The accuracy of all-atom and coarse-grained molecular dynamics (MD) simulations depends both on the accuracy of the parameters used in interatomic potentials. However, the determination of these parameters for any soft- and/or hard-materials systems is a challenging task. Here, we have developed a computational framework to build coarse-grained (CG) embedded atom method (EAM) potentials for the face-centered cubic (FCC) metals. Specifically, we have developed CG models of Palladium (Pd), Gold (Au), Nickel (Ni), Copper (Cu), Aluminum (Al), and Platinum (Pt). This framework consists of the following steps: (i) Initially we have used Particle Swarm Optimization (PSO) integrated CG MD approach to identify a 14 parameter space defined using EAM potential that can reproduce experimental and/or computational methods (e.g. atomic simulations and quantum methods) properties of respective metals. (ii) The parameter set obtained in (i) was then used to perform a sensitivity analysis using the Sobol sampling method. (iii) A Gaussian Process (GP) based machine learning model to be used as a surrogate model for MD simulations was developed on this data. (iv) The Bayesian parameter estimation and further uncertainty quantification (UQ) of the model were then carried out by integrating the GP model with PEUQSE, a Bayesian parameter estimation tool. The metal properties predictions of the Bayesian refined CG model were within reasonable confidence levels of experimental data. With this robust computational framework, hard- and soft-material interatomic potentials can be developed with excellent reliability and accuracy.

Computational framework to develop coarse-grained embedded atom method (EAM) potentials for metals and their Bayesian uncertainty quantification

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.