The Van der Waerden conjecture for mixed discriminants (Q818049)

If \(A^ k=(a^ k_{ij})\) are \(n\times n\) complex matrices with \(k=1,2,\dots,n\), then their mixed discriminant \(D(A^ 1,\dots,A^ n)\) is \(\frac{1}{n!} \sum_{\sigma \in S_ n}\det (a_{ij}^{\sigma (j)})\), where \(S_ n\) is the symmetric group of degree n. If all the \(A^ k\) are equal this turns out to be \(\det A\), whereas if each \(A^ k\) is a diagonal matrix the mixed discriminant equals \(\frac{1}{n!}\text{per}(a^ j_{ii})\), where \(\text{per}(X)\) is the permanent of the matrix \(X\). The author proves that the mixed discriminant of doubly stochastic \(n\)-tuples of semidefinite Hermitian \(n\times n\) (complex) matrices is bounded by \(\frac {n!}{n^n}\) and that this bound is uniquely attained at the \(n\)-tuple \((\frac 1nI_n,\dots, \frac 1nI_n)\), where \(I_n\) is the \(n\times n\) identity matrix. This result is an analogue of the (classical) verified Van der Waerden conjecture that \(\{\text{per}(A)| A\) is a doubly stochastic matrix of degree \(n\}\) attains its mimimum \(\frac {n!}{n^n}\) at \(J_n\), where \(J_n\) is the doubly stochastic matrix with all entries equal to \(\frac 1n\). This result in the real matrix case was conjected by \textit{R.~B. Bapat} [Linear Algebra Appl. 126, 107--124 (1989; Zbl 0696.15007)]. The real symmetric doubly stochastic \(n\)-tuples was proved by the author and \textit{A. Samorodnisky} [Proceedings of the 32 ACM Symposium on Theory of Computing, ACM Press (2000)].