Predicting Material Properties from Electronic Band Structures: Integrating Machine Learning with Computational Modeling Techniques

The project involves the development of a machine learning model which can predict targeted properties of materials from their band structure. Computational modeling methods, such as ab initio (quantum chemistry) calculations, Density Functional Theory and Molecular Dynamics, enable us to calculate properties of materials from their crystalline structure. The band structure is a representation of how the electrons are distributed in the material and how they behave. This is linked to many properties of the material, such that visualizing the band structure can elucidate the underlying electronic properties of materials. However, establishing such a correlation is far from trivial and necessitates extensive and intensive calculations. We envision that machine learning could bridge easily computed band structures with electronic properties of interest, namely total magnetization, piezoelectric and dielectric constants.

Faculty Supervisor:

Bettina Kemme

Student:

Partner:

Chemia Discovery Inc

Discipline:

Computer science

Sector:

Clean Technology; Green/Alternative Energy; Artificial Intelligence; Quantum Science

University:

McGill University

Program: