Bourbaki

§ 1. Fibred products and cartesian squares

1. Structure of a B-space

Let B be a topological space.

Definition 1 01L9

A topological B-space (or simply a B-space) is a topological space X, endowed with a continuous mapping pp of X into B. The mapping pp is called the projection of the B-space X.

Let X, X′X' be B-spaces and p,p′p,p' their respective projections. A B-morphism of X into X′X' is a continuous mapping ff of X into X′X' such that p′∘f=pp'\circ f=p.

It is sometimes convenient to denote by (X,p)(X, p) the B-space obtained by endowing the topological space X with the continuous mapping pp.

The composite of two B-morphisms is a B-morphism. The isomorphisms of B-spaces, also called B-isomorphisms, are the B-morphisms which are homeomorphisms.

Thus, if one calls a B-space structure on a set X the datum of a topology on X and of a continuous mapping p:X→Bp: X\rightarrow B, one can take the B-morphisms as morphisms of the B-space structure (E, IV, p. 11).

© N. Bourbaki and Springer-Verlag Berlin Heidelberg 2 016

N. Bourbaki, Algebraic Topology, DOI 10.1007/978-3-662-49361-8_1 1

Let X, X′X' be B-spaces. We denote by CB(X;X′)\mathscr{C}_B(X; X') the set of B-morphisms of X into X′X', by IsomB(X;X′)_B(X; X') the set of B-isomorphisms of X into X′X', and by AutB(X)_B(X) the set of B-automorphisms of X, that is to say of B-isomorphisms of X into X.

Let X be a B-space and pp its projection. If B is endowed with the B-space structure whose projection is IdB_B, the set CB(B;X)\mathscr{C}_B(B; X) is the set of continuous sections (E, II, p. 18, Definition 11) of pp.

Let X be a B-space, pp its projection and bb a point of B. The subspace p−1(b)\overset{-1}{p}(b) of X is called the fibre of X at bb (or the fibre of pp at bb) and denoted by XbX_b. In order that a continuous mapping ff of X into a B-space X′X' be a B-morphism, it is necessary and sufficient that f(Xb)f(X_b) be contained in Xb′X'_b for every b∈Bb\in B.

Let X be a B-space and pp its projection. Let ff be a continuous mapping of a topological space B′B' into B. A continuous mapping g:B′→Xg: B'\rightarrow X such that p∘g=fp\circ g=f is called a continuous lifting of ff to X. In other words, if B′B' is endowed with the B-space structure of projection ff, the continuous liftings of ff to X are the B-morphisms of B′B' into X.

2. Operations on B-spaces

Let B be a topological space.

Let X be a B-space and pp its projection. Every topological subspace Y of X is endowed with the B-space structure whose projection is p∣Yp|Y. Let A be a subspace of B; endowed with the mapping pA:p−1(A)→Ap_A:\overset{-1}{p}(A)\rightarrow A deduced from pp by passing to subsets, the topological space p−1(A)\overset{-1}{p}(A) is an A-space. It is called the A-space induced by (X,p)(X, p) over A and is sometimes denoted by XAX_A.

Let (Xi)i∈I(X_i)_{i\in I} be a family of B-spaces, and let pip_i be the projection of XiX_i. The sum space X=∐i∈IXiX =\coprod_{i\in I}X_i (TG, I, p. 15), endowed with the mapping p:X→p: X\rightarrow B defined by p(i,x)=pi(x)p(i, x) =p_i(x) (for i∈i\in I and x∈Xix\in X_i), is a B-space called the sum of the family of B-spaces (Xi)i∈I(X_i)_{i\in I}. The canonical injections Xi→XX_i\rightarrow X are B-morphisms.

Let X be a B-space, pp its projection, and let R be an equivalence relation on X. Let us denote by X/RX/R the quotient space (TG, I, p. 20, def. 3). If the mapping p:X→Bp: X\rightarrow B is compatible with the relation R (E, II, p. 44), the mapping p′:X/R→Bp': X/R\rightarrow B deduced from pp by passing to the quotient is continuous (TG, I, p. 21, prop. 6); the B-space obtained by endowing X/RX/R with the projection p′p' is then called the quotient B-space of X by the relation R.

3. Fibred product of two B-spaces

Let B be a topological space, X and X′X' B-spaces, pp and p′p' their respective projections. Let us denote by X×BX′X\times_BX' the topological subspace of X×X′X\times X' formed by the pairs (x,x′)(x, x') such that p(x)=p′(x′)p(x) =p'(x'). The mapping q:X×BX′→Bq: X\times_BX'\rightarrow B defined by q(x,x′)=p(x)q(x, x') =p(x) is continuous.

Definition 2 01LA

The topological space X×BX′X\times_BX' is called the fibred product of X and X′X' over B. The B-space obtained by endowing X×BX′X\times_BX' with the mapping qq is called the product B-space of X and X′X'.

The restrictions to X×BX′X\times_BX' of the projections of X×X′X\times X' into X and into X′X' are again denoted by pr1_1 and pr2_2 and called the first and second projections of the fibred product. They are continuous and are B-morphisms, for one has q=p∘q=p\circ pr1=p′∘_1=p'\circ pr2_2.

It should be noted that X×BX′X\times_BX' may be empty even if X and X′X' are non-empty: in fact, the relation X×BX′=∅X\times_BX'=\emptyset is equivalent to saying that p(X)p(X) and p′(X′)p'(X') are disjoint.

Let Y be a B-space and u:Y→X,u′:Y→X′u: Y\rightarrow X,u': Y\rightarrow X' B-morphisms. There exists a unique B-morphism v:Y→X×BX′v: Y\rightarrow X\times_BX' such that pr1∘v=u_1\circ v=u and pr2∘v_2\circ v = u′u' (universal property of the product B-space of two B-spaces): this is the mapping y↦(u(y),u′(y))y\mapsto (u(y), u'(y)) of Y into X×BX′X\times_BX', which is sometimes denoted by (u,u′)(u, u').

Let X,X′,Y,Y′X,X',Y,Y' be B-spaces, let f:X→Y,f′:X′→Y′f: X\rightarrow Y,f': X'\rightarrow Y' be B-morphisms. The mapping (x,x′)↦(f(x),f′(x′))(x, x')\mapsto (f(x), f'(x')) is a B-morphism of X×BX′X\times_BX' into Y×BY′Y\times_BY', which is denoted by f×Bf′f\times_Bf' and is called the extension of ff and f′f' to fibred products.

Example 1 01LB

Let X and X′X' be B-spaces; then the mapping (x,x′)↦(x′,x)(x, x')\mapsto (x', x) defines a B-isomorphism of X×BX′X\times_BX' onto X′×BXX'\times_BX.

Example 2 01LC

Let X, X′,X′′X',X'' be B-spaces and let p,p′,p′′p,p',p'' be their respective projections. The product B-space (X×BX′)×BX′′(X\times_BX')\times_BX'' is the topological subspace X×BX′×BX′′X\times_BX'\times_BX'' of X×X′×X′′X\times X'\times X'' formed by the triples (x,x′,x′′)(x, x', x'') such that p(x)=p′(x′)=p′′(x′′)p(x) =p'(x') =p''(x''), endowed with the projection q:X×BX′×BX′′→Bq: X\times_BX'\times_BX''\rightarrow B defined by q(x,x′,x′′)=p(x)q(x, x', x'') =p(x).

Example 3 01LD

Let X be a B-space and pp its projection. The fibred product X×BXX\times_BX of X and X over B is called the fibred square of X. It is the subspace of X×XX\times X formed by the pairs (x,x′)(x, x') such that p(x)=p(x′)p(x) =p(x'). It is endowed with the structure of a B-space whose projection is the mapping (x,x′)↦(x, x')\mapsto p(x)p(x). The diagonal ΔX\Delta_X of X×XX\times X (E, II, p. 13) is contained in X×BXX\times_BX; it is also called the diagonal of X×BXX\times_BX. The mapping x↦(x,x)x\mapsto (x, x) of X into X×BXX\times_BX is a B-morphism, called the diagonal B-morphism and often denoted by δX\delta_X; it defines a B-isomorphism of X onto ΔX\Delta_X (TG, I, p. 25, cor. 2).

Example 4 01LE

Let (Bi)i∈I(B_i)_{i\in I} be a family of topological spaces and let, for each i∈Ii\in I, (Xi,pi)(X_i, p_i) and (Yi,qi)(Y_i, q_i) be BiB_i-spaces. Put B=∏i∈IBiB =\prod_{i\in I}B_i, X=∏i∈IXiX =\prod_{i\in I}X_i and Y=∏i∈IYiY =\prod_{i\in I}Y_i. Endowed with the continuous mapping p=∏ipip=\prod_ip_i (resp. q=∏iqiq=\prod_iq_i), the topological space X (resp. Y) is a B-space. By the associativity isomorphism of the topological products of ∏i(Xi×Yi)\prod_i(X_i\times Y_i) onto (∏iXi)×(∏iYi)=X×Y(\prod_iX_i)\times (\prod_iY_i) = X\times Y (TG, I, p. 25, prop. 2), the subspace ∏i(Xi×BiYi)\prod_i(X_i\times_{B_i}Y_i) of ∏i(Xi×Yi)\prod_i(X_i\times Y_i) is identified with X×BYX\times_BY.

Example 5 01LF

Let (Xi)i∈I(X_i)_{i\in I} and (Yj)j∈J(Y_j)_{j\in J} be families of B-spaces. Let X and Y be their sums. For each (i,j)∈I×J(i, j)\in I\times J, the mapping (x,y)↦(x, y)\mapsto ((i,x),(j,y))((i, x),(j, y)) is a B-isomorphism of Xi×BYjX_i\times_BY_j onto the subspace ({i}×Xi)×B({j}×Yj)(\{i\} \times X_i)\times_B(\{j\} \times Y_j) of X×BYX\times_BY. Since the latter form a partition of X×BYX\times_BY into open subsets, the mapping

h:∐(i,j)∈I×J(Xi×BYj)→X×BY h:\coprod_{(i,j)\in I\times J}(X_i\times_BY_j)\rightarrow X\times_BY

defined by h((i,j),(x,y))=((i,x),(j,y))h((i, j),(x, y)) = ((i, x),(j, y)) is a B-isomorphism.

4. Change of base

Let B, B′B' be topological spaces and f:B′→Bf: B'\rightarrow B a continuous mapping. Let X be a B-space. The mapping ff endows B′B' with a structure of B-space, which makes it possible to define the fibred product B′×BXB'\times_BX. The latter, endowed with the mapping pr1:B′×BX→B′_1: B'\times_BX\rightarrow B', is a B′B'-space called the B′B'-space deduced from the B-space X by the base change f:B′→Bf: B'\rightarrow B (or by base change from B to B′B' following ff). It is also called the B′B'-space inverse image of X by ff. It is denoted by f∗(X)f^*(X), or sometimes XB′X_{B'} when there is no possible confusion about the mapping ff.

When B′B' is a subspace of B and f:B′→Bf: B'\rightarrow B is the canonical injection, the mapping (b′,x)↦x(b', x)\mapsto x of B′×BX′B'\times_BX' into p−1(B′)\overset{-1}{p}(B') (where pp is the projection of X) is a B′B'-isomorphism of f∗(X)f^*(X) onto the B′B'-space induced by X over B′B'.

Let Y be a second B-space and u:X→Yu: X\rightarrow Y a B-morphism. The mapping IdB′×Bu:B′×BX→B′×BY_{B'}\times_Bu: B'\times_BX\rightarrow B'\times_BY is a B′B'-morphism, called the B′B'-morphism deduced from the B-morphism uu by the base change f:B′→Bf: B'\rightarrow B and sometimes denoted by f∗(u)f^*(u), or uB′u_{B'} when there is no possible confusion about the mapping ff. It is the unique B′B'-morphism vv of B′×BXB'\times_BX into B′×BYB'\times_BY such that pr2∘v=u∘_2\circ v=u\circ pr2_2.

Let B′′B'' be a topological space and let g:B′′→B′g: B''\rightarrow B' be a continuous mapping. Then the mapping given by (b′′,(b′,x))↦(b′′,x)(b'',(b', x))\mapsto (b'', x) is an isomorphism of B′′B''-spaces of g∗(f∗(X))g^*(f^*(X)) onto (f∘g)∗(X)(f\circ g)^*(X), which will be said to be canonical.

5. Fibred product of a family of B-spaces

Let (Xi)i∈I(X_i)_{i\in I} be a family of B-spaces. Let pi:Xi→Bp_i: X_i\rightarrow B be their projections. Let ∏BXi\prod_BX_i denote the topological subspace of B×∏i∈IXiB\times \prod_{i\in I}X_i formed by the pairs (b,(xi)i∈I)(b,(x_i)_{i\in I}) such that pi(xi)=bp_i(x_i) =b for all i∈Ii\in I. The mapping p:∏BXi→Bp:\prod_BX_i\rightarrow B defined by p(b,(xi)i∈I)=bp(b,(x_i)_{i\in I}) =b is continuous.

Definition 3 01LG

The topological space ∏BXi\prod_BX_i is called the fibred product of the family (Xi)i∈I(X_i)_{i\in I} over B. The B-space obtained by endowing ∏BXi\prod_BX_i with the mapping pp is called the product B-space of the family (Xi)i∈I(X_i)_{i\in I}.

Let j∈Ij\in I. The mapping (b,x)↦(b, x)\mapsto prj(x)_j(x) of ∏BXi\prod_BX_i into XjX_j is called the projection of index jj of the fibred product and is again denoted by prj_j. It is continuous. It is a B-morphism, for one has p=pj∘p=p_j\circ prj_j.

Let Y be a B-space and qq its projection. For each i∈i\in I, let ui:Y→Xiu_i: Y\rightarrow X_i be a B-morphism. There exists a unique B-morphism v:Y→v: Y\rightarrow ∏BXi\prod_BX_i such that pri∘v=ui_i\circ v=u_i for each i∈Ii\in I (universal property of the product B-space): it is the mapping of Y into ∏BXi\prod_BX_i, defined by y↦(q(y),(ui(y))i∈I)y\mapsto (q(y),(u_i(y))_{i\in I}), which is sometimes denoted by (ui)i∈I(u_i)_{i\in I}.

Let (Xi)i∈I(X_i)_{i\in I} and (Yi)i∈I(Y_i)_{i\in I} be families of B-spaces and, for each i∈Ii\in I, let fi:Xi→Yif_i: X_i\rightarrow Y_i be a B-morphism. The mapping (b,(xi)i∈I)↦(b,(x_i)_{i\in I})\mapsto (b,(fi(xi))i∈I)(b,(f_i(x_i))_{i\in I}) is a B-morphism of ∏BXi\prod_BX_i into ∏BYi\prod_BY_i which is denoted by ∏Bfi\prod_Bf_i and is called the extension of the family (fi)i∈I(f_i)_{i\in I} to the fibred products.

Example 1 01LH

When the set I is empty, the set ∏i∈IXi\prod_{i\in I}X_i consists of a single element and the B-space ∏BXi\prod_BX_i is identified with B (endowed with the projection IdB_B).

Example 2 01LI

When I is not empty, one deduces from the mapping (b,x)↦x(b, x)\mapsto x of B×∏i∈IXiB\times \prod_{i\in I}X_i into ∏i∈IXi\prod_{i\in I}X_i, by passing to subspaces, a homeomorphism of ∏BXi\prod_BX_i onto the subspace of ∏i∈IXi\prod_{i\in I}X_i formed by the families (xi)i∈I(x_i)_{i\in I} such that pi(xi)=pj(xj)p_i(x_i) =p_j(x_j) for all i,j∈Ii, j\in I. This subspace will be called, by abuse, the fibred product of the family (Xi)i∈I(X_i)_{i\in I}.

Example 3 01LJ

When the set I is a set with one element α\alpha (resp. with two elements α\alpha and β\beta; resp. with three elements α,β,γ\alpha ,\beta ,\gamma ), the mapping prα_{\alpha} (resp. (prα_{\alpha}, prβ_{\beta}) ; resp. (prα_{\alpha}, prβ_{\beta}, prγ_{\gamma})) of ∏BXi\prod_BX_i into XαX_{\alpha} (resp. into Xα×BXβX_{\alpha}\times_BX_{\beta}; resp. into Xα×BXβ×BXγX_{\alpha}\times_BX_{\beta}\times_BX_{\gamma}) is a B-isomorphism. This will allow us to deduce the properties of the fibred product of two or three B-spaces from those of the fibred product of families of B-spaces.

Let (Xi)i∈I(X_i)_{i\in I} be a family of B-spaces and let J be a subset of I. From the mapping IdB×_B\times prJ_J of B×∏i∈IXiB\times \prod_{i\in I}X_i into B×∏i∈JXiB\times \prod_{i\in J}X_i, by passing to subsets, one deduces a B-morphism ∏i∈IBXi→∏i∈JBXi\prod_{i\in IB}X_i\rightarrow \prod_{i\in JB}X_i. One

again denotes it by prJ_J and calls it the projection of index J of the fibre product.

Let (Xi)i∈I(X_i)_{i\in I} be a family of B-spaces. Let (Jλ)λ∈L(J_{\lambda})_{\lambda\in L} be a partition

of I. The mapping (prJλ_{J_{\lambda}})λ∈L_{\lambda\in L} of ∏i∈IBXi→∏λ∈LB(∏i∈JλBXi)\prod_{i\in IB}X_i\rightarrow \prod_{\lambda\in LB}(\prod_{i\in J_{\lambda}B}X_i) is a

B-isomorphism ("associativity" of fibre products of B-spaces).

6. Cartesian Squares

Let B, B′B', X, X′X' be topological spaces and let f:B′→Bf: B'\rightarrow B, f′:X′→X,p:X→B,p′:X′→B′f': X'\rightarrow X,p: X\rightarrow B,p': X'\rightarrow B' be continuous mappings. Such a quadruple (f,f′,p,p′)(f, f', p, p') may be represented by a diagram

X′f′{X'}^{f'} X

p′p(1) p'p \tag{1}

B′f{B'}^f B

(E, II, p. 14). One then says: "Consider square diagram (1)," or simply "square (1)," instead of saying: "Consider the quadruple (f,f′,p,p′)(f, f', p, p') of continuous mappings." One says that square (1) is commutative if the equality

f∘p′=p∘f′ f\circ p'=p\circ f'

is satisfied. In this case, one often endows B′B', X and X′X' with the structures of B-spaces defined by the mappings f,pf,p and f∘p′=p∘f′f\circ p'=p\circ f' respectively; the mappings p′p' and f′f' are then B-morphisms.

Definition 4 01LK

One says that square (1) is a Cartesian square of topological spaces (or, simply, that it is Cartesian) if it is commutative and if, for every topological space Y and every pair of continuous mappings u:Y→B′,v:Y→Xu: Y\rightarrow B',v: Y\rightarrow X such that f∘u=p∘vf\circ u=p\circ v, there exists a unique continuous mapping w:Y→X′w: Y\rightarrow X' such that p′∘w=up'\circ w=u and f′∘w=vf'\circ w=v.

For square (1) to be Cartesian, it is necessary and sufficient that the square

X′p′B′ {X'}^{p'}B'
(1′)f′f (1')f'f

X p^p B

be Cartesian.

Proposition 1 01LL

For square (1) to be Cartesian, it is necessary and sufficient that it be commutative and that, for every B-space Y and every pair of B-morphisms u:Y→B′,v:Y→Xu: Y\rightarrow B',v: Y\rightarrow X, there exist a unique B-morphism w:Y→X′w: Y\rightarrow X' such that p′∘w=up'\circ w=u and f′∘w=vf'\circ w=v.

Suppose that square (1) is Cartesian. Let Y be a B-space and let u:Y→B′,v:Y→Xu: Y\rightarrow B',v: Y\rightarrow X be B-morphisms. The mappings f∘uf\circ u and p∘vp\circ v are both equal to the projection of the B-space Y; the unique continuous mapping ww such that p′∘w=up'\circ w=u and f′∘w=vf'\circ w=v is then a B-morphism. This proves the necessity of the condition.

Conversely, suppose this condition satisfied. Let Y be a topological space and let u:Y→B′,v:Y→Xu: Y\rightarrow B',v: Y\rightarrow X be continuous mappings such that f∘u=p∘vf\circ u=p\circ v. When Y is endowed with the structure of a B-space defined by f∘u,uf\circ u,u and vv are B-morphisms. Every continuous mapping w:Y→X′w: Y\rightarrow X' such that p′∘w=up'\circ w=u and f′∘w=vf'\circ w=v being a B-morphism, there exists one and only one.

Proposition 2 01LM

Let B, B′B' and X be topological spaces and let p:X→B,f:B′→Bp: X\rightarrow B,f: B'\rightarrow B be continuous mappings.

a) The square

B′×BXpr2B'\times_BX^{pr_2} X

(2) pr1p_{_1}p

B′f{B'}^f B

is a Cartesian square.

b) For every commutative square

X′f′{X'}^{f'} X

p′p(3) p'p \tag{3}

B′f{B'}^f B

there exists a unique continuous mapping h:X′→B′×BXh: X'\rightarrow B'\times_BX such that pr1∘h=p′_1\circ h=p' and pr2∘h=f′_2\circ h=f'.

c) The commutative square (3) is cartesian if and only if hh is a homeomorphism.

Assertion a) results from Prop. 1 and from the universal property of the product B-space of two B-spaces (I, p. 3). Assertion b) follows from it.

If square (3) is cartesian, there exists a unique continuous mapping h′:B′×BX→X′h': B'\times_BX\rightarrow X' such that f′∘h′f'\circ h' = pr2_2 and p′∘h′=p'\circ h'= pr1_1. We have f′∘h′∘h=f′f'\circ h'\circ h=f' and p′∘h′∘h=p′p'\circ h'\circ h=p', whence h′∘h=h'\circ h= IdX′_{X'} since square (3) is cartesian. We have pr1∘h∘h′_1\circ h\circ h' = pr1_1 and pr2∘h∘h′_2\circ h\circ h' = pr2_2, whence h∘h′=h\circ h'= IdB′×BX_{B'\times_BX} since square (2) is cartesian. This proves that hh is a homeomorphism.

Conversely, suppose that hh is a homeomorphism; since square (2) is cartesian, square (3) is also cartesian.

The mapping h:X′→B′×BXh: X'\rightarrow B'\times_BX whose existence and uniqueness is asserted by assertion b) of the preceding proposition will be said to be canonical: it is the mapping denoted by (p′,f′)(p', f') in I, p. 3.

Proposition 3 01LN

Let

X′f′{X'}^{f'} X

p′pp'p

B′f{B'}^f B

be a cartesian square. For every continuous section ss of p′p', the mapping f′∘sf'\circ s is a continuous lifting of ff to X. The mapping s↦f′∘ss\mapsto f'\circ s is a bijection of CB′(B′;X′)\mathscr{C}_{B'}(B'; X') onto CB(B′;X)\mathscr{C}_B(B'; X).

If s:B′→X′s: B'\rightarrow X' is a continuous section of p′p', we have p∘f′∘s=p\circ f'\circ s= f∘p′∘s=ff\circ p'\circ s=f, which proves that f′∘sf'\circ s is a continuous lifting of ff to X, i.e. a B-morphism from B′B' into X. Conversely, let g:B′→Xg: B'\rightarrow X be a B-morphism. We have f∘f\circ IdB′=p∘g_{B'}=p\circ g; there therefore exists, by definition of a cartesian square, a unique continuous mapping s:B′→X′s: B'\rightarrow X' such that p′∘s=p'\circ s= IdB′_{B'} and f′∘s=gf'\circ s=g, whence the proposition.

7. Cartesian Squares Constructed by Passing to Subspaces

Proposition 4 01LO

Let

X′f′{X'}^{f'} X

p′p(4) p'p \tag{4}

B′f{B'}^f B

be a cartesian square, and let B0,B0′,X0B_0,B'_0,X_0 be subspaces of B, B′B' and X respectively. Suppose that we have f(B0′)⊂B0,p(X0)⊂B0f(B'_0)\subset B_0,p(X_0)\subset B_0 and put X0′=(p−1′)(B0′)∩(f−1′)(X0)X'_0=(\overset{-1}{p}')(B'_0)\cap (\overset{-1}{f}')(X_0). Then the square

X0′f0′X0 {X'_0}^{f'_0}X_0
(4′)p0′p0 (4')p'_{_0}p_{_0}
B0′f0B0 {B'_0}^{f_0}B_0

(where the mappings f0,f0′,p0,p0′f_0,f'_0,p_0,p'_0 are deduced from f,f′,p,p′f,f',p,p' respectively by passing to the subsets) is cartesian.

Consider the canonical mapping h:X′→B′×BXh: X'\rightarrow B'\times_BX deduced from commutative diagram (4). Since square (4) is Cartesian, hh is a

homeomorphism. By construction, one has X0′=h−1(B0′×B0X0)X'_0=\overset{-1}{h}(B'_0\times_{B_0}X_0) and the mapping h0:X0′→B0′×B0X0h_0: X'_0\rightarrow B'_0\times_{B_0}X_0 deduced from commutative diagram (4′4') is deduced from hh by passing to subsets. It is therefore a homeomorphism and square (4′4') is Cartesian (Prop. 2).

Corollary 01LP

Let

X′f′{X'}^{f'} X

p′pp'p

B′f{B'}^f B

be a Cartesian square.

a) For every point b′b' of B′B', the mapping f′f' induces a homeomorphism of the fibre Xb′′X'_{b'} of p′p' onto the fibre Xf(b′)X_{f(b')} of pp.

b) If the mapping pp is injective (resp. surjective, resp. bijective), the same is true of p′p'.

Let b′b' be a point of B′B'. Put b=f(b′)b=f(b'). For every x′∈Xb′′x'\in X'_{b'}, one has p(f′(x′))=f(p′(x′))=f(b′)=bp(f'(x')) =f(p'(x')) =f(b') =b, whence f′(x′)∈Xbf'(x')\in X_b. This proves that

one has Xb′′⊂(f−1′)(Xb)X'_{b'}\subset (\overset{-1}{f}')(X_b). In Prop. 4, take B0={b},B0′={b′}B_0=\{b\},B'_0=\{b'\} and X0=XbX_0= X_b; one then has X0′=Xb′′X'_0= X'_{b'}, whence assertion a).

For the mapping pp to be injective (resp. surjective, resp. bijective), it is necessary and sufficient that the cardinal of each of its fibres be less than (resp. greater than, resp. equal to) 1. Assertion b) follows.

Example 01LQ

Let (X,p)(X, p) be a B-space and A a subspace of B. The square

p−1(A)j\overset{-1}{p}(A)^j X

pAp(5) p_{_A}p \tag{5}

A i^i B

(where ii and jj are the canonical injections) is Cartesian.

In particular, if A and A′A' are subspaces of the topological space B, the square

A∩A′A′ A\cap A'A'

(6)

A B

(where the arrows are the canonical injections) is Cartesian.

8. Cartesian Squares Constructed by Products, Fibre Products and Sums

Proposition 5 01LR

Let I be a set and, for every i∈Ii\in I, let

Xi′fi′Xi {X'_i}^{f'_i}X_i
pi′pi(7) p'_{_i}p_{_i} \tag{7}
Bi′fiBi {B'_i}^{f_i}B_i

be a Cartesian square. The square

∏i∈IX′if′∏i∈IXi \prod i\in IX'if'\prod i\in IXi
(7′)p′p (7')p'p
∏i∈IB′if∏i∈IBi \prod i\in IB'if\prod i\in IBi

(where f,f′,p,p′f,f',p,p' are the extensions of the families (fi)(f_i), (fi′)(f'_i), (pi)(p_i), (pi′)(p'_i) to the products) is Cartesian.

Let Y be a topological space, let u:Y→∏iBi′u: Y\rightarrow \prod_iB'_i and v:Y→∏iXiv: Y\rightarrow \prod_iX_i be continuous mappings such that f∘u=p∘vf\circ u=p\circ v. For i∈Ii\in I, put uiu_i = pri∘u_i\circ u and vi=v_i= pri∘v_i\circ v; one has fi∘ui=pi∘vif_i\circ u_i=p_i\circ v_i and there exists a unique continuous mapping wi:Y→Xi′w_i: Y\rightarrow X'_i such that pi′∘wi=uip'_i\circ w_i=u_i and fi′∘wi=vif'_i\circ w_i=v_i. Then the mapping w=(wi)w= (w_i) is a continuous mapping of Y into ∏iXi′\prod_iX'_i such that p′∘w=up'\circ w=u and f′∘w=vf'\circ w=v, and it is the only one having these properties.

Corollary 1 01LS

Let X be a B-space, let pp be its projection, and let F be a topological space. The square

X×F⟶pr1Xp×IdF↓↓pB×F⟶pr1B(8)\begin{array}{ccc} X\times F & \overset{\mathrm{pr}_1}{\longrightarrow} & X \\ {\scriptstyle p\times \mathrm{Id}_F}\big\downarrow & & \big\downarrow{\scriptstyle p} \\ B\times F & \overset{\mathrm{pr}_1}{\longrightarrow} & B \end{array} \tag{8}

is cartesian.

Let P be a topological space consisting of a single point. Corollary 1 follows from Prop. 5 applied to the cartesian squares

X⟶IdXXF⟶Pp↓↓pandIdF↓↓IdPB⟶IdBBF⟶P.\begin{array}{ccccccc} X & \overset{\mathrm{Id}_X}{\longrightarrow} & X & & F & \longrightarrow & P \\ {\scriptstyle p}\big\downarrow & & \big\downarrow{\scriptstyle p} & \text{and} & {\scriptstyle \mathrm{Id}_F}\big\downarrow & & \big\downarrow{\scriptstyle \mathrm{Id}_P} \\ B & \overset{\mathrm{Id}_B}{\longrightarrow} & B & & F & \longrightarrow & P \end{array}.

Let B and B′B' be topological spaces and let f:B′→f: B'\rightarrow B be a continuous mapping. Let I be a set and, for each i∈Ii\in I, let XiX_i be a B-space, Xi′X'_i a B′B'-space, and fi′:Xi′→Xif'_i: X'_i\rightarrow X_i a continuous mapping such that the square

Xi′fi′Xi {X'_i}^{f'_i}X_i

(9)

B′f{B'}^f B

be commutative. There exists a unique continuous mapping

f′:∏i∈IB′Xi′→∏i∈IBXi f':\prod_{i\in IB'}X'_i\rightarrow \prod_{i\in IB}X_i

such that pri∘f′=fi′∘_i\circ f'=f'_i\circ pri_i for every i∈Ii\in I and such that the square

i∏∈IB′X′if′i∏∈Ii\prod\in IB'X'if'i\prod\in IB XiXi

(9′) (9')

B′f{B'}^f B be commutative (this last condition resulting from the others if I =∅̸=\not\emptyset ) :: it is the mapping deduced from the mapping

f×∏i∈Ifi′:B′×∏i∈IXi′→B×∏i∈IXi f\times \prod_{i\in I}f'_i: B'\times \prod_{i\in I}X'_i\rightarrow B\times \prod_{i\in I}X_i

by passing to the subsets. With these notations:

Corollary 2 01LT

If the square (9) is cartesian for every i∈Ii\in I, the square (9′9') is cartesian.

From Prop. 5 one deduces a cartesian square

B′×∏iXi′IdB×∏ifi′B×∏iXi B'\times \prod_i{X'_i}^{Id_B\times}\prod_{_i}^{f'_i}B\times \prod_iX_i
B′×(B′)If×∏ifB×BI B'\times (B')^{If\times}\prod_{_i}^fB\times B^I

Let ΔB′\Delta_{B'} and ΔB\Delta_B denote the diagonals of B′×(B′)IB'\times (B')^I and B×BIB\times B^I. The diagram

i∏∈IB′X′if′i∏∈Ii\prod\in IB'X'if'i\prod\in IB XiXi

ΔB′ΔB \Delta_{B'}\Delta_B

deduced from the preceding one by passing to the subspaces is cartesian (I, p. 9, Prop. 4). It is identified with diagram (9′9').

Example 1 01LU

Let

X′f′{X'}^{f'} X

p′pp'p

B′f{B'}^f B

be a cartesian square. Then Corollary 2 yields a cartesian square

X′×B′X′φX×BX X'\times_{B'}{X'}^{\varphi}X\times_BX

B′f{B'}^f B .

We have the relation φ−1(ΔX)=ΔX′\overset{-1}{\varphi}(\Delta_X) = \Delta_{X'}. In fact, by prop. 2 of I, p. 8, it is enough to consider the case where X′=B′×BXX'= B'\times_BX. Let then ((b,x),(b,x′))((b, x),(b, x')) be an element of X′×B′X′X'\times_{B'}X' with b∈B′b\in B' and x,x′∈Xx, x'\in X. This element belongs to φ−1(ΔX)\overset{-1}{\varphi}(\Delta_X) if and only if x=x′x=x'.

Proposition 6 01LV

Let I be a set and, for every i∈Ii\in I, let

Xi′fi′{X'_i}^{f'_i} X

pi′p(10) p'_{_i}p \tag{10}

Bi′fi{B'_i}^{f_i} B

be a cartesian square. Let X′X' and B′B' be the sum spaces of the families (Xi′)(X'_i) and (Bi′)(B'_i) respectively. Let f:B′→B,f′:X′→Xf: B'\rightarrow B,f': X'\rightarrow X and p′:X′→B′p': X'\rightarrow B' be the mappings deduced from the families (fi)(f_i), (fi′)(f'_i) and (pi′)(p'_i) respectively. The square

X′f′{X'}^{f'} X

(10′)p′p (10')p'p

B′f{B'}^f B

is cartesian.

The mappings f,f′f,f' and p′p' are continuous. The commutativity of square (10′10') follows from its definition. Let us denote by hih_i the canonical homeomorphism of Xi′X'_i onto Bi′×BXB'_i\times_BX (I, p. 8, prop. 2) and by h:X′→B′×BXh: X'\rightarrow B'\times_BX the canonical mapping (loc. cit.). We have h=h′′∘h′h=h''\circ h' where h′:X′→∐(Bi′×BX)h': X'\rightarrow \coprod(B'_i\times_BX) is the homeomorphism deduced from the hih_i and h′′:∐(Bi′×BX)→(∐Bi′)×BXh'':\coprod(B'_i\times_BX)\rightarrow (\coprod B'_i)\times_BX is that defined in example 5 of I, p. 4. We conclude by prop. 2 of I, p. 8.

Example 2 01LW

Let (X,p)(X, p) be a B-space and let (Ak)k∈K(A_k)_{k\in K} be a family of subspaces of B. Let A be the sum space of the family (Ak)k∈K(A_k)_{k\in K} and let Y be the sum space of the family (p−1(Ak))k∈K(\overset{-1}{p}(A_k))_{k\in K}; let us denote by i:A→Bi: A\rightarrow B, j:Y→Xj: Y\rightarrow X and p′:Y→Ap': Y\rightarrow A the mappings deduced from the canonical injections of AkA_k into B, the canonical injections of p−1(Ak)\overset{-1}{p}(A_k) into X, and the mappings pAk:p−1(Ak)→Akp_{A_k}:\overset{-1}{p}(A_k)\rightarrow A_k, for k∈Kk\in K. The square

Y j^j X

p′p(11) p'p \tag{11}

A i^i B is cartesian; this follows from the example of I, p. 10 and from prop. 6.

9. Composition of cartesian squares

Proposition 7 01LX

Let

X′′g′X′X′f′{X''}^{g'}X'{X'}^{f'} X

(12) p′′p′p''p' and (13) p′pp'p

B′′gB′B′f{B''}^gB'{B'}^f B

be commutative squares; consider the square

X′′f′∘g′{X''}^{f'\circ g'} X

p′′p(14) p''p \tag{14}

B′′f∘g{B''}^{f\circ g} B .

It is commutative. If squares (12) and (13) are cartesian, the same is true of square (14). If squares (13) and (14) are cartesian, the same is true of square (12).

Square (14) is commutative, for we have p∘f′∘g′=f∘p′∘g′=f∘g∘p′′p\circ f'\circ g'=f\circ p'\circ g'=f\circ g\circ p''.

Let us denote by h′:X′′→B′′×B′X′,h:X′→B′×BXh': X''\rightarrow B''\times_{B'}X',h: X'\rightarrow B'\times_BX and h′′:X′′→B′′×BXh'': X''\rightarrow B''\times_BX the canonical continuous mappings deduced from the commutative squares (12), (13) and (14). Moreover, let us denote

j:B′′×BX→B′′×B′(B′×BX) j: B''\times_BX\rightarrow B''\times_{B'}(B'\times_BX)

the continuous mapping which associates to (b′′,x)(b'', x) the element (b′′,g(b′′),x)(b'', g(b''), x). It is a homeomorphism and one has j∘h′′=j\circ h''= (IdB′′×Bh_{B''}\times_Bh)∘h′\circ h'.

Suppose that square (13) is cartesian. Then hh is a homeomorphism (I, p. 8, prop 2), hence the mapping IdB′′×Bh_{B''}\times_Bh is a homeomorphism, and h′′h'' is a homeomorphism if and only if h′h' is one. This means that square (12) is cartesian if and only if square (14) is cartesian (loc. cit.).

With the notation of proposition 7, one sometimes says that square (14) is the square composed of squares (13) and (12). The first assertion expresses that the square composed of two cartesian squares is cartesian. In particular, the B′′B''-spaces g∗(f∗(X))g^*(f^*(X)) and (f∘g)∗(X)(f\circ g)^*(X) are isomorphic.

Remark 1 01LY

It may happen that squares (12) and (14) are cartesian without square (13) being so (I, p. 139, exerc. 2).

Remark 2 01LZ

Let p:X→Bp: X\rightarrow B and f:B′→Bf: B'\rightarrow B be continuous mappings. The mapping g:B′→B′×Bg: B'\rightarrow B'\times B defined by g(b′)=(b′,f(b′))g(b') = (b', f(b')) is a homeomorphism of B′B' onto the graph G of the mapping ff (TG, I, p. 25, cor. 2), and the fibred product B′×BXB'\times_BX is identified (I, p. 6) with the subspace of B′×XB'\times X, inverse image of G under the mapping IdB′×p:B′×X→B′×B_{B'}\times p: B'\times X\rightarrow B'\times B. By the example of I, p. 10 and cor. 1 of prop. 5 (I, p. 11), the squares

B′×BXiB′×XB′×Xpr2B'\times_BX^iB'\times XB'\times X^{pr_2} X

pr1_1 IdB′×p_{B'}\times p and IdB′×pp_{B'}\times pp

B′gB′×BB′×Bpr2{B'}^gB'\times BB'\times B^{pr_2} B

(where ii denotes the canonical injection) are cartesian, and the cartesian square

B′×BXpr2B'\times_BX^{pr_2} X

pr1p_1p

B′f{B'}^f B

is their composite. In other words, every cartesian square is identified with the square composed of a cartesian square obtained by product (I, p. 11, cor. 1 of prop. 5, square (8)) and of a cartesian square obtained by passing to subspaces (I, p. 10, example, square (5)).

Remark 3 01M0

Let

X1g1X_1^{g_1} X X2g2X_2^{g_2} X

(15) p1pp_{_1}p and (16) p2pp_{_2}p

B1f1B_1^{f_1} B B2f2B_2^{f_2} B

be cartesian squares. Consider the square

X1×XX2gX_1\times_XX_2^g X

p′p(17) p'p \tag{17}

B1×BB2fB_1\times_BB_2^f B where ff (resp. gg) is the mapping which defines the structure of B-space (resp. of X-space) of the fibred product B1×BB2B_1\times_BB_2 (resp. of the fibred product X1×XX2X_1\times_XX_2), and where p′p' is the mapping deduced from p1×p2p_1\times p_2 by passing to subsets. It is cartesian (I, p. 13, cor. 2 of prop. 5).

Consider then the following two commutative squares:

X1×XX2⟶pr1X1p′↓↓p1B1×BB2⟶pr1B1(18)\begin{array}{ccc} X_1\times_XX_2 & \overset{\mathrm{pr}_1}{\longrightarrow} & X_1 \\ {\scriptstyle p'}\big\downarrow & & \big\downarrow{\scriptstyle p_1} \\ B_1\times_BB_2 & \overset{\mathrm{pr}_1}{\longrightarrow} & B_1 \end{array} \tag{18}

and

X1×XX2⟶pr2X2p′↓↓p2B1×BB2⟶pr2B2.(19)\begin{array}{ccc} X_1\times_XX_2 & \overset{\mathrm{pr}_2}{\longrightarrow} & X_2 \\ {\scriptstyle p'}\big\downarrow & & \big\downarrow{\scriptstyle p_2} \\ B_1\times_BB_2 & \overset{\mathrm{pr}_2}{\longrightarrow} & B_2 \end{array}. \tag{19}

Square (17) is composed of squares (15) and (18), as well as of squares (16) and (19). By prop. 7, squares (18) and (19) are cartesian.

10. Strict Mappings

Proposition 8 01M1

Let

X′⟶f′Xp′↓↓pB′⟶fB\begin{array}{ccc} X' & \overset{f'}{\longrightarrow} & X \\ {\scriptstyle p'}\big\downarrow & & \big\downarrow{\scriptstyle p} \\ B' & \overset{f}{\longrightarrow} & B \end{array}

be a cartesian square. If the mapping pp is open ( resp. is proper, resp. has in a neighbourhood of each point a continuous section), the same is true of p′p'.

By remark 2, I, p. 16, it is enough to prove the proposition for cartesian squares of the following type:

p−1(A)⟶jXX×F⟶pr1XpA↓↓pandp×IdF↓↓pA⟶iBB×F⟶pr1B,\begin{array}{ccccccc} \overset{-1}{p}(A) & \overset{j}{\longrightarrow} & X & & X\times F & \overset{\mathrm{pr}_1}{\longrightarrow} & X \\ {\scriptstyle p_A}\big\downarrow & & \big\downarrow{\scriptstyle p} & \text{and} & {\scriptstyle p\times \mathrm{Id}_F}\big\downarrow & & \big\downarrow{\scriptstyle p} \\ A & \overset{i}{\longrightarrow} & B & & B\times F & \overset{\mathrm{pr}_1}{\longrightarrow} & B \end{array},

where F is a topological space, A a subspace of B, and i,ji,j the canonical injections. If the mapping pp is open, the mappings pAp_A and p×p\times IdF_F are open (TG, I, p. 30, prop. 2 and TG, I, p. 34, prop. 8). If the mapping pp is proper, the mappings pAp_A and p×p\times IdF_F are proper (TG, I, p. 72, prop. 3 and TG, I, p. 72, def. 1). If U is an open subset of B and s:U→Xs: U\rightarrow X a continuous section of pp over U, the mapping s∣(A∩U)s|(A\cap U) is a continuous section of pAp_A over A∩UA\cap U and the mapping s×s\times IdF_F is a continuous section of p×p\times IdF_F over U×FU\times F.

Remark 1 01M2

With the notation of prop. 8, if pp is a closed mapping, this is not necessarily so for p′(p'(cf. TG, I, p. 72, example). However, if the mapping pp is closed and if A is a subspace of B, the mapping pA:p−1(A)→Ap_A:\overset{-1}{p}(A)\rightarrow A is closed (TG, I, p. 30, prop. 2, a)).

Definition 5 01M3

Let X and Y be topological spaces and f:X→Yf: X\rightarrow Y a mapping. Let R be the equivalence relation associated with ff, and

X→X/R−→gf(X)→Y X\rightarrow X/R-\overset{g}{\rightarrow}f(X)\rightarrow Y

the canonical decomposition of ff (E, II, p. 44). One says that the mapping ff is strict if gg is a homeomorphism, when X/RX/R is endowed with the quotient topology and f(X)f(X) with the topology induced by that of Y.

A strict mapping is continuous.

Let us recall (TG, I, p. 22, prop. 8) that, in order that a mapping ff be strict, it is necessary and sufficient that ff be continuous and that for every open (resp. closed) saturated subset A of X, the set f(A)f(A) be open (resp. closed) in f(X)f(X).

Example 1 01M4

The composite of two strict mappings is not necessarily strict. In fact, every continuous mapping f:X→Yf: X\rightarrow Y is the composite of the mapping pr2:X×Y→_2: X\times Y\rightarrow Y and of the mapping x↦(x,f(x))x\mapsto (x, f(x)) of X into X×YX\times Y, which are both strict (TG, I, p. 26, prop. 5 and TG, I, p. 25, cor. 2). On the other hand, the composite of two strict and injective (resp. surjective) mappings is a strict mapping.

Example 2 01M5

A continuous mapping which is open, or closed, or which possesses a continuous section, is strict. This follows from prop. 3 of TG, I, p. 32 and from prop. 9 of TG, I, p. 22.

Example 3 01M6

In order that a continuous homomorphism of one topological group into another be a strict morphism (TG, III, p. 16, def. 1), it is necessary and sufficient that it be a strict mapping in the sense of Definition 5.

Proposition 9 01M7

Let X, Y and Z be topological spaces. Let f:X→Yf: X\rightarrow Y be a continuous surjective mapping and g:Y→Zg: Y\rightarrow Z a mapping.

a) If ff is strict and if g∘fg\circ f is continuous, the mapping gg is continuous.

b) If gg is continuous and if g∘fg\circ f is strict, the mapping gg is strict.

c) If ff and g∘fg\circ f are strict, gg is strict.

Let us prove assertion a). Let R denote the relation associated with ff in X; by hypothesis, the mapping of X/RX/R onto Y deduced from ff by passing to the quotient is a homeomorphism. The first assertion then follows from prop. 6 of I, p. 21.

Let us prove b). Let B be a closed saturated subset of Y for the relation defined by gg and let A=f−1(B)A =\overset{-1}{f}(B). Since ff is continuous, A is closed in X, and A is saturated for the equivalence relation defined by g∘fg\circ f. Since g∘fg\circ f is strict and ff is surjective, g(B)=g∘f(A)g(B) =g\circ f(A) is therefore closed in Z. The continuous mapping gg is therefore strict.

Assertion c) results immediately from assertions a) and b).

Proposition 10 01M8

Let X be a topological space and let R be an equivalence relation in X. Let Y be a locally compact topological space. Let S be the equivalence relation in X×YX\times Y which is the product of the equivalence relation R in X and of the equality relation in Y. The canonical bijection (X×Y)/S→(X/R)×Y(X\times Y)/S\rightarrow (X/R)\times Y is a homeomorphism.

Recall that if U and V are topological spaces, Cc(U;V)\mathscr{C}_c(U; V) denotes the set of continuous mappings of U into V, endowed with the topology of compact convergence (TG, X, p. 26, def. 1).

Let p:X→X/Rp: X\rightarrow X/R and q:X×Y→(X×Y)/Sq: X\times Y\rightarrow (X\times Y)/S be the canonical surjections. Let g:(X×Y)/S→(X/R)×Yg: (X\times Y)/S\rightarrow (X/R)\times Y denote the canonical bijection. It is continuous; let hh denote its inverse and let us prove that it is continuous.

The mapping i:X→Cc(Y;X×Y)i: X\rightarrow \mathscr{C}_c(Y; X\times Y) such that, for every x∈x\in X, i(x)i(x) is the mapping defined by y↦(x,y)y\mapsto (x, y), is continuous (TG, X, p. 28, th. 3). The mapping q~:Cc(Y;X×Y)→Cc(Y;(X×Y)/S)\widetilde{q}:\mathscr{C}_c(Y; X\times Y)\rightarrow \mathscr{C}_c(Y; (X\times Y)/S) which associates with a continuous mapping φ\varphi the mapping q∘φq\circ \varphi, is continuous (TG, X, p. 29, prop. 9). Consequently, the mapping q~∘i:X→Cc(Y;(X×Y)/S)\widetilde{q}\circ i: X\rightarrow \mathscr{C}_c(Y; (X\times Y)/S) is continuous. It is compatible with the equivalence relation R; the unique mapping j:X/R→Cc(Y;(X×Y)/S)j: X/R\rightarrow \mathscr{C}_c(Y; (X\times Y)/S) such that j∘p=q~∘ij\circ p=\widetilde{q}\circ i is therefore continuous. We have h(ξ,y)=j(ξ)(y)h(\xi , y) =j(\xi )(y) for every pair (ξ,y)∈(X/R)×Y(\xi , y)\in (X/R)\times Y. Since Y is locally compact, it then follows from TG, X, p. 28, th. 3, that the mapping hh is continuous.

The conclusion of Proposition 10 is no longer necessarily true if Y is not locally compact (TG, I, p. 96, exerc. 6).

Corollary 01M9

Let X and Y be topological spaces and let f:X→f: X\rightarrow Y be a continuous mapping. Let T be a locally compact topological space. If the mapping ff is strict, the same is true of the mapping f×f\times IdT_T of X×TX\times T into Y×TY\times T.

11. Universally strict mappings

Let

X′f′{X'}^{f'} X

p′pp'p

B′f{B'}^f B

be a cartesian square, where the mapping pp is strict. The mapping p′p' is not necessarily strict (cf. TG, I, p. 96, exerc. 6). However, if A is an open or closed subspace of B, the mapping pA:p−1(A)→Ap_A:\overset{-1}{p}(A)\rightarrow A is strict (TG, I, p. 23, cor. 1).

Definition 6 01MA

Let p:X→Bp: X\rightarrow B be a continuous mapping. One says that the mapping pp is universally strict if for every cartesian square

X′f′{X'}^{f'} X

p′pp'p

B′f{B'}^f B

the mapping p′p' is strict.

A universally strict mapping is strict. With the notations of Definition 6, if the mapping pp is universally strict, the same is true of p′p'. This results immediately from def. 6 and prop. 7 of I, p. 15.

Corollary 01MB

A continuous mapping which is open, or proper, or which admits in a neighbourhood of every point a continuous section, is universally strict (prop. 8 and example 2).

Example 01MC

Let B be a topological space and let (Ai)i∈I(A_i)_{i\in I} be a covering of B; let A denote the sum space of the family (Ai)i∈I(A_i)_{i\in I}, and let p:A→Bp: A\rightarrow B be the canonical mapping. The mapping pp is universally strict under each of the following two assumptions:

(i) For every point b∈Bb\in B, there exists i∈Ii\in I such that bb is an interior point of AiA_i;

(ii) The family (Ai)(A_i) is locally finite and the AiA_i are closed subsets of B.

Under assumption (i), the mapping pp possesses in a neighbourhood of every point a continuous section. Under assumption (ii), the mapping pp is proper by definition of the sum space (resp. from TG, I, p. 6, prop. 4 and TG, I, p. 75, th. 1). It is therefore universally strict.

Remark 01MD

A closed mapping is not necessarily universally strict (cf. TG, I, p. 96, exerc. 6).

Proposition 11 01ME

Let

X′f′{X'}^{f'} X

p′pp'p

B′f{B'}^f B

a cartesian square.

a) Suppose the mapping ff strict and surjective. Then, if the mapping p′p' is open ( resp. closed, resp. proper), the same is true of pp.

b) Suppose that the mapping ff is closed and surjective. Then, if the mapping p′p' is strict, the same is true of pp.

c) Suppose that the mapping ff is universally strict and surjective. Then, if the mapping p′p' is strict, the same is true of pp.

Let us first remark that for every subset A of X, one has

p′((f−1′)(A))=f−1(p(A))(20) p'((\overset{-1}{f}')(A)) =\overset{-1}{f}(p(A)) \tag{20}

In fact, if x′∈(f−1′)(A)x'\in (\overset{-1}{f}')(A), then f(p′(x′))=p∘f′(x′)∈p(A)f(p'(x')) =p\circ f'(x')\in p(A). Conversely, if b′∈f−1(p(A))b'\in \overset{-1}{f}(p(A)), let x∈Ax\in A be such that f(b′)=p(x)f(b') =p(x). By definition of a cartesian square, there exists a unique x′∈X′x'\in X' such that f′(x′)=xf'(x') =x and p′(x′)=b′p'(x') =b', so that b′∈p′((f−1′)(A))b'\in p'((\overset{-1}{f}')(A)).

Let us prove a). Suppose first that p′p' is an open (resp. closed) mapping and let A be an open (resp. closed) set in X. The set (f−1′)(A)(\overset{-1}{f}')(A) is open (resp. closed) in X′X'. Hence the set p′((f−1′)(A))p'((\overset{-1}{f}')(A)) is open (resp. closed) in B′B'. By relation (20), it is also saturated with respect to the equivalence relation defined by ff. Since the mapping ff is supposed surjective, one then has p(A)=p(A) =

f(p′((f−1′)(A)))f(p'((\overset{-1}{f}')(A))), and since it is strict, the set p(A)p(A) is open (resp. closed) in B. The mapping pp is therefore open (resp. closed).

In order that a continuous mapping be proper, it is necessary and sufficient that it be closed and that its fibres be quasi-compact (TG, I, p. 75, th. 1). If the mapping p′p' is proper, the mapping pp is closed by what precedes. Let us study its fibres: if b∈Bb\in B, let b′∈B′b'\in B' be such that f(b′)=bf(b') =b. By the corollary of I, p. 10, the mapping f′f' induces a homeomorphism Xb′′→XbX'_{b'}\rightarrow X_b. Since p′p' is proper, Xb′′X'_{b'} is quasi-compact, hence XbX_b is so also. The mapping pp is therefore proper.

Let us prove b) and c). Let Y = p(X)p(X) and Y′=p′(X′)Y'=p'(X') ; relation (20) applied to X implies that Y′=f−1(Y)Y'=\overset{-1}{f}(Y). Let gg denote the mapping of Y′Y' into Y induced by ff. In case b), the mapping gg is strict by remark 1, I, p. 18. In case c), it is strict by virtue of definition 6 since the square

Y′Y' Y

B′B' B

is cartesian.

Let qq and q′q' denote the mappings of X into Y and of X′X' into Y′Y' induced by pp and p′p', so that the square

X′f′{X'}^{f'} X

q′qq'q

Y′g{Y'}^g Y

is cartesian. Since the mappings gg and q′q' are both strict and surjective, their composite g∘q′g\circ q' is strict. Hence q∘f′q\circ f' is strict and p∘f′p\circ f' is itself strict. Since ff is surjective, f′f' is so also and pp is strict by prop. 9, b) of I, p. 18.

Corollary 1 01MF

Suppose that the mapping ff is universally strict and surjective and that the mapping p′p' is universally strict. Then the mapping pp is universally strict.

Let

Y g′^{g'} X

qpqp

C g^g B

be a cartesian square; it is a question of proving that the mapping qq is strict. By remark 3 of I, p. 16, the squares

X′×XYpr1X′ X'\times_XY^{pr_1}X'
rp′rp'
B′×BCpr1B′ B'\times_BC^{pr_1}B'

and

X′×XYpr2X'\times_XY^{pr_2} Y

rq(21) rq \tag{21}

B′×BCpr2B'\times_BC^{pr_2} C

are cartesian, where r:X′×XY→B′×BCr: X'\times_XY\rightarrow B'\times_BC denotes the mapping induced by (p′,q)(p', q). Since the mapping ff is universally strict and surjective, the same is true of the mapping pr2:B′×BC→C_2: B'\times_BC\rightarrow C (I, p. 20, def. 6 and I, p. 10, cor. of prop. 4). On the other hand, the mapping rr is strict, since p′p' is assumed universally strict. By prop. 11, c) applied to the cartesian square (21), the mapping qq is strict.

Corollary 2 01MG

Let B and X be topological spaces and let p:X→Bp: X\rightarrow B be a continuous mapping. Let (Ai)i∈I(A_i)_{i\in I} be a family of subsets of B which is an open covering of B, or else a locally finite closed covering of B. If for every i∈i\in I, the mapping pAi:p−1(Ai)→Aip_{A_i}:\overset{-1}{p}(A_i)\rightarrow A_i is strict ( resp. universally strict), the mapping pp is strict ( resp. universally strict).

For every i∈Ii\in I, put Yi=p−1(Ai)Y_i=\overset{-1}{p}(A_i) and pi=pAip_i=p_{A_i}. Let A be the sum space of the family (Ai)i∈I(A_i)_{i\in I}, Y the sum space of the family (Yi)i∈I(Y_i)_{i\in I}; let us denote by f:A→Bf: A\rightarrow B (resp. g:Y→X,q:Y→Ag: Y\rightarrow X,q: Y\rightarrow A) the mapping deduced from the family of canonical injections Ai→BA_i\rightarrow B (resp. from the canonical injections Yi→XY_i\rightarrow X, from the mappings pip_i). The square

Y g^g X

qpqp

A f^f B

is a cartesian square (examples, I, p. 10 and p. 14). By the corollary, I, p. 20, the mapping ff is universally strict. By proposition 11, it is therefore enough to prove that the mapping qq is strict (resp. universally strict) if the mappings pip_i, for i∈Ii\in I, are so. We are thus reduced to proving the corollary when the sets AiA_i, i∈Ii\in I, constitute a partition of the space B into open and closed subsets, which we shall henceforth suppose.

Suppose that each of the mappings pi,i∈Ip_i,i\in I, is strict. If U is an open subset of X and saturated for the equivalence relation defined by pp, the set pi(Xi∩U)p_i(X_i\cap U) is open in AiA_i and p(U)=p(U) = ⋃i∈Ipi(Xi∩U)\bigcup_{i\in I}p_i(X_i\cap U) is open in B. The mapping pp is therefore strict.

Suppose now that each of the mappings pi,i∈Ip_i,i\in I, is universally strict, and now prove that the mapping pp is universally strict. Let C be a topological space and h:C→Bh: C\rightarrow B a continuous mapping. It is a question of proving that the mapping pr1:X×BC→C_1: X\times_BC\rightarrow C is strict. The space C is identified with the sum space of the family of the Ci=h−1(Ai)C_i=\overset{-1}{h}(A_i) and the space X×BCX\times_BC is identified with the sum space of the family of the X×BCi=Xi×AiCiX\times_BC_i= X_i\times_{A_i}C_i. Since pip_i is universally strict, the mapping pr1:Xi×AiCi→Ci_1: X_i\times_{A_i}C_i\rightarrow C_i is strict. By what precedes, the mapping pr1:X×BC→C_1: X\times_BC\rightarrow C is strict. This proves that the mapping pp is universally strict.

Exercises

See the exercises for § 1.

Machine translation of the English, by gpt-5.4, against glossary version 34. Not checked by a person.