Bourbaki

Proposition 10

Tag 0ARY, permanent. Cite it as [Bourbaki, Tag 0ARY].

Lie Groups and Lie Algebras, chapter V, § 4. The geometric representation of a Coxeter group

The group W provided with the discrete topology acts properly on A; it is generated by orthogonal reflections; it is infinite, irreducible and essential (§ 3, no. 7). The intersection C ∩ A is a chamber of A for W. If Ls denotes the hyperplane of A formed by the intersection of A with the hyperplane of E* orthogonal to e_s, the L_s for s ∈ S are the walls of C ∩ A. If εs\varepsilon_s is the unit vector of T orthogonal to L_s on the same side of L_s as C ∩ A, then (εs∣εt)=−cos⁡(π/m(s,t))(\varepsilon_s|\varepsilon_t) = -\cos(\pi/m(s,t)) (for s, t ∈ S) and the Coxeter matrix of W (§ 3, no. 4) is identified with M.*

By Cor. 3 of Th. 1, W is discrete in GL(E)\mathbf{GL}(E), and hence in G, and acts properly on A. Let s∈Ss \in S. Since Card S ≥ 2, the hyperplane of E* orthogonal to ese_s is not orthogonal to v and its intersection L_s with A is indeed a hyperplane. The displacement corresponding to s is thus a displacement of order 2 leaving fixed all the points of L_s: it is necessarily the orthogonal reflection associated to L_s. It follows that W is generated by orthogonal reflections. Th. 2 now shows that it is infinite and Prop. 7 that it is essential and irreducible.

Since C is an open simplicial cone, whose walls are the hyperplanes with equations ⟨x∗,es⟩=0\langle x^*, e_s \rangle = 0 (for s∈Ss \in S), the intersection C ∩ A is a convex, hence connected, open and closed subset of the complement of the union of the L_s in A. Moreover, C ∩ A is non-empty, for if x∗∈Cx^* \in C we have ⟨x∗,v⟩=∑svs⟨x∗,es⟩>0\langle x^*, v \rangle = \sum_s v_s \langle x^*, e_s \rangle > 0 and ⟨x∗,v⟩−1x∗∈C∩A\langle x^*, v \rangle^{-1} x^* \in C \cap A. It follows that C ∩ A is a chamber of A relative to the system of the L_s. Moreover, w(C∩A)∩Ls=∅w(C \cap A) \cap L_s = \varnothing for all w∈Ww \in W (cf. no. 4, property (P_n)) and it follows that C ∩ A is a chamber of A relative to the system consisting of the transforms of the L_s by the elements of W; by the Cor. of Th. 1 of § 3, no. 2, it follows that C ∩ A is a chamber of A relative to W.

Let as∗a_s^* be the vertex of the simplex C ∩ A not in L_s. We have

⟨as∗,et⟩=0 \langle a_s^*, e_t \rangle = 0

for s,t∈Ss, t \in S and s≠ts \neq t, and

⟨as∗,es⟩=vs−1⟨as∗,v⟩=vs−1. \langle a_s^*, e_s \rangle = v_s^{-1} \langle a_s^*, v \rangle = v_s^{-1}.

Let εs\varepsilon_s be the vector in T defined by the relations:

(εs∣as∗−at∗)=vs−1for t∈S, t≠s. (\varepsilon_s|a_s^* - a_t^*) = v_s^{-1} \quad \text{for } t \in S, \ t \neq s.

The vector εs\varepsilon_s is orthogonal to LsL_s and is on the same side of the hyperplane LsL_s as C∩AC \cap A. Moreover

(εs∣as∗−at∗)=⟨es,as∗−at∗⟩for all s,t∈S (\varepsilon_s | a_s^* - a_t^*) = \langle e_s, a_s^* - a_t^* \rangle \quad \text{for all } s, t \in S

which shows that εs\varepsilon_s is the image of the class of ese_s under the isomorphism from E/E0E/E^0 to TT given by the quadratic form BMB_M. It follows that

(εs∣εt)=BM(es,et). (\varepsilon_s | \varepsilon_t) = B_M(e_s, e_t).

Consequently εs\varepsilon_s is a unit vector and the last assertion of Prop. 10 is proved.

The euclidean affine space AA provided with the group WW is called the space associated to the Coxeter matrix MM and we denote it by AMA_M.

Proposition 10 admits a converse:

Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.

Cited by

Nothing.

Cites

Transcribed from the printed volume. Not translated.