Abstract

Development and Optimization of Empirical Potentials as Surrogate Models for Energy Landscapes

Richard Hennig
University of Florida

Empirical potential provide the opportunity for rapid exploration of energy landscapes at the cost of being an approximation. In this tutorial, I will provide an overview of the choices of functional forms for empirical potentials for various materials that are motivated by the chemistry of bonding and discuss optimization and validation strategies for empirical energy models. To stimulate discussion, I will present open questions such as how to select the optimal functional form or an energy model, how to choose a materials data sets for parameter optimization, how to identify tradeoffs between conflicting model predictions.