G. Li,J. Li,J. Merten,Y. Xu,S. Zhu
Abstract:In this paper, we propose an adaptive finite element method for computing the first eigenpair of the p-Laplacian problem. We prove that starting from a fine initial mesh our proposed adaptive algorithm produces a sequence of discrete first eigenvalues that converges to the first eigenvalue of the continuous problem and the distance between discrete eigenfunctions and the normalized eigenfunction set with respect to the first eigenvalue in $W^{1,p}$-norm also tends to zero. Extensive numerical examples are provided to show the effectiveness and efficiency.
What problem does this paper attempt to address?
The core problem that this paper attempts to solve is to propose an adaptive finite element method (AFEM) for calculating the first eigenpair (i.e., the first eigenvalue and the corresponding eigenfunction) related to the p - Laplacian operator. Specifically, the author focuses on the p - Laplacian eigenvalue problem with homogeneous Dirichlet boundary conditions:
\[
\begin{cases}
-\nabla \cdot (|\nabla u|^{p - 2}\nabla u)=\lambda|u|^{p - 2}u & \text{in }\Omega,\\
u = 0 & \text{on }\partial\Omega,
\end{cases}
\]
where $\Omega$ is a bounded Lipschitz polygonal or polyhedral region, and $1 < p<\infty$.
### Main problems
1. **Accuracy of numerical calculation**: Due to the degenerate structure of the p - Laplacian operator and the possible existence of interior corner points in the computational domain, the solution has local singularities. Therefore, the traditional uniform meshing method is difficult to solve this problem efficiently and accurately.
2. **Effectiveness of the adaptive algorithm**: In order to improve the computational efficiency and accuracy, the author proposes an adaptive finite element method (AFEM) in a standard form, which approximates the first eigenpair by gradually refining the mesh. The specific steps include:
- **SOLVE**: Solve the discretized eigenvalue problem on the current mesh.
- **ESTIMATE**: Calculate the error estimator.
- **MARK**: Mark the elements that need further subdivision.
- **REFINE**: Refine the mesh according to the marking results.
3. **Convergence proof**: The author proves that starting from the initial fine mesh, the sequence of discrete eigenvalues generated by the proposed adaptive algorithm converges to the first eigenvalue of the continuous problem, and the discrete eigenfunctions also converge to the set of normalized eigenfunctions in the $W^{1,p}$ norm.
### Key contributions
- **Theoretical analysis**: The author not only proposes an adaptive algorithm but also proves its convergence through strict mathematical analysis.
- **Numerical experiments**: The effectiveness and efficiency of the algorithm are verified by several numerical examples, especially in computational domains of different shapes and with different values of $p$.
### Formula summary
- Weak - form equation:
\[
\int_{\Omega}|\nabla u|^{p - 2}\nabla u\cdot\nabla v\,dx=\lambda\int_{\Omega}|u|^{p - 2}uv\,dx\quad\forall v\in V.
\]
- Rayleigh quotient minimization problem:
\[
\lambda_1=\inf_{v\in V\setminus\{0\}}J(v):=\frac{\int_{\Omega}|\nabla v|^p\,dx}{\int_{\Omega}|v|^p\,dx}.
\]
- Discretized problem:
\[
\mu_T := J(u_T)=\inf_{v\in V_T\setminus\{0\}}J(v).
\]
- Local error indicator:
\[
\eta_T^q(\mu_T, u_T; T):=h_T^q\|R_T(\mu_T, u_T)\|_{L^q(T)}^q+\sum_{F\in\partial T\cap\Omega}h_F\|J_F(u_T)\|_{L^q(F)}^q,
\]
where $q = \frac{p}{p - 1}$.
Through these formulas and methods, the author successfully solves the numerical calculation problem of the p - Laplacian eigenvalue problem and provides theoretical guarantees and verification of the effectiveness in practical applications.