scientific article; zbMATH DE number 169204

From MaRDI portal
Publication:4037852





Investigations on inequalities for singular values of the product of two quaternion matrices are presented and thereafter, making use of the relation between singular values and eigenvalues, the authors give an estimation for each eigenvalue of the product of two self-conjugate matrices \(A\) and \(B\) for \(A\geq 0\), \(B\geq 0\) or \(B>0\). To be specific, let \(Q\) be the quaternion field, \(A\in Q^{m\times n}\) and \(GL_ n(Q)\), \(SC_ n(Q)\), \(GU_ n(Q)\) be the set of nonsingular, self-conjugate and generalized unitary matrices, respectively, and \(\sigma_ i(A)\), \(\lambda_ i(A)\) be the singular values and eigenvalues, respectively, then: 1) If \(A\in Q^{m\times n}\), \(B\in Q^{n\times s}\), then \(\sigma_ k(AB)\leq \min_{t=1,2,\dots,n-k+1} \{\sigma_ t(A)\sigma_{k- t+1}(B)\}\), \(k=1,2,\dots\), \(min(m,n,s)\); 2) If \(A,B\in GL_ n(Q)\), then \(\sigma_ k(AB)\geq\max_{t=1,2,\dots,n-k+1} \{\sigma_{k+t-1}(A) \sigma_{n- t+1}(B)\}\), \(k=1,2,\dots,n\); 3) If \(A,B\in SC_ n(Q)\), \(A\geq 0\), \(B\geq 0\), \(\text{rank }AB=m\), \(\text{rank }A=r\), \(\text{rank }B=s\), then \(\lambda_ k(AB)\leq \displaystyle{ \min_{{t=1,2,\dots,k} \atop {i=1,2,\dots,k-t+1}}} \{(\lambda_ t(A) \lambda_{k-t-i+2}(A))^{1/2} \lambda_ i(B),(\lambda_ t(B) \lambda_{k-t-i+2}(B))^{1/2} \lambda_ i(A)\}\), \(k=1,2,\dots,n\), and \(\lambda_ k(AB)\geq\displaystyle{\max_{{t=1,2,\dots,m-k+1} \atop {l=1,2,\dots,t}}} \{(\lambda_{k+t-1+r-m}(A)\lambda_{r- l+1}(A))^{1/2} \lambda_{m-t+l+n-r}(B)\), \((\lambda_{k+t-1+n-r}(B) \lambda_{m-l+1+n-r}(B))^{1/2} \lambda_{r-t+l}(A)\), \((\lambda_{k+t-1+s-m}(B)\lambda_{s- t+1}(B))^{1/2} \lambda_{m-t+l+n-s}(A)\), \((\lambda_{k+1-1+n-s}(A) \lambda_{m-l+1+n-s}(A))^{1/2} \lambda_{s-t+l}(B)\}\), \(k=1,2,\dots,m\). A few misprints are obvious.












This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4037852)