Mechanistic Insights and SISSO Machine Learning Prediction of Activation Energies in Phosphate Triester Hydrolysis

Phosphate triester hydrolysis is investigated using density functional theory (DFT) calculations combined with interpretable machine learning (ML) to elucidate reaction mechanisms, structure-reactivity relationships and develop predictive models of activation barriers. Using dimethyl 4-nitrophenyl phosphate with three explicit water molecules as a model, both concerted and stepwise hydrolysis pathways were examined. The concerted pathway proceeds through a single transition state (TS) with a high activation energy of 49.1 kcal/mol. In contrast, the stepwise mechanism involves nucleophilic attack at phosphorus to form a pentacoordinate intermediate, followed by P─O bond cleavage, with activation energies of 29.4 and 8.4 kcal/mol, respectively. These results indicate that intermediate formation is rate-determining and that the stepwise pathway is kinetically favored. Structure-activity relationship analysis further reveals that phosphate triesters bearing electron-deficient aromatic substituents, such as nitro- or chloro-substituted groups, together with phenoxide leaving groups, exhibit enhanced hydrolytic reactivity. Finally, a predictive model based on the Sure Independence Screening and Sparsifying Operator (SISSO) method is developed using descriptors derived solely from isolated phosphate triester molecules, enabling reliable prediction of activation barriers with significantly reduced computational cost.