Global diabatic potential energy surfaces of the BeH2 system constructed using neural network method

High-quality diabatic potential energy surfaces (PESs) for the 11A' and 21A' electronic states of the BeH2 system were constructed using a neural network approach incorporating symmetry-restricted functions, based on 13 172 high-level ab initio energy points. The ab initio calculations were performed at the MRCI-F12/AVQZ level of theory. After obtaining the diabatic PES matrix, a detailed examination of the elements of the diabatic coupling matrix was carried out. The results indicate that the newly developed diabatic PESs provide a reasonable description of the transitions between the electronic states. To further validate the diabatic PESs, dynamical calculations for the Be(1S) + H2 (v0 = 0, j0 = 0) reaction were performed based on both the adiabatic and diabatic PESs. A comparison between the adiabatic and nonadiabatic dynamical results reveals that, due to nonadiabatic effects, an energy barrier emerges along the reaction path, leading to a higher reaction threshold in the nonadiabatic case than in the adiabatic case. Furthermore, over the investigated collision energy range, the adiabatic results are significantly higher than the nonadiabatic results, underscoring the important role played by nonadiabatic effects in the reaction process, which should not be neglected.