Convergence analysis of waveform relaxation methods for neutral differential-functional systems (Q953388): Difference between revisions

Ths discussion in this paper focuses on the convergence and superlinear convergence of a continuous-time waveform relaxation (WR) method applied to Volterra type systems of neutral functional-differential equations (NFDEs). In the introduction the authors provide a review of results from earlier work (on WR method and NFDEs) by \textit{Z. Bartoszewski} and \textit{M. Kwapisz} [Comput. Math. Appl. 48, No. 12, 1877--1892 (2004; Zbl 1066.34068)] and \textit{Z. Jackiewicz} and \textit{M. Kwapisz} [SIAM J. Numer. Anal. 33, No. 6, 2303--2317 (1996; Zbl 0889.34064)]. In this paper a WR iteration scheme is applied which differs from that in the cited earlier work. Convergence analysis of the new scheme, under a time-dependent Lipschitz condition, is discussed in Section 2 and the error is established. Following a further condition, uniform convergence of the new method in \([0,T]\) is proved, and an error estimate is given. The results are sharper than pre-existing ones, e.g. the condition \(\beta(t) \leq t\) has been replaced by \(\beta(t)<t\). The case when \(\beta(t) =t\) is discussed and, under stated conditions, the authors establish that the WR method converges uniformly in each subinterval \([t_i, t_{i+1}]\). Conditions for superlinear convergence are established in Section 3. The paper concludes in Section 4 with two test problems to verify the newly established results. The potential of the method to achieve faster convergence is demonstrated.