(i) The set R=R(g,h) is a reduced root system in h∗.
(ii) Let α∈R. The map sα,Hα:λ→λ−λ(Hα)α from h∗ to h∗ is the unique reflection s of h∗ such that s(α)=−α and s(R)=R. For all t∈k∗, s is the transpose of θα(t)∣h.
First, R generates h∗, for if h∈h is such that α(h)=0 for all α∈R, then ad h=0 and hence h=0 since the centre of g is zero. By definition, 0∈/R. Let α∈R. Since α(Hα)=2,s=sα,Hα is a reflection such that s(α)=−α. Then s(R)=R by Lemma 1 (iii), and β(Hα)∈Z for all β∈R (Prop. 2 (i)). This shows that R is a root system in h∗. For all h∈h and λ∈h∗,
⟨s(λ),h⟩=⟨λ−λ(Hα)α,h⟩=⟨λ,h−α(h)Hα⟩=⟨λ,θα(t)h⟩
so s is the transpose of θα(t)∣h. Finally, we show that the root system R is reduced. Let α∈R and y∈g2α. Since 3α∈/R (Chap. VI, §1, no. 3, Prop. 8),[Xα,y]=0; on the other hand, [X−α,y]∈g−α+2α=gα=kXα, so [Xα,[X−α,y]]=0; thus
4y=2α(Hα)y=[Hα,y]=−[[Xα,X−α],y]=0
so y=0 and g2α=0. In other words, 2α is not a root. Q.E.D.
Identify h canonically with h∗∗. With the notations of Chap. VI, §1, no. 1, we then have, by Th. 2 (ii),
Hα=α∨ for all α∈R. (2)
The Hα thus form the root system R∨ in h inverse to R.
We shall call R(g,h) the root system of (g,h). The reflections sα,Hα will be denoted simply by sα. The Weyl group, group of weights, Coxeter number ... of R(g,h) are called the Weyl group, group of weights, Coxeter number ... of (g,h). As in Chap. VI, §1, no. 1, we consider the Weyl group as operating not only on h∗, but also on h by transport of structure, so that sα=θα(t)∣h. Since the θα(t) are elementary automorphisms of g, we have:
Bourbaki sets the proof directly after the statement, and the corpus does not separate them, so what is above is the statement and its proof together.