Quantitative Prediction of the Transformation Potential of Polyfluoroalkyl Substances: A Computational and Machine Learning Study of •OH‑Initiated Initial Reaction Kinetics

Abstract

Per- and polyfluoroalkyl substances (PFAS) are ubiquitous persistent organic pollutants, and hydroxyl radical (•OH)-initiated abiotic transformation dominates the environmental fate of polyfluoroalkyl substances. However, quantitative relationships between molecular structures and the •OH-initiated transformation reactivity remain poorly understood. Herein, we combined density functional theory (DFT) calculations and machine learning (ML) modeling to establish a predictive framework for the •OH-initiated initial transformation potential of polyfluoroalkyl substances. The Gibbs free energy barriers of the rate-determining hydrogen atom abstraction (HAA) step were computed at the M06-2X/6-311++G(2d,2p) level with the SMD solvation model for polyfluoroalkyl substances with varied functional groups, perfluoroalkyl chain lengths, C-H bond positions, alkyl spacers, and branched configurations. The results revealed that perfluoroalkyl chain length imposed negligible effects on HAA barriers, whereas functional groups governed the transformation reactivity in the order of $-P{O_4}^{2-}>-OH≈-CO{O}^->-S{O_3}^-.$ The positive charge density of aliphatic C-H hydrogens served as a decisive descriptor, in which stronger C-H polarization induced higher HAA barriers. Alkyl spacers shielded the electron-withdrawing effect of fluorine, while branched CF moieties enhanced C-H polarization. The optimal machine learning model (RF-MACCS) achieved excellent predictive performance $(R^2=0.91),$ and its practical application on environmentally relevant polyfluoroalkyl substances yielded predictions highly consistent with DFT results. SHAP analysis verified the mechanistic consistency between model predictions and DFT-derived structure-reactivity relationships. This work establishes a robust predictive framework, providing a high-throughput tool for the environmental risk assessment and preliminary screening of polyfluoroalkyl alternatives with controlled transformation behavior.