Cyclic symmetry and adic convergence in Lagrangian Floer theory (Q1958473)

Let \(L\) be a compact relatively spin Lagrangian submanifold in a symplectic manifold \((M,\omega)\), where \(M\) is either compact or convex at infinity. It was shown that \(H(L;\mathbb{Q})\otimes\Lambda_{0,\text{nov}}\) has a structure of a unital filtered \(A_\infty\)-algebra which is well defined up to isomorphism ([Theorem A: \textit{K. Fukaya}, \textit{Y.-G. Oh}, \textit{H. Ohta} and \textit{K. Ono}, CRM Proceedings and Lecture Notes 49, 201--227 (2009; Zbl 1187.53085)], hereafter referred to as [1]). Here, \(\Lambda_{0, \text{now}}\) is the universal Novikov ring ([1], cf. \S6, Def.~6.3). In this paper, this theorem is improved as follows: Theorem 1.1. \(H(L;\Lambda_{0,\text{nov}})= H(L;\mathbb{R})\otimes \Lambda_{0, \text{nov}}\) has a structure of a unital filtered cyclic \(A_\infty\)-algebra which is well defined up to isomorphism (Definition of cyclic filtered \(A_\infty\)-algebra is given in \S6, Def.~6.1). This theorem allows to improve an adic convergence condition of some infinite series of elements of \(H(L;\Lambda_{0,\text{nov}})\). Roughly speaking, a filtered \(A_\infty\)-structure defines a map \[ m_k: B_k(H(L;\Lambda_{0, \text{nov}}))\to H(L; \Lambda_{0,\text{nov}}), \] see [1], where \(B_k(V)= \overbrace{V\otimes\cdots\otimes V}^{k}\). \(m_k\) satisfies a kind of coboundary condition, and if the filtered \(A_\infty\)-structure is cyclic, then \[ \langle m_k(x_1,\dots, x_k),x_0\rangle= (-1)^*\langle m_k(x_0, x_1,\dots, x_{k-1}), x_k\rangle. \] Here, \(*=(\deg x_0+ 1)(\deg x_1+\cdots+ \deg x_k+ k)\), and \[ \langle u,v\rangle= (-1)^{\deg u\deg v+\deg u}\int_L u\wedge v. \] Let \(P_N(x)\) be \(\sum^N_{k=0} m_k(x,\dots, x)\). Then, if \(\lim_{N\to\infty}P_N(x)\) exists, it is denoted as \(m(e^x)\) and used to define a Maurer-Cartan equation and its solution. Previously, the existence of \(\lim_{N\to\infty}P_N(x)\) was proved assuming \(x\in\Lambda^+_{0, \text{nov}}\), where \(\Lambda^+_{0,\text{nov}}\) is a proper ideal of \(\Lambda_{0,\text{nov}}\) consisting of elements having nonzero degree in the filtration of \(\Lambda_{0,\text{nov}}\) (cf. Def.~6.3). This result is improved in this paper proving convergence of \(\lim_{N\to\infty}P_N(x)\) for any \(x\in\Lambda_{0,\text{nov}}\) (Th.~1.2). A module \({\mathcal M}(L)_\delta\) of a family of strict and unital cyclic filtered \(A_\infty\)-algebras is also constructed in Theorem 1.2. The author says \({\mathcal M}(L)_\delta\) is a chart of a certain rigid analytic space. Then, extending the construction of Th.~1.2, another Lagrangian submanifold \(L'\) is shown to give an object of coherent sheaves on the rigid analytic space (cf. [1], Conjecture U). According to the author, this is a step towards realizing the project to construct a homological mirror functor by using the family of Floer cohomology [cf. \textit{K. Fukaya}, J. Algebr. Geom. 11, No.~3, 393--512 (2002; Zbl 1002.14014); \textit{M. Kontsevich} and \textit{Y. Soibelamn}, Symplectic geometry and mirror symmetry. Proceedings of the 4th KIAS annual international conference, Seoul, South Korea, 2000. Singapore: World Scientific. 203--263 (2001; Zbl 1072.14046)]. The author also says that another application of Theorems 1.1 and 1.2 defines a numerical invariant of a Lagrangian submanifold in a Calabi-Yau 3-fold, which is a rational homology sphere, by counting the number of pseudoholomorphic discs with appropriate wedge. This is invariant under perturbations and other choices, but depends on the almost complex structure and its rigorous definition is essential to find a perturbation that is compatible with forgetful maps [\textit{K. Fukaya}, Counting pseudo-holomorphic discs in Calabi-Yau 3-fold. \url{arXiv:0908.0148v1}]. The definition of the universal Novikov ring is done by taking a subset \(G\subset\mathbb{R}_{\geq 0}\times 2\mathbb{Z}\). Theorem 1.1 is proved taking \(G\) to be the submonoid of \(\mathbb{R}_{\geq 0}\times 2\mathbb{Z}\) generated by the set \[ \{(\beta\cap \omega, \mu(\beta))\mid \beta\in H_2(M, L;\mathbb{Z}),{\mathcal M}_0(\beta)\neq \emptyset\}. \] Here \(\mu(\beta)\) is the Maslov index, \({\mathcal M}_0(\beta)={\mathcal M}_{0,0}(\beta)\), where \({\mathcal M}_{\ell, k}(\beta)\) is the moduli space of bordered \(J\)-holomorphic curves of genus 0 with \(\ell\) interior marked points and \(k\) boundary marked points, one boundary component, and of homology class \(\beta\). Therefore, a study of \({\mathcal M}_{\ell,k}\) is necessary including the case \(k= 0\) (Previously, \({\mathcal M}_{\ell,k}\) was studied assuming \(k\geq 1\)). The \({\mathcal M}_{\ell, k}\) have singularities. To control the singularities, a Kuranishi structure, which is reviewed in \S2, is used, showing the existence of a Kuranishi structure compatible with the forgetful map \[ \mathfrak{forget}_{k+1,1}:{\mathcal M}_{\ell,k+ 1}(\beta)\to{\mathcal M}_{\ell, 1}(\beta) \] (Cor.3.1; (1)). Then, this result is extended to a system of continuous families of multisections on \({\mathcal M}_{\ell,k+1}(\beta)\) (Cor.~5.1(2)). A review on continuous families of multisections is given in \S4. By using these results, Th.~1.1 is proved modulo \(T^E\) (Th. 7.1). Here, \(E\) is a filtration of \(\Lambda_{0,\text{nov}}\) and modulo \(T^E\) means a truncation by \(E\). To deduce Th.~1.1 from Th.~7.1, precise studies of the homological algebra of a cyclic filtered \(A_\infty\)-algebra is needed. They are done in \S8--\S11. (The results are summarized in \S8. \S9--\S10 are devoted to an algebraic study. Geometric realizations are done in \S11.) After these preparations, Th.~1.1 is proved in \S12, except the uniqueness part, which is proved in \S14, the last section. The most essential part of Th.~1.2 is the change of variables \(x_i\to y_i= e^{e_i}\). This becomes possible after constructing a Kuranishi structure and its perturbation in a way compatible with the forgetful map. Then, \(m_{k,\beta}\), the contribution of \(\beta\) to \(m_k\) satisfies useful formulas such as \[ m_{3,\beta}(e_i, e_i, x)+ m_{3,\beta}(e_i, x,e_i)+ m_{3,\beta}(x,e_i, e_i)= {1\over 2!}(\beta\cap e_i)^2 m_{1,\beta}(x). \] Here, the \(e_i\)'s are the basis of \(H(L;\mathbb{R})\) \((x=\sum x_i e_i)\). These formulas justify the notation \(\lim_{N\to\infty} P_N(x)= e^x\). Th.1.1 is proved by using these facts and constructions in \S13.