Hermite interpolation in the roots of unity (Q1907512): Difference between revisions

Let \(A_0\) denote the class of functions analytic \(|z|<1\) and continuous in \(|z|\leq 1\). If \(L_{n-1, \rho} (f,z)\) is the Lagrange interpolant to \(f\in A_0\) on the zeros of \(z^n- \rho^n\) \((0< \rho< 1)\) and if \(s_{n-1} (f, z)\) is the Taylor section of \(f\), then it is known that \(\lim_{n\to \infty} L_{n-1, \rho} (f, z)- s_{n-1} (f, z)=0\) for \(|z|< 1/\rho\). In 1989, K. G. Ivanov and E. B. Saff proved the following interesting relation between \(L_{n-1, \rho} (f,z)\) and \(s_{n-1} (f,z)\): Theorem A. Let \(\Lambda\) be any subset of \(\mathbb{N}\) and let \(m\in \mathbb{N}\). Then the following are equivalent: (a) There exists an \(f\in A_0\) such that the first \(m\) coefficients \(c(j, n)\), \(j= n-1, \dots, n-m\) of \(L_{n-1, \rho} (f, z)\) are zero for every \(n\in \Lambda\). (b) There exists distinct points \(\omega_j\), \(|\omega_j |>1\), \(j= 1, 2, \dots, m\) and \(g\in A_0\) such that \(L_{n-1} (g, \omega_j) \equiv 0\), \(j=1, \dots, m\) for every \(n\in \Lambda\). Here the above result is extended to Hermite interpolants \(H_{r,n} (f, z)\) for functions \(f\in A_{r-1}\). \(A_{r-1}\) denotes the class of functions \(f\) which together with its \(r-1\) derivatives are analytic in \(|z|<1\) and \(f^{(r)}\) is continuous in \(|z|\leq 1\). A special case of this for \(r=2\) was proved by the authors in 1992 [Ganita 43, No. 1-2, 171-180 (1992; Zbl 0832.41004)].