On families of degenerate representations of GL n ​ ( F ) \mathrm{GL}_{n}(F)

Johannes Girsch, David Helm · 2025

Read On families of degenerate representations of GL n ​ ( F ) \mathrm{GL}_{n}(F) by Johannes Girsch, David Helm (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-2504-02098.txt.

Introduction

A layer of pure algebra under the representation theory of a matrix group is exposed in a 2025 paper by Johannes Girsch and David Helm. They study smooth representations of the group GL_n over a p-adic field, sorted by a ladder built from degenerate Whittaker models.

Each step of the ladder, when the lower steps are peeled away, turns out to be a module category over a commutative ring. Those rings are endless products of smooth algebras over the ground field. The authors describe the rings plainly using the Zelevinsky classification of representations. The picture closely echoes a classic description of the Bernstein center, the ring that governs the whole category.

Catalog record

CREATORJohannes Girsch, David Helm
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE138 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-02098

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

On families of degenerate representations of GL n ​ ( F ) \mathrm{GL}_{n}(F) Johannes Girsch, David Helm Abstract . We consider the stratification of the category of smooth representations of GL n ​ ( F ) \mathrm{GL}_{n}(F) (for F F a p p -adic field) induced by degenerate Whittaker models. We show that, remarkably, over algebraically closed fields of characteristic zero, the successive quotients in this stratification turn out to be module categories over commutative rings. In fact, they are infinite products of rings of functions on smooth varieties over the ground field. We further obtain explicit descriptions of these in terms of the Zelevinsky classification; these descriptions closely resemble the rings appearing in the Bernstein-Deligne description of the Bernstein center. 1. Introduction Let F be a finite extension of \mathbb{Q}_{p} and {\mathbb{G}} a connected reductive group over F . Let G denote the group of F -points of {\mathbb{G}} . If \pi is an irreducible smooth representation \pi of G , then one may associate to \pi a natural invariant known as its wave front set . This is a subset of the set of nilpotent orbits in \mathrm{Lie}(G) , consisting of the orbits {\mathcal{O}} that are maximal (with respect to orbit closure) among those for which the coefficients c_{\mathcal{O}} of the local character expansion of \pi is nonzero. Another way to associate a set of nilpotent orbits to \pi is via the theory of degenerate Whittaker models. For each nilpotent orbit {\mathcal{O}} one has a space of (compactly supported) degenerate Whittaker functions W_{\mathcal{O}} (we refer the reader to section 4 for the construction of this space in the case {\mathbb{G}}=\mathrm{GL}_{n} ), and one can ask for which {\mathcal{O}} there is a nonzero map W_{\mathcal{O}}\rightarrow\pi . A fundamental result of Moeglin-Waldspurger [ MW ] for p odd, and Varma [ Va ] for p=2 , relates these two questions. More precisely, this result states that the wave front set of \pi coincides with the set of maximal orbits {\mathcal{O}} such that \mathrm{Hom}_{G}(W_{\mathcal{O}},\pi) is nonzero, and moreover that for such orbits the dimension of this space is equal to the coefficient c_{\mathcal{O}} . When {\mathbb{G}}=\mathrm{GL}_{n} these results can be sharpened significantly. In this setting nilpotent orbits are indexed by partitions, and Moeglin-Waldspurger show that the wave front set of any irreducible representation \pi consists of a single orbit. The partition corresponding to this orbit is thus a natural invariant of \pi . Moreover, in this setting the relevant Hom-spaces are all one-dimensional; that is, when \lambda(\pi) is the partition associate to \pi , then \mathrm{Hom}_{\mathrm{GL}_{n}(F)}(W_{{\mathcal{O}}_{\lambda(\pi)}},\pi) has dimension one. The partition \lambda(\pi) has several alternative interpretations. In terms of the local Langlands correspondence, a result of Gourevich-Sahi ( [ GS ] , Theorem B) shows that it is equal to the classification partition of \pi (that is, the partition conjugate to that describing the orbit of the monodromy operator on the Langlands parameter for the Zelevinsky dual of \pi ). On the other hand there is also an interpretation in terms of Bernstein-Zelevinsky derivatives: if \lambda(\pi) is given by the sequence \lambda_{1}\geq\lambda_{2}\geq\dots\geq\lambda_{r} , then for each i , \lambda_{i} is the largest integer such that the iterated derivative (((\pi^{(\lambda_{1})})^{(\lambda_{2})})^{\dots})^{(\lambda_{i})} is nonzero. We thus refer to \lambda as the highest derivative partition of \pi ; these are discussed in detail in section 2 , below. The association \pi\mapsto\lambda(\pi) naturally defines a stratification of the category \operatorname{Rep}(\mathrm{GL}_{n}(F)) of smooth representations of \mathrm{GL}_{n}(F) . Indeed, for each \lambda one can define the full subcategories \operatorname{Rep}(\mathrm{GL}_{n}(F))^{\preceq\lambda} (resp. \operatorname{Rep}(\mathrm{GL}_{n}(F))^{\prec\lambda} ) consisting of smooth representations V for which every irreducible subquotient \pi of V satisfies \lambda(\pi)\preceq\lambda (resp. \lambda(\pi)\prec\lambda ). (Here \preceq is the natural dominance order on partitions, which corresponds to the closure ordering on nilpotent orbits.) From these two subcategories one can also form the Serre quotient: \operatorname{Rep}(\mathrm{GL}_{n}(F))^{\lambda}:=\operatorname{Rep}(\mathrm{GL}_{n}(F))^{\preceq\lambda}/\operatorname{Rep}(\mathrm{GL}_{n}(F))^{\lambda}. One can think of this quotient as a summand of the “associated graded” category of \operatorname{Rep}(\mathrm{GL}_{n}(F)) with respect to this stratification. The goal of this paper is to study the structure of the categories \operatorname{Rep}(\mathrm{GL}_{n}(F))^{\lambda} . To this end we first construct projective objects W^{\prime}_{\lambda} of \operatorname{Rep}(\mathrm{GL}_{n}(F)) , that are closely related to (in fact direct summands of) the spaces of compact degenerate Whittaker functions W_{{\mathcal{O}}_{\lambda}} . Like the latter, the W^{\prime}_{\lambda} have the property that \mathrm{Hom}(W^{\prime}_{\lambda},\pi) is zero unless \lambda(\pi)\preceq\lambda , and is one-dimensional if \lambda(\pi)=\lambda . Unlike the W_{{\mathcal{O}}_{\lambda}} , the representations W^{\prime}_{\lambda} have good finiteness properties (their projections to individual Bernstein blocks are finitely generated), and this allows us to easily establish their projectivity. (It is possible that the W_{{\mathcal{O}}_{\lambda}} are also projective but this is not known. If so, one could argue with the W_{{\mathcal{O}}_{\lambda}} in place of the W^{\prime}_{\lambda} in what follows.) From the W^{\prime}_{\lambda} we construct projective objects \mathcal{P}_{\lambda} of \operatorname{Rep}(\mathrm{GL}_{n}(F))^{\lambda} . Indeed, the inclusion of this category in \operatorname{Rep}(\mathrm{GL}_{n}(F)) is exact, and thus its left adjoint V\mapsto V^{\preceq\lambda} that takes projectives to projectives. We set \mathcal{P}_{\lambda}=(W^{\prime}_{\lambda})^{\preceq\lambda} . This is a projective object of \operatorname{Rep}(\mathrm{GL}_{n}(F))^{\preceq\lambda} , and the functor \mathrm{Hom}(\mathcal{P}_{\lambda},-) is an equivalence between the Serre quotient \operatorname{Rep}(\mathrm{GL}_{n}(F))^{\lambda} and the category of modules over the endomorphism ring E_{\lambda} of \mathcal{P}_{\lambda} . In light of this, the remainder of the paper is devoted to a study of the rings E_{\lambda} . Remarkably, in many settings these rings turn out to be commutative! Our main result, Theorem 5.4 , covers the case where the coefficient field is algebraically closed of characteristic zero. In this context the rings in question turn out to be (infinite products of) rings of functions on smooth varieties over the ground field, and can be described in a very concrete and explicit fashion, that closely resembles the Bernstein-Deligne description of the Bernstein center. Our constructions work well in a modular or mixed-characteristic setting as well, and we conjecture that the rings E_{\lambda} remain commutative in these contexts, although this seems much harder to prove. The proof of this result in characteristic zero occupies sections 6 through 11 of the paper, together with the appendix. The most difficult part of the argument is the commutativity; in section 6 we work locally, and reduce this statement to a computation of I^{\preceq\lambda} , where I is a certain parabolic induction. This computation (Theorem 6.3 ) is the technical heart of the paper, and proceeds via Bernstein-Zelevinsky theory; since the approach to this computation is largely disjoint from the rest of the paper we postpone the proof of this theorem to an appendix (section 12 .) Once commutativity is established we build a theory of families of objects of \operatorname{Rep}(\mathrm{GL}_{n}(F))^{\preceq\lambda} in section 7 , in a manner that exactly follows the development of the theory of co-Whittaker families in earlier work of the second author [ He2 ] . (Indeed, the theory of co-Whittaker families manifests in our more general setting, precisely as the case where \lambda is the one-element partition.) The theory of families in particular gives us an interpretation of tangent vectors to \operatorname{Spec}E_{\lambda} as families over a ring of dual numbers, which we use to give a description of this tangent space in terms of self-extensions of a certain “standard module” that arises as the parabolic induction of a tensor product of representations associated to Zelevinsky segments. We compute these self-extensions

Read the complete text