Accelerating electrocatalyst design for CO2 conversion through machine learning: Interpretable models and data-driven innovations

Electrocatalytic conversion of CO2 into valuable products is a promising approach toward mitigating climate change and energy crisis. However, the product diversity and multi-electron transfer pathways drive the development of numerous strategies for catalyst component and active site modifications, leading to a long journey toward rational electrocatalyst design. The integration of machine learning (ML) with experimental workload provides an opportunity to speed up the materials discovery by automatically exploiting trends and patterns from database. This review focuses on the interpretability of ML models in electrocatalyst design, and demonstrates a reliable workflow based on adequate catalytic data with refined descriptors, and satisfactory configuration of model with appropriate human intervention. Moreover, the combination of data-driven techs and cutting-edge methodologies for the discovery of catalyst is also discussed, which can serve as bridge between contemporary catalysis and quantum chemistry. This review may provoke more ML-based innovations toward the rationalization of electrocatalyst design and the discovery of novel materials for net-zero energy industries.