Abstract
This paper is concerned with approximations and related discretization error estimates for the normal derivatives of solutions of linear elliptic partial differential equations. In order to illustrate the ideas, we consider the Poisson equation with homogeneous Dirichlet boundary conditions and use standard linear finite elements for its discretization. The underlying domain is assumed to be polygonal but not necessarily convex. Approximations of the normal derivatives are introduced in a standard way as well as in a variational sense. On general quasi-uniform meshes, one can show that these approximate normal derivatives possess a convergence rate close to one in $L^2$ as long as the singularities due to the corners are mild enough. Using boundary concentrated meshes, we show that the order of convergence can even be doubled in terms of the mesh parameter while increasing the complexity of the discrete problems only by a small factor. As an application, we use these results for the numerical analysis of Dirichlet boundary control problems, where the control variable corresponds to the normal derivative of some adjoint variable.
Introduction
In many engineering and physics applications, the primary objective is not just the solution to a partial differential equation (PDE), but its normal derivative at the boundary. This quantity represents physical fluxes, such as heat transfer through a wall or fluid pressure against a surface. When using the Finite Element Method (FEM) on standard, quasi-uniform grids, calculating these derivatives is notoriously difficult if the domain has corners. Corners create "singularities" where the solution's gradient behaves poorly, causing the error in the normal derivative to vanish at a rate of only $O(h)$ on standard grids, or even slower for non-convex domains. This paper proposes a strategy to restore high-order accuracy without needing an impossibly fine mesh everywhere.
Methodology: Boundary Concentrated Meshes
The authors focus on the Poisson equation on polygonal domains, comparing classical direct derivatives ($\partial_n u_h$) with more stable variational approximations ($\partial_n^h u_h$). The core innovation is the use of Boundary Concentrated Meshes. Instead of a uniform grid, the mesh is isotropically refined toward the entire boundary: elements touching the boundary have a diameter of $O(h^2)$, while interior elements remain $O(h)$.
To prove the effectiveness of this strategy, the authors employ weighted $L^2$ error estimates and a dyadic decomposition of the domain. As illustrated in the figure above, the domain $\Omega$ is partitioned into a corner region $\Omega_c$ and boundary-adjacent regions $\Omega_{\Gamma_i}$, allowing for precise management of weights near geometric singularities.
Analysis of Convergence Rates
The primary advantage of boundary concentration is a "free" accuracy boost. While the total number of elements increases only by a logarithmic factor $|\ln h|$, the convergence rate $s$ can effectively double.
As shown in the figure above, for domains with interior angles $\omega < 120^\circ$ ($2\pi/3$), the $L^2$ error rate for the normal derivative reaches order 2 on concentrated meshes (blue line), compared to order 1 on quasi-uniform meshes (red line). Even in non-convex domains (e.g., L-shaped rooms where $\omega > \pi$), the convergence rate is significantly improved. For instance, where a standard mesh might yield a rate of 0.33, the concentrated mesh provides 0.66. The figure also highlights "superconvergence" effects (black line) which offer intermediate improvements for specific angular ranges.
Applications and Conclusion
These results have immediate implications for Optimal Control. In Dirichlet boundary control problems, the error in the control variable is directly tied to the error of the normal derivative of the "adjoint state." By employing boundary concentrated meshes, engineers can achieve high-order accuracy in control solutions without the computational burden of global refinement. While currently limited to polygonal domains and linear elements, this work provides a robust theoretical foundation for high-precision flux calculations in complex geometries.
Summary updated