On the genus two free energies for semisimple Frobenius manifolds (Q1933341)

This paper decomposes the genus-two free energy of a semisimple Frobenius manifold \(M\) into two parts. The first part is a linear combination of products of genus-zero or genus-one free energies and their derivatives with respect to the flat coordinates. The second part is a function of the canonical coordinates derived from the semisimple Frobenius manifold \(M\), called the genus-two \(G\)-function. The first part can be written as a sum over 16 graphs, with each graph representing one type of contribution from lower-genus free energies, which is a usual way in the study of Gromov-Witten invariants. These graphs have some notable features; among other things, they are dual graphs of stable curves of arithmetic genus-two, and they are so-called one-particle irreducible graphs in the physical terminology. Moreover, restricting to the small phase space, the contribution of each of the 16 graphs is zero. The proof of the decomposition is based on the general formula of the all-genus free energies of semisimple Frobenius manifolds in [\textit{B. Dubrovin} and \textit{Y. Zhang}, ``Normal forms of hierarchies of integrable PDEs, Frobenius manifolds and Gromov-Witten invariants'', Preprint, \url{arXiv:math/0108160}]. A decomposition of the genus-one free energy with similar properties has appeared in [\textit{B. Dubrovin} and \textit{Y. Zhang}, Commun. Math. Phys. 198, No. 2, 311--361 (1998; Zbl 0923.58060)]. For a special class of semisimple Frobenius manifolds corresponding to the genus-zero FJRW invariants of ADE singularities, or the genus-zero Gromov-Witten invariants of \(\mathbb{P}^1\)-orbifolds of ADE types, this paper proposes a vanishing conjecture for the genus-two \(G\)-function. For these cases of semisimple Frobenius manifolds, an identity for the first 12 graphs' contributions is proved, which is also valid for all two-dimensional semisimple Frobenius manifolds and some special three-dimensional ones. Case by case computations are given to support this conjecture. The conjecture has been proved for the type-A singularities in [\textit{Y. Fu} et al., J. Geom. Phys. 76, 10--24 (2014; Zbl 1288.53080)], for all ADE singularities in [\textit{X. Liu} and \textit{X. Wang}, Trans. Am. Math. Soc. 367, No. 12, 8975--8996 (2015; Zbl 1334.53094)], and for \(\mathbb{P}^1\)-orbifolds of ADE types in [\textit{X. Liu} and \textit{X. Wang}, ``Genus-2 G-function for \(\mathbb{P}^1\) orbifolds'', Adv. Math. 274, 631--650 (2015; \url{doi:10.1016/j.aim.2015.01.011})].