Revisit the actual roles of catalytic sites in a Fenton-like system

Abstract

The perception of catalytic site contributions in peroxymonosulfate (PMS)/biochar systems is biased due to the neglect of active site interactions. Here, random forest regression (RF), supporting vector regression (SVR), XGBoost (XGB), and gradient boosting decision tree (GBDT) are selected to construct models using active sites (A model) instead of elements (E model) as input features to revisit the relationship between C, N, and O-containing sites on biochar surface and Fenton-like activity. Consequently, the A models achieve twice the accuracy of the E models. For individual sites, a low CC or high CN concentration promotes the degradation of electron-donating organics in PMS systems, while system activity initially increases and then declines with rising CO concentration. Considering site interactions, CC&CO, CC&CN, and CO&CN show excellent synergy for PMS activation. Specifically, the Gibbs free energy (ΔG) of PMS at CC&CO (0.44 eV), CN&CO (0.78 eV), and CC&CN (0.82 eV) is significantly lower than that of single CC (2.26 eV), CN (1.44 eV), and CO&CO (1.14 eV) during the transition state formation process. The reduced ΔG facilitates OO bond cleavage, enhancing the generation of active species. This study employs A_SVR model and density functional theory (DFT) calculations to clarify the structure–activity relationships of biochar by considering the synergistic effects of active site. The results contribute to the precise design of Fenton-like catalysts for targeted pollutant degradation, improving water purification efficiency.