Optimizing multimetallic nanoparticle compositions with machine learning towards high entropy alloys: case study of PtRuPdRhAu for the CO oxidation reaction Vladislav A. Mints 1 , Jack Pedersen 2 , Alexander Bagger 2 , Jonathan Quinson 2,3 , Andy S. Anker 2 , Kirsten M. Ø. Jensen 2 , Jan Rossmeisl 2 and Matthias Arenz 1 1 Department of Chemistry, Biochemistry and Pharmaceutical Sciences, University of Bern, Switzerland, 2 Department of Chemistry, University of Copenhagen, Denmark, 3 Department of Biochemical and Chemical Engineering, University of Aarhus, Denmark The studies of multi metallic and high entropy alloys (HEAs) face a major challenge due to the vast experimental space available for such materials [1-5]. The large number of different compositions in an alloy severely increases with the number of elements considered. This composition space can lead to materials with as many different properties. To best explore this range of potential compositions, machine learning methods become very attractive, to guide the synthesis of multi metallic nanomaterials in a time and resource efficient manner and so to rapidly design an optimal catalyst form a given application [6, 7]. In this work [8], we have optimized the bifunctional PtRuPdRhAu hydrogen oxidation catalyst for the CO oxidation reaction. The nanoparticles were simply obtained as room temperature using alkaline methanol solutions to reduce the metal precursors direclty on a carbon support. The optimization was carried out with Bayesian optimization in which the precursor ratio in the synthesis was optimized with respect to the CO oxidation onset potential. Subsequently, the obtained library of different samples was evaluated with EDX, which was used to construct machine learning models that correlate the elemental relationship to the CO oxidation reaction. Finally, the obtained correlations were elucidated with density functional theory. The machine learning models of PtRuPdRhAu, showed that ruthenium is affecting the CO oxidation reaction the most, and that it is ideally alloyed with platinum and rhodium. Using DFT it was pointed that the observed trends are correlated with the formation of *OH on ruthenium at lower potentials, which then oxidizes the adsorbed CO. References 1. George, Raabe, Ritchie, Nat. Rev. Mater. 2019 , 4, 515-534 2. Faraday Discuss. 2018 , 208, 471-495 3. Wu, et al. J. Am. Chem. Soc. 2022 , 144, 3365–3369
4. Yao, et al., Science 2022 , 376, eabn3103 5. Batchelor, et al. Joule 2019 , 3, 834-845 6. Pedersen, et al. Angew. Chem. Int. Ed. 2021 ,60,24144-24152 7. Zhang, et al. Npj Comput. Mater. 2022 , 8, 89 8. Mints, et al., ChemRxiv 2021 : DOI: 10.26434/chemrxiv-2021-zpbqb-v3
P27
© The Author(s), 2022
Made with FlippingBook Learn more on our blog