Dominant Factors Governing Benzene Adsorption in Soils: Thermodynamic Analysis and Predictive Modeling

Abstract

Understanding the processes controlling benzene adsorption in soils is critical for predicting its environmental fate and associated risks. However, the adsorption behavior of benzene across different soil components and under varying environmental conditions remains insufficiently constrained. In this study, adsorption thermodynamics of benzene on representative soil components, including humic acid, kaolinite, montmorillonite, birnessite, and goethite, were systematically investigated, together with the effects of temperature, pH, and coexisting Pb²⁺. Batch adsorption experiments were integrated with machine learning approaches to quantify adsorption behavior and identify key controlling factors.Results showed that humic acid exhibited a substantially higher benzene adsorption capacity than mineral components, with saturated adsorption capacities at 25 °C following the order: humic acid > birnessite > montmorillonite > goethite > kaolinite. Thermodynamic analysis indicated that benzene adsorption on all components was spontaneous, exothermic, and entropy-decreasing, suggesting a process dominated by physical adsorption. Among the examined environmental factors, temperature exerted a significantly stronger influence on adsorption equilibrium than pH and coexisting Pb²⁺, with increasing temperature markedly suppressing benzene adsorption, particularly for humic acid. A machine learning prediction model was constructed using 395 experimental datasets. Among the tested models, the Random Forest model showed the best predictive performance (R² = 0.97, RMSE = 1.12 mg g⁻¹). Feature importance analysis revealed that benzene addition concentration, specific surface area, micropore volume, and total pore volume were the dominant factors controlling adsorption behavior, collectively accounting for over 80% of the adsorption behavior. These findings provide process-based insights into soilbenzene interactions and offer a favorable predictive tool for assessing the environmental behavior and remediation potential of benzene-contaminated soils.