Atomistic Modelling of Barocaloric Plastic Crystals using Equivariant Graph Neural Network Interatomic Potentials
||Atomistic Modelling of Barocaloric Plastic Crystals using Equivariant Graph Neural Network Interatomic Potentials|
||Johan Klarbring <firstname.lastname@example.org>|
||2023-04-11 – 2023-11-01|
This project aims to probe the usefulness of the recently developed Allegro  equivariant graph Neural Network interatomic potential in modelling phase transformations in so-called plastic crystals. Plastic crystals are crystals that contain molecular units that are randomly oriented, but whose centers-of-mass retain the long-range ordering characteristic of crystalline solids. These systems often have phase transformations from an ordered low-temperature state to a high-temperature phase with disordered molecular units. Since these transformations are highly sensitive to external pressure, they can be used to produce efficient solid-state cooling devices through the barocaloric effect . Ideally, atomistic modelling of such systems would employ DFT-based ab-initio molecular dynamics (AIMD) simulations based on density functional theory (DFT). The time- and length scales that need to be simulated are, however, prohibitively expensive with this methodology. The development of accurate machine learning interatomic potentials (MLIPs) over the last ~decade show promise in resolving this issue.
In this project, we will use pre-existing DFT-based data-sets for a few plastic crystals and train Allegro  MLIP’s. For other classes of crystalline materials, Allegro shows great promise in modelling the required length and timescales . We will first sweep a space of hyper-parameter’s looking for suitable accuracy-efficiency tradeoffs. We will then perform large-scale MD simulations on GPU’s in LAMMPS using the existing allegro-LAMMPS interface , to directly probe if the relevant phase-transformations are correctly described in relation to known experimental results.
Allegro, and the allegro-LAMMPS interface has been developed and tested for NVIDIA V100 and A100 GPU’s and so should be suitable for use on Berzelius.
We apply for the default allocation.
 https://github.com/mir-group/allegro; https://www.nature.com/articles/s41467-023-36329-y