Full-dimensional neural network potential energy surface for the multi-channel photodissociation of HNSiO via its S1 band

Abstract

A full-dimensional potential energy surface (PES) for the first excited state S₁(¹A″) of HNSiO was built using the neural network method based on more than 91 000 ab initio points, covering six possible product channels for photodissociation reactions: HNSiO + hv → H(²S) + NSiO(²Π)/HN(¹Δ) + SiO(¹Σ⁺)/HNSi(³Π) + O(³P)/NSi(²Σ) + OH(²Π)/N(²D) + SiOH(²A′)/HSi(²Σ) + NO(²Π). Seventeen stationary points consisting of five local minima and twelve transition states were found on the S₁ PES along the six dissociation pathways. Based on the newly constructed S₁ PES, the quasi-classical trajectory calculations were carried out to study the photodissociation dynamics of HNSiO(S₁) in the total energy range of 5.5–7.7 eV. It was found that although six product channels are energetically available at high total energies, the dominant (>90%) products are HN + SiO and H + NSiO. At low total energies, the major products are HN + SiO, whereas for a total energy of >7.1 eV, the H + NSiO product channel becomes increasingly favored owing to the faster movement of the light H atom. This scenario is quite analogous to the photodissociation of HNCO in the S₁ band. In particular, a high proportion of trajectories undergo isomerization during dissociation because of the relatively low barriers separating the isomers. Owing to the much shallower HOSiN and NSiOH wells compared with the HNSiO wells on the S₁ PES, the transferred H atom after isomerization returns to the HNSiO wells or dissociates directly, which leads to quite small percentages of other products, namely, N + SiOH, NSi + OH, and HSi + NO.