Supervised AI and Deep Neural Networks to Evaluate High-Entropy Alloys as Reduction Catalysts in Aqueous Environments

Competitive surface adsorption energies on catalytic surfaces constitute a fundamental aspect of modeling electrochemical reactions in aqueous environments. The conventional approach to this task relies on applying density functional theory, albeit with computationally intensive demands, particularly when dealing with intricate surfaces. In this study, we present a methodological exposition of quantifying competitive relationships within complex systems. Our methodology leverages quantum-mechanical-guided deep neural networks, deployed in the investigation of quinary high-entropy alloys composed of Mo–Cr–Mn–Fe–Co–Ni–Cu–Zn. These alloys are under examination as prospective electrocatalysts, facilitating the electrochemical synthesis of ammonia in aqueous media. Even in the most favorable scenario for nitrogen fixation identified in this study, at the transition from O and OH coverage to surface hydrogenation, the probability of N2 coverage remains low. This underscores the fact that catalyst optimization alone is insufficient for achieving efficient nitrogen reduction. In particular, these insights illuminate that system consideration with oxygen- and hydrogen-repelling approaches or high-pressure solutions would be necessary for improved nitrogen reduction within an aqueous environment.