More results on singular value inequalities of matrices (Q2496642)

The well-known arithmetic-geometric mean inequalities due to \textit{R. Bhatia} and \textit{F. Kittaneh} [SIAM J. Matrix Anal. Appl. 11, No. 2, 272--277 (1990; Zbl 0704.47014)] state that \[ 2s_j(AB^*)\leq s_j(A^*A+B^*B), \quad j=1,2, \dots,n, \] for \(n\times n\) complex matrices \(A,B\), where \(s_j(M)\) denotes the \(j\)th largest eigenvalue of \((AA^*)^{1/2}\). The author proves the new equivalent form: \[ 2s_j(K)\leq s_j\left[ \begin{smallmatrix} M & K \\ K^* & N\end{smallmatrix}\right], \quad j=1,2,\dots,r, \] where all matrices are complex, \(M\) is \(m\times m\), \(N\) is \(n\times n\), \(r=\min\{m,n\}\), and the block matrix is positive semidefinite. Some special cases and related generalizations are derived.