On (0,2) interpolation for the Laguerre abscissas (Q1096817): Difference between revisions

Let \((0<)x_ 1<x_ 2<...<x_ n\) be the roots of the Laguerre polynomials \(L_ n^{(\alpha)}(x)\) with \(\alpha >1\), and let \(f(x)\) be a continuous function on the half-line \([0,\infty)\), then the \((0,2)\) interpolating polynomials \(R_ n(f,x)=\sum^{n}_{k=1}f(x_ k)r_ k(x)\) of degree at most 2n-1 associated with f(x) are defined by the relations (1) \(R_ n(f,x_ k)=f(x_ k)\), \(R_ n''(f,x_ k)=0\), \(k=1,2,...,n\), (2) \(r_ k(x_ j)=\delta_{kj}\), \(r_ k''(x_ j)=0\), \(k,j=1,2,...,n\). The authors prove that the necessary and sufficient condition for this (0,2) interpolation problem to have a unique solution is that \(P_ n^{(\alpha,\alpha +1/2-n)}(-1/3)\neq 0\), where \(P_ n^{(\alpha,\beta)}(x)\) is the generalized Jacobi polynomial of degree n with parameters \(\alpha\) and \(\beta\). The following sufficient condition for unique solvability \[ n>\frac{(3\alpha -1)(\alpha +3)}{8}\text{ when \(\gamma =\frac{\alpha+1}{2}\) is an integer, }\frac{\gamma +(2\gamma +)[\gamma]-[\gamma]^ 2}{\gamma -[\gamma]}\text{ otherwise } \] is given. They also develop a method for finding the fundamental polynomials \(r_ k(x)\) satisfying (2) when \(\alpha =2k+1\), \(k=0,1,2,..\). but discuss the case for \(\alpha =1\); only as for the other values of \(\alpha\) are the formulas for \(r_ k(x)\) very complicated.