Accelerated discovery of M₆@g-N₄ catalysts for CO₂ electroreduction via machine learning and DFT: descriptor...

Graphitic nitrogen-doped graphene (g-N₄)-supported M₆ metal clusters are promising candidates for efficient CO₂ electroreduction (CO₂RR). However, traditional trial-and-error experiments and computationally intensive DFT calculations hinder the rapid development of high-performance catalysts. Herein, we integrate machine learning (ML) with DFT to screen and predict the CO₂RR performance of M₆@g-N₄ catalysts, where M represents 36 transition/main group metals (excluding unstable Na/K/Ir/Hg clusters). A DFT-derived dataset covering 16 structural, electronic, and physicochemical descriptors was constructed, and 8 ML algorithms were systematically evaluated. Ridge regression (RR) emerged as the optimal model, achieving a high coefficient of determination (R² = 0.963) and low root mean square error (RMSE = 0.228) with strong anti-multicollinearity and interpretability. Pearson correlation and RR-based feature importance analyses revealed that Bader charge transfer, hydrogen evolution reaction (HER) competition, and CO₂ structural distortion (∠OCO and C-O bond length) are the dominant activity descriptors. The ML predicted top-performing catalysts for CO₂RR were Cd₆@g-N₄, Zn₆@g-N₄, and Sn₆@g-N₄, which were further validated by additional DFT calculations on the *CO₂ → *CO reaction pathway. This work demonstrates that the integration of ML and DFT provides a data-driven route to accelerate the discovery of high-performance CO₂RR catalysts, offering quantitative guidance for materials design and contributing to climate-change mitigation.