$\newcommand{\vx}{\mathbf{x}}$ Let $e_k(\vx)$ denote the elementary symmetric polynomial, defined for $k=0,1,\ldots,n$ over a vector $\vx=(x_1,\ldots,x_n)$ by

\begin{equation*} e_k(\vx) := \sum_{1 \le l_1 < l_2 < \cdots < l_k \le n} x_{l_1}x_{l_2}\cdots x_{l_k}. \end{equation*}

To avoid messy conjugates in what follows, let me limit $x$ to be a real vector.

[EDIT]: Changed notation below to highlight that size of $M_k$ is independent of size of the vectors being used to define it.

Definition (GC-matrix) Let $(\vx^1,\ldots,\vx^m)$ be vectors in $\mathbb{R}^{n}$ such that each component $|x^p_j| < 1$ (for $1\le p \le m$, $1\le j \le n$). Denote by $\vx^i \circ \vx^j$ the vector formed by taking elementwise products; also, let $\mathbf{1}$ denote the length $n$ vector of all ones.

Define now the $m\times m$ Generalized Cauchy (GC) matrix $M_k$, for $1\le k\le n$, as \begin{equation*} M_k := \left[\frac{1}{e_k(\mathbf{1}-\vx^i\circ \vx^j)}\right]_{i,j=1}^m. \end{equation*}

Example For $e_1(\vx)=\sum_lx_l$, the GC matrix above essentially (up to scaling) boils down to a standard Cauchy matrix, of the form \begin{equation*} \left[\frac{1}{1-\lambda_i\lambda_j}\right], \end{equation*} for suitable scalars $\lambda_i$; while for $e_n(\vx)=\prod_l x_l$, the GC matrix reduces to a Hadamard product of ordinary Cauchy matrices.

Question. The above example shows that $M_1$ and $M_n$ are positive semidefinite (since the involved Cauchy matrices are). So my question is there a slick representation for $e_k$ that I could use to conclude positive semidefiniteness of $M_k$?


PS: I suspect the answer follows immediately from well-known (in the correct circles) results; but I need some MO help here to put me rapidly in the correct direction!

  • 1
    Unfortunately, this question has seems to have a negative answer (as proved to me very recently by one of the mathematicians with whom I discussed it; will update this question once I get a chance.) – Suvrit Oct 28 '14 at 3:32

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Browse other questions tagged or ask your own question.