Prediction of CO₂ Reduction Reaction Intermediates and Products on Transition Metal-Doped γ-GeSe Monolayers: A Combined DFT and Machine Learning Approach

Accurate prediction of free energy changes (ΔG) for the vast network of reaction intermediates in the electrocatalytic CO2 reduction reaction (CO2RR) is essential for evaluating catalytic performance. We combined density functional theory (DFT) and machine learning (ML) to screen 25 single-atom catalysts (SACs) on defective γ-GeSe monolayers for CO2 reduction to methanol, methane, and formic acid. Among nine ML models evaluated with 14 features, the XGBoost performed best (R2 = 0.92 and MAE = 0.24 eV), identifying Ni, Ru, and Rh@GeSe as prospective catalysts. Feature importance analysis highlighted CO2 activation with ∠O–C–O and IPC–O1 as the key attributes. The trained ML model's ΔG predictions closely match DFT-calculated values for the reported Ti@N4–C, Fe@g-C6N6, and Co@g-C6N6. Incorporating non-DFT-based features enabled rapid predictions while retaining model performance. This work identifies effective SACs for CO2RR and offers insights for efficient catalyst design.