A characterization on singular value inequalities of matrices (Q2305185)

The authors recap some inequality bounds for singular values of block matrices. If \( s_1(X), \dots, s_n(X) \) denote the singular values of a matrix \(X\in \mathbb{M}_n(\mathbb{C})\) in decreasing order, the following non-trivial inequality \[ s_j(B)\le \dfrac{1}{2}s_j(D)\le s_j(A\oplus C), \, \,\, j=1,\dots,n, \] where \(D=\begin{pmatrix}A&B\\B^*&C\end{pmatrix}\) is positive semi-definite with \(n\times n\) size blocks, is discussed (see [\textit{X. Zhan}, SIAM J. Matrix Anal. Appl. 22, 819--823 (2001; Zbl 0985.15016); \textit{Y. Tao}, Linear Algebra Appl. 416, 724--729 (2006; Zbl 1106.15013)]). The main theorem stated in the paper is the following. Theorem. Given \(p\) \(n\times n \) matrices \(A_i\), \(i=1,\dots,p\), and the \(p\)-tuple \((\lambda_1,\dots,\lambda_p)\in \mathbb{C}^p\) such that \(\sum_{i=1}^p|\lambda_i|\le 1\), there holds \[ s_j(\sum_{i=1}^p \lambda_i A _i)\le s_j(\bigoplus_{i=1}^p A _i), \,\,\, j=1,\dots,n. \] See [\textit{O. Hirzallah} and \textit{F. Kittaneh}, Linear Algebra Appl. 424, 71--82 (2007; Zbl 1116.47012)] for related inequalities.