Calabi-Yau/Landau-Ginzburg Correspondence for Weil-Peterson Metrics and $tt^*$ Structures
Extracted source text · arXiv 2205.05791
\documentclass[11pt]{amsart}
\allowdisplaybreaks[4]
\linespread{1.2}
\usepackage{amssymb}
\usepackage{amsfonts}
\usepackage{graphicx}
\usepackage{epstopdf}
\usepackage{dcolumn}
\usepackage{amsmath}
\usepackage{enumerate}
\usepackage{latexsym,bm}
\usepackage{tikz-cd}
\usepackage{slashed}
\usepackage{float}
\usepackage{cite}
\usepackage{CJK}
\usepackage[all,cmtip]{xy}
\usepackage{appendix}
\usepackage[colorlinks,
linkcolor=black,
citecolor=red
]{hyperref}
\setlength{\textwidth}{6.7in}
\setlength{\hoffset}{-0.90in} %{-0.65in}
\setlength{\voffset}{-0.90in} %{-0.8in}
\setlength{\textheight}{9.2in}
%%%%%%%%%%%%%%%%%%%%%% Defintions %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\newcommand{\op}{\operatorname}
\newcommand{\C}{\mathbb{C}}
\newcommand{\N}{\mathbb{N}}
\newcommand{\R}{\mathbb{R}}
\newcommand{\Q}{\mathcal{Q}}
\newcommand{\Z}{\mathbb{Z}}
\newcommand{\T}{\mathcal{T}}
\newcommand{\V}{\mathcal{V}}
%\newcommand{\I}{\mathcal{I}}
\renewcommand{\H}{\mathbb{H}}
\newcommand{\Etau}{{\text{E}_\tau}}
\newcommand{\F}{\mathbf{F}}
\newcommand{\G}{\mathbf{G}}
\newcommand{\eps}{\epsilon}
\newcommand{\g}{\mathbf{g}}
\newcommand{\im}{\op{im}}
\newcommand{\h}{\mathbf{h}}
\newcommand{\J}{\mathscr{J}}
\newcommand{\CC}{\mathscr{C}}
\newcommand{\K}{\mathcal{K}}
\newcommand{\Gmax}[1]{G_{#1}}
\newcommand{\AW}{E}
%%%%%%%%%%%%%%%%%%%%%% Functions %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\providecommand{\abs}[1]{\left\lvert#1\right\rvert}
\providecommand{\norm}[1]{\left\lVert#1\right\rVert}
\newcommand{\abracket}[1]{\left\langle#1\right\rangle}
\newcommand{\bbracket}[1]{\left[#1\right]}
\newcommand{\fbracket}[1]{\left\{#1\right\}}
\newcommand{\bracket}[1]{\left(#1\right)}
\newcommand{\aabracket}[1]{\left\llangle#1\right\rrangle}
\newcommand{\ket}[1]{|#1\rangle}
\newcommand{\bra}[1]{\langle#1|}
\newcommand{\ora}[1]{\overrightarrow#1}
\newcommand{\llb}{\llbracket}
\newcommand{\rrb}{\rrbracket}
\newcommand{\llp}{(\!(}
\newcommand{\rrp}{)\!)}
\def\tgm{\gamma^+}
\providecommand{\from}{\leftarrow}
\newcommand{\bl}{\textbf}
\newcommand{\mbf}{\mathbf}
\newcommand{\mbb}{\mathbb}
\newcommand{\mf}{\mathfrak}
\newcommand{\mc}{\mathcal}
\newcommand{\mr}{\mathrm}
%\newcommand{\bb}{\mathbb}
\newcommand{\cinfty}{C^{\infty}}
\newcommand{\p}{\partial}
\newcommand{\prm}{\prime}
\newcommand{\hotimes}{\hat\otimes}
\newcommand{\BV}{Batalin--Vilkovisky }
\newcommand{\CE}{Chevalley--Eilenberg }
\newcommand{\suml}{\sum\limits}
\newcommand{\prodl}{\prod\limits}
\newcommand{\into}{\hookrightarrow}
\newcommand{\Ol}{\mathcal O_{\mr{loc}}}
\newcommand{\mD}{{\mathcal D}}
\newcommand{\iso}{\cong}
\newcommand{\dpa}[1]{{\pa\over \pa #1}}
\newcommand{\PP}{\mathscr{P}}
\newcommand{\Kahler}{K\"{a}hler }
\newcommand{\0}{\mathbf{0}}
\newcommand{\wt}{\widetilde}
\newcommand{\wh}{\widehat}
\newcommand{\ol}{\overline}
\newcommand{\bb}{\mathbf{b}}
%\newcommand{\B}{\mathcal{B}}
%\newcommand{\HH}{\mathcal{H}}
\DeclareMathOperator{\HH}{H}
\newcommand{\W}{\mathbf W}
\renewcommand{\Im}{\op{Im}}
\renewcommand{\Re}{\op{Re}}
\renewcommand{\P}{\mathcal{P}}
\newcommand{\B}{\mathcal{B}}
%%%%%%%%%%%%%%%%%%%%%% Math Operators %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\DeclareMathOperator{\mHom}{\mathcal{H}om}
\DeclareMathOperator{\Aut}{Aut}
\DeclareMathOperator{\End}{End}
\DeclareMathOperator{\Supp}{Supp}
\DeclareMathOperator{\Sym}{Sym}
\DeclareMathOperator{\Hom}{Hom}
\DeclareMathOperator{\Spec}{Spec}
\DeclareMathOperator{\Deg}{Deg}
\DeclareMathOperator{\Diff}{Diff}
\DeclareMathOperator{\Lie}{Lie}
\DeclareMathOperator{\Vol}{Vol}
\DeclareMathOperator{\Tr}{Tr}
\DeclareMathOperator{\Or}{Or}
\DeclareMathOperator{\Ker}{Ker}
\DeclareMathOperator{\Mat}{Mat}
\DeclareMathOperator{\Ob}{Ob}
\DeclareMathOperator{\Isom}{Isom}
\DeclareMathOperator{\PV}{PV}
\DeclareMathOperator{\Der}{Der}
\DeclareMathOperator{\GM}{GM}
\DeclareMathOperator{\Eu}{Eu}
\DeclareMathOperator{\prim}{prim}
\DeclareMathOperator{\Res}{Res}
\DeclareMathOperator{\Gr}{Gr}
\DeclareMathOperator{\Jac}{Jac}
\DeclareMathOperator{\ord}{ord}
\DeclareMathOperator{\res}{res}
\DeclareMathOperator{\Id}{Id}
\DeclareMathOperator{\FM}{FM}
\DeclareMathOperator{\Graphs}{Graphs}
\DeclareMathOperator{\Dens}{Dens}
\DeclareMathOperator{\Poin}{Poin}
\DeclareMathOperator{\Conf}{Conf}
\DeclareMathOperator{\w}{wt}
\DeclareMathOperator{\Span}{Span}
\DeclareMathOperator{\Def}{Def}
\DeclareMathOperator{\GG}{G_H}
\DeclareMathOperator{\SF}{SF}
\newcommand{\A}{\mathcal A}
\newcommand{\M}{\mathcal M}
\newcommand{\E}{\mathbb E}
\renewcommand{\S}{\mathcal S}
\renewcommand{\L}{\mathcal L}
\renewcommand{\H}{\mathcal H}
\newcommand{\la}{\langle}
\newcommand{\ra}{\rangle}
\def\p{\partial}
\def\bap{\bar{\partial}}
\def\I{\mathcal{I}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%% Theorems type enviroments %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\theoremstyle{plain}
\newtheorem{thm}{Theorem}[section]
\newtheorem{thm-defn}{Theorem/Definition}[section]
%\newtheorem{lem}{Lemma}[section]
\newtheorem{lem}[thm]{Lemma}
%\newtheorem{lem-defn}{Lemma/Definition}[section]
\newtheorem{lem-defn}[thm]{Lemma/Definition}
\newtheorem{prop-defn}[thm]{Proposition/Definition}
\newcommand{\be}{\begin{equation}}
\newcommand{\ee}{\end{equation}}
%\newtheorem{prop}{Proposition}[section]
\newtheorem{prop}[thm]{Proposition}
%\newtheorem{cor}{Corollary}[section]
\newtheorem{cor}[thm]{Corollary}
\newtheorem{conj}{Conjecture}%[section]
\newtheorem{prob}{Problem}%[section]
\newtheorem*{claim}{Claim}
%\theoremstyle{definition}
%\newtheorem{defn}{Definition}[section]
\newtheorem{defn}[thm]{Definition}
%\newtheorem{notn}{Notation}[section]
\newtheorem{notn}[thm]{Notation}
%\newtheorem{eg}{Example}[section]
\newtheorem{eg}[thm]{Example}
\theoremstyle{remark}
%\newtheorem{rem}{Remark}[section]
\newtheorem{rem}[thm]{Remark}
\newtheorem{warn}{Warning}%[section]
\newcommand{\xinxing}[1]{(\textcolor{orange}{Xinxing: #1})}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Allow display breaks within equations %%%%%%%%%%%%%%%%%%%%
\allowdisplaybreaks[4] %%%%%%%%%%% choose 1-4, where 4 is the strongest desire to breack
\def\curved{\tikz[baseline=.1ex]{
\draw[ ->] (0,0.32) arc (30:330:0.5);}
}
\def\anticurved{\tikz[baseline=.1ex]{
\draw[ ->] (0,0.32) arc (150:-150:0.5);}
}
\begin{document}
\title{Calabi-Yau/Landau-Ginzburg Correspondence for Weil-Peterson Metrics and $tt^*$ Structures}
\author{Xinxing Tang and Junrong Yan}
\address{
X. Tang: Yanqi Lake Beijing Institute of Mathematical Sciences And Applications, Beijing, China;
}
\email{tangxinxing@bimsa.cn}
\address{
J. Yan: Department of Mathematics, University of California, Santa Barbara, US;
}
\email{j\_yan@math.ucsb.edu}
\date{}
\maketitle
\begin{abstract} The goal of this paper is to establish the Calabi-Yau/Landau-Ginzburg (CY/LG) correspondence for the $tt^*$ geometry structure, which is thought to hold all genus $0$ information about B-models. More explicitly, given a non-degenerate homogeneous polynomial $f\in\C[z_1,\ldots,z_n]$ of degree $n$, one can investigate the Landau-Ginzburg B-model, which concerns the deformation of singularities. Its zero set, on the other hand, defines a Calabi-Yau hypersurface $X_f$ in $\mathbb{P}^{n-1}$, whereas the Calabi-Yau B-model is concerned with the deformation of the complex structure on $X_f$. Both LG B-model and CY B-model's genus 0 information can be captured by the $tt^*$ geometry structure. In this paper, we construct a map between the $tt^*$ structures on CY and LG's sides, and by a careful study of the period integrals, we build the isomorphism of $tt^*$ structures between the CY B-model and the LG B-models.\end{abstract}
\tableofcontents
\section{Introduction}
\subsection{Overview}
%Gromov's pioneering work \cite{Grom1985} on pseudoholomorphic curves revolutionized the field of symplectic geometry. The pseudoholomorphic curves are solutions to Cauchy-Riemann equation
%$$\bar{\partial}u=0,$$
%where $u$ is a map from a Riemann surface to a K\"{a}hler manifold. The path integral in the A-twisted sigma models for a K\"{a}hler manifold reduces to the integral over the moduli space. The integral of appropriate classes over moduli space defines the Gromov-Witten invariants, which is known as the A-model Gromov-Witten Theory.
In the 1990s, physicists discovered mirror symmetry, a relationship between the symplectic geometry (A-model) of a Calabi-Yau manifold $X$ and the complex geometry (B-model) of its mirror $X^{\vee}$. Since its formulation over three decades ago \cite{Candelas1991APO,Greene1990DualityI}, the mirror symmetry conjecture has had a profound influence on mathematics. The A-model on a Calabi-Yau manifold $X$ is the famous Gromov–Witten theory of $X$. The genus 0 theory of Calabi-Yau B-model is related to the variation of Hodge structure, whereas the higher genus B-model theory is known as BCOV theory \cite{Bershadsky1994KodairaSpencerTO}, \cite{CostLi}.
Simultaneously, physicists extended the above discussion to singularity theory, which became known as the Landau-Ginzburg model. A Landau-Ginzburg model is defined on the pair $(X,f)$, where $X$ is a complete noncompact K\"{a}hler manifold and $f$ is a holomorphic function. For the A-side, the Landau-Ginzburg A-model is constructed by Fan-Jarvis-Ruan \cite{FJR2013} following Witten's proposal \cite{Witten1993}, which is known as FJRW-theory. Moreover, Chang-Li-Li \cite{li2013primitive} recently formulated FJRW-theory in an algebraic geometric formulation. On the B-side, Saito's theories of primitive form \cite{1981Primitive} and the higher residue pairing \cite{1983higher} generate the structure of a Frobenius manifold on the singularity's universal deformation space \cite{SaitoFM}, which yield the genus-0 theory of Landau-Ginzburg B-model. Also, some Hodge theoretical aspect in LG B-model, called the $tt^*$-geometry, is discovered by Cecotti-Vafa \cite{cecotti1991topological}, whose integrability structure is studied by Dubrovin \cite{Dub1993tt}. Later, Hertling \cite{Hert2003tt} carefully researched the $tt^*$ geometric structure and organized a variety of known structures into the so-called TERP structure. Additionally, Fan \cite{fan2011schr} introduced an analytic approach in the spirit of $N=2$ geometry by studying the spectral theory of a twisted Laplacian operator. Besides, Li-Wen \cite{li2019l2} used the $L^2$ Hodge theory to give a Frobenius manifold structure for the case $f$ with compact critical locus.
Furthermore, motivated by the Virasoro equations and localization calculation in the A-model at a higher genus, Givental \cite{Giv2001b} gave a remarkable formula for the partition function in the semi-simple case.
The Calabi-Yau/Landau-Ginzburg correspondence connects nonlinear sigma models on Calabi-Yau manifolds with Landau-Ginzburg models. It turns out that CY/LG correspondence and mirror symmetry have served as guidelines in the study of many branched of mathematics (see the following diagram):
%In mathematical physics, mirror symmetry is a relationship between two theories on Calabi-Yau (CY) manifolds, which are called A-model and B-model. Since it was formulated more than thirty years ago \cite{Candelas1991APO,Greene1990DualityI}, the mirror symmetry conjecture has had a profound influence on mathematics.
%Moreover, there is another model, which is closely related to Sigma models on Calabi-Yau manifolds, called Landau-Ginzburg (LG) model. An LG model consists of a noncompact K\"ahler manifold with a holomorphic function (superpotential) on it. People then realized that, apart from the equivalence between A theory and B theory of the Calabi-Yau mirror pair, there are relations between the A and B theories of the CY model and the LG model (Calabi-Yau/Landau-Ginzburg correspondence), which can be briefly described by the following diagram:
%\[\begin{tikzcd}
%\mbox{LG A theory} \arrow[rrrr, "\mbox{mirror map}"] \arrow[ddd, "CY/LG"] & & & & \mbox{LG B theory} \arrow[ddd, "CY/LG"] \\
% & & & & \\
% & & & & \\
%\mbox{CY A theory} \arrow[rrrr, "\mbox{mirror map}"] & & & & \mbox{CY B theory}
%\end{tikzcd}\]
\vskip 0.2cm
\begin{center}
\sffamily
\footnotesize
\begin{tikzpicture}[auto,
%decision/.style={diamond, draw=black, thick, fill=white,
%text width=8em, text badly centered,
%inner sep=1pt, font=\sffamily\small},
block_center/.style ={rectangle, draw=black, thick, fill=white,
text width=12em, text centered,
minimum height=3em,inner sep=6pt,outer sep=4pt},
block_left/.style ={rectangle, draw=black, thick, fill=white,
text width=12em, text centered, minimum height=3em, inner sep=6pt,outer sep=4pt}]
% outlining the flowchart using the PGF/TikZ matrix funtion
\matrix [column sep=25mm,row sep=12mm] {
% enrollment - row 1
\node [block_left] (CYA) {Calabi-Yau A-model};
& \node [block_center] (LGA) {Landau-Ginzburg A-model};\\
% enrollment - row 2
% enrollment - row 3
\node [block_left] (CYB) {Calabi-Yau B-model};
& \node [block_center] (LGB) {Landau-Ginzburg B-model};\\
};% end matrix
% connecting nodes with paths
\path[draw=black,latex-latex,thick] (CYA) -- node[above,font=\tiny,outer sep=2pt]{CY/LG}(LGA);
\path[draw=black,latex-latex,thick] (LGA) -- node[left, font=\tiny, align=left]{Mirror Symmetry}(LGB);
%\path[draw=black,latex-,thick] (mc2) -- node[right,font=\tiny,align=left] {Topological descendant}(cme);
\path[draw=black,latex-latex,thick] (CYB) -- node[below,font=\tiny,outer sep=2pt] {CY/LG} (LGB);
\path[draw=black,latex-latex,thick] (CYA) -- node[right,font=\tiny,align=left] {Mirror Symmetry} (CYB);
\end{tikzpicture}
\end{center}
%It is well known that the CY B-model concerns the deformation of the complex structure. The genus 0 theory is equivalent to the variation of Hodge structure. While the LG B-model concerns the deformation of singularities. Moreover, its genus 0 theory is given by Saito's theory of primitive forms and higher residue pairing(c.f. \cite{Saito1982PrimitiveFF,Saito}), originated from Saito's study of period integrals over vanishing cycles associated to an isolated singularity.
The CY/LG correspondence for the $A$ model is partially done by Chiodo-Iritani-Ruan \cite{chiodo2014landau}. The question of whether the CY/LG correspondence holds for the B-model is both intriguing and difficult. To address this question for the genus 0 cases, we show CY/LG correspondence for $tt^*$ structures. In fact, Carlson-Griffiths \cite{carlson1980infinitesimal} compared the cup product and symplectic pairing on the CY and LG sides. Also, Cecotti \cite{CECOTTI1991N} investigated the $tt^*$ geometry structure on both sides from a physical standpoint. %{\color{red} is it OK to add this sentense? where some definitions and proofs are not sufficiently explicit}.
Besides, Fan-Lan-Yang \cite{Fan2020LGCYCB} partially proved that the two $tt^*$ structures are isomorphic via the CY/LG correspondence except for the real structures. Here we use different methods than \cite{Fan2020LGCYCB} to show the full CY/LG correspondence: we introduce two $U(1)$ actions (see Section \ref{2.4}), which act as certain bi-grading for LG B-models. We also establish CY/LG correspondence for Weil-Peterson type metrics using these two $U(1)$ actions. Furthermore, the Agmon estimate derived in \cite{DY2020cohomology} plays an essential role in our method (See also Section \ref{witagm}).
\subsection{$tt^*$ geometry structure on Calabi-Yau manifolds}
Let $X$ be a compact Calabi-Yau $(n-2)$-fold and $\mathcal{M}$ be the moduli stack of the complex structure of $X$ with the universal family
$\pi: \mathcal{X}\rightarrow\mathcal{M}$.
By Tian-Todorov Theorem, the universal deformation space $\mathcal{M}$ of $X$ is smooth of dimension $\dim H^1(X,T_{X})$.
Let $H_{\prim}^{n-2}$ be the Hodge bundle of primitive forms over $\mathcal{M}$, i.e., for any $[X]\in\mathcal{M}$, the fiber $H_{\prim}^{n-2}\big|_{[X]}$ is the space of primitive $(n-2)$-forms on $X$. Then we have the Hodge decomposition and Hodge filtration:
\begin{align*}
&H^{n-2}_{\prim}=H^{n-2,0}_{\prim}\oplus H^{n-3,1}_{\prim}\oplus\cdots\oplus H^{0,n-2}_{\prim},\quad \overline{H^{n-2-p,p}_{\prim}}=H^{p,n-2-p}_{\prim};\\
&\F^{n-2}H^{n-2}_{\prim}\subset\cdots\subset \F^0H^{n-2}_{\prim}=H^{n-2}_{\prim},\quad \F^kH^{n-2}_{\prim}=\oplus_{p=0}^{n-2-k}H_{\prim}^{n-2-p,p}.
\end{align*}
The theory of variation of Hodge structure tells us such a bundle is equipped with
\begin{itemize}
\item a flat Gauss-Manin connection $\nabla^{\GM}$ which satisfies the Griffiths transversality conditions:
$$\nabla^{\GM}_{v}\F^p\subset \F^{p-1}, \quad v\in \mathcal{T}_{\mathcal{M}}.$$
\item a pairing $g: H^{n-2}_{\prim}\otimes H^{n-2}_{\prim}\rightarrow C^{\infty}(\mathcal{M})$
$$g(\alpha,\beta)=i^{2p-(n-2)}\int_{X}\alpha\wedge\bar{\beta},\quad \alpha,\beta\in H^{n-2-p,p}_{\prim}$$
gives a positive Hermitian metric on $H^{n-2}_{\prim}$.
%$$\nabla^{\GM}(g)=0.$${\color{red} $\nabla^{GM}$ preserve the symmetric pairing but may not preserve hermition metric $g.$}
We call it the $tt^*$ metric in the CY B-model.
\end{itemize}
In particular, $H^{n-2,0}_{\prim}$ forms a line bundle over $\mathcal{M}$. Let $\Omega_u$ be a local holomorphic section of $H^{n-2,0}_{\prim}$, which, fiberwisely, is a holomorphic volume form on $X$. The K\"{a}hler potential $K(u,\bar{u})$ is defined via
$$e^{-K(u,\bar{u})}=i^{-n-2}\int_{X}\Omega_u\wedge\overline{\Omega_u}.$$
Moreover, the Weil-Peterson metric is defined by
$$G^{CY}=\partial\bar{\partial}K(u,\bar{u}).$$
It can be checked easily that $G^{CY}$ is independent of the choice of the local holomorphic section of $H^{n-2,0}_{\prim}$.
%By Griffiths transversality, one can {\color{red} wirite down the formula directly......}
The variation of Hodge structure is related to the so-called $tt^*$ geometry structure:
\def\l{\lambda}
\begin{defn}[$tt^*$ geometry] A $tt^*$ geometry structure $(K\rightarrow M, \kappa, g, D, \bar{D}, C, \bar{C})$ consists of the following data
\begin{itemize}
\item $M$ is a complex manifold, $K\rightarrow M$ is a smooth vector bundle,
\item a complex anti-linear involution $\kappa: K\rightarrow K$, i.e. $\kappa^2=\Id$, $\kappa(\l\alpha)=\bar{\l}\kappa(\alpha)$, $\forall~ \l\in\C$,
\item a Hermitian metric $g$ on $K$,
\item a one parametric family of flat connections $\nabla^z=D+\bar{D}+\frac{1}{z}C+z\bar{C}$, where $z\in\C^*$, $D+\bar{D}$ is the Chern connection of $g$, $C$, $\bar{C}$ are the $C^{\infty}(M)$-linear map
$$C:C^{\infty}(K)\rightarrow C^{\infty}(K)\otimes_{C^{\infty}(M)}\mathcal{A}^{1,0}(M),\quad \bar{C}:C^{\infty}(K)\rightarrow C^{\infty}(K)\otimes_{C^{\infty}(M)}\mathcal{A}^{0,1}(M),$$
\end{itemize}
satisfying
\begin{enumerate}
\item $g$ is real with respect to $\kappa: g(\kappa(u),\kappa(v))=\overline{g(u,v)}$,
\item $D+\bar{D}$ respects the Hermitian metric $g$,
\item $(D+\bar{D})(\kappa)=0$, $\bar{C}=\kappa\circ C\circ\kappa$,
\item $\bar{C}$ is the adjoint of $C$ with respect to $g$, i.e. $g(C_Xu,v)=g(u,\bar{C}_{\bar{X}}v)$, $X\in\mathcal{T}_M$.
\end{enumerate}
Then we say such structure $(K\rightarrow M, \kappa, g, D, C, \bar{C})$ is a $tt^*$ geometry structure.
\end{defn}
%{\color{red}We should give the definition of $tt^*$ str and the definition of morphism of $tt^*$ str here. Then we could talk about the $tt^*$ str of CY.}
Now consider the $tt^*$ geometry structure induced by the variation of the Hodge structure on the Calabi-Yau B model. To begin, the $tt^*$ connection is the Chern connection of the $tt^*$ metric; denote the $(1,0)$ and $(0,1)$ parts of the $tt^*$ connection by $D^{CY}$ and $\bar{D}^{CY}$, respectively.
Following that, let $$C^{CY}:T_{\mathcal{M}}\longrightarrow \bigoplus_p\Hom(H^{n-2-p,p}_{\prim},H^{n-2-p-1,p+1}_{\prim})\subset\End(H^{n-2}_{\prim})
$$
be the Kodaira-Spencer map and $\bar{C}^{CY}$ be its complex conjugate.
Then the Griffiths transversality implies that
$$\nabla^{\GM,(1,0)}=D^{CY}+C^{CY},\quad \nabla^{\GM,(0,1)}=\bar{D}^{CY}+\bar{C}^{CY}.$$
The flatness of $\nabla^{\GM}$ is equivalent to the so-called $tt^*$ equations.
\begin{prop}[$tt^*$ equations] The operators $D^{CY},\bar{D}^{CY}$, $C^{CY}, \bar{C}^{CY}$ satisfy the following equations
\begin{enumerate}
\item $[C^{CY},C^{CY}]=0, \quad [\bar{C}^{CY}, \bar{C}^{CY}]=0,\quad [D^{CY},\bar{C}^{CY}]=[\bar{D}^{CY},C^{CY}]=0;$
\item $[D^{CY},C^{CY}]=[D^{CY},C^{CY}], \quad [\bar{D}^{CY},\bar{C}^{CY}]=[\bar{D}^{CY},\bar{C}^{CY}]$;
\item $[D^{CY}, D^{CY}]=0,\quad [\bar{D}^{CY},\bar{D}^{CY}]=0,\quad [D^{CY},\bar{D}^{CY}]=-[C^{CY},\bar{C}^{CY}].$
\end{enumerate}
\end{prop}
Combining with the complex conjugation and the $tt^*$ metric $g$, the above data define a $tt^*$ structure on CY's side.
Let $u=(u_1,...,u_s)$ be local coordinates of the moduli $\mathcal{M}$, and $C^{CY}_i:=C^{CY}(\frac{\p}{\p u_i})$, $G^{CY}_{i\bar{j}}=\p_{u_i}\bar{\p}_{\bar{u}_j}\log K(u,\bar{u}) $.
By Griffiths transversality, one can show that $G_{i\bar{j}}$ is related to the $tt^*$ metric as follows:
$$G_{i\bar{j}}^{CY}=\frac{g\left(C_i^{CY}u,C_j^{CY}u\right)}{g(u,u)}.$$
%{\color{red} Say a word about real str, and the above data define $tt^*$ str on CY's side.}
%\subsection{Calabi-Yau/Landau-Ginzburg Correspondence: the Mixed Hodge Structure}
\subsection{Calabi-Yau/Landau-Ginzburg correspondence}
A Landau-Ginzburg model is defined by the pair $(X, f)$, where $X$ denotes a noncompact complete K\"ahler manifold and $f$ denotes a holomorphic function on $X$. The corresponding B-model is related to the deformation theory of singularities.
It is well-known that the Landau-Ginzburg (LG) B-model is closely
related to Sigma B-models on Calabi-Yau manifolds. For instance, if we consider the LG model $(\mathbb{C}^5, f(\textbf{z})=z_1^5+\cdots+z_5^5)$, regarding $[z_1,\ldots,z_5]$ as the homogeneous coordinate on $\mathbb{P}^4$, then $f=0$ defines a hypersurface $X_f$ (hence is a Calabi-Yau 3-fold) in $\mathbb{P}^4$. In particular, in the physical literature, the central charge $\hat{c}_f$ (defined below) of $f$ is given by $5-2=3$, which is equal to the dimension of the hypersurface. In fact, more quantum information of $X_f$ can be read form $(\mathbb{C}^5, f)$. Furthermore, consider the deformation
$$F=f+u\prod_{i=1}^5z_i, \quad u\in\mathbb{C}.$$
For sufficiently small $|u|$, $F$ determines a family of Calabi-Yau hypersurfaces $X_{F(\cdot,u)}$ parametrized by $u$, i.e., a deformation of $f$ induces a deformation of complex structure of $X_f$. In the physical literature, this phenomenon is referred to as the Calabi-Yau/Landau-Ginzburg correspondence. Before we continue, let us briefly review some well-known results.
%The Landau-Ginzburg B-model at genus-zero is related to Saito's theory of primitive form and higher residue [?], [?], which induces the structure of a Frobenius manifold on the universal deformation space of the singularity [?]. There is also a Hodge theoretical aspect (i.e. $tt^*$ geometry structure) in Landau-Ginzburg B-model which is discovered by Cecotti-Vafa [?], whose integrability structure is studied by Dubrovin [?]. Later, C. Hertling [?] studied the $tt^*$ geometric structure systematically and organized various known structures into the so-called TERP structure. H. Fan presented an analytic approach by considering the spectrum theory of a twisted Laplacian operator in the spirit of $N=2$ geometry. We will recall such a construction in the next section.
%It is natural to analyze the relation of the two $tt^*$ geometry structures arising from the Calabi-Yau B-model and Landau-Ginzburg B-model.
\subsubsection{State spaces and the residue map}
%\textcolor{blue}{More words here.}
Consider the pair $(\C^n,f)$, where $f:\C^n\rightarrow \C$ is a quasi-homogeneous polynomial with isolated singularity at the origin, i.e. there exist $q_1,\ldots,q_n\in\mathbb{Q}$ such that for any $\lambda\in\C^*$,
$$f(\lambda^{q_1}z_1,\ldots, \lambda^{q_n}z_n)=\lambda f(z_1,...,z_n).$$
Each $q_i$ is referred to as the weight of $z_i$, denoted by the expression $\w(z_i)=q_i$. The system of equations $\partial_i f=0$ has only one solution: zero (with multiplicity).
It is known that, as a vector space, the state space of Landau-Ginzburg model is isomorphic to the Jacobi ring $\mathrm{Jac}(f)$ of $f$, where
$$\Jac(f)=C[z_1,\ldots,z_n]/\langle\p_1f,\ldots,\p_nf\rangle.$$
For simplicity, now assume that $f$ is homogeneous of degree $n,$ then $$X_f:=\{[z_1,...,z_n]\in \mathbb{P}^n:f(z_1,...,z_n)=0\}$$
is a Calabi-Yau $(n-2)$-fold. It is well known that the state space of CY B-model on $X_f$ is give by $$H^{n-2}(X_f).$$ It follows from Proposition \ref{res} that
the (small) state spaces of LG B-model and CY B-model are related via the residue map.
More explicitly, let $$\Omega=\sum_{i=1}^n(-1)^{i+1}z_idz_1\wedge\cdots\wedge\widehat{dz_i}\wedge\cdots\wedge dz_n,$$
be a nonzero section of $\Omega_{\mathbb{P}^{n-1}}\cong\mathcal{O}_{\mathbb{P}^{n-1}}(-n)$,
and assume that $\phi$ is a homogeneous polynomial of degree $kn-n$, then a rational $(n-1)$-form with polar locus $X_f$ can be constructed as:
$$\Omega_{\phi}=\frac{\phi\Omega}{f^k}\in \Omega^{n-1}(\mathbb{P}^{n-1}-X_f).
$$%H^0(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}^{n-1}(kX_f))$$
%defines a rational $(n-1)$-form with polar locus $X_f$. %The polynomial in the numerator is called an
There is a well known pole order filtration on the space of rational differential $(n-1)$-forms,
$$\F_{\ord}^1\subset\F_{\ord}^2\subset\cdots\subset\F_{\ord}^{n-1},$$
where $\F_{\ord}^k$ is the subgroup with a pole of order $\leq k$ along $X_f$.
The residue map is defined to be
\begin{align*}
\res:H^{n-1}(\mathbb{P}^{n-1}-X_f)&\rightarrow H^{n-2}(X_f,\C)\\
\frac{1}{2\pi i}\int_{\tau(\gamma)}\frac{\phi\Omega}{f^k}&=\int_{\gamma}\res\left(\frac{\phi\Omega}{f^k}\right)
\end{align*}
where $\tau:H_{n-2}(X_f)\rightarrow H_{n-1}(\mathbb{P}^{n-1}-X_f)$ is the Leray coboundary map, i.e. $\tau(\gamma)$ is a tube neighborhood of $\gamma$.
\begin{prop}[\cite{carlson1980infinitesimal}]\label{res} The residue map satisfies the following properties:
\begin{enumerate}
\item The image of the residue map is the $(n-2)$-th primitive cohomology of $X_f$;
\item The residue map takes the pole order filtration on the space of rational differentials to the Hodge filtration on $H^{n-2}(X_f;\C)_{\prim}$, i.e., if $\phi$ is a homogeneous polynomial of degree $n(k-1),$
$$\res \left(\frac{\phi\Omega}{f^k}\right)\in\F^{n-k-1}H^{n-2}(X_f;\C)_{\prim}.$$
%{\color{red}where is k in the left hand side?}
Moreover, the residue of a form $\Omega_{\phi}=\frac{\phi\Omega}{f^k}$ has Hodge level $n-k$ %\textcolor{blue}{I think it's right, only need to check that the case $k=1$}
if and only if $\phi$ lies in the Jacobian ideal $\langle\p_1f,\cdots,\p_nf\rangle$.
\end{enumerate}
\end{prop}
\subsubsection{Cup product and the pairing}
The Griffiths-Carlson theory
\cite{carlson1980infinitesimal} provides the CY/LG correspondence for the cup product and the pairing:
\begin{prop}[\cite{carlson1980infinitesimal}]\label{carlson} Let $\phi$ and ${\psi}$ be homogeneous polynomials of degree $n(a-1)$ and $n(b-1)$ respectively, then
\begin{enumerate}
\item $\res\Omega_{\phi}\wedge\res\Omega_{\psi}=0$ if and only if $\phi\psi$ lies in the Jacobian ideal $\langle\p_1f,\cdots,\p_nf\rangle$;
\item Let $\delta:H^{n-2}(X_f)\rightarrow H^{n-1}(\mathbb{P}^{n-1})$ be the coboundary operator in the Poincar\'{e} residue sequence, then the cup product in the projective space is given by
$$\delta(\res\Omega_{\phi}\wedge\res\Omega_{\psi})=c_{ab}\frac{\phi\psi\Omega}{\p_1f\cdots\p_nf},$$
where $c_{ab}=\frac{(-1)^{a(a-1)/2+b(b-1)/2+n+(b-1)^2}}{(a-1)!(b-1)!}\deg f$.
\item %The pairing is related as follows% Let $\lambda$ be the residue of the fundamental class{\color{red}which fundamental class, why it has a pole?} of $\mathbb{P}^{n-1}$, viewed in $H^{n-1}(\mathbb{P}^{n-1},\C)$.{\color{red} You means $H^{n-1}(\mathbb{P}^{n-1}-X_f,\C)$? } Then
$$\int_{X_f}\res\Omega_{\phi}\wedge\res\Omega_{\psi}=\frac{\tilde{c}_{ab}}{(2\pi i)^n}\int_{\Gamma(\epsilon)}\frac{\phi\psi dz_1\wedge\cdots\wedge dz_n}{\p_1f\cdots\p_nf},$$
where $\tilde{c}_{ab}=\frac{(-1)^{a(a-1)/2+b(b-1)/2+n+(b-1)^2}}{(a-1)!(b-1)!},$ $\Gamma(\epsilon)=\{|\p_if|=\epsilon: i=1,\ldots,n\}$, and the right hand side is the Grothendieck residue symbol.%{\color{red} we need to address the order pole filtration explicitly or at least give an reference. We need explicitly expression of $c_a$ in thm 4.7, also cite this thm in the proof of thm4.7 instead.}
\end{enumerate}
\end{prop}
According to the above result, for a non-degenerate homogeneous degree $n$ polynomial $f$, we consider the subspace $\Jac(f)'$ of the Jacobi ring of $f$, given by
$$\Jac(f)'=\Span\left\{\phi_i~\bigg|~\frac{\deg\phi_i}{n}\in\mathbb{N}, \{\phi_i\}_{i=0}^{\mu-1} \text{ is a homogeneous basis of $\Jac(f)$}\right\}.$$
%Set $\mu'=\dim\Jac(f)$.
%$\phi_i$, with $n|\deg\phi_i$.{\color{red}we should mension $\phi$ is a basis somewhere.}
Then the proposition above would be needed when we establish the CY/LG correspondence between $\Jac(f)'$ and $H^{n-2}(X_f)_{\prim}$.
%\begin{rem} \textcolor{blue}{Say some words about MHS?}
%\end{rem}
%\textcolor{blue}{The correspondence between the small state space.}
%The Hodge structure on $H^{n-1}(\mathbb{P}^{n-1}-X_f)$ is induced by the trivial filtration on the $\C^{\infty}$ logarithm de Rham complex (fine resolution of holomorphic logarithm de Rham complex):
%$$F^p\Omega_{\mathbb{P}^{n-1}}^*(\log X_f)=\{0\rightarrow \Omega_{\mathbb{P}^{n-1}}^p(\log X_f)\rightarrow \Omega_{\mathbb{P}^{n-1}}^{p+1}(\log X_f)\rightarrow\cdots\}.$$
%The induced Hodge filtration is the limit filtration in the collapsed spectral sequence:
%$$E_1^{p,q}=H^q(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}^q(\log X_f))~\Rightarrow~ H^{p+q}(\mathbb{P}^{n-1}-X_f,\mathbb{C}).$$
%By Deligne Theorem, $E_1=E_{\infty}$.
%
%
%
%The computation of $H^q(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}^q(\log X_f))$ is related to the filtration by order of pole along $X_f$. First, there is an exact sequence
%$$0\rightarrow \Omega_{\mathbb{P}^{n-1}}^p(\log X_f)\rightarrow\Omega_{\mathbb{P}^{n-1}}^p(X_f)\stackrel{d}{\rightarrow} \Omega_{\mathbb{P}^{n-1}}^{p+1}(2X_f)/\Omega_{\mathbb{P}^{n-1}}^{p+1}(X_f)\stackrel{d}{\rightarrow} \Omega_{\mathbb{P}^{n-1}}^{p+2}(3X_f)/\Omega_{\mathbb{P}^{n-1}}^{p+2}(2X_f)\rightarrow\cdots$$
%By Bott's vanishing theorem, Griffiths proved that the Hodge filtration on $H^{n-1}(\mathbb{P}^{n-1}-X_f,\C)$ can be represented by
%$$F^{n-p}H^{n-1}(\mathbb{P}^{n-1})=\mbox{Im}\left(H^0(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}(pX_f))\rightarrow H^{n-1}(\mathbb{P}^{n-1}-X_f,\C)\right).$$
%
%
%The element of $H^0(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}(kX_f))$ can be written as a rational differential $n-1$-form
%$$\frac{A(z)\Omega}{f(z)^k},\quad \Omega=\sum_{i=1}^n(-1)^{i+1}z_idz_1\wedge\cdots\wedge\widehat{dz_i}\wedge\cdots\wedge dz_n$$
%and $A(x)$ is a homogeneous polynomial of degree $kd-(n-1)$.
%
%
%\begin{thm}[Griffiths] (1) $H^0(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}((n-1)X_f))=H^0(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}(*X_f))$;\\
%(2) There is a natural injection $H^0(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}(k X_f))\subset H^0(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}((k+1)X_f))$ for each $k\geq1$.
%\end{thm}
%
%
%%There is a natural pole order filtration [\textcolor{blue}{Griffith}] of $H^0(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}(*X_f))$ given by
%%\begin{equation}\label{poleorder}
%%H^0(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}(X_f))\subset H^0(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}(2X_f))\subset %H^0(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}(3X_f))\subset\cdots\subset \subset H^0(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}((n-1)X_f))=H^0(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}(*X_f)).
%%\cdots\subset H^0(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}((n-1)X_f))=H^0(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}(*X_f)).
%%\end{equation}
%
%
%
%\begin{defn} The residue map is defined to be
%\begin{align*}
%\res:H^{0}(\mathbb{P}^{n-1},\mathbb{P}^{n-1}((n-1)X_f))&\rightarrow H^{n-2}(X_f,\C)\\
%\frac{1}{2\pi i}\int_{\tau(\gamma)}\frac{F(x)}{G(x)^k}\Omega_n&=\int_{\gamma}\Res\left(\frac{F(x)\Omega_n}{G(x)^k}\right)
%\end{align*}
%where $\tau:H_{n-2}(X_f)\rightarrow H_{n-1}(\mathbb{P}^{n-1}-X_f)$ be the Leray coboundary map, i.e. $\tau(\gamma)$ is the tube over $\gamma$.
%\end{defn}
%
%
%
%\begin{prop} The residue map descends to an injective map
%$$\res:H^{n-1}(\mathbb{P}^{n-1}-X_f,\C)\rightarrow H^{n-2}(X_f,\C),$$
%with the image being the $(n-2)$-th primitive cohomology of $X_f$. Moreover, the residue map $\Res$ takes the pole order filtration
%$$H^0(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}(X_f))\subset H^0(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}(2X_f))\subset
%\cdots\subset H^0(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}((n-1)X_f))=H^0(\mathbb{P}^{n-1},\Omega_{\mathbb{P}^{n-1}}(*X_f))$$
%onto the Hodge filtration on $H^{n-2}(X_f,\C)_{\prim}$
%$$\F^{n-2}H^{n-2}(X_f,\C)_{\prim}\subset\F^{n-3}H^{n-2}(X_f,\C)_{\prim}\subset\cdots\subset\F^{0}H^{n-2}(X_f,\C)_{\prim}=H^{n-2}(X_G,\C)_{\prim}.$$
%In particular, if $d=n$,
%$$\res\left(\frac{\Omega}{f}\right)\in \F^{n-2}H^{n-2}(X_f,\C)_{\prim}=H^{n-2,0}(X_{f},\C)_{\prim}$$
%is a nowhere vanishing holomorphic volume form on $X_f$.
%\end{prop}
%
%
%The Griffiths-Carlson theory provides a correspondence between the cup product and the pairing on both sides.
%
%
%
%
%For any $t\in \C^*$, $V_t:=\{z\in\C^n:f(z)=t\}.$ Let us consider a projective compactification of $V_{1}$ in a weighted projective space. Let $q_{i}=a_{i} / b_{i}$ with $\left(a_{i}, b_{i}\right)=1$, and $d= \mathrm{lcm}\left(b_{1}, \ldots, b_{n}\right)$, i.e. $d$ is the least common multiple of $b_1,...,b_n$. We put $Q_{i}=q_{i} d$. Let $z=(z_1,...,z_n)$, and $Z=(z,z_{n+1})$, then the polynomial
%$$
%F_{t}(Z) \equiv f\left(z\right)-t z_{n+1}^{d}.
%$$
%is also quasi-homogeneous. Let $M=\mathbb{P}\left(Q_{1}, Q_{2}, \ldots, Q_{n}, 1\right)$ be the weight projective space with weights $(Q_1,Q_2,...,Q_n,1)$, then $$\bar{V}_t:=\{[Z]\in \mathbb{P}\left(Q_{1}, Q_{2}, \ldots, Q_{n}, 1\right):F_t(Z)=0\}.$$
%If we identify $\C^{n}$ with the open set in $M$ where $z_{n+1} \neq 0$, we see that $V_{t}=\bar{V}_t \cap C^{n}$. Generically, $\bar{V}_t$ is not smooth: it is an orbifold. Let $M_{\infty}$ be the hypersurface in $M$ given by $z_{n+1}=0$. Then $M_\infty\cong\mathbb{P}\left(Q_{1}, Q_{2}, \ldots, Q_{n}\right)$. Consider the part of $\bar{V}_t$ at infinity,
%$$
%V_{\infty} \equiv \bar{V} \backslash V_{t}=\bar{V} \cap M_{\infty},
%$$
%i.e. the hypersurface in $M_{\infty}$ is given by the equation
%$$
%f(z)=0.
%$$
%Again, $V_{\infty}$ is an orbifold (in general).
%It follows from the Mayer-Vietoris exact sequence of the pair $\left(\bar{V}_t, V_{t}\right)$ and the Thom isomorphism $H^{*}\left(\bar{V}_t, V_{t}\right) \simeq$ $H^{*-2}\left(V_{\infty}\right)$
%that
%\begin{equation}\label{gysin}
% \cdots \rightarrow H^{i}(\bar{V}_t) \rightarrow H^{i}\left(V_{t}\right) \rightarrow H^{i-1}\left(V_{\infty}\right) \rightarrow H^{i+1}(\bar{V}_t) \rightarrow \cdots
%\end{equation}
%
%
%Hence
%$$
%H^{n-1}\left(V_{t}\right) \simeq H^{n-1}(\bar{V}_t)_{0} \oplus H^{n-2}\left(V_{\infty}\right)_{0},
%$$
%where
%$$
%\begin{array}{c}
%H^{n-1}(\bar{V})_{0}=\operatorname{coker}\left(H^{n-3}\left(V_{\infty}\right) \rightarrow H^{n-1}(\bar{V}_t)\right) \\
%H^{n-2}\left(V_{\infty}\right)_{0}=\operatorname{ker}\left(H^{n-2}\left(V_{\infty}\right) \rightarrow H^{n}(\bar{V}_t)\right) .
%\end{array}
%$$
%From the Lefschetz theorem it is clear that $H^{n-2}\left(V_{\infty}\right)_{0}$ is just the primitive cohomology of the K?hler orbifold $V_{\infty}$. The weight graduation of $H^{n-1}\left(V_{t}\right)$ is
%\begin{equation}\label{shift}
%\begin{array}{c}
%\mathrm{Gr}_{n-1} H^{n-1}\left(V_{t}\right) \simeq H^{n-1}(\bar{V})_{0} \\
%\mathrm{Gr}_{n} H^{n-1}\left(V_{t}\right) \simeq H^{n-2}\left(V_{\infty}\right)_{0}[-1]\footnote{$[-1]$ means "Tate twist", i.e., a degree shift by $-1$ in the Hodge filtration:
%$$
%{F}^{p} \mathrm{Gr}_{n} H^{n-1}\left(V_{t}\right) \simeq {F}^{p-1} H^{n-2}\left(V_{\infty}\right)_{0},
%$$
%which defines a morphism of Hodge structures of type $(-1,-1)$.} .
%\end{array}
%\end{equation}
%Hence $H^{n-1}(\bar{V})_{0}$ and $H^{n-2}\left(V_{\infty}\right)_0$ carry pure Hodge structures which define the Hodge filtration of the group $H^{n-1}\left(V_{t}\right)$.
%
%
%As vector spaces, (c.f. \cite{DY2020cohomology, Fan2020LGCYCB,fan2011schr}),
%
%$$H^{n-1}(V_t)\simeq \Omega^n(\C^n)/df\wedge\Omega^{n-1}(\C^n)\simeq \Jac(f).$$
%
%Now we describe the mixed Hodge structure on $\Omega^n(\C^n)/df\wedge\Omega^{n-1}(\C^n)\simeq \Jac(f)$ directly.
%
%Fix a homogeneous basis $\{\phi_a\}_{a=0,\ldots,\mu-1}$ of $\Jac(f)$ with $\phi_0=1$, $\deg(\phi_a)\leq n-2$, and let $A_a:=\phi_a dz_1\ldots dz_n\in \Omega^n(\C^n)/df\wedge\Omega^{n-1}(\C^n)$.
%
%\begin{defn} \label{l}For $A=z_1^{\beta_1}\cdots z_n^{\beta_n}dz_1\wedge\cdots\wedge dz_n$, we define
%$$l(A):=\sum_{i=1}^n(\beta_i+1)Q_i.$$
%\end{defn}
%
%{\bf{Now assume that $f$ is homogeneous of degree $n$.}} In this case, $Q_i\equiv1.$
%
%Let $\Jac(f)':=\Span\{\phi_a:n|l(A_a)\}$, then by \cite{steenbrink1977intersection}, as a vector space
%$$ \mathrm{Gr}_{n} H^{n-1}\left(V_{t}\right) \simeq\Jac(f)'.$$ Moreover, one also has
%\[{F}^{p} \mathrm{Gr}_{n} H^{n-1}\left(V_{t}\right) \simeq\Span\{\phi_a\in \Jac(f)':l(A_a)\leq n(n-1-p)\}.\]
%
%In Section ?, we will define a map $R:\Jac(f)'\to H^{n-2}(V_\infty)_0[-1]$ that preserves the filtration. In particular, $R(1)\in \Omega^{n-2,0}(V_\infty).$
%
\subsection{Main theorems}
%The Landau-Ginzburg B-model at genus-zero is related to Saito's theory of primitive form \cite{1981Primitive} and higher residue \cite{1983higher}, which induces the structure of a Frobenius manifold on the universal deformation space of the singularity \cite{Saito2008}. There is also a Hodge theoretical aspect (i.e. $tt^*$ geometry structure) in Landau-Ginzburg B-model discovered by Cecotti-Vafa \cite{cecotti1991topological}. Its integrability structure is studied by Dubrovin \cite{Dub1993tt}. Later, C. Hertling \cite{Hert2003tt} studied the $tt^*$ geometric structure systematically, and organized various known structures into the so-called TERP structure. H. Fan \cite{fan2011schr} presented an analytic approach by considering the $L^2$ Hodge theory of a twisted Laplacian operator in the spirit of $N=2$ geometry. For the case $f$ with critical locus, Li-Wen \cite{li2019l2} also used the $L^2$ Hodge theory to give a Frobenius manifold structure.
In this paper, we investigate the $tt^*$ structure of the Landau-Ginzburg B-model and CY/LG correspondence. We study the $tt^*$ geometric structure in this paper from the viewpoint of Schr\"{o}dinger representations of 1d Landau-Ginzburg model \cite{cecotti1991topological} and \cite{fan2011schr}, see next section for more details.
Consider the real twisted Laplacian operator $\Delta_f$ associated to $(\C^n,f)$ (See Section \ref{Hodgedecomposition} for the definition of $\Delta_f$). The space of harmonic forms $H_f$ (with respect to $\Delta_f$) is called the vacuum space in the physical literature, and is isomorphic to the Jacobi ring of the polynomial $f$ via the so-called spectral flow $S$ (See the definitions in subsection \ref{S}). The complex conjugate on differential forms gives a natural real structure on $H_f$.
Furthermore, consider the marginal deformation $F$ of $f$ parametrized by the space $u\in M$
$$F(z,u)=f(z)+\sum_{i=1}^su_i\psi_i,$$
where $\psi_i$ has the same weight as $f$ (see Definition \ref{quasi}).
The Hodge bundle is defined over the deformation space $M$. Fiberwise, at $u\in M$, it is given by the harmonic forms of $\Delta_{F(\cdot,u)}$. The $tt^*$ metric on $M$ is defined as
$$h(\alpha,\beta)=\int_{\C^n}\alpha\wedge*\bar{\beta},\quad \alpha,\beta\in\ker(\Delta_{F}). $$
Next, we construct a $C^{\infty}(M)$-linear isomorphism (see subsection \ref{S} for details)
$$S:C^{\infty}(M)\otimes\Omega_{\C^n\times M/M}^{n}/dF\wedge\Omega_{\C^n\times M/M}^{n-1}\longrightarrow\ker\Delta_F.
$$%{\color{red}(Maybe wite down the definition of small tt str somewhere or remove this sentnece)Then we discuss the corresponding $tt^*$ connection, Higgs operator and anti-Higgs operator. }
Via the isomorphism $S$ and some $U(1)$ action analysis, we will study the vacuum line bundle and Weil-Peterson-type metric in Lanudau-Ginzburg B-model. %of Calabi-Yau side{\color{red}{(This sentence looks weird)}},
We will derive the similar relation between $tt^*$ metric and Weil-Peterson metric as in Calabi-Yau's case:
Let $u=(u_1,...,u_s)$ be local coordinates of $M$, and $\p_i:=\partial_{u_i} $.
\begin{thm} \label{131}The Weil-Peterson-type metric $G_{i\bar{j}}$ is related to the $tt^*$ metric $h$ in the following way:
$$G_{i\bar{j}}=\frac{h(w_i,w_j)}{h(w_0,w_0)},$$
where
\begin{itemize}
\item $w_0=S(dz_1\wedge\cdots\wedge dz_n)$ is a holomorphic section of the vacuum line bundle, and
$w_i=S(\psi_idz_1\wedge\cdots\wedge dz_n)$ is the holomorphic section corresponding to the deformation $\psi_i$.
\item $G_{i\bar{j}}:=-\partial_i\bar{\p}_{\bar{j}}\log h(w_0,w_0)$.
\end{itemize}
\end{thm}
Then, to understand $tt^*$ sturcture, we investigate the following two types of period integrals (see subsection \ref{periodintegral} for more details).
$$\int_{\gamma^-}e^{F+\bar{F}}w\quad\text{and}\quad \int_{\gamma^-}e^FA,$$
where $w=S(A)$.
By choosing a basis $\{\gamma_k^-\}$ (Lefschetz thimble) in the homology class $H_n(\C^n, F^{-\infty};\mathbb{Z})$ and a holomorphic basis $\{A_a\}$ of $C^{\infty}(M)\otimes\Omega_{\C^n\times M/M}^n/dF\wedge\Omega_{\C^n\times M/M}^{n-1}$, there exists a matrix-valued function $\T(u,\bar{u})$ such that
$$\int_{\gamma_k^-}e^FA_a=\sum_b \T_{ab}(u,\bar{u})\int_{\gamma_k^-}e^{F+\bar{F}}w_b,$$
where $w_a=S(A_a).$
In \cite{cecotti1991topological}, using Leznov-Saveliev method, Cecotti-Vafa shows that there exist (c.f. (A.9) in \cite{cecotti1991topological}) a block-diagonal and holomorphic matrix $\mathcal{F}$ and a unit lower triangular matrix $\mathcal{N}$, such that $\T=e^{\mathcal{F}}\mathcal{N}.$ As a result, there exists a holomorphic function $\lambda(u)$, such that
\[\int_{\gamma_k^-}e^Fdz_1\wedge\cdots\wedge dz_n=e^{\lambda(u)}\int_{\gamma_k^-}e^{F+\bar{F}}w_0.\]
However, by introducing some $U(1)$ actions and using the Agmon estimate (see, for example, \cite{DY2020cohomology} or Section \ref{witagm}), we have the following structure theorem for the matrix-valued functions $\T$. Moreover, we show $\lambda(u)\equiv0:$
\begin{thm}\label{lgcy1} When we choose a specific monomial representation of a basis in $\Jac(f)$ and required the elements of that basis listed in a specific order (See Section \ref{ttgeometry}), then the matrix-valued function $\T$ is a (block) unit lower triangular matrix. More explicitly,
\begin{align*}
(\T)_{aa}&=1 \mbox{ for~ $0\leq i\leq \mu-1$};\\
(\T)_{ab}&=0 \mbox{ if~ $a<b$};\\
(\T)_{ab}&=\begin{cases}
0, &\mbox{ if~ $\frac{l(A_a)-l(A_b)}{n}\notin \mathbb{Z}^+$};\\
\frac{(-1)^{l_{ab}}}{l_{ab}!}\int_{\C^n}\bar{f}^{l_{ab}}A_a\wedge*\bar{w}_b, &\mbox{ if~ $l_{ab}:=\frac{l(A_a)-l(A_b)}{n}\in \mathbb{Z}^+$.}
\end{cases}
\end{align*}
In particular,
\[\int_{\gamma_k^-}e^{F+\bar{F}}w_0=\int_{\gamma_k^-}e^Fdz_1\wedge\cdots\wedge dz_n.\]
%\[\int_{\tilde{\gamma}_k}e^{-F-\bar{F}}*w_0=\int_{\tilde{\gamma}_k}e^{-F}*\dz,\]
\end{thm}
Let
$$X_F=\{[z]\in \C P^{n-1}:F(z,u)=0 \}.$$ In Section 4, we define a map $R_u:\Jac(F(\cdot,u))'\to H^{n-2}(X_F)_{\prim}[-1]$ that preserves the Hodge filtration. In particular, $R(1)$ is a nowhere vanishing holomorphic $(n-2)$-form on $X_F$. Now let $G^{CY}:=-\partial\bar{\p}\log\int_{X_F}R_u(1)\wedge*\overline{R_u(1)}$, then $G^{CY}$ is the Weil-Peterson metric on $M.$ It follows from Theorem \ref{lgcy1}, that
\begin{thm}\label{133}
$G=G^{CY}$(See Theorem \ref{131} for the definition of the metric $G$).
\end{thm}
Finally, we show the full Calabi-Yau/Landau-Ginzburg correspondence
\begin{thm}\label{134}
The small $tt^*$ structure on LG's side (See Definition \ref{smallLG}) and the $tt^*$ structure on CY's side (See in the last subsection) are isomorphic.
\end{thm}
\subsection{Organizations}
The following is the organization of the paper. In Section 2, we introduce the $tt^*$ geometric structure on the Hodge bundle in the Landau-Ginzburg B-model and use the spectral flow $S$ to establish a relationship between the space of harmonic forms and the Jacobi ring. Then, we explore the vacuum line bundle and the Weil-Peterson metric on Landau-Ginzburg's side. Moreover, we would like to emphasis that following this section, we choose a specific monomial representation of basis of $\Jac(f)$ and required them listed in a specific order (See Section \ref{ttgeometry}). Section 3 discusses the Lefschetz thimbles and investigates two different types of period integrals. Lastly, we prove the Calabi-Yau/Landau-Ginzburg correspondence for $tt^*$ structure in Section 4 using period integrals. In the appendix, we summarize some useful results about Agmon estimate, and prove the Riemann bilinear formula.\\
\noindent \textbf{Acknowledgments}. The authors thank Huijun Fan for useful discussions on related subjects. The work of X.T. is partially supported by Tsinghua Postdoc Grant 100410058 and research funding at BIMSA. The work of J.Y. is partially supported by Mathematics Department Block Grant Fellowship at UCSB. Part of this work was done while J.Y. was visiting Peking University in Spring 2021. J.Y. thanks for their hospitality and provision of excellent working environment.
\section{Deformation Theory and Weil-Peterson Metric on the Marginal Deformation Space}
\subsection{Hodge decomposition}\label{Hodgedecomposition}
\begin{defn}[Quasi-homogeneous polynomials]\label{quasi}
We say a polynomial $f: \C^n\rightarrow\C$ is quasi-homogenous (of $(q_1,...,q_n)$ for $q_1,\ldots,q_n\in\mathbb{Q}$), if for any $\lambda\in\C^*$,
$$f(\lambda^{q_1}z_1,\ldots, \lambda^{q_n}z_n)=\lambda f(z_1,...,z_n).$$
Each $q_i$ is called the weight of $z_i$, and let $\w(z_i):=q_i$. We extend the definition of $\w$ to the monomial $\prod_{i=1}^nz_i^{k_i}$, such that $\w(\prod_{i=1}^nz_i^{k_i})=\sum_{i=1}^nk_iq_i$. In particular,
$\w(f)=1$.
In the physical literature, the central charge $\hat{c}_f$ is defined to be
$$\hat{c}_f=\sum_{i=1}^n(1-2q_i).$$
\end{defn}
%Consider a pair $(\C^n,f)$, where $f$ is quasi-homogeneous.
While, for convenience, we assume that $f$ is non-degenerate, i.e. we require that
\begin{enumerate}
\item $f$ contains no monomial of the form $z_iz_j$ for $i\neq j$,
\item $f$ has only an isolated singularity at the origin.
\end{enumerate}
Next, we'll go over some physcis terminology. We consider the pair $(\C^n,f)$. The polynomial $f$ is called the superpotential for LG model. The Schr\"{o}dinger representation of 1d LG model consists of a Hilbert space, which is the space of $L^2$-integrable forms $L^2\Lambda^*(\C^n)$, and four charge operators $Q_{+},Q_{-}, Q_{+}^{\dag},Q_{-}^{\dag}$, which are operators on $L^2\Lambda^*(\C^n)$ given by
\begin{flalign*}
&Q_{+}=\bar{\p}_{f}=\bar{\p}+df\wedge,\quad Q_{-}=\p_{f}=\p+d\bar{f}\wedge,\\
&Q_{+}^{\dag}=\bar{\p}_{f}^{\dag}=-\ast \p_{-f}\ast, \quad Q_{-}^{\dag}=\p_{f}^{\dag}=-\ast \bar{\p}_{-f}\ast.
\end{flalign*}
Here $\ast$ is the Hodge star operator with respect to the standard hermitian metric on $\C^n$. Then we have the following relation, called supersymmetric algebra structure in the physical literature
\begin{flalign*}
&\p_{f}^2=\p_{f}^{\dag 2}=\overline{\p}_{f}^2=\overline{\p}_{f}^{\dag 2}=0,\\
&\{\p_{f},\p_{f}^{\dag}\}=\{\overline{\p}_{f},\overline{\p}_{f}^{\dag}\}=\Delta_{f},\\
&\{\overline{\p}_{f},\p_{f}\}=\{\overline{\p}_{f}^{\dag},\p_{f}^{\dag}\}=\{\p_{f},\overline{\p}_{f}^{\dag}\}=\{\overline{\p}_{f},\p_{f}^{\dag}\}=0,
\end{flalign*}
where $\{\cdot,\cdot\}$ is the anti-commutative bracket, that is, $\{P,Q\}=PQ+QP$.
In the spirit of Cecotti-Vafa's paper\cite{cecotti1991topological}, Klimek-Lesniewski\cite{Klimek1991local} and Fan\cite{fan2011schr} used the $L^2$ analysis to study the $L^2$-cohomology $H^{*}_{(2),\overline{\p}_{f}}(\C^n)$ of $(L^2\Lambda^*(\C^n), \overline{\p}_{f})$ and the harmonic forms $H_f^*:=\ker\Delta_{f}$. Let us recall their results as follows.
It is easy to check that $df$ satisfy the elliptic condition\footnote{Our $df$ is of the form $f=\sum_{j=1}^nf_j(z)dz_j$ in \cite{Klimek1991local}.} in \cite{Klimek1991local}, that is
\begin{enumerate}
\item $|df(z)|^2:=\sum_{\nu=1}^n|\p_{z_{\nu}}f(z)|^2\rightarrow\infty$, as $|z|\rightarrow\infty$;
\item For any $\epsilon>0$, there is a constant $C$ such that
$$|\p_{z_{\nu_1}}\p_{z_{\nu_2}}f(z)|\leq\epsilon|df(z)|^2+C,\quad\text{for all }z\in\C^n,\text{ and }\nu_1,\nu_2=1,\ldots,n.$$
\end{enumerate}
Then we see that
\begin{thm}[Hodge Theorem \cite{Klimek1991local} or \cite{fan2011schr}]\label{Hodgethm} Let $f\in\C[z_1,\ldots,z_n]$ be a non-degenerate quasi-homogeneous polynomial. Then
\begin{enumerate}
\item $\dim H_f^*\leq\infty$.
\item \emph{(Hodge decomposition)} There are orthogonal decompositions
$$L^{2}\Lambda^{*}(\C^n)=H_{f}^{*}\oplus\Im(\overline{\p}_{f})\oplus \Im(\overline{\p}_{f}^{\dag}).$$
More precisely, there is a self-adjoint compact operator $G_{f}$ on $L^2\Lambda^*(X)$ such that
$$L^2\Lambda^k(\C^n)=H_f^k\oplus \overline{\p}_{f}\left(\overline{\p}_{f}^{\dag}G_{f}L^2\Lambda^k(\C^n)\right)\oplus \overline{\p}_{f}^{\dag}\left(\overline{\p}_{f}G_{f}L^2\Lambda^k(\C^n)\right).$$
%\item There is a canonical isomorphism
%$$H^k_{(2),\overline{\p}_{f_{0}}}(X)\cong H^k_{\overline{\p}_{f_{0}}}(X),\quad 0\leq k\leq 2n.$$
\item \emph{(Vanishing Theorem)}
\begin{equation}
\label{eq:abs}
H_{f}^{k}\cong H^k_{(2),\overline{\p}_{f}}(\C^n)\cong
\begin{cases}
0& \text{if $k \neq n$}\\
\Omega^{n}(\C^n)/df\wedge\Omega^{n-1}(\C^n)\cong\Jac(f) & \text{if $k=n$}.
\end{cases}
\end{equation}
\end{enumerate}
\end{thm}
\begin{rem}\label{KHD} One can also introduce the Witten deformation $d_{2\Re(f)}:=d+2d\Re(f)\wedge$ of de Rham differential $d$. By K\"ahler identity,
$$\Delta_{2\Re(f)}=\{d_{2\Re(f)},d_{2\Re(f)}^{\dagger}\}=2\Delta_f.$$
Moreover, we have the Kodaira-Hodge decomposition, see for example, \cite{DY2020cohomology},
$$L^{2}\Lambda^{*}(\C^n)=\ker(\Delta_{2\Re(f)})\oplus \Im(d_{2\Re(f)})\oplus \Im(d_{2\Re(f)}^{\dag}).$$
\end{rem}
%\textcolor{blue}{Maybe also see some words about $d+df\wedge$}
\subsection{$tt^*$ geometry}\label{ttgeometry}
Fix a monomial representation $\{\phi_a\}_{a=0}^{\mu-1}$ of basis of $\Jac(f)$, where $\mu=\dim\Jac(f)$ and $\phi_0=1$. Moreover, we assume that $\w(\phi_a)\leq \w(\phi_b)$ if $a<b$ (See Definition \ref{quasi} for the definition of $\w$). In particular, let $\{\psi_i:=\phi_{a_i}\}_{i=1}^s$ be the subset of $\{\phi_a\}_{a=0}^{\mu-1}$, such that each $\psi_i$ has weight $1$, $i=1,\ldots,s$. We will study marginal deformation of $f$
$$F(z,u)=f(z)+\sum_{i=1}^{s}u_{i}\psi_i(z).$$
%be the marginal deformation of $f$, i.e. $\w(\phi_i)=\w(f)=1$. In particular, we choose $\phi_i$ to be %{\color{red}The definition of wt(f) is not clear}, $i=1,\ldots,s$.
We denote the space of parameters ${u_{i}}$ by $M$, which could be taken as a small neighborhood of the origin in $\C^s$. Therefore, we have a family of supersymmetric algebra operators $\bar{\p}_{F},\p_{F}, \bar{\p}^{\dag}_{F},\p_{F}^{\dag},\Delta_{F}$ parameterized by $u\in M$. When $|u|$ is small enough, the results in the previous subsection apply to $\Delta_F$. In particular, $\Jac(F(\cdot, u))$ is isomorphic to $\Jac(f)$ as a vector space.
We have the trivial complex Hilbert bundle $L^2\Lambda^{*}(\C^n)\times M\rightarrow M$. For simplicity, denote by $L^2\mathcal{A}$ its $L^2$-integrable section space. There are two natural parings,
\begin{flalign*}
h: ~&L^2\A\times L^2\mathcal{A}\longrightarrow C^{\infty}(M)\qquad h(\alpha,\beta)=\int_{X}\alpha\wedge\ast\bar{\beta},\\
\eta: ~&L^2\mathcal{A}\times L^2\mathcal{A}\longrightarrow C^{\infty}(M)\qquad \eta(\alpha,\beta)=\int_{X}\alpha\wedge\ast\beta.
\end{flalign*}
Note that there is a canonical real structure on the Hilbert bundle induced from the complex conjugate in $L^2\Lambda^*(\C^n)$, which was denoted by $\tau_{R}$. Then $h(\alpha,\beta)=\eta(\alpha,\tau_R\beta)$.
%Let $\{\a_{a}\}_{a=1}^{\infty}$ be a basis of $L^{2}\Lambda^{*}$ and then $\a_{\bar{a}}:=\overline{\a_{a}}=\tau_{R}\cdot\a_{a}$ also forms a basis of $L^{2}\Lambda^{*}$.
Let $H^n:=H_{F}^{n}$ be the Hodge bundle over $M$. Fiberwisely, $H^n\big|_u$ is the space of harmonic forms of $\Delta_{F(u)}$. We denote the space of its section by $\H^n$.
Let $\Pi_u: L^2\A\rightarrow \H^n$ be the harmonic projection, $G$ be the inverse of $\Delta_{F}$ on $\Im(\bar{\p}_{F})\oplus\Im(\bar{\p}_{F}^{\dag})$, then $G$ commutes with the operators $\bar{\p}_{F},\p_{F}, \bar{\p}^{\dag}_{F},\p_{F}^{\dag},\Delta_{F}$, and the Hodge decomposition reads
$$\Id=\Pi_u+\Delta_{F}G=\Pi_u+G\Delta_{F}.$$
Now we introduce some important operators on the Hodge bundle that give the $tt^*$ structures. Let $u=(u_1,...,u_s)$ be local coordinates of $M$, and $\partial_i:=\partial_{u_i}$, $D_i=D_{\frac{\p}{\p u_i}}$.
\vskip 0.1cm
\noindent(1) The connection $D$, $\bar{D}$
\vskip 0.1cm
Notice that the Hodge bundle is embedded into the trivial Hilbert bundle, so we can define $D$, $\bar{D}$ in a natural way:
$$D_i=\Pi_u\circ \p_i,\quad \bar{D}_{\bar{i}} =\Pi_u\circ\bar{\p}_{\bar{i}}\quad i=1,...,s.$$
\noindent(2) The operators $C_i, \bar{C}_{\bar{i}}$
\vskip 0.1cm
We define $C_{i}=\Pi_u\circ\p_{i}F=\Pi_u\circ\psi_i$, $\bar{C}_{\bar{i}}=\Pi_u\circ\overline{\p_iF}=\Pi_u\circ\overline{\psi_i}$. We can easily show that
$$C_{i}\alpha_{a}=(\p_{i}F\alpha_{a})-\bar{\p}_{F}\bar{\p}^{\dag}_{F}G(\p_{i}F)\alpha_{a}, \quad \bar{C}_{\bar{i}}\alpha_{a}=(\overline{\p_{i}F}\alpha_{a})-\p_{ F}\p^{\dag}_{F}G(\overline{\p_{i}F})\alpha_{a}.$$
By definition, $\bar{C}_{\bar{i}}$ is the adjoint operator of $C_i$ with respect to the $tt^*$ metric $h$, i.e.
$$h(C_i\alpha_a,\alpha_b)=h(\alpha_a,\bar{C}_{\bar{i}}\alpha_b).$$
%Or equivalently we define the components
%$$C_{ia\bar{b}}=g(\p_if \a_{a},\a_{b}),\quad \bar{C}_{\bar{i}a\bar{b}}=g(\overline{\p_i f}\a_{a},\a_{b}).$$
\begin{prop}[$tt^*$ equation] $D_i,\bar{D}_{\bar{j}}$, $C_i, \bar{C}_{\bar{j}}$ satisfy the following equations
\begin{enumerate}
\item $[C_{i},C_{j}]=0, \quad [\bar{C}_{\bar{i}}, \bar{C}_{\bar{j}}]=0,\quad [D_{i},\bar{C}_{\bar{j}}]=[\bar{D}_{\bar{i}},C_{j}]=0;$
\item $[D_{i},C_{j}]=[D_{j},C_{i}], \quad [\bar{D}_{\bar{i}},\bar{C}_{\bar{j}}]=[\bar{D}_{\bar{j}},\bar{C}_{\bar{i}}]$;
\item $[D_i, D_j]=0,\quad [\bar{D}_{\bar{i}},\bar{D}_{\bar{j}}]=0,\quad [D_{i},\bar{D}_{\bar{j}}]=-[C_{i},\bar{C}_{\bar{j}}].$
\end{enumerate}
\end{prop}
\begin{proof} See, for example, \cite{Tang2017ttGS}.
\end{proof}
\subsection{The isomorphism $S$}\label{S}
%{\color{red}To be more clear, maybe replace $X$ with $\C^n$, $X\backslash \mbox{Crit}(F)$ with $(\C^N)^*:=\C^n-\{0\}.$ The definition of $Crit(F)$ is not clear.}
Now take a $C^{\infty}(M)$-basis $\{A_{a}(t)\}_{a=1}^{\mu}$ of $C^{\infty}(M)\otimes\Omega^{n}_{\C^n\times M/M}/dF\wedge\Omega^{n-1}_{\C^n\times M/M}$\footnote{where $\Omega^*_{\C^n\times M/M}$ is the sheaf of relative holomorphic differential forms.}, we will construct a basis of $\H^n$ from it. To do this, we follow the method in \cite{li2013primitive} to show that
$$H^{n}_{\bar{\p}_{F}}(\C^n)\cong H^{n}_{c,\bar{\p}_{F}}(\C^n),$$
where the left side is the cohomology of smooth complex, and the right side is the cohomology of smooth complex having a compact support.
Let $(\C^n)^*:=\C^n-\{0\}$, we define the operator
$$V_{F}=\sum_{\nu=1}^{n}\frac{\overline{\p_{\nu}F}}{|\nabla F|^2}(dz_{\nu})^{\dagger}=\frac{(\p F)^{\dagger}}{|\nabla F|^2}:\quad \mathcal{A}^{*,*}((\C^n)^*)\rightarrow \mathcal{A}^{*-1,*}((\C^n)^*).$$
Next, given a cut-off function $\rho$, which is compactly supported in a small disc around the origin, we define
$$T_{\rho}: \mathcal{A}^{*}(\C^n)\rightarrow \mathcal{A}_{c}^{*}(\C^n),\qquad R_{\rho}:\mathcal{A}^{*}(\C^n)\rightarrow \mathcal{A}^{*}(\C^n)$$
by
$$T_{\rho}(A)=\rho A+(\bar{\p}\rho)V_{F}\frac{1}{1+[\partial, V_{F}]}(A),\qquad R_{\rho}(A)=(1-\rho)V_{F}\frac{1}{1+[\partial, V_{F}]}(A).$$
\begin{lem}[\cite{li2013primitive}]
$[\bar{\p}_{F}, R_{\rho}]=1-T_{\rho}$ on $\mathcal{A}^*(\C^n)$. Moreover, the embedding $(\mathcal{A}_{c}^*(\C^n),\bar{\p}_{F})\hookrightarrow (\mathcal{A}^*(\C^n),\bar{\p}_{F})$ is a quasi-isomorphism.
\end{lem}
Let $A$ be a representative of $C^{\infty}(M)\otimes\Omega^n_{\C^n\times M/M}/dF\wedge\Omega^{n-1}_{\C^n\times M/M}$, then $T_{\rho}(A)$ is $L^2$-integrable, thus we can define $w$ to be its harmonic projection, that is,
$$w=\Pi_u(T_{\rho}(A))=(\mbox{Id}-G\Delta_{F})T_{\rho}(A)=A-\bar{\p}_{F}R_{\rho}(A)-\bar{\p}_{F}G\bar{\p}_{F}^{\dag}T_{\rho}(A).$$
For simplicity, we write it as $w=A+\bar{\p}_{F}\nu$, where, by construction, $\nu$ is a $n-1$ form with polynomial growth. We define the following map
\begin{align*}
S:C^{\infty}(M)\otimes\Omega^n_{\C^n\times M/M}/dF\wedge\Omega^{n-1}_{\C^n\times M/M}&\longrightarrow \mathcal{H},\\
A&\longmapsto w.
\end{align*}
\begin{prop}\label{spectral} The map $S$ satisfies the following properties:
\begin{enumerate}
\item $S$ is a $C^{\infty}(M)$-linear map;
\item $S$ is well defined, that is, $S$ is independent of the choice of the representative in $C^{\infty}(M)\otimes\Omega^n_{X\times M/M}/dF\wedge\Omega^{n-1}_{X\times M/M}$ and also independent of the cut-off function $\rho$.
\end{enumerate}
\end{prop}
\begin{proof} \
\begin{enumerate}[(1)]
\item The $C^{\infty}(M)$-linearity follows from the definitions of $T_{\rho}$ and $\Pi_u$.
\item
\begin{itemize}
\item The independence of the choice of the representative:
Let $A'=A+dF\wedge B$, for some $B\in C^{\infty}(M)\otimes\Omega_{\C^n\times M/M}^{n-1}$ with polynomial growth. By definition, there exist $n-1$ forms $\nu$ and $\nu'$ with polynomial growth such that
$$w:=S(A)=A+\bar{\p}_F\nu,\quad w':=S(A')=A'+\bar{\p}_F\nu'.$$
Moreover, since $B$ is holomorphic in the $\C^n$ direction,
$$w':=S(A')=A+dF\wedge B+\bar{\p}_F\nu'=A+\bar{\p}_F(B+\nu').$$
%Let $w$ and $w'$ be the image of $A$ in the above way corresponding to the cut-off functions $\rho$ and $\rho'$, respectively, i.e.
%$$A=w-\bar{\p}_{F(\cdot,u)}\nu=w'-\bar{\p}_{F(\cdot,u)}\nu'.$$
Then one has
\[w-w'=\bar{\p}_{F(\cdot,u)}(\nu+B-\nu'),\]
where $(\nu+B-\nu')$ has polynomial growth. Then by Agmon estimate (Lemma \ref{epde12}) and integration by parts, one has
\[h(w-w',w-w')=h(w-w',\bar{\p}_{F(\cdot,u)}(\nu-\nu'))=h(\bar{\p}_{F(\cdot,u)}^{\dagger}(w-w'),\nu-\nu')=0,\]
which implies that $w=w'$.\\
\item The independence of the choice of the cut-off function:
Let $S$ and $S'$ be the map with respect to the cut-off functions $\rho$ and $\rho'$ respectively. Then
$$A=S(A)-\bar{\p}_{F(\cdot,u)}\nu=S'(A)-\bar{\p}_{F(\cdot,u)}\nu'$$ for some differential forms $\nu$ and $\nu'$ with polynomial growth.
Using the same argument as above, one can easily verify that $S(A)=S'(A)$.
\end{itemize}
\end{enumerate}
\end{proof}
\begin{rem} In the physical literature, such as in \cite{CECOTTI1991N}, $S$ is called the spectral flow.
\end{rem}
Let $u=(u_1,...,u_s)$ be local coordinates of $M$, and $\partial_i:=\partial_{u_i}$.
If $A$ is holomorphic in $u\in M$, then $w:=S(A)=A+\bar{\p}_F\nu$ is holomorphic. Since
$$\bap_{\bar{i}}w=\bap_{\bar{i}}A+\bap_{\bar{i}}\bar{\p}_{F}\nu=\bar{\p}_F\bap_{\bar{i}}\nu\in\Im(\bar{\p}_{F})\quad (\mbox{Note that $[\bap_{\bar{i}}, \bar{\p}_{F}]=0$}).$$
In particular, when $|u|$ is small, we can simply choose
$$A_a=\phi_a dz_1\wedge\cdots\wedge dz_n, \quad \{\phi_a\}_{a=0,\ldots,\mu-1}\text{ is a monomial basis of}\Jac(f).$$
Moreover, $\phi_0=1$. Then $A_a$'s are $\bar{\p}_F$-closed and one can find a unique harmonic form $w_a=S(A_a)$ such that
\begin{equation}\label{dec}
\phi_a dz_1\wedge\ldots\wedge dz_n=w_a-\bar{\p}_F\nu_a
\end{equation}
for some $n-1$ form $\nu_a$, $a=0,\ldots,\mu-1$. Here $\nu_a$ has at most polynomial growth.
Thus, we get a holomorphic basis $\{w_a\}$ for $\H^n$.
Now we would like to explore the relationship between the operator $C$ and the product in the Jacobi ring of $F$:
On the one hand, by the definition, $\partial_{i}Fw_{a}=C_{ia}^{b}w_{b}+\bar{\partial}_{F}\nu$ for some differential forms $\nu$ with polynomial growth. On the other hand,
\begin{align*}
\partial_{i}Fw_{a}&=\partial_{i}F(A_{a}+\bar{\partial}_{F}\nu_a)\\
&=\partial_{i}FA_{a}+\bar{\partial}_{F}\partial_iF\nu_a\\
%&=(\widetilde{C}_{ia}^{b}A_{b}+df\wedge\delta_{a})+\bar{\p}_{\b f}R_{\rho}(C_{ia}^{b}A_{b}+df\wedge\delta_{a})+\bar{\p}_{\b f}\nu\\
&\overset{(a)}=(\widetilde{C}_{ia}^{b}A_{b}+dF\wedge\delta_{a})+\bar{\partial}_{F}\nu_a\\
&\overset{(b)}=\widetilde{C}_{ia}^{b}\left(w_{b}-\bar{\partial}_{F}\nu_b+\bar{\partial}_{F}\delta_a\right)+\bar{\partial}_{F}\nu_a=\widetilde{C}_{ia}^{b}w_{b}+\bar{\partial}_{F}\widetilde{\nu},
\end{align*}
where in the equality (a), $\partial_{i}FA_{a}=\widetilde{C}_{ia}^{b}A_{b}+dF\wedge\delta_{a}$ for some $\delta_{a}\in\Omega_{X\times M/M}^{n-1}$ and $\widetilde{C}_{ia}^{b}\in \mathcal{O}_{M}$, then in the equality (b), notice that $dF\wedge\delta_{a}=\bar{\partial}_{F}\delta_{a}$.
%Using the Hodge decomposition{\color{red}To be more clear, not the Hodge decomposition, since $\tilde{v}$ is not in $L^2$, we can say
By the similar arguments as in Proposition \ref{spectral}, we get $C_{ia}^{b}=\widetilde{C}_{ia}^{b}$, i.e., the operator $C$ coincides with the product in the Jacobi ring $\Jac(F)$.%{\color{red} This sentence looks weird, since "correspond" is an intransitive verb}.
%\begin{prop}[\cite{Fan2016pairings}]\label{etares}
%With the construction above, there exists a constant $C_{n}$, depending only on $n=\dim X$, such that
%$$\eta(\alpha_{a},\alpha_{b})=C_{n}\Res(A_{a},A_{b}).$$
%\end{prop}
\subsection{$U(1)$ action}\label{2.4}
To further study the structure of the Hodge bundle, let us introduce the $U(1)$ action associated to $f$ on differential forms, which is closely related to the pole order filtration in Section 1.
First, assume that the weight $q_i=\frac{a_i}{b_i}$ with $(a_i,b_i)=1$, and $d=\mathrm{lcm}\left(b_{1}, \ldots, b_{n}\right)$, i.e. $d$ is the least common multiple of $b_1,...,b_n$. We put $Q_{i}=q_{i} d$.
Let us consider the $U(1)$ action on $\mathbb{C}^n$ which is related to the weight of the variables:
$$e^{i\theta}\cdot(z_1,\ldots,z_n)=\left(e^{iQ_1\theta}z_1,\ldots,e^{iQ_n\theta}z_n\right).$$
In particular, if $f$ is homogeneous,
$$e^{i\theta }\cdot(z_1,\ldots,z_n)=\left(e^{i\theta }z_1,\ldots,e^{i\theta }z_n\right).$$
%Since $\deg(f)=d$, we have
%$$e^{i\theta}\cdot f=e^{-id\theta} f.$$
This action induces a unitary action $\T$ of $U(1)$ on differential forms, such that for any $(p,q)$ form $\alpha=\alpha(z_1,...,z_n,\bar{z}_1,...\bar{z}_n)dz_{i_1}...dz_{i_p}d\bar{z}_{j_1}...d\bar{z}_{j_q}$, $$\T(\theta)\alpha=e^{i\theta\left(-\sum_{k=1}^pQ_{i_k}+\sum_{k=1}^q{Q_{j_k}}\right)}\alpha(e^{-i\theta Q_1}z_1,...,e^{-i\theta Q_n}z_n,e^{i\theta Q_1}\bar{z}_1,...,e^{i\theta Q_n}\bar{z}_n)dz_{i_1}...dz_{i_p}d\bar{z}_{j_1}...d\bar{z}_{j_q}.$$
Note that $f$ is a $(0,0)$-form, thus,
$$\T(\theta)f=e^{-id\theta}f.$$
Similarly, if $f$ is homogeneous,
$$\T(\theta)\alpha=e^{i(q-p)\theta}\alpha(e^{-i\theta}z_1,...,e^{-i\theta}z_n,e^{i\theta}\bar{z}_1,...,e^{i\theta}\bar{z}_n)dz_{i_1}...dz_{i_p}d\bar{z}_{j_1}...d\bar{z}_{j_q}.$$
Now consider the following shifted $U(1)$ action $\P$ on $L^2\Lambda^{p,q}(\C^n)$, $$\P(\theta)(\alpha)=e^{idp\theta}\T(\theta)\alpha,\quad \alpha\in L^2\Lambda^{p,q}(\C^n).$$
Similarly, we have another shifted $U(1)$ action $\Q$ on $L^2\Lambda^{p,q}(\C^n)$,
$$\Q(\theta):=e^{idq\theta}\T(-\theta).$$
Immediately, we have,
\begin{lem}\label{Atheta} The $U(1)$ actions $\P$ and $\Q$ satisfy the following properties:
\begin{enumerate}[(1)]
\item $\P$ and $\Q$ are unitary, i.e.,
$$h(\Q(\theta)\alpha,\Q(\theta)\beta)=h(\P(\theta)\alpha,\P(\theta)\beta)=h(\alpha,\beta), \quad \alpha,\beta\in L^2\Lambda^{p,q}(\C^n).$$
\item \label{sf6}$[\P(\theta),\bar{\partial}_F]=0,~[\P(\theta),\bar{\partial}_F^{\dagger}]=0,~[\P(\theta),\Delta_F]=0$.
\item $\P(\theta){\partial}_F\P(\theta)^{-1}=e^{id\theta}{\partial}_F,$ $\P(\theta){\partial}_F^{\dagger}\P(\theta)^{-1}=e^{-id\theta}{\partial}_F^{\dagger}.$
\item\label{sf7} $[\Q(\theta),{\partial}_F]=0,~[\Q(\theta),{\partial}_F^{\dagger}]=0,~[\Q(\theta),\Delta_F]=0$.
\item $\Q(\theta)\bar{\partial}_F\Q(\theta)^{-1}=e^{id\theta}\bar{\partial}_F,$ $\P(\theta)\bar{\partial}_F^{\dag}\P(\theta)^{-1}=e^{-id\theta}\bar{\partial}_F^{\dagger}.$
\end{enumerate}
In particular, (\ref{sf6}) and (\ref{sf7}) imply that $\P:\H\to\H$, $\Q:\H\to\H.$
\end{lem}
\begin{rem} Let $X$ be a compact Calabi-Yau manifold, one can define the $U(1)$ actions $\P^{CY}$ and $\Q^{CY}$ as follows, for $\alpha\in \A^{p,q}(X)$,
$$\P^{CY}(\theta)\alpha:=e^{ip\theta}\alpha,\quad \Q^{CY}(\theta)\alpha:=e^{iq\theta}\alpha.$$
It is easy to see that $\P^{CY}$ and $\Q^{CY}$ are unitary; $\P^{CY}$ commutes with $\bar{\partial}$; $\P^{CY}(\theta) \p\P^{CY}(\theta)^{-1}=e^{i\theta}\p.$
Thus one can regard $\P$ and $\Q$ as the bi-grading in the Landau-Ginzburg models. More explicitly, the infinitesimal action of $\P$ and $\Q$ plays the role of holomorphic and anti-holomorphic degrees in Landau-Ginzburg models.
\end{rem}
\begin{defn} For $A=z_1^{\beta_1}\cdots z_n^{\beta_n}dz_1\wedge\cdots\wedge dz_n$, we define
$$l(A):=\sum_{i=1}^n(\beta_i+1)Q_i.$$
We also call $l(A)$ the $U(1)$ charge of $A$.
\end{defn}
For the homogeneous basis $\{A_a\}_{a=0}^{\mu-1}$, set $l_a=l(A_a).$
\begin{lem}\label{uone} The $U(1)$ actions $\P$ and $\Q$ commute with the map $S$, that is
$$S(\P(\theta)A)=\P(\theta)(S(A)),\quad A\in C^{\infty}(M)\otimes\Omega^n_{\C^n\times M/M}/dF\wedge\Omega^{n-1}_{\C^n\times M/M}$$
and
$$S(\Q(\theta)A)=\Q(\theta)(S(A)),\quad A\in C^{\infty}(M)\otimes\Omega^n_{\C^n\times M/M}/dF\wedge\Omega^{n-1}_{\C^n\times M/M}.$$
More explicitly, let $w_a=S(A_a)$, then
\begin{align*}
\P(\theta)(w_a)&=e^{i(nd-l_a)\theta}w_a,\\
\Q(\theta)(w_a)&=e^{il_a\theta}w_a.
\end{align*}
Thus, we could define the $U(1)$ charge of $w_a$ to be $l_a$. This lemma could be rephrased as follows: if a harmonic form $w$ has $U(1)$ charges $l$, that is, $w=S(A)$ for some homogeneous $n$ form $A$ with $l(A)=l$, then
\begin{align*}
\P(\theta)(w)&=e^{i(nd-l)\theta}w,\\
\Q(\theta)(w)&=e^{il\theta}w.
\end{align*}
\end{lem}
\begin{proof} It suffices to check the result for $A_a=\phi_adz_1\wedge\cdots\wedge dz_n$. On the one hand,
\begin{equation}\label{sf4}\P(\theta)A_a=e^{i(nd-l_a)\theta}A_a.\end{equation}
On the other hand, since
\begin{equation} \label{sf1}
A_a=S(A_a)-\bar{\p}_{F}\nu_a,
\end{equation}
by Lemma \ref{Atheta}, we have
\begin{equation}\label{sf2}
\P(\theta)A_a=\P(\theta)S(A_a)-\bar{\p}_F(\P(\theta)\nu_a);
\end{equation}
By \eqref{sf4}, the left hand side of \eqref{sf2} is
\begin{equation}\label{sf3}
\P(\theta)A_a=e^{i(nd-l_a)\theta}A_a=e^{i(nd-l_a)\theta}S(A_a)-\bar{\p}_F\left(e^{i(nd-l_a)\theta}\nu_a\right).
\end{equation}
Then (\ref{sf2}) and (\ref{sf3}) imply that
$$e^{i(nd-l_a)\theta}S(A_a)-\P(\theta)S(A_a)=\bar{\p}_F\mu_a$$
for some at most polynomial growth differential form $\mu_a$.
By Agmon estimate (Lemma \ref{epde12}) and the fact that $[\P(\theta),\Delta_F]=0$ , we get
$$\P(\theta)S(A_a)=e^{i(nd-l_a)\theta}S(A_a)=S(\P(\theta)A_a).$$
\end{proof}
\def\lan{\langle}
\def\ran{\rangle}
\def\dz{dz_1...dz_n}
%\begin{rem}
% If $f$ contains a unique isolated singularity at the origin, the strong nullstellensatz implies the existence of a sufficiently large $N$ such that $z_i^N\in\lan \partial_{z_1}f,...,\partial_{z_n}f\ran$ for all $i=1,2,...,n$. As a result, if $\phi$ is homogeneous with a sufficient large degree $D$, $\phi \in\lan \partial_{z_1}f,...,\partial_{z_n}f\ran$. In fact, Lemma \ref{Atheta} could provide an explicit description of $D.$ To keep things simple, we will consider only the case where $f$ satisfies the Calabi-Yau condition. A harmonic form's $U(1)$ charge cannot exceed $n(n-1)$ in this case. Thus, if $w=S(\phi \dz)$ for some homogeneous polynomial $\phi$ of degree $n(n-2)+1$, $w$ must have a $U(1)$ charge of $n(n-1)+1$. As a result, $w$ should be trivial, implying that $\phi\in \lan \partial_{z_1}f,...,\partial_{z_n}f\ran. $ As a result, we have $D=n(n-2)+1$ in this case.
%\end{rem}
In CY's case, if $u\in \A^{p,n-p}(Y),v\in \A^{p',n-p'}(Y)$, then $\int_Yu\wedge*\bar{v}\neq0$ iff $p=p'$. Similarly,
by unitary property of $\P$ and Lemma \ref{uone}, it is easy to obtain
\begin{cor}\label{wawb} If $h(w_a, w_b)\neq0$, then $l_a=l_b$.
\end{cor}
\begin{proof}
By Lemma \ref{uone},
\[h(w_a,w_b)=h(\P(\theta) w_a,\P(\theta) w_b)=h(e^{i(nd-l_a)\theta}w_a,e^{i(nd-l_b)\theta}w_b)=e^{i(l_b-l_a)\theta}h(w_a,w_b).\]
Hence, $h(w_a,w_b)\neq 0$ implies $l_a=l_b.$
\end{proof}
Now we investigate the relation between $U(1)$ charges and the real structures in the Landau-Ginzburg B-model. Since $w_a$ is a basis of $\mathcal{H}^n$, there exists a matrix-valued function $(M_{\bar{a}b}(u,\bar{u}))$ such that
%Since the $tt^*$ metric is defined through a real structure (complex conjugate on harmonic forms), now let us discuss the relation of $U(1)$ charge within the real structure.{\color{red}This sentence looks weird. Maybe just compare with CY's case, and then say similarly, in LG's case, one has.....}
%By using the isomorphism $S$, we can induce a real structure{\color{red}(not a real str indeed, just a map)} on $C^{\infty}(M)\otimes\Omega^n_{X\times M/M}/dF\wedge\Omega^{n-1}_{X\times M/M}$ from the one on the harmonic forms. If we write{\color{red}Why M exists? To be more clear, we should indicate that $\{w_a\}$ is a basis}
$$\tau_R(w_a):=\bar{w}_a=\sum_bM_{\bar{a}b}(u,\bar{u})w_b,$$
then we can define a complex anti-linear map
$$\kappa:C^{\infty}(M)\otimes\Omega^n_{\C^n\times M/M}/dF\wedge\Omega^{n-1}_{\C^n\times M/M}\longrightarrow C^{\infty}(M)\otimes\Omega^n_{\C^n\times M/M}/dF\wedge\Omega^{n-1}_{\C^n\times M/M}$$ via the basis
\begin{equation}\label{realform}
\kappa(A_a):=\sum_bM_{\bar{a}b}(u,\bar{u}) A_b.
\end{equation}
Since $S$ is $C^{\infty}(M)$-linear, we have
\begin{equation}\label{taukappa}
S(\kappa(A_a))=\sum_bM_{\bar{a}b}(u,\bar{u})S(A_b)=\sum_bM_{\bar{a}b}(u,\bar{u})w_b=\tau_R(S(A_a)).
\end{equation}
Moreover, we have
\begin{prop}\label{conjcharge} Suppose $A\in C^{\infty}(M)\otimes\Omega^n_{\C^n\times M/M}/dF\wedge\Omega^{n-1}_{\C^n\times M/M}$ has (quasi-)homogeneous charge $l(A)$, then $\kappa(A)$ is also (quasi-)homogeneous with charge $l(\kappa(A))$. Furthermore,
$$l(A)+l(\kappa(A))=nd.$$
\end{prop}
\begin{proof}
It suffices to prove it for $\{A_a\}_{a=0}^{\mu-1}$. Let $w_a=S(A_a)$, i.e.
$$w_a=A_a+\bar{\p}_F\nu_a \quad \text{for some $(n-1)$-form }\nu_a \text{ (with polynomial growth)}.$$
Then
\begin{equation}\label{bara}
\bar{w}_a=\bar{A}_a+\p_F\bar{\nu}_a.
\end{equation}
First by Lemma \ref{Atheta}, (\ref{bara}), and $\P(\theta)\p_F=e^{i\theta}\p_F\P(\theta)$, proceed as what we did in Lemma \ref{uone}, one has \begin{equation}\label{bara1}\P(\theta)\bar{w}_a=e^{il_a\theta}\bar{w}_a=e^{il_a\theta}\sum_bM_{\bar{a}b}w_b.
\end{equation}
On the other hand,
$$\bar{w}_a=\kappa(A_a)+\bar{\p}_F\tilde{\nu}_a \quad \text{for some $(n-1)$-form } \tilde{\nu}_a \text{ (with polynomial growth)},$$
and $\kappa(A_a)=\sum_bM_{\bar{a}b}(u,\bar{u})A_b$, we have
\begin{equation}\label{bara2}
\P(\theta)\bar{w}_a=\sum_be^{i(nd-l_b)\theta}M_{\bar{a}b}w_b.
\end{equation}
(\ref{bara1})-(\ref{bara2}), one has
\[\sum_b(e^{i(nd-l_b)\theta}-e^{il_a\theta})M_{\bar{a}b}w_b=0.\]
Since $\{w_a\}$ is a basis, one has $M_{\bar{a}b}=0$ if $nd-l_b\neq l_a$, which implies that $\kappa(A_a)$ is homogeneous of degree $nd-l_a.$
\end{proof}
%\begin{proof}
%{\color{red}I think my orginal proof is better, since it's not clear to see that $\kappa(A)$ is homogeneous. Then delete the next Cor and write it down somewhere in the statement of this prop.}
%First, we have
%$$w=S(A)=A+\bar{\p}_F\nu \quad \text{for some $(n-1)$-form }\nu \text{ with polynomial growth}.$$
%Then
%$$\bar{w}=\bar{A}+\p_F\bar{\nu}.$$
%By Lemma \ref{Atheta} (3), that is, $\P(\theta)\p_F=e^{id\theta}\p_F\P(\theta)$, proceed as what we did in Lemma \ref{uone}, one has
%$$\P(\theta)\bar{w}=e^{il(A)}\bar{w}.$$
%
%On the one hand, by \eqref{taukappa},
%$$\bar{w}=\tau_R(S(A))=S(\kappa(A)).$$
%It means that $\kappa(A)$ is homogeneous and
%$$\P(\theta)\kappa(A)=e^{il(A)\theta}\kappa(A).$$
%On the other hand, by Lemma \ref{uone},
%$$\P(\theta)\kappa(A)=e^{i(nd-l(\kappa(A)))\theta}\kappa(A).$$
%Thus, $\kappa(A)$ has homogeneous degree $l(\kappa(A))=nd-l(A)$.
%\end{proof}
%Applying this lemma to the basis $\{A_a\}$, we get the following result for the matrix $M(u,\bar{u})$.
%\begin{cor} Let $\{A_a\}_{a=0}^{\mu-1}$ be the homogeneous $C^{\infty}(M)$-basis as above, then
%$$\kappa(A_a)=\sum_{b,l_b+l_a=nd}M_a^b(u,\bar{u})A_b.$$
%\end{cor}
\begin{rem}\label{n-2}
If $f$ contains a unique isolated singularity at the origin, the strong nullstellensatz implies the existence of a sufficiently large $N$ such that $z_i^N\in\lan \partial_{z_1}f,...,\partial_{z_n}f\ran$ for all $i=1,2,...,n$. As a result, if $\phi$ is homogeneous with a sufficient large degree $K$, $\phi \in\lan \partial_{z_1}f,...,\partial_{z_n}f\ran$. In fact, Lemma \ref{Atheta} could provide an explicit description of $K$. To keep things simple, we will consider only the case where $f$ is homogeneous of degree $n$. A harmonic form's $U(1)$ charge cannot exceed $n(n-1)$ in this case. Thus, if $w=S(\phi \dz)$ for some homogeneous polynomial $\phi$ of degree $n(n-2)+1$, $w$ must have a $U(1)$ charge of $n(n-1)+1$. As a result, $w$ should be trivial, implying that $\phi\in \lan \partial_{z_1}f,...,\partial_{z_n}f\ran. $ As a result, we have $K=n(n-2)+1$ in this case.
\end{rem}
Now suppose $f\in\C[z_1,\ldots,z_n]$ is homogeneous of degree $n$. Then the above analysis implies that:
\begin{prop-defn}[small $tt^*$ structure in the Landau-Ginburg B models \cite{Fan2020LGCYCB}]\label{smallLG}
Let $f$ be a non-degenerate homogeneous degree $n$ polynomial and $F$ be its marginal deformation with parameter space $M$. Let $\H'\subset\H$ be the subbundle generated by $w_{k_a}=S(A_{k_a})$, where $l(A_{k_a})/n\in\mathbb{Z}.$ By Proposition \ref{conjcharge}, restriction of $\tau_R,h,D,C,\bar{C}$ to $\H'$ defines a $tt^*$ structure, called small $tt^*$ geometry structure in the Landau-Ginzburg B models.
\end{prop-defn}
\begin{proof}
Let $u=(u_1,...,u_s)$ be local coordinates of $M$, and $\p_i=\p_{u_i} $.
The restriction gives a well defined $tt^*$ structure, since
\begin{enumerate}
\item $D_i$ preserves the $U(1)$ charge of each $w_a$:\\%{{\color{red} what's the meaning of U(1) charges of a harmonic form? We should explain it somewhere}}: \\
Since $w_a$ is holomorphic, i.e. $\bar{D}_{\bar{i}}w_a=0$, we have
$$\p_ih(w_a,w_b)=h(D_iw_a,w_b)+h(w_a,\bar{D}_{\bar{i}}w_b)=h(D_iw_a,w_b).$$
By the non-degeneracy of $h$ and Corollary \ref{wawb}, we can see that $D_iw_a$ and $w_a$ have the same $U(1)$ charge.
\item $C_i$ increases the $U(1)$ charge by $n$, since the operator $C_i$ is induced from multiplication by $\psi_i$ in $\Jac(F)$, where $\psi_i$ has the same $U(1)$ charge $n$ as $f$.
\item $\mathcal{H}'$ is stable under $\tau_R$, by Proposition \ref{conjcharge}. Combining with (1) and (2), it implies that $\mathcal{H}'$ is stable under $\bar{D}_{\bar{i}}$ and $\bar{C}_{\bar{i}}$.
\end{enumerate}
%First, Proposition \ref{conjcharge} implies that the subbundle $\mathcal{H}'$ is stable under $\tau_R$. {\color{red}Next, by the relation between $C$ and the product in the Jacobi ring $\Jac(F)$ and the fact that the deformation is marginal}, $\mathcal{H}'$ is stable under $C$ and $\bar{C}$. Combining with Corollary \ref{wawb}, and the fact that $\bar{C}_{\bar{i}}$ is the adjoint operator of $C_i$ with respect to the $tt^*$ metric $h$, one can easily prove that $\mathcal{H}'$ is stable under $D_i$, $\bar{D}_{\bar{i}}$ and $\bar{C}_{\bar{i}}${\color{red}(We should say $D$ preserve the degree instead)}.
\end{proof}
It follows from the above analysis that one can also endow the following filtration on $\mathcal{H}'$
$$\F^0\H'\subset\F^1\H'\subset\cdots\subset \F^{n-2}\H'=\H'$$
with the gradation $\F^{k}\H'/\F^{k-1}\H'$ generated by
$$\{w_{a}=S(A_a)~|~l_a=n(k+1)\}.$$
It follows from the definition that under the isomorphism $S$, such filtration is the same as the pole order filtration on the rational differential forms after a degree shift.
The above analysis implies that $\nabla^{(1,0)}=D+C$ satisfies the Griffiths transversality.
Next, we denote $\Jac(F)'$ to be the subspace of $\Jac(F)$ that generated by the monomial basis $\phi$, with $n|\deg(\phi)$. We have,
$$\Jac(F)'\stackrel{\cong}{\longrightarrow}\mathcal{H}',\quad \phi\mapsto S(\phi dz_1\wedge\cdots\wedge dz_n).$$
Denote $\dim\Jac(F)'=\mu'$. Moreover, we let $\phi_0=1$.
\subsection{The vacuum line bundle and the Weil-Peterson metric}
In this subsection, we assume that $f$ is homogeneous of degree $n$. Let us introduce the vacuum line bundle and Weil-Peterson metric in the LG models. Recall that $A_0=dz_1\wedge\cdots\wedge dz_n$ is the unique representative of lowest $U(1)$-charges $l_0=n$. As before, set $w_0=S(A_0)$ and $h_{0\bar{0}}=h(w_0,w_0)$.
Let $u=(u_1,...,u_s)$ be local coordinates of $M$, and $\p_{i}=\p_{u_i} $.
By Corollary \ref{wawb}, we have
\begin{align*}
&\p_ih(w_0,w_0)=h(D_iw_0,w_0)+h(w_0,\bar{D}_{\bar{i}}w_0)=h(D_iw_0,w_0),\\
&0=\p_ih(w_0,w_a)=h(D_iw_0,w_a)+h(w_0,\bar{D}_{\bar{i}}w_a)=h(D_iw_0,w_a),\quad a>0.
\end{align*}
Hence, we have
\begin{equation}\label{diazero}
D_i w_0=\left(h_{\bar{0}0}^{-1}\p_ih_{0\bar{0}}\right)w_0,
\end{equation}
which means that the line subbundle $\mathcal{L}$ of $\H^n$ generated by $w_0$ has Chern connection $D+\bar{D}$. It is called the vacuum line bundle in the physical literature.
The Weil-Peterson metric is given by
$$G_{i\bar{j}}=-\p_i\bar{\p}_{\bar{j}}\log h_{0\bar{0}}.$$
\begin{thm} The Weil-Peterson metric $G_{i\bar{j}}$ is related to the $tt^*$ metric as follows:
$$G_{i\bar{j}}=\frac{h(C_iw_0,C_jw_0)}{h(w_0,w_0)}.$$
\end{thm}
\begin{proof}
First, since $w_0=S(dz_1\wedge\cdots\wedge dz_n)$ has the lowest $U(1)$ charge and $C_i$ increases the $U(1)$ charge by $n$, by a similar argument as in Lemma \ref{uone}
\[h(w_0,C_iw_a)=0.\]
While $h(\bar{C}_{\bar{i}}w_0,w_a)=h(w_0,C_iw_a)$
implies that $\bar{C}_{\bar{i}}w_0=0.$
Then,
\begin{align*}
\bar{\p}_{\bar{j}}\p_i\log h(w_0,w_0)&=-\frac{h(\bar{D}_{\bar{j}}D_iw_0,w_0)+h( D_iw_0,D_jw_0)}{h(w_0,w_0)}+\frac{h( D_iw_0,w_0)h(w_0,D_jw_0)}{h(w_0,w_0)^2}\\
&=-\frac{h([\bar{D}_{\bar{j}},D_i]w_0,w_0)}{h(w_0,w_0)}-h_{\bar{0}0}^{-1}\p_ih_{0\bar{0}}\cdot\overline{h_{\bar{0}0}^{-1}\p_jh_{0\bar{0}}}+ h_{\bar{0}0}^{-1}\p_ih_{0\bar{0}}\cdot\overline{h_{\bar{0}0}^{-1}\p_jh_{0\bar{0}}}\\
&=-\frac{h([C_i,\bar{C}_{\bar{j}}]w_0,w_0)}{h(w_0,w_0)}\\
&=\frac{h( C_iw_0, C_jw_0)}{h(w_0,w_0)}.
%=-\frac{h(w_i,w_j)}{h(w_0,w_0)}.
%&=g_{i\bar{j}}/g_{0\bar{0}}.
\end{align*}
\end{proof}
\section{Lefschetz thimble, Riemann Bilinear Formula and Period Integral}
In this section, we express the $tt^*$ metric in terms of period integrals using the Riemann bilinear relation. Next, we present an explicit construction of the Lefschetz thimble in order to study the period integral. Lastly, we investigate two types of period integrals.
\subsection{Lefschetz thimble and intersection matrix}
Let $a>0$, and consider the sets
$$F^{\geq a}=\{z\in\C^n\big|\Re(F(z,u))\geq a\},\quad F^{\leq -a}=\{z\in\C^n\big|\Re(F(z,u))\leq-a\}.$$
By Morse theory, we know that if there are no critical values of $\Re(F(z,u))$ between $[a,b]$, then the set $F^{\leq a}$ is the deformation kernel of $F^{\leq b}$ by the flow generated by the vector field $\nabla\Re(F(z,u))/|\nabla\Re(F(z,u))|$. So if $a$ is large enough, there are no critical points in $F^{\leq-a}$ and each $F^{\leq-a}$ has the same homotopy type. We denote the equivalence class by $F^{-\infty}$. Similarly, we have $F^{+\infty}$. Then we can consider the relative homology groups $H_*(\C^n, F^{-\infty};\Z)$ and $H_*(\C^n,F^{+\infty};\Z)$.
\begin{thm} $H_k(\C^n,F^{-\infty};\Z)$ is trivial except for $k=n$. $H_n(\C^n,F^{-\infty};\Z)$ is a free abelian group, and it is generated by the Lefschetz thimble $\{\gamma_k^-\}$. The same holds true for $H_*(\C^n, F^{+\infty};\Z)$.
\end{thm}
\begin{proof} See, for example, \cite{Fan2020LGCYCB}.
\end{proof}
Before moving on, let us give an explicit construction of the Lefschetz thimble $\gamma_k^-\in H_n(\C^n,F^{-\infty};\Z)$.
Let $V_t$ be the level manifold $V_t=F^{-1}(t)$, then the hypersurface $X_{F}\subset\mathbb{P}^{n-1}$ could be regarded as the complement component of the projection of $V_t$:
Consider $G(z_1,\ldots, z_n,z_{n+1})=F-tz_{n+1}^d$, $\bar{V}_t:=G^{-1}(0)\subset\mathbb{P}^n$, then $$X_F=\bar{V}_t-V_t.$$
%{\color{red}I don't think write it on the foot note is a good idea.}
%\textcolor{blue}{give the definition of $V_t$ and $V_{\infty}$}
Let $c_t(s):=te^{2\pi i s}$, $0\leq s\leq1$, then $c_t$ induces a monodromy operator
$$M_{c_t}: H_{n-1}(V_t)\to H_{n-1}(V_t).$$
Following \cite{Fan2020LGCYCB}, we fix a basis $\{\sigma_k\}_{k=0}^{\mu-1}$ of $H_{n-1}(V_{-1})$, such that $\sigma_{k}\in\ker(M_{c_t}-\Id)$ for $0\leq k\leq\mu'-1$. Hence, there exists $\delta_k\in H_{n-2}(X_F)_0:=(H^{n-2}(X_F)_{\prim})^*$, such that $\sigma_k=\tau(\delta_k)$ for $0\leq k\leq\mu'-1$ (c.f., dualizing (3.54) in \cite{CECOTTI1991N}), where $\tau: H_{n-2}(X_F)\to H_{n-1}(\bar{V}_{-1}-X_F)=H_{n-1}(V_{-1})$ is the Leray coboundary map. For $t>0$, let $\Phi_t:V_{-t}\to V_{-1}$ be the map $(z_1,..,z_n)\to(t^{-\frac{1}{n}}z_1,...,t^{-\frac{1}{n}}z_n)$, and set
\[\gamma_k^-:=\cup_{t>0}(\Phi_t)^*\sigma_k,\]
then one can see that $\{\gamma_k^-\}_{k=0}^{\mu-1}$ is a basis of $H_n(\C^n,F^{-\infty};\mathbb{Z})$.
Similarly, one can construct a basis $\{\gamma_k^+\}_{k=0}^{\mu-1}$ of $H_n(\C^n,F^{+\infty};\mathbb{Z}).$
Now we would like to define intersection matrix between $H_n(\C^n,F^{-\infty};\mathbb{Z})$ and $H_n(\C^n,F^{+\infty};\mathbb{Z})$.
First note that
$$d_F=\bar{\p}_F+\p_F=e^{-F-\bar{F}}\circ d\circ e^{F+\bar{F}},$$
then for any $\Delta_F$-harmonic form $w$, the modified forms
$$e^{F+\bar{F}}w\quad\text{and}\quad e^{-F-\bar{F}}* w$$
are $d$-closed. In particular, they give the representatives of $H^n(\C^n, F^{-\infty};\C)$ and $H^n(\C^n, F^{+\infty};\C)$, respectively.
For each $\gamma_k^-\in H_n(\C^n,F^{-\infty};\mathbb{Z})$, the map
$$w\mapsto \int_{\gamma_k^-}e^{F+\bar{F}}w,\quad w\in \ker{\Delta_{F(\cdot,u)}}$$
is linear. Hence by Riesz representation, there exists $\alpha_k\in \ker{\Delta_{F(\cdot,u)}}$, such that
\[\int_{\gamma_k^-}e^{F+\bar{F}}w=\int_{\C^n}w\wedge*\alpha_k=\int_{\C^n}e^{F+\bar{F}}w\wedge e^{-F-\bar{F}}*\alpha_k.\]
Then set
$$PD(\gamma_k^-):=e^{-F-\bar{F}}*\alpha_k\in H^n(\C^n,F^{+\infty};\C).$$
Similarly, we can define $PD(\gamma_k^+),$ such that
\[\int_{\gamma_k^+}e^{-F-\bar{F}}*w=\int_{\C^n}PD(\gamma^+_k)\wedge e^{-F-\bar{F}}*w\]
for all $w\in\ker{\Delta_{F(\cdot,u)}}.$
\begin{defn}\label{intermatrix}
The intersection matrix $I=(I_{kl})_{\mu\times\mu}$ is defined by $$I_{kl}=\int_{\gamma_k^-}PD(\gamma_l^+)=\int_{\C^n}PD({\gamma}_l^+)\wedge*PD({\gamma}_k^-).$$
for $0\leq k,l\leq \mu-1.$
While $I'$ is a submatrix of $I$, that is $I_{kl}'=I_{kl}$ for $0\leq k,l\leq \mu'-1.$
Now set $\I=I^{-1},\I'=(I')^{-1}.$
\end{defn}
By definition, the forms $PD(\gamma_k^-)$ and $PD(\gamma_k^+)$ depend on $u$, however,
\begin{prop}
The intersection matrix $I$ is locally constant, i.e. when $|u|$ is small, $\p I=\bar{\p} I=0.$
\end{prop}
\begin{proof}
Suppose $w$ is $\Delta_{F(\cdot,u)}$-harmonic, then if $|u|$ is small, by Agmon estimate (Lemma \ref{epde12}),
$$w':=e^{(F(\cdot,u)-f)+\overline{F(\cdot,u)-f}}w$$
has exponential decay and is $d_f$-closed. Then proceed as what we did in Section 7.5 of \cite{DY2020cohomology}, we can find $\nu$, s.t. $w'-\Pi_0 w'=d_f\nu$, and $\nu$ has exponential decay. Here $\Pi_0$ is the projection $L^2\to \ker{\Delta_f}$. Hence,
\[\int_{\gamma_k^-}e^{F+\bar{F}}w=\int_{\gamma_k^-}e^{f+\bar{f}}w'=\int_{\gamma_k^-}e^{f+\bar{f}}\Pi_0w'.\]
Again, as what we did in Section 7.5 of \cite{DY2020cohomology},
$$\Pi_0\circ e^{(F(\cdot,u)-f)+\overline{F(\cdot,u)-f}}:H_{(2)}^*(\C^n,d_{F(\cdot,u)})\to H_{(2)}^*(\C^n,d_f)$$
is isomorphic, for the similar reason, one can see that $PD(\gamma_k^-)(u,\bar{u})-PD(\gamma_k^-)(0)=d \mu$ for some differential form $\mu$ with exponential decay on $\Re(f)<0.$
Hence, $I$ is locally constant.
\end{proof}
By the definition of intersection matrix, the following Riemann bilinear formula is straightforward:
\begin{prop}\label{riebil}
If $w$ and $w'$ are harmonic, then
$$\int_{\C^n}w\wedge*w'=\sum_{k,l}\left(\int_{\gamma_k^-}e^{F+\bar{F}}w\right)(\I)_{kl}\left(\int_{\gamma_l^+}e^{-F-\bar{F}}w'\right).$$
\end{prop}
%Actually, one also have
%\begin{prop}\label{riebil1}
%If $w$ is harmonic,
% $$\int_{\C^n}e^{-\bar{F}}A_a\wedge*w=\sum_{k,l}\left(\int_{\gamma_k^-}e^{F}A_a\right)(\I)_{kl}\left(\int_{\gamma_l^+}e^{-F-\bar{F}}w\right).$$
%\end{prop}
%\begin{proof}
%
%
%
%It suffices to prove the case of $u=0$.
%
%Choose a smooth function $\tilde\eta_R \in C^\infty(\R)$, such that
%\begin{itemize}
% \item $\tilde\eta_R\equiv 0$ when $|x|\leq R$,
% \item $\tilde\eta_R(x)=1$ when $|x|\geq R+1$.
%\end{itemize}
%Let
%$$\eta_{R}(z):=\tilde\eta_R(|z|).$$
%
%Fix $C_0>0$, s.t. $2\rho(z)\leq C_0|z|^n$, where $\rho$ is the Agmon distance with respect to $2|\nabla\Re(f)|^2g$. Since $\rho(z)\geq 2|\Re(f)|$, one also have $$C_0|z|^n\geq 4|\Re(f)|.$$
%
%
%
%Then for any $\epsilon>0$, there exists an $R_0$, s.t. whenever $R\geq R_0$,
%\begin{align}
%\left|\int_{\C^n}e^{-C_0\eta_R|z|^n}e^{-\bar{f}}A_a\wedge*w-\int_{\C^n}e^{-\bar{f}}A_a\wedge*w\right|\leq \epsilon,\label{cpt1}\\
%\left|\int_{\gamma_k^-}e^{-C_0\eta_R|z|^n}e^{f}A_a-\int_{\gamma_k^-}e^{f}A_a\right|\leq \epsilon.\label{cpt2}
%\end{align}
%
%Moreover, by the construction of $C_0$, $e^{-C_0\eta_R|z|^n}e^{-\bar{f}}A_a$ is $L^2$-integrable. Thus, it's reasonable to consider the Kodaira-Hodge decomposition
%\begin{equation}\label{kodaira}
%e^{-C_0\eta_R|z|^n}e^{-\bar{f}}A_a=w_{R}+d_{2\Re(f)}\alpha+d_{2\Re(f)}^*\beta,
%\end{equation}
%where $w_{R}$ is harmonic.
%
%Moreover, we can choose $\alpha\in \Im(d_{2\Re(f)}^{\dagger})\cap\ker(\Delta_{2\Re(f)})^\perp,\beta\in \Im(d_{2\Re(f)})\cap\ker(\Delta_{2\Re(f)})^\perp.$
%
%Then one can see that, by Proposition \ref{riebil}
%\begin{equation}\label{cpt3}
%\int_{\C^n}e^{-C_0\eta_R|z|^n}e^{-\bar{f}}A_a\wedge*w=\int_{\C^n}w_{R}\wedge*w=\sum_{k,l}\left(\int_{\gamma_k^-}e^{f+\bar{f}}w_{R}\right)(\I)_{kl}\left(\int_{\gamma_l^+}e^{-f-\bar{f}}w\right).
%\end{equation}
%
%Now we claim that
%\begin{equation}\label{claim51}
%\left(\int_{\gamma_k^-}e^{f+\bar{f}}d_{2\Re(f)}\alpha\right)=0
%\end{equation}
%and
%\begin{equation}\label{claim52}
%\left(\int_{\gamma_k^-}e^{f+\bar{f}}d_{2\Re(f)}^*\beta\right)<\epsilon\quad \text{if $R$ is big enough.}
%\end{equation}
%
%Hence, by (\ref{cpt1}), (\ref{cpt2}), (\ref{cpt3}), (\ref{claim51}), and (\ref{claim52}), for any $\epsilon>0$, we have
%
%\[\left|\int_{\C^n}e^{-\bar{f}}A_a\wedge*w-\sum_{k,l}\left(\int_{\gamma_k^-}e^{f}A_a\right)(\I)_{kl}\left(\int_{\gamma_l^+}e^{-f-\bar{f}}w\right)\right|<C\epsilon.\]
%
%Let $\epsilon\to0$, then we are done.
%
%Now it suffices to prove the claim:\\
%By (\ref{kodaira}) and our choice of $\alpha$,
%\begin{align*}d_{2\Re(f)}^*e^{-C_0\eta_R|z|^n}e^{-\bar{f}}A_a&=d_{2\Re(f)}^*(w_{R}+d_{2\Re(f)}\alpha+d_{2\Re(f)}^*\beta)\\
%&=d_{2\Re(f)}^*d_{2\Re(f)}\alpha=\Delta_{2\Re(f)}\alpha,\end{align*}
%
%By Lemma 7.2 in \cite{DY2020cohomology} and construction of $C_0$, $\alpha$ has exponential decay. By Stoke formula, we have (\ref{claim51}).
%
%By (\ref{kodaira}) and our choice of $\beta$,
%\begin{equation} \label{cptbeta}d_{2\Re(f)}e^{-C_0\eta_R|z|^n}e^{-\bar{f}}A_a=\Delta_{2\Re(f)}\beta,\end{equation}
%
%Notice that $d_{2\Re(f)}e^{-C_0\eta_R|z|^n}e^{-\bar{f}}A_a$ has support inside $B_R^c:=\{z:|z|\geq R\}$. Hence, one can see easily that
%\[\|d_{2\Re(f)}e^{-C_0\eta_R|z|^n}e^{-\bar{f}}A_a\|_{L^2(\C^n)}=\|d_{2\Re(f)}e^{-C_0\eta_R|z|^n}e^{-\bar{f}}A_a\|_{L^2(B_R^c}\leq Ce^{-\frac{C_0R^n}{2}}.\]
%
%Since $\beta\in \ker(\Delta_{2\Re(f)})^{\perp}$, there exists $c>0$, s.t.
%\[\|\beta\|_{L^2}\leq c\|\Delta_{2\Re(f)} \beta\|_{L^2}\leq c\|d_{2\Re(f)}e^{-C_0\eta_R|z|^n}e^{-\bar{f}}A_a\|_{L^2} \leq Ce^{-\frac{C_0R^n}{2}}.\]
%By Lemma 7.2 in \cite{DY2020cohomology},
%\[|\beta|\leq Ce^{-a\rho}e^{-\frac{C_0R^n}{2}}\]
%for some $a\in(0,1), C>0$.
%
%Hence, we have (\ref{claim52}).
%
%
%
%\end{proof}
\subsection{Two types of integrals: $\int_{\gamma^-}e^{F}A $ vs $\int_{\gamma^-}e^{F+\bar{F}}w$}\label{periodintegral}
This subsection is dedicated to compare the two kinds of period integrals in the LG B-model
$$\int_{\gamma^-}e^{F}A \quad\text{and}\quad\int_{\gamma^-}e^{F+\bar{F}}w,$$
i.e., to prove Theorem \ref{lgcy1} (or Theorem \ref{132} below).
%First note that
%$$d_F=\bar{\p}_F+\p_F=e^{-F-\bar{F}}\circ d\circ e^{F+\bar{F}},$$
%then for any $\Delta_F$-harmonic form $w$, the modified forms
%$$e^{F+\bar{F}}w\quad\text{and}\quad e^{-F-\bar{F}}* w$$
%are $d$-closed. In particular, they give the representatives of $H^n(\C^n, F^{-\infty};\C)$ and $H^n(\C^n, F^{+\infty};\C)$, respectively.
Let $w_a$ be the holomorphic basis constructed from $A_a$. We will study
$$\A_{ka}=\int_{\gamma_k^-}e^{F+\bar{F}}w_a, \quad \tilde{\A}_{ka}=\int_{\gamma_k^+}e^{-F-\bar{F}}* w_a.$$
%It is easy to check that the period integrals serve as the flat section of Gauss-Manin connections.
%\begin{prop} The matrix-valued functions $\A=(\A_{ka}), \widetilde{\A}=(\widetilde{\A_{ka}})$ are invertible, and
%$$(D+\bar{D}-C-\bar{C})\A=0,\quad (D+\bar{D}+C+\bar{C})\widetilde{\A}=0.$$
%\end{prop}
On the other hand,
$$e^FA_a\quad \text{and}\quad e^{-F}* A_a$$
are also $d$-closed $n$-forms, which also give a basis of $H^n(\C^n, F^{-\infty};\C)$ and $H^n(\C^n, F^{+\infty};\C)$ respectively. We can also consider the following integral
$$\B_{ka}=\int_{\gamma_k^-}e^FA_a,\quad \tilde{\B}_{ka}=\int_{\gamma_k^+}e^{-F}* A_a.$$
Considering the pairing of these two types of integrals, the following Riemann bilinear relations are established.
\begin{prop}\label{riebil1}
If $w$ is harmonic, then
$$\int_{\C^n}e^{-\bar{F}}A_a\wedge*w=\sum_{k,l}\left(\int_{\gamma_k^-}e^{F}A_a\right)(\I)_{kl}\left(\int_{\gamma_l^+}e^{-F-\bar{F}}w\right).$$
\end{prop}
We will prove this proposition in the appendix.
Since $\A$ and $\widetilde{\A}$ are invertible for $u\in M$, there exist matrix-valued functions $\T$ and $\widetilde{\T}$, such that $\B=\T\A$ and $\widetilde\B=\widetilde\T\tilde\A.$ More specifically,
\[\int_{\gamma_k^-}e^{F}A_a=\sum_a(\T)_{ab}\int_{\gamma_k^-}e^{F+\bar{F}}w_b\].
%Moreover, we have the following structure theorem for the matrix-valued functions $\T$ and $\widetilde{\T}$.
\begin{thm}
%\label{lgcy1}
\label{132}
$\T$ and $\tilde{\T}$ are (block) unit lower triangular matrix. More explicitly,
\begin{align*}
(\T)_{aa}&=1 \mbox{ for~ $0\leq i\leq \mu-1$};\\
(\T)_{ab}&=0 \mbox{ if~ $a<b$};\\
(\T)_{ab}&=\begin{cases}
0, &\mbox{ if~ $\frac{l(A_a)-l(A_b)}{n}\notin \mathbb{Z}^+$};\\
\frac{(-1)^{l_{ab}}}{l_{ab}!}\int_{\C^n}\bar{f}^{l_{ab}}A_a\wedge*\bar{w}_b, &\mbox{ if~ $l_{ab}:=\frac{l(A_a)-l(A_b)}{n}\in \mathbb{Z}^+$.}
\end{cases}
\end{align*}
%(We will see that by Agmon estimate, $\int_{\C^n}\bar{f}^{l_{ab}}A_a\wedge*\bar{w}_b<\infty.$ )
In particular,
\[\int_{\gamma_k^-}e^{F+\bar{F}}w_0=\int_{\gamma_k^-}e^Fdz_1\wedge\cdots\wedge dz_n,\]
\[\int_{\gamma_k^+}e^{-F-\bar{F}}*w_0=\int_{\gamma_k^+}e^{-F}*dz_1\wedge\cdots\wedge dz_n,\]
\end{thm}
\begin{proof}
By Riemann bilinear formula, on the one hand,
\begin{equation}\label{rieb1}
\left(\int_{\gamma_k^-}e^{F}A_a\right)\I_{kl}\overline{\left(\int_{\gamma_l^+}e^{-F-\bar{F}}*{w}_c\right)}=\int_{\C^n}e^{-\bar{F}}A_a\wedge*\overline{w_c}=h(e^{-\bar{F}}A_a,w_c).
\end{equation}
On the other hand,
\begin{equation}\label{rieb2}
\begin{aligned}
&\int_{\gamma_k^-}e^FA_a\I^{kl}\overline{\left(\int_{\gamma_l^+}e^{-F-\bar{F}}* w_c\right)}\\
=~&\left(\sum_b\T_{ab}\int_{\gamma_k^-}e^{F+\bar{F}}w_b\right)\I_{kl}\overline{\left(\int_{\gamma_l^+}e^{-F-\bar{F}}* w_c\right)}\\
=~&\sum_b\T_{ab}\int_{\C^n}w_b\wedge*\overline{w_c}\\
=~&\sum_{b:l_b=l_c}\T_{ab}h(w_b,w_c).
\end{aligned}
\end{equation}
%\begin{equation}\label{rieb2}\left(\int_{\gamma_k}e^{F}A_a\right)\I_{kl}\overline{\left(\int_{\tilde{\gamma}_l}e^{-F-\bar{F}}*w_c\right)}=\sum_{b:l(A_c)=l(A_b)}(\T)_{ab}h(w_b,w_c).
%\end{equation}
Let $t=e^{i\theta}$, then
\begin{equation} \label{htheta1}h(e^{-\bar{F}}A_a,w_c)=h(\P(\theta)e^{-\bar{F}}A_a,\P(\theta)w_c)=t^{l_c-l_a}h(e^{-t^n\bar{F}}A_a,w_c).
\end{equation}
\begin{itemize}
\item If $\frac{l_c-l_a}{n}\notin \Z$, then take $t$ to be the $n$-th root of $1$, such that $t^{l_c-l_a}\neq1$, by (\ref{htheta1}), we see that
$$h(e^{-\bar{F}}A_a,w_c)=0.$$
This together with (\ref{rieb1}), (\ref{rieb2}) and the fact that the matrix $(h(w_b,w_c))_{l_b=l_c}$ is invertible imply that
$$(\T)_{ab}=0,\quad\text{if }\frac{l_a-l_b}{n}\notin\Z.$$
\item If $l_c-l_a=nl_{ca}$ for some integer $l_{ca}$, then we consider the function
$$v_{ac}(s)=h(e^{-s\bar{F}}A_a,w_c).$$
By Lemma \ref{Agmonhol} below, it is holomorphic in $s$ when $|s|<2$. Then
\begin{itemize}
\item if $l_{ca}>0$: when $|s|=1$, by (\ref{htheta1}) $s^{l_{ca}}v_{ac}(s)=v_{ac}(1)$, which implies $v_{ac}(s)=s^{-l_{ca}}v_{ac}(1)$ when $|s|=1$. However, by the properties of holomorphic function, we must have $v_{ac}(s)=s^{-l_{ca}}v_{ac}(1)$ when $0<|s|<2.$ Since $v_{ac}$ is holomorphic function on $|s|<2$, we must have $v_{ac}(1)=h(e^{-\bar{F}}A_a,w_c)=0.$
$$\T_{ab}=0,\quad \mbox{ if } l_b=l_c.$$
\item if $l_{ca}\leq0$: if follows from the same argument as above that $v_{ac}(s)=s^{-l_{ca}}v_{ac}(1)$ when $|s|<2$. Applying $\left(\frac{d}{ds}\right)^{-l_{ca}}\bigg|_{s=0}$ to both sides, we obtain
%\[v_{ac}(1)=\frac{1}{(-l_{ca})!}\left(\frac{d}{ds}\right)^{-l_{ca}}|_{s=0}v_{ac}(s),\]
$$h(e^{-\bar{F}}A_a,w_c)=\frac{1}{(-l_{ca})!}\int_{\C^n}\bar{F}^{-l_{ca}}A_a\wedge*\bar{w}_c.$$
In particular, when $l_{ca}=0$, $$h(e^{-\bar{F}}A_a,w_c)=h(A_a,w_c)=h(w_a,w_c)-h(\bar{\p}_{F(\cdot,u)}\nu_a,w_c)=h(w_a,w_c).$$
\end{itemize}
\end{itemize}
As a result, Theorem \ref{132} follows.
\end{proof}
\begin{lem}\label{Agmonhol}
The function $v_{ac}(s)=\int_{\C^n}e^{-s\bar{F}}A_a\wedge*\bar{w}_c$ is holomorphic on the disk $|s|<2.$
\end{lem}
\begin{proof}
Without loss of generality, it suffices to prove it when $u=0$. Note that $$\{\bar{\p}_{f},\bar{\p}_{f}^{\dagger}\}=\frac{1}{2}\{d_{2\Re(f)},d_{2\Re(f)}^{\dagger}\}.$$
By Agmon estimate, there exists $C_b>0$ for any $b\in(0,1)$, such that
\[|w_c|\leq C_be^{-b\rho}\|w_c\|_{L^2},\]
where $\rho(z)$ is the Agmon distance between $z$ and $0$ with respect to Agmon metric $2|\nabla \Re(f)|^2g_0$ (See appendix for more details). It follows from the properties of Agmon distance and holomorphic functions that $\rho(z)\geq 2|\Re(e^{i\theta}f)|$ for any $\theta\in\R$. As a result, for any $a<b$,
if $|s|\leq 2a$,
\begin{align*}
\int_{\C^n}e^{-s\bar{f}}A_a\wedge*\bar{w}_c&\leq C\int_{\C^n} e^{2a\Re(f)}e^{-b\rho}\mathrm{dvol}_{\C^n}\leq C\int_{\C^n} e^{-(b-a)\rho}\mathrm{dvol}_{\C^n}<\infty.
\end{align*}
Hence $v_{ac}(s)$ is holomorphic when $|s|<2.$
\end{proof}
%\section{Example}
%
%The simplest case: the period domain $D$ is a symmetric domain (It is related to the central charge $c$). If $\sum_iq_i\in\mathbb{N}$, $\phi$ is onto and hence we can identify $\widetilde{S}$ with $D$. That is, up to a change of coordinates in coupling constant, $\phi$ is the identity.
%
%
%Compute the WP metric for the example $X^3+Y^3+Z^3+bXYZ$.
%
%Let us first recall the period integral of $A_n$ type, i.e. $f=x^{n+1}$. The computation of $X^3+Y^3+Z^3$ will be related to the $A_2$ type via the Fubini Theorem of oscillatory integral.
%
%Let $f=x^{n+1}$, then $0$ is the unique critical point of multiplicity $n$. Let $\epsilon_1,\ldots,\epsilon_{n+1}$ be the unit root of $x^{n+1}=-1$, i.e.
%$$\epsilon_j=e^{\frac{\pi i+2j\pi i}{n+1}},\quad j=1,\ldots,n+1.$$
%Let $\Gamma_j$ be the segment from $\epsilon_{j+1}$ to $\epsilon_j$, and $A_j=x^{j-1}dx$, then
%\begin{lem}[\cite{AGVsingularities}] For $1\leq j,l\leq n$,
%$$\int_{[\Gamma_j]}e^{\tau f}A_l=\tau^{-\frac{l}{n+1}}\left(\epsilon_{j+1}^l-\epsilon_j^l\right)\frac{\Gamma(l/(n+1))}{n+1}.$$
%\end{lem}
%
%The integral of the form $\int_{[\Gamma_j]}e^{\tau f}A_l$ with $l\geq n+1$ can be reduced to the above computation via the integration by parts.
%\begin{cor} For $q\leq j\leq n$, $l\geq1$,
%$$\int_{[\Gamma_j]}e^{\tau f}A_l=.$$
%\end{cor}
%
%
%Let $f: (\C^l,0)\rightarrow(\C,0)$ and $g:(\C^m,0)\rightarrow(\C,0)$ be two functions holomorphic at the origin, and have at the origin a critical point of finite multiplicity. Let $\Gamma_1\subset\C^l$, $\Gamma_2\subset\C^m$ be admissible chains, respectively for the critical points of $f$ and $g$. Let us suppose that
%\begin{align*}
%&\Re f\bigg|_{\partial \Gamma_1}=\Re g\bigg|_{\partial\Gamma_2}=-a\\
%&a/2>\Re f\bigg|_{\Gamma_1},\Re g\bigg|_{\Gamma_2}\geq-a
%\end{align*}
%for a sufficiently small $a>0$. Then the chain
%$$\Gamma_1\times\Gamma_2\subset\C^{l+m}$$
%is the admissible chain for the critical points of $f+g:(\C^{l+m},0)\rightarrow(\C,0)$.
%
%
%
%\begin{lem}[Fubini Theorem \cite{AGVsingularities}]\label{Funibi} Let $f,g$ and $\Gamma_1$, $\Gamma_2$ be as above. Let $\omega$ be a holomorphic $l$-form given in the neighborhood of the origin in $\C^l$ and $\eta$ be a holomorphic $m$-form given in the neighborhood of the origin in $\C^m$, then
%\begin{equation}\label{Fubini}
%\int_{\Gamma_1\times\Gamma_2}e^{\tau (f+g)}\omega\wedge \eta=\int_{\Gamma_1}e^{\tau f}\omega \int_{\Gamma_2}e^{\tau g}\eta.
%\end{equation}
%\end{lem}
%
%
%
%Now let us return to our case, we want to compute $\int_{\Gamma_j}e^{X^3+Y^3+Z^3+uXYZ}dX\wedge dY\wedge dZ$, which for small $u$ reduces to the computation of
%$$\int_{\Gamma}e^{X^3+Y^3+Z^3}(XYZ)^kdX\wedge dY\wedge dZ,\quad k\geq0.$$
%By the above Fubini Theorem, it furthermore reduces to compute the computation of $A_2$ type
%$$\int_{\Gamma}e^{x^3}x^kdx.$$
\section{Calabi-Yau/Landau-Ginzburg Correspondence}
In this section, we assume that
$$f:\C^n\rightarrow\C\quad n\geq3$$
is a non-degenerate homogeneous polynomial of degree $n$. Regarding $[z_1,z_2,\ldots,z_n]$ as the homogeneous coordinate in $\mathbb{P}^{n-1}$, the zero set of $f$, denoted by $X_{f}$, defines a Calabi-Yau hypersurface in $\mathbb{P}^{n-1}$.
Now consider the marginal deformation $F=f+\sum_{i=1}^s u_i\psi_i$ of $f$, then for $|u|$ small enough, $F$ determines a family of Calabi-Yau hypersurface $X_{F(z,u)}$ parametrized by $u\in M$, which gives a deformation of $X_f$.
\subsection{Residue map}
\def\res{\operatorname{res}}
For each $\phi_a\in \Jac(F)'$, then $\frac{l_a}{n}\in \Z^+$. Set $n_a=\frac{l_a}{n}$, Let
$$\Omega:=\sum_{i=1}^n(-1)^{i+1}z_{i} d z_{1} \wedge \ldots \widehat{dz_i}\ldots \wedge d z_{n},$$
and denote $\Omega_a:=\frac{\phi_a\Omega}{F(z,u)^{n_a}}.$ Then one can see that $\Omega_a\in H^{n-1}(\mathbb{P}^{n-1}-X_{F(z,u)})$.
Let $\tau: H_{n-2}(X_{F(z,u)})\to H_{n-1}(\mathbb{P}^{n-1}-X_{F(z,u)})$ be the Leray coboundary map, one can define the residue map $\res:H^{n-1}(\mathbb{P}^{n-1}-X_{F(z,u)})\to H^{n-2}(X_{F(z,u)})$ by
\[2\pi i\int_{\delta}\res(\Omega_a)=\int_{\tau(\delta)}\Omega_a.\]
It was shown that (see, for example, \cite{carlson1980infinitesimal})
\begin{prop} The residue map
$$\res: H^{n-1}(\mathbb{P}^{n-1}-X_{F(z,u)},\mathbb{C})\longrightarrow H^{n-2}(X_{F(z,u)},\C)$$
is injective, with the image being the $(n-2)$-th primitive cohomology of $X_{F(z,u)}$. Moreover, the residue map preserves the Hodge structure. That is,
\begin{equation}\label{fil}\res(\Omega_a)\in \F^{n-n_a-1}H^{n-2}(X_{F(z,u)})_{\prim}.
\end{equation}
%where $H^{n-2}(X_f)_{\prim}$ denotes the primitive form in $H^{n-2}(X_f)$.
In particular, $\res(\Omega_0)$ is a nowhere vanishing holomorphic $(n-2,0)$ form on $X_{F(z,u)}$.
\end{prop}
%To discuss the Calabi-Yau/Landau-Ginzburg correspondence in our setup, let us introduce the following def
\begin{defn}\label{resmap}
We define the map $R_u:\Jac(F)'\to H^{n-2}(X_F)$ via
\[R_u(\phi_a):=\res\left(\frac{\phi_a\Omega}{F^{n_a}}\right).\]
In particular, $R_u(1)$ is a nowhere vanishing holomorphic $(n-2,0)$ form on $X_F$.
\end{defn}
%Recall that the Hodge filtration on $H^{n-1}(\mathbb{CP}^{n-1}-X_F,\C)$ is induced by the trivial filtration on the $C^{\infty}$ logarithm de Rham complex (fine resolution of holomorphic logarithm de Rham complex), thus is also induced by
%$$F^p\Omega_{\mathbb{CP}^{n-1}}^*(\log X_F)=\{0\rightarrow \Omega_{\mathbb{CP}^{n-1}}^p(\log X_F)\rightarrow \Omega_{\mathbb{CP}^{n-1}}^{p+1}(\log X_F)\rightarrow\cdots\}.$$
%The induced Hodge filtration is the limit filtration in the collapsed spectral sequence:
%$$E_1^{p,q}=H^q(\mathbb{CP}^{n-1},\Omega_{\mathbb{CP}^{n-1}}^q(\log X_F))~\Rightarrow~ H^{p+q}(\mathbb{CP}^{n-1}-X_F,\mathbb{C}).$$
%By Deligne [?], $E_1=E_{\infty}$.
%\subsection{The correspondence between the chiral ring structure: cup product and the pairing}
\subsection{The correspondence between the period integral}
By Theorem \ref{132}, we also have
\begin{equation}\label{haref}\int_{\gamma_k^-}e^{F+\bar{F}}w_a=\int_{\gamma_k^-}\left(e^FA_a-\sum_{b:l_b<l_a}(\T)_{ab}(u,\bar{u})e^FA_b\right).\end{equation}
\begin{defn}
We define the map $r_u:\H'\to H^{n-2}(X_F)_{\prim}$ via
\[r_u(w_a)=2\pi i\left((-1)^{n_a-1}(n_a-1)!R_u(\phi_a)-\sum_{b:l_b<l_a}(\T)_{ab}(u,\bar{u})(-1)^{n_b-1}(n_b-1)!R_u(\phi_b)\right),\]
where $X_F:=\{[z]\in\C P^{n-1}:F(z,u)=0\}.$
\end{defn}
\begin{rem}\label{rsha}
By (\ref{fil}), one can see that if $l(A)=l$ and $\frac{l}{n}\in\Z$, then
$$r_u(S(A))\in \F^{n-1-\frac{l}{n}}H^{n-2}(X_F)_{\prim}.$$
\end{rem}
\begin{lem}\label{ruan}
For $0\leq k\leq \mu'-1$, $\phi_a\in\Jac(F)'$, $A_a=\phi_adz_1\wedge\cdots\wedge dz_n$,
\[\int_{\gamma_k^-}e^{F}A_a=2\pi i(-1)^{n_a-1}(n_a-1)!\int_{\delta_k}R_u(\phi_a),\]
and
\[\int_{\gamma_k^+}e^{-F}*A_a=2\pi i(-1)^{n_a}(n_a-1)!\int_{\delta_k}*R_u(\phi_a).\]
\end{lem}
\begin{proof}
It suffices to compute at $u=0.$
Let $\sigma_k(t)=(\Phi_T)^*\sigma_k\in H_n(V_{-t}).$ For each $t\neq 0$, let
$$\tilde{c}_t(\theta):=-t+\frac{te^{i\theta}}{2},\quad T(\sigma_k)(t):=\cup_{\theta}P_{\theta}(\sigma_k(-\frac{t}{2})),$$
where $P_{\theta}: H_{n}(V_{-\frac{t}{2}})\to H_n(V_{\tilde{c}_t(\theta)})$ is the parallel transport along $\tilde{c}_t.$
Then
\begin{align}
\begin{split}
\int_{\gamma_k^-}e^{f}A_a&=\int_0^\infty e^{-t}\int_{\sigma_k(t)}\frac{A_a}{df}dt\\
&=\frac{1}{2\pi i}\int_0^\infty e^{-t}\int_{T(\sigma_k(t))}\frac{A_a}{f+t}dt.
\end{split}
\end{align}
Let $v(t)=\int_{T(\sigma_k(t))}\frac{A_a}{f+t},$ then $v(t)=t^{n_a-1}v(1)$ (recall that $n_a=\frac{l_a}{n}$), proceed as in the proof of Lemma A.2 in \cite{chiodo2014landau}, and take the differential of $t$, one has
\[v(t)=(-1)^{n_a-1}t^{n_a-1}\int_{T(\sigma_k(t))}\frac{A_a}{(f+t)^{n_a}}\]
It follows from Theorem 4.2 in \cite{Fan2020LGCYCB} that
\[\int_{T(\sigma_k(t))}\frac{A_a}{(f+t)^{n_a}}=\lim_{t\to 0}\int_{T(\sigma_k(t))}\frac{A_a}{(f+t)^{n_a}}=(2\pi i)^2\int_{\delta_k}R_0(\phi_a).\]
\end{proof}
\begin{prop}\label{rpqdegree}
For $0\leq k\leq\mu'-1$, one has
\begin{equation}\label{periodlgcy}\int_{\gamma_k^-}e^{F+\bar{F}}w=\int_{\delta_k} r_u(w)\end{equation}
and \begin{equation}\label{periodlgcy0}\int_{\gamma_k^+}e^{-F-\bar{F}}*w=\int_{\delta_k} *r_u(w)\end{equation}
for any $w\in\H'.$
Moreover, $r_u(w_a)\in H^{n-n_a-1,n_a-1}(X_F)_{\prim}$ and $r_u(\bar{w}_a)=\overline{r_u(w_a)}$ represent the same class in $X_F.$
\end{prop}
\begin{proof}
By Lemma \ref{ruan} and (\ref{haref}), for $0\leq k\leq\mu'-1$, any $w\in\H'$,
\begin{equation}\label{ruan11}
\int_{\gamma_k^-}e^{F+\bar{F}}w=\int_{\delta_k} r_u(w).
\end{equation}
As a result,
\begin{equation*}
\int_{\gamma_k^-}e^{F+\bar{F}}\bar{w}_a={\int_{\delta_k} r_u(\bar{w}_a)}.
\end{equation*}
Moreover,
\begin{equation*}
\int_{\gamma_k^-}e^{F+\bar{F}}\bar{w}_a=\overline{\int_{\gamma_k^-}e^{F+\bar{F}}{w}_a}=\overline{\int_{\delta_k} r_u({w}_a)}={\int_{\delta_k} \overline{r_u({w}_a)}}.
\end{equation*}
Thus, \begin{equation}\label{ruan1}r_u(\bar{w}_a)=\overline{r_u(w_a)}\in \overline{\F^{n-n_a-1}H^{n-2}(X_F)_{\prim}},\end{equation}
where $n_a:=\frac{l_a}{n}.$
One the other hand, since $\bar{w}_a=S(\kappa(\phi_adz_1\wedge\cdots\wedge dz_n))$, by Remark \ref{rsha},
$$r_u(\bar{w}_a)\in \F^{n_a-1}H^{n-2}(X_F)_{\prim}.$$
As a result,
\begin{equation} \label{rbwa}
r_u(\bar{w}_a)\in \overline{\F^{n-n_a-1}H^{n-2}(X_F)_{\prim}}\cap \F^{n_a-1}H^{n-2}(X_F)_{\prim}=H^{n_a-1,n-n_a-1}(X_F)_{\prim}.
\end{equation}
(\ref{ruan1}) and (\ref{rbwa}) tell us that $r_u(w_a)\in H^{n-n_a-1,n_a-1}(X_F)_{\prim}.$
\end{proof}
\begin{thm}\label{fanlan}
If $\phi_a,\phi_b\in \Jac(f)'$, then
$$\int_{\C^n}w_a\wedge*w_b=\frac{(2\pi)^2}{2^{n-2}}\int_{X_F}r_u(w_a)\wedge*r_u(w_b).$$
\end{thm}
\begin{proof}
By Lemma \ref{wawb} and Lemma \ref{conjcharge}, if $l_a+l_b\neq n^2$, we have
$$\int_{\C^n}w_a\wedge*w_b=0.$$
While by Proposition \ref{rpqdegree}, if $l_a+l_b\neq n^2$, we also have
$$\int_{X_F}r_u(w_a)\wedge*r_u(w_b)=0.$$
Hence it suffices to prove the case of $l_a+l_b=n^2$. Since
$$r_u(w_a)=(2\pi i)(-1)^{n_a-1}(n_a-1)!R_u(\phi_a)^{n-n_a-1,n_a-1},$$
(by (\ref{fil}), Proposition \ref{rpqdegree} and the definition of $r_u$ and $R_u$), by Theorem 3.4 in \cite{shen2016explicit} and Proposition \ref{carlson},
\begin{align*}\int_{\C^n}w_a\wedge*w_b&=\frac{(-1)^{(n_a-1)(n_b-1)+1}(n_a-1)!(n_b-1)!}{2^{n-2}}\int_{X_F}R_u(\phi_a)^{n-n_a-1,n_a-1}\wedge*R_u(\phi_b)^{n-n_b-1,n_b-1}\\
&=\frac{(2\pi)^2}{2^{n-2}}\int_{X_F} r_u(w_a)\wedge*r_u(w_b).\end{align*}
\end{proof}
%\textcolor{blue}{Maybe think about rewriting this proof.}
\subsection{The correspondence for intersection matrices}
Let $(I)_{kl}$ be the intersection matrix, $\mathcal{I}=I^{-1}$; $I_{kl}'$ be a submatrix of ${I}$, and $\mathcal{I}'=(I')^{-1};$
$(I^{CY})_{kl}=\delta_{k}\cap\delta_l$ be the intersection matrix, and $\mathcal{I}^{CY}=(I^{CY})^{-1}.$
If $\phi_a\in \Jac(F)'$, for $k\geq\mu'$,
\[\int_{\gamma_k^-}e^{F}A_a=\int_0^\infty e^{-t}\int_{\sigma_k}\frac{A_a}{dF},\]
where $\frac{A_a}{dF}$ is the Gelfand-Leray form (c.f. Lemma 10.3 in \cite{arnold2012singularities}).
Integration by substitution tells us that
$$\int_{\sigma_k}\frac{A_a}{dF}=0,\quad \text{if } M_{c_t}\sigma_k\neq\sigma_k,$$
which implies if $k> \mu'$
\begin{equation}\label{monovan}\int_{\gamma_k^-}e^{f}A_a=0\end{equation}
\begin{thm}\label{intersec}
$\I'=\frac{\pi^2}{2^{n-4}}\I^{CY}.$
\end{thm}
\begin{proof}
On the one hand, by (\ref{monovan}) and Riemann bilinear formula,
\begin{align*}
\int_{\C^n}w_a\wedge*w_b&=\left(\int_{\gamma_k^-}e^{F+\bar{F}}w_a\right)(\I')_{kl}\left(\int_{\gamma_l^+}e^{-F-\bar{F}}*w_b\right)
\\
&=\left(\int_{\delta_k}r_u(w_a)\right)(\I')_{kl}\left(\int_{{\delta}_l}*r_u(w_b)\right)\mbox{ (By (\ref{haref}) and Lemma \ref{ruan})}.
\end{align*}
On the other hand, by Theorem \ref{fanlan},
\begin{align*}
\int_{\C^n}w_a\wedge*w_b&=\frac{\pi^2}{2^{n-4}}\int_{X_F}r_u(w_a)\wedge*r_u(w_b)\\
&=\frac{\pi^2}{2^{n-4}}\left(\int_{\delta_k}r_u(w_a)\right)(\I^{CY})_{kl}\left(\int_{{\delta}_l}*r_u(w_b)\right).
\end{align*}
Hence, $\I'=\frac{\pi^2}{2^{n-4}}\I^{CY}.$
\end{proof}
%\begin{lem}\label{wef}
% If $l_a+l_b< n^2$, then
% \[h(w_a,e^{F}\bar{A}_b)=\int_{\C^n}w_a\wedge* e^{-\bar{F}}A_b=0.\]
% If $l_a+l_b=n^2$, then
%\[h(w_a,e^{F}\bar{A}_b)=h(w_a,\bar{A}_b)=h(w_a,\bar{w}_b)=\eta(w_a,w_b).\]
%\end{lem}
%\begin{proof}
%Let $t=e^{i\theta}$, then
%\begin{align*}
%h(w_a,e^{-F}\bar{A}_b)&=h(\Q(\theta)w_a,\Q(\theta)e^{-F}\bar{A}_b)\\
%&=h(t^{l_a}w_a,e^{-t^nF}t^{n^2-l_b}\bar{A}_b)\\
% &=t^{(l_a+l_b-n^2)}h(w_a,e^{-t^nF}\bar{A}_b).
%\end{align*}
%Proceed as in the proof of Theorem \ref{}, the Lemma follows.
%\end{proof}
\begin{proof}[Proof of Theorem \ref{133}]
By Theorem \ref{intersec} and Lemma \ref{ruan}, it's easy to see that
\[\int_{\C^n}w_0\wedge*\bar{w}_0=\frac{\pi^2}{2^{n-4}}\int_{X_F}r_u(w_0)\wedge *\overline{r_u(w_0)}.\]
Then $G_{i\bar{j}}=G_{i\bar{j}}^{CY}$ follows from the definition
$$G_{i\bar{j}}=-\p_i\bar{\p}_{\bar{j}}\log(\int_{\C^n}w_0\wedge*\bar{w}_0),\quad G^{CY}=-\p_i\bar{\p}_{\bar{j}}\log(\int_{X_F}r_u(w_0)\wedge *\overline{r_u(w_0)}).$$
\end{proof}
To show Calabi-Yau/Landau-Ginzburg correspondence for $tt^*$ geometry, we first show that:
\begin{lem}\label{gmc}
Let $u=(u_1,...,u_s)$ be local coordinates of $M$, and $\p_i=\p_{u_i}$.
\[\partial_i\int_{\gamma_k^-}e^{F+\bar{F}}w_a=\int_{\gamma_k^-}e^{F+\bar{F}}(D_{i}+C_{i})w_a,\]
\[\bar{\p}_{\bar{i}}\int_{\gamma_k^-}e^{F+\bar{F}}w_a=\int_{\gamma_k^-}e^{F+\bar{F}}(\bar{D}_{\bar{i}}+\bar{C}_{\bar{i}})w_a.\]
\end{lem}
\begin{proof}
First, notice that
\begin{equation}\label{dplusc1}\partial_i\int_{\gamma_k^-}e^{F+\bar{F}}w_a=\int_{\gamma_k^-}\partial_i(e^{F+\bar{F}}w_a)=\int_{\gamma_k^-}e^{F+\bar{F}}(\partial_i+\psi_i)w_a.\end{equation}
Since $e^{F+\bar{F}}w_a$ is $d$-closed for all $u$ and $\partial_{i}$ commutes with $d$, $e^{F+\bar{F}}(\partial_i+\psi_i)w_a=\partial_{i}(e^{F+\bar{F}}w_a)$ is also $d$-closed, which implies that $(\partial_i+\psi_i)w_a$ is $d_{2\Re(F)}$-closed. Proceed as in the proof of Proposition \ref{riebil1}, one shows that \[(\partial_i+\psi_i)w_a=\Pi_u((\partial_i+\psi_i)w_a)+d_{2\Re(F)}\alpha=(D_i+C_i)w_a+d_{2\Re(F)}\alpha\]
for some differential form $\alpha$ with exponential decay.
As a result,
\begin{equation}\label{dplusc2}\int_{\gamma_k^-}e^{F+\bar{F}}(\partial_i+\psi_i)w_a
=\int_{\gamma_k^-}e^{F+\bar{F}}(D_{i}+C_{i})w_a.
\end{equation}
By (\ref{dplusc1}) and (\ref{dplusc2}), one has
\[\partial_i\int_{\gamma_k^-}e^{F+\bar{F}}w_a=\int_{\gamma_k^-}e^{F+\bar{F}}(D_{i}+C_{i})w_a.\]
Similarly, one can show that
\[\bar{\p}_{\bar{i}}\int_{\gamma_k^-}e^{F+\bar{F}}w_a=\int_{\gamma_k^-}e^{F+\bar{F}}(\bar{D}_{\bar{i}}+\bar{C}_{\bar{i}})w_a.\]
\end{proof}
\begin{proof}[Proof of Theorem \ref{134}]
First, let $r_u'=2^{\frac{4-n}{2}}\pi r_u.$
By Theorem \ref{fanlan} and Proposition \ref{rpqdegree}, it suffices to prove
\begin{flalign}
\begin{split}\label{gaussmanin}
r_u'(Dw_a)&=D^{CY}r_u'(w_a),\quad r_u'(\bar{D}w_a)=\bar{D}^{CY}r_u'(w_a),\\
r_u'(Cw_a)&=C^{CY}r_u'(w_a),\quad r_u'(\bar{C}w_a)=\bar{C}^{CY}r_u'(w_a)
\end{split}
\end{flalign}
for $\phi_a\in\Jac(F)'.$
Then $r_u':\H'\to H^{n-2}(X_F)_{\prim}$ induces an isomorphism of $tt^*$ structure.
By (\ref{periodlgcy}), one has
\[\partial_i\int_{\gamma_k^-}e^{F+\bar{F}}w=\partial_i\int_{\delta_k} r_u'(w),\]
\[\bar{\p}_{\bar{i}}\int_{\gamma_k^-}e^{F+\bar{F}}w=\bar{\p}_{\bar{i}}\int_{\delta_k} r_u'(w).\]
Hence by Lemma \ref{gmc}, Proposition \ref{rpqdegree} and the definition of Gauss-Manin connection $D^{CY}+C^{CY}$,
\begin{equation}\label{gaussmanin1}
\begin{aligned}
&r_u'\left((D+C)w_a\right)=(D^{CY}+C^{CY})r_u'(w_a),\\
&r_u'\left((\bar{D}+\bar{C})w_a\right)=(\bar{D}^{CY}+\bar{C}^{CY})r_u'(w_a).
\end{aligned}
\end{equation}
Suppose $w_b=S(A_b)$ such that $l(A_b)=l(A_a)$, then
by Riemann bilinear formula and Proposition \ref{rpqdegree},
\[\int_{\C^n} w_a\wedge*\bar{w}_b=\int_{X_F}r_u'(w_a)\wedge*\overline{r_u'(w_b)}.\]
As a result,
\[\partial_i\int_{\C^n} w_a\wedge*\bar{w}_b=\partial_i\int_{X_F}r_u'(w_a)\wedge*\overline{r_u'(w_b)}.\]
Then notice that
\[\partial_i\int_{\C^n} w_a\wedge*\bar{w}_b=\int_{\C^n} D_iw_a\wedge*\bar{w}_b,\]
and
\[\partial_i\int_{X_F}r_u'(w_a)\wedge*\overline{r_u'(w_b)}=\int_{X_F}D^{CY}_ir_u'(w_a)\wedge*\overline{r_u'(w_b)}.\]
As a result,
\[r_u'(Dw_a)=D^{CY}r_u'(w_a).\]
Similarly,
\[r_u'(\bar{D}w_a)=\bar{D}^{CY}r_u'(w_a).\]
Together with (\ref{gaussmanin1}), we have (\ref{gaussmanin}).
\end{proof}
Finally, we summarize our results. Let $F$ be the marginal deformation of a non-degenerate homogeneous polynomial $f$ of degree $n$ on $\C[z_1,\ldots,z_n]$. The (small) $tt^*$ structure in the Calabi-Yau B-model and Landau-Ginzburg B-model are given by
\begin{itemize}
\item $tt^*$ structure in the Calabi-Yau B model:
$$\left(H^{n-2}(X_F)_{\prim}, D^{CY}, \bar{D}^{CY}, C^{CY},\bar{C}^{CY},g\right).$$
\item small $tt^*$ structure in the Landau-Ginzburg B model:
$$\left(\mathcal{H}', D,\bar{D}, C,\bar{C},h\right).$$
\end{itemize}
The map $r_u':\mathcal{H}'\rightarrow H^{n-2}(X_F)_{\prim}$ is an isomorphism of two $tt^*$ geometry structures above.
\appendix
\section{Agmon estimate}
In this section, we let $V\in C^\infty(\R^n)$ be a nonnegative function with finite isolated zeros. Moreover
\[\lim_{|x|\to\infty}\frac{|\nabla V|}{(V+1)^{3/2}}(x)=0.\]
The metric $\tilde{g}:=Vg_0$ is called Agmon metric with respect to $V$, where $g_0$ is the standard metric on $\R^n$. Let $\tilde{d}$ be the distance function induced by the Agmon metric, and $\rho(x):=\tilde{d}(x,0).$
We summarize several nice properties of Agmon distance here (c.f. \cite{DY2020cohomology} )
\begin{lem}\label{propagm}
\begin{enumerate}
\item $|\nabla \rho|^2=V$, almost everywhere;
\item If $|\nabla f|\leq V$, then $|f(x)-f(y)|\leq \tilde{d}(x,y)$. In particular, if $f(0)=0$, $|f(x)|\leq \rho(x).$
\end{enumerate}
\end{lem}
\begin{lem}\label{Agmonre1}
\def\dvol{\mathrm{dvol}}
Assume that $w\in L^2(\R^n),0\leq u\in L^2(M)$, such that $(\Delta +V)u\leq w$ outside a compact subset $K\subset M$ in the weak sense (where the interior of $K$ contains all the zeros of $V$). That is
\[\int_{\R^n-K}\left(\nabla u\nabla v+V u v\right) \dvol\leq\int_{\R^n-K}w\cdot v~\dvol, \ \ \forall\ 0\leq v\in C_c^\infty(M-K).\]
If $\int_{\R^n-K}|w|^2V^{-1}\exp(2b\rho)\dvol<\infty$ for some $b\in(0,1)$, then for any compact set $L$ such that $L^\circ\supset K$, one has
\begin{align}\begin{split}\label{techeq}\int_{\R^n-L}V|u|^2\exp(2b\psi)\dvol&\leq C(b,K,L) \int_{\R^n-K}|w|^2V^{-1}\exp(2b\psi)\dvol \\
&+ C(b,K,L) \int_{L-K}V|u|^2\exp(2b\psi)\dvol. \end{split}\end{align}
%for some $a\in(0,\frac{\sqrt{2}}{2})$,
\end{lem}
\begin{proof}
This is exactly Lemma 3.1 in \cite{DY2020cohomology}.
\end{proof}
First, let us recall the De Giorgi-Nash-Moser theorem:
\begin{thm}\label{moserde}
Let $B_r:=\{x\in\R^n:|x|<r\}.$
Suppose that $u\in L^2(B_r)$, $w\in L^N(B_r)$ for some $N>n/2$, s.t. $\Delta u\leq w$ in the weak sense (and $u\geq 0.$).
Then
\[\sup_{y\in B_r}u(y)\leq {C}(r^{-n/2}\|u\|_{L^2(B_{2r})}+r^{-n/N}\|w\|_{L^N(B_{2r})}),\]
where $C>0$ is a constant depending on $n$ and $N.$
\end{thm}
\begin{proof}
See Theorem 4.1 in \cite{han2011elliptic} for a reference.
\end{proof}
\begin{lem}\label{epde12}
Suppose $u,w$ satisfy the same conditions as in Lemma \ref{Agmonre1} for any compact set containing the zeros of $V$. Moreover, assume that
$w$ and satisfies
\[\|w\|_{L^{N}_{wt}}^{N}:=\int_{M}|w|^{N}\exp(Nb\rho)dvol<\infty,\]
\[\|w\|_{L^{2}_{wt}}^{2}:=\int_{M}|w|^{2}\exp(2b\rho)dvol<\infty,\]
where $N^*=\frac{N}{N-1}.$
Then for $a\in(0,b)$,
\[|u|\leq C(\psi,V,n,N,a,b)\left(\|u\|_{L^2}+\|w\|_{L^{N}_{wt}}+\|w\|_{L^{2}_{wt}}\right)\exp(-a\rho).\]
\end{lem}
\begin{proof}
First, we fix a compact subset ${K}$, such that outside ${K}$, $\frac{|\nabla V|}{(V+1)^{3/2}}\leq\frac{(b-a)}{2},$
and let $L:=\{x\in\R^n: \tilde{d}(x,{K})\leq3\}.$
\def\tb{\tilde{B}}
Denoted $\tb_r(x):=\{y\in\R^n:\tilde{d}(y,x)<r\}$.
For $x_0\notin L$, set $l=\sup_{x\in \tb_2(x_0)}V(x)$, and $r=1/(2l).$ Then one can easily verify that $B_{2r}(x_0)\subset \tb_1(x_0)$.
Choose $y_0\in \overline{\tb_2(x_0)}$ so that $V(y_0)\in (l/2,l].$
By Lemma \ref{Agmonre1} and De Giorgi-Nash-Moser estimate (Theorem \ref{moserde}),
\begin{align*}
&|u(x_0)|^2e^{2b\rho(x_0)}\leq C(n,N)(r^{-n}\|u\|^2_{L^2(B_{2r}(x_0))}e^{2b\rho(x_0)}+r^{-2n/N}\|w\|^2_{L^N(B_{2r}(x_0))}e^{2b\rho(x_0)})\\
&\leq{C(n,N,b)}\left({r^{-n}}\int_{\tb_1(x_0)}|u|^2(y)e^{2b\rho(y)}dy +r^{-2n/N}(\int_{\tilde{B}_1(x_0)}|w(y)|^Ne^{Nb\rho(y)}dy)^{2/N}\right)\\
&\leq C(n,N,b,a,V)\left({r^{-n}}\int_{L-K}|u|^2e^{2b\rho}dy+r^{-n}\int_{\R^n-K}|w|^2e^{2b\rho}dy +r^{-2n/N}\|w\|_{L^N_{wt}}^2\right)\\
&\leq C(n,N,b,a,V)\left(|V(y_0)|^n\|u\|_{L^2}^2+|V(y_0)|^{n}\|w\|_{L^2_{wt}}^2 +|V(y_0)|^{2n/N}\|w\|_{L^N_{wt}}^2\right)
\end{align*}
Proceeding as in \cite{DY2020cohomology}, one has \begin{equation}|V(y_0)|^2 \leq C(V,a,b)\exp((b-a) \rho(y_0)).\end{equation}
Hence, the result follows.
For $x_0\in L$, we have classical De Giorgi-Nash-Moser estimate
\[|u(x_0)|\leq C(a,b,V,N,n)(\|u\|_{L^2}+\|w\|_{L^N}). \]
Since in $L$, $\exp(-a \rho)$ has an upper and a lower bound depending on $a,b$ and $V$, the result follows.
\end{proof}
\subsection{Witten deformation and Agmon estimate}\label{witagm}
Suppose that $f$ is a non-degenerate homogeneous polynomial on $\C^n$, then for any $a\in(0,1)$, there exists $r_0:=r_0(a)>0$, s.t outside $B_{r_0}:=\{x\in \R^n:|x|\leq r_0\}$, the Witten Laplacian $\Delta_{2\Re(f)}\geq \Delta+a|2\nabla \Re(f)|^2$ (c.f. \cite{DY2020cohomology} and \cite{zhang2001lectures}), i.e. for any smooth form $w$,
\[g_0(\Delta_{2\Re(f)}w,w)(p)\geq g_0((\Delta+a|2\nabla \Re(f)|^2)w,w)(p)\]
for any point $p\notin B_{r_0}.$
Then if $\Delta_{2\Re(f)} u=v$ for some differential forms $u$ and $v$, Bochner formula and Kato's
inequality tells us that
\be\label{boc}(\Delta+a|2\nabla \Re(f)|^2)|u|\leq |v|\ee
weakly (outside $B_{r_0}$).
In this case, let $\rho$ be the Agmon distance with respect to $V:=|\nabla 2\Re(f)|^2$. The Agmon estimate discussed in the previous section is applicable for Witten Laplacian.
\section{Proof of Proposition \ref{riebil1}}
It suffices to prove the case of $u=0$.
Recall that by our construction, $\gamma_k=\sup_{t>0}\Phi_t^*(\sigma_k)$ for some $\sigma_k\in H_{n-2}(V_{-1}).$
Let \[\epsilon_1:=(2\sup_{z\in\cup_{k=0}^{\mu-1}\sigma_k}|z|)^{-n}.\]
Let \[V:=\{z\in\C^n:|z|=1, |\Re(f)|\geq \epsilon_1\},\] and $U$ be the cone
\[U:=\{z\in\C^n:\frac{z}{|z|}\in V\},\]
then one has $\cup_{k=0}^{\mu-1}\gamma_k\subset U.$
\def\tf{\tilde{f}}
\begin{lem}\label{tff4}
For every sufficiently small $\epsilon>0$, there exists a smooth function $\tf:\C^n\to \R$, such that
\begin{itemize}
\item $\tf\geq |\Re(f)|.$ Moreover, $\tf=|\Re(f)|$ in $U.$
\item $|\nabla \tf|\leq (1+c(n,f)\epsilon)|\nabla \Re(f)|$ for some $c(n,f)>0.$ Hence by Proposition \ref{propagm}, we also have \be\label{tfrho}\tf\leq\frac{(1+c(n,f)\epsilon)}{2}\rho.\ee
\item $\tf\geq \epsilon |z|^n$.
\end{itemize}
\end{lem}
\begin{proof}
Let $\eta\in C^\infty(\R)$, $\epsilon<\epsilon_1$, such that
\begin{itemize}
\item $\epsilon/2\leq\eta(x)\leq \epsilon$ if $|x|\leq\epsilon,$ and $\eta(x)=|x|$, if $|x| \geq \epsilon$;
\item $\eta\geq \epsilon/2$;
\item $|\eta'|\leq 1$.
\end{itemize}
Let $(r,\theta)$ be the polar coordinates of $\C^n$, and $\nabla^{\theta}$ be gradient with respect to the standard metric on $S^{2n-1}:=\{z\in\C^n:|z|=1\}$. Now set $\tf(r,\theta):=r^n\eta\circ\Re(f)(1,\theta)$.
By our construction, one can see that
\begin{itemize}
\item \be\label{tff1}\mbox{$\epsilon|z|^n/2\leq\tf\leq\epsilon|z|^n$ if $z\notin U$, and $\tf(z)=|\Re(f)|(z)$ if $z\in U$}\ee
\item \be\label{tff2}\tf\geq \epsilon |z|^n/2;\ee
\item \be\label{tff}|\nabla^{\theta} \tf|\leq |\nabla^{\theta} \Re(f)|.\ee
\end{itemize}
Note that in $U$, $|\nabla \tf|= |\nabla \Re(f)|$.
In polar coordinates,
\begin{align}\begin{split}\label{tff3}&|\nabla \Re(f)(r,\theta)|^2=n^2r^{2n-2}|\Re(f)(1,\theta)|^2+r^{2n-2}|\nabla^{\theta} \Re(f)(1,\theta)|^2,\\
&|\nabla \tf(r,\theta)|^2=n^2r^{2n-2}\tf(1,\theta)+r^{2n-2}|\nabla^{\theta} \tf (1,\theta)|^2.\end{split} \end{align}
Let $\epsilon_2:=\inf_{|z|=1}|\nabla \Re(f)|$, then by (\ref{tff1}), (\ref{tff}) and (\ref{tff3}), one can see easily that outside $U$,
\[|\nabla \tf|\leq (1+\frac{\epsilon}{n\epsilon_2})|\nabla \Re(f)|.\]
\end{proof}
Notice that $e^{-\bar{f}}A_a$ is $d_{2\Re(f)}$ closed, although $e^{-\bar{f}}A_a$ is not $L^2$ integrable, one still have the following formulation of Hodge decomposition:
\begin{lem}\label{decom}
One has the following decomposition
\[e^{-\bar{f}}A_a=w'+d_{2\Re(f)}\beta,\]
where $w'$ is a harmonic form. Moreover, $\beta$ satisfies
\begin{itemize}
\item $e^{f+\bar{f}}\beta$ has exponential decay on $U$, i.e., there exists $c,C>0$, such that $|e^{f+\bar{f}}\beta|\leq Ce^{-c|z|^n}$ in $U.$
\item There exist $a\in(0,1), C>0$, such that $|\beta|\leq Ce^{a\rho},$ where $\rho$ is the Agmon distance.
\end{itemize}
\end{lem}
Once we have the decomposition in the lemma, one can see
\[\int_{\gamma_k^-}e^{f}A_a=\int_{\gamma_k^-}e^{f+\bar{f}}w'\]
and for any harmonic form $w$,
\[\int_{\C^n} e^{-\bar{f}}A_a\wedge*w=\int_{\C^n}w'\wedge*w.\]
(This is because, since $\beta$ satisfied the estimate above, by Stoke formula
\[\int_{\gamma_k^-}d e^{f+\bar{f}}\beta=0,\]
and integration by parts,
\[\int_{\C^n}d_{2\Re(f)}\beta\wedge*w=\int_{\C^n}\beta\wedge* d_{2\Re(f)}^\dagger w=0.)\]
By Proposition \ref{riebil}, one has Proposition \ref{riebil1}.
Now it suffices to prove Lemma \ref{decom}.
\begin{proof}[Proof of Lemma \ref{decom}]
We fix a function $\tf$ that satisfies the conditions in Lemma \ref{tff4} for a fixed $\epsilon<\min\{\epsilon_1,\frac{1}{16c(n,f)}\}$.
\begin{enumerate}[Step 1.]
\item Let
$$d_{tw,0}:=e^{-2\Re(f)-3\tf/2}\circ d\circ e^{2\Re(f)+3\tf/2}=e^{-3\tf/2}\circ d_{2\Re(f)}\circ e^{3\tf/2},$$
and
$$\Delta_{tw,0}:=d_{tw,0}d_{tw,0}^{\dagger}+d_{tw,0}^{\dagger}d_{tw,0}$$ be the Witten Laplacian with respect to $d_{tw,0}$. Then one can see that $e^{-3\tf/2-\bar{f}}A_a\in L^2(\C^n)$ (since $|e^{-3\tf/2-\bar{f}}A_a|\leq Ce^{\tf/2}|z|^{l_a}\leq Ce^{-\epsilon|z|^n/2}|z|^{l_a}$),
and \[d_{tw}e^{-3\tf/2-\bar{f}}A_a=0.\] As a reselt, we have Hodge decomposition (c.f. \cite{DY2020cohomology})
\[e^{-3\tf/2-\bar{f}}A_a={w}_0+d_{tw,0}\beta_0,\]
where $w_0$ is $\Delta_{tw,0}$-harmonic, and we can also assume that $\beta_0\in \Im(d_{tw,0}^{\dagger}).$
Since $\left|\nabla \left(2\Re(f)+3\tf/2\right)\right|\geq 2|\nabla \Re(f)|-3/2|\nabla \tf|\geq 13/32|\nabla \Re(f)|$, Agmon estimate tells us that
\be\label{wzero}|w_0|\leq Ce^{-3\rho/16}\ee
for some $C>0$.
Moreover, by our choice of $\beta_0$, one has
\[d_{tw,0}^{\dagger}e^{-3\tf/2-\bar{f}}A_a=d_{tw,0}^{\dagger}\tilde{w}_0+d_{tw,0}^{\dagger}d_{tw,0}\beta_0=d_{tw,0}^{\dagger}d_{tw,0}\beta_0=\Delta_{tw,0}\beta_0.\]
Let $\epsilon_3:=n\sup_{|z|=1}|\nabla 2\Re(f)|$, then $\rho\leq \epsilon_3|z|^n$. Hence, there exists $b>0$, such that
$e^{b\rho}d_{tw,0}^{\dagger}e^{-3\tf/2-\bar{f}}A_a$ is $L^2$ and $L^N$ integrable for some $N>n.$
Hence, by Agmon estimate again, there exists $c<b$, such that
\be|\beta_0|\leq Ce^{-c\rho}.\ee
Now let $\tilde{\beta}_0:=e^{3\tf/2}\beta_0$, $\tilde{w}_0=e^{3\tilde{f}/2}w_0$, then
$e^{-\bar{f}}A_a=\tilde{w}_0+d_{2\Re(f)}\tilde{\beta}_0.$
Moreover, $\tilde{w}_0$ is $d_{2\Re(f)}$-closed. In $U$, $\tilde{f}=|\Re(f)|$, hence for some $c'>0$, \[|e^{f+\bar{f}}\tilde{\beta}_0|\leq Ce^{-f/2-c\rho}\leq Ce^{-c'|z|^n}.\]
In addition, \[|\tilde\beta_0|\leq Ce^{3\tf/2-c\rho}\leq Ce^{(3(1+\epsilon c(n,f))/4-c)\rho}.\] Hence $\tilde{\beta}_0$ satisfies the estimate stated above.
\item Although $\tilde{w}_0$ is $d_{2\Re(f)}$-closed, it is not harmonic. Also, $\tilde{w}_0$ may not be $L^2$ integrable. To continue, we will use the techniques similar to those used in Section 7.5 of \cite{DY2020cohomology}.
Let $$d_{tw,1}=e^{-\tf/4}\circ d_{tw,0}\circ e^{\tf/4}=e^{-5/4\tf-2\Re(f)}\circ d\circ e^{5/4\tf+2\Re(f)},$$ and $\Delta_{tw,1}$ be the Witten Laplacian with respect to $d_{tw,1}.$
By (\ref{wzero}) and (\ref{tfrho}), $\alpha_1:=e^{\tf/4}w_0$ is $d_{tw,1}$ closed and $L^2$-integrable. Hence, we have Hodge decomposition
\[\alpha_1=w_1+d_{tw,1}\beta_1,\]
where $w_1$ is $\Delta_{tw,1}$ harmonic, and we may assume $\beta_1\in \Im(d_{tw,1}^{\dagger})$.
Since
\begin{align*}
\left|\nabla \left(5/4\tf+2\Re(f)\right)\right|&\geq 2|\nabla \Re(f)|-5/4|\nabla \tf|\\
&\geq2|\nabla \Re(f)|-3/2|\nabla \tf|\geq 13/32|\nabla \Re(f)|,
\end{align*}
we also have $|w_1|\leq C e^{-3\rho/16}.$ Similarly, $|\beta_1|\leq Ce^{-c\rho}$ for some $c>0.$
Then, let $\tilde{w}_1=e^{5\tf/4}w_1,$ $\tilde{\beta}_1=e^{5\tf/4}\beta_1$, then we have
\[\tilde{w}_0=\tilde{w}_1+d_{2\Re(f)}\tilde{\beta}_1.\]
Here $\tilde{\beta}_1$ satisfies the estimates stated above, and $\tilde{w}_1$ is $d_{2\Re(f)}$ closed (but may not be $L^2$ integrable).
\item Now let $d_{tw,k}=e^{-k\tf/4}\circ d_{tw,0}\circ e^{k\tf/4}$. Repeating the arguments in Step 2 for 6 times, eventually we get
$\alpha_6=e^{\tilde{f}/4} w_5$ is $d_{tw,6}(=d_{2\Re(f)})$ closed and $L^2$ integrable. Hence we have Hodge decomposition
\[\alpha_6=w_6+d_{2\Re(f)}\beta_6,\] where $w_6$ is $\Delta_{2\Re(f)}$ harmonic, and $\beta_6$ satisfies the estimates stated above.
\end{enumerate}
Eventually, set $w'=w_6$, $\beta:=\tilde{\beta}_0+\cdots+\tilde{\beta}_5+\beta_6$, we finish the proof.
\end{proof}
\bibliography{lib}
\bibliographystyle{plain}
\end{document}
\begin{thebibliography}{10}
\bibitem{arnold2012singularities}
V.~I. Arnold, A.~N. Varchenko, and S.~M. Gusein-Zade.
\newblock {\em Singularities of differentiable maps: Volume {II} Monodromy and
asymptotic integrals}, volume~83.
\newblock Springer Science \& Business Media, 2012.
\bibitem{Bershadsky1994KodairaSpencerTO}
M.~Bershadsky, M.~Cecotti, H.~Ooguri, and C.~Vafa.
\newblock {K}odaira-{S}pencer theory of gravity and exact results for quantum
string amplitudes.
\newblock {\em Communications in Mathematical Physics}, 165:311--427, 1994.
\bibitem{Candelas1991APO}
P.~Candelas, X.~Ossa, P.~Green, and L.~Parkes.
\newblock A pair of {Calabi-Yau} manifolds as an exactly soluble superconformal
theory.
\newblock {\em Nuclear Physics}, 359:21--74, 1991.
\bibitem{carlson1980infinitesimal}
J.~Carlson and P.~Griffiths.
\newblock Infinitesimal variations of {H}odge structure and the global
{T}orelli problem.
\newblock In {\em Journ{\'e}es de G{\'e}ometrie Alg{\'e}brique dAngers, Juillet
1979/Algebraic Geometry, Angers, 1979}, pages 51--76, 1980.
\bibitem{CECOTTI1991N}
S.~Cecotti.
\newblock ${N}=2$ {L}andau-{G}inzburg {vs}. {C}alabi-{Y}au $\sigma$-models:
nonperturbative aspects.
\newblock {\em International Journal of Modern Physics A}, 06(10):1749--1813,
1991.
\bibitem{cecotti1991topological}
S.~Cecotti and C.~Vafa.
\newblock Topological--anti-topological fusion.
\newblock {\em nuclear Physics B}, 367(2):359--461, 1991.
\bibitem{chiodo2014landau}
A.~Chiodo, H.~Iritani, and Y.~Ruan.
\newblock {Landau-Ginzburg/Calabi-Yau} correspondence, global mirror symmetry
and {Or}lov equivalence.
\newblock {\em Publications math{\'e}matiques de l'IH{\'E}S}, 119(1):127--216,
2014.
\bibitem{CostLi}
K.~Costello and S.~Li.
\newblock Quantum {BCOV theory on Calabi-Yau} manifolds and the higher genus
{B}-model.
\newblock {\em arXiv:1201.4501}.
\bibitem{DY2020cohomology}
X.~Dai and J.~Yan.
\newblock Witten deformation for noncompact manifolds with bounded geometry.
\newblock {\em Journal of the Institute of Mathematics of Jussieu}, pages
1--38, 2021.
\bibitem{Dub1993tt}
B.~Dubrovin.
\newblock Geometry and integrability of topological-antitopological fusion.
\newblock {\em Comm. Math. Phys.}, 152(3):539--564, 1993.
\bibitem{fan2011schr}
H.~Fan.
\newblock Schr$\mathrm{\ddot{o}}$dinger equations, deformation theory and $
tt^* $-geometry.
\newblock {\em arXiv:1107.1290}.
\bibitem{FJR2013}
H.~Fan, T.~Jarvis, and Y.~Ruan.
\newblock The {W}itten equation, mirror symmetry, and quantum singularity
theory.
\newblock {\em Ann. Math}, (178):1--106, 2013.
\bibitem{Fan2020LGCYCB}
H.~Fan, T.~Lan, and Z.~Yang.
\newblock {LG/CY} correspondence between $ tt^* $ geometries.
\newblock {\em arXiv:2011.14658}.
\bibitem{Giv2001b}
A.~Givental.
\newblock Semisimple {F}robenius structures at higher genus.
\newblock {\em Internat. Math. Res. Notices}, 23:1265--1286, 2001.
\bibitem{Greene1990DualityI}
B.~Greene and M.~Plesser.
\newblock Duality in {Calabi-Yau} moduli space.
\newblock {\em Nuclear Physics}, 338:15--37, 1990.
\bibitem{han2011elliptic}
Q.~Han and F.~Lin.
\newblock {\em Elliptic partial differential equations}, volume~1.
\newblock American Mathematical Soc., 2011.
\bibitem{Hert2003tt}
C.~Hertling.
\newblock $tt^*$ geometry, {F}robenius manifolds, their connections, and the
construction for singularities.({E}nglish {S}ummarty).
\newblock {\em J. Reine Angew. Math.}, 555:77--161, 2003.
\bibitem{Klimek1991local}
S.~Klimek and A.~Lesniewski.
\newblock Local rings of singularities and {$N=2$} supersymmetric quantum
mechanics.
\newblock {\em Communications in Mathematical Physics}, 136(2):327--344, 1991.
\bibitem{li2013primitive}
C.~Li, S.~Li, and K.~Saito.
\newblock Primitive forms via polyvector fields.
\newblock {\em arXiv preprint arXiv:1311.1659}, 2013.
\bibitem{li2019l2}
S.~Li and H.~Wen.
\newblock On the ${L}^2$-{H}odge theory of {Landau-Ginzburg} models.
\newblock {\em arXiv:1903.02713}.
\bibitem{1981Primitive}
K.~Saito.
\newblock Primitive forms for a universal unfolding of a function with an
isolated critical point.
\newblock {\em Faculty of Science, The University of Tokyo}, (3):775--792,
1982.
\bibitem{1983higher}
K.~Saito.
\newblock {\em The higher residue pairings {$\mathcal{K}_F$} for a family of
hypersurface singular points.}
\newblock 1983.
\bibitem{SaitoFM}
K.~Saito.
\newblock From primitive forms to {F}robenius manifolds.
\newblock {\em Proc. Sympos. Pure Math.}, 78, 2008.
\bibitem{shen2016explicit}
Y.~Shen and H.~Fan.
\newblock Explicit isomorphisms between pairings in quantum singularity theory.
\newblock {\em SCIENTIA SINICA Mathematica}, 46(5):533--548, 2016.
\bibitem{Tang2017ttGS}
X.~Tang.
\newblock $tt^*$ geometry, singularity torsion and anomaly formulas.
\newblock {\em arXiv:1710.03915}.
\bibitem{Witten1993}
E.~Witten.
\newblock Algebraic geometry associated with matrix models of two-dimensional
gravity.
\newblock {\em Topological methods in modern mathematics}, pages 235--269,
1993.
\bibitem{zhang2001lectures}
W.~Zhang.
\newblock {\em Lectures on {Chern-Weil} theory and {Witten} deformations},
volume~4.
\newblock World Scientific, 2001.
\end{thebibliography}