Machine Learning Accelerated Discovery of Subnanoparticles for Electrocatalytic Hydrogen Evolution

Metal and alloy subnanoparticles (SNPs) have been anticipated to be a class of promising catalysts because of their fundamental difference from nanoparticles (NPs). In general, the interaction among the surface and bulk atoms of SNPs is significant due to the higher degree of alloying in SNPs than that in NPs counterparts. This study compared the SNPs and NPs concerning their electrocatalytic activities of hydrogen evolution reaction (HER) to understand the essential difference between alloy SNPs and NPs by using machine learning. Phase segregation does not occur on multimetallic subnanoparticles (SNPs), but easily occurs on nanoparticles (NPs), which led to the homogeneously alloying of SNPs and unpredictable phase segregation of NPs. Because of this distinct behavior between SNPs and NPs, the HER activity of SNPs should be easier to predict than NPs. Therefore, machine learning was applied to accelerate the discovery of SNPs.