Adaptive mesh refinement for the Landau-Lifshitz-Gilbert equation

Jan Bohn,Willy Dörfler,Michael Feischl,Stefan Karch

2023-03-14

Abstract:We propose a new adaptive algorithm for the approximation of the Landau-Lifshitz-Gilbert equation via a higher-order tangent plane scheme. We show that the adaptive approximation satisfies an energy inequality and demonstrate numerically, that the adaptive algorithm outperforms uniform approaches.

Numerical Analysis

What problem does this paper attempt to address?

### What problems does this paper attempt to solve?

This paper aims to solve the computational efficiency and accuracy problems in the numerical simulation of the Landau - Lifshitz - Gilbert (LLG) equation. Specifically, the paper proposes a new adaptive algorithm for approximately solving the LLG equation through a high - order tangent plane scheme. The paper mainly solves the following problems:

1. **Inefficiency of non - adaptive simulations**:
- Magnetic effects in magnetic materials have strong temporal and spatial locality (such as domain walls and fast switching processes), which makes non - adaptive simulations very inefficient.
- Classical methods are based on uniform grids and often cannot achieve meaningful accuracy when dealing with these complex phenomena, even on the largest computers.

2. **Selection of time step size**:
- Even solving the LLG equation with a uniform time step size is challenging. Although weak convergence was proven as early as 2008, it was only recently that the strong a priori convergence of high - order time step size schemes satisfying physical energy bounds was proven.
- The paper uses a high - order time step size scheme to estimate the time approximation error and constructs an adaptive time step size scheme.

3. **Spatial discretization and grid adaptation**:
- A spatial error estimation method combined with a gradient recovery estimator is proposed, thereby deriving a fully adaptive LLG integrator.
- Through the Adaptive Mesh Refinement (AMR) technique, the grid density can be automatically increased in areas requiring higher resolution, thereby improving computational efficiency and accuracy.

4. **Satisfaction of energy inequality**:
- The paper proves that the adaptive approximation satisfies the energy inequality, ensuring the physical rationality of the numerical solution.
- This property is not always guaranteed in other high - order methods.

5. **Verification by numerical experiments**:
- Through multiple numerical experiments, the effectiveness of the adaptive algorithm is demonstrated, showing that it can provide smaller errors under the same workload and is superior to the method with uniform step size.

### Formula summary

- **LLG equation**:
\[
\alpha \partial_t m + m\times\partial_t m = P(m)H_{\text{eff}}(m)\quad\text{on}\ (0,T)\times\Omega
\]
\[
\partial_n m = 0\quad\text{on}\ (0,T)\times\partial\Omega
\]
\[
m(0) = m_0\quad\text{in}\ \Omega
\]
where \(P(m)=I_d-\frac{m\otimes m}{|m|^2}\), and the effective field \(H_{\text{eff}}(m)=C_e\Delta m + H_{\text{ext}}\).

- **Energy inequality**:
\[
E(t)=C_e\|\nabla m(t)\|^2_{L^2(\Omega)}+\frac{\alpha}{2}\int_0^t\|\partial_t m(s)\|^2_{L^2(\Omega)}ds
\]
\[
\leq C_e\|\nabla m(0)\|^2_{L^2(\Omega)}+\frac{1}{2\alpha}\int_0^t\|H_{\text{ext}}(s)\|^2_{L^2(\Omega)}ds
\]

Through these improvements, the paper provides a new solution for efficiently and accurately simulating micromagnetic phenomena in magnetic materials.