Bourbaki

Proposition 17

Tag 0AJ4, permanent. Cite it as [Bourbaki, Tag 0AJ4].

Lie Groups and Lie Algebras, chapter III, § 9. Commutators, centralizers and normalizers in a Lie group

Let gg be a finite-dimensional nilpotent Lie algebra over KK and (g0,g1,…,gn)(g_0, g_1, \ldots, g_n) a decreasing sequence of ideals of gg such that g0=g,gn={0}g_0 = g, g_n = \{0\} and [g,gi]⊂gi+1[g, g_i] \subset g_{i+1} for 0≤i<n0 \leq i < n. Let a1,a2,…,apa_1, a_2, \ldots, a_p be vector subspaces of gg such that each gig_i is the direct sum of its intersections with the aja_j. Let gg be given the Hausdorff law of composition HH. Let ϕ\phi be the mapping

(x1,x2,…,xp)↦x1Hx2H⋯Hxp (x_1, x_2, \ldots, x_p) \mapsto x_1 \mathbf{H} x_2 \mathbf{H} \cdots \mathbf{H} x_p

of a1×a2×⋯×apa_1 \times a_2 \times \cdots \times a_p into gg.

(i) ϕ\phi is a bijection of a1×a2×⋯×apa_1 \times a_2 \times \cdots \times a_p onto gg; (ii) ϕ\phi and ϕ−1\phi^{-1} are polynomial mappings; (iii) the mapping (x,y)↦ϕ−1(ϕ(x)⋅ϕ(y)−1)(x, y) \mapsto \phi^{-1}(\phi(x) \cdot \phi(y)^{-1}) of (a1×a2×⋯×ap)2(a_1 \times a_2 \times \cdots \times a_p)^2 into a1×a2×⋯×apa_1 \times a_2 \times \cdots \times a_p is polynomial.

The proposition is obvious for dim⁡g=0\dim g = 0. Suppose that dim⁡g>0\dim g > 0 and that the proposition has been established for dimensions <dim⁡g< \dim g. It can be assumed that gn−1≠{0}g_{n-1} \neq \{0\} and gn−1g_{n-1} is then a non-zero central ideal of gg. There exists an index jj such that h=gn−1∩aj≠{0}h = g_{n-1} \cap a_j \neq \{0\}. Let g′=g/hg' = g/h, θ\theta be the canonical morphism of gg onto g′g', gi′=θ(gi)g'_i = \theta(g_i) and ai′=θ(ai)a'_i = \theta(a_i). Then (g0′,g1′,…,gn′)(g'_0, g'_1, \ldots, g'_n) is a decreasing sequence of ideals of g′g' such that g0′=g′,gn′={0}g'_0 = g', g'_n = \{0\},

[g′,gi′]⊂gi+1′ [g', g'_i] \subset g'_{i+1}

for 0≤i<n0 \leq i < n and each gi′g'_i is the direct sum of its intersections with the aj′a'_j. Let ϕ′\phi' be the mapping

(x1′,x2′,…,xp′)↦x1′Hx2′H⋯Hxp′ (x'_1, x'_2, \ldots, x'_p) \mapsto x'_1 \mathbf{H} x'_2 \mathbf{H} \cdots \mathbf{H} x'_p

of a1′×a2′×⋯×ap′a'_1 \times a'_2 \times \cdots \times a'_p into g′g'. By the induction hypothesis, ϕ′\phi' is bijective and ϕ′,ϕ′−1\phi', {\phi'}^{-1} are polynomial mappings.

Let x∈gx \in g. We write

(1)

ϕ′−1(θ(x))=(x1′(x),x2′(x),…,xp′(x)). {\phi'}^{-1}(\theta(x)) = (x'_1(x), x'_2(x), \ldots, x'_p(x)).
Then (2)
θ(x)=x1′(x)Hx2′(x)H⋯Hxp′(x). \theta(x) = x'_1(x) \mathbf{H} x'_2(x) \mathbf{H} \cdots \mathbf{H} x'_p(x).
Let h1h_1 be a vector subspace supplementary to hh in gg, the sum of the aka_k with k≠jk \neq j and a supplement of hh in aja_j. There exists a bijection η\eta of g′g' onto h1h_1 such that θ∘η=Idg′\theta \circ \eta = \mathrm{Id}_{g'}. For x∈gx \in g, we write (3)
ζ(x)=η(x1′(x))Hη(x2′(x))H⋯Hη(xp′(x))∈g. \zeta(x) = \eta(x'_1(x)) \mathbf{H} \eta(x'_2(x)) \mathbf{H} \cdots \mathbf{H} \eta(x'_p(x)) \in g.
(4)
y(x)=ζ(x)−1Hx=(−ζ(x))⋅x. y(x) = \zeta(x)^{-1} \mathbf{H} x = (-\zeta(x)) \cdot x.

By (2) and (3), θ(ζ(x))=θ(x)\theta(\zeta(x)) = \theta(x) and hence y(x)∈hy(x) \in \mathfrak{h}. Finally we write

(5)ψ(x)=(η(x1′(x)),…,η(xp′(x))+y(x),…,η(xp′(x)))∈a1×⋯×ap. (5) \quad \psi(x) = (\eta(x_1'(x)), \ldots, \eta(x_p'(x)) + y(x), \ldots, \eta(x_p'(x))) \in a_1 \times \cdots \times a_p.
As y(x)y(x) is central in g\mathfrak{g},
ϕ(ψ(x))=η(x1′(x))H⋯Hη(xp′(x))H⋯Hη(xp′(x))Hy(x)=ζ(x)Hy(x)by (3)=xby (4). \begin{align*} \phi(\psi(x)) &= \eta(x_1'(x)) \mathbin{\mathbf{H}} \cdots \mathbin{\mathbf{H}} \eta(x_p'(x)) \mathbin{\mathbf{H}} \cdots \mathbin{\mathbf{H}} \eta(x_p'(x)) \mathbin{\mathbf{H}} y(x) \\ &= \zeta(x) \mathbin{\mathbf{H}} y(x) & \text{by (3)} \\ &= x & \text{by (4).} \end{align*}
Hence ϕ∘ψ=Idg\phi \circ \psi = \mathrm{Id}_g. Now let (x1,x2,…,xp)∈a1×a2×⋯×ap(x_1, x_2, \ldots, x_p) \in a_1 \times a_2 \times \cdots \times a_p and write x=ϕ(x1,x2,…,xp)=x1Hx2H⋯Hxpx = \phi(x_1, x_2, \ldots, x_p) = x_1 \mathbin{\mathbf{H}} x_2 \mathbin{\mathbf{H}} \cdots \mathbin{\mathbf{H}} x_p. Then
θ(x)=θ(x1)Hθ(x2)H⋯Hθ(xp), \theta(x) = \theta(x_1) \mathbin{\mathbf{H}} \theta(x_2) \mathbin{\mathbf{H}} \cdots \mathbin{\mathbf{H}} \theta(x_p),
hence xi′(x)=θ(xi)x_i'(x) = \theta(x_i) for 1≤i≤p1 \leq i \leq p and therefore
ζ(x)=x1Hx2H⋯H(ηθ(xj))H⋯Hxpy(x)=xj−ηθ(xj). \begin{align*} \zeta(x) &= x_1 \mathbin{\mathbf{H}} x_2 \mathbin{\mathbf{H}} \cdots \mathbin{\mathbf{H}} (\eta \theta(x_j)) \mathbin{\mathbf{H}} \cdots \mathbin{\mathbf{H}} x_p \\ y(x) &= x_j - \eta \theta(x_j). \end{align*}
Then by (5)
ψ(x)=(x1,…,ηθ(xj)+xj−ηθ(xj),…,xp)=(x1,x2,…,xp). \psi(x) = (x_1, \ldots, \eta \theta(x_j) + x_j - \eta \theta(x_j), \ldots, x_p) = (x_1, x_2, \ldots, x_p).
Hence ψ∘ϕ=Ida1×…×ap\psi \circ \phi = \mathrm{Id}_{a_1 \times \ldots \times a_p}. This proves (i). As the Hausdorff law is polynomial, ϕ\phi is polynomial. By the induction hypothesis, ϕ′−1{\phi'}^{-1} is polynomial; by formula (1), the functions xj′x_j' are polynomial, hence ζ\zeta is polynomial (formula (3)), yy is polynomial (formula (4)) and ψ\psi is polynomial (formula (5)). This proves (ii). Assertion (iii) follows from (i) and (ii) and the fact that the Hausdorff law is polynomial.

Example of a nilpotent Lie group. Let GG be the lower strict triangular subgroup of GL(n,K)\mathbf{GL}(n, K). It is a Lie subgroup of GL(n,K)\mathbf{GL}(n, K) and L(G)⊂gl(n,K)L(G) \subset \mathfrak{gl}(n, K) is the Lie algebra of lower triangular matrices with zero diagonal (§ 3, no. 10, Proposition 36). By Chapter II, § 4, no. 6, Remark, GG is nilpotent. Suppose henceforth that K=RK = \mathbf{R} or C\mathbf{C}. As GG is homeomorphic to Kn(n−1)/2K^{n(n-1)/2}, GG is simply connected. The exponential mapping of L(G)L(G) into GG is just the mapping

u↦exp⁡u=∑k≥0ukk!=∑k=0n−1ukk! u \mapsto \exp u = \sum_{k \geq 0} \frac{u^k}{k!} = \sum_{k=0}^{n-1} \frac{u^k}{k!}
(§ 6, no. 4, Example). By Proposition 13, the exponential is an isomorphism of the manifold L(G)L(G) onto the manifold GG. Proposition 17 of § 6, no. 9 gives the inverse bijection log⁡\log. We give KnK^n a norm. By Spectral Theories, Chapter I, § 4, no. 9, for g∈Gg \in G and ∥g−1∥<1\|g - 1\| < 1,
log⁡g=∑k≥1(−1)k−1k(g−1)k \log g = \sum_{k \geq 1} \frac{(-1)^{k-1}}{k} (g - 1)^k

that is

log⁡g=∑k=1n−1(−1)k−1k(g−1)k. \log g = \sum_{k=1}^{n-1} \frac{(-1)^{k-1}}{k} (g-1)^k.
But the two sides of (6) are analytic functions of gg for g∈Gg \in G and are therefore equal for all g∈Gg \in G.

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

Cited by

Nothing.

Cites

Transcribed from the printed volume. Not translated.