DPG Phi
Verhandlungen
Verhandlungen
DPG

Regensburg 2022 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

MA: Fachverband Magnetismus

MA 11: Computational Magnetism 1

MA 11.2: Talk

Monday, September 5, 2022, 15:15–15:30, H48

AiiDA-UppASD: Automatic workflow engine for high-throughput mag-netic simulations and machine learning — •Qichen Xu1,2,3, Jonathan Chico4, Manuel Pereiro2, Danny Thonig5, Erik Sjöqvist2, Olle Eriksson2, Anders Bergman2, and Anna Delin1,31KTH Royal Institute of Technology, Stockholm,Sweden — 2Uppsala University, Uppsala, Sweden — 3Swedish e-Science Research Center, Stockholm,Sweden — 4Sandvik Coromant, Stockholm, Sweden — 5Örebro University,Örebro,Sweden

The ever-raising theoretical peak performance and accessibility of supercomputer resources bring automated simulations to a more important position for studies of magnetic materials properties. Recently, Huber et al. developed the AiiDA framework. In order to perform high-throughput atomic spin dynamics (ASD) simulations and take advantage of AiiDA platform and its build-in DFT plugins, we introduce the AiiDA-UppASD plugin which allows users to access to the majority of the functionalities of the UppASD code within the AiiDA framework via a Python package. In addition, several robust built-in workflows are also provided for managing ASD simulations, handling possible errors, and providing common modular workchains. With these workflows, complex tasks like high-throughput simulations of magnetodynamic properties as well as the determination of spin-wave excitation spectra and magnetic phase diagrams can be performed in a very efficient manner. Meanwhile, a machine learning (ML) prepared data generation workflow is also designed in order to offer ASD-related databases for the ML community to benchmark and develop methods.

100% | Mobile Layout | Deutsche Version | Contact/Imprint/Privacy
DPG-Physik > DPG-Verhandlungen > 2022 > Regensburg