On Landau-Ginzburg models for quadrics and flat sections of Dubrovin connections (Q304090)

Let \(X\) be a smooth projective complex Fano variety of dimension \(N\). Choose an effective anticanonical divisor \(D\) of \(X\). The mirror to \(X\) is then a \textit{Landau-Ginzburg model} (LG-model) \(f : M \to \mathbb{C}\), with \(f\) called the \textit{superpotential}. If \(F\) denotes a generic fiber of \(f\), then \(F\) is mirror to \(D\). In particular, different choices of \(D\) lead to different Landau-Ginzburg models, and \(M\) is the mirror to the open Calabi-Yau manifold \(X\backslash D\). Moreover, in a suitable sense (in terms of oscillatory integrals, cf. below), \(X\) is mirror to the \textit{critical locus} of \((M,f)\). Hence, given the same choice of \(D\), different LG-models may be considered. Mirrors to toric Fano manifolds were first constructed by Givental. Subsequently, Hori-Vafa provided a physics derivation for Givental's construction. There are various ways in which the mirror relationship between \(X\) and \((M,f)\) is described. The \(A\)-model data of virtual curve counts is encoded in the \(J\)-function, which is a certain generating function of Gromov-Witten invariants. In the case of projective Fano complete intersections, Givental expressed their \(J\)-functions in terms of oscillatory integrals on their mirror LG-models. Subsequently, generalizations of Givental's mirror conjecture were considered, cf. below. As in the paper under review, we now restrict to the case of Fano quadrics \(X=Q_N\subset \mathbb{P}^{N+1}\). The authors review and improve the known mirror symmetry results for odd \(N\), and they push forward the understanding of mirror symmetry for even \(N\). The authors start with reviewing and comparing the known LG-model constructions for \(Q_N\) and any \(N\): The original construction by Givental. A version with superpotential given as a Laurent polynomial on \((\mathbb{C}^*)^N\). A Lie-theoretic mirror. By passing through certain quivers, they obtain combinatorial models for the Laurent polynomial superpotential. This should be of independent interest in the light of the program by Coates-Corti-Galkin-Golyshev-Kasprzyk to classify Fano manifolds by their LG-model Laurent polynomials. Then, the authors introduce a new mirror LG-model, which they call the canonical mirror and denote by \((\check{X}_{\mathrm{can}},\mathcal{W}_q)\). It is a partial compactification of the other mirrors. Moreover, \(\check{X}_{\mathrm{can}}\) is the complement of an anticanonical divisor in a Langlands dual quadric and this LG-model has the advantage that \(\mathcal{W}_q\) is written in terms of coordinates that are identified with a cohomology basis of \(X\). The \(A\)-model information considered then is the \(D\)-module \(M_A\). This is a trivial \(H^*(X,\mathbb{C})\)-bundle on the 2-dimensional torus \(\mathbb{C}^*_q\times\mathbb{C}^*_{\hbar}\) with coordinates \(q\) and \(\hbar\). Using the flat connection constructed by Dubrovin in terms of Gromov-Witten invariants of \(X\), this gives \(M_A\) the structure of a \(D\)-module, where \(D=\mathbb{C}[\hbar^{\pm1},q^{\pm1}]\langle\partial_\hbar,\partial_q\rangle\). Note that for odd \(N\), the coefficients of the \(J\)-function generate \(M_A\), but this no longer holds true for even \(N\). The mirror \(B\)-model data considered is the \(D\)-module \(M_B\) obtained from the Gauss-Manin system associated to \((\check{X}_{\mathrm{can}},\mathcal{W}_q)\). The mirror symmetry conjecture states that as \(D\)-modules, \(M_A\) and \(M_B\) are isomorphic. While this was already known for odd \(N\), the description in terms of canonical coordinates allows the authors to write an explicit natural isomorphism. For even \(N\), they construct an explicit isomorphism from \(M_A\) to a natural sub-\(D\)-module of \(M_B\) and then conjecture that the image in fact is all of \(M_B\). Then, for any \(N\), they go deeper and more explicitly in the mirror symmetry relationship by describing the flat sections of the dual Dubrovin connection in terms of oscillatory integrals of \((\check{X}_{\mathrm{can}},\mathcal{W}_q)\). They explicitly calculate a particular section, the hypergeometric series of \(X\). As an application, this then allows them to compute some Gromov-Witten invariants of \(X\). The paper is well-written and contains an extensive introduction, which should be accessible to non-experts of the field. The work is well-referenced and contains links to other areas of mathematics including combinatorics, Lie theory and representation theory.