On the weights of certain quadratures for the numerical evaluation of Cauchy principal value integrals and their derivatives (Q1074042)

We consider interpolatory quadrature rules of the type \[ \frac{d^ k}{d\lambda^ k}\int^{1}_{-1}(1- x)^{\alpha}(1+x)^{\beta}\frac{f(x)}{\quad x- \lambda}dx=\sum^{n}_{i=1}w_{ni}^{(k)}(\lambda)f(x_{ni})+R_ n^{(k)}(\quad f),\quad k=0,1,..., \] where the nodes \(\{x_{ni}\}\) are the zeros of the Jacobi polynomial \(P_ n^{(\alpha,\beta)}(x)\). In particular we prove that \(\sum^{n}_{i=1}| w_{ni}^{(k)}(\lambda)| =O(n^ k \log n)\) uniformly in \(-1+\epsilon \leq \lambda \leq 1-\epsilon\). Furthermore we show that when \(k=1\), \(f(x)\in C^ p[-1,1]\), \(p\geq 1\), and \(f^{(p)}(x)\in H_{\mu}[-1,1]\), \(0<\mu \geq 1\), we have \(R_ n^{(1)}(f)=O(n^{-p-\mu +1+\delta})\), \(\delta >0\) small as we like, uniformly in \(-1+\epsilon \leq \lambda \leq 1-\epsilon\).