1: \documentclass[12pt,reqno]{amsart}

2: \usepackage{amssymb}

3: 

4: %\input iso.def

5: \input epsf.tex

6: \newdimen\xsize

7: \newdimen\oldbaselineskip

8: \newdimen\oldlineskiplimit

9: \xsize=.7\hsize

10: \def\nolineskip{\oldbaselineskip=\baselineskip\baselineskip=0pt%

11: \oldlineskiplimit=\lineskiplimit\lineskiplimit=0pt}

12: \def\restorelineskip{\baselineskip=\oldbaselineskip%

13: \lineskiplimit=\oldlineskiplimit}

14: %

15: \def\putm[#1][#2]#3{% #1 X-axis  #2 Y-axis(downwords)  #3 $"mathobject"$

16: \hbox{\vbox to 0pt{\parindent=0pt%

17: \vskip#2\xsize\hbox to0pt{\hskip#1\xsize $#3$\hss}\vss}}}%

18: %\enddef\putm

19: %

20: %\def\putm[#1][#2]#3{% #1 X-axis  #2 Y-axis(downwords)  #3 $"mathobject"$

21: %\vbox to 0pt{\noindent\hskip#1\xsize\lower#2\xsize\hbox to0pt{$#3$\hss}\vss}}%

22: %\enddef\putm

23: %

24: \def\putt[#1][#2]#3{% #1 X-axis  #2 Y-axis(downwords)  #3 "text"

25: \vbox to 0pt{\noindent\hskip#1\xsize\lower#2\xsize%

26: \vtop{\restorelineskip#3}\vss}}

27: %\enddef\putt

28: 

29: 

30: %

31: \font\Bigbf=cmbx12 scaled 1200

32: \font\bigbf=cmbx12

33: %

34: \thinmuskip = 2mu

35: \medmuskip = 2.5mu plus 1.5mu minus 2.1mu  %minus 2.5mu

36: \thickmuskip = 4mu plus 6mu

37: %

38: %\font\teneusm=eusm10

39: %\font\seveneusm=eusm7

40: %\font\fiveeusm=eusm5

41: %\newfam\eusmfam

42: %\textfont\eusmfam=\teneusm

43: %\scriptfont\eusmfam=\seveneusm

44: %\scriptscriptfont\eusmfam=\fiveeusm

45: %\def\cal#1{{\fam\eusmfam\relax#1}}

46: %

47: %\def\cal#1{{\mathcal{#1}}}

48: 

49: 

50: 

51: %\font\tenscr=rsfs10 % scaled \magstep1

52: %\font\sevenscr=rsfs7 % scaled \magstep1

53: %\font\fivescr=rsfs5 % scaled \magstep1

54: %\skewchar\tenscr='177 \skewchar\sevenscr='177 \skewchar\fivescr='177

55: %\newfam\scrfam 

56: %\textfont\scrfam=\tenscr 

57: %\scriptfont\scrfam=\sevenscr

58: %\scriptscriptfont\scrfam=\fivescr

59: %\def\scr#1{{\fam\scrfam \tenscr{#1}}}

60: \DeclareFontFamily{U}{rsf}{\skewchar\font'177}%

61: \DeclareFontShape{U}{rsf}{m}{n}{<-6>rsfs5<6-8>rsfs7<8->rsfs10}{}%

62: \DeclareFontShape{U}{rsf}{b}{n}{<-6>rsfs5<6-8>rsfs7<8->rsfs10}{}%

63: \DeclareMathAlphabet\RSFS{U}{rsf}{m}{n}

64: \SetMathAlphabet\RSFS{bold}{U}{rsf}{b}{n}

65: \DeclareOption{psamsfonts}{%

66:   \DeclareFontFamily{U}{rsf}{}%

67:   \DeclareFontShape{U}{rsf}{m}{n}{<-6>rsfs5<6-8>rsfs7<8->rsfs10}{}%

68:   \DeclareFontShape{U}{rsf}{b}{n}{<-6>rsfs5<6-8>rsfs7<8->rsfs10}{}%

69: }

70: \ProcessOptions

71: 

72: \def\rfs#1{{\RSFS{#1}}}

73: \let\scr=\rfs

74: %\let\scr=\mathcal 

75: %\def\scr#1{{\mathcal{#1}}}

76: 

77: 

78: 

79: %

80: %\font\tenmib=cmmib10

81: %\font\sevenmib=cmmib7

82: %\font\fivemib=cmmib5

83: %\newfam\mibfam

84: %\textfont\mibfam=\tenmib

85: %\scriptfont\mibfam=\sevenmib

86: %\scriptscriptfont\mibfam=\fivemib

87: %\def\mib{\fam\mibfam}

88: \def\mib#1{\boldsymbol{#1}}

89: 

90: %\font\tensl=cmsl10

91: %\font\sevensl=cmsl7 %cmsl10 scaled 694

92: %\font\fivesl=cmsl5 %cmsl10 scaled 482

93: %

94: %\font\sevensl=cmsl10 scaled 694

95: %\font\fivesl=cmsl10 scaled 482

96: %\newfam\slfam

97: %\textfont\slfam=\tensl

98: %\scriptfont\slfam=\sevensl

99: %\scriptscriptfont\slfam=\fivesl

100: %\def\sl{\fam\slfam\tensl}

101: 

102: 

103: %

104: %\font\tensf=cmss10

105: %\font\sevensf=cmss8 scaled 833

106: %\font\fivesf=cmss8 scaled 578

107: %\newfam\sffam

108: %\textfont\sffam=\tensf

109: %\scriptfont\sffam=\sevensf

110: %\scriptscriptfont\sffam=\fivesf

111: \def\sf#1{{\mathsf{#1}}}%\def\sf{\fam\sffam}

112: 

113: %

114: %\font\tenfr=eufm10 scaled 1200

115: %\font\sevenfr=eufm7 scaled 1200 

116: %\font\fivefr=eufm5 scaled 1200

117: %\newfam\frfam

118: %\textfont\frfam=\tenfr

119: %\scriptfont\frfam=\sevenfr

120: %\scriptscriptfont\frfam=\fivefr

121: %\def\fr#1{{\mathfrak{#1}}

122: 

123: 

124: 

125: %

126: %\font\mathnine=cmmi9

127: %\font\rmnine=cmr9

128: %\font\cmsynine=cmsy9

129: %\font\cmexnine=cmex9 %cmex10 scaled 913

130: %\font\ninesl=cmsl9

131: %\font\ninesf=cmss9

132: %\font\ninemib=cmmib9

133: \def\msmall#1{\mathchoice{\hbox{\small$\displaystyle {#1}$}}{#1}{#1}{#1}}        

134: %\def\msmall#1{\hbox{$\displaystyle

135: %\textfont0=\rmnine   \textfont1=\mathnine

136: %\textfont2=\cmsynine \textfont3=\cmexnine

137: %\textfont\slfam=\ninesl \font\tensf=cmss9

138: %\font\tenmib=cmmib9

139: %{#1}$}}

140: 

141: 

142: %\font\mathnine=cmmi9

143: %\font\rmnine=cmr9

144: %\font\cmsynine=cmsy9

145: %\font\cmexnine=cmex10 scaled 905

146: %\def\msmall#1{\hbox{$\displaystyle \textfont0=\rmnine

147: %\textfont1=\mathnine \textfont2=\cmsynine \textfont3=\cmexnine

148: %{#1}$}}

149: %

150: \hyphenation{Lip-schit-zian Lip-schitz com-pact-ness}

151: %

152: \let\3=\ss

153: \def\cc{{\mathbb C}}

154: \def\jj{{\mathbb J}}

155: \def\rr{{\mathbb R}}

156: \def\nn{{\mathbb N}}

157: \def\pp{{\mathbb P}}

158:   \def\cp{\cc\pp}

159: \def\zz{{\mathbb Z}}

160: \def\sss{{\mathbb S}}

161: \def\ttt{{\mathbb T}}

162: \def\bfg{{\mathbf G}}

163: \def\cinf{C^\infty}

164: %

165: \def\arg{{\sf{Arg}}}

166: \def\aut{{\sf{Aut}}}

167: \def\loc{{\sf{loc}}}

168: \def\st{_{\mathsf{st}}}

169: %\def\st{\mathchoice{_\sf{st}}{_\sf{st}}%

170: %{_{\hbox{\fivesf st}}}{_{\hbox{\fivesf st}}}}

171: \def\area{\sf{area}}

172: \def\sfb{\sf{b}}

173: \def\br{\sf{Br}}

174: \def\cosh{\sf{cosh}}

175: \def\cos{\sf{cos}}

176: \def\coker{\sf{Coker}\,}

177: \def\codim{\sf{codim}\,}

178: \def\deg{\sf{deg}\,}

179: \def\diam{\sf{diam}\,}

180: \def\dim{\sf{dim}\,}

181: \def\dimc{\dim_\cc}

182: \def\dimr{\dim_\rr}

183: \def\dfrm{\sf{dfrm}}

184: \def\dist{\sf{dist}\,}

185: \def\endo{\sf{End}}

186: \def\endr{\sf{End}\vph_\rr}

187: \def\endrn{\endr(\cc^n)}

188: \def\endc{\sf{End}\vph_\cc}

189: \def\ev{\sf{ev}}

190: \def\gh{\sf{gh}}

191: \def\ind{\sf{ind}}

192: \def\sym{\sf{Sym}}

193: \def\sind{\sf{S\hbox{-}ind}\,}

194: \def\exp{\sf{exp}}

195: \def\gcd{\sf{gcd}}

196: \def\sfh{\sf{H}}

197: \def\hom{\sf{Hom}\,}

198: \def\homr{\sf{Hom}\vph_\rr}

199: \def\id{\sf{Id}}

200: \def\im{\sf{Im}\,}

201: \def\bint{{-}\mathchoice{\mkern-18mu}{\mkern-15mu}{\mkern-15mu}{\mkern-15mu}\int}

202: \def\sfl{\sf{L}}

203: \def\re{\sf{Re}\,}

204: \def\ker{\sf{Ker}\,}

205: \def\lim{\mathop{\sf{lim}}}

206: \def\limsup{\mathop{\sf{lim\,sup}}}

207: \def\log{\sf{log}\,}

208: \def\max{\sf{max}}

209: \def\min{\sf{min}}

210: \def\nod{{\sf{nod}}}

211: \def\ord{\sf{ord}}

212: \def\osc{\sf{osc}}

213: \def\op{\sf{op}}

214: \def\pr{\sf{pr}}

215: \def\pt{\sf{pt}}

216: \def\rank{\sf{rank}}

217: \def\res{\sf{Res}}

218: \def\sinh{\sf{sinh}}

219: \def\sin{\sf{sin}}

220: %\def\spec{\sf{Sp}}

221: \def\supp{\sf{supp}\,}

222: \def\sup{\sf{sup}\,}

223: \def\frs{{\frak S}}

224: \def\vol{\sf{Vol}}

225: \def\sfz{\sf{Z}}

226: 

227: 

228: \def\bfone{\boldsymbol{1}}

229: \def\mbfe{{\mib{e}}}

230: \def\mbfk{{\mib{k}}}

231: \def\mbfl{{\mib{l}}}

232: \def\mbfp{{\mib{p}}}

233: \def\mbfs{{\mib{s}}}

234: \def\mbfu{{\mib{u}}}

235: \def\mbfv{{\mib{v}}}

236: \def\mbfw{{\mib{w}}}

237: \def\mbfx{{\mib{x}}}

238: \def\mbfz{{\mib{z}}}

239: 

240: \def\vpp{\,\,\vec{\vphantom{p}}\!\!p'{}}

241: \def\vdp{\,\,\vec{\vphantom{d}}\!\!d'{}}

242: 

243: \def\eps{\varepsilon}

244: \def\epsi{\varepsilon}

245: \let\vkappa=\varkappa

246: \def\<{\langle}\let\la=\<

247: \def\>{\rangle}\let\ra=\>

248: \def\bss{\backslash} \let\bs=\bss

249: \def\comp{\Subset}

250: \def\d{\partial}

251: \def\dbar{{\barr\partial}}

252: \def\dbarj{\dbar_J}

253: \def\ddef{\mathrel{{=}\raise0.3pt\hbox{:}}}

254: \def\deff{\mathrel{\raise0.3pt\hbox{\rm:}{=}}}

255: \def\inv{^{-1}}

256: \def\hook{\hookrightarrow}

257: \def\fraction#1/#2{\mathchoice{{\msmall{ #1\over#2}}}%

258: {{ #1\over #2 }}{{#1/#2}}{{#1/#2}}}

259: \def\norm#1{\Vert #1 \Vert}

260: \def\half{{\fraction1/2}}

261: \def\le{\leqslant}

262: \def\vph{^{\mathstrut}}

263: \def\lrar{\longrightarrow}

264: \def\emptyset{\varnothing}

265: \def\scirc{\mathop{\mathchoice{\hbox{\small$\circ$}}{\hbox{\small$\circ$}}%

266: {{\scriptscriptstyle\circ}}{{\scriptscriptstyle\circ}}}}

267: %

268: \def\longpoints{\leaders\hbox to 0.5em{\hss.\hss}\hfill \hskip0pt}

269: \def\stateskip{\smallskip}

270: \def\state#1. {\stateskip\noindent{\bf#1. }} %\medskip

271: \def\statep#1. {\stateskip\noindent{\bf#1 }} %\medskip

272: \def\step#1{{\sl Step #1)}}

273: \def\proof{\state Proof. \.}

274: \def\Chi{\raise 2pt\hbox{$\chi$}}

275: \def\eg{\hskip1pt plus1pt{\sl{e.g.\/\ \hskip1pt plus1pt}}}

276: \def\ie{\hskip1pt plus1pt{\sl i.e.\/\ \hskip1pt plus1pt}}

277: \def\iff{if and only if }

278: %{\hskip1pt plus1pt{\sl iff\/\hskip2pt plus1pt }}

279: \def\wrt{with respect to }

280: \def\isl{{\mathrm i}}

281: \def\sli{{\sl i)} } \def\slip{{\sl i)}}

282: \def\slii{{\sl i$\!$i)} } \def\sliip{{\sl i$\!$i)}}

283: \def\sliii{{\sl i$\!$i$\!$i)} }\def\sliiip{{\sl i$\!$i$\!$i)}}

284: \def\sliv{{\sl i$\!$v)} } \def\slivp{{\sl i$\!$v)}}

285: \def\slv{{\sl v)} } \def\slvp{{\sl v)}}

286: \def\slvi{{\sl v$\!$i)} }

287: \def\reg{^\sf{reg}}

288: \def\sing{^\sf{sing}}

289: \def\mat{\sf{Mat}}

290: \def\matn{\mat(n,\cc)}

291: \def\diff{{\scr D}\mskip -2mu i\mskip -5mu f\mskip-6.3mu f}

292: \def\star{\mathop{\msmall{*}}}

293: \def\barr#1{\mskip1mu\overline{\mskip-1mu{#1}\mskip-1mu}\mskip1mu}

294: %\def\barr#1{\overline{#1}\vphantom{#1}}

295: \def\usig{U\!\bfsig}

296: 

297: %

298: \def\Chi{\raise 2pt\hbox{$\chi$}}

299: \def\yps{\Upsilon}

300: \let\phI=\phi\let\phi=\varphi\let\varphi=\phI\def\vphi{\varphi}

301: %

302: \def\igam{{\mathchar"0100}} % Italic \Gamma

303: \def\idel{{\mathchar"0101}} % Italic \Delta

304: \def\ithe{{\mathchar"0102}} % Italic \Theta

305: \def\ilam{{\mathchar"0103}} % Italic \Lambda

306: \def\ixi{{\mathchar"0104}} % Italic \Xi

307: \def\ipi{{\mathchar"0105}} % Italic \Pi

308: \def\isig{{\mathchar"0106}} % Italic \Sigma

309: \def\iyp{{\mathchar"0107}} % Italic \Upsilon

310: \def\iphi{{\mathchar"0108}} % Italic \Phi

311: \def\ipsi{{\mathchar"0109}} % Italic \Psi

312: \def\iom{{\mathchar"010A}} % Italic \Omega

313: 

314: %

315: \def\bfalpha{{\boldsymbol\alpha}}%

316: \def\bfbeta{{\boldsymbol\beta}}%

317: \def\bfgamma{{\boldsymbol\gamma}}%

318: \def\bfdelta{{\boldsymbol\delta}}%

319: \def\bfepslon{{\boldsymbol\varepsilon}}%

320: \def\bfvarepsilon{{\boldsymbol\epslon}}%

321: \def\bfzeta{{\boldsymbol\zeta}}%

322: \def\bfeta{{\boldsymbol\eta}}%

323: \def\bftheta{{\boldsymbol\theta}}%

324: \def\bfvartheta{{\boldsymbol\vartheta}}%

325: \def\bfiota{{\boldsymbol\iota}}%

326: \def\bfkappa{{\boldsymbol\varkappa}}%

327: \def\bfvarkappa{{\boldsymbol\kappa}}%

328: \def\bflambda{{\boldsymbol\lambda}} \let\bflamb=\bflambda%

329: \def\bfmu{{\boldsymbol\mu}}%

330: \def\bfnu{{\boldsymbol\nu}}%

331: \def\bfxi{{\boldsymbol\xi}}%

332: \def\bfpi{{\boldsymbol\pi}}%

333: \def\bfvarpi{{\boldsymbol\varpi}}%

334: \def\bfrho{{\boldsymbol\rho}}%

335: \def\bfvarrho{{\boldsymbol\varrho}}%

336: \def\bfsigma{{\boldsymbol{\sigma}}}

337:     \def\bfsig{\bfsigma}%

338: \def\bfvarsigma{{\boldsymbol\varsigma}}%

339: \def\bftau{{\boldsymbol\tau}}%

340: \def\bfupsilon{{\boldsymbol\upsilon}}%

341: \def\bfphi{{\boldsymbol\varphi}}%

342: \def\bfvarphi{{\boldsymbol\phi}}%

343: \def\bfchi{{\boldsymbol\chi}}%

344: \def\bfpsi{{\boldsymbol\psi}}%

345: \def\bfomega{{\boldsymbol\omega}}%

346: \def\bfell{{\boldsymbol\ell}}%

347: %

348: \def\scra{\scr{A}}

349: \def\scrb{\scr{B}}

350: \def\scrc{\scr{C}}

351: \def\scrd{\scr{D}}

352: \def\scrh{\scr{H}}

353: \def\scre{\scr{E}}

354: \def\scrf{\scr{F}}

355: \def\scri{\scr{I}}

356: \def\scrj{\scr{J}}

357: \def\scrk{\scr{K}}

358: \def\scrl{\scr{L}}

359: \def\scrm{\scr{M}}

360:   \def\whcalm{\wh{\scr M}}

361:   \def\barm{{\mskip5mu\barr{\mskip-5mu\scr M}}{}}

362:   \def\frm{{\frak{M}}}

363: \def\scrn{{\scr N}}

364: \def\scro{{\scr O}}

365: \def\scrp{{\scr P}}

366: \def\scrpp{{\scr{P\!\!P}}}

367: %  \def\whcalp{\wh{\scr P}}

368: \def\scrq{{\scr Q}}

369: \def\scrs{{\scr S}}

370: \def\scrt{{\scr T}}

371: \def\scru{{\scr U}}

372: \def\scrv{{\scr V}}

373: \def\scrw{{\scr W}}

374: \def\scrx{{\scr X}}

375: \def\scry{{\scr Y}}

376: \def\scrz{{\scr Z}}

377: 

378: \def\trans{\pitchfork}

379: \def\eps{\varepsilon}

380: \def\epsi{\varepsilon}

381: \def\bs{\backslash}

382: \def\ogran{{\hskip0.7pt\vrule height6pt depth2.5pt\hskip0.7pt}}

383: \def\comp{\Subset}

384: \def\d{\partial}

385: \def\dbar{{\barr\partial}}

386: \def\1{{1\mkern-5mu{\rom l}}}

387: \def\dbarj{\dbar_J}

388: \def\ge{\geqslant}

389: \def\inv{^{-1}}

390: \let\wh=\widehat

391: \let\wt=\widetilde

392: \def\fraction#1/#2{\mathchoice{{\msmall{ #1\over#2}}}%

393: {{ #1\over #2 }}{{#1/#2}}{{#1/#2}}}

394: \def\half{{\fraction1/2}}

395: \def\le{\leqslant}

396: \def\vph{^{\mathstrut}}

397: \def\emptyset{\varnothing}

398: \def\.{\thinspace}

399: \def\parnin{\par\noindent}

400: 

401: \def\ti#1{{\tilde{#1}}}

402: \def\term#1#2{\underline{#2}_{[{#1}]}}

403: 

404: \def\qed{\ \ \hfill\hbox to .1pt{}\hfill\hbox to .1pt{}\hfill $\square$\par}

405: 

406: \def\comment#1\endcomment{}

407: 

408: \abovedisplayskip=5pt plus3pt minus2pt

409:  \belowdisplayskip=\abovedisplayskip

410: 

411: %\show\abovedisplaylimit

412: %\def\o#1{\buildrel\circ \over{#1}}$\o u$

413: 

414: 

415: %

416: %%%

417: 

418: \def\lineeqqno(#1){\hfill\llap{\vbox to 10pt%

419: {\vss\begin{align} \eqqno(#1)\end{align}\vss}}\vskip1pt}

420: 

421: \textwidth= 6.3 truein

422: \textheight=9.4 truein

423: \vsize=9.5 truein

424: \voffset= -.5truein

425: \oddsidemargin 0.1in

426: \evensidemargin 0in

427: \advance\headheight 1.2pt

428: 

429: \numberwithin{equation}{subsection}

430: %\renewcommand{\theequation}{\thesection\thesubsection\arabic{equation}}

431: 

432: 

433: \def\newsection[#1]#2{\section{#2}\label{sec:#1}}

434:    \def\refsection#1{{\sl Section \ref{sec:#1}$\,$}}

435: \def\newsubsection[#1]#2{\subsection{#2}\label{sec:#1}\showlabel{\tt #1}}

436:    \def\refsubsection#1{{\sl Paragraph \ref{sec:#1}$\,$}}

437:    

438: 

439: 

440: 

441: \def\inDex#1{\index{#1}}

442: \def\Zpage#1{ page\ #1}

443: 

444: \newtheorem{thm}{Theorem}[subsection]

445:    \def\newthm#1{\begin{thm} \showlabel{\tt #1}\label{#1}%

446:       \index{{\sl Theorem \ref{#1}},  \ label {\tt #1}|Zpage}} 

447:    \def\refthm#1{{\sl Theorem \ref{#1}$\,$}}

448: \newtheorem{lem}[thm]{Lemma}

449:    \def\newlemma#1{\begin{lem} \showlabel{\tt #1}\label{#1}%

450:       \inDex{{\sl Lemma \ref{#1}},  \ label {\tt #1}|Zpage}} 

451:    \def\lemma#1{{\sl Lemma \ref{#1}$\,$}}

452: \newtheorem{prop}[thm]{Proposition}

453:    \def\newprop#1{\begin{prop} \showlabel{\tt #1}\label{#1}%

454:       \inDex{{\sl Proposition \ref{#1}},  \ label {\tt #1}|Zpage}}

455:    \def\propo#1{{\sl Proposition \ref{#1}$\,$}}

456: \newtheorem{corol}[thm]{Corollary}

457:    \def\newcorol#1{\begin{corol} \showlabel{\tt #1}\label{#1}%

458:       \inDex{{\sl Corollary \ref{#1}},  \ label {\tt #1}|Zpage}}

459:    \def\refcorol#1{{\sl Corollary \ref{#1}$\,$}}

460: \newtheorem{defi}{Definition}[subsection]

461:    \def\newdefi#1{\begin{defi} \showlabel{\tt #1}\label{#1}\rm %

462:       \inDex{{\sl Definition \ref{#1}},  \ label {\tt #1}|Zpage}}

463:    \def\refdefi#1{{\sl Definition \ref{#1}$\,$}}

464: 

465: \def\eqqno(#1){\label{eq#1}%

466:       \index{{\sl  Eqtn \ref{eq#1}},  \ label {\tt(#1)}|Zpage}}

467: \def\eqqref(#1){\eqref{eq#1}}

468: %\def\eeq{\end{equation}}

469: 

470: 

471: %\def\showlabel#1{}

472: \def\showlabel#1{}

473: % Uncomment this line if you want to see labels 

474: %\def\showlabel#1{{\rm[}{\tt #1}{\rm]}}

475: 

476: 

477: \makeindex{}

478: 

479: \begin{document}

480: \baselineskip=14.0pt plus 2pt

481: 

482: 

483: 

484: \title[Pseudoholomorphic curves and the symplectic isotopy problem]%

485: %[Symplectic isotopy and pseudoholomorphic curves]%

486: {Pseudoholomorphic curves 

487: \\[5pt]

488: and 

489: the symplectic isotopy problem%

490: }

491: \author[V.~Shevchishin]{Vsevolod V.~Shevchishin}

492: \address{Fakult\"at f\"ur Mathematik\\

493: Ruhr-Universit\"at Bochum\\ 

494: Universit\"atsstrasse 150\\

495: 44780 Bochum\\

496: Germany}

497: \email{sewa@@cplx.ruhr-uni-bochum.de}

498: %\curraddr{}

499: \dedicatory{}

500: \subjclass{}

501: %\thanks{The author was supported in part by ...}

502: %\date{This version: October 2000. 

503: %3rd version: March 2000. 

504: %2nd version: October 1999.

505: %1st version: June 1999.}

506: \keywords{}

507: \begin{abstract}

508: The deformation problem for pseudoholomorphic curves and related

509: geometrical properties of the total moduli space of pseudoholomorphic curves

510: are studied. A sufficient condition for the saddle point property

511: of the total moduli space is established. The local symplectic isotopy problem

512: is formulated and solved for the case of imbedded pseudoholomorphic curves. 

513: It is shown that any two symplectically imbedded surfaces $\Sigma_0, \Sigma_1

514: \subset \cp^2$ of the same degree $d\le 6$ are symplectically isotopic.

515: \end{abstract}

516: \maketitle

517: %\pagenumbering{roman}

518: \setcounter{tocdepth}{2}

519: %\setcounter{tocdepth}{1}

520: %\tableofcontents

521: 

522: 

523: 

524: \setcounter{section}{-1}

525: %\pagebreak[1]

526: 

527: \newsection[intro]{Introduction}

528: 

529: \baselineskip =14.0pt plus .5pt

530: The symplectic isotopy problem can be formulated as follows:

531: 

532: {\it For a given  symplectic 4-dimensional manifold $(X,\omega)$ and symplectically

533: imbedded compact connected oriented surfaces $\Sigma_0, \Sigma_1 \subset X$ 

534: in the same homology class $[A] \in \sfh_2(X,\zz)$, does there exist an isotopy 

535: $\{\Sigma_t\}_{t\in  [0,1]}$ connecting $\Sigma_0$ with $\Sigma_1$ such that all\/ 

536: $\Sigma_t$ are also symplectically imbedded?} 

537: 

538: In this case $\Sigma_0$ and $\Sigma_1$ are called {\sl symplectically isotopic}.

539: Note that by the {\sl genus formula} for pseudoholomorphic curves (see 

540: \refsection{1}) $\Sigma_0$ and $\Sigma_1$ have the same genus. 

541: The example of Fintushel and Stern \cite{Fi-St} (see \refsection{6} for details)

542: shows that in general the answer is negative. On the other hand, Sikorav 

543: \cite{Sk-3} gives an affirmative answer in the case of surfaces of positive degree

544: $d\le 3$ in $\cp^2$. So it is natural to ask under which conditions on $(X,\omega)$

545: and $[A] \in \sfh_2(X,\zz)$ the symplectic isotopy does exist.

546: 

547: In this paper techniques involving pseudoholomorphic curves are developed in

548: directions needed for solving the symplectic isotopy problem. The main result 

549: is the following 

550: 

551: \state Theorem 1. {\it Two symplectically imbedded compact connected oriented 

552: surfaces $\Sigma_0, \Sigma_1 \subset \cp^2$ of the same positive degree $d\le 6$ 

553: are symplectically isotopic}.

554: 

555: \smallskip

556: The proof is based on the solution of two problems which are closely related to 

557: the symplectic isotopy problem and concern the geometry of the total moduli space

558: of pseudoholomorphic curves. The first result is a sufficient condition for the 

559: {\it saddle point property} of the total moduli space of pseudoholomorphic 

560: curves. This removes one of the obstacles to the existence of a symplectic isotopy 

561: and is applied to solve the {\it local symplectic isotopy problem} for imbedded 

562: pseudoholomorphic curves. The latter appears as a necessary part of the global 

563: problem.

564: 

565: 

566: \smallskip

567: The obtained progress gives hope that the symplectic isotopy problem has

568: an affirmative solution in the case $c_1(X,\omega)[A] >0$. Note that compact 

569: symplectic 4-manifolds with this property are classified: Up to the case when 

570: $[A]$ is represented by an exceptional sphere, these are symplectic blow-ups of 

571: a rational or ruled symplectic manifold, see \cite{McD-Sa-3}, {\sl Corollary 1.5},

572: and also \refsection{6}.

573: 

574: \newsubsection[0.1]{Overview of main results} There is essentially only one known

575: method of constructing symplectic isotopies. Having its origin in Gromov's 

576: celebrated article \cite{Gro}, it utilizes moduli spaces of pseudoholomorphic 

577: curves. One fixes a homotopy $h(t) \deff J_t$, $t\in [0,1]$, of 

578: $\omega$-tame almost complex structures and considers the relative moduli space,

579: $$

580: \scrm_h \deff \bigr\{ (C,t) : 

581: C \text{ is an imbedded $J_t$-holomorphic 

582: curve in the homology class }[A] 

583: \bigl\}, 

584: $$

585: equipped with a natural topology and with the projection $\pi_h: (C,t) \in \scrm_h 

586: \mapsto t \in [0,1]$. It follows essentially from \cite{Gro} that for a generic 

587: homotopy $h(t) = J_t$ the space $\scrm_h$ is a smooth manifold of expected 

588: dimension $\dimr\scrm_h = 2c_1(X,\omega)[A] + 2g-1$ and the projection $\pi_h$ is 

589: also smooth. Moreover, if this dimension is positive, then there exists a generic 

590: path $h(t)=J_t$ such that the original surface $\Sigma_0$ (resp.\ $\Sigma_1$) is 

591: $J_0$-holomorphic (resp.\ $J_1$-holomorphic) so that $(\Sigma_i, i) \in \scrm_h$ 

592: for $i=0,1$. Using a trivial but crucial observation that for every $(C,t) \in 

593: \scrm_h$ the curve $C$ is an $\omega$-symplectic real surface, one tries to find 

594: the desired isotopy $\Sigma_t$ by constructing a continuous section $\sigma: t 

595: \in [0,1] \mapsto (\Sigma_t, t) \in \scrm_h$ connecting $(\Sigma_0,0)$ with 

596: $(\Sigma_1,1)$.

597: 

598: One can easily see possible obstacles to the existence of such a section $\sigma:

599:  [0,1]\to \scrm_h$. The first one is that the projection $\pi_h: \scrm_h \to [0,1]$,

600: considered as a function, can have local maxima and minima. Indeed, if some 

601: $(C^*, t^*)\in \scrm_h$ appears as a local maximum of $\pi_h$, then for all $t>

602: t^*$ there exists no $J_t$-holomorphic curves sufficiently close to $C^*$. Observe 

603: that the mere fact of existence of a $J_t$-holomorphic curve $C_t$ 

604: for $t>t^*$ does not help much, because this does not imply that such a curve $C_t$ 

605: is symplectically isotopic to $\Sigma_0$ or $\Sigma_1$. Note that exactly the 

606: existence of $J$-holomorphic curves is the main technical tool in the Gromov's 

607: article \cite{Gro}. 

608: 

609: On the other hand, this obstacle does not appear in the case when the projection 

610: $\pi_h$ has the following {\sl saddle point property}: the Hesse matrix 

611: of $\pi_h$ at any critical point has at least one positive and at least one negative

612: eigenvalues. In \refsection{3} we prove 

613: 

614: \state Theorem 2. {\it Assume that $c_1(X,\omega)[A]>0$. Then for a generic 

615: homotopy $h(t)$ all critical points of the projection $\pi_h: \scrm_h \to 

616:  [0,1]$ are saddle}. 

617: 

618: \smallskip

619: On the other hand, we also show that in the case $c_1(X,\omega)[A]\le 0$ local 

620: maxima and minima of the projection $\pi_h: \scrm_h \to [0,1]$ do exist in the case

621: of an apropriate generic choice of $h$. Note that in the example of Fintushel 

622: and Stern \cite{Fi-St} one has $c_1(X,\omega)[A]= 0$.

623: 

624: \medskip

625: The next obstacle to existence of the desired section $\sigma: [0,1]\to \scrm_h$

626: comes from the fact that in general $\scrm_h$ is not compact and the projection

627: $\pi_h$ is not proper. This means that while attempting to construct the section 

628: $\sigma: [0,1]\to \scrm_h$ we go to ``infinity'' in the space $\scrm_h$. The 

629: Gromov compactness theorem provides control on the limiting behavior of curves 

630: $C_t$ in this case. It says that some subsequence, say $C_{t_n}$, converges in 

631: a certain weak sense to a pseudoholomorphic curve $C^*$ in the same homology class

632: $[A]$. The curve $C^*$ can have several irreducible components, some of them 

633: multiple, and also several singular points. Thus we come to a problem of 

634: describing  symplectic isotopy classes of imbedded pseudoholomorphic

635: curves close to a given singular curve $C^*$.

636: 

637: Here we have essentially two different difficulties. The first one comes from

638: multiple components. At the moment, we have no remedy for this problem. So we 

639: appempt to avoid the appearence of multiple components. This is done

640: by imposing the following additional constraint. We consider the curves $C_t$ 

641: which pass through fixed points $\mbfx=(x_1,\ldots,x_k)$ on $X$. This means, 

642: however, that now we must consider a new moduli space $\scrm_{h,\mbfx}$ with a 

643: new projection $\pi_{h,\mbfx}: \scrm_{h,\mbfx} \to [0,1]$. In \refsection{3}

644: we also show that if the number $k$ of the fixed points is strictly less than

645: $c_1(X,\omega)[A]$, then the saddle point property for $\pi_{h,\mbfx}$ remains

646: valid. An easy calculation shows that fixing $k=3d-1$ generic points on $\cp^2$, 

647: we can avoid the appearance of multiple components in pseudoholomorphic curves 

648: of degree $d\le6$. This explains the restriction in the main theorem.

649: 

650: \smallskip

651: In the case when $C^*$ has no multiple components we still have to describe the

652: possible symplectic isotopy classes of imbedded pseudoholomorphic curves close to 

653: $C^*$. Recall that by the result of Micallef and White (see \refsubsection{1.2}) 

654: every singular point of $C^*$ is isolated and topologically equivalent to a 

655: singular point of a usual holomorphic curve. In this way we come to the {\sl local

656: symplectic isotopy problem} which asks about possible symplectic isotopy types of 

657: imbedded pseudoholomorphic curves in a neighborhood of a given isolated 

658: singularity, see \refsubsection{6.2} for details. 

659: 

660: The solution of the local symplectic isotopy problem is based on the simple

661: observation that for {\sl holomorphic curves} this problem has a trivial solution.

662: Namely, a generic holomorphic deformation of a given (local) holomorphic curve 

663: $C^*$ gives a non-singular curve $C$, and the set of such curves is open and 

664: connected. Using this fact, we expoit essentially the same method as in the case 

665: of the global problem and prove that it is possible to deform isotopically an 

666: imbedded pseudoholomorphic curve $C$ sufficiently close to a given singular curve 

667: $C^*$ into a genuine holomorphic curve. In this way we prove

668: 

669: \state Theorem 3. {\it There exists a unique symplectic isotopy class of 

670: non-singular pseudoholomorphic curves which are close to a given pseudoholomorphic 

671: curve $C^*$ without multiple components}.

672: 

673: \smallskip

674: It should be noted that in the proof the saddle point property from {\sl Theorem 

675: 2\/} is used in an essential way. {\sl Theorem 1\/} follows now by 

676: the procedure of avoiding multiple components as it is explained above. 

677: 

678: \medskip

679: Further technical results of the paper are as follows. \refsection{4} is devoted 

680: to the deformation problem of pseudoholomorphic maps with prescribed singularities.

681: It is shown that the subspace of such maps is an immerced submanifold of expected 

682: codimension in the total moduli space of pseudoholomorphic maps. In \refsection{3}

683: the second variation of the $\dbar$-equation is computed. The result establishes 

684: the relationship between the geometry of a pseudoholomorphic curve $C^*$ 

685: corresponding to a critical point of the projection $\pi_h: \scrm_h \to [0,1]$

686: and the eigenvalues of the Hesse matix $d^2\pi_h$ at this point. Combined with

687: transversality results, this yields the proof of {\sl Theorem 2}. Finally,

688: in \refsection{5} the problem of smoothing of nodal points on pseudoholomorphic 

689: curves is studied.

690: 

691: 

692: \bigskip

693: {\it Acknowledgements.}

694: The author would like to express his gratitude

695: to  A.~Huckle\-berry, S.~Ivash\-ko\-vich,

696: St.~Ne\-mi\-rov\-ski, St.~Orev\-kov, B.~Sie\-bert and J.-C.~Siko\-rav for 

697: numerous useful conversations, suggestions and remarks.

698: 

699: %\newpage

700: %\tableofcontents

701: %\newpage

702: 

703: %\bigskip

704: \newsection[1]{Deformation and the normal sheaf of pseudoholomorphic curves}

705: 

706: In this section we give a brief description of pseudoholomorphic %(ps.-hol.)

707: curves and the related deformation theory.

708: 

709: \newsubsection[1.1]{Pseudoholomorphic curves}

710: First we collect some facts from the Gromov's theory. 

711: %It was introduced as a tool for studying symplectic manifolds in \cite{Gro}, 

712: %and appears to be very effective. 

713: Since there are several books devoted to or

714: treating this theme (see \eg \cite{McD-Sa-1} or \cite{McD-Sa-2})

715: we only mention the basic definitions and results we shall use later.

716: 

717: \newdefi{def1.1.0} 

718: An {\sl almost complex structure} on a manifold $X$ is an

719: endomorphism  $J \in TM$ of the tangent bundle such that $J^2 = - \id$.

720: The pair $(X,J)$ is called an {\sl almost complex manifold}.

721: One of the most important classes of such manifolds appears in symplectic

722: geometry. An almost complex structure on a symplectic manifold $(X,\omega)$ is

723: called {\sl$\omega$-tame} if $\omega(v,Jv)>0$ for any non-zero tangent vector

724: $v$. It is well-known that the set $\scrj_\omega$ of $\omega$-tame almost complex

725: structures is non-empty and contractible, (see \eg \cite{Gro},

726: \cite{McD-Sa-1}, or \cite{McD-Sa-2}). In

727: particular, any two $\omega$-tame almost complex structures $J_0$ and $J_1$ can be

728: connected by a homotopy (path) $J_t, t\in [0,1],$ inside $\scrj_\omega$.

729: \end{defi}

730: 

731: \newdefi{def1.1.1} A {\sl parameterized $J$-holomorphic curve} in an almost 

732: complex manifold $(X,J)$ is given by a (connected) Riemann surface $S$ with a

733: complex structure $J_S$ on $S$ and a non-constant $C^1$-map $u: S \to X$ 

734: satisfying the Cauchy-Riemann equation

735: \begin{equation}

736:  du + J \scirc du \scirc J_S =0.

737: \eqqno(1.1.1)

738: \end{equation}

739: In this case we call $u$ a $(J_S, J)$-holomorphic map, or simply

740: $J$-holomorphic map. Here we use the fact that if $u$ is not constant, then

741: such a structure $J_S$ is unique. We shall also use the notion {\sl $J$-curve}

742: which, depending on the context, will mean a map $u:S \to X$, \ie a

743: parameterized curve in $X$, or an image $u(S)$ of $J$-holomorphic map, taken

744: with appropriate multiplicity, \ie a {\sl non-parameterized $J$-curve}.

745: \end{defi}

746: 

747: 

748: The equation \eqqref(1.1.1) is elliptic with the Cauchy-Riemann symbol. This 

749: provides regularity properties for $u$. In particular, $u$ is H\"older 

750: $C^{l+1,\alpha}$-smooth, $u\in C^{l+1,\alpha}(S,X)$, provided $J\in C^{l,\alpha}$

751: with integer $l\ge 1$ and $0< \alpha <1$. To simplify the notations we set $\ell

752: \deff l+ \alpha$ and write $C^\ell$ to indicate $C^{l,\alpha}$-smoothness.

753: In what follows we shall assume that almost complex structures $J$ on $X$ are

754: $C^\ell$-smooth for some fixed sufficiently big non-integer $\ell$.

755: 

756: 

757: 

758: An easy consequence of the tameness condition is that any $J$-holomorphic

759: {\sl imbedding} $u: S \to X$ with $J \in \scrj_\omega$ is {\sl symplectic}

760: \ie the pull-back $u^*\omega$ is non-degenerate on $S$. The converse is also

761: true: Any $C^{\ell+1}$-smooth symplectic imbedding $u:S \to X$ with $\ell >1$

762: is $J$-holomorphic for some $C^\ell$-smooth $\omega$-tame structure

763: $J$. For {\sl immersions} the situation is more complicated. We state a result 

764: in a setting which will be relevant later on 

765: (see, \eg \cite{Gro} or \cite{McD-Sa-2} for details). 

766: %We state a related result only for the special case we are interested in.

767: 

768: \newlemma{lem1.1.1}

769: Let $(X,\omega)$ be a symplectic manifold with $\dim_\rr X =4$, and $u: S \to X$ 

770: an $\omega$-symplectic $C^1$-map such that $u(S)$ has only simple transversal

771: {\sl positive} self-intersection points.

772: 

773: Then there exist an $\omega$-tame almost complex structure $J$ on $X$ and a complex

774: structure $J_S$ on $S$ and making $u$ a $J$-holomorphic map.

775: \end{lem}

776: 

777: %The claim is standard, see \eg \cite{Gro} or \cite{McD-Sa-2} for details. 

778: It is worth to make the following remark. If $x \in X$ is a self-intersection 

779: point of $u$, $x= u(z_1)= u(z_2)$ with $z_1 \not= z_2$, such that the tangent 

780: planes $du(T_{z_i}) \subset T_xX$ are transversal and {\sl complex} \wrt some 

781: structure $J_x$ in $T_xX$, then the intersection index of planes $du(T_{z_i})$ 

782: in $x$ is {\sl positive}. However, it is possible that two {\sl symplectic} planes

783: $L_i$ in $(\rr^4, \omega)$ have {\sl negative} intersection index.

784: 

785: 

786: \smallskip

787: More detailed considerations lead to 

788: %Making similar considerations in more details one obtains 

789: the {\sl genus formula} (also called {\sl adjunction formula}) for immersed 

790: symplectic surfaces in symplectic four-folds. For this let $(X,\omega)$ be 

791: a symplectic 

792: manifold of dimension 4, $S\deff \bigsqcup_{j=1}^dS_j$ a compact oriented surface 

793: and $u:  S \to X$ an immersion with only transversal self-intersection points.

794: Denote by $g_j$ the genus of $S_j$, by $[C]$ the homology class of the image

795: $C \deff u(S)$, $[C]^2$ the homological self-intersection number of $[C]$, and

796: by $c_1(X)$ the first Chern class of $(X, \omega)$. Define the {\sl geometric

797: self-intersection number} $\delta$ of $M=u(S)$ as the algebraic number of pairs

798: $z' \not= z'' \in S$ with $u(z') = u(z'')$, taken with the sign corresponding

799: to the intersection index.

800: 

801: 

802: \newlemma{lem1.1.2}

803: Suppose that $u: S \to X$ is a symplectic immersion which is 

804: compatible with the orientation on each component $S_j$ of $S$. Then

805: \begin{equation}

806: \sum_{j=1}^d g_j = {[C]^2 - c_1(X)[C]\over2} + d - \delta .

807: \eqqno(1.1.3)

808: \end{equation}

809: \end{lem}

810: 

811: \smallskip

812: An elementary proof uses the fact that for a {\sl symplectic} immersion

813: $u:S \to (X,\omega)$ one has $c_1(X)[C]= \chi(S) + \chi(N)$, where

814: $N$ is the normal bundle and $\chi$ denotes the Euler characteristic. Finally,

815: one observes that $\chi(N)= [C]^2 -2\delta$. For details, see \cite{Iv-Sh-1}.

816: 

817: \newsubsection[1.2]{Local structure of pseudoholomorphic curves}

818: For the most results of this paragraph we refer to \cite{Mi-Wh} where a very 

819: precise description of the local structure of pseudoholomorphic  curves is 

820: given. As a rough summary, one can say that the 

821: %The main pathos is that a 

822: local behavior of (non-parameterized) pseudoholomorphic  curves is essentially 

823: the same as for usual holomorphic curves.

824: 

825: \newlemma{lem1.2.1} {\rm (\cite{Mi-Wh})} Let $(X,J)$ be an almost complex

826: manifold of $\dimc X 

827: \allowbreak

828: =n$, $u: S \to X$  a~$J$-holomorphic map, and $x\in X$ a 

829: point. Suppose that $J\in C^2$ and that for any $x'\in X$ sufficiently close to

830: $x$ the pre-image $u\inv(x)$ is finite. Then there exist neighborhoods $U

831: \subset X$ of $x$, $U' \subset \cc^n$ of\/ $0\in \cc^n$ and a 

832: $C^1$-diffeomorphism $\phi: U \to U'$ such that $C' \deff \phi\bigl( u(S) 

833: \cap U\bigr)$ is a proper analytic curve in $U'$ and such that $\phi_*(J_x)=J\st$,

834: where $J\st$ is the standard complex structure in $\cc^n$.

835: \end{lem}

836: 

837: In particular, the notion of a (local) irreducible component of a

838: $J$-holomorphic curve $C=u(S)$ is well-defined. Further, in the special  case

839: when $(X,J)$ is an {\sl almost complex surface} one can correctly define

840: 

841: \noindent

842: \sli  {\sl the intersection index $\delta_{ij}(x)\in \nn$} of two local

843: components $C_i$ and $C_j$ at $x\in X$, and

844: 

845: \noindent

846: \slii {\sl the nodal number $\delta_i(x) \in \nn$} of a local component

847: $C_i$ at $x\in X$ (see \cite{Mil}, \S\.10 and \refdefi{def6.2.1}).

848: 

849: The main properties of these local invariants are summarized in

850: 

851: \newlemma{lem1.2.2}

852: \sli If $x\in C_i \cap C_j$, then $\delta_{ij}(x)

853: \ge 1$. The equality holds \iff $C_i$ and $C_j$ are smooth and

854: intersect transversally in $x$;

855: 

856: \slii The set $\{ z \in S\;:\; \delta_i(u(z)) >0 \}$ is discrete in $S$;

857: 

858: \sliii Suppose additionally that $S= \sqcup_{j=1}^d S_j$ is a closed surface

859: and $u: S \to X$ is an imbedding almost everywhere on $S$. Set $C\deff u(S)$.

860: Denote by $\delta$ the sum of all local intersection indices $\delta_{ij}(x)$ and

861: all local nodal numbers $\delta_i(x)$, the homology class of $C$ by $[C]$, and

862: the genera of particular components $S_j$ by $g_j$. Then

863: \begin{equation}

864: {\textstyle \sum_{j=1}^d} g_j = {[C]^2 - c_1(X)[C]\over2} + d -\delta.

865: \eqqno(1.2.1)

866: \end{equation}

867: \end{lem}

868: 

869: The formula \eqqref(1.2.1) is the {\sl genus formula for pseudoholomorphic

870: curves}. We shall also apply a local version of this result. Here we say that a 

871: pseudoholomorphic curve $C$ in a symplectic manifold $X$ is {\sl parameterized by 

872: a real surface $S$} if there exists a map $u: S \to X$ which is an imbedding 

873: outside a discrete subset in $S$. Such a surface $S$, possibly not connected, can

874: be constructed as the normalization of $C \subset X$.

875: 

876: \newlemma{lem1.2.2a} Let $B \subset \rr^4$ be the unit ball equipped with the 

877: standard symplectic structure $\omega\st$, and $C_1$, $C_2$ pseudoholomorphic 

878: curves in $B$.

879: Assume that the boundaries of the curves $\d C_i$ are imbedded in the boundary of

880: the ball $\d B$, are sufficiently close to each other, and that every $C_i$ meets

881: transversally $\d B$. Denote by $\Chi_i$ the Euler characteristic of the surface 

882: $S_i$ parameterizing $C_i$ and by $\delta_i$ the sum of the nodal number of 

883: singular points of $C_i$. Then

884: \begin{equation}\eqqno(1.2.1a)

885:  \Chi_1 - 2\delta_1 = \Chi_2 - 2\delta_2

886: \end{equation}

887: \end{lem}

888: 

889: 

890: 

891: \proof It is shown in \cite{Iv-Sh-1} that every $C_i$ can be perturbed 

892: to a nearby pseudoholomorphic curve $C'_i$ in such a way that every singular point 

893: $x \in C_i$

894: with the nodal number $\delta_x(C_i)\ge 2$ ``splits'' into $\delta_x(C_i)$ {\sl 

895: nodal points} on the perturbed curve $C'_i$, \ie the points where exactly two 

896: branches of $C'_i$ meet transversally. By this procedure the topology of each $S_i$ 

897: and the whole nodal number of every $C_i$ remain unchanged. After this, one can 

898: replace a sufficiently small neighborhood of every nodal point $x \in C'_i$ by a 

899: symplectically imbedded handle. This ``symplectic surgery of $C'_i$'' produce {\sl 

900: imbedded} pseudoholomorphic curves $C''_i$ with $\Chi(C''_i)= \Chi_i - 2\delta_i$.

901: 

902: 

903: Moreover, all this can be carried out with the boundaries $\d C_i$ unchanged. 

904: Further, the hypothesis of the lemma implies that the boundaries $\d C_i$ are

905: transversal to the standard symplectic structure on $\d B=S^3$ and are isotopic

906: as {\sl transversal links}, see \cite{Iv-Sh-1} and \cite{Eli}. Now one applies 

907: the theorem of Bennequin \cite{Bn}, see also \cite{Eli}, which claims that, up

908: to sign convention, $\Chi(C''_i)$ is the {\sl Bennequin index of $\d C_i$} and

909: depends only on the {\sl transversal isotopy} class of $\d C_i$. The lemma follows.

910: \qed

911: 

912: \medskip

913: The result of Micallef and White, \lemma{lem1.2.1}, is not sufficient for

914: our purpose, because it does not allow us to control local structure of

915: pseudoholomorphic  curves under deformation. A necessary tool is provided

916: by the following statement proven in \cite{Iv-Sh-1}. Here and thereafter 

917: $\Delta$ denotes the unit disc in $\cc$ equipped with the standard complex 

918: structure.

919: 

920: \newlemma{lem1.2.3} 

921: Suppose that a $f\in L^{1,2}_\loc(\Delta, \cc^n)$ is not identically $0$ and 

922: satisfies {\sl a.e.} 

923: the~inequality

924: \begin{equation}

925: | \dbar f(z)| \le h(z)\cdot |z|^k \cdot |f(z)| 

926: \eqqno(1.2.2)

927: \end{equation}

928: for some $k\in \nn$ and nonnegative $h\in L^p_\loc(\Delta)$ with $2<p<\infty$.

929: Then

930: \begin{equation}

931: f(z)=z^\mu\bigl(P^{(k)}(z) + z^k g(z)\bigr),

932: \eqqno(1.2.3)

933: \end{equation}

934: where $\mu\in\nn$, $P^{(k)}$ is a polynomial in $z$ of degree $\le k$ with

935: $P^{(k)}(0) \not=0$, and $g\in L^{1, p}_\loc(\Delta,\cc^n)\hookrightarrow C^{0,

936: \alpha}$, $\alpha=1-{2\over p}$, with $g(0)=0$.

937: \end{lem}

938: 

939: 

940: \medskip

941: Using this result one can obtain the following  description of the local

942: behavior of a pseudoholomorphic  map. Note that on a given almost complex

943: manifold $(X,J)$ in a neighborhood of a given point $x_0\in X$ there exist

944: an (integrable) complex structure $J^*$ with $J^*(x_0)= J(x_0)$ and

945: $J^*$-holomorphic coordinates $w_1, \ldots, w_n$, $n=\dimc X$.

946: 

947: \newlemma{lem1.2.4} \sli Assume that $J$ is $C^1$-smooth and $u: \Delta \to

948: X$ is a non-constant $J$-holomorphic map with $u(0)=x_0$. Then in coordinates

949: $w_1, \ldots, w_n$ chosen as above in a neighborhood of $x_0\in X$ the map $u$

950: has the form

951: \begin{equation}

952: u(z) = z^\mu\cdot P^{(\mu-1)}(z) + z^{2\mu-1} \cdot v(z),

953: \eqqno(1.2.4)

954: \end{equation}

955: where $\mu\in \nn$, $P^{(\mu-1)}(z)$ is a complex $\cc^n$-valued polynomial

956: of degree $\le \mu-1$ with $P^{(\mu-1)}(0)\not=0$, and $v(z) \in

957: L^{1,p}(\Delta, \cc^n)$ with $v(0)=0$. 

958: 

959: \slii  Assume that $J$ is $C^1$-smooth and let $u_1, u_2: \Delta \to X$

960: be $J$-holomorphic maps such that $u_1$ is an immersion and $u_2(0) \in 

961: u_1(\Delta)$. Then there exists $r\in ]0,1[$ such that either $u_2(\Delta(r))

962: \subset u_1(\Delta)$ or $u_2(\Delta(r)) \cap u_1(\Delta)= u_2(0)$.

963: 

964: \sliii Let $J$ be a $C^\ell$-smooth almost complex structure

965: on the ball $B \subset \cc^n$ with $J(0) = J\st(0)$, and let $u_1, u_2 : 

966: \Delta \to B$ be $J$-holomorphic maps with $u_1(0) =u_2(0) =0 \in B$,

967: such that $u_1 \not = u_2$.

968: 

969: Then there exists a uniquely defined $\nu \in \nn$ and $w(z) \in C^1(

970: \Delta, \cc^n)$ such that 

971: \begin{equation}

972: u_1(z) - u_2(z) = z^\nu w(z).

973: \eqqno(1.2.5)

974: \end{equation}

975: \end{lem}

976: 

977: \proof The first and second parts of the lemma are proven in \cite{Iv-Sh-1}. 

978: For the third see {\sl Remark 1.6} and {\sl Section 6} of \cite{Mi-Wh}.

979: 

980: 

981: 

982: \smallskip

983: \newdefi{def1.2.1} If for an appropriate local complex coordinate $z$ on $S$ 

984: a $J$-holomorphic map $u:S \to X$ has the form \eqqref(1.2.4), then we call 

985: the (uniquely defined) $\mu$ the {\sl multiplicity of $u$ at the point $z=0$}.

986: \end{defi}

987: 

988: \newdefi{def1.2.2} A $J$-holomorphic map $u: S \to X$ is {\sl multiple}

989: if there exists a non-empty $U \subset S$ such that the restriction

990: $u\ogran_U$ can be represented as a composition $u\ogran_U = u' \scirc \phi$

991: where $u': \Delta \to X$ is a $J$-holomorphic map and $\phi: U \to \Delta$ is

992: a (branched) covering of degree $m\ge 2$. In other words, $u$ is multiple

993: if some part of the image $u(S)$ is multiply covered by $u$. 

994: \end{defi}

995: 

996: 

997: 

998: \newsubsection[1.3]{Deformation of pseudoholomorphic maps and the Gromov operator

999: $D_{u, J}$}

1000: Roughly speaking, the main idea of the Gromov's theory is to construct and

1001: study $J$-holomorphic curves in a symplectic manifold $(X, \omega)$ for some

1002: special (\eg {\sl integrable}) $J$. Often, one can show the existence of a

1003: $J_0$-holomorphic curve with some other almost complex structure $J_0$, see \eg

1004: {\sl Lemma \ref{lem1.1.1}}. If both $J$ and $J_0$ are $\omega$-tame, then 

1005: there exists a homotopy $\{J_t\}_{t\in [0,1]}$ from $J_0$ to $J_1=J$. Hence 

1006: one could try to deform the constructed $J_0$-holomorphic map $u_0:S \to X$ 

1007: into a $J_1$-holomorphic one using the continuity principle. The first step in 

1008: this direction is to study the linearization of (\ie the first variation) the 

1009: equation \eqqref(1.1.1). This means that we are interested in the first

1010: differential of the section $\sigma_\dbar$.

1011: 

1012: \smallskip

1013: Fix a compact surface $S$ of genus $g$. Denote by $\scrj_S$ the Banach

1014: manifold of $C^{1,\alpha}$-smooth complex structures on $S$ with some fixed

1015: $\alpha \in\; ]0,1[$. Thus

1016: \begin{equation}

1017: \scrj_S = \{ J_S\in C^{1,\alpha}(S, \endo(TS)):

1018: J_S^2=-\id \}

1019: \end{equation}

1020: and the tangent space to $\scrj_S$ at $J_S$ is

1021: \begin{equation}

1022: T_{J_S}\scrj_S = \{ I\in C^{1,\alpha}(S, \endo(TS)) :

1023: J_S I + I J_S = 0\} \equiv C^{1,\alpha}(S, \Lambda^{0,1}S\otimes TS),

1024: \end{equation}

1025: where $\Lambda^{0,1}S$ denotes the line bundle of $(0,1)$-form on $S$.

1026: 

1027: Let $\scrj$ be an open {\sl connected} subset in the Banach manifold of all

1028: $C^\ell$-smooth almost complex structures on $X$ for some fixed non-integer $\ell>

1029: 2$. In our context the most interesting example is the set $\scrj_\omega$ of

1030: $C^\ell$-smooth $\omega$-tame almost complex structures on $X$. The tangent space

1031: to $\scrj$ at $J$ consists of $C^\ell$-smooth $J$-antilinear endomorphisms of

1032: $TX$,

1033: \begin{equation}

1034: T_J\scrj = \{ I\in C^\ell(X, \endo(TX)) : JI + IJ = 0\}

1035: \equiv C^\ell(X, \Lambda^{0,1}X \otimes TX),

1036: \end{equation}

1037: where $\Lambda^{0,1}X$ denotes the complex bundle of $(0,1)$-forms on $X$.

1038: 

1039: Fix $p$ with $2<p<\infty$. Then the set $L^{1,p}(S, X)$ of all Sobolev

1040: $L^{1,p}$-smooth maps from $S$ to $X$ is a Banach manifold. For $u\in L^{1,p}(

1041: S, X)$ we denote

1042: \begin{equation}

1043: E_u \deff u^*TX.

1044: \eqqno(1.3.1)

1045: \end{equation}

1046: In this notation, the tangent space at $u \in L^{1,p}(S, X)$ is $T_uL^{1,p}(S,

1047: X) = L^{1,p}(S, E_u)$, the space of $L^{1,p}$-smooth sections

1048: of the pulled-back tangent bundle of $X$.

1049: 

1050: Fix a homology class $[C] \in \sfh_2(X, \zz)$ and consider the set

1051: \begin{equation}

1052: \scrs = \{ u\in L^{1,p}(S, X):u(S)\in [C]\}

1053: \end{equation}

1054: of maps $u$ representing the class $[C]$. Then $\scrs$ is open

1055: in $L^{1,p}(S, X)$ and has the same tangent space, $T_u\scrs = L^{1,p}(S,

1056: E_u)$. Since $\scrj$ is connected, the first Chern class $c_1(X,J)$ is

1057: constant on $\scrj$. We shall denote it simply by $c_1(X)$. Set

1058: \begin{equation}

1059: \mu \deff \la c_1(X), [C] \ra.

1060: \eqqno(1.3.1a)

1061: \end{equation}

1062: 

1063: 

1064: \smallskip

1065: Consider the subset $\scrp\subset \scrs\times \scrj_S \times \scrj$

1066: consisting of all triples $(u,J_S,J)$ with $u$ being $(J_S,J)$-holomorphic,

1067: \begin{equation}

1068: \scrp = \{(u, J_S, J)\in \scrs\times \scrj_S\times \scrj: du +

1069: J\scirc du\scirc  J_S = 0 \}.

1070: \eqqno(1.3.2)

1071: \end{equation}

1072: 

1073: Let $\nabla$ be some symmetric connection on $TX$. Covariant differentiation

1074: of \eqqref(1.1.1) gives the equation for the tangent space to $\scrp$.

1075: Namely, a vector $(v,\dot J_S,\dot J)$ is tangent to $\scrp$ at the point

1076: $(u,J_S,J)$ if it satisfies the equation

1077: \begin{equation}

1078: \nabla v + J\scirc \nabla v\scirc J_S + (\nabla_vJ)\scirc (du\scirc J_S) +

1079: J\scirc du\scirc \dot J_S + \dot J\scirc du\scirc J_S = 0. \eqqno(1.3.3)

1080: \end{equation}

1081: 

1082: \newdefi{def1.3.1}\.{\bf a)} For a complex bundle $E$ over $S$ let 

1083: \begin{equation}

1084: L^p_{(0,1)}(S, E)\deff L^p(S, E\otimes \Lambda^{(0,1)}S)

1085: \end{equation}

1086: denote the Banach space of $L^p$-integrable $E$-valued (0,1)-forms on $S$.

1087: 

1088: \statep b). Let $u$ be a $J$-holomorphic curve in $X$. Define the operator

1089: $D_{u, J}: L^{1,p}(S, E_u) \to L^p_{(0,1)}(S, E_u)$ as

1090: \begin{equation}

1091: D_{u, J}(v) \deff \nabla v + J\scirc\nabla v\scirc J_S

1092: + (\nabla_vJ) \scirc du\scirc J_S

1093: \eqqno(1.3.4)

1094: \end{equation}

1095: 

1096: \statep c). Define complex Banach bundles $\scre$ and $\scre'$ over $\scrs

1097: \times \scrj_S \times \scrj$ by 

1098: \begin{equation}

1099: \scre_{(u,J_S,J)} \deff L^{1,p}(S, E_u)

1100: \quad\hbox{and}\quad

1101: \scre'_{(u,J_S,J)} \deff L^p_{(0,1)}(S, E_u).

1102: \eqqno(1.3.5)

1103: \end{equation}

1104: These bundles are $C^\ell$-smooth and the formula \eqqref(1.3.4) defines a

1105: $\rr$-linear homomorphism $D=D_{u,J_S,J}: \scre \to \scre'$ which is $C^{\ell-1}

1106: $-smooth. The bundle $\scre$ is essentially the tangent bundle to $\scrs$, whereas 

1107: $\scre'$ appears as the space where the equation \eqqref(1.1.1) ``lives''. 

1108: More precisely, \eqqref(1.1.1) defines a section $\sigma_\dbar$ of

1109: $\scre'$,

1110: \begin{equation}

1111: \sigma_\dbar: (u,J_S,J) \in \scrs \times \scrj_S \times \scrj

1112: \;\mapsto \;

1113:  (du + J \scirc du \scirc J_S) \in \scre'_{(u,J_S,J)},

1114: \eqqno(1.3.6)

1115: \end{equation}

1116: such that the equation \eqqref(1.1.1) reads $\sigma_\dbar(u, J_S, J)=0$. 

1117: The space $\scr P$ appears then as the zero set of the section $\sigma_\dbar$.

1118: \end{defi}

1119: 

1120: \state Remark. Here and thereafter we use the normalization ${\d \over \d\bar z} 

1121: = {\d\over \d x} + \isl {\d\over \d y}$, deviating from the usual convention 

1122: ${\d \over \d\bar z} = \mathbf{\half} \cdot ({\d\over \d x} +\isl {\d\over\d y})$.

1123: The same normalization is used for all operators with Cauchy-Riemann symbol.

1124: 

1125: 

1126: \newlemma{lem1.3.1}

1127: Let $\scrx$ be a Banach manifold, $\scre\to \scrx$

1128: and $\scre'\to\scrx$ $C^1$-smooth Banach bundles over $\scrx$, $\nabla$ and

1129: $\nabla'$ linear connections in $\scre$ and $\scre'$ respectively, $\sigma$

1130: a $C^1$-smooth section of $\scre$, and $D: \scre \to \scre'$ a $C^1$-smooth

1131: bundle homomorphism.

1132: 

1133: \sli If $\sigma(x)=0$ for some $x\in \scrx$, then the map $\nabla\sigma_x:

1134: T_x\scrx \to \scre_x$ is independent of the choice of the connection $\nabla$

1135: in $\scre$;

1136: 

1137: \slii Set $K_x\deff \ker(D_x:\scre_x\to \scre'_x)$ and $Q_x\deff \coker(D_x:

1138: \scre_x\to \scre'_x)$ with the corresponding imbedding $i_x:K_x \to \scre_x$

1139: and projection $p_x:\scre'_x \to Q_x$. Let $\nabla^\hom$ be the connection in

1140: $\hom(\scre,\scre')$ induced by the connections $\nabla$ and $\nabla'$. Then

1141: the map

1142: \begin{equation}

1143: p_x \circ(\nabla^\hom\!\! D_x)\scirc i_x:

1144: T_x\scrx\to \hom(K_x,Q_x)

1145: \end{equation}

1146: is independent of the choice of connections

1147: $\nabla$ and $\nabla'$.

1148: \end{lem}

1149: 

1150: \state Remark. Taking this lemma into account, we shall use the following

1151: notation. For $\sigma\in \Gamma(\scrx, \scre)$, $D\in \Gamma(\scrx,

1152: \hom(\scre, \scre'))$ and $x\in \scrx$ as in the hypothesis of the lemma we

1153: shall denote by $\nabla\sigma_x:T_x\scrx \to \scre_x$ and ${\nabla\!D}:

1154: T_x\scrx\times \ker D_x \to \coker D_x$ the corresponding operators without

1155: pointing out which connections were used to define them.

1156: 

1157: \proof \sli Let $\wt\nabla$ be another connection in $\scre$. Then

1158: $\wt\nabla$ has the form $\wt\nabla=\nabla +A$ for some

1159: $A\in\Gamma(\scrx,\hom(T\scrx,\endo(\scre)))$. So for $\xi\in T_x\scrx$

1160: we obtain $\wt\nabla_\xi\sigma -\nabla_\xi\sigma = A(\xi,\sigma(x))=0$.

1161: 

1162: \slii Similarly, let $\wt\nabla'$ be another connection in $\scre'$, and let

1163: $\wt\nabla^\hom$ be the connection in $\hom(\scre,\scre')$ induced by

1164: $\wt\nabla$ and $\wt\nabla'$. Then $\wt\nabla'$ also has the form $\wt\nabla=

1165: \nabla +A'$ for some $A'\in\Gamma(\scrx,\hom(T\scrx, \endo(\scre')))$.

1166: So for $\xi\in T_x\scrx$ we obtain $\wt\nabla_\xi^\hom D -\nabla_\xi^\hom D=

1167: A'(\xi)\scirc D_x - D_x \scirc A(\xi)$. The statement of the lemma now follows

1168: from the identities $p_x\scirc D_x=0$ and $D_x\scirc i_x=0$.

1169: \qed

1170: 

1171: \smallskip

1172: \state Remark. The operator $D_{u, J}$ is the linearization of the equation

1173: \eqqref(1.1.1). Thus \lemma{lem1.3.1} shows that the definition of $D_{u, J}$

1174: is independent of the choice of $\nabla$. In particular, one can also use

1175: non-symmetric connections, \eg those compatible with $J$, as it is done in

1176: \cite{Gro}. However, it is convenient to have a fixed

1177: connection considering varying almost complex structures $J$ on $X$. But this 

1178: is impossible for $\nabla$ compatible with $J$. On the other hand, with a

1179: symmetric connection computations become simpler.

1180: 

1181: \smallskip

1182: The operator $D_{u, J}$, as well as the equation \eqqref(1.1.1) itself, is 

1183: elliptic

1184: of order 1 with the Cauchy-Riemann symbol. This implies standard regularity 

1185: properties for $D_{u, J}$. In particular, the kernel and the cokernel are of 

1186: finite dimension. The Riemann-Roch formula gives the index of $D_{u, J}$:

1187: \begin{equation}

1188: \dim_\rr \ker D_{u, J} - \dim_\rr \coker D_{u, J} = 2\cdot

1189: \bigl(\mu  + n (1-g) \bigr),

1190: \eqqno(1.3.7)

1191: \end{equation}

1192: where $\mu \deff c_1(X) \cdot [u(S)]$, $g$ is the genus of $S$, and $n$ the

1193: {\sl complex} dimension of $X$, \ie $n \deff \half \dim_\rr X$. The factor

1194: 2 appears because we compute {\sl real, not complex} dimensions of the

1195: (co)kernel.

1196: 

1197: 

1198: \newsubsection[1.4]{Holomorphic structure on the induced bundle}

1199: Now we want to understand the structure of the operator $D_{u,J}$ in more

1200: detail. Note that the pulled-back bundle $E_u = u^*TX$ carries a complex

1201: structure, namely $J$ itself, or more accurately $u^*J$. However, the

1202: operator $D\deff D_{u,J}$ is only $\rr$-linear. So we decompose it into

1203: $J$-linear and $J$-antilinear parts. Namely, for $v\in

1204: L^{1,p}(S, E)$ we write $D v = \half \bigl( D v - JD(Jv)\bigr) +

1205: \half (D v + JD (Jv)\bigr) = \dbar_{u, J}v+ R(v)$.

1206: 

1207: \newdefi{def1.4.1} The $J$-linear part $\dbar_{u, J}$ of the operator $D_{u, J}$

1208: is called the {\sl $\dbar$-operator associated with a $J$-holomorphic map $u$}.

1209: \end{defi}

1210: 

1211: \smallskip

1212: By the definition, the operator $\dbar_{u, J} : L^{1,p}(S, E_u) \to L^p

1213: _{(0,1)}(S, E_u)$ is $J$-linear. The following  statement is well

1214: known in the smooth case.

1215: 

1216: \newlemma{lem1.4.1A} Let $S$ be a Riemann surface with a complex

1217: structure $J_S$ and $E$ a $L^{1,p}$-smooth complex vector bundle of rank $r$

1218: over $S$. Let also $\dbar_E : L^{1,p}(S, E) \to L^p_{(0,1)}(S,E)$ be a complex

1219: linear differential operator satisfying the condition

1220: \begin{equation}

1221: \dbar_E(f\xi) = \dbar_S f \otimes \xi

1222: + f\cdot \dbar_E\xi,  \eqqno(1.3.6a)

1223: \end{equation}

1224: where $\dbar_S$ is the Cauchy-Riemann operator on $S$ associated to $J_S$.

1225: Then the sheaf

1226: \begin{equation}

1227: U\subset S \mapsto \scro(E)(U) := \{\,\xi \in  L^{1,p}(U, E) \, :\,

1228: \dbar_E\xi=0 \,\}   \eqqno(1.3.7a)

1229: \end{equation}

1230: is coherent and locally free of rank $r$. This defines a holomorphic

1231: structure on $E$ for which $\dbar_E$ is the associated Cauchy-Riemann

1232: operator.

1233: \end{lem}

1234: 

1235: 

1236: \state Remark. The condition $\eqqref(1.3.6a)$ means that $\dbar_E$ is of 

1237: order 1 and has the Cauchy-Riemann symbol. For the proof we refer to 

1238: \cite{Iv-Sh-1} and \cite{Iv-Sh-2} for the general case, or to \cite{H-L-S} 

1239: for  the case of line bundles.

1240: 

1241: 

1242: \smallskip

1243: Thus, according to \lemma{lem1.4.1A}, the operator $\dbar_{u, J}$ defines

1244: a holomorphic structure on the bundle $E_u$. We shall denote by $\scro(E_u)$

1245: the sheaf of holomorphic sections of $E_u$. The tangent bundle $TS$ to our

1246: Riemann surface also carries a natural holomorphic structure. We shall

1247: denote by $\scro(TS)$ the~corresponding coherent sheaf.

1248: 

1249: \medskip

1250: Denote by $N_J(v, w)$ the Nijenhuis torsion tensor of the almost complex

1251: structure $J$, (see \eg \cite{Ko-No}, Vol.II., p.123.)\.\footnote{\.

1252: Note that in \cite{Ko-No} another normalization constant is used. However,

1253: this is not essential for our purpose.}

1254: 

1255: \newlemma{lem1.4.1}

1256: \sli The $J$-antilinear part $R$ of $D_{u,J}$ is related to $u$ and $J$ by

1257: the formula

1258: \begin{equation}

1259: R(v)(\xi ) = N_J(v, du(\xi ))\qquad \xi \in TS.

1260: \eqqno(1.4.1)

1261: \end{equation}

1262: Thus $R$ is a continuous $J$-antilinear operator from $E$ to $\Lambda^{0,1}S

1263: \otimes E_u$ of order zero which satisfies $R\scirc du \equiv 0 $, \ie

1264: $R( du(\eta), \xi )=0$ for all $\eta ,\xi \in TS$.

1265: 

1266: \smallskip

1267: \slii If $u$ is non-constant, then $du$ defines an injective analytic

1268: morphism of coherent sheaves

1269: \begin{equation}

1270: 0\lrar \scro(TS) \buildrel du \over \lrar \scro(E_u). \eqqno(1.4.2)

1271: \end{equation}

1272: \end{lem}

1273: 

1274: \proof \sli Formula \eqqref(1.4.1) can be found in \cite{McD-2}. The rest of

1275: part \sli follows from the well-known fact that $N_J(v, w)$ is skew-symmetric

1276: and $J$-antilinear in both arguments.

1277: 

1278: \noindent \slii The fact that $du: TS \to E_u$ defines a morphism between

1279: coherent sheaves $\scro(TS)$ and $\scro(E_u)$ means that $du$ is

1280: a {\sl holomorphic} section of $T^*S \otimes E_u$. This is equivalent to

1281: relation

1282: \begin{equation}

1283: du \scirc \dbar_S = \dbar_{u,J} \scirc du.

1284: \end{equation}

1285: For the proof of this fact we refer to \cite{Iv-Sh-1} and \cite{Iv-Sh-2}.

1286: 

1287: Injectivity of the sheaf homomorphism $du$ is equivalent to its

1288: nondegeneracy which is the case in our context. \qed

1289: 

1290: 

1291: 

1292: \smallskip

1293: The zeros of the analytic morphism $du : \scro(TS) \to \scro (E_u)$ are

1294: isolated. So we obtain

1295: 

1296: \newcorol{cor1.4.2} {\rm(\cite{Mi-Wh}, \cite{Sk-1})}

1297: The set of critical points of a $J$-holo\-mor\-phic map is discrete, 

1298: provided $J$ is of class $C^1$.

1299: \end{corol}

1300: 

1301: \newdefi{def1.4.2} By the {\sl order of zero $\ord_p du$} of the

1302: differential $du$ at a point $p\in S$ we shall understand the order of

1303: vanishing at $p$ of the holomorphic morphism $du : \scro(TS)\to \scro(E_u)$.

1304: \end{defi}

1305: 

1306: It follows from \lemma{lem1.4.1} that $\ord_p du$ is a well-defined

1307: non-negative integer.

1308: 

1309: 

1310: \newsubsection[1.5]{The normal sheaf of a pseudoholomorphic curve}

1311: From \eqqref(1.4.2) we obtain the following short exact sequence of coherent 

1312: sheaves

1313: \begin{equation}

1314: 0\lrar \scro(TS) \buildrel du \over\lrar \scro(E_u)

1315: \lrar \scrn_u\lrar 0,

1316: \eqqno(1.5.1)

1317: \end{equation}

1318: where $\scrn_u\deff \scro(E)/du(\scro(TS))$ is the quotient sheaf. It follows from

1319: \lemma{lem1.4.1} \slii that there is a decomposition $\scrn_u= \scro(N_u)\oplus 

1320: \scrn_u\sing$ where $N_u$ is a holomorphic

1321: vector bundle and $\scrn_u\sing = \bigoplus_{z\in S} \cc_z^{\ord_z du}$ is a

1322: discrete sheaf with support in the set of critical points $a_i$ of $u$

1323: with the stalk $\cc^{n_i}$ of dimension $n_i \deff \ord_{a_i}du$ at every

1324: such point $a_i$. 

1325: 

1326: \newdefi{def1.5.1} The quotient sheaf $\scrn_u \deff \scro(E)/du(TS)$ is called 

1327: the {\sl normal sheaf of a $J$-curve $u:S \to X$}, $N_u$ the {\sl normal 

1328: bundle to the $J$-curve $u:S \to X$}, and $[A] \deff \sum n_i[a_i]$ 

1329: the {\sl branching divisor of the $J$-curve $u:S \to X$}.

1330: \end{defi}

1331: 

1332: 

1333: Denote by $\scro([A])$ the sheaf of meromorphic functions on $S$ having poles

1334: of order at most $n_i$ at $a_i$. Then \eqqref(1.5.1) gives rise to the exact 

1335: sequence of coherent sheaves

1336: \begin{equation}

1337: 0\lrar \scro(TS)\otimes \scro([A])\buildrel{du}\over

1338: {\lrar} \scro(E_u)

1339: \lrar \scro(N_u)\lrar 0.

1340: \eqqno(1.5.2)

1341: \end{equation}

1342: 

1343: The holomorphic structure in $N_u$ defines the Cauchy-Riemann operator

1344: $\dbar_N :L^{1,p}(S, N_u) \allowbreak \lrar L^p_{(0,1)}(S, N_u)$.

1345: \lemma{lem1.4.1} implies that

1346: the homomorphism $R: E_u \to E_u \otimes \Lambda^{(0,1)}S$

1347: induces a $J$-antilinear bundle homomorphism $R_N : N_u \to N_u \otimes

1348: \Lambda^{(0,1)}S$. Define the operator

1349: \begin{equation}

1350: D_{u, J}^N : L^{1,p}(S, N_u) \allowbreak \lrar L^p_{(0,1)}(S, N_u)

1351: \quad\hbox{by}\quad

1352: D_{u, J}^N \deff\dbar_N + R_N.

1353: \eqqno(1.5.3)

1354: \end{equation}

1355: 

1356: 

1357: \smallskip

1358: \newdefi{def1.5.2} Let $E$ be a holomorphic vector bundle over a compact Riemann 

1359: surface $S$ of genus $g$ and let $D:L^{1,p}(S, E)\to L^p(S, \Lambda^{0,1}S \otimes 

1360: E)$ be an operator of the~form $D=\dbar + R$ where $R\in L^p\bigl(S,\,\hom_\rr(E,

1361: \,\Lambda^{0, 1}S\otimes E) \bigr)$ with $2<p<\infty$. Define $\sfh^0_D(S, E)\deff 

1362: \ker D$ and $\sfh^1_D(S, E) \deff \coker D$.  The groups $\sfh^i_D(S, E)$ are 

1363: referred to as {\sl $D$-cohomology groups of $E$}.

1364: \end{defi}

1365: 

1366: \medskip

1367: The Riemann-Roch formula gives the {\sl index} of $D$,

1368: \begin{equation}

1369: \ind_\rr D \deff \dimr\sfh^0_D(S, E) - \dimr\sfh^1_D(S, E)=

1370: 2\bigl( c_1(E) + \rank(E) (1-g)\bigr).

1371: \eqqno(1.5.?4)

1372: \end{equation}

1373: 

1374: 

1375: \state Remark. Taking into account the elliptic regularity of the

1376: Cauchy-Riemann operator, for given $S$, $E$ and $R\in L^p$, $2<p<\infty$, one

1377: can define $\sfh^i_D(S, E)$ as the (co)kernel of the operator $\dbar +R:

1378: L^{1,q}(S, E) \to L^q(S,\,\Lambda^{0, 1}S\otimes E)$ for any $q\in\; ]1, p]$.

1379: So the definition is independent of the choice of the functional spaces. Note also

1380: that the $\sfh^i_D(S, E)$ are of finite dimension provided that $S$ is closed.

1381: For details, see \eg \cite{Iv-Sh-1}.

1382: 

1383: \smallskip

1384: The following lemmas contain main properties of $D$-cohomologies which will

1385: be used later. For complete proofs we refer to \cite{Iv-Sh-1} and \cite{Iv-Sh-2}.

1386: 

1387: \newlemma{lem1.5.1} {\sl (Serre duality for $D$-cohomologies.)}

1388: Let $E$ be a holomorphic vector bundle over a~compact Riemann

1389: surface $S$ and let $D:L^{1,p}(S, E)\to L^p_{(0,1)}(S, E)$ be

1390: an operator of the~form $D=\dbar + R$, where $R\in L^p\bigl(S,\,

1391: \hom_\rr(E,\,\Lambda^{0, 1}S\otimes E) \bigr)$ with $2<p <\infty$. Let

1392: $K_S\deff \Lambda^{1, 0}S$ be the~canonical holomorphic line

1393: bundle of $S$. Then there exists the~naturally defined operator

1394: \begin{equation}

1395: D^*=\dbar- R^*: L^{1,p}(S, E^* \otimes K_S)

1396: \to L^p_{(0,1)}(S, E^* \otimes K_S)

1397: \end{equation}

1398: with $R^* \in L^p\bigl(S,\,\homr(E^*\otimes K_S,\,

1399: \Lambda^{0, 1}S \otimes E^*\otimes K_S) \bigr)$ and the~natural isomorphisms

1400: \begin{equation}

1401: \matrix

1402: \sfh^0_D(S,\, E)^* &\cong& \sfh^1_{D^*}(S,\, E^*\otimes K_S),

1403: \cr

1404: \sfh^1_D(S,\, E)^* &\cong& \sfh^0_{D^*}(S,\, E^*\otimes K_S),

1405: \endmatrix

1406: \eqqno(1.5.4)

1407: \end{equation}

1408: induced by the pairings

1409: \begin{equation}

1410: \begin{matrix}%\format\l\ \ &\l\ \ &\l\\

1411: \phi \in \sfh^0_D(S,\, E), \quad

1412: \psi \in L^p_{(0,1)}(S,\, E^*\otimes K_S)&

1413: \mapsto \<\phi,\psi\> &

1414: \deff \re \int_S \psi \scirc \phi

1415: \cr

1416: \psi \in \sfh^0_D(S,\, E^*\otimes K_S), \quad

1417: \phi \in L^p_{(0,1)}(S,\, E)&

1418: \mapsto \<\phi,\psi\> &

1419: \deff \re \int_S \psi \scirc \phi

1420: \end{matrix}

1421: \eqqno(1.5.5)

1422: \end{equation}

1423: If, in addition, $R$ is $\cc$-antilinear, then $R^*$ is also

1424: $\cc$-antilinear.

1425: \end{lem}

1426: 

1427: 

1428: \state Remark. The lemma expresses the well-known relation $\ker(D^*) =(\im D)

1429: ^\perp$ between a linear operator $D$ and its adjoint. It is worth observing

1430: that the spaces themselves and the duality are defined only over the real numbers

1431: $\rr$ and not over $\cc$.

1432: 

1433: 

1434: \smallskip

1435: \newlemma{lem1.5.2}

1436: {\rm(\cite{H-L-S}, {\sl Vanishing theorem for $D$-cohomologies.})}

1437: Let $S$ be a closed Riemann surface of genus $g$ and $L$ 

1438: a~holomorphic {\sl line} bundle over $S$, equipped with a~differential

1439: operator $D=\dbar + R$ with $R\in L^p\bigr(S, \homr(L,\, \Lambda^{0,

1440: 1}S\otimes L) \bigl)$, $p>2$. If $c_1(L)<0$, then $\sfh^0_D (S,\, L)=0$. If

1441: $c_1(L)>2g-2$, then $\sfh^1_D (S,\, L)=0$.

1442: \end{lem}

1443: 

1444: 

1445: 

1446: \medskip

1447: The importance of the operator $D=\dbar + R$ lies in the fact that we can

1448: associate with the short exact sequence \eqqref(1.5.1) the long exact 

1449: sequence of $D$-cohomo\-logies. Note, that due to \lemma{lem1.4.1} we obtain

1450: the short exact sequence of complexes

1451: \begin{equation}

1452: \eqqno(1.5.6)

1453: \def\normalbaselines{\baselineskip20pt\lineskip3pt \lineskiplimit3pt }

1454: \setbox1=\hbox{$\lrar$}

1455: \def\mapright#1{\!\!\!\smash{\mathop{{\hbox to

1456: \wd1{\hss\hbox{$\displaystyle\longrightarrow$}\hss}}}\limits^{#1}}\!\!\!}

1457: %

1458: \def\mapdown#1{\Big\downarrow\rlap{$\vcenter{\hbox{$\scriptstyle#1$}}$}}

1459: %

1460: \begin{matrix}

1461: \llap{$0\lrar$}  L^{1,p}(S, TS)&

1462: \mapright{du}&

1463: L^{1,p}(S, E)&

1464: \mapright{\barr\pr}&

1465: L^{1,p}(S, E)\bigm/ du(L^{1,p}(S, 

1466: \rlap{$TS))\lrar0$}

1467: \\

1468: \mapdown{\barr \partial_S}\qquad&& \mapdown{D}\qquad&&

1469: \mapdown{\barr D}\qquad\qquad\qquad\qquad

1470: \\

1471: \llap{$0\lrar$} L^p_{(0,1)}(S, TS)&

1472: \mapright{du}&

1473: L^p_{(0,1)}(S, E)&

1474: \mapright{\barr\pr}&

1475: L^p_{(0,1)}(S, E)\bigm/

1476: du(L^p_{(0,1)}(S, 

1477: \rlap{$TS)) \lrar0$}

1478: \end{matrix}

1479: \end{equation}

1480: where $\barr D$ is induced by $D\equiv D_{u, J}$.

1481: 

1482: 

1483: \newlemma{lem1.5.3} For $\barr D$ as in \eqqref(1.5.6), $\ker \barr D =

1484: \sfh^0_D (S, N_u) \oplus \sfh^0(S, \scrn_u\sing)$ and $\coker \barr D

1485: \allowbreak= \sfh^1_D (S, N_u)$.

1486: \end{lem}

1487: 

1488: 

1489: \proof For an open set $U \subset S$ let $\Gamma_D(U, E_u) \deff \{ v\in L^{1,p}

1490: _\loc(U, E_u) \;:\; Dv = 0\}$. Use the analogous notation for $N_u$. Consider the 

1491: sheaves $U \mapsto \Gamma\, (U, \scro(TS)\,)$, $U \mapsto \Gamma_D (U, E_u)$, and 

1492: $U \mapsto \Gamma_D(U, N_u) \oplus \Gamma(U, \scrn_u \sing)$. It is easy to show 

1493: that the first two columns of the diagram \eqqref(1.5.6)

1494: define fine resolutions of the sheaves $\scro(TS)$ and $\Gamma_D (\, \cdot\,,

1495: E_u)$. Moreover, $du$ defines injective homomorphisms between these sheaves

1496: and between their resolutions. An explicit computation shows that $\Gamma_D(\,

1497: \cdot \,, N_u) \oplus \Gamma(\,\cdot\,, \scrn_u\sing)$ is the

1498: corresponding quotient sheaf and that the third column of \eqqref(1.5.6) is its

1499: resolution. For details, see \cite{Iv-Sh-1}. \qed

1500: 

1501: \newcorol{cor1.5.4} The~short exact sequence $\eqqref(1.5.1)$ induces

1502: the~long exact sequence of $D$-cohomologies

1503: \begin{equation*}

1504: \def\normalbaselines{\baselineskip20pt\lineskip3pt \lineskiplimit3pt }

1505: \def\mapright#1{\;\smash{\mathop{\longrightarrow}\limits^{#1}}\;}

1506: \def\mapdown{\Big\downarrow}

1507: %

1508: \matrix%\format\c&\c&\c&\c&\c&\c&\l&\c&\l\\

1509: 0& \mapright{}& \sfh^0(S, TS) &\mapright{} & \sfh^0_D(S, E)

1510:  & \mapright{}& \sfh^0_D(S, N_u)\oplus \sfh^0(S, \scrn_u\sing)

1511:  &\mapright{\delta} &\vphantom{\mapdown}\\

1512: %

1513:  & \mapright{}& \sfh^1(S, TS) &\mapright{} & \sfh^1_D(S, E)

1514:  & \mapright{}& \sfh^1_D(S, N_u)         &\mapright{} &0.

1515: \endmatrix

1516: \end{equation*}

1517: \end{corol}

1518: 

1519: \medskip

1520: \newsection[2]{The total moduli space of pseudoholomorphic curves}

1521: 

1522: \newsubsection[2.1]{Transversality}

1523: Any deformation of a given $J$-holomorphic map $u: S \to X$ defines a path in

1524: the space $\scrp$ of pseudoholomorphic  maps. Thus to construct such a deformation 

1525: we want to equip the space $\scrp$ with a structure of a smooth Banach manifold.

1526: 

1527: Note that by definition the set $\scrp$ is the zero set of the

1528: section $\sigma_\dbar$ of the bundle $\scre' $, \ie the intersection of

1529: the images of $\sigma_\dbar$ and the zero-section $\sigma_0$.

1530: Thus we are interested in which points these sections meet

1531: transversally. The analysis of the problem leads to the following

1532: 

1533: \newdefi{def2.1.1} Let $\scrx$, $\scry$, and $\scrz$ be Banach manifolds

1534: with $C^\ell$-smooth maps $f:\scry \to \scrx$ and $g:\scrz\to\scrx$, $\ell

1535: \ge1$. Define the {\sl fiber product} $\scry\times_\scrx \scrz $ by setting

1536: $\scry\times_\scrx \scrz \deff \{\,(y,z)\in \scry\times\scrz \;:\; f(y)= g(z)

1537: \,\}$. The map $f$ is called {\sl transversal to $g$} at a point $(y,z)\in

1538: \scry\times_\scrx \scrz$ with $x\deff f(y)=g(z)$, and $(y,z)$ is called a

1539: {\sl transversality point}, if the map $df_y\oplus dg_z: T_y\scry \oplus

1540: T_z\scrz \to T_x\scrx$ is {\sl surjective} and admits a {\sl closed

1541: complement} to its kernel. The set of transversality points $(y,z)\in \scry

1542: \times_\scrx \scrz$ will be denoted by $\scry\times^\trans_\scrx \scrz$, with

1543: $\trans$ symbolizing the transversality condition.

1544: 

1545: We say that $f:\scry \to \scrx$ is {\sl transversal} to $g:\scrz\to\scrx$ if

1546: every point $(y,z) \in \scry\times_\scrx \scrz$ is a transversality point. In

1547: particular, if $\scry$ consists of a point $x \in \scrx$ and the imbedding $\{

1548: x\} \hook \scrx$ is transversal to $g$, we call $x$ a {\sl regular value of

1549: $g$}. Note that by this definition any $x\in \scrx\bs g(\scrz)$ is a regular

1550: value of $g$.

1551: 

1552: In the special case when the map $g:\scrz \to \scrx$ is a closed imbedding,

1553: the fiber product $\scry\times_\scrx \scrz$ is simply the preimage

1554: $f\inv\scrz$ of $\scrz\subset\scrx$. In particular, every point

1555: $(y,z)\in\scry\times_\scrx \scrz$ is completely defined by its component

1556: $y\in\scry$, $z=f(y)\in\scrz\subset \scrx$. In this case we simply say that

1557: $f:\scry \to \scrx$ is {\sl transversal to $\scrz$ at $y\in\scry$},

1558: \iff $(y,f(y))$ is a transversal point of $\scry\times_\scrx \scrz\cong

1559: f\inv\scrz$.

1560: \end{defi}

1561: 

1562: 

1563: \newlemma{lem2.1.1}

1564: The set $\scry\times^\trans_\scrx \scrz$ is open in $\scry\times_\scrx \scrz$

1565: and is a $C^\ell$-smooth Banach manifold with tangent space

1566: \begin{equation}

1567: T_{(y,z)}\scry\times^\trans_\scrx \scrz = \ker\bigl(

1568: df_y\oplus d(-g_z): T_y\scry \oplus T_z\scrz \to T_x\scrx \bigr).

1569: \end{equation}

1570: \end{lem}

1571: 

1572: 

1573: \proof Fix $w_0\deff(y_0,z_0)\in \scry\times^\trans_\scrx \scrz$

1574: and set $K_0 \deff \ker( df_{y_0}\oplus dg_{z_0}:

1575: T_{y_0}\scry \oplus T_{z_0}\scrz \to T_x\scrx \bigr)$. Let $Q_0$ be a closed

1576: complement to $K_0$. Then the map $df_{y_0}\oplus dg_{z_0}:Q_0\to T_x\scrx$

1577: is an isomorphism.

1578: 

1579: Due to the choice of $Q_0$, there exists a neighborhood $V\subset

1580: \scry\times\scrz$ of $(y_0,z_0)$ and $C^\ell$-smooth maps $w':V\to K_0$ and

1581: $w'':V\to Q_0$, such that $dw'_{w_0}$ (resp.\ $dw''_{w_0}$) is the projection

1582: from $T_{y_0}\scry \oplus T_{z_0}\scrz$ onto $K_0$ (resp.\ onto $Q_0$),

1583: so that $(w',w'')$ are coordinates in some smaller  neighborhood $V_1\subset

1584: \scry\times\scrz$ of $w_0=(y_0,z_0)$. It remains to consider the equation

1585: $f(y)=g(z)$ in new coordinates $(w',w'')$ and apply the implicit function

1586: theorem.

1587: %

1588: \qed

1589: 

1590: \medskip

1591: Due to \lemma{lem2.1.1}, the set $\scrp$ is a Banach manifold at those points

1592: $(u,J_S,J)\in\scrp$ where $\sigma_\dbar$ is transversal to $\sigma_0$.

1593: However, at any point $(u,J_S,J;0)$ on the zero section $\sigma_0$ of

1594: $\scre'$ we have the natural decomposition

1595: \begin{equation}

1596: T_{(u,J_S,J;0)}\scre' = d\sigma_0\bigl( T_{(u,J_S,J)}

1597: (\scrs\times \scrj_S\times \scrj) \bigr)

1598: \oplus \scre'_{(u,J_S,J)},

1599: \end{equation}

1600: where the first component is the tangent space to the zero section of $\scre'$

1601: and the second one is the tangent space to the fiber $\scre'_{(u,J_S,J)}$.

1602: 

1603: Let $p_2$ denote the projection on the second component. Then the

1604: transversality $\sigma_\dbar$ and $\sigma_0$ is equivalent to the surjectivity

1605: of the map $p_2\scirc d\sigma_\dbar: T_{(u,J_S,J)}(\scrs\times \scrj_S \times

1606: \scrj)\to \scre'_{(u,J_S,J)}$, \ie to the surjectivity of the operator

1607: %\begin{equation}

1608: \begin{align*}

1609: \nabla\sigma_\dbar:\;&

1610:  T_uL^{1,p}(S,X)\oplus T_{J_S}\ttt_g \oplus T_J\scrj \lrar

1611: \scre'_{(u,J_S,J)}

1612: \\

1613: \nabla\sigma_\dbar:\;&(v,\dot J_S,\dot J) \longmapsto

1614: D_{(u,J)}v + J\scirc du \scirc \dot J_S

1615: + \dot J\scirc du \scirc J_S.

1616: \end{align*}

1617: %\end{equation}

1618: By \refdefi{def1.5.2}, the quotient of $\scre'_{(u,J_S,J)}$ with respect

1619: to the image of $D_{u,J}$ is $\sfh^1_D(S,E_u)$. The induced map

1620: \begin{equation}

1621: T_{J_S}\scrj_S \ni \dot J_S \mapsto J\scirc du \scirc \dot J_S

1622: \in \sfh^1_D(S,E_u)

1623: \eqqno(2.1.1)

1624: \end{equation}

1625: is also easy to describe. Recall that for a given complex

1626: structure $J_S$ on $S$ one has the Dolbeault isomorphism

1627: \begin{equation}

1628: \sfh^1(S,TS)= \coker \bigl(\,\dbar :

1629: C^{2,\alpha}(S, TS) \lrar C^{1,\alpha}_{(0,1)}(S, TS) \,\bigr)

1630: \end{equation}

1631: with the operator $\dbar$ associated to $J_S$. Recall also that $C^{1,\alpha}

1632: _{(0,1)}(S, TS)$ is the tangent space $T_{J_S}\scrj_S$. This shows that the map 

1633: \eqqref(2.1.1) is the same as the map $J\scirc du: \sfh^1(S,TS)

1634: \to \sfh^1_D(S,E_u)$ and, due to identity $J\scirc du = du\scirc J_S$ and 

1635: \refcorol{cor1.5.4}, its cokernel is $\sfh^1_D(S,N_u)$.

1636: 

1637: It remains to study the image of $T_J\scrj$ in $\sfh^1_D(S,N_u)$.

1638: 

1639: \newdefi{def2.1.2} For $(u,J_S,J)\in \scrp$ we define  $\Psi=

1640: \Psi_{(u,J)}:T_J\scrj \to \scre'_{(u,J_S,J)}$ by setting $\Psi_{(u,J

1641: )}( \dot J) \deff \dot J\scirc du \scirc J_S$. Let $\barr\Psi=\barr\Psi_{(u,

1642: J)}:T_J\scrj \to \sfh^1_D(S,N_u)$ be induced by $\Psi$. Finally, define

1643: $\scrp^* \deff \{ (u,J_S,J) \in \scrp: u \hbox{ is injective in generic }z

1644: \in S\}$.

1645: \end{defi}

1646: 

1647: \state Remark. One can show that $\scrp\bs \scrp^*$ consists of {\sl multiple

1648: curves} for which the map $u:(S,J_S) \to (X,J)$ admits a factorization

1649: $u= u' \scirc g$ for some non-trivial holomorphic branched covering $g:(S,J_S)

1650: \to (S',J'_S)$ and a $J$-holomorphic map  $u':(S',J'_S) \to X$. On the other

1651: hand, for any $(u,J_S,J) \in \scrp^*$ the map $u$ is a smooth imbedding outside 

1652: finitely many points in $S$. For details see \cite{Mi-Wh} or \cite{Iv-Sh-1}.

1653: 

1654: 

1655: \newlemma{lem2.1.2}  {\sl(Infinitesimal transversality)}.

1656: Let $(u,J_S,J)\in \scrp^*$. Then the operator $\barr\Psi:T_J\scrj \to

1657: \sfh^1_D(S, N_u)$ is surjective.

1658: \end{lem}

1659: 

1660: 

1661: \proof Choose some nonempty open set $V\subset S$, such that

1662: $u\ogran_V$ is an imbedding. Use Serre duality (\lemma{lem1.5.1}) and

1663: find a basis $\psi_1,\ldots \psi_l\in  \sfh^0_D(S, N^*\otimes K_S)\cong

1664: \sfh^1_D(S, N)^*$.

1665: 

1666: Note that $\psi_i$ satisfy the equation $D \psi_i =0$, where the operator

1667: $D= D_{N^*\otimes K_S}$ is of the form $\dbar + R$. One can show

1668: (see \eg \cite{Iv-Sh-1} or \cite{H-L-S}) that any solution $v$ of the equation

1669: $(\dbar + R)v=0$ is $L^{1,p}$-smooth and furthermore such a $v$ is either

1670: identically zero or has isolated zeros.

1671: 

1672: This implies that there exist $I_1,\ldots I_l\in  C^\ell(S, N \otimes

1673: \Lambda^{(0,1)} )$ with supports $\supp(I_i)$ in $V$ such that the

1674: matrix $\bigl(\re\int_S \psi_i \scirc I_j\bigr)_{i,j=1} ^l$ is non-degenerate.

1675: Since $u\ogran_V$ is a $C^{\ell+1}$-smooth imbedding, any such $I_i$ can be

1676: represented in the form $I_i = \dot J_i \scirc du \scirc J_S$ with some

1677: $\dot J_i \in C^\ell(X, \endo(TX))$ with $ J \scirc \dot J_i + \dot J_i\scirc

1678: J =0$. The latter relation means that $\dot J_i \in T_J\scrj$. \qed

1679: 

1680: 

1681: \smallskip

1682: \newcorol{cor2.1.3} $\scrp^*$ is a $C^\ell$-smooth Banach manifold

1683: with the tangent space

1684: \begin{equation}

1685: T_{(u,J_S,J)} \scrp^* = \bigl\{ (v,\dot J_S, \dot J) \;:\;

1686: D_{u,J}v + J \scirc du \scirc \dot J_S + \dot J \scirc du \scirc J_S

1687: =0 \;\bigr\}.

1688: \eqqno(2.1.2)

1689: \end{equation}

1690: \end{corol}

1691: 

1692: \state Remark. $\scrp^*$ is in general smaller than the set $\scrp^\trans

1693: \deff \sigma_\dbar \times^\trans_{\scre'} \sigma_0$ of all transversality

1694: points of $\scrp$. On the other hand, it is sufficient for applications to

1695: consider only the space $\scrp^*$.

1696: 

1697: 

1698: 

1699: \newsubsection[2.2]{Moduli space of pseudoholomorphic curves}

1700: The space $\scrp$ (see \eqqref(1.3.2)) of all pseudoholomorphic maps is too big.

1701: Indeed, one has a natural right action of the group $\diff _+(S)$ of the

1702: orientation preserving $C^{2,\alpha}$-smooth diffeomorphisms of $S$ on the

1703: product $\scrs \times \scrj_S \times \scrj$ given by formula

1704: \begin{equation}

1705: (u,J_S,J) \in \scrs \times \scrj_S \times \scrj, g\in \diff_+(S)

1706: \; \lrar \; (u,J_S,J)\cdot g \deff (u\scirc g, J_S \scirc g, J),

1707: \end{equation}

1708: such that $\scrp$ and $\scrp^*$ are invariant \wrt this action. It is natural

1709: to consider $(u,J_S,J) \in \scrp$ and $(u,J_S,J) \cdot g$ as two

1710: parameterization of the same a $J$-curve. In other words, we are interested

1711: in the quotient space $\scrp/ \diff_+(S)$ rather than the space $\scrp$

1712: itself. As in Yang-Mills theory one can treat the group

1713: $\diff_+(S)$ as the gauge group of the problem and the quotient as

1714: the corresponding moduli space. Again as in Yang-Mills theory it is useful to

1715: know in which points the quotient spaces $\scrp/\diff_+(S)$ is

1716: a Banach manifold.

1717: 

1718: \smallskip

1719: First we consider the action of the group $\diff_+(S)$ on the space $\scrj_S$.

1720: Note that if $J'_S, J''_S \in \scrj_S$ are related by $J''_S = J'_S \scirc g$

1721: for some $C^1$-diffeomorphism $g:S \to S$, then $g$ is $(J'_S, J''_S)

1722: $-holomorphic. Since $J'_S$ and $J''_S$ are $C^{1,\alpha}$-smooth, elliptic

1723: regularity implies that $g$ is $C^{2,\alpha}$-smooth.

1724: 

1725: Further, it is known that the action of $\diff_+(S)$ on $\scrj_S$

1726: admits a global finite-dimen\-sional slice. To describe this slice we recall

1727: some standard facts from Teichm\"uller theory.

1728: 

1729: \smallskip

1730: Denote by $\ttt_g$ the Teichm\"uller space of marked complex structures on $S$.

1731: This is a complex manifold of dimension

1732: \begin{equation}

1733: \dim_\cc \ttt_g = \cases  0      &\text{ if $g=0$;}  \\

1734:                        1      &\text{ if $g=1$;}  \\

1735:                        3g-3   &\text{ if $g\ge2$;}

1736:                \endcases

1737: \end{equation}

1738: which can be completely characterized in the following way.

1739: 

1740: \newprop{prop2.2.A1} The product

1741: $S\times \ttt_g$ admits a (non-unique) complex (\ie holomorphic) structure

1742: $J_{S\times \ttt}$ such that:

1743: 

1744: \sli The natural projection $\pi_\ttt: S\times \ttt_g \to \ttt_g$ is

1745: holomorphic, so that for any $\tau\in \ttt_g$ the identification $S \cong S

1746: \times \{\tau\}$ induces the complex structure $J_S(\tau)\deff J_{S\times

1747: \ttt}\ogran_{S\times \{\tau\}}$ on $S$;

1748: 

1749: \slii For any complex structure $I_S$ on $S$ there exist a uniquely

1750: defined $\tau\in\ttt_g$ and a diffeomorphism $f: S\to S$ homotopic to the

1751: identity map $\id_S: S\to S$ such that $I_S = f^*J_S(\tau)$;

1752: 

1753: \sliii Moreover, for any finite-dimen\-sional manifold $Y$ and any smooth

1754: map $H: Y \to \scrj_S$ there exist maps $F:Y \to \diff_+(S)$ and $h:Y \to

1755: \ttt$ such that $H(y) = (F(y))^* h(y)$;

1756: 

1757: \sliv The group $\bfg$ of automorphisms of $S\times \ttt_g$ preserving the

1758: projection onto $\ttt_g$ is

1759: \begin{equation}

1760: \bfg=\cases \mathbf{PGl}(2,\cc)       &\text{ for $g=0$,} \\

1761:             \mathbf{Sp}(2,\;\;\zz) \ltimes T^2   &\text{ for $g=1$,} \\

1762:             \hbox{\rm discrete}                  &\text{ for $g\ge2$;}

1763: \endcases

1764: \eqqno(2.2.1)

1765: \end{equation}

1766: 

1767: \slv The tangent space to $\ttt_g$ at $\tau$ is canonically isomorphic to

1768: $\sfh^1(S, TS)$ where $S$ is equipped with the structure $J_S(\tau)$. The

1769: group $\sfh^0(S, TS)$ is canonically isomorphic to the Lie algebra of

1770: $\bfg$.

1771: \end{prop}

1772: 

1773: \smallskip

1774: We shall assume that such a structure $J_{S \times \ttt}$ is fixed. Then we

1775: obtain an imbedding $\ttt \hook \scrj_S$ given by $\tau \in \ttt \mapsto

1776: J_S(\tau) \in \scrj_S$. Using it, we identify $\ttt$ with its image in

1777: $\scrj_S$. For any $J_S \in \ttt_g$ this induces a monomorphism

1778: $T_{J_S}\ttt_g \hook T_{J_S}\scrj_S=C^{1,\alpha}_{(0,1)}(S, TS)$. Now, the

1779: isomorphism $T_{J_S}\ttt_g \cong \sfh^1(S, TS)$ mentioned in

1780: \slv is obtained as the composition

1781: \begin{equation}

1782: T_{J_S}\ttt_g \hook T_{J_S}\scrj_S=C^{1,\alpha}_{(0,1)}(S, TS) \lrar

1783: C^{1,\alpha}_{(0,1)}(S, TS)/\dbar\bigl(C^{2,\alpha}(S, TS)\bigr) =

1784: \sfh^1(S, TS).

1785: \eqqno(2.2.2)

1786: \end{equation}

1787: 

1788: \smallskip

1789: By our construction, any $\diff_+(S)$-orbit in $\scrj_S$ intersects

1790: $\ttt$. This implies that instead of $\scrp/\diff_+(S)$ we can consider the

1791: quotient $\scrp \cap (\scrs \times \ttt \times \scrj_S)$ by the action of

1792: $\bfg$.

1793: 

1794: \newdefi{def2.2.1} Let $\whcalm \deff \scrp^* \cap (\scrs \times

1795: \ttt \times \scrj_S)$ and use the same the notations to the restrictions of

1796: the bundles $\scre$ and $\scre'$ onto $\whcalm$ and for the induced

1797: operator $D: \scre\to \scre'$. The quotient $\scrm \deff \whcalm/ \bfg$ is

1798: the {\sl total moduli space of parameterized pseudoholomorphic curves}. It

1799: is equipped with the projection $\pr_{\!\!\scrj}:\scrm \to \scrj$. Elements of 

1800: $\scrm$ are denoted by $[u,J]$. To indicate the surface $S$, the ambient manifold

1801: $X$, and the homology class $[C]\in \sfh_2(X, \zz)$ involved in the definition

1802: of $\scrm$ we shall also use the notation $\scrm(S, X, [C])$. The same meaning

1803: have the notations $\whcalm(S, X, [C])$ and $\scrp(S, X, [C])$.

1804: \end{defi}

1805: 

1806: \newlemma{lem2.2.1}.

1807: \sli The projection $\wh\pr:\whcalm \lrar \scrm$ is a principal $\bfg$-bundle.

1808: 

1809: \slii The bundles\/ $\scre$ and $\scre'$ over $\whcalm$ admit a natural

1810: $C^\ell$-smooth $\bfg$-action such that $D:\scre \to \scre'$ is

1811: $\bfg$-invariant.

1812: 

1813: \sliii For any $J \in \scrj$ and any non-multiple $J$-holomorphic map $u: S

1814: \to X$ there exists a diffeomorphism $\phi: S \to S$ such that $[u\scirc \phi,

1815: J]$ lies in $\scrm$. Moreover, such element of $\scrm$ is unique.

1816: \end{lem}

1817: 

1818: 

1819: \state Remarks.~1. Part \slii of the lemma is equivalent to the existence

1820: of $C^\ell$-smooth bundles $\scre_\scrm$ and $\scre'_\scrm$ over $\scrm$ and

1821: a $C^{\ell-1}$-smooth bundle homomorphism $D_\scrm: \scre_\scrm \to \scre'_\scrm$

1822: which lift to the corresponding objects over $\whcalm$. Later on we drop the

1823: sub-index $_\scrm$, so that, \eg $\scre$ will denote also the corresponding 

1824: bundle over $\scrm$.

1825: 

1826: \state 2. Our main interest is the space $\scrm$. However, in the proofs below

1827: we shall mostly work with $\whcalm$. The reason is that an element $(u, J_S,

1828: J) \in \whcalm$ fixes a parameterization of a pseudoholomorphic curve, whereas

1829: $[u,J] \in \scrm$ defines only an appropriate equivalence class of

1830: parameterizations.

1831: 

1832: 

1833: 

1834: 

1835: \proof Part \slip.

1836: 

1837: {\sl Case $g\ge2$}. It is known that in this case $\bfg$ acts properly

1838: discontinuously on $\ttt_g$. This implies that the same is true for the action

1839: of $\bfg$ on $\whcalm$. Moreover, it is clear that $\bfg$ acts freely on $\whcalm$.

1840: %Moreover, the definition of $\whcalm$ provides that $\bfg$ acts freely on 

1841: %this space. 

1842: Consequently, the map $\whcalm\lrar \scrm = \whcalm/\bfg$ is simply an (unbranched)

1843: covering.

1844: 

1845: \smallskip

1846: {\sl Case $g=0$}. In this case $S=S^2$, $\ttt_0=\{J\st\}$, and the action of

1847: $\bfg$ on $S$ is generated by holomorphic vector fields, \ie by the space

1848: $\sfh^0(S,TS)$. One can show that the action of $\bfg$ on

1849: $\whcalm$ is generated by vector fields

1850: \begin{equation}

1851: (u, J\st,J)\in \whcalm \mapsto

1852: (du(v), 0,0) \in T_{(u, J\st, J)}\whcalm

1853: \quad\hbox{with $v\in \sfh^0(S,TS)$ fixed}.

1854: \end{equation}

1855: In particular, the action is {\sl smooth} or, more precisely, $C^\ell$-smooth.

1856: 

1857: Consequently, for a given $(u^0,J\st,J^0)\in \whcalm$ we can find a closed

1858: complementing space $\scrv \subset T_{(u^0,J\st,J^0)}\whcalm$ to $(u^0 (\sfh

1859: ^0(S,TS)),0,0)$. Represent it as the tangent space of a submanifold $\scrw

1860: \subset \whcalm$ through $(u^0,J\st,J^0)$, $T_{(u^0,J\st,J^0)} \scrw=\scrv$.

1861: If $\scrv$ is chosen sufficiently small, then it intersects every orbit $\bfg

1862: \cdot (u,J\st,J)$ transversally in exactly one point. Moreover, we have a

1863: $\bfg$-invariant diffeomorphism $\bfg\cdot \scrw \cong \bfg

1864: \times \scrw$, so that $\scrw$ is a local slice of $\bfg$-action at $(u^0,

1865: J\st,J^0)$. This equips the quotient $\whcalm/\bfg$ with a structure

1866: of a smooth Banach manifold such that the projection $\scrw \to \scrm=

1867: \whcalm/\bfg$ is a $C^\ell$-smooth chart.

1868: 

1869: \smallskip

1870: {\sl Case $g=1$} is a combination of the above two cases. First we consider

1871: the action of $T^2 \triangleleft \bfg$. The existence of a local $T^2$-slice

1872: $\scrw$ through any given $(u^0,J_S^0, J^0)\in \whcalm$ can be

1873: shown by copying the construction of {\sl Case $g=0$}. This implies that

1874: $\whcalm \lrar \whcalm/T^2$ is a principle $T^2$-bundle. Then we repeat

1875: the argument of {\sl Case $g\ge2$} and show that $\whcalm/T^2 \to

1876: \whcalm/\bfg$ is an unbranched covering with the group $\mathbf{Sp}(2,\zz) =

1877: \bfg/ T^2$. This completes the proof of part \slip.

1878: 

1879: \medskip {\sl Part} \sliip.

1880: The action of $\bfg$ extends in a natural way to an action on

1881: $\scrz\deff S \times \scrs \times \ttt \times \scrj$. The evaluation map

1882: $\ev: \scrz \to X$, $\ev(z, u, J_S, J) \deff u(z)$ is $\bfg$-equivariant. 

1883: Consequently, the bundle $E\deff \ev^* TX$ over $\scrz$ is equipped

1884: with the natural $\bfg$-action. The action of $\bfg$ on $E$ induces the

1885: actions on section spaces $\scre$ and $\scre'$. Since all constructions are

1886: natural, $D: \scre \to \scre'$ is $\bfg$-invariant.

1887: 

1888: Finally, it remains to note that the action of $\bfg$ on the bundles $\scre$

1889: and $\scre'$ over $\whcalm$ is $C^\ell$-smooth.

1890: 

1891: \smallskip\noindent

1892: {\sl Part} \sliii of the lemma states the universality property of

1893: $\scrm$ which easily follows from the definitions.

1894: \qed

1895: 

1896: \newcorol{cor2.2.2} $\scrm$ is a $C^\ell$-smooth Banach

1897: manifold and $\pi_{\!\!\scrj}:\scrm \to \scrj$ is a Fredholm map. For

1898: $[u,J]\in\scrm$ there exist natural isomorphisms

1899: %\begin{equation}

1900: \begin{align*}

1901: \ker(d\pi_{\!\!\scrj}:T_{[u,J]}\scrm \to T_J\scrj)&

1902: \;\cong\; \sfh^0_D(S, \scrn_u),

1903: \\

1904: \coker(d\pi_{\!\!\scrj}:T_{[u,J]} \scrm \to T_J\scrj)

1905: &\;\cong\; \sfh^1_D(S, \scrn_u).

1906: \end{align*}

1907: %\end{equation}

1908: In particular, the index of $\pi_{\!\!\scrj}$ is equal to

1909: \begin{equation}

1910: \ind_\rr \pi_{\!\!\scrj} = \chi_\rr(\scrn_u)= 2\bigl(\mu + (n-3)(1-g) \bigr),

1911: \eqqno(2.2.3)

1912: \end{equation}

1913: where $\mu = \la c_1(X), [C] \ra$.

1914: \end{corol}

1915: 

1916: \proof The $C^\ell$-smooth structure on $\scrm$ is is the quotient structure 

1917: defined by the the $C^\ell$-smooth $\bfg$-action on $\whcalm$.

1918: %provided by the fact that the action on $\bfg$ on $\whcalm$ is $C^\ell$-smooth.

1919: 

1920: \smallskip

1921: Using \refcorol{cor2.1.3} we see that the tangent space

1922: to $\whcalm$ is

1923: \begin{equation}

1924: T_{(u,J_S,J)} \whcalm = \bigl\{ (v,\dot J_S, \dot J) \,:\,

1925: \dot J_S \in T_{J_S}\ttt,\;\,

1926: D_{u,J}v + J \scirc du \scirc \dot J_S + \dot J \scirc du \scirc J_S

1927: =0 \,\bigr\}.

1928: \eqqno(2.2.4)

1929: \end{equation}

1930: Consider the natural projection $\pi_\scrp: \scrp^* \to \scrj$,  $(u,J_S,J)

1931: \mapsto J$ with the differential $d\pi_\scrp: T_{(u,J_S,J)}\scrp^* \to

1932: T_J\scrj$ given by $(v,\dot J_S,\dot J)\in T_{(u,J_S,J)}\scrp^*\mapsto

1933: \dot J\in  T_J\scrj$.

1934: 

1935: The kernel $\ker(d\pi_\scrp)$ consists of solutions $v\in \scre_{(u,J)}$

1936: of the equation

1937: \begin{equation}

1938: D_{u,J}v +  J\scirc du \scirc \dot J_S =0

1939: \end{equation}

1940: with $\dot J_S\in T_{J_S}\ttt$. Since the map

1941: $\wh\pi: \wh\scrm\to \scrm$ is a principle $\bfg$-bundle, the kernel

1942: $\ker( d\pi_{\!\!\scrj}:T_{(M,J)}\scrm \to T_J\scrj)$ is obtained from $\ker(d\pi)$

1943: by taking the quotient by the tangent space to the fiber $\bfg

1944: \cdot (u,J_S,J)$ which is equal to $du(\sfh^0(S,TS))$. Using the relations

1945: $\sfh^0(S,TS)= \ker(\dbar_{TS}:L^{1,p}(S,TS)\to L^p(S,TS\otimes

1946: \Lambda^{(0,1)}S)$, $T_{J_S}\ttt_g\cong\sfh^1(S,TS)=\coker(\dbar_{TS})$, and

1947: $du\scirc \dbar_{TS}= D_{(u,J)}\scirc du$, we conclude that the space

1948: $\ker(d\pi_{\!\!\scrj})$ is isomorphic to the quotient

1949: \begin{equation}

1950: \{v\in L^{1,p}(S,E_u)\;:\; Dv=du(\phi) \hbox{ for some }\phi\in

1951: L^p(S,TS\otimes\Lambda^{(0,1)}S)\}

1952: \!\!\bigm/\!\!

1953: du\bigl(L^{1,p}(S,TS)\bigr).

1954: \end{equation}

1955: Hence, by \lemma{lem1.5.3}, $\ker(d\pi_{\!\!\scrj}:T_{[u,J]}\scrm\to T_J\scrj)

1956: \cong \sfh^0_D(S,\scrn_u)$. In particular, $\ker(d\pi_{\!\!\scrj})$ is finite 

1957: dimensional.

1958: 

1959: \smallskip

1960: Similarly, the image of $d\pi_{\!\!\scrj}$ consists of those $\dot J$ for which

1961: the equation

1962: \begin{equation}

1963: D_{u,J}v + J\scirc du \scirc \dot J_S + \dot J\scirc

1964: du \scirc J_S =0

1965: \end{equation}

1966: has a solution $(v,\dot J_S)$. Using \eqqref(1.5.6) and \lemma{lem1.5.3}

1967: we obtain the relations $\im(d\pi_{\!\!\scrj})=\ker\barr\Psi$ and

1968: $\coker(d\pi) \cong \sfh^1_D(S,N_u)$.

1969: 

1970: This implies the Fredholm property for the projection $\pi_{\!\!\scrj}:\scrm \to

1971: \scrj$ and the formula $\ind(d\pi_{\!\!\scrj}) = \ind (\scrn_u)$.

1972: \qed

1973: 

1974: \medskip

1975: We conclude the paragraph with a description of the deformations of non-closed 

1976: pseudoholomorphic curves. 

1977: 

1978: \newdefi{def2.2.2} Let $\barr S=S \cup \d S$ be a compact {\sl non-closed} 

1979: oriented surface with the boundary $\d S$ consisting of finitely many circles.

1980: Denote by $\scrj_S$ the (Banach) space of complex structure on $S$ which are 

1981: compatible with the orientation of $S$ and $C^\ell$-smooth up to boundary 

1982: $\d S$. As usual let 

1983: \begin{equation}

1984: \eqqno(2.2.11)

1985: \scrp(S, X) \deff \{ (u,J_S,J) \in L^{1,p}(S, X) \times \scrj_S \times 

1986: \scrj: \dbar_{J_S,J} u=0\},

1987: \end{equation}

1988: the space of pseudoholomorphic maps equipped it with the natural projections

1989: $\pr_{\scrj_S}: \scrp(S, X) \to \scrj_S$ and $\pr_{\!\!\scrj}: \scrp(S, X) \to 

1990: \scrj$. The fibers of the projections are denoted by $\scrp(S, X, J)=\pr_{\!\!\scrj}

1991: \inv(J)$, $\scrp(S,J_S, X)=\pr_{\scrj_S}\inv(J_S)$, and $\scrp(S,J_S, X,J)= 

1992: \scrp(S,J_S, X) \cap \scrp(S, X, J)$ respectively.

1993: \end{defi}

1994: 

1995: \newlemma{lem2.2.4} \sli Let $S$ be a non-closed oriented surface. Then

1996: 

1997: \sli the space $\scrp(S, X)$ is a Banach submanifolds of $L^{1,p}(S, X) 

1998: \times \scrj_S\times \scrj$;

1999: 

2000: \slii For any $(u,J_S, J)\in \scrp(S, X)$, the operators $d\pr_{\!\!\scrj}: T_{(u,

2001: J_S,J)}\scrp(S, X) \to T_J\scrj$ and $d\pr_{\scrj_S}: T_{(u,J_S,J)} \scrp(S, 

2002: X) \to T_{J_S}\scrj$ are surjective and split. In particular, $\scrp(S; X,J)$ 

2003: and $\scrp(S,J_S; X)$ are are Banach submanifolds of $\scrp(S, X)$;

2004: 

2005: \sliii For any $(u,J_S, J)\in \scrp(S, X)$, the submanifolds $\scrp(S; X,J)$ 

2006: and $\scrp(S,J_S; X)$ are transversal in $(u,J_S,J)$; in particular, $\scrp(S, J_S; 

2007: X,J)= \scrp(S; X,J) \cap \scrp(S,J_S; X)$ is also a Banach submanifold;

2008: 

2009: \sliv The the tangent spaces are given by

2010: \begin{align}

2011: T_{(u,J_S,J)}\scrp(S, X) &= \bigl\{ (v,\dot J_S,\dot J) \in T_uL^{1,p}(S, X)

2012: \times T_{J_S}\scrj_S \times T_J\scrj : 

2013: \notag\\

2014: &\qquad\quad D_{u,J}v + \dot J \scirc du \scirc J_S + 

2015: J \scirc du \scirc \dot J_S =0\bigr\};

2016: \eqqno(2.2.12)\\

2017: T_{(u,J_S)}\scrp(S; X,J)& = \bigl\{ (v,\dot J_S,\dot J) \in 

2018: T_{(u,J_S,J)}\scrp(S, X): \dot J=0 \bigr\};

2019: \eqqno(2.2.13)\\

2020: T_{(u,J)}\scrp(S,J_S; X)& = \bigl\{ (v,\dot J_S,\dot J) \in 

2021: T_{(u,J_S,J)}\scrp(S, X): \dot J_S=0 \bigr\};

2022: \eqqno(1.2.14)\\ 

2023: T_u\scrp(S,J_S; X,J)& = \bigl\{ (v,\dot J_S,\dot J) \in 

2024: T_{(u,J_S,J)}\scrp(S, X): \dot J_S=0 = \dot J \bigr\}.

2025: \eqqno(2.2.15) 

2026: \end{align}

2027: \end{lem}

2028: 

2029: \proof The lemma is obtained by the transversality techniques of 

2030: \refsubsection{2.1} using the following claim: {\sl For any $(u,J_S,J) \in 

2031: \scrp(S,X)$ the operator $D_{u,J} : L^{1,p}(C, E_u) \to L^p_{(0,1)}(C, E_u)$ 

2032: is surjective}. Since $D_{u,J}$ is elliptic, this is a standard fact following 

2033: from compactness and non-closedness of $S$. \qed

2034: 

2035: 

2036: \newsubsection[2.3]{Transversality I$\!$I}

2037: Before stating further results, we introduce some new notation. Here $S$ is a 

2038: {\sl closed} real surface.

2039: 

2040: \newdefi{def2.3.1}

2041: Let $Y$ be a $C^\ell$-smooth finite-dimen\-sional manifold, possibly with

2042: non-empty $C^\ell$-smooth boundary $\d Y$, and $h:Y \to \scrj$ a $C^\ell

2043: $-smooth map. Define the {\sl relative Moduli space}

2044: \begin{equation}

2045: \scrm_h \deff Y\times_{\!\!\scrj} \scrm \cong

2046: \{\,(u,J_S,y)\in \scrs\times \ttt_g\times Y\,:\, (u,J_S, h(y))\in \scrp^*

2047: \,\}/\bfg

2048: \end{equation}

2049: with the natural projection $\pi_h:\scrm_h \to Y$. In the special case

2050: $Y=\{J\}\hook \scrj$, we obtain the Moduli space of $J$-holomorphic curves

2051: $\scrm_J\deff \pi_{\!\!\scrj}\inv(J)$. The projection $\pi_h:\scrm_h \to Y$

2052: is a fibration with a fiber $\pi_h\inv(y)=\scrm_{h(y)}$. We shall denote

2053: elements of $\scrm_h$ by $[u,y]$, where $u:S\to X$ is a $h(y)$-holomorphic map.

2054: \end{defi}

2055: 

2056: \medskip

2057: The next two lemmas follow from the transversality theory.

2058: 

2059: \newlemma{lem2.3.1}

2060: Let $Y$ be a $C^\ell$-smooth finite-dimen\-sional

2061: manifold, and $h:Y \to \scrj$ a $C^\ell$-smooth map. Then $\scrm_h$ is a

2062: $C^\ell$-smooth manifold in some neighborhood of a point $[u,y] \in

2063: \scrm_h$ with $J \deff h(y)$ \iff the map $\barr\Psi_{u, J} \scirc dh

2064: : T_uY \to \sfh^1_D(S, N_u)$ is surjective.

2065: In this case the tangent space to $\scrm_h$ is

2066: \begin{equation}

2067: T_{[u,y]}\scrm_h=\ker\bigl(D\;\oplus\; \Psi\scirc dh:

2068: \scre_{u,h(y)}\oplus T_yY \lrar \scre'_{u,h(y)} \bigr)

2069: \bigm/ du(\sfh^0(S, TS))

2070: \eqqno(2.3.1)

2071: \end{equation}

2072: \end{lem}

2073: 

2074: 

2075: \proof We reformulate the transversality condition and

2076: use \lemma{lem2.1.1}.

2077: \qed

2078: 

2079: \newlemma{lem2.3.2}

2080: \sli There exists a Baire subset $\scrj\reg \subset

2081: \scrj$ such that any $J \in \scrj$ is a regular value of $\pi_{\!\!\scrj} :\scrm

2082: \to \scrj$.

2083: 

2084: \slii There exists a Baire subset $\scrv$ in the space $C^\ell([0,1], \;

2085: \scrj)$, such that any map $h:[0,1]\to \scrj$ from $\scrv$ is transversal to

2086: $\pi_{\!\!\scrj} :\scrm \to \scrj$ and both $h(0)$ and $h(1)$ are

2087: regular values of $\pi_{\!\!\scrj}$.

2088: \end{lem}

2089: 

2090: 

2091: \state Remark. In general, for any finite-dimen\-sional manifold $Y$ with

2092: boundary $\d Y$ there exists a Baire subset $\scrv \subset C^\ell(Y,\; \scrj)$

2093: such that any $h \in \scrv$, as well as its restriction $h\ogran_{\d Y}$ are

2094: transversal to $\pi_{\!\!\scrj}$. The proof uses the Sard lemma.

2095: 

2096: \newlemma{lem2.3.3}

2097: Suppose that $S$ is the sphere $S^2$ and $\dim_\rr(X) =4$.

2098: Then the exists a {\sl connected} Baire subset $\scrj\reg \subset

2099: \scrj$ such that any $J \in \scrj$ is a regular value of $\pi_{\!\!\scrj} :\scrm

2100: \to \scrj$. Moreover, any $J_0, J_1 \in \scrj\reg$ can be connected by

2101: a smooth path $h: [0,1] \to \scrj\reg$.

2102: \end{lem}

2103: 

2104: 

2105: \proof By \lemma{lem2.3.2}, there exists a Baire subset $\scrj\reg

2106: \subset\scrj$ such that any $J \in \scrj\reg$ is a regular value of

2107: $\pi_{\!\!\scrj}$. Further, any $J_0, J_1 \in \scrj\reg$ can be connected by a

2108: smooth path $h: [0,1] \to \scrj$, transversal to $\pi_{\!\!\scrj}$.

2109: 

2110: For any such path $h: [0,1] \to \scrj$ and any $[u,t] \in \scrm_h$ with

2111: $h(t)=J$ the map $\barr\Psi_{u,J}\scirc dh: T_t[0,1]\cong\rr \; \lrar

2112: \sfh^1(S, N_u)$ is surjective by \lemma{lem2.3.1}. Consequently,

2113: for such $u$ and $J$ we have $\dimr \sfh^1_D(S, \scrn_u)= \dimr \sfh^1_D(S,

2114: N_u) \le 1$.

2115: 

2116: Recall that the difference $\dimr\sfh^0_D(S, N_u) -\dimr\sfh^1_D(S, N_u)$ is

2117: {\sl even}, see \eqqref(1.5.?4). Hence, if $\dimr \sfh^1(S, N_u)= 1$ then

2118: $\sfh^1_D(S, N_u)$ should also be nontrivial. On the other hand, the condition

2119: $\dimr(X)=4$ implies that $N_u$ is a {\sl line bundle}. But, in view of {\sl

2120: Lemma 1.5.2}, on the sphere $S=S^2$ one of the spaces $\sfh^i_D(S, N_u)$ must

2121: be trivial.

2122: 

2123: Thus, we see that for any such path $h$ and any $[u,t]\in \scrm_h$ one has

2124: $\sfh^1_D(S, \scrn_u)=0$. This means that $h$ takes values in $\scrj\reg$.

2125: \qed

2126: 

2127: \smallskip

2128: For higher genus $g(S) \ge1$ we have a similar, but weaker result.

2129: 

2130: \newlemma{lem2.3.4}

2131: Assume that $\dim_\rr X =4$ and $g\deff g(S) \ge1$

2132: and set $\mu \deff c_1(X)[u(S)]$. Then

2133: \begin{equation}

2134: \mu  \le \dim_\cc \sfh^0(S, \scrn\sing_u) \le \mu + g-1

2135: \eqqno(2.3.2)

2136: \end{equation}

2137: for any $[u,J] \in \scrm$ with $\sfh^1(S, N_u)\cong \rr$.

2138: \end{lem}

2139: 

2140: 

2141: \proof The condition $\dim_\rr X =4$ means that $N_u$ is a {\sl line} bundle. 

2142: Thus, by \lemma{lem1.5.2},  $\sfh^1(S, N_u)\cong \rr$ implies $c_1(N_u) \le 2g 

2143: -2$. Further, since $\dim_\rr \sfh^1(S, N_u)=1$ and $\ind_\rr D^N_{u,J} = \dimr

2144: \sfh^0_D(S, N_u) -\dimr\sfh^1_D(S, N_u)$ is {\sl even}, see \eqqref(1.5.?4), we 

2145: conclude that $\dimr\sfh^0_D(S, N_u) \ge 1$. Consequently, $\ind_\rr D^N_{u,J} 

2146: \ge 0$. The formula \eqqref(1.5.?4) for $\ind_\rr D^N_{u,J}$ yields $c_1(N_u) 

2147: \ge g-1$. Finally, the definition of $\scrn_u$ yields the relation $\mu= c_1(X)

2148:  [u(S)] = c_1(E_u) = c_1(TS) + c_1(N_u) + \dimc \sfh^0(S, \scrn\sing_u)$. \qed

2149: 

2150: \newsubsection[2.4]{Pseudoholomorphic curves through fixed points} In this 

2151: paragraph we consider the total moduli space of pseudoholomorphic curves passing 

2152: through given fixed points $x_1, \ldots, x_m \allowbreak 

2153: \in X$. Gromov in \cite{Gro} proposed a method to reduce 

2154: this problem to the one of pseudoholomorphic curves

2155: without such constraints. The idea is to blow up $X$ in the points $x_1, \ldots, 

2156: x_m$ and consider the curves on the blown-up space $\ti X$. He has shown that a 

2157: $C^\ell$-smooth almost complex structure $J$ on $X$ lifts to a $C^{\ell-1}$-smooth 

2158: almost complex structure $\ti J$ on $\ti X$ such that the natural

2159: projection $\pr: \ti X \to X$ is holomorphic and such that every $J$-holomorphic 

2160: curve $C$ in $X$ passing through $x_1, \ldots, x_m$ lifts to a unique $\ti 

2161: J$-holomorphic curve $\ti C$ in $\ti X$ with $C= \pr(\ti C)$. 

2162: 

2163: 

2164: 

2165: \smallskip

2166: Our aim here is to make an explicit construction for the moduli space of 

2167: pseudoholomorphic curves passing through fixed points. Since the construction is 

2168: simply a modification of the case $m=0$ where no points are marked, we shall 

2169: mostly skip or merely indicate proof of claims.

2170: 

2171: We begin by introducing some notation. Denote by $\mbfx=(x_1, \ldots x_m)$ the

2172: tuple of fixed points on $X$, which are supposed to be pairwise distinct. 

2173: Also fix a tuple $\mbfz=(z_1, \ldots, z_m)$ of pairwise distinct points on the

2174: surface $S$. Define 

2175: $$

2176: \scrs(\mbfz, \mbfx) \deff \{ u \in \scrs=L^{1,p}(S, X): u(z_i)=x_i \};

2177: $$

2178: $$

2179: \scrp(S,\mbfz; X, \mbfx) \deff \{ (u, J_S, J) \in \scrp: u\in \scrs(\mbfz, 

2180: \mbfx) \};

2181: $$

2182: $$

2183: \scrp^*(S,\mbfz; X, \mbfx) \deff  \scrp(S,\mbfz; X, \mbfx) \cap \scrp^*(S,X).

2184: $$

2185: The linearization of the conditions $u(z_i)=x_i$ yields the equations $v(z_i)=0$

2186: for $v \in T_u\scrs= L^{1,p}(S, E_u)$. Denote as above $E=E_u \deff u^*TX$, and

2187: set $E_i= E_{u,i} \deff (E_u)_{z_i} = T_{u(z_i)}X$. Then we obtain the bundle

2188: $E_\mbfz$ over $\scrs$ with a fiber $(E_\mbfz)_u\deff \oplus E_{u,i}$ equipped with

2189: the natural {\sl evaluation homomorphism} $\ev_\mbfz: \scre \to E_\mbfz$

2190: $$

2191: \ev_{\mbfz}: v \in \scre_u=L^{1,p}(S, E) \mapsto (v(z_1), \ldots, v(z_m)) \in 

2192: E_\mbfz.

2193: $$

2194: It is easy to see that $\ev_\mbfz: \scre \to E_\mbfz$ is surjective. This means

2195: that the equations $u(z_i)=x_i$ are transversal and implies that $\scrs(\mbfz, 

2196: \mbfx)$ is a Banach submanifold of $\scrs$ with the tangent space

2197: \begin{equation}\eqqno(2.4.1)

2198: T_u\scrs(\mbfz, \mbfx) = \{ v \in T_u\scrs= L^{1,p}(S, E_u): v(z_i)=0 \}.

2199: \end{equation}

2200: The same argument shows that $\scrp^*(S,\mbfz; X, \mbfx)$ is also a Banach 

2201: submanifold of $\scrp^*(S,X)$ with the tangent space

2202: $$

2203: T_u\scrp^*(S,\mbfz; X, \mbfx) =  \{ (v, \dot J_S, \dot J) \in  

2204: T_u\scrp^*(S,X): v(z_i)=0 \}.

2205: $$

2206: 

2207: \smallskip

2208: Let $\diff_+(S, \mbfz)$ be the subgroup of those $g\in \diff_+(S)$ which fix the 

2209: marked points $z_1,\ldots, z_m$. Then $\diff_+(S, \mbfz)$ leaves the subsets 

2210: $\scrs(\mbfz, \mbfx) \subset \scrs$ and $\scrp^*(S,\mbfz; X, \mbfx) \subset 

2211: \scrp(S,X)$ invariant. So we can define the {\sl total moduli space of 

2212: pseudoholomorphic curves through the given points $x_1, \ldots, x_m$} as the 

2213: quotient $\scrm(\mbfx) \deff \scrp^*(S,\mbfz; X, \mbfx) /\diff_+(S, \mbfz)$. This

2214: space is equipped with the natural projection $\pi_\scrj: \scrm(\mbfx) \to \scrj$ 

2215: defined in an obvious way.

2216: 

2217: \smallskip

2218: The smooth structure on $\scrm(\mbfx)$ is constructed in the same way as it was 

2219: for $\scrm$. First, one constructs a global slice on the action of $\diff_+

2220:  (S, \mbfz)$ on $\scrj_S$. To do this, we consider the action of the component 

2221: of $\diff_0(S, \mbfz)$ the group $\diff_+(S, \mbfz)$ containing the identity. The 

2222: quotient $\scrj_S/ \diff_0(S, \mbfz)$ is the {\sl Teichm\"uller space $\ttt

2223: _{g,m}$ of complex structures on a Riemann surface of genus $g=g(S)$ with $m$ 

2224: punctures}. The marked points $z_1,\ldots, z_m$ are the positions of punctures. 

2225: 

2226: As in the case $m=0$, one can imbed $\ttt_{g,m}$ in $\scrj_S$ in such a way that

2227: the composition $\ttt_{g,m} \hook \scrj_S \twoheadrightarrow \scrj_S/ \diff_0(S, 

2228: \mbfz) \cong \ttt_{g,m}$ is the identity map. This imbedding $\ttt_{g,m} \hook 

2229: \scrj_S$ is the desired slice. The choice of such imbedding $\ttt_{g,m} \hook 

2230: \scrj_S$ is equivalent to the choice of the complex structure $J_{\ttt, S, \mbfz}$

2231: on the product $\ttt_{g,m} \times S$. One considers $\ttt_{g,m} \times S$ with

2232: this complex structure and with the holomorphic projection $\pr: \ttt_{g,m} 

2233: \times S \to \ttt_{g,m}$ as the universal family corresponding to $\ttt_{g,m}$.

2234: 

2235: 

2236: After the choice of the slice $\ttt_{g,m} \hook \scrj_S$, $\scrm(\mbfx)$ is

2237: obtained as the quotient of the space 

2238: $$

2239: \wh\scrm(\mbfx) \deff \{ (u, J_S, J) \in \scrp^*(S,\mbfz; X, \mbfx):

2240: J_S \in \ttt_{g,m} \}

2241: $$

2242: by the group $\bfg$ of biholomorphisms of $\ttt_{g,m} \times S$ preserving the 

2243: projection $\pr: \ttt_{g,m} \times S \to \ttt_{g,m}$. It is discrete except the

2244: cases where $g=0$ and $m=1$ or $2$. In the case $m=1$, $S\bs \{z_1\}$ is the 

2245: complex plane $\cc$ and $\bfg$ is its automorphism group $\cc^* \ltimes \cc$. 

2246: Similarly, in  the case $m=2$, $S\bs \{z_1, z_2\}$ is the punctured complex plane

2247: $\cc^*$ and $\bfg= \zz_2 \ltimes \cc^*$ is likewise its automorphism group. In 

2248: either case one can construct a local slice for the action of $\bfg$ on $\wh\scrm

2249:  (\mbfx)$ by repeating the arguments of \refsubsection{2.2}. So the quotient 

2250: $\whcalm(\mbfx)/\bfg$ is a $C^\ell$-smooth Banach manifold.

2251: 

2252: \medskip

2253: Now we define the notion of the normal sheaf of a pseudoholomorphic curve passing 

2254: through fixed points on $X$. In this new situation, the linearization of 

2255: $\dbar$-equations leads to the operator

2256: \begin{equation}\eqqno(2.4.1a)

2257: D=D_{u,J}: \{ v \in L^{1,p}(S,E_u): v(z_i)=0 \text{ for }i=1,\ldots,m\} 

2258: \to L^p_{(0,1)}(S, E),

2259: \end{equation}

2260: which is the usual Gromov operator $D=D_{u,J}$, but now considered with 

2261: a new domain of definition

2262: \begin{equation*}

2263: \scre_{u,\mbfx} \deff \{ v \in L^{1,p}(S,E_u): v(z_i)=0  

2264: \text{ for }i=1,\ldots,m\}.

2265: \end{equation*}

2266: The space $\scre_{u,\mbfx}$ is the kernel of the evaluation homomorphism

2267: $\ev_z: \scre_u \to E_\mbfz$ and is the tangent plane to $\scrs(\mbfz, \mbfx)$,

2268: see \eqqref(2.4.1). 

2269: 

2270: We now describe the structure of the operator \eqqref(2.4.1a). Recall that we have

2271: the 

2272: decomposition $D_{u,J}= \dbar_{u,J} + R_{u,J}$, see \refsubsection{1.4}. Observe

2273: that the sheaf $\scro(E_u)[-\mbfz]$ of holomorphic sections of $\scro(E_u)$

2274: vanishing at the points $z_1, \ldots, z_m \in S$ is locally free and hence 

2275: corresponds to a holomorphic bundle. Let us denote this bundle by $E_{u, -\mbfz}$.

2276: 

2277: \newlemma{lem2.4.1} \sli The (co)kernel of the operator 

2278: \begin{equation}\eqqno(2.4.2a)

2279: \dbar_{u,J}: \{ v \in L^{1,p}(S,E_u): v(z_i)=0 \text{ for }i=1,\ldots,m\} 

2280: \to L^p_{(0,1)}(S, E),

2281: \end{equation}

2282: is canonically isomorphic to the cohomology groups $\sfh^0_\dbar(S, E_{u,-\mbfz})$

2283: and $\sfh^1_\dbar(S, E_{u,-\mbfz})$. 

2284: 

2285: \slii The operator $D_{u,J}$ induces the operator

2286: $$

2287: D_{u,-\mbfz, J}: L^{1,p}(S, E_{u,-\mbfz}) \to L^p_{(0,1)}(S, E_{u,-\mbfz})

2288: $$

2289: which is of the form $D_{u,-\mbfz, J}= \dbar_{u,-\mbfz, J} + R_{u,-\mbfz, J}$, where

2290: $\dbar_{u,-\mbfz, J}$ is the Cauchy-Riemann operator corresponding to the natural 

2291: holomorphic structure in $E_{u,-\mbfz}$ and $R_{u,-\mbfz, J}$ is a $\cc$-antilinear

2292: $L^\infty$-bounded bundle homomorphism, \ie 

2293: $$

2294: R_{u,-\mbfz, J} \in L^\infty\bigl(S, \barr\hom_\cc(E_{u,-\mbfz},

2295:  E_{u,-\mbfz} \otimes \Lambda^{(0,1)}S)\bigr).

2296: $$

2297: 

2298: \sliii The (co)kernel of the operator 

2299: \begin{equation}\eqqno(2.4.3a)

2300: D_{u,J}: \{ v \in L^{1,p}(S,E_u): v(z_i)=0 \text{ for }i=1,\ldots,m\} 

2301: \to L^p_{(0,1)}(S, E)

2302: \end{equation}

2303: is canonically isomorphic to the cohomology groups $\sfh^0_D(S, E_{u,-\mbfz})$

2304: and $\sfh^1_D(S, E_{u,-\mbfz})$ corresponding to the operator $D_{u,-\mbfz, J}$.

2305: \end{lem}

2306: 

2307: \proof Fix local holomorphic coordinates $\zeta_i$ on $S$, each centered at

2308: the corresponding marked point $z_i$. Consider the natural inclusions 

2309: \begin{align}

2310: \eqqno(2.4.4a)

2311: j^0:L^{1,p}(S, E_{u,-\mbfz}) &\hook 

2312: \{ v \in L^{1,p}(S,E_u): v(z_i)=0 \text{ for }i=1,\ldots,m\},

2313: \\

2314: \eqqno(2.4.5a)

2315: j^1: L^p_{(0,1)}(S, E_{u,-\mbfz}) &\hook L^p_{(0,1)}(S, E_u).

2316: \end{align}

2317: Observe that $v\in L^{1,p}(S,E_u)$ with $v(z_i)=0$ belongs to $L^p_{(0,1)}(S, 

2318: E_{u,-\mbfz})$ \iff locally near every $z_i$ it has the form $v(\zeta_i)= 

2319: \zeta_i w(\zeta_i)$ for some (uniquely defined!) $L^{1,p}$-section $w(\zeta_i)$ 

2320: of $E_{u,-\mbfz}$. This is equivalent to the condition $\zeta_i\inv \dbar_{u,J} 

2321: v(\zeta_i) \in L^p$ as well as to the condition $\zeta_i\inv D_{u,J} v(\zeta_i) 

2322: \in L^p$. Consequently, $D_{u,J}$ restricted to $L^{1,p}(S,E_{u,-\mbfz})$ takes 

2323: values in $L^p_{(0,1)}(S, E_{u,-\mbfz})$. This yields the operator $D_{u,-\mbfz, 

2324: J}$. Moreover,  $D_{u,-\mbfz, J}$ is of order 1 and has the Cauchy-Riemann symbol.

2325: Consequently, it has the form $D_{u,-\mbfz, J}= \dbar_{u,-\mbfz, J} +R_{u,-\mbfz,

2326: J}$, where $\dbar_{u,-\mbfz, J}$ is the Cauchy-Riemann operator corresponding

2327: to the holomorphic structure in $E_{u,-\mbfz}$, and $R_{u,-\mbfz, J}$ is the

2328: $\cc$-antilinear part of $D_{u,-\mbfz, J}$.

2329: 

2330: Let $v_1(\zeta_i), \ldots, v_n(\zeta_i)$ be a local holomorphic frame of $E_u$ in 

2331: a neighborhood of $z_i$ and $R_{\alpha\beta}(\zeta_i)$ the matrix of $R_{u,J}$ in 

2332: this frame. Then $\zeta_iv_1(\zeta_i), \ldots, \zeta_iv_n(\zeta_i)$ is a local 

2333: frame of $E_{u,-\mbfz}$. From $\cc$-antilinearity of $R_{u,J}$ we obtain

2334: $$

2335: \textstyle

2336: R_{u,J}(\zeta_i\,v_\alpha(\zeta_i))= 

2337: \sum_\beta R_{\alpha\beta}(\zeta_i)\bar \zeta_i v_\beta(\zeta_i)=

2338: \sum_\beta \msmall{\bar \zeta_i \over \zeta_i} R_{\alpha\beta}(\zeta_i)

2339: \cdot \zeta_i v_\beta(\zeta_i).

2340: $$

2341: This shows that ${\bar \zeta_i \over \zeta_i} R_{\alpha\beta}(\zeta_i)$ is the 

2342: matrix of $R_{u, -\mbfz, J}$ in the frame $\zeta_iv_1(\zeta_i), \ldots, \zeta_i

2343: v_n(\zeta_i)$. Recall that $R_{u,J}$ is a continuous bundle homomorphism (see 

2344: \lemma{lem1.4.1}, \slip). So we see that $R_{u, -\mbfz, J}$ is also continuous 

2345: outside the marked points $z_i$ and has singularities of the form ${\bar 

2346: \zeta_i \over \zeta_i} R_{u,J}$ at $z_i$. In particular, $R_{u, -\mbfz, J}$ is of

2347: type $L^\infty$, but is not continuous in general. 

2348: 

2349: 

2350: \smallskip

2351: The equality of the kernels of the operators in \sli and \slii with the

2352: corresponding 0-cohomology groups follows directly from the definition of the 

2353: operators $\dbar_{u,-\mbfz,J}$ and $D_{u,-\mbfz,J}$. The equality for 

2354: 1-cohomology groups will be shown only for the operator $D_{u,-\mbfz,J}$, the 

2355: other one is carried out in the same manner. So let $\phi \in L^p_{(0,1)}(S, E_{u, 

2356: -\mbfz})$. If $\phi = D_{u,-\mbfz,J}(v)$ for $v \in L^{1,p}(S, E_{u, -\mbfz})$,

2357: then $v \in L^{1,p}(S, E_u)$ and $j^1\phi = D_{u,J}(v)$, or more precisely $j^1

2358: \phi = D_{u,J}(j^0(v))$. This shows that the inclusion $j^1$ in \eqqref(2.4.5a) 

2359: induces a well-defined homomorphism from $\sfh^1_D(S, E_{u, -\mbfz})$ to the 

2360: cokernel of \eqqref(2.4.3a). Moreover, $\phi \in L^p_{(0,1)}(S, E_{u, -\mbfz})$ 

2361: induces the zero class in the cokernel of \eqqref(2.4.3a) \iff $j^1(\phi) = 

2362: D_{u,J}(v)$ for some $v \in L^{1,p}(S, E_u)$ with $v(z_i)=0$. But then locally

2363: $$

2364: D_{u,J}(\zeta_i\inv v(\zeta_i)) = 

2365: \zeta_i\inv D_{u,-\mbfz,J}(v(\zeta_i)) =

2366: \zeta_i\inv j^1 \phi(\zeta_i) \in L^p

2367: $$

2368: by the definition of the inclusion $E_{u, -\mbfz} \hook E_u$. This implies

2369: that $v \in L^{1,p}(S, E_{u-\mbfz})$. Thus the homomorphism induces by $j^1$

2370: is injective.

2371: 

2372: Further, for any $\phi \in L^p_{(0,1)}(S, E_u)$ there exists $v \in L^{1,p}(S, 

2373: E_u)$ which vanishes at all $z_i$ and solves the equation $\phi = D_{u,J}(v)$

2374: in a neighborhood of every $z_i$. Then $\phi - D_{u,J}(v)$ represents the same

2375: class in the cokernel of \eqqref(2.4.3a) and is of the form $\phi - D_{u,J}(v)

2376: = j^1(\psi)$ for some $\psi \in L^p_{(0,1)}(S, E_{u, -\mbfz})$. This finishes

2377: the proof of the claim \sliiip. \qed

2378: 

2379: 

2380: \medskip

2381: Now we define the normal sheaf of a pseudoholomorphic curve passing through fixed 

2382: points. The construction is completely analogous to that in the  case of no fixed

2383: points. Here, instead of the tangent bundle $TS$ we use the bundle related 

2384: the new situation. This is the bundle $TS_{-\mbfz}$ associated to the locally

2385: free coherent sheaf $\scro(TS)[-\mbfz]$ of local holomorphic sections of $TS$ 

2386: vanishing at the points $z_i$. One can prove the analog of \lemma{lem2.4.1}

2387: for $TS_{-\mbfz}$. Observe however, that such a result follows immediately

2388: from that lemma if we set $X=S$ and $u=\id_S$.

2389: 

2390: As in \refsubsection{1.5}, we obtain the sheaf homomorphism $du: \scro(TS

2391: _{-\mbfz}) \to \scro(E_{u, -\mbfz})$, which is injective for non-constant $u:

2392: S \to X$. Now, the {\sl normal sheaf to curve $C=u(S)$ passing through the 

2393: points $\mbfx=(x_1,\ldots, x_m)$} is defined as the quotient $\scrn_{u,\mbfx} 

2394: \deff \scro(E_{u, -\mbfz}) / du\bigl(\scro(TS_{-\mbfz}) \bigr)$ together with

2395: the exact sequence

2396: \begin{equation}\eqqno(2.4.7)

2397: 0 \lrar \scro(TS_{-\mbfz}) \buildrel du \over \lrar \scro(E_{u, -\mbfz}) 

2398: \lrar \scrn_{u,\mbfx} \lrar 0.

2399: \end{equation}

2400: The sheaf $\scrn_{u,\mbfx}$ can be decomposed into its {\sl regular part} $\scrn

2401: \reg_{u,\mbfx}$ and its {\sl singular part} $\scrn\sing_{u,\mbfx}$, where 

2402: $\scrn \reg_{u,\mbfx}$ is locally free and $\scrn\sing_{u, \mbfx}$ is a torsion 

2403: sheaf. Then $\scrn \reg_{u,\mbfx}$ is a sheaf of local holomorphic sections of 

2404: the {\sl normal bundle $N_{u,\mbfx}$ to curve $C=u(S)$ passing through the points

2405: $\mbfx=(x_1,\ldots, x_m)$}, so that $\scrn\reg_{u,\mbfx}= \scro(N_{u,\mbfx})$.

2406: 

2407: As in \refsubsection{1.5}, we also obtain the exact sequence

2408: \begin{equation}\eqqno(2.4.8)

2409: 0 \lrar \scro(TS_{-\mbfz}) \otimes \scro([A])

2410: \buildrel du \over \lrar \scro(E_{u, -\mbfz}) 

2411: \scro(N_{u,\mbfx})

2412: \lrar 0.

2413: \end{equation}

2414: where $[A]$ is the branching divisor of $du$ (see \refdefi{def1.5.1}). This

2415: implies that the regular part $\scro(N_{u,\mbfx})$ is the quotient

2416: $$

2417: \scro(N_{u,\mbfx})= 

2418: \scro(E_{u,-\mbfz}) / du\bigl(\scro(TS_{-\mbfz}) \otimes \scro([A])\bigr).

2419: $$

2420: From the definition of $E_{u,-\mbfz}$ and $TS_{-\mbfz}$ we obtain the isomorphism

2421: $$

2422: \scro(N_{u,\mbfx})\cong \scro(N_u) \otimes \scro([A]).

2423: $$

2424: On the other hand, the singular part remains the same as is the case without 

2425: constraints:

2426: $$

2427: \scrn\sing_{u,\mbfx}\cong \scrn\sing_u \cong \scro /\scro(-[A]).

2428: $$

2429: 

2430: Further, we observe that the operators $\dbar$ on $TS_{-\mbfz}$ and $D_{u,-\mbfz,

2431: J}$ in $E_{u,-\mbfz}$ commute with the homomorphism $du: TS_{-\mbfz} \to E_{u,

2432: -\mbfz}$. Consequently, $D_{u,-\mbfz, J}$ induces the operator 

2433: $$

2434: D^N_{u,-\mbfz, J}: L^{1,p}(S, N_{u, -\mbfx}) \to L^p_{0,1)}(S, N_{u, -\mbfx}) 

2435: $$

2436: with the properties similar to ones of \eqqref(1.5.3). Further, as in 

2437: \lemma{lem1.5.3} and \refcorol{cor1.5.4} we obtain a long exact sequence of

2438: $D$-cohomologies. 

2439: 

2440: \newprop{prop2.4.2} The~short exact sequence $\eqqref(2.4.7)$ induces

2441: the~long exact sequence of $D$-cohomologies

2442: \begin{equation*}

2443: \def\normalbaselines{\baselineskip20pt\lineskip3pt \lineskiplimit3pt }

2444: \def\mapright#1{\smash{\mathop{\longrightarrow}\limits^{#1}}}

2445: \def\mapdown{\Big\downarrow}

2446: %

2447: \matrix%\format\c&\c&\c&\c&\c&\c&\l&\c&\l\\

2448: 0& \mapright{}& \sfh^0(S, TS_{-\mbfz}) &\mapright{} 

2449:  & \sfh^0_D(S, E_{u,-\mbfz})

2450:  & \mapright{}& \sfh^0_D(S, N_{u,-\mbfz})\oplus \sfh^0(S, \scrn_u\sing)

2451:  &\mapright{\delta} &\vphantom{\mapdown}\\

2452: %

2453:  & \mapright{}& \sfh^1(S, TS_{-\mbfz}) &\mapright{} 

2454:  & \sfh^1_D(S, E_{u,-\mbfz})

2455:  & \mapright{}& \sfh^1_D(S, N_{u,-\mbfz})         &\mapright{} &0.

2456: \endmatrix

2457: \end{equation*}

2458: \end{prop}

2459: 

2460: \medskip

2461: Finally, we note that the results of {\sl Paragraphs \ref{sec:2.2}\/} and

2462: {\sl\ref{sec:2.3}\/} remain valid, after an appropriate modification,

2463: also for curves passing through fixed points. We state without the proof 

2464: the summary of results which will be used later.

2465: 

2466: \newthm{thm2.4.3} \sli The total moduli space $\scrm_\mbfx$ of pseudoholomorphic 

2467: curves in a given homology class $[C] \in \sfh_2(X, \zz)$ passing through fixed 

2468: pairwise distinct points $\mbfx=(x_1, \ldots, \allowbreak

2469: x_m)$ on $X$ is a $C^\ell$-smooth 

2470: Banach submanifold of $\scrm$ of real codimension $2m$. In particular, the 

2471: projection $\pi_{\!\!\scrj}: \scrm_\mbfx \to \scrj$ is a $C^\ell$-smooth Fredholm 

2472: map of index 

2473: $$

2474: 2(c_1(X)[C] + (n-3)(1-g) -m).

2475: $$

2476: 

2477: \slii For a generic $J \in \scrj$ and a generic $C^\ell$-smooth path $h:[0,1] \to

2478: \scrj$ the fiber

2479: $$

2480: \scrm_{J,\mbfx} \deff \pi_\scrj\inv(J)

2481: $$

2482: and the {\sl relative moduli space}

2483: $$

2484: \scrm_{h,\mbfx} \deff [0,1] \times_\scrj \scrm_\mbfx 

2485: $$

2486: are $C^\ell$-smooth manifolds of expected dimension $2(c_1(X)[C] + (n-3)(1-g) -m)$

2487: and $2(c_1(X)[C] + (n-3)(1-g) -m) +1$ respectively.

2488: \end{thm}

2489: 

2490: 

2491: 

2492: \medskip

2493: \newsection[4]{Cusp-curves in the moduli space.}

2494: In this section we study the problem of deformation of pseudoholomorphic curves

2495: with prescribed singularities and develop the techniques required for controlling 

2496: their singularities under deformation. As the main result of this section we show 

2497: that the locus of pseudoholomorphic curves with a prescribed type of singularity 

2498: is a smooth 

2499: Banach submanifold of expected codimension in the total moduli space of

2500: pseudoholomorphic curves. This improvement of the result of 

2501: Micallef and White (see \lemma{lem1.2.1}) plays a crucial role below in 

2502: \refsection{3} in the proof of the {\sl saddle point property}.

2503: 

2504: Recall that our moduli space $\scrm$ consists of parameterized non-multiple

2505: pseudoholomorphic curves, \ie pseudoholomorphic maps from a fixed real surface 

2506: $S$ modulo reparameterizations.

2507: 

2508: \newdefi{def4.0.1} 

2509: A point $z \in S$ on a $J$-holomorphic curve $u: S \to X$ is a {\sl cusp}, or a 

2510: {\sl cuspidal point}, if $\ord_z du >0$.

2511: The number $\ord_z du$ is called the {\sl order of the cusp} of $u$ at $z$.

2512: A $J$-holomorphic curve $u: S \to X$ containing cuspidal points is called

2513: a {\sl cusp curve}. 

2514: \end{defi}

2515: 

2516: Note that in the literature on pseudoholomorphic curves the notion ``cusp curve''

2517: has a different meaning. Our terminology agrees rather with the one used in

2518: algebraic geometry where the notion ``cusp'' means a ``peak'', \ie an irreducible

2519: singularity. This describes the situation at hand more accurately.

2520: %In the author's opinion, such terminology is more correct.

2521: 

2522: 

2523: \newsubsection[4.1]{Deformation of pseudoholomorphic maps} 

2524: Explicit construction of deformations is needed to obtain local charts for 

2525: subspaces of curves with prescribed singularities.

2526: 

2527: \newlemma{lem4.1.1} Let $B \subset \rr^{2n} \cong \cc^n$ be the unit ball, 

2528: $\scry$ a Banach manifold, $\{J_{\eta,t}\}_{\eta\in \scry, t\in  [0,1]}$ a 

2529: family of homotopies of almost complex structures in $B$ with parameterized by $

2530: \scry$ and depending $C^{\ell-1}$-smoothly on $(\eta,t)\in \scry\times [0,1]$.

2531: Further, let $u_{\eta,0}: \Delta \to B$, $\eta\in \scry$, be a $C^{\ell-1}

2532: $-smooth family of $J_{\eta,0}$-holomorphic map, such that $u_{\eta,0}(\Delta) 

2533: \subset B(\half)$, and $\ord_0(du_{\eta,t})= \mu$. 

2534: 

2535: Then for any family $v_\eta\in \rr^{2n}$ depending $C^{\ell-1}$-smoothly on 

2536: $\eta \in \scry$ and any $\nu \in \nn$ there exists $t^*=t^*(J_t, u_0,v, \mu, 

2537: \nu)>0$, a neighborhood $U_\scry$ of a given $\eta^*\in \scry$, and a 

2538: $C^{\ell-1}$-smooth family of homotopies $\{ w_{\eta,t} \}_{\eta\in U_\scry, 

2539: t\in [0, t^*]}$ with $w_{\eta,t}\in L^{1,p}(\Delta, \rr^{2n})$ such that the 

2540: maps $u_{\eta,t}: \Delta \to B$ given by

2541: \begin{equation}

2542: u_{\eta,t}(z) = u_{\eta,0}(z) + z^\nu (t\,v_\eta + w_{\eta,t}(z))

2543: \eqqno(4.1.1)

2544: \end{equation}

2545: 

2546: \noindent

2547: \sli are $J_{\eta,t}$-holomorphic if $\nu \le 2\mu +1$, and

2548: 

2549: \noindent

2550: \slii are $J_{\eta,0}$-holomorphic if $\nu >2\mu +1$.

2551: 

2552: Moreover, for $z\not =0$ the function $w_{\eta,t}(z)$ depends 

2553: $C^{\ell-1}$-smoothly on $(\eta, t, z)$.

2554: \end{lem}

2555: 

2556: \state Remarks.~1. 

2557: In other words, there exists a pseudoholomorphic deformation $u_t$ of a given map

2558: $u_0$ in a given direction ${d \over dt}u_t\ogran_{t=0}=z^\nu v+ O(|z|^{\nu+

2559: \alpha})$; and moreover, for smaller $\nu$ it is possible to deform 

2560: simultaneously the almost complex structure. Furthermore, if the initial data 

2561: depend smoothly on the parameter $\eta$, then the corresponding constructions give 

2562: a smooth dependence of the maps on $\eta$.

2563: 

2564: \state 2. The loss of smoothness from $C^\ell$ to $C^{\ell-1}$ is due to the fact 

2565: that the Gromov operator $D_{u,J}$ depends only $C^{\ell-1}$-smoothly

2566: on $u$. Indeed, $D_{u,J}$ is the derivative of the $\dbar$-operator $u \mapsto 

2567: \dbar_J u$ in the $u$-direction, which is only $C^\ell$-smooth.

2568: 

2569: \proof We give only a sketch. First, we fix a family $\phi_{\eta,t}$ of affine 

2570: transformations of $\rr^{2n}$ with depend $C^{\ell-1}$-smoothly on $(\eta,t)$ 

2571: such that $\phi_{\eta,t} \scirc u_{\eta,0} (0) =0 \in B$ and $\phi_{\eta,t}\scirc 

2572: J_{\eta,t} \scirc \phi_{\eta,t}\inv$ coincide with $J\st$ in $u_{\eta,0}(0)$. 

2573: Setting $\ti u_{\eta,0} \deff \phi_{\eta,t} \scirc u_{\eta,0}$ and $\ti J_{\eta,t}

2574: \deff \phi_{\eta,t} \scirc J_{\eta,t} \scirc \phi_{\eta,t}\inv$ we reduce the 

2575: problem to the case where $\ti u_{\eta,0}(0) = 0$ and $\ti J_{\eta,t}(0) = J\st$. 

2576: %the standard structure in $\rr^{2n} \cong \cc^n$. 

2577: 

2578: Now we assume that there is no dependence on the parameter and drop the index 

2579: $\eta$. Using \eqqref(4.1.1) one writes the equation $\dbar_{J_t} u_t =0$ in the 

2580: form

2581: \begin{equation}

2582: (x + y J\st)^{-\nu} \dbar_{J_t}

2583: \bigl(u_0(z) + (x + y J\st)^\nu (t\,v + w_t(z)) \bigr)=0

2584: \eqqno(4.1.2)

2585: \end{equation}

2586: with $x +\isl y =z$ the standard coordinates on $\Delta$, and considers 

2587: \eqqref(4.1.2) as an equation for $w_t(z)$. Then one shows that under the 

2588: hypotheses of the lemma the linearization of \eqqref(4.1.2) has the form

2589: \begin{equation}

2590: (\dbar^{(\nu)}_{u_t, J_t} + R^{(\nu)}_{u_t, J_t}) \dot w_t(z) =

2591: \psi^{(\nu)}_{u_t, J_t}(\dot J_t)(z),

2592: \eqqno(4.1.3)

2593: \end{equation}

2594: where $\dot w_t(z)= {d \over dt}w_t(z)$ and $\psi^{(\nu)}_t(\dot J_t)(z) \in 

2595: L^\infty(\Delta, \cc^n)$. Thus it is sufficient to find a right inverse

2596: $T^{(\nu)}_{u_t, J_t}$ of the Gromov type operator $D^{(\nu)}_{u_t, J_t} = 

2597: \dbar^{(\nu)}_{u_t, J_t} + R^{(\nu)}_{u_t, J_t}$ with an additional condition 

2598: $\dot w_t(0)=0$. We refer to \cite{Iv-Sh-1}, {\sl Lemma 3.3.1}, for the 

2599: explicit construction of such a right inverse $T^{(\nu)}_{u, J}$. Moreover, 

2600: the operator $T^{(\nu)}_{u, J}$ and the inhomogeneity term $\psi^{(\nu)}_{u,J}$

2601: depend smoothly on $u$ and $J$. As a consequence, the solution $w_t$ of 

2602: \eqqref(4.1.2) depends $C^{\ell-1}$-smoothly on the parameter $\eta \in \scry$.

2603: \qed

2604: 

2605: 

2606: \newdefi{def4.1.1} Let $B \subset \rr^{2n} \cong \cc^n$ be a ball, $J_0$ a 

2607: $C^\ell$-smooth almost complex structure in $B$, $u_0: \Delta \to B$ a 

2608: $J_0$-holomorphic map and $\nu\ge 1$ an integer exponent. Denote

2609: by $\dfrm_\nu (u, J; v)$ the a map depending $C^{\ell-1}$-smoothly on

2610: 

2611: \begin{itemize}

2612: \item

2613: a $C^\ell$-smooth almost complex structure $J$ in $B$, sufficiently close to $J_0$;

2614: 

2615: \item a $J$-holomorphic map $u$, sufficiently close to $u_0$;

2616: 

2617: \item a vector $v \in \rr^{2n}$, sufficiently close to $0$;

2618: \end{itemize}

2619: 

2620: \noindent 

2621: such that $\ti u\deff \dfrm_\nu (u, J; v)$ is a $J$-holomorphic map of the form

2622: $\ti u(z)= u(z) + z^\nu v + O(|z|^{\nu + \alpha})$. Note that the choice of such

2623: a map $\dfrm_\nu$ is not unique.

2624: \end{defi}

2625: 

2626: 

2627: \smallskip

2628: \lemma{lem4.1.1} allows us to construct local deformations of pseudoholomorphic 

2629: maps with appropriate types of singularities. To obtain a global deformation,

2630: we use 

2631: 

2632: \newlemma{lem4.1.2} Let $u_0: S \to X$ be a non-multiple $J_0$-holomorphic 

2633: map, $z_1, \ldots, z_m$ fixed points on $S$, and $U_1, \ldots, U_m \subset S$

2634: disjoint neighborhoods of these points. Further, let $\{J_t\}_{t\in [0,1]}$

2635: be a given $C^{\ell-1}$-smooth homotopy of almost complex structures on $X$, and

2636: $\{u_{i,t}\}_{t\in [0,1]}$ given $C^{\ell-1}$-smooth homotopies of

2637: $J_t$-holomorphic maps $u_{i,t}: U_i \to X$.

2638: 

2639: Then there exist $t^*>0$, a $C^{\ell-1}$-smooth homotopy $\{\ti J_t\}_{t

2640: \in [0,t^*]}$ of almost complex structures on $X$, a $C^{\ell-1}$-smooth homotopy

2641: $\{\ti u_t\}_{t\in [0,t^*]}$ of $\ti J_t$-holomorphic maps $\ti u_t: S \to X$

2642: such that $u_t$ coincides with each $u_{i,t}$ in some (possibly smaller)

2643: neighborhood of $z_i$ and $\ti J_t$ coincides with $J_t$ in some

2644: neighborhood of each $x_i\deff u_0(z_i)$.

2645: \end{lem}

2646: 

2647: The proof of the lemma is left to the reader.

2648: 

2649: \smallskip

2650: Refining the result of \lemma{lem4.1.1} we show that the condition $u_1(z)-

2651: u_2(z)= o(|z|^k)$ of \lemma{lem1.2.4} defines a submanifold in the spaces of 

2652: pairs of pseudoholomorphic maps.

2653: 

2654: 

2655: \newdefi{def4.2b.1} Define {\sl the spaces of pairs of pseudoholomorphic maps 

2656: coinciding up to order $k$ at $z=0$} as $\scrpp_k(\Delta, X) \deff $

2657: \begin{equation}

2658: \bigl\{ (u', u'', J)\in 

2659: L^{1,p}(\Delta, X) \times L^{1,p}(\Delta, X) \times \scrj: 

2660: \dbar_Ju'=0=\dbar_Ju'', u'(z) - u''(z) = o(|z|^k) \bigr\},

2661: \eqqno(4.2b.1)

2662: \end{equation}

2663: where the condition $u'(z) - u''(z) = o(z^k)$ is related to any local 

2664: coordinate system on $X$ in a neighborhood of the point $u'(0)= u''(0) \in X$.

2665: \end{defi}

2666: 

2667: The structure of $\scrpp_m(\Delta, X)$ for the cases $k=0$ and $k=1$ is easily

2668: obtained from transversality techniques. In general we have

2669: 

2670: \newthm{thm4.2b.1} Assume that $\scrj$ consists of $C^\ell$-smooth structures 

2671: with $\ell\ge2$. Then the space $\scrpp_k(\Delta, X)$ is a $C^{\ell-1}

2672: $-submanifold of the fiber product $\scrp(\Delta, X) \times_{\!\!\scrj} 

2673: \scrp(\Delta, X)$ of codimension of $2n(k+1)$ with the the tangent space 

2674: \begin{multline}

2675: T_{(u',u'',J)}\scrpp_k(\Delta, X) = 

2676: \\

2677: \bigl \{ (v',v'',\dot J) \in 

2678: T_{(u',u'',J)}\bigl(\scrp(\Delta, X) \times_{\!\!\scrj} \scrp(\Delta, X) 

2679: \bigr) :j^k(v' -v'') = 0 \bigr\}.

2680: \eqqno(4.2b.2)

2681: \end{multline}

2682: 

2683: Moreover, for $k=0$ and $k=1$ the space $\scrpp_k(\Delta, X)$ is well-defined

2684: and $C^\ell$-smooth also for $\ell \ge1$.

2685: \end{thm}

2686: 

2687: \proof It follows from \lemma{lem2.2.4} that $\scrp(\Delta, X) 

2688: \times_{\!\!\scrj} \scrp(\Delta, X)$ is a $C^\ell$-smooth Banach manifold with 

2689: the tangent space 

2690: \begin{multline}

2691: T_{(u',u'',J)}\bigl(\scrp(\Delta, X) \times_{\!\!\scrj} \scrp(\Delta, X) 

2692: \bigr)=

2693: \\

2694: \bigl\{ (v', v'', \dot J) : (v',\dot J) \in T_{(u',J)}\scrp(\Delta, X), 

2695: (v'',\dot J) \in T_{(u'',J)}\scrp(\Delta, X) \bigr\},

2696: \eqqno(4.2b.3)

2697: \end{multline}

2698: so that $D_{u',J}v' + \dot J \scirc du' \scirc J_\Delta =0$ and similarly for 

2699: $v''$. 

2700: 

2701: Fix $(u'_0, u''_0, J_0) \in \scrpp_0(\Delta, X)$ and local coordinates $(w_i)$ in

2702: a neighborhood $U \subset X$ of $x^* \deff u'_0(0)= u''_0(0) \in X$. Then there 

2703: exists $r>0$ such that for any pair $(u', u'')$ of $L^{1,p}(\Delta, X)$-maps

2704: sufficiently close to $(u'_0, u''_0)$ we  have $u'(\Delta(r)) \subset U$ and 

2705: $u''(\Delta(r)) \subset U$. The coordinates in $U$ induce the linear structure.

2706: Thus we can consider the difference $u'(z)- u''(z)$ having in mind that it 

2707: is well-defined only for $z\in \Delta(r)$. 

2708: 

2709: The subspace $\scrpp_0(\Delta, X)$ is defined by the condition $u'(0)=u''(0)$ 

2710: for $(u',u'',J) \in \scrp(\Delta, X) \times_{\!\!\scrj} \scrp(\Delta, X)$. Setting 

2711: $F(u', u'', J) \deff u'(0) -u''(0)$ we obtain a $C^\ell$-smooth function, which

2712: is well-defined in a neighborhood of $(u'_0, u''_0, J_0)$ and is a 

2713: local defining function for $\scrpp_0(\Delta, X)$. The differential of $F$ in 

2714: $(u',u'',J) \in \scrpp_0(\Delta, X)$,

2715: $$

2716: dF: T_{(u',u'',J)}\bigl(\scrp(\Delta, X) \times_{\!\!\scrj} \scrp(\Delta, X) 

2717: \bigr) \to T_{u'(0)} X,

2718: $$ 

2719: is given by the formula $dF(v',v'', \dot J) = v'(0)-v''(0)$ and is

2720: a surjective map. Thus $\scrpp_0(\Delta, X)$ is a $C^\ell$-smooth submanifold

2721: of $\scrp(\Delta, X) \times_{\!\!\scrj} \scrp(\Delta, X)$ of codimension $2n =

2722: \dimr X$. 

2723: 

2724: \smallskip

2725: Considering the $C^\ell$-smooth map $\ev_0: \scrpp_0(\Delta, X) \to X$ with 

2726: $$\ev_0(u',u'', J) \deff u'(0) = u''(0),$$ we obtain  a $C^\ell$-smooth bundle

2727: $E^{(0)}$ over $\scrpp_0(\Delta, X)$ with fiber $E^{(0)}_{(u',u'', J)}= {u'}^*

2728: T_{u'(0)}X$. The formulas $\sigma'(u',u'', J) \deff du'(0)$ and $\sigma''(u',u'',

2729:  J) \deff du''(0)$ define $C^\ell$-smooth sections of the bundle $T^*_0\Delta 

2730: \otimes E^{(0)}$ over $\scrpp_0(\Delta, X)$. Thus the condition $du'(0)= du''(0)$

2731: is equivalent to the vanishing of $\sigma' -\sigma''$. Consequently, $\scrpp_1

2732:  (\Delta, X)$ is a $C^\ell$-smooth submanifold of $\scrpp_0(\Delta, X)$ of 

2733: codimension $2n$. 

2734:  

2735: \smallskip

2736: We proceed further by induction using the case $k=0$ as the base. Our notation is

2737: as follows. For a triple $(u',u'', J)\in \scrpp_0(\Delta,X)$ we 

2738: consider the (integrable) complex structure $J\st$ in $U$ with coincides with 

2739: $J$ at the point $u'(z)= u''(z)$ and is constant \wrt the coordinates in $U$. 

2740: Note that $J\st$ depends $C^\ell$-smoothly on $(u',u'', J) \in \scrpp_k(\Delta,

2741: X)$. Thus we can regard $U$ as an open subset in $\cc^n$.

2742: 

2743: For a pair $(u',u'')$ of $J$-holomorphic maps with values in $U \subset X$ we 

2744: obtain

2745: \begin{align}

2746: 0& = \dbar_J u' -\dbar_J u'' = \bigl( (\d_x u' - J(u')\cdot \d_y u') -

2747: (\d_x u'' - J(u'')\cdot \d_y u'') 

2748: \notag\\

2749: &= \d_x(u'-u'') + J(u')\cdot \d_y(u' -u'') +

2750: \bigl(J(u') -J(u'')\bigr)\cdot \d_y u'' 

2751: \notag\\

2752: &= \dbar_{J(u')}(u' -u'') +\bigl(J(u') -J(u'')\bigr)\cdot \d_y u'' .

2753: \eqqno(4.2b.4)

2754: \end{align}

2755: Consequently,

2756: \begin{align}

2757: \llap{$\dbar_{J\st}$}(u'-u'')&

2758: = \dbar_{J\st}(u'-u'') - (\dbar_J u' -\dbar_J u'')

2759: \notag\\

2760: &=\bigl(J\st - J(u')\bigr)\cdot \d_y (u'-u'')- 

2761: \bigl(J(u') -J(u'')\bigr)\cdot \d_y u'' .  

2762: \eqqno(4.2b.5)

2763: \end{align}

2764: Let us denote the last expression by $H_{u',u'',J}(z)$

2765: 

2766: \smallskip

2767: Now suppose that $(u',u'', J)$ varies in $\scrpp_k(\Delta,X)$ with $k\ge1$.

2768: We can assume by induction that $\scrpp_k(\Delta,X)$ is a $C^{\ell-1}$-smooth 

2769: manifold. We claim that for any $p<\infty$

2770: \begin{equation}

2771: \eqqno(4.2b.7)

2772: \llap{$f_k(z) \deff $}z^{-(k+1)}(u'(z) - u''(z)) 

2773: \end{equation}

2774: is a well-defined $L^{1,p}(\Delta(r), \cc^n)$-valued function depending $C^{\ell-

2775: 1}$-smoothly on $(u',u'',J) \in \scrpp_k(\Delta,X)$. The claim implies the 

2776: theorem. Indeed, the function $F_k$ given by $F_k: (u',u'', J) \in \scrpp_k(

2777: \Delta, X) \mapsto f_k(0) \in \cc^n$ is then a local defining function for

2778: $\scrpp_{k-1}(\Delta,X)$ inside $\scrpp_k(\Delta,X)$, whereas non-degeneracy

2779: of $dF_k$ can be easily obtained from \lemma{lem4.1.1}.

2780: 

2781: Again by induction, we can suppose that $f_{k-1}(z)= z^{-k}(u'(z) - u''(z))$ is a 

2782: well-defined $L^{1,p}(\Delta(r), \cc^n)$-valued function depending $C^{\ell-1}

2783: $-smoothly on $(u',u'',J) \in \scrpp_{k-1}(\Delta,X)$. Note that $f_{k-1}(0)$

2784: vanishes identically on $\scrpp_k(\Delta,X)$. Further, for any exponents $p, p'$

2785: with $2<p'<p <\infty$ the map $f(z)\in L^{1,p}(\Delta, \cc^n) \mapsto z\inv (f(z)

2786: -f(0)) \in L^{p'}(\Delta, \cc^n)$ is linear and bounded. Consequently, for any

2787: $p<\infty$ the function $f_k(z) = z\inv f_{k-1}(z)$ lies in $L^p(\Delta, \cc^n)$

2788: and depends $C^{\ell-1}$-smoothly on $(u',u'',J) \in \scrpp_k(\Delta,X)$ \wrt

2789: the $L^p$-topology.

2790: 

2791: \smallskip

2792: Without loss of generality we may assume that $U$ is convex. The identity 

2793: $$

2794: J(w) = J(w^*) + \int_{t=0}^1 \d_t J(w^* +t(w-w^*)) dt

2795: $$

2796: for $(w,w^*) \in U \times U$ implies the relation $J(w)= J(w^*) +\sum_i (w_i-

2797: w^*_i) S_i(w,w^*) = S(w,w^*; \allowbreak w-w^*)$ with the function $S(w,w^*; 

2798: \ti w)$ depending $C^\ell$-smoothly on $J \in \scrj$, $C^{\ell-1}$-smoothly on 

2799: $(w,w^*) \in U\times U$ and $\rr$-linearly on $\ti w\in \cc^n$. Substituting 

2800: $u''(z) = u'(z) + z^{k+1} f_k(z)$ in $(J(u') -J(u''))\cdot \d_y u''$ we obtain

2801: $$

2802: (J(u'(z)) -J(u''(z)))\cdot \d_y u''(z) = S\bigl(u''(z), u'(z); z^{k+1}f_k(z) 

2803: \bigr)\cdot \d_y u''(z) 

2804: $$

2805: By apriori regularity estimates, for $r<1$ we can consider $du''(z)$ as a map from 

2806: $\scrpp_{\!k}(\Delta,X)$ to $C^0(\Delta(r),\cc^n)$ which depends $C^\ell$-smoothly 

2807: on $(u',u'',J)$. Thus we have represented the term $(J(u') -J(u''))\cdot \d_y u''$

2808: as a composition of the $C^{\ell-1}$-smooth map 

2809: $$

2810: (u',u'',J)\in \scrpp_k(\Delta,X) \;\mapsto\; S(u''(z),u'(z); f_k(z)) \cdot

2811: \d_y u''(z) \in L^p(\Delta(r), \cc^n)

2812: $$

2813: and the linear bounded map

2814: $$

2815: S(u''(z),u'(z); f_k(z)) \d_y u''(z)\;\mapsto\; z^{-(k+1)} \cdot 

2816: S(u''(z),u'(z); z^{k+1} \cdot f_k(z)) \d_y u''(z).

2817: $$

2818: Thus $(J(u') -J(u''))\cdot \d_y u''$ depends $C^\ell$-smoothly on $(u',u'',J)\in

2819: \scrpp_k(\Delta,X)$ \wrt the norm topology in $L^p(\Delta(r), \cc^n)$. 

2820: Consequently, the formula 

2821: $$

2822: (u',u'', J) \in \scrpp_k(\Delta,X) \;\mapsto \;

2823: z^{-(k+1)} \cdot (J\st - J(u'(z))) \cdot \d_y (u'(z)-

2824: u''(z))

2825: $$

2826: defines a $L^p(\Delta(r), \cc^n)$-valued map depending $C^{\ell-1}

2827: $-smoothly on $(u',u'', J) \in \scrpp_k(\Delta,X)$. 

2828: 

2829: Similar estimates can be be carried out for the first term $(J\st - J(u'))\cdot 

2830: \d_y(u'-u'')$ in \eqqref(4.2b.5). Together, this implies that $h_k(z) \deff z^{-k}

2831: H_{u',u'',J}(z)$ lies in $L^p(\Delta(r),\cc^n)$ and depends $C^{\ell-1}$-smoothly

2832: on $(u',u'', J) \in \scrpp_k(\Delta,X)$ \wrt $L^p$-topology. Now let $f_{\dbar, 

2833: k}(z)$ be a solution of the equation $\dbar_{J\st}f_{\dbar, k}(z) = h_k(z)$

2834: depending $C^{\ell-1}$-smoothly on $(u',u'', J) \in \scrpp_k(\Delta,X)$ \wrt 

2835: the $L^{1,p}$-topology. Then $(u'(z) -u''(z)) - z^{k+1} f_{\dbar, k}(z)$ is 

2836: a holomorphic $\cc^n$-valued function, depending $C^{\ell-1} $-smoothly on 

2837: $(u',u'', J) \in \scrpp_k(\Delta,X)$ \wrt $L^{1,p}$-topology and vanishing in 

2838: $z=0$ up to order $k+1$. Consequently, 

2839: $$

2840: (u'(z) -u''(z)) - z^{k+1} f_{\dbar, k}(z)= z^{k+1}f_{\scro, k}(z)

2841: $$

2842: and $f_k(z) = f_{\scro, k}(z) + f_{\dbar, k}(z)$ possesses the property claimed

2843: above. \qed

2844: 

2845: 

2846: \smallskip

2847: 

2848: \newsubsection[4.2a]{Curves with prescribed cusp order}

2849: In this paragraph we show that $J$-curves with cusps of given order form a 

2850: Banach submanifold of the moduli space and compute its codimension.

2851: 

2852: 

2853: %\nobreak

2854: \newdefi{def4.2.1} For a given natural $m$ we denote by $\mbfk$ an $m$-tuple 

2855: $(k_1, \ldots, \allowbreak

2856: k_m)$ with $k_i \ge1$ and set $|\mbfk| \deff \sum_i

2857: k_i$. The $m$-tuple $(1,\ldots,1)$ is denoted $\bfone_m$.

2858: Define the {\sl moduli space $\scrm_\mbfk$ of pseudoholomorphic curves

2859: with a given cusp order $\mbfk$} as the set of classes $[u,J, \mbfz]$ such

2860: that $[u,J] \in \scrm$ and $u$ has $m$ (marked) cusp-points $\mbfz= \{z^*_1,

2861: \ldots, z^*_m \}$ with $\ord_{z^*_i} \ge k_i$. Two triples $(u,J, \mbfz)$

2862: and $(\ti u, \ti J, \ti\mbfz)$ define the same class $[u,J, \mbfz] = [\ti u,

2863:  \ti J, \ti\mbfz] \in \scrm_\mbfk$ \iff there exists $g \in \bfg$ such that

2864: $\ti u = u \scirc g$ and $\ti z^*_i = g(z^*_i)$.

2865: \end{defi}

2866: 

2867: 

2868: \smallskip

2869: The main result of this paragraph is

2870: 

2871: \newthm{thm4.2.1} The set $\scrm_\mbfk$ is a $C^\ell$-smooth

2872: manifold and the natural map $\scrm_\mbfk \lrar \scrm$ given by $[u,J, \mbfz]

2873: \mapsto [u,J]$ of $\scrm$ is an immersion of codimension $2\, (n\, |\mbfk|

2874: -m)$, where $n= \dimc X$ and $m$ is the number of marked cusp-points.

2875: \end{thm}

2876: 

2877: \smallskip

2878: We divide the proof in several steps. First we consider the corresponding

2879: problem for $\whcalm$. The reason is that it is more convenient to work

2880: with maps, \ie elements of

2881: $\whcalm$, than with parameterized curves, \ie elements of $\scrm$. This

2882: means that we are interested in the set

2883: \begin{equation}

2884: \whcalm_\mbfk \deff \left\{ (u,J_S, J; z^*_1, \ldots z^*_m) \in

2885: \whcalm \times (S)^m :

2886: \msmall{ \matrix %\format\c\\

2887:  z^*_i\text{ are pairwise distinct, }\cr

2888: \ord_{z^*_i} du \ge k_i% \text{\rmnine has a cusp in $z^*_i$ of order}\;

2889: \endmatrix}

2890: \right\},

2891: \eqqno(4.2.1)

2892: \end{equation}

2893: where $(S)^m= S \times \cdots \times S$ is the $m$-fold product of $S$.

2894: Obviously, the projection from $\whcalm \times (S)^m$ onto $\whcalm$ and then

2895: onto $\scrm$ maps $\whcalm_\mbfk$ onto $\scrm_\mbfk$. In our proof of

2896: \refthm{thm4.2.1} we shall show that this map $\whcalm_\mbfk \to

2897: \scrm_\mbfk$ is a principle $\bfg$-bundle.

2898: 

2899: \newdefi{def4.2.1n} Set 

2900: \begin{equation}

2901: \whcalm^{(m)} \deff \{ (u,J_S, J; z^*_1, \ldots z^*_m) \in

2902: \whcalm \times (S)^m : z^*_i \not = z^*_j \text{ for every } i\not=j\,\}

2903: \eqqno(4.2.2n)

2904: \end{equation}

2905: denoting by $S_i$ the $i$-th factor in $(S)^m$. Equip $\whcalm^{(m)}$ with the 

2906: maps $\ev_i: \whcalm^{(m)} \to X^m$ defined by $\ev_i(u,J_S, J; z^*_1, \ldots, 

2907: z^*_m) \deff u(z^*_i)$. Denote by $E_i$ the pulled-back bundles $\ev_i^*TX$ and

2908: $\ev^{(m)}{}^* T(X^m)$ over $\whcalm^{(m)}$. The fiber of $E_i$ over $(u, J_S, 

2909: J; \mbfz)$ is $(E_i)_{(u, J_S, J; \mbfz)}= T_{u(z^*_i)}X$.

2910: \end{defi}

2911: 

2912: Obviously, the space $\whcalm^{(m)}$ is a $C^\ell$-smooth Banach manifold, 

2913: $\ev_i: \whcalm^{(m)} \to X^m$ are $C^\ell$-smooth maps, and $E_i$ are 

2914: $C^\ell$-smooth bundles over $\whcalm^{(m)}$. Note that we also have line

2915: bundles $TS_i$ and $T^*S_i$ over $\whcalm^{(m)}$ which are defined in an obvious 

2916: way as the (co)tangent bundles to each $S_i$.

2917: 

2918: \newlemma{lem4.2.2n} The formula $\yps(u,J_S, J; z^*_1, \ldots z^*_m) \deff

2919:  (du(z^*_1), \ldots, du(z^*_m)) \in \bigoplus_i T^*S_i \otimes E_i$ defines

2920: a $C^\ell$-smooth section of\/ $\bigoplus_i T^*S_i \otimes E_i$ over 

2921: $\whcalm^{(m)}$, transversal to the zero section. The zero-set of $\yps$

2922: coincides with the space $\whcalm_{\bfone_m}$ of maps having cups in each

2923: marked $z^*_i$. Thus $\whcalm_{\bfone_m}$ is a $C^\ell$-smooth Banach 

2924: submanifold of $\whcalm^{(m)}$ of codimension $2nm$.

2925: \end{lem}

2926: 

2927: Before starting the proof we introduce some new notation.

2928: 

2929: \newdefi{def4.2.2n} Let $\scry$ be a $C^\ell$-smooth Banach manifold and $f: 

2930: \scry \to \ttt_g \times S$ a $C^\ell$-smooth map of the form $f(y)= (J_S(y), 

2931: z^*(y))$. Set $F(y) \deff (y, z^*(y))$ so that $F: \scry \to \scry \times S$ is 

2932: an imbedding. A {\sl local $J_S(y)$-holomorphic coordinate (or simply 

2933: a {\sl$J_S$-holomorphic coordinate}) on $\scry \times S$ centered at $z^*$} 

2934: is a $C^\ell$-smooth $\cc$-valued function $z$ defined in some neighborhood $U

2935: \subset \scry \times S$ of $F(\scry)$ which vanishes along $F(\scry)$ and is

2936: $J_S(y)$-holomorphic along each $\{y\} \times S$. One can use \lemma{lem4.1.1}

2937: for a proof of the existence of such a local holomorphic coordinate. 

2938: \end{defi}

2939: 

2940: \statep Proof of. \lemma{lem4.2.2n}. It is obvious that $\yps$ is 

2941: well-defined. To show the $C^\ell$-smoothness of $\yps$, for any $i=1,\ldots,m$,

2942: we fix some local coordinate $z_i$ on $\whcalm^{(m)}$ which is 

2943: $J_S$-holomorphic along $S_i$ and centered at $z^*_i \in S_i$. Now we can

2944: find a local frame $\bfxi=(\xi_1, \ldots, \xi_n)$ of $T^*S_i \otimes E_i$ which

2945: depends $C^\ell$-smoothly on $(u,J_S, J) \in \whcalm$ and

2946: holomorphically on the coordinate $z_i$. The existence of such a frame follows

2947: from \refdefi{def1.4.1} and a parametric version of \lemma{lem1.4.1A}.

2948: The coefficients of $du \in T^*S_i \otimes E_i$ \wrt such a frame 

2949: $\bfxi$ depend $C^\ell$-smoothly on $(u,J_S, J) \in \whcalm$ and holomorphically

2950: on $z_i$. Consequently, the $du(z^*_i)$ depend $C^\ell$-smoothly on $(u,J_S, J; 

2951: \mbfz) \in \whcalm^{(m)}$. Thus $\yps$ is $C^\ell$-smooth.

2952: 

2953: The transversality of $\yps$ to the zero-section of $\bigoplus_i T^*S_i \otimes

2954: E_i$ follows immediately from results of \refsubsection{4.1}. In particular,

2955: $\whcalm^{(m)}_{\bfone_m}$ is the $C^\ell$-smooth Banach submanifold of

2956: $\whcalm^{(m)}$. The corresponding codimension is $\rank_\rr\left(

2957: \bigoplus_i T^*S_i \otimes E_i\right) =2nm$. \qed

2958: 

2959: 

2960: \newdefi{def4.2.3n} For (finite-dimensional) complex vector spaces $V$, $W$, 

2961: and $k\in \nn$ denote by $j^k(V,W)$ the vector space of polynomial maps 

2962: $f:V \to W$ of degree $\deg f\le k$ with $f(0)=0$, considered as the space of 

2963: $k$-jets of holomorphic maps $F: V \to W$. For $l\ge k$ the natural projection

2964: $\pr: j^l(V,W) \to j^k(V,W)$ is well-defined. Let $j^{k,l}(V,W)$ denote 

2965: its kernel. Similar notation for complex bundles is used. Note that 

2966: $j^1(V,W) = \hom(V,W) = V^* \otimes W$.

2967: \end{defi}

2968: 

2969: \newlemma{lem4.2.3n} \sli For any $(u, J_S, J; \mbfz) \in \whcalm_\mbfk$ the

2970: jet $j^{2k_i+1}u(z^*_i)$ is a well-defined element of $j^{2k_i+1} (T_{z^*_i}, 

2971: T_{u(z^*_i)}X) = j^{2k_i+1} (TS_i, E_i)_{(u, J_S, J; \mbfz)}$. 

2972: 

2973: \slii Moreover, $j^{2k_i+1}u(z^*_i) \in j^{k_i+1, 2k_i+1} (T_{z^*_i}, 

2974: T_{u(z^*_i)}X)= j^{k_i+1, 2k_i+1} (TS_i, E_i)_{(u, J_S, J; \mbfz)}$.

2975: 

2976: \sliii Set $\yps_\mbfk(u, J_S, J; \mbfz) \deff \bigl(j^{k_1+1, 2k_1+1} u(z^*_1), 

2977: \ldots, j^{k_m+1, 2k_m+1}u(z^*_m) \bigr)$. Then $\yps_\mbfk: \whcalm_\mbfk \to 

2978: \bigoplus_i j^{k_i+1, 2k_i+1} (TS_i, E_i)$ is a section which is $C^\ell$-smooth 

2979: and transversal to the zero-section.

2980: \end{lem}

2981: 

2982: 

2983: \proof Assertions \sli and \slii follow essentially from \lemma{lem1.2.4}. 

2984: The nontrivial points here are the following. First, the jet 

2985: $j^{2k_i+1}u(z^*_i)$ is defined even if the structure $J$ is $C^\ell$-smooth 

2986: with $\ell<2k_i$ and the map $u$ is $C^{\ell+1}$-smooth, since in general there

2987: are no higher smoothness for $u$. Second, the jet $j^{2k_i+1}

2988: u(z^*_i)$ is a {\sl complex} polynomial. Finally, the jet $j^{2k_i+1}u(z^*_i)$

2989: is independent of the choice of the integrable structure $J\st$ and 

2990: $J\st$-holomorphic coordinates in a neighborhood of $u(z^*_i)$ used in 

2991: \lemma{lem1.2.4} for definition of the jet. Let us give a proof of the latter

2992: property.

2993: 

2994: Let $J'$ and $J''$ be integrable complex structures in a neighborhood of 

2995: $u(z^*_i)$ such that $J'(u(z^*_i)) = J''(u(z^*_i)) = J(u(z^*_i))$. Find local

2996: complex coordinate systems $\mbfw'= (w'_1, \ldots,  \allowbreak

2997: w'_n)$ and $\mbfw''= (w''_1, \ldots,w''_n)$ which are centered in $u(z^*_i)$ and 

2998: holomorphic \wrt $J'$ and

2999: $J''$ respectively. Without loss of generality we may assume that the frames 

3000: $({\d \over \d w'_1}, \ldots,  {\d \over \d w'_n})$ and $({\d \over \d w''_1}, 

3001: \ldots,  {\d \over \d w''_n})$ coincide in $u(z^*_i)\in X$. Consequently, 

3002: we can express one system by another using the formula $\mbfw''= \mbfw' + 

3003: F(\mbfw')$ with 

3004: \begin{equation}

3005: F(\mbfw') = O(|\mbfw'|^2) \qquad \text{and} \qquad 

3006: dF(\mbfw') = O(|\mbfw'|).

3007: \eqqno(4.2.3n)

3008: \end{equation}

3009: Let $u'(z)$ and $u'(z)$ be the local expressions of $u: S \to X$ in the local 

3010: coordinate systems $\mbfw'$ and $\mbfw''$ respectively. Then $u''(z)= u'(z) + 

3011: F(u'(z))$. So from \eqqref(4.2.3n) and $u'(z)= O(|z|^{k_i+2})$ we see that

3012: coefficients of polynomials $j^{2k_i+1}u'(z)$ and $j^{2k_i+1}u''(z)$ coincide. 

3013: 

3014: \smallskip

3015: To show the smoothness of the section $\yps_\mbfk$ we fix an element 

3016: $(u_0,J_{S,0}, J_0; \mbfz_0)\in \whcalm_\mbfk$, $\mbfz_0=(z^*_{1,0}, \ldots,

3017: z^*_{m,0})$, and a sufficiently small neighborhood $\scry \subset \whcalm

3018: _\mbfk$ of $y_0 \deff (u_0, J_{S,0}, J_0; \mbfz_0)$.

3019: In what follows, for any $i=1, \ldots,m$, we fix families of certain structures 

3020: on various spaces. We assume that the members of the families are parameterized

3021: by and depend $C^\ell$-smoothly on $y=(u,J_S, J; \mbfz)\in \scry$. The 

3022: families are:

3023: \begin{enumerate}

3024: \item integrable complex structures $J'_i$ in a neighborhood 

3025: of each $u(z^*_{i,0})$ such that each $J'_i$ coincides with $J$ in $u(z^*_i)$;

3026: \item local complex coordinate systems $\mbfw'_i= (w'_{i,1}, \ldots, 

3027: w'_{i,n})$ on $X$ centered in $u(z^*_i)$ and holomorphic \wrt $J'_i$;

3028: \item local frames $\bfxi_i= (\xi_{i,1},\ldots ,\xi_{i,n})$ of the bundles 

3029: $E_i$ which are defined in a neighborhood of $z^*_i$ and holomorphic along 

3030: $S_i$; 

3031: \item  local $J_S$-holomorphic coordinates $z_i$ on $S_i$ centered in 

3032: $z^*_i$.

3033: \end{enumerate}

3034: Further, we assume that every coordinate $z_i$ has image the whole disc $\Delta$.

3035: Note that pulling back the frames $({\d \over \d w'_{i,1}}, \ldots, {\d \over 

3036: \d w'_{i,n}} )$ we obtain local frames $\left(u^*({\d \over \d w'_{i,1}}), \ldots, 

3037: u^*({\d \over \d w'_{i,n}}) \right)$ of $E_i$ which depend $C^\ell$-smoothly 

3038: on $y=(u,J_S, J; \mbfz)\in \scry$. Now, the expression of $u(z)$ 

3039: in the local coordinate system $\mbfw'_i$ yields an element $u'_i(z_i) \in

3040: L^{1,p}(\Delta, \cc^n)$ which depends $C^\ell$-smoothly on $y\in \scry$ 

3041: \wrt the standard smooth structure in $L^{1,p}(\Delta, \cc^n)$. Deriving, we 

3042: obtain an element $du'_i(z_i)\in L^p(\Delta, \cc^n\otimes_\rr T^*\Delta)$ which

3043: depends $C^\ell$-smoothly on $y\in \scry$ \wrt the standard smooth structure 

3044: in $L^p(\Delta, \cc^n\otimes_\rr T^*\Delta)$. 

3045: 

3046: Consider now $du'_i$ as a 

3047: section of $E_i\otimes T^*S_i$, and its coefficients of $du'_i$ in the frame 

3048: $\left(u^*({\d \over \d w'_{i,1}}) \otimes dz_i, \ldots, u^*({\d \over 

3049: \d w'_{i,n}}) \otimes dz_i \right)$ as $L^p(\Delta, \cc)$-functions. Thus we

3050: can conclude that the coefficients of $du'_i$ depend $C^\ell$-smoothly on 

3051: $y\in \scry$ \wrt the standard smooth structure in $L^p(\Delta, \cc)$.

3052: Consequently, the same is true for the coefficients of $du'_i$ in the frame 

3053: $(\xi_{i,1}\otimes dz_i,\ldots, \xi_{i,n}\otimes dz_i)$. Since the latter frame

3054: is holomorphic, the coefficients of the jet $j^{2k_i}du(z^*_i)$ depend 

3055: $C^\ell$-smoothly on $y\in \scry$. This provides the desired smoothness 

3056: property of $\yps_\mbfk$.

3057: 

3058: \smallskip

3059: Finally, note that the transversality of $\yps_\mbfk$ to the zero-section 

3060: follows from results of \refsubsection{4.1}. \qed

3061: 

3062: 

3063: \newcorol{cor4.2.4n} For any $\mbfk=(k_1,\ldots, k_m)$ with $k_i \ge 1$ 

3064: the space $\whcalm_\mbfk$ is a $C^\ell$-submanifold of $\whcalm^{(m)}$ of 

3065: codimension $2|\mbfk|n$.

3066: \end{corol}

3067: 

3068: \proof Assume that for a given $\mbfk=(k_1,\ldots, k_m)$ with 

3069: $k_i \ge 1$ the claim holds. Fix some $\mbfk^+=(k^+_1,\ldots, k^+_m)$ with 

3070: $k_i \le k^+_i \le 2 k_i$ and consider truncated section $\yps_{\mbfk,\mbfk^+}

3071: : \whcalm_\mbfk \to \bigoplus_i j^{k_i+1, k^+_i+1} (TS_i, E_i)$ given

3072: by 

3073: $$

3074: \yps_{\mbfk,\mbfk^+}(u, J_S, J; \mbfz) \deff \bigl(j^{k_1+1, k^+_1+1} 

3075: u(z^*_1), \ldots, j^{k_m+1, k^+_m+1}u(z^*_m) \bigr).

3076: $$ 

3077: Then $\whcalm_{\mbfk^+}$ is identified with the zero set of $\yps_{\mbfk,

3078: \mbfk^+}$. By \lemma{lem4.2.3n}, $\yps_{\mbfk,\mbfk^+}$ is transversal to 

3079: the zero-section. Thus $\whcalm_{\mbfk^+}$ is a $C^\ell$-smooth submanifold

3080: of $\whcalm_\mbfk$ of codimension equal to $\rank_\rr \bigoplus_i j^{k_i+1, 

3081: k^+_i+1} (TS_i, E_i) = 2n(|\mbfk^+| -|\mbfk|)$. So we can apply the induction.

3082: \qed

3083: 

3084: \smallskip

3085: \newlemma{lem4.2.5n} The natural projection $\wh\pr_\mbfk : \whcalm_\mbfk 

3086: \to \whcalm$ given by the formula $\wh\pr_\mbfk(u, J_S, J; \mbfz) 

3087: \allowbreak

3088: \deff  (u, J_S, J)$ is an immersion of codimension $2(|\mbfk|n -m)$. 

3089: \end{lem}

3090: 

3091: \proof The differential of the projection $\wh\pr_\mbfk$ is given by 

3092: $$

3093: d\wh\pr_\mbfk: (v, \dot J_S, \dot J; \dot \mbfz)\in 

3094: T_{(u, J_S, J; \mbfz)}\whcalm_\mbfk \mapsto 

3095: (v, \dot J_S, \dot J) \in T_{(u, J_S, J)}\whcalm.

3096: $$

3097: Thus the kernel $\ker d\wh\pr_\mbfk$ consists of vectors of the form $(0,0,0; 

3098: \dot \mbfz)$ with $\dot \mbfz_i = (\dot z^*_1, \ldots, \dot z^*_m) \in 

3099: \bigoplus T_{z^*_i}S_i$ and we must show that $\ker d\wh\pr_\mbfk$ is trivial.

3100: Intuitively this is obvious, since elements of the kernel correspond to 

3101: deformations leaving $(u,J_S,J)$ unchanged but moving cusp-points $z^*_i$ on 

3102: $S$ and this is impossible.

3103: 

3104: For a rigorous proof we use conclusions of the proof of \lemma{lem4.2.3n}. 

3105: Consider $du(z_i)$ as a holomorphic section of $T^*S_i 

3106: \otimes E_i$. Then $du(z_i)$ vanishes in $z^*_{i,t}$ up to the order $\ge k_i$ 

3107: and there are no other zeros of $du(z_i)$ in a neighborhood of $z^*_i$. Thus 

3108: we can locally express $z^*_i$ as the zero set of $du(z_i)$. This implies

3109: that locally there exists $C^\ell$-smooth functions $F_i$ of $(u,J_S,J)\in 

3110: \whcalm$ such that $F_i(u,J_S,J) =z^*_i$ for $(u,J_S,J)\in \whcalm_\mbfk$.

3111: Thus $\wh\pr_\mbfk : \whcalm_\mbfk \to \whcalm$ is an immersion.

3112: 

3113: To compute the codimension of $\wh\pr_\mbfk: \whcalm_\mbfk \hook \whcalm$

3114: one represents $\wh\pr_\mbfk$ as the composition $ \whcalm_\mbfk \hook 

3115: \whcalm^{(m)} \buildrel \pr \over \lrar \whcalm$.

3116: \qed

3117: 

3118: 

3119: 

3120: \medskip

3121: Now we can finish

3122: 

3123: \nobreak

3124: \statep Proof of. \refthm{thm4.2.1}. Consider the action of $\bfg$ on 

3125: $\whcalm$ and the diagonal action of $\bfg$ on $\whcalm \times (S)^m$. The 

3126: both actions are $C^\ell$-smooth, free, and commute with the projection $\pr: 

3127: \whcalm \times (S)^m \to \whcalm$. Moreover, for every $\mbfk=(k_1,\ldots, 

3128: k_m)$ with $k_i\ge 1$ the submanifold $\whcalm_\mbfk \hook \whcalm \times 

3129:  (S)^m$ is $\bfg$-invariant \wrt the diagonal action of $\bfg$. 

3130: For the quotient $\scrm_\mbfk= \whcalm_\mbfk /

3131: \bfg$ one can construct a $C^\ell$-smooth atlas in the same way as it was

3132: done for $\scrm= \whcalm /\bfg$. The construction shows that the map 

3133: $\scrm_\mbfk \to \scrm$ is a $C^\ell$-smooth immersion of codimension

3134: equal to the codimension of $\wh\pr_\mbfk: \whcalm_\mbfk \hook \whcalm$.

3135: \qed

3136: 

3137: \medskip

3138: Summarizing the results and notation of this paragraph, we obtain

3139: 

3140: 

3141: \newcorol{cor4.2.6n} The maps $\ev_\mbfk : \scrm_\mbfk \to X^m$ and $\ev_i:

3142: \scrm_\mbfk \to X$ given by $\ev_\mbfk([u,J, \mbfz]) \deff (u(z^*_1), \ldots , 

3143: u(z^*_m))$ and $\ev_i([u,J, \mbfz]) \deff u(z^*_i)$ are well-defined and 

3144: $C^\ell$-smooth. This yields $C^\ell$-smooth bundles $E_i \deff \ev_i^* TX$ 

3145: with a fiber $(E_i)_{[u,J, \mbfz]} = T_{u(z^*_i)}X$. The bundles $T_{z^*_i}

3146: S_i$ over $\whcalm_\mbfk$ induce $C^\ell$-smooth bundles $L_i$ over $\scrm

3147: _\mbfk$ with the fiber $(L_i) _{[u,J, \mbfz]}= T_{z^*_i} S_i$.

3148: The section $\yps_\mbfk: \whcalm_\mbfk \to \bigoplus_i j^{2k_i+1}(TS_i, E_i)$

3149: induces the section $\yps_\mbfk: \scrm_\mbfk \to \bigoplus_i j^{2k_i+1}

3150:  (L_i, E_i)$ with $\yps_\mbfk([u,J, \mbfz]) \deff \yps_\mbfk(u,J_S,J; \mbfz)$. 

3151: \end{corol}

3152: 

3153: 

3154: \proof The claim follows from the fact that all the constructions are 

3155: compatible with $\bfg$-action.

3156: \qed

3157: 

3158: \smallskip

3159: \newdefi{def4.2.4n} %{def4.3.2a}

3160: For a given $\mbfk=(k_1, \ldots, k_m)$ we set

3161: \begin{align}

3162: \whcalm_{=\mbfk} &\deff \bigl\{ (u, J_S, J; \mbfz) \in  \whcalm_\mbfk

3163: \;:\; \ord_{z^*_i} du =k_i \,\bigl\};

3164: \\

3165: \scrm_{=\mbfk} &\deff \bigl\{ [u, J; \mbfz] \in  \scrm_\mbfk

3166: \;:\; \ord_{z^*_i} du =k_i \,\bigl\}.

3167: \end{align}

3168: \end{defi}

3169: 

3170: \newlemma{lem4.2.7n} %{lem4.3.3}

3171: \sli The set $\whcalm_{=\mbfk}$ is an open $C^{\ell-1}$-smooth submanifold 

3172: of $\whcalm_\mbfk$ invariant \wrt the natural action of\/ $\bfg$ on 

3173: $\whcalm_\mbfk$.

3174: 

3175: \slii The image of the projection of $\whcalm_{=\mbfk}$ to $\whcalm$ is an

3176: {\sl imbedded} submanifold of $\whcalm$, and the projection is a non-ramified

3177: covering over the image.

3178: 

3179: \sliii There exists a $C^{\ell-1}$-smooth bundle $N$ over $\whcalm_{=\mbfk}

3180: \times S$ whose restriction onto  $\{(u, J_S, 

3181: \allowbreak

3182: J; \mbfz)\} \times S$ coincides with $N_u$. The diagonal action 

3183: of\/ $\bfg$ on $\whcalm_{=\mbfk} \times S$ lifts canonically to the action 

3184: on the bundle $N$.

3185: 

3186: \sliv The bundle $N$ induces Banach bundles $L^{1,p}(S, N)$ and $L^p_{(0,1)}

3187:  (S, N)$ over $\whcalm_{=\mbfk}$ with fibers $L^{1,p}(S, N_u)$ and $L^p_{(0,1)}

3188:  (S, N_u)$ over $(u, J_S, J, \mbfz) \in \whcalm_{=\mbfk}$ respectively. The

3189: operators $D^N_{u,J}: L^{1,p}(S, N_u) \to L^p_{(0,1)}(S, N_u)$ induce a 

3190: $C^{\ell-1}$-smooth bundle homomorphism $D^N: L^{1,p}(S, N) \to L^p_{(0,1)}

3191:  (S, N)$.

3192: \end{lem}

3193: 

3194: \proof \sli The complement $\whcalm_\mbfk \bs \whcalm_{=\mbfk}$ is of the 

3195: union of (the projections of) the spaces $\whcalm_{\mbfk'}$ such that either 

3196: $\mbfk' =(k_1, \ldots, k_m, 1)$, or $\mbfk' =(k'_1, \ldots, k'_m)$ with $k'_i 

3197: \ge k_i$ and $k'_{i_0} > k_{i_0}$ for some $i_0$. In other words, we have 

3198: either at least one additional cusp-point or a higher order cusp in at least 

3199: one point. Obviously, these conditions define closed subsets in 

3200: $\whcalm_\mbfk$. The $\bfg$-invariance of $\whcalm_{=\mbfk}$ follows from 

3201: the definition.

3202: 

3203: \slii The set $\whcalm_{=\mbfk}$ admits a finite transformation group $\aut(

3204: \mbfk)$ generated by transpositions of marked cusp-points $z^*_i$ and

3205: $z^*_j$ with $k_i = k_j$. The rest of part \slii follows.

3206: 

3207: \sliii Let $z_i$ be a local $J_S$-holomorphic coordinate on $\whcalm_{=\mbfk}

3208: \times S$ centered at $z^*_i$ as in \refdefi{def4.2.2n}. It follows from the

3209: proof of \lemma{lem4.2.3n} that $z_i^{-k_i} du(z_i)$ is a well-defined 

3210: {\sl non-vanishing} local section of $\hom(TS, E_u)$, 

3211: which depends $C^{\ell-1}$-smoothly on $(u,J_S, J)$ and holomorphically on

3212: $z_i$. This provides the existence on $N$ with the stated property,

3213: at least locally in a neighborhood of $(u,J_S, J; z^*_i)$. The globalization

3214: of $N$ is trivial. Since the constructions involved are natural, the

3215: $\bfg$-action admits the desired lift.

3216: 

3217: \sliv One uses the fact that the constructions of the bundles $L^{\!1,p}

3218: \!(S, N)$, $L^p_{(0,1)}\!(S, N)$, and the operator $D^N$ are natural. This

3219: implies $C^{\ell-1}$-smoothness of the obtained objects. \qed

3220: 

3221: 

3222: \smallskip

3223: \state Remark. One could explain the meaning of \lemma{lem4.2.7n} as follows.

3224: First, we note that for the globalization of normal bundles $N_u$ to $\scrm$

3225: we should use not the Cartesian product $\scrm \times S$, but the

3226: $\bfg$-twisted product $\scrm \ltimes S$, \ie

3227: $\whcalm \times_\bfg S \deff \bigl( \whcalm \times S\bigr)/\bfg$. Second, we

3228: must choose a stratification of $\scrm$ by strata where $N_u$ does not

3229: ``jump''. By the definition of $N_u$ such strata are exactly $\scrm_{=\mbfk} 

3230: = \whcalm_{=\mbfk} / \bfg$ where there is no ``jump'' of

3231: the cusp-order.

3232: 

3233: 

3234: 

3235: \medskip

3236: Another application of the techniques used in the proof of \refthm{thm4.2.1}

3237: is a local version of the theorem. Below $\scrp(\Delta, X)$ denotes the Banach 

3238: space of pseudoholomorphic maps between the unit disc $\Delta$ with the standard 

3239: structure $J\st$ and $X$, \ie $\scrp(\Delta, X) = \{ (u,J) \in L^{1,p}(\Delta,

3240:  X) \times \scrj: \dbar_{J\st,J} u=0 \}$. 

3241: 

3242: \newlemma{lem4.2.8n} \sli For any given integer $k\ge1$ the set 

3243: \begin{equation}

3244: \eqqno(4.2.4n)

3245: \llap{$\scrp_k($}\Delta,0; X) \deff

3246: \{ (u,J) \in \scrp(\Delta, X) : \ord_{z=0}(du) \ge k \}

3247: \end{equation}

3248: is a $C^\ell$-smooth submanifold of $\scrp(\Delta, X)$ of 

3249: real codimension $2kn$, $n\deff \dimc X= \half \dimr X$, with tangent

3250: space

3251: \begin{equation}

3252: \eqqno(4.2.5n)

3253: T_{(u,J)}\scrp_k(\Delta,0; X) = 

3254: \{ (v,\dot J) \in T_uL^{1,p}(\Delta, X)\times T_J\scrj : 

3255: D_{u,J}v=0, j^k(v(z)-v(0)) =0 \}.

3256: \end{equation}

3257: \end{lem}

3258: 

3259: 

3260: \newsubsection[4.2s]{Curves with prescribed secondary cusp index}

3261: %In this paragraph we prove some claims to be used later.

3262: Recall that by \lemma{lem1.2.3} for a pseudoholomorphic map $u:(S,J_S) \to (X,J)$ 

3263: with cusp order $k$ at $z^*\in S$ the jet $j^{2k+1}u(z^*)$ is well-defined. As we 

3264: shall see, the part of the jet $j^{2k+1}u(z^*)$ invariant under reparameterization 

3265: plays an important role for determining the type of critical points on

3266: moduli spaces (see \refsubsection{3.3}).

3267: 

3268: \newdefi{def4.2s.1} \sli Let $u:(S,J_S) \to (X,J)$ be pseudoholomorphic map with

3269: a cusp of order $k\deff \ord_{z^*} du$ at $z^*\in S$, $\pr_N: E_u \to N_u$

3270: the projection to the normal bundle, and $z$ a local holomorphic coordinate

3271: on $S$ centered at $z^*$. Define the {\sl secondary cusp index 

3272: $l$ of $u$ at $z^*\in S$} by setting $l\deff k$ if $\pr_N \scirc j^{2k+1}

3273: u(z^*)$ is zero polynomial and $l\deff \ord_{z=0} \pr_N \scirc j^{2k+1} 

3274: u(z^*) -k-1$ otherwise.

3275: 

3276: \slii For a given $m$-tuple $\mbfk=(k_1, \ldots, k_m)$ of prescribed

3277: orders of cusps we consider $m$-tuples $\mbfl= (l_1, \ldots, l_m)$ 

3278: with $0 \le l_i \le k_i$ and set $|\mbfl| \deff \sum_i l_i$. 

3279: Define the {\sl moduli space $\scrm_{\mbfk, \mbfl}$ of pseudoholomorphic maps

3280: with cusps of given order and secondary index $(\mbfk, \mbfl)$} as the set of 

3281: $[u,J, \mbfz] \in \scrm_\mbfk$ such that $\ord_{z^*_i} du =k_i$

3282: and the secondary cusp index of $u$ at $z^*_i$ is at least $l_i$. Set

3283: \begin{equation}

3284: \eqqno(4.2s.1)

3285: \whcalm_{\mbfk, \mbfl} \deff \{ (u, J_S, J, \mbfz) \in \whcalm_{=\mbfk}

3286: \;:\; [u, J, \mbfz] \in \scrm_{\mbfk, \mbfl} \}.

3287: \end{equation}

3288: \end{defi}

3289: 

3290: 

3291: \smallskip

3292: \newthm{thm4.2s.1} The space $\scrm_{\mbfk, \mbfl}$ is a closed $C^{\ell-1}

3293: $-smooth submanifold of $\scrm_{=\mbfk}$ of codimension $2(n-1)|\mbfl|$.

3294: \end{thm}

3295: 

3296: \state Remark.

3297: The meaning of the notion of secondary cusp index can be explained as follows.

3298: One expects that for a $J$-holomorphic map $u:S \to X$ with a cusp of order $k=

3299: \ord_{z^*}du$ at $z^* \in S$ the polynomial $\pr_N \scirc j^{2k+1} u(z^*)$ has 

3300: vanishing order $k+1$. Thus the secondary cusp index $l$ is the order of deviation 

3301: from this condition. The content of \refthm{thm4.2s.1} is that for generic 

3302: $[u,J;\mbfz] \in \scrm_{=\mbfk}$ there is no deviation and that the space of 

3303: curves with cusps of prescribed degeneration order is of expected codimension.

3304: 

3305: We note also that the range $0\le l_i \le k_i=\ord_{z^*_i}du$ is the maximal 

3306: one where the secondary cusp index is well-defined: The higher order terms

3307: of $\pr_N \scirc du$, as well as the coefficients of $du$ (considered as a 

3308: holomorphic section of $T^*S \otimes E_u$), depend on the choice of the local

3309: holomorphic coordinate $z_i$ centered at $z^*_i\in S$.

3310: 

3311: \proof We maintain the notation of \lemma{lem4.2.3n}. Now, for any $(u,J_S,J; 

3312: \mbfz)\in \whcalm_{=\mbfk}$,  $\mbfz = (z^*_1,\ldots, z^*_m)$, the jets 

3313: $j^{2k_i+1} u(z^*_i) \in j^{2k_i+1} (TS_i, E_i)_{(u,J_S, J; \mbfz)}$ are 

3314: well-defined and depend $C^{\ell-1}$-smoothly on $(u,J; \mbfz)$. By 

3315: \lemma{lem4.2.7n}, for any $i=1,\ldots,m$ the formula $(N_i)_{(u,J_S, J; 

3316: \mbfz)} \deff (N_u)_{z^*_i}$ defines a $C^{\ell-1}$-smooth bundle $N_i$ over 

3317: $\whcalm_{=\mbfk}$ with the projection $\pr_N : E_i \to N_i$. 

3318: This yields the compositions $\pr_N \scirc j^{2k_i+1}u(z^*_i) \in j^{2k_i+1} 

3319:  (TS_i, N_i)_{(u,J_S, J; \mbfz)}$ which depend $C^{\ell-1}$-smoothly on 

3320: $(u,J_S,J; \mbfz)$. Thus we obtain a $C^{\ell-1}$-smooth bundle 

3321: $$

3322: \bigoplus\nolimits_{i=1}^m j^{k_i+1, k_i+l_i+1}(TS_i, N_i)

3323: _{(u,J_S, J; \mbfz)}

3324: $$

3325: over $\whcalm_{=\mbfk}$ of rank $2(n-1)|\mbfl|$ and a $C^{\ell-1}$-smooth 

3326: section 

3327: $$

3328: \yps^N_{\mbfk,\mbfl} \deff

3329: (\pr_N \scirc j^{k_i+1, k_i+l_i+1} u(z^*_i))_{i=1}^m.

3330: $$

3331: Observe that $\whcalm_{\mbfk,\mbfl}$ is defined in $\whcalm_{=\mbfk}$ as the 

3332: zero set of $\yps^N_{\mbfk,\mbfl}$. It follows from \lemma{lem4.2.3n} that

3333: $\yps^N_{\mbfk,\mbfl}$ is transversal to the zero section. Consequently,

3334: $\whcalm_{\mbfk,\mbfl}$ is a submanifold of $\whcalm_{=\mbfk}$ of codimension 

3335: $2(n-1)|\mbfl|$. The claim of the theorem follows now by taking the 

3336: $\bfg$-quotient.\qed

3337: 

3338:  

3339: 

3340: 

3341: \newsubsection[4.3a]{Curves with cusps of prescribed type}

3342: In this paragraph we give a construction of $J$-curves of any given cusp type,

3343: completing the result of Micallef and White. In particular, we obtain a

3344: more direct and constructive proof of \lemma{lem1.2.1} without referring

3345: to local structure of minimal surfaces, as  is done in \cite{Mi-Wh}. Then 

3346: we show that the set of cusp-curves with prescribed cusp type is a Banach 

3347: submanifold of the total moduli space and compute its codimension.

3348: 

3349: \smallskip

3350: Let $J$ be an almost complex structure on the ball $B \subset \cc^n$ such that 

3351: $J(0) =J\st(0)$. We assume that $J$ is $C^\ell$-smooth with $\ell \ge 2$.

3352: First we consider the local structure of multiple maps.

3353: 

3354: \newlemma{lem4.3a.1} Let $u: \Delta \to B$ be a non-constant $J$-holomorphic 

3355: map with $u(0)=0 \in B$. Then there exist a radius $r>0$, a uniquely defined 

3356: $\nu\in \nn$, and a non-multiple $J$-holomorphic map $u': \Delta(r^\nu) \to B$

3357: such that $u(z) = u'(z^\nu)$ for $z\in \Delta(r)$. 

3358: \end{lem}

3359: 

3360: \proof By \lemma{lem1.2.4}, $u(z) = v \cdot z^\mu + O(|z|^{\mu + \alpha})$

3361: with some $v\in T_0B=\cc^n$, positive $\mu \in\nn$, and $\alpha>0$. If $\mu 

3362: =1$, then $u$ is already non-multiple in some $\Delta(r)$ and there are

3363: nothing to prove. Thus we may assume that $\mu \ge 2$. 

3364: 

3365: Take a sufficiently small $\rho_0>0$ and consider $U\deff u\inv(B(\rho_0))$.

3366: By the first part of \lemma{lem1.2.4}, $U$ is a disc and $u$ is an immersion 

3367: in $U\bs \{0\}$. Using the second part of \lemma{lem1.2.4} it is not difficult

3368: to show that $u(U \bs \{0\})$ is an immersed $J$-holomorphic punctured disc 

3369: in $B$. Therefore the restriction $u\ogran_U$ is a composition 

3370: of a non-multiple $J$-holomorphic map and a covering branched only in 

3371: $0\in U$.\qed%\hfill$\square$

3372: 

3373: \medskip

3374: It is known that any non-multiple holomorphic map $u: \Delta \to \cc^n$, 

3375: in appropriate holomorphic coordinates on $\Delta$ and $\cc^n$,

3376: has locally the form

3377: \begin{equation}

3378: \notag%\eqqno(4.3a.0a)

3379: u(z) = \sum_{i=0}^l v_i z^{p_i},

3380: \end{equation}

3381: with the following properties. $p_0 = \ord_0(du) +1$, $v_0 \not=0$, the vectors

3382: $v_i \in \cc^n$ are linearly independent of $v_0$ for $i>0$, and $\gcd(p_0,\ldots, 

3383: p_l)=1$. We want to establish a similar result for pseudoholomorphic curves, 

3384: replacing the operation $u_{i-1}(z) \mapsto u_{i-1}(z) + v_i z^{p_i}$ by 

3385: $u_{i-1}(z) \mapsto \dfrm_{p_i}(u_{i-1}, v_i)$. 

3386: 

3387: \newlemma{lem4.3a.2} Let $B \subset \cc^n$ be the unit ball, $J$ a $C^\ell$-smooth

3388: almost complex structure on $B$ with $J(0)= J\st$, and $u: \Delta \to B$ a 

3389: non-multiple $J$-holomorphic map such that $u(z)= v_0 z^{p_0}  +o(z^{p_0})$ for 

3390: some $p_0>1$ and $v_0 \not=0 \in \cc^n$. Take a divisor $d>1$ of $p_0$ and 

3391: denote by $\eta$ a primitive $d$-th root of unity. Then there exist an integer 

3392: $q>0$, a vector $v\in \cc^n$, and a complex polynomial $\psi(z)$ such that

3393: 

3394: \sli $q$ is {\sl not} a multiple of $d$;

3395: %\lineeqqno(4.3a.1a)

3396: 

3397: \slii $v$ is $\cc$-linearly independent of $v_0;$ in particular, $v \not=0;$

3398: 

3399: \sliii $\psi(z)= z + o(z)$ and $\deg\psi(z)\le q;$ 

3400: 

3401: \sliv $u(\eta z) = u( \psi(z))  + z^q \cdot v + o(z^q).$ 

3402: \lineeqqno(4.3a.2a)

3403: 

3404: %\slv Moreover, for any complex polynomial $\phi(z)$ with $\phi(z)=z +o(z)$ 

3405: %\begin{equation}

3406: %\eqqno(4.3a.2b)

3407: %u(\eta \phi(z))= u( \psi(z))  + z^q \cdot v + o(z^q)

3408: %\end{equation}

3409: \end{lem}

3410: 

3411: \proof Denote by $v_0^\perp\subset \cc^n$ a complex orthogonal complement to 

3412: $v_0$, and by $B^\perp(\rho)$ the ball of radius $\rho$ in $v_0^\perp$. 

3413: Note that we can canonically identify the space $v_0^\perp$ with the fiber 

3414: $(N_u)_{z=0}$ of the normal bundle $N_u$ of $u$ (see \refdefi{def1.5.1}). 

3415: Fixing a holomorphic frame $w_1(z), \ldots w_{n-1}(z)$ of $N_u$ we can

3416: identify $\Delta \times v_0^\perp$ with the total space of $N_u$ over $\Delta$

3417: and use $(z, w_1, \ldots w_{n-1})$ as coordinates in $\Delta \times v_0^\perp$.

3418: Denote by $J\st$ the standard integrable complex structure in $\Delta  \times 

3419: v_0^\perp$. It coincides with the canonical holomorphic structure in $N_u$. Set

3420: $$

3421: U_{r,\rho} \deff \Delta(r) \times B^\perp(\rho)

3422: $$

3423: Fix a holomorphic splitting $F_0: N_u \to E_u$ of the projection $\pr_N : E_u 

3424: \to N_u$. We shall identify $N_u$ as a subbundle of $E_u$ by means of $F_0$.

3425: Define the map $F: U_{r,\rho}  \to \cc^n$ as the composition

3426: $$

3427: (z,w) \mapsto F_0(z)(w) \in (E_u)_z = T_{u(z)}B=\cc^n  

3428: \mapsto F(z,w) \deff u(z) + F_0(z)(w).

3429: $$

3430: It is not difficult to see that for sufficiently small $r$ and $\rho$ the map 

3431: $F=F(z,w)$ takes values in $B$ and has the following properties:

3432: 

3433: \begin{itemize}

3434: \item $F(z,w)$ is $C^1$-smooth;

3435: \item $F(z,0) = u(z)$ and $\nabla_{\dot w}F(z,0)=\dot w$; or more precisely,

3436: $\nabla_{\dot w}F(z,0)=F_0(z)(\dot w)$;

3437: \item the pulled-back structure $\ti J \deff F^*J$ coincides with $J\st$ along

3438: the set $\check\Delta \times \{0\}$, \ie 

3439: \begin{equation}

3440: \ti J(z,0) = J\st(z,0).

3441: \eqqno(4.3a.3a)

3442: \end{equation}

3443: \end{itemize}

3444: 

3445: 

3446: From \eqqref(4.3a.3a) we obtain a uniform estimate 

3447: \begin{equation}

3448: \eqqno(4.3a.4a)

3449: \bigl|\ti J(z,w) - J\st(z,w)\bigr| \le C \cdot |w|.

3450: \end{equation}

3451: Further, $\eta^{p_0}=1$ obviously gives $u(\eta z) -u(z)= o(z^{p_0}) = o(z^d)$. 

3452: This implies that for sufficiently 

3453: small $r'$ we can represent $u(z)$ in the form $u(z) = F(\zeta(z), \ti w(z))$ 

3454: with uniquely defined $C^1$-smooth $\zeta(z): \Delta(r') \to \Delta(r)$ and 

3455: $\ti w(z): \Delta(r') \to B^\perp(\rho)$ fulfilling the condition $\zeta(z)= z 

3456: +o(z)$. Further, $\ti w(z)= o(z^d)$.

3457: 

3458: Set $\ti u(z) \deff \bigl( \zeta(z), \ti w(z) \bigr)$. We obtain a $C^1$-smooth 

3459: map $\ti u(z): \Delta(r') \to U_{r,\rho}$, for which

3460: $$

3461: |J\st(\ti u(z)) - \ti J(\ti u(z)) | = 

3462: |J\st (\zeta(z), \ti w(z))- \ti J(\zeta(z), \ti w(z)) | 

3463: \le C'\cdot |\ti w(z)|.

3464: $$

3465: Consequently

3466: $$

3467: \bigl|\dbar_{J\st} \ti u(z) \bigr|= 

3468: \bigl|\dbar_{J\st} \ti u(z) - \dbar_{\ti J} \ti u(z) \bigr|= 

3469: \bigl|\bigl(J\st(\ti u(z)) - \ti J(\ti u(z)) \bigr) \d_y\ti u(z)\bigr|

3470: \le C''\cdot |\ti w(z)|,

3471: $$

3472: or explicitly for components $\zeta(z)$ and $\ti w(z)$

3473: \begin{align}

3474: \bigl|\dbar_{J\st} \ti w(z) \bigr| & \le C''\cdot |\ti w(z)|;

3475: \eqqno(4.3a.5a)\\

3476: \bigl|\dbar_{J\st} \zeta(z) \bigr| & \le C''\cdot |\ti w(z)|.

3477: \eqqno(4.3a.6a)

3478: \end{align}

3479: Observe that $\ti w(z)$ is not identically zero. Otherwise we would obtain

3480: that $u(\eta z) = u(\zeta(z))$, which would contradict the condition of 

3481: non-multiplicity of $u(z)$.

3482: 

3483: Hence, by \lemma{lem1.2.1}, $\ti w(z) = z^q v + o(z^q)$ and $\zeta(z)= \psi(z)

3484: + o(z^q)$ for some $q>0$, non-zero $v\in v_0^\perp$, and a complex polynomial 

3485: $\psi(z)$ of degree $\le q$. Substituting these relations in $u(z)= 

3486: F(\ti u(z))$ we obtain \eqqref(4.3a.2a).

3487: 

3488: Finally, the identity $\sum_{j=1}^d \bigl(u(\eta^j z) - u(\eta^{j-1} z) \bigr)

3489: \equiv 0$ together with \eqqref(4.3a.2a) implies that $\sum_{j=1}^d (\eta^{j-1}

3490: z)^q \cdot v =0$. Thus $\sum_{j=1}^d \eta^{jq} =0$ which is possible \iff $q$

3491: is not a multiple of $d$.  \qed

3492: 

3493: \medskip

3494: Iterating the construction of \lemma{lem4.3a.2}, we obtain

3495: 

3496: \newcorol{cor4.3a.3} Let $B \subset \cc^n$ be the unit ball, $J$ a $C^\ell

3497: $-smooth almost complex structure in $B$ with $J(0)= J\st$, and $u: \Delta \to B$ 

3498: a non-multiple $J$-holomorphic map with $u(0)=0$.

3499: 

3500: Then there exist uniquely defined sequences $(p_0, p_1, \ldots, p_l)$ and 

3501: $(d_0, d_1, \ldots, d_l)$ of positive integers with the following properties:

3502: 

3503: \sli $p_0 = \ord_0 du +1$, so that $u(z) = z^{p_0} v_0 + o(z^{p_0})$ with 

3504: non-zero $v_0 \in \cc^n$;

3505: 

3506: \slii $d_i =  \gcd(p_0, \ldots, p_i);$

3507: 

3508: \sliii $p_i< p_{i+1}$, $d_i > d_{i+1}$, and $d_l=1;$ 

3509: in particular, $p_{i+1}$ is not a multiple of $d_i$;

3510: \lineeqqno(4.3a.7a)

3511: 

3512: \sliv if $\eta_i$ is the primitive $d_i$-th root of unity, then 

3513: $$

3514: u(\eta_i z) = u(\psi_i(z)) \cdot v_0 +  z^{p_{i+1}} \cdot v_{i+1}

3515: + o(z^{p_{i+1}})

3516: $$

3517: for appropriate complex polynomials $\psi_i(z)$ with $\psi_i(z)=z+o(z)$, and 

3518: vector $v_{i+1}\in \cc^n$, $\cc$-linearly independent of $v_0$.

3519: \end{corol}

3520: 

3521: %\proof The claim follows immediately from \lemma{lem4.3a.2}. \qed

3522: 

3523: \newdefi{def4.3a.1} \sli To any increasing sequence of positive integers

3524: $1\le p_0 < p_1 < \cdots < p_l$ we associate the {\sl sequence of divisors}

3525: $d_i \ge d_1 \ge \cdots \ge d_l$ defined by $d_i =  \gcd(p_0, \ldots, p_i)$.

3526: In particular, $d_0 =p_0$. 

3527: 

3528: \slii A sequence $\vec p= (p_0,p_1,\ldots, p_l)$ of positive integer exponents

3529: is called a {\sl cusp type} if $p_i$ and the associate divisors $d_i = \gcd

3530:  (p_0, \ldots, p_i)$ satisfy the condition \eqqref(4.3a.7a). In the situation of 

3531: \refcorol{cor4.3a.3}, the sequence $\vec p= (p_0,p_1, \ldots, p_l)$ is called

3532: the {\sl cusp type of $u$ at $z=0$}, $p_i$ the {\sl critical exponents of $u$

3533: at $z=0$}, and $\vec d =(d_i)$ the {\sl sequence of divisors of $u$ at $z=0$}. 

3534: 

3535: \sliii For a given cusp type $\vec p= (p_0,p_1, \ldots, p_l)$, an integer $p'$ 

3536: is called an {\sl admissible exponent} if $p'$ equals $p_l$ or is of the 

3537: form $p'= p_i+ j\cdot d_i$ for some $i=0,\ldots,l-1$ and $j=0,\ldots, l_i$, 

3538: $l_i \deff \left[ {p_{i+1} - p_i \over d_i } \right]$. Thus all critical 

3539: exponents are admissible and there are exactly $l_i$ non-critical admissible 

3540: exponents between $p_i$ and $p_{i+1}$. Denote by $\vpp = (p'_0, \ldots, 

3541: p'_{l'})$ the sequence of the admissible exponents ordered by growth. Its 

3542: length  is $l' = l + \sum_{i=0}^{l-1} l_i = l + \sum_{i=0}^{l-1}\left[ 

3543:  {p_{i+1} - p_i \over d_i } \right]$.

3544: 

3545: Note that the corresponding sequence of divisors $d'_j \deff \gcd(p'_0, 

3546: \ldots, p'_j)$ consists of divisors $d_i$ of critical exponents, repeated $l_i+

3547: 1$ times. Vice versa, an admissible exponent $p'_j>p'_0=p_0$ is critical \iff 

3548: $d'_j < d'_{j-1}$.

3549: \end{defi}

3550: 

3551: \smallskip

3552: \newthm{thm4.3a.4} Let $B \subset \cc^n$ be the unit ball, $J$ an almost complex 

3553: on $B$ with $J(0)= J\st$, and $u: \Delta \to B$ a non-multiple $J$-holomorphic map

3554: such that $u(0)= 0$. Further, let $\vec p= (p_0, \ldots, p_l)$ and $\vpp= (p_0, 

3555: \ldots, p_{l'})$ be the sequences of critical and resp.\ admissible exponents 

3556: of\/ $u$ at\/ $z=0$, and $\vdp =(d'_0, \ldots, d'_{l'})$ the corresponding 

3557: sequence of divisors.

3558: 

3559: Then there exist a sequence $(v_0, \ldots, v_{l'})$ of vectors in $\cc^n$ (one 

3560: $v_j$ for each $p'_j$), a complex polynomial $\phi(z)$, and a radius $r>0$, such 

3561: that the following holds.

3562: 

3563: \sli $u(z)= z^{p_0}\cdot v_0 + o(z^{p_0})$; $v_0 \not=0$, $v_1, \ldots, v_{l'}$ 

3564: are complex orthogonal to $v_0;$

3565: \lineeqqno(4.3a.2)

3566: 

3567: 

3568: \slii $\phi(z)=z + o(z)$ and $\deg \phi(z) \le p_l-p_0 +1;$

3569: \lineeqqno(4.3a.3)

3570: 

3571: 

3572: \sliii for appropriately chosen maps $\dfrm$, the recursive formula

3573: \begin{equation}

3574: \eqqno(4.3a.4)

3575: u_j(z) \deff \dfrm_{p'_j/d'_j}\bigl(u_{j-1}(z^{d'_{j-1}/d'_j}), J; v_j\bigr) 

3576: \quad j=0,1,\ldots,l'

3577: \end{equation}

3578: beginning from $u_{-1}(z) \equiv 0$ yields a sequence of well-defined 

3579: $J$-holomorphic maps $u_j: \Delta(r^{d'_j}) \to B$ with the property

3580: \begin{equation}

3581: \eqqno(4.3a.5)

3582: u(\phi(z)) - u_j(z^{d'_j}) = v_{j+1} z^{p'_{j+1}} + o(z^{p'_{j+1}}).

3583: \end{equation}

3584: 

3585: \noindent

3586: Moreover, such $v_j$ and $\phi(z)$ are uniquely defined. Further, $v_j$ is 

3587: non-zero if $p'_j$ is  critical.

3588: \end{thm}

3589: 

3590: \proof The choice of the maps $\dfrm_d $ ensuring that at each recursive step 

3591: the right hand side of \eqqref(4.3a.5) is well-defined will be made below. Now

3592: we assume that for given $j<l'$ we have constructed a $J$-holomorphic map $u_j:

3593: \Delta(r^{d'_j}) \to X$ and a polynomial $\phi(z)$ such that $\phi(z)= z+o(z)$

3594: and $u(\phi(z)) = u_j(z^{d'_j}) +o(z^{p'_j})$. Then by \lemma{lem1.2.4},

3595: \begin{equation}

3596: \eqqno(4.3a.6)

3597: u(\phi(z)) = u_j(z^{d'_j}) + z^q w +  o(z^q)

3598: \end{equation}

3599: for some non-zero $w \in \cc^n$ and $q>p'_j$. Represent $w\in \cc^n$ in the form 

3600: $w= a \cdot v_0 + w'$ and replace $\phi$ by $\phi'(z) \deff \phi(z)- {a\over p_0}

3601: \cdot z^{q-p_0 +1}$. The relations $u(z) = z^{p_0} v_0 + o(z^{p_0})$ and 

3602: \eqqref(4.3a.6) yield

3603: \begin{equation}

3604: \eqqno(4.3a.6')

3605: u(\phi'(z)) = u_j(z^{d'_j}) + z^q w' +  o(z^q).

3606: \end{equation}

3607: If $w' =0$, we can consider \eqqref(4.3a.6) with some $q'>q$. Thus we may assume

3608: that $w' \not=0$. Moreover, we see that $\phi(z)$ is defined uniquely by 

3609: \eqqref(4.3a.5) up to degree $p'_j -p_0 +1$.

3610: 

3611: Denote by $\eta_j$ the primitive $d'_j$-th root of unity. Then 

3612: by \lemma{lem4.3a.2},

3613: \begin{equation}

3614: \eqqno(4.3a.7)

3615: u(\eta_j z) = u(\psi_j(z)) + v z^p + o(z^p)

3616: \end{equation}

3617: for an appropriate polynomial $\psi_j(z) =z +o(z)$ and $v$ linearly independent 

3618: of $v_0$. Moreover, $p$ is the first critical exponent after $p'_j$ in the 

3619: sequence $\vpp$ of the admissible exponents of $u(z)$ at $z=0$. In particular, $p$ 

3620: is not a multiple of $d'_j$. Set $\hat\phi_{j+1}(z) \deff \eta_j\inv\phi_{j+1}

3621:  (\eta_j z)$. Then we obtain $\hat\phi_{j+1}(z) = z+ o(z)$ and

3622: \begin{align}

3623: u(\phi_{j+1}(\eta_j z)) &= u(\eta_j\hat\phi_{j+1}(z)) = 

3624: u(\psi_j(\hat\phi_{j+1}(z))) + z^p v + o(z^p)

3625: \notag\\

3626: &= u(\phi_{j+1}(\hat\psi_j(z))) + z^p v + o(z^p),

3627: \eqqno(4.3a.8)

3628: \end{align}

3629: where $\hat\psi_j(z)$ is a polynomial with $\hat\psi_j(z)= z +o(z)$ and 

3630: $\hat\psi_j(\hat\phi_{j+1}(z)) = \phi_{j+1}(\hat\psi_j(z)) +o(z^p)$.

3631: Substitution of \eqqref(4.3a.6') in \eqqref(4.3a.8) together with the identity

3632: $\eta_j^{d'_j}=1$ yields

3633: \begin{equation}

3634: \eqqno(4.3a.9)

3635: u_j( z^{d'_j})  + \eta_j^q z^q w' = 

3636: u_j\bigl( \hat\psi_j^{d'_j}(z) \bigr) 

3637: + z^q w' + v z^p + o(z^{\min(p,q)}).

3638: \end{equation}

3639: Further, since $\hat\psi_j(z)= z +o(z)$, we can find a polynomial $\ti\psi_j(z)$ 

3640: with the properties $\ti\psi_j(z)= z +o(z)$ and $\hat\psi_j^{d'_j}(z)= \ti\psi_j

3641:  (z^{d'_j})$. For such $\ti\psi_j(z)$, the relation \eqqref(4.3a.9) transforms to

3642: \begin{equation}

3643: \eqqno(4.3a.10)

3644: u_j( z^{d'_j})  =

3645: u_j\bigl( \ti\psi_j(z^{d'_j}) \bigr) 

3646: + (1 - \eta_j^q) z^q w' + v z^p + o(z^{\min(p,q)}).

3647: \end{equation}

3648: Assume that $q<p$. Then $q$ is a multiple of $d'_j$. In particular, $q\ge 

3649: p'_{j+1}$. In the case $q> p'_{j+1}$ we simply set $v_{j+1} \deff 0$ and obtain

3650: the relation \eqqref(4.3a.5). In the case $q= p'_{j+1}$ we set $v_{j+1} \deff w'$

3651: and come to the relation \eqqref(4.3a.5) again. The case $p<q$ is impossible since

3652: $p$ is not a multiple of $d'_j$. 

3653: 

3654: In the remaining case $q=p$ we have two subcases, $p'_{j+1} <p$ and $p'_{j+1} 

3655: =p$. Then we set $v_{j+1} \deff 0$ or respectively $v_{j+1} \deff w'$ and obtain 

3656: \eqqref(4.3a.5) from \eqqref(4.3a.6'). 

3657: 

3658: 

3659: \medskip

3660: Now we construct the maps $\dfrm_{p'_j/d_j}$ with the desired properties. The idea

3661: is to rescale the maps $u_j$ making the norms $\norm{du_j}_{L^2}$ sufficiently 

3662: small and obtaining a recursive apriori estimate on $v_j$. For this fix some 

3663: $r \in\; ]0,1[$ and maps $\wt \dfrm_p$ with the properties listed in 

3664: \refdefi{def4.1.1}. Then the substitutions $\ti u(z) \deff u(r z)$, 

3665: $\ti u_j(z) \deff u_j(r^{d'_j} z)$, 

3666: $\ti v_j \deff r^{p'_j}v_j$, and $ \ti \phi_j(z) \deff r\inv \phi_j(r z)$

3667: transform \eqqref(4.3a.4) and \eqqref(4.3a.5) into recursive relations

3668: \begin{equation}

3669: \eqqno(4.3a.a6)

3670: \llap{$\ti u_j(z)= $}\;\wt\dfrm_{p'_j/d'_j}

3671: \bigl(\ti u_{j-1}(z^{d'_{j-1}/d'_j}), J; 

3672: \ti v_j\bigr),

3673: \end{equation}

3674: \begin{equation}

3675: \eqqno(4.3a.a7)

3676: \ti u(\ti \phi(z)) - \ti u_j(z^{d'_j})= 

3677: \ti v_{j+1} z^{p'_{j+1}} + o(z^{p'_{j+1}}).

3678: \end{equation}

3679: for $J$-holomorphic maps $\ti u_j: \Delta \to B$. Note that $\norm{d\ti u}

3680: _{L^2(\Delta)} = \norm{du}_{L^2(\Delta(r))}$ will be arbitrarily small for $r$ 

3681: small enough. Choosing an appropriate $r\ll1$ and using induction, on can obtain 

3682: sufficiently small upper bounds on $\ti v_j$, ensuring that \eqqref(4.3a.a6) is

3683: well-defined for $j=0,\ldots, l$. For such $r$, we define $\dfrm_{p'_j/d_j}$

3684: by the reverse substitutions in \eqqref(4.3a.a6).

3685: \qed

3686: 

3687: 

3688: \smallskip

3689: \state Remark. For almost complex surface, \ie in the case $n=2$, the critical 

3690: exponents determine a topological type of a cusp. In particular, under 

3691: hypotheses of \refthm{thm4.3a.4}, the intersection of the image $u(\Delta)$ 

3692: with the sphere $S^3_r$ of a sufficiently small radius $r>0$ is an iterated 

3693: toric knot $\gamma$ transversal to the 2-plane distribution $\xi$ on $S^3_r$ 

3694: given by $\xi_x \deff T_xS^3_r \cap J(x) T_xS^3_r$. Thus the Bennequin index 

3695: $\beta(\gamma, \xi)$ is well-defined. We refer to \cite{Iv-Sh-1} for the proof

3696: of the formula $\beta= 2\delta -1$ relating the Bennequin index $\beta$ of 

3697: $\gamma$ and the nodal number $\delta$ of $u(\Delta)$ in $0\in B$. On the 

3698: other hand, $\delta$ can be computed by the formula 

3699: \begin{equation}

3700: \eqqno(4.3a.11a)

3701: \delta = \sum_{i=1}^m (d_{i-1}- d_i)(p_i -1),

3702: \end{equation}

3703: see \cite{Rf} or \cite{Mil}. In the higher dimensional setting, 

3704: \ie for $n\ge3$, the topological type of the 

3705: cusp $u(\Delta)$ is not determined by the critical exponents and depends on 

3706: additional information encoding further linear relations between $v_j$. For 

3707: example, the condition {\sl $v_2$ and $v_1$ are linearly dependent} defines a 

3708: proper subset in $\scrp_{\vec p}(\Delta, 0;B)$. Moreover, using the techniques

3709: of this paragraph one can show that this subset is a $C^{\ell-1}$-smooth 

3710: submanifold in $\scrp_{\vec p}(\Delta, 0;B)$. Details can be recovered by an 

3711: interested reader.

3712: 

3713: 

3714: 

3715: 

3716: \smallskip

3717: \newthm{thm4.3a.5} Let $B \subset \cc^n$ be the unit ball, $\vec p= (p_0,\ldots,

3718: p_l)$ a cusp type, $\vpp= (p_0, \ldots, p_{l'})$ the corresponding sequences of 

3719: admissible exponents, and $\vdp =(d'_0, \ldots, d'_{l'})$ the sequence of 

3720: divisors associated with $\vpp$. Then the set

3721: \begin{equation}

3722: \eqqno(4.3a.11)

3723: \llap{$\scrp_{\vec p} (\Delta, 0$};B) \deff 

3724: \bigl\{ (u,J) \in \scrp_{p_0-1} (\Delta, 0; B):

3725: u \text{ \rm has a cusp type $\vec p$ in $z=0$}\; \bigr\}

3726: \end{equation}

3727: is a $C^{\ell-1}$-smooth submanifold of $\scrp_{p_0-1} (\Delta, 0;B)$ of real

3728: codimension $2(n-1)(p_l -p_0-l')$.

3729: \end{thm}

3730: 

3731: Note that $p_0=p'_0$ and $p_l= p'_{l'}$.

3732: 

3733: \proof Let $u:\Delta(r) \to B$ be a $J$-holomorphic map, $r>0$, and let

3734: \begin{equation}

3735: \scrp_k(\Delta, u; B,J) \deff 

3736: \{ u'\in \scrp(\Delta; B, J) : u(z) -u'(z) = o(z^k) \} 

3737: \notag

3738: \end{equation}

3739: By \refthm{thm4.2b.1}, $\scrp_k(\Delta, u; B,J)$ is a $C^{\ell-1} $-smooth 

3740: submanifold of $\scrp(\Delta; B, J)$ of codimension $2n(k+1)$. For $l>k$ it 

3741: follows that $\scrp_l(\Delta, u; B,J)$ has codimension $2n(l-k)$ in $\scrp_k(

3742: \Delta, u; B,J)$. Moreover, if $J_y$ is a $C^\ell$-smooth family of almost complex 

3743: structures in $B$ parameterized by a (Banach) manifold $\scry$ and $u_y\in 

3744: L^{1,p}(\Delta(r), B)$ a $C^{\ell-1}$-smooth family of $J_y$-holomorphic maps, 

3745: then $\cup_{y\in \scry} \scrp_k(\Delta, u_y; B,J_y)$ is a $C^{\ell-1} $-smooth 

3746: manifold.

3747: 

3748: \smallskip

3749: For a given $(u^*, J^*) \in \scrp_{\vec p} (\Delta, 0;B)$, let $v^*_0, \ldots, 

3750: v^*_{l'}$ and $\phi^*(z) = z + \phi^*_2 z^2 + \phi^*_3 z^3+ \cdots$  be the

3751: parameters of $u^*$ constructed in \refthm{thm4.3a.4}. Define $\scry$ to be

3752: the space of small deformations of $v^*_j$ and $\phi^*_i$. This means that $y

3753: \in \scry$ is a tuple $(v_0, \ldots, v_{l'}; \phi_2, \ldots, \phi_{p_l-p_0+1})$

3754: with $v_j \in \cc^n$ and $\phi_i \in \cc$ satisfying $|v_j - v^*_j| <\epsi$ 

3755: and $|\phi_i - \phi^*_i| < \epsi$ with $\epsi$ sufficiently small. Further, 

3756: let $U$ denote a sufficiently small neighborhood of $J^*$ in the space of 

3757: $C^\ell$-smooth almost complex structures in $B$. For $y=(v_0, \ldots;\phi_2, 

3758: \ldots) \in \scry$ and $J\in U$ we construct the maps $u_{y,J; j}$, $j=0, 

3759: \ldots, l'$, using the recursive relation \eqqref(4.3a.4) and set $\phi_y(z) 

3760: \deff z + \phi_2 z^2 + \cdots +\phi_{p_l -p_0 +1} z^{p_l -p_0 +1}$. Then for 

3761: $|z| < r' \ll1$ the inverse map $\phi_y\inv(z)$ is well-defined and 

3762: holomorphic. Define $u_{y,J}(z) \deff u_{y,J; l'}(\phi\inv(z))$. We obtain a 

3763: $C^{\ell-1}$-smooth family of pseudoholomorphic maps $u_{y,J}: \Delta(r') \to B$ 

3764: parameterized by $(y,J) \in \scry\times U$. 

3765: 

3766: Note that by \refthm{thm4.3a.4} every $(u,J) \in \scrp_{\vec p} (\Delta, 0;B)$

3767: sufficiently close to $(u^*, J^*)$ lies in $\scrp_{p_l}(\Delta, u_{y,J}; B,J)$ 

3768: for an appropriate $y\in \scry$, and such $y\in \scry$ is uniquely defined. 

3769: Thus the union $\cup_{(y,J)\in \scry\times U} \scrp_{p_l}(\Delta, u_{y,J}; B,

3770: J)$ is a local $C^{\ell-1}$-smooth chart for $\scrp_{\vec p} (\Delta, 0;B)$. 

3771: 

3772: Finally, note that the union $\cup_{(y,J)\in \scry\times U} \scrp_{p_0-1}

3773:  (\Delta, u_{y,J}; B,J)$ is naturally isomorphic to $\scrp_{p_0}(\Delta, 0; B)

3774: \times \scry$. Computing the number of parameters we obtain the codimension of

3775: the imbedding $\scrp_{\vec p} (\Delta, 0;B) \hook \scrp_{p_0} (\Delta, 0;B)$. 

3776: \qed

3777: 

3778: Globalizing \refthm{thm4.3a.5} we obtain

3779: 

3780: \newcorol{cor4.3a.6} Let $\vec \mbfp =(\vec p_1, \ldots, \vec p_m)$ be a

3781: sequence of cusp types, $\vec p_i= (p_{i,0}, \ldots p_{i, l_i})$. Set 

3782: $k_i \deff p_{i,0} -1$ and $\mbfk\deff (k_1, \ldots, k_m)$. Then the space

3783: $$

3784: \scrm_{\vec \mbfp} \deff \{ [u,J; z^*_1, \ldots z^*_m] \in \scrm_\mbfk:

3785: u \text{ \sl has cusp type $\vec p_i$ in }z^*_i\; \}

3786: $$

3787: is a $C^{\ell-1}$-smooth submanifold of $\scrm_\mbfk$ of codimension

3788: $2(n-1)\sum_{i=0}^m (p_{i,l_i} -p_{i,0}-l'_i)$.

3789: \end{corol}

3790: 

3791: 

3792: \medskip

3793: \newsection[3]{Saddle points in the moduli space}

3794: 

3795: \newsubsection[3.1]{Critical and saddle points in the moduli space}

3796: In application of the continuity method for constructing $J$-holomorphic

3797: curves two main difficulties occur. The first one appears in the

3798: proof of the ``closedness part'', when one tries to extend a deformation $[u_t, 

3799: J_t] \in \scrm, t\in [0,t'[\;$ into the endpoint $t'$. This difficulty is 

3800: connected with the fact that the projection $\pi_{\!\!\scrj}: \scrm \to \scrj$ 

3801: is, in general, not proper. In particular, for a path $J_t\in \scrj, t\in [0,t']$ 

3802: there may not exist a lift $[u_t, J_t]$ to $\scrm$, and the fibers

3803: $\scrm_{\!\!\scrj} = \pi_J\inv (J)$ can be non-compact. Gromov's compactness

3804: theorem (\cite{Gro}, see also  \cite{Iv-Sh-3}) gives a fiberwise topological

3805: compactification of $\scrm$ by adding certain degenerate curves $C$.

3806: However, for the moment we neglect this difficulty assuming we can avoid it in 

3807: our case.

3808: 

3809: The second main difficulty appears in the proof of the ``openness part'' when one

3810: tries to extend a lift $[u_t, J_t] \in \scrm$, $t \in[0, t']$, of a

3811: path of $J_t\in \scrj, t\in [0, 1]$ to a bigger interval $t \in[0, t''[$ with

3812: some $t''>t'$. Obviously, this difficulty can appear only if $[u_{t'}, J_{t'}]$ 

3813: is a {\sl critical point} of $\pi _{\!\!\scrj}$, \ie when the differential

3814: of the projection $d\pi_{\!\!\scrj}$ is not surjective in $[u_{t'}, J_{t'}]$.

3815: Thus it is desirable to find conditions on the critical points of

3816: $\pi_{\!\!\scrj}$ which ensure the existence of such a lift.

3817: 

3818: \smallskip

3819: Assume additionally that the given path $h: [0,1] \to \scrj$, $h(t)\deff

3820: J_t$, is $C^2$-smooth and transversal to $\pi_h: \scrm \to \scrj$, \ie

3821: $\scrm_h$ is a manifold. Let $\pi_h: \scrm_h \to I$, $I\deff [0,1]$, be the

3822: projection. Then we have a well-defined $C^1$-smooth bundle homomorphism

3823: $d\pi_h: \scrm_h \to \pi_h^*(TI) \cong \rr$. Further, $d\pi_h$ vanishes

3824: exactly at critical points of $\pi_h$ and, by \lemma{lem1.3.1}, at each such

3825: point $p\deff [u,h(t)]$ we have a well-defined quadratic form $\nabla

3826: d\pi_h(p): T_p\scrm_h \to \rr\cong T_tI$. For our purpose it is

3827: sufficient to show that each critical point $p$ is a {\sl saddle}, \ie

3828: the quadratic form $\nabla d\pi_h(p)$ has at least one positive and one

3829: negative eigenvalue.

3830: 

3831: 

3832: \smallskip

3833: It turns out that this condition depends only on the geometry of the projection

3834: $\pi_{\!\!\scrj}:\scrm \to \scrj$ at $p=[u,h(t)]$, and not on the particular choice

3835: of a transversal map $h:I \to \scrj$. In more detail, the situation is as

3836: follows.

3837: 

3838: 

3839: \smallskip

3840: First, since $\scrm$ is $C^\ell$-smooth with $\ell\ge2$, the map $\pi_{\!\!\scrj}:

3841: \scrm \to \scrj$ defines a $C^1$-smooth homomorphism of Banach bundles $d\pi_{\!\!

3842: \scrj}: T\scrm \to \pi_{\!\!\scrj}^*(T\scrj)$. \refcorol{cor2.2.2} relates

3843: the (co)kernel of $d\pi_{\!\!\scrj}$ for a given $[u,J] \in \scrm$ with $\sfh^i(S,

3844: \scrn_u)$, and \lemma{lem1.3.1} provides a well-defined bilinear map

3845: \begin{equation}

3846: \Phi=\Phi_{[u,J]} \deff \nabla d \pi_{\!\!\scrj}:

3847: T_{[u,J]}\scrm \times \sfh^0(S, \scrn_u)

3848: \to \sfh^1(S, \scrn_u).

3849: \eqqno(3.1.1)

3850: \end{equation}

3851: 

3852: The situation remains essentially the same if we consider a relative moduli

3853: space $\scrm_h= Y\times_h\scrm$ with a $C^\ell$-smooth map $h: Y \to \scrj$

3854: transversal to $\pi_{\!\!\scrj}$. Indeed, one can easily see that for the natural

3855: projection $\pi_h: \scrm_h \to Y$ and a point $[u,y] \in \scrm_h$ with $h(y)

3856: \ddef J$ one has the natural isomorphisms

3857: \begin{equation}

3858: \matrix%\format\r\;&\c\;&\r\;&\c&\l\\

3859: \ker(d\pi_h :T_{[u,y]}\scrm_h \to T_yY)

3860: &\cong& \ker(d\pi_{\!\!\scrj} :T_{[u,J]}\scrm \to T_J\scrj)

3861: &\cong \sfh^0_D(S, \scrn_u),

3862: \cr

3863: \coker(d\pi_h :T_{[u,y]}\scrm_h \to T_yY)

3864: &\cong& \coker(d\pi_{\!\!\scrj} :T_{[u,J]}\scrm \to T_J\scrj)

3865: &\cong \sfh^1(S, \scrn_u).

3866: \endmatrix

3867: \eqqno(3.1.2)

3868: \end{equation}

3869: Further, the relation between $\Phi=\nabla d \pi_{\!\!\scrj}$ and $\nabla d \pi_h$

3870: is given by the following

3871: 

3872: \newlemma{lem3.1.1}

3873: \sli The isomorphism $\coker(d\pi_h)\cong \sfh^1(S,

3874: \scrn_u)$ is induced by composition $T_yY \buildrel dh \over \lrar T_J\scrj

3875: \buildrel \barr \Psi_{u,J} \over {\relbar\!\!\relbar\!\!\lrar}

3876: \sfh^1(S, \scrn_u)$.

3877: 

3878: \smallskip

3879: \baselineskip=14pt

3880: \slii The bilinear map $\nabla d \pi_h: T_{[u,y]}\scrm_h \times \sfh^0_D(S,

3881: \scrn_u) \to \sfh^1(S, \scrn_u)$ is induced by the composition $T_{[u,y]}

3882: \scrm_h \hook T_{[u,J]}\scrm \oplus T_yY \twoheadrightarrow T_{[u,J]}\scrm$

3883: and the bilinear map $\Phi: T_{[u,J]}\scrm \times \sfh^0_D(S,  \scrn_u) \to

3884: \sfh^1(S, \scrn_u)$.

3885: \end{lem}

3886: 

3887: 

3888: \medskip

3889: Summing up, we obtain the following situation in the most important case $Y=I$.

3890: 

3891: \newlemma{lem3.1.2}

3892: For a map $h: I \to \scrj$ transversal to $\pi_J$,

3893: the singular points of the projection $\pi_h: \scrm_h \to I$ are exactly those

3894: $[u,t] \in \scrm_h$ for which $\sfh^1(S, \scrn_u)=\rr$.

3895: 

3896: For such $[u,t] \in \scrm_h$ with $J\deff h(t)$ one has the equality

3897: $T_{[u,t]} \scrm_h = \sfh^0_D(S, \scrn_u)$ and the isomorphism $\barr\Psi

3898: _{[u,J]}: T_tI \buildrel \cong \over \lrar \sfh^1(S, \scrn_u)$. Moreover,

3899: the quadratic form $\Phi_{[u,J]}: \sfh^0_D(S, \scrn_u)\to \sfh^1(S, \scrn_u)$

3900: equals to the composition of the quadratic form $\nabla d\pi_h: T_{[u,t]}

3901: \scrm_h  \to T_tI$ with $\barr\Psi_{[u,J]}: T_tI \to \sfh^1(S, \scrn_u)$.

3902: \end{lem}

3903: 

3904: 

3905: \newcorol{cor3.1.3} The nullity, rank and signature of $\Phi_{[u,J]}$ and

3906: $\nabla d\pi_h$ coincide.

3907: \end{corol}

3908: 

3909: \newdefi{def3.1.1} Let $Q$ be a quadratic form defined on a

3910: (finite-dimen\-sional) vector space $V$ and taking values in a vector space $W$

3911: with $\dimr W=1$. Define the {\sl saddle index of $Q$} by $\sind Q \deff

3912: \min \{ \ind_+ Q, \ind_- Q \}$, where $\ind_\pm Q$ are respectively the

3913: positive and negative indices of $Q$ \wrt some (in fact, any) orientation of

3914: $W$. For a critical point $[u,J]\in \scrm$ with $\sfh^1(S, \scrn_u) \cong \rr$,

3915: call $\sind \Phi_{[u,J]}$ the {\sl saddle index of $[u,J]$}. A point

3916: $[u,J]\in \scrm$ is a {\sl saddle point} of the moduli space $\scrm$ \iff

3917: $\sind \Phi_{[u,J]}$ is strictly positive.

3918: \end{defi}

3919: 

3920: 

3921: 

3922: \newsubsection[3.2]{Second variation of the $\dbar$-equation}

3923: To find saddle points of $\scrm$ we need to find an explicit formula for the

3924: form $\Phi$ in \eqqref(3.1.1). Note that, since the space $\scrp$ appears as 

3925: the zero-set of the $\dbar$-equation \eqqref(1.1.1), the description of the 

3926: tangent space $T\scrp$ is given by the variation of the $\dbar$-equation. 

3927: Similarly, we show that the form $\Phi$ is essentially the part of the second 

3928: variation of the $\dbar$-equation invariant \wrt the choice of a connection 

3929: being used.

3930: 

3931: \smallskip

3932: Let $[u,J]\in\scrm$ be represented by $(u,J_S,J)\in\whcalm$. Recall the

3933: description of $T_{(u,J_S,J)}\whcalm$ given in \eqqref(2.2.4). Moreover, since

3934: $du$ is non-vanishing at a generic point, $\dot J_S$ is determined by $v$ and

3935: $\dot J$. Note that the tangent space to an orbit $\bfg\cdot(u,J) \subset\whcalm$

3936: can be identified with  $du(\sfh^0(S,TS)) \subset \scre_{(u,J_S,J)}$. This

3937: defines a subbundle of $\scre$ which we also denote by $du(\sfh^0(S,TS))$.

3938: Thus we obtain the isomorphism

3939: \begin{equation}

3940: T_{[u,J]}\scrm\cong T_{(u,J_S,J)}\whcalm/du(\sfh^0(S,TS)).

3941: \eqqno(3.2.1)

3942: \end{equation}

3943: Explicitly, the tangent space $T_{[u,J]}\scrm$ consists of triples

3944: $([v], \dot J_S, \dot J)$ for which $(v, \dot J_S,\dot J) \in T_{(u,J_S,J)}

3945: \whcalm$ and $[v]$ is the equivalence class $v + du\bigl(\sfh^0(S,TS) \bigr)$.

3946: 

3947: \newdefi{def3.2.1} Set

3948: \begin{equation}

3949:  \wh\scre_{(u,J_S,J)} \deff

3950: \bigl(\scre_{(u,J_S,J)} /du(\sfh^0(S,TS) \bigr) \oplus \sfh^1(S,TS).

3951: \eqqno(3.2.2)

3952: \end{equation}

3953: Recall the canonical isomorphism $T_{J_S}\ttt_g \cong \sfh^1(S,TS)$ given

3954: by \eqqref(2.2.2). For $(u,J_S,J)\allowbreak\in \whcalm$ define the operator

3955: \begin{equation}

3956: \wh D=\wh D_{u,J}: \wh\scre_{(u,J_S,J)} \to  \scre'_{(u,J_S,J)}

3957: \qquad

3958: \wh D ([v], [I_S]) \deff Dv + J\scirc du \scirc I_S.

3959: \eqqno(3.2.3)

3960: \end{equation}

3961: \end{defi}

3962: 

3963: 

3964: \newlemma{lem3.2.1}

3965: Formula \eqqref(3.2.2) defines a $C^\ell$-smooth Banach bundle $\wh\scre$ 

3966: over $\scrm$ with the fiber $\wh\scre_{(u,J_S,J)}$

3967: over $[u,J]\in \scrm$. The tangent bundle $T\scrm$ can be included in the

3968: following exact sequence of bundles over $\scrm$

3969: \begin{equation}

3970: 0\to T\scrm \buildrel \alpha \over\lrar

3971: \wh\scre \oplus \pi_{\!\!\scrj}^*T\scrj

3972: \buildrel \beta \over\lrar \scre' \to0,

3973: \eqqno(3.2.4)

3974: \end{equation}

3975: where the homomorphisms $\alpha=(\alpha_1,\alpha_2)$ and $\beta=(\beta_1,

3976: \beta_2)$ are given by

3977: \begin{equation}

3978: \matrix%\format \r& \;\c\;& \l \\

3979: \alpha_1([v],\dot J_S, \dot J) &\deff&

3980: ([v], [\dot J_S]) \in \wh\scre=

3981: \scre/du\bigl(\sfh^0(S,TS)\bigr)\bigoplus \sfh^1(S,TS)

3982: \cr

3983: \alpha_2 &\deff& d\pi_{\!\!\scrj}: T_{[u,J]}\scrm \to T_J\scrj

3984: \cr

3985: \beta_1 &\deff& \wh D_{u,J}: \wh\scre_{(u,J_S,J)}\to\scre'_{(u,J_S,J)}

3986: \cr

3987: \beta_2 &\deff& \Psi_{u,J}: T_J\scrj\to\scre'_{(u,J_S,J)}

3988: \endmatrix

3989: \end{equation}

3990: \end{lem}

3991: 

3992: \proof It is easy to show that $\sfh^0(S,TS)$ and $\sfh^1(S,TS)$ can

3993: be considered as smooth bundles over $\whcalm$ equipped with the natural

3994: $\bfg$-action. Then $du$ defines a $\bfg$-equivariant homomorphism between the

3995: bundles $\sfh^0(S, TS)$ and $\scre$. Hence, using formula \eqqref(3.2.2), we 

3996: can construct a bundle $\wh\scre_{\whcalm}$ over $\whcalm$ with the induced

3997: $\bfg$-action. By \lemma{lem2.2.1}\.\slip, this is equivalent to the first

3998: assertion of the lemma.

3999: 

4000: The exactness of \eqqref(3.2.4) follows from relations \eqqref(2.2.4) and 

4001: (\ref{eq3.2.1}--\ref{eq3.2.3}).

4002: \qed

4003: 

4004: \smallskip

4005: \newlemma{lem3.2.2}

4006: The homomorphisms $\alpha_1$ and $\beta_2$ yield isomorphisms

4007: \begin{equation}

4008: \sfh^0_D(S, \scrn_u) \cong \ker \alpha_2 \buildrel

4009: \alpha_1 \over \cong \ker \beta_1

4010: \quad\text{and}\quad

4011: \sfh^1_D(S, \scrn_u) \cong \coker \alpha_2 \buildrel

4012: \beta_2 \over \cong \coker \beta_1,

4013: \eqqno(3.2.5)

4014: \end{equation}

4015: inducing the identity

4016: \begin{equation}

4017: \Phi_{u,J}= -\nabla \wh D: T_{[u,J]}\scrm \times \sfh^0_D(S, \scrn_u)

4018: \to \sfh^1_D(S, \scrn_u).

4019: \eqqno(3.2.6)

4020: \end{equation}

4021: \end{lem}

4022: 

4023: 

4024: \proof The isomorphisms \eqqref(3.2.5) follow from definitions and

4025: \refcorol{cor2.2.2}. Moreover, we can identify $\sfh^0_D(S, \scrn_u)$ with

4026: $\ker\bigl( \wh D_{u,J}: \wh\scre_{(u, J_S, J)} \to \scre'_{(u, J_S, J)}

4027: \bigr)$.

4028: 

4029: Let $i: \sfh^0_D(S, \scrn_u) \to T_{[u,J]}\scrm$ and $p:\scre'_{(u,J_S,J)} \to

4030: \sfh^1_D(S, \scrn_u)$ denote the corresponding inclusion and projection. Fix

4031: some connections on $T\scrm$, $\pi^*_{\!\!\scrj} T\scrj$, $\wh\scre$, and $\scre'$,

4032: and denote all of them simply by $\nabla$. Covariant differentiation of the

4033: relation $\beta_1 \scirc \alpha_1 + \beta_2 \scirc \alpha_2=0$ gives

4034: \begin{equation}

4035: \nabla\beta_1 \scirc \alpha_1 + \nabla\beta_2 \scirc \alpha_2 +

4036: \beta_1 \scirc \nabla\alpha_1 + \beta_2 \scirc \nabla\alpha_2 = 0,

4037: \end{equation}

4038: which together with $\alpha_2\scirc i=0$ and $p\scirc \beta_1=0$ yields

4039: \begin{equation}

4040:   p\scirc \nabla\beta_1 \scirc \alpha_1\scirc i=

4041: - p\scirc \beta_2 \scirc \nabla \alpha_2\scirc i.

4042: %\eqNno\square

4043: \end{equation}

4044: \qed

4045: 

4046: \newdefi{def3.2.2} Using the isomorphisms $\ker \bigl( \wh D_{u,J}: \wh

4047: \scre_{(u,J_S,J)} \to \scre'_{(u,J_S,J)} \bigr) \cong \sfh^0_D(S, \scrn_u)$

4048: from \eqqref(3.2.5) and $\sfh^1(S, TS) \cong T_{J_S} \ttt_g$ from \eqqref(2.2.2),

4049: redefine

4050: \begin{equation}

4051: \sfh^0_D(S, \scrn_u) \deff \bigl\{\, ([v], I_S)\in \wh\scre_{(u,J_S,J)}

4052: \oplus  T_{J_S} \ttt_g\;:\; Dv + J \scirc du \scirc I_S =0 \;\bigr\}.

4053: \eqqno(3.2.7)

4054: \end{equation}

4055: Then the projection $\sfh^0_D(S, \scrn_u) \to \sfh^0_D(S, N_u)$ is given by

4056: the formula $([v], I_S) \mapsto \pr_N(v)$ with $\pr_N: E_u \to N_u$

4057: defined by \eqqref(1.5.2).

4058: \end{defi}

4059: 

4060: \smallskip

4061: Now assume that some symmetric connections on $TX$ and $TS$ are fixed.

4062: They induce connections on $\scre$ and $\scre'$, on the tangent bundles

4063: $T\whcalm$ and $T\scrm$, and so on. We shall use the same notation $\nabla$

4064: for all these connections. Further, denote by $R^X(\cdot,\cdot;\cdot)$ the

4065: curvature operator of the connection $\nabla$ on $X$.

4066: 

4067: \newlemma{lem3.2.3}

4068: For $([v], \dot J_S, \dot J) \in T_{[u,J]}\scrm$

4069: and $([w], I_S) \in \sfh^0_D(S, \scrn_u) \subset \wh\scre_{(u, J_S, J)}\!\!$

4070: 

4071: \smallskip

4072: $\qquad

4073: \bigl(\nabla_{([v],\dot J_S, \dot J)}\wh D\bigr) ([w],I_S)=

4074: \term1{R^X(v,du ;w)} + \term2{J \scirc R^X(v, du\scirc J_S;w)} +

4075: $

4076: \lineeqqno(3.2.8)

4077: 

4078: \smallskip

4079: $\qquad

4080: + \term3{\nabla_vJ\scirc \nabla w \scirc J_S} +

4081: \term4{\nabla^2_{v,w}J \scirc du \scirc J_S} +

4082: \term5{\nabla_wJ\scirc \nabla v \scirc J_S} +

4083: \term6{\dot J\scirc \nabla w \scirc J_S} +

4084: $

4085: 

4086: \smallskip

4087: $\qquad

4088: + \term7{\nabla_w\dot J\scirc du \scirc J_S} +

4089: \term8{J\scirc \nabla w \scirc \dot J_S} +

4090: \term9{\nabla_wJ\scirc du \scirc \dot J_S} +

4091: \term{10}{\nabla_vJ \scirc du \scirc I_S} +

4092: $

4093: 

4094: \smallskip

4095: $\qquad

4096: + \term{11}{J \scirc \nabla v \scirc I_S} +

4097: \term{12}{\dot J \scirc du \scirc I_S}.

4098: $

4099: \end{lem}

4100: 

4101: 

4102: \state Remark. The numerical subscripts on the various terms are for future 

4103: reference.  

4104: %The numeration of particular terms is done for convenience.

4105: 

4106: \proof Consider the bundle $\wt\scre\deff \scre \oplus \sfh^1(S,TS)$

4107: over $\whcalm$ with the bundle homomorphism $\wt D: \wt\scre \to \scre'$,

4108: $\wt D(w, [I_S]) \deff Dw + J \scirc du \scirc I_S$. We claim that for the

4109: covariant derivative $\bigl(\nabla_{([v],\dot J_S, \dot J)}\wt D\bigr)(w,

4110: I_S)$ we obtain the same expression as in the statement of the lemma. Obviously,

4111: this would imply the lemma.

4112: 

4113: The only nontrivial point here is to compute the derivative of the operator of

4114: covariant differentiation $\nabla^\op\deff \nabla: L^{1,p}(S, u^*TX) \to L^p(S,

4115: u^*TX \otimes T^*S)$ in the direction given by some $v \in T_u L^{1,p}(S, X)=

4116: L^{1,p}(S, u^*TX)$. To do this, we fix a smooth vector field $\xi$ on $S$ and

4117: a local section $\mbfw$ of $\scre$. Then $\bigl( \nabla^\op\mbfw \bigr) (\xi)=

4118: \nabla_\xi \mbfw$ is a local section of a Banach bundle with the fiber $L^p(S,

4119: u^*TX)$ over $u\in L^{1,p}(S, X)$.

4120: 

4121: Differentiation in the direction $v$ yields

4122: \begin{equation}

4123: \nabla _v \bigl( \nabla^\op \mbfw  \bigr) (\xi)

4124: = \nabla^2 _{v,\xi} \mbfw=

4125: R^X(v, du(\xi); \mbfw) + \nabla^2_{\xi,v}\mbfw =

4126: \end{equation}

4127: \begin{equation}

4128: R^X(v, du; \mbfw)(\xi) +  \bigl(\nabla^\op_\xi(\nabla  \mbfw) \bigr)(v).

4129: \end{equation}

4130: Thus we obtain the formula $\nabla_v (\nabla^\op)= R^X(v, du; \cdot)$. Besides,

4131: we have the relation $\nabla_v du= \nabla v$, which was already used for

4132: deriving \eqqref(1.3.3) from \eqqref(1.1.1). Now, the proof of the lemma 

4133: can be completed by explicit calculations. \qed

4134: 

4135: Using \eqqref(3.2.8) we can describe in more detail the structure of $\pi

4136: _{\!\!\scrj} :\scrm \to \scrj$ at critical points with $\sfh^1_D(S, N_u) \cong \rr$.

4137: Note that the term $[4]$ in \eqqref(3.2.8) is the only one that depends on 

4138: second order derivatives of $J$. Further, the operator $D=D_{u,J}$ is also 

4139: independent of second order derivatives. Thus, deforming $J$ and preserving 

4140: the order one jet $j^1\!J|_{u(S)}$, the map $u:S \to X$ remains $J$-holomorphic

4141: with same the $D$-cohomology groups $\sfh^i(S, \scrn_u)$. The result of such

4142: changes of $J$ is given by

4143: 

4144: \newlemma{lem3.2.4} Let $[u,J] \in \scrm$ with $\sfh^1_D(S, N_u)\cong

4145: \rr$ and a quadratic form $\ti \Phi: \sfh^0_D(S, N_u)\to \sfh^1_D(S, N_u)$ be

4146: given. Then there exists a $C^1$-small perturbation $\ti J \in \scrj$ of $J$

4147: such that $j^1\! J|_{u(S)} = j^1\! \ti J|_{u(S)}$ and the restriction of

4148: $\Phi_{u, \ti J}$ to $\sfh^0_D(S, N_u)$ equals the given $\ti \Phi$.

4149: Moreover, such a perturbation $\ti J$ of $J$ can be realized in an arbitrarily

4150: small neighborhood of a given point $x \in u(S)$.

4151: \end{lem}

4152: 

4153: 

4154: \proof Let $U\subset X$ be a neighborhood of the given $x$. Find

4155: $U' \subset U$ such that $u\inv(U') \not= \emptyset$ and $u$ is an imbedding

4156: on $u\inv(U')$. Obviously, it is sufficient to find an appropriate jet $j^2\!

4157: \ti J|_{u(S)}$ which differs from $j^2\! J|_{u(S)}$ only in $U'\cap u(S)$.

4158: Then $j^2\! \ti J|_{u(S)}$ can be extended to $\ti J$ with the desired

4159: properties.

4160: 

4161: Covariant differentiation of the identity $J^2 = -\id$ gives the relations

4162: $\nabla_v J \scirc J + J \scirc \nabla_v J=0$ and

4163: \begin{equation}

4164: \nabla^2 _{v_1,v_2}J \scirc J + \nabla_{v_1} J \scirc \nabla_{v_2} J +

4165: \nabla_{v_2} J \scirc \nabla_{v_1}J  + J \scirc \nabla^2 _{v_1,v_2}J =0,

4166: \qquad v_1,v_2 \in T_xX.

4167: \end{equation}

4168: Consequently, we have the following description of the possible choice for 

4169: $j^2\! \ti J|_{u(S)}$ with $j^1\! \ti J|_{u(S)}= j^1\!  J|_{u(S)}$. The tensor

4170: field $u(S) \ni x \mapsto \Theta_x$ defined by

4171: \begin{equation}

4172: v_1, v_2, w \in T_xX \mapsto \Theta_x(v_1, v_2; w)

4173: \deff \nabla^2 _{v_1,v_2}(\ti J -J) (w)\in  T_xX

4174: \end{equation}

4175: must be supported in $U'$, symmetric%

4176: \footnote{\.Obviously, $\nabla^2 _{v_1,v_2}(\ti J -J)- \nabla^2 _{v_2,

4177: v_1} (\ti J -J)$ can be expressed via $\ti J -J$ and the curvature tensor

4178: $R^X(\cdot,\cdot; \cdot)$ of $\nabla$. But $\ti J -J$ vanishes on $u(S)$.}

4179: %

4180: in $v_1$ and $v_2$, $J$-antilinear in $w$, and zero for $v_1,v_2\in T_x(u(S))

4181: \subset T_xX$. Vice versa, any tensor field $\Theta$ with these properties has

4182: the form $\Theta(v_1, v_2; w) = \nabla^2 _{v_1,v_2}(\ti J -J) (w)$ for an

4183: appropriate $\ti J$ with $j^1\! \ti J|_{u(S)}= j^1\! J|_{u(S)}$.

4184: 

4185: The condition that $\Theta_x(v_1, v_2; w)$ vanishes for $v_1, v_2\in T_x(u(S)

4186: )$  means that for $x\in U'\cap u(S)$ we can consider $\Theta_x(v_1, v_2;w)$

4187: as a tensor with arguments $v_1, v_2$ varying in the normal bundle $N_u$.

4188: Now, for $\ti J$ as above, $v\in \sfh^0_D(S, N_u)$, and $\psi \in \sfh^0_D(S,

4189: N_u \otimes K_S) \allowbreak \cong \sfh^1(S, N_u)^*$ we obtain the relation

4190: \begin{equation}

4191: \la\psi, \Phi_{u,\ti J}(v,v) \ra =  \la\psi, \Phi_{u,J}(v,v) \ra +

4192: \re \int_S \psi \scirc \Theta(v,v; du).

4193: \end{equation}

4194: Finally, observe that any quadratic form on a {\sl finite dimensional}

4195: space $\sfh^0_D(S, N_u)$ can be realized as $\re \int_S \psi \scirc \Theta(v,v;

4196: du)$ with $\Theta$ satisfying the conditions stated above. \qed

4197: 

4198: 

4199: 

4200: \newsubsection[3.3]{Second variation at cusp-curves}

4201: Our aim in this paragraph is to find conditions ensuring that a critical 

4202: point $[u,J] \in \scrm$ with $\sfh^1(S, N_u)\cong \rr$ is a saddle point. 

4203: \lemma{lem3.2.4} shows that such critical points with $\scrn\sing_u \cong 0$ 

4204: are ``hopeless'' from this point of view. Hence, we need to understand in more

4205: detail the structure of the bilinear operator $\Phi$ on the component $\sfh^0

4206:  (S, \scrn_u \sing) \subset \sfh^0_D(S, \scrn_u)$.

4207: 

4208: %%\newdefi{def3.3.1} 

4209: Recall that by the definition of the normal sheaf the stalk $(\scrn_u\sing)_z$

4210: at $z \in S$ is non-trivial exactly when $z$ is a cusp-point of $u: S \to X$

4211: and in this case $\dimc (\scrn_u\sing)_z = \ord_z du$. Thus we want to 

4212: understand the structure of the moduli space at critical points corresponding

4213: to cusp-curves. The following two lemmas contain technical results needed for

4214: this purpose. Recall that the holomorphic line bundle $\scro([A])$ was

4215: introduced in \refdefi{def1.5.1}. We maintain the notation $\nabla$ and

4216: $R^X( \cdot{,} \cdot{;} \cdot)$ from \lemma{lem3.2.3}. In particular, we

4217: have $\nabla_\xi J= \nabla_{du(\xi )}J$, $\nabla^2_{\xi, \eta}v - \nabla^2

4218: _{\eta,\xi} v = R^X(du(\xi), du(\eta); v)$, and other similar relations.

4219: Further, we assume that $\nabla J_S=0$.

4220: 

4221: 

4222: \newlemma{lem3.3.1} An element $([w], I_S) \in \sfh^0_D(S, \scrn_u)$

4223: lies in $\sfh^0(S, \scrn\sing_u)$ \iff $w=du(\ti w)$ for some $\ti w \in

4224: L^{1,p}_\loc (S\bs \supp(\scrn_u\sing), TS)$ that extends to

4225: $\ti w \in L^{1,p}(S, TS \otimes \scro([A])\mkern .5mu)$.

4226: 

4227: \smallskip

4228: In this case, outside the zero-set of $du$ one has

4229: \begin{equation}

4230: \dbar \ti w \equiv  \nabla\ti w +J_S \scirc \nabla\ti w \scirc J_S

4231: = - J_S \scirc I_S

4232: \eqqno(3.3.1)

4233: \end{equation}

4234: and

4235: \begin{align}

4236: 0\equiv &\nabla_{\ti w}(D_{u,J}v + \dot J \scirc du \scirc J_S

4237: + J \scirc du \scirc \dot J_S)  =\term{D}{D_{u,J}(\nabla_{\ti w}v) }+

4238: \eqqno(3.3.2) \\

4239: &\term{1'}{R^X(w,du ;v)} + \term{2'}{J \scirc R^X(w, du\scirc J_S;v)} +

4240: \term5{\nabla_wJ\scirc \nabla v \scirc J_S} +

4241: \term{4'}{\nabla^2_{w,v}J \scirc du \scirc J_S} +

4242: \notag\\

4243: &\term{3'}{\nabla_vJ\scirc \nabla_{\ti w}du \scirc J_S} +

4244: \term{6'}{\dot J\scirc \nabla_{\ti w}du \scirc J_S} +

4245: \term7{\nabla_w\dot J\scirc du \scirc J_S} +

4246: \term{8'}{J\scirc \nabla_{\ti w}du \scirc \dot J_S} +

4247: \notag\\

4248: &\term9{\nabla_wJ\scirc du \scirc \dot J_S} +

4249: \term{13}{J\scirc du \scirc \nabla_{\ti w} \dot J_S} -

4250: \term{14}{\nabla v \scirc \nabla \ti w} -

4251: \term{15}{J \scirc \nabla v \scirc \nabla \ti w  \scirc J_S}.

4252: \notag

4253: \end{align}

4254: \end{lem}

4255: 

4256: 

4257: \proof Relation \eqqref(3.3.1) follows from the equality

4258: %\begin{equation}

4259: \begin{align*}

4260: 0&= D_{u,J}(du(\ti w))+  J \scirc du \scirc I_S

4261: = du(\dbar\ti w)+  du \scirc J_S \scirc I_S =

4262: \cr

4263:  &=du \Bigl( ( \nabla \ti w + J_S \scirc \nabla \ti w\scirc J_S)

4264: + J_S \scirc I_S \Bigr).

4265: \end{align*}

4266: %\end{equation}

4267: Using $J_S^2 = -\id$ we can write the relation in the form

4268: $I_S = J_S \scirc \nabla \ti w - \nabla \ti w \scirc J_S$.

4269: 

4270: \medskip

4271: To show \eqqref(3.3.2) we start with the computation of 

4272: $D_{u,J}(\nabla _{\ti w} v)$:

4273: \begin{gather}

4274: \eqqno(3.3.2a)

4275: D_{u,J}(\nabla_{\ti w}v) = \nabla(\nabla_{\ti w}v) +

4276: J \scirc \nabla(\nabla_w v) \scirc J_S +

4277: \nabla J({\nabla_w v}, du \scirc J_S)=

4278: \\

4279: \notag

4280: \term{16}{\nabla^2_{\cdot, \ti w}v} \!+

4281: \term{14}{\nabla v \scirc \nabla \ti w \mathstrut} \!+

4282: \term{17}{J \scirc \nabla^2_{\cdot,\ti w}v \scirc J_S} \!+

4283: \term{15}{J \scirc \nabla v \scirc \nabla \ti w \scirc J_S \mathstrut} 

4284: \!+

4285: \term{18}{\nabla J({\nabla_{\ti w} v}, du \scirc J_S)} .

4286: \end{gather}

4287: \pagebreak[3]

4288: Similarly,

4289: \begin{align}

4290: \eqqno(3.3.2b)

4291: &\nabla_{\ti w}(D_{u,J}v + \dot J \scirc du \scirc J_S

4292: + J \scirc du \scirc \dot J_S)=

4293: \\

4294: &\nabla_{\ti w}( \nabla v + J\scirc \nabla v \scirc J_S +

4295: \nabla_v J \scirc du \scirc J_S +

4296: \dot J \scirc du \scirc J_S + J \scirc du \scirc \dot J_S)=

4297: \notag\\

4298: &\term{16'}{\nabla^2_{\ti w,\cdot}v} +

4299: \term{17'}{J \scirc \nabla^2_{\ti w,\cdot}v \scirc J_S} +

4300: \term{5}{\nabla_wJ \scirc \nabla v \scirc J_S} +

4301: \term{4'}{\nabla^2_{w,v}J \scirc du \scirc J_S} +

4302: \notag\\

4303: &\term{18}{\nabla J(\nabla_{\ti w}v;  du \scirc J_S)} +

4304: \term{3'}{\nabla_v J \scirc \nabla_{\ti w}du \scirc J_S} +

4305: \term{7}{\nabla_w\dot J \scirc du \scirc J_S} +

4306: \notag\\

4307: &\term{6'}{\dot J \scirc \nabla_{\ti w}du \scirc J_S} +

4308: \term{9}{\nabla_w J \scirc du \scirc \dot J_S} +

4309: \term{8'}{J \scirc \nabla_{\ti w}du \scirc \dot J_S} +

4310: \term{13}{J \scirc du \scirc \nabla_{\ti w}\dot J_S}.

4311: \notag

4312: \end{align}

4313: Comparing the terms [16] and $[16']$, it follows that $\nabla^2_{\cdot, \ti w}v -

4314: \nabla^2_{\ti w,\cdot}v = R^X(du, w;v)$. A similar relation holds for the

4315: terms [17] and $[17']$. The equality \eqqref(3.3.2) of the lemma is obtained

4316: by subtracting \eqqref(3.3.2a) from \eqqref(3.3.2b).

4317: %

4318: \qed

4319: 

4320: \newlemma{lem3.3.2}

4321: \sli In the situation of \lemma{lem3.3.1}, let $z^*

4322: \in S$ be a cusp-point. Consider $\ti w$ as a section of $TS$ with poles. Set

4323: $k \deff \ord_{z^*}du = \dimc (\scrn_u\sing)_{z^*}$ and choose a local complex

4324: coordinate $z$ on $S$ centered in $z^*$. Fix additionally $([v], \dot J_S, \dot

4325:  J) \in T_{[u, J]} \scrm$ and $\psi \in \sfh^0_D(S, N_u^* \otimes K_S)$. Then 

4326: locally in a neighborhood of $z^*$

4327: \begin{equation}

4328: \matrix%\format \r &\;\c \;& \l &\; \l & \; \l & \; \l\\

4329: %               w(z)  =   t0   t1   tk    t*

4330: z^k \cdot \ti w(z) &=&

4331: w_0 & + z\cdot w_1 &+ \cdots + z^k\cdot w_k  &+ z^k\cdot w^*(z),

4332: \cr

4333: v(z) &=&

4334: v_0 & + z\cdot v_1 &+ \cdots +z^k \cdot v_k &+ z^k \cdot v^*(z),

4335: \cr

4336: \psi(z) &=&

4337: \psi_0 & + z\cdot \psi_1 &+ \cdots +z^k \cdot \psi_k &

4338: + z^k \cdot \psi^*(z),

4339: \endmatrix

4340: \eqqno(3.3.2a1)

4341: \end{equation}

4342: where $w^*(z)$, $v^*(z)$, and $\psi^*(z)$ are $L^{1,p}$-smooth local sections

4343: of the corresponding bundles vanishing at $z=0$.

4344: 

4345: \slii The polynomials in \eqqref(3.3.2a1) can be considered as the order 

4346: $k$ jets of

4347: the following local holomorphic objects: a section of $TS$ for $w_0 + \cdots +

4348: z^k\cdot w_k$, a $(E_u)_{z^*}$-valued function for $v_0+ \cdots +z^k \cdot

4349: v_k$, and resp.\ $(N_u^*)_{z^*}$-valued $(0,1)$-form for $\psi_0 +\cdots +z^k

4350: \cdot \psi_k$. In particular, the coefficients can be considered as

4351: well-defined elements

4352: \begin{equation}

4353: \matrix%\format \r & \;\c\;& \r  &\; \c\; & \; \l \\

4354: %              w_i    =     \na     \in    T^*S

4355: w_i &=& \left({\d \over \d z}\right)^i(z^k \ti w(z))\ogran_{z=0}

4356: &\in& (T^*_{z^*} S)^{\otimes i-k} \otimes T_{z^*} S,

4357: \cr

4358: v_i &=& \nabla^i(v(z))\ogran_{z=0}  &\in&

4359: (T^*_{z^*} S)^{\otimes i} \otimes (E_u)_{z^*},

4360: \cr

4361: \psi_i &=& \nabla^i(\psi(z))\ogran_{z=0}  &\in&

4362: (T^*_{z^*} S)^{\otimes i} \otimes (N^*_u \otimes K_S)_{z^*}.

4363: \cr

4364: \endmatrix

4365: \eqqno(3.3.2b1)

4366: \end{equation}

4367: \end{lem}

4368: 

4369: 

4370: \proof \sli It follows from \refdefi{def1.5.1} that $du$, considered as a 

4371: {\sl holomorphic} section of the bundle $\hom_\cc(TS, E_u)$, locally has the 

4372: form $du(z)= z^k s(z)$ for some local holomorphic non-vanishing section $s$. 

4373: Consequently, $\scro([A]) = \scro( k\cdot [z^*])$ in a neighborhood of $z^*$. 

4374: Thus by \lemma{lem3.3.1} $\ti w$ can be locally

4375: represented in the form $\ti w(z) = z^{-k} \cdot \hat w(z)$ for some local

4376: $L^{1,p}_\loc$-smooth section of $TS$. Equation \eqqref(3.3.1) is equivalent

4377: to $\dbar \hat w = -\isl\cdot z^k \cdot I_S$ and implies the estimate $|\dbar

4378: \hat w(z)| \le |z^k|\cdot |I_S(z)|$. Now we use \lemma{lem1.2.3}.

4379: 

4380: The same argument applies to $v(z)$ and $\psi(z)$. Indeed, equation 

4381: \eqqref(2.2.4) on $v$ and relation \eqqref(1.4.1) imply the inequality 

4382: $|\dbar v(z)| \le c\cdot |z^k|$ with some constant $c$. A similar inequality for 

4383: $\psi(z)$ follows from \eqqref(1.5.3).

4384: 

4385: \smallskip

4386: \slii This part of the lemma can be reformulated in terms of the transformation

4387: of coefficients $w_i$, $v_i$, and $\psi_i$ under the change of a local

4388: holomorphic coordinate $z$ on $S$ and local coordinates on $X$. The claim

4389: concerning the change of $z$ is obvious.

4390: 

4391: Considering changes of coordinates on $X$ we make the following observation.

4392: If $\mbfx'=(x'_1, \ldots, x'_{2n})$ and $\mbfx''=(x''_1, \ldots,

4393: x''_{2n})$ are two systems of coordinates on $X$ centered in $u(z^*)$, then

4394: $\mbfx''=L(\mbfx') + Q(\mbfx') + \cdots$, where $L: \rr^{2n} \to

4395: \rr^{2n}$ (resp. $Q: \rr^{2n} \to \rr^{2n}$) is an appropriate linear (resp.\

4396: quadratic) map, and so on. In particular, $\mbfx'' - L(\mbfx')= O(

4397: |\mbfx'|^2)$. Consequently, for local frames $\d_{\mbfx'}= (\d_{x'_1},

4398: \ldots,\d_{x'_{2n}})$ and $\d_{\mbfx''}= (\d_{x''_1}, \ldots,\d_{x''_{2n}}

4399: )$ of $TX$ we obtain the relation $\d_{\mbfx''}(\mbfx') - L^{\sf t}(

4400: \d_{\mbfx''}) (\mbfx') = O(|\mbfx'|)$. Thus, for the pulled-back

4401: frames $u^*\d_{\mbfx'}$ and $u^*\d_{\mbfx''}$ of $E_u$ we have

4402: $u^*\d_{\mbfx'}(z) - L^{\sf t}(u^*\d_{\mbfx''})(z)= O(|z|^{k+1})$.

4403: This implies that the change of local coordinates on $X$ induces only a linear

4404: transformation of the $k$-jet of $v$, \ie the $k$-jet of $v$ behaves like a

4405: $k$-jet of a $T_{u(z^*)}$-valued function. The same argumentations

4406: can be applied to the $k$-jet of $\psi$.

4407: \qed

4408: 

4409: \newlemma{lem3.3.3}

4410: For $[u,J] \in \scrm\!\!$, $([w], I) \in \sfh^0_D(S,

4411: \scrn_n\sing)$, $([v], \dot J_S, \dot J) \in T_{[u,J]}\scrm\!\!$, and

4412: $\psi \in \sfh^0_D(S, N_u \otimes K_S) \cong \sfh^1_D(S, N_u)^*$

4413: it follows that %one has the relation

4414: \begin{equation}

4415: \Bigl\la \psi ,

4416: \Phi_{u,J}\bigl( ([v], \dot J_S, \dot J),\; ([w],I_S) \bigr) \Bigr\ra=

4417: \re \res_S(\psi\scirc \nabla_{\ti w}v),

4418: \eqqno(3.3.3)

4419: \end{equation}

4420: where $\res_S(\psi\scirc \nabla_{\ti w}v)$ denotes the residual type sum

4421: \begin{equation}

4422: \res_S(\psi\scirc \nabla_{\ti w}v) \deff

4423: \mathop{\textstyle\sum}\limits_{du(z^*_i)=0}\;

4424: \lim\limits_{\epsi \lrar 0}\;

4425: \int_{|z - z^*_i|=\epsi} \psi\scirc \nabla_{\ti w}v

4426: \eqqno(3.3.4)

4427: \end{equation}

4428: over all cusp-points $z^*_i\in S$ of $u$.

4429: 

4430: Moreover, if $\dot J=0$ and $([v], \dot J_S) \in \sfh^0(S, \scrn_u\sing)$,

4431: then $v=du(\ti v)$ with $\ti v \in L^{1, p}(S, TS \otimes \scro([A])\,)$

4432: and

4433: \begin{equation}

4434: \Bigl\la \psi ,

4435: \Phi_{u,J}\bigl( ([v], \dot J_S, 0),\; ([w],I_S) \bigr) \Bigr\ra=

4436: \re \res_S(\psi\scirc \nabla du(\ti w, \ti v)).

4437: \eqqno(3.3.5)

4438: \end{equation}

4439: \end{lem}

4440: 

4441: 

4442: \proof First, we note that by \lemma{lem3.3.2} the formulas

4443: (\ref{eq3.3.3}--\ref{eq3.3.5}) are well-defined.

4444: 

4445: Now, compute the subtraction of \eqqref(3.3.2) from \eqqref(3.2.8). The terms

4446: $[5]$, $[7]$, and $[9]$ cancel. To simplify further terms we use the Bianchi

4447: identity and antisymmetry of $R^X(\cdot {,} \cdot{;}\cdot)$ in the first two

4448: arguments. The difference of terms $[1] +[2] +[4] -[1'] -[2'] -[4']$ is zero:

4449: \begin{equation}

4450: \term1{R^X( v, du; w)} - \term{1'}{R^X( w, du; v)} +

4451: \term2{J \scirc R^X( v, du \scirc J_S; w)} -

4452: \end{equation}

4453: \begin{equation}

4454: \term{2'}{J \scirc R^X( w, du\scirc J_S; v)} +

4455: \term4{\nabla^2_{v,w}J \scirc du \scirc J_S} -

4456: \term{4'}{\nabla^2_{w,v}J \scirc du \scirc J_S} =

4457: \end{equation}

4458: \begin{equation}

4459: = R^X( v, du; w) + R^X( du, w; v) +

4460: J \scirc R^X( v, du \scirc J_S; w) +

4461: \end{equation}

4462: \begin{equation}

4463: J \scirc R^X( du\scirc J_S ,w ; v) +

4464: R^X(v,w; J \scirc du \scirc J_S) -

4465: J \scirc R^X(v,w; du \scirc J_S)  =

4466: \end{equation}

4467: \begin{equation}

4468: =R^X( v, du; w) + R^X( du, w; v) + R^X(w,v; du) +

4469: \end{equation}

4470: \begin{equation}

4471: J \scirc R^X( v, du \scirc J_S; w) +

4472: J \scirc R^X( du\scirc J_S, w; v) +

4473: J \scirc R^X( w, v; du \scirc J_S)  = 0.

4474: \end{equation}

4475: 

4476: \smallskip

4477: In the differences $[3] -[3']$, $[6] -[6']$, and $[8] -[8']$ respectively,

4478: we use the relation

4479: \begin{equation}

4480: \nabla(w)=\nabla(du(\ti w))= \nabla_{\ti w}du + du\scirc

4481: \nabla \ti w.

4482: \eqqno(3.3.6)

4483: \end{equation}

4484: This yields

4485: \begin{equation}

4486: \term3{\nabla_vJ \scirc \nabla w \scirc J_S} -

4487: \term{3'}{\nabla_vJ \scirc \nabla_{\ti w}du \scirc J_S} =

4488: \term{3''}{\nabla_vJ \scirc du \scirc \nabla \ti w \scirc J_S},

4489: \end{equation}

4490: and similar equalities for $[6] -[6']$ and $[8] -[8']$. Thus we obtain

4491: \begin{equation}

4492: \nabla_{([v],\dot J_S, \dot J)}(\wh D)([w],I_S) =

4493: \term{3''}{\nabla_vJ\scirc du\scirc \nabla\ti w \scirc J_S} +

4494: \term{6''}{\dot J\scirc du\scirc \nabla\ti w \scirc J_S} +

4495: \end{equation}

4496: \begin{equation}

4497: \term{8''}{J\scirc du\scirc \nabla\ti w \scirc \dot J_S}

4498: + \term{10}{\nabla_vJ \scirc du \scirc I_S}

4499: + \term{11}{J \scirc \nabla v \scirc I_S}

4500: + \term{12}{\dot J \scirc du \scirc I_S}

4501: \end{equation}

4502: \begin{equation}

4503: - \term{13}{J\scirc du \scirc \nabla_{\ti w} \dot J_S} +

4504: \term{14}{\nabla v \scirc \nabla \ti w} +

4505: \term{15}{J \scirc \nabla v \scirc \nabla \ti w  \scirc J_S}

4506: - \term{D}{D_{u,J}(\nabla_{\ti w}v) }.

4507: \end{equation}

4508: 

4509: \smallskip

4510: Further simplification uses the relation $I_S = J_S \scirc \nabla \ti w -

4511: \nabla \ti w \scirc J_S$. This gives

4512: \begin{equation}

4513: \term{3''}{\nabla_vJ \scirc du\scirc \nabla\ti w \scirc J_S} +

4514: \term{10}{\nabla_vJ \scirc du \scirc I_S} =

4515: \nabla_vJ\scirc du\scirc \nabla\ti w \scirc J_S +

4516: \end{equation}

4517: \begin{equation}

4518: \nabla_vJ \scirc du \scirc

4519: (J_S \scirc \nabla \ti w - \nabla \ti w \scirc J_S)

4520: = \term{3'''}{\nabla_vJ \scirc du\scirc J_S \scirc \nabla \ti w},

4521: \end{equation}

4522: and similarly,

4523: \begin{equation}

4524: \term{6''}{\dot J\scirc du\scirc \nabla\ti w \scirc J_S} +

4525: \term{12}{\dot J \scirc du \scirc I_S} =

4526: \term{6'''}{\dot J\scirc du\scirc J_S \scirc \nabla \ti w},

4527: \end{equation}

4528: \begin{equation}

4529: \term{11}{J \scirc \nabla v \scirc I_S} +

4530: \term{15}{J \scirc \nabla v \scirc \nabla \ti w  \scirc J_S}

4531: = \term{15'}{J \scirc \nabla v \scirc J_S \scirc \nabla \ti w}.

4532: \end{equation}

4533: 

4534: \smallskip

4535: Now we put together the terms $[3''']$, $[6''']$, $[14]$, and $[15']$.

4536: Because of the relation

4537: \begin{equation}

4538: \nabla v + J \scirc \nabla v \scirc J_S + \nabla_vJ \scirc du\scirc J_S +

4539: \dot J \scirc du\scirc J_S + J\scirc du\scirc \dot J_S =0

4540: \end{equation}

4541: this yields

4542: \begin{equation}

4543: \term{3'''}{\nabla_vJ \scirc du\scirc J_S \scirc \nabla \ti w}+

4544: \term{6'''}{\dot J\scirc du\scirc J_S \scirc \nabla \ti w} +

4545: \term{14}{\nabla v \scirc \nabla \ti w} +

4546: \term{15'}{J \scirc \nabla v \scirc J_S \scirc \nabla \ti w}=

4547: \end{equation}

4548: \begin{equation}

4549: \bigl( \nabla_vJ \scirc du\scirc J_S +

4550: \dot J\scirc du\scirc J_S +

4551: \nabla v +

4552: J \scirc \nabla v \scirc J_S \bigr)\scirc \nabla \ti w =

4553: - \term{8'''}{J\scirc du\scirc \dot J_S \scirc \nabla \ti w}.

4554: \end{equation}

4555: 

4556: \smallskip

4557: Finally, we conclude that outside the zero-set of $du$ one has

4558: \begin{equation}

4559: \nabla_{([v],\dot J_S, \dot J)}(\wh D)([w],I_S) =

4560: \end{equation}

4561: \begin{equation}

4562: J\scirc du\scirc \bigl( \term{8''}{\nabla\ti w \scirc \dot J_S}

4563: - \term{8'''}{\dot J_S \scirc \nabla \ti w}

4564: - \term{13}{\nabla_{\ti w} \dot J_S} \bigr)

4565: - \term{D}{D_{u,J}(\nabla_{\ti w}v) }.

4566: \end{equation}

4567: Note that $\psi \scirc J \scirc du = \psi \scirc du \scirc J_S =0$, since

4568: $\psi$ vanishes on $du(TS)$. Consequently,

4569: \begin{equation}

4570: \Bigl\la \psi , \Phi_{u,J}\bigl( ([v], \dot J_S, \dot J),

4571: \; ([w],I_S) \bigr)  \Bigr\ra=

4572: \re \int_S \psi \scirc

4573: \Bigl( - \nabla _{([v], \dot J_S, \dot J)} \wh D\Bigr)([w], I_S)=

4574: \end{equation}

4575: \begin{equation}

4576: \re \lim\limits_{\epsi \to 0} \int_{S \bs \cup \Delta(z_i, \epsi)}

4577: \psi \scirc \Bigl( -\nabla _{([v], \dot J_S, \dot J)} \wh D\Bigr)([w], I_S)

4578: =

4579: \re \lim\limits_{\epsi \to 0} \int_{S \bs \cup \Delta(z_i, \epsi)}

4580: \psi \scirc D_{u,J}(\nabla_{\ti w}v).

4581: \end{equation}

4582: Integrating by parts and using $D \psi=0$ we obtain the desired formula 

4583: \eqqref(3.3.3).

4584: 

4585: To obtain \eqqref(3.3.5) we use \eqqref(3.3.6) and relation $\psi \scirc du =0$.

4586: \qed

4587: 

4588: \smallskip

4589: Now we can describe the structure of $\Phi_{u,J}$ for cusp-curves. Here we 

4590: restrict ourselves to the case when $(X,J)$ is an almost complex surface. The 

4591: point is that, unlike to the higher dimensional situation, in this dimension 

4592: there are {\sl topological} reasons for the existence of cusp-curves, see

4593: \lemma{lem2.3.4}. 

4594: 

4595: Other than the restriction on dimension, our setting is as follows. $[u,J] \in 

4596: \scrm$ is a $J$-holomorphic curve with $\sfh^1(S, N_u) \cong \rr$, $z^*\in S$ a 

4597: cuspidal point, $k\deff \ord _{z^*} du$, $z$ a local complex coordinate centered 

4598: at $z^*$, $J^*$ a local (integrable) complex structure in a neighborhood of

4599: $u(z^*)$, and $(w^1, w^2)$ a local system of $J^*$-holomorphic coordinates on

4600: $X$ centered at $u(z^*)$. Finally, we fix some {\sl non-zero} $\psi \in

4601: \sfh^0(S, N_u^* \otimes K_S) \cong \sfh^1(S, N_u)^*$ and denote by $\ord_{z^*}

4602: \psi$ the order of vanishing of $\psi$ at $z^*$.

4603: 

4604: \newlemma{lem3.3.4}

4605: \sli After a polynomial transformation of the coordinates $(w^1, 

4606: \allowbreak

4607: w^2)$, chosen above, the map $u$ will have the form

4608: \begin{equation}

4609: u(z)=\bigr( z^{k+1} P_1(z), z^{k+l +2} P_2(z) \bigl) + z^{2k+1} g(z),

4610: \eqqno(3.3.7)

4611: \end{equation}

4612: such that $0\le l \le k$, $P_1$ is a polynomial of degree $\le k$ with $P_1(0)

4613: \not =0$, $P_2$ is a polynomial of degree $\le k-l-1$, trivial if $l=k$ or with

4614: $P_2(0) \not =0$ otherwise, and $g(z)$ is an $L^{1,p}$-smooth $\cc^2$-valued

4615: function.

4616: 

4617: \slii The integers $\ord_{z^*} \psi$ and $l$ do not depend on the particular

4618: choice of coordinates $(w^1, w^2)$ and $\psi\in \sfh^0(S, N_u^* \otimes K_S)$.

4619: For the restriction of $\Phi_{u,J}$ on the stalk $(\scrn_u\sing)_{z^*}

4620: \subset \sfh^0 (S, \scrn\sing_u)$, it follows that

4621: \begin{equation}

4622: \matrix %\format\r& \,\c\,&\r& \,\c\, \cr \\

4623: \ind_+ \bigl( \Phi_{u,J}\ogran_{(\scrn_u\sing)_{z^*} } \bigr)&=&

4624: \ind_- \bigl( \Phi_{u,J}\ogran_{(\scrn_u\sing)_{z^*} } \bigr)=\cr

4625: \sind \bigl( \Phi_{u,J}\ogran_{(\scrn_u\sing)_{z^*} } \bigr)&=&

4626: \max\,( 0, k-l -\ord_{z^*} \psi).

4627: \endmatrix

4628: \eqqno(3.3.8)

4629: \end{equation}

4630: 

4631: \sliii If $z^*_1$ and $z^*_2$ are distinct cusp-points of $u:S \to X$,

4632: then the stalks $(\scrn_u\sing)_{z^*_1}$ and $(\scrn_u\sing)_{z^*_2}$ are

4633: $\Phi$-orthogonal, \ie

4634: \begin{equation}

4635: \Phi_{u,J}\bigl( (\scrn_u\sing)_{z^*_1}, (\scrn_u\sing)_{z^*_2} \bigr)=0.

4636: \eqqno(3.3.9)

4637: \end{equation}

4638: \end{lem}

4639: 

4640: \proof {\sl Part} \sli follows immediately from \lemma{lem1.2.4}. It

4641: simply says that if $\ord_{z^*}du=k$, then the jet $j^{2k+1}u$ is well-defined

4642: and {\sl holomorphic}, \ie can be represented by a complex polynomial.

4643: Note that the theorem of \cite{Mi-Wh} (see \lemma{lem1.2.1}) says that {\sl

4644: topologically} one can also define higher terms which determine the whole

4645: behavior of $u$ at $z^*$.

4646: 

4647: {\sl Part} \sliii can be easily obtained from \eqqref(3.3.4) and 

4648: \eqqref(3.3.5). It remains to consider

4649: 

4650: {\sl Part \sliip}. First, we observe that the integer $l$ is the secondary cusp 

4651: index of $u$ at $z^*$ (see \refdefi{def4.2s.1}). It follows then from the

4652: results of \refsubsection{4.2s} that this integer is well defined and 

4653: independent of the choice of $(w^1, w^2)$. The independence of $\ord_{z^*} \psi$ of

4654: the choice of $(w^1, w^2)$ and $\psi$ is obvious. 

4655: 

4656: Let $J^*$ and $(w^1, w^2)$

4657: be a complex structure and $J^*$-holomorphic coordinates in a neighborhood

4658: of $u(z^*)$, such that $J^*(u(z^*)) = J(u(z^*))$ and $u$ has the local form

4659: \eqqref(3.3.7). Differentiating \eqqref(3.3.7) we see that in the coordinates 

4660: $(w^1, w^2)$

4661: \begin{equation}

4662: du(z)=\bigr( z^k P'_1(z), z^{k+l +1} P'_2(z) \bigl) + z^{2k} g'(z),

4663: \eqqno(3.3.10)

4664: \end{equation}

4665: with polynomials

4666: \begin{equation}P'_1(z)= (k+1)P_1(z) + z {\textstyle{d \over dz}} P_1(z)

4667: \quad\text{and}\quad

4668: P'_2(z)= (k+l+2)P_2(z) + z {\textstyle{d \over dz}} P_2(z),

4669: \eqqno(3.3.11)

4670: \end{equation}

4671: of degree $\le k$ and $\le k-l-1$ respectively

4672: and with $g'(z) = (2k+1) g(z) dz + z dg(z)$ being $L^p$-bounded.

4673: 

4674: From the definition of the Nijenhuis torsion tensor $N_J$ of $J$ we obtain a

4675: pointwise estimate $|\dbar_J w_\alpha| \le |N_J|$. Further,

4676: \begin{equation}

4677: \dbar(w_\alpha \scirc u) = (dw_\alpha \scirc du)^{(0,1)} =

4678: \dbar_J w_\alpha  \scirc du,

4679: \end{equation}

4680: since $u$ is $J$-holomorphic. Consequently, we obtain a pointwise  estimate

4681: \begin{equation}

4682: |\dbar(w_\alpha \scirc u)(z)| \le c\cdot |z^k|

4683: \eqqno(3.3.12)

4684: \end{equation}

4685: with some constant $c$. Let $\{ e^*_\alpha\}_{\alpha=1,2}$ be the local

4686: $J^*$-complex frame of $T^*X$ dual to the frame $\{ dw_\alpha\}_{\alpha=1,2}$.

4687: Then there exists a local $J$-complex frame $\{ e_\alpha\}_{\alpha=1,2}$ of

4688: $T^*X$ with pointwise estimates

4689: \begin{equation}

4690: |e^*_\alpha(w) -e_\alpha(w)| \le c\cdot |w|

4691: \quad\text{and}\quad

4692: |\nabla e^*_\alpha(w) - \nabla e_\alpha(w)| \le c,

4693: \eqqno(3.3.13)

4694: \end{equation}

4695: where $|w|^2 = |w_1|^2 + |w_2|^2$ and $c$ is some constant.

4696: Using (\ref{eq3.3.10}--\ref{eq3.3.13}) and the estimates 

4697: $|u(z)| \le c\cdot |z^{k+1}|$ and $|du(z)|

4698: \le c\cdot |z^k|$ we conclude that

4699: 

4700: {\sl a)\.} $\mbfe_\alpha \deff u^* e_\alpha$ is a local complex frame of

4701: $E_u = u^*TX$ with a pointwise estimate

4702: \begin{equation}

4703: |\dbar_{u,J} \mbfe_\alpha(z)| \le c\cdot |z^k|;

4704: \end{equation}

4705: 

4706: {\sl b)\.} $du$, considered as a section of $E_u\otimes T^*S$ with the frame

4707: $\mbfe_\alpha\otimes dz$, has local form \eqqref(3.3.10), possibly with another

4708: $g'(z) \in L^p$. Moreover, since $du$ is a holomorphic section and

4709: $\mbfe_\alpha$ are sufficiently regular, this new $g'(z)$ is $C^1$-smooth.

4710: Further, since $z dg(z)= g'(z) - (2k+1)zg(z) dz$ is continuous and $dg(z) \in

4711: L^p$ with $p>2$, we conclude that $z dg(z)$ vanishes at $z=0$. This gives

4712: an additional relation $g'(0)=0$.

4713: 

4714: Differentiating \eqqref(3.3.10) we obtain that in the frame 

4715: $\mbfe_\alpha\otimes dz^2$

4716: \begin{equation}

4717: \nabla du(z)=\bigr( z^{k-1} P''_1(z), z^{k+l } P''_2(z) \bigl) +

4718: z^{2k-1} g''(z),

4719: \eqqno(3.3.14)

4720: \end{equation}

4721: with polynomials

4722: \begin{equation}

4723: P''_1(z)= k\,P'_1(z) + z {\textstyle{d \over dz}} P'_1(z)

4724: \quad\text{and}\quad

4725: P''_2(z)= (k+l+1)P_2(z) + z {\textstyle{d \over dz}} P'_2(z),

4726: \eqqno(3.3.15)

4727: \end{equation}

4728: of degree $\le k$ and $\le k-l-1$ respectively and with

4729: \begin{equation}

4730: g''(z) = (2k+1) g'(z) \otimes dz + z \nabla g'(z)

4731: \end{equation}

4732: continuous and vanishing at $z=0$. By our construction, $P''_1(0)= (k+1)k

4733: P_1(0) \not =0$ and $P''_2(0)= (k+l+2)(k+l+1) P_1(0)$ vanishes \iff $l=k$.

4734: 

4735: Since the projection $\pr: E_u \to N_u$ is obtained as the quotient \wrt

4736: the image of $z^{-k} du \sim (P'_1(z), z^{l+1} P'_2(z))$, we have the following

4737: form for the composition:

4738: \begin{equation}

4739: \pr \scirc \nabla du(z)= P'''(z) + g'''(z),

4740: \end{equation}

4741: where $P'''(z)$ is a polynomial $P'''(z)$ of degree $\le k-l-1$ given by 

4742: the relation

4743: \begin{equation}

4744: z^{k+l}P'''(z) = z^{k+l} P''_2(z) -

4745: {z^{k+l+1} P'_2(z) \cdot z^{k-1} P''_1(z) \over

4746: z^k P'_1(z)} + o(z^{2k-1}).

4747: \end{equation}

4748: In particular, $P'''(0)= (k+l+1)(l+1)P_2(0)$ vanishes \iff $l=k$.

4749: 

4750: Denoting $\nu \deff \ord_{z^*} \psi$ we obtain that

4751: \begin{equation}

4752: \psi \scirc \nabla du(z) = a z^{k+l+\nu} + o(z^{k+l+\nu})

4753: \end{equation}

4754: with $a$ vanishing \iff $l=k$. The proof of part \slii of the lemma can be now

4755: finished using the following algebraic result. \qed

4756: 

4757: \newlemma{lem3.3.5} For a given polynomial $P(z)= a_0 + a_1 \,z

4758: + \cdots + a_{ k-l-1} z^{ k-l-1}$ with $a_0 \not=0$ and $0\le l <k$

4759: the quadratic form

4760: \begin{equation}

4761: (w_0,\ldots,w_k) \in \cc^{k+1}

4762: \mapsto \re\res_{z=0} \left( { z^{k+l}\, P(z)\,

4763: \bigl(\sum_{i=0}^k w_i\, z^i\bigr)^2  \over z^{2k} } dz \right) \in \rr

4764: \end{equation}

4765: is equivalent to the quadratic form

4766: \begin{equation}

4767: (w_0,\ldots,w_k) \in \cc^{k+1}

4768: \mapsto \re\res_{z=0} \left( { z^{k+l}\, a_0\,

4769: \bigl(\sum_{i=0}^k w_i\, z^i\bigr)^2  \over z^{2k} } dz \right)\in \rr

4770: \end{equation}

4771: and satisfies the index relations

4772: \begin{equation}

4773: \ind_+ Q = \ind_- Q =  \sind Q = k-l.

4774: \eqqno(3.3.16)

4775: \end{equation}

4776: \end{lem}

4777: 

4778: 

4779: %\newdefi{def3.3.2} In the situation of \lemma{lem3.3.4} the integer $l$ 

4780: %is called the {\sl secondary cusp index of $u$ at  point $z^*$}.

4781: %\end{defi} ***!!!

4782: 

4783: 

4784: \newsubsection[4.4]{Critical points and cusp-curves in the moduli space}

4785: Recall that in \lemma{lem3.3.4} we found two obstructions for existence of

4786: saddle points. They are encoded in the secondary cusp-indices $l_i$ of

4787: cusp-points $z^*_i$ of $u$ (see \refdefi{def4.2s.1}) and the vanishing

4788: order at $z^*_i$ of a generic $\psi \in \sfh^0_D(S, N_u \otimes K_S) \cong

4789: \bigl(\sfh^1_D(S, N_u) \bigr) ^*$. The behavior $l_i$ under deformation was

4790: studied in \refsubsection{4.2s}. 

4791: In this paragraph we describe the behavior of $\sfh^0_D(S, N_u \otimes K_S)$.

4792: Our main interest is, of course, $[u,J] \in \scrm$ with $\dimr\sfh^1_D (S, N_u)

4793: =1$, because these are candidates for saddle points. We start with

4794: 

4795: \newlemma{lem4.4.1} Let $\mbfk=(k_1, \ldots, k_m)$ and $h^1\in \nn$ 

4796: be given. Then the set

4797: \begin{equation}

4798: \whcalm_{=\mbfk, h^1} \deff \bigl\{ (u, J_S, J; \mbfz) \in \whcalm

4799: _{=\mbfk} \;:\; \dimr \sfh^1_D(S, N_u) = h^1\,\bigr\}

4800: \subset \whcalm_{=\mbfk}

4801: \end{equation}

4802: is a $C^{\ell-1}$-smooth submanifold of codimension $h^0{\cdot} h^1$ where

4803: $h^0 \deff \dimr \sfh^0_D(S, N_u)$.

4804: 

4805: The set $\whcalm_{=\mbfk, h^1}$ is $\bfg$-invariant and the projection $\pr:

4806: \whcalm_{=\mbfk, h^1} \lrar \scrm_{=\mbfk, h^1} \deff \whcalm_{=\mbfk, h^1}

4807: /\bfg$ is a $C^{\ell-1}$-smooth principle $\bfg$-bundle.

4808: \end{lem}

4809: 

4810: 

4811: \state Remark. The definition \eqqref(1.5.1) of $\scrn_u$ and the index formula

4812: \eqqref(2.2.3) imply that $h^0= h^1 +2(\mu +(g-1)(3-n) -|\mbfk|)$, where $n=\half

4813: \dimr X$ and $\mu \deff \la c_1(X), [u(S)] \ra$. So $h^0=\dimr \sfh^0_D(S,

4814: N_u)$ is constant along $\whcalm_{=\mbfk, h^1}$ and

4815: \begin{equation}

4816: \whcalm_{=\mbfk} = \mathop{\textstyle\bigsqcup\,}\nolimits

4817: _{h^1 =0} ^\infty \;\whcalm_{=\mbfk, h^1}

4818: \end{equation}

4819: is a stratification of $\whcalm_{=\mbfk}$ indexed by $h^1 =\dimr \sfh^1_D(S,

4820: N_u)$. Taking the $\bfg$-quotients, we obtain a similar stratification of $\scrm

4821: _{=\mbfk}$. Another stratification, more interesting for our purpose, is

4822: \begin{equation}

4823: \scrm_{h^1} = \mathop{\textstyle\bigsqcup\,}\nolimits

4824: _{\mbfk}\scrm_{=\mbfk, h^1}

4825: \end{equation}

4826: with $h^1=1$.

4827: Note that if for given $\mbfk$ and $h^1$ the expected value of $h^0$ is

4828: negative, then $\whcalm_{=\mbfk, h^1}$ is empty.

4829: 

4830: \proof Consider the Banach bundles $L^{1,p}(S, N)$, $L^p_{(0,1)}(S, N)$

4831: over $\whcalm_{=\mbfk}$, and the bundle homomorphism $D^N: L^{1,p}(S, N) \to

4832: L^p_{(0,1)}(S, N)$ constructed in \lemma{lem4.2.7n}. Then $\sfh^i_D(S, N_u)$

4833: is the (co)kernel of $D^N$. From \lemma{lem1.3.1} we obtain the map

4834: \begin{equation}

4835: \nabla_{(v, \dot J_S, \dot J)}D^N: \sfh^0_D(S, N_u) \to \sfh^1_D(S, N_u)

4836: \eqqno(4.4.1),

4837: \end{equation}

4838: which is bilinear in $(v, \dot J_S, \dot J) \in T_{(u, J_S, J)}\whcalm_{=\mbfk}$ 

4839: and $w \in \sfh^0_D(S, N_u)$. It is not difficult to see that the map 

4840: \eqqref(4.4.1) can be computed using \eqqref(3.2.8) and that it coincides with 

4841: the restriction of $\Phi$ from \eqqref(3.1.1) to the corresponding spaces.

4842: 

4843: The key point of the proof is to show the surjectivity of the induced map

4844: \begin{equation}

4845: \Phi: T_{(u, J_S, J)}\whcalm_{=\mbfk} \;\lrar\; \homr \bigl(

4846: \sfh^0_D(S, N_u),\;  \sfh^1_D(S, N_u) \bigr)

4847: \end{equation}

4848: Then the claim of the lemma will follow from  the implicit function theorem.

4849: 

4850: Fix bases $(w_1, \ldots, w_{h^0})$ of $\sfh^0_D(S, N_u)$ and $(\psi_1, \ldots,

4851: \psi_{h^1})$ of $\sfh^0_D(S, N_u \otimes K_S) \cong \bigl(\sfh^1(S,N_u)

4852: \bigr)^*$. The last isomorphism is the Serre duality from \lemma{lem1.5.1}.

4853: We must find tangent vectors $(v_{ij}, \dot J_{S,ij}, \dot J_{ij}) \in T

4854: _{(u, J_S, J)}\whcalm_{=\mbfk}$, $i=1,\ldots, h^0$, $j=1,\ldots, h^1$

4855: obeying the relation

4856: \begin{equation}

4857: \bigl\la \psi_{j'},\,

4858: \Phi\bigl( (v_{ij}, \dot J_{S,ij}, \dot J_{ij}), w_{i'} \bigr) \bigr\ra

4859: = \delta_{ii'} \delta_{jj'}

4860: \eqqno(4.4.2)

4861: \end{equation}

4862: with $\la\cdot\,, \cdot \ra$ denoting the pairing from \eqqref(1.5.5).

4863: 

4864: The main idea is to find solutions of \eqqref(4.4.2) in the special form such that

4865: $v_{ij}$ and $\dot J_{S,ij}$ are identically zero, and $\dot J_{ij}$ vanish

4866: along $u(S)$ and in a neighborhood of all cusp-points on $u(S)$. This

4867: assumption implies that all the terms in \eqqref(3.2.8) except $[7]$ vanish.

4868: Thus \eqqref(4.4.2) reduces to

4869: \begin{equation}

4870: \re \int_S \psi_{j'} \scirc \nabla_{w_{i'}} \dot J_{ij}\scirc du

4871: \scirc J_S = \delta_{ii'} \delta_{jj'}.

4872: \end{equation}

4873: From this point we can use the arguments either of \lemma{lem2.1.2} or {\sl

4874: Lemma 3.2.4}. Note that we can arrange $\dot J_{ij}$ to have support in

4875: any given open subset $U \subset X$ with $U\cap u(S) \not = \emptyset$. \qed

4876: 

4877: \smallskip

4878: The big freedom in the choice of $\dot J_{ij}$ implies the following

4879: 

4880: \newcorol{cor4.4.2} Let $\dimr X =4$, \ie $X$ is an almost complex

4881: surface. Then the intersection of $\whcalm_{=\mbfk, \mbfl}$ and

4882: $\whcalm_{=\mbfk, h^1}$ is transversal, so that the set

4883: \begin{equation}

4884: \whcalm_{=\mbfk, \mbfl, h^1}\deff \whcalm_{=\mbfk, \mbfl} \cap

4885: \whcalm_{=\mbfk, h^1}

4886: \end{equation}

4887: is a $C^{\ell-1}$-smooth submanifold of $\whcalm_{=\mbfk}$ of codimension

4888: $2|\mbfl| + h^0{\cdot}h^1$. A similar result also holds for

4889: $\scrm_{=\mbfk, \mbfl, h^1} \deff \whcalm_{=\mbfk, \mbfl, h^1} /\bfg =

4890: \scrm_{=\mbfk, \mbfl} \cap \scrm_{=\mbfk, h^1}$.

4891: \end{corol}

4892: 

4893: 

4894: \medskip

4895: Now we will study the behavior of zeros of a non-trivial $\psi \in \sfh^0_D(S,

4896: N_u \otimes K_S)$ for $[u, J] \in \scrm_{=\mbfk, h^1=1}$. Note that, modifying

4897: the construction from \lemma{lem4.4.3}, we obtain a bundle $N^* \otimes K_S$

4898: over $\whcalm_{=\mbfk} \times S$, $C^{\ell-1}$-smooth Banach

4899: bundles $L^{1,p}(S, N^* \otimes K_S)$ and $L^p_{(0,1)}(S, N^* \otimes K_S)$

4900: over $\whcalm_{=\mbfk}$, and a $C^{\ell-1}$-smooth bundle homomorphism

4901: \begin{equation}

4902: (D^N)^*: L^{1,p}(S, N^* \otimes K_S) \to L^p_{(0,1)}(S, N^* \otimes K_S).

4903: \end{equation}

4904: Since the kernel of $(D^N)^*$ is of constant dimension on each $\whcalm_{=\mbfk,

4905: h^1}$, we obtain a $C^{\ell-1}$-smooth bundle $\sfh^0_D(S, N^* \otimes K_S)$ of

4906: $\rank_\rr =h^1$ on $\whcalm_{=\mbfk, h^1}$. This means that there exists a

4907: (local) frame $\psi_1, \ldots, \psi_{h^1}$ of $\sfh^0_D(S, N^* \otimes K_S)$

4908: which depends $C^{\ell-1}$-smoothly on $(u, J_S, J) \in \whcalm_{=\mbfk, h^1}$.

4909: 

4910: In the particular case $h^1=1$ we obtain a (local) $C^{\ell-1}$-smooth family of

4911: non-trivial $\psi \in \sfh^0_D(S,N^* \otimes K_S)$ such that for every $(u,

4912: J_S, J)$ the corresponding $\psi$ is defined uniquely up to a constant factor.

4913: \lemma{lem1.2.3} ensures that the zero-divisor of such $\psi$ is well-defined

4914: and has degree

4915: $c_1(N^* \otimes K_S)$. By \lemma{lem2.3.4}, the possible range for $c_1(N^*

4916: \otimes K_S)$ is the interval between 0 and $g-1$. We are interested in the

4917: distribution of the zeros of $\psi$, especially at cusp-points of $u$. For

4918: a given $\mbfk=(k_1, \ldots, k_m)$ we consider  $m$-tuples $\bfnu=(\nu_1,

4919: \ldots, \nu_m)$ with $0 \le \nu_i \le k_i$, $i=1, \ldots, m$. Denote

4920: $|\bfnu| \deff \sum_{i=1}^m \nu_i$.

4921: 

4922: \newlemma{lem4.4.3}

4923: \sli The set $\whcalm_{=\mbfk, \bfnu} \deff \bigl\{ (u, J_S, J) \in

4924: \whcalm_{=\mbfk, h^1=1} \;:\; \ord_{z^*_i} \psi \ge \nu_i \,\bigr\} \subset

4925: \whcalm_{=\mbfk, h^1=1}$ is a $C^{\ell-1}$-smooth submanifold of codimension

4926: $2(n-1)|\bfnu|$, $n=\half\,\dimr X$.

4927: 

4928: \slii  Let $n=2$, \ie $X$ is an almost complex surface. Then the intersection of

4929:  $\whcalm_{=\mbfk, \mbfl}$ and $\whcalm_{=\mbfk, \bfnu}$ is transversal, so

4930: that the set

4931: \begin{equation}

4932: \whcalm_{=\mbfk, \mbfl, \bfnu} \deff \whcalm_{=\mbfk, \mbfl, h^1=1} \cap

4933: \whcalm_{=\mbfk, \bfnu} \;\subset\; \whcalm_{=\mbfk, h^1=1}

4934: \end{equation}

4935: is a $C^{\ell-1}$-smooth submanifold of codimension $2|\mbfl| + |\bfnu|$.

4936: 

4937: 

4938: \sliii Similar results hold for $\scrm_{=\mbfk, \mbfl, \bfnu} \deff

4939: \whcalm_{=\mbfk, \mbfl, \bfnu}/\bfg= \scrm_{=\mbfk, \mbfl,h^1=1} \cap

4940: \scrm_{=\mbfk, \bfnu}$.

4941: \end{lem}

4942: 

4943: 

4944: \proof \slip.

4945: Fix $(u_0, J_{S,0}, J_0) \in \whcalm_{=\mbfk, \bfnu}$. Let $z_i$

4946: be a local $J_S$-holomorphic coordinate on $\whcalm_{=\mbfk, h^1=1}$ in the

4947: sense of \refdefi{def4.2.2n}, centered at the cusp-point $z^*_i$ of $(u,

4948: J_S, J)$, $i=1, \ldots, m$. Further, let $\psi$ be a local $C^{\ell-1}$-smooth

4949: family of non-trivial elements of $\sfh^0_D(S,N^* \otimes K_S)$. Then by

4950: \lemma{lem3.3.2}, for each $i=1, \ldots, m$ and each $(u, J_S, J) \in

4951: \whcalm_{=\mbfk, h^1=1}$ we can construct the jets $j^{k_i\!}\psi =

4952: \sum_{j=0} ^{k_i} \psi_{i,j} \cdot z_i^{\;j}$ of $\psi$ at $z^*_i$.

4953: 

4954: Repeating the arguments used in the proof of \lemma{lem4.2.3n} %%?? Check!

4955: we can show

4956: that the coefficients $\psi_{i,j} \in (T^*_{z^*_i})^{\otimes j} \otimes (N^*

4957: \otimes K_S)_{z^*_i}$ depend $C^{\ell-1}$-smoothly on $(u, J_S, J) \in \whcalm

4958: _{=\mbfk, h^1=1}$. This means that $\whcalm_{=\mbfk, \bfnu}$ is the zero set

4959: of the (locally defined) function $\yps^\psi_\bfnu$ on $\whcalm_{=\mbfk,

4960: h^1=1}$ given by the first $\nu_i$ coefficients of each $j^{k_i\!}\psi$,

4961: $i=1, \ldots,  m$, \ie

4962: \begin{equation}

4963: \yps^\psi_\bfnu(u, J_S, J) =

4964: (\psi_{1,0}, \ldots, \psi_{1, \nu_1 -1}, \ldots, \psi_{m,0}, \ldots,

4965: \psi_{m, \nu_m -1}).

4966: \end{equation}

4967: 

4968: Consequently, it is sufficient to show the surjectivity of the differential 

4969: $d\yps^\psi_\bfnu$ at the fixed $(u_0, J_{S,0}, J_0)$. But first we must compute 

4970: $d\yps^\psi_\bfnu$ for a given $(v, \dot J_S, \dot J) \in T_{(u_0, J_{S,0}, J_0)} 

4971: \whcalm_{=\mbfk, h^1=1}$. Let $\gamma(t)= (u_t, J_{S,t}, J_t)$ be a curve in 

4972: $\whcalm_{=\mbfk, h^1=1}$ which starts at $(u_0, J_{S,0}, J_0)$ and has the 

4973: tangent vector $(v, \dot J_S, \dot J)$ at $t=0$. Then we obtain a family $\psi_t$ 

4974: of non-trivial $\psi_t \in \sfh^0_D(S, N_{u_t}^* \otimes K_S)$. In particular, 

4975: for each $t$ we obtain the relation $D^*_t \psi_t=0$, where  $D^*_t$ denotes 

4976: the operator $D^{N^* \otimes K_S}$ corresponding to $(u_t, J_{S,t}, J_t)$.

4977: 

4978: Fix some symmetric connections on $X$ and $S$. As in \refsubsection{3.2}, we 

4979: obtain induced connections for all (usual and Banach) bundles involved in our

4980: computations. We use the same notation $\nabla$ for all these connections,

4981: in particular, for the connection in the bundle  $L^{1,p}(S, N_u)$ with

4982: the fiber $L^{1,p}(S, N_{u_t})$ over $(u_t, J_{S,t}, J_t)$. Hence for any

4983: $w_0 \in L^{1,p}(S, N_{u_0})$ we can construct a family  $w_t \in L^{1,p}(S,

4984:  N_{u_t})$ which is covariantly constant. This yields a covariantly constant

4985: trivialization of the Banach bundle $L^{1,p}(S, N_u)$ along $\gamma$.

4986: 

4987: For every such family $w_t$ we have the relation

4988: \begin{equation}

4989: \la w_t, D^*_t \psi_t \ra=0.

4990: \eqqno(4.4.3)

4991: \end{equation}

4992: Vice versa, a family $\psi_t \in L^{1,p}(S, N^*_{u_t} \otimes K_S)$ lies

4993: in $\sfh^1_D(S, N^*_{u_t} \otimes K_S)$ if \eqqref(4.4.3) holds. Rewrite 

4994: \eqqref(4.4.3) in the form

4995: \begin{equation}

4996: \la \psi_t, D_t w_t \ra=0

4997: \eqqno(4.4.4)

4998: \end{equation}

4999: with $D_t$ denoting the operator $D^N_{u_t, J_t}: L^{1,p}(S, N_{u_t}) \to

5000: L^p_{(0,1)}(S, N_{u_t})$. After covariant differentiation in $t$ we obtain

5001: $\la \dot \psi_t, D_t w_t \ra + \la \psi_t, (\nabla_{(v_t, \dot J_{S,t},

5002: \dot J_t)} D_t) w_t \ra =0$. The latter is equivalent to

5003: \begin{equation}

5004: \la  D^*_t \dot\psi_t,  w_t \ra +

5005: \la \psi_t, (\nabla_{(v_t, \dot J_{S,t}, \dot J_t)} D_t) w_t \ra =0.

5006: \eqqno(4.4.5)

5007: \end{equation}

5008: 

5009: Now we can give the description of $d\yps^\psi_\bfnu$ at $(u_0, J_{S,0}, J_0) \in 

5010: \whcalm_{=\mbfk, h^1=1}$. For a given tangent vector $(v, \dot J_S, \dot J)$ we 

5011: find $\dot \psi \in L^{1,p}(S, N_{u_0} \otimes K_S)$ such that \eqqref(4.4.5)

5012: holds for every $w \in L^{1,p}(S, N_{u_0} \otimes K_S)$. The existence of such

5013: $\dot \psi$ is equivalent to the condition that $(v, \dot J_S, \dot J)$ is

5014: tangent to $\whcalm_{=\mbfk, h^1=1}$. Such $\dot \psi$ is unique up to

5015: addition of $\psi \in \sfh^0_D(S, N_{u_0} \otimes K_S)$. The

5016: jets $j^{k_i\!} \dot \psi= \sum_{j=0} ^{k_i} \dot \psi_{i,j} \cdot z_i^{\;j}$

5017: of such $\dot \psi$ at $z^*_i$ are well-defined and

5018: \begin{equation}

5019: d\yps^\psi_\bfnu(v, \dot J_S, \dot J) =

5020: (\dot \psi_{1,0}, \ldots, \dot \psi_{1, \nu_1 -1}, \ldots,

5021: \dot \psi_{m,0}, \ldots,

5022: \dot \psi_{m, \nu_m -1}).

5023: \end{equation}

5024: $d\yps^\psi_\bfnu$ is independent of the choice of $\dot \psi$ provided

5025: $(u_0, J_{S,0}, J_0) \in \whcalm_{=\mbfk, \bfnu}$.

5026: 

5027: \smallskip

5028: To show the surjectivity of $d\yps^\psi_\bfnu$ we must invert the construction

5029: above. Let $j^{\nu_i-1\!} \dot \psi$ be given jets. Extend them to jets

5030: $j^{k_i\!} \dot \psi$. Note that by definition the operator $D^*_0 = D^{N^*

5031: \otimes K_S} _{u_0, J_0}$ has the form $\dbar^{N^* \otimes K_S}_{u_0, J_0} +

5032: R^{N^* \otimes K_S}_{u_0, J_0}$ where

5033: \begin{equation}

5034: R^{N^* \otimes K_S}_{u_0, J_0}: N^* \otimes K_S \to

5035: N^* \otimes K_S \otimes \Lambda^{(0,1)}

5036: \end{equation}

5037: is a continuous bundle homomorphism. Consider the equations

5038: \begin{equation}

5039: z_i^{-k_i}\bigl(\dbar^{N^* \otimes K_S}_{u_0, J_0} +

5040: R^{N^* \otimes K_S}_{u_0, J_0} \bigr)

5041: \bigl(j^{k_i\!} \dot \psi + z_i^{k_i}  \phi_i(z_i)\bigr)=0

5042: \eqqno(4.4.6)

5043: \end{equation}

5044: for unknown $\phi_i(z_i)$ defined in a neighborhood of $z^*_i$. Using

5045: \lemma{lem1.4.1} we obtain pointwise estimates $|R^{N^* \otimes K_S}_{u_0,

5046: J_0}(z_i)| \le C\cdot |z_i|^{k_i}$. Thus equation \eqqref(4.4.6) is equivalent to

5047: \begin{equation}

5048: \left(\dbar^{N^* \otimes K_S}_{u_0, J_0} +

5049: \left(\msmall{\bar z_i \over z_i}\right)^{k_i}

5050: R^{N^* \otimes K_S}_{u_0, J_0} \right) \phi_i(z_i)

5051: + z_i^{- k_i} R^{N^* \otimes K_S}_{u_0, J_0}j^{k_i\!}=0.

5052: \eqqno(4.4.7)

5053: \end{equation}

5054: The existence of solutions of \eqqref(4.4.7) can be deduced from the

5055: surjectivity of the operator $\dbar + R: L^{1,p}(\Delta,\cc^n) \to L^p(\Delta,

5056: \cc^n)$ with $R \in L^p$, $p>2$.

5057: We refer to \cite{Iv-Sh-1} for the construction of a right inverse for such

5058: $\dbar + R$. This implies the local existence of solutions $\phi_i(z_i)$

5059: of \eqqref(4.4.6).

5060: 

5061: The regularity property of $R^{N^* \otimes K_S}_{u_0, J_0}$ implies that the $z_i

5062: ^{k_i} \phi_i(z_i)$ are $C^{\ell-1}$-smooth. Thus we can construct a $\dot \psi 

5063: \in C^{\ell-1}(S, N^*_{u_0} \otimes K_S)$ which locally near $z^*_i$ has the form 

5064: $\dot \psi(z_i) = j^{k_i\!} \dot \psi + z_i^{k_i} \phi_i(z_i)$ and satisfies 

5065: \eqqref(4.4.6). Now, the surjectivity of $\yps^\psi_\bfnu$

5066: will follow from the existence of $(v, \dot J_S, \dot J)\in T_{(u_0, J_{S,0},

5067: J_0)}  \whcalm_{=\mbfk, h^1=1}$ such that for the constructed $\dot \psi$ and

5068: a fixed non-zero $\psi_0 \in \sfh^0_D(S, N^*_{u_0} \otimes K_S)$

5069: the relation \eqqref(4.4.5) holds for any $w \in L^{1,p}(S, N_{u_0})$.

5070: 

5071: Now observe that we can use \eqqref(3.2.8) to compute $\nabla_{ (v, \dot J_v, 

5072: \dot J)} D^N_{u_0, J_0}$. This implies that we can use the trick from the proof 

5073: of \lemma{lem4.4.1}.  Namely, we look for the desired $(v, \dot J_S, \dot J)$

5074: in the special form, such that $v$ and $\dot J_S$ vanish identically, and

5075: $\dot J$ vanishes  along $u_0(S)$ and in some neighborhoods of cusp-points of

5076: $u(S)$. Now all terms in \eqqref(3.2.8) except $[7]$ vanish, and 

5077: \eqqref(4.4.5) is equivalent to

5078: \begin{equation}

5079: D^{N^* \otimes K_S}_{u_0, J_0} \dot \psi +

5080: \psi_0 \scirc \nabla \dot J \scirc du_0 \scirc J_S =0.

5081: \end{equation}

5082: To finish the construction of $\dot J$ we use the fact that $D^{N^* \otimes

5083: K_S}_{u_0, J_0} \dot \psi$ vanishes in a neighborhood of each cusp-point

5084: $z^*_i$. This yields the surjectivity of $\yps^\psi_\bfnu$ and

5085: the first assertion of the lemma.

5086: 

5087: \smallskip

5088: The second and third assertions follow from previous considerations. \qed

5089: 

5090: 

5091: 

5092: \newsubsection[4.5]{(Non)existence of saddle points in the moduli space}

5093: The results obtained above in this section allow us to prove the main technical

5094: result of the paper. Let $X$ be a manifold of dimension $2n$,

5095: $\scrj$ an open connected set in the space of $C^\ell$-smooth almost complex

5096: structures on $X$ with $\ell >2$ non-integer, $S$ a closed surface of genus

5097: $g\ge1$, and $[C] \in \sfh_2(X,\zz)$ a homology class.

5098: 

5099: \newdefi{def4.5.1} A pseudoholomorphic  map $u:S \to X$ has an {\sl

5100: ordinary cusp} at $z^*\in S$ if for appropriate coordinates $z$ on $S$ and

5101: $(w_1, w_2)$ on $X$

5102: \begin{equation}

5103: u(z)=\bigl(z^2 +O(|z|^3), z^3 + O(|z|^{3+\alpha}) \bigr).

5104: \end{equation}

5105: This property is equivalent to the condition that $u$ has a cusp of order 1

5106: and the secondary cusp-index 0 at $z^*$.

5107: \end{defi}

5108: 

5109: \newthm{thm4.5.1} Let $h(t)=J_t$, $t\in I=[0,1]$, be a {\sl generic}

5110: path in $\scrj$ and $\scrm_h$ the corresponding relative moduli space of

5111: parameterized pseudoholomorphic curves of genus $g\ge1$ in the homology class

5112: $[C]$.

5113: 

5114: \sli If $n\ge 3$, then every critical point of the projection $\pi_h: \scrm_h

5115: \to I$ is represented by an {\sl imbedded} curve $C=u(S)$, $u: S \to X$;

5116: 

5117: \slii If $n=2$, then every critical point of the projection $\pi_h: \scrm_h

5118: \to I$ is represented by a curve $C=u(S)$ such that:

5119: \begin{itemize}

5120: \item the only singularities on $C$ are nodes or ordinary cusps;

5121: %nodal an ordinary cuspidal ones;

5122: \item the possible number of cuspidal points $\vkappa$ on $C$ is 

5123: \begin{equation}\eqqno(4.5.2)

5124: \mu \le \vkappa \le \mu +g-1,

5125: \end{equation}

5126: where $\mu \deff \la c_1(X), [C] \ra$ and $g$ is the (geometric) genus of $C$,

5127: $g=g(S)$;

5128: \item the saddle index of $d^2\pi_h$ at $C$ is at least $\vkappa$, \ie

5129: $$

5130: \sind_C d^2\pi_h \ge \vkappa \ge \mu .

5131: $$

5132: \end{itemize}

5133: 

5134: \sliii In the case when the inequality \eqqref(4.5.2) is a contradiction, 

5135: the claim \slii has the following meaning:

5136: \begin{itemize} 

5137: \item If $g=0$, then $\pi_h$ has no critical points;

5138: \item If $\mu +g -1<0$, then the space $\scrm_h$ is empty for generic $h$.

5139: \end{itemize}

5140: \end{thm}

5141: 

5142: Before giving the proof we must specify the meaning of the notion {\sl generic

5143: path}. One of the most reasonable conditions is that any two {\sl regular}

5144: almost complex structures $J_0,\; J_1 \in \scrj$ (see \S\.{\sl2.3}) can be

5145: connected by a path $\{J_t\}_{t\in I=[0,1]}$ with the property stated in

5146: the theorem. To ensure this we need the following easy

5147: 

5148: \newprop{prop4.5.2} Let $F: \scrx \to \scry$ be a $C^1$-smooth

5149: Fredholm map between separable Banach manifolds. Assume that $\scry$ is

5150: connected and that the index of $F$ is at most $-2$. Then the set $\scry \bs

5151: F(\scrx)$ is path-connected.

5152: \end{prop}

5153: 

5154: \state Remark. The proposition generalizes the obvious fact that submanifolds 

5155: of codimension at least 2 do not divide the ambient manifold. Note that one can 

5156: have at most countably many connected components of $\scrx$ and that on these 

5157: components the index of $F$ can vary from component to component.

5158: 

5159: \statep Proof. \refthm{thm4.5.1}. We already know from \refsection{2} that for 

5160: a generic path $h(t) =J_t$ in $\scrj$ the set $\scrm_h$ is a manifold. In previous

5161: paragraphs of this section we have showed that critical points $[u,J]$ of the

5162: projection $\pi_h: \scrm_h \to I$ have an intrinsic description independent of the

5163: particular choice of the path $J_t$. Moreover, the quadratic form $d^2\pi_h$ at 

5164: these points also admits a similar intrinsic description. Furthermore, we have 

5165: found a stratification of the set of ``suspicious''

5166: points $[u,J] \in \scrm$ by submanifolds and estimated their codimension. It

5167: remains to find the strata with the Fredholm index $\le -2$ over $\scrj$ and

5168: apply \propo{prop4.5.2}.

5169: 

5170: The ``suspicious'' points $[u,J]$ on $\scrm$ are those with $\sfh^1_D(S,

5171: \scrn_u) \cong \rr$. They can be separated into classes according to the

5172: structure of the normal sheaf $\scrn_u$. Since the singular part $\scrn_u\sing$

5173: of $\scrn_u$ reflects the cusp-curves we are led to the spaces $\scrm_{=\mbfk}$

5174: of curves with prescribed order of cusps.

5175: 

5176: Denote by $\ind$ the index of the projection $\pr _{\!\!\scrj}: \scrm \to \scrj$,

5177: so that $\ind = 2(\la c_1(X), [C] \ra + (g-1)(3-n))$. If $\ind < 0$, then

5178: for a generic path $h(t)=J_t$ the set $\scrm_h$ is empty and the claim of the

5179: theorem holds. Thus we may assume that $\ind \ge0$. By \refthm{thm4.2.1}, we

5180: must ``pay'' at least $2(n-1)|\mbfk|$ dimensions to go to $\scrm_{=\mbfk}$. By

5181: \lemma{lem4.4.1}, we must ``pay'' further $\ind-2|\mbfk| +1$ dimensions to 

5182: obtain the condition $\sfh^1_D(S, \scrn_u) \cong \rr$. Note that $2(n-1)|\mbfk|

5183: \ge 2|\mbfk| +2$ if $n\ge 3$ and $\mbfk$ is non-trivial. Thus in the case $n\ge3$ 

5184: we ``overdraw'' our ``credit'' $\ind$ at least by $3$. This means that for 

5185: non-trivial $\mbfk$ the index of the projection from $\scrm_{=\mbfk,h^1=1}$ to 

5186: $\scrj$ is at most $-3$ and we can apply \propo{prop4.5.2}. Thus for $n\ge3$ any 

5187: critical point of $\scrm_h$ is represented by an immersion $u:S \to X$. 

5188: 

5189: \medskip

5190: In the case $n=2$ we can ``strike the balance'' in a similar way. Indeed, we

5191: come to the ``overdraw'' of at least $3$ dimensions in each of the following

5192: cases:

5193: 

5194: {\sl a)} $\sfh^1_D(S, \scrn_u) \cong \rr$ and there exists at least one

5195: cusp-point of cusp-order$\ge2$;

5196: 

5197: {\sl b)} $\sfh^1_D(S, \scrn_u) \cong \rr$ and there exists at least one

5198: cusp-point the secondary cusp-index $\ge1$;

5199: 

5200: {\sl c)} $\sfh^1_D(S, \scrn_u) \cong \rr$ and a non-trivial $\psi \in

5201: \sfh^1_D(S, N_u^* \otimes K_S )$ vanishes in at least one cusp-point.

5202: 

5203: \noindent

5204: Thus for generic $h(t)=J_t$ we can exclude all these possibilities. The

5205: remaining case admits only cusps of order 1 with the secondary cusp-index 0.

5206: This means that $u$ has only ordinary cusps. Since possibility

5207: {\sl c)} is excluded, each such cusp gives input 1 into the saddle index by

5208: \lemma{lem3.3.4}. Finally, we estimate the number of such cusps using

5209: \lemma{lem2.3.4}. 

5210: 

5211: \medskip

5212: Now we show that for $n\ge3$ and generic $h$ any critical point of $\scrm_h$ is 

5213: represented by an {\sl imbedding} $u:S \to X$. Since this result will be not used

5214: in the sequel, we give only a sketchy proof. 

5215: 

5216: Denote by $\scrm_{\sf{imm}}$ 

5217: the total moduli space of {\sl immersed pseudoholomorphic curves} with the same 

5218: topological data $g=g(S)$ and $[C] \in \sfh_2(X, \zz)$ as usual. In other words,

5219: $\scrm_{\sf{imm}}= \scrm_{=\mbfk}$ with trivial $\mbfk$. It follows easily from 

5220: \refsection{4} that this is an open set in the whole space $\scrm$. The space 

5221: $\scrm_{\sf{imm}}$ admits a natural stratification in which every stratum contains

5222: curves with the same number and type of multiple point on the image $C=u(S)$. 

5223: Obviously, the biggest stratum is the subspace of {\sl imbedded} curves, and this 

5224: is an open subset in $\scrm_{\sf{imm}}$. The next biggest stratum consists of 

5225: curves with exactly one transversal double point on $C=u(S)$. Let us denote it by 

5226: $\scrm_{\sf{imm}}^\times$ with the character $\times$ symbolizing a transversal 

5227: self-intersection of exactly 2 branches of $C=u(S)$. %%here

5228: 

5229: Locally, $\scrm_{\sf{imm}}^\times$ is defined by the condition $u(z_1) =u(z_2)$ 

5230: for some $z_1 \not= z_2 \in S$. Linearization of this condition is the equation 

5231: $$

5232: \pr_{N^\times}(v(z_1) -v(z_2)) =0

5233: $$

5234: on $[v, \dot J_S, \dot J] \in T_{[u,J]}\scrm_{\sf{imm}}$, where $N^\times$ denotes

5235: the plane in $T_{x^\times}X$ normal to both branches of $C=u(S)$ at the 

5236: point $x^\times=u(z_1)=u(z_2)$, \ie $N^\times \deff T_{x^\times}

5237: X /\bigl(du(T_{z_1}S) \oplus du(T_{z_2} S) \bigr)$. It is easy to see that this 

5238: condition is transversal. Thus $\scrm_{\sf{imm}}^\times$ is a $C^\ell$-smooth 

5239: submanifold of real codimension $2(n-2)$. Moreover, it follows from the proof of 

5240: \lemma{lem4.4.1} that biggest stratum is transversal to the subspace $\scrm

5241: _{\sf{imm}, h^1=1}$ of immersed curves with $\sf{h}^1(S, \scrn_u)=1$. Consequently,

5242: the space $\scrm^\times_{\sf{imm}, h^1=1} \deff \scrm_{\sf{imm}}^\times \cap \scrm

5243: _{\sf{imm}, h^1=1}$ of immersed curves with $\sf{h}^1(S, \scrn_u)=1$ and with 

5244: exactly one transversal self-intersection point is a $C^\ell$-smooth submanifold 

5245: of real codimension $2(n-2)$ in $\scrm_{\sf{imm}, h^1=1}$, and of real codimension

5246: $\ind + 1+ 2(n-2)$ in $\scrm$. Since $2(n-2) \ge2$ for $n\ge3$, we can apply 

5247: \propo{prop4.5.2}. 

5248: 

5249: The complementary strata $\scrm _{\sf{imm}}^{\mib{a}}$ of $\scrm_{\sf{imm}}$ 

5250: consist of curves having either several double points, or one double point with 

5251: tangency of higher degree, or even more complicated multiple points, with the index

5252: $\mib{a}$ encoding the number and the type of multiple points. A similar argument 

5253: shows that these strata $\scrm_{\sf{imm}}^{\mib{a}}$ are transversal to $\scrm

5254: _{\sf{imm},h^1 =1}$ and that the intersections $\scrm _{\sf{imm}, h^1=1}^{\mib{a}} 

5255: \deff \scrm _{\sf{imm}}^{\mib{a}} \cap \scrm _{\sf{imm}, h^1=1}$ are transversal. 

5256: The computation of the number of conditions shows that these strata have even 

5257: higher codimension in $\scrm_{\sf{imm}}$. So \propo{prop4.5.2} still applies. 

5258: This finishes the proof of the theorem. \qed

5259: 

5260: \medskip

5261: In applications, one needs a version of \refthm{thm4.5.1} for the case of curves

5262: passing through given fixed points $\mbfx=(x_1,\ldots,x_m)$ on $X$. Recall that

5263: for a $C^\ell$-smooth map $h: I\deff[0,1] \to \scrj$ we denote by $\scrm_{h,\mbfx}$

5264: the relative moduli space of $J_t=h(t)$-holomorphic curves passing through $\mbfx=

5265:  (x_1,\ldots,x_m)$ (see \refsubsection{2.4}). Let $\pi_{h,\mbfx}: \scrm_{h,\mbfx}

5266: \to I$ be the corresponding projection. We also assume that $\dimr X=4$.

5267: 

5268: \newthm{thm4.5.3} For a generic $h$ every critical point of the projection 

5269: $\pi_{h, \mbfx}: \scrm_{h,\mbfx} \to I$ is represented by a curve $C$ such that:

5270: \begin{itemize}

5271: \item the only singularities on $C$ are nodes or ordinary cusps;

5272: \item the marked points $(x_1,\ldots, x_m)$ are smooth points of $C=u(S)$;

5273: \item the possible number of cuspidal points $\vkappa$ on $C$ is 

5274: \begin{equation}\eqqno(4.5.3)

5275: \mu -m \le \vkappa \le \mu -m +g-1

5276: \end{equation}

5277: where $g$ is the (geometric) genus of $C$;

5278: \item the saddle index of $d^2\pi_h$ at $C$ is at least $\vkappa$, \ie

5279: $$

5280: \sind_C d^2\pi_h \ge \vkappa \ge \mu -m.

5281: $$

5282: \end{itemize}

5283: 

5284: In the case when the inequality \eqqref(4.5.3) is a contradiction

5285: the claim has the following meaning:

5286: \begin{itemize} 

5287: \item If $g=0$, then $\pi_{h, \mbfx}$ has no critical points;

5288: \item If $\mu -m +g -1 <0$, then the space $\scrm_h$ is empty for generic $h$.

5289: \end{itemize}

5290: 

5291: 

5292: \end{thm}

5293: 

5294: \proof The main observation in the proof is that after an appropriate modification

5295: all the results of this section remain valid also for curves passing through 

5296: fixed points. In particular, the most important formulas \eqqref(3.3.3) and 

5297: \eqqref(3.3.5) from \lemma{lem3.3.3} holds after replacing $\scrn\sing_u$ by 

5298: $\scrn\sing_{u,\mbfx}$. To show this we note first that {\sl Lemmas 

5299: \ref{lem3.2.3}, \ref{lem3.3.1}}, and {\sl \ref{lem3.3.2}} can be applied without

5300: any modification. After this, the proof of \lemma{lem3.3.3} applies with

5301: the only difference that the usual Gromov operator $D_{u,J}$ acting in $E$

5302: should be replaced by the operator $D_{u,-\mbfz,J}$ acting in $E_{-\mbfz}$.

5303: The validation of such a replacement is justified in \refsubsection{2.4}. Indeed,

5304: by the very definition, $D_{u,-\mbfz,J}$ is the restriction of $D_{u,J}$ to

5305: the subspace of sections of the subbundle $E_{u,-\mbfz} \subset E_u$.

5306: In a similar way one modifies the argumentation of \refsubsection{4.4}.

5307: 

5308: Finally, we note that the condition of coincidence of some cusp point of $C=u(S)$

5309: with some of marked points $x_1,\ldots,x_m$ defines a subset in $\scrm_\mbfx$

5310: which has a natural stratification into submanifolds of codimension $\ge2$.

5311: Every such stratum is defined by the cusp order $\mbfk$ of $C=u(S)$ and indication

5312: of the those cuspidal points which pass through the marked points $x_1,\ldots,

5313: x_m$. This means that every such stratum is a submanifold of the space $\scrm

5314: _{=\mbfk}$. Moreover, the codimension of every such stratum in $\scrm_{=\mbfk}$ is 

5315: $4a$, where $a$ is the number of cusps lying in the marked points. As in the case 

5316: $m=0$ above, one can show that the intersection of such a stratum with the space

5317: $\scrm_{=\mbfk, h^1=1}$ is transversal and has the expected codimension. 

5318: Hence we may conclude that for generic $h$ such a coincidence can not occur in 

5319: the critical points of $\pi_{h,\mbfx}$. The same argument is applied to show that 

5320: for generic $h$ there are no coincidence of the marked points $x_1,\ldots,x_m$ 

5321: with nodal points of the curve $C=u(S)$ representing a critical point of $\pi_h$. 

5322: \qed

5323: 

5324: 

5325: \newsection[5]{Deformation of nodal curves}

5326: 

5327: \newsubsection[5.2]{Nodal curves and Gromov compactness theorem}

5328: The total moduli space $\scrm$ constructed in {\sl Section 2} is not complete.

5329: More precisely, the projection $\pi_{\!\!\scrj}: \scrm \to \scrj$ is, in 

5330: general, not proper. This means that there exists a sequence $[u_i, J_i] \in 

5331: \scrm$ such that $J_i$ converges to $J_\infty \in \scrj$ but no subsequence of

5332: $\{ u_i \}$ converges in $L^{1,p}(S, X)$-topology, even after 

5333: reparameterization. Gromov compactness theorem ensures that there still exists 

5334: subsequence of $\{ u_i \}$ which converges \wrt the Gromov topology, which is 

5335: weaker that the Sobolev $L^{1,p}$-topology.

5336: 

5337: 

5338: \smallskip

5339: In the literature one can find several non-equivalent definitions for Gromov

5340: topo\-logy. In this paper  we shall use that one which is equivalent to the

5341: original definition of Gromov (\cite{Gro}). However, our version is more

5342: detailed in the sense that it is based on the notion of {\sl stable maps}.

5343: This notion for curves in a complex algebraic manifold $X$ was introduced by

5344: Kontsevich in \cite{K}, see also \cite{K-M}. Our definition of stable maps over 

5345: $(X,J)$ is simply a translation of this notion to almost complex manifolds.

5346: 

5347: \newdefi{def5.2.1} The {\sl standard node} is the complex analytic

5348: set

5349: \begin{equation}

5350: \scra_0 \deff \{ (z_1,z_2)\in \Delta^2 : z_1\cdot z_2 =0\}.

5351: \eqqno(5.2.1)

5352: \end{equation}

5353: A point on a complex curve is called a {\sl nodal point}, if has a

5354: neighborhood biholomorphic to the standard node. A {\sl nodal curve} $C$ is a

5355: complex analytic space of pure dimension 1 with only nodal points as

5356: singularities.

5357: \end{defi}

5358: 

5359: \newdefi{def5.2.1a} An annulus $A$ with a complex structure $J$ has

5360: {\sl conformal radius} $R>1$ if $A$ is biholomorphic to $A(1,R) \deff \{ z\in

5361: \cc \,:\, 1 <|z| < R \}$. Define a {\sl cylinder} $Z(a,b) \deff S^1

5362: \times [a,b] = \{ (\theta, t) : 0 \le \theta \le 2\pi,\; a\le t \le b \}$, $a<

5363: b$, with the complex structure $J_Z ({\d\over \d\theta}) \deff {\d\over \d t}$.

5364: Obviously, $Z(a,b)$ is also an annulus $A$ of conformal radius $R= e^{b-a}$.

5365: Also denote $Z_k\deff Z(k, k+1)$.

5366: \end{defi}

5367: 

5368: 

5369: \smallskip

5370: In other terminology, nodal curves are called {\sl prestable}. We shall

5371: always suppose that $C$ is connected and has a ``finite topology", \ie $C$ has

5372: finitely many irreducible components, finitely many nodal points, and that

5373: $C$ has a smooth boundary $\d C$ consisting of finitely many smooth circles

5374: $\gamma_i$, such that $\barr C \deff C \cup \d C$ is compact.

5375: 

5376: 

5377: \newdefi{def5.2.2} A real oriented surface with boundary $(\Sigma,

5378: \d\Sigma)$ {\sl parameterizes} a complex nodal curve $C$ if there

5379: is a continuous map $\sigma :\barr\Sigma \to \barr C$ such that:

5380: 

5381: \sli if $a\in C$ is a nodal point, then $\gamma_a = \sigma\inv(a)$ is a

5382: smooth imbedded circle in $\Sigma \bs \d \Sigma $, and if $a\not= b$ then

5383: $\gamma_a \cap \gamma_b= \emptyset$;

5384: 

5385: \slii $\sigma :\barr\Sigma \bs \bigcup_{i=1}^N\gamma_{a_i}\to \barr C \bs \{

5386: a_1,\ldots ,a_N\} $ is a diffeomorphism, where $a_1,\ldots ,a_N$ are the

5387: nodes of $C$.

5388: \end{defi}

5389: 

5390: 

5391: \bigskip

5392: \vbox{\xsize=.54\hsize\nolineskip\rm

5393: \putm[.17][-.01]{\gamma_1}%

5394: \putm[.31][.27]{\gamma_2}%

5395: \putm[.53][0.275]{\gamma_3}

5396: %

5397: \putm[.71][.265]{\gamma_4}%

5398: \putm[.87][.22]{\gamma_5}%

5399: \putt[1.1][0]{\advance\hsize-1.1\xsize%

5400: \centerline{Fig.~1}\smallskip%\parindent=0pt

5401: Circles $\gamma_1,..., \gamma_5$ are contracted by the parameterization

5402: map $\sigma$ to nodal points  $a_1, \ldots a_5$.

5403: }%

5404: \putm[.56][.37]{\bigg\downarrow\sigma}%

5405: \epsfxsize=\xsize\epsfbox{pic1.eps}%%

5406: \vskip.15\xsize

5407: %%  second picture

5408: \putm[.20][.05]{a_1}%

5409: \putm[.36][.255]{a_2}%

5410: \putm[.49][0.27]{a_3}%

5411: \putm[.725][.23]{a_4}%

5412: \putm[.86][.19]{a_5}%

5413: \epsfxsize=\xsize\epsfbox{pic2.eps}

5414: }

5415: 

5416: 

5417: \smallskip

5418: Note that such a parameterization is not unique: if $g:\barr\Sigma \to

5419: \barr\Sigma$ is any orientation preserving diffeomorphism then $\sigma \scirc

5420: g: \barr\Sigma \to \barr C$ is again a parameterization.

5421: 

5422: A parameterization of a nodal curve $C$ by a real surface can be considered as

5423: a method of ``smoothing'' of $C$. We shall also use an alternative method of

5424: ``smoothing'', the normalization.

5425: Consider the normalization $\hat C$ of $C$. Mark on each component of this

5426: normalization the pre-images (under the normalization map $\pi_C: \hat C \to

5427: C$) of nodal points of $C$. Let $\hat C_i$ be a component of $\hat C$. We can

5428: also obtain $\hat C_i$ by taking an appropriate irreducible component $C_i$,

5429: replacing nodes contained in $C_i$ by pairs of discs with marked points, and

5430: marking the remaining nodal points. Since it is convenient to consider components

5431: in this form, we make the following

5432: 

5433: \newdefi{def5.2.3} A {\sl component $C'$} of a nodal curve $C$ is the 

5434: normalization of an irreducible component of $C$ with marked points selected

5435: as above.

5436: \end{defi}

5437: 

5438: \smallskip

5439: This definition allows us to introduce Sobolev and H\"older spaces of

5440: functions and (continuous) maps of nodal curves.

5441: 

5442: \newdefi{def5.2.3a} A continuous map $u: C \to X$ is Sobolev $L^{1,p}

5443: $-smooth, $u\in  L^{1,p}(C, X)$ if the induced maps $u_i \deff u\ogran_{C_i}

5444: : C_i \to X$ of all of its components $C_i$ are $L^{1,p}$-smooth.

5445: % if so are all its restrictions on components of $C$. 

5446: The notion of $J$-holomorphic maps $u: C \to X$ is similarly defined. 

5447: For $u \in L^{1,p}(C, X)$ define $E_u \deff u^*TX$. Thus $E_u$ is determined 

5448: by restricting to each component and by identifying fibers over pairs $(z', z'')$ 

5449: of marked points corresponding to nodal points. An $L^{1,p}$-smooth section $v$ of 

5450: $E_u$ over $C$ is given by a collection of sections $v_{C_i} \in L^{1,p}(C_i, 

5451: E_u)$, one for every component $C_i$ of $C$, such that $v(z')=v(z'')$ for each pair

5452: $(z', z'')$ of marked points chosen as above. Denote by $L^{1,p}(C, E_u)$ the space

5453: of $L^{1,p}$-sections of $E_u$.

5454: \end{defi}

5455: 

5456: 

5457: \smallskip

5458: \newdefi{def5.2.4} The {\sl energy} or the {\sl area} of a

5459: continuous $L_\loc ^{1,2}$-smooth map \wrt a metric $h$ on $X$ is defined as

5460: \begin{equation}

5461: \area_h(u) \deff \norm{du}^2_{L^2(C)} = \int_C |du|^2_h

5462: \eqqno(5.2.2)

5463: \end{equation}

5464: \end{defi}

5465: 

5466: This definition depends only on the complex structure on $C$ but not on the

5467: choice of a metric on $C$ in the given conformal class. If an $\omega$-tame

5468: almost complex structure $J$ is given, there is a prefered choice of a metric

5469: $h$ on $X$ defined by $h(v,v) \deff \omega(v, Jv)$ for $v\in TX$.

5470: 

5471: \state Remark. Our definition of the area uses the following fact. Let $g$ be

5472: a Riemannian

5473: metric on $C$ compatible with $j_C$, $h$ a Riemannian metric on $X$, and $u:C

5474: \to X$ a $J$-holomorphic immersion. Then $\norm{du}^2_{L^2(C)}$ is independent

5475: of the choice of $g$ and coincides with the area of the image $u(C)$ \wrt

5476: the metric $h_J(\cdot, \cdot) \deff \half(h(\cdot, \cdot) + h(J\cdot, J\cdot)

5477: )$. The metric $h_J$ here can be seen as a ``Hermitization'' of $h$ \wrt

5478: $J$. It is well-known that $\norm{du}^2_{L^2(C)}$ is independent of the choice

5479: of a metric $g$ on $C$ in the same conformal class, see \eg \cite{S-U}.

5480: %The independence of $\norm{du}^2_{L^2(C)}$ of the choice of a metric $g$

5481: %on $C$ in the same conformal class is a well-known fact, see \eg \cite{S-U}. 

5482: Thus we can use the flat metric $dx^2 + dy^2$ to compare area and energy.

5483: For a $J$-holomorphic map we obtain

5484: \begin{equation}

5485: \norm{du}^2_{L^2(C)}= \int_C |\d_x u|_h^2 + |\d_y u|_h^2 =

5486: \int_C |\d_x u|_h^2 + |J \d_x u|_h^2 = \int_C |du|_{h_J}^2 = \area_{h_J}(u(C)),

5487: \end{equation}

5488: where the last equality is another well-known result, see \eg \cite{Gro}. Since we

5489: consider varying almost complex structures on $X$, it is useful to know that

5490: we can use any Riemannian metric on $X$ having a reasonable notion of area.

5491: 

5492: \smallskip

5493: \newdefi{def5.2.5} A {\sl stable curve over $(X,J)$} is a pair

5494: $(C,u)$, where $C$ is a nodal curve and $u: C\to X$ is a $J$-holomorphic map

5495: satisfying the following condition: If $C'$ is a closed component of $C$ such

5496: that $u$ is constant on $C'$, then there exist only finitely many biholomorphisms

5497: of $C'$ which preserve the marked points of $C$. In this case $u$ is

5498: called a {\sl stable map}.

5499: \end{defi}

5500: 

5501: 

5502: \state Remark. One can see that stability condition is nontrivial only in the

5503: following cases:

5504: 

5505: \begin{itemize}

5506: \item[{\sl 1)}] some component $C'$ is biholomorphic to $\cp^1$ with 1 or 2

5507: marked points; in this case $u$ should be non-constant on any such component $C'$;

5508: \item[{\sl 2)}] some irreducible component $C'$ of $C$ is $\cp^1$ or a torus

5509: without nodal points.

5510: \end{itemize}

5511: 

5512: \noindent

5513: Since we consider only connected nodal curves, case {\sl 2)} can occur only

5514: if $C$ irreducible, \ie $C'=C$. In this case $u$ must be non-constant on

5515: $C$.

5516: 

5517: \newdefi{def5.2.6} A component $C'$ of a nodal curve $C$ is called

5518: {\sl non-stable} in the following cases:

5519: 

5520: { 1)} $C'$ is $\cp^1$ and has one or two marked points;

5521: 

5522: { 2)} $C'$ is $\cp^1$ or a torus and has no marked points.

5523: 

5524: \noindent

5525: Let $u: C \to X$ be a pseudoholomorphic  map. An irreducible component $C'$ of $C$

5526: is a {\sl ghost} component (\wrt $u$) if $u$ is constant on $C'$. The {\sl ghost 

5527: part $C^{gh}$ of $C$ (\wrt $u$)} is the union of all ghost components. In this 

5528: paper we shall deal only with the case when all ghost components are closed.

5529: 

5530: A map $u$ is {\sl non-multiple} if, except finitely many points $z\in C$, 

5531: one has $u\inv(u(z)) =\{z \}$. Note that this condition excludes also ghost 

5532: components. 

5533: \end{defi}

5534: 

5535: 

5536: \smallskip

5537: Now we are going to describe the Gromov topology on the space of stable curves 

5538: over $X$ introduced in \cite{Gro}. Let $\{J_n\}$ be a sequence of continuous 

5539: almost complex structures on $X$ which converges to $J_\infty$ in the $C^0

5540: $-topology. Furthermore, let $(C_n, u_n)$ be a sequence of stable curves over 

5541: $(X, J_n)$, such that all $C_n$ are parameterized by the same real surface $S$.

5542: 

5543: \newdefi{def5.2.7} We say that $(C_n,u_n)$ {\sl converges in the Gromov 

5544: topology to a stable $J_\infty$-holomorphic curve $(C_\infty,u_\infty)$ over 

5545: $X$} if the parameterizations $\sigma_n: \barr S \to \barr C_n$ and 

5546: $\sigma_\infty: \barr S \to \barr C_\infty$ can be chosen in such a way 

5547: that the following holds:

5548: 

5549: \sli $u_n\scirc \sigma_n$ converges to $u_\infty\scirc \sigma_\infty$ in the

5550: $C^0(S, X)$-topology;

5551: 

5552: \slii if $\{ a_k \}$ is the set of nodes of $C_\infty$ and $\{\gamma_k\}$ are

5553: the corresponding circles in $S$, then on any compact subset $K\comp

5554: S \bs \cup_k \gamma_k$ the convergence $u_n\scirc \sigma_n \to u_\infty

5555: \scirc \sigma_\infty$ is $L^{1,p}(K, X)$ for all $p< \infty$;

5556: 

5557: \sliii for any compact subset $K\comp \barr S \bs \cup_k\gamma_k$ there

5558: exists $n_0=n_0(K)$ such that $ \sigma_n^{-1}(\{ a_k \}) \cap K= \emptyset$

5559: for all $n\ge n_0$ and the complex structures $\sigma_n^*j_{C_n}$ converge

5560: smoothly to $\sigma_\infty^*j_{C_\infty}$ on $K$;

5561: 

5562: \sliv the structures $\sigma_n^*j_{C_n}$ are independent of $n$ near the

5563: boundary $\d S$.

5564: \end{defi}

5565: 

5566: 

5567: \medskip

5568: Condition \sliv is trivial if $S $ is closed, but it is useful when

5569: one considers the ``free boundary case'', \ie when $S$ (and thus all

5570: $C_n$) are not closed and no boundary condition is imposed.

5571: 

5572: The reason for introducing the notion of a curve stable over $X$ is similar

5573: to the one for the Gromov topology. We are looking for a completion of the

5574: space of smooth imbedded pseudoholomorphic  curves which has ``nice''

5575: properties, namely: \.1) such a completion should contain the limit of a

5576: subsequence of every sequence of smooth curves which is bounded in an appropriate

5577: sense; \.2) such a limit should also exist for every sequence in the

5578: completed space; \.3) such a limit should be unique. The Gromov's compactness

5579: theorem ensures us that the space of curves stable over $X$ has these nice

5580: properties.

5581: 

5582: 

5583: \newdefi{def5.2.8} Let $C_n$ be a sequence of nodal curves,

5584: parameterized by the same real surface $S$. We say that the complex

5585: structures on $C_n$ {\sl do not degenerate near boundary}, if there exist

5586: $R>1$, such that for any $n$ and any boundary circle $\gamma_{n, i}$ of $C_n$

5587: there exist an annulus $A_{n,i} \subset C_n$ adjacent to $\gamma_{n, i}$,

5588: such that all $A_{n,i}$ are mutually disjoint, do not contain nodal points of

5589: $C_n$, and have the same conformal radius $R$.

5590: \end{defi}

5591: 

5592: Since the conformal radii of all $A_{n, i}$ are all the same, we can identify them

5593: with $A(1,R)$. This means that all changes of complex structures of $C_n$

5594: take place away from boundary. The condition is trivial if $C_n$ and $S$

5595: are closed, $\d S = \d C_n = \emptyset$.

5596: 

5597: \state Remark. Changing our parameterizations $\sigma_n: S \to C_n$, we

5598: may suppose that for any $i$ the pre-image $\sigma_n\inv (A_{n,i} )$ is the

5599: same annulus $A_i$ independent of $n$.

5600: 

5601: \medskip

5602: Now we state Gromov's compactness theorem for stable curves. Assume that

5603: $X$ is a compact manifold and fix some Riemannian metric $h$ on $X$.

5604: 

5605: \newthm{thm5.2.1} Let $(C_n,u_n)$ be a sequence of stable

5606: $J_n$-holo\-mor\-phic curves over $X$ with parameterizations $\delta_n: 

5607: S \to C_n$. Suppose that:

5608: 

5609: \begin{itemize}

5610: \item[{\sl a)}] $\{J_n\}$ is a sequence of continuous almost complex structures 

5611: on $X$, which converges to $J_\infty$ in the $C^0$-topology;

5612: 

5613: \item[{\sl b)}] there is a  constant $M$ such that $\area_h [u_n (C_n)]\le M$

5614: for all $n$;

5615: 

5616: \item[{\sl c)}] complex structures on the $C_n$ do not degenerate near the

5617: boundary.

5618: 

5619: \end{itemize}

5620: 

5621: Then there is a subsequence $(C_{n_k},u_{n_k})$ and parameterizations $\sigma

5622: _{n_k}: S \to C_{n_k}$, such that $(C_{n_k}, u_{n_k}, \sigma_{n_k})$

5623: converges to a $J_\infty$-ho\-lo\-mor\-phic curve $(C_\infty,

5624: u_\infty, \sigma_\infty)$ stable over $X$. 

5625: 

5626: Moreover, the limit curve $(C_{n_k}, u_{n_k})$ is unique up to the choice

5627: of the parameterization $\sigma_\infty$.

5628: 

5629: Furthermore, if the structures $\delta_n^*j_{C_n}$ are constant on the fixed

5630: annuli $A_i$, each adjacent to a boundary circle $\gamma_i$ of $S$, then

5631: the new parameterizations $\sigma_{n_k}$ can be taken equal to $\delta_{n_k}$

5632: on some subannuli $A'_i \subset A_i$, also adjacent to $\gamma_i$.

5633: \end{thm}

5634: 

5635: A detailed proof of the theorem in the stated form can be found in \cite{Iv-Sh-3}.

5636: We also refer to the original proof of Gromov in \cite{Gro}.

5637: 

5638: \smallskip

5639: Gromov's compactness theorem induce a natural completion of the moduli space 

5640: $\scrm$. Let a closed real surface $S$ of genus $g$ and a homology class $A \in 

5641: \sfh_2(X,\zz)$ be given.

5642: 

5643: \newdefi{def5.2.9} Nodal $J$-holomorphic curves $u': C' \to X$ and $u'': C'' \to 

5644: X$ are {\sl equivalent} if there exists a biholomorphism $\phi: C' \to C''$ with 

5645: $u' = u'' \scirc \phi$. The {\sl total moduli space $\barm^{st}$ of stable nodal 

5646: curves over $X$} is the set of equivalence classes $[C,u, J]$ with $J \in \scrj$ 

5647: and $u: C \to X$ a stable $J$-holomorphic curve representing a given class $A\in 

5648: \sfh_2(X,\zz)$. The space $\barm^{st}$ is equipped with the {\sl Gromov topology}

5649: in which a sequence $[C_n,u_n, J_n]$ converges to $[C_\infty,u_\infty, J_\infty]$ 

5650: if $J_n$ converges to $J_\infty$ in the $C^\ell$-topology and $(C_n, u_n)$ to $(C

5651: _\infty, u_\infty,)$ in the sense of \refdefi{def5.2.7}. Denote by $\pr_{\!\!

5652: \scrj}^{st}$ the natural projection $\pr_{\!\!\scrj}^{st}:  [C, u,J] \in \barm

5653: ^{Gr} \mapsto J \in \scrj$.

5654: 

5655: Define the {\sl Gromov compactification} $\barm^{Gr}$ of the total moduli space 

5656: $\scrm$ of pseudoholomorphic curves $X$ as the closure of $\scrm$ in $\barm^{st}$.

5657: \end{defi}

5658: 

5659: Note that every fiber $\barm^{Gr}_J \deff \barm^{Gr} \cap \bigl(\pr_{\!\!\scrj}

5660: ^{st} \bigr)\inv (J)$ is compact. Note also that in general $\barm^{st} \not=

5661: \barm^{Gr}$. This means that there are stable curves $[C,u,J] \in \barm^{st}$

5662: which can not be reached from $\scrm$.

5663: 

5664: 

5665: \newsubsection[5.2c] {The cycle topology for pseudoholomorphic curves} The 

5666: Gromov compactness theorem gives a precise description of the behavior of {\sl 

5667: parameterized} pseudoholomorphic curves at ``infinity'' of the total moduli

5668: space. However, what we are really interested in is not a pseudoholomorphic map

5669: $u:C \to X$ itself but rather the image $u(C)\subset X$, \ie a {\sl 

5670: non-pa\-ra\-me\-te\-rized} pseudoholomorphic curve. The natural space

5671: where non-pa\-ra\-me\-te\-rized curves ``live'' is the space $\scrz_2(X)$ of 

5672: 2-currents on the ambient manifold $X$. Recall that $\scrz_2(X)$ is the dual space

5673: to the space $C^\infty(X, \Lambda^2X)$ of smooth 2-forms on $X$ (see \eg

5674: \cite{Gr-Ha}, {\sl Chapter 3}). 

5675: 

5676: 

5677: \newdefi{def5.2c.1} Let $X$ be a manifold, $C$ an abstract nodal curve with the 

5678: smooth boundary $\d C$ such that $\barr C \deff C \cup \d C$ is compact, and $u: 

5679: C \to X$ a map which is $L^{1,p}$-smooth up to boundary. Define the {\sl cycle 

5680: $u[C]$ associated with the map $u: C \to X$} as the current whose pairing with 

5681: a smooth $2$-form $\phi$ on $X$ equals $\< u(C), \phi \> \deff \int_C u^*\phi$. In

5682: this case we also say that $u[C]$ is {\sl represented by the map $u: C \to X$}.

5683: 

5684: If additionally $u: C \to X$ is $J$-holomorphic \wrt some an almost complex 

5685: structure on $X$, we call $C' \deff u[C]$ an {\sl $J$-holo\-mor\-phic curve $C$ 

5686: \underline{\it in} $X$}. In this case we say that the $J$-holomorphic curve $(C, 

5687: u)$ \underline{\it over} $X$ and the map $u:C\to X$ {\sl represent the curve 

5688: $C'= u[C]$}. The set $u(C)$ is called the {\sl support of $C'$} and denoted 

5689: by $\supp(C')$.

5690: 

5691: A curve $C'$ in $X$ is {\sl non-multiple} if it can be represented by a 

5692: non-multiple pseudoholomorphic map $u:C\to X$ (see {\sl Definitions 

5693: \ref{def1.2.2}} and {\sl\ref{def5.2.6}}). In this case we identify the set $u(C)$ 

5694: and the current $u[C]$ and use the same notation $u(C)$.

5695: 

5696: A sequence of cycles $u_n[C_n]$  {\sl converges to a cycle $u_\infty[C_\infty]$}

5697: if $\< u_n(C_n), \phi \>$ converges to $\< u_\infty(C_\infty), \phi \>$ for any 

5698: smooth smooth $2$-form $\phi$ on $X$. In other words, the cycle topology is the 

5699: topology induced from the space of currents $\scrz_2(X)$.

5700: \end{defi}

5701: 

5702: \smallskip

5703: \newlemma{lem5.2c.1} \sli Let $(X,J)$ be an almost complex manifold

5704: and $(C, u)$, $(C', u')$ closed $J$-holomorphic curves over $X$. Assume that

5705: \begin{itemize}

5706: \item $C$ and $C'$ are parameterized by the same closed surface $S$;

5707: \item $(C, u)$ contains no multiple and ghost components;

5708: \item the associated cycles $u[C]$ and $u'[C']$ coincide. 

5709: \end{itemize}

5710: Then $(C, u)$ are $(C', u')$ equivalent.

5711: 

5712: \slii Let $J_n$ be a sequence of continuous almost complex 

5713: structures on $X$ which converges to an almost complex structure $J_\infty$ in the

5714: $C^0$-topology, and $(C_n, u_n)$ a sequence of stable $J_n$-holomorphic closed 

5715: curves over $X$ which converges to $(C_\infty, u_\infty)$ in the Gromov topology. 

5716: Then $u_n[C_n]$ converges to $u_\infty[C_\infty]$ in the cycle topology.

5717: 

5718: \sliii Let $J_n$ be a sequence of continuous almost complex 

5719: structures on $X$ which converges to an almost complex structure $J_\infty$ in the

5720: $C^0$-topology, $(C_n, u_n)$ a sequence of stable $J_n$-holomorphic closed 

5721: curves over $X$, and $(C_\infty, u_\infty)$ a parameterized $J_\infty$-holomorphic 

5722: curve. Assume that 

5723: \begin{itemize}

5724: \item $u_n[C_n]$ converges to $u_\infty[C_\infty]$ in the cycle topology;

5725: \item $C_n$ and $C_\infty$ are parameterized by the same closed surface $S$;

5726: \item $(C_\infty, u_\infty)$ contains no multiple and ghost components;

5727: \item $J_\infty$ is $C^1$-smooth.

5728: \end{itemize}

5729: Then $(C_n, u_n)$ converges to $(C_\infty, u_\infty)$ in the Gromov topology. 

5730: \end{lem}

5731: 

5732: \proof {\sl Part \slip}. The hypotheses on $(C, u)$ and $(C', u')$ imply

5733: that $(C', u')$ also contains no multiple and ghost components. The claim then 

5734: follows from the unique continuation property of pseudoholomorphic curves

5735: (see \lemma{lem1.2.4}).

5736: 

5737: {\sl Part \sliip}. This follows from the definition of the Gromov 

5738: topology and the description of the convergence at nodes given in 

5739: {\sl Step 0)} of the proof of \lemma{lem5.2.2} below.

5740: 

5741: {\sl Part \sliiip}. 

5742: Fix a $J_\infty$-Hermitian metric $h$ on $X$. Let $\omega$ be the associated 

5743: 2-form, $\omega(v,w) \deff h(J_\infty v,w)$. Then the structures $J_n$ are 

5744: $\omega$-tame for $n\gg1$.

5745: Note that even if $\omega$ is apriori only continuous and not closed, the notion 

5746: of $\omega$-tameness is still meaningful. Moreover, the $\omega$-tameness provides

5747: a uniform bound of $h$-area of $u_n[C_n]$. Consequently, some subsequence

5748: $(C_{n'}, u_{n'})$ of $(C_n, u_n)$ converges in the Gromov topology to 

5749: a stable $J_\infty$-holomorphic curve $(C'_\infty, u'_\infty)$. The hypotheses

5750: of the corollary imply that $(C'_\infty, u'_\infty)$ is equivalent to 

5751: $(C_\infty, u_\infty)$ and the result follows.\qed

5752: 

5753: 

5754: \state Remark. The meaning of \lemma{lem5.2c.1} is that, in the absence of 

5755: multiple and ghost components, the notions of pseudoholomorphic curves {\sl over} 

5756: $X$ and {\sl in} $X$ essentially coincide. The same also holds for the Gromov 

5757: and the cycle topologies. Note also that several authors (\cite{Ye}, 

5758: \cite{Pa-Wo}, \cite{Hum}) considered a weaker version of the Gromov compactness

5759: theorem where the cycle topology is used instead of the Gromov one. 

5760: 

5761: 

5762: \newdefi{def5.2c.2} Define the {\sl cycle compactification $\barm$} of the total 

5763: moduli space $\scrm$ as the set of pairs $(C,J)$, where $J \in \scrj$ and $C$ is a

5764: $J$-holomorphic curve {\it in} $X$ which considered as the cycle $u[C']$ 

5765: represented 

5766: by some $J$-holomorphic map $u: C' \to X$. Equip the space $\barm$ with the 

5767: {\sl cycle topology} in which a sequence $(C_n, J_n)$ converges to $(C_\infty, 

5768: J_\infty)$ if $J_n$ converges to $J_\infty$ in the $C^\ell$-topology and $C_n$ 

5769: converges to $C_\infty$ in the sense of \refdefi{def5.2.7}. Denote by $\pr_{\!\!

5770: \scrj}$ the natural projection $\pr_{\!\!\scrj}:  (C,J) \in \barm \mapsto J \in 

5771: \scrj$. Define the natural projection $\pr^{Gr}:\barm^{Gr} \to \barm$ by

5772: $\pr^{Gr}: (C, u,J) \in \barm^{Gr} \mapsto u[C]\in \barm$.

5773: \end{defi}

5774: 

5775: 

5776: \newdefi{def5.2c.3} A {\sl normal parameterization} of a $J$-holomorphic curve 

5777: $C$ in $X$ is given by a Riemann surface $S$ (possibly not connected) and a map $u:

5778: S \to X$ such that 

5779: \begin{itemize}

5780: \item[1)] $u[S]=C$;

5781: \item[2)] $u$ is $J$-holomorphic and $L^{1,p}$-smooth up to boundary;

5782: \item[3)] the restriction of $u$ to every connected component of $S$ is 

5783: non-multiple; in particular, there are no ghost components,  \ie $u$ is 

5784: non-constant on every connected component of $S$;

5785: \item[4)] the number of boundary circles of $S$ is as small as possible.

5786: \end{itemize}

5787: \end{defi}

5788: 

5789: \state Remark. Without condition (3) one could add new ghost spheres to $S$ 

5790: and make the Euler characteristic $\Chi(S)$ arbitrarily large. Condition (4) 

5791: excludes the possibility of dividing components of $S$ into pieces which also 

5792: allows to increase $\Chi(S)$.

5793: 

5794: \newlemma{lem5.2c.2} Let $J$ be a $C^1$-smooth almost complex structure on $X$,

5795: $C$ an abstract nodal curve, and $u: C\to X$ a $J$-holomorphic map which is an 

5796: imbedding near the boundary $\d C$. Then, up to a diffeomorphism, there exists 

5797: a unique normal parameterization $\ti u: S \to X$ of $u[C]$.

5798: \end{lem}

5799: 

5800: \proof Let $C= \cup_i C_i$ be the decomposition of $C$ into irreducible 

5801: components. Denote by $m_i$ the degree $u$ on $C_i$. This means that

5802: \begin{itemize}

5803: \item $m_i=0$ if $C_i$ is a ghost component, \ie $u$ is constant on $C_i$;

5804: \item $m_i$ is the number of points in the preimage $u\inv(x)$ for a

5805: generic $x\in u(C_i)$ otherwise.

5806: \end{itemize}

5807: For every non-zero $m_i$, denote by $S_i$ the normalization of the image $u(C_i)$.

5808: Denote by $\ti u_i: S_i \to u(C_i)$ the corresponding normalization maps. In 

5809: particular, $m_i=1$ for every non-closed component $C_i$ and in this case $S_i$ 

5810: is the normalization of $C_i$. Define $S$ as the disjoint union of the surfaces

5811: $S_i$, each taken $m_i$ times. Let $\ti u: S \to X$ be the map which coincides 

5812: with the composition $\ti u_i: S_i \to u(C_i) \hook X$ on every copy of $S_i$.

5813: One can see that $\ti u: S \to X$ is a normal parameterization of $u[C]$.

5814: 

5815: The uniqueness of such a normal parameterization follows from \lemma{lem1.2.4}

5816: \qed

5817: 

5818: \state Remark. Let us give an example showing that the condition on the behavior 

5819: of $u$ at the boundary imposed in \lemma{lem5.2c.2} is necessary. Define curves

5820: $C'$ and $C''$ as the disjoint unions $C'\deff \{ z\in \cc: |z|<2\} \sqcup \{ z\in

5821: \cc: 1<|z|<3\}$ and $C''\deff \{ z\in \cc: |z|<3\} \sqcup \{z\in \cc:

5822: 1<|z|<2\}$. Let $u': C' \to \cc$ and $u'': C'' \to \cc$ be the maps which are the 

5823: standard imbeddings on every component of $C'$ and $C''$. Then

5824: obviously $(C',u')$ and $(C'',u'')$ are not equivalent in the sense of 

5825: \refdefi{def5.2.9} and define non-equivalent normal parameterizations

5826: of $u'[C'] =u''[C'']$.

5827: 

5828: 

5829: \newcorol{cor5.2c.3} Under the hypotheses of \lemma{lem5.2c.2}, the curve $u[C]$,

5830: considered as a current $u[C]\in \scrz_2(X)$, admits a unique representation in 

5831: the form 

5832: $$

5833: \textstyle

5834: u[C]= \sum_i m_i u_i[C_i], 

5835: $$

5836: where the $u_i: C_i \to X$ are $J$-holomorphic maps and the $u_i(C_i)$ are 

5837: the irreducible components of $\supp(u[C])$. 

5838: \end{corol}

5839: 

5840: The corollary ensures that the notions of an irreducible component and the 

5841: multiplicity of a closed pseudoholomorphic curve {\it in} $X$ are well-defined.

5842: 

5843: \medskip

5844: The importance of the notion of a normal parameterization lies in the fact that it 

5845: allows us to define a natural stratification of the cycle compactification $\barm$

5846: of the total moduli space. Let $S$ be a given connected real surface $S$ of genus 

5847: $g$ and $[C] \in \sfh_2(X, \zz)$ a homology class, and $\barm=\barm(S, X, [C])$ 

5848: the cycle compactification of the total space $\scrm=\scrm(S, X, [C])$ of 

5849: irreducible pseudoholomorphic curves of genus $g$ in the homology class $[C]$. 

5850: Take $(C,J) \in \scrm(S, X, [C])$ and consider a normal parameterization $u': S' 

5851: \to X$ of $C$. Let $C= \sum_i m_i C_i$ be the decomposition into irreducible 

5852: components in the sense of \refcorol{cor5.2c.3}. Restricting $u'$ to 

5853: appropriate connected components we obtain normal parameterizations $u'_i: S'_i

5854: \to X$ of the corresponding $C_i$. 

5855: 

5856: \newdefi{def5.2c.4} The {\sl topological type of a component $C_i$} is the triple

5857: $(S'_i, m_i, [C_i])$, where $[C_i]$ denotes the homology class of $C_i$. The {\sl

5858: topological type $\bftau$ of a curve $(C,J) \in \scrm(S, X, [C])$} is the sequence

5859: of all topological types of components $(S'_i, m_i, [C_i])$ defined up to 

5860: permutation. 

5861: \end{defi}

5862: 

5863: \newlemma{lem5.2c.4} \sli The space $\scrm_\bftau$ of pseudoholomorphic curves 

5864: $(C,J) \in \barm(S, X, [C])$ of a given topological type $\tau$ is a 

5865: $C^\ell$-smooth Banach manifold. The natural projection $\pr_{\!\!\scrj}: \scrm

5866: _\bftau \to \scrj$ is a $C^\ell$-smooth Fredholm map.

5867: 

5868: \slii The space $\barm=\barm(S, X, [C])$ is the union of subspaces

5869: $\scrm_\bftau$. 

5870: \end{lem}

5871: 

5872: \proof 

5873: The decomposition $C= \sum_i m_i C_i$ of every $(C,J) \in \scrm_\bftau$

5874: shows that $\scrm_\bftau$ is the fiber product of the spaces $\scrm(S'_i, X, 

5875:  [C_i])$ over all triples $(S'_i, m_i, [C_i]) \in \bftau$ taken over the space

5876: $\scrj \!\!$, 

5877: $$

5878: \textstyle

5879: \scrm_\bftau=\prod_{(S'_i, m_i, [C_i]) \in \bftau } 

5880: \scrm(S'_i, X, [C_i]) \bigm/ \!\!\!\scrj\!\!.

5881: $$

5882: Checking the transversality condition, one obtains the desired  differentiable 

5883: structure on $\scrm_\bftau$.

5884: 

5885: The second assertion of the lemma is obvious. \qed 

5886: 

5887: \smallskip

5888: 

5889: 

5890: \newsubsection[5.2a]{Fine apriori estimates for convergence at a node}

5891: For the purpose of this paper we need a refined version of the {\sl Second

5892: apriori estimate} given in \cite{Iv-Sh-3}, {\sl Lemma 3.4}. This gives a 

5893: precise description with estimates of the Gromov convergence in neighborhoods 

5894: of the contracted circles.

5895: 

5896: \newlemma{lem5.2.2} Let $X$ be a compact manifold $X$, $J^*$ a

5897: $C^{0,s}$-smooth almost complex structure on $X$ with $s>0$, and $h$ a metric on 

5898: $X$. Then there exist constants $\epsi=\epsi(X,h, J^*, s)>0$ and $C <\infty$ 

5899: such that for any $C^{0,s}$-smooth almost complex structure $J$ with 

5900: \begin{equation}

5901: \norm{J -J^*}_{C^{0,s}(X)} \le \epsi

5902: \eqqno(5.2a.1)

5903: \end{equation}

5904: and any $J$-holomorphic map $u: Z(0, l) \to X$ the condition

5905: \begin{equation}

5906: \norm{du}_{L^2(Z_k)} \le \epsi \quad \text{for any }k\in[0, l-1]

5907: \eqqno(5.2.3)

5908: \end{equation}

5909: implies the uniform estimate

5910: \begin{equation}

5911: \norm{du}^2_{L^2(Z_k)} \le

5912: C \cdot e^{-2k} \cdot \norm{du}^2_{L^2(Z(0, 2))} +

5913: C \cdot e^{-2 (l-k)} \cdot\norm{du}^2_{L^2(Z(l-2, l))}

5914: \eqqno(5.2.4)

5915: \end{equation}

5916: for any $k\in[1, l-2]$.

5917: \end{lem}

5918: 

5919: 

5920: 

5921: \proof \step0.\. {\sl Lemma 3.3} in \cite{Iv-Sh-3} states that

5922: under hypotheses of the lemma one has a ``local'' estimate

5923: \begin{equation}

5924: \norm{du}^2_{L^2(Z_k)} \le \msmall{\gamma \over 2}

5925: \left( \norm{du}^2_{L^2(Z_{k-1})} +

5926: \norm{du}^2_{L^2(Z_{k+1})}  \right ) \qquad \text{for any $k\in[1, l-2]$}

5927: \eqqno(5.2.5)

5928: \end{equation}

5929: with a universal constant $\gamma <1$. Then in {\sl Corollary 3.4} in

5930: \cite{Iv-Sh-3} it is shown that \eqqref(5.2.5) implies the estimate

5931: \begin{equation}

5932: \norm{du}^2_{L^2(Z_k)} \le

5933: e^{-2\alpha (k-1)} \cdot \norm{du}^2_{L^2(Z(0, 2))} +

5934: e^{-2\alpha (l-2-k)} \cdot\norm{du}^2_{L^2(Z(l-2, l))}

5935: \eqqno(5.2.6)

5936: \end{equation}

5937: for any $k\in[1, l-2]$ with a constant $\alpha >0$ related to $\gamma$

5938: by $\gamma = {1\over \cosh(2\alpha)}$.

5939: 

5940: \state Remark. Note that in the proof of the estimates \eqqref(5.2.5) and 

5941: \eqqref(5.2.6) are proven in \cite{Iv-Sh-3} under the following assumption:

5942: It is supposed that $J^*$ and $J$ in question are only continuous and that

5943: $\norm{J- J^*}_{C^0(X)} \le \epsi'$ for some $\epsi'=\epsi'(X,J^*,h)>0$

5944: independent of $J$. 

5945: 

5946: \medskip

5947: From the relation $\gamma = {1\over \cosh(2\alpha)}$ we see that the smaller the $

5948: \gamma$ we have the bigger the $\alpha$ in 

5949: \eqqref(5.2.6) we obtain. For our purpose it would be sufficient to prove 

5950: estimate \eqqref(5.2.5) with the parameter $\gamma^* \deff {1\over \cosh2}$. 

5951: Note however that in the ``ideal'' case when $(X,J,h)$ is $\cc^n$ with the 

5952: standard complex and Hermitian structures, $\gamma^*$ is exactly the best 

5953: possible constant, see the proof of {\sl Lemma 3.3} in \cite{Iv-Sh-3}. Thus one 

5954: can not expect that estimate \eqqref(5.2.5) holds with {\sl uniform} $\gamma \le 

5955: \gamma^*$. The idea is to consider \eqqref(5.2.5) with parameters $\gamma_k$ 

5956: depending on $k$ and to estimate the difference $\gamma_k -\gamma^*$.

5957: 

5958: \medskip\noindent

5959: \step1.\. {\sl Under hypotheses of the lemma, for any $k\in[1, l-2]$,

5960: one has the estimate

5961: \begin{equation}

5962: \norm{du}^2_{L^2(Z_k)} \le 

5963: \msmall{\gamma_k \over 2} 

5964: \cdot

5965: \left( \norm{du}^2_{L^2(Z_{k-1})} +

5966: \norm{du}^2_{L^2(Z_{k+1})}  \right )

5967: \eqqno(5.2.7)

5968: \end{equation}

5969: for 

5970: \begin{equation}

5971: \gamma_k \deff \gamma^* + 

5972: C_1\cdot\left(e^{-\alpha s k} + e^{-\alpha s (l-k)} \right)

5973: \eqqno(5.2a.2)

5974: \end{equation}

5975: with the parameter $\alpha >0$ as in \step0 and some constant $C_1$

5976: depending only on $X$, $h$, $J^*$, and $s$.}

5977: 

5978: \smallskip

5979: While proving this estimate we shall denote by $C$ a constant whose particular

5980: value is not important and which may not be the same in different formulas.

5981: The main condition is that these constants are {\sl uniform}, \ie independent

5982: of $J$, $u$, and $l$, and depend only on $X$, $h$, $J^*$, and $s$.

5983: 

5984: 

5985: Estimates  \eqqref(5.2.3) and  \eqqref(5.2.6) together with apriori 

5986: estimates show that

5987: \begin{equation}

5988: \diam(u(Z(k-1, k+2)) \le C \cdot 

5989: \left(e^{-\alpha k} + e^{-\alpha (l-k)} \right).

5990: \end{equation}

5991: Consequently, due to a uniform H\"older $C^{0,s}$-estimate on $J$,

5992: for the oscillation of $J$ on the image $u(Z(k-1, l-k))$ we obtain

5993: \begin{equation}\eqqno(5.2a.2a)

5994: \osc(J, u(Z(k-1, k+2))) \le C \cdot 

5995: \left(e^{-\alpha s k} + e^{-\alpha s (l-k)} \right).

5996: \end{equation}

5997: This implies that in a neighborhood of each $u(Z(k-1, k+2))$ there exist an 

5998: integrable structure $J\st$ and a flat (\ie Euclidean) metric $h\st$ such that

5999: \begin{equation}

6000: \norm{J -J\st}_{L^\infty(u(Z_k))} + \norm{h -h\st}_{L^\infty(u(Z_k))}

6001: \le C\cdot \left(e^{-\alpha s k} + e^{-\alpha s (l-k)} \right).

6002: \end{equation}

6003: Using this we obtain estimates

6004: \begin{align}

6005: \norm{\dbar\st u}_{L^2(Z_k)}

6006: = \norm{\dbar\st u - \dbar_J u}_{L^2(Z_k)}

6007: &\le \norm{J -J\st}_{L^\infty(u(Z_k))} \cdot \norm{du}_{L^2(Z_k)} \le

6008: \notag

6009: \\

6010: &\le C\cdot \left(e^{-\alpha s k} + e^{-\alpha s (l-k)} \right);

6011: \end{align}

6012: \begin{equation}

6013: \Bigl| \norm{du}_{L^2(Z_k), h} 

6014: -\norm{du}_{L^2(Z_k), h\st} \Bigr| \le

6015: C\cdot \left(e^{-\alpha s k} + e^{-\alpha s (l-k)} \right)

6016: \cdot \norm{du}_{L^2(Z_k), h}.

6017: \end{equation}

6018: In particular, we can use $h\st$ instead of $h$ in our estimates.

6019: 

6020: Now consider $U$ as a subset on $\cc^n$ with the standard $J\st$ and $h\st$.

6021: Then we can find $u_\dbar\in L^{1,2}(Z(k-1, k+2), \cc^n)$ such that $\dbar\st

6022: u_\dbar= \dbar\st u$ and

6023: \begin{equation}

6024: \norm{du_\dbar}_{L^2(Z(k-1, k+2))} \le C \norm{\dbar\st u}

6025: _{L^2(Z(k-1, k+2))}.

6026: \eqqno(5.2a.3)

6027: \end{equation}

6028: Set $u_\scro \deff u -u_\dbar$, so that $u_\scro$ is $J\st$-holomorphic.

6029: It follows that %Then we obtain

6030: \begin{equation}

6031: \norm{du_\scro}^2_{L^2(Z_k)} \le \msmall{\gamma^* \over2}

6032: \left( \norm{du_\scro}^2_{L^2(Z_{k-1})} +

6033: \norm{du_\scro}^2_{L^2(Z_{k+1})}  \right).

6034: \eqqno(5.2.8)

6035: \end{equation}

6036: Together with the estimates on $u_\dbar$, \eqqref(5.2.8) implies 

6037: \eqqref(5.2.7).

6038:           

6039: \medskip\noindent

6040: \step2.\. {\sl There exist a uniform $k_0=k_0(X, h, J^*,s)$ and $A^\pm_k$, 

6041: $k=k_0, \ldots,l-k_0 $ with the properties

6042: 

6043: \sli $A^\pm_k$ are ``supersolutions'' of \eqqref(5.2.7), \ie

6044: \begin{equation}

6045: \eqqno(5.2a.4a)

6046: A^\pm_k \ge \msmall{\gamma_k \over2} (A^\pm_{k-1} + A^\pm_{k+1})

6047: \end{equation}

6048: 

6049: \slii  $A^\pm_k$ have the desired exponential decay}

6050: \begin{equation}

6051: \eqqno(5.2a.5a)

6052: A^+_k \le C\cdot e^{-2k},

6053: \qquad\qquad

6054: A^-_k \le C\cdot e^{-2(l-k)}.

6055: \end{equation}

6056: 

6057: \sliii $\gamma_k <1$ for $k \in [k_0, l-k_0]$.

6058: 

6059: 

6060: \smallskip

6061: Fix $k^*\in \zz$ such that $l-1 \le k^* <l+1$, so that $k^* \approx {l\over2}$.

6062: Set 

6063: \begin{align}

6064: A^+_k & \deff \cases 

6065:       e^{-2k -{1\over k}} & 0\le k\le k^* \\

6066:       e^{-2k -{1\over k^*} + {1\over l-k}- {1\over l-k^*}} 

6067:    \hskip7pt& k^*\le k \le l

6068: \endcases  \\

6069: A^- _k& \deff \cases 

6070:       e^{-2(l-k) -{1\over l-k^*} + {1\over k}-{1\over k^*}}

6071:     & 0\le k\le k^* \\

6072:       e^{-2(l-k) -{1\over l-k}} & k^*\le k \le l

6073: \endcases  

6074: \end{align}

6075: Making the Taylor expansion in $k\inv$ we obtain 

6076: $$

6077: \msmall{ 2 A^\pm_k \over A^\pm_{k-1} + A^\pm_{k+1}}= 

6078: \cases

6079:    \msmall { 1\over \cosh(2)}+  

6080:    \msmall { \sinh(2) \over \cosh^2(2)} \cdot k^{-2} + O(k^{-3})

6081:  & \text{for }0< k \le k^*;

6082: \\

6083:    \msmall { 1\over \cosh(2)}+  

6084:    \msmall { \sinh(2) \over \cosh^2(2)} \cdot (l-k)^{-2} + O((l-k)^{-3})

6085:  & \text{for }k^*\le k <l;

6086: \endcases

6087: $$

6088: So the existence of the desired $k_0(s)$ follows from the asymptotic behavior

6089: $C_1 e^{-\alpha s k} = o(k^{-2})$ for $k\lrar \infty$.

6090: 

6091: 

6092: \medskip\noindent

6093: \step3. {\sl There exists a constant $C_2= C_2(X,h, J^*,s)$ such that 

6094: 

6095: %\begin{equation}

6096: $\qquad

6097: \norm{du}^2_{L^2(Z_k)} \le C_2\cdot \left(

6098: A^+_k \cdot \norm{du}^2_{L^2(Z(0,2))} +

6099: A^-_k \cdot\norm{du}^2_{L^2(Z(l-2,l))} 

6100: \right)

6101: %\eqqno(5.2.9)

6102: $%\end{equation}

6103: \lineeqqno(5.2.9)

6104: 

6105: \smallskip\noindent

6106: for any $k\in[k_0, l-k_0]$ with the uniform constant $k_0=k_0(s)$ chosen as 

6107: above.}

6108: 

6109: 

6110: \smallskip

6111: Obviously, \eqqref(5.2.9) implies the claim of the lemma. Set 

6112: $$

6113: A^*:= C_2 \cdot \left(A^+_k \cdot \norm{du}^2_{L^2(Z(0,2))} +

6114: A^-_k \cdot\norm{du}^2_{L^2(Z(l-2,l))} \right)

6115: $$

6116: and choose a constant $C_2$ so that 

6117: $$

6118: A^*_{k_0} \ge \norm{du}^2_{L^2(Z_{k_0})}

6119: \qquad \text{and} \qquad

6120: A^*_{l-k_0} \ge

6121: \norm{du}^2_{L^2(Z_{l-k_0})}. 

6122: $$

6123: Then by \eqqref(5.2.7) and \eqqref(5.2a.4a) 

6124: \begin{equation}\notag

6125: \norm{du}^2_{L^2(Z_k)} - A^*_k \le 

6126: \msmall{\gamma_k \over2} \cdot 

6127: \left(\norm{du}^2_{L^2(Z_{k-1})} - A^*_{k-1}

6128: +\norm{du}^2_{L^2(Z_{k+1})} - A^*_{k+1} \right).

6129: \end{equation}

6130: Find $k_\max\in [k_0, l-k_0]$ realizing the maximum of

6131: $\norm{du}^2_{L^2(Z_k)} - A^*_k$. Then

6132: \begin{equation}

6133: \matrix%\format\l&\;\c\;&\l\\

6134: \norm{du}^2_{L^2(Z_{k_\max})} - A^*_{k_\max}

6135: &\le& \msmall{\gamma_{k_\max} \over2}

6136: \left(\norm{du}^2_{L^2(Z_{k_\max-1})} - A^*_{k-1}

6137: +\norm{du}^2_{L^2(Z_{k+1})} - A^*_{k_\max+1} \right)

6138: \cr\noalign{\vskip3pt}

6139: &\le&

6140: \gamma_{k_\max} \cdot( \norm{du}^2_{L^2(Z_{k_\max})} - A^*_{k_\max}).

6141: \endmatrix

6142: \end{equation}

6143: Since $\gamma_{k_\max}<1$, the last inequality holds only if

6144: $\norm{du}^2_{L^2(Z_{k_\max})} \le A^*_{k_\max}$. Thus

6145: \begin{equation}

6146: \norm{du}^2_{L^2(Z_k)} \le A^*_k

6147: \qquad\text{for any }k\in [k_0, l-k_0].

6148: \end{equation}

6149: This finishes the proof. \qed

6150: 

6151: 

6152: \smallskip

6153: \newthm{thm5.2.3} Let $J^*$ be a $C^{0,s}$-smooth almost complex structure on the 

6154: ball $B\subset \rr^{2n}$ with $0<s<1$. Then there exists $\epsi^*= \epsi^* (J^*,

6155: s)$ with the following property. For any almost complex  structure $J$ on $B$ with 

6156: $\norm{J-J^*}_{C^{0,s}(B)} \le \epsi^*$ and any $J$-holomorphic map 

6157: $u: Z(0,l) \to B(\half)$ with $l\ge3$ satisfying the condition 

6158: $$

6159: \norm{du}_{L^2(Z_k)} \le \epsi^*\qquad\text{for any }k\in[1,l]

6160: $$

6161: there exist a {\sl linear} complex structure $J\st$ in $\rr^{2n}$ and vectors 

6162: $v^+, v^0, v^- \in \rr^{2n}$ such that 

6163: \begin{multline}\eqqno(5.2.10)

6164: \left\Vert

6165: u - \left( e^{-t+J\st\theta}v^+ + v^0 + e^{t-J\st\theta}v^- \right)

6166: \right\Vert^2_{L^{1,2}(Z_k)} \le

6167: \\

6168: \le C^*  \cdot\bigl(k \;e^{-2(1+s)k}\norm{du}^2_{L^2(Z(0,2))} + 

6169: (l-k)\;e^{-2(1+s)(l-k)}\norm{du}^2_{L^2(Z(l-2,l))} 

6170: \bigr)\ 

6171: \end{multline}

6172: for any $k=1,\ldots, l-1$ with a constant $C^*=C^*(J^*,s) <\infty$ independent of 

6173: $J$, $l$, and $u$.

6174: \end{thm}

6175: 

6176: 

6177: \proof In fact, we prove that for the constant $\epsi^*$ one can take the 

6178: $\epsi$ from \lemma{lem5.2.2}. The proof also exploits the same ideas which were 

6179: used in the proof of that lemma. 

6180: 

6181: \smallskip\noindent {\sl Step 1}. 

6182: Let $J$ and $u: Z(0,l) \to B(\half)$ be as in the hypotheses of the theorem. For 

6183: $k= 1,\ldots,l$, let $x_k \in B(\half)$ be the average value of $u$ on $Z_k$ \wrt

6184: the cylinder metric, \ie

6185: $$

6186: x_k \deff \bint_{Z_k} u \deff

6187: \msmall{1\over 2\pi}\int_{(t,\theta) \in Z_k} u(t,\theta) 

6188: dt\,d\theta.

6189: $$

6190: Define the complex structures $J_k$ by $J_k\deff J(u(k,0))$. We consider every 

6191: $J_k$ as a {\sl linear} complex structure in $\rr^{2n}$, \ie constant in $x\in 

6192: \rr^{2n}$. Further, any $k=1,\ldots,l$ we define the metric $g_k$ setting $g_k(v, 

6193: w) \deff \half (g\st(v,w) + g\st(J_kv, J_kw))$, where $g\st$ denotes the standard 

6194: Euclidean metric in $\rr^{2n}$. Then $g_k$ are linear in the same sense as $J_k$. 

6195: In  computing various norms related to $Z_k$ or $Z(k-2, k+1)$, we shall use the 

6196: metric $g_k$ without indicating this in the notation. Observe that all $g_k$ are 

6197: equivalent since the $J_k$ are uniformly bounded. Further, convergence of 

6198: $J_{k_\nu}$ implies convergence of $g_{k_\nu}$.

6199: 

6200: For any $k=1,\ldots,l$ there exist uniquely defined vectors 

6201: $v^+_k, v^0_k, v^-_k \in \rr^{2n}$ such that for the function 

6202: $$

6203: v_k(t, \theta) \deff  e^{-t+J_k\theta}v^+_k  

6204: +v^0_k+  e^{t-J_k\theta}v^-_k 

6205: $$

6206: the norm $\norm{u -v_k}_{L^{1,2}(Z_k)}$ (computed with $g_k$) attains the minimum.

6207: 

6208: We claim that under the hypotheses of the theorem there exist a constant $C_1=

6209: C_1(J^*, s)$ and and an integer $k_0=k_0(J^*, s)$ such that for any integer 

6210: $k=k_0,\ldots, l-k_0$

6211: \begin{multline}\eqqno(5.2.11)

6212: \norm{u -v_k}^2_{L^{1,2}(Z_k)} +

6213: \norm{v_{k-1} -v_k}_{L^{1,2}(Z_k)}^2 +

6214: \norm{v_{k+1} -v_k}_{L^{1,2}(Z_k)}^2 \le

6215: \\

6216: \le \gamma_s \cdot\left(

6217:    \norm{u -v_{k-1}}^2_{L^{1,2}(Z_{k-1})} +

6218:    \norm{u -v_{k+1}}^2_{L^{1,2}(Z_{k+1})} \right)  +

6219: \\

6220: +C_1  \cdot \bigl(e^{-2(1+s)k} \norm{du}^2_{L^2(Z(0,2))} + 

6221: e^{-2(1+s)(l-k)}\norm{du}^2_{L^2(Z(l-2,l))}\bigr) \ 

6222: \end{multline}

6223: with the parameter $\gamma_s \deff {1\over \cosh(2+2s)}\cdot\;$ 

6224: Assuming the contrary, there must exist sequences of %involved objects

6225: \begin{itemize}

6226: \item integers  $l_\nu\lrar \infty$;

6227: \item integers $k_\nu \lrar \infty$ with $l_\nu -k_\nu \lrar \infty$;

6228: \item structures $J_\nu$ in $B$ with $\norm{J_\nu -J^*}_{C^{0,s}(B)} \le \epsi^*$;

6229: \item $J_\nu$-holomorphic maps $u_\nu: Z(0,l_\nu) \to B(\half)$ with 

6230: $\sup_{k=0,\ldots,l_\nu} \norm{du_\nu}_{L^2(Z_k)} \lrar 0$

6231: \end{itemize}

6232: with the following property. For the points $x_{\nu,k} \deff \bint_{Z_k} u_\nu$, 

6233: the linear complex structure $J_{\nu,k} \deff J_\nu(x_{\nu,k})$, the corresponding 

6234: metrics $g_{\nu,k}$, and vectors $v^+_{\nu,k}, v^0_{\nu,k}, v^-_{\nu,k} \in \rr

6235: ^{2n}$ constructed as above for every $u_\nu \ogran_{Z_k}$ with $k=1,\ldots,

6236: l_\nu$, at the position $k=k_\nu$ we obtain the inequality in the opposite 

6237: direction: 

6238: \begin{multline}\eqqno(5.2.12)

6239: \norm{u_\nu -v_{\nu,k_\nu}}^2_{L^{1,2}(Z_{k_\nu})} +

6240: \norm{v_{k_\nu-1} -v_{k_\nu}}^2_{L^{1,2}(Z_{k_\nu})} +

6241: \norm{v_{k_\nu+1} -v_{k_\nu}}^2_{L^{1,2}(Z_{k_\nu})} \ge

6242: \\

6243: \ge \gamma_s \cdot\left(

6244:    \norm{u_\nu -v_{\nu,k_\nu-1}}^2_{L^{1,2}(Z_{k_\nu-1})} +

6245:    \norm{u_\nu -v_{\nu,k_\nu+1}}^2_{L^{1,2}(Z_{k_\nu+1})} \right)  +

6246: \\

6247: +\nu  \cdot \bigl(e^{-2(1+s)k_\nu} \norm{du_\nu}^2_{L^2(Z(0,2))} + 

6248: e^{-2(1+s)(l-k_\nu)}\norm{du_\nu}^2_{L^2(Z(l_\nu-2,l_\nu))}\bigr). \ 

6249: \end{multline}

6250: 

6251: 

6252: \smallskip

6253: Let us estimate the behavior of $u_\nu -v_{\nu,k}$ in $Z_k$ for $k\approx k_\nu$.

6254: Set

6255: $$

6256: A_{\nu, k} \deff e^{-k} \norm{du_\nu}_{L^2(Z(0,2))} + 

6257: e^{-(l-k)}\norm{du_\nu}_{L^2(Z(l_\nu-2,l_\nu))}.

6258: $$

6259: Then by \lemma{lem5.2.2} we have $\norm{du_\nu}_{L^2(Z_k)} \le C\cdot A_{\nu,k}$. 

6260: This yields a similar estimate on the diameter: $\diam(u_\nu(Z_k)) \le C \cdot 

6261: A_{\nu,k}$, possibly with a new constant $C$. Further, for a linear complex 

6262: structure $J'$ with the corresponding operator $\dbar' \deff \dbar_{J'}$ we obtain

6263: the pointwise estimate

6264: \begin{multline*}%\eqqno(5.2.13)

6265: \left|\dbar' u_\nu \right| = 

6266: \left|\dbar' u_\nu - \dbar_{J_\nu} u_\nu \right| = 

6267: \left| (\d_x u_\nu - J' \cdot \d_y u_\nu) - 

6268:  (\d_x u_\nu - J_\nu(u_\nu) \cdot \d_yu_\nu)\right| 

6269: \\

6270: \le \left|J' - J_\nu \scirc u_\nu\right| \cdot \left|du_\nu \right|.

6271: \qquad

6272: \end{multline*}

6273: For $J_{\nu,k}$ this yields the estimate

6274: \begin{equation}\eqqno(5.2.14)

6275: \norm{\dbar_{J_{\nu,k}} u_\nu}_{L^2(Z(k-2,k+1))} \le 

6276: C \bigl(\diam(u_\nu(Z(k-2,k+1))) \bigr)^s \cdot 

6277: \norm{du_\nu}_{L^2(Z(k-2,k+1))} 

6278: \le C'\cdot A^{1+s}_{\nu,k}.

6279: \end{equation}

6280: 

6281: By construction, $J_{\nu,k}$ are uniformly bounded. This implies that we can 

6282: represent $u_\nu\ogran_{Z_k}$ in the form $u_\nu\ogran_{Z_k} = w_{\nu,k} + 

6283: f_{\nu,k}$, where $w_{\nu,k}$ is $J_{\nu,k}$-holomorphic and $f_{\nu,k}$ 

6284: is estimated as 

6285: $$

6286: \norm{f_{\nu,k}} _{L^{1,2}(Z_k)} \le C\cdot A^{1+s}_{\nu,k}.

6287: $$

6288: 

6289: \smallskip

6290: Define the positive $\eta_\nu$ by the relation 

6291: $$

6292: \eta_\nu^2= \norm{u_\nu -v_{\nu,k_\nu}}^2_{L^{1,2}(Z_{k_\nu})} +

6293: \norm{v_{k_\nu-1} -v_{k_\nu}}^2_{L^{1,2}(Z_{k_\nu})}  + 

6294: \norm{v_{k_\nu+1} -v_{k_\nu}}^2_{L^{1,2}(Z_{k_\nu})} 

6295: $$

6296: and set

6297: $$

6298: \ti u_\nu(t, \theta) \deff 

6299: \msmall{1\over \eta_\nu} u_\nu(t+k_\nu, \theta), 

6300: \qquad\qquad 

6301: \ti w_{\nu,k}(t, \theta) \deff 

6302: \msmall{1\over \eta_\nu} w_{\nu,k+k_\nu}(t+k_\nu, \theta),

6303: $$

6304: $$

6305: \ti f_{\nu,k}(t, \theta) 

6306: \deff \msmall{1\over \eta_\nu} f_{\nu,k+k_\nu}(t+k_\nu, \theta),

6307: \qquad

6308: \ti J_{\nu, k} \deff J_{\nu, k+k_\nu},

6309: \qquad 

6310: \ti v^\iota_{\nu,k} \deff 

6311: \msmall{1\over \eta_\nu} v^\iota_{\nu,k+k_\nu},\; \iota=+,0,-,

6312: $$

6313: In other words, we shift all the picture from $Z_{k_\nu}$ to $Z_0$ and

6314: rescale the maps $u_\nu$, the vectors $v^\iota_{\nu,k}$, $\iota=+,0,-$, and so on

6315: in a way as to make the left hand side of \eqqref(5.2.12) equal to $1$. 

6316: 

6317: It follows from \eqqref(5.2.12) that $A^{1+s}_{\nu, k+k_\nu} \le C \nu^{-\half} 

6318: \eta_\nu =o(\eta_\nu)$ for any fixed $k$. Consequently, $\norm{\ti f_{\nu, k}}

6319: _{L^{1,2} (Z_k)} \lrar 0$ for any fixed $k$ and $\nu\lrar \infty$. This implies 

6320: that the norms $\norm{\ti w_{\nu, k} - \ti v_{\nu, k}}_{L^{1,2}(Z_k)}$ remain 

6321: uniformly bounded in $\nu$ for any fixed $k$. 

6322: 

6323: Represent every $\ti w_{\nu, k}$ as the Laurent series 

6324: \begin{equation}\eqqno(5.2.14a)

6325: \textstyle

6326: \ti w_{\nu, k}(t, \theta)= \sum_{m=-\infty}^{+\infty}  

6327: e^{m(-t + J_{\nu, k}\theta)} w^m_{\nu, k}

6328: \end{equation}

6329: and denote by $\ti w'_{\nu, k}$ the sum of terms with $m=0, \pm1$, \ie 

6330: $$

6331: \ti w'_{\nu, k}(t, \theta)\deff 

6332: e^{t - J_{\nu, k}\theta} w^{-1}_{\nu, k} +  w^0_{\nu, k} +

6333: e^{-t + J_{\nu, k}\theta} w^1_{\nu, k}.

6334: $$

6335: It follows from the construction of $\ti v_{\nu,k}$ that 

6336: \begin{equation}\eqqno(5.2.15)

6337: \norm{\ti w'_{\nu, k} - \ti v_{\nu,k} }_{L^{1,2}(Z_k)}= 

6338: O(\norm{\ti f_{\nu, k}}_{L^{1,2}(Z_k)}) \lrar 0

6339: \end{equation}

6340: for any fixed $k$. Indeed, $\ti v_{\nu,k}$, considered as a function in $\theta$, 

6341: is a linear combination of a constant and the trigonometric functions $\cos\theta$

6342: and $\sin\theta$. So it is orthogonal to the remaining terms $e^{m(-t + J_{\nu, k}

6343: \theta)} w^m_{\nu, k}$, $|m|\ge2$. Thus $\ti w'_{\nu, k}$ is the best 

6344: approximation of $\ti w_{\nu, k}$ by such linear combinations, whereas the 

6345: difference $\ti w'_{\nu, k} - \ti v_{\nu,k}$ appears as the best approximation of 

6346: $\ti f_{\nu, k}$. 

6347: 

6348: Since $\ti J_{\nu, 0}$ is bounded uniformly in $\nu$, there exists a subsequence,

6349: still indexed by $\nu$, which converges to a linear complex structure $\ti J$.

6350: It follows from the definition of $\ti J_{\nu, k}$ and the estimate on the

6351: diameter of $u_\nu(Z_k)$ for $k\approx k_\nu$ that for any fixed $k$

6352: the structures $\ti J_{\nu, k}$ also converge to $\ti J$. 

6353: 

6354: Now we show that, after going to a subsequence, $\ti u_{\nu, 0} - \ti v_{\nu, 0}$ 

6355: converges weakly in the $L^{1,2}(Z(-2,1))$-topology to a $\ti J$-holomorphic 

6356: function, and that this convergence is strong in the $L^{1,2}(Z_0)$-topology. The 

6357: inequality \eqqref(5.2.12) together with the choice of $\eta_\nu$ and the

6358: construction of $\ti u_{\nu, k}$ gives boundedness of the norms $\norm{\ti u_{\nu,

6359: 0} - \ti v_{\nu, 0} }_{L^{1,2}(Z(-2,1))}$ uniform in $\nu$. So the weak 

6360: convergence follows. From \eqqref(5.2.14) and $A^{1+s}_{\nu,k} =o(\eta_\nu)$ we 

6361: obtain the vanishing $\norm{\dbar_{\ti J _{\nu, 0}} \ti u_{\nu, 0} }_{L^{1,2} 

6362:  (Z(-2,1))} \lrar0$. The estimate \eqqref(5.2.15) and $\norm{\ti f_{\nu, k}}

6363: _{L^{1,2} (Z_k)} \lrar 0$ yield $\norm{\dbar_{\ti J _{\nu, 0}} \ti v_{\nu, 0} }

6364: _{L^{1,2} (Z(-2,1))} \lrar0$. Now the desired strong $L^{1,2}(Z_0)$-convergence 

6365: follows from elliptic regularity of $\ti J _{\nu, 0}$. In the same way for $k= 

6366: \pm1$ we obtain the weak $L^{1,2}$-convergence of $\ti u_{\nu, k} -\ti v_{\nu, k}$

6367: in $Z_k$.  

6368: 

6369: For $k=0,\pm1$, let $\ti u_k \deff \lim \ti u_{\nu, k} -\ti v_{\nu, k}$ be the 

6370: limit functions obtained above. Since $\norm{\ti f_{\nu, k}} _{L^{1,2} (Z_k)} 

6371: \lrar 0$, these are $\ti J$-holomorphic functions in $Z_k$, $\ti J = \lim 

6372: \ti J_{\nu,k}$.

6373: 

6374: Observe that the functions $\ti v_{\nu,\pm1} -\ti v_{\nu,0}$ are linear 

6375: combinations of constants and the function $e^{\pm t}\cos\theta$, $e^{\pm t}\sin

6376: \theta$ which are uniformly bounded in the $L^{1,2}(Z_0)$-norm. Consequently,

6377: after taking a subsequence, we also obtain the strong $L^{1,2}$-convergence

6378: in $Z_0$. This implies the strong $L^{1,2}$-convergence in $Z(-2,1)$. Finally, from

6379: \eqqref(5.2.15) we conclude that the Laurent series for each $\ti u_k$ does not

6380: contain terms of degree $m=0,\pm1$, \ie  a constant term and a multiple of

6381: $e^{\pm(-t + \ti J \theta)}$. This, in turn, implies that, first, the $\ti u_k(t, 

6382: \theta)$ are restrictions to $Z_k$ of the same $\ti J$-holomorphic function 

6383: $\ti u$, and second, $\lim \ti v_{\nu,-1} -\ti v_{\nu,0} = \lim \ti v_{\nu,1} -

6384: \ti v_{\nu,0} =0$.

6385: 

6386: Substituting into \eqqref(5.2.12), we see that $\ti u$ satisfies the inequality

6387: \begin{equation}\eqqno(5.2.16)

6388: \norm{\ti u}^2_{L^{1,2}(Z_0)} 

6389: \ge \gamma_s \cdot\left(

6390:    \norm{\ti u}^2_{L^{1,2}(Z_{-1})} +

6391:    \norm{\ti u}^2_{L^{1,2}(Z_1)} \right).

6392: \end{equation}

6393: On the other hand, the absence of the terms of degree $m=0,\pm1$ in the Laurent 

6394: decomposition of type \eqqref(5.2.14a) for $\ti u$ implies the inequality

6395: \begin{equation}\eqqno(5.2.17)

6396: \norm{\ti u}^2_{L^{1,2}(Z_0)} 

6397: \le \gamma_2 \cdot\left(

6398:    \norm{\ti u}^2_{L^{1,2}(Z_{-1})} +

6399:    \norm{\ti u}^2_{L^{1,2}(Z_1)} \right).

6400: \end{equation}

6401: with $\gamma_2 \deff {1\over \cosh(4)}$. This inequality is easily obtained for 

6402: mutually orthogonal terms $\ti u^m e^{m(-t+\ti J \theta)}$. However, since $s<1$,

6403: $\gamma_2 < \gamma_s= {1\over \cosh(2 +2s)}$, which is a contradiction.

6404: 

6405: This implies the validity of \eqqref(5.2.11) for all $k= k_0,\ldots, l-k_0$

6406: with $k_0$ independent of $J$, $l$, and $u$.

6407: 

6408: \smallskip\noindent {\sl Step 2}. 

6409: We now turn back to the proof of the theorem. To show that \eqqref(5.2.11) 

6410: implies \eqqref(5.2.10), we set for $k=0, \ldots, l$

6411: \begin{multline}

6412: A'_k \deff k\cdot \msmall{\cosh(2+2s) \over \sinh(2+2s)} \cdot C_1 \cdot

6413: e^{-2(1+s)k} \norm{du}^2_{L^2(Z(0,2))}+ 

6414: \\ 

6415: + (l-k)\cdot \msmall{\cosh(2+2s) \over \sinh(2+2s)} \cdot C_1 \cdot

6416: e^{-2(1+s)(l-k)} \norm{du}^2_{L^2(Z(l-2,l))}.\ 

6417: \end{multline}

6418: Then $A'_k$ satisfies the equality

6419: $$

6420: A'_k = \msmall{\gamma_s \over 2} \cdot (A'_{k-1} + A'_{k+1}) 

6421: + C_1  \cdot \bigl(e^{-2(1+s)k} \norm{du}^2_{L^2(Z(0,2))} + 

6422: e^{-2(1+s)(l-k)}\norm{du}^2_{L^2(Z(l-2,l))}\bigr). 

6423: $$

6424: Consequently,

6425: \begin{multline}\eqqno(5.2.18)

6426: \norm{u -v_k}^2_{L^{1,2}(Z_k)} - A'_k +

6427: \norm{v_{k-1} -v_k}_{L^{1,2}(Z_k)}^2 +

6428: \norm{v_{k+1} -v_k}_{L^{1,2}(Z_k)}^2 \le

6429: \\

6430: \le \gamma_s \cdot\left(

6431:    \norm{u -v_{k-1}}^2_{L^{1,2}(Z_{k-1})} - A'_{k-1} +

6432:    \norm{u -v_{k+1}}^2_{L^{1,2}(Z_{k+1})} - A'_{k+1})\right).\quad

6433: \end{multline}

6434: As in the proof of \lemma{lem5.2.2}, \eqqref(5.2.18) implies the estimate

6435: \begin{multline}\eqqno(5.2.19)

6436: \norm{u -v_k}^2_{L^{1,2}(Z_k)} - A'_k \le 

6437: \\

6438: \le C_2  \cdot \bigl(e^{-2(1+s)k} \norm{du}^2_{L^2(Z(0,2))} + 

6439: e^{-2(1+s)(l-k)}\norm{du}^2_{L^2(Z(l-2,l))}\bigr)

6440: \end{multline}

6441: for all $k= k_0,\ldots, l-k_0$ with $k_0$ independent of $J$, $l$, and $u$.

6442: Substitution the definition of $A'_k$ yields

6443: \begin{multline}\eqqno(5.2.20)

6444: \norm{u -v_k}^2_{L^{1,2}(Z_k)} +

6445: \norm{v_{k-1} -v_k}_{L^{1,2}(Z_k)}^2 +

6446: \norm{v_{k+1} -v_k}_{L^{1,2}(Z_k)}^2 \le 

6447: \\

6448: \le C_3  \cdot \bigl(k\cdot e^{-2(1+s)k} \norm{du}^2_{L^2(Z(0,2))} + 

6449: (l-k) \cdot e^{-2(1+s)(l-k)}\norm{du}^2_{L^2(Z(l-2,l))}\bigr)

6450: \end{multline}

6451: 

6452: 

6453: \smallskip\noindent {\sl Step 3}. For concrete $J$, $l$, and $u$ as in the 

6454: hypotheses of the theorem, find $k^*$ for which the right hand side of 

6455: \eqqref(5.2.20) takes its minimum. Set $J\st \deff J_{k^*} = J(x_{k^*})$, 

6456: and $v^\iota \deff v^\iota_{k^*}$, $\iota=-,0,+$, and $v(t,\theta) \deff v_{k^*}(t,

6457: \theta)$. In view of \eqqref(5.2.20), for the proof of the theorem it is 

6458: sufficient to estimate $\norm{v_k - v}_{L^{1,2}(Z_k)}$.

6459: 

6460: We do this by descending recursion starting from $k=k^*$. Assume that we have 

6461: shown that 

6462: \begin{equation}\eqqno(5.2.21)

6463: \norm{v_k - v}_{L^{1,2}(Z_k)} \le 

6464: C_4 \, k^{\half}\, e^{-(1+s)k}\,\norm{du}_{L^2(Z(0,2))}

6465: \end{equation}

6466: for all $k=k_1+1,\ldots, k^*$ with the constant $C_4$ to be chosen below. 

6467: By our choice of $k^*$, for $k=1,\ldots, k_1$ we obtain from 

6468: \eqqref(5.2.20)

6469: $$

6470: \norm{v_k - v_{k+1}}_{L^{1,2}(Z_k)} \le 

6471: 2\,C_3\, k^{\half}\, e^{-(1+s)k}\,\norm{du}_{L^2(Z(0,2))}.

6472: $$

6473: Observe that for any function $w(t,\theta)$ of the form 

6474: $$

6475: w(t,\theta)= w^0 

6476: + (e^t w^+_c + e^{-t} w^-_c) \cos(\theta) 

6477: + ( e^t w^+_s + e^{-t} w^-_s) \sin(\theta) 

6478: $$

6479: with constant vectors $w^0, w^\pm_c, w^\pm_s \in \rr^{2n}$---so are all our 

6480: differences $v_k - v_{k'}$---we have the estimate

6481: $$

6482: \norm{w}_{L^{1,2}(Z_k)} \le e \cdot \norm{w}_{L^{1,2}(Z_{k+1})}.

6483: $$

6484: Applying this, we obtain

6485: \begin{align*}

6486: \norm{v_{k_1} - v}_{L^{1,2}(Z_{k_1})} 

6487: &\le \norm{v_{k_1} - v_{k_1+1}}_{L^{1,2}(Z_{k_1})} + 

6488: \norm{v_{k_1+1} -v }_{L^{1,2}(Z_{k_1})} 

6489: \\

6490: \le& 2\,C_3\, k_1^{\half}\, e^{-(1+s)k_1} \,\norm{du}_{L^2(Z(0,2))}+ 

6491: e \cdot \norm{v_{k_1+1} -v }_{L^{1,2}(Z_{k_1+1})}

6492: \\

6493: \le& 2\,C_3\, k_1^{\half}\, e^{-(1+s)k_1} \,\norm{du}_{L^2(Z(0,2))}+ 

6494: C_4\, (k_1+1)^{\half}\, e^{1-(1+s)(k_1+1)}\,\norm{du}_{L^2(Z(0,2))}

6495: \\

6496: =&2\,C_3\, k_1^{\half}\, e^{-(1+s)k_1} \,\norm{du}_{L^2(Z(0,2))}+ 

6497: e^{-s} \,C_4\, (k_1+1)^{\half}\, e^{-(1+s)k_1} \,\norm{du}_{L^2(Z(0,2))}.

6498: \end{align*}

6499: Assume additionally that $e^{-s/2}(k_1+1)^{\half} \le k_1^{\half}$. Then 

6500: setting $C_4 \deff {2C_3 \over 1- e^{-s/2}}$ we can conclude that \eqqref(5.2.21) 

6501: also holds for $k=k_1$. Since the condition $e^{-s/2}(k_1+1)^{\half} 

6502: \le \cdot k_1^{\half}$ is equivalent to $k_1 \ge { 1 \over e^s-1 }$, our recursive 

6503: construction implies \eqqref(5.2.21) for all $k\in \left[{ 1 \over e^s-1 }, k^*

6504: \right]$. For the remaining $k\in \left[1, { 1 \over e^s-1 }\right]$ the 

6505: estimate \eqqref(5.2.21) follows from \lemma{lem5.2.2}.

6506: 

6507: Making a similar recursive construction for $k=k^*, \ldots, l$

6508: we obtain the estimate 

6509: \begin{equation}\eqqno(5.2.22)

6510: \norm{v_k - v}_{L^{1,2}(Z_k)} \le 

6511: C_4 \, (l-k)^{\half}\, e^{-(1+s)(l-k)}\,\norm{du}_{L^2(Z(l-2,l))}

6512: \end{equation}

6513: for all $k=k^*,\ldots, l-1$ with the the same constant $C_4$. Now \eqqref(5.2.21),

6514: \eqqref(5.2.22), and \eqqref(5.2.20) imply the desired estimate 

6515: \eqqref(5.2.10). \qed 

6516: 

6517: 

6518: \newsubsection[5.3]{Deformation of a node and gluing}

6519: The cycle topology on $\barm$, introduced in \refsubsection{5.2c},

6520: has the nice property that $\pr_{\!\!\scrj}: \barm \to \scrj$ is continuous and 

6521: {\sl proper}. The last property is follows from the Gromov compactness for 

6522: closed curves. However, it is desirable to have a better understanding of the 

6523: topological structure of $\barm$. Recall that in \refsubsection{5.2c} we obtained

6524: a natural stratification of $\barm$ in which the strata are distinguished by a 

6525: topological type of curves. Moreover, every stratum $\scrm_\bftau$ has a natural 

6526: structure of a $C^\ell$-smooth Banach manifold such that the restricted

6527: projection $\pr_{\!\!\scrj}: \scrm_\bftau \to \scrj$ is Fredholm. 

6528: So to understand of the topology of $\barm$

6529: means to describe how different strata are attached to each other.

6530: The most important problem is to describe deformations of the standard node. 

6531: Let us formulate the question as follows:

6532: 

6533: \state Gluing problem. {\sl Let $J_0\in \scrj$ be an almost complex structure and 

6534: $u_0: \scra_0 \to X$ a $J_0$-holomorphic map. Describe possible $J$-holomorphic 

6535: maps $u: Z(0,l) \to X$ with $J\in \scrj$ sufficiently close to $J_0$ and $l\gg0$ 

6536: which are sufficiently close to $u_0$ \wrt the Gromov topology}. In other words, 

6537: we try to reverse the bubbling and construct a single map $u$ of a long cylinder 

6538: $Z(0,l)$ by gluing together the components $u'_0, u''_0: \Delta \to X$ of the map 

6539: $u_0$. 

6540: 

6541: \smallskip

6542: Moreover, one would like to have a smooth structure on the set of such deformation,

6543: so that the transversality techniques could be applied. This means that one seeks

6544: a family of deformations of a given $u_0:  \scra_0 \to (X, J_0)$ depending

6545: smoothly on the parameter.

6546: 

6547: \smallskip

6548: As the main result of this paragraph we give a satisfactory solution to

6549: the {\sl Gluing problem}. Let us start with introducing some notation.

6550: 

6551: \newdefi{def5.3.1} For a fixed sufficiently small $\epsi>0$, let

6552: \begin{equation}

6553: \scra \deff \{\, (z^+, z^-) \in \Delta^2\;:\;

6554: |z^+| \cdot |z^-| < \epsi \,\}

6555: \end{equation}

6556: with the projection

6557: \begin{equation}

6558: \pr_\scra : \scra \to \Delta(\epsi),

6559: \quad \pr_\scra(z^+, z^+) = \lambda(z^+, z^+) \deff z^+ \cdot z^- .

6560: \end{equation}

6561: Further, for $\lambda\in \Delta(\epsi)$ define the analytic sets

6562: \begin{equation}

6563: \scra_\lambda \deff \{\, (z^+, z^-) \in \Delta^2\;:\;

6564: z^+ \cdot z^- = \lambda\,\} = \pr_\scra \inv(\lambda).

6565: \end{equation}

6566: For $\lambda=0$ this is the standard node and for $\lambda \not=0$ a cylinder of 

6567: conformal radius $R= \log{1\over|\lambda|}$. Define 

6568: $$

6569: \scra^\pm_\lambda \deff

6570: \{(z^+,z^-) \in \scra_\lambda: |\lambda| \le z^\pm <1 \}.

6571: $$

6572: Then $\scra^\pm_\lambda$ are subannuli for $\lambda \not=0$ and $\scra^\pm_0$ are

6573: discs $\Delta^\pm$, the irreducible components of $\scra_0$. In any case, $\scra

6574: _\lambda = \scra^+_\lambda \cup \scra^-_\lambda$. 

6575: 

6576: To describe a Hermitian metric on a complex manifold $X$, it is sufficient to 

6577: indicate only the corresponding K\"ahler form $\omega$. In this case we shall say

6578: that $\omega$ {\sl induces} a metric on $X$ or even that $\omega$ {\sl is} a 

6579: metric on $X$. The author begs the reader's pardon for such informality in the

6580: terminology. The advantage of such notation is that the restriction of a metric on 

6581: a complex 

6582: submanifold is given by the restriction of the corresponding K\"ahler form.

6583: In this notation, the standard metric on the disc $\Delta$ with the coordinate

6584: $z$ is given by the form ${\isl \over 2}dz \wedge d\bar z$. 

6585: 

6586: We equip $\scra_\lambda$ with the Riemannian metric induced from $\Delta^2$. This

6587: gives the standard metric ${\isl \over 2}dz^\pm \wedge d\bar z^\pm$ on each 

6588: component $\Delta^\pm$ of $\scra_0$ and the {\sl hyperbola metric} 

6589: \begin{equation}\eqqno(5.3.0.1)

6590: {\isl \over 2}\Bigl(1 + {|\lambda|^2 

6591: \over |z^+|^4}\Bigr) dz^+ \wedge d\bar z^+ = 

6592: {\isl \over 2}\Bigl(1 + {|\lambda|^2 

6593: \over |z^-|^4}\Bigr) dz^- \wedge d\bar z^-

6594: \end{equation}

6595: on $\scra_\lambda$ with $\lambda \not=0$. 

6596: 

6597: Set

6598: \begin{equation*}

6599: \check\scra \deff  \{\, (z^+, z^-) \in \scra\;:\;

6600: z^+ \cdot z^- \not=0 \,\} = 

6601: \sqcup_{\lambda \in \check\Delta(\epsi)} \scra_\lambda 

6602: \end{equation*}

6603: and

6604: \begin{equation*}

6605: V^\pm \deff \{ 1-\epsi < |z^\pm | < 1 \}, \qquad V \deff V^+ \sqcup V^-.

6606: \end{equation*}

6607: For a given $\lambda$ we have the canonical imbedding $V \to \scra_\lambda$,

6608: defined by the coordinate functions $z^\pm$ on $\scra_\lambda$ and on $V^\pm$.

6609: This imbedding defines the restriction map

6610: \begin{equation*}

6611: u \in L^{1,p}(\scra_\lambda, X) \mapsto u\ogran_V \in L^{1,p}( V, X)

6612: \qquad

6613: u \mapsto (u(z^+)\ogran_{V^+}, \;  u(z^-)\ogran_{V^-}).

6614: \end{equation*}

6615: \end{defi}

6616: 

6617: \newdefi{def5.3.2} For a nodal curve $C$ with smooth boundary $\d C=

6618: \sqcup_i \gamma_i$, $\gamma_i \cong S^1$, let $\scrp(C)$ be the set of

6619: {\sl stable pseudoholomorphic maps between $C$ and $X$},

6620: \begin{equation}

6621: \scrp(C) \deff \{\, (u, J) \in L^{1,p}(C, X) \times \scrj

6622: \;:\; \dbar_J u=0,\; \text{ $u$ is stable}\;\}.

6623: \end{equation}

6624: Equip $\scrp(C)$ with the topology induced from $L^{1,p}(C, X) \times \scrj$.

6625: In particular, $\scrp(V)$ consists of triples $(u^+, u^-, J)$, where

6626: $u^\pm: V^\pm \to X$ is $L^{1,p}$-smooth $J$-holomorphic map.

6627: Denote by $\scrp^*(C)$ the subset of $(u,J) \in \scrp(C)$ for which

6628: $u$ is {\sl non-multiple on the union of compact components of $C$}.

6629: Further, we define

6630: \begin{equation}

6631: \scrp(\scra) \deff \sqcup_{\lambda \in \Delta(\epsi)}

6632: \scrp(\scra_\lambda) ,

6633: \qquad

6634: \scrp(\check\scra) \deff \sqcup_{\lambda \in \check\Delta(\epsi)}

6635: \scrp(\scra_\lambda)

6636: \end{equation}

6637: and equip this spaces with the topology induced by the Gromov convergence in the 

6638: interior of $\scra_\lambda$ and $L^{1,p}$-convergence near boundary. This means 

6639: that $(u_n, J_n, \lambda_n)$ converges to $(u_\infty, J_\infty, \lambda_\infty)$ 

6640: if $(J_n, \lambda_n)$ converges to $(J_\infty,  \lambda_\infty)$ in $\scrj \times

6641: \Delta (\epsi)$, the restrictions $u_n\ogran_V$ converges to $u_\infty \ogran_V$

6642: \wrt $L^{1,p}$-norm, and $u_n$ converges to $u_\infty$ in the sense of {\sl

6643: Definition 5.2.7}. Elements of $\scrp(\scra)$ will be denoted by $(u, J,

6644: \lambda)$. As usual, $\pr_{\!\!\scrj}$ stands for the natural projections from

6645: $\scrp(C)$ or $\scrp(\scra)$ to $\scrj$.

6646: \end{defi}

6647: 

6648: 

6649: \newthm{thm5.3.1} The natural map $\pr_V: \scrp(\scra) \to \scrp(V)

6650: \times \Delta(\epsi)$, defined by

6651: \begin{equation}

6652: \pr_V(u, J, \lambda) \deff (u(z^+)\ogran_{V^+}, u(z^-)\ogran_{V^-}, J;

6653: \lambda)

6654: \end{equation}

6655: is an imbedding of a {\sl topological} Banach submanifold. 

6656: 

6657: Moreover, for every $(u_0,J_0) \in \scrp(\scra_0)$ there exists a neighborhood 

6658: $\scru \subset \scrp(\scra_0)$ of $(u_0,J_0)$, an $\epsi'>0$, and a map

6659: $\Phi: \scru \times \Delta(\epsi') \to \scrp(\scra)$ such that

6660: \begin{itemize}

6661: \item $\Phi$ is a homeomorphism onto the image;

6662: \item for every $\lambda \in \Delta(\epsi')$ the restricted map $\Phi_\lambda 

6663: \deff \Phi\ogran_{\scru \times \{\lambda\}}: \scru \to \scrp(V) \times \Delta(

6664: \epsi)$takes values in $\scrp(\scra_\lambda) \subset \scrp(\scra)$ and is a 

6665: $C^1$-diffeomorphism;

6666: \item the family of maps $\Psi_\lambda \deff \pr_V \scirc \Phi_\lambda: \scru \to 

6667: \scrp(V)$ depends continuously on $\lambda \in \Delta(\epsi')$ \wrt the 

6668: $C^1$-topology. 

6669: \end{itemize} 

6670: \end{thm}

6671: 

6672: In other words, the theorem statess that $\scrp(\scra) =\sqcup \scrp(\scra_\lambda)$

6673: is a continuous family of $C^1$-submanifolds. Before proving this result, we state 

6674: and prove a corollary which provides a technique which allows one to smooth nodal 

6675: points on pseudoholomorphic curves.

6676: 

6677: \newthm{thm5.3.1a} Let $C^*$ be a closed connected nodal curve parameterized by

6678: a real surface $S$, $J^*\in \scrj$, $u^*: C^* \to X$ a $J^*$-holomorphic map, and 

6679: $(C^*,u^*,J^*) \in \barm^{Gr}=\barm^{Gr}(S, X, [C^*])$ the corresponding

6680: element of the Gromov compactification of the total moduli space. Assume

6681: that the map $u^*:C^* \to X$ is non-multiple.

6682: 

6683: Then there exist $(C', u', J') \in \scrm(S, X, [C^*])$ arbitrarily close to

6684: $(C^*,u^*,J^*)$ \wrt the Gromov topology such that $C'$ is a smooth curve.

6685: \end{thm}

6686: 

6687: The notation used in this theorem  was introduced in {\sl Definitions 

6688: \ref{def5.2.6}} and {\sl\ref{def5.2c.2}}. Note that the condition of 

6689: non-multiplicity of $u^*:C^* \to X$ is equivalent to the absence of ghost and

6690: multiple components. 

6691: 

6692: \proof Let $\{ z^*_1, \ldots z^*_k\}$ be the set of nodal points of $C^*$.

6693: For every $z^*_i$ fix a neighborhood $V_i$ isomorphic to the standard node.

6694: We may also assume that the sets $V_i$ are pairwise disjoint. 

6695: Let $u^*_i$ denote the restriction of $u^*$ to $V_i$. Applying \refthm{thm5.3.1} 

6696: we can perturb $V_i\cong \scra_0$ to an annulus $V'_i = \scra_{\lambda_i}$ and 

6697: $u^*_i: V_i \to X$ to a $J^*$-holomorphic map $u'_i: V'_i \to X$. If these

6698: perturbations $(V'_i, u'_i)$ are made small enough, then we can adjust the

6699: structure $J^*$ and the map $u^*$ on the remaining part of the curve $C^*$ in a way

6700: yielding the desired $(C', u', J') \in \scrm(S, X, [C^*])$. \qed

6701: 

6702: 

6703: Modifying the proof of \refthm{thm5.3.1a} one can also obtain 

6704: 

6705: \newprop{prop5.3.1b} Let $C^*$ be a closed connected nodal curve parameterized by

6706: a real surface $S$, $J^*\in \scrj$, $u^*: C^* \to X$ a $J^*$-holomorphic map, and 

6707: $(C^*,u^*,J^*) \in \barm^{Gr}=\barm^{Gr}(S, X, [C^*])$ the corresponding

6708: element of the Gromov compactification of the total moduli space. Assume

6709: that the map $u^*:C^* \to X$ is non-multiple. 

6710: 

6711: Then in a neighborhood of $(C^*,u^*,J^*)$ the space $\barm^{Gr}$ is a topological

6712: Banach manifold and  the natural projection $\pr^{Gr}:\barm^{Gr} \to \barm$ 

6713: a homeomorphism. 

6714: \end{prop}

6715: 

6716: 

6717: Since the result of \propo{prop5.3.1b} is not needed for the 

6718: purposes of this paper, we leave it without a proof. 

6719: 

6720: 

6721: \medskip

6722: The proof of \refthm{thm5.3.1} is divided in the subsequent lemmas. The first two 

6723: are simple but useful technical results.

6724: 

6725: \newlemma{lem5.3.2} Let $C$ be a nodal curve without closed compact components,

6726: and $E$ a holomorphic vector bundle over $C$. Then any operator $D: L^{1,p}(C,

6727: E) \to L^p_{(0,1)}(C, E)$ of the form $D= \dbar_E + R$ with $R \in L^p$ is

6728: surjective and its kernel $\sfh^0_D(C, E)$ admits a closed complement.

6729: \end{lem}

6730: 

6731: \proof Imbed $C$ into a compact nodal curve $\wt C$ and extend $E$ to

6732: a holomorphic vector bundle $\wt E$ over $\wt C$. Without loss of generality

6733: we may assume that the Chern numbers $\la c_1(\wt E), \wt C_i\ra$ are

6734: sufficiently large for each component $\wt C_i$ of $\wt C$. Now extend

6735: $R \in L^p(C, \homr(E, E\otimes \Lambda^{(0,1)}))$ to $\wt R \in L^p(\wt C,

6736: \homr(\wt E, \wt E\otimes \Lambda^{(0,1)}))$ and set $\wt D \deff \dbar

6737: _{\wt E} + \wt R$. Adjusting $\wt R$ on the complement $\wt C \bs C$ we may

6738: assume that $\wt D: L^{1,p}(\wt C, \wt E) \to L^p_{(0,1)}(\wt C, \wt E)$

6739: is surjective. The sufficient condition for existence of such an adjustment

6740: is provided by the condition on the Chern numbers of $\wt E$. Since any $\eta

6741: \in L^p_{(0,1)}(C, E)$ extends to $\wt \eta \in L^p_{(0,1)}(\wt C, \wt E)$,

6742: the surjectivity of $\wt D$ implies the surjectivity of $D$.

6743: 

6744: The existence of a closed complement to the kernel of $D$ is shown in 

6745: \cite{Iv-Sh-2} in the case when $D= \dbar$. This proof applies also in our case 

6746: with only minor adjustments.

6747: %with necessary minor corrections. \ 

6748: \qed

6749: 

6750: \state Remark. The existence of a closed complement to the kernel of 

6751: $D$ allows us 

6752: %ensures possibility 

6753: to apply the implicit function theorem.

6754: 

6755: \smallskip 

6756: \newlemma{lem5.3.3}

6757: \sli The space $L^{1,p}(C, X)$ is a smooth Banach

6758: manifold with tangent space

6759: \begin{equation}

6760: T_uL^{1,p}(C, X)=L^{1,p}(C, E_u).

6761: \end{equation}

6762: 

6763: \slii The space $\scrp^*(C)$ is a $C^\ell$-smooth

6764: submanifold of $L^{1,p}(C, X) \times \scrj$ with tangent space

6765: \begin{equation}

6766: T_{(u,J)}\scrp^*(C)= \{ (v, \dot J)\in L^{1,p}(C, E_u) \times T_J\scrj\,:\,

6767: D_{u,J}v + \dot J \scirc du \scirc J_C =0 \}.

6768: \end{equation}

6769: 

6770: \sliii $\pr_V: \scrp(\scra_0) \to \scrp(V)$ and $\pr_V: \scrp(\check\scra)

6771: \to \scrp(V) \times \check\Delta(\epsi)$ are $C^\ell$-smooth imbeddings on

6772: Banach submanifolds.

6773: \end{lem}

6774: 

6775: The definitions of the spaces which are involved here are given in 

6776: \refdefi{def5.2.3a} and \refdefi{def5.3.2}.

6777: 

6778: \proof Let $C= \cup C_i$ be the decomposition of $C$ into components

6779: and $\{(z'_a, z''_a)\}$ the set of pairs of points on the normalization

6780: $\hat C$ corresponding to the nodal points.

6781: 

6782: \sli The space $L^{1,p}(C, X)$ is a subset of a smooth Banach

6783: manifold $\prod_i L^{1,p}(C_i, X)$ defined by equations $u(z'_a) =u(z''_a)$.

6784: One checks the transversality condition and computes the tangent space.

6785: 

6786: \slii One can use the same arguments as in part \slip.

6787: 

6788: \sliii First we note that \lemma{lem1.2.4} \slii implies the following unique 

6789: continuation property: Any

6790: $J$-holomorphic map $u$, defined on an open set $U$ of a nodal curve $C$ admits

6791: at most one $J$-holomorphic extension to an irreducible component $C'$ of $C$ 

6792: provided $C'\cap U \not =\emptyset$. This shows that the restriction maps $F_0: 

6793: \scrp(\scra_0) \to \scrp(V)$ and $\check F: \scrp(\check\scra) \to \scrp(V)\times

6794: \check\Delta(\epsi)$ are set-theoretically injective. Note that we have

6795: introduced new notation, $F_0$ and $\check F$, for the restrictions of

6796: the map $\pr_V$ to the corresponding definition domains.

6797: 

6798: Further,  $F_0: \scrp(\scra_0) \to \scrp(V)$ is obviously $C^\ell$-smooth

6799: and the differential $dF_0: T_u\scrp(\scra_0) \to T_u\scrp(V)$ is simply

6800: the restriction map $dF_0: (v, \dot J) \mapsto (v\ogran_V, \dot J)$.

6801: \lemma{lem5.3.2} shows that the restriction map

6802: \begin{equation}

6803: \{\, v \in L^{1,p}(\scra_0, E_u)\;:\; D_{u,J}v=0\, \}\mapsto

6804: \{\, v \in L^{1,p}(V, E_u)\;:\; D_{u,J}v=0\, \}

6805: \eqqno(5.3.2)

6806: \end{equation}

6807: is a closed imbedding and splits, \ie admits a closed complement.

6808: 

6809: \smallskip

6810: The claim about $\check F: \scrp(\check\scra) \to \scrp(V) \times

6811: \Delta(\epsi)$ is proven in a similar way. Details are left to the reader.

6812: \qed

6813: 

6814: \smallskip

6815: A crucial point in the proof of \refthm{thm5.3.1} is to find an apriori estimate

6816: for the operator $D_{u,J,\lambda}$ which is uniform as $\lambda\lrar0$. Because of

6817: local nature of the estimate it is sufficient to work with the ball $B\subset 

6818: \rr^{2n} \cong \cc^n$ equipped with the standard complex structure $J\st$. We 

6819: start with introducing a chart for the space $L^{1,p}(\scra, \cc^n) \deff \sqcup

6820: _{\lambda \in \Delta(\epsi)} L^{1,p}(\scra_\lambda, \cc^n)$.

6821: 

6822: \newdefi{def5.3.3} For a nodal complex curve $C$ and a complex manifold $X$ let

6823: $\scrh(C, X)$ denote the space of holomorphic maps $f: C \to X$ which are

6824: $L^{1,p}$-smooth up to the boundary $\d C$. In the case $X=\cc$ we abbreviate

6825: the notation to $\scrh(C)$.

6826: \end{defi}

6827: 

6828: \newlemma{lem5.3.3a}%

6829: \footnote{\.The results presented in the lemma have been obtained jointly with

6830: S.~Ivashkovich.} There exist families of homomorphisms $T_\lambda: L^p_{(0,1)} 

6831:  (\scra_\lambda{,} \cc)

6832: \to L^{1,p}(\scra_\lambda, \cc)$ and isomorphisms $\sfl_\lambda: \scrh(\scra_0) 

6833: \to \scrh(\scra_\lambda)$ and $Q_\lambda: L^p_{(0,1)} (\scra_0, \cc) \to 

6834: L^p_{(0,1)} (\scra_\lambda, \allowbreak

6835: \cc)$ with the following properties:

6836: 

6837: \sli the homomorphisms $T_\lambda$ are right inverses of $\dbar: L^{1,p}(\scra

6838: _\lambda, \cc) \to L^p_{(0,1)} (\scra_\lambda, \cc)$;

6839: 

6840: \slii the norms of $T_\lambda$, $\sfl_\lambda$, and $Q_\lambda$, as well as  

6841: $\sfl_\lambda\inv$ and $Q_\lambda\inv$, are uniformly bounded;

6842: 

6843: \sliii the homomorphisms $T_\lambda$, $\sfl_\lambda$, and $Q_\lambda$ depend

6844: smoothly on $\lambda\not=0$. 

6845: \end{lem}

6846: 

6847: \proof 

6848: By the definition, the hyperbola metric ${\isl \over 2}\left(1 + {|\lambda|^2 \over

6849: |z^+|^4} \right) dz^+ \wedge d\bar z^+$ on $\scra_\lambda$ is the sum of the 

6850: standard flat metrics ${\isl \over 2}dz^+ \wedge d\bar z^+$ and ${\isl \over 2}dz^-

6851: \wedge d\bar z^-$. Note also that the function $\left(1 + {|\lambda|^2 \over |z^+|

6852: ^4}\right)$, restricted to the subannulus $\scra^+_\lambda= \{ |\lambda|^\half <

6853: |z^+| <1\} \subset \scra_\lambda$, takes values in the interval $[1, 2]$. This 

6854: implies that in every subannulus $\scra^\pm_\lambda$ the metric ${\isl \over 2}

6855: \left(1 + {|\lambda|^2 \over |z|^4}\right) dz^+ \wedge d\bar z^+$  is equivalent to

6856: the disc metric ${\isl \over 2}dz^\pm \wedge d\bar z^\pm$. In particular, the norm 

6857: $\norm{v}_{L^{1,p}(\scra_\lambda)}$ is equivalent to the norm

6858: \begin{equation}

6859: \left(\int_{|\lambda|^ < |z^+|^2 <1} (|v| + |dv|)^p\;

6860: \msmall{\isl \over 2}dz^+ \wedge d\bar z^+

6861: \,+\,

6862: \int_{|\lambda|^ < |z^-|^2 <1} (|v| + |dv|)^p\;

6863: \msmall{\isl \over 2}dz^- \wedge d\bar z^-

6864: \right)^{1\over p}.

6865: \end{equation}

6866: 

6867: We start with construction of $T_\lambda$. For the discs $\Delta^\pm$ with the 

6868: coordinates $z^\pm$ respectively we define $\ti T^\pm: L^p_{(0,1)} (\Delta^\pm,

6869: \cc) \to L^{1,p}(\Delta^\pm, \cc)$ to be the Cauchy-Green operators, \ie 

6870: $$

6871: \ti T^+(\phi^+)(z^+) \deff 

6872: \msmall{1\over 2\pi \isl} \int_{\zeta \in \Delta} 

6873: \msmall{d\zeta \wedge \phi^+(\zeta) \over \zeta -z^+}

6874: $$

6875: and similarly for $\ti T^-$. Then we set

6876: $$

6877: T^\pm(\phi^\pm)(z^\pm) \deff \ti T^\pm(\phi^\pm)(z^\pm) - 

6878: \ti T^\pm(\phi^\pm)(0)    

6879: $$

6880: So $T^\pm$ are normalizations of $\ti T^\pm$ respectively to the condition 

6881: $T^\pm(\phi^\pm)(0) =0$. For a form $\phi \in L^p_{(0,1)} (\scra_\lambda, \cc)$ we 

6882: denote by $\phi^\pm (z^\pm)$ its restriction to $\scra^\pm \subset \Delta^\pm$ 

6883: extended by $0$ to the whole discs $\Delta^\pm$, and set  

6884: $$

6885: T_\lambda(\phi) \deff T^+(\phi^+)(z^+) + T^-(\phi^-)(z^-) 

6886: $$

6887: For the special case $\lambda=0$ this construction should be modified follows:

6888: Every form $\phi \in L^p_{(0,1)} (\scra_0, \cc)$ has two components $\phi^\pm

6889:  (z^\pm)$ corresponding to the decomposition 

6890: $$

6891: L^p_{(0,1)} (\scra_0, \cc)= L^p_{(0,1)}  (\Delta^+, \cc) \oplus 

6892: L^p_{(0,1)} (\Delta^-, \cc),

6893: $$

6894: and the operator $T_0$ transforms $\phi^\pm$ into functions $f^\pm(z^\pm)\deff

6895: T^\pm_0(\phi^\pm)(z^\pm)$ which satisfy $f^+(0)= f^-(0) =0$. Then $f\deff (f^+,f^-)

6896: \in L^{1,p}(\scra_0, \cc)$. The desired properties of $T_\lambda$ can be seen 

6897: in a straightforward way.

6898: 

6899: \medskip

6900: Defining the operators $\sfl_\lambda$, we recall the identification

6901: $$

6902: \scrh(\scra_0) =  \{ (g^+(z^+), g^-(z^-)) \in 

6903: \scrh(\Delta^+) \oplus \scrh(\Delta^-) : g^+(0) =g^-(0) \}.

6904: $$

6905: We set $\sfl_0 = \id: \scrh(\scra_0) \to \scrh(\scra_0)$ and 

6906: $$

6907: \sfl_\lambda: g=(g^+(z^+), g^-(z^-)) \in \scrh(\scra_0) \mapsto 

6908: g^+(z^+) + g^-(z^-) - g(0)

6909: $$

6910: for $\lambda \not=0$. The desired properties of $\sfl_\lambda$ are obvious as well.

6911: 

6912: Note also that for $\lambda \not=0$ the inverse operator $\sfl_\lambda\inv$

6913: essentially gives the Laurent decomposition of functions $g \in \scrh(\scra

6914: _\lambda)$.

6915: 

6916: 

6917: \medskip

6918: The definition of $Q_\lambda$ is more subtle. For $\lambda \not=0 \in 

6919: \Delta(\epsi)$ we set 

6920: \begin{equation}\eqqno(5.3.a1)

6921: \rho_\lambda(r) \deff 

6922: \msmall{ r^2 - {|\lambda|^2 \over r^2} \over 1-|\lambda|^2}\cdot

6923: \end{equation}

6924: Then every $\rho_\lambda$ induces a diffeomorphism of the interval $[|\lambda|,1]$ 

6925: onto  $[-1,1]$, such that $[|\lambda|, |\lambda|^\half]$ and $[|\lambda|^\half,1]$

6926: are mapped onto the intervals $[-1,0]$ and $[0,1]$ respectively. The inverse map 

6927: is given by 

6928: \begin{equation}\eqqno(5.3.a2)

6929: r= R_\lambda(\rho) = \sqrt{

6930: \msmall{ \rho(1-|\lambda|^2) + \sqrt{ \rho^2(1-|\lambda|^2)^2 +4\,|\lambda|^2}

6931:           \over2}} \cdot

6932: \end{equation}

6933: For $\lambda \not=0\in \Delta(\epsi)$ we define the maps $\sigma^\pm_\lambda: 

6934: Z(-1,1) = [-1,1] \times S^1 \to \scra_\lambda$ which are given in the coordinates

6935: $z^\pm= r^\pm \, e^{\isl\theta^\pm}$ by relations $r^\pm = R_\lambda(\rho)$ and 

6936: $\theta^\pm =\theta$ respectively, so that

6937: $$

6938: \sigma^\pm_\lambda: (\rho, \theta)\in Z(-1,1)  \mapsto 

6939: z^\pm= R_\lambda(\rho) e^{\isl\theta} \in \scra_\lambda.

6940: $$

6941: Now we can explain the reason for the choice of \eqqref(5.3.a1), which at first 

6942: glance probably seems rather wild. %at first look seems probably rather wild. 

6943: Our choice is made to insure that the pull-back 

6944: by $\sigma_\lambda$ of the natural volume form of $\scra_\lambda$ is a constant 

6945: multiple of the standard volume form on $Z(-1,1)$, \ie

6946: \begin{equation}\eqqno(5.3.a3)

6947: \sigma_\lambda^*\left( \left(1+ {|\lambda|^2 \over (r^+)^4}  \right)

6948: \msmall{\isl\over2} dz^+ \wedge d\bar z^+ \right)

6949: = (1-|\lambda|^2) d\rho \wedge d\theta .

6950: \end{equation}

6951: Moreover, $1-|\lambda|^2$ remains uniformly bounded as $\lambda$ varies

6952: in $\Delta(\epsi)$, so that the volume forms in the right hand side of 

6953: \eqqref(5.3.a3) are equivalent. 

6954: 

6955: The behavior of $\sigma^\pm_\lambda$ by $\lambda$ close to $0$, which is rather 

6956: delicate, can be described as follows. Denote $Z^+ \deff Z(0,1)$ and $Z^- \deff 

6957: Z(-1,0)$. Then for $\lambda \lrar 0$ we have convergence of $\sigma^+_\lambda$ on 

6958: $Z^+$ to a map $\sigma^+_0: Z^+ \to \scra^+_0 =\Delta^+$, and resp.\ convergence 

6959: of $\sigma^-_\lambda$ on $Z^-$ to $\sigma^-_0: Z^- \to \scra^-_0$, which are given

6960: by 

6961: \begin{align*}

6962: \sigma^+_0: &(\rho, \theta) \mapsto z^+=\sqrt\rho \, e^{\isl\theta},

6963: \\

6964: \sigma^-_0: &(\rho, \theta) \mapsto z^-=\sqrt\rho \, e^{\isl\theta}.

6965: \end{align*}

6966: Observe also that we obtain the map $R_0(\rho) = \sqrt\rho$ in the limit of

6967: \eqqref(5.3.a2) as $\lambda\lrar0$. The convergence $\sigma^\pm_\lambda \lrar

6968: \sigma^\pm_0$ is in the $C^\infty$-sense in the interiors of $Z^\pm$, and in

6969: the $C^0$-topology up to boundary of $Z^\pm$.

6970: 

6971: On the other hand, there is no convergence of $\sigma^\pm_\lambda$ on $Z^\mp$. 

6972: The topological reason for the absence of the convergence of $\sigma^\pm_\lambda

6973: \ogran_{Z^\mp}$ is that, making with $\lambda$ a small bypass around $0$, we 

6974: perform a {\sl Dehn twist} with $\scra_\lambda$.\footnote{\.The author is

6975: indebted to Bernd Siebert for this remark.}

6976: 

6977: Now, we define $Q_\lambda$ representing the components $\phi^\pm$ of every $\phi\in

6978: L^p_{(0,1)} (\scra_\lambda, \cc) = L^p_{(0,1)} (\Delta^+, \cc) \oplus L^p_{(0,1)} 

6979:  (\Delta^-, \cc)$ in the form $\phi^\pm(z^\pm) = f^\pm(r^\pm, \theta^\pm) d\bar

6980: z^\pm$ with $L^p$-functions $f^\pm(r^\pm, \allowbreak

6981: \theta^\pm)$ and setting 

6982: \par\noindent

6983: $

6984: Q_\lambda (\phi) \deff \begin{cases}

6985: f^+\Bigl(\! R_\lambda\bigl( (r^+)^2\bigr), \theta^+\Bigr)d\bar z^+ 

6986: & \!\!\!\!\text{at the point $z^+= r^+ e^{\isl\theta^+}$ with 

6987: $r^+ \in [|\lambda|^\half, 1]$};

6988: \\

6989: f^-\Bigl(\! R_\lambda\big( (r^-)^2 \bigr), \theta^-\Bigr)d\bar z^-

6990: & \!\!\!\!\text{at the point $z^-= r^- e^{\isl\theta^-}$ with 

6991: $r^- \in [|\lambda|^\half, 1]$}. 

6992: \end{cases}$ \lineeqqno(5.3.a4)

6993: 

6994: \smallskip

6995: Let us explain the meaning of the construction of $Q_\lambda$ given in

6996: \eqqref(5.3.a4). The first point is that we essentially transform every form

6997: $\phi \in L^p_{(0,1)}(\scra_\lambda, \cc)$ into a function $f\in L^p(\scra

6998: _\lambda, \cc)$. This is done by representing $\phi$ in the form $\phi(z^+) = f^+

6999:  (z^+) d\bar z^+$ on $\scra^+_\lambda$ and in the form $\phi(z^-) = f^-(z^-) 

7000: d\bar z^-$ on $\scra^-_\lambda$. Observe that we use different coordinates $z^\pm$

7001: on the different parts $\scra^\pm_\lambda$ of $\scra_\lambda$. Independently of

7002: this, we obtain a well-defined $L^p$-function $f$ on $\scra_\lambda$, and hence a

7003: family of well-defined maps $F_\lambda: L^p_{(0,1)}(\scra_\lambda, \cc) \to L^p(

7004: \scra_\lambda, \cc)$. It is not difficult to see that the $F_\lambda$ are complex 

7005: linear isomorphisms and that the operator norms $\norm{F_\lambda}_{\sf{op}}$

7006: and $\norm{F_\lambda\inv}_{\sf{op}}$ are bounded uniformly in $\lambda$.

7007: 

7008: The second point is the observation that for the definition of a space $L^p(Y)$ 

7009: only the involved measure $\mu$ on $Y$ is essential. In particular, if a measurable

7010: map $g: (Y_1, \mu_1) \to (Y_2, \mu_2)$ induces an equivalence of measures, \ie 

7011: $g^*\mu_2 = e^\rho(y) \mu_1$ for some bounded $\rho \in L^\infty(Y_1, \mu_1)$,

7012: then the induced map $g^*: L^p(Y_2, \mu_2) \to  L^p(Y_1, \mu_1)$ is an isomorphism

7013: of Banach spaces. Thus our construction exploits the fact that the measures

7014: $$

7015: (\sigma^\pm_\lambda)^* \left( \left(1+ {|\lambda|^2 \over (r^\pm)^4}  \right)

7016: \msmall{\isl\over2} dz^\pm \wedge d\bar z^\pm \right)

7017: $$

7018: on $Z^\pm$ are equivalent.

7019: 

7020: \newcorol{cor5.3.3b} The Banach spaces $L^{1,p}(\scra_0, \cc^n)$ and $L^{1,p}

7021:  (\scra_\lambda, \cc^n)$ with $\lambda \not=0$ are isomorphic.

7022: \end{corol}

7023: 

7024: \proof One uses $T_\lambda$ to split the exact sequences

7025: $$

7026: 0 \lrar \scrh(\scra_\lambda, \cc^n) \lrar L^{1,p}(\scra_\lambda, \cc^n)

7027: \buildrel \dbar \over \lrar L^p_{(0,1)}(\scra_\lambda, \cc^n)

7028: \lrar 0

7029: $$

7030: for $\lambda=0$ and $\lambda \not=0$. Then one applies $\sfl_\lambda$ and 

7031: $Q_\lambda$ to identify $\scrh(\scra_0, \cc^n)$ with $\scrh(\scra_\lambda, \cc^n)$

7032: and respectively $L^p_{(0,1)}(\scra_0, \cc^n)$ with $L^p_{(0,1)}(\scra_\lambda, 

7033: \cc^n)$. \qed

7034: 

7035: \medskip

7036: Let us explain now the main difficulty in the proof of \refthm{thm5.3.1}. Several 

7037: authors (see \eg \cite{Sie}, \cite{Li-T},  \cite{Ru}) have approached the 

7038: {\sl Gluing problem} %tried to solve the 

7039: by making an appropriate local imbedding $\scrp(\scra) \hook L^{1,p} (\scra, 

7040: \cc^n)= \sqcup_{\lambda \in \Delta(\epsi)} L^{1,p}(\scra_\lambda, \cc^n)$

7041: and showing that, roughly speaking, the $\dbar$-equation induces an operator which 

7042: defines $\scrp(\scra)$ and has a certain (rather weak) smoothness property. 

7043: This property is sufficient for showing that for any fixed $J \in \scrj$ the 

7044: compactification $\barr\scrm_J$ has a well-defined ``virtual fundamental class'',

7045: the basic object to define in the theory of Gromov-Witten invariants. 

7046: 

7047: The difficulty with this approach is that the smooth structure in $L^{1,p}(\scra,

7048: \cc^n)$ given by the isomorphism $L^{1,p}(\scra, \cc^n) \cong L^{1,p}(\scra_0, 

7049: \cc^n) \times \Delta(\epsi)$ depends heavily on both the linear and the complex 

7050: structures in $\cc^n$. Moreover, one can show that for a generic smooth 

7051: diffeomorphism $g: \cc^2 \to \cc^2$ the induced map $u \in L^{1,p}(\scra, \cc^n) 

7052: \mapsto u\scirc g \in L^{1,p}(\scra, \cc^n)$ is not even Lipschitzian at generic 

7053: $u \in L^{1,p}(\scra_0, \cc^n)$. 

7054: 

7055: To the contrary, our approach of imbedding $\scrp(\scra)$ into $\scrp(V)\times

7056: \Delta(\epsi)$ by means of tracing the restriction $u\ogran_V$ has the

7057: advantage that the smooth structure in $\scrp(V)$ is natural and canonical. 

7058: Here we shall use \lemma{lem5.3.3a} rather in a different way: It 

7059: serves for us as an approximative description of the behavior of the Gromov 

7060: operator $D_{u,J,\lambda}$ as $\lambda \lrar 0$. 

7061: 

7062: \newdefi{def5.3.4} Let $X$ be a manifold with a fixed symmetric connection

7063: $\nabla$. Then for $\lambda \in \Delta(\epsi)$, $u\in L^{1,p}(\scra_\lambda, X)$, 

7064: and a $C^1$-smooth almost complex structure on $X$ we denote by $J_\lambda$ the

7065: complex structure on $\scra_\lambda$ and by 

7066: $$

7067: D_{u,J,\lambda}: L^{1,p}(\scra_\lambda, E_u) \to 

7068: L^p_{(0,1)}(\scra_\lambda, E_u) 

7069: $$

7070: the operator given by 

7071: \begin{equation}\eqqno(5.3.a5)

7072: D_{u,J,\lambda}(v) \deff \nabla v + J \scirc \nabla v \scirc J_\lambda

7073: + \msmall{1\over2} \bigl( \nabla_vJ \scirc du \scirc J_\lambda -

7074: J \scirc \nabla_vJ \scirc du \bigr).

7075: \end{equation}

7076: Observe that in the definition of the space $L^p_{(0,1)}(\scra_\lambda, E_u)$ we 

7077: equip $E_u \deff u^*TX$ with the structure $u^*J$. The construction of 

7078: $D_{u,J,\lambda}$ is an extension of the definition of the Gromov operator, because

7079: for $(u, J)\in \scrp(\scra_\lambda)$ the definition \eqqref(5.3.a5) coincides

7080: with the original one in \eqqref(1.3.4).

7081: \end{defi}

7082: 

7083: 

7084: \newlemma{lem5.3.4} Let $B\subset \rr^{2n}$ be the ball and $J^*$ a  $C^1$-smooth 

7085: almost complex structure in $B$. Then there exist constants  $\epsi=\epsi(J)>0$ 

7086: and $C =C(J) < \infty$ such that for any 

7087: almost complex structure $J$ with

7088: $$

7089: \norm{J -J^*}_{C^1(B)} \le \epsi,

7090: $$

7091: any $\lambda \in\Delta(\epsi)$, and any $J$-holomorphic map $u: \scra_\lambda \to

7092: B(\half)$ with

7093: $$

7094: \norm{du}_{L^p(\scra_\lambda)} \le \epsi,

7095: $$

7096: 

7097: \smallskip

7098: \sli 

7099: one has a uniform estimate

7100: \begin{equation}

7101: \norm{v}_{L^{1,p}(\scra_\lambda)} \le C \cdot 

7102: \bigl(\norm{v}_{L^{1,p}(V)} + 

7103: \norm{D_{u,J,\lambda}v}_{L^p(\scra_\lambda)} \bigr)

7104: \eqqno(5.3.4)

7105: \end{equation}

7106: for any $v\in L^{1,p}(\scra_\lambda, E_u)$;

7107: 

7108: \slii there exists an operator $T_{u,J,\lambda}: L^p_{(0,1)}(\scra_\lambda, E_u)

7109: \to L^{1,p}(\scra_\lambda, E_u)$ with $D_{u,J,\lambda} \scirc T_{u,J,\lambda}= id$;

7110: 

7111: \sliii moreover, the family of operators $T_{u,J,\lambda}$ depends continuously 

7112: on $(u,J,\lambda)$. 

7113: \end{lem}

7114: 

7115: 

7116: \proof Recall that on each half-annulus $\scra^\pm_\lambda$ the hyperbola metric

7117: \eqqref(5.3.0.1) is equivalent to the (flat) disc metric ${\isl\over2} dz^\pm

7118: \wedge \bar z^\pm$. Thus we can apply the Morrey estimate

7119: $$

7120: \diam(u(\scra_\lambda^\pm)) \le 

7121: C \cdot \norm{du}_{L^p(\scra^\pm_\lambda)}

7122: $$

7123: which is uniform in $\lambda$.  This gives

7124: \begin{equation}\eqqno(5.3.a6)

7125: \diam(u(\scra_\lambda)) \le C \cdot \norm{du}_{L^p(\scra_\lambda)} 

7126: \end{equation}

7127: again uniformly in $\lambda$. 

7128: 

7129: \smallskip

7130: Consequently, $\diam(u(\scra_\lambda))$ is sufficiently small. Let $J\st$ be a 

7131: linear complex structure in $\rr^{2n}$ with coincides with $J$ at some point $x_0=

7132: u(z_0)\in u(\scra_\lambda)$. Then $\norm{J\scirc u -J\st}_{C^0(\scra_\lambda)}$ is 

7133: also small enough. 

7134: 

7135: The canonical trivialization of the tangent bundle $TB$ yields the canonical

7136: trivialization of $E_u = u^*TB$ and the identification $L^{1,p}(\scra_\lambda, E_u)

7137: =L^{1,p}(\scra_\lambda, \cc^n)$. By \refcorol{cor5.3.3b} we obtain a structure of 

7138: a continuous Banach bundle on the union $\sqcup_{(u,\lambda)\in L^{1,p}(\scra,B)}

7139: L^{1,p}(\scra_\lambda, E_u)$.

7140: 

7141: A similar identification for $L^p_{(0,1)}(\scra_\lambda, E_u)$ requires some 

7142: modification, since in the definition of this space one

7143: involves the complex structure $J_u \deff u^*J = J\scirc u$. Therefore we must fix 

7144: a {\sl complex} isomorphism $\phi$ of $(TB, J)$ with the trivial complex bundle 

7145: $(TB, J\st)$ over $B$. This means that $\phi$ is an $\rr$-linear endomorphism of 

7146: $TB$ satisfying $\phi \scirc J = J\st \scirc \phi$. Since $J$ coincides with 

7147: $J\st$ at $x_0=u(z_0)$ with $z_0 \in \scra_\lambda$, we may also 

7148: assume that $\norm{\phi - \id_{TB}}_{C^0(u(\scra_\lambda))}$ is small enough. Using

7149: $\phi$, we obtain an isomorphism $\phi_*: L^p_{(0,1)}(\scra_\lambda, E_u)\buildrel 

7150: \cong \over \lrar L^p_{(0,1)} (\scra_\lambda, \cc^n)$. Moreover, the norms 

7151: $\norm{\phi_*}_{\sf{op}}$ and $\norm{\phi_*\inv}_{\sf{op}}$ are bounded uniformly 

7152: in $\lambda \in \Delta(\epsi)$. Now the composition $\phi_* \scirc D_{u,J,\lambda}$

7153: is a homomorphism between $L^{1,p}(\scra_\lambda, \cc^n)$ and $L^p_{(0,1)} 

7154:  (\scra_\lambda, \cc^n)$. 

7155: 

7156: Using the trivialization $(u^*TB, u^*J\st)$ we define the operator 

7157: $$

7158: \dbar: L^{1,p}(\scra_\lambda, \cc^n) \to L^p_{(0,1)} (\scra_\lambda, \cc^n).

7159: $$

7160: By construction, $\norm{ \dbar - \phi_* \scirc D_{u,J,\lambda}}_{\sf{op}}$

7161: is small enough. Consequently, the restriction of $D_{u,J,\lambda}$ to the image

7162: of the operator $T_\lambda$ constructed in \lemma{lem5.3.3a} is an isomorphism.

7163: Now the existence of $T_{u,J,\lambda}$ follows from the ``closed graph theorem'' 

7164: of Banach. 

7165: 

7166: Note that the choice of the point $x_0$ and the isomorphism $\phi$ used in the 

7167: construction of $T_{u,J,\lambda}$ can be made continuous on $(u,J,\lambda)$.

7168: This means that there exist families $x_0(u,J,\lambda)$ and $\phi(u,J,\lambda)$

7169: which depend continuously on  $(u,J,\lambda)$ and have the properties needed above.

7170: For example, one can set $x_0(u,J,\lambda) \deff u(1,\lambda)$, the image of

7171: the point $(1,\lambda) \in \scra_\lambda$. For such a choice of $x_0(u,J,\lambda)$

7172: and $\phi(u,J,\lambda)$ the family $T_{u,J,\lambda}$ also depends 

7173: continuously on $(u,J,\lambda)$.

7174: 

7175: \smallskip

7176: As a consequence of \sliip, it is sufficient to prove \eqqref(5.3.4) 

7177: under the additional condition $D_{u,J,\lambda}v =0$. But then 

7178: \begin{multline*}

7179: \norm{v}_{L^{1,p}(\scra_\lambda)} \le C_2 \cdot 

7180: \bigl(\norm{v}_{L^{1,p}(V)} + 

7181: \norm{\dbar v}_{L^p(\scra_\lambda)} \bigr)=

7182: C_2 \cdot \bigl(\norm{v}_{L^{1,p}(V)} + 

7183: \norm{\dbar v-  \phi_* \scirc D_{u,J,\lambda}v}_{L^p(\scra_\lambda)} \bigr)

7184: \\

7185: \le 

7186: C_2 \cdot \bigl(\norm{v}_{L^{1,p}(V)} 

7187:   + \norm{\dbar -  \phi_* \scirc D_{u,J,\lambda}}_{\sf{op}} 

7188:   \cdot \norm{v}_{L^{1,p}(\scra_\lambda)}\bigr). 

7189: \end{multline*}

7190: This yields \eqqref(5.3.4) provided $C_2 \cdot \norm{\dbar -  \phi_* \scirc 

7191: D_{u,J,\lambda}}_{\sf{op}} \le \half$.

7192: 

7193: \smallskip

7194: Note once more that all the estimates in the proof are uniform in $\lambda \in

7195: \Delta(\epsi)$. \qed

7196: 

7197: 

7198: \smallskip

7199: \newlemma{lem5.3.7} The union $\sqcup_{\lambda \in \Delta(\epsi)} T\scrp(

7200: \scra_\lambda)$ is a continuous locally trivial Banach bundle over 

7201: $\scrp(\scra)$.

7202: \end{lem}

7203: 

7204: \proof We use the map $\pr_V$ to identify $\sqcup_{\lambda \in \Delta(\epsi)} 

7205: \scrp(\scra_\lambda)$ with its image in $\scrp(V) \times \Delta(\epsi)$. By 

7206: \lemma{lem5.3.4} \sliiip, we can we consider $\sqcup_{\lambda \in \Delta(\epsi)} 

7207: T\scrp(\scra_\lambda)$ as subbundle of the bundle $T\scrp(V) \times \Delta(\epsi)$

7208: over $\scrp(V) \times \Delta(\epsi)$. This defines the canonical topology on 

7209: $\sqcup_{\lambda \in \Delta(\epsi)} T\scrp(\scra_\lambda)$. Moreover, 

7210: \lemma{lem5.3.4} \sliii implies the claim at all $(u,J,\lambda)$ with

7211: $\lambda\not=0$. Hence it remains to show the local triviality of the bundle

7212: in question in a neighborhood of a given $(u_0, J_0, 0) \in \scrp(\scra_0)$.

7213: 

7214: Consider first the special case when $\norm{du_0}_{L^p(\scra_0)}$ is small 

7215: enough. Under this assumption, \lemma{lem5.3.4} provides existence of 

7216: a continuous family of splittings $L^{1,p}(\scra_\lambda, E_u) = \ker( D_{u,J,

7217: \lambda}) \oplus \im(T_{u,J,\lambda})$ defined in a neighborhood of $(u_0,J_0,0)$ 

7218: in $\sqcup_{\lambda \in \Delta(\epsi)} \scrp(\scra_\lambda)$. Moreover, it follows

7219: from the construction of $T_{u,J,\lambda}$ that the map $\dbar:\im(T_{u,J,\lambda})

7220: \to L^p_{(0,1)}(\scra_\lambda, \cc^n)$ is an isomorphism. By \refcorol{cor5.3.3b},

7221: this implies local triviality of $\im(T_{u,J,\lambda})$. In a similar way one

7222: shows that for a continuous family of isomorphisms $\phi_{u,J,\lambda}: (TB, J) 

7223: \to  (TB, J\st)$ the operators

7224: $$

7225: v \in \ker( D_{u,J, \lambda}) \subset L^{1,p}(\scra_\lambda, E_u) 

7226: \mapsto \bigl(\id - T_\lambda \scirc \dbar\bigr)\phi_{u,J,\lambda} v

7227: \in \scrh(\scra_\lambda, \cc^n)\subset L^{1,p}(\scra_\lambda,  \cc^n)

7228: $$

7229: are isomorphisms. This gives the local triviality of $\ker( D_{u,J, \lambda})$.

7230: 

7231: Now observe that the bundle

7232: \begin{equation}\eqqno(5.3.f1)

7233: \{ (v, \dot J) \in T_{(u,J)}\scrp(\scra_\lambda):

7234: v = -T_{u,J,\lambda}(\dot J \scirc du \scirc J_\lambda) \}

7235: \end{equation}

7236: is a complement to $\ker( D_{u,J, \lambda})$ in $T_{(u,J)}\scrp(\scra_\lambda)$.

7237: Since the bundle \eqqref(5.3.f1) is isomorphic to the lift of $T\!\!\scrj(B)$ onto

7238: $\scrp(\scra_\lambda)$, it is locally trivial over the whole union $\sqcup_\lambda

7239: \scrp(\scra_\lambda)$. 

7240: 

7241: \smallskip

7242: Turn back to the general case of the lemma. For a given $(u_0, J_0, 0) \in 

7243: \scrp(\scra_0)$, find a radius $r>0$ such that for the subnode 

7244: $$

7245: \scra'_0 \deff \{ (z^+,z^-) \in \scra_0 : |z^+|<r,\; |z^-|<r\,\}

7246: = \Delta^+(r) \cup \Delta^-(r)

7247: $$

7248: the norm $\norm{du_0}_{L^p(\scra'_0)}$ is small enough. Define 

7249: $$

7250: \textstyle

7251: A^+\deff \{ z^+ \in \Delta^+: {r \over2} <|z^+| <1\,\},

7252: \qquad

7253: A^-\deff \{ z^- \in \Delta^-: {r \over2} <|z^-| <1\,\},

7254: $$ 

7255: $$ 

7256: \scra'_\lambda \deff \{ (z^+,z^-) \in \scra_\lambda: |z^+|<r,\; |z^-|<r\,\}.

7257: $$

7258: and 

7259: $$

7260: V' \deff A^+ \cup A^-,

7261: \qquad

7262: V'' \deff V' \cap \scra'_0.

7263: $$

7264: For $|\lambda|< {r \over2}$ we identify $A^\pm$ with the corresponding subsets in 

7265: $\scra_\lambda$ by means of the coordinates $z^\pm$. Then for $|\lambda|< {r \over

7266: 4}$ $A^+$ is disjoint from $A^-$. This gives a family of coverings $\scra_\lambda=

7267: \scra'_\lambda \cup V'$ parameterized by $\lambda$ with $|\lambda|<  {r \over4}$. 

7268: Moreover, we can consider $V'$ and $V''$ as constant (\ie independent of $\lambda$)

7269: complex curves. Now, for any $(u,J,\lambda) \in \scrp(\scra)$ sufficiently

7270: close to $(u_0,J_0,0)$ we obtain the sequence

7271: \begin{equation}\eqqno(5.3.f2)

7272: 0 \to T_{(u,J,\lambda)} \scrp(\scra_\lambda) 

7273: \buildrel \alpha_{u,J,\lambda} \over \lrar

7274: T_{(u,J,\lambda)} \scrp(V') 

7275: \oplus T_{(u,J,\lambda)} \scrp(\scra'_\lambda) 

7276: \buildrel \beta_{u,J,\lambda} \over \lrar

7277: T_{(u,J,\lambda)} \scrp(V'') 

7278: \to 0,

7279: \end{equation}

7280: where we set

7281: $$

7282: \alpha_{u,J,\lambda}(v) \deff 

7283: \bigl(v\ogran_{V'}, v\ogran_{\scra'_\lambda} )

7284: \qquad

7285: \beta_{u,J,\lambda}(v, w) 

7286: \deff v\ogran_{V''} - w\ogran_{V''}.

7287: $$

7288: We claim that the sequence \eqqref(5.3.f2) is exact and splits.

7289: Since $\alpha_{u,J,\lambda}$ is obviously injective, it is sufficient

7290: to construct a right inverse to each $\beta_{u,J,\lambda}$, \ie a homomorphism

7291: $$

7292: \gamma_{u,J,\lambda}: T_{(u,J,\lambda)} \scrp(V'') \to

7293: T_{(u,J,\lambda)} \scrp(V') 

7294: \oplus T_{(u,J,\lambda)} \scrp(\scra'_\lambda)

7295: $$

7296: such that $\beta_{u,J,\lambda} \scirc \gamma_{u,J,\lambda}= \id$. Furthermore,

7297: since $\beta_{u,J,\lambda}$ depends continuously on $(u,J,\lambda) \in \scrp(\scra

7298: )$ close to the $(u_0,J_0,0)$, it is sufficient to construct a right inverse only 

7299: for the $\beta_{u_0,J_0,0}$. 

7300: 

7301: \smallskip

7302: Consider the restriction map between $L^{1,p}(\scra_0)$ and $L^{1,p}(V'')$ induced

7303: by the imbedding $V'' \hook \scra_0$. It is well-known that this map admits a left 

7304: inverse, \ie a map $Q:L^{1,p}(V'') \to L^{1,p}(\scra_0)$ such that $Q(v'')\ogran

7305: _{V''} =v''$ for any $v''\in L^{1,p}(V'')$. Let us use the same notation for the 

7306: restriction of $Q$ onto $T_{(u_0,J_0,0)}\scrp(V'') \subset L^{1,p}(V'')$. Recall

7307: that in \lemma{lem5.3.4} we have constructed the operator $T_{u_0,J_0,0}: L^p_{

7308:  (0,1)}(\scra_0) \to L^{1,p}(\scra_0)$ which is a right inverse to the operator 

7309: $D_{u_0,J_0,0}: L^{1,p}(\scra_0) \to L^p_{(0,1)}(\scra_0)$. Denote by $\Chi

7310: _{\scra'_0}: L^p_{(0,1)}(\scra_0, E_{u_0}) \to L^p_{(0,1)}(\scra_0, E_{u_0})$ the 

7311: operator given the multiplication on the characteristic function of the subset 

7312: $\scra'_0 \subset \scra_0$. Now, we obtain a left inverse $\gamma_{u_0,J_0,0}$ to 

7313: the operator $\beta_{u_0,J_0,0}$ setting 

7314: $$

7315: \gamma_{u_0,J_0,0} (v'') \deff

7316: \biggl(\!\!

7317: \bigl( T_{u_0,J_0,0} \scirc \Chi_{\scra'_0} \scirc D_{u_0,J_0,0}

7318: \scirc  Q(v'') \bigr)\ogran_{V'},

7319: \bigl((T_{u_0,J_0,0} \scirc \Chi_{\scra'_0} \scirc D_{u_0,J_0,0} - \id)

7320: \scirc  Q(v'') \bigr)\ogran_{\scra'_0}

7321: \biggr)

7322: $$

7323: for $v'' \in T_{(u_0,J_0,0)} \scrp(V'')$. Let us check that $\gamma_{u_0,J_0,0}$ 

7324: has the desired properties. For a given $v'' \in T_{(u_0,J_0,0)} \scrp(V'')$,

7325: set $\ti v'' \deff Q(v'')$. Then $D_{u_0,J_0,0}\ti v''$ vanishes on $V''$.

7326: Consequently, $D_{u_0,J_0,0}\ti v''$ is the sum of the forms $\phi, \psi

7327: \in L^p_{(0,1)}(\scra_0, E_{u_0})$ with the supports in $V'\bs V''$ and $\scra'_0

7328: \bs V''$ respectively. Moreover, $\psi = \Chi_{\scra'_0} \scirc D_{u_0,J_0,0} 

7329: \ti v''$. From the relation $D_{u_0,J_0,0}\scirc T_{u_0,J_0,0} =\id$ we obtain

7330: $$

7331: D_{u_0,J_0,0} \bigl( T_{u_0,J_0,0} \scirc \Chi_{\scra'_0} \scirc D_{u_0,J_0,0}

7332: \scirc  Q(v'') \bigr)\ogran_{V'} =

7333: \bigl( \Chi_{\scra'_0} \scirc D_{u_0,J_0,0}

7334: \scirc  Q(v'') \bigr)\ogran_{V'} = 0.

7335: $$

7336: This means that the first component $\bigl( T_{u_0,J_0,0} \scirc \Chi

7337: _{\scra'_0} \scirc D_{u_0,J_0,0} \scirc  Q(v'') \bigr)\ogran_{V'}$ of $\gamma

7338: _{u_0,J_0,0} (v'')$ satisfies $D_{u_0,J_0,0}v=0$. Thus the first component

7339: of $\gamma_{u_0,J_0,0}$ takes values in $T_{(u_0,J_0,0)}\scrp(V')$. In the same

7340: way one can show that the second component of $\gamma_{u_0,J_0,0}$ takes values 

7341: in $T_{(u_0,J_0,0)}\scrp(\scra'_0)$. Finally, it is obvious that the difference

7342: of the components of $\gamma_{u_0,J_0,0} (v'')$ is $v''$. 

7343: The lemma follows.\qed

7344: 

7345: 

7346: \statep Proof. of \refthm{thm5.3.1}. Take some $(u_0, J_0) \in \scrp(\scra_0)$.

7347: Using \lemma{lem5.3.3} \sliiip, we identify $\scrp(\scra_0)$ with its image 

7348: $\pr_V\bigl( \scrp(\scra_0) \bigr) \subset \scrp(V)$ and $T_{(u_0, J_0)}\scrp

7349:  (\scra_0)$ with its image  $\pr_V\bigl( T_{(u_0, J_0)}\scrp (\scra_0) \bigr) 

7350: \subset T_{(u_0, J_0)}\scrp(V)$.

7351: 

7352: We claim that there exists a closed complement to $\pr_V\bigl( T_{(u_0, J_0)}

7353: \scrp (\scra_0)\bigr)$ 

7354: in $T_{(u_0, J_0)}\scrp(V)$. To show this, let us fix a closed nodal complex curve

7355: $C$ and an imbedding $\scra_0 \hook C$. Then there exists the unique open set $\wt

7356: V \subset C$ such that $\wt V \cap \scra_0 =V$ and $\wt V \cup \scra_0 =C$. Extend

7357: the bundle $E_{u_0} = u_0^*\cc^n$ to an $L^{1,p}$-smooth complex bundle $\wt E$ 

7358: over $C$. Also extend the operator $D_{u_0,J_0,0}: L^{1,p}(\scra_0, E_{u_0}) \to 

7359: L^p_{(0,1)} (\scra_0, E_{u_0})$ to an operator $\wt D: L^{1,p}(C, \wt E) \to L^p

7360: _{(0,1)} (C, \wt E)$ of the form $\wt D= \dbar_{\wt E} + \wt R$ where $\dbar_{\wt 

7361: E}$ is the Cauchy-Riemann operator corresponding to some holomorphic structure in 

7362: $\wt E$ and $\wt R$ is a $\cc$-antilinear $L^p$-integrable homomorphism between 

7363: $\wt E$ and $\wt E \otimes \Lambda^{(0,1)}C$, \ie $\wt R \in L^p\bigl(C, \barr

7364: \hom_\cc (\wt E, \wt E \otimes \Lambda^{(0,1)}C) \bigr)$. It is not difficult to 

7365: show that the extensions $\wt E$ and $\wt D$ can be made in such a way that the 

7366: operator $\wt D$ is an isomorphism. In particular, there exists the inverse 

7367: operator $\wt T: L^p_{(0,1)}(C, \wt E) \to L^{1,p}(C, \wt E)$. For an open set 

7368: $U \subset C$ define

7369: $$

7370: \scrh(U) \deff \{ v \in L^{1,p}(U, \wt E) : \wt Dv=0\}.

7371: $$

7372: In particular, we have $\scrh(\scra_0)=\pr_V\bigl( T_{(u_0,J_0)}\scrp(\scra_0)

7373: \bigr)$ and

7374: similar identifications for $V$ and $V'$. Define the operator $\beta: \scrh(\wt V)

7375: \oplus \scrh(\scra_0) \to \scrh(V)$ setting $\beta(v,w) \deff v\ogran_V -

7376: w\ogran_V$. Then there exists an operator $\gamma: \scrh(V)\to \scrh(\wt V)

7377: \oplus \scrh(\scra_0)$ which is left inverse to $\beta$. Indeed, 

7378: the construction of the operator $\gamma_{u_0,J_0,0}$ made in the proof of

7379: \lemma{lem5.3.7} can be applied with appropriate modifications. In particular,

7380: one uses $\wt T$ instead of $T_{u_0,J_0,0}$. 

7381: 

7382: Observe that the kernel $\ker(\beta)$ can be identified in a natural way with 

7383: the kernel of $\wt D: L^{1,p}(C, \wt E) \to L^p_{(0,1)} (C, \wt E)$. This implies

7384: that $\beta$ is an isomorphism and $\gamma$ its inverse. Consequently, every

7385: $v\in \scrh(V)$ can be uniquely represented in the form $v=v_1 + v_2$ such that 

7386: $v_1$ extends to $\ti v_1 \in \scrh(\wt V)$ and $v_1$ extends to $\ti v_2 \in 

7387: \scrh(\scra_0)$. The set of all $v_1\in \scrh(V)= T_{u_0,J_0}\scrp(V)$ obtained in

7388: this way forms the desired closed compliment to $\pr_V\bigl( T_{(u_0, J_0)}\scrp 

7389:  (\scra_0) \bigr)$ in $T_{(u_0, J_0)}\scrp(V)$.

7390: 

7391: \medskip

7392: The existence of such a complement implies that there exists a small ball $\scru 

7393: \subset \scrp(\scra_0)$ centered at $(u_0, J_0)$ and an open imbedding $F: \scru 

7394: \times \scrb \hook \scrp(V)$ with the following properties:

7395: \begin{itemize}

7396: \item $\scrb$ is a small ball in a closed complement of $\pr_V \bigl(T_{(u_0, 

7397: J_0)}\scrp(\scra_0) \bigr)$ in $T_{\pr_V(u_0, J_0)} \scrp(V)$;

7398: \item the map $F$ is a $C^1$-diffeomorphism onto image;

7399: \item the restricted map $F\ogran_{\scru \times \{0\}}: \scru \to \pr_V(\scru)

7400: \subset \pr_V(\scrp(\scra_0)) \subset \scrp(V)$ coincides with $\pr_V$.

7401: \end{itemize} 

7402: In other words, $\scru \times \scrb$ appears as a chart for $\scrp(V)$ in

7403: which $\pr_V(\scru)= F(\scru \times \{0\})$ is a linear subspace.

7404: 

7405: Now consider the natural projection $\pi_\scru: \scru \times \scrb \to \scru$

7406: restricted to $F\inv\bigl(\pr_V(\scrp(\scra_\lambda))\bigr)$. By 

7407: \lemma{lem5.3.7} this is a $C^1$-diffeomorphism for all sufficiently small 

7408: $\lambda$, provided the ball $\scru$ is small enough. Denote by 

7409: $$

7410: \Phi_\lambda: \scru \to \pr_V\inv\bigl(\pr_V(\scrp(\scra_\lambda)) \cap 

7411:  F(\scru \times \scrb)  \bigr) \subset \scrp(\scra_\lambda)

7412: $$ 

7413: the inverse map. We claim that the family $\Phi_\lambda$, parameterized by $\lambda

7414: \in\Delta(\epsi')$ with a sufficiently small $\epsi'>0$, has the properties

7415: stated in \refthm{thm5.3.1}. In fact, it remains to check only the fact that

7416: the whole map 

7417: $$

7418: \Phi: \scru \times \Delta(\epsi') \to \sqcup_{\lambda

7419: \in\Delta(\epsi')} \scrp(\scra_\lambda)

7420: $$

7421: is continuous. However, this follows from the construction of $\Phi$. \qed

7422: 

7423: 

7424: \newsection[6]{Symplectic isotopy problem in $\cp^2$}

7425: 

7426: \newsubsection[6.1]{Symplectic isotopy problem}

7427: %The symplectic isotopy problem comes in question as one of possible applications 

7428: %of \refthm{thm4.5.1}. We describe it as follows. 

7429: Let $\Sigma, \; \Sigma'$ be two (connected) symplectically imbedded surfaces in a 

7430: symplectic 4-fold $(X, \omega)$. Assume that they have the same homology class. 

7431: Then they have the same genus, see \lemma{lem1.1.2}. Thus one can ask whether or 

7432: not there exists an isotopy $\{ \Sigma_t \}_{t \in [0, 1]}$ from $\Sigma$ to 

7433: $\Sigma'$ such that all $\Sigma_t$ are also symplectically imbedded.

7434: This is refered to as the symplectic isotopy problem.

7435: 

7436: The example of Fintushel and Stern \cite{Fi-St} shows that there is no hope 

7437: %strangles the hope 

7438: to obtain a results of this type in the case when $\la c_1(X), [\Sigma] \ra \le 0$.

7439: Namely, they proved that under certain conditions on a symplectic 4-fold $(X, 

7440: \omega)$ there exists an infinite 

7441: collection of symplectic imbeddings $\Sigma_i \hook X$, such that $\Sigma_i$ 

7442: represent the same homology class $[C] \in \sfh_2(X, \zz)$ but are pairwise 

7443: non-isotopic, even smoothly. Moreover, the class of symplectic 4-folds with these 

7444: conditions is sufficiently wide, so that one has enough examples of this type.

7445: 

7446: \smallskip

7447: On the other hand, \refthm{thm4.5.1} hints that a satisfactory solution for the 

7448: symplectic isotopy problem in the case $\la c_1(X), [C] \ra \ge 1$ is possible. 

7449: We state the problem in a more precise form.

7450: 

7451: \state Conjecture 1. {\sl (Symplectic isotopy problem). \it Let $(X,\omega)$ be 

7452: a compact symplectic $4$-dimensional manifold and $[C]\in \sfh_2(X,\zz)$ a 

7453: homology class with $\la c_1(X), [C] \ra \ge1$. Then every two symplectically {\sl

7454:  immersed} surfaces $\Sigma$ and $\Sigma'$ in the class $[C]$ are symplectically 

7455: isotopic provided they have the same genus $g$ and the only singularities are 

7456: {\sl positive} nodal points.}

7457: 

7458: 

7459: Recall, there exists a complete classification of compact symplectic 4-folds

7460: $X$ which come in question. Namely, {\bf Corollary 1.5} in \cite{McD-Sa-3}, 

7461: claims

7462: 

7463: \newprop{prop6.1.0a}

7464: Let $X$ be a symplectic manifold and $\Sigma

7465: \subset X$ a symplectically imbedded surface which is not an exceptional

7466: sphere. Then $X$ is the blow-up of a rational or ruled manifold.

7467: \end{prop}

7468: 

7469: The complete description of possible symplectic structures on such $X$ was done in 

7470: \cite{McD-4}, \cite{La-McD}, and \cite{McD-Sa-3}, see also \cite{Li-Liu}, 

7471: \cite{Liu}.

7472: 

7473: 

7474: As the main result of this paper we give a positive solution of the symplectic

7475: isotopy problem for imbeddings of low degree in $\cp^2$.

7476: 

7477: \newthm{thm6.1.1} Any two symplectically imbedded surfaces $\Sigma,\; \Sigma'

7478: \subset \cp^2$ of the same degree $d\le6$ are symplectically isotopic.

7479: \end{thm}

7480: 

7481: The case $d=1$ and $2$ of the theorem has been proven by Gromov in \cite{Gro},

7482: the case $d=3$ by Sikorav \cite{Sk-3}.

7483: 

7484: \smallskip

7485: In this connection a result of S.~Finashin about (non-symplectic) isotopy problem 

7486: in $\cp^2$ should be mentioned. He proves in \cite{Fin} that for any

7487: even degree $d=2k\ge6$ there exist infinitely

7488: many isotopy classes of imbedded real surfaces in $\cp^2$ having the degree $d$

7489: and the genus $g$ given by the genus formula, \ie $g={(d-1)(d-2) \over2}\cdot\;$

7490: Note that \refthm{thm6.1.1} claims that for $d=6$ only one of these isotopy 

7491: classes is realizable by a {\sl symplectic} imbedding.

7492: 

7493: \medskip

7494: Let us explain main ideas of the proof of \refthm{thm6.1.1}. First we observe

7495: that existence of a symplectic isotopy $\{ \Sigma_t \}_{t \in [0,1]}$ between

7496: surfaces $\Sigma,\; \Sigma'$ in a symplectic manifold $(X,\omega)$ implies  

7497: existence of an ``accompanying'' homotopy $\{ J_t \}_{t \in [0, 1]}$ of tame 

7498: almost complex structures, such that the imbeddings $\Sigma_t \hook X$ are 

7499: $J_t$-holomorphic. Conversely a homotopy of $\omega$-tame $J_t$-holomorphic 

7500: imbeddings is necessarily a symplectic isotopy. So given $\Sigma _0$ and 

7501: $\Sigma_1$, the natural thing to do is to outfit them with compatible structures 

7502: $J_0$ and $J_1$, take a generic curve $J_t$ and attempt to find appropriate 

7503: liftings $\Sigma _t$. We do this using the following theorem of Harris \cite{Ha}

7504: for an intermediate construction.

7505: 

7506: 

7507: \newprop{prop6.1.2} Any two irreducible nodal algebraic curves $C_0$ and $C_1$ in 

7508: $\cp^2$ of the same degree $d$ and the same geometric genus $g$ are {\sl 

7509: holomorphically isotopic}, \ie can be connected by an isotopy $\{C_t\}_{t\in [0,

7510: 1]}$ consisting of nodal algebraic curves.

7511: \end{prop}

7512: 

7513: By this result, in order to construct the symplectic isotopy, it is enough

7514: to construct a lifting as above for the case where $J_1$ is the standard

7515: integrable structure on $\cp^2$ and $\Sigma_1$ is some smooth algebraic curve.

7516: 

7517: \medskip

7518: Obviously, \refthm{thm4.5.1} would imply existence of symplectic isotopy if we 

7519: could show that for a generic path $\{J_t \}_{t \in [0,1]}$ the moduli space $\scrm

7520: _{J_t}$ is non-empty. An obstruction to this is the fact that the projection $\pi

7521: _\scrj:\scrm \to \scrj$ is not proper. This means that we must understand the 

7522: structure of the total moduli space $\scrm$ ``at infinity''. In 

7523: \refsubsection{5.2c} we have constructed a completion $\barm$ of $\scrm$ and 

7524: eqquiped it with a natural stratification such that every stratum is a smooth

7525: Banach manifold. In particular, the transversality technique developed in 

7526: \refsection{2} can be applied to every such stratum. 

7527: 

7528: The next idea in the proof of \refthm{thm6.1.1} is to construct a path $\ti\gamma

7529: _t \deff(C_t, J_t) \in \barr\scrm$ which goes piecewisely along some strata and 

7530: which can be ``pushed'' into the ``main stratum'' $\scrm$ yielding the desired 

7531: isotopy $(\Sigma_t, J_t)$. The main difficulty in realization this idea is to 

7532: ensure that pushing $\ti\gamma_t$ into $\scrm$ we still remain in the same 

7533: connected component of $\scrm$ so that the symplectic isotopy class is preserved.

7534: This means that we are interested in describing possible connected components of 

7535: $\scrm$ in a neighborhood of a given curve $(C^*, J^*) \in \barr\scrm$. Moreover, 

7536: the positive solution of a symplectic isotopy problem would follow immediately from

7537: the fact that locally exactly one such component exists. Indeed, it would be then

7538: sufficient to construct {\sl any} path $\ti\gamma_t\deff(C_t, J_t) \in \barr\scrm$

7539: connecting $\Sigma$ and $\Sigma'$. But existence of such a path follows easily

7540: from \refthm{thm4.5.1} in the case $c_1(X)[C]>0$. 

7541: 

7542: \smallskip

7543: The result of \refthm{thm6.1.1} is obtained via the proof of the local uniqueness

7544: of such a component of $\scrm$ near a given $(C^*, J^*) \in \barr\scrm$ in the

7545: special case when $C^*$ contains no multiple components. The restriction $d\le6$

7546: in the theorem comes from the fact that in this case it is possible to avoid the 

7547: appearance of multiple components in $C^*$. We are able to do so by demanding that 

7548: the pseudoholomorphic curves $\Sigma_t$ in the isotopy path pass through fixed 

7549: generic $3d-1$ points on $X=\cp^2$. Note that the number $3d-1$ is the maximal 

7550: possible in \refthm{thm4.5.3}.

7551: 

7552: 

7553: 

7554: \newsubsection[6.2]{Local symplectic isotopy problem} As we have explained

7555: in the previous paragraph, we are interested in the possible symplectic isotopy

7556: classes of pseudoholomorphic curves $C$ in a neighborhood of a given singular curve

7557: $C^*$ with no multiple components. The main difficulty in this case is, of course,

7558: to understand the local behavior of curves $C$ near singular points of $C^*$. 

7559: In this way we come to the following question.

7560: 

7561: \state The Local Symplectic Isotopy Problem. 

7562: {\sl Let $B$ be the unit ball in $\rr^4$ equipped with the standard symplectic 

7563: structure $\omega\st$, $J^*$ an $\omega\st$-tame almost complex structure, and 

7564: $C^*\subset B$ a connected $J^*$-holomorphic curve in $B$ with a unique isolated 

7565: singularity at $0\in B$ and without multiple components. Describe the possible 

7566: symplectic isotopy classes of curves $C$ in $B$ which lie sufficiently close to 

7567: $C^*$ \wrt the cycle topology and which have prescribed singularities, \eg

7568: prescribed number of nodes and ordinary cusps.} 

7569: 

7570: 

7571: \medskip

7572: We start with a construction of certain symplectic isotopy classes of nodal

7573: pseudoholomorphic curves. For $C^*$ as above, let $C^*= \cup_i C^*_i$ be the 

7574: decomposition

7575: into irreducible components. Then there exist $J^*$-holomorphic parameterizations 

7576: $u^*_i : S_i \to B$, $u^*_i(S_i) = C^*_i$. Shrinking $C^*_i$, if needed, we may 

7577: assume that all $S_i$ are compact and smooth boundaries $\d S_i$, each 

7578: consisting of finitely many circles. Note that the images of the boundary

7579: circles are imbedded in $B$ and mutually disjoint. Further, we can also suppose 

7580: that $u^*_i$ are $L^{1,p}$-smooth up to boundaries $\d S_i$. Set $S\deff \sqcup 

7581: S_i$ and define $u^*: S \to B$ by $u^*\ogran_{S_i} \deff u^*_i$. Denote by $J^*

7582: _S$ the complex structure on $S$ induced by $u^*: S \to B$ from $C^*$.

7583: 

7584: 

7585: \newlemma{lem6.2.1} \sli The set $\scrp(S,B)_\nod$ of those $(u, J_S, J) \in 

7586: \scrp(S,B)$ for which the map $u:S \to B$ is an immersion and the singularities 

7587: of the image $C \deff u(S)$ are only nodal points is open and dense in $\scrp(S,B)$

7588: and is connected;

7589: 

7590: \slii For $(u', J'_S, J')$ and $(u'', J''_S, J'') \in \scrp(S,B)_\nod$, 

7591: sufficiently close to $(u^*, J^*_S, J^*) \in \scrp(S,B)$, the pseudoholomorphic 

7592: curves $C' \deff u'(S)$ and $C' \deff u'(S)$ are symplectically isotopic;

7593: 

7594: \sliii For a fixed $J_S \in \scrj_S$ and $J\in \scrj(B)$, the subspace of nodal 

7595: curves in each of the spaces $\scrp(S; B, J)$, $\scrp(S,J_S; B)$, and $\scrp(S,J_S; 

7596: B, J)$ is open and dense in thecorresponding space.  

7597: \end{lem}

7598: 

7599: \proof By results of \refsection{4}, the complement to $\scrp(S,B)_\nod$ in 

7600: $\scrp(S,B)$ is closed and consists of submanifolds of real codimension at least

7601: 2. This shows \sli and implies \slii. Part \sliii is obtained similarly. 

7602: \qed

7603: 

7604: \newdefi{def6.2.1} In the situation of \lemma{lem6.2.1}, we call $C_\nod=u(S)$ a 

7605: {\sl maximal nodal deformation of $C^*$} and the number $\delta$ of nodes on $C

7606: _\nod$ the {\sl nodal number of $C^*$} at the singular point $0\in C^*$. In other 

7607: words, a maximal nodal deformation is a nodal pseudoholomorphic curve obtained from

7608: $C^*= u^*(S)$ by a (sufficiently small) generic deformation of the parameterization

7609: map $u^*: S \to X$, $C^*= u^*(S)$. 

7610: 

7611: Further, a {\sl canonical smoothing of $C^*$} is a $J^*$-holomorphic curve 

7612: $C^\dag$ obtained from a maximal nodal nodal deformation $C_\nod$ by smoothing of 

7613: all nodes. We use the notion of canonical smoothing for both the construction and 

7614: the resulting curve. Further, we shall always assume that a canonical smoothing 

7615: $C^\dag$ is sufficiently close to $C^*$ \wrt the cycle topology. 

7616: \end{defi}

7617: 

7618: It follows immediately from \lemma{lem6.2.1} that the symplectic isotopy class

7619: of a canonical smoothing of $C^*$ is well-defined.

7620: 

7621: 

7622: \newprop{prop6.2.1a} Any two curves $C^\dag_1$ and $C^\dag_2$ obtained from $C^*$

7623: by the construction of canonical smoothing are symplectically isotopic. Moreover,

7624: such an isotopy can be be carried out sufficiently close to the identity map.

7625: \end{prop}

7626: 

7627: \smallskip

7628: Note that the number $\delta(C_\nod)$ of nodes on a maximal nodal deformation

7629: $C_\nod$ of $C^*$ equals the nodal number $\delta(0, C^*)$ of $C^*$ at $0$. Observe

7630: also that one can smooth {\sl some} number of nodes on $C_\nod$ producing further 

7631: symplectic isotopy classes. It is easy to show that these new classes are 

7632: determined by the set of the nodes on $C_\nod$ which are smoothed. We conjecture 

7633: that these are all possible symplectic isotopy classes of nodal curves in a 

7634: neighborhood of $C^*$ \wrt the cycle topology. 

7635: 

7636: \state Conjecture 2. {\sl (Local symplectic isotopy problem for nodal curves).

7637: \it Let $J^*$ be a $C^2$-smooth $\omega\st$-tame almost

7638: complex structure in $B \subset \rr^4$ and $C^* \subset B$ a $J^*$-holomorphic 

7639: curve with a unique isolated singular point at $0\in B$ and without multiple 

7640: components. Assume that $J$ is an almost complex structure in $B$ which is $C^{0,

7641: \alpha}$-smooth for $\alpha>0$ and sufficiently close to $J^*$ \wrt the $C^{0,

7642: \alpha}$-topology.

7643: 

7644: Then any {\it nodal} $J$-holomorphic curve $C$ sufficiently close to $C^*$ \wrt 

7645: the cycle topology is symplectically isotopic to a $J^*$-holomorphic curve 

7646: obtained from a maximal nodal deformation $C_\nod$ of $C^*$ by smoothing some 

7647: number of nodes on $C_\nod$.}

7648: 

7649: 

7650: \medskip

7651: We give a proof the conjecture for the case of {\sl imbedded} curves. Observe 

7652: that here we have only one candidate, namely the canonical smoothing.

7653: 

7654: 

7655: \newthm{thm6.2.2} In the situation described in {\sl Conjecture 2}, let $C^\dag$ 

7656: be $J^*$-holomorphic curve obtained by the canonical smoothing of $C^*$.

7657: 

7658: Let $J$ be an almost complex structure on $B$ sufficiently close to $J^*$ \wrt the

7659: $C^{0,\alpha}$-topology and $C$ an imbedded $J$-holomorphic $C$ sufficiently close

7660: to $C^*$ \wrt the cycle topology. Then there exist a homotopy $J_t$ which is

7661: $C^0$-sufficiently close to $J^*$ and connects $J^*$ with $J$, and an isotopy

7662: $C_t$ of $J_t$-holomorphic curves which connects $C^\dag$ with $C$ and is 

7663: sufficiently close to $C^*$ \wrt the cycle topology.

7664: \end{thm}

7665: 

7666: \smallskip

7667: The proof will be given after some preparatory results. We shall always assume

7668: that the hypotheses of the theorem are fulfilled. Denote by $S$ the real surface

7669: parameterizing $C$. In other words $S$ is the curve $C$, considered as 

7670: real oriented surface without complex structure.

7671: 

7672: \smallskip

7673: Our first observation is that the theorem holds in the case when $C^*$ and the

7674: approximating curve $C$ are holomorphic in the usual sense. The result is 

7675: well-known, see \eg \cite{Mil}. Its proof is based on the main advantage of the 

7676: holomorphic case: the fact that one can represent a holomorphic curve as the zero 

7677: divisor of a holomorphic function. 

7678: 

7679: \newlemma{lem6.2.2a} Let $f^*$ be a holomorphic function in the ball $B$ in 

7680: $\cc^2$ whose zero divisor is a holomorphic curve $C^*$ with a single singular 

7681: point at $0 \in B$ and without multiple components. Assume that $f^*$ and $C^*$

7682: are sufficiently smooth also at the boundary $\d B$. Then

7683: \begin{itemize}

7684: \item[\slip] a canonical smoothing $C$ is obtained as the zero divisor of

7685: a sufficiently small perturbation $f$ of $f^*$;

7686: \item[\sliip] for two generic sufficiently small perturbations $f_1$ and $f_2$

7687: of $f^*$ their zero divisors $C_0$ and $C_1$ are non-singular and {\sl 

7688: holomorphically isotopic}, \ie can be connected by a homotopy consisting

7689: of holomorphic non-singular curves $C_t$.

7690: \end{itemize}

7691: \end{lem}

7692: 

7693: 

7694: \smallskip

7695: Denote by $\delta^*$ the nodal number of $C^*$ at $0\in C^*$. We may assume 

7696: inductively that the claim of \refthm{thm6.2.2} holds for all curves $C'$ which 

7697: satisfy the hypotheses of the theorem but have the nodal number $\delta(C')$ at

7698: $0\in C'$ which is strictly less than $\delta^*$. Further, we assume that $\delta^*

7699: \ge2$, since otherwise $\delta^*=1$ and $0\in C^*$ is a nodal point, 

7700: the case covered by \refsubsection{5.3}.

7701: 

7702: \medskip

7703: Recall that by the theorem of Micallef and White (see \lemma{lem1.2.1}) in a 

7704: neighborhood of $0\in B$ there exists a $C^1$-diffeomorphism $\phi$ of $B\subset 

7705: \rr^4$ such that $\phi(0)=0$, $\phi_*(J^*(0))=J\st$, the standard complex structure

7706: in $\rr^4=\cc^2$, and such that $\phi(C^*)$ is a $J\st$-holomorphic curve. 

7707: Obviously, we may also assume that $d\phi : T_0B \to T_0B$ is the identity map. 

7708: This means that the form $\phi_*\omega\st$ coincides with $\omega\st$ at $0\in B$,

7709: $\phi_* (\omega\st)\ogran_{T_0B}=\omega\st$, and similarly $\phi_*(J^*(0))=J^*(0)

7710: =J\st$. Consequently, $\phi_*(J^*)$ is $\omega\st$-tame in a sufficiently small 

7711: ball $B(r)$, $r\ll1$. Let us fix such a radius $r$. 

7712: 

7713: Moreover, since $C^*\subset B$ is imbedded outside $0$, we can additionally assume

7714: that $\phi$ is {\sl smooth} outside $0\in B$.

7715: 

7716: Below, we translate the original situation by means of such $\phi$ and work with 

7717: a {\sl holomorphic} curve $\phi(C^*)\cap B(r)$. This leads to the difficulty that 

7718: $\phi_*(J^*)$ is apriori only {\sl continuous} at $0\in B(r)$. This requires an

7719: additional control on the behavior of pairs $(C,J) \in \scrp(B)$ approximating

7720: $(C^*, J^*)$.

7721: 

7722: 

7723: 

7724: \newlemma{lem6.2.3} Let $(u_n, J_{S,n}, J_n) \in \scrp(S, B)$ be a sequence such 

7725: that $J_n$ converges to $J^*$ in the $C^{0,\alpha}$-topology with $0<\alpha <1$, 

7726: and $C_n \deff u_n(S)$ converges to $C^*$ \wrt the cycle topology and \wrt the

7727: $L^{1,p}$-topology near boundary $\d C_n = u_n(\d S)$. Further, let $\phi: B \to

7728: B$ be the diffeomorphism introduced above. Then for all sufficiently big $n$

7729: 

7730: \sli $u_n: S \to B$ is an imbedding;

7731: 

7732: \slii there exists a sequence $J^*_n$ of $C^\ell$-smooth almost complex structures 

7733: in $B$ such that

7734: \begin{itemize}

7735: \item $u_n$ are $(J_{S,n}, J^*_n)$-holomorphic;

7736: \item $\phi_*(J^*_n)$ converges to $J\st$ in the $C^0$-topology in $B(r)$ and

7737: in the $C^{0, \alpha}$-topology outside $0\in B(r)$.

7738: \end{itemize}

7739: \end{lem}

7740: 

7741: \proof The first part follows from \lemma{lem1.2.2a}, applied to a smaller

7742: ball $B(\rho)$, $\rho<1$, and curves $C^* \cap B(\rho)$, $C_n \cap B(\rho)$.

7743: 

7744: \smallskip

7745: Define $J^\sharp$ as the pull-back of $J\st$ \wrt $\phi$, $J^\sharp \deff \phi^*(

7746: J\st)$. Then the second part is equivalent to the convergence $J^*_n \lrar 

7747: J^\sharp$ in the appropriate topology. 

7748: 

7749: Fix some sufficiently small $\epsi>0$. Since $J^*(0)=J\st=J^\sharp(0)$, there 

7750: exists a positive radius $\rho\ll r$ such that $\norm{J^* -J^\sharp}_{C^0(B(\rho))} 

7751: < \epsi$. This implies that $\norm{J_n -J^\sharp}_{C^0(B(\rho))} < \epsi$ for all 

7752: sufficiently big $n$.

7753: 

7754: Now observe that in $B\bs B(\rho)$ we have the $C^{1,\alpha}$ convergence $C_n 

7755: \lrar C^*$. In particular, in $B\bs B(\rho)$ we have $C^{0,\alpha}$-convergence of 

7756: tangent bundles $TC_n \lrar TC^*$. This implies that for $n \gg1$

7757: we can extend every $J_n$ from $B(\rho)$ to $B(r)$ as a $C^\ell$-smooth structure 

7758: $J^*_n$ which is defined {\sl along} $C_n$ and obeys the estimate

7759: \begin{subequations}

7760: \eqqno(6.2.1)

7761: \begin{align}

7762: &\norm{J^*_n -J^\sharp}_{C^0(C_n\cap B(r))} < \epsi,

7763: \\

7764: &\norm{J^*_n -J^\sharp}_{C^{0,\alpha}(C_n \cap ( B(r) \bs B(2\rho) ))} 

7765: < \epsi.

7766: \end{align}

7767: \end{subequations}

7768: Finally we extend the constructed $J^*_n$ from $C_n \cup B(\rho)$ to the whole

7769: ball $B$ preserving the estimates \eqqref(6.2.1). \qed

7770: 

7771: \state Remark. In fact, below we shall merely make use of the weaker $C^0

7772: $-convergence $\phi_*(J^*_n) \to J\st$. The H\"older $C^{0,\alpha} $-convergence 

7773: $J_n \to J^*$ was used only to provide the $C^0$-convergence of 

7774: tangent bundles $TC_n \to TC^*$ outside $0\in C^*$. In particular, it would be 

7775: sufficient to have only $C^0$-convergence $J_n \to J^*$ in $B$ and the $C^{0,

7776: \alpha}$-convergence outside $0\in B$. On the other hand, in the case when the 

7777: convergence $J_n \to J^*$ is better, say in the $C^\ell$-topology with 

7778: non-integer $\ell >1$, we could achieve just as as well the $C^\ell$-convergence in 

7779: $B(r)$ outside $0$.

7780: 

7781: \medskip

7782: \lemma{lem6.2.3} insures that we can reduce the problem to the case when $C^*$ is

7783: holomorphic in the usual sense, \ie \wrt the structure $J\st$. Further, observe

7784: that for the proof of \refthm{thm6.2.2} it is sufficient to show that for any 

7785: sequence $(u_n, J_{S,n}, J_n)$ satisfying the hypotheses of \lemma{lem6.2.3} the 

7786: curves $C_n \deff u_n(S)$ are symplectically isotopic to $C^\dag$ for $n\gg1$. An 

7787: equivalent problem is to show that $\phi(C_n)$ are symplectically isotopic to 

7788: $\phi(C^\dag)$ in $B(r)$. Thus we can replace our initial objects by their 

7789: $\phi$-images in $B(r)$. For the sake of simplicity we maintain the original 

7790: notations for these new objects, \eg $B$ for $B(r)$, $C^*$ and $C_n$ for 

7791: respectively $\phi(C^*) \cap B(r)$ and $\phi(C_n) \cap B(r)$, $J^*$ and $J_n$ for 

7792: respectively $\phi_*(J^*)\ogran_{B(r)}$ and $\phi_* (J_n)\ogran_{B(r)}$, and so 

7793: on. On the other hand, $J\st$ and $\omega\st$ remain the standard structures in 

7794: $B$. Observe that now we have the weaker $C^0$-convergence $J_n \lrar J^*$. 

7795: 

7796: \smallskip

7797: Imbed $B$ in $\cp^2$ in the standard way so that $J\st$ becomes the standard

7798: integrable structure, still denoted by $J\st$. Then we can extend $\omega$ to a 

7799: global symplectic form on $\cp^2$ taming $J\st$. We maintain the notation $\omega$

7800: for this extension.

7801: 

7802: We claim that $C^*$ also extends to $\cp^2$ as a compact closed pseudoholomorphic 

7803: curve. Moreover, we claim that there exists an extension $\ti C^*$ with the 

7804: following properties

7805: \begin{itemize}

7806: \item all irreducible components of $\ti C^*$ are rational, \ie 

7807: parameterized by the sphere $S^2$;

7808: \item except for the original singularity at $0\in \ti C^*$, all new singularities 

7809: are only nodal points.

7810: \end{itemize}

7811: 

7812: Indeed, every irreducible component of $C^*\subset B$ is

7813: parameterized by a holomorphic map $u_i=u_i(z): \Delta \to B$ with $u_i(0)=0$. For

7814: every $u_i(z)$ we take the Taylor polynomials $u_i^{(d)}(z)$ of degree $d$ chosen 

7815: sufficiently high to satisfy the following conditions: 

7816: \begin{itemize}

7817: \item every $u_i^{(d)}(z)$ is non-multiple;

7818: \item the images $u_i^{(d)}(\Delta)$ are pairwise distinct holomorphic discs.

7819: \end{itemize}

7820: 

7821: Then every $u_i^{(d)}(z)$ can be considered as an algebraic map $f_i$ from $\cp^1=

7822: S^2$ to $\cp^2$. Making a generic perturbation of $f_i$ outside $B$, we obtain 

7823: desired curve $\ti C^* \subset \cp^2$ as the union of the images $\ti f_i(S^2)$ of

7824: the perturbed maps. Observe that $d$ appears as the degree of every component 

7825: $\ti C^*_i \deff \ti f_i(S^2)$. 

7826: 

7827: \newlemma{lem6.2.4} There exist an almost complex structure $\ti J^*$ and points 

7828: $x_\alpha$ on $\ti C^*$ satisfying the following conditions:

7829: \begin{itemize}

7830: \item[{\sl (a)}] the points $x_\alpha$ are pairwise distinct, and there 

7831: are exactly $3d-1$ of them on every component $\ti C^*_i$;

7832: \item[{\sl (b)}] $\ti J^*$ is $C^\ell$-smooth and $\omega$-tame, $\ti C^*$ is 

7833: $\ti J^*$-holomorphic, and $\ti J^*$ coincides with $J^*$ on $B$;

7834: \item[{\sl (c)}] any $\ti J^*$-holomorphic curve $C'$ which

7835:   \begin{itemize}

7836:   \item passes through the fixed points $x_\alpha$;

7837:   \item is sufficiently close to $\ti C^*$ \wrt the cycle topology;

7838:   \item has the same number of singular points as $\ti C^*$;

7839:   \item[$*$] has a singular point $x'\in C'$ with the nodal number $\delta^*$

7840:  at $x'$

7841:   \end{itemize} 

7842:   must coincide with $\ti C^*$.

7843: \end{itemize}

7844: \end{lem}

7845: 

7846: The last property asserts that every pseudoholomorphic curve $C'\not=\ti C^*$ with 

7847: the properties {\sl(c)} except $(*)$ has simpler singularities than $\ti C^*$. So 

7848: the induction assumption can be applied to such a $C'$. 

7849: 

7850: \proof We use the results of {\sl Sections \ref{sec:2}} and {\sl\ref{sec:4}}. 

7851: Fix non-singular points $x_\alpha$ on $\ti C^*$ such that condition {\sl(a)} 

7852: is fulfilled. Let $\mbfx_i$ be the $(3d-1)$-tuple of the points lying on the

7853: component $\ti C^*_i$. Denote by $\scrm'$ the space of pairs $(C',J')$, 

7854: where $J'\in \scrj$ and $C'$ is $J'$-holomorphic curve $C'$ satisfying properties 

7855: {\sl(c)} except $(*)$. Then by the genus formula \eqqref(1.2.1) any such curve 

7856: $C'$ has only rational irreducible components $C'_i$, the number of which is the

7857: same as for $\ti C^*$, and the degree of 

7858: every component $C'_i$ is $d$. This means that $\scrm'$ is the fiber product

7859: of the spaces $\scrm(S^2, \cp^2, d, \mbfx_i)$ of rational pseudoholomorphic curves

7860: of degree $d$ in $\cp^2$ passing through $\mbfx_i$. The product is taken

7861: over the space $\scrj$ of almost complex structure in $\cp^2$. By the transversality

7862: technique of \refsection{2}, the space $\scrm'$ a Banach manifold. To compute

7863: the Fredholm index of the natural projection $\pi'_\scrj: \scrm' \to \scrj$ observe

7864: that the expected dimension of rational $J$-holomorphic curves in $\cp^2$ of 

7865: degree $d$ passing through $3d-1$ fixed distinct points is $0$. This implies that

7866: the index of the projection $\pi'_\scrj: \scrm' \to \scrj$ is also $0$.

7867: 

7868: Further, by results of \refsection{4} the condition $(*)$ defines a proper 

7869: $C^\ell$-smooth submanifold $\scrm^*$ in $\scrm'$ of finite codimension, say $m$. 

7870: Consequently, the index of the corresponding projection $\pi^*_\scrj\deff \pi'

7871: _\scrj\ogran_{\scrm^*}: \scrm^* \to \scrj$ is negative. Using the transversality 

7872: technique of \refsection{2} we can construct a $C^\ell$-smooth submanifold $Y 

7873: \subset \scrm^*$ of dimension $m$ such that 

7874: \begin{itemize}

7875: \item $(\ti C^*, \ti J^*)\in Y$ for some $\ti J^*$ obeying the condition 

7876: {\sl(b)} of the lemma;

7877: \item $Y$ is transversal to $\scrm^*$;

7878: \item the restricted projection $\pi'_\scrj\ogran_Y: Y \to \scrj$ is an 

7879: imbedding.

7880: \end{itemize}

7881: Then $(\ti C^*, \ti J^*)$ is an isolated point of the intersection $Y \cap 

7882: \scrm^*$. But this means that $\ti J^*$ has the desired properties. \qed 

7883: 

7884: 

7885: 

7886: \medskip

7887: Below we shall need a property which is a bit sharper than {\sl(c)} in 

7888: \lemma{lem6.2.4}. Roughly speaking, it claims that one can recover a 

7889: pseudoholomorphic curve $C$ in $B$ knowing its part $\bigl(\barr B \bs B(\half)

7890: \bigr) \cap C$. 

7891: 

7892: \newdefi{def6.2.2} Denote by $A$ the spherical annulus $\barr B \bs B(\half)$.

7893: It is a closed subset of the closed unit ball $\barr B \subset \cp^2$. For

7894: closed subsets $Y_1, Y_2 \subset \cp^2$ we denote by $\dist_A(Y_1, Y_2)$ the

7895: Hausdorff distance between $Y_1 \cap A$ and $Y_2 \cap A$,

7896: $$

7897: \dist_A(Y_1, Y_2) \deff \dist(Y_1\cap A, Y_2\cap A),

7898: $$

7899: if both $Y_1 \cap A$ and $Y_2 \cap A$ are non-empty. The standard distance 

7900: function in $\cp^2$ is used as the base. If exactly one of the set $Y_i \cap A$ 

7901: is empty, we set $\dist_A(Y_1, Y_2) \deff \diam(\cp^2)$. If $Y_1 \cap A= Y_2 

7902: \cap A =\emptyset$, we define $\dist_A(Y_1, Y_2) \deff 0$. We call $\dist_A$

7903: the {\sl $A$-distance}. 

7904: \end{defi}

7905: 

7906: It is easy to see that $\dist_A$ is only a pseudo-distance function, \ie it is 

7907: non-negative, symmetric, and has the triangle inequality property, but does not 

7908: distinguish all closed subsets $Y_1 \not= Y_2 \subset \cp^2$ in general. It turns 

7909: out that it induces the cycle topology on the set of pseudoholomorphic curves lying

7910: in a sufficiently small  $\dist_A$-neighborhood of $\ti C^*$ provided only $C^1

7911: $-smooth almost complex structures $J$ are used. More precise statement is given in

7912: 

7913: \newlemma{lem6.2.4a} There exists an $\epsi>0$ with the following property.

7914: 

7915: Let $J \in \scrj$ be a $C^1$-smooth almost complex structure which satisfies the

7916:  condition $\norm{J-J^*}_{C^0(\cp^2)} \le \epsi$ and $C$ a $J$-holomorphic curve 

7917: which is homologous to $\ti C^*$ and satisfies the condition $\dist_A(C, \ti C^*)

7918: \le \epsi$. Then for any sequence $J_n$ of continuous almost complex 

7919: structures $J_n$ converging to $J$ in the $C^0$-topology, $\norm{J_n-J}_{C^0(\cp

7920: ^2)} \lrar 0$, and any sequence of $J_n$-holomorphic curves $C_n$ the condition 

7921: $\dist_A(C_n, C) \lrar 0$ implies that $C_n$ converges to $C$ in the cycle 

7922: topology.

7923: \end{lem}

7924: 

7925: \proof Consider a sequence of almost complex structures $J_n$ in $\cp^2$ which

7926: converges to $J^*$ in the $C^0$-topology, and a sequence $C_n$ of closed $J_n

7927: $-holomorphic curves homologous to $C^*$, for which $\lim \dist_A(C_n, \ti C^*)=

7928: 0$. Then $J_n$ are $\omega\st$-tame for all $n\gg1$. Hence we can apply the Gromov

7929: compactness theorem (see \refthm{thm5.2.1}). This means that some subsequence,

7930: still denoted $C_n$, converges to a $J^*$-holomorphic curve $C^+$ \wrt the cycle

7931: topology. The condition $\lim \dist_A(C_n, \ti C^*)= 0$ implies that $\dist_A(C^+,

7932: \ti C^*)=0$, which means that $\ti C^* \cap A = C^+ \cap A$.

7933: 

7934: Observe now that by the construction of $\ti C^*$ every irreducible component $\ti 

7935: C^*_i$ of $\ti C^*$ meets the interior $\sf{Int}(A)$ of $A$. By the unique

7936: continuation property of pseudoholomorphic curves proven in \lemma{lem1.2.4} 

7937: \sliip, every component $\ti C^*_i$ is contained in $C^+$. Thus $\ti C^* \subset 

7938: C^+$. Since $C^+$ is homologous to $\ti C^*$, we must have equality $\ti C^*= C^+$.

7939: This means that $C_n$ converges to $\ti C^*$ in the cycle topology. In particular,

7940: for every sufficiently big $n$ every irreducible component of $C_n$ meets the 

7941: interior $\sf{Int}(A)$ of $A$. 

7942: 

7943: The latter property shows that the same argumentation can be used if we replace

7944: $\ti C^*$ by any $C_n$ with $n\gg1$ and the lemma follows. \qed

7945: 

7946: \medskip

7947: Now we are ready to complete the

7948: 

7949: \statep Proof. of \refthm{thm6.2.2}. It follows from the construction

7950: of the extension $\ti C^*$ that the sequence $(C_n, J_n)$ can be extended 

7951: to a sequence $(\ti C_n, \ti J_n)$ such that $\ti J_n$ is a sequence of

7952: $\omega$-tamed almost complex structures in $\cp^2$ converging to $\ti J^*$

7953: and $\ti C_n$ is a sequence of compact (\ie closed) $\ti J_n$-holomorphic curves 

7954: converging to $\ti C^*$. Moreover, we may additionally assume that the curves

7955: $\ti C_n$ pass through the marked points $\mbfx$ for all sufficiently big $n$.

7956: 

7957: Observe that all $\ti C_n$ are symplectically isotopic. We denote by $\ti S$ 

7958: the closed oriented real surface parameterizing $\ti C_n$. It can be obtained

7959: from the surface $S$ parameterizing $C_n$ by gluing in discs to fill out the holes 

7960: in $S$.

7961: 

7962: 

7963: Fix a sequence of homotopies $\{\ti J_{n,t}\}_{t\in [0,1]}$ of almost complex 

7964: structures with the following properties:

7965: \begin{itemize}

7966: \item all $\ti J_{n,t}$ are $C^\ell$-smooth and depend $C^\ell$-smoothly on $t$;

7967: \item every initial structure $J_{n,0}$ is $\ti J_n$;

7968: \item for some small $\epsi_0>0$ the structures $\ti J_{n,t}$ are integrable in 

7969: $B$ for all $t\in [1-\epsi_0, 1]$;

7970: \item as $n$ goes to infinity, the structures $\ti J_{n,t}$ converge to $\ti J^*$

7971: in the $C^0$-topology uniformly in $t\in [0,1]$, \ie 

7972: $$

7973: \lim_{n\lrar\infty}\; \mathop{\sup}\limits_{t\in [0,1]}\;

7974: \norm{\ti J_{n,t} - \ti J^*}_{C^0(\cp^2)} =0;

7975: $$

7976: \item the homotopy $\{\ti J_{n,t}\}_{t \in [0,1]}$ is generic for every $n$.

7977: \end{itemize} 

7978: 

7979: \smallskip

7980: Now let us try to deform continuously every $\ti C_n$ inside a family $\ti J_{n,

7981: t}$-holomorphic curves preserving the isotopy class. Since we want to control also

7982: the local isotopy class we must impose the condition that the curves in the family

7983: lie sufficiently close to $\ti C^*$. Apriori, it can occur that such a curve does 

7984: not exist for all $t\in [0,1]$. Nevertheless, we can find the maximal subinterval 

7985: where such a family of curves exists. Moreover, we allow that under the deformation

7986: some nodal points appear. Let us formalize this observation.

7987: 

7988: \newprop{prop6.2.6} Fix a sufficiently small $\epsi>0$. Then for every $n\gg1$ 

7989: there exists a $t^+_n \in (0,1]$ which is {\sl maximal} \wrt the following 

7990: condition: 

7991: 

7992: For any $t <t^+_n$ there exists a $\ti J_{n,t}$-holomorphic curve $\ti C

7993: _{n,t}$ such that

7994: \begin{itemize}

7995: \item $\ti C_{n,t}$ passes through the fixed points $\mbfx$ on $\cp^2$;

7996: \item the curve $\ti C'_{n,t}$, obtained from $\ti C_{n,t}$ by smoothing of all 

7997: singular points contained in $B$, is symplectically isotopic to $\ti C_n$;

7998: \item $\dist_A(\ti C_{n,t}, C^*) < \epsi$.

7999: \end{itemize}

8000: \end{prop}

8001: 

8002: Recall that for $t$ sufficiently close to $1$ the structures $\ti J_{n,t}$ are 

8003: integrable in $B$. So if $t^+_n =1$ for some $n$, then for some $t$ close to $1$ 

8004: we obtain a {\sl holomorphic} curve $C_{n,t} \deff \ti C_{n,t} \cap B$ whose

8005: smoothing is symplectically isotopic to the original curve $C_n$. In this case 

8006: \refthm{thm6.2.2} follows from \lemma{lem6.2.2a}. We claim that it is always 

8007: the case for $n\gg1$.

8008: 

8009: \smallskip

8010: To show this, let us analyze the possible reasons which could cause the strict

8011: inequality $t^+_n<1$ for a given $n\gg1$. 

8012: %To do it let us fix some $n\gg1$ and 

8013: Consider

8014: an increasing sequence of parameters $t_\nu$ approaching to $t^+_n$. Then there

8015: exists a sequence of $\ti J_{n, t_\nu}$-holomorphic curves $\ti C_{n, t_\nu}

8016: $ with the properties from \propo{prop6.2.6}. In particular, all $\ti C_{n, 

8017: t_\nu}$ are homologous to $\ti C^*$. Taking a subsequence, we may assume that

8018: $\ti C_{n, t_\nu}$ converges to a $\ti J_{n, t^+_n}$-holomorphic curve $\ti C^+_n$

8019: in the cycle topology. Note $\dist_A(\ti C^+_n, \ti C^*)\le \epsi$ by our 

8020: construction. By \lemma{lem6.2.4a}, $\ti C^+_n$ is sufficiently close to 

8021: $\ti C^*$ also \wrt the cycle topology. Consequently, near every nodal point of 

8022: $\ti C^*$ there is exactly one nodal point of $\ti C^+_n$. 

8023: 

8024: Observe that $\ti C^+_n$ has no singular point $x^+_n \in \ti C^+_n$ with the 

8025: nodal number $\ge \delta^*$ at $x^+_n$. Indeed, otherwise we can repeat the 

8026: argumentation from the proof of \lemma{lem6.2.4} and show that $\ti C^+_n$ must 

8027: consist of rational components the number of which is the same as that for 

8028: $\ti C^*$. But the expected 

8029: dimension of such curves in the space $\scrm'_{\ti J_{n, t^+_n}, \mbfx}$ with a 

8030: singular point of this type is negative and less then $-1$. So the existence of 

8031: $x^+_n\in \ti C^+_n$ with $\delta(x^+_n, \ti C^+_n) \ge \delta^*$ contradicts the 

8032: genericity of the path $\ti J_{n, t}$. Thus all singularities of $\ti C^+_n$ are 

8033: simpler than those of $\ti C^*$. By the induction assumption, the curve $\ti C'_n$ 

8034: obtained as the canonical smoothing of all singular points of $\ti C^+_n$ 

8035: contained in $B$ is symplectically isotopic to $\ti C_n$.

8036: 

8037: Let $u^+_n: S^+_n \to \cp^2$ be a {\sl normal} parameterization of $\ti C^+_n$

8038: (see \refdefi{def5.2c.3}).

8039: Consider the relative moduli space $\scrm_{h_n, \mbfx}(S^+_n, \cp^2)$ of $h_n(t)= 

8040: \ti J_{n,t}$-holomorphic curves which are parameterized by $S^+_n$, are in the 

8041: homology class $[\ti C^*]$, and pass through the fixed points $\mbfx$. This space

8042: is non-empty since it contains $(\ti C^+_n, t^+_n)$. \refthm{thm4.5.3} provides 

8043: that for some interval $t\in [t^+_n, t^{++}_n]$ with $t^{++}_n >t^+_n$ we can 

8044: construct a path of $\ti J_{n,t}$-holomorphic curves $\ti C_{n, t}$ which lies in 

8045: $\scrm_{h_n, \mbfx}(S^+_n, \cp^2)$ and starts at $\ti C^+_n$. Then the curves 

8046: obtained from such $\ti C_{n, t}$ by smoothing of all singular points contained 

8047: in $B$ will be symplectically isotopic to $\ti C_n$. Moreover,

8048: if we would additionally have the strict inequality $\dist_A(\ti C^+_n, \ti C^*)< 

8049: \epsi$, then $\dist_A(\ti C_{n, t}, \ti C^*)< \epsi$ for some $t\in ]t^+_n, 

8050: t^{++}_n[$, and this would contradict the maximality of $t^+_n$.

8051: 

8052: \smallskip

8053: Thus we may assume that $\dist_A(\ti C^+_n, \ti C^*)= \epsi$ for every $n$. Then 

8054: for every $n\gg1$ we can fix $t^-_n \in [0, t^+_n]$ and a $\ti J_{n,t^-_n}

8055: $-holomorphic curve $\ti C_{n,t^-_n}$ which has properties from \propo{prop6.2.6} 

8056: and satisfies the additional condition 

8057: $$

8058: \msmall{\epsi\over2} \le \dist_A(\ti C_{n, t^-_n}, \ti C^*) \le \epsi.

8059: $$ 

8060: Taking a subsequence, we may assume that $\ti C_{n,t^-_n}$ converges to a $\ti J

8061: $-holomorphic curve $\ti C^+$ in the cycle topology. Then 

8062: $$

8063: \msmall{\epsi\over2} \le \dist_A(\ti C^+, \ti C^*) 

8064: \le \epsi.

8065: $$

8066: By {\sl Lemmas \ref{lem6.2.4}} and {\sl\ref{lem6.2.4a}}, $\ti C^+$ must have 

8067: simpler singularities than $\ti C^*$ provided the constant $\epsi$ was chosen 

8068: small enough. By the induction assumption, the curve $\ti C'$ obtained by canonical 

8069: smoothing of all singular points of $\ti C^+$ lying in $B$ is symplectically 

8070: isotopic to every $\ti C_n$, as also to every $\ti C_{n,t^-_n}$. On the other 

8071: hand, $\ti J^*$ coincide in $B$ with the standard structure $J\st$. Thus $C' 

8072: \deff \ti C' \cap B$ is a canonical smoothing of $C^*$ by \lemma{lem6.2.2a}. 

8073: 

8074: 

8075: 

8076: \newsubsection[6.3]{Global symplectic isotopy in $\cp^2$} In this paragraph give

8077: 

8078: \nobreak

8079: \statep Proof. of \refthm{thm6.1.1}. We proceed by making appropriate 

8080: modifications of the argumentation used in the proof of \refthm{thm6.2.2}. Let 

8081: $\Sigma$ be an imbedded surface in $\cp^2$ of degree $d\le 6$, such that $\omega\st

8082: \ogran_{\Sigma}$ is non-degenerate. By \propo{prop6.1.2}, to prove the theorem it 

8083: is sufficient to show that $\Sigma$ is symplectically isotopic to a non-singular 

8084: algebraic curve of degree $d$.

8085: 

8086: \smallskip 

8087: Find an $\omega\st$-tame almost complex structure $J_0$ making $\Sigma$ a $J_0

8088: $-holomorphic curve, denoted by $C_0$. Fix $3d-1$ distinct points $\mbfx=(x_1,

8089: \ldots, x_{3d-1})$ on $C_0$. Perturbing $C_0$ and the points, we may assume that 

8090: $x_1,\ldots, x_{3d-1}$ are in generic position \wrt the standard structure $J\st$ 

8091: in the following sense. For any positive degree $d'\le d$ and any closed oriented 

8092: surface $S$, not necessary connected, the moduli space $\scrm_{J\st, \mbfx}(S, 

8093: \cp^2; d')$ of $J\st$-holomorphic (and hence {\sl algebraic}) curves of degree 

8094: $d'$ with normalization $S$ passing through $\mbfx$ is a (possibly empty)

8095: complex space of the expected dimension.

8096: 

8097: Fix a generic path $h(t)$ of $\omega\st$-tame almost complex structures $J_t \deff 

8098: h(t)$ connecting $J_0$ with $J\st=J_1$. Without loss of generality we may assume

8099: that all $J_t$ are $C^\ell$-smooth and depend $C^\ell$-smoothly on $t$. 

8100: 

8101: \newprop{prop6.3.1} There exists a $t^+ \in (0,1]$ which is {\sl maximal} \wrt 

8102: the following condition: 

8103: 

8104: For any $t <t^+$ there exists a $J_t$-holomorphic curve $C_t$ such that 

8105: \begin{itemize}

8106: \item[\slip] $C_t$ passes through the fixed points $\mbfx$ on $\cp^2$;

8107: \item[\sliip] $C_t$ is non-multiple, but not necessarily irreducible;

8108: \item[\sliiip] the curve $C'_t$, obtained from $C_t$ by smoothing of all singular

8109: points, is symplectically isotopic to $C_0$.

8110: \end{itemize}

8111: \end{prop}

8112: 

8113: To prove the theorem, we must show that $t^+=1$ and that there exist a $J\st

8114: $-holomorphic curve $C_1$ with the properties \slip--\slii in the proposition. 

8115: 

8116: \smallskip

8117: Let $t_n$ be an increasing sequence converging to $t^+$. Fix $J_{t_n}$-holomorphic

8118: curves $C_n$ with these properties. Property \slii implies that the $C_n$ have

8119: the same degree $d$. Going to a subsequence we may assume that they converge

8120: to a $J_{t^+}$-holomorphic curve $C^+$ in the cycle topology. 

8121: 

8122: We claim that $C^+$ contains no multiple components. To show this, it is sufficient

8123: to consider only the case when $C^+$ has only two components $C^+_1$ and $C^+_2$ 

8124: with multiplicities $m_1=1$ and $m_2=2$ respectively. Let $d_i$ be the degree of 

8125: $C^+_i$, so that $d_1+2d_2=d$. Then the geometric genus $g_i$ of every $C_i$ is at

8126: most $g_i \le{(d_i-1)(d_i-2) \over 2}\cdot\;$ By the genericity of the path $h(t)=

8127: J_t$, each $C_i$ can contain at most $k_i \le 3d_i -1 + g_i \le{d_i(d_i+3) \over 

8128: 2}$ of the fixed points $\mbfx$, see \refsubsection{2.4}. Thus $C^+$ can contain

8129: at most $\le{d_1(d_1+3) + d_2(d_2+3) \over 2}$ points. It is easy to show that for

8130: $d\le 6$ this number is strictly less then $3d-1$. For example, in the worst case 

8131: with $d=6$, $d_1=4$, and $d_2=1$ we would have on $C^+$ at most ${4\cdot(4+3) 

8132: \over 2} + {1\cdot (1+3)\over 2}= 14 +2 =16$ the marked points $\mbfx$ instead 

8133: of the necessary $3\cdot 6-1 =17$. Observe, that this argument remains valid also 

8134: in the case $t^+=1$ and $J_1=J\st$.

8135: 

8136: Now the results of \refsubsection{6.2} show that the curve $C'$ obtained 

8137: from $C^+$ by the 

8138: canonical smoothing of all singular points is symplectically isotopic to $C_0$.

8139: This implies the theorem in the case $t^+=1$. Indeed, in this case $C^+$ is the 

8140: zero divisor of a homogeneous polynomial $F^+$ of degree $d$. Making a generic

8141: perturbation of coefficients of $F^+$ we obtain a polynomial $F'$ whose

8142: zero divisor is an algebraic (and hence $J\st$-holomorphic) curve $C'$ which

8143: is symplectically isotopic to $C_0$.

8144: 

8145: In the case $t^+ <1$ we must show that for some $t^{++}>t^+$ there exists a 

8146: $J_{t^{++}}$-holomorphic curve $C^{++}$ with the properties given in 

8147: \propo{prop6.3.1}. To do this we fix a normal parameterization $u^+: S^+ \to 

8148: C^+ \subset \cp^2$ (see \refdefi{def5.2c.3}) and consider the relative moduli 

8149: space $\scrm_{h, \mbfx}(S^+, \cp^2, d)$ of $J_t=h(t)$-holomorphic curves which 

8150: are parameterized by $S^+$, 

8151: pass through $\mbfx$ and have the degree $d$. This space is non-empty because it 

8152: contains $C^+$. The results of \refsubsection{4.5} imply that for some interval 

8153: $t\in [t^+_n, t^{++}_n]$ with $t^{++}_n >t^+_n$ we can construct a path of 

8154: $J_t$-holomorphic curves $C^+_t$ which lies in $\scrm_{h, \mbfx}(S^+, \cp^2, d)$ 

8155: and starts at $C^+$. By \refsubsection{6.2}, the $C^+_t$ have the the properties 

8156: \slip--\sliii from \propo{prop6.3.1}. This contradicts the maximality of $t^+$ 

8157: and implies the statement of \refthm{thm6.1.1}. \qed

8158: 

8159: \medskip

8160: \state Remark. In fact, the real homotopy $C_t$ from $C_0=\Sigma$ to an algebraic

8161: curve $C_1$ has the property described in \refsubsection{6.1}. Namely, after 

8162: fixing a generic homotopy $h(t)=J_t$, one tries to construct {\sl any} path $C_t$

8163: of {\sl imbedded} $J_t$-holomorphic curves $C_t$. Such a path exists for some

8164: interval $t\in [0, t')$. The saddle point property proven in

8165: \refsubsection{4.5} removes the main difficulty in the construction of $C_t$:

8166: the presence of local maxima in the corresponding moduli space $\scrm$. This means

8167: that at end of this interval, when $t \lrar t'$, the curves $C_t$ go to infinity in

8168: $\scrm$. By Gromov compactness, going along some sequence $t'_n \lrar 

8169: t'$, we approach a $J_{t'}$-holomorphic curve $C'$ lying on some infinity 

8170: stratum of $\barr \scrm$ parameterized by a new moduli space $\scrm'$. As we have 

8171: shown in the proof, one can avoid the strata $\scrm'$ corresponding to curves with

8172: multiple components. Now we continue to deform $C'$ as a path $C'_t$ along 

8173: $\scrm'$, having in mind that the canonical (in fact, any) smoothing of singular 

8174: points of $C'_t$ gives curves symplectically isotopic to $C_0$. The new path $C'_t$

8175: continues until we come to the next infinity stratum $\scrm''$ of $\barr \scrm$,

8176: and so on.

8177: 

8178: \medskip

8179: We finish the paper with a remark on {\sl Conjecture 2} about the local symplectic

8180: isotopy problem for nodal curves. The proof of this result would follow from the 

8181: corresponding result for {\sl holomorphic} curves, which is essentially a local 

8182: version of the Severi problem, see \propo{prop6.1.2}. Indeed, the proof of 

8183: \refthm{thm6.2.2} could be applied after appropriate modification. 

8184: 

8185: \smallskip

8186: \state Conjecture 3. {\sl (Local Severi-Harris problem). \it Let $C^*$ be 

8187: a holomorphic curve in the ball $B \subset \cc^2$ with a unique isolated singular 

8188: point at $0\in B$ and without multiple components. 

8189: 

8190: Then any {\it nodal} holomorphic curve $C\subset B$ sufficiently close to $C^*$ 

8191: \wrt the cycle topology is holomorphically isotopic to a holomorphic curve 

8192: $C^\dag$ obtained from a maximal nodal deformation $C_\nod$ of $C^*$ by smoothing 

8193: some number of nodes on $C_\nod$.}

8194: 

8195: \smallskip

8196: In view of the main results of this paper, the validity of {\sl Conjecture 3}, 

8197: and hence of {\sl Conjecture 2}, seems quite plausible.

8198: 

8199: 

8200: \vskip1cm plus 1cm

8201: 

8202: 

8203: \newpage

8204: 

8205: %\input iso_bib.tex

8206: \ifx\undefined\bysame

8207: \newcommand{\bysame}{\leavevmode\hbox to3em{\hrulefill}\,}

8208: \fi

8209: 

8210: \begin{thebibliography}{sh-sh-sh}

8211: %\show\bibitem

8212: %\makeatletter

8213: %\show\@bibitem

8214: 

8215: 

8216: 

8217: \def\entry#1#2#3#4\par{\bibitem[#1]{#1}

8218: {\textsc{#2 }}{\sl{#3}}#4\par\vskip2pt}

8219: %{ref}{author}{title}.

8220: 

8221: \entry{Abi}%

8222: {Abikoff, W.}% 

8223: {The real analytic theory of Teichm\"uller space.} 

8224: Lecture Notes in Mathematics, {\bf820}. Springer, Berlin, 1980. vii+144 pp.,

8225: {\bf Math.\ Rev.:\.}82a:32028.

8226: 

8227: \entry{Ar}{Aronszajn, N.:}{A unique continuation theorem for elliptic

8228: differential equation or inequalities of the second order.} J.\ Math.\ Pures

8229: Appl., {\bf36}, 235--339, (1957).

8230: 

8231: %\entry{B-K}{Bedford, E., Klingenberg W.:}{On the~envelope of meromorphy of

8232: %a 2-sphere in $\cc^2$.}  J.\ Amer.\ Math.\ Soc., {\bf4}, 623-646, (1981).

8233: 

8234: 

8235: \entry{B-P-V}{Barth W., Peters, C., Van de Ven, A.:}{Compact complex

8236: surfaces.} Springer Verlag, (1984).

8237: 

8238: %

8239: \entry{Bi}{Birman, J.:}{Braids, links, and mapping class groups.}

8240: Annals of Math.\ Studies., {\bf82}(1975), Princeton Univ.\ Press and Univ.\ 

8241: of Tokyo Press., 229 p., (1975).

8242: 

8243: \entry{Bn}{Bennequin, D.:}{Entrelacement et \'equation de Pfaff.}

8244: Ast\'erisque {\bf107--108}(1982), 87--161.

8245: 

8246: 

8247: \entry{Ch-Sp}{Chern, S.-S., Spanier, E.:}{A theorem on orientable surfaces

8248: in four-dimensional space.} Comment.\ Math.\ Helv., {\bf25}(1951), 205--209.

8249: 

8250: 

8251: %\entry{Ch}{Chirka E.:}{On generalized Hartogs lemma.} To appear in the

8252: %volume dedicated to B.\.V.\.Shabat, PHASIS, Moscow.

8253: 

8254: %\entry{Ch-St}{Chirka, E., Stout L.:}{A Kontinuit\"atssatz.} In ``Topics in

8255: %Complex Analysis'', Banach Center Publications, {\b31}, 143--350, (1995).

8256: 

8257: \entry{Diaz}%

8258: {Diaz, Steven}%

8259: {Exceptional Weierstrass points and the divisor on moduli space that they 

8260: define.} Mem.\ AMS, {\bf56}(1985), No.\.327, iv+69 pp.,  

8261: {\bf Math.\ Rev.:\.} 86j:14022 

8262: 

8263: \entry{Eli}{Eliashberg, Y.:}{Filling by holomorphic discs and its applications.}

8264: London Math.\ Soc.\ Lecture Notes, {\bf151}, Geometry of low dimensional

8265: manifolds, (1991).

8266: 

8267: \entry{Fi-St}%

8268: {Fintushel, Ronald; Stern, Ronald J.:}%

8269: {Symplectic surfaces in a fixed homology class.}

8270: J.\ Diff.\ Geom., {\bf52}(1999), No.\.2, 203--222.

8271: 

8272: \entry{Fin}%

8273: {Finashin, Sergey:}%

8274: {Knotting of Algebraic Curves in $\cp^2$.}

8275: Preprint, {\tt arXiv:math.GT/9907108}.

8276: 

8277: %\entry{F}{Forstneri\v c, F.:}{Complex tangents of real surfaces in complex

8278: %Ssurfacea.} Duke Math.\ J., {\bf67}, 353--376, (1992).

8279: 

8280: %\entry{Gra}{Grauert, H.:}{\"Uber Modifikationen und exzeptionelle analytische

8281: %Mengen.} Math.Ann. {\bf146}, 331--368 (1962).

8282: 

8283: \entry{Gr-Ha}{Griffiths, P., Harris, J.:}{Principle of algebraic geometry.}

8284: John Wiley \& Sons, N.-Y., (1978).

8285: 

8286: 

8287: \entry{Gro}{Gromov, M.:}{Pseudo holomorphic curves in symplectic

8288: manifolds.} Invent.\ math. {\bf82}, 307--347 (1985).

8289: 

8290: \entry{Ha}{Harris, J.:}{On the Severi problem.} Invent.\ Math., 

8291: {\bf84}(1986), 445-461.

8292: 

8293: \entry{Ha-Mo}%

8294: {Harris, J; Morrison, I.:}%

8295: {Moduli of curves.} 

8296: Graduate Texts in Mathematics, {\bf187}, 

8297: Springer-Verlag, 1998, xiv+366 pp., ISBN 0-387-98438-0,

8298: {\bf Math.\ Rev.:}\.99g:14031.

8299: 

8300: %\entry{Hv-Po}{Harvey, R., Polking, J.:}{Removable singularities of

8301: %solutions of partial differential equations.} Acta Math. {\bf125}, 209--226

8302: %(1970).

8303: 

8304: 

8305: \entry{Hrt-W}{Hartman, P., Winter, A.:}{On the~local behavior of solutions

8306: of non-parabolic partial differential equations.} Amer.\ J.\ Math.,

8307: {\bf75}, 449--476, (1953).

8308: 

8309: %

8310: 

8311: \entry{Hf}{Hofer, H.:}{Pseudoholomorphic curves in symplectizations with

8312: applications to the Weinstein conjecture in dimension three.} Invent.\ Math.,

8313: {\bf114}, 515--563 (1993).

8314: %

8315: 

8316: \entry{H-L-S}{Hofer, H., Lizan, V., Sikorav, J.-C.:}{On genericity for

8317: holomorphic curves in four-dimensional almost-complex manifolds.} J.\ of Geom.\

8318: Anal., {\bf7}, 149--159, (1998).

8319: 

8320: \entry{Hum}%

8321: {Hummel, Christoph.:}%

8322: {Gromov's compactness theorem for pseudo-holomorphicmorphic curves.} 

8323: Progress in Mathematics, {\bf151}, 

8324: Birkh\"auser Verlag, Basel, 1997, viii+131pp., 

8325: {\bf Math.\ Rev.:}\.98k:58032. 

8326: 

8327: %\entry{Iv-1}{Ivashkovich, S.:}{Extension of analytic objects by the method

8328: %of Cartan-Thullen.}In Proceedings of Conference on Complex Analysis and Math.

8329: %Phys., 53--61, Krasnojarsk, (1988).

8330: 

8331: %\entry{Iv-2}{Ivashkovich, S.:}{The Hartogs-type extension theorem for

8332: %meromorphic maps into compact K\"ahler manifolds.} Invent. Math. {\bf109},

8333: %47--54 (1992).

8334: 

8335: \entry{Iv-Sh-1}{Ivashkovich, S., Shevchishin, V.:}{Pseudoholomorphic curves

8336: and envelopes of meromorphy of two-spheres in $\cc\pp^2$.} Preprint, Bochum

8337: (1995); available as e-print {\tt ArXiv:math.CV/9804014}.

8338: 

8339: \entry{Iv-Sh-2}{Ivashkovich, S., Shevchishin, V.:}%

8340: {Deformations of noncompact complex curves, and envelopes of meromorphy of 

8341: spheres.} (Russian) Mat.\ Sb.\ {\bf189}(1998), No.9, 

8342: 23--60; translation in Sb.\. Math.\ {\bf189}(1998), No.9--10, 1335--1359.

8343: 

8344: 

8345: \entry{Iv-Sh-3}{Ivashkovich, S., Shevchishin, V.:}{Gromov compactness theorem

8346: for stable curves.} Preprint, Bochum, (1999); available at 

8347: {\tt ArXiv:math.DG/9903047}, to appear in Int.\ Math.\ Res.\ Notes.

8348: 

8349: \entry{Ko-No}{Kobayashi, S., Nomizu, K.:}{Foundations of differential

8350: geometry.} Vol.II, Interscience Publishers, (1969).

8351: 

8352: \entry{K}{Kontsevich M.:}{Enumeration of rational curves via torus actions.}

8353: Proc.\ Conf.\ ``The moduli spaces of curves'' on Texel Island, Netherland.

8354: Birkh\"auser Prog.\ Math., {\bf129}(1995), 335--368.

8355: 

8356: \entry{K-M}{Kontsevich M., Manin Yu.:}{Gromov-Witten classes, quantum 

8357: cohomology, and enumerative geometry.} Comm.\ Math.\ Phys., {\bf164}(1994),

8358: 525--562.

8359: 

8360: 

8361: \entry{Knu}%

8362: {Knudsen, Finn F.:}%

8363: {The projectivity of the moduli space of stable curves}

8364: {\sl I.:} Math.\ Scand., {\bf39} (1976), No. 1, 19--55; 

8365: {\bf Math.\ Rev.:}\.55\#10465; \ 

8366: {\sl II.:} Math.\ Scand., {\bf52}(1983), No. 2, 161--199. 

8367: {\bf Math.\ Rev.:}\.85d:14038a; \ 

8368: {\sl II.:} Math.\ Scand., {\bf52}(1983), No. 2, 200--212.

8369: {\bf Math.\ Rev.:}\.85d:14038b. \ 

8370: 

8371: \entry{La-McD}{Lalond, F., Mcduff, D.:}{The classification of ruled

8372: symplectic manifolds.} Math.\ Res.\ Lett., {\bf3}, 769--778, (1996).

8373: 

8374: \entry{Li-Liu}{Li, T.-J., Liu, A.-K.:}{Symplectic structure on ruled surfaces

8375: and a generalized adjunctionformula.} Math.\ Res.\ Lett., {\bf2}, 453--471,

8376: (1995).

8377: 

8378: \entry{Li-T}%

8379: {Li, Jun; Tian, Gang}%

8380: {Virtual moduli cycles and Gromov-Witten invariants of general symplectic 

8381: manifolds.} 

8382: Topics in symplectic $4$-manifolds, (Irvine, CA, 1996), 47--83, 

8383: First Int. Press Lect. Ser., I, 

8384: Internat. Press, Cambridge, MA, 1998,

8385: {\bf Math.\ Rev.:}2000d:53137.

8386: 

8387: 

8388: \entry{Liu}{Liu, A.-K.:}{Some new applications of general wall crossing

8389: formula, Gompf's conjecture and its applications.}

8390: Math.\ Res.\ Lett., {\bf3}, 569--585, (1996).

8391: 

8392: 

8393: \entry{Lich}{Lichnerovicz, A:}{Global theory of connections and holonomy

8394: groups.} Noordhoff Int. Publishing, Leiden, (1976).

8395: 

8396: \entry{Mi-Wh}{Micallef, M., White, B.:}{The structure of branch points in

8397: minimal surfaces and in pseudoholomorphic curves.} Ann.\ Math., {\bf 139},

8398: 35-85 (1994).

8399: 

8400: \entry{Mil}{Milnor, J.:}{Singular points of complex hypersurfaces.} Annals of 

8401: Mathematics Studies., {\bf61},. Princeton, N.J.: Princeton University Press and the

8402: University of Tokyo Press. 122 p., (1968). 

8403: 

8404: \entry{McD-1}{McDuff, D.:}{The local behavior of holomorphic curves  in

8405: almost complex 4-manifolds.} J.Diff.Geom. {\bf 34}(1991), 143-164.

8406: 

8407: \entry{McD-2}{McDuff, D.:}{Examples of symplectic structures.} Invent.\

8408: Math., {\bf 89}(1987), 13--36.

8409: 

8410: \entry{McD-3}{McDuff, D.:}{Singularities and positivity of intersections of

8411: $J$-holomorphic curves.} In {\it``Holomorphic curves in symplectic

8412: geometry".} Ed. by M.~Audin and J.~Lafontaine, Birkh\"auser, (1994).

8413: 

8414: \entry{McD-4}{McDuff, D.:}{The structure of rational and ruled

8415: symplectic manifolds.} J.\  AMS, {\bf3}(1990), 679--712; Erratum:

8416: J.\  AMS, {\bf5}(1992), 987--988,.

8417: 

8418: \entry{McD-5}{McDuff, D.:}{Blow ups and symplectic embeddings in dimension 4.}

8419: Topology, {\bf30}(1991), 409-421.

8420: 

8421: \entry{McD-Sa-1}{McDuff, D., Salamon, D.:}{Introduction to symplectic

8422: topology.} Clarendon Press, Oxford, 425+viii p., (1995).

8423: 

8424: \entry{McD-Sa-2}{McDuff, D., Salamon, D.:}{Introduction to symplectic

8425: topology.} {\it 2nd edition}, (1998).

8426: 

8427: \entry{McD-Sa-3}{McDuff, D., Salamon, D.:}{A survey of symplectic

8428: $4$-manifolds with $b_+ =1$.} Turk.\ J.\ Math. {\bf20}, 47--60, (1996).

8429: 

8430: \entry{Mo}{Morrey, C.:}{Multiple integrals in the calculus of variations.}

8431: Springer Verlag, (1966).

8432: 

8433: \entry{MFK}%

8434: {Mumford, D., Fogarty, J., Kirwan, F.:}%

8435: {Geometric invariant theory,} 

8436: third edition, Springer-Verlag, Berlin, 1994. xiv+292 pp.,

8437: ISBN 3-540-56963-4, 

8438: {\bf Math.\ Rev.:}\.95m:14012 

8439: 

8440: 

8441: \entry{N-1}{Nemirovsky S.:}{Stein domains on algebraic surfaces.} Russian

8442: Math.\ Notes, {\bf60}(1996), 295--298 (1996).

8443: 

8444: \entry{N-2}{Nemirovsky S.:}{Holomorphic functions and imbedded real surfaces.}

8445: Math.\ Notes, {\bf63}(1998), 527--532.   

8446: 

8447: \entry{Pa-Wo}{Parker, Th.\ H., Wolfson, J.\ G.:}{Pseudo-holomorphic maps

8448: and bubble trees.} J.~Geom.\ Anal., {\bf3}, 63-98, (1993).

8449: 

8450: \entry{Ra}{Ramis J.-P.:}{Sous-ensembles analytiques d'une vari\'et\'e

8451: banachique complex.} Springer, Berlin (1970).

8452: 

8453: %\entry{Rm}{Remmert, R.:}{Holomorphe und meromorphe Abbildungen komplexer

8454: %R\"aume.} Math.\ Ann., {\bf133}, 328--370 (1957).

8455: 

8456: \entry{Rf}{Rolfsen, D.:}{Knots and links.} 2nd print, Math.\ Lect.\  Series, 

8457: {\bf N7}, xiv+439pp., Publish or Perish, (1990).

8458: 

8459: %\entry{R}{Rosay J.-P.:}{An example to the question of Chirka.} Preprint.

8460: 

8461: \entry{Ru}%

8462: {Ruan, Yongbin}%

8463: {Virtual neighborhoods and pseudo-holomorphic curves.}

8464: Proceedings of 6th G\"okova Geometry-Topology Conference. 

8465: Turkish J.\ Math., {\bf23}(1999), 161--231, 

8466: {\bf Math.\ Rev.:} CMP 1\.701\.645 (2000:03).

8467: 

8468: \entry{S-U}{Sacks J., Uhlenbeck K.:}{Existence of minimal immersions of

8469: two-spheres} Annal.\ Math., {\bf113}(1981), 1--24.

8470: 

8471: \entry{Sk-1}{Sikorav, J.-C.:}{Some properties of holomorphic curves in almost

8472: complex manifolds.} In \it ``Holomorphic curves in symplectic geometry .''

8473: \rm Ed. by M.~Audin and J.~Lafontaine, Birkh\"auser, (1994).

8474: 

8475: \entry{Sk-2}{Sikorav, J.-C.:}{Singularities of $J$-holomorphic curves.}

8476: Math.\ Z., {\bf226}, 359--373, (1997).

8477: 

8478: \entry{Sk-3}{Sikorav, J.-C.:}{The gluing construction for normally generic 

8479: $J$-holomorphic curves.} Preprint No.\.UMPA--2000-n$^\circ$264.

8480: 

8481: \entry{Sie}%

8482: {Siebert, Bernd}%

8483: {Symplectic Gromov-Witten invariants.}

8484: New trends in algebraic geometry, (Warwick, 1996), 375--424, 

8485: London Math. Soc. Lecture Note Ser., {\bf264}, 

8486: Cambridge Univ. Press, Cambridge, 1999, 

8487: {\bf Math.\ Rev.:\.}1 714 832 (2000:02). 

8488: 

8489: 

8490: %\entry{Si-1}{Siu, Y.-T.:}{Extension of meromorphic maps into K\"ahler

8491: %manifolds.} Annals of Math. {\bf102}, 421--462 (1975).

8492: 

8493: %\entry{Si-2}{Siu, Y.-T.:}{Every Stein subvariety admits a Stein

8494: %neighborhood.} Invent.math. {\bf38}, 89--100 (1976).

8495: 

8496: \entry{Sh}{Shafarevich I.:}{Basic algebraic geometry. Second edition.}

8497: Springer-Verlag.

8498: 

8499: %\entry{Shch}{Shcherbina, N.:}{On the~polynomial hull of a two-dimensional

8500: %sphere in $\cc^2$.} Soviet Math.\ Doklady, {\bf49}, 629--632, (1991).

8501: 

8502: \entry{Wn}{Weinstein, A.:}{Lectures on symplectic manifolds.} CBMS. Reg.

8503: Conf. Series, {\bf N29}, AMS(1977).

8504: 

8505: \entry{Wa}{Walker, R.:}{Algebraic curves.} Springer Verlag, (1978).

8506: 

8507: \entry{Ye}%

8508: {Ye, Rugang}%

8509: {Gromov's compactness theorem for pseudo holomorphic curves.} 

8510: Trans.\ AMS, {\bf342}(1994), 671--694,

8511: {\bf Math.\ Rev.:}\.94f:58030. 

8512: 

8513: 

8514: 

8515: %%%

8516: \end{thebibliography}

8517: 

8518: %%% Local Variables: 

8519: %%% mode: latex

8520: %%% TeX-master: "iso"

8521: %%% End: 

8522: 

8523: 

8524: 

8525: \newpage

8526: \tableofcontents

8527: 

8528: %\printindex

8529: \end{document}

8530: 

8531: \font\sevensl=cmsl7 %cmsl10 scaled 694

8532: \font\fivesl=cmsl5 %cmsl10 scaled 482

8533: 

8534: 

8535: % LocalWords: resp eps

8536: 

8537: %%% Local Variables: 

8538: %%% mode: latex

8539: %%% TeX-master: t

8540: %%% End: 

8541: