Machine-Learning Accelerated DFT to Predict Activation Energies on Amorphous Nanocatalysts for the Chlorine Evolution Reaction

In order to improve the production efficiency of chlorine (Cl2) and reduce production costs, electrocatalysts with high activity are essential. The conventional catalysts such as dimensionally stable anode (DSA) have the disadvantages of high cost and poor selectivity. Amorphous alloy is a promising candidate for a low-cost and efficient chlorine evolution reaction (CER) electrocatalyst. In this letter, we predicted an efficient amorphous NiFeP catalyst by using machine-learning (ML) accelerated density functional theory (DFT). There is a problem of insufficient adsorption sites in traditional DFT calculations. To deal with this insufficiency, we developed a distance contribution descriptor for ML feature engineering and calculated the Gibbs free energies (Delta G Cl) of 50400 Cl* binding sites on the surface of Ni40Fe40P20 by ML-accelerated DFT. At the same time, the analysis of catalytic reaction pathways, bonding configurations, and surface adsorption capacity shows that the CER is more inclined to the Volmer-Heyrovsky pathway, and the Ni and Fe atoms contribute to the adsorption of Cl-, while the P atom contributes to the desorption of Cl-. These research methods provide an idea for predicting the activation energies of amorphous nanocatalysts.