(257d) Benchmark Simulation of Ag Nanoparticles on Supported Graphene Substrates

Nanoparticles have a wide range of applications (thermal-, electro-, and photocatalysis, sensors, etc.) but their in-situ characterization and computational modeling prove challenging. Current nanoparticle models typically use low-index periodic models with defect-free surface due to computational cost reasons. The recent development of machine learning force fields (MLFFs) provide computationally viable approximations of density functional theory calculations while maintaining similar levels of accuracy in energies and forces to model nanoparticles as is. To benchmark the validity of MLFFs for nanoparticle simulations, we build upon the recent work of Campbell et al⁽¹⁾ that gives an experimental understanding of the adhesion of Ag nanoparticles to Ni(111) supported graphene and their chemical potential as a function of particle size. In this work, we train MLFFs using Allegro via an iterative committee training scheme to identify poorly predicted structures and improve the training set. Hybrid Grand Canonical Monte Carlo and Molecular Dynamics simulations are performed to explore favorable geometries of nanoparticles at various temperatures to make comparison to Campbell’s results and also compare the resulting optimal nanoparticle shapes with nanoparticles formed via the Winterbottom construction. This work marks the first progress toward understanding how supported nanoparticles look like under experimentally relevant conditions as we move from the nanocluster (diameter below 1 nm) to the nanoparticle size regime (diameter above 1 nm).

(1) Rumptz, J. R.; Mao, Z.; Campbell, C. T. Size-Dependent Adsorption and Adhesion Energetics of Ag Nanoparticles on Graphene Films on Ni(111) by Calorimetry. ACS Catalysis 2022, 12 (5), 2888-2897.