No label defined (Q4387004)

The \((0,2)\) interpolation is a well-studied special case of Birkhoff (or lacunary) interpolation: Let \((a,b)\) be an interval, \(n\) a positive integer and \(a <x_{n,n} < \dots <x_{1,n} <b\) distinct points. Determine a polynomial \(R_n\) of lowest degree satisfying \(R_n(x_{k,n})=y_{k,n}\), \(R_n''(x_{k,n})=y_{k,n}''\) with arbitrarily given real numbers \(y_{k,n}\), and \(y_{k,n}''\). In the weighted interpolation problem the second condition is replaced by \((wR_n)''=y_{x,n}''\) with a given weight function \(w\). The paper under review gives a well-written survey on results for \((0,2)\) interpolation. The situation is rather more complicated than for Lagrange or Hermite interpolation. If \(n\) is odd and the points \(x_{k,n}\) are symmetric with respect to 0, then there is in general no polynomial \(R_n\) of degree \(2n-1\) satisfying the interpolation conditions. If there is such a polynomial, then there are infinitely many. The situation when \(n\) is even and the \(x_{k,n}\) are the roots of one of the familiar orthogonal polynomials, is completely solved. Then in most cases solutions exist and even explicit representations are known. Uniform convergence and convergence in the mean can be established for \((0,2)\) interpolating polynomials on the roots of the integrated Legendre polynomials for functions \(f\) with sufficiently smooth derivative \(f'\). Weighted \((0,2)\) interpolation can be treated in a uniform way on the roots of all classical orthogonal polynomials with a suitably chosen weight \(w\) (which is not the same as the weight \(\rho\) used for the orthogonalization). In general there is no interpolating polynomial of degree \(2n-1\), but a polynomial of degree \(2n\) can be constructed in a relatively simple way. Uniform convergence has been proved for weighted \((0,2)\) interpolation on the roots of several classes of orthogonal polynomials under the hypothesis that \(f'\) is smooth, e.g., \(f'\) is in a Lipschitz class. Some of these results can be generalized to other types of lacunary interpolation.