Abstract:This paper discusses the theory and numerical method of two-scale analysis for the multiscale Landau-Lifshitz-Gilbert equation in composite ferromagnetic materials. The novelty of this work can be summarized in three aspects: Firstly, the more realistic and complex model is considered, including the effects of the exchange field, anisotropy field, stray field, and external magnetic field. The explicit convergence orders in the $H^1$ norm between the classical solution and the two-scale solution are obtained. Secondly, we propose a robust numerical framework, which is employed in several comprehensive experiments to validate the convergence results for the Periodic and Neumann problems. Thirdly, we design an improved implicit numerical scheme to reduce the required number of iterations and relaxes the constraints on the time step size, which can significantly improve computational efficiency. Specifically, the projection and the expansion methods are given to overcome the inherent non-consistency in the initial data between the multiscale problem and homogenized problem.
What problem does this paper attempt to address?
This paper attempts to solve the theoretical and numerical analysis problems of the multi - scale Landau - Lifshitz - Gilbert (LLG) equations in composite ferromagnetic materials. Specifically, the paper mainly focuses on the following aspects:
1. **Consideration of complex models**:
- The paper introduces more realistic and complex models, including the effects of exchange fields, anisotropic fields, stray fields, and external magnetic fields.
- The explicit convergence order of the classical solution and the two - scale solution in the \(H^1\) norm is clearly obtained.
2. **Robust numerical framework**:
- A robust numerical framework is proposed, and the convergence results under periodic boundary conditions and Neumann boundary conditions are verified through multiple comprehensive experiments.
3. **Improved implicit numerical scheme**:
- An improved implicit numerical scheme is designed to reduce the number of required iterations and relax the restrictions on the time step, thereby significantly improving computational efficiency.
- In particular, projection methods and extension methods are proposed to overcome the inherent inconsistency of the initial data between multi - scale problems and homogenization problems.
### Formula summary
- **Multi - scale LLG equations**:
\[
\frac{\partial m^\varepsilon}{\partial t}-\alpha m^\varepsilon\times\frac{\partial m^\varepsilon}{\partial t}=-\frac{1}{1 + \alpha^2}m^\varepsilon\times h^\varepsilon_{\text{eff}}(m^\varepsilon)
\]
where \(h^\varepsilon_{\text{eff}}(m^\varepsilon)\) is the effective field, including exchange fields, anisotropic fields, stray fields, and external magnetic fields.
- **Two - scale approximation**:
\[
m^\varepsilon(x)\approx m_0(x)+\varepsilon\chi\left(\frac{x}{\varepsilon}\right)\cdot\nabla m_0(x)
\]
- **Neumann correction**:
\[
m^\varepsilon(x)\approx m_0(x)+(\Phi^\varepsilon(x)-x)\cdot\nabla m_0(x)
\]
### Convergence analysis
- **Periodic boundary conditions**:
\[
\|m^\varepsilon(x, t)-m_0(x, t)\|_{L^2(\Omega)}\leq C\varepsilon^{5/6}\ln(\varepsilon^{-1}+ 1)
\]
\[
\|m^\varepsilon(x, t)-m_0(x, t)-\varepsilon\chi\left(\frac{x}{\varepsilon}\right)\nabla m_0(x, t)\|_{H^1(\Omega)}\leq C\varepsilon^{1/2}\ln(\varepsilon^{-1}+ 1)
\]
- **Neumann boundary conditions**:
\[
\|m^\varepsilon(x, t)-m_0(x, t)\|_{L^2(\Omega)}\leq C\varepsilon^{5/6}\ln(\varepsilon^{-1}+ 1)
\]
\[
\|m^\varepsilon(x, t)-m_0(x, t)-(\Phi^\varepsilon - x)\nabla m_0(x, t)\|_{H^1(\Omega)}\leq C