Permanent magnets are key elements of modern society. Important application areas are energy conversion including eco-efficient transport, hydro- and wind power. A promising magnetic material is MnAl-C.  Although it contains no ferromagnetic elements such as iron, nickel or cobalt, the so-called tau-MnAl-C is ferromagnetic up to high temperatures and has all properties which are prerequisites for high performance permanent magnets. The tau-MnAl-C contains no critical elements and therefore the long term use of this material is environmentally sustainable, in stark contrast to that of rare earth magnets such as Nd-Fe-B. In addition, tau-MnAl-C has a low physical density, which is a significant advantage for transport and aerospace applications.
In this project, a novel approach combining state of the art characterisation techniques with cutting edge computer simulations will be used to obtain quantitative information concerning the effect of interfaces on magnetisation reversal in tau-MnAl-C.


07.07.2020: The research results on extracting local nucleation fields in permanent magnets by machine learning were published in the Nature Partner Journal Computational Materials ( In addition, the successful project work was honored with an invited lecture at the congress: International Workshop on Rare-Earth and Future Permanent Magnets and their Applications (REPM2020, in Baltimore.

Local nucleation fields: (a) ground truth, (b) prediction, (c) absolute deviation

An abstract has been submitted for the conference: Magnetism and Magnetic Materials Materials (MMM2020,, which will take place in virtual form from 2 to 6 November 2020, originally planned in Florida.

„Macromagnetic Simulations by Micromagnetic Superposition

Micromagnetic simulations are typically limited to a few micrometers due to the high demand on computing resources. But in applications and experiments the specimen size is usually orders of magnitude larger. While the micromagnetic simulations reproduce trends nicely, the absolute value of the results differ from the experiments [1]. Often the size limitation in the simulations is the reason for this deviation. Normally this is overcome either by artificial scaling [2] or by a reduced-order model [3]. In this work we introduce a type of reduced-order model to bridge the length scale from micromagnetism to experiments. In contrast to the work of Blank [4], we also consider the microstructural features of the magnet. We subdivide the computation of a large sample, e.g. from experimental measurements, into multiple independent feasible-sized subsets. For each subset, the nucleation field is either calculated by micromagnetic simulation or predicted by a trained machine learning model [5] (Fig. 1). The subsets along with their microstructure and nucleation fields are fed into a newly developed python code to spatially reassemble the entire sample and compute the overall hysteresis (Fig. 2). To take the microstructure into account, each subset is further discretized in cubic elements with their own anisotropic easy axis. In each calculation step, the magnetostatic field and the exchange field of the entire sample is calculated based on the magnetic moments of these elements. The two field contributions and an increasing external field are summed up to a total field. Taking this total field into account, we compute the reversible rotations with the Stoner-Wohlfarth model and use the prestored nucleation fields to irreversibly switch the respective subset.

We gratefully acknowledge the financial support of the Austrian ScienceFund (FWF), Project: I 3288-N36 and the German Research Foundation (DFG), Project: 326646134.

[1] X. Tang, et al., Acta Mater. 144, 884-895 (2018)
[2] S. Bance, et al., Acta Mater. 131, 48-56 (2017)
[3] L. Exl, et al., J. Phys. Mater. 2, 014001 (2019)
[4] R. Blank, J. Magn. Magn. Mater. 101, 1-3, 317-322 (1991)
[5] M. Gusenbauer, et al., accepted npj Comput. Mater., (2020)”