Artificial intelligence and machine learning in photocatalytic nanomaterials: applications across diverse environmental pollutants

Escalating environmental pollution, particularly from persistent organic pollutants, dyes, pharmaceuticals, and other recalcitrant contaminants in wastewater and air, demands innovative, efficient, and sustainable remediation technologies. Nanomaterials-based photocatalysis is a solar-driven advanced oxidation process that generates reactive electron-hole pairs that initiate redox reactions for pollutant degradation. However, this conventional method relies heavily on resource-intensive, time-consuming trial-and-error experimentation. This approach struggles with the complex, non-linear interplay among nanomaterial properties, operational parameters, and environmental conditions. Machine learning (ML) and artificial intelligence (AI) are emerging as transformative tools to overcome these barriers. By leveraging large experimental and computational datasets, ML models enable accurate prediction of photocatalytic degradation efficiency, identification of key performance-influencing factors via feature importance analysis, and rational design of optimized photocatalysts. This review synthesizes recent advances (primarily from 2020 to 2026) in integrating ML/AI with photocatalytic nanomaterials for various environmental pollutant degradation. The review found that boosting-based ensemble algorithms and hybrid machine learning models possess the highest predictive performance in water and air pollutant degradations. They also predict degradation rate constants and concentrations. This is due to their generalization capability for the non-linear complex photocatalytic process. However, the scarcity and inconsistency of datasets, as well as the difficulty in identifying the input variables that may result in black-box problems, remain major challenges.