The remodeling conjecture and the Faber-Pandharipande formula (Q1941041)

For a toric Calabi-Yau \(3\)-fold \(X\) the mirror theory lives on a family of curves in \((\mathbb{C}^*)^2\) called the ``mirror curve''. The Eynard-Orantin topological recursion applied to a smooth complex curve produces an infinite tower of free energies \(F_g\) and meromorphic differentials \(W_n^g\). The remodeling conjecture states that the mirror map applied to \(F_g\) and \(W_n^g\) produces the generating functions of the closed and open Gromov-Witten invariants of \(X\) respectively. The open part of the conjecture for \(X=\mathbb{C}^3\) was proved independently by Chen and Zhou. In this note the authors complete the proof for \(\mathbb{C}^3\) by demonstrating that the free energies reproduce the closed invariants given by the celebrated Faber-Pandharipande formula. The proof relies on methods of Chen and Zhou. The mirror curve to \(\mathbb{C}^3\) is a sphere with \(3\) punctures (pair of pants), and Chen and Zhou expressed its \(W_n^g\) in terms of Hodge integrals. The authors reduce \(F_g\) to Hodge integrals as well, but the computation is more subtle. A subtlety appears also in the relation to matrix models. The Eynard-Orantin recursion applied to the spectral curve of a matrix model exactly reproduces its correlation functions. However, there is a normalization ambiguity in the computation that may lead to discrepancy in the contributions of the constant maps to free energies. The authors show explicitly that this discrepancy is in fact present for the pair of pants, and in contrast to the open part of the conjecture the matrix model machinery can not be used. In conclusion they ask if there is a more direct computation of \(F_g\) that relies on geometry of the pair of pants rather than reduction to Hodge integrals.