Cohomology of the hyperplane complement of a quaternionic reflection group

Stephen Griffeth, David Guevara · 2025

Read Cohomology of the hyperplane complement of a quaternionic reflection group by Stephen Griffeth, David Guevara (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-2510-18607.txt.

Introduction

Stephen Griffeth and David Guevara compute a quantity in algebra connected to quaternionic reflection groups, special symmetry groups whose elements act like reflections but use quaternion numbers. Each such group leaves a set of hyperplanes fixed, and mathematicians study the space formed by removing those hyperplanes. The cohomology of that space, a graded algebraic summary of its shape, is what the authors determine.

Their main result gives the graded rank of that cohomology and shows it splits into irreducible pieces whose coefficients are all positive whole numbers. For an irreducible group, nearly every piece has the simplest, linear form. At most one piece is quadratic, and that exception happens for exactly three of the exceptional groups, a tidy pattern that the computation makes visible.

Catalog record

CREATORStephen Griffeth, David Guevara
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE114 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2510-18607

Browse the full library · About the Tower of Babel

Full text

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

Cohomology of the hyperplane complement of a quaternionic reflection group Thanks: We thank Gwyn Bellamy, Gerhard Röhrle, Johannes Schmitt, Don Taylor, and Ulrich Thiel for comments on a preliminary version of this article. Stephen Griffeth, David Guevara Abstract . We compute the graded rank of the cohomology of the hyperplane complement associated with a quaternionic reflection group, and observe that it factors into irreducible factors with positive integer coefficients. For an irreducible group, these irreducible factors are all linear except at most one irreducible quadratic factor, which occurs for precisely three of the exceptional groups. 1. Introduction 1.1. Reflection groups Each compact Lie group K with maximal torus T produces a Weyl group W , which is a finite linear group W=N_{K}(T)/T acting on the dual of the Lie algebra of T and stabilizing the lattice of characters of T . This representation of the Weyl group is in particular defined over \mathbf{Q} , and one shows that W is generated by the reflections it contains. In this way compact Lie groups are related to reflection groups defined over \mathbf{Q} , with many of the properties of K encoded by the corresponding Weyl group W . Extending scalars from \mathbf{Q} to \mathbf{R} and considering those finite \mathbf{R} -linear groups generated by reflections produces the class of finite Coxeter groups, and further enlarging the class of groups under study to allow a group W generated by reflections acting on a \mathbf{C} -vector space V results in complex reflection groups, characterized as those finite linear groups whose corresponding quotient space V/W is smooth. The sequence \mathbf{Q}\subseteq\mathbf{R}\subseteq\mathbf{C} of fields may be extended to \mathbf{Q}\subseteq\mathbf{R}\subseteq\mathbf{C}\subseteq\mathbf{H} by considering the division ring \mathbf{H} of quaternions, and while the resulting class of reflection groups is far less well-studied than the class of complex reflection groups, it has recently developed (see [ CaGr ] ) that the natural generality for the study of diagonal coinvariant rings of complex reflection groups seems to be in the context of these quaternionic reflection groups (also known as symplectic reflection groups ). The purpose of this paper is to explore another aspect of the theory of complex reflection groups that, quite mysteriously, also extends to quaternionic reflection groups. Specifically, we compute the Poincaré polynomials of the hyperplane complements for quaternionic reflection groups and observe that they exhibit some extremely striking behavior similar to the case of complex reflection groups (as first proved in [ OrSo ] ): they factor into irreducible (in \mathbf{Z}[t] ) factors of degree at most two with positive integer coefficients. Our calculation relies on Cohen’s classification [ Coh ] of quaternionic reflection groups (see also [ Tay ] and [ Wal ] ). Specifically, we show that for the infinite family W_{n}(\Gamma,\Delta) (see 3.4 for the definition), the hyperplane complement is an iterated fiber bundle, while for the exceptional groups we compute the structure of the lattice of parabolic subgroups explicitly enough to determine the Poincaré polynomial. This calculation should also be of use in questions having to do with the structure of the class of parabolic subgroups, for instance in the study of symplectic reflection algebras. 1.2. Numerical invariants from polynomial invariants and the coinvariant ring Given a complex reflection group W with reflection representation \mathfrak{h} , the ring \mathbf{C}[\mathfrak{h}]^{W} of W -invariant polynomial functions on \mathfrak{h} is generated by n=\mathrm{dim}(\mathfrak{h}) algebraically independent homogeneous polynomials of degrees d_{1}\leq d_{2}\leq\cdots\leq d_{n} . This sequence d_{1}\leq d_{2}\leq\cdots\leq d_{n} is independent of the choice of generators, and the integers d_{1},\dots,d_{n} are called the degrees of W . The ring \mathbf{C}[\mathfrak{h}]^{W} is the ring of functions on the quotient scheme \mathfrak{h}/W , and the coinvariant ring of W is the ring \mathbf{C}[\pi^{-1}(0)] of functions on the scheme-theoretic fiber \pi^{-1}(0) of the quotient map \pi:\mathfrak{h}\to\mathfrak{h}/W . The structure of the graded W -module \mathbf{C}[\pi^{-1}(0)] provides another family of numerical invariants of W : upon forgetting the grading it is isomorphic to the regular representation \mathbf{C}W and so for each irreducible \mathbf{C}W -module E there is a well-determined sequence m_{1}\leq m_{2}\leq\cdots\leq m_{p} of p=\mathrm{dim}(E) non-negative integers known as the exponents of E with E occurring in degrees m_{1},m_{2},\dots,m_{p} in \mathbf{C}[\pi^{-1}(0)] . In particular, the representation \mathfrak{h} appears in degrees d_{1}-1,d_{2}-1,\dots,d_{n}-1 in \mathbf{C}[\pi^{-1}(0)] , and the numbers e_{1}=d_{1}-1,\dots,e_{n}=d_{n}-1 are known simply as the exponents of W . The co-exponents of W are the integers 1=f_{1}\leq f_{2}\leq\cdots\leq f_{n} so that \mathfrak{h}^{*} appears in degrees f_{1},f_{2},\dots,f_{n} in \mathbf{C}[\pi^{-1}(0)] . If \mathfrak{h}\cong\mathfrak{h}^{*} , or in other words if W is a real reflection group, then the exponents and co-exponents coincide. 1.3. Exponents and co-exponents via topology of the hyperplane complement and the combinatorics of the intersection lattice These numerical invariants arise in another context. Namely, given a complex reflection group W we define two polynomials c_{W}(t)=\sum_{w\in W}t^{\mathrm{codim}(\mathrm{fix}(w))} and p_{W}(t)=\sum_{d=0}^{\infty}\mathrm{dim}(H^{d}(\mathcal{M}_{W}))t^{d}, where \mathcal{M}_{W} is the hyperplane complement for the group W , obtained by removing all its reflecting hyperplanes from the space V on which it acts. These polynomials factor in terms of the exponents and coexponents of W : c_{W}(t)=\prod_{i=1}^{n}(1+e_{i}t)\quad\text{and}\quad p_{W}(t)=\prod_{i=1}^{n}(1+f_{i}t). Thus if W is a real reflection group, we have p_{W}(t)=c_{W}(t) . Both of these definitions still make sense for a quaternionic reflection group, and the purpose of this paper is to point out that while the algebraic definitions of the exponents and coexponents no longer work, the polynomials c_{W}(t) and (especially!) p_{W}(t) exhibits fascinating and mysterious behavior very similar to the case of a complex reflection group. 1.4. The codimension and Poincaré polynomials for a quaternionic reflection group Given a quaternionic reflection group W with reflection representation V , we search for numerical invariants. The paper [ CaGr ] introduced three variants of the Coxeter number h of an irreducible real reflection group, g=2N/n, h=(N+N^{*})/n , and k=2N^{*}/n , where N is the number of reflections, N^{*} is the number of reflecting hyperplanes and n is the rank. Here we compute the ranks of the cohomology groups with coefficients in \mathbf{Z} of the hyperplane complement V^{\circ} associated with W , with the hope of discovering an analog of the coexponents. We encode this information in the Poincaré polynomial p_{W}(t) , which is the generating function for these ranks: p_{W}(t)=\sum_{d=0}^{\infty}\mathrm{rk}\left(H^{3d}(V^{\circ},\mathbf{Z})\right)t^{d}. We note that the cohomology groups are always free \mathbf{Z} -modules, and are zero in degrees not divisible by 3 (essentially because the 3 -sphere S^{3} has these properties), so p_{W}(t) contains all the information about the cohomology groups up to isomorphism. We also define the codimension generating function c_{W}(t) just as in the complex case (but taking \mathbf{H} -codimension). Our main theorem is the computation of these polynomials for all quaternionic reflection groups. Since they are multiplicative, p_{W_{1}\times W_{2}}=p_{W_{1}}p_{W_{2}} and c_{W_{1}\times W_{2}}=c_{W_{1}}c_{W_{2}} , it suffices to compute them for irreducible groups, which we do using Cohen’s [ Coh ] classification. We may assume our group W is not complex. Theorem 1.1 . (a) For a quaternionic group W of rank 2 containing N reflections and N^{*} reflecting hyperplanes, p_{W}(t)=(1+t)(1+(N^{*}-1)t)\quad\text{and}\quad c_{W}(t)=1+Nt+(|W|-N-1)t^{2}. (b) For a positive integer n\geq 3 , a finite non-abelian subgroup \Gamma of \mathbf{H}^{\times} with order |\Gamma|=m , and a normal subgroup \Delta of \Gamma such that \Gamma/\Delta is abelian, the Poincaré polynomial p_{W}(t) of W_{n}(\Gamma,\Delta) is p_{W}(t)=\left(1+t\right)\left(1+(1+m)t\right)\left(1+(1+2m)t\right)\cdots\left(1+(1+(n-1)m)t\right), and the polynomial c_{W}(t) is c_{W}(t)=\left(1+(m-1)t\right)\left(1+(2m-1)t\right)\cdots\left(1+((n-1)m-1)t\right)\left(1+(np-1)t\right), where

Read the complete text