Machine Learning Assisted Materials Classification to Boost Catalyst Design for Electrochemical Oxidation

Abstract

Machine learning (ML) is revolutionizing materials science with electrocatalysis emerging as a particularly promising area. While numerous noble and non-noble materials have been explored for chlorine evolution reactions (CER), identifying robust and readily accessible electrocatalysts for broader electrochemical oxidation remains a significant challenge. In this study, we leverage ML to address this gap and identify such materials. We examine the complex relationships between the Fermi level and conduction band position of cobalt-based oxides, alongside various uncommon descriptors such as formation energy, energy above the hull, density, number of magnetic sites, and total magnetization. The data underwent careful cleaning and feature engineering using different ML processes to ensure accuracy. Models were trained on 70% of the data and tested on the remaining 30%. Using a Random Forest classifier, we analyzed electrochemical data and identified Co₃O₄ as a cost-effective and scalable material for electrochemical oxidation. This machine-learning-driven approach revealed that Co₃O₄ is more susceptible to oxidation, leading to high reaction efficiency in synthetic applications such as arene chlorination and epoxide conversion.