Abstract
The efficient harnessing of solar power for water treatment via photocatalytic processes has long been constrained by the challenge of understanding and optimizing the interactions at the photocatalyst surface, particularly in the presence of nontarget cosolutes. The adsorption of these cosolutes, such as natural organic matter, onto photocatalysts can inhibit the degradation of pollutants, drastically decreasing the photocatalytic efficiency. In the present work, computational methods are employed to predict the inhibitory action of a suite of small organic molecules during TiO(2) photocatalytic degradation of para-chlorobenzoic acid (pCBA). Specifically, tryptophan, coniferyl alcohol, succinic acid, gallic acid, and trimesic acid were selected as interfering agents against pCBA to observe the resulting competitive reaction kinetics via bulk and surface phase reactions according to Langmuir-Hinshelwood adsorption dynamics. Experiments revealed that trimesic and gallic acids were most competitive with pCBA, followed by succinic acid. Density functional theory (DFT) and machine learning interatomic potentials (MLIPs) were used to investigate the molecular basis of these interactions. The computational findings showed that while the type of functional group did not directly predict adsorption affinity, the spatial arrangement and electronic interactions of these groups significantly influenced adsorption dynamics and corresponding inhibitory behavior. Notably, MLIPs, derived by fine-tuning models pretrained on a vastly larger dataset, enabled the exploration of adsorption behaviors over substantially longer periods than typically possible with conventional ab initio molecular dynamics, enhancing the depth of understanding of the dynamic interaction processes. Our study thus provides a pivotal foundation for advancing photocatalytic technology in environmental applications by demonstrating the critical role of molecular-level interactions in shaping photocatalytic outcomes.