Stability analysis of numerical methods for linear neutral Volterra delay-integro-differential system

https://doi.org/10.1016/j.amc.2004.08.003Get rights and content

Abstract

This paper is concerned with the stability of numerical methods for linear neutral Volterra delay-integro-differential system. A sufficient condition such that the system is asymptotically stable is derived. Furthermore, it is proved that every linear θ-method with θ  (1/2, 1] and A-stable BDF method preserve the delay-independent stability of its exact solutions. Finally, the numerical experiments are given to demonstrate the conclusions.

Introduction

In this paper, we consider the delay-integro-differential equationsMy(t)=ft,y(t),y(t-τ),Gt,y(t-τ),y(t-τ),t-τtg(t,s,y(s))ds,where f:R+ × Rd × Rd × Rd  Rd is continuous, M  Rd×d may be singular.

This system can be found in a wide variety of scientific and engineering fields such as biology, physics, ecology and so on (cf. [6], [9]). Particularly, it is observed that the delay-integro-differential-algebraic system (when M is singular) plays an important role in modelling many different phenomena of circuit analysis and chemical process simulation, which have a comprehensive list in [15]. Recently, there is a growing interest in developing numerical methods for solving the system (1.1) (cf. [2], [3], [4]).

As for the linear delay-integro-differential equationsdudt=λu(t)+μu(t-τ)+κt-τtu(σ)dσ,Koto dealt with the stability of θ-methods and Runge–Kutta methods with a given quadrature formula (cf. [16], [17]).

Moreover, Zhang studied the nonlinear Volterra delay-integro equationsy(t)=g(t)+0tf(ξ,y(ξ),y(ξ-τ))dξ,the criteria on stability of BDF method is given (cf. [21]). In Ref. [22], [23], Zhang extended the research to Volterra delay-integro-differential equationsy(t)=ft,y(t),Gt,y(t-τ),t-τtg(t,s,y(s))ds,by investigating the BDF methods and Runge–Kutta methods.

Up to now, few papers discuss the stability analysis of neutral delay-integro-differential system (cf. [9], [10]).

For the system (1.1) without integro term, Ref. [24] considered the asymptotic stability of some numerical methods for the linear delay-differential-algebraic equationAx(t)+Bx(t)+Cx(t-τ)+Dx(t-τ)=0,where the matrix A is singular.

In this paper, we focus our attention on the stability of numerical methods for the linear neutral Volterra delay-integro-differential system. The stability analysis of exact solutions to the equation is considered. A sufficient condition is given for the neutral Volterra delay-integro-differential system. Especially, this condition is also necessary for the system with algebraic constraints. Furthermore, the linear θ-method and BDF method are applied to the system, respectively. It is proved that the numerical method preserve the delay-independent stability of system if the linear θ-method satisfies θ  (1/2,1] and the BDF method is A-stable. Finally, the numerical experiments are presented to illustrate the conclusions.

Section snippets

Asymptotic stability of neutral Volterra delay-integro-differential system

In this paper, we consider the linear neutral Volterra delay-integro-differential system:Au(t)+Bu(t)+Cu(t-τ)+Du(t-τ)+Gt-τtu(x)dx=0,t>0,u(t)=ϕ(t),-τt0,where A, B, C, D, G  Rd×d, τ > 0 and the matrix A may be singular. Particularly, this system is the delay-integro-differential-algebraic equation if the matrix A is singular.

The existence and uniqueness of solutions to Eq. (2.1) with singular A has been studied by [20], and the Ref. [14] concerns with the case for A is nonsingular.

For

Stability of the linear θ-method

For the general neutral integro-differential system with constant delayddt[Ay(t)+W(t,y(t),y(t-τ))]=F(t,y(t),y(t-τ),t-τtV(t,x,y(x))dx),the linear θ-method, presented in [19], gives out the recurrence relation:Ayn+1-Ayn+W(tn+1,yn+1,yh(tn+1-τ))-W(tn,yn,yh(tn-τ))=h(1-θ)F(tn,yn,yh(tn-τ),Vn)+hθF(tn+1,yn+1,yh(tn+1-τ),Vn+1),where θ  [0,1], yn, yh(t) and Vn can be interpreted as the approximations to y(tn), y(t) and tn-τtnV(tn,x,y(x))dx, respectively.

Let δ=m-τh[0,1) with m be the smallest integer

Stability of the BDF method

For the linear scalar autonomous test equationy(t)=λy(t),with Rλ<0, the BDF method (cf. Section 5.1 of Chapter V in [11]) gives out the recurrence relation:i=0sαiyn-i=hλyn,where yn can be interpreted as the approximation to y(tn).

Here we quote two important definitions (cf. Definition 8.2.2 in [5] or Definition 3.3 in [11]).

Definition 4.1

The A-stability region of a numerical method for ordinary differential Eq. (4.1) is the set SA of complex numbers α =  such that the numerical solution yn obtained with

Numerical experiments

For illuminating the applicability of Theorem 3.1, Theorem 4.1, we give the following example.

Consider the linear delay-integro-differential-algebraic systemAu(t)+Bu(t)+Cu(t-1)+Du(t-1)+Gt-1tu(x)dx=0,whereA=100000010,B=0.10000-0.400.30.5,C=-0.5000000-0.50,D=0.120.20000.100.1-0.1,G=0.1600000.1600.08-0.2,with the initial function value ϕ(t)=(sint,cost,sint)T for t  [−1,0).

Here the matrix coefficients satisfy the sufficient and necessary conditions in Theorem 2.1. Hence, the system (5.1) is

References (24)

  • L.V. Ahlfors

    Complex Analysis

    (2004)
  • H. Ardt et al.

    Runge–Kutta Formulae Applied to Volterra Functional Equations with a Fixed Delay

    (1988)
  • Cited by (37)

    • Convergence of one-leg methods for neutral delay integro-differential equations

      2017, Journal of Computational and Applied Mathematics
    • Convergence of Runge-Kutta methods for neutral delay integro-differential equations

      2016, Applied Mathematics and Computation
      Citation Excerpt :

      For convenience, we shall always use the symbol R(α1, α2, β, γ) to denote the problem class consisting of all the problems (1.1) satisfying the conditions (1.3) and (1.4). It is obvious that the problem class R(α1, α2, β, γ) is more general than those of delay differential equations (DDEs), delay integro-differential equations (DIDEs), neutral delay differential equations (NDDEs) and so on, which have been in-depth studied by many authors and a great deal of interesting results have been obtained, as reflected in the recent monograph by Bellen and Zennaro [14], Brunner [15], Kuang [16] and in many references such as Enright [2], Zennaro [3], Vermiglio [7], Huang [4,5], Zhang [8–10], Zhao [11], Gan [12], Tian [13], Koto [6] etc. In 2007, Yu et al. [24] investigated the stability of Runge–Kutta methods for (1.1) ∈ R(α1, α2, β, γ).

    • Nonlinear stability of one-leg methods for neutral Volterra delay-integro-differential equations

      2014, Mathematics and Computers in Simulation
      Citation Excerpt :

      Brunner and Vermiglio [6] considered the stability of continuous Runge–Kutta methods for NVDIDEs of the “Hale's form”. Several researchers investigated the stability of numerical method for nonlinear NVDIDEs where the kernel K(t, θ, y) does not depend on y′ and its linear version (see [31–33,37,13,38,29]). However, few studies have been done on the stability of numerical methods for nonlinear NVDIDEs (1.1) in which the kernel K(t, θ, y, y′) also depends on y′.

    • Asymptotic stability of exact and discrete solutions for neutral multidelay-integro-differential equations

      2011, Applied Mathematical Modelling
      Citation Excerpt :

      We have found some results in the literature that directly deal with systems as general as (1.1), (1.2). In 2005, Zhao, Xu and Liu [28] discussed the asymptotic stability of the exact and discrete solutions to linear version of (1.1), (1.2) with p = 1. Yu [18] (see also [29]) gave the sufficient conditions for the numerical stability of one-leg methods (OLMs), RKMs and general linear methods (GLMs) for nonlinear NMIDEs (1.1), (1.2) with p = 1, but not give the sufficient condition for the asymptotic stability of the exact solution to nonlinear NMIDEs (1.1), (1.2) with p = 1.

    View all citing articles on Scopus

    This paper is supported by the National Natural Science Foundation of China (10271036).

    View full text