High-performance predictions of electron and phonon transport from first principles

Abstract

The electrical and thermal transport properties of materials are central to their use in everyday technology. Using modern computational methods, theoretical predictions of transport can be made accurately for many materials. These calculations rely on knowledge of the electron and phonon states of materials, and in particular, their interaction and scattering rates. While first-principles methods based on density functional theory can now provide such material-specific quasiparticle properties, using this information to calculate transport coefficients is computationally demanding. To address this challenge, we developed and released a new software package, Phoebe, which efficiently predicts electron and phonon transport by solving the Boltzmann transport equation (BTE) from a full scattering matrix formalism. Using this new framework, we can predict a range of properties including electrical, thermal, and thermoelectric effects using an extensive set of BTE solvers suited to different situations. Phoebe utilizes MPI-OpenMP hybrid parallelization as well as GPU acceleration to manage the cost of these calculations and take advantage of modern HPC architectures. At the same time, a strongly object-oriented implementation makes it readily extensible to new theoretical developments. Through this thesis, I will describe the motivation, background knowledge, computational theory and implementation structure behind the Phoebe code, as well as some results demonstrating successful phonon and electron transport predictions.