Guaranteed eigenvalue bounds for the Steklov eigenvalue problem

Chun’guang You, Hehu Xie, Xuefeng Liu

Abstract

To provide mathematically rigorous eigenvalue bounds for the Steklov eigenvalue problem, an enhanced version of the eigenvalue estimation algorithm developed by the third author is proposed, which removes the requirements of the positive definiteness of bilinear forms in the formulation of eigenvalue problems. In practical eigenvalue estimation, the Crouzeix--Raviart finite element method (FEM) along with quantitative error estimation is adopted. Numerical experiments for eigenvalue problems defined on a square domain and an L-shaped domain are provided to validate the precision of computed eigenvalue bounds.

Introduction

In computational physics and engineering, eigenvalue problems are central to understanding phenomena such as fluid vibrations and crack propagation. While standard numerical methods, such as the conforming Finite Element Method (FEM), typically provide upper bounds for these eigenvalues, obtaining a mathematically guaranteed lower bound remains a significant challenge. Without both bounds, the exact range containing the true physical eigenvalue remains uncertain.

This research addresses the Steklov eigenvalue problem, where the eigenvalue parameter appears in the boundary conditions. These eigenvalues are fundamental to determining constants in the trace theorem, a cornerstone of mathematical analysis. Existing techniques for lower bounds often rely on restrictive assumptions, such as specific domain shapes or the positive definiteness of bilinear forms. This paper proposes a generalized framework to overcome these limitations.

Methodology

The authors introduce an enhanced algorithm that generalizes the framework established by Liu (2015). The approach integrates several key components:

  1. Crouzeix–Raviart Finite Element Method (FEM): The method utilizes non-conforming elements that maintain continuity only at edge midpoints. This flexibility is essential for deriving lower bounds.
  2. Generalized Theoretical Framework: By removing the requirement for bilinear forms to be positive definite, the authors extend the applicability of the Liu framework to a broader class of partial differential equations (PDEs).
  3. Quantitative Error Estimation: A computable lower bound formula is derived: $\lambda_i \geq \frac{\lambda_{h,i}}{1 + C_h^2 \lambda_{h,i}}$. Here, $\lambda_{h,i}$ is the approximate eigenvalue and $C_h$ is a projection error constant determined by the mesh size $h$.
Figure 1: Illustration of a triangular element $K$ and edge $e$ within a unit square domain, used for finite element discretization.
  1. Verified Computing: To ensure mathematical rigor, the team employed interval arithmetic via the INTLAB toolbox. This approach accounts for all potential rounding errors during numerical processing, transforming simulation results into formal mathematical proofs.

Numerical Experiments and Results

The algorithm was validated using two benchmark geometric domains: a unit square and an L-shaped domain. The latter is particularly challenging due to re-entrant corners that create singularities in the solution.

  • Precision and Convergence: For the unit square with a mesh size of $h=1/64$, the first eigenvalue $\lambda_1$ was rigorously bounded between 0.239 and 0.241. While upper bounds (calculated via standard Lagrange FEM) showed a convergence rate of $O(h^2)$, the new lower bounds converged at $O(h)$, primarily due to error terms on the boundary.
  • Robustness: The method successfully provided tight, verified bounds for the first five eigenvalues of the L-shaped domain, demonstrating its effectiveness even in the presence of geometric singularities.

Conclusion

The proposed method provides a robust "safety net" for Steklov eigenvalue problems by delivering guaranteed lower and upper bounds. By linking the first Steklov eigenvalue to the trace theorem constant ($C = 1/\sqrt{\lambda_1}$), the research offers a practical tool for functional analysis. Future work will focus on improving the convergence rate of lower bounds through the Lehmann–Goerisch method and targeted mesh refinement near boundaries.

Summary updated

Abstract

To provide mathematically rigorous eigenvalue bounds for the Steklov eigenvalue problem, an enhanced version of the eigenvalue estimation algorithm developed by the third author is proposed, which removes the requirements of the positive definiteness of bilinear forms in the formulation of eigenvalue problems. In practical eigenvalue estimation, the Crouzeix--Raviart finite element method (FEM) along with quantitative error estimation is adopted. Numerical experiments for eigenvalue problems defined on a square domain and an L-shaped domain are provided to validate the precision of computed eigenvalue bounds.
BibTeX
@misc{you2018guaranteed,
  author = {Chun’guang You and Hehu Xie and Xuefeng Liu},
  title = {{Guaranteed eigenvalue bounds for the Steklov eigenvalue problem}},
  year = {2018},
  eprint = {1808.08148},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1808.08148},
}
Figure