Bourbaki

Proposition 4

Tag 01GP, permanent. Cite it as [Bourbaki, Tag 01GP].

Lie Groups and Lie Algebras, chapter IX, § 7. Irreducible representations of connected compact Lie groups

Let Γ\Gamma be the Casimir element of G.

a) If τ\tau is an irreducible representation of G,Γ~(τ),\widetilde{\Gamma}(\tau ) is real and positive. If τ\tau is not the trivial representation, Γ~(τ)>0\widetilde{\Gamma}(\tau )>0.

b) There exists a unique quadratic form QΓQ_{\Gamma} on X(T)⊗RX(T)\otimes \mathbf{R} such that, for every irreducible representation τ\tau of G,

Γ~(τ)=QΓ(λ+ρ)−QΓ(ρ) \widetilde{\Gamma}(\tau ) = Q_{\Gamma}(\lambda +\rho )-Q_{\Gamma}(\rho )

where λ\lambda is the highest weight of τ\tau. The form QΓQ_{\Gamma} is positive, separating and invariant under W.

Let F be a separating negative symmetric bilinear form on g\mathfrak{g} defining Γ\Gamma. Let τ:G→GL(V)\tau : G\rightarrow \mathbf{G}\mathbf{L}(V) be an irreducible representation of G; let ⟨,⟩\langle ,\rangle be a Hilbert scalar product on V invariant under G (§1, no. 1), and (ei)(e_i) a basis of g\mathfrak{g} such that F(ei,ej)=−δijF(e_i, e_j) =-\delta_{ij}. Then, for every element vv of V not invariant under G we have

Γ~(τ)⟨v,v⟩=⟨v,ΓV(v)⟩=−∑i⟨v,(ei)V2v⟩=∑i⟨v,(ei)V∗(ei)Vv⟩ \widetilde{\Gamma}(\tau )\langle v, v\rangle =\langle v, \Gamma_V(v)\rangle =-\sum_i\langle v,(e_i)^2_Vv\rangle =\sum_i\langle v,(e_i)^*_V(e_i)_Vv\rangle
=∑i⟨(ei)Vv,(ei)Vv⟩>0 =\sum_i\langle (e_i)_Vv,(e_i)_Vv\rangle >0

hence a).

Let B be the inverse form on tC∗\mathfrak{t}^*_{\mathbf{C}} of the restriction to tC\mathfrak{t}_{\mathbf{C}} of the bilinear form on gC\mathfrak{g}_{\mathbf{C}} induced by F by extension of scalars. By the Cor. of Prop. 7 of Chap. VIII, §6, no. 4, we have5Γ~(τ)=B(δ(λ),δ(λ+2ρ))^5\widetilde{\Gamma}(\tau ) = B(\delta (\lambda ), \delta (\lambda +2\rho )). Extend δ:X(T)→\delta : X(T)\rightarrow tC∗\mathfrak{t}^*_{\mathbf{C}} to an R\mathbf{R}-linear map from X(T)⊗RX(T)\otimes \mathbf{R} to tC∗\mathfrak{t}^*_{\mathbf{C}} and let QΓQ_{\Gamma} be the quadratic form x→B(δ(x),δ(x))x \rightarrow B(\delta (x), \delta (x)) on X(T)⊗RX(T)\otimes \mathbf{R}; it is separating and invariant under W, and

Γ~(τ)=B(δ(λ+ρ),δ(λ+ρ))−B(δ(ρ),δ(ρ))=QΓ(λ+ρ)−QΓ(ρ) \widetilde{\Gamma}(\tau ) = B(\delta (\lambda +\rho ), \delta (\lambda +\rho ))-B(\delta (\rho ), \delta (\rho )) = Q_{\Gamma}(\lambda +\rho )-Q_{\Gamma}(\rho )

We show that the form QΓQ_{\Gamma} is positive. Indeed, if x∈X(T)⊗Qx\in X(T)\otimes \mathbf{Q}, the element δ(x)\delta (x) of tC∗\mathfrak{t}^*_{\mathbf{C}} takes purely imaginary values on t\mathfrak{t}, hence real values on iti\mathfrak{t}; we conclude by remarking that, for y∈ity\in i\mathfrak{t}, we have F(y,y)≥0F(y, y)\geq 0.

It remains to prove the uniqueness assertion in b). Let Q be a quadratic form on X(T)⊗RX(T)\otimes \mathbf{R} satisfying the required condition, and let Φ\Phi (resp. ΦΓ\Phi_{\Gamma}) be the bilinear form associated to Q (resp. QΓQ_{\Gamma}). For λ,μ∈X++\lambda , \mu\in X_{++}, we have

Φ(λ,μ)=(Q(λ+μ+ρ)−Q(ρ))−(Q(λ+ρ)−Q(ρ))−(Q(μ+ρ)−Q(ρ)) \Phi (\lambda , \mu) = (Q(\lambda +\mu+\rho )-Q(\rho ))-(Q(\lambda +\rho )-Q(\rho ))-(Q(\mu+\rho )-Q(\rho ))
=ΦΓ(λ,μ) =\Phi_{\Gamma}(\lambda , \mu)

Since X(T)++X(T)_{++} generates the R\mathbf{R}-vector space X(T)⊗RX(T)\otimes \mathbf{R}, we have Φ=ΦΓ\Phi =\Phi_{\Gamma}, so Q=QΓQ = Q_{\Gamma}.

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.

Cited by

Cites

Transcribed from the printed volume. Not translated.