An Algebraic Study of Gauss-Kronrod Quadrature Formulae for Jacobi Weight Functions (Q3800970)

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 4068153
Language Label Description Also known as
default for all languages
No label defined
    English
    An Algebraic Study of Gauss-Kronrod Quadrature Formulae for Jacobi Weight Functions
    scientific article; zbMATH DE number 4068153

      Statements

      0 references
      0 references
      1988
      0 references
      Gauss-Kronrod quadrature formula
      0 references
      Gauss-nodes
      0 references
      Kronrod nodes
      0 references
      Jacobi weights
      0 references
      Gegenbauer weights
      0 references
      An Algebraic Study of Gauss-Kronrod Quadrature Formulae for Jacobi Weight Functions (English)
      0 references
      The Gauss-Kronrod quadrature formula for integrating f with respect to a fixed n is of the form \[ (*)\int^{b}_{a}w(t)f(t)dt=\sum^{n+1}_{j=1}a_ jf(t_ j)+\sum^{n+1}_{k=1}a^*_ kf(t^*_ k)+R_ n(f), \] where the \(t_ j\) are the ordinary Gauss-nodes (with respect to the given function w), and the ``Kronrod nodes'' \(t^*_ j\) and the coefficients \(a_ j\), \(a^*_ j\) are determined in such a way that the order of the formula (*) is as high as possible which results in \(R_ n(f)=0\) for all \(f\in \prod_{3n+1}\). In this general setting it is known that the Kronrod nodes \(t^*_ j\) are uniquely defined as zeros of a certain polynomial \(p\in \prod_{n+1}\). However, the zeros are not necessarily in [a,b] nor real. Therefore the following problems are immediate. Which weight functions w guarantee the following properties: a) The Gauss and Kronrod nodes interlace, i.e. the Kronrod nodes are all real and between two neighboring Gauss nodes there is exactly one Kronrod node and between two neighboring Kronrod nodes there is exactly one Gauss node. b) The nodes interlace and are all in ]a,b[. c) The nodes interlace and the coefficients \(a_ j\) are all positive (it is already known that positivity of \(a^*_ j\) is equivalent to the interlacing property). d) All nodes are real (whether interlacing or not). The authors investigate these problems for the so-called Jacobi weights \(w(t)=(1-t)^{\alpha}(1+t)^{\beta},\) \(t\in [-1,1]\), \(\alpha,\beta >-1\) and for the special case \(\alpha =\beta =\lambda -1/2\), \(\lambda >-1/2\), which results in the so-called Gegenbauer weights. The authors first treat the Gegenbauer case. By computing the corresponding polynomials for \(n=1(1)4\) explicitly the the authors manage to determine those \(\lambda\) 's for which properties a), b), c), d) hold. From these results they deduce two conjectures for the general case (still Gegenbauer) which is supported by further numerical result for \(n=1(1)20(4)40\) given in an appendix. In the further part of their paper the authors investigate properties a), b) and c) (deleting d) except for \(n=1)\) for the Jacobi weights. For \(n=1\) the problem is solved completely by analytic means. For \(n=2(1)10\) the regions of validity of the three properties a), b), c) are sketched in corresponding \(\alpha\)-\(\beta\) plane for each n there are two sketches) by numerical means. For two special weights namely Jacobi weight with \(\beta =1/2\) and for the weight \(w=| t|^{\gamma}(1- t^ 2)^{\alpha}\) there are explicit results in terms of nodes and coefficients.
      0 references

      Identifiers

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