Preliminary Exam Seminar: William Zummo
Event Date: | January 29, 2025 |
---|---|
Time: | 3pm |
Location: | DLR 221 or via WebEx |
Priority: | No |
School or Program: | Materials Engineering |
College Calendar: | Show |
"Machine Learning Interatomic Potentials for Large Scale Molecular Dynamics"
William Zummo, MSE PhD Candidate
Advisor: Professor Strachan
ABSTRACT
Molecular dynamics is an atomistic simulation technique that has allowed researchers to study material properties with atom-scale resolution. Accurate predictions for material properties are entirely dependent on the parameterization of the interatomic potential used. Interatomic potentials are functions that describe the energy and forces of atoms based on their positions. The forces are obtained by differentiation of the potential energy surface predicted by the interatomic potential. The need for an accurate and scalable potential has long been of interest to the material science community. Although we have obtained accurate results using techniques such as density functional theory, the computational cost becomes too great when scaling to larger systems. When using simple interatomic potential functions molecular dynamics can scale to millions of atoms. Bridging the gap between the accuracy of density functional theory and the scalability of molecular dynamics is the aim of machine learning interatomic potentials. These machine learning models aim to learn the mapping of a set of atoms to their density functional theory predicted energies while scaling linearly with the number of atoms for large-scale molecular dynamics simulations.
2025-01-29 15:00:00 2025-01-29 16:00:00 America/Indiana/Indianapolis Preliminary Exam Seminar: William Zummo DLR 221 or via WebEx