Machine Learning-Optimized Overpotential in Zr-Doped CeO2 Electrocatalysts for Enhanced OER Efficiency

Abstract

Despite a wide range of noble metal-free electrocatalysts having been developed for efficient oxygen evolution reactions (OERs) to date, there is a need to optimize (i.e., dopant concentration, material deposited, drying time, etc.) these electrocatalysts to get the best overpotential. Thus, herein, we have studied the impact of Zr doping concentration (1, 2, 3, 4, and 5%) in CeO₂ along with experimental condition optimization as a function of overpotential through machine learning (ML) to design a highly efficient OER electrocatalyst. Our results demonstrated that the ML-optimized 3% Zr-doped CeO₂ electrode showed maximum electrocatalytic activity with lower onset potential (1.39 V vs RHE), overpotential (380 mV at 10 mA cm^–2), and Tafel slope (85.7 mV dec^–1) compared to CeO₂ doped with various concentrations of Zr. Additionally, ML-optimized 3% Zr-doped CeO₂ has shown better stability with 84% current density retention after 48 h, thus suggesting the reliability of our designed system.