Uniform convergence of derivatives of extended Lagrange interpolation (Q1183034)

For a weight function \(w\) on \([-1,1]\) and \(w_ 1(x)=(1-x)w(x)\), \(w_ 2(x)=(1+x)w(x)\), consider the sequences of orthogonal polynomials \(\{p_ m(w)\}^ \infty_{m=0}\) and \(\{p_ m(w_ k)\}^ \infty_{m=0}\), corresponding to \(w\) and \(w_ k\), \(k=1,2\), respectively. In Theorems 2.2-2.5 of this paper the authors prove that the zeros of \(p_ m(w)\) interlace with the zeros of \(p_ m(w_ k)\) and the same is true for the zeros of \(p_{m+1}(w)\) and \(p_ m(w_ k)\), \(k=1,2\). In Theorems 2.6 and 2.7 one gives uniform estimates for the distance between two consecutive zeros of \(p_ m(w)p_ m(w_ k)\) or of \(p_{m+1}(w)p_ m(w_ k)\), \(k=1,2\). Because of the density of zeros of orthogonal polynomials, such estimates are very important in the numerical solving of singular integral equations by collocation methods. Section 3 and 4 are concerned with Lagrange and Lagrange-Hermite interpolation with the zeros of \(p_ m(w)p_ m(w_ k)\) and \(p_{m+1}(w)p_ m(w_ k)\), \(k=1,2\), as nodes, for \(w\) a generalized Jacobi weight, i.e. \(w(x)=\varphi(x)(1-x)^ \alpha(1+x)^ \beta\), \(\alpha,\beta>-1\), \(\varphi>0\) and \(\int_{-1}^ 1 \omega(\varphi;\delta)\delta^{-1}d\delta<\infty\), where \(\omega\) is the ordinary modulus of continuity. It is proved that, by a suitable choice of additional nodes (near the points \(\pm1\)) or of the exponents of the generalized Jacobi weight, the approximation of a function \(f\in C^ q[- 1,1]\) and of its derivatives by Lagrange or Lagrange-Hermite interpolation polynomials is optimal. Uniform evaluations of the error in terms of \(\omega(f^{(q)};1/m)\) are given.