Reconsideration of the multivariate moment problem and a new method for approximating multivariate integrals

Abstract

Due to its intimate relation to Spectral Theory and Schrödinger operators, the multivariate moment problem has been a subject of many researches, so far without essential success (if one compares with the one--dimensional case). In the present paper we reconsider a basic axiom of the standard approach - the positivity of the measure. We introduce the so--called pseudopositive measures instead. One of our main achievements is the solution of the moment problem in the class of the pseudopositive measures. A measure \ $μ$ is called pseudopositive if its Laplace-Fourier coefficients $μ_{k,l}(r) ,$ $r\geq0,$ in the expansion in spherical harmonics are non--negative. Another main profit of our approach is that for pseudopositive measures we may develop efficient ''cubature formulas'' by generalizing the classical procedure of Gauss--Jacobi: for every integer \ $p\geq1$ we construct a new pseudopositive measure $ν_{p}$ having ''minimal support'' and such that $μ(h) =ν_{p}(h) $ for every polynomial $h$ with $Δ^{2p}h=0.$ The proof of this result requires application of the famous theory of Chebyshev, Markov, Stieltjes, Krein for extremal properties of the Gauss-Jacobi measure, by employing the classical orthogonal polynomials $p_{k,l;j},$ $j\geq0,$ with respect to every measure $μ_{k,l}.$ As a byproduct we obtain a notion of multivariate orthogonality defined by the polynomials $p_{k,l;j}$. A major motivation for our investigation has been the further development of new models for the multivariate Schrödinger operators, which generalize the classical result of M. Stone saying that the one--dimensional orthogonal polynomials represent a model for the self--adjoint operators with simple spectrum.