Discovering Descriptors for CO 2 ‐to‐Methanol Electrosynthesis on Cu‐Based Alloy Catalysts by Machine Learning

Abstract The electrochemical CO 2 reduction reaction to methanol (CH 3 OH) suggests an attractive approach toward carbon neutrality, while the activity and selectivity of CO 2 ‐to‐CH 3 OH are substantially low compared to most of the other C 1 products like CH 4 , largely due to the much less preferred thermodynamic CH 3 OH formation pathway and the limited capability of controlling the adsorption of key intermediates toward CH 3 OH. Herein, a machine learning approach with first‐principles calculations is demonstrated to identify key descriptors that govern the protonation of * CO intermediates and prevent C─C coupling, which are critical factors in determining product selectivity toward the CH 3 OH formation. Datasets linking the intrinsic structural properties of heteroatom‐doped Cu catalysts to their catalytic performances are developed, among which the Bader charge, bond length, and d ‐band center are determined to be the key descriptors. The electronic structure analysis reveals that RhCu alloys demonstrate both enhanced selectivity and catalytic activity for the CO 2 ‐to‐CH 3 OH synthesis, with the d ‐band center shifting toward the Fermi level by 0.18 eV. This shift mitigates the competing hydrogen evolution reaction and simultaneously enhances the adsorption of key intermediates ( * CHO and * CO). This work presents a cost‐effective data‐driven approach for designing high‐performance Cu‐based catalysts toward selective CO 2 ‐to‐CH 3 OH electrosynthesis.