(544bb) Developing First-Principles Based Embedded Atom Method Potentials for Metal Clusters Using Bayesian Statistics

Fundamental understanding of the electronic and catalytic properties of metallic nanoparticles has been widely studied using density-functional-theory (DFT) calculations over last few decades. However, the high cost of computations has hindered employing first-principles DFT calculations for extremely large number of atoms (over a thousand) in real-world catalyst systems. Many highly efficient methods of calculating the energetic and structural properties of atomistic systems have been developed to overcome the size limitation of DFT method by employing semi-empirical interatomic potentials. One of the most commonly used semi-empirical potential formalism for metals is the embedded-atom method (EAM). There are large repositories providing EAM potentials for various bulk metallic systems. However, since the local environment of atoms in nanoparticles differs from the bulk, EAM potentials that are parameterized to bulk material properties generally do not reproduce accurate binding energies when applied to metallic nanoparticles.

The goal of our current study is to to parameterize new EAM models based on an extensive set of DFT calculations of metal clusters. To implement the parametrization, as a new approach, we apply Bayesian statistics as it update the distribution of model parameters (posterior) by combining existing knowledge on the distribution of those parameters (prior) with a measure of how good the simulated and DFT calculated data fit with each other. The improved potentials with higher accuracy for different nanoparticle systems obtained from this parametrization can be used to study large systems catalytic activity.