A theorem of Grünwald-type for Hermite-Fejér interpolation of higher order (Q1340074)

For a function \(f\in C[- 1,1]\) and triangular matrix of nodes \(X: x_{ni}\in [- 1,1]\), \(i= 1,\dots, n\); \(n= 1,2,\dots\) it is considered \((0, 1,\dots, m- 1)\) a Hermite-Fejér interpolation polynomial of the form \[ H_{nm}(f, x)= \sum^ n_{i= 1} f(x_ i) A_ i(x), \] where \[ A_ i(x)= \sum^{m- 1}_{k= 0} b_{ik}(x- x_ i)^ k \ell^ m_ i(x),\quad i= 1,\dots, n, \] \(\ell_ i(x)\) is a well-known standard polynomial. It is proved that if \(X\) is a strongly normal matrix of nodes, then \[ \lim_{n\to \infty} \| H_{nm}(f)- f\|= 0 \] for all \(f\in C[-1,1]\), where \(\| \cdot\|\) is the Chebyshev norm. In case \(m=2\) this result was established more than 50 years ago by \textit{G. Grünwald} [Acta Math. 75, 219-245 (1943; Zbl 0028.05001)]. In the proof the author uses the idea of Grünwald's proof. However, it is shown, that the condition of strong normality of nodes matrix, used by the author, differs from the corresponding definition, used by Grünwald. In conclusion an example of a strongly normal matrix of nodes in case \(m=4\) is given.