about

Our work on atomistic simulations aims to help us better understand the fundamental mechanisms that underlie the complex processes in metallic alloys. We have worked on aspects related to interfacial structure and kinetics, phase transformations, mechanical properties.

A key area of interest for us is to extend atomistic simulations to move beyond “atomistic” timescale – timescales dictated by the frequency of atomic vibration. In this area we have worked to develop and use a variety of coarse graining techniques to access the dynamics at long (relative to brute force atomistic simulation) timescales.

Our work has explored field theory based approaches such as the phase field crystal method and alloy diffusive molecular dynamics approach as well as our most recent work coupling graph convolution neural networks to markov state models. See references on the right for examples

GdyNet x Markov State Models
S. Soltani et al. “Exploring glassy dynamics with Markov state models from graph dynamical neural networks“

Alloy Diffusive Molecular Dynamics
E. Dontsova et al. “Solute-defect interactions in Al-Mg alloys from diffusive variational Gaussian calculations“
E. Dontsova et al. “Solute segregation kinetics and dislocation depinning in a binary alloy“

Phase Field Crystal Method
J. Berry et al. “Phase field crystal modeling as a unified atomistic approach to defect dynamics“
M. Greenwood et al. “Phase field crystal model of solute drag“

Example Presentations:

Presentation by collaborator Prof. Jörg Rottler (UBC Physics) at the AISSAI Workshop : Machine Learning Glasses in Paris, November 2022 – Presenting work on GdyNet/Markov State Modelling applied to glass dynamics