Modeling electrochemical nitrogen reduction

Abstract

Despite intense research efforts both computational and experimental, a catalyst able to electrochemically reduce atmospheric nitrogen to ammonia in aqueous media has not been identified. While rigid protocols have been implemented on the experimental side, a lack of agreement between theory and experiments persists. Here, we critically assess the methodology and assumptions employed in constructing the free energy landscape in the bulk of theoretical studies on the electrochemical nitrogen reduction reaction (NRR) with the aim of contributing to better agreement with experiments. The focus is specifically on the treatment of non-electrochemical reaction steps. Moreover, we discuss the use of machine learning models such as deep neural networks (DNNs) for catalyst discovery and point out common pitfalls. Our work shows the promise of DNNs if they are used correctly but also highlights their limitations and the necessity of high-quality data for training. Finally, we gauge the feasibility of the NRR, using high-entropy alloys as a case study.