Accelerating the convergence of spectral deferred correction methods (Q2489683)

This paper is concerned with the numerical solution of stiff initial value problems for ordinary differential equations (ODEs). The proposed approach, called the spectral deferred correction, is a variant of the deferred correction method based on the integral equation formulation instead of the standard differential form, together with a Gaussian quadrature approximation for the integral. It is shown that for linear ODEs the iterations obtained with this technique is equivalent to construct a preconditioned Neumann series expansion for the solution of the standard collocation discretization of the ODE. This fact allows the authors to propose for linear problems an accelerating technique by using the generalized minimal residual (GMRES) Krylov subspace method. In addition, this approach is extended to nonlinear problems by using linearly implicit methods. The paper includes also some numerical experiments to show a favourable behaviour of these methods with a fixed step size for some linear and nonlinear problems.