Some norm inequalities for operators (Q782506)

The article shows log-convexity of some functions defined using some unitarily invariant norm of operators on Hilbert space. Several classical inequalities are then proved based on the log-convexity property. From Lemma~2 to the end of Section~2, the results could be presented in a more concise and complete form of log-majorization. Theorems 5 and 6 are well known in the literature, and the proofs do not seem to offer new insights. We list the latter as example. Theorem 6. Let \(A, B\) be \(n\times n\) positive definite matrices. Then, for \(t\in[0,1]\), \[ \prod_{i=1}^k \lambda_i(A\#_tB)\le \prod_{i=1}^k \lambda_i(A^{1-t} B^t), \quad k=1, \dots n. \] The equality case due to determinant consideration is not listed in the theorem. For better readability, the \(t\)-geometric mean \(A \#_t B\) should be defined. Indeed, Theorem~6 follows from a result of [\textit{H. Araki}, Lett. Math. Phys. 19, No. 2, 167--170 (1990; Zbl 0705.47020)]. Theorem. Let \(A, B\) be \(n\times n\) positive definite matrices. For any \(t \in [0, 1]\) and \(s > 0\), \(\lambda(A\#_{t}B) \prec_{\log} \lambda\left( A^{(1-t)s}B^{ts} \right)^{1/s}\). See also [\textit{M. Liao} et al., J. Lie Theory 24, No. 3, 725--736 (2014; Zbl 1331.15011)] for an extension in the context of symmetric space, which has a geometric flavour.