Machine learning-assisted screening of copper–cerium bimetallic catalysts for tetracycline degradation

Abstract

Bimetallic catalysts have shown significant potential for refractory organic pollutant removal. However, the individual preparation and testing of candidate materials are both time-consuming and impractical. In this study, a d-band center guided machine learning (ML) approach was proposed to rapidly screen copper–cerium bimetallic catalysts for H₂O₂ activation. Integrating density functional theory (DFT) with ML models, over 2000 bimetallic combinations were evaluated using the d-band center as the key descriptor. The ML model achieved excellent prediction for ·OH adsorption energies (R² = 0.913). CuCe exhibited the largest d-band center shift, indicating superior electronic tunability. DFT calculations further revealed that the optimal Cu₁Ce₃ catalyst possesses a d-band center of −1.32 eV and a ·OH adsorption energy of −1.37 eV. Experimentally, the Cu₁Ce₃/H₂O₂ system achieved 99.4% tetracycline degradation and maintained 72% removal after eight cycles. This study provides a practical ML-assisted strategy for designing high-performance bimetallic catalysts in advanced oxidation processes.