Abstract by Nathan Foulk

Personal Infomation

Presenter's Name

Nathan Foulk

Degree Level



Jeremy Jorgensen

Abstract Infomation


Physics and Astronomy

Faculty Advisor

Gus Hart


Discovering the Materials of Tomorrow — Faster


The future of materials discovery lies in machine learning. However, current datasets of materials calculations are very small compared to the massive amounts needed for robust machine learning. The only viable way to expand these datasets is to make current density functional theory (DFT) codes much faster. We look to speedup Brillouin zone integration, the most computationally expensive part of DFT, by interpolating the Hamiltonians for an arbitrary set of k-points in the Brillouin zone. By projecting the atomic wavefunctions onto a significantly smaller basis set of pseudo atomic orbitals (PAOs), the diagonalization of these interpolated Hamiltonians becomes much more computationally feasible. Therefore, we can afford to integrate the Brillouin zone over a much denser k-point mesh.  Speeding up Brillouin zone integration will be a critical step towards discovering the materials of the future.