Density Functional Theory and Machine Learning-Based Quantitative Structure-Activity Relationship Models Enabling Prediction of Contaminant Degradation Performance with Heterogeneous Peroxymonosulfate Treatments.

Abstract

Heterogeneous peroxymonosulfate (PMS) treatment is recognized as an effective advanced oxidation process (AOP) for the treatment of organic contaminants. Quantitative structure-activity relationship (QSAR) models have been applied to predict the oxidation reaction rates of contaminants in homogeneous PMS treatment systems but are seldom applied in heterogeneous treatment systems. Herein, we established QSAR models updated with density functional theory (DFT) and machine learning approaches to predict the degradation performance for a series of contaminants in heterogeneous PMS systems. We imported the characteristics of organic molecules calculated using constrained DFT as input descriptors and predicted the apparent degradation rate constants of contaminants as the output. The genetic algorithm and deep neural networks were used to improve the predictive accuracy. The qualitative and quantitative results from the QSAR model for the degradation of contaminants can be used to select the most appropriate treatment system. A strategy for selection of the optimum catalyst for PMS treatment of specific contaminants was also established according to the QSAR models. This work not only increases our understanding of contaminant degradation in PMS treatment systems but also highlights a novel QSAR model to predict the degradation performance in complicated heterogeneous AOPs.