Stability Analysis of Linearly Implicit One-Step Interpolation Methods for Stiff Retarded Differential Equations

Authors:
  • KSK. Strehmel
    K. Strehmel
  • RWR. Weiner
    R. Weiner
  • HCH. Claus
    H. Claus

Abstract
The aim of this paper is to analyze the stability properties of the modified problem and of linearly implicit Runge–Kutta interpolation methods for stiff retarded differential equations with respect to the model equation \[y'(t) = py + qy(t - \tau ),\quad t \geqq 0, \]\[ y(t) = \varphi (t)\quad t \leqq 0 \] where p and q are co mplex constants and $\tau > 0$ is a constant d. Necessary and sufficient conditions for the stability of the modified problem (i.e., the retarded term $y(t - \tau )$ of the model equation is replaced by an interpolation polynomial) and of linearly implicit one-step interpolation methods are presented. These conditions allow us to compute the boundary locus of the stability region. It is shown that any modified problem and any linearly implicit Runge–Kutta–Lagrange method with appropriate stability function are $P(\beta )$-stable and that linearly implicit Runge–Kutta–Lagrange methods possess better stability properties than linearly implicit Runge–Kutta–Hermite methods. Furthermore, some P- and $P_0 $-stable discretization methods are given. The numerical experiments show the advantage of $P(\beta )$-stable linearly implicit Runge–Kutta–Lagrange methods.

This publication has 21 references indexed in Scilit: