| Abstract: |
| The precise capture of interfacial magnetization dynamics, such as domain wall motion and skyrmion evolution at heterostructure interfaces, remains a formidable challenge in computational spintronics due to the high non-linearity and multiscale nature of the governing equations. This talk presents recent progress in developing a high-performance simulation framework for the Landau-Lifshitz-Gilbert equation. We focus on a synergistic approach that integrates high-order structure-preserving numerical schemes with GPU-accelerated heterogeneous computing. The numerical framework consists of a structure preserving temporal discretization and finite element spatial discretization. A treecode algorithm is designed and implemented in GPU to significantly accelerate the simulations. Numerical experiments demonstrate that our framework provides competitive performance and flexibility on handling complex geometry of the domain, for resolving the dynamics of the domain wall, as well as skyrmion structures. |
|