BookR: School of Health Core Facilities Booking

The lack of active and selective catalysts hinders transition to the electrochemical nitrogen reduction reaction (NRR) as an environmentally friendly route for on-site ammonia production under ambient conditions. In search of active NRR electrocatalysts, we generated a dataset of surface and ordered intermetallic alloys using density functional theory (DFT) and trained an artificial neural network (ANN) using characteristics of the active-site transition-metal electronic d-states. The electrochemical limiting potential of ~350 surface alloys with varying surface configurations was predicted using our developed ANN, with a mean absolute error of 0.23 eV comparable to that of DFT. The full energy profiles of Au@Au₃Re and Au@Au₃Mo as potential candidates were further characterized and compared with those of pure Re(001) and Mo(110) via DFT calculations to understand the enhanced catalytic activity rooted in the electronic structure of the catalyst. Moreover, charge analysis showed significant charge transfer from Re and Mo to Au in these alloys, resulting in a change of their electronic structure, and improvement of their catalytic activity. Finally, the selectivity of the alloys was investigated by comparing the adsorption free energy of nitrogen and hydrogen adatoms, and the resulting theoretical faradaic efficiency. This work further confirms that alloying is an effective approach to enhance the catalytic activity of transition metals and highlights how machine learning algorithms trained with physically intuitive features of the materials can efficiently screen the chemical space of bimetallic alloys and predict the limiting potential for a reaction such as the NRR over these alloys, thereby reducing the computational cost of alloy catalyst design and providing an affordable path to electrocatalytic materials discovery.