Bourbaki

4. Product of topological spaces

1. Product spaces

Definition 1

Given a family (Xi)i∈I(X_i)_{i \in I} of topological spaces, the product space of this family is the product set X=∏i∈IXiX = \prod_{i \in I} X_i with the topology which is the product of the topologies of the XiX_i (§ 2, no. 3, Example 3). The spaces Xi(i∈I)X_i (i \in I) are called the factors of X.

By virtue of § 2, no. 3, Proposition 4, the product topology on XX has as a base the set B\mathfrak{B} of finite intersections of sets of the form pr‾i−1(Ui)\overline{\mathrm{pr}}_i^{-1}(U_i), where UiU_i is open in XiX_i; these sets are products ∏i∈IAi\prod_{i \in I} A_i, where AiA_i is open in XiX_i for each i∈Ii \in I and Ai=XiA_i = X_i for all but a finite number of indices. These sets will be called elementary sets.

If Bi\mathfrak{B}_i is a base of the topology of XiX_i (for each i∈Ii \in I), it is clear that the elementary sets ∏i∈IAi\prod_{i \in I} A_i such that Ai∈BiA_i \in \mathfrak{B}_i for each index ii such that Ai≠XiA_i \neq X_i form another base of the product topology. The elementary sets of this type which contain a given point x∈Xx \in X thus form a fundamental system of neighbourhoods of xx (§1§ 1, no. 3, Proposition 3).

If II is a finite set, the construction of the product topology from the topologies of the factors XiX_i is simpler: the elementary sets are just products ∏i∈IAi\prod_{i \in I} A_i, where AiA_i is any open subset of XiX_i, for each i∈Ii \in I (cf. Exercise 9).

Example

\* The product Rn\mathbf{R}^n of nn spaces identical with the real line R\mathbf{R} is called real number space of nn dimensions; R2\mathbf{R}^2 is also called the real plane (cf. Chapter VI, § 1). Likewise, starting from the rational line Q\mathbf{Q}, we define the rational number space of nn dimensions Qn\mathbf{Q}^n (rational plane for n=2n = 2).

The topology of the space Rn\mathbf{R}^n has as a base the set of all products of nn open intervals in R\mathbf{R}, which are called open boxes of nn dimensions. The open boxes which contain a point x∈Rnx \in \mathbf{R}^n form a fundamental system of neighbourhoods of this point. Likewise the products of nn closed intervals in R\mathbf{R} are called closed boxes of nn dimensions. The closed boxes to which xx is interior also form a fundamental system of neighbourhoods of xx. There are analogous results for Qn\mathbf{Q}^n. \

Proposition 1

Let f=(fi)f = (f_i) be a mapping of a topological space YY into a product space X=∏i∈IXiX = \prod_{i \in I} X_i. Then ff is continuous at a point a∈Ya \in Y if and only if fif_i is continuous at aa for each i∈Ii \in I.

Since fi=pri∘ff_i = \mathrm{pr}_i \circ f, this is just a particular case of Proposition 4 of § 2, no. 3.

Corollary 1

Let (Xi)i∈I,(Yi)i∈I(X_i)_{i \in I}, (Y_i)_{i \in I} be two families of topological spaces with the same set of indices. For each i∈Ii \in I, let fif_i be a mapping of XiX_i into YiY_i. In order that the product mapping f:(xi)→(fi(xi))f : (x_i) \to (f_i(x_i)) of

∏i∈IXi into ∏i∈IYi \prod_{i \in I} X_i \text{ into } \prod_{i \in I} Y_i
should be continuous at a point a=(ai)a = (a_i) it is necessary and sufficient that fif_i is continuous at aia_i for each i∈Ii \in I.

f can be written as x→(fi(pri(x)))x \to (f_i(\mathrm{pr}_i(x))), so that by Proposition 1 the condition is sufficient. Conversely, for each x∈Ix \in I let gxg_x be the mapping of XxX_x into ∏i∈IXi\prod_{i \in I} X_i such that prx(gx(xx))=xx\mathrm{pr}_x(g_x(x_x)) = x_x and pri(gx(xx))=ai\mathrm{pr}_i(g_x(x_x)) = a_i whenever i≠xi \neq x; then gxg_x is continuous at the point axa_x, by Proposition 1. Since fx=prx∘f∘gxf_x = \mathrm{pr}_x \circ f \circ g_x it follows that if ff is continuous at aa, then fxf_x is continuous at axa_x.

Corollary 2

Let X,YX, Y be two topological spaces. In order that a mapping f:X→Yf : X \to Y should be continuous it is necessary and sufficient that the mapping g:x→(x,f(x))g : x \to (x, f(x)) is a homeomorphism of XX onto the graph GG of ff (considered as a subspace of the product space X×YX \times Y).

Since f=pr2∘gf = \mathrm{pr}_2 \circ g, the condition is sufficient. It is also necessary, for if ff is continuous, then gg is bijective and continuous (Proposition 1) and the inverse of gg is the restriction of pr1\mathrm{pr}_1 to GG, which is continuous (cf. Set Theory, Chapter IV, § 2, no. 4, criterion CST 17).

Proposition 2 (Associativity of topological products)

Let (Xi)i∈I(X_i)_{i \in I} be a family of topological spaces, (Jx)x∈K(J_x)_{x \in K} a partition of the set II, and for each x∈Kx \in K let Xx′=∏i∈JxXiX'_x = \prod_{i \in J_x} X_i be the product of the spaces XiX_i for i∈JKi \in J_K. Then the canonical mapping of the product space*

∏i∈IXi onto the product space ∏x∈KXx′ \prod_{i \in I} X_i \text{ onto the product space } \prod_{x \in K} X'_x
*is a homeomorphism.

This is a particular case of transitivity of initial topologies (§\S 2, no. 3, Proposition 5; cf. Set Theory, Chapter IV, § 2, no. 4, criterion CST 13).

Generally we identify the product spaces ∏i∈IXi\prod_{i \in I} X_i and ∏x∈KXx′\prod_{x \in K} X'_x by means of the canonical mapping.

Corollary

Let σ\sigma be a permutation of the set II. Then the mapping (xi)→(xσ(i))(x_i) \to (x_{\sigma(i)}) is a homeomorphism of*

∏i∈IXi onto ∏i∈IXσ(i). \prod_{i \in I} X_i \text{ onto } \prod_{i \in I} X_{\sigma(i)}.

Take K=IK = I and Ji={σ(i)}J_i = \{ \sigma(i) \} for each i∈Ii \in I in Proposition 2.

Proposition 3

Let XX be a set, (Yi)i∈I(Y_i)_{i \in I} a family of topological spaces, and for each i∈Ii \in I let fif_i be a mapping of XX into YiY_i. Let ff be the mapping x→(fi(x))x \to (f_i(x)) of XX into Y=∏i∈IYiY = \prod_{i \in I} Y_i, and let T\mathfrak{T} be the coarsest topology on XX for which the mappings fif_i are continuous. Then T\mathfrak{T} is the inverse image under ff of the topology induced on f(X)f(X) by the product topology on YY.

This is another particular case of transitivity of initial topologies (\S\ 2, no. 3, Proposition 5; cf. Set Theory, Chapter IV, \S\ 2, no. 4, criterion CST 15).

Corollary

For each i∈Ii \in I let AiA_i be a subspace of YiY_i. Then the topology induced on A=∏i∈IAiA = \prod_{i \in I} A_i by the product topology on ∏i∈IYi\prod_{i \in I} Y_i is the product of the topologies of the subspaces AiA_i.

Let jij_i be the canonical injection Ai→YiA_i \to Y_i, and apply Proposition 3 to the mappings fi=ji∘prif_i = j_i \circ \mathrm{pr}_i (cf. Set Theory, Chapter IV, \S\ 2, no. 4, criterion CST 14).

2. Section of an open set; section of a closed set; projection of an open set. Partial continuity

Proposition 4

Let X1,X2X_1, X_2 be two topological spaces; then for each a1∈X1a_1 \in X_1, the mapping x2→(a1,x2)x_2 \to (a_1, x_2) is a homeomorphism of X2X_2 onto the subspace {a1}×X2\{a_1\} \times X_2 of X1×X2X_1 \times X_2.

This is a particular case of Corollary 1 of Proposition 1 of no. 1 applied to the constant function x2→a1x_2 \to a_1.

The mapping x2→(a1,x2)x_2 \to (a_1, x_2) is a continuous section (\S\ 3, no. 5) with respect to the equivalence relation pr2z=pr2z′\mathrm{pr}_2 z = \mathrm{pr}_2 z' in X1×X2X_1 \times X_2; the quotient space X1×X2X_1 \times X_2 by this equivalence relation is therefore homeomorphic to X2X_2.

Corollary

The section A(x1)A(x_1) of an open (resp. closed) set AA of the product X1×X2X_1 \times X_2 at an arbitrary point x1∈X1(∗)x_1 \in X_1 (*) is open (resp. closed) in X2X_2.

Proposition 5

The projection of an open set UU of the product X1×X2X_1 \times X_2 onto either factor is an open set.

For example, we have pr2U=⋃x1∈X1U(x1)\mathrm{pr}_2 U = \bigcup_{x_1 \in X_1} U(x_1), and the proposition follows from the Corollary to Proposition 4 and axiom (O_1).

Remark 1

The projection of a closed subset of X1×X2X_1 \times X_2 onto a factor need not be closed. For example, in the rational plane Q2\mathbf{Q}^2, the hyperbola whose equation is x1x2=1x_1 x_2 = 1 is a closed set, but both its projections are equal to the complement of the point 0 in Q\mathbf{Q}, and this is not a closed set.

(*) By the section A(x1)A(x_1) of AA at x1x_1 is meant the set of all x2∈X2x_2 \in X_2 such that (x1,x2)∈A(x_1, x_2) \in A (cf. Set Theory R, \S\ 3, no. 7).

Proposition 6

Let X1,X2,YX_1, X_2, Y be three topological spaces, ff a mapping of the product space X1×X2X_1 \times X_2 into YY. If ff is continuous at the point (a1,a2)(a_1, a_2) then the partial mapping x2→f(a1,x2)x_2 \to f(a_1, x_2) of X2X_2 into YY is continuous at the point a2a_2.

For this mapping is the composition of ff and the mapping x2→(a1,x2)x_2 \to (a_1, x_2); hence the result follows from Proposition 4.

Proposition 6 is often expressed by saying that a continuous function of two variables is continuous with respect to each of them separately.

Remark 2

It is possible for all the partial mappings determined by a map f:X1×X2→Yf : X_1 \times X_2 \to Y to be continuous without ff being continuous on X1×X2X_1 \times X_2 (cf. Chapter IX, § 5, Exercise 23). \* For example if ff is the mapping of the real plane R2\mathbf{R}^2 into R\mathbf{R} defined by

f(x,y)=xy/(x2+y2)if(x,y)≠(0,0) f(x, y) = xy/(x^2 + y^2) \quad \text{if} \quad (x, y) \neq (0, 0)
and f(0,0)=0f(0, 0) = 0, then all the partial mappings are continuous; but ff is not continuous at (0,0)(0, 0), since f(x,x)=1/2f(x, x) = 1/2 if x≠0x \neq 0.

If gg is a mapping of X1X_1 into YY, continuous at a point a1a_1, then the mapping (x1,x2)→g(x1)(x_1, x_2) \to g(x_1) of X1×X1→YX_1 \times X_1 \to Y is continuous at all points (a1,x2)(a_1, x_2), for it is the composition of gg and the projection onto X1X_1.

The results of this subsection are easily extended to an arbitrary product ∏i∈IXi\prod_{i \in I} X_i of topological spaces by remarking that this product is homeomorphic to the product (∏i∈JXi)×(∏i∈KXi)\left( \prod_{i \in J} X_i \right) \times \left( \prod_{i \in K} X_i \right) for any partition (J,K)(J, K) of II (no. 1, Proposition 2).

3. Closure in a product

Proposition 7

In a product space ∏i∈IXi\prod_{i \in I} X_i, the closure of a product of sets ∏i∈IAi\prod_{i \in I} A_i is the same as the product ∏i∈IA‾i\prod_{i \in I} \overline{A}_i of their closures.

Suppose that a=(ai)a = (a_i) lies in the closure of ∏i∈IAi\prod_{i \in I} A_i; then for each x∈Ix \in I, ax=pr⁡xaa_x = \operatorname{pr}_x a is in the closure of AxA_x because of the continuity of pr⁡x\operatorname{pr}_x (§ 2, no. 1, Theorem 1) and therefore a∈∏i∈IA‾ia \in \prod_{i \in I} \overline{A}_i. Conversely, let b=(bi)∈∏i∈IA‾ib = (b_i) \in \prod_{i \in I} \overline{A}_i, and let ∏i∈IVi\prod_{i \in I} V_i be any elementary set containing bb; for each i∈Ii \in I, ViV_i contains a point xi∈Aix_i \in A_i; hence ∏i∈IVi\prod_{i \in I} V_i contains the point (xi)∈∏i∈IAi(x_i) \in \prod_{i \in I} A_i and therefore bb lies in the closure of ∏i∈IAi\prod_{i \in I} A_i.

Corollary

A product* ∏i∈IAi\prod_{i \in I} A_i of non-empty sets is closed in the product space ∏i∈IXi\prod_{i \in I} X_i if and only if AiA_i is closed in XiX_i for each i∈Ii \in I.

If II is finite, a product ∏i∈IAi\prod_{i \in I} A_i is open provided that AiA_i is open in XiX_i for each i∈Ii \in I; but this is no longer so if II is infinite.

Proposition 8

Let* a=(ai)a = (a_i) be any point of a product space X=∏i∈IXiX = \prod_{i \in I} X_i; then the set DD of points x∈Xx \in X such that prix=ai\mathrm{pr}_i x = a_i except for a finite number of indices ii is dense in XX.

For each x∈Xx \in X and each elementary set V=∏i∈IUiV = \prod_{i \in I} U_i which contains xx, we have Ui=XiU_i = X_i except for indices ii belonging to a finite subset JJ of II; if we take yi=xiy_i = x_i for i∈Ji \in J and yi=aiy_i = a_i for i∉Ji \notin J, it is clear that

y=(yi)∈Dandy∈V; y = (y_i) \in D \quad \text{and} \quad y \in V;
hence the result.

4. Inverse limits of topological spaces

Let II be a partially ordered (but not necessarily directed) set (*), in which the order relation is written α≤β\alpha \leq \beta. For each α∈I\alpha \in I, let XαX_\alpha be a topological space, and for each pair (α,β)(\alpha, \beta) such that α≤β\alpha \leq \beta let fαβf_{\alpha \beta} be a mapping of XβX_\beta into XαX_\alpha. We say that (Xα,fαβ)(X_\alpha, f_{\alpha \beta}) is an inverse system of topological spaces if: 1) (Xα,fαβ)(X_\alpha, f_{\alpha \beta}) is an inverse system of sets; 2) fαβf_{\alpha \beta} is a continuous mapping whenever α≤β\alpha \leq \beta. Let XX denote the set lim←⁡Xα\varprojlim X_\alpha, and for each α∈I\alpha \in I let fαf_\alpha be the canonical mapping X→XαX \to X_\alpha; then the coarsest topology on XX for which the fαf_\alpha are continuous is said to be the inverse limit (with respect to the fαβf_{\alpha \beta}) of the topologies of the XαX_\alpha, and the set XX with this topology is called the inverse limit of the inverse system of topological spaces (Xα,fαβ)(X_\alpha, f_{\alpha \beta}). Whenever we speak of lim←⁡Xα\varprojlim X_\alpha as a topological space, it is always to be understood that the topology of this space is the inverse limit of the topologies of the XαX_\alpha, unless the contrary is expressly stated.

The set XX is the subset of the product ∏α∈IXα\prod_{\alpha \in I} X_\alpha consisting of those points xx such that

(1)(1)
prα(x)=fαβ(prβ(x)) \mathrm{pr}_\alpha(x) = f_{\alpha \beta}(\mathrm{pr}_\beta(x))
whenever α≤β\alpha \leq \beta. It follows from Proposition 3 of no. 1 that the inverse limit of the topologies of the XαX_\alpha is the same as the topology induced on XX by the topology of the product space ∏α∈IXα\prod_{\alpha \in I} X_\alpha. If, for each α∈I\alpha \in I, YαY_\alpha is

() That is, a set endowed with a reflexive and transitive relation (Set Theory* R, § 6, no. 1).

a subspace of XαX_\alpha such that the YαY_\alpha form an inverse system of subsets of the XαX_\alpha (Set Theory chapter III, , n° 2), then it is clear that the topological space lim←⁡Yα\varprojlim Y_\alpha is a subspace of lim←⁡Xα\varprojlim X_\alpha

Let (Xα′,fαβ′)(X'_\alpha, f'_{\alpha\beta}) be another inverse system of topological spaces indexed by the same set II, and for each α∈I\alpha \in I let uα:Xα→Xα′u_\alpha : X_\alpha \to X'_\alpha be a continuous mapping such that (uα)(u_\alpha) is an inverse system of mappings; then u=lim←⁡uαu = \varprojlim u_\alpha is a continuous mapping of X=lim←⁡XαX = \varprojlim X_\alpha into X′=lim←⁡Xα′X' = \varprojlim X'_\alpha. For if fα′f'_\alpha is the canonical mapping X′→Xα′X' \to X'_\alpha, we have fα′∘u=uα∘fαf'_\alpha \circ u = u_\alpha \circ f_\alpha, so that fα′∘uf'_\alpha \circ u is continuous for each α∈I\alpha \in I, and the assertion follows from Proposition 4 of § 2, no. 3.

Finally, suppose II is a directed set, and let JJ be a cofinal subset of II; let ZZ be the inverse limit of the inverse system of topological spaces (Xα,fαβ)α∈J,β∈J(X_\alpha, f_{\alpha\beta})_{\alpha \in J, \beta \in J}. Then the canonical bijection g:X→Zg : X \to Z (Set Theory, Chapter III, § 7, no. 2, proposition 3) is a homeomorphism. For we have prλ(g(x))=prλ(x)\mathrm{pr}_\lambda(g(x)) = \mathrm{pr}_\lambda(x) for each λ∈J\lambda \in J; hence gg is continuous (no. 1, Proposition 1); and if hh is the inverse of gg, then for each α∈I\alpha \in I there exists λ∈J\lambda \in J such that α≤λ\alpha \leq \lambda, and therefore prα(h(z))=fαλ(prλ(z))\mathrm{pr}_\alpha(h(z)) = f_{\alpha\lambda}(\mathrm{pr}_\lambda(z)), which shows that hh is continuous (no. 1, Proposition 1), since the fαλf_{\alpha\lambda} are continuous.

Proposition 9

Let II be a directed set and JJ a cofinal subset of II. Let (Xα,fαβ)(X_\alpha, f_{\alpha\beta}) be an inverse system of topological spaces indexed by II; let X=lim←⁡XαX = \varprojlim X_\alpha and let fα:X→Xαf_\alpha : X \to X_\alpha be the canonical mapping. Then the family of sets fα−1(Uα)f^{-1}_\alpha(U_\alpha), where α\alpha runs through JJ and UαU_\alpha runs through a base Bα\mathfrak{B}_\alpha of the topology of XαX_\alpha for each α∈J\alpha \in J, is a base of the topology of XX.

From § 2, no. 3 we know that the finite intersections of sets of the form fα−1(Uα)f^{-1}_\alpha(U_\alpha) (α∈I,Uα\alpha \in I, U_\alpha open in XαX_\alpha) form a base of the topology of XX. If (αi)1≤i≤n(\alpha_i)_{1 \leq i \leq n} is a finite family of indices of II, then there exists γ∈J\gamma \in J such that αi≤γ\alpha_i \leq \gamma for 1≤i≤n1 \leq i \leq n; hence fαi=fαiγ∘fγf_{\alpha_i} = f_{\alpha_i\gamma} \circ f_\gamma; if we put

Vγ=⋂ifαiγ−1(Uαi), V_\gamma = \bigcap_i f^{-1}_{\alpha_i\gamma}(U_{\alpha_i}),

then we have

fγ−1(Vγ)=⋂ifαi−1(Uαi); f^{-1}_\gamma(V_\gamma) = \bigcap_i f^{-1}_{\alpha_i}(U_{\alpha_i});

but VγV_\gamma is open and is therefore a union of sets belonging to Bγ\mathfrak{B}_\gamma. Hence the result.

Corollary

Let AA be a subset of XX and let AαA_\alpha denote fα(A)f_\alpha(A) for each α∈I\alpha \in I. Then:

(i) The AαA_\alpha (resp. the A‾α\overline{A}_\alpha) form an inverse system of subsets of the XαX_\alpha, and

A‾=⋂αfα−1(A‾α)=lim←⁡A‾α. \overline{A} = \bigcap_\alpha f^{-1}_\alpha(\overline{A}_\alpha) = \varprojlim \overline{A}_\alpha.

(ii) If AA is closed in XX, A=lim←⁡Aα=lim←⁡A‾αA = \varprojlim A_\alpha = \varprojlim \overline{A}_\alpha.

The first assertion of (i) follows from the relations fα=fαβ∘fβf_\alpha = f_{\alpha\beta} \circ f_\beta for α≤β\alpha \leq \beta and from the continuity of the fαβf_{\alpha\beta} (§ 2, no. 1, Theorem 1). Let A′A' denote

⋂αf‾α(A‾α); \bigcap_\alpha \overline{f}_\alpha(\overline{A}_\alpha);
then it is clear that A′A' is closed and contains AA, so that A‾⊂A′\overline{A} \subset A'. Conversely, let x∈A′x \in A'; we have to show that xx lies in the closure of AA. By virtue of Proposition 9 it is enough to prove that every neighbourhood of xx which is of the form f‾α(Uα)\overline{f}_\alpha(U_\alpha), with α∈I\alpha \in I and UαU_\alpha open in XαX_\alpha, meets AA. Now, by hypothesis, fα(x)∈Uαf_\alpha(x) \in U_\alpha, and since fα(x)∈A‾f_\alpha(x) \in \overline{A} we have Uα∩Aα≠∅U_\alpha \cap A_\alpha \neq \varnothing, which means that A∩f‾α(Uα)A \cap \overline{f}_\alpha(U_\alpha) is not empty.

To establish (ii) it is enough to remark that, without any restriction on AA, we have A⊂lim⁡Aα⊂lim⁡A‾αA \subset \lim A_\alpha \subset \lim \overline{A}_\alpha; now if AA is closed, then from (i)

A=lim⁡A‾α A = \lim \overline{A}_\alpha
and (ii) follows.

Example

Let II be a directed set and (Xα)α∈I(X_\alpha)_{\alpha \in I} a family of subsets of a set YY, such that Xα⊃XβX_\alpha \supset X_\beta whenever α≤β\alpha \leq \beta. For each α∈I\alpha \in I let Tα\mathfrak{T}_\alpha be a topology on XαX_\alpha such that Tβ\mathfrak{T}_\beta is finer than the topology induced on XβX_\beta by Tα\mathfrak{T}_\alpha whenever α≤β\alpha \leq \beta. If we take fαβf_{\alpha\beta} to be the canonical injection Xβ→XαX_\beta \to X_\alpha for α≤β\alpha \leq \beta, then lim⁡Xα\lim X_\alpha may be identified canonically with the intersection XX of the XαX_\alpha, with the topology which is the least upper bound (§ 2, no. 3, Example 2) of the topologies induced on XX by the Tα\mathfrak{T}_\alpha.

Exercises

See the exercises for § 4.

Transcribed from the printed volume. Not translated.