math0512286/analysis.tex
1: \section{Analytic input}
2: \label{sec:Analysis}
3: 
4: We will be concerned in this section with gluing results for
5: pseudo-holomorphic disks.  Consider an $\ell$-pointed Heegaard diagram
6: $(\Sigma,\alphas,\betas,\{w_1,...,w_\ell\})$, and let $d=g+\ell-1$.
7: Given $\x,\y\in\Ta\cap\Tb$ and a homology class of Whitney disk
8: $\phi\in\pi_2(\x,\y)$, we can form the moduli space $\ModFlow(\phi)$.
9: 
10: Recall that these moduli spaces have Gromov compactifications,
11: cf.~\cite{Gromov}, \cite{McDuffSalamon}, \cite{FloerLag}, \cite{FOOO}.
12: For a given moduli space of pseudo-holomorphic Whitney disks,
13: these Gromov compactifications include possibly 
14: moduli spaces of pseudo-holomorphic
15: Whitney disks connecting other intersection points,  moduli
16: spaces of pseudo-holomorphic spheres, and finally also
17: moduli spaces of further degenerate disks called
18: {\em boundary degenerations}. More formally,
19: given a point $\x\in\Ta$, we let $\pi_2^{\beta}(\x)$ denote the space
20: of homology classes of maps $$\left\{ u\colon [0,\infty)\times
21:   \R\longrightarrow \Sym^g(\Sigma)\Bigg|
22: \begin{array}{l}
23: u(\{0\}\times \R)\subset \Tb \\
24: \lim_{z\goesto \infty}u(z)=\x 
25: \end{array}
26: \right\}.$$
27: Such a map is called a $\beta$-boundary degeneration.
28: We let $\pi_2^{\alpha}(\x)$ denote the set of $\alpha$-boundary
29: degenerations, defined analogously.
30: 
31: 
32: Loosely speaking, Gromov's compactness theorem states that a sequence
33: of pseudo-holomorphic curves representing $\phi$, has a subsequence
34: which converges locally to a ``broken flow-line'', consisting of
35: collection of pseudo-holomorphic flow-lines $\{\phi_i\}$, a collection
36: of $\alpha$- and $\beta$-boundary degenerations $\{\psi_j\}$, and
37: finally a collection of pseudo-holomorphic spheres $\{S_k\}$ with
38: $$\sum\cald(\phi_i)+\sum\cald(\psi_j)+\sum\cald(S_k)=\cald(\phi).$$
39: 
40: We will also find it necessary at several future points to study ends
41: of moduli spaces as the Heegaard surface is degenerated. We turn to
42: this more formally as follows.
43: 
44: \subsection{Gluing moduli spaces}
45: \label{subsec:FiberedProducts}
46: 
47: Let $(\Sigma_1,\alphas_1,\betas_1,z_1)$ and
48: $(\Sigma_2,\alphas_2,\betas_2,z_2)$ be a pair of Heegaard diagrams,
49: where here $\alphas_i$ and $\betas_i$ are $d_i$-tuples of attaching
50: circles in $\Sigma_i$. We can form their connected sum at $z_1\in
51: \Sigma_1$ and $z_2\in \Sigma_2$ to obtain a new surface
52: $\Sigma=\Sigma_1\#\Sigma_2$, endowed with sets of attaching circles
53: $\alphas_1\cup\alphas_2$ and $\betas_1\cup\betas_2$. We will assume
54: that $d_i\geq g_i$, and write $d=d_1+d_2$ and $g=g_1+g_2$.
55: 
56: We will need here descriptions of the moduli spaces of flowlines in
57: the connected sum diagram, in terms of moduli spaces for the two
58: pieces.
59: 
60: Fix points $\x_i,\y_i\in{\mathbb T}_{\alpha_i}\cap{\mathbb
61: T}_{\beta_i}\subset \Sym^{d_i}(\Sigma_i)$, which in turn give rise to
62: points $\x_1\times \x_2$, $\y_1\times\y_2$ in
63: $\Sym^{d_1+d_2}(\Sigma_1\#\Sigma_2)$.  Fix a pair of homology
64: classes of Whitney disks
65: $\phi_i\in\pi_2(\x_i,\y_i)$ with
66: $n_{z_1}(\phi_1)=n_{z_2}(\phi_2)$. These can be combined naturally to
67: form a homology class
68: $\phi_1\#\phi_2\in\pi_2(\x_1\times\x_2,\y_1\times\y_2)$.
69: Specifically, the local multiplicities of $\phi_1\#\phi_2$ at each
70: domain $\cald\subset \Sigma_1\#\Sigma_2$ is the local multiplicity of
71: $\phi_i$ at the corresponding $\cald\subset \Sigma_1$ or $\Sigma_2$.
72: 
73: Conversely, each homology class
74: $\phi\in\pi_2(\x_1\times\y_1,\x_2\times\y_2)$ can be uniquely
75: decomposed as $\phi=\phi_1\#\phi_2$ for some pair of
76: $\phi_i\in\pi_2(\x_i,\y_i)$ with $n_{z_1}(\phi_1)=n_{z_2}(\phi_2)$.
77: 
78: Moreover, given complex structures $j_1$ and
79: $j_2$ on $\Sigma_1$ and $\Sigma_2$, we can form a complex structure 
80: $J(T)$ with neck-length $T$ as follows: find conformal 
81: disks $D_1$ and $D_2$ about
82: $z_1$ and $z_2$, and form the connected sum
83: $$\Sigma(T)=(\Sigma_1-D_1)\# \left([-T-1,T+1]\times S^1\right) \# (\Sigma_2-D_2),$$
84: under identifications $\partial D_1\cong \{-T\}\times S^1$
85: and $\{T\}\times S^1 \cong \partial D_2$. As $T\goesto\infty$, the
86: conformal structure on $\Sigma(T)$ converges to
87: the nodal curve $\Sigma_1\vee\Sigma_2$.
88: 
89: 
90: 
91: \begin{theorem}
92: \label{thm:GenericFiberedProduct}
93: Fix diagrams $(\Sigma_i,\alphas_i,\betas_i,z_i)$ for $i=1,2$ as above,
94: where here $\alphas_i$ and $\betas_i$ are $d_i$-tuples of attaching
95: circles.  Given a homology class $\phi=\phi_1\#\phi_2$ for the
96: connected sum of the two diagrams, we have that
97: $$\Mas(\phi)=\Mas(\phi_1)+\Mas(\phi_2)-2k,$$ 
98: where $k=n_{z_1}(\phi)=n_{z_2}(\phi)$. Suppose that
99: $\ModFlow(\phi)\neq \emptyset$
100: for a sequence of almost-complex structures $J(T_i)$ 
101: with $T_i\goesto \infty$. Then, the moduli spaces of broken pseudo-holomorphic
102: flowlines representing $\phi_1$ and $\phi_2$ (i.e. the Gromov
103: compactifications of these two moduli spaces) are non-empty. Finally,
104: suppose that $\Mas(\phi_1)=1$, $\Mas(\phi_2)=2k$, and also that $d_2>g_2$;
105: and consider the
106: maps
107: \begin{eqnarray*}
108: \rho_1\colon \ModFlow(\phi_1) \longrightarrow \Sym^{k}(\CDisk) 
109: &{\text{and}}&
110: \rho_2\colon \ModFlow(\phi_2) \longrightarrow \Sym^{k}(\CDisk),
111: \end{eqnarray*}
112: where here
113: $$\rho_i(u)=u^{-1}(\{z_i\}\times \Sym^{d_i-1}(\Sigma_i)).$$
114: If the fibered product of $\ModFlow(\phi_1)$ and $\ModFlow(\phi_2)$
115: $$\ModFlow(\phi_1)\times_{\Sym^k(\CDisk)} \ModFlow(\phi_2)
116: =\{u_1\times u_2 \in\ModFlow(\phi_1)\times \ModFlow(\phi_2)\big|
117: \rho_1(u_1)=\rho_2(u_2)\}$$
118: is a smooth manifold, then there is an identification of this moduli
119: space with the moduli space $\ModFlow(\phi)$, for sufficiently 
120: long connected sum length.
121: \end{theorem}
122: 
123: 
124: There are three assertions in the above theorem: one concerns the
125: Maslov index, one the existence of weak limits, and the third is a
126: gluing result. They are arranged in order of difficulty; the third
127: requires the most work.  One could approach this problem from the
128: point of view of degenerating $\Sym^{d}(\Sigma_1\#\Sigma_2)$ as the
129: connected sum degenerates into $\Sigma_1\vee\Sigma_2$, following the
130: approach to stabilization invariance from~\cite{HolDisk}.  In this
131: case, the limiting symplectic space is
132: $\Sym^{d}(\Sigma_1\vee\Sigma_2)$, which has a fairly complicated
133: singular set (consisting of those $d$-tuples where one element is the
134: singular point $p\in \Sigma_1\vee\Sigma_2$). In particular, the
135: singular set is, in itself, not a symplectic manifold but rather a
136: singular space whose singularities consist of those $d$-tuples where
137: at least two elements are the point $p$.  Under the present
138: circumstances, holomorphic disks we wish to resolve -- whose boundary
139: lies from in the the top stratum
140: $\Sym^{d_1}(\Sigma_1)\times\Sym^{d_2}(\Sigma_2)\subset
141: \Sym^{d}(\Sigma_1\vee\Sigma_2)$ -- meet the singular stata (consisting of
142: those tuples where at least one point is the singular point $z\in
143: \Sigma_1\vee\Sigma_2$) in a complex codimension one subset, i.e. where
144: at least two coordinates agree with this singular point.
145: 
146: A much simpler approach can be given using Lipshitz's cylindrical
147: reformulation of Heegaard Floer homology, cf.~\cite{LipshitzCyl}.
148: With this reformulation, then, the gluing problem takes place 
149: in a four-manifold, along a singular set which is a manifold,
150: (placing it on roughly an equal footing with the proof of stabilization
151: invariance for cylindrical reformulation, cf.~\cite{LipshitzCyl}). 
152: This kind of degeneration has been extensively studied in the literature,
153: cf.~\cite{IonelParker}, \cite{LiRuan}, \cite{EGH}, \cite{BEHWZ},
154: and of course~\cite{LipshitzCyl}. We
155: turn to this approach, first recalling the basic set-up of Lipshitz's picture.
156: 
157: \subsection{Lipshitz's cylindrical formulation of Heegaard Floer homology}
158: 
159: The starting point of Lipshitz's formulation is that a holomorphic
160: disk $u\colon\CDisk\longrightarrow \Sym^d(\Sigma)$ corresponds to a
161: holomorphic curve in $\CDisk\times\Sigma$, which, for any
162: $p\in\CDisk$, meets the fiber $\Sigma\times \{p\}$ in the 
163: set of $d$ points $u(p)$.  Thus, one could reformulate the chain
164: complex defining Heegaard Floer homology as counting certain
165: pseudo-holomorphic surfaces in $\CDisk\times \Sigma$. 
166: 
167: This can be made more precisely as follows.
168: Consider 
169: the four-manifold 
170: $$W=\Sigma\times[0,1]\times\R,$$ equipped
171: with two projection maps
172: \begin{eqnarray*}
173:   \pi_{\Sigma}\colon W \longrightarrow \Sigma
174:   &{\text{and}}&
175:   \pi_{\CDisk}\colon W\longrightarrow [0,1]\times\R.
176: \end{eqnarray*}
177: (As usual, we think of the unit disk in the complex plane $\CDisk$
178: as the conformal compactification of the infinite strip $[0,1]\times \R$,
179: obtained by adding points $\pm i$ at infinity.)
180: Endow $W$ with an almost-complex structure $J$ tamed by a natural split
181: symplectic form on $W$, which is translation invariant in the
182: $\R$-factor, and for which the projection $\pi_{\CDisk}$ is
183: a pseudo-holomorphic map.
184: For example, the product complex structure -- which will be called
185: a {\em split} complex structure -- satisfies these conditions,
186: but it is often useful to perturb this. However, to ensure
187: positivity, it is convenient to choose points $z\in \Sigma$,
188: and require that $J$ is split in a neighborhood of $z$ 
189: (in $\Sigma$) times $[0,1]\times \R$. Such an almost-complex structure
190: $J$ is called {\em split near $z$}.
191: 
192: Consider next a Riemann surface $S$ with boundary,
193: $d$ ``positive'' punctures $\{p_1,...,p_d\}$ and $d$ ``negative''
194: punctures $\{q_1,...,q_d\}$ on its boudnary. 
195: 
196: Lipshitz considers pseudo-holomorphic 
197: maps $$\uL\colon S\longrightarrow W,$$
198: satisfying the following conditions:
199: \begin{itemize}
200:     \item $\uL$ is smooth.
201:     \item $\uL(\partial S)\subset \left(\alphas\times\{1\}\times \R\right)\cup
202:       \left(\betas\times\{0\}\times\R\right)$.
203:     \item No component in the image of $\uL(S)$ is contained in a fiber of 
204:       $\pi_{\CDisk}$.
205:     \item For each $i$, $\uL^{-1}(\alpha_i\times\{1\}\times \R)$
206:       and $\uL^{-1}(\beta_i\times\{0\}\times \R)$
207:       consist of exactly one component of $\partial S-\{p_1,...,p_d,q_1,...,q_d\}$
208:     \item The energy of $\uL$ is finite.
209:     \item $\uL$ is an embedding.
210:     \item Any sequence of points in $S$ converging to $q_i$ resp. $p_i$
211: 	is mapped under $\pi_{\CDisk}$ to a sequence of points
212: 	whose second coordinate converges to $\{-\infty\}$ resp $\{+\infty\}$.
213: \end{itemize}
214: We call holomorphic curves of this type {\em cylindrical flow-lines}.
215: Thinking of the complex disk $\CDisk$ as a compactification of
216: $[0,1]\times \R$, a map $\uL$ as above
217: can be extended continuously to a map of the closure of $S$ into
218: ${\overline W}=\Sigma\times\CDisk$.  We say that $\uL$ connects $\x$
219: to $\y$ if the image of this extension meets $\Sigma\times \{-i\}$
220: in the points $\x\times\{-i\}$ and it meets $\Sigma\times\{i\}$ in the
221: points $\y\times \{i\}$.
222: 
223: Projecting such a map $\uL$ onto $\Sigma$, we obtain a relative
224: two-chain in $\Sigma$ relative to $\alphas\cup\betas$, whose local
225: multiplicity at some point $z\in\Sigma$ is given by the intersection
226: number $$\nL_z(\uL)=\#\Big(\uL\cap
227: \left(\{z\}\times[0,1]\times\R\right)\Big).$$ Conversely, given
228: $\phi\in\pi_2(\x,\y)$, let $\ModFlowL(\phi)$ denote the moduli space
229: of cylindrical flow-lines $\uL$ which induce the same two-chain as
230: $\phi$. 
231: 
232: It is sometimes also useful to consider the analogue of boundary degenerations
233: in this cylindrical context. 
234: 
235: \begin{defn}
236: Consider a Riemann surface $S$ with
237: boundary and $d$ punctu      
238: res $\{p_1,...,p_d\}$ on its boundary. Consider now
239: pseudo-holomorphic maps ${\widetilde u}\colon S \longrightarrow \Sigma
240: \times [0,\infty)\times \R$ which are finite energy, smooth
241: embeddings, sending the boundary of $\Sigma$ into $\betas$, containing
242: no component in the fiber of the projection to $[0,\infty)\times\R$,
243: so that each component of ${\widetilde
244:   u}^{-1}(\beta_i\times\{0\}\times\R)$ consists of exactly one
245: component of $\partial S-\{p_1,...,p_d\}$. Such a map is called
246: a {\em cylindrical boundary degeneration}. For such a map, the point
247: at infinity is mapped into a fixed $\x\in\Tb$. These maps can be
248: organized into moduli spaces of $\ModDegL(\psi^\beta)$ indexed by homology
249: classes $\psi\in H_2(\Sigma,\betas)$. A corresponding definition can also
250: be made with $\alphas$ playing the role of $\betas$.
251: \end{defn}
252: 
253: In~\cite{LipshitzCyl}, Lipshitz sets up a theory analogous to Heegaard
254: Floer homology (in the case where $d=g$), counting elements of
255: $\ModFlowL(\phiL)$. In setting up this theory, he establishes the
256: necessary transversality properties: for a generic choice of
257: almost-complex structure $J$ over $W$ as above, all the moduli spaces
258: $\ModFlowL(\phiL)$ with $\Mas(\phiL)<0$ are empty; non-empty moduli
259: spaces $\ModFlowL(\phiL)$ with $\Mas(\phiL)=0$ consist of constant
260: flowlines, and moduli spaces with $\Mas(\phi)=1$ are smooth
261: one-manifolds. He also shows that these moduli spaces have the
262: necessary Gromov compactifications analogous to those for the moduli
263: spaces of holomorphic disks in a more traditional Lagrangian set-up.
264: Indeed, in~\cite{LipshitzCyl} Appendix A, Lipshitz establishes
265: identifications $\ModFlowL(\phi)\cong \ModFlow(\phi)$ for suitably
266: generic choices of almost-complex structures, in cases where
267: $\Mas(\phi)=1$. In particular, if we consider the map defined as in
268: Equation~\eqref{eq:DefD}, only using moduli spaces
269: $\ModFlowL(\phi)/\R$ in place of $\ModFlow(\phi)/\R$, then the two
270: maps actually agree.  Although Lipshitz considers
271: the case where $d=g$, the logic here applies immediately to prove
272: the corresponding result in the case where $d>g$, as well.
273: 
274: With this cylindrical formulation, now the denegeration considered in
275: Theorem~\ref{thm:GenericFiberedProduct} becomes more transparent.
276: Specifically, the degeneration of $\Sigma$ to $\Sigma_1\vee\Sigma_2$
277: corresponds to a generation of $W$ into
278: $W_1=\Sigma_1\times[0,1]\times\R$ and
279: $W_2=\Sigma_2\times[0,1]\times\R$, two symplectic manifolds which meet
280: along the hypersurface $[0,1]\times\R$, joined along
281: $\{p_1\}\times[0,1]\times\R\subset W_1$ and
282: $\{p_2\}\times[0,1]\times\R\subset W_2$. (Compare
283: also~\cite{IonelParker}, \cite{LiRuan}, \cite{EGH}, \cite{BEHWZ}.
284: Finally, observe that this is precisely the set-up which Lipshitz uses
285: in his proof of stabilization invariance for the cylindrical theory,
286: see \cite{LipshitzCyl}).  
287: 
288: More precisely, we start with almost-complex structures $J_1$ and
289: $J_2$ on $W_1$ and $W_2$ and neighborhoods $D_1$ and $D_2$ of $z_1$
290: and $z_2$.  We assume that $J_1$ and $J_2$ which are split on
291: $D_1\times[0,1]\times\R$ and $D_2\times [0,1]\times\R$ respectively.
292: From this, we construct a complex structure $J(T)$ on
293: $$W(T)=\Sigma(T)\times[0,1]\times\R,$$
294: which agrees with $J_1$ near $(\Sigma_1-D_1)\times[0,1]\times \R$,
295: and $J_2$ near $(\Sigma_2-D_2)\times[0,1]\times\R$,  and which 
296: is split over $S^1\times [-T-1,T+1]\times [0,1]\times\R$. 
297: 
298: 
299: \begin{defn}
300:         \label{def:PreGlued}
301:         A {\em pre-glued flowline representing the homology class
302:           $\phi=\phi_1\#\phi_2\in\pi_2(\x_1\times\y_1 ,\x_2\times
303:           \y_2)$} is a pair of cylindrical flow-lines ${\widetilde
304:           u}_1\in\ModFlowL(\phiL_1)$ and ${\widetilde
305:           u}_2\in\ModFlow(\phiL_2)$ satisfying the matching condition
306:         $$(\pi_{\CDisk}\circ\uL_1)
307:         \left(\pi_{\Sigma_1}\circ\uL_1)^{-1}(\{z_1\})\right) =
308:         (\pi_{\CDisk}\circ\uL_2)\left(\pi_{\Sigma_2}\circ
309:           \uL_2)^{-1}(\{z_2\})\right).$$  Similarly, a {\em pre-glued
310:           $\alpha$-boundary degeneration} is a pair of
311:         $\alpha$-boundary degenerations ${\widetilde
312:           v}_1\in\ModFlow(\phi_1)$ and ${\widetilde
313:           v}_2\in\ModFlow(\phi_2)$ satisfying the analogous matching
314:         condition $(\pi_{\CDisk}\circ {\widetilde
315:           u})\left(\pi_{\Sigma_1}\circ {\widetilde
316:             v}_1)^{-1}(\{z_1\})\right) = (\pi_{\CDisk}\circ
317:         {\widetilde u}\left(\pi_{\Sigma_2}\circ {\widetilde
318:             v}_2)^{-1}(\{z_2\})\right)$. A similar definition can also
319:         be made for $\beta$-boundary degenerations.
320: \end{defn}
321: 
322: The curves $\alpha_1\cup...\cup \alpha_{g+\ell-1}$ divide $\Sigma$ into
323: $\ell$ regions $A_1,...,A_\ell$. 
324: Choose reference points $w_i$, one in each $A_i$.
325: 
326: It will be useful to have the following:
327: 
328: \begin{lemma}
329:   \label{lemma:MaslovBoundaryDegenerations}
330:   Given ${\widetilde \psi}\in\pi_2^\beta(\x)$, we have that 
331:   $\Mas({\widetilde\psi})=2\sum_{i=1}^\ell n_{w_i}({\widetilde \psi})$.
332: \end{lemma}
333: 
334: \begin{proof}
335:   This follows readily from the excision principle for the 
336:   the linearized $\dbar$ operator, to reduce to the case of a disk.
337:   (See the proof of Theorem~\ref{thm:CountBoundaryDeg} for a more
338:   detailed discussion of a related problem.)
339: \end{proof}
340: 
341: We interrupt now our path to Theorem~\ref{thm:GenericFiberedProduct},
342: paying off an earlier debt, supplying the following quick consequence
343: of the above lemma:
344: 
345: \vskip.2cm
346: \noindent{\bf{Proof of Proposition~\ref{prop:MaslovIndexFormula}}}
347: According to Equation~\eqref{eq:GroupPerDoms}, the homology classes of
348: $\phi$ and $\phi'$ differ by the juxtaposition of a boundary
349: degeneration with $\cald(\psi)=\sum_i (n_{w_i}(\phi)-n_{w_i}(\phi'))$,
350: whose index, according to
351: Lemma~\ref{lemma:MaslovBoundaryDegenerations}, is given by
352: $2\sum_i (n_{w_i}(\phi)-n_{w_i}(\phi'))$.  The result now follows
353: from the additivity of the index under juxtaposition.
354: \qed \vskip.2cm
355: 
356: We give another consequence of Lemma~\ref{lemma:MaslovBoundaryDegenerations}.
357: By arranging for the almost-complex structure on $W$ to be split near
358: neighborhoods of all the $w_i$, one can arrange for the usual
359: positivity principle to hold: if a moduli space $\ModDeg(\psi^\beta)$
360: is non-empty, then all the $n_{w_i}(\psi^\beta)\geq 0$.  It follows
361: from this, together with
362: Lemma~\ref{lemma:MaslovBoundaryDegenerations}, that if $\psi$ is a
363: homology class of boundary degenerations which contains a non-constant
364: pseudo-holomorphic representative, then $\Mas(\psi)\geq 2$. This 
365: principle is used in the following:
366: 
367: 
368: \vskip.2cm
369: \noindent{\bf{Proof of the cylindrical analogue of
370: Theorem~\ref{thm:GenericFiberedProduct}}.}
371: We turn to Theorem~\ref{thm:GenericFiberedProduct}, using 
372: moduli spaces of cylindrical flowlines in place of pseudo-holomorphic
373: Whitney disks. In this context, the restriction maps are to be replaced
374: by maps
375: $$\rhoL_i\colon \ModFlowL(\phiL_i)\longrightarrow \Sym^k(\CDisk)$$
376: given by
377: $$\rhoL_i(\uL)=(\pi_{\CDisk}\circ \uL)\left((\pi_{\Sigma}\circ \uL)^{-1}(\{z_i\})\right).$$
378: 
379: 
380: The formula for the Maslov index follows readily from the excision
381: principle for the linearized $\DBar$ operator, using the cylindrical
382: formulation. Consider a sequence of pseudo-holomorphic curves
383: $\{v_t\}_{t\in\Z}\in\ModFlowL_{J(t)}(\phi)$, where here the subscript
384: $J(t)$ denotes the almost-complex structure induced on $W(t)$ with
385: neck-length $t$ as described earlier. Using Gromov's compactness
386: theorem, after passing to a subsequence, $v_{t}$ converges locally to
387: a pseudo-holomorphic curve in the symplectic manifold $$W(\infty)
388: \cong \Big(W_1-\{z_1\}\times[0,1]\times \R\Big) \coprod
389: \Big(W_2-\{z_2\}\times[0,1]\times \R\Big),$$
390: which in turn can be
391: completed to a pseudo-holomorphic curve in $W_1$ and one in $W_2$.
392: More precisely, we obtain a broken flow-line whose components consist
393: of pre-glued flowlines and boundary degenerations, finally also nodal
394: curves supported entirely inside fibers $\Sigma_1\vee\Sigma_2$.  The
395: representatives in the moduli spaces for $\phi_1$ and $\phi_2$
396: respectively are gotten by ignoring the matching conditions.
397: 
398: In the case where $\Mas(\phi_1)=1$, the limiting process generically
399: gives rise rise to an unbroken, preglued flowline, according to the
400: following dimension counts. Specifically, taking a Gromov
401: compactification, we obtain a pseudo-holomorphic representative
402: $\uL_1$ of $\phi_1$, and also a possibly 
403: broken flow-line representing $\phi_2$.
404: 
405: Assuming that this broken flow-line contains no components which are
406: closed curves, there is some component of it $u_2$ with the property
407: that $u_1$ and $u_2$ represents a pre-glued flowline in the sense of
408: Definition~\ref{def:PreGlued}.  We claim that $u_2$ in fact represents
409: $\phi_2$.  If $\uL_2$ did not represent $\phi_2$, then it represents a
410: homology class $\phi'$ which is a component in the Gromov
411: compactification of $\phi'$.  If this compactification contains
412: additional boundary degenerations, the Maslov index of $\phi'$ is at
413: least $2$ smaller than the Maslov index of $\phi_2$ (in view of
414: Lemma~\ref{lemma:MaslovBoundaryDegenerations}). Moreover, if the
415: compactification contains other flows, those will serve only to
416: further decrease the Maslov index of $\phi'$ relative to that of
417: $\phi$. In sum, in the case where $\phi'$ does not agree with
418: $\phi_2$, its Maslov index $\Mas(\phi')<2k$. But given $\Delta\in
419: \Sym^k(\CDisk)$, the moduli space
420: $$\rho_2^{-1}(\Delta)=\{\uL\in\ModFlowL(\phiL')\big| 
421: (\pi_\CDisk\circ \uL_2)\left(\pi_\Sigma\circ \uL_2)^{-1}(\{z_2\})\right)=\Delta\} $$
422: has expected dimension $\Mas(\phi')-2k<0$. Thus, for a generic choice
423: of $\Delta$ (gotten by $\rho_1(\phi_1)$), this space is empty. 
424: 
425: It is easy to rule out also the case that the Gromov limiting
426: representing representing $\phi_2$ cannot have any closed
427: components. To this end, suppose it has some components which
428: represent the homology class $m[\Sigma_2]$ for some $m>2$. After
429: deleting these components, we are left with a homology class $\phi'$
430: with $\Mas(\phi_2')=2(d_2-g_2+1)$
431: (c.f. Lemma~\ref{lemma:MaslovBoundaryDegenerations}). Some component
432: $u_2$ of this Gromov compactification has $\rho_2(u_2)=\Delta'$, where
433: $\Delta'$ is obtained from $\rho_1(\phi_1)$ by deleting $m$ points.
434: But the moduli space of such points has expected dimension given by
435: \begin{equation}
436: \label{eq:eDimModSp}
437: \Mas(\phi_2')-2(k-m) 
438: = 2k-2(k-m)-2(d_2+1-g_2)m
439: =-2m(d_2-g_2)<0,
440: \end{equation}
441: and hence it is empty (here, of course, is where
442: we used our assumption that  $d_2>g_2$). 
443: 
444: Thus, we have established that a sequence of holomorphic
445: representatives for $\phi$ has a Gromov limit (as we stretch the neck)
446: to a pre-glued flowline representing $\phi_1$ and $\phi_2$.
447: Conversely, given a pre-glued flowline, one can obtain a
448: pseudo-holomorphic curve in $\ModFlowL(\phi)$ by gluing
449: (cf.~\cite{LipshitzCyl} for further details on this gluing problem,
450: and~\cite{BEHWZ} for a discussion of gluing in a very general
451: context).
452: \qed
453: 
454: 
455: \subsection{Counting boundary degenerations}
456: \label{subsec:BoundaryDegenerations}
457: 
458: We let $\ModDegL(\psi)$ denote the moduli
459: space of pseudo-holomorphic boundary degenerations
460: in the homology class of  $\psi$.  Note that
461: ${\mathbb P}{\mathrm SL}_2(\R)$ acts on the this moduli space, and we
462: let $\UnparModDegL(\psi)$ denote the quotient by this action.
463: The principles used in proving Theorem~\ref{thm:GenericFiberedProduct}
464: can also be used to count boundary degenerations.
465: 
466: \begin{theorem}
467: \label{thm:CountBoundaryDeg}
468: Consider $\Sigma$, a surface of genus $g$, equipped with a
469: set of attaching circles $\alpha_1,...,\alpha_{g+\ell-1}$ for a handlebody.
470: If $\cald(\psi)\geq 0$ and $\Mas(\psi)=2$, then
471: $\cald(\psi)=A_i$ for some $i$; and indeed in this case
472: \begin{equation}
473:   \label{eq:CountBoundaryDegenerations}
474:   \#\UnparModDegL(\psi)\equiv 
475:   \left\{\begin{array}{ll}
476:       0 \pmod{2} & {\text{if $\ell=1$}} \\
477:       1\pmod{2} & {\text{if $\ell>1$.}}
478:     \end{array}\right.
479: \end{equation}
480: \end{theorem}
481: 
482: \begin{proof}
483:   It follows readily from
484:   Lemma~\ref{lemma:MaslovBoundaryDegenerations} that if $\psi$
485:   is a non-zero homology class of boundary degenerations with 
486:   $\cald(\psi)\geq 0$, then $\Mas(\psi)\geq 2$, with equality iff
487:   $\cald(\psi)=A_i$. In the case where $\cald(\phi)=A_i$, it remains
488:   to verify Equation~\eqref{eq:CountBoundaryDegenerations}.  The case
489:   where $\ell=1$ has already been established
490:   in~\cite[Proposition~\ref{HolDisk:thm:GromovInvariant}]{HolDisk} for
491:   the usual Heegaard-Floer moduli spaces and~\cite{LipshitzCyl} for
492:   the cylindrical version.
493:   Consider some region $A_i$, which we re-name simply $A'$. Recall
494:   that this is a Riemann surface with boundary, equipped with $m+p$
495:   curves $\alpha_1,...,\alpha_{m+p}$, of which the first $m$ comprise
496:   its boundary, and the rest are pairwise disjoint, embedded circles in
497:   the interior. We have fixed also
498:   $\x\in\alpha_1\times...\times \alpha_{m+p}$.  We reduce to the case where
499:   $p=0$, by de-stabilizing.  Next, we reduce to the case where the
500:   number of boundary circles is one. To this end, we can write $A'=A\#
501:   D$, where $D$ here is a disk with boundary $\alpha_{m}$, and $A$ is
502:   a planar surface-with-boundary $\alpha_1,...,\alpha_{m-1}$.  Denote
503:   the connected sum point in $A$ by $z_1$ and the one in $D$ by $z_2$.
504:   Degenerating the connected sum tube, we obtain a fibered product
505:   description
506:   $\ModDegL(A')=\ModDegL(A)\times_{\Sym^1(\CDisk)}\ModDegL(D)$, where the
507:   fibered product is taken over the maps
508:   \begin{eqnarray*}
509:     \rho_1\colon \ModDegL(A)\longrightarrow \Sym^1(\CDisk)
510:     &{\text{and}}&
511:     \rho_2\colon \ModDegL(D)\longrightarrow \Sym^1(\CDisk)
512:   \end{eqnarray*}
513:   defined as before.
514:   It is easy to see, though that $\ModDegL(D)$ is smooth, and the 
515:   map $\rho_2$ is in fact a diffeomorphism. Thus, it follows
516:   that $\ModDegL(A')\cong \ModDegL(A)$. In this manner, we have reduced
517:   to the case where $p=0$ and $m=1$, which is, once again, the case where
518:   $A=D$. Now, since the map $\rho_2$ is ${\mathbb P}{\mathrm SL}_2(\R)$-equivariant, we see
519:   that $\#\UnparModDeg(D)$ is precisely the degree of the map $\rho_2$,
520:   which is one.
521: \end{proof}
522: 
523: \subsection{Notational remark}
524: 
525: The notation of Theorem~\ref{thm:GenericFiberedProduct} suggests using
526: moduli spaces of pseudo-holomorphic Whitney disks, while its proof
527: uses cylindrical flow-lines.  One could close the gap by either
528: appealing to an identification between the two kinds of moduli spaces,
529: cf.~\cite{LipshitzCyl}, or simply adopting the cylindrical point of
530: view instead; either approach has the same final outcome. In view of
531: this remark, we henceforth drop the notational distinction between
532: cylindrical and more traditional moduli spaces.
533: 
534: 
535: