ICYS Annual Report 2023

12/50

ICYS Annual Report 2023The development of environmental benign technologies with zero CO2 emission and deep integration of information technologies into daily life are the main trends for Japan toward the Society 5.0. Magnetic materials play one of the major roles in this process, for example, high-performance permanent magnets and magnetocaloric materials are essential for green energy conversion, while hard disk drives are the most reasonable choice for storing big data. To realize such a societal transformation, the performance of magnetic materials should be improved closer to their physical limits that can be accelerated by advanced computer simulations and data science.The main bottleneck for heat-assisted magnetic recording (HAMR) to achieve a potential areal density of 4 Tb/in2 is the difficulty in obtaining L10-FePt nanogranular media with perfectly isolated grains of a small size D < 6 nm. To address this, a fully automated routine has been developed, that combines a convolutional neural network and machine vision to enable data mining from transmission electron microscopy images of FePt-C nanogranular media [1]. This allowed to generate a dataset and implement a machine learning optimization model that guides process parameters the desired nanostructure, i.e., small grain size with unimodal distribution and a large coercivity, which was successfully validated Research Digest 1. Outline of Research2. Research ActivitiesAnton BOLYACHKINReferences1) N. Kulesh et al., Acta Mater. 255, 119039 (2023).2) Xin Tang, et al., NPG Asia Materials. 15, 50 (2023).Fig. 1. The workflow for the FePt-C nanogranular media which starts with deep learning (DL) segmentation of top-view TEM images by U-NET 3+, followed by the extraction of microstructural features: distributions of grain size (D), roundness (R), and pitch distance (Dp). These data are then combined with sputtering conditions and magnetic properties of the FePt-C media, such as coercivity (Hc), to form a dataset and train regressors. Finally, an optimization routine is launched to optimize the media, i.e., reduce D while maintaining high Hc.to achieve experimentally (Fig. 1).Another magnetic material is Nd-Fe-B-based permanent magnets which are widely used for energy conversion applications. However, their usage at elevated temperatures is hindered by a relatively low coercivity (Hc) with respect to the anisotropy field (HA) of the Nd2Fe14B compound. In this work [2], it was found that the coercivity of a (Nd0.8Dy0.2)-Fe-B sintered magnet could reach 0.4HA, which was twice as high as its Dy-free counterpart. Detailed microstructural that of characterization, first principles calculations, and micromagnetic simulations showed that the large value of coercivity originated not only from the enhanced HA of the main phase (intrinsic factor) but also from the reduced magnetization of the thin intergranular phase (extrinsic factor). The latter was attributed to the dissolution of 4 at.% Dy in the intergranular phase that anti-ferromagnetically coupled with Fe. The reduction in the magnetization of the intergranular phase resulted in a change in the angular dependence of coercivity from the Kondorsky-type for the Dy-free magnet to the Stoner-Wohlfarth-type for the Dy-containing magnet, indicating that the typical pinning-controlled coercivity mechanism began to show nucleation features as the magnetization of the intergranular phase was reduced by Dy substitution. 10Data Driven Development of Magnetic Materials