The Z-linear mapping (7) is bijective.
We set P=lim(Eα⊗AαFα) and, for all α∈I, let hα:Eα⊗AαFα→P be the canonical mapping. On the other hand, for all α∈I, let
tα:Eα×Fα→Eα⊗AαFα be the canonical
Z-bilinear mapping; for
α≤β,
tβ(fβα(xα),gβα(yα))=fβα(xα)⊗gβα(yα)=hβα(tα(xα,yα)) and hence
(tα) is a direct system of mappings. Canonically identifying
lim(Eα×Fα) with
E×F (
Set Theory, III, § 7, no. 7, Proposition 10), we derive a mapping
t=limtα:E×F→P such that
t(fα(xα),gα(yα))=hα(tα(xα,yα))=hα(xα⊗yα). Taking account of
Set Theory, III, § 7, no. 5, Lemma 1, it is immediately seen that
t is
Z-bilinear; moreover, for
x∈E,y∈F,λ∈A, there exists
α∈I such that
x=fα(xα),y=gα(yα),λ=ρα(λα) with
λα∈Aα,xα∈Eα,yα∈Fα (
Set Theory, III, § 3, no. 7, Lemma 1); whence
t(xλ,y)=hα((xαλα)⊗yα)=hα(xα⊗(λαyα))=t(x,λy). Hence there exists one and only one
Z-linear mapping
π′:E⊗AF→P such that
π′(x⊗y)=t(x,y) (§ 3, no. 1, Proposition 1). Moreover, by definition
π′(π(hα(xα⊗yα)))=π′(fα(xα)⊗gα(yα))=hα(xα⊗yα) π(π′(fα(xα)⊗gα(yα)))=π(hα(xα⊗yα))=fα(xα)⊗gα(yα) and as the elements of the form
fα(xα)⊗gα(yα) (resp.
hα(xα⊗yα)) generate the
Z-module
E⊗AF (resp.
P),
π′∘π and
π∘π′ are the identity mappings.
Loosely speaking, Proposition 7 may be expressed by saying that tensor products commute with direct limits and usually the two sides of (7) are identified by means of the isomorphism π.
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.