Unlocking the Nitrogen\nReduction Electrocatalyst with\na Dual-Metal–Boron System: From High-Throughput Screening to\nMachine Learning

Recently, dual-metal catalysts have attracted much attention\ndue\nto their abundant active sites and tunable chemical properties. On\nthe other hand, metal borides have been widely applied in splitting\nthe inert chemical bonds in small molecules (such as N<sub>2</sub>) because of their excellent catalytic performances. As a combination\nof the above two systems, in this work, 11 kinds of transition metal\natoms (TM = Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, Mo, and W) were selected\nto embed in boron-doped graphene (BG) to construct 66 dual-metal–boron\nsystems, and their performances toward the N<sub>2</sub> reduction\nreaction (NRR) were examined using first-principles simulations. Our\nresults revealed that such a dual-TM@BG system exhibits excellent\nthermodynamic and electrochemical stabilities, which facilitate the\nexperimental synthesis. In particular, Fe–Fe- and Fe–Co-doped\nBG exhibit excellent performance for NRR, with the limiting potentials\nof −0.29 and −0.32 V, respectively, and both of them\nexhibit inhibitory effects on the H<sub>2</sub> evolution reaction.\nRemarkably, the microkinetic modeling analysis revealed that the turnover\nfrequency for the NH<sub>3</sub> production on FeFe@BG reaches up\nto 7.27 × 10<sup>8</sup> s<sup>–1</sup> site<sup>–1</sup> at 700 K and 100 bar, which further confirms its ultrafast reaction\nrate. In addition, the machine learning method was employed to further\nunderstand the catalytic mechanism, and it is found that the NRR performances\nof dual-TM@BG catalysts are closely related to the sum of radii of\ntwo TM atoms. Therefore, our work not only proposed two promising\nelectrocatalysts for NRR but also verified the feasibility for the\napplication of a dual-metal–boron system in NRR.