A Machine Learning-Enhanced Framework for the Accelerated Development of Spinel Oxide Electrocatalysts

Abstract

The surging demand for sustainable energy has spurred intensive research into electrochemical conversion devices such as fuel cells, water splitting, and metal-air batteries. The performance of oxygen electrocatalysts significantly impacts overall electrochemical efficiency. Recently, spinel oxides (AB2O4) have emerged as promising candidates; however, the scarcity of prior studies underscores the need for a thorough and comprehensive exploration. This study presents a computational framework that integrates machine learning and density functional theory (DFT) calculations for the systematic screening of 1240 spinel oxides. The data scarcity is addressed while enhancing prediction accuracy. Selected candidates are identified to outperform the benchmarking perovskite oxide. Additionally, their potential as mixed ionic and electronic conductors with a 3D network of ion diffusion pathways is highlighted. To further enhance the understanding and prediction of stability, catalytic activity, and reaction mechanisms, a new undemanding descriptor is introduced: the covalency indicator. This study offers a design principle for the development of high-performance spinel oxide oxygen electrocatalysts. Interest in spinel oxide oxygen electrocatalysts has recently surged, leading to concentrated research efforts in the materials discovery. An efficient computational framework, assisted by machine learning, is proposed for systematically screening 1240 spinel oxides. This framework guides the materials discovery and the principles of material design for the development of high-performance spinel oxide oxygen electrocatalysts. image