1: 

2: \documentclass[leqno,10pt]

3: {article}

4: \usepackage{amsmath, amsfonts, amssymb}

5: \usepackage{psfig}

6: 

7:  \def\firstpage{1}\def\lastpage{1000}

8: %

9: 

10: %\makeatletter

11: %\def\magnification{\afterassignment\m@g\count@}

12: %\def\m@g{\mag=\count@\hsize6.5truein\vsize8.9truein\dimen\footins8truein}

13: %\makeatother

14: 

15: %%% Choose 10pt/12pt:

16: \oddsidemargin1.66cm\evensidemargin1.66cm\voffset1.2cm%10pt

17: %\magnification1200\oddsidemargin.41cm\evensidemargin.41cm\voffset-.75cm%12pt

18: 

19: \textwidth13.0cm\textheight19.5cm

20: 

21: 

22: \pagestyle{myheadings}

23: \pagenumbering{arabic}

24: \setcounter{page}{\firstpage}

25: 

26:  \def\sign{\mathrm {sign}}

27: \def\Fc{{\mathcal F}}

28: \def\Pc{{\mathcal P}}

29: \def\Cc{\mathcal{C}}

30: \def\a{\alpha}

31: \def\ora{\overrightarrow}

32: \def\ola{\overleftarrow}

33: \def\g{\gamma}

34: \def\d{\delta}

35: \newcommand{\up}{{}^}

36: \newcommand{\ointw}{\int\limits_{|w|=1}}

37: \newcommand\hook{\mathbin{\hbox{\vrule height .5pt width 3.5pt depth 0pt

38: \vrule height 6pt width .5pt depth 0pt}}}

39: \def\ThetaH{\Theta^{H_2}}

40: \newcommand{\qed}{\hfill {$\blacktriangleright$}}

41: \newcommand{\fW}{{\mathfrak W}}

42: \newcommand{\fA}{{\mathfrak A}}

43: \newcommand{\fP}{{\mathfrak P}}

44: \newcommand{\fK}{\mathfrak {K}}

45: \newcommand{\fF}{{\mathfrak F}}

46: \newcommand{\fD}{{\mathfrak D}}

47: \newcommand{\fL}{{\mathfrak L}}

48: \newcommand{\fH}{{\mathfrak H}}

49: \newcommand{\fE}{{\mathfrak E}}

50: \newcommand{\NOV}{{\mathrm {NOV}}}

51: \newcommand{\cont}{{\mathrm {cont}}}

52: \newcommand{\restr}{\mathrm {restr}}

53:  \newcommand{\bx}{{\mathbf x}}

54: \newcommand{\by}{{\mathbf y}}

55: \newcommand{\bz}{{\mathbf z}}

56: \newcommand{\bgg}{{\mathbf g}}

57: \newcommand{\bff}{{\mathbf f}}

58: \newcommand{\bg}{\boldsymbol{\gamma}}

59: \newcommand{\bo}{\boldsymbol{\omega}}

60: \newcommand{\bmu}{\boldsymbol{\mu}}

61: \newcommand{\bphi}{{\mathbf \phi}}

62: \newcommand{\bA}{{\mathbf A}}

63: \newcommand {\bE}{\mathbf{E}}

64: \newcommand{\bC}{{\mathbf C}}

65: \newcommand{\bh}{{\mathbf h}}

66: \newcommand{\bF}{{\mathbf F}}

67: \newcommand{\bG}{{\mathbf G}}

68: \newcommand{\bt}{{\mathbf t}}

69: \newcommand{\bH}{{\mathbf H}}

70: \newcommand{\bK}{{\mathbf K}}

71: \newcommand{\bk}{{\mathbf k}}

72: \newcommand{\bS}{{\mathbf S}}

73: \newcommand{\bV}{ \mathbf {V}}

74:  \newcommand{\CZ}{{\mathrm {CZ}}}

75:  \newcommand{\SFT}{{\mathrm {SFT}}}

76:  \newcommand{\RSFT}{{\mathrm {RSFT}}}

77: \newcommand{\bPhi}{\boldsymbol{\Phi}}

78: \newcommand{\bPsi}{\boldsymbol{\Psi}}

79: \newcommand{\bXi}{\boldsymbol{\Xi}}

80: \newcommand{\bTheta}{\boldsymbol{\Theta}}

81: \newcommand{\btheta}{\boldsymbol{\theta}}

82: \newcommand{\bd}{\boldsymbol{\delta}}

83: \newcommand{\wt}{\widetilde}

84: \newcommand{\cT}{{\mathcal T}}

85: \newcommand{\Mc}{\mathcal {M}}

86: \newcommand{\cM}{\mathcal {M}}

87: \newcommand{\cD}{\mathcal {D}}

88: \newcommand{\cS}{{\mathcal S}}

89: \newcommand{\cK}{{\mathcal K}}

90: \newcommand{\cP}{{\mathcal P}}

91: \newcommand{\cE}{{\mathcal E}}

92: \newcommand{\cF}{{\mathcal F}}

93: \newcommand{\ch}{{\mathcal h}}

94: 

95: \newcommand{\CR}{\mathcal{CR}}

96: \newcommand{\mc}{{\mathcal c}}

97: \newcommand{\ve}{\varepsilon}

98: \newcommand{\Ker}{{\mathrm {Ker}}}

99: %\newcommand{\Im}{{\rm Im}}

100: \newcommand{\cV}{{\mathcal V}}

101: \newcommand{\cN}{{\mathcal N}}

102: \newcommand{\cW}{{\mathcal W}}

103:  \newcommand{\dist}{{\mathrm {dist}}}

104: \newcommand{\Coker}{{\mathrm {Coker}}}

105: \newcommand{\Morse}{{\mathrm {Morse}}}

106: \newcommand{\Int}{{\mathrm {Int}}}

107: %\newcommand{\Ker}{{\mathrm {Ker}}}

108: \newcommand{\Id}{{\mathrm {Id}}}

109: \newcommand{\ind}{\mathrm{ind\,}}

110: 

111: \newcommand{\im}{\mathrm {Im}}

112: \newcommand{\comp}{\mathrm {comp}}

113: \newcommand{\D}{{\rm D}}

114: \newcommand{\TT}{T\Theta_+\Theta_-}

115: \newcommand{\TTH}{\TT^{H_2}}

116: \newcommand{\contr}{{\mathrm {contr}}}

117: \newcommand{\punct}{{\mathrm {punct}}}

118: \newcommand{\even}{{\mathrm {even}}}

119: \newcommand{\cl}{{\mathrm {closed}}}

120:  \newcommand{\proof}{{\sl Proof. }}

121: %

122: \newcommand{\1}{{{\mathchoice {\rm 1\mskip-4mu l} {\rm 1\mskip-4mu l}

123: {\rm 1\mskip-4.5mu l} {\rm 1\mskip-5mu l}}}}

124: \newcommand{\N}{{\mathbb N}}

125: \newcommand{\Z}{{\mathbb Z}}

126: \newcommand{\R}{{\mathbb R}}

127: \newcommand{\Q}{\mathbb {Q}\,}

128: \newcommand{\C}{\mathbb {C}\,}

129: \newcommand{\const}{{\mathrm const}}

130: %

131: \newcommand{\Om}{{\Omega}}

132: \newcommand{\om}{{\omega}}

133: \newcommand{\eps}{{\varepsilon}}

134: \newcommand{\si}{{\sigma}}

135: \newcommand{\ups}{{\upsilon}}

136: \renewcommand{\i}{{\iota}}

137: \newcommand{\p}{{\partial}}

138: \newcommand{\ba}{{\mathbf a}}

139: \newcommand{\bb}{{\mathbf b}}

140: \newcommand{\bc}{{\mathbf c}}

141:  \newcommand{\bX}{{\mathbf X}}

142:  \newcommand{\bY}{{\mathbf Y}}

143:  \newcommand{\DM}{{\overline{\cM}}}

144:    \newcommand{\pr}{\mathrm{pr}}

145:  \newcommand{\ct}{\mathrm{ct}}

146:  \newcommand{\ev}{\operatorname{ev}}

147:  \newcommand{\disk}{\mathrm{disk}}

148: \newcommand{\gt}{{\tau}}

149: \newcommand{\go}{{\omega}}

150: \newcommand{\gd}{{\delta}}

151: \newcommand{\gb}{{\beta}}

152: \newcommand{\w}{{\wedge}}

153: \newcommand{\bs}{ \mathbf {s}}

154: %

155: 

156: 

157:  \newtheorem{theorem}{Theorem}[subsection]

158: \newtheorem{corollary}[theorem]{Corollary}

159: \newtheorem{lemma}[theorem]{Lemma}

160: \newtheorem{proposition}[theorem]{Proposition}

161: \newtheorem{conjecture}[theorem]{Conjecture}

162: \newtheorem{question}[theorem]{Question}

163: \newtheorem{problem}[theorem]{Problem}

164: \newtheorem{criterion}[theorem]{Criterion}

165: \newtheorem{summary}[theorem]{Summary}

166: \newtheorem{remark}[theorem]{Remark}

167: \newtheorem{definition}[theorem]{Definition}

168: 

169: %\theoremstyle{definition}

170: 

171: 

172: \newtheorem{examples}[theorem]{Examples}

173: \newtheorem{example}[theorem]{Example}

174: \newtheorem{exercise}[theorem]{Exercise}

175: 

176: \newtheorem{generalization}[theorem]{Generalization}

177: %\numberwithin{section}{chapter}

178: %$\numberwithin{equation}{chapter}

179: %\numberwithin{figure}{chapter}

180: \begin{document}

181: %%%%% ------------- fill in your data below this line  -------------------

182: %%%%%    The following lines \Title ... \EndAddress must ALL be present

183: %%%%%    and in the given order.

184: \title{Introduction to Symplectic Field Theory}

185: \author{Yakov Eliashberg\\Stanford

186: University \and Alexander Givental \\  UC  Berkeley

187: \and Helmut Hofer\\New York University}

188: \date{October 2000}

189:  \maketitle

190: 

191:   %

192: \begin{abstract}

193: 

194: We sketch in this article  a new theory, which we call {\sl

195: Symplectic Field Theory } or SFT,  which provides an approach to

196: Gromov-Witten invariants of symplectic manifolds and their

197: Lagrangian submanifolds in the spirit of topological field theory,

198: and at the same time

199:  serves as a rich source

200: of new invariants of contact manifolds and their Legendrian

201: submanifolds. Moreover, we  hope that the applications of  SFT

202: go far beyond this framework.\footnote{The research is   partially

203: supported by the National Science Foundation.}

204: \end{abstract}

205: 

206:  \tableofcontents

207:  \bigskip

208: 

209:  \centerline{\bf Disclosure}

210:  \medskip

211: 

212: Despite its length, the current paper presents only a very sketchy

213: overview of  Symplectic Field Theory. It contains practically no proofs,

214: and in a few places where the proofs  are given

215: their role is    just to

216: illustrate the involved ideas,  rather than to give  complete rigorous

217: arguments.

218: 

219:  The ideas, the algebraic formalism, and some of the applications

220:  of this new theory were presented  and popularized

221:  by the authors at several conferences

222:  and seminars (e.g. \cite{Eliash-ICM}). As a result,

223:    currently there exists a  significant

224:  mathematical community which is in some form familiar with

225:    the subject.

226:   Moreover, there are

227:   many mathematicians, including several former

228:    and current students of the authors,

229:   who are actively working  on foundational aspects

230:   of the theory and its applications, and    even

231:   published papers on this subject. Their results

232:     show that already the simplest versions of the theory

233:   have some remarkable corollaries (cf. \cite{Ustilovsky}).

234:    We hope that   the present  paper will help

235:    attracting even more people

236:    to   SFT.

237: 

238:    Of course, our ideas give just a small new twist

239:     to many other active directions of research in

240:     Mathematics and Physics

241:     (Symplectic topology, Gromov-Witten invariants and quantum cohomology,

242:     Floer homology theory, String theory,

243:      just to mention few), pioneered by  V.I. Arnold,

244:      C.~Conley-E.~Zehnder,

245:       M.~Gromov, S.K. Donaldson, E.~Witten, A.~Floer,

246:         M.~Kontsevich and others

247:      (see \cite{Arnold-conjecture, Conley-Zehnder,

248:      Gromov-holomorphic, Donaldson-invariants, Floer, Ruan, Witten1, Witten2, Kontsevich-CP2,

249:      Konts-Manin}).

250:      Many people independently contributed results and ideas,

251:       which may be considered as  parts of  SFT.

252:       Let us just mention here the work of  Yu.Chekanov

253:       \cite{Chekanov},

254:        K.Fukaya--K.Ono--Y.-G.Oh--H.Ohta  \cite{FKO3},

255:        A.~Gathmann  \cite{Gathmann}, E.Ionel--T.Parker

256:       \cite{Ionel-Parker, Ionel}  , Y.Ruan--A.-M.Li  \cite{Ruan-Li}.

257:       It also  draws on  other results of   the  current authors

258:       and their coauthors (see \cite{EHS, EH-3ball, Givental, Givental-Maslov,

259:       Givental-toric, Hofer-Weinstein, Hofer-Wysocki-Zehnder, AbbasHofer,

260:       HoferWysockiZehnder:94d, HoferWysockiZehnder:99}). The contact-geometric

261:       ingredient of our work is greatly motivated by two

262:       outstanding conjectures in contact geometry: Weinstein's conjecture about periodic

263:       orbits of Reeb fields  \cite{Weinstein},

264:        and  Arnold's chord conjecture \cite{Arnold-steps}.

265: 

266: 

267:    Presently, we are working on a series of papers devoted

268:    to the foundations,

269:    applications,  and further development of  SFT.  Among the applications,

270:    some of which are mentioned in this paper, are new invariants of contact manifolds

271:    and Legendrian knots and links, new methods for computing Gromov-Witten invariants,

272:    new restrictions on the topology of Lagrangian submanifolds, new non-squeezing type

273:    theorems in contact geometry etc. We are expecting new links with the low-dimensional topology

274:    and, possibly, Physics. It seems,

275:    however,

276:    that what we see at the moment is just a tip of an iceberg.

277:    The main   body of Symplectic Field Theory  and its applications is yet to be discovered.

278: 

279:      \bigskip

280:      \noindent{\bf Guide for an impatient reader.}

281:      The paper consists of two parts. The first part, except Section \ref{sec:Floer}

282:      and the end of

283:        Section \ref{sec:orientation},

284:      contains some  necessary   background symplectic-geometric and analytic information.

285:      An impatient reader can try to begin reading with Section

286:      \ref{sec:Floer},and use

287:      the rest of the first part for  the references.

288:      The second part begins with its own introduction

289:      (Section \ref{sec:informal})  where we present a very rough sketch of SFT. At the end of Section

290:          \ref{sec:informal} we describe the plan of the remainder of the paper.

291: 

292:    \bigskip

293:    \noindent{\bf Acknowledgements.} The authors benefited a lot from discussions

294:     with many mathematicians, and from the ideas which they corresponded to

295:     us. We are especially grateful to C. Abbas,  P. Biran, F. Bourgeois, K. Cieliebak,

296:     T. Coates, T. Ekholm, K. Fukaya, E. Getzler,

297:     M. Hutchings, E.-N.

298:     Ionel, V.M. Kharlamov,

299:     K. Mohnke, L. Polterovich, D. Salamon, M. Schwarz, K. Wysocki and E. Zehnder.

300:       A  part of the paper

301:     was written,  when  the first author visited

302:     RIMS at Kyoto University.

303:      He wants  to thank  RIMS  and  K. Fukaya and K. Saito,

304:     the organizers of a special program

305:     in Geometry and String Theory,   for the  hospitality. He also thanks

306:     T. Tsuboi and J.-L. Brylinski for organizing   cycle of lectures on  SFT at

307:      the    University of

308:     Tokyo and the Pennsylvania State University. The second author is thankful to

309:      A. Kirillov, R. Donagi, and

310:      the Department of Mathematics of the University

311:     of Pennsylvania  for the possibility to present   SFT in a series of lectures.

312:      The third author acknowledges the hospitality of   FIM at the ETH Z\"urich, where   some

313:      of the work was

314: carried

315: out. Our special thanks to N.M. Mishachev for drawing the

316:     pictures and to J. Sabloff for

317:     the computer verification of the formula

318:     (\ref{eq:CP2-final}).

319: 

320: 

321: \section{Symplectic  and analytic setup}

322: 

323: 

324: \subsection{Contact preliminaries}\label{sec:prelim}

325: A $1$-form $\a$ on a $(2n-1)$-dimensional manifold $V$

326: is called {\it contact} if the

327: restriction of $d\a$ to  the    $(2n-2)$-dimensional tangent

328: distribution $\xi =

329: \{\a = 0\}$ is non-degenerate (and hence symplectic).

330: A  codimension $1$ tangent distribution $\xi$ on $V$ is called

331:  a {\it contact structure} if

332: it can be locally (and in the co-orientable case globally) defined

333: by  the Pfaffian equation $\a

334: = 0$ for some choice of a contact form $\a$.  The pair  $(V,\xi)$   is called

335: a {\it contact manifold}.  According to  Frobenius' theorem the contact

336: condition is a condition of maximal

337:  non-integrability of the tangent hyperplane field $\xi$. In particular, all integral

338: submanifolds of $\xi$ have

339: dimension $\le n-1$. On the other hand, $(n-1)$-dimensional integral submanifolds, called

340: {\it Legendrian}, always exist in abundance.    We will be dealing in this paper

341: only with co-orientable, and moreover co-oriented contact structures.

342: Any non-coorientable contact structure can be canonically double-covered

343:  by a coorientable one.

344: If a contact form $\a$ is fixed then one can associate with it the {\it Reeb vector field}

345: $R_\a$, which is transversal

346: to the contact structure $\xi=\{\alpha=0\}$.

347: The field $R_\a$ is uniquely determined by the equations

348: $R_\a\hook d\a=0;\,\a(R_\a)=1\,.$   The flow of $R_\a$ preserves the contact form $\a$.

349: 

350:  The  $2n$-dimensional manifold $M=(T(V)/\xi)^*  \setminus V$ ,  called the

351: {\it symplectization} of $(V,\xi)$, carries a natural

352: symplectic structure $\omega$ induced by an embedding $M\to  T^*(V)$

353:  which assigns to each linear form $T(V) / \xi \to \R$ the

354: corresponding form $T(V) \to T(V)/\xi \to \R$. A choice of a contact

355: form $\a$ (if $\xi$ is co-orientable) defines a splitting $M=V\times(\R\setminus 0)$. As     $\xi$ is

356:  assumed  to be co-oriented  we can

357:  pick

358:  the positive half  $V\times\R_+$ of  $M$, and call it symplectization

359: as well.  The symplectic structure $\omega$ can be written in terms of this splitting as

360: $d(\tau\a), \tau>0$. It will be more convenient for us, however, to use additive notation

361: and write $\omega$ as $d(e^t\a),\,t\in\R$, on $M=V\times\R$.

362: Notice that the vector field $T=\frac{\p}{\p t}$ is conformally symplectic:

363:  we have ${\cal L}_T\omega=\omega$, as well as ${\cal L}_T

364:  (e^t\a)=e^t\a$,

365:  where ${\cal L}_T$ denotes the Lie derivative along the vector

366:  field $T$.

367: All the notions of contact geometry can be formulated as the

368: corresponding symplectic notions, invariant or equivariant with

369: respect to this conformal action. For instance, any contact

370: diffeomorphism of $V$ lifts to an equivariant symplectomorphism of

371: $M$;  contact vector fields on $V$ (i.e. vector fields preserving

372: the contact structure) are   projections of $\R$-invariant

373: symplectic (and automatically Hamiltonian) vector fields on $M$;

374: Legendrian submanifolds in $M$ correspond to cylindrical (i.e.

375: invariant with respect to the $\R$-action) Lagrangian submanifolds

376: of $M$.

377: 

378:  Notice that the Hamiltonian vector field on $V\times\R$, defined by the Hamiltonian function

379: $H=e^t$ is invariant under translations $t\mapsto t+c$, and projects to the Reeb vector field

380: $R_\alpha$ under the projection $V\times\R\to \R$.

381: 

382: The symplectization of a contact manifold is an example of a

383: symplectic manifold with {\it cylindrical} (or rather conical)

384: ends. We mean by that    a possibly non-compact symplectic

385: manifold $(W,\omega)$ with    ends of the form

386: $E^+=V^+\times[0,\infty)$ and  $E^-=V^-\times(-\infty,0]$, such

387: that  $V^\pm$ are compact manifolds, and

388: $\omega|_{V^\pm}=d(e^t\alpha^\pm)$, where $\a^\pm$ are  contact

389: forms on $V^\pm$. In other words,  the ends  $E^\pm$ of

390: $(W,\omega)$ are symplectomorphic, respectively,

391:   to the positive or negative halves of the

392: symplectizations of    contact manifolds $(V^\pm,\xi^\pm=\{\a^\pm=0\})$.

393: We will consider   the splitting of the ends and the  the contact forms

394: $\a^\pm$ to be parts of the structure of a symplectic manifold with cylindrical ends.

395: We will also call

396: $(W,\omega)$ a {\it directed symplectic cobordism} between

397:  the contact

398: manifolds $(V^+,\xi^+)$ and $(V^-,\xi^-)$, and denote it  by $\ora{V^-V^+}$.

399: 

400: Sometimes we will have to consider the compact part

401: $W^0=W\setminus(\Int E^+\cup\Int E^-)$ of a directed symplectic

402: cobordism $\ora{V^-V^+}$.  If it is not clear from the context we

403: will refer to $W^0$   as a {\it compact}, and to $W$ as a {\it

404: completed} symplectic cobordism.

405: 

406: Let us point out that ``symplectic cobordism" {\it is not an

407: equivalence relation, but rather a partial order. }

408: Existence of a directed

409: symplectic cobordism $\ora{V^-V^+}$  does not imply

410: the existence of a directed symplectic cobordism   $\ora{V^+V^-}$,

411: even if one does not fix contact forms for the contact structures $\xi^\pm$. On the other

412: hand,

413: directed

414: symplectic cobordisms $\ora{V_0V_1}$ and $\ora{V_1V_2}$ can be glued, in an obvious

415: way, into a directed symplectic cobordism  $\ora{V_0V_2}=\ora{V_0V_1}

416: \circledcirc\ora{V_1V_2}$.

417: 

418: 

419: 

420: Contact  structures have no local invariants. Moreover, any contact form is locally

421: isomorphic to the

422: form $\alpha_0= dz-\sum\limits_1^{n-1}y_idx_i$ (Darboux' normal form).

423: The contact structure $\xi_0$ on $\R^{2n-1}$ given by the form $\a_0$ is called {\it

424: standard}.

425:  The standard contact structure on $S^{2n-1}$ is formed by complex tangent hyperplanes to the

426: unit sphere in $\C^n$. The standard contact structure on $S^{2n-1}$

427: is isomorphic in the complement of a point to the standard contact structure  on

428: $\R^{2n-1}$.

429: According to a theorem of J. Gray (see \cite{Gray}) contact structures

430:  on closed

431: manifolds   have  the following stability property:

432: {\it Given a family  $\xi_t$, $t\in[0,1]$, of contact  structures on  a closed manifold

433: $M$, there exists

434: an isotopy $f_t:M\to M$,

435: such that $df_t(\xi_0)=\xi_t; t\in [0,1]$.}

436: Notice that for contact {\it forms} the analogous statement is  wrong. For

437: instance,

438: the topology of the $1$-dimensional foliation determined by the  Reeb vector field $R_\a$

439: is very sensitive to deformations of the contact form $\a$.

440: 

441: The conformal class of the symplectic form $d\a|_\xi$ depends only on

442: the cooriented contact structure $\xi$ and not on the

443: choice of the contact form $\a$. In particular, one can associate with $\xi$

444:  an almost complex structure $J:\xi\to\xi$, compatible with $d\a$

445: which means that $d\alpha(X,JY); X,Y\in\xi,$ is an Hermitian metric on $\xi$.

446: The space of   almost complex structures $J$ with this property is contractible, and hence

447:  the choice of $J$ is homotopically canonical.

448: Thus a  co-oriented contact structure  $\xi$ defines on $M$ a {\it

449:  stable almost complex

450: structure}

451:  $\widetilde J=\widetilde J_\xi$,

452: i.e. a splitting of

453: the tangent bundle $T(V)$ into the Whitney sum of a complex bundle of (complex) dimension

454: $(n-1)$ and a trivial $1$-dimensional real bundle.

455: The existence of a

456:  stable almost complex

457: structure  is  necessary for  the existence of a contact structure on $V$. If $V$

458: is open (see \cite {Gro-PDR}) or ${\mathrm {dim}}\,V=3$ (see \cite{Martinet,Lutz})

459:  this property is also sufficient for  the existence of a contact structure in the

460: prescribed homotopy

461: class. It  is still unknown

462: whether this condition

463: is sufficient for  the existence of a contact structure on a closed manifold of dimension

464: $>3$. However,   a positive answer  to this question is extremely unlikely.

465: The homotopy class of $\widetilde J_\xi$, which we denote by

466: $[\xi]$ and call the {\it formal }homotopy class  of $\xi$,

467: serves as an

468: invariant of $\xi$.

469: For an open $V$ it is a complete invariant (see \cite {Gro-PDR})   up to   homotopy

470: of contact structures,

471: but not up to a contact diffeomorphism. For closed manifolds this is known to be false

472: in many, but not all dimensions. The theory discussed in this

473: paper serves as a rich source of contact invariants, both of

474: closed and open contact manifolds.

475: 

476: \subsection{Dynamics of Reeb vector fields}\label{sec:dynamics}

477: 

478: Let $(V,\xi)$ be a $(2n-1)$-dimensional manifold with  a

479: co-orientable contact structure with a fixed contact form $\alpha$.

480:  For a generic choice of $\a$ there are only countably many periodic trajectories

481:   of the

482: vector field $R_\a$.  Moreover, these trajectories can be assumed {\it non-degenerate}

483: in the sense  that the linearized Poincar\'e return map $A_\g$

484:  along any  closed

485:  trajectory $\g$, including multiples,

486: has no eigenvalues equal to $1$.  Let  us denote by $\Pc=\Pc_\a$

487:   the set of  all periodic trajectories of $R_\a$, including multiples.

488:   \footnote{As it is explained below in Section \ref{sec:orientation}

489:   the orientation issues require us to exclude certain multiple periodic orbits out

490:   of consideration. Namely, let us recall that real eigenvalues of symplectic matrices

491:   different from $\pm 1$ come in pairs $\lambda,\lambda^{-1}$.

492:  Let $\g\in \Pc$ be a simple periodic orbit and $A_\g$ its linearized Poincar\'e return map.

493: If

494:   the total multiplicity of   eigenvalues of $A_\g$ from the interval

495:   $(-1,0)$ is odd, then we   exclude from $\Pc$ all even

496:   multiples of $\g$.}

497: 

498: The reason for a such choice

499: is discussed in Section \ref{sec:orientation} below. We will also fix a point

500: $m_\gamma$ on each {\it simple} orbit from $\Pc$.

501: Non-degenerate trajectories can be divided into {\it odd} and {\it even} depending

502: on the sign of the Lefshetz number ${\mathrm {det}}  (I- A_\g)$.

503: Namely, we call $\g$ odd if $  {\mathrm {det}}  (I-  A_\g)<0$, and even otherwise.

504: The parity of  a periodic orbit $\g$  agrees   with the parity of a certain

505: integer grading  which is defined   if certain additional choices are made, as

506: it is described below.

507: %%%%

508: 

509: If $H_1(V)=0$

510:  then for each $\g\in\Pc$ we can choose  and

511:  fix a surface $F_{\g}$ spanning

512: the trajectory $\g$ in $V$.

513:   We will allow the case $H_1(V)\neq 0$,

514:    but will require in most of the paper that the torsion part is

515:    trivial

516:    \footnote{ The case when $H_1(V)$ has torsion elements is

517:    discussed in Section \ref{sec:torsion} below.}.

518: In this case   we choose a basis of $H_1(V)$,

519:   represent    it by oriented curves ${C}_1,\dots,{C}_K$, and    choose

520:   a symplectic trivialization of the bundle $\xi|_{{C}_i}$ for each chosen curve.

521:   We recall that the bundle $\xi$ is endowed with the symplectic

522:   form $d\alpha$ whose   conformal class depends only on $\xi$.

523:        For  any periodic orbit $\gamma\in\Pc$

524:    let us choose a surface $F_\gamma$ with $[\partial F_\gamma]=

525:    [\gamma]-\sum n_i[{C}_i]$.

526:   The coefficients $n_i$ are uniquely defined because of our  assumption that $H_1(V)$

527:   is

528:   torsion-free.

529: 

530: 

531: 

532: The above choices enable us to define the {\it Conley-Zehnder

533: index} $\CZ(\g)$ of  $\g$ as follows. Choose a homotopically

534: unique trivialization of the symplectic vector bundle $(\xi,d\a)$

535: over each trajectory $\g\in\Pc$ which extends to $\xi|_{F_{\g}}$

536: (and coincides with a chosen trivialization of $\xi|_{{C}_i}$

537: if ${C}_i$ is not homologically trivial). The linearized flow

538: of $R_\a$ along $\g$ defines then a path in the group

539: $Sp(2n-2,\R)$ of symplectic matrices, which begins at the unit

540: matrix and ends at a matrix with all eigenvalues different from

541: $1$. The Maslov index of this path (see \cite{Arnold-Maslov,

542: Robbin-Salamon}) is, by  the definition, the Conley-Zehnder index

543: $\CZ(\g)$ of the trajectory $\g$. See also

544:  \cite{HoferWysockiZehnder:94d}, Section 3, for an axiomatic

545:         description of the Conley-Zehnder index using our normalization

546:         conventions.

547: 

548: Notice that by changing the spanning surfaces for the trajectories

549: from $\Pc$ one can change Conley-Zehnder indices by  the value of

550: the cohomology class $2c_1(\xi)$, where $c_1(\xi)$ is the first

551: Chern class of the contact bundle $\xi$. In particular, ${\mathrm

552: {mod}}\,2$ indices can be defined independently of any spanning

553: surfaces, and even in the case when $H_1(V)\neq 0$. In fact,

554: $$(-1)^{\CZ(\g)}=(-1)^{n-1}\sign \left({\mathrm

555: {det}}(I-A_\gamma)\right).$$

556: 

557: 

558: 

559: 

560: 

561: 

562:   \subsection{Splitting of a symplectic manifold along \\

563:   a contact submanifold}

564: \label{sec:splitting} Let $V$ be a hypersurface of contact type,

565: or in a different terminology, a symplectically convex

566: hypersurface  in a symplectic manifold $(W,\omega)$. This means

567: that $\omega$ is exact, $\omega=d\beta$, near $V$, and the

568: restriction $\alpha=\beta|_V$ is a contact form on $V$.

569: Equivalently, one can say that the conformally symplectic vector

570: field $X$, $\omega$-dual to $\beta$, is transversal to $V$. Let us

571: assume that $V$ divides $W,\, W=W_+\cup W_-$, where

572:  the notation of the parts are chosen in such a way that   $X$

573: serves as an inward transversal for $W_+$, and an outward

574: transversal for $W_-$. The manifolds $W_\pm$ can be viewed as

575: compact directed symplectic cobordisms such that $W_-$ has only positive

576: contact boundary $(V,\alpha)$, while the same contact manifold

577: serves as a negative boundary of $W_+$.

578: 

579: Let $$(W_-^\infty,\omega_-^\infty)=(W_-,\omega)\cup\left(V\times[0,\infty),d(e^t\alpha)\right)$$ and

580: $$(W_+^\infty,\omega_+^\infty)=\left(V\times(-\infty,0],d(e^t\alpha)\right)\cup(W_+,\omega)$$

581:  be the  completions, and

582: $$(W_-^\tau,\omega_-^\tau)=(W_-,\omega)\cup\left(V\times[0,\tau],d(e^t\alpha)\right)$$ and

583: $$(W_+^\tau,\omega_+^\tau)=\left(V\times[ -\tau,0],d(e^t\alpha)\right)\cup(W_+^0,\omega)$$

584:   {\it partial completions} of

585: $W_\pm$. Let us observe that the symplectic manifolds

586: $$(W_-,e^{-\tau}\omega_-^\tau),\quad (V\times [-\tau,\tau],

587: d(e^t\alpha))\quad\hbox{and}\quad (W_+,e^\tau\omega_+^\tau)$$ fit

588: together into a symplectic manifold $(W^\tau, \omega^\tau)$, so

589: that $W^0=W$. Hence when $\tau\to\infty$ the deformation

590: $(W^\tau,\omega^\tau)$ can be viewed as a decomposition of the

591: symplectic manifold $W=W^0$ into the union   of two completed

592: symplectic cobordisms  $W^\infty_+$ and $ W^\infty_-$ . We will

593: write $W=W_-\circledcirc W_+$ and also

594: $W^\infty=W_-^\infty\circledcirc W_+^\infty$.

595: %%%%%%%%%%%%%%%%

596: 

597: \begin{figure}

598: \centerline{\psfig{figure=sf3.eps,height=60mm}}

599:  \caption{\small Splitting of a closed symplectic manifold $W$

600:  into two completed symplectic cobordisms $W_-^\infty$ and $W^\infty_+$}

601: \label{fig:splitting}

602: \end{figure}

603: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

604:  Let us give here two important examples of the above

605:  splitting construction.

606: 

607:  \begin{example}\label{ex:cotangent}

608:  {\rm Suppose $L\subset W$ is a Lagrangian submanifold.

609:   Its neighborhood  is symplectomorphic to a

610:    neighborhood of the $0$-section  in the cotangent bundle

611:    $T^*(L)$. The boundary $V$ of an  appropriately chosen neighborhood

612:    has  contact type, and thus we can apply    along $V$

613:    the  above splitting construction. As the result we split

614:    $W$ into  $W_+^\infty$ symplectomorphic

615:    to $W\setminus L$, and $W_-^\infty$ symplectomorphic to $T^*(L)$.

616:     }

617:     \end{example}

618:     \begin{example}\label{ex:section}

619:         {\rm Let $M$ be a hyperplane section of a K\"ahler manifold

620:    $W$, or

621:    more generally a symplectic hyperplane section of a

622:    symplectic manifold

623:    $W$, in the sense of Donaldson (see \cite{Donaldson}). Then $M$ has a neighborhood

624:    with a contact boundary $V$. The affine part

625:    $W\setminus M$ is a Stein manifold in the K\"ahlerian case,

626:    and in any case

627:    has a structure of   a symplectic Weinstein

628:    manifold $\widetilde W$ (notice that the symplectic structure

629:    of $\widetilde W$ does not coincide with  the induced

630:    symplectic structure

631:    on $W\setminus M$ but contains $W\setminus M$ as an

632:    open symplectic

633:    submanifold).

634:    The Weinstein manifold $W\setminus M$ contains

635:    an isotropic  deformation retract  $\Delta$. The splitting of $W$ along $V$

636:    produces $W^\infty_-$ symplectomorphic to

637:     $\widetilde W,$ and $W^\infty_+$

638:     symplectomorphic to $W\setminus \Delta$.

639:     If $\Delta$ is a smooth Lagrangian submanifold, then

640:     we could get the same decomposition by

641:     splitting along the boundary of a tubular neighborhood

642:     of $L$, as in Example \ref{ex:cotangent}.  }

643:               \end{example}

644: 

645: \subsection{Compatible almost complex structures}\label{sec:almost}

646: 

647: According to M. Gromov (see \cite{Gromov-holomorphic}) an almost

648: complex structure $J$ is called {\it tamed} by a  symplectic form

649: $\omega$ if $\omega$ is positive on complex lines. If, in

650: addition,  one adds the calibrating condition that $\omega$ is

651: $J$-invariant, then $J$ is   said {\it compatible} with $\omega$.

652: For symplectic manifolds with cylindrical ends one needs further

653: compatibility conditions at infinity, as it is described below.

654: 

655: At each positive, or negative end $\left(V\times\R_\pm,d(e^t\alpha

656: )\right)$ we require   $J$ to be invariant with respect to

657: translations $t\mapsto t \pm c$, $c>0$ at least for sufficiently

658: large $t$. We also require   the contact structure

659: $\xi^\pm|_{V\times t}$ to be invariant under $J$, and define

660: $J\frac{\partial}{\partial t}=R_\alpha$, where

661:  $R_\alpha$ is

662:  the Reeb vector field (see \ref{sec:dynamics} above)

663:   of the contact form $\alpha$. In the case when

664: $W=V\times\R$ is the symplectization of a manifold $V$, i.e. $W$ is    a

665: cylindrical manifold, we additionally  require   $J$ to be globally

666: invariant under all translations along the second factor.

667: 

668: To define a compatible almost complex structure $J$ in the above

669: Examples \ref{ex:cotangent} and \ref{ex:section}

670:  one needs to specify a contact form $\alpha$ on the contact manifold

671:  $V$. In the case of the boundary of a tubular neighborhood    of a Lagrangian

672:  submanifold $L$ a natural choice of a contact form is provided by

673:  a  Riemannian metric on $L$. The Reeb vector field for such a

674:  form $\alpha$ generates on $V$ the geodesic flow of the metric.

675: 

676:  When $V$ is the boundary of a neighborhood of a hyperplane section  $M$

677:  then there exists another good choice of a contact form. It is

678:  a $S^1$-invariant connection form $\a$ on the principal $S^1$-bundle

679:  $V\to M$, whose curvature equals the symplectic form $\omega|_M$.

680:  The contact manifold $(V,\xi=\{\a=0\})$ is called the {\it

681:  pre-quantization} of the symplectic  manifold $(M,\omega)$.

682:  Orbits of the  Reeb field $R_\a$ are all closed and coincide

683:  with the fibers of the fibration, or their multiples. Notice that though

684:  the Reeb flow in

685:  this case looks extremely nice and simple, all its  periodic orbits are

686:  highly degenerate, see Section \ref{sec:Bott} below.

687:  Notice that  the symplectization $W$  of $V$

688:  can be viewed  as the total space of a complex  line bundle

689:  $L$ associated with

690: the $S^1$-fibration $V\to M$ with the zero-section removed.

691: It is possible

692:  and convenient to choose

693: $J$  compatible with the structure of this bundle,

694: and in such a way that the projection   $W\to M$

695: becomes holomorphic with respect to a certain almost

696: complex structure $J_M$ on $M$

697: compatible with $\omega$.

698: 

699: 

700:    Let us describe now what the symplectic splitting

701:    construction from  Section \ref{sec:splitting}

702:    looks like from the point of view of  a compatible almost

703:    complex structure.

704: 

705:    First, we assume that the original almost complex structure

706:    $J$ on $W$ is chosen in such a way

707:    that     the contact structure $\xi=\{\alpha=0\}$ on $V$ consists

708:    of complex tangencies to $V$, and that $JX=R_\alpha$, where

709:    $X$ is a conformally symplectic vector field, $\omega$-dual

710:    to $\alpha$, and $R_\alpha$ is the Reeb vector field of $\alpha$.

711:    Next we define an almost complex structure

712:    $J^\tau$ on $W^\tau=W_-\cup V\times[-\tau,\tau]\cup W_+$ by setting

713:    $J^\tau|_{W_\pm}=J$ and requiring  $J^\tau$ to be independent

714:    of $t\in[-\tau,\tau]$ on  $V\times[-\tau,\tau]$.

715:    When $\tau\to\infty$  the almost complex structure

716:    $J_-^\tau$ on

717:    $W_-^\tau=W_- \cup V\times[-\tau,\tau]$   converges

718:     to an almost complex

719:    structure $ J^\infty_-$ on $W^\infty_- $ compatible   with

720:    $\omega^\infty_-$, and  $J_+^\tau$ on

721:    $W_+^\tau=V\times[-\tau,\tau]\cup W_+$   converges

722:     to an almost complex

723:    structure $ J^\infty_+$ on $W^\infty_+$ compatible   with

724:    $\omega^\infty_+$.

725: 

726:    \subsection{Holomorphic curves in symplectic

727:    cobordisms}\label{sec:holomorphic}

728: 

729: 

730:   Let $(V,\alpha)$ be a contact manifold with a fixed contact form

731: and $(W=V\times\R,\omega=d(e^t\alpha))$ its symplectization. Let

732: us denote by $\pi_\R$ and $\pi_V$ the projections $W\to\R$ and $W\to

733: V$, respectively. For a map $f:X\to W$ we write $ f_\R$ and $f_V$

734: instead of $\pi_\R\circ f$ and $\pi_V\circ f$.

735: 

736: Notice that given a trajectory $\gamma$ of the Reeb field

737: $R_\a$,

738: the cylinder $\R\times\gamma\subset W$ is a $J$-holomorphic curve.

739:  Let us also observe that

740: \begin{proposition}\label{prop:positivity}

741:   For a $J$-holomorphic curve  $C\subset W$

742:  the restriction $d\a|_C$ is non-negative, and if $d\a|_C\equiv 0$

743:  then $C$ is a (part of a) cylinder $\R\times\gamma$ over a

744:  trajectory $\gamma$ of the Reeb field $R_\a$.

745:   \end{proposition}

746: 

747:  Given a $J$-holomorphic map $f$ of a punctured disk $D^2\setminus

748:  0\to W$ we say that  the map $f$ is {\it asymptotically

749:  cylindrical } over a periodic orbit $\g$ of the Reeb field

750: $R_\a$ at $+\infty$ (resp. at $-\infty$) if

751: $\mathop{\lim}\limits_{r\to 0} f_\R(re^{i\theta})=+\infty$ (resp.

752: $=-\infty$), and $\mathop{\lim}\limits_{r\to

753: 0}f_V(re^{i\theta})=\bar f(\theta)$, where the map $\bar

754: f:[0,2\pi]\to V$ parameterizes the trajectory $\gamma$.

755: 

756:  %%%%%

757:  \bigskip

758: 

759:  The almost complex manifold $(W,J)$ is bad from the point of view of

760: the theory of

761: holomorphic curves: it has a {\it pseudo-concave} end

762: $V\times(-\infty,0)$, or using Gromov's terminology its geometry

763: at this end is not bounded. However, it was shown in

764: \cite{Hofer-Weinstein} that   Gromov compactness theorem can be

765: modified to accommodate this situation, see Theorems

766: \ref{thm:comp1} and \ref{thm:comp2}

767: below. We will mention in this

768: section only the following fact related to compactness, which motivates the usage

769: of holomorphic curves asymptotically cylindrical over orbits from

770: $\Pc_\a$.

771: \begin{proposition}

772:  Suppose that all periodic orbits of the Reeb field $R_\alpha$ are

773: non-degenerate.

774:  Let  $C$ be  a non-compact Riemann surface  without boundary and $f:C\to W$

775:    a proper $J$-holomorphic curve. Suppose that there

776: exists a constant $K>0$ such that

777: $\int\limits_{C}f^*d\a  < K$.

778: Then

779:  $C$ is conformally equivalent to a compact Riemann surface $S_g$

780:  of genus $g$ with $s^++s^-$ punctures

781:  $$ x^+_1,\dots,

782:  x^+_{s^+},x^-_1\dots,x^-_{s^-}\in S_g,$$

783:   such that near the punctures ${\bx^+}=(x^+_1,\dots,x^+_{s^+})$

784:   the map

785:  $f$ is asymptotically cylindrical over periodic orbits

786:   ${\Gamma^+}=\{\g^+_1,\dots,\g^+_{s^+}\}$ at $+\infty$, and near

787:  the punctures ${\bx^-}=\{x^-_1,\dots,x^-_{s^-}\}$ the map

788:  $f$ is asymptotically cylindrical over periodic orbits

789:   ${\Gamma^-}=\{\g^-_1,\dots,\g^-_{s^-}\}$ at $-\infty$.

790: \end{proposition}

791: 

792: 

793:  %%%%%

794:  Thus holomorphic maps of punctured Riemann surfaces,

795:    asymptotically cylindrical over periodic

796: orbits of the Reeb vector field $R_\alpha$, form a natural class of

797: holomorphic curves to consider in symplectizations

798: as well as  more general symplectic manifolds with cylindrical ends.  We

799: will define now moduli spaces of such curves.

800: 

801: Let $W =\ora{V^-V^+}$ be a (completed) directed cobordism,

802: $\alpha^\pm$    corresponding contact forms on $V^-$ and $V^+$,

803:  $\Pc^\pm$  the sets of all

804: periodic orbits (including multiple ones) of the Reeb vector

805: fields

806: $R_{\alpha^\pm}$. We assume that $\alpha^\pm$ satisfies

807: the genericity assumptions from Section \ref{sec:dynamics}.

808: Choose a compatible almost complex  structure  $J$   on $W$.

809:  Let $\Gamma^\pm$ be   ordered sets of trajectories from

810:  $\Pc^\pm$ of cardinality $s^\pm$.

811:  We also assume that every {\it simple} periodic orbit $\gamma$ from

812:  $\Pc^\pm$ comes with a fixed marker

813:  $m_\gamma\in\g$.

814: 

815:  Let $S=S_g$ be a compact Riemann surface of genus $g$ with a conformal

816:  structure $j$, with $s^+$ punctures $\bx^+=\{x^+_1,\dots,x^+_{s^+}\}$, called

817:  positive,

818:  $s^-$ punctures $\bx^-=\{x^-_1,\dots,x^-_{s^-}\}$, called negative, and

819:  $r$ marked points $\by=\{y_1,\dots,y_r\}$.

820:   We will also   fix an {\it asymptotic marker} at each

821:    puncture. We mean by that  a ray originating at each

822:    puncture.  Alternatively, if one

823:    takes the cylinder $S^1\times[0,\infty)$ as a conformal model of

824:    the punctured disk $D^2\setminus 0$ then an asymptotic marker

825:    can be viewed as a point on the circle at infinity.

826:    If a holomorphic map $f:D^2\setminus 0\to V^\pm\times\R_\pm$ is asymptotically

827:    cylindrical over a periodic orbit $\g$,  we say that

828:     a marker $\mu=\{\theta=\theta_0\}$

829:     is mapped by $f$ to the

830:    marker $m_\gamma\in\overline\gamma$, where $\overline\g$ is the simple

831:      orbit which  underlines $\g$, if $\mathop{\lim}\limits_{r\to

832:    0}f_{V^\pm}(re^{i\theta_0})=m_\g$.

833:   %%%%%

834: Let us recall (see Section \ref{sec:dynamics} above) that we

835: provided each periodic orbit from $\Pc^\pm$ with a ``capping"

836: surface. This surface bounds $\g\in\Pc^\pm$ in $V_\pm$ if $\g$ is

837: homologically trivial, or realizes a homology    between $\g$ and

838: the corresponding linear combination of  basic curves

839: ${C}_i^\pm$. We will  continue to rule out torsion elements in

840: the first homology

841:  (see the discussion of torsion in Section \ref{sec:torsion} below)   and choose

842:   curves ${C}_k\subset W$ which represent a basis of

843: the image $H_1(V^-\cup V^+)\to H(W)$ and for each curve

844: ${C}^\pm_i$ fix a surface $G^\pm_i$ which realizes a homology

845: in $W$

846: between ${C}^\pm_i$ and the corresponding linear combination of

847: curves ${C}_k$.

848: All the choices enable us to associate with a  relative

849:  homology class $A'\in H_2(W,\Gamma^-\cup\Gamma^+)$,

850:  $\Gamma^\pm\subset\Pc^\pm$,

851:  an absolute integral  class

852:  $A\in H_2(W)$, which we will view as an element of $H_2(W;\C)$.

853: 

854:   %%%%%

855: Let us denote by $\Mc^A_{g,r}(\Gamma^-,\Gamma^+;W,J)$ the moduli

856: space of $(j,J)$-holomorphic curves

857: $S_g\setminus(\bx^-\cup\bx^+)\to W$ with $r$ marked points, which

858: are asymptotically cylindrical over the periodic orbit $\g^+_i$

859: from $\Gamma^+$ at    the positive end at the puncture $x^+_i$,

860: and asymptotically cylindrical over the periodic orbit $\g^-_i$

861: from $\Gamma^-$ at    the negative end  at the puncture $x^-_i$,

862: and which send asymptotic markers to the

863: markers on the

864: corresponding periodic orbits. The curves from

865: $\Mc^A_{g,r}(\Gamma^-,\Gamma^+;W,J)$ are additionally required to

866: satisfy a stability condition, discussed in the next section. We

867: write $\Mc^A_{g}(\Gamma^-,\Gamma^+;W,J)$ instead of

868: $\Mc^A_{g,0}(\Gamma^-,\Gamma^+;W,J)$, and

869: $\Mc^A_{g,r}(\Gamma^-,\Gamma^+)$ instead of

870: $\Mc^A_{g,r}(\Gamma^-,\Gamma^+;W,J)$ if   it is clear from the

871: context which target almost complex  manifold $(W,J)$ is

872: considered.

873: 

874:  Notice, that we are not fixing $j$, and the configurations of

875:  punctures, marked points  or asymptotic markers.

876:  Two maps are called equivalent if they differ by a conformal map

877:  $S_g\to S_g$ which preserves all punctures, marked points and

878:  asymptotic markers.

879:      When the manifold $W=V\times\R$ is cylindrical,

880:      and hence the almost complex structure      $J$ is invariant under

881:      translations along the second factor, then

882:      all the moduli spaces $\Mc^A_{g,r}(\Gamma^-,\Gamma^+;W,J)$ inherit the $\R$-action.

883:      We will denote the quotient moduli space  by

884:        $\Mc^{A}_{g,r}(\Gamma^-,\Gamma^+;W,J)/\R$, and by

885: $\Mc^A_{g,r,s^-,s^+}(W,J)$ the union $\bigcup

886: \Mc^A_{g,r}(\Gamma^-,\Gamma^+;W,J)$ taken over

887: all sets of periodic orbits $\Gamma^\pm\subset\Pc^\pm$   with the

888: prescribed numbers $s^\pm$ of elements.

889: We will also need to consider the moduli space of disconnected

890:   curves of  Euler characteristic $2-2g$, denoted by

891:   $\wt{\Mc}_{g,r}^A(\Gamma^-,\Gamma^+)$.

892: 

893: 

894: 

895: 

896: 

897: 

898: 

899: 

900: \subsection{Compactification of the moduli spaces

901: $\Mc_{g,r}^A(\Gamma^-,\Gamma^+)$}\label{sec:compact}

902: 

903: To describe the compactification we need an appropriate notion of

904: a {\it stable holomorphic curve}.

905: 

906: Given a completed symplectic cobordism $W=\ora{V^-V^+}$ we first

907: define a {\it   stable curve of height $1$}, or a {\it $1$-story stable curve }

908:  as a ``usual" stable curve in a sense of M. Kontsevich (see \cite{Kontsevich-CP2}),

909: i.e. a collection of of holomorphic curves $h_i:S_i\to

910: W$  from  moduli spaces

911: $\Mc_{g_i,r}^A(\Gamma^-_i,\Gamma^+_i)$ for various genera $g_i$

912:  which realize  homology classes

913: $A_i$, and collections of periodic orbits $\Gamma_i^\pm$, such

914: that certain pairs of marked points  (called special) on these

915: curves are required to be mapped to one point in $W$. The

916: stability condition means the absence of infinitesimal symmetries

917: of the moduli space. Let us point out, however, that in the case

918: when

919:  $W$ is a cylindrical cobordism, and in particular the almost complex structure

920:  $J$ is translationally invariant, we would need to consider along

921:  with the above moduli space its quotient under the $\R$-action.

922:  The stability for this new moduli space still means an absence of

923:  infinitesimal deformations, but it translates into an additional restriction

924:  on holomorphic curves.

925:   Namely, in the first case the stability

926: condition means that each constant curve has, after   removal

927: of the marked points, a negative Euler characteristic. In the second

928: case it additionally requires  that when {\it all} connected

929: components of the curve are straight cylinders $\g\times\R,

930: \g\in\Pc$ then  at least one of these cylinders should have a

931: marked point.

932: 

933:   One can define an arithmetic genus $g$ of the

934: resulting curve, the total sets $\bx^\pm$ and $\by$ (equal to the

935: union of sets $\bx_i^\pm$ and $\by_i$ for the individual curves of

936: the collection), and the total  absolute  homology class  $A\in

937: H_2(W )$ (see the discussion in Section \ref{sec:holomorphic}

938: above), realized by the union of all curves of the collection.

939: 

940: Moduli of stable curves of height $1$, denoted by

941: ${}_1\Mc_{g,r}^A(\Gamma^-,\Gamma^+)$, form  a part of the

942: compactification of the moduli space

943: $\Mc_{g,r}^A(\Gamma^-,\Gamma^+)$.

944:   However, unlike the case of  closed symplectic manifolds, the

945:   stable curves of height $1$ are not   sufficient  to describe  the

946:  compactification of the moduli space $\Mc_{g,r}^A(\Gamma^-,\Gamma^+)$.

947: 

948: A finite sequence  $(W_1,\dots, W_k)$ of symplectic manifolds with

949: cylindrical ends is called a {\it chain} if  the positive end of

950: $W_i$ matches with the negative end of $W_{i+1}$, $i=1,\dots,k-1$.

951: This means that all data, assigned to an end, i.e. a contact

952: form, marking of periodic orbits, and an almost complex structure,

953: are the same for the matching ends.

954: 

955: Let us first suppose that

956:  none of the cobordisms  which form a chain $(W_1,\dots, W_k)$

957: is cylindrical.    Then

958: a {\it stable curve of height

959: $k$}, or a $k$-story stable curve  in the

960: chain $(W_1,\dots, W_k)$ is a $k$-tuple

961: $f=(f_1,\dots,f_k)$, where $f_{

962: i } \in

963: {}_1\wt{\Mc}_{g_i,r_i}^{A_i}(\Gamma_{i}^-,\Gamma_{i}^+;W_i,J_i)$,

964:  such that the  boundary data  of the curve $f_i$ at the positive end

965: match  the boundary data of $f_{i+1}$ on the negative one.

966:  One

967: also needs to impose the following additional equivalence relation regarding

968: the asymptotic markers on multiple orbits.

969:                     Suppose that $\g $ is

970:    a $k$-multiple periodic orbit, so that the holomorphic curve $f_i$ is

971:    asymptotically cylindrical over $\g$ at the positive end at a puncture $x^+ $, and

972:    $f_{i+1}$ is asymptotically cylindrical over $\g$ at  the negative end  at a puncture $x^-$.

973:     There are $k$ possible  positions $\mu^+_1,\dots,\mu^+_k$

974:     and $\mu^-_1,\dots,\mu^-_k$

975:   of   asymptotic  markers at  each of the punctures $x^\pm$.

976:    We assume here that the markers are numbered cyclically

977:   with respect to the orientation defined by the Reeb vector field

978:   at each of the punctures, and that the markers $\mu^+_1$ and

979:   $\mu^-_1$ are chosen for the curves $f_i$ and

980:   $f_{i+1}$.         Then we identify

981:   $f=\{\dots,f_i,     f_{i+1},\dots\}$

982:    with $(k-1)$ other stable curves of height $k$

983:   obtained by simultaneous cyclic  shift of

984:    the asymptotic markers at the punctures $x^+$ and $x^-$.

985: 

986:    The curves $f_i$, which form  a $k$-story stable curve $f=(f_1,\dots,f_k)$ are called

987: {\it floors},  or {\it levels} of $f$.

988: \medskip

989: 

990:    If some of the cobordisms which form the chain   $(W_1,\dots, W_k)$, say $W_{i_1},\dots,W_{i_l}$,

991: are cylindrical  then we will assume that the corresponding floors

992: of a $k$-story curve in $W=(W_1,\dots,W_k)$ are defined {\it only

993: up to translation.} In other words, if $W_i$ is cylindrical for some $i=1,\dots, k$

994: (i.e. $W_i=V_i\times\R$ and $J_i$ is translationally invariant) then

995: $f_i$ should be viewed as an element of

996: ${}_1\wt{\Mc}_{g_i,r_i}^{A_i}(\Gamma_{i}^-,\Gamma_{i}^+;W_i,J_i)/\R$,

997: rather than ${}_1\wt{\Mc}_{g_i,r_i}^{A_i}(\Gamma_{i}^-,\Gamma_{i}^+;W_i,J_i)$.

998:  It will be convenient for us, however, to introduce the following convention.

999:  When speaking about stable holomorphic  curves  in chains which contain

1000:  cylindrical cobordisms, we will treat the corresponding floors as curves representing

1001:  their equivalence classes from

1002:   ${}_1\wt{\Mc}_{g_i,r_i}^{A_i}(\Gamma_{i}^-,\Gamma_{i}^+;W_i,J_i)/\R$. Any statement about such

1003:   curves should be understood in the sense, that {\it there exist} representatives for

1004:   which the statement is true.

1005: 

1006: 

1007: \bigskip

1008: Let us define now the meaning of convergence of a

1009: sequence of holomorphic curves to a stable curve of height $l$.

1010: For $l=1$ this is   Gromov's standard  definition (see \cite{Gromov-holomorphic}). Namely, with

1011: each stable curve

1012:  $h=\{S_i,h_i\}\in {}_1\Mc_{g,r}^A(\Gamma^-,\Gamma^+)$   of height $1$ we associate a nodal

1013:    surface $\widehat S$

1014: obtained by identifying special pairs of marked points on $\coprod S_i$.  The maps $h_i$ fits together

1015: to a continuous map $\widehat S\to W$ for which we will keep the notation

1016: $h$.

1017:  Let us  consider

1018: also a smooth surface $S$ obtained by smoothing the nodes

1019: of $\widehat S$. There exist a partitioning of  $S$

1020:  by circles

1021:  into  open parts   diffeomorphic to  surfaces $S_i$    with removed special points, and a map

1022:   $g:S\to\widehat S$  which is a diffeomorphism from the complement $\tilde S$

1023:   of the dividing circles

1024:   in $S$ to the complement of  the double points

1025:     in $\hat  S$, and  which collapses the partitioning circles to

1026:   double points.

1027:   A sequence of holomorphic  $\varphi_l:(S,j_l)\to (W,J)$ is said to converge

1028:    to a stable curve

1029:  $h$

1030:     if

1031:       the sequence $\varphi_l|_{\tilde S}$ converges

1032:       to $h_i\circ g|_{\tilde S}$, and

1033:       $j_l$ converges to $g^*(j)$  uniformly on compact sets,

1034:       where $j$ is the conformal structure on the stable curve.

1035:       Of course, we also require  convergence of marked points and asymptotic markers.

1036:     A sequence of stable curves

1037:     $h^j=\{S^j_i,h^j_i\}_{i=1,\dots,k}\in {}_1\Mc_{g,r}^A(\Gamma^-,\Gamma^+), j=1,\dots,$  is said to converge

1038:     to a stable curve $h$, if $h$ can be presented as a collection

1039:     of stable curves $h_i$, $i=1,\dots, k$, such that  $h^j_i$ converges to $  h_i$ in the above

1040:     sense for each $i=1,\dots,k$.

1041:     \medskip

1042: 

1043:  The convergence of

1044:   a sequence of smooth curves to a stable curve of height  $l>1$  is understood

1045:   in a similar sense.

1046:   Let us  assume here     for simplicity that $l=2$ and that the   floors

1047:     $f_{1} :     S_1\to W_1$ and $f_2 :

1048:   S_2\to W_2$  of  a stable curve $f$ in a chain $(W_1,W_2)$

1049:   are smooth, i.e.

1050:   have no  special marked points.

1051:   As in the height $1$ case let us consider

1052:       \begin{description}

1053:       \item{-} the smooth surface $S$

1054:       partitioned  according to the combinatorics of our stable curve

1055:       by circles into two open  (possibly disconnected) parts $U_1$ and $U_2$  diffeomorphic to

1056:       the punctured surfaces $S_1$

1057:        and $S_2$,

1058:        \item{-} the surface

1059:         $\hat S$ with double points obtained by collapsing

1060:         these circles to points, and

1061:         \item{-} the projection $g:S\to\hat S$.

1062:       \end{description}

1063:        Let $(W,J)=(W_1,J_1)\circledcirc (W_2,J_2)$

1064:    be the composition of (completed) directed symplectic    cobordisms $W_1$ and $W_2$

1065:    with compatible almost complex structures $J_1$ and $J_2$. This means that

1066:    \begin{description}

1067:    \item{-}    there exists a contact    hypersurface $V\subset W$ which splits

1068:       $W$ into two cobordisms $W^0_1$ and $W^0_2$;

1069:       \item{-} $W_1=W^0_1\cup V\times[0,\infty),\;W_2=V\times(-\infty,0]\cup W^0_2$;

1070:       \item{-} $J|_{W^0_j}=J_1|_{W^0_j},\;\; j=1,2$;

1071:       \item{-} $J_1$ and $J_2$ are translationally invariant at the ends

1072:       $V\times[0,\infty)$ and $(-\infty,0]\times V$.

1073:       \end{description}

1074: 

1075:       We denote  by

1076:             $W^k $, $k=1,\dots,$ the quotient space of the disjoint union

1077:             $$W^0_1\coprod V\times[-k,k]\coprod W^0_2$$ obtained by identifying $V=\partial W^0_1$

1078:             with  and $V\times( -k)$ and $V=\partial W^0_2$

1079:             with  and $V\times k$,  and extend the   almost

1080:       complex  structures $J_1|_{W^0_1}$ and     $J_2|_{W^0_2}$

1081:       to the unique almost complex structure  $J^k$ on $W^k$ which is translationally

1082:        invariant on $V\times[-k,k]$. We also consider $W^k_1$ obtained

1083:        by gluing $W^0_1$ and $V\times[0,k]$ along $V=\partial

1084:        S^0_1=V\times 0$,

1085:        and $W^k_2$ glued in a similar way from $V\times[-k,0]$ and $W^0_2$.

1086:         We have $W_j=\bigcup\limits_{k=0}^\infty

1087:        W^k_j$, $j=1,2$. On the other hand, $W^k_1$ and $W^k_2$

1088:         can be viewed as submanifolds

1089:        of $W^k$.

1090: 

1091: 

1092:       \begin{definition}\label{def:convergence}

1093:       {\rm

1094:                 Suppose that we are  given a sequence $j^k$

1095:                  of conformal structures on the surface $S$

1096:        and a sequence  of  $1$-story $(j^k,J^k)$-holomorphic curves

1097:        $f^k:S\to W^k$.

1098:        We say that this sequence converges to a stable curve $f=(f_1,

1099:        f_2)$

1100:        of height   $2$

1101:        in $(W_1,W_2)$ if     there     exist two sequences of domains

1102:        $U_1^1\subset\dots \subset U_1^i\subset \dots\subset U_1$

1103:        and $U_2^1\subset\dots \subset U_2^i\subset \dots\subset U_2$,

1104:         such that

1105:         \begin{description}

1106:         \item{} $\bigcup\limits_{k=1}^\infty U^k_i=U_i,\; i=1,2$ ;

1107:         \item{} $f^k(U^k_i)\subset W^k_i$ for $i=1,2,\;\;k=1,\dots$;

1108:         \item{}  for $i=1,2$  the  holomorphic curves

1109:         $f^k|_{U^k_i}$ converges to $f_i\circ g: U_i\to W_i$,

1110:                 and the conformal structures   $j^k|_{U^k_1}$ converge to $g^*{j_i} $

1111:                 when $k\to\infty$

1112:          uniformly on

1113:         compact sets.

1114: As in the case of stable curves of level $1$

1115: we also require

1116: convergence of marked points and asymptotic markers.

1117:                 \end{description} }

1118:   \end{definition}

1119:   Let us emphasize that  when some of the cobordisms  are cylindrical

1120:   then according to  the convention which we introduced above

1121:     one is allowed to compose  the corresponding curves  with translations to satisfy

1122:   the above definition.

1123: 

1124:  \bigskip

1125:  Notice that if the  cobordism $W_2$ is cylindrical, i.e.

1126:  $W_2=V\times\R$ and $J_2$ is translationally invariant,

1127:  then $W_1\circledcirc W_2$ can be identified with $W_1$, and thus one can talk about

1128:  convergence of a sequence of curves

1129:   $f^k\in {}_1{\Mc}_{g,r}^{A}(\Gamma^-,\Gamma^+;W_,J_1)$

1130:   (where the almost complex structure

1131:   $J_1$ is fixed!)

1132:  to   a $2$-story curve $(f_1,f_2)$, where

1133:  $f_1 \in{}_1\wt{\Mc}_{g_1,r_1}^{A_1}(\Gamma^-,\Gamma;W_1,J_1)$,

1134:   $f_2\in{}_1\wt{\Mc}_{g_2,r_2}^{A_2}(\Gamma,\wt\Gamma^+;V\times\R,J_2)/\R$,

1135:   $g=g_1+g_2,\,r=r_1+r_2, A=A_1+A_2$, and $J_2$ is translationally invariant.

1136:  It is important to stress the point that the curve   $f_2$ is defined only up to translation.

1137: 

1138: 

1139: 

1140:        \begin{theorem}\label{thm:comp1}

1141:        Let $f_k\in {}_1\Mc_g^A(\Gamma^-,\Gamma^+)$, $k=1,\dots, $ be a

1142:        sequence of stable holomorphic curves in

1143:        a (complete)  directed symplectic cobordism $W$.

1144:         Then there exists a chain of  directed symplectic

1145:         cobordisms $$A_1,\dots, A_a,W,B_1,\dots,B_b,$$ where all

1146:         cobordisms $A_i$ and $B_i$ are cylindrical,

1147:         and a stable curve $f_\infty$ of height $a+b+1$ in this chain

1148:         such that a subsequence of $ \{f_i\}$  converges to

1149:         $f_\infty$. {\rm See Fig. \ref{fig:split1}}.

1150:         \end{theorem}

1151: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1152: 

1153: \begin{figure}

1154: \centerline{\psfig{figure=sf2.eps,height=100mm}}

1155: \caption{\small A possible splitting of a sequence of holomorphic curves in

1156:  a completed symplectic cobordism}

1157: \label{fig:split1}

1158: \end{figure}

1159: \begin{figure}

1160: \centerline{\psfig{figure=sf4.eps,height=80mm}} \caption{\small A

1161: possible splitting of a sequence of holomorphic curve   when

1162: $J_k\to J_\infty$ } \label{fig:split2}

1163: \end{figure}

1164: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1165:         \begin{theorem}\label{thm:comp2}

1166:         Let $W$ be a completed directed symplectic cobordism,

1167:         $V\subset W$ a contact hypersurface, and $J_k$    a

1168:         sequence of compatible almost complex structures on $W$

1169:         which realizes the splitting of $W$ along $V$ into two

1170:         directed symplectic cobordisms $W_-^\infty$ and

1171:         $W_+^\infty$

1172:         {\rm (see Section \ref{sec:almost} above)}.

1173:         Let $f_k$ be a sequence of stable $J_k$-holomorphic curves

1174:           from $ {}_1\Mc_g^A(\Gamma^-,\Gamma^+;W,J_k)$. Then there exists a

1175:         chain of directed symplectic cobordisms

1176:         $$A_1,\dots,A_a,W_-,B_1,\dots,B_b,W_+,C_1,\dots,C_c$$  where

1177:         all cobordisms $A_i, B_j, C_l$ are cylindrical, such that a

1178:         subsequence of

1179:         $\{f_i\}$   converges to a stable curve of height $a+b+c+2$

1180:         in

1181:          the chain $$A_1,\dots,A_a,W_-,B_1,\dots,B_b,W_+, C_1,\dots,

1182:          C_c.$$ {\rm See Fig. \ref{fig:split2}. The reader

1183:          may consult \cite{HoferWysockiZehnder:99}

1184:          for the analysis of splitting $\C P^2$

1185:          along the boundary of a tubular

1186:           neighborhood

1187:          of $\C P^1\subset\C P^2$.}

1188:         \end{theorem}

1189: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1190: 

1191: The definition of convergence can be extended in an obvious way to a sequences of stable curves

1192: of height $l>1$. Namely, we say that a sequence of $l$-story curves

1193: $f^k=(f^k_1,\dots,f^k_l)$, $k=1,\dots,\infty$, in a chain $(W_1,\dots,W_l)$ converge to a stable $L$-story,

1194: $L=m_1+\dots+m_l$, curve  $f=(f_{11},\dots,,f_{1m_1},\dots,f_{l_1}\dots f_{lm_l}) $ in a chain

1195:  $$(W_{11},\dots,W_{1m_1},\dots,W_{l_1}\dots W_{lm_l})$$

1196: if for each $i=1,\dots,l$  the cobordism $W_i$ splits into the composition

1197: $$W_i=W_{i1}\circledcirc\dots\circledcirc W_{im_i}$$

1198: and the sequence $f^k_i$, $k=1,\dots,\infty$, of stable curves of height $1$ converges

1199: to the $m_i$-story curve $f_i=(f_{i1},\dots,f_{im_i})$ in the chain $(W_{i1},\dots, W_{im_i})$ in the sense of Definition

1200: \ref{def:convergence}.

1201: \bigskip

1202: 

1203: %%%%%%%%%%%%%%%%%%%%%%%%%

1204: It is important to combine Theorems \ref{thm:comp1} and

1205: \ref{thm:comp2} with the following

1206: observation which is a

1207: corollary of  Stokes' theorem combined with Proposition \ref{prop:positivity}.

1208: 

1209: 

1210: \begin{proposition}\label{prop:one-end}

1211:  A holomorphic curve in an exact

1212:   directed symplectic cobordism

1213:  (and in particular in a cylindrical one)

1214:  must have at least one positive puncture.

1215:  \end{proposition}

1216: 

1217:  In particular,  we have

1218:  \begin{corollary}\label{cor:pos-puncture}

1219: Let $f^n\in {\Mc}_0(W,J)$  be a sequence of rational holomorphic curves with  one positive, and possibly

1220: several negative punctures.

1221:  Suppose that the sequence  converges to a stable

1222:  curve $$F=\{g_1,\dots,g_a,f,h_1,\dots,h_b\}$$ of height $a+b+1$ in a chain

1223:  $$A_1,\dots,A_a,W,B_1,\dots,B_b.$$ Then the  $W$-component   $f\in {\Mc}_0(W,J)$ of the stable curve

1224:  $F$ has precisely

1225:  one positive puncture as well.

1226:  \end{corollary}

1227: 

1228: 

1229:  \subsection{Dimension of the moduli spaces

1230: $\Mc^A_{g,r}(\Gamma^-,\Gamma^+)$}

1231: One has the following index formula for the corresponding

1232: $\bar\partial$-problem which compute the dimension

1233: of the moduli space $\Mc^A_{g,r}(\Gamma^-,\Gamma^+;W,J)$ for a {\it

1234: generic}

1235:  choice of $J$.

1236: \begin{proposition}

1237: 

1238:   \begin{equation}\label{eq:dim}

1239:   \begin{split}

1240:  & {\mathrm {dim}}\Mc^A_{g,r}(\Gamma^-,\Gamma^+;W,J)

1241:  =\sum\limits_1^{s^+}\CZ(\gamma^+_i)-

1242:  \sum\limits_1^{s^-}\CZ(\gamma^-_k)\\

1243:  &  +  (n-3)(2-2g-s^+-s^-)+2c_1(A)+2r,\\

1244:  \end{split}

1245:  \end{equation}

1246:   where

1247: $s^\pm$ are the cardinalities of the sets $\Gamma^\pm$, and $c_1\in

1248: H^2(W)$ is the first Chern class of the almost complex manifold

1249: $(W,J)$

1250: \end{proposition}

1251: 

1252:   Making the moduli spaces non-singular by picking

1253:    generic $J$ is needed for the purpose of curve counting but

1254:             does not always work properly. It is therefore crucial

1255:              that the moduli spaces of stable $J$-holomorphic

1256: curves are non-singular {\em virtually}. This means that for {\em

1257: any} $J$ the moduli spaces, being generally speaking singular, can

1258: be equipped with some canonical additional structure that make

1259: them {\em function in the theory the same way as if they were

1260: orbifolds with boundary and had the dimension prescribed by the

1261: Fredholm index}. In particular, the moduli spaces come equipped

1262: with rational fundamental cycles relative to the boundary (called

1263: {\em virtual fundamental cycles}) which admit pairing with

1264: suitable de Rham cochains and allow us to use the Stokes

1265: integration formula.

1266: 

1267: 

1268: 

1269: Technically the virtual smoothness is achieved by a

1270: finite-dimensional reduction of the following picture: a singular

1271: moduli space is the zero locus of a section defined by the

1272: Cauchy-Riemann operator in a suitable orbi-bundle over a moduli

1273: orbifold of stable $C^{\infty}$-maps. More general {\em virtual

1274: transverality} properties for families of

1275:  $J$'s also hold true

1276: (cf. \cite{Fukaya-Ono, Fukaya-Ono2, Liu-Tian, Li-Tian, Ruan-virtual, Siebert, McDuff} at al.).

1277: We

1278: are reluctant to provide in this quite informal exposition

1279:  precise

1280: formulations   because of numerous not entirely innocent

1281: subtleties this would entail.  Fortunately, what we intend to say in the rest of this

1282: paper does not depend much on the details we are omitting.

1283: 

1284: 

1285: \bigskip

1286: 

1287: 

1288: As it was explained in Section \ref{sec:compact} above the moduli

1289: space $\Mc^A_{g,r}(\Gamma^-,\Gamma^+;W,J)$  can be compactified by

1290: adding strata which consist of stable holomorphic curves of

1291: different height.  This compactification looks quite similar to

1292: the Gromov-Kontsevich compactification  of   moduli spaces of

1293: holomorphic curves in a closed symplectic manifold  with a

1294: compatible almost complex structure.

1295: There is, however, a major difference. In the case of a closed

1296: manifold   all the strata   which one needs to add to compactify the moduli space

1297: of smooth holomorphic curves have (modulo virtual cycle complications) codimension

1298: $\geq 2$. On the other hand  in our case  the  codimension one strata are

1299: present {\it generically}. Thus in this case the boundary of the compactified moduli space,

1300: rather than the moduli space  itself, carries  the fundamental

1301: cycle.

1302: 

1303:  In particular, this boundary

1304:   is tiled by codimension one

1305:  strata represented by stable curves $(f_-,f_+)$ of {\it height two}.

1306: Each such a stratum can be described by the constraint matching

1307: the positive ends of $f_-$ with the negative ends of $f_+$ in the

1308: Cartesian product of the moduli spaces $\Mc_{\pm}$ corresponding to   the

1309: curves

1310: $f_{\pm}$  separately.

1311: Proposition \ref{prop:boundary} below describes these top-dimensional boundary

1312:  strata more precisely

1313: in two important for our purposes situations.

1314: Let us point out that  \ref{prop:boundary} literally holds only under certain

1315:  transversality conditions.

1316: Otherwise it should be understood only virtually.

1317: 

1318: 

1319:  \begin{proposition}\label{prop:boundary}

1320:      \begin{description}

1321:      \item{1.} Let $(W=V\times\R,J)$ be a cylindrical cobordism.

1322:      Then any

1323:     top-dimensional stratum  $\cS$ on the boundary  of the compactified moduli space

1324:      $\overline{\Mc^A_{g,r}(\Gamma^-, \Gamma^+;W,J)/\R}$ consists

1325:      of

1326:          stable curves  $(f_-,f_+)$   of height   two,

1327:           $f_\pm\in\Mc_\pm/\R$, where

1328:           $$\Mc_-=\wt\Mc^{A_-}_{g_-,r_-}(\Gamma^-,\Gamma;W,J)/\R ,

1329:           \quad \Mc_+=\wt\Mc^{A_+}_{g_+,r_+}(\Gamma,\Gamma^+;W,J)/\R ,$$

1330:            $g=g_-+g_+, \;r=r_-+r_+,\;A=A_-+A_+$,   $\Gamma=\{\g_1,\dots,\g_l\}\subset\Pc$.

1331: 

1332:                  All  but one

1333:           connected components  of  each of the curves $f_-$ and $f_+$

1334:            are trivial cylinders

1335:           (i.e.  have the form   $\g\times\R, \g\in\Pc$) without

1336:           marked points.

1337:           \item{2.} Let $(W=\ora{V^-V^+},J)$  be any    cobordism,

1338:           and $(W_\pm,J_\pm)= (V^\pm\times\R,J_\pm)$

1339:           be the cylindrical cobordisms associated to its

1340:           boundary.

1341:           Then any  top-dimensional strata $\cS$ on the boundary  of the

1342:           compactified moduli space

1343:      $\overline{\Mc^A_{g,r}( \Gamma^-, \Gamma^+;W,J) }$ consists

1344:      of

1345:          stable curves  $(f_-,f_+)$   of height   two,

1346:           $f_\pm\in\Mc_\pm$, where either

1347:           $$\Mc_-=\wt\Mc^{A_-}_{g_-,r_-}(\Gamma^-,\Gamma;W_-,J_-)/\R,\;

1348:  \Mc_+=\wt\Mc^{A_+}_{ g_+,r_+}(\Gamma,\Gamma^+;W,J)\;\hbox{

1349:  and}\;

1350:   \Gamma\subset\Pc^-,$$ or

1351: $$\Mc_+=\wt\Mc^{A_+}_{g_-,r_-}(\Gamma ,\Gamma^+;W_+,J_+)/\R,\;

1352:  \Mc_-=\wt\Mc^{A_-}_{g_-,r_-}(\Gamma^-,\Gamma;W,J ) \;\hbox{

1353:  and}\;

1354:   \Gamma\subset\Pc^+.$$

1355:    In both cases  we have  $$g=g_-+g_+,

1356:  \;r=r_-+r_+,\;\;\hbox{and}\;\;A=A_-+A_+.$$

1357:   The part of the stable curve

1358:  $(f_-,f_+)$ which is  contained in in  $W_\pm$ must have

1359:  precisely one non-cylindrical connected component, while there

1360:  are no restrictions on the number and the character of connected components or the

1361:  other part of the stable curve.

1362:  \item{~~} In both cases the stratum

1363:          $\cS=\cS(\Gamma,g_-|g_+,r_-|r_+,A_-|A_+)$ is

1364:          diffeomorphic to a $\kappa$-multiple cover of the product

1365:          $\Mc_-\times\Mc_+$, where the multiplicity $\kappa$ is

1366:          determined by the multiplicities of periodic orbits from

1367:          $\Gamma$.

1368:  \end{description}

1369:               \end{proposition}

1370: 

1371: Proposition \ref{prop:boundary} is not quite sufficient for our

1372: purposes, as we also needs to know the structure of the boundary

1373: of   moduli spaces of $1$-parametric families of holomorphic

1374: curves. However, we are not  formulating the corresponding

1375: statement in this paper, because it  is   intertwined  in a much more serious way

1376: with

1377: the virtual cycle techniques and terminology. An algebraic description

1378: of this boundary is given in

1379:    Theorem

1380: \ref{thm:SFT-chain}

1381: below.

1382: 

1383: 

1384: 

1385: \bigskip

1386: 

1387: 

1388: Let us consider some special cases of the formula (\ref{eq:dim}).

1389: Suppose, for instance, that $W$ is the cotangent bundle of a manifold

1390: $L$. Then $W$ is a symplectic manifold which has  only a positive

1391: cylindrical end.

1392:  If $L$ is orientable then there is a canonical way to

1393: define Conley-Zehnder indices. Namely, one takes any

1394: trivialization along orbits, which is  tangent to vertical

1395: Lagrangian fibers. The resulting index is independent of a

1396: particular  trivialization. For this trivialization, and a choice

1397: of a contact form corresponding to  a metric on $L$ we have

1398:  \begin{proposition}

1399: Periodic orbits of the Reeb flow are lifts of closed geodesics,

1400: and if $L$ is orientable their Conley-Zehnder indices   are equal

1401: to Morse indices of the corresponding geodesics and we have

1402:  $${\mathrm {dim}}\Mc_g^A( \Gamma^+)= \sum\limits_i\Morse(\g^+_i)+ (n-3)(2-2g-s^+) .$$

1403: \end{proposition}\label{prop:Morse-Zehnder}

1404: Notice that for a metric on  $L$ of  non-positive curvature we have  $s^+>1$, because in this case

1405: there are no

1406: contractible geodesics. Moreover, if the metric has negative curvature then all geodesics

1407: have  Morse

1408: indices equal to $0$.  Hence, we get

1409: \begin{corollary}\label{cor:isolated} In  the cotangent bundle of a  negatively curved

1410: manifold  of dimension $>2$ there  could be only isolated

1411: holomorphic curves. If, in addition, $n\neq 3$ then these curves are spheres

1412: with two positive punctures. Each of these curves is asymptotically

1413: cylindrical at punctures over   lifts of the same geodesic with

1414: opposite orientations.

1415: \end{corollary}

1416: %CHANGE

1417: Let us point out that the orientability is not required in

1418: Corollary \ref{cor:isolated}. The corresponding result for a

1419: non-orientable manifold  follows from \ref{prop:Morse-Zehnder}

1420: applied to its orientable double  cover.

1421: 

1422: \subsubsection*{Absence of hyperbolic Lagrangian submanifolds  in uniruled manifolds}

1423: 

1424: As the first  application of the above compactness theorems let us

1425: prove here the following theorem of C. Viterbo. Let us recall that

1426: a  complex projective manifold $W$ is called uniruled, if there is a

1427: rational holomorphic curve

1428:   through each point of $W$. For instance,  according to

1429:   Y. Myaoka--S. Mori

1430:  \cite{Mori-Miayoka}  and J. Kollar  \cite{Kollar}  Fano manifolds are uniruled.

1431: 

1432: 

1433: 

1434: \begin{theorem} {\rm (C. Viterbo, \cite{Viterbo})}

1435: Let $W$ be a uniruled manifold of complex dimension $>2$,  $\omega$ its K\"ahler

1436: sympletic form, and $L\subset W$ an embedded Lagrangian

1437: submanifold. Then $L$ does not admit a Riemannian metric of

1438: negative sectional  curvature.

1439: \end{theorem}

1440: \proof J. Kollar  \cite{Kollar}  and in a more general case Y.

1441: Ruan   \cite{Ruan-virtual} proved  that there exists a homology

1442: class $A\in H_2(W)$, such that for any almost complex structure

1443: compatible with $\omega$ and any point $z\in W$ there exists

1444: $f\in\Mc_{0,1}^A(W,J)$ with $f(y)=z$, where $y$ is the marked

1445: point. Let us identify a neighborhood $U$ of $L$ in $W$ with  a

1446: neighborhood of the zero-section in $ T^*(L)$. Suppose $L$ admits

1447: a Riemannian metric of negative

1448:   curvature. We can assume that  $U$ is the round neighborhood of radius $1$ in

1449:   $T^*(L)$. Let us consider a sequence $J^m$ of almost complex structures

1450:   on $W$, which realizes the splitting along the contact  type hypersurface

1451:   $(V=\partial U,\alpha=pdq|_V)$.

1452:   (see Section \ref{sec:almost}). Then according to Example \ref{ex:cotangent}

1453:   $W$ splits into $W_-=T^*(L)$ and $W_+=W\setminus L$. The almost

1454:   complex structure $J_-$ on $T^*(L)$ is compatible at infinity  with the

1455:   contact $1$-form   $\alpha=pdq|_{V}$. According to Corollary

1456:   \ref{cor:isolated} for any choice $\Gamma=\{\g_1,\dots,\g_k\}$ and

1457:   any $g\geq 0$ the

1458:   moduli spaces $\Mc_g(\Gamma;W_-,J_-)$ are empty, or

1459:   $0$-dimensional. One the other hand, Theorem \ref{thm:comp2}

1460:   together with Ruan's theorem guarantee the existence of a

1461:   rational holomorphic curve  with punctures through every point

1462:   of $L$. This contradiction proves that

1463:   $L$ cannot admit a metric

1464:   of negative curvature.  \qed

1465: 

1466: %%%%%%%%%%%%%%%%%%%%%%%%

1467: \subsection{ Coherent orientation of the moduli spaces

1468:   of holomorphic curves}\label{sec:orientation}

1469: 

1470: 

1471:     To get started with the algebraic formalism,

1472:     one first needs to orient moduli spaces

1473:     $\Mc(\Gamma^-,\Gamma^+)$ of holomorphic curves with punctures.

1474:     This problem is much easier in the case

1475:     of moduli spaces of closed holomorphic curves, because in that

1476:     case moduli spaces are even-dimensional and carry a canonical almost

1477:      complex structure (see Section \ref{sec:coherent-closed} below).

1478:     In our case we have to adapt the philosophy of {\it coherent orientations}

1479:      of the moduli spaces

1480:     borrowed from Floer homology theory (see \cite{Floer-Hofer}).

1481:     We sketch this approach in this section.

1482: 

1483: 

1484: \subsubsection{Determinants}

1485: 

1486: In order to separate the problems of orientation and

1487: transversality we are going to orient the determinant line bundles

1488: of the linearized $\overline\partial$-operators, rather than  the moduli spaces themselves.

1489: 

1490:  For a linear Fredholm operator $F:A\rightarrow

1491: B$ between Banach spaces we can define its determinant line

1492: $\det(F)$ by $$ \det(F)= (\Lambda^{max} \Ker (F))\otimes

1493: (\Lambda^{max} \Coker (F))^{\ast}. $$ We note that for the trivial

1494: vector space $\{0\}$ we have $\Lambda^{max}\{0\}= {\R}$.  An

1495: orientation for $F$ is by definition an orientation for the line

1496: $\det(F)$. In particular, given an isomorphism $F$ we can define a canonical

1497: orientation given through the vector $1\otimes 1^{\ast}\in {\R

1498: }\otimes {\R}^{\ast}$.

1499: 

1500: Given a continuous family $F=\{F_y\}_{y\in Y}$ of Fredholm

1501: operators, parameterized by a topological space $Y$,  the

1502: determinants of operators $F_y$ form   a line bundle

1503: $\det(F)\rightarrow Y$. The fact that this is a line bundle in a

1504: natural way might be surprising since the dimensions of kernel and

1505: cokernel vary in general. This is however a standard fact, see for

1506: example \cite{Floer-Hofer}.

1507: 

1508: \subsubsection{Cauchy-Riemann Type Operators on Closed Riemann Surfaces}

1509: \label{sec:coherent-closed} Let $(S,j)$ be a closed, not

1510: necessarily connected Riemann surface and $E\rightarrow S$ a

1511: complex vector bundle. Denote by $X_E\rightarrow S$ the complex

1512: $n$-dimensional vector bundle whose fiber over $z\in S$ consists

1513: of all complex ant-linear maps $$ \phi:T_zS\rightarrow E_z,\ z\in

1514: S,\ \hbox{i. e.}\ J\circ\phi+\phi\circ j=0, $$ where $J$ is the

1515: complex structure on $E$. Fixing a connection $\nabla$ and a

1516: smooth $a\in\hbox{Hom}_{\mathbb R}(E,X_E)$ we can define a

1517: Cauchy-Riemann type operator $$L:C^{\infty}(E)\rightarrow

1518: C^{\infty}(X_E)$$ by the formula $$

1519: (Lh)(X)=\nabla_Xh+J\nabla_{jX}h +(ah)(X), $$ where $X$ is an

1520: arbitrary vector field on $S$. Since the space of connections is

1521: an affine space we  immediately see that the set ${\mathcal O}_E$

1522: of all Cauchy-Riemann type operators on $E$ is convex. For a

1523: proper functional analytic set-up, where we may chose H\"older or

1524: Sobolev spaces, the operator $L$ is Fredholm. By elliptic

1525: regularity theory the kernel and cokernel would be spanned always

1526: by the same smooth functions, regardless which choice we have

1527: made. The index of $L$ is given by the Riemann-Roch formula $$

1528: \hbox{ind}(L)= (1-g)  \hbox{dim}_{\mathbb R}(E) + 2 c(E), $$ where

1529: $c(E)$ the first Chern number $c_1(E)(S)$ of $E$. Here we assume

1530: $S$ to be a connected closed surface of genus $g=g(S)$.

1531: 

1532: 

1533:  Let   $\phi:(S,j)\rightarrow (T,i)$

1534:  be a biholomorphic map and $\Phi:E\rightarrow F$ a ${\mathbb

1535: C}$-vector bundle isomorphism covering $\phi$. Then $\Phi$ induces

1536: an isomorphism $$ { \Phi}_{\ast}:{\mathcal O}_E\rightarrow

1537: {\mathcal O}_F $$ in the obvious way. The operators   $(E,L)$ and

1538: $(F,K)$ are called isomorphic  if there exists ${

1539: \Phi}:E\rightarrow F$, so that ${\bf \Phi}_{\ast}(L)=K$. We will

1540: denote by $[E,L]$ the equivalence class  of an operator  $(E,L)$

1541: which consists of operators $(F,K)$, equivalent to $(E,L)$ under

1542: isomorphisms, {\it isotopic to the identity}, and by $[[E,L]]$ the

1543: equivalence class under the action of the full group of

1544: isomorphisms. The moduli space of equivalence classes $[[E,L]]$

1545: will be denoted by $\CR$, and the ``Teichmuller space" which

1546: consists of classes $[E,L]$ will be denoted by $\wt{\CR}$. An

1547: isomorphism $\Phi$ induces an isomorphism between the kernel

1548: (cokernel) of $L$ and ${ \Phi}_{\ast}L$ for every $L\in {\mathcal

1549: O}_E$, and hence one can canonically associate the determinant

1550: line to an isomorphism class, and thus define the {\it determinant

1551: line bundle} $\cV$ over the moduli space $\CR$.

1552:  Given an orientation $o$ for $L$ we obtain an induced

1553: orientation ${ \Phi}_{\ast}(o)$.  Let us note the following

1554: \begin{lemma}

1555:  The bundle $\cV$ is orientable.

1556: \end{lemma}

1557: \proof The lift $\wt\cV$ of the bundle $\cV$ to the Teichmuller

1558: space $\wt\CR$ is obviously orientable, because each connected

1559: component of the space $\wt\CR$ is contractible. However, one

1560: should check that an arbitrary isomorpism $\Phi:(E,L)\to (F,K)$

1561: preserves the orientation. This follows from the following

1562: observation. Any connected component of $\wt\CR$ contains an

1563: isomorphism class of a complex linear operator $(E,L_0)$,  and any

1564: two complex linear operators representing points in a given

1565: component of $\wt\CR$ are homotopic in the class of complex linear

1566: operators. The determinant of $(E,L_0)$ can be oriented

1567: canonically by observing that its kernel and cokernel are complex

1568: spaces.  Any isomorphism  maps a   complex linear operator to a

1569: complex linear operator and preserves its complex orientation.

1570: Hence, it preserves an orientation of the determinant line  of any

1571: operator $(E,L)$.

1572: \qed

1573: \medskip

1574: 

1575: We will call an orientation of $\cV$ {\it complex} if it coincides

1576: with    the complex orientation of   determinants   of complex

1577: linear operators.

1578: 

1579: 

1580: \bigskip

1581: 

1582: The  components of the space $\CR$ are

1583: parameterized by the topological type of the underlying surface

1584: $S$ and the isomorphism class of the bundle $E$.

1585: It turns out that the complex orientation of $\cV$ satisfies three

1586: {\it coherency} Axioms A1--A3 which we formulate below.

1587: They relate  orientations of $\cV$ over different components of

1588: $\CR$.

1589: Conversely, we will see that these axioms determine the

1590: orientation uniquely up to a certain normalization.

1591: 

1592: 

1593: 

1594: Given $(E,L)$ and $(F,K)$ over surfaces $\Sigma_0$ and $\Sigma_1$

1595:  we define a {\it disjoint union}

1596: $$ [E,L]\dot{\cup}[F,K] := [G,M]  $$ of $(E,L)$ and $(F,K)$ as a pair

1597: $(G,M)$, where $G$ is a bundle over     the disjoint union

1598: $\Sigma=\Sigma_0\coprod\Sigma_1$, so that $(G,M)|_{\Sigma_0}$ is

1599: isomorphic to $(E,L)$ and $(G ,M)_{\Sigma_1}$  is isomorphic to

1600: $(F,K)$. Clearly, the

1601: isomorphism class of a disjoint union is uniquely determined by the classes

1602: of $(E,L)$ and $(F,K)$.

1603: Thus, we have a well-defined construction called {\it disjoint

1604: union}:

1605:  The determinant $\det\Sigma$ is canonically isomorphic to $\det

1606:  L\otimes\det K$, and hence the orientations    $o_K$ and $o_L$

1607:  define

1608:  an orientation $o_K\otimes o_L$ of    $ \Sigma $.

1609:  Our first axiom  reads

1610:  \medskip

1611: 

1612: \noindent{\textsf{Axiom C1}}. For any disjoint union $

1613: [G,M]=[E,L]\dot{\cup}[F,K]$ the orientation $o_M$ equals

1614:  $o_K\otimes o_L$.

1615:  \bigskip

1616: 

1617: 

1618: 

1619: Given $(E,L)$ and $(F,K)$, where $E$ and $F$ are  bundles over $S$

1620: of possibly different rank, we can define an operator $(E\oplus

1621: F,L\oplus K)$. There is a canonical map $$

1622: \hbox{det}(L)\otimes\hbox{det}(K)\rightarrow \hbox{det}(L\oplus

1623: K),

1624: $$ and thus  given orientations $o_L$ and $o_K$ we obtain $o_L\oplus o_K$.

1625: 

1626: 

1627: \bigskip

1628: \noindent{\textsf{Axiom C2}}.$$o_{L\oplus K}=o_L\oplus o_K.$$

1629:  \bigskip

1630: 

1631: To formulate the third axiom, we need a construction, called {\it

1632: cutting and pasting}.

1633: 

1634: Let $(E,L)$ be given and assume that

1635: $\gamma_1,\gamma_2:S^1\rightarrow S$ be real analytic embeddings

1636: with mutually disjoint images. Assume that $\Phi:E|_{\gamma_1}

1637: \rightarrow E|_{\gamma_2}$ is a complex vector bundle isomorphism

1638: covering $\sigma=\gamma_2\circ\gamma_{1}^{-1}$. The maps

1639: $\gamma_1$ and $\gamma_2$   extends as holomorphic

1640: embeddings $\bar{\gamma}_j:[-\varepsilon,\varepsilon]\times S^1

1641: \rightarrow S$ for a suitable small $\varepsilon>0$, so that the

1642: images are still disjoint. Locally, near $\gamma_j$ we can

1643: distinguish the left and the right side of $\gamma_j$. These sides

1644: correspond to the left or the right part of the annulus

1645: $[-\varepsilon,\varepsilon]\times S^1$. Cutting $S$ along the

1646: curves $\gamma_j$ we obtain a compact Riemann surface $\bar{S}$

1647: with boundary. Its boundary components are $\gamma_{j}^{\pm}$,

1648: $j=1,2$, where $\gamma^{\pm}_{j}$ is canonically isomorphic to

1649: $\gamma_j$. The vector bundle $E$ induces a vector bundle

1650: $\bar{E}\rightarrow \bar{S}$. We define a space of smooth sections

1651: $\Gamma_{\Delta}(\bar{E})$ as follows. It consists of all

1652: smooth sections $\bar{h}$ with the property that $$

1653: \bar{h}|_{\gamma_{j}^{-} }= \bar{h}|_{\gamma_{j}^{+}}\ \hbox{for}\

1654: j=1,2. $$ Then $L$ induces an operator

1655: $\bar{L}:\Gamma_{\Delta}(\bar{E}) \rightarrow

1656: \Gamma(X_{\bar{E}})$. The operators  $L$ and $\bar{L}$

1657: have naturally isomorphic kernel and cokernel. So an orientation $o$ of $\hbox{det}(L)$

1658: induces one of $\hbox{det}(\bar{L})$. The boundary condition

1659: $\Delta$ can be written in the form

1660: \begin{eqnarray*}

1661: \left[\begin{array}{cc} 1&0\\ 0&1

1662: \end{array}\right]

1663: \cdot\left[\begin{array}{c}

1664: \Phi(\gamma_{1}(t)\bar{h}\circ\gamma_{1}^{-}(t)\\

1665: \bar{h}\circ\gamma_{2}^{-}(t)

1666: \end{array}\right]

1667: =\left[\begin{array}{c}

1668: \Phi(\gamma_1(t))\bar{h}\circ\gamma^{+}_{1}(t)\\

1669: \bar{h}\circ\gamma^{+}_{2}(t)

1670: \end{array}\right]

1671: \end{eqnarray*}

1672: We introduce a parameter depending boundary condition by

1673: \begin{eqnarray*}

1674: \left[\begin{array}{cc} cos(\tau)&sin(\tau)\\ -sin(\tau)&cos(\tau)

1675: \end{array}\right]

1676: \cdot\left[\begin{array}{c}

1677: \Phi(\gamma_{1}(t)\bar{h}\circ\gamma_{1}^{-}(t)\\

1678: \bar{h}\circ\gamma_{2}^{-}(t)

1679: \end{array}\right]

1680: =\left[\begin{array}{c}

1681: \Phi(\gamma_1(t))\bar{h}\circ\gamma^{+}_{1}(t)\\

1682: \bar{h}\circ\gamma^{+}_{2}(t)

1683: \end{array}\right]

1684: \end{eqnarray*}

1685: for $\tau\in [0,\frac{\pi}{2}]$. For all these boundary conditions

1686: $L$ induces an operator, which is again Fredholm of the same

1687: index. For every $\tau$ we obtain a Cauchy-Riemann type operator

1688: from $\Gamma_{\Delta_{\tau}}(\bar{E}) $ to $\Gamma(X_{\bar{E}})$.

1689: Note that for a section $h$ satisfying the boundary condition

1690: $\Delta_{\tau}$ the section $ih$ satisfies the same boundary condition.

1691: On the other hand for

1692: $\tau=\frac{\pi}{2}$ we obtain a Fredholm operator whose

1693: kernel and cokernel naturally isomorphic to the kernel on cokernel of

1694:  a   Fredholm operator

1695: on a new closed surface. Namely identify $\gamma^{+}_{1}$ with

1696: $\gamma^{-}_{2}$ and $\gamma^{+}_{2}$ with $\gamma^{-}_{1}$. For

1697: the bundle $\bar{E}$ we identify the part above $\gamma^{+}_{1}$

1698: via $\Phi$ with the part over $\gamma_{2}^{-}$ and we identify the

1699: part above $\gamma^{-}_{1}$ with $-\Phi$ to the part above

1700: $\gamma^{+}_{2}$. The latter surface and bundle we denote by

1701: $E_{\Phi}\rightarrow S_{\Phi}$ and the corresponding operator by

1702: $L_{\Phi}$. Letting the parameter run we obtain starting with an

1703: orientation $o$ for $L$ an orientation $o_{\Phi}$ for $L_{\Phi}$.

1704: If $o$ is the complex orientation it is easily verfied that

1705: $o_{\Phi}$ is the complex orientation as well. We  say that the operator

1706: $L_{\Phi}$  is an operator obtained

1707: from $L$ by cutting and pasting.  This operator $L_\Phi$ has the

1708: same index as $L$,  and the component of $[L,\Phi]$ in $\wt\CR$

1709: depends only the isotopy classes of the embeddings $\gamma_1$ and

1710: $\gamma_2$.

1711: 

1712: 

1713: 

1714: \bigskip

1715: 

1716: \noindent{\textsf{Axiom C3}}. $$o_{L_\Phi}=o_\Phi\, .$$

1717:  \bigskip

1718: 

1719:  Note that we have to require here that the parts of $L$ over the curves

1720:   $\gamma_1$ and $\gamma_2$ are isomorphic via the gluing data.

1721: It is straightforward to check that

1722: \begin{theorem}\label{thm:complex-coherent}

1723: The complex orientation of $\cV$ is coherent, i.e. it satisfies

1724: Axioms C1--C3.

1725: \end{theorem}

1726: 

1727: Let us point out a simple

1728:  \begin{lemma}

1729:   Let $(E,L)$ be  an isomorphism then the orientation by $1\otimes 1^{\ast}$

1730:   of the $\det

1731: L={\mathbb R}\otimes {\mathbb R}^{\ast}$ defines the complex

1732: orientation of $\cV$ over the component of $[E,L]$.

1733: \end{lemma}

1734:  The following

1735: theorem gives the converse of Theorem \ref{thm:complex-coherent}.

1736: 

1737: \begin{theorem}\label{thm:coherent-complex}

1738: Suppose that a coherent orientation of $\cV$  coincides with

1739: the complex orientation for  the trivial line bundle over $S^2$

1740: and for

1741: the line bundle over $S^2$ with Chern number $1$. Then the

1742: orientation is complex.

1743: \end{theorem}

1744: 

1745: \proof

1746: Let us first observe that according to Theorem

1747: \ref{thm:complex-coherent} the disjoint union, direct sum and

1748: cutting and pasting procedures preserve the class of complex

1749: orientations.  Consider the pair $(E_0,L_0)$, where  $E_0$

1750: is the trivial bundle $ S^2\times {\mathbb C}\to S^2$ and $L_0$ is the

1751: standard Cauchy-Riemann operator. Then the $\ind L_0=2$. Take

1752: small loops around north pole and south pole on $S^2$ and

1753:  identify the trivial bundles over these loops. Now

1754: apply the cutting and pasting procedure and Axiom C3 to obtain the

1755: disjoint union of the trivial bundle over the torus and the

1756: trivial bundle over $S^2$. Hence we can use Axiom C1 to obtain an

1757: induced orientation for the Cauchy-Riemann operator on the trivial

1758: bundle over $T^2$. Taking appropriate loops we obtain orientations

1759: for all trivial line bundles over Riemann surfaces of arbitrary

1760: genus. Using direct sums

1761:  and disjoint unions constructions, and applying Axioms C1 and C2  we see that  the orientation of

1762:   all trivial bundles

1763: of arbitrary dimensions over Riemann surfaces of arbitrary genus

1764: are complex. Let $E_1$ be the bundle over $S^2$ with Chern number

1765: $1$.  Then we can use C3 to glue  two copies of $(E_1,L_1)$ to

1766: obtain the complex orientation of the disjoint union of a Cauchy

1767: Riemann operator on the trivial bundle and one on the bundle with

1768: Chern number $2$. Now it is clear that  the given coherent

1769: orientation has to  be complex over all components of the moduli space

1770: $\CR$. \qed

1771: 

1772: In the next section we extend the coherent orientation from

1773: Cauchy-Riemann type operators over closed surfaces to

1774:   a  special class of Cauchy-Riemann type

1775: operators on Riemann surfaces with punctures.

1776: 

1777: \subsubsection{A special class of Cauchy-Riemann type

1778: operators on punctured Riemann surfaces}\label{sec:coherent-punct}

1779: 

1780: Let us view ${\mathbb C}^n$ as a real vector space equipped with

1781: the Euclidean  inner product which is the real part of the

1782: standard Hermitian inner product. We define a class of

1783: self-adjoint operators as follows. Their domain in

1784: $L^2(S^1,{\mathbb C}^n)$

1785:  is

1786: $H^{1,2}(S^1,{\mathbb C}^n)$ of  Sobolev maps $h:S^1\rightarrow

1787: {\mathbb C}^n$.  The operators have the form

1788: \begin{equation}\label{eq:asymptotic}

1789: (Ax)(t)=-i\frac{dx}{dt}-a(t)x,

1790: \end{equation}

1791:  where $a(t)$ is a smooth loop of

1792: real linear self-adjoint maps. We assume that $A$ is

1793: non-degenerate in the sense that $Ah=0$ only has the trivial

1794: solution, which just means that the time-one map $\psi(1)$ of the

1795: Hamiltonian flow

1796: \begin{equation}\label{eq:asympt-flow}

1797: \begin{split}

1798:  \dot{\psi}(t)=&i a(t)\psi(t),\\

1799:  \psi(0)=&\Id\\

1800:  \end{split}

1801:  \end{equation}

1802:  has no eigenvalues equal  to $1$.

1803:   In particular, $A:H^{1,2}\rightarrow L^2$ is an

1804: isomorphism.

1805: 

1806: Given a smooth vector bundle $E\rightarrow S^1$ we can define

1807: $H^{1,2}(E)$ and $L^2(E)$ and a class of operators $B$ by

1808: requiring that $A=\Phi B\Phi^{-1}$ for  an Hermitian

1809: trivialization $\Phi$ of the bundle $E$.  We shall call such

1810: operators {\it asymptotic}, for reasons which will become clear

1811: later.

1812: 

1813:  As it was defined in Section \ref{sec:holomorphic} above,

1814:   an asymptotically marked

1815: punctured Riemann surface is a triplet $(S, \bx , \mu )$, where

1816: $S=(S,j)$ is a closed Riemann surface,

1817:  $\bx =\{x _1,\dots,x _{s }\}$

1818: is the set  of  punctures, some of them called positive, some

1819: negative, and $\bmu=\{\mu_1,\dots,\mu_s\}$   is the set of

1820: asymptotic markers, i.e. tangent rays,

1821:  or equivalently oriented

1822: tangent lines at the punctures.

1823: 

1824: One can introduce near each puncture $x_k\in\bx$ a holomorphic

1825: parameterization, i.e. a holomorphic map $h_k:D\to S$ of the unit

1826: disk $D$ such that $h_k(0)=x_k$ and the asymptotic marker $\mu_k$

1827: is tangent to the ray $h_k(r),\;r\geq 0$. We assume that the

1828: coordinate neighborhoods $\cD_k=h_k(D)$ of all the punctures are

1829: disjoint. Then we define $\sigma_k:{\mathbb R}^+\times

1830: S^1\rightarrow {\mathcal D}\setminus\{0\}$ by $$

1831: \sigma_k(s,t)=h_k(e^{\pm{2\pi(s+it)}}), $$ where the sign $-$ is

1832: chosen if the  puncture $x_k$ is positive, and the sign $+$ for

1833: the negative puncture. We will refer to $\sigma_k$ as holomorphic

1834: polar coordinates adapted to $(x_k,\mu_k)$. Given two adapted

1835: polar coordinate systems $\sigma $ and $\sigma'$ near the same

1836: puncture $x\in\bx$

1837:   we observe that the transition map

1838: (defined for $R$ large enough) $$ \sigma

1839: ^{-1}\circ\sigma':[R,\infty)\times S^1 \rightarrow

1840: [0,\infty)\times S^1 $$ satisfies for every multi-index $\alpha$

1841: $$ D^{\alpha}[\sigma^{-1}\circ\sigma'(s,t)-(c+s,t)] \rightarrow 0

1842: $$ uniformly for $s\rightarrow\infty$, where $c$ is a suitable

1843: constant. The main point is the fact that there is no phase shift

1844: in the $t$-coordinate.

1845: 

1846: Given $(S, \bx , \mu )$ we associate to it a smooth surface

1847: $\bar{S}$ with boundary compactifying the punctured Riemann

1848: surface $S\setminus\bx$ by adjoining a circle for every puncture.

1849: Each circle has a distinguished point $0\in S^1={\mathbb

1850: R}/{\mathbb Z}$. Namely for each   positive puncture we compactify

1851: ${\mathbb R}^+\times S^1$ to $[0,\infty]\times S^1$, where

1852: $[0,\infty]$ has the smooth structure making the map $$

1853: [0,\infty]\rightarrow [0,1]:s\rightarrow s(1+s^2)^{-\frac{1}{2}},\

1854: \infty\rightarrow 1 $$ a diffeomorphism. We call

1855: $S^{+}_{k}=\{\infty\}\times S^1$ the circle at infinity associated

1856: to $(x_k,\mu_k)$. For negative punctures we compactify at

1857: $-\infty$ in a similar way.

1858: 

1859: \begin{definition}

1860: A smooth complex vector bundle $E\rightarrow  (S, \bx , \mu )$ is

1861: a smooth vector bundle over $\hat{S}$ together with Hermitian

1862: trivializations $$ \Phi_k:E|_{S_{k}}\rightarrow S^1\times{\mathbb

1863: C}^n. $$ An isomorphism between two bundles $E$ and $F$ over

1864: surfaces $S$ and $T$ is a a complex vector bundle isomorphism

1865: $\Psi:E\to F$ which covers   a biholomorphic map

1866: $\phi:(S,j)\rightarrow (T,i)$, preserves punctures and the

1867: asymptotic markers  (their numbering and signs)  and respects the

1868: asymptotic trivializations.

1869: \end{definition}

1870: 

1871: Define as in Section \ref{sec:coherent-closed} above   the bundle

1872: $X_E\rightarrow \bar{S}$. Set $\dot{S}=S\setminus\Gamma$. We

1873: introduce the Sobolev space $H^1(E)$ which  consists of all

1874: sections $h$ of $E\rightarrow \dot{S}$ of class $H^{1,2}_{loc}$

1875: with the following behavior  near   punctures.

1876:   Suppose, that $x$ is a positive puncture and   $\sigma$

1877: is an adapted system of  holomorphic polar coordinates. Pick a

1878: smooth   trivialization $\psi$ of $E\rightarrow \bar{S}$ over

1879: $[0,\infty]\times S^1$ (in local coordinates) compatible with the

1880: given asymptotic trivialization. Then the map $(s,t)\rightarrow

1881: \psi(s,t)h\circ\sigma(s,t)$ is assumed to belong to

1882: $H^{1,2}({\mathbb R}^+\times S^1,{\mathbb C}^n)$. A similar

1883: condition is required for   negative punctures. In a similar

1884: way we define  the space $L^2(X_E)$. Observe that defacto we use

1885: measures which are infinite on $\dot{S}$ and that the

1886: neighborhoods of punctures look  like half-cylinders.

1887: 

1888: A Cauchy-Riemann type operator $L$ on $E$ has the form $$ (Lh)X =

1889: \nabla_Xh+J\nabla_{jX}h +(ah)X, $$ where $X$ is a vector field on

1890: $S$.   We require, however a particular behaviour of $L$  near the

1891: punctures. Namely, regarding   $E$ as a trivial bundle

1892: $[0,\infty]\times \C^n$ with respect to the chosen polar

1893: coordinates and trivialization near say a positive  puncture we

1894: require that $$ (Lh)(  s,t )(\frac{\partial }{\partial s})

1895: =\frac{\partial h}{\partial s} - A(s)h, $$ where $A(s)\rightarrow

1896: A_{\infty}$ for an asymptotic operator $A_{\infty}$,   as  it was

1897: previously introduced.

1898: \begin{theorem}

1899: The operator $L$ is Fredholm.

1900: \end{theorem}

1901:  The index of $L$ can be computed in terms of Maslov indices of

1902:  the asympotic operators (and, of course, the first Chern class of

1903:  $E$ and the topology of $S$).

1904: 

1905: 

1906: 

1907: Similar to the case of closed surfaces we define the notion of

1908: isomorphic pairs $(E,L)$ and $(F,K)$, where we emphasize the

1909: importance of the compatibility of the asymptotic trivializations,

1910: define the moduli space $\CR_{\punct} \supset \CR_{\cl}$ and the

1911: Teichmuller spaces $\wt\CR_{\punct} \supset\wt\CR_{\cl}$, and

1912: extend the determinant line bundle $\cV$ to $\CR_{\punct}$ and

1913: $\wt\cV$ to $\wt\CR_{\punct}$. The bundle $\wt\cV$ is orientable

1914: by the same reason as in the

1915:   case of closed surfaces: each component of the space $\wt\CR_{\punct}$ is

1916: contractible. However, unlike the closed case,  there is no

1917: canonical (complex) orientation of $\wt\cV$. Still due to the

1918: requirement

1919:  that

1920: isomorphisms  preserves the end structure of the operators, one

1921: can deduce the fact that even isotopically non-trivial isomorphisms

1922: preserve the orientation of $\wt\cV$, which shows that the bundle

1923: $\cV$ over $\CR_{\punct}$ is orientable.

1924: 

1925: 

1926:  Let us review now Axioms C1--C3 for the line bundle

1927: $\cV$ over $\CR_\punct$. The formulation of Disjoint Union Axiom

1928: C1

1929:  should be appended by the  following requirement. Let $(E,L)$ and $(F,K)$ be

1930:  operators

1931:  over the punctured

1932:  Riemann surfaces $(S,\bx=\{x_1,\dots,x_s\})$ and $(T,\by=\{y_1,\dots,y_t\})$, respectively.

1933:  Then $(E,L)\dot{\cup}(F,K)$ is an operator over the surface

1934:  $S\coprod T$ with the set of punctures ${\mathbf z}=\{x_1,\dots,x_s,y_1,\dots,y_t\}$.

1935:  The disjoint union operation is associative, but not necessarily

1936:  commutative (unlike the case of closed surfaces).

1937:   Axioms C2 and C3  we formulate without any changes compared to the closed case.

1938:  By a coherent orientation of the bundle $\cV$ over $\CR_{\punct}$

1939:  we will mean any orientation of $\cV$ which satisfies Axioms

1940:  C1--C3.

1941: 

1942: Take  the trivial (and globally trivialized) line bundle $E_0=

1943: \C\times\C$ over the 1-punctured  Riemann sphere $\C=\C P^1\setminus \infty$.

1944: For any

1945: admissible asymptotic operator $A$ we choose a Cauchy-Riemann

1946: operator $L^\pm_A$ on $E_0$ which has $A$ as its asymptotics at $\infty$. The

1947: superscript $\pm$ refers to the choice of $\infty$ as the positive

1948: or negative puncture.

1949:  Note, that the   component  of $([E_0,L^\pm_A)$ in the  moduli spaces

1950:   $\CR$   is uniquely determined

1951:   by the  homotopy class $[A]$ of  the asymptotic operator $A$ in the space

1952:   of {\it non-degenerate} asymptotic operators.

1953: 

1954: The following theorem describes all possible coherent orientations

1955: of  the line bundle $\cV$ over $\CR$.

1956: 

1957: 

1958: \begin{theorem}\label{thm:coherent}

1959: Let us choose an  orientation $o^\pm_A$ of the operator

1960: $(E_0,L^\pm_A)$ for a representative $A$ of each homotopy class

1961: $[A]$ of non-degenerate  asymptotic operators. Then this choice

1962: extends to the unique coherent orientation of the bundle $\cV$

1963: over $\CR_{\punct}$, which coincide with the complex orientation

1964: over $\CR_{\cl}$.

1965: \end{theorem}

1966: Thus  there are infinitely many coherent orientations of $\cV$

1967: over $\CR_{\punct}$ unlike the  case of closed surfaces, when there are

1968: precisely four.

1969: 

1970: We sketch below  the proof of Theorem \ref{thm:coherent}.

1971:  First, similar to the case of closed surfaces, it is sufficient

1972:  to consider only operators on the trivial, and even globally

1973:  trivialized bundles.

1974:  Next take the disjoint union of

1975:  $(E_0,L^-_A)$ and $ (E_0,L^+_A)$, consider two circles $\g^\pm$ around the  punctures

1976:  in the two copies of $\C$ and apply the cutting/pasting

1977:  construction along these circles. As the result we get a disjoint

1978:  union of an operator $\wt L_A$ on the trivial line  bundle over the closed Riemann

1979:  sphere, and an operator $\overline L_A$ over  the cylinder $C=S^1\times\R$,

1980:  which we view as the Riemann sphere

1981:  with  two

1982:  punctures $x_1=\infty$ and $x_2=0$ and consider $x_1$ as a positive puncture

1983:  and $x_2$ as a negative one. The operator $\overline L_A$ has the

1984:  same asymptotic operator $A$ at both punctures.

1985:  Then Axioms C1 and C3 determine the orientation of $\overline

1986:  L_A$, because for  the operator $\widetilde L_A$ we have chosen

1987:  the complex orientation. Notice that if one glue $L_A^\pm$ in the opposite order,

1988:  then we get an operator $\overline L_A'$ which has the reverse

1989:   numbering of the punctures.

1990:   The orientation of $\overline L_A'$  determined by the gluing  may be the same, or opposite as

1991:   for the operator $\overline L_A$, depending on the parity of the Conley-Zehnder

1992:   index of the asymptotic operator $A$.\footnote{

1993:   The operator $\overline L_A$

1994:  is homotopic to an isomorphism, and thus has a canonical

1995:  orientation

1996:  $1\otimes 1^{\ast}$. If we insist on that normalization, than

1997:  our construction would determine the orientation of $L_A^-$ in

1998:  terms of $L_A^+$.}

1999: 

2000: 

2001: Consider now an arbitrary operator  $(E,L)$  acting on sections of a complex line bundle

2002: $E$ over   a punctured

2003: Riemann surface

2004:  $(S,\bx,\bmu)$ with $\bx=\{x_1,\dots,x_s\}$, $E=S\times\C^n$, and the  asymptotic operators

2005:  $A_1,\dots,A_s$ at the corresponding punctures. For each

2006:  $i=1,\dots,s$ consider an operator $(E_0,L_i=L_{A_i}^\pm)$, where

2007:  $E_0=\C\times\C^n$,  the sign $+$ is chosen if the puncture

2008:  $x_i$ is negative, and the sign $+$ is chosen  otherwise.

2009:  Using Axiom C1 we orient the operator $

2010:  (E,L)\dot{\cup}(E_0,L_s)$, and then choosing circles around the

2011:  puncture $x_s$ and $\infty$ apply the cutting/pasting procedure.

2012:  As the result we get the disjoint union of an operator $L'$ over

2013:  the Riemann surface with punctures $(x_1,\dots,x_{s-1})$ and the

2014:  operator $\overline L_A$, or $\overline L_A'$ depending on

2015:  whether the puncture $x_s$ was negative, or positive. Hence

2016:  Axioms C1 and C3 determine the orientation of $L'$ in terms of the orientation

2017:   of $L$. Repeating the

2018:  procedure for the punctures $x_{s-1},\dots,x_1$ we express

2019:   the orientation of $L$ in terms of the complex

2020:  orientation of an operator over the closed surface.

2021: 

2022:  It remains to observe that if $E$ is a trivial complex bundle of rank $r>1$,

2023:   then any asymptotic operator

2024:  $A$ can be deformed through non-degenerate asymptotic operators to an operator

2025:   $\widetilde A$ which is split into the direct

2026:  sum of asymptotic operators on the trivial complex line. Hence we can use the direct sum

2027:   axiom C2 to orient determinants of  operators  acting on bundles of arbitrary rank.

2028:  \bigskip

2029: 

2030: \subsubsection{Remark about  the coherent orientation for

2031:  asymptotic  operators with symmetries}

2032: Let $A$ be an asymptotic operator

2033: given by the formula (\ref{eq:asymptotic}),

2034: where the loop $a(t),\,t\in S^1=\R/\Z$, of symmetric matrices has a symmetry

2035: $a(t+1/2)=a(t),\;t\in\R/\Z$. Let $L$ be a Cauchy-Riemann type

2036: operator on a bundle $E\to S$, which has $A$ as its

2037: asymptotic

2038: operator at a puncture $x\in S$ with an asymptotic marker $\mu$.

2039: Let $L'$ be an operator which differs from $L$ by

2040: rotating by the angle $\pi$ the marker $\mu$ to a marker $\mu'$,

2041: with the corresponding change of the trivialization near the

2042: puncture.

2043:  Let $h:S\to S$ be  a diffeomorphism  which rotates

2044:  the polar coordinate

2045:  neighborhood $\cD$ of the punctures $x$ by $\pi$, and is fixed

2046:  outside a slightly larger neighborhood.

2047:  Then the operator $h_*L'$ has the same asymptotic data as $L$ and

2048:  the  isomorphism classes $[E,L]$ and $[E,h_*L']$ belongs to the

2049:  same component of the space $\wt\CR$. Given a coherent

2050:  orientation of $\cV$, do the orientations $o_L$ and $o_{h_*L}$

2051:  coincide? It turns out that

2052:  \begin{lemma}\label{lm:bad-orbits}

2053:  Let $\Psi$ be the time-one map of the linear Hamiltonian flow $\psi(t)$,

2054:  defined by the equation {\rm (\ref{eq:asympt-flow})}. The orientations

2055:  $o_L$ and $o_{h_*L}$ coincide if and only if the number of  real eigenvalues  of  $\Psi$

2056:  (counted with multiplicities)  from the interval $(-1,0)$ is

2057:  even.

2058:  \end{lemma}

2059: This lemma is the reason why we excluded certain periodic orbits

2060: from $\Pc$ in Section \ref{sec:dynamics} above. See also Remarks

2061: \ref{rem:bad-orbits1} and \ref{rem:bad-orbits2}.

2062: 

2063: 

2064: \subsubsection{Coherent orientations of moduli spaces}

2065: The moduli spaces of holomorphic curves which we need to orient are zero sets of nonlinear

2066: Cauchy-Riemann type operators, whose linearizations are related to

2067: operators of the kind we described (see below for more details).

2068: In general, the moduli spaces are neither manifolds nor orbifolds,

2069: due to the fact  that Fredholm sections cannot be made transversal

2070: to the zero section by changing natural parameters like the almost

2071: complex structure or the contact form. Such a transversality will

2072: only be achievable by making abstract perturbations, leading to

2073: virtual moduli spaces. Those virtual spaces will be the moduli

2074: spaces which will provide us with the data for our constructions.

2075: Nevertheless the Fredholm operators occurring in the description

2076: of the virtual moduli spaces will only be compact perturbations of

2077: the Cauchy-Riemann type operators, and hence the orientation scheme

2078: for these virtual moduli spaces does not differ from the case of moduli spaces

2079: of holomorphic curves.

2080: 

2081: 

2082: A moduli space  $\Mc(\Gamma^+,\Gamma^-;W,J)$ of  holomorphic

2083: curves in a directed symplectic cobordism $(W=\ora{V^-V^+},J)$ is

2084: a fiber bundle over the corresponding moduli space of Riemann

2085: surfaces. Its base is   a complex orbifold, and hence it is

2086:  canonically oriented, while the

2087: fiber over a point $S$, where $S$ is a Riemann surface with a

2088: fixed conformal structure

2089:  and positions

2090: of punctures,  can be viewed as the space  solutions of the $\overline

2091: \partial_J$-equation.

2092:  If the  transversality is achieved than the  tangent bundle  of a

2093:  moduli space  $\Mc(\Gamma^+,\Gamma^-;W,J)$  arise as  the kernel of the linearized

2094:  surjective  operator $\overline\partial_J$.

2095: The linearization  of $\overline

2096: \partial_J$

2097:   at a point $f\in \Mc(\Gamma^+,\Gamma^-;W,J)$

2098: is a Fredholm operator in a suitable functional analytic setting.

2099: This set-up involves Sobolev spaces with suitable asymptotic

2100: weights derived from the non-degeneracy properties of the periodic

2101: orbits. It is a crucial observation, again a corollary of the

2102: behaviour near the punctures, that up to a compact perturbation,

2103: the operator $L$ splits into two operators $L'$ and $L''$, where

2104: $L'$ is   a complex linear operator acting on the complex line

2105: bundle $T(S)$ of the Riemann surface $S$, and $L''$ is a

2106: Cauchy-Riemann  type on the the bundle $E$, such that $T(S)\oplus

2107: E=f^*(TM)$. This operator is usually only real linear, but most

2108: importantly it is of the kind we just described in our linear

2109: theory.

2110:   The trivialization of $E$ near the punctures

2111: is determined by the chosen in \ref{sec:holomorphic}

2112: trivialization

2113:  of the contact structure near

2114: periodic orbit of   the Reeb vector fields on $V^\pm$, and the

2115: asymptotic operators are determined by the linearized Reeb flow

2116: near the periodic orbits. We have $\det L=\det L'\otimes\det L''$.  But  $\det L'$ has a canonical complex

2117: orientation, and   hence the orientation for $\det L$ is  determined by   the orientation of

2118:  $\det L''$. Therefore, a choice of

2119:   a coherent orientation of $\cV$ over $\CR$ determines in the

2120: transversal case the orientation of  all the moduli spaces

2121: $\Mc(\Gamma^+,\Gamma^-;W,J)$.

2122: 

2123: 

2124: %%%%%%%%%%%%%%%%%%%%%%%

2125: 

2126: 

2127: 

2128: 

2129: 

2130: 

2131: 

2132: 

2133:    \subsection{First attempt of algebraization:

2134:    Contact Floer homology}  \label{sec:Floer}

2135:    \subsubsection{Recollection of finite-dimensional Floer theory} \label{sec:Morse}

2136:    Let us  first recall   the basic steps in defining a Floer homology theory in the simplest case of a

2137: Morse function $f$ on a finite-dimensional orientable closed manifold

2138: $M$. We refer the reader

2139: to  Floer's  original papers (see, for instance, \cite{Floer}), as well as an excellent exposition by D. Salamon  \cite{Salamon:Floer}

2140: for the general theory.

2141: 

2142:  First,  one forms a  graded complex $C(f,g)$ generated by

2143: critical points  $c_1,

2144: \dots,c_N$ of      $f$, where the grading is given by the Morse index of critical points.

2145:  Next, we  choose a  generic Riemannian metric $g$ on $M$  which satisfy the Morse-Smale condition of transversality

2146:  of stable and unstable varieties

2147:  of critical points. This enables us to define  a differential $d=d_{f,g}:C(f,g)\to C(f,g)$ by counting  gradient trajectories

2148:  connecting critical points   of neighboring indices:

2149:  $$d(c_i)=\sum L_i^jc_j,$$ where the sum is taken

2150:  over all critical points $c_j$  with   $\ind c_j=\ind c_j-1$. The coefficient  $L_i^j$

2151:  is the {\it  algebraic number}  of    trajectories    connecting $c_i$ and $c_j$.

2152:   This means that the trajectories are counted

2153:  with signs.   In the finite-dimensional case the signs could  be determined

2154:  as follows. For each critical point we orient arbitrarily its stable manifold. Together with

2155:  the orientation of $M$ this allows us to orient  all unstable manifolds,

2156:   as well as the intersections

2157:  of stable and unstable ones.   If $\ind c_j=\ind c_i-1$ then

2158:  the stable manifold of $c_i$ and the unstable manifold

2159:  of $c_j$ intersect along finitely many trajectories which   we want to count,

2160:   and hence each of these trajectories

2161:  gets an orientation. Comparing this orientation with the one given by the direction of the gradient $\nabla f$ we can associate

2162:  with every trajectory a sign. \footnote{The generalization of this procedure to an infinite-dimensional

2163:  case is not straightforward, because stable and unstable manifolds not only can become

2164:  infinite-dimensional, but in most interesting cases cannot  be defined at all.  On the other hand,

2165:   the moduli spaces of

2166:  gradient trajectories connecting pairs of critical points  (which in the finite-dimensional case

2167:  coincide with the intersection

2168:  of stable and unstable manifolds of the critical points) are often defined, and one can use the coherent

2169:  orientation scheme, similar to the one described in Section \ref{sec:orientation}

2170:  above for the moduli spaces of holomorphic curves,

2171:   to define their orientation.}

2172: 

2173:   To show that $d^2=0$, which then  would allow us to define the homology group

2174:   $H_*(C(f,g),d)$, we proceed as follows.

2175: Let us observe that  the coefficients $K_i^j$ in the expansion $d^2(c_i)=\sum K_i^jc_j$

2176:   count  the algebraic number of broken gradient trajectories  $(\d_{il},\d_{lj})$  passing through an intermidiate

2177:   critical point $c_l$, $l=1,\dots N$.

2178:  But each broken trajectory    $(\d_{il},\d_{lj})$, which connects critical points whose

2179:  indices differ by $2$, is a boundary point

2180:  of the $1$-dimensional manifold of smooth trajectories connecting $c_i$ and $c_j$. The  algebraic number of boundary points of a compact

2181:  $1$-dimensional  manifold is, of course, equal to $0$. Hence $K_i^j=0$, and thus $d^2=0$.

2182: 

2183: 

2184:  Next we want to show that the homology group $H_*(C(f,g),d)$ is  an invariant of the manifold $M$

2185:  (of course, in the case we consider it is just $H_*(M)$), i.e. it is independent of the choice of the function

2186:  $f$ and the Riemannian metric $g$.  The proof of the invariance consists of three steps.

2187: \medskip

2188: 

2189: \noindent \textsf{Step 1.}

2190: Let us show that given a homotopy  of functions $F=\{f_t\}_{t\in[0,1]}$,   and a homotopy  of Riemannian metrics

2191:  $G=\{g_t\}_{t\in[0,1]}$, one can define a homomorphism $\Phi=\Phi_{F,G}:C(f_1,g_1)\to C(f_0,g_0)$ which commutes with

2192:  the boundary homomorphisms $d_0=d_{f_0,g_0}$ and $d_1=d_{f_1,g_1}$, i.e.

2193:   \begin{equation}\label{eq:Morse-boundary}

2194:   \Phi\circ d_1-d_0\circ\Phi=0\,.

2195:   \end{equation}

2196:          To construct $\Phi$ we consider the product

2197:          $W=M\times\R $ and, assuming that the homotopies $\{f_t\}$ and $\{g_t\}$ are

2198:          extended to all $t\in\R$  as independent  of $t$ on

2199:          $(-\infty,-1]\cup[1,\infty)$, we

2200:           define on $W$

2201:         a function, still denoted by $F$,

2202:         by the formula

2203:          \begin{equation*}

2204:            F(x,t)=

2205:            \begin{cases}

2206:            f_0(x)+ct,&  \;\;t\in (\infty,0) ;\\

2207:            f_t(x)+ct,&  \;\;t\in[0,1]\;\;  ;\\

2208:            f_1(x)+ct,&  \;\;t\in(0,\infty) ,\\

2209:                         \end{cases}

2210:            \end{equation*}

2211: where the constant $c$ is chosen to ensure that $\frac{\partial

2212: F}{\partial t}>0$. Similarly, we use the family of Riemannian

2213: metrics $g_t$

2214: to define a metric $G$ on $W$  which is equal to $g_t$ on $M\times

2215: t$ for all $t\in\R$, and such that $\frac{\partial}{\partial t}$ is the unit vector field

2216: orthogonal to the slices $M\times t,t\in\R$.

2217: The gradient trajectories of $\nabla F$ converge to

2218: critical points of $f_1$ at $+\infty$ , and  to the critical points

2219: of $f_0$ at $-\infty$. For a generic choice of $G$ the moduli space of the

2220: (unparameterized)

2221: trajectories connecting two critical points, $c^1$ of $f_1$ and

2222: $c^0 $ of $f_0$, is a compact $k$-manifold with boundary with corners,

2223: where $k=\ind c^1-\ind c^0$. Hence,

2224: similarly

2225:  to the above definition of the differential $d$, we can define a homomorphism

2226: $\Phi:C(f_1,g_1)\to C(f_0,g_0)$ by  taking an algebraic count of gradient trajectories between

2227: the critical point of $f_1$ and $f_0$ of the same Morse index,

2228: i.e.

2229: $\Phi(c^1_j)=\sum \wt L^i_jc^0_j$. The identity

2230: (\ref{eq:Morse-boundary}) comes from the  description of the  boundary

2231: of the $1$-dimensional moduli spaces of trajectories of $\nabla

2232: F$. Notice that the function $F$ has no critical points, and hence

2233: a family of gradient trajectories cannot converge to a broken

2234: trajectory in a usual sense. However, this can happen {\it at

2235: infinity}. Let us recall that the function $F$ and the metric $G$

2236: are cylindrical outside of  $M\times[-1,1]$. Hence

2237: away from a compact set  a gradient

2238: trajectory of $F$  projects to a gradient

2239: trajectory of $f_0$ or $f_1$. When  the projection, say at $+\infty$,

2240:  of a sequence  $\d_n:\R\to W$ of trajectories of $\nabla F$   converges  to

2241: a broken trajectory  of    $\nabla f_1$ this can be interpreted as a

2242: splitting at $+\infty$. This phenomenon is very similar to  the one  described

2243: for the moduli spaces of holomorphic curves in Section

2244: \ref{sec:compact}. Namely, there exist    gradient  trajectories $\d:\R\to W $  of $\nabla F$,

2245: and $\d':\R\to M_1$ of $\nabla f_1$,  such that

2246: \begin{description}

2247: \item{$-\;$} $\d_n\to\d$ uniformly on $(-\infty,C]$ for all $C$;

2248: \item{$-\;$} there exists a sequence $C_n\to+\infty$ such that

2249: $\d'_n(t)=\d_n(t+C_n)$ converges to $(\d'(t),t)$ uniformly on all

2250: subsets $[-C,\infty)$.

2251:  \end{description}

2252: 

2253: In this sense broken trajectories of the form $(\d,\d')$ and

2254: $(\d'',\d)$, where $\d''$ is a   trajectory of $\nabla f_0$ form

2255: the boundary of the $1$-dimensional moduli spaces of trajectories

2256: of $\nabla F$ connecting critical points  $c^1$ of $f_1$ and

2257: $c^0 $ of $f_0$  with $\ind c^1-\ind c^0=1$. Therefore the

2258: algebraic number of these trajectories equals $0$. On the other hand,

2259: this number is equal   to $\Phi\circ d_1-d_0\circ\Phi $  which yields

2260: the identity (\ref{eq:Morse-boundary}).

2261: \medskip

2262: 

2263: \noindent\textsf{Step 2.} Our next goal is to check that  if

2264: $(F_u,G_u), u\in[0,1],$  is a homotopy of homotopies which is

2265: constant outside of a compact subset of $W$, then

2266:   the homomorphisms $\Phi_0=\Phi_{F_0,G_0}$ and $\Phi_1=\Phi_{F_1,G_1}$ are related via the

2267: chain homotopy formula

2268: \begin{equation}\label{eq:Morse-chain}

2269: \Phi_1-\Phi_0=K\circ d_1+d_0\circ K,

2270: \end{equation}

2271: for a homomorphism $K:C(f_1,g_1)\to C(f_0,g_0)$. The space of all homotopies $(F,G)$

2272: connecting given pairs $(f_0,g_0)$ and $(f_1,g_1)$ is contractible,

2273:  and hence (\ref{eq:Morse-chain}) implies

2274:   that the homomorphism $\Phi_*:H_*(C(f_1,g_1),d_1)\to

2275: H_*(C(f_0,g_0),d_0)$ is independent of the choice of a homotopy

2276: $(F,G)$.

2277: 

2278: To prove (\ref{eq:Morse-chain}) one studies moduli spaces  of

2279: gradient trajectories  of the whole $1$-parametric family of

2280: functions $F_u$. For a generic choice of the homotopy one has

2281: isolated critical values of the parameter $u$  when appear

2282: {\it handle-slides}, i.e. gradient connections between critical

2283: points with the index difference $-1$.  By  counting these

2284: trajectories

2285: one can then define a

2286: homomorphism $K:C(f_1,g_1)\to\C(f_0,g_0)$ in exactly the same way as the homomorphism $\Phi$

2287: was  defined in Step 1

2288:    by counting   trajectories   with the index difference  $0$.

2289: 

2290:  The identity (\ref{eq:Morse-chain}) expresses  the fact that the broken

2291:  trajectories  of the form $(\d,\d')$ and $(\d'',\d)$, where $\d$

2292:  is a handle-slide trajectory and $\d'$ is a trajectory of $\nabla

2293:  f_1$, form the boundary of the moduli space  of   index $0$

2294:  trajectories in the family $(F_u,G_u)$. The difference in signs

2295:  in formulas (\ref{eq:Morse-boundary}) and (\ref{eq:Morse-chain})

2296:  is a reflection of the fact that the homomorphism $K$ raises the grading by

2297:  $1$, while $\Phi$ leaves it unchanged.

2298: 

2299: 

2300: 

2301: 

2302: 

2303: \medskip

2304: \noindent\textsf{Step 3.} Finally we need to show that

2305: \begin{equation}

2306: \label{eq:Morse-composition}

2307: (\Phi_{F,G})_*=(\Phi_{F',G'})_*\circ(\Phi_{F'',G''})_*,

2308: \end{equation}

2309: if

2310:  $(F,G)=\{f_t,g_t\}_{t\in[0,2]}$ is the composition of homotopies

2311:  $(F'',G'')=\{f''_t,g''_t\}_{t\in[0,1]}$ and

2312:  $(F',G')=\{f'_t,g'_t\}_{t\in[1,2]}$. To prove this we  view,

2313:  as in Step 1, the  homotopy $(F,G)$ as a function and a metric

2314:  on the cylinder $W=M\times\R$. Consider a deformation $(F_T,G_T)$ of $F,G$, by

2315:  cutting $W$ open along $M\times 1$ and inserting a cylinder

2316:  $M\times[0,T]$ of growing height $T$ with the function and the

2317:  metric independent of the coordinate $t$. When $T\to+\infty$

2318:  the gradient trajectories of $F_T$ with respect to $G_T$  split  in

2319:  a  sense, similar to  the one explained in Step 2,

2320:  \footnote{See

2321:  also the discussion of  a similar phenomenon for the

2322:  moduli spaces of holomorphic curves   in Section \ref{sec:compact}

2323:  above.} into a ''broken trajectory"

2324:  $(\d'',\d')$, where $\d'$  (resp.  $\d"$) is a   trajectory of $\nabla_{G'}F'$

2325: ( resp. $\nabla_{G''}F''$).  Consider the $1$-dimensional moduli  space $\Mc$ of

2326: trajectories  of $\nabla_{G_T}F_T, T\in[0,\infty),$ connecting a

2327: fixed

2328: critical point  $c=c^2$ of $f_2$  with an arbitrary critical point

2329: $c^0 $ of $f_0$  with $\ind c-\ind c^0=1$. Then the boundary of

2330: $\Mc$ consists of

2331: \begin{description}

2332: \item{ a) } all the trajectories of

2333: $\nabla_{G_0}F_0=\nabla_G F$ connecting $c^0$ and $c$; they are

2334: given by the expression $\Phi(c)$;

2335: \item{ b) } all the broken trajectories $(\d'',\d')$ described

2336: above, such that $\d''$ begins at $c^0$ and ends  at a critical point

2337: $c^1$ of $f^1$ which is, necessarily, of the same Morse index as

2338: $c^0$ and $c^2$, $\d'$ begins at $c^1$ and ends at

2339: $c$; these broken trajectories are described by the

2340: expression

2341: $\Phi_{F',G'}\big(\Phi_{F'',G''}(c)\big)$;

2342: \item{ c) } broken trajectories defined according to Step 2 for the

2343: $1$-dimensional family $F_T, T\in[0,\infty)$; they are described

2344: by the expression $K(d_0(c))+d_2( K(c))$ for some homomorphism

2345: $K:C(f_2,g_2)\to C(f_0,g_0)$.

2346: \end{description}

2347:  Thus the sum (taken with appropriate signs) of the three

2348:  expressions defined in a)--c)  equals $0$, and thus we get

2349:  \begin{equation*}

2350: \Phi_{F',G'}\big(\Phi_{F'',G''}(c)\big)-\Phi(c )=K\circ d_0(c)+d_2\circ

2351: K(c),

2352: \end{equation*}

2353: i.e. the homomorphisms $\Phi$ and

2354: $\Phi_{F',G'}\circ\Phi_{F'',G''}$ are chain homotopic, which

2355: yields formula (\ref{eq:Morse-composition}).

2356: 

2357: We can finish now the proof that the homology group  $H_*(C(f,g),d)$

2358: is independent of the choice of $f$ and $g$ as follows.

2359:  Given two pairs   $(f_0,g_0)$ and  $(f_1,g_1)$ we

2360:  first take any homotopy $(F,G)$  connecting  $(f_0,g_0)$ with  $(f_1,g_1)$, and also take  the

2361:  inverse homotopy $(\overline F,\overline G)$ connecting  $(f_1,g_1)$ with

2362:  $(f_0,g_0)$.

2363:   The composition $(\wt F,\wt G)$ of  the  homotopies $(F,G)$ and  $(\overline F,\overline G)$

2364:   connects  the pair $(f_0,g_0)$ with itself. According to Step 3

2365:   we have

2366:   $(\Phi_{\wt F,\wt G})_*=(\Phi_{F,G})_*\circ(\Phi_{\overline

2367:   F,\overline

2368:   G})_*$. On the other hand, we have  shown in Step 3  that the

2369:   homomorphism $(\Phi_{\wt F,\wt G})_*$ is independent of the

2370:   choice of a homotopy, connecting $(f_0,g_0)$ with itself, and

2371:   hence it equals the identity. Therefore, we conclude that $(\Phi_{F,G})_*$ is

2372:  surjective, while $ (\Phi_{\overline

2373:   F,\overline

2374:   G})_*$ is injective. Taking the composition of homotopies  $(\overline F,\overline G)$

2375:   and $(F,G)$ in the opposite order we prove that both

2376:   homomorphisms are bijective.

2377:   \bigskip

2378: 

2379:  A. Floer  discovered  that the

2380:     finite-dimensional scheme  which we explained  in this section  works,

2381:      modulo some

2382:     analytic  complications, for several geometrically interesting functional on infinite-dimensional

2383:     spaces. For instance,

2384:   in the  symplectic Floer homology theory one deals with

2385:   critical points of the action functional. Its critical points are periodic orbits of

2386:   a Hamiltonian system, while for an appropriate choice of

2387:   a metric  and an almost complex structure the  gradient trajectories    can be interpreted

2388:   as holomorphic

2389:   cylinders   which connect these trajectories.

2390:   The role of broken trajectories is played here  by split holomorphic cylinders,

2391:    and    finite-dimensional  compactness theorems are replaced by  the  highly

2392:    non-trivial Gromov  compactness theorem for holomorphic curves.

2393: 

2394:    In the  rest of this section we explore  the  Floer-theoretic approach for

2395:    the problem of defining invariants of contact manifolds. We will see that this approach

2396:    works only  in a very special and restrictive   situation. However, the  general

2397:    algebraic formalism of SFT, though quite different, has  a distinctive flavor    of

2398:    a Floer homology theory.

2399: 

2400: 

2401: 

2402: \subsubsection{Floer homology for the Action functional}

2403:    Let us make an attempt to define invariants of contact manifolds

2404:    in the spirit of Floer homology theory. Let $(V,\xi)$     be a contact manifold

2405:          with a fixed contact form $\alpha$ and an  almost complex structure

2406:          $J:\xi\to\xi$,   compatible

2407:     with the symplectic form $d\alpha|_\xi$.  Then $J$ and $d\alpha$

2408:     define a Riemannian metric on the vector bundle $\xi$

2409:     by the formula $g(X,Y)=d\alpha(X,JY)$ for any vectors $X,Y\in\xi$.

2410:     We extend $g$ to the whole tangent bundle $T(V)$ by declaring the vector field $R_\alpha$ to be the unit normal

2411:     field to $\xi$.  Consider the free loop space

2412:     $$\Lambda(V)=\{u:S^1=\R/\Z\to V\},$$

2413: and define the {\it action functional}

2414:    \begin{equation}

2415:    S:\Lambda(V)\to\R\quad\hbox{  by the formula}\quad

2416:     S(\g) =\int\limits_\gamma \alpha .

2417:     \end{equation}

2418:    The least action principle tells us that the critical points of the functional $S$

2419:    are, up to parameterization,   the

2420:    periodic orbits of the Reeb field $R_\a$.

2421: 

2422: 

2423:      The metric $g$ on $T(V)$ defines a        metric on $\Lambda(V)$

2424:       and thus allows us to consider

2425:     gradient trajectories of the action functional connecting critical points of $V$.

2426:       The gradient direction $\nabla S(u)$, $u\in \Lambda(V)$, is given

2427: by the vector field $J\pi( \frac{du}{dt})$, where $\pi:T(V)\to\xi$

2428: is the projection along the Reeb direction, so that a gradient

2429: trajectory $u(t,s),\,t\in\R/\Z,s\in\R,$ is given by the equation

2430: \begin{equation}\label{eq:CR}

2431: \frac{\partial u}{\partial s}(t,s)=J\pi\big(\frac{\partial u}{\partial

2432: t}(t,s)\big)\,.

2433:  \end{equation}

2434: Equation (\ref{eq:CR}) has a flavor of a Cauchy-Riemann equation.

2435: We want to modify it into a genuine one.

2436: Namely, consider the Cauchy-Riemann equation

2437:  \begin{equation*}

2438:   \frac{\partial U}{\partial t}(t,s)=J\frac{\partial U}{\partial

2439:   s}(t,s)

2440:   \end{equation*}

2441:   for $U(s,t)=   \big(u(s,t),\varphi(s,t)\big)\in V\times\R$.

2442:   It  can be rewritten

2443:   as a system

2444: \begin{equation}\begin{split}\label{eq:CR2}

2445: \frac{\partial u}{\partial s}(t,s)&=J\pi\big(\frac{\partial u}{\partial

2446: t}(t,s)\big)+\frac{\partial \varphi}{\partial t}(t,s)R_\a(u(t,s)\\

2447:  \frac{\partial \varphi}{\partial s}(s,t)&=-\big\langle\frac{\partial

2448: u}{\partial t}(t,s),R_\a(u(t,s))\big\rangle\,.\\

2449: \end{split}

2450:  \end{equation}

2451: Notice that $dS(\nabla S+\psi R_\a)\geq 0$ for any function $\psi(t,s)$.

2452:  Hence,

2453:    the first

2454:   equation of the system (\ref{eq:CR2})  can be viewed as the flow equation of the

2455:   gradient-like vector-field $\nabla S+\frac{\partial \varphi}{\partial

2456:   t} R_\a$. Trajectories  of  this gradient like field   connecting critical points

2457:   $\gamma^-,\gamma^+$ of the action functional   correspond to

2458:    elements

2459:   of the moduli space  $\Mc_0(\gamma^-,\gamma^+;W,J)$, and

2460:   therefore the Floer homology philosophy (\cite{Floer}),

2461:   which we described above in the finite-dimensional

2462:   case, suggests the following construction.

2463: 

2464: 

2465: 

2466: 

2467: 

2468:   Let us associate a variable $q_\gamma$ with every periodic orbit $\gamma\in\Pc_\a$

2469:   and assign to it the grading $$\deg{ q_\gamma}=\CZ(\gamma)+(n-3).$$

2470:   The choice of the   constant $n-3$ is not important

2471:   for purposes of this definition, but

2472:  it will become important for generalizations considered in the second

2473:  part of this paper.

2474: 

2475:   Let  $A$ be  the group algebra $ \C\,[H_2(V)]$. We will fix a basis

2476:   $A_1,\dots, A_N$ of $H_2(V;\C\,)$ and identify each homology class $\sum d_iA_i$ with its {\it degree}

2477:   $d=(d_1,\dots, d_N)$. Thus  we can view the algebra $A$ as

2478:   the  algebra of  Laurent polynomials

2479:    of $N$ variables $z_1,\dots,z_N$ with complex coefficients, and write its elements

2480:    in the form $\sum a_dz^d,$ where $z^d=z_1^{d_1}\cdots

2481:    z_N^{d_N}$. The variables $z_i$ are also considered graded,

2482:    $\deg\,z_i=-2c_1(A_i),\,i=1,\dots,N$.

2483:    Consider a complex $\fF$ generated by the (infinitely many) graded variables $q_\gamma$ with coefficients

2484:   in the graded algebra $A$,

2485:   and define  a differential $\partial: \fF\to\fF $ by the formula:

2486:   \begin{equation} \label{eq:Floer1}

2487:   \partial q_\gamma= \sum\limits_{\g',d}\frac{n_{\gamma,\gamma',d}}{\kappa_{\g'}}z^d  q_{\gamma'},

2488:   \end{equation}

2489:   where $\kappa_{\gamma'}$ denotes the multiplicity of the orbit

2490:   $\gamma'$,

2491:    the sum is taken over all trajectories

2492:   $\g'\in\Pc_\a $ and $d=(d_1,\dots,d_N)$ with $$\CZ(\g')=\CZ(\g)+2\langle

2493:   c_1,d\rangle-1,$$

2494:    and the coefficient $n_{\gamma,\gamma',d}$ counts

2495:   the algebraic number of components of the $0$-dimensional moduli space

2496:   $\Mc^d_0(\gamma',\gamma;W,J)/\R

2497:   $.\footnote{

2498:   Let us recall that  according to our definition of the moduli

2499:   space $\Mc^d_0(\gamma',\gamma;W,J)/\R

2500:   $ the coefficient $n_{\gamma,\gamma',d}$ counts

2501:   equivalence classes of holomorphic curves {\it with asymptotic

2502:   markers}, and hence     each holomorphic cylinder connecting $\g$ and

2503:   $\g'$ is counted $\kappa_\g\kappa_{\g'}$ times, unless the

2504:   cylinder itself is

2505:   multiply covered. The role of the denominators $\kappa_{\g'}$

2506:   in formula (\ref{eq:Floer1}), as well in a similar

2507:   formula (\ref{eq:Floer2}) below, is to correct this ``over-counting". }

2508:   Notice that the Liouville flow of the vector field  $\frac{\partial}{\partial t}$

2509:     defines  a $\R$-action on

2510:   the moduli spaces    $\Mc^d_0(\gamma',\gamma;W,J)$, which makes the

2511:    $1$-dimensional components of the moduli

2512:   spaces

2513:     canonically oriented.

2514:    Comparing this orientation

2515:   with the coherent orientation we produce  signs   which we use in the formula

2516:   (\ref{eq:Floer1}).

2517: 

2518: To simplify the  assumptions in the propositions which we formulate below

2519: we will  assume for the rest of this section that $c_1|_{\pi_2(V)}=0$.

2520:  This assumption allows us to define for any

2521:  contractible periodic orbit $\g$ the Conley-Zehnder index

2522:  $\CZ_{\disk}(\g)$ computed with respect to {\it any disk}

2523:  $\Delta$ spanned by $\g$ in $V$. We denote

2524:  $\deg_{\disk}(\g)=\CZ_{\disk}(\g)+n-3$.

2525: 

2526:   \begin{proposition}\label{prop:Floer1}

2527:   If for a  contact form $\alpha$ the Reeb

2528:   field $R_\alpha$ has no contractible periodic orbits $\g\in\Pc_\a$  with

2529:   $\deg_{\disk}(\g)=1$, then

2530:   $\partial^2=0$.

2531:   \end{proposition}

2532: 

2533: {\sl Sketch of the proof}. Similarly to the finite-dimensional

2534: case considered in Section \ref{sec:Morse} above the identity

2535: $\partial^2=0$ in Floer homology  is equivalent to the fact that

2536: the codimension $1$ stratum of the compactified moduli spaces

2537: $\Mc^d_0(\gamma',\gamma)$   consists of broken

2538: trajectories, which in our case are represented by the height $2$

2539: stable curves $(f_1,f_2)$, $f_1\in\Mc^{d'}_0(\gamma',\gamma'')/\R$, $f_2\in

2540: \Mc^{d''}_0(\gamma'',\gamma)/\R$, where $d=d'+d''$. However, in  the general case a sequence of

2541: holomorphic cylinders in $\Mc^d_0(\gamma',\gamma)$ can split into curves

2542: different from cylinders, as it is stated in Proposition \ref{prop:boundary} and  Corollary

2543: \ref{cor:pos-puncture}. But

2544:  if this happens then the  first-floor curve

2545: $f_1$ must have  a component which is conformally equivalent to $\C\,$ and

2546: asymptotically cylindrical over a  contractible orbit at

2547: $+\infty$. Moreover, if $(f_1,f_2)$ belongs to a top-dimensional

2548: stratum of the boundary of the moduli space

2549: $\Mc^d_0(\gamma',\gamma)$, then $\deg_{\disk}(\g)=1$, which

2550: contradicts our assumption.

2551: 

2552: \begin{remark}\label{rem:bad-orbits1}{\rm

2553: Let us recall that we excluded from $\Pc$ certain ``bad" periodic

2554: orbits (see  the footnote in Section \ref{sec:dynamics}).

2555: However on the boundary of the moduli space

2556: $\Mc^d_0(\gamma',\gamma)$  there could be a stratum

2557:  which consists of height $2$

2558: stable curves $(f_1,f_2)$, $f_1\in\Mc^{d'}_0(\gamma',\gamma'')$, $f_2\in

2559: \Mc^{d''}_0(\gamma'',\gamma')$, where the orbit $\gamma''$ is

2560: one of  the bad orbits which we excluded from $\Pc$. The

2561: orbit $\g''$ has even multiplicity $2k$, and hence on the boundary

2562: of   $\Mc^d_0(\gamma',\gamma)$ there are $2k$ strata which

2563: correspond

2564: to

2565: $2k$ different possible positions of the asymptotic marker at

2566: the punctures mapped to $\g''$. The Poincar\'e return map of the Reeb flow

2567: along the orbit $\g''$ has an odd number of eigenvalues in the interval $(-1,0)$,

2568: and hence  according to Lemma \ref{lm:bad-orbits}  the coherent orientation

2569: will automatically assign to these orbits opposite signs, which means that these strata

2570: will not contribute to the sum  (\ref{eq:Floer1}). This explains why the exclusion of bad orbits

2571: is {\em possible}. Remark \ref{rem:bad-orbits2} below explains why this

2572: exclusion is {\em necessary}.}

2573: \end{remark}

2574: 

2575: Now we follow Steps 1--3 in Section \ref{sec:Morse} above to show the independence

2576: of   the homology group   $$\oplus H_k(\fF ,\partial )= \Ker\partial /{\mathrm {Im}}\partial, $$

2577:     graded by the degree $k$,  of the choice of

2578:   a  nice contact form $\a$ and  a compatible almost complex

2579:   structure  $J$.

2580: 

2581: Suppose now that we have a  directed symplectic cobordism

2582: $W=\ora{V^-V^+}$,  and $J$  is a compatible almost complex structure on $W$.

2583: Suppose that the inclusions $V^\pm\hookrightarrow W$ induce

2584: isomorphisms on $2$-dimensional homology.  Then we can define

2585:   a homomorphism $\Phi=\Phi_W:\fF^+\to\fF^-$ by the formula

2586: \begin{equation}\label{eq:Floer2}

2587: \Phi( q_\gamma)= \sum\limits_{\g',d} \frac{1}{\kappa_{\gamma' }}n_{\gamma,\gamma',d}z^dq_{\gamma'},

2588: \end{equation}

2589:   where the sum is taken over all trajectories $\g'\in\Pc^- $  and $d$

2590:   with

2591:   $\CZ(\g')=\CZ(\g)+2\langle c_1,d\rangle$, and

2592:   the coefficient $n_{\gamma,\gamma',d}$ counts

2593:   the algebraic number of points of the compact $0$-dimensional moduli space $\Mc^d_0(\gamma',\gamma;W,J)

2594:   $.

2595:    If the   condition on the second homology is not satisfied then the above construction

2596:   gives us only a correspondence, rather than

2597:   a homomorphism. See Section \ref{sec:composition}  for the discussion of a

2598:    more general case.

2599: 

2600: \begin{proposition}\label{prop:Floer2}

2601: Suppose that the contact forms   $\a^\pm$ associated to the

2602: ends satisfy the condition  $\deg_{\disk}(\g)\neq 0,1$ for any {\em contractible in $W$} periodic orbit

2603:  $\g\in\Pc^\pm$. Then the homomorphism $\Phi_W$ commutes

2604: with $\partial $.

2605: \end{proposition}

2606: \begin{proposition}\label{prop:Floer3}

2607:  Let $J_t,t\in[0,1] $, be a family of  almost complex structures compatible with

2608:  the directed symplectic cobordism

2609: $W=\ora{V^-V^+}$. Suppose  that the forms   $\a^\pm$ associated to the

2610: ends satisfies the condition  $\deg_{\disk}(\g)\neq -1,0,1$ for any  {\em contractible in $W$} periodic orbit

2611:  $\g\in\Pc^\pm$. Then the  homomorphisms

2612:  $\Phi_0=\Phi_{W,J_0}$ and  $\Phi_1=\Phi_{W,J_1}$  are chain

2613:  homotopic, i.e there exists a homomorphism $\Delta:\fF^+\to\fF^-$

2614:  such that $\Phi_1-\Phi_0=\partial \Delta +\Delta\partial$.

2615:  \end{proposition}

2616: \begin{proposition}\label{prop:Floer4}

2617: Given two cobordisms $W_1$ and $W_2$, and a compatible almost complex structure

2618: $J$ on the composition $W_1\circledcirc W_2$, the homomorphism

2619: $\Phi_{W_1\circledcirc W_2}$ is chain-homotopic to

2620: $\Phi_{W_1}\circ\Phi_{W_2}$.

2621: \end{proposition}

2622: 

2623: Together with an obvious remark that for the cylindrical cobordism   $W_0$

2624: the homomorphism $\Phi_{W_0}$ is the identity,

2625:   Propositions \ref{prop:Floer1}--\ref{prop:Floer3}

2626:   imply that if a contact structure $\xi$ on $V$ admits a {\it

2627:   nice} contact form, i.e. a form without contractible periodic

2628:   orbits of index $-1,0$ and $1$, then the {\it contact homology

2629:   group}

2630:   $$\oplus HC_k(V,\xi)=\oplus H_k(\fF ,\partial ) $$

2631:    is well defined and

2632:     independent of the choice of

2633:   a  nice contact form and  a compatible almost complex

2634:   structure

2635:   (however, if $H_2(V)\neq 0$ and/or $H_1(V)\neq0$ it depends on a

2636:   choice of spanning surfaces $F_\g$ and the  framing of the bundle

2637:   $\xi$ over basic loops).

2638:    Similarly to what was explained in the sketch of

2639:      the proof  Proposition

2640:       \ref{prop:Floer1} the  ``niceness" assumptions   guarantees

2641:       that the  top codimension strata   on the boundary of the involved

2642:       moduli spaces consist of height 2 cylindrical curves,

2643:       and thus         the proofs of Propositions  \ref{prop:Floer2}--\ref{prop:Floer4}

2644:       may precisely follow the standard scheme of the Floer

2645:       theory (see \cite{Floer, Salamon:Floer}).

2646: \begin{remark}\label{rem:bad-orbits2}{\rm

2647: Similarly to what we explained in Remark \ref{rem:bad-orbits1} the

2648: coefficient $n_{\g,\g'}$ in the definition (\ref{eq:Floer2}) of

2649: $\Phi$ equals $0$ if at least one of the orbits $\g,\g'$ is ``bad''.

2650: Hence, in the presence of ``bad" orbits the

2651: homomorphism $\Phi$  could never be  equal to the  identity, even

2652: for the cylindrical cobordism. This explains why

2653: the exclusion of ``bad" periodic orbits is {\em necessary}.}

2654: \end{remark}

2655: 

2656: 

2657: 

2658: 

2659:     Besides the degree (or Conley-Zehnder)  grading,

2660:    the contact homology group is graded by elements of $H_1(V)$, because

2661:    the boundary operator preserves the homology class of a periodic orbit.  We will denote the part of  $HC_*(V,\xi)$

2662:    which correspond to a class $a\in H_1(V)$ by $HC_*(V,\xi|a)$.

2663:   One can  similarly construct a contact homology group

2664:   $HC^{\contr}_*(V,\xi)$,

2665:   generated  only by contractible periodic orbits,  which is another invariant

2666:   of

2667:   the contact manifold $(V,\xi)$.

2668: 

2669: Contact structures which admit nice contact forms do exist, as it

2670: is

2671: illustrated by examples in Section \ref{ex:Floer}  below. However, the

2672: condition of existence of a nice form is too restrictive. The

2673: general case leads to an algebraic formalism developed in

2674: Sections \ref{sec:af-contact}--\ref{sec:composition} below.

2675: 

2676: \subsubsection{Examples}\label{ex:Floer}

2677: 

2678: 

2679: 

2680: \noindent 1. \textsf { Contact homology of the standard contact sphere $S^{2n-1}$.}

2681: 

2682:  Take the $1$-form $\alpha=\frac 12\sum (x_idy_i-y_idx_i)$,

2683:  which is a primitive of the standard symplectic structure

2684:  in $\R^{2n}$. Its restriction  to a generic  ellipsoid

2685:  $$S=\{\sum\frac{x_i^2+y_i^2}{a_i^2}=1\}$$ is a  nice contact form for the standard contact structure

2686:  $\xi$ on the sphere $S=S^{2n-1}$. The form $\alpha|_S$ has

2687:  precisely one periodic orbit for each  Conley-Zehnder index

2688:  $n +2i-1$ for $i=1,\dots,$. Hence the  contact homology group

2689:  $HC_*(S,\xi)$ has one generator in each  dimension $2i, i\geq n-1$.

2690:  See also the discussion in Section \ref{sec:Bott} below.

2691:  \medskip

2692: 

2693: \noindent 2. \textsf{ Contact homology of Brieskorn spheres.}

2694: 

2695:  Ilya Ustilovsky  computed (\cite{Ustilovsky}) the

2696:  contact homology of certain Brieskorn spheres.

2697: 

2698:  Let us  consider the Brieskorn manifold

2699:  \begin{equation*}

2700:  \Sigma(p,\underbrace{2,\dots,2}_n)=

2701:  \{z_0^p+\sum\limits_1^n  z_j^2=0\}\cap \{\sum\limits_0^n|z_j|^2=1\}\subset \C^{n+1}.

2702:  \end{equation*}

2703:  $\Sigma(p,\underbrace{2,\dots,2}_n)$ carries a

2704:   canonical contact structure  as a strictly pseudo-convex hypersurface in a complex manifold.

2705: 

2706:  Suppose that $n=2m+1$ is odd, and $p\equiv 1\,\mod\,8.$

2707:  Under this assumption  $\Sigma(p,\underbrace{2,\dots,2}_n)$

2708:  is diffeomorphic to $S^{2n-1}$ (see \cite{Brieskorn}).

2709:   However, the following theorem of Ustilovsky implies that     the contact structures on

2710:    Brieskorn spheres

2711:  $\Sigma(p,\underbrace{2,\dots,2}_n)$ and $\Sigma(p',\underbrace{2,\dots,2}_n)$ are not isomorphic, unless $p=p'$.

2712: This result should be confronted with a computation of Morita

2713: (\cite{Morita} ), which implies

2714:  that  the formal homotopy class (see Section \ref{sec:prelim} above) of the contact structure on $\Sigma(p,\underbrace{2,\dots,2}_n)$

2715:   is   standard, provided $p\equiv 1\,\mod\, 2(2m!)$. Hence, Ustilovsky's theorem provides infinitely many non-isomorphic contact

2716:   structures on $S^{4m+1}$ in the standard formal homotopy class.

2717: 

2718:   \begin{theorem} {\rm (I. Ustilovsky, \cite{Ustilovsky})}

2719:   The contact homology $$HC_*\left(\Sigma(p,\underbrace{2,\dots,2}_n)\right)$$ is defined, and

2720:   the dimension  $$c_k=\dim\,HC_k\left( \Sigma(p,\underbrace{2,\dots,2}_n)\right) $$

2721:   is given by the formula

2722:   \begin{equation*}

2723:   c_k  =

2724:   \begin{cases}

2725:   0,& k \;\;\hbox {is odd or}\;\;k<2n-4,\\

2726:   2,& k=2\left[ \frac{2N}{p}\right] +2(N+1)(n-2),\;\hbox{for  }\;\;

2727:   N\geq 1,\,2N+1\notin p\Z,\\

2728:   1,& \hbox{in all other cases.}\\

2729:   \end{cases}

2730:   \end{equation*}

2731:   \end{theorem}

2732: 

2733: \medskip

2734: 

2735: \noindent 3. \textsf{ Contact homology of boundaries

2736: of subcritical Stein manifolds.}

2737: 

2738:  A co-oriented contact manifold $(V,\xi)$ is called Stein-fillable if it can be realized as a strictly pseudoconvex

2739:  boundary of a complex manifold $W$,

2740:  whose interior is Stein, and if the co-oriented contact structure $\xi$ coincides with the canonical contact structure

2741:  of a strictly pseudo-convex hypersurface.   We say that $(V,\xi)$ admits a

2742:  subcritical Stein filling if the corresponding Stein manifold $\Int W$ admits an exhausting plurisubharmonic function

2743:  without critical points of dimension $\dim_\C(W)$. If $\dim V>3$ then one can equivalently require

2744:  that $W$ deformation  retracts to   a CW-complex of dimension $<\dim_\C W$ (see \cite{Eliash-Stein}).

2745: 

2746:  Mei-Lin Yau studied in her  PhD thesis \cite{MLYau} contact homology of contact manifolds

2747:  admitting a subcritical Stein filling.

2748:  Here is her result.

2749:  \begin{theorem}\label{prop:MLYau} {\rm (Mei-Lin Yau,  \cite{MLYau})}

2750:  Let $(V,\xi)$ be a  contact manifold of dimension $2n-1$ which admits a subcritical Stein filling $W$.

2751:  Suppose that $c_1(V)=0$ and $H_1(V)=0$. Let $c_1,\dots, c_k$ be generators of $H_*(W)$.

2752:   Then  the contact homology $HC_*(V)$ is defined  and generated

2753:   by elements

2754:   $ q_{i,j}$ of  degree $\deg q_{i,j}= 2(n+i-2)-\dim c_j$, where  $j=1,\dots, k$, and

2755:   $i\geq 1$.

2756:   \end{theorem}

2757: 

2758: \medskip

2759: 

2760: \noindent 4. \textsf{ Contact homology of $T^3$ and its coverings.}

2761: \medskip

2762: 

2763:  Set $\a_n=\cos 2\pi nz\,dx+\sin 2\pi nz\,dy$. This contact form descend

2764:  to the $3$-torus $T^3=\R^3/\Z^3$ and defines there a contact structure $\xi_n$.

2765:  The structure $\xi_1$ is  just the canonical contact structure on $T^3$ as the space

2766:  of co-oriented contact elements of $T^2$. The form  $\a_n$ for $n>1$  is

2767:  equal to the pull-back $\pi^*_n (\a_1)$, where $\pi_n:T^3\to T^3$

2768:  is the covering

2769:  $(x,y,z)\mapsto(x,y,nz)$. Notice that all structures $\xi_n$

2770:  are homotopic as plane field to the foliation $dz=0$.

2771: 

2772:  \begin{theorem}\label{thm:T3}

2773:  The contact homology group $HC_*(T^3,\xi_n|w)$, where $w$ is the homology class $(p,q,0)\in H_1(T^3)$,

2774:  is isomorphic to $\Z^{2n}$.

2775:  \end{theorem}

2776:  In particular we get as a corollary a theorem of E. Giroux:

2777:  \begin{corollary} {\rm (E.  Giroux, \cite{Giroux})}

2778:  The contact structures $\xi_n$, $n=1,\dots,$ are pairwise non-isomorphic.

2779:  \end{corollary}

2780: 

2781:   The contact manifold $(T^3, \xi_1)$

2782:  is foliated by pre-Lagrangian

2783:  tori $L_{p,q}$, indexed by simple homology classes $ (p,q)\in H_1(T^2)$. Each torus

2784:  $L_{p,q}$ is foliated by  the $S^1$-family of lifts of

2785:   closed geodesics which represent the class $(p,q)$.

2786:   Thus for any given $ (p,q)\in H_1(T^2)$ (even when  $(p,q)$ have common divisors)

2787:    the set of closed orbits

2788:   in  $\Pc_{\a_1}$ which represent the class $(p,q,0)\in H_1(T^3)$

2789:    is  a circle $S_{p,q}$, and for any $n\geq 1$ the set of closed orbits

2790:   in  $\Pc_{\a_n}$ which represent the class $(p,q,0)\in H_1(T^3)$

2791:   consists of $n$ copies $S^1_{p,q},\dots,  S^n_{p,q}$ of such circles.

2792:    The forms $\a_n$ have no contractible periodic orbits, but of course, they

2793:     are degenerate. To compute the contact homology groups, one can   either work directly with these

2794:     degenerate forms, as it is explained in Section \ref{sec:Bott} below, and

2795:     show that  $HC_*(T^3,\xi_n|w)=H_*(\bigcup\limits_1^n S^i_{p,q})=\Z^{2n}$, or  first  perturb the form $\alpha_1$,

2796:   and respectively all its covering forms $\a_n= \pi_n^*(\a_1)$, in order to substitute each circle $S^i_{p,q}$

2797:   by two non-degenerate periodic orbits, and then show that the orbits from  each of these pairs  are connected by precisely

2798:   two holomorphic cylinders, which cancel each other in the formula for the boundary operator $\partial$.

2799: 

2800: 

2801: \subsubsection {Relative contact homology and contact non-squeezing theorems}

2802: \label{sec:squeezing}

2803: Let us observe that  the  complex $(\fF,\partial)$  is filtrated

2804: by the values of the action  functional $S$, $\fF=\mathop{\bigcup}\limits_{a\in\R}\fF^a$, where

2805: the complex $\fF^a$

2806: is generated  by variables $q_\gamma$ with $S(\g)\leq a$. The differential

2807: $\partial$ respects this filtration, and hence descends to $\fF^b/\fF^a, a<b$.

2808:  Hence, one can define  the homology

2809: $H_*^{(a,b]}(\fF,\partial)=H_*(\fF^b/\fF^a,\partial)$ in the window $(a,b]\subset\R$. Of

2810: course, $H_*^{(a,b]}$ depends on a choice of a particular nice form

2811: $\alpha$. If $\alpha>\beta$ then we  have a map $\Phi_*:H_*^{(a,b]}(\fF,\partial;\alpha)\to

2812: H_*^{(a,b]}(\fF,\partial;\beta)$. We write $H^a$ instead of $H^{(-\infty,a]}$.

2813: 

2814: Consider now a contact manifold $(V,\xi)$ which is

2815:  either closed, or satisfies

2816: the following

2817:   {\it pseudo-convexity}  condition at infinity.

2818: A contact manifold $(V,\xi=\{\a=0\})$  with a fixed contact form $a$

2819: is called pseudo-convex at infinity

2820: if there exists a compatible almost complex structure

2821:  $J$ on the symplectization $V\times\R$ for which $V$ can be exhausted

2822:  by compact domains $V_i$ with smooth  pseudo-convex

2823:  boundary. A sufficient condition for pseudo-convexity is existence of

2824:   an exhaustion $V=\bigcup V_i$, such that for each $i=1,\dots, $

2825:  trajectories of the Reeb field

2826:  $R_\a|_{V_i}$ do not have interior tangency points with $\partial  V_i$.  For instance, for

2827:  the standard contact form  $\a=dz-\sum y_idx_i$

2828:  on $\R^{2n+1}$ the latter condition is satisfied for an exhaustion of $\R^{2n+1}$

2829:  by round balls, and  ence the standard contact form on $\R^{2n+1}$ is pseudo-convex at

2830:   infinity.

2831: 

2832: 

2833: 

2834: 

2835: Our goal is to define  a relative contact homology group $HC_*(V,U,\xi)$

2836:  for a relatively compact open subset

2837: $U\subset V$, so that  this group would be invariant    under a contact isotopy of $U$ in $V$.

2838: 

2839: Let us fix a contact form $\alpha$ on $V$  which satisfies the above pseudo-convexity

2840:  condition.

2841: Let us denote by $\Fc_{U,\a}$ the set of  $C^\infty$-functions $f:V\to[0,\infty)$

2842: which  are $\leq 1$ on   $U$,   and for which  the contact form

2843: $f\alpha$ is nice and pseudo-convex at infinity.\footnote{ Of course, the set

2844:  $\Fc_{U,\a}$ may be

2845: empty, because the niceness condition is very restrictive.

2846: In this case one needs to employ

2847: a more general construction from Section \ref{sec:3algebras}.}

2848:  Take a strictly increasing sequence

2849: of functions

2850: $f_i\in \Fc_{U,\a}$, such that

2851: \begin{description}

2852: \item{a)}

2853: $\max\limits_K f_i \mathop{\to}\limits_{i\to\infty}\infty$ for   each compact

2854:  set $K\subset (V\setminus

2855: \overline{U})$;

2856: \item{b)} $f_i|_U \mathop{\to}\limits_{i\to\infty} 1$ uniformly on

2857: compact sets.

2858: \end{description}

2859: 

2860: \begin{proposition}\label{prop:rel-homol}

2861: The limit

2862: \begin{equation}

2863: HC_*(V,U,\xi)=\lim\limits_{a\to+\infty}

2864: \lim\limits_{\leftarrow}HC^a_*(V,f_i\alpha)

2865: \end{equation}

2866: is independent of $\alpha$,

2867:  and

2868: thus it is  an invariant of  the contact pair $(V,U)$.

2869: A contact isotopy  $f_t:V\to V$ induces a  family of

2870: isomorphisms $(f_t)_*:H_*(V,U)\to H_*(V, f_t(U))$.

2871: An inclusion $i:U_1\mapsto U_2$ induces a homomorphism

2872: \begin{equation*}

2873: i_*: HC_*(V,U_1,\xi)\to HC_*(V,U_2,\xi).

2874: \end{equation*}

2875: \end{proposition}

2876: 

2877: 

2878: 

2879: \bigskip

2880: 

2881: One of the most celebrated results in Symplectic topology is

2882: Gromov's non-squeezing theorem which  states that one cannot

2883: symplectically embed a  $2n$-ball of radius $1$ into

2884: $\D^2_r\times\R^{2n-2}$ for $r<1$. Here $D^2_r$ denotes a

2885:  $2$-disk of radius $r$ and $\D^2_r\times\R^{2n-2}$ is endowed

2886:  with the product of standard symplectic structures.

2887:   Because of the conformal character of contact geometry one

2888:   cannot expect as strong non-squeezing results

2889:   for

2890:   contact manifolds: one can embed any domain in the standard

2891:   $\R^{2n-1}$ in an arbitrary small ball. However, it turns out

2892:   that it is not always possible  to realize a contact squeezing via a contact

2893:   isotopy inside a manifold with  a non-trivial first Betti number.

2894: 

2895: As an example, consider the $1$-jet bundle $$V=J^1(\R^n,S^1)=T^*(\R^n)\times S^1$$

2896:  of

2897: $S^1$-valued functions with its standard contact structure $\xi$,

2898: given by the contact form $\alpha=dz-\sum y_idx_i$, $(x,y)\in

2899: \R^{2n}=T^*(\R^{n}),\, z\in S^1 =\R/\Z$.

2900: The contact form $\alpha$ satisfies the condition

2901: of  pseudo-convexity   at infinity

2902: and it is nice: the Reeb field equals $\frac{\partial}{\partial z}$,

2903: and thus it has no contractible periodic orbits.

2904:  Let us consider the class $\cN$ of domains $\Omega\subset V$

2905:  which are images of the split domains $U\times S^1\subset V$

2906:  under a contact isotopy of $V$, where $U$ is connected.

2907: For any $\Omega\in\cN$ the relative contact  homology

2908: group $HC_*(V,\Omega)$ is  well defined

2909:  because   for any function $f:\R^{2n}\to\R$

2910:  the form $f(x,y)\alpha$ is nice.

2911: 

2912: Let us denote by $\cE_r(\Omega)$, $\Omega\in\cN$,

2913: the space of contact embeddings $ D_r\times S^1\to

2914:  \Omega\times   S^1$, contact

2915:  isotopic in $V$  to the inclusion

2916:  $$D_r\times S^1\hookrightarrow \R^{2n}\times S^1=V\,.$$

2917: Notice, that for any two embeddings $f,g\in\cE_r(\Omega)$

2918: there exists  a positive $\rho\leq r$, such that the

2919: restrictions

2920: $f|_{D_\rho\times S^1}$ and $g|_{D_\rho\times S^1}$

2921: are isotopic       via a {\it contact} isotopy.

2922: 

2923: 

2924: 

2925: Given a  contact embedding $f\in\cE_r(\Omega)$

2926:  there is defined a homomorphism

2927: \begin{equation*}

2928: f_*: HC_*(V,D_r\times S^1,\xi)\to HC_*(V,\Omega,\xi)\,.

2929: \end{equation*}

2930:   Let us choose a

2931:     symplectic trivialization

2932:   of the contact bundle $\xi$ induced by the projection

2933:   $V\to\R^{2n}$. We will assume that indices of periodic orbits,

2934:   and hence the grading of contact homology groups,

2935:   are associated with this trivialization.

2936: 

2937: For each homology class $k\in \Z=H_1(D_r\times S^1)$ let us

2938: consider the maximal $l=l(f,k)$ such that

2939: $$\Ker\big( f_*|_{HC_l(V,D_r\times S^1,\xi|k)}\}\neq 0 ,,$$

2940:     and define an invariant $w_{\cont}(V,\Omega)$, called

2941:     the relative {\it

2942: contact width} by the formula

2943: \begin{equation}

2944: w_{\cont}(V,\Omega)=\sup \limits_{k,r>0,f\in

2945: \cE_r(U)}\frac{2 k}{l(f,k)}\;.

2946: \end{equation}

2947: 

2948: 

2949:   S.-S. Kim has computed $w_{\cont}(V,\Omega)$ for

2950:   certain domains $\Omega$. In particular,

2951:    she proved

2952: \begin{proposition}\label{prop:Kim}

2953: \begin{eqnarray*}

2954: w_{\cont}(V,D^{2n}_r\times S^1)= \pi r^2;  \\

2955: w_{\cont}(V,D^2_r\times D^{2n-2}_R\times S^1) =\pi r^2,

2956: \end{eqnarray*}

2957: if $R\geq r$.

2958: \end{proposition}

2959:     The contact width is clearly a monotone invariant, i.e.

2960: $$ w_{\cont}(V,U_1\times S^1)\leq w_\cont(V,U_2\times S^1)$$

2961: if $U_1\subset U_2$. Hence Proposition \ref{prop:Kim} implies

2962: 

2963:   \begin{corollary}\label{cor:non-squeezing}

2964:   Suppose that $r<\min(r',R).$ Then there is no  contact isotopy

2965:    $f_t:D^{2n}_{r'}\times S^1\to V$ such that $f_0$

2966:    is the inclusion, and

2967:   $$f_1( D^{2n}_{r'}\times S^1)\subset D^2_{r}\times D_R^{2n-2}\times S^1 .$$

2968:   \end{corollary}

2969: 

2970: 

2971: 

2972: 

2973:      \begin{problem}  Suppose there exists a contact isotopy

2974:       $f_t:V=\R^{2n}\times S^1\to V$ with

2975:      $f_0=\Id$ and $f_1(U_1\times S^1)\subset U_2\times S^1$. Does

2976:      there exist

2977:      a Hamiltonian isotopy $g_t:\R^{2n}\to\R^{2n}$ such that

2978:       $g_0=\Id$ and $g_1(U_1)\subset U_2$?

2979:       \end{problem}

2980:       Notice that the converse is obviously true.

2981: 

2982: \section{Algebraic formalism}\label{sec:algebraic}

2983: 

2984: \subsection{Informal introduction}\label{sec:informal}

2985: 

2986: The Floer-theoretic formalism described in Section \ref{sec:Floer}

2987:  is applicable only to a very limited class

2988: of contact manifolds. As it follows from Theorem \ref{thm:comp1}

2989: the boundary of the moduli space   of holomorphic cylinders in the

2990: symplectization may   consist of stable curves, different from broken

2991: cylinders; for instance,  it may contain    height 2 stable curves

2992: which consist  of a pair of pants on the upper level, and a copy

2993: of $\C$ plus a trivial cylinder at the bottom one. Hence the

2994: minimal class of holomorphic curves

2995:  in symplectizations which

2996: has the property that the  stable curves  of height $>1$ on the boundary

2997: of the corresponding moduli space  are made of curves from the same class,

2998: must contain all  rational curves with one positive   and an arbitrary

2999:  number of  negative punctures.

3000:  The counting of curves  with one positive and arbitrary number of negative punctures

3001:  can still be interpreted as a differential, but this time defined

3002:  not on the {\it vector space} generated by    periodic trajectories but on the   graded {\it algebra}

3003:   which they

3004:  generate. Thus

3005:  this  leads to a straightforward generalization of the Floer type formalism considered

3006:    in    Section \ref{sec:Floer}

3007:  when instead of  the additive  Floer complex   $\fF$  generated by the

3008:  variables $q_\g$,

3009:  one considers a graded commutative algebra  $\fA$

3010:  generated by these variables, and when  instead of the  formula (\ref{eq:Floer1})

3011:   the differential  $\partial q_\g$ is defined

3012:  as a polynomial of  a higher degree, rather than a linear

3013:  expression as in the Floer homology case. Namely, we define

3014:  \begin{equation}\label{eq:d}

3015: \partial q_\g= \sum \frac{n_{\Gamma,I,d}}{k!\prod^k_1

3016: i_j!\kappa_{\g_j}^{i_j}}q_{\g_1}^{i_1}\dots q_{\g_k}^{i_k}z^d,

3017:    \end{equation}

3018:    where the sum is taken over all ordered

3019:    \footnote{The coefficient $\frac{1}{k!}$ is the price we pay

3020:    for taking {\it ordered} sets of periodic orbits.} sets of

3021:    different periodic orbits $\Gamma=\{\g_1,\dots,\g_k\} $,   multi-indices

3022: $d=(d_1,\dots, d_N)$ and $I=(i_1,\dots,i_k), i_j\geq 0, $

3023:  and where the coefficient $n_{\Gamma,I,d}$  counts

3024: the algebraic number of elements of the moduli space

3025: $$\Mc^d_0(\g; \underbrace{\g_1,\dots,\g_1}_{i_1},\dots,\underbrace{\g_k,\dots,\g_k}_{i_k})/\R,$$ if

3026: this space is $0$-dimensional, and equals $0$ otherwise.

3027: The differential $\partial$ extends to the algebra $\fF$ according

3028: to the graded Leibnitz rule.

3029: Roughly speaking, $\partial q_\g$ is a polynomial,

3030: whose monomomials $q_{\g_1}\dots q_{\g_l}$ are in 1-1

3031: correspondence with rigid, up to translation,   rational

3032: holomorphic   curves with one positive  cylindrical end

3033: over $\g$ and $l$ negative cylindrical ends over

3034: trajectories $\g_1,\dots,\g_l$.

3035: 

3036: 

3037:  It turns out that the  quasi-isomorphism class

3038:  of the differential algebra $(\fA,\partial)$

3039:  is  independent of all extra choices (see Section \ref{sec:3algebras} below).

3040:  In particular, the {\it contact homology algebra} $ H_*(\fA,\partial)$

3041:   is an invariant of the contact manifold

3042:  $(M,\xi)$.

3043: 

3044:  Having included into the picture the moduli spaces of rational

3045:  curves with one positive and several negative punctures, one may

3046:  wonder, what is the role of rational curves with an arbitrary

3047:  number of positive and negative punctures. One can try to

3048:  interpret the

3049:  counting of rational holomorphic curves with

3050: fixed number of  positive and an arbitrary number of negative

3051: punctures  as a sequence of bracket type  operations on the

3052: algebra $\fA$. These operations  satisfy an infinite system of

3053: indentities, which remind the formalism of homotopy Lie algebras.

3054: 

3055:  However,  there is a more adequate algebraic formalism

3056:  for this picture. Let us associate   with  each periodic

3057:   orbit $\g$ two graded variables $p_\g$ and $q_\g$ of the

3058:   same parity (but   of

3059:   different integer grading, as we will see in Section

3060:   \ref{sec:correlators} below),

3061:   and consider an algebra $\fP$ of formal power series

3062:   $\sum f_\Gamma(q)p^\Gamma$, where $f_\Gamma(q)$

3063:     are polynomials of $q=\{q_\g\}$ with coefficients

3064:   in (a completion of) the group algebra  of $H_2(V)$.

3065:   It is useful to think about the algebra $\fP$ as the graded

3066:   Poisson

3067:    algebra of functions

3068:   on the     infinite-dimensional symplectic super-space $\bf{V}$ with the even

3069:   symplectic form $\sum\limits_{\g\in\Pc} {\kappa_\g}^{-1} dp_\g\wedge dq_\g$, or rather on

3070:   its formal analog along the $0$-section $\{p=\{p_\g\}=0\}$.

3071:   With each $0$-dimensional moduli space

3072:   $\Mc^d_0(\Gamma^-,\Gamma^+)/\R$, $$\Gamma^\pm=\{\underbrace{\g^\pm_1,\dots\g^\pm_1}_{i_1^\pm},\dots,\underbrace{\g^\pm_{s^\pm}\dots

3073:   \g^\pm_{s^\pm}}_{i^\pm_{s^\pm}}\},$$

3074:   we associate a monomial

3075:   $$\frac{n_{\Gamma^-,\Gamma^+,d}}{s^-!s^+!(i_1^-)!\dots(i_{s^-}^-)!(i_1^+)!\dots(i_{s^+}^+)!}

3076:     q_{\g^-}^{I^-}p_{\g^+}^{I^+}z^d,$$

3077:   where $q_{\g^-}^{I^-}=(q_{\g^-_{1}})^{i^-_{1}}\dots(q_{\g^-_{s^-}})^{i^-_{s^-}}$, $p_{\g^+}^{I^+}=(p_{\g^+_1})^{i^+_1}\cdots

3078:   (p_{\g^+_{s^+}})^{i^+_{s^+}}$,  and $n_{\Gamma^-,\Gamma^+,d}$ is

3079:    the algebraic number of elements of the moduli

3080:   space     $\Mc^d_0(\Gamma^-,\Gamma^+)/\R$.

3081: 

3082: The sum of all these monomials over all $1$-dimensional moduli

3083: spaces    $\Mc^d_0(\Gamma^-,\Gamma^+) $ for   all ordered sets

3084:     $\Gamma^-,\Gamma^+$ of periodic orbits

3085: is  an {\it odd} element $\bh\in\fP$. All the operations on the

3086: algebra $\fA$ which we mentioned above appear as the expansion

3087: terms of $\bh$ with respect to $p$-variables. It turns out that

3088: the   infinite system of  identities  for

3089: operations on $\fA$, which we mentioned above, and which is

3090:  defined by counting holomorphic curves with  a

3091: certain fixed number of positive punctures, can be encoded into a single

3092: equation $\{\bh,\bh\}=0$. Then the differentiation with respect to

3093: the Hamiltonian vector field, defined by the Hamiltonian $\bh$:

3094: \begin{equation*}

3095: d^{\bh}(g)=\{\bh,g\},\;g\in\fP,

3096: \end{equation*}

3097: defines a differential $d=d^{\bh}:\fP\to\fP$, which satisfies the

3098: equation $d^2=0$. Thus one can define the homology $H_*(\fP,d^{\bh})$

3099: which inherits the structure of a graded Poisson algebra.

3100: 

3101: 

3102:  The identities, like  $d^2=0$ and $\partial^2=0$,

3103:  encode in algebraic terms information about the structure of the  top-dimensional strata on the boundary of

3104:  compactified moduli spaces of holomorphic curves,

3105: as it is described in Proposition \ref{prop:boundary} above. For

3106: instance, the codimension $1$ strata on the boundary of the moduli

3107: space $\Mc^d_0(\Gamma^-,\Gamma^+)/\R$ consists of height two stable

3108: rational curves $(f_1,f_2)$.  Each floor of

3109:  this curve  may be disconnected,

3110: but  precisely one of its components differs from the straight cylinder.

3111:   Each   connected component of $f_1$  can be glued with a component of $f_1$ only

3112: along one of their ends. One can easily see that the combinatorics  of  such gluing   precisely corresponds to

3113: the Poisson bracket formalism and that the algebraic sum of the monomials associated to all

3114: stable curves of height two equals $\{\bh,\bh\}$. On the other hand,

3115:  the algebraic number of such height 2 curves equals

3116:  $0$ because they form   the  boundary

3117: of the a compactified  $1$-dimensional moduli space of holomorphic curves.

3118: Hence we

3119:   get the identity $\{\bh,\bh\}=0$. The identity is

3120: not tautological due to   the fact that $\bh$ is odd.   In view of the super-Jacobi identity

3121: it is equivalent to the identity $(d^{\bh})^2=0$.

3122: 

3123: 

3124: One can go further and include into the picture  moduli spaces of punctured holomorphic

3125: curves

3126: of higher genus.  Introducing a new variable, denoted $\hbar$, to keep track

3127: of the genus, one can associate with each $0$-dimensional moduli

3128: space

3129:  $\Mc^d_g(\Gamma^-,\Gamma^+)/\R$

3130:    a monomial  $$\frac{n_{\Gamma^-,\Gamma^+,d,g}}{s^-!s^+!(i_1^-)!\dots(i_{s^-}^-)!(i_1^+)!\dots(i_{s^+}^+)!}

3131:     q_{\g^-}^{I^-}p_{\g^+}^{I^+}\hbar^{g-1}z^d,$$

3132:   and form a generating function $\bH=\hbar^{-1}\sum\limits_{g=0}^\infty\bH_g\hbar^{g}$

3133:   counting all rigid holomorphic curves of arbitrary

3134:   genus, whose term $\bH_0$ coincides with $\bh$.    Again,

3135:   the codimension $1$ strata of the boundary of the moduli spaces

3136:   $\Mc_g(\Gamma^-,\Gamma^+)$ consists of height 2 stable curves, but unlike the case of  rational

3137:   curves, two connected components

3138:   on different levels can be glued together along an arbitrary number of ends.

3139:    The combinatorics of such gluing

3140:   can be described by the formalism of algebra of higher order differential operators.

3141:   Fig. \ref{fig:combinatorics}

3142:   illustrates how the composition

3143: of

3144:   differential operators can be interpreted via gluing of Riemann

3145:   surfaces with punctures. A letter $p_i$ in the picture represents a differential

3146:   operator

3147:              $\hbar\frac{\partial}{\partial q_i}$, and  a surface of genus $g$ with upper punctures

3148:     $p_{i_1},\dots,p_{i_k}$ and lower punctures $q_{j_1},\dots,q_{j_l}$ represents

3149:     a differential operator

3150:         $$\hbar^{g-1}q_{j_1}\dots q_{j_l}p_{i_1}\dots p_{i_k}=\hbar^{g-1}q_{j_1}\dots q_{j_l}

3151:         \big(\hbar\frac{\partial}{\partial q_{i_1}}\big)

3152:         \dots \big(\hbar\frac{\partial}{\partial q_{i_k}}\big).$$

3153:                  Thus

3154:          we are led to consider $\bH$ as an element of the Weyl super-algebra

3155:   $\fW$. This algebra should be

3156:    viewed

3157:   as a quantization of the Poisson algebra $\fP$, so that the description of the boundary of the

3158:   moduli spaces is given by the equation $[\bH,\bH]=0$, where   $[\;,\;]$ denotes the commutator in $\fW$.

3159:   As in  the rational case, this  identity is equivalent

3160:   to the identity $D^2_{\bH}=0$  for the differential

3161:   $D^\bH(f)=[H,f]$.

3162:   Hence we can define the

3163:   homology algebra $H_*(\fW,D^{\bH})$, which   also turns out to be an invariant of the contact manifold

3164:   $(V,\xi)$. Similarly to the standard Gromov-Witten theory for closed

3165:   symplectic manifolds one can develop an even more general

3166:   formalism  by encoding in $\bH$ information about higher-dimensional moduli spaces

3167:   of holomorphic curves. This leads to a deformation of the differential

3168:   algebra $(\fW,D^{\bH})$ along the space of closed

3169:   forms on $V$. The corresponding family of homology algebras   is

3170:   then  parameterized by $H^*(V)$.

3171: 

3172: %%%%%%%%%%%%%%%%%%%%

3173: 

3174: \begin{figure}

3175: \centerline{\psfig{figure=sf1.eps,height=170mm}} \caption{\small

3176:          There are  four different way to glue the lower and upper surfaces

3177:     on the picture along their matching ends, i.e. the ends denoted by $p$'s

3178:     and $q$'s with the same index.  These 4 ways correspond to  4

3179:     terms in the composition formula for differential operators:

3180:                              $(\hbar^{-1}p_1 p_2 p_3) \circ(\hbar^{-1}q_1 q_2 p_1)=p_1 p_3+

3181:                                 \hbar^{-2} q_1 q_2 p_1^2 p_2 p_3 +

3182:                 \hbar^{-1}q_1 p_1^2 p_3+\hbar^{-1}q_2 p_1 p_2 p_3 \;.

3183:                $  We  are ignoring here the sign issues and assuming all  the boundary components

3184:                  to be simple orbits.

3185:                                  }

3186:         \label{fig:combinatorics}

3187: \end{figure}

3188: %\end{document}

3189: %%%%%%%%%%%%%%%%%%%%%%

3190:   After going that far it is natural to  make the above algebraic structure

3191:   for contact manifolds a part

3192:   of a formalism in the spirit of topological field theory, which we call

3193:    {\it Symplectic Field Theory},

3194:   and which also includes the theory of Gromov-Witten invariants of closed manifolds.

3195:   To do that one considers moduli spaces of holomorphic curves with cylindrical ends

3196:   in directed symplectic cobordisms $W=\ora{V^-V^+}$. The generating function

3197:   counting rational holomorphic curves

3198:   in $W$ can be naturally written as a function  $\bff(q^-,p^+ )$ of $p^+$-variables

3199:   associated with the positive end $V^+$, and

3200:   $q^-$-variables associated with the negative end $V^-$ of the cobordism $W$.

3201:   It turns out that the Lagrangian submanifold

3202:   in $(-\mathbf{V}^-)\times\mathbf{V}^+$ generated by the function $\bff$

3203:   defines a {\it Lagrangian correspondence}

3204: $  \mathbf{L}_W\subset(-\mathbf{V}^-)\times(\mathbf{V}^+) $

3205:   which transforms the Hamiltonian

3206:   functions $\bh^+$ and  $\bh^-$ to each other, i.e.

3207:   \begin{equation}\label{eq:HJ-informal}

3208:   \left(\bh^-(p^-,q^-)-\bh^+(p^+,q^+)\right)|_{\mathbf{L}_W}=0,

3209:   \end{equation}

3210:   where

3211:   \begin{equation*}

3212:   L_W=

3213:   \begin{cases}

3214:   q^+_{\g^+}=&\kappa_{\g^+}\frac{\partial \bff}{\partial p^+_{\g^+}}(q^-,p^+));\\

3215: p^-_{\g^-}=&\kappa_{\g^-}\frac{\partial \bff}{\partial

3216: q^-_{\g^-}}(q^-,p^+)).\\

3217: \end{cases}

3218: \end{equation*}

3219:    We recall that $\kappa_{\g^\pm}$ denotes the multiplicity of the orbit $\g^\pm$.

3220: 

3221:   The composition of symplectic cobordisms produces

3222:   the composition of Lagrangian correspondences,

3223:   so that  if one  consider a ``Heegard splitting" of  a closed symplectic

3224:   manifold $W$ along a contact hypersurface $V$, then  the computation of Gromov-Witten invariants

3225:    of $W$ can be viewed

3226:   as a Lagrangian intersection problem in the   symplectic super-space

3227:   $\mathbf{V}$ associated to the contact manifold $V$.

3228: 

3229:   After what was said it should not come as a surprise that in

3230:    the quantized picture Lagrangian correspondences

3231:   are being replaced by Fourier integral operators, and the

3232:    composition of correspondences by the convolution

3233:   of the corresponding operators.

3234: 

3235:  We describe below the SFT-formalism with more details. We treat

3236: contact manifolds in Section \ref{sec:af-contact} and  symplectic

3237: cobordisms in Section \ref{sec:af-cobordism}. Section

3238: \ref{sec:chain}  is devoted to  the   SFT-version of the  chain  homotopy statement

3239:  in Floer homology theory.   In Section  \ref{sec:composition}  we introduce the composition

3240:  formula for the SFT-invariants of symplectic cobordisms. In Section \ref{sec:contact-invariants}

3241:  we discuss    how the   introduced algebraic structures of contact manifolds

3242:  depend on extra choices.  Section \ref{sec:master}  is devoted

3243:  to a differential equation for the  potential $\bF$ of a directed symplectic

3244:  cobordism    with  a {\it non-empty } boundary.  Together with the gluing formula from Section

3245:  \ref{sec:composition} this  equation  provides an effective tool for  computing

3246:   Gromov-Witten invariants.

3247: The remainder

3248: of the paper has even  more sketchy character than the rest of

3249: the paper.

3250:  Section \ref{sec:Legendrian}   is

3251: devoted to invariants of Legendrian submanifolds via SFT. Section

3252: \ref{sec:remarks} is devoted to various examples and possible

3253: generalizations of  SFT. In particular, in Section

3254: \ref{sec:Bott} we discuss   how  one can adapt the theory to

3255: include an important for applications, though non-generic, case of

3256: contact forms with continuous families of periodic orbits. In

3257: Section \ref{sec:computing} we describe a  new recursive

3258: procedure for computing rational  Gromov-Witten invariants of $\C

3259: P^n$. Finally, in Section \ref{sec:satellites} we  just touch the

3260: wealth of other  invariants which exist in  Symplectic Field

3261: Theory.

3262: 

3263:  \subsection{Contact manifolds}\label{sec:af-contact}

3264: 

3265:  \subsubsection{Evaluation maps}\label{sec:contact-eval}

3266:  Let $(V,\xi)$ be a contact manifold with a fixed contact form

3267:  $\alpha$, $(W=V\times\R,d(e^t\alpha))$ the symplectization of

3268:  $(V,\xi)$, and $J$  a compatible almost complex structure.  As in Section \ref{sec:holomorphic}

3269:   we denote  by $f_V$ and $f_\R$

3270:   the $V $- and $\R$-components of  a

3271:  $J$-holomorphic curve $f$ in $W$, and

3272:  by $\Mc_{g,r,s^-,s^+}(W,J)$ the  disjoint union

3273:    $$\bigcup\Mc_{g,r}^A(\Gamma^-, \Gamma^+),$$

3274:    which is taken over all $A\in H_2(V)$, and all

3275:     sets $ \Gamma^-,\Gamma^+\subset\Pc_\alpha$ of cardinalities $s^\pm$.

3276: 

3277:    %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3278: Let us view the set  $\Pc=\Pc_\a$ of periodic orbits of  the Reeb fields $ R_{\alpha }$ as a discrete topological space.

3279:  It naturally splits into the   disjoint

3280: union

3281: $$\coprod\limits_{k=1}^{\infty}\Pc _k,$$ of identical subspaces, where $\Pc_k$ is the space of

3282:   periodic orbits of

3283: multiplicity $k$.

3284: 

3285:            Consider now three sets of evaluation maps:

3286:             $$ev^0_i: \Mc_{g,r,s^-,s^+}/\R \to V,\;i=1,\dots,r,$$

3287:             $$ev^+_j:\Mc_{g,r,s^-,s^+ }/\R    \to \Pc ,\;j=1,\dots      ,s^+,$$

3288:             and

3289:               $$ev^-_k:\Mc_{g,r,s^-,s^+ }/\R    \to \Pc ,\;k=1,\dots      ,s^-,$$

3290:               where $ev_i^0$ is the  evaluation map $f_V(y_i)$

3291:                at the $i$-th marked point $y_i$, while

3292:                             $ev^\pm_j$   are the evaluation maps

3293:                             at asymptotic markers $\mu^{\bx^\pm}_j$.

3294:                More precise,

3295:                let

3296:                            $$\overline{f}=(f,j,\bx^-,\bx^+,\by,\mu^{\bx^-},\mu^{\bx^+})\in

3297:                             \Mc_{g,r,s^-,s^+ },$$

3298:               and $f$ be asymptotically cylindrical over a

3299:                             periodic orbit $\gamma^\pm_j\in \Pc $ at $\pm\infty$ at the

3300:               puncture   $x^\pm_j$. Then $ev_j^\pm(\overline{f})$ is a point of $

3301:               \Pc $ representing the orbit $\g^\pm_j$ (comp. Section \ref{sec:Bott} below).

3302: 

3303:               All the above evaluation maps can be combined into a map

3304:               $$ev:\Mc_{g,r,s^-,s^+ }/\R\to V^r\times (\Pc^-)^{s^-}\times (\Pc^+)^{s^+}\,,$$

3305: which extends to the compactified moduli space

3306: $\overline{\Mc_{g,r,s^-,s^+ }/\R}$.

3307:    %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

3308: \subsubsection{Correlators}\label{sec:correlators}

3309: %%%%%

3310:           Now we are ready to define

3311:   correlators.

3312:  Given $r$    differential

3313: forms $\theta_1,\dots,\theta_r$ on $V$ and $s^\pm$ ($0$-dimensional) cohomology classes $\alpha^\pm_1,\dots,

3314: \alpha^\pm_{s^\pm}\in H^*(\Pc )$  we define the    degree $-1$, or contact {\it

3315: correlator}

3316: \begin{eqnarray}\up{-1}\langle\,\theta_1,\dots,\theta_r;\alpha^-_1,\dots,

3317: \alpha^-_{s^-};\alpha^+_1,\dots,

3318: \alpha^+_{s^+} \rangle^A_{g}=\\

3319:  \nonumber \int\limits_{\overline{ {\Mc_{g,r,s^-,s^+ }^A}/\R}}ev^*(\theta_1\otimes\dots

3320: \otimes \theta_r\otimes\alpha^-_1\otimes\dots\otimes\alpha^-_{s^-}

3321: \otimes\alpha^+_1\otimes\dots\otimes\alpha^+_{s^+} ).

3322: \end{eqnarray}

3323: 

3324: Usually we will assume that the forms $\theta_1,\dots,\theta_r$

3325: are closed, but even in this case the above  correlator  depends

3326: on the actual forms, and not just their cohomology classes,

3327: because the domain of integration may have a boundary.  As we will see

3328: below the

3329: superscript $-1$ in $\up{-1}\langle\dots\rangle $ corresponds

3330: to the grading of the generating function for these

3331: correlators. It also refers to the

3332: enumerative meaning of the correlators: they count

3333: components of $1$-dimensional moduli spaces of holomorphic curves.

3334: We will consider below also correlators $\up0\langle\dots\rangle $

3335: and $\up1\langle\dots\rangle$, counting $0$-dimensional and

3336: $-1$-dimensional (i.e. appearing in 1-dimensional families) moduli

3337: spaces of holomorphic curves.

3338: 

3339: 

3340:  If we are given $K$ linearly independent differential forms

3341:   $\Theta_1,\dots,\Theta_K$,

3342:   then it is convenient

3343:   to introduce a ``general form" $t=\sum\limits_1^Kt_i\Theta_i$

3344:   from the space $L=L(\Theta_1,\dots,\Theta_K)$

3345:   generated by the  chosen forms, and view $t_i$ as graded variables

3346:   with  $\deg(t_i) =\deg(\Theta_i)-2$. In particular, all terms in

3347:   the sum $\sum\limits_1^Kt_i\Theta_i$ have even degrees.

3348: 

3349: 

3350:   Let us   consider two copies

3351:  $\Pc^+$ and $\Pc^-$ of the $0$-dimensional space

3352:  $\Pc$, one   associated with the positive end of $W$, the other with

3353:   the negative one. Cohomology classes  in $\Pc_+$ we will denote

3354:   by $p$, and in $\Pc_-$ by $q$, and will write

3355:   $$p=\sum_{\g\in\Pc}\frac{1}{\kappa_\g}p_\g[\g],\quad q=

3356:   \sum_{\g\in\Pc}\frac{1}{\kappa_\g}q_\g[\g],$$

3357:    where   $\kappa_\g$ is

3358:  the multiplicity of $\g$, and the cohomology classes $[\g]$

3359:   form the canonical basis of $H^*(\Pc)$, dual to points in $\Pc$.

3360:   Of course, speaking about cohomology classes of a discrete space may sound

3361:   somewhat ridiculous. However, this point of view   is useful, especially

3362:    in preparation for a more general  case when

3363:     some periodic orbits may be degenerate and  thus the spaces  $\Pc^\pm$

3364:       need not to be  anymore discrete, see Section \ref{sec:Bott} below.

3365:  We will also fix a basis $A_1,\dots,A_N$ of

3366:  $H_2(V)$. The coordinate vector  $d=(d_1,\dots,d_N)$ of a class $A$

3367:  is called the degree.   Here $d_j$ are integers, while we   consider

3368:  $t,p,q $ as graded variables,

3369:   where the  degrees of the variables $p,q$ are   defined by the  formulas

3370:   $$\deg (p_\gamma )  =-\CZ(\gamma)+(n-3),\, $$

3371:  $$ \deg (q_\gamma )  =+\CZ(\gamma)+(n-3).$$

3372:  The choice

3373:   of grading,

3374:    somewhat strange at the first glance,  is explained by Proposition \ref{thm:finiteness-of-H} below.

3375: 

3376:      The correlators

3377:   $$ \langle \underbrace{t,\dots,t}_r;\underbrace{q,\dots,q  }_{s^-}

3378:      ;\underbrace{p,\dots,p}_{s^+}\rangle^d_g$$ with  different

3379:      $r,d,g$ determine all the correlators involving forms from the

3380:      space $L$.

3381: 

3382: 

3383: \subsubsection{Three differential algebras}\label{sec:3algebras}

3384: 

3385:    Similar to the theory of Gromov-Witten invariants of   closed symplectic manifolds we

3386:    will organize all correlators

3387:       into a generating function, called {\it Hamiltonian},

3388:    $${\bH}=\frac{1}{\hbar}\sum\limits_{g=0}^\infty{\bH}_g\hbar^g,$$

3389:    where,

3390: \begin{equation}

3391:  {\bH}_g=\sum\limits_d\sum\limits_{r,s^\pm=0}^\infty\frac{1}{r!s^-!s^+!}

3392:      \up{-1}\langle \underbrace{t,\dots,t}_r;\underbrace{q,\dots,q  }_{s^-}

3393:      ;\underbrace{p,\dots,p}_{s^+}\rangle^d_g z^d,

3394:      \end{equation}

3395: and  $t=\sum_1^Kt_i\Theta_i$. We will assume throughout the paper,

3396: that

3397: all forms $\Theta_1,\dots,\Theta_K$ are closed (see,

3398: however, Remarks \ref{rem:non-closed1} and \ref{rem:non-closed2}, and

3399: Section \ref{sec:master} below).

3400:  The variables $\hbar$ and $z=(z_1,\dots,z_N)$ are also considered as graded

3401:  with $\deg \hbar= 2(n-3)$ and $\deg (z_i)=- 2c_1(A_i)$, where $c_1$ is the first

3402:  Chern class of  the almost complex structure $J$.

3403: 

3404: 

3405: 

3406: \begin{proposition} \label{thm:finiteness-of-H}

3407: \begin{description}

3408:   \item{a)}  For each $g=0,\dots$  the series $\bH_g$

3409:    can be viewed as  formal power series in variables $p_\g$   with coefficients

3410:   which are   polynomials   of variables $q_\g$ and formal power series of $t_i$ \footnote{

3411:   In fact, $\bH_g$ depends polynomially on all variables $t_i$ of degree $\neq 0$.

3412:   The degree $0$ variables, i.e. the variables

3413:   associated with $2$-forms, enter into the constant part of $\bH_g$ (i.e.

3414:   the part describing constant holomorphic curves) polynomially, while

3415:    the non-constant part    of $\bH_g$ depends

3416:    polynomially on $e^{t_i}$. This fact is similar to the standard Gromov-Witten theory

3417:   and  will not discussed  in the present paper.}

3418:   with coefficients

3419:   in the group algebra $\C[H_2(V)]$

3420:     {\rm (which we identify with the algebra

3421:   of Laurent polynomials of $z$ with complex coefficients)};

3422:   \item{b)} All terms of $\bH$  have degree $-1$;

3423:   \item{c)} $\bH\big|_{p=0}=\bH_{\const}$, where

3424:    $$\bH_{\const}=\hbar^{-1}\sum\limits_{g,r=0}^\infty

3425:    \frac{1}{r!}\up{-1}

3426:    \langle\underbrace{t,\dots,t}_r\rangle^0_g\hbar^g$$

3427:    accounts for the contribution of constant holomorphic curves.

3428:    In particular, $\bH\big|_{p=0}$ is independent of $q$ and $z$.

3429:   \end{description}

3430:   \end{proposition}

3431: The polynomial dependence of $\bH_g$  on variables $q_\g$ and $z$   in a  geometric language just means that

3432: the union $\overline{\Mc^d_g(\Gamma^+)}$, $\Gamma^+=\g_1\dots\g_{s^+}$,

3433: of the  compactified moduli spaces of holomorphic curves  of a fixed genus of any degree

3434:  with prescribed

3435: positive ends  $\g^+_1,\dots,  \g_{s^+}$

3436: is compact, and in particular that there are only finitely many possibilities for the degrees

3437: and the negative

3438: ends of these curves.

3439:  This follows from the fact  that  for each curve $C\in \Mc_g(\Gamma^-,\Gamma^+)$ we have

3440:  \begin{equation}\label{eq:pos-ends}

3441:  0\leq \int\limits_C d\alpha= \sum\limits_{\g_i\in\Gamma^+}\int\limits_{\g_i}\alpha-

3442:  \sum\limits_{\g_j\in\Gamma^-}\int\limits_{\g_j}\alpha\leq

3443:   \sum\limits_{\g_i\in\Gamma^+}\int\limits_{\g_i}\alpha,

3444:   \end{equation}

3445:   the fact   that there exists a constant $m>0$ such that $\int\limits_{\g}\a>m$ for

3446:   any periodic   orbit $\g\in\Pc_\a$ and Theorem \ref{thm:comp1}  above.

3447:      Proposition \ref{thm:finiteness-of-H}b)   follows from the formula

3448:   (\ref{eq:dim}) for the dimension of the moduli spaces of holomorphic curves, our degree

3449:   convention   and the fact that a correlator

3450:       $\up{-1}\langle\,\theta_1,\dots,\theta_r;\g^-_1,\dots,

3451: \g^-_{s^-};\g^+_1,\dots,

3452: \g^+_{s^+} \rangle^A_{g}$  may be different from $0$ only if the total dimension of the forms

3453: $\theta_1,\dots\theta_r$ equals the dimension of the moduli  space   $\Mc^A_{g,r}( \g^-_1,\dots,

3454: \g^-_{s^-};\g^+_1,\dots,

3455: \g^+_{s^+})/\R $.

3456:   Proposition \ref{thm:finiteness-of-H}c) just means that every non-constant holomorphic curves should have at least

3457:   one positive end, which follows from   inequality  (\ref{eq:pos-ends}), or alternatively the

3458:   maximum principle for holomorphic curves.

3459:   \medskip

3460: 

3461:  Let us consider the {\it Weyl super-algebra}

3462:   $\fW=\{\sum\limits_{\Gamma,g} f_{\Gamma,g }(q,t )p^\Gamma\hbar^g \}$,

3463:   where

3464:  $$\Gamma=(\g_1,\dots,\g_a),\gamma_i\in\Pc,\,a=1,\dots,\, p^{\Gamma}=p_{\g_1}\dots p_{\g_a},$$

3465:   and        $f_{\Gamma,g }(q,t)$ are polynomials of variables

3466:   $q_\gamma$ and formal power series of $t_i$\footnote{See the previous footnote.}. Proposition

3467:   \ref{thm:finiteness-of-H}a) states that $\bH\in\hbar^{-1}\fW$.

3468: 

3469:   The product operation $F\circ G$ in $\fW$ is associative and satisfies the following commutation

3470:   relations: all variables

3471:   are super-commute (i.e. commute or anti-commute according to their grading), except

3472:   $p_\gamma$ and $q_\gamma$ which correspond to the same periodic orbit $\gamma$. For these pairs of variables

3473:    we have the following commutation relation:

3474:  \begin{equation}

3475:  [p_\gamma,q_\gamma]=p_\g\circ q_\g-(-1)^{\deg p_\g\deg q_\g}q_\g\circ p_\g=

3476:  {\kappa_\gamma}\hbar \,

3477:  \end{equation}

3478:   where $\kappa_\g$ is the multiplicity of the orbit $\g$.

3479:   The algebra $\fW$ can be represented as an  algebra of formal differential   operators  with respect

3480:   to

3481:   $q$-variables acting  {\it on the left} on the space of polynomials

3482:     $  f(q,z,\hbar) $, by setting

3483:     $$p_\g= {\kappa_\g}\hbar\ora{\frac{\partial}{\partial

3484:     q_\g}}.$$

3485:     Alternatively by setting

3486:     $$q_\g={\kappa_\g}\hbar\ola{\frac{\partial}{\partial p_\g}}$$ we can

3487:     represent

3488:     $\fW$   as an algebra  of  polynomial differential operators  acting {\it on the right} on the algebra $\{\sum

3489:     \limits_{\Gamma,g}f_{\Gamma,g}(q,z)\hbar^gp^\Gamma\}$ of formal power series

3490:     of $\hbar$ and the $p$-variables.

3491: 

3492:   Notice that  the commutator $[F,G]$ of  two homogeneous elements $F,G\in\fW$

3493:   equals $F\circ G-(-1)^{\deg F\deg G}G\circ F$, and hence if $F\in\fW$ is an odd element (i.e. all its

3494:   summands are odd) then

3495:      $[F,F] =2F\circ F,$

3496:      and $[F,F]=0$ if $F$ is even. For any two elements

3497:      $F,G\in\fW$ the commutator $[F,G]$ belongs to the ideal

3498:      $\hbar\fW$.

3499:                     According to Proposition \ref{thm:finiteness-of-H}   the Hamiltonian

3500:      $\bH$ can be viewed as an element of $\frac{1}{\hbar}\fW$, and hence

3501:      the above remark shows that for $F\in\fW$

3502:      we have $[\bH,F]\in\fW$.

3503:   \begin{theorem} \label{thm:HH}

3504:  The Hamiltonian  $\bH$ satisfies the identity

3505:   \begin{equation}\label{eq:HH}

3506:  \bH\circ \bH=0.

3507:  \end{equation}

3508:  \end{theorem}

3509: 

3510: This theorem (as well as Theorems

3511: \ref{thm:SFT-cobordism}, \ref{thm:SFT-chain} and \ref{thm:SFT-composition}

3512: below) follows from the description of the boundary of the

3513: corresponding moduli spaces of holomorphic curves. As it was stated

3514: in Proposition \ref{prop:boundary}

3515:   this boundary

3516:   is tiled by codimension one

3517:  strata represented by stable curves of height $2$, so that the

3518: (virtual) fundamental cycles of the boundary of the compactified  moduli

3519: spaces

3520:   can be

3521: symbolically written as   $\partial [\Mc]=\kappa\sum [\Mc_{-}]\times [\Mc_{+}]$,

3522: where $[\Mc_\pm]$ are chains represented by the corresponding moduli spaces and  where

3523:  the coefficient

3524:  $\kappa$ depends on multiplicities of orbits along which the two levels of the corresponding

3525:  stable curve are glued. Together

3526: with the Stokes formula $\int\limits_{[\Mc]} d\omega = \int\limits_{\partial [\Mc]} \omega

3527: $, and the fact that the integrand is a closed form, we obtain

3528: identity

3529: (\ref{eq:HH}).

3530: 

3531: \begin{remark}\label{rem:non-closed1}

3532: {\rm The same argument shows that when the forms $\Theta_i$

3533: generating the space $L$  are not necessarily closed

3534: we get the following  equation

3535: \begin{equation}\label{eq:non-closed1}

3536: d\bH+\frac12[\bH,\bH]=0,

3537:  \end{equation}

3538:  which

3539: generalizes (\ref{eq:HH}) and can be interpreted as the

3540: zero-curvature equation for the connection $d +[\bH,\cdot\,]$. We

3541: denote here by $d$ the de Rham differential, i.e.

3542:  \begin{equation}

3543:  \begin{split}

3544:  d\bH&=d\Big(\sum\limits_{d,g}\sum\limits_{r,s^\pm=0}^\infty\frac{1}{r!s^-!s^+!}\\

3545:    & \quad \big\langle \underbrace{\sum\limits_1^Kt_i\Theta_i,\dots,\sum\limits_1^Kt_i\Theta_i}_r;\underbrace{q,\dots,q  }_{s^-}

3546:      ;\underbrace{p,\dots,p}_{s^+}\big\rangle^d_g

3547:      z^d\hbar^{g-1}\Big)\\

3548:       \end{split}

3549:      \end{equation}

3550:  \begin{equation*}

3551:  \begin{split}

3552:      \;\;&=\sum\limits_{d,g}\mathop{\sum\limits_{s^\pm=0,}}\limits_{r=1}^\infty\frac{1}{(r-1)!s^-!s^+!}\\

3553:     &\quad \big\langle\sum\limits_1^K t_i

3554:      d\Theta_i, \underbrace{\sum\limits_1^Kt_i\Theta_i,

3555:      \dots,\sum\limits_1^Kt_i\Theta_i}_{r-1};\underbrace{q,\dots,q  }_{s^-}

3556:      ;\underbrace{p,\dots,p}_{s^+}\big\rangle^d_g

3557:      z^d\hbar^{g-1}\,.\\

3558:      \end{split}

3559:      \end{equation*}

3560:  }

3561: \end{remark}

3562: \bigskip

3563: 

3564:  The identity $\bH\circ\bH=0$

3565:  is equivalent to $[\bH,\bH]=0$, because $\bH$ is an odd element.

3566:  Let  us  define the differential

3567:  $D=D^{\bH}:\fW\to\fW$

3568:  by the formula

3569:  \begin{equation}

3570:  D^{\bH}(f)=[\bH,f]\quad\hbox{ for }\quad f\in\fW .

3571:  \end{equation}

3572: Then Theorem \ref{thm:HH} translates into the identity $D ^2=0$. The differential

3573:  $D^{\bH}$ satisfies the Leibnitz rule, and thus

3574:  $(\fW,D )$

3575:  is a        differential Weyl (super-)algebra.

3576: In particular, one can define the homology algebra $H_*(\fW,D)$,

3577: which inherits  its multiplication operation from  the Weyl

3578: algebra $\fW$.   The differential $D^{\bH}$ extends in an obvious way to the modules

3579: $\hbar^{-k}\fW$, $k=1,\dots.$

3580: 

3581: \begin{example}\label{ex:circle}

3582: {\rm Let

3583:  $V=S^1$. We have  in this case

3584:  \begin{equation}\label{eq:circle}

3585:  \bH=\hbar^{-1}\big(\frac{t_1t_0^2}{2}+ t_1\sum p_kq_k-\frac{t_1\hbar }{24}\big),

3586:  \end{equation}

3587:   where

3588: $ t=t_01+t_1d\phi$ is a general harmonic differential form on $S^1$,

3589: so that $\deg t_1=-1$,   $\deg t_0=-2$, $\deg\hbar=-4$ and  $\deg p_k,\deg q_k=-2$,

3590: which corresponds to the convention

3591: that the Maslov index of any path in the

3592: $1$-point group $Sp(0)$ equals 0.

3593: The term $t_1t_0^2/2=\int\limits_{S^1} t^{\wedge 3}/6$ is the contribution of the moduli

3594: space $S^1$ of constant maps $\C P^1\to \R\times S^1$ with $3$ marked points.

3595:  The term $-\frac{t_1\hbar}{24}$ is accounted for

3596: the contribution of constant curves of genus $1$ (see

3597: \cite{Witten2}), and the term $t_1p_kq_k$ represents the

3598: contribution $t_1=\int\limits_{S^1} t$ of trivial curves of multiplicity

3599: $k$ with one marking. All other curves do not contribute to $\bH$

3600: for dimensional reasons and because $t_1^2=0$.

3601: 

3602: Notice that if we organize all variables $p_k,q_k$   into

3603:  formal Fourier series  (comp. \cite{Getzler})

3604: \begin{equation}

3605: u(x)= \sum\limits_{k=1}^\infty(p_ke^{ixk}+q_ke^{-ixk}),

3606: \end{equation}

3607:  then the term accounting for

3608:   the contribution of rational curves in the formula (\ref{eq:circle})

3609:   takes the form

3610:   %CHANGE

3611:   \begin{equation}

3612:      \frac{t_1}{4\pi }\int\limits_0^{2\pi} (t_0+   u(x)) ^2  \ dx,

3613: \end{equation}

3614: see further discussion of this $u$-formalism in Section \ref{sec:Bott} below.

3615:  }

3616: \end{example}

3617: 

3618: 

3619:  We will  associate now with $(\fW,D)$    two other differential algebras, $(\fP,d)$ and $(\fA,\partial)$, which can be viewed

3620:  as {\it semi-classical} and {\it classical} approximations of the Weyl  differential algebra.

3621: 

3622: Let us denote  by

3623: \begin{description}

3624:  \item{-} $\fP$ -- a graded Poisson algebra  of formal power series in variables

3625:   $p_\g $ with coefficients which are polynomials of $q_\g, z_j,

3626:   z_j^{-1}$, and formal power series of $t_i$,

3627:   \footnote{See the first footnote in Section \ref{sec:3algebras}.} and by

3628: \item{-} $\fA $ --  a graded commutative algebra generated by variables

3629:  $q=\{q_\g\}_{ \g\in\Pc}$   with coefficients in   the algebra

3630:     $\C[H_2(V)][[t]]$.

3631:     \end{description}

3632:   The Poisson bracket on $\fP$ is defined by the formula

3633:   \begin{equation}\label{eq:Poisson}

3634:   \{h,g\}= \sum\limits_\g  {\kappa_\g}\left(\frac{\partial h}{\partial p_\g}

3635:   \frac{\partial g}{\partial q_\g}-(-1)^{{\deg h}{\deg g}}

3636:    \frac{\partial g}{\partial p_\g}\frac{\partial h}{\partial q_\g}

3637:  \right)    ,

3638:  \end{equation}

3639:  assuming that $h$ and $g$ are {\it monomials}.

3640:  When computing partial derivatives, like $\frac{\partial h}{\partial q_\g}$,

3641:   one should remember that we are working in the super-commutative environment,

3642:  and in particular the operator $\frac{\partial }{\partial q_\g}$

3643:  has the same parity as  the variable $q_\g$.

3644:  \begin{remark}\label{rem:u-notation}

3645:  {\rm

3646:   Notice, that if  similarly to  Example \ref{ex:circle} above

3647:  we organize the variables $p_k=p_{\g_k},q_k=q_{\g_k}$ corresponding to multiples

3648:  of each simple periodic orbit $\g=\g_1$ into  a

3649:  Fourier series

3650:  $$u_\g=\sum\limits_{k=1}^\infty(p_ke^{ixk}+q_ke^{-ixk}),$$

3651:  then the value of the Poisson tensor

3652:  (\ref{eq:Poisson}) on covectors $\delta u,\delta v$ takes      the

3653:   form

3654:  \begin{equation}

3655:  \label{eq:u-notation}

3656:  \frac{1}{2\pi i}\int\limits_0^{2\pi}  (\d u)'\d vdx  \,.

3657:  \end{equation}

3658:     }

3659:  \end{remark}

3660:  \bigskip

3661: 

3662:     In order to define differentials

3663:      on the algebras $\fA$ and $\fP$ let us  first     make

3664: the following observation.

3665: \begin{lemma} \label{lm:HH-Weyl}

3666: We have $$[\bH,\bH]=\frac{1}{\hbar}\{\bH_0,\bH_0\}+\dots\,,$$

3667: and for any $f=\frac{1}{\hbar}\sum f_g\hbar^g\in \frac{1}{\hbar}\fW$

3668:  we have

3669: $$D^\bH(f) = \frac{1}{\hbar}\{\bH_0,f_0\}+\dots\,.$$ In

3670: particular, $\bH_0$ satisfies the equation

3671: $\{\bH_0,\bH_0\}=0.$

3672: \end{lemma}

3673: 

3674: To cope with a growing number of indices we will rename $\bH_0$

3675: into $\bh$.

3676: %%%%%%%%%

3677:    Lemma \ref{lm:HH-Weyl} allows us to define the differential

3678:    $d=d^{\bh}:\fP\to\fP $ by the formula

3679:    \begin{equation}

3680:    dg=\{\bh,g\} \quad\hbox{ for }\quad g\in\fP .

3681: \end{equation}

3682: Theorem \ref{thm:HH} then implies

3683:    \begin{proposition}\label{prop:dd-Poisson} We have

3684:    $d^2=0$ and $d\{f,g\}=\{df,g\}+ (-1)^{\deg f}\{f,dg\}$ for any homogeneous element

3685:    $f\in\fP$. In other words,   $(\fP,d)$ is a

3686:    graded differential Poisson algebra with  unit.

3687:    \end{proposition}

3688:    Proposition \ref{prop:dd-Poisson}

3689:    enables us to define the homology $H_*(\fP,d)$ which inherits  from

3690:     $\fP$ the structure of a graded Poisson algebra with unit.

3691: 

3692:      Let us recall that according to \ref{thm:finiteness-of-H}

3693:      $\bh|_{p=0}=\bh_{\const}$,

3694:      where $\bh_{\const}$ accounts for constant rational

3695:      holomorphic curves, and thus it is independent of $q$-variables.

3696:      In fact, $$\bh_{\const}(t)=\frac16\sum\limits_{i,j,k=1}^K c^{ijk}t_it_jt_k,$$

3697:      where $c^{ijk}=\int\limits_V\Theta_i\wedge\Theta_j\wedge\Theta_k$ are

3698:      the structural

3699:      coefficients of the cup-product.

3700:           Hence,  $$\bh=\bh_{\const}+\sum h_{\g }(q,t,z)p_\g   +\dots,$$

3701:      where $\dots$ denote terms at least quadratic in $p_\g$.  Thus

3702:      we have

3703:       \begin{equation}

3704:      \{\bh,\bh\} = 2\sum\limits_{\g ,\g'\in\Pc }\kappa_{\g'}

3705:       h_{\g'  }(q,t)\frac {\partial h_{\g  }}{\partial q_{\g'}}(q,t)p_{\g}+ o(p)=0\,.

3706:       \end{equation}

3707:         Therefore,

3708:         \begin{equation}\label{eq:d2}

3709:          \sum\limits_{\g'  \in\Pc }\kappa_{\g'}

3710:       h_{\g' }(q,t)\frac{\partial h_{\g  }}{\partial

3711:       q_{\g'}}(q,t)=0

3712:       \end{equation}

3713:       for all $t$ and all $\g\in\Pc$.

3714: 

3715: 

3716:    Let us define a differential $\partial: \fA\to\fA$ by the

3717:       formula

3718: 

3719:       \begin{equation}

3720:       \partial f=\{\bh,f\}|_{\{p=0\}}=\sum\limits_{\g\in\Pc}\kappa_\g h_\g\frac{\partial

3721:       f}{\partial q_\g}.

3722:       \end{equation}

3723:       Then the equation (\ref{eq:d2}) is equivalent to

3724: 

3725:       \begin{proposition}\label{prop:zero-section}

3726:      $\partial^2=0$, and hence $(\fA,\partial)$ is a graded

3727:      commutative

3728:      differential algebra with  unit.

3729:      \end{proposition}

3730: 

3731:    The homology group $H_*(\fA,\partial)$

3732:      inherits the structure

3733:  of a graded commutative algebra with unit.

3734: 

3735: As it was already mentioned in Section \ref{sec:informal},

3736:  it is convenient to view the Poisson algebra $\fP$ as an algebra

3737: of functions on an infinite-dimensional symplectic super-space $\bV$ with the even symplectic form

3738: $\bo=\sum k_\g^{-1}dp_\g\wedge dq_\g$.

3739: Then  the differential $d^\bh$ is   the  Hamiltonian  vector field on $\bV$ generated by the  Hamiltonian

3740: function $\bh$. One should remember, however, that the $p$-variables are formal,

3741: so all that we have is the infinite jet of the symplectic space

3742: $\bV$ along the $0$-section.  The equation $\bh|_{p=0}=\bh_{\const}$ translates into

3743:  the fact that the vector field $d^\bh$

3744: is tangent to the $0$-section, and the differential $\partial$ is just the restriction of this vector field

3745: to the $0$-section.

3746:   The higher order terms in the expansion of $\bh$ with respect

3747:   to $p$-variables define a sequence  of (co-)homological operations on the algebra $\fA$.

3748:       \medskip

3749: 

3750:     Notice  also that the differentials $D,d$ and $\partial  $ do not involve any differentiation

3751:     with respect to $t$. Hence the  differential algebras

3752:       $(\fW,D^{\bH})$, $(\fP,d^{\bh})$ and

3753:        $(\fA,\partial)$ can be viewed

3754:              as {\it families} of differential algebras, parameterized

3755:       by $t\in H^*(V)$, and in particular,

3756:       one can compute the homology at any fixed $t\in H^*(V)$.   We will

3757:       sometimes denote

3758:        the corresponding

3759:        algebras and their homology

3760:        groups with the subscript $t$, i.e.

3761:         $(\fW,D)_t$,  $H_*(\fP,d)_t,$ etc., and call them

3762:        {\it specialization} at the point $t\in H^*(V)$.

3763:   We will also use the notation

3764:   $$H_*^{\SFT}(V,\xi|\,J,\a ),\;\;

3765:     H_*^{\RSFT}(V,\xi|\,J,\a  ), \quad\hbox{and }\quad

3766:   H_*^{\cont}(V,\xi|\,J,\a )$$

3767:      instead of

3768:    $H_*(\fW,\partial), H_*(\fP,\partial)$ and $H_*(\fA,\partial)$,

3769:    and

3770:    will usually omit the   extra data $J,\a$ from the notation:

3771:   as we will see in Section \ref{sec:composition} below all these

3772:   homology algebras are independent  of  $J,\a$ and other extra

3773:   choices, like

3774:   closed forms representing cohomology classes of $V$, a coherent orientation of the moduli spaces,

3775:   etc. The

3776:   abbreviation $\RSFT$ stands here for  {\it Rational Symplectic

3777:   Field Theory}.

3778: \begin{remark}\label{rem:grading}

3779: {\rm It is   important to observe that the algebras $\fW,\fP $ and

3780: $\fA$ have an additional grading by elements of $H_1(V)$ (comp. Section

3781: \ref{sec:Floer} above). This

3782: grading is also inherited by the corresponding homology algebras. However, this grading carries a

3783: non-trivial information only when we consider homology of algebras, specialized at points

3784: $t=\sum\limits_1^K t_i\Theta_i$ with  $t_i=0$ for at least some of the

3785:   coordinates $t_i$ corresponding

3786: $1$-dimensional forms.  Otherwise

3787: all cycles in these algebras are graded by  the $0$-class from $H_1(V)$.}

3788: \end{remark}

3789: 

3790: \subsection{  Symplectic cobordisms}\label{sec:af-cobordism}

3791: 

3792: 

3793: 

3794: \subsubsection{Evaluation maps and correlators}

3795:  \label{sec:sympl-eval}

3796: Let us now repeat the constructions of the previous section for a

3797: general directed symplectic cobordism $W=\ora{V^-V^+}$

3798:  between  two contact manifolds

3799:  $V^-$ and $V^+$ with fixed contact forms $\alpha^-$ and $\alpha^+$.

3800:  As in Section \ref{sec:contact-eval} we consider the sets $\Pc^\pm$

3801:   of periodic orbits of the Reeb fields $R^\pm=R_{\alpha^\pm}$ as   discrete topological spaces.

3802: 

3803: We denote by $\Mc^A_{g,r,s^-,s^+}( W,J)$ the  disjoint union

3804:    $$\bigcup\Mc_{g,r}^A(\Gamma^-, \Gamma^+;W,J),$$

3805:    where the union is taken over all sets $(\Gamma^-, \Gamma^+)$ of cardinality $(s^-,s^+) $, and

3806:   consider  three sets of evaluation maps:

3807:             $$ev^0_i: \Mc^A_{g,r,s^-,s^+}( W,J) \to W,\;i=1,\dots,r,$$

3808:             $$ev^\pm_j:\Mc^A_{g,r,s^-,s^+}(W,J)    \to \Pc^\pm,\;j=1,\dots      ,s^\pm,$$

3809:               where $ev_i^0$ is the  evaluation map $f (y_i)$ of

3810:                the map $f$  at the $i$-th marked point $y_i$, while

3811:                             $ev^\pm_j$   are the evaluation maps

3812:                             at asymptotic markers

3813:                             $\mu^{\bx^\pm}_j$, i.e.

3814:   $ev_j^\pm(f)$ is a point of $

3815:               \Pc^\pm$ representing the orbit $\g^\pm_j$, which contains the image of the

3816:               corresponding marker.

3817:                The evaluation maps $ev^0_i$ and $ev^\pm_j$ can be combined into a map

3818:               $$ev:\Mc^A_{g,r,s^-,s^+}(W,J)\to W^r\times (\Pc^+)^{s^+}\times (\Pc^-)^{s^-}\,.$$

3819: 

3820:             Now we are ready to define   degree $0$, or symplectic

3821:    correlators.  We will have to consider on $W$ differential

3822:   forms, which do not necessarily have compact support, but

3823:  have,   however, {\it cylindrical ends}. We say that a

3824:    differential  form  $\theta$ on $W$ is said to have cylindrical ends

3825:   if it satisfies the following

3826:   condition:

3827: \smallskip

3828: 

3829: \noindent there exists  $C>0$  such that

3830: $$\theta|_{V^-\times(-\infty,-C)}=(\pi^-)^*(\theta^-)\quad\hbox{ and}\quad

3831:  \theta|_{V^+\times(C,\infty)}=(\pi^+)^*(\theta^+),$$ where

3832: $\theta^\pm$ are forms on $V^\pm$, and $\pi^\pm$ are the

3833: projections of the corresponding ends to $V^{\pm}$.

3834:  We will denote the forms $\theta^\pm$ also by  $\restr^\pm(\theta)$, or $\theta|_{V^\pm}$.

3835:   In what follows we assume that all considered differential  forms on $W$ have   cylindrical ends.

3836:  \smallskip

3837: 

3838:   Given $r$    differential

3839: forms $\theta_1,\dots,\theta_r$ on $W$ and $s^\pm$   cohomology

3840: classes $$\alpha^\pm_1,\dots, \alpha^\pm_{s^\pm}\in

3841: H^*(\Pc^\pm)=H^*_0(\Pc^\pm)$$  we define the degree $0$

3842: correlator

3843: \begin{eqnarray}

3844: \up{0}\langle\,\theta_1,\dots,\theta_r;\alpha^-_1,\dots,

3845: \alpha^-_{s^-};\alpha^+_1,\dots,

3846: \alpha^+_{s^+} \rangle^A_{g}= \hfill \\

3847: \nonumber  \int\limits_{\overline{\Mc_{g,r,s^-,s^+ }^A}}ev^*(\theta_1\otimes\dots

3848: \otimes \theta_r\otimes\alpha^-_1\otimes\dots\otimes\alpha^-_{s^-}

3849: \otimes\alpha^+_1\otimes\dots\otimes\alpha^+_{s^+} ).

3850: \end{eqnarray}

3851: 

3852: Similar to

3853:   Section \ref{sec:contact-eval} above, we denote  the cohomology classes in

3854:   $H^*(\Pc^+)= H^0(\Pc^+)$

3855:   (resp. in   $H^*(\Pc^-) = H^0(\Pc^-)$) by

3856:   $p^+$ (resp.   $q^-$), and write $$p^+=\sum\limits_{\g\in\Pc^+}k_g^{-1}p^+_\g[\g]\quad

3857:   \hbox{\big(resp.}\quad q^-=\sum\limits_{\g\in\Pc^-}k_g^{-1}q^-_\g[\g]\,\big).$$

3858:    We will also fix a basis $A_1,\dots,A_N$ of

3859:  $H_2(W)$ and denote

3860:  by $d=(d_1,\dots,d_N)$ the degree of $A$ in this basis.

3861: %CHANGE

3862: 

3863: Let us call a system of linearly independent

3864:  closed forms  $\theta_1,\dots,\theta_m $

3865:   on $W$ with cylindrical ends

3866:  {\it basic}, if

3867:            \begin{equation}\label{eq:basic}

3868:  \begin{split}

3869:  &\hbox{a) the image}\quad

3870:     \restr^\pm(L(\theta_1,\dots,\theta_m ))\\

3871:     &\quad\hbox{ generates }\quad \im (H^*(W)\to H^*(V^\pm)); \\

3872:     & \hbox{b) the homomorphism}

3873:  \quad \Ker\big((\restr^+\oplus \restr^-)|_L\big)\to

3874:  H_{\mathrm{comp}}^*(W)\\

3875:  &\quad\;\hbox {is bijective.}\\

3876:  \end{split}

3877:  \end{equation}

3878:         Here we denote by $L(\theta_1,\dots,\theta_m )$ the subspace

3879:  generated by the forms $\theta_1,\dots,\theta_m $, and by

3880:  $H_{\mathrm{comp}}^*(W)$ the cohomology with compact support.

3881:  Equivalently, one can say that a basic system of forms consists

3882:  of a basis of $H^*(W)$ together with a basis of

3883:  $\Ker\big(H^*_{\comp}(W)\to H^*(W)\big)$.

3884: 

3885: A general point     $t\in L(\theta_1,\dots,\theta_m )$  we will

3886:    write in the form

3887:     $t=\sum\limits_1^mt_i\theta_i.$

3888:    The grading of the variables

3889:  $t,p^+,q^- $  is defined   as in Section

3890:  \ref{sec:correlators}:

3891:  \begin{equation}

3892:  \begin{split}

3893:  \deg(t_i) =&\;\deg(\theta_i)-2;\,\\

3894:   \deg (p_\gamma^+) =&-\CZ(\gamma^+)+(n-3),\\

3895:    \deg (q_\gamma^-) =&\; \CZ(\gamma^-)+(n-3)\\

3896:    \end{split}

3897:    \end{equation}

3898: 

3899: 

3900: 

3901: 

3902: 

3903: 

3904: \subsubsection{Potentials of symplectic cobordisms}\label{sec:relative}

3905:   Let us now organize the  correlators

3906:    into a generating function, called the { \it potential} of the

3907:    symplectic

3908:    cobordism $(W=\ora{V^-V^+},J,\a^\pm)$

3909:   \begin{equation}\label{eq:GW-potential}     \begin{split}

3910:   &{ { \bF}}=

3911:    {  \bF}_{W,J,\alpha^\pm} =\frac1\hbar

3912:    \sum\limits_{g=0}^\infty {\bF}_g\hbar^g,   \\

3913:    &\quad\hbox{  where }\\

3914:   &{\bF }_g=\sum\limits_d\sum\limits_{r  ,s^\pm=0}\frac{1}{r!s^+!s^-!}\up0\langle

3915:     \underbrace{t,\dots,t}_r;

3916:  \underbrace{q^-,\dots,q^-

3917:  }_{s^-};\underbrace{p^+,\dots,p^+}_{s^+}

3918:  \rangle^d_{g}z^d\,.\\

3919:  \end{split}

3920:  \end{equation}

3921: 

3922: When $W$ is a closed symplectic manifold, then the potential $F$

3923: is just the Gromov-Witten invariant of the symplectic manifold

3924: $W$. However, if $W$ is not closed, then $F$ itself is not an

3925: invariant. It depends on    particular forms

3926: $\theta_i$, rather than their cohomology classes, on $J$, on a

3927: coherent orientation, and several other choices. We will see,

3928: however, that the {\it homotopy class} of $F$, which we define in

3929: Section \ref{sec:chain} below, is independent of most of these

3930: choices.

3931: \medskip

3932: 

3933: In order to make sense out of the expression for

3934: $\bF$ let us consider a graded commutative

3935:  algebra $ \fD=\fD({W,\a^\pm })$  which consists of

3936: power series of the form

3937: \begin{equation}

3938: \sum\limits_{\Gamma,d,g}\varphi_{\Gamma,d,g} (q^-,t ) z^d(p^+)^{\Gamma}\hbar^g,

3939: \end{equation}

3940: where $\varphi_{\Gamma,d,g}$ are polynomials

3941:  of $q_\g$, formal power series of variables

3942:   $t_i$,\footnote{See the first footnote in Section \ref{sec:3algebras}.}

3943:  and where $\Gamma$ and

3944: $d$ satisfies the following Novikov type inequality:

3945: \begin{equation}\label{eq:Novikov}

3946: [\omega](d)=\sum

3947: d_i\int\limits_{A_i}\omega>-\left|\Gamma\right|=-\sum\limits_{i=1}^k|\gamma_i|,

3948: \end{equation}

3949: where

3950: $\Gamma=\{\g_1,\dots,\g_k\}$, and

3951: $|\gamma_i|=\int\limits_{\gamma_i}\alpha^+$ is the period of the

3952: periodic orbit $\g_i\in\Pc^+$, or equivalently its action. Recall

3953: that $(p^+)^\Gamma=p^+_{\g_1}\dots p^+_{\g_k}$.

3954: 

3955: 

3956: 

3957:  \begin{proposition}

3958: We have

3959: $$\bF_{W,J,\alpha^\pm}

3960: \in\frac1{\hbar}\fD({W,\alpha^\pm}).$$

3961: \end{proposition}

3962: Let us also consider a bigger algebra $\fD\fD$ which consists

3963: of series

3964: \begin{equation}

3965: \sum\limits_{\Gamma,d}\varphi_{\Gamma,d} (q^-,t,\hbar ) z^d(p^+)^{\Gamma},

3966: \end{equation}

3967: where $\varphi_{\Gamma,d}$ are polynomials of $q^-_\g$, formal power series of $t_i$ and formal {\it Laurent

3968: series}

3969: of $\hbar$, while $\Gamma$ and $d$  still satisfy

3970:  condition

3971: (\ref{eq:Novikov}).

3972: For instance, for any element $f\in \hbar^{-1}\fD$ we have $e^f\in\fD\fD$.

3973: 

3974: The algebra $\fD\fD=\fD\fD(W,J,\alpha^\pm)$ has a structure

3975: of a {\it left $D$-module} over the Weyl algebra

3976: $\fW^-=\fW(V^-,J,\alpha^-)$, and of a {\it right $D$-module} over the Weyl algebra

3977: $\fW^+=\fW(V^+,J,\alpha^+)$.

3978: Indeed,  we first associate with an element

3979: $$\Delta^-= \sum\limits_{\Gamma=\{\g_1,\dots,\g_m\},\Gamma',d,g }

3980:   {\delta}^-_{\Gamma',\Gamma,d,g  }( t )

3981:    (q^-)^{\Gamma'}(p^-)^{\Gamma}z^d\hbar^g\in\fW^-$$

3982:    a differential operator

3983: 

3984: \begin{equation}

3985: \sum\limits_{\Gamma=\{\g_1,\dots,\g_m\},\Gamma',d,g }

3986:  {\delta}^-_{\Gamma,\Gamma',d,g }(t )

3987:    (q^-)^{\Gamma'}{\hbar^{m+g}}\prod\limits_{i=1}^m

3988:    \kappa_{\g_i}

3989:    \overrightarrow{\frac{\partial}

3990: {\partial q^-_{\g_i}}}z^d,

3991: \end{equation}

3992: then  change   the coefficient ring via  the inclusion homomorphism

3993: $H_2(V^-)\to

3994: H_2(W)$,  and finally lift functions

3995: $ {\delta}^-_{\Gamma,\Gamma',d,g }(t )$ to the space of forms with cylindrical

3996: ends on $W$ via the restriction map $t\mapsto t|_{V^-}$.

3997:  We will   denote  the resulting operator by

3998: $\overrightarrow{\Delta^-}$. Similarly we associate with

3999: $\Delta^+\in\fW^+$ an operator $ \overleftarrow{\Delta^+}$ by

4000: first quantizing $q^+_\g\Rightarrow  \hbar

4001: \kappa_\g\overleftarrow{\frac{\partial}{\partial p^+_\g}}$  and then making an

4002: appropriate change of the coefficient ring.

4003:                                    It is straightforward to verify that

4004: for $f\in \fD$ we have $\overrightarrow{\Delta^-}f,\;f \overleftarrow{\Delta^+}\in \fD$,

4005: and for $F\in\fD\fD$ we have

4006: $\overrightarrow{\Delta^-}F,\; F\overleftarrow{\Delta^+}\in \fD\fD$.

4007: 

4008: \medskip

4009: 

4010: Let us denote the Hamiltonians (see Section \ref{sec:3algebras} above)  in

4011: $\fW^\pm$ by $\bH^\pm$ and define a map

4012: $D_W=D_W:\fD\fD\to\fD\fD$ by the formula

4013: 

4014: \begin{equation}\label{eq:DW}

4015: D_W(G)= \overrightarrow{

4016: \bH^-}G-(-1)^{\deg G}G\overleftarrow{\bH^+},\quad G\in\fD\fD,

4017: \end{equation}

4018: where we assume $G$ dimensionally homogeneous.

4019: Clearly, Theorem \ref{thm:HH} implies that $D_W^2=0$. However, the differential algebra

4020: $(\fD\fD,D_W)$ is too big

4021: and instead of considering its homology we will define a differential on the algebra $\fD$,

4022: or which is equivalent but more convenient, on the module $\hbar^{-1}\fD$.

4023: 

4024: For an {\it even} element $F\in\hbar^{-1}\fD$  let us define  a map

4025:  $D^F=T_FD_W:\hbar^{-1}\fD\to\hbar^{-1}\fD$

4026: by the formula

4027: \begin{equation} \label{eq:DF}

4028: D^F(g)=e^{-F}[D_W,g](e^F)=e^{-F}\big(D_W(ge^F)-(-1)^{\deg g}g D_W(e^F)\big),

4029: \quad g\in\hbar^{-1}\fD.

4030: \end{equation}

4031:       The map    $T_FD_W$ is         the linearization of  the map

4032:        $\wt D_W:\hbar^{-1}\fD\to\hbar^{-1}\fD$,

4033:       defined by the formula $$\wt D_W(F)=e^{-F}D_W(e^F),\quad F\in\hbar^{-1}\fD\,.$$

4034:             at the point $ F\in\hbar^{-1}\fD$.  Notice that   if $D_W(e^F)=0$ then

4035:             $D^F(g)=e^{-F}D_W(ge^F)$.

4036:       Let us first state a purely algebraic

4037:       \begin{proposition} \label {prop:SFT-cobordism}

4038:       Suppose that     for $F\in\hbar^{-1}\fD$ we have $D_W(e^F)=0$. Then

4039:    \begin{description}

4040:     \item{1.}   $(D^F)^2=0;$

4041: \item{2.}  The homology algebra         $H_*(\fD,D^{F})$

4042: inherits the structure of a left module over the homology

4043: algebra

4044: $ H_*(\fW^-,D^-)$, and

4045:  the structure of a right module over the homology

4046: algebra $ H_*(\fW^+,D^+)$;

4047:   \item{3.} The homomorphisms

4048: $F^\pm:  \fW^{\pm}  \to \fD $,

4049: defined by  the formulas

4050: \begin{equation} \label{eq:end-homomorphisms}

4051: \begin{split}

4052: &f\mapsto e^{-F}

4053: \overrightarrow{f}e^{F},\,f\in\fW^-,\quad\hbox{and}\\

4054: & f\mapsto  e^{F} \overleftarrow{f}e^{-F},\,f\in\fW^+\,,\\

4055: \end{split}

4056: \end{equation}

4057: commute with the  boundary maps of chain complexes, i.e.

4058: $$F^\pm\circ D^\pm= D^{F}\circ F^\pm,$$ and thus induce

4059: homomorphisms of homology

4060:  $$F^\pm_*:H_*(\fW^{\pm},D^\pm)

4061:  \to H_*(\fD,D^{F}),$$

4062:  as modules over $H_*(\fW^{\pm},D^\pm)$.

4063:  \end{description}

4064:  \end{proposition}

4065: 

4066: \begin{theorem}\label{thm:SFT-cobordism}

4067: The   potential $ \bF\in\hbar^{-1}\fD$  defined  above by the formula

4068: (\ref{eq:GW-potential})

4069: satisfies the equation

4070: \begin{equation}\label{eq:SFT-cycle}

4071: D_We^{\bF}=0 ,

4072: \end{equation} and hence all conclusions of Proposition \ref{prop:SFT-cobordism}

4073: hold for $\bF$.

4074:  \end{theorem}

4075: 

4076:   The appearance of $e^{\bF}$ in  equation

4077: (\ref{eq:SFT-cycle}) has the following reason.

4078:    Similar to Theorem

4079: \ref{thm:HH} above, equation (\ref{eq:SFT-cycle}) follows from the

4080: description of codimension $1$ strata on the boundary of the

4081: moduli space of holomorphic curves in the cobordism $W$, see

4082:  Proposition \ref{prop:boundary} above.   Notice

4083: that $e^{\bF}$ is the generating function  counting possibly disconnected

4084:  holomorphic curves in $W$. Thus the identity

4085: $$\overrightarrow{\bH^-} e^{\bF}- e^{\bF}\overleftarrow{\bH^+}=0$$

4086: asserts, in agreement with Proposition \ref{prop:boundary},

4087: that the  codimension $1$ strata on the  boundary of the moduli space

4088: $\wt\Mc(W)$

4089: of not necessarily connected curves

4090: in $W$ correspond

4091:   to  stable

4092: curves $(f_1,f_2)$ of height $2$, where one of the  curves $f_1,f_2$ belongs to $\wt\Mc(W)$,

4093:  while the second one

4094: is contained in the symplectization of $V^\pm$  and has

4095: precisely one component different from the   straight cylinder

4096: over a periodic orbit from $\Pc^\pm$.

4097: 

4098: 

4099: \medskip

4100: 

4101: \begin{remark}\label{rem:non-closed2}

4102: {\rm

4103: (Comp. Remark \ref{rem:non-closed1} above) The   potential

4104: $\bF$, extended to the space of all, not necessarily closed

4105:  differential forms

4106:  satisfies the   equation

4107: \begin{equation}\label{eq:non-closed2}

4108: d(e^\bF)=D_We^{\bF},

4109: \end{equation}

4110: where $d$ is the de Rham differential.

4111: This equation generalizes

4112: equation   (\ref{eq:SFT-cycle}).

4113: }

4114: \end{remark}

4115:  Following the scheme of Section \ref{sec:3algebras} above we will associate

4116:  now with the cobordism $W$ two other left-right modules, one over

4117:  the

4118:  Poisson algebras $\fP^\pm$, and another over the graded differential

4119:  algebras $\fA^\pm$.

4120: 

4121:  Consider the graded commutative algebra  $\fL=\fL({W,\a^\pm })$

4122:   of

4123: power series of the form

4124: \begin{equation}

4125: \sum\limits_{\Gamma,d}\varphi_{\Gamma,d} (q^-,t ) z^d(p^+)^{\Gamma},

4126: \end{equation}

4127: where $\varphi_{\Gamma,d}$ are polynomials of $q_\g^-$ and formal power seies of $t_i$, while $\Gamma$ and

4128: $d$  satisfies the  above inequality (\ref{eq:Novikov}).

4129: Let us also consider the larger graded commutative algebra

4130: \begin{equation}\label{eq:fL}

4131: \widehat{\fL}=\{\sum\limits_{\Gamma^+,\Gamma^-,d}

4132: \varphi_{\Gamma^+,\Gamma^-,d} (q^-,q^+,t )

4133: z^d(p^+)^{\Gamma}(p^-)^{\Gamma'}\},

4134: \end{equation}

4135: where the Novikov condition (\ref{eq:Novikov}) is satisfied for both pairs

4136: $(d,\Gamma)$ and $(d,\Gamma')$.

4137:    The algebra $\widehat{\fL}$

4138:  has a natural Poisson  bracket so that the homomorphisms

4139: $f\mapsto \widehat{f}$,

4140:  where we  denote by $ \widehat{f}$

4141: the image in $\widehat{\fL}$ of an element $f\in\fP^\pm$ under the

4142: coefficient homomorphism, are Poisson homomorphisms.  We set

4143: $\widehat{\bh}=\widehat{\bh^-}-\widehat{\bh^+}$, and   for any $f\in\fL$ denote

4144: by $L_{f}$ the ``Lagrangian variety''

4145:  $$\{p^-_\g=\kappa_\g\frac{\partial

4146: \widehat{f}}{\partial q^-_\g}, q^+_\g=\kappa_\g\frac{\partial

4147: {\widehat f }}{\partial p^+_\g}\}. $$ Strictly speaking $L_f$ is

4148: an ideal in the Poisson algebra $\widehat{L}$. However, it is

4149: useful to think about $L_f$ as a Lagrangian variety in the

4150: symplectic super-space $(\bV^-)\oplus\bV^+$ with the symplectic

4151: form $$\sum \kappa^{-1}_{\g^-}dp^-_{\g^-}\wedge

4152: dq^-_{\g^-}+\kappa^{-1}_{\g^+}dq^+_{\g^+}\wedge dp^+_{\g^+}\,,$$

4153: and with an appropriate change of the coefficient ring.

4154: 

4155: For any function $f\in\fL$, which satisfies the Hamilton-Jacobi

4156: equation

4157: \begin{equation}\label{eq:HJ1}

4158: \widehat\bh|_{L_f}=0

4159: \end{equation}

4160:  the Hamiltonian vector

4161: field defined by the Hamiltonian $\widehat{\bh}$ is tangent to

4162: $L_{f}$, and hence

4163:   the differential $d^f:\fL\to\fL$, defined by the formula

4164:   $$d^f(g)=\{\widehat\bh,g\}|_{L_f}$$ has the following meaning: we identify

4165:   $\fL$ with the space of functions on $L_f$ and differentiate

4166:   them along the Hamiltonian vector field determined by $\widehat\bh$.

4167: 

4168: Here is an analog of Proposition \ref{prop:RSFT-cobordism}

4169: for the algebra

4170: $\fL$.

4171: \begin{proposition}\label{prop:RSFT-cobordism}

4172:  Suppose that    $\widehat\bh|_{L_f}=0$.

4173:  Then

4174:  \begin{description}

4175:  \item{1.} $(d^f)^2=0;$

4176:  \item{2.} The maps $f^\pm:\fP^\pm\to\fL$, defined by the

4177: formula $g\mapsto {\widehat{g}}|_{L_{f}}$, are homomorphisms of

4178: chain complexes, i.e. $$d^{f}\circ f^\pm=f^\pm\circ

4179: d^\pm;$$

4180: \item{3.} If  $g_1,g_2\in\fP^\pm$ Poisson commute with $\bh^\pm$

4181:  or, in other words, if $g_1,g_2\in{\mathrm {Ker}}\,d^\pm$

4182: then

4183: \begin{equation*}

4184: \{f^\pm(g_1),f^\pm(g_2)\}=f^\pm(\{g_1,g_2\}).

4185: \end{equation*}

4186: where the left-side Poisson bracket is taken  in the algebra $\widehat{\fL}$.

4187: \end{description}

4188: \end{proposition}

4189: Let us recall that $\bF=\bF_W\in\fD$ has the form

4190: $\bF=\hbar^{-1}\sum\limits_{g=0}^\infty\bF_g\hbar^g$.

4191: Again, to simplify the notation we will write $\bff $ instead of

4192: $\bF_0$.

4193: The following theorem is the reduction of Theorem

4194: \ref{thm:SFT-cobordism} to the level of rational Gromov-Witten theory.

4195: \begin{theorem}\label{thm:RSFT-cobordism}

4196: The series $\bff(q^-,p^+,t)$ belongs to the algebra $\fL$ and satisfies

4197: the  equation

4198: \begin{equation}\label{eq:Lagr}

4199: \widehat{\bh}|_{L_{\bff}}=0.

4200: \end{equation}

4201: \end{theorem}

4202: In particular, all statements of  the above Proposition

4203: \ref{prop:RSFT-cobordism} hold for $\bff$, and  this allows

4204: us    to   define    the homology

4205: \begin{equation*}

4206: {}_{\bff}H^{\RSFT}_*(W|J,\a^\pm)=H_*(\fL,d^{\bff}).

4207: \end{equation*}

4208:   The chain maps $\bff^{\pm}$ induce homomorphism of Poisson homology

4209: algebras

4210: \begin{equation}

4211: (\bff^\pm)_*:H^{\RSFT}_*(V^\pm|J,\a^\pm)=H_*(\fP^\pm,d^\pm)\to

4212: {}_{\bff}H^{\RSFT}_*(W|J,\a^\pm)\,.

4213: \end{equation}

4214: 

4215: \bigskip

4216: For the rest of this section we assume that $W$ is a  rational

4217: homology

4218: cobordism, i.e. the restriction maps $$H^*(V^-;\R)\leftarrow

4219: H^*(W;\R)\rightarrow H^*(V^+,\R)$$ are isomorphisms.  Equivalently,

4220: this means that

4221:   the inclusions $V^\pm\to W$ induce isomorphisms of rational

4222:   homology groups.

4223: 

4224: 

4225: The potential $\bff\in\fL$ which we defined above can be written in the

4226:   form

4227: \begin{equation}

4228: \bff=\sum\limits_i\sum\limits_{|\Gamma^+|=i}f_{\Gamma}^i(q^-,t)(p^+)^{\Gamma^+}.

4229: \end{equation}

4230:                         Notice that the assumption that $W$

4231:                          is a homology cobordism implies

4232:   that $f^0(q^-,t)$ is independent of $q^-$.

4233:  Let us now define a homomorphism

4234:  $\bPsi : \fA^+\to \fA^-$

4235:  by the formula

4236:  \begin{equation}

4237:  \bPsi (q^+_{\g } )=f^1_{ \g }(q^-,t)\in\fA^-

4238:  \end{equation}

4239: on the generators $q^+_\g,\g\in\Pc^+$,

4240: of the algebra $\fA^+$ and then   extend by

4241: linearity.

4242: 

4243: \begin{theorem}   The homomorphism

4244: $\bPsi : \fA^+\to \fA^-$ commutes with the boundary operators

4245: $\partial^\pm$, i.e. $\partial^-\circ\bPsi =\bPsi

4246: \circ\partial^+$, and in particular defines a homomorphism of

4247: homology algebras

4248: \begin{equation*}

4249: (\bPsi )_*:H_*(\fA^+,\partial^+)\to H_*(\fA^-,\partial^-).

4250: \end{equation*}

4251: \end{theorem}

4252: 

4253: Without the assumption that $W$ is a homology cobordism

4254: one gets only a correspondence between the algebras $\fA^+$ and $\fA^-$, similar to the

4255: ``semi-classical" case considered above.

4256: 

4257: \subsection{Chain homotopy}\label{sec:chain}

4258: Let $W=\ora{V^-V^+}$ be a directed symplectic cobordism with fixed

4259: contact forms $\a^\pm$ on $V^\pm$. We will discuss in this section

4260: how the function $\bF=\bF_{W,J,\a^\pm}(p^+,q^-,t)$ and other

4261: associated structures change when one replaces  $J$ with another

4262: compatible almost complex structure $J'$ and replaces $t$ with a

4263: form $t'=t+d\theta$ where $\theta$ has   compact support in $W$.

4264: 

4265: Let us begin with some algebraic preliminaries. Two series

4266: $F_0,F_1\in\hbar^{-1}\fD$ are called {\it homotopic}, if they can

4267: be included into  a family $F_s\in\hbar^{-1}\fD,\,s\in[0,1]$,

4268: which satisfies the following differential equation

4269: \begin{equation} \label{eq:homotopy}

4270:     \frac{d F_s}{ds }=

4271:     e^{-F_s}

4272:     \left(\ora{[\bH^-,K_s]}e^{F_s}+

4273:      e^{F_s}\overleftarrow{[

4274:      K_s,\bH^+]}\right),\,\,s\in[0,1],

4275:     \end{equation}

4276:      for a family $K_s\in\hbar^{-1}\fD$.

4277:      Here  $[\bH^-,K_s]$ and $[K_s ,\bH^+]$ are commutators in the

4278: algebra $\hbar^{-1}\widehat{\fD}$, defined similar to

4279: $\widehat{\fL}$ in (\ref{eq:fL}) above, i.e.

4280: \begin{equation}

4281: \widehat{\fD}=\{\sum\limits_{\Gamma^+,\Gamma^-,d,g}

4282: f_{\Gamma^+,\Gamma^-,d} (q^-,q^+,t )

4283: z^d\hbar^g(p^+)^{\Gamma^+}(p^-)^{\Gamma^-}\},

4284: \end{equation}

4285: where the Novikov condition (\ref{eq:Novikov}) is satisfied for

4286: both pairs $(d,\Gamma^+)$ and $(d,\Gamma^-)$.

4287:       In other words, we view $K_s$ as an operator on

4288:        $\hbar^{-1}\fD$, acting by the multiplication

4289:     by the series $K_s$,

4290:     and  view $\bH^-$ and $ \bH^+  $

4291:     as left and right differential operators.

4292: 

4293: 

4294: Notice that the family $ K_s\in\hbar^{-1}\fD, s\in [0,1],$

4295: defines a flow $\Phi^s=\Phi_K^s: \fD\fD\to\fD \fD$,

4296:   by a differential equation

4297:    \begin{equation} \label{eq:chain-flow}

4298:     \frac{d \Phi^s(G) }{ds }=\cK_s\big(\Phi^s(G)\big),

4299:  \end{equation}

4300:  where we set

4301: \begin{equation*}

4302:     \cK_s(G)=\left(\ora{[\bH^-,K_s]}G+

4303:     G\overleftarrow{[ K_s,\bH^+]}\right),\,\,s\in[0,1].

4304:  \end{equation*}

4305: The linear operators $\Phi^s$ preserve  the ``submanifold" $\fE=e^{

4306: \hbar^{-1}\fD_{\even}}$, where $\fD_\even$ is the even part of

4307: $\fD$, and we have $$\Phi^s(e^{F_0})=e^{F_s},$$ where

4308: the family $F_s$ satisfies the equation (\ref{eq:homotopy}).

4309: 

4310: The tangent space to $\fE$ at a point $e^F$, $F\in \hbar^{-1}\fD_\even$, consists of

4311: series $fe^F$, $f\in\hbar^{-1}\fD_\even$, and thus it is naturally parameterized by

4312: $\hbar^{-1}\fD_\even$. With respect to this parameterization  the

4313: differential  of the flow $\Phi^s|_\fE$ defines a family of maps

4314: $  T_F^s:\hbar^{-1}\fD_\even\to\hbar^{-1}\fD_\even$, $F\in \hbar^{-1}\fD_\even$, by the formula

4315:  \begin{equation}\label{eq:lin-flow}

4316:     T_F^s(f)=e^{-F_s}\Phi^s(fe^F),\;\;\hbox{ where

4317:     }\;\;F_s=\Phi^s(F);

4318:   \end{equation}

4319:   We extend   $T_F^s$ to the whole

4320:    $\hbar^{-1}\fD$ by the same formula

4321:    (\ref{eq:lin-flow}).

4322: Let us list some properties of the flows $\Phi^s$ and $T^s_F$

4323: \begin{proposition}\label{prop:SFT-chain}

4324: Suppose that    for an  element  $F\in\hbar^{-1}\fD_\even$ we

4325:  have $$D_W(e^F)=\ora{\bH^-}e^F -e^F\ola{\bH^+}=0.$$

4326:     Then

4327: \begin{description}

4328: \item{1.}

4329:        The   flow  $T^s_F :\hbar^{-1}\fD\to\hbar^{-1}\fD$

4330: satisfies the equation

4331: \begin{equation}

4332:  T_F^s\circ D^{F}=D^{F_s}\circ T_F^s.

4333:   \end{equation}

4334:   for all $s\in[0,1].$

4335:   In particular,    $D_W(F_s)=0$  for all $s\in[0,1]$, and $T_F^s$

4336:        defines a family of isomorphisms

4337: $ H_*(\fD,D^{F})\to

4338:  H_*(\fD,D^{F_s})$.

4339: \item{2.} The homology class

4340: $[e^{F_s}]\in H_*(\fD \fD,D_W)$ is independent of  $s$.

4341:                     \item{3.} The diagram

4342: \bigskip\bigskip

4343: 

4344: \centerline{$ \mathop{\fD\qquad}\limits_{\;

4345:  \; \nwarrow \, F^\pm}

4346: \mathop{\longrightarrow}\limits^{T_F^s}

4347: \mathop{\fD}\limits_{\nearrow \, F_s^\pm}$}

4348: 

4349: \centerline{$\fW^\pm$}

4350: 

4351: homotopically commutes, i.e. there exist operators

4352: $A_s^\pm:\fW^\pm\to\fD$,

4353: such that

4354: \begin{equation}\label{eq:SFT-diagram}

4355: (T_F^s)^{-1}\circ F_s^\pm- F^\pm=D^{F}\circ A_s^\pm+ A_s^\pm\circ

4356: D^\pm\,.

4357: \end{equation}

4358: In particular,

4359: this diagram commutes on the level of homology algebras.

4360:   \end{description}

4361:   \end{proposition}

4362:   %%%%%%%%%%%%%%%%%%%%%%%%%

4363: The proof of this proposition is a  straightforward computation by

4364: differentiating  the corresponding equations.

4365: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4366: To illustrate the argument, let us   verify

4367: (\ref{eq:SFT-diagram})

4368: in \ref{prop:SFT-chain}.

4369: 

4370: Take, for instance, $f\in\fW^-$ and

4371: set

4372:  \begin{equation*}

4373: \begin{split}

4374: G_s(f)=&(T_F^s)^{-1}(e^{-F_s}\ora fe^{F_s}))-e^{-F}\ora fe^F\\&=

4375: e^{-F}(\Phi^s)^{-1}\left(\ora fe^{F_s}\right)-e^{-F}\ora fe^F.\\

4376: \end{split}

4377: \end{equation*}

4378: 

4379: 

4380: 

4381: Then we have $G_0(f)=0 $ and

4382: 

4383: \begin{equation*}

4384: \begin{split}

4385: \frac{dG_s(f)}{ds}

4386: &= -e^{-F}\left((\Phi^s)^{-1}

4387:  (  \cK_s \ora{f}  e^{F_s }  )\right)

4388: +  e^{-F}

4389: (\Phi^s)^{-1}\left(

4390:   \overrightarrow{f} \cK_s\left( e^{F_s }\right)\right)

4391:  \\

4392: & =           e^{-F}(\Phi^s)^{-1}\left(

4393:            \left[\cK_s,\overrightarrow{f}\right]

4394:             e^{F} \right).  \\

4395: \end{split}

4396: \end{equation*}

4397: Now remember that $\cK_s=[\widehat{\bH}  ,K_s]$,

4398:  where

4399: $\widehat{\bH}=\overrightarrow{\bH^-}-\overleftarrow

4400: {\bH^+ }$, and using the Jacobi identity we get

4401: \begin{equation*}

4402: \begin{split}

4403:   \frac{dG_s(f)}{ds}

4404:   &=  e^{-F} (\Phi^s)^{-1}\left(

4405:            \left[[\widehat{\bH}  ,K_s],

4406:            \overrightarrow{f}\right]

4407:             e^{F_s}\right) \\

4408:       & =

4409:  e^{-F}(\Phi^s)^{-1}\left(

4410:            \left[[K,\overrightarrow{f}],

4411:            \widehat{\bH}  \right]

4412:             e^{F_s }  \right)\\

4413:             &+

4414:               e^{-F} (\Phi^s)^{-1}\left(

4415:            \left[[\overrightarrow{f},

4416:            \widehat{\bH}  ],K_s\right]

4417:              e^{F_s } \right). \\

4418:                            \end{split}

4419:          \end{equation*}

4420:          Let us define now a linear operator

4421:          $B_s:\hbar^{-1}\fW_-\to\hbar^{-1}\fD $ by the formula

4422:          \begin{equation}

4423:          B_s(g)=   e^{-F}(\Phi^s)^{-1}\left(

4424:          \overrightarrow{[g,K_s] }

4425:           e^{F_s}\right).

4426:          \end{equation}

4427:           for $g\in \hbar^{-1}\fW_-$.

4428:         Recall that

4429:          $D_W(e^F)=\widehat{\bH}e^{F}=0$, and observe that

4430:          $\widehat{\bH}$ commutes with

4431:          $\cK_s=[\widehat{\bH},K_s]$ because $\widehat{\bH}\circ

4432:           \widehat{\bH}=0$.  We also have

4433:                     $[\widehat{\bH},f]= D^-f$ for

4434:                      $f\in\hbar^{-1}\fW_-$.

4435:           Hence

4436:            $ \frac{dG_s(f)}{ds}$ can be rewritten as

4437:           \begin{equation}\label{eq:derivative}

4438:           \frac{dG_s(f)}{ds}= D^{F}(B_s(f))    +

4439:             B_s(D^-(f))

4440:           \end{equation}

4441: 

4442:         Finally we integrate $B_s$ into the required linear operator

4443:          $A^-:\hbar^{-1}\fW^-\to\hbar^{-1}\fD$:

4444:          \begin{equation}

4445:         A_s^-(g)=\int\limits_0^s B_s(g)ds,

4446:         \quad\hbox{for}\quad g\in\hbar^{-1}\fW^-.

4447:         \end{equation}

4448: 

4449:          In view of   equation

4450:          (\ref{eq:derivative})

4451:          we get

4452:   \begin{equation}

4453: (\Phi^s)^{-1}\circ(F^s)^- -(F)^-=D^F\circ A_s^-+

4454: A_s^-\circ D^-\,.

4455: \end{equation}

4456:  \qed

4457: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

4458: 

4459: \bigskip

4460:  Let us consider now a generic family $J^\tau$, $\tau\in[0,1]$,     of

4461: compatible almost complex structures on $W$ connecting $J^0=J$

4462: with $J^1=J'$. We  assume that the deformation $J^\tau$ is

4463: fixed outside of a compact subset of $W$.

4464: 

4465: \medskip

4466: Set

4467: \begin{equation}

4468: \Mc_{g,r,s^+,s^-}(W,\{J^\tau\}) =\bigcup\limits_{\tau\in[0,1]}

4469: \Mc_{g,r,s^+,s^-}(W,J^\tau).

4470: \end{equation}

4471: 

4472: The evaluation maps   defined for each $\tau$ can be

4473: combined into the evaluation map

4474: \begin{equation}

4475: ev:\Mc_{g,r,s^+,s^-}(W,\{J^\tau\})\to

4476: (W\times I)^r\times\left(\Pc^-\right)^{s^-}\times\left(\Pc^+\right)^{s^+}.

4477: \end{equation}

4478: 

4479: 

4480: Consider  closed forms

4481: $\widehat {\theta}_1,\dots \widehat{\theta}_r$    on

4482:   $W\times I$, such that

4483:   $\widehat{\theta}_i=\wt{\theta}_i+d\beta_i$, $i=1,\dots, r$, where

4484:   $\wt{\theta}_i$ is the pull-back of  a form $\theta_i$ on $W$

4485:   with cylindrical ends, and

4486:   $\beta_i$ has   compact support in $W\times \Int I$.

4487: 

4488: 

4489: 

4490: Similarly to correlators  of degree $-1$ and $0$

4491: (see \ref{sec:correlators} and \ref{sec:sympl-eval})

4492: we  can define correlators of degree $1$, or

4493: $1$-parametric correlators by the formula

4494: \begin{eqnarray}

4495:  \up1\langle\widehat {\theta}_1,\dots \widehat{\theta}_r\,;\theta^-_1,\dots,

4496: \theta^-_{s^-};\theta^+_1,\dots,

4497: \theta^+_{s^+} \rangle^A_{ g}  =

4498:   \int\limits_{\overline{\Mc_{g,r,s^+,s^-}(W,\{J^\tau\})} }\\

4499: \nonumber   ev^*(\widehat{\theta}_1\otimes\dots

4500: \otimes \widehat{\theta}_r

4501: \otimes\theta^-_1\otimes\dots\otimes\theta^-_{s^-}

4502: \otimes\theta^+_1\otimes\dots\otimes\theta^+_{s^+} ),

4503: \end{eqnarray}

4504:   for a homology class $A\in H_2(W)$, and cohomology

4505: classes $$\theta^\pm_i\in H^*(\Pc^\pm),\; i=1,\dots,s^\pm.$$

4506: 

4507: 

4508: 

4509: Consider a closed form $T=\widetilde{t}+d\beta$,

4510:  where  the notation $\wt t$ and $\beta$

4511:   have the same meaning as

4512:  above, i.e. $\wt t$ is the pull-back of a form $t$

4513:  on $W$ with cylindrical ends, and $\beta$ has

4514:    compact support in $ W\times\Int I$.

4515: We can organize the correlators

4516:  $$\up1\langle T,\dots,T\,; q^-,\dots,q^-,p^+,\dots, p^+\rangle^A_{

4517:  g}$$

4518:     into a generating function

4519: $\bK=\frac 1{\hbar}\sum\limits_{g=0}^\infty\bK_g\hbar^g\in\frac 1{\hbar}\fD$, defined

4520: by the formula

4521: \begin{equation}

4522: \bK=\sum\limits_d\sum\limits_{g,r

4523: ,s^\pm=0}^\infty

4524: \frac{1}{r!s^+!s^-!}\up1\langle \underbrace{T,\dots,T}_r;

4525:  \underbrace{q^-,\dots,q^-

4526:  }_{s^-};\underbrace{p^+,\dots,p^+}_{s^+}\rangle^d_{ g}\hbar^gz^d\,

4527:  \end{equation}

4528: 

4529: 

4530:   \medskip

4531:  Let us define an operator $\cK:\fD\fD\to\fD\fD$

4532:  by the formula

4533:  \begin{equation}

4534:  \cK(G)=  \overrightarrow{ [\bH^+,\bK]}G +G

4535: \overleftarrow{[\bK ,\bH^-]},\; G\in\fD\fD .

4536:   \end{equation}

4537: 

4538: Similar to Theorems \ref{thm:HH} and \ref{thm:SFT-cobordism}

4539: the next theorem  can be viewed

4540: as an algebraic    description of the boundary of the compactified

4541: moduli

4542: space        $\overline{\Mc_{g,r,s^+,s^-}(W,\{J^\tau\})}$.

4543: \begin{theorem}\label{thm:SFT-chain}

4544: For a generic family $J_\tau,\tau\in[0,1]$, of compatible almost complex

4545: structures on $W$ we have

4546: \begin{equation}\label{eq:SFT-chain}

4547: e^{\bF^1}= e^{\cK }(e^{\bF^0}),

4548: \end{equation}

4549: where

4550: $\bF^0=\bF_{W,J_0}(T|_{W\times 0}),\,\bF^1 =\bF_{W,J_1}(T|_{W\times 1}),$ and

4551: $\cK=\cK(T).$

4552: \end{theorem}

4553:  Notice that if we define $\bF^s$,

4554:  $s\in [0,1]$, by the formula

4555: \begin{equation}

4556: e^{\bF^s}= e^{s\cK }(e^{\bF^0}),

4557: \end{equation}

4558: then the flow $\Phi^s(F)=F^s$   satisfies the differential

4559: equation (\ref{eq:chain-flow})

4560:  with $K(s)\equiv \bK$. Hence $\bPhi^0$ and $\bPhi^1$

4561:  are homotopic, and therefore

4562:   Theorem

4563: \ref{thm:SFT-chain} and

4564: Proposition \ref{prop:SFT-chain} imply

4565: \begin{corollary}\label{cor:SFT-chain}

4566: \begin{description}

4567: \item{1.} The homology class $[e^{\bF}]\in H_*(\fD

4568: \fD,D_W)$ is

4569: independent of the choice of a compatible almost complex structure

4570: $J$ on $W$ and of the choice of

4571: $t\,{\mathrm {mod}}\left(d\left(\Omega_{\mathrm

4572: {comp}}(W)\right)\right)$, where $ {\Omega}_{\mathrm

4573: {comp}}(W)$ is the space of forms with compact support.

4574: \item{2.} For a generic compatible deformation $J^\tau$, $\tau\in[0,1]$,

4575: the isomorphism  $T :\fD\to\fD$

4576:    defined by the formula

4577: \begin{equation}

4578:     T (f)= e^{-\bF_1}

4579: e^{\cK}(fe^{\bF^0 } )

4580: \end{equation}

4581:  satisfies the equation

4582: \begin{equation}

4583:  T \circ D^{\bF^0}=D^{\bF^1}\circ T  ,

4584:  \end{equation}

4585:  and thus defines an isomorphism

4586: $ H_*(\fD,D^{\bF^0})\to

4587:  H_*(\fD,D^{\bF^1})$.

4588: \item{3.} The diagram

4589: \bigskip\bigskip

4590: 

4591: \centerline{$ \mathop{\fD\qquad}\limits_{\;

4592:  \; \nwarrow \, \left(\bF^0\right)^\pm}

4593: \mathop{\longrightarrow}\limits^{T }

4594: \mathop{\fD}\limits_{\nearrow \, \left(\bF^1\right)^\pm}$}

4595: 

4596: \centerline{$\fW^\pm$}

4597: 

4598: homotopically commutes, i.e. there exist operators

4599: $A^\pm:\fW^\pm\to\fD$,

4600: such that

4601: \begin{equation}

4602: T^{-1} \circ\left(\bF^1\right)^\pm-

4603: \left(\bF^0\right)^\pm=D^{\bF^0}\circ A^\pm+

4604: A^\pm\circ D^\pm\,.

4605: \end{equation}

4606:   \end{description}

4607: \end{corollary}

4608: 

4609: \bigskip

4610: Consider now the equivalence relation for

4611:  rational   potentials.

4612: 

4613:  Two series

4614:  $ f_0,f_1\in\fL$ are called {\it homotopic}

4615: if there exist families

4616: $f_s, k_s\in\fL$, $s\in[0,1]$, so that the family

4617:  $f_s$ connects $f_0$ and $f_1$ and satisfies

4618: the following Hamilton-Jacobi differential

4619: equation

4620: \begin{equation}\label{eq:HJ}

4621:  \frac{\partial f_s( p^+,q^-,t)}{\partial s}=

4622:  \bG(p^+,\frac{\partial f_s ( p^+,q^-,t)}

4623:  {\partial p^+},

4624:  \frac{\partial f_s ( p^+,q^-,t)}

4625:  {\partial q^-} ,q^-),

4626: \end{equation}

4627:  where

4628:   \begin{eqnarray}

4629:  \nonumber \bG(p^+,q^+,p^-,q^-,t)=\{\bh^+-\bh^-,k_s\}=\\

4630:  \sum\limits_{

4631:  \g^+\in\Pc^+,\g^-\in\Pc^-} \kappa_{\g^-}\frac{\partial \bh^-(p^-,q^-,t)}

4632:  {\partial p^-_{\g^-}}

4633:  \frac {\partial k_s( p^+,q^-,t)}{\partial  q^-_{\g^-}}+\\

4634:   \nonumber \kappa_{\g^+}\frac{\partial k_s( p^+,q^-,t)}{

4635:   \partial  p^+_{\g^+}}\frac {\partial \bh^+(p^+,q^+,t)}

4636:  {\partial q^+_{\g^+}} \,.

4637:  \end{eqnarray}

4638:   Here $\kappa_\g$ denotes, as usual,  the multiplicity of

4639:   $\g$.

4640: 

4641: 

4642: We can view  the correspondence

4643:   \begin{equation*}

4644:   f ( p^+,q^-,t)\mapsto f_s( p^+,q^-,t),

4645:   \end{equation*}

4646:   where $f_s$ is the solution of

4647:   the above equation (\ref{eq:HJ}) with the initial data

4648:   $f_0=f$,

4649:   as a non-linear operator $ S^s:\fL\to\fL$.

4650:   Let us denote by

4651:   $T^s_f $ the linearization of $ S^s$ at a point

4652:   $f$.

4653:   The next proposition is a rational version of Proposition

4654:   \ref{prop:SFT-chain}. It can be either deduced from

4655: \ref{prop:SFT-chain}, or similarly  proven by    differentiation

4656: with respect to the parameter $s$.

4657:   Denote by $\cS$ the subspace of $\fL$ which consists of solutions of

4658:     the Hamilton-Jacobi equation

4659: (\ref{eq:HJ1}), i.e.

4660:  \begin{equation*}

4661:  \widehat\bh|_{L_f}=0,\quad\hbox{  where}\quad

4662:   L_f=\{p^-_\g=\kappa_\g\frac{\partial

4663:  {f}}{\partial q^-_\g}, q^+_\g=\kappa_\g\frac{\partial

4664: {  f }}{\partial p^+_\g}\}.

4665: \end{equation*}

4666: 

4667: 

4668: \begin{proposition}\label{prop:RSFT-chain}

4669:   \begin{description}

4670:   \item{1.} The subspace  $\cS\subset\fL$ is invariant

4671:     under the flow $S^s$.

4672:   \item{2.}   For $f\in \cS$

4673:   the isomorphism  $T^s_f:\fL\to\fL$

4674:  satisfies the equation

4675: \begin{equation}

4676:  T^s_f\circ d^{f }=d^{S^s(f)}\circ T^s,

4677:  \end{equation}

4678:  and thus defines an isomorphism

4679: $H_*(\fL,d^{f})\to

4680: H^*(\fL,d^{S^s(f)} )$.

4681:   \item{3.}  For $f\in\cS$ the diagram

4682:    \bigskip\bigskip

4683: 

4684: \centerline{$ \mathop{\fL\qquad}\limits_{\;

4685:  \; \nwarrow \, \left(f\right)^\pm}

4686: \mathop{\longrightarrow}\limits^{T^s_f}

4687: \mathop{\fL}\limits_{\nearrow \, \left(S^s(f)\right)^\pm}$}

4688: 

4689: \centerline{$\fP^\pm\;$}

4690: 

4691: \bigskip

4692: 

4693:   homotopically commutes.

4694:   \end{description}

4695: \end{proposition}

4696: 

4697:  Theorem     \ref{thm:SFT-chain} reduces on the level of

4698:  rational curves

4699:  to the following

4700: 

4701: \begin{theorem}\label{thm:RSFT-chain}

4702: Let $\bF^0$, $\bF^1$ and $\bK$ be as in Theorem

4703: \ref{thm:SFT-chain}. Set $\bff^0=\bF^0_0,\;\bff^1=\bF^1_0,\;

4704: \bk=\bK_0.$ Then

4705: $\bff_0$ and $\bff_1$ are homotopic, i.e. they can be included into a family

4706: $\bff_s,\,s\in [0,1]$, such that the Hamilton-Jacobi equation

4707: (\ref{eq:HJ}) holds with $f_s=\bff_s$ and  $k_s\equiv\bk$.

4708: \end{theorem}

4709: 

4710: Hence, Proposition \ref{prop:RSFT-chain} implies

4711: \begin{corollary}\label{cor:RSFT-chain}

4712:  For a generic compatible deformation $J_s$,

4713: $s\in[0,1]$, we have

4714: \begin{description}

4715: \item{1.}

4716:  The operator  $S^1:\fL\to\fL$ defines an automorphism of

4717:  the space  of solutions of (\ref{eq:HJ1});

4718:     \item{2.} The linearization $T^1 =T^1_{\bff^0}$ of $S^1$ at the point

4719:     $\bff_0$ satisfies the

4720:   equation

4721: \begin{equation}

4722:  T^1\circ d^{\bff^0}=d^{\bff^1}\circ T^1,

4723:  \end{equation}

4724:  and thus defines an isomorphism

4725: $H_*(\fL,d^{\bff^0})\to

4726: H^*(\fL,d^{\bff^1})$.

4727:   \item{2.} The diagram

4728:    \bigskip\bigskip

4729: 

4730: \centerline{$ \mathop{\fL\qquad}\limits_{\;

4731:  \; \nwarrow \, \left(\bff^0\right)^\pm}

4732: \mathop{\longrightarrow}\limits^{T^1}

4733: \mathop{\fL}\limits_{\nearrow \, \left(\bff^1\right)^\pm}$}

4734: 

4735: \centerline{$\fP^\pm$}

4736: 

4737: \bigskip

4738: 

4739:   homotopically commutes.

4740:   \end{description}

4741: \end{corollary}

4742: 

4743: 

4744: To formulate the ``classical level'' corollary of Theorem

4745: \ref{thm:RSFT-chain}

4746: we assume, as usual, that $W$ is a homology cobordism.

4747: \begin{theorem}\label{thm:classical-chain}

4748: The homomorphisms  $\Psi^1_{J_0} ,\Psi^1_{J_1}: \fA^+\to\fA^-$ associated

4749: to two compatible almost complex structures $J_0$ and $J_1$ are

4750: homotopic, i.e. there exists a map $\Delta:\fA^+\to\fA^-$ such

4751: that

4752: $$\Psi^1_{J_1} -\Psi^1_{J_1} =\partial^-\circ\Delta+\Delta\circ\partial^+.$$

4753:       \end{theorem}

4754: The map $\Delta$ can be

4755: expressed through the function

4756:   $\bk\in\fL$. However, unlike the case of usual Floer homology

4757:   theory, $\Delta$ and $\bk$ are related via a first order non-linear PDE

4758:   (which can be deduced from the equation (\ref{eq:HJ})),

4759:      and thus  one cannot write a general explicit formula  relating

4760:      $\Delta$ and $\bk$.

4761: 

4762: 

4763:      \subsection{Composition of cobordisms}\label{sec:composition}

4764: In this section we   study the behavior of

4765: potentials and associated algebraic structures under

4766: the operation of composition of directed symplectic cobordisms.

4767: 

4768:     Let    us recall (see Section \ref{sec:splitting}) that   given

4769:     a dividing contact type hypersurface $V$ in a

4770:     directed symplectic cobordism $W=\ora{V^-V^+}$ one can split $W$

4771:     into a composition $W=W_-\circledcirc W_+$ of cobordisms $W_-=\ora{V^-V}$ and

4772:     $W_+=\ora{VV^+}$. From the point of view of an almost

4773:      complex structure        the process of splitting consists of deforming

4774:      an original almost complex structure $J=J^0$ to an

4775:      almost complex structure $J^\infty$, such that the restrictions

4776:      $J_\pm=J^\infty|_{W_\pm}$   are compatible  with the structure

4777:      of (completed) directed symplectic cobordisms $W_\pm$.

4778: 

4779: Conversely,  directed symplectic cobordisms

4780:  $W_-=\ora{V^-V}$ and

4781:     $W_+=\ora{VV^+}$ with matching data

4782:     on the common boundary    can be glued into

4783:     a cobordism         $W=\ora{V^-V^+}$ in the following sense:

4784:     there exists a family  $J^\tau$ of almost complex structures

4785:     on $W$ which in the limit  splits $W$ into the

4786:        composition of cobordisms $W_-=\ora{V^-V}$ and

4787:     $W_+=\ora{VV^+}$.

4788: 

4789: 

4790: 

4791:     \bigskip

4792: %%%%CONT

4793: In order to write the formula relating   the potentials

4794: of $W$ and $W_\pm$ we  first need to make more explicit the

4795: relation  between $2$-dimensional homology classes realized  by

4796: holomorphic curves in $W_\pm$ and $W$. We will keep

4797:  assuming that there are no torsion

4798: elements in $H_1$.

4799: Let us recall (see Sections \ref{sec:dynamics} and

4800: \ref{sec:holomorphic} above) that  we have chosen curves ${C}^i_-\subset

4801: W_-,i=1,\dots, m_-,\;\;{C}^j_+\subset W_+,j=1,\dots,m_+,$

4802: and ${C}^k

4803: \subset W,k=1,\dots m,$ which represent bases of first homology of

4804: the respective cobordisms. We also have chosen

4805: for every periodic orbit $\g\in

4806: \Pc_\a$ of the Reeb field $R_\a$ on $V$ a surface

4807: $F^\g_\pm$ realizing homology in $W_\pm$ between $\g$ and a linear

4808: combination of basic curves ${C}^i_\pm$.

4809: For our current purposes we have to make one extra choice: for

4810: each curve $ {C}^i_\pm$ we choose a surface $S^i_\pm$ which

4811: realizes homology in $W$ between ${C}^i_\pm$ and the

4812: corresponding linear combination of the curves ${C}^1,\dots,

4813: {C}^m$.

4814: All the choices enable us to associate with every orbit

4815: $\g\in\Pc_\a$ a homology class $C^\g$ which is realized by the chain

4816:  $$F^\g_+-F^\g_-+\sum\limits_1^{m_+} n^+_jS^j_+-\sum\limits_1^{m_-}

4817: n^-_iS^i_-\,,$$

4818: where $\partial F^\g_\pm=[\gamma]-\sum\limits_1^{m_\pm}

4819: n^{\pm}_j{C}^j_\pm$. We will denote by $d^\g$ the degree of

4820: $C^\g$, i.e. the string of its coordinates in the

4821: chosen basis

4822: $A_1,\dots,A_N\in H_2(W)$.

4823: 

4824: 

4825: 

4826:     Let us define an operation

4827:  \begin{equation}

4828:  \star:\fD\fD_-\otimes \fD\fD_+ \to \fD\fD,

4829:  \end{equation}

4830:   where   $\fD\fD_\pm=\fD\fD_{W_\pm}$ and

4831: $\fD\fD= \fD\fD_{W} $.

4832:   For

4833:    $F=\sum\limits_{\Gamma} f_{\Gamma}(t^-,q^-,\hbar,z^-)

4834:    p^{\Gamma}\in\fD\fD_-$ and

4835:     $

4836:   G=\sum\limits_{\Gamma^+} g_{\Gamma^+}

4837:  (t^+,q ,\hbar,z^+)(p^+)^{\Gamma^+} \in\fD\fD_+$ we set

4838:   \begin{equation}

4839:  \begin{split}

4840:   &F  {\star} G(t,q^-,p^+,\hbar,z)=\\

4841:   &\Big(

4842:  \sum\limits_{\Gamma }\wt f_\Gamma(t,q^-,\hbar,z)

4843:  \hbar^s  \prod\limits_{i=1}^s

4844:  \kappa_{\g_{i }}z^{d^{\g_i}}

4845:  \overrightarrow{\frac{\partial}{\partial

4846:  q_{\g_i}}}

4847: \sum\limits_{\Gamma^+} \wt g_{\Gamma^+}

4848: (t,q ,\hbar,z)(p^+)^{\Gamma^+}\Big)\Big|_{q=0}\,. \\

4849: \end{split}

4850: \end{equation}

4851: Here we denote by $\wt f_\Gamma$ and $\wt

4852: g_{\Gamma^+}$ the images

4853: of $ f_\Gamma$ and $ g_{\Gamma^+}$ under the coefficient

4854: homomorphisms $H_2(W_\pm)\to H_2(W)$.

4855: Let us explain what happens with the

4856:  variables $t$ and $z$ in more details.  Let

4857:   $A_1^\pm,\dots, A^\pm_{N^\pm},$

4858:   and

4859:    $A_1,\dots, A_{N},$

4860:     be   the chosen bases in     $H_2(W_\pm)$ and $ H_2(W)$.

4861:                    Then we have $$i^\pm_*(A^\pm_k)=\sum\limits_{j=1}^{N^\pm}n^\pm_{kj} A_j,$$

4862:    where $k=1,\dots, N$,  $(n^\pm_{kj})$ are integer matrices,

4863:    and $i^\pm:W_\pm\to W$ the inclusion maps.

4864: 

4865: We have

4866:  \begin{equation*}\begin{split}

4867:   & f_{\Gamma}(t^-,q^-,\hbar,z^-)

4868:    = \sum\limits_{d=(d_1,\dots, d_{N^-})}f_{\Gamma,d}(t^-,q^-,\hbar)(z_1^-)^{d_1}\dots

4869:    (z_{N^-}^-)^{d_{N^-}},\\

4870:   &g_{\Gamma^+}

4871:  (t^+,q ,\hbar,z^+) = \sum\limits_{d=(d_1,\dots, d_{N^+})}g_{\Gamma^+,d}(t^+,q,\hbar)(z_1^+)^{d_1}\dots

4872:    (z_{N^+}^+)^{d_{N^+}}  ,\\

4873:  \end{split}

4874:  \end{equation*} where

4875:     we denote by $z_\pm$ the ``$z$-variables" in $W_\pm$.

4876:     Then

4877:     \begin{equation*}

4878:     \begin{split}

4879:    & \wt f_\Gamma(t,q^-,\hbar,z)=

4880:           \sum\limits_{d=(d_1,\dots,d_{N_-})} \wt f_{\Gamma,d}(t|_{W_-},q^-,\hbar)z_1^{M_1^-}

4881:          \dots z_N^{M^-_{N}}, \\

4882:      & \wt g_{\Gamma^+}(t,q,\hbar,z)=

4883:           \sum\limits_{d=(d_1,\dots,d_{N_+})} \wt g_{\Gamma^+,d}(t|_{W_+},q^-,\hbar)z_1^{M_1^+}

4884:          \dots z_N^{M^+_{N}}, \\

4885:           \end{split}

4886:           \end{equation*}

4887:         where $M^\pm_j=   \sum\limits_{k=1}^{N^\pm}  n^\pm_{kj}d_k$, $j=1,\dots,N$.

4888: 

4889:    \bigskip

4890:   Let us  observe

4891:   \begin{lemma}\label{lm:star}

4892:   \begin{description}

4893:   \item{1.} The operation $\star$ is associative.

4894:   \item{2.}   For $F\in\hbar^{-1 }\fD_-,G\in\hbar^{-1 }\fD_+$ there exists a unique function $H\in \hbar^{-1 }\fD$,

4895:   such that

4896:  $$

4897:   e^H=e^F\star e^G.

4898:  $$

4899:  \end{description}

4900:   \end{lemma}

4901:   We will denote this $H$ by $F\lozenge G$, so that we have

4902:   $e^{F\lozenge G}=e^F\star e^G$.

4903: We will also consider the maps

4904: $$\lozenge G:\hbar^{-1}\fD_-\to \hbar^{-1}\fD,\;\;\lozenge

4905: G(F)=F\lozenge G,\,F\in\fD_-,$$

4906: and

4907: $$F\lozenge:\hbar^{-1}\fD_+\to \hbar^{-1}\fD,\;\;F\lozenge

4908:  (G)=F\lozenge G,\,G\in\fD_+,$$

4909:  and for even $F,G$ their linearizations:

4910:               \begin{equation*}

4911:  \begin{split}

4912: & T_F(\lozenge G):\hbar^{-1}\fD_-\to \hbar^{-1}\fD,\\

4913:  &T_F(\lozenge G)(f)=\frac{d\big(F+\varepsilon

4914:  f)\lozenge G\big)}{d\varepsilon}\big|_{\varepsilon=0}=e^{-F\lozenge

4915:  G}\big((fe^F)\star e^G\big),\;\; f\in\hbar^{-1} \fD_-,\\

4916:  \end{split}

4917:  \end{equation*}

4918: and

4919: \begin{equation*}

4920:  \begin{split}

4921: & T_G(F\lozenge  ):\hbar^{-1}\fD_+\to \hbar^{-1}\fD,\\

4922:  &T_G(F\lozenge  )(g)=\frac{d\big(F \lozenge (G+\varepsilon

4923:  g)\big)}{d\varepsilon}\big|_{\varepsilon=0}=e^{-F\lozenge

4924:  G}\big( e^F \star (ge^G)\big),\;\; g\in\hbar^{-1} \fD_+,\\

4925:  \end{split}

4926:  \end{equation*}

4927: 

4928: Let us first formulate an algebraic

4929: \begin{proposition}\label{prop:SFT-composition}

4930: Suppose that $F\in \hbar^{-1}\fD_-$ and $G\in \hbar^{-1}\fD_+$ are

4931: even elements, which

4932: satisfy the equations

4933: $$D_{W_-}(e^F)=\overrightarrow{\bH^-}e^F-e^F\overleftarrow{\bH}=0$$  and $$D_{W_+}(e^G)=

4934: \overrightarrow{\bH}e^F-e^F\overleftarrow{\bH^+}=0,$$

4935: where $\bH^\pm=\bH_{V^\pm},\,\bH=\bH_V$.

4936: Then we have

4937: \begin{description}

4938: \item{1.} $D_W(e^{F\lozenge G})=\ora{\bH^-}\, e^{F\lozenge G}-e^{F\lozenge

4939: G}\,\ola{\bH^+}=0$.

4940: \item{2.} The homomorphisms

4941: $ T_G(F\lozenge  ):\hbar^{-1}\fD_+\to \hbar^{-1}\fD$ and

4942: $T_F(\lozenge G):\hbar^{-1}\fD_-\to \hbar^{-1}\fD$ satisfy

4943: the equations $$T_G(F\lozenge  )\circ

4944: D^{G}=D^{F\lozenge G}\circ T_G(F\lozenge )$$ and $$T_F(

4945: \lozenge G )\circ D^{F}=D^{F\lozenge G}\circ T_F(

4946: \lozenge G),$$ and in particular they define homomorphisms

4947: of the corresponding homology algebras:

4948: $$ \left( T_G(F\lozenge  )\right)_*:H_*( \fD_+,D^G)\to H_*(  \fD,D^

4949:  {F\lozenge G})$$ and

4950: $$ \left( T_F( \lozenge G  )\right)_*:H_*( \fD_-,D^G)\to H_*(  \fD,D^

4951:  {F\lozenge G}).$$

4952: \item{3.} $$ T_F( \lozenge G)\circ F^-   =

4953: (F \lozenge G)^-$$ and $$  T_G( F\lozenge)

4954: \circ G^+= (F \lozenge G)^+.$$

4955: \item{4.}

4956: Suppose we are given three cobordisms $W_1,W_2,W_3$ with matching ends

4957:  so that we can  form  the  composition

4958: $W_{123}=W_1\circledcirc W_2 \circledcirc W_3$, and series

4959: $F_i\in\hbar^{-1}\fD_i=\hbar^{-1}\fD_{W_i},\,i=1,2,3$, such that

4960: $D_{W_i}e^{F_i}=0$, $i=1,2,3$.

4961: Then

4962: \begin{equation}

4963: T_{F_1\lozenge F_2}(\lozenge F_3)\circ T_{F_1}(\lozenge

4964: F_2)=T_{F_1}(\lozenge (F_2\lozenge F_3)).

4965: \end{equation}

4966: \end{description}

4967: \end{proposition}

4968: The proof of this proposition is immediate from the definition of

4969: the corresponding operations. For instance, to prove

4970: \ref{prop:SFT-composition}.1 we write

4971: \begin{equation*}

4972: \begin{split}

4973:  \ora{\bH^-}\,\big(e^{F\lozenge G}\big)-\big(e^{F\lozenge

4974: G}\big)\,\ola{\bH^+}&=\ora{\bH^-}\,e^F\star e^G -e^F\star e^G\ola

4975: {\bH^+}\\

4976: &=\left(\ora{\bH^-}\,e^F\right)\star e^G -e^F\star\left( e^G\ola

4977: {\bH^+}\right)\\

4978: &=\left(e^F\,\ola{\bH}\right)\star e^G

4979: -e^F\star\left( \ora{\bH}e^G\right).\\

4980: \end{split}

4981: \end{equation*}

4982: To finish the argument

4983: let us consider a cylindrical cobordism $W_0=V\times\R,$

4984: take the function

4985: $I=\sum \kappa_\g^{-1}p_\g q_\g$. Taking into account associativity

4986: of $\star$ (see \ref{lm:star}) we have

4987: \begin{equation}

4988: \ora {f}e^G=(fe^I)\star e^G,\quad\hbox{and}\quad

4989:  e^F\ola{f}=e^F\star (fe^I)\,.

4990:  \end{equation}

4991:  Hence, we have

4992:  \begin{equation*}

4993:  \begin{split}

4994:  & \left(e^F\,\ola{\bH}\right)\star e^G

4995: -e^F\star\left( \ora{\bH}e^G\right)\\

4996: &=e^F\star(\bH e^I)\star e^G-  e^F\star(\bH e^I)\star e^G=0.\\

4997:  \end{split}

4998: \end{equation*}

4999: \bigskip

5000: 

5001: Any cohomology class from $H^*(W)$

5002: can be represented by a form $t$ which splits into the sum of forms

5003: $t_\pm$  with cylindrical ends on $W_\pm$,

5004:     so that we have $t_\pm|_{V}=t_V$.

5005:      Let us define now a series

5006:       $\bF^\infty\in\hbar^{-1}\fD$

5007:       by the formula

5008:     \begin{equation}\label{eq:SFT-gluing}

5009:  {\bF^\infty(q^-,p^+,t)}=  {\bF_-(q^-,p,t_-)}

5010:  \lozenge

5011:    {\bF_+(q,p^+,t_+)} ,

5012:    \end{equation}

5013:     where $p,q$      are variable associated to the space $H^*(\Pc)$

5014:     of   periodic orbits of the Reeb vector field

5015:     $R_\a$ of the contact form $\alpha$ on $V$.

5016: 

5017:      The following theorem is the main claim of this section, and

5018:      similar to Theorems \ref{thm:HH} and \ref{thm:SFT-cobordism}

5019:      and \ref{thm:SFT-chain} it  is a statement about the

5020:      boundary of an appropriate moduli space of holomorphic curves.

5021:      This time we deal with limits of $J^s$-holomorphic curves

5022:      in $W$ when $s\to\infty$,

5023:      i.e. when the family $J^s$ realizes the splitting

5024:      of $W$ into the composition $W_-\circledcirc W_+$,

5025:       see Theorem \ref{thm:comp2} above.

5026: 

5027: 

5028:     \begin{theorem}\label{thm:SFT-composition}

5029: The element $\bF^\infty$ is homotopic to the

5030:    potential

5031:    $\bF=\bF_{W,J ,\a^\pm} $

5032:    for any generic  compatible almost complex structure $J$ on $W$.

5033:    \end{theorem}

5034:   \bigskip

5035:   Let us now describe the above results on the level of rational potentials.

5036: Let  $W_-=\ora{V^-V }$, $W_+=\ora{VV^+}$ and

5037:  $W=W_-\circledcirc W_+=\ora{V^-V^+}$ be as above. Set

5038: $\bh^\pm=\bh_{V^\pm},\bh=\bh_V$,

5039: $\widehat\bh_-=\bh^- -\bh$, $\widehat\bh_+=\bh  -\bh^+$,

5040: $\widehat\bh=\bh^+-\bh^-$,

5041: $\fL_\pm=\fL_{W_{\pm}}$ and $ \fL=\fL_W$.

5042: 

5043:  The operation

5044: $\lozenge:\hbar^{-1}\fD_-\times\hbar^{-1}\fD_+\to\hbar^{-1}\fD$ defined above reduces on the

5045: rational level to the

5046:    operation $$\sharp:\fL_-\times\fL_+\to\fL,$$ defined as follows.

5047:    For $ f_\pm\in\fL_\pm$ we set

5048: \begin{equation}

5049: f_-\sharp f_+(q^-,p^+)=\big(f_-(q^-,p)+f_+(q,p^+)

5050: -\sum_{\g\in\Pc}\kappa_\g^{-1}z^{-d_{\g}}q_\g p_\g\big)\big|_{L},

5051: \end{equation}

5052: where

5053: \begin{equation*}

5054: L=

5055: \begin{cases}

5056: q_\g=&\kappa_\g z^{d_\g}\frac{\partial f_-}{\partial p_\g};\\

5057: p_\g=&\kappa_\g z^{d_\g}\frac{\partial f_+}{\partial q_\g}.\\

5058: \end{cases}

5059: \end{equation*}

5060: Notice that given series $$F_-=\hbar^{-1}\sum\limits_0^\infty

5061: (F_-)_g\hbar^g\in\hbar^{-1}\fD^-\quad\hbox{ and}\quad

5062:  F_+=\hbar^{-1}\sum\limits_0^\infty (F_+)_g\hbar^g\in\hbar^{-1}\fD^+$$  with

5063: $F_-\lozenge F_+= \hbar^{-1}\sum\limits_0^\infty (F_-\lozenge

5064: F_+)_g\hbar^g\in\hbar^{-1}\fD $ then  $$(F_-\lozenge F_+)_0=(F_-)_0\sharp

5065: (F_+)_0.$$

5066:  We will also consider the operations

5067: \begin{equation*}

5068: \begin{split}

5069: &\sharp f_+:\fL_-\to\fL,\;\; \sharp f_+(f_-)=f_-\sharp

5070: f_+\;\;\hbox{and}\;\;

5071: \sharp f_-:\fL_+\to\fL,\;\; \sharp f_-(f_+)=f_-\sharp

5072: f_+,\\

5073: \end{split}

5074: \end{equation*}

5075: and their linearizations

5076: \begin{equation*}

5077: \begin{split}

5078: &T_{f_-}(\sharp f_+):\fL_-\to\fL,\;\; T_{f_-}(\sharp f_+)(g)=

5079: (g|_{L_+})\sharp f_+

5080: \quad\hbox{and}\\

5081: &T_{f_+}(\sharp f_-):\fL_+\to\fL,\;\; T_{f_+}(\sharp f_-)(g)=f_-\sharp

5082: (g|_{L_-})\,.\\

5083: \end{split}

5084: \end{equation*}

5085: Here we view $g|_{L_+}$ (resp. $g|_{L_-}$) as an element of

5086: $\fL_+$  which depends on variables $q^-$ as parameters (resp. an element of

5087: $\fL_-$, which depends on $p^+$ as parameters).

5088: %\end{proposition}

5089:   We have the following rational version

5090: of Theorem \ref{prop:SFT-composition}.

5091: \begin{proposition}\label{prop:RSFT-composition}

5092: Suppose that even elements $f_\pm\in\fL_\pm=\fL_{W_{\pm}}$ satisfy

5093: equation (\ref{eq:HJ1}), i.e.

5094: \begin{equation*}

5095: \widehat\bh_\pm|_{L_{f_\pm}}= 0 ,

5096: \end{equation*}

5097: where

5098: \begin{equation*}

5099: L_{f_-}=

5100: \begin{cases}

5101: q_\g&=k_\g\frac{\partial f_-}{\partial p_\g};\;\;\g\in\Pc\\

5102: p^-_{\g^-}& =\kappa_{\g^-}\frac{\partial f_-}{\partial q^-_{\g^-}},\;\;\g^-\in\Pc^-,\\

5103: \end{cases}

5104: \end{equation*}

5105: and

5106: \begin{equation*}

5107: L_{f_+}=

5108: \begin{cases}

5109: p_\g&=k_\g\frac{\partial f_+}{\partial q_\g};\;\;\g\in\Pc\\

5110: q^+_{\g^+}&=\kappa_{\g^+}\frac{\partial f_+}{\partial p^+_{\g^+}},\;\;\g^+\in\Pc^+.\\

5111: \end{cases}

5112: \end{equation*}

5113: Then

5114: \begin{description}

5115: \item{1.} The function $f_-\sharp f_+$ satisfies the Hamilton-Jacobi

5116: equation

5117: $$\widehat\bh_{L|_{f_-\sharp f_+}}=0;$$

5118: \item{2.} The homomorphisms $T_{f_-}(\sharp f_+):\fL_-\to\fL$ and

5119: $T_{f_+}( f_-\sharp):\fL_+\to\fL$ satisfy the equations

5120: \begin{equation*}

5121: \begin{split}

5122: T_{f_-}(\sharp f_+)\circ d^{f_-}=&d^{f_-\sharp f_+}\circ T_{f_-}(\sharp

5123: f_+),\\

5124: T_{f_+}( f_-\sharp)\circ d^{f_+}=&d^{f_-\sharp f_+}\circ T_{f_+}(\sharp

5125: f_-)\,, \\

5126: \end{split}

5127: \end{equation*}

5128: and hence define homomorphisms of the corresponding

5129: homology algebras:

5130: \begin{equation*}

5131: \begin{split}

5132: \big(T_{f_-}(\sharp f_+)\big)_*:&H_*(\fL_-,d^{f_-})\to H_*(\fL,d^{f_-\sharp f_+}),\\

5133: \big(T_{f_+}(f_-\sharp )\big)_*:&H_*(\fL_-,d^{f_-})\to H_*(\fL,d^{f_-\sharp f_+});\\

5134: \end{split}

5135: \end{equation*}

5136: \item{3.}

5137: \begin{equation*}

5138: \begin{split}

5139: T_{f_-}(\sharp f_+)\circ (f_-)^-=& (

5140: f_-\sharp f_+)^-\,, \\

5141: T_{f_+}(\sharp f_-)\circ (f_+)^-=& (

5142: f_-\sharp f_+)^+

5143: \,; \\

5144: \end{split}

5145: \end{equation*}

5146: \item{4.}

5147: Suppose we are given three cobordisms $W_1,W_2,W_3$ with matching ends

5148:  so that we can  form  the  composition

5149: $W_{123}=W_1\circledcirc W_2 \circledcirc W_3$, and series

5150: $f_i\in\hbar^{-1}\fL_i= \fL_{W_i}$  which satisfy

5151: Hamilton-Jacobi equations

5152: $\widehat{\bh}_{W_i}|_{L_{f_i}}=0,\,i=1,2,3$.

5153: Then

5154: \begin{equation*}

5155: T_{f_2}(\sharp f_3)\circ T_{f_1}(\sharp

5156: f_2)=T_{f_1}(\sharp (f_2\sharp f_3)).

5157: \end{equation*}

5158: \end{description}

5159: \end{proposition}

5160: 

5161: Let $t$ be a closed form on $W$ which is split into two forms $t_\pm$ on $W_\pm$ with

5162: cylindrical ends. Set $\bff_\pm=\bff_{W_\pm}$ and

5163: \begin{equation}

5164: \bff^\infty(q^-,p^+,t)=\bff_-(q^-,p,t_-)\sharp \bff_+(q,p^+,t_+).

5165: \end{equation}

5166: Alternatively $\bff^\infty$ can be defined as

5167: the first term  in the expansion

5168: $\bF^\infty=\hbar^{-1}\sum\limits_0^\infty\bF^\infty_g\hbar^g$.

5169:  The following theorem is a rational analog, and a direct

5170:  corollary of Theorem \ref{thm:SFT-composition}.

5171: \begin{theorem}

5172: \label{thm:RSFT-composition}

5173:   The series $\bff^\infty(q^-,p^+,t)$  and  $\bff_{W,J}(q,p,t)$

5174:   are  homotopic  for any generic  compatible almost complex structure

5175:   $J$ on $W$.

5176: \end{theorem}

5177: 

5178: Coming down to the   ``classical" level, let us assume that

5179:  $W,W_-$ and $W_+$ are homology cobordisms (see \ref{sec:3algebras}

5180:  above). Thus there are   defined  the homomorphisms  $\bPsi  :\fA^+\to\fA^-$,

5181:  $\bPsi_+ :\fA^+\to\fA $ and $\bPsi_- :\fA \to\fA^-$,

5182:     see Section \ref{sec:relative} above. Set

5183:  $\bPsi_\infty=\bPsi_+\circ\bPsi_-$.

5184:  Then we have

5185:  \begin{theorem}\label{thm:classical-composition}

5186:  For any generic compatible almost complex structure $J$ on $W$

5187:  homomorphisms

5188:  $\bPsi_1=\bPsi_J,\bPsi_\infty:\fA^+\to\fA^-$ are chain

5189:  homotopic.

5190: \end{theorem}

5191: 

5192: \subsection{Invariants of contact

5193: manifolds}\label{sec:contact-invariants}

5194: 

5195: 

5196: 

5197:  Theorem \ref{thm:SFT-composition}  allows us to define

5198:  SFT-invariants of contact manifolds.

5199:  Let $(V,\xi)$ be a contact manifold,  and

5200:  $\alpha^+$ and $\alpha^-$

5201:  two contact forms for $\xi$,

5202:  such that $\alpha^+>\alpha^-$, i.e.

5203:  $\alpha^+=f\alpha^-$, for a  function $f>1$.

5204:    Then for an appropriately chosen function $\zeta:V\times\R\to(0,\infty)$

5205:    the form

5206:  $\omega=d(\zeta\alpha^-)$ on $W=V\times\R$ is symplectic, and

5207:  $(W,\omega)$ is a directed symplectic cobordism between $(V,\alpha^-)$

5208:  and $(V,\alpha^+)$.  Let  $t^\pm$ be two cohomologous forms on $V$, and $t$ be a  closed

5209:  form  on $W$ with cylindrical ends which restricts to $t^\pm$ on $V^\pm$.

5210:  Suppose we are also given almost complex structures $J^\pm$ on $V$,

5211:   compatible with $\alpha^\pm$, which

5212:  are extended to  a compatible almost complex structure $J$ on $W$.

5213:  We will call a  directed symplectic cobordism  $(W,J,t)$,

5214:   chosen in the  above way, a {\it concordance} between

5215:   the data on its boundary. Notice, that concordance becomes   an

5216:   equivalence relation if we identify contact forms proportional

5217:   with a constant factor. A concordance $(W,J,t)$ is called {\it

5218:   trivial} if  $W=V\times\R$, the almost complex structure

5219:    $J$ is translationally  invariant, and

5220:   $t$ is the pull-back of a form $t_+$ under the projection $W\to

5221:   V$.

5222: 

5223: 

5224:  Let us denote, as usual,   by $(\fW^\pm,D^\pm)$  the differential Weyl algebras

5225:   associated to the data at the

5226:  ends of the cobordism $W$, by $(\fD,D_W)$ the $D$-module

5227:  $\fD(W,J,\a^\pm)$,  by $\bF\in\fW$

5228:  the    potential of the cobordism $W$, and by

5229:  $\bF^\pm:(\fW^\pm,D^\pm)\to(\fD,D_W)$ the

5230:  corresponding homomorphisms

5231:  of differential algebras defined in (\ref{eq:end-homomorphisms}).

5232: 

5233:  \begin{theorem}\label{thm:SFT-invariance}

5234:   For any concordance $(W,J,t)$ the  homomorphisms $$\bF^\pm:(\fW^\pm,d^\pm)\to(\fD,D_W)$$

5235:     are quasi-isomorphisms

5236:   of differential algebras. In particular, the homology algebras

5237:   $H_*(\fW^-,D^-)$ and

5238:   $H_*(\fW^+,D^+)$ are isomorphic.

5239:   \end{theorem}

5240: 

5241:  \noindent\proof We will prove   \ref{thm:SFT-invariance}  in three

5242:   steps.

5243: 

5244: \noindent Step 1. Let us begin with the trivial concordance $(W ,J

5245: ,t  )$. In this case $\fD$ can be identified with $\fW^\pm$ and we

5246: have $\bF(q,p,t)=\hbar^{-1}\sum\kappa_\g^{-1}q_\g p_\g$. Hence

5247: \begin{equation*}

5248: \bF^-(f)=e^{-\bF}\ora{f}e^{\bF}=

5249: f=e^{-\bF}\left(

5250: e^{\bF}\ola{f}\right)=\bF^+(f).

5251: \end{equation*}

5252: 

5253:          \noindent Step 2. If we add now to $t$ a form $d\theta$, where

5254: $\theta$ has a compact support, and change $J$ in a compact part

5255: of $W$ then according to Theorem \ref{thm:SFT-chain}

5256:  the potential $\bF_{W,J}(t+d\theta)$ remains the same

5257:  up to homotopy, and hence Corollary \ref{cor:SFT-chain} implies

5258: that the homomorphisms  $\bF^\pm_*$   induced on homology remain

5259:  unchanged.

5260: 

5261: 

5262: 

5263: \noindent Step 3. Now assume that $(W,J,t)=(W^1,J^1,t^1)$ is a general

5264: concordance. Consider the reversed concordance $(W^2,J^2,t^2)$, so

5265: that the compositions $$(W^{12}=W^1\circledcirc W^2,J^{12}=J^1\circledcirc

5266: J^2,t^{12}=

5267: t_1\circledcirc t_2)$$ and

5268:   $$(W^{21}=W^2\circledcirc W^1,J^{21}=J^2\circledcirc J^1,t^{21}=

5269: t_2\circledcirc t_1)$$ of concordances $(W^1 ,J^1 ,t^1 )$

5270: and $(W^2 ,J^2 ,t^2 )$   are   as in Step 2.

5271:  Then according to Theorem

5272: \ref{thm:SFT-composition}

5273: $\bF_{W^{12},J^{12}}( t^{12})$ is homotopic to

5274: $$\bF_{W^1,J^1}(t^1)\lozenge\bF_{W^2,J^2}(t^2)=

5275: \bF^1\lozenge\bF^2$$ and

5276:  $\bF^{21}=\bF_{W^{21},J^{21}}( t^{21})$ is homotopic to

5277: $$\bF_{W^2,J^2}(t^2)\lozenge\bF_{W^1,J^1}(t^1)=\bF^2\lozenge\bF^1.$$ Hence

5278: Proposition \ref{prop:SFT-composition}

5279: implies

5280: \begin{equation*}

5281: \Id=\left(\bF_{W^1\circledcirc W^2 }\right)^-_*

5282: =\left(T_{\bF^1}(\lozenge\bF^2) \right)_*\circ (\bF^1)^-_*

5283:   \end{equation*}

5284:   and

5285:   \begin{equation*}

5286:   \begin{split}

5287: & \Id= \left(T_{\bF^1}\left(\lozenge(\bF^2\lozenge\bF^1) \right)

5288: \right)_* \\

5289: &= \left(T_{\bF^1\lozenge\bF^2}(\lozenge \bF^1)\right)_*

5290: \circ \left(T_{\bF^1} (\lozenge  \bF^2)\right)_* .\\

5291: \end{split}

5292: \end{equation*}

5293: Hence

5294: $\left(T_{\bF^1}(\lozenge\bF^2) \right)_*$,  $\;(\bF^1)^-_*$,

5295: and similarly    $(\bF_1)^+_* $

5296: are isomorphisms. \qed

5297: 

5298: \bigskip

5299: 

5300: 

5301: The following rational and classical analogs of Theorem

5302: \ref{thm:SFT-invariance}

5303: can be either deduced directly from Theorem

5304: \ref{thm:SFT-invariance}, or alternatively can be proven similarly

5305: to \ref{thm:SFT-invariance}  using \ref{thm:RSFT-chain} (resp. \ref{thm:classical-chain}) and

5306: \ref{thm:RSFT-composition} (resp. \ref{thm:classical-composition}).

5307: 

5308: \begin{theorem}\label{thm:RSFT-invariance}

5309:     For any concordance $(W,J,t)$ the  homomorphisms $$\bff^\pm:(\fP^\pm,D^\pm)\to(\fL,D_W)$$

5310:      are quasi-isomorphisms

5311:   of differential algebras. In particular,  Poisson homology

5312:    algebras $H_*(\fP^\pm,d^\pm)$  are isomorphic.

5313:   \end{theorem}

5314: 

5315: 

5316: 

5317: \begin{theorem}\label{cor:classical-invariance}

5318:  For any concordance $(W,J,t)$ the  homomorphism  $$\bPsi:(\fA^+,\partial^+)\to

5319:  (\fA^-,\partial^-)$$

5320:   is a quasi-isomorphism

5321:   of differential algebras.

5322: \end{theorem}

5323: \bigskip

5324: 

5325: The definition of the differential algebras $(\fW,D),\;(\fP,d)$ and $(\fA,\partial)$

5326:  depends on the choice

5327: of a coherent orientation (see Section \ref{sec:orientation}), and

5328:   of spanning surfaces and framings

5329: of periodic orbits

5330: (see Section \ref{sec:dynamics}).

5331: As it is stated in Theorem \ref{thm:coherent}  a coherent orientation

5332: is determined by a choice of asymptotic operators associated

5333: with each periodic orbit $\g$. Let $\bH'$ be the new  Hamiltonian which

5334: one gets by changing the orientation of the asymptotic operator

5335: associated with a fixed  periodic orbit $\g $. One can then check that the

5336: change of variables  $(p_\g, q_\g)\mapsto(-p_\g,-q_\g)$ is an

5337: isomorphism  between the differential algebras $(\fW,D^{\bH})$ and

5338: $(\fW,D^{\bH'})$.

5339:   Different choices for spanning   surfaces and framings of periodic orbits do not affect

5340: $\mod\;2$ grading but change the integer grading of the

5341: differential algebras.

5342: 

5343: \begin{remark}\label{rem:other-choices}

5344: {\rm

5345:   An accurate introduction of  virtual cycle techniques

5346: would reveal that even more choices should be made.   However,  an independence

5347: of  all these extra choices can be also established following the scheme of this section.

5348: }

5349: \end{remark}

5350: 

5351: \subsection{A differential equation for potentials of symplectic cobordisms} \label{sec:master}

5352: 

5353: In this section we describe  differential equations for

5354: the  potentials $\bF_W$ and $\bff_W$ of a  directed symplectic

5355: cobordism with a non-empty boundary. These equations completely

5356: determine  the potentials, and

5357:  in

5358: combination with gluing Theorems  \ref{thm:SFT-composition}  and

5359: \ref{thm:RSFT-composition} they provide in many cases   an effective

5360:  recursive

5361: procedure for computing     potentials $\bF_W$ and $\bff_W$,

5362:  and even  Gromov-Witten invariants of closed symplectic manifolds  $W$

5363: (see some examples in Section \ref{sec:computing} below).

5364: 

5365: \bigskip

5366: Let us assume for simplicity that $W$ has only a positive end

5367: $E^+=V\times(0,\infty)$, and choose a basic system

5368: $\Delta_1,\dots,\Delta_k$, $\Theta_1,\dots,\Theta_m$ of closed

5369: forms so that the   following conditions are satisfied:

5370: \begin{description}

5371: \item{a)} $\Delta_1,\dots,\Delta_k$ form a basis of $H^*(W)$, and the restrictions

5372: $\delta_i=\Delta_i|_V$, $i=1,\dots,l$ for $l\leq k$ form a basis

5373: of    $\im\big(H^*(W)\to H^*(V)\big)$;

5374: \item{b)}  $\Theta_1,\dots,\Theta_m$ are compactly

5375: supported and    represent a basis of

5376: $\Ker\big(H^*_{\comp}(W)\to H^*(W)\big)$,

5377: \item{c)} there exist  forms $\theta_1,\dots,\theta_m$ on $V$ and

5378: a compactly supported $1$-form $\rho$ on $(0,+\infty)$, such that

5379: $\Theta_j=e_*\big(\rho\wedge \pi^*(\theta_j)\big),\;j=1,\dots, m$, where

5380: 

5381: $\pi$ is the projection $E=V\times(0,\infty)\to V$ and $e:E\hookrightarrow W$ is the inclusion.

5382: In  other words, $\Theta_j$ equals $\rho\wedge \pi^*(\theta_j)$

5383: viewed as a form on $W$.

5384: \end{description}

5385: \medskip

5386: 

5387: \begin{theorem}\label{thm:SFT-master}

5388: Let $\bH=\bH_{V,\a,J}$ be the Hamiltonian associated with the

5389: contcat manifold $V$.

5390: Set $$\bH^j(t_1,\dots,t_l,q,p)=\Big(

5391: \frac{\partial \bH}{\partial s_j}(\sum\limits_{i=1}^l

5392: t_i\delta_i+s_j\theta_j,q,p)\Big)\Big|_{s_j=0}

5393: ,\;j=1,\dots m,$$

5394: $$\bF^0(t_1,\dots,t_k, p)=\bF_{W,J}(\sum

5395:  t_i\Delta_i,p),$$

5396: and define

5397:  $\bF(t_1,\dots,t_k,\tau_1,\dots,\tau_m, p)$

5398:  by the formula

5399:  \begin{equation}

5400:  e^{\bF(t_1,\dots,t_k,\tau_1,\dots,\tau_m,

5401:  p)}=e^{\bF^0(t_1,\dots,t_k,p)}

5402:  \overleftarrow\bG(t_1,\dots,t_l,\tau_1,\dots,\tau_m,p),

5403:  \end{equation}  where we denote by

5404:  $\overleftarrow\bG$ the operator obtained from

5405:  \begin{equation}

5406:  %\begin{split}

5407:   \bG(t_1,\dots,t_l,\tau_1,\dots,\tau_m,q,p)

5408:   = e^{\tau_m\bH^m(t_1,\dots,t_l,q,p)}\dots

5409:  e^{\tau_1\bH^1(t_1,\dots,t_l,q,p)}

5410:  %\end{split}

5411:  \end{equation}

5412:  by quantizing   $q_\g=\kappa_\g

5413:  \hbar\overleftarrow{\frac{\partial }{\partial

5414:  p_\g}}$.

5415:  Then  $\bF(t_1,\dots,t_k,\tau_1,\dots,\tau_m, p)$  is

5416:   homotopic to  $$\bF_{W,J}(\sum\limits_{i=1}^k

5417:  t_i\Delta_i+\sum\limits_{r=1}^m\tau_r\Theta_r,p).$$

5418:   \end{theorem}

5419:   \proof

5420:  Set  $$T^j(s)={\bF_{W,J}(\sum

5421:  t_i\Delta_i+\sum\limits_{r=1}^{j-1}\tau_r\Theta_r+s\tau_j\Theta_j,p) }.$$

5422:   We have $$T^j(1)=T^{j+1}(0)\quad\hbox{

5423:  for}\;\;j=1,\dots,m-1 ,     $$

5424:   $$

5425:   T^m(1)= {\bF_{W,J}(\sum\limits_{i=1}^k

5426:  t_i\Delta_i+\sum\limits_{r=1}^m\tau_r\Theta_r,p)},$$

5427:  and

5428:  $$T^1(0)={\bF_{W,J}(\sum

5429:  t_i\Delta_i,p) }={\bF^0(t_1,\dots,t_k, p)}.$$

5430:  Let $S^j\in\hbar^{-1}\fD$ be defined from the equation

5431:  $$e^{S^j}=e^{T^j(0)}e^{\tau_j\ola{\bH^j(t_1,\dots,t_l,q,p)} }.$$

5432:  It is enough to prove that

5433:  $  T^j(1)$ is homotopic    to $S^j$ for $j=1,\dots,m$.

5434:  We have

5435:  \begin{equation*}

5436:  \begin{split}

5437:  &\frac{\partial e^{T^j(s)}}{\partial s}=

5438:   \frac{\partial T^j(s)}{\partial s} e^{T^j(s)}  \\

5439:   &=  e^{T^j(s)}\sum\limits_d\sum\limits_{g,u  ,v

5440:   =0}^\infty     \frac{1}{u!v ! }

5441:    \up0\langle \tau_j\Theta_j,

5442:     \underbrace{\sum\limits_{i=1}^k

5443:  t_i\Delta_i+\sum\limits_{r=1}^{j-1}\tau_r\Theta_r+s\tau_j\Theta_j}_u;

5444:  \underbrace{p ,\dots,p }_{v }

5445:  \rangle^d_{g}z^d \hbar^{g-1}.\\

5446:  \end{split}

5447:  \end{equation*}

5448:   The compactly supported form $\Theta_j$ is exact in $W$,

5449:   $$\Theta_j=d\wt\Theta_j,$$

5450: where $\wt\Theta_j$ is closed at infinity, has a cylindrical    end,

5451: and $\wt\Theta_j|_V=\theta_j$.

5452: Hence,

5453:  \begin{equation*}

5454:  % \begin{split}

5455:  \frac{\partial T^j(s)}{\partial s}  =

5456: \ \sum\limits_d\sum\limits_{g,u  ,v =0}^\infty

5457: \frac{1}{u!v !}

5458:   \up0\langle \tau_jd\wt \Theta_j,

5459:     \underbrace{\sum\limits_{i=1}^k

5460:  t_i\Delta_i+\sum\limits_{r=1}^{j-1}\tau_r\Theta_r+s\Theta_j}_u;

5461:   \underbrace{p ,\dots,p }_{v }

5462:  \rangle^d_{g}z^d \hbar^{g-1}

5463: % \end{split}

5464:  \end{equation*}

5465:   \begin{equation*}

5466:  \begin{split}

5467:  &= d\big( \sum\limits_d\sum\limits_{g,u  ,v=0}^\infty  \frac{1}{u!v

5468:  !}

5469:   \up0\langle \tau_j\wt \Theta_j,

5470:     \underbrace{\sum\limits_{i=1}^k

5471:  t_i\Delta_i+\sum\limits_{r=1}^{j-1}\tau_r\Theta_r+s\tau_j\Theta_j}_u;

5472:   \underbrace{p ,\dots,p }_{v }

5473:  \rangle^d_{g}z^d \hbar^{g-1}\big)\\

5474:   &= \frac{\partial}{\partial u}

5475:   \Big( d\big( \sum

5476:   \limits_d\sum\limits_{g,u  ,v=0}  \frac{1}{u!v!}\\

5477:  &\up0\langle

5478:     \underbrace{\sum\limits_{i=1}^k

5479:  t_i\Delta_i+\sum\limits_{r=1}^{j-1}\tau_r\Theta_r+s\tau_j\Theta_j

5480:  +u\tau_j\wt \Theta_j}_u;

5481:  \underbrace{p ,\dots,p }_{v}

5482:  \rangle^d_{g}z^d \hbar^{g-1}\big)\Big)\Big|_{u=0}\\

5483: & =  \frac{\partial}{\partial u}\Big( d\big(

5484: \bF_{W,J}(\sum\limits_{i=1}^k

5485:  t_i\Delta_i+\sum\limits_{r=1}^{j-1}\tau_r\Theta_r+

5486:  s\tau_j\Theta_j+u\tau_j\wt \Theta_j,p)\big)\Big)\Big|_{u=0} ,

5487:     \end{split}

5488:  \end{equation*}

5489:    where $d$ denotes the de Rham differential.

5490:    Using  equation (\ref{eq:non-closed2}) we get

5491:     \begin{equation*}

5492:  \begin{split}

5493: &  d\big(

5494: \bF_{W,J}(\sum\limits_{i=1}^k

5495:  t_i\Delta_i+\sum\limits_{r=1}^{j-1}\tau_r\Theta_r+

5496:  s\tau_j\Theta_j+u\tau_j\wt \Theta_j,p)\big) =\\

5497:  &e^{-\bF_{W,J}}\big( e^{\bF_{W,J}}\ola\bH(\sum\limits_{i=1}^l t_i\Delta_i +

5498:  u\tau_j\theta_j,p) \big),\\

5499:      \end{split}

5500:  \end{equation*}

5501:  and hence

5502:  \begin{equation*}

5503:  \begin{split}

5504:    &\frac{\partial T^j(s)}{\partial s}=

5505:    \frac{\partial}{\partial u}\Big( e^{-\bF_{W,J}}

5506:    \big( e^{T^j(s)}\ola\bH(\sum\limits_{i=1}^l t_i\Delta_i +

5507:  u\tau_j\theta_j,q,p) \big)\Big)\Big|_{u=0}\\

5508:  &= \tau_j e^{-T^j(s)}\Big(-\bF^j  (t_1,\dots,t_k,\tau_1,\dots,\tau_j,s,p)

5509:   \big(e^{T^j(s) }\ola\bH(\sum\limits_{i=1}^l t_i\Delta_i ,q,p)\big)\\

5510:  &+ \big( e^{T^j(s)} \bF^j  (t_1,\dots,t_k,\tau_1,\dots,\tau_j,s,p)\big)

5511:   \ola\bH(\sum\limits_{i=1}^l t_i\Delta_i ,q,p)\\

5512:  & +e^{T^j(s)}

5513:  \ola\bH^j(t_1,\dots,t_l,q,p)\Big) =

5514:  \tau_j e^{-T^j(s)}\Big(\big( e^{T^j(s)} [\bF^j , \ola\bH]

5515:  +e^{T^j(s)}

5516:  \ola\bH^j\Big),  \\

5517:     \end{split}

5518:  \end{equation*}

5519:     where

5520:     \begin{equation*}

5521:     \begin{split}

5522:     &\bF^j  (t_1,\dots,t_k,\tau_1,\dots,\tau_j,s,p) \\

5523:  &=\frac{ \partial   }{\partial u} \big(

5524:  \bF(\sum

5525:  t_i\Delta_i+\sum\limits_{r=1}^{j-1}\tau_r\Theta_r+

5526:  s\tau_j\Theta_j+u\wt\Theta_j,p)\big)\big|_{u=0}

5527:  \end{split}

5528:  \end{equation*}

5529:  and

5530:  $$\bH^j( t_1,\dots,t_l,q,p)=   \frac

5531:  { \partial   }{\partial u} \big(

5532:  \bH(\sum t_i\Delta_i +

5533:  u\theta_j,q,p)   \big)\big|_{u=0}\,.$$

5534:  Therefore,

5535:  \begin{equation*}

5536: \frac{\partial e^{T^j(s)}}{\partial s}=  e^{T^j(s)}

5537:   \frac{\partial T^j(s)}{\partial s}

5538:   =\tau_je^{T^j(s)}\big([\bF^j , \ola\bH]+\ola\bH^j\big)\,.

5539:  \end{equation*}

5540: 

5541: Let us define now a family

5542:  $U^j(s)\in \hbar^{-1}\fD,s\in [0,1]$

5543:  by the formula

5544:  \begin{equation}

5545:  e^{U^j(s)}=e^{T^j(s)}e^{(1-s)\tau_j\ola

5546:  {\bH^j}(t_1,\dots,t_l,q,p)}\,.

5547:      \end{equation}

5548:      Then $U^j(s)$ is a homotopy

5549:      between $S^j$ and $T^j(1)$. Indeed,   we have

5550:  $U^j(0)=S^j$ and $U^j(1)=T^j(1)$.

5551:  On the other hand we get an equation

5552:  \begin{equation*}

5553: % \begin{split}

5554: \frac {\partial e^{U^j(s)}}{\partial s} =\tau_je^{U^j(s)}

5555:  \big(-\ola\bH^j+ [\bF^j

5556: , \ola\bH]+\ola\bH^j\big) =e^{U^j(s)}[\tau_j\bF^j , \ola\bH]\;,

5557: %\end{split}

5558: \end{equation*}

5559: which is the definition of homotopy (see Section

5560: \ref{sec:chain} above).\qed

5561: 

5562: \bigskip

5563: 

5564: 

5565: 

5566: 

5567: We formulate now a version of Theorem \ref{thm:SFT-master} for

5568: rational potentials.

5569: 

5570: Set $$\bh^j(t_1,\dots,t_l,q,p)=\frac{\partial \bh}{\partial

5571: s_j}(\sum t_i\delta_i+s_j\theta_j,q,p) ,\;j=1,\dots m,$$ and for

5572: any $\bgg\in \fL$ we denote by $L_{\bgg}$ the Lagrangian variety of

5573: $\bV$, defined by the equation

5574: \begin{equation}

5575: L_\bgg=\{q_\g= \kappa_\g\frac{\partial\bgg}

5576: {\partial p_\g}\}.

5577: \end{equation}

5578: \begin{theorem}\label{thm:RSFT-master}

5579:  Let $\bff(t_1,\dots,t_k,\tau_1,\dots,\tau_m, p)$

5580: be the solution of the   Hamilton-Jacobi equation

5581: \begin{equation} \label{eq:RSFT-master}

5582: \frac{\partial \bff}{\partial \tau_j} (t_1,\dots,t_k,

5583: \tau_1,\dots,\tau_m)=

5584: \bh^j(t_0,\dots,t_l,q,p)|_{L_{\bff}}

5585: \end{equation}

5586: with the initial condition

5587:  $$\bff|_{\tau_j=0}=\bff_{W,J}(\sum

5588:  t_i\Delta_i+\sum\limits_{r\neq j}\tau_r\Theta_r,p).$$

5589: Then $$\bff(t_1,\dots,t_k,\tau_1,\dots,\tau_m, p)$$ is

5590:   homotopic to $$\bff_{W,J}(\sum

5591:  t_i\Delta_i+\sum\limits_{r=1}^m\tau_r\Theta_r,p).$$

5592:  \end{theorem}

5593: 

5594: 

5595: \subsection{Invariants    of Legendrian

5596: knots}\label{sec:Legendrian}

5597: Symplectic Field Theory can be extended to include Gromov-Witten

5598: invariants of pairs $(W,L)$, where $L$ is a Lagrangian

5599: submanifold of a symplectic manifold $W$.   The corresponding relative object is a

5600: pair $(W,L)$, where $W=\ora {V^-V^+}$ is a directed symplectic cobordism between

5601: contact manifolds $(V^\pm,\a^\pm)$, and $L$ is a Lagrangian cobordism between Legendrian

5602: submanifolds $\Lambda^\pm\subset V^\pm$.  More precisely, we assume that Lagrangian submanifold

5603: $L$ is cylindrical at infinity over $\Lambda^\pm$, i.e. there exists $C>0$, such that

5604: $\;L\cap V^-\times (-\infty, -C]=\Lambda^-\times (-\infty, -C]\;$ and

5605: $\; L\cap V^+\times (C,\infty]=

5606: \Lambda^+\times (C,\infty]\;$.

5607:                                   In other words, we require $L$

5608:                                    to coincide at infinity with  symplectizations

5609: of Legendrian submanifolds $\Lambda^\pm$.

5610: 

5611: The moduli space of holomorphic curves to be considered in this case

5612: consists of  equivalence classes of     holomorphic curves   with boundary which can have

5613: punctures of two types,

5614:  interior  and at the boundary.  The boundaries of holomorphic curves

5615: are required to be mapped to the Lagrangian submanifold $L$,   the holomorphic  curves

5616: should be cylindrical over periodic orbits from $\Pc^\pm$ at interior punctures,

5617:  while at boundary punctures

5618: we require them to be asymptotically cylindrical over Reeb chords connecting points  of the Legendrian submanifolds

5619: $\Lambda^\pm\subset V^\pm$.  A  more precise definition is given below.

5620: The algebraic structure arising from the stratification of boundaries of these moduli spaces is more complicated

5621: than in the closed case. First of all, unlike the interior punctures the    punctures

5622:  at the boundary are cyclically ordered, which leads to associative, rather than graded commutative algebras.

5623:  Second, the ``usual" cusp degenerations of curves with boundary

5624:  at boundary points (see \cite{Gromov-holomorphic}) has

5625:  in this case codimension

5626:  $1$, rather than $2$ as in the  closed case, and hence the  combinatorics of such

5627:   degenerations  should also be a part of the

5628:  algebraic formalism.

5629: 

5630:  We will sketch in this paper only the simplest of three cases of SFT, namely the ``classical case",

5631:  which corresponds to the theory of moduli spaces of holomorphic

5632:  disks with only $1$ positive puncture at the boundary.

5633:  \bigskip

5634: 

5635:          Let $(V,\xi=\{\a=0\})$ be a contact  manifold with a fixed

5636:          contact form $\a$, $W=V\times \R$ its symplectization

5637:          with a compatible almost complex structure $J$,

5638:          $\Lambda\subset V$ a  compact Legendrian submanifold,

5639:          and $L=\Lambda\times\R\subset W$ the symplectization of $\Lambda$, i.e.

5640:          the corresponding Lagrangian cylinder in $W$.   We  assume that all periodic orbits

5641:          of  the Reeb   vector field $R_\a$ are non-degenerate

5642:            and fix a marker

5643:          on every periodic orbit. We also consider the set $\Cc$   of Reeb chords

5644:           connecting points on $\Lambda$, and impose an extra

5645:            non-degeneracy condition along the   chords from $\Cc$

5646:           by requiring that the linearized flow of $R_\a$  along

5647:                    a chord   $c\in\Cc$  connecting points $a,b\in\Lambda$

5648:                      sends  the tangent  space $T_a(\Lambda)$ to

5649: a space transversal to $T_b(\Lambda)$.

5650: We also require that none of

5651: the chords from $\Cc$ be a part of an orbit from

5652: $\Pc$.

5653:  Under these assumptions, the set $\Cc$ is finite: $\Cc=\{c_1,\dots,c_m\}$.

5654: 

5655:  We will restrict the consideration to the case when

5656:  \begin{equation}

5657:  \pi_1(V)=0,\; \pi_2(V,\Lambda)=0,\quad\hbox{ and}\quad w_2(\Lambda)=0.

5658:  \end{equation}

5659:  First two assumptions are    technical and can be removed (comp. Section \ref{sec:dynamics} above).

5660:   However, the third one

5661:  is essential for orientability of the involved moduli spaces of holomorphic curves. Moreover, the invariants

5662:   we define depends on a particular choice of a spin-structure on $\Lambda$.

5663:   \footnote{We thank K. Fukaya for pointing this out.}

5664: 

5665:  As in  Section \ref{sec:dynamics} we choose capping surfaces $F_\g$ for $\g\in\Pc$, and  for each chord

5666:  $c\in \Cc$ we  also choose   a surface $G_c$

5667: which is bounded by  a curve

5668:   $c\cup\delta_c, $ where $\delta_c\subset \Lambda$.

5669:    The choice of  surfaces $F_\g,\,\g\in\Pc,$  allows us to define

5670:       Conley-Zehnder indices of  periodic orbits

5671:    (see Section \ref{sec:dynamics} above). Similarly, surfaces $G_c$ enable us to define

5672:    {\it Maslov indices} $\mu(c) ,\,c\in\Cc$.   Namely,  let us consider a Lagrangian subbundle of $\xi|_{\partial G_c}$,

5673:    which consists of  the Lagrangian sub-bundle

5674:   $T\Lambda|_{ \delta_c}\subset \xi|_{\delta_c}$ over $\delta_c$, together with

5675:    the family  of Lagrangian  planes  $T_u\subset \xi_u,u\in c$, which are images of $T_a(\Lambda)$ under the linearized flow of

5676:    the Reeb field $R_\a$.  Choose a symplectic trivialization of $\xi|_{ \partial G_c}$ which extends

5677:    to $G_c$.   With respect to this trivialization the above sub-bundle can be viewed

5678:    as a path of Lagrangian planes in a

5679:    symplectic vector space.  The Maslov index of such path

5680:     is defined as in \cite{Robbin-Salamon}.

5681: 

5682:     \medskip

5683:    Consider a unit disk

5684:    $D\subset\C$ with punctures $$\left(\{z^+,z_1^-,\dots,z_\sigma^-\}\cup\{x^-_1,\dots, x^-_s\}\right),$$

5685:    where  $\bz=\{z^+,z_1^-,\dots,z_\sigma^-\}$, $0\leq\sigma\leq m$,

5686:    is  a   counter-clockwise ordered set of punctures on $\partial D$,

5687:    and  $\bx=\{x^-_1,\dots, x^-_s\}$ is an ordered set of interior punctures.

5688:     As usual  we provide  interior punctures with asymptotic markers.

5689: 

5690:                         Let us denote by

5691:                          $\Mc^A(\{c_{i_1},\dots,c_{i_\sigma}\},\{\g_1,\dots,\g_s\},c_i;W,\Lambda,J)$

5692:                          the moduli space of

5693: $J$-holomorphic maps

5694: \begin{equation*}

5695: (D\setminus \left(\bz\cup\bx\right),

5696: \partial\left(D\setminus \left(\bz\cup\bx\right)\right) \to (W,L),

5697: \end{equation*}

5698:  which

5699: are asymptotically cylindrical at    the negative end over the periodic orbit $\g^-_{k}$

5700:   at the puncture $x^-_{i_k}$, and over the chord $c_{i_k}$ at the puncture $z_k$,

5701: asymptotically cylindrical at    the positive  end over the chord $c_i$

5702:    at the puncture $z^+$, and which send

5703: asymptotic markers of interior punctures to

5704: the  markers on the  corresponding

5705: periodic orbits.

5706: 

5707:   Two maps are called equivalent if they differ by a conformal map

5708:  $D\to D$ which preserves all punctures, marked points and     asymptotic markers.

5709:     Each moduli space  $\Mc^A(c_i,\{c_{i_1},\dots,c_{i_\sigma}\},\{\g_1,\dots,\g_s\};W,\Lambda,J)$ is invariant under

5710:     translations  $V\times\R\to V\times\R$ along the factor  $\R$, and we

5711:    denote the corresponding quotient moduli space  by

5712:      $$\Mc^A(c_i,\{c_{i_1},\dots,c_{i_\sigma}\},\{\g_1,\dots,\g_s\};W,\Lambda,J)/\R\;.$$

5713: 

5714: Let $(\fA,\partial)=(\fA(V,\alpha),\partial_J)$ be the graded commutative differential algebra

5715: defined in Section

5716: \ref{sec:3algebras} above, or rather its specialization at the point $0$.

5717:  Consider a graded  associative  algebra $\fK=\fK(V,\Lambda,\alpha)$ generated

5718: by elements $c_i\in \Cc$ with coefficients in the algebra $\fA$.

5719: We define a differential $\partial_{\Lambda}=\partial_{\Lambda,J}$

5720: on $\fK$ first on the generators $c_i$ by the formula

5721: \begin{equation}\label{eq:d-leg}

5722:  \partial_{\Lambda}(c_i)=

5723:  \sum \frac{n_{\Gamma,I,d}}

5724:  {k!\prod_1^k \kappa_{\g_j}^{i_j}i_j!}

5725:   c_{j_1}\dots c_{j_\sigma}

5726:  q_{\g_1}^{i_1}\dots q_{\g_k}^{i_k}z^d,

5727:   \end{equation}

5728:    where the sum is taken over all $d\in H_2(V)$, all ordered sets of

5729:    different periodic orbits $\Gamma=\{\g_1,\dots,\g_k\} $,

5730:     all multi-indices  $J=(j_1,\dots,j_\sigma),\,1\leq j_i\leq m$,

5731: and $I=(i_1,\dots,i_k)$, $i_j\geq 0, $ and where the coefficient

5732: $n_{\Gamma,I,d}$  counts the algebraic number of elements of the

5733: moduli space $$\Mc^d\big(c_i,\{c_{j_1},\dots,c_{j_\sigma}\},

5734: \{\underbrace{\g_1,\dots,\g_1}_{i_1},\dots,

5735: \underbrace{\g_k,\dots,\g_k}_{i_k}\}\big)/\R,$$ if this space is

5736: $0$-dimensional, and equals $0$ otherwise. The differential

5737: extends to the whole  algebra $\fK$ by the graded Leibnitz rule.

5738: However, it does not treat coefficients as constants: we have

5739: $\partial_\Lambda(q_\g)=\partial(q_\gamma)$, where  $\partial$ is

5740: the differential defined on the algebra $\fA$.

5741: 

5742: Then we have

5743: \begin{proposition}

5744: $$  \partial_{\Lambda}^2=0.$$

5745: \end{proposition}

5746: Given a family   of contact forms $\Lambda_\tau,

5747: \alpha_\tau,

5748: J_\tau\;\tau\in[0,1]$

5749: %CHANGE

5750: of Legendrian submanifolds, contact forms, and compatible almost complex structures one can define, similar to the  case of closed

5751: contact manifolds (see Sections \ref{sec:relative}  and Section \ref{sec:chain} above)

5752: a   homomorphism of differential algebras

5753: $$\Psi_S:\fK(V, \Lambda_0,\alpha_0)\to\fK(V, \Lambda_1,\alpha_1),$$

5754: which is independent up to homotopy of the choice of a connecting

5755: homotopy.

5756: Composition of homotopies generates composition of homomorphisms, and hence one conclude

5757: \begin{proposition}

5758: The quasi-isomorphism type of the differential algebra

5759:   $$(\fK(V,\Lambda,\alpha),\partial_{\Lambda,J})$$ depends only on the contact structure $\xi$

5760:   and the Legendrian isotopy class of $\Lambda$. The

5761:   %CHANGE

5762:   {\em Legendrian contact homology} algebra

5763:   $H_*(\fK,\partial_\Lambda)$ has a structure of a module over

5764:   the contact homology algebra

5765:   $H^\cont_*(V,\xi)=H_*(\fA,d)$, and it is an invariant of the Legendrian knot (or link) $\Lambda$.

5766:   \end{proposition}

5767: 

5768: 

5769:     The theory looks especially simple when the contact structure $\xi$ on

5770:     $V$ admits a contact form $\alpha$ such that the Reeb vector

5771:     field $R_\a$ has no closed periodic orbits. If, in addition

5772:     the space of trajectories is a  manifold $M$ (e.g. when

5773:     $V=J^1(N)=T^*(N)\times\R$ with a contact form $dz+pdq$), then

5774:      $W$ is automatically symplectic, and the projection $\pi:W\to V$ sends the Legendrian submanifold

5775:     $\Lambda\subset V$ to an immersed {\it Lagrangian} submanifold

5776:     $L\looparrowright M$ with transverse self-intersection points. These points

5777:     correspond to Reeb chords  $c_i$

5778:     connecting points on $\Lambda$.

5779:     Hence,  the algebra $\fK$ in this case is just a free associative

5780:     algebra, generated over $\C$ (or $\C$) by the

5781:     self-intersection points of $L$. It is possible

5782:      to choose a compatible  almost complex structures

5783:     $J$ on  the symplectization  $W=V\times\R$ and  $J_M$ on $M$

5784:     to make

5785:     the projection $W\to M$ holomorphic (comp. Section \ref{sec:Bott} below). Then  punctured

5786:     holomorphic disks in $W$  from moduli spaces

5787:     $\Mc^A(c_i,\{c_{i_1},\dots,c_{i_\sigma}\} ;W,\Lambda,J)$

5788:     project to $J_M$-holomorphic disks in $M$ with boundary in the

5789:     immersed Lagrangian  manifold $L$. Conversely, one can

5790:     check that each such disk  lifts to a disk from  the

5791:    corresponding moduli space $\Mc^A(c_i,\{c_{i_1},\dots,c_{i_\sigma}\}

5792:    ;W,\Lambda,J)$, uniquely, up to translation along

5793:     the $\R$-factor in $W=V\times\R$. This is especially useful when $\dim M=2$.  In

5794:    this case $L$ is an immersed curve, and the holomorphic disks

5795:    are precisely immersed, or branched disks with their boundaries in $L$.

5796:    Moreover, branched disks are never rigid, because the branching

5797:    point may vary. Hence, the  differential

5798:    $\partial:\fK\to\fK$ can be defined in this case in a pure

5799:    combinatorial way, just   summing the terms  corresponding

5800:    to  all  appropriate immersed disks whose boundary consists of arcs of $L$, and

5801:     which are locally convex near their corner.

5802: 

5803:     Yu. Chekanov independently realized  (see \cite{Chekanov}) this program  for

5804:     Legendrian links in the standard contact $\R^3$.

5805:     He was also  motivated by the  hypothetical description of the

5806:     compactification of the  moduli spaces of holomorphic discs,

5807:     but has chosen to prove the  invariance of the quasi-isomorphism type

5808:     of the differential algebra $(\fK,\partial)$ in a pure

5809:     combinatorial way.  In fact, he proved   a potentially stronger    form  of equivalence

5810:     of

5811:     differential algebras of isotopic Legendrian links,

5812:     which

5813:     he called     stable tamed isomorphism.  Stable

5814:     tame

5815:     isomorphism implies quasi-isomorphism, but we do not know

5816:     whether it is indeed stronger. Let also note that

5817:      Chekanov considered a   $\Z_2$-version of the theory.

5818:     In some examples it works better

5819:       the  $\Q$-version, which is provided by our formalism.

5820:     J. Etnyre--J. Sabloff (\cite{Etnyre-Sabloff}) and L. Ng (\cite{Ng}) worked out

5821:      the combinatorial

5822:     meaning of signs dictated by the coherent orientation

5823:     theory (see Section \ref{sec:orientation} above), and proved

5824:     invariance of the

5825:     stable tame type of the differential algebra $(\fK,\partial)$

5826:     {\it defined over $ \Z$}. Note that Chekanov's paper \cite{Chekanov}

5827:     also contains   examples  which show that the  stable  tame

5828:     $\Z_2$-isomorphism type do distinguish some Legendrian knots,

5829:     which could not be  previously distinguished.

5830: 

5831: 

5832: 

5833: 

5834:  Similar to the absolute case of SFT, one can define   further invariants of Legendrian

5835:   submanifolds by including in the formalism higher-dimensional moduli spaces. For instance,  by introducing

5836:     marked points on the boundary

5837:   of the disk

5838:   one gets a non-commutative deformation of Legendrian  contact homology

5839:   along the homology of Legendrian

5840:   manifolds. This is useful, in particular, to define invariants of

5841:     Legendrian links with ordered components.

5842:   However, the full-scale generalization of Symplectic Field Theory  to

5843:   directed symplectic-Lagrangian cobordisms between pairs

5844:   of contact manifolds and their Legendrian submanifolds, which would formalize

5845:   information  about moduli spaces of   holomorphic curves of arbitrary genus and arbitrary number

5846:   of positive and negative punctures, is not straightforward due to existence of different type

5847:   of codimension $1$ components on the boundary of the corresponding moduli spaces.

5848:   We will discuss this theory

5849:   in one of our future papers.

5850: 

5851: 

5852: 

5853: 

5854: \subsection{Remarks, examples, and further algebraic constructions in

5855: SFT}\label{sec:remarks}

5856:  \subsubsection{Dealing with torsion elements in $H_1$} \label{sec:torsion}

5857: Let us discuss grading issues for a contact manifold

5858: $(V,\xi=\{\a=0\})$ in the case when

5859:   the torsion part  of $H_1(V)$ is

5860: non-trivial. As we will see it is impossible to assign in a coherent way

5861: an integer grading to torsion elements and still keep the property

5862: that the Hamiltonian $\bH$ has total grading $-1$.  We will

5863: deal with this problem by assigning to some elements fractional

5864: degrees, and thus obtain a rational grading, incompatible with the

5865: canonical $\Z_2$-grading. In fact the

5866: term ``grading" is  misleading in this case, and more appropriately one

5867: should talk about

5868:      an Euler vector field with rational coefficients.

5869: 

5870: Let us split  $H_1(V)$ as $T\oplus F$, where $T$  and $F$ are   the

5871: torsion and free parts, respectively. As in Section \ref{sec:dynamics}

5872: above let us  choose curves ${C}_1,\dots,{C}_k$ representing a basis

5873: of $F$, fix a trivialization of the bundles $\xi_{{C}_i}$,

5874: for any periodic orbit $\g\in\Pc_\a$ with $[\g]\in F$ choose a

5875: surface $F_\g$ which realizes the homology between $[\g]$ and a

5876: linear combinations $\sum n_i [{C}_i] $, and trivialize the

5877: bundle $\xi|_\g$ accordingly.

5878: For any other  periodic orbit $\g$ let $\g_l$ be its smallest multiple which

5879: belong to  $F$. In particular, the bundle $\xi_{\g_l}$ is

5880: already trivialized by a framing $f$. The problem is that in general there is no

5881: framing over $\g$ which would produce $f$ over $\g_l$. Choose then

5882: an arbitrary framing $g$  over $\g$ and denote by $g_l$ the

5883: resulting framing over $\g_l$.  Let

5884: $2m(g_l,f)\in\pi_1(Sp(2n-2,\R))=\Z$ be the Maslov class of the

5885: framing $g_l$ with respect to $f$. The Conley-Zehnder indices of

5886: $\g_l$ with respect to these two gradings are then

5887:  related  by the formula

5888: \begin{equation*}

5889: \CZ(\g|f)=\CZ(\g|g_l)+2m(g_l,f).

5890: \end{equation*}

5891: We then assign to $\g$ the fractional  degree

5892: \begin{equation}

5893: \deg \g=\CZ(\g|g)-\frac{2m(g_l,f)}{l}\,.

5894: \end{equation}

5895: With this modification SFT can be extended to the case of contact

5896: manifolds with no restrictions on $H_1$. However, the price we pay

5897: is that this grading, even if integer, may not be compatible with

5898: the universal $\Z_2$-grading which determines the sign rules.

5899: 

5900: 

5901: 

5902: \subsubsection{Morse-Bott formalism}\label{sec:Bott}

5903: 

5904: 

5905:  Our assumption that all periodic orbits from $\Pc_\a$ for the

5906:  considered contact forms $\a$

5907:  are non-degenerate, though generic,  but is very inconvenient for

5908:  computations: in many interesting examples periodic orbits

5909:  come in continuous families. Sometimes  the  Reeb

5910: flow is periodic, and it sounds quite stupid to destroy this beautiful

5911: symmetry.

5912: 

5913: In fact the above formalism can be adapted to this ``Morse-Bott"

5914: case. We sketch below how this could be done   for the periodic

5915: Reeb flow of an $S^1$-invariant form of a

5916:  pre-quantization space.

5917: We consider below only the ``semi-classical" case which concerns

5918:  moduli spaces of rational holomorphic curves.

5919: 

5920: 

5921: 

5922: Let $(M,\omega)$ be a symplectic manifold  of dimension $2n-2$

5923: with  an integral cohomology class $[\omega]\in H^2(M)$. We will assume for simplicity

5924: that $H_1(M)=0$.

5925:  The

5926: pre-quantization space $V$ is a circle bundle over $M$ with

5927: first Chern class equal to $[\omega]$. The fibration $\pi:V\to M$

5928: admits a   $S^1$-connection form $\alpha$ whose curvature is

5929: $\omega$. It   defines a $S^1$-invariant contact structure  $\xi$

5930: on $V$, transversal to the fibers  of the fibration. The Reeb flow

5931: of $R_\a$ is periodic, so all its trajectories

5932:   are closed and coincide with the fibers of the fibration $\pi$, or their multiples.

5933: 

5934: 

5935:   The fiber of the fibration $V$ is a torsion element

5936: in $H_1(V)$, and if $l$ is the greatest divisor of the class $[\omega]$ then

5937: the  $l$-multiple of the fiber is homological to $0$.

5938: 

5939: 

5940: Consider the cylindrical cobordism  (the symplectization)

5941: $W=V\times\R$ with an almost

5942:  complex structure $J$ compatible with $\alpha$

5943: and denote by $\Mc_{0,r}(s|W,J,\a)$ the moduli  space of rational

5944: holomorphic curves in $W$ with $s$ punctures and $r$ marked points.

5945: Near punctures the curves are required to be  asymptotically

5946: cylindrical over some fibers of $V$, or their multiples.

5947: However, we do not specify to which particular fiber they are

5948:  being asymptotic, or whether this fiber is

5949:  considered

5950: on the positive, or negative end of $W$. We

5951:  do not equip   curves from $\Mc_{0,r}(s|W,J,\a)$

5952: with asymptotic markers of punctures, because   we  cannot

5953:   fix  in a continuous way  points on each simple periodic orbit,

5954: as we did in the non-degenerate case.

5955: 

5956: As it was already mentioned in Section \ref{sec:almost} above,

5957:  $W$ can be viewed  as the total space of the complex  line bundle  $L$ associated with

5958: the $S^1$-fibration $V\to M$, with the zero-section removed, and

5959: the almost complex structure $J$ can be chosen

5960:   compatible with the structure of this bundle, so that the projection   $W\to M$

5961: becomes holomorphic with respect to a certain almost complex structure $J_M$ on $M$

5962: compatible with $\omega$.

5963: Then automatically the   bundle induced over  any holomorphic curve in the base

5964: has a structure of a holomorphic line bundle.

5965: With this choice of $J$ each holomorphic curve   $f\in \Mc_{0,r}(s|W,J,\a)$ projects

5966: to a $J_M$-holomorphic sphere $\overline{f}:

5967: \C P^1\to M$,   and  can be viewed as a meromorphic section of the   induced

5968: bundle $(\overline{f})^*L$ over $\C P^1$.  This bundle is ample, and hence  poles of its

5969: sections

5970: correspond to   the  {\it negative} ends of $f$, while zeroes  correspond to  the positive ones.

5971: Notice that although  the moduli spaces

5972:  $\Mc_{0,r}(s|W,J,\a)$ can be identified with the moduli spaces

5973: of closed  holomorphic curves in a $\C P^1$-bundle over $M$

5974: with prescribed tangencies to two divisors,

5975:  their compactifications are different, and in particular the

5976:  compactification of the first moduli space may have codimension one strata

5977:  on its boundary.

5978: 

5979: 

5980: 

5981: The correspondence $f\mapsto\overline{f}$ define a fibration

5982:  $$\pr:\Mc_{0,r}(s|W,J,\a)/\R\to \Mc_{0,r+s}(M,J_M).$$

5983:  The fiber $\pr^{-1}(\overline{f})$ is   the

5984:   union of  (an infinite number of) disjoint circles, which are indexed by

5985:  sequences of integers $(k_1,\dots,k_{s+r})$ with $\sum k_i=d_0

5986:  =\int\limits_{A}\omega$,

5987:  where $A\in H_2(M)$ is the homology class realized by $\overline{f}$,

5988:   and where  there are precisely

5989:  $s$  non-zero coefficients  $k_i$.

5990: 

5991: Let us  consider two copies $\Pc^\pm$ of  the space $\Pc=\Pc_\a$ of periodic orbits,

5992:   as we need to differentiate

5993: between positive and negative ends of holomorphic curves. We will write $\ddot{\Pc}=\Pc^+\cup\Pc^-$

5994:  and define  the   evaluation maps:

5995: \begin{equation}

5996: ev_0:\Mc_{0,r}(s|W,J,\a)/\R\to V^{\times r}\;\; \hbox{ and}\;\;

5997:  ev^+_-:\Mc_{0,r}(s|W,J,\a/\R)\to\ddot{\Pc}^{\times s}\,.

5998: \end{equation}

5999: 

6000: Here $ev^+_-$ associates with each puncture the corresponding

6001: point of $\ddot\Pc$.

6002: The space $\Pc^\pm$  can be presented as

6003: $    \coprod\limits_{k=1}^{\infty}\Pc^\pm_k, $

6004:     where each $\Pc^\pm_k$ is a copy   of $M$, associated

6005:     with $k$-multiple orbits.

6006: 

6007: We will denote forms on $\Pc^+$ by $p$, on $\Pc^-$ by $q$,

6008:   denote by $p_k, q_k$ their restrictions to $\Pc_k^\pm$, and organize them

6009:   into   Fourier series

6010:   $u=\sum\limits_{k=1}^\infty(p_ke^{ikx}+q_ke^{-ikx})$. If we are given a

6011:   basis of $H^*(M)$ represented by forms $\Delta_1,\dots,\Delta_a$

6012:   we will consider only forms from the space generated by this

6013:   basis, and write $p_k=\sum\limits_{i=1}^a p_{k,i}\Delta_i,

6014: q_k=\sum\limits_{i=1}^a q_{k,i}\Delta_i$ and denote by

6015: $u_i$ the  $\Delta_i$-component of $u$, i.e.

6016:  $$u_i=\sum\limits_{k=1}^\infty(p_{k,i}e^{ikx}+q_{k,i}e^{-ikx})\;\;\hbox{

6017:  and}\;\;

6018:  u=\sum\limits_1^a u_i\Delta_i.$$

6019: 

6020: 

6021: Given a  closed form $t$ on $V$ and  a class $A\in H_2(V)$

6022:  we define the correlator

6023:  \begin{equation}

6024:  \begin{split}

6025:  & \up{-1}\langle

6026: \underbrace{t,\dots,t}_r; \underbrace{u,\dots,u}_s\rangle^A_{0} =\\

6027: &  \int\limits_{\overline{\Mc^A_{0,r}(s|W,J,\a)/\R} } ev_0^*\big(t\otimes \dots

6028: \otimes t\big)\wedge (ev^+_-)^*\big(u \otimes\dots

6029: \otimes u\big) \bigg{|}_{x=0}.\\

6030: \end{split}

6031: \end{equation}

6032: 

6033: Let us  choose a basis $A_0,\dots, A_N$ in

6034:   $H_2(M)$ in such a way that $\int\limits_{A_0}\omega=l>0$ and

6035:   $\int\limits_{A_i}\omega= 0$ for $i=1,\dots, N$.

6036:   Then the classes $A_i,i\geq 1 ,$ lift to classes $\wt{A}_i\in H_2(V)$  which

6037:   %CHANGE

6038: under the assumption $H_1(M)=0$ form

6039:    a basis of $

6040:   H_2(V)$. The   degree

6041:   $d=(d_1,\dots,d_N )$ of a class

6042:   $A\in H_2(V)$ is a vector of its coordinates in this basis.

6043: 

6044: 

6045: To associate an absolute   homology

6046: class with a holomorphic curve

6047: we  pick   the $l$-multiple (recall that   $l$ denotes the greatest divisor of $\omega$)

6048: of the fiber   $\g$ over a point $x\in M$

6049:  and

6050: choose a lift of the surface representing the class

6051:  $A_0$ with $\int\limits_{A_0}\omega =l$ as a spanning surface

6052: $F_\g$. Any other $m$-multiple of  $\g$ we will cap with the chain

6053:  $\frac ml[F_\g]$. However, to fix a spanning surface

6054: for a    fiber over any other point $y\in  M$  or

6055: its multiples, one needs to make some extra choices, for instance   fix

6056:  a path connecting $x$ and $y$. The condition $H_1(M)=0$ guarantees

6057:   independence of the homology class of the resulting surface of

6058:  the choice of this connecting path.  Notice that   with this choice,

6059:  the degree of $f\in \Mc_{0,r}(s|W,J,\a)/\R$  equals $(d_1,\dots,d_N)$, if the degree of its projection

6060:  $\pr(f)\in\Mc_{0,r+s}(M,J_m)$ is equal to $(d_0,d_1,\dots,d_N)$.

6061: 

6062: 

6063: In this notation the rational Hamiltonian $\bh=\bh_{V,J,\a }$ is defined

6064:  by the formula

6065: \begin{equation}

6066: \bh(t,u)=\sum\limits_{d}\sum\limits_{r,s=0}^\infty\frac{1}{r!s!}

6067:     \up {-1} \langle \underbrace{t,\dots,t}_r;\underbrace{u,\dots,u  }_{s}

6068:      \rangle_{0}^{d }  z^d.

6069:      \end{equation}

6070: 

6071:    Suppose that a basis of $H^*(M)$, represented by

6072:      closed forms   $\Delta_1,\dots,\Delta_a$, is  chosen in such

6073:      a way that   for $b\leq

6074: a$  the system forms $\wt{\Delta}_j=\pi^*(\Delta_j),\,j=1,\dots,b$, generate

6075:  the image $\pi^*(H^*(M))\subset H^*(V)$, and the forms

6076:  $\wt{\Theta}_1,\dots,\wt{\Theta}_c$,

6077:   complete  it

6078:    to a basis of $H^*(V)$.

6079:     We will denote  (graded) coordinates in the space generated by the forms

6080:      $\wt{\Delta}_j,\,j=1,\dots, b$ and

6081:   $\wt{\Theta}_1,\dots,\wt{\Theta}_c$ by

6082:     $ (t,\tau)=(t_1,\dots,t_b,\tau_1,\dots, \tau_c)$.

6083: 

6084: 

6085: As usual, the Hamiltonian $\bh$ is viewed as an element  of  a  graded

6086:  commutative Poisson

6087:  algebra $\fP$,

6088:  which consists of formal power series of coordinates of vectors $p_k$

6089:  and  $T=(t,\tau)=

6090:  (t_1,\dots,t_b,\tau_1,\dots,\tau_c)$

6091:   with coefficients

6092:  which are polynomials of coordinates of vectors $q_k=(q_{k,1},\dots,q_{k,a})$.

6093:   The coefficients of these polynomials

6094:  belong to a certain completion (see condition (\ref{eq:Novikov}) above) of the group algebra

6095:  of $H_2(V)$.

6096:  All  the variables $p_{k,i},q_{k,i}$  have in this case

6097:  the same parity as forms $\Delta_i$, the parity of variables $t_i$ and $\tau_j$ is the same as the  degree of

6098:  the corresponding

6099:  forms $\wt \Delta_i$ and $\wt\Theta_j$.

6100:   If $l=1$, i.e. when $H_1(V)=0$, then  the integer grading of variables which corresponds

6101:   to the   choice of capping surfaces described above  is defined

6102:    as follows:

6103: \begin{equation}\label{eq:dim-Bott}

6104: \begin{split}

6105: \deg t_i&=\deg \wt{\Delta}_i-2;\\

6106: \deg \tau_i&=\deg\wt{\Theta}_i-2;\\

6107: \deg q_{k,i}&=\deg\Delta_i-2+2ck;\\

6108: \deg p_{k,i}&=\deg\Delta_i-2-2ck;\\

6109: \deg z_i&==2c_1(A_i)\\

6110: \end{split}

6111: \end{equation}

6112: where $c=c_1(A_0)$.

6113: As it was explained in Section \ref{sec:torsion}  if $l>1$ one can only define fractional degrees, given by the above

6114: formulas (\ref{eq:dim-Bott}) with $c=\frac{c_1(A_0)}{l}$.

6115: 

6116: The following proposition is useful for applications (see below the discussion

6117: of Biran-Cieliebak conjecture about subcritical symplectic mainfolds).

6118: It follows from the fact that all the  moduli spaces

6119: $\Mc_{g,r}(s|W,J,\a)$ which we defined above  are even-dimensional.

6120: 

6121: \begin{proposition}\label{prop:even-dim}

6122: Let $(V,\xi)$ be the contact pre-quantization space for a

6123: symplectic manifold $(M,\omega)$. Then all contact homology algebras

6124: $$H_*^{\SFT}(V,\xi)\big|_{t=0},H_*^{\RSFT}(V,\xi)\big|_{t=0},H_*^{\cont}(V,\xi)\big|_{t=0}$$

6125: specialized  at $0\in H^*(V)$

6126: are

6127: free graded, respectively  Weyl, Poisson, or commutative algebras,

6128: generated by  elements

6129: $$p_{k,i},q_{k,i},\;\;i=1,\dots,a,\;k=1,\dots\,.$$ In particular, the parts of all these homology

6130: algebras graded by the homology class $w\in H_1(V)$ ({\rm see \ref{rem:grading} above})

6131: are   non-trivial.

6132:  \end{proposition}

6133: 

6134: The Poisson tensor on $\fP$  is determined in the ``$u$-notation" by

6135: the  following generalization of the   formula

6136:  (\ref{eq:u-notation}):

6137: \begin{equation}\label{eq:u-vector}

6138: %CHANGE

6139:  \frac{1}{2\pi i}\int\limits_0^{2\pi}\langle(\delta u)',\delta

6140: v\rangle dx,

6141: \end{equation}

6142:  where $\langle\;,\;\rangle$ denotes Poincar\'e pairing

6143: on cohomology $H^*(M)$,  which is given in the basis $\Delta_1,\dots, \Delta_a$ by the matrix

6144: $$\eta_{ij}=\langle\Delta_i,\Delta_j\rangle=\int\limits_M\Delta_i\wedge\Delta_j.$$

6145: The Poisson tensor  can be written in

6146: $(p,q)$-coordinates as

6147: \begin{equation*}

6148: \sum\limits_{k=1}^\infty k\sum\limits_{i,j=1}^{a }\eta_{ij}\frac{\partial}{\partial

6149: p_{k,i}}\wedge\frac{\partial}{\partial

6150: q_{k,j}}\,.

6151: \end{equation*}

6152: It can be shown that the  above Hamiltonian $\bh$

6153: satisfies the identity $\{\bh,\bh\}=0,$  and  that the differential

6154: Poisson algebra $(\fP,d^{\bh})$ is quasi-isomorphic to the

6155: corresponding  differential Poisson algebra defined  in Section

6156: \ref{sec:3algebras} for any non-degenerate contact form for the

6157: same contact structure  $\xi$ on $V$.

6158: \bigskip

6159: 

6160:  The following formula (\ref{eq:Bourg}), which sometimes allows to compute

6161:  the Hamiltonian $\bh$  of $V$

6162:   in terms of the Gromov-Witten invariant $\bff=\bff_{M,J_M}$ of $M$, emerged   in a discussion  of

6163:   the authors with T. Coates and  F. Bourgeois.

6164: 

6165:     \begin{proposition}\label{prop:Coates}

6166:     Set

6167:     $$

6168:     \bh^{j}_{W,J}(t,q,p,z)=\frac{\partial\bh}

6169:     {\partial\tau_j}(\sum\limits_1^b

6170:     t_i\wt\Delta_i+\tau_j\wt\Theta_j,q,p,z)

6171:     \big|_{\tau_j=0}\;,

6172:     $$

6173:     and

6174:     $$\widehat\bff^j(t,z)=\frac{\partial\bff}{\partial s}\big(

6175:     \sum\limits_1^a t_i\Delta_i +s\pi_*\wt\Theta_j,z\big)\big|_{s=0}

6176:     \,,$$

6177:     for $j=1,\dots,c$

6178:           (comp. Theorem \ref{thm:RSFT-master}).

6179:    Then   we have

6180:    \begin{equation}

6181:       \label{eq:Bourg}

6182:       \begin{split}

6183:     &\bh^j_{W,J}(t_1,\dots,t_b,q,p,z)\\&

6184:     =\frac{1}{2\pi }\int\limits_0^{2\pi}

6185:     \widehat\bff^j(t_1+u_1(x),\dots,t_b+u_b(x), u_{b+1}(x),\dots,u_a(x),\tilde z)dx

6186:     \end{split}

6187:         \end{equation}

6188:      where $z=(z_1,\dots,z_N)$,

6189:       $\tilde z=(e^{-ilx},z_1,\dots,z_N)$ and

6190:       $l$ is the greatest divisor of $\omega$.

6191:      \end{proposition}

6192:      To prove    (\ref{eq:Bourg}) one just observes that

6193:       the correlator

6194:           \begin{equation*}

6195:  \up{-1}\langle

6196: \wt \Delta_{j_1},\dots,\wt\Delta_{j_r},\wt\Theta_j;

6197: u_{i_1}\Delta_{i_i},\dots,u_{i_v}\Delta_v\rangle^d_{0}

6198: \end{equation*}

6199: equals  the Fourier coefficient with $e^{ilx}$ of

6200: the       correlator

6201:    \begin{equation*}

6202:  \up{0}\langle

6203: {\Delta_{j_1},\dots,\Delta_{j_r},\pi_*\wt\Theta_j}, {

6204: u_{i_1}\Delta_{i_i},\dots,u_{i_v}\Delta_v}\rangle^{\tilde

6205: d}_{0}\,.

6206: \end{equation*}

6207: 

6208:       Notice that if $\wt\Theta_j$ is an odd form, then

6209:       $$\bh(\sum\limits_1^b

6210:     t_i\wt\Delta_i+\tau_j\wt\Theta_j,q,p,\tilde z)=

6211:     \tau_j\bh^j(t,q,p,z),$$

6212:     because all terms of $\bh$ must contain $\tau_j$.

6213:                                         In particular,

6214:      for $M=\C P^{n-1}$  the manifold $V$ is a rational homology sphere, and thus

6215:      a volume form $\Theta$  on $S^{2n-1}$

6216:      generates the odd part

6217:      of $H^*(V;\R)$.

6218:      Hence, the formula

6219:      (\ref{eq:Bourg}) completely determines $\bh$.

6220:          Namely, let

6221:          $\bff(t,z)$ be the Gromov-Witten invariant of $\C P^{n-1}$,

6222:          and let $\Delta_{2i},i=0,\dots,n-1$, be closed forms generating

6223:      $H^*(\C P^{n-1})$, so that $\Delta_{2n-2}=\pi_*(\wt\Theta)$.

6224:       Set $\wt\Delta_0=\pi^*(\Delta_0)$

6225:       and $$\widehat\bff(t,z)=\widehat\bff^{2n-2}(t,z)=\frac{\partial\bff(t,z)}{\partial t_{2n-2}}.$$

6226: 

6227:      Then we have

6228:      \begin{equation}\label{eq:Bourg2}

6229:   \bh(t_0\wt\Delta_0+\tau\Theta,q,p) =

6230:    \frac{\tau}{2\pi } \int\limits_0^{2\pi}

6231: {\widehat\bff}( t_0\Delta_0+u, e^{-ix})  dx .

6232:  \end{equation}

6233: 

6234: 

6235: Let us consider some applications of the formula (\ref{eq:Bourg2}).

6236: 

6237: %%%%%%%

6238: 

6239: 

6240: 

6241: 

6242: \bigskip

6243: 

6244: \noindent\textsf {Contact homology of the standard contact $3$-sphere}

6245: \medskip

6246: 

6247:   Let $\pi:V=S^3\to M=\C P^1$ be the Hopf

6248:  fibration. $V$ is the pre-quantization space for $(S^2,\omega)$ with

6249:  $\int\limits_{S^2}\omega=1$. The $0$-form $\Delta_0=1$ and

6250:    the  symplectic  $2$-form $\Delta_2=\omega$

6251:    generate $H^*(M)$, the $0$-form

6252:        $\wt{\Delta}_0=\pi^*(\Delta_0)$ and the volume form $\wt{\Theta}_3$

6253:         with $\pi_*(\wt{\Theta}_3)=\Delta_2$ on $S^3$ generate $H^*(S^3)$.

6254:          Thus we have  functional variables

6255:                  \begin{equation*}

6256:            u_j(w)=\sum\limits_{k=1}^\infty \left(p_{k,j}e^{ikx}+q_{k,j}e^{-ikx}\right),

6257:  \end{equation*}

6258: associated  to the classes $\Delta_j,\; j=0,2,$ and variables $t_0$ and $\tau$

6259:  associated to $\wt{\Delta}_0$ and $\wt{\Theta}_3$. According to (\ref{eq:dim-Bott}) we have

6260: \begin{equation*}

6261: \begin{split}

6262: &\deg q_{k,0}=-2+4k,\quad \deg q_{k,2}= 4k,

6263: \quad\deg p_{k,0}=-2-4k,\\

6264: &\quad\deg p_{k,2}=- 4k,\quad

6265:  \deg t_0=-2,\quad\deg \tau=1.\\

6266: \end{split}

6267: \end{equation*}

6268:  The  potential $\bff$ for $M=\C P^1$ can be written, as it well known (see

6269:  also Section \ref{sec:computing} below),

6270: as

6271: \begin{equation}\label{eq:CP2}

6272: \bff=\frac{t_0^2t_2}{2} +e^{t_2}z,

6273: \end{equation}

6274:   and hence

6275:   \begin{equation*}

6276: \widehat\bff=\frac{t_0^2}{2} +e^{t_2}z,

6277: \end{equation*}

6278: Thus applying (\ref{eq:Bourg}) we get the following expression for the rational Hamiltonian

6279: $\bh$ for $S^3$:

6280: \begin{equation}\label{eq:3-sphere}

6281: \begin{split}

6282:  \bh&=\frac{\tau}{2\pi i}\int\limits_0^{2\pi} \left(\frac{(t_0+u_0)^2}{2}+

6283: e^{u_2-ix}\right)dx = \tau\Big(\frac{t_0^2}{2}+\sum\limits_{k\geq 1} q_{k,0}p_{k,0}\\

6284: &+\sum\limits_{t,s\geq 0}\mathop{\mathop{\sum

6285: \limits_{i_1,\dots,i_s\geq 0}}\limits_{j_1,\dots j_t\geq 0}}

6286: \limits_{\sum\limits_1^s li_l-\sum\limits_1^tmj_m=1}\frac{q_{1,2}^{i_1}\dots

6287: q_{s,2}^{i_s}p_{1,2}^{j_1}\dots p_{t,2}^{i_t}}{i_1!\dots i_s!j_1!\dots

6288: j_t!}\Big)\,.\\

6289: \end{split}

6290: \end{equation}

6291: 

6292: 

6293: Let us use (\ref{eq:3-sphere})

6294: to compute the contact homology algebra $$H^{\cont}_*(S^3,\xi_0)=H_*(\fA(S^3,J,\alpha),\partial).$$

6295: The part of $\bh$  linear in the $p$-variables has the form

6296: \begin{equation*}

6297: \tau\sum\limits_1^\infty\left( p_{k,0}q_{k,0}+ p_{k,2}h_k(q_{1,2},\dots,q_{k-1,2})\right)\;,

6298: \end{equation*}

6299: so that the differential $\partial :\fA\to\fA$

6300: is given by the formulas

6301: \begin{equation*}

6302: \partial q_{k,2}=k\tau q_{k,0},\quad \partial q_{k,0}=k\tau h_k(q_{1,2},\dots,q_{k-1,2})\;.

6303: \end{equation*}

6304: %CHANGE

6305: Here are

6306: few first polynomials $h_k$:

6307:  \begin{equation*}

6308:  \begin{split}

6309: h_1=&1,\\

6310: h_2=&q_{1,2},\\

6311: h_3=&q_{2,2}+\frac12q_{1,2}^2,\\

6312: h_4=&q_{3,2}+  q_{2,2}q_{1,2}+\frac16q_{1,2}^3.\\

6313: \end{split}

6314: \end{equation*}

6315: Notice that $\im\partial$ coincides with the ideal $I(\tau)$

6316: generated by $\tau$. Hence,  the homology algebra

6317: $H_*(\fA,\partial)$  specialized over a point $t=(t_0,0)$ is a

6318: free graded commutative algebra $ \fA_0$ generated by variables

6319: $q_{k,0},q_{k,2},\;k=1,\dots,$  and in particular it has one

6320: generator in each even dimension. On the other hand, over any

6321: point $t=(t_0,\tau)$ with  $\tau\neq 0$ the algebra

6322: $H_*(\fA,\partial)$ is isomorphic to a  proper subalgebra $\fA_1$

6323: of $ \fA_0$. It       has its first non-trivial generator $g_1=

6324: q_{1,2}-\frac12 q_{1,0}^2 $ in dimension $4$.

6325: \bigskip

6326: 

6327: \begin{remark}\label{rem:lens-space}

6328: {\rm The contact homology of the Lens space $V=L(l,1)$ which is

6329:  the pre-quantization space for $(S^2,\omega)$ with

6330:  $\int\limits_{S^2}\omega=l$ can be computed similarly.

6331:  The variables   $p_{k,0}, q_{k,0}, p_{k,2}, q_{k,2}, t_0 $ and $\tau$ have the same

6332:  $\Z_2$-grading, as in the case of $S_3$, i.e. all of them, except  $\tau$ are even.

6333:  However, the grading assigned by the Euler field to $p_{k,0}, q_{k,0}, p_{k,2}, q_{k,2}$

6334:   is fractional in this case

6335:  and given by formulas

6336: \begin{equation*}

6337: \begin{split}

6338: &\deg q_{k,0}=-2+\frac{4k}{l},\quad \deg q_{k,2}= \frac{4k}{l},

6339: \quad\deg p_{k,0}=-2-\frac{4k}{l},\\

6340: &\quad\deg q_{k,0}=- \frac{4k}{l} .\\

6341: \end{split}

6342: \end{equation*}

6343: 

6344: The formula for $\bh$ takes the form

6345: \begin{equation}\label{eq:lens-space}

6346: \bh=\frac{\tau}{2\pi i}\int\limits_0^{2\pi} \left(\frac{(t_0+u_0)^2}{2}+

6347: e^{u_2-ilx}\right)dx

6348: \end{equation}

6349: 

6350: We will not carry on here the computation of the contact homology  of

6351: the Lens space $L(l,1)$, and only note, that  as in the case of

6352: $S^3$

6353:   the homology algebra $H_*(\fA,\partial)$  specialized over a point $t=0$ is a free

6354: graded commutative algebra $ \fA_0$ generated by variables

6355: $q_{k,0},q_{k,2},\;k=1,\dots,$

6356:  and over any  point $t \neq 0$  the

6357: algebra $H_*(\fA,\partial)$ is isomorphic to a  proper subalgebra $\fA_1$ of $ \fA_0$.

6358: In particular, {\it over any point the contact homology algebra $H_*(\fA,\partial)$

6359:   has no odd elements}.}

6360: \end{remark}

6361:  %%%%%%

6362: \noindent\textsf{ Distinguishing contact structures on

6363: pre-quantizations spaces}

6364:  \medskip

6365: 

6366:  Formula (\ref{eq:Bourg}) can be used

6367:    for distinguishing contact structures on pre-quantization

6368:    spaces

6369:    of certain symplectic manifolds, which have different

6370:    Gromov-Witten invariants.

6371:                          Here is an example.

6372:                          \begin{proposition} \label{prop:Barlow}

6373:                          Let $(M_1,\omega_1)$

6374:                          and $(M_2,\omega_2,J_2)$ be two

6375:                          symplectic $4$-manifolds with  integral

6376:                          cohomology classes of their symplectic forms.

6377:                          Suppose that for compatible

6378:                          almost complex structures $J_1$ on $M_1$ and $J_2$

6379:                          on $M_2$ there are no non-constant  $J_1$-holomorphic spheres

6380:                          in

6381:                         $ M_1 $,

6382:                          while $M_2$ contains an embedded

6383:                          $J_2$-holomorphic $(-1)$-sphere $S$.

6384:                           Then  the pre-quantization spaces $(V_1,\xi_1)$ and $(V_2,\xi_2)$

6385: are not contactomorphic.

6386: \footnote{

6387: Yongbin Ruan proved in \cite{Ruan-virtual}

6388: that under the assumptions of Proposition \ref{prop:Barlow}   the

6389: symplectic manifolds $(M_1,\omega_1)\times \C P^1$ and

6390:  $(M_2,\omega_2)\times \C P^1$ are not symplectomorphic

6391:  (and  not even deformationally equivalent), despite   the fact

6392:  that    for a certain choice of $M_1$ and $M_2$

6393:  (e.g. $M_1$ is the Barlow surface and $M_2=\C P^2\#8\overline{\C P^2}$), and for

6394:  appropriate symplectic forms $\omega_1$ and $\omega_2$

6395:  the underlying manifolds $V_1$ and $V_2$ are diffeomorphic. }

6396: \end{proposition}

6397: \begin{remark}

6398: {\rm Even when the manifolds $M_1$ and $M_2$ are homeomorphic,

6399: the prequantization spaces $V_1$ and $V_2$ are not diffeomorphic

6400: (even not homotopy equivalent!)  for most choices   of symplectic forms

6401: $\omega_1$ and $\omega_2$, and hence the statement of the theorem is trivial in these

6402: cases. However, one can show that

6403: for homeomorphic $M_1$ and $M_2$ the

6404: symplectic forms can always be deformed in the class of symplectic forms compatible with

6405: the chosen almost complex structures $J_1$ and $J_2$, in order

6406:   to make $V_1$ and $V_2$ diffeomorphic.}

6407: \end{remark}

6408: To prove Proposition \ref{prop:Barlow}   we will ahow that

6409:    the ``classical"   contact homology algebras

6410:  $H^{\cont}_*(V_1,\xi_1) $ and $H^{\cont}_*(V_2,\xi_2) $ are not isomorphic.

6411: 

6412: Let  $S_0=S,S_1,\dots, S_m$ be the exceptional $J_2$-holomorphic

6413: spheres in $M_2$. Then the cohomology classes $ S^*_0,\dots,

6414: S^*_m,[\omega]$ are linearly independent, where  we denote by

6415: $S^*_i$ the cohomology class Poincar\'e-dual to $[S_i]\in

6416: H_2(M_2)$. Hence there exists a class $X\in H^2(M_2)$, such that

6417: \begin{equation}

6418: XS^*_0=1, \;X[\omega]=0,\;\;\hbox{and}\;\;XS^*_i=0\;\;\hbox{for}\;\; 1\geq i\geq m.

6419: \end{equation}

6420: Then the  potential $\bff_{M_2,J_2}(tX)$

6421: coincides with $$\bff_{S,J_2|_S}(tX|_{S})=e^tz.$$

6422:  Let us choose a basis of closed forms $\Delta_i, i=0,\dots,a$,

6423: generating

6424:   generating $H^*(M_2)$, so that one of the forms, say $\Delta_1$

6425:   represents the class $X$. The form $\Delta_1$ then

6426:      lifts to a  form $\wt \Delta_1$ such that

6427: $\pi_*\wt \Delta_1=\Delta_1$.

6428:  According to the formula

6429: (\ref{eq:Bourg2}) we have

6430: \begin{equation}

6431: \bh_V( \tau_1\wt\Delta_1,u)= \frac{\tau_1}{2\pi } \int\limits_0^{2\pi}

6432: e^{u_1-ilx} dx ,

6433: \end{equation}

6434: where $u=\sum\limits_0^au_j\Delta_j,\,u_j=\sum\limits_1^\infty

6435:  (p_{k,j}e^{ikx}+q_{k,j}e^{-ikx}),\;l=\int\limits_{S}\omega.$

6436: 

6437:  Hence, the contact homology algebra $H^{\cont}_*(M_2,\xi_2)$,

6438:  specialized     at the point $\tau\wt\Delta$ for $\tau\neq 0$

6439:  is isomorphic to the contact homology algebra of the  standard

6440:  contact

6441:  Lens space $L(l,1)=\pi^{-1}(S)\subset V$, specialized    at the volume

6442: form $\tau\wt\Delta|_{L(l,1)}$. It follows then from  Remark \ref{rem:lens-space}

6443:    that $H^{\cont}_*(V_2,\xi_2)$,

6444:  specialized     at any point  has

6445:  no odd elements.   On the other hand, for any $2$-dimensional

6446:  cohomology class $t\in H_2(M_1)$ we have

6447:  $\bff_{M_1,J_1}(t)=0$, and hence for any $3$-form

6448:  $\wt\Delta$ on $V_1$,

6449:  the formula (\ref{eq:Bourg2}) implies that the Hamiltonian

6450:  $\bh_{V_1} (\wt\Delta, u )$ vanishes as well, and  therefore

6451:  the contact homology algebra  $H^{\cont}_*(M_1,\xi_1)$,

6452:  specialized     at the point $\tau\wt\Delta,\tau\neq 0$

6453: is a free graded commutative algebra generated by the odd variable

6454: $\tau$, and even variables $p'_{k,j},q'_{k,j }$, which correspond

6455: to even dimensional generators of $H^*(M_1)$. Hence the contact

6456: manifolds $(M_1,\xi_1)$ and $(M_2,\xi_2)$ are not isomorphic. \qed

6457: 

6458: 

6459: 

6460:   \bigskip

6461: 

6462:  \noindent\textsf{ Subcritical symplectic manifolds}

6463:   \medskip

6464: 

6465:    The content of this example is a result of our discussion

6466:    with P. Biran and K. Cieliebak.

6467: 

6468:   In \cite{Donaldson} S. Donaldson  generalized  the Kodaira embedding theorem by

6469:   proving that for any closed  symplectic manifold $(W,\omega)$

6470:    with an integral cohomology class of the symplectic form there

6471:    exists an integer $l>0$ such that the homology class dual to $l[\omega]$

6472:    can be represented by an embedded symplectic hypersurface $W_0$.

6473:   In fact, S. Donaldson proved a stronger result, which  together

6474:   with

6475:    an improvement by  Biran-Cieliebak   asserts that for a sufficiently large $l$

6476:    the hypersurface $W_0$ can be chosen in such a way that in the complement

6477:    $W\setminus W_0$ there exists a vector field $X$ with the following properties:

6478:    \begin{description}

6479:    \item{} $L_X\omega=-\omega$, where $L_X$ denotes the Lie derivative

6480:    along $X$; in other words, $X$ is conformally symplectic and   contracting;

6481:     \item{}  $X$ is forward integrable;

6482:     \item{} $X$ is gradient-like for a Morse function $\phi:W\setminus W_0\to \R_+$,

6483:     which coincides with $-\log\dist^2$ near $W$, where $\dist(x)$ is the     distance

6484:     function from a point $x$ to $W_0$ with respect to some Riemannian metric.

6485:     \end{description}

6486: 

6487:        The vector field $X$ retracts $W\setminus W_0$ to the Morse complex $K$

6488:        of the function $\phi$,

6489:        which is automatically isotropic for the symplectic form $\omega$

6490:        (see  \cite{Eliash-Gromov-convex}), and, in particular, $\dim K\leq n$.

6491:          Biran-Cieliebak call the pair $(W,W_0)$ {\it subcritical}

6492:            if $\dim K<n$.

6493:       They constructed  in \cite{Biran-Cieliebak}

6494:     several interesting  examples of subcritical pairs, and conjectured

6495:     that {\it if $(W,W_0)$ is subcritical, then $l=1$}.

6496:     We sketch below  the proof of this conjecture.

6497: 

6498: First, let us observe that the contact structure $\xi$, defined by

6499: the contact form $\a=

6500:  X\hook\omega$ on the boundary  $V$ of a  small tubular neighborhood of

6501: $W_0$, is equivalent to the contact structure which is defined on $V$ as the

6502: pre-quantization space   of the symplectic manifold

6503: $(W_0,l\omega)$. On the other hand, the condition, that the pair

6504: $(W,W_0)$ is sub-critical implies that the contact manifold $(V,\xi)$

6505: is itself  subcritical in the sense of Examlpe \ref{ex:Floer}.4 above,

6506: i.e. it is  isomorphic to the strictly pseudo-convex

6507: boundary  of a sub-critical Stein (or Weinstein) manifold with its canonical complex structure.

6508: Let us recall (see \ref{rem:grading} above) that all SFT-objects, in particular Floer contact

6509: homology $HC_*(V,\xi)$ and the contact homology algebra $H^{\cont}_*(V,\xi)$ are

6510: graded by elements of $H_1(V)$.

6511: Using  arguments  as in  the theorem of Mei-Lin Yau (see \ref{prop:MLYau} above) one can show

6512:  that all  non-trivial elements

6513:  in  the

6514: contact homology algebra $H^{\cont}_*(V,\xi)$ of a subcritical contact manifold

6515: $(V,\xi)$ may correspond only to $0\in H_1(V)$.

6516:   On the other hand,

6517:  it follows from Proposition \ref{prop:even-dim} above that $H^{\cont}_*(V,\xi)$ specialized at $0\in H^*(V)$ has

6518:  non-trivial elements which correspond to the homology class of       the fiber in

6519:  $H_1(V)$. Therefore, $l=1$.

6520: 

6521: \subsubsection{Computing  rational Gromov-Witten

6522: invariants of $\C P^n$}\label{sec:computing}

6523: We will  show  in this section how SFT  can be used for computing      rational

6524:  Gromov-Witten invariants of $\C P^n$. Our method differs from

6525:  traditional ways (see \cite{Kontsevich-CP2,Fulton-Pand,Graber-Pand,Vakil,Ruan-Tian})   for this computation.

6526:   We will be simultaneously computing the rational potential of $\C^n$ and the rational

6527: Gromov-Witten invariant of $\C P^n$ by a recursion

6528: using Theorem \ref{thm:RSFT-master}

6529: 

6530: Let us choose basic forms in $\C^n$ as in the previous section,

6531: i.e $\Delta =1$, and $\Theta$ is a volume form with  compact

6532: support in $C^n\setminus 0=S^{2n-1}\times(0,\infty)$

6533:   with $\int\limits_{\C^n}\Theta =1$. We denote by $\delta$ the restriction of

6534:   $\Delta$ to $S^{2n-1}$.

6535: We also assume that $\Theta$ splits into a product $\hat\theta\wedge\rho$, where $\hat\theta$

6536: is pull-back of a unit volume form $\theta$ on $S^{2n-1}$, and $\rho$ is a

6537: compactly supported form in $(0,\infty)$.

6538: Set

6539: \begin{equation}

6540: \begin{split}

6541: \bh^1(t_0,q,p))&=\frac{\partial \bh}{\partial

6542: \tau}(t_0\delta+\tau\theta,q,p)|_{\tau=0}\\ &=\frac{1}{2\pi

6543: }\int\limits_0^{2\pi}\widehat\bff_{\C P^{n-1}}(t_0+u_0(x),u_2,\dots,

6544: u_{2n-2},e^{-ix})dx,\\

6545: \end{split}

6546: \end{equation}

6547:  where  $$\widehat\bff_{\C P^{n-1}}(t_0,\dots,t_{2n-2},z)=

6548:  \frac{\partial\bff_{\C P^{n-1}}}{\partial t_{2n-2}}(t_0,\dots,t_{2n-2},z),$$

6549:  $\bff_{\C P^{n-1}}(t_0,\dots,t_{2n-2},z)$ is the rational

6550: Gromov-Witten invariant  of $\C P^{n-1}$, and

6551: $$u_{2j}(x)=\sum\limits_1^\infty

6552: p_{k,2j}\,e^{ikx}+q_{k,2j}\,e^{-ikx}.$$ Then the equation

6553: (\ref{eq:RSFT-master}), which determines $

6554: \bff(t_0,t_{2n},p)=\bff_{\C^n}(t_0\Delta+t_{2n}\Theta,p)$ takes

6555: the form

6556: \begin{equation}\label{eq:Cn}

6557: %\begin{split}

6558: \frac{\partial\bff}{\partial t_{2n}}(t_0,t_{2n},p) =\frac{1}{2\pi

6559: }\int\limits_0^{2\pi} \widehat\bff_{\C P^{n-1}}(t_0+u_0(x),u_2,\dots,

6560: u_{2n-2},e^{-ix})dx\big|_{L_\bff},

6561: %\end{split}

6562: \end{equation}

6563: where $$L_{\bff}=\{q_{k,2j}=k\frac{\partial \bff}{\partial

6564: p_{k,2n-2j-2}}(t_0,t_{2n},p)\}.$$

6565:  Together with the initial data

6566: \begin{equation}

6567: \bff(t_0,0,p)=

6568: \begin{cases}

6569: p_{1,0},&\hbox{if}\;\;n=1;\\

6570: 0,&\hbox {otherwise} \\

6571: \end{cases}

6572: \end{equation}

6573: the equation (\ref{eq:Cn}) provides a recursive procedure for

6574: computing coefficients $f_j(t_0,p)$ of the expansion

6575: $$\bff(t_0,t_{2n},p)=\sum\limits_0^\infty f_j(t_0,p)t_{2n}^j.$$

6576: 

6577: For instance for $n=1$ we have (see Example \ref{ex:circle})

6578: $\bh^1=\frac{t_0^2}{2}+\sum\limits_1^\infty p_kq_k$, where we

6579: write

6580: $p_k,q_k$ instead of $p_{k,0},q_{k,0}$, and hence the equation

6581: (\ref{eq:Cn}) takes the form

6582: \begin{equation}

6583: \frac{\partial\bff}{\partial t_{2 }}(t_0,t_{2 },p)=\frac{t_0^2}{2}

6584: + \sum\limits_0^\infty k p_k\frac{\partial\bff}{\partial

6585: p_k}(t_0,t_{2 },p)

6586: \end{equation}

6587: with the initial data $\bff(t_0,0,p)=p_1$. This linear first order

6588: PDE is straightforward to solve, and we get $$\bff(t_0,t_{2

6589: },p)=\frac{t_2t_0^2}{2}+e^{t_2}p_1.$$

6590: \bigskip

6591: 

6592: For $n=2$ the Hamiltonian $\bh$ is given by the formula

6593: (\ref{eq:3-sphere}), and we have

6594: $$\bh(t_0,\tau,p)=\tau\bh^1(t_0,p).$$ Thus the equation for the

6595: potential of $\C^2$ has the form

6596: \begin{equation}\label{eq:C2}

6597: \begin{split}

6598: &\frac{\partial\bff}{\partial t_{4}}(t_0,t_{4},p)

6599:  =\frac{t_0^2}{2}+\sum\limits_{k\geq 1}k

6600: \frac{\partial\bff}{\partial p_{k,2}}p_{k,0}\\

6601: &+\sum\limits_{t,s\geq 0}\mathop{\mathop{\sum

6602: \limits_{i_1,\dots,i_s\geq 0}}\limits_{j_1,\dots j_t\geq 0}}

6603: \limits_{\sum\limits_1^s li_l-\sum\limits_1^tmj_m=1} \frac{ \left(

6604: \frac{\partial\bff} {\partial p_{1,0}}\right)^{i_1}\dots

6605:  \left(s \frac{\partial\bff}{\partial p_{s,0}}\right)^{i_s}p_{1,2}^{j_1}\dots

6606:  p_{t,2}^{i_t}}{i_1!\dots i_s!j_1!\dots

6607: j_t!};\\

6608: &\bff(t_0,0,p)=0\,.\\

6609: \end{split}

6610: \end{equation}

6611: Hence, we get

6612: \begin{equation}

6613: \begin{split}

6614: \bff(t_0,t_4,p)=&t_4(\frac{t_0^2}{2}+p_{1,2})+\frac{t_4^2}{2!}p_{1,0}+

6615: \frac{t_4^3}{3!}(p_{2,2}+\frac12 p_{1,2}^2)\\

6616: &+\frac{t_4^4}{4!}(2p_{2,0}+p_{1,2}p_{1,0})+\dots\;.

6617: \end{split}

6618: \end{equation}

6619: 

6620: To compute $\bff_{\C^n}$ for $n>2$ we need to know $\bff_{\C

6621: P^{n-1}}$. So to complete the recursion we will explain now how to

6622: express the rational Gromov-Witten invariant $\bff_{ \C P^{n-1}}$

6623: through $\bff_{\C^n}$.

6624: 

6625: \bigskip

6626: First of all we split, as it is described in Example

6627: \ref{ex:section}

6628:  above, $\C P^n$ along the boundary of a tubular neighborhood

6629: of $\C P^{n-1}\subset \C P^n$ into two completed symplectic

6630: cobordism $W_1=\C^n$ and $W_2=\C P^{n }\setminus {x}$, where we

6631: introduce on $W_2$ a complex structure of the holomorphic line

6632: bundle over $\C P^{n-1}$ determined by the hyperplane section $\C

6633: P^{n-2}\subset\C P^{n-1}$. We will denote by $\bff_1$ and $\bff_2$

6634: the potentials for $W_1$ and $W_2$, respectively.

6635: 

6636: Let $\Delta_0,\dots,\Delta_{2 n-2}$ be closed forms representing

6637: the standard basis of $H^*(\C P^{n-1})$. We will keep the same

6638: notation for the pull-backs of these forms to $W_2$. Let

6639: $\Delta_{2n}$ be  a closed $2n$-form with a compact support, which

6640: generates $$\Ker\big(H^*_{\comp}(W_2)\to H^*(W_2)\big).$$

6641:  We are interested

6642: in the potential

6643: $\bff_2(t_0,\dots,t_{2n},q)=\bff_2(\sum\limits_{i=1}^{n }

6644: t_{2i}\Delta_{2i},q)$. First of all notice that by dimensional

6645: reasons the moduli spaces of holomorphic curves which project  to

6646: non-constant curves in $\C P^{n-1}$ do not contribute to the

6647: potential $$\bff_2(t_0,\dots,t_{2n-2},0,q),$$ and hence we have

6648: \begin{equation}

6649: \bff_2(\sum\limits_{i=1}^{n-1} t_{2i}\Delta_{2i},q)= z\sum

6650: \limits_{i=0}^{n-1}q_{1,2i} \mathop{\sum\limits_{(s_1,\dots,s_{n-1})}}\limits_{  \sum

6651: s_j(j-1)=n-i-1}\prod\limits_{j=1}^{n-1}\frac{t_{2j}^{s_j}}{s_j!}\;.

6652: \end{equation}

6653: In particular, for $n=2$ we get

6654: \begin{equation*}

6655: \bff_2(t_0\Delta_0+t_2\Delta_2,q)=ze^t_2q_{1,2}

6656: \end{equation*}

6657:  One can recover $\bff_2(t_0,\dots,t_{2n},q)$ for $t_{2n}\neq 0$

6658: using the equation (\ref{eq:RSFT-master}), as we did it above for

6659: $W_1=\C^n$. However, for the purpose of our  computation of

6660: Gromov-Witten invariant $\bff_{\C P^n}$ this is not necessary, as

6661: we can proceed as follows.

6662: 

6663: Notice that the above chosen  forms

6664: $\Delta_2, \Delta_{2n-2}$ extend to $\ CP^n$. On the other hand, we

6665: will choose a  volume form $\Delta_{2n}$ on $\C P^n$ to be

6666: supported in the affine part, so that the restriction $\Delta_{2n}|_{\C^n}$

6667: coincides with the form $\Theta$ introduced above. Then Theorem

6668: (\ref{thm:RSFT-composition}) implies that

6669: \begin{equation}\label{eq:CPn}

6670: \begin{split}

6671: &\bff_{\C P^n}(\sum\limits_{i=1}^{n } t_{2i}\Delta_{2i})=\\

6672:  &\big(\bff_1(t_0,t_{2n},p)+

6673:  \bff_2(t_0,\dots,t_{2n-2},0,q)-\sum\limits_{i+j=n-1}

6674: \sum\limits_1^\infty \frac1{k} p_{k,2i}q_{k,2j}\big)\big|_L\;\;,\\

6675: \end{split}

6676: \end{equation}

6677: where

6678: \begin{equation}\label{eq:Lag-CPn}

6679: L=

6680: \begin{cases}

6681: p_{1,2i}=& z  \mathop{\sum\limits_{(s_1,\dots,s_{n-1})}}\limits_{

6682: \sum

6683: s_j(j-1)=i}\prod\limits_{j=1}^{n-1}\frac{t_{2j}^{s_j}}{s_j!}\;;\\

6684: p_{k,2i}=&0,\;\;\hbox{if}\;\; k>1\;; \\

6685: q_{k,2i}=&k\frac{\partial\bff_1}{\partial

6686: p_{k,2(n-i-1)}}(t_0,t_{2n}, p)\,.\\

6687: \end{cases}

6688: \end{equation}

6689: 

6690: Plugging expressions from (\ref{eq:Lag-CPn}) into equation

6691: (\ref{eq:CPn}) we get

6692: \begin{equation}\label{eq:CPn-Cn}

6693: \bff_{\C

6694: P^n}(t_0,\dots,t_{2n})=\bff_{\C^n}(t_0,t_{2n},p)\big|_{L_1}\;,

6695: \end{equation}

6696: where

6697: \begin{equation*}

6698: L_1=

6699: \begin{cases} p_{1,2i}&= z  \mathop{\sum\limits_{(s_1,\dots,s_{n-1})}}\limits_{

6700: \sum

6701: s_j(j-1)=i}\prod\limits_{j=1}^{n-1}\frac{t_{2j}^{s_j}}{s_j!}\;;\\

6702: p_{k,2i}&=0\;\;\hbox{if}\;\;

6703: k>1\,.\\

6704: \end{cases}

6705: \end{equation*}

6706: Indeed, two last terms in the formula (\ref{eq:CPn}) cancel each

6707: other (as it always happens when $\bff_2$ is linear with respect

6708: to $q$-variables).

6709:  For instance, for $n=1$ we get

6710: \begin{equation}

6711: \bff_1(t_0,t_2,p)=\big(\frac{t_0^2t_2

6712: }2+e^{t_2}p_1\big)\big|_{p_1=z}=\frac{t_0^2t_2}2+e^{t_2}z\,.

6713: \end{equation}

6714: 

6715: 

6716: For $n=2$ we have

6717: \begin{equation}

6718: L_1=

6719: \begin{cases}

6720:  p_{1,0}&= ze^{t_2}\,;\\

6721: p_{k,i}&= 0,\quad\hbox{ for all other}\quad k,i\,,\\

6722: \end{cases}

6723: \end{equation}

6724: and hence

6725: \begin{equation}\label{eq:CP2-final}

6726: \bff_{\C P^2}(t_0,t_2,t_4)=\bff_{\C^n}(t_0,t_4,

6727:  ze^{t_2},0,\dots)\,.

6728: \end{equation}

6729: 

6730: 

6731: \begin{remark} {\rm The method which we used  above

6732:  for computing of the rational

6733: potential of $\C P^n$, when  applied to an arbitrary

6734: symplectic manifold $W$,   allows us to express

6735:   $\bff_W$ through the potential of the affine part $W\setminus

6736:   M$.

6737:   The latter computation seems tractable when  the Weinstein manifold

6738:   $W\setminus M$ is

6739:   subcritical (see Section \ref{sec:splitting} above), i.e. when

6740:   its isotropic skeleton does not have Lagrangian cells.

6741:   On the other hand, when Lagrangian cells are present

6742:   this problem is related to central questions of Symplectic

6743:   topology.}

6744:   \end{remark}

6745: 

6746: 

6747: 

6748: \subsubsection{Satellites}\label{sec:satellites}

6749: Let $(V,\xi=\{\a=0\})$ be a  contact manifold, $(W

6750: = V\times \R, d(e^t\a))$  its symplectization, and   $J$   a

6751: compatible translation-invariant almost complex structure on $W$.

6752:    In this section  we will show that the

6753:  homological Poisson super-algebra $H_*(\fP ,  d^{\bh}  )$ comes equipped with some

6754: additional structures, rather unfamiliar in abstract Poisson

6755: geometry. Namely, the counting  of genus $g$ curves with a fixed

6756: complex structure and with a fixed configuration of $n$ points

6757: gives rise to an odd $n$-linear totally symmetric poly-form

6758: $\bh^{g,n}$ on the Poisson super-space $\bV$ underlying $\fP $. The

6759: poly-form descends well to the homology and thus yields another

6760: invariant of the contact structure which we call the genus-$g$

6761: $n$-point {\em satellite} of the Poisson structure.

6762: 

6763: Let us denote by

6764: $\overline{\cM_{g,m}(V)/\R}$   the compactified moduli space of stable connected

6765: $J$-holomorphic curves in $W$ which are characterized by the

6766: arithmetical genus $g$ and by the total number $m$ of punctures and

6767: marked points numbered somehow by the indices $1,...,m$ (see Section \ref{sec:compact} above). We

6768: emphasize that the moduli space in question contains equivalence

6769: classes of all such curves, and in particular, may have infinitely

6770: many connected components corresponding to different homotopy

6771: types of curves in $W$ and different numbering of the $m$

6772: markings. Let $\DM_{g,n}$ be the Deligne-Mumford compactification

6773: of the moduli space of genus $g$ Riemann surfaces with $n$ marked

6774: points. For any $g,n$ with $2g-2+n >0$ and $l\geq 0$ there is a

6775: natural {\em contraction map} $\ct :\overline{\cM_{g,n+l}(V)/\R}\to \DM_{g,n}$

6776: defined by forgetting the map to $W$ and the last $l$ markings

6777: followed by the contraction of those components of the curve which

6778: have become unstable.  Given a differential form

6779: $\tau$ on $\DM_{g,n}$ we will denote by $\ct^*\gt$ its  pull-back

6780:  to $\overline{\cM_{g,m}(V)/\R}$.

6781: 

6782: Let $u =(p,q,t)$ denote a point in $\bV$, that is $p,q$ and $t$ are

6783: (closed) differential forms on $\cP^{-},\cP^{+}$ and $V$

6784: respectively. We will denote $\ev_i^*{u}, \ i=1,...,m$, the

6785: pull-back by the evaluation map $$\ev_i: \overline{\cM_{g,m}(V)/\R}\to

6786: (\cP^{-}\cup \cP^{+}\cup V)$$ at the $i$-th marking. Let us emphasize the point that we

6787: are

6788: treating here  the marked points and punctures on equal footing.

6789: 

6790: Let $\gd u\in \bV$ be a tangent vector to $\bV$ at a point $u\in \bV$.

6791: We introduce the formal function

6792:  \begin{equation}

6793:  %\begin{split}

6794: \bh^{g,n}_{\gt}:=\frac{1}{n!}

6795: \sum_{l=0}^{\infty}\frac{1}{l!}\int\limits_{\overline{\cM_{g,n+l}(V)/\R }}\

6796: \ct^*{\gt}\w \ev_1^*\gd u\w...\w\ev^*_n\gd

6797: u\w\ev_{n+1}^*u\w...\w\ev_{n+l}^*u .

6798: %\end{split}

6799: \end{equation}

6800:  It is a super-symmetric  $n$-linear

6801: form in $\gd u$ with coefficients depending on the application

6802: point $u$.

6803: 

6804: Let $d^{\bh}(f)$ denote the Lie derivative of a tensor field $f$

6805:  along the odd Hamiltonian

6806: vector field $d^{\bh}$ on $\bV$ with the Hamilton function $\bh$.

6807: 

6808: \begin{proposition}

6809: Let $\gt $ be a top degree form on $\DM_{g,n}$. Then

6810: $d^{\bh}(\bh_{\gt}^{g,n})=0$. If the top degree form $\gt=d\a$ is exact

6811: then $\bh^{\gt}_{g,n}=d^{\bh}( \bh^{\a}_{g,n})$. In particular, the

6812: tensor field $\bh^{g,n}_{\gt}$ descends to the homology    algebra

6813: $H_*(\fP, \p)$ into a satellite which depends only on the total

6814: volume of $\gt$.

6815: \end{proposition}

6816: 

6817: This follows from the Stokes formula applied to

6818: $\bh^{d\gt}_{g,n}=0$ and respectively  to $\bh^{d\a}_{g,n}$.

6819: Codimension $1$ boundary strata of the moduli space $\cM_{g,m}(V)/\R$

6820: are formed by  stable curves of height $2$. Most of the strata do not contribute to

6821: the Stokes formula since they are mapped by the contraction map to

6822: complex codimension $1$ strata of the Deligne-Mumford space

6823: $\DM_{g,n}$, where $\gt$ and   $\a$ restrict to $0$ for

6824: dimensional reasons. Exceptions to this rule occur only if one of

6825: the two curves which form the stable curve is to be contracted. It is therefore a

6826: sphere with  glued to the other level of the stable curve along at precisely

6827: one end, and which have at most one marked points or ends with the index $\leq n$,

6828:  and with any

6829: number of ends or marked points with indices $>n$. Contributions

6830: of such curves to the Stokes formula is expressed bi-linearly via

6831: the $1$-st or $2$-nd derivatives of the Hamilton function $\bh$

6832: and the satellite. It is easy to see that the whole expression is

6833: interpreted correctly as the  Lie derivative of the tensor field

6834: $\bh^{\gt}_{g,n}$ along the Hamiltonian vector field $d^{\bh}$. \qed

6835: 

6836: 

6837: \medskip

6838: 

6839: We will assume further on that $\gt $ is normalized to the total

6840: volume $1$ and will often drop it from the notation for the

6841: satellite $\bh^{g,n}$.

6842: 

6843: \medskip

6844: 

6845: Let us consider now a directed symplectic  cobordism

6846: $W=\ora{V_-V_+}$ between two contact boundaries $V_{\pm}$. Then we

6847: have the Hamilton function $\widehat\bh

6848: =\bh^{+}-\bh^{-}$ and the satellites

6849: $\widehat\bh^{g,n}=\big(\bh^{g,n}\big)^+-

6850: \big(\bh^{g,n}\big)^-$ defined  as elements of

6851: the algebra $\widehat \fL$, which in the case when the cobordism

6852: is a concordance just equal to the tensor product of the Poisson

6853: algebras $\fP_{\pm}$. Also, we have the potential $\bff

6854: (p_{-},q_{+},t) $ counting rational $J$-holomorphic curves in $W$

6855: which defines a Lagrangian correspondence between $\fP^{\pm}$

6856: invariant under the vector field $ d^{\widehat{\bh}}$ with the Hamilton function

6857: $\widehat\bh$. Finally, using the moduli spaces $\cM_{g,m}(W)$ of

6858: $J$-holomorphic curves in the cobordism, we can introduce the

6859: satellites $\bff^{g,n}_{\gt}$ as symmetric $n$-forms on the space

6860: $(p_{-},q_{+},t)$-space parameterizing the Lagrangian

6861: correspondence. Then the arguments similar to the above proof of

6862: the proposition, but applied this time to $\bff^{g,n}_{d\gt}=0$,

6863: show that {\em the restriction of $\widehat\bh_{\gt}^{g,n}$ to the

6864: Lagrangian correspondence defined by $\bff$ coincides with the Lie

6865: derivative of $\bff^{g,n}_{\gt}$ along the vector field $d^{\widehat\bh}$

6866: restricted to the Lagrangian correspondence}

6867: (comp. Theorem \ref{thm:RSFT-cobordism} above).

6868: In this sense the

6869: Lagrangian correspondences defined by symplectic cobordisms

6870: preserve the satellite structures defined by $\big(\bh^{g,n}\big)^{\pm}$ on

6871: the homology $H_*(\fP_{\pm},d^{\bh^{\pm}})$. In particular, {\em the

6872: satellite structures of a contact manifold $V$ on the homology

6873: $H_*(\fP,d^{\bh})$ depend only on the contact structure.}

6874: 

6875: \medskip

6876: 

6877: The following discussion is the first steps in the study of the

6878: geometric structure defined by the satellites.

6879: 

6880: Let $\bh^{g,n}_{\a_1,...,\a_n}$ denote components of the satellite

6881: tensors on $\fP$. Using the Poisson tensor $\pi^{\mu\nu}$ we can

6882: couple two satellites with respect to some indices:

6883: $$\bh^{g',n'+1}_{...\mu}\pi^{\mu\nu}\bh^{g'',n''+1}_{\nu ...}.$$

6884: Similarly, we can couple two indices in $\bh^{g-1,n+2}$ with two

6885: indices in the $2$-nd differential $\gd^2\bh $ of the Hamilton

6886: function $\bh$.

6887: 

6888: \begin{proposition} If $g=g'+g''>0$ then the

6889: coupling of $\bh^{g',n'}$ and $\bh^{g'',n''}$ is a Lie derivative

6890: along $\p$ and thus vanishes in the homology $H_*(\fP ,\p )$.

6891: Similarly, the coupling of $\bh^{g-1,n+2}$ with $\gd^2\bh$

6892: vanishes in the homology $H_*(\fP,\p)$. \end{proposition}

6893: 

6894: The proof is based on some famous but non-trivial property of the

6895: spaces $\DM_{g,n}$ with $g>0$ to have complex codimension one

6896: strata homologically independent. Such strata correspond to

6897: different ways of cutting a $(g,n)$-surface along one circle and

6898: can be identified either with $\DM_{g',n'+1}\times

6899: \DM_{g'',n''+1}$ where $g'+g''=g, n'+n''=n$ or with

6900: $\DM_{g-1,n+2}$. The independence property implies that a volume

6901: form $\gt$ on the stratum, say  $\gt'\otimes\gt''$ in the first

6902: case, can be obtained as the restriction of a closed codegree two

6903: form $\go$ on $\DM_{g,n}$ which have exact (or even zero, for

6904: suitable choices of $\gt$) restrictions to all other

6905: codimension-$1$ strata in $\DM_{g,n}$. Applying the Stokes formula

6906: to $0=\bh^{g,n}_{d\go}$ we find that the coupling of

6907: $\bh^{g',n'+1}_{\gt'}$ and $\bh^{\gt'',n''+1}_{\gt''}$ (or --- in

6908: the second case --- of $\bh^{g-1,n+2}_{\gt}$ and $\gd^2\bh$)

6909: equals $d^{\bh}( \bh^{g,n}_{\go})$.

6910: 

6911: \medskip

6912: 

6913: \begin{remark} {\em

6914: To the contrary, coupling $\bh^{0,3}$ with itself via one index is

6915: not, in general, a $d^{\bh}$-derivative, but instead the following

6916: triple sum is:

6917: % I am afraid, the signs might be wrong

6918: \begin{equation*}

6919: \begin{split}

6920: & \bh^{0,3}_{\a\gb\mu}\pi^{\mu\nu}\bh^{0,3}_{\nu\g\d}+

6921:    (-1)^{(\deg{\a}+\deg{\gb})\deg{\g}}

6922:    \bh^{0,3}_{\g\a\mu}\pi^{\mu\nu}\bh^{0,3}_{\nu\gb\d}+\\

6923:   & (-1)^{\deg{\a}(\deg{\gb}+\deg{\g})}

6924:    \bh^{0,3}_{\gb\g\mu}\pi^{\mu\nu}\bh^{0,3}_{\nu\a\d} \equiv 0

6925:   \, .\\

6926:    \end{split}

6927:    \end{equation*}

6928: This follows from the property of the $3$ boundary strata in

6929: $\DM_{0,4}$ to represent the same homology class (use the Stokes

6930: formula for $\go =1$). In fact $\bh^{0,3}$ coincides with the

6931: $3$-rd differential $\gd^3\bh /6$ of the Hamilton function, and

6932: the above Jacobi-like identity can be derived by $4$

6933: differentiations of $\{ \bh ,\bh \}=0$ in $\a,\gb,\g,\d$. One can

6934: interpret the integrability property $(d^{\bh})^2=0$ of the odd vector

6935: field $d^{\bh}$ on $\bV$ as a homotopy Lie super-algebra structure on

6936: $\Pi \bV^*$, the dual space with changed parity. The identity in

6937: question corresponds to the Jacobi identity for the remnant Lie

6938: super-algebra structure in homology. }

6939:  \end{remark}

6940: 

6941: 

6942: It is sometimes convenient to extend the definition of genus $0$

6943: satellites to unstable values of $n$ by $\bh^{0,n}=\gd^n\bh /n!$

6944: for $n=0,1,2$. Also, one can define the function $\bh^{1,0}$ at

6945: least locally as a potential for $\bh^{1,1}$, using the following

6946: 

6947: \begin{proposition} The differential $1$-form $\bh^{1,1}$ is closed.

6948: \end{proposition}

6949: 

6950: Indeed, the partial derivatives $\gd_{\mu}\bh^{1,1}_{\nu}$ and

6951: $\gd_{\nu}\bh^{1,1}_{\mu}$ are identified with the satellites

6952: $(\bh^{1,2}_{\go})_{\mu\nu}$ corresponding to the $2$-form $\go $

6953: on $\DM_{1,2}$ pulled-back from $\DM_{1,1}$ by forgetting the

6954: $1$-st and respectively the $2$-nd marked point. But the two maps

6955: $\DM_{1,2}\to \DM_{1,1}$ coincide.

6956: 

6957: \medskip

6958: 

6959: It would be interesting to study other general properties of

6960: satellites which may depend on more sophisticated geometry of

6961: Deligne-Mumford compactifications. For instance, what can be said

6962: about Poisson brackets among the functions $\bh^{g,0}$?

6963: 

6964: \medskip

6965: 

6966: We complete the section with the computation of the satellites in the

6967: example $V=S^1$. Let $t=t_01+t_1d\phi $ denote the general

6968: harmonic form on $S^1$, $\gd t=\gt_0 1+\gd t_1\d\phi $,

6969: $u(x)=t_0+\sum p_ke^{ikx}+ q_ke^{-ikx}$, $\gd u = \gd t_0+\sum \gd

6970: p_ke^{ikx}+\gd q_k e^{-ikx}$.

6971: 

6972: \begin{proposition}

6973: For $2g-2+n \geq 0$ we have

6974: \[ \bh^{g,n+1}\ =\ \frac{\gd t_1}{2\pi\ n!}\int\limits_0^{2\pi} \ (u_{xx})^g \

6975: (\gd u)^n \ dx .\]

6976: \end{proposition}

6977: 

6978: Let us begin with the remark that the formula does not (and

6979: cannot) contain $t=t_0+t_1\phi$ because $\deg t < 2$, and

6980: therefore pushing forward from $\cM_{g,m+1}(V)/\R \to

6981: \cM_{g,m}(V)/\R$ by forgetting the corresponding marked point

6982: would send $t$ to $0$. Exceptions to this rule could occur only if

6983: $\cM_{g,m}(V)$ were ill-defined, that is only in the case of

6984: constant maps with "unstable" indices, $2g-2+m\leq 0$, which has

6985: no effect on the satellites with "stable" indices. On the other

6986: hand the factor $\d t_1$ is (and must be) present in the formula

6987: since the dimension of the moduli spaces is odd. With this

6988: information in mind, the enumerative question equivalent to

6989: computation of the satellites can be described as follows. On a

6990: Riemann surface $\Sigma$ of genus $g$, we are given a divisor $D$

6991: of $n$ distinct points with (possibly zero) multiplicities

6992: $m_1,...,m_n$. We have to count the divisors $l_1P_1+...+l_gP_g$

6993: which in the sum with $D$ form the divisor of a rational function.

6994: (In particular, the degree $\sum m_i+\sum l_j$ of the total

6995: divisor must vanish.)   The answer to this question is equal to

6996: the degree of the Abel-Jacobi map $\Sigma^g\to J_{\Sigma}$ defined

6997: by integration of holomorphic differentials

6998: $\vec{\go}=(\go_1,...,\go_g)$ on $\Sigma$ as

6999: \[ (P_1,...,P_g)\mapsto l_1\int^{P_1}\vec{\go}+...+l_g \int^{P_g}\vec{\go}

7000: .\] When the multiplicities $(l_1,...,l_g)=(1,...,1)$, the degree

7001: equals $g!$ (it is well-known that $S^g\Sigma^g\to J_{\Sigma}$ is

7002: a bi-rational isomorphism). For arbitrary $(l_1,...l_g)$ the

7003: Abel-Jacobi map has the Jacobi matrix $[l_j \go_i (P_j)]$. Thus

7004: the degree equals $l_1^2...l_g^2 g!$. Taking these answers as the

7005: coefficients in the generating function on the variables $t_0,

7006: p_l, q_{-l}$ corresponding to $l=0,l>0$ and $l<0$ we arrive at the

7007: factor $u_{xx}^g$. The other factor $(\gd u)^n/n!$ is similarly

7008: accountable for all possible choices of multiplicities

7009: $m_1,...,m_n$ in the divisor $D$. The contour integration of the

7010: product couples the choices with $m_1+...+m_n+l_1+...+l_g = 0$.

7011: 

7012: 

7013: \begin{thebibliography}{99}

7014: 

7015: \bibitem{AbbasHofer}

7016: C.~Abbas and H.~Hofer,

7017:   Holomorphic curves and global questions in contact geometry,

7018:   to appear in Birkh\"auser.

7019: \bibitem{Arnold-conjecture} V.I. Arnold,        {Sur une propri\'et\'e topologique des

7020: applications globalement canoniques de la m\'echanique classique},

7021:       {\it C.\ R.\ Acad. Paris}, 261(1965), 3719--3722.

7022: \bibitem{Arnold-Maslov} V. I. Arnold, On a characteristic class entering in quantization

7023: conditions,

7024: {\it Funct. Anal. and Applic.}, 1(1967), 1--14.

7025: \bibitem{Arnold-steps}  V. I. Arnold,       {First steps in symplectic topology},

7026: {\it Russian Math. Surveys}, 41(1986), 1--21.

7027: 

7028: \bibitem{Bennequin} D. Bennequin,

7029: Entrelacements et \'equations de Pfaff, {\it Ast\'erisque}, 106--107(1983).

7030: \bibitem{Biran-Cieliebak} P. Biran and K. Cieliebak, Symplectic

7031: topology on subcritical manifolds, preprint 2000.

7032: \bibitem{Brieskorn} E. Brieskorn, Beispiele zur

7033: Differentialtopologie von Singularit\"aten, {\it Invent. Math.},

7034: {\bf 2}(1966), 1--14.

7035: \bibitem{Chekanov} Yu. Chekanov, Differential algebra of a Legendrian link.

7036: Preprint 1997.

7037: \bibitem{Conley-Zehnder} C. Conley and E. Zehnder,

7038:         {The {B}irkhoff--{L}ewis fixed point theorem and a conjecture of {V}.{I}.

7039: {A}rnold},

7040:     {\it Invent. Math.}, 73(1983), 33--49.

7041:     \bibitem{Donaldson-invariants} S.K. Donaldson,

7042:     Polynomial invariants for smooth four-manifolds,

7043: {\it Topology}, {\bf 29}(1990),  257--315.

7044: 

7045:      \bibitem{Donaldson} S.K. Donaldson, Symplectic

7046: submanifolds and almost-complex geometry, {\it J. Diff. Geom.},

7047: {\bf 44}(1996), 666-705.

7048: \bibitem{Eliash-ICM} Y. Eliashberg, Invariants in contact

7049: topology, {\it Proc. of ICM-98, Berlin}, Doc. Math., 1998,

7050: 327-338.

7051: 

7052: \bibitem{Eliash-Stein} Y. Eliashberg, Topological characterization of Stein manifolds

7053: of complex dimension $>2$, {\it Int. J. of Math}, {\bf 1}(1991),

7054: 29--46.

7055: 

7056:  \bibitem{Eliash-Gromov-convex} Y. Eliashberg and M. Gromov, Convex symplectic manifolds, {\it Proc.

7057: of Symp. in Pure Math.}, vol.52(1991), Part 2, 135--162.

7058:  \bibitem{EHS} Y. Eliashberg, H. Hofer and S. Salamon, Lagrangian intersections

7059: in contact geometry, {\it Geom. and Funct. Anal.}, 5(1995), 244--269.

7060: \bibitem{Etnyre-Sabloff} J. Etnyre and J. Sabloff,

7061: Coherent Orientations and Invariants of Legendrian Knots, preprint

7062: 2000.

7063:    \bibitem{EH-3ball} Y. Eliashberg, H. Hofer,  A Hamiltonian characterization

7064:    of the three-ball, {\it Differential Integral Equations},

7065:    {\bf 7}(1994),  1303--1324.

7066: 

7067: \bibitem{Floer} A. Floer,

7068: The unregularised gradient flow of the symplectic action,

7069:       {\it Comm. Pure Appl. Math.},

7070:      {\bf 41}(1988),

7071:   775-813.

7072: \bibitem{Floer-Hofer} A. Floer,  H. Hofer, Coherent

7073: orientations for periodic orbit

7074: problems in symplectic geometry, {\it Math. Z.}, {\bf  212}(1993), 13--38.

7075: \bibitem{Fukaya-Ono}

7076:  K. Fukaya and K. Ono, Arnold conjecture and Gromov-Witten invariants, preprint 1996.

7077:  \bibitem{Fukaya-Ono2} K. Fukaya,  K. Ono,  Arnold conjecture and Gromov-Witten invariant

7078: for general symplectic

7079: manifolds, in {\it The Arnoldfest (Toronto, ON, 1997)}, 173--190, Fields

7080: Inst. Commun., 24, Amer. Math. Soc., Providence, RI,

7081: 1999.

7082: 

7083:   \bibitem{FKO3}   K.Fukaya, K.Ono, Y.-G.Oh and H.Ohta, to appear.

7084:   \bibitem{Fulton-Pand} W. Fulton and R. Pandharipande, Notes on stable

7085:   maps and quantum cohomology, {\it Proc. of Symp. in Pure Math.},

7086:   {\bf 62}(1995), Part 2, 45--96

7087:   \bibitem{Gathmann}A. Gathmann,

7088:   Absolute and relative Gromov-Witten invariants of very ample

7089:   hypersurfaces, preprint 1999.

7090:   \bibitem{Getzler} E. Getzler, Topological recursion relations in genus $2$, in {\it Integrable systems

7091:    and algebraic geometry} (Kobe/Kyoto, 1997), 73--106, World Sci. Publishing  1998.

7092: \bibitem{Giroux} E. Giroux, Une structure de contact,  m\^eme tendue est plus ou

7093: moins tordue,

7094: {\it Ann. Scient. Ec. Norm. Sup.}, {\bf  27}(1994), 697--705.

7095: 

7096: \bibitem{Givental-toric} A. Givental,

7097: A symplectic fixed point theorem for

7098:  toric manifolds,

7099:  {\it The Floer memorial volume}, 445--481,

7100:  Progr. Math.,

7101:  133, Birkhäuser, Basel, 1995.

7102:  \bibitem{Givental-Maslov} A. Givental,

7103:  Nonlinear generalization of the Maslov index, {\it Adv. Soviet Math.}, {\bf 1}(1990),

7104:   71--103.

7105: \bibitem{Givental} A. Givental.

7106: Homological geometry and mirror symmetry. Proc. Int. Congress of

7107: Math., Z\"urich-1994. Birkh\"auser, 1995, v. 1, 472-480.

7108: 

7109: 

7110: \bibitem{Givental2} A. Givental.

7111: Homological geometry I: Projective hypersurfaces. Selecta Math.,

7112: New series 1 (2), 1995, 325-345.

7113: 

7114: \bibitem{Givental3}A. Givental.

7115:  Equivariant Gromov - Witten invariants. Intern. Math. Res. Notices, 1996,

7116: No.13, 613 - 663.

7117: \bibitem{Givental1} A. Givental and B. Kim.

7118: Quantum cohomology of flag manifolds and Toda lattices. Commun.

7119: Math. Phys. 168 (3), 1995, 609-641.

7120: 

7121: \bibitem{Graber-Pand} T. Graber and R. Pandharipande, Localization

7122: of virtual classes, {\it Invent. Math.},{\bf  135}(1999),    487--518.

7123: \bibitem{Gray} J.W. Gray, Some global properties of contact structures,

7124: {\it Annals of Math.}, {\bf 69}(1959), 421--450.

7125: \bibitem{Gromov-holomorphic} M. Gromov, Pseudo-holomorphic curves in symplectic manifolds,

7126: {\it Invent. Math.}, {\bf 82}(1985), 307--347.

7127: \bibitem{Gro-PDR} M. Gromov, Partial Differential Relations, Springer-Verlag,

7128: 1986.

7129: \bibitem{Hofer-Weinstein} H. Hofer, Pseudo-holomorphic curves and Weinstein conjecture

7130: in dimension three, {\it Invent. Math.}, {\bf 114}(1993), 515--563.

7131:  \bibitem{Hofer-Wysocki-Zehnder}    H. Hofer, K. Wysocki and E. Zehnder, Properties of

7132: pseudo-holomorphic curves in symplectisations. I. Asymptotics.

7133: {\it  Ann. Inst. H. Poincar\'e, Anal. Non Lin\'eaire}, {\bf

7134: 13}(1996), 337--379.

7135: 

7136: \bibitem{HoferWysockiZehnder:94d}

7137: H.~Hofer, K.~Wysocki, and E.~Zehnder,

7138:   The dynamics on a strictly convex energy surface in {${\bf R}^4$},

7139:   {\it Annals of Math.}, {\bf 148}(1998),  197-289.

7140: \bibitem{HoferWysockiZehnder:99}

7141: H. Hofer, K. Wysocki, and E. Zehnder,

7142:   Finite Energy Foliations of Tight Three-Spheres and Hamiltonian

7143: Dynamics, preprint 1999.

7144: \bibitem{Ionel} E.-N. Ionel, Topological recursive relations in

7145: $H^{2g}(M_{g,n})$, preprint 1999.

7146: \bibitem{Ionel-Parker2} E.-N. Ionel and T.H. Parker

7147:  Gromov-Witten invariants of symplectic sums.{\it  Math. Res. Lett.}

7148:  {\bf 5} (1998),  563--576.

7149:  \bibitem{Ionel-Parker} E.-N. Ionel and T.H. Parker,

7150:  Relative Gromov-Witten Invariants, preprint 1999.

7151: \bibitem{Kollar} J. Kollar,

7152: Rational curves on algebraic varieties, Springer-Verlag, 1996.

7153: 

7154: \bibitem{Kontsevich-quant} M. Kontsevich, Deformation quantization of Poisson manifolds, I,

7155: preprint 1997.

7156: \bibitem{Kontsevich-CP2} M. Kontsevich,  Enumeration of rational

7157: curves via torus action, in {\it The Moduli Space of Curves}, R.

7158: Dijgraaf, C. Faber and G. van der Geer eds., Birkhauser,1995,

7159: 335--368.

7160: \bibitem{Konts-Manin} M. Kontsevich, Yu. Manin,

7161:  Gromov-Witten classes, quantum cohomology, and enumerative

7162:  geometry, {\it Commun.Math.Phys.}, {\bf  164} (1994) 525-562.

7163: 

7164: \bibitem{Li-Tian} J. Li and G. Tian, Virtual moduli cycles and Gromov-Witten invariants of

7165: general symplectic

7166: manifolds, in {\it Topics in symplectic $4$-manifolds (Irvine, CA, 1996)}, 47--83,

7167: First Int. Press Lect. Ser., I, Internat. Press,

7168: Cambridge, MA, 1998.

7169: \bibitem{Liu-Tian} G. Liu and G. Tian, Floer homology and Arnold conjecture,

7170: {\it J. Diff. Geom.}, {\bf 49}(1998),  1--74.

7171: \bibitem{Lutz} R. Lutz,  Structures de contact sur les fibr\'es principaux en cercles de

7172: dimension $3$

7173: {\it  Ann. Inst. Fourier},   XXVII, {\bf 3}(1977), 1--15.

7174:               \bibitem{Martinet}J. Martinet, Formes de contact sur les vari\'et\'es de dimension $3$,

7175: {\it  Lecture Notes in Math.,} {\bf 209}(1971), 142--163.

7176: \bibitem{McDuff} D. McDuff, The virtual moduli cycle, in{\it  Northern California Symplectic

7177: Geometry Seminar}, 73--102,

7178: Amer. Math. Soc. Transl. Ser. 2, 196, Amer. Math. Soc., Providence, RI,

7179: 1999.

7180:     \bibitem{Mori-Miayoka} Y. Myayoka and S. Mori, A numerical

7181:     criterion for uniruleness, {\it Annals of Math}, {\bf

7182:     124}(1986), 65--69.

7183:     \bibitem{Morita} S. Morita, A topological classification of

7184:     complex structures on $S^1\times \Sigma^{2n-1}$,

7185:     {\it Topology}, {\bf 14}(1975), 13--22.

7186: \bibitem{Ng} L. Ng, On Invariants of Legendrian Knots, preprint

7187: 2000.

7188: \bibitem{Robbin-Salamon} J.  Robbin, D. Salamon,The Maslov index for

7189: paths, {\it Topology}, {\bf  32}(1993), 827--844.

7190: 

7191: \bibitem{Ruan} Y. Ruan,  Topological sigma model and

7192: Donaldson-type invariants in Gromov theory. {\it Duke Math. J.,},

7193: {\bf 83}(1996),  461--500.

7194:             \bibitem{Ruan-Li} Y. Ruan and A.-M.

7195:             Li,

7196:              Symplectic surgery and Gromov-Witten invariants of Calabi-Yau 3-folds

7197:              I, preprint 1999

7198:     \bibitem{Ruan-virtual} Y. Ruan,  Virtual

7199:     neighborhoods and pseudo-holomorphic curves, {\it Proceedings

7200:     of 6th G\"okova Geometry-Topology Conference, Turkish J. Math.}, {\bf  23} (1999),

7201:     161--231.

7202: \bibitem{Ruan-Tian} Y. Ruan and G. Tian, A mathematical theory of

7203: quantum cohomology, {\it J. Diff. Geom.}, {\bf 42}(1995),259--367.

7204: \bibitem{Siebert} B. Siebert,

7205: Gromov-Witten invariants of general symplectic manifolds, preprint 1997.

7206:    \bibitem{Salamon:Floer} D. Salamon. Lectures on Floer homology,

7207: in ``Symplectic Geometry and Topology'', IAS/Park City Mathematics Series, v. 7, AMS/IAS

7208: 1999, 144 -- 229.

7209:  \bibitem{Ustilovsky}  I. Ustilovsky, Infinitely many contact

7210:  structures on $S^{4m+1}$, {\it Int. Math. Res. Notices}, {\bf

7211:  14}(1999), 781--791.

7212: 

7213:  \bibitem{Vakil} R. Vakil, The enumerative geometry of rational

7214:  and elliptic curves in projective space, preprint 1997.

7215:  \bibitem{Viterbo} C. Viterbo, in preparation.

7216:  \bibitem{Weinstein}  A. Weinstein,

7217:  On the hypotheses of Rabinowitz's periodic orbits theorems

7218: {\it J. Diff. Eq. }, {\bf 33}(1979), 353--358.

7219:  \bibitem{Witten1} E. Witten,   Supersymmetry and Morse theory,

7220:  {\it J. Diff. Geom.}, {\bf 17}(1982), 661--692.

7221:  \bibitem{Witten2} E. Witten, Two-dimensional gravity and

7222:  intersection theory on moduli space, {\it Surveys in Diff.

7223:  Geom.}, {\bf 1}(1991), 243--310.

7224: \bibitem{MLYau} M.-L. Yau, Contact homology of subcritical Stein

7225: manifolds, thesis, Stanford University, 1999.

7226: 

7227: 

7228: 

7229:            \end{thebibliography}

7230: 

7231: \end{document}

7232: %%%% To be continued        1)

7233: 

7234: 2) Fukaya, Kenji; Ono, Kaoru Arnold conjecture and Gromov-Witten

7235: invariant.

7236: 

7237: 3) Li, Jun; Tian, Gang

7238: 

7239: 4)   dg-ga/9608005

7240: 

7241:     Notes: LaTeX2e with AMS-fonts, 73 pages, 3 figures, full revision

7242:            12/1997

7243: 

7244: 5)

7245: