Machine learning-driven high-throughput screening of electrocatalysts and electrolytes for electrochemical surfaces and interfaces

Machine learning (ML) establishes a new paradigm for electrocatalyst and electrolyte research by coupling high-throughput screening (HTS) with a data-driven understanding of electrochemical surfaces and interfaces. This review presents an end-to-end ML-HTS pipeline that unifies catalyst and electrolyte/additive screening at electrochemical interfaces and integrates thermodynamic and kinetic modeling, data realism, and descriptor universality. We systematically classify the workflow components, including database construction and descriptor design. Specifically, the descriptors correlated with the activity, selectivity, and stability of materials are categorized as geometric, electronic, energetic, and integrated descriptors. On this basis, the typical cases of high-throughput screening and ML model training are enumerated for single-atom and dual-atom catalysts, high-entropy alloys, and electrolytes and additives. In the end, we discuss current challenges, including database quality, model transferability, and the lack of standardization, benchmarking, and reproducibility. Ultimately, this review highlights that coupling ML-driven HTS with surface- and interface-level understanding accelerates the rational design of electrocatalysts and electrolytes.