Speaker
Description
Aiming at data-driven design of magnetic materials as a demonstration of using NOMAD to integrate automated workflows, metadata formulation, and machine learning, we elucidate how research data management can be implemented for first-principles calculations on magnetic materials. On the one hand, we have established workflows to perform high-throughput calculations on the intrinsic magnetic properties, including the magnetic ground state, saturation magnetization, Curie/Neel temperature, magneto-crystalline anisotropy, topological transport, and spectroscopic properties, for both crystalline and chemically disordered materials. On the other hand, extensive machine learning algorithms have been implemented to map out the structure-property relationships, and also to bridge to experimental simulations. We are going to show how the research data management can be performed for our data on NOMAD using pre-defined and custom schemas that it is Findable, Accessible, Interoperable, and Reusable (FAIR), with illustrative machine learning demos, facilitated by a local NOMAD Oasis at TU Darmstadt.