A global fundamental invariant neural network potential energy surface and dynamics study of the hydroxyl radical and ammonia reaction

Abstract

The reaction between the hydroxyl radical (OH) and ammonia (NH₃) is of great significance to combustion and atmospheric chemistry, and serves as a typical six-atom reaction system involving two heavy atoms. In this study, a global ground electronic state potential energy surface (PES) for the OH + NH₃ reaction was constructed via the fundamental invariant neural network (FI-NN) fitting method. A total of 77 330 ab initio energy points were calculated at the AE-UCCSD(T)-F12a/AVTZ theoretical level. The overall root-mean-square error (RMSE) is 10.55 meV. The PES accurately describes not only the hydrogen abstraction channel, but also the NH₂OH formation channel. Employing the quasi-classical trajectory (QCT) method, we calculated integral cross-sections for the two reaction channels and analyzed product state distributions based on the new PES. Rate coefficients of the hydrogen abstraction channel were further obtained within the temperature range from 300 K to 2000 K.