Unraveling the Stability of Trimetallic Nanoparticles with Machine Learning

Metal nanoparticles have gained immense interest due to their wide application in various fields spanning from catalysis to nanoelectronics and drug delivery. With the vast configurational space of multimetallic nanoparticles, it is difficult to implement Density Functional Theory calculations to find the most thermodynamically stable chemical ordering (arrangement of different metals) for a specific nanoparticle. Here, we apply the Bond Centric Model (BCM)^1,2 coupled with an in-house developed Genetic Algorithm (GA)², which can accurately and efficiently capture the stability of nanoparticles of any size, shape and metal composition. We use a proof of concept on how the BCM + GA captures trimetallic nanoparticle stability of PdPtAu³, to extend its usage to four more trimetallic nanoparticles of AgPdPt, AgAuPd, AgAuPt and AgCuPd made up of 2869 atoms, of cuboctahedral symmetry, with varying metal compositions. Our study reveals the total chemical ordering and, importantly, the surface composition of the nanoparticles resulting in the lowest cohesive energy (most thermodynamically preferred). We observe that for AuPdPt, Au tends to segregate to the surface, whereas Pd and Pt tend to reside in the bulk and subsurface. These observations for AuPdPt agree with other computational studies. We observed similar trends for AgPdPt; at high concentrations of Ag, there was more Ag on the surface, whereas Pd and Pt were in the bulk and subsurface, in agreement again with other computational studies. AgAuPt and AgAuPd showed trends that Ag and Au segregate to the surface while Pt and Pd occupy the bulk and subsurface layers. Our AgAuPt findings had similar surface composition trends compared to other computational studies and the AgAuPd findings agreed with experiments. We further extended the study to AgCuPd and we found that Ag resided on the surface, whereas the other two metals were in the bulk and subsurface, in agreement with experiments. Our work demonstrates the feasibility of using the BCM and the developed GA to reveal the exact chemical ordering of highly thermodynamically stable trimetallic nanoparticles and reveals the importance of metal cohesion and bimetallic bond strength in the overall chemical ordering trends.

References:

1. Z. Yan, M. G. Taylor, A. Mascareno, G. Mpourmpakis., Nano Lett. 2018, 18, 4, 2696-2704.
2. J. Dean, M. J. Cowan, J. Estes, M. Ramadan, G. Mpourmpakis, ACS Nano. 2020, 14, 8171-8180.
3. D. J. Loevlie, B. Ferreira, and G. Mpourmpakis, Acc. Chem. Res. 2023, 56, 3, 248-257.