Bourbaki

Lemma 2

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

Lie Groups and Lie Algebras, chapter VII, § 5. Decomposable linear Lie algebras

2^2 (i) Let W be a vector subspace of V of dimension dd, D the line ⋀dW⊂⋀dV,θ\bigwedge^dW\subset \bigwedge^dV,\theta the canonical representation of gl(V)\mathfrak{g}\mathfrak{l}(V) on ⋀V\bigwedge V (Chap. III, App.). Let x∈gl(V)x\in \mathfrak{g}\mathfrak{l}(V). Then x(W)⊂Wx(W)\subset W if and only if θ(x)(D)⊂D\theta (x)(D)\subset D.

(ii) Let (e1,...,en)(e_1, . . . , e_n) be the canonical basis of kn,θk^n,\theta the canonical representation of gl(n,k)\mathfrak{g}\mathfrak{l}(n, k) on ⋀(kn)\bigwedge(k^n), and x∈gl(n,k)x\in \mathfrak{g}\mathfrak{l}(n, k). Then x∈n(n,k)x\in \mathfrak{n}(n, k) if and only if

θ(x)(en−d+1∧⋯∧en)=0 \theta (x)(e_{n-d+1}\wedge \cdots \wedge e_n) = 0

for 1≤d≤n1\leq d\leq n.

(i) If x(W)⊂Wx(W)\subset W, it is clear that θ(x)D⊂D\theta (x)D\subset D. Conversely, assume that θ(x)D⊂D\theta (x)D\subset D. Let uu be a non-zero element of D and let y∈Wy\in W. Then y∧u=0y\wedge u= 0. Since θ(x)\theta (x) is a derivation of ⋀V\bigwedge V, this implies

θ(x)y∧u+y∧θ(x)u=0 \theta (x)y\wedge u+y\wedge \theta (x)u= 0

Now θ(x)u∈ku\theta (x)u\in ku, so y∧θ(x)u=0y\wedge \theta (x)u= 0 and consequently θ(x)y∧u=0\theta (x)y\wedge u= 0. By Algebra, Chap. III, §7, no. 9, Prop. 13, this implies that θ(x)y∈W\theta (x)y\in W, i.e. x(y)∈Wx(y)\in W, which proves that x(W)⊂Wx(W)\subset W.

(ii) The condition stated in (ii) is clearly necessary for x∈n(n,k)x\in \mathfrak{n}(n, k). Assume that it is satisfied. By (i), xx leaves

2^2 In this lemma, kk can be an arbitrary (commutative) field.

ken−d+1+⋯+ken ke_{n-d+1}+\cdots +ke_n

stable, and since this holds for d=1,...,n,xd= 1, . . . , n,x is lower triangular. Put

x=(xij)1≤i,j≤n x= (x_{ij})_{1\leq i,j\leq n}

We have 0=x(en)=xnnen0 =x(e_n) =x_{nn}e_n, so xnn=0x_{nn}= 0. Let i<ni < n, and assume that we have proved that xjj=0x_{jj}= 0 for j>ij > i. Then

0=θ(x)(ei∧ei+1∧⋯∧en)=xii(ei∧ei+1∧⋯∧en) 0 =\theta (x)(e_i\wedge e_{i+1}\wedge \cdots \wedge e_n) =x_{ii}(e_i\wedge e_{i+1}\wedge \cdots \wedge e_n)

so xii=0x_{ii}= 0. Thus, x∈n(n,k)x\in \mathfrak{n}(n, k).

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

  • Algebra, Chap. III, §7, no. 9, Prop. 13 not in this corpus
Transcribed from the printed volume. Not translated.