Inverse design and discovery of high entropy alloy catalysts for efficient oxygen evolution reaction

High-entropy alloys (HEAs) consisting of earth-abundant elements can serve as replacement for expensive, industrial grade catalysts, such as iridium dioxide IrO2, for oxygen evolution reactions (OERs). However, navigating the exponential configuration space of possible HEAs and unknown inverse structure-property relationship has hindered the development of HEAs for efficient electrocatalytic applications. In this paper, using OH adsorption energy as a descriptor of optimal catalyst property, machine-learning-based frameworks are utilized for discovering HEA catalysts. Each designed HEA is composed of 5 elements chosen from Mn, Fe, Co, Ni, Cu, Zn, and Mo specific to OER. Calculated theoretical overpotential for one HEA catalyst composed of MnCoNiCuZn is found to be ηtheor = 0.67 V. However, deviations from scaling results can be as high as 0.99 V when compared to explicit calculations. After screening more than 163 × 109 HEA configurations using a regression model, MnCoNiCuZn with a stoichiometric ratio of 2, 2, 8, 12, and 12 displayed the highest likelihood of positive OH binding energy/active sites. To investigate the intrinsic catalytic activity of the MnCoNiCuZn HEA in an operational OER environment, a computational hydrogen electrode model was constructed to represent the oxyhydroxide layer of OER catalysts, with the MnCoNiCuZn placed at the surface. Analysis indicates that upon consistent comparison, the most stable rutile IrO2 (η = 0.33 V) exhibits superior performance relative to the proposed MnCoNiCuZn catalyst (η = 0.57 V). However, the present methodology, which resulted in statistically favorable adsorption sites, could be successfully used to discover highly active sites within a compositional space.