The full‐dimensional potential energy surface of He2H− using fundamental invariant neural network method

Abstract

AbstractSo far, numerous researches on the He2H− system have still been restricted to several analytical forms of the potential energy interaction, where the calculations in the three‐body interaction may be not precise. With that in mind, we have provided an accurate, global, full‐dimensional potential energy surface (PES) for the ground state of the He2H− system by the fundamental invariant neural network (FI‐NN) fitting based on roughly 45 000 ab initio data points. The energy points are calculated with the coupled‐cluster singles and doubles with perturbative triples (CCSD[T]) method using the augmented Dunning‐type correlation‐consistent (aug‐cc‐pVTZ, aug‐cc‐pVQZ and aug‐cc‐pV5Z) basis sets, respectively. Based on adoption and comparison of three different extrapolation formulas, we obtained even accurate data points, which are extrapolated to the complete basis set (CBS) limit. The overall root mean square error of the fitting PES is only about 6.314 × 10−3 cm−1. It can be expected that this accurate PES would provide insights to further interesting discoveries in dynamics or spectroscopies for the relevant molecular systems.