1)^-\cup \{p_1\}$ may coincide with the

1822: intersection of $M$ with a local complex curve transverse to $M$ at

1823: $p_1$, hence $C$ is not locally removable. In this (trivial) example,

1824: the condition $\mathcal{ F}^c\{C\}$ is not satisfied and this

1825: justifies a more refined geometrical analysis to chase a suitable

1826: special point $p_{\rm sp}\in C$ to be removed locally.

1827: This is the main purpose of Sections~4 and~5 below.

1828: 

1829: \section*{\S4.~Construction of a semi-local half wedge}

1830: 

1831: \subsection*{4.1.~Preliminary}

1832: Let $M$ be a $\mathcal{ C}^{2, \alpha}$ globally minimal generic

1833: submanifold of $\C^n$ of CR dimension $m=1$ and of codimension $d=n-m

1834: \geq 1$ and let $M^1$ be a $\mathcal{ C}^{2,\alpha}$-smooth

1835: one-codimensional submanifold which is generic in $\C^n$. Let

1836: $\gamma:[-1,1] \to M^1$ be a $\mathcal{ C}^{2, \alpha}$-smooth curve,

1837: embedding the segment $[-1,1]$ into $M$. Later, in Section~5 below, we

1838: shall exploit the geometric condition formulated in Theorem~2.1' that

1839: such a characteristic curve should satisfy, but in the present

1840: Section~4, we shall not at all take account of this geometric

1841: condition. Our goal is to construct a semi-local half-wedge attached

1842: to a one-sided neighborhood of $M^1$ along $\gamma$ with the property

1843: that holomorphic functions in the neighborhood $\Omega$ of

1844: $M\backslash C$ in $\C^n$ do extend holomorphically to this

1845: half-wedge. First of all, we need to define what we understand by the

1846: term ``half-wedge''. Although all the geometric considerations of

1847: this section may be generalized to the CR dimension $m\geq 2$ case

1848: with slight modifications, we choose to endeavour the exposition in

1849: the case $m=1$, because our constructions are essentially needed only

1850: for this case (however in Section~10 below, since we aim to present a

1851: new proof of Theorem~1.4, we shall also use the notion of a local

1852: half-wedge, without any characteristic curve $\gamma$).

1853: 

1854: \subsection*{4.2.~Three equivalent definitions of attached 

1855: half-wedges} First of all, we define the notion of a local

1856: half-wedge. We shall denote by $\Delta_n (p,\delta)$ the open polydisc

1857: centered at $p \in \C^n$ of radius $\delta >0$. Let $p_1\in M^1$, and

1858: let ${\sf C}_1$ be an open infinite cone in the normal space $T_{p_1}

1859: \C^n/ T_{ p_1}M$. Classically, a {\sl local wedge of edge $M$ at

1860: $p_1$} is a set of the form: $\mathcal{ W}_{p_1}:= \{p+ {\sf c}_1: \,

1861: p \in M, \, {\sf c}_1 \in { \sf C}_1\}\cap \Delta_n (p_1, \delta_1 )$,

1862: for some $\delta_1 >0$. Sometimes, we shall use the following

1863: terminology: if $v_1$ is a nonzero vector in $T_{p_1} \C^n /T_{

1864: p_1}M$, we shall say that $\mathcal{ W}_{p_1}$ is a {\sl local wedge

1865: at $(p_1, v_1)$}. This definition seems to be misleading in the sense

1866: that different vectors $v_1$ seem to yield local wedges with different

1867: directions, however, there is a concrete geometric meaning in this

1868: definition that should be reminded: the positive half-line $\R^+ \cdot

1869: v_1$ generated by the vector $v_1$ is locally contained in the wedge

1870: $\mathcal{ W}_{p_1}$.

1871: 

1872: For us, a {\sl local half-wedge of edge $M$ at $p_1$} will be a set of

1873: the form

1874: \def\theequation{4.3}\begin{equation}

1875: \mathcal{ HW}_{p_1}^+:= \left\{p+{\sf c}_1: \,

1876: p\in U_1\cap (M^1)^+, \, {\sf c}_1 \in {\sf C}_1\right\}

1877: \cap \Delta_n(p_1,\delta_1). 

1878: \end{equation}

1879: This yields a first definition and we shall delineate two further

1880: definitions. Let $\Delta$ denote the unit disc in $\C$, let $\partial

1881: \Delta$ denote its boundary, the unit circle and let $\overline{

1882: \Delta}= \Delta \cup \partial \Delta$ denote its closure.

1883: Throughout this paper, we shall denote by $\zeta=\rho e^{i\theta}$

1884: the variable of $\overline{\Delta}$ with $0 \leq \rho \leq 1$ and

1885: with $\vert \theta \vert \leq \pi$.

1886: 

1887: In fact, our local half-wedges (to be constructed in this section)

1888: will be defined by means of a $\mathcal{ C}^{2, \alpha-0}$-smooth

1889: $\C^n$-valued (remind $\mathcal{ C}^{2,\alpha-0}\equiv \bigcap_{\beta

1890: < \alpha} \mathcal{ C}^{2,\beta}$) mapping $(t, \chi, \nu, \rho)

1891: \longmapsto \mathcal{ Z}_{t,\chi, \nu} (\rho)$, which comes from

1892: parametrized family of analytic discs of the form $\zeta \mapsto

1893: \mathcal{ Z}_{ t,\chi,\nu} (\zeta)$, where the parameters $t \in

1894: \R^{n-1}$, $\chi \in \R^n$, $\nu \in \R$ satisfy $\vert t \vert<

1895: \varepsilon$, $\vert \chi \vert < \varepsilon$, $\vert \nu \vert <

1896: \varepsilon$ for some small $\varepsilon >0$, and where $\mathcal{

1897: Z}_{ t,\chi,\nu} (\zeta)$ is holomorphic with respect to $\zeta$ in

1898: $\Delta$. This mapping will satisfy the following three properties:

1899: 

1900: \smallskip

1901: \begin{itemize}

1902: \item[{\bf (i)}]

1903: $(\chi,\nu) \mapsto \mathcal{ Z}_{0,\chi,\nu}(1)$ is a 

1904: diffeomorphism onto a neighborhood of $p_1$ in $M$,

1905: the mapping $\chi \mapsto \mathcal{ Z}_{0,\chi,0}(1)$ is a 

1906: diffeomorphism onto a neighborhood of $p_1$ in $M^1$

1907: and $(M^1)^+$ corresponds to $\nu>0$ in the diffeomorphism

1908: $(\chi,\nu) \mapsto \mathcal{ Z}_{0,\chi,\nu}(1)$.

1909: \item[{\bf (ii)}]

1910: $\mathcal{ Z}_{t,0,0}(1)=p_1$ and the half-boundary

1911: $\mathcal{ Z}_{t, \chi, \nu}\left(\left\{ e^{i\theta} : \ 

1912: \vert \theta \vert \leq \frac{\pi}{2} \right\}\right)$ is contained in

1913: $M$ for all $t$, all $\chi$ and

1914: all $\nu$.

1915: \item[{\bf (iii)}]

1916: The vector $v_1:= \frac{ \partial \mathcal{ Z}_{0, 0,0}}{ \partial

1917: \theta} (1)\in T_{p_1 }\C^n$ is nonzero and belongs to $T_{p_1}M^1$.

1918: Furthermore, the rank of the $\R^{ n-1}$-valued $\mathcal{

1919: C}^{1,\alpha-0}$-smooth mapping 

1920: \def\theequation{4.4}\begin{equation}

1921: \R^{n-1}\ni t \longmapsto 

1922: \frac{\partial \mathcal{ Z}_{t,0,0}}{\partial \theta}(1)

1923: \in T_{p_1}M^1

1924: \ {\rm mod} \ \left( T_{p_1}M^1\cap T_{p_1}^cM

1925: \right) \cong \R^{n-1}

1926: \end{equation}

1927: is maximal equal to $(n-1)$ at $t=0$.

1928: \end{itemize}

1929: 

1930: \smallskip

1931: \noindent

1932: By holomorphicity of the map $\zeta\mapsto \mathcal{

1933: Z}_{t,\chi,\nu}(\zeta)$, we have $\frac{\partial \mathcal{

1934: Z}_{t,\chi,\nu}}{ \partial \theta}(1) =J\cdot \frac{\partial \mathcal{

1935: Z}_{t,\chi,\nu}}{ \partial \rho}(1)$, where $J$ denotes the complex

1936: structure of $T\C^n$. Consequently, because $J$ induces an isomorphism

1937: from $T_{p_1}M/T_{p_1}^cM\to T_{ p_1}\C^n/T_{p_1}M$, it follows from

1938: property {\bf (iii)} above that the vectors $\frac{\partial \mathcal{

1939: Z}_{t,0,0}}{\partial \rho}(1)$ cover an open cone containing $Jv_1$ in

1940: the quotient space $T_{p_1}M/ T_{p_1}^cM$, as $v$ varies. Then a {\sl

1941: local half-wedge of edge $(M^1)^+$ at $p_1$} will be a set of the form

1942: \def\theequation{4.5}\begin{equation}

1943: \mathcal{ HW}_{p_1}^+:=\left\{

1944: \mathcal{ Z}_{t,\chi,\nu}(\rho)\in\C^n: \ 

1945: \vert t \vert < \varepsilon, \ 

1946: \vert \chi \vert < \varepsilon, \

1947: 0 < \nu <\varepsilon, \ 

1948: 1-\varepsilon < \rho < 1

1949: \right\}.

1950: \end{equation}

1951: We notice that a complete local wedge of edge $M$ at $p_1$ is also

1952: associated to such a family $\mathcal{ Z}_{t,\chi, \nu}(\zeta)$ and may

1953: be defined as $\mathcal{ W}_{p_1}:= \left\{ \mathcal{

1954: Z}_{t,\chi,\nu}(\rho): \ \vert t \vert < \varepsilon, \ \vert \chi

1955: \vert < \varepsilon, \ \vert \nu \vert < \varepsilon, \ 1-\varepsilon

1956: < \rho < 1 \right\}$.

1957: 

1958: As may be checked, this second definition of a half-wege is {\sl

1959: essentially equivalent} to the first one, in the sense that a half

1960: wedge in the sense of the first definition always contains a

1961: half-wedge in the sense of the second definition, and vice versa,

1962: after appropriate shrinkings of open neighborhoods and cones.

1963: 

1964: Furthermore, we may distinguish two cases: either the vector $v_1= 

1965: \frac{ \partial \mathcal{ Z}_{0, 0,0}}{ \partial

1966: \theta} (1)$ is

1967: not complex tangent to $M$ at $p_1$ or it is complex tangent to $M$ at

1968: $p_1$. In the first case, after possibly shrinking $\varepsilon>0$, it

1969: may be checked that a local half-wedge of edge $(M^1 )^+$ coincides

1970: with the intersection of a (full) local wedge $\mathcal{ W}_{ p_1}$ of

1971: edge $M$ at $p_1$ with a one-sided neighborhood $(N^1)^+$ of a local

1972: hypersurface $N^1$ which intersects $M$ locally transversally along

1973: $M^1$ at $p_1$, as drawn in the left hand side of the following

1974: figure, where $M$ is of codimension two.

1975: 

1976: \bigskip

1977: \begin{center}

1978: \input half-wedge.pstex_t

1979: \end{center}

1980: 

1981: In the second case, the vector $v_1= \frac{ \partial \mathcal{ Z}_{0,

1982: 0,0}}{ \partial \theta} (1)$ is complex tangent to $M$ at $p_1$, hence

1983: belongs to the characteristic direction $T_{p_1}M^1\cap T_{p_1}^cM$,

1984: so the vector $-Jv_1$ which is interiorly tangent to the disc

1985: $\mathcal{ Z}_{0,0,0}(\Delta)$, is tangent to $M$ at $p_1$, is not

1986: tangent to $M^1$ at $p_1$, but points towards $(M^1)^+$ at $p_1$. It

1987: may then be checked that a local half-wedge of edge $(M^1)^+$

1988: coincides with the intersection of a local wedge $\mathcal{

1989: W}_{p_1}^1$ of edge $M^1$ at $(p_1,-Jv_1)$ which contains the side

1990: $(M^1)^+$, as drawn in the right hand side of {\sc Figure~5} above, in

1991: which $M$ is of codimension one. This provides the third and the most

1992: intuitive definition of the notion of local half-wedge.

1993: 

1994: Finally, we may define the desired notion of a semi-local attached

1995: half-wedge. Let $\gamma: [-1,1]\to M^1$ be an embedded $\mathcal{

1996: C}^{2, \alpha}$-smooth segment in $M^1$. Since the normal bundle to

1997: $M^1$ in $M$ is trivial, we can choose a coherent family of one-sided

1998: neighborhoods $(M_\gamma^1)^+$ of $M^1$ in $M$ along $\gamma$. A {\sl

1999: half-wedge attached to a one-sided neighborhood $(M_\gamma^1)^+$ of

2000: $M^1$ along $\gamma$} is a domain $\mathcal{ HW}_\gamma^+$ which

2001: contains a local half-wedge of edge $(M^1)^+$ at $\gamma(s)$ for every

2002: $s\in [-1,1]$. Another essentially equivalent definition is to

2003: require that we have a family $\mathcal{ Z}_{t,\chi, \nu:s}(\rho)$ of

2004: mappings smoothly varying with the parameter $s$ such that at each

2005: point $\gamma(s)=\mathcal{ Z}_{t,\chi, \nu:s}(1)$, the three

2006: conditions {\bf (i)}, {\bf (ii)} and {\bf (iii)} introduced above to

2007: define a local half-wedge are satisfied. Intuitively speaking, the

2008: direction of the cones defining the local half wedge at the point

2009: $\gamma(s)$ are smoothly varying with respect to $s$.

2010: 

2011: \bigskip

2012: \begin{center}

2013: \input attached-half-wedge.pstex_t

2014: \end{center}

2015: 

2016: We can now state the main proposition of this section, which will be

2017: of crucial use for the proof of Theorem~3.19 {\bf (i)}. Forgetting

2018: for a while the complete content of the geometric condition $\mathcal{

2019: F}_{M^1}^c\{C\}$ formulated in the assumptions of Theorem~1.2', which

2020: we will analyze thoroughly in Section~5 below, we shall only assume

2021: that we are given a characteristic segment $\gamma : [-1,1] \to M^1$

2022: in the following proposition, whose proof is the main goal of this

2023: Section~4.

2024: 

2025: \def\theproposition{4.6}\begin{proposition} 

2026: Let $M$ be a $\mathcal{ C}^{2,\alpha}$ globally minimal generic

2027: submanifold of $\C^n$ of CR dimension $m=1$ and of codimension

2028: $d=n-m\geq 1$ and let $M^1$ be a $\mathcal{ C}^{2,\alpha}$-smooth

2029: one-codimensional submanifold which is generic in $\C^n$. Let $\gamma:

2030: [-1,1]\to M^1$ be an arbitrary $\mathcal{ C}^{2,\alpha}$-smooth

2031: curve. Then there exist a neighborhood $V_\gamma$ of $\gamma[-1,1]$ in

2032: $M$ and a semi-local one-sided neighborhood $(M_\gamma^1)^+$ of $M^1$

2033: in $M$ along $\gamma$ which is the intersection of $V_\gamma$ with a

2034: side $(M_\gamma^1)^+$ of $M^1$ along $\gamma$ and there exists a

2035: semi-local half-wedge $\mathcal{ HW }_\gamma^+$ attached to

2036: $(M_\gamma^1 )^+ \cap V_\gamma$ with $\Omega \cap \mathcal{

2037: HW}_\gamma^+$ connected {\rm (}shrinking $\Omega$ if necessary{\rm )}

2038: such that for every holomorphic function $f\in \mathcal{ O} (\Omega)$,

2039: there exists a holomorphic function $F\in \mathcal{ O} \left(

2040: \mathcal{ HW }_\gamma^+ \cup \Omega \right)$ with $F\vert_\Omega=f$.

2041: \end{proposition}

2042: 

2043: To build $\mathcal{ HW}_\gamma^+$, we shall construct families of

2044: analytic discs with boundaries in $(M_\gamma^1)^+$. First of all, we

2045: need to formulate a special, adapted version of the so-called {\sl

2046: approximation theorem} of~\cite{ bt}.

2047: 

2048: \subsection*{4.7.~Local approximation theorem}

2049: As noted in~\cite{m2}, ~\cite{mp1} and~\cite{mp3}, when dealing with

2050: some natural geometric assumptions on the singularity to be

2051: removed\,--\,for instance, a two-codimensional singularity $N\subset

2052: M$ with $T_pN \supset T_p^cM$ for some points $p\in N$ or metrically

2053: thin singularities $E\subset M$ with $H^{2m+d-2}(E)=0$\,--\,it is

2054: impossible to show {\it a priori}\, that continuous CR functions on

2055: $M$ minus the singularity are approximable by polynomials, which

2056: justifies the introduction of deformations and the use of the

2057: continuity principle in~\cite{m2}, \cite{mp1}, \cite{mp3}. On the

2058: contrary the genericity of the submanifold $M^1$ containing the

2059: singularity $C$ enables us to get an approximation Lemma~4.8 just

2060: below, in the spirit of~\cite{bt}. Together with the existence of

2061: Bishop discs attached to $M^1$, the validity of this approximation

2062: lemma on $M\backslash M^1$ is the second main reason for the relative

2063: simplicity of the geometric proofs of Theorem~1.4 provided

2064: in~\cite{j4}, \cite{m2}, \cite{p1}, in comparison the 

2065: proof of Theorem~1.2' to be conducted in this paper.

2066: 

2067: \def\thelemma{4.8}\begin{lemma}

2068: Let $M$ be a $\mathcal{ C}^{2, \alpha}$-smooth globally minimal

2069: generic submanifold of $\C^n$ of CR dimension $m \geq 1$ and of

2070: codimension $d = n-m \geq 1$, let $M^1 \subset M$ be a $\mathcal{

2071: C}^{2, \alpha}$-smooth one-codimensional submanifold which is generic

2072: in $\C^n$ which divides locally $M$ in two open sides $(M^1)^-$ and

2073: $(M^1)^+$ and let $p_1\in M^1$. Then there exist two neighborhoods

2074: $U_1$ and $V_1$ of $p_1$ in $M$ with $V_1 \subset \subset U_1$ such

2075: that for every continuous CR function $f \in \mathcal{ C}_{CR}^0

2076: \left(( M^1 )^+\cap U_1 \right)$, there exists a sequence of

2077: holomorphic polynomials $(P_\nu)_{\nu \in\N}$ wich converges uniformly

2078: towards $f$ on $(M^1)^+\cap V_1$. Of course, the same property holds

2079: in the side $(M^1)^-$ instead of $(M^1)^+$.

2080: \end{lemma}

2081: 

2082: \proof

2083: The proof is a slight modification of~Proposition~5B in~\cite{m2} and

2084: we summarize it, taking for granted that the reader is acquainted with

2085: the approximation theorem proved in~\cite{bt} ({\it see}

2086: also~\cite{j5}). Let $L_0^1$ be a maximally real submanifold passing

2087: through $p_1$ and contained in $M^1\cap U_1$, for a sufficiently small

2088: neighborhood $U_1$ of $p_1$ in $M^1$, possibly to be shrunk later. In

2089: coordinates $z=(z_1,\dots,z_n)=x+iy\in\C^n$ vanishing at $p_1$, we can

2090: assume that the tangent plane to $L_0^1$ at $p_1$ identifies with

2091: $\R^n=\{y=0\}$. As the codimensions of $L_0^1$ in $M^1$ and in $M$ are

2092: equal to $(d-1)$ and to $d$, we can include $L^1$ in a $d$-parameter

2093: family of submanifolds $L_t^1$, where $t\in \R^d$ is small, so that

2094: $L_t^1\cap V_1$ makes a foliation of $M\cap V^1$ by maximally real

2095: $\mathcal{ C}^{2,\alpha}$-smooth submanifolds, for some neighborhood

2096: $V_1\subset\subset U^1$ of $p^1$ in $M$, such that $L_t^1$ is

2097: contained in $M^1$ for $t=(t_1,\dots, t_{d-1},0)$, {\it i.e.} for

2098: $t_d=0$, such that $L_t^1\cap V_1$ is contained in $(M^1)^+$ for

2099: $t_d>0$ and such that $L_t^1 \cap V_1$ is contained in $(M^1)^-$ for

2100: $t< 0$. In addition, we can assume that all the $L_t^1$ coincide with

2101: $L_0^1$ in a neighborhood of $\partial U_1$.

2102: 

2103: We shall first treat the case where $f$ is of class $\mathcal{

2104: C}^1$. Thus, let $f$ be a $\mathcal{ C}^1$-smooth CR function on

2105: $\left( M \backslash M^1\right) \cap U_1$, let $\tau \in \R$ with

2106: $\tau > 0$, fix $t_0\in \R^d$ small with $t_{d;0} >0$, whence

2107: $L_{t_0}^1\cap V^1$ is contained in $(M^1)^+$, let $\widehat{z} \in (

2108: M^1)^+ \cap V_1$ be an arbitrary point and consider the following

2109: integral which consists of the convolution of $f$ with the Gauss

2110: kernel:

2111: \def\theequation{4.9}\begin{equation}

2112: G_\tau f(\widehat{z}):=

2113: \left(\frac{\tau}{\pi}\right)^{n/2}

2114: \int_{U_1\cap L_{t_0}} \, 

2115: e^{-\tau(z-\widehat{z})^2}\, 

2116: f(z) \, dz,

2117: \end{equation}

2118: where $(z- \widehat{z})^2:= (z_1- \widehat{ z}_1)^2 + \cdots+

2119: (z_n-\widehat{ z}_n)^2$ and $dz := dz_1 \wedge \cdots \wedge d z_n$.

2120: The point $\widehat{ z}$ belongs to some maximally real submanifold

2121: $L_{\widehat{ t}}$ with $\widehat{ t}_d>0$. We now claim that the

2122: value of $G_\tau f( \widehat{ z})$ is the same if we replace the

2123: integration on the fixed submanifold $U_1 \cap L_{ t_0}$ in the

2124: integral~\thetag{4.9} by an integration over $U_1 \cap L_{ \widehat{

2125: t}}$. This key argument will follows from Stoke's theorem, from the

2126: fact that $f$ is CR on $M\cap U_1$ and from the important fact that

2127: between $L_{ t_0}$ and $L_{\widehat{ t}}$, we can construct a $(n+

2128: 1)$-dimensional submanifold $\Sigma$ with boundary $\partial \Sigma =

2129: L_{t_0} - L_{ \widehat{ t}}$ {\it which is entirely contained in

2130: $(M^1)^+\cap U_1$}, thanks to the fact that $M^1$ is generic and of

2131: codimension $1$; indeed, we may compute

2132: \def\theequation{4.10}\begin{equation}

2133: \left\{

2134: \aligned

2135: G_\tau f(\widehat{ z})=

2136: & \

2137: \left(\frac{\tau}{\pi}\right)^{n/2}

2138: \int_{U_1\cap L_{\widehat{ t}}}

2139: e^{-\tau(z-\widehat{ z})^2}\, 

2140: f(z) \, dz+\left(\frac{\tau}{\pi}\right)^{n/2}

2141: \int_{\Sigma}\, 

2142: d\left(

2143: e^{-\tau (z-\widehat{ z})^2}\, f(z) \, dz

2144: \right) \\

2145: =

2146: & \

2147: \left(\frac{\tau}{\pi}\right)^{n/2}

2148: \int_{U_1\cap L_{\widehat{ t}}}

2149: e^{-\tau(z-\widehat{ z})^2}\, 

2150: f(z) \, dz,

2151: \endaligned\right.

2152: \end{equation}

2153: noticing that the second integral in the right hand side of the first

2154: line vanishing, because $f$ and $e^{-(z-\widehat{ z})^2}$ being CR and

2155: of class at least $\mathcal{ C}^1$, one has $d\left( e^{-\tau (z-

2156: \widehat{ z})^2}\, f(z) \, dz \right)=0$. This proves the claim.

2157: 

2158: By analyzing the real and the imaginary part of the phase function

2159: $-\tau(z-\widehat{ z})^2$ on $L_{\widehat{ t}}$, one can show by means

2160: of a standard argument (convolution with Gauss' kernel is an

2161: approximation of the Dirac mass) that the integral over $U_1 \cap

2162: L_{\widehat{ t}}$ tends towards $f(\widehat{ z})$ as $\tau$ tends to

2163: $\infty$, provided that the submanifold $U_1\cap L_{\widehat{ t}}$ is

2164: sufficiently close to the real plane $\R^n$ in $\mathcal{ C}^1$

2165: norm. Finally, by developing in power series and truncating the

2166: exponential in the first expression~\thetag{4.9} which defines $G_\tau

2167: f(\widehat{ z})$ and by integrating termwise, one constructs the

2168: desired sequence of polynomials $(P_\nu (z))_{\nu\in\N}$.

2169: 

2170: The case where $f$ is only continuous follows from standard arguments

2171: from the theory of distributions. This completes the proof of

2172: Lemma~4.8.

2173: \endproof

2174: 

2175: \subsection*{4.11.~A family of straightenings} 

2176: Our main goal in the remainder of this section is to construct a

2177: semi-local half-wedge attached to a one-sided neighborhood

2178: $(M_\gamma^1)^+$ of $M^1$ in $M$ along $\gamma$ which consists of

2179: analytic discs attached to $(M_\gamma^1)^+$. First of all, we need a

2180: convenient family of normalizations of the local geometries of $M$ and

2181: of $M^1$ along the points $\gamma(s)$ of our characteristic curve

2182: $\gamma$, for all $s$ with $-1\leq s \leq 1$.

2183: 

2184: Let $\Omega$ be a thin neighborhood of $\gamma( [-1, 1])$ in $\C^n$,

2185: say a union of polydiscs of fixed radius centered at the points

2186: $\gamma (s)$. Then there exists $n$ real valued $\mathcal{ C}^{2,

2187: \alpha}$-smooth functions $r_1 (z, \bar z), \dots, r_n (z,\bar z)$

2188: defined in $\Omega$ such that $M \cap \Omega$ is given by the $(n-1)$

2189: Cartesian equations $r_2 (z,\bar z)= \cdots= r_n (z,\bar z)=0$ and

2190: such that moreover, $M^1\cap \Omega$ is given by the $n$ Cartesian

2191: equations $r_1(z,\bar z)=r_2(z,\bar z)=\cdots=r_n(z,\bar z)=0$. We

2192: first center the coordinates at $\gamma(s)$ by setting $z':=

2193: z-\gamma(s)$. Then the defining functions centered at $z'=0$ become

2194: \def\theequation{4.12}\begin{equation}

2195: r_j\left(z'+

2196: \gamma(s), \bar z'+\overline{\gamma(s)}\right)-r_j

2197: \left(\gamma(s),\overline{\gamma(s)}\right)=:

2198: r_j'(z',\bar z':s),

2199: \end{equation}

2200: for $j=1,\dots,n$, and they are parametrized by $s\in [-1,1]$. Now,

2201: we drop the primes on coordinates and we denote by $r_j(z,\bar z:s)$,

2202: $j=1,\dots,n$, the defining equations for the new 

2203: $M_s$ and $M_s^1$, which

2204: correspond to the old $M$ and $M^1$ locally in a neighborhood of

2205: $\gamma(s)$. Next, we straighten the tangent planes by using the

2206: linear change of coordinates $z'=A_s\cdot z$, where the $n\times n$

2207: matrix $A_s$ is defined by $A_s:= 2i \, \left(\frac{\partial

2208: r_j}{\partial z_k}(0,0:s) \right)_{1\leq j,k\leq n}$. Then the

2209: defining equations for the two transformed 

2210: $M_s'$ and for $M_s^{1'}$ are given by

2211: \def\theequation{4.13}\begin{equation}

2212: r_j'(z',\bar z':s):= r_j\left(A_s^{-1} \cdot z', \overline{A}_s^{-1}

2213: \cdot \overline{ z}': s\right),

2214: \end{equation}

2215: and we check immediately that the matrix $\left( \frac{ \partial

2216: r_j'}{ \partial z_k} (0,0:s) \right)_{1 \leq j,k \leq n}$ is equal to

2217: $2i$ times the $n\times n$ identity matrix, whence $T_0 M_s'= \{y_2'=

2218: \cdots = y_n' =0\}$ and $T_0{ M_s' }^1=\{y_1'=y_2'= \cdots =

2219: y_n'=0\}$. It is important to notice that the matrix $A_s$ only

2220: depends $\mathcal{ C}^{1,\alpha}$-smoothly with respect to $s$.

2221: Consequently, if we now drop the primes on coordinates, the defining

2222: equations for $M_s$ and for $M_s^1$ are of class $\mathcal{ C}^{2,

2223: \alpha}$ with respect to $(z,\bar z)$ and only of class $\mathcal{

2224: C}^{1, \alpha}$ with respect to $s$.

2225: 

2226: Applying then the $\mathcal{ C}^{2,\alpha}$-smooth

2227: implicit, we deduce that there exist $(n-1)$ functions

2228: $\varphi_j(x,y_1:s)$, $j=2,\dots,n$, which are all of class $\mathcal{

2229: C}^{2,\alpha}$ with respect to $(x,y_1)$ in a real cube

2230: $\I_{n+1}(2\rho_1):=\{(x,y_1)\in \R^n \times \R: \ \vert x \vert <

2231: 2\rho_1, \ \vert y_1 \vert < 2\rho_1\}$, for some $\rho_1>0$, which

2232: are uniformly bounded in $\mathcal{ C}^{2,\alpha}$-norm as the

2233: parameter $s$ varies in $[-1,1]$, which are of class $\mathcal{

2234: C}^{1,\alpha}$ with respect to $s$, such that $M_s$ may be represented

2235: in the polydisc $\Delta_n (\rho_1)$ by the $(n-1)$ graphed equations

2236: \def\theequation{4.14}\begin{equation}

2237: y_2=\varphi_2(x,y_1: s), \dots\dots, \, 

2238: y_n=\varphi_n(x,y_1:s),

2239: \end{equation}

2240: or more concisely $y'=\varphi'(x,y_1:s)$, if we denote the coordinates

2241: $(z_2,\dots,z_n)$ simply by $z'=x'+iy'$. Here, by construction, we

2242: have the normalization conditions $\varphi_j(0:s)=\partial_{x_k}

2243: \varphi_{j}(0:s)=\partial_{y_1} \varphi_{j}(0:s)=0$, for $j=2,\dots,n$

2244: and $k=1,\dots,n$. Sometimes in the sequel, we shall use the notation

2245: $\varphi_j(z_1,x':s)$ instead of $\varphi_j(x,y_1:s)$. Similarly,

2246: again by means of the implicit function theorem, we obtain $n$

2247: functions $h_k(x:s)$, for $k=1,\dots,n$ , which 

2248: are of class $\mathcal{ C}^{2,\alpha}$

2249: in the cube $\I_n(2\rho_1)$ (after possibly shrinking $\rho_1$)

2250: enjoying the same regularity property with respect 

2251: to $s$, such that $M_s^1$ is

2252: represented in the polydisc $\Delta_n(\rho_1)$ by the $n$ graphed

2253: equations

2254: \def\theequation{4.15}\begin{equation}

2255: y_1=h_1(x:s), \

2256: y_2=h_2(x:s), \

2257: \dots\dots, \ y_n=h_n(x:s).

2258: \end{equation} 

2259: In addition, we can assume that

2260: \def\theequation{4.16}\begin{equation}

2261: h_j(x:s)\equiv

2262: \varphi_j\left(x,h_1(x:s):s\right), \ \ \ \ \ \ \

2263: j=2,\dots,n.

2264: \end{equation}

2265: Here, by construction, we have the normalization conditions

2266: $h_k(0:s)=\partial_{x_l}

2267: h_k(0:s)=0$ for $k,l=1,\dots,n$.

2268: 

2269: In the sequel, we shall denote by $\widehat{ z}=\Phi_s(z)$ the

2270: final change of coordinates which is centered at $\gamma(s)$ and

2271: which straightens simultaneously the tangent planes to 

2272: $M$ at $\gamma(s)$ and to $M^1$ at $\gamma(s)$ and we shall

2273: denote by $M_s$ and by $M_s^1$ the transformations of

2274: $M$ and of $M^1$.

2275: 

2276: Also, we shall remind that the following regularity

2277: properties hold for the functions $\varphi_j (x,y_1:s)$ and

2278: $h_k(x:s)$:

2279: \begin{itemize}

2280: \item[{\bf (a)}]

2281: For fixed $s$, they are of class $\mathcal{ C}^{2,\alpha}$ with

2282: respect to their principal variables, namely excluding the parameter

2283: $s$.

2284: \item[{\bf (b)}]

2285: They are of class $\mathcal{ C}^{1,\alpha}$ with respect to all their

2286: variables, including the parameter $s$.

2287: \item[{\bf (c)}]

2288: Each of their first order partial derivative with respect to one of

2289: their principal variables is of class $\mathcal{ C}^{1,\alpha}$ with

2290: respect to all their variables, including the parameter $s$.

2291: \end{itemize}

2292: Indeed, these properties are clearly satisfied for the

2293: functions~\thetag{4.13} and they are inherited after the two

2294: applications of the implicit function theorem which yielded the

2295: functions $\varphi_j(x,y_1:s)$ and $h_k(x:s)$.

2296: 

2297: \subsection*{4.17.~Contact of a small ``round'' analytic

2298: disc with $M^1$} Let $r\in \R$ with $0\leq r \leq r_1$, where $r_1$ is

2299: small in comparison with $\rho_1$. Then the ``round'' analytic disc

2300: $\overline{\Delta} \ni \zeta \to \widehat{ Z}_{1;r}(\zeta):=

2301: ir(1-\zeta)\in \C$ with values in the complex plane equipped with the

2302: coordinate $z_1=x_1+iy_1$ is centered at the point $ir$ of the

2303: $y_1$-axis, is of radius $r$ and is contained in the open upper half

2304: plane $\{z_1\in \C: y_1> 0\}$, except its boundary point $\widehat{

2305: Z}_{1;r}(1)=0$. In addition, the tangent direction

2306: $\frac{\partial}{\partial \theta} \widehat{ Z}_{1;r}(1)=r$ is directed

2307: along the positive $x_1$-axis, {\it see} in advance {\sc Figure~7}

2308: below.

2309: 

2310: As in~\cite{tu2}, \cite{tu3}, \cite{mp1}, \cite{mp3}, we denote by

2311: $T_1$ the Hilbert transform (harmonic conjugate operator) on $\partial

2312: \Delta$ vanishing at $1$, namely $(T_1 X) (1)=0$, whence

2313: $T_1(T_1(X))=-X+X(1)$.

2314: 

2315: By lifting this disc contained in the complex tangent space

2316: $T_{p_1}^cM\equiv \C_{z_1}\times \{0\}$, we may define an analytic

2317: disc parametrized by $r$ and $s$ which is attached to $M$ 

2318: of the form

2319: \def\theequation{4.18}\begin{equation}

2320: \widehat{ Z}_{r:s}(\zeta)=\left(

2321: ir(1-\zeta), \widehat{ Z}_{r:s}'(\zeta)

2322: \right)\in \C\times \C^{n-1}

2323: \end{equation}

2324: where the real part $\widehat{ X}_{r:s}'(\zeta)$ of $\widehat{

2325: Z}_{r:s}'(\zeta)$ satisfies the following Bishop type equation on

2326: $\partial \Delta$

2327: \def\theequation{4.19}\begin{equation}

2328: \widehat{ X}_{r:s}'(\zeta)= -\left[T_1

2329: \varphi'\left(\widehat{ Z}_{1;r}(\cdot), 

2330: \widehat{ X}_{r:s}'(\cdot):s\right)\right](\zeta), \ \ \ \ \ \

2331: \zeta\in\partial\Delta.

2332: \end{equation}

2333: By~\cite{tu1}, \cite{tu3}, if $r_1$ is sufficiently small, there

2334: exists a solution which is $\mathcal{ C}^{2,\alpha-0}$-smooth with

2335: respect to $(r,\zeta)$, but only $\mathcal{ C}^{1, \alpha-0}$-smooth

2336: with respect to $(r,\zeta,s)$. Notice that for $r=0$, the disc

2337: $\widehat{ Z}_{1;0}(e^{i\theta })$ is constant equal to $0$ and by

2338: uniqueness of the solution of~\thetag{4.19}, it follows that

2339: $\widehat{ Z}_{0:s}'\left(e^{i\theta}\right) \equiv 0$. It follows

2340: trivially that $\partial_\theta \widehat{ X}_{0:s}\left(e^{i\theta}

2341: \right)\equiv 0$ and that $\partial_\theta\partial_\theta \widehat{

2342: X}_{0:s} \left(e^{i\theta}\right)\equiv 0$, which 

2343: will be used in a while. Notice also that

2344: $\widehat{ X}_{r:s}(1)=0$ for all $r$ and all $s$.

2345: 

2346: On the other hand, since by assumption, we have $h_1(0:s)=0$ and

2347: $\partial_{x_k}h_1(0:s)=0$ for $k=1,\dots,n$, it follows from the

2348: chain rule that if we set

2349: \def\theequation{4.20}\begin{equation}

2350: F(r,\theta:s):= h_1\left(

2351: \widehat{ X}_{r:s}\left(e^{i\theta}\right):s\right)

2352: \end{equation}

2353: where $\theta$ satisfies $0\leq \vert \theta \vert \leq \pi$,

2354: then

2355: the following four equations hold

2356: \def\theequation{4.21}\begin{equation}

2357: F(0,\theta:s)\equiv 0, \ \ \ \ \

2358: F(r,0:s)\equiv 0, \ \ \ \ \

2359: \partial_\theta F(0,\theta:s)\equiv 0, \ \ \ \ \ 

2360: \partial_\theta F(r,0:s)\equiv 0.

2361: \end{equation}

2362: We claim that there exists a constant $C>0$ such that

2363: the following five inequalities hold

2364: for $0\leq \vert \theta \vert \leq \pi$, for

2365: $0\leq r \leq r_1$, for $s\in [-1,1]$ and

2366: for $\vert x \vert \leq \rho_1$:

2367: \def\theequation{4.22}\begin{equation}

2368: \left\{

2369: \aligned

2370: {}

2371: & \

2372: \left\vert

2373: \widehat{ X}_{r:s}\left(e^{i\theta}\right) 

2374: \right\vert\leq C\cdot r, \\

2375: & \

2376: \left\vert 

2377: \partial_\theta \widehat{ X}_{r:s}\left(e^{i\theta}\right)

2378: \right\vert \leq C\cdot r, \\

2379: & \

2380: \left\vert

2381: \partial_\theta \partial_\theta \widehat{ X}_{r:s}\left(e^{i\theta}

2382: \right) 

2383: \right\vert \leq C\cdot r^{\frac{\alpha}{2}}, \\

2384: & \

2385: \sum_{k=1}^n 

2386: \left\vert

2387: \partial_{x_k} h_1(x) 

2388: \right\vert\leq C\cdot \vert x \vert, \\

2389: & \

2390: \sum_{k_1,k_2=1}^n\, 

2391: \left\vert

2392: \partial_{x_{k_1}}\partial_{x_{k_2}}

2393: h_1(x)

2394: \right\vert \leq C.

2395: \endaligned\right.

2396: \end{equation}

2397: The best constants for each inequality are {\it a priori} distinct,

2398: but we simply take for $C$ the largest one. Indeed, the first, the

2399: second and the third inequalities are elementary consequences of the

2400: (uniform with respect to $s$) $\mathcal{ C}^{2,

2401: \frac{\alpha}{2}}$-smoothness of $\widehat{

2402: X}_{r:s}\left(e^{i\theta}\right)$ with respect to $(r,\theta)$, and of

2403: the normalization conditions~\thetag{4.21}. The fourth and the fifth

2404: inequalities are consequences of the $\mathcal{

2405: C}^2$-smoothness of $h_1$ and of its first order normalizations

2406: (complete argument may be easily be provided by imitating the

2407: reasonings of the elementary Section~6 below). 

2408: 

2409: Computing now the second derivative

2410: of $F(r,\theta:s)$ with respect to $\theta$, we obtain

2411: \def\theequation{4.23}\begin{equation}

2412: \small

2413: \left\{

2414: \aligned

2415: \partial_\theta\partial_\theta

2416: F(r,\theta:s)

2417: & 

2418: =

2419: \sum_{k=1}^n\, 

2420: \partial_{x_k} h_1 \left(

2421: \widehat{ X}_{r:s}\left(e^{i\theta}\right):s

2422: \right)\cdot

2423: \partial_\theta \partial_\theta

2424: \widehat{ X}_{k;r:s}\left(e^{i\theta}\right) + \\

2425: &

2426: +

2427: \sum_{k_1,k_2=1}^n\, 

2428: \partial_{x_{k_1}}\partial_{x_{k_2}} 

2429: h_1\left(

2430: \widehat{ X}_{r:s}\left(e^{i\theta}\right)\right)\cdot 

2431: \partial_\theta \widehat{ X}_{k_1;r,s}\left(e^{i\theta}\right)\cdot 

2432: \partial_\theta \widehat{ X}_{k_2;r,s}\left(e^{i\theta}\right),

2433: \endaligned\right.

2434: \end{equation}

2435: and we may apply the majorations~\thetag{4.22} to get

2436: \def\theequation{4.24}\begin{equation}

2437: \left\{

2438: \aligned

2439: \left\vert

2440: \partial_\theta \partial_\theta

2441: F(r,\theta:s)

2442: \right\vert\leq

2443: & \

2444: C\cdot \left\vert

2445: \widehat{ X}_{r:s}(e^{i\theta}) \right\vert

2446: \cdot C \cdot r^{\frac{\alpha}{2}}

2447: +C\cdot (C\cdot r)^2 \\

2448: \leq

2449: & \

2450: r\cdot C^3\left[

2451: r^{\frac{\alpha}{2}}+r^2

2452: \right].

2453: \endaligned\right.

2454: \end{equation}

2455: 

2456: We can now state and prove a lemma which shows that the disc

2457: boundaries $\widehat{ Z}_{r:s}(\partial\Delta)$ touches $M^1$ only at

2458: $p_1$ and lies in $(M^1)^+\cup \{p_1\}$.

2459: 

2460: \def\thelemma{4.25}\begin{lemma}

2461: If $r_1 \leq \min\left( 1, \ \left( \frac{1}{4C^3 \pi^2}

2462: \right)^{\frac{2}{\alpha}}\right)$, then $\widehat{ Z}_{r:s}(\partial

2463: \Delta \backslash \{1\})$ is contained in $(M_s^1)^+$ for all $r$ with

2464: $0 < r \leq r_1$ and all $s$ with $-1\leq s \leq 1$.

2465: \end{lemma}

2466: 

2467: \proof

2468: In the polydisc $\Delta_n(\rho_1)$, the positive half-side $(M_s^1)^+$

2469: in $M$ is represented by the single equation $y_1>h_1(x:s)$, hence we

2470: have to check that $\widehat{ Y}_{1;r}\left(e^{i\theta}\right) >

2471: \left\vert h_1\left(\widehat{ X}_{r:s}\left(e^{i\theta}\right):s\right)

2472: \right\vert$, for all $\theta$ with $0< \vert \theta \vert \leq \pi$.

2473: According to~\thetag{4.18}, the $y_1$-component $\widehat{

2474: Y}_{1;r}\left(e^{i\theta}\right)$ of $\widehat{

2475: Z}_{r:s}\left(e^{i\theta}\right)$ is given by

2476: $r\left(1-\cos \theta\right)$.

2477: 

2478: On the first hand, we observe the elementary minoration $r(1-\cos

2479: \theta) \geq r\cdot \theta^2\cdot \frac {1}{\pi^2}$, valuable for

2480: $0\leq \vert \theta \vert \leq \pi$.

2481: 

2482: On the second hand, taking account of the second and fourth

2483: relations~\thetag{4.21}, Taylor's integral formula now yields

2484: \def\theequation{4.26}\begin{equation} 

2485: F(r,\theta:s)=\int_0^\theta \,

2486: (\theta-\theta') \cdot \partial_\theta 

2487: \partial_\theta F\left(r,\theta':s\right) \cdot

2488: d\theta'.

2489: \end{equation}

2490: Observing that $r^2\leq r^{\frac{ \alpha}{2}}$, since $0<r \leq r_1\leq

2491: 1$, and using the majoration~\thetag{4.24}, we may estimate, taking

2492: account of the assumption on $r_1$ written in the statement of the

2493: lemma:

2494: \def\theequation{4.27}\begin{equation}

2495: \left\vert

2496: F(r,\theta:s) 

2497: \right\vert\leq r\cdot \frac{\theta^2}{2}

2498: \cdot C^3 [2r^{\frac{\alpha}{2}}]

2499: \leq r\cdot \theta^2 

2500: \cdot \frac{1}{4\pi^2}.

2501: \end{equation}

2502: The desired inequality $r(1-\cos\theta) > \left\vert

2503: F(r,\theta:s)

2504: \right\vert$ for all $0< \vert \theta \vert \leq \pi$ is

2505: proved.

2506: \endproof

2507: 

2508: We now fix once for all a radius $r_0$ with $0< r_0 \leq r_1$. In the

2509: remainder of the present Section~4, we shall deform the disc

2510: $\widehat{ Z}_{r_0:s}(\zeta)$ by adding many more parameters. We

2511: notice that for all $\theta$ with $0\leq \vert \theta \vert \leq

2512: \frac{\pi}{4}$, we have the trivial minoration $\partial_\theta

2513: \partial_\theta \widehat{ Y}_{1;r_0}\left(e^{i\theta}\right) = r_0

2514: \cos \theta\geq \frac{ r_0}{\sqrt{2}}$. Also, by~\thetag{ 4.24} and

2515: by the inequality on $r_1$ written in the statement of Lemma~4.25, we

2516: deduce $\left\vert \partial_\theta \partial_\theta h_1\left( \widehat{

2517: X}_{r_0:s}\left(e^{i\theta}\right) \right) \right\vert\leq \frac{

2518: r_0}{2\pi^2}$ for all $\theta$ with $0\leq \vert \theta \vert \leq

2519: \pi$. Since we shall need a generalization of Lemma~4.25 in

2520: Lemma~4.51 below, let us remember these two interesting inequalities,

2521: valid for $0\leq \vert \theta \vert \leq \frac{\pi}{4}$:

2522: \def\theequation{4.28}\begin{equation}

2523: \left\{

2524: \aligned

2525: \partial_\theta\partial_\theta

2526: \widehat{ Y}_{1;r_0}\left(e^{i\theta}\right) \geq 

2527: & \

2528: \frac{r_0}{\sqrt{2}}, \\

2529: \left\vert

2530: \partial_\theta \partial_\theta h_1\left(

2531: \widehat{ X}_{r_0:s}\left(e^{i\theta}\right)

2532: \right)

2533: \right\vert\leq 

2534: & \

2535: \frac{r_0}{2\pi^2},

2536: \endaligned\right.

2537: \end{equation}

2538: noticing of course that $\frac{r_0}{2\pi^2}< \frac{r_0}{\sqrt{2}}$.

2539: 

2540: \subsection*{4.29.~Normal deformations of the disc 

2541: $\widehat{ Z}_{r:s}(\zeta)$} So, we fix $r_0$ small with $0 < r_0 \leq

2542: r_1$ and we consider the disc $\widehat{ Z}_{r_0:s}(\zeta)$ for

2543: $\zeta\in\overline{\Delta}$. Then the point $\widehat{

2544: Z}_{r_0:s}(-1)$ belongs to $(M_s^1)^+$ for each $s$ and stays at a

2545: positive distance from $M_s^1$ as $s$ varies in $[-1,1]$. It follows

2546: that we can choose a subneighborhood $\omega_s$ of $\widehat{

2547: Z}_{r_0:s}(-1)$ in $\C^n$ which is contained in $\Omega$ and whose

2548: diameter is uniformly positive with respect to $s$.

2549: 

2550: \bigskip

2551: \begin{center}

2552: \input normal-deformations.pstex_t

2553: \end{center}

2554: 

2555: Following~\cite{tu2} and~\cite{mp1}, we shall introduce {\sl normal

2556: deformations} of the analytic discs $\widehat{ Z}_{r_0 :s}(\zeta)$

2557: parametrized by $s$ as follows. Let $\kappa: \R^{ n-1} \to \R^{n-1}$

2558: be a $\mathcal{ C}^{2, \alpha}$-smooth mapping fixing the origin and

2559: satisfying $\partial_{ x_k} \kappa_{j}(0)=\delta_j^k$ for $j,k=1,

2560: \dots, n-1$, where $\delta_j^k$ denotes Kronecker's symbol. For

2561: $j=2,\dots,n$, let $\eta_j = \eta_j( z_1, x':s)$ be a real-valued

2562: $\mathcal{ C}^{2, \alpha}$-smooth function compactly supported in a

2563: neighborhood of the point of $\R^{n+1}$ with coordinates $\left(

2564: \widehat{ Z}_{ 1;r_0:s }(-1), \widehat{ X}_{ r_0:s}'(-1) \right)$ and

2565: equal to $1$ at this point. We then define the $\mathcal{ C}^{2,

2566: \alpha}$-smooth deformed generic submanifold $M_{ s,t}$ of equations

2567: \def\theequation{4.30}\begin{equation}

2568: \left\{

2569: \aligned

2570: y'

2571: & \

2572: =\varphi'(z_1,x': s) +\kappa(t) \cdot 

2573: \eta'(z_1,x': s) \\

2574: & \

2575: =:\Phi'(z_1,x',t:s).

2576: \endaligned\right.

2577: \end{equation}

2578: Notice that $M_{ s,0} \equiv M_s$ and that $M_{ s,t}$ coincides with

2579: $M_s$ in a small neighborhood of the origin, for all $t$. If $\mu=

2580: \mu \left( e^{i\theta}: s\right)$ is a real-valued nonnegative

2581: $\mathcal{ C}^{2, \alpha}$-smooth function defined for $e^{i \theta}

2582: \in \partial \Delta$ and for $s \in [-1,1]$ whose support is

2583: concentrated near the segment $\{-1\} \times [-1,1]$, then applying

2584: the existence Theorem~1.2 of~\cite{tu3}, for each fixed $s\in [-1,1]$,

2585: we deduce the existence of a $\mathcal{ C}^{2,\alpha-0}$-smooth

2586: solution of the Bishop type equation

2587: \def\theequation{4.31}\begin{equation}

2588: \widehat{ X}_{r_0,r:s}'\left(e^{i\theta}\right)=

2589: -\left[T_1

2590: \Phi'\left(

2591: \widehat{ Z}_{1;r_0:s}(\cdot), 

2592: \widehat{ X}_{r_0,t:s}'(\cdot), 

2593: t\mu(\cdot : s):s\right)

2594: \right]\left(e^{i\theta}\right),

2595: \end{equation}

2596: which enable us to construct a deformed family of analytic disc

2597: \def\theequation{4.32}\begin{equation}

2598: \widehat{ Z}_{r_0,t:s}\left(e^{i\theta}\right):= 

2599: \left(

2600: \widehat{ Z}_{1;r_0:s}(e^{i\theta}), 

2601: \widehat{ X}_{r_0,t:s}'(e^{i\theta})+i

2602: T_1\left[\widehat{ X}_{r_0,t:s}'(\cdot)\right]

2603: \left(e^{i\theta}\right)

2604: \right)

2605: \end{equation}

2606: whose boundaries are contained in $M\cup \omega_s$, by construction.

2607: By an inspection of Theorem~1.2 in~\cite{tu3}, taking account of the

2608: regularity properties {\bf (a)}, {\bf (b)} and {\bf (c)} stated

2609: after~\thetag{4.16}, one

2610: can show that

2611: the general solution $\widehat{ Z}_{r_0,t:s}(\zeta)$ enjoys regularity

2612: properties which are completely similar:

2613: \begin{itemize}

2614: \item[{\bf (a)}]

2615: For fixed $s$, it is of class $\mathcal{ C}^{2,\alpha-0}$ with respect

2616: to $(t,\zeta)$.

2617: \item[{\bf (b)}]

2618: It is of class $\mathcal{ C}^{1,\alpha-0}$ with respect to 

2619: all the variables $(t,\zeta,s)$.

2620: \item[{\bf (c)}]

2621: Each of its first order partial derivative

2622: with respect to the principal variables

2623: $(t,\zeta)$ is of class $\mathcal{ C}^{1,\alpha-0}$

2624: with respect to all the variables $(t,\zeta,s)$.

2625: \end{itemize}

2626: 

2627: Since the solution is $\mathcal{ C}^{1,\alpha-0}$-smooth

2628: with respect to $s$, it crucially follows that 

2629: the vector

2630: \def\theequation{4.33}\begin{equation}

2631: v_{1:s}:= - \frac{\partial \widehat{ Z}_{r_0,t:s}}{\partial \rho}(1),

2632: \end{equation}

2633: which points inside the analytic disc, varies continuously with

2634: respect to $s$. The following key proposition may be established (up

2635: to a change of notations) just by reproducing the proof of Lemma~2.7

2636: in~\cite{mp1}, taking account of the uniformity of all

2637: differentiations with respect to the curve parameter $s$.

2638: 

2639: \def\thelemma{4.34}\begin{lemma}

2640: There exists a real-valued nonnegative $\mathcal{

2641: C}^{2,\alpha}$-smooth function $\mu=\mu(e^{i\theta}:s)$ defined for

2642: $e^{i\theta}\in\partial \Delta$ and $s\in [-1,1]$ of support

2643: concentrated near $\{-1\}\times [-1,1]$ such that the

2644: mapping

2645: \def\theequation{4.35}\begin{equation}

2646: \R^{n-1}\ni t \longmapsto 

2647: \left.\frac{\partial \widehat{ X}_{r_0,t:s}'}{\partial 

2648: \theta}\left(e^{i\theta}\right)\right\vert_{

2649: \theta=0}\in \R^{n-1}

2650: \end{equation}

2651: is maximal equal to $(n-1)$ at $t=0$.

2652: \end{lemma}

2653: 

2654: Geometrically speaking, since the vector $\left. \frac{ \partial

2655: \widehat{ X}_{1;r_0:s }}{ \partial \theta} \left(e^{i\theta} \right)

2656: \right\vert_{ \theta=0}$ is nonzero, it follows that

2657: when the parameter $t$ varies, then the set of lines

2658: generated by the vectors $\left.\frac{ \partial \widehat{ X}_{ r_0,

2659: t:s}}{\partial \theta}\left(e^{i \theta}\right) \right \vert_{

2660: \theta=0}$ covers an open cone in the space $T_{ p_1}M^1\equiv \R^n$

2661: equipped with coordinates $(x_1,x' )$, {\it see} again {\sc Figure~7}

2662: above for an illustration.

2663: 

2664: \subsection*{4.36.~Adding pivoting and translation parameters}

2665: Let $\chi=(\chi_1,\chi')\in \R\times \R^{n-1}$ 

2666: and $\nu \in \R$ satisfying 

2667: $\vert \chi \vert < \varepsilon$ and $\vert \nu \vert < \varepsilon$

2668: for some small $\varepsilon>0$. Then the mapping

2669: \def\theequation{4.37}\begin{equation}

2670: \left\{

2671: \aligned

2672: \R^{n+1}\ni

2673: (\chi_1,\chi',\nu)\longmapsto 

2674: & \

2675: \left( \,

2676: \chi_1+i[h_1(\chi:s)+\nu], \chi'+i\varphi'

2677: (\chi, h_1(\chi:s)+\nu:s) \,

2678: \right) \\

2679: & \ 

2680: =: \widehat{p}(\chi,\nu:s)\in M_s

2681: \endaligned\right.

2682: \end{equation}

2683: is a $\mathcal{ C}^{2,\alpha}$-smooth diffeomorphism

2684: onto a neighborhood of the origin in $M_s$ with the

2685: property that 

2686: \begin{itemize}

2687: \item[{\bf (a)}]

2688: $\nu >0$ if and only if $\widehat{ p}(\chi,\nu:s)\in (M_s^1)^+$.

2689: \item[{\bf (b)}]

2690: $\nu=0$ if and only if $\widehat{ p}(\chi,\nu:s)\in M_s^1$.

2691: \item[{\bf (c)}]

2692: $\nu<0$ if and only if $\widehat{ p}(\chi,\nu:s)\in (M_s^1)^-$.

2693: \end{itemize}

2694: If $\tau\in \R$ with $\vert \tau \vert < \varepsilon$ is a

2695: supplementary parameter, we may now define a crucial 

2696: deformation of the first

2697: component $\widehat{ Z}_{1; r_0:s}\left(e^{i \theta} \right)$ by setting

2698: \def\theequation{4.38}\begin{equation}

2699: \widehat{ Z}_{1; r_0,\tau,\chi,\nu:s}\left(e^{i\theta}\right):=

2700: ir_0\left(1-e^{i\theta}\right)[1+i\tau]+\chi_1+

2701: i[h_1(\chi:s)+\nu].

2702: \end{equation}

2703: Of course, we have $\widehat{ Z}_{ 1; r_0, 0,0,0:s} \left( e^{i

2704: \theta} \right) \equiv \widehat{ Z}_{ 1;r_0 :s} \left( e^{i \theta }

2705: \right)$. Geometrically speaking, this perturbation corresponds to

2706: add firstly a small ``rotation parameter'' $\tau$ which rotates (and

2707: slightly dilates) the disc $ir_0\left(1-e^{i\theta}\right)$ passing

2708: through the origin in $\C_{ z_1}$, to add secondly a small

2709: ``translation parameter $(\chi_1, \chi')$ which will enable to cover a

2710: neighborhood of the origin in $M_s^1$ and to add thirdly a small

2711: translation parameter $\nu$ along the $y_1$-axis. Consequently, with

2712: this first component $\widehat{ Z}_{1; r_0, \tau, \chi, \nu :s} \left(

2713: e^{i \theta}\right)$, we can construct a $\C^n$-valued analytic disc

2714: $\widehat{ Z}_{r_0, t, \tau, \chi, \nu: s} (\zeta)$ satisfying the

2715: important property

2716: \def\theequation{4.39}\begin{equation}

2717: \widehat{ Z}_{r_0,t,\tau,\chi,\nu:s}(1)=\widehat{ p}(\chi,\nu:s),

2718: \end{equation}

2719: simply by solving the perturbed Bishop type equation

2720: which extends~\thetag{ 4.31}

2721: \def\theequation{4.40}\begin{equation}

2722: \widehat{ X}_{r_0,t,\tau,\chi,\nu:s}'

2723: \left(e^{i\theta}\right)=-

2724: \left[

2725: T_1\left(

2726: \Phi'\left(\widehat{ Z}_{1;r_0,\tau,\chi,\nu:s}(\cdot), 

2727: \widehat{ X}_{r_0,t,

2728: \tau,\chi,\nu:s}'(\cdot), 

2729: t\mu(\cdot:s): s\right)

2730: \right)

2731: \right]\left(e^{i\theta}\right).

2732: \end{equation}

2733: Of course, thanks to the sympathetic stability of Bishop's equation

2734: under perturbation, the solution exists and satisfies smoothness

2735: properties entirely similar to the ones stated after~\thetag{4.32}.

2736: We can summarize the description of our final family of analytic discs

2737: \def\theequation{4.41}\begin{equation}

2738: \widehat{ Z}_{r_0,t,\tau,\chi,\nu:s}(\zeta):

2739: \left\{

2740: \aligned

2741: r_0 = 

2742: & \ 

2743: \text{\rm approximate radius}. \\

2744: t =

2745: & \

2746: \text{\rm normal deformation parameter}. \\

2747: \tau =

2748: & \

2749: \text{\rm pivoting parameter}. \\

2750: \chi =

2751: & \

2752: \text{\rm parameter of translation along} \ M^1. \\

2753: \nu =

2754: & \

2755: \text{\rm parameter of translation in} \ M \ 

2756: \text{\rm transversally to} \ M^1. \\

2757: s =

2758: & \

2759: \text{\rm parameter of the characteristic curve} \ \gamma. \\

2760: \zeta = 

2761: & \ 

2762: \text{\rm unit disc variable}.

2763: \endaligned\right.

2764: \end{equation}

2765: 

2766: For every $t$ and every $\chi$, we now want to adjust the pivoting

2767: parameter $\tau$ in order that the disc boundary $\widehat{ Z}_{r_0,

2768: t, \tau, \chi, 0:s} \left( e^{i\theta}\right)$ for $\nu=0$ is tangent

2769: to $M_s^1$. This tangency condition will be useful in order to derive

2770: the crucial Lemma~4.51 below.

2771: 

2772: \def\thelemma{4.42}\begin{lemma}

2773: Shrinking $\varepsilon$ if necessary, there exists a unique $\mathcal{

2774: C}^{1,\alpha-0}$-smooth map $(t,\chi,s) \mapsto \tau(t, \chi:s)$

2775: defined for $\vert t \vert <\varepsilon$, for $\vert \chi \vert <

2776: \varepsilon$ and for $s \in [-1,1]$ satisfying $\tau(0,

2777: 0:s)=\partial_{ t_j} \tau( 0,0:s)= \partial_{ \chi_k}\tau (0,0:s)=0$

2778: for $j=1, \dots, n-1$ and $k=1, \dots,n$, such that the vector

2779: \def\theequation{4.43}\begin{equation}

2780: \left.

2781: \frac{\partial}{\partial \theta}\right\vert_{\theta=0}

2782: \widehat{ Z}_{r_0,t,\tau(t,\chi:s),\chi,0:s}\left(e^{i\theta}\right)

2783: \end{equation}

2784: is tangent to $M_s^1$ at the point $\widehat{ Z}_{ r_0, t, \tau(t,

2785: \chi:s), \chi, 0:s}(1)= \widehat{ p} ( \chi, 0:s) \in M_s^1$.

2786: \end{lemma}

2787: 

2788: \proof

2789: We remind that $M_s$ is represented by the $(n-1)$ scalar equations

2790: $y'= \varphi'( x, y_1: s)$ and that $M_s^1$ is represented by the $n$

2791: equations $y_1=h_1(x:s)$ and $y'=\varphi '(x, h_1 (x: s):s)\equiv

2792: h'(x':s)$. We can therefore compute the Cartesian equations of the

2793: tangent plane to $M_s^1$ at the point $\widehat{p}(\chi,0:s)=

2794: \chi+ih(\chi:s))$:

2795: \def\theequation{4.44}\begin{equation}

2796: \left\{

2797: \aligned

2798: {\sf Y}_1 -h_1(\chi:s)=

2799: & \

2800: \sum_{k=1}^n\, 

2801: \partial_{x_k}h_{1}(\chi:s) \, \left[

2802: {\sf X}_x-\chi_k\right], \\

2803: {\sf Y}'-\varphi'(\chi,h_1(\chi:s):s) =

2804: & \

2805: \sum_{k=1}^n\, 

2806: \left(

2807: \partial_{x_k}\varphi'+

2808: \partial_{y_1}\varphi'\cdot

2809: \partial_{x_k}h_{1}

2810: \right)

2811: \left[

2812: {\sf X}_k-\chi_k

2813: \right].

2814: \endaligned\right.

2815: \end{equation}

2816: On the other hand, we observe that the tangent vector

2817: \def\theequation{4.45}\begin{equation}

2818: \left. \frac{\partial }{\partial \theta}\right\vert_{\theta=0}

2819: \widehat{ Z}_{r_0,t,\tau,\chi,0:s}\left(e^{i\theta}\right)= \left(

2820: r_0[1+i\tau], \ \left. \frac{\partial }{\partial

2821: \theta}\right\vert_{\theta=0} \widehat{ Z}_{r_0,t,\tau,\chi,0:s}'

2822: \left(e^{i\theta}\right) \right)

2823: \end{equation}

2824: is already tangent to $M_s$ at the point $\widehat{ p}(\chi,0:s)$, 

2825: because $M_{s,t}\equiv M_s$ in a neighborhood of the origin.

2826: More precisely, since $\Phi'\equiv \varphi'$

2827: in a neighborhood of the origin, 

2828: we may differentiate with respect to 

2829: $\theta$ at $\theta=0$ the relation

2830: \def\theequation{4.46}\begin{equation}

2831: \widehat{ Y}_{r_0,t,\tau,\chi,0:s}'\left(e^{i\theta}\right)\equiv 

2832: \varphi'\left(\widehat{ X}_{r_0,\tau,\chi,0:s}\left(e^{i\theta}\right), 

2833: \widehat{ Y}_{1;r_0,\tau,\chi,0:s}\left(e^{i\theta}\right),

2834: \right)

2835: \end{equation}

2836: which is valid for $e^{i\theta}$ close to

2837: $1$ in $\partial \Delta$, noticing in advance that it

2838: follows immediately from~\thetag{4.38} that

2839: \def\theequation{4.47}\begin{equation}

2840: \left.

2841: \frac{\partial }{\partial 

2842: \theta} \right\vert_{\theta=0}

2843: \widehat{ X}_{1;r_0,\tau,\chi,0:s}

2844: \left(e^{i\theta}\right)=r_0 \ \ \ \ \ 

2845: {\rm and} \ \ \ \ \ 

2846: \left.

2847: \frac{\partial }{\partial 

2848: \theta} \right\vert_{\theta=0}

2849: \widehat{ Y}_{1;r_0,\tau,\chi,0:s}

2850: \left(e^{i\theta}\right)=r_0\tau,

2851: \end{equation}

2852: hence

2853: we obtain by a direct application of the chain rule

2854: \def\theequation{4.48}\begin{equation}

2855: \left\{

2856: \aligned

2857: \left.

2858: \frac{\partial }{\partial \theta} 

2859: \right\vert_{\theta=0}\widehat{ Y}_{r_0,t,

2860: \tau,\chi,0:s}'\left(e^{i\theta}\right)=

2861: & \

2862: \partial_{y_1}\varphi'\cdot r_0\tau+ \\

2863: & \

2864: +

2865: \sum_{k=1}^n\,

2866: \partial_{x_k}\varphi'\cdot \left(

2867: \left.

2868: \frac{\partial }{\partial \theta} 

2869: \right\vert_{\theta=0}

2870: \widehat{ X}_{k;r_0,t,\tau,\chi,0:s}

2871: \left(e^{i\theta}\right)

2872: \right).

2873: \endaligned\right.

2874: \end{equation}

2875: On the other hand, the vector~\thetag{4.45} belongs to the tangent

2876: plane to $M_s^1$ at $\widehat{ p}(\chi,0:s)$ whose equations are

2877: computed in~\thetag{4.44} if and only if the following two conditions

2878: are satisfied

2879: \def\theequation{4.49}\begin{equation}

2880: \left\{

2881: \aligned

2882: {}

2883: &

2884: r_0\tau=

2885: \sum_{k=1}^n\, 

2886: \partial_{x_k}h_1(\chi:s)\left[

2887: \left.

2888: \frac{\partial }{\partial \theta} 

2889: \right\vert_{\theta=0}

2890: \widehat{ X}_{k;r_0,t,\tau,

2891: \chi,0:s}\left(e^{i\theta}\right)

2892: \right], \\

2893: &

2894: \left.

2895: \frac{\partial }{\partial \theta} 

2896: \right\vert_{\theta=0}

2897: \widehat{ Y}_{r_0,t,\tau,\chi,0:s}'

2898: \left(e^{i\theta}\right)= \\

2899: &

2900: \ \ \ \ \ \ \ \ \ \ \ \ 

2901: =

2902: \sum_{k=1}^n\, 

2903: \left(

2904: \partial_{x_k}\varphi'+\partial_{y_1}

2905: \varphi'\cdot \partial_{x_k}

2906: h_1

2907: \right)\cdot \left[

2908: \left.

2909: \frac{\partial }{\partial \theta} 

2910: \right\vert_{\theta=0}

2911: \widehat{ X}_{k;r_0,t,\tau,\chi,0:s}

2912: \left(e^{i\theta}\right) \right].

2913: \endaligned\right.

2914: \end{equation}

2915: We observe that the first line of~\thetag{4.49} together with the

2916: relation~\thetag{4.48} already obtained implies the second line

2917: of~\thetag{4.49} by an obvious linear combination. Consequently, the

2918: vector~\thetag{4.45} belongs to the tangent plane to $M_s^1$ at

2919: $\widehat{ p}( \chi, 0:s)$ if and only if the first line

2920: of~\thetag{4.49} is satisfied. As $r_0$ is nonzero, as the first

2921: order derivative $\partial_{x_k}h_1 (\chi :s)$ are of class $\mathcal{

2922: C}^{1, \alpha}$ and vanish at $x=0$ and as $\left. \frac{\partial

2923: }{\partial \theta} \right \vert_{ \theta=0} \widehat{ X}_{k; r_0, t,

2924: \tau, \chi,0:s} \left(e^{ i \theta } \right)$ is of class $\mathcal{

2925: C}^{1, \alpha-0}$ with respect to all variables $(t, \tau, \chi,s)$,

2926: it follows from the implicit function theorem that there exists a

2927: unique solution $\tau= \tau(t, \chi: s)$ of the first line

2928: of~\thetag{4.49} which satisfies in addition the normalization

2929: conditions $\tau( 0,0: s)= \partial_{ t_j}\tau ( 0, 0:s)= \partial_{

2930: \chi_k}\tau (0, 0:s)=0$ for $j=1, \dots, n-1$ and $k=1, \dots, n$.

2931: This completes the proof of Lemma~4.42.

2932: \endproof

2933: 

2934: We now define the analytic disc 

2935: \def\theequation{4.50}\begin{equation}

2936: \widehat{\mathcal{ Z}}_{t,\chi,\nu:s}(\zeta):=

2937: \widehat{ Z}_{r_0,t,\tau(t,\chi:s),\chi,\nu:s}(\zeta).

2938: \end{equation}

2939: 

2940: \def\thelemma{4.51}\begin{lemma}

2941: Shrinking $\varepsilon$ if necessary, 

2942: the following two properties are satisfied{\rm :}

2943: \begin{itemize}

2944: \item[{\bf (1)}]

2945: $\widehat{\mathcal{ Z}}_{

2946: t,\chi,0:s}(\partial \Delta \backslash \{1\}) \subset (M_s^1)^+$

2947: for all $t$, $\chi$, $\nu$ and $s$ with $\vert t \vert< \varepsilon$,

2948: with

2949: $\vert \chi \vert < \varepsilon$, with

2950: $\vert \nu \vert < \varepsilon$ and with

2951: $-1\leq s \leq 1$.

2952: \item[{\bf (2)}]

2953: If $\nu$

2954: satisfies $0< \nu < \varepsilon$, then 

2955: $\widehat{\mathcal{ Z}}_{t,\chi,\nu:s}(\partial \Delta)\subset

2956: (M_s^1)^+$ for all $t$, $\chi$ and $s$ with 

2957: $\vert t \vert <\varepsilon$, with $\vert \chi \vert

2958: < \varepsilon$ and with $-1\leq s \leq 1$.

2959: \end{itemize}

2960: \end{lemma}

2961: 

2962: \proof

2963: To establish property {\bf (1)}, we first observe that the disc

2964: $\widehat{\mathcal{ Z}}_{0, 0,0 :s} \left(e^{i \theta}\right)$

2965: identifies with the disc $\widehat{ Z}_{ r_0 :s}(e^{i \theta})$

2966: defined in \S4.29. According to Lemma~4.25, we know that

2967: $\widehat{\mathcal{ Z}}_{0,0,0:s}(\partial\Delta \backslash\{1\})$ is

2968: contained in $(M_s^1)^+$. By continuity, if $\varepsilon$ is

2969: sufficiently small, we can assume that for all $t$ with $\vert t\vert

2970: < \varepsilon$, for all $\chi$ with $\vert \chi \vert < \varepsilon$

2971: and for all $\theta$ with $\frac{\pi}{4}\leq \vert \theta \vert \leq

2972: \pi$, the point $\widehat{\mathcal{ Z}}_{ t,\chi,0:s}\left(

2973: e^{i\theta}\right)$ is contained in $(M_s^1)^+$. It remains to

2974: control the part of $\partial \Delta$ which corresponds to $\vert

2975: \theta \vert \leq \frac{\pi}{4}$.

2976: 

2977: Since the disc $\widehat{\mathcal{ Z}}_{

2978: t, \chi, \nu:s}\left(e^{i \theta}\right)$ is of

2979: class $\mathcal{ C}^2$ with respect to all its principal variables

2980: $\left(t, \chi, \nu,e^{i\theta}\right)$, 

2981: if $\vert t \vert < \varepsilon$, if

2982: $\vert \chi \vert < \varepsilon$ and if $0 \leq \vert \theta \vert

2983: \leq \frac{\pi }{4}$, for sufficiently small $\varepsilon$, then the

2984: inequalities~\thetag{4.28} are just perturbed a little bit, so we can

2985: assume that

2986: \def\theequation{4.52}\begin{equation}

2987: \left\{

2988: \aligned

2989: \partial_\theta\partial_\theta

2990: \widehat{\mathcal{ Y}}_{1; t,\chi,0:s}\left(e^{i\theta}\right) \geq

2991: & \

2992: r_0, \\

2993: \left\vert

2994: \partial_\theta \partial_\theta

2995: h_1\left(

2996: \widehat{\mathcal{ X}}_{t,\chi,0:s}\left(e^{i\theta}\right)

2997: \right)\right\vert \leq 

2998: & \

2999: \frac{r_0}{2}.

3000: \endaligned\right.

3001: \end{equation}

3002: We claim that the inequality 

3003: \def\theequation{4.53}\begin{equation}

3004: \widehat{\mathcal{ Y}}_{1;t,\chi,0:s}\left(e^{i\theta}\right) >

3005: \left\vert

3006: h_1\left(\widehat{\mathcal{ X}}_{t,\chi,0:s}\left(e^{i\theta}\right)

3007: \right)

3008: \right\vert

3009: \end{equation}

3010: holds for all $0<\vert \theta \vert \leq \frac{\pi}{4}$, which 

3011: will complete the proof of property {\bf (1)}. 

3012: 

3013: Indeed, we first remind that the tangency to $M_s^1$ of the vector

3014: $\left. \frac{ \partial }{\partial \theta} \right \vert_{ \theta=0}

3015: \widehat{\mathcal{ Z}}_{ t,\chi,0:s}\left(e^{ i \theta}\right)$ at the

3016: point $\widehat{ p} ( \chi,0:s)$ is equivalent to the first

3017: relation~\thetag{4.49}, which may be rewritten in terms of the

3018: components of the disc $\widehat{\mathcal{ Z}}_{ t,\chi,0:s}\left(

3019: e^{i\theta}\right)$ as follows

3020: \def\theequation{4.54}\begin{equation}

3021: \partial_\theta \widehat{\mathcal{ Y}}_{1;t,\chi,0:s}(1)=

3022: \sum_{k=1}^n\, [\partial_{x_k} \, 

3023: h_1]\left(\widehat{\mathcal{ X}}_{t,\chi,0:s}(1)\right)\cdot

3024: \partial_\theta \widehat{\mathcal{ X}}_{k;t,\chi,0:s}(1).

3025: \end{equation}

3026: Substracting this relation from~\thetag{4.53} and

3027: substracting also the relation $\widehat{ Y}_{1;t,\chi,0:s}(1)=

3028: h_1\left(\mathcal{ X}_{t,\chi,0:s}(1)

3029: \right)$, we see that it suffices to establish that

3030: for all $\theta$ with $0< \vert \theta \vert \leq \frac{\pi}{4}$, 

3031: we have the strict inequality

3032: \def\theequation{4.55}\begin{equation}

3033: \left\{

3034: \aligned

3035: {}

3036: &

3037: \widehat{\mathcal{ Y}}_{1;t,\chi,0:s}\left(e^{i\theta}\right)-

3038: \widehat{\mathcal{ Y}}_{1;t,\chi,0:s}(1)-

3039: \theta \cdot \partial_\theta 

3040: \widehat{\mathcal{ Y}}_{1;t,\chi,0:s}(1)

3041: >\\

3042: & \

3043: >

3044: \left\vert

3045: h_1\left(

3046: \widehat{\mathcal{ X}}_{t,\chi,0:s}\left(e^{i\theta}\right)

3047: \right)

3048: -h_1\left(\widehat{\mathcal{ X}}_{t,\chi,0:s}(1)

3049: \right) - \right. \\

3050: & 

3051: \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

3052: \left. 

3053: -\sum_{k=1}^n\, [\partial_{x_k} \, 

3054: h_1]\left(\widehat{\mathcal{ X}}_{t,\chi,0:s}(1)\right)\cdot

3055: \partial_\theta \widehat{\mathcal{ X}}_{k;t,\chi,0:s}(1)

3056: \right\vert

3057: \endaligned\right.

3058: \end{equation}

3059: However, by means of Taylor's integral

3060: formula, this last inequality may be rewritten 

3061: as

3062: \def\theequation{4.56}\begin{equation}

3063: \small

3064: \int_0^\theta \, 

3065: (\theta -\theta')\cdot

3066: \partial_\theta\partial_\theta

3067: \widehat{\mathcal{ Y}}_{1;t,\chi,0:s}

3068: \left(e^{i\theta'}\right) \cdot d\theta' >

3069: \left\vert

3070: \int_0^\theta \, 

3071: (\theta-\theta') \cdot

3072: \partial_\theta \partial_\theta

3073: \left[

3074: h_1\left(

3075: \widehat{\mathcal{ X}}_{t,\chi,0:s}\left(e^{i\theta'}\right)

3076: \right)

3077: \right]\cdot d\theta'

3078: \right\vert

3079: \end{equation}

3080: and it follows immediately by means of~\thetag{4.52}.

3081: 

3082: Secondly, to check property {\bf (2)}, we observe that by

3083: the definition~\thetag{4.37}, the parameter $\nu$ corresponds to a

3084: translation of the $z_1$-component of the disc boundary $\widehat{

3085: \mathcal{ Z}}_{ t, \chi, 0:s} (\partial \Delta)$ along the $y_1$ axis.

3086: More precisely, we have

3087: \def\theequation{4.57}\begin{equation}

3088: \frac{\partial }{\partial \nu}\, 

3089: \widehat{ \mathcal{ Y}}_{1; t,\chi,\nu:s}(\zeta) \equiv 1, 

3090: \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ 

3091: \frac{\partial }{\partial \nu}

3092: \widehat{ \mathcal{ X}}_{1; t,\chi,\nu:s}(\zeta)\equiv 0.

3093: \end{equation}

3094: On the other hand, differentiating Bishop's equation~\thetag{4.40},

3095: and using the smallness of the function $\Phi'$, it may be checked

3096: that

3097: \def\theequation{4.58}\begin{equation}

3098: \left\vert 

3099: \frac{\partial }{\partial \nu}

3100: \widehat{ Z}_{r_0,t,\tau,\chi,\nu:s}'(e^{i\theta})

3101: \right\vert < < 1, 

3102: \end{equation}

3103: if $r_0$ and $\varepsilon$ are sufficiently small. It follows that

3104: the disc boundary $\widehat{ \mathcal{ Z}}_{t, \chi, \nu:s}( \partial

3105: \Delta)$ is globally moved in the direction of the $y_1$-axis as

3106: $\nu>0$ increases, hence is contained in $(M_s^1 )^+$.

3107: 

3108: The proof of Lemma~4.51 is complete.

3109: \endproof

3110: 

3111: \subsection*{4.59.~Local half-wedges}

3112: As a consequence of Lemma~4.34, of~\thetag{4.39} and of property {\bf

3113: (2)} of Lemma~4.51, we conclude that for every $s\in [-1,1]$, our

3114: discs $\widehat{\mathcal{ Z}}_{ t,\chi,\nu:s}(\zeta)$ satisfy all the

3115: requirements {\bf (i)}, {\bf (ii)} and {\bf (iii)} of \S4.2 insuring

3116: that the set defined by

3117: \def\theequation{4.60}\begin{equation}

3118: \mathcal{HW}_s^+:=\left\{

3119: \widehat{\mathcal{ Z}}_{t,\chi,\nu:s}(\rho):\, 

3120: \vert t\vert < \varepsilon, \ 

3121: \vert \chi \vert < \varepsilon, \

3122: 0 < \nu < \varepsilon, \

3123: 1-\varepsilon < \rho < 1 

3124: \right\}

3125: \end{equation}

3126: is a local half-wedge of edge $(M_s^1)^+$ at the origin

3127: in the $\widehat{ z}$-coordinates, which corresponds 

3128: to the point $\gamma(s)$ in the $z$-coordinates. Coming back to 

3129: the $z$ coordinates, we may define the family of analytic discs

3130: \def\theequation{4.61}\begin{equation}

3131: \mathcal{ Z}_{t,\chi,\nu:s}(\zeta):= 

3132: \Phi_s^{-1}\left(\widehat{ \mathcal{ Z}}_{t,\chi,\nu:s}(\zeta)\right)

3133: \end{equation}

3134: and we shall construct the desired semi-local attached

3135: half-wedge $\mathcal{ HW}_\gamma^+$ of Proposition~4.6.

3136: 

3137: \subsection*{4.62.~Holomorphic extension to an 

3138: attached half-wedge} Indeed, we can now complete the proof of

3139: Proposition~4.6. Let us denote by $\widehat{z}=\Phi_s(z)$ the

3140: parametrized change of coordinates defined in \S4.11, where the point

3141: $\gamma(s)$ in $z$-coordinates corresponds to the origin in

3142: $\widehat{z}$-coordinates. Given an arbitrary holomorphic function

3143: $f\in \mathcal{ O}(\Omega)$ as in Proposition~4.6, by the change of

3144: coordinates $\widehat{ z}=\Phi_s(z)$ and by restriction to

3145: $(M_s^1)^+$, we get a CR function $\widehat{f}_s\in \mathcal{

3146: C}_{CR}^0\left((M_s^1)^+\cap U_1\right)$, for some small neighborhood

3147: $U_1$ of the origin in $\C^n$, whose size is uniform with respect to

3148: $s$. Thanks to an obvious generalization of the approximation

3149: Lemma~4.8 with a supplementary parameter $s\in [-1,1]$, we know that

3150: there exists a second uniform neighborhood $V_1\subset \subset U_1$ of

3151: the origin in $\C^n$ such that every continuous CR function in

3152: $\mathcal{ C}_{CR}^0\left( (M_s^1)^+ \cap U_1 \right)$ is uniformly

3153: approximable by polynomials on $(M_s^1)^+ \cap V_1$. In particular,

3154: this property holds for the CR function $\widehat{f}_s$. Furthermore,

3155: choosing $r_0$ and $\varepsilon$ sufficiently small, we can insure

3156: that all the discs $\widehat{\mathcal{ Z}}_{t, 

3157: \chi,\nu:s}(\zeta)$ are attached

3158: to $(M_s^1)^+ \cap V_1$. It then follows from the maximum principle

3159: applied to the approximating sequence of polynomials that for each

3160: $s\in [-1,1]$, the function $\widehat{f}_s$ extends holomorphically to

3161: the half-wedge defined by~\thetag{4.60}. Finally, we deduce that the

3162: holomorphic function $f\in \mathcal{ O}(\Omega)$ extends

3163: holomorphically to the half-wedge attached to the one-sided

3164: neighborhood $(M_\gamma^1)^+$ defined by

3165: \def\theequation{4.63}\begin{equation}

3166: \left\{

3167: \aligned

3168: \mathcal{HW}_\gamma^+:=\left\{

3169: \mathcal{ Z}_{t,\chi,\nu:s}(\rho):\, 

3170: \vert t\vert < \varepsilon, \ 

3171: \vert \chi \vert < \varepsilon, \ 

3172: 0< \nu < \varepsilon, \right. \\

3173: \left. 

3174: 1-\varepsilon < \rho < 1, \ 

3175: -1\leq s \leq s

3176: \right\}.

3177: \endaligned\right.

3178: \end{equation}

3179: Without shrinking $\Omega$ near the points $\mathcal{ Z}_{t, \chi,

3180: \nu:s}(-1)$ (otherwise, the crucial rank property of Lemma~4.34 would

3181: degenerate), we can shrink the open set $\Omega$ in a very thin

3182: neighborhood of the characteristic segment $\gamma$ in $M$ and we can

3183: shrink $\varepsilon >0$ if necessary in order that the intersection

3184: $\Omega \cap \mathcal{ HW}_\gamma^+$ is connected. By the principle of

3185: analytic continuation, this implies that there exists a well-defined

3186: holomorphic function $F\in \mathcal{ O}\left( \Omega \cup \mathcal{ HW

3187: }_\gamma^+ \right )$ with $F \vert_\Omega = f$.

3188: 

3189: The proof of Proposition~4.6 is complete.

3190: \qed

3191: 

3192: \subsection*{4.64.~Local half-wedge in CR dimension $m\geq 2$}

3193: Repeating the above constructions in the simpler case where the

3194: curve $\gamma$ degenerates to the point $p_1$ of Theorem~3.22 and

3195: adding some supplementary parameters along the $(m-1)$ remaining

3196: complex tangent directions of $M$ at $p_1$ for the constructions of

3197: analytic discs, we can show that under the assumptions of

3198: Theorem~3.22, there exists a local half-wedge $\mathcal{ HW}_{p_1}^+$ 

3199: at $p_1$ to which (shrinking $\Omega$ if

3200: necessary) every holomorphic function $f\in \mathcal{ O}(\Omega)$

3201: extends holomorphically. We shall not write down the details. 

3202: 

3203: \subsection*{4.65.~Transition} In the

3204: case $m=2$, using such a local half-wedge and applying the continuity

3205: principle along analytic discs of whose one boundary part is contained

3206: in $M$ and whose second boundary part is contained in the local

3207: half-wedge $\mathcal{ HW}_{p_1}^+$, we shall establish that $p_1$ is

3208: $\mathcal{ W}$-removable in Section~10 below. At present, in the more

3209: delicate case $m=1$, we shall pursue in Section~5 below our geometric

3210: constructions for the choice of a special point $p_{\rm sp}\in C$

3211: which satisfies the conclusion of Theorem~3.19 {\bf (i)}.

3212: 

3213: \section*{\S5.~Choice of a special point of $C_{\rm nr}$ 

3214: to be removed locally}

3215: 

3216: \subsection*{5.1.~Choice of a first supporting hypersurface} 

3217: Continuing with the proof of Theorem~3.19 {\bf (i)}, we shall now

3218: analyze and use the important geometric 

3219: condition $\mathcal{ F}_{M^1}^c\{C\}$

3220: defined in Theorem~1.2'. We first delineate a convenient geometric

3221: situation.

3222: 

3223: \def\thelemma{5.2}\begin{lemma} 

3224: Under the assumptions of Theorem~3.19 {\bf (i)}, there exists a

3225: $\mathcal{ C}^{2, \alpha}$-smooth segment $\gamma: [-1,1]\to M^1$

3226: running in characteristic directions, namely satisfying $d \gamma(s)/

3227: ds\in T_{ \gamma(s)}^c M\cap T_{\gamma (s)}M^1 \backslash \{0\}$ such

3228: that $\gamma (-1) \not \in C$, $\gamma(0) \in C$, $\gamma (1) \not \in

3229: C$, and there exists a $\mathcal{ C}^{ 1,\alpha}$-smooth hypersurface

3230: $H^1$ of $M^1$ with $\gamma\subset H^1$ which is foliated by

3231: characteristic segments close to $\gamma$, such that locally in a

3232: neighborhood of $H^1$, the closed subset $C$ is contained in

3233: $\gamma\cup (H^1)^-$, where $(H^1)^-$ denotes an open 

3234: one-sided neighborhood of $H^1$ in $M^1$.

3235: \end{lemma}

3236: 

3237: \proof

3238: By the assumption $\mathcal{ F}_{M^1}^c\{C\}$, there exists a

3239: characteristic curve $\widetilde{\gamma}: [-1,1]\to M^1$ with

3240: $\widetilde{\gamma}(-1)\not\in C$, $\widetilde{\gamma}(0)\in C$ and

3241: $\widetilde{\gamma}(1)\not\in C$, there exists a neighborhood

3242: $V_{\widetilde{\gamma}}^1$ of $\widetilde{\gamma}$ in $M^1$ and there

3243: exists a local $(n-1)$-dimensional submanifold $R^1$ passing through

3244: $\widetilde{\gamma}(0)$ which is transversal to $\widetilde{\gamma}$

3245: such that the semi-local projection $\pi_{\mathcal{ F}_{M^1}^c}:

3246: V_{\widetilde{\gamma}}^1\to R^1$ parallel to the characteristic curves

3247: maps $C$ onto the closed subset $\pi_{\mathcal{ F}_{M^1}^c}(C)$ with

3248: the property that $\pi_{\mathcal{ F}_{M^1}^c}(\widetilde{\gamma})$

3249: lies on the {\it boundary of $\pi_{\mathcal{ F}_{M^1}^c}(C)$ with

3250: respect to the topology of $R^1$}. This property is illustrated in the

3251: right hand side of the following figure.

3252: 

3253: \bigskip

3254: \begin{center}

3255: \input tube-ellipse.pstex_t

3256: \end{center}

3257: 

3258: However, we want in addition a foliated supporting hypersurface $H^1$,

3259: which does not necessarily exist in a neighborhood of

3260: $\widetilde{\gamma}$. To construct $H^1$, let us first straighten the

3261: characteristic lines in a neighborhood of $\widetilde{ \gamma}$,

3262: getting a product $[-1,1]\times [-\delta_1,\delta_1]^{n-1}$, for some

3263: $\delta_1>0$, equipped with coordinates

3264: $(s,\chi)=(s,\chi_2,\dots,\chi_n)\in \R\times \R^{n-1}$, so that the

3265: lines $\{\chi={\rm ct.}\}$ correspond to characteristic lines. Such a

3266: straightening is only of class $\mathcal{ C}^{1,\alpha}$, because the

3267: line distribution $T^cM\vert_{M^1} \cap TM^1$ is only of class

3268: $\mathcal{ C}^{1,\alpha}$. Clearly, we may assume that $\delta_1$ is

3269: so small that there exists $s_1$ with $0< s_1 <1$ such that the two

3270: cubes $[-1,-s_1]\times [-\delta_1,\delta_1]^{n-1}$ and $[s_1,1]\times

3271: [-\delta_1,\delta_1]^{n-1}$ do not meet the singularity $C$, as drawn

3272: in {\sc Figure~8} above.

3273: 

3274: We may identify the transversal $R^1$ with $[- \delta_1,

3275: \delta_1]^{n-1}$; then the projection of $\widetilde{\gamma}$ 

3276: is the origin of

3277: $R^1$. By assumption, $\pi_{ \mathcal{ F}_{M^1}^c}(C)$ is a proper

3278: closed subset of $R^1$ with the origin lying on its boundary. We can

3279: therefore choose a point $\chi_0$ in the interior of $R^1$ lying

3280: outside $\pi_{ \mathcal{ F}_{M^1 }^c }(C)$. Also, we can choose a

3281: small open $(n- 1)$-dimensional ball $Q_0$ centered at this point

3282: which is contained in the complement $R^1\backslash \pi_{ \mathcal{

3283: F}_{M^1 }^c}(C)$. Furtermore, we can include this ball in a one

3284: parameter family of $\mathcal{ C}^{1, \alpha}$-smooth domains $Q_\tau

3285: \subset R^1$, for $\tau \geq 0$, which are parts of ellipsoids

3286: stretched along the segment which joins the point $\chi_0$ with the

3287: origin of $R^1$.

3288: 

3289: We then consider the tube domains $[-1,1] \times Q_\tau$ in $[-1, 1]

3290: \times [- \delta_1, \delta_1]^{ n-1}$. Clearly, there exists the

3291: smallest $\tau_1>0$ such that the tube $[-1,1]\times Q_{\tau_1}$ meets

3292: the singularity $C$ on its boundary $[-1,1]\times \partial

3293: Q_{\tau_1}$. In particular, there exists a point $\chi_1\in \partial

3294: Q_{\tau_1}$ such that the characteristic segment $[-1,1]\times

3295: \{\chi_1\}$ intersects $C$. Increasing a little bit the curvature of

3296: $\partial Q_{\tau_1}$ in a neighborhood of $\chi_1$ if necessary, we

3297: can assume that $\pi_{\mathcal{ F}_{M^1}^c}(C)\cap

3298: \overline{Q_{\tau_1}}=\{\chi_1\}$ in a neighborhood of $\chi_1$.

3299: Moreover, since by construction the two segments $[-1,-s_1]\times

3300: \{\chi_1\}\cup [s_1,1]\times \{\chi_1\}$ do not meet $C$, we can

3301: reparametrize the characteristic segment $[-1,1]\times \{\chi_1\}$ as

3302: $\gamma: [-1,1]\to M^1$ with $\gamma(-1)\not\in C$, $\gamma(0)\in C$

3303: and $\gamma(1)\not\in C$. Since all characteristic lines are

3304: $\mathcal{ C}^{2,\alpha}$-smooth, we can choose the parametrization to

3305: be of class $\mathcal{ C}^{2,\alpha}$. For the supporting hypersurface

3306: $H^1$, it suffices to choose a piece of $[-1,1]\times \partial

3307: Q_{\tau_1}$ near $[-1,1]\times \{\chi_1\}$. By construction, this

3308: supporting hypersurface is only of class $\mathcal{ C}^{1,\alpha}$ and

3309: we have that $C$ is contained in $\gamma \cup (H^1)^-$ semi-locally in

3310: a neighborhood of $\gamma$, as desired. This completes the proof of

3311: Lemma~5.2.

3312: \endproof

3313: 

3314: \subsection*{5.3.~Field of cones on $M^1$}

3315: With the characteristic segment $\gamma$ constructed in Lemma~5.2, by

3316: an application of Proposition~4.6, we deduce that there exists a

3317: semi-local half-wedge $\mathcal{ HW}_\gamma^+$ attached to

3318: $(M_\gamma^1)^+\cap V_\gamma$, for some neighborhood $V_\gamma$ of

3319: $\gamma$ in $M$, to which $\mathcal{ O}(\Omega)$ extends

3320: holomorphically.

3321: 

3322: Then, we remind that by~\thetag{4.37}, \thetag{4.39}

3323: and~\thetag{4.50}, for all $t$ with $\vert t \vert < \varepsilon$, the

3324: point $\widehat{ \mathcal{ Z}}_{ t,\chi,0:s}(1)$ identifies with the

3325: point $\widehat{ p}( \chi,0:s)\in M_s^1$ defined in~\thetag{4.37}

3326: (which is independent of $t$) and the mapping $\chi \mapsto \widehat{

3327: \mathcal{ Z}}_{t,\chi,0:s}(1)\in M_s^1$ is a local diffeomorphism.

3328: 

3329: Sometimes in the sequel, we shall denote the disc $\mathcal{ Z}_{t,

3330: \chi, \nu:s} (\zeta) \equiv \Phi_s^{-1} \left( \widehat{ \mathcal{

3331: Z}}_{t,\chi, \nu:s}(\zeta) \right)$ defined in~\thetag{4.61} by

3332: $\mathcal{ Z}_{t, \chi_1, \chi',\nu:s}(\zeta)$, where $\chi'= (

3333: \chi_2, \dots, \chi_n) \in \R^{n-1}$. Since the characteristic curve

3334: is directed along the $x_1$-axis, which is transversal in $T_0M_s^1$

3335: to the space $\{(0,\chi')\}$, it follows that the mapping $(s, \chi')

3336: \longmapsto \mathcal{ Z}_{

3337: t,0,\chi',0:s}(1)=\Phi_s^{-1}\left(\widehat{ p}(0,\chi',0:s)\right)$

3338: is, independently of $t$, a diffeomorphism onto its image for $s\in

3339: [-1,1]$ and for $\chi'$ close to the origin in $\R^{n-1}$. To fix

3340: ideas, we shall let $\chi'$ vary in the {\it closed}\, cube

3341: $[-\varepsilon,\varepsilon]^{n-1}$ (analogously to the fact that $s$

3342: runs in the {\it closed}\, interval $[-1,1]$) and we shall denote by

3343: $V_\gamma^1$ the closed subset of $M^1$ which is the image of this

3344: diffeomorphism.

3345: 

3346: At every point $p:=\mathcal{ Z}_{t,0,\chi',0:s}(1)= \mathcal{

3347: Z}_{0,0,\chi',0:s}(1)$ of this neighborhood $V_\gamma^1$, we define an

3348: open infinite oriented cone contained in the $n$-dimensional linear

3349: space $T_pM^1$ by

3350: \def\theequation{5.4}\begin{equation}

3351: {\sf C}_p:=\R^+\cdot 

3352: \left\{

3353: \frac{\partial \mathcal{ Z}_{

3354: t,0,\chi',0:s}}{\partial \theta} 

3355: (1): \ \vert t \vert < \varepsilon

3356: \right\}.

3357: \end{equation}

3358: The fact that ${\sf C}_p$ is indeed an open cone follows

3359: from Lemma~4.34, from~\thetag{4.61} and

3360: from the fact that $\Phi_s^{-1}$ is a biholomorphism. 

3361: This cone contains in its interior the nonzero vector 

3362: \def\theequation{5.5}\begin{equation}

3363: v_p^0:= \frac{\partial \mathcal{ Z}_{0,0,\chi',0:s}}{\partial \theta}

3364: (1)\in {\sf C}_p\subset T_pM^1 \backslash \{0\}.

3365: \end{equation}

3366: We shall say that the cone ${\sf C}_p$ is the {\sl cone created at $p$

3367: by the semi-local attached half-wedge $\mathcal{ HW}_\gamma^+$}

3368: (more precisely, by the family of analytic discs which covers

3369: this semi-local half-wedge).

3370: 

3371: As $p$ varies, the mapping $p\mapsto {\sf C}_p$ constitutes what we

3372: shall call a {\sl field of cones over $V_\gamma^1$}, {\it see} {\sc

3373: Figure~2} in Section~2 above and {\sc Figure~9} just below for

3374: illustrations.

3375: 

3376: \bigskip

3377: \begin{center} 

3378: \input field-cones.pstex_t

3379: \end{center}

3380: 

3381: The mapping $p\mapsto v_p^0$ defines a $\mathcal{

3382: C}^{1,\alpha-0}$-smooth vector field tangent to $M^1$. This vector

3383: field is contained in the field of cones $p\mapsto {\sf C}_p$. Over

3384: $V_\gamma^1$, we can also consider a nowhere zero characteristic

3385: vector field $X$ whose direction agrees with the orientation of

3386: $\gamma$ and which satisfyies $\exp(sX)(\gamma(0))=\gamma(s)$ for all

3387: $s\in [-1,1]$. Furthermore, for every $p\in V_\gamma^1$, we define the

3388: {\sl filled cone}

3389: \def\theequation{5.6}\begin{equation}

3390: {\sf FC}_p:=\R^+\cdot \left\{

3391: \lambda \cdot X_p+

3392: (1-\lambda)\cdot v_p: \ 

3393: 0\leq \lambda < 1, \ 

3394: v_p\in {\sf C}_p

3395: \right\}.

3396: \end{equation}

3397: In the right hand side of {\sc Figure~10} just below, in the tangent

3398: space $T_pM^1$ equipped with linear coordinates $(x_1,\dots,x_n)$ such

3399: that the characteristic direction $T_p^cM\cap T_pM^1$ is directed

3400: along the $x_1$-axis, we draw ${\sf C}_p$, its filling ${\sf FC}_p$

3401: and its projection $\pi'({\sf C}_p)$ onto the $(x_2,,\dots,x_n)$-space

3402: parallel to the $x_1$-axis.

3403: 

3404: \bigskip

3405: \begin{center}

3406: \input filled-cone.pstex_t

3407: \end{center}

3408: 

3409: Given an arbitrary nonzero vector $v_p \in {\sf C}_p$, where $p\in

3410: V_\gamma^1$, it may be checked that a small neighborhood of the origin

3411: in the positive half-line $\R^+\cdot Jv_p$ generated by $Jv_p$, where

3412: $J$ denotes the complex structure of $T\C^n$, is contained in the

3413: attached half-wedge $\mathcal{ HW}_\gamma^+$. More generally, the same

3414: property holds for every nonzero vector $v_p$ which belongs to the

3415: filled cone ${\sf FC}_p$. In fact, we shall establish that analytic

3416: discs which are half-attached to $M^1$ at $p$, namely send

3417: $\partial^+\Delta:=\{\zeta\in \partial \Delta: \ {\rm Re}\, \zeta \geq

3418: 0\}$ to $M^1$, which are closed to the complex line $v_p+Jv_p$ in

3419: $\mathcal{ C}^1$ norm and which are sufficiently small are contained

3420: in the attached half-wedge $\mathcal{ HW}_\gamma^+$. The interest of

3421: this property and the reason why we have defined fields of cones and

3422: their filling will be more appearant in Sections~8 and~9 below, where

3423: we apply the continuity principle to achieve the proof of Theorem~3.19

3424: {\bf (i)}. A precise statement, involving families of analytic discs

3425: $A_c(\zeta)$ which will be constructed in Section~7 below, is as

3426: follows. For the proof of a more precise statement involving families

3427: of analytic discs $A_{x,v:c}^1(\zeta)$, we refer to Section~7 and

3428: especially to Lemma~8.3 {\bf ($\mathbf{9_1}$)} below.

3429: 

3430: \def\thelemma{5.7}\begin{lemma}

3431: Fix a point $p\in V_\gamma^1$ and a vector $v_p$ in the cone ${\sf

3432: C}_p$ created by the semi-local attached half-wedge $\mathcal{

3433: HW}_\gamma^+$ at $p$. Suppose that there exist two constants $c_1>0$

3434: and $\Lambda_1 >1$ such that for every $c$ with $0< c \leq c_1$, there

3435: exists a $\mathcal{ C}^{2,\alpha-0}$-smooth analytic disc $A_c(\zeta)$

3436: with $A_c (\partial^+\Delta) \subset M^1$,

3437: such that

3438: \begin{itemize}

3439: \item[{\bf (i)}]

3440: The positive half-line generated by the boundary of $A_c$ at $\zeta=1$

3441: coincides with the positive half-line generated by $v_p$, namely

3442: $\R^+\cdot \frac{\partial A_c}{\partial \theta}(1) \equiv \R^+ \cdot

3443: v_p$.

3444: \item[{\bf (ii)}]

3445: $\left\vert A_c (\zeta) \right\vert \leq c^ 2\cdot \Lambda_1$ for all

3446: $\zeta \in \overline{ \Delta}$ and $c \cdot \frac{ 1}{ \Lambda_1} \leq

3447: \left\vert \frac{ \partial A_c}{\partial \theta}(1) \right\vert \leq

3448: c\cdot \Lambda_1$.

3449: \item[{\bf (iii)}]

3450: $\left\vert \frac{\partial A_c}{\partial \theta}\left( \rho

3451: e^{i\theta}\right) - \frac{\partial A_c}{\partial \theta}(1)

3452: \right\vert\leq c^2 \cdot \Lambda_1$ for all $\zeta=\rho e^{i\theta} \in

3453: \overline{\Delta}$.

3454: \end{itemize}

3455: If $c_1$ is sufficiently small, then for every $c$ with $0< c \leq

3456: c_1$, the set $A_c\left(\overline{\Delta} \backslash \partial^+\Delta

3457: \right)$ is contained in the semi-local half-wedge $\mathcal{

3458: HW}_\gamma^+$. Furthermore, the same conclusion holds if the nonzero

3459: vector $v_p$ belongs to the filled cone ${\sf FC}_p$.

3460: \end{lemma}

3461: 

3462: \subsection*{5.8.~Choice of the special point $p_{\rm sp}$ in 

3463: the CR dimension $m=1$ case} We can now answer the question raised

3464: after the statement of Theorem~3.19 {\bf (i)}, which was the main

3465: purpose of the present Section~5: {\it How to choose the special point

3466: $p_{\rm sp}$ to be removed locally}? In the following statement,

3467: property {\bf (ii)} will be really crucial for the removal of $p_{\rm

3468: sp}$, {\it see}\, in advance Proposition~5.12 below.

3469: 

3470: \def\thelemma{5.9}\begin{lemma}

3471: Let $\gamma$ be the characteristic segment constructed in Lemma~5.2

3472: above. Let $\mathcal{ HW }_\gamma^+$ be the semi-local attached

3473: half-wedge of edge $(M_\gamma^1 )^+\cap V_\gamma$ constructed in

3474: Proposition~4.6 above, and let $p\mapsto {\sf FC }_p$ be the filled

3475: field of cones created in a closed neighborhood $V_\gamma^1$ of

3476: $\gamma$ in $M^1$ by this semi-local

3477: attached half-wedge $\mathcal{ HW }_\gamma^+$. Then

3478: there exists a special point $p_{\rm sp}\in V_\gamma^1$ such that the

3479: following two geometric properties are fulfilled{\rm :}

3480: \begin{itemize}

3481: \item[{\bf (i)}]

3482: There exists a $\mathcal{ C}^{2, \alpha}$-smooth local supporting

3483: hypersurface $H_{ \rm sp}$ of $M^1$ passing through $p_{\rm sp}$ such

3484: that, locally in a neighborhood of $p_{ \rm sp}$, the closed subset

3485: $C$ is contained in $(H_{ \rm sp})^-\cup \{p_{ \rm sp}\}$, where

3486: $(H_{\rm sp})^-$ denotes an open one-sided neighborhood of

3487: $H_{\rm sp}$ in $M^1$.

3488: \item[{\bf (ii)}]

3489: There exists a nonzero vector $v_{\rm sp}\in T_{p_{\rm sp}}H_{\rm sp}$

3490: which belongs to the filled cone ${\sf FC}_{p_{\rm sp}}$.

3491: \end{itemize}

3492: \end{lemma}

3493: 

3494: \proof

3495: According to Lemma~5.2, the singularity $C$ is contained

3496: in $\gamma\cup (H^1)^-$, where $H^1$ is a $\mathcal{ C}^{

3497: 1,\alpha}$-smooth hypersurface containing $\gamma$ which is foliated

3498: by characteristic segments.

3499: If $\lambda\in [0,1)$ is very close to $1$, the vector field

3500: over $V_\gamma^1$ defined by

3501: \def\theequation{5.10}\begin{equation}

3502: p\longmapsto 

3503: v_p^\lambda:= \lambda\cdot X_p +(1-\lambda)\cdot v_p

3504: \in T_pM^1

3505: \end{equation}

3506: is very close to the characteristic vector field $X_p$, so the

3507: integral curves of $p\mapsto v_p^\lambda$ are very close to the

3508: integral curves of $p\mapsto X_p$, which are the characteristic

3509: segments. If $\lambda$ is sufficiently close to $1$, we can choose a

3510: subneighborhood $V_\gamma^\lambda\subset V_\gamma^1$ of $\gamma$ which

3511: is foliated by integral curves of $p\mapsto v_p^\lambda$. As in

3512: Lemma~5.2, let us fix an $(n-1)$-dimensional submanifold $R^1$

3513: transversal to $\gamma$ and passing through $\gamma(0)$. Since the

3514: vector field $p\mapsto v_p^\lambda$ is very close to the

3515: characteristic vector field, it follows that after projection onto

3516: $R^1$ parallelly to the integral curves of $p\mapsto v_p^\lambda$, the

3517: closed set $C\cap V_\gamma^\lambda$ is again a proper closed subset of

3518: $R^1$. We notice that, by its very definition, the vector

3519: $v_p^\lambda$ belongs to the filled cone ${\sf FC}_p$ for all $p

3520: \in V_\gamma^\lambda$.

3521: 

3522: We can proceed exactly as in the

3523: proof of Lemma~5.2 with the foliation of $V_\gamma^\lambda$ induced by

3524: the integral curves of the vector field $p\mapsto v_p^\lambda$,

3525: instead of the characteristic foliation, except that we want a

3526: supporting hypersurface $H_{\rm sp}$ which is of class $\mathcal{

3527: C}^{2,\alpha}$. Consequently, we first approximate the vector field

3528: $p\mapsto v_p^\lambda$ by a new vector field $p\mapsto \widetilde{

3529: v}_p^\lambda$ whose coefficients are of class $\mathcal{ C}^{2,

3530: \alpha}$ (with respect to every local graphing function of $M^1$) and

3531: which is very close to the vector field $p\mapsto v_p^\lambda$ in

3532: $\mathcal{ C}^1$-norm. Again, we get a subneighborhood $\widetilde{

3533: V}_\gamma^\lambda \subset V_\gamma^\lambda$ of $\gamma$ which is

3534: foliated by integral curves of $p\mapsto \widetilde{ v}_p^\lambda$ and

3535: a projection of $C\cap \widetilde{ V}_\gamma^\lambda$ which is a

3536: proper closed subset of $R^1$. Moreover, if the approximation is

3537: sufficiently sharp, we still have $\widetilde{ v}_p^\lambda \in {\sf

3538: FC}_p$ for all $p \in \widetilde{ V}_\gamma^\lambda$. Then by

3539: repeating the reasoning which yielded Lemma~5.2, using the foliation

3540: of $\widetilde{ V}_\gamma^\lambda$ induced by $p \mapsto \widetilde{ v

3541: }_p^\lambda$, we deduce that there exists an integral curve

3542: $\widetilde{ \gamma}$ of the vector field $p\mapsto \widetilde{

3543: v}_p^\lambda$ satisfying (after reparametrization) $\widetilde{\gamma}

3544: (-1) \not \in C$, $\widetilde{\gamma}(0)\in C$ and

3545: $\widetilde{\gamma}(1) \not\in C$, together with a $\mathcal{ C}^{

3546: 2,\alpha}$-smooth supporting hypersurface $\widetilde{ H}$ of

3547: $\widetilde{ V}_\gamma^\lambda$ which contains $\widetilde{ \gamma}$

3548: such that $C$ is contained in $\widetilde{ \gamma}\cup (\widetilde{

3549: H})^-$. The fact that $\widetilde{ H}$ is of class $\mathcal{

3550: C}^{2,\alpha}$ is due to the $\mathcal{ C}^{2,\alpha}$-smoothness of

3551: the foliation on $\widetilde{ V}_\gamma^\lambda$ induced by the vector

3552: field $p\mapsto \widetilde{ v }_\gamma^\lambda$. In {\sc Figure~11}

3553: just below, suited for the case where $M^1$ is two-dimensional, we

3554: have drawn as a dotted line the limiting integral curve $\widetilde{

3555: \gamma}$ of $p\mapsto \widetilde{ v}_p^\lambda$ having the property

3556: that $C$ lies in one closed side of $\widetilde{ \gamma}$ in

3557: $\widetilde{ V}_\gamma^\lambda$.

3558: 

3559: \bigskip

3560: \begin{center}

3561: \input perturbed-vector-field.pstex_t

3562: \end{center}

3563: 

3564: To conclude the proof of Lemma~5.9, for the desired special point $p_{

3565: \rm sp}$, it suffices to choose $\widetilde{ \gamma}(0)$. For the

3566: desired local supporting hypersurface $H_{p_{\rm sp}}$, we cannot

3567: choose directly a piece of $\widetilde{ H}$ passing through $p_{\rm

3568: sp}$, because an open interval contained in $C\cap \widetilde{

3569: \gamma}$ may well be contained in $\widetilde{ H}$ by the construction

3570: in the proof of Lemma~5.2 that we have just reapplied. Fortunately,

3571: since we know that locally in a neighborhood of $p_{\rm sp}$, the

3572: closed subset $C$ is contained in $(\widetilde{ H})^-\cup \widetilde{

3573: \gamma}$, it suffices to choose for the desired supporting

3574: hypersurface $H_{p_{\rm sp}}\subset M^1$ a piece of a $\mathcal{

3575: C}^{2,\alpha}$-smooth hypersurface passing through $p_1$, tangent to

3576: $\widetilde{ H}$ at $p_1$ and satisfying $H_{p_{\rm sp}}\backslash

3577: \{p_{\rm sp}\}\subset (\widetilde{ H})^+$ in a neighborhood of $p_{\rm

3578: sp}$. Finally, for the nonzero vector $v_{\rm sp}$, it suffices to

3579: choose any positive multiple of the vector $\widetilde{ v}_{ p_{\rm

3580: sp}}^\lambda$. This completes the proof of Lemma~5.9.

3581: \endproof

3582: 

3583: In Section~8 below, property {\bf (ii)} of Lemma~5.9 together with the

3584: observation made in Lemma~5.7 will be crucial for the local $\mathcal{

3585: W}$-removability of the special point $p_{\rm sp}$.

3586: 

3587: \subsection*{5.11.~Main removability proposition in the 

3588: CR dimension $m=1$ case} We can now formulate the main removability

3589: proposition to which Theorem~3.19 {\bf (i)} is now reduced. From now

3590: on, we localize the situation at $p_{\rm sp}$, we denote this point

3591: simply by $p_1$, we denote its supporting hypersurface simply by $H^1$

3592: and we denote its associated vector simply by $v_1$. Furthermore, we

3593: localize at $p_1$ the family of analytic discs considered in Section~4

3594: for the construction of the semi-local attached half-wedge $\mathcal{

3595: HW}_\gamma^+$, hence we get a family of analytic discs $\mathcal{

3596: Z}_{t, \chi, \nu }( \zeta)$ which satisfy properties {\bf (i)}, {\bf

3597: (ii)} and {\bf (iii)} of \S4.2 and which generate a local half-wedge

3598: $\mathcal{ HW}_{p_1}^+ \subset \mathcal{ HW}_\gamma^+$ as defined

3599: in~\thetag{4.5}. At present, we deduce from our constructions achieved

3600: in Section~4 and in the beginning of Section~5 that Theorem~3.19 {\bf

3601: (i)} is now a consequence of the following main Proposition~5.12 just

3602: below, to the proof of which Sections~6, 7, 8 and~9 below are

3603: devoted. From now on, all our geometric considerations will be

3604: localized at the special point $p_1$.

3605: 

3606: \def\theproposition{5.12}\begin{proposition} 

3607: Let $M$ be a $\mathcal{ C}^{2, \alpha}$-smooth generic submanifold of

3608: $\C^n$ of CR dimension $m =1$ and of codimension $d = n-1 \geq 1$, let

3609: $M^1\subset M$ be a $\mathcal{ C}^{2, \alpha }$-smooth

3610: one-codimensional submanifold which is generic in $\C^n$, let $p_1 \in

3611: M^1$, let $H^1\subset M^1$ be a $\mathcal{ C}^{2, \alpha}$-smooth

3612: one-codimensional submanifold of $M^1$ passing through $p_1$ and let

3613: $(H^1 )^-$ denote an open local one-sided neighborhood of $H^1$ in

3614: $M^1$. Let $C \subset M^1$ be a {\rm nonempty} proper closed subset of

3615: $M^1$ with $p_1 \in C$ which is situated, locally in a neighborhood of

3616: $p_1$, only in one side of $H^1$, namely $C \subset (H^1)^- \cup \{

3617: p_1\}$. Let $\Omega$ be a neighborhood of $M \backslash C$ in $\C^n$,

3618: let $\mathcal{ HW}_{ p_1}^+$ be a local half-wedge of edge $(M^1)^+$

3619: at $p_1$ generated by a family of analytic discs $\mathcal{

3620: Z}_{t,\chi,\nu}(\zeta)$ satisfying the properties {\bf (i)}, {\bf

3621: (ii)} and {\bf (iii)} of \S4.2, let ${\sf C}_{p_1}\subset T_{p_1}M^1$

3622: be the cone created by $\mathcal{ HW}_{p_1}^+$ at $p_1$ and let ${\sf

3623: FC}_{p_1}$ be its filling. {\rm As a main assumption}, suppose that

3624: there exists a nonzero vector $v_1\in T_{p_1} H^1$ which belongs to

3625: the filled cone ${\sf FC}_{p_1}$.

3626: \begin{itemize}

3627: \item[{\bf (I)}]

3628: If $v_1$ does not belong to $T_{p_1}^cM$, then there exists a local

3629: wedge $\mathcal{ W}_{ p_1}$ of edge $M$ at $(p_1,Jv_1)$ with

3630: $\mathcal{ W}_{ p_1} \cap \left[\Omega \cup \mathcal{

3631: HW}_{p_1}^+\right]$ connected {\rm (}shrinking $\Omega\cup \mathcal{

3632: HW}_{p_1}^+$ if necessary{\rm )} such that for every holomorphic

3633: function $f\in \mathcal{ O} \left( \Omega\cup \mathcal{ HW}_{p_1}^+

3634: \right)$, there exists a holomorphic function $F \in \mathcal{ O}

3635: \left( \mathcal{ W}_1 \cup \left[\Omega \cup \mathcal{ HW}_1^+

3636: \right]\right)$ with $F\vert_{\Omega\cup \mathcal{ HW}_{p_1}^+} =f$.

3637: \item[{\bf (II)}]

3638: If $v_1$ belongs to $T_{p_1}^cM$, then there exists a neighborhood

3639: $\omega_{p_1}$ of $p_1$ in $\C^n$ with $\omega_{p_1} \cap \left[\Omega

3640: \cup \mathcal{ HW}_{p_1}^+\right]$ connected {\rm (}shrinking

3641: $\Omega\cup \mathcal{ HW}_{p_1}^+$ if necessary{\rm )} such that for

3642: every holomorphic function $f\in \mathcal{ O} \left( \Omega\cup

3643: \mathcal{ HW}_{p_1}^+ \right)$, there exists a holomorphic function $F

3644: \in \mathcal{ O} \left( \omega_{p_1} \cup \left[\Omega \cup \mathcal{

3645: HW}_1^+ \right]\right)$ with $F\vert_{\Omega\cup \mathcal{

3646: HW}_{p_1}^+} =f$.

3647: \end{itemize}

3648: \end{proposition}

3649: 

3650: In the CR dimension $m\geq 2$ case, we observe that an

3651: analogous main removability proposition may be formulated simply by

3652: adding to the assumptions of Proposition~3.22 a local half-wedge

3653: $\mathcal{ HW}_{p_1}^+$, whose existence was established in \S4.64

3654: above. The remainder of Section~5, and then Section~6, Section~7,

3655: Section~8 and Section~9 below will be entirely devoted to the proof of

3656: Proposition~5.12.

3657: 

3658: \subsection*{5.13.~A dichotomy}

3659: Under the assumptions of Proposition~5.12, we shall indeed 

3660: distinguish two cases: 

3661: \begin{itemize}

3662: \item[{\bf (I)}]

3663: The nonzero vector $v_1$ does not belong to the characteristic

3664: direction $T_{p_1}M^1\cap T_{p_1}^cM$.

3665: \item[{\bf (II)}]

3666: The nonzero vector $v_1$ belongs to the characteristic direction

3667: $T_{p_1}M^1\cap T_{p_1}^cM$.

3668: \end{itemize}

3669: 

3670: We must clarify the main assumption that $v_1$ belongs to the filling

3671: ${\sf FC}_{p_1}$

3672: of the cone ${\sf C }_{ p_1}\subset T_{p_1} M^1$ created by the local

3673: half-wedge $\mathcal{ HW}_{ p_1}^+$. As we have observed in \S4.2, in

3674: the (generic) situation of Case {\bf (I)}, a local half-wedge may be

3675: represented geometrically as the intersection of a (complete) local

3676: wedge of edge $M$ at $p_1$, with a local one-sided neighborhood $(N^1

3677: )^+$ of a hypersurface $N^1$ passing through $p_1$, which is

3678: transversal to $M$ and which satisfies $N^1 \cap M \equiv M^1$ in a

3679: neighborhood of $p_1$. The slope of the tangent space $T_{p_1} N^1$ to

3680: $N^1$ at $p_1$ with respect to the tangent space $T_{p_1} M$ to $M$ at

3681: $p_1$ may be understood in terms of the cone ${\sf C}_{p_1}$, as we

3682: will now explain. Afterwards, we shall consider Case {\bf (II)}

3683: separately.

3684: 

3685: \subsection*{5.14.~Cones, filled cones, subcones and 

3686: local description of half-wedges in Case (I)} As in some of the

3687: assumptions of Proposition~5.12, let $M$ be a $\mathcal{ C}^{2,

3688: \alpha}$-smooth generic submanifold of $\C^n$ of CR dimension $m=1$

3689: and of codimension $d=n-1 \geq 1$, let $p_1 \in M$, let $M^1$ be a

3690: $\mathcal{ C}^{2, \alpha}$-smooth one-codimensional submanifold of $M$

3691: passing through $p_1$. For the sake of concreteness, it will be

3692: convenient to work in a holomorphic coordinate system $z=(z_1, \dots,

3693: z_n)= (x_1+i y_1, \dots, x_n+ iy_n)$ centered at $p_1$ in which $T_{

3694: p_1}M=\{y_2= \cdots =y_n =0\}$ and $T_{ p_1}M^1=\{y_1= y_2= \cdots=

3695: y_n=0\}$ (the existence of such a coordinate system which straightens

3696: both $T_{p_1}M$ and $T_{p_1}M^1$ is a direct consequence of the

3697: considerations of \S4.11). Let $\pi': T_{p_1}M^1\to T_{p_1}M^1/

3698: (T_{p_1}M^1\cap T_{p_1}^cM)$ denote the canonical projection, namey

3699: $\pi'(x_1,x_2,\dots,x_n)=(x_2,\dots,x_n)$. Sometimes, we shall denote

3700: the coordinates by $(z_1,z')= (x_1+iy_1,x'+iy')\in \C\times

3701: \C^{n-1}$. In these coordinates, the characteristic direction is given

3702: by the $x_1$-axis and we may assume that the tangent plane at $p_1$ of

3703: the one-sided neighborhood $(M^1)^+$ is given by

3704: $T_{p_1}(M^1)^+=\{y'=0, \ y_1>0\}$.

3705: 

3706: Let ${\sf C}_{ p_1} \subset T_{p_1} M^1$ be the infinite open cone

3707: created by $\mathcal{ HW}_{ p_1}^+$ at $p_1$ and let ${\sf FC}_{ p_1}

3708: \subset T_{ p_1} M^1$ be its filling. Let ${\sf C}_{ p_1}':= \pi'

3709: \left( {\sf C}_{ p_1} \right)$ be its projection onto the $x'$-space,

3710: which yields an $(n-1)$-dimensional infinite cone in the $x'$-space,

3711: open with respect to its topology. Notice that, by the

3712: definition~\thetag{ 5.6} of the filling (along the characteristic

3713: direction) of a cone, the two projections $\pi'({\sf C}_{p_1})$ and

3714: $\pi'({\sf FC}_{p_1})$ are identical. We now need to explain how

3715: these three cones ${\sf C}_{p_1}$, ${\sf FC}_{p_1}$, ${\sf C}_{p_1}'$

3716: and the nonzero vector $v_1\in {\sf FC}_{p_1}$ are disposed,

3717: geometrically, {\it see} {\sc Figure~12} just below.

3718: 

3719: \bigskip

3720: \begin{center}

3721: \input cones-half-wedge.pstex_t

3722: \end{center}

3723: 

3724: Because the disc $\mathcal{ Z}_{ t, \chi, \nu}$ of Proposition~5.12

3725: (which is a localization in a neighborhood of the special point of the

3726: discs constructed in Section~4) is small, the tangent vector

3727: $\frac{\partial \mathcal{ Z}_{0, 0,0 }}{ \partial \theta} (1)$ is

3728: necessarily close to the complex tangent plane $T_{p_1 }^cM$: this may

3729: be checked directly by differentiating Bishop's equation~\thetag{4.40}

3730: with respect to $\theta$, using the fact that the $\mathcal{

3731: C}^1$-norm of $\Phi'$ is small. Moreover, since this vector $\frac{

3732: \partial \mathcal{ Z}_{ 0,0,0 }}{ \partial \theta }(1)$ also belongs

3733: to $T_{p_1} M^1$, it is in fact close to the positive

3734: $x_1$-axis. Furthermore, since the vector $v_1$ belongs to the filling

3735: of the open cone ${\sf C}_{ p_1}$ which contains the vector $\frac{

3736: \partial \mathcal{ Z}_{ 0,0,0 }}{ \partial \theta }(1)$, and since in

3737: the proof of Lemma~5.9 above we have chosen the special point, the

3738: supporting hypersurface and the vector $v_1$ with a parameter

3739: $\lambda$ very close to $1$, so that the vector field $p \mapsto

3740: \widetilde{ v}_p^\lambda$ was very close to the characteristic vector

3741: field $p \mapsto X_p$, it follows that the vector $v_1\equiv

3742: \widetilde{ v}_{ p_{ \rm sp}}^\lambda$ is even closer to the positive

3743: $x_1$-axis. However, we suppose in Case {\bf (I)} that $v_1$ is {\it

3744: not}\, directed along the $x_1$-axis, so $v_1$ has coordinates

3745: $(v_{1;1},v_{ 2;1}, \dots,v_{ n;1}) \in \R^n$ with $v_{1;1}>0$, with

3746: $\vert v_{j;1}\vert < < v_{ 1;1}$ for $j=2, \dots,n$ and with at least

3747: one $v_{j;1}$ being nonzero.

3748: 

3749: We need some general terminology. Let ${\sf C}$ be an open infinite

3750: cone in a real linear subspace $E$ of dimension $q\geq 1$. We say

3751: that ${\sf C}'$ is a {\sl proper subcone} and we write ${\sf

3752: C}'\subset \subset {\sf C}$ ({\it see} the left hand side of {\sc

3753: Figure~10} above for an illustration) if the intersection of ${\sf

3754: C}'$ with the unit sphere of $E$ is a relatively compact subset of the

3755: intersection of ${\sf C}$ with the unit sphere of $E$, this property

3756: being independent of the choice of a norm on $E$. We say that ${\sf

3757: C}$ is a {\sl linear cone} if it may be defined by ${\sf C}=\{x\in E:

3758: \ \ell_1(x)>0,\dots,\ell_q(x)>0\}$ for some $q$ linearly independent

3759: real linear forms $\ell_1,\dots,\ell_q$ on $E$.

3760: 

3761: In the $(x_2,\dots,x_n)$-space, we now choose an open infinite

3762: strictly convex linear proper subcone ${\sf C}_1'\subset \subset {\sf

3763: C }_{ p_1}'$ with the property that $v_1$ belongs to its filling ${\sf

3764: FC}_1'$, {\it cf.} {\sc Figure~12} above. Here, we may assume that

3765: ${\sf C}_1'$ is described by $(n-1)$ strict inequalities $\ell_1'(x')

3766: >0, \dots, \ell_{n-1}'(x')>0$, where the $\ell_j'( x')$ are linearly

3767: independent linear forms. It then follows that there exists a linear

3768: form $\sigma(x_1,x')$ of the form $\sigma(x_1, x')= x_1+a_2x_2+ \cdots

3769: +a_nx_n$ such that the original filled cone ${\sf FC}_{p_1}$ is

3770: contained in the linear cone

3771: \def\theequation{5.15}\begin{equation}

3772: {\sf C}_1:= \left\{

3773: (x_1,x')\in \R^n : \

3774: \ell_1'(x')>0,\dots,

3775: \ell_{n-1}'(x')>0, \ 

3776: \sigma(x_1,x')>0

3777: \right\},

3778: \end{equation}

3779: which contains the vector $v_1$. This cone is automatically filled,

3780: namely ${\sf C}_1\equiv {\sf FC}_1$.

3781: 

3782: We remind that by genericity of $M$, the complex structure $J$ of

3783: $T\C^n$ induces an isomorphim $T_{p_1}M/T_{p_1}^cM \to

3784: T_{p_1}\C^n/T_{p_1}M$. Hence $J{\sf C}_{p_1}'$ and $J{\sf C}_1'$ are

3785: open infinite strictly convex linear proper cones in

3786: $T_{p_1}\C^n/T_{p_1}M\cong \{(0,y')\in \C^n\}$. Since $J{\sf C}_1'$

3787: is a proper subcone of $J{\sf C}_{p_1}'$ and since in the classical

3788: definition of a wedge, only the projection of the cone on the quotient

3789: space $T_{p_1}M/T_{p_1}^cM$ has a contribution to the wedge, it then

3790: follows that the complete wedge $\mathcal{ W}_{p_1}$ associated to the

3791: family $\mathcal{ Z}_{t,\chi,\nu}(\zeta)$ ({\it cf.} the paragraph

3792: after~\thetag{4.5}) contains a wedge of the form

3793: \def\theequation{5.16}\begin{equation}

3794: \mathcal{ W}_1:=\left\{

3795: p+{\sf c}_1': \ p\in M, \ 

3796: {\sf c}_1' \in J{\sf C}_1'

3797: \right\}\cap \Delta_n(p_1,\delta_1),

3798: \end{equation}

3799: for some $\delta_1$ with $0 < \delta_1 < \varepsilon$, where

3800: $\varepsilon$ is as in \S4.2. Furthermore, as observed in \S4.2,

3801: there exists a $\mathcal{ C}^{2, \alpha}$-smooth hypersurface $N^1$ of

3802: $\C^n$ passing through $p_1$ with the property that $N^1 \cap M \equiv

3803: M^1$ locally in a neighborhood of $p_1$ such that, shrinking $\delta_1

3804: >0$ if necessary, the local half-wedge $\mathcal{ HW}_{ p_1}^+$

3805: contains a local half-wedge $\mathcal{ HW}_1^+$ of edge $(M^1)^+$ at

3806: $p_1$ which is described as the geometric intersection of the complete

3807: wedge $\mathcal{ W}_{p_1}$ with a one-sided neighborhood $(N^1)^+$,

3808: namely

3809: \def\theequation{5.17}\begin{equation}

3810: \mathcal{ HW}_1^+:=

3811: \mathcal{ W}_1\cap (N^1)^+.

3812: \end{equation}

3813: An illustration for the case $n =2$ where $M \subset \C^2$ is a

3814: hypesurface is provided in the left hand side of {\sc Figure~12}. In

3815: addition, it follows from the definition of $\mathcal{ HW}_{p_1}^+$ by

3816: means of the segments $\mathcal{ Z}_{t, \chi, \nu} \left ((1-

3817: \varepsilon, 1) \right)$ that we can assume that

3818: \def\theequation{5.18}\begin{equation}

3819: T_{p_1}(N^1)^+=T_{p_1}M \oplus J(\Sigma^1)^+,

3820: \end{equation}

3821: where $(\Sigma_1 )^+$ is the hyperplane one-sided neighborhood $\{(

3822: x_1, x'): \ \sigma (x_1, x')>0\} \subset T_{p_1} M^1$. Equivalently,

3823: $T_{p_1}(N^1)^+$ is represented by the inequality

3824: $y_1+a_2y_2+\cdots+a_ny_n>0$. Consequently, there exists a $\mathcal{

3825: C}^{2,\alpha}$-smooth function $\psi(x,y')$ with

3826: $\psi(0)=\partial_{x_k}\psi(0)= \partial_{y_j}\psi(0)=0$ for

3827: $k=1,\dots,n$ and $j=2,\dots,n$ such that $N^1$ is represented by the

3828: equation $y_1+a_2y_2+\cdots+a_ny_n=\psi(x,y')$ and $(N^1)^+$ by the

3829: inequation $y_1+a_2y_2+\cdots+a_ny_n>\psi(x,y')$.

3830: 

3831: \subsection*{5.19.~Cones, filled cones, subcones and 

3832: local description of half-wedges in Case (II)} Secondly, we assume

3833: that the nonzero vector $v_1$ of Proposition~5.12 belongs to the

3834: characteristic direction $T_{p_1 }M^1 \cap T_{p_1 }^cM$. In this

3835: case, as observed in \S4.2, the half-wedge $\mathcal{ HW}_{p_1}^+$

3836: coincides with a local wedge of edge $M^1$ at $(p_1,Jv_1)$. After a

3837: real dilatation of the $z_1$-axis, we can assume that

3838: $v_1=(1,0,\dots,0)$. Choosing an open infinite strictly convex linear

3839: proper subcone ${\sf C}_2\subset \subset {\sf C}_{p_1}\subset

3840: T_{p_1}M^1=\R_x^n$ defined by $n$ strict inequalities

3841: $\ell_1(x)>0,\dots, \ell_n(x)>0$, where the $\ell_j(x)$ are linearly

3842: independent real linear forms\,--\,of course with ${\sf C}_2$

3843: containing the vector $v_1$\,--\,it follows that there exists

3844: $\delta_1>0$ such that the half-wedge $\mathcal{ HW}_{p_1}^+$ contains

3845: the following local wedge of edge $M^1$ at $p_1$:

3846: \def\theequation{5.20}\begin{equation}

3847: \mathcal{ W}_2:=\left\{

3848: p+{\sf c}_2: \ 

3849: p\in M^1, \

3850: {\sf c}_2 \in J{\sf C}_2

3851: \right\}\cap \Delta_n

3852: (p_1,\delta_1).

3853: \end{equation}

3854: We remind that it was observed in \S4.2 ({\it cf.} especially the

3855: right hand side of {\sc Figure}~5) that $\mathcal{ W}_2$ contains

3856: $(M^1)^+$ locally in a neighborhood of $p_1$. In \S5.22 below, we

3857: shall provide a more concrete representation of $\mathcal{ W}_2$

3858: in an appropriate system of coordinates.

3859: 

3860: \subsection*{5.21.~A trichotomy}

3861: Let us pursue this discussion more concretely by introducing further

3862: normalizations. Our goal will now be to construct appropriate

3863: normalized coordinate systems. Analyzing further the dichotomy 

3864: introduced in \S5.13 by taking account of the 

3865: presence of the one-codimensional 

3866: submanifold $H^1\subset M^1$,

3867: we shall distinguish three cases by dividing Case {\bf (I)} in two

3868: subcases {\bf ($\mathbf{ I_1}$)} and {\bf ($\mathbf{ I_2}$)} as

3869: follows:

3870: \begin{itemize}

3871: \item[{\bf ($\mathbf{I_1}$)}]

3872: The nonzero vector $v_1$ does not belong to the characteristic

3873: direction $T_{p_1}M^1\cap T_{p_1}^cM$ and

3874: $\dim_\R \left(T_{p_1}H^1 \cap T_{p_1}^cM \right)=0$.

3875: \item[{\bf ($\mathbf{I_2}$)}]

3876: The nonzero vector 

3877: $v_1$ does not belong to the characteristic direction 

3878: $T_{p_1}M^1\cap T_{p_1}^cM$ and

3879: $\dim_\R \left(T_{p_1}H^1 \cap T_{p_1}^cM \right)=1$ 

3880: (this possibility can only occur when $n\geq 3$).

3881: \item[{\bf (II)}]

3882: The nonzero vector $v_1$ belongs to the

3883: characteristic direction $T_{p_1}M^1\cap 

3884: T_{p_1}^cM$.

3885: \end{itemize}

3886: 

3887: In case {\bf ($\mathbf{I_1}$)}, we notice that the

3888: assumption $T_{p_1}H^1\cap T_{p_1}^cM=\{0\}$ implies that

3889: $v_1$ does not belong to the characteristic direction, 

3890: because $v_1\in T_{p_1}H^1$. Also, 

3891: in case {\bf (II)}, we notice that $\dim_\R \left(T_{ p_1} H^1 \cap

3892: T_{ p_1}^c M \right)=1$ because $v_1 \in T_{ p_1} H^1$, because

3893: $T_{p_1}H^1 \subset T_{ p_1}M^1$ and because the characteristic

3894: direction $T_{p_1}M^1\cap T_{p_1}^cM$ is one-dimensional.

3895: 

3896: In each of the above three cases, it will be convenient in

3897: Section~8 below to work with simultaneously normalized defining

3898: (in)equations for $M$, for $M^1$, for $(M^1)^+$, for $H^1$, for $(H^1

3899: )^+$, for ${\sf C }_1'$, for $v_1$, for ${\sf C }_1$, for $(N^1)^+$

3900: and for $\mathcal{ HW}_{ p_1}^+$, in a single coordinate system

3901: centered at $p_1$. In the next paragraphs, we shall set up further

3902: elementary normalization lemmas in a {\it common system of

3903: coordinates}, firstly for Case {\bf ($\mathbf{I_1}$)}, secondly for

3904: Case {\bf ($\mathbf{I_2}$)} and thirdly for Case {\bf (II)}.

3905: 

3906: First of all, in the above coordinate system $(z_1,z')$ with $T_{p_1

3907: }M=\{ y_2= \cdots= y_n =0\}$ and with $T_{p_1 }M^1=\{y_1 = y_2=

3908: \cdots= y_n=0\}$, by means of the implicit function theorem, we can

3909: represent locally $M$ by $(n-1)$ grahed equations of the form $y_2=

3910: \varphi_2 (x,y_1),\dots, y_n= \varphi_n (x,y_1)$, where

3911: the $\varphi_j$ are $\mathcal{ C}^{2, \alpha}$-smooth functions

3912: satisfying $\varphi_j(0)= \partial_{ x_k} \varphi_j(0)= \partial_{y_1}

3913: \varphi_j (0)= 0 $ for $j=2,\dots,n$, $k=1,\dots,n$ and

3914: we can represent $M^1$ by $n$ graphed equations $y_1= h_1(x), y_2=

3915: h_2(x), \dots,y_n= h_n(x)$, where the $h_j$ are $\mathcal{

3916: C}^{2,\alpha}$-smooth functions satisfying

3917: $h_j(0)=\partial_{x_k}h_j(0)=0$ for $j,k=1, \dots,n$.

3918: 

3919: \subsection*{5.22.~First order normalizations in Case 

3920: {\bf ($\mathbf{I_1}$)}} Thus, let us deal first with Case {\bf

3921: ($\mathbf{I_1}$)}. After a possible permutation of coordinates, we can

3922: assume that $T_{p_1}H^1$, which is a one-codimensional subspace of

3923: $T_{p_1}M^1$, is given by the equations

3924: \def\theequation{5.23}\begin{equation}

3925: x_1=b_2x_2+\cdots+b_nx_n, \ 

3926: y_1=0,y'=0,

3927: \end{equation}

3928: for some real numbers $b_2,\dots,b_n$. If we define the linear

3929: invertible transformation $\widehat{z}_1:=z_1-b_2z_2-\cdots-b_nz_n$,

3930: $\widehat{z}':=z'$, then the plane $T_{ p_1} H^1$ written

3931: in~\thetag{5.23} clearly transforms to the plane $\widehat{

3932: x}_1=\widehat{ y}_1= \widehat{ y}'=0$, and (fortunately) $T_{ p_1}M$

3933: and $T_{ p_1} M^1$ are left unchanged, namely $T_{p_1} \widehat{M}=\{

3934: \widehat{y}'=0\}$ and $T_{p_1}\widehat{ M^1}= \{\widehat{ y}_1=

3935: \widehat{ y}'=0\}$.

3936: 

3937: Dropping the hats on coordinates, we have $T_{p_1}M=\{y'=0\}$,

3938: $T_{p_1}M^1=\{y_1=y'=0\}$, $T_{p_1}H^1=\{x_1=y_1=y'=0\}$. Let ${\sf

3939: C}_1'\subset \subset {\sf C}_{p_1}'$ be the open infinite strictly

3940: convex linear cone introduced in \S5.14, which is contained in the

3941: real $(n-1)$-dimensional space $\{(0,x')\}$ and which is defined by

3942: $(n-1)$ strict inequalities $\ell_1'(x')>0,\dots,

3943: \ell_{n-1}'(x')>0$. By means of a real linear invertible

3944: transformation of the form $\widehat{ z}_1:= z_1$, $\widehat{ z}':=

3945: A'\cdot z'$, where $A'$ is an $(n-1)\times (n-1)$ real matrix, we can

3946: transform ${\sf C}_1'$ to a cone $\widehat{\sf C}_1'$ defined by the

3947: simpler inequalities $\widehat{ x}_2>0,\dots, \widehat{ x}_n>0$.

3948: Fortunately, this transformation stabilizes $T_{p_1}M$, $T_{p_1}M^1$

3949: and $T_{p_1}H^1$.

3950: 

3951: Dropping the hats on coordinates, we now have $T_{p_1}M=\{y'=0\}$,

3952: $T_{p_1}M^1=\{y_1=y'=0\}$, $T_{p_1}H^1=\{x_1=y_1=y'=0\}$ and ${\sf

3953: C}_1'=\{(0,x'): \ x_2>0,\dots,x_n>0\}$. Then the nonzero vector

3954: $v_1\in T_{p_1}H^1$ which belongs to ${\sf C}_1'$ has coordinates

3955: $v_1=(0,v_{2;1},\dots, v_{n;1})\in\R^n$, where

3956: $v_{2;1}>0,\dots,v_{n;1}>0$. By means of real dilatations or real

3957: contractions of the real axes $\R_{x_2},\dots, \R_{x_n}$ (a

3958: transformation which does not perturb the previously achieved

3959: normalizations), we can assume that $v_1=(0,1,\dots,1)$ and that

3960: $T_{p_1}(M^1)^+=\{y'=0, y_1>0\}$, $T_{p_1}(H^1)^+=\{y=0, \ x_1>0\}$.

3961: 

3962: Finally, the linear one-codimensional subspace $\sigma_1\subset

3963: T_{p_1}M^1$ introduced in \S5.14 which does not contain the

3964: characteristic direction $T_{p_1}M^1\cap T_{p_1}^cM \equiv \R_{x_1}$

3965: may be represented by an equation of the form

3966: $\sigma(x_1,x'):=x_1+a_2x_2+\cdots+a_nx_n=0$, for some real numbers

3967: $a_2,\dots,a_n$. By~\thetag{5.15}, the vector $v_1$ belongs to the

3968: cone ${\sf C}_1$, hence $a_2+\dots+a_n>0$. After a dilatation of the

3969: $x_1$-axis, we can even assume that $a_2+\dots+a_n=1$. We remind that

3970: by~\thetag{5.18}, the half-space $T_{p_1}(N^1)^+$ is given by

3971: $y_1+a_2y_2+\cdots+a_ny_n>0$, hence there exists a $\mathcal{

3972: C}^{2,\alpha}$-smooth function $\psi(x,y')$ with

3973: $\psi(0)=\partial_{x_k}\psi(0)= \partial_{y_j}\psi(0)=0$ for

3974: $k=1,\dots,n$ and $j=2,\dots,n$ such that $(N^1)^+$ is represented by

3975: the inequation $y_1+a_2y_2+\cdots+a_ny_n>

3976: \psi(x,y')$. Consequently, in this coordinate system, we may

3977: represent concretely the local half-wedge $\mathcal{ HW}_1^+\subset

3978: \mathcal{ HW}_{p_1}^+$ constructed in \S5.14 as

3979: \def\theequation{5.24}\begin{equation}

3980: \left\{

3981: \aligned

3982: {}

3983: &

3984: \mathcal{ HW}_1^+=

3985: \left\{

3986: (z_1,z')\in \C^n:

3987: \vert z_1 \vert < \delta_1, \

3988: \vert z' \vert < \delta_1, \right. \\

3989: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

3990: y_1+a_2y_2+\dots+a_ny_n-\psi(x,y')>0,\\

3991: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

3992: \left.

3993: y_2-\varphi_2(x,y_1) > 0, \dots, 

3994: y_n-\varphi_n(x,y_1) >0 \ 

3995: \right\}.

3996: \endaligned\right.

3997: \end{equation}

3998: For the continuation of the proof of Proposition~5.12, it will also be

3999: convenient to proceed to further second order normalizations of the

4000: totally real submanifolds $M^1$ and $H^1$. These normalizations will

4001: all be tangent to the identity tranformation, hence they will leave the

4002: previously achieved normalizations unchanged.

4003: 

4004: \subsection*{5.25.~Second order normalizations in Case 

4005: ($\mathbf{ I_1}$)}

4006: Let us then perform a second order Taylor development of the defining

4007: equations of $M^1$

4008: \def\theequation{5.26}\begin{equation}

4009: y=h(x)=\sum_{k_1,k_2=1}^n\, 

4010: a_{k_1,k_2}\, x_{k_1} x_{k_2}+o(\vert x \vert^2),

4011: \end{equation}

4012: where the $a_{k_1,k_2}=\frac{1}{2}

4013: \partial_{x_{k_1}}

4014: \partial_{x_{k_2}}h (0)$ are vectors of $\R^n$.

4015: If we define the quadratic invertible transformation

4016: \def\theequation{5.27}\begin{equation}

4017: \widehat{z}:= z-i\sum_{k_1,k_2=1}^n\, 

4018: a_{k_1,k_2}\, z_{k_1} z_{k_2}=\Phi(z), 

4019: \end{equation}

4020: which is tangent to the identity mapping at the origin, 

4021: then for $x+iy=x+ih(x)\in M^1$, 

4022: we have by replacing~\thetag{5.26} in the imaginary 

4023: part of $\widehat{z}$ given by~\thetag{5.27}

4024: \def\theequation{5.28}\begin{equation}

4025: \left\{

4026: \aligned

4027: \widehat{y}= 

4028: & \

4029: y - \sum_{k_1,k_2=1}^n\, 

4030: a_{k_1,k_2}\, x_{k_1}x_{k_2}+

4031: \sum_{k_1,k_2=1}^n\, a_{k_1,k_2}\, y_{k_1}y_{k_2} \\

4032: = 

4033: & \ 

4034: {\rm o}(\vert x \vert^2) \\

4035: = 

4036: & \

4037: {\rm o}\left(\left\vert

4038: {\rm Re}\, \Phi^{-1}(\widehat{z})

4039: \right\vert^2\right)= {\rm o}\left( \left\vert

4040: (\widehat{x}, \widehat{y})

4041: \right\vert^2 \right),

4042: \endaligned\right.

4043: \end{equation}

4044: whence by applying the $\mathcal{ C}^{2,\alpha}$ implicit function

4045: theorem to solve~\thetag{5.28} in terms of $\widehat{ y}$, we find

4046: that $\widehat{ M^1}:=\Phi( M^1)$ may be represented by an equation of

4047: the form $\widehat{ y}= \widehat{ h}(\widehat{ x})$, for some

4048: $\R^n$-valued local $\mathcal{ C}^{2,\alpha}$-smooth mapping

4049: $\widehat{ h}$ which satisfies $\widehat{ h}( \widehat{ x})= {\rm o}(

4050: \vert \widehat{ x} \vert^2)$.

4051: 

4052: Finally, dropping the hats on coordinates, we can assume that the

4053: functions $h_1, \dots, h_n$ vanish at the origin to second order.

4054: Since $T_{p_1}H^1=\{y=0, \ x_1=0\}$, there exists a $\mathcal{

4055: C}^{2,\alpha}$-smooth function $g(x')$ with $g(0)=\partial_{x_k}

4056: g(0)=0$ for $k=2,\dots,n$ such that $(H^1)^+$ is given by the equation

4057: $x_1 >g(x')$. We want to normalize also the defining equation

4058: $x_1=g(x')$ of $H^1$. Instead of requiring, similarly as for

4059: $h_1,\dots,h_n$, that $g$ vanishes to second order at the origin

4060: (which would be possible), we shall normalize $g$ in order that

4061: $g(x')=-x_1^2-\cdots - x_n^2+ {\rm o}\left(\vert x' \vert^{2+\alpha}

4062: \right)$ (which will also be possible, thanks to the total reality of

4063: $H^1$). The reason why we want $g$ to be strictly concave is a trick

4064: that will be useful in Section~8 below.

4065: 

4066: Thus, we now perform a second

4067: order Taylor development of the defining equations of $H^1$

4068: \def\theequation{5.29}\begin{equation}

4069: \left\{

4070: \aligned

4071: {}

4072: &

4073: x_1=g\left(x'\right)=\sum_{k_1,k_2=2}^n\, 

4074: b_{k_1,k_2} \, x_{k_1}x_{k_2}+ o (\left\vert

4075: x'

4076: \right\vert^2),\\

4077: &

4078: y=h\left(g(x'),x'\right)=:k\left(x'\right)=

4079: {\rm o}(\left\vert x' \right\vert^2),

4080: \endaligned\right.

4081: \end{equation}

4082: where the $b_{k_1,k_2}=\frac{1}{2}

4083: \partial_{x_{k_1}}

4084: \partial_{x_{k_2}} g(0)$

4085: are real numbers. If we define the quadratic invertible transformation

4086: \def\theequation{5.30}\begin{equation}

4087: \left\{

4088: \aligned

4089: {}

4090: &

4091: \widehat{z}_1:= z_1-\sum_{k_1,k_2=2}^n\, 

4092: b_{k_1,k_2}\, z_{k_1}z_{k_2} -z_2^2-\cdots-z_n^2, \\

4093: &

4094: \widehat{z}':= z', 

4095: \endaligned\right.

4096: \end{equation}

4097: which is tangent to the identity mapping, then 

4098: for $\left(g(x')+ik_1(x'),x'+ik'(x')\right)\in H^1$,

4099: we have by replacing~\thetag{5.29} in the 

4100: real part of $\widehat{z}_1$, given by~\thetag{5.30}:

4101: \def\theequation{5.31}\begin{equation}

4102: \left\{

4103: \aligned

4104: \widehat{x}_1=

4105: & \

4106: x_1-\sum_{k_1,k_2=2}^n\, b_{k_1,

4107: k_2}\, x_{k_1}x_{k_2}+\sum_{k_1,k_2=2}^n\, 

4108: b_{k_1,k_2}\, y_{k_1}y_{k_2}-

4109: \sum_{k=2}^n\, x_k^2+\sum_{k=2}^n\, 

4110: y_k^2, \\

4111: =

4112: & \

4113: -x_2^2-\cdots-x_n^2+{\rm o}\left(

4114: \left\vert x' \right\vert^2\right) \\

4115: =

4116: & \

4117: -\widehat{x}_2^2-\cdots-\widehat{x}_n^2-{\rm o}\left(\left\vert

4118: (\widehat{x},\widehat{y})\right\vert^2\right).

4119: \endaligned\right.

4120: \end{equation}

4121: Similarly (dropping the elementary computations), we may obtain for

4122: the imaginary part of $\widehat{z}_1$ and for the imaginary part of

4123: $\widehat{z}'$

4124: \def\theequation{5.32}\begin{equation}

4125: \widehat{y}_1={\rm o}\left(\left\vert

4126: (\widehat{x},\widehat{y})\right\vert^2\right) \ \ \ \ \ 

4127: {\rm and} \ \ \ \ \ 

4128: \widehat{y}'={\rm o}\left(\left\vert

4129: (\widehat{x},\widehat{y})\right\vert^2\right),

4130: \end{equation}

4131: whence by applying the $\mathcal{ C}^{2,\alpha}$ implicit function

4132: theorem to solve the system~\thetag{5.31}, \thetag{5.32} in terms

4133: of $\widehat{x}_1$, $\widehat{y}_1$ and $\widehat{y}'$, we find

4134: that $\widehat{ H}^1:=\Phi(H^1)$ may be represented 

4135: by equations

4136: of the form

4137: \def\theequation{5.33}\begin{equation}

4138: \left\{

4139: \aligned

4140: \widehat{x}_1=

4141: & \

4142: \widehat{g}\left(\widehat{x}'\right)=

4143: -\widehat{x}_2^2-\cdots-\widehat{x}_n^2+

4144: {\rm o}\left(

4145: \left\vert

4146: \widehat{x}'

4147: \right\vert^2

4148: \right), \\

4149: \widehat{y}=

4150: & \

4151: \widehat{k}\left(\widehat{x}'\right)=o\left(

4152: \left\vert

4153: \widehat{x}'

4154: \right\vert^2

4155: \right).

4156: \endaligned\right.

4157: \end{equation}

4158: It remains to check that the above transformation has not perturbed

4159: the previous second order normalizations of $h_1, \dots, h_n$ (this is

4160: important), which is easy: replacing $y$ by $h(x) = {\rm o} (\vert x

4161: \vert^2)$ in the imaginary parts of $\widehat{z}_1$ and of

4162: $\widehat{z}'$ defined by the transformation~\thetag{5.30}, we get

4163: firstly

4164: \def\theequation{5.34}\begin{equation}

4165: \left\{

4166: \aligned

4167: \widehat{y}_1 =

4168: & \

4169: y_1 -\sum_{k_1,k_2}^n\, 

4170: b_{k_1,k_2}\, \left(

4171: x_{k_1}y_{k_2}+y_{k_1}x_{k_2}

4172: \right)-2\sum_{k=2}^n\, 

4173: x_ky_k \\

4174: =

4175: & \

4176: {\rm o}(\vert x\vert^2) \\

4177: =

4178: & \

4179: {\rm o}\left(\left\vert

4180: {\rm Re}\, \Phi^{-1}(\widehat{z})

4181: \right\vert^2\right)= {\rm o}\left( \left\vert

4182: (\widehat{x}, \widehat{y})

4183: \right\vert^2 \right),

4184: \endaligned\right.

4185: \end{equation}

4186: and similarly 

4187: \def\theequation{5.35}\begin{equation}

4188: \widehat{y}'=o\left( \left\vert

4189: (\widehat{x}, \widehat{y})

4190: \right\vert^2 \right),

4191: \end{equation}

4192: whence by applying the $\mathcal{ C}^{2,\alpha}$ implicit function

4193: theorem to solve the system~\thetag{5.34}, \thetag{5.35} in terms of

4194: $\widehat{y}$, we find that $\widehat{ M}^1 :=\Phi(M^1)$ may be

4195: represented by equations of the form $\widehat{y}=\widehat{ h} \left(

4196: \widehat{x} \right)= {\rm o}\left(\left \vert \widehat{ x}\right \vert^2

4197: \right)$. Thus, after dropping the hats on coordinates, all the

4198: desired normalizations are satisfied. We shall now summarize these

4199: normalizations and we shall formulate just afterwards the analogous

4200: normalizations for Cases {\bf ($\mathbf{I_2}$)} and {\bf (II)}.

4201: 

4202: \subsection*{5.36.~Simultaneous normalization lemma}

4203: In the following lemma, the final choice of sufficiently small radii

4204: $\rho_1>0$ and $\delta_1>0$ is made after that all the biholomorphic

4205: changes of coordinates and all the applications of the implicit function

4206: theorem are achieved.

4207: 

4208: \def\thelemma{5.37}\begin{lemma}

4209: Let $M$, $M^1$, $p_1$, $H^1$, $(H^1)^+$, $\mathcal{ HW}_{p_1}^+$,

4210: ${\sf C}_{p_1}$ and ${\sf FC}_{p_1}$ be as in Proposition~5.12. Then

4211: there exists a sub-half-wedge $\mathcal{ HW}_1^+$ contained in

4212: $\mathcal{ HW}_{p_1}^+$ such that the following normalizations hold in

4213: each of the three cases {\bf ($\mathbf{I_1}$)}, {\bf

4214: ($\mathbf{I_2}$)} and

4215: {\bf (II)}{\rm :}

4216: 

4217: \smallskip

4218: \begin{itemize}

4219: \item[{\bf ($\mathbf{I_1}$)}]

4220: If $\dim_\R \left( T_{ p_1} H^1 \cap T_{p_1 }^cM \right) =0$, then

4221: there exists a system of holomorphic coordinates $z= (z_1, \dots,

4222: z_n)=(x_1+i y_1,\dots, x_n+iy_n)$ vanishing at $p_1$ with the vector

4223: $v_1$ equal to $(0,1,\dots,1)$, there exists positive numbers $\rho_1$

4224: and $\delta_1$ with $0< \delta_1 < \rho_1$, there exist $\mathcal{

4225: C}^{2,\alpha}$-smooth functions $\varphi_2,\dots, \varphi_n$, $h_1,

4226: \dots, h_n$, $g$, $k_1, \dots,k_n,\psi$, all defined in real cubes of

4227: edge $2 \rho_1$ and of the appropriate dimension, and there exist real

4228: numbers $a_1,\dots,a_n$ with $a_2+\cdots+a_n=1$, such that, if we

4229: denote $z':=(z_2, \dots, z_n)=x' +iy'$, then $M$, $M^1$, $(M^1)^+$,

4230: $H^1$, $(H^1)^+$ and $N^1$ are represented in the polydisc of radius

4231: $\rho_1$ centered at $p_1$ by the following graphed (in)equations and

4232: the sub-half-wedge $\mathcal{ HW}_1^+ \subset \mathcal{ HW}_{p_1}^+$ is

4233: represented in the polydisc of radius $\delta_1$ centered at $p_1$ by

4234: the following inequations

4235: \def\theequation{5.38}\begin{equation}

4236: \left\{

4237: \aligned

4238: M: \ \ \ \ \ 

4239: & \

4240: y_2=\varphi_2(x,y_1), \dots\dots, \

4241: y_n=\varphi_n(x,y_1), \\

4242: M^1: \ \ \ \ \ 

4243: & \ 

4244: y_1=h_1(x),y_2=h_2(x),\dots\dots, \ 

4245: y_n=h_n(x), \\

4246: (M^1)^+: \ \ \ \ \ 

4247: & \ 

4248: y_1>h_1(x),y_2=\varphi_2(x,y_1),\dots\dots, \ 

4249: y_n=\varphi_n(x,y_1), \\

4250: H^1: \ \ \ \ \ 

4251: & \

4252: x_1=g(x'), \

4253: y_1=k_1(x'), \dots\dots, \ 

4254: y_n=k_n(x'), \\

4255: (H^1)^+: \ \ \ \ \ 

4256: & \

4257: x_1>g(x'), \

4258: y_1=h_1(x),y_2=h_2(x), \dots\dots, \ 

4259: y_n=h_n(x), \\

4260: N^1 : \ \ \ \ \ 

4261: & \

4262: y_1+a_2y_2+\cdots+a_ny_n=\psi(x,y'), \\

4263: \mathcal{ HW}_1^+: \ \ \ \ \

4264: & \

4265: y_1+a_2y_2+\cdots+a_ny_n>\psi(x,y'), \\

4266: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ 

4267: y_2>\varphi_2(x,y_1),\dots, y_n>\varphi_n(x,y_1), 

4268: \endaligned\right.

4269: \end{equation}

4270: where we can assume that $M^1$ coincides with the intersection $M\cap

4271: \{y_1=h_1(x)\}$, that $H^1$ coincides with the intersection

4272: $M^1\cap \{x_1=g(x')\}$ and that $N^1$ contains $M^1$, which yields at

4273: the level of defining equations the following three collections of

4274: identities

4275: \def\theequation{5.39}\begin{equation}

4276: \left\{

4277: \aligned

4278: h_2(x)\equiv 

4279: & \

4280: \varphi_2(x,h_1(x)),\dots\dots, h_n(x)\equiv 

4281: \varphi_n(x,h_1(x)), \\

4282: k_1(x')\equiv

4283: & \

4284: h_1(g(x'),x'), \dots\dots,

4285: k_n(x')\equiv h_n(g(x'),x'), \\

4286: \psi(x,h'(x)) \equiv

4287: & \

4288: h_1(x)+a_2h_2(x)+\cdots+

4289: a_nh_n(x),

4290: \endaligned\right.

4291: \end{equation}

4292: and where the following normalizations hold (where $\delta_a^b$, equal

4293: to $0$ if $a\neq b$ and to $1$ if $a=b$, denotes Kronecker's

4294: symbol){\rm :}

4295: \def\theequation{5.40}\begin{equation}

4296: \left\{

4297: \aligned

4298: {}

4299: &

4300: \varphi_{j}(0)=\partial_{x_k}\varphi_j(0)

4301: =\partial_{y_1}\varphi_j(0)=0, \ \ \ \ \ j=2,\dots,n, \

4302: k=1,\dots,n,\\

4303: &

4304: h_j(0)=\partial_{x_k}

4305: h_j(0)=\partial_{x_{k_1}}

4306: \partial_{x_{k_2}}h_j(0)=0, \ \ \ \ \ \ \ \ \ \

4307: j,k,k_1,k_2=1,\dots,n, \\

4308: &

4309: g(0)=\partial_{x_k}

4310: g(0)=k_j(0)=\partial_{x_k}k_j(0)=0, \ \ \ \ \ 

4311: j=1,\dots,n, \ k=2,\dots,n, \\

4312: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

4313: \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

4314: \partial_{x_{k_1}}\partial_{x_{k_2}}

4315: g(0)=-\delta_{k_1}^{k_2}, \ \ \ \ \ \ \ 

4316: k_1,k_2=2,\dots,n, \\

4317: &

4318: \psi(0)=\partial_{x_k}\psi(0)=\partial_{y_j}\psi(0)=0, \ \ \ \ \

4319: k=1,\dots,n, \ j=2,\dots,n.

4320: \endaligned\right.

4321: \end{equation}

4322: In other words, $T_0M=\{y'=0\}$ (hence $T_0^cM$ coincides with the

4323: complex $z_1$-axis), $T_0N^1=\{ y_1+a_2y_2+\cdots+a_ny_n=0\}$ and the

4324: second order Taylor approximations of the defining equations of $M^1$,

4325: of $H^1$ and of $(H^1)^+$ are the quadrics

4326: \def\theequation{5.41}\begin{equation}

4327: \left\{

4328: \aligned

4329: T_{p_1}^{(2)}M^1: \ \ \ \ \ 

4330: & \

4331: y_1=0, \dots\dots, \ y_n=0,

4332: \\

4333: T_{p_1}^{(2)}H^1: \ \ \ \ \ 

4334: & \

4335: x_1=-x_2^2-\cdots-x_n^2, \ y_1=0,\dots\dots, \ y_n=0, \\

4336: T_{p_1}^{(2)}(H^1)^+: \ \ \ \ \ 

4337: & \

4338: x_1>-x_2^2-\cdots-x_n^2, \ y_1=0,\dots\dots, \ y_n=0.

4339: \endaligned\right.

4340: \end{equation}

4341: \item[{\bf ($\mathbf{I_2}$)}]

4342: Similarly, if $\dim_\R \left( T_{p_1}H^1\cap T_{p_1}^c M\right) = 1$

4343: and if $v_1$ is not complex tangent to $M$ {\rm (}this possibility can

4344: only occur in the case $n \geq 3${\rm )}, then there exists a system

4345: of holomorphic coordinates $z= (z_1,\dots,z_n)=

4346: (x_1+iy_1,\dots,x_n+iy_n)$ vanishing at $p_1$ with $v_1$ equal to

4347: $(1,\dots,1,0)$, there exists positive numbers $\rho_1$ and $\delta_1$

4348: with $0< \delta_1 < \rho_1$, there exist $\mathcal{ C}^{

4349: 2,\alpha}$-smoooth functions $\varphi_2,\dots,\varphi_n, h_1,\dots,

4350: h_n, g, k_1,\dots,k_n,\psi$ all defined in real cubes of edge

4351: $2\rho_1$ and of the appropriate dimension, such that if we denote

4352: $z'':=(z_1,\dots,z_{n-1})=x''+iy''$ and $z'=(z_2,\dots,z_n)=

4353: x'+iy'$, then $M$, $M^1$, $(M^1)^+$,

4354: $H^1$, $(H^1)^+$ and $N^1$ are represented in the polydisc of radius

4355: $\rho_1$ centered at $p_1$ by the following graphed (in)equations and

4356: the sub-half-wedge $\mathcal{ HW}_1^+\subset \mathcal{ HW}_{p_1}^+$ is

4357: represented in the polydisc of radius $\delta_1$ centered at $p_1$ by

4358: the following inequations

4359: \def\theequation{5.42}\begin{equation}

4360: \left\{

4361: \aligned

4362: M: \ \ \ \ \ 

4363: & \

4364: y_2=\varphi_2(x,y_1), 

4365: \dots\dots, \

4366: y_n=\varphi_n(x,y_1), \\

4367: M^1: \ \ \ \ \ 

4368: & \ 

4369: y_1=h_1(x),y_2=h_2(x),\dots\dots, \ 

4370: y_n=h_n(x), \\

4371: (M^1)^+: \ \ \ \ \ 

4372: & \ 

4373: y_1>h_1(x),y_2= \varphi_2

4374: (x,y_1), \dots\dots, \ 

4375: y_n=\varphi_n(x,y_1), \\

4376: H^1: \ \ \ \ \ 

4377: & \

4378: x_n=g(x''), \

4379: y_1=k_1(x''), \dots\dots, \ 

4380: y_n=k_n(x''), \\

4381: (H^1)^+: \ \ \ \ \ 

4382: & \

4383: x_n>g(x''), \

4384: y_1=h_1(x),y_2=h_2(x), \dots\dots, \ 

4385: y_n=h_n(x), \\

4386: N^1 : \ \ \ \ \ 

4387: & \

4388: y_2+\cdots+y_{n-1}-y_n=

4389: \psi(x,y'), \\

4390: \mathcal{ HW}_1^+: \ \ \ \ \

4391: & \

4392: y_2+\cdots+y_{n-1}-y_n>

4393: \psi(x,y'), \\

4394: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ 

4395: y_1>\varphi_1(x,y_1),\dots, 

4396: y_{n-1}>\varphi_{n-1}(x,y_1), 

4397: \endaligned\right.

4398: \end{equation}

4399: where we can assume that $M^1$ coincides with the intersection $M\cap

4400: \{y_1=h_1(x)\}$, that $H^1$ coincides with the intersection

4401: $M^1\cap \{x_1=g(x')\}$ and that $N^1$ contains $M^1$, which yields at

4402: the level of defining equations the following three collections of

4403: identities

4404: \def\theequation{5.43}\begin{equation}

4405: \left\{

4406: \aligned

4407: h_2(x)\equiv 

4408: & \

4409: \varphi_2(x,h_1(x)),\dots\dots, h_n(x)\equiv 

4410: \varphi_n(x,h_1(x)), \\

4411: k_1(x'')\equiv

4412: & \

4413: h_1(x'',g(x'')), \dots\dots,

4414: k_n(x'')\equiv h_n(x'',g(x'')), \\

4415: \psi(x,h'(x)) \equiv

4416: & \

4417: h_1(x)+h_2(x)+\cdots+h_{n-1}(x)-

4418: h_n(x),

4419: \endaligned\right.

4420: \end{equation}

4421: and where the following normalizations hold{\rm :}

4422: \def\theequation{5.44}\begin{equation}

4423: \left\{

4424: \aligned

4425: {}

4426: &

4427: \varphi_{j}(0)=\partial_{x_k}\varphi_j(0)

4428: =\partial_{y_1}\varphi_j(0)=0, \ \ \ \ \ j=2,\dots,n, \

4429: k=2,\dots,n, \\

4430: &

4431: h_j(0)=\partial_{x_k}

4432: h_j(0)=\partial_{x_{k_1}}

4433: \partial_{x_{k_2}}h_j(0)=0, \ \ \ \ \ \ \ \ \ \

4434: j,k,k_1,k_2=1,\dots,n, \\

4435: &

4436: g(0)=\partial_{x_k}

4437: g(0)=k_j(0)=\partial_{x_k}k_j(0)=0, \ \ \ \ \ 

4438: j=1,\dots,n, \ k=1,\dots,n-1, \\

4439: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

4440: \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

4441: \partial_{x_{k_1}}\partial_{x_{k_2}}

4442: g(0)=-\delta_{k_1}^{k_2}, \ \ \ \ \ \ \ 

4443: k_1,k_2=1,\dots,n-1, \\

4444: &

4445: \psi(0)=\partial_{x_k}\psi(0)=\partial_{y_j}\psi(0)=0, \ \ \ \ \

4446: k=1,\dots,n, \ j=2,\dots,n.

4447: \endaligned\right.

4448: \end{equation}

4449: In other words, $T_0M=\{y'=0\}$ {\rm (}hence $T_0^cM$ coincides with

4450: the complex $z_1$-axis{\rm )}, $T_0N^1=\{

4451: y_1+y_2+\cdots+y_{n-1}-y_n=0\}$ and the second order Taylor

4452: approximations of the defining equations of $M^1$, of $H^1$ and of

4453: $(H^1)^+$ are the quadrics

4454: \def\theequation{5.45}\begin{equation}

4455: \left\{

4456: \aligned

4457: T_{p_1}^{(2)}M^1: \ \ \ \ \ 

4458: & \

4459: y_1=0, \dots\dots, \ y_n=0,

4460: \\

4461: T_{p_1}^{(2)}H^1: \ \ \ \ \ 

4462: & \

4463: x_n=-x_1^2-\cdots-x_{n-1}^2, \ y_1=0,\dots\dots, \ y_n=0, \\

4464: T_{p_1}^{(2)}(H^1)^+: \ \ \ \ \ 

4465: & \

4466: x_n>-x_1^2-\cdots-x_{n-1}^2, \ y_1=0,\dots\dots, \ y_n=0.

4467: \endaligned\right.

4468: \end{equation}

4469: \item[{\bf (II)}]

4470: Finally, if $\dim_\R\left( T_{p_1}H^1\cap T_{p_1}^c M\right) = 1$ and

4471: if $v_1$ is complex tangent to $M$ {\rm (}this possibility can occur

4472: in all cases $n\geq 2${\rm )}, then there exists a system of holomorphic

4473: coordinates $z=(z_1,\dots,z_n)= (x_1+iy_1,\dots,x_n+iy_n)$ vanishing

4474: at $p_1$ with $v_1$ equal to $(1,0,\dots,0)$, there exist positive

4475: numbers $\rho_1$ and $\delta_1$ with $0< \delta_1 < \rho_1$, there

4476: exist $\mathcal{ C}^{ 2,\alpha}$-smoooth functions

4477: $\varphi_2,\dots,\varphi_n, h_1,\dots, h_n, g, k_1,\dots,k_n$ all

4478: defined in real cubes of edge $2\rho_1$ and of the appropriate

4479: dimension, such that if we denote $z'':=(z_1,\dots,z_{n-1})=x''+iy''$

4480: and $z'=(z_2,\dots,z_n)= x'+iy'$, then $M$, $M^1$, $(M^1)^+$, $H^1$

4481: and $(H^1)^+$ are represented in the polydisc of radius $\rho_1$

4482: centered at $p_1$ by the first five (in)equations of~\thetag{5.42}

4483: together with the normalizations~\thetag{5.45} and such that the local

4484: wedge $\mathcal{ W}_2 \subset \mathcal{ HW}_{p_1}^+$ of edge $M^1$

4485: at $p_1$ is represented in the polydisc of radius $\delta_1$ centered

4486: at $p_1$ by the following inequations

4487: \def\theequation{5.46}\begin{equation}

4488: \left\{

4489: \aligned

4490: \mathcal{ W}_2: \ \ \ \ \

4491: & \

4492: y_1-h_1(x)>-\left[y_2-h_2(x)\right], \dots\dots, \

4493: y_1-h_1(x)>-\left[y_n-h_n(x)\right], \\

4494: & \ \ \ 

4495: y_1-h_1(x)>y_2-h_2(x)+\cdots+

4496: y_n-h_n(x).

4497: \endaligned\right.

4498: \end{equation}

4499: \end{itemize}

4500: \end{lemma}

4501: 

4502: \subsection*{5.47.~Summarizing figure and proof of Lemma~5.37}

4503: As an illustration for this technical lemma, by specifying the value

4504: $n=3$, we have drawn in the following figure the cones ${\sf C}_1$ and

4505: ${\sf C}_2$ together with the vector $v_1$, the tangent plane

4506: $T_{p_1}H^1$ and the hyperplane $\Sigma^1$ in the three cases {\bf

4507: ($\mathbf{I_1}$)}, {\bf ($\mathbf{I_2}$)} and {\bf (II)}. In the left

4508: part of this figure, the cone ${\sf C}_1$ is given by $x_2>0$,

4509: $x_3>0$, $x_1>-\frac{1}{2}x_2-\frac{1}{2}x_3$, namely we have chosen

4510: the values $a_2=a_3=\frac{1}{2}$ for the drawing; in the central part,

4511: the cone ${\sf C}_1$ is given by $x_1>0$, $x_2>0$, $x_2>x_3$; in the

4512: right part, the cone ${\sf C}_2$ is given by $x_1>-x_2$, $x_1>-x_3$,

4513: $x_1>x_2+x_3$.

4514: 

4515: \bigskip

4516: \begin{center}

4517: \input three-cones.pstex_t

4518: \end{center}

4519: 

4520: \proof

4521: Case {\bf ($\mathbf{I_1}$)} has been completed before

4522: the statement of Lemma~5.37. 

4523: 

4524: For Case {\bf ($\mathbf{I_2}$)}, we reason similarly, as follows. We

4525: start with the normalizations $T_{p_1} M=\{y'=0\}$ and $T_{p_1} M^1=

4526: \{y=0\}$ as in the end of \S5.21. By assumption, $T_{p_1}H^1$ contains

4527: the characteristic direction, which coincides with the $x_1$-axis. By

4528: means of an elementary real linear transformation of the form

4529: $\widehat{ z}_1:= z_1$, $\widehat{ z}'=A'\cdot z'$, we may first

4530: normalize $T_{p_1} H^1$ to be the hyperplane (after dropping the hats

4531: on coordinates) $\{x_n=0, \ y=0\}$. 

4532: Similarly, we may normalize $v_1$ to be

4533: the vector $(1,1,\dots,1,0)$. Let again $\pi': (x_1,x') \mapsto x'$

4534: denote the canonical projection on the $x'$-space. Then

4535: $\pi'(v_1)=(1,\dots,1,0)$. Using again a real linear transformation of

4536: the form $\widehat{ z}_1:= z_1$, $\widehat{ z}'=A'\cdot z'$, we can

4537: assume that the proper subcone ${\sf C}_1'\subset \subset {\sf

4538: C}_{p_1}'\equiv \pi'({\sf C}_{p_1})$ which contains the vector $v_1$ is

4539: given (after dropping the hats on coordinates) by

4540: \def\theequation{5.48}\begin{equation}

4541: {\sf C}_1': \ \ \ \ \ 

4542: x_2>0,\dots,x_{n-1}>0, \ 

4543: x_2+\dots+x_{n-1}>x_n.

4544: \end{equation}

4545: Following \S5.14 ({\it cf.} {\sc Figure}~12), we choose a linear cone

4546: ${\sf C}_1\subset\subset {\sf FC}_{p_1}$ defined by the $(n-1)$

4547: inequations of ${\sf C}_1'$ plus one inequation of the form

4548: $x_1>a_2x_2+\cdots+a_nx_n$ with $1>a_2+\cdots+a_{n-1}$, since $v_1$

4549: belongs to ${\sf C}_1$. Then by means of a real linear transformation

4550: of the form $\widehat{ z}_1:=z_1 + a_2z_2+\cdots +a_nz_n$, $\widehat{

4551: z}':= z'$, which stabilizes $\pi'(v_1)$ and the

4552: inequations~\thetag{5.48} of ${\sf C}_1'$, we can assume that the

4553: supplementary inequation for ${\sf C}_1$, namely the inequation for

4554: $(\Sigma^1)^+$, is simply (after dropping the hats on coordinates)

4555: $x_1>0$. Then the vector $v_1$ is mapped to the vector of coordinates

4556: $(1-a_2-\cdots-a_n,1,\dots,1,0)$, which we map to the vector of

4557: coordinates $(1,1,\dots,1,0)$ by an obvious positive scaling of the

4558: $x_1$-axis. In conclusion, in the final system of coordinates,

4559: the cone ${\sf C}_1$ is given by

4560: \def\theequation{5.49}\begin{equation}

4561: {\sf C}_1: \ \ \ \ \ 

4562: x_1>0,x_2>0,\dots,x_{n-1}>0, \ 

4563: x_2+\cdots+x_{n-1}-x_n>0.

4564: \end{equation}

4565: This implies that the half-wedge $\mathcal{ HW}_1^+\subset \mathcal{

4566: HW}_{p_1}^+$ may be represented by the inequations of the last two line

4567: of~\thetag{5.42}. To conclude the proof of Case {\bf

4568: ($\mathbf{I_2}$)} of Lemma~5.37, it suffices to observe that, as in

4569: Case {\bf ($\mathbf{I_1}$)}, the further second order normalizations

4570: do not perturb the previously achieved first order normalizations,

4571: because the transformations are tangent to the identity mapping at the

4572: origin.

4573: 

4574: Finally, we treat Case {\bf (II)} of Lemma~5.37, starting with the

4575: system of coordinates $(z_1,\dots,z_n)$ of the end of \S5.21. After

4576: an elementary real linear transformation stabilizing the characteristic

4577: $x_1$-axis, we can assume that $v_1=(1,0,\dots,0)$ and that the

4578: convex infinite linear cone ${\sf C}_2$ introduced in \S5.19 which

4579: contains $v_1$ is given by the inequations

4580: \def\theequation{5.50}\begin{equation}

4581: x_1>-x_2,\dots\dots,\ x_1>-x_n,\ 

4582: x_1>x_2+\cdots+x_n.

4583: \end{equation}

4584: This implies that the local wedge $\mathcal{ W}_2\subset \mathcal{

4585: HW}_{p_1}^+$ of edge $M^1$ at $p_1$ introduced in \S5.19 may be

4586: represented by the inequations~\thetag{5.46}. Finally, the second

4587: order normalizations, which are tangent to the identity mapping, are

4588: achieved as in the two previous cases {\bf ($\mathbf{I_1}$)} and {\bf

4589: ($\mathbf{I_2}$)}.

4590: 

4591: The proof of Lemma~5.37 is complete.

4592: \endproof

4593: 

4594: \section*{\S6.~Three preparatory lemmas on H\"older spaces}

4595: 

4596: In this section, we first collect a few very elementary lemmas that

4597: will be useful in our geometric construction of half-attached analytic

4598: discs which will be achieved in Section~7 below. From now on, we

4599: shall admit the convenient index notation $g_{x_k}$ for the partial

4600: derivative which was denoted up to now by $\partial_{x_k}g$.

4601: 

4602: \subsection*{6.1.~Local growth of $\mathcal{ C}^{2,\alpha}$-smooth 

4603: mappings} Let $n\in \N$, $n \geq 1$ and let $x= (x_1, \dots, x_n) \in

4604: \R^n$. We shall use the norm $\vert x \vert := \max_{1 \leq k\leq n}\,

4605: \vert x_k \vert$. If $g= g(x)$ is an $\R^n$-valued $\mathcal{

4606: C}^1$-smooth mapping on the real cube $\{x\in \R^n: \ \vert x \vert <

4607: 2\rho_1\}$, for some $\rho_1>0$, and if $\vert x'\vert, \vert x''

4608: \vert \leq \rho$, for some $\rho<2\rho_1$, we have the trivial

4609: estimate

4610: \def\theequation{6.2}\begin{equation}

4611: \left\vert g(x')- g(x'') \right\vert \leq \left\vert 

4612: x' -x'' \right\vert \cdot \left(

4613: \sum_{k=1}^n \, \sup_{\vert x 

4614: \vert \leq \rho} \, \vert g_{x_k}(x)

4615: \vert \right),

4616: \end{equation} 

4617: where we denote by $g_{j,x_k}$ the partial derivative $\partial

4618: g_j/\partial x_k$. Notice that by the definition of the norm $\vert

4619: \cdot \vert$, we have in~\thetag{6.2} that $\vert g(x) \vert \equiv

4620: \max_{1\leq k\leq n} \, \vert g_j (x) \vert$ and that $\vert g_{x_k}

4621: (x)\vert \equiv \max_{1 \leq k \leq n}\, \vert g_{j,x_k}(x) \vert$.

4622: 

4623: Let $\alpha$ with $0 < \alpha < 1$ and let $h= h(x) = (h_1(x), \dots,

4624: h_n(x ))$ be an $\R^n$-valued mapping which is of class $\mathcal{

4625: C}^{2, \alpha}$ on the real cube $\{x\in \R^n: \ \vert x \vert <

4626: 2\rho_1\}$, for some $\rho_1 >0$. For every $\rho< 2 \rho_1$, we

4627: consider the $\mathcal{ C}^{2, \alpha}$ norm of $h$ over $\{ \vert x

4628: \vert \leq \rho\}$ which is defined precisely as:

4629: \def\theequation{6.3}\begin{equation}

4630: \left\{

4631: \aligned

4632: \vert \vert h \vert \vert_{\mathcal{ C}^{2,\alpha}(

4633: \{\vert x \vert \leq \rho\})}:= 

4634: & \

4635: \sup_{\vert x \vert \leq \rho}\, 

4636: \vert h(x)\vert+

4637: \sum_{k=1}^n\, 

4638: \sup_{\vert x \vert \leq \rho}\, 

4639: \left\vert

4640: h_{x_k}(x)\right\vert+

4641: \sum_{k_1,k_2=1}^n\, 

4642: \left\vert h_{x_{k_1}x_{k_2}}(x)\right\vert+\\

4643: & \

4644: +\sum_{k_1,k_2=1}^n\, 

4645: \sup_{\left\vert x'\right\vert, \,

4646: \left\vert x''\right\vert \leq \rho, \ 

4647: x'\neq x''}\ 

4648: \frac{

4649: \left\vert h_{x_{k_1}x_{k_2}}(x')-

4650: h_{x_{k_1}x_{k_2}}(x'')

4651: \right\vert

4652: }

4653: {\left\vert x' - x'' \right\vert^\alpha}<\infty,

4654: \endaligned\right.

4655: \end{equation}

4656: and which is finite. With these definitions at hand, the

4657: following lemma can easily be established by means of~\thetag{6.2}.

4658: 

4659: \def\thelemma{{\bf 6.4}}\begin{lemma}

4660: Under the above assumptions, let 

4661: \def\theequation{6.5}\begin{equation}

4662: K_1:= \vert \vert h \vert \vert_{

4663: \mathcal{ C}^{2, \alpha}(\{ \vert x \vert \leq \rho_1\})}<\infty

4664: \end{equation}

4665: be the $\mathcal{ C}^{2, \alpha}$ norm of $h$ over the cube $\{\vert x

4666: \vert \leq \rho_1\}$ and assume that $h_j (0)=0$, $h_{j, x_k} (0)=0$

4667: and $h_{j, x_{k_1}x_{k_2}}(0)=0$, for all $j,k, k_1,

4668: k_2=1,\dots,n$. Then the following three inequalities hold for $\vert

4669: x \vert \leq \rho_1${\rm :}

4670: \def\theequation{6.6}\begin{equation}

4671: \left\{

4672: \aligned

4673: {\bf [1]:}

4674: & \ \ \ \ \ 

4675: \vert h(x)\vert 

4676: \leq 

4677: \vert x\vert^{2+\alpha} \cdot K_1,\\

4678: {\bf [2]:} 

4679: & \ \ \ \ \ 

4680: \sum_{k=1}^n\, \left\vert

4681: h_{x_k}(x) \right\vert 

4682: \leq 

4683: \vert x \vert^{1+\alpha} \cdot K_1, \\

4684: {\bf [3]:} 

4685: & \ \ \ \ \ \

4686: \sum_{k_1,k_2=1}^n \, 

4687: \left\vert

4688: h_{x_{k_1}x_{k_2}}(x)

4689: \right\vert 

4690: \leq 

4691: \vert x \vert^\alpha \cdot K_1.

4692: \endaligned\right.

4693: \end{equation} 

4694: \end{lemma}

4695: 

4696: \subsection*{6.7.~A $\mathcal{ C}^{1,\alpha}$ estimate for

4697: composition of mappings} Recall that $\Delta$ is the open unit disc in

4698: $\C$ and that $\partial \Delta$ is its boundary, namely the unit

4699: circle. We shall constantly denote the complex variable in $\overline{

4700: \Delta}:= \Delta \cup \partial \Delta$ by $\zeta=\rho \, e^{i

4701: \theta}$, where $0 \leq \rho \leq 1$ and where $\vert \theta \vert

4702: \leq \pi$, except when we consider two points $\zeta'=e^{i\theta'}$,

4703: $\zeta''=e^{i\theta''}$, in which case we may obviously choose $\vert

4704: \theta ' \vert, \ \vert \theta'' \vert \leq 2\pi$ with $0\leq \vert

4705: \theta ' -\theta'' \vert \leq \pi$. Let now $X(\zeta)=\left(X_1

4706: (\zeta), \dots, X_n (\zeta)\right)$ be an $\R^n$-valued mapping which

4707: is of class $\mathcal{ C}^{1,\alpha}$ on the unit circle $\partial

4708: \Delta$. We define its $\mathcal{ C}^{1, \alpha}$-norm precisely by

4709: \def\theequation{6.8}\begin{equation}

4710: \left\{

4711: \aligned

4712: \vert \vert X \vert \vert_{\mathcal{ C}^{1,\alpha}(\partial 

4713: \Delta)}:= 

4714: & \

4715: \sup_{\vert \theta \vert \leq \pi}\, 

4716: \left\vert X\left(e^{i\theta} \right) \right\vert+ 

4717: \sup_{\vert \theta \vert \leq \pi}\, 

4718: \left\vert

4719: \frac{d X\left(e^{i\theta}\right)}{d \theta}

4720: \right\vert+ \\

4721: & \ \ \ \ \

4722: +\sup_{0 < \left\vert \theta' - \theta''\right\vert \leq \pi} \ 

4723: \frac{

4724: \left\vert\frac{dX(e^{i\theta'})}{d\theta}-

4725: \frac{dX(e^{i\theta''})}{d\theta}

4726: \right\vert

4727: }

4728: {\left\vert \theta' - \theta''\right\vert^\alpha},

4729: \endaligned\right.

4730: \end{equation}

4731: and we define its $\mathcal{ C}^1$-norm $\vert \vert X \vert

4732: \vert_{\mathcal{ C}^1(\partial \Delta)}$ by keeping only the first two

4733: terms. Let $h$ be as in Lemma~6.5.

4734: 

4735: \def\thelemma{{\bf 6.9}}\begin{lemma} Under the above assumptions, if

4736: moreover $\left\vert X \left(e^{i\theta}\right) \right\vert \leq\rho$

4737: for all $\theta$ with $\vert \theta \vert \leq \pi$, where $\rho \leq

4738: \rho_1$, then we have the following three estimates{\rm :}

4739: \def\theequation{6.10}\begin{equation}

4740: \small

4741: \left\{

4742: \aligned

4743: \vert \vert h(X) \vert \vert_{\mathcal{ 

4744: C}^{1,\alpha}(\partial \Delta)}

4745: \leq 

4746: & \

4747: \sup_{\vert x \vert \leq \rho} \, 

4748: \vert h(x) \vert +

4749: \left(

4750: \sum_{k=1}^n\, 

4751: \sup_{\vert x \vert \leq \rho}\, 

4752: \left\vert h_{x_k}(x) \right\vert

4753: \right)\cdot \vert \vert 

4754: X \vert \vert_{\mathcal{ C}^1(\partial \Delta)}+\\

4755: & \

4756: +\left(

4757: \sum_{k_1,k_2=1}^n\, \sup_{\vert x\vert\leq \rho}

4758: \left\vert h_{x_{k_1}x_{k_2}} (x) \right\vert

4759: \right)\cdot \pi^{1-\alpha} \cdot 

4760: \left[

4761: \vert \vert X \vert \vert_{\mathcal{ C}^1(\partial \Delta)}

4762: \right]^2+ \\

4763: & \

4764: +

4765: \left(

4766: \sum_{k=1}^n\, 

4767: \sup_{\vert x \vert\leq \rho} \, 

4768: \vert h_{x_k} (x) \vert

4769: \right)\cdot \vert \vert X 

4770: \vert \vert_{\mathcal{ C}^{1,\alpha}

4771: (\partial \Delta)}. \\

4772: \sum_{k=1}^n \, 

4773: \left\vert \left\vert h_{x_k}(X) \right\vert

4774: \right\vert_{\mathcal{ C}^\alpha

4775: (\partial \Delta)}\leq 

4776: & \

4777: \sum_{k=1}\, \sup_{\vert x \vert \leq \rho} \, 

4778: \left\vert h_{x_k} (x) \right\vert + \\

4779: & \ \

4780: +

4781: \left(

4782: \sum_{k_1,k_2=1}^n\, 

4783: \sup_{\vert x \vert \leq \rho} \, 

4784: \left\vert h_{x_{k_1}x_{k_2}}(x) \right\vert

4785: \right)\cdot \pi^{1-\alpha} \cdot

4786: \vert \vert X \vert \vert_{\mathcal{ C}^1(\partial 

4787: \Delta)},

4788: \\

4789: \sum_{k_1,k_2=1}^n \, 

4790: \left\vert 

4791: \left\vert

4792: h_{x_{k_1}x_{k_2}}(X)

4793: \right\vert

4794: \right\vert_{\mathcal{ C}^\alpha(\partial \Delta)} \leq

4795: & \

4796: \sum_{k_1,k_2=1}^n\, 

4797: \sup_{\vert x \vert \leq \rho} \,

4798: \left\vert

4799: h_{x_{k_1}x_{k_2}}(x)

4800: \right\vert + \\

4801: & \ \

4802: +\vert\vert

4803: h \vert \vert_{\mathcal{ C}^{2,\alpha}(

4804: \{\vert x \vert \leq \rho\})}\cdot

4805: \left(

4806: \vert \vert X \vert \vert_{\mathcal{C}^1(\partial \Delta)}

4807: \right)^\alpha.

4808: \endaligned\right.

4809: \end{equation}

4810: \end{lemma}

4811: 

4812: \proof

4813: We summarize the computations. Applying the definition~\thetag{6.8},

4814: using the chain rule for the calculation of

4815: $dh\left(X\left(e^{i\theta}\right)\right)/ d\theta$, and using the

4816: trivial inequality $\left\vert a'b'-a''b''\right\vert \leq \left\vert

4817: a' \right\vert \cdot \left\vert b'-b''\right\vert+ \left\vert b''

4818: \right\vert\cdot \left\vert a' -a'' \right\vert$, we may majorize

4819: \def\theequation{6.11}\begin{equation}

4820: \small

4821: \left\{

4822: \aligned

4823: \vert \vert h(X) \vert \vert_{\mathcal{ 

4824: C}^{1,\alpha}(\partial \Delta)}

4825: \leq 

4826: & \

4827: \sup_{\vert \theta \vert \leq \pi}\, 

4828: \left\vert h(X(e^{i\theta})) \right\vert + 

4829: \left(\sum_{k=1}^n\, 

4830: \sup_{\vert \theta \vert\leq \pi}\, 

4831: \left\vert h_{x_k}(X(e^{i\theta})) \right\vert

4832: \right)\cdot \max_{1\leq k\leq n}\, 

4833: \sup_{\vert \theta \vert \leq \pi}\, 

4834: \left\vert

4835: \frac{dX_k(e^{i\theta})}{d\theta}

4836: \right\vert+ \\

4837: & \

4838: \sup_{0 < \vert \theta' -\theta'' 

4839: \vert\leq \pi} \, \sum_{k=1}^n\, 

4840: \frac{

4841: \left\vert

4842: h_{x_k}(X(e^{i\theta'}))-

4843: h_{x_k}(X(e^{i\theta''}))

4844: \right\vert

4845: }{\vert \theta' -\theta ''\vert^\alpha} \cdot

4846: \max_{1\leq k\leq n}\, 

4847: \sup_{\vert \theta ' \vert \leq \pi}\, 

4848: \left\vert

4849: \frac{dX_k(e^{i\theta'})}{d\theta}

4850: \right\vert + \\

4851: & \

4852: \left(

4853: \sum_{k=1}^n\, \sup_{\vert \theta''\vert\leq \pi}\, 

4854: \left\vert h_{x_k}(e^{i\theta''}) \right\vert

4855: \right) \cdot \left(

4856: \max_{1\leq k\leq n}\, 

4857: \sup_{0< \vert \theta' -\theta''\vert\leq \pi}\,

4858: \frac{

4859: \left\vert

4860: \frac{dX_k(e^{i\theta'})}{d\theta}-

4861: \frac{dX_k(e^{i\theta''})}{d\theta}

4862: \right\vert

4863: }

4864: {\vert \theta' -\theta '' \vert^\alpha}\right),

4865: \endaligned\right.

4866: \end{equation}

4867: which yields the first inequality of~\thetag{ 6.10} after

4868: using~\thetag{ 6.2} for the second line of~\thetag{ 6.11} and the

4869: trivial majoration $\left\vert \theta' - \theta''

4870: \right\vert^{1-\alpha}\leq \pi^{1-\alpha}$. The second and the third

4871: inequalities of~\thetag{6.10} are established similarly, which

4872: completes the proof.

4873: \endproof

4874: 

4875: The following direct consequence will be strongly used in 

4876: Section~7 below.

4877: 

4878: \def\thelemma{{\bf 6.12}}\begin{lemma} Under the above assumptions,

4879: suppose that there exist constants $c_1>0$, $K_2>0$ with $c_1 K_2 \leq

4880: \rho_1$ such that for each $c\in\R$ with $0\leq c \leq c_1$, there

4881: exists $X_c\in\mathcal{ C}^{1,\alpha}(\partial \Delta,\R^n)$ with

4882: $\vert \vert X_c \vert \vert_{\mathcal{ C}^{1,\alpha}(\partial

4883: \Delta)}\leq c \cdot K_2$. Then there exists a constant $K_3>0$ such

4884: that the following three estimates hold{\rm :}

4885: \def\theequation{6.13}\begin{equation}

4886: \left\{

4887: \aligned

4888: \vert \vert h(X_c)\vert 

4889: \vert_{{\mathcal C}^{1,\alpha}(\partial 

4890: \Delta)} \leq 

4891: & \

4892: c^{2+\alpha} \cdot K_3, \\

4893: \sum_{k=1}^n\, 

4894: \vert \vert h_{x_k}(X_c) 

4895: \vert \vert_{\mathcal{ C}^\alpha(

4896: \partial \Delta)} \leq 

4897: & \

4898: c^{1+\alpha} \cdot K_3, \\

4899: \sum_{k_1,k_2=1}^n\, 

4900: \left\vert \left\vert

4901: h_{x_{k_1}x_{k_2}} (X_c)

4902: \right\vert \right\vert_{\mathcal{ 

4903: C}^\alpha(\partial \Delta)} \leq

4904: & \

4905: c^\alpha \cdot K_3.

4906: \endaligned\right.

4907: \end{equation}

4908: \end{lemma}

4909: 

4910: \proof

4911: Applying Lemmas~6.4 and~6.9, we see that it suffices to choose 

4912: \def\theequation{6.14}\begin{equation}

4913: K_3:=\max \left(

4914: K_1K_2^{2+\alpha}(3+\pi^{1-\alpha}), \ 

4915: K_1K_2^{1+\alpha}(1+\pi^{1-\alpha}), \ 

4916: 2K_1K_2^\alpha

4917: \right),

4918: \end{equation}

4919: which completes the proof.

4920: \endproof

4921: 

4922: Up to now, we have introduced three positive constants $K_1$, $K_2$,

4923: $K_3$. In Sections~7, 8 and~9 below, we shall introduce further

4924: positive constants $K_4$, $K_5$, $K_6$, $K_7$, $K_8$, $K_9$, $K_{10}$,

4925: $K_{11}$, $K_{12}$, $K_{13}$, $K_{14}$, $K_{15}$, $K_{16}$, $K_{17}$,

4926: $K_{18}$ and $K_{19}$, whose precise value will not be important.

4927: 

4928: \section*{\S7.~Families of analytic discs half-attached 

4929: to maximally real submanifolds}

4930: 

4931: \subsection*{7.1.~Preliminary}

4932: Let $E\subset \C^n$ be an arbitrary subset and let $A:

4933: \overline{\Delta}\to \C^n$ be a continuous mapping, holomorphic in

4934: $\Delta$. If $\partial^+\Delta:= \{\zeta\in\partial \Delta: {\rm Re}\,

4935: \zeta \geq 0\}$ denotes the {\sl positive half-boundary} of $\Delta$,

4936: we say that $A$ is {\sl half-attached} to $E$ if $A(\partial^+

4937: \Delta)\subset E$. Such analytic discs which are glued in part to a

4938: geometric object were studied by S.~Pinchuk in~\cite{p} to establish a

4939: boundary uniqueness principle about continuous functions on a

4940: maximally real submanifold of $\C^n$ which extend holomorphically to a

4941: wedge. Further works on the CR edge of the wedge theorem using discs

4942: partly attached to generic submanifolds were achieved by

4943: R.~Ayrapetian~\cite{a} and by A.~Tumanov~\cite{tu2}.

4944: 

4945: In this section, we shall construct local families of analytic discs

4946: $Z_{c,x,v}^1(\zeta): \overline{\Delta} \to \C^n$, where $c\in \R^+$ is

4947: small, where $x\in\R^n$ is small and where $v\in\R^n$ is small, which

4948: are half-attached to a $\mathcal{ C}^{2,\alpha}$-smooth maximally real

4949: submanifold $M^1$ of $\C^n$, which satisfy $Z_{c,0,v}^1(1)\equiv

4950: p_1\in M^1$, such that the boundary point $Z_{c,x,v}^1(1)$ covers a

4951: neighborhood of $p_1$ in $M^1$ as $x$ varies ($c$ and $v$ being fixed)

4952: and such that the tangent vector $\frac{\partial Z_{c,0,v}^1}{\partial

4953: \theta}(1)$ at the fixed point $p_1$ covers a cone in

4954: $T_{p_1}M^1$. These families will be used in Sections~8 and~9 below

4955: for the final steps in the proof of the main Proposition~5.12. With

4956: this choice, when $x$ varies, $v$ varies and $\zeta$ varies (but $c$

4957: is fixed), the set of points $Z_{c,x,v}^1(\zeta)$, covers a thin wedge

4958: of edge $M^1$ at $p_1$. Similar families of analytic discs were

4959: constructed in~\cite{ber} to reprove S.~Pinchuk's boundary uniqueness

4960: theorem, with $M^1$ of class $\mathcal{ C}^\infty$, using a method

4961: (implicit function theorem in Banach spaces) which in the case where

4962: $M$ is of class $\mathcal{ C}^{\kappa,\alpha}$ necessarily induces a

4963: loss of smoothness, yielding families of analytic discs which are only

4964: of class $\mathcal{ C}^{\kappa-1,\alpha}$. Since we want our families

4965: to be of class at least $\mathcal{ C}^2$ and since $M$ is only of

4966: class $\mathcal{ C}^{2,\alpha}$, we shall have to proceed differently.

4967: 

4968: To summarize symbolically the structure of the desired family:

4969: \def\theequation{7.2}\begin{equation}

4970: Z_{c,x,v}^1(\zeta): \left\{

4971: \aligned

4972: c= 

4973: & \

4974: \text{\rm small scaling factor}, \\

4975: x=

4976: & \

4977: \text{\rm translation parameter}, \\

4978: v=

4979: & \

4980: \text{\rm rotation parameter}, \\

4981: \zeta=

4982: & \

4983: \text{\rm unit disc variable}. \\

4984: \endaligned

4985: \right.

4986: \end{equation} 

4987: We shall begin our constructions in the ``flat'' case where the

4988: maximally real submanifold $M^1$ coincides with $\R^n$ and then

4989: perform a pertubation argument, using the scaling parameter $c$ in an

4990: essential way.

4991: 

4992: \subsection*{7.3.~A family of analytic

4993: discs sweeping $\R^n \subset \C^n$ with prescribed first order jets}

4994: We denote the coordinates over $\C^n$ by $z= x+iy= (x_1 + iy_1, \dots,

4995: x_n+ iy_n)$. Let $c\in \R$ with $c \geq 0$ be a ``scaling factor'',

4996: let $n \geq 2$, let $x= (x_1, \dots, x_n) \in \R^n$, let $v= (v_1,

4997: \dots, v_n) \in \R^n$ and consider the algebraically parametrized

4998: family of analytic discs defined by

4999: \def\theequation{7.4}\begin{equation} 

5000: B_{c,x,v}(s+it):=

5001: \left(x_1+cv_1(s+it),\dots, 

5002: x_n+cv_n(s+it)\right),

5003: \end{equation}

5004: where $s+it \in \C$ is the holomorphic variable. For $c \neq 0$, the

5005: map $B_{c, x, v}$ embeds the complex line $\C$ into $\C^n$ and sends

5006: $\R$ into $\R^n$ with {\it arbitrary first order jet at $0$}: center

5007: point $B_{c, x, v}(0) =x$ and tangent direction $\partial B_{c, x, v}

5008: (s) /\partial s \vert_{ s=0}= cv$.

5009: 

5010: To localize our family of analytic discs, we restrict the

5011: map~\thetag{7.4} to the following specific set of values: $0\leq c\leq

5012: c_0$ for some $c_0>0$; $\vert x \vert \leq c$~; $\vert v \vert \leq

5013: 2$~; and $\vert s+it \vert \leq 4$. To localize $\R^n$, we shall

5014: denote $M^0:= \{x\in\R^n: \,\vert x \vert \leq \rho_0\}$, where

5015: $\rho_0 >0$, and we notice that $B_{c,x,v}(\{\vert s+it \vert \leq

5016: 4\}) \subset M^0$ for all $c$, all $x$ and all $v$ provided that $c_0

5017: \leq \rho_0/9$.

5018: 

5019: We then consider the mapping $(s+it) \longmapsto B_{c,x, v} (s+it)$ as

5020: a local (nonsmooth) analytic disc defined on the rectangle

5021: $\{s+it\in\C: \, \vert s \vert \leq 4, 0\leq t\leq 4\}$ whose bottom

5022: boundary part $B_{c,x, v}([-4,4])$ is a small real segment contained

5023: in $\R^n$.

5024: 

5025: \subsection*{7.5.~A useful conformal equivalence}

5026: Next, we have to get rid of the corners of the rectangle $\{s+it\in

5027: \C: \, \vert s \vert \leq 4, 0\leq t \leq 4\}$. We proceed as

5028: follows. In the complex plane equipped with coordinates $s+it$, let

5029: $\mathcal{D}(i\sqrt{3},2)$ be the open disc of center $i\sqrt{3}$ and

5030: of radius $2$. Let $\mu: (-2,2)\to [0,1]$ be an {\it even}\,

5031: $\mathcal{ C}^\infty$-smooth function satisfying $\mu(s)=0$ for $0\leq

5032: s\leq 1$; $\mu(s)>0$ and $d\mu(s)/ds>0$ for $1< s < 2$; and $\mu(s)=

5033: \sqrt{3}-\sqrt{4-s^2}$ for $\sqrt{3} \leq s < 2$. The simply connected

5034: domain $C^+\subset \{t>0\}$ which is represented in {\sc Figure~1}

5035: just below may be formally defined as

5036: \def\theequation{7.6}\begin{equation}

5037: \left\{

5038: \aligned

5039: C^+\cap \{t\geq \sqrt{3}-1\}

5040: := \mathcal{D}(i\sqrt{3},2)\cap \{t\geq

5041: \sqrt{3}-1\}, \\

5042: C^+\cap \{0<t< \sqrt{3}-1\}

5043: := \{s+it\in\C: \, t>

5044: \mu(s)\}.

5045: \endaligned\right.

5046: \end{equation}

5047: 

5048: \bigskip

5049: \begin{center}

5050: \input conformal.pstex_t

5051: \end{center}

5052: 

5053: \noindent

5054: Let $\Psi: \Delta\to C^+$ be a conformal equivalence (Riemann's

5055: theorem). Since the boundary $\partial C^+$ is $\mathcal{

5056: C}^\infty$-smooth, the mapping $\Psi$ extends as a $\mathcal{

5057: C}^\infty$-smooth diffeomorphism $\partial \Delta \to \partial C^+$.

5058: Remind that $\partial^+ \Delta:=\{ \zeta \in \C: \vert \zeta \vert =1,

5059: \ {\rm Re}\, \zeta \geq 0\}$ is the positive half-boundary of

5060: $\Delta$. Then after a reparametrization of $\Delta$, we can (and we

5061: shall) assume that $\Psi(\partial^+\Delta)=[-1,1]$, $\Psi (1) =0$ and

5062: $\Psi(\pm i)=\pm 1$. It follows that $d\Psi \left(e^{i \theta}\right)

5063: /d \theta$ is a positive real number for all

5064: $e^{i\theta}\in\partial^+\Delta$. Although the precise shape of $C^+$

5065: and the specific expression of $\Psi$ will not be crucial in the

5066: sequel, it will be convenient to fix them once for all.

5067: 

5068: \subsection*{7.7.~Flat families of half-attached analytic discs}

5069: Thanks to $\Psi$, we can define a family of small analytic

5070: discs which are half-attached to the ``{\sl flat}'' maximally real

5071: manifold $M^0\equiv \{x\in \R^n: \ \vert x \vert \leq \rho_0

5072: \}$ as follows

5073: \def\theequation{7.8}\begin{equation}

5074: Z_{c,x,v}^0(\zeta):= 

5075: B_{c,x,v} \left (\Psi (\zeta) \right)=

5076: \left(

5077: x+cv\Psi(\zeta)

5078: \right).

5079: \end{equation}

5080: We then have $Z_{c, x, v}^0( \partial^+ \Delta) \subset M^0$ and $Z_{

5081: c, x, v }^0 (1) = x$. Notice that every disc $Z_{c,x,v}

5082: \left(\overline{\Delta}\right)$ 

5083: is contained in a single complex line. Starting

5084: with a maximally real submanifold of $\C^n$, as in Proposition~5.12, but

5085: dealing with the ``flat'' maximally real submanifold $M^0 \equiv

5086: \R^n$, we first construct a ``flat model'' of the desired family of

5087: analytic disc.

5088: 

5089: \def\thelemma{{\bf 7.9}}\begin{lemma} 

5090: Let $M^0=\{x\in \R^n: \ \vert x \vert \leq \rho_0\}$ be the ``flat''

5091: local maximally real submanifold defined above, let $p_0 \equiv 0 \in

5092: M^0$ denote the origin and let $v_0\in T_{p_0}M^0$ be a tangent vector

5093: with $\vert v^0\vert =1$. Then there exists a constant $\Lambda_0>0$

5094: and there exists a $\mathcal{ C }^\infty$-smooth family $A_{c,x,v}^0

5095: (\zeta)$ of analytic discs defined for $c\in \R$ with $0 \leq c \leq

5096: c_0$ for some $c_0>0$ satisfying $c_0\leq \rho_0/9$, for $x\in \R^n$

5097: with $\vert x \vert \leq c$ and for $v\in \R^n$ with $\vert v \vert

5098: \leq c$ which enjoy the following six properties{\rm :}

5099: 

5100: \smallskip

5101: \begin{itemize}

5102: \item[{\bf ($\mathbf{1_0}$)}] \

5103: $A_{c,0,v}^0(1)=p_0=0$ for all $c$ and all $v$.

5104: \item[{\bf ($\mathbf{2_0}$)}] \

5105: $A_{c,x,v}^0: \overline{\Delta}\to \C^n$ is

5106: an embedding and

5107: $\left\vert A_{c,x,v}^0(\zeta) \right \vert \leq c\cdot \Lambda_0$ 

5108: for all $c$, all $x$, all $v$ and all $\zeta$.

5109: \item[{\bf ($\mathbf{3_0}$)}] \

5110: $A_{c,x,v}^0(\partial^+\Delta)

5111: \subset M^0$ for all $c$, all $x$ and all $v$.

5112: \item[{\bf ($\mathbf{4_0}$)}] \

5113: $\frac{\partial A_{c,0,0}^0}{\partial 

5114: \theta}(1)$ is a positive multiple of 

5115: $v_0$ for all $c\neq 0$.

5116: \item[{\bf ($\mathbf{5_0}$)}] \

5117: For all $c$, all $v$ and all $e^{i\theta}\in \partial^+\Delta$, the

5118: mapping $x\longmapsto A_{c,x,v}^0\left(e^{i\theta}\right) \in M^0$ is

5119: of rank $n$.

5120: \item[{\bf ($\mathbf{6_0}$)}] \

5121: For all $e^{i \theta} \in \partial^+ \Delta$, all $c \neq 0$ and all

5122: $x$, the mapping $v \longmapsto \frac{ \partial A_{c, x, v}^0}{

5123: \partial \theta }\left(e^{ i\theta}\right)$ is of rank $n$ at $v=0$.

5124: Consequently, the positive half-lines $\R^+ \cdot \frac{ \partial

5125: A_{c,0,v}}{ \partial \theta} (1)$ describe an open infinite cone

5126: containing $v_0$ with

5127: vertex $p_0$ in $T_{p_0} M^0$ when $v$ varies.

5128: \end{itemize}

5129: \end{lemma}

5130: 

5131: \proof

5132: Proceeding as in the proof of Lemma~5.37, we can find a new affine

5133: coordinate system centered at $p_0$ and stabilizing $\R^n$, which we

5134: shall still denote by $(z_1,\dots,z_n)$, 

5135: in which the vector $v_0$

5136: has coordinates $(0,\dots,0,1)$. In this coordinate system, we then

5137: construct the family $Z_{c,x,v}^0(\zeta)$ as in~\thetag{7.8} above and

5138: we define the desired family simply as follows:

5139: \def\theequation{7.10}\begin{equation}

5140: A_{c,x,v}^0(\zeta):= Z_{c,x,v_0+v}^0(\zeta),

5141: \end{equation}

5142: where we restrict the variations of the parameter $v$ to $\vert v

5143: \vert \leq c$. Notice that every disc $A_{c,x,v}^0

5144: \left(\overline{\Delta}\right)$ is contained in a single complex line.

5145: All the properties are then elementary consequences of the explicit

5146: expression~\thetag{7.8} of $Z_{c,x,v}^0(\zeta)$.

5147: 

5148: Finally, we notice that it follows from properties {\bf

5149: ($\mathbf{5_0}$)} and {\bf ($\mathbf{6_0}$)} that the set of points

5150: $A_{c,x,v}^0(\zeta)$, where $c>0$ is fixed, where $x$ varies, where $v$

5151: varies and where $\zeta$ varies covers a local wedge of edge $M^0$ at

5152: $p_0$. The proof of Lemma~7.9 is complete.

5153: \endproof

5154: 

5155: \subsection*{7.11.~Curved families of half-attached analytic discs}

5156: Our main goal in this section is to obtain a statement similar to

5157: Lemma~7.9 after replacing the ``flat'' maximally real submanifold

5158: $M^0\cong \R^n$ by a ``curved'' $\mathcal{ C}^{2, \alpha}$-smooth

5159: maximally real submanifold $M^1$. We set up a formulation which will

5160: be appropriate for the achievement of the end of the proof of

5161: Proposition~5.12 in the next Sections~8 and~9. In particular, we shall

5162: have to shrink the family of half-disc $Z_{c, x,v}^1(\zeta)$ which we

5163: will construct as a perturbation of the family $Z_{c, x,v}^0(\zeta)$

5164: in \S7.50 below, and we shall construct discs of size $\leq c^2 \cdot

5165: \Lambda_1$ for some constant $\Lambda_1>0$, instead of requiring that

5166: their size is $\leq c \cdot \Lambda_1$, which would be the property

5167: analogous to {\bf ($\mathbf{2_0}$)}. Also, we shall loose the

5168: $\mathcal{ C}^{ 2, \alpha-0}$-smoothness with respect to the scaling

5169: parameter $c$.

5170: 

5171: \def\thelemma{7.12}\begin{lemma}

5172: Let $M^1$ be $\mathcal{ C}^{2,\alpha}$-smooth maximally real

5173: submanifold of $\C^n$, let $p_1\in M^1$ and let $v_1\in T_{p_1} M^1$

5174: be a tangent vector with $\vert v_1 \vert =1$. Then there exists

5175: a positive constant $\Lambda_1>0$ and there exists $c_1\in \R$ with

5176: $c_1>0$ such that for every $c\in \R$ with $0 < c \leq c_1$, there

5177: exists a family $A_{x,v:c}^1 (\zeta)$ of analytic discs defined for

5178: $x\in \R^n$ with $\vert x \vert \leq c^2$ and for $v\in \R^n$ with

5179: $\vert v \vert \leq c$ which is $\mathcal{ C }^{2,\alpha-0}$-smooth

5180: with respect to $(x,v,\zeta)$ and which enjoys the following six

5181: properties{\rm :}

5182: 

5183: \smallskip

5184: \begin{itemize}

5185: \item[{\bf ($\mathbf{1_1}$)}] \

5186: $A_{0,v:c}^1(1)=p_1$ for all $v$.

5187: \item[{\bf ($\mathbf{2_1}$)}] \

5188: $A_{x,v:c}^1: \overline{\Delta}\to \C^n$ is

5189: an embedding and

5190: $\left\vert A_{x,v:c}^1 (\zeta) \right \vert \leq c^2\cdot \Lambda_1$ 

5191: for all $x$, all $v$ and all $\zeta$.

5192: \item[{\bf ($\mathbf{3_1}$)}] \

5193: $A_{x,v:c}^1(\partial^+\Delta)

5194: \subset M^1$ for all $x$ and all $v$.

5195: \item[{\bf ($\mathbf{4_1}$)}] \

5196: $\frac{\partial A_{0,0:c}^1}{\partial 

5197: \theta}(1)$ is a positive multiple of 

5198: $v_1$ for all $c\neq 0$.

5199: \item[{\bf ($\mathbf{5_1}$)}] \ 

5200: The mapping $x\longmapsto A_{x,0:c}^1(1)\in M^1$ is of

5201: rank $n$.

5202: \item[{\bf ($\mathbf{6_1}$)}] \

5203: The mapping $v \longmapsto \frac{ \partial A_{0, v:c}^1}{ \partial

5204: \theta }\left(e^{ i\theta}\right)$ is of rank $n$ at $v=0$.

5205: Consequently, as $v$ varies, the positive half-lines $\R^+ \cdot

5206: \frac{ \partial A_{0,v:c}^1 }{ \partial \theta} (1)$ describe an open

5207: infinite cone containing $v_1$ with vertex $p_1$ in $T_{p_1} M^1$ and

5208: the set of points $A_{x,v:c}^1(\zeta)$, as $\vert x\vert \leq c$,

5209: $\vert v\vert \leq c$ and $\zeta\in \Delta$ vary, covers a wedge of

5210: edge $M^1$ at $(p_1,Jv_1)$.

5211: \end{itemize}

5212: \end{lemma}

5213: 

5214: In Figure~16 drawn in Section~8 after Lemma~8.3 below, we have drawn

5215: the property that the tangent direction $\frac{ \partial A_{0,v: c}^1}{

5216: \partial \theta} (1)$ describes an open cone in $T_{p_1}M^1$ with

5217: vertex $p_1$. The remainder of Section~3 is devoted to complete the

5218: proof of Proposition~7.12.

5219: 

5220: \subsection*{7.13.~Perturbed family of analytic discs half-attached to 

5221: a maximally real submanifold} Thus, let $M^1\subset \R^n$ be a locally

5222: defined maximally real $\mathcal{ C}^{2,\alpha}$-smooth submanifold

5223: passing through the origin. We can assume that $M^1$ is represented by

5224: $n$ Cartesian equations

5225: \def\theequation{7.14}\begin{equation}

5226: y_1=h_1(x_1,\dots,x_n),\cdots\cdots, y_n=h_n(x_1,\dots,x_n),

5227: \end{equation} 

5228: where $z_k= x_k+i y_k\in\C$, for $k= 1,\dots, n$, where $\vert x \vert

5229: \leq \rho_1$ for some $\rho_1 >0$, where $h= h(x)$ is of class

5230: $\mathcal{ C}^{2, \alpha}$ in $\{\vert x \vert < 2\rho_1\}$, and

5231: where, importantly, $h_j(0)=h_{j,x_k}(0)= h_{j, x_{k_1}

5232: x_{k_2}}(0)=0$, for all $j, k, k_1,k_2 =1, \dots,n$. As

5233: in~\thetag{6.5}, we set $K_1:=\vert\vert h \vert \vert_{ \mathcal{

5234: C}^{2,\alpha}(\{ \vert x \vert \leq \rho_1\})}$. Also, we can assume

5235: that $v_1=(0,\dots,0,1)$.

5236: 

5237: Our goal is to show that we can produce a $\mathcal{

5238: C}^{2,\alpha-0}$-smooth (remind $\mathcal{ C}^{2,\alpha-0} \equiv

5239: \bigcap_{ \beta < \alpha} \, \mathcal{ C}^{2, \beta}$) family of

5240: analytic discs $Z_{c,x,v}^1(\zeta)$ which is half-attached to $M^1$

5241: and which is sufficiently close, in $\mathcal{ C}^2$ norm, to the

5242: original family $Z_{c,x,v}^0(\zeta)$. After having constructed the

5243: family $Z_{c,x,v}^1(\zeta)$, we shall define the desired family

5244: $A_{x,v:c}^1(\zeta)$.

5245: 

5246: Let $d\in\R$ with $0\leq d \leq 1$ and let the maximally real

5247: submanifold $M^d$ (like ``$M$ {\it d}eformed'') be defined precisely

5248: as the set of $z= x+iy \in \C^n$ with $\vert x \vert \leq \rho_1$

5249: which satisfy the $n$ Cartesian equations

5250: \def\theequation{7.15}\begin{equation}

5251: y_1=d\cdot h_1(x_1,\dots,x_n),\cdots\cdots, \, 

5252: y_n=d\cdot h_n(x_1,\dots,x_n).

5253: \end{equation}

5254: Notice that $M^0 \equiv\{ x\in \R^n: \ \vert x \vert \leq \rho_1\}$ is

5255: essentially the same piece $M^0$ of $\R^n$ as in Lemma~7.9 (which

5256: contains the $M^0$ of Lemma~7.9 if we choose $\rho_0\leq \rho_1$) and

5257: notice that $\left. M^d\right \vert_{d=1} \equiv M^1$. Even better, we

5258: shall construct for each $d$ with $0 \leq d\leq 1$ a one-parameter

5259: family of analytic dics $Z_{c,x, v}^d (\zeta)$ which is of class at

5260: least $\mathcal{ C}^{2, \alpha-0}$ with respect to all variables and

5261: which is half-attached to $M^d$, by proceeding as follows.

5262: 

5263: First of all, the analytic disc $Z_{c,x, v}^d (\zeta)=: X_{c,x,

5264: v}^d (\zeta)+ iY_{c,x, v}^d( \zeta)$ is half-attached to $M^d$ if

5265: and only if

5266: \def\theequation{7.16}\begin{equation}

5267: Y_{c,x,v}^d(\zeta)=d\cdot h\left(

5268: X_{c,x,v}^d(\zeta)

5269: \right), \ \ \ \ \ 

5270: {\rm for} \ \zeta\in\partial^+\Delta.

5271: \end{equation}

5272: Furthermore, $Y_{c,x, v}^d$ should be a harmonic conjugate of

5273: $X_{c,x, v}^d$. However, the condition~\thetag{7.16} does not give any

5274: relation between $X_{ c,x, v}^d$ and $Y_{c,x, v}^d$ on the

5275: negative part $\partial^-\Delta$ of the unit circle. To fix this

5276: point, we shall assign the following more

5277: complete equation

5278: \def\theequation{7.17}\begin{equation}

5279: Y_{c,x,v}^d(\zeta)= d\cdot h\left(

5280: X_{c,x,v}^d(\zeta)\right)+Y_{c,x,v}^0(\zeta), 

5281: \ \ \ \ \

5282: \text{\rm for {\it all}} \ \zeta\in\partial\Delta,

5283: \end{equation}

5284: which coincides with~\thetag{7.16} for $\zeta\in

5285: \partial^+ \Delta$, since we have $Z_{c, x, v}^0( \partial^+

5286: \Delta) \subset \R^n$ by construction ({\it cf.}~\thetag{7.8}).

5287:  

5288: As in~\cite{tu2}, \cite{tu3}, \cite{mp1}, \cite{mp3}, we denote by

5289: $T_1$ the Hilbert transform (harmonic conjugate operator) on $\partial

5290: \Delta$ vanishing at $1$, namely $(T_1 X) (1)=0$, whence

5291: $T_1(T_1(X))=-X+X(1)$. By Privalov's theorem, for every integer

5292: $\kappa \geq 0$ and every $\alpha\in\R$ with $0< \alpha < 1$, its norm

5293: $\vert \vert \vert T_1 \vert \vert \vert_{ \kappa, \alpha}$ as an

5294: operator $\mathcal{ C}^{ \kappa, \alpha}( \partial \Delta, \R^n) \to

5295: \mathcal{ C}^{\kappa, \alpha} (\partial \Delta, \R^n)$ is finite and

5296: explodes as $\alpha$ tends either to $0$ or to $1$. Also, we shall

5297: require that $X_{c,x, v}^d (1) =x$, whence $Y_{ c,x, v}^d (1)= d

5298: \cdot h(x)$.

5299: 

5300: With this choice, the mapping $\zeta \mapsto Y_{c,x, v}^d (\zeta)$

5301: should necessarily coincide with the harmonic conjugate $\zeta \mapsto

5302: \left[T_1 X_{c,x, v}^d\right] 

5303: (\zeta)+ d\cdot h(x)$ (this property is already

5304: satisfied for $d=0$) and we deduce that $X_{c, x,v}^d( \zeta)$

5305: should satisfy the following Bishop type equation

5306: \def\theequation{7.18}\begin{equation}

5307: X_{c,x, v}^d(\zeta) =

5308: -T_1\left[

5309: d \cdot h\left(

5310: X_{c,x,v}^d

5311: \right)

5312: \right](\zeta)+X_{c,x,v}^0(\zeta), 

5313: \ \ \ \ \ 

5314: \text{\rm for all} \ \zeta\in\partial\Delta.

5315: \end{equation}

5316: Conversely, if $X_{c,x, v}^d$ is a solution of this functional

5317: equation, then setting $Y_{c,x, v}^d(\zeta):= T_1 X_{c,x, v}^d

5318: (\zeta)+ d\cdot h(x)$, it is easy to see that the analytic disc $Z_{c,x,

5319: v}^d (\zeta):= X_{c,x,v}^d (\zeta)+ iY_{c,x, v}^d(\zeta)$ is

5320: half-attached to $M^d$ and more precisely, satisfies the

5321: equation~\thetag{7.16}.

5322: 

5323: Applying now Theorem~1.2 of \cite{tu3}, we deduce that if the given

5324: positive number $c_1$ is sufficiently small, and if $c$ satisfies

5325: $0\leq c \leq c_1$, there exists a unique solution $X_{c,x,

5326: v}^d(\zeta)$ to~\thetag{7.18} which is of class $\mathcal{ C}^{2,

5327: \alpha}$ with respect to $\zeta$ and of class $\mathcal{ C}^{2,\alpha

5328: -0}$ with

5329: respect to all variables $(c,x, v, \zeta)$ with $0\leq c \leq c_1$,

5330: $\vert x\vert \leq c$, $\vert v \vert \leq 2$ and $\zeta\in

5331: \overline{\Delta}$. We shall now estimate the difference $\vert \vert

5332: Z_{c,x,v}^d-Z_{c,x,v}^0\vert \vert_{ \mathcal{ C}^{1,\alpha} (\partial

5333: \Delta)}$ and prove that it is bounded by a constant times

5334: $c^{2+\alpha}$. In particular, if $c_1$ is sufficiently small, this

5335: will imply that $Z_{c,x,v}^d$ is nonconstant.

5336: 

5337: \subsection*{7.19.~Size of the solution 

5338: $X_{c,x,v}^d(\zeta)$ in $\mathcal{ C}^{1,\alpha}$ norm} Following the

5339: beginning of the proof of Theorem~1.2 in~\cite{tu3}, we introduce the

5340: mapping

5341: \def\theequation{7.20}\begin{equation}

5342: F: \ X(\zeta)\longmapsto X_{c,x,v}^0(\zeta)

5343: -T_1\left[

5344: d\cdot h(X)

5345: \right](\zeta)

5346: \end{equation}

5347: from a neighborhood of $0$ in $\mathcal{ C}^{1,\alpha}( \partial

5348: \Delta, \R^n)$ to $\mathcal{ C}^{1,\alpha}(\partial \Delta, \R^n)$,

5349: and then, as in~\cite{b}, we introduce a Picard iteration processus by

5350: defining $X\{0\}_{c,x,v}^d(\zeta):= X_{c,x,v}^0(\zeta)$ and for every

5351: integer $\nu\geq 0$

5352: \def\theequation{7.21}\begin{equation}

5353: X\{\nu+1\}_{c,x,v}^d(\zeta):= 

5354: F\left(X\{\nu\}_{c,x,v}^d(\zeta)\right).

5355: \end{equation}

5356: In a first moment, A.~Tumanov proves in~\cite{tu3} that the sequence

5357: $\left( X\{\nu \}_{c,x, v}^d (\zeta) \right)_{ \nu \in \N}$ converges

5358: towards the unique solution $X_{c,x, v}^d(\zeta)$ of~\thetag{7.18} in

5359: $\mathcal{ C}^{1, \alpha}( \partial \Delta)$. Admitting this

5360: convergence result, we need to extract the supplementary information

5361: that $\left \vert \left \vert X_{c,x, v}^d \right \vert\right\vert_{

5362: \mathcal{ C}^{1, \alpha}( \partial \Delta)} \leq c\cdot K_2$ for some

5363: positive constant $K_2$, which will play the role of the constant

5364: $K_2$ of Lemma~6.12.

5365: 

5366: To get this information, we observe that by construction

5367: ({\it cf.}~\thetag{7.8}) there exists a constant $K_4>0$ such that

5368: \def\theequation{7.22}\begin{equation}

5369: \left\vert \left\vert X_{c,x,v}^0\right\vert 

5370: \right\vert_{\mathcal{

5371: C}^{2,\alpha}(\partial \Delta)}\leq c \cdot K_4.

5372: \end{equation}

5373: Also, we set $K_5:= K_1 (3+ \pi^{1-\alpha}) \vert \vert \vert T_1

5374: \vert \vert \vert_{ \mathcal{ C}^{1, \alpha}( \partial \Delta)}$.

5375: 

5376: \def\thelemma{{\bf 7.23}}\begin{lemma}

5377: With these notations, if 

5378: \def\theequation{7.24}\begin{equation}

5379: c_1\leq\min\left( 

5380: \frac{\rho_1}{2K_4}, \ 

5381: \left(

5382: \frac{1}

5383: {2^{2+\alpha}\, K_4^{1+\alpha} \, K_5}

5384: \right)^{\frac{1}{1+\alpha}}\right),

5385: \end{equation}

5386: then the solution of~\thetag{7.18} satisfies $\left\vert X_{c,x, v}^d

5387: \left( e^{i \theta }\right) \right\vert \leq \rho_1$ for all

5388: $e^{i\theta}\in\partial \Delta$ and there exists a constant $K_2>0$

5389: such that

5390: \def\theequation{7.25}\begin{equation}

5391: \left

5392: \vert \left \vert X_{c,x, v}^d 

5393: \right \vert \right \vert_{ \mathcal{

5394: C}^{1, \alpha} ( \partial \Delta)} \leq c \cdot K_2. 

5395: \end{equation}

5396: In fact, it

5397: suffices to choose $K_2:= 2 K_4$.

5398: \end{lemma}

5399: 

5400: \proof

5401: Indeed, using Lemmas~6.4 and~6.9, if $X \in \mathcal{ C}^{1, \alpha}(

5402: \partial \Delta, \R^n)$ satisfies $\left\vert 

5403: X\left( e^{i \theta }\right) \right\vert \leq

5404: \rho_1$ for all $e^{i\theta}\in\partial\Delta$ and $\vert \vert X \vert

5405: \vert_{ \mathcal{ C}^{1, \alpha} (\partial \Delta )} \leq c \cdot

5406: 2K_4$ for all $c \leq c_1$, where $c_1$ is as in~\thetag{7.24}, we

5407: may estimate (remind $0 \leq d \leq 1$)

5408: \def\theequation{7.26}\begin{equation}

5409: \left\{

5410: \aligned

5411: \vert \vert F(X) \vert \vert_{\mathcal{ C}^{1, 

5412: \alpha}(\partial \Delta)} 

5413: & \

5414: \leq

5415: \left\vert \left\vert X_{c,x,v}^0 \right\vert 

5416: \right\vert_{\mathcal{ C}^{1,\alpha}

5417: (\partial \Delta)}+

5418: \vert \vert \vert T_1 \vert \vert \vert_{\mathcal{ C}^{1, \alpha}

5419: (\partial \Delta)} \cdot

5420: \vert \vert h(X) \vert \vert_{\mathcal{ C}^{1, \alpha}

5421: (\partial \Delta)} \\

5422: & \

5423: \leq 

5424: c\cdot K_4+

5425: \vert \vert \vert T_1 \vert \vert \vert_{\mathcal{ C}^{1, \alpha}

5426: (\partial \Delta)} \cdot K_1

5427: ( c \cdot 2K_4)^{2+\alpha}(

5428: 3+\pi^{1-\alpha}) \\

5429: & \

5430: = 

5431: c\cdot \left(

5432: K_4+c^{1+\alpha} 2^{2+\alpha} K_4^{2+\alpha} K_5

5433: \right) \\

5434: & \

5435: \leq 

5436: c\cdot (K_4+c_1^{1+\alpha} 

5437: 2^{2+\alpha}K_4^{2+\alpha} K_5) \\

5438: & \

5439: \leq c\cdot 2K_4.

5440: \endaligned\right.

5441: \end{equation}

5442: Notice that from the last inequality, it also follows that $\left\vert

5443: F\left( X\left( e^{i\theta} \right) \right)\right\vert \leq \rho_1$

5444: for all $e^{i\theta}\in\partial \Delta$. Consequently, the processus

5445: of successive approximations~\thetag{7.21} is well defined for each

5446: $\nu\in\N$ and from the inequality~\thetag{7.26}, we deduce that the

5447: limit $X_{c,x,v}^d$ satisfies the desired estimate $\left\vert

5448: \left\vert X_{c,x, v}^d \right\vert \right\vert_{ \mathcal{ C}^{1,

5449: \alpha}( \partial \Delta )}\leq c \cdot 2K_4$, which completes the

5450: proof.

5451: \endproof

5452: 

5453: \def\thecorollary{{\bf 7.27}}\begin{corollary}

5454: Under the above assumptions, there exists a constant

5455: $K_6>0$ such that

5456: \def\theequation{7.28}\begin{equation}

5457: \left\vert \left\vert X_{c,x,v}^d - X_{c,x,v}^0 

5458: \right\vert \right\vert_{

5459: \mathcal{ C}^{1,\alpha}(\partial \Delta)} \leq

5460: c^{2+\alpha} \cdot K_6.

5461: \end{equation}

5462: \end{corollary}

5463: 

5464: \proof

5465: We estimate

5466: \def\theequation{7.29}\begin{equation}

5467: \left\{

5468: \aligned

5469: \left\vert \left\vert X_{c,x,v}^d - X_{c,x,v}^0 

5470: \right\vert \right\vert_{

5471: \mathcal{ C}^{1,\alpha}(\partial \Delta)} 

5472: & \

5473: \leq

5474: \vert \vert \vert T_1 \vert \vert \vert_{\mathcal{ C}^{1,\alpha}

5475: (\partial \Delta)} \cdot 

5476: \left\vert \left\vert h

5477: \left(X_{c,x,v}^d\right) \right\vert 

5478: \right\vert_{\mathcal{ C}^{1,\alpha}

5479: (\partial \Delta)} \\

5480: & \ 

5481: \leq 

5482: \vert \vert \vert T_1 \vert \vert \vert_{\mathcal{ C}^{1,\alpha}

5483: (\partial \Delta)} \cdot 

5484: K_1 (c\cdot 2K_4)^{2+\alpha}(3+\pi^{1-\alpha}) \\

5485: & \

5486: \leq

5487: c^{2+\alpha} \cdot K_5 (2K_4)^{2+\alpha}.

5488: \endaligned\right.

5489: \end{equation}

5490: so that it suffices to set 

5491: $K_6:= K_5(2K_4)^{2+\alpha}$.

5492: \endproof

5493: 

5494: \subsection*{7.30.~Smallness of the deformation in $\mathcal{ C}^2$

5495: norm} As was already noticed (and admitted), the solution

5496: $X_{c,x,v}^d(\zeta)$ is in fact $\mathcal{ C}^{2,\alpha}$-smooth with

5497: respect to $\zeta$ and $\mathcal{ C}^{2,\alpha-0}$-smooth with respect

5498: to all variables $(d,c,x,v,\zeta)$. We can therefore differentiate

5499: twice Bishop's equation~\thetag{7.18}. First of all, if $X\in\mathcal{

5500: C}^{2,\alpha-0}( \partial \Delta, \R^n)$, we remind the commutation

5501: relation $\frac{\partial }{\partial \theta} (TX)= T\left(

5502: \frac{\partial X}{\partial \theta} \right)$, whence

5503: \def\theequation{7.31}\begin{equation}

5504: \frac{\partial}{\partial \theta} \left(

5505: T_1X

5506: \right)= 

5507: T\left(

5508: \frac{\partial X}{\partial \theta}

5509: \right),

5510: \end{equation}

5511: since $T_1X= TX-TX(1)$. We may then compute the first order 

5512: derivative of~\thetag{7.18} with respect 

5513: to $\theta$:

5514: \def\theequation{7.32}\begin{equation}

5515: \left\{

5516: \aligned

5517: \frac{\partial }{\partial \theta} 

5518: X_{c,x,v}^d\left(e^{i\theta}\right) -

5519: & \

5520: \frac{\partial }{\partial \theta}

5521: X_{c,x,v}^0\left(e^{i\theta}\right)=

5522: -T\left[

5523: d\cdot \sum_{l=1}^n\, 

5524: \frac{\partial h}{\partial x_l} \left(X_{c,x,v}^d\right) \

5525: \frac{\partial X_{l;c,x,v}^d}{\partial \theta}

5526: \right]\left(e^{i\theta}\right).

5527: \endaligned\right.

5528: \end{equation}

5529: and then its second order partial derivatives $\partial^2 / \partial

5530: v_k \partial \theta$, for $k=1,\dots,n$, without writing the argument

5531: $e^{i\theta}$:

5532: \def\theequation{7.33}\begin{equation}

5533: \left\{

5534: \aligned

5535: \frac{\partial^2 X_{c,x,v}^d}{\partial v_k\partial \theta}

5536: -

5537: \frac{\partial^2X_{c,x,v}^0}{\partial v_k

5538: \partial \theta}

5539: = 

5540: & \

5541: -T\left[

5542: d\cdot \sum_{l_1,l_2=1}^n\,

5543: \frac{\partial^2 h}{\partial x_{l_1} \partial x_{l_2}}\left(

5544: X_{c,x,v}^d\right) \

5545: \frac{\partial X_{l_1; c,x,v}^d}{\partial v_k} \,

5546: \frac{\partial X_{l_2; c,x,v}^d}{\partial \theta} + \right. \\

5547: & \

5548: \left. \ \ \ \ \ \ \ \ \ \

5549: +d\cdot \sum_{l=1}^n\, 

5550: \frac{\partial h_j}{\partial x_l} \left(

5551: X_{c,x,v}^d

5552: \right) \

5553: \frac{\partial^2 X_{l;c,x,v}^d}{\partial v_k 

5554: \partial \theta}

5555: \right].

5556: \endaligned\right.

5557: \end{equation}

5558: Let now $K_2$ be as in~\thetag{7.25} and let 

5559: $K_3$ be as in Lemma~6.12, applied to 

5560: $X_{c,x,v}^d(\zeta)$.

5561: 

5562: \def\thelemma{7.34}\begin{lemma}

5563: If in addition to the inequality~\thetag{7.24}, the constant

5564: $c_1$ satisfies the inequality

5565: \def\theequation{7.35}\begin{equation}

5566: c_1\leq \left(

5567: \frac{1}{2K_3 \vert \vert \vert 

5568: T \vert \vert \vert_{

5569: \mathcal{ C}^\alpha(\partial \Delta)}}

5570: \right)^{\frac{1}{1+\alpha}},

5571: \end{equation}

5572: then there exists a positive constant $K_7>0$

5573: such that for all $d$, all $c$, all $x$, all $v$, and for

5574: $k=1,\dots,n$, the following two estimates hold{\rm :}

5575: \def\theequation{7.36}\begin{equation}

5576: \left\{

5577: \aligned

5578: \left\vert\left\vert

5579: \frac{\partial^2 X_{c,x,v}^d}{

5580: \partial v_k \partial \theta}-

5581: \frac{\partial^2 X_{c,x,v}^0}{

5582: \partial v_k \partial \theta}

5583: \right\vert\right\vert_{\mathcal{ 

5584: C}^\alpha(\partial \Delta)}\leq 

5585: & \

5586: c^{2+\alpha} \cdot K_7, \\

5587: \left\vert\left\vert

5588: \frac{\partial^2 X_{c,x,v}^d

5589: }{\partial \theta^2}-

5590: \frac{\partial^2 X_{c,x,v}^0

5591: }{\partial \theta^2}

5592: \right\vert\right\vert_{\mathcal{ 

5593: C}^\alpha(\partial \Delta)}\leq 

5594: & \

5595: c^{2+\alpha} \cdot K_7.

5596: \endaligned\right.

5597: \end{equation}

5598: \end{lemma}

5599: 

5600: \proof

5601: We check only the first inequality, the proof of the second being

5602: totally similar. According to Lemma~1.6 in~\cite{ tu3}, 

5603: there exists a solution 

5604: $\frac{\partial^2 X_{c,x,v}^d}{

5605: \partial v_k \partial \theta}$

5606: to the linearized Bishop equation~\thetag{ 7.33},

5607: hence it suffices to make an estimate.

5608: 

5609: Introducing for the second line of~\thetag{7.33} a

5610: new simplified notation $\mathcal{ R}:= -T\left[ d\cdot

5611: \sum_{l_1,l_2=1}^n\, \frac{\partial^2 h}{ \partial x_{l_1} \partial

5612: x_{l_2} }\left( X_{c,x,v}^d \right) \ \frac{\partial X_{l_1;

5613: c,x,v}^d}{\partial v_k} \, \frac{\partial X_{l_2; c,x,v}^d}{ \partial

5614: \theta}\right]$ and setting further obvious simplifying changes of

5615: notation, we can rewrite~\thetag{7.33} more concisely as

5616: \def\theequation{7.37}\begin{equation}

5617: \mathcal{ X}^d-\mathcal{ X}^0= 

5618: \mathcal{ R}- T\left[

5619: d\cdot \mathcal{ H} \mathcal{ X}^d

5620: \right].

5621: \end{equation}

5622: Here, thanks to the inequality $\left\vert \left\vert X_{c,x, v}^d

5623: \right\vert \right\vert_{\mathcal{ C}^{1, \alpha} (\partial

5624: \Delta)}\leq c\cdot K_2$ already established in Lemma~7.23 and thanks

5625: to Lemma~6.12, we know that the vector $\mathcal{ R}\in \mathcal{

5626: C}^{\alpha }( \partial \Delta, \R^n)$ and the matrix $\mathcal{ H}\in

5627: \mathcal{ C}^{1, \alpha}( \partial \Delta, \mathcal{ M}_{n\times

5628: n}(\R))$ are small and more precisely, they satisfy the following two

5629: estimates

5630: \def\theequation{7.38}\begin{equation}

5631: \left\{

5632: \aligned

5633: \vert \vert \mathcal{ R} \vert \vert_{

5634: \mathcal{ C}^\alpha(\partial \Delta)}\leq

5635: & \

5636: c^{2+\alpha} \cdot 

5637: \vert \vert \vert T \vert \vert \vert_{\mathcal{ C}^\alpha

5638: (\partial \Delta)} K_3 (K_2)^2

5639: \\

5640: \vert \vert

5641: \mathcal{ H} \vert \vert_{\mathcal{ C}^\alpha(

5642: \partial \Delta)}\leq 

5643: & \

5644: c^{1+\alpha} \cdot K_3.

5645: \endaligned\right.

5646: \end{equation}

5647: We can rewrite~\thetag{7.37} under the form

5648: \def\theequation{7.39}\begin{equation}

5649: \mathcal{ X}^d -\mathcal{ X}^0 =

5650: \mathcal{ S} -T\left[

5651: d\cdot \mathcal{ H}(\mathcal{ X}^d-

5652: \mathcal{ X}^0)

5653: \right],

5654: \end{equation}

5655: with $\mathcal{ S}:= \mathcal{ R}-T \left[ d\cdot \mathcal{ H}

5656: \mathcal{ X}^0 \right]$. Using the inequality $\left\vert \left\vert

5657: \mathcal{ X}^0 \right\vert \right\vert_{ \mathcal{ C}^\alpha( \partial

5658: \Delta)} \leq c\cdot K_4$ which is a direct consequence

5659: of~\thetag{7.22} and taking the previous estimates~\thetag{7.38}

5660: into account, we deduce the inequality

5661: \def\theequation{7.40}\begin{equation}

5662: \vert \vert \mathcal{ S}

5663: \vert \vert_{\mathcal{ C}^\alpha(\partial \Delta)}\leq

5664: c^{2+\alpha}\cdot

5665: \vert \vert \vert T \vert \vert \vert_{\mathcal{ C}^\alpha

5666: (\partial \Delta)}\left[

5667: K_3(K_2)^2+

5668: K_3K_4

5669: \right].

5670: \end{equation} 

5671: Taking the $\mathcal{ C}^\alpha(\partial \Delta)$ norm of both sides

5672: of~\thetag{7.40}, we deduce the estimate

5673: \def\theequation{7.41}\begin{equation}

5674: \left\{

5675: \aligned

5676: \left\vert \left\vert 

5677: \mathcal{ X}^d - \mathcal{ X}^0 \right\vert 

5678: \right\vert_{

5679: \mathcal{ C}^\alpha(\partial \Delta)} \leq

5680: & \ 

5681: c^{2+\alpha} \cdot \frac{

5682: \vert \vert \vert T \vert \vert \vert_{\mathcal{ C}^\alpha

5683: (\partial \Delta)}\left[

5684: K_3(K_2)^2+

5685: K_3K_4

5686: \right]

5687: }{

5688: 1- c^{1+\alpha}

5689: \cdot \vert \vert \vert T \vert \vert \vert_{\mathcal{ C}^\alpha

5690: (\partial \Delta)} K_3} \\

5691: \leq

5692: & \

5693: c^{2+\alpha} \cdot 2\vert 

5694: \vert \vert T \vert \vert \vert_{\mathcal{ C}^\alpha

5695: (\partial \Delta)}\left[

5696: K_3(K_2)^2+

5697: K_3K_4

5698: \right]

5699: \endaligned\right.

5700: \end{equation}

5701: where we use the assumption~\thetag{7.35} on $c_1$

5702: to obtain the second second inequality. It suffices

5703: to set $K_7:= 2\vert 

5704: \vert \vert T \vert \vert \vert_{\mathcal{ C}^\alpha

5705: (\partial \Delta)}\left[

5706: K_3(K_2)^2+

5707: K_3K_4

5708: \right]$, which completes the proof.

5709: \endproof

5710: 

5711: \subsection*{7.42.~Adjustment of the tangent vector}

5712: Let $v_1\in T_{p_1}M^1$ with $\vert v_1 \vert =1$, as in Lemma~7.12.

5713: Coming back to the first family $Z_{c,x, v}^0 (\zeta)$ defined

5714: by~\thetag{7.8}, we observe that

5715: \def\theequation{7.43}\begin{equation}

5716: \left\{

5717: \aligned

5718: \frac{\partial Z_{j;c,0,v_1}^0}{\partial x_k}

5719: (1)=

5720: & \

5721: \delta_k^j, 

5722: \ \ \ \ \ 

5723: j,k=1,\dots,n, \\

5724: \frac{\partial^2 Z_{j;

5725: c,0,v_1}^0}{\partial v_k 

5726: \partial \theta}(1)=

5727: & \

5728: c\, \frac{\partial \Psi}{\partial \theta}

5729: \left(e^{i\theta}\right) \, 

5730: \delta_k^j, 

5731: \ \ \ \ \ 

5732: j,k=1,\dots,n.

5733: \endaligned\right.

5734: \end{equation}

5735: From now on, we shall set $d=1$ and we shall only consider the family

5736: $Z_{c,x,v}^1(\zeta)$. Thanks to the estimates~\thetag{7.28}

5737: and~\thetag{7.36}, we deduce that if $c_1$ is sufficiently small, then

5738: for all $c$ with $0< c \leq c_1$, the two Jacobian matrices

5739: \def\theequation{7.44}\begin{equation}

5740: \left(

5741: \frac{\partial Z_{j;c,0,v_1}^1}{\partial x_k}(1)

5742: \right)_{1\leq j,k\leq n} \ \ \ \ \

5743: {\rm and} \ \ \ \ \ \

5744: \left(

5745: \frac{\partial^2 Z_{j;c,0,v_1}^1}{\partial v_k\partial \theta}(1)

5746: \right)_{1\leq j,k\leq n}

5747: \end{equation}

5748: are invertible. It would follow that if we would set $A_{x, v:c}^1

5749: (\zeta):= Z_{c, x, v_1 +v}^1(\zeta)$, similarly as in~\thetag{7.10},

5750: then the disc $A_{x, v:c}^1 (\zeta)$ would satisfy the two rank

5751: properties {\bf ($\mathbf{ 5_1}$)} and {\bf ($\mathbf{ 6_1}$)} of

5752: Lemma~7.12. However, the tangency condition {\bf ($\mathbf{ 4_1}$)}

5753: would certainly not be satisfied, because as $d$ varies from $0$ to

5754: $1$, the disc $Z_{c,x,v}^d(\zeta)$ undergo a nontrivial deformation.

5755: 

5756: Consequently, for every $c$ with $0< c \leq c_1$, 

5757: we have to adjust the ``cone

5758: parameter'' $v$ in order to maintain the tangency 

5759: condition.

5760: 

5761: \def\thelemma{7.45}\begin{lemma}

5762: For every $c$ with $0< c \leq c_1$, there exists a

5763: vector $v(c)\in \R^n$ such that

5764: \def\theequation{7.46}\begin{equation}

5765: \frac{\partial Z_{c,0,v_1

5766: +v(c)}^1}{\partial \theta} 

5767: (1)=

5768: \frac{\partial Z_{c,0,

5769: v_1}^0}{\partial \theta}(1)=c \cdot

5770: \frac{\partial \Psi}{\partial \theta}(1) \cdot v_1.

5771: \end{equation}

5772: Furthermore, there exists a constant $K_8>0$ such that

5773: $\vert v(c) \vert \leq c^{1+\alpha} \cdot K_8$.

5774: \end{lemma}

5775: 

5776: \proof

5777: Unfortunately, we cannot apply the implicit function theorem, because

5778: the mapping $Z_{c,x,v}^1$ is identically zero when $c=0$, so we have

5779: to proceed differently. First, we set

5780: \def\theequation{7.47}\begin{equation}

5781: C_1:= \frac{\partial \Psi}{\partial \theta}(1), \ \ \ \ \ 

5782: {\rm and}

5783: \ \ \ \ \

5784: C_2:= \vert\vert

5785: \Psi \vert\vert_{\mathcal{ C}^2\left(\overline{\Delta}\right)}.

5786: \end{equation}

5787: The constant $C_2$ will be used only in Section~8 below.

5788: Choose $K_8\geq \frac{2K_6}{C_1}$. According to the explicit

5789: expression~\thetag{7.8}, the set of points

5790: \def\theequation{7.48}\begin{equation}

5791: \left\{

5792: \frac{\partial X_{c,0,v_1+v}^0}{\partial \theta}(1)\in \R^n: \ 

5793: \vert v \vert \leq c^{1+\alpha} \cdot K_8

5794: \right\}

5795: \end{equation}

5796: covers a cube in $\R^n$ centered at the point $\frac{ \partial X_{

5797: c,0, v_1 }^0}{\partial \theta}(1)$ of radius $c^{2+ \alpha} \cdot C_1

5798: K_8$.

5799: Thanks to the estimate~\thetag{7.28}, we deduce

5800: that the (deformed) set of points 

5801: \def\theequation{7.49}\begin{equation}

5802: \left\{

5803: \frac{\partial X_{c,0,v_1+v}^1}{\partial \theta}(1)\in \R^n: \ 

5804: \vert v \vert \leq c^{1+\alpha} \cdot K_8

5805: \right\}

5806: \end{equation}

5807: covers a cube in $\R^n$ centered at the same

5808: point $\frac{ \partial X_{

5809: c,0, v_1 }^0}{ \partial \theta}(1)$, but of radius 

5810: \def\theequation{7.50}\begin{equation}

5811: c^{2+ \alpha} \cdot C_1

5812: K_8-c^{2+\alpha} \cdot K_6\geq c^{2+\alpha} \cdot K_6.

5813: \end{equation}

5814: Consequently, there exists at least one $v(c) \in \R^n$ with $\vert

5815: v(c) \vert \leq c^{ 1+\alpha} \cdot K_8$ such that~\thetag{7.46}

5816: holds, which completes the proof. 

5817: \endproof

5818: 

5819: \subsection*{7.51.~Construction of the family $A_{x, v:c}^1 (\zeta)$}

5820: We can now complete the proof of the main Lemma~7.12 of the present

5821: section. First of all, with $\Psi (\zeta)$ as in \S7.5 ans in {\sc

5822: Figure~14}, we consider the composed conformal mapping

5823: \def\theequation{7.52}\begin{equation}

5824: \zeta \longmapsto c\Psi(\zeta) \longmapsto 

5825: \frac{i-c\Psi(\zeta)}{i+c\Psi(\zeta)}=:

5826: \Phi_c(\zeta).

5827: \end{equation}

5828: The image $\Phi_c ( \zeta)$ of the unit disc is a small domain

5829: contained in $\Delta$ and concentrated near $1$. More precisely,

5830: assuming that $c$ satifies $0< c \leq c_1$ with $c_1<<1$ as in the

5831: previous paragraphs, and taking account of the definition of $\Psi

5832: (\zeta)$, it can be checked easily that $\Phi_c(1) =1$, that $\Phi_c

5833: (\partial^+ \Delta)$ is contained in $\{e^{ i\theta} \in

5834: \partial^+\Delta: \ \vert \theta \vert < 10c\}$, and that

5835: \def\theequation{7.53}\begin{equation}

5836: \Phi_c\left(\overline{\Delta}\backslash 

5837: \partial^+\Delta\right)\subset 

5838: \{ \zeta \in \Delta : \ 

5839: \vert \zeta -1 \vert < 8c\} \subset 

5840: \{\rho e^{i\theta} \in \Delta : \

5841: \vert \theta \vert < 10c, \

5842: 1-10c < \rho < 1\}.

5843: \end{equation}

5844: the second inclusion being trivial. Here is an illustration:

5845: 

5846: \bigskip

5847: \begin{center}

5848: \input sub-half-disc.pstex_t

5849: \end{center}

5850: 

5851: We can now define the final desired family of analytic discs, 

5852: writing the parameter $c$ after a semi-colon, since we lose the

5853: $\mathcal{ C}^{2,\alpha-0}$-smoothness with respect to 

5854: $c$ after the application of Lemma~7.45:

5855: \def\theequation{7.54}\begin{equation}

5856: A_{x,v:c}^1(\zeta):= Z_{c,x,v_1+v(c)+v}^1\left(

5857: \Phi_c(\zeta)\right).

5858: \end{equation}

5859: We restrict the variation of the parameters $x$ to $\vert x \vert \leq

5860: c^2$ and $v$ to $\vert v \vert\leq c$. Property {\bf ($\mathbf{

5861: 4_1}$)} holds immediately, thanks to the choice of $v(c)$. Properties

5862: {\bf ($\mathbf{1_1}$)}, {\bf ($\mathbf{ 3_1}$)}, {\bf ($\mathbf{

5863: 5_1}$)} and {\bf ($\mathbf{ 6_1}$)} as well as the embedding property

5864: in {\bf ($\mathbf{2_1}$)} are direct consequences of the similar

5865: properties~\thetag{7.44} satisfied by $Z_{ c,x, v_1+ v(c)+v}^1 (\zeta)$,

5866: using the chain rule and the nonvanishing of the partial derivative

5867: $\frac{ \partial \Phi_c }{ \partial \theta}(1)$. The size estimate in

5868: {\bf ($\mathbf{2_1}$)} follows from~\thetag{7.25}, from~\thetag{7.28},

5869: from the restriction of the domains of variation of $x$ and of $v$ and

5870: from~\thetag{7.53}. This completes the proof of Lemma~7.12.

5871: \hfill 

5872: \qed

5873: 

5874: \section*{\S8.~Geometric properties of families of half-attached 

5875: analytic discs}

5876: 

5877: \subsection*{8.1.~Preliminary}

5878: By Lemma~7.12, for every $c$ with $0 < c \leq c_1$, the family of

5879: half-attached analytic discs $A_{x, v:c}^1 (\zeta)$ covers a local

5880: wedge of edge $M^1$ at $p_1$. However, not only we want the family

5881: $A_{x, v:c}^1$ to cover a local wedge of edge $M^1$ at $p_1$, but we

5882: certainly want to remove the point $p_1$ by means of the continuity

5883: principle, under the assumptions of the main Proposition~5.12, a final

5884: task which will be achieved in Section~9 below. Consequently, in each

5885: one of the three geometric situations {\bf ($\mathbf{I_1}$)}, {\bf

5886: ($\mathbf{I_2}$)} and {\bf (II)} which we have normalized in

5887: Lemma~5.37 above, we shall firstly deduce from the tangency condition

5888: {\bf ($\mathbf{4_1}$)} of Lemma~7.12 that the blunt half-boundary

5889: $A_{0,0:c}^1( \partial^+ \Delta \backslash \{1\})$ is contained in the

5890: open side $(H^1)^+$ (this is why we have normalized in Lemma~5.37 the

5891: second order terms of the supporting hypersurface $H^1$ in order that

5892: $(H^1)^+$ is strictly concave; the reason why we require that

5893: $A_{0,0:c}^1(\partial^+\Delta\backslash \{1\})$ is contained in

5894: $(H^1)^+$ will be clear in Section~9 below). Secondly, we shall show

5895: that for all $x$ with $\vert x \vert \leq c^2$, the disc interior

5896: $A_{x,0:c}(\Delta)$ is contained in the local half-wedge $\mathcal{

5897: HW}_1^+$ in the cases {\bf ($\mathbf{I_1}$)}, {\bf ($\mathbf{I_2}$)}

5898: and is contained in the wedge $\mathcal{ W}_2$ in case {\bf (II)}.

5899: 

5900: \subsection*{8.2.~Geometric disposition of the

5901: discs with respect to $H^1$ and to $\mathcal{ HW}_1^+$ or to

5902: $\mathcal{ W}_2$} We remember that the positive $c_1$ of Lemmas~7.12,

5903: 7.23 and 7.34 was shrunk explicitely, in terms of the constants

5904: $K_1,K_2,K_3,\dots$. In this section, we shall again shrink $c_1$

5905: several times, but without mentioning all the similar explicit

5906: inequalities which will appear. The precise statement of the main

5907: lemma of this section, which is a continuation of Lemma~7.12, is as

5908: follows; whereas we can essentially gather the three cases in the

5909: formal statement of the lemma, it is necessary to treat them

5910: separately in the proof, because the normalizations of

5911: Lemma~5.37 differ.

5912: 

5913: \def\thelemma{8.3}\begin{lemma}

5914: Let $M$, let $M^1$, let $p_1$, let $H^1$, let $v_1$, let $(H^1)^+$,

5915: let $\mathcal{ HW }_1^+$ or let $\mathcal{ HW }_2$ and let a

5916: coordinate system $z= (z_1, \dots, z_n)$ vanishing at $p_1$ be as in

5917: Case {\bf ($\mathbf{I_1}$)}, as in Case {\bf ($\mathbf{ I_2 }$)} or as

5918: in Case {\bf (II)} of Lemma~5.37. Choose as a local one-dimensional

5919: submanifold $T^1 \subset M^1$ transversal to $H^1$ in $M^1$ and

5920: passing through $p_1$ the submanifold $T_1:=\{\left( x_1,0, \dots,0) +

5921: ih(x_1, 0, \dots, 0) \right)\}$ in Case {\bf

5922: ($\mathbf{I_1}$)} and the submanifold $T_1:=\{\left( 0, \dots,0,x_n) +

5923: ih(0, \dots, 0,x_n) \right)\}$ in Cases {\bf

5924: ($\mathbf{I_2}$)} and {\bf (II)}. For every $c$ with $0 < c \leq

5925: c_1$, let $A_{x,v:c}^1(\zeta)$ be the family of analytic discs

5926: satisfying properties {\bf ($\mathbf{1_1}$)}, {\bf ($\mathbf{2_1}$)},

5927: {\bf ($\mathbf{3_1}$)}, {\bf ($\mathbf{4_1}$)}, {\bf ($\mathbf{5_1}$)}

5928: and {\bf ($\mathbf{6_1}$)} of Lemma~7.12. Shrinking $c_1$ if

5929: necessary, then for every $c$ with $0< c\leq c_1$, the following three

5930: further properties hold

5931: \begin{itemize}

5932: \item[{\bf ($\mathbf{7_1}$)}] $A_{0, 0:c}^1( \partial^+ \Delta

5933: \backslash\{ 1\}) \subset (H^1)^+$.

5934: \item[{\bf ($\mathbf{8_1}$)}]

5935: $A_{x,0:c}^1(\partial^+\Delta)$ is contained

5936: in $(H^1)^+$ for all $x$ such that the point 

5937: $A_{x,0:c}^1(1)$ belongs to 

5938: $T^1\cap (H^1)^+$.

5939: \item[{\bf ($\mathbf{9_1}$)}]

5940: $A_{x,v:c}^1(\overline{\Delta}\backslash\partial^+\Delta)$ is

5941: contained in the half-wedge

5942: $\mathcal{ HW}_1^+$ or in the wedge $\mathcal{ W}_2$ for all

5943: $x$ and all $v$.

5944: \end{itemize}

5945: \end{lemma}

5946: 

5947: \proof

5948: For the three new properties {\bf ($\mathbf{ 7_1}$)}, {\bf ($\mathbf{

5949: 8_1}$)}, and {\bf ($\mathbf{ 9_1}$)}, we study thoroughly only Case

5950: {\bf ($\mathbf{I_1}$)}, because the other two cases can be treated in

5951: a totally similar way. {\sc Figure~16} just below illustrates

5952: properties {\bf ($\mathbf{ 7_1}$)} and {\bf ($\mathbf{ 8_1}$)} and

5953: also properties {\bf ($\mathbf{ 1_1}$)}, {\bf ($\mathbf{ 5_1}$)} and

5954: {\bf ($\mathbf{ 6_1}$)} of Lemma~7.12.

5955: 

5956: \bigskip

5957: \begin{center}

5958: \input half-boundaries.pstex_t

5959: \end{center}

5960: 

5961: Intuitively, the reason why property {\bf ($\mathbf{ 7_1}$)} holds

5962: true is clear: the open set $(H^1)^+$ is strictly concave and the

5963: small segment $A_{0,0:c}^1(\partial^+\Delta)$ is tangent to $H^1$ at

5964: $p_1$; also, the reason why property {\bf ($\mathbf{ 8_1}$)} holds

5965: true is equally clear: when $x$ varies, the small segments

5966: $A_{x,0:c}^1(\partial^+\Delta)$ are essentially translated (inside

5967: $M^1$) from $p_1$ by the vector $x\in\R^n$; and finally, the reason

5968: why property {\bf ($\mathbf{ 8_1}$)} holds true has a simple geometric

5969: interpretation: if the scaling parameter $c_1$ is small enough, the

5970: small analytic disc $A_{x,v:c}^1(\overline{\Delta})$ is essentially a

5971: slightly deformed small part of the straight complex line

5972: $\C\cdot(v_1+Jv_1)$, where the half-wedge $\mathcal{ HW}_1^+$ or the

5973: wedge $\mathcal{ W}_2$ is directed by the vector $Jv_1$ according to

5974: Lemma~5.37. The next paragraphs are devoted to some elementary

5975: estimates which will establish these properties rigorously.

5976: 

5977: Firstly, let us prove property {\bf ($\mathbf{ 7_1}$)} in Case {\bf

5978: ($\mathbf{I_1}$)}. According to Lemma~5.37, the vector $v_1$ is given

5979: by $(0,1,\dots,1)$ and the side $(H^1)^+\subset M^1$ is defined by

5980: $x_1>g(x')=-x_2^2-\cdots-x_n^2+ \widehat{ g}(x')$, where the

5981: $\mathcal{ C}^{2,\alpha}$-smooth function $\widehat{ g}(x')$ vanishes

5982: to second order at the origin, thanks to the normalization

5983: conditions~\thetag{5.40}. By Lemma~6.4, the remainder $\widehat{

5984: g}(x')$ then satisfies an inequality of the form $\vert \widehat{

5985: g}(x') \vert \leq K_9 \cdot \vert x' \vert^{2+\alpha}\leq K_9 \cdot

5986: \left( \sum_{j=2}^n\, x_j^2 \right)^{\frac{ \alpha+2}{2}}$, for some

5987: constant $K_9>0$. Since the strictly concave open subset

5988: $(\widetilde{ H}^1)^+$ of $M^1$ with $\mathcal{ C}^{2,\alpha}$-smooth

5989: boundary defined by the inequality $x_1>-x_1^2-\cdots-x_n^2 + K_9

5990: \cdot \left( \sum_{j=2}^n \, x_j^2 \right)^{\frac{2+\alpha}{2}}$ is

5991: contained in $(H^1)^+$, it suffices to prove property {\bf ($\mathbf{

5992: 7_1}$)} with $(H^1)^+$ replaced by $(\widetilde{ H}^1)^+$.

5993: 

5994: By construction, the disc boundary $A_{0,0:c}(\partial\Delta)$ is

5995: tangent at $p_1$ to $H^1$, hence also to $\widetilde{

5996: H}^1$. Intuitively, it is clear that the blunt disc half-boundary

5997: $A_{0,0:c}(\partial^+\Delta\backslash\{1\})$ should then be contained

5998: in the strictly concave open subset $(\widetilde{ H}^1)^+$, 

5999: {\it see}\, {\sc Figure~16} above. 

6000: 

6001: To proceed rigorously, we shall come back to the definition~\thetag{

6002: 7.53} which yields $A_{0, 0:c}^1( \zeta) \equiv Z_{c,0, v_1+ v(c)}^1

6003: \left( \Phi_c(\zeta)\right)$, with the tangency condition~\thetag{

6004: 7.46} satisfied. First of all, denoting the $n$ components of $v(c)$

6005: by $(v_1(c), \dots, v_n(c))$, we may compute the second order

6006: derivatives of the similar discs attached to $M^0$:

6007: \def\theequation{8.4}\begin{equation}

6008: \left\{

6009: \aligned

6010: \frac{\partial^2 Z_{1;c,0,

6011: v_1+v(c)}^0}{\partial \theta^2}(1)= 

6012: & \

6013: c\cdot \frac{\partial^2\Psi}{

6014: \partial \theta^2}\left(e^{i\theta}\right)\cdot

6015: v_1(c), \\

6016: \frac{\partial^2 Z_{j;c,0,

6017: v_1+v(c)}^0}{\partial \theta^2}(1)= 

6018: & \

6019: c\cdot \frac{\partial^2\Psi}{\partial 

6020: \theta^2}\left(e^{i\theta}\right)\cdot

6021: (1+v_j(c)), \ \ \ \ \ \ \ 

6022: j=2,\dots,n.

6023: \endaligned\right.

6024: \end{equation}

6025: Using the definition~\thetag{ 7.47}, the inequality

6026: $\vert v(c) \vert \leq c^{1+\alpha} \cdot K_8$ and the 

6027: second estimate~\thetag{7.36}, we deduce that

6028: \def\theequation{8.5}\begin{equation}

6029: \left\{

6030: \aligned

6031: \left\vert

6032: \frac{\partial^2 Z_{1;c,0,

6033: v_1+v(c)}^1}{\partial \theta^2}(1)\right\vert \leq

6034: & \

6035: c^{2+\alpha}\cdot K_7+c^{2+\alpha}\cdot C_2 K_8

6036: =: c^{2+\alpha} \cdot 2K_{10} 

6037: \\

6038: \left\vert

6039: \frac{\partial^2 Z_{j;c,0,

6040: v_1+v(c)}^1}{\partial \theta^2}(1)\right\vert \leq

6041: & \

6042: c\cdot 2C_2, 

6043: \ \ \ \ \ \ \

6044: j=2,\dots,n.

6045: \endaligned\right.

6046: \end{equation}

6047: Applying then Taylor's integral formula $F (\theta)=F(0) +\theta \cdot

6048: F'(0)+ \int_0^\theta \, (\theta- \theta') \cdot \partial_\theta

6049: \partial_\theta F(\theta') \cdot d \theta'$ to $F(\theta):= X_{1;c,

6050: 0,v_1 +v(c)}^1\left(e^{i\theta}\right)$ and afterwards to $F(\theta):=

6051: X_{j;c, 0,v_1 +v(c)}^1\left(e^{i\theta}\right)$ for $j=2,\dots,n$,

6052: taking account of the tangency conditions

6053: \def\theequation{8.6}\begin{equation}

6054: \frac{\partial X_{1;c,0,v_1+v(c)}^1}{\partial \theta}(1)=0, 

6055: \ \ \ \ \ \ \ \ \ \ \ \ 

6056: \frac{\partial X_{j;c,0,v_1+v(c)}^1}{\partial \theta}(1)=

6057: c\cdot C_1, \ \ \ \ \ \ \ 

6058: j=2,\dots,n,

6059: \end{equation}

6060: (a simple rephrasing of~\thetag{7.46}) 

6061: and using the inequalities~\thetag{8.5}, 

6062: we deduce that

6063: \def\theequation{8.7}\begin{equation}

6064: \left\{

6065: \aligned

6066: \left\vert

6067: X_{1;c,0,v_1+v(c)}^1\left(e^{i\theta}\right)

6068: \right\vert \leq

6069: & \

6070: \theta^2\cdot c^{2+\alpha} \cdot K_{10}, 

6071: \\

6072: \left\vert

6073: X_{j;c,0,v_1+v(c)}^1\left(e^{i\theta}\right) -\theta \cdot c \cdot C_1

6074: \right\vert \leq

6075: & \

6076: \theta^2 \cdot

6077: c\cdot C_2, 

6078: \ \ \ \ \ 

6079: j=2,\dots,n.

6080: \endaligned\right.

6081: \end{equation}

6082: Recall that

6083: \def\theequation{8.8}\begin{equation}

6084: x_1>\widetilde{ g}(x'):= -x_2^2-\cdots-x_n^2+K_9\left(

6085: \sum_{j=2}^n\, x_j^2

6086: \right)^{\frac{2+\alpha}{2}}

6087: \end{equation}

6088: denotes the equation of $(\widetilde{ H}^1)^+$. We now claim that if

6089: $c_1$ is sufficiently small, then for every $\theta$ with $0<\vert

6090: \theta \vert < 10c$, we have

6091: \def\theequation{8.9}\begin{equation}

6092: X_{1;c,0,v_1+v(c)}^1\left(e^{i\theta}\right) >

6093: \widetilde{ g}\left(

6094: X_{2;c,0,v_1+v(c)}^1\left(e^{i\theta}\right),\dots\dots,

6095: X_{n; c,0,v_1+v(c)}^1\left(e^{i\theta}\right)

6096: \right).

6097: \end{equation}

6098: Since $\Phi_c( \partial^+\Delta)$ is 

6099: contained in $\{e^{ i\theta}\in \partial^+

6100: \Delta: \ \vert \theta \vert <

6101: 10c\}$, this will imply the desired inclusion

6102: for proving {\bf ($\mathbf{ 7_1}$)}:

6103: \def\theequation{8.10}\begin{equation}

6104: \left\{

6105: \aligned

6106: A_{x,v:c}^1(\partial^+\Delta\backslash \{1\})=

6107: & \

6108: Z_{c,0,v_1+v(c)}^1\left(

6109: \Phi_c(\partial^+\Delta\backslash\{1\})

6110: \right)\subset \\

6111: \subset 

6112: & \

6113: Z_{c,0,v_1+v(c)}^1\left(

6114: \{e^{i\theta}\in \partial^+\Delta: \ 0 <\vert \theta \vert \leq 10c\}

6115: \right)

6116: \subset (\widetilde{ H}^1)^+.

6117: \endaligned\right.

6118: \end{equation}

6119: To prove the claim, we notice a minoration 

6120: of the left hand side of~\thetag{8.9}, using~\thetag{8.7}

6121: \def\theequation{8.11}\begin{equation}

6122: X_{1;c,0,v_1+v(c)}^1\left(e^{i\theta}\right) 

6123: \geq -\theta^2\cdot c^{2+\alpha}

6124: \cdot K_{10}.

6125: \end{equation}

6126: On the other hand, using two

6127: inequalities which are direct consequences of

6128: the second line of~\thetag{8.7}, provided that 

6129: $10c_1\cdot C_2\leq \frac{C_1}{2}$:

6130: \def\theequation{8.12}\begin{equation}

6131: \left\{

6132: \aligned

6133: \left\vert

6134: X_{j;c,0,v_1+v(c)}^1\left(e^{i\theta}\right)

6135: \right \vert \leq 

6136: & \

6137: \vert \theta\vert \cdot c \cdot \left(

6138: C_1+ \vert \theta \vert \cdot C_2\right)

6139: \leq \vert \theta \vert 

6140: \cdot c\cdot \frac{ 3C_1}{2}, \\

6141: \left[

6142: X_{j;c,0,v_1+v(c)}^1

6143: \right]^2\geq

6144: & \

6145: \theta^2\cdot c^2\cdot 

6146: (C_1-\vert \theta\vert \cdot C_2)^2 \geq

6147: \theta^2\cdot c^2 \cdot

6148: \frac{ C_1^2}{4}, 

6149: \endaligned\right.

6150: \end{equation}

6151: for $j=2,\dots,n$, we deduce the following

6152: majoration of the 

6153: right hand side of~\thetag{8.9}

6154: \def\theequation{8.13}\begin{equation}

6155: \left\{

6156: \aligned

6157: \widetilde{ g}

6158: & \

6159: \left(

6160: X_{2;c,0,v_1+v(c)}^1\left(e^{i\theta}\right),\dots\dots,

6161: X_{n;c,0,v_1+v(c)}^1\left(e^{i\theta}\right)

6162: \right)= \\

6163: & \

6164: =-\sum_{j=2}^n\, \left[

6165: X_{j;c,0,v_1+v(c)}^1

6166: \right]^2+K_9\left(

6167: \sum_{j=2}^n\, 

6168: \left[

6169: X_{j;c,0,v_1+v(c)}^1\left(e^{i\theta}\right)\right]^2

6170: \right)^{\frac{2+\alpha}{2}} \\

6171: & \

6172: \leq

6173: -\theta^2\cdot c^2 \cdot \frac{C_1^2}{4}

6174: (n-1)+

6175: \vert \theta \vert^{2+\alpha} \cdot c^{2+\alpha} \cdot

6176: \left(\frac{(n-1)9C_1^2}{4}\right)^{\frac{2+\alpha}{2}} K_9 \\

6177: & \

6178: \leq 

6179: -\theta^2\cdot c^2\left(

6180: \frac{ C_1^2}{4}(n-1)-

6181: c^\alpha\cdot \left(\frac{(n-1)9C_1^2}{4}\right)^{

6182: \frac{2+\alpha}{2}} K_9

6183: \right).

6184: \endaligned\right.

6185: \end{equation}

6186: Thanks to the minoration~\thetag{8.11} and to the

6187: majoration~\thetag{8.13}, in order that the inequality~\thetag{8.9}

6188: holds for all $\theta$ with $0< \vert \theta \vert \leq 10c$, it

6189: suffices that the right hand side of~\thetag{8.11} be greater than the

6190: last line of~\thetag{8.13}. By writing this strict inequality

6191: and clearing the factor $\theta^2\cdot c^2$, we see

6192: that it suffices that

6193: \def\theequation{8.14}\begin{equation}

6194: -K_{10}\cdot c^\alpha > -\left( \frac{ C_1^2}{4}

6195: (n-1)-c^\alpha\cdot

6196: \left(

6197: \frac{(n-1)9C_1^2}{4}

6198: \right)^{\frac{2+\alpha}{2}}K_9

6199: \right),

6200: \end{equation}

6201: or equivalently

6202: \def\theequation{8.15}\begin{equation}

6203: c_1 <

6204: \left(

6205: \frac{

6206: \frac{ C_1^2}{4}(n-1)

6207: }{

6208: K_{10}+\left(\frac{ (n-1)9C_1^2}{4}\right)^{

6209: \frac{2+\alpha}{2}} K_9

6210: }

6211: \right)^{\frac{1}{\alpha}}.

6212: \end{equation}

6213: This completes the proof of property 

6214: {\bf ($\mathbf{7_1}$)}.

6215: 

6216: Secondly, let us prove property {\bf ($\mathbf{ 8_1}$)} in Case {\bf

6217: ($\mathbf{I_1}$)}, proceeding similarly. As above, we come back to

6218: the definition $A_{x, 0 :c}^1 (\zeta):= Z_{c, x,v_1 +v(c)}^1 \left(

6219: \Phi_c (\zeta) \right)$ and we remind that $A_{x, 0: c}^1 (1)= Z_{c,

6220: x,v_1+ v(c)}^1(1)= x+ih(x)$, which follows by putting $d=1$ and

6221: $\zeta=1$ in~\thetag{ 7.18}. Thanks to the inclusion $\Phi_c(

6222: \partial^+ \Delta)\subset \{e^{i\theta}\in \partial^+\Delta : \ \vert

6223: \theta \vert < 10c\}$, it suffices to prove that the segment $Z_{c, x,

6224: v_1+v(c)}\left( \left\{ e^{i\theta}: \ \vert \theta \vert < 10c

6225: \right\} \right)$ is contained in the open side $( \widetilde{ H}^1)^+

6226: \subset (H^1 )^+$ defined by the inequation~\thetag{8.8}, if the point

6227: $x+ i h(x)$ belongs to the transversal half-submanifold $T^1 \cap

6228: (H^1)^+$, namely if $x=(x_1, 0, \dots,0)$ with $x_1>0$. In the sequel,

6229: we shall denote the disc $Z_{c,x, v_1+ v(c)}^1(\zeta)$ by $Z_{c, x_1,

6230: x',v_1+ v(c) }^1( \zeta)$, emphasizing the decomposition $x= (x_1,

6231: x')\in \R \times \R^{n-1}$, and we shall also use the convenient

6232: notation

6233: \def\theequation{8.16}\begin{equation}

6234: Z_{c,x_1,x',v_1+v(c)}^{'1}\left(\rho e^{i\theta}\right):=

6235: \left(

6236: Z_{2;c,x_1,x',v_1+v(c)}^1\left(\rho e^{i\theta}\right),\dots\dots, 

6237: Z_{n;c,x_1,x',v_1+v(c)}^1\left(\rho e^{i\theta}\right)

6238: \right).

6239: \end{equation}

6240: So, we have to show that for all $c$ with $0 < c \leq c_1$, for all

6241: $x_1$ with $0 < x_1 \leq c^2$ and for all $\theta$ with $\vert \theta

6242: \vert < 10c$, then the following strict inequality holds true

6243: \def\theequation{8.17}\begin{equation}

6244: X_{1;c,x_1,0,v_1+v(c)}^1\left(e^{i\theta}\right) > 

6245: \widetilde{ g}\left(

6246: X_{c;x_1,0,v_1+v(c)}^{1'}\left(e^{i\theta}\right)

6247: \right),

6248: \end{equation}

6249: First of all, coming back to the family of discs attached to $M^0$, we

6250: see by differentiating~\thetag{ 7.8} twice with respect to $x_1$ that

6251: $\frac{\partial^2 Z_{ c,x_1,0,v_1+v(c)}^0}{\partial x_1^2}(\zeta)

6252: \equiv 0$. Next, by differentiating twice Bishop's equation~\thetag{

6253: 7.18} with respect to $x_1$ and by reasoning as in Lemma~7.34, we

6254: deduce the estimate

6255: \def\theequation{8.18}\begin{equation}

6256: \left\vert

6257: \left\vert

6258: \frac{\partial^2 Z_{c,x_1,0,v_1+v(c)}^1}{

6259: \partial x_1^2}

6260: \right\vert

6261: \right\vert_{\mathcal{ C}^\alpha(\partial \Delta)}

6262: \leq c^{2+\alpha} \cdot K_7,

6263: \end{equation}

6264: say, with the same constant $K_7>0$ as in Lemma~7.34, after enlarging

6265: it if necessary. Applying then Taylor's integral formula $F(x_1)=

6266: F(0) +x_1 \cdot \partial_{x_1}F(0)+ \int_0^{ x_1}(x_1- \widetilde{

6267: x}_1) \cdot \partial_{x_1} \partial_{ x_1} F( \widetilde{ x}_1) \cdot

6268: d\widetilde{ x}_1$ to the function $F(x_1 ):= X_{1;c, x_1,0, v_1+

6269: v(c)}^1 \left(e^{ i\theta}\right)$, we deduce the minoration

6270: \def\theequation{8.19}\begin{equation}

6271: X_{1;c,x_1,0,v_1+v(c)}^1\left(e^{i\theta}\right)\geq 

6272: X_{1;c,0,0,v_1+v(c)}^1\left(e^{i\theta}\right)+

6273: x_1 \cdot

6274: \frac{\partial X_{1;c,0,0,v_1+v(c)}^1}{

6275: \partial x_1}\left(e^{i\theta}\right)

6276: - x_1^2 \cdot c^{2+\alpha} \cdot \frac{K_7}{2}.

6277: \end{equation}

6278: On the other hand, by differentiating Bishop's equation~\thetag{7.18}

6279: with respect to $x_1$ at $x=0$, the derivative $\partial_{x_1} x$

6280: yields the vector $(1,0,\dots,0)$ and we obtain

6281: \def\theequation{8.20}\begin{equation}

6282: \small

6283: \left\{

6284: \aligned

6285: \frac{\partial X_{c,0,0,v_1+v(c)}}{\partial x_1}

6286: \left(e^{i\theta} \right)=

6287: & \

6288: -T_1\left[

6289: \sum_{l=1}^n\, 

6290: \frac{\partial h}{\partial x_l}

6291: \left(

6292: X_{c,0,0,v_1+v(c)}^1(\cdot)

6293: \right) \, 

6294: \frac{\partial X_{l;c,0,0,v_1+v(c)}^1}{\partial x_1}(\cdot)

6295: \right](e^{i\theta})+ \\

6296: & \ \

6297: +

6298: (1,0,\dots,0).

6299: \endaligned\right.

6300: \end{equation}

6301: Using then the second inequality~\thetag{ 6.13} and

6302: the estimate~\thetag{ 7.25}, we deduce from~\thetag{ 8.20}

6303: \def\theequation{8.21}\begin{equation}

6304: \left\{

6305: \aligned

6306: \left\vert\left\vert

6307: \frac{\partial X_{1;c,0,0,v_1+v(c)}^1}{\partial x_1}(\cdot)-1

6308: \right\vert\right\vert_{

6309: \mathcal{ C}^\alpha(\partial \Delta)}\leq

6310: & \

6311: c^{2+\alpha} \cdot \vert\vert\vert

6312: T_1 \vert \vert \vert_{\mathcal{ C}^\alpha(\partial \Delta)} 

6313: K_2 K_3, \\

6314: \left\vert\left\vert

6315: \frac{\partial X_{j;c,0,0,v_1+v(c)}^1}{\partial x_1}(\cdot)

6316: \right\vert\right\vert_{

6317: \mathcal{ C}^\alpha(\partial \Delta)}\leq

6318: & \

6319: c^{2+\alpha} \cdot \vert\vert\vert

6320: T_1 \vert \vert \vert_{\mathcal{ C}^\alpha(\partial \Delta)} 

6321: K_2 K_3, \ \ \ \ \ \ \ 

6322: j=2,\dots,n.

6323: \endaligned\right.

6324: \end{equation}

6325: Thanks to the first line of~\thetag{8.21}, we can refine the

6326: minoration~\thetag{8.19} by replacing the first order partial

6327: derivative $\frac{ \partial X_{1;c, 0,0, v_1+v(c)}^1 }{ \partial x_1

6328: }\left (e^{i \theta} \right)$ in the right hand side of~\thetag{8.19}

6329: by the constant $1$, modulo an error term and also, we can use the

6330: trivial minoration $-x_1^2\geq -x_1$, which yields a new, more

6331: interesting minoration of the form

6332: \def\theequation{8.22}\begin{equation}

6333: X_{1;c,x_1,0,v_1+v(c)}^1\left(e^{i\theta}\right)\geq 

6334: X_{1;c,0,0,v_1+v(c)}^1\left(e^{i\theta}\right)+

6335: x_1 - x_1 \cdot c^{2+\alpha} \cdot K_{11}, 

6336: \end{equation}

6337: for some constant $K_{ 11}>0$. On the other hand, using the

6338: inequalities $\vert \partial_{ x_j} \widetilde{ g} (x') \vert \leq

6339: \vert x' \vert+ K_9 \cdot \vert x' \vert^{1+\alpha} \cdot \left(1+

6340: \frac{ \alpha }{2}\right) (n-1)^{ \frac{ \alpha }{2}}$ for $j=

6341: 2,\dots,n$, using the estimate~\thetag{ 7.25} and using~\thetag{ 6.2},

6342: we deduce an inequality of the form

6343: \def\theequation{8.23}\begin{equation}

6344: \widetilde{ g}\left(

6345: X_{c,x_1,0,v_1+v(c)}^{'1}\left(e^{i\theta}\right)

6346: \right)\leq 

6347: \widetilde{ g}\left(

6348: X_{c,0,0,v_1+v(c)}^{'1}\left(e^{i\theta}\right)

6349: \right)+ x_1 \cdot c \cdot K_{12}, 

6350: \end{equation}

6351: for some constant $K_{12}>0$. Finally, putting together the two

6352: inequalities~\thetag{ 8.22} and~\thetag{ 8.23}, and using the

6353: following inequality, which is an immediate consequence

6354: of the strict inequality~\thetag{ 8.9}:

6355: \def\theequation{8.24}\begin{equation}

6356: X_{1;c,0,0,v_1+v(c)}\left(e^{i\theta}\right)

6357: \geq \widetilde{ g}\left(

6358: X_{c,0,0,v_1+v(c)}^{'1}\left(

6359: e^{i\theta}

6360: \right)

6361: \right),

6362: \end{equation}

6363: valuable for all $\theta$ with $\vert \theta \vert < 10c$, we deduce

6364: the desired inequality~\thetag{ 8.17} as follows:

6365: \def\theequation{8.25}\begin{equation}

6366: \left\{

6367: \aligned

6368: X_{1;c,x_1,0,v_1+v(c)}^1\left(e^{i\theta} \right) 

6369: \geq 

6370: & \

6371: X_{1;c,0,0,v_1+v(c)}^1\left(e^{i\theta} \right)+

6372: x_1-x_1\cdot c^{2+\alpha} \cdot K_{11} \\

6373: \geq 

6374: & \ \widetilde{ g}\left(

6375: X_{c,0,0,v_1+v(c)}^{'1}\left(

6376: e^{i\theta}

6377: \right)

6378: \right)+ x_1-x_1\cdot c^{2+\alpha} \cdot K_{11} \\

6379: \geq 

6380: & \

6381: \widetilde{ g}\left(

6382: X_{c,x_1,0,v_1+v(c)}^{'1}\left(

6383: e^{i\theta}

6384: \right)

6385: \right)+ x_1-x_1\cdot c \cdot K_{11}-x_1\cdot c \cdot K_{12} \\

6386: > 

6387: & \

6388: \widetilde{ g}\left(

6389: X_{c,x_1,0,v_1+v(c)}^{'1}\left(

6390: e^{i\theta}

6391: \right)

6392: \right),

6393: \endaligned\right.

6394: \end{equation}

6395: for all $x_1$ with $0< x_1 \leq c^2$, all $\theta$ 

6396: with $\vert \theta \vert < 10c$ and all $c$ with 

6397: $0 < c \leq c_1$, 

6398: provided

6399: \def\theequation{8.26}\begin{equation}

6400: c_1 \leq \frac{ 1/2}{K_{11}+K_{12}}.

6401: \end{equation}

6402: This completes the proof of property {\bf ($\mathbf{ 8_1}$)}.

6403: 

6404: Thirdly, let us prove property {\bf ($\mathbf{ 9_1}$)} in Case {\bf

6405: ($\mathbf{I_1}$)}. The half-wedge $\mathcal{ HW}_1^+$ is defined by

6406: the $n$ inequalities of the last two lines of~\thetag{5.38}, where

6407: $a_2+\cdots+a_n=1$. For notational reasons, it will be convenient to

6408: set $a_1:=1$ and to write the first inequality defining $\mathcal{

6409: HW}_1^+$ simply as $\sum_{j=1}^n\, a_j y_j> \psi(x,y')$.

6410: 

6411: Because $\Phi_c \left( \overline{ \Delta} \backslash \partial^+ \Delta

6412: \right)$ is contained in the open sector $\{\rho e^{i \theta} \in

6413: \overline{ \Delta}: \ \vert \theta \vert < 10c, \ 1-10c < \rho < 1$,

6414: taking account of the definition~\thetag{7.53} of $A_{x,v

6415: :c}^1(\zeta)$, in order to check property {\bf ($\mathbf{ 9_1}$)}, it

6416: clearly suffices to show that $Z_{c,x, v_1+ v(c)+ v}^1 \left( \left\{

6417: \rho e^{i\theta}\in \Delta : \ 1-10c < \rho < 1, \ \vert \theta \vert

6418: < 10c \right\} \right)$ is contained in $\mathcal{ HW}_1^+$, which

6419: amounts to establish that for all $x$ with $\vert x \vert \leq c^2$,

6420: all $v$ with $\vert v \vert \leq c$, all $\rho e^{i \theta}$ with

6421: $1-10c < \rho < 1$ and with $\vert \theta \vert < 10c$, the following

6422: two collections of strict inequalities hold true

6423: \def\theequation{8.27}\begin{equation}

6424: \left\{

6425: \aligned

6426: \sum_{k=1}^n\, a_k 

6427: Y_{j;c,x,v_1+v(c)+v}^1\left(\rho e^{i\theta}\right) 

6428: > 

6429: & \

6430: \psi\left(

6431: X_{c,x,v_1+v(c)+v}^1\left(\rho e^{i\theta}\right), \

6432: Y_{c,x,v_1+v(c)+v}^{'1}\left(\rho e^{i\theta}\right)

6433: \right), \\

6434: Y_{j; c,x,v_1+v(c)+v}^1\left(

6435: \rho e^{i\theta}

6436: \right) > 

6437: & \

6438: \varphi_j\left(

6439: X_{c,x,v_1+v(c)+v}^1\left(

6440: \rho e^{i\theta}

6441: \right), \ 

6442: Y_{1;c,x,v_1+v(c)+v}^1\left(

6443: \rho e^{i\theta}

6444: \right)

6445: \right),

6446: \endaligned\right.

6447: \end{equation}

6448: for $j=2,\dots,n$, 

6449: provided $c_1$ is sufficiently small, where we use the

6450: notation~\thetag{ 8.16}.

6451: 

6452: We first treat the collection of $(n-1)$ strict inequalities

6453: in the second line of~\thetag{ 8.27}.

6454: First of all, by differentiating~\thetag{ 7.8}

6455: twice with respect to $\theta$, we obtain

6456: \def\theequation{8.28}\begin{equation}

6457: \frac{\partial^2 Z_{c,x,v_1+v(c)+v}^0}{\partial \theta^2}\left(

6458: e^{i\theta}

6459: \right)= c\cdot \frac{

6460: \partial^2\Psi}{\partial \theta^2}

6461: \left(

6462: e^{i\theta}

6463: \right) \cdot

6464: \left[

6465: v_1+v(c)+v

6466: \right].

6467: \end{equation}

6468: Using the second estimate~\thetag{7.36}, we deduce that

6469: there exists a constant $K_{13}>0$ such that

6470: \def\theequation{8.29}\begin{equation}

6471: \left\vert

6472: \frac{\partial^2 Z_{c,x,v_1+v(c)+v}^1}{

6473: \partial \theta^2}\left(e^{i\theta}\right)

6474: \right\vert\leq c\cdot K_{13}.

6475: \end{equation}

6476: Using the inequality~\thetag{ 6.2}, using~\thetag{ 8.29}, and then

6477: taking account of the inequalities $\vert \theta \vert < 10c$, $\vert

6478: x \vert \leq c^2$ and $\vert v \vert < c$, we deduce the following

6479: inequality

6480: \def\theequation{8.30}\begin{equation}

6481: \left\{

6482: \aligned

6483: \left\vert

6484: \frac{\partial Z_{c,x,v_1+v(c)+v}^1}{\partial \theta}

6485: \left( e^{i\theta}

6486: \right)-

6487: \frac{\partial Z_{c,0,v_1+v(c)}^1}{\partial \theta} (1)

6488: \right\vert

6489: \leq 

6490: & \

6491: c \cdot \left(

6492: \vert \theta \vert + \vert x \vert + \vert v \vert

6493: \right) \\

6494: \leq 

6495: & \

6496: c^2\cdot K_{14},

6497: \endaligned\right.

6498: \end{equation}

6499: for some constant $K_{14} >0$. On the other hand, by

6500: differentiating~\thetag{ 7.8} with respect to $\theta$ at $\theta=0$

6501: and applying the inequality~\thetag{ 7.28}, we obtain

6502: \def\theequation{8.31}\begin{equation}

6503: \left\vert

6504: \frac{\partial Z_{c,0,v_1+v(c)}^1}{\partial \theta} (1)

6505: -c\cdot C_1\cdot (0,1,\dots,1)

6506: \right\vert\leq c^{2+\alpha}\cdot K_6, 

6507: \end{equation}

6508: where $C_1=\frac{ \partial \Psi }{ \partial \theta}(1)$, as defined

6509: in~\thetag{7.47}. We remind that for every $\mathcal{ C}^1$-smooth

6510: function $Z$ on $\overline{ \Delta}$ which is holomorphic in $\Delta$,

6511: we have $i \frac{ \partial }{ \partial \theta} Z ( e^{i\theta})=

6512: -\frac{ \partial }{\partial \rho} Z( e^{i\theta})$. Consequently, we

6513: deduce from~\thetag{ 8.30} the following first (among three)

6514: interesting inequality

6515: \def\theequation{8.32}\begin{equation}

6516: \left\vert

6517: -\frac{\partial Z_{c,x,v_1+v(c)+v}^1}{\partial \rho}

6518: \left(

6519: e^{i\theta}

6520: \right)- c\cdot C_1 \cdot (0,i,\dots,i)

6521: \right\vert\leq 

6522: c^2 \cdot K_{15}, 

6523: \end{equation}

6524: for some constant $K_{15}>0$.

6525: 

6526: According to the definition~\thetag{ 7.8}, 

6527: we may compute 

6528: \def\theequation{8.33}\begin{equation}

6529: \frac{ \partial^2 Z_{c,x,v_1+v(c)+v}^0}{\partial \rho^2}

6530: \left(

6531: \rho e^{i\theta}\right)= c\cdot 

6532: \frac{\partial^2 \Psi}{\partial \rho^2}

6533: \left(

6534: \rho e^{i\theta} 

6535: \right)\cdot

6536: (v_1+v(c)+v)

6537: \end{equation}

6538: By reasoning as in the proof of Lemma~7.34, we may obtain an

6539: inequality similar to~\thetag{ 7.36}, with the second order partial

6540: derivative $\partial^2/ \partial \theta^2$ replaced by the second

6541: order partial derivative $\partial^2 / \partial \rho^2$. Putting this

6542: together with~\thetag{ 8.33}, we deduce that there exists a constant

6543: $K_{16} >0$ such that

6544: \def\theequation{8.34}\begin{equation}

6545: \left\vert

6546: \frac{\partial^2 Z_{c,x,v_1+

6547: v(c)+v}^1}{\partial \rho^2}

6548: \left(

6549: \rho e^{i\theta}

6550: \right)

6551: \right\vert\leq 

6552: c \cdot 2K_{16},

6553: \end{equation}

6554: for some constant $K_{16}>0$. Applying then Taylor's integral formula

6555: $F (\rho) = F(1) + (\rho -1) \cdot \partial_\rho F(1)+ \int_1^\rho (

6556: \rho- \widetilde{ \rho})\cdot \partial_\rho \partial_\rho F(

6557: \widetilde{ \rho}) \cdot d\widetilde{ \rho}$ to the functions $F(

6558: \rho):= Y_{k;c, x,v_1 +v(c) +v}^1\left( \rho e^{ i\theta} \right)$ for

6559: $k=1,\dots,n$, we deduce the second interesting collection of

6560: inequalities

6561: \def\theequation{8.35}\begin{equation}

6562: \left\{

6563: \aligned

6564: {}

6565: &

6566: \left\vert

6567: Y_{k;c,x,v_1+v(c)+v}^1

6568: \left(

6569: \rho e^{i\theta}

6570: \right)

6571: -Y_{k;c,x,v_1+v(c)+v}^1\left(

6572: e^{i\theta} 

6573: \right) - \right. \\

6574: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ 

6575: \left.

6576: -(\rho-1)\cdot

6577: \frac{\partial Y_{k;c,x,v_1+v(c)+v}^1}{\partial \rho }

6578: \left(

6579: e^{i\theta}

6580: \right)

6581: \right\vert 

6582: \leq (1-\rho)^2 \cdot c \cdot K_{16},

6583: \endaligned\right.

6584: \end{equation}

6585: for $k=1,\dots,n$.

6586: 

6587: On the other hand, thanks to the normalizations of the functions

6588: $\varphi_j (x, y_1)$ given in~\thetag{ 5.40}, namely $\varphi_j (0)=

6589: \partial_{ x_k} \varphi_j (0)= \partial_{ y_1} \varphi_j (0)=0$, $j=2,

6590: \dots, n$, $k=1, \dots, n$, we see that, possibly after increasing the

6591: constant $K_1 >0$ of Lemma~6.4, we have inequalities of the form

6592: \def\theequation{8.36}\begin{equation}

6593: \left\{

6594: \aligned

6595: {}

6596: &

6597: \sum_{k=1}^n\, 

6598: \left\vert

6599: \varphi_{j,x_k}(x,y_1)

6600: \right\vert+

6601: \left\vert

6602: \varphi_{j,y_1}(x,y_1)

6603: \right\vert\leq (\vert x \vert 

6604: + \vert y_1 \vert) \cdot

6605: K_1, \\

6606: &

6607: \left\vert \varphi_j(x,y_1) - 

6608: \varphi_j\left(\widetilde{ x}, 

6609: \widetilde{ y}_1\right)

6610: \right\vert \leq 

6611: \left(

6612: \left\vert 

6613: x-\widetilde{ x}

6614: \right\vert+

6615: \left\vert

6616: y_1- \widetilde{ y}_1 

6617: \right\vert

6618: \right)\cdot

6619: \left(

6620: \sum_{k=1}^n\, 

6621: \sup_{\vert x \vert, \, 

6622: \vert y_1 \vert \leq c\cdot K_2}

6623: \left\vert

6624: \varphi_{j,x_k}(x,y_1)

6625: \right\vert+ \right. \\

6626: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

6627: \left. 

6628: +\sup_{\vert x \vert, \, 

6629: \vert y_1 \vert \leq c\cdot K_2}

6630: \left\vert

6631: \varphi_{j,y_1}(x,y_1)

6632: \right\vert

6633: \right),

6634: \endaligned\right.

6635: \end{equation}

6636: for $j=2, \dots, n$, provided $\vert x\vert, \, \vert \widetilde{ x}

6637: \vert, \, \vert y_1 \vert, \, \vert \widetilde{ y_1} \vert \leq c\cdot

6638: K_2$. On the other hand, computing $\frac{ \partial Z_{c,x, v_1+

6639: v(c)+v}^0}{\partial \rho} \left( \rho e^{ i\theta} \right)$

6640: in~\thetag{ 7.8}, using~\thetag{ 7.25}, \thetag{ 7.28} and an

6641: inequality of the form~\thetag{ 6.2}, we deduce that there exists a

6642: constant $K_{17}>0$ such that

6643: \def\theequation{8.37}\begin{equation}

6644: \left\vert

6645: Z_{c,x,v_1+v(c)+v}^1\left(

6646: \rho e^{i\theta} 

6647: \right)-

6648: Z_{c,x,v_1+v(c)+v}^1\left(

6649: e^{i\theta} 

6650: \right) 

6651: \right\vert\leq

6652: (1-\rho) \cdot c \cdot K_{17}.

6653: \end{equation}

6654: Finally, using the inequality $\vert Z_{ c,x, v_1+ v(c)+v}^1\left(

6655: \rho e^{i \theta} \right) \vert \leq c \cdot K_2$ obtained in~\thetag{

6656: 7.25}, using the collection of inequalities~\thetag{ 8.36} and using

6657: the inequality~\thetag{ 8.37}, we may deduce the third (and last)

6658: interesting inequality for $j=2, \dots, n$:

6659: \def\theequation{8.38}\begin{equation}

6660: \left\{

6661: \aligned

6662: {}

6663: &

6664: \left\vert

6665: \varphi_j\left(

6666: X_{c,x, v_1+v(c)+v}^1\left(

6667: \rho e^{i\theta}

6668: \right), \

6669: Y_{1;c,x, v_1+v(c)+v}^1

6670: \left(\rho e^{i\theta} \right)

6671: \right)- \right. \\

6672: &

6673: \ \ \ \ \ \ \ \ \ \ \ \ \ 

6674: \left.

6675: -\varphi_j\left(

6676: X_{c,x, v_1+v(c)+v}^1\left(

6677: e^{i\theta}

6678: \right), \

6679: Y_{1;c,x, v_1+v(c)+v}^1

6680: \left(e^{i\theta} \right)

6681: \right)

6682: \right\vert\leq \\

6683: &

6684: \left(

6685: \left\vert

6686: X_{c,x,v_1+v(c)+v}^1\left(

6687: \rho e^{i\theta}

6688: \right)- X_{c,x,v_1+v(c)+v}^1\left(

6689: e^{i\theta}\right)

6690: \right\vert+ \right. \\

6691: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

6692: \left.

6693: +\left\vert

6694: Y_{1;c,x, v_1+v(c)+v}^1\left(

6695: \rho e^{i\theta}

6696: \right)- Y_{1;c,x, v_1+v(c)+v}^1\left(

6697: e^{i\theta}\right)

6698: \right\vert

6699: \right)\cdot \\

6700: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

6701: \cdot 

6702: \left(

6703: \sum_{k=1}^n\, 

6704: \sup_{\vert x \vert, \, 

6705: \vert y_1 \vert \leq c\cdot K_2}

6706: \left\vert

6707: \varphi_{j,x_k}(x,y_1)

6708: \right\vert

6709: +\sup_{\vert x \vert, \, 

6710: \vert y_1 \vert \leq c\cdot K_2}

6711: \left\vert

6712: \varphi_{j,y_1}(x,y_1)

6713: \right\vert

6714: \right) \leq \\

6715: & 

6716: \leq (1-\rho) \cdot c^2 \cdot K_{18}, 

6717: \endaligned\right.

6718: \end{equation}

6719: for some constant $K_{18}>0$.

6720: 

6721: We can now complete the proof of the collection of inequalities in the

6722: second line of~\thetag{ 8.27}. As before, 

6723: let $c$ with $0 < c \leq c_1$, let $\rho$ with $10c < \rho

6724: < 1$, let $\theta$ with $\vert \theta \vert < 10 c$, 

6725: let $x$ with $\vert x \vert \leq c^2$, let $v$ with

6726: $\vert v \vert \leq c$ and let $j=2, \dots, n$. Starting

6727: with~\thetag{ 8.35}, using~\thetag{ 8.32}, using the fact that

6728: $Z_{c,x,v_1+ v(c)+v}^1 \left( \partial^+\Delta \right) \subset M^1

6729: \subset M$ and using~\thetag{ 8.38}, we have

6730: \def\theequation{8.39}\begin{equation}

6731: \left\{

6732: \aligned

6733: {}

6734: &

6735: Y_{j;c,x,v_1+v(c)+v}^1\left(

6736: \rho e^{i\theta} 

6737: \right) \geq \\

6738: & \ \ \ \

6739: \geq Y_{j;c,x,v_1+v(c)+v}^1

6740: \left(

6741: e^{i\theta} 

6742: \right)+ \\

6743: & \ \ \ \ \ \ \ 

6744: +(\rho-1) \cdot 

6745: \frac{\partial Y_{j;c,x,

6746: v_1+v(c)+v}^1}{\partial

6747: \rho}\left(

6748: e^{i\theta}

6749: \right)- (1-\rho)^2\cdot c \cdot K_{16} \geq \\

6750: & \ \ \ \ \

6751: \geq 

6752: Y_{j;c,x,v_1+v(c)+v}^1

6753: \left(

6754: e^{i\theta} 

6755: \right)+

6756: (1-\rho)\cdot

6757: c\cdot C_1

6758: -(1-\rho) \cdot c^2 \cdot K_{15} - 

6759: (1-\rho)^2\cdot c \cdot K_{16} \\

6760: & \ \ \ \ \

6761: = 

6762: \varphi_j\left(

6763: X_{1;c,x,v_1+v(c)+v}^1\left(

6764: e^{i\theta} 

6765: \right), \

6766: Y_{1;c,x,v_1+v(c)+v}^1\left(

6767: e^{i\theta}

6768: \right)

6769: \right)+

6770: (1-\rho)\cdot

6771: c\cdot C_1 - \\

6772: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ 

6773: \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

6774: \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

6775: \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

6776: -(1-\rho)\cdot c^2 

6777: \cdot K_{15} - (1-\rho)^2\cdot c \cdot K_{16} \\

6778: & \ \ \ \ \ 

6779: \geq 

6780: \varphi_j\left(

6781: X_{1;c,x,v_1+v(c)+v}^1\left(

6782: \rho e^{i\theta} 

6783: \right), \

6784: Y_{1;c,x,v_1+v(c)+v}^1\left(

6785: \rho e^{i\theta}

6786: \right)

6787: \right)+

6788: (1-\rho)\cdot

6789: c\cdot C_1 - \\

6790: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ 

6791: \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

6792: \ \ \ \ \

6793: -(1-\rho) \cdot c^2 \cdot K_{15} - (1-\rho)^2\cdot c \cdot K_{16}-

6794: (1-\rho)\cdot c^2\cdot K_{18} \\

6795: & \ \ \ \ \ 

6796: \geq 

6797: \varphi_j\left(

6798: X_{1;c,x,v_1+v(c)+v}^1\left(

6799: \rho e^{i\theta} 

6800: \right), \

6801: Y_{1;c,x,v_1+v(c)+v}^1\left(

6802: \rho e^{i\theta}

6803: \right)

6804: \right)+

6805: (1-\rho)\cdot

6806: c\cdot \left[C_1 - \right.\\

6807: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ 

6808: \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

6809: \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

6810: \ \ \ \ \ \ \ \ \ \

6811: \left.

6812: -c \cdot K_{15} - 10c \cdot K_{16}-

6813: c\cdot K_{18}\right]\\

6814: & \ \ \ \ \

6815: \geq 

6816: \varphi_j\left(

6817: X_{1;c,x,v_1+v(c)+v}^1\left(

6818: \rho e^{i\theta} 

6819: \right), \

6820: Y_{1;c,x,v_1+v(c)+v}^1\left(

6821: \rho e^{i\theta}

6822: \right)

6823: \right)+

6824: (1-\rho)\cdot

6825: c\cdot \frac{ C_1}{2} \\

6826: & \ \ \ \ \ 

6827: > 

6828: \varphi_j\left(

6829: X_{1;c,x,v_1+v(c)+v}^1\left(

6830: \rho e^{i\theta} 

6831: \right), \

6832: Y_{1;c,x,v_1+v(c)+v}^1\left(

6833: \rho e^{i\theta}

6834: \right)

6835: \right),

6836: \endaligned\right.

6837: \end{equation}

6838: provided that

6839: \def\theequation{8.40}\begin{equation}

6840: c_1 \leq 

6841: \frac{ C_1/2}{K_{15}+10K_{16}+K_{18}}.

6842: \end{equation}

6843: This yields the collection of

6844: inequalities in the second line of~\thetag{ 8.27}.

6845: 

6846: For the first inequality~\thetag{ 8.27}, we proceed similarly. Recall

6847: that $v_1= (0,1, \dots,1)$, that $a_1=1$ and that $a_2+ \cdots+a_n=1$.

6848: Since $Z_{c, x, v_1+ v(c)+v }^1( \partial^+\Delta)\subset M^1 \subset

6849: N^1$, we have for all $\theta$ with $\vert \theta \vert \leq

6850: \frac{\pi}{2}$ the following relation

6851: \def\theequation{8.41}\begin{equation}

6852: \sum_{k=1}^n\, 

6853: a_k\, 

6854: Y_{k;c,x,v_1+v(c)+v}^1\left(

6855: e^{i\theta}

6856: \right)= \psi\left(

6857: X_{c,x,v_1+v(c)+v}^1\left(

6858: e^{i\theta}

6859: \right), \ 

6860: Y_{c,x,v_1+v(c)+v}^{'1} \left(

6861: e^{i\theta}\right)

6862: \right).

6863: \end{equation}

6864: Using that $\psi$ vanishes to order one at the origin by the

6865: normalization conditions~\thetag{ 5.40} and proceeding as in the

6866: previous paragraph concerning the functions $\varphi_j$, we obtain an

6867: inequality similar to~\thetag{ 8.38}:

6868: \def\theequation{8.42}\begin{equation}

6869: \left\{

6870: \aligned

6871: {}

6872: &

6873: \left\vert

6874: \psi\left(X_{c,x,v_1+v(c)+v}^1\left(

6875: \rho e^{i\theta} 

6876: \right), \

6877: Y_{c,x,v_1+v(c)+v}^{'1}\left(

6878: \rho e^{i\theta}

6879: \right)\right)- \right . \\

6880: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

6881: \left.

6882: -\psi\left(X_{c,x,v_1+v(c)+v}^1\left(

6883: e^{i\theta} 

6884: \right), \

6885: Y_{c,x,v_1+v(c)+v}^{'1}\left(

6886: e^{i\theta}

6887: \right)\right)

6888: \right\vert\leq 

6889: (1-\rho)\cdot c^2 \cdot K_{19}, 

6890: \endaligned\right.

6891: \end{equation}

6892: for some constant $K_{19}>0$.

6893: 

6894: As before, let $c$ with $0 < c \leq c_1$, let $\rho$ with $10c < \rho

6895: < 1$, let $\theta$ with $\vert \theta \vert < 10 c$, let $x$ with

6896: $\vert x \vert \leq c^2$ and let $v$ with $\vert v \vert \leq c$. Using

6897: then~\thetag{ 8.35}, \thetag{ 8.32}, \thetag{ 8.41} and~\thetag{

6898: 8.42}, we deduce the desired strict inequality

6899: \def\theequation{8.43}\begin{equation}

6900: \small

6901: \left\{

6902: \aligned

6903: {}

6904: &

6905: \sum_{k=1}^n\, 

6906: a_k \, 

6907: Y_{k; c,x,v_1+v(c)+v}^1\left(

6908: \rho e^{i\theta}

6909: \right)\geq 

6910: \sum_{k=1}^n\, 

6911: a_k \, 

6912: Y_{k; c,x,v_1+v(c)+v}^1\left(

6913: e^{i\theta}

6914: \right) +\\

6915: & \ \ \ \ \ \ \ \ 

6916: +(1-\rho) \left[

6917: \sum_{k=1}^n\, 

6918: a_k\left(

6919: -\frac{\partial Y_{k;c,x,v_1+v(c)+v}^1}{\partial

6920: \rho}\left(

6921: e^{i\theta}

6922: \right)

6923: \right)

6924: \right]-

6925: (1-\rho)^2 \cdot c \cdot \left( \sum_{k=1}^n\, a_k\right)K_{16} \\

6926: & \ \ \ \ \ \ \ \ \ 

6927: \geq 

6928: \sum_{k=1}^n\, 

6929: a_k \, 

6930: Y_{k; c,x,v_1+v(c)+v}^1\left(

6931: e^{i\theta}

6932: \right) +

6933: (1-\rho) \left[

6934: \sum_{j=2}^n\,

6935: a_j\cdot 

6936: c\cdot C_1 - \sum_{k=1}^n\, a_k \cdot c^2 \cdot

6937: K_{15} 

6938: \right] - \\

6939: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ 

6940: \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

6941: -(1-\rho)^2\cdot c \cdot 2K_{16} \\

6942: & \ \ \ \ \ \ \ \ \ 

6943: \geq 

6944: \sum_{k=1}^n\, 

6945: a_k \, 

6946: Y_{k; c,x,v_1+v(c)+v}^1\left(

6947: e^{i\theta}

6948: \right) +

6949: (1-\rho)\cdot c \cdot C_1- \\

6950: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

6951: \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ 

6952: \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

6953: -(1-\rho)\cdot c^2

6954: \cdot 2K_{15}- (1-\rho)^2 \cdot c \cdot 2K_{16} \\

6955: & \ \ \ \ \ \ \ \ \ 

6956: =

6957: \psi\left(

6958: X_{c,x,v_1+v(c)+v}^1\left(

6959: e^{i\theta}

6960: \right), \ 

6961: Y_{c,x,v_1+v(c)+v}^{'1} \left(

6962: e^{i\theta}\right)

6963: \right) + \\

6964: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

6965: \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ 

6966: \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

6967: + (1-\rho) \cdot c \cdot \left[

6968: C_1-c \cdot 2K_{15}- 10c \cdot 2K_{16}

6969: \right] \\

6970: & \ \ \ \ \ \ \ \ \ 

6971: \geq 

6972: \psi\left(

6973: X_{c,x,v_1+v(c)+v}^1\left(

6974: \rho e^{i\theta}

6975: \right), \ 

6976: Y_{c,x,v_1+v(c)+v}^{'1} \left(

6977: \rho e^{i\theta}\right)

6978: \right) + \\

6979: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

6980: \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ 

6981: \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

6982: + (1-\rho) \cdot c \cdot \left[

6983: C_1-c \cdot 2K_{15}- 10c \cdot 2K_{16}-c\cdot K_{19}

6984: \right] \\

6985: & \ \ \ \ \ \ \ \ \ 

6986: \geq 

6987: \psi\left(

6988: X_{c,x,v_1+v(c)+v}^1\left(

6989: \rho e^{i\theta}

6990: \right), \ 

6991: Y_{c,x,v_1+v(c)+v}^{'1} \left(

6992: \rho e^{i\theta}\right)

6993: \right) +

6994: (1-\rho) \cdot c \cdot \frac{ C_1}{2} \\

6995: & \ \ \ \ \ \ \ \ \ 

6996: >

6997: \psi\left(

6998: X_{c,x,v_1+v(c)+v}^1\left(

6999: \rho e^{i\theta}

7000: \right), \ 

7001: Y_{c,x,v_1+v(c)+v}^{'1} \left(

7002: \rho e^{i\theta}\right)

7003: \right), 

7004: \endaligned\right.

7005: \end{equation}

7006: provided

7007: \def\theequation{8.44}\begin{equation}

7008: c_1\leq \frac{ C_1/2}{2K_{15}+20K_{16}+K_{19}}

7009: \end{equation}

7010: This yields the first inequality of~\thetag{ 8.27} and

7011: completes the proof of {\bf ($\mathbf{ 9_1}$)} in 

7012: Case {\bf ($\mathbf{ I_1}$)}.

7013: 

7014: The proof of Lemma~8.3 is complete, because by following the

7015: normalizations of Lemma~5.37 and by formulating analogous inequalities,

7016: cases {\bf ($\mathbf{ I_2}$)} and {\bf (II)} are achieved in a totally

7017: similar way.

7018: \endproof

7019: 

7020: \section*{\S9.~End of proof of Theorem~1.2': application

7021: of the continuity principle}

7022: 

7023: \subsection*{9.1.~Preliminary}

7024: In this section, we shall now complete the proof of Proposition~5.12

7025: (at last!), hence the proof of Theorem~3.19 {\bf (i)} (which is a

7026: direct consequence of Proposition~5.12, as was explained in Section~5)

7027: and hence also the proof of Theorem~1.2', modulo supplementary

7028: arguments postponed to \S9.27 below. By means of a deformation

7029: $A_{x,v,u:c}^1(\zeta)$ (we add a real parameter

7030: $u$) of the family of analytic discs

7031: $A_{x,v:c}^1(\zeta)$ satisfying properties {\bf ($\mathbf{ 1_1}$)} to

7032: {\bf ($\mathbf{ 9_1}$)} of Lemmas~7.12 and~8.3, and by means of the

7033: continuity principe, we shall show that, in Cases {\bf ($\mathbf{

7034: I_1}$)} and {\bf ($\mathbf{ I_2}$)}, there exists a local wedge

7035: $\mathcal{ W}_{p_1}$ of edge $M$ at $p_1$ to which all holomorphic

7036: functions in $\mathcal{ O}\left( \Omega \cup \mathcal{ HW}_1^+\right)$

7037: extend holomorphically and we shall show that in Case {\bf (II)},

7038: there exists a neighborhood $\omega_{p_1}$ of $p_1$ in $\C^n$ to which

7039: all holomorphic functions in $\mathcal{ O}\left( \Omega \cup \mathcal{

7040: W}_2\right)$ extend holomorphically. To organize this last main

7041: step of the proof of Proposition~5.12, we shall consider jointly

7042: Cases~{\bf ($\mathbf{ I_1}$)}, {\bf ($\mathbf{ I_2}$)} and then

7043: afterwards Case~{\bf (II)} separately in \S9.22 below.

7044: 

7045: \subsection*{9.2.~Isotopies of analytic

7046: discs and continuity principle} To begin with, we shall formulate a

7047: convenient version of the continuity principle. If $E\subset \C^n$ is

7048: an arbitrary subset, we denote by

7049: \def\theequation{9.3}\begin{equation}

7050: \mathcal{V}_{\C^n}(E,\rho)=\bigcup_{p\in E}\, 

7051: \{z\in \C^n: \vert z - p \vert < \rho\}

7052: \end{equation}

7053: the union of polydiscs of radius $\rho>0$ centered at points of $E$.

7054: We then have the following lemma, extracted from~\cite{m2}, which

7055: applies to families of analytic discs $A_\tau(\zeta)$ which are

7056: embeddings of $\overline{\Delta}$ into $\C^n$:

7057: 

7058: \def\thelemma{9.4}\begin{lemma}

7059: \text{\rm (\cite{m2}, Proposition~3.3)}

7060: Let $\mathcal{ D}$ be a nonempty domain in $\C^n$ and let $A_\tau:

7061: \overline{\Delta} \to \C^n$ be a one-parameter family of analytic

7062: discs, where $\tau\in\R$ satisfies $0\leq \tau \leq 1$. Assume that

7063: there exist constants $c_\tau$ and $C_\tau$ with $0 < c_\tau < C_\tau$

7064: such that 

7065: \def\theequation{9.5}\begin{equation} c_\tau \vert

7066: \zeta_1 - \zeta_2 \vert < \vert A_\tau(\zeta_1) 

7067: - A_\tau (\zeta_2) \vert <

7068: C_\tau \vert \zeta_1 - \zeta_2 \vert,

7069: \end{equation}

7070: for all distinct points $\zeta_1,\zeta_2 \in \overline{ \Delta}$ and

7071: all $0\leq \tau \leq 1$. Assume that $A_1(\overline{ \Delta} )

7072: \subset \mathcal{ D}$, set $\rho_\tau := \inf\{ \vert t - A_\tau

7073: (\zeta)\vert: \, t\in \partial \mathcal{ D},\, \zeta \in \partial

7074: \Delta\}$, namely $\rho_\tau$ is the polydisc distance between $A_\tau

7075: (\partial \Delta)$ and $\partial \mathcal{ D}$, assume

7076: $\rho_\tau>0$ for all $\tau$ with $0\leq \tau \leq 1$,

7077: and set $\sigma_\tau

7078: := \rho_\tau c_\tau / 2C_\tau$. Then for every holomorphic function

7079: $f \in \mathcal{O} (\mathcal{ D})$, and for every $\tau\in [0, 1]$,

7080: there exists a holomorphic function $F_\tau \in \mathcal{ O}\left(

7081: \mathcal{ V}_{\C^n} (A_\tau\left( \overline{ \Delta}\right),

7082: \sigma_\tau)\right)$ such that $F_\tau =f$ in $\mathcal{ V}_{\C^n}

7083: (A_\tau ( \partial \Delta), \sigma_\tau)\subset \mathcal{ D}$.

7084: \end{lemma}

7085: 

7086: Two analytic discs $A', A'': \overline{\Delta} \to \C^n$ which are of

7087: class $\mathcal{ C}^1$ over $\overline{\Delta}$ and holomorphic in

7088: $\Delta$ and which are both {\it embeddings}\, of $\overline{\Delta}$

7089: into $\C^n$ are said to be {\sl analytically isotopic} if there exists

7090: a $\mathcal{ C}^1$-smooth family of analytic discs $A_\tau: \overline{

7091: \Delta}\to \C^n$ which are of class $\mathcal{ C}^1$ over

7092: $\overline{\Delta}$ and holomorphic in $\Delta$ such that $A_0=A'$,

7093: such that $A_1=A''$ and such that $A_\tau$ is an {\it embedding}\, of

7094: $\overline{ \Delta}$ into $\C^n$ for all $\tau$ with $0\leq

7095: \tau \leq 1$.

7096: 

7097: \subsection*{9.6.~Translations of $M^1$ in $M$}

7098: According to Lemma~5.37, in Case~{\bf ($\mathbf{ I_1}$)}, the

7099: one-codimensional submanifold $M^1\subset M$ is given by the equations

7100: $y'=\varphi'(x,y_1)$ and $x_1=g(x')$. If $u\in \R$ is a small real

7101: parameter, we may define a ``translation'' $M_u^1$ of $M^1$ in $M$ by

7102: the $n$ equations

7103: \def\theequation{9.7}\begin{equation}

7104: M_u^1: \ \ \ \ \ 

7105: y'=\varphi'(x,y_1), \ \ \ \ \ 

7106: x_1=g(x')+u.

7107: \end{equation}

7108: Clearly, we have $M_0^1 \equiv M^1$, we have $M_u^1 \subset (M^1 )^+$

7109: if $u>0$ and we have $M_u^1 \subset (M^1)^-$ if $u <0$. We may perturb

7110: the family of analytic discs $Z_{c, x,v}^d ( \zeta)$ attached to $M^1$

7111: and satisfying Bishop's equation~\thetag{ 7.18} by requiring that it

7112: is attached to $M_u^1$. We then obtain a new family of analytic discs

7113: $Z_{c,x, v,u}^d (\zeta)$ which is half-attached to $M_u^1$ and which

7114: is of class $\mathcal{ C}^{2,\alpha-0}$ with respect to all variables

7115: $(c,x, v, u, \zeta)$, thanks to the stability under perturbation of

7116: the solutions to Bishop's. For $u=0$, this solution coincides with the

7117: family $Z_{c, x, v}^d ( \zeta)$ constructed in \S7.13. Using a similar

7118: definition as in~\thetag{7.53}, namely setting $A_{x,v, u:c}^1

7119: (\zeta):= Z_{c,x, v_1+v(c) +v,u}^1 \left( \Phi_c (\zeta) \right)$, we

7120: obtain a new family of analytic discs which coincides, for $u=0$, with

7121: the family of analytic discs $A_{x, v: c}^1 (\zeta)$ of Lemmas~7.12

7122: and~8.3. Similarly, in Case~{\bf ($\mathbf{ I_2}$)}, taking account

7123: of the normalizations stated in Lemma~5.37, we can also construct an

7124: analogous family of analytic discs $A_{x,v,u:c}^1(\zeta)$. From now on,

7125: we shall fix the scaling parameter $c$ with $0 < c \leq c_1$, so that

7126: the nine properties {\bf ($\mathbf{ 1_1}$)} to {\bf ($\mathbf{ 9_1}$)}

7127: of Lemmas~7.12 and~8.3 are satisfied.

7128: 

7129: \subsection*{9.8.~Definition of a local wedge of edge $M$ at $p_1$

7130: in Cases~($\mathbf{ I_1}$) and~($\mathbf{I_2}$)} First of all, in

7131: Cases~{\bf ($\mathbf{I_1}$)} and~{\bf ($\mathbf{ I_2}$)}, we shall

7132: restrict the variation of the parameter $v$ to a certain

7133: $(n-2)$-dimensional linear subspace $V_2$ of $T_{p_1} \R^n \cong \R^n$

7134: as follows. By hypothesis, the vector $v_1$ does not belong to the

7135: characteristic direction $T_{p_1}M^1 \cap T_{p_1}^cM$, so the real

7136: vector space $\left( \R \cdot v_1 \right) \oplus \left( T_{p_1}M^1

7137: \cap T_{p_1 }^cM \right) \subset T_{ p_1} M^1$ is $2$-dimensional. We

7138: choose an arbitrary $(n-2)$-dimensional real vector subspace

7139: $V_2\subset T_{p_1}M^1$ which is a supplementary in $T_{p_1}M^1$ to

7140: $\left( \R\cdot v_1\right) \oplus \left( T_{p_1} M^1 \cap T_{ p_1}^cM

7141: \right)$ and we shall let the parameter $v$ vary

7142: only in $V_2$.

7143: 

7144: From the rank properties {\bf

7145: ($\mathbf{ 5_1}$)} and {\bf ($\mathbf{ 6_1}$)} of Lemma~7.12 and from

7146: the definition of $V_2$, we deduce that there exists $\varepsilon>0$

7147: small enough with $\varepsilon < < c^2$ such that the mapping

7148: \def\theequation{9.9}\begin{equation}

7149: (x,v,u,\rho)\longmapsto 

7150: A_{x,v,u:c}^1(\rho)

7151: \end{equation}

7152: is a {\it one-to-one immersion}\, from the open set $\{(x,v,u,\rho)\in

7153: \R^n\times V_2 \times \R\times \R: \ \vert x \vert < \varepsilon,

7154: \ \vert v \vert < \varepsilon, \ \vert u \vert < \varepsilon, \

7155: 1-\varepsilon < \rho < 1\}$ into $\C^n$. This property will be

7156: important for uniqueness of the holomorphic extension in our

7157: application of the continuity principle to be conducted in Lemma~9.20

7158: below. Furthermore, shrinking $\varepsilon >0$ if necessary, we can

7159: insure that the open subset

7160: \def\theequation{9.10}\begin{equation}

7161: \left\{

7162: \aligned

7163: {}

7164: &

7165: \mathcal{ W}_{p_1} :=\left\{

7166: A_{x,v,u:c}^1(\rho)\in \C^n: 

7167: (x,v,u,\rho)\in

7168: \R^n\times V_2 \times \R\times \R, \right. \\

7169: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

7170: \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

7171: \left.

7172: \vert x \vert < \varepsilon,

7173: \ \vert v \vert < \varepsilon, \ 

7174: \vert u \vert < \varepsilon, \

7175: 1-\varepsilon < \rho < 1

7176: \right\}

7177: \endaligned\right.

7178: \end{equation}

7179: is a local wedge of edge $M$ at $(p_1, Jv_1)$, with $\mathcal{ W}_{

7180: p_1} \cap M=\emptyset$.

7181: 

7182: Let the closed subset $C$ with $p_1 \in C$ and $C \backslash \{ p_1\}

7183: \subset (H^1)^-$, let the neighborhood $\Omega$ of $M \backslash C$ in

7184: $\C^n$, let the half-wedge $\mathcal{ HW }_{p_1 }^+$ be as in

7185: Proposition~5.12, and let the sub-half-wedge $\mathcal{ HW}_1^+\subset

7186: \mathcal{ HW}_{ p_1 }^+$ be as in \S5.14 and Lemma~5.37. 

7187: In Cases {\bf ($\mathbf{ I_1

7188: }$)} and {\bf ($\mathbf{ I_2}$)}, we shall consider the envelope of

7189: holomorphy of the open subset $\Omega \cup \mathcal{ HW }_1^+ $. We

7190: shall prove in the next paragraphs that, after possibly shrinking it a

7191: little bit, its envelope of holomorphy contains the wedge $\mathcal{ W}_{ p_1 }$.

7192: 

7193: \subsection*{9.11.~Boundaries of analytic discs}

7194: Since we want to apply the continuity principle Lemma~9.4, we have to

7195: show that most discs $A_{x, v, u:c}^1 ( \zeta)$ have their boundaries

7196: in $\Omega \cup \mathcal{ HW}_1^1+$. 

7197: To this aim, it will be useful to decompose the boundary

7198: $\partial \Delta$ in three closed parts $\partial \Delta = \partial^1

7199: \Delta \cup \partial^2 \Delta \cup \partial^3 \Delta$, where

7200: \def\theequation{9.12}\begin{equation}

7201: \left\{

7202: \aligned

7203: \partial^1\Delta := 

7204: & \

7205: \left\{

7206: e^{i\theta} \in \partial\Delta : \ 

7207: \vert \theta \vert \leq \pi/2 -\varepsilon

7208: \right\} \subset \partial^+\Delta, \\

7209: \partial^2\Delta := 

7210: & \

7211: \left\{

7212: e^{i\theta} \in \partial\Delta : \ 

7213: \pi/2 +\varepsilon \leq

7214: \vert \theta \vert \leq \pi

7215: \right\} \subset \partial^-\Delta, \\

7216: \partial^3\Delta := 

7217: & \

7218: \left\{

7219: e^{i\theta} \in \partial\Delta : \ 

7220: \pi/2-\varepsilon \leq 

7221: \vert \theta \vert \leq \pi/2 +\varepsilon

7222: \right\} \subset \partial\Delta, \\

7223: \endaligned\right.

7224: \end{equation}

7225: where $\varepsilon$ with $0< \varepsilon < < c^2$ is as in \S9.8 just

7226: above. This decomposition is illustrated in the left hand side of {\sc

7227: Figure~17} below. Next, we observe that the two points $A_{0,

7228: 0,0:c}^1(i)$ and $A_{ 0,0, 0:c}^1(-i)$ belong to $(H^1 )^-\subset M

7229: \backslash C\subset \Omega$, hence there exists a fixed open

7230: neighborhood of these two points which is contained in $\Omega$. We

7231: shall denote by $\omega^3$ such a (disconnected) neighborhood, for 

7232: instance the union of two small open polydiscs centered at 

7233: these two points. To

7234: proceed further, we need a crucial geometric information about the

7235: boundaries of the analytic discs $A_{x,v,u:c}^1 (\zeta)$ with $u\neq

7236: 0$.

7237: 

7238: \def\thelemma{9.13}\begin{lemma}

7239: Under the assumptions of Cases~{\bf ($\mathbf{ I_1}$)} and~{\bf

7240: ($\mathbf{ I_2}$)} of Proposition~5.12, after

7241: shrinking $\varepsilon >0$ if necessary, then 

7242: \def\theequation{9.14}\begin{equation}

7243: A_{x,v,u:c}^1\left(\partial \Delta

7244: \right)\subset \Omega \cup 

7245: \mathcal{ HW}_1^+,

7246: \end{equation}

7247: for all $x$ with $\vert x \vert < \varepsilon$, for all $v$ with

7248: $\vert v \vert < \varepsilon$ and for all {\rm nonzero} $u\neq 0$ with

7249: $\vert u \vert < \varepsilon$. 

7250: \end{lemma}

7251: 

7252: \proof 

7253: Firstly, since $A_{0, 0, 0:c}^1 (\pm i) \in \omega^3$, it

7254: follows just by continuity of the family $A_{x, v, u: c}^1(\zeta)$

7255: that, after possibly shrinking $\varepsilon >0$, the closed arc $A_{x,

7256: v,u: c}^1 \left( \partial^3 \Delta \right)$ is contained in

7257: $\omega^3$, for all $x$ with $\vert x \vert < \varepsilon$, for all

7258: $v$ with $\vert v \vert < \varepsilon$ and for all $u$ with $\vert u

7259: \vert < \varepsilon$. Secondly, since $A_{0, 0, 0:c}^1( \partial^2

7260: \Delta) \subset A_{0, 0, 0: c}^1 \left( \partial^-\Delta \backslash

7261: \{i,-i\}\right) \subset \mathcal{ HW }_1^+$, then by property {\bf

7262: ($\mathbf{ 9_1 }$)} of Lemma~8.3, it follows just by continuity of the

7263: family $A_{x, v, u:c}^1(\zeta)$ that, after possibly shrinking

7264: $\varepsilon >0$, the closed arc $A_{x, v, u:c}^1 \left( \partial^2

7265: \Delta \right)$ is contained in $\mathcal{ HW}_1^+$, for all $x$ with

7266: $\vert x \vert < \varepsilon$, for all $v$ with $\vert v \vert <

7267: \varepsilon$ and for all $u$ with $\vert u \vert < \varepsilon$.

7268: Thirdly, it follows from the inclusion $A_{x, v, u: c}^1 \left(

7269: \partial^1 \Delta \right) \subset A_{x, v,u:c}^1 \left( \partial^+

7270: \Delta \right) \subset M_u^1$ and from the inclusion $M_u^1 \subset

7271: \Omega$ for all $u\neq 0$ that, after possibly shrinking

7272: $\varepsilon>0$, the closed arc $A_{x, v, u:c}^1 \left( \partial^1

7273: \Delta \right)$ is contained in $\Omega$, for all $x$ with $\vert x

7274: \vert < \varepsilon$, for all $v$ with $\vert v \vert < \varepsilon$

7275: and for all $u$ with $\vert u \vert < \varepsilon$ and $u \neq 0$. 

7276: This completes the proof of Lemma~9.13.

7277: \endproof

7278: 

7279: \bigskip

7280: \begin{center}

7281: \input three-boundary.pstex_t

7282: \end{center}

7283: 

7284: In addition to this lemma, we notice that it follows immediately from

7285: properties {\bf ($\mathbf{8_1}$)} and {\bf ($\mathbf{9_1}$)} of

7286: Lemma~8.3 that $A_{x,0,0:c}^1(\overline{\Delta})$ is contained in

7287: $\Omega \cup \mathcal{ HW}_1^+$ for all $x$ with $\vert x \vert <

7288: \varepsilon$ such that $A_{x,0,0:c}^1(1)\in T^1\cap (H^1)^+$. This

7289: property and Lemma~9.13 are illustrated in the right hand side of {\sc

7290: Figure~17} just above.

7291: 

7292: \subsection*{9.15.~Analytic isotopies}

7293: Next, in Case {\bf ($\mathbf{ I_1}$)}, we fix some $x_0= (x_{1; 0},0,

7294: \dots, 0) \in \R^n$ with $0< x_{1;0} < \varepsilon$. Then $A_{x_0, 0,

7295: 0:c }^1 (1)= x_0+ i h(x_0)$ belongs to $T^1 \cap (

7296: H^1)^+$. Analogously, in Cases {\bf ($\mathbf{ I_2}$)}, we fix some

7297: $x_0=(0,\dots,0, x_{ n;0})\in \R^n$ with $0< x_{n;}0 <

7298: \varepsilon$. Then in this second case, the point $A_{x_0, 0,

7299: 0:c}^1(1)=x_0+ ih(x_0)$ also belongs to $T^1 \cap (H^1 )^+$. We fix

7300: the disc $A_{x_0, 0,0:c}^1( \zeta )$, which satisfies $A_{x_0,

7301: 0,0:c}^1( \overline{ \Delta })\subset \Omega \cup \mathcal{ HW 

7302: }_1^+$.

7303: 

7304: \def\thelemma{9.16}\begin{lemma}

7305: In Cases~{\bf ($\mathbf{ I_1})$} and~{\bf ($\mathbf{ I_2}$)}, every

7306: disc $A_{x, v, u:c}^1 (\zeta)$ with $\vert x \vert < \varepsilon$,

7307: $\vert v \vert < \varepsilon$, $\vert u \vert < \varepsilon$ and

7308: $u\neq 0$ is analytically isotopic to the disc $A_{x_0, 0,

7309: 0:c}^1(\zeta)$, with the boundaries of the analytic discs of the

7310: isotopy being all contained in $\Omega \cup \mathcal{ HW}_1^+$.

7311: \end{lemma}

7312: 

7313: \proof

7314: Indeed, since the set $\{u=0\}$ is a hyperplane, there clearly exists

7315: a $\mathcal{ C}^{2, \alpha- 0}$-smooth curve $\tau \mapsto (x (\tau),v

7316: (\tau),u (\tau))$ in the parameter space which joins a given arbitrary

7317: point $(x^*, v^*,u^*)$ with $u^*\neq 0$ to the point $(x_0, 0,0)$

7318: without meeting the hyperplane $\{u= 0\}$, except at its endpoint

7319: $(x_0, 0,0)$. According to the previous Lemma~9.13, each boundary

7320: $A_{x (\tau), v (\tau), u (\tau) :c}^1( \partial \Delta)$ is then

7321: automatically contained in $\Omega \cup \mathcal{ HW }_1^+$, which

7322: completes the proof.

7323: \endproof

7324: 

7325: \subsection*{9.17.~holomorphic extension 

7326: to a local wedge of edge $M$ at $p_1$} In Cases {\bf ($\mathbf{

7327: I_1}$)} and~{\bf ($\mathbf{ I_2}$)}, we define the

7328: following $\mathcal{ C}^{2,\alpha-0}$-smooth connected

7329: hypersurface of $\mathcal{ W}_{p_1}$:

7330: \def\theequation{9.18}\begin{equation}

7331: \left\{

7332: \aligned

7333: \mathcal{ M}_{p_1}:= 

7334: \left\{

7335: A_{x,v,0:c}^1(\rho): \ 

7336: (x,v,\rho)\in \R^n\times V_2 \times \R, \right. \\

7337: \left.

7338: \vert x \vert < \varepsilon, \ 

7339: \vert v \vert < \varepsilon, \

7340: 1-\varepsilon < \rho < \varepsilon

7341: \right\}, 

7342: \endaligned\right.

7343: \end{equation}

7344: together with the following closed subset of 

7345: $\mathcal{ M}_{p_1}$:

7346: \def\theequation{9.19}\begin{equation}

7347: \left\{

7348: \aligned

7349: \mathcal{ C}_{p_1} := 

7350: & \

7351: \left\{

7352: A_{x,v,0:c}^1(\rho): \ 

7353: (x,v,\rho)\in \R^n\times V_2 \times \R, \right. \\

7354: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

7355: \left.

7356: A_{x,v,0:c}^1(\partial^+ \Delta) \cap 

7357: C= \emptyset, \ 

7358: \vert x \vert < \varepsilon, \ 

7359: \vert v \vert < \varepsilon, \

7360: 1-\varepsilon < \rho < \varepsilon

7361: \right\}.

7362: \endaligned\right.

7363: \end{equation}

7364: Since $A_{x, 0, 0: c}^1( \partial^+ \Delta)$ is contained in $(

7365: H^1)^+$ for all $x$ such that $A_{ x, 0, 0}^1(1) \in T^1 \cap (

7366: H^1)^+$, the closed subset $\mathcal{ C}_{ p_1}$ of $\mathcal{ M}_{

7367: p_1}$ is a {\it proper}\, closed subset of $\mathcal{ M}_{ p_1}$.

7368: The following figure provides a geometric illustration.

7369: 

7370: \bigskip

7371: \begin{center}

7372: \input wedge-half-wedge.pstex_t

7373: \end{center}

7374: 

7375: We can now state the main lemma of this section, which will complete

7376: the proof of Proposition~5.12 in Cases~{\bf ($\mathbf{ I_1}$)}

7377: and~{\bf ($\mathbf{ I_2}$)}.

7378: 

7379: \def\thelemma{9.20}\begin{lemma}

7380: After possibly shrinking $\Omega$ in a small neighborhood of $p_1$ and

7381: after possibly shrinking $\varepsilon >0$, the set $\mathcal{ W}_{

7382: p_1} \cap \left[ \Omega \cup \mathcal{ HW}_1^+ \right]$ is connected

7383: and for every holomorphic function $f \in \mathcal{ O}\left( \Omega

7384: \cup \mathcal{ HW }_1^+ \right)$, there exists a holomorphic function

7385: $F \in \mathcal{ O} \left( \Omega \cup \mathcal{ HW }_1^+ \cup

7386: \mathcal{ W}_{p_1} \right)$ such that $F \vert_{ \Omega \cup \mathcal{

7387: HW}_1^+}=f$.

7388: \end{lemma}

7389: 

7390: \proof

7391: Remind that $\varepsilon < < c^2$ and remind that the wedge $\mathcal{

7392: W}_{p_1}$ with $\mathcal{ W}_{p_1}\cap M= \emptyset$ in the two cases

7393: is of size ${\rm O}( \varepsilon)$. Since $C$ is contained in

7394: $(H^1)^-\cup \{ p_1\}\subset M^1$, we observe that $M \backslash C$ is

7395: locally connected at $p_1$. Since the half-wedge $\mathcal{ HW }_1^+$

7396: defined in Lemma~5.37 by simple inequalities is of size ${\rm O} (

7397: \delta_1)$, if moreover $\varepsilon < < \delta_1$, after shrinking

7398: $\Omega$ if necessary in a smal neighborhood of $p_1$ whose size

7399: is ${\rm

7400: O}(\varepsilon)$, it follows that we can assume that $\mathcal{

7401: W}_{p_1} \cap \left[ \Omega \cup \mathcal{ HW}_1^+ \right]$ is

7402: connected. However, in {\sc Figure~18} above, because we draw $M$ as

7403: if it were one-dimensional, the intersection $\mathcal{ W}_{p_1} \cap

7404: \left[ \Omega \cup \mathcal{ HW }_1^+ \right]$ appears to be

7405: disconnected, which is a slight incorrection.

7406: 

7407: Let $f$ be an arbitrary holomorphic function in $\mathcal{ O} \left(

7408: \Omega \cup \mathcal{ HW}_1^+ \right)$. Thanks to the isotopy

7409: Lemma~9.16 and thanks to the continuity principle Lemma~9.4, we deduce

7410: that $f$ extends holomorphically to a neighborhood in $\C^n$ of every

7411: disc $A_{x, v, u: c}( \overline{ \Delta})$ whose boundary $A_{ x, v,

7412: u:c} (\partial \Delta)$ is contained in $\Omega \cup \mathcal{

7413: HW}_1^+$. Using the fact that the mapping~\thetag{ 9.9} is one-to

7414: one, we deduce that we can extend $f$ uniquely by means of Cauchy's

7415: formula at points of the form $A_{x,v,u:c}^1(\rho)$ with 

7416: such values $\vert x

7417: \vert < \varepsilon$, $\vert v \vert < \varepsilon$, $\vert u \vert <

7418: \varepsilon$ and $1-\varepsilon < \rho < \varepsilon$ for 

7419: which $A_{x,v,u:c}^1(\partial \Delta) \subset

7420: \Omega \cup \mathcal{ HW}_1^+$, simply as follows

7421: \def\theequation{9.21}\begin{equation}

7422: f\left(

7423: A_{x,v,u:c}^1(\rho)

7424: \right):=\int_{\partial \Delta} \, 

7425: \frac{ f\left(

7426: A_{x,v,u:c}^1(\widetilde{ \zeta})

7427: \right)}{\widetilde{ \zeta} -\rho} \, 

7428: d \widetilde{ \zeta}.

7429: \end{equation}

7430: With this definition, we extend $f$ holomorphically and uniquely to

7431: the domain $\mathcal{ W}_{ p_1} \backslash \mathcal{ C}_{p_1}$, where

7432: $\mathcal{ C}_{p_1}$ is the proper closed subset, defined by~\thetag{

7433: 9.21}, of the $\mathcal{ C}^{2,\alpha-0}$-smooth hypersurface

7434: $\mathcal{ M}_{ p_1 }$ defined by~\thetag{ 9.20}. Let $F \in \mathcal{

7435: O} \left(\mathcal{ W}_{p_1} \backslash \mathcal{ C}_{p_1} \right)$

7436: denote this holomorphic extension. Since $\mathcal{ W}_{p_1} \cap

7437: \left[ \Omega \cup \mathcal{ HW}_1^+ \right]$ is connected, it follows

7438: that $\left[ \mathcal{ W}_{p_1} \backslash \mathcal{ C}_{p_1}\right]

7439: \cap \left[ \Omega \cup \mathcal{ HW }_1^+ \right]$ is also

7440: connected. By Lemma~9.4, the function $f$ and its holomorphic

7441: extension $F$ coincide in a neighborhood of every boundary

7442: $A_{x,v,u:c}^1 (\partial \Delta)$ which is contained in the domain

7443: $\Omega \cup \mathcal{ HW}_1^+$. From the analytic continuation

7444: principle, we deduce that there exists a well-defined function, still

7445: denoted by $F$, which is holomorphic in $\left[ \mathcal{ W}_{p_1}

7446: \backslash \mathcal{ C}_{p_1}\right] \cap \left[ \Omega \cup \mathcal{

7447: HW}_1^+ \right]$ and which extends $f$, namely $F\vert_{\Omega \cup

7448: \mathcal{ HW}_1^+}=f$.

7449: 

7450: To conclude the proof of Proposition~5.12 in Cases {\bf ($\mathbf{

7451: I_1}$)} and~{\bf ($\mathbf{ I_2}$)}, it suffices to extend $F$

7452: holomorphically through the closed subset $\mathcal{ C}_{p_1}$ of the

7453: connected hypersurface $\mathcal{ M}_{p_1}$ in the domain $\mathcal{

7454: W}_{p_1} \subset \C^n$. Since $n\geq 2$, we notice that we are

7455: exactly in the situation of Theorem~1.4 in the special, much simpler

7456: case where the generic submanifold $M$ is replaced by a domain of

7457: $\C^n$. It may therefore appear to be quite satisfactory to have

7458: reduced Theorem~1.2' to the CR dimension $\geq 2$ version Theorem~1.4,

7459: in an open subset of $\C^n$ (notice however that Theorem~1.4 as well

7460: as Theorems~1.2 and~1.2' are stated in positive codimension, since the

7461: case where $M$ is replaced by a domain of $\C^n$ is relatively trivial

7462: in comparison). The removability of such a proper closed subset

7463: contained in a connected hypersurface of a domain in $\C^n$ is known,

7464: follows from~\cite{j4} and is explicitely stated and proved as

7465: Lemma~2.10, p.~842 in~\cite{ mp1}. However, we shall still provide

7466: another different geometric proof of this simple removability result

7467: in Lemma~10.10 below, using fully the techniques developed in the

7468: previous sections.

7469: 

7470: The proofs of Lemma~9.20 together with Cases~{\bf ($\mathbf{I_1}$)}

7471: and~{\bf ($\mathbf{ I_2}$)} of Proposition~5.12 are complete now.

7472: \endproof

7473: 

7474: \subsection*{9.22.~End of proof of Proposition~5.12 in 

7475: Case {\bf (II)}} According to Lemma~5.37, in Case {\bf (II)}, the

7476: one-codimensional totally real submanifold $M^1 \subset M$ is given by

7477: the equations $y'=\varphi'(x,y_1)$ and $x_n=g(x'')$. If $u\in \R$ is

7478: a small real parameter, we may define a ``translation'' $M_u^1$ of

7479: $M^1$ in $M$ by the equations

7480: \def\theequation{9.23}\begin{equation}

7481: y'=\varphi'(x,y_1), \ \ \ \ \ \ \ \ 

7482: x_n=g\left(x''\right)+u.

7483: \end{equation}

7484: Similarly as in \S9.6, we may construct a family of analytic discs

7485: $A_{x,v,u:c}^1(\zeta)$ half-attached to $M_u^1$. We then we fix a

7486: small scaling parameter $c$ with $0 < c \leq c_1$ so that properties

7487: {\bf ($\mathbf{ 1_1}$)} to {\bf ($\mathbf{ 9_1}$)} of Lemmas~7.12

7488: and~8.3 hold true. Similarly as in \S9.8, we shall restrict 

7489: the variation

7490: of the parameter $v$ to an arbitrary $(n-1)$-dimensional subspace

7491: $V_1$ of $T_{p_1}M^1\cong \R^n$ which is supplementary to the real

7492: line $\R\cdot v_1$ in $T_{p_1}M^1$. If $\varepsilon>0$ is small

7493: enough with $\varepsilon < < c^2$, it follows that the mapping

7494: \def\theequation{9.24}\begin{equation}

7495: (x,v,u,\rho) \longmapsto 

7496: A_{x,v,u:c}^1(\rho) 

7497: \end{equation}

7498: is a {\it one-to-one immersion}\, from the open set $\{(x,v,\rho)\in

7499: \R^n\times V_1 \times \R\times \R: \ \vert x \vert < \varepsilon, \

7500: \vert v \vert < \varepsilon, \ 1-\varepsilon < \rho < 1\}$ into

7501: $\C^n$. Thanks to the choice of the linear subspace $V_1$, shrinking

7502: $\varepsilon >0$ if necessary, it follows that for every $u$ with

7503: $\vert u \vert < \varepsilon$, the open subset

7504: \def\theequation{9.25}\begin{equation}

7505: \left\{

7506: \aligned

7507: {}

7508: &

7509: \mathcal{ W}_u^1 :=\left\{

7510: A_{x,v,u:c}^1(\rho)\in \C^n: 

7511: (x,v,\rho)\in

7512: \R^n \times \R\times \R, \right. \\

7513: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

7514: \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

7515: \left.

7516: \vert x \vert < \varepsilon,

7517: \ \vert v \vert < \varepsilon, \ 

7518: 1-\varepsilon < \rho < 1

7519: \right\}

7520: \endaligned\right.

7521: \end{equation}

7522: is a local wedge of edge $M_u^1$. Clearly, this wedge 

7523: $\mathcal{ W}_u^1$ depends $\mathcal{ C}^{2,\alpha-0}$-smoothly 

7524: with respect to $u$. 

7525: 

7526: Using the fact that in Case~{\bf (II)} we have

7527: \def\theequation{9.26}\begin{equation}

7528: \frac{\partial A_{0,0,0:c}^1}{\partial \theta}(1)=

7529: v_1=(1,0,\dots,0)\in T_{p_1}M^1\cap T_{p_1}^cM,

7530: \end{equation}

7531: one can prove that Lemma~9.13 holds true with $\mathcal{ HW}_1^+$

7532: replaced by $\mathcal{ W}_2$ in~\thetag{ 9.14} and also that

7533: Lemma~9.16 holds true, again with $\mathcal{ HW}_1^+$ replaced by

7534: $\mathcal{ W}_2$. Similarly as in the proof of Lemma~9.20, applying

7535: then the continuity principle and using the fact that the

7536: mapping~\thetag{ 9.24} is one-to-one, after possibly shrinking

7537: $\Omega$ in a neighborhood of $p_1$, and shrinking $\varepsilon>0$, we

7538: deduce that for each $u\neq 0$, there exists a holomorphic function

7539: $F\in \mathcal{ O}\left(\Omega \cup \mathcal{ W}_2 \cup

7540: \mathcal{ W}_u^1 \right)$ with 

7541: $F_{\vert \Omega \cup \mathcal{ W}_2}=f$.

7542: 

7543: To conclude the proof of Proposition~5.12 in Case {\bf (II)}, it

7544: suffices to observe that for every fixed small $u$ with $

7545: -\varepsilon < < u < 0$, the wedge $\mathcal{ W}_u^1$ contains in fact

7546: a neighborhood $\omega_{p_1}$ of $p_1$ in $\C^n$.

7547: 

7548: The proofs of Proposition~5.12 and of Theorem~3.19 {\bf (i)} 

7549: are complete now. \qed

7550: 

7551: \subsection*{9.27.~End of proof of Theorem~1.2'}

7552: In order to derive Theorem~1.2' from Theorem~3.19 {\bf (i)}, we now

7553: remind the necessity of supplementary arguments about the stability of

7554: our constructions under deformation. Coming back to the strategy

7555: developped in \S3.16, we had a wedge $\mathcal{ W}_1$ attached to $M

7556: \backslash C_{\rm nr}$. Using a partition of unity, we may introduce a

7557: one-parameter $\mathcal{ C}^{ 2, \alpha}$-smooth family of generic

7558: submanifolds $M^d$, $d\in \R$, $d\geq 0$, with $M^0 \equiv M$, with

7559: $M^d$ containing $C_{\rm nr}$ and with $M^d \backslash C_{ \rm nr}$

7560: contained in $\mathcal{ W}_1$. In the proof of Theorem~3.19 {\bf (i)},

7561: thanks to this deformation, the wedge $\mathcal{ W}_1$ was replaced by

7562: a neighborhood $\Omega$ of $M\backslash C_{\rm nr}$ in $\C^n$.

7563: 

7564: In Sections~4 and~5, we constructed an important semi-local half-wedge

7565: $(\mathcal{ HW}_\gamma^+)^d$ attached to a one-sided neighborhood of

7566: $(M^1)^d$ in $M^d$ along a characteristic segment $\gamma^d$ of

7567: $M^d$. Now, we make the crucial claim that, after possibly adapting

7568: the deformation $M^d$, we may achieve that the geometric extent of

7569: this semi-local half-wedge be uniform as $d>0$ tends to zero, namely

7570: $(\mathcal{ HW}_\gamma^+)^d$ tends to a semi-local half-wedge

7571: $(\mathcal{ HW}_\gamma^+)^0$ attached to a one-sided neighborhood of

7572: $M^1$ in $M$ along $\gamma$, as $d$ tends to zero. Indeed, in

7573: Section~4 we have constructed a family of analytic discs $(\mathcal{

7574: Z}_{t,\chi,\nu:s}(\zeta))^d$ ({\it cf.}~\thetag{ 4.61}) which covers

7575: the half-wedge $(\mathcal{ HW}_\gamma^+)^d$. Thanks to the stability

7576: of Bishop's equation under $\mathcal{ C}^{2,\alpha}$-smooth

7577: perturbations, the deformed family $(\mathcal{

7578: Z}_{t,\chi,\nu:s}(\zeta))^d=: \mathcal{ Z}_{t,\chi,\nu:s}^d(\zeta)$ is

7579: also of class $\mathcal{ C}^{2,\alpha-0}$ with respect to the

7580: parameter $d$. We remind that for every $d>0$, the family $\mathcal{

7581: Z}_{t,\chi,\nu:s}^d(\zeta)$ was in fact constructed by means of a

7582: family $\widehat{ Z}_{ r_0,t,\tau,\chi,\nu:s}^d(\zeta)$ obtained by

7583: solving Bishop's equation~\thetag{ 4.40}, where we now add the

7584: parameter $d$ in the function $\Phi'$. In order to construct the

7585: semi-local attached half-wedge, we have used the rank property stated

7586: in Lemma~4.34. This rank property relied on the possibility of

7587: deforming the disc $\widehat{ Z}_{r_0,t:s}(\zeta)$ near the point

7588: $\widehat{ Z}_{r_0,t:s}^d(-1)$ in the open neighborhood

7589: $\Phi_s(\Omega) \equiv \Phi_s\left(\mathcal{ W}_1 \right)$ of $

7590: \Phi_s\left(M^d \right)$. As $d>0$ tends to zero, if $M^d$ tends to

7591: $M$, the size of the neighborhood $\Phi_s\left(\mathcal{ W}_1 \right)$

7592: shrinks to zero, hence it could seem that the we have no control on

7593: the semi-local attached half-wege $(\mathcal{ HW}_\gamma^+)^d$ as

7594: $d>0$ tends to zero. Fortunately, since the points $\widehat{

7595: Z}_{r_0,0:s}^d(-1)$ in a neighborhood of which we introduce the

7596: deformations~\thetag{ 4.30} are at a uniformly positive distance

7597: $\delta >0$ from $\gamma$, we may choose the deformation $M^d$ of $M$

7598: to tend to $M$ as $d$ tends to zero only in a small neighborhood of

7599: $\gamma$, whose size is small in comparison to this distance

7600: $\delta$. By smoothness with respect to $d$ of the family $\mathcal{

7601: Z}_{t,\chi,\nu:s}^d (\zeta)$, we then deduce that the semi-local

7602: half-wedge $(\mathcal{ HW}_\gamma^+)^d$ tends to a nontrivial

7603: semi-local half-wedge $(\mathcal{ HW}_\gamma^+)^0$ as $d$ tends to

7604: zero, which proves the claim.

7605: 

7606: Next, again thanks to the stability of Bishop's equation under

7607: perturbation, all the constructions of Sections~5, 6, 7, 8 and~9 above

7608: may be achieved to depend $\mathcal{ C}^{2, \alpha-0}$-smoothly, hence

7609: uniformly, with respect to $d$. Importantly, we observe that if the

7610: deformation $M^d$ is chosen so that $M^d$ tends to $M$ only in a small

7611: neighborhood of $p_1$ of size $ < < \varepsilon$, then the shrinking

7612: of $\varepsilon$ which occurs in Lemma~9.13 may be achieved to be

7613: uniform as $d$ tends to zero, because the part

7614: $A_{x,v,u:c}(\partial^3\Delta)$ stays in a uniform compact subset of

7615: $\Omega$, as $d$ tends to zero. At the end of the proof of

7616: Proposition~5.12, we then obtain univalent holomorphic extension to a

7617: local wedge $\mathcal{ W}_{p_1}^d$ of edge $M^d$ or to a neighborhood

7618: $\omega_{ p_1 }^d$ of $M^d$ in $\C^n$, and they tend smoothly to a

7619: wedge $\mathcal{ W }_{ p_1}^0$ of edge $M$ at $p_1$ or to a

7620: neighborhood $\omega_{ p_1 }$ of $p_1$ in $\C^n$.

7621: 

7622: The proof of Theorem~1.2' is complete.

7623: \qed

7624: 

7625: \section*{\S10.~Three proofs of Theorem~1.4}

7626: 

7627: \subsection*{10.1.~Preliminary}

7628: Theorem~1.4 may be established by means of the processus of

7629: minimalization of generic submanifods developed by B.~J\"oricke

7630: in~\cite{j2}, as is done effectively in~\cite{j4} in the hypersurface

7631: case and then in~\cite{p2} in arbitrary codimension. In this section,

7632: we shall suggest three more different proofs of Theorem~1.4. As

7633: explained in Section~3, it suffices to treat $\mathcal{

7634: W}$-removability, and essentially to prove Proposition~3.22. The first

7635: proof appears already in~\cite{ m2} and also in~\cite{p1}. The second

7636: proof consists in repeating some of the constructions of the previous

7637: sections, using the fact that $M^1$ is of positive CR dimension in

7638: order to simplify substantially the reasonings. The third proof

7639: consists of a slicing argument showing that {\it Theorem~1.4 is in

7640: fact a logical consequence of Theorem~1.2'}. In fact, because these

7641: three proofs are already written elsewhere or very close to the

7642: constructions developed in the previous sections, we shall only

7643: provide summaries here.

7644: 

7645: \subsection*{10.2.~Normal deformation of analytic discs attached 

7646: to $M^1$} Firstly, in the situation of Proposition~3.22, because $M^1$

7647: is of positive CR dimension, we can construct a small analytic disc $A

7648: (\zeta)$ attached to $M^1$ which satisfies $A(1) = p_1 \in M^1$ and $A

7649: (\partial \Delta \backslash \{1\}) \subset (H^1)^+$. As in~\cite{m2},

7650: \cite{ mp1}, using normal deformations of $A$ near $A(-1)$, we may

7651: include $A$ in a $\mathcal{ C}^{ 2, \alpha-0}$-smooth parametrized

7652: family $A_v (\zeta)$ of analytic discs attached to $M^1$, where $v \in

7653: \R^{d+1}$ is small, so that the rank at $v=0$ of the mapping $v

7654: \longmapsto -\frac{ \partial A_v}{ \partial \rho}(1) \in T_{ p_1} \C^n

7655: \ {\rm mod} \ T_{p_1} M^1 \cong \R^{ d+1}$ is maximal equal to $(d

7656: +1)$, the codimension of $M^1$ in $\C^n$. For this, we use a

7657: deformation lemma which is essentially due to A.~Tumanov~\cite{tu3},

7658: which appears as Lemma~2.7 in~\cite{ mp1} and which was already used

7659: above (with a supplementary parameter $s$) in Lemma~4.34. Then we add

7660: a ``translation'' parameter $x\in \R^{ 2m+d-1}$, getting a family of

7661: analytic discs $A_{x,v} (\zeta)$ with $A_{x,v} (\partial

7662: \Delta)\subset M^1$ so that the rank at $x=0$ of the mapping $x

7663: \longmapsto A_{x,0} (1) \in M^1$ is maximal equal to $(2m+d-1)$, the

7664: dimension of $M^1$. Finally, we introduce some ``translations''

7665: $M_u^1$ of $M^1$ in $M$, where $u \in \R$, and we obtain a family $A_{

7666: x, v,u} (\zeta)$ of analytic discs attached to $M_u^1$. For $u\neq 0$,

7667: since the discs $A_{x,v,u}(\zeta)$ are attached to $M_u^1$, their

7668: boundaries are contained in $M\backslash C$. Applying the

7669: approximation Lemma~4.8, we deduce that for every $u\neq 0$, all

7670: holomorphic functions in the open set $\Omega$ of Proposition~3.22

7671: (which contains $M\backslash C$) extend holomorphically to the local

7672: wedge $\mathcal{ W }_u := \{A_{x, v,u} (\rho): \ \vert x \vert <

7673: \varepsilon, \ \vert v \vert < \varepsilon, \ 1-\varepsilon < \rho <

7674: 1\}$ of edge $M_u^1$. To control the univalence of the holomorphic

7675: extension, it suffices to shrink $\Omega$ a little bit in a

7676: neighborhood of $p_1$. To conclude the proof, one observes as

7677: in~\cite{ j4}, \cite{ cs} that the union $\bigcup_{ u \neq 0} \,

7678: \mathcal{ W}_u^1$ contains a local wedge of edge $M$ at $p_1$. This

7679: processus is sometimes called ``sweeping out by wedges''. We notice

7680: that this proof is geometrically much simpler than the proof of

7681: Theorem~1.2' achieved in the previous sections.

7682: 

7683: \subsection*{10.3.~Half-attached analytic discs}

7684: Secondly, we may generalize our constructions achieved in the previous

7685: sections from the CR dimension $m=1$ case to the CR dimension $m\geq

7686: 2$ case, as follows. Let $M$, $M^1$, $p_1$, $H^1$ and $C$ be as in

7687: Proposition~3.22. The fact that $M^1$ is of positive CR dimension

7688: will provide substantial geometric simplifications for essentially two

7689: reasons. Indeed, as the hypersruface $H^1\subset M^1$ is generic and

7690: at least of real dimension $n$, there exists a (in fact infinitely many)

7691: maximally real submanifold $K^1$ passing through $p_1$ which is

7692: contained in $H^1$. Then

7693: \begin{itemize}

7694: \item[{\bf (i)}]

7695: Applying the considerations of Section~7, we may construct families of

7696: small analytic discs $A_{x,v} (\zeta)$ which are half-attached to

7697: $K^1$ and which cover a local wedge of edge $K^1$ at $p_1$, when the

7698: translation parameter $x$ and the rotation parameter $v$ vary. We

7699: notice that this would be impossible in the case $m=1$, because in

7700: this case $H^1$ is of real dimension $(n-1)$, hence does {\it not}\,

7701: contain any maximally real submanifold.

7702: \item[{\bf (ii)}]

7703: We can even prescribe the direction $\frac{ \partial A_{ 0,0} }{

7704: \partial \theta} (1)$ as an arbitrary given nonzero vector $v_1\in T_{

7705: p_1}K^1$. Again, this would be impossible in the CR dimension $m=1$

7706: case.

7707: \end{itemize}

7708: Let $\mathcal{ HW }_1^+$ be a local half-wedge of edge $(M^1)^+$ at

7709: $p_1$, whose construction is suggested in \S4.64. Since $K^1$ is

7710: generic, we may choose a nonzero vector $v_1\in T_{ p_1}K^1$ with the

7711: property that $\mathcal{ HW }_1^+$ is directed by $J

7712: v_1$. Generalizing Lemma~8.3, we see that $A_{x,v} \left( \overline{

7713: \Delta } \backslash \partial^+ \Delta \right)$ is contained in

7714: $\mathcal{ HW}_1^+$. We notice that in the CR dimension $m=1$ case,

7715: $H^1$ is {\it not}\, generic, and we remember that the choice of a

7716: special point $p_1$ to be removed locally and the choice of a

7717: supporting hypersurface $H^1 \subset M^1$ was much more subtle,

7718: because we had to insure that there exists a vector $v_1 \in T_{

7719: p_1}H^1$ such that $\mathcal{ HW }_1^+$ is directed by $J v_1$.

7720: 

7721: Next, we can translate $K^1$ in $M^1$ by means of a small parameter

7722: $t\in \R^{d-1}$ and then $M^1$ in $M$ by means of a small parameter

7723: $u\in \R$. By stability of Bishop's equation, we get a family of

7724: analytic discs $A_{x,v,t,u}(\zeta)$ half-attached to the translation

7725: $K_{t,u}^1$ of $K^1$. Nine properties analogous to properties {\bf

7726: ($\mathbf{ 1_1}$)} to {\bf ($\mathbf{ 9_1}$)} of Lemmas~7.12 and~8.3

7727: are then satisfied and we conclude the proofs of Proposition~3.22 and

7728: of Theorem~1.4 in essentially the same way as in Section~9

7729: above. We shall not write down all the details.

7730: 

7731: We notice that this second strategy of proof is much more complicated

7732: than the first (known) proof summarized in \S10.2 just above.

7733: 

7734: \subsection*{10.4.~Slicing argument: reduction of Theorem~1.4 to 

7735: Theorem~1.2'} Thirdly, we may provide a new proof of the central

7736: Theorem 1.4 valid in CR dimension $m\geq 2$, in order to illustrate

7737: how our results in CR dimension 1 can be applied to the general case

7738: via slicing techniques. We shall see that Theorem~1.4 is a logical 

7739: consequence of Theorem~1.2'.

7740: 

7741: Let $M$, $M^1$ and $C$ be as in Theorem~1.4. To begin with, we shall

7742: treat the three notions of removability (CR-, $L^{\sf p}$- and

7743: $\mathcal{ W}$-) commonly. However, we remind that CR-removability is

7744: immediately reduced to $\mathcal{ W}$-removability thanks to

7745: Lemma~3.5, hence it suffices to consider only $L^{\sf p}$- and

7746: $\mathcal{ W}$-removability.

7747: 

7748: Arguing by contradiction, we see as in the

7749: previous parts of this paper that we lose essentially nothing if we

7750: consider $C$ to be the minimal nonremovable subset. Also, we may

7751: assume holomorphic extension to a wedge attached to $M\backslash C$. 

7752: As explained in \S3.16 it is enough to remove one single point of $C$.

7753: 

7754: As in \S3.16 ({\it see} especially the statement of Proposition~3.22),

7755: we can show that there is a $\mathcal{ C}^{2,\alpha}$-smooth

7756: hypersurface $H$ of $M$ which is generic in $\C^n$, transversal to

7757: $M^1$, and which has the following property: $H$ contains some point

7758: $p_1\in C$, and we can choose a small neighborhood of $H$ in $M$ in

7759: which $H$ has two connected open sides $H^+$ and $H^-$ such that $C$

7760: is contained in $H^-\cup\{p_1\}$ locally in a neighborhood of

7761: $p_1$. Notice that the hypersurface $H^1$ constructed in Lemma~3.21 is

7762: simply the intersection of $M^1$ with such a hypersurface $H$.

7763: 

7764: Next, in a neighborhood of $p_1$, we construct a local foliation of

7765: $M$ by generic submanifolds of CR dimension 1 (hence of dimension

7766: $n+1$) as follows. We notice that $\dim_\R H=2m+d-1$. We first choose

7767: a local $\mathcal{ C}^{2,\alpha}$-smooth foliation of $H$ by generic

7768: submanifolds of CR dimension 1 (hence of real dimension $n+1$) which

7769: we denote by $\widetilde{ M}_s$, where the transversal parameters $s=

7770: (s_1,\ldots,s_{m-2})\in \I_{m-2}(\rho_1)$ belong to a cube in

7771: $\R^{m-2}$ of radius some $\rho_1>0$. Afterwards, we extend this

7772: foliation to a local $\mathcal{ C}^{2,\alpha}$-smooth foliation

7773: $\widetilde{ M}_{s,t}$ of a neighborhood of $p_1$ in $M$, where $s\in

7774: \I_{m-2} (\rho_1)$, where $t\in \I(\rho_1)$ and where $\widetilde{

7775: M}_{ s,0}\equiv \widetilde{ M}_s$; if $m=2$, we notice that the

7776: parameter $s$ disappears and that we have only one real parameter $t$.

7777: Also, we can assume that $\widetilde{ M}_{s,t}$ is contained in 

7778: $H^+$ if and only if $t>0$.

7779: 

7780: By genericity, the submanifolds $\widetilde{ M}_{s,t}$ may be chosen

7781: in addition to be transversal to the one-codimensional submanifold

7782: $M^1\subset M$ containing the singularity $C$ with of course $p_1\in

7783: \widetilde{ M}_{0,0}$. Then for all parameters $s$ and $t$ the

7784: intersections $\widetilde{ M}_{s,t}^1:=M^1\cap \widetilde{ M}_{s,t}$

7785: are maximally real submanifolds of $\C^n$. Considering $\widetilde{

7786: M}_{s,t}^1$ as a maximally real one-codimensional submanifold of

7787: $\widetilde{ M}_{ s,t}$, a characteristic foliation is induced on each

7788: $\widetilde{ M}_{s,t}^1$. After contraction around $p_1$ we can

7789: assume that these characteristic foliations all have trivial topology:

7790: their leaves are the level sets of an $\R^{m-1}$-valued submersion. As

7791: the intersection $\widetilde{ M}_{0,0}^1 \cap C$ is the singleton

7792: $\{p_1\}$ (by construction of $H$), there exists $\varepsilon>0$ such

7793: that for all $s$ with $\vert s \vert < \varepsilon$ and all $t$ with

7794: $\vert t \vert < \varepsilon$, the closed subsets

7795: $C_{s,t}:=\widetilde{ M}_{s, t}^1\cap C$ are compact in $\widetilde{

7796: M}_{s, t}^1$. Notice that $C_{s,t}$ is even empty if 

7797: $t>0$, because $C$ is contained in $H^-\cup \{p_1\}$.

7798: The following simple fact shows that $C_{s,t}$ satisfies

7799: the nontransversality condition $\mathcal{ F}^c_{

7800: \widetilde{ M}_{s,t}^1}\{C_{s,t}\}$ of

7801: Theorem~1.2', for all $s$ with $\vert s \vert < \varepsilon$ and all

7802: $t$ with $-\varepsilon < t \leq 0$.

7803: 

7804: \def\thelemma{10.5}\begin{lemma}

7805: Let $\mathcal{ F }_\mathcal{ M}$ be a $\mathcal{ C}^{ 1,

7806: \alpha}$-smooth foliation by curves on some $m$-dimensional $\mathcal{

7807: C}^{ 2, \alpha}$-smooth real manifold $\mathcal{ M}$ defined by a

7808: surjective $\mathcal{ C}^{ 1, \alpha}$-smooth submersion $F: \mathcal{

7809: M} \rightarrow \I_{ m-1}( \rho_1)$. Then every compact set $\mathcal{

7810: C} \subset \mathcal{ M}$ satisfies the nontransversality condition

7811: $\mathcal{ F}_\mathcal{ M}\{ \mathcal{ C} \}$.

7812: \end{lemma}

7813: 

7814: \proof

7815: Let $\mathcal{ C}'$ be an arbitrary compact subset of $\mathcal{ C}$.

7816: As $\mathcal{ C}'$ is compact, there exists the smallest $\rho_2<

7817: \rho_1$ with $\mathcal{ C}' \subset F^{-1}( \overline{ \I_{ m-1}(

7818: \rho_2)})$. The semi-local projection $\pi_{ \mathcal{ F}_{\mathcal{

7819: M}}}$ along the leaves of $\mathcal{ F }$ may of course be identified

7820: with $F$. Thus, $\pi_{ \mathcal{ F}_{ \mathcal{ M } }} ( \mathcal{

7821: C}')$ is contained in $\overline{ \I_{m-1 }( \rho_2)}$ and meets the

7822: boundary $\partial \I_{ m-1} (\rho_2 )$ of the cube $\I_{m-1}

7823: (\rho_2)$. Also, by compactness, the set

7824: $\mathcal{ C}'$ cannot contain a fiber of $F$ in the whole. 

7825: This completes the proof.

7826: \endproof

7827: 

7828: We can now show that Theorem~1.4 is a logical consequence of

7829: Theorem~1.2'. In fact, we cannot insure that the generic submanifolds

7830: $\widetilde{ M}_{s,t}$ of CR dimension $1$ defined above are all

7831: globally minimal, hence it seems that Theorem~1.2' itself does not

7832: apply. However, we notice that the wedge attached to $M\backslash C$

7833: restricts to a wedge attached to $\widetilde{ M}_{s,t}\backslash

7834: C_{s,t}$, for all $s$ and $t$. Hence, we can observe that everything

7835: that was needed in the proof of Theorem~1.2' was the existence of a

7836: wedge attached to $M\backslash C$ to which holomorphic extension is

7837: already assumed. One can even formulate a slightly more general

7838: version of Theorem~1.2', where the global minimality assumption is

7839: replaced by the assumption of holomorphic extension to a wedge

7840: attached to $M\backslash C$. Of course, with this more general

7841: assumption, $M$ may consists of several CR orbits, but thanks to

7842: Lemma~3.5 about stability of CR orbits, one may check that the proof

7843: of the main Theorem~3.19 {\bf (i)} and of the main Proposition~5.12

7844: remain unchanged, in the case where holomorphic extension is assumed

7845: in a wedgelike domain over $M\backslash C$ and not only in wedgelike

7846: domains attached to the CR orbits of $M\backslash C$. 

7847: 

7848: Thus this slight generalization of Theorem~1.2' together with the

7849: observation made in Lemma~10.5 just above yield that for all $s$ with

7850: $\vert s \vert < \varepsilon$ and for all $t$ with $-\varepsilon < t

7851: \leq 0$, the closed subset $C_{ s,t}$ is $\mathcal{ W}$-removable in

7852: the generic submanifold $\widetilde{ M}_{s, t}$. We deduce that for

7853: every $(s, t)$ with $\vert s \vert < \varepsilon$ and $-\varepsilon <

7854: t \leq 0$, we get holomorphic extension from the given restricted

7855: wedge attached to $\widetilde{ M}_{ s, t} \backslash C_{ s, t}$ into

7856: an open wedge $\widetilde{ \mathcal{ W}}_{ s,t}$ attached $\widetilde{

7857: M}_{ s, t}$. Notice that this does not immediately achieves the proof,

7858: since the direction of the $\widetilde{ \mathcal{ W }}_{s, t}$ need

7859: not depend continuously on $(s, t)$. In fact, the proof of the slight

7860: generalization of Theorem~1.2' contains arguments (for example the

7861: localization near a very special point) which do not depend nicely on

7862: external parameters. Hence the attached wedges $\widetilde{ \mathcal{

7863: W}}_{s,t}$ may well be completely unrelated.

7864: 

7865: To overcome this difficulty we proceed in the following way, already

7866: argued in~\cite{ m2}, \cite{ mp1} (Lemma~2.7) in slightly different

7867: contexts. We first construct a regular family $A_{ x, v}( \zeta)$ of

7868: analytic discs attached to $\widetilde{ M}_{ 0, 0 }\cup \Omega$ whose

7869: size is small in comparison to the basis of the wedge $\widetilde{

7870: \mathcal{ W} }_{0, 0}$ and which sweep out a local wedge $\mathcal{ W}

7871: (A_{ x, v})$ of edge $\widetilde{ M}_{ 0, 0}$ at $p_1$. Here, the

7872: parameter $x\in \R^{ n+ 1}$ corresponds to translations in

7873: $\widetilde{ M}_{ 0, 0}$ and $v \in \R^{ n-2}$ to normal deformations

7874: in a neighborhood of the point $A_{ 0, 0} (-1)\in \Omega$. Deforming

7875: this family thanks to the flexibility of Bishop's equation, we

7876: construct a family $A_{x, v,s,t} (\zeta)$ attached to $\widetilde{

7877: M}_{ s,t} \cup \Omega$, still sweeping out a local wedge $\mathcal{ W

7878: }(A_{ x, v, s, t})$ of edge $\widetilde{ M}_{ s, t}$. This family is

7879: of class $\mathcal{ C}^{2,\alpha-0}$ with respect to all

7880: parameters. Using the $\mathcal{ W}$-removability of $C_{s,t}$, we can

7881: introduce for every $(s, t)$ a one-parameter deformation $\widetilde{ M

7882: }_{ s,t}^d$ of $\widetilde{ M}_{s,t}$ which is contained in the

7883: attached wedge $\widetilde{ W}_{ s,t}$ whenever $d>0$ and which

7884: coincides with $\widetilde{ M}_{s,t}$ when $d=0$. Thanks to the

7885: flexibility of Bishop's equation with parameters, we get a deformed

7886: family $A_{x,v, s, t: d}(\zeta)$ of analytic discs. Since the wedges

7887: $\widetilde{ W}_{ s,t}$ are {\it a priori}\, unrelated, we loose the

7888: smoothness with respect to all variables, including $d$. Fortunately,

7889: by an application of the continuity principle, for every $d>0$, we

7890: deduce holomorphic extension to the wedge generated by the family

7891: $A_{x, v,s, t:d} (\zeta)$. If we let $d$ tend to zero, fixing $(s,t)$,

7892: we obtain univalent holomorphic extension to the wedge generated by

7893: the family $A_{x,v,s,t}(\zeta)$. Finally, as $(s,t)$ varies, the

7894: wedges $\mathcal{ W}( A_{x,v,s,t})$ varies smoothly and covers a local

7895: wedge $\mathcal{ W}$ of edge $M$ at $p_1$. By the continuity

7896: principle, we may verify that we obtain univalent holomorphic

7897: extension to $\mathcal{ W}$.

7898: 

7899: Secondly, we explain how $L^{\sf p}$-removability of the

7900: point $p_1\in C$ in $M$ follows logically from the $L^{\sf

7901: p}$-removability of every $C_{s,t}$ in $\widetilde{ M}_{s,t}$. The

7902: main argument relies on the following simple but useful fact: if $M$

7903: is a generic CR submanifold of $\C^n$ and if $N\subset M$ is a lower

7904: dimensional submanifold which is itself a generic CR submanifold of

7905: $\C^n$ of positive CR dimension, then differentiable CR functions on

7906: $M$ obviously restrict to CR functions on $N$. More generally, a

7907: foliated version of this observation with lower regularity assumptions

7908: is as follows.

7909: 

7910: \def\thelemma{10.6}\begin{lemma}

7911: Let $M\subset\C^n$ be a generic submanifold of class $\mathcal{

7912: C}^{2,\alpha}$ and of CR dimension $m\geq 2$.

7913: 

7914: \begin{itemize}

7915: \item[{\bf (a)}]

7916: If $p_1\in M$ and

7917: if $M$ carries a local $\mathcal{ C }^{ 2, \alpha}$-smooth foliation

7918: by a family $N_u$ of generic submanifolds of CR dimension $1$ where

7919: $u \in \R^{m-1}$ is a small parameter and where

7920: $p_1 \in N_0$, then for every CR function $f

7921: \in L_{ loc}^{\sf p} (M)$, ${\sf p}\geq 1$, and for almost every $u\in

7922: \R^{ m-1}$, its restriction $f|_{ N_u}$ is an $L_{ loc}^{\sf p} ( N_u)$

7923: function which is CR on $N_u$.

7924: \item[{\bf (b)}]

7925: Conversely, if $p_1 \in M$ and if $M$ carries $m$ local $\mathcal{

7926: C}^{2, \alpha}$-smooth foliations by families $N_{u_j}^j$,

7927: $j=1, \dots,m$, $u_j\in \R^{m-1}$, of generic submanifolds of CR

7928: dimension $1$ satisfying $p_1\in N_0^j$ for $j=1,\dots,m$ and

7929: \def\theequation{10.7}\begin{equation}

7930: T_{p_1}N_0^1+\cdots + T_{p_1}N_0^m =T_{p_1}M, 

7931: \end{equation}

7932: then a function $f\in L_{ loc}^{ \sf p }(M)$ is CR in a neighborhood of

7933: $p_1$ if and only if for all $j =1, \dots,m$ and for almost every

7934: $u_j \in \R^{ m-1}$, its restriction $f\vert_{ N_{ u_j}^j}$ is CR on

7935: $N_{ u_j }^j$.

7936: \end{itemize}

7937: \end{lemma}

7938: 

7939: \proof 

7940: Of course, property {\bf (a)} only makes sense for an everywhere

7941: defined representative of $f$ and the nullset of excluded parameters

7942: $t$ depends on the choice of the representative of $f$.

7943: 

7944: To establish {\bf (a)}, we choose a small box-neighborhood $U\cong

7945: \I_{ m-1}( \rho_1)\times \widetilde{ N}$, where $\I_{ m-1}(\rho_1)$ is

7946: a cube of some positive radius $\rho_1 >0$ in $\R^{ m-1}$, such that

7947: every plaque $\{ \upsilon \}\times \widetilde{N}$ is an open subset of

7948: some leaf $N_u$. By the $L^{\sf p}$ version of the

7949: approximation theorem ({\it cf.}~\cite{j5}, \cite{p1}, \cite{mp1}),

7950: the restriction $f|_U$ is the limit in the $L^{ \sf p}$ norm of the

7951: restrictions of holomorphic polynomials $(P_\nu )_{ \nu \in

7952: \N}$. Thanks to Fubini's theorem, we deduce that for almost every

7953: $\upsilon \in \I_k ( \rho_1)$, the restriction $P_\nu |_{ \{\upsilon

7954: \} \times \widetilde{ N }}$ converges in $L^{\sf p}$ norm to $f|_{ \{

7955: \upsilon \} \times \widetilde{ N }}$. Hence for such parameters

7956: $\upsilon$, the restriction $f \vert_{\{\upsilon \} \times \widetilde{

7957: N}}$ is CR, which completes the proof of {\bf (a)}.

7958: 

7959: To establish {\bf (b)}, we observe first that the ``only if'' part is

7960: a direct consequence of {\bf (a)}. To prove the ``if'' part, we may

7961: introduce for every $j=1,\dots,m$ and for every $u_j\in \R^{m-1}$ a

7962: $(0,1)$ vector field $\overline{L}_{u_j}^j$ tangent to $N_{u_j}^j$ and

7963: $\mathcal{ C}^{1,\alpha}$-smoothly parameterized by $u_j$. The

7964: geometric assumption~\thetag{ 10.7} entails that the $m$ vector fields

7965: $\overline{L}_{u_1}^1,\dots,\overline{L}_{u_m}^m$ generate the CR

7966: bundle $T^{0,1}M$ in a neighborhood of $p_1$. By assumption, the

7967: $L_{loc}^{\sf p}$ function $f$ is annihilated in the distributional

7968: sense by these $m$ vector field, hence it is CR. This completes the

7969: proof of {\bf (b)}.

7970: \endproof

7971: 

7972: We can now prove that the $L^{ \sf p}$-removability of $C$ in $M$

7973: follows from an application of Theorem~1.2'. Let a function $f\in

7974: L^{\sf p}_{ loc}(M)$ which is CR on $M \backslash C$. Coming back to

7975: the construction of the submanifolds $\widetilde{ M}_{s,t}$ achieved

7976: in the paragraphs before Lemma~10.5, it is clear that for almost every

7977: $(s,t) \in \R^{ m-1}$, the restriction $f\vert_{ \widetilde{

7978: M}_{s,t}}$ is $L_{ loc}^{ \sf p}$-integrable. More generally,

7979: proceeding as in the paragraph before Lemma~10.5, we may construct $m$

7980: such families $\widetilde{ M}_{j; s_j,t_j}$ for $j=1, \dots, m$ with

7981: $p_1\in \widetilde{ M}_{ j; 0,0}$ and

7982: \def\theequation{10.8}\begin{equation}

7983: T_{p_1}\widetilde{ M}_{1;s_1,t_1}+\cdots+

7984: T_{p_1}\widetilde{ M}_{m;s_m,t_m}=

7985: T_{p_1}M,

7986: \end{equation}

7987: without changing the conclusion that the corresponding closed subsets

7988: $C_{s_j, t_j}^j$ are $L^{\sf p}$-removable in $\widetilde{ M }_{j;

7989: s_j,t_j}$ for $j =1, \dots, m$. Applying Lemma~10.7 just above, we

7990: finally deduce that $f$ is CR on $M$, as desired.

7991: 

7992: This completes the description of the reduction of

7993: Theorem~1.4 to Theorem~1.2' via a slicing argument.

7994: 

7995: \subsection*{10.9.~Version of Theorem~1.4 in an open subset 

7996: of $\C^n$} To conclude this section, we remind that in the end of the

7997: proof of Proposition~5.12 in Cases {\bf ($\mathbf{ I_1}$)} and {\bf

7998: ($\mathbf{ I_2}$)}, we came down to the removability of a proper

7999: closed subset $\mathcal{ C}_{p_1}$ of a one-codimensional submanifold

8000: $\mathcal{ M}_{p_1}$ of an open subset of $\C^n$ ($n\geq 2$), namely

8001: the wedge $\mathcal{ W}_{p_1}$ (remind {\sc Figure~18} above), which

8002: amounts exactly to prove Theorem~1.4 in the case where the generic

8003: submanifold $M$ is replaced by an open subset of $\C^n$. We may

8004: formulate this result as the following lemma. To our knowledge, its

8005: first known proof is given in~\cite{ j4}. Here, we provide a slightly

8006: different proof, using half-attached analytic discs.

8007: 

8008: \def\thelemma{10.10}\begin{lemma}

8009: Let $\mathcal{ D}\subset \C^n$ be a domain, let $\mathcal{ M}^1

8010: \subset \mathcal{ D}$ be a connected $\mathcal{ C}^{2,\beta}$-smooth

8011: hypersurface with $0 < \beta < 1$ and let $\mathcal{ C}$ be a proper

8012: closed subset of $\mathcal{ M}^1$ which does not contain any CR orbit

8013: of $\mathcal{ M}^1$. Then for every holomorphic function $f \in

8014: \mathcal{ O} \left( \mathcal{ D} \backslash \mathcal{ C} \right)$,

8015: there exists a holomorphic function $F \in \mathcal{ O}(\mathcal{ D})$

8016: such that $F \vert_{\mathcal{ D} \backslash \mathcal{ C }} = f$.

8017: \end{lemma}

8018: 

8019: \proof

8020: We summarize the proof, which anyway is very similar to the proof of

8021: Proposition~3.22 delineated in \S10.3 above. Reasoning by

8022: contradiction and constructing a supporting hypersurface, we come down

8023: to the local removability of a single point $p_1$ in a geometric

8024: situation analogous to the one described in Proposition~3.22, with $M$

8025: replaced by the domain $\mathcal{ D}$, with $M^1$ replaced by

8026: $\mathcal{ M}^1$ and with a generic $\mathcal{ C}^{2,\beta}$-smooth

8027: submanifold $\mathcal{ H}^1\subset \mathcal{ M}^1$ such that, locally

8028: in a neighborhood of $p_1$, we have $\mathcal{ C}\subset (\mathcal{

8029: H}^1)^-\cup \{p_1\}$, where we use the same notation $\mathcal{ C}$

8030: for the smallest non-removable subset of the original $\mathcal{ C}$.

8031: 

8032: Notice that $\mathcal{ H}^1$ is of codimension $2$. Let $\mathcal{

8033: K}^1\subset \mathcal{ H}^1$ be a maximally real submanifold passing

8034: through $p_1$. We may translate $\mathcal{ K}^1$ in $\mathcal{ M}^1$

8035: by means of a small parameter $t\in \R^{n-1}$ and then $\mathcal{

8036: M}^1$ in $\mathcal{ D}$ by means of a small parameter $u\in \R$.

8037: Following Section~7, we then construct a small family of analytic

8038: discs $\mathcal{ A}_{x,v,t,u}^1(\zeta)$ half-attached to the

8039: ``translations'' $\mathcal{ K}_{t,u}^1$. Nine properties analogous to

8040: properties {\bf ($\mathbf{ 1_1}$)} to {\bf ($\mathbf{ 9_1}$)} of

8041: Lemmas~7.12 and~8.3 are then satisfied and we conclude the proof 

8042: in essentially the same way as in Section~9 above.

8043: \endproof

8044: 

8045: \section*{\S11~$\mathcal{ W}$-removability implies

8046: $L^{\sf p}$-removability}

8047: 

8048: \subsection*{11.1.~Preliminary}

8049: This section is devoted to prove Lemma~3.15 about $L^{\sf

8050: p}$-removability of the proper closed subset $C\subset M^1$, granted

8051: it is $\mathcal{ W}$-removable. More generally, we shall establish the

8052: $L^{\sf p}$-removability of certain proper closed subsets $\Phi$ of

8053: $M$ that are nullsets with respect to the Lebesgue measure of $M$.

8054: 

8055: As a preliminary, we remind that if $M'$ is a globally minimal

8056: $\mathcal{ C}^{2,\alpha}$-smooth generic submanifold of $\C^n$ of CR

8057: dimension $m \geq 1$ and of codimension $d= n-m\geq 1$, there exists a

8058: wedge $\mathcal{ W}'$ attached to $M'$ constructed by means of

8059: analytic discs successively glued to $M'$ and to conelike submanifolds

8060: attached to $M'$ consisting of parametrized families of pieces of

8061: analytic discs. By means of the approximation theorem of~\cite{ bt},

8062: one deduces classically that continuous CR functions on $M'$ extend

8063: holomorphically to $\mathcal{ W}'$, and continuously to $M'\cup

8064: \mathcal{ W}'$.

8065: 

8066: For the holomorphic extension of the $L_{loc}^{\sf p}$ CR functions to

8067: a wedge attached to $M'$, some supplementary routine, though not

8068: obvious, work has to be achieved. Firstly, using a convolution with

8069: Gauss' kernel as in~\cite{bt}, one shows that on a $\mathcal{

8070: C}^2$-smooth generic submanifold $M'$ of $\C^n$, every $L_{loc}^{\sf

8071: p}$ CR function on $M'$ is locally the limit, in the $L^{\sf p}$ norm,

8072: of a sequence of polynomials ({\it see} Lemma~3.3 in~\cite{j5}). In

8073: the case where $M'$ is a hypersurface, studied in~\cite{j5}, the wedge

8074: $\mathcal{ W}'$ is in fact a one-sided neighborhood attached to $M'$,

8075: which we will denote by $\mathcal{ S}'$. The theory of Hardy spaces on

8076: the unit disc transfers to parameterized families of small analytic

8077: discs glued to $M'$ which cover local one-sided neighborhoods of a

8078: hypersurface, provided the boundaries of these discs foliate an open

8079: subset of $M'$. Using in an essential way L.~Carleson's imbedding

8080: theorem, B.~J\"oricke established in~\cite{j5} that every $L^{\sf

8081: p}_{loc}$ CR function on a globally minimal $\mathcal{ C}^2$-smooth

8082: hypersurface $M'$ extends holomorphically in the Hardy space $H^{\sf

8083: p}(\mathcal{ S}')$ of holomorphic functions defined in the one-sided

8084: neighborhood $\mathcal{ S}'$, with

8085: $L^{\sf p}$ boundary values on the hypersurface $M'$. 

8086: In his thesis~\cite{p1}, the second author

8087: of the present paper has built the theory in higher codimension,

8088: introducing the Hardy space $H^{\sf p}(\mathcal{ W}')$ of functions

8089: holomorphic in the wedge $\mathcal{ W}'$ attached to $M'$, with

8090: $L^{\sf p}$ boundary values on the edge $M'$.

8091: 

8092: At present, these background statements about holomorphic

8093: extendability of $L_{loc}^{\sf p}$ CR functions on globally minimal

8094: generic submanifolds may be reproved in a more elegant way than by

8095: going through the rather complicated technology dispersed in the

8096: articles~\cite{tu1}, \cite{tu2}, \cite{m1}, \cite{j2}, thanks to a

8097: simplification of the wedge extendability theorem obtained recently by

8098: the second author of this paper, which treats in an unified way local

8099: and global minimality. We refer the reader to the work in

8100: preparation~\cite{p3} for a substantial cleaning of the theory.

8101: 

8102: \subsection*{11.2.~$L^{\sf p}$-removability of nullsets}

8103: Let us say that a subset $\Phi$ of a $\mathcal{ C}^{2,\alpha}$-smooth

8104: generic submanifold is {\sl stably $\mathcal{ W}$-removable} if it is

8105: $\mathcal{ W}$-removable on every compactly supported sufficiently

8106: small $\mathcal{ C}^{2,\alpha}$-smooth deformation $M^d$ of $M$

8107: leaving $\Phi$ fixed. In the situations of Theorems~1.2' and~1.4, the

8108: assumptions of Lemma~11.3 just below are satisfied with $\Phi=C$,

8109: taking account of the fact that we have already established the

8110: $\mathcal{ W}$-removability of $C$ and that for logical reasons only,

8111: the closed set $C$ in the statements of Theorems~1.2' and~1.4 is

8112: obviously stably removable.

8113: 

8114: \def\thelemma{11.3}\begin{lemma}

8115: Let $M$ be a $\mathcal{ C}^{2, \alpha}$-smooth generic submanifold of

8116: $\C^n$ of CR dimension $m \geq 1$ and of codimension $d= n-m\geq 1$,

8117: hence of dimension $(2m+ d)$, let $\Phi \subset M$ be a nonempty

8118: proper closed subset whose $(2m +d)$-dimensional Hausdorff measure is

8119: equal to zero. Assume that $M \backslash \Phi$ is globally minimal and

8120: let $\mathcal{ W}$ be a wedge attached to $M\backslash \Phi$ such that

8121: every function in $L_{loc}^{\sf p}(M) \cap CR(M\backslash \Phi)$

8122: extends holomorphically as a function in the Hardy space $H^{\sf

8123: p}(\mathcal{ W})$. If $\Phi$ is stably 

8124: $\mathcal{ W}$-removable, then $\Phi$

8125: is $L^{\sf p}$-removable.

8126: \end{lemma}

8127: 

8128: Before giving the proof,

8129: let us summarize intuitively the reason why this strong $L^{\sf

8130: p}$-removability result Lemma~11.3 holds. Indeed, let $f\in L_{loc

8131: }^{\sf p}(M) \cap CR (M \backslash \Phi)$. As soon as wedge extension

8132: over points of $\Phi$ is known, thanks to the fact that we can deform

8133: $M$ over $\Phi$ in the wedgelike domain, thus erasing the singularity

8134: $\Phi$, we get a $L_{loc}^{ \sf p}$ CR function $f^d$ on the deformed

8135: manifold $M^d$, without singularities anymore, and in additition, we

8136: can let the deformation $M^d$ tend to $M$ with a uniform with $L^{\sf

8137: p}$ control of the extension $f^d$, which therefore tends to a CR

8138: extension of $f$ through $\Phi$.

8139: 

8140: \proof

8141: First of all, we remind that for every ${\sf p}$ with $1\leq {\sf p}

8142: \leq \infty$, the space $L_{loc}^{\sf p}(M)$ is contained in

8143: $L_{loc}^1(M)$. We claim that it follows that $\Phi$ is $L^{\sf

8144: p}$-removable for every ${\sf p}$ with $1\leq {\sf p} \leq \infty$ if

8145: and only if $\Phi$ is $L^1$-removable. Indeed, suppose that $\Phi$ is

8146: $L^1$-removable, namely for every function $f\in L_{loc}^1(M)\cap {\rm

8147: CR}(M\backslash \Phi)$, and every $\mathcal{ C}^1$-smooth

8148: $(n,m-1)$-form with compact support, we have $\int_M \, f \cdot

8149: \overline{\partial} \psi =0$. In particular, since $L_{loc}^{\sf p}$

8150: is contained in $L_{loc}^1$ by H\"older's inequality, this property

8151: holds for every function $g\in L_{loc}^{\sf p}(M) \cap CR(M\backslash

8152: \Phi)$, hence $\Phi$ is $L^{\sf p}$-removable, as

8153: claimed. Consequently, it suffices to show that $\mathcal{

8154: W}$-removability implies $L^1$-removability.

8155: 

8156: Let $f\in L_{loc}^1(M\backslash \Phi)\cap L^1(M)$ be an arbitrary

8157: function. The goal is to show that $f$ is in fact CR on $\Phi$. Of

8158: course, it suffices to show that $f$ is CR locally at every point of

8159: $\Phi$. So, we fix an arbitrary point $q\in \Phi$. If $\psi$ is an

8160: arbitrary $(n,m-1)$-form of class $\mathcal{ C}^1$ supported in a

8161: sufficiently small neighborhood of $q$, we have to prove that $\int_M

8162: \, f \cdot \overline{\partial} \psi =0$.

8163: 

8164: We may also fix a small open polydisc $\mathcal{ V}_q$ centered at

8165: $q$. We shall first argue that we can assume that the $L_{loc}^1$

8166: function $f$ is holomorphic in a neighborhood of $\left(M\backslash

8167: \Phi\right)\cap \mathcal{ V}_q$ in $\C^n$. Indeed, since $M \backslash

8168: \Phi$ is globally minimal, there exists a wedge $\mathcal{ W}$

8169: attached to $M\backslash \Phi$ such that every $L_{loc}^1$ CR function

8170: on $M \backslash \Phi$, and in particular $f$, extends holomorphically

8171: as a function which belongs to the Hardy space $H^1(\mathcal{ W})$. By

8172: slightly deforming $\left(M\backslash \Phi\right) \cap \mathcal{ V}_q$

8173: into $\mathcal{ W}$ along Bishop discs glued to $M\backslash \Phi$,

8174: keeping $\Phi$ fixed, using the theory of Hardy spaces in wedges

8175: developed in~\cite{p1}, we may obtain the following deformation result

8176: with $L^1$ control, a statement which is a particular case of

8177: Proposition~1.16 in~\cite{mp1}.

8178: 

8179: \def\theproposition{11.4}\begin{proposition}

8180: For every $\varepsilon > 0$, every $\beta <\alpha$, there exists a

8181: small $\mathcal{ C}^{ 2, \beta}$-smooth deformation $M^d$ of $M$ with

8182: support contained in $\overline{ \mathcal{ V }_q}$ and there exists a

8183: function $f^d \in L_{ loc}^1 \left(M^d \right)\cap CR \left( M^d

8184: \backslash \Phi \right)$, such that

8185: \begin{itemize}

8186: \item[{\bf (1)}]

8187: $M^d \cap \mathcal{V}_q \supset \Phi \cap \mathcal{ V}_q\ni q$.

8188: \item[{\bf (2)}]

8189: $\left(M^d\backslash \Phi \right)\cap \mathcal{ V}_q 

8190: \subset \mathcal{ W}\cap \mathcal{ V}_q$.

8191: \item[{\bf (3)}]

8192: $f^d$ is holomorphic in the neighborhood

8193: $\mathcal{ W}\cap \mathcal{ V}_q$ of $(M^d\backslash \Phi)\cap 

8194: \mathcal{ V}_q$ in $\C^n$.

8195: \item[{\bf (4)}]

8196: $M\cap \mathcal{ V}_q$ and $M^d\cap \mathcal{ V}_q$ are graphed over

8197: the same $(2m+d)$ linear real subspace and $\left\vert\left\vert M^d\cap

8198: \mathcal{ V}_q - M \cap \mathcal{ V}_q

8199: \right\vert\right\vert_{\mathcal{ C}^{2,\beta}} \leq \varepsilon$.

8200: \item[{\bf (5)}]

8201: The volume forms of $M\cap \mathcal{ V}_q$ and of $M^d \cap \mathcal{

8202: V}_q$ may be identified and $\left\vert f-f^d

8203: \right\vert_{L^1(M\cap \mathcal{ V}_q)} \leq \varepsilon$. 

8204: \end{itemize}

8205: \end{proposition}

8206: 

8207: Since it will suffice to have a control of the

8208: deformation $M^d$ only in $\mathcal{ C}^2$ norm, we shall

8209: replace $\mathcal{ C}^{2,\beta}$ and $\mathcal{ C}^{1,\beta}$

8210: by $\mathcal{ C}^2$ and $\mathcal{ C}^1$ in the sequel.

8211: 

8212: Let us be more explicit about conditions {\bf (4)} and {\bf

8213: (5)}. Without loss of generality, we can assume that in coordinates

8214: $(z,w)= (x+iy, u+iv)\in \C^m\times \C^d$ centered at $q$, we have

8215: $T_qM= \{v=0\}$, hence the generic submanifolds $M$ and $M^d$ are

8216: represented locally by vectorial equations $v=\varphi(x,y,u)$ and $v=

8217: \varphi^d (x,y,u)$, where $\varphi$ and $\varphi^d$ are defined in the

8218: real cube $\I_{2m+d}(2\rho_1)$, for some small $\rho_1>0$ and that

8219: $\mathcal{ V}_q$ is the polydisc $\Delta_n(\rho_1)$ of radius

8220: $\rho_1$. Then condition {\bf (4)} simply means that $\vert \vert

8221: \varphi^d - \varphi \vert \vert_{\mathcal{ C}^2 (\I_{2m+d}

8222: (\rho_1) )}\leq \varepsilon$ and condition {\bf (5)} is clear if we

8223: choose $dxdydu$ as the volume form on $M$ and on $M^d$.

8224: 

8225: Suppose that for every $\varepsilon>0$ and for every 

8226: deformation $M^d$, we can show that

8227: the function $L_{loc}^1$ function $f^d$ on $M^d$ is in fact CR over

8228: $M^d\cap \Delta_n(\rho_1)$. Then we claim that $f$ is CR in a

8229: neighborhood of $q$.

8230: 

8231: Indeed, to begin with, let us denote by

8232: $\overline{L}_1,\dots,\overline{L}_m$ a basis of $(0,1)$ vector fields

8233: tangent to $M$, having coefficients depending on the first order

8234: derivatives of $\varphi$. More precisely, in slightly abusive matrix

8235: notation, we can choose the basis $\overline{L}:=

8236: \frac{\partial}{\partial \bar z}+2(i-\varphi_u)^{-1} \, \varphi_{\bar

8237: z} \frac{\partial}{\partial \bar w}$. Let us denote this basis

8238: vectorially by $\overline{L}=\frac{\partial}{\partial \bar z}+ A \,

8239: \frac{\partial}{\partial \bar w}$. To compute the formal adjoint of

8240: $\overline{L}$ with respect to the local Lebesgue measure $dx dy du$

8241: on $M$, we choose two $\mathcal{ C}^1$-smooth functions $\psi$, $\chi$

8242: of $(x,y,u)$ with compact support in $\I_{2m+d}(\rho_1)$. Then the

8243: integration by part $\int \overline{L} (\psi) \cdot \chi \cdot dxdydu=

8244: \int\, \psi\cdot {}^T \overline{L}

8245: (\chi)\cdot dx dy du$ yields the explicit

8246: expression ${}^T\overline{L} (\chi):= -\overline{L}(\chi)-A_{\bar w}

8247: \cdot \chi$ of the formal adjoint of $\overline{L}$.

8248: 

8249: It follows immediately that if we denote by ${}^T(\overline{L}^d)$ the

8250: formal adjoint of the basis of CR vector fields tangent to $M^d$, then

8251: we have an estimate of the form $\vert \vert {}^T(\overline{L}^d) -

8252: {}^T(\overline{L})\vert \vert_{ \mathcal{ C}^1} \leq {\rm C} \cdot

8253: \varepsilon$, for some constant $C>0$. Recall that $f^d$ is assumed to

8254: be CR in $M^d\cap \Delta_n(\rho_1)$. Equivalently, we have $\int \,

8255: f^d \cdot \, {}^T(\overline{L}^d)(\psi)\cdot dxdydu=0$ for every

8256: $\mathcal{ C}^1$-smooth function $\psi$ with compact support in the

8257: cube $\I_{2m+d}(\rho_1)$. Then we deduce that (some explanation

8258: follows)

8259: \def\theequation{11.5}\begin{equation}

8260: \left\{

8261: \aligned

8262: {}

8263: &

8264: \left\vert

8265: \int\, f\cdot {}^T\overline{L}(\psi) \cdot dxdydu \right\vert=

8266: \left\vert

8267: \int\, 

8268: \left[

8269: f\cdot {}^T\overline{L}(\psi)-

8270: f^d\cdot {}^T(\overline{L}^d)(\psi)

8271: \right] \cdot dxdydu \right\vert \\

8272: & \

8273: \leq 

8274: \left\vert

8275: \int\, 

8276: \left[

8277: f\cdot {}^T\overline{L}(\psi)-f\cdot {}^T(\overline{L}^d)(\psi)+

8278: f\cdot {}^T(\overline{L}^d)(\psi)-

8279: f^d\cdot {}^T(\overline{L}^d)(\psi)

8280: \right] \cdot dxdydu \right\vert

8281: \\

8282: & \

8283: \leq 

8284: C_1(\psi) \cdot \varepsilon \cdot 

8285: \int_{\I_{2m+d}(\rho_1)} \, 

8286: \vert f \vert \cdot dxdydu + C_2(\psi) \cdot

8287: \int_{\I_{2m+d}(\rho_1)}\, 

8288: \vert f - f^d \vert \cdot dxdydu

8289: \\

8290: & \

8291: \leq C(\psi,f,\rho_1)\cdot \varepsilon,

8292: \endaligned\right. 

8293: \end{equation}

8294: taking account of property {\bf (5)} of Proposition~11.4 for the

8295: passage from the third to the fourth line, where $C(\psi,f,\rho_1)$ is

8296: a positive constant depending only on $\psi$, $f$ and $\rho_1$. As

8297: $\varepsilon$ was arbitrarily small, it follows that $\int\, f\cdot

8298: {}^T\overline{L}(\psi) \cdot dxdydu =0 $ for every $\psi$, namely $f$

8299: is CR on $M\cap \Delta_n(\rho_1)$, as was claimed.

8300: 

8301: It remains to show that $f^d$ is CR on $M^d\cap \Delta_n(\rho_1)$.

8302: First of all, we need some observations. For every compactly

8303: supported small deformation $M^d$ stabilizing $\Phi$, the wedge

8304: $\mathcal{ W}$ attached to $M\backslash \Phi$ is still a wedge

8305: attached to $M^d\backslash \Phi$. In addition, this wedge contains a

8306: neighborhood of $\left(M^d \backslash \Phi\right)\cap

8307: \Delta_n(\rho_1)$ in $\C^n$ by property {\bf (3)} of Proposition~11.4.

8308: As $\Phi$ was supposed to be stably removable, it follows that there

8309: exists a wedge $\mathcal{ W}_1$ attached to $M^d$ (including points of

8310: $\Phi$) to which holomorphic functions in $\mathcal{ W}$ extend

8311: holomorphically.

8312: 

8313: Consequently, replacing $M^d\cap \Delta_n(\rho_1)$ by $M$, we are led

8314: to prove the following lemma, which, on the geometric side, is totally

8315: similar to Lemma~11.3, except that the wedge $\mathcal{ W}$ attached

8316: to $M\backslash \Phi$ appearing in the formulation of Lemma~11.3 is

8317: now replaced by a neighborhood $\Omega$ of $M\backslash \Phi$ in

8318: $\C^n$.

8319: 

8320: \def\thelemma{11.6}\begin{lemma} 

8321: Let $M$ be a $\mathcal{ C}^{2, \alpha}$-smooth generic submanifold of

8322: $\C^n$ of CR dimension $m \geq 1$ and of codimension $d= n-m\geq 1$, 

8323: let $\Phi \subset M$ be a nonempty

8324: proper closed subset whose $(2m +d)$-dimensional Hausdorff measure is

8325: equal to zero. Let $\Omega$ be a neighborhood of $M\backslash \Phi$ in

8326: $\C^n$ and let $\mathcal{ W}_1$ be a wedge attached to $M$, including

8327: points of $\Phi$.

8328: Let $f\in L_{loc}^1(M)$ and assume that its restriction to

8329: $M\backslash \Phi$ extends as a holomorphic function $f'\in \mathcal{

8330: O}(\Omega \cup \mathcal{ W}_1)$. Then $f$ is CR all over $M$.

8331: \end{lemma}

8332: 

8333: \proof

8334: It suffices to prove that $f$ is CR at every point of $\Phi$. Let

8335: $q\in \Phi$ be arbitrary and let $\mathcal{ W}_q$ be a local wedge of

8336: edge $M$ at $q$ which is contained in $\mathcal{ W}_1$. Without loss

8337: of generality, we can assume that in coordinates

8338: $(z,w)=(x+iy,u+iv)\in\C^m\times \C^d$ vanishing at $q$ with

8339: $T_qM=\{v=0\}$, the generic submanifold $M$ is represented locally in

8340: the polydisc $\Delta_n(\rho_1)$ by $v=\varphi(x,y,u)$ for some

8341: $\mathcal{ C}^{2,\alpha}$-smooth $\R^d$-valued mapping $\varphi$

8342: defined on the real cube on $\I_{2m+d}(\rho_1)$. First of all, we

8343: construct a family of analytic discs half attached to $M$ whose

8344: interior is contained in the local wedge $\mathcal{ W}_q\subset

8345: \mathcal{ W}_1$.

8346: 

8347: \def\thelemma{11.7}\begin{lemma}

8348: There exists a family of analytic

8349: discs $A_s(\zeta)$, with $s\in \R^{2m+d-1}$, $\vert s \vert \leq

8350: 2\delta$ for some $\delta>0$, and $\zeta \in\overline{\Delta}$, which

8351: is of class $\mathcal{ C}^{2,\alpha-0}$ with respect to all variables,

8352: such that

8353: \begin{itemize}

8354: \item[{\bf (1)}]

8355: $A_0(1)=q$.

8356: \item[{\bf (2)}]

8357: $A_s(\overline{\Delta})\subset \Delta_n(\rho_1)$. 

8358: \item[{\bf (3)}]

8359: $A_s(\Delta) \subset \mathcal{ W}_q\cap \Delta_n(\rho_1)$.

8360: \item[{\bf (4)}]

8361: $A_s(\partial^+\Delta)\subset M$.

8362: \item[{\bf (5)}]

8363: $A_s(i)\in M\backslash \Phi$ and $A_s(-i)\in M\backslash\Phi$

8364: for all $s$. 

8365: \item[{\bf (6)}]

8366: The mapping

8367: $[-2\delta,2\delta]^{2m+d-1}\times [-\pi/2,\pi/2]

8368: \ni (s,\theta)\longmapsto 

8369: A_s(e^{i\theta})\in M$ is an embedding onto 

8370: a neighborhood of $q$ in $M$.

8371: \item[{\bf (7)}]

8372: There exists $\rho_2>0$ such that the image of

8373: $[- \delta, \delta]^{ 2m+d-1} \times [-\pi /4,\pi /4]$ through 

8374: this mapping contains $M \cap \Delta_n (\rho_2)$.

8375: \end{itemize}

8376: \end{lemma} 

8377: 

8378: \proof

8379: Let $M^1$ be a $\mathcal{ C}^{2,\alpha}$-smooth maximally real

8380: submanifold of $M$ passing through $q$ such that $M^1\cap \Phi$ is of

8381: zero measure with respect to the Lebesgue measure of $M^1$. Let $t\in

8382: \R^d$ and include $M^1$ in a parametrized family of maximally real

8383: submanifolds $M_t^1$ which foliates a neighborhood of $q$ in $M$.

8384: Starting with a family of analytic discs $A_{c,x,v}^1(\zeta)$ which

8385: are half-attached to $M^1$ as constructed in Lemma~7.12 above, 

8386: we first choose the rotation

8387: parameter $v_0$ and a sufficiently small scaling factor $c_0$ in order

8388: that $A_{c_0,0,v_0}^1(\pm i)$ does not belong to $\Phi$. In fact, this

8389: can be done for almost every $(c_0,v_0)$, because the

8390: mapping $(c,v)\mapsto A_{c,0,v}^1(\pm i)$ is of rank $n$ at

8391: every point $(c,v)$ with $c\neq 0$ and $v\neq 0$. In addition, we

8392: adjust the rotation parameter $v_0$ in order that the vector $Jv_0$

8393: points inside a proper subcone of the cone which defines the wedge

8394: $\mathcal{W}_q$. If the scaling parameter $c$ is sufficiently small,

8395: this implies that $A_{c_0,0,v_0}^1(\Delta)$ is contained in $\mathcal{

8396: W}_q\cap \Delta_n(\rho_1)$, as in Lemma~8.3 above. The translation

8397: parameter $x$ runs in $\R^n$ and we may select a $(n-1)$-dimensional

8398: parameter subspace $x'$ which is transversal in $M^1$ to the half

8399: boundary $A_{c_0, 0,v_0}^1(\partial^+ \Delta)$. With such a choice,

8400: there exists $\delta>0$ such that the mapping $[-2 \delta,

8401: 2\delta]^{n-1}\times [-\pi/2, \pi/2 ]\ni (x', \theta) \longmapsto

8402: A_{c_0, x', v_0}^1(e^{i\theta})$ is a diffeomorphism onto a

8403: neighborhood of $q$ in $M^1$. Finally, using the stability of Bishop's

8404: equation under perturbations, we can deform this family of discs by

8405: requiring that it is half attached to $M_t^1$, thus obtaining a family

8406: $A_s(\zeta) := A_{c_0, x',v_0, t}^1( \zeta)$ with $s:= (x',t) \in\R^{

8407: 2m+d-1}$. Shrinking $\delta$ if necessary, we can check as

8408: in the proof of Lemma~8.3 {\bf ($\mathbf{9_1}$)} that

8409: condition {\bf (5)} holds. This completes the proof.

8410: \endproof

8411: 

8412: Let now $f \in L_{ loc}^1 (M)$ and let $f' \in \mathcal{ O} ( \Omega

8413: \cup \mathcal{ W }_1)$. Thanks to the foliation propery {\bf (6)} of

8414: Lemma~11.7, it follows from Fubini's theorem that for almost every

8415: translation parameter $s$, the mapping $e^{ i\theta} \mapsto f \left(

8416: A_s (e^{ i\theta })\right)$ defines a $L^1$ function on $\partial^+

8417: \Delta$. In addition, the restriction of the function $f' \in

8418: \mathcal{ O}( \Omega\cup \mathcal{ W}_1)$ to the disc $A_s( \Delta)

8419: \subset \mathcal{ W}_q \subset \mathcal{ W}_1$ yields a holomorphic

8420: function $f' \left( A_s( \zeta) \right)$ in $\Delta$.

8421: 

8422: \def\thelemma{11.8}\begin{lemma}

8423: For almost every $s$ with $\vert s \vert \leq 2 \delta$, the function

8424: $f' \left( A_s (\zeta) \right)$ belongs to the Hardy space $H^1 (

8425: \Delta )$.

8426: \end{lemma}

8427: 

8428: \proof

8429: Indeed, for almost every $s$, the restriction $f\left (A_s(e^{

8430: i\theta})\right)$ belongs to $L^1(\partial^+ \Delta)$. We can also

8431: assume that for almost every $s$, the intersection $\Phi \cap

8432: A_s(\partial^+ \Delta)$ is of zero one-dimensional measure. By the

8433: assumption of Lemma~11.6, the restriction of $f\circ A_s$ and of

8434: $f'\circ A_s$ to $\partial^+\Delta \backslash \Phi$ coincide. Recall

8435: that $\partial^-\Delta=\{\zeta\in \partial \Delta: \, {\rm Re}\, \zeta

8436: \leq 0\}$. Since $A_s( \pm i)$ does not belong to $\Phi$ and since

8437: $A_s \left(e^{i\theta}\right)$ belongs to $\mathcal{ W}_q$ for all

8438: $\theta$ with $\pi/2 < \vert \theta \vert \leq \pi$, it follows that

8439: $f \circ A_s \vert_{\partial^+ \Delta}$ and $f' \circ A_s \vert_{

8440: \partial^-\Delta}$ (which is holomorphic in a neighborhood of

8441: $\partial^-\Delta$ in $\C$) match together in a function which is

8442: $L^1$ on $\partial \Delta$. Let us denote this function by $f_s$.

8443: Furthermore, $f_s$ extends holomorphically to $\Delta$ as $f'\circ A_s

8444: \vert_\Delta$. Consequently, $f'\circ A_s\vert_\Delta$ belongs to the

8445: Hardy space $H^1( \Delta)$, which proves the lemma.

8446: \endproof

8447: 

8448: Since we have now established that the boundary value of

8449: $f'$ on $M\backslash \Phi$ along the family

8450: of discs $A_s(\zeta)$ coincides with $f$, we can

8451: now denote both functions by the same letter $f$. 

8452: 

8453: For $\varepsilon\geq 0$ small, let now $\chi_\varepsilon \left(s,e^{i

8454: \theta} \right)$ be a $\mathcal{ C}^2$-smooth function on $[ -2\delta,

8455: 2\delta]\times \partial \Delta$ which equals $\varepsilon$ for $\vert

8456: s \vert \leq \delta$ and for $\theta \in [-\pi/4,\pi/4]$ and which

8457: equals $0$ if either $\pi/2 \leq \vert \theta \vert \leq \pi$ or

8458: $\vert s \vert \geq 2\delta/3$. We may require in addition that $\vert

8459: \vert \chi_\varepsilon \vert \vert_{\mathcal{ C}^2}\leq

8460: \varepsilon$. We define a deformation $M^\varepsilon$ of $M$ compactly

8461: supported in a neighborhood of $q$ by pushing $M$ inside $\mathcal{

8462: W}_q$ along the family of discs $A_s(\zeta)$ as follows:

8463: \def\theequation{11.9}\begin{equation}

8464: M^\varepsilon:= \{A_s\left([1-\chi_\varepsilon (s,e^{i\theta})]\, 

8465: e^{i\theta}\right): \, 

8466: \vert \theta \vert \leq \pi/2, \ 

8467: \vert s \vert \leq 2\delta\}.

8468: \end{equation}

8469: Notice that $M^\varepsilon$ coincides with $M$ outside a small

8470: neighborhood of $q$. Then we have $\vert \vert M^\varepsilon - M \vert

8471: \vert_{\mathcal{ C}^2}\leq C\cdot \varepsilon$, for some constant

8472: $C>0$ which depends only on the $\mathcal{ C}^2$ norms of $A_s(\zeta)$

8473: and of $\chi_\varepsilon(s,e^{i\theta})$. If the radius $\rho_2$ is as

8474: in Property {\bf (7)} of Lemma~11.7 above, the deformation

8475: $M^\varepsilon \cap \Delta_n(\rho_2)$ is entirely contained in

8476: $\mathcal{ W}_q$ and since $f$ is holomorphic in $\mathcal{ W}_q$,

8477: its restriction to $M^\varepsilon \cap \Delta_n (\rho_2)$ is obviously

8478: CR. An illustration is provided in the right hand side of the

8479: following figure.

8480: 

8481: \bigskip

8482: \begin{center}

8483: \input M-deformation-discs.pstex_t

8484: \end{center}

8485: 

8486: As in~\cite{j5}, \cite{p1}, \cite{mp1}, we notice that for every $s$

8487: and every $\varepsilon$, the one-dimensional Lebesgue measure on the

8488: arc 

8489: \def\theequation{11.10}\begin{equation}

8490: \Gamma_{\varepsilon,s}:=\{[1-\chi_\varepsilon

8491: (s,e^{i\theta})]e^{i\theta}\in \Delta: \vert

8492: \theta \vert \leq \pi\}

8493: \end{equation}

8494: is a Carleson measure. Thanks to the geometric uniformity of these

8495: arcs $\Gamma_{\varepsilon,s}$, it follows from an inspection of the

8496: proof of L.~Carleson's imbedding theorem that there exists a (uniform)

8497: constant $C$ such that for all $s$ with $\vert s \vert \leq 2\delta$

8498: and all $\varepsilon$, one has the estimate

8499: \def\theequation{11.11}\begin{equation}

8500: \int_{\Gamma_{\varepsilon,s}}\, 

8501: \left\vert f\left(

8502: A_s\left(

8503: \left[

8504: 1-\chi_\varepsilon(s,e^{i\theta})

8505: \right]\, e^{i\theta}

8506: \right)

8507: \right)\right\vert \cdot d\theta

8508: \,

8509: \leq \, C 

8510: \int_{\partial\Delta}\, 

8511: \vert f \vert \cdot d\theta.

8512: \end{equation} 

8513: 

8514: We are now ready to complete the proof of Lemma~11.6. Let

8515: $\pi_{x,y,u}$ denote the projection parallel to the $v$-space onto the

8516: $(x, y,u)$-space. The mapping $(s, \theta) \mapsto \pi_{x, y,u }\left(

8517: A_s ( \theta) \right)$ may be used to define new coordinates in a

8518: neighborhood of the origin in $\C^m\times \R^d$, an open subset above

8519: which $M$ and $M^\varepsilon$ are graphed. We shall now work with

8520: these coordinates. With respect to the coordinates $(s,\theta)$, on

8521: $M$ and on $M^\varepsilon$, we have formal adjoints ${}^T \overline{

8522: L}$ and ${}^T( \overline{L }^\varepsilon)$ of the basis of CR vector

8523: fields with an estimation of the form $\left\vert \left\vert {}^T(

8524: \overline{ L }^\varepsilon)- {}^T \overline{ L} \right \vert \right

8525: \vert_{ \mathcal{ C}^1}\leq C \cdot \varepsilon$, for some constant

8526: $C>0$. Let now $\psi= \psi (s,\theta)$ be $\mathcal{ C}^1$-smooth

8527: function with compact support in the set $\{\vert s \vert < \delta, \,

8528: \vert \theta \vert \leq \pi/4\}$. By construction, the subpart of

8529: $M^\varepsilon$ defined by $\widetilde{ M }^\varepsilon:= \{A_s

8530: \left([1-\chi_\varepsilon (s,e^{ i \theta })]\, e^{ i\theta} \right):

8531: \, \vert \theta \vert \leq \pi/4, \ \vert s \vert \leq \delta\}$ is

8532: contained in the wedge $\mathcal{ W}_q$, hence the restriction of the

8533: holomorphic function $f\in \mathcal{ W}_q$ to $\widetilde{

8534: M}^\varepsilon$ is obviously CR on $\widetilde{ M}^\varepsilon$.

8535: 

8536: For simplicity of notation, we shall denote $f \left(A_s(e^{ i\theta})

8537: \right)$ by $f_s(\theta)$ and $f\left( A_s \left( \left[ 1-

8538: \chi_\varepsilon(s, e^{i\theta }) \right]\, e^{ i\theta} \right)

8539: \right)$ by $f_s^\varepsilon ( \theta)$. Since by construction

8540: for every $\varepsilon >0$, the $L^1$ function $(s,\theta)\mapsto

8541: f_s^\varepsilon (\theta)$ is annihilated in the distributional sense

8542: by the CR vector fields $\overline{ L}^\varepsilon$ on $\widetilde{

8543: M}^\varepsilon$, we may compute (not writing the arguments $(s,

8544: \theta)$ of $\psi$)

8545: \def\theequation{11.12}\begin{equation}

8546: {\small

8547: \left\{

8548: \aligned

8549: {}

8550: &

8551: \left\vert

8552: \int_{\vert s \vert \leq \delta}\,\int_{\vert

8553: \theta \vert \leq \pi/4}\, 

8554: f_s(\theta)\cdot 

8555: {}^T\overline{L}(\psi) \cdot dsd\theta \right\vert=\\

8556: & \

8557: =\left\vert

8558: \int_{\vert s \vert \leq \delta}\,\int_{\vert

8559: \theta \vert \leq \pi/4}\, 

8560: \left[

8561: f_s(\theta)

8562: \cdot {}^T\overline{L}(\psi)-

8563: f_s^\varepsilon(\theta)

8564: \cdot {}^T(\overline{L}^\varepsilon)(\psi)

8565: \right] \cdot dsd\theta \right\vert \\

8566: & \

8567: \leq 

8568: \left\vert

8569: \int_{\vert s \vert \leq \delta} \left(

8570: \int_{\vert \theta \vert \leq \pi/4}\,

8571: \left[

8572: f_s(\theta)\cdot {}^T\overline{L}(\psi)-f_s(\theta)\cdot 

8573: {}^T(\overline{L}^\varepsilon)(\psi)+ \right.\right.\right.\\

8574: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

8575: \ \ \ \ \ \ \

8576: \left.\left.\left.

8577: +f_s(\theta)\cdot {}^T(

8578: \overline{L}^\varepsilon)(\psi)-

8579: f_s^\varepsilon(\theta)

8580: \cdot {}^T(\overline{L}^\varepsilon)(\psi)

8581: \right] \cdot d\theta \right)\cdot ds \right\vert

8582: \\

8583: & \

8584: \leq 

8585: C_1(\psi) \cdot \varepsilon \cdot 

8586: \int_{\vert s \vert \leq \delta}\,\int_{\vert

8587: \theta \vert \leq \pi/4}\, 

8588: \vert f_s(\theta) \vert \cdot dsd\theta + \\

8589: & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ 

8590: +C_2(\psi) \cdot

8591: \int_{\vert s \vert \leq \delta}\,\int_{\vert

8592: \theta \vert \leq \pi/4}\,

8593: \vert f_s(\theta) - f_s^\varepsilon(\theta) \vert \cdot dsd\theta

8594: \\

8595: & \

8596: \leq C_1(\psi,f,\delta)\cdot \varepsilon+

8597: C_2(\psi,\delta)\cdot \max_{\vert s \vert \leq \delta}\, \int_{\vert

8598: \theta \vert \leq \pi/4} \, 

8599: \vert f_s(\theta) - f_s^\varepsilon(\theta) \vert \cdot dsd\theta.

8600: \endaligned\right. 

8601: }

8602: \end{equation}

8603: However, thanks to the estimate~\thetag{11.11} and thanks to

8604: Lebesgue's dominated convergence theorem, the last integral tends to

8605: zero as $\varepsilon$ tends to zero. It follows that the integral in

8606: the first line of~\thetag{11.12} can be made arbitrarily small, hence

8607: it vanishes. This proves that $f$ is CR in a neighborhood

8608: of $q$ and completes the proof of Lemma~11.6.

8609: \endproof

8610: 

8611: The proof of Lemma~11.3 is complete.

8612: \endproof

8613: 

8614: \section*{\S12.~Proofs of Theorem~1.1 and of Theorem~1.3}

8615: 

8616: \subsection*{12.1.~Tree of separatrices linking hyperbolic points}

8617: Let $M\subset \C^2$ be a globally minimal $\mathcal{

8618: C}^{2,\alpha}$-smooth hypersurface, let $S\subset M$ be a $\mathcal{

8619: C}^{2,\alpha}$-smooth open surface (without boundary) and let

8620: $K\subset S$ be a proper compact subset of $S$. Assume that $S$ is

8621: totally real outside a discrete subset of complex tangencies which are

8622: hyperbolic in the sense of E.~Bishop. Since we aim to remove the

8623: compact subset $K$ of $S$, we can shrink the open surface

8624: $S$ around $K$ in order that $S$ contains {\it 

8625: only finitely many}\, such hyperbolic

8626: complex tangencies, which we shall

8627: denote by $\{h_1,\dots,h_\lambda\}$, where

8628: $\lambda$ is some integer, possibly zero. Furthermore, we can

8629: assume that $\partial S$ is of class $\mathcal{ C}^{2,\alpha}$. As a

8630: corollary of the qualitative theory of planar vector fields, due to

8631: H.~Poincar\'e and I.~Bendixson, we know that

8632: \begin{itemize}

8633: \item[{\bf (i)}]

8634: The hyperbolic points $h_1, \dots, h_\lambda$ are singularities of the

8635: characteristic foliation $\mathcal{ F}_S^c$.

8636: \item[{\bf (ii)}]

8637: Incoming to every hyperbolic point $h_1,\dots,h_\lambda$, there

8638: are exactly four $\mathcal{ C}^{2,\alpha}$-smooth open {\sl

8639: separatrices} (to be defined precisely below).

8640: \item[{\bf (iii)}]

8641: After perturbing slightly the boundary $\partial S$ if necessary,

8642: these separatrices are all transversal to $\partial S$ and the union of

8643: all separatrices together with all hyperbolic points makes a {\sl

8644: finite tree without cycles in $S$} (to be defined below).

8645: \end{itemize} 

8646: 

8647: Precisely, by an (open) {\sl separatrix}, we mean a $\mathcal{

8648: C}^{2,\alpha}$-smooth curve $\tau: (0,1) \to S$ with $\frac{ d

8649: \tau}{ds}(s) \in T_{\tau(s)}S \cap T_{\tau(s)}^c M\backslash \{0\}$

8650: for every $s\in (0,1)$, namely its tangent vectors are all nonzero and

8651: characteristic, such that one limit point, say $\lim_{s\to 0} \,

8652: \tau(s)$ is a hyperbolic point, and the other $\lim_{s\to 1} \,

8653: \tau(s)$ either belong to the boundary $\partial S$ or is a second

8654: hyperbolic point.

8655: 

8656: From the local study of saddle phase diagrams ({\it cf.}~\cite{

8657: ha}), we get in addition:

8658: 

8659: \begin{itemize}

8660: \item[{\bf (iv)}] 

8661: There exists $\varepsilon >0$ and for every $l = 1, \dots, \lambda$,

8662: there exist two curves $\gamma_l^1, \gamma_l^2 : (- \varepsilon,

8663: \varepsilon ) \to S$ which are of class $\mathcal{ C }^{ 1, \alpha}$,

8664: {\it not more}, with $\gamma_l^i (0) = h_l$ and $\frac{ d \gamma_l^1

8665: }{ dt} (s) \in T_{ \gamma_l (s)} S \cap T_{ \gamma_l(s) }^c M

8666: \backslash \{0\}$ for every $s \in (- \varepsilon, \varepsilon)$ and

8667: for $i= 1,2$, such that the four open segments $\gamma_l^1 (

8668: -\varepsilon,0)$, $\gamma_l^1 ( 0, \varepsilon )$, $\gamma_l^2 (

8669: -\varepsilon, 0)$ and $\gamma_l^2 ( 0, \varepsilon )$ cover the four

8670: pieces of open separatrices incoming at $h_l$.

8671: \end{itemize}

8672: 

8673: Let $\tau_1,\dots,\tau_\mu: (0, 1) \to S$ denote all the separatrices of

8674: $S$, where $\mu$ is some integer, possibly equal to zero. By the

8675: {\sl finite hyperbolic tree $T_S$ of $S$}, we mean:

8676: \def\theequation{12.2}\begin{equation}

8677: T_S := \{h_1,\dots,h_\lambda\} 

8678: \bigcup_{1\leq k\leq \mu} \ \tau_k(0,1).

8679: \end{equation} 

8680: We say that $T_S$ has {\sl no cycle} if it does not contain any subset

8681: homeomorphic to the unit circle. For instance, in the case where

8682: $S\equiv D$ is diffeomorphic to a real disc (as in the assumptions of

8683: Theorem~1.1), its hyperbolic tree $T_D$ necessarily has no

8684: cycle. However, in the case where $S$ is an annulus (for instance),

8685: there is a trivial example of a characteristic foliation with two

8686: hyperbolic points and a circle in the hyperbolic tree.

8687: 

8688: \subsection*{12.3.~Hyperbolic decomposition in the disc case}

8689: Let the real disc $D$ and the compact subset $K \subset D$ be as in

8690: Theorem~1.1. As in \S12.1 just above, we shrink $D$ slightly and

8691: smooth out its boundary, so that its hyperbolic tree

8692: $T_D$ is finite and has

8693: no cycle. We may decompose $D$ as the disjoint union

8694: \def\theequation{12.4}\begin{equation}

8695: D= T_D \cup D_o,

8696: \end{equation}

8697: where the complement of the hyperbolic tree $D_o: = D \backslash T_D$

8698: is an open subset of $D$ entirely contained in the totally real part

8699: of $D$. Then $D_o$ has finitely many connected

8700: components $D_1,\dots,D_\nu$, the {\sl hyperbolic sectors of $D$}.

8701: Then, for $j=1,\dots,\nu$, we define the proper closed subsets $C_j:=

8702: D_j \cap K$ of $D_j$ as illustrated in the left hand side of the

8703: following figure.

8704: 

8705: \bigskip

8706: \begin{center}

8707: \input hyperbolic.pstex_t

8708: \end{center}

8709: 

8710: Again from H.~Poincar\'e and I.~Bendixson's 

8711: theory, we know that for every

8712: component $D_j$ (in which the characteristic foliation is

8713: nonsingular), the proper closed subset $C_j$ satisfies the nontransversality

8714: condition $\mathcal{ F}_{ D_j}^c\{ C_j\}$ formulated in Theorem~1.2.

8715: In {\sc Figure~20} just above, we have drawn the characteristic curves

8716: only for the two sectors $D_4$ and $D_6$. One may observe that

8717: $\mathcal{ F}_{D_4}\{ C_4\}$ and $\mathcal{ F}_{D_6}\{ C_6\}$ hold

8718: true. Also, $K\cap T_D$ is a proper closed subset of the hyperbolic

8719: tree of $D$.

8720: 

8721: \subsection*{12.5.~Global minimality of some complements}

8722: Before proceeding to the deduction of Theorems~1.1 and~1.3 from

8723: Theorem~1.2, we must verify that the complement $M\backslash K$ is

8724: also globally minimal. Here, we state a generalization of Lemma~3.5 to

8725: the case where some hyperbolic complex tangencies are allowed. Its

8726: proof is not immediate.

8727: 

8728: \def\thelemma{12.6}\begin{lemma}

8729: Let $M$ be a $\mathcal{ C}^{2,\alpha}$-smooth hypersurface in $\C^2$

8730: and let $S\subset M$ be $\mathcal{ C}^{2,\alpha}$-smooth surface which

8731: is totally real outside a discrete subset of hyperbolic complex

8732: tangencies. Assume that the hyperbolic tree $T_S$ of $S$ has no cycle.

8733: Then for every compact subset $K\subset S$ and for

8734: an arbitrary point $p\in

8735: M\backslash K$, its CR orbit in $M\backslash K$ coincides with its CR

8736: orbit in $M$, minus $K$, namely

8737: \def\theequation{12.7}\begin{equation}

8738: \mathcal{ O}_{CR}(M\backslash K, p)= 

8739: \mathcal{ O}_{CR}(M, p) \backslash K.

8740: \end{equation}

8741: \end{lemma}

8742: 

8743: \proof

8744: Of course, we may assume that $S$ coincides with the shrinking of a

8745: slightly larger surface and has finitely many hyperbolic points

8746: $\{h_1,\dots,h_\lambda\}$, as described in \S12.1, with the same

8747: notation. Let $K_{T_S}:= K \cap T_S$ be the track of $K$ on the

8748: hyperbolic tree $T_S$. Since the intersection of $K_{T_S}$ with any

8749: open separatrix may in general coincide

8750: with any arbitrary closed subset of an

8751: interval, in order to fix ideas, it will be convenient to deal with an

8752: enlargement $\overline{K}$ of $K_{T_S}$, simply defined by

8753: filling the possible holes of $K_{T_S}$ in $T_S$: more

8754: precisely, $\overline{K}$

8755: should contain all hyperbolic points together with all separatrices

8756: joining them and for every separatrix $\tau_k(0,1)$ with right limit

8757: point $\lim_{s\to 1} \, \tau_k(s)$ belonging to the boundary of $S$,

8758: we require that $\overline{K}$ contains the segment $\tau_k[0, r_1]$,

8759: where $r_1<1$ is close enough to $1$ in order that $\overline{ K}$

8760: effectively contains $K_{T_S}$. 

8761: 

8762: Obviously, from the inclusions

8763: \def\theequation{12.8}\begin{equation}

8764: K_{T_S} \subset \overline{ K} \subset K,

8765: \end{equation}

8766: we deduce that for every point $p\in M\backslash K$, we have

8767: the reverse inclusions

8768: \def\theequation{12.9}\begin{equation}

8769: \mathcal{ O}_{CR}(M\backslash K,p) \subset

8770: \mathcal{ O}_{CR}(M\backslash \overline{ K},p) \subset

8771: \mathcal{ O}_{CR}(M\backslash K_{T_S},p).

8772: \end{equation}

8773: The main step in the proof of Lemma~12.6 will be to establish the

8774: following two assertions, implying the third, desired

8775: assertion, already stated as~\thetag{12.7}.

8776: 

8777: \begin{itemize}

8778: \item[{\bf (A1)}]

8779: For every point $q\in M\backslash \overline{ K}$, we have

8780: $\mathcal{ O}_{CR}(M\backslash \overline{ K}, q)= 

8781: \mathcal{ O}_{CR}(M, q) \backslash \overline{K}$.

8782: \item[{\bf (A2)}]

8783: For every point $r\in M\backslash K_{T_S}$, we have $\mathcal{

8784: O}_{CR}(M\backslash K_{T_S}, r)= \mathcal{ O}_{CR}(M, r) \backslash

8785: K_{T_S}$.

8786: \end{itemize}

8787: 

8788: Indeed, taking these two assertions for granted, let us conclude the

8789: proof of Lemma~12.6. Let $p\in M\backslash K$ and decompose $K$ as a

8790: disjoint union $K= K_{T_S}\cup C'$, where $C':= K\backslash K_{T_S}$

8791: is a relatively closed subset of the hypersurface $M':= M\backslash

8792: K_{T_S}$. Notice that $C'$ is contained in the totally real 

8793: part of $S$.

8794: Again thanks to foliation theory, we see that

8795: the assumption that $K_{T_S}$ does not contain any cycle

8796: entails that $C'$ does not contain maximal characteristic lines of the

8797: totally real surface $S\backslash K_{T_S}$. Consequently, all the

8798: assumptions of Lemma~3.5 are satisfied, hence by applying it to $p$, we

8799: deduce that $\mathcal{ O}_{CR}(M' \backslash C', p) = \mathcal{

8800: O}_{CR}(M', p) \backslash C'$. By developing in length this identity

8801: between sets, we get

8802: \def\theequation{12.10}\begin{equation}

8803: \aligned

8804: \mathcal{ O}_{CR}(M\backslash K, p)

8805: & \ 

8806: = 

8807: \mathcal{ O}_{CR}\left((M\backslash K_{T_S})

8808: \backslash C', p\right) \\

8809: & \

8810: =

8811: \left[\mathcal{ O}_{CR}(M\backslash K_{T_S},p)

8812: \right]\backslash C' \\

8813: & \ 

8814: = 

8815: \left[\mathcal{ O}_{CR}(M, p)\backslash K_{T_S}

8816: \right]\backslash C' \\

8817: & \ 

8818: =

8819: \mathcal{ O}_{CR}(M,p) \backslash K,

8820: \endaligned

8821: \end{equation}

8822: where, for the passage from the second to the third line, we

8823: use {\bf (A2)}. This is~\thetag{ 12.7}, as desired. 

8824: 

8825: Thus, the (main) remaining task is to establish the 

8826: assertion {\bf (A2)}, with {\bf (A1)} being a preliminary step.

8827: 

8828: First of all, we show how to deduce {\bf (A2)} from {\bf (A1)}. Pick

8829: an arbitrary point $r \in M\backslash K_{ T_S}$. Of course, we have

8830: the trivial inclusion $\mathcal{ O}_{ CR}(M \backslash K_{ T_S},r)

8831: \subset \mathcal{ O}_{CR} (M, r) \backslash K_{ T_S}$ and we want an

8832: equality. As $\overline{ K}$ contains $K_{ T_S}$, we have either $r

8833: \in M \backslash \overline{ K}$ (first case) or $r \in \overline{ K}

8834: \backslash K_{T_S}$ (second case), {\it see}\, {\sc Figure~21} just

8835: below, where $r$ is located in $\overline{ K}\backslash K_{T_S}$.

8836: 

8837: \bigskip

8838: \begin{center}

8839: \input kts.pstex_t

8840: \end{center}

8841: 

8842: Since the second case makes it impossible to apply {\bf (A1)}, we need

8843: to find another point $r'$ in the CR orbit of $r$ in $M\backslash

8844: K_{T_S}$ such that $r'$ belongs to $M\backslash \overline{ K}$. This

8845: is elementary. We make a dichotomy: either $r=h_l$ is a hyperbolic

8846: point or it belongs to an open separatrix $\tau_k(0,1)$. If $r=h_l\in

8847: M\backslash K_{T_S}$ is a hyperbolic point, we may use one of the two

8848: $\mathcal{ C}^{2,\alpha}$-smooth complex tangent curves $\gamma_l^1$

8849: or $\gamma_l^2$ passing through $h_l$ and running in $M\backslash

8850: K_{T_S}$ to join $r$ with another point which belongs to an open

8851: separatrix and which obviously lies in the same CR orbit $\mathcal{

8852: O}_{CR }( M \backslash K_{T_S}, r)$. Hence, we may assume that $r\in

8853: \overline{ K} \backslash K_{T_S}$ belongs to an open separatrix. Since

8854: $S$ is now maximally real near $r$, we may choose a $T^cM$-tangent

8855: vector field $Y$ defined in a neighborhood of $r$ which is transversal

8856: to $S$ at $r$. Then for all $\delta>0$ small enough, the point $r':

8857: =\exp (\delta Y)(r)$ is outside $S$, hence does not belong to

8858: $\overline{ K}$ and clearly lies in the same CR orbit $\mathcal{ O

8859: }_{CR }( M \backslash K_{ T_S}, r)$.

8860: 

8861: In summary, when $r \in \overline{ K} \backslash K_{ T_S}$, we have

8862: exhibited a point $r'\in \mathcal{ O }_{CR}(M \backslash K_{T_S}, r)$

8863: with $r '\in M \backslash \overline{ K}$ so that it suffices now to

8864: show that for every point $r\in M \backslash \overline{ K}$, we have

8865: $\mathcal{ O}_{ CR}(M \backslash K_{T_S}, r)= \mathcal{ O}_{ CR} (M,r)

8866: \backslash K_{ T_S}$.

8867: 

8868: Using a trivial inclusion and

8869: applying {\bf (A1)}, we deduce that

8870: \def\theequation{12.11}\begin{equation}

8871: \mathcal{ O}_{CR}(M\backslash K_{T_S}, r) \supset

8872: \mathcal{ O}_{CR}(M\backslash \overline{ K}, r) = 

8873: \mathcal{ O}_{CR}(M, r) \backslash \overline{ K}.

8874: \end{equation}

8875: Unfortunately, there may well exist points of $\mathcal{ O}_{CR}(M,

8876: r)$ belonging to $\overline{ K} \backslash K_{T_S}$, so that it

8877: remains to show that {\it every}\, point $r'\in \mathcal{ O}_{CR}(M,

8878: r)\cap [ \overline{ K} \backslash K_{T_S}]$ also belong to $\mathcal{

8879: O}_{ CR}(M \backslash K_{T_S}, r)$. Again, this last step is

8880: elementary and totally analogous to the above argument: we first claim

8881: that we can join such a point $r'$ to a point $r'''\in M\backslash

8882: \overline{ K}$ by means of a piecewise smooth CR curve running in

8883: $M\backslash K_{T_S}$. Indeed, if $r'=h_l$ is a hyperbolic point, we

8884: may first use one of the two $\mathcal{ C}^{2, \alpha}$-smooth complex

8885: tangent curves $\gamma_l^1$ or $\gamma_l^2$ passing through $h_l$ and

8886: running in $M \backslash K_{ T_S}$ to join $r'$ with another nearby

8887: point $r''$ which belongs to an open separatrix. If $r'$ already

8888: belongs to an open separatrix, we simply set $r'':= r'$. Since $S$ is

8889: now maximally real near $r''$, we may choose a $T^cM$-tangent vector

8890: field $Y$ defined in a neighborhood of $r''$ which is transversal to

8891: $S$ at $r''$. Then for all $\delta>0$ small enough, the point

8892: $r''':=\exp( \delta Y )(r'')$ satisfies $r''' \not \in S$, whence

8893: $r''' \not \in \overline{ K}$. Of course, by choosing $r''$ and $r'''$

8894: sufficiently close to $r'$, it follows that the piecewise smooth CR

8895: curve joining them does not meet $K_{T_S}$. We deduce that $r'\in

8896: \mathcal{O}_{CR} (M\backslash K_{ T_S}, r''')$.

8897: 

8898: Since by assumption $r' \in \mathcal{ O}_{CR}(M, r)$, it follows that

8899: $r'''\in \mathcal{ O}_{CR}(M,r)$ and then $r'''\in \mathcal{ O}_{

8900: CR}(M,r)\backslash \overline{ K}$. By means of the

8901: supclusion~\thetag{ 12.11}, we deduce that $r'''\in \mathcal{

8902: O}_{CR}(M\backslash K_{T_S},r)$.

8903: 

8904: Finally, from the two relations $r'''\in \mathcal{ O}_{CR}(M\backslash

8905: K_{T_S},r)$ and $r'\in \mathcal{O}_{CR} (M\backslash K_{ T_S}, r''')$,

8906: we conclude immediately that $r'\in \mathcal{ O}_{CR}(M\backslash

8907: K_{T_S}, r)$. We have thus shown that the supclusion in~\thetag{

8908: 12.11} is an equality.

8909: 

8910: This completes the deduction of {\bf (A2)}

8911: from {\bf (A1)}.

8912: 

8913: It remains now to establish {\bf (A1)}. We remind that in 

8914: Section~3, we derived Lemma~3.5 from Lemma~3.7.

8915: By means of a totally similar argument, which we shall

8916: not repeat, one deduces {\bf (A1)} from the following

8917: assertion. Remind that as $M$ is a hypersurface in 

8918: $\C^2$, its CR orbits are of dimension either $2$ or

8919: $3$.

8920: 

8921: \def\thelemma{12.12}\begin{lemma}

8922: Let $M$, $S$, $T_S$ and $\overline{ K} \subset T_S$ be as above.

8923: There exists a connected submanifold $\Omega$ embedded in $M$

8924: containing the hyperbolic tree $T_S$ such that

8925: \begin{itemize}

8926: \item[{\bf (1)}]

8927: $\Omega$ is a $T^c M$-integral manifold, namely $T_p^cM \subset

8928: T_p \Omega$ for all $p \in \Omega$.

8929: \item[{\bf (2)}]

8930: $\Omega$ is contained in a single CR orbit of $M$.

8931: \item[{\bf (3)}]

8932: $\Omega\backslash \overline{ K}$ is also 

8933: contained in a single CR orbit of $M \backslash

8934: \overline{ K}$.

8935: \end{itemize}

8936: More precisely, $\Omega$ is an open neighborhood of $T_S$ if it is of

8937: real dimension $3$ and a complex curve surrounding $T_S$ if it is of

8938: dimension $2$.

8939: \end{lemma}

8940: 

8941: \proof

8942: We shall construct $\Omega$ by means of a flowing procedure, 

8943: starting from a local piece of it. We start

8944: locally in a neighborhood of a fixed point

8945: $p_0 \in \overline{ K} \backslash \{ h_1, \dots, h_\lambda \}$,

8946: whose precise

8947: choice does not matter. Since $S$ is totally real in a neighborhood of

8948: $p_0$, there exists a locally defined $T^cM$-tangent vector field $Y$

8949: which is transversal to $S$ at $p_0$. Consequently, for $\delta >0$

8950: small enough, the small segment $I_0:= 

8951: \{ \exp(sY) (p_0): \, - \delta < s < \delta\}$ is transversal to 

8952: $S$ at $p_0$ and moreover,

8953: the

8954: two half-segments

8955: \def\theequation{12.13}\begin{equation}

8956: I_0^\pm:= \left\{

8957: \exp(s Y)(p_0): \, 

8958: 0 < \pm s < \delta 

8959: \right\}

8960: \end{equation}

8961: lie in $M\backslash S$. Since $p_0$ belongs to some $T^cM$-tangent

8962: open separatrix $\tau_k(0,1)$, there exists a $\mathcal{ C}^{1,

8963: \alpha}$-smooth vector field $X$ defined in a neighborhood of $p_0$ in

8964: $M$ which is tangent to $S$ and whose integral curve passing through

8965: $p_0$ is a piece of $\tau_k(0,1)$. Since $Y$ is transversal to $S$ at

8966: $p_0$, it follows that the set $\omega_0 := \{\exp(s_2 X)(\exp ( s_1

8967: Y) (p_0)): \, -\delta < s_1, \, s_2 < \delta\}$ is a well-defined

8968: $\mathcal{ C}^{1, \alpha}$-smooth codimension one small submanifold

8969: passing through $p_0$ which is transversal to $S$ at $p_0$. Clearly,

8970: we even have $T_{p_0}\omega_0 = T_{ p_0}^cM$. Thanks to the fact that

8971: the flow of $X$ stabilizes $S$, we see that the integral curves $s_2

8972: \mapsto \exp(s_2 X)(\exp (s_1 Y) (p_0))$ are contained in $M

8973: \backslash S$ for every starting point $\exp (s_1 Y) (p_0)$ in the

8974: segment $I_0$ which does not lie in $S$, namely for all $s_1 \neq

8975: 0$. We deduce that the two open halves of $\omega_0$ defined by

8976: \def\theequation{12.14}\begin{equation}

8977: \omega_0^\pm:=\{

8978: \exp(s_2

8979: X)(\exp (s_1 Y)(p_0)): \, 

8980: 0 < \pm s_1 < \delta, \ 

8981: -\delta < s_2 < \delta

8982: \}

8983: \end{equation}

8984: are contained in a single CR orbit of $M \backslash \overline{ K}$.

8985: 

8986: To begin with, assume that the CR orbit in $M \backslash \overline{

8987: K}$ the point $q_0^+:= \exp\left(\frac{ \delta}{2} Y \right) (p_0)$

8988: which belong to $\omega_0^+$, as drawn in {\sc Figure~22} below,

8989: is of real dimension $2$. Afterwards, we shall treat the case where

8990: its CR orbit in $M\backslash \overline{ K}$ is of dimension $3$.

8991: 

8992: Since $M$ is a hypersurface in $\C^2$ and since we have just proved

8993: that the CR orbit $\mathcal{ O}_{ CR}(M \backslash \overline{ K},

8994: q_0^+)$ already contains the $2$-dimensional half piece $\omega_0^+$,

8995: we deduce that $\omega_0^+$ is a piece of complex curve whose boundary

8996: $\partial \omega_0^+$ is (by construction) contained in the separatrix

8997: $\tau_k (0,1)$. Since $\tau_k (0,1)$ is an embedded segment, we may

8998: suppose from the beginning that the vector field $X$ is defined in a

8999: neighborhood of $\tau_k (0,1)$ in $M$. Using then the flow of $X$, we

9000: may easily prolong the small piece $\omega_0^+$ to get a semi-local

9001: $\mathcal{ C }^{1, \alpha}$-smooth submanifold $\omega_k^+$ stretched

9002: along $\tau_k(0,1)$, which constitutes its boundary. Again, this

9003: piece $\omega_k^+$ is (by construction) contained in the CR orbit of

9004: $q_0^+$ in $M\backslash \overline{K}$. By the fundamental stability

9005: property of CR orbits under flows, we deduce that $\omega_k^+$ is in

9006: fact a piece of complex curve with boundary $\tau_k(0,1)$.

9007: 

9008: Remind that by definition of separatrices, the point $\tau_k(0)$ is

9009: always a hyperbolic point. There is a dichotomy: either $\tau_k(1)$ is

9010: also a hyperbolic point or it lies in $\partial S$. If $\tau_k(1)$ is

9011: a hyperbolic point, then by the definition of $\overline{ K}$, the

9012: complete boundary $\partial \omega_k^+ =\tau_k(0,1)$ is contained in

9013: $\overline{ K}$, hence it may {\it not}\, be crossed by means of a CR

9014: curve running in $M \backslash \overline{ K}$. Since the piece

9015: $\omega_k^+$ will be flowed all around $T_S$, our filling $\overline{

9016: K}$ of $K_{T_S}$ was motivated by the desire of simplifying the

9017: geometric situation without having to discuss whether $K_{T_S}$

9018: contains or does not contain the whole segment $\tau_k (0,1)$, for

9019: each $k=1, \dots, \mu$.

9020: 

9021: Before studying the case where $\tau_k (1) \in \partial S$, 

9022: let us analyze the local situation in a neighborhood of the

9023: hyperbolic point $\tau_k(0)=:h_l$, for some $l$ with $1\leq l\leq

9024: \lambda$. 

9025: 

9026: As a preliminary, in order to understand clearly the situation, 

9027: let us assume that the two characteristic curves $\gamma_l^1$ and

9028: $\gamma_l^2$ passing through $h_l$ are of class $\mathcal{

9029: C}^{ 2,\alpha}$, an assumption which would be satisfied if

9030: we had assumed that $M$ and $S$ are of class $\mathcal{ C}^{3,

9031: \alpha}$. By a straightening of $\gamma_l^1$ and of $\gamma_l^2$, 

9032: which induces a loss of one derivative,

9033: we easily show that there exist two linearly independent $\mathcal{

9034: C}^{ 1,\alpha}$-smooth $T^cM$-tangent vector fields $X_1$ and $X_2$

9035: whose integral curves issued from $h_l$ coincide with $\gamma_l^1$ and

9036: $\gamma_l^2$. After possibly renumbering and reversing $X_1$ and $X_2$

9037: and also reparametrizing $\gamma_l^1$ and

9038: $\gamma_l^2$, we may assume that

9039: $\gamma_l^1(s) =\exp(sX_1) (h_l)$ and that 

9040: $\gamma_l^2(s) =\exp(sX_2) (h_l)$, for all small

9041: $s>0$. Furthermore, we may assume that the direction of $X_2$ in a

9042: neighborhood of $h_l$ is the same as the direction from 

9043: $\partial \omega_k^+$ to $\omega_k^+$.

9044: 

9045: \bigskip

9046: \begin{center}

9047: \input tree-orbit.pstex_t

9048: \end{center}

9049: 

9050: As in {\sc Figure~22} just above, let $\tau_j(0,1)$ be the separatrix

9051: issued from $h_l$ in the positive direction of $X_2$. 

9052: We may assume that $\tau_j (0) = h_l$.

9053: Thanks to the

9054: flow of the vector field $X_2$, we may now propagate the piece of

9055: complex curve $\omega_k^+$ by stretching it along $\tau_j(0,1)$ in a

9056: neighborhood of $h_l$. Using then the flow of a semi-locally defined

9057: complex tangent vector field defined in a neighborhood of

9058: $\tau_j(0,1)$, we may extend this local piece as a complex curve

9059: $\omega_j^+$ with boundary $\tau_j(0,1)$. Finally, $\omega_k^+$ and

9060: $\omega_j^+$ glue together as a complex curve with boundary

9061: $\tau_k(0,1) \cup \{h_l\} \cup \tau_j(0,1)$ and corner $h_l$.

9062: 

9063: However, by {\bf (iv)} above, $\gamma_l^1$ and $\gamma_l^2$ are only

9064: of class $\mathcal{ C}^{1, \alpha}$. Examples for which this

9065: regularity is optimal are easily found. Straightening them is again

9066: possible, but the vector fields $X_1$ and $X_2$ would be of class

9067: $\mathcal{ C}^\alpha$, and we would lose the uniqueness of their

9068: integral curves as well as the regularity of their flow. Consequently,

9069: to prove that $\omega_k^+$ propagates along the second separatrix, with

9070: its boundary contained in it, we must proceed differently~:

9071: the proof is longer and we need one more diagram.

9072: 

9073: In {\sc Figure~23} just below, we draw the sadlle-looking surface $S$

9074: in the $3$-dimensional space $M$~; the horizontal plane passing

9075: through $h_l$ is thought to be the complex tangent plane $T_{

9076: h_l}^cM$.

9077: 

9078: \bigskip

9079: \begin{center}

9080: \input saddle-orbit.pstex_t

9081: \end{center}

9082: 

9083: Let us introduce two $T^cM$-tangent vector fields $X_1$ and $X_2$

9084: defined in a neighborhood of $h_l$ with $X_1 (h_l)$ directed along

9085: $\tau_k$ in the sense of increasing $s$ and $X_2 (h_l)$ directed along

9086: $\tau_j$ in the sense of increasing $s$. Let $Z$ denote the vector

9087: field $X_1 +X_2$, as shown in the top of the left hand side of {\sc

9088: Figure~23} above. Using the flow of $Z$ we can begin by extending the

9089: banana-looking piece $\omega_k^+$ of complex curve by introducing the

9090: submanifold $\omega$ consisting of points

9091: \def\theequation{12.15}\begin{equation}

9092: \exp(s_2 Z ) (\tau_k(s_1)),

9093: \end{equation}

9094: where $0 < s_1 < \delta$ and $0 < s_2 < \delta$, for some small

9095: $\delta > 0$. One checks that all these points stay in $M \backslash

9096: S$, hence are contained in the same CR orbit as

9097: $\omega_k^+$ in $M \backslash \overline{ K}$. By the stability

9098: property of CR orbits, it follows of course that $\omega$ is a piece

9099: of complex curve contained in $M$.

9100: 

9101: For $0 < s < \delta$, let $\mu(s) := \exp (sZ) (h_l)$ denote the CR

9102: curve lying ``between'' $\tau_k$ and $\tau_j$ and which

9103: constitutes a part of the boundary of $\omega$. Let $p$ be an

9104: arbitrary point of this curve, close

9105: to $h_l$. 

9106: 

9107: \def\thelemma{12.16}\begin{lemma}

9108: The integral curve $s\mapsto

9109: \exp (-s X_1)(p)$ of $-X_1$ issued from $p$ necessarily

9110: intersects $S$ at a point $q$ close to $h_l$ and close to $\tau_j$

9111: {\rm (}{\it cf.} {\sc Figure~23}{\rm )}.

9112: \end{lemma}

9113: 

9114: \proof

9115: First of all, we need some preliminary.

9116: 

9117: Thanks to the existence of a ``$1/8$ piece'' of complex curve $\omega$

9118: with $h_l\in \overline{ \omega}$ which is contained in the

9119: hypersurface $M$, we see that $M$ is necessarily Levi-degenerate at

9120: $h_l$.

9121: 

9122: Next, we introduce local holomorphic coordinates $(z, w) = (x+ iy,

9123: u+iv)\in \C^2$ vanishing at $h_l$ in which the hypersurface $M$ is

9124: given as the graph $v= \varphi (x, y, u)$, where $\varphi$ is a

9125: $\mathcal{ C}^{ 2, \alpha}$-smooth function. Since $M$ is Levi

9126: degenerate at $h_l$, we may assume that $\varphi (x, y, u) \vert \leq

9127: C \cdot \left(\vert x \vert + \vert y \vert + \vert u \vert \right)^{

9128: 2+ \alpha}$. We may also assume that the surface $S$, as a subset of

9129: $M$, is represented by one supplementary equation of the form $u = h

9130: (x,y)$, where the $\mathcal{ C}^{ 2,\alpha}$-smooth

9131: function $h$ satisfies

9132: \def\theequation{12.17}\begin{equation}

9133: \left\{

9134: \aligned

9135: h(x,y)

9136: & \

9137: = z \bar z+ \gamma \, (z^2+ \bar z^2)+ 

9138: {\rm O}\left( \vert z \vert^{ 2+ \alpha} \right) \\

9139: & \

9140: =

9141: (2\gamma + 1) \, x^2 - 

9142: (2\gamma -1) \, y^2+ {\rm O} \left( \vert z \vert^{ 2+ \alpha}

9143: \right),

9144: \endaligned\right.

9145: \end{equation}

9146: and where $\gamma >\frac{ 1}{2}$ is E.~Bishop's invariant. Then the

9147: tangents at $h_l$ to the two half-separatrices $\tau_k$ and $\tau_j$

9148: are given respectively by the linear (in)equations $x>0$, $y = -

9149: \frac{ 2\gamma + 1}{ 2\gamma -1} \, x$, $u=0$ and $x< 0$, $y = -

9150: \frac{ 2\gamma + 1}{ 2\gamma -1} \, x$, $u=0$. In {\sc Figure~23},

9151: where we do not draw the axes, the $u$-axis is vertical, the $y$ axis

9152: points behind $h_l$ and the $x$-axis is horizontal, from left to

9153: right.

9154: 

9155: Expressing the two $T^cM$-tangent vector fields $X_1$ and $X_2$ in the

9156: (natural) real coordinates $(x,y,u)$ over $M$, we may write them as

9157: \def\theequation{12.18}\begin{equation}

9158: \left\{

9159: \aligned

9160: X_1

9161: & \

9162:  = 

9163: \frac{ \partial}{ \partial x} -

9164: \left( 

9165: \frac{ 2\gamma + 1}{ 2\gamma -1}\right) \, 

9166: \frac{ \partial }{ \partial y} + 

9167: A_1 (x,y,u)\, 

9168: \frac{ \partial }{ \partial u}, \\ 

9169: X_2 

9170: & \

9171: = 

9172: -

9173: \frac{ \partial}{ \partial x} -

9174: \left( 

9175: \frac{ 2\gamma + 1}{ 2\gamma -1}\right) \, 

9176: \frac{ \partial }{ \partial y} + 

9177: A_2 (x,y,u)\, 

9178: \frac{ \partial }{ \partial u}. 

9179: \endaligned\right.

9180: \end{equation}

9181: Since $\varphi$ vanishes to second order at $h_l$, the two

9182: $\mathcal{ C}^{ 1,\alpha}$-smooth

9183: coefficients $A_1$ and $A_2$ satisfy an estimate

9184: of the form

9185: \def\theequation{12.19}\begin{equation}

9186: \left\vert 

9187: A_1, A_2 (x,y,u) 

9188: \right\vert < C \cdot \left(

9189: \vert x \vert + \vert y \vert + 

9190: \vert u \vert

9191: \right)^{ 1+ \alpha}.

9192: \end{equation}

9193: 

9194: Now, we come back to the integral curve of Lemma~12.16. It is

9195: contained in the real $2$-surface passing through $h_l$ defined by

9196: \def\theequation{12.20}\begin{equation}

9197: \Sigma:= \{\exp(-s_2 X_1)

9198: (\exp(s_1 Z)(h_l)): \, -\delta < s_1, s_2 <

9199: \delta\},

9200: \end{equation}

9201: for some $\delta >0$. Because the vector fields $X_1$, $X_2$ and $Z=

9202: X_1 + X_2$ have $\mathcal{ C}^{ 1,\alpha}$-smooth coefficients, the

9203: surface $\Sigma$ is only $\mathcal{ C}^{ 1,\alpha}$-smooth in

9204: general. In $M$ equipped with the three real coordinates $(x,y,u)$, we

9205: may parametrize $\Sigma$ by a mapping of the form

9206: \def\theequation{12.21}\begin{equation}

9207: (s_1, s_2) \longmapsto 

9208: \left(

9209: s_2 - 

9210: 2s_1 \left(

9211: \frac{ 2\gamma +1}{ 2\gamma -1}

9212: \right), \ s_2

9213: \left(

9214: \frac{ 2\gamma+ 1}{ 2\gamma -1}

9215: \right), \ 

9216: u( s_1,s_2)

9217: \right), 

9218: \end{equation}

9219: where $u$ is of class $\mathcal{ C}^{ 1,\alpha}$. It is clear that

9220: $u(0)= \partial_{ s_1} u(0) = \partial_{ s_2} u( 0) =0$, so that

9221: there is a constant $C$ such that

9222: \def\theequation{12.22}\begin{equation}

9223: \left \vert u (s_1,\, s_2) 

9224: \right \vert < C \cdot

9225: \left(

9226: \vert s_1 \vert + \vert s_2 \vert

9227: \right)^{ 1+ \alpha},

9228: \end{equation}

9229: since $u$ is of class $\mathcal{ C}^{ 1,\alpha}$. Furthermore, by

9230: inspecting the flows appearing in~\thetag{ 12.20}, taking account of

9231: the estimates~\thetag{ 12.19}, we claim that $u$ satisfies the

9232: better estimate

9233: \def\theequation{12.23}\begin{equation}

9234: \vert u (s_1, s_2) \vert < C \cdot

9235: \left(

9236: \vert s_1 \vert + \vert s_2 \vert 

9237: \right)^{ 2+ \alpha},

9238: \end{equation}

9239: for some constant $C>0$.

9240: In other words, $\Sigma$ osculates the complex tangent plane $T_{

9241: h_l}^cM$ to second order at $h_l$~: $\Sigma$ is more flat than $S$ at

9242: $h_l$. One may check the estimate~\thetag{ 12.23} is sufficient to

9243: establish Lemma~12.16, because the second jet of the saddle function

9244: $h(x,y)$ does not vanish at the origin.

9245: 

9246: To prove the claim, we formulate the main argument as an independent

9247: assertion. Mild modifications of this argument apply to our case, but

9248: we shall not provide all the details.

9249: 

9250: Let $L_1 := \frac{ \partial }{ \partial x} + A_1 (x,y, u) \, \frac{

9251: \partial }{ \partial u}$ and $L_2 := \frac{ \partial }{ \partial y} +

9252: A_2 (x,y, u) \, \frac{ \partial}{ \partial u}$ be two vector fields

9253: having $\mathcal{ C}^{ 1,\alpha }$-smooth coefficients

9254: satisfying~\thetag{ 12.19}. Denote 

9255: by $s_1 \longmapsto (s_1, \lambda (s_1), \mu(s_1))$

9256: the integral curve of $L_1$ passing through the origin. It is

9257: $\mathcal{ C}^{ 2,\alpha}$-smooth and

9258: we have 

9259: \def\theequation{12.24}\begin{equation}

9260: \left \vert

9261: \lambda (s_1)

9262: \right \vert < C \cdot \vert s_1 \vert^{ 2+\alpha}

9263: \ \ \ \ \ \ \

9264: {\rm and} 

9265: \ \ \ \ \ \ \

9266: \left \vert

9267: \mu (s_1)

9268: \right \vert < C \cdot \vert s_1 \vert^{ 2+\alpha},

9269: \end{equation}

9270: for some constant $C>0$. Consider the composition of flows $\exp (s_2

9271: L_2) (\exp (s_1 L_1)(0))$. We have to solve the system of ordinary

9272: differential equations

9273: \def\theequation{12.25}\begin{equation}

9274: \frac{ dx}{ ds_2}= 0, 

9275: \ \ \ \ \ \ \

9276: \frac{ dy}{ ds_2} = 1, 

9277: \ \ \ \ \ \ \

9278: \frac{ du}{ ds_2}= A_2 (x,y,u)

9279: \end{equation}

9280: with initial conditions

9281: \def\theequation{12.26}\begin{equation}

9282: x(0)= s_1, 

9283: \ \ \ \ \ \ \

9284: y(0) = \lambda( s_1), 

9285: \ \ \ \ \ \ \ 

9286: u(0) = \mu(s_1).

9287: \end{equation}

9288: This yields $x(s_1,s_2) =s_1$, $y(s_1,s_2)= \lambda (s_1)+ s_2$ and

9289: the integral equation

9290: \def\theequation{12.27}\begin{equation}

9291: u(s_1,s_2) = 

9292: \mu(s_1)+ 

9293: \int_0^{ s_2} \, 

9294: A_2 \left(

9295: s_1, s_2'+ \lambda (s_1), u(s_1, s_2')

9296: \right) \, ds_2'.

9297: \end{equation}

9298: Since $u$ is at least $\mathcal{ C}^{ 1,\alpha}$-smooth and vanishes

9299: to order $1$ at $(s_1,s_2)= (0,0)$, we know already that it

9300: satisfies~\thetag{ 12.22}. Using~\thetag{ 12.19}, it is now

9301: elementary to provide an upper estimate of the right hand side

9302: of~\thetag{ 12.27} which yields the desired estimate~\thetag{ 12.23}.

9303: 

9304: The proof of Lemma~12.16 is complete.

9305: \endproof

9306: 

9307: So, for various points $p= \mu(s)$ close to $h_l$ the intersection

9308: points $q\in S$ exist. If all points $q$ belong to $\tau_j$, we are

9309: done: the piece $\omega$ extends a $1/4$ piece of complex curve with

9310: boundary $\tau_k \cup \tau_j$ near $h_l$ and corner $h_l$.

9311: 

9312: Assume therefore that one such point $q$ does not belong to $\tau_j$,

9313: as drawn in the left hand side of {\sc Figure~23} above. Suppose that

9314: $q$ lies above $\tau_j$, the case where $q$ lies under $\tau_j$ being

9315: similar and in fact simpler. The characteristic curve $\gamma'\subset

9316: S$ passing through $q$ stays above $\tau_j$ and is nonsingular.

9317: Propagating the complex curve $\omega$ in $M\backslash \overline{ K}$

9318: by means of the flow of $-X_1$, we deduce that there exists at $q$ a

9319: local piece $\omega_q^+$ of complex curve with boundary contained in

9320: $\gamma'$ which is contained in the same CR orbit as $\omega$. Using

9321: then the flow of a CR vector having $\gamma'$ as an integral curve, we

9322: can propagate $\omega_q^+$ along $\gamma'$, which yields a long thin

9323: banana-looking complex curve with boundary in $\gamma'$. However, this

9324: piece may remain too thin. Fortunately, thanks to the flow of $X_1 -

9325: X_2$, we can extend it as a piece $\omega'$ of complex curve with

9326: boundary $\gamma'$ which goes over $h_l$, with respect to a

9327: complex projection onto $T_{h_l}^cM$, as illustrated in {\sc

9328: Figure~23} above. We claim that this yields a contradiction.

9329: 

9330: Indeed, as $\omega$ and $\omega'$ are complex curves, they are locally

9331: defined as graphs of holomorphic functions $g$ and $g'$ defined in

9332: domains $D$ and $D'$ in the complex line $T_{h_l}^cM$. By

9333: construction, there exists a point in $r \in D\cap D'$ at which the

9334: values of $g$ and $g'$ are distinct. However, since by construction

9335: $g$ and $g'$ coincide in a neighborhood of the CR curve joining $p$ to

9336: $q$, they must coincide at $r$ because of the principle of analytic

9337: continuation: this is a contradiction. In conclusion, the CR orbit

9338: passes through the hyperbolic point $h_l$, in a neighborhood of which

9339: it consists of a cornered complex curve with boundary $\tau_k\cup

9340: \tau_j$.

9341: 

9342: We can now continue the proof. Since the hyperbolic tree $T_S$ does

9343: not contain any cycle, by proceeding this way we claim that the small

9344: piece of complex curve $\omega_0^+$ propagates all around $T_S$ and

9345: matches up as a smooth complex curve $\Omega$ containing the

9346: hyperbolic tree. Indeed, in the case where $\tau_k( 1)$ is not a

9347: hyperbolic point, recall that we arranged at the beginning that

9348: $\overline{ K}\cap \tau_k(0,1)= \tau_k(0,r_1]$, where $r_1<1$. It is

9349: then crucial that when a limit point $\tau_k(1)$ belongs to $\partial

9350: S$, we escape from $\overline{ K}$ and using a local CR vector field

9351: $Y$ transversal to $S$, we may cross the separatrix $\tau_k(0,1)$ at a

9352: point $\tau_k(r_2)$ where $r_2$ satisfies $r_1 < r_2 < 1$. Hence, we

9353: pass to the other side of $S$ in $M$ and then, by means of a futher

9354: flowing, we turn around to the other side of $\tau_k(0,1)$. Also, the

9355: two pieces in either side of $\tau_k(0,1)$ match up at least

9356: $\mathcal{ C}^{1,\alpha}$-smoothly. Then thanks to the stability

9357: property of orbits under flows, we deduce that these two pieces match

9358: up as a piece of complex curve containing $\tau_k(0,1)$ in its

9359: interior.

9360: 

9361: We thus construct the complex curve $\Omega$ surrounding $T_S$, which

9362: is obviously contained in a single CR orbit of $M$. Also, by

9363: construction, $\Omega\backslash \overline{ K}$ is contained in a

9364: single CR orbit of $M\backslash \overline{ K}$. Thus, we have

9365: established Lemma~3.12 under the assumption that the CR orbit of

9366: $q_0^+$ is two-dimensional.

9367: 

9368: Assume finally that the CR orbit of $q_0^+$ is $3$-dimensional. By a

9369: similar propagation procedure, we easily construct a neighborhood

9370: $\Omega$ in $M$ of the hyperbolic tree satisfying conditions {\bf

9371: (1)}, {\bf (2)} and {\bf (3)} of Lemma~12.12. This complete its proof.

9372: \endproof

9373: 

9374: The proof of Lemma~12.6 is complete.

9375: \endproof

9376: 

9377: \subsection*{12.28.~Proofs of Theorems~1.1 and 1.3}

9378: We can now prove Theorem~1.1. In fact, we shall directly prove the

9379: more general version stated as Theorem~1.3 which implies Theorem~1.1

9380: as a corollary, thanks to the geometric observations of \S12.3.

9381: 

9382: First of all, we notice that as $M\backslash K$ is globally minimal,

9383: there exists a wedge attached to $M\backslash K$ to which continuous

9384: CR functions on $M\backslash K$ extend holomorphically. Hence, the

9385: CR-removability of $K$ is a consequence of its $\mathcal{

9386: W}$-removability. Also, Lemma~11.3 shows that the 

9387: $L^{\sf p}$-removability of $K$ is a consequence of

9388: its $\mathcal{ W}$-removability. Consequently, it suffices

9389: to establish that $K$ is $\mathcal{ W}$-removable in 

9390: Theorem~1.3.

9391: 

9392: Let $T_S$ be the hyperbolic tree of (a suitable shrinking of) $S$,

9393: which contains no cycle by assumption. Let $\omega_1$ be a one-sided

9394: neighborhood of $M \backslash K$ in $\C^2$. Because the

9395: nontransversality condition $\mathcal{ F }_{ S \backslash T_S}^c \{ K

9396: \cap (S\backslash T_S) \}$ holds true by assumption, we may apply

9397: Theorem~1.2 to the totally real surface $S \backslash T_S$ in the

9398: globally minimal (thanks to Lemma~12.6) hypersurface $M \backslash K_{

9399: T_S}$ to remove the proper closed subset $K\cap (S \backslash

9400: T_S)$. We deduce that there exists a one-sided neighborhood $\omega_2$

9401: of $M\backslash K_{ T_S}$ in $\C^2$ such that (after shrinking

9402: $\omega_1$ if necessary), holomorphic functions in $\omega_1$ extend

9403: holomorphically to $\omega_2$. Then we slightly deform $M$ inside

9404: $\omega_2$ over points of $K\cap (S \backslash T_S)$. We obtain a

9405: $\mathcal{ C}^{2, \alpha}$-smooth hypersurface $M^d$ with

9406: $M^d\backslash K_{ T_S} \subset \omega_2$. Also, by stability of

9407: global minimality under small perturbations, we can assume that $M^d$

9408: is also globally minimal. By construction, we obtain holomorphic

9409: functions in the neighborhood $\omega_2$ of $M^d \backslash K_{T_S}$

9410: in $\C^2$.

9411: 

9412: Since $M$ and $M^d$ are of codimension $1$, the union of a one-sided

9413: neighborhood $\omega^d$ of $M^d$ in $\C^2$ together with $\omega_2$

9414: constitutes a one-sided neighborhood of $M$ in $\C^2$. To conclude the

9415: proof of Theorem~1.1, it suffices therefore to show that the closed

9416: set $K_{T_S}$ is $\mathcal{ W}$-removable.

9417: The reader is referred to {\sc Figure~20} above for

9418: an illustration.

9419:  

9420: Reasoning by contradiction (as for the proof of Theorem~1.2'), let

9421: $K_{\rm nr}\subset K_{T_S}$ denote the smallest nonremovable subset of

9422: $K_{T_S}$. If $K_{\rm nr}$ is empty, we are done, gratuitously.

9423: Assume therefore that $K_{\rm nr}$ is nonempty. Let $T'$ be a

9424: connected component of the minimal subtree of $T$ containing $K_{\rm

9425: nr}$. By a {\sl subtree} of a tree $T$ defined as in~\thetag{12.2}

9426: above, we mean of course a finite union of some of the separatrices

9427: $\tau_k(0,1)$ together with all hyperbolic points which are endpoints

9428: of separatrices. Since $T'$ does not contain any subset homeomorphic

9429: to the unit circle, there exists at least one extremal branch of $T'$,

9430: say $\tau_1(0,1)$ after renumbering, with $\tau_1(1)\in \partial

9431: S$. To reach a contradiction, we shall show that at least one point of

9432: the nonempty set $K_{\rm nr} \cap T'$ is in fact $\mathcal{

9433: W}$-removable.

9434: 

9435: If the subtree $T'$ consists of the single branch $\tau_1 (0,1)$

9436: together with the single elliptic point $\tau_1 (0)$, thanks to

9437: properties {\bf (iii)} and {\bf (iv)} of \S12.1, we can enlarge a

9438: little bit this branch by prolongating the curve $\tau_1 (0,1)$ to an

9439: open $\mathcal{ C}^{ 2, \alpha}$-smooth Jordan arc $\tau_1

9440: (-\varepsilon, 1+ \varepsilon)$, for some $\varepsilon>0$, with the

9441: appendix $\tau_1(1,1+\varepsilon)$ outside $S$ (in the slightly larger

9442: surface containing $S$). Then we remind that by a special case of

9443: Theorem~4 (ii) of~\cite{mp1}, every proper closed subset of

9444: $\tau_1(-\varepsilon, 1+\varepsilon)$ is $\mathcal{ W}$-removable. It

9445: follows that the proper closed subset $K_{\rm nr}$ of the Jordan arc

9446: $\tau_1(-\varepsilon, 1+\varepsilon)$ is removable, which yields the

9447: desired contradiction in the case where $T'$ consists of a single

9448: branch together with a single elliptic point.

9449: 

9450: If $T'$ consists of at least two branches, again with $\tau_1(1)\in

9451: \partial S$, then applying Theorem~4 (ii) of~\cite{mp1}, we may at

9452: least deduce that $K_{\rm nr} \cap \tau_1(0,1)$ is $\mathcal{

9453: W}$-removable, since $K_{\rm nr}\cap \tau_1(0,1)$ is contained in

9454: $\tau_1(0, r_1]$, for some $r_1<1$. But possibly, this set $K_{\rm nr}

9455: \cap \tau_1(0,1)$ could be empty.

9456: 

9457: However, we claim that it is

9458: nonempty. Indeed, otherwise, if $K_{ \rm nr} \cap \tau_1[0,1)$

9459: consists of the single point $\tau_1(0)$, which is a hyperbolic point,

9460: then $K_{ \rm nr}$ is in fact contained in the smaller subtree $T''$

9461: defined by $T'' := T' \backslash \tau_1 (0,1)$ (here, we use that

9462: $\tau_1 (0)$ is a hyperbolic point, hence there exists another branch

9463: $\tau_k (0,1)$ with $\tau_k (1)= \tau_1 (0)$ or $\tau_k (0)=

9464: \tau_1(0)$). This contradicts the assumption that $T'$ is the minimal

9465: subtree containing $K_{\rm nr}$. Then $K_{\rm nr} \cap \tau_1 (0,1)$

9466: is nonempty and removable, which contradict the assumption that

9467: $K_{\rm nr}$ is the smallest nonremovable subset of $K$.

9468: 

9469: The proofs of Theorems~1.1 and~1.3 are complete.

9470: \endproof

9471: 

9472: \section*{\S13.~Applications to the edge of the wedge theorem}

9473: 

9474: In this section we formulate three versions of the edge of the wedge

9475: theorem for holomorphic and meromorphic functions, two of which are

9476: based upon an application of our removable singularities theorems.

9477: Let us begin with some definition.

9478: 

9479: \subsection*{13.1.~Preliminary}

9480: Let $E$ be a generic submanifold of $\C^n$, which may be maximally

9481: real. By a {\sl double wedge attached to $E$}, we mean a pair

9482: $(\mathcal{ W}_1, \mathcal{ W }_2)$ of disjoint wedges attached to $E$

9483: which admit a nowhere vanishing continuous vector field $v: E

9484: \rightarrow T \C^n\vert_E / T E$ such that $ Jv ( p)$ points into

9485: $\mathcal{ W }_1$ and $-Jv(p)$ into $\mathcal{ W }_2$, for every $p \in

9486: E$.

9487: 

9488: In the case where $E=\R^n$, the classical edge of the wedge theorem

9489: states that there exists a neighborhood $\mathcal{ D}$ of $E$ in

9490: $\C^n$ such that every function which is continuous on 

9491: $\mathcal{ W}_1 \cup E \cup 

9492: \mathcal{ W }_2$ and holomorphic in $\mathcal{

9493: W}_1 \cup \mathcal{ W }_2$ extends holomorphically to $\mathcal{ D}$.

9494: Also, generalizations are known in the case where $f|_{ \mathcal{

9495: W}_1}$ and $f|_{\mathcal{ W }_2}$ have coinciding distributional

9496: boundary values on $E$.

9497: 

9498: The assumption about the matching up of boundary values along $E$ from

9499: $\mathcal{ W}_1$ and from $\mathcal{ W}_1$ is really needed, even if

9500: the two boundary values coincide on a thick subset of $E$. To support

9501: this observation, consider the following elementary example: the

9502: complex hyperplane $H := \{z_n=0\} \subset \C^n$ and the maximally

9503: real plane $E:= \R^n \subset \C^n$ intersect transversally in the

9504: $(n-1)$-dimensional totally real plane $C:= \{y=0, \ x_n =0\}$; the

9505: pair of wedges $\mathcal{ W}_1 := \{y_1 >0, \dots, y_n >0\}$ and

9506: $\mathcal{ W }_2:= \{y_1<0, \dots, y_n <0\}$ clearly form a double wedge

9507: attached to $E$; the function $\exp(-1 /z_n)$ restricted to the two

9508: wedges is holomorphic there, has coinciding boundary values on the

9509: thick set $E \backslash C$, but does not extend holomorphically to a

9510: neighborhood of $E$ in $\C^n$. Evidently, the envelope of holomorphy

9511: of the union of $\mathcal{ W}_1 \cup \mathcal{ W}_2$ together with a

9512: thin neighborhood of $E \backslash C$ in $\C^n$ does

9513: {\it not}\, contain any neighborhood of $E$ in $\C^n$.

9514: 

9515: Thus, in order to apply our removability theorems (which are

9516: essentially statements about envelopes of holomorphy), the first

9517: question is how to impose coincidence of boundary values on the

9518: edge. We shall first see that the fact that $C\subset E$ is exactly of

9519: codimension one in the above example is the limiting case for the

9520: obstruction to holomorphic extension. Since we want to treat

9521: also meromorphic extension, let us remind some definitions.

9522: 

9523: \subsection*{13.2.~Meromorphic functions and envelopes}

9524: Let $U$ be a domain in $\C^n$. A meromorphic function $f \in \mathcal{

9525: M} ( U)$ is a collection of equivalence classes of quotients of

9526: locally defined holomorphic functions. It defines a $P_1 (\C)$-valued

9527: function, which is single-valued only on some Zariski dense open

9528: subset $D_f \subset U$. More geometrically, we may represent $f$ by

9529: the closure $\Gamma_f$ of its graph $f|_{ D_f}$ over $D_f$, which

9530: always constitutes an irreducible $n$-dimensional complex analytic

9531: subset of $U \times P_1 (\C)$ with surjective, almost everywhere

9532: biholomorphic projection onto $U$ (equivalent definition). It is well

9533: known that the {\sl indeterminacy set} of $f$, namely the set of $z

9534: \in U$ over which the whole fiber $\{ z\} \times P_1 (\C)$ is

9535: contained in $\Gamma_f$, is an analytic subset of $U$ of codimension

9536: at least 2. It is the only set where $f$ is multivalued.

9537: 

9538: We shall constantly apply a theorem due to P.~Thullen

9539: (generalized by S.~Ivashkovitch in~\cite{ i} in the

9540: context of K\"ahler manifolds) according to which 

9541: the envelope of holomorphy of a domain in $\C^n$

9542: coincides with its envelope of meromorphy. As holomorphic

9543: functions are meromorphic, we shall state

9544: Lemma~13.4, Corollary~13.8 and Corollary~13.11 below

9545: directly for meromorphic functions.

9546: 

9547: \subsection*{13.3.~Edge of the wedge theorem over a maximally 

9548: real edge} Let $E \subset \C^n$ $(n \geq 2)$ be a real analytic

9549: maximally real submanifold, let $( \mathcal{ W }_1, \mathcal{

9550: W}_2)$ be a double wedge attached to $E$ and let $f_1$, $f_2$ be two

9551: meromorphic functions in $\mathcal{ W}_1$, $\mathcal{ W}_2$.

9552: 

9553: We need an assumption which tames the behaviour of their indeterminacy

9554: sets, as one tends towards the edge $E$ from either $\mathcal{ W}_1$

9555: or $\mathcal{ W}_2$. It will be sufficient to impose a matching up of

9556: their boundary values on the complement of a closed subset $C$ whose

9557: $(n-1)$-dimensional Hausdorff measure vanishes. Let ${\sf H}_d$ denote

9558: the $d$-dimensional Hausdorff measure.

9559: 

9560: \def\thelemma{13.4}\begin{lemma} If there is a closed subset $C\subset

9561: E$ with ${\sf H }_{ n-1} (C) =0$ such that both $\overline{

9562: \Gamma_{f_1 }}\cap [(E \backslash C) \times P_1(\C)]$ and $\overline{

9563: \Gamma_{ f_2 }} \cap [( E \backslash C)\times P_1(\C)]$ coincide with

9564: the graph of a {\rm (}single{\rm )} continuous mapping from $E

9565: \backslash C$ to $ P_1 (\C)$, then there exists a neighborhood

9566: $\mathcal{ D}$ of $E$ in $\C^n$ which depends only on $(\mathcal{ W}_1, 

9567: \mathcal{ W}_2)$ and a meromorphic function 

9568: \def\theequation{13.5}\begin{equation}

9569: f\in \mathcal{ M}\left(\mathcal{ D} \cup

9570: \mathcal{ W}_1 \cup \mathcal{ W}_2\right), 

9571: \end{equation}

9572: extending the $f_j$, namely such that $f|_{ \mathcal{ W }_j}

9573: =f_j$, for $j= 1,2$.

9574: \end{lemma}

9575: 

9576: \proof

9577: First of all, the assumption of continuous coincidence of boundary

9578: values enables us to apply the classical edge of the wedge theorem at

9579: each point of $E \backslash C$. This yields a neighborhood $\mathcal{

9580: D }_0$ of $E \backslash C$ in $\C^n$ and a meromorphic extension $f_0

9581: \in \mathcal{ M}( \mathcal{ D}_0 \cup \mathcal{ W}_1 \cup \mathcal{

9582: W}_2)$. We claim that the envelope of meromorphy of

9583: $\mathcal{ D}_0 \cup \mathcal{ W}_1 \cup \mathcal{ W}_2$ contains a

9584: neighborhood $\mathcal{ D}_1$ of $E$ in $\C^n$.

9585: 

9586: Indeed, this follows from a very elementary application of the

9587: continuity principle. Let $p\in C$ be arbitrary. After a local

9588: straightening, we may insure that $p$ is the origin, that $E=\R^n$, that

9589: $\mathcal{ W}_1$ contains $\{y_1 >0, \dots, y_n >0 \}$ and that

9590: $\mathcal{ W}_2$ contains $\{y_1 < 0, \dots, y_n < 0\}$.

9591: 

9592: Let us introduce the trivial family of analytic discs

9593: \def\theequation{13.6}\begin{equation}

9594: A_{c,x,v}( \zeta) :=

9595: \left(x_1+c(1+v_1)\zeta,\dots,

9596: x_n+(1+v_n)\zeta\right), 

9597: \end{equation}

9598: where $c >0$ is a sufficiently small fixed scaling factor, where $x

9599: \in \R^n$ is a small translation parameter and where $v\in \R^n$ is a

9600: small pivoting parameter. Clearly, $A_{c, x,v}( \partial^+ \Delta)$ is

9601: contained in $\mathcal{ W }_1$ and $A_{c, x,v} (\partial^- \Delta)$ is

9602: contained in $\mathcal{ W }_2$. However $A_{c, x,v}( \pm 1)$ may

9603: encounter $C$.

9604: 

9605: First of all, using the submersiveness of the two mappings $v \mapsto

9606: A_{c,0,v}(\pm 1)\in E$, we may find $v_0$ arbitrarily 

9607: close to the origin in $\R^n$ such that $A_{c,0,v_0}(\pm

9608: 1)$ does not belong to $C$. It follows that for all small

9609: translation vectors $q\in \C^n$, the disc boundary

9610: $A_{c,0,v_0}(\partial \Delta)+q$ is contained in the domain $\mathcal{

9611: D}_0\cup \mathcal{ W}_1 \cup \mathcal{ W}_2$.

9612: 

9613: Furthermore, because $C$ is of Hausdorff $(n-1)$-dimensional measure

9614: zero, for almost all $x\in \R^n$, the segment $A_{c,x,v_0}([-1,1])$

9615: does not meet $C$. It follows that for such $x$, the disc

9616: $A_{c,x,v_0}(\overline{ \Delta})$ is contained in the domain

9617: $\mathcal{ D}_0\cup \mathcal{ W}_1 \cup \mathcal{ W}_2$. We deduce

9618: that every disc $A_{c,0,v_0}(\partial \Delta)+q$ is analytically

9619: isotopic to a point in $\mathcal{ D}_0\cup \mathcal{ W}_1 \cup

9620: \mathcal{ W}_2$. An application of the continuity principle yields

9621: meromorphic extension to a neighborhood of 

9622: $p=A_{c,0,0}(0)\in C$.

9623: 

9624: In sum, we have constructed a neighborhood $\mathcal{ D}_1$ of $E$ in

9625: $\C^n$ and a meromorphic extension $f\in \mathcal{ M}( \mathcal{

9626: D}_1)$. But $\mathcal{ D}_1$ is not independent of $(f_1,f_2)$, since

9627: it depends on $C$. Fortunately, once we know meromorphic extension to

9628: a neighborhood $\mathcal{ D}_1$ of $E$ in $\C^n$, we may reemploy the

9629: analytic disc technique of the classical edge of the wedge theorem to

9630: describe a neighborhood $\mathcal{ D}$ of $E$ in $\C^n$ which depends

9631: only on $(\mathcal{ W}_1, \mathcal{ W}_2)$ ({\it see}\, the end of the

9632: proof of Corollary~13.8 below for more arguments). This completes the

9633: proof of Lemma~13.4.

9634: \endproof

9635: 

9636: \subsection*{13.7.~Edge of the wedge theorem over an edge of positive

9637: CR dimension} Let $M$ be a $\mathcal{ C}^{2,\alpha}$-smooth generic

9638: submanifold of $\C^n$ of positive CR dimension and let $C$ be a proper

9639: closed subset of $M$ such that $M$ and $M\backslash C$ are globally

9640: minimal. In~\cite{ mp3}, Theorem~1.1, it was shown as a main theorem

9641: that every such closed subset $C$ of $M$ is $CR$-, $\mathcal{ W}$- and

9642: $L^{\sf p}$-removable. We may formulate the following application,

9643: where, for simplicity, we assume local minimality at every point.

9644: 

9645: \def\thecorollary{13.8}\begin{corollary}

9646: Let $E \subset \C^n$ $( n \geq 2)$ be a generic manifold of class

9647: $\mathcal{ C }^{ 2, \alpha}$ of positive CR dimension which is locally

9648: minimal at every point, let $(\mathcal{ W}_1, \mathcal{ W}_2)$ a

9649: double wedge attached to $E$ and let two meromorphic functions $f_j\in

9650: \mathcal{ M} ( \mathcal{ W}_j)$ for $j=1,2$. If there is a closed

9651: subset $C\subset E$ with ${\sf H }_{ n-1} (C) =0$ such that both

9652: $\overline{ \Gamma_{f_1 }} \cap [(E \backslash C) \times P_1( \C)]$

9653: and $\overline{ \Gamma_{ f_2 }} \cap [( E \backslash C)\times P_1(

9654: \C)]$ coincide with the graph of a {\rm (}single{\rm )} continuous

9655: mapping from $E \backslash C$ to $ P_1 (\C)$, then there exists a

9656: neighborhood $\mathcal{ D }$ of $E$ in $\C^n$ which depends only on

9657: $(\mathcal{ W}_1, \mathcal{ W}_2)$ and a meromorphic function

9658: \def\theequation{13.9}\begin{equation}

9659: f\in \mathcal{ M} \left(\mathcal{ W}_1 \cup

9660: \mathcal{ D} \cup \mathcal{ W}_2\right), 

9661: \end{equation}

9662: extending the $f_j$, namely such that $f|_{ \mathcal{ W }_j}

9663: =f_j$, for $j = 1,2$.

9664: \end{corollary}

9665: 

9666: \proof

9667: Applying the classical edge of the wedge theorem, we get a meromorphic

9668: extension $f_0 \in \mathcal{ M}( \mathcal{ W}_1 \cup \mathcal{ D}_0

9669: \cup \mathcal{ W}_2)$, where $\mathcal{ D}_0$ is some open

9670: neighborhood of $E \backslash C$ in $\C^n$. Next, we include $E$ in a

9671: CR manifold $M$ with $M \subset \mathcal{ W}_1 \cup E \cup \mathcal{

9672: W}_2$ and $\dim_\R M = 1 + \dim_\R E$, as shown in the following

9673: figure.

9674: 

9675: \bigskip

9676: \begin{center}

9677: \input double-wedge.pstex_t

9678: \end{center}

9679: 

9680: Of course, the domain $\mathcal{ W}_1 \cup \mathcal{ D}_0 \cup

9681: \mathcal{ W}_2$ constitues a (rather thick) wedge attached to

9682: $M\backslash C$. Since $E$ is locally minimal at every point and

9683: since one CR tangential direction of $M$ is transversal to $E$, both

9684: $M$ and $M \backslash \Sigma$ are both globally minimal. Applying

9685: Theorem~1.1 in~\cite{ mp2}, we deduce that there exists a wedge

9686: $\mathcal{ W}$ attached to $M$ which is contained in the envelope of

9687: meromorphy of $\mathcal{ W}_1 \cup \mathcal{ D}_0 \cup \mathcal{

9688: W}_2$. We then claim that there exists a neighborhood $\mathcal{ D}$

9689: of $p$ in $\C^n$, which depends only on $(\mathcal{ W}_1, \mathcal{

9690: W}_2)$ such that $\mathcal{ D}$ is contained in the envelope of

9691: meromorphy of $\mathcal{ W}_1 \cup \mathcal{ W} \cup \mathcal{ W}_2$.

9692: 

9693: Indeed, by deforming slightly $M$ inside $\mathcal{ W}$ near $E$, we

9694: get a $\mathcal{ C}^{2, \alpha}$-smooth generic submanifold $M^d

9695: \subset \mathcal{ W}_1 \cup \mathcal{ W} \cup \mathcal{ W}_2$. Instead

9696: of functions meromorphic in the disconnected open set $\mathcal{ W}_1

9697: \cup \mathcal{ W}_2$, we now consider meromorphic functions in the

9698: {\it connected}\, open set $\mathcal{ W}_1 \cup \mathcal{ W} \cup

9699: \mathcal{ W}_2$, which is a neighborhood of $M^d$ in $\C^n$. Then by

9700: following the proof of the edge of the wedge theorem given in~\cite{a}

9701: and applying the continuity principle, one deduces meromorphic

9702: extension to a neighborhood $\mathcal{ D }^d$ in $\C^n$ of the

9703: deformed submanifold $M^d$. Since the size of $\mathcal{ W }_1$ and

9704: the size of $\mathcal{ W }_2$ are uniform with respect to $d$, the

9705: size of the domain $\mathcal{ D }^d$ is also uniform with respect to

9706: $d$, as follows from the stability of the edge of the wedge theorem

9707: established in~\cite{ a}, since it relies on E.~Bishop's equation.

9708: Hence for $M^d$ sufficiently close to $M$, the domain $\mathcal{ D

9709: }^d$ contains a neighborhood of $p$ in $\C^n$. This completes the

9710: proof of Corollary~13.8.

9711: \endproof

9712: 

9713: \subsection*{13.10.~Hartogs-Bochner phenomenon and edge of the wedge 

9714: theorem} Next we turn to the question whether a Hartogs-Bochner

9715: phenomenon holds in presence of a double wedge. More

9716: precisely we ask when it is sufficient to require coincidence of

9717: boundary values only outside some compact $K\subset E$. Let us first

9718: look at a prototypical case where the answer is particularly neat,

9719: thanks to Theorem~1.1. Obviously, the proof is totally similar to the

9720: proof of Corollary~13.8 and will not be repeated.

9721: 

9722: \def\thecorollary{13.11}\begin{corollary}

9723: Let $E \subset \C^2$ be an embedded real analytic totally real disc,

9724: let $(\mathcal{ W }_1, \mathcal{ W }_2)$ a double wedge attached to

9725: $E$ and let two meromorphic functions $f_j\in \mathcal{ M} ( \mathcal{

9726: W}_j)$ for $j=1,2$. If there is a compact subset $K\subset E$ such

9727: that both $\overline{ \Gamma_{f_1 }} \cap [(E \backslash K) \times

9728: P_1( \C)]$ and $\overline{ \Gamma_{ f_2 }} \cap [( E \backslash

9729: K)\times P_1( \C)]$ coincide with the graph of a {\rm (}single{\rm )}

9730: continuous mapping from $E \backslash K$ to $ P_1 (\C)$, then there

9731: exists a neighborhood $\mathcal{ D}$ of $E$ in $\C^2$ which depends

9732: only on $(\mathcal{ W}_1,\mathcal{ W}_2)$ and a meromorphic function

9733: \def\theequation{13.12}\begin{equation}

9734: f\in \mathcal{ M}\left(\mathcal{ W}_1 \cup

9735: \mathcal{ D} \cup \mathcal{ W}_2\right), 

9736: \end{equation}

9737: extending the $f_j$, namely such that $f|_{ \mathcal{ W }_j}

9738: =f_j$, for $j = 1,2$.

9739: \end{corollary}

9740: 

9741: In order to find the most general application of Theorems~1.2 and

9742: 1.2', we first remark that in {\sc Figure~24}, we have a considerable

9743: freedom in the choice of the generic submanifolds $M$ of CR dimension

9744: $1$ with $E\subset M \subset \mathcal{ W}_1 \cup E \cup \mathcal{ W}_2$,

9745: depending on the aperture of $\mathcal{ W }_1$ and $\mathcal{ W

9746: }_2$. Since $\dim_\R M = 1+ \dim_\R E$, the tangent space

9747: to such an $M$ at

9748: a point $p\in E$ is uniquely determined by some nonzero vector $v_p\in

9749: T_p \C^n / T_p E$. In order that $M$ is locally contained in

9750: $\mathcal{ W}_1 \cup E \cup \mathcal{ W}_2$, it is necessary and

9751: sufficient that either $Jv$ points in $\mathcal{ W}_1$ and $-Jv$

9752: points in $\mathcal{ W}_2$, or vice versa, depending on the

9753: orientations of $\mathcal{ W}_1$ and $\mathcal{ W}_2$ with respect to

9754: $E$. Without loss of generality, after a possible shrinking, we can

9755: therefore assume that the cones of $\mathcal{ W}_1$ and of $\mathcal{

9756: W}_2$ are exactly opposite to each other at every point of $E$;

9757: indeed, it would be impossible to construct an $M$ locally contained

9758: in $\mathcal{ W}_1 \cup E \cup \mathcal{ W}_2$ which satisfies $T_p M

9759: = T_p E \oplus \R v_p$ at a point $p\in E$, in the case where the

9760: vector $Jv_p$ points in the cone at $p$ of $\mathcal{ W}_1$ but $-J

9761: v_p$ lies outside the cone at $p$ of $\mathcal{ W}_2$, or

9762: vice versa.

9763: 

9764: Assuming $\mathcal{ W}_2$ to be opposite to $\mathcal{ W}_1$, let us

9765: define an induced field of open cones $p\mapsto {\sf C }_p^{ \mathcal{

9766: W }_1, \mathcal{ W}_2 }$ as follows: a nonzero vector $v_p \in T_p E

9767: \backslash \{0\}$ belongs to ${\sf C }_p^{ \mathcal{ W }_1, \mathcal{

9768: W}_2 }$ if either $Jv$ or $-Jv$ points into $\mathcal{ W }_1$. A

9769: nowhere vanishing vector field $p \mapsto v (p)$ is said to be {\sl

9770: directed by ${\sf C }_p^{ \mathcal{ W }_1, \mathcal{ W}_2 }$} if $v(p)

9771: \in {\sf C}_p^{ \mathcal{ W }_1, \mathcal{ W}_2 }$ for every $p \in

9772: E$. Clearly, for every vector field $p\mapsto v(p)$ directed by $

9773: p\mapsto {\sf C }_p^{ \mathcal{ W }_1, \mathcal{ W}_2 }$, we may

9774: construct a $\mathcal{ C}^{2,\alpha}$-smooth semi-local generic

9775: submanifold $M$ containing $E$, contained in $\mathcal{ W}_1 \cup E

9776: \cup \mathcal{ W}_2$ which satisfies $T_p M = T_p E \oplus J v(p)$ at

9777: every point $p\in E$.

9778: 

9779: In the statement of Theorem~1.2', we defined the condition $\mathcal{

9780: F}_{M^1}^c\{C\}$ with respect to some CR manifold $M$ containing the

9781: totally real manifold $M^1$. But $M$ entered in the definition only

9782: via the characteristic foliation induced on $M^1$. Hence its

9783: reasonable to define a more general nontransversality property, by

9784: replacing the characteristic foliation by the foliation induced by any

9785: vector field directed by the field of cones $p\mapsto {\sf

9786: C}_p^{\mathcal{ W}_1, \mathcal{ W}_2}$, as follows:

9787: 

9788: \smallskip

9789: \begin{itemize}

9790: \item[

9791: $\mathcal{ F}_{ \mathcal{ W}_1, \mathcal{ W}_2 } \{C\}:$] For every

9792: closed subset $C' \subset C$ there is a smooth vector field $ p

9793: \mapsto v(p)$ directed by the field

9794: of cones $p\mapsto {\sf C }_p^{ \mathcal{ W}_1, \mathcal{ W

9795: }_2}$ such that there exists a simple $\mathcal{ C}^{2,

9796: \alpha}$-smooth curve $\gamma': [-1,1] \to E$ whose range

9797: $\gamma'([-1,1])$ is contained in a single integral curve of $p

9798: \mapsto v (p)$ with $\gamma'(-1) \not \in C'$, $\gamma' (0) \in C'$

9799: and $\gamma' (1) \not \in C'$, there exists a local $(n -

9800: 1)$-dimensional transversal $R \subset E$ to $\gamma'$ passing through

9801: $\gamma' (0)$ and there exists a thin open neighborhood $V$ of

9802: $\gamma' ( [-1, 1])$ in $E$ such that if $\pi: V \to R$ denotes the

9803: semi-local projection parallel to the flow lines of $v$, then

9804: $\gamma'(0)$ lies on the boundary, relatively to the topology of $R$,

9805: of $\pi (C' \cap V)$.

9806: \end{itemize}

9807: \smallskip

9808: 

9809: The proper application of Theorem~1.2' is the following.

9810: Its proof follows by a direct examination of the

9811: proof of Theorem~1.2'.

9812: 

9813: \def\thecorollary{13.13}\begin{corollary}

9814: Let $E \subset \C^n$ $(n \geq 2)$ be a real analytic maximally real

9815: submanifold, let $(\mathcal{ W }_1, \mathcal{ W }_2)$ a double wedge

9816: attached to $E$. Let $C$ be a proper closed subset of $E$ satisfying

9817: the nontransversality property $\mathcal{ F}_{ \mathcal{ W }_1,

9818: \mathcal{ W}_2} \{C\}$ above. Let two meromorphic functions $f_j\in

9819: \mathcal{ M} ( \mathcal{ W}_j)$ for $j=1,2$ such that both $\overline{

9820: \Gamma_{f_1 }} \cap [(E \backslash C) \times P_1( \C)]$ and

9821: $\overline{ \Gamma_{ f_2 }} \cap [( E \backslash C)\times P_1( \C)]$

9822: coincide with the graph of a {\rm (}single{\rm )} continuous mapping

9823: from $E \backslash C$ to $ P_1 (\C)$, Then there exists a neighborhood

9824: $\mathcal{ D}$ of $E$ in $\C^2$ which depends

9825: only on $(\mathcal{ W}_1,\mathcal{ W}_2)$ and a meromorphic function

9826: \def\theequation{13.14}\begin{equation}

9827: f\in \mathcal{ M}\left(\mathcal{ W}_1 \cup

9828: \mathcal{ D} \cup \mathcal{ W}_2\right), 

9829: \end{equation}

9830: extending the $f_j$, namely such that $f|_{ \mathcal{ W }_j} =f_j$,

9831: for $j = 1,2$. 

9832: \end{corollary}

9833: 

9834: \subsection*{13.15.~Further applications}

9835: We now conclude this section by suggesting two applications of

9836: Theorem~1.2' in higher dimensions, in the case where $M^1$ is not

9837: everywhere totally real. However, we must mention that we did not try

9838: to generalize the results of E.~Bishop to understand the local

9839: geometry of complex tangencies of generic submanifolds of CR dimension

9840: $1$ in $\C^n$, for $n\geq 3$. Consequently, our formulations

9841: should be considered as mild generalizations of

9842: Theorem~1.1 and~1.3.

9843: 

9844: Thus, let $M$ be a $\mathcal{ C }^{2, \alpha}$-smooth generic

9845: submanifold of $\C^n$ {\rm (}$n \geq 2${\rm )} of CR dimension $1$,

9846: let $M^1$ be a codimension one submanifold of $M$ which is maximally

9847: real except at every point of some proper closed subset $E \subset

9848: M^1$. Let $C$ be a proper closed subset of $M^1$. For simplicity, we

9849: assume that $M$ is locally minimal at every point, an assumption which

9850: insures that for every closed subset $\widetilde{ C}$ of $M$, 

9851: both $M$ and $M \backslash \widetilde{ C}$ are globally

9852: minimal.

9853: 

9854: Firstly, applying Theorem~1.2' to remove

9855: $C \cap (M^1 \backslash E)$ and then Theorem~1.1 of~\cite{ mp2}

9856: to remove $C \cap E$, we deduce the following. 

9857: 

9858: \def\thecorollary{13.16}\begin{corollary}

9859: Assume that $E$ is of vanishing $(n - 1)$-dimensional Hausdorff

9860: content, and that the nontransversality condition $\mathcal{ F}_{M^1

9861: \backslash E}^c \{C \cap (M^1 \backslash E)\}$ holds. Then $C$ is

9862: CR-, $\mathcal{ W}$- and $L^{ \sf p}$-removable.

9863: \end{corollary}

9864: 

9865: Secondly, we may generalize the notion of hyperbolic tree and assume

9866: that $E$ consists of finitely many compact submanifolds of codimension

9867: $2$ in $M^1$ joined by a collection of finitely many codimension one

9868: submanifolds of $M^1$ with boundaries in $E$ which are foliated by

9869: characteristic curves. Under some easily found assumptions, one could

9870: formulate a second corollary analogous to Theorem~1.3.

9871: 

9872: \section*{\S14.~An example of a nonremovable three-dimensional torus}

9873: 

9874: This final section is devoted to exhibit a crucial example of a closed

9875: subset $C$ violating the main nontransversality condition $\mathcal{

9876: F}_{M^1}^c\{ C\}$ of Theorem~1.2' such that $C$ is truly nonremovable.

9877: In addition, we may require that $M$ and $M^1$ have the simplest

9878: possible topology.

9879: 

9880: \def\thelemma{14.1}\begin{lemma}

9881: There exists a triple $(M, M^1, C)$, where

9882: \begin{itemize}

9883: \item[{\bf (i)}]

9884: $M$ is a $\mathcal{ C}^\infty$-smooth generic submanifold in

9885: $\C^3$ of CR dimension $1$, diffeomorphic to a real $4$-ball{\rm ;}

9886: \item[{\bf (ii)}]

9887: $M^1$ is a $\mathcal{ C}^\infty$-smooth one-codimensional

9888: submanifold of $M$ which is maximally real in $\C^n$ and diffeomorphic

9889: to a real $3$-ball{\rm ;}

9890: \item[{\bf (iii)}]

9891: $C$ is a compact subset of $M^1$

9892: diffeomorphic to a real three-dimensional torus which is everywhere

9893: transversal to the characteristic foliation $\mathcal{ F}_{M^1}^c$, 

9894: hence the nontransversality condition $\mathcal{ F}_{M^1}^c\{C\}$ of

9895: Theorem~1.2' clearly does not hold{\rm ;}

9896: \item[{\bf (iv)}]

9897: $M$ of finite type $4$ in the sense of

9898: T.~Bloom and I.~Graham at every point, hence globally minimal,

9899: \end{itemize} 

9900: such that $C$ is neither CR- nor 

9901: $\mathcal{ W}$- nor $L^{\sf p}$-removable with respect to $M$.

9902: \end{lemma}

9903: 

9904: By type 4 at a point $p\in M$, we mean of course that

9905: the Lie brackets of the complex tangent bundle $T^cM$ up to length $4$

9906: generate $T_pM$.

9907: 

9908: \subsection*{14.2.~The geometric recipe}

9909: We first construct the $3$-torus $C$, then construct the maximally

9910: real $M^1$ and finally define $M$ as a certain thickening of

9911: $M^1$. The argument for insuring global minimality of $M$ involves

9912: computations with Lie brackets and is postponed to the

9913: end.

9914: 

9915: Firstly, in $\R^3 =\R^3\oplus i \{0\}\subset \C^3$ equipped with the

9916: coordinates $(x_1,x_2,x_3)$, where $x_j={\rm Re}\, z_j$ for $j=1,2,3$,

9917: pick the ``standard'' $2$-dimensional torus $T^2$ of Cartesian

9918: equation

9919: \def\theequation{14.3}\begin{equation}

9920: \left(\sqrt{x_1^2+x_2^2}-2\right)^2+x_3^2=1.

9921: \end{equation}

9922: This torus is stable under the rotations directed by the $x_3$-axis;

9923: its intersection with the $(x_1,x_3)$-plane consists of two circles of

9924: radius $1$ centered at the points $x_1=2$ and $x_1=-2$; it bounds a

9925: three-dimensional open ``full'' torus $T^3$; both $T^2$ and $T^3$ are

9926: contained in the ball $B^3$ of radius $5$ centered at the origin.

9927: 

9928: It is better to drop the square root: one checks that the equations

9929: of $T^2$ and $T^3$ are equally given by $T^2:=\{\rho =0\}$ and

9930: $T^3:= \{\rho<0\}$, by means of the {\it polynomial}\, defining 

9931: function

9932: \def\theequation{14.4}\begin{equation}

9933: \rho(x_1,x_2,x_3):=

9934: (x_1^2+x_2^2+x_3^2+3)^2-16\, (x_1^2+x_2^2),

9935: \end{equation}

9936: which has nonvanishing differential at every point of $T^2$.

9937: Consequently, the extrinsinc complexification of $T^2$, namely the

9938: complex hypersurface defined by

9939: \def\theequation{14.5}\begin{equation}

9940: \Sigma:= \{(z_1,z_2,z_3)\in \C^3: \

9941: \rho(z_1,z_2,z_3)=0\}

9942: \end{equation}

9943: cuts $\R^3$ along $T^2$ with the transversality property $T_x \R^3

9944: \cap T_x\Sigma= T_x T^2$ for every point $x\in T^2$.

9945: 

9946: Secondly, according to G.~Reeb ({\it see}\, \cite{ cln}, pp.~25--27;

9947: {\it see}\, also the figures there), by considering the space

9948: $\R^3\equiv S^3 \backslash \{ \infty\}$ as a punctured

9949: three-dimensional sphere $S^3$, one may glue a second

9950: three-dimensional full torus $\widetilde{ T}^3$ to $T^3$ along $T^2$

9951: with $\infty\in \widetilde{ T}^3$ and then construct a foliation of

9952: $S^3$ by $2$-dimensional surfaces all of whose leaves, except one, are

9953: diffeomorphic to $\R^2$, are contained in either $T^3$ or in

9954: $\widetilde{ T}^3$ and are accumulating on $T^2$, and finally, whose

9955: single compact leaf is the above $2$-torus $T^2$. This yields the

9956: so-called {\it Reeb foliation} of $S^3$, which is $\mathcal{

9957: C}^\infty$-smooth and orientable. Consequently, there exists a

9958: $\mathcal{ C}^\infty$ smooth vector field $L= a_1(x)\, \partial_{x_1}+

9959: a_2(x) \, \partial_{x_2} + a_3(x) \, \partial_{x_3}$ of norm $1$,

9960: namely $a_1(x)^2+ a_2(x)^2+ a_3(x)^2=1$ for every $x\in \R^3$, which

9961: is everywhere orthogonal (with respect to the standard Euclidean

9962: structure) to the leaves of the Reeb foliation. Geometrically, the

9963: integral curves of $L$ accumulate asymptotically on the two nodal

9964: (central) circles of $T^3$ and of $\widetilde{ T}^3$.

9965: 

9966: The open ball $B^3\subset \R^3$ of radius $5$ centered at the origin

9967: will be our maximally real submanifold $M^1$. The two-dimensionaly

9968: torus $T^2$ will be our nonremovable closed set $C$. The integral

9969: curves of the vector field $L$ will be our characteristic lines.

9970: Since $L$ is orthogonal to $T^2$, these characteristic lines will of

9971: course be everywhere transverse to $C$, so that $\mathcal{

9972: F}_{M^1}^c\{C\}$ clearly does not hold.

9973: 

9974: Thirdly, it remains to construct the generic submanifold $M$ of

9975: CR dimension $1$ containing $M^1$ and to check that $C$ will be

9976: nonremovable.

9977: 

9978: First of all, we notice that $L$ provides the characteristic

9979: directions of $M^1$ if and only if $T_x M= T_p\R^3\oplus \R \, J\,

9980: L(x)$ for every point $x\in M^1\equiv B^3$. Consequently, all

9981: submanifolds $M\subset \C^3$ obtained by slightly thickening $M^1$ in

9982: the direction of $J\, L(x)$ will be convenient; in other words, only

9983: the first jet of $M$ along $M^1$ is prescribed by our choice of the

9984: characterisctic vector field $L$. Notice that all such thin strips

9985: $M$ along $M^1$ will be diffeomorphic to a real $4$-ball.

9986: 

9987: The fact that $C$ is nonremovable for all such generic submanifolds

9988: $M$ is now clear: the hypersurface $\Sigma=\{z\in \C^3 : \, \rho(z)

9989: =0\}$ satisfying $T_x\Sigma= T_xT^2\oplus \R \, J\, T_xT^2$ for all

9990: $x\in T^2$ and $L$ being transversal to $T^2$, we easily deduce the

9991: transversality property $T_x \Sigma + T_x M= T_x\C^3$ for all $x\in

9992: T^2$, a geometric property which insures that the holomorphic function

9993: $1/\rho(z)$, which is CR on $M\backslash C$, does not extend

9994: holomorphically to any wedge of edge $M$ at any point of $C$.

9995: Intuitively, $T_x\Sigma/T_xM$ absorbs all the normal space 

9996: $T_x\C^3/T_x M$ at every point $x\in T^2$, leaving no room

9997: for any open cone.

9998: 

9999: Finally, to fulfill all the hypotheses of Theorem~1.2' (except of

10000: course $\mathcal{ F}_{M^1}^c\{C\}$), we have to insure that $M$ is

10001: globally minimal. We claim that by bending strongly the second and the

10002: fourth order jet of $M$ along $M^1$ (without modifying the first order

10003: jet which must be prescribed by $J\, L$), one may insure that $M$ is

10004: of type 4 in the sense of T.~Bloom and I.~Graham at every point of

10005: $M^1$; since being of finite type is an open property, it follows that

10006: $M$ is finite type at every point provided that, as a strip, $M$ is

10007: sufficiently thin along $M^1$. As is known, finite-typeness at every

10008: point implies local minimality at every point which in turn implies

10009: global minimality. This completes the recipe.

10010: 

10011: We would like to mention that by following a similar recipe, one may

10012: construct an elementary example of a non-removable compact subset of a

10013: generic submanifold of codimension one diffeomorphic to a $4$-ball

10014: lying in a globally minimal hypersurface in $\C^3$ which is (also)

10015: diffeomorphic to a $5$-ball ({\it cf.} \cite{ js}).

10016: 

10017: \subsection*{14.6.~Finite-typisation}

10018: Thus, it remains to construct a generic submanifold $M\subset \C^3$ of

10019: CR dimension $1$ satisfying $T_x M = T_xM^1 \oplus \R \, J \, L(x)$

10020: for every $x\in M^1$, which is of {\it type $4$ at every point $x\in

10021: M^1$}.

10022: 

10023: First of all, let us denote by $L=a_1(x)\, \partial_{x_1}+ a_2(x) \,

10024: \partial_{x_2}+ a_3(x) \, \partial_{x_3}$ the unit vector field which

10025: was constructed as a field orthogonal to the Reeb foliation: it is

10026: defined over $\R^3$ and has $\mathcal{ C}^\infty$-smooth coefficients

10027: satisfying $a_1(x)^2+ a_2(x)^2+ a_3(x)^2= 1$ for all $x\in\R^2$. The

10028: two-dimensional quotient vector bundle $T\R^3 / (\R L)$ with

10029: contractible base being necessarily trivial, it follows that we can

10030: complete $L$ by two other $\mathcal{ C }^\infty$-smooth unit

10031: vector fields $K^1$ and $K^2$ defined over $\R^3$ such that the triple

10032: $(L(x), K^1(x), K^2(x))$ forms a direct orthonormal frame at every

10033: point $x\in \R^3$. Let us denote the coefficients of $K^1$ and of

10034: $K^2$ by

10035: \def\theequation{14.7}\begin{equation}

10036: \aligned

10037: K^1 

10038: & \

10039: = \rho_1 \, \partial_{x_1}+ 

10040: \rho_2\, \partial_{x_2} + 

10041: \rho_3 \, \partial_{x_3},\\

10042: K^2 

10043: & \

10044: =

10045: r_1 \, \partial_{x_1}+ 

10046: r_2 \, \partial_{x_2}+ 

10047: r_3 \, \partial_{x_3},

10048: \endaligned

10049: \end{equation}

10050: where $\rho_j$ and $r_j$ for $j=1,2,3$ are $\mathcal{

10051: C}^\infty$-smooth functions of $x\in \R^3$ satisfying $\rho_1^2+

10052: \rho_2^2+ \rho_3^2=1$ and $r_1^2+r_2^2+r_3^2=1$. In our case, $K^1$

10053: and $K^2$ may even be constructed directly by means of a

10054: trivialization of the bundle tangent to the Reeb foliation.

10055: 

10056: Let $P>0$ be a constant, which will be chosen later to be large. Since

10057: by construction we have the two orthogonality relations $a_1 \rho_1 +

10058: a_2 \rho_2 + a_3 \rho_3= 0$ and $a_1 r_1 + a_2 r_2 + a_3 r_3= 0$, it

10059: follows that every generic submanifold $M_P\subset 

10060: \C^3$ defined by the two

10061: Cartesian equations

10062: \def\theequation{14.8}\begin{equation}

10063: \aligned

10064: 0 

10065: & \

10066: = \rho 

10067: = 

10068: y_1 \, \rho_1(x) + 

10069: y_2 \, \rho_2(x) +

10070: y_3 \, \rho_3(x) + 

10071: P 

10072: \left[

10073: y_1^2+ y_2^2+y_3^2

10074: \right], \\

10075: 0 

10076: & \

10077: = r 

10078: = 

10079: y_1\, r_1(x) + 

10080: y_2 \, r_2(x) + 

10081: y_3 \, r_3(x)+

10082: P^3

10083: \left[

10084: y_1^4 + y_2^4 + y_3^4

10085: \right]

10086: \endaligned

10087: \end{equation}

10088: enjoys the property that the vector field $J L(x)= a_1(x) \,

10089: \partial_{y_1}+ a_2(x) \, \partial_{x_2}+ a_3(x) \, \partial_{x_3}$ is

10090: tangent to $M_P$ at every $x\in \R^3$. As desired, we deduce that

10091: $T_x^c M = \R L(x) \oplus J \R L(x)$ for every $x\in \R^3$, a property

10092: which insures that $\R L(x)$ is the characteristic direction of $M^1$

10093: in $M_P$, independently of $P$.

10094: 

10095: To complete the final minimalization argument for the construction of

10096: a nonremovable compact set $C:= T^2 \subset M^1\subset M$ which

10097: appears in the Introduction, it suffices now to apply the following

10098: lemma with $R= 5$. Though calculatory, its proof is totally

10099: elementary.

10100: 

10101: \def\thelemma{14.9}\begin{lemma}

10102: For every $R>0$, there exist $P>0$ sufficiently large

10103: such that $M_P$ is of type $4$ at every point

10104: $x\in \R^3$ with $x_1^2+ x_2^2+ x_3^2\leq R^2$. 

10105: \end{lemma}

10106: 

10107: \proof

10108: As above, let $M_P = \{z\in \C^3: \, \rho =r =0\}$. By writing the

10109: tangency condition, one checks immediately that the one-dimensional

10110: complex vector bundle $T^{1, 0}M_P$ is generated over $\C$ by the

10111: vector field $\LL := A_1 \, \partial_{z_1}+ A_2 \, \partial_{ z_2}+

10112: A_3\, \partial_{ z_3}$, with the explicit expressions

10113: \def\theequation{14.10}\begin{equation}

10114: \aligned

10115: A_1:= 

10116: & \ 

10117: 4\rho_{z_3}r_{z_2}- 4 \rho_{z_2} r_{z_3}, \\

10118: A_2:= 

10119: & \ 

10120: 4\rho_{z_1}r_{z_3}- 4 \rho_{z_3} r_{z_1}, \\

10121: A_3:= 

10122: & \ 

10123: 4\rho_{z_2}r_{z_1}- 4 \rho_{z_1} r_{z_2}. \\

10124: \endaligned

10125: \end{equation}

10126: Using the expressions~\thetag{ 14.8} for $\rho$ and $r$, we see that

10127: these three components restrict on $\{y=0\}$ as the Pl\"ucker

10128: coordinates of the bivector $(K^1,K^2)$, namely

10129: \def\theequation{14.11}\begin{equation}

10130: \aligned

10131: A_1 \vert_{y=0}=

10132: & \

10133: \rho_2 r_3- \rho_3 r_2=:

10134: \Delta_{2,3}, \\

10135: A_2\vert_{y=0}=

10136: & \

10137: \rho_3 r_1- \rho_1 r_3

10138: =:\Delta_{3,1}, \\

10139: A_3 \vert_{y=0}=

10140: & \

10141: \rho_1 r_2- \rho_2 r_1

10142: =:\Delta_{1,2}.

10143: \endaligned

10144: \end{equation}

10145: As $K^1$ and $K^2$ are of norm $1$ and orthogonal at every point, it

10146: follows by direct computation that $\Delta_{2,3}^2 + \Delta_{3,1}^2+

10147: \Delta_{ 1,2}^2 = 1$ and that the vector of coordinates $(\Delta_{

10148: 2,3}, \Delta_{3,1}, \Delta_{1,2})$ is orthogonal to both $K^1$ and

10149: $K^2$. Moreover, as the orthonormal trihedron $(L(x), K^1(x), K^2(x))$

10150: is direct at every point, we deduce that necessarily

10151: \def\theequation{14.12}\begin{equation}

10152: \Delta_{2,3} \equiv a_1, \ \ \ \ \

10153: \Delta_{3,1} \equiv a_2, \ \ \ \ \

10154: \Delta_{1,2} \equiv a_3.

10155: \end{equation}

10156: 

10157: Next, we compute in length $A_1$, $A_2$ and $A_3$ using~\thetag{

10158: 14.8}. As their complete explicit development will not be crucial for

10159: the sequel and as we shall perform with them differentiations and

10160: linear combinations yielding relatively complicated expressions, let

10161: us adopt the following notation: by $\mathcal{ R}^0$, we denote

10162: various expressions which are polynomials in the jets of the functions

10163: $\rho_1,\rho_2,\rho_3$ and $r_1,r_2,r_3$. Similarly, by $\mathcal{

10164: R}^{I}$, by $\mathcal{ R}^{II}$, by $\mathcal{ R}^{III}$ and by

10165: $\mathcal{ R}^{IV}$, we denote polynomials in the transverse variables

10166: $(y_1,y_2,y_3)$ which are homogeneous of degree $1$, $2$, $3$ and $4$

10167: and have as coefficients various expressions $\mathcal{ R}^0$.

10168: 

10169: Importantly, we make the convention that such expressions $\mathcal{

10170: R}^0$, $\mathcal{ R}^I$, $\mathcal{ R}^{II}$, $\mathcal{ R}^{III}$ and

10171: $\mathcal{ R}^{IV}$ should be totally independent of the constant

10172: $P$. Consequently, if $P$ appears somehow, we shall write it as a

10173: factor, as for instance in $P \, \mathcal{ R}^I$ or in $P^3 \,

10174: \mathcal{ R}^{III}$.

10175: 

10176: With this convention at hand, we may develope~\thetag{ 14.10} using the

10177: expressions~\thetag{ 14.8} by writing out only the terms which will be

10178: useful in the sequel and by treating the rest as controlled

10179: remainders. Let us detail the computation of $A_1$:

10180: \def\theequation{14.13}\begin{equation}

10181: \aligned

10182: A_1 

10183: & \

10184: =

10185: 4 \, 

10186: \left[

10187: -\frac{ i}{2} \rho_3 - iP y_3 + \mathcal{ R}^I

10188: \right] 

10189: \left[

10190: -\frac{ i}{2} r_2 - 2i P^3 y_2^3 + 

10191: \mathcal{ R}^I

10192: \right]- \\

10193: & \ \ \ \ \ \ \ 

10194: - 

10195: 4\, 

10196: \left[

10197: -\frac{ i}{2} \rho_2 - iP y_2 + \mathcal{ R}^I

10198: \right] 

10199: \left[

10200: - \frac{ i}{2} r_3 - 2i P^3 y_3^3 + 

10201: \mathcal{ R}^I

10202: \right] \\

10203: & 

10204: \

10205: =

10206: -\rho_3 r_2 -4 P^3 \rho_3y_2^3+ \mathcal{ R}^I - 

10207: 2P r_2y_3+ P^4 \mathcal{ R}^{IV}+ P \mathcal{ R}^I + 

10208: \mathcal{ R}^I + P^3\mathcal{ R}^{IV}+ 

10209: \mathcal{ R}^{II} \\

10210: & \ \ \ \ \

10211: +

10212: \rho_2 r_3 + 4 P^3 \rho_2 y_3^3+

10213: \mathcal{ R}^I +

10214: 2P r_3y_2+ P^4 \mathcal{ R}^{IV}+ P \mathcal{ R}^I + 

10215: \mathcal{ R}^I + P^3\mathcal{ R}^{IV}+ 

10216: \mathcal{ R}^{II} \\

10217: & \

10218: =

10219: \rho_2 r_3 - \rho_3 r_2 + 

10220: 2P r_3 y_2 - 2P r_2 y_3 + 

10221: 4P^3 \rho_2 y_3^3 - 4 P^3 \rho_3 y_2^3 + \\

10222: & \ \ \ \ \

10223: +

10224: \mathcal{ R}^I + \mathcal{ R}^{II}+

10225: P \mathcal{ R}^{II}+

10226: P^3 \mathcal{ R}^{IV}+

10227: P^4 \mathcal{ R}^{IV}.

10228: \endaligned

10229: \end{equation}

10230: In the development, before simplification, we firstly write out in 

10231: lines 3 and 4 all

10232: the $9\times 2$ terms of the two product: for instance, the third term

10233: of the first product, namely 

10234: $4(-\frac{i}{ 2} \rho_3 )(\mathcal{ R}^I)$, yields

10235: a term $\mathcal{ R}^I$ whereas the fifth term $4(-iPy_3)(-2iP^3

10236: y_2^3)$ yields a term $P^4 \mathcal{ R}^{IV}$; secondly, we simplify

10237: the obtained sum: by our convention, $\mathcal{ R}^{I}+ \mathcal{

10238: R}^I=\mathcal{ R}^I$, whereas $\mathcal{ R}^I+ P\mathcal{ R}^I$ cannot

10239: be simplified, since the large constant $P$ will be chosen later.

10240: With these technical explanations at hand, we shall not provide any

10241: intermediate detail for the further computations, whose rules are

10242: totally analogous. For $A_1$, $A_2$ and $A_3$, we obtain

10243: \def\theequation{14.14}\begin{equation}

10244: \left\{

10245: \aligned

10246: A_1 

10247: & \

10248: =

10249: \rho_2 r_3 - \rho_3 r_2 + 

10250: 2P r_3 y_2 - 2P r_2 y_3 + 

10251: 4P^3 \rho_2 y_3^3 - 4 P^3 

10252: \rho_3 y_2^3 + \\

10253: & \ \ \ \ \

10254: +

10255: \mathcal{ R}^I + \mathcal{ R}^{II}+

10256: P \mathcal{ R}^{II}+

10257: P^3 \mathcal{ R}^{IV}+

10258: P^4 \mathcal{ R}^{IV}, \\

10259: A_2

10260: & \

10261: =

10262: \rho_3 r_1 - \rho_1 r_3 + 

10263: 2P r_1 y_3 - 2P r_3 y_1 + 

10264: 4P^3 \rho_3 y_1^3 - 4 P^3 

10265: \rho_1 y_3^3 + \\

10266: & \ \ \ \ \

10267: +

10268: \mathcal{ R}^I + \mathcal{ R}^{II}+

10269: P \mathcal{ R}^{II}+

10270: P^3 \mathcal{ R}^{IV}+

10271: P^4 \mathcal{ R}^{IV}, \\

10272: A_3

10273: & \

10274: =

10275: \rho_1 r_2 - \rho_2 r_1 + 

10276: 2P r_2 y_1 - 2P r_1 y_2 + 

10277: 4P^3 \rho_1 y_2^3 - 4 P^3 

10278: \rho_2 y_1^3 + \\

10279: & \ \ \ \ \

10280: +

10281: \mathcal{ R}^I + \mathcal{ R}^{II}+

10282: P \mathcal{ R}^{II}+

10283: P^3 \mathcal{ R}^{IV}+

10284: P^4 \mathcal{ R}^{IV}.

10285: \endaligned\right.

10286: \end{equation}

10287: 

10288: Now that we have written the complex vector field $\LL$ and its

10289: coefficients $A_1$, $A_2$ and $A_3$, in order to establish Lemma~14.9,

10290: it suffices to choose $P>0$ sufficiently large in order that the four

10291: complex vector fields

10292: \def\theequation{14.15}\begin{equation}

10293: \overline{ \LL}\, \vert_{y=0} , 

10294: \ \ \ \ \

10295: \LL \vert_{y=0}, 

10296: \ \ \ \ \

10297: \left[

10298: \overline{ \LL}, \LL

10299: \right]\, \vert_{y=0}, 

10300: \ \ \ \ \

10301: \left[

10302: \overline{\LL}, 

10303: \left[

10304: \overline{\LL}, 

10305: \left[

10306: \overline{ \LL}, \LL

10307: \right]\right]\right]

10308: \vert_{y=0}

10309: \end{equation}

10310: are linearly independent at every point $x \in \R^3$ with $x_1^2 +

10311: x_2^2 + x_3^2 \leq R^2$. At the end of the proof, we shall explain why

10312: we cannot insure type $3$ at every point, namely why the consideration

10313: of $\left[ \overline{\LL}, \left[ \overline{ \LL}, \LL

10314: \right]\right]\vert_{y=0}$ instead of the length four last Lie bracket

10315: in~\thetag{ 14.15} would fail.

10316: 

10317: As promised, we shall now summarize all the subsequent

10318: computations. As we aim to restrict the last Lie bracket to $\{y=0\}$

10319: which is of length four and whose coefficients involve derivatives of

10320: order at most three of the coefficients $A_1$, $A_2$ and $A_3$, we can

10321: already neglect the last two remainders $P^3\mathcal{ R}^{IV}$ and

10322: $P^4 \mathcal{ R}^{IV}$ in~\thetag{ 14.14}. In other words, we can

10323: consider $A^1$, $A^2$ and $A^3 \ {\rm mod} (IV)$. Similarly, in the

10324: computation of the Lie bracket

10325: \def\theequation{14.16}\begin{equation}

10326: \left[\overline{\LL},

10327: \LL\right]=:

10328: C_1 \, \partial_{z_1}+

10329: C_2 \, \partial_{z_2}+

10330: C_3 \, \partial_{z_3}- 

10331: \overline{C_1} \, \partial_{\bar z_1}- 

10332: \overline{C_2}\, \partial_{\bar z_2} - 

10333: \overline{C_3} \, \partial_{\bar z_3},

10334: \end{equation}

10335: before restriction to $\{y=0\}$, we can restrict our task to

10336: developing the coefficients

10337: \def\theequation{14.17}\begin{equation}

10338: \aligned

10339: C_1 

10340: & \

10341: :=

10342: \overline{A_1} A_{1,\bar z_1}+

10343: \overline{A_2} A_{1,\bar z_2}+ 

10344: \overline{A_3} A_{1,\bar z_3}, \\

10345: C_2

10346: & \

10347: :=

10348: \overline{A_1} A_{2,\bar z_1}+

10349: \overline{A_2} A_{2,\bar z_2}+ 

10350: \overline{A_3} A_{2,\bar z_3}, \\

10351: C_3

10352: & \

10353: :=

10354: \overline{A_1} A_{3,\bar z_1}+

10355: \overline{A_2} A_{3,\bar z_2}+ 

10356: \overline{A_3} A_{3,\bar z_3} \\

10357: \endaligned

10358: \end{equation}

10359: only modulo order $(III)$, which yields by means of

10360: the expressions~\thetag{ 14.14}

10361: \def\theequation{14.18}\begin{equation}

10362: \aligned

10363: C_1 \ {\rm mod} \, 

10364: (III) 

10365: & \

10366: \equiv 

10367: -iP \rho_1 +

10368: 6i P^3 a_3 \rho_2 y_3^2 -

10369: 6i P^3 a_2 \rho_3 y_2^2 +

10370: \mathcal{ R}^0 +

10371: \mathcal{ R}^I+ \\

10372: & \ \ \ \ \ \ 

10373: +

10374: P\mathcal{ R}^I+

10375: P^2 \mathcal{ R}^I

10376: +\mathcal{ R}^{II}+

10377: P\mathcal{ R}^{II} +

10378: P^2 \mathcal{ R}^{II}, \\ 

10379: C_2 \ {\rm mod} \, 

10380: (III) 

10381: & \

10382: \equiv 

10383: -iP \rho_2 +

10384: 6i P^3 a_1 \rho_3 y_1^2 -

10385: 6i P^3 a_3 \rho_1 y_3^2 +

10386: \mathcal{ R}^0 +

10387: \mathcal{ R}^I+ \\

10388: & \ \ \ \ \ \ 

10389: +

10390: P\mathcal{ R}^I+

10391: P^2 \mathcal{ R}^I

10392: +\mathcal{ R}^{II}+

10393: P\mathcal{ R}^{II} +

10394: P^2 \mathcal{ R}^{II}, \\

10395: C_3 \ {\rm mod} \, 

10396: (III) 

10397: & \

10398: \equiv 

10399: -iP \rho_3 +

10400: 6i P^3 a_2 \rho_1 y_2^2 -

10401: 6i P^3 a_1 \rho_2 y_1^2 +

10402: \mathcal{ R}^0 +

10403: \mathcal{ R}^I+ \\

10404: & \ \ \ \ \ \ 

10405: +

10406: P\mathcal{ R}^I+

10407: P^2 \mathcal{ R}^I

10408: +\mathcal{ R}^{II}+

10409: P\mathcal{ R}^{II} +

10410: P^2 \mathcal{ R}^{II}. 

10411: \endaligned

10412: \end{equation} 

10413: We must mention the use of natural rule hold for computing the partial

10414: derivatives $A_{j, \bar z_k}$: we have for instance $\partial_{\bar

10415: z_k} \left( \mathcal{ R}^{II} \right)= \mathcal{ R}^I+ \mathcal{

10416: R}^{II}$. Also, we have used the hypothesis that $(L(x), K^1(x),

10417: K^2(x))$ provides a direct orthonormal frame at every $x\in \R^3$,

10418: which yields in particular the three relations

10419: \def\theequation{14.19}\begin{equation}

10420: a_2 r_3 - a_3 r_2=-\rho_1, \ \ \ \ \ 

10421: a_3 r_1-a_1 r_3= -\rho_2, \ \ \ \ \ 

10422: a_1r_2-a_2r_1= -\rho_3.

10423: \end{equation}

10424: After mild computation, the coefficients 

10425: $F_1$, $F_2$ and $F_3$ of the length four Lie bracket

10426: \def\theequation{14.20}\begin{equation}

10427: \left[

10428: \overline{\LL}, 

10429: \left[

10430: \overline{\LL}, 

10431: \left[

10432: \overline{\LL}, \LL

10433: \right]

10434: \right]

10435: \right]=

10436: F_1 \, \partial_{z_1}+

10437: F_2 \, \partial_{z_2}+

10438: F_3 \, \partial_{z_3}+

10439: G_1 \, \partial_{\bar z_1}+

10440: G_2 \, \partial_{\bar z_2}+

10441: G_3 \, \partial_{\bar z_3}

10442: \end{equation}

10443: are given, after restriction to $\{y=0\}$, by 

10444: \def\theequation{14.21}\begin{equation}

10445: \aligned

10446: F_1\vert_{y=0} 

10447: & \

10448: =

10449: 3i P^3 a_2^3\rho_3 -3i P^3 a_3^3 \rho_2 +

10450: \mathcal{ R}^0+ 

10451: P\mathcal{R}^0 +

10452: P^2\mathcal{ R}^0, \\

10453: F_2\vert_{y=0} 

10454: & \

10455: =

10456: 3i P^3 a_3^3\rho_1 -3i P^3 a_1^3 \rho_3 +

10457: \mathcal{ R}^0+ 

10458: P\mathcal{R}^0 +

10459: P^2\mathcal{ R}^0, \\

10460: F_3\vert_{y=0} 

10461: & \

10462: =

10463: 3i P^3 a_1^3\rho_2 -3i P^3 a_2^3 \rho_1 +

10464: \mathcal{ R}^0+ 

10465: P\mathcal{R}^0 +

10466: P^2\mathcal{ R}^0, \\

10467: \endaligned

10468: \end{equation}

10469: 

10470: We can now complete the proof of Lemma~14.9. In the basis

10471: $(\partial_{z_1}, \partial_{z_2}, \partial_{z_3}, \partial_{\bar z_1},

10472: \partial_{\bar z_2}, \partial_{\bar z_3})$, the $4\times 6$ matrix

10473: associated with the four vector fields~\thetag{ 14.15} (without

10474: mentioning $\vert_{y=0}$)

10475: \def\theequation{14.22}\begin{equation}

10476: \left(

10477: \begin{array}{cccccc}

10478: 0 & 0 & 0 & a_1 & a_2 & a_3 \\

10479: a_1 & a_2 & a_3 & 0 & 0 & 0 \\

10480: C_1 & C_2 & C_3 & -\overline{C_1} 

10481: & - \overline{C_2} & -\overline{C_3} \\

10482: F_1 & F_2 & F_3 & G_1 & G_2 & G_3

10483: \end{array}

10484: \right)

10485: \end{equation}

10486: has rank four at a point $x\in\R^3$ if and only if the $3\times 3$

10487: determinant in the left low corner is nonvanishing, namely if and only

10488: if the developped expression

10489: \def\theequation{14.23}\begin{equation}

10490: \aligned

10491: {}

10492: & \

10493: \left\vert

10494: \begin{array}{ccc}

10495: a_1 & a_2 & a_3 \\

10496: -iP\rho_1+\mathcal{ R}^0 &

10497: -iP\rho_2+\mathcal{ R}^0 &

10498: -iP\rho_3+\mathcal{ R}^0 \\

10499: 3iP^3a_2^3\rho_3-3iP^3a_3^3\rho_2+ &

10500: 3iP^3a_3^3\rho_1-3iP^3a_1^3\rho_3+ &

10501: 3iP^3a_1^3\rho_2-3iP^3a_2^3\rho_1+ \\ 

10502: +\mathcal{ R}^0+P\mathcal{ R}^0+P^2\mathcal{ R}^0 &

10503: +\mathcal{ R}^0+P\mathcal{ R}^0+P^2\mathcal{ R}^0 &

10504: +\mathcal{ R}^0+P\mathcal{ R}^0+P^2\mathcal{ R}^0 

10505: \end{array}

10506: \right\vert \\

10507: & \

10508: =

10509: 3P^4 \left(

10510: r_3 [a_1^3 \rho_2 -a_2^3 \rho_1]+

10511: r_2[a_3^3\rho_1 -a_1^3 \rho_3]+

10512: r_1 [a_2^3 \rho_3 -a_3^3 \rho_2]

10513: \right)+ \\

10514: & \ \ \ \ \ \

10515: +

10516: \mathcal{ R}^0+

10517: P\mathcal{ R}^0+

10518: P^2 \mathcal{ R}^0+

10519: P^3\mathcal{ R}^0 +

10520: P^4\mathcal{ R}^0 \\

10521: & \

10522: =

10523: 3P^4 \left(a_1^4+a_2^4+a_3^4\right)+

10524: \mathcal{ R}^0+

10525: P\mathcal{ R}^0+

10526: P^2 \mathcal{ R}^0+

10527: P^3\mathcal{ R}^0 +

10528: P^4\mathcal{ R}^0

10529: \endaligned

10530: \end{equation}

10531: is nonvanishing. 

10532: 

10533: At this point, the conclusion of the lemma is now an immediate

10534: consequence of the following trivial assertion: {\it Let $a_1$, $a_2$

10535: and $a_3$ be $\mathcal{ C }^\infty$-smooth functions on $\R^3$

10536: satisfying $a_1 (x)^2+ a_2 (x)^2+a_3(x)^2=1$ for all $x \in \R^3$ and

10537: let $\mathcal{ R }_0^0$, $\mathcal{ R}_1^0$, $\mathcal{ R }_2^0$,

10538: $\mathcal{ R }_3^0$ and $\mathcal{ R}_4^0$ be $\mathcal{ C

10539: }^\infty$-smooth functions on $\R^3$. For every $R>0$, there exists a

10540: constant $P>0$ large enough so that the function

10541: \def\theequation{14.24}\begin{equation}

10542: 3P^4 \left(

10543: a_1^4+a_2^4+a_3^4\right)+

10544: \mathcal{ R}_0^0+P\mathcal{ R}_1^0+

10545: P^2\mathcal{ R}_2^0 +

10546: P^3 \mathcal{ R}_3^0+

10547: P^4\mathcal{ R}_4^0 

10548: \end{equation}

10549: is positive at every $x\in \R^3$ with $x_1^2 + x_2^2+ x_3^2 \leq R^2$}.

10550: 

10551: If we had put $y_1^3+ y_2^3+ y_3^3$ instead of $y_1^4+ y_2^4+ y_3^4$

10552: in the second equation~\thetag{14.8}, we would have considered the

10553: length three Lie bracket $\left[ \overline{\LL}, \left[

10554: \overline{\LL}, \LL \right] \right] \vert_{y=0}$ instead of the length

10555: four Lie bracket in~\thetag{14.15}, and hence instead of the quartic

10556: $a_1^4+ a_2^4+ a_3^4$ in~\thetag{ 14.24}, we would have obtained the

10557: cubic $a_1^3+a_2^3+a_3^3$, a function which (unfortunately) vanishes,

10558: for instance if $a_1(x) =\frac{ 1}{\sqrt{ 2}}$, $a_2(x)= -\frac{ 1}{

10559: \sqrt{ 2}}$ and $a_3(x) =0$. We notice that in our example, this value

10560: of $(a_1,a_2,a_3)$ is indeed attained at the point $x \in T^2$ of

10561: coordinates $(\frac{ 3}{\sqrt{2}}, -\frac{ 3}{ \sqrt{ 2}}, 0)$, whence

10562: the necessity of passing to type 4.

10563: The proof of Lemma~14.9 is complete.

10564: \endproof

10565: 

10566: \begin{thebibliography}{XL}

10567: 

10568: \bibitem[A]{a}

10569: {\sc Ayrapetian}, R.A.: 

10570: {\em Extending CR functions from piecewise smooth CR manifolds}. 

10571: Mat. Sbornik {\bf 134} (1987), 108--118. Trad. in english in Math. 

10572: {\sc Ussr} Sbornik {\bf 62} (1989), 1, 111--120.

10573: 

10574: \bibitem[BT]{bt} 

10575: {\sc Baouendi}, M.S.;

10576: {\sc Treves}, F: 

10577: {\em A property of the functions and distributions annihilated by a

10578: locally integrable system of complex vector fields},

10579: Ann. of Math.

10580: {\bf 113} (1981), no.2, 387--421.

10581: 

10582: \bibitem[BER]{ber} 

10583: {\sc Baouendi}, M.S.;

10584: {\sc Ebenfelt}, P.; {\sc Rothschild}, L.P.: 

10585: {\em Real submanifolds in complex space and their mappings}.

10586: Princeton Mathematical Series, {\bf 47}, Princeton University Press,

10587: Princeton, NJ, 1999, xii+404 pp.

10588: 

10589: \bibitem[BK]{bk}

10590: {\sc Bedford}, E.; {\sc Klingenberg}, W.:

10591: {\em On the envelope of holomorphy of a $2$-sphere in $\C^2$}, 

10592: J. Amer. Math. Soc. {\bf 4} (1991), 623--646.

10593: 

10594: \bibitem[Be]{be}

10595: {\sc Bennequin}, D.:

10596: {\em Entrelacements et \'equations de Pfaff}, Ast\'erisque 107/108, 

10597: 87--161, 1983.

10598: 

10599: \bibitem[B]{b}

10600: {\sc Bishop}, E.:

10601: {\em Differentiable manifolds in complex Euclidean space}, 

10602: Duke Math. J. {\bf 32} (1965), 1--22.

10603: 

10604: \bibitem[Bo]{bo} 

10605: {\sc Boggess, A.}:

10606: {\em CR manifolds and the tangential Cauchy-Riemann complex}. 

10607: Studies in Advanced Mathematics. CRC Press,

10608: Boca Raton, FL, 1991, xviii+364 pp. 

10609: 

10610: \bibitem[CLN]{cln}

10611: {\sc Camacho}, C.; {\sc Lins Neto}, A.:

10612: {\em Geometric theory of foliations}, 

10613: Birkh\"auser, Boston, 1985.

10614: 

10615: \bibitem[CS]{cs} 

10616: {\sc Chirka}, E.M.; {\sc Stout}, E.L.:

10617: {\em Removable singularities in the boundary}. 

10618: Contributions to complex analysis and analytic geometry, 43--104,

10619: Aspects Math., E26, Vieweg, Braunschweig, 1994.

10620: 

10621: \bibitem[DS]{ds} 

10622: {\sc Dinh, T.C.}; {\sc Sarkis, F.}: {\it Wedge removability of

10623: metrically thin sets and application to the CR meromorphic

10624: extension}. Math. Z. {\bf 238} (2001), no.3, 639--653.

10625: 

10626: \bibitem[D]{d}

10627: {\sc Duval}, J.: 

10628: {\em Surfaces convexes dans un bord pseudoconvexe}, 

10629: Colloque d'Analyse Complexe et G\'eom\'etrie, Marseille, 1992~;

10630: Ast\'erisque {\bf 217}, Soc. Math. France, 

10631: Montrouge, 1993, 6, 103--118.

10632: 

10633: \bibitem[E]{e}

10634: {\sc Eliashberg}, Y.:

10635: {\em Classification of overtwisted contact structures

10636: on 3-manifolds}, 

10637: Invent. Math. {\bf 98} (1989), no.~3, 623--637.

10638: 

10639: \bibitem[FM]{fm}

10640: {\sc Forn{\ae}ss}, J.E.; {\sc Ma}, D.:

10641: {\em A $2$-sphere in $\C^2$ that cannot be filled

10642: in with analytic discs}, International Math. Res. 

10643: Notices, 1995, no.~1, 17--22.

10644: 

10645: \bibitem[FS]{fs}

10646: {\sc Forstneri$\check{\text{\sc c}}$}, F.; {\sc Stout}, E.L.:

10647: {\em A new class of polynomially convex sets}, Ark. Mat. {\bf 29}

10648: (1991), 52--62.

10649: 

10650: \bibitem[HT]{ht} 

10651: {\sc Hanges}, N; {\sc Treves}, F.:

10652: {\em Propagation of holomorphic extendability of CR functions},

10653: Math. Ann. {\bf 263} (1983), 157--177.

10654: 

10655: \bibitem[Ha]{ha}

10656: {\sc Hartman, P.}:

10657: {\em Ordinary Differential Equations}.

10658: Birkh\"auser, Boston 1982.

10659: 

10660: \bibitem[HL]{hl} 

10661: {\sc Harvey, R.}; {\sc Lawson, B.}:

10662: {\em On boundaries of complex analytic varieties}. Ann. of Math., I:

10663: {\bf 102} (1975), no.2, 233--290; II: {\bf 106} (1977), no.2, 213--238.

10664: 

10665: \bibitem[I]{i} 

10666: {\sc Ivashkovitch}, S.M.; 

10667: {\em The Hartogs-type extension theorem for meromorphic maps 

10668: into compact K\"ahler manifolds}, 

10669: Invent. Math., {\bf 109} (1992), no.1, 47--54.

10670: 

10671: \bibitem[J1]{j1}

10672: {\sc J\"oricke}, B.:

10673: {\em Removable singularities of CR-functions}, 

10674: Ark. Mat. {\bf 26} (1988), 117--143.

10675: 

10676: \bibitem[J2]{j2}

10677: {\sc J\"oricke, B.}: 

10678: {\em Deformation of CR-manifolds, minimal points and 

10679: CR-manifolds with the microlocal analytic extension property},

10680: J. Geom. Anal. {\bf 6} (1996), no.4, 555--611.

10681: 

10682: \bibitem[J3]{j3}

10683: {\sc J\"oricke}, B.: 

10684: {\em Local polynomial hulls of discs near isolated parabolic points},

10685: Indiana Univ. Math. J. {\bf 46} (1997), no.3, 789--826. 

10686: 

10687: \bibitem[J4]{j4}

10688: {\sc J\"oricke}, B.: 

10689: {\em Boundaries of singularity sets, removable

10690: singularities, and CR-invariant subsets of CR-manifolds},

10691: J. Geom. Anal. {\bf 9} (1999), no.2, 257--300.

10692: 

10693: \bibitem[J5]{j5}

10694: {\sc J\"oricke, B.}: 

10695: {\em Removable singularities of $L^{\sf p}$ 

10696: CR functions on hypersurfaces},

10697: J. Geom. Anal. {\bf 9} (1999), no.3, 429--456.

10698: 

10699: \bibitem[JS]{js}

10700: {\sc J\"oricke}, B.; {\sc Shcherbina}, N.:

10701: {\em A nonremovable generic 4-ball in the unit sphere of $\C^3$},

10702: Duke Math. J. {\bf 102} (2000), no.1, 87--100.

10703: 

10704: \bibitem[JP]{jp}

10705: {\sc J\"oricke, B.}; {\sc Porten}, E.:

10706: {\em Hulls and analytic extension from nonpseudoconvex boundaries}, 

10707: U.U.M.D. Report 2002:19, Uppsala University, 38~pp., 2002.

10708: 

10709: \bibitem[KR]{kr}

10710: {\sc Kytmanov}, A.M.; {\sc Rea}, C.:

10711: {\em Elimination of $L^1$ singularities on H\"older peak 

10712: sets for CR functions}, Ann. Scuola Norm. Sup. Pisa, Classe 

10713: di Scienze, {\bf 22} (1995), 211--226.

10714: 

10715: \bibitem[L]{l}

10716: {\sc Laurent-Thi\'ebaut}, C.:

10717: {\em Sur l'extension des fonctions CR dans une vari\'et\'e

10718: de Stein}, Ann. Mat. Pura Appl. {\bf 150} (1988), 141--151.

10719: 

10720: \bibitem[LP]{lp}

10721: {\sc Laurent-Thi\'ebaut}, C.; {\sc Porten}, E.:

10722: {\em Analytic extension from nonpseudoconvex boundaries and 

10723: $A(D)$-convexity}, Ann. Inst. Fourier (Grenoble), 2003, to appear.

10724: 

10725: \bibitem[Lu]{lu}

10726: {\sc Lupacciolu}, G.:

10727: {\em Characterization of removable sets in strongly pseudoconvex

10728: boundaries}, Ark. Mat. {\bf 32} (1994), 455--473.

10729: 

10730: \bibitem[M1]{m1} 

10731: {\sc Merker}, J.:

10732: {\em Global minimality of generic manifolds and holomorphic

10733: extendibility of CR functions}. Internat. Math. Res. Notices 1994,

10734: no.8, 329--342.

10735: 

10736: \bibitem[M2]{m2} 

10737: {\sc Merker}, J.:

10738: {\em On removable singularities for CR functions 

10739: in higher codimension}, 

10740: Internat. Math. Res. Notices 1997, no.1, 21--56.

10741: 

10742: \bibitem[MP1]{mp1}

10743: {\sc Merker, J.}; {\sc Porten, P.}:

10744: {\em On removable singularities for integrable CR functions},

10745: Indiana Univ. Math. J. {\bf 48} (1999), no.3, 805--856. 

10746: 

10747: \bibitem[MP2]{mp2}

10748: {\sc Merker, J.}; {\sc Porten, P.}:

10749: {\em On the local meromorphic extension of CR meromorphic functions}.

10750: Complex analysis and applications (Warsaw,

10751: 1997). Ann. Polon. Math. {\bf 70} (1998), 163--193.

10752: 

10753: \bibitem[MP3]{mp3}

10754: {\sc Merker}, J; {\sc Porten}, E.:

10755: {\em On wedge extendability of CR meromorphic functions}, 

10756: Math. Z. {\bf 241} (2002) 485--512.

10757: 

10758: \bibitem[P]{p}

10759: {\sc Pinchuk}, S.: 

10760: {\em A boundary uniqueness theorem for holomorphic functions of

10761: several complex variables}, 

10762: Mat. Zametki {\bf 15} (1974), 205--212.

10763: 

10764: \bibitem[P1]{p1}

10765: {\sc Porten}, P.:

10766: {\em Analytic extension and removable singularities of the integrable

10767: CR-functions}, 

10768: Preprint, Humboldt-Universit\"at zu Berlin, 2000, no.~3, 18~pp.

10769: 

10770: \bibitem[P2]{p2}

10771: {\sc Porten}, P.:

10772: {\em Totally real discs in nonpseudoconvex boundaries}, 

10773: Ark. Mat. {\bf 41} (2003), no.1, 133--150.

10774: 

10775: \bibitem[P3]{p3}

10776: {\sc Porten}, P.:

10777: {\em Habilitationsschrift} (in preparation).

10778: 

10779: \bibitem[Sa]{sa} 

10780: {\sc Sarkis, F.}:

10781: {\em CR-meromorphic extension and the non embeddability of the

10782: Andreotti-Rossi CR structure in the projective space},

10783: Int. J. Math. {\bf 10} (1999), no.7, 897--915.

10784: 

10785: \bibitem[Sl]{sl}

10786: {\sc Slapar}, M:

10787: {\em On Stein neighborhood basis of real surfaces},

10788: e-print {\tt arXiv:math.CV/0309218}, september 2003.

10789: 

10790: \bibitem[St]{st}

10791: {\sc Stout}, E.L.:

10792: {\em Removable singularities for the boundary values of holomorphic

10793: functions}. Several Complex Variables (Stockholm, 1987/1988),

10794: 600--629, {\it Math. Notes}, {\bf 38}, Princeton Univ. Press,

10795: Princeton, NJ, 1993.

10796: 

10797: \bibitem[Su]{su} 

10798: {\sc Sussmann}, H.J.: 

10799: {\em Orbits of families of vector fields and integrability of 

10800: distributions},

10801: Trans. Amer. Math. Soc. {\bf 180} (1973), 171--188.

10802: 

10803: \bibitem[Tr1]{tr1}

10804: {\sc Tr\'epreau}, J.-M.: 

10805: {\em Sur le prolongement holomorphe des fonctions CR d\'efinies sur une

10806: hypersurface r\'eelle de classe ${\bf C}^{2}$}, Invent. Math.

10807: {\bf 83} (1986), 583--592.

10808: 

10809: \bibitem[Tr2]{tr2}

10810: {\sc Tr\'epreau}, J.-M.:

10811: {\em Sur la propagation des singularit\'es dans les vari\'et\'es CR},

10812: Bull. Soc. Math. Fr. {\bf 118} (1990), no.4, 403--450.

10813: 

10814: \bibitem[Trv]{trv} 

10815: {\sc Treves}, F.: 

10816: {\em Approximation and

10817: representation of functions and distributions annihilated by a system

10818: of complex vector fields},

10819: Palaiseau, \'Ecole

10820: Polytechnique, Centre de Math\'ematiques, 1981.

10821: 

10822: \bibitem[Tu1]{tu1}

10823: {\sc Tumanov}, A.E.: 

10824: {\em Extending CR-functions into a wedge},

10825: Mat. Sbornik {\bf 181} (1990), 951--964. Trad. in English in Math.

10826: {\sc Ussr} Sbornik {\bf 70} (1991), 2, 385--398.

10827: 

10828: \bibitem[Tu2]{tu2}

10829: {\sc Tumanov}, A.E.:

10830: {\em Connections and propagation of analyticity for CR functions},

10831: Duke Math. J. {\bf 73} (1994), no.1, 1--24.

10832: 

10833: \bibitem[Tu3]{tu3}

10834: {\sc Tumanov, A.E.}:

10835: {\em On the propagation of extendibility of CR functions}. Complex

10836: analysis and geometry (Trento, 1993), 479--498, Lecture Notes in Pure

10837: and Appl. Math., {\bf 173},

10838: Dekker, New York, 1996.

10839: 

10840: \end{thebibliography}

10841: 

10842: \vfill

10843: \end{document}

10844: 

10845: \endproof

10846: 

10847: \endproof