Machine Learning-Assisted Design of Novel High-Entropy Alloy Catalysts for Enhanced Hydrogen Evolution Reaction in Alkaline Media

Alkaline water electrolysis represents a pivotal technology for scalable green hydrogen production; however, its overall efficiency is severely constrained by the sluggish kinetics of the hydrogen evolution reaction (HER), specifically the slow Volmer water-dissociation step. High-entropy alloys (HEAs) offer unprecedented compositional flexibility and tailored electronic structures, yet navigating their vast multi-dimensional chemical space via conventional trial-and-error synthesis remains prohibitively resource-intensive. Here, we present an integrated machine learning (ML)-assisted framework combined with density functional theory (DFT) to screen, design, and synthesize novel quinary non-noble-metal HEA electrocatalysts for alkaline HER. A gradient boosting regressor, trained on electronic and thermodynamic descriptors from 1,200 DFT surface configurations, accurately predicted the hydrogen adsorption free energy (ΔG_H*) and water dissociation barrier (ΔG_diss). Among candidate configurations, a single-phase face-centered cubic (FCC) FeCoNiCrMo alloy was identified as optimal and successfully synthesized via magnetron sputtering. In 1.0 M KOH, the optimized FeCoNiCrMo film exhibited remarkable HER performance, demanding an overpotential of merely 41 mV to achieve a current density of 10 mA cm⁻² with a Tafel slope of 33.8 mV dec⁻¹, rivaling commercial 20 wt.% Pt/C. In-situ electrochemical characterization and extended chronopotentiometry confirmed exceptional durability exceeding 120 hours of continuous operation at 100 mA cm⁻² without apparent structural degradation. This work demonstrates the power of ML-accelerated discovery in pioneering cost-effective, high-performance multi-principal element catalysts for industrial clean energy transition.