Decoding Interfacial Interactions Governing Organic Compounds Removal: Transfer of Feature Engineering from TiO 2 to Clay Minerals

Abstract

Interfacial interactions between organic molecules and inorganic substrates play an important role in ion exchange, immobilization, and adsorption, which affect contaminant removal in aquatic environments. Here, we develop a transferable model using readily accessible descriptors that can be applied across distinct organic–inorganic interfacial systems. The model was trained on density functional theory (DFT) data of three antibiotics (tetracycline, ciprofloxacin, and berberine hydrochloride) adsorbed on the modified TiO2 surface, and implemented to predict adsorption energies of antibiotics on three other kinds of clay minerals (rectorite, montmorillonite, and vermiculite) with satisfactory performance (R2 = 0.75). The uppermost Ti atoms on the TiO2 surface are the potential active sites for hosting organic molecules, whereas the surface oxygen atoms on the clay minerals act as the primary interaction sites for adsorption. Using the Sure Independence Screening and Sparsifying Operator (SISSO) algorithm, we constructed composite descriptors of interfacial charge transfer and hydrogen bonding preference to rationalize the experimental removal efficiency of tetracycline on various substrates (with versus without light illumination) as well as the dye on TiO2. Applying this framework to methylene blue on P25 TiO2, we show that surface polarization and the dye’s electronic structure properties dictate adsorption and photocatalytic efficiency, which can be further enhanced through Bayesian-optimized reaction conditions. This study provides a transferable, physically interpretable framework for understanding and predicting complex organic–inorganic interfacial interactions, and offers mechanistic guidance for designing functional interfaces for contaminant removal.