Reidemeister torsion in Floer-Novikov theory and counting pseudo-holomorphic tori. II (Q2369458)

This is the continuation of the review of Part I [\textit{Y. J. Lee}, J. Symplectic Geom. 3, 221--311 (2006; Zbl 1093.53091)]. Hereafter referred to as Part I of this series of papers. Same terminologies and notations as in Part I are used in this review. In Part I of this series of papers, a Floer-theoretic invariant \(I_F\) is introduced for (monotone) symplectic manifolds and its invariance is shown under general symplectic isotopies. The author proved the perturbation of a 1-parameter family of Floer systems into an admissible \((J,X)\)-homotopy, \(J\) an \(\omega\)-compatible complex structure of \(K\), the subbundle of \(TT_f\) consisting of tangent vectors to the fibre and \(X\) is a symplectic vector field (Part~I, \S6). This homotopy is assumed to have good properties (RHFS and NEP, given in \S 4 of Part~I). This assumption is proved in Part~I. Now, in Part~II, for an admissible \((J,X)\)-homotopy \((J^\Lambda,X^\Lambda)\) connecting two regular parts \((J_1,X_1)\), \((J_2,X_2)\), the following is shown: 1. (Corner structures of parametric moduli spaces): Codirectional homotopy of Floer systems (CHFS) generated by \((J^\Lambda,X^\Lambda)\) satisfies properties of regular homotopy of Floer systems (RHFS2c and RHFS3c, Part I, \S4.3.5 and 4.3.6). 2. (Orientation): The parametrized moduli spaces \({\mathcal M}_P^{\Lambda, +}\) and \(\widetilde{\mathcal M}_O^{\Lambda,1,+}\) may be, respectively, given coherent and grading compatible orientations such that (RHFS4, Part I, \S4.3.7) holds. 3. (Existence of nonequivariant perturbations (NEP)): (NEP) (Part I, \S4.4.5) holds for all Type II handleslides (Part I, \S4.3.5) in the CHFS generated by \((J^\Lambda,X^\Lambda)\). (1) follows from gluing theorems which are proved in \S\S2--6. (2) is proved in the last section (\S7) and (3) is proved in \S6.2-6.3. Gluing theory is the unifying theme of this paper. It is useful for studying the local structure of a stratified moduli space, usually coming from compactification. A typical gluing theorem constructs a map from \(\Xi (\mathbb{S})\), a gluing parameter space associated with a codimension \(>0\) stratum \(\mathbb{S}\), to a neighborhood of \(\mathbb{S}\) in the moduli space of solutions to a PDE \({\mathcal F}(w)=0\), which is a local diffeomorphism. In Floer theory, \({\mathcal F}=\partial_s +{\mathcal V}\) and \(\mathbb{S}\) is a strutum in a moduli space of broken trajectory or broken orbits. Hence it is a product of reduced moduli spaces \[ \mathbb{S}=\widehat {\mathcal M}_0\times\widehat {\mathcal M}\times\cdots\times\widehat{\mathcal M}_k,\quad\text{or}\quad \widehat{\mathcal M}_1\times\cdots\times\widehat{\mathcal M}_k/\mathbb{Z}/k \mathbb{Z}. \] A typical gluing theorem is proved by the following four steps: Step 1. For each gluing parameter \(\chi\), construct an approximate solution \(w_\chi\) of \({\mathcal F}\) which varies smoothly with \(\chi\), and show \({\mathcal F}(w_\chi)\) is sufficiently small estimating the pregluing map \(\chi\to w_\chi\) (error estimate). At deaths, nondegenerate elements of \({\mathcal P}^\Lambda\) which give local maxima of \(\Pi_\Lambda\), the projection from parametrized moduli spaces to \(\Lambda\), this step is carried out in \S2 mainly by using results in \S5 of Part I. Step 2. Let \({\mathfrak D}_w:E\to F\) be the linearization of \({\mathcal F}\) at \(w\) (deformation operator). It is shown \({\mathfrak D}_w=\partial_s+A(s)\) is a Fredholm operator having a right inverse bounded uniformly in \(\chi\) (Prop. 3.1.1). This structure is said to be the Kuranishi structure \((K\)-model). Precisely, a \(K\)-model \({\mathfrak D}\:K\to C|_B\) for a Fredholm operator \({\mathfrak D}\:E\to F\) is a triple \(K,C,B\), where \(K,C\) are finite-dimensional subspaces \(K\subset E\), \(C\subset F\), respectively, and \(B \subset E\) is a closed subspace such that \({\mathfrak D}|_B:B\to (B)\) is an isomorphism and there is a decomposition \[ E=K \oplus B,\quad F=C \oplus{\mathfrak D}(B). \] At deaths, these are discussed in \S3. Step 3. The gluing map is defined correspondingly for each gluing parameter \(\chi\) assigning \(w_\chi+P_\chi\eta_\chi\), \[ {\mathcal F}(w)={\mathcal F}(w_\chi)+ {\mathfrak D}_{w_\lambda}\xi+N_{w_\lambda}(\xi),\quad w=w_\chi+ \xi. \] It is a smooth map from the space of gluing parameters to the moduli space. Its properties are derived from a quadratic bound on the nonlinear part of \({\mathcal F}\) (\S4. cf. \S5 of Part I). At deaths, the gluing map is defined in \S4. Step 4. A gluing theorem is obtained showing the gluing map is a local diffeomorphism to a neighborhood of \(S\). This is proved for gluing at deaths in \S2--\S4 (Prop. 2.1) and gluing at births, nondegenerate elements of \(P^\lambda\) which are local minima of \(\Pi_\Lambda\), in \S5 (Prop. 5.1). Discussions at birth are similar to the case of deaths, but simpler than the case of deaths. Because we need only glue a single flow and the generalized cokernel (Def. 3.1.2) in this case is trivial. An outline of the proof of the gluing theorem, according to the above 4 steps, is sketched in \S1. The proof of statement 1 is given in \S6 by using the gluing theorem together with relevant \(K\)-models given in \S7.3.2 and 7.3.3 (Prop. 6.1.1). Verifying the bifurcation behavior at handle-slides, statement 3 is also proved in \S (Prop. 6.1.2). Analysis is done first discussing a simpler situation in which the nonequivariant perturbation may be obtained from a vector field on a finite-cyclic covering of \(M\) (\S6.2), then the general case is treated applying Fredholm theory and a transversality argument in Part I (\S6.3). Since the coherence of orientation is determined by linearized versions of the gluing theorems, discussions on orientations and sign (statement 2, Part I, \S4.3.7) are done in \S7, the last paragraph.