1:45 PM - 2:30 PM
▲ [23p-B200-2] Investigation of magnetic properties using an explainable machine learning outputted energy landscape model
Keywords:Magnetic properties, Material informatics, Energy landscape
Coercivity is expressed as a complex correlation between magnetization and microstructure; moreover it is an important property for describing material functions. In real materials, metallography highly influences the magnetic properties owing to the various processes as defect pinning of domains walls. To improve the electrical devices, the coercivity mechanism must be clarified. However, owing to multiple intrinsic origins, coercivity cannot be easily described in the framework of conventional Ginzburg-Landau (GL) theory.
To elucidate the magnetization process, we use topological data analysis and other feature extraction techniques combined with dimension reduction algorithm, as principal component analysis (PCA), to draw a realistic energy landscape. The landscape map enables the visualization of the energy and magnetic loop as a function of feature space components, which allows us to identify the correlation between the morphological components and the physical properties. In our model, we aim to combine the microstructure and magnetization to describe the properties and functions of actual materials.
To elucidate the magnetization process, we use topological data analysis and other feature extraction techniques combined with dimension reduction algorithm, as principal component analysis (PCA), to draw a realistic energy landscape. The landscape map enables the visualization of the energy and magnetic loop as a function of feature space components, which allows us to identify the correlation between the morphological components and the physical properties. In our model, we aim to combine the microstructure and magnetization to describe the properties and functions of actual materials.