On the spectral radius and the spectral norm of Hadamard products of nonnegative matrices (Q616280)

Let \(A_1,\dots,A_k,A,B\) be nonnegative \(n\times n\) matrices, and let \(\circ\), \(\rho\) and \(\|.\|\) denote the Hadamard product, spectral radius and spectral norm, respectively. The author proves the following: {\parindent6.5mm \begin{itemize}\item[(1)] \(\rho(A_1\circ\cdots\circ A_k)\leq\rho(A_1\cdots A_k)\), conjectured for \(k=2\) by \textit{X.~Zhan} [Advanced Workshop on Trends and Developments in Linear Algebra, ICTP, Trieste (2009)], proved in this case by \textit{K.~M.~R.~Audenaert} [Linear Algebra Appl.~432, 366--368 (2010; Zbl 1205.15034)] and \textit{R.~A.~Horn} and \textit{F.~Zhang} [Electron. J. Linear Algebra 20, 90--94 (2010; Zbl 1191.15015)]; \item[(2)] \(\|A_1\circ\cdots\circ A_k\|^2\leq\rho(A_1A_1^T\cdots A_kA_k^T)\); \item[(3)] \(\|A_1\circ\cdots\circ A_k\|^2\leq\rho(A_1^TA_2\cdots A_{k-1}^TA_k)\rho(A_1A_2^T\cdots A_{k-1}A_k^T)\) if \(k\) is even, \newline \(\|A_1\circ\cdots\circ A_k\|^2\leq\rho(A_1A_2^T\cdots A_{k-2}A_{k-1}^TA_kA_1^TA_2\cdots A_{k-2}^TA_{k-1}A_k^T)\) if \(k\) is odd. \end{itemize}} It follows from (3) that \(\|A\circ B\|\leq\rho(A^TB)\), improving Schur's inequality \(\|A\circ B\|\leq\|A\|\|B\|\). Finally, let \(A\) and \(B\) be complex \(n\times n\) matrices. Denote by \(\|.\|\), \(s_j\) and \(\prec_w\) a unitarily invariant norm, the \(j\)'th largest singular value and weak majorization, respectively. The author disproves the inequality {\parindent6.5mm \begin{itemize}\item[(4)] \((s_j^2(A\circ B))\prec_w(s_j(A\circ\bar{A})s_j(B\circ\bar{B}))\), conjectured by \textit{R.~Huang} [Linear and Multilinear Algebra~56, 543--554 (2008; Zbl 1153.15024)], and the stronger inequality \item[(5)] \(\|(A\circ B)(A\circ B)^*\|\leq\|(A\circ\bar{A})(B\circ\bar{B})^T\|\), conjectured by Zhan [op. cit.]. \end{itemize}} The question on the validity of (4) for nonnegative matrices remains open.