Active learning-driven multi-objective design of high-entropy alloy catalysts for saline water electrolysis

초록

Saline water electrolysis has emerged as a promising strategy for sustainable hydrogen production due to the abundance of seawater resources. However, commercialization is hindered by reliance on precious-metal catalysts and the difficulty of maintaining selective reaction pathways under saline conditions. Here, we introduce a high-entropy alloy (HEA) catalyst design strategy enabled by an active learning, multi-objective framework that simultaneously optimizes activity, selectivity, and cost. By combining Gaussian process regression with the expected improvement acquisition, we efficiently explore the vast compositional space of HEAs using only a limited number of density functional theory calculations, achieving high predictive accuracy at ∼1.2% of the computational cost of full-space exploration. Computational results reveal distinct optimal compositions for seawater electrolysis. Cu-rich HEA compositions at the anode maximize oxygen evolution reaction activity while effectively suppressing the competing chlorine evolution reaction, providing a cost-effective alternative to conventional noble-metal-based catalysts. In addition, Ni-rich compositions exhibit hydrogen evolution reaction activity comparable to that of pure Pt while significantly reducing precious-metal content at the cathode. Overall, this active learning-driven, multi-objective protocol accelerates HEA discovery for seawater electrolysis and is readily transferable to other electrochemical energy-conversion and catalyst-design problems requiring simultaneous optimization of multiple competing objectives. © 2026 Elsevier B.V.

키워드