Gromov-Witten invariants of \(\mathbb{P}^{1}\) and Eynard-Orantin invariants (Q463139)

For every \((g, n)\in \mathbb{Z}^2 \) with \(g \geq 0\) and \(n >0\) the Eynard-Orantin invariant of a Torelli marked Riemann surface with meromorphic functions \((C, x, y)\) is a multidifferential \(\omega_n^g(p_1,\dots, p_n)\) for \(p_i \in C\). When \(2g - 2 + n > 0\), \(\omega_n^g(p_1,\dots, p_n)\) is defined recursively in terms of the local information around the poles of \(\omega_{n^\prime}^{g^\prime}(p_1,\dots, p_n^\prime)\) for \(2g^\prime+ 2 - n^\prime< 2g -2 +n\). By taking sequences of meromorphic functions, one can extend the definition to allow \(y\) to be any analytic function defined on a domain of \(C\) containing the branch points of \(x\). Different choices of \((C, x, y)\) give different Eynard-Orantin invariants. These invariants have been so far used to recover information about intersection numbers over the moduli space of curves, simple Hurwitz numbers, a count of lattice points in the moduli space of curves, (conjecturally) local Gromov-Witten invariants of (non-compact) toric Calabi-Yau 3-folds, and Chern-Simons invariants of 3-manifolds. The paper under review shows that the Gromov-Witten invariants of \(\mathbb{P}^1\) arise as Eynard-Orantin invariants, and use this to obtain some information about the structure of the Gromov-Witten invariants. For this, the authors choose \(x=z+1/z\) and \(y=\ln z\), where \(C \) is defined via the meromorphic function \(x(z)\). The first main result of the paper shows that for \(g = 0\) and \(1\) and \(2g - 2 + n > 0\), the Eynard-Orantin invariants of this curve \(C\) agree with the generating function for the stationary Gromov-Witten invariants of \(\mathbb{P}^1\) assembled as \[ \Omega_n^g(x_1,\dots,x_n)=\sum_{b_1,\dots,b_n}\langle \prod_{i=1}^n \tau_{b_i}(\omega)\rangle^g_d \prod_{i=1}^n (b_i+1)!x_i^{-b_i-2}dx_i. \] This result proves an efficient way to calculate the Gromov-Witten invariants of \(\mathbb{P}^1\). As an application, the authors show that tautological intersection numbers on the moduli space of curves arise in the asymptotics of large degree Gromov-Witten invariants of \(\mathbb{P}^1\). In fact, using the polynomial behavior of Eynard-Orantin invariants and the result above, they prove in this paper that \[ \Big\langle \prod_{i=1}^k \tau_{2u_i}(\omega) \prod_{i=k+1}^n \tau_{2u_i-1}(\omega)\Big\rangle^g=\frac{u_{k+1}\cdots u_n}{\prod_{i=1}^nu_i!^2}p_{n,k}^g(u_1,\dots,u_n), \] where \(p_{n,k}^g(u_1,\dots,u_n)\) is a polynomial of degree \(3g-3+n\) in the \(u_i\), symmetric in the first \(k\) and the last \(n-k\) variables, with top coefficient of \(u^{b_1} \cdots u^{b_n}\) given by \(2^g\int_{\overline{M}_{g,n}}\psi_1^{b_1}\cdots \psi_n^{b_n}\) with \(\sum b_i=3g-3+n\).