Seminar: Machine Learning for Force Field Parameterization – Application to 2D Materials

Title:  Machine learning for force field parameterization — Application to 2D materials

Speaker:  Horacio Espinosa (Northwestern University)

Abstract:

The parameterization of interatomic potentials for molecular dynamics (MD) simulations has long been a highly-specialized endeavor requiring strong domain expertise and in most cases deep chemical intuition. We propose a robust approach incorporating multi-objective genetic algorithms and machine-learning-inspired protocols. Using monolayer MoSe2 as a testbed, we demonstrate the effectiveness of the proposed approach in parametrizing interatomic potentials with different levels of complexities for structural and mechanical properties in both the equilibrium and non-equilibrium regimes. Applications to flexible electronics, heat transfer, surface stability, as well as force field transferability will be discussed.

Brief bio:

Horacio Espinosa is the James and Nancy Farley Professor of Manufacturing and Entrepreneurship, Mechanical Engineering, and the Director of the Theoretical and Applied Mechanics Program at Northwestern University. He made key contributions in the areas of deformation and failure of materials, design of micro- and nano-systems, and in-situ microscopy characterization of nanomaterials. Espinosa received numerous awards and is a member of the National Academy of Engineering (NAE), foreign member of Academia Europaea, the Russian Academy of Engineering, and Fellow of AAAS, ASME, SEM, and AAM.