摘要:

The oxidation of organic compounds by nitrate radicals (NO3) in the atmospheric aqueous phase makes significant contributions to the production of secondary organic aerosols (SOA) and brown carbon (BrC). However, relevant kinetic parameters remain scarce, particularly in aerosol liquid water (ALW), where high concentrations of inorganic ions coexist. In this study, we developed a predictive model to estimate the aqueous-phase reaction rate constants between NO3 and organic compounds (kNO3) using a novel machine learning (ML)-based approach. By simultaneously considering the chemical properties of the reactants and experimental conditions, this model enables an accurate prediction of kNO3 across ionic strength (I) ranges of 0–6 M, while also accounting for the influence of different ionic species. The model's predictive accuracy, generalization ability, and applicability domain (AD) are evaluated, followed by a mechanistic interpretation via Shapley additive explanation (SHAP) analysis. In summary, this study provides a valuable supplementary tool for estimating kinetic parameters in both cloud droplets and ALW. As an application, the model is employed to predict kNO3 values for phenolic compounds emitted from biomass burning (BB), extending the currently available data set for atmospheric modeling.

访问链接