A higher-rank rigidity theorem for convex real projective manifolds (Q6142263)

The paper proves a rigidity theorem for higher-rank convex projective manifolds. A properly convex domain in \(\mathbb{P}(\mathbb{R}^d)\) is higher-rank if for any two points in the domain there exists a properly embedded simplex of dimension at least two that contains the projective line segment between the two points. Reducible and symmetric properly convex domains form two basic families of domains of higher-rank. The main theorem states that an irreducible properly convex domain of higher-rank with a cocompact action by a discrete group of projective transformations is already symmetric, meaning that there exists a semi-simple Lie group which preserves the domain and acts transitively on it. Thus for divisible domains the above two families constitute the only examples of higher-rank domains. This result should be thought of as a convex projective analogue of the higher-rank rigidity theorem of \textit{W. Ballmann} [Ann. Math. (2) 122, 597--609 (1985; Zbl 0585.53031)], \textit{K. Burns} and \textit{R. Spatzier} [Publ. Math., Inst. Hautes Étud. Sci. 65, 35--59 (1987; Zbl 0643.53037)] and \textit{K. Burns} and \textit{R. Spatzier} [Publ. Math., Inst. Hautes Étud. Sci. 65, 5--34 (1987; Zbl 0643.53036)], that states that every compact irreducible Riemannian manifold with non-positive curvature and higher rank is a locally symmetric space. Furthermore, for irreducible properly convex divisible domains, the author provides eleven ``higher-rank'' conditions that are each equivalent to the domain being symmetric of real rank at least two. The difficult part is to prove that any one of these eleven conditions implies symmetry of the domain. For this the author studies special semi-groups in \(\mathbb{P}(\textrm{End}(\mathbb{R}^d))\) associated to each boundary face. An important tool is the following dichotomy due to \textit{Y. Benoist} [Duke Math. J. 120, No. 1, 97--120 (2003; Zbl 1037.22022)]: An irreducible properly convex divisible domain is either symmetric or the group dividing it is Zariski-dense in \(\textrm{PGL}(d,\mathbb{R})\).