Accelerating the discovery of active and selective CO2RR catalysts using a high-throughput virtual screening strategy

Developing efficient and selective catalysts for CO2 reduction reaction (CO2RR) is crucial for mitigating climate change. This work presents a high-throughput virtual screening (HTVS) workflow combining machine learning (ML) models and CO2RR selectivity maps to identify promising CO2RR catalysts. The workflow utilizes a structure-free active motif-based representation (DSTAR) for binding energy prediction, enabling the exploration of a vast chemical space encompassing 2,463,030 active motifs. A potential-dependent 3D selectivity map, utilizing three binding energies (ΔE_CO, ΔE_OH, ΔE_H), accurately predicts the activity and selectivity of various catalysts. Experimental validation of Cu-Pd and Cu-Ga alloys confirmed the high selectivity for C1+ and formate, respectively, aligning with the predictions. The HTVS strategy significantly accelerates the discovery of active and selective CO2RR catalysts.

D. H. Mok, H. Lee, G. Zhang, C. Li, Kun Jiang, Seoin Back