Unveiling the Role of Solvent in Solution Phase Chemical Reactions Using Deep Potentials-based Enhanced Sampling Simulations

We have used the deep learning-based active learning strategy to develop ab initio accurate machine-learned (ML) potential for a solution phase reactive system. This approach enables us to study the effect of solvents on chemical reactions. As a paradigmatic example, we have investigated the Menshutkin reaction, a classic bimolecular nucleophilic substitution SN2 reaction in aqueous medium. Enhanced sampling simulations using the ML potential enabled efficient sampling of multiple transitions between the reactant and the product states, allowing us to calculate the converged free energy surface. Our analysis revealed that water stabilizes the product state, facilitating the reaction and making it more spontaneous. Our approach expands the scope of studying chemical reactions in explicit solvents at finite temperatures, closely mimicking experiments.