Machine Learning Interatomic Potentials
Machine learning interatomic potentials carry first-principles accuracy into the long molecular-dynamics trajectories required for transport calculations. We train a DeepMD potential to reproduce the energetics and forces of point defects in zirconium carbide (ZrC), enabling long-timescale simulations of defect-controlled thermal conductivity through the Green–Kubo formalism at high temperatures.
- Defect-Controlled Thermal Conductivity of ZrC at High Temperatures from First Principles — In Preparation (2026).