Let G be a Lie group and (U,ϕ,E) a chart on G such that ϕ(e)=0. Let V be an open neighbourhood of e such that V2⊂U. Let m be the analytic mapping (a,b)↦ϕ(ϕ−1(a)ϕ−1(b)) of ϕ(V)×ϕ(V) into E. Let
m=i,j≥0∑mi,j
be the expansion of m as an integral series about (0, 0), where mi,j is a bihomogeneous continuous-polynomial of bidegree (i,j) on E×E with values in E.
(i) mi,0=m0,j=0 for all i=1 and j=1. (ii) m1,0(a,b)=a and m0,1(a,b)=b for all a∈E,b∈E. (iii) *Let ψ:Te(G)→E be the differential of ϕ at e. For all u,v in Te(G),
ψ([u,v])=m1,1(ψ(u),ψ(v))−m1,1(ψ(v),ψ(u)).
m(a,0)=a, m(0,b)=b for all a,b in ϕ(V), which proves (i) and (ii). Let u,v be in Te(G). Let T0(E) be identified with E and hence ψ with Te(ϕ): The images of u and v under Te(ϕ) are ψ(u) and ψ(v). The tensor product point distribution of these images is the symmetric product of (ψ(u),0) and (0,ψ(v)) in TS(E×E)=TS(E)×TS(E), that is
(ψ(u),0)⊗(0,ψ(v))+(0,ψ(v))⊗(ψ(u),0).
Hence ϕ∗(u∗v) is the image of the above element under the mapping m of ϕ(V)×ϕ(V) into E. Its component of degree 1 in TS(E) is, by Lemma 1,
x=⟨m1,1,(ψ(u),0)⊗(0,ψ(v))+(0,ψ(v))⊗(ψ(u),0)⟩.
We define a bilinear mapping n:(E×E)2→E by
n((a,b),(a′,b′))=m1,1(a,b′).
Then n((a,b),(a,b))=m1,1(a,b) and hence
x=⟨n,(ψ(u),0)⊗(0,ψ(v))+(0,ψ(v))⊗(ψ(u),0)⟩=m1,1(ψ(u),ψ(v))+m1,1(0,0)=m1,1(ψ(u),ψ(v)).
Similarly, ϕ∗(v∗u) admits m1,1(ϕ(v),ϕ(u)) as component of degree 1 in TS(E). As ϕ([u,v]) is of degree 1, this proves (iii).
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.