1: 

2: \documentclass[10pt,twoside]{article}

3: \usepackage{Latex-document}

4: \usepackage{amsmath,amsthm,amsfonts,eucal,epsfig}

5: \usepackage[matrix,arrow]{xy}

6: 

7: \newcommand{\R}{\mathbb{R}}

8: \newcommand{\Z}{\mathbb{Z}}

9: \newcommand{\Q}{\mathbb{Q}}

10: \newcommand{\C}{\mathbb{C}}

11: \newcommand{\N}{\mathbb{N}}

12: 

13: \newcommand{\iso}{\cong}           %isomorphism sign

14: \newcommand{\htp}{\simeq}

15: \newcommand{\smooth}{C^\infty}

16: \newcommand{\CP}[1]{\C {\mathrm P}^{#1}}

17: 

18: \newcommand{\re}{\mathrm{re}}

19: \newcommand{\im}{\mathrm{im}}

20: \renewcommand{\ker}{\mathrm{ker}}

21: \newcommand{\coker}{\mathrm{coker}}

22: \newcommand{\mymod}{\quad\text{mod }}

23: \newcommand{\Hom}{\mathrm{Hom}}

24: \newcommand{\End}{\mathrm{End}}

25: 

26: \renewcommand{\o}{\omega}

27: %

28: %------------- Theorem environments ----------------------

29: %

30: \newtheoremstyle{my}{1.5em}{0.5em}{\em}{}{\sc}{.}{0.5em}{}

31: \newtheoremstyle{mydef}{1.5em}{0.5em}{}{}{\sc}{.}{0.5em}{}

32: % #1 = name

33: % #2 = preskip

34: % #3 = postskip

35: % #4 = bodyfont

36: % #5 = noindent?

37: % #6 = headfont

38: % #7 = headpunct, e.g. "."

39: % #8 = labelsep (between label and statement}

40: % #9 = apparently overrides the whole header

41: 

42: \theoremstyle{my}

43: \newtheorem{thm}{Theorem}

44: \newtheorem{theorem}[thm]{Theorem}

45: \newtheorem{conjecture}[thm]{Conjecture}

46: \newtheorem{examples}[thm]{Examples}

47: %\theoremstyle{mydef}

48: \newtheorem{assumption}[thm]{Assumption}

49: 

50: \parskip1em

51: \renewcommand{\theequation}{\arabic{equation}}

52: 

53: \newcommand{\A}{\mathcal A}

54: \newcommand{\E}{\mathcal E}

55: \newcommand{\F}{\mathcal F}

56: \newcommand{\K}{\mathcal K}

57: \renewcommand{\O}{\mathcal O}

58: 

59: \begin{document}

60: 

61: \title{Fukaya categories and deformations}

62: \author{Paul Seidel}

63: \date{June 15, 2002}

64: \maketitle

65: 

66: Soon after their first appearance \cite{fukaya93}, Fukaya categories

67: were brought to the attention of a wider audience through the

68: homological mirror conjecture \cite{kontsevich94}. Since then Fukaya

69: and his collaborators have undertaken the vast project of laying down

70: the foundations, and as a result a fully general definition is

71: available \cite{fooo,fukaya01b}. The task that symplectic geometers

72: are now facing is to make these categories into an effective tool,

73: which in particular means developing more ways of doing computations

74: in and with them.

75: 

76: For concreteness, the discussion here is limited to projective

77: varieties which are Calabi-Yau (most of it could be carried out in

78: much greater generality, in particular the integrability assumption

79: on the complex structure plays no real role). The first step will be

80: to remove a hyperplane section from the variety. This makes the

81: symplectic form exact, which simplifies the pseudo-holomorphic map

82: theory considerably. Moreover, as far as Fukaya categories are

83: concerned, the affine piece can be considered as a first

84: approximation to the projective variety. This is a fairly obvious

85: idea, even though its proper formulation requires some algebraic

86: formalism of deformation theory. A basic question is the

87: finite-dimensionality of the relevant deformation spaces. As

88: Conjecture \ref{conj:hh} shows, we hope for a favourable answer in

89: many cases. It remains to be seen whether this is really a viable

90: strategy for understanding Fukaya categories in interesting examples.

91: 

92: Lack of space and ignorance keeps us from trying to survey related

93: developments, but we want to give at least a few indications. The

94: idea of working relative to a divisor is very common in symplectic

95: geometry; some papers whose viewpoint is close to ours are

96: \cite{ionel-parker98, li-ruan98, eliashberg-givental-hofer02,

97: ozsvath-szabo01}. There is also at least one entirely different

98: approach to Fukaya categories, using Lagrangian fibrations and Morse

99: theory \cite{fukaya-oh98, kontsevich-soibelman00, fukaya02}. Finally,

100: the example of the two-torus has been studied extensively

101: \cite{polishchuk-zaslow98}.

102: 

103: {\em Acknowledgements.} Obviously, the ideas outlined here owe

104: greatly to Fukaya and Kontsevich. The author is equally indebted to

105: Auroux, Donaldson, Getzler, Joyce, Khovanov, Smith, and Thomas (an

106: incomplete list), all of whom have influenced his thinking

107: considerably. The preparation of this talk at the Institute for

108: Advanced Study was supported by NSF grant DMS-9729992.

109: 

110: \section{Symplectic cohomology}

111: 

112: We will mostly work in the following setup:

113: 

114: \begin{assumption} \label{as:projective}

115: $X$ is a smooth complex projective variety with trivial canonical

116: bundle, and $D$ a smooth hyperplane section in it. We take a suitable

117: small open neighbourhood $U \supset D$, and consider its complement

118: $M = X \setminus U$. Both $X$ and $M$ are equipped with the

119: restriction of the Fubini-Study K{\"a}hler form. Then $M$ is a

120: compact exact symplectic manifold with contact type boundary,

121: satisfying $c_1(M) = 0$.

122: \end{assumption}

123: 

124: Consider a holomorphic map $u: \Sigma \rightarrow X$, where $\Sigma$

125: is a closed Riemann surface. The symplectic area of $u$ is equal (up

126: to a constant) to its intersection number with $D$. When counting

127: such maps in the sense of Gromov-Witten theory, it is convenient to

128: arrange them in a power series in one variable $t$, where the $t^k$

129: term encodes the information from curves having intersection number

130: $k$ with $D$. The $t^0$ term corresponds to constant maps, hence is

131: sensitive only to the classical topology of $X$. Thus, for instance,

132: the small quantum cohomology ring $QH^*(X)$ is a deformation of the

133: ordinary cohomology $H^*(X)$.

134: 

135: As we've seen, there are only constant holomorphic maps from closed

136: Riemann surfaces to $M = X \setminus D$. But one can get a nontrivial

137: theory by using punctured surfaces, and deforming the holomorphic map

138: equation near the punctures through an inhomogeneous term, which

139: brings the Reeb dynamics on $\partial M$ into play. This can be done

140: more generally for any exact symplectic manifold with contact type

141: boundary, and it leads to the symplectic cohomology $SH^*(M)$ of

142: Cieliebak-Floer-Hofer \cite{cieliebak-floer-hofer95} and Viterbo

143: \cite{viterbo97a,viterbo97b}. Informally one can think of $SH^*(M)$

144: as the Floer cohomology $HF^*(M \setminus \partial M,H)$ for a

145: Hamiltonian function $H$ on the interior whose gradient points

146: outwards near the boundary, and becomes infinite as we approach the

147: boundary. For technical reasons, in the actual definition one takes

148: the direct limit over a class of functions with slower growth (to

149: clarify the conventions: our $SH^k(M)$ is dual to the $FH^{2n-k}(M)$

150: in \cite{viterbo97a}). The algebraic structure of symplectic

151: cohomology is different from the familiar case of closed $M$, where

152: one has large quantum cohomology and the WDVV equation. Operations

153: $SH^*(M)^{\otimes p} \rightarrow SH^*(M)^{\otimes q}$, for $p \geq 0$

154: and $q>0$, come from families of Riemann surfaces with $p+q$

155: punctures, together with a choice of local coordinate around each

156: puncture. The Riemann surfaces may degenerate to stable singular

157: ones, but only if no component of the normalization contains some of

158: the first $p$ and none of the last $q$ punctures. This means that if

159: we take only genus zero and $q = 1$ then no degenerations at all are

160: allowed, and the resulting structure is that of a Batalin-Vilkovisky

161: (BV) algebra \cite{getzler94}. For instance, let $M = D(T^*L)$ be a

162: unit cotangent bundle of an oriented closed manifold $L$. Viterbo

163: \cite{viterbo97b} computed that $SH^*(M) \iso H_{n-*}(\Lambda L)$ is

164: the homology of the free loop space, and a reasonable conjecture says

165: that the BV structure agrees with that of Chas-Sullivan

166: \cite{chas-sullivan99}.

167: 

168: Returning to the specific situation of Assumption

169: \ref{as:projective}, and supposing that $U$ has been chosen in such a

170: way that the Reeb flow on $\partial M$ becomes periodic, one can use

171: a Bott-Morse argument \cite{pozniak} to get a spectral sequence which

172: converges to $SH^*(M)$. The starting term is

173: \begin{equation} \label{eq:ss}

174: E_1^{pq} = \begin{cases}

175:  H^q(M) & p = 0, \\

176:  H^{q+3p}(\partial M) & p < 0.

177: \end{cases}

178: \end{equation}

179: It might be worth while to investigate this further, in order to

180: identify the differentials (very likely, a version of the relative

181: Gromov-Witten invariants \cite{ionel-parker98} for $D \subset X$).

182: But even without any more effort, one can conclude that each group

183: $SH^k(M)$ is finite-dimensional. In particular, assuming that

184: $dim_{\C}(X) > 2$ (and appealing to hard Lefschetz, which will be the

185: only time that we use any algebraic geometry) one has

186: \begin{equation} \label{eq:b2}

187: dim\, SH^2(M) \leq b_2(M) + b_0(\partial M) = b_2(X).

188: \end{equation}

189: 

190: 

191: \section{Fukaya categories}

192: 

193: $M$ (taken as in Assumption \ref{as:projective}) is an exact

194: symplectic manifold, and there is a well-defined notion of exact

195: Lagrangian submanifold in it. Such submanifolds $L$ have the property

196: that there are no non-constant holomorphic maps $u: (\Sigma,\partial

197: \Sigma) \rightarrow (M,L)$ for a compact Riemann surface $\Sigma$,

198: hence a theory of ``Gromov-Witten invariants with Lagrangian boundary

199: conditions'' would be trivial in this case. To get something

200: interesting, one removes some boundary points from $\Sigma$, thus

201: dividing the boundary into several components, and assigns different

202: $L$ to them. The part of this theory where $\Sigma$ is a disk gives

203: rise to the Fukaya $A_\infty$-category $\F(M)$.

204: 

205: The basic algebraic notion is as follows. An $A_\infty$-category $\A$

206: (over some field, let's say $\Q$) consists of a set of objects

207: $Ob\,\A$, and for any two objects a graded $\Q$-vector space of

208: morphisms $hom_\A(X_0,X_1)$, together with composition operations

209: \begin{align*}

210:  & \mu^1_\A : hom_\A(X_0,X_1) \longrightarrow hom_\A(X_0,X_1)[1], \\

211:  & \mu^2_\A : hom_\A(X_1,X_2) \otimes hom_\A(X_0,X_1) \longrightarrow

212:  hom_\A(X_0,X_2), \\

213:  & \mu^3_\A : hom_{\A}(X_2,X_3) \otimes hom_{\A}(X_1,X_2) \otimes

214:  hom_{\A}(X_0,X_1) \rightarrow \\ & \qquad \qquad \longrightarrow

215:  hom_{\A}(X_0,X_3)[-1], \quad \dots

216: \end{align*}

217: These must satisfy a sequence of quadratic ``associativity''

218: equations, which ensure that $\mu^1_\A$ is a differential, $\mu^2_\A$

219: a morphism of chain complexes, and so on. Note that by forgetting all

220: the $\mu^d_\A$ with $d \geq 3$ and passing to $\mu^1_\A$-cohomology

221: in degree zero, one obtains an ordinary $\Q$-linear category, the

222: induced cohomological category $H^0(\A)$ -- actually, in complete

223: generality $H^0(\A)$ may not have identity morphisms, but we will

224: always assume that this is the case (one says that $\A$ is

225: cohomologically unital).

226: 

227: In our application, objects of $\A = \F(M)$ are closed exact

228: Lagrangian submanifolds $L \subset M \setminus \partial M$, with a

229: bit of additional topological structure, namely a grading

230: \cite{kontsevich94,seidel99} and a $Spin$ structure \cite{fooo}. If

231: $L_0$ is transverse to $L_1$, the space of morphisms $hom_\A(L_0,L_1)

232: = CF(L_0,L_1)$ is generated by their intersection points, graded by

233: Maslov index. The composition $\mu_\A^d$ counts ``pseudo-holomorphic

234: $(d+1)$-gons'', which are holomorphic maps from the disk minus $d+1$

235: boundary points to $M$. The sides of the ``polygons'' lie on

236: Lagrangian submanifolds, and the corners are specified intersection

237: points; see Figure \ref{fig:polygon}. There are some technical issues

238: having to do with transversality, which can be solved by a small

239: inhomogeneous perturbation of the holomorphic map equation. This

240: works for all exact symplectic manifolds with contact type boundary,

241: satisfying $c_1 = 0$, and is quite an easy construction by today's

242: standards, since the exactness condition removes the most serious

243: problems (bubbling, obstructions).

244: %

245: \begin{figure}[htb] \begin{center}

246: \epsfig{file = polygon.eps}

247: \caption{\label{fig:polygon}}

248: \end{center} \end{figure}

249: 

250: It is worth while emphasizing that, unlike the case of Gromov-Witten

251: invariants, each one of the coefficients which make up $\mu^d_\A$

252: depends on the choice of perturbation. Only by looking at all of them

253: together does one get an object which is invariant up to a suitable

254: notion of quasi-isomorphism. To get something which is well-defined

255: in a strict sense, one can descend to the cohomological category

256: $H^0(\F(M))$ (which was considered by Donaldson before Fukaya's work)

257: whose morphisms are the Floer cohomology groups, with composition

258: given by the ``pair-of-pants'' product; but that is rather a waste of

259: information.

260: 

261: At this point, we must admit that there is essentially no chance of

262: computing $\F(M)$ explicitly. The reason is that we know too little

263: about exact Lagrangian submanifolds; indeed, this field contains some

264: of the hardest open questions in symplectic geometry. One way out of

265: this difficulty, proposed by Kontsevich \cite{kontsevich94}, is to

266: make the category more accessible by enlarging it, adding new objects

267: in a formal process, which resembles the introduction of chain

268: complexes over an additive category. This can be done for any

269: $A_\infty$-category $\A$, and the outcome is called the

270: $A_\infty$-category of twisted complexes, $Tw(\A)$. It contains the

271: original $A_\infty$-category as a full subcategory, but this

272: subcategory is not singled out intrinsically, and very different $\A$

273: can have the same $Tw(\A)$. The cohomological category $D^b(\A) =

274: H^0(Tw(\A))$, usually called the derived category of $\A$, is

275: triangulated (passage to cohomology is less damaging at this point,

276: since the triangulated structure allows one to recover many of the

277: higher order products on $Tw(\A)$ as Massey products). For our

278: purpose it is convenient to make another enlargement, which is

279: Karoubi or idempotent completion, and leads to a bigger

280: $A_\infty$-category $Tw^{\pi}(\A) \supset Tw(\A)$ and triangulated

281: category $D^\pi(\A) = H^0(Tw^{\pi}(\A))$. The main property of

282: $D^\pi(\A)$ is that for any object $X$ and idempotent endomorphism

283: $\pi: X \rightarrow X$, $\pi^2 = \pi$, there is a direct splitting $X

284: = im(\pi) \oplus ker(\pi)$. The details, which are not difficult,

285: will be explained elsewhere.

286: 

287: \section{Picard-Lefschetz theory}

288: 

289: We will now restrict the class of symplectic manifolds even further:

290: 

291: \begin{assumption} \label{as:lefschetz-fibre}

292: In the situation of Assumption \ref{as:projective}, suppose that $X$

293: is itself a hyperplane section in a smooth projective variety $Y$,

294: with $\K_Y \iso \O_Y(-X)$. Moreover, $X = X_0$ should be part of a

295: Lefschetz pencil of such sections $\{X_z\}$, whose base locus is $D =

296: X_0 \cap X_{\infty}$.

297: \end{assumption}

298: 

299: This gives a natural source of Lagrangian spheres in $M$, namely the

300: vanishing cycles of the Lefschetz pencil. Recall that to any

301: Lagrangian sphere $S$ one can associate a Dehn twist, or

302: Picard-Lefschetz monodromy map, which is a symplectic automorphism

303: $\tau_S$. The symplectic geometry of these maps is quite rich, and

304: contains information which is not visible on the topological level

305: \cite{seidel97,seidel98b,seidel99}. The action of $\tau_S$ on the

306: Fukaya category is encoded in an exact triangle in $Tw(\F(M))$, of

307: the form

308: \begin{equation} \label{eq:triangle}

309: \xymatrix{

310:  {L} \ar[rr] && {\tau_S(L)} \ar[dl]^-{[1]} \\

311:  & {\!\!\!\! HF^*(S,L) \otimes S \!\!\!\!} \ar[ul]

312: }

313: \end{equation}

314: for any $L$, and where the $\otimes$ is just a direct sum of several

315: copies of $S$ in various degrees. This is a consequence of the long

316: exact sequence in Floer cohomology \cite{seidel01}.

317: 

318: In the situation of Assumption \ref{as:lefschetz-fibre}, if we choose

319: a distinguished basis of vanishing cycles $S_1,\dots,S_m$ for the

320: pencil, the product of their Dehn twists is almost the identity map.

321: More precisely, taking into account the ``grading'' of the objects of

322: the Fukaya category, one finds that

323: \[

324: \tau_{S_1}\dots\tau_{S_m}(L) \iso L[2]

325: \]

326: where $[2]$ denotes change in the grading by 2. By combining this

327: trick with \eqref{eq:triangle} one can prove the following result:

328: 

329: \begin{thm} \label{th:generators}

330: $S_1,\dots,S_m$ are split-generators for $D^\pi(\F(M))$. This means

331: that any object of $Tw^\pi(\F(M))$ can be obtained from them, up to

332: quasi-isomorphism, by repeatedly forming mapping cones and idempotent

333: splittings.

334: \end{thm}

335: 

336: \section{Hochschild cohomology}

337: 

338: The Hochschild cohomology $HH^*(\A,\A)$ of an $A_\infty$-category

339: $\A$ can be defined by generalizing the Hochschild complex for

340: algebras in a straightforward way, or more elegantly using the

341: $A_\infty$-category $fun(\A,\A)$ of functors and natural

342: transformations, as endomorphisms of the identity functor. A

343: well-known rather imprecise principle says that ``Hochschild

344: cohomology is an invariant of the derived category''. In a rigorous

345: formulation which is suitable for our purpose,

346: \begin{equation} \label{eq:hh}

347: HH^*(\A,\A) \stackrel{?}{\iso} HH^*(Tw^\pi(\A),Tw^\pi(\A)).

348: \end{equation}

349: This is unproved at the moment, because $Tw^\pi(\A)$ itself has not

350: been considered in the literature before, but it seems highly

351: plausible (a closely related result has been proved in

352: \cite{keller98}). Hochschild cohomology is important for us because

353: of its role in deformation theory, see the next section; but we want

354: to discuss its possible geometric meaning first.

355: 

356: Let $M$ be as in Assumption \ref{as:projective} (one could more

357: generally take any exact symplectic manifold with contact type

358: boundary and vanishing $c_1$). Then there is a natural ``open-closed

359: string map'' from the symplectic cohomology to the Hochschild

360: cohomology of the Fukaya category:

361: \begin{equation} \label{eq:open-closed}

362: SH^*(M) \longrightarrow HH^*(\F(M),\F(M)).

363: \end{equation}

364: This is defined in terms of Riemann surfaces obtained from the disk

365: by removing one interior point and an arbitrary number of boundary

366: points. Near the interior point, one deforms the holomorphic map

367: equation in the same way as in the definition of $SH^*(M)$, using a

368: large Hamiltonian function; otherwise, one uses boundary conditions

369: as for $\F(M)$. Figure \ref{fig:polygon2} shows what the solutions

370: look like.

371: %

372: \begin{figure}[htb] \begin{center}

373: \epsfig{file=polygon2.eps} \caption{\label{fig:polygon2}}

374: \end{center} \end{figure}

375: 

376: $HH^*(\A,\A)$ for any $\A$ carries the structure of a Gerstenhaber

377: algebra, and one can verify that \eqref{eq:open-closed} is a morphism

378: of such algebras. Actually, since $SH^*(M)$ is a BV algebra, one

379: expects the same of $HH^*(\F(M),\F(M))$. This should follow from the

380: fact that $\F(M)$ is a cyclic $A_\infty$-category in some appropriate

381: weak sense, but the story has not yet been fully worked out (two

382: relevant papers for the algebraic side are \cite{tradler01} and

383: \cite{tamarkin-tsygan00}).

384: 

385: \begin{conjecture} \label{conj:hh}

386: If $M$ is as in Assumption \ref{as:lefschetz-fibre},

387: \eqref{eq:open-closed} is an isomorphism.

388: \end{conjecture}

389: 

390: Assumption \ref{as:lefschetz-fibre} appears here mainly for the sake

391: of caution. There are a number of cases which fall outside it, and to

392: which one would want to extend the conjecture, but it is not clear

393: where to draw the line. Certainly, without some restriction on the

394: geometry of $M$, there can be no connection between the Reeb flow on

395: $\partial M$ and Lagrangian submanifolds?

396: 

397: \section{Deformations of categories\label{sec:def}}

398: 

399: The following general definition, due to Kontsevich, satisfies the

400: need for a deformation theory of categories which should be

401: applicable to a wide range of situations: for instance, a deformation

402: of a complex manifold should induce a deformation of the associated

403: differential graded category of complexes of holomorphic vector

404: bundles. By thinking about this example, one quickly realizes that

405: such a notion of deformation must include a change in the set of

406: objects itself. The $A_\infty$-formalism, slightly extended in an

407: entirely natural way, fits that requirement perfectly. The relevance

408: to symplectic topology is less immediately obvious, but it plays a

409: central role in Fukaya, Oh, Ohta and Ono's work on ``obstructions''

410: in Floer cohomology \cite{fooo} (a good expository account from their

411: point of view is \cite{fukaya00}).

412: 

413: For concreteness we consider only $A_\infty$-deformations with one

414: formal parameter, that is to say over $\Q[[t]]$. Such a deformation

415: $\E$ is given by a set $Ob\,\E$ of objects, and for any two objects a

416: space $hom_\E(X_0,X_1)$ of morphisms which is a free graded

417: $\Q[[t]]$-module, together with composition operations as before but

418: now including a $0$-ary one: this consists of a so-called

419: ``obstruction cocycle''

420: \begin{equation} \label{eq:obstruction}

421: \mu^0_\E \in hom_\E^2(X,X)

422: \end{equation}

423: for every object $X$, and it must be of order $t$ (no constant term).

424: There is a sequence of associativity equations, extending those of an

425: $A_\infty$-category by terms involving $\mu^0_\E$. Clearly, if one

426: sets $t = 0$ (by tensoring with $\Q$ over $\Q[[t]]$), $\mu^0_\E$

427: vanishes and the outcome is an ordinary $A_\infty$-category over

428: $\Q$. This is called the special fibre and denoted by $\E_{sp}$. One

429: says that $\E$ is a deformation of $\E_{sp}$.

430: 

431: A slightly more involved construction associates to $\E$ two other

432: $A_\infty$-categories, the global section category $\E_{gl}$ and the

433: generic fibre $\E_{gen}$, which are defined over $\Q[[t]]$ and over

434: the Laurent series ring $\Q[t^{-1}][[t]]$, respectively. One first

435: enlarges $\E$ to a bigger $A_\infty$-deformation $\E_c$ by coupling

436: the existing objects with formal connections (the terminology comes

437: from the application to complexes of vector bundles). Objects of

438: $\E_c$ are pairs $(X,\alpha)$ consisting of $X \in Ob\,\E$ and an

439: $\alpha \in hom^1_\E(X,X)$ which must be of order $t$. The morphism

440: spaces remain the same as in $\E$, but all the composition maps are

441: deformed by infinitely many contributions from the connection. For

442: instance,

443: \begin{equation} \label{eq:deformed-obstruction}

444: \mu^0_{\E_c} = \mu^0_\E + \mu^1_\E(\alpha) + \mu^2_\E(\alpha,\alpha)

445: + \dots \in hom^2_{\E_c}((X,\alpha),(X,\alpha)) = hom^2_\E(X,X).

446: \end{equation}

447: $\E_{gl} \subset \E_c$ is the full $A_\infty$-subcategory of objects

448: for which \eqref{eq:deformed-obstruction} is zero; and $\E_{gen}$ is

449: obtained from this by inverting $t$. The transition from $\E_{sp}$ to

450: $\E_{gl}$ and $\E_{gen}$ affects the set of objects in the following

451: way: if for some $X$ one cannot find an $\alpha$ such that

452: \eqref{eq:deformed-obstruction} vanishes, then the object is

453: ``obstructed'' and does not survive into $\E_{gl}$; if on the other

454: hand there are many different $\alpha$, a single $X$ can give rise to

455: a whole family of objects of $\E_{gl}$. Finally, two objects of

456: $\E_{gen}$ can be isomorphic even though the underlying objects of

457: $\E_{sp}$ aren't; this happens when the isomorphism involves negative

458: powers of $t$.

459: 

460: The classification of $A_\infty$-deformations of an

461: $A_\infty$-category $\A$ is governed by its Hochschild cohomology, or

462: rather by the dg Lie algebra underlying $HH^{*+1}(\A,\A)$, in the

463: sense of general deformation theory \cite{goldman-millson88}. We

464: cannot summarize that theory here, but as a simple example, suppose

465: that $HH^2(\A,\A) \iso \Q$. Then a nontrivial $A_\infty$-deformation

466: of $\A$, if it exists, is unique up to equivalence and change of

467: parameter $t \mapsto f(t)$ (to be accurate, $f(t)$ may contain roots

468: of $t$, so the statement holds over $\Q[[t,t^{1/2},t^{1/3},\dots]]$).

469: The intuitive picture is that the ``versal deformation space'' has

470: dimension $\leq 1$, so that any two non-constant arcs in it must

471: agree up to reparametrization.

472: 

473: In the situation of Assumption \ref{as:projective}, the embedding of

474: our exact symplectic manifold $M$ into $X$ should give rise to an

475: $A_\infty$-deformation $\F(M \subset X)$. We say ``should'' because

476: the details, which in general require the techniques of \cite{fooo},

477: have not been carried out yet. Roughly speaking one takes the same

478: objects as in $\F(M)$ and the same morphism spaces, tensored with

479: $\Q[[t]]$, but now one allows ``holomorphic polygons'' which map to

480: $X$, hence may intersect the divisor $D$. The numbers of such

481: polygons intersecting $D$ with multiplicity $k$ will form the $t^k$

482: term of the composition maps in $\F(M \subset X)$. Because there can

483: be holomorphic discs bounding our Lagrangian submanifolds in $X$,

484: nontrivial obstruction cocycles \eqref{eq:obstruction} may appear.

485: 

486: The intended role of $\F(M \subset X)$ is to interpolate between

487: $\F(M)$, which we have been mostly discussing up to now, and the

488: Fukaya category $\F(X)$ of the closed symplectic manifold $X$ as

489: defined in \cite{fooo,fukaya01b}. The $t^0$ coefficients count

490: polygons which are disjoint from $D$, and these will automatically

491: lie in $M$, so that

492: \[

493: \F(M \subset X)_{sp} \iso \F(M).

494: \]

495: The relation between the generic fibre and $\F(X)$ is less

496: straightforward. First of all, $\F(M \subset X)_{gen}$ will be an

497: $A_\infty$-category over $\Q[t^{-1},t]]$, whereas $\F(X)$ is defined

498: over the Novikov ring $\Lambda_t$. Intuitively, one can think of this

499: difference as the consequence of a singular deformation of the

500: symplectic form. Namely, if one takes a sequence of symplectic forms

501: (all in the same cohomology class) converging towards the current

502: $[D]$, the symplectic areas of holomorphic discs $u$ would tend to

503: the intersection number $u \cdot D$. A more serious issue is that

504: $\F(M \subset X)_{gen}$ is clearly smaller than $\F(X)$, because it

505: contains only Lagrangian submanifolds which lie in $M$. However, that

506: difference may disappear if one passes to derived categories:

507: 

508: \begin{conjecture} \label{conj:generic-fibre} In the situation of

509: Assumption \ref{as:lefschetz-fibre}, there is a canonical equivalence

510: of triangulated categories

511: \[

512: D^\pi(\F(M \subset X)_{gen} \otimes_{\Q[t^{-1}][[t]]} \Lambda_t) \iso

513: D^\pi(\F(X)).

514: \]

515: \end{conjecture}

516: 

517: In comparison with the previous conjecture, Assumption

518: \ref{as:lefschetz-fibre} is far more important here. The idea is that

519: there should be an analogue of Theorem \ref{th:generators} for

520: $D^\pi(\F(X))$, saying that this category is split-generated by

521: vanishing cycles, hence by objects which are also present in $\F(M

522: \subset X)$.

523: 

524: To pull together the various speculations, suppose that $Y =

525: \CP{n+1}$ for some $n \geq 3$; $X \subset Y$ is a hypersurface of

526: degree $n+2$; and $D \subset X$ is the intersection of two such

527: hypersurfaces. Then $D^\pi(\F(M))$ is split-generated by finitely

528: many objects, hence $Tw^\pi(\F(M))$ is at least in principle

529: accessible to computation. Conjecture \ref{conj:hh} together with

530: \eqref{eq:b2}, \eqref{eq:hh} tells us that $HH^2(\F(M),\F(M)) \iso

531: HH^2(Tw^\pi(\F(M)),Tw^\pi(\F(M)))$ is at most one-dimensional, so an

532: $A_\infty$-deformation of $Tw^\pi(\F(M))$ is unique up to a change of

533: the parameter $t$. From this deformation, Conjecture

534: \ref{conj:generic-fibre} would enable one to find $D^\pi(\F(X))$,

535: again with the indeterminacy in the parameter (fixing this is

536: somewhat like computing the mirror map).

537: \providecommand{\bysame}{\leavevmode\hbox

538: to3em{\hrulefill}\thinspace}

539: \providecommand{\MR}{\relax\ifhmode\unskip\space\fi MR }

540: % \MRhref is called by the amsart/book/proc definition of \MR.

541: \providecommand{\MRhref}[2]{%

542:   \href{http://www.ams.org/mathscinet-getitem?mr=#1}{#2}

543: } \providecommand{\href}[2]{#2}

544: \begin{thebibliography}{10}

545: 

546: \bibitem{chas-sullivan99}

547: M.~Chas and D.~Sullivan, \emph{String topology}, Preprint

548: math.GT/9911159.

549: 

550: \bibitem{cieliebak-floer-hofer95}

551: K.~Cieliebak, A.~Floer, and H.~Hofer, \emph{Symplectic homology {II}:

552: a general

553:   construction}, Math. Z. \textbf{218} (1995), 103--122.

554: 

555: \bibitem{eliashberg-givental-hofer02}

556: Ya. Eliashberg, A.~Givental, and H.~Hofer, \emph{Introduction to

557: symplectic

558:   field theory}, Geom. Funct. Anal. \textbf{Special Volume, Part II} (2000),

559:   560--673.

560: 

561: \bibitem{fukaya02}

562: K.~Fukaya, \emph{Asymptotic analysis, multivalued {M}orse theory, and

563: mirror

564:   symmetry}, Preprint 2002.

565: 

566: \bibitem{fukaya00}

567: \bysame, \emph{Deformation theory, homological algebra, and mirror

568: symmetry},

569:   Preprint, December 2001.

570: 

571: \bibitem{fukaya01b}

572: \bysame, \emph{Floer homology and mirror symmetry {II}}, Preprint

573: 2001.

574: 

575: \bibitem{fukaya93}

576: \bysame, \emph{Morse homotopy, {$A_\infty$}-categories, and {F}loer

577:   homologies}, Proceedings of {GARC} workshop on Geometry and Topology (H.~J.

578:   Kim, ed.), Seoul National University, 1993.

579: 

580: \bibitem{fukaya-oh98}

581: K.~Fukaya and Y.-G. Oh, \emph{Zero-loop open strings in the cotangent

582: bundle

583:   and {M}orse homotopy}, Asian J. Math. \textbf{1} (1998), 96--180.

584: 

585: \bibitem{fooo}

586: K.~Fukaya, Y.-G. Oh, H.~Ohta, and K.~Ono, \emph{Lagrangian

587: intersection {F}loer

588:   theory - anomaly and obstruction}, Preprint, 2000.

589: 

590: \bibitem{getzler94}

591: E.~Getzler, \emph{{B}atalin-{V}ilkovisky algebras and 2d

592: {T}opological {F}ield

593:   {T}heories}, Commun. Math. Phys \textbf{159} (1994), 265--285.

594: 

595: \bibitem{goldman-millson88}

596: W.~Goldman and J.~Millson, \emph{The deformation theory of the

597: fundamental

598:   group of compact {K}{\"a}hler manifolds}, IHES Publ. Math. \textbf{67},

599:   43--96.

600: 

601: \bibitem{ionel-parker98}

602: E.-N.~Ionel and T.~Parker, \emph{Gromov-{W}itten invariants of

603: symplectic

604:   sums}, Math. Res. Lett. \textbf{5} (1998), 563--576.

605: 

606: \bibitem{keller98}

607: B.~Keller, \emph{Invariance and localization for cyclic homology of

608: {DG}

609:   algebras}, J. Pure Appl. Alg. \textbf{123} (1998), 223--273.

610: 

611: \bibitem{kontsevich94}

612: M.~Kontsevich, \emph{Homological algebra of mirror symmetry},

613: Proceedings of

614:   the International Congress of Mathematicians (Z{\"u}rich, 1994),

615:   Birkh{\"a}user, 1995, pp.~120--139.

616: 

617: \bibitem{kontsevich-soibelman00}

618: M.~Kontsevich and Y.~Soibelman, \emph{Homological mirror symmetry and

619: torus

620:   fibrations}, Symplectic geometry and mirror symmetry, World Scientific, 2001,

621:   pp.~203--263.

622: 

623: \bibitem{li-ruan98}

624: A.-M. Li and Y.~Ruan, \emph{Symplectic surgery and {G}romov-{W}itten

625: invariants

626:   of {C}alabi-{Y}au 3-folds}, Invent. Math. \textbf{145} (2001), 151--218.

627: 

628: \bibitem{ozsvath-szabo01}

629: P.~Ozsvath and Z.~Szabo, \emph{Holomorphic disks and topological

630: invariants for

631:   rational homology three-spheres}, Preprint math.SG/0101206.

632: 

633: \bibitem{polishchuk-zaslow98}

634: A.~Polishchuk and E.~Zaslow, \emph{Categorical mirror symmetry: the

635: elliptic

636:   curve}, Adv. Theor. Math. Phys. \textbf{2} (1998), 443--470.

637: 

638: \bibitem{pozniak}

639: M.~Po{\'z}niak, \emph{Floer homology, {N}ovikov rings and clean

640: intersections},

641:   Northern California Symplectic Geometry Seminar, Amer. Math. Soc., 1999,

642:   pp.~119--181.

643: 

644: \bibitem{seidel97}

645: P.~Seidel, \emph{Floer homology and the symplectic isotopy problem},

646: Ph.D.

647:   thesis, Oxford University, 1997.

648: 

649: \bibitem{seidel98b}

650: \bysame, \emph{Lagrangian two-spheres can be symplectically knotted},

651: J.

652:   Differential Geom. \textbf{52} (1999), 145--171.

653: 

654: \bibitem{seidel99}

655: \bysame, \emph{Graded {L}agrangian submanifolds}, Bull. Soc. Math.

656: France

657:   \textbf{128} (2000), 103--146.

658: 

659: \bibitem{seidel01}

660: \bysame, \emph{A long exact sequence for symplectic {F}loer

661: cohomology},

662:   Preprint math.SG/0105186.

663: 

664: \bibitem{tamarkin-tsygan00}

665: D.~Tamarkin and B.~Tsygan, \emph{Noncommutative differential

666: calculus, homotopy

667:   {BV} algebras and formality conjectures}, Preprint math.KT/0002116.

668: 

669: \bibitem{tradler01}

670: T.~Tradler, \emph{Infinity-inner-products on {A}-infinity-algebras},

671: Preprint

672:   math.\-AT/0108027.

673: 

674: \bibitem{viterbo97a}

675: C.~Viterbo, \emph{Functors and computations in {F}loer homology with

676: applications,

677:   {P}art {I}}, Geom. Funct. Anal. \textbf{9} (1999), 985--1033.

678: 

679: \bibitem{viterbo97b}

680: \bysame, \emph{Functors and computations in {F}loer homology with

681:   applications, {P}art {II}}, Preprint 1996.

682: \end{thebibliography}

683: \end{document}

684: