A first-Principles and machine learning study of electrocatalytic nitrate reduction to ammonia on single-atom covalent organic frameworks

Electrocatalytic nitrate reduction to ammonia (NO 3 RR) is critically important for human health, agriculture production and industry. However, this complex process involves multi-proton/electron transfers and byproduct formation, driving the requirement for highly efficient and selective electrocatalysts. In this work, we performed a comprehensive investigation on the catalytic activities of a series of TM-COFs by combining density functional theory (DFT) and machine learning. Ti–COF was identified as the optimal NO 3 RR electrocatalyst, owing to its low limiting potential (−0.19 V), high NH 3 selectivity, and excellent thermodynamic and electrochemical stability, which jointly ensure superior performance. This exceptional activity originates from balanced electronic property coupled with the favorable local structure and chemical environment of the Ti active site. Additionally, employing Sure Independence Screening and Sparsifying Operator, we established explicit structure-activity relationships through two analytical formulas linking adsorption energy of NO 3 − and limiting potential to critical features. This work establishes a computational strategy for rapid screening and rational design of electrocatalysts. • Ti–COF is the optimal NO 3 RR electrocatalyst due to its low limiting potential, high selectivity, and robust stability. • The activity of Ti–COF stems from its balanced electronic structure and favorable Ti coordination environment. • Using SISSO, we derived two analytical formulas that explicitly define the NO 3 RR structure-activity relationship.