Machine-Learning-Assisted Discovery of High-Efficient Oxygen Evolution Electrocatalysts

Iridium oxide (IrO2) is the predominant electrocatalyst for the oxygen evolution reaction (OER), but its low efficiency and high cost limit its applications. In this work, we have developed a strategy by combination of high-throughput density functional theory (DFT) and machine learning (ML) techniques for material discovery on IrO2-based electrocatalysts with enhanced OER activity. A total of 36 kinds of metal dopants are considered to substitute for Ir to form binary and ternary metal oxides, and the most stable surface structures are selected from a total of 4648 structures for OER activity evaluation. Utilizing the neural network language model (NNLM), we associate the atomic environment with the formation energies of crystals and free energies of OER intermediates, and finally a series of potential candidates have been screened as the superior OER catalysts. Our strategy could efficiently explore promising electrocatalysts, especially for evaluating complex multi-metallic compounds.

Automated summary

In this work, we developed a strategy combining high-throughput density functional theory (DFT) and machine learning (ML) techniques to discover IrO2-based electrocatalysts with enhanced oxygen evolution reaction (OER) activity. We considered 36 metal dopants to substitute for Ir and evaluated the most stable surface structures from a total of 4648 structures for OER activity. By using a neural network language model (NNLM), we associated the atomic environment with the formation energies of crystals and free energies of OER intermediates, and screened a series of potential candidates as superior OER catalysts. Our strategy efficiently explores promising electrocatalysts, particularly for evaluating complex multi-metallic compounds.