Group actions and Helly's theorem (Q734822)

This is a remarkable paper introducing a new notion and proving several interesting results. In particular, the author defines a sequence of properties \(FA_n\) which generalize the classical property FA (Fixing property on Arbres) due to J.-P. Serre. The author defines : A group \(\Gamma\) has property \(FA_n\) if any isometric action of \(\Gamma\) on a CAT(0) cell complex \(X\) has a fixed point. We recall that the CAT(k)-spaces are some special metric spaces and were so named by M. Gromov after E. Cartan, A.D. Aleksandrov and V. A. Toponogov. The validity of the property \(FA_n\) for a group puts severe restrictions on it like the existence of only finitely many conjugacy classes of irreducible \((n+1)\)-dimensional representations over an algebraically closed field. The property FA was useful in distinguishing lattices in rank \(1\) groups and in higher rank groups. The notion of \(FA_n\) introduced in this paper makes it possible to distinguish between subgroups of higher rank groups as well. This notion of \(FA_n\) as well as the methods in this paper have already been used by other authors (M. Bridson, L. Funar and A. Barnhill) to prove other interesting results since this paper became available as a preprint. For instance, Bridson has proved that the mapping class group \(M_g\) has the strong \(FA_g\) property; apparently, it is an open problem to determine the maximal \(n(g)\) for which \(M_g\) has the strong \(FA_{n(g)}\) property -- one knows \(n(g) \leq 6g-7\). The property \(FA_1\) is the classical property FA. After introducing this property \(FA_n\), the author first observes : (i) \(FA_n\) implies \(FA_r\) for each \(r<n\). (ii) If \(G\) has \(FA_n\), then so does each quotient. (iii) Property \(FA_n\) for a normal subgroup \(N\) of \(G\) and for the quotient \(G/N\) implies its truth for \(G\) itself. (iv) If a subgroup of finite index in \(G\) has \(FA_n\), then so does \(G\). The author also defines a property called `Strong \(FA_n\) property' which implies \(FA_n\) (by a result due to Bridson) but we do not recall it here. The results in this paper apply to higher \(\mathbb {Q}\)-rank \(S\)-arithmetic groups, to simplex reflection groups (including some nonarithmetic ones constructed by Vinberg), and to Chevalley groups of higher rank over polynomial rings and Laurent series rings over \(\mathbb {Z}\) or over fields. More precisely, here are some results from this paper : (i) Let \(\Gamma\) be an \(S\)-arithmetic subgroup of an absolutely simple, simply connected algebraic group \(G\) over a number field \(k\) with \(r:= k\)-rank(\(G) \geq 2\). Then \(\Gamma\) has the strong property \(FA_{r-1}\). This includes, in particular, the assertions that \(SL_n(\mathbb {Z})\) and \(SL_n(\mathbb {Z}[1/p])\) have property \(FA_{n-2}\) when \(n \geq 3\). It should be noted that \(SL_n(\mathbb {Z}[1/p])\) does NOT have property \(FA_{n-1}\) as implied by its action on the Bruhat-Tits building of \(SL_n\) over \(\mathbb {Q}_p\). Note also that we have a way to distinguish between \(SL_n(\mathbb {Z}[1/p])\) and \(SL_m(\mathbb {Z}[1/p])\) (although the same cannot be said about \(SL_n(\mathbb {Z})\) and \(SL_m(\mathbb {Z})\)) for \(m \neq n\) now. (ii) Let \(R\) be a commutative ring, \(\Phi\) be a reduced, irreducible root system and \(E(\Phi,R)\) the elementary subgroup of a Chevalley group of type \(\Phi\) over \(R\). If \(\Phi\) has rank \(\geq 2\), then \(E(\Phi,R)\) has the strong property \(FA_{r-1}\). (iii) Let \(\Gamma_n\) be any discrete group generated by reflections in the sides of an \(n\)-simplex of finite volume in Euclidean or hyperbolic \(n\)-space. Then, \(\Gamma_n\) has the strong property \(FA_{n-1}\). Note that the groups \(\Gamma_n\) include some nonarithmetic lattices in \(O(n,1)\). The method of proof is also very interesting and leads to an assertion which is a common generalization of the first two statements. The generalization alluded to is : Let \(\Gamma\) be a finitely generated group and let \(\{\Gamma_1, \cdots, \Gamma_{r+1} \}\) be a collection of finitely generated nilpotent groups which generate a subgroup of finite index in \(\Gamma\). Suppose they also satisfy : (a) Each proper subset of this collection generates a nilpotent group, and: (b) There is \(m>0\) such that for each \(x \in \bigcup_i \Gamma_i\), there is a proper, nilpotent subgroup \(N\) of \(\Gamma\) satisfying \(x^m \in [N,N]\). Then, \(\Gamma\) has the strong property \(FA_{r-1}\). In view of the above statement, it is not surprising that the proof involves combinatorics of generators. There is a classical theorem of Helly which concerns the combinatorics of compact, convex sets in a Euclidean space. The classical theorem shows that if there is a finite family of compact, convex sets in \(\mathbb {R}^n\) so that any \(n+1\) of them intersect, then they must all intersect. The theorem had been generalized by several people and the author uses a topological generalization due to Debrunner which asserts : Let \(X\) be a normed topological space with the property that every nonempty open subset \(Y\) has reduced homology \(\tilde{H}_q(Y)=0\) for all \(q \geq n\). Let \(\{X_i \}\) be a finite collection of nonempty closed homology cells in \(X\). Suppose any \(r\) sets (with \(r \leq n+1\)) from this family intersect in a (nonempty) homology cell. Then, \(\bigcap_i X_i\) is a homology cell. For proving the generalization mentioned above, the author considers the action of \(\Gamma\) on an \((r-1\))-dimensional CAT(0)-complex and applies the above topological Helly theorem to the family \(X_i = Fix(\Gamma_i)\) by verifying that the necessary hypothesis holds. This paper is bound to be extensively cited in the future.