In this talk, we consider the numerical solution of a singularly perturbed reaction-diffusion problem:
Using a standard Galerkin method, it is natural to conduct the analysis with respect to the -weighted norm (energy norm), . This has been done in several studies, such as [2], where the method is implemented with bilinear elements on a specially constructed piecewise-uniform Shishkin mesh. However, the energy norm is weak for this problem, since and are both . Thus, as , . This is particularly problematic, since it is easy to show that, for sufficiently small , if the FEM solution, , is computed on a uniform mesh, then is (meaning there is no convergence, since the layers are not resolved), yet . Even with the layer resolving Shishkin mesh, the error estimate in the energy norm is -dependent: .
More recently, Lin and Stynes [1] have proposed the use of a ``balanced norm'', which arises in the analysis of a mixed finite-element method and requires conforming spaces. In this talk we extend these results in the framework of a non-symmetric least-squares formulation, which finds the solution in an conforming finite-element space. The resulting finite-element solutions yield -independent convergence in a balanced norm, resolving the boundary layers accurately and efficiently, yet are also amenable to solution using scalable algebraic multigrid methods.