Physical science of materials

Determination of the proportions of platinum atoms in agglomerates of bimetallic nanoparticles using machine learning methods

Ya. N. Gladchenko-Djevelekis, D. B. Tolchina, S. V. Belenov, V. V. Srabionyan, V. A. Durymanov, I. A. Viklenko, L. A. Avakyan, A. A. Alekseenko, L. A. Bugaev

Southern Federal University, Rostov-on-Don

Abstract: In this paper, we consider the applicability of machine learning methods, in particular, artificial neural networks, to obtain information on the distribution of target substance atoms in aggregates of bimetallic nanoparticles of various architectures. To solve the problem, we use data on paired radial distribution functions of atoms, the direct sources of which are experimental methods of X-ray diffraction and X-ray absorption spectroscopy from an extended energy region of the spectrum. The trained model of the artificial neural network demonstrates high accuracy in determining the proportions of platinum atoms in the composition of nanoparticles of various architectures in the agglomerate (determination coefficient $\sim$0.98). To verify the trained model, experimental data for catalysts containing bimetallic PtCu nanoparticles were used. Verification showed a high generalisability of the model, which indicates the promising application of this approach to the determination of platinum consumption efficiency in the synthesis of platinum-containing nanoparticle-based catalysts.

Keywords: core-shell nanoparticles, gradient nanoparticles, RDF, exafs, artificial neural networks, CatBoost.

Received: 24.10.2024
Revised: 11.03.2025
Accepted: 01.04.2025

DOI: 10.61011/JTF.2025.08.60901.365-24

Bibliographic databases:

•