The numerical solution of a parabolic Volterra integro-differential equation with a memory term on a one-dimensional unbounded spatial domain is considered. A quasi-wavelet based numerical method is proposed to handle the spatial discretisation, the Crank-Nicolson scheme is used for the time discretisation, and second-order quadrature to approximate the integral term. Some numerical examples are presented to illustrate the efficiency and accuracy of this approach.

This article discusses key characteristics of a semi-adaptive finite difference method for solving singular degenerate reaction-diffusion equations. Numerical stability, monotonicity, and convergence are investigated. Numerical experiments illustrate the discussion. The study reconfirms and improves several of our earlier results.

We prove a posteriori error estimates of optimal order for linearSchrödinger-type equations in the L ∞(L 2)- and theL ∞(H 1)-norm. We discretize only in time by theCrank-Nicolson method. The direct use of the reconstructiontechnique, as it has been proposed by Akrivis et al. in [Math. Comput.75 (2006) 511–531], leads to a posteriori upper bounds thatare of optimal order in the L ∞(L 2)-norm, but ofsuboptimal order in the L ∞(H 1)-norm. The optimality inthe case of L ∞(H 1)-norm is recovered by using anauxiliary initial- and boundary-value problem.