Progress on three-body tight binding for materials prediction

Parameterized tight binding models have the potential to fill an important niche in materials modeling when both explicit treatment of electronic degrees of freedom and computational efficiency are paramount. Examples where tight binding is potentially useful include surfaces, defects, nanostructures, and disordered materials, all of which require large unit cells and treating broken or modified bonding. However, difficulty in fitting and testing such models has limited their applicability in practice. In this talk, we present progress in the ThreeBodyTB.jl parameterized tight binding project, which has a goal of developing a predictive and transferable model for most of the periodic table. The project includes both two-body and three-body effects when constructing the tight binding Hamiltonian, as well as a self-consistent treatment of the Coulomb interaction. We present new results on optimizing the overlap matrix and the treatment of charge transfer, as well as testing on new classes of materials systems.