On the Steinberg module of Chevalley groups. (Q1772062): Difference between revisions

Let \(G\) be a simple Chevalley group with root system \(\Phi\). If \(K\) is a field, then the Steinberg module \(S_K\) for \(G(K)\) is the top homology of the Tits building \(T_K\). It contains important cohomological information. By the Solomon-Tits theorem \(S\) is a cyclic module for the group ring \(K[G(K)]\). Now suppose \(\mathcal D\) is a Euclidean domain with field of fractions \(K\). It is then of interest to know that \(S_K\) is also cyclic as a \(K[G(\mathcal D)]\)-module. This has been shown by Ash and Rudolph for \(G=PSL_n\) and by Gunnells for the symplectic case. See \textit{A. Ash} and \textit{L. Rudolph} [Invent. Math. 55, 241-250 (1979; Zbl 0426.10023)] and \textit{P. E. Gunnells} [Duke Math. J. 102, No. 2, 329-350 (2000; Zbl 0988.11023)]. The author now gives new proofs, valid for all \(A\), \(B\), \(C\), \(D\) types and also for types \(E_6\), \(E_7\). What these types have in common is that there is a maximal parabolic subgroup with Abelian unipotent radical. First one needs a completely new presentation of \(S_K\) as a \(K[G(K)]\)-module. This presentation involves only \(SL_2\) relations and works for any \(\Phi\). To get to the main result requires extensive `word processing'. A representative in \(K[G(K)]\) of an element of \(S_K\) is modified repeatedly until it lies in \(K[G(\mathcal D)]\). Although this word processing is guided by the Euclidean norm amongst other things, the method probably extends to cases where \(\mathcal D\) is only a UFD.