Machine Learning-Assisted DFT Screening of Nitrogen-Doped Graphene Diatomic Catalysts for Nitrogen Reduction Reaction

This research seeks to investigate extremely efficient catalysts for the nitrogen reduction process (NRR), utilizing machine learning (ML)-aided density functional theory (DFT) computations. Specifically, we investigate dual transition metal atoms anchored on hexagonal nitrogen-doped graphene (TM₁-TM₂@N₆G) as prospective high-activity catalysts for the NRR. The findings indicate that the synergistic effect of dual transition metal atoms in the TM₁-TM₂@N₆G catalyst overcomes the intrinsic constraints of the linear relationship among intermediates, facilitating the activation and adsorption of N₂, thereby exhibiting significant potential for ammonia synthesis through N₂ reduction. Particularly, four catalysts screened by ML and DFT exhibit good stability and excellent selectivity and activation towards N₂. Among them, the catalysts Ti-Cr@N₆G, Ti-Mo@N₆G, and Ti-Pd@N₆G possess two reaction pathways with minimum reaction energies of 0.55 eV, 0.50 eV, and 0.40 eV, respectively. Remarkably, Ti-Co@N₆G, which features a single reaction pathway, exhibits a reaction energy lower than 0.05 eV, allowing the NRR to proceed spontaneously. It is noteworthy that incorporating ML into DFT calculations facilitates the rapid screening of all transition metal combinations, significantly accelerating the research on catalytic performance and optimizing the selection of catalysts.