Machine Learning for Determining the Architecture of Ensembles of Bimetallic PtCu Nanoparticles Based on Atomic Radial Distribution Functions

Abstract

It is known that the catalytic properties of materials based on bimetallic PtCu nanoparticles depend on both the composition and the distribution of atoms in these particles. Therefore, the development of new materials with improved catalytic properties requires the application of an accurate and reliable experimental method for determining the architecture of nanoparticles (NPs) (random solid solution, Janus, core–shell or “gradient”). Our previous study demonstrated through machine-learning simulations that the architecture of single bimetallic nanoparticles can be determined using accurate theoretically calculated paired atomic radial distribution functions (RDFs), which can also be obtained from the most common sources of NP structural information, such as the X-ray absorption spectroscopy (XAS) and X-ray diffraction (XRD) techniques. This work is a logical continuation of the research mentioned above and is devoted to a theoretical study of the influence of errors in determining the RDFs, as well as the influence of the size and composition distributions of nanoparticles on the possibility of determining the architecture of nanoparticles from their RDFs.