(666g) Simulating Chemical Reactions at Working Electrode Interfaces with Constant Potential Neural Network Potentials

Simulating charge transfer reactions at electrochemical interfaces poses a unique set of challenges. It is essential to capture (1) the motion of charges within the metal electrode, electrolyte solution and between the two phases, (2) how this changes as an electric potential is applied to the electrodes and the total number of electrons in the system fluctuates, and (3) the specific and dynamically changing chemical bonding and adsorption interactions within the electrolyte and between the electrolyte and electrode as chemical reactions are occurring. Moreover, describing reactions requires us to sample a large ensemble of rare reactive events requiring extremely efficient computational methods. In this talk, we will describe how High Dimensional Neural Network Potential Energy models can be combined with Classical Constant Electric potential methods used to treat charged conductors to simultaneously address these challenges and allow us to simulate heterogeneous electrochemical reactions at working electrodes with near ab initio accuracy at costs near classical molecular dynamics. This new simulation technique allows us to apply powerful tools from statistical mechanical rate theories to rigorously study electrochemical reactions at atomic resolution.