Accelerated Discovery of High Performance Ni3S4/Ni3Mo HER Catalysts via Bayesian Optimization

The hydrogen evolution reaction (HER) in alkaline media is a promising strategy for sustainable hydrogen production, but the exploration of efficient and durable HER electrocatalysts is often hindered by the empirical and time-consuming nature of traditional synthesis. Herein, a machine learning (ML)–driven strategy combining Bayesian optimization is introduced to achieve the rational design of Ni 3 S 4 /Ni 3 Mo heterostructures for alkaline HER. By coupling predictive modeling with experimental feedback, this approach efficiently navigated a complex synthesis space and identified conditions yielding a structurally and electronically optimized catalyst. The optimized Ni 3 S 4 /Ni 3 Mo exhibits a vibrant morphological evolution—from compact buds to blooming petal-like structures—enabling enriched active sites and accelerated mass transport. Guided by Bayesian optimization, the optimized Ni 3 S 4 /Ni 3 Mo achieves a 10.5-fold enhancement in C dl and delivers an exceptionally low overpotential of 18.2 mV at 100 mA cm −2 , outperforming most reported transition-metal catalysts and even surpassing commercial Pt/C. Mechanistic insights from in situ Raman and DFT reveal that interfacial charge redistribution between Ni 3 S 4 and Ni 3 Mo optimizes H* adsorption (Δ G H* ≈ 0.04 eV) and significantly reduces the water-dissociation barrier (0.08 eV), thereby accelerating reaction kinetics. This work demonstrates how the ML-guided optimization can synergistically couple morphology control, interfacial engineering with electronic tuning, offering a generalizable framework for intelligent catalyst discovery and mechanistic understanding in electrochemical energy conversion.