The latest advancements in materials synthesis have allowed for the realization of a plethora of novel materials and their alloys. These nowadays contain the majority of the periodic table of elements, show extraordinary properties, and are gradually finding their ways into numerous applications including energy, sustainability, medicine, novel computation, etc.

A major attribute that determines the usability of many of these materials is their electronic properties. Their accurate assessment and prediction, however, is a highly difficult task. The reason is that the highly advantageous properties that these materials exhibit, are a consequence of the complexity of their electronic structure, with multiple anisotropic bands, tunable bandgap and effective masses, topologically protected bands, etc. Thus, computational methods to evaluate electronic properties are tremendously complex and computationally expensive.

The project uses Machine Learning (ML) to predict, accelerate, and scale the computation of electronic properties of complex materials and their alloys. In a typical calculation we use Boltzmann transport simulators, which take as input the electronic structure (from DFT) and the scattering rates that the electrons experience as they propagate in the material (extracted from ab initio or from effective Hamiltonians). The latter is the most computationally expensive part, which requires the evaluation of the interaction of electrons and phonons. The focus of this project is to accelerate this through ML, such that it can be scaled for a wide range of new materials and their alloys. The project will accelerate the Boltzmann transport simulations as well, closing the complete path of feasible and scalable computations from crystal structure all the way to electronic transport properties based on ML. The project partners with experimental collaborators that will test model predictions for high mobility materials.

Project Objectives and Outcomes

1. Development of a high-throughput computational framework that evaluates scattering rates for electronic transport in novel materials and their alloys.
2. The acceleration of the framework developed in item 1 using ML techniques.
3. Scalable studies for the electronic properties of materials and prediction of possible high mobility materials that hopefully are tested experimentally.
4. Development of ML methods for semiclassical transport that can bypass Boltzmann transport simulations.