Capturing Dichotomic\nSolvent Behavior in Solute–Solvent\nReactions with Neural Network Potentials

Simulations of chemical reactivity in condensed phase\nsystems represent\nan ongoing challenge in computational chemistry, where traditional\nquantum chemical approaches typically struggle with both the size\nof the system and the potential complexity of the reaction. Here,\nwe introduce a workflow aimed at efficiently training neural network\npotentials (NNPs) to explore energy barriers in solution at the hybrid\ndensity functional theory level. The computational burden associated\nwith training at the PBE0-D3(BJ) level is bypassed through the use\nof active and transfer learning techniques, whereas extensive sampling\nof the transition state region is accelerated by well-tempered metadynamics\nsimulations using multiple time step integration. These NNPs serve\nto explore a puzzling solute–solvent reactivity route involving\nthe ring opening of <i>N</i>-enoxyphthalimide experimentally\nobserved in methanol but not in 2,2,2-trifluoroethanol (TFE). This\nreaction represents a challenging example characterized by intricate\nhydrogen bonding networks and structurally ambiguous solvent-sensitive\ntransition states. The methodology successfully delivers detailed\nfree energy surfaces and relative energy barriers in agreement with\nexperiment. These barriers are associated with an ensemble of transition\nstates involving the direct participation of up to five solvent molecules.\nWhile this picture contrasts with the single transition state structure\nassumed by current static models, no drastic qualitative difference\nis observed between the formed hydrogen bonding networks and the number\nof participating solvent molecules in methanol or TFE. The dichotomy\nbetween the two solvents thus essentially arises from an electronic\neffect (i.e., distinct nucleophilicity) and from the larger conformational\nentropy contributions in methanol. This example underscores the critical\nrole that dynamic simulations at the ab initio levels play in capturing\nthe full complexity of solute–solvent interactions.