Our research vision is to understand, and ultimately to control, materials structure on the atomic scale. We combine quantum mechanics with machine learning (ML) to study relationships of structure, bonding, and properties. Our work is theoretical and computational, but is done in close collaboration with experimental partners, and with practical applications in mind.
Machine-learning approaches to inorganic chemistry
Computer simulations based on the laws of quantum mechanics are a cornerstone of materials research – but they are severely limited by their computational cost. We develop and apply interatomic potential models that "learn" from quantum-mechanical data, enabling accurate simulations that are many orders of magnitude faster. We are especially interested in building optimised and efficient databases for ML potential fitting, and in ML tools for chemical discovery.
Structural chemistry of amorphous solids
Amorphous (non-crystalline) materials are a frontier of current materials research: their disordered structures are difficult to determine experimentally, and they also pose large challenges for computational modelling. We use machine-learning-driven computer simulations to explore amorphous structures on the atomic scale. In this part of the group's research, we are purely driven by curiosity – we aim to create the most realistic models of amorphous structures, and we aim to understand the subtle rules that guide their formation.
We study the connections between microscopic structure and macroscopic functionality in inorganic solids. We are interested in functional materials for a variety of applications: phase-change materials for digital memories, battery electrodes, and graphene oxide materials.