Let (g,h) be a split semi-simple Lie algebra. Then (g,h) possesses a Chevalley system (§2, no. 4, Def. 3).
Let R be the root system of (g,h). For all α∈R, let Xα be a non-zero element of gα. Assume that the Xα are chosen so that [Xα,X−α]=−Hα for all α∈R(§2, no. 4, Lemma 2). Let B be a basis of R and θ the automorphism of g such that θ(Xα)=X−α for all α∈B∪(−B). We have θ∣h=−Idh. Hence, for all α∈R there exists tα∈k∗ such that θXα=tαX−α. We have
tαt−αHα=[tαX−α,t−αXα]=[θXα,θX−α]=θ([Xα,X−α])
=θ(−Hα)=Hα
so tαt−α=1 for all α∈R. Introduce the Nαβ as in §2, no. 4. If α,β,α+β∈R,
N−α,−βtαtβX−α−β=tαtβ[X−α,X−β]=[θXα,θXβ]=θ([Xα,Xβ])
=NαβθXα+β=Nαβtα+βX−α−β
so, in view of §2, no. 4, Lemma 4,
(q+1)2tαtβ=Nαβ2tα+β
where q is an integer. It follows that if tα and tβ are squares in k∗, so is tα+β. Since tα=1 for all α∈B, Prop. 19 of Chap. VI, §1, no. 6, proves that tα is a square for all α∈R. Choose, for all α∈R, a uα∈k such that uα2=tα. This choice can be made so that uαu−α=1 for all α∈R. Put Xα′=uα−1Xα. Then, for all α∈R,
Xα′∈gα,[Xα′,X−α′]=[Xα,X−α]=−Hα
and θ(Xα′)=θ(uα−1Xα)=uα−1tαX−α=uαX−α=uαu−αX−α′=X−α′, so that (Xα′)α∈R is a Chevalley system of (g,h).
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.