Machine Learning-Enhanced Screening of Single-Atom Alloy Clusters for Nitrogen Reduction Reaction

The electrochemical nitrogen reduction reaction (eNRR) under ambient conditions is a promising method to generate ammonia (NH₃), a crucial precursor for fertilizers and chemicals, without carbon emissions. Single-atom alloy catalysts (SAACs) have reinvigorated catalytic processes due to their high activity, selectivity, and efficient use of active atoms. Here, we employed density functional theory (DFT) calculations integrated with machine learning (ML) to investigate dodecahedral nanocluster-based SAACs for analyzing structure-activity relationships in eNRR. Over 300 nanocluster-based SAACs were screened with all the transition metals as dopants to develop an ML model predicting stability and catalytic performance. Facet sites were identified as optimal doping positions, particularly with late group transition metals showing superior stability and activity. Utilizing DFT+ML, we identified 8 highly suitable SAACs for eNRR. Interestingly, the number of valence d-electrons in dopants proved crucial in screening for eNRR activity. These catalysts exhibited low activity in hydrogen evolution reaction, further enhancing their suitability for eNRR. This successful ML-driven approach accelerates catalyst design and discovery, holding significant practical implications.