Bimetal-Anchored Single-Atom Catalysts for Superior Nitrogen Reduction: Synergistic DFT and Machine Learning Design

Electrochemical N₂-to-NH₃ conversion is attracting intense interest, yet catalysts that combine high activity, selectivity, and stability remain scarce. Here, machine-learning (ML) surrogate models with density-functional theory (DFT) were integrated to rationally design bimetallic-supported single-atom catalysts (M₃@M₁M₂). We propose a fast, two-step screening protocol: (1) train a machine-learning surrogate on a small DFT data set to rank hundreds of bimetallic-supported single-atom candidates in seconds, and (2) recheck the top hit with full DFT calculations, delivering an accurate, low-cost pathway to discover high-performance NRR electrocatalysts. For this kind of catalyst, the DFT calculation shows that the first or last step (N₂ → NNH or NH₂ → NH₃) is the potential limiting step of NRR reaction. The correlation coefficient is more than 0.95, and the mean square error is less than 0.05. Machine-learning predictions singled out Ru@AgNi as the most active NRR catalyst; this was immediately confirmed by DFT, which gave a potential limiting Gibbs free energy hurdle of only 0.454 eV. Ru@AgNi sits at the summit of the NRR activity volcano, outperforming all of the surveyed catalysts and underscoring its exceptional nitrogen-reduction performance. This machine-learning-guided breakthrough enlarges the design space for high-performance, low-cost nitrogen-reduction electrocatalysts and sheds light on the application of bimetal-anchored single-atom catalysts in nitrogen reduction reaction.