Positivity of the weights of interpolatory quadrature formulae with Bernstein-Szegő abscissae (Q1014757): Difference between revisions

Given an interpolatory quadrature formula \(\int_{-1}^1 f(t)w(t)dt\approx \sum_{k=1}^n \lambda_{n,k}f(x_{n,k})\), one of the main questions to bear in mind is that we can not assure that all the coefficients \(\lambda_{n,k}\) are positive. The paper under review considers interpolatory quadrature formulae relative to the weight function \(w(t)=1\), whose nodes coincide with the zeros of the \(n\)-th degree orthogonal polynomial relative to the Bernstein-Szegő weight functions \(w_\gamma^{(\pm 1/2)}(t)=(1-t^2)^{\pm 1/2}/K(\gamma)\), where \(K(\gamma)=1-4\gamma t^2/(1+\gamma^2)\). Semi-explicit formulae are obtained for the quadrature weights, and it is shown that the weights are all positive, except for \(w_{(n+1)/2}^{(-1/2)}\) when \(n=4m+1\), \(m=1,2,\dots\), in which case it is positive if and only if \(\gamma_n^{(-1/2)}<\gamma<1\), where \[ \gamma_n^{(-1/2)}=M((n-1)/2)/M((n-3)/2), \] and \(M(s)=1-2\sum_{k=1}^s (-1)^k/(4k^2-1)\)