Matrix inequalities including Furuta inequality via Riemannian mean of \(n\)-matrices (Q2914892)

Let \(P_m(\mathbb{C})\) denote the set of complex positive definite \(m\times m\) matrices. For \(A,B\in P_m(\mathbb{C})\), define the Riemannian metric \(\delta_2(A,B)=\|\log{A^{-\frac{1}{2}}BA^{-\frac{1}{2}}}\|_2\), where \(\|\cdot\|_2\) stands for the Euclidean norm. Let \(\omega=(w_1,\dots,w_n)\) be a probability vector, and let \(A_1,\dots,A_n\in P_m(\mathbb{C})\). The weighted Riemannian mean \(\mathfrak{S}_\delta(\omega;A_1,\dots,A_n)\) is the matrix \(X\in P_m(\mathbb{C})\) minimizing \(w_1\delta_2^2(A_1,X)+\dots+w_n\delta_2^2(A_n,X)\).NEWLINENEWLINEFor \(p_1,\dots,p_n\geq 1\), let \(\omega'\) be the probability vector parallel with \((\frac{w_1}{p_1},\dots,\frac{w_n}{p_n})\). The author proves that if \(\mathfrak{S}_\delta(\omega;A_1,\dots,A_n)\leq I\) (Löwner partial order), then NEWLINE\[NEWLINE \mathfrak{S}_\delta(\omega';A_1^{p_1},\dots,A_n^{p_n})\leq\mathfrak{S}_\delta(\omega;A_1,\dots,A_n). NEWLINE\]NEWLINE This extends the generalizations of the Ando-Hiai inequality by \textit{T. Yamazaki} [Oper. Matrices 6, No. 3, 577--588 (2012; Zbl 1477.15018)] and by \textit{M. Fujii} and \textit{E. Kamei} [Linear Algebra Appl. 416, No. 2--3, 541--545 (2006; Zbl 1110.47011)].NEWLINENEWLINENow, for \(w_i>0\), \(p_i'\geq p_i>q_i\), \(i=1,\dots,n\), let \(\omega\) and \(\omega'\) be probability vectors parallel with \((\frac{w_1}{p_1-q_1},\dots,\frac{w_n}{p_n-q_n})\) and \((\frac{w_1}{p_1'-q_1},\dots,\frac{w_n}{p_n'-q_n})\), respectively. The main result (Theorem 3.3) gives a condition under which NEWLINE\[NEWLINE \mathfrak{S}_\delta(\omega';A_1^{p_1'},\dots,A_n^{p_n'})\leq\mathfrak{S}_\delta(\omega;A_1^{p_1},\dots,A_n^{p_n})\leq A_n^{q_n}. NEWLINE\]NEWLINE This extends the Furuta inequality as formulated by \textit{M. Fujii} et al. [Math. Jap. 45, No. 3, 519--525 (1997; Zbl 0904.47017)].