1: \documentclass[11pt]{article}

2: \setlength{\textwidth}{14cm}

3: \setlength{\oddsidemargin}{1cm}

4: 

5: \usepackage{amssymb,latexsym,epsfig,epsf,xypic}

6: \usepackage[mathscr]{euscript}

7: 

8: \makeatletter \@addtoreset{equation}{section} \makeatother

9: \renewcommand{\theequation}{\thesection.\arabic{equation}}

10: 

11: \newtheorem{Lemma}[equation]{Lemma}

12: \newtheorem{Theorem}[equation]{Theorem}

13: \newtheorem{Prop}[equation]{Proposition}

14: \newtheorem{Definition}[equation]{Definition}

15: \newtheorem{Corollary}[equation]{Corollary}

16: \newtheorem{Conj}[equation]{Conjecture}

17: \newenvironment{Proof}{\noindent\emph{Proof\ }}{\hfill$\square$\\}

18: \newenvironment{Remarks}{\noindent\textbf{Remarks}\ }

19: 

20: \newcommand{\rt}[1]{\stackrel{#1\,}{\rightarrow}}

21: \newcommand{\Rt}[1]{\stackrel{#1\,}{\longrightarrow}}

22: \newcommand\A{\mathcal A}

23: \newcommand\db{\bar\partial}

24: \newcommand\R{\mathbb R}

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

26: \newcommand\Z{\mathbb Z}

27: \newcommand\Pee{\mathbb P}

28: \newcommand\PD{\mathrm{PD}\,}

29: \newcommand\OO{\mathscr O}

30: \newcommand\I{\mathcal I}

31: \newcommand\To{\longrightarrow}

32: \newcommand\ot{\leftarrow}

33: \newcommand\Ot{\longleftarrow}

34: \newcommand\into{\hookrightarrow}

35: \newcommand\res{\arrowvert_}

36: \newcommand\tr{\mathrm{tr}}

37: \newcommand\id{\mathrm{id}}

38: \newcommand\End{\mathrm{End}\,}

39: \newcommand\ip{{\mbox\,\_\hspace{-1.5mm}\shortmid\hspace{1.5pt}}}

40: \newcommand\ImO{\,\mathrm{Im}\,\Omega}

41: 

42: \title{\textbf{Moment maps, monodromy and mirror manifolds}}

43: \author{R.\,P. Thomas}

44: \date{}

45: 

46: \begin{document}

47: 

48: \maketitle

49: \begin{abstract} \noindent

50: Via considerations of symplectic reduction, monodromy, mirror

51: symmetry and Chern-Simons functionals, a conjecture is proposed on

52: the existence of special Lagrangians in the

53: hamiltonian deformation class of a given Lagrangian submanifold of a

54: Calabi-Yau manifold. It involves a stability condition

55: for graded Lagrangians, and can be proved for the simple case of

56: $T^2$.

57: \end{abstract}

58: 

59: \section{Introduction}

60: Just as explicit solutions of the Einstein and Hermitian-Yang-Mills

61: equations exist only on spaces that are either low dimensional,

62: non-compact and/or highly symmetric,

63: so the equations for special Lagrangian (SLag) cycles,

64: also important in physics, have the same properties.

65: Physically there are also similarities in that we have two first order

66: supersymmetric minimal energy equations (HYMs and SLag)

67: implying the more standard second order equations (YMs and minimal

68: volume equations).

69: 

70: There are powerful existence results of Calabi and Yau (and more

71: recently Tian, Donaldson and others) for the Einstein equations, and of

72: Donaldson-Uhlenbeck-Yau for the HYM equations, so long as we are on a

73: K\"ahler (or projective) manifold; this often reduces

74: an infinite dimensional problem in PDEs to a finite dimensional

75: problem in linear algebra. Producing many K\"ahler-Einstein

76: (e.g. Calabi-Yau) manifolds becomes trivial, and dealing with

77: Hermitian-Yang-Mills connections requires only algebraic

78: computations; in both cases the complicated role of the K\"ahler form

79: and/or metric is almost removed. This can be thought of as possible

80: because of the existence of some infinite dimensional geometry recasting

81: the equations in terms of moment maps and symplectic reduction. A

82: similar situation for SLags would therefore be highly desirable.

83: In particular it might give a way of studying SLags using only

84: Lagrangians and symplectic geometry, much as HYM connections are

85: studied via stable bundles and algebraic geometry.

86: 

87: This paper explores the mirror symmetry of holomorphic bundles

88: (on a Calabi-Yau $3$-fold $M$,

89: often referred to here as `the complex side') and Lagrangians

90: (on the mirror Calabi-Yau 3-fold $W$, 'the symplectic side', known

91: as the K\"ahler side in the physics literature). Many

92: people have worked and are still working on proving some kind of

93: direct correspondence between such objects given an

94: SYZ torus fibration \cite{SYZ}; see for example \cite{AP}, \cite{BBHM},

95: \cite{Ch}, \cite{Fu1}, \cite{Gr}, \cite{LYZ}, \cite{PZ}, \cite{Ty}, and

96: see \cite{MMM} for a review of this and many many more issues in mirror

97: symmetry. Here, however, we work purely formally without

98: reference to a particular pair of mirror manifolds, without

99: worrying about what mirror symmetry might rigorously mean, and we will

100: not try to give any explicit correspondence. Using mirror symmetry

101: merely as motivation, we point out some similar structures on

102: both sides of the mirror map. Under some conditions (in some

103: `large complex structure' or `semi-classical' or somesuch limit)

104: these structures might be genuinely dual; again it does not matter

105: if they are not in general. For instance, physics \cite{MMMS},

106: \cite{DFR} predicts that one should consider not the HYM equations and

107: slope but some perturbation of them away from the large complex

108: structure limit; however these equations also come from a moment map

109: and, conjecturally, a stability condition (for a discussion of such

110: matters see \cite{Le} or \cite{T3}). So while the slope and phase of

111: Lagrangians discussed below might not be exactly mirror to slope of

112: bundles, it should be mirror to something with analogous properties

113: and significance.

114: 

115: Loosely, we would like to think of submanifolds in a fixed homology class

116: as mirror to connections on a fixed

117: topological complex bundle (with Chern classes mirror to the homology

118: class); then Lagrangians should correspond to holomorphic

119: connections (i.e. integrable connections; those with no

120: $(0,2)$-curvature) and special Lagrangians to those with HYM

121: curvature. These last two conditions should be stability conditions

122: for the group actions of hamiltonian deformations and complex gauge

123: transformations, respectively. The full picture is much more complicated,

124: involving triangulated categories and so forth, as envisaged some six

125: years ago in the seminal conjecture of Kontsevich \cite{K}; we can

126: ignore this in only using mirror symmetry

127: as motivation. It could be noted, however, that the functionals defined

128: below are additive under exact sequences of holomorphic vector bundles

129: and sums of Lagrangians, so should extend to the derived category

130: of coherent sheaves and the derived Fukaya category of Lagrangians

131: respectively.

132: 

133: First note that while the connections side has a complex

134: structure and a complex gauge group involved, the Lagrangian side needs

135: complexifying. So motivated by Kontsevich \cite{K} and by physics

136: (e.g. \cite{SYZ}) we add in connections on the submanifolds (which

137: will later reduce to flat connections on Lagrangians). The dictionary

138: we are aiming towards, much of which is already standard, is the

139: following in the 3-dimensional case; all the terms used will be defined

140: in due course. \newpage

141: 

142: \renewcommand\arraystretch{2}

143: $$

144: \begin{array}{|c|c|}

145: \hline \text{Complex side }M & \text{Symplectic side }W \\

146: \hline\hline \Omega=\Omega^{ }_M\in H^{3,0} & \omega=\omega^{ }_W\in

147: H^{1,1} \\ \hline

148: H^{\mathrm{ev}} & H^3 \\ \hline

149: \mathrm{Connections\ }A\mathrm{\ on\ a\ fixed\ }C^\infty\mathrm{\ complex}

150: & \mathrm{Submanifolds/cycles\ }L\mathrm{\ in\ a\ fixed

151: \ class} \vspace{-4mm} \\

152: \mathrm{bundle\ }E;\ \,v(E)=ch(E)\sqrt{\mathrm{Td\,}X}\in H^{\mathrm{ev}} &

153: [L]\in H^3,\mathrm{\ with\ a\ connection \ on\ }\C\times L \\ \hline

154: CS_\C(A=A_0+a)= \hspace{25mm} &

155: f_\C(A,L)=\int_{L_0}^L(F+\omega)^2 \hspace{2cm}\vspace{-3mm} \\

156: \hspace{6mm} \frac1{4\pi^2}\int_M\tr\left(\db_{A_0}a\wedge

157: a+\frac23a^{\wedge3}\right)\wedge\Omega &

158: \hspace{1cm}=\int_{L_0}^L(F^2+\omega^2)+2

159: \int_{L_0}^L\omega\wedge F \\ \hline

160: \mathrm{Critical\ points:\ }F_A^{0,2}=0 &

161: \mathrm{Critical\ points:\ }\omega\res L=0,\ F_A=0 \vspace{-4mm} \\

162: \mathrm{holomorphic\ bundles\ } &

163: \mathrm {Lagrangians\ +\ flat\ line\ bundles\ } \\ \hline

164: \text{Holomorphic Casson invariant \cite{T1}} &

165: \text{Counting SLags \cite{J}} \\ \hline

166: \mathrm{Gauge\ group} & U(1)\mathrm{\ gauge\ group\ on\ }L \\ \hline

167: \mathrm{Complexified\ gauge\ group} & \mathrm{Hamiltonian\ 

168: deformations} \\ \hline

169: \omega=\omega^{\ }_M\in H^{1,1} & \Omega=\Omega^{\ }_W\in H^{3,0} \\ \hline

170: \mathrm{Moment\ map\ \ }F_A\wedge\omega^{n-1} &

171: \mathrm{Moment\ map\ \,}\ImO\res L \\ \hline

172: \mathrm{Stability,\ slope\ }\mu=\frac1{\mathrm{rk\,}E}

173: \int\tr\,F_A\wedge\omega^{n-1} & \mathrm{Stability,\ slope\ }

174: \mu=\frac1{\mathrm{vol\,}(L)}\int_L\ImO \\ \hline

175: \end{array}

176: $$

177: (In the fourth line, $v(E)$ is the Mukai vector of $E$; in the last

178: line, vol$\,(L)$ is the cohomological volume measured with respect

179: to Re$\,\Omega$.) A SLag cycle (of phase $\phi$) in a Calabi-Yau

180: is a Lagrangian with Im\,$(e^{-i\phi}\Omega)\res L\equiv0$ \cite{HL};

181: then Re\,$(e^{-i\phi}\Omega)\res

182: L$ is the Riemannian volume form on $L$ induced by the Ricci-flat

183: metric on the Calabi-Yau. Obviously, rotating $\Omega$ by $e^{-i\phi}$

184: gives SLags in the more traditional sense of phase zero.

185: The part of the theory to do with SLags will apply in all dimensions;

186: it is only the functionals that are special to Calabi-Yau 3-folds.

187: 

188: We will partially justify the above table, though the symplectic

189: structure and moment map give problems that will appear in due course.

190: However we can derive enough to arrive at

191: a conjecture about Lagrangians and SLags for which evidence will

192: be given by using monodromy and mirror symmetry from \cite{ST}

193: to interpret an example of Lawlor and Joyce \cite{J}. \\

194: 

195: 

196: \noindent \textbf{Acknowledgements.} The debt of any ex-student of

197: Simon Donaldson writing a paper on moment maps should be clear.

198: This work is also more immediately influenced by the papers

199: \cite{D},\,\cite{J},\,\cite{K}. In particular I was surprised to see the

200: Lagrangian condition coming from a moment map in \cite{D},\,\cite{H},

201: which does not fit into the

202: scheme I always supposed was true. So the purpose of this paper, apart

203: from trying to set a record for the number of m's in a title, is to

204: expand on that scheme and to try to get the \emph{special} condition

205: from a moment map instead. This paper was finished in the summer of 2000

206: and reported on in \cite{T3}; since then exciting new ideas have

207: appeared in physics \cite{Do} and mathematics \cite{KS} better explaining

208: mirror symmetry. I would like to thank S.-T. Yau, C. H.

209: Taubes and Harvard University for support, and Yi Hu,

210: Albrecht Klemm, Ivan Smith and Xiao Wei Wang for

211: useful conversations. Communications with Mike Douglas, Dominic Joyce,

212: Paul Seidel, S.-T. Yau and Eric Zaslow have been extremely influential.

213: 

214: 

215: \section{Chern-Simons-type functionals and critical points}

216: 

217: Consider the space $\A$ of $(0,1)$-connections $A$ on a fixed complex

218: bundle on a Calabi-Yau 3-fold $M$. This infinite dimensional space

219: has a natural complex structure, with respect to which it admits a

220: holomorphic functional, Witten's holomorphic Chern-Simons functional

221: \cite{W1},\,\cite{DT},

222: $$

223: CS_\C(A=A_0+a)=\frac1{4\pi^2}\int_M\tr\left(\db_{A_0}a\wedge

224: a+\frac23a\wedge a\wedge a\right)\wedge\Omega,

225: $$

226: where $\Omega$ is the holomorphic (3,0)-form. It is infinitesimally

227: gauge-invariant (gauge transformations not homotopic to the identity

228: can give periods to $CS_\C$) and its gradient is $F_A^{0,2}$,

229: with zeros the integrable connections. That is, after dividing by

230: gauge equivalence (under which grad$\,CS_\C$ is invariant), the critical

231: points of $CS_\C$ form the space of holomorphic bundles of the same

232: topological type. As critical points of a functional, moduli of

233: holomorphic bundles have virtual dimension zero, and one might try to

234: make sense of counting them -- a holomorphic Casson invariant \cite{T1}.

235: This is independent of deformations of the complex structure, but can

236: have wall-crossing changes as the K\"ahler form varies. (This is because

237: we count only stable bundles, and the notion of stability depends on a

238: K\"ahler form.)

239: 

240: On the other hand, on a different Calabi-Yau 3-fold $W$ (for instance

241: the mirror, in some situation where this makes sense),

242: Lagrangians are the critical points of a functional

243: too, on the space of all 3-dimensional submanifolds (or cycles):

244: $$

245: f_\R(L)=\int_{L_0}^L\omega\wedge\omega,

246: $$

247: where $\omega$ is the symplectic form on $W$. Here $L_0$ is a fixed cycle

248: in the same homology class, and we integrate over a 4-chain

249: with boundary $L-L_0$; the functional $f_\R$ is invariant under the choice

250: of different, homologous, 4-chains (picking non-homologous 4-chains can give

251: periods to $f_\R$). It is invariant under

252: deformations of $L$ pulled back from hamiltonian deformations of $W$

253: (deformations generated by vector fields $v$ on $W$ whose

254: contraction with $\omega$ is exact $v\ip\omega=dh$ at each point in

255: time) as $\int_L\omega\wedge dh=0$, and its

256: gradient is $\omega|_L$. Thus its critical points are Lagrangian

257: submanifolds. We would like to think of $f_\R$ as mirror

258: to $CS_\C$, but to do so we must complexify it.

259: 

260: Thus we work on the space $\A$ of submanifolds $L$ of $W$ with $U(1)$

261: connections

262: $A$ on the trivial bundle $\C\times L$ on $L$.  Notice these submanifolds

263: are not parameterised by a map of a real 3-manifold into $W$; we are only

264: interested in the image $L$. \emph{From now on we shall restrict

265: attention to smooth Lagrangian submanifolds.} Formally, we

266: consider the tangent space to $\A$ at a point $(A,L\subset W)$ to be

267: \begin{equation} \label{tsp}

268: \Omega^1(L;\R)\oplus \Omega^1(L;\R),

269: \end{equation}

270: at least for those $L$ with no $J$-invariant subspaces of its tangent

271: spaces ($J$ is the complex structure on $W$, and this is reasonable since

272: we are looking for Lagrangian submanifolds after all). The first factor

273: is the obvious tangent space to the connections on $L$; the second gives

274: deformations of $L$ via the vector fields produced by contracting with

275: the K\"ahler form $\omega$ on $W$. That is, we use the metric on $W$ to

276: map $\Omega^1(L)$ to $\Omega^1(W)|_L$, then use the isomorphism

277: provided by $\omega$ to get a vector field along $L$. Equivalently,

278: using the metric on $W$, we may think of one-forms on $L$ as tangent

279: vectors to $L$, then apply the complex structure $J$ on $W$ to give

280: $W$-vector fields on $L$. We denote this map from one-forms to normal

281: vector fields by

282: \begin{equation}\label{ip}

283: \Omega^1(L)\to TW\res L, \qquad \sigma\mapsto\sigma\ip\omega^{-1}.

284: \end{equation}

285: Connections on $L$ are carried along by the

286: vector field to connections on nearby cycles, and we are identifying

287: the space of $U(1)$ connections with $i\Omega^1(L;\R)$.

288: 

289: There is a natural almost complex structure on $\A$, acting as

290: \renewcommand\arraystretch{1}

291: $$

292: J=\left(\begin{array}{cc} 0 & 1 \\ \!-1 & 0 \end{array}\right),

293: $$

294: with respect to the splitting (\ref{tsp}) of the tangent spaces. With

295: respect to this we claim to have the following holomorphic functional

296: $$

297: f_\C(A,L)=\int_{L_0}^L(F+\omega)^2=\int_{L_0}^L(F^2+\omega^2)\ +\ 2

298: \int_{L_0}^L\omega\wedge F.

299: $$

300: Here we have extended $A$ to a connection on the trivial bundle on the

301: whole of $W$ (restricting to a fixed connection $A_0$ on $L_0$, and to

302: $A$ on $L$) and taken its curvature form $F$. We have again picked

303: a 4-cycle bounding $L-L_0$; because $F$ and $\omega$ are closed the

304: resulting functional is independent of different homologous choices

305: of the 4-cycle, and in general well defined up to the addition of some

306: discrete periods. It is also (again) independent of hamiltonian isotopies

307: of $L$. Notice that the $\int_{L_0}^LF^2$ term is just the

308: \emph{real} Chern-Simons functional $CS_\R$ of the connection $A$ on $L$,

309: whose critical points are well known to be flat connections. As pointed

310: out to me by Eric Zaslow, the real and complex Chern-Simons

311: functionals already appear in \cite{W1} and \cite{Va} as possible

312: mirror partners (this is partially justified in \cite{LYZ}), but without

313: the terms in the symplectic form (and including instanton corrections

314: from holomorphic discs which we are ignoring for our rough analogy).

315: Asking for a real function to be equal to a complex one is possible

316: when one restricts attention to a real slice such as the space of

317: \emph{Lagrangian} submanifolds in $\A$; deforming within this space

318: the imaginary part of $f_\C$ remains constant and it reduces to

319: $CS_\R$. But allowing the imaginary counterparts to these real

320: deformations the right functional to consider is $f_\C$. Notice also

321: that if $\omega/2\pi$ is integral, so that we can pick a connection $B$

322: with curvature $-i\omega$, then the action functional

323: can be written in the more familiar looking Chern-Simons form

324: $$

325: f_\C(A,L)=\int_L(B+iA)\wedge d(B+iA)=\int_LCdC

326: $$

327: for the `complexified connection' $C=B+iA$ (a $\C^{\!\times}$-connection,

328: instead of a $U(1)$-connection.) This makes more contact with the

329: physics literature and allows one to extend the identification of

330: $CS_\R$ and $CS_\C$ in \cite{LYZ} to non Lagrangian sections, giving

331: complex values. Tian has informed me that he and Chen have also

332: considered the functional $f_\C(A,L)$ \cite{Ch}.

333: 

334: Mirror symmetry should relate Lagrangians not just to bundles but the

335: whole derived category. For Riemann surfaces $C\subset M$, for instance,

336: there is a functional in \cite{DT}, \cite{W2} rather like $f_\R$

337: above:

338: $$

339: \int_{C_0}^C\Omega

340: $$

341: is formally holomorphic and has as critical points the \emph{holomorphic}

342: curves $C$. Similarly for four-manifolds $S\subset M$ with connections on

343: them the following functional (formally similar to $f_\C$)

344: $$

345: \int_{S_0}^S\tr\,F\wedge\Omega

346: $$

347: has critical points the holomorphic surfaces with flat connection on them.

348: Alternatively, as $CS_\C$ is additive under extensions of

349: bundles it does extend to the derived category. (Whether these two

350: approaches are compatible; i.e. whether or not the functional

351: associated to a curve or surface is the same as $CS_\C$ applied

352: to a locally free resolution of its structure sheaf, up to a constant,

353: seems to not have been worked out.)

354: 

355: That $f_\C$ is holomorphic follows from the computation that the derivative

356: of $f_\C$ down $a\in\Omega^1(L)\oplus0$ (that only changes the connection

357: $A\mapsto A+\delta a$) is $\int_L2F\wedge ia+2\omega\wedge ia$,

358: while the derivative down $-Ja\in0\oplus\Omega^1(L)$, i.e. down the

359: vector field $a\ip\omega^{-1}$, is $\int_L2\omega\wedge a+2a\wedge F$.

360: The second expression is $-i$ times

361: the first, so the derivative is complex linear and $f$ is holomorphic.

362: Equivalently we are saying that $df_\C$ is the 2-form

363: $$

364: 2i\,(F+\omega)\ \oplus\ 2\,(F+\omega),

365: $$

366: which pairs with the tangent space (\ref{tsp}) by integration over $L$ to

367: give a form of type (1,0) on (\ref{tsp}).

368: 

369: Thus critical points of the functional are Lagrangian cycles with flat line

370: bundles on them: exactly the basic building blocks of the objects proposed

371: in \cite{K} to be mirror dual to the holomorphic bundles that are the

372: critical points of $CS$. So this ties

373: in three well known moduli problems of virtual dimension zero (i.e. with

374: deformation theories whose Euler characteristic vanishes) -- flat bundles

375: on 3-manifolds, holomorphic bundles on Calabi-Yau 3-folds, and Lagrangians

376: (up to hamiltonian deformation) in symplectic 6-manifolds.

377: 

378: So as mirror to \cite{T1} one would like to count Lagrangians

379: (up to hamiltonian deformations) plus flat line bundles on them, and

380: this is what Joyce's work \cite{J} has begun to tackle (in the rigid

381: case of $L$ being a homology sphere). Mirroring precisely the behaviour

382: of the holomorphic Casson invariant this count appears to be independent

383: of deformations of the K\"ahler form and to have wall-crossing changes

384: as the complex structure varies.

385: 

386: \section{Gauge equivalence and moment maps}

387: 

388: In fact what Joyce is proposing to count is \emph{special} Lagrangian

389: spheres with flat line bundles on them

390: (hence the otherwise anomalous dependence on the complex

391: structure), while \cite{T1} counts \emph{stable} bundles (i.e. by

392: Donaldson-Uhlenbeck-Yau, modulo the technicalities of polystable and

393: non-locally-free sheaves, we count \emph{Hermitian-Yang-Mills}

394: connections; hence the dependence on the K\"ahler form). (Tyurin

395: \cite{Ty} was perhaps the first to suggest

396: that the holomorphic Casson invariant should be related by mirror

397: symmetry to the real Casson invariant (here the $U(1)$ Casson

398: invariant) of SLag submanifolds.)

399: 

400: The link should be, of course, that we want to consider holomorphic

401: connections on one side, up to complex gauge equivalence, and

402: Lagrangians on the other side, up to hamiltonian isotopy, and in both cases

403: we try to do this by picking distinguished representatives of equivalence

404: classes by the usual method of symplectic reduction.

405: Bringing in a K\"ahler structure on the complex side, we get a moment

406: map for the gauge group action, whose zeros give the HYM equations.

407: Dually, we would like to bring in the holomorphic 3-form on the

408: symplectic (K\"ahler) side, and get a complex group to act.

409: So again complexify by adding flat line bundles: consider the critical

410: points of the functional $f$ of the last section, i.e.

411: the space

412: $$

413: \mathcal Z=\{(L,A)\,:\,L\subset W\mathrm{\ is\ Lagrangian,\ }A

414: \mathrm{\ is\ a\ flat\ connection\ on\ }L\,\}

415: $$

416: (\emph{not} up to gauge equivalence). In fact consider this space

417: on a Calabi-Yau manifold $W$ of any dimension $n$. It has tangent space

418: $$

419: T_{(L,A)}\mathcal Z=Z^1(L)\oplus Z^1(L)

420: $$

421: ($Z^1(L)$ denotes closed real one-forms on $L$), the first being tangent

422: to the space of flat connections, the second giving

423: normal vector fields (by contracting with $\omega^{-1}$) preserving the

424: Lagrangian condition. We have an obvious almost complex structure

425: \begin{equation} \label{aha}

426: J=\left(\begin{array}{cc} 0 & 1 \\ -1 & 0 \end{array}\right).

427: \end{equation}

428: 

429: Then the real group $C^\infty(L;\R)/\R$ acts as the Lie algebra to the

430: group of gauge transformations on

431: the flat line bundles (taking $d$ and adding to the connection)

432: whose complexification $C^\infty(L;\C)/\C$

433: acts complex linearly: the imaginary part $C^\infty(L;\R)/\R$

434: acts by hamiltonian deformations through the normal vector field

435: $$

436: h\mapsto dh\ip\omega^{-1}.

437: $$

438: Unfortunately, without using a metric this vector field is only

439: defined up to the addition of tangent vector fields to $L$; the map

440: (\ref{ip}) is really a map to $(TW\res L)/TL$ which we have lifted to

441: $TW\res L$ using the metric. (Equivalently we can extend

442: $h$ to a first formal neighbourhood of $L$ in different ways to get a

443: different vector field.) How we pick this alters how we carry the flat

444: connection along with $L$, and how the almost complex structure

445: (\ref{aha}) acts. For instance suppose we are in the rather artificial

446: case of $L$ being transverse to an SYZ $T^n$-fibration. Then

447: we can carry $L$ and the flat connection up the fibres

448: and identify the functions $C^\infty(L)$ from Lagrangian

449: to Lagrangian using the projection. Thus the group remains constant

450: as $L$ moves (effectively what we are doing is extending

451: functions from $L$ to a neighbourhood of $L$ in $W$ by pulling up

452: along the SYZ fibres). This does not work when $L$ branches over the

453: base of such a fibration. One can instead use the metric to define

454: normal vector fields, but then identifying the Lie algebra

455: $C^\infty(L)$ with a fixed $C^\infty(L_0)$ for all $L$ becomes difficult.

456: 

457: This problem is perhaps not so surprising -- the moment the Lagrangian

458: has branching over the base of an SYZ fibration simple explicit

459: correspondences between Lagrangians and vector bundles (such as \cite{LYZ})

460: also break down due to our ignoring important holomorphic

461: disc instanton corrections that appear in the physics. For instance

462: recent work of Fukaya, Oh, Ohta and Ono \cite{FO3}, surveyed in \cite{Fu1},

463: show these provide the obstructions mirror to those of deformations

464: of holomorphic bundles \cite{T2} -- one should not in general consider

465: all (S)Lags (which are unobstructed) as mirror to holomorphic bundles,

466: but only those whose Floer cohomology (whose definition involves holomorphic

467: discs) is well defined.

468: 

469: However, what is clear is the totality of the group action,

470: even if identifying individual elements causes problems, and this is

471: all we really need. For instance in the $K3$ (or $T^4$) case one can get

472: the same total group orbit, with a genuine

473: fixed group acting, by hyperk\"ahler rotating a construction due to

474: Donaldson \cite{D}. The end result is that

475: one considers parametrised Lagrangian embeddings $f$ from a Riemann

476: surface $L$ into the $K3$ such that the pullback of Re$\,\Omega$ is a

477: fixed symplectic form on $L$. Then the group of exact

478: symplectomorphisms of

479: $$

480: (L,f^*\mathrm{Re\,}\Omega)

481: $$

482: provide a symmetry group of the space of maps $f$, which also carries

483: a natural K\"ahler structure. Complexified orbits give hamiltonian

484: deformations, and the moment map is $m(f)=f^*\ImO$. The connection

485: with our construction is that after fixing a line bundle $\eta$

486: and connection with curvature $f^*\mathrm{Re\,}\Omega$, an infinitesimal

487: symplectomorphism $\phi$ induces a flat connection, via parallel transport

488: and pull back, on the bundle $\eta\otimes\phi^*\eta^*$. Globally the

489: action is different (this action has non-zero Lie bracket, for instance,

490: and a fixed group) but the total group orbit and the moment map

491: (see below) are the same.

492: 

493: In general it is clear that the problem of identifying the group

494: for different embeddings of $L$ should be resolved by working with

495: the space of maps from a fixed $L_0$ to $W$, and enlarging the

496: group by including diffeomorphisms of $L_0$, giving a semi-direct

497: product of Diff$(L_0)$ and $U(1)$ gauge transformations on $L_0$.

498: Then the moment map for the diffeomorphism

499: part of the total group would be the Lagrangian condition as in

500: \cite{D}, and the problems we are encountering would come from

501: the fact that the group is a semi-direct product and not a product,

502: so that we cannot separate the two out and divide by them separately,

503: as in effect we have been trying to do. Unfortunately, I have not found the

504: correct formulation of the problem, but it is not so important for

505: follows.

506: 

507: So we shall not worry too much about whether the complex structure

508: defined above is integrable, the group is fixed, or the symplectic

509: structure below is closed. In 1 complex dimension it is trivial, in

510: 2 we can use Donaldson's picture, and in 3 dimensions

511: we could either try to use an abstract SYZ fibration to deform and

512: identify Lagrangians transverse to it, or take everything

513: in this section as motivation for finding the stability condition for

514: Lagrangians of the next section.

515: 

516: Fix a homology class of Lagrangians and multiply $\Omega$ by a unit

517: norm complex number so that $\int_L\ImO=0$.

518: We induce a symplectic structure on $\mathcal Z$ from $J$ and the

519: following metric on the tangent space

520: $$

521: \langle a,b\rangle=\int_La\wedge((b\ip\omega^{-1})\ip\ImO),

522: $$

523: for $a,\,b$ closed 1-forms. A computation in local coordinates

524: shows this is symmetric in $a$ and $b$; in fact it can be written as

525: \begin{equation} \label{lc}

526: \int_La\wedge(\widetilde b\ip\mathrm{\,Re\,}\Omega)=\int_L\cos\theta

527: \,(a\wedge*b),

528: \end{equation}

529: where $\,\widetilde{\ }\,$ is the isomorphism $T^*L\to TL$ set up by

530: the induced metric on $L$, $\Omega\res L=e^{i\theta}\mathrm{vol}_L$,

531: and vol$_L$ the Riemannian volume form on $L$ induced by the Ricci-flat

532: metric. Thus for Lagrangians with $\theta\in(-\pi/2,\pi/2)$, i.e. 

533: those for which Re$\,\Omega$ restricts to a nowhere vanishing volume

534: form on $L$ and so are not too far from being SLag ($\theta\equiv0$),

535: this gives a non-degenerate metric.

536: 

537: The symplectic form is invariant under the group action, and formally

538: the moment map is indeed $m(L,A)=\ImO$ in the dual $\Omega^n(L)_0$ of

539: the Lie algebra (i.e. $n$-forms on $L$ with integral zero). This follows

540: from the computation

541: $$

542: X\int_Lh\ImO=\int_Lh\,d\,(X\ip\ImO)=

543: \int_Ldh\wedge((\sigma\ip\omega^{-1})\ip\ImO),

544: $$

545: where $X=\sigma\ip\omega^{-1}$ is a normal vector field to the Lagrangian

546: $L$ down which we compute the derivative of the hamiltonian $\int_L

547: h\ImO=\langle m(L,A),h\rangle$ for the infinitesimal

548: action of $h$. Here have extended $h$ to a first-order neighbourhood

549: of $L\subset W$ so that it is constant in the direction of

550: $X=\sigma\ip\omega^{-1}$. Then the right hand side of the above

551: equation is the pairing using the symplectic form of $dh$ and $\sigma$,

552: as required.

553: 

554: Infinitesimally we can see the moment map interpretation very

555: easily, and fitting naturally with the mirror bundle point of view.

556: Deformations of holomorphic connections $A$ modulo

557: complex gauge equivalence are given by a ker$\,\db_A$/im$\,\db_A$\,

558: first cohomology group, related to deformations

559: ker$\,\db_A\,\cap$\,ker$\,\db_A^*$ of the HYM equations

560: (modulo \emph{unitary} gauge transformations)

561: via Hodge theory, with the moment

562: map equation providing the $d^*=0$ slice to the imaginary part of the

563: linearised group action. Similarly, deformations of

564: Lagrangians are given by closed 1-forms

565: ker$\,d:\,\Omega^1(L;\R)\to\Omega^2(L;\R)$, so that dividing by hamiltonian

566: deformations we get

567: $$

568: H^1(L)=\mathrm{ker}\,d/\mathrm{im}\,d.

569: $$

570: If instead of dividing we impose the \emph{special}

571: condition, we get a ker$\,d^*$ slice

572: $$

573: H^1(L)=\mathrm{ker}\,d\cap\mathrm{ker}\,d^*,

574: $$

575: to the (imaginary) deformations (real deformations are given by

576: changing the flat $U(1)$ connection that can be incorporated into this).

577: 

578: 

579: \subsection*{A symplectic example}

580: 

581: To motivate a guess at the correct definition of stability for

582: Lagrangians, we expand on an example of Lawlor and Joyce

583: (\cite{J} Sections 6 and 7, building on work of \cite{Ha}, \cite{L};

584: see also a similar example in \cite{SV} that is studied in

585: \cite{TY}), explaining its relevance to mirror symmetry, and giving a

586: simple example in algebraic geometry that mirrors it.

587: 

588: First define the pointwise phase $\theta$ of a submanifold $L$: we may

589: write

590: $$

591: \Omega\res L=e^{i\theta}\mathrm{\,vol\,}

592: $$

593: where vol is the Riemannian volume form on $L$ induced by Yau's

594: Ricci-flat metric \cite{Y} on $W$. Thus vol provides a (local) orientation

595: for $L$, and reversing its sign alters the phase $\theta$ by $\pi$. A

596: SLag is a Lagrangian with \emph{constant} phase $\theta$.

597: 

598: At first sight $\theta$ is multiply-valued; we always choose it to be a 

599: fixed single-valued function to $\R$, lifting $e^{i\theta}:\,L\to S^1$

600: and thus providing the Lagrangian with a \emph{grading} as introduced by

601: Kontsevich \cite{K},\,\cite{S2}. Thus \emph{we only consider Lagrangians

602: of vanishing Maslov class} -- for a Calabi-Yau this is the winding class

603: $\pi_1(L)\to\pi_1(S^1)$ of the phase map

604: $$

605: L\Rt{e^{i\theta}}S^1,

606: $$

607: which of course vanishes for a SLag. (The definition of grading in

608: \cite{K}, \cite{S2} is topological and uses the universal

609: $\mathbb Z$-cover of the bundle of Lagrangian Grassmannians; here we

610: first pass to the $\mathbb Z/2$ orientation cover of the Grassmannian,

611: choosing an orientation of our Lagrangians,

612: and then use a complex structure to pass to the universal

613: $\mathbb Z$-cover of this. The two definitions

614: are of course equivalent.)

615: 

616: Similarly we can define a kind of average phase $\phi=\phi(L)$ of a

617: submanifold (or homology class) $L\subset W$ by

618: $$

619: \int_L\Omega=A\,e^{i\phi(L)},

620: $$

621: for some real number $A$; we then use Re\,$(e^{-i\phi(L)}\Omega\res L)$

622: to orient $L$. Reversing the sign of $A$ alters the phase by $\pi$

623: and reverses the orientation. Again for a \emph{graded} Lagrangian

624: $L=(L,\theta)$, and \emph{we will always implicitly assume a grading},

625: $\phi(L)$ is canonically a real number (rather than $S^1$-valued).

626: Shifting the grading $[\,2n\,]:\,\theta\mapsto\theta+2n\pi$ gives a

627: similar shift to the phase $\phi(L)$.

628: 

629: The terminology comes from the fact that if there is a submanifold

630: in the same homology class as $L$ that is SLag with respect to some

631: rotation of $\Omega$, then it is with respect to $e^{-i\phi(L)}\Omega$.

632: Slope, which we define as

633: $$

634: \mu(L):=\tan(\phi(L))=\frac1{\int_L\mathrm{Re\,}\Omega}\int_L\ImO,

635: $$

636: is defined independently of grading, is monotonic in $\phi$ in the

637: range $(-\pi/2,\pi/2)$, and is invariant under change of orientation

638: $\phi\mapsto\phi\pm\pi$. This agrees with the slope of a straight

639: line SLag in the case of $T^2$, as featured in \cite{PZ}, and we think

640: of it as mirror to the slope of a mirror sheaf, as is shown for tori

641: in \cite{PZ} (see \cite{DFR} for corrections in higher dimensions away

642: from the large complex structure limit).

643: 

644: Joyce describes examples of SLags which we interpret as follows.

645: We have a family of Calabi-Yau 3-folds $W^t$ as $t$ ranges through (a

646: small open subset of) the moduli space of complex structures on $W$

647: \emph{with fixed symplectic structure}. That is, the holomorphic 3-form

648: $\Omega^t$ varies with $t$, but the K\"ahler form $\omega$ is fixed.

649: We also have a family of SLag homology 3-spheres

650: $L_1^t,\ L_2^t\subset W^t$ such that $L^t_1$ and $L^2_t$ intersect

651: at a point. If we choose a rotation of $\Omega^t$ such that $L_2^t$

652: always has phase $\phi_2^t\equiv0$ (this is possible locally at least;

653: in the family described later it will have to be modified slightly),

654: then we are interested as $t$ varies only in the complex number

655: $$

656: \int_{L_1^t}\Omega^t=R^te^{\phi^t_1}

657: $$

658: and its polar phase $\phi=\phi^t_1$; we plot this (i.e. the projection

659: from the complex structure moduli space to $\C$ via this map) in Figure 1.

660: 

661: Then in Joyce's example, for $\phi<0$ (and $R^t>0$)

662: there is a SLag $L^t$ (of some phase $\phi^t$)

663: in the homology class $[L^t]=[L^t_1]+[L^t_2]$,

664: such that as $\phi\uparrow0$, this degenerates to a singular union

665: of SLags of the same phase $L^t=L^t_1\cup L^t_2$ and then disappears

666: for $\phi>0$.

667: 

668: Most importantly, where $L^t$ exists as a smooth SLag ($\phi<0$) we have

669: the slope (and phase) inequality

670: \begin{equation} \label{ineq}

671: \mu^t_1<\mu^t_2, \qquad \mathrm{i.e.\ \,}\phi=\phi^t_1<\phi^t_2\equiv0;

672: \end{equation}

673: at $t=0,\ L^t$ becomes the singular union of $L^t_1$ and $L^t_2$,

674: with

675: $$

676: \mu^t_1=\mu^t_2\ (\phi^t_1=\phi^t_2);

677: $$

678: then there is no SLag in $L$'s homology class for

679: $$

680: \mu^t_1>\mu^t_2\ (\phi^t_1>\phi^t_2),

681: $$

682: \emph{though there is a Lagrangian}, of course -- the symplectic

683: structure has not changed. Though

684: we have been using slope $\mu$ in order to strengthen the analogy

685: with the mirror (bundle) situation, from now on we shall use only

686: the phase (lifted to $\R$ using the grading). While each is

687: monotonic in the other for small phase (as $\tan\phi=\mu$), slope

688: does not see orientation as phase does; reversing orientation adds

689: $\pm\pi$ to the phase but leaves $\mu$ unchanged. This is related to

690: the fact that we should really be working with complexes and so forth

691: on the mirror side (the bundle analogy is too narrow) and changing

692: orientation has no mirror analogue in terms of only stable bundles; it

693: corresponds to shifting (complexes of) bundles by one place

694: in the derived category. While slopes of bundles cannot go past

695: infinity (without moving degree in the derived category at least), for

696: Lagrangians they certainly can, and phase $\phi$ continues monotonically

697: upwards as its slope $\tan\phi$ becomes singular and then negative.

698: 

699: Importantly, we can think of the various SLags as independent of time

700: when thought of as Lagrangians in the fixed symplectic manifold $W^t$:

701: 

702: \begin{Lemma}

703: For $t>0$ the SLags $L^t$ are all in the same hamiltonian deformation

704: class. Similarly for $L_1^t,\ L^t_2$, and for $t<0$.

705: \end{Lemma}

706: 

707: \begin{Proof}

708: Now choosing the phase of $\Omega^t$ such that $\phi(L^t)\equiv0$,

709: \begin{equation} \label{fred}

710: \int_L\frac{d}{dt}(\ImO^t)=\int_L\,\mathrm{Im}\,\dot\Omega^t=0.

711: \end{equation}

712: To show this deformation preserves the hamiltonian class of L,

713: we need to find a corresponding first order hamiltonian deformation

714: $dh\ip\omega^{-1}$ under which the change in $\ImO^t$,

715: $$

716: \mathcal L_{dh\_\hspace{-.4mm}\shortmid\hspace{1pt}\omega^{-1}}

717: (\ImO^t)\res L=d((dh\ip\omega^{-1})\ip\ImO^t)\res L,

718: $$

719: is $-$Im\,$\dot\Omega^t\res L$. But as Re\,$\Omega^t\res L$ is the induced

720: Riemannian volume form vol$^t$ on $L$, this means we want to solve

721: $$

722: -\mathrm{Im\,}\dot\Omega^t\res L=d(J(dh\ip\omega)\ip\mathrm{\,Re\,}

723: \Omega^t\res L)=d(\widetilde{dh}\ip\mathrm{\,vol}^t)=d(*dh)=\Delta(*h),

724: $$

725: where $J$ is the complex structure and $\,\widetilde{\ }\,$ is the

726: isomorphism $T^*L\to TL$ set up by the induced metric on $L$. So the

727: equation has a solution by the Fredholm alternative and (\ref{fred}).

728: \end{Proof}

729: 

730: Thus for $\phi>0$ we consider the $L^t$s as the \emph{same} as

731: Lagrangian submanifolds (up to hamiltonian deformation) in the

732: \emph{fixed} symplectic manifold

733: $W^t$; it is only the SLag representative that changes as $\Omega^t$

734: varies. We think of this as mirror to a \emph{fixed} holomorphic bundle

735: in a fixed complex structure, with varying HYM connection as the mirror

736: K\"ahler form changes.

737: 

738: \begin{Lemma} \label{lem} In the analogous 2-dimensional situation of

739: SLags in a $K3$ or abelian surface, the obstruction does not occur.

740: \end{Lemma}

741: 

742: \begin{Proof}

743: Choose a real path of complex structures $W^t,\

744: t\in(-\epsilon,\epsilon)$ in complex structure moduli space such that

745: there is a nodal SLag $L^0=L_1^0\cup L_2^0$ in $W^0$. Without loss of

746: generality we can choose the

747: phase of $\Omega^t$ so that both $\omega$ and Im$\,\Omega^t$ pair to

748: zero on the homology class of $L^0$. Now hyperk\"ahler rotate the

749: complex structures so that instead the new Re$\,\Omega^t$ and

750: Im$\,\Omega^t$ pair to zero on the homology class of $L^0$ for all

751: $t$. $L^0$ is now a nodal holomorphic curve $C$ in the central $K3$.

752: We can understand deformations of $C$ via

753: deformations of the ideal sheaf $\mathcal J_C$, with obstructions in

754: $$

755: \mathrm{Ext}^2(\mathcal J_C,\mathcal J_C)\to H^{0,2}(W)\cong\C,

756: $$

757: where the arrow is the trace map and is an isomorphism by Serre

758: duality. Standard deformation theory shows the obstruction is

759: purely cohomological -- it

760: is the derivative of the $H^{0,2}$-component of the class

761: $$

762: [C]\in H^2(W)\cong H^{2,0}(W)\oplus H^{1,1}(W)\oplus H^{0,2}(W).

763: $$

764: But we have fixed this to remain zero by the phase condition,

765: so the curve deforms to all $t$ (really we should assume the family

766: is analytic in $t$ here and extend to $t\in\C$, or just work with

767: first order deformations). Hyperk\"ahler unrotating gives back a

768: family of SLags.

769: \end{Proof}

770: 

771: There is a notion of connect summing Lagrangian submanifolds intersecting

772: in a single point (probably due to Polterovich) -- see for instance

773: Appendix A of \cite{S1} -- which

774: we claim gives the smoothings $L^t$ of the singular $L^0=L_1\cup L_2$.

775: This follows by comparing the local models \cite{J}, \cite{S1} for

776: the Lagrangians; see \cite{TY} where it is studied in more detail for

777: a related purpose, and our conventions (slightly different from those

778: of \cite{S2}) are described. While

779: topologically we are just connect summing $L_1$ and $L_2$ by removing a

780: small 3-ball containing the intersection point from each and gluing the

781: resulting boundary $S^3$s (there are two ways, depending on orientation),

782: symplectically the construction does not explicitly use orientations of

783: the submanifolds. (Effectively we are using their relative orientation --

784: the canonical orientation of the sum of the tangent spaces of $L_1,\ L_2$

785: at the intersection point given by the symplectic structure.)

786: 

787: Giving $L_1$ and $L_2$ in that order produces a Lagrangian, well

788: defined up to hamiltonian isotopy (this will be shown in Section

789: \ref{Ko} in more generality; see (\ref{proj})),

790: $$

791: L_1\#L_2,

792: $$

793: with the singular union $L_1\cup L_2$ a limit point in the hamiltonian

794: isotopy class, which is \emph{not} itself hamiltonian isotopic

795: to $L_1\#L_2$ (we have seen a family of hamiltonian

796: deformations which has limit $L_1\cup L_2$, but the deformations

797: are singular at this limit).

798: 

799: There is also an obvious notion of \emph{graded connect sum},

800: which is in fact what we shall always mean by $\#$. There is a unique

801: grading on $L_1$ compatible with a fixed grading on $L_2$ such that

802: we can give a (continuous) grading to the smoothing $L_1\#L_2$. In the

803: case of connect summing at multiple intersection points (Section

804: \ref{Ko}) there is \emph{at most} one such grading; in general the

805: graded connect sum may not exist.

806: 

807: In $n$ dimensions, if $L_1$ and $L_2$ are graded such that $L_1\#L_2$

808: exists, then on reversing the order of the $L_i$, the graded sum that

809: exists is

810: \begin{equation} \label{+-}

811: L_2\#(L_1[2-n]) \quad\text{in the homology class}\quad

812: [L_2]+(-1)^n[L_1].

813: \end{equation}

814: Here $L[m]$ means the graded Lagrangian $L$ with its grading changed by

815: adding $m\pi$ to $\theta$, and the homology class of $L_1\#L_2$ is

816: $[L_1]+[L_2]$ using the orientations on the $L_i$s induced by the gradings.

817: 

818: This is closely related, as we shall see, to Joyce's obstruction,

819: and the lack of it in dimension 2 (Lemma \ref{lem}). In 2 dimensions,

820: $L_1\#L_2$ and $L_2\#L_1$ are in the same homology class, though

821: by a result of Seidel \cite{S1} \emph{not in general in the same hamiltonian

822: isotopy class},

823: $$

824: L_1\#L_2\not\approx L_2\#L_1,

825: $$

826: importantly (we use $\approx$ to denote equivalence up to hamiltonian

827: deformations). For $t>0$ in the above family $L^t$ is in the constant

828: hamiltonian deformation class of $L_1\#L_2$, for $t<0$ it is in the

829: different class of $L_2\#L_1$, and at $t=0$ it is $L_1\cup L_2$ -- in

830: neither class but in the closure of both. (For

831: complex $t$ the symplectic structure is no longer constant like it is

832: for $t\in\R$, as one can see by following through the hyperk\"ahler

833: rotation; thus we do not get a contradiction to the above statement

834: by going round $t=0$ in $\C$.) In 3 dimensions, however, the

835: corresponding obvious choice for a SLag on the other side of the

836: $\pi^t_1=0$ wall, $L_2\#L_1[-1]$, is in the wrong homology class.

837: 

838: In the case that the $L_i$ are Lagrangian spheres we can see this

839: by going round the wall

840: \begin{equation} \label{wall}

841: \phi(L_1^t)=0\simeq\phi(L_2^t),

842: \end{equation}

843: and using monodromy. In the 2-dimensional $K3$ or $T^4$ case this

844: works as follows.

845: 

846: \begin{figure}[h]

847: \vspace{4mm}

848: \center{

849: \input{k3.pstex_t}

850: \caption{$\left(\int_{L_1}\Omega\right)$-space, as $\Omega$

851: on $K3$ varies, with polar coordinates $(R,\,\phi(L_1))$}}

852: \vspace{2mm}

853: \end{figure}

854: 

855: Consider a disc in complex structure moduli space over which the

856: family of K\"ahler $K3$ surfaces (with constant K\"ahler form)

857: degenerates at the origin to a $K3$ with an ordinary double point

858: (ODP) with the Lagrangian $L_1\cong S^2$ as vanishing cycle.

859: A local model is the standard K\"ahler structure on $x^2+y^2+z^2=u$,

860: over the parameter $u$ in the unit disc in $\C$. Now base-changing

861: by pulling back to the double cover in $u$, $u\mapsto u^2$, we

862: get the 3-fold 

863: $$

864: x^2+y^2+z^2=u^2,

865: $$

866: with a 3-fold ODP which has a small

867: resolution at the origin putting in a holomorphic sphere resolving

868: the central $K3$ fibre $u=0$. Choosing a nowhere-zero holomorphic

869: section $\Omega_u$ of the fibrewise $(2,0)$-forms

870: (using the fact that the relative canonical bundle of either family is

871: trivial), this restricts to zero on the exceptional $\Pee^1$ (which is

872: homologous to the vanishing cycle $L_1$). Therefore the function

873: \begin{equation} \label{zero}

874: \int_{L_1}\Omega_u

875: \end{equation}

876: \emph{has a simple zero at} $u=0$, i.e. it vanishes to order 1 in $u$.

877: (The same expression vanished only as $\sqrt u$ in the original family

878: with the singular fibre, and as such its sign was not well defined; this

879: is because the class $[L_1]$ was defined globally only up to the

880: monodromy $T_{L_1}[L_1]=-[L_1]$, i.e. up to a sign. In our new family

881: the monodromy action $T^2_{L_1}$ is trivial on homology so it makes sense

882: to talk about the homology class $[L_1]$ in any fibre, and (\ref{zero})

883: is single valued.)

884: 

885: Then our loop of complex structures is given by taking the loop

886: $u=e^{it}$ and setting $\Omega^t=\Omega_{e^{it}}$. Pulling back the

887: K\"ahler form from the original family, we get a locally

888: trivial fibre bundle of symplectic manifolds over the

889: circle whose monodromy is the Dehn twist $T_{L_1}^2$ (since the

890: monodromy round the un-base-changed loop is $T_{L_1}$ \cite{S1}). As

891: the family no longer has a singular fibre this monodromy is trivial as

892: a diffeomorphism, but it is a result of \cite{S1},\,\cite{S2} that as a

893: symplectic automorphism \emph{it is non-trivial}. This is possible

894: because although the family is a locally trivial bundle of symplectic

895: manifolds over the punctured disc, over $u=0$ the symplectic form

896: becomes degenerate since it was pulled back via the resolution map.

897: 

898: Measuring $[L_1]$ against $\Omega_u$ as in (\ref{zero}) we see a

899: principle familiar in physics (in issues of `marginal stability',

900: and taught to me by Eric Zaslow) -- we detect a monodromy,

901: like the degree 1 map $S^1\to\C^{\!\times}$ given by $t\mapsto\int_{L_1}

902: \Omega^t$ in this example, by counting wall crossing where a certain

903: real part (here $\int_{L_1}\ImO^t$, or the phase $\phi_1^t$) hits

904: $0\simeq\phi^t_2$.

905: 

906: (Here we can no longer choose the phase of $\Omega$

907: such that $\phi^t_2=\phi(L^t_2)\equiv0$ in the whole family, as the

908: homology class of $L_2$ is not preserved in the family:

909: $$

910: [T_{L_1}^2L_2]=[L_2]+2[L_1].

911: $$

912: However, for a sufficiently small loop about the ODP, i.e. for

913: $\big|\int_{L_1}\Omega\big|$ sufficiently small, this will not

914: affect us much and we can write $\phi^t_2\simeq0$: we are only

915: interested in topological information like winding numbers and

916: $\phi^t_1$ crossing the wall at $\phi^t_2\simeq0$, which are

917: unaffected by small perturbations.)

918: 

919: So instead of going through the $\phi(L_1^t)=\phi(L_2^t)\simeq0$

920: wall we can go round it. If the loop is sufficiently

921: small we do not encounter any more walls where the homology class

922: $[L_1]+[L_2]$ can be split into classes of the same phase to possibly

923: make the SLag a singular union of distinct SLags of equal phase. For

924: instance the wall at phase 0 does not extend past $u=0$ to phase

925: $\phi^t_1=\pi$ (even though there $\mu^t_1=0$) --

926: the phase of $L_1$ is not zero but $\pi$, and is only zero for $L_1$ with

927: the opposite orientation, so it does not exist as a SLag (e.g. in the

928: hyperk\"ahler rotated situation, we are saying there is no complex curve in

929: $L_1$'s homology class to possibly make $L$ the nodal union of $L_1$

930: and something else, there is only an anti-complex curve).

931: So we really can go round the wall; it ends at $u=0$.

932: 

933: So this monodromy

934: description shows that on the other $t\uparrow2\pi$ side of the wall

935: the SLag deforming $L_2\cup L_1$ is in the hamiltonian deformation

936: class

937: \begin{equation} \label{giggs}

938: T_{L_1}^2L\,=\,T_{L_1}^2(L_1\#L_2)\,=\,T_{L_1}^2(T_{L_1}^{-1}L_2)\,

939: \approx\,T_{L_1}L_2\,\approx\,L_2\#L_1,

940: \end{equation}

941: as claimed (for the above equalities see \cite{S1}, \cite{S2}).

942: 

943: Notice that the alternative connect sum description of the

944: above Lagrangian

945: \begin{equation} \label{decomp}

946: L_2\#L_1\,=\,T_{L_1}^2(L_1\#L_2)\,\approx\,T_{L_1}^2(L_1)\#

947: T_{L_1}^2(L_2)\,\approx\,L_1[-2]\#T_{L_1}^2(L_2),

948: \end{equation}

949: does not violate the phase inequality to (\ref{ineq}), as

950: $$

951: -2\pi+\epsilon\approx\phi(L_1[-2])<\phi(T_{L_1}^2(L_2))\simeq0.

952: $$

953: This is why it is important here to keep track of gradings --

954: assigning the phase

955: $\epsilon$ to $\phi(T_{L_1}^2(L_1))$ would give the opposite

956: inequality, but one would not be able to form the above graded

957: connect sum without also shifting

958: the phase of $T_{L_1}^2(L_2)$ by $-2\pi$.

959: 

960: \begin{figure}[h]

961: \vspace{4mm}

962: \center{

963: \input{m5.pstex_t}

964: \caption{$\left(\int_{L_1}\Omega\right)$-space, as $\Omega$

965: on a 3-fold varies, with polar coordinates $(R,\,\phi(L_1))$}}

966: \end{figure}

967: 

968: The 3-fold case (which Dominic Joyce has also, independently, considered)

969: is slightly different; we need only take a single Dehn twist

970: $T_{L_1}$ corresponding to the local family

971: $$

972: x^2+y^2+z^2=u,

973: $$

974: over $u\in\C$ to get a winding number one loop in the phase of $L_1$.

975: This is because

976: $$

977: T_{L_1}L_1\approx L_1[1-n]

978: $$

979: in dimension $n$, so in 3 dimensions the homology class $[L_1]$ is

980: preserved instead of being reversed. The corresponding picture is

981: displayed in Figure 2.

982: 

983: Again there is a SLag on the other side of the $\phi=0$ wall, but

984: it is in the wrong homology class $[L_2]$:

985: \begin{equation} \label{cole}

986: T_{L_1}L\approx L_2.

987: \end{equation}

988: Analogously to (\ref{decomp}) this has a number of decompositions

989: as connect sums induced by monodromy,

990: $$

991: T_{L_1}(L_1\#L_2)\,\approx\,L_1[-2]\#(L_2\#(L_1[-1]))\,\approx\,L_2

992: \,\approx\,(L_1\#L_2)\#(L_1[\,1\,]),

993: $$

994: none of which violate the phase inequality (\ref{ineq}). The only other

995: obvious choice for a (S)Lag on the other side of the $\phi=0$ wall

996: (given the $K3$ result) is $T_{L_1}^2(L_1\#L_2)\approx L_2\#(L_1[-1])$;

997: this however is also in the wrong homology class, and in any case

998: does violate (\ref{ineq}) and so, by Joyce's analysis, should not

999: be represented by a SLag. Thinking of $T_{L_1}^2$ as rotating through

1000: $-4\pi$ in Figure 2, it is at roughly $-3\pi$ that the phase inequality

1001: (\ref{ineq}) gets violated, and the $-\pi$ rotation of $L_2$ splits

1002: as a SLag into the union of the $-\pi$ rotations of $(L_1\#L_2)$ and

1003: $L_1[\,1\,]$: these both have phase approximately zero.

1004: 

1005: 

1006: \subsection*{A holomorphic bundle example}

1007: 

1008: These phenomena are similar to wall-crossing in bundle theory on the

1009: complex side -- in a real one-parameter family of K\"ahler forms, for

1010: fixed complex structure, stable holomorphic bundles for $t>0$ can become

1011: semistable at $t=0$ and unstable for $t<0$.

1012: 

1013: An example that mirrors Joyce's is the following. Suppose we

1014: have two stable bundles (or coherent sheaves) $E_1$ and $E_2$ with

1015: $$

1016: \mathrm{Ext}^1(E_2,E_1)\cong\C.

1017: $$

1018: This is $H^1(E_1\otimes E_2^*)$ in the case of

1019: bundles and is the mirror \cite{K} of the one dimensional

1020: Floer cohomology $HF^*(L_2, L_1)\cong\C$ that is defined by the single

1021: intersection point of $L_1$ and $L_2$ (see Section \ref{Ko} for more

1022: details of this, and an explanation of why we are dealing with Ext$^1$

1023: and $HF^1$ here). We then form $E$ from this extension class

1024: \begin{equation} \label{E}

1025: 0\to E_1\to E\to E_2\to0.

1026: \end{equation}

1027: Take a family of K\"ahler forms $\omega^t$ such that

1028: $\mu^t(E_2)-\mu^t(E_1)$ is the same sign as $t$ (here $\mu^t(F)=

1029: c_1(F)\,.\,(\omega^t)^{n-1}/$\,rk\,$(F)$ is the slope of $F$ with

1030: respect to $\omega^t$). Supposing that the $E_i$ are stable for all

1031: $t\in(-\epsilon,\epsilon)$, we claim that $E$ is stable

1032: for sufficiently small $t>0$, while it is destabilised by $E_1$ for

1033: $t\le0$.  Without loss of generality take $\mu^t(E_2)=\mu$ fixed,

1034: and $\mu^t(E_1)=\mu-t$. As $E_2$ is stable, for $t$ sufficiently small

1035: there are no subsheaves of $E_2$ of slope greater than $\mu-t$, so for

1036: any stable destabilising subsheaf $F$ of $E$, the composition

1037: $$

1038: F\into E\to E_2

1039: $$

1040: cannot be an injection (unless it is an isomorphism, but (\ref{E}) does

1041: not split. So $F\cap E_1\ne0$, and the quotient $Q=F/(F\cap E_1)$ has

1042: slope $\mu(Q)>\mu(F)>\mu-t$ by the stability of $F$ and

1043: instability of $E$. But $Q$ injects into $E_2$, which we know is

1044: impossible.

1045: 

1046: In the 2-dimensional case, by Serre duality Ext$^1(E_1,E_2)

1047: \cong\,$Ext$^1(E_2,E_1)^*\cong\C$on $K3$ or $T^4$, so for $t<0$ we

1048: can instead form an extension

1049: \begin{equation} \label{ext}

1050: 0\to E_2\to E'\to E_1\to0,

1051: \end{equation}

1052: to give a new bundle $E'$ which is also stable, and has the same Mukai

1053: vector

1054: $$

1055: v(E')=v(E_1)+v(E_2);

1056: $$

1057: compare (\ref{+-}). At $t=0$ we take the (polystable)

1058: bundle

1059: $$

1060: E_1\oplus E_2.

1061: $$

1062: This is because the semistable extension (\ref{E}) no longer admits a

1063: Hermitian-Yang-Mills metric, but $E_1\oplus E_2$ does. Also, the algebraic

1064: geometry of the moduli problem shows that while a semistable bundle gets

1065: identified in the moduli space with the other (``S-equivalent'') sheaves

1066: in the closure of its gauge group orbit, there is a distinguished

1067: representative of its

1068: equivalence class -- the polystable direct sum (of the Jordan-H\"older

1069: filtration, which here is $E_1\oplus E_2$).

1070: 

1071: Thus, while the HYM connections vary, the bundle

1072: has only 3 different holomorphic structures -- for $t>0,\ t=0,$ and $t<0$.

1073: Put another way (to spell out the analogy with the Lagrangians $L^t,\ 

1074: L_1,\ L_2$) as $\omega_t$ varies with $t>0$ we take different points in

1075: a fixed complexified gauge group orbit, and at $t=0$ we take as limit point

1076: something in a different orbit that is nonetheless in the closure of the

1077: $t>0$ (and $t<0$) orbit. The \emph{stable} deformations of the polystable

1078: $E_1\oplus E_2$ (which we are thinking of as the mirror of the singular

1079: union $L_1\cup L_2$, of course) are precisely (\ref{E}) for $t>0$ and

1080: (\ref{ext}) for $t<0$.

1081: 

1082: In the 3-fold case, however, Serre duality gives Ext$^2(E_1,E_2)\cong\,

1083: $Ext$^1(E_2,E_1)^*\cong\C$ instead, and so no stable extension (\ref{ext}).

1084: In fact one would expect there to be no stable bundle with the right Chern

1085: classes; instead the one dimensional Ext$^2$ gives us a complex $E'$ in

1086: the derived category $D^b(M)$ fitting into an exact sequence of complexes

1087: $$

1088: 0\to E_2\to E'\to E_1[-1]\to0,

1089: $$

1090: where $E_1[-1]$ is $E_1$ shifted in degree by one place to the right as

1091: a complex. This has Mukai vector

1092: $$

1093: v(E')=v(E_2)-v(E_1),

1094: $$

1095: compare (\ref{+-}). Thus, just as in the

1096: case of SLags, as we pass through $t=0$ there is no natural stable

1097: object on the other side in the same homology class in 3 dimensions

1098: (though there is in 2 dimensions) and so an element of the appropriate

1099: moduli space disappears.

1100: 

1101: In fact, as in the Lagrangian example, the natural stable object

1102: on the other side of the wall is $E_2$ if we consider monodromy.

1103: The mirror of the symplectic Dehn twists of above are described

1104: in \cite{ST} (in the case that the bundles $E_i$ are \emph{spherical}

1105: in the sense of \cite{ST}: Ext$^k(E_i,E_i)\cong H^k(S^n;\C)$; this is

1106: the natural mirror analogue of the

1107: $L_i$s being spheres). These are the twists $T_{E_1}$ of \cite{ST}

1108: on the derived category of the Calabi-Yau that act on the extension

1109: bundle $E$ of (\ref{E}) to give precisely the extension (\ref{ext}),

1110: $$

1111: T_{E_1}^2E=E'

1112: $$

1113: (compare (\ref{giggs})), as a short calculation using \cite{ST}

1114: shows. Similarly

1115: $$

1116: T_{E_1}E=E_2,

1117: $$

1118: the analogue of (\ref{cole}). (In both of these calculations it is

1119: important to calculate this monodromy in the derived

1120: category; in the $K3$ case the action of $T_{E_1}^2$ is trivial on

1121: K-theory and cohomology, and we cannot distinguish between (\ref{E})

1122: and (\ref{ext}), but they are very different as holomorphic bundles

1123: and as elements of the derived category.)

1124: 

1125: The mirror wall crossing, with a SLag splitting into two and then

1126: disappearing, is interpreted in \cite{DFR} (and in \cite{SV} in a

1127: different case) as the

1128: state it represents decaying as we reach a point of `marginal

1129: stability'. Despite this dealing with only SLags (and so with only

1130: a priori \emph{stable} Lagrangians in our mathematical sense of

1131: stability), this suggestive language does in fact have something to say

1132: about the stability, in our sense of group actions, of (non-special)

1133: Lagrangians, by considering the nodal limit $L_1\cup L_2$ to be a

1134: \emph{semistable} Lagrangian.

1135: 

1136: Thus the Lagrangian $L_1\#L_2$ (which

1137: \emph{always exists as a Lagrangian} as the complex

1138: structure varies with fixed K\"ahler form) becomes semistable

1139: at $t=0$ and is represented by something in a different orbit of the

1140: hamiltonian deformation symmetry group (but in the closure of the

1141: original orbit), and is unstable for $t<0$ so exists there only as a

1142: Lagrangian and \emph{not} as a SLag. This, and the bundle analogue

1143: described above, leads us to think of the Lagrangian $L_1$

1144: as destabilising $L=L_1\#L_2$ when $\phi(L_1)\ge\phi(L_2)$. This

1145: motivates the now obvious definition of stability in Section \ref{st};

1146: first we explain more about the connections to mirror symmetry, and

1147: generalisations to connect sums at more intersection points.

1148: 

1149: 

1150: \section{Relationship to Kontsevich's mirror conjecture} \label{Ko}

1151: 

1152: The inspiration behind most of this paper is of course Kontsevich's

1153: mirror conjecture \cite{K}. In particular, Kontsevich proposes that

1154: the graded vector spaces Ext$^*$ and $HF^*$ should be isomorphic for

1155: mirror choices of bundles $E_i$ and graded Lagrangians $L_i$

1156: (or more exotic objects in their derived categories)

1157: $$

1158: HF^*(L_2,L_1)\cong\,\mathrm{Ext}^*(E_2,E_1);

1159: $$

1160: this corresponds to the equality of (graded) morphisms on both sides.

1161: Here $HF^*$ is Floer cohomology \cite{Fl} -- a symplectic refinement

1162: of the intersection number of $L_1$ and $L_2$ -- which can be $\mathbb

1163: Z$-graded for graded Lagrangians \cite{S2}, whenever it is defined

1164: \cite{FO3}, \cite{Fu1}. (More precisely it is the cohomology of a chain

1165: complex built out of the free vector space generated by the intersection

1166: points, with the differential defined by counting holomorphic

1167: discs with boundary in the Lagrangians running from one intersection

1168: point to another.) In mirror symmetry, and so in this paper, one

1169: should only really consider those Lagrangians whose Floer cohomology is

1170: well defined \cite{Fu1}.

1171: 

1172: Thus the point of intersection of the $L_1$ and $L_2$ of the last

1173: section define the Floer cohomology $HF^*(L_2,L_1)\cong\C$, and the

1174: grading of \cite{S2} is designed specifically so that $L_1\#L_2$ can

1175: be graded precisely when the relative gradings of the $L_i$ force

1176: the Floer cohomology to be concentrated in degree 1; $HF^*(L_2,L_1)=

1177: HF^1(L_2,L_1)$. We then think of the connect sum $L_1\#L_2$ as being

1178: mirror to the extension (\ref{E}) defined by Ext$^1(E_2,E_1)\cong\C$.

1179: Fukaya, Seidel, and perhaps others have also proposed that Lagrangian

1180: connect sum should be mirror to extensions \cite{Fu2}, \cite{S3}.

1181: 

1182: We also consider connect sums of Lagrangians intersecting at $n$

1183: points $p_i$. Then the connect sum is not unique up to hamiltonian

1184: deformation: $H^1$ is added to the Lagrangian as loops between the

1185: intersection points, giving additional deformations of its hamiltonian

1186: isotopy class. The upshot is that there is a scaling of the neck of

1187: the connect sum at each intersection point; we denote any such

1188: resulting Lagrangian by $L_1\#L_2$. Since we insist on all intersection

1189: points having Floer (Maslov) index one (so that the connect sum can be

1190: graded), the Floer differential vanishes in this case, and

1191: these scalings define a class in $HF^1(L_2,L_1)$.

1192: 

1193: Deformations (up to those which are hamiltonian) as such a connect sum

1194: are given by the elements of

1195: $$

1196: H^1(L_1\#L_2)\cong H_{n-1}(L_1\#L_2)

1197: $$

1198: spanned by the $S^{n-1}$ vanishing cycles $S_i$ at the points of

1199: intersection $p_i\in L_1\cap L_2$. Given a particular connect sum,

1200: the deformation represented by $\sum_ia_iS_i$ simply scales the local

1201: gluing parameter in a Darboux chart around each $p_i$ by a factor

1202: $(1+a_i)$ (here $a_i$ is considered to be infinitesimal). Since

1203: the sum of these spheres

1204: separates $L_1\#L_2$ into $L_1\backslash\cup\{p_i\}$ and

1205: $L_2\backslash\cup\{p_i\}$ and so is zero in homology

1206: $$

1207: \sum_i[S_i]=\pm\partial[L_1\backslash\cup\{p_i\}]=

1208: \mp\partial[L_2\backslash\cup\{p_i\}]=0\in H_{n-1}(L_1\#L_2),

1209: $$

1210: the infinitesimal deformation represented by $\sum_iS_i$ is zero

1211: (it is pure hamiltonian) and dividing out gives the projectivisation

1212: \begin{equation} \label{proj}

1213: \Pee(\oplus_i\R_{p_i}).

1214: \end{equation}

1215: (Replace $\R$ by $\C$ when including flat bundles and their gluing

1216: parameters at the $p_i$s.)  This

1217: explains the earlier claim that connect sums at one intersection

1218: point are uniquely defined up to hamiltonian deformations.

1219: More precisely, when holomorphic discs are

1220: taken into account and we consider only those Lagrangians whose Floer

1221: cohomology is defined \cite{FO3}, hamiltonian deformation classes of

1222: connect sums \emph{whose Floer cohomology can be defined}

1223: should be parameterised by $\Pee(HF^1(L_2,L_1))$.

1224: (On the mirror side \emph{isomorphism classes} of extensions of $E_2$

1225: by $E_1$ are parametrised by $\Pee\mathrm{\,Ext}^1(E_2,E_1)$.)

1226: 

1227: We would then expect that the resulting connect sum has a canonical

1228: homomorphism from $L_1$; that is there should be a canonical element

1229: $$

1230: \id_{L_1}\in HF^0(L_1,L_1\#L_2)

1231: $$

1232: for any graded Lagrangians $L_i$ for which the graded connect sum

1233: exists. While a local model suggests this is true (see for instance

1234: \cite{TY}), a complete proof is

1235: still not available. This homomorphism we think of as expressing $L_1$

1236: as a \emph{subobject} of $L_1\#L_2$; i.e. as giving an injection.

1237: It should be emphasised that subobject does not make sense in a

1238: triangulated category such as the derived Fukaya category of Lagrangians;

1239: in the context of the derived category of sheaves, subobject

1240: only makes sense for an abelian category such as that of the sheaves

1241: themselves (i.e. complexes with cohomology in degree zero only).

1242: What we are proposing is that it also makes sense in the category of

1243: (complexes of sheaves mirror to) graded Lagrangians, and is vital to

1244: make definitions of stability (which involve such subobjects). While

1245: there are now more Homs to consider, in particular those of higher

1246: order (i.e. Homs to Lagrangians shifted in phase by some $2\pi n$),

1247: the targets of these Homs have higher phase and so do not disturb the

1248: definition of stability below -- this is seemingly a huge piece of luck

1249: that means we can extend the stability condition for bundles to all

1250: Lagrangians. For similar reasons, the many

1251: connect sum decompositions of the $L_i$s given in the last section

1252: also do not destabilise them.

1253: 

1254: There are other operations, however, which can also be thought of

1255: as Ext$^1$-type extensions. For instance, taking the product of a

1256: single Lagrangian curve $L_1$ in $T^2$ with a (graded) connect sum

1257: $L_2\#L_3$ in another $T^2$, we get a

1258: Lagrangian $L_1\times(L_2\#L_3)$ in $T^4$ which is some kind of

1259: extension of the Lagrangians $L_1\times L_2$ and $L_1\times L_3$ in

1260: $T^4$. Supposing that the $L_i$s are mirror to some (complexes of)

1261: sheaves $E_i$, and that the connect sum $L_2\#L_3$ is mirror to an

1262: extension represented by an element $e\in$\,Ext$^1(E_3,E_2)$. Then

1263: by the K\"unneth formula for sheaf cohomology, we see that

1264: $L_1\times(L_2\#L_3)$ is indeed mirror to an extension

1265: $$

1266: \id\,\otimes e\in\mathrm{Hom\,}(E_1,E_1)\otimes\mathrm{Ext}^1

1267: (E_3,E_2)=\mathrm{Ext}^1(E_1\boxtimes E_3,E_1\boxtimes E_2),

1268: $$

1269: and so this sort of relative connect sum (which is not $\#$ on

1270: $T^4$: $L_1\times L_2$ and $L_1\times L_3$ do not intersect

1271: transversely) should also be considered.

1272: 

1273: So we consider Lagrangians $L_1,\,L_2$ intersecting \emph{cleanly}

1274: (see e.g. \cite{S1} Definition 2.1), that is $N=L_1\cap L_2$ is a smooth

1275: submanifold, and $TN=TL_1\res N\cap TL_2\res N$. Basic results of

1276: Weinstein allow us to identify a neighbourhood of $N$ with a

1277: neighbourhood of the zero section $N$ in $T^*N\oplus E$, where the total

1278: space of $T^*N$ has its canonical symplectic structure, and

1279: $$

1280: E\equiv(TL_1\res N)/TN\,\oplus\,(TL_2\res N)/TN

1281: $$

1282: is the annihilator, under the symplectic form, of $TN\subset TX\res N$

1283: (to which the symplectic form therefore descends, making $E$ a

1284: symplectic bundle).

1285: 

1286: Choosing a metric on $E$, compatible with its symplectic structure,

1287: such that its transverse

1288: subbundles $(TL_1\res N)/TN,\ (TL_2\res N)/TN$ are orthogonal, we can now

1289: perform the family connect sum of these, over the base $N$, since

1290: the local model in \cite{S1} is $O(n)$ invariant. As before we insist

1291: that this can be compatibly graded again denote it by $\#$; given a

1292: grading on $L_1$ there will be at most one grading on $L_2$ such that

1293: this graded relative connect sum exists.

1294: 

1295: It should be noted that although such a clean intersection could be

1296: hamiltonian isotoped to be transverse, the resulting intersection points

1297: would not necessarily all be of Floer/Maslov index one, and so the

1298: pointwise graded connect sum could not be formed at every point; we

1299: would end up with an immersed Lagrangian. Studying which immersed

1300: Lagrangians should be included in the Fukaya category, and which

1301: embedded Lagrangians they should be considered equivalent to, is

1302: an important part of mirror symmetry and will need to be better understood

1303: to refine our conjecture. For instance forming extensions of bundles

1304: which also have nonzero homorphisms between them would appear to

1305: be mirror to forming connect sums between graded Lagrangians at index

1306: one intersection points, leaving the index zero intersection points

1307: immersed. In general one would like to consider two objects of the Fukaya

1308: category to be equivalent if their Floer cohomologies with any other

1309: objects are the same. This would include hamiltonian deformation

1310: equivalence, but also more exotic equivalences for immersed

1311: Lagrangians (thanks to Paul Seidel for pointing this out to me).

1312: A start in understanding the Floer cohomology of immersed

1313: Lagrangians is \cite{Ak}; in the present paper we are largely ignoring

1314: singularities.

1315: 

1316: 

1317: \section{Stability} \label{st}

1318: 

1319: \begin{Definition}

1320: Take graded Lagrangians $(L_1,\theta_1)$ and $(L_2,

1321: \theta_2)$, hamiltonian isotoped to intersect cleanly, and such

1322: that the graded (relative) Lagrangian connect sums

1323: $(L_1\#L_2,\theta_1\#\theta_2)$ exist as above. Then a

1324: Lagrangian $L$ of Maslov class zero is said to be

1325: destabilised by the $L_i$ if it is hamiltonian isotopic to such an

1326: $L_1\#L_2$, and the phases (\emph{real} numbers, induced by the

1327: gradings) satisfy

1328: $$

1329: \phi(L_1)\ge\phi(L_2).

1330: $$

1331: If $L$ is not destabilised by any such $L_i$ then it is called stable.

1332: \end{Definition}

1333: 

1334: \begin{Remarks}

1335: \begin{description} \item[\ \ $\bullet$\ \ ]

1336: There is an obvious notion of a flux homomorphism for isotopies

1337: of smooth Lagrangians, taking a deformation to an element of

1338: $H^1(L;\R)$ (and linearising to give the usual deformation theory of

1339: Lagrangians). Namely, take a deformation $\Phi_t(L)$ through a

1340: vector field $X_t,\ t\in[0,1]$ to the one form

1341: $$

1342: \int_0^1(X_t\ip\omega)dt\in H^1(L;\R).

1343: $$

1344: Alternatively, the homomorphism takes a loop $\gamma\subset L$,

1345: tracing out the 2-cycle $f(\gamma\times[0,1])$ in $W$ under the

1346: isotopy, to the real number $\int_{\gamma\times[0,1]}\omega$.

1347: (See Chapter 10 of \cite{MS} for the analogous map for

1348: symplectomorphisms.) If the isotopy $\Phi_t$ is hamiltonian, the

1349: flux is zero; the converse is also easily proved using the methods of

1350: (\cite{MS} Theorem 10.12): we may assume without loss of generality

1351: that the 1-form $\int_0^1X_t\ip\omega$ is identically zero in

1352: $\Omega^1(L)$. [To see this, write the 1-form as

1353: $d\phi$, and compose the deformation with

1354: the time one map of the hamiltonian flow with vector field

1355: $d\phi\ip\omega^{-1}$; this does not alter the flux in $H^1(L;\R)$ or

1356: the property of being hamiltonian.] Then let $\Sigma^s$ be the closed

1357: 1-form on $L$ defined by

1358: $$

1359: \Sigma^s=\int_0^sX_t\ip\omega dt,

1360: $$

1361: and let $\Psi^s_t$ be the corresponding flow through time $t$. Then

1362: the flow $\phi_t=\Psi^t_t\circ\Phi_t$ is the corresponding

1363: \emph{hamiltonian} flow from $\Phi_0(L)$ to $\Phi_1(L)$; see

1364: \cite{MS}.

1365: 

1366: Thus it is not too hard to check if two Lagrangians

1367: are hamiltonian deformations of each other, at least through

1368: smooth Lagrangians, if we know they are deformations of each

1369: other as Lagrangians. This second condition, however, is harder

1370: to test, as the results of \cite{S1} demonstrate.

1371: 

1372: \item[\ \ $\bullet$\ \ ] As mentioned in the last section,

1373: holomorphic discs are

1374: crucial in both mirror symmetry and Floer cohomology; thus one

1375: should perhaps restrict attention in the above definition to

1376: those Lagrangians whose Floer cohomology is defined \cite{FO3},

1377: 

1378: \item[\ \ $\bullet$\ \ ]As pointed out to me by Conan Leung, this

1379: definition and the resulting conjecture below may only be reasonable

1380: close to the

1381: large complex structure limit point where the mirror symmetric

1382: arguments used to motivate the conjecture are most valid.

1383: \end{description}

1384: \end{Remarks}

1385: 

1386: \begin{Conj}

1387: A Lagrangian of Maslov class zero has a special Lagrangian in its

1388: hamiltonian deformation class if and only if it is stable, and this

1389: SLag representative is unique.

1390: \end{Conj}

1391: 

1392: Again, we have been vague about singularities: which we allow, and what

1393: hamiltonian deformation equivalence would mean for them. We might also

1394: want to restrict attention to those Lagrangians whose Floer cohomology

1395: exists \cite{FO3}, and whose Oh spectral sequence $H^*(L)\Rightarrow

1396: HF^*(L,L)$ \cite{Oh} degenerates; this will be discussed more in \cite{TY}.

1397: We might also want to restrict to Lagrangians whose phase function

1398: varies only by a certain bounded amount; in the example worked out in

1399: \cite{TY} this is required. In \cite{TY}

1400: it is shown there that the gradient of the

1401: norm-squared $|m|^2$ of the moment map can be taken to be the mean

1402: curvature vector of the Lagrangian, so mean curvature flow (which

1403: is hamiltonian for Maslov class zero) should

1404: converge to this SLag representative if the Lagrangian is stable

1405: and the phase satisfies certain bounds.

1406: 

1407: 

1408: \section{The 2-torus}

1409: 

1410: Everything works rather simply on $T^2$; Grayson \cite{G}, building

1411: on work of Gage, Hamilton and others (e.g. \cite{GH}), has shown that

1412: mean curvature flow for curves (of Maslov class zero) converges to

1413: straight lines and so we get the mirror symmetric analogue 

1414: of Atiyah's classification \cite{At} of sheaves on an elliptic curve

1415: -- they are basically all sums of stable sheaves. The only exceptions

1416: are the non-trivial extensions of certain

1417: sheaves by themselves; these correspond to thickenings of the

1418: corresponding special Lagrangian (giving fat SLags, as they are known

1419: in Britain, or multiply-wrapped cycles in physics speak).

1420: 

1421: \begin{figure}[ht]

1422: \vspace{1mm}

1423: \center{\input{T2.pstex_t}}

1424: \caption{$L_1\# L_2$ and $L_2\#(L_1[1])$, equivalent SLags, and

1425: their mirror sheaves}

1426: \vspace{1mm}

1427: \end{figure}

1428: 

1429: We give an example to demonstrate why one cannot form smooth

1430: \emph{unstable} Lagrangians on $T^2$ in Figure 3.

1431: First, giving $L_1$ and $L_2$ the gradings such that their

1432: phases are 0 and $\pi/4$, we expect $L_1\#L_2$ to be stable, and

1433: indeed we see it is hamiltonian deformation equivalent to the slope

1434: $1/2$ SLag mirror to the stable extension $E$ of $\OO$ by $\OO(p)$

1435: (where $p$ is a basepoint of $T^2$ with corresponding line bundle mirror

1436: to the diagonal SLag drawn).

1437: 

1438: If one then tries to form an unstable SLag $L_2\#L_1$, the \emph{graded}

1439: connect sum does not exist -- the phase would become

1440: discontinuous. To form $L_2\#L_1$ we see from the diagram that we have to

1441: take the phase of $L_1$ to be $\pi$, thus reversing its orientation, and

1442: in fact forming $L_2\#(L_1[1])$.

1443: Then the stability inequality (\ref{ineq}) is not violated, and in fact

1444: this Lagrangian is stable and hamiltonian deformation equivalent to

1445: the SLag in $T^2$ represented by the vertical edge of the square (and so

1446: drawn with a little artistic license in Figure 3). Under the mirror

1447: map this corresponds to replacing the extension Ext$^1$ class by a Hom

1448: (as we have shifted complexes of sheaves by one place) and taking the

1449: cone of this in the derived

1450: category; this is the cokernel $\OO_p$ of Figure 3.

1451: 

1452: As pointed out to me by Markarian and Polishchuk, one can play with

1453: lots of pictures of connect sums on tori to recover descriptions

1454: of certain moduli of sheaves, their special cycles (for instance

1455: where one connect-sum neck parameter goes to zero), and so forth,

1456: giving results similar to some of those in \cite{FO}.

1457: 

1458: This example can be extended to show that we cannot form the graded

1459: connect sum $L_1\#L_2$ of any two Lagrangians (via a class in

1460: $HF^*(L_1,L_2)$) if $\phi(L_1)>\phi(L_2)$. Namely, replace

1461: $L_1$ and $L_2$ by their hamiltonian deformation equivalent SLag

1462: representatives, which are straight lines of constant phase $\theta_i=

1463: \phi(L_i)$. As Figure 3 shows, $L_1\#L_2$ can be compatibly

1464: graded about an intersection point if and only if we have the local

1465: inequalities

1466: $$

1467: \theta_2>\theta_1>\theta_2-\pi.

1468: $$

1469: Thus we require $\phi(L_2)>\phi(L_1)$. (We will explain this kind

1470: of phenomenon more generally in \cite{TY} in terms of the grading

1471: on Floer cohomology.) Each intersection point is

1472: Floer coclosed since the Floer grading is the same

1473: as the relative orientation of the Lagrangians, mod 2, and this

1474: is the same at each intersection point of the straight lines. So each

1475: possible connect sum of the SLags defines a class in $HF^*$, and

1476: any other connect sum, defined on hamiltonian deformations of $L_1$

1477: and $L_2$ by a class in $HF^*$, will be hamiltonian deformation

1478: equivalent to the appropriate connect sum of the SLags, and so

1479: satisfy the same phase inequality.

1480: 

1481: If two smooth Lagrangians have the same phase then their

1482: representative SLags will either be the same or disjoint parallel

1483: SLags. Either way there are no connect sums (though as mentioned above

1484: to account for the mirror symmetry of bundles one should also include

1485: non-trivial thickenings of SLags in the Fukaya category).

1486: 

1487: So unstable Lagrangians do not exist, and by the result of \cite{G}

1488: mentioned earlier, the conjecture is true on $T^2$.

1489: 

1490: Thus complex dimension 1 is too simple -- in trying to make the phase

1491: of one Lagrangian become larger than the phase of another, the two must

1492: cross, thus reversing their relative orientations and changing the

1493: order of the connect sum. Far more complicated phenomena arise

1494: in 2 and 3 dimensions, however.

1495: 

1496: 

1497: \begin{thebibliography}{MMMS}

1498: 

1499: \bibitem[Ak]{Ak} M. Akaho,

1500: \textit{An intersection theory of Lagrangian immersions}, preprint.

1501: 

1502: \bibitem[At]{At} M. F. Atiyah,

1503: \textit{Vector bundles over an elliptic curve},

1504: Proc. London Math. Soc. \textbf{7} (1957), 414--452.

1505: 

1506: \bibitem[AP]{AP} D. Arinkin and A. Polishchuk,

1507: \textit{Fukaya category and Fourier transform}, preprint

1508: math.AG/9811023.

1509: 

1510: \bibitem[BBHM]{BBHM}

1511: C. Bartocci, U. Bruzzo, D. Hernandez Ruiperez and J. Munoz Porras,

1512: \textit{Mirror symmetry on $K3$ surfaces via Fourier-Mukai transform},

1513: Comm. Math. Phys. \textbf{195} (1998), 79--93.

1514: 

1515: \bibitem[Ch]{Ch} J. Chen,

1516: \textit{Lagrangian sections and holomorphic $U(1)$-connections}, preprint.

1517: 

1518: \bibitem[D]{D} S. K. Donaldson, 

1519: \textit{Moment maps and diffeomorphisms}, Asian Jour. Math. \textbf{3}

1520: (1999), 1--16.

1521: 

1522: \bibitem[DT]{DT} S. K. Donaldson and R. P. Thomas,

1523: \textit{Gauge theory in higher dimensions.} In: The Geometric Universe:

1524: Science, Geometry and the work of Roger Penrose, S. A. Huggett et al

1525: (eds), Oxford University Press, 1998.

1526: 

1527: \bibitem[Do]{Do} M. R. Douglas,

1528: \textit{D-branes, Categories and N=1 Supersymmetry},

1529: preprint hep-th/0011017.

1530: 

1531: \bibitem[DFR]{DFR} M. R. Douglas, B. Fiol, and C. R\"omelsberger,

1532: \textit{Stability and BPS branes}, preprint hep-th/0002037.

1533: 

1534: \bibitem[FO]{FO}

1535: B. Feigin and A. Odesskij,

1536: \textit{Vector bundles on an elliptic curve and Sklyanin algebras},

1537: q-alg/9509021. In: Topics in quantum groups and finite-type invariants,

1538: B. Feigin et al (eds), AMS Transl. 185 (38), 65--84, 1998.

1539: 

1540: \bibitem[Fl]{Fl} A. Floer,

1541: \textit{Morse theory for Lagrangian intersections},

1542: Jour. Diff. Geom. \textbf{28} (1988), 513--547.

1543: 

1544: \bibitem[Fu1]{Fu1} K. Fukaya,

1545: \textit{Floer homology and Mirror symmetry I}, preprint (1999)

1546: http://www.kusm.kyoto-u.ac.jp/$\sim$fukaya/fukaya.html.

1547: 

1548: \bibitem[Fu2]{Fu2} K. Fukaya,

1549: \textit{Mirror symmetry of Abelian variety and multi theta functions},

1550: preprint (2000) http://www.kusm.kyoto-u.ac.jp/$\sim$fukaya/fukaya.html.

1551: 

1552: \bibitem[FO3]{FO3} K. Fukaya, Y.-G. Oh, H. Ohta, and K. Ono,

1553: \textit{Lagrangian intersection Floer theory -- anomoly and obstruction},

1554: book to appear in International Press.

1555: 

1556: \bibitem[GH]{GH} M. Gage and R. Hamilton,

1557: \textit{The shrinking of convex plane curves by the heat equation},

1558: Jour. Diff. Geom. \textbf{23} (1986), 69--96.

1559: 

1560: \bibitem[G]{G} M. Grayson,

1561: \textit{Shortening embedded curves},

1562: Ann. of Math. \textbf{129} (1989), 71--111.

1563: 

1564: \bibitem[Gr]{Gr} M. Gross,

1565: \textit{Special Lagrangian Fibrations I: Topology.} In: Integrable

1566: systems and algebraic geometry (Kobe/Kyoto, 1997), 156--193,

1567: World Sci. Publishing, 1998.

1568: 

1569: \bibitem[Ha]{Ha} F. R. Harvey,

1570: \textit{Spinors and Calibrations.} Academic Press, San Diego, 1990.

1571: 

1572: \bibitem[HL]{HL} F. R. Harvey and H. B. Lawson,

1573: \textit{Calibrated Geometries}, Acta Math. \textbf{148} (1982), 47--157.

1574: 

1575: \bibitem[H]{H} N. Hitchin,

1576: \textit{Lectures on Special Lagrangian Submanifolds},

1577: preprint math.DG/9907034.

1578: 

1579: \bibitem[J]{J} D. Joyce,

1580: \textit{On counting special Lagrangian homology 3-spheres},

1581: preprint hep-th/9907013.

1582: 

1583: \bibitem[K]{K} M. Kontsevich, 

1584: \textit{Homological Algebra of Mirror Symmetry}, International Congress

1585: of Mathematicians, Z\"urich 1994. Birkh\"auser, 1995.

1586: 

1587: \bibitem[KS]{KS} M. Kontsevich and Y. Soibelman,

1588: \textit{Homological mirror symmetry and torus fibrations}, preprint

1589: math.SG/0011041.

1590: 

1591: \bibitem[L]{L} G. Lawlor,

1592: \textit{The angle criterion}, Inv. Math. \textbf{95} (1989), 437--446.

1593: 

1594: \bibitem[Le]{Le} N. C. Leung,

1595: \textit{Mirror symmetry without corrections}, preprint math.DG/0009235.

1596: 

1597: \bibitem[LYZ]{LYZ} N. C. Leung, S.-T. Yau and E. Zaslow,

1598: \textit{From Special Lagrangian to Hermitian-Yang-Mills via Fourier-Mukai

1599: Transform}, preprint math.DG/0005118.

1600: 

1601: \bibitem[MS]{MS} D. McDuff and D. Salamon,

1602: \textit{Introduction to symplectic topology}, 2nd edition, Oxford

1603: University Press, 1998.

1604: 

1605: \bibitem[MMM]{MMM} Y. Manin,

1606: \textit{Moduli, Motives, Mirrors}, preprint math.AG/0005144.

1607: 

1608: \bibitem[MMMS]{MMMS} M. Marino, R. Minasian, G. Moore and A. Strominger,

1609: \textit{Nonlinear Instantons from Supersymmetric p-Branes},

1610: preprint hep-th/9911206.

1611: 

1612: \bibitem[Oh]{Oh} Y.-G. Oh,

1613: \textit{Floer cohomology, spectral sequence and the Maslov class of

1614: Lagrangian embeddings}, Internat. Math. Res. Notices \textbf{7} (1996),

1615: 305--346.

1616: 

1617: \bibitem[PZ]{PZ} A. Polishchuk and E. Zaslow,

1618: \textit{Categorical mirror symmetry: the elliptic curve},

1619: Adv. Theor. Math. Phys. \textbf{2} (1998), 443--470.

1620: 

1621: \bibitem[S1]{S1} P. Seidel,

1622: \textit{Lagrangian two-spheres can be symplectically knotted}, Jour.

1623: Diff. Geom. \textbf{52} (1999), 145--171.

1624: 

1625: \bibitem[S2]{S2} P. Seidel,

1626: \textit{Graded Lagrangian submanifolds}, Bull. Soc. Math. France.

1627: \textbf{128} (2000), 103--146.

1628: 

1629: \bibitem[S3]{S3} P. Seidel, private communication.

1630: 

1631: \bibitem[ST]{ST} P. Seidel and R. P. Thomas,

1632: \textit{Braid group actions on derived categories of sheaves},

1633: pre\-print math.AG/0001043.

1634: 

1635: \bibitem[SV]{SV} A. Shapere and C. Vafa,

1636: \textit{BPS structure of Argyres-Douglas superconformal theories},

1637: preprint hep-th/9910182.

1638: 

1639: \bibitem[SYZ]{SYZ} A. Strominger, S.-T. Yau and E. Zaslow,

1640: \textit{Mirror Symmetry is T-Duality}, Nucl. Phys. \textbf{B479}

1641: (1996), 243--259.

1642: 

1643: \bibitem[T1]{T1} R. P. Thomas, 

1644: \textit{A holomorphic Casson invariant for Calabi-Yau 3-folds, and

1645: bundles on K3 fibrations}, preprint math.AG/9806111.

1646: 

1647: \bibitem[T2]{T2} R. P. Thomas, 

1648: \textit{An obstructed bundle on a Calabi-Yau 3-fold},

1649: Adv. Theor. Math. Phys. \textbf{3} (1999).

1650: 

1651: \bibitem[T3]{T3} R. P. Thomas, 

1652: \textit{D-branes: Lagrangians, bundles and derived categories},

1653: to appear as a chapter in the book \textit{``Mirror Symmetry''},

1654: ed. C. Vafa and E. Zaslow. Available at

1655: http://www.ma.ic.ac.uk/$\sim$rpwt.

1656: 

1657: \bibitem[TY]{TY} R. P. Thomas and S.-T. Yau, 

1658: \textit{Special Lagrangians, stable bundles and mean curvature flow},

1659: preprint.

1660: 

1661: \bibitem[Ty]{Ty} A. N. Tyurin,

1662: \textit{Non-abelian analogues of Abel's theorem}, ICTP preprint

1663: (1997).

1664: 

1665: \bibitem[Va]{Va} C. Vafa,

1666: \textit{Extending Mirror Conjecture to Calabi-Yau with Bundles},

1667: Comm. Contemp. Math. \textbf{1} (1999), 65--70.

1668: 

1669: \bibitem[W1]{W1} E. Witten,

1670: \textit{Chern-Simons gauge theory as a string theory},

1671: The Floer memorial volume, 637--678. Progr. Math., 133,

1672: Birkh\"auser, Basel, 1995.

1673: 

1674: \bibitem[W2]{W2} E. Witten,

1675: \textit{Branes And The Dynamics Of QCD},

1676: Nucl. Phys. \textbf{B}507 (1997), 658--690.

1677: 

1678: \bibitem[Y]{Y} S.-T. Yau,

1679: \textit{On the Ricci curvature of a compact K\"ahler manifold and the

1680: complex Monge-Ampere equation}, Comm. Pure Appl. Math.

1681: \textbf{31} (1978), 339--411.

1682: 

1683: \end{thebibliography}

1684: 

1685: \noindent {\tt thomas@math.harvard.edu} \newline

1686: \noindent Department of Mathematics, Harvard University, One

1687: Oxford Street, Cambridge MA 02138. USA.

1688: 

1689: \end{document}

1690: 

1691: 

1692: % Local Variables: 

1693: % TeX-parse-self: t

1694: % TeX-auto-save: t

1695: % mode: latex

1696: % TeX-master: "thesis"

1697: % End: