Self-single-atomization of active sites in electrochemical ammonia synthesis unveiled by machine learning potential

Electrochemical nitric oxide (NO) reduction reaction (eNORR) has emerged as a promising route for the synthesis of green ammonia. Fe was observed experimentally with great activity and selectivity for eNORR to ammonia. However, we found that electrochemical performance over iron is not due to its intrinsic activity, but from an adaptive capability of self-single-atomization in operando conditions. We first performed Grand Canonical Monte Carlo simulations driven by a machine learning potential, where we explored ∼2.5 million possible active structures and their evolution with increasing electrode potential. Microkinetic modeling indicates that Fe surface will be predominantly covered by adsorbed N* species in reaction, resulting in the formation of FeN _x motifs by self-single-atomization. Statistical results indicate that the top sites of isolated FeN _x motifs contribute higher activity of ammonia synthesis than bridge and hollow sites. Furthermore, based on the self-single-atomized active sites, we reproduced and rationalized the variation trends of experimental Faradaic efficiency of different products with electrode potential, verifying the reliability of the proposed active sites herein. Hence, we propose to make use of the self-single-atomization capability of catalysts under working conditions to construct active sites beyond ammonia synthesis.