TensorNetwork MLIPs

Molecular dynamics simulations predict how individual atoms in a protein or material (or many other systems) evolve over time, allowing the detailed in silico study of atomistic processes. The most important input into those simulations is the force field, which specifies the interaction between individual atoms. For many technologically important processes such as electro-catalysis, or biochemical reactions only quantum mechanical models of interaction provide sufficient precision, but too computationally expensive. This project will contribute to the development of machine learning interatomic potentials, which “learn” the interaction between atoms from quantum-mechanical models but with significantly reduced computational demands. This technology opens up new abilities for scientific and industrial research to understand and control the behaviour of materials and of biological systems at the atomistic level.

Faculty Supervisor:

Christoph Ortner

Student:

Partner:

University of Cambridge

Discipline:

Mathematics

Sector:

Education

University:

The University of British Columbia

Program: