Show Summary Details
More options …

# Journal of Group Theory

Editor-in-Chief: Parker, Christopher W.

Managing Editor: Wilson, John S. / Khukhro, Evgenii I. / Kramer, Linus

6 Issues per year

IMPACT FACTOR 2017: 0.581

CiteScore 2017: 0.53

SCImago Journal Rank (SJR) 2017: 0.778
Source Normalized Impact per Paper (SNIP) 2017: 0.893

Mathematical Citation Quotient (MCQ) 2017: 0.45

Online
ISSN
1435-4446
See all formats and pricing
More options …
Volume 21, Issue 4

# The Varchenko determinant of a Coxeter arrangement

Götz Pfeiffer
• School of Mathematics, Statistics and Applied Mathematics, National University of Ireland Galway, University Road, Galway, Ireland
• Email
• Other articles by this author:
/ Hery Randriamaro
Published Online: 2018-04-10 | DOI: https://doi.org/10.1515/jgth-2018-0009

## Abstract

The Varchenko determinant is the determinant of a matrix defined from an arrangement of hyperplanes. Varchenko proved that this determinant has a beautiful factorization. It is, however, not possible to use this factorization to compute a Varchenko determinant from a certain level of complexity. Precisely at this point, we provide an explicit formula for this determinant for the hyperplane arrangements associated to the finite Coxeter groups. The intersections of hyperplanes with the chambers of such arrangements have nice properties which play a central role for the calculation of their associated determinants.

## 1 Introduction

Let $x=\left({x}_{1},\mathrm{\dots },{x}_{n}\right)$ be a variable of the Euclidean space ${ℝ}^{n}$, and ${a}_{1},\mathrm{\dots },{a}_{n},b$ real coefficients such that $\left({a}_{1},\mathrm{\dots },{a}_{n}\right)\ne \left(0,\mathrm{\dots },0\right)$. A hyperplane H of ${ℝ}^{n}$ is an $\left(n-1\right)$-dimensional affine subspace $H:=\left\{x\in {ℝ}^{n}\mid {a}_{1}{x}_{1}+\mathrm{\dots }+{a}_{n}{x}_{n}=b\right\}$. An arrangement of hyperplanes in ${ℝ}^{n}$ is a finite set of hyperplanes. For example, the best known hyperplane arrangement is certainly

${\mathcal{𝒜}}_{{A}_{n-1}}={\left\{\left\{x\in {ℝ}^{n}\mid {x}_{i}-{x}_{j}=0\right\}\right\}}_{1\le i

associated to the Coxeter group ${A}_{n-1}$.

A chamber of a hyperplane arrangement $\mathcal{𝒜}$ is a connected component of the complement ${ℝ}^{n}\setminus {\bigcup }_{H\in \mathcal{𝒜}}H$. Denote the set of all chambers of $\mathcal{𝒜}$ by $ℭ\left(\mathcal{𝒜}\right)$.

Assign a variable ${a}_{H}$ to each hyperplane H of an arrangement $\mathcal{𝒜}$. Let

${R}_{\mathcal{𝒜}}=ℤ\left[{a}_{H}\mid H\in \mathcal{𝒜}\right]$

be the ring of polynomials in variables ${a}_{H}$. The module of ${R}_{\mathcal{𝒜}}$-linear combinations of chambers of the hyperplane arrangement $\mathcal{𝒜}$ is

${M}_{\mathcal{𝒜}}:=\left\{\sum _{C\in ℭ\left(\mathcal{𝒜}\right)}{x}_{C}C\mid {x}_{C}\in {R}_{\mathcal{𝒜}}\right\}.$

Let $\mathcal{ℋ}\left(C,D\right)$ be the set of hyperplanes separating the chambers C and D in $ℭ\left(\mathcal{𝒜}\right)$. The ${R}_{\mathcal{𝒜}}$-bilinear symmetric form $𝖡:{M}_{\mathcal{𝒜}}×{M}_{\mathcal{𝒜}}\to {R}_{\mathcal{𝒜}}$ on the hyperplane arrangement $\mathcal{𝒜}$ defined by Varchenko [12] is

The matrix ${\left(𝖡\left(C,D\right)\right)}_{C,D\in ℭ\left(\mathcal{𝒜}\right)}$ associated to the bilinear symmetric form $𝖡$ is called the Varchenko matrix of the hyperplane arrangement $\mathcal{𝒜}$. In terms of Markov chains, it is the matrix of random walks on $ℭ\left(\mathcal{𝒜}\right)$ whose walk probability from the chamber C to the chamber D is equal to $𝖡\left(C,D\right)$. The Varchenko determinant of the hyperplane arrangement $\mathcal{𝒜}$ is the determinant

$det\mathcal{𝒜}:=det{\left(𝖡\left(C,D\right)\right)}_{C,D\in ℭ\left(\mathcal{𝒜}\right)}.$

One of the first appearances of this bilinear form was in the work of Schechtman and Varchenko [11, Section 1], in the implicit form of a symmetric bilinear form on a Verma module over a $ℂ$-algebra. It appeared more explicitly later, one year after that publication, albeit as a very special case, when Zagier studied a Hilbert space $ℍ$ together with a nonzero distinguished vector $|0〉$, and a collection of operators ${a}_{k}:ℍ\to ℍ$ satisfying the commutation relations

$a\left(l\right){a}^{†}\left(k\right)-q{a}^{†}\left(k\right)a\left(l\right)={\delta }_{k,l},$

and the relation $a\left(k\right)|0〉=0$. To demonstrate the realizability of its model, he defined an inner product space $\left(ℍ\left(q\right),〈\cdot ,\cdot 〉\right)$ with basis B consisting of n-particle states ${a}^{†}\left({k}_{1}\right)\mathrm{\dots }{a}^{†}\left({k}_{n}\right)|0〉$, and proved that [13, Theorems 1, 2]

$det{\left(〈u,v〉\right)}_{u,v\in B}=\prod _{k=1}^{n-1}{\left(1-{q}^{{k}^{2}+k}\right)}^{\frac{n!\left(n-k\right)}{{k}^{2}+k}}.$

It is the Varchenko determinant of ${\mathcal{𝒜}}_{{A}_{n-1}}$ with all hyperplanes weighted by q. Using the diagonal solutions of the Yang–Baxter equation, Duchamp et al. computed [6, Section 6.4.2]

$det{\mathcal{𝒜}}_{{A}_{n-1}}=\prod _{\begin{array}{c}I\in {2}^{\left[n\right]}\\ |I|\ge 2\end{array}}{\left(1-\prod _{\left\{i,j\right\}\in \left(\genfrac{}{}{0pt}{}{I}{2}\right)}{a}_{{H}_{i,j}}^{2}\right)}^{\left(|I|-2\right)!\left(n-|I|+1\right)!},$

each hyperplane $\left\{x\in {ℝ}^{n}\mid {x}_{i}-{x}_{j}=0\right\}$ having its own weight ${a}_{{H}_{i,j}}$ this time.

An edge of a hyperplane arrangement $\mathcal{𝒜}$ is a nonempty intersection of some of its hyperplanes. Denote the set of all edges of $\mathcal{𝒜}$ by $L\left(\mathcal{𝒜}\right)$. The weight $𝖺\left(E\right)$ of an edge E is

$𝖺\left(E\right):=\prod _{\begin{array}{c}H\in \mathcal{𝒜}\\ E\subseteq H\end{array}}{a}_{H}.$

The multiplicity $l\left(E\right)$ of an edge E is a positive integer computed as follows [5, Section 2]: First choose a hyperplane H of $\mathcal{𝒜}$ containing E. Then $l\left(E\right)$ is half the number of chambers C of $ℭ\left(\mathcal{𝒜}\right)$ which have the property that E is the minimal intersection containing $\overline{C}\cap H$.

Due to Varchenko [12, Theorem 1.1], the formula for the determinant of a hyperplane arrangement $\mathcal{𝒜}$ is

$det\mathcal{𝒜}=\prod _{E\in L\left(\mathcal{𝒜}\right)}{\left(1-𝖺{\left(E\right)}^{2}\right)}^{l\left(E\right)}.$

From this formula, we see that we can get a more explicit or computable value of the determinant of an arrangement if we have computable forms of the $𝖺\left(E\right)$’s and the $l\left(E\right)$’s. In this article, we prove that this is the case for the arrangements associated to finite Coxeter groups. Recall that a reflection in ${ℝ}^{n}$ is a linear map sending a nonzero vector α to its negative while fixing pointwise the hyperplane H orthogonal to α. The finite reflection groups are also called finite Coxeter groups, since they have been classified by Coxeter [4]. Coxeter groups find applications in practically all areas of mathematics. They are particularly studied in great depth in algebra [7], combinatorics [2], and geometry [1]. And they are the foundation ingredients of mathematical theories like the descent algebras, the Hecke algebras, or the Kazhdan–Lusztig polynomials. A finite Coxeter group W has the presentation

We begin with some definitions relating to the finite Coxeter group W.

The elements of the set $S:=\left\{{s}_{1},{s}_{2},\mathrm{\dots },{s}_{n}\right\}$ are called the simple reflections of W. The set of all reflections of W is denoted by $T:=\left\{{s}_{i}^{x}\mid {s}_{i}\in S,x\in W\right\}$.

As usually, for a subset J of S, ${W}_{J}$ is the parabolic subgroup $〈J〉$ of W, ${T}_{J}$ the set of reflections $T\cap {W}_{J}$ of ${W}_{J}$, ${X}_{J}$ the set of coset representatives of minimal length of ${W}_{J}$, and $\left[J\right]$ its Coxeter class that is the set of W-conjugates of J which happen to be subsets of S.

One says that J is irreducible if the relating parabolic subgroup ${W}_{J}$ is irreducible.

For a subset K of J, we write ${N}_{{W}_{J}}\left({W}_{K}\right)$ for the normalizer of ${W}_{K}$ in ${W}_{J}$, and $X\left(J,K\right)$ for the set of double coset representatives

$X\left(J,K\right):=\left\{w\in {W}_{J}\cap {X}_{K}\cap {X}_{K}^{-1}\mid {K}^{w}=K\right\}.$

The normalizers of the parabolic subgroups were determined by Howlett [8]. He also proved that ${N}_{{W}_{J}}\left({W}_{K}\right)={W}_{K}\cdot X\left(J,K\right)$, see [8, Corollary 3].

Let ${s}_{{i}_{1}}\mathrm{\dots }{s}_{{i}_{l}}$ be a reduced expression of an element x of W. The support of x is

$J\left(x\right):=\left\{{s}_{{i}_{1}},\mathrm{\dots },{s}_{{i}_{l}}\right\}$

which is independent of the choice of the reduced expression [1, Proposition 2.16]. One says that x has full support if $J\left(x\right)=S$.

We finish with the set

Now, we come to the hyperplane arrangement associated to a finite Coxeter group called Coxeter arrangement. Let ${H}_{t}$ be the hyperplane $\mathrm{ker}\left(t-1\right)$ of ${ℝ}^{n}$ whose points are fixed by each element t of T. The hyperplane arrangement associated to the finite Coxeter group W is ${\mathcal{𝒜}}_{W}:={\left\{{H}_{t}\right\}}_{t\in T}$. In this case, explicit formulas for $𝖺\left(E\right)$ and $l\left(E\right)$ can be given, and it is the aim of this article. For a subset U of T, we write ${E}_{U}$ for the edge

${E}_{U}:=\bigcap _{u\in U}{H}_{u}.$

#### Theorem 1.1.

Let E be an edge of ${\mathcal{A}}_{W}$. Then $l\mathit{}\mathrm{\left(}E\mathrm{\right)}\mathrm{\ne }\mathrm{0}$ if and only if there exist an irreducible subset J of S and an element w of W such that $E\mathrm{=}{E}_{{T}_{J}^{w}}$. In this case, we clearly have $\mathrm{a}\mathit{}\mathrm{\left(}{E}_{{T}_{J}^{w}}\mathrm{\right)}\mathrm{=}{\mathrm{\prod }}_{u\mathrm{\in }{T}_{J}}{a}_{{H}_{{u}^{w}}}$. Moreover, let ${t}_{J}$ be a reflection with support J, ${s}_{J}$ a simple reflection and v an element of W such that a reduced expression of ${t}_{J}$ is ${s}_{J}^{v}$. Then

$l\left({E}_{{T}_{J}^{w}}\right)=|⌊{t}_{J}⌉|\cdot |\left[J\right]|\cdot |X\left(S,J\right)|\cdot |X\left(J,\left\{{s}_{J}\right\}\right)|.$

Let ${Y}_{J}$ be the set of the cosets of ${N}_{W}\left({W}_{J}\right)$, and $\mathcal{ℐ}\left(S\right)$ the set of the Coxeter classes $\left[J\right]$ of W such that J is irreducible. We deduce that the Varchenko determinant of the arrangement associated to a finite Coxeter group W is

$\prod _{\left[J\right]\in \mathcal{ℐ}\left(S\right)}\prod _{w\in {Y}_{J}}{\left(1-\prod _{u\in {T}_{J}}{a}_{{H}_{{u}^{w}}}^{2}\right)}^{|⌊{t}_{J}⌉|\cdot |\left[J\right]|\cdot |X\left(S,J\right)|\cdot |X\left(J,\left\{{s}_{J}\right\}\right)|}.$

We compute the Varchenko determinants associated to the irreducible Coxeter groups in Section 6. We deduce the Varchenko determinant associated to a finite Coxeter group from the following remark: If $W={W}_{1}{W}_{2}$, where ${W}_{1}$ and ${W}_{2}$ are two irreducible Coxeter groups, then

$det{\mathcal{𝒜}}_{W}={\left(det{\mathcal{𝒜}}_{{W}_{1}}\right)}^{|{W}_{2}|}{\left(det{\mathcal{𝒜}}_{{W}_{2}}\right)}^{|{W}_{1}|}.$

Recall that there is a one-to-one correspondence between the elements of W and the chambers of ${\mathcal{𝒜}}_{W}$ such that: if the chamber C corresponds to the neutral element e and ${C}_{x}$ to another element x, then ${C}_{x}=Cx$ with $x={t}_{1}\mathrm{\dots }{t}_{r}$, the ${H}_{{t}_{i}}$’s being the hyperplanes one goes through from C to ${C}_{x}$, see [1, Theorem 1.69].

Let $〈{\overline{C}}_{x}\cap {H}_{t}〉$ be the subspace generated by the closed face ${\overline{C}}_{x}\cap {H}_{t}$ of the chamber ${C}_{x}$. Determining the multiplicity of the edge E contained in the hyperplane ${H}_{t}$ consists of counting the half of the chambers ${C}_{x}$ which have the property that E is the minimal edge containing $〈{\overline{C}}_{x}\cap {H}_{t}〉$. For the proof of Theorem 1.1, we need to introduce the three propositions that we prove in the next three sections.

#### Proposition 1.2.

Let E be an edge of $\mathcal{A}$. Then $l\mathit{}\mathrm{\left(}E\mathrm{\right)}\mathrm{\ne }\mathrm{0}$ if and only if there exist an irreducible subset J of S and an element w of W such that $E\mathrm{=}{E}_{{T}_{J}^{w}}$.

Let us introduce the set $L\left(E,t\right):=\left\{x\in W\mid 〈{\overline{C}}_{x}\cap {H}_{t}〉=E\right\}$.

#### Proposition 1.3.

Let J be an irreducible subset of S, t a reflection with support J, s a simple reflection, and v an element of W such that a reduced expression of t is ${s}^{v}$. For a conjugate K of J, let ${c}_{K\mathrm{,}J}$ be an element of W such that ${K}^{{c}_{K\mathrm{,}J}}\mathrm{=}J$, and for a conjugate u of t with support J, let ${c}_{u\mathrm{,}t}$ be an element of W such that ${u}^{{c}_{u\mathrm{,}t}}\mathrm{=}t$. Then,

$L\left({E}_{{T}_{J}},t\right)=\bigsqcup _{K\in \left[J\right]}{c}_{K,J}X\left(S,J\right)\left(\bigsqcup _{u\in ⌊t⌉}{c}_{u,t}{N}_{{W}_{J}}{\left({W}_{\left\{s\right\}}\right)}^{v}\right).$

#### Proposition 1.4.

Let $u\mathrm{,}v\mathrm{\in }T$ and let E be an edge contained in both ${H}_{u}$ and ${H}_{v}$. Then

$|L\left(E,u\right)|=|L\left(E,v\right)|.$

## 2 The Coxeter complex

Not all edges are relevant, or in other words, there are some edges whose multiplicities are null. We develop the condition for an edge E to be relevant which means $l\left(E\right)\ne 0$.

#### Lemma 2.1.

A finite Coxeter group W is irreducible if and only if W has a reflection of full support.

#### Proof.

If W is irreducible, then W has a highest root [9, Sections 2.10, 2.13], and the reflection corresponding to the highest root has full support.

Suppose that W is the product of nontrivial Coxeter groups ${W}_{1}$ and ${W}_{2}$, and let t be a reflection in W. Without loss of generality, we can suppose that t is a conjugate of a simple reflection s of ${W}_{1}$, hence t lies in ${W}_{1}$ and can not have full support. ∎

We continue our investigation by using the Coxeter complex. Recall that the Coxeter complex $\mathcal{𝒞}$ of W is the set $\left\{{W}_{J}x\mid J\subseteq S,x\in W\right\}$ of faces. The Coxeter complex is a combinatorial setup which permits to study the geometrical structure of ${\mathcal{𝒜}}_{W}$. Indeed, ${ℝ}^{n}$ is partitioned by $\mathcal{𝒞}$, see [9, Section 1.15].

The chamber ${C}_{x}$ is identified with the singleton $\left\{x\right\}$. The coset ${W}_{J}w$ is a face of $\left\{x\right\}$ if and only if ${W}_{J}w={W}_{J}x$. Then the closure of ${C}_{x}$ is

${\overline{C}}_{x}:=\left\{{W}_{J}x\mid J\subseteq S\right\}.$(2.1)

More generally, ${W}_{K}w$ is a face of ${W}_{J}x$ if and only if ${W}_{J}x\subseteq {W}_{K}w$ and ${W}_{K}w={W}_{K}x$. Then the closure of a face ${W}_{J}x$ of ${C}_{x}$ is

$\overline{{W}_{J}x}:=\left\{{W}_{K}x\mid J\subseteq K\subseteq S\right\}.$(2.2)

Hyperplanes, and more generally edges, can be described as collections of the faces they consist of. We extend the definition of $J\left(x\right)$ to a subset X of W in the following way:

$J\left(X\right):=\bigcup _{x\in X}J\left(x\right).$

#### Lemma 2.2.

Let $t\mathrm{\in }T$. Then ${H}_{t}\mathrm{=}\mathrm{\left\{}{W}_{J}\mathit{}w\mathrm{\mid }J\mathrm{\subseteq }S\mathrm{,}w\mathrm{\in }W\mathrm{,}J\mathit{}\mathrm{\left(}{t}^{{w}^{\mathrm{-}\mathrm{1}}}\mathrm{\right)}\mathrm{\subseteq }J\mathrm{\right\}}$.

#### Proof.

As W acts by right multiplication, ${W}_{J}w\in {H}_{t}$ if and only if ${W}_{J}wt={W}_{J}w$, i.e. ${W}_{J}^{w}t={W}_{J}^{w}$, i.e. $t\in {W}_{J}^{w}$, i.e. ${t}^{{w}^{-1}}\in {W}_{J}$, i.e. $J\left({t}^{{w}^{-1}}\right)\subseteq J$. ∎

#### Lemma 2.3.

Let $x\mathrm{\in }W$ and $t\mathrm{\in }T$. Then, ${\overline{C}}_{x}\mathrm{\cap }{H}_{t}\mathrm{=}\overline{{W}_{J\mathit{}\mathrm{\left(}{t}^{{x}^{\mathrm{-}\mathrm{1}}}\mathrm{\right)}}\mathit{}x}$.

#### Proof.

We have

${\overline{C}}_{x}\cap {H}_{t}=\left\{{W}_{J}x\mid J\subseteq S\right\}\cap \left\{{W}_{J}w\mid J\subseteq S,w\in W,J\left({t}^{{w}^{-1}}\right)\subseteq J\right\}$$=\left\{{W}_{J}x\mid J\left({t}^{{x}^{-1}}\right)\subseteq J\right\}$$=\overline{{W}_{J\left({t}^{{x}^{-1}}\right)}x},$

the first equality being a consequence of (2.1) and Lemma 2.2, and the last equality of (2.2). ∎

#### Lemma 2.4.

Let ${W}_{J}\mathit{}x$ be a face of the Coxeter complex. Then, the subspace $\mathrm{〈}{W}_{J}\mathit{}x\mathrm{〉}$ generated by ${W}_{J}\mathit{}x$ is the edge ${E}_{{T}_{J}^{x}}$.

#### Proof.

It is clear that the subspace generated by $\overline{{W}_{J}x}$ is equal to the subspace generated by ${W}_{J}x$. From the proof of Lemma 2.2, we know that ${H}_{t}$ being a hyperplane containing ${W}_{J}x$ is equivalent to $J\left({t}^{{x}^{-1}}\right)\subseteq J$ which is equivalent to $t\in {T}_{J}^{x}$. ∎

Let us take a reflection t of U. Recall that $l\left({E}_{U}\right)$ is half the number of chambers ${C}_{x}$ which have the property that ${E}_{U}$ is the minimal intersection containing $〈{\overline{C}}_{x}\cap {H}_{t}〉$. Minimality implies equality, that is, for the chamber ${C}_{x}$ to be counted, we must have $〈{\overline{C}}_{x}\cap {H}_{t}〉={E}_{U}$, i.e. (by Lemma 2.3)

$〈\overline{{W}_{J\left({t}^{{x}^{-1}}\right)}x}〉={E}_{U},$

i.e. (by Lemma 2.4)

${E}_{{T}_{J\left({t}^{{x}^{-1}}\right)}^{x}}={E}_{U}.$

Moreover, since ${t}^{{x}^{-1}}$ is a full support reflection of the group ${W}_{J\left({t}^{{x}^{-1}}\right)}$, we know from Lemma 2.1 that $J\left({t}^{{x}^{-1}}\right)$ is an irreducible subset of S. So ${E}_{U}$ is of the form ${E}_{J}^{x}$, where J is irreducible, otherwise $l\left({E}_{U}\right)=0$. That proves Proposition 1.2.

## 3 The chambers to consider

For a relevant edge E and a hyperplane ${H}_{t}$ containing E, we determine the chambers ${C}_{x}$ such that $〈{\overline{C}}_{x}\cap {H}_{t}〉=E$.

#### Lemma 3.1.

Let J be an irreducible subset of S, t a reflection with support J, and E the relevant edge ${E}_{{T}_{J}}$. For each K in the Coxeter class of J, let ${c}_{K\mathrm{,}J}$ be the coset of minimal length of ${N}_{W}\mathit{}\mathrm{\left(}{W}_{J}\mathrm{\right)}$ such that ${K}^{{c}_{K\mathrm{,}J}}\mathrm{=}J$. Then,

$L\left(E,t\right)=\bigsqcup _{K\in \left[J\right]}{c}_{K,J}\left\{x\in {N}_{W}\left({W}_{J}\right)\mid J\left({t}^{{x}^{-1}}\right)=J\right\}.$

#### Proof.

We have

$L\left(E,t\right)=\left\{x\in W\mid 〈{\overline{C}}_{x}\cap {H}_{t}〉=E\right\}$$=\left\{x\in W\mid 〈\overline{{W}_{J\left({t}^{{x}^{-1}}\right)}x}〉=E\right\}$$=\left\{x\in W\mid {E}_{{T}_{J\left({t}^{{x}^{-1}}\right)}^{x}}={E}_{{T}_{J}}\right\},$

the second and third equalities being consequences of Lemma 2.3 and Lemma 2.4, respectively. Denoting $J\left({t}^{{x}^{-1}}\right)$ by K, we see that the equality ${E}_{{T}_{J\left({t}^{{x}^{-1}}\right)}^{x}}={E}_{{T}_{J}}$ means that

• $|J\left({t}^{{x}^{-1}}\right)|=|J|$,

• $K\in \left[J\right]$, and

• $x\in {c}_{K,J}{N}_{W}\left({W}_{J}\right)$.

Hence

$L\left(E,t\right)=\bigsqcup _{K\in \left[J\right]}\left\{x\in {c}_{K,J}{N}_{W}\left({W}_{J}\right)\mid |J\left({t}^{{x}^{-1}}\right)|=|J|\right\}.$

Let $x={c}_{K,J}y$ with $y\in {N}_{W}\left({W}_{J}\right)$. Note that $|J\left({t}^{{x}^{-1}}\right)|=|J|$ if and only if $|J\left({t}^{{y}^{-1}}\right)|=|J|$ if and only if $J\left({t}^{{y}^{-1}}\right)=J$. We thus obtain the result. ∎

Recall that $X\left(S,J\right):=\left\{x\in {X}_{J}\mid {J}^{x}=J\right\}$. We introduce the set

${W}_{J}\left(t\right):=\left\{x\in {W}_{J}\mid J\left({t}^{{x}^{-1}}\right)=J\right\}.$

#### Lemma 3.2.

Let J be an irreducible subset of S. Consider a reflection t of ${W}_{J}$ with support J. Then,

$\left\{x\in {N}_{W}\left({W}_{J}\right)\mid J\left({t}^{{x}^{-1}}\right)=J\right\}=X\left(S,J\right)\cdot {W}_{J}\left(t\right).$

#### Proof.

We have ${N}_{W}\left({W}_{J}\right)={W}_{J}\cdot X\left(S,J\right)$, see [8, Corollary 3]. So

$\left\{x\in {N}_{W}\left({W}_{J}\right)\mid J\left({t}^{{x}^{-1}}\right)=J\right\}$$\mathrm{ }=\left\{yz\mid y\in {W}_{J},z\in X\left(S,J\right),J\left({t}^{{z}^{-1}{y}^{-1}}\right)=J\right\}.$

Since we always have ${t}^{{z}^{-1}}=t$, the remaining condition is $J\left({t}^{{y}^{-1}}\right)=J$. ∎

We write $C\left(x\right)$ for the centralizer of the element x of W.

#### Lemma 3.3.

Let J be an irreducible subset of S. Consider a reflection $t\mathrm{=}{s}^{v}$ with support J where s is a simple reflection and v an element of ${W}_{J}$. For another reflection u of ${W}_{J}$ with support J and conjugate to t, let ${c}_{u\mathrm{,}t}$ be the coset of minimal length of $C\mathit{}\mathrm{\left(}t\mathrm{\right)}$ such that ${u}^{{c}_{u\mathrm{,}t}}\mathrm{=}t$. Then,

${W}_{J}\left(t\right)=\bigsqcup _{u\in ⌊t⌉}{c}_{u,t}{N}_{{W}_{J}}{\left({W}_{\left\{s\right\}}\right)}^{v}.$

#### Proof.

The equality $J\left({t}^{{x}^{-1}}\right)=J$ means that ${x}^{-1}\in C\left(t\right){c}_{t,u}$ or $x\in {c}_{u,t}C\left(t\right)$, where u is a conjugate of t with support J. Then,

${W}_{J}\left(t\right)=\bigsqcup _{u\in ⌊t⌉}{c}_{u,t}C\left(t\right).$

Since $C\left(t\right)=C{\left(s\right)}^{v}$ and $C\left(s\right)={N}_{{W}_{J}}\left({W}_{\left\{s\right\}}\right)$, we get the result. ∎

Proposition 1.3 is a combination of Lemma 3.1, Lemma 3.2, and Lemma 3.3.

## 4 Invariance of the multiplicity

For the calculation of the multiplicity of an edge E, one has to choose a hyperplane containing E. We prove that the result of the calculation is independent of the choice of the hyperplane containing E.

#### Lemma 4.1.

Consider two hyperplanes ${H}_{u}$ and ${H}_{v}$ of ${\mathcal{A}}_{W}$ associated with the orthogonal vectors ${\alpha }_{u}$ and ${\alpha }_{v}$, respectively, of equal length and from the positive root system. Let r be the rotation of ${\mathrm{R}}^{n}$, not necessarily in W, which transforms ${\alpha }_{u}$ to ${\alpha }_{v}$. Then, $r\mathit{}\mathrm{\left(}{\mathcal{A}}_{W}\mathrm{\right)}\mathrm{=}{\mathcal{A}}_{W}$.

#### Proof.

Let ${E}_{u,v}$ be the 2-dimensional subspace $〈{\alpha }_{u},{\alpha }_{v}〉$:

• on ${E}_{u,v}$, the map r is the rotation of angle $\theta =\mathrm{arccos}\frac{{\alpha }_{u}.{\alpha }_{v}}{\parallel {\alpha }_{u}\parallel \parallel {\alpha }_{v}\parallel }$,

• on ${E}_{u,v}^{\perp }$, the map r is the identity map.

We have ${\mathcal{𝒜}}_{W}={\mathcal{𝒜}}_{1}\bigsqcup {\mathcal{𝒜}}_{2}$ with

${\mathcal{𝒜}}_{1}:=\left\{H\in \mathcal{𝒜}\mid {E}_{u,v}\subset H\right\}\mathit{ }\text{and}\mathit{ }{\mathcal{𝒜}}_{2}:=\left\{H\in \mathcal{𝒜}\mid dim{E}_{u,v}\cap H=1\right\}.$

• For all H in ${\mathcal{𝒜}}_{1}$, we have $r\left(H\right)=H$.

• Let p be the projection on the subspace ${E}_{u,v}$. The arrangement $p\left({\mathcal{𝒜}}_{2}\right)$ is the arrangement of a dihedral group whose angle between two certain hyperplanes is θ. Then, $r\left(p\left({\mathcal{𝒜}}_{2}\right)\right)=p\left({\mathcal{𝒜}}_{2}\right)$. Hence, for any H in ${\mathcal{𝒜}}_{2}$, we have

$r\left(H\right)=r\left(p\left(H\right)\oplus {E}_{u,v}^{\perp }\right)=r\left(p\left(H\right)\right)\oplus {E}_{u,v}^{\perp }$

which still belongs to ${\mathcal{𝒜}}_{2}$.∎

We prove Proposition 1.4 now. Consider two hyperplanes ${H}_{u}$ and ${H}_{v}$ containing the edge E. Let ${\alpha }_{u}$ and ${\alpha }_{v}$ be the unit vectors of the positive root system associated to W which are orthogonal to ${H}_{u}$ and ${H}_{v}$, respectively. We use the rotation r of Lemma 4.1 transforming ${\alpha }_{u}$ to ${\alpha }_{v}$, and leaving $\mathcal{𝒜}$ invariant. For a given w in $L\left(E,u\right)$, we have

${\overline{C}}_{w}\cap {H}_{u}={\overline{C}}_{w}\cap E$$\left({\overline{C}}_{w}\cap {H}_{u}\right)r=\left({\overline{C}}_{w}\cap E\right)r$$\left({\overline{C}}_{w}\right)r\cap \left({H}_{u}\right)r=\left({\overline{C}}_{w}\right)r\cap \left(E\right)r$$\left({\overline{C}}_{w}\right)r\cap {H}_{v}=\left({\overline{C}}_{w}\right)r\cap E.$

Hence $L\left(E,u\right)r=L\left(E,v\right)$, and $|L\left(E,u\right)|=|L\left(E,v\right)|$, which is Proposition 1.4.

## 5 The multiplicity of an edge

We establish a formula for the multiplicity of a relevant edge of ${\mathcal{𝒜}}_{W}$ in this section.

We begin with the proof of Theorem 1.1. From Proposition 1.2, we know that the relevant edges are the intersections of hyperplanes ${E}_{{T}_{J}^{w}}$ with the condition that J is irreducible. The weight of ${E}_{{T}_{J}^{w}}$ is obviously ${\prod }_{u\in {T}_{J}}{q}_{{u}^{w}}$ so that the real problem concerns $l\left({E}_{{T}_{J}^{w}}\right)$. Fixing a hyperplane ${H}_{t}$ containing ${E}_{{T}_{J}^{w}}$, we see that the set of chambers taken into account to determine $l\left({E}_{{T}_{J}^{w}}\right)$ is $L\left({E}_{{T}_{J}^{w}},t\right)$. But Proposition 1.4 allows us to choose any hyperplane containing ${E}_{{T}_{J}^{w}}$. Hence the multiplicity of ${E}_{{T}_{J}^{w}}$ is

$l\left({E}_{{T}_{J}^{w}}\right)=\frac{1}{2}|L\left({E}_{{T}_{J}^{w}},t\right)|.$

#### Lemma 5.1.

Let J be an irreducible subset of S, and w an element of W. Then,

$L\left({E}_{{T}_{J}^{w}},{t}^{w}\right)=L\left({E}_{{T}_{J}},t\right)w.$

#### Proof.

Let $x\in L\left({E}_{{T}_{J}},t\right)$. We have

${\overline{C}}_{x}\cap {H}_{t}={\overline{C}}_{x}\cap {E}_{{T}_{J}}$$\left({\overline{C}}_{x}\cap {H}_{t}\right)w=\left({\overline{C}}_{x}\cap {E}_{{T}_{J}}\right)w$$\left({\overline{C}}_{x}\right)w\cap \left({H}_{t}\right)w=\left({\overline{C}}_{x}\right)w\cap \left({E}_{{T}_{J}}\right)w$${\overline{C}}_{xw}\cap {H}_{{t}^{w}}={\overline{C}}_{xw}\cap {E}_{{T}_{J}^{w}}.$

Then

$L\left({E}_{{T}_{J}},t\right)w\subseteq L\left({E}_{{T}_{J}^{w}},{t}^{w}\right).$

With the same argument, we obtain also $L\left({E}_{{T}_{J}^{w}},{t}^{w}\right){w}^{-1}\subseteq L\left({E}_{{T}_{J}},t\right)$. Hence $L\left({E}_{{T}_{J}^{w}},{t}^{w}\right)=L\left({E}_{{T}_{J}},t\right)w$. ∎

We see in Lemma 5.1 that we just need to investigate $L\left({E}_{{T}_{J}},t\right)$ for each Coxeter class $\left[J\right]$. From Proposition 1.3, we get

$l\left({E}_{{T}_{J}}\right)=\frac{1}{2}|L\left({E}_{{T}_{J}},t\right)|$$=\frac{1}{2}|\bigsqcup _{K\in \left[J\right]}{c}_{K,J}X\left(S,J\right)\left(\bigsqcup _{u\in ⌊t⌉}{c}_{u,t}{N}_{{W}_{J}}{\left({W}_{\left\{s\right\}}\right)}^{v}\right)|$$=|⌊t⌉|\cdot |\left[J\right]|\cdot |X\left(S,J\right)|\cdot |X\left(J,\left\{s\right\}\right)|$

since

$\frac{1}{2}|{N}_{{W}_{J}}\left({W}_{\left\{s\right\}}\right)|=|X\left(J,\left\{s\right\}\right)|,$

which finishes the proof of Theorem 1.1.

## 6 Computing the determinants of finite Coxeter groups

Before computing the Varchenko determinants of the irreducible finite Coxeter groups, we first have to determine their numbers of full support reflections.

The graph $\mathrm{\Gamma }=\left(S,M\right)$ associated to a finite Coxeter group W is defined as follows:

• the vertex set S is the simple reflection set of W,

• the edge set M is composed of the pairs of simple reflections $\left\{{s}_{i},{s}_{j}\right\}$ such that ${m}_{ij}\ge 3$, and these edges $\left\{{s}_{i},{s}_{j}\right\}$ are labeled by ${m}_{ij}$.

We only consider the graphs associated to the irreducible finite Coxeter groups [7, Table 1.1]. The reflections of an irreducible finite Coxeter group W form a single conjugacy class if and only if the edge labels of its associated graph are all odd [3, n${}^{\circ }$1.3, Proposition 3]. This is the case of the Coxeter groups ${A}_{n-1}$, ${D}_{n}$, ${E}_{6}$, ${E}_{7}$, ${E}_{8}$, ${H}_{3}$, ${H}_{4}$, and ${I}_{2}\left(m\right)$ (m odd).

For the case of the Coxeter groups ${B}_{n}$, ${F}_{4}$, and ${I}_{2}\left(m\right)$ (m even), removing the even labeled edge from their associated graphs leaves two graphs of type A. So their reflections form two conjugacy classes.

We can see the numbers of reflections of the irreducible finite Coxeter groups in the book of Björner and Brenti [2, Appendix A1] for example. Using the Principle of Inclusion and Exclusion, applied to the irreducible maximal parabolic subgroups, and Pascal’s triangle in the form $\left(\genfrac{}{}{0pt}{}{n}{2}\right)-\left(\genfrac{}{}{0pt}{}{n-1}{2}\right)=n-1$, we get the number of full support reflections of the irreducible finite Coxeter groups in Table 1.

Table 1

Number of full support reflections of the irreducible finite Coxeter groups.

Table 2

Multiplicities of the Coxeter classes – part 1.

Table 3

Multiplicities of the Coxeter classes – part 2.

We are now able to compute the Varchenko determinant of a finite Coxeter group by using Theorem 1.1. The ingredients necessary to calculate this determinant are given in Tables 23 for all irreducible finite Coxeter groups. They are obtained with tools of Table 1, those in [7, Propositions 2.3.8, 2.3.10, 2.3.13], and [7, Tables A.1, A.2].

Let $\left[±n\right]:=\left\{-n,\mathrm{\dots },-2,-1,1,2,\mathrm{\dots },n\right\}$. We write $\overline{{2}^{\left[±n\right]}}$ for the subset of ${2}^{\left[±n\right]}$ having the following properties:

• the elements of $\overline{{2}^{\left[±n\right]}}$ are the elements $\left\{{i}_{1},\mathrm{\dots },{i}_{t}\right\}$ of ${2}^{\left[±n\right]}$ such that $|{i}_{r}|\ne |{i}_{s}|$ if $r\ne s$; and

• if $\left\{{i}_{1},\mathrm{\dots },{i}_{t}\right\}\in \overline{{2}^{\left[±n\right]}}$, then $\left\{-{i}_{1},\mathrm{\dots },-{i}_{t}\right\}\notin \overline{{2}^{\left[±n\right]}}$.

Using Theorem 1.1 and Tables 23, we refind, for example, the determinants

$det{\mathcal{𝒜}}_{{B}_{n}}=\prod _{\begin{array}{c}J\in \overline{{2}^{\left[±n\right]}}\\ |J|\ge 2\end{array}}{\left(1-\prod _{\left\{i,j\right\}\in \left(\genfrac{}{}{0pt}{}{J}{2}\right)}{a}_{{H}_{i,j}}^{2}\right)}^{{2}^{n-|J|+1}\left(|J|-2\right)!\left(n-|J|+1\right)!}$$\mathrm{ }×\prod _{\begin{array}{c}I\in {2}^{\left[n\right]}\\ |I|\ge 1\end{array}}{\left(1-\prod _{i\in I}{a}_{{H}_{i}}^{2}\prod _{\left\{i,j\right\}\in \left(\genfrac{}{}{0pt}{}{I}{2}\right)}{a}_{{H}_{i,j}}^{2}{a}_{{H}_{-i,j}}^{2}\right)}^{{2}^{n-1}\left(|I|-1\right)!\left(n-|I|\right)!},$

computed by Randriamaro [10] with combinatorial methods.

## References

• [1]

P. Abramenko and K. S. Brown, Buildings. Theory and Applications, Grad. Texts in Math. 248, Springer, New York, 2008.  Google Scholar

• [2]

A. Björner and F. Brenti, Combinatorics of Coxeter Groups, Grad. Texts in Math. 231, Springer, New York, 2005.  Google Scholar

• [3]

N. Bourbaki, Éléments de mathématique. Groupes et algèbres de Lie. Chapitres 4, 5 et 6, Masson, Paris, 1981.  Google Scholar

• [4]

H. S. M. Coxeter, Discrete groups generated by reflections, Ann. of Math. (2) 35 (1934), no. 3, 588–621.

• [5]

G. Denham and P. Hanlon, Some algebraic properties of the Schechtman–Varchenko bilinear forms, New Perspectives in Algebraic Combinatorics (Berkeley 1996/97), Math. Sci. Res. Inst. Publ. 38, Cambridge University Press, Cambridge (1999), 149–176.  Google Scholar

• [6]

G. Duchamp, A. Klyachko, D. Krob and J.-Y. Thibon, Noncommutative symmetric functions. III. Deformations of Cauchy and convolution algebras, Discrete Math. Theor. Comput. Sci. 1 (1997), no. 1, 159–216.  Google Scholar

• [7]

M. Geck and G. Pfeiffer, Characters of Finite Coxeter Groups and Iwahori–Hecke Algebras, London Math. Soc. Monogr. (N. S.) 21, Oxford University Press, New York, 2000.  Google Scholar

• [8]

R. B. Howlett, Normalizers of parabolic subgroups of reflection groups, J. Lond. Math. Soc. (2) 21 (1980), no. 1, 62–80.  Google Scholar

• [9]

J. E. Humphreys, Reflection Groups and Coxeter Groups, Cambridge Stud. Adv. Math. 29, Cambridge University Press, Cambridge, 1990.  Google Scholar

• [10]

H. Randriamaro, Computing the Varchenko determinant of a bilinear form, preprint (2014), https://arxiv.org/abs/1411.1647.

• [11]

V. V. Schechtman and A. N. Varchenko, Quantum groups and homology of local systems, Algebraic Geometry and Analytic Geometry (Tokyo 1990), ICM-90 Satell. Conf. Proc., Springer, Tokyo (1991), 182–197.  Google Scholar

• [12]

A. Varchenko, Bilinear form of real configuration of hyperplanes, Adv. Math. 97 (1993), no. 1, 110–144.

• [13]

D. Zagier, Realizability of a model in infinite statistics, Comm. Math. Phys. 147 (1992), no. 1, 199–210.

Revised: 2018-03-07

Published Online: 2018-04-10

Published in Print: 2018-07-01

This research was supported through the programme “Oberwolfach Leibniz Fellows” by the Mathematisches Forschungsinstitut Oberwolfach in 2017.

Citation Information: Journal of Group Theory, Volume 21, Issue 4, Pages 651–665, ISSN (Online) 1435-4446, ISSN (Print) 1433-5883,

Export Citation

© 2018 Walter de Gruyter GmbH, Berlin/Boston.