Adaptive refinement for convection-diffusion problems based on a defect-correction technique and finite difference method (Q2365377)

This paper presents a finite difference method, based on a defect-correction technique for a singularly perturbed convection-diffusion problem. Discretization error estimates of higher order (order \(p\)), which hold uniformly in the singular perturbation parameter \(\varepsilon\) are obtained. The numerical difficulties due to the fact that the solution has a boundary (or interior) layer are circumvented by appropriately grading and patching the given mesh. The meshsizes vary between \(h\) and \(\varepsilon^{3/2}h\) when \(p=1\), where \(h\) is the meshsize, used in that part of the domain where the solution is smooth, and \(\varepsilon^{3/2}h\) is the final meshsize in the boundary (or interior) layer. The correction operator is a monotone operator and this enables the estimate of an optimal order error in maximum norm. The total number of meshpoints is \(O(\varepsilon^{-1/s})h^{-d}+h6{-d}\) for a \(d\)-dimensional problem, where \(s=p\) and \(s=2p\), respectively in a boundary and interior layer. Numerical results that verify the theoretical conclusions are described.