An explicit decomposition of higher Deligne-Lsuztig representations
Ben Liu, Sian Nie · 2025
Read An explicit decomposition of higher Deligne-Lsuztig representations by Ben Liu, Sian Nie (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-12918.txt.
Introduction
Ben Liu and Sian Nie make a decomposition of higher Deligne-Lusztig representations fully explicit in this 2025 paper. These representations are built from geometry and sit deep in the theory of symmetry groups over finite fields. An earlier paper split an elliptic one into basic blocks.
Those blocks rest on a geometric cousin of a classic representation. The note shows the cousin differs from the classic only by a character. Under a mild size condition on the field, that character is one built by other authors. The payoff is an explicit list of the blocks in the split. The team also draw out which standard types appear. Representation theorists gain formulas they can compute with directly.
Catalog record
| CREATOR | Ben Liu, Sian Nie |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 43 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-12918
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
An explicit decomposition of higher Deligne-Lsuztig representations Ben Liu, Sian Nie Abstract . In a previous paper [ 22 ] , the second named author obtains a decomposition of an elliptic higher Deligne-Lusztig representation into irreducible summands, which are built in the same way as Yu types using a geometric analog κ ′ \kappa^{\prime} of the Weil-Heisenberg representation κ \kappa . In this note, we show that κ ′ \kappa^{\prime} and κ \kappa differs by a character χ \chi . Moreover, under a mild condition on the cardinality q q of the residue field (for instance q > 3 q>3 ), we show that χ \chi equals the sign character constructed by Fintzen-Kaletha-Spice [ 15 ] , which gives an explicit irreducible decomposition result on elliptic higher Deligne-Lusztig representations. As an application, we deduce (under the mild condition on q q ) that each unramified Yu type appears in the cohomology of higher Deligne-Lusztig varieties, and each unramified Kaletha’s regular supercuspidal representation is the compact induction of a specified higher Deligne-Lusztig representation up to a sign. 1. Introduction In [ 21 ] , Lusztig introduced higher Deligne-Lusztig representations for parahoric subgroups, which are natural extensions of the celebrated Deligne-Lusztig representations for finite groups of Lie type [ 11 ] . Motivated by the work of Gérardin [ 16 ] , Lusztig conjectured that compact inductions of higher Deligne-Lusztig representations will realize supercuspidal representations of p -adic groups. Recently, there have been intensive studies on Lusztig’s conjecture and further applications to (modular) local Langlands program. Let us mention the work [ 2 ] , [ 3 ] , [ 6 ] , [ 9 ] , [ 10 ] , [ 7 ] , [ 8 ] , [ 22 ] , [ 17 ] , [ 18 ] and references therein. 1.1. Main result The most interesting higher Deligne-Lusztig representations are those attached to attached to elliptic tori, which are referred to as elliptic higher Deligne-Lusztig representations. All the other higher Deligne-Lusztig representations are parabolic inductions of elliptic ones. The main purpose of note is to give an explicit algebraic characterization of elliptic higher Deligne-Lusztig representation. Let p\neq\ell be two different prime numbers. Let k be a non-archimedean local field with a finite residue field of cardinality q and of characteristic p . Let G be a connected k -rational reductive group which splits over a maximal unramifed extension {\breve{k}} of k . Let Z_{G} denote the center of G . Fix a point {\mathbf{x}} in the Bruhat-Tits building \mathcal{B}(G,k) of G over k . Denote by \mathcal{G}=\mathcal{G}_{\mathbf{x}} the attached (connected) parahoric subgroup over the integer ring \mathcal{O}_{k} of k . We recall the following condition on p : \text{ $p\neq 2$ is not a bad prime for $G$ and $p\nmid|\pi_{1}(G_{\rm der})|\cdot|\pi_{1}(\widehat{G}_{\rm der})|$ }. Here G_{\rm der} and \widehat{G}_{\rm der} are the derived subgroups of G and its dual group \widehat{G} respectively. Let T\subseteq G be a k -rational and {\breve{k}} -split maximal elliptic torus, whose apartment over k contains {\mathbf{x}} . \phi:T(k)\overline{\mathbb{Q}}_{\ell}^{\times} of depth \leqslant r\in\mathbb{Z}_{\geqslant 0} . Choose a {\breve{k}} -rational maximal unipotent subgroup U\subseteq G normalized by T . Following [ 21 ] and [ 5 ] , one can associate an \overline{\mathbb{F}}_{q} -variety X=X_{T,U,x,r}, called a higher Deligne-Lusztig variety. Applying Deligne-Lusztig cohomological induction to X , one obtains a virtual smooth Z_{G}(k)\mathcal{G}(\mathcal{O}_{k}) -module R_{T,r}^{G}(\phi)=R_{T,U,{\mathbf{x}},r}^{G}(\phi), which is referred to as a higher Deligne-Lusztig representation. If r=0 , R_{T,r}^{G}(\phi) is by definition a classical Deligne-Lusztig representation. Suppose that \phi has a Howe factorization (\Lambda,\phi_{-1}) in the sense of [ 19 ] , where \Lambda=(G^{i},\phi_{i},r_{i})_{0\leqslant i\leqslant d} is a generic datum and \phi_{-1}:T(k)\overline{\mathbb{Q}}_{\ell}^{\times} a depth 0 character. Under the condition (*), the second named author [ 22 ] proved the following result R_{T,r}^{G}(\phi)={\rm ind}_{\mathcal{K}(\mathcal{O}_{k})}^{\mathcal{G}(\mathcal{O}_{k})}\kappa_{\Lambda}\otimes R_{T,0}^{G^{0}}(\phi_{-1}), where \mathcal{K} is the identity component of a Yu-type \mathcal{O}_{k} -subgroup associated to (\Lambda,{\mathbf{x}}) ; \kappa_{\Lambda} is an irreducible \mathcal{K}(\mathcal{O}_{k}) -module obtained through a cohomological method; and R_{T,0}^{G^{0}}(\phi_{-1}) is the classical Deligne-Lusztig representation. Thus, to give a complete description of R_{T,r}^{G}(\phi) , it remains to determine the representation \kappa_{\Lambda} arising from geometry. To this end, we recall two classical representations of Yu’s subgroup \widetilde{K}\supseteq Z_{G}(k)\mathcal{K}(\mathcal{O}_{k}) associated to (\Lambda,{\mathbf{x}}) . The first one is the Weil-Heisenberg representation \widetilde{\kappa}(\Lambda) introduced by Yu [ 23 ] , which play an essential role in the construction of supercuspidal representations for p -adic groups. The other is the (inflated) quadratic character \widetilde{\epsilon}_{\Lambda} constructed by Fintzen, Kaletha and Spice [ 15 ] , which paly a delicate role in constructing supercuspidal L -packets (see also the work by Chan and Oi [ 7 ] when \phi is toral). Both \widetilde{\kappa}(\Lambda) and \widetilde{\epsilon}_{\Lambda} are constructed in a purely algebraic way and have very explicit descriptions. Put \kappa(\Lambda)=\widetilde{\kappa}(\Lambda)|_{Z_{G}(k)\mathcal{K}(\mathcal{O}_{k})} and \epsilon_{\Lambda}=\widetilde{\epsilon}_{\Lambda}|_{Z_{G}(k)\mathcal{K}(\mathcal{O}_{k})} for simplicity. Now we state the main result of this note. Theorem 1.1 . Assume p\neq 2 . Then there exists a character \psi of Z_{G}(k)\mathcal{K}(\mathcal{O}_{k}) such that (-1)^{d_{\Lambda}}\kappa_{\Lambda}=\kappa(\Lambda)\otimes\epsilon_{\Lambda}\otimes\psi, where d_{\Lambda}\in\{\pm 1\} is a sign associated to \Lambda (see Theorem 4.1 ). If, moreover, q\geqslant c_{\Lambda} then \psi=1 . Here 2\leqslant c_{\Lambda}\leqslant 4 is an integer defined in Proposition 5.3 . Remark 1.2 . It would be interesting to extend Theorem 1.1 to the modular case. This would give an explicit description of modular higher Deligne-Lusztig representations, see [ 18 ] . Combining [ 22 , Theorem 1.6] and Theorem 1.1 we have an explicit irreducible decomposition result. Corollary 1.3 . Assume p satisfies (*) and q\geqslant c_{\Lambda} . Then R_{T,r}^{G}(\phi)=(-1)^{d_{\Lambda}}\sum_{\rho}m_{\rho}{\rm ind}_{\mathcal{K}(\mathcal{O}_{k})}^{\mathcal{G}(\mathcal{O}_{k})}\kappa(\Lambda)\otimes\epsilon_{\Lambda}\otimes\rho, where \rho runs through irreducible factors up to isomorphism of R_{T,0}^{G^{0}}(\phi_{-1}) with multiplicity m_{\rho}\in\mathbb{Z} . Moreover, the \mathcal{G}(\mathcal{O}_{k}) -modules {\rm ind}_{\mathcal{K}(\mathcal{O}_{k})}^{\mathcal{G}(\mathcal{O}_{k})}\kappa(\Lambda)\otimes\epsilon_{\Lambda}\otimes\rho on the right hand side of (*) are irreducible and non-isomorphic with each other. Remark 1.4 . When a preliminary version of this note is finished, we noticed that Chan and Oi [ 8 ] also obtains Corollary 1.3 under certain largeness condition on q . Their approach is quite different from ours, which is based on studies of the regular case and Green functions of higher Deligne-Lusztig representations. Note that our assumption on q may be weaker than theirs in some cases. For instance, if \phi is toral, then c_{\Lambda}=2 by definition and hence there is no restriction on q in Corollary 1.3 . 1.2. Application Now we discuss some applications of the main result. In [ 23 ] , Yu introduced the notion of (tame) cuspidal G -data, and associated with each cuspidal G -datum \Sigma an irreducible supercuspidal representation \pi_{\Sigma}^{\rm Yu} of G(k) . Conversely, thanks to work of Kim [ 20 ] and Fintzen [ 14 ] , all the irreducible supercuspidal representations are exhausted by Yu’s representations \pi_{\Sigma}^{\rm Yu} , provided p does not divide the order of the absolute Weyl group of G . Recall briefly that a cupidal G -datum \Sigma consists of a generic datum \Lambda=(G^{i},\phi_{i},r_{i})_{0\leqslant i\leqslant d} , a point {\mathbf{x}}\in\mathcal{B}(G^{0},k) and a depth 0 irreducible representation \widetilde{\rho} of G^{0}(k)_{[{\mathbf{x}}]} satisfying certain conditions. Here G^{0}(k)_{[{\mathbf{x}}]} denotes the stabilizer in G^{0}(k) of the image [{\mathbf{x}}] of {\mathbf{x}} in \mathcal{B}(G^{0}_{\rm der},k) . Then \pi_{\Sigma}^{\rm Yu} and its twisted company \pi_{\Sigma}^{\rm FKS} are given by \pi_{\Sigma}^{\rm Yu}=\text{c-}{\rm ind}_{\widetilde{K}}^{G(k)}\widetilde{\kappa}(\Lambda)\otimes\widetilde{\rho},\quad\pi_{\Sigma}^{\rm FKS}=\text{c-}{\rm ind}_{\widetilde{K}}^{G(k)}\widetilde{\kappa}(\Lambda)\otimes\widetilde{\epsilon}_{\Lambda}\otimes\widetilde{\rho}. Now we assume the cuspidal datum \Sigma=(\Lambda,{\mathbf{x}},\widetilde{\rho}) above is unramified, that is, the