Singular value inequalities and applications (Q2045892)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Singular value inequalities and applications |
scientific article |
Statements
Singular value inequalities and applications (English)
0 references
16 August 2021
0 references
Let \({\mathbb M}_{n}\) be the algebra of \(n\times n\) complex matrices. The eigenvalues of \(|A|=(A^*A)^{\frac{1}{2}}\) are the singular values of \(A\in {\mathbb M}_{n}\) which are arranged as \(s_1(A)\geq s_2(A)\geq \cdots\geq s_n(A).\) A norm \(|\!|\!|\cdot |\!|\!|\) in \(M_n(\mathbb{C})\) is said to be \textit{unitarily invariant} if \(|\!|\!|UAV|\!|\!|=|\!|\!|A|\!|\!|\) for any \(A,U,V\in M_n(\mathbb{C})\) with \(U,V\) unitary. Examples of unitarily invariant norms are the \textit{Schatten \(p\)-norms} given by \(\|A\|_{p}=\left(\sum_{i=1}^n s^p_i(A)\right)^{\!\frac{1}{p}},\) \(p\geq 1, \) and the \textit{Fan \(k\)-norms} defined by \(\|A\|_{(k)}=\sum_{i=1}^k s_i(A),\) \(k = 1, \dots, n, \) including \(\|A\|=s_1(A)\). \textit{F. Kittaneh} [Positivity 10, No. 2, 251--260 (2006; Zbl 1127.47010)] proved that if \(A,B\in{\mathbb M}_n\) are positive semidefinite then \[ s_j(A+B)\leq s_j\left(\left(A+\left|B^{\frac{1}{2}}A^{\frac{1}{2}}\right|\right)\oplus\left(B+\left|A^{\frac{1}{2}}B^{\frac{1}{2}}\right|\right)\right), \tag{1} \] for \(j=1,\ldots,2n\). The previous inequality can be extended to unitarily invariant norms. For \(j=1,\) \((1)\) is the spectral norm inequality \[ \|A+B\|\leq \max\left\{\left(\| A+\left|B^{\frac{1}{2}}A^{\frac{1}{2}}\right|\|\right),\|B+\left|A^{\frac{1}{2}}B^{\frac{1}{2}}\right| \|\right\}, \tag{2} \] and specifying \((1)\) to the Schatten \(p\)-norms, we get \[ \|A+B\|_p\leq \left(\| A+\left|B^{\frac{1}{2}}A^{\frac{1}{2}}\right|\|_p^p+\|B+\left|A^{\frac{1}{2}}B^{\frac{1}{2}}\right| \|_p^p\right)^{\frac{1}{p}}, \tag{3} \] for \(p\geq 1\). In this paper, the author gives generalizations of the inequalities \((1)\), \((2)\) and \((3)\). Among other inequalities, the author proves that if \(A,B\) and \(X\) are \(n\times n\) complex matrices such that \(A\) and \(B\) are positive semidefinite, then \[ s_j(A\,X - X\,B)\leq s_j\left(\left(\frac{1}{2} A+\frac{1}{2}A^{\frac{1}{2}}\left|X^*\right|^2 A^{\frac{1}{2}}\right)\oplus\left(\frac{1}{2} B+\frac{1}{2}B^{\frac{1}{2}}\left|X\right|^2 B^{\frac{1}{2}}\right)\right), \] for \(j=1,\ldots,2n\).
0 references
concave function
0 references
positive semidefinite matrix
0 references
singular value
0 references
unitarily invariant norm
0 references
inequality
0 references
0 references