Ultimate sensitivity of radial distribution functions to architecture of PtCu bimetallic nanoparticles

Abstract

Bimetallic nanoparticles containing platinum and another d-metal are highly perspective catalysts with stability and activity superior to a single-metal platinum materials. It is known that the improvement of catalytic properties depends both from the composition and from the structural arrangement of atoms in bimetallic nanoparticles. This leads to importance of the experimental determination of the nanoparticles architecture (random solid solution, Janus, core–shell or “gradient”) for the search of novel bimetallic systems. We considered the platinum–copper nanoparticles synthesized by simultaneous or multistage sequential depositions of metals. The insight of the architecture of bimetallic PtCu nanoparticles was obtained by the study of radial distribution functions (RDFs) of metal atoms. The RDFs were obtained both theoretically, using molecular dynamics simulations, and experimentally, from the analysis of the extended X-ray absorption fine structure (EXAFS) spectra at Pt L3- and Cu K- edges. Machine learning (ML) algorithms revealed the outstanding sensitivity of the theoretical RDFs to the architecture of the bimetallic nanoparticles: the correct architecture can be determined with 99 % confidence in terms of F1 score. The application of the variety of ML classification methods to the experimental RDFs showed the benefit K-Neighbors classification method.