Descriptors of Electrocatalysis for CO <sub>2</sub> Reduction to C <sub>2+</sub> Products Formation
S. K. Midya, Arpan Chakraborty, Subhendu Mishra, Abhishek K. Singh
Abstract
The electrochemical reduction of CO 2 (CO 2 RR) into multicarbon (C 2+ ) products holds great promise for renewable fuel and chemical production. A major challenge, however, lies in designing catalysts that selectively promote C–C bond formation while suppressing competing reactions. To address this, researchers increasingly rely on material descriptors, quantitative properties that relate catalyst structure to performance, to rationalize reactivity trends and streamline computational screening. These descriptors span structural (e.g., coordination environment, geometric confinement), electronic (e.g., oxidation state, d-band center, Bader charge), and thermodynamic (e.g., *CO and *C–C intermediate binding energies) categories, each offering insights into catalytic behavior along the CO 2 RR pathway. Intricate relationships among descriptors make it nearly impossible to identify a single optimal descriptor for CO 2 reduction product selectivity. Complementing these physics-driven approaches, machine learning (ML) has emerged as a transformative tool, enabling rapid prediction of key properties across vast material spaces. ML models trained on DFT or experimental data can reveal hidden correlations. Further, target specific catalysts structures can be generated using the inverse design technique. This accelerates discovery well beyond traditional trial-and-error. The synergy between descriptor-based frameworks and data-driven algorithms is reshaping catalyst design, paving the way for efficient, selective, and scalable CO 2 electroreduction. Looking ahead, integrating ML with dynamic, operando descriptors and real-time experiments will further close the loop on intelligent catalyst optimization.