Data-driven optimisation of more sustainable high-to-medium entropy Perovskites for oxygen evolution catalysts.

An initial library of thirty (minimal ruthenium) high-to-medium entropy La(RuwNixMnyFez)O3 oxygen evolution reaction (OER) catalysts were synthesised via a two-step process involving continuous hydrothermal flow synthesis (co-precipitation), followed by heat-treatment in air at 800 °C for 8 h. Structural analysis using powder X-ray diffraction (XRD) suggested the formation of a phase-pure orthorhombic LaFeO3-type Perovskite structure (Pnma) in all samples except the most Ru-rich composition, which additionally showed minor rutile RuO2 peaks. Electrochemical OER screening identified two initial champions that were La0.94Ru0.46Ni0.10Mn0.16Fe0.34O3 (exhibited high performance with a moderate Ru content) and La1.00Ru0.16Ni0.30Mn0.22Fe0.32O3 (low Ru content catalyst), with the latter delivering an overpotential of 325 mV at 10 mA cm-2 and excellent stability of 70 h at 10 mA cm-2 (only 4% decay). Subsequently, the electrochemical data were used to inform machine learning (ML)-guided synthesis of twelve new compositions spanning moderate- and low-Ru contents. This approach resulted in samples with improved OER performance, which included sample La1.00Ru0.10Ni0.38Mn0.18Fe0.34O3 (329 mV, Tafel slope 79 mV dec-1, 93% voltage retention after 106 h at 10 mA cm-2; durability up to 30 mA cm-2) and sample La1.02Ru0.16Ni0.36Mn0.12Fe0.34O3 (269 mV, Tafel slope 62 mV dec-1, > 90% voltage retention after 85 h at 10 mA cm-2 and with outstanding stability with 92% voltage retention after 45 h at 30 mA cm-2). This work demonstrates an efficient strategy for generating phase-pure Perovskite catalyst libraries and shows that a coupled electrochemical screening and ML-guided synthesis process can accelerate the discovery of stable, sustainable OER catalysts containing less critical elements but with high performance.