SKM 2023 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
MA: Fachverband Magnetismus
MA 2: Skyrmions I
MA 2.3: Vortrag
Montag, 27. März 2023, 10:15–10:30, HSZ 02
Machine learning based skyrmion detection with Kerr microscopy data — Isaac Labrie-Boulay1, Thomas Winkler1, Daniel Franzen2, •Kilian Leutner1, Alena Romanova1, Hans Fangohr3,4, and Mathias Kläui1 — 1Johannes Gutenberg University, Mainz, Institute of Physics, Staudinger Weg 7, Germany — 2Johannes Gutenberg University, Mainz, Institute of Informatics, Staudinger Weg 9, Germany — 3Max-Planck Institute for the Structure and Dynamics of Matter, Luruper Chaussee 149, 22761 Hamburg, Germany — 4University of Southampton, SO17 1BJ, Southampton, United Kingdom
Magnetic skyrmions are topologically stabilized quasi-particles and are a potential enabler for unconventional computing devices [1]. A common method for detecting skyrmions is to use a Kerr microscope. Experimental data is affected by noise, low contrast, intensity gradients, or defects. Therefore, manual data treatment is necessary to evaluate the observations. To automatize Kerr microscopy data analysis, we have used a special type of convolutional neural network, called U-Net, to determine the shapes and positions of skyrmions [2]. Different methods were used to optimize the classification and to detect the skyrmions quickly with high reliability and to minimize manual work [3]. Our approach can also be extended to other magnetic structures, such as stripe domains or vortices.
[1] Klaus Raab et al., Nat. Commun. 13, 6982 (2022)
[2] Olaf Ronneberger et al., arXiv:1505.04597 [cs.CV] (2015)
[3] Isaac Labrie-Boulay et al. (in preparation)