Colloquium: Dr. Kayahan Saritas, ORNL
Wednesday, October 2, 2019 · 3:30 PM - 4:30 PM
TITLE: Quantum Monte Carlo from Benchmarking to Understanding Nanoscale Materials
Density Functional Theory (DFT) is currently the most popular method for atomic scale modeling and rational design of new materials. However, transition metal oxides are particularly problematic for DFT, due to the strong many-body interactions in the d-orbitals. Although DFT calculations can be adjusted by empirical parameters, such as Hubbard-U, the transferability of these parameters is questionable. Diffusion Monte Carlo (DMC) is a variational method that treats electrons explicitly, solving the many-body Schrödinger equation with systematically improvable approximations. Although DMC has the identical scaling O(N^3) with DFT, it comes with a large prefactor that is nearly ~10^3 times that of DFT. Although DMC calculations on large systems are challenging, systems with 500-1000 electrons can be routinely performed on an institution sized cluster using the QMCPACK code.
In this presentation, we will discuss ground and excited state properties of increasingly complex bulk materials from main-group compounds to transition metal oxides. The examples will include a high throughput DMC method we developed to calculate formation energies and build phase diagrams of closed shell materials. This can be considered mainly a benchmarking effort. We then discuss our DMC effort towards understanding the structure and magnetism in bulk and uniaxially strained LaCoO3, which has been a challenging problem for DFT. Finally, we will make a brief mention on the several developments in the QMCPACK code that enables publication quality calculations in quite modest systems.