An effective theory of GW and FJRW invariants of quintics Calabi-Yau manifolds (Q2135648)

The paper is a part of the authors' program to develop a rigorous geometric theory that formalizes Witten's idea of a phase transition between Calabi-Yau and Landau-Ginzburg models in string theory. The geometric correlates of the two are, respectively, the Gromov-Witten (GW) and Fan-Jarvis-Ruan-Witten (FJRW) theories, and the goal is to find wall-crossing formulas between the corresponding invariants. In a prior work [Camb. J. Math. 7, No. 3, 319--364 (2019; Zbl 1430.14104)], the authors introduced the theory of Mixed-Spin-P (MSP) fields with a field parameter \(\nu\) that ``quantizes'' Witten's phase transition parameter. In this paper, they derive virtual localization formulas for MSP moduli, and apply them to devise a recursive procedure for computing GW and FJRW invariants of the Fermat quintic. Based on Theorem 1 below, the first author et al. [Ann. Math. (2) 194, No. 3, 585--645 (2021; Zbl 1478.14084)] were able to prove a number of structural results for the quintic GW Feynman potentials \(F_g\), including Yamaguchi-Yau's finite generation conjecture, BCOV Feynman rules, and the holomorphic anomaly equations. The MSP fields consist of pointed twisted curves \(\Sigma^{\mathscr C}\subset{\mathscr C}\), line bundles \({\mathscr L}\) and \({\mathscr N}\), fields \(\varphi\in H^0({\mathscr L}^{\oplus 5})\), \(\rho\in H^0({\mathscr L}^{\vee\otimes5}\otimes \omega^{\log}_{\mathscr L})\), and the projectivized spin (scaling) field \(\nu\in H^0({\mathscr L}\otimes{\mathscr N}\oplus {\mathscr N})\), subject to some non-degeneracy condition. MSP moduli were proved to be Deligne-Mumford \(\mathbb{G}_m\)-stacks with properly supported \(\mathbb{G}_m\)-equivariant cycles obtained by the cosection localized virtual cycle construction. In this paper the authors derive a class of their vanishings which provide, by virtual localization, polynomial relations between GW and FJRW invariants of the Fermat quintic. Based on these relations, they describe an algorithm to calculate all genus GW invariants of quintic threefolds, and all genus FJRW invariants of Fermat quintics with \(\frac{2}{5}\)-insertions from certain initial data \(\Big(\)a \(\frac{2}{5}\)-insertion is a marking \(\Sigma_i\subset\Sigma^{\mathscr C}\) with the monodromy representation of \({\mathscr L}\) along \(\Sigma_i\) equal to \(\left(\exp{\frac{2\pi i}{5}}\right)^{\!2}\)\,\(\Big)\). More precisely, let \(N_{g,d}\) be the genus \(g\) degree \(d\) GW invariants of the quintic threefold, and let \(\Theta_{g,k}\) be the genus \(g\) FJRW invariants of \(([\mathbb{C}^5/\mathbb{Z}_5],x_1^5+\cdots+x_5^5)\) with \(k\) \(\frac{2}{5}\)-insertions. The main results are summarized in the following two theorems. {Theorem 1.} Let \(d\geq g\geq1\). Then \(N_{g,d}\) can be computed recursively from:\\ \textrm{1)} \(N_{g',d'}\) for \(g'\leq g\), \(d'<d\); and\\ \textrm{2)} \(\Theta_{g,k}\) for \(k\leq2g-2\) and \(\Theta_{g',k}\) for \(g'<g\), \(k\leq2g-4\). {Theorem 2.} Let \(g\geq 1\) and \(k\ge 7g-2\). Then \(\Theta_{g,k}\) can be computed recursively from \(\Theta_{g,k'}\) for \(k'< k\) and \(\Theta_{g',k'}\) for \(g'<g\) and \(k'\leq k-2\). Since FJRW invariants \(\Theta_{0,k}\) have been previously computed [\textit{A. Chiodo} and \textit{Y. Ruan}, Invent. Math. 182, No. 1, 117--165 (2010; Zbl 1197.14043)] the algorithm allows to calculate all \(\Theta_{1,k}\) of \(([\mathbb{C}^5/\mathbb{Z}_5],x_1^5+\cdots+x_5^5)\). More generally, determination of the initial data remains an open problem and the authors formulate a conjecture in that direction.