Inequalities and identities for generalized matrix functions (Q1057339): Difference between revisions

Let \(\lambda\) be an irreducible complex character of \(S_ n\). The Schur polynomial afforded by \(\lambda\) is \[ s_{\lambda}(y_ 1,...,y_ n)=\frac{1}{n!}\sum_{\sigma}\lambda (\sigma)\prod^{n}_{i=1}y_ t^{t(\sigma)}, \] where \(t(\sigma)=the\) number of cycles of length t in \(\sigma\). The immanent afforded by \(\lambda\) is defined on n-by-n complex matrices by \(d_{\lambda}(A)=\sum_{\sigma}\lambda (\sigma)\prod^{n}_{i=1}a_{i\sigma (i)}.\) In 1918, Schur proved that the following inequality holds for all positive semidefinite Hermitian \(A: d_{\lambda}(A)/\lambda (id)\geq \det (A).\) Actually, Schur proved this result (and more) for characters of any subgroup G of \(S_ n\); which implied the Fischer and Hadamard determinantal inequalities. Since a 1964 paper of M. Marcus that established a permanental analog of the Hadamard determinant theorem, it has been conjectured that the following permanental analog of Schur's result holds for all positive semidefinite \((*)\quad A:\quad per(A)\geq d_{\lambda}(A)/\lambda (id).\) \textit{E. Lieb} [J. Math. Mech. 16, 127-134 (1966; Zbl 0144.268)] showed that (*) holds when \(\lambda =1\), the principal character, and G is a direct product of \(S_ k\) with \(S_{n-k}\). This gave a permanental analog to the Fischer inequality. \textit{T. Pate} [Linear Multilinear Algebra 11, 1-17 (1982; Zbl 0476.15005)] showed that (*) holds for the principal character when G is a Wreath product of \(S_ k\) with \(S_{n/k}\) (with some restrictions on k). In this paper the authors obtain two inequalities of type (*) for immanents and \(\lambda\) \(\neq 1\); the first results of this kind. Specifically, in Theorem 1 (of 8), they obtain that (*) holds when \(\lambda\) is a ''single hook'' character corresponding to the partitions (n-1,1) or (n-2,1,1). The technique used is to exploit the Littlewood correspondence between identities for Schur polynomials and immanents, which allows immanents to be expressed as sums of products of principal subpermanents or subdeterminants. The technique should be useful in establishing other results of type (*) for immanents.