Let E be a finite dimensional simple g-module, λ its highest weight and (⋅∣⋅) a W-invariant non-degenerate positive symmetric bilinear form on h∗. Then:
∏⟨λ+ρ,Hα⟩∏((λ∣α))
dim E = = 1 +.
α∈R+⟨ρ,Hα⟩α∈R+(ρ∣α)
Let T be an indeterminate. For all ν∈P, denote by fν the homomorphism from Z[P] to R[[T]] that takes eμ to e(ν∣μ)T for all μ∈P. Then dim E is the constant term of the series fν(ch E).
For all μ,ν∈P, we have
fν(J(eμ))=w∈W∑ε(w)e(ν∣wμ)T
=w∈W∑ε(w)e(w−1ν∣μ)T=fμ(J(eν))
In particular, in view of Chap. VI, §3, no. 3, formula (3),
fρ(J(eμ))=fμ(J(eρ))=e(μ∣ρ)Tα∈R+∏(1−e−(μ∣α)T)
Hence, setting Card(R+) =N,
fρ(J(eμ))≡TN∏α∈R(μ∣α) (mod TN+1R[[T]]).
The equality J(eλ+ρ)= ch(E).J(eρ) (Th. 1) thus implies that
TN∏α∈R+(λ+ρ∣α)≡fρ(ch E).TN∏α∈R+(ρ∣α) (mod TN+1R[[T]])
so
dim E=((∏α∈R+(λ+ρ∣α))/((∏α∈R+(ρ∣α))=∏α∈R+(1+(ρ∣α)(λ∣α)).
Now, if α∈R+,α can be identified with an element of hR proportional to Hα, so
(λ+ρ∣α)/(ρ∣α)=⟨λ+ρ,Hα⟩/⟨ρ,Hα⟩
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.