The Lebesgue function and Lebesgue constant of Lagrange interpolation for Erdős weights (Q1270277): Difference between revisions

An Erdős weight is of the form \(W(x)=\exp(-Q(x))\) where \(Q(x)\) is even and grows faster than any polynomial at infinity. For a given weight \(W\) and a given set of nodes \(\{\xi_{1n},\ldots,\xi_{nn}\}\subset{\mathbb R}\), the Lebesgue function is \(\Lambda_n(x)=W(x)W_n(x)\) where \(W_n\) is the Lagrange interpolating polynomial of \(W^{-1}\) for the given nodes. The Lebesgue constant is \(\| \Lambda_n\| _\infty\). It measures how much the Lagrange interpolating polynomial for a continuous function \(f\) is away from the best polynomial approximation. Thus if estimates for an upper bound of the Lebesgue constant are known, then it allows to deduce theorems about uniform convergence of Lagrange interpolation. In this paper upper bounds for the Lebesgue constant are derived (behaving like \(n^{1/6}\) at \(\infty\)) in the case where the nodes are the zeros of orthogonal polynomials with respect to the weight \(W^2\) for some particular \(W\). When the set of nodes is extended with two special points, then much better estimates for the Lebesgue constant can be given (behaving like \(\log n\)). Also pointwise estimates (upper bounds) are given for the Lebesgue function but not all the details of the proof are included.