Semigroups of distributions with linear Jacobi parameters (Q1930535): Difference between revisions

Probability measures with finite moments on the real line can be realized as spectral measures of tridiagonal matrices and the corresponding entries are called Jacobi parameters. They also occur in the recursion formula for the orthogonal polynomials of the measure. In this paper, the authors study the problem to find all convolution semigroups of probability measures with Jacobi parameters linear in the convolution parameter. It turns out that this property is another characterization of the Meixner class. Allowing polynomial instead of linear dependence does not increase the class of measures. This analysis is worked out for classical convolution, for free convolution and for two-state (conditionally) free convolution. In the latter case, this class has not been explicitly described before and the authors provide a Laha-Lukacs type characterization in terms of certain conditional expectations, showing that this class also has Meixner type properties. The investigations are based on a detailed study of the relations between moments, cumulants, partitions, Jacobi parameters and continued fractions.