Some inequalities involving Heron and Heinz means of two convex functionals (Q2003781)

For \(\lambda \in[0,1]\), let \(a\nabla_\lambda b=(1-\lambda)a+\lambda b\) and \(a\sharp_\lambda b=a^{1-\lambda}b^\lambda \) be the \(\lambda\)-weighted arithmetic and geometric means of the numbers \(a,b>0\), respectively. The Heron and Heinz means of these numbers are defined by \(K_\lambda (a,b)=(1-\lambda)\sqrt{ab}+\lambda \frac{a+b}2\) and \(HZ_\lambda (a,b)=2^{-1}\big(a^{1-\lambda} b^\lambda+ a^{\lambda}b^{1-\lambda}\big)\), respectively.\par Many inequalities are known to hold between these means as well as between their operator analogs. The authors consider in this paper some functional analytic versions of these means by replacing \(a,b\) by two convex lsc proper functions \(f,g:H\to\mathbb{R}\cup\{\infty\}\), where \(H\) is a Hilbert space, as they were introduced in \textit{M. Raïssouli} and \textit{H. Bouziane} [Ann. Sci. Math. Qué. 30, No. 1, 79--107 (2006; Zbl 1222.49025)]. The functional arithmetic and geometric means are defined by \(\mathcal{A}_\lambda(f,g)=(1-\lambda)f+\lambda g\) and \[ \mathcal{G}_\lambda(f,g)=\frac{\sin(\pi\lambda)}{\pi}\int_0^1t^{\lambda-1}(1-t)^{-\lambda}\big((1-t)f^*+tg^*\big)^*dt, \] respectively, where \(h^*\) denotes the Fenchel conjugate of a convex lsc proper function \(h:H\to\mathbb{R}\cup\{\infty\},\, h^*(x^*)=\sup\{\Re e\langle x,x^*\rangle -h(x): x\in H\},\, x^*\in H\). The following inequalities hold: \(\mathcal{H}_\lambda(f,g)\le \mathcal{G}_\lambda(f,g)\le \mathcal{A}_\lambda(f,g),\) where \(\mathcal{H}_\lambda(f,g)=\big((1-t)f^*+tg^*\big)^*\) is the \(\lambda\)-weighted functional harmonic mean of \(f,g\). The functional Heron and Heinz means are defined by \(\mathcal{K}_\lambda(f,g)=(1-\lambda)\mathcal{G}(f,g)+\lambda \mathcal{A}(f,g)\) and \(\mathcal{HZ}_\lambda=2^{-1}\big(\mathcal{G}_\lambda(f,g)+\mathcal{G}_{1-\lambda}(f,g)\big)\), respectively. The authors prove some inequalities involving these means as well as for their operator versions. As they mention in the abstract: ``The operator versions of our theoretical functional results are immediately deduced. We also obtain new refinements of some known operator inequalities via our functional approach in a fast and nice way.''