No label defined (Q4268136)

The author answers the question posed in the paper by \textit{G. Howell} and \textit{A. K. Varma} [J. Approximation Theory 58, No. 1, 58-67 (1989; Zbl 0687.41015)] on the best error bound for the first derivative of a function in \(W^5_{\infty}[0,1]\) interpolated by a quartic spline. An exact statement of the result proven is as follows: Theorem. Let \(f \in W^5_{\infty}[0,1]\) and let \(s \in C^2[0,1]\) be a quartic spline such that \[ s(x_i)=f(x_i), \quad i=0,1, \ldots, N, \] at grid points \(x_i\), \(0=x_0<x_1 \ldots < x_N=1\); \[ \begin{gathered} s(z_i)=f(z_i) \quad z_i=\frac{x_{i-1}+x_i}{2}, \quad i=1, 2, \ldots, N; \\ s'(0)=f'(0), \quad s'(1)=f'(1). \end{gathered} \] Then \[ \|s'-f'\|_{C}\leq\frac{h^4}{6!}\|f^{V}\|_{\infty}, \] where \[ h=\max_{0\leq i \leq N-1} x_{i+1}-x_i. \] The estimate obtained cannot be improved.