Stability of Finite-Difference Models Containing Two Boundaries or Interfaces (Q3763516)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 4020387
Language Label Description Also known as
default for all languages
No label defined
    English
    Stability of Finite-Difference Models Containing Two Boundaries or Interfaces
    scientific article; zbMATH DE number 4020387

      Statements

      1985
      0 references
      wave packets
      0 references
      parasitic waves
      0 references
      finite-difference methods
      0 references
      hyperbolic system
      0 references
      wave solutions
      0 references
      rightgoing and leftgoing waves
      0 references
      reflection matrix
      0 references
      GKS stability
      0 references
      P-stability
      0 references
      practical stability
      0 references
      instability
      0 references
      0 references
      Stability of Finite-Difference Models Containing Two Boundaries or Interfaces (English)
      0 references
      This interesting paper is concerned with stability of finite-difference methods for a hyperbolic system (1) \(u_ t=Au_ x\), where u(x,t) is N- vector, \(t>0\), x is a real number, A is a constant \(N\times N\) matrix. The problem (1) is approximated using the following \(s+2\)-level finite- difference formula \((2)\quad Q_{-1}v^{n+1}=\sum^{S}_{\sigma =0}Q_{\sigma}v^{n-\sigma},\) where \(v^ n_ j=v(jh,nk)\), k is the time step, h is the space step and each \(Q_{\sigma}\) is a spatial finite-difference operator with constant matrix coefficient of dimension \(N\times N.\) The wave solutions to (2) of the form \((3)\quad v^ n_ j=V\kappa^ jz^ n,\) where \(V\in {\mathbb{R}}^ N\), \(\kappa =e^{i\xi h}\), \(z=e^{i\omega k}\) are investigated. The numbers \(\xi\in {\mathbb{R}}\), \(\omega\in {\mathbb{C}}\) are related by a dispersion relation \(\omega =\omega (\xi)\), and the group velocity is defined by \(C=-d\omega /d\xi\). A solution of (3) with \(| \kappa | =| z| =1\) and \(C\leq 0\) (resp. \(C\geq 0)\) or with \(| z| \geq 1\), \(| \kappa | >1\) (resp. \(| z| \geq 1\), \(| \kappa | <1)\) is called leftgoing (resp. rightgoing). The general solution to (2) is a linear combination of rightgoing and leftgoing waves with coefficients \(\alpha_ m\) and \(\beta_ m\), respectively. If the vectors \(\alpha\) and \(\beta\) are related by \(\alpha =B(z)\beta\), then B(z) is the so-called reflection matrix. The author considers various connections between GKS stability introduced by \textit{B. Gustafsson}, \textit{H.-O. Kreiss}, and \textit{A. Sundström} [ibid. 26, 649-686 (1972; Zbl 0293.65076)] and the so-called P-stability (practical stability). Unfortunately, both are too complicated to reproduce here. The numerical schemes considered contain two boundaries or two or more interfaces. It is a source of appearance of some additional ``parasitic'' waves which may cause instability. The author obtains several interesting results. Here only two of them are quoted: the first one says that for the stability of some multi-interface model it is not sufficient that the individual interfaces be stable. The second one establishes stability in certain two-interface problem if each interface individually is stable and the reflection matrices have norms less than one. The explicit examples of numerical schemes play an important role in the paper. Some of them are rather artificial but still well working.
      0 references

      Identifiers