Asymptotic positivity of Hurwitz product traces: two proofs (Q846296): Difference between revisions

\textit{D. Bessis, P. Moussa} and \textit{M. Villani} (BMV) conjectured [J. Math. Phys. 16, 2318--2325 (1975; Zbl 0976.82501)] that for any Hermitian \(n\times n\) matrices \(A\) and \(B\), the function \(\mu(t)=tr\exp (A-tB)\) with \(t\in\mathbb{R}\) is the Laplace transform of a positive measure on \([0,+\infty)\), provided \(B\) is positive. The above conjecture is proved by \textit{E. H. Lieb} and \textit{R. Seiringer} [J. Stat. Phys. 115, No.~1--2, 185--190 (2004; Zbl 1157.81313)] to be equivalent to the assertion that, for all positive integers \(m\) and all positive Hermitian \(A\) and \(B\), the polynomial \( tr(A+tB)^m=\sum_k trS_{m,k}(A,B)t^k\) has nonnegative coefficients only, where the Hurwitz product \(S_{m,k}(A,B)\) equals the sum of all words in \(A\) and \(B\), containing \(m-k\) letters \(A\) and \(k\) letters \(B\). The present authors investigate the behaviour of the terms \(trS_{m,k}(A,B)\) for large \(m\) instead of small \(m\) which the most of the authors used to do. The main result of the paper is that given two positive Hermitian matrices \( A,B\in\mathbb{R}^{n\times n}\) and \(k\in\mathbb{N}\) there is some \(m_{0}\in\mathbb{N}\), such that: a) \(AB=0\Rightarrow \) \(trS_{m,k}(A,B)\) for any integer \(m\neq k\neq 0\), and b) \(AB\neq 0\Rightarrow \) \(trS_{m,k}\left( A,B\right) >0\) for any integer \(m\geq m_{0}\) where \(m_{0}\) depends on \(A,B,k\). This result is proved by two different ways: one using complex analysis methods and one using combinatorics.