Automated Exploration of Reaction Network and Mechanism via Meta-dynamics Nanoreactor

Abstract

We developed an automated approach to construct the complex reaction network and explore the reaction mechanism for several reactant molecules. The nanoreactor type molecular dynamics was employed to generate possible chemical reactions, in which the meta-dynamics was taken to overcome reaction barriers and the semi-empirical GFN2-xTB method was used to reduce computational cost. The identification of reaction events from trajectories was conducted by using the hidden Markov model based on the evolution of the molecular connectivity. This provided the starting points for the further transition state searches at the more accurate electronic structure levels to obtain the reaction mechanism. Then the whole reaction network with multiply pathways was obtained. The feasibility and efficiency of this automated construction of the reaction network was examined by two examples. The first reaction under study was the HCHO + NH3 biomolecular reaction. The second example focused on the reaction network for a multi-species system composed of dozens of HCN and H2O compounds. The result indicated that the proposed approach was a valuable and effective tool for the automated exploration of reaction networks.