Machine Learning-Driven Design of Cr-Doped CeO₂ Electrocatalysts for the Efficient Oxygen Evolution Reaction

Despite tremendous progress in developing noble-metal-free doped electrocatalysts for the oxygen evolution reaction (OER), optimizing dopant concentration remains challenging. Unfortunately, traditional trial-and-error methods often yield suboptimal performance with high overpotentials and poor stability, underscoring the need for systematic, data-driven optimization strategies. To meet this challenge, herein, we have systematically optimized doping concentrations (1, 3, 5, 7, and 10%) in CeO 2, alongside experimental conditions (KOH electrolyte concentration), using machine learning (ML) techniques to predict their influence on overpotentials. The integration of ML enables rational and efficient design of active and stable OER electrocatalysts, highlighting the novelty and significance of our approach. Our results demonstrated that ML-optimized Cr 0.05 -doped Ce 0.95 O 2 coated on nickel foam (NF) showed high electrocatalytic activity by exhibiting a lower overpotential (280 mV) and Tafel slope (86.9 mV/dec) compared to Cr 0.03 -doped Ce 0.97 O 2 /NF (380 mV, 126.6 mV/dec), CeO 2 /NF (460 mV, 182 mV/dec), and NF (540 mV, 223 mV/dec). Interestingly, ML-optimized Cr 0.05 -doped Ce 0.95 O 2 /NF showed better stability, retaining 14.61 mA cm –2 current density after 100 h, thus suggesting the reliability of our designed system.