Bourbaki

Proposition 6

Tag 01OO, permanent. Cite it as [Bourbaki, Tag 01OO].

Topologie algébrique, chapter I, § 3. Faisceaux

Let B be a topological space, F\mathscr{F} a sheaf on B and (Ai)i∈I(A_i)_{i\in I} a locally finite closed covering of B. In order that the sheaf F\mathscr{F} be soft, it is necessary and sufficient that, for every i∈Ii\in I, the induced sheaf FAi\mathscr{F}_{A_i} be soft.

The condition is obviously necessary. Let us prove that it is sufficient. Let us denote by p:E→Bp: E\rightarrow B the étale B-space EFE_{\mathscr{F}} associated with the sheaf F\mathscr{F}. Let A be a closed subspace of B and s:A→Es: A\rightarrow E a continuous section of pp over A; it is a question of proving that ss possesses a continuous extension to B. For every subset J of I, put AJ=⋃i∈JAiA_J=\bigcup_{i\in J}A_i; the set AJA_J is closed in B (TG, I, p. 6, prop. 4).

Let S\mathscr{S} be the set of pairs (J,t)(J, t) where J is a subset of I and tt a continuous section of E over AJA_J which coincides with ss in A∩AJA\cap A_J. Let us make S\mathscr{S} into an ordered set by the ordering relation denoted by ⩽\leqslant for which (J,t)⩽(J′,t′)(J, t)\leqslant (J', t') if J⊂J′J\subset J' and t′∣AJ=tt'|A_J=t. For σ=(J,t)∈S\sigma = (J, t)\in \mathscr{S}, we write Jσ=JJ_{\sigma}= J and tσ=tt_{\sigma}=t. Let us show that the ordered set S\mathscr{S} is inductive. Let S be a totally ordered subset of S\mathscr{S}. Put J=⋃σ∈SJσJ =\bigcup_{\sigma\in S}J_{\sigma}; this is a subset of I. One then defines a section tt of E over AJA_J by setting t(x)=tσ(x)t(x) =t_{\sigma}(x), if x∈AJσx\in A_{J_{\sigma}}; thus t∣A∩AJ=st|A\cap A_J=s. Let j∈Jj\in J and let σ∈S\sigma \in S be such that j∈Jσj\in J_{\sigma}; since t∣Aj=tσ∣Ajt|A_j=t_{\sigma}|A_j, the restriction of tt to AjA_j is continuous. It then follows from TG, I, p. 19, Prop. 4 that tt is continuous. Thus, (J,t)(J, t) is an element of S\mathscr{S}; by construction, it is an upper bound of S. This proves that the set S\mathscr{S} is inductive. It therefore possesses a maximal element (J,t)(J, t) (E, III, p. 20, th. 2).

Let us reason by contradiction, supposing that J =I̸= I\not . Let ii be an element of I-J. Put A′=(Ai∩A)∪(Ai∩AJ)A'= (A_i\cap A)\cup (A_i\cap A_J) and define a section s′s' of E over A′A' by:

′s(a)'s(a) for a∈Ai∩Aa\in A_i\cap A,

s(a)= s(a) =

t(a)t(a) for a∈Ai∩AJa\in A_i\cap A_J,

which is possible since ss and tt coincide on A∩AJA\cap A_J. Moreover, since Ai∩AA_i\cap A and Ai∩AJA_i\cap A_J are closed, the section s′s' is continuous (TG, I, p. 19, prop. 4). By hypothesis, there exists a continuous section si:Ai→Es_i: A_i\rightarrow E extending s′s'. Since the restrictions of sis_i and tt to AJ∩AiA_J\cap A_i are equal, the continuous section t′:AJ∪{i}→Et': A_{J\cup \{i\}}\rightarrow E which coincides with tt on AJA_J and with sis_i on AiA_i is a continuous section of pp over AJ∪{i}A_{J\cup \{i\}}, extending s∣A∩AJ∪{i}s|A\cap A_{J\cup \{i\}}. We then have (J,t)<(J∪{i},t′)(J, t)<(J\cup \{i\}, t'), which contradicts the hypothesis that (J,t)(J, t) is maximal.

Thus, J = I, hence AJ=BA_J= B and tt is a continuous section of E over B extending ss.

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.

Cited by

Nothing.

Cites

Nothing.

Machine translation of the English, by gpt-5.4, gpt-5-6-mini, gpt-5-mini, against glossary version 34. Not checked by a person. That model is a cut down version of the one the rest of this language was written by.