Bourbaki

Proposition 10 (elimination theorem)

Tag 0A39, permanent. Cite it as [Bourbaki, Tag 0A39].

Lie Groups and Lie Algebras, chapter II, § 2. Free Lie algebras

Let X be a set, S a subset of X and T the set of sequences (s1,…,sn,x)(s_1, \ldots, s_n, x) with n≥0,s1,…,snn \geq 0, s_1, \ldots, s_n in S and x in X −S.- S.†

(a) The module L(X)L(X) is the direct sum of the subalgebra L(S)L(S) of L(X)L(X) and the ideal aa of L(X)L(X) generated by X−SX - S.

(b) There exists a Lie algebra isomorphism ϕ\phi of L(T)L(T) onto aa which maps (s1,…,sn,x)(s_1, \ldots, s_n, x) to (ad⁡s1∘⋯∘ad⁡sn)(x)(\operatorname{ad} s_1 \circ \cdots \circ \operatorname{ad} s_n)(x).

Let gg be the Lie algebra constructed as in Proposition 9 given

S1=S,S2=T,d(s,t)=(s,s1,…,sn,x)∈T⊂L(T) S_1 = S, \quad S_2 = T, \quad d(s, t) = (s, s_1, \ldots, s_n, x) \in T \subset L(T)

for t=(s1,…,sn,x)t = (s_1, \ldots, s_n, x) in T and s∈S1s \in S_1. We identify L(S)L(S) and L(T)L(T) with their canonical images in gg (Proposition 9 (a)).

Let ψ\psi be the mapping (s1,…,sn,x)↦(ad⁡s1∘⋯∘ad⁡sn)(x)(s_1, \ldots, s_n, x) \mapsto (\operatorname{ad} s_1 \circ \cdots \circ \operatorname{ad} s_n)(x) of T into L(X)L(X). Obviously ψ(d(s,t))=[s,ψ(t)]\psi(d(s, t)) = [s, \psi(t)] for s∈Ss \in S and t∈Tt \in T and hence there exists a homomorphism α:g→L(X)\alpha : g \to L(X) whose restriction to S is the identity and whose restriction to T is ψ\psi. Now X−S⊂TX - S \subset T, whence there is a homomorphism β:L(X)→g\beta : L(X) \to g whose restriction to X=S∪(X−S)X = S \cup (X - S) is the identity.

We show that α\alpha is an isomorphism and β\beta the inverse isomorphism. As ψ(x)=x\psi(x) = x for xx in X−SX - S, we see that α∘β\alpha \circ \beta coincides with the identity on X, whence α∘β=IdL(X)\alpha \circ \beta = \mathrm{Id}_{L(X)}. On the other hand, [s,t]=d(s,t)[s, t] = d(s, t) in gg for s∈S,t∈Ts \in S, t \in T by construction; it follows that t=(s1,…,sn,x)t = (s_1, \ldots, s_n, x) is equal in gg to (ad⁡s1∘⋯∘ad⁡s2)(x)(\operatorname{ad} s_1 \circ \cdots \circ \operatorname{ad} s_2)(x) whence t=β(α(t))t = \beta(\alpha(t)). As β(α(s))=s\beta(\alpha(s)) = s for s∈Ss \in S and S∪TS \cup T generates gg, β∘α=Idg\beta \circ \alpha = \mathrm{Id}_g.

As α\alpha is an isomorphism of gg onto L(X)L(X), Proposition 9 shows that the restric-

† For n=0n = 0, we obtain the elements of X−SX - S, whence X−S⊂TX - S \subset T.

tion of α\alpha to L(T)L(T) is an isomorphism ϕ\phi of L(T)L(T) onto an ideal bb of L(X)L(X) such that the module L(X)L(X) is the direct sum of L(S)L(S) and bb. Obviously

ϕ(s1,…,sn,x)=(ad s1∘⋯∘ad sn)(x) \phi(s_1, \ldots, s_n, x) = (\mathrm{ad}\ s_1 \circ \cdots \circ \mathrm{ad}\ s_n)(x)
for (s1,…,sn,x)(s_1, \ldots, s_n, x) in TT.

Hence ϕ(T)⊂a\phi(T) \subset a, whereas b⊂ab \subset a since ϕ(T)\phi(T) generates the subalgebra bb of L(X)L(X). But bb is an ideal and X−S⊂ϕ(T)⊂bX - S \subset \phi(T) \subset b, whence a⊂ba \subset b.

Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.

Cited by

Cites

Transcribed from the printed volume. Not translated.