Fehlerabschätzungen für Gauß-Quadraturformeln. (Error estimates for Gauss quadrature formulae) (Q1064488)

Let \(Q_ n:=\int^{1}_{-1}w(x)f(x)dx\) for an even weight function w, \(f(z)=\sum^{\infty}_{K=0}\alpha^ f_ Kz^ K\) be analytic in \(K_ r:=\{z\in {\mathbb{C}}:| z| <r\}\), \(r>1\), and \(| f|_ r:=\sup_{K\geq n}\{| \alpha^ f_{2K}| r^{2K}\}<\infty.\) The error functional \(R_ n:=I-Q_ n\) is continuous with respect to \(| \cdot |_ r\) and the relation \(\| R_ n\| =\sum^{\infty}_{K=0}[R_ n(q_{2K})/r^{2K}],\quad q_{2K}(x):=x^{2K},\) holds. This paper contains estimates of \(\| R_ n\|\). The author first derives two new representations of \(\| R_ n\|\) and then estimates \(\| R_ n\| =r^ 2R_ n(\phi)\), with \(\phi (x):=1/(r^ 2-x^ 2)\), in various ways by using the best uniform approximation of \(\phi\) in \(P_{2n-1}\) and the expansion of \(\phi\) with respect to Chebyshev polynomials. For \(w(x)=(1-x^ 2)^{\alpha}\), \(\alpha =\pm 1/2\), \(\| R_ n\|\) is calculated. The asymptotic behaviour, for \(r\to 1^+\), of \(\| R_ n\|\) and of the derived error bounds is also discussed. Finally, he compares different error bounds and gives numerical examples.