An extension theorem for non-compact split embedded Riemannian symmetric spaces and an application to their universal property (Q2663680)

This article deals with general symmetric spaces, general in the sense that no differential and or topological structure is assumed to exist on the underlying space. These symmetric spaces are defined in [\textit{W. Freyn} et al., Münster J. Math. 13, No. 1, 1--114 (2020; Zbl 1491.53068)] using nearly the same axioms as in the classical theory [\textit{O. Loos}, Symmetric spaces. I: General theory. II: Compact spaces and classification. New York etc.: W. A. Benjamin (1969; Zbl 0175.48601)] (the only difference is that the authors assume here that a reflection has only its center as fixed point, whereas this property is classically assumed only in a neighborhood of the center). Given an inclusion \(Y \subset X\) of symmetric spaces, the transvection group \(\operatorname{Trans}(Y \subset X)\) is the subgroup of the automorphism group of \(X\) generated by the elements \(s_{y_1} \circ s_{y_2}\) for \(y_1,y_2 \in Y\). The first main result is that if \(X\) is a general symmetric space (even more generally, the axiom about the fixed points of the reflections can be dropped) and \(Y\) is an embedded Riemannian symmetric space of non-compact split type, then \(\operatorname{Trans}(Y \subset X)\) is a perfect central extension of \(\operatorname{Trans}(Y):=\operatorname{Trans}(Y \subset Y)\). Assuming now that \(X\) is a symmetric space corresponding to an irreducible algebraically simply connected two-spherical split real Kac-Moody group \(G\), it is shown that \(X\) is a colimit of the hyperbolic planes and the rank-two Riemannian symmetric spaces of non-compact split type it contains. The proof of this second result is based on the main result in [\textit{P. Abramenko} and \textit{B. Mühlherr}, C. R. Acad. Sci., Paris, Sér. I, Math. 325, No. 7, 701--706 (1997; Zbl 0934.20024)] stating that \(G\) is a colimit of groups isomorphic to \(\mathrm{SL}_2(\mathbb{R})\). It also makes use of the fact that the universal central extensions of \(\operatorname{Trans}(Y \subset X)\), for \(Y\) varying among the symmetric spaces of rank one or two embedded in \(X\), satisfy a universal property which allows amalgamating them to a group acting on \(X\) (whereas the groups \(\operatorname{Trans}(Y \subset X)\) themselves don't: this is why the first result is needed).