1: \section{Analytical Aspects}
2: \label{sec:Analysis}
3:
4: Lagrangian Floer homology (see~\cite{FloerLag}) is a homology theory
5: associated to a pair $L_0$ and $L_1$ of Lagrangian submanifolds in a
6: symplectic manifold. Its boundary map counts certain
7: pseudo-holomorphic disks whose boundary is mapped into the union of
8: $L_0$ and $L_1$. Our set-up here differs slightly from Floer's: we
9: are considering a pair of totally real submanifolds, $\Ta$ and $\Tb$,
10: in the symmetric product. It is the aim of this section to show that
11: the essential analytical aspects -- the Fredholm theory,
12: transversality, and compactness -- carry over to this context. We
13: then turn our attention to orientations. In the final subsection, we
14: discuss certain disks, whose boundary lies entirely in either $\Ta$ or
15: $\Tb$.
16:
17: \subsection{Nearly symmetric almost-complex structures}
18:
19: We will be counting pseudo-holomorphic disks in $\Sym^g(\Sigma)$,
20: using a restricted class of almost-complex structures over
21: $\Sym^g(\Sigma)$ (which can be thought of as a suitable elaboration of
22: the taming condition from symplectic geometry).
23:
24: Recall that an almost-complex structure $J$ over a symplectic manifold
25: $(M,\omega)$ is said to {\em tame} $\omega$ if $\omega(\xi,J\xi)>0$
26: for every non-zero tangent vector $\xi$ to $M$. This is an open
27: condition on $J$.
28:
29: The quotient map $$\pi\colon \ProdSig{g} \longrightarrow
30: \Sym^g(\Sigma)$$ induces a covering space of
31: $\Sym^g(\Sigma)-{\Diag}$, where $\Diag\subset \ProdSig{g}$ is the
32: diagonal, see Subsection~\ref{subsec:SymmetricProducts}. Let $\eta$ be
33: a K\"ahler form over $\Sigma$, and $\omega_0 =\eta^{\times
34: g}$. Clearly, $\omega_0$ is invariant under the covering action, so it
35: induces a K\"ahler form $\pi_*(\omega_0)$ over $\Sym^g(\Sigma)-\Diag$.
36:
37: \begin{defn}
38: \label{def:NearlyStandard}
39: Fix a K\"ahler structure $(\sj,\eta)$ over $\Sigma$,
40: a finite collection of points
41: $$\{z_i\}_{i=1}^m\subset \CurveComp,$$
42: and an open set $V$ with
43: $$\left(\{z_i\}_{i=1}^m\times \Sym^{g-1}(\Sigma) \bigcup \Diag\right)\subset V \subset
44: \Sym^g(\Sigma)$$
45: and
46: $${\overline V}\cap (\Ta\cup\Tb)=\emptyset.$$
47: An almost-complex structure $J$ on
48: $\Sym^g(\Sigma)$ is called $(\sj,\eta,V)$-{\em nearly symmetric}
49: if \begin{itemize}
50: \item $J$ tames $\pi_*(\omega_0)$ over $\Sym^g(\Sigma)-{\overline V}$
51: \item $J$ agrees with $\Sym^g(\sj)$ over $V$
52: \end{itemize}
53: The space of $(\sj,\eta,V)$-nearly symmetric almost-complex
54: structures will be denoted $\AlmostCx(\sj,\eta,V)$.
55: \end{defn}
56:
57: Note that since $\Ta$ and $\Tb$ are Lagrangian with respect to
58: $\pi_*(\omega_0)$, and $J$ tames $\pi_*(\omega_0)$, the tori $\Ta$ and
59: $\Tb$ are totally real for $J$.
60:
61: The space $\AlmostCx(\sj,\eta,V)$ is a subset of the set of all
62: almost-complex structures, and as such it can be endowed with Banach
63: space topologies ${\mathcal C}^\ell$ for any $\ell$. In fact,
64: $\Sym^g(\sj)$ is $(\sj,\eta,V)$-nearly symmetric for any choice of $\eta$
65: and $V$; and the space $\AlmostCx(\sj,\eta,V)$ is an open neighborhood
66: of $\Sym^g(\sj)$ in the space of almost-complex structures which agree
67: with $\Sym^g(\sj)$ over $V$.
68:
69: Unless otherwise specified, we choose the points $\{z_i\}_{i=1}^m$
70: so that there is some
71: $z_i$ in each connected component of $\CurveComp$.
72:
73: \subsection{Fredholm Theory}
74: \label{subsec:FredholmTheory}
75:
76: We recall the Fredholm theory for pseudo-holomorphic disks, with appropriate
77: boundary conditions. For more details, we refer the reader
78: to~\cite{FloerUnregularized}, see also
79: \cite{OhFloer},
80: \cite{FloerHoferSalamon}, and~\cite{FOOO}.
81:
82: To set this up we assume that $\Ta$ and $\Tb$ meet transversally,
83: i.e. that each $\alpha_{i}$ meets each $\beta_{j}$ transversally.
84:
85: We consider the moduli space of holomorphic strips connecting $\x$
86: to $\y$, suitably generalized as follows.
87: Let $\Strip=[0,1]\times i\R \subset \C$ be the
88: strip in the complex plane. Fix a path $J_s$
89: of almost-complex structures over $\Sym^g(\Sigma)$.
90: Let $\ModFlow_{J_s}(\x,\y)$ be the set of
91: maps satisfying the following conditions:
92: $$\ModFlow_{J_s}(\x,\y)=\left\{u\colon \Strip\longrightarrow
93: \Sym^g(\Sigma)\Bigg|
94: \begin{array}{ll}
95: u(\{1\}\times\R)\subset \Ta \\
96: u(\{0\}\times\R)\subset \Tb \\
97: \lim_{t\goesto -\infty} u(s+it)=\x \\
98: \lim_{t\goesto +\infty} u(s+it)=\y \\
99: \frac{du}{ds}+J(s) \frac{du}{dt}=0 &
100: \end{array}\right\}. $$
101: For $\phi\in\pi_2(\x,\y)$, the space $\ModFlow_{J_s}(\phi)$ denotes
102: the subset consisting of maps as above which represent the given
103: homotopy class (or equivalence class, when $g=2$) $\phi$.
104: The translation action on $\Strip$ endows this moduli space with an
105: $\R$ action. The space of {\em unparameterized $J_s$-holomorphic disks} is
106: the quotient
107: $$\UnparModSp_{J_s}(\phi)=\frac{\ModSp_{J_s}(\phi)}{\R}.$$ The word
108: ``disk'' is used, in view of the holomorphic identification of the
109: strip with the unit disk in the complex plane with two
110: boundary points removed (and maps in the moduli space
111: extend across these points, in view of the asymptotic conditions).
112:
113: We will be considering moduli space $\ModFlow_{J_s}(\x,\y)$, where
114: $J_s$ is a one-parameter family of nearly symmetric almost-complex structures: i.e. where we
115: have some fixed $(\sj,\eta,V)$ for which each $J_s$ is
116: $(\sj,\eta,V)$-nearly symmetric (see
117: Definition~\ref{def:NearlyStandard}) for each $s\in[0,1]$.
118:
119: In the definition of nearly-symmetric almost-complex structure, the
120: almost-complex structure in a neighborhood of $\Diag$ is fixed to help
121: prove the required energy bound, c.f. Subsection~\ref{subsec:EnergyBounds}.
122: Moreover, the complex structure in a neighborhood of the
123: $\{z_i\}_{i=1}^m\times\Sym^{g-1}(\Sigma)$ is fixed to establish the following:
124:
125: \begin{lemma}
126: \label{lemma:NonNegativity}
127: If $u\in\ModSp_{J_s}(\phi)$ is any $J_s$-holomorphic disk, then
128: $\cald(u)\geq 0$.
129: \end{lemma}
130:
131: \begin{proof}
132: In a neighborhood of $\{z_i\}_{i=1}^m\times \Sym^{g-1}(\Sigma)$, we are using an
133: integrable complex structure, so the disk $u$ must either be contained
134: in the subvariety (which is excluded by the boundary conditions) or it
135: must meet it non-negatively.
136: \end{proof}
137:
138: Let $E$ be a vector bundle over $[0,1]\times \R$ equipped with a
139: metric and compatible connection,
140: $p$, $\delta$ be positive real numbers,
141: and $k$ be a non-negative integer. The
142: $\delta$-weighted Sobolev space of sections of $E$, written
143: $\Sobol{p}{k,\delta}([0,1]\times \R,E)$, is the space of sections
144: $\sigma$ for which the norm
145: $$\|\sigma\|_{\Sobol{p}{k,\delta}(E)}=
146: \sum_{\ell=0}^{k}
147: \int_{[0,1]\times \R}|\nabla^{(\ell)}
148: \sigma(s+it)|^{p}e^{\delta \tau(t)}ds\wedge dt$$
149: is finite. Here, $\tau\colon \R\longrightarrow \R$ is a
150: smooth function with $\tau(t)=|t|$ provided that $|t|\geq 1$.
151:
152: Fix some $p>2$.
153: Let $\Paths_{\delta}(\x,\y)$ denote the space of maps
154: $$u\colon [0,1]\times \R\longrightarrow\Sym^{g}(\Sigma)$$
155: in $\Sobol{p}{1,\loc}$,
156: satisfying the boundary conditions
157: \begin{eqnarray*}
158: u(\{1\}\times \R)\subset \Ta, &{\text{and}}&
159: u(\{0\}\times \R)\subset \Tb,
160: \end{eqnarray*}
161: which are asymptotic to $\x$ and $\y$ as $t\goesto -\infty$ and
162: $+\infty$, in the following sense.
163: There is a real number $T>0$ and sections
164: \begin{eqnarray*}
165: \xi_{-}\in
166: \Sobol{p}{1,\delta}\Big([0,1]\times (-\infty,-T], T_{\x}\Sym^{g}(\Sigma)\Big)
167: &{\text{and}}&
168: \xi_{+}\in
169: \Sobol{p}{1,\delta}\Big([0,1]\times [T,\infty), T_{\y}\Sym^{g}(\Sigma)\Big)
170: \end{eqnarray*}
171: with the property that
172: \begin{eqnarray*}
173: u(s+it)= \exp_{\x}(\xi_{-}(s+it))
174: &{\text{and}}&
175: u(s+it)= \exp_{\y}(\xi_{+}(s+it),
176: \end{eqnarray*}
177: for all $t<-T$ and $t>T$ respectively. Here,
178: $\exp$ denotes the usual exponential map for some Riemannian
179: metric on $\Sym^{g}(\Sigma)$.
180: Note that $\Paths_{\delta}(\x,\y)$ can be naturally given the
181: structure of a Banach manifold, whose tangent space at any
182: $u\in\Paths_{\delta}(\x,\y)$ is given by
183: $$\Sobol{p}{1,\delta}(u)
184: := \left\{\xi\in \Sobol{p}{1,\delta}([0,1]\times \R,
185: u^{*}(T\Sym^{g}(\Sigma)))\Big|
186: \begin{array}{l}
187: \xi(1,t)\in T_{u(1+it)}(\Ta), \forall t\in\R\\
188: \xi(0,t)\in T_{u(0+it)}(\Tb), \forall t\in\R
189: \end{array}\right\}.$$
190: Moreover, at each $u\in\Paths_{\delta}(\x,\y)$, we denote the
191: space of sections
192: $$\Sobol{p}{\delta}(\Lambda^{0,1}u)
193: := \Sobol{p}{\delta}\Big([0,1]\times \R, u^{*}(T\Sym^{g}(\Sigma))\Big)
194: $$
195: These Banach spaces fit together to form a bundle ${\mathcal
196: L}^{p}_{\delta}$ over $\Paths_{\delta}(\x,\y)$.
197: At each
198: $u\in\Paths_{\delta}(\x,\y)$,
199: $\DBar_{J_s}u=\frac{d}{ds}+J(s)\frac{d}{dt}$ lies in the space $\Sobol{p}{\delta}(\Lambda^{0,1}(u))$
200: and is zero exactly
201: when $u$ is a $J_s$-holomorphic map. (Note that our definition of
202: $\DBar_{J_s}$ implicitly uses the natural trivialization of the
203: the bundle $\Lambda^{0,1}$ over $\Strip$, which is why the bundle does
204: not appear in the definition of $\Sobol{p}{\delta}(\Lambda^{0,1}u)$, but does appear in its notation.)
205: This assignment fits together over $\Paths_{\delta}(\x,\y)$ to induce
206: a Fredholm section of ${\mathcal L}^{p}_{\delta}$. The linearization
207: of this section is denoted $$D_{u}
208: \colon \Sobol{p}{1,\delta}(u)\longrightarrow
209: \Sobol{p}{\delta}(\Lambda^{0,1}u),$$
210: and it is given by the formula
211: $$D_u(\nu)=\frac{d\nu}{ds} + J(s)\frac{d\nu}{dt} + (\nabla_\nu J(s))\frac{du}{dt}.$$
212: Since the intersection of $\Ta$ and $\Tb$ is transverse,
213: this linear map is Fredholm for all sufficiently small non-negative
214: $\delta$. Indeed, there is some $\delta_{0}>0$ with the property
215: that any map $u\in\ModFlow(\x,\y)$ lies in
216: $\Paths_{\delta}(\x,\y)$, for all $0\leq \delta < \delta_{0}$.
217:
218: The components of $\Paths_{\delta}(\x,\y)$ can be partitioned
219: according to homotopy classes $\phi\in\pi_{2}(\x,\y)$. The index of
220: $D_u$, acting on the unweighted space ($\delta=0$) descends to a
221: function on $\pi_{2}(\x,\y)$. Indeed, the index is calculated by the
222: Maslov index $\Mas$ of the map $u$ (see~\cite{FloerMaslov},
223: \cite{SalamonZehnder}, \cite{SpecFlow},
224: \cite{Viterbo}). We conclude the subsection with a result about the
225: Maslov index which will be of relevance to us later:
226:
227: \begin{lemma}
228: \label{lemma:MasClass}
229: Let $S\in\pi_2'(\Sym^g(\Sigma))$ be the positive generator. Then for
230: any $\phi\in\pi_2(\x,\y)$, we have that
231: $$\Mas(\phi+k[S])=\Mas(\phi)+2k.$$
232: In particular, if $O_\x\in \pi_2(\x,\x)$ denotes the class of the
233: constant map, then
234: $$\Mas(O_\x+kS)=2k.$$
235: \end{lemma}
236:
237: \begin{proof} It follows from the excision principle for the index that attaching a
238: topological sphere $Z$ to a disk changes the Maslov index by $2\langle c_1, [Z]\rangle$
239: (see \cite{FloerMaslov}, \cite{McDuffSalamon}).
240: On the other hand for the positive
241: generator we have $\langle c_1, [S]\rangle=1$ according
242: to Lemma~\ref{lemma:ChernClass}.
243: \end{proof}
244:
245: \subsection{Transversality}
246: \label{subsec:Transversality}
247:
248: Given a Heegaard diagram $(\Sigma,\alphas,\betas)$
249: for which all the $\alpha_{i}$ meet the $\beta_{j}$ transversally,
250: the tori $\Ta$ and $\Tb$ meet transversally, so
251: the holomorphic disks connecting
252: $\Ta$ with $\Tb$ are naturally endowed with a Fredholm deformation
253: theory.
254:
255: Indeed, the usual arguments from Floer theory
256: (see~\cite{FloerUnregularized}, \cite{OhFloer} and
257: \cite{FloerHoferSalamon}) can be modified to prove the following result:
258:
259: \begin{theorem}
260: \label{thm:Transversality}
261: Fix a Heegaard diagram $(\Sigma,\alphas,\betas)$ with the
262: property that each $\alpha_i$ meets $\beta_j$ transversally,
263: and fix $(\sj,\eta,V)$ as in
264: Definition~\ref{def:NearlyStandard}. Then, for a dense set of
265: paths $J_s$ of $(\sj,\eta,V)$-nearly symmetric almost-complex
266: structures, the moduli
267: spaces $\ModFlow_{J_s}(\x,\y)$ are all smoothly cut out by the
268: defining equations.
269: \end{theorem}
270:
271:
272: In the above statement, ``dense'' is meant in the $C^{\infty}$
273: topology on
274: the path-space of $\AlmostCx(\sj,\eta,V)$.
275:
276: \vskip.5cm
277: \noindent{\bf{Proof of Theorem~\ref{thm:Transversality}}.}
278: This is a modification of the usual proof of transversality,
279: see~\cite{FloerUnregularized}, \cite{OhFloer} and
280: \cite{FloerHoferSalamon}.
281:
282: Recall (see for instance Theorem~5.1 of~\cite{OhFloer}) that if $u$ is
283: any non-constant holomorphic disk, then there is a dense set of points
284: $(s,t)\in [0,1]\times \R$ satisfying the two conditions that
285: $du_{(s,t)}\neq 0$ and $u(s,t)\cap u(s,\R-\{t\})=\emptyset$. By
286: restricting to an open neighborhood of the boundary of $\Strip$ (note
287: that we have assumed that ${\overline V}$ is disjoint from $\Ta$ and
288: $\Tb$), it follows that we can find such an $(s,t)$ with
289: $u(s,t)\not\in {\overline V}$. By varying the path $J_s$ in a
290: neighborhood of $u(s,t)$, the usual arguments show that $u$ is a
291: smooth point for the parameterized moduli space ${\mathfrak M}$,
292: consisting of pairs $(J_s,u)$ for which $\DBar_{J_s}u=0$. The result
293: then follows from the Sard-Smale theorem, applied to the Fredholm
294: projection from ${\mathfrak M}$ to the space of paths of
295: nearly-symmetric almost-complex structures.
296: \qed
297: \vskip.3cm
298:
299: Under certain topological hypotheses, one can achieve
300: transversality by placing the curves $\alphas$ and $\betas$ in general
301: position, but leaving the almost-complex structure fixed: indeed,
302: letting $J_s$ be the constant path $\Sym^g(\sj)$. We return to this in
303: Proposition~\ref{prop:MoveToriTransversality}, after setting up more of the
304: theory of holomorphic disks in $\Sym^g(\Sigma)$.
305:
306:
307: \subsection{Energy bounds}
308: \label{subsec:EnergyBounds}
309:
310: Let $\Omega$ be a domain in $\C$.
311: Recall that the energy of a
312: map $u\colon \Omega\rightarrow X$ to a Riemannian manifold $(X,g)$ is
313: given by $$\Energy(u)=\OneHalf\int_{\Omega}|d u|^2.$$
314:
315: Fix
316: $\phi\in \pi_2(\x,\y).$
317: In order to get the usual compactness results for holomorphic disks
318: representing $\phi$, we need an
319: {\em a priori} energy bound for any holomorphic representative $u$ for
320: $\phi$.
321:
322: Such a bound exists in the symplectic context. Suppose that
323: $(X,\omega)$ is a compact symplectic manifold, with a tame
324: almost-complex structure $J$, then there a constant $C$ for which
325: $$\Energy(u)\leq C \int_{\Omega}u^*(\omega),$$ for each
326: $J$-holomorphic map $u$. When the $u$ has Lagrangian boundary
327: conditions, the integral on the right-hand-side depends only on the
328: homotopy class of the map. This principle holds in our context as
329: well, according to the following lemma.
330:
331: \begin{lemma}
332: \label{lemma:EnergyBound}
333: Fix a path $J_s$ in the space of nearly-symmetric almost-complex
334: structures. Then, for each pair of intersection points
335: $\x,\y\in\Sym^g(\Sigma)$, and $\phi\in
336: \pi_2(\x,\y)$, there is an upper bound on the energy of any holomorphic
337: representative of $\phi$.
338: \end{lemma}
339:
340: \begin{proof}
341: Given
342: $$u\colon (\Strip,\partial \Strip)\longrightarrow
343: (\Sym^g(\Sigma),\Ta\cup \Tb),$$ we consider the
344: lift
345: $${\widetilde u}\colon (\brDisk,\partial
346: \brDisk)\longrightarrow (\Sigma^{\times g},\pi^{-1}(\Ta\cup \Tb))$$
347: obtained by pulling back the branched covering space $\pi\colon
348: \Sigma^{\times g}\longrightarrow \Sym^g(\Sigma)$. (That is to say, $F$
349: is defined to be the covering space of the image of
350: $u$ away from the diagonal
351: $\Diag\subset \Sym^g(\Sigma)$,
352: and in a
353: neighborhood of $\Diag$, $F$ is defined as a subvariety of
354: $\Sigma^{\times g}$ -- it is here that we are using the fact that each of the
355: $J_s$ agree with the standard complex structure near $\Diag$.)
356:
357: We break the energy integral into two regions:
358: $$\Energy(u)=\int_{u^{-1}(\Sym^g(\Sigma)-V)}|du|^2 +
359: \int_{u^{-1}(V)}|du|^2.$$
360: To estimate the integral on $\Sym^g(\Sigma)-V$, we use the fact that
361: each $J_s$ tames $\pi_*(\omega_0)$, from which it follows that there is a
362: constant $C_1$ for which
363: \begin{eqnarray}
364: \Energy(u|_{\Sym^g(\Sigma)-V})&\leq& C_1
365: \int_{u^{-1}(\Sym^g(\Sigma)-V)}u^*(\pi_*(\omega_0))
366: =
367: \frac{C_1}{g!}\int_{{\widetilde u}^{-1}(\ProdSig{g}-{\widetilde V})}{\widetilde
368: u}^{*}(\omega_0),
369: \label{ineq:AwayFromDiag}
370: \end{eqnarray}
371: where ${\widetilde V}=\pi^{-1}({\overline V})$.
372:
373: To estimate the other integrand, choose a K\"ahler form $\omega$ over
374: $\Sym^g(\Sigma)$. Over $V$ all the $J_s$ agree with
375: $\Sym^g(\sj)$, so
376: $u$ is $\Sym^g(\sj)$-holomorphic in that region,
377: so there is some constant $C_2$ with
378: the property that
379: \begin{equation}
380: \label{eq:KahlerDominates}
381: \Energy(u|_V) \leq C_2 \int_{u^{-1}(V)} u^*(\omega)
382: \end{equation}
383: (the constant $C_2$ depends on the Riemannian metric used
384: over $\Sym^g(\Sigma)$ and the choice of K\"ahler form $\omega$).
385: Moreover, the right hand side can be calculated using ${\widetilde u}$
386: according to the
387: following formula:
388: \begin{equation}
389: \label{eq:BranchFormula}
390: \int_{u^{-1}(V)}u^*(\omega)=\frac{1}{g!}\int_{{\widetilde
391: u}^{-1}({\widetilde V})}{\widetilde u}^*(\pi^*(\omega)).
392: \end{equation}
393:
394: Now,
395: fix any two-form $\omega_1$ over $\ProdSig{g}$.
396: Then there is a constant $C_3$ with the
397: following property. Let $${\widetilde u}\colon F
398: \longrightarrow \ProdSig{g}$$ be any map which is
399: $\sj^{\times g}$-holomorphic on ${\widetilde u}^{-1}({\widetilde V})$.
400: Then
401: we have the inequality
402: \begin{equation}
403: \label{ineq:BoundEnergyDiag}
404: \int_{{\widetilde u}^{-1}({\widetilde V})}{\widetilde u}^*(\omega_1) \leq C_3
405: \int_{{\widetilde u}^{-1}({\widetilde V})}
406: {\widetilde u}^*(\ProductForm).
407: \end{equation}
408: This holds for the constant with the property that for each tangent
409: vector $\xi$ to $\ProdSig{g}$ and
410: $$
411: \omega_1(\xi,J\xi)\leq C_3\ProdForm(\xi,J\xi),
412: $$
413: where $J=\sj^{\times g}$.
414: Such a constant can be found since $\ProdSig{g}$ is compact and
415: $\ProdForm(\cdot ,J \cdot)$ determines a non-degenerate quadratic form on
416: each tangent space $T\ProdSig{g}$.
417:
418: Applying Inequality~\eqref{ineq:BoundEnergyDiag} for the form
419: $\omega_1=\pi^*(\omega)$, and combining with
420: Inequality~\eqref{ineq:AwayFromDiag}, we find a constant $C_0$ with the property that
421: \begin{equation}
422: \label{eq:TopologicalBound}
423: \Energy(u)\leq C_0\int_{\brDisk}{\widetilde u}^*(\omega_0).
424: \end{equation}
425:
426: Moreover, with respect to the symplectic form $\ProductForm$, the preimage
427: under $\pi$ of $\Ta$ and $\Tb$ are both Lagrangian. This
428: gives a topological interpretation to the right-hand-side of
429: Equation~\eqref{eq:TopologicalBound}:
430: \begin{equation}
431: \label{eq:CohomInterp}
432: \int_{\brDisk}{\widetilde u}^*(\ProductForm)=\langle \ProductForm,
433: [\brDisk,\partial\brDisk]\rangle,
434: \end{equation}
435: which makes sense since $\ProductForm$ defines a relative cohomology
436: class in $H^2(\ProdSig{g},\pi^{-1}(\Ta\cup \Tb))$. Note that the correspondence
437: $u\mapsto{\widetilde u}$ induces a right inverse,
438: up to a multiplicative constant, to the map on
439: homology
440: $$\pi_*\colon H_2(\ProdSig{g},\pi^{-1}(\Ta\cup\Tb))\longrightarrow H_2(\Sym^g(\Sigma),\Ta\cup\Tb);$$
441: thus,
442: the homology class $[\brDisk,\partial\brDisk]$ depends only on
443: the relative homology class of $u$, thought of as a class in
444: $H_2(\Sym^g(\Sigma),\Ta\cup\Tb)$ --
445: in particular, it depends only on the
446: equivalence class $\phi\in\pi_{2}(\x,\y)$ of $u$).
447:
448: Thus, given a class $\phi\in\pi_2(\x,\y)$,
449: this gives us an {\em a priori}
450: bound on the $\ProductForm$-energy of the (branched) lift of any
451: holomorphic disk $u\in \ModFlow_{J_s}({\phi})$, combining
452: Inequality~\eqref{eq:KahlerDominates},
453: Equation~\eqref{eq:BranchFormula}, Inequality~\eqref{eq:TopologicalBound},
454: and Equation~\eqref{eq:CohomInterp}, we get that
455: \begin{equation}
456: \label{eq:EnergyBound}
457: \Energy(u)\leq C_0 \langle
458: \ProductForm, [\brDisk,\partial\brDisk]\rangle,
459: \end{equation}
460: (for some constant $C_0$ independent of the class $\phi\in\pi_2(\x,\y)$).
461: \end{proof}
462:
463: \subsection{Holomorphic disks in the symmetric product}
464:
465: Suppose that the path $J_s$ is constant, and it is given by
466: $\Sym^g(\sj)$ for some complex structure $\sj$ over $\Sigma$. Then, the
467: space of holomorphic disks connecting $\x,\y$ can be given an alternate
468: description, using only maps between one-dimensional complex manifolds.
469:
470: \begin{lemma}
471: \label{lemma:Correspondence}
472: Given any holomorphic disk $u\in\ModFlow(\x,\y)$, there is a
473: branched $g$-fold covering space $p\colon {\widehat
474: \Strip}\longrightarrow \Strip$ and a holomorphic map ${\widehat
475: u}\colon {\widehat \Strip}\longrightarrow \Sigma $, with the
476: property that for each $z\in\Strip$, $u(z)$ is the image under
477: ${\widehat u}$ of the pre-image $p^{-1}(z)$.
478: \end{lemma}
479:
480: \begin{proof}
481: Given a holomorphic map $u\colon \Strip\longrightarrow
482: \Sym^g(\Sigma)$ which does not lie in the diagonal,
483: we can find a branched $g!$-fold cover
484: $p\colon {\widetilde \Strip}\longrightarrow \Strip$ pulling back the canonical $g!$-fold
485: cover $\pi\colon \Sigma^{\times g}\longrightarrow \Sym^g(\Sigma)$,
486: i.e. making the following diagram commutative:
487: $$\begin{CD}
488: {\widetilde \Strip}@>{\widetilde u}>> \Sigma^{\times g} \\
489: @V{p}VV @V{\pi}VV \\
490: \Strip @>{u}>> \Sym^g(\Sigma).
491: \end{CD}
492: $$
493: Indeed, ${\widetilde \Strip}$ inherits an action by the symmetric
494: group on $g$ letters $S_g$, and ${\widetilde u}$ is equivariant for
495: the action (and its quotient is $u$). Let $\Pi_1\colon \Sigma^{\times
496: g}\longrightarrow \Sigma$ denote projection onto the first
497: factor. Then, the composite map
498: $$\Pi_1\circ {\widetilde u}\colon {\widetilde \Strip}\longrightarrow
499: \Sigma$$ is invariant under the action of $S_{g-1}\subset S_g$
500: consisting of permutations which fix the first letter. Then, we let
501: ${\widehat \Strip}={\widetilde \Strip}/S_{g-1}$, and ${\widehat u}$ be
502: the induced map from ${\widehat \Strip}$ to $\Sigma$. It is easy to
503: verify that ${\widehat u}$ has the desired properties.
504: \end{proof}
505:
506: \begin{remark}
507: It is straightforward to find appropriate topological conditions
508: on ${\widehat u}|{\partial{\widehat \Strip}}$ to give a one-to-one
509: correspondence between flows in $\ModFlow(\x,\y)$ and certain
510: pairs $(p\colon {\widehat \Strip}\longrightarrow \Strip, {\widehat
511: u}\colon {\widehat \Strip}\longrightarrow \Sigma)$.
512: \end{remark}
513:
514: Let $\cald _1,...,\cald _m$ denote the connected components of $\Sigma
515: -\alpha _1,-...-\alpha _g-\beta_1,...-\beta _g$. Fix a basepoint $z_i$
516: inside each $\cald _i$. Then for any $\phi \in \pi_2(\x,\y)$ we define
517: the {\it domain} associated to $\phi$, as a formal linear combination
518: of components:
519: $$\cald(\phi)=\sum _{i=1} ^m n_{z_i}(\phi)\cdot \cald _i.$$
520: Similarly the {\it area} of $\phi$ is given by
521: $${\mathcal A}(\phi)= \sum _{i=1} ^m n_{z_i}(\phi)\cdot {\rm
522: Area}_{\eta}({\cald _i}),$$
523: where $\eta$ is the K\"ahler form on $\Sigma$.
524: This area gives us a concrete way to understand the energy bound from
525: the
526: previous section since, as is easy to verify,
527: $$\int_{F}{\widetilde u}^*(\omega_0)=(g!){\mathcal A}(\phi).$$
528:
529: As an application of Lemma~\ref{lemma:Correspondence}, we observe that
530: for certain special homotopy classes of maps in $\pi_2(\x,\y)$
531: transversality can also be achieved by moving the curves $\alphas$ and
532: $\betas$, following the approach of Oh~\cite{OhTransversality}. (This
533: observation will prove helpful in the explicit calculations of
534: Section~\ref{sec:HandleSlides} and also Section~\ref{HolDiskTwo:sec:Examples}
535: of~\cite{HolDiskTwo}.)
536:
537: To state it, we need the following:
538:
539: \begin{defn}
540: A domain $\cald(\phi)$ is called {\em $\alpha$-injective} if all of its
541: multiplicities are $0$ or $1$, if its interior (i.e. the interior of
542: the region with multiplicity $1$) is
543: disjoint from each $\alpha_i$ for $i=1,...,g$, and its boundary
544: contains intervals in each $\alpha_i$.
545: \end{defn}
546:
547: \begin{prop}
548: \label{prop:MoveToriTransversality}
549: Let $\phi\in\pi_2(\x,\y)$ be an $\alpha$-injective homotopy class, and
550: fix a complex structure $\sj$ over $\Sigma$. Then, for generic
551: perturbations of the $\alphas$, the moduli space $\ModFlow(\phi)$ of
552: $\Sym^g(\sj)$-holomorphic disks is smoothly cut out by its defining
553: equation.
554: \end{prop}
555:
556: \begin{proof}
557: The hypotheses
558: ensure that for all $t\in \R$, we have that
559: $u(1+it)=\{a_1,...,a_g\}\in\Ta$ where $a_i\not\in u(1+it')$ for any
560: $t'\neq t$. This is true because the $\alpha$-injectivity hypothesis
561: ensures that the corresponding map ${\widehat u}\colon
562: \brDisk\longrightarrow \Sigma$, coming from
563: Lemma~\ref{lemma:Correspondence}, is injective (with injective
564: linearization, by elementary complex analysis) on the region mapping
565: to the $\alpha$-curves $p^{-1}(\{1\}\times \R)$. Thus,
566: following~\cite{OhTransversality}, by varying the $\alpha_i$ in a
567: neighborhood of the $a_i$, one can see that the map $u$ is a smooth
568: point in a parameterized moduli space (parameterized now by variations
569: in the curves). Thus, according to the Sard-Smale theorem, for generic
570: small variations in the $\alphas$, the corresponding moduli spaces are
571: smooth.
572: \end{proof}
573:
574: \subsection{Orientability}
575: \label{subsec:Orientability}
576:
577: In this subsection, we show that the moduli spaces of flows
578: $\ModFlow(\x,\y)$ are orientable. As is usual
579: in the gauge-theoretic set-up,
580: this is done by proving triviality of the determinant line bundle of the
581: linearization of the equations (the $\DBar$-equation)
582: which cut out the moduli spaces. A thorough treatment of orientability
583: in general can be found in~\cite{FOOO}.
584:
585: For some fixed $p>2$ and some real $\delta>0$ (both of which we
586: suppress from the notation),
587: consider the space ${\mathcal B}(\x,\y)={\mathcal B}_{\delta}(\x,\y)$
588: of maps discussed in Subsection~\ref{subsec:FredholmTheory}.
589: The moduli spaces of holomorphic disks
590: are finite-dimensional subspaces of this Banach
591: manifold.
592:
593: Recall that for a family $F_{x}$ of Fredholm operators parameterized by an auxiliary space
594: $X$, the virtual vector spaces
595: $\ker F_{x}-\CoKer F_{x}$ naturally fit together
596: to give rise to an element in the $K$-theory of $X$
597: (see~\cite{KTheory}), the {\em virtual index bundle}.
598: The determinant of this is a real line bundle over $X$, the {\em
599: determinant line bundle} of the family $F_{x}$.
600:
601: \begin{prop}
602: \label{prop:Orientable} There is a trivial line bundle over
603: ${\mathcal B}(\x,\y)$ whose restriction to the moduli space
604: $\ModFlow_{J_s}(\x,\y)\subset {\mathcal B}(\x,\y)$ is naturally
605: identified with the determinant line bundle for the linearization
606: $\det(D_{u})$, where $J_s$ is any path of $(\sj,\eta,V)$-nearly
607: symmetric almost-complex structures.
608: \end{prop}
609:
610:
611:
612: As we shall see, the main ingredient in the above proposition is the fact
613: that the totally real subspaces $\Ta$ and $\Tb$ have trivial tangent
614: bundles. We shall give the proof after a
615: preliminary discussion.
616:
617:
618: Let $L_{0}(t)$ and $L_{1}(t)$ be a pair of paths of totally real
619: subspaces of $\C^{n}$, indexed by $t\in \R$ which are asymptotically
620: constant as $t\goesto\pm \infty$, i.e. there are totally real subspaces
621: $L_{0}^{-}$, $L_{1}^{-}$, $L_{0}^{+}$, and $L_{1}^{+}$ with the
622: property that
623: \begin{eqnarray*}
624: \lim_{t\goesto \pm \infty}L_{0}(t)=L_{0}^{\pm} &&
625: \lim_{t\goesto \pm \infty}L_{1}(t)=L_{1}^{\pm}.
626: \end{eqnarray*}
627: Suppose moreover that $L_{0}^{-}$ and $L_{1}^{-}$ are transverse, and
628: similarly $L_{0}^{+}$ and $L_{1}^{+}$ are transverse, too.
629: Then, the $\DBar$ on $\C^{g}$-valued functions on
630: the strip, satisfying boundary conditions specified by
631: the paths $L_{0}(t)$ and $L_{1}(t)$
632: $$\DBar\colon
633: \left\{f \in \Sobol{p}{1}([0,1]\times \R; \C^{g})
634: \Bigg|
635: \begin{array}{ll}
636: f(1+it)\in L_{0}(t), \\
637: f(0+it)\in L_{1}(t), \\
638: \DBar f = 0
639: \end{array}\right\}
640: \longrightarrow \Sobol{p}{}([0,1]\times \R;\C^{g})$$
641: is Fredholm.
642: Thus, the $\DBar$ operator induces a family of Fredholm operators
643: indexed by the space
644: $${\mathcal P}=
645: \left\{L_{0},L_{1}\colon [0,1] \longrightarrow \TRGras(g)
646: \Bigg| \begin{array}{l}
647: L_{0}(0)=L_{0}^{-}, \\
648: L_{1}(0)=L_{1}^{-}, \\
649: L_{0}(1)=L_0^+, \\
650: L_1(1)=L_1^+
651: \end{array}
652: \right\}
653: $$
654: (after reparameterizing the paths in ${\mathcal P}$ to be indexed by
655: $\R\cup\{\pm\infty\}$ rather than $[0,1]$),
656: where $\TRGras(g)$ denotes the Grassmannian of totally real
657: $g$-dimensional subspaces of $\C^{g}$.
658:
659: The index of the linearization $D_{u}$ of the $\DBar_{J_s}$ operator on maps
660: of the disk into $\Sym^{g}(\Sigma)$ can be related to index of the
661: $\DBar$-operators over ${\mathcal P}$, as follows. First observe that
662: $D_u$ depends on a path $J_s$ of almost-complex structures.
663: However, we can
664: connect the family to the constant path $\Sym^g(\sj)$,
665: without changing
666: the index bundle. Next, fix a contraction of the unit disk to
667: $-i$.
668: Together with a connection over $T\Sym^g(\Sigma)$,
669: this induces a trivialization for any $u\in
670: {\mathcal B}(\x,\y)$ of the pull-back of the complex tangent bundle of
671: $\Sym^{g}(\Sigma)$ (induced from $\Sym^g(\sj)$). Via
672: these trivializations, the one-parameter family of totally real
673: subspaces
674: \begin{eqnarray*}
675: \left\{t\mapsto T_{u(1+it)}\Ta\subset
676: T_{u(1+it)}\Sym^{g}(\Sigma)\right\}, &&
677: \{t\mapsto T_{u(0+it)}\Tb\subset T_{u(0+it)}\Sym^{g}(\Sigma)\}
678: \end{eqnarray*}
679: induce one-parameter families $L_{0}(t)$ and $L_{1}(t)$ of totally
680: real subspaces of $\C^{g}$. Indeed, if we use a connection over
681: $T\Sym^g(\Sigma)$ which trivial along $\Tb$,
682: and we choose the
683: contraction of our disk to preserve the left arc, both $t=0$ and $t=1$
684: endpoints of the families can be viewed as a fixed (i.e. independent of
685: the particular choice of $u$). Thus, we have a map
686: $$\Psi\colon {\mathcal B}(\x,\y)\longrightarrow
687: {\mathcal P},$$ together with an identification between the pull-back
688: of the (virtual) index bundle for $\DBar$ and the (virtual) index
689: bundle for $D_{u}$ (over the moduli space $\ModFlow(\x,\y)\subset
690: {\mathcal B}(\x,\y)$).
691:
692:
693: We wish to study the index bundle over ${\mathcal P}$.
694: There is a ``difference'' map
695: $$\delta\colon \TRGras(g)\times \TRGras(g) \longrightarrow
696: \frac{\Gl_{g}(\C)}{\Gl_{g}(\R)},$$
697: where $\delta(L_{0},L_{1})$ is the equivalence class of any matrix
698: $A\in\Gl_{g}(\C)$ with the property that $A L_{0}=L_{1}$. (By
699: taking the difference with $\R^{g}\subset \C^{g}$, we obtain a
700: diffeomorphism between $\TRGras(g)$ and the homogeneous space
701: $\frac{\Gl_{g}(\C)}{\Gl_{g}(\R)}$.) In this space, we have a
702: Maslov cycle
703: $$Z_{\mu}\subset \frac{\Gl_{g}(\C)}{\Gl_{g}(\R)}= \{[A]\big|
704: \R^{g}+A(\R^{g})\neq \C^{n}\}.$$
705: Of course, $L_{0}$ and $L_{1}$ meet transversally if and only if their
706: difference $\delta(L_{0},L_{1})$ does not lie in the Maslov cycle.
707:
708: Let $[a_{0}]=\delta(L_{0}^{-}, L_{1}^{-})$, $[a_1]=\delta(L_0^+,L_1^+)$.
709: The difference map gives us a map
710: $$\Phi\colon {\mathcal P} \longrightarrow {\mathcal Q} = \{A \colon [0,1]\longrightarrow
711: \frac{\Gl_{g}(\C)}{\Gl_{g}(\R)} \big| A(0)=[a_{0}], A(1)=[a_1]\}.$$
712: In this notation,
713: then, the numerical index of the $\DBar$ operator associated to a pair
714: of paths $L_{0}(t)$ and $L_{1}(t)$ in ${\mathcal P}$ is calculated by
715: the intersection number of the difference with the Maslov cycle:
716: $$\ind(\DBar(L_{0}(t),L_{1}(t)))=\delta(L_{0}(t),L_{1}(t))\cap
717: Z_{\mu}.$$ Moreover, we could work entirely over ${\mathcal Q}$:
718: ${\mathcal Q}$ is identified with the subspace of ${\mathcal P}$ where
719: $L_{0}(t)\equiv
720: \R^{g}$, so there is an index bundle over ${\mathcal Q}$, and the
721: index bundle for $\DBar$ over ${\mathcal P}$ is easily seen to be the
722: pull-back of this index bundle over ${\mathcal Q}$ (since the index
723: bundle over ${\mathcal P}$ is trivial over the
724: fiber of $\Phi$).
725:
726: Clearly,
727: $$\pi_{2}\left(\frac{\Gl_{g}(\C)}{\Gl_{g}(\R)}\right)\cong
728: \pi_{1}({\mathcal Q}).$$
729: Now, if $g\geq 2$, it is easy to see that
730: $$\pi_{2}\left(\frac{\Gl_{g}(\C)}{\Gl_{g}(\R)}\right)\cong\Zmod{2}.$$
731: Thus, there is no {\em a priori} reason for the determinant line bundle
732: $\det(\DBar)\longrightarrow {\mathcal Q}$ to be trivial: its first
733: Stiefel-Whitney class may evaluate non-trivially on the non-trivial
734: homotopy class $\Zmod{2}$. Proposition~\ref{prop:Orientable}
735: is established by giving a suitable lift of the composite map
736: $\Phi\circ\Psi$.
737:
738: \vskip0.3cm
739: \noindent{\bf{Proof of Proposition~\ref{prop:Orientable}.}}
740: Continuing the above notation,
741: fix matrices $a_{0}, a_1 \in\Gl_{g}(\C)$, and
742: consider the space
743: $${\widetilde {\mathcal Q}}=\{A\colon [0,1]\longrightarrow \Gl_{g}(\C)\big|
744: A(0)=a_{0}, A(1)=a_1 \}.$$
745: Since $\pi_{2}(\Gl_{g}(\C))=0$, we
746: see that the index bundle of the $\DBar$ operator over ${\widetilde Q}$ is
747: orientable. Thus, to establish orientability of the determinant line
748: bundle over the moduli spaces of flows, we lift
749: $\Phi\circ \Psi$ to a map
750: $${\widetilde \Phi}\colon
751: {\mathcal B}(\x,\y)\longrightarrow {\widetilde Q}.$$
752: To define this lift, note that the tangent spaces to $\Ta$ and $\Tb$
753: respectively can be
754: trivialized by ordering and orienting the attaching circles $\alphas$
755: and $\betas$. This in turn gives rise to a complex trivialization of
756: the restrictions of $T\Sym^{g}(\Sigma)$ to $\Ta$ and $\Tb$ respectively
757: (induced from the identifications
758: $T \Sym^{g}(\Sigma)|_{\Ta}\cong \Ta\otimes_{\R}\C$,
759: $T \Sym^{g}(\Sigma)|_{\Tb}\cong \Tb\otimes_{\R}\C$ arising from the
760: corresponding totally
761: real structures).
762: Given a holomorphic disk $u$, then, we let $A(t)$
763: denote the matrix corresponding to the linear transformation from
764: $\C^{g}$ to itself given by parallel transporting the vector space
765: $\C^{g}\cong T_{u(1+it)}\Ta\otimes_{\R}\C$ to
766: $T_{u(1)}\Ta\otimes_{\R}\C\cong \C^{g}$,
767: using the arc which is the image under $u$
768: of the path prescribed by the fixed contraction of
769: $\Strip$.
770: Now, the composite $\Phi\circ\Psi$ factors through
771: the projection from ${\widetilde {\mathcal Q}}$
772: to ${\mathcal Q}$, so the pull-back of the determinant of the
773: index bundle
774: is trivial since it is trivial over ${\widetilde{\mathcal Q}}$.
775: \qed
776: \vskip0.3cm
777:
778: We would like to choose orientations for all moduli spaces in a
779: consistent manner. To this end, we construct ``coherent orientations''
780: closely following~\cite{FloerHofer}. Note that splicing gives an
781: identification $$\det(u_1)\wedge\det(u_2)\cong \det(u_1 * u_2),$$
782: where $u_1\in\pi_2(\x,\y)$ and $u_2\in\pi_2(\y,\w)$ are a pair of
783: maps.
784:
785: \begin{defn}
786: \label{defn:CoherentSystem}
787: A {\em coherent system of orientations for $\spinc$}, ${\mathfrak o}$,
788: is a choice of non-vanishing sections ${\mathfrak o}(\phi)$ of the determinant line
789: bundle over each $\phi\in\pi_2(\x,\y)$
790: for each $\x,\y\in{\mathcal S}$ and each
791: $\phi\in\pi_2(\x,\y)$, which are compatible with gluing in the sense that
792: $${\mathfrak o}({\phi_1})\wedge {\mathfrak o}({\phi_2})={\mathfrak o}({\phi_1 *\phi_2}),$$
793: under the identification coming from splicing, and
794: $${\mathfrak o}({u*S})={\mathfrak o}(u),$$
795: under the identification coming from the canonical orientation for
796: the moduli space of holomorphic spheres.
797: \end{defn}
798:
799: To construct these it is useful to have the following:
800:
801: \begin{defn}
802: \label{def:CompSetPath}
803: Let $(\Sigma,\alphas,\betas,z)$ be a Heegaard diagram representing
804: $Y$, and let $\spinc$ be a $\SpinC$ structure for $Y$. A {\em
805: complete set of paths for $\spinc$} is an enumeration
806: $\{\x_0,...,\x_m\}={\mathcal S}$ of all the intersection points of
807: $\Ta$ with $\Tb$ representing $\spinc$, and a collection of homotopy classes
808: $\theta_i\in\pi_2(\x_0,\x_i)$ for $i=1,...,m$ with $n_z(\theta_i)=0$.
809: \end{defn}
810:
811: Fix periodic classes $\phi_1,...,\phi_b\in\pi_2(\x,\x)$ representing a
812: basis for $H^1(Y;\Z)$, and non-vanishing
813: sections of the determinant line bundle for bundle for the
814: homotopy classes $\theta_1,...,\theta_m$ and $\phi_1,...,\phi_b$.
815: These data uniquely determine coherent system of orientations by splicing,
816: since any homotopy class $\phi\in\phi_2(\x_i,\x_j)$ can be uniquely
817: written as $$\phi=a_1\phi_1+...+a_b\phi_b-\theta_i+\theta_j.$$
818:
819:
820:
821: \subsection{Degenerate disks}
822:
823: Fix a nearly symmetric almost-complex structure $J$ over
824: $\Sym^g(\Sigma)$. For each $\x\in\Ta\cap \Tb$, the moduli space of
825: {\em $\alpha$-degenerate disks} is the set of maps
826: $$\ModDeg_J(\x)=\left\{ u\colon [0,\infty)\times \R\longrightarrow
827: \Sym^g(\Sigma)\Bigg|
828: \begin{array}{l}
829: u(\{0\}\times \R)\subset \Ta \\
830: \lim_{z\goesto \infty}u(z)=\x \\
831: \frac{du}{ds}+J\frac{du}{dt}=0
832: \end{array}
833: \right\}.$$
834: Equivalently, we can think of $\ModDeg_J(\x)$ as the moduli space of
835: $J$-holomorphic maps of the unit disk $\CDisk$ in $\C$ to
836: $\Sym^g(\Sigma)$, which carry $\partial\CDisk$ into $\Ta$, and $i$ to
837: $\x$. This also gives rise to a finite-dimensional moduli space,
838: partitioned according to the homotopy classes of maps satisfying these
839: boundary conditions, a set which we denote by $\pi_2(\x)$ (again, when
840: $g=2$, we divide out by the larger equivalence relation as in the
841: definition of $\pi_2(\x,\y)$). Suppose that $g>1$. Since the map
842: $\pi_1(\Ta)\longrightarrow
843: \pi_1(\Sym^g(\Sigma))$ is injective, it follows that $\pi_2(\x)\cong\Z$
844: under the map $n_z(u)$; equivalently, if $O_\x\in\pi_2(\x)$ is the
845: homotopy class of the constant, then any other is given by $O_\x+k[S]$
846: for $k\in\Z$. Of course, when $g=1$, there is only one homotopy class, and
847: that is $O_\x$.
848:
849: Note that there is a two-dimensional automorphism group acting on
850: $\ModDeg_J(\x)$ (pre-composing $u$ by either a purely imaginary
851: translation or a real dilation), and we denote the quotient space by
852: $\UnparModDeg_J(\x)$. If $\phi\in\pi_2(\x)$ is a homotopy class,
853: then we let $\ModDeg_J(\phi)$, resp. $\UnparModDeg_J(\phi)$, denote
854: its corresponding component in $\ModDeg_J(\x)$,
855: resp. $\UnparModDeg_J(\x)$.
856:
857: In studying smoothness properties of $\ModDeg_J(\x)$, it is useful to
858: have the following result concerning the complex structures $\Sym^g(\sj)$:
859:
860: \begin{lemma}
861: \label{lemma:Genericj}
862: Given a finite collection of points $\{\x_i\}_{i=1}^n$ in
863: $\Sym^g(\Sigma)$, the set of complex structures $\sj$ over $\Sigma$ for
864: which there is a $\Sym^g(\sj)$-holomorphic sphere containing at least
865: one of the $\x_i$ has real codimension two.
866: \end{lemma}
867:
868: \begin{proof}
869: The spheres in $\Sym^g(\Sigma)$ for the complex structure $\Sym^g(\sj)$
870: are all contained in the set of critical points for the Abel-Jacobi
871: map $$\AbelJacobi\colon \Sym^g(\Sigma)\longrightarrow
872: \Pic{g}(\Sigma)\cong H^1(\Sigma;S^1),$$ which is a degree one holomorphic map.
873: Thus, the set of spheres is contained in a subset of real
874: codimension two.
875: \end{proof}
876:
877: \begin{prop}
878: \label{prop:CompactDegDisk}
879: Suppose $\x\in\Ta$ is not contained in any $\Sym^g(\sj)$-holomorphic sphere in $\Sym^g(\Sigma)$.
880: Then, there is a contractible neighborhood ${\mathcal U}$ of $\Sym^g(\sj)$ in
881: $\AlmostCx(\sj,\eta,V)$ with the property that for generic
882: $J\in{\mathcal U}$, the moduli space
883: $\UnparModDeg_J(O_\x+[S])$ is a compact, formally zero-dimensional
884: space which is smoothly cut out by its defining equations.
885: \end{prop}
886:
887: \begin{proof}
888: To investigate compactness, note first that a sequence of elements in
889: $\UnparModDeg_J(\x)$ has a subsequence which either bubbles off spheres, or
890: additional disks with boundaries lying in $\Ta$. However,
891: it is impossible for a sequence to bubble off a
892: null-homotopic disk with boundary lying in $\Ta$, since such disks must
893: be constant, as they have no energy (according to the
894: proof of Lemma~\ref{lemma:EnergyBound}, see
895: Equation~\eqref{eq:EnergyBound}). Moreover, sequences in $O_\x+[S]$ cannot bubble off
896: homotopically non-trivial disks, because then one of the components in the
897: decomposition would have negative $\omega$-integral, and such homotopy
898: classes have no holomorphic representatives.
899:
900: This argument also rules out bubbling off spheres, except in the
901: special case where the subsequence converges to a single sphere (more
902: precisely, the constant disk mapping to $\x$, attached to some
903: sphere). But this is ruled out by our hypothesis on $\sj$, which ensures that
904: for any $J$ sufficiently close to $\Sym^g(\sj)$, the $J$-holomorphic
905: spheres are disjoint from $\x$.
906:
907: To prove smoothness, note first that any holomorphic disk in
908: $\UnparModDeg(\x)$ for $\Sym^g(\sj)$ has a dense set of injective
909: points. To see this, fix any point $z'\in
910: \Sigma-\alpha_1-...-\alpha_g$. The intersection number of
911: $\{z'\}\times\Sym^{g-1}(\Sigma)$ with $u$ is $+1$, and both are
912: varieties; it follows that there is a single point of intersection,
913: i.e. there is only one $(s,t)$ for which $u(s,t)\in \{z'\}\times
914: \Sym^{g-1}(\Sigma)$.
915: Thus, $u$ is injective in a neighborhood of
916: $(s,t)$. It follows that for any $J$ sufficiently close to
917: $\Sym^g(\sj)$, all the $J$-holomorphic degenerate disks are injective in
918: a neighborhood of $u(s,t)$. Thus, according to the usual proof of
919: transversality, these pairs are all smooth points in the parameterized
920: moduli space. Thus, the result follows from the Sard-Smale theorem.
921: \end{proof}
922:
923: Proceeding as earlier, we can orient the moduli spaces
924: $\UnparModDeg_J$.
925: Our aim now is to prove the following:
926:
927: \begin{theorem}
928: \label{thm:GromovInvariant}
929: Fix a finite collection $\{\x_i\}$ of points in $\Sym^{g}(\Sigma)$,
930: and an almost-complex structure $\sj$ over $\Sigma$
931: for which each
932: $\Sym^{g}(\sj)$-holomorphic sphere misses $\{\x_i\}$.
933: Then, there is a contractible open neighborhood of $\Sym^g(\sj)$,
934: ${\mathcal U}$, in the space
935: of nearly-symmetric almost-complex structures, with the property that
936: for generic $J\in {\mathcal U}$, the total signed number of points in
937: $\UnparModDeg_J(O_{\x_i}+[S])$ is zero.
938: \end{theorem}
939:
940: Thinking of $\Sigma$ as the connected sum of $g$ tori, each of which
941: contains exactly one $\alpha_i$, we can endow $\Sigma$ with a complex
942: structure with long connected sum tubes.
943:
944:
945: \begin{prop}
946: \label{prop:ModDegEmpty}
947: If $\sj$ is sufficiently stretched out along the connected sum tubes,
948: then the moduli space ${\widehat \ModDeg}_\sj(O_\x+[S])$ is empty for any $\x\in\Ta$.
949: \end{prop}
950:
951: \begin{proof}
952: Fix a genus one Riemann surface $E$. Let $\sj_t$
953: denote the complex structure on $\Sigma$, thought of as the connected
954: sum of $g$ copies of $E$, connected along cylinders isometric to
955: $S^1\times [-t,t]$. As $t\goesto\infty$, the Riemann surface degenerates to the wedge
956: product of $g$
957: copies of $E$, $\bigvee_{i=1}^g E_i$.
958:
959: If for each $\sj_t$, the moduli space were non-empty, we could take
960: the Gromov limit of a sequence $u_t$ in ${\widehat
961: \ModDeg}_{\sj_t}(O_\x+[S])$
962: to obtain a holomorphic map $u_\infty$ into $\Sym^g(E_1\vee...\vee
963: E_g)$ (a linear chain of $g$ tori meeting in $g-1$ nodes). (In this
964: argument, we have a one-parameter family of symmetric products, which
965: we can embed into a fixed K\"ahler manifold, where we can apply the
966: usual Gromov compactness theorem, see also
967: Section~\ref{sec:Stabilization}.) The latter symmetric product
968: decomposes into irreducible components $$\bigcup_{\{k_1,...,k_g \in \Z
969: \big| 0\leq k_i \leq g,~~k_1+...+k_g=g\}}
970: \Sym^{k_1}(E_1)\times...\times\Sym^{k_g}(E_g).$$
971: These components meet along loci containing the connected sum points
972: for the various $E_i$. Moreover, the torus $\Ta$ can be viewed as a
973: subset of the irreducible component $E_1\times...\times E_g$
974: (corresponding to all $k_i=1$). The Gromov limit $u_\infty$ then
975: consists of a holomorphic disk $v$ with boundary mapping into $\Ta$,
976: and a possible collection of spheres bubbling off into the other
977: irreducible components. But $\pi_2(E_1\times...\times
978: E_g,\alpha_1\times...\times\alpha_g)=0$, so it follows that $v$ is
979: constant, mapping to $\x\in\Ta$ (which is disjoint from the connected
980: sum points). Since $v$ misses the other components of the symmetric
981: product, it cannot meet any of the spheres, so $v$ is the Gromov limit
982: of the $u_t$. But, $n_z(v)=0$, while we have assumed that
983: $n_z(u_t)=1$.
984: \end{proof}
985:
986: \begin{lemma}
987: \label{lemma:CompactCobordism}
988: Let $\x$, $\sj$, and ${\mathcal U}$
989: be as in Proposition~\ref{prop:CompactDegDisk}.
990: Suppose that $J_1,J_2\in {\mathcal U}$ are a pair of generic
991: almost-complex structures, in the sense that $\UnparModDeg_{J_s}(\x)$
992: is smooth for $s=0$ and $1$. Then, these moduli
993: spaces are compactly cobordant.
994: \end{lemma}
995:
996: \begin{proof}
997: We connect $J_1$ and $J_2$ by a generic path $\{J_s\}$ in ${\mathcal
998: U}$. As in the proof of Proposition~\ref{prop:CompactDegDisk}, this
999: gives rise to the required compact cobordism. Note that the
1000: possibility of bubbling off a sphere is ruled out, choosing ${\mathcal
1001: U}$ small enough to ensure that $\x$ is disjoint from all
1002: $J$-holomorphic spheres with $J\in{\mathcal U}$.
1003: \end{proof}
1004:
1005: \vskip.3cm
1006: \noindent{\bf Proof of Theorem~\ref{thm:GromovInvariant}.}
1007: Let $\sj$ be any complex structure over $\Sigma$ for which the
1008: $\Sym^{g}(\sj)$-holomorphic spheres miss the $\{\x_i\}$. Let
1009: ${\mathcal U}\subset \bigcap_{i=1}^n {\mathcal U}_i$ be a
1010: contractible, open subset of the the open subsets ${\mathcal U}_i$
1011: given to us by Proposition~\ref{prop:CompactDegDisk} for the points
1012: $\x_i\in\Ta$. According to Lemma~\ref{lemma:CompactCobordism}, the
1013: number of points $\#\UnparModDeg_J(O_{\x_i}+[S])$ is independent of
1014: $J$, i.e. it depends only on the complex structure $\sj$ over
1015: $\Sigma$. In fact, if $J$ is a generic $\sj$-nearly-symmetric
1016: almost-complex structure, and $J'$ is a sufficiently close
1017: $\sj'$-nearly-symmetric almost-complex structure, then the moduli
1018: spaces are identified. It follows that
1019: $\#\UnparModDeg_J(O_{\x_i}+[S])$ is a locally-constant function of the
1020: complex structure $\sj$. Since the space of complex structures for
1021: which the $\Sym^{g}(\sj)$-holomorphic spheres miss $\{\x_i\}$ is
1022: connected, the theorem follows from
1023: Proposition~\ref{prop:ModDegEmpty}.
1024: \qed
1025: \vskip.3cm
1026:
1027: \subsection{Structure of moduli spaces}
1028:
1029: \begin{theorem}
1030: \label{thm:StructMod}
1031: Let $(\Sigma,\alphas,\betas)$ be a Heegaard diagram with curves in
1032: general position. For a generic path $J_s$ of nearly-symmetric
1033: almost-complex structures, we have the following. There is no
1034: non-constant $J_s$-holomorphic disk $u$ with $\Mas(u)\leq 0$. Moreover
1035: for each $\phi\in\pi_2(\x,\y)$ with
1036: $\Mas(\phi)=1$, the quotient space
1037: $$\UnparModSp(\phi)=\frac{\ModSp(\phi)}{\R}$$ is a compact,
1038: zero-dimensional manifold.
1039: \end{theorem}
1040:
1041: \begin{proof}
1042: The first part follows directly from the Theorem~\ref{thm:Transversality}.
1043:
1044: Compactness follows from the usual compactification theorem for
1045: holomorphic curves
1046: (see~\cite{FloerUnregularized} and also~\cite{Gromov},
1047: \cite{ParkerWolfson}, \cite{Ye}), which holds thanks to the energy bound
1048: (Lemma~\ref{lemma:EnergyBound}).
1049:
1050: Specifically, the compactness theorem says that a sequence of points
1051: in the moduli spaces converges to an ideal disk, with possible broken
1052: flowlines, boundary degenerations, and bubblings of
1053: spheres. Broken flowlines are excluded by the additivity of the Maslov
1054: index, and the transversality result
1055: Theorem~\ref{thm:Transversality}. Spheres and boundary
1056: degenerations both carry Maslov index at
1057: least two, so these kinds of degenerations are excluded as well.
1058: \end{proof}
1059: