Extreme points of the Vandermonde determinant on the sphere and some limits involving the generalized Vandermonde determinant (Q1980748)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Extreme points of the Vandermonde determinant on the sphere and some limits involving the generalized Vandermonde determinant
scientific article

    Statements

    Extreme points of the Vandermonde determinant on the sphere and some limits involving the generalized Vandermonde determinant (English)
    0 references
    0 references
    0 references
    0 references
    8 September 2021
    0 references
    Let \(\mathbf{x}_n=(x_j)\in\mathbb{K}^n\) and \(\mathbf{a}_m=(a_j)\in\mathbb{K}^m\), where \(\mathbb{K}\) is \(\mathbb{R}\) or \(\mathbb{C}\). The generalized Vandermonde matrix \(G_{mn}(\mathbf{x}_n,\mathbf{a}_m)\in\mathbb{K}^{m\times n}\) with \((i,j)\)-th entry \(x_j^{a_i}\). The determinant of \(G_n(\mathbf{x}_n,\mathbf{a}_n)=G_{nn}(\mathbf{x}_n,\mathbf{a}_n)\) is the generalized Vandermonde determinant: \(g_n(\mathbf{x}_n,\mathbf{a}_n)=\det{G_n(\mathbf{x}_n,\mathbf{a}_n)}\). The Vandermonde matrix \(V_{mn}(\mathbf{x}_n)=G_{mn}(\mathbf{x}_n,\mathbf{b}_m)\) and the Vandermonde determinant \(v_n(\mathbf{x}_n)=g_n(\mathbf{x}_n,\mathbf{b}_n)\), where \(\mathbf{b}_m=(0,1,\dots,m-1)\). The authors first visualize \(v_3(\mathbf{x}_3)\) and also \(g_3(\mathbf{x}_3,\mathbf{a}_3)\) for certain choices of \(\mathbf{a}_3\). Second, they prove that a point \(\mathbf{x}_n\) on the unit sphere of \(\mathbb{R}^n\) is an extreme point of \(v_n\) if and only if \(x_1,\dots,x_n\) are distinct roots of the rescaled Hermite polynomial \[ P_n(x)=\frac{H_n(c_nx)}{(2c_n)^n},\quad c_n=\sqrt{\frac{n(n-1)}{2}}. \] They credit this result to \textit{G. Szegõ} [Orthogonal polynomials, Amer. Math. Soc. (1939; Zbl 0023.21505)], but their proof is different. They also give a recursive formula for the coefficients of \(P_n(x)\) and find the roots explicitly for \(n\le 7\). Third, the authors prove that \(\mathbf{ x}_n\) maximizes \(|v_n(\mathbf{x}_n)|\) over the real unit sphere if and only if \[ \sum_{\substack{i,j=1\\ i<j}}^n\frac{1}{(x_j-x_i)^2}=\frac{1}{2}\Big(\frac{n(n-1)}{2}\Big)^2,\quad \sum_{i=1}^nx_i^2=1. \] Fourth, they study limits involving these matrices or determinants. For example, they show that \[ G_{mn}(\mathbf{x}_n,\mathbf{a}_m)=\lim_{k\to\infty}V_{km}(\mathbf{a}_m)D_kV_{kn}(\log{\mathbf{x}_n}), \] where the limit is entrywise, \[ D_k=\textrm{diag}\,\Big(\frac{1}{0!},\frac{1}{1!},\dots,\frac{1}{(k-1)!}\Big), \] each \(x_j\ne 0\), and \(\log{\mathbf{x}_n}=(\log{x_1},\dots,\log{x_n})\). (In the complex case, the branch of logarithm is fixed and defines \(x_j^{a_i}=e^{a_i\log{x_j}}\).) For the entire collection see [Zbl 1467.16001].
    0 references
    Vandermonde matrix
    0 references
    Vandermonde determinant
    0 references

    Identifiers