Exhaustive state-specific dissociation study of the N2(Σg+1)+N(S4) system using QCT combined with a neural network method.

This work studies the exhaustive rovibrational state-specific collision-induced dissociation properties of the N2+N system by QCT (quasi-classical trajectory) combined with a neural network method based on the ab initio PES recently published by Varga et al. [Phys. Chem. Chem. Phys. 23, 26273 (2021)]. The QCT combined with a neural network for state-specific dissociation (QCT-NN-SSD) model is developed and used to predict the dissociation cross sections and their energy dependence on the thermal range from a sparsely sampled noisy dataset. It is conservatively estimated that using this method can reduce the cost of the calculation by 96.5%. The rate coefficient of thermal non-equilibrium between different energy modes is obtained by combining the QCT-NN-SSD model with the multi-temperature model. The results show that, for the equilibrium state, dissociation mainly occurs at high vibrational and moderately low rotational levels. When the system is in non-equilibrium, there is no obvious vibrational level preference and highly rotationally excited molecules play a major role in promoting the dissociation by compensating for the lack of vibrational energy. The use of neural network training to generate complete datasets based on limited and discrete data provides an economical and reliable way to obtain a complete kinetic database needed to accurately simulate non-equilibrium flows.