Let (aσ)σ∈Γ be a family of elements of N. For b to exist in N such that aσ=σ(b)−b for all σ∈Γ it is necessary and sufficient that aστ=aσ+σ(aτ) for σ,τ in Γ.
We have στ(b)−b=[σ(b)−b]+σ[τ(b)−b] for all b in N and σ,τ in Γ, whence the necessity.
Conversely suppose that aστ=aσ+σ(aτ) for any σ and τ in Γ. Put Uσ=(10aσ1) for σ∈Γ; then we have Uστ=Uσ⋅σ(Uτ) for σ,τ in Γ; by Prop. 9, there exists thus a matrix A=(xzyt) with non-zero determinant such that σ(A)=AUσ for all a∈Γ; writing down the relation σ(A)=AUσ we find
(σ(x)σ(z)σ(y)σ(t))=(xzxaσ+yzaσ+t)(σ∈Γ). In particular,
x and
z belong to
K and we have
σ(y)=xaσ+y,σ(t)=zaσ+t(σ∈Γ). If
x=0 we have
a,=σ(b)−b with
b=x−1y; if
z=0, we have the same relation with
b=z−1t. Now
x and
z cannot both be zero because
xt−yz=detA=0.
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.