Double piling structure of matrix monotone functions and of matrix convex functions (Q732072)

Let \(J\) be a (nontrivial) interval of the real line \(\mathbb{R}\) and let \(M_n\) be the algebra of \(n\times n\) matrices. A function \(f\) is called \(n\)-monotone in \(J\) if the inequality \(f(a) \leq f(b)\) holds for every pair of selfadjoint matrices \(a, b\in M_n\) such that \(a \leq b\) and all eigenvalues of \(a\) and \(b\) are contained in \(J\). The notion of \(n\)-convex function on \(J\) is similarly defined. The authors consider the following assertions: (i) \(f(0) \leq 0\) and \(f\) is \(n\)-convex in \([0, \alpha)\). (ii) For each matrix \(a\) whose spectrum is contained in \([0, \alpha)\) and each contraction \(c \in M_n\), \(f(c^*ac) \leq c^*f(a)c\). (iii) The function \(f(t)/t(=g(t))\) is \(n\)-monotone in \((0, \alpha)\). The authors prove that, for any \(n\in \mathbb{N}\), the two conditions (ii) and (iii) are equivalent; the assertion that \(f\) is \(n\)-convex with \(f(0)\leq 0\) implies that \(g(t)\) is \((n - 1)\)-monotone holds; the implication from (iii) to (i) does not hold even for \(n = 1\); and, under some conditions, (i) implies (ii). Similar assertions to (i), (ii) and (iii) for operator monotone and operator convex functions are well-known [see \textit{J.\,Pečarić}, \textit{T.\,Furuta}, \textit{H.\,Mičić Hot} and \textit{Y.\,Seo}, ``Mond--Pečarić method in operator inequalities.\ Inequalities for bounded selfadjoint operators on a Hilbert space'' (Monographs in Inequalities 1; Zagreb:\ ELEMENT) (2005; Zbl 1135.47012)].