On the multistep time discretization of linear\newline initial-boundary value problems and their boundary integral equations

Ch. Lubich
DOI: https://doi.org/10.1007/s002110050033
IF: 2.5
1994-04-01
Numerische Mathematik
Abstract:Summary.Convergence estimates in terms of the data are shown formultistep methods applied to non-homogeneous linear initial-boundaryvalue problems. Similar error bounds are derivedfor anew class of time-discrete andfully discrete approximationschemes for boundary integral equations of suchproblems, e.g., for the single-layer potentialequation of the wave equation. In both cases,the results are obtained from convergence andstability estimates for operational quadratureapproximations of convolutions.These estimates, which are also proved here, depend on bounds of theLaplace transform of the (distributional)convolution kernel outside the stability region scaledby the time stepsize, and on the smoothness of thedata.
mathematics, applied
What problem does this paper attempt to address?