Convex rank 1 subsets of Euclidean buildings (of type \(A_{2}\)) (Q2481652): Difference between revisions

Euclidean buildings share many properties with symmetric spaces of noncompact type: They are CAT(0) length spaces (``Hadamard spaces'') with a boundary at infinity (a spherical building) and they contain sufficiently many flats (``appartments''). A general question for such spaces is: Given a convex subset \(A\) in the infinite boundary \(\partial X\) of a Hadamard space \(X\), can it be filled, i.e., is there some \(C \subset X\) with \(\partial C = A\)? In the present article this question is posed and answered when \(X\) is a Euclidean building of type \(A_2\) and \(A\) a finite zero-dimensional subbuilding, i.e. any pair of points in \(A\) is opposite (for some appartment containing both points). There is a necessary condition which turns out to be also sufficient. Any three points \(\xi,\eta,\zeta \in A\) must be mid points in their Weyl chambers and bound a tripod: a geodesic tree \(T \subset X\) with \(\partial T = \{\xi,\eta,\zeta\}\). Such \(T\) exists for \(\xi,\eta,\zeta\) iff the so-called holonomy is trivial which is defined as follows. Since \(\xi,\eta\) are opposite, the set \(X_\xi\) and \(X_\eta\) of rays parallel to \(\xi\) and to \(\eta\) respectively can be identified, and similarly there are isomorphisms between \(X_\eta\), \(X_\zeta\) and between \(X_\zeta\), \(X_\xi\). Composing these three isomorphisms yields an automorphism of \(X_\xi\) called holonomy which in the \(A_2\)-case is just a shift on a real line; triviality of holonomy means that this is the zero shift. There is an extension to infinite sets \(A\) under additional conditions.