Asymptotic Expansion with Boundary Layer Analysis for Strongly Anisotropic Elliptic Equations

Ling Lin, Xiang Zhou

Abstract

In this article, we derive the asymptotic expansion, up to an arbitrary order in theory, for the solution of a two-dimensional elliptic equation with strongly anisotropic diffusion coefficients along different directions, subject to the Neumann boundary condition and the Dirichlet boundary condition on specific parts of the domain boundary, respectively. The ill-posedness arising from the Neumann boundary condition in the strongly anisotropic diffusion limit is handled by the decomposition of the solution into a mean part and a fluctuation part. The boundary layer analysis due to the Dirichlet boundary condition is conducted for each order in the expansion for the fluctuation part. Our results suggest that the leading order is the combination of the mean part and the composite approximation of the fluctuation part for the general Dirichlet boundary condition.

Introduction

Strongly anisotropic diffusion is a hallmark of many complex physical systems, ranging from heat conduction in magnetized fusion plasmas and groundwater flow in porous media to advanced image processing techniques. In these systems, diffusion occurs significantly faster along a specific direction—such as a magnetic field line—than across it.

Mathematically, this behavior is modeled using a small parameter $\epsilon$, where the diffusion coefficient in the "fast" direction scales as $1/\epsilon^2$. As $\epsilon$ approaches zero, the governing partial differential equations (PDEs) become increasingly "stiff," leading to numerical instabilities and ill-posedness in standard finite difference or finite element schemes. While existing asymptotic preserving methods handle simple boundary conditions well, they often struggle with non-constant Dirichlet boundary conditions, which trigger sharp transitions known as boundary layers. This paper provides a rigorous mathematical framework to resolve these layers and achieve high-order accuracy in the strongly anisotropic limit.

Methodology

The authors employ a multi-stage analytical approach to construct a robust solution that remains valid even as anisotropy becomes infinite.

1. Solution Decomposition

To address the ill-posedness of the Neumann boundary conditions in the limit, the solution $u_\epsilon$ is decomposed into two distinct components:

  • The Mean Part: Representing the average value of the solution along the fast $x$-direction.
  • The Fluctuation Part: Representing the local variations that deviate from the mean.

2. Matched Asymptotic Expansion

Using Van Dyke’s method, the fluctuation part is solved through two separate expansions:

  • Outer Expansion: This captures the behavior in the "bulk" of the domain, where the influence of the top and bottom boundaries is less immediate.
  • Inner Expansion: By utilizing a stretched coordinate ($\xi = y/\epsilon$), the authors "zoom in" on the top and bottom boundaries. This allows for the precise modeling of boundary layers that are only $O(\epsilon)$ thick.

3. Fourier Analysis and Composite Approximation

The inner equations are solved using Fourier cosine series, which inherently satisfy the zero-flux Neumann conditions on the lateral boundaries. Finally, a composite approximation is formed by stitching the inner and outer solutions together and removing overlapping terms, resulting in a single, continuous formula valid across the entire domain.

Theoretical Insights: The Failure of Simple Averaging

A common intuition in anisotropic problems is that if diffusion is infinitely fast in the $x$-direction, the solution should effectively become a function of $y$ alone (the average). However, the authors demonstrate that this intuition fails near the boundaries when Dirichlet conditions vary along the fast direction.

Through a Feynman-Kac interpretation, the authors view the PDE as a random walk. A particle moves extremely fast horizontally but at a standard rate vertically. Boundary layers emerge because particles starting near the top or bottom edges are likely to hit the boundary before they have sufficient time to explore the horizontal range and reach a state of "equilibrium" or averaging. Consequently, the boundary values are "imprinted" on the solution before averaging can occur.

Results and Validation

The framework was validated using a synthetic benchmark on a unit square with a source term $f(x, y) = \sin(\pi(x^2 + y^2))$. The asymptotic approximations were compared against a high-resolution numerical "gold standard" ($512 \times 2048$ grid).

  • Accuracy: For an anisotropy level of $\epsilon^2 = 0.05$, the mean part alone is insufficient. However, the first-order approximation ($u^{[0]}$) significantly improves accuracy, and the second-order approximation ($u^{[2]}$) is virtually indistinguishable from the exact solution.
  • Convergence: Numerical tests confirmed that a $2n$-th order expansion yields an error of $O(\epsilon^{2(n+1)})$. On log-log scales, the 0th-order expansion followed a slope of 1 relative to $\epsilon^2$, while the 2nd-order expansion followed a slope of 2, confirming the theoretical $O(\epsilon^2)$ and $O(\epsilon^4)$ accuracy.
  • Efficiency: By reducing a complex 2D problem into a sequence of 1D problems, the method offers significant computational speedups and is highly amenable to parallelization.

Conclusion and Future Work

This work provides a rigorous foundation for handling boundary layers in strongly anisotropic elliptic equations. While highly effective for rectangular domains with straight diffusion lines, future research is required to extend this analysis to curvilinear coordinates—common in plasma physics where magnetic field lines are curved—and to explore the implications of periodic boundary conditions in the strongly anisotropic limit.

Summary updated

Abstract

In this article, we derive the asymptotic expansion, up to an arbitrary order in theory, for the solution of a two-dimensional elliptic equation with strongly anisotropic diffusion coefficients along different directions, subject to the Neumann boundary condition and the Dirichlet boundary condition on specific parts of the domain boundary, respectively. The ill-posedness arising from the Neumann boundary condition in the strongly anisotropic diffusion limit is handled by the decomposition of the solution into a mean part and a fluctuation part. The boundary layer analysis due to the Dirichlet boundary condition is conducted for each order in the expansion for the fluctuation part. Our results suggest that the leading order is the combination of the mean part and the composite approximation of the fluctuation part for the general Dirichlet boundary condition.
BibTeX
@inproceedings{lin2018asymptotic,
  author = {Ling Lin and Xiang Zhou},
  title = {{Asymptotic Expansion with Boundary Layer Analysis for Strongly Anisotropic Elliptic Equations}},
  booktitle = {Communications},
  year = {2018},
  doi = {10.4310/cms.2018.v16.n3.a3},
}

Loading related content...

Figure