No label defined (Q3972192)

[For the entire collection see Zbl 0695.00019.] For a positive integer \(m\) let \(x_ k=k\pi/m\), \(0\leq k<2m\), be equidistant nodes in \([0,2\pi)\) and for \(\varepsilon=0\) or \(\varepsilon=1\) define \(T_{m,\varepsilon}\) as the space of all trigonometric polynomials of the form \[ t_ m(x)=a_ 0+\sum_{k=1}^{m-1} (a_ k\cos kx+b_ k \sin kx)+a_ m \cos\left(mx+{\varepsilon\pi \over 2}\right). \] In the paper the following \((0;M)\) interpolation problem is considered: given \(2m\) complex numbers \(\{\alpha_ \nu\}_{0\leq\nu<m}\) and \(\{\beta_ \nu\}_{0\leq\nu<m}\), is there a unique \(t_ m\in T_{m,\varepsilon}\) such that \[ t_ m(x_{2\nu})=\alpha_ \nu, \qquad t_ m^{(M)}(x_{2\nu+1})=\beta_ \nu \] hold for all \(0\leq\nu<m\)? The authors completely solve the problem by showing that the answer is yes if and only if \(\varepsilon=1\) and \(m\) is odd if \(M\) is odd, and \(\varepsilon=0\) (and \(m\) is arbitrary) if \(M\) is even. They also explicitly determine the fundamental polynomials \(r_ j(x)\) of this \((0;M)\)-interpolation, and they also discuss the convergence properties of the associated operators \[ (L_ nf)(x)=\sum_{j=0}^{m- 1} f(x_{2j})r_{2j}(x)+\sum_{j=0}^{m-1} f^{(M)} (x_{2j+1}) r_{2j+1}(x) \] and \[ (L_ nf)(x)=\sum_{j=0}^{m-1} f(x_{2j}) r_{2j}(x)+\sum_{j=0}^{m-1} \beta_{2j+1}r_{2j+1}(x), \] where in the latter case the \(\beta_ j\)'s are associated with \(f\) in a prescribed way (say \(\beta_ j\equiv 0\)).