Relative Serre Duality for Coxeter Groups
Colton Sandvik · 2026
Read Relative Serre Duality for Coxeter Groups by Colton Sandvik (2026) 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-2604-06084.txt.
Introduction
Sandvik's 2026 proof settles a claim by Gorsky, Hogancamp, Mellit, and Nakagane on relative Serre duality for Coxeter groups. It links the left and right adjoints of the parabolic induction functor between homotopy categories of Soergel bimodules for a finite Coxeter group. It says those adjoints relate through the relative full twist. Known cases covered symmetric, crystallographic, and dihedral groups. Sandvik settles it in complete generality using Abe-Bott-Samelson bimodules and the Achar-Riche-Vay mixed derived category.
Catalog record
| CREATOR | Colton Sandvik |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 50030 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.06084
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Relative Serre Duality for Coxeter Groups Colton Sandvik Abstract . It was conjectured by Gorsky, Hogancamp, Mellit, and Nakagane that the left and right adjoints of the parabolic induction functor between homotopy categories of Soergel bimodules associated to a finite Coxeter group are related by the relative full twist. Several cases of this conjecture are known including for symmetric groups, crystallographic Coxeter groups, and dihedral groups. We prove this conjecture in complete generality using the theory of Abe–Bott–Samelson bimodules and the Achar–Riche–Vay mixed derived category. 1. Introduction 1.1. Bott–Samelson Bimodules Let (W,S) be a Coxeter system and \mathfrak{h} be a realization of W over a noetherian domain \mathbbm{k} of finite global dimension. Associated to the pair (\mathfrak{h},W) is the category of Abe–Bott–Samelson bimodules \mathcal{A}^{\oplus}(\mathfrak{h},W) introduced by [ Abe1 ] . This is a (non-full) additive subcategory of graded R=\Sym(\mathfrak{h}^{*}) -bimodules which recovers the usual Bott–Samelson bimodules when \mathfrak{h} is reflection faithful. Abe–Bott–Samelson bimodules can be used to define the mixed derived category , D^{m}(\mathfrak{h},W)\coloneq K^{b}\mathcal{A}^{\oplus}(\mathfrak{h},W). This category admits a monoidal structure \star inherited from the tensor product of graded R -bimodules. Based on similar constructions using parity sheaves, Achar–Riche–Vay [ ARV ] gave an in-depth study of the mixed derived category. In particular, they showed that many constructions and features encountered in the theory of mixed \ell -adic sheaves have purely combinatorial analogues for the mixed derived category. The mixed derived category was first studied in knot theory. For example, Rouquier gave an action of the braid group on the mixed derived category [ Rou ] . When W=S_{n} , this action is the foundation for the construction of link invariants such as Khovanov–Rozansky homology [ Kho , KR ] . Since their introduction, the mixed derived category has also seen considerable study in modular representation theory. For example, it is a crucial component in proving the tilting character formulas for reductive groups developed by Achar–Makisumi–Riche–Williamson [ AMRW ] . 1.2. Relative Serre Duality Let I\subseteq S be a subset of simple reflections, and write W_{I} for the parabolic subgroup generated by I . The realization \mathfrak{h} for W can also be viewed as a realization for W_{I} . There is a monoidal functor \iota:D^{m}(\mathfrak{h},W_{I})\to D^{m}(\mathfrak{h},W) called parabolic induction which is induced by the inclusion W_{I}\subseteq W . It is shown in [ GHMN , Corollary A.8] , that parabolic induction admits a left adjoint \iota^{L} and a right adjoint \iota^{R} . Another perspective on these adjoints comes from the recollement formalism of [ ARV ] which we heavily use in this paper. In [ GHMN ] , it was conjectured that the left and right adjoints are related under \star by the relative full twist – a certain object \FT_{W,I}\in D^{m}(\mathfrak{h},W) defined using Rouquier’s braid group action [ Rou ] . More precisely, they conjectured the following. Conjecture 1.1 ( [ GHMN , Conjecture 1.8] ) . Assume that W is finite. (1) There are natural isomorphisms of functors \iota^{L}(\FT_{W,I}\star-)\cong\iota^{R}\cong\iota^{L}(-\star\FT_{W,I}). (2) \FT_{W,I} naturally commutes with objects of D^{m}(\mathfrak{h},W_{I}) . We will briefly discuss some history on known cases of Conjecture 1.1 . (1) Let W=S_{n} and W_{I}=S_{r}\times(S_{1})^{n-r} for some r\leq n . Consider the realization \mathfrak{h}=\mathbbm{k}^{\oplus n} where W acts on \mathfrak{h} by permuting the coordinates. In this case, Conjecture 1.1 was proved in [ GHMN ] . (2) Q. Ho and P. Li [ HL3 ] made substantial progress on Conjecture 1.1 using their mixed sheaf formalism introduced in [ HL1 ] . In particular, they proved in [ HL3 ] that Conjecture 1.1 (1) holds when W is the Weyl group of a reductive algebraic group and \mathfrak{h} is its associated Cartan realization over \mathbb{C} . (3) When W is a dihedral group, Conjecture 1.1 (1) was proved by C. Li in [ Li ] under some minor non-degeneracy conditions. (4) Conjecture 1.1 (2) is already known provided W does not have a parabolic subgroup of type H_{3} . More precisely, Elias and Hogancamp [ EH ] proved the much stronger statement that the full twist \FT_{W} naturally commutes with objects of D^{m}(\mathfrak{h},W) . The restriction on W is a facet of working with the Elias–Williamson diagrammatic category where the 3-color relation is not known in type H_{3} . In this paper, we prove Conjecture 1.1 in its entirety without any assumptions on W or \mathfrak{h} other than the minor assumptions on \mathfrak{h} needed to have a well-behaved theory of Abe–Bott–Samelson bimodules (cf., [ Abe2 ] ). We refer to Theorem 3.1 and Corollary 3.1 for precise statements. Before continuing, it is worth noting that the author expects that Elias and Hogancamp’s result on the centrality of the full twist [ EH ] should also hold in the setting of this paper. It seems likely that this can be proved using similar methods to loc. cit. using Abe–Bott–Samelson bimodules instead of the diagrammatic category. Conjecture 1.2 . \FT_{W} naturally commutes with objects of D^{m}(\mathfrak{h},W) under \star . The overarching strategy for proving Conjecture 1.1 is heavily influenced by [ HL3 ] . The mixed derived category formalism of [ ARV ] will allow many of their arguments to be adapted even when W is not crystallographic. We briefly mention where our proofs differ. (1) The argument for D^{m}(\mathfrak{h},W) being a rigid monoidal category in [ HL3 ] heavily depends on geometry which cannot be used for general Coxeter groups. Instead, we give an elementary proof of rigidity using the theory of Abe–Bott–Samelson bimodules. As part of our proof, we also give some applications to the structure of the “sheaf-functors” in mixed derived categories. (2) The present paper is written solely using ordinary categories rather than \infty -categories. This decision is largely necessitated by [ ARV ] working with triangulated categories. The author expects that all the constructions in loc. cit. can be adapted to use \infty -categories, but this is beyond the scope of this paper. The only place where \infty -categorical machinery is crucially used in [ HL3 ] is to reduce proving their main result to the case of I=\varnothing . Instead, our argument is more closely based on an earlier pre-print [ HL2 ] which did not make such a reduction. Unfortunately, the argument in the original pre-print has an error which has necessitated new proofs in some places. 1.3. Acknowledgements The author would like to thank Simon Riche for suggesting this project, numerous valuable comments, and carefully reading an earlier draft. The author also thanks Pramod Achar for helpful discussions. The author was partially supported by NSF Grant No. DMS-2231492. 2. Hecke Categories For this section, we will allow (W,S) to be an arbitrary (possibly infinite) Coxeter system. Let (\mathfrak{h},\{\alpha_{s}\}_{s\in S},\{\alpha_{s}^{\vee}\}_{s\in S}) be a realization of W over \mathbbm{k} satisfying [ Abe2 , Assumption 3.5] . Consider the graded algebra R=\Sym(\mathfrak{h}^{*}) where \mathfrak{h}^{*} is placed in degree 2 . We will also consider the localization Q of the ring R with respect to the multiplicative subset generated by \{w(\alpha_{s}):s\in S,w\in W\} . This ring is \mathbb{Z} -graded where \frac{1}{w(\alpha_{s})} is in degree -2 . Let (X,\preceq) be a poset. We will say that a subset Y\subseteq X is closed if for all x,x^{\prime}\in X with x^{\prime}\in Y and x\preceq x^{\prime} we have x\in