Let $X$ be a complete CAT(0) space with a proper and cocompact group action by isometries, and suppose there are $\xi, \xi' \in \partial X$ with $\angle (\xi, \xi') < \pi$. Using proposition 9.5 (3) of Chapter II.9 from the book Metric spaces of non-positive curvature by Bridson and Haefliger, there exist two points $\\eta, \eta' \in \partial X$ and a point $y\in X$ such that $\angle_y (\eta,\eta') =\angle(\xi,\xi')$. Then by corollary 9.9 of the same chapter, the convex hull of two geodesic rays $c, c'$ issuing from $y$ with $c(\infty) = \eta$, $c'(\infty) = \eta'$ is isometric to a sector in $\mathbb{E}^2$ bounded by two rays meeting at the angle $\angle (\xi, \xi')$.

The above result gives geodesic rays with possibly different endpoints. So I would like to know whether we can have a stronger result that there are geodesic rays $c_1, c'_1$ issuing from some point $y_1\in X$ with endpoints $\xi,\xi'$ such that the convex hull of these rays is isometric to a flat sector. I think this should be true in general, but maybe there is a counter-example that I do not know.

  • 1
    $\begingroup$ I think one can prove that there is a isometric embedding of a round circle into $\partial X$ containing $\xi, \xi'$ with respect to the Tits metric (there is a unique isometric arc connecting $\xi, \xi'$ of length $\angle(\xi, \xi')$, which will be an arc of this circle. So I think one could answer your question if you could prove that there is a geodesic connecting points of Tits distance $=\pi$. $\endgroup$
    – Ian Agol
    Commented Jan 28, 2012 at 7:07

1 Answer 1


Xie proved that this is true if $X$ is a 2 dimensional complex. I recommend reading his article: "The Tits Boundary of a CAT(0) 2-Complex", Trans. AMS., 357, no. 4, 1627-1661. I think that the answer to the more general question above is unknown. The techniques used in Xie's paper are further developed in an article by Bestvina, Kleiner, and Sageev: "Quasi-flats in CAT(0) complexes", arXiv:0804.2619v1 [math.GR].


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.