On the convergence of Gromov-Witten potentials and Givental's formula (Q887907)

Let \(X\) be a smooth projective variety, and let \(\mathcal{Z}_X=\exp (\sum_{g\geq 0} h^{g-1}\mathcal{F}^g_X)\) be the total descendant potential of \(X\), where \(\mathcal{F}^g_X\) is a generating function for the genus \(g\) Gromov-Witten invariants of \(X\). \(\mathcal{Z}_X\) is a formal power series in \(h, h^{-1}\) and infinitely many variables keeping tracks of cohomology classes on \(X\). The paper under review studies the question of convergence for \(\mathcal{Z}_X\) and \(\mathcal{F}^g_X\) which is relevant to Ruan's work on Gromov-Witten theory and binational geometry. The notion of convergence that is mainly used here is called NF-convergence which roughly says \(\mathcal{Z}_X\) is convergent if for each \(g\), \(\mathcal{F}^g_X\) is convergent on an infinite dimensional polydisc of a specific form. The Gromov-Witten invariants of \(X\) are assumed to satisfy three conditions: 1) Formal semisimplicity, 2) Genus 0 convergence, and 3) Analytic semisimplicity. The conditions 1) \(\&\) 2) implies 3), and 2) \(\&\) 3) implies 1). If \(X\) is a compact toric variety or a complete flag variety then mainly from mirror symmetry and reconstruction theorems for logarithmic Frobenius manifolds it is known that \(X\) satisfies these conditions. The main result of the paper under review proves that \(\mathcal{Z}_X\) is NF-convergent if three conditions above are satisfied. In fact a more general version of this result is proven for the total ancestor potential of \(X\). An essential ingredient of the proof is \textit{A. Givental}'s [in: Topics in singularity theory. V. I. Arnold's 60th anniversary collection. Providence, RI: American Mathematical Society. 103--115 (1997; Zbl 0895.32006)] and \textit{C. Teleman}'s [Invent. Math. 188, No. 3, 525--588 (2012; Zbl 1248.53074)] work on higher genus potentials for target spaces with semisimple quantum cohomology.