Alan F. Hegarty,Eugene O'Riordan
Abstract:A Petrov-Galerkin finite element method is constructed for a singularly perturbed elliptic problem in two space dimensions. The solution contains a regular boundary layer and two characteristic boundary layers. Exponential splines are used as test functions in one coordinate direction and are combined with bilinear trial functions defined on a Shishkin mesh. The resulting numerical method is shown to be a stable parameter-uniform numerical method that achieves a higher order of convergence compared to upwinding on the same mesh.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to construct a higher - order numerical method for singularly perturbed elliptic problems in two - dimensional space. Specifically, the paper focuses on solving the following form of singularly perturbed elliptic equation on the unit square region \(\Omega\):
\[ Lu := -\varepsilon \Delta u + a(x, y) u_x = f(x, y), \quad (x, y) \in \Omega; \quad 0 < \varepsilon \leq 1; \]
\[ u(x, y) = 0, \quad (x, y) \in \partial \Omega; \]
where \(a(x, y)>\alpha > 0\) and \(a(x, y)\) and \(f(x, y)\) have certain smoothness. The solution \(u\) contains a regular boundary layer and two characteristic boundary layers.
### Main problems
1. **High - order convergence**: Existing numerical methods usually can only achieve first - order convergence when dealing with such problems. The goal of the paper is to construct a numerical method that can achieve higher - order convergence, especially when the parameter \(\varepsilon\) is very small.
2. **Stability and parameter consistency**: The paper hopes to construct a numerical method that remains stable when the parameter \(\varepsilon\) changes, and this method has uniform convergence under different parameters.
### Solutions
The paper proposes a numerical method based on the Petrov - Galerkin finite element method, using exponential splines as test functions and combining bilinear trial functions defined on a Shishkin grid. This method can achieve the following goals:
- **Stability**: By choosing appropriate basis functions and grids, ensure that the numerical method maintains inverse monotonicity at all grid points.
- **High - order convergence**: On the Shishkin grid, this method can achieve a higher convergence order than existing methods.
### Mathematical background
- **Shishkin grid**: This grid uses a fine grid in the boundary layer region and a coarse grid in other regions to adapt to the influence of the boundary layer.
- **Petrov - Galerkin method**: By choosing different test functions and trial functions, the stability and accuracy of the numerical method can be improved.
### Main results
- **Global error bound**: The paper proves that the proposed numerical method has a convergence order of \(O(N^{- 2}(\ln N)^2)\) in global error, where \(N\) is the number of elements in the grid.
- **Numerical experiments**: Through numerical experiments of several test problems, the correctness of the theoretical analysis is verified, and the effectiveness of the method in practical applications is demonstrated.
In conclusion, this paper solves the challenges in high - order convergence and parameter consistency of singularly perturbed elliptic problems by constructing a new numerical method.