Phase redistribution of chlorobenzene in compacted low-permeability soil under coupled environmental conditions experiments, machine learning, and molecular dynamics.

Phase redistribution of organic contaminants in low-permeability soils strongly influences their persistence, mobility, and remediation, its dependence on coupled environmental conditions remains poorly understood. In this study, chlorobenzene was selected as a representative dense non-aqueous phase liquid contaminant, and its redistribution among vapor, free, dissolved, and adsorbed phases in low-permeability soil was investigated by varying temperature, relative humidity, and dry density. A sequential multi-phase extraction framework was developed to quantify phase-specific fractions, while tree-based machine learning was used to predict redistribution and identify the relative importance of environmental drivers. Molecular dynamics simulations were further conducted to interpret observed trends from the perspective of adsorption-desorption and confined diffusion at the clay scale. Results showed that the vapor phase dominated under most conditions, whereas the free phase remained suppressed. Temperature exerted the strongest control by promoting volatilization and weakening interfacial retention, while relative humidity and dry density jointly regulated aqueous and adsorption domains through changes in water film development and pore structure. This integrated framework provides a quantitative basis for evaluating contaminant persistence and optimizing remediation strategies. In particular, the vapor-phase dominance under thermal driving indicates that vapor extraction and capture should be prioritized during thermally enhanced remediation of chlorobenzene-contaminated low-permeability soils.