Spherical subcomplexes of spherical buildings (Q1945757)

Every spherical building \(\Delta\) (those whose associated Coxeter group is finite) is topologically spherical (that is, \((n-1)\)-connected if the building is \(n\)-dimensional or, equivalently, of rank \(n+1\)) by the Solomon-Tits theorem. In the article under review it is shown that certain subcomplexes of \(\Delta\) have this property, too. Theorem A states that if \(M \subseteq \Delta\) is open and convex, then \(\Delta \setminus M\) is spherical. The full subcomplex of simplices contained in \(\Delta \setminus M\) is a deformation retract and therefore spherical as well. The proof uses a clever filtration of the apartments of \(\Delta\) due to \textit{A. von Heydebreck} [Isr. J. Math. 133, 369--379 (2003; Zbl 1046.55011]. Theorem B explains when and how the analogous statement can fail if \(M\) is a closed ball of radius \(\pi/2\). Regarding the center \(n\) of \(M\) as the north pole, \(\Delta \setminus M\) can be regarded as the open (southern) hemisphere (though, of course, it is not a topological hemisphere) and its boundary as the equator. The full subcomplex of simplices contained in the open hemisphere is called the open hemisphere complex. In this case, \(\Delta \setminus M\) is still spherical of the dimension of \(\Delta\) unless \(\Delta\) is reducible and a whole join factor is contained in the equator. In that case, the dimension of the open hemisphere complex drops accordingly: it is the dimension of the join of all join factors not contained in the equator. Theorem B states that the open hemisphere complex is spherical of this dimension. This result can be seen in contrast to the situation of the complex of chambers opposite a fixed chamber, which is known not to be spherical in small cases [\textit{P. Abramenko}, Twin buildings and applications to \(S\)-arithmetic groups. Berlin: Springer (1996; Zbl 0908.20003)]. Beyond their intrinsic beauty, these results come with a concrete application in mind. Namely, if a Euclidean building is filtered using as Morse function a Busemann function (or a function that locally looks like a Busemann function) then the descending links are just hemisphere complexes (the north pole being the gradient of the Morse function). Via this connection Schulz's work has triggered and is key ingredient to, progress in the understanding of finiteness properties of lattices on Euclidean buildings, notably \textit{K.-U. Bux} and \textit{K. Wortman} [Invent. Math. 185, No. 2, 395--419 (2011; Zbl 1237.20041)] and \textit{K.-U. Bux} et al. [Ann. Math. (2) 177, No. 1, 311--366 (2013; Zbl 1290.20039)]. The proof of Theorem B takes up most of the paper and is more intricate than may be apparent from the statement. Not surprisinlgy, it involves a filtration from the open hemisphere complex to the closed hemisphere complex (the full subcomplex of simplices contained in the closed hemisphere, which is covered by Theorem A). But along the filtration, it is not true (and cannot be true) that the descending links have the right connectivity. Instead, the author retracts boundary spheres inside the sublevel sets. This involves an well-chosen ordering of the simplices which is closely related to the one later discovered for Euclidean buildings by Bux and Wortman [loc. cit.].