Invariant means (Q1614704)

Among others the following result is offered. Let \(m: ]0,\infty[^2 \to]0,\infty[\) be symmetric, strictly increasing in both variables and a strict mean (\(\min(x,y)\leq m(x,y)\leq\max(x,y)\) with \(=\) iff \(x=y\)). Then there exists a unique solution \(M: ]0,\infty[^3 \to ]0,\infty[\) of the functional equation \(M[m(x,y),m(y,z),m(z,x)]=M(x,y,z) (x>0, y>0, z>0).\) This \(M\) is also a symmetric strict mean, strictly increasing in all three variables. This follows by a Gaussian mixing iteration argument [see, e.g., \textit{J. M. Borwein} and \textit{P. M. Borwein}, ``Pi and the AGM. A study in analytic number theory and computational complexity'' (1987; Zbl 0611.10001)]. A partly similar result is offered for the functional equation \(M[x,y,m(x,y)]=m(x,y).\) Conversely, for every symmetric, strictly increasing mean \(M\) there exists a symmetric mean \(m\) that satisfies the second equation but a counter example shows that this is not true for the first equation. Particular attention is given to the logarithmic mean \(L(x,y)=(y-x)/(\log y-\log x).\) Several open problems are stated.