Let(Eα,fβα)and(Eα′,fβα′)be two direct systems of sets relative to I, and for eachα∈Iletuαbe a mapping ofEαintoEα′such that theuαform a direct system of mappings. Letu=limuα. If eachuαis injective (resp. surjective) thenuis injective (resp. surjective).
Let E=limEα, E′=limEα′, and let fα:Eα→E, fα′:Eα′→E′ be the canonical mappings. Suppose that each uα is injective. To show that u is injective it is enough, by Proposition 6, to verify that if x∈Eα and y∈Eα are such that fα′(uα(x))=fα′(uα(y)), then there exists β⩾α such that fβα(x)=fβα(y). Now the hypothesis implies (no. 6, Lemma 1) that there exists β⩾α such that
With the notation of Proposition 7, let Mα be a subset of Eα for each α∈I; if we have fβα(Mα)∈Mβ whenever α⩽β, the family (Mα)α∈I is said to be a direct system of subsets of theEα. Let gβα (where α⩽β) be the mapping of Mα into Mβ whose graph is the same as that of the restriction of fβα to Mα. Then it is clear that (Mα,gβα) is a direct system of sets; and Proposition 7, applied to the canonical injections jα:Mα→Eα, allows us to identifyM=limMα with a subset of E by means of the injection j=limjα.
Bourbaki sets the proof directly after the statement, and the corpus does not separate them, so what is above is the statement and its proof together.