Hausdorff means and moment sequences (Q539268)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Hausdorff means and moment sequences
scientific article

    Statements

    Hausdorff means and moment sequences (English)
    0 references
    0 references
    0 references
    27 May 2011
    0 references
    A moment sequence (m.s for short) \(\mu=(\mu_0\mu_1,\dots)\), defined by \(\mu_n=\int_I t^n\,d\mu(t)\) \((n=0,1,\dots)\) (here \(d\mu(t)\) is a finite non-negative Borel measure on \(I\)), is said to be a Hausdorff m.s if \(I=[0,1]\). In the case where \(I\) is \((-\infty,\infty)\) or \([0,\infty)\), \(\mu\) is called a Hamburger and Stieltjes m.s, respectively. By the well-known positivity criterion, a sequence \(\mu\) is a m.s (associated with the interval \(I\)) if and only if \(\sum_{n=0}^N c_n\mu_n\geq 0\) for every polynomial \(\sum_{n=0}^N c_n x^n\) that is non-negative on \(I\). Applying this criterion, the author shows that if \(\sigma\) and \(\tau\) are m.ss of the same type, then so are their Hausdorff convolutions \(\sum_{k=0}^n h_{nk}(\mu)\sigma_k\tau_{n-k}\), where \(H_\mu=(h_{nk}(\mu))\) is a Hausdorff mean (Theorem 2). Then, by using this result, he proves the following. Theorem 1. If \(a_1,\dots,a_N\) are real numbers and \(b_1,\dots,b_N\) are positive, then the sequence \(q\), defined formally by \[ \prod_{m=1}^N\frac{1}{(1-a_m t)^{b_m}}= \sum_{n=0}^\infty\binom {b_1+\cdots+b_N+n-1}{n} q_n t^n, \] is a Hamburger m.s. If, moreover, \(a_1,\dots,a_N\) are nonnegative, then \(q\) is a Stieltjes m.s. The author gives an integral representation for \(q\) and discusses some nice applications of Theorem 1 to Fibonacci and Stirling numbers, Haber's inequality \((\frac{x+y}{2})^n\leq \frac{x^n+x^{n-1} y+\cdots+y^n}{n+1}\) \((x,y\geq 0)\), positivity of polynomials, counting sequences, etc. By using the positivity criterion, he gives the following characterization of triangular matrices \(A=(a_{nk})\) that preserve Stieltjes m.ss. Theorem 6. Suppose that for every Stieltjes m.s \(\sigma\), the sequence \(\rho\), defined by \(\rho_n=\sum_{k=0}^n a_{nk}\sigma_k\) \((n=0,1,\dots)\), is a Stieltjes m.s. Then so is \(\sum_{k=0}^n a_{nk}\sigma_k\tau_{n-k}\), the \(A\)-convolution of Stieltjes m.ss \(\sigma\) and \(\tau\). Theorem 6 applies when \(A\) is the Stirling kernel of the first or second kind, or if it is the Eulerian kernel.
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    moment sequences
    0 references
    Hausdorff mean
    0 references
    complete symmetric function
    0 references
    0 references