Research Methods
Our group is primarily computational/theoretical, with a small but growing experimental presence in collaboration with Prof. Jay Whitacre, focused on integrating machine learning approaches with automated experimentation. We have a strong commitment to using and contributing to open-source software development efforts. We’re always learning more, but the current methods/tools most commonly used in the group are summarized here.
- Electronic structure calculations: Density Functional Theory using GPAW and the Atomic Simulation Environment (
ASE)
- Scientific Machine Learning in both Julia and Python
- Phase-field modeling: MOOSE
- Crystal structure search/prediction: CALYPSO
- Bayesian optimization with Dragonfly
- Nearly all of the above rely on high-performance computing resources on the shared College of Engineering Cluster mananged by our group, Arjuna, as
well as at national leadership computing facilities such as NERSC and XSEDE