Quasi-classical trajectory study of the CN− + CH3I reaction on a high-dimensional neural network PES

The gas-phase dynamics of bimolecular nucleophilic substitution (SN2) reactions have been extensively studied by both experimental and theoretical groups due to their broad applicability and the emergence of new mechanistic insights. The reaction between CN− and CH3I is particularly intriguing, as it can yield two isomeric products (NCCH3 or CNCH3 + I−) owing to the ambident nature of the CN− nucleophile. Previous velocity map imaging experiments revealed predominantly direct rebound dynamics and high internal energy excitation in the reaction products. In another study, direct dynamics simulations were performed employing the PM7 semi-empirical method and B3LYP/aug-cc-pVDZ/ECP density functional theory (DFT); however, these approaches were limited by an insufficient number of DFT trajectories (due to high computational costs) and an overestimation of hydrogen transfer reactions with PM7. In the present study, a high-dimensional neural network-based potential energy surface (PES) was developed using extensive electronic structure data. The PES was rigorously validated using established benchmarks and subsequently employed in quasi-classical trajectory simulations. A substantial number of reactive trajectories were generated, enabling a detailed analysis of the reaction dynamics. The simulation results are consistent with previous findings and provide new insights into the mechanistic pathways of this SN2 reaction.