The advent of exascale computers enables us to accelerate insight into the nature of materials and the discovery of new ones with desired properties. To take advantage of the exascale era, comes with several challenges. For the computational materials-science community this means dramatic speed-ups of “legacy codes”, such to optimally scale up to hundreds of thousands of cores; high-throughput calculations of large fractions of the (in principle infinite) materials space; and handling extreme-scale data with novel data-analytics and AI tools. These components cannot stand by themselves but need to be orchestrated by workflows, forming the “glue” in between. This concerns input generation, convergence of results with respect to computational parameters, job and error handling, as much as suggesting materials to be considered in the next step. In this presentation, I will address all these aspects with the examples of spectroscopy and solar-cell data, and show how the challenges can be tackled by bringing together developments from different research projects [1,2,3].
[1] NOMAD Center of Excellence (https://nomad-coe.eu) – EC HORIZON 2020 funded project for exascale computing in ab initio computational materials science
[2] NOMAD data infrastructure (https://nomad-lab.eu), developed within the FAIRmat consortium (https://www.fairmat-nfdi.eu/fairmat)
[3] FONDA – Foundations of Workflows for Large-Scale Scientific Data Analysis, Collaborative Research Center at Humboldt-Universität zu Berlin (https://fonda.hu-berlin.de/)
Copyright. All Rights Reserved.