Work

Rapid Atomic Structure Prediction of Multimetallic Nanoparticles with Physics-Based Machine Learning

Year: 2025

Type: article

Abstract: Metal nanoparticles (NPs) find tremendous application in various fields, including catalysis, biomedicine, and electronics, due to their unique physicochemical properties arising from their morphology (i.e., size and shape) and composition. The chemical ordering of NPs, consisting of more than one metal, is crucial for optimizing their application performance, including stability. Traditionally, Density Functional Theory (DFT) has been used to investigate NP stability, but it is computationally expensive, limited to small systems, and cannot be applied to multimetallic NPs, which have enormous materials space. To address this limitation, recent efforts coupled a physics-based model (Bond-Centric Model) with a developed genetic algorithm to optimize the chemical ordering of NPs, leading to minimum (most exothermic) cohesive energies. Central to this approach is the calculation of weighting factors that scale the monometallic bond strength to describe that of the bimetallic bond. Herein, we perform a critical analysis and set some rules on how to apply these methods for rapid and accurate chemical ordering prediction of multimetallic NPs. Specifically, we optimized the chemical ordering of 2869-atom cuboctahedron NPs across 15 different bimetallic combinations and at varying metal compositions. In comparison with both experimental and computational results, our findings indicate that the use of small metal dimers for the calculation of the weighting factors leads to accurate and computationally efficient chemical ordering and stability predictions for a wide range of NP compositions. We further extended our investigation to 6 trimetallic NPs with a tremendously large materials space, testing our model's capability to predict chemical ordering patterns in multimetallic systems and demonstrating its power as a rapid and accurate computational method. This methodology can facilitate the design of thermodynamically stable multimetallic NPs and predict the distribution of different metal atoms from the core to the surface, which is central to any nanotechnological application.

Source: ACS Omega

DOI: 10.1021/acsomega.5c04082

Authors: Bassel Alkhatib, Maya Salem, Klaertje Kiyora Hesselink, Giannis Mpourmpakis

Institution: University of Pittsburgh

Language: English

Topic: Machine Learning in Materials Science

Subfield: Materials Chemistry

Field: Materials Science

Domain: Physical Sciences

Open Access status: gold

Estimated APC paid: $1,935

Funders: Division of Chemical, Bioengineering, Environmental, and Transport Systems, Division of Engineering Education and Centers