Theoretical High-Throughput Screening of Single Atom CO2 Electroreduction Catalysts to Methanol Using Active Learning

Industrial decarbonization is critical for achieving net-zero goals, with the CO2 electrochemical reduction reaction (CO2RR) being a promising approach to converting CO2 into high-value chemicals and offering the potential for decarbonizing industrial processes toward a sustainable, carbon-neutral future. However, developing CO2RR catalysts with high selectivity and activity remains a challenge due to the complexity of finding such catalysts and the inefficiency of traditional computational or experimental approaches. We present a methodology integrating density functional theory (DFT) calculations, deep learning models, and active learning strategies to rapidly screen high-performance catalysts and demonstrate it on graphene-based single-atom catalysts for selective CO2 electroreduction to methanol. First, we conduct systematic binding energy calculations for 3,045 single-atom catalysts to identify thermodynamically stable catalysts as the design space. We then use a graph neural network, fine-tuned with a specialized adsorption energy database, to predict the relative activity and selectivity of candidate catalysts. An autonomous active learning framework facilitated the exploration of designs. After six learning cycles and 2,180 adsorption calculations across 15 intermediates, we developed a surrogate model that identified four novel catalysts on the Pareto front of activity and selectivity. Our work demonstrates the effectiveness of combining finetuning large foundation model and an active learning framework, and it can significantly accelerate the discovery of high-performance CO2RR catalysts.