SKM 2023 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

MA: Fachverband Magnetismus

MA 32: Magnetic Imaging Techniques I

MA 32.5: Vortrag

Mittwoch, 29. März 2023, 16:00–16:15, HSZ 401

Deep learning assisted reconstruction of the magnetization from the 2D antiferromagnetic van der Waals material CrSBr — •Riccardo Silvioli, Michele Bissolo, Kartikay Tehlan, Martin Schalk, Ferdinand Menzel, Nathan P. Wilson, Andreas V. Stier, and Jonathan J. Finley — Walter Schottky Institute and TUM School of Natural Sciences, Technische Universität München

We investigate the layered antiferromagnet (AFM) CrSBr, a material with three phase transitions. Order within the layers (T_intra∼160K), order between the layers (T_N∼135K) and a low T phase close to 40K that is speculated to originate from the ordering of Br vacancies. We use widefield nitrogen vacancy (NV) vector magnetometry to investigate the magnetic phases of this material. We image the 3D magnetic stray field in the plane of the NV centers, located 100nm from the surface of the diamond. Retrieving information on the magnetization (M) from an NV measurement requires reconstruction of M. For out-of-plane M, this is an analytically solvable problem, whereas for in-plane M this problem is ill-posed. We employ a deep learning (DL) approach based on a convolutional neural network (CNN), in order to solve the inverse problem, and determine M from the data. We apply additional constrains to the CNN to follow Maxwells equations by incorporating micro magnetic simulations in the computation of loss during the training phase. We discuss the advantages of this physics informed CNN training approach and compare it to conventional CNN methods as well as reconstruction efforts.