Improving the accuracy of collocation solutions of Volterra integral equations of the first kind by local interpolation (Q2266578): Difference between revisions

We study the numerical solution of Volterra integral equations of the first kind by collocation methods using piecewise polynomials of degree p. We show that when superconvergence occurs, the error in the collocation solution at the node points alternates in sign, independent from the kernel or the exact solution. Evaluation at certain points of interpolating polynomials of degree \(p+2\) annihilates the leading term in the asymptotic expansion of the error, thus yielding convergence of order \(p+3\). The derivation of the asymptotic expansion of the error for the general equation is reduced to the special case of numerical differentiation by the use of an asymptotic continuity result for the projectors associated with the collocation method under compact perturbations. We present some numerical results on the improved order of convergence for \(p=2\) and \(p=3\).