Dual analysis of degradation mechanisms: Unified models integrate the oxidation–reduction potentials of oxidants with the molecular structures of volatile organic compounds to improve predictive performance

Abstract

The rate constants (k) for the reaction of oxidants with volatile organic pollutants (VOCs) are key parameters to assess the ecological risk of VOCs. Herein, the k datasets of four representative oxidants, namely hydroxyl radicals (·OH), chlorine radicals (·Cl), nitrate radicals (·NO₃), and ozone (O₃), were systematically integrated into a comprehensive Unified dataset. Molecular descriptors, quantum chemical descriptors and Fukui indices were employed to fully characterize VOCs structure. Models for predicting k for four oxidants were developed using six machine learning algorithms (Categorical boosting, Gradient boosting decision tree, Light gradient boosting machine, Random Forest, Extreme gradient boosting and Support vector machine), and the feasibility of improving model performance by using oxidation reduction potential (ORP) as a link to integrate four single oxidant datasets into one unified dataset was explored. Model evaluation indicated that the predictive performance of the unified models (R²_ext = 0.884–0.948) was superior to that of the single oxidant (R²_ext = 0.591–0.884), with the best performance of the unified XGB model (R²_adj = 0.996, Q²_ext = 0.948). According to Shapley’s additive interpretation, the degradation efficiency of VOCs by different oxidants follows the order: ·Cl > ·OH > ·NO₃ > O₃. The degradation process was primarily influenced by the ability of transferring and receiving electrons, van der Waals volume, ionization potential and electronegativity of VOCs. This study demonstrated the effectiveness of combining small datasets of k values to improve the performance of models, providing a useful tool for predicting k values in the atmosphere.