Runge–Kutta Defect Control Using Hermite–Birkhoff Interpolation (Q3977641)

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 13035
Language Label Description Also known as
default for all languages
No label defined
    English
    Runge–Kutta Defect Control Using Hermite–Birkhoff Interpolation
    scientific article; zbMATH DE number 13035

      Statements

      Runge–Kutta Defect Control Using Hermite–Birkhoff Interpolation (English)
      0 references
      0 references
      25 June 1992
      0 references
      defect control
      0 references
      Runge-Kutta method
      0 references
      stepsize selection
      0 references
      Hermite-Birkhoff interpolation
      0 references
      algorithm
      0 references
      numerical tests
      0 references
      0 references
      0 references
      In a single step of a Runge-Kutta method applied to an initial value problem \(y'(x)=f(x,y(x))\), which takes the solution approximation from \(x_ n\) to \(x_{n+1}=x_ n+h\), the defect is defined as the value of \(P'(x)-f(x,P(x))\), where \(P\) is a suitable interpolant to the numerical solution computed within the step. Controlling the defect by controlling the value of \(h\) is a possible means of stepsize selection. What is required is an appropriate choice of \(P\). Amongst the requirements laid down for \(P\) are its global continuity (continuity at least \(C^ 1\) is suggested) as well as accuracy and efficiency. The current paper proposes the use of the Hermite-Birkhoff interpolation scheme as a means of achieving these goals. For an \(s\)-stage method of order \(p\), in addition to the usual stage values, the solution is approximated at values \(\gamma_ 3,\gamma_ 4,\dots,\gamma_ r\) of the way through the step. From the derivatives computed at \(\gamma_ 1=0\) (the initial point of the step) and \(\gamma_ 2=1\) (the end of the step), and at the \(r-2\) additional values, an interpolation scheme of the form \(H(x_ n+\tau h)=d_ 1(\tau)y(x_ n)+d_ 2(\tau)y(x_{n+1})+h\sum^ r_{i=1} e_ i(\tau)y'(x_ n+h\gamma_ i)\) is proposed. The paper discusses a number of practical details concerning the selection of \(\gamma_ 3,\gamma_ 4,\dots,\gamma_ r\). Even for such a low order as \(p=5\), where \(r=5\) is the appropriate choice, the relative merits of various options turn out to be exceedingly complicated. Using values of \(\gamma_ 3,\gamma_ 4,\dots,\gamma_ r\) which are in some sense optimal, and a standard embedded Runge-Kutta method [cf. \textit{J. R. Dormand} and \textit{P. J. Prince}, J. Comput. Appl. Math. 6, 19-26 (1980; Zbl 0448.65045)], a specific algorithm is developed. This compares favourably on numerical tests with a known method based on Hermite interpolation at the points \(x_ n\), \(x_{n+1/2}\) and \(x_{n+1}\).
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references
      0 references
      0 references
      0 references