A characterization of convex functions and its application to operator monotone functions

Abstract

We give a characterization of convex functions in terms of difference among values of a function. As an application, we propose an estimation of operator monotone functions: If $A \geq B \ge 0$, $A-B$ is invertible and $f$ is operator monotone on $(0, \infty)$, then $ f(A) - f(B) \ge f(\|B\|+ \epsilon) - f(\|B\|) > 0$, where $\epsilon = \|(A-B)^{-1}\|^{-1}$. Moreover it gives a simple proof to Furuta's theorem: If $\log A$ is strictly greater than $\log B$ for invertibel operators $A, \ B \geq 0$ and $f$ is operator monotone on $(0, \infty)$, then there exists a positive number $\beta$ such that $ f(A^\alpha)$ is strictly greater than $f(B^\alpha)$ for all positive numbers $ \alpha \le \beta$.