Reveal the main factors and adsorption behavior influencing the adsorption of pollutants on natural mineral adsorbents: Based on machine learning modeling and DFT calculation

Abstract

Montmorillonite, as a natural mineral adsorption material that has high research value in water pollution treatment. However, the adsorption capacity varies depending on the type of pollutant and the properties of the montmorillonite material, and the factors controlling adsorption are not yet clear. Herein, we investigated the adsorption behavior of pollutants on montmorillonite materials using density functional theory (DFT) calculations and machine learning modeling. Furthermore, it explores the main factors influencing their adsorption. The machine learning results indicate that the gradient boosting decision tree (GBDT) model exhibits a better fit to the experimental data compared to the other five machine learning models (R2 = 0.79). The higher pH levels and larger relative molecular mass of pollutants have a positive impact on montmorillonite adsorption. However, an increase in the proportion of oxygen atoms in the adsorbent material and longer hydrothermal preparation time show a trend of initially positive and then negative effects on the predicted results. The influence of pH on the adsorption capacity of montmorillonite adsorbents was further analyzed using density functional theory (DFT). Density functional theory (DFT) studies reveal that montmorillonite primarily removes protonated sulfamethoxazole (SMZ) through hydrogen bonding (N-H…O) interactions, accompanied by van der waals (O-O) and ionic bond (C-O…Al) forces under different pH conditions. The partial density of states (PDOS) reveals that the LUMO orbital of montmorillonite has a higher electron accepting ability than the HOMO orbital of SMZ (p orbital peak is greater than the S orbital) in the actual electron transfer process. This study provided some support for the investigation of montmorillonite-based adsorbent materials through the combination of machine learning and theoretical calculations.