Unitriangular shape of decomposition matrices of unipotent blocks (Q2206348)

Let \(\mathbf{G}\) be a connected reductive algebraic group defined over a field \(K\) with a Frobenius endomorphism \(F\) and let \(G:=\mathbf{G}^{F}\) be the group of fixed points. Let \(\ell \) be a prime not equal \(p:=char(K)\). An important object in representation theory is the decomposition matrix which describes the way in which ordinary characters for \(G\) decompose under \(\ell \) -reduction. In his 1990 thesis [Verallgemeinerte Gelfand-Graev Charaktere und Zerlegungszahlen endlicher Gruppen vom Lie-Typ. (Generalized Gelfand-Graev characters and decomposition numbers of finite groups of Lie type). Aachen: RWTH Aachen, Math.-Naturwiss. Fak./FB 1 (1990; Zbl 0729.20005)] \textit{M. Geck} made a conjecture about the form of the decomposition matrix of the unipotent blocks for a finite reductive group when \(\ell \) is large enough. The present paper proves this conjecture in a precise form. \textit{G. Lusztig} [Characters of reductive groups over a finite field. Princeton University Press, Princeton, NJ (1984; Zbl 0556.20033)] defines a partition \(\mathrm{Fam}(W)\) of the irreducible characters of the Weyl group \(W\) of \(\mathbf{G}\). Let Fam\((W)^{F} \) be the fixed points under the action induced from the natural action of \(F\) on \(W\). For each \(\mathcal{F}\in \mathrm{Fam}(W)\) there is a corresponding unipotent conjugacy class \(\mathcal{O}_{\mathcal{F}}\) of \( \mathbf{G}\) related via the Springer correspondence. When \(\mathcal{F}\in \mathrm{Fam}(W)^{F}\) there is a corresponding family \(UCh(\mathcal{F})\) of ordinary characters of \(G\) consisting of all unipotent characters with unipotent support \(\mathcal{O}_{\mathcal{F}}\). We set \(\mathcal{O}_{\mathcal{F} }\preccurlyeq \mathcal{O}_{\mathcal{F}^{\prime }}\) if \(\mathcal{O}_{\mathcal{ F}}\) is contained in the Zariski closure of \(\mathcal{O}_{\mathcal{F} ^{\prime }}\). The main theorem of the present paper is as follows. (Theorem A): Assume that \(\ell \) and \(p\) are good for \(\mathbf{G}\) and that \(\ell \) does not divide the order of \(Z(\mathbf{G)}_{F}\), the largest quotient of the centre on which \(F\) acts trivially. Let \(\mathcal{F}_{1}\leq \mathcal{F}_{2}\leq ...\leq \mathcal{F}_{r}\) be a total ordering of \(\mathrm{Fam}(W)^{F}\) such that \( \mathcal{F}_{i}\leq \mathcal{F}_{j}\) whenever \(\mathcal{O}_{\mathcal{F} _{i}}\preccurlyeq \mathcal{O}_{\mathcal{F}_{j}}\). Then the modular irreducible unipotent representations of \(G\) can be ordered such that the decomposition matrix for unipotent \(\ell \)-blocks of \(G\) is a lower triangular block matrix of the form \[ \left[ \begin{array}{cccc} D_{\mathcal{F}_{1}} & 0 & \cdots & 0 \\ \ast & D_{\mathcal{F}_{2}} & \ddots & \vdots \\ \vdots & \ddots & \ddots & 0 \\ \ast & \cdots & \ast & D_{\mathcal{F}_{r}} \\ \ast & \cdots & \cdots \ddots & \ast \end{array} \right] \] where each block \(D_{\mathcal{F}_{i}}\) is an identity matrix whose rows are indexed by the elements of \(UCh(\mathcal{F}_{i})\). The proof of Theorem A is based on the approach in [\textit{M. Geck} and \textit{G. Hiss}, Prog. Math. 141, 195--249 (1997; Zbl 0867.20014)]; it involves construction of a character which contain a unipotent character of a given family with multiplicity \(1\) and other unipotent characters only from larger families. Kawanaka used a version of generalized Gelfand-Graev representations to construct projective modules conjectured to have this property [\textit{N. Kawanaka}, Proc. Symp. Pure Math. None, 147--163 (1987; Zbl 0654.20046)]. A large part of the present paper is a proof of Kawanaka's conjecture (Theorem B) which is then applied to prove Theorem A. The paper includes a short section which discusses further applications of the two theorems.