Micromagnetic Modeling and Prediction of Magnetic Phenomena: Domain Wall Dynamics and Hysteresis Loops

Abstract

The micromagnetics is a computational methodology for the comprehension of sub-micro-sized magnetization configurations. The usefulness of the method is verified by plenty of research about spintronics and magnetization dynamics. Nevertheless, the limitation of micromagnetics is explicit. First, it simply calculates the numerical dynamics of the individual magnetic moments; hence, the interpretation of the dynamic behavior of magnetization configuration is a separate issue. The scale of micromagnetics is also a drawback. Since it focuses on the calculation of sub-micro scale systems, it cannot cover larger subjects, like bulk-sized magnets. In this thesis, we suggest complementary methods that can be integrated with micromagnetics to overcome its limitations. The first topic is about the resonance of the magnetic domain wall (DW). The resonance of DW accompanies the complex oscillation of its shape and anomalous velocity increment, which cannot be explained by known analytic models. We introduced the new analytic model to describe the dissipation of DW based on the classical Lagrange mechanics and the collective coordinate. The merge of micromagnetics and analytic models effectively depicts anomalous velocity increment and the manifestation of resonance. The second topic is about the hysteresis of large-scale magnets. For the implementation of simulation beyond the limitation of micromagnetics, machine learning (ML) is adopted to be integrated. We decomposed the simulation system into grain-by-grain local networks and used them as the source of ML. The model can reconstruct a macro-sized soft magnet based on the micro-sized micromagnetic effects. The ML fills up the gap between micromagnetics and macro-scale magnets. This research introduces the integration of different methodologies into a single phenomenon to overcome the limitations of individual tools. Such approaches will offer hints to multi-scaled computational frameworks to cover the research procedure from the physical and analytic interpretation of phenomena to the statistical behavior to maximize the effectiveness of numerical simulation.