Existence and uniqueness of the Karcher mean on unital \(C^\ast\)-algebras (Q2287221)

The purpose of this paper is to extend the theory of the Karcher means to the cone of invertible positive elements in a unital \(C^{\ast}\)-algebra \(\mathbb{A}\). Let \(\mathcal{H}(\mathbb{A})\) be the set of all Hermitian elements endowed with the norm metric and topology and let \(\Omega\) be the open cone of positive invertible elements in \(\mathbb{A}\), endowed with the Thompson metric \(d(x,y)=\left\Vert \log(x^{-1/2}yx^{-1/2})\right\Vert \). This metric makes \(\Omega\) a complete metric space and, moreover, the Thompson metric topology agrees with the relative norm topology. The Karcher mean (or least squares mean) \(x=\Lambda(\mathbf{\omega},\mathbf{a})\) of a string \(\mathbf{a}=(a_{1},\dots,a_{n})\in\) \(\Omega^{n}\) and weight \(\mathbf{\omega}=(w_{1},\dots,w_{n})\) (with entries non-negative real numbers summing to 1) is defined via Theorem 7.4 as the unique solution in \(\Omega\) of the Karcher equation \(\sum_{i=1}^{n}w_{i}\log\left( x^{-1/2}a_{i}x^{-1/2}\right) =0\). Let \(\mathcal{P}^{1}(\Omega)\) be the set of Borel probability measures \(\mu\) on \(\Omega\) with finite first moment, that is, for some (and hence for all) \(y\in\Omega,\) we have \(\int_{\Omega}d(x,y)\,\mathrm{d}\mu(x)<\infty\). The Karcher mean can be extended by density from the set of finitely supported probability measures to \(\mathcal{P}^{1}(\Omega)\) (see the Appendix), thus providing a contractive map \(\Lambda:\mathcal{P}^{1}(\Omega)\rightarrow\Omega\) (from the Wasserstein metric to the Thompson metric) called the Karcher barycenter map. In addition, if \(\mu\in \mathcal{P}^{1}(\Omega)\), then \(x=\Lambda(\mu)\) is the unique solution of the Karcher equation \(\int_{\Omega}\log (x^{-1/2}yx^{-1/2})\,\mathrm{d}\mu(y)=0\). One basic tool is the fact (proved in Section~6) that the Karcher mean is separately real analytic in each of its coordinates. As the author mentions in the Introduction, this extension was also proved, using a different argument, by \textit{Y.-D. Lim} and \textit{M. Pálfia} [Adv. Math. 376, Article ID 107435, 46 p. (2021; Zbl 1482.47143)].