Neural network enhanced thermo-chemical reaction cross section and broad temperature vibrational relaxation times for N2 + N

Abstract

For the N2 + N process, the quasi-classical trajectory method based on the first-principles potential energy surface is combined with the neural network method to construct a high-precision cross section model for inelastic collision, the Zeldovich exchange reaction, and the dissociation reaction. The rate coefficients at relatively low temperatures (<4000 K) are obtained by integrating the derived cross section model, and an accurate prediction of the vibrationally relaxation time of N2 is achieved by solving the corresponding master equations. It is found that the rate coefficient from the first vibrational excited state to the ground state can be accurately fitted by a second-order Arrhenius formula, and below 2000 K, the vibrational relaxation times calculated by the two-state method are consistent with those of the full vibrational ladder method (Landau–Teller formula and the e-folding method). The logarithm of the vibrational relaxation times turns out to show a linear relationship with Ttr−2/3, rather than the traditionally expected Ttr−1/3. The obtained vibrational relaxation times of the N2 + N process span nine orders of magnitude in the range of 1000–30000 K and provide a database for the study of the relaxation dynamics of N2 molecules in hypersonic flows.