Higher generation by subgroups (Q5906504)

The paper under review is motivated by the following question: ``How well'' can the structure of a group \(G\) be described by a given family \(\mathcal H\) of subgroups of \(G\)? This question is closely related to the topological properties of \(N({\mathcal H})\), the nerve of the covering of \(G\) by all cosets \(gH\), \(H \in {\mathcal H}\). The authors call \(\mathcal H\) \(n\)- generating if \(N(\mathcal H)\) is \((n-1)\)-connected. This definition generalizes well known geometric criteria for \(G\) being generated by the elements of \(\mathcal H\) \((n = 1)\) and for \(G\) being the amalgamated product of the elements of \(\mathcal H\) (\(n = 2\)). The authors also derive a homological condition for \(\mathcal H\) being \(n\)-generating if \(n > 2\) (cf. Theorem 2.4(c)). The case \(n = 3\) leads to identities among relations and is treated in some detail in Section 4. The result obtained in Proposition 4.12 is motivated by considering certain homotopy colimits but derived purely algebraically. It is useful either in order to compute identities for presentations of \(G\) induced by presentations of the elements of \(\mathcal H\) or to test whether \(\mathcal H\) is 3-generating. Proposition 4.12 is interpreted in terms of ``pictures'' by \textit{W. A. Bogley} and \textit{S. J. Pride} [Lond. Math. Soc. Lect. Note Ser. 197, 157- 188 (1993)]. Two interesting applications are discussed in Section 3 of the paper. The first is concerned with Tits systems and buildings. It is shown (cf. Theorem 3.3) that the family of all rank \(m\) standard parabolic subgroups of a group with a \(BN\)-pair is \(m\)-generating. This generalizes a well known result due to J. Tits stating that a group with a \(BN\)-pair is the amalgamated product of its rank 2 standard parabolic subgroups. The second application results from specializing \textit{K. S. Brown}'s \(FP_ n\)-criterion [J. Pure Appl. Algebra 44, 45-75 (1987; Zbl 0613.20033)] to the group \(G\) acting on \(N({\mathcal H})\) in case \(\mathcal H\) is finite. Provided that the elements of \(\mathcal H\) have ``good'' finiteness properties \(G\) is of type \(FP_ n\) if \(\mathcal H\) is \(n\)-generating. By this method, it can be proved that \(G_ 5(\mathbb{Z}[1/p])\) is of type \(FP_ 3\), where \(G_ n\) denotes the (solvable algebraic) group of upper triangular \(n\times n\)-matrices with \(g_{11} = g_{nn} = 1\). The reader is referred to the thesis of S. Holz for the details of this proof. (By the way, in the paper under review, last line of \S 3.15, \(H(R)\) has to be replaced by \(T(R)H(R)\).) A different approach due to \textit{H. Abels} and \textit{K. S. Brown} [J. Pure Appl. Algebra 44, 77-83 (1987; Zbl 0617.20020)] yields that, more generally, \(G_ n(\mathbb{Z}[1/p])\) is of type \(FP_{n-2}\) but not \(FP_{n - 1}\) for all integers \(n \geq 2\). Recently, J. Harlander and H. Meinert (preprint 1994) deduced finiteness properties and estimations for virtual cohomological dimensions of graph products by using families of subgroups which are infinitely generating. Further applications of the very natural concept of \(n\)-generatedness are to be expected.