Bourbaki

§ 2. Étale mappings

1. Separated mappings

Proposition 1 01MH

Let X and Y be topological spaces and let f:X→f: X\rightarrow Y be a continuous mapping. The following properties are equivalent:

(i) The diagonal ΔX\Delta_X of the fibred product X×YXX\times_YX is a closed subspace;

(ii) For every topological space W and every pair of continuous mappings (g1,g2)(g_1, g_2) from W into X such that f∘g1=f∘g2f\circ g_1=f\circ g_2, the set of points w∈Ww\in W such that g1(w)=g2(w)g_1(w) =g_2(w) is closed in W;

(iii) For every pair (x1,x2)(x_1, x_2) of points of X such that x1≠x2x_1=\not x_2 and f(x1)=f(x2)f(x_1) =f(x_2), there exists a neighbourhood V1V_1 of x1x_1 in X and a neighbourhood V2V_2 of x2x_2 in X such that V1∩V2=∅V_1\cap V_2=\emptyset.

(i)⇒\Rightarrow(ii) : Let g1,g2g_1,g_2 be continuous mappings from W into X such that f∘g1=f∘g2f\circ g_1=f\circ g_2, and let g:W→X×YXg: W\rightarrow X\times_YX be the mapping deduced from g1g_1 and g2g_2. The set of points w∈Ww\in W such that g1(w)=g2(w)g_1(w) =g_2(w) is g−1(ΔX)\overset{-1}{g}(\Delta_X). Since gg is continuous, it is therefore closed if the diagonal ΔX\Delta_X is closed.

(ii)⇒\Rightarrow(i) : The diagonal ΔX\Delta_X is the set of points z∈X×YXz\in X\times_YX such that pr1(z)=_1(z) = pr2(z)_2(z). It follows from (ii) applied to W=X×YXW = X\times_YX and to the pair of mappings (pr1_1, pr2_2) that the diagonal ΔX\Delta_X is closed in X×YXX\times_YX.

(i)⇔\Leftrightarrow(iii) : Let (x1,x2)(x_1, x_2) be a point of X×YXX\times_YX and let V1V_1 and V2V_2 be neighbourhoods of x1x_1 and x2x_2 respectively. The condition V1∩V2=∅V_1\cap V_2=\emptyset is equivalent to the condition (V1×V2)∩ΔX=∅(V_1\times V_2)\cap \Delta_X=\emptyset, that is, to (V1×V2)∩(V_1\times V_2)\cap (X×YX)∩ΔX=∅(X\times_YX)\cap \Delta_X=\emptyset. Since the sets (V1×V2)∩(X×YX)(V_1\times V_2)\cap (X\times_YX) form a basis of neighbourhoods of (x1,x2)(x_1, x_2) in X×YXX\times_YX, this proves the equivalence of (i) and (iii).

Definition 1 01MI

Let X and Y be topological spaces. A continuous mapping f:X→Yf: X\rightarrow Y is said to be separated if it satisfies the equivalent conditions of proposition 1.

Proposition 2 01MJ

Let X, Y, Z be topological spaces and let f:X→Yf: X\rightarrow Y and g:Y→Zg: Y\rightarrow Z be continuous mappings.

a) If ff and gg are separated, then g∘fg\circ f is separated.

b) If g∘fg\circ f is separated, then ff is separated.

c) Suppose moreover that the mapping ff is proper and surjective. Then, if g∘fg\circ f is separated, gg is separated.

Consider, in X×XX\times X, the subspaces ΔX\Delta_X (diagonal), X×YXX\times_YX and X×ZXX\times_ZX. These are respectively the sets of points uu of X×XX\times X such that pr1(u)=_1(u) = pr2(u),f∘_2(u),f\circpr1(u)=f∘_1(u) =f\circpr2(u),g∘f∘_2(u),g\circ f\circpr1(u)=g∘f∘_1(u) =g\circ f\circpr2(u)_2(u). If g∘fg\circ f is separated, ΔX\Delta_X is closed in X×ZXX\times_ZX, hence also in X×YXX\times_YX, whence b).

By prop. 1, (ii), applied to W=X×ZX,g1=f∘W = X\times_ZX,g_1=f\circ pr1_1 and g2=f∘g_2=f\circ pr2,X×YX_2,X\times_YX is closed in X×ZXX\times_ZX if gg is separated. If moreover ff is separated, ΔX\Delta_X is closed in X×YXX\times_YX (prop. 1, (i)), hence in X×ZXX\times_ZX, whence a).

Finally, let us prove c). The mapping (f,f):X×X→Y×Y(f, f): X\times X\rightarrow Y\times Y is proper (TG, I, p. 73, prop. 4). The subspace X×ZXX\times_ZX is the inverse image of Y×ZYY\times_ZY by (f,f)(f, f) ; by TG, I, p. 72, prop. 3, the mapping u:X×Zu: X\times_Z X→Y×ZYX\rightarrow Y\times_ZY deduced from (f,f)(f, f) is proper. Since g∘fg\circ f is separated, the diagonal ΔX\Delta_X is closed in X×ZXX\times_ZX. Consequently, u(ΔX)u(\Delta_X) is closed in Y×ZYY\times_ZY. Since ff is surjective, u(ΔX)u(\Delta_X) is the diagonal of Y×ZYY\times_ZY. This shows that gg is separated.

Remark 1 01MK

An injective continuous mapping is separated.

Remark 2 01ML

For a topological space X to be separated (TG, I, p. 52, def. 1), it is necessary and sufficient that the mapping of X into a space reduced to a point be separated. In this case, every continuous mapping of X into a topological space is separated (prop. 1, (iii)).

Let f:X→f: X\rightarrow Y be a continuous and separated mapping. For every point yy of Y, the fibre f−1(y)\overset{-1}{f}(y) is a separated topological space (loc. cit.). There exist nevertheless continuous mappings which are not separated but all of whose fibres are separated topological spaces (I, p. 140, exerc. 1).

Remark 3 01MM

Let f:X→Yf: X\rightarrow Y be a continuous and separated mapping. If the space Y is separated, the space X is separated. This follows from remark 2 and proposition 2 applied with a space Z reduced to a point.

Remark 4 01MN

Let f:X→Yf: X\rightarrow Y be a continuous and separated mapping, yy a point

of Y and A a finite subset of f−1(y)\overset{-1}{f}(y). Let us show that there exists a family (Va)a∈A(V_a)_{a\in A} of pairwise disjoint sets such that for each a∈Aa\in A, the set VaV_a is a neighbourhood of aa in X. For this, for each subset {a,b}\{a, b\} of A with two elements, let us choose a neighbourhood V(a,b)V_{(a,b)} of aa and a neighbourhood V(b,a)V_{(b,a)} of bb in X such that V(a,b)∩V(b,a)=∅V_{(a,b)}\cap V_{(b,a)}=\emptyset (prop. 1, (iii)). Let VaV_a denote the intersection of the family consisting of X and the sets V(a,b)V_{(a,b)} for b∈A,b=a̸b\in A,b=\not a. The set VaV_a is a neighbourhood of aa in X, and if a,ba,b are two distinct elements of A, the set Va∩VbV_a\cap V_b is contained in V(a,b)∩V(b,a)V_{(a,b)}\cap V_{(b,a)}, hence is empty.

Remark 5 01MO

Let Y be a topological space. Let (Ai)i∈I(A_i)_{i\in I} be a family of subsets of Y and let X be the sum space of the family (Ai)i∈I(A_i)_{i\in I}. The canonical mapping of X into Y is separated.

Proposition 3 01MP

Let f:X→Yf: X\rightarrow Y be a continuous and separated mapping. Every continuous section of ff induces a homeomorphism of Y onto a closed subset of X.

Let s:Y→Xs: Y\rightarrow X be a continuous section of ff. The mapping ss induces a homeomorphism of Y onto the subspace s(Y)s(Y) of X (TG, I, p. 22, prop. 9). The identity mapping of X and the mapping s∘fs\circ f are continuous and one has f∘f\circ IdX=f∘(s∘f)_X=f\circ (s\circ f). Since the mapping ff is separated, the set s(Y)s(Y), which is the set of points xx of X such that x=s∘f(x)x=s\circ f(x), is closed in X (prop. 1, (ii)).

Proposition 4 01MQ

Let

X′f′{X'}^{f'} X

p′pp'p

B′f{B'}^f B

be a cartesian square. If the mapping pp is separated, the same is true of p′p'. Conversely, if p′p' is separated and if ff is universally strict and surjective, then pp is separated.

Consider the square

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

B′f{B'}^f B

where the mapping φ\varphi is induced by the mapping f′×f′:X′×X′→X×Xf'\times f': X'\times X'\rightarrow X\times X. Recall (I, p. 13, example 1) that this square is cartesian and that one has φ−1(ΔX)=ΔX′\overset{-1}{\varphi}(\Delta_X) = \Delta_{X'}.

If the mapping pp is separated, the set ΔX\Delta_X is closed in X×BXX\times_BX; consequently, ΔX′\Delta_{X'} is a closed subset of X′×B′X′X'\times_{B'}X', which proves that the mapping p′p' is separated.

If the mapping ff is surjective, the mapping φ\varphi is surjective (I, p. 10, cor.) ; if ff is universally strict, φ\varphi is strict (I, p. 20, def. 6). Suppose the mapping p′p' separated. Then ΔX′\Delta_{X'} is closed in X′×B′X′X'\times_{B'}X'. Since ΔX′=φ−1(ΔX)\Delta_{X'}=\overset{-1}{\varphi}(\Delta_X) and since φ\varphi is surjective and strict, ΔX\Delta_X is closed in X×BXX\times_BX, which proves that the mapping pp is separated.

Proposition 5 01MR

Let X, Y, Z be topological spaces and let f:X→Y,g:Y→Zf: X\rightarrow Y,g: Y\rightarrow Z be continuous mappings. Suppose the mapping gg separated and the mapping g∘fg\circ f proper. Then the mapping ff is proper.

Consider the following cartesian square:

X×ZYpr2X\times_ZY^{pr_2} Y

pr1g_1g

X g∘fZ^{g\circ f}Z.

Let s:X→X×ZYs: X\rightarrow X\times_ZY be the mapping x↦(x,f(x))x\mapsto (x, f(x)). It is a continuous section of pr1:X×ZY→X_1: X\times_ZY\rightarrow X. By prop. 4, the mapping pr1_1 is separated; it follows that the mapping ss is proper (I, p. 27, prop. 3 and TG, I, p. 72, prop. 2). On the other hand, the mapping pr2_2 is proper (I, p. 17, prop. 8). Hence the mapping ff, which is equal to pr2∘s_2\circ s, is proper (TG, I, p. 73, prop. 5).

2. Etale mappings

Definition 2 01MS

Let E and B be topological spaces, let p:E→Bp: E\rightarrow B be a mapping and let xx be a point of E. One says that the mapping pp is topologically etale at xx if there exists a neighbourhood U of xx in E and a neighbourhood V of p(x)p(x) in B such that pp induces a homeomorphism of U onto V.

One says that the mapping pp is topologically etale if it is topologically etale at every point xx of E.

When no confusion is possible, cf. example 3 below, one shall say etale instead of topologically etale. Instead of saying that p:E→Bp: E\rightarrow B is an etale mapping, one also says that the B-space (E,p)(E, p) is an etale B-space, or simply that E is a space spread over B, when no doubt is possible as to the mapping pp.

Remark 1 01MT

The set of points of E at which a mapping p:E→Bp: E\rightarrow B is etale is an open set U of E, and the restriction of pp to U is an etale mapping of U into B.

Remark 2 01MU

An etale mapping is continuous and open (TG, I, p. 33, prop. 5) ; in particular, the image of an etale mapping is open. Conversely, if p:E→Bp: E\rightarrow B is a continuous and open mapping, and if every point xx of E possesses a neighbourhood V such that the mapping p∣Vp|V is injective, the mapping pp is etale. The fibres of an etale mapping are discrete.

Example 1 01MV

Let B be a topological space and (Ui)i∈I(U_i)_{i\in I} a family of subsets of B. Let E denote the sum space of the family (Ui)i∈I(U_i)_{i\in I} and p:E→Bp: E\rightarrow B the mapping deduced from the canonical injections of the UiU_i into B. In order that the mapping pp be etale, it is necessary and sufficient that the Ui,i∈IU_i,i\in I, all be open.

Example 2 01MW

Let U be an open set of C\mathbf{C}. In order that a holomorphic function f:U→Cf: U\rightarrow \mathbf{C} be an etale mapping, it is necessary and sufficient that its derivative nowhere vanish.

Example 3 01MX

An etale morphism of varieties (VAR, R, 5.7.8) is a topologically etale mapping, but there exist morphisms of real varieties which are topologically etale and which are not etale morphisms. This is the case, for example, of the mapping x↦x3x\mapsto x^3 of R\mathbf{R} into R\mathbf{R}. However, a morphism of complex analytic varieties which is topologically etale is an etale morphism (I, p. 141, exerc. 6).

Proposition 6 01MY

Let X, Y, Z be topological spaces and let f:X→Y,g:Y→Zf: X\rightarrow Y,g: Y\rightarrow Z be mappings.

a) Suppose that ff and gg are etale; then g∘fg\circ f is etale.

b) Suppose that gg is étale, ff is continuous and g∘fg\circ f is open. Then ff is open.

c) Suppose that g∘fg\circ f and gg are étale and that ff is continuous. The mapping ff is then étale.

d) Suppose that g∘fg\circ f is étale and that the mapping ff is continuous and open; then ff is étale and gg is étale at every point of f(X)f(X).

Let us prove a). Suppose the mappings ff and gg are étale. Then they are continuous and open, hence the mapping g∘fg\circ f is continuous and open. Let xx be a point of X. There exists a neighbourhood W of f(x)f(x) in Y such that the mapping g∣Wg|W is injective, and a neighbourhood V of xx

in X, contained in f−1(W)\overset{-1}{f}(W), such that the mapping f∣Vf|V is injective; the mapping (g∘f)∣V(g\circ f)|V is then injective. This proves that the mapping g∘fg\circ f is étale (remark 2).

Let us prove b). Let xx be a point of X and W an open neighbourhood of f(x)f(x) such that gg induces a homeomorphism of W onto the open set g(W)g(W). Let V be a neighbourhood of xx such that f(V)f(V) is contained in W; then g∘f(V)g\circ f(V) is a neighbourhood of g∘f(x)g\circ f(x), hence f(V)f(V) is a neighbourhood of f(x)f(x) in the open set W, and also in Y. This proves that the mapping ff is open (TG, I, p. 33, prop. 5).

Let us prove c). By b),f),f is open. Let xx be a point of X. Since g∘fg\circ f is étale, there exists a neighbourhood V of xx in X such that g∘f∣Vg\circ f|V is injective. Thus, f∣Vf|V is injective and ff is étale (I, p. 29, remark 2).

Let us prove d). Let xx be a point of X and put y=f(x)y=f(x). There exists an open neighbourhood V of xx such that g∘fg\circ f induces a homeomorphism of V onto the open set g∘f(V)g\circ f(V). Since the mapping ff is open, f(V)f(V) is an open neighbourhood of yy. The mapping f∣V:V→f(V)f|V: V\rightarrow f(V) deduced from ff by passing to subspaces is continuous, open and bijective; hence it is a homeomorphism, so that ff is étale at xx. Moreover, the mapping gg induces a homeomorphism of f(V)f(V) onto g∘f(V)g\circ f(V), hence is étale at yy.

Corollary 1 01MZ

Let B be a topological space. A B-morphism of one étalé B-space into another is étale.

This follows from assertion c) of prop. 6.

Corollary 2 01N0

Let B be a topological space; a bijective B-morphism of one étalé B-space into another is a B-isomorphism.

By corollary 1, such a morphism is étale; it is therefore open (I, p. 29, remark 2). If it is bijective, it is a B-isomorphism.

Corollary 3 01N1

Let p:E→Bp: E\rightarrow B be an étale mapping. Every continuous section of pp induces a homeomorphism of B onto an open subset of E.

In fact, such a section is étale (corollary 1), hence open.

Corollary 4 01N2

Let p:E→Bp: E\rightarrow B be an étale and separated mapping. Suppose that E is connected and that pp admits a section. Then pp is a homeomorphism.

Let ss be a section of pp; since pp is étale, the image of ss is open in E (Corollary 3); it is also closed, since pp is separated (I, p. 27, Prop. 3). Since E is connected, we have s(B)=Es(B) = E and pp is a homeomorphism.

Proposition 7 01N3

Let p:E→Bp: E\rightarrow B be a continuous mapping. In order that the mapping pp be étale, it is necessary and sufficient that it be open and that the diagonal ΔE\Delta_E of E×BEE\times_BE be open in E×BEE\times_BE.

Suppose first that the mapping pp is étale. Then it is open and every point of E possesses a neighbourhood V such that p∣Vp|V is injective, which amounts to saying that one has (V×V)∩(E×BE)⊂ΔE(V\times V)\cap (E\times_BE)\subset \Delta_E. Hence ΔE\Delta_E is an open subset of E×BEE\times_BE.

Conversely, suppose that pp is an open mapping and that ΔE\Delta_E is an open subset of E×BEE\times_BE. Let xx be a point of E. Let V be an open neighbourhood of xx in E such that (V×V)∩(E×BE)(V\times V)\cap (E\times_BE) is contained in ΔE\Delta_E. Then p∣Vp|V is injective. By Remark 2, I, p. 29, the mapping pp is étale.

Proposition 8 01N4

Let

E′f′{E'}^{f'} E

p′pp'p

B′f{B'}^f B

be a cartesian square. If the mapping pp is étale, the mapping p′p' is étale. Conversely, if the mapping p′p' is étale and the mapping ff is universally strict and surjective, the mapping pp is étale.

Suppose that the mapping pp is étale. It is in particular open and p′p' is open (I, p. 17, Prop. 8). Consider then the cartesian square (I, p. 13, Example 1)

E′×B′E′φE×BE E'\times_{B'}{E'}^{\varphi}E\times_BE
q′qq'q

B′fB{B'}^fB. One has ΔE′=φ−1(ΔE)\Delta_{E'}=\overset{-1}{\varphi}(\Delta_E) (loc. cit.). Moreover, the diagonal ΔE\Delta_E is open in E×BEE\times_BE (Prop. 7), hence the diagonal ΔE′\Delta_{E'} is open in E′×B′E′E'\times_{B'}E'. This proves that the mapping p′p' is étale (loc. cit.).

Suppose now that the mapping ff is surjective and universally strict and that the mapping p′p' is étale. Then p′p' is open, hence pp is open (I, p. 21, Prop. 11, a)). On the other hand ΔE′\Delta_{E'} is an open subset of E′×B′E′E'\times_{B'}E'. Since ΔE′=φ−1(ΔE)\Delta_{E'}=\overset{-1}{\varphi}(\Delta_E) and the mapping φ\varphi is surjective and strict (I, p. 20, Def. 6), ΔE\Delta_E is open in E×BEE\times_BE. By Prop. 7, the mapping pp is étale.

Corollary 01N5

Let B be a topological space. The fibre product of two étale B-spaces is an étale B-space.

Let (E,p)(E, p) and (E′,p′)(E', p') be étale B-spaces. The mapping pr1:E×B_1: E\times_B E′→EE'\rightarrow E is étale (prop. 8), hence the mapping p∘p\circ pr1:E×BE′→B_1: E\times_BE'\rightarrow B is étale (prop. 6, a)).

Remark 3 01N6

Let

E′f′{E'}^{f'} E

p′pp'p

B′f{B'}^f B

be a cartesian square. If the mapping pp is étale and if the mapping ff is strict, the mapping f′f' is strict. In fact, under these hypotheses, every point of E possesses an open neighbourhood U such that pp induces a homeomorphism of U onto the open set p(U)p(U). The mapping p′p' then induces

a homeomorphism of (f−1′)(U)(\overset{-1}{f}')(U) onto f−1(p(U))\overset{-1}{f}(p(U)). The mapping ff induces

a strict mapping of f−1(p(U))\overset{-1}{f}(p(U)) onto p(U)p(U) (I, p. 20) and the mapping

f′f' therefore induces a strict mapping of (f−1′)(U)(\overset{-1}{f}')(U) into U. It follows that the mapping f′f' is strict ( I, p. 23, corollary 2).

3. Local sections of étale mappings

Let E and B be sets, A a subset of B and p:E→Bp: E\rightarrow B a mapping. A section of pp over A (or on A) is a mapping s:A→Es: A\rightarrow E such that p∘sp\circ s is the canonical injection of A into B. To give a section ss of pp over A amounts to giving a section of the mapping pA:p−1(A)→Ap_A:\overset{-1}{p}(A)\rightarrow A deduced from pp. If ss is a section of pp over A and if A′A' is a subset of A, the restriction s′s' of ss to A′A' is a section of pp over A′A'. One then says that ss is an extension of s′s' to A.

When E and B are topological spaces and p:E→p: E\rightarrow B a continuous mapping, the set CB(A;E)\mathscr{C}_B(A; E) (I, p. 2) of continuous sections of pp over A is also denoted by S(A;p)\mathscr{S}(A;p) or S(A;E)\mathscr{S}(A; E). Let ss be a continuous section of pp over A. The mapping ss induces a homeomorphism of A onto s(A)s(A), and pp induces the reciprocal homeomorphism. By definition 2, I, p. 28, we therefore have:

Proposition 9 01N7

Let p:E→Bp: E\rightarrow B be a continuous mapping. In order that the mapping pp be étale, it is necessary and sufficient that every point of E possess an open neighbourhood which is the image of a continuous section of pp over an open subset of B.

Remark 1 01N8

Let p:E→Bp: E\rightarrow B be a continuous and open mapping. In order that an open set U of E be the image of a continuous section of pp over an open set of B, it is necessary and sufficient that the restriction of pp to U be injective.

Remark 2 01N9

Let p:E→p: E\rightarrow B be a continuous mapping. Suppose that every point of B possesses an open neighbourhood V such that there exists a continuous section of pp over V. Such a mapping pp is not necessarily étale, nor even open. It is nevertheless surjective and universally strict (I, p. 20, corollary).

4. Continuous liftings of étale mappings

Let p:E→Bp: E\rightarrow B and f:Z→Bf: Z\rightarrow B be continuous mappings. A continuous lifting of ff to E means a continuous mapping g:Z→Eg: Z\rightarrow E such that p∘g=fp\circ g=f. The set of continuous liftings of ff to E is none other than CB(Z;E)\mathscr{C}_B(Z; E). It is also identified with the set S(Z;Z×BE)\mathscr{S}(Z; Z\times_BE) of continuous sections of the projection of the Z-space (Z×BE(Z\times_BE, pr1)_1).

If T is a subset of Z, a continuous lifting of ff to E defined on T means a continuous lifting of f∣Tf|T to E:

E↗↓pT⊂Z→ f B. \begin{array}{ccc} & & E\\ & \nearrow & \Big\downarrow{\scriptstyle p}\\ T\subset Z & \xrightarrow{\ f\ } & B. \end{array}

Proposition 10 01NA

Let E, B and Z be topological spaces, p:E→Bp: E\rightarrow B an étale mapping and f:Z→Bf: Z\rightarrow B a continuous mapping.

Let z∈Zz\in Z and x∈Ex\in E be points such that f(z)=p(x)f(z) =p(x). There exists a neighbourhood W of zz in Z and a continuous lifting gg of ff to E, defined on W, such that g(z)=xg(z) =x.

There exists an open neighbourhood V of p(x)p(x) in B and a continuous section ss of pp over V such that s(p(x))=xs(p(x)) =x (prop. 9). The set

W=f−1(V)W =\overset{-1}{f}(V) is an open neighbourhood of zz and the mapping g=s∘(f∣W)g=s\circ (f|W) is a continuous lifting of ff to E, defined on W and such that g(z)=xg(z) =x.

Proposition 11 01NB

Let E, B and Z be topological spaces, p:E→Bp: E\rightarrow B and f:Z→Bf: Z\rightarrow B continuous mappings. Let gg and g′g' be continuous liftings of ff to E. Let W be the set of points of Z where gg and g′g' coincide.

a) If the mapping pp is separated, W is closed.

b) If the mapping pp is étale, W is open.

Let us denote by hh the continuous mapping (g,g′):Z→E×E(g, g'): Z\rightarrow E\times E. The image of hh is contained in E×BEE\times_BE since p∘g=f=p∘g′p\circ g=f=p\circ g' and the set W of points of Z where gg and g′g' coincide is the inverse image under hh of the diagonal ΔE\Delta_E.

If the mapping pp is étale, the diagonal ΔE\Delta_E is open in E×BEE\times_BE (I, p. 31, prop. 7), therefore the set W is open in Z.

If the mapping pp is separated, the diagonal ΔE\Delta_E is closed in E×BEE\times_BE (I, p. 25, def. 1) and the set W is closed in Z.

Corollary 1 01NC

If the space Z is connected and if the mapping pp is étale and separated, continuous liftings of ff to E which coincide at one point are equal.

Corollary 2 01ND

Let p:E→Bp: E\rightarrow B be an étale and separated mapping. If the space E is connected, the group AutB(E)_B(E) operates freely on E.

In fact, if f:E→Ef: E\rightarrow E is a B-morphism, the set of points where ff and IdE_E coincide is equal to E or to the empty set by Cor. 1.

Corollary 3 01NE

Let E, B and Z be topological spaces, let p:E→Bp: E\rightarrow B be an étale and separated mapping, and let f:Z→Bf: Z\rightarrow B be a continuous mapping. For i=1i= 1, 2, let UiU_i be an open subset ( resp. closed subset) of Z and gi:Ui→Eg_i: U_i\rightarrow E a continuous lifting of ff defined on UiU_i. Suppose that the intersection U1∩U2U_1\cap U_2 is connected and that there exists a point zz of U1∩U2U_1\cap U_2 such that g1(z)=g2(z)g_1(z) =g_2(z). Then there exists a continuous lifting of ff defined on U1∪U2U_1\cup U_2 which extends g1g_1 and g2g_2.

By Corollary 1, the restrictions of g1g_1 and g2g_2 to U1∩U2U_1\cap U_2 are equal. The mapping g:U1∪U2→Eg: U_1\cup U_2\rightarrow E defined by g(z)=gi(z)g(z) =g_i(z) for z∈Ui(i=1z\in U_i(i= 1, 2) is a continuous lifting of ff defined on U1∪U2U_1\cup U_2 (TG, I, p. 19, Prop. 4).

In the preceding results, the particular case where Z = B and f=f= IdB_B is important: the continuous liftings of ff are then the continuous sections of pp.

Proposition 12 01NF

Let p:E→Bp: E\rightarrow B be an étale mapping. In order that the mapping pp be separated, it is necessary and sufficient that for every open subset V of B and every pair (s,s′)(s, s') of continuous sections of pp over V, the set of points where ss and s′s' coincide be closed in V.

The condition is necessary (Proposition 11, I, p. 34). Let us prove that it is sufficient. Let bb be a point of B, let xx and x′x' be two distinct points of E such that p(x)=p(x′)=bp(x) =p(x') =b. Let V be an open neighbourhood of bb and s,s′s,s' continuous sections of pp over V such that s(V)s(V) and s′(V)s'(V) are open neighbourhoods of xx and x′x' respectively (Prop. 9). By hypothesis, the set W of points x∈Vx\in V such that s(x)≠s′(x)s(x)=\not s'(x) is open in V, hence in B. The sets s(W)s(W) and s′(W)s'(W) are open neighbourhoods of xx and x′x' respectively; they are disjoint by construction. This proves that the mapping pp is separated (I, p. 25, Definition 1).

5. Construction of continuous sections of étale mappings

Theorem 1 01NG

Let X, Y, Z be topological spaces, let f:Z→f: Z\rightarrow X×YX\times Y be an étale and separated mapping, and let y0y_0 be a point of Y. Let s:X×Y→Zs: X\times Y\rightarrow Z be a section of ff. Suppose that the restriction of ss to X×{y0}X\times \{y_0\} is continuous, as are analogously the restrictions of ss to {x}×Y\{x\} \times Y for every x∈Xx\in X. If the space Y is connected and locally connected (TG, I, p. 84, def. 4), the mapping ss is continuous.

Lemma 01NH

Let U be an open set of X, let V be a connected open set of Y, and let (x1,y1)∈U×V(x_1, y_1)\in U\times V. Suppose that the restriction of ss to U×{y1}U\times \{y_1\} is continuous. Let σ\sigma be a continuous section of ff over U×VU\times V such that σ(x1,y1)=s(x1,y1)\sigma (x_1, y_1) =s(x_1, y_1). There exists a neighbourhood U′U' of x1x_1 such that s=σs=\sigma on U′×VU'\times V. In particular, ss is continuous in a neighbourhood of (x1,y1)(x_1, y_1).

Since the restriction of ss to U×{y1}U\times \{y_1\} is continuous and the mapping ff is étale, there exists a neighbourhood U′U' of x1x_1 such that s(x,y1)=σ(x,y1)s(x, y_1) =\sigma (x, y_1) for every x∈U′(x\in U'(I, p. 34, prop. 11, b)). Let x∈U′x\in U'; since the restriction of ss to {x}×V\{x\} \times V is continuous and the mapping ff is étale and separated, it follows from Corollary 1, I, p. 34, that σ(x,y)=s(x,y)\sigma (x, y) =s(x, y) for every y∈Vy\in V. Thus ss and σ\sigma coincide on U′×VU'\times V.

We now prove the theorem. First let us prove that the mapping ss is continuous in a neighbourhood of every point of X×{y0}X\times \{y_0\}. Let x0∈x_0\in X; one can choose an open neighbourhood U of x0x_0 in X, a connected open neighbourhood V of y0y_0 in Y, and a continuous section σ:U×V→Z\sigma : U\times V\rightarrow Z of ff such that σ(x0,y0)=s(x0,y0)\sigma (x_0, y_0) =s(x_0, y_0). By the above lemma, ss is continuous in a neighbourhood of (x0,y0)(x_0, y_0).

Let us now prove that the mapping ss is continuous at every point of X×YX\times Y. For x0∈Xx_0\in X, let us denote by C(x0)C(x_0) the set of points y∈y\in Y such that ss is continuous in a neighbourhood of (x0,y)(x_0, y). The set C(x0)C(x_0) is open in Y by definition, and contains y0y_0 by what precedes.

Let us prove that it is closed in Y. Consider a point y1y_1 of Y adherent to C(x0)C(x_0), an open neighbourhood U of x0x_0 in X, an open neighbourhood V of y1y_1 in Y, and a continuous section σ:U×V→Z\sigma : U\times V\rightarrow Z of ff such that σ(x0,y1)=s(x0,y1)\sigma (x_0, y_1) =s(x_0, y_1). Since the restriction of ss to {x0}×V\{x_0\} \times V is continuous and ff is étale, there exists an open neighbourhood V′V' of y1y_1 in Y such that σ(x0,y)=s(x0,y)\sigma (x_0, y) =s(x_0, y) for every y∈V′y\in V' (prop. 11 of I, p. 34). Since Y is locally connected, we may suppose that V′V' is connected. Since y1y_1 is adherent to C(x0)C(x_0), the set C(x0)∩V′C(x_0)\cap V' is not empty; let y2y_2 be a point of C(x0)∩V′C(x_0)\cap V'. By the lemma applied to (x0,y2)(x_0, y_2), there exists a neighbourhood U′U' of x0x_0 contained in U such that s=σs=\sigma over U′×V′U'\times V'. This proves that C(x0)C(x_0) is closed, for U′×V′U'\times V' is a neighbourhood of (x0,y1)(x_0, y_1). Since Y is connected, we have C(x0)=YC(x_0) = Y, and this for every x0∈Xx_0\in X, whence the theorem.

Corollary 1 01NI

Let X, Y, E, B be topological spaces. Let p:E→Bp: E\rightarrow B be an étale and separated mapping, let h:X×Y→Bh: X\times Y\rightarrow B be a continuous mapping and let y0y_0 be a point of Y. Let g:X×Y→Eg: X\times Y\rightarrow E be a mapping such that p∘g=hp\circ g=h. Suppose that the restrictions of gg to X×{y0}X\times \{y_0\} on the one hand and, for every x∈Xx\in X, to {x}×Y\{x\} \times Y on the other hand, are continuous. If the space Y is connected and locally connected, the mapping gg is continuous.

Let us denote by Z the fibre product (X×Y)×BE(X\times Y)\times_BE and by f:Z→X×Y,h′:Z→Ef: Z\rightarrow X\times Y,h': Z\rightarrow E the canonical projections. The mapping ff is étale (I, p. 31, prop. 8) and separated (I, p. 27, prop. 4). The mapping s:X×Y→s: X\times Y\rightarrow Z defined by s(x,y)=((x,y),g(x,y))s(x, y) = ((x, y), g(x, y)) is a section of ff satisfying the hypotheses of theorem 1. It is therefore continuous, as is g=h′∘sg=h'\circ s.

Remark 01NJ

If, in theorem 1, it is not supposed that the space Y is locally connected, the conclusion of the theorem is no longer necessarily true (I, p. 141, exerc. 7).

Theorem 2 01NK

Let B be a topological space and A a subspace of B. Make one of the following hypotheses:

(i) A admits a fundamental system of paracompact neighbourhoods;

(ii) B is paracompact and A closed;

(iii) B is metrizable;

(iv) A is compact and two distinct points of A possess disjoint neighbourhoods in B.

Then every continuous section over A of an étale mapping p:E→Bp: E\rightarrow B extends to a continuous section of pp over a neighbourhood of A.

Consider the following property (PCV) of the pair (B,A)(B,A):

(PCV) For every covering (Ui)i∈I(U_i)_{i\in I} of A by open subsets

of B, there exists a neighbourhood V of A and a locally

finite family (Fj)j∈J(F_j)_{j\in J} of closed subsets of V covering V, such that

each of the FjF_j is contained in one of the UiU_i.

Theorem 2 follows from Lemmas 1 and 3 below.

Lemma 1 01NL

Each of properties (i) to (iv) of theorem 2 implies property (PCV).

Let (Ui)i∈I(U_i)_{i\in I} be an open covering of A by open sets of B. Under hypothesis (i), there exists a paracompact neighbourhood V of A contained in the union of the UiU_i, an open covering (Uj′)j∈J(U'_j)_{j\in J} of the space V, locally finite and finer than the covering (V∩Ui)i∈I(V\cap U_i)_{i\in I}, and an open covering (Uj′′)j∈J(U''_j)_{j\in J} of the space V such that for every j∈Jj\in J, the closure FjF_j of Uj′′U''_j in V is contained in Uj′U'_j (TG, IX, p. 49, prop. 4 and p. 48, cor. 1 to th. 3). Hence property (PCV) in this case.

Condition (ii) implies condition (i) by Corollary 2 of TG, IX, p. 50. Analogously, condition (iii) implies condition (i): in fact, if B is metrisable, every neighbourhood of A is metrisable, hence paracompact (TG, IX, p. 51, th. 4).

Lastly, suppose that condition (iv) is satisfied, and let us show that property (PCV) holds. Since A is compact, it is enough to consider the case of a finite covering (Ui)0⩽i⩽n(U_i)_{0\leqslant i\leqslant n}. We then construct by induction open sets V0,...,VnV_0, . . . ,V_n of B satisfying the following conditions for 0⩽i⩽n:0\leqslant i\leqslant n:

α)A⊂V0∪⋯∪Vi∪Ui+1∪⋯∪Un \alpha )A\subset V_0\cup \cdots \cup V_i\cup U_{i+1}\cup \cdots \cup U_n
β)Vi‾∩A⊂Ui \beta )\overline{V_i}\cap A\subset U_i

Suppose that V0,...,Vr−1V_0, . . . ,V_{r-1} have been constructed satisfying the above conditions for 0⩽i⩽r−10\leqslant i\leqslant r-1, and let us construct VrV_r. The sets K = A∩∁UrA\cap \complement U_r and L=A∩∁(V0∪⋯∪Vr−1∪Ur+1∪⋯∪Un)L = A\cap \complement (V_0\cup \cdots \cup V_{r-1}\cup U_{r+1}\cup \cdots \cup U_n) are closed in A, hence compact. By virtue of (α)(\alpha ), they are disjoint. By hypothesis, for every point aa of L and every point bb of K, there exist disjoint neighbourhoods of aa and bb in B. By Lemma 2 below, there exist disjoint open sets VrV_r and W in B such that L⊂VrL\subset V_r and K⊂WK\subset W. From the inclusions L⊂VrL\subset V_r and A⊂V0∪⋯∪Vr−1∪Ur∪⋯∪UnA\subset V_0\cup \cdots \cup V_{r-1}\cup U_r\cup \cdots \cup U_n and from the definition of L, it follows that (α)(\alpha ) holds for i=ri=r. On the other hand, we have Vr‾∩K=∅\overline{V_r}\cap K =\emptyset, hence Vr‾∩A⊂Ur\overline{V_r}\cap A\subset U_r.

The set M=⋃0⩽i⩽n(Vi∩∁Ui)M =\bigcup_{0\leqslant i\leqslant n}(V_i\cap \complement U_i) is closed and does not meet A by (β)(\beta ). By (α)(\alpha ), the set V=((⋃0⩽i⩽nVi‾)−V =((\bigcup_{0\leqslant i\leqslant n}\overline{V_i})- M is a neighbourhood of A in B. For i=0,...,ni= 0, . . . , n, put Fi=V∩Vi‾:F_i= V\cap \overline{V_i}: this is a closed subset of V, contained in UiU_i. The family (Fi)0⩽i⩽n(F_i)_{0\leqslant i\leqslant n} is a covering of V, whence property (PCV).

Lemma 2 01NM

Let B be a topological space, and let K and L be quasi-compact subsets of B. Suppose that for every point aa of K and every point bb of L, there exist disjoint neighbourhoods of aa and bb in B. Then there exist two disjoint open sets U and V in B such that K⊂UK\subset U and L⊂VL\subset V.

Let aa be a point of K. For every b∈Lb\in L, let Ua,bU_{a,b} and Va,bV_{a,b} be disjoint open neighbourhoods of aa and of bb. For fixed aa, the family (Va,b)b∈L(V_{a,b})_{b\in L} is an open covering of L. Since this space is quasi-compact, there exists a finite family Ta⊂LT_a\subset L such that the union VbV_b of the Va,bV_{a,b} for b∈Tab\in T_a contains L. The intersection UaU_a of the Ua,bU_{a,b} for b∈Tab\in T_a is an open neighbourhood of aa, since TaT_a is finite; moreover, UaU_a and VaV_a are disjoint. The (Ua)a∈A(U_a)_{a\in A} form an open covering of K. Since K is quasi-compact, there exists a finite family S⊂KS\subset K such that U=⋃a∈SUaU =\bigcup_{a\in S}U_a contains K. Then V=⋂a∈SVaV =\bigcap_{a\in S}V_a is an open subset of B which contains L. The lemma is proved.

Lemma 3 01NN

Let B be a topological space and A a subspace of B such that the pair (B,A)(B,A) satisfies property (PCV) of I, p. 37. Then every continuous section over A of an étale mapping p:E→Bp: E\rightarrow B extends to a continuous section of pp over a neighbourhood of A.

Let p:E→Bp: E\rightarrow B be an étale mapping and s:A→Es: A\rightarrow E a continuous section of pp over A. For every point aa of A, there exists an open neighbourhood UaU_a of aa and a continuous section sa:Ua→Es_a: U_a\rightarrow E which coincides with ss on Ua∩AU_a\cap A (I, p. 34, Prop. 10 and I, p. 34, Prop. 11). The family (Ua)a∈A(U_a)_{a\in A} is a covering of A by open subsets of B. Let V be a neighbourhood of A in B, let (Fj)j∈J(F_j)_{j\in J} be a locally finite family of closed subsets of V covering V, and let a:J→Aa: J\rightarrow A be a mapping such that FjF_j is contained in Ua(j)U_{a(j)} for every j∈Jj\in J (property (PCV)). Let sjs_j denote the restriction of sa(j)s_{a(j)} to FjF_j. Let W be the set of points b∈Vb\in V satisfying sj(b)=sk(b)s_j(b) =s_k(b) for every pair (j,k)∈J×J(j, k)\in J\times J such that b∈Fj∩Fkb\in F_j\cap F_k. One defines a section s′:W→Es': W\rightarrow E by s′(b)=sj(b)s'(b) =s_j(b) if b∈Fjb\in F_j. We have A⊂WA\subset W and s′∣A=ss'|A =s. The section s′s' is continuous by Prop. 4 of I, p. 19.

It remains to prove that W is a neighbourhood of A in B. For every pair (j,k)∈J×J(j, k)\in J\times J, let TjkT_{jk} denote the set of points b∈Fj∩Fkb\in F_j\cap F_k such that sj(b)≠sk(b)s_j(b)=\not s_k(b). The set TjkT_{jk} is closed in Fj∩FkF_j\cap F_k (Prop. 11, b) of I, p. 34), hence in V, and the sets TjkT_{jk} form a locally finite family of subsets of V. The union T of the family (Tjk)(T_{jk}) is therefore a closed set of V (TG, I, p. 6, Prop. 4). Now W is, by definition, the complement of T in V. It is therefore a neighbourhood of A in V and consequently in B.

6. Upper bound for the cardinal of the fibres of an étale and separated mapping

Theorem 3 01NO

Let E and B be topological spaces and let p:E→Bp: E\rightarrow B be an étale and separated mapping. Suppose that E is connected and that B is locally connected. Let W\mathscr{W} be a basis for the topology of B. Then, for every point aa of B, one has Card(Ea_a)⩽\leqslant sup(Card(W\mathscr{W}), Card(N\mathbf{N})).

Since the space B is locally connected, its topology has a basis W′\mathscr{W}' consisting of connected open sets, for example the connected components of the open sets of W\mathscr{W} (TG, I, p. 85, Prop. 11). By virtue of Lemma 4 below, there exists a basis V\mathscr{V} for the topology of B consisting of connected open sets and such that Card(V\mathscr{V})⩽\leqslant Card(W\mathscr{W})2^2.

Lemma 4 01NP

Let B be a topological space, and let W\mathscr{W} and W′\mathscr{W}' be bases for the topology of B. There exists a subset V\mathscr{V} of W′\mathscr{W}' which is a basis for the topology of B and such that Card(V\mathscr{V})⩽\leqslant Card(W\mathscr{W})2^2.

Let A⊂W×W\mathscr{A}\subset \mathscr{W}\times \mathscr{W} be the set of pairs (W1,W2)(W_1,W_2) for which there exists W′∈W′W'\in \mathscr{W}' such that W1⊂W′⊂W2W_1\subset W'\subset W_2; let φ:A→W′\varphi :\mathscr{A}\rightarrow \mathscr{W}' be a mapping such that one has W1⊂φ(W1,W2)⊂W2W_1\subset \varphi (W_1,W_2)\subset W_2 for every pair (W1,W2)∈A(W_1,W_2)\in \mathscr{A}, and let V⊂W′\mathscr{V}\subset \mathscr{W}' be the image of φ\varphi. One has

Card(V\mathscr{V})⩽\leqslant Card(A\mathscr{A})⩽\leqslant Card(W×W\mathscr{W}\times \mathscr{W})

(E, III, p. 25, Prop. 3). Let us prove that V\mathscr{V} is a basis for the topology of B. Let xx be a point of B and U a neighbourhood of xx. By hypothesis, xx has a neighbourhood W2∈WW_2\in \mathscr{W} contained in U, a neighbourhood W′∈W′W'\in \mathscr{W}' contained in W2W_2 and a neighbourhood W1∈WW_1\in \mathscr{W} contained in W′W'. Thus W1⊂W′⊂W2⊂UW_1\subset W'\subset W_2\subset U. Then (W1,W2)∈A(W_1,W_2)\in \mathscr{A} and φ(W1,W2)\varphi (W_1,W_2) is a neighbourhood of xx contained in U. By Prop. 3 of TG, I, p. 5, the set V\mathscr{V} is a basis for the topology of B.

Let V\mathscr{V} be a basis for the topology of B consisting of nonempty connected open sets and such that Card(V\mathscr{V})⩽\leqslant Card(W\mathscr{W})2^2. Let U\mathscr{U} be the set of open sets U of E such that pp induces a homeomorphism of U onto an open set V belonging to V\mathscr{V}. Every element of U\mathscr{U} is a connected nonempty open set. Since the mapping pp is étale, by Definition 2 (I, p. 28) and Prop. 3 of TG, I, p. 5, the set U\mathscr{U} is a basis for the topology of E. Let us call a caterpillar any finite sequence (U0,...,Un)(U_0, . . . ,U_n) of elements of U\mathscr{U} such that for 1⩽i⩽n,Ui−1∩Ui1\leqslant i\leqslant n,U_{i-1}\cap U_i is nonempty and p(Ui−1)∩p(Ui)p(U_{i-1})\cap p(U_i) is connected. Let S denote the set of finite sequences of elements of V\mathscr{V} and, for every caterpillar c=(U0,...,Un)c= (U_0, . . . ,U_n), let us denote by p(c)p(c) the sequence (p(U0),...,p(Un))(p(U_0), . . . , p(U_n)).

Lemma 5 01NQ

a) Let c=(U0,...,Un)c= (U_0, . . . ,U_n) and c′=(U0′,...,Um′)c'= (U'_0, . . . ,U'_m) be caterpillars such that U0=U0′U_0= U'_0 and p(c)=p(c′)p(c) =p(c'). Then c=c′c=c'.

b) Let U and U′U' be elements of U\mathscr{U}. Then there exists a caterpillar c=(U0,...,Un)c= (U_0, . . . ,U_n) such that U0=UU_0= U and Un=U′U_n= U'.

a) Necessarily, m=nm=n. For every integer ii such that 0⩽i⩽n0\leqslant i\leqslant n, put Vi=p(Ui)V_i=p(U_i) and let sis_i and si′s'_i denote the continuous sections of pp over ViV_i with respective images UiU_i and Ui′U'_i. To prove the equality c=c′c=c', we shall prove, by induction on ii, the equality si=si′s_i=s'_i for every integer ii such that 0⩽i⩽n0\leqslant i\leqslant n. By hypothesis, one has s0=s0′s_0=s'_0. Let ii be such that 1⩽i⩽n1\leqslant i\leqslant n and si−1=si−1′s_{i-1}=s'_{i-1}. Since Ui−1∩UiU_{i-1}\cap U_i contains a point xx, the continuous sections si−1s_{i-1} and sis_i coincide at p(x)p(x), hence at every point of Vi−1∩ViV_{i-1}\cap V_i, since it is connected (Corollary 1 of I, p. 34). The same is true of si−1′s'_{i-1} and si′s'_i. For the same reason, sis_i and si′s'_i, which coincide on Vi−1∩ViV_{i-1}\cap V_i, also coincide on ViV_i, whence the result.

b) If xx and yy are points of E, we say that a caterpillar c=(U0,...,Un)c= (U_0, . . . ,U_n) joins xx to yy if x∈U0x\in U_0 and y∈Uny\in U_n. In the set E, let R be the relation “there exists a caterpillar which joins xx to yy”. The relation R is reflexive since U\mathscr{U} is a covering of E. It is obviously symmetric. Let us prove that it is transitive: let x,y,zx,y,z be three points of E, (U0,...,Un)(U_0, . . . ,U_n) and (U0′,...,Um′)(U'_0, . . . ,U'_m) caterpillars joining xx to yy and yy to zz respectively. Let U∈UU\in \mathscr{U} be a neighbourhood of yy contained in Un∩U0′U_n\cap U'_0; we have p(Un)∩p(U)=p(U0′)∩p(U)=p(U)p(U_n)\cap p(U) =p(U'_0)\cap p(U) =p(U), and, since U is connected , the sequence (U0,...,Un,U,U0′,...,Um′)(U_0, . . . ,U_n,U,U'_0, . . . ,U'_m) is a caterpillar which joins xx to zz. The equivalence classes under R are open, hence also closed. Since E is connected, it follows that any two points of E are always joined by a caterpillar.

Let now U and U′U' be elements of U,x\mathscr{U},x a point of U and x′x' a point of U′U'. There exists a caterpillar (U2,...,Un−2)(U_2, . . . ,U_{n-2}) joining xx to x′x'. Let U1U_1 and Un−1U_{n-1} be open sets of U\mathscr{U} such that x∈U1,x′∈Un−1x\in U_1,x'\in U_{n-1}, U1⊂U∩U2,Un−1⊂U′∩Un−2U_1\subset U\cap U_2,U_{n-1}\subset U'\cap U_{n-2}. Put U0=UU_0= U, Un=U′U_n= U'. Then the sequence (U0,...,Un)(U_0, . . . ,U_n) is a caterpillar.

Let us now prove the theorem. Let U be an element of U\mathscr{U} and let C be the set of caterpillars (U0,...,Un)(U_0, . . . ,U_n) such that U0=UU_0= U. The mapping of C into S which, to a caterpillar c=(U0,...,Un)c= (U_0, . . . ,U_n) of C, associates the sequence p(c)=(p(U0),...,p(Un))p(c) = (p(U_0), . . . , p(U_n)) is injective (lemma 5, a)), hence

(1) Card(C) ⩽\leqslant Card(S).

The mapping of C into U\mathscr{U} which, to c=(U0,...,Un)∈Cc= (U_0, . . . ,U_n)\in C, associates the open set UnU_n is surjective by the second part of lemma 5, hence (2) Card(U\mathscr{U})⩽\leqslant Card(C).

For every point xx of E, let us choose an open set UxU_x of U\mathscr{U} such that x∈Uxx\in U_x. If xx and yy are distinct points of one and the same fibre EaE_a of pp, we have Ux=U̸yU_x= U\not_y since p∣Uxp|U_x is injective. Hence, for a∈Ba\in B, (3) Card(Ea_a)⩽\leqslant Card(U\mathscr{U}).

Finally, if V\mathscr{V} is a finite set, the set S of finite sequences of elements of V\mathscr{V} is countable (E, III, p. 49, prop. 1), hence

(4) Card(S) ⩽\leqslant Card(N\mathbf{N}).

If V\mathscr{V} is infinite, we have

(5) Card(S) = Card(V\mathscr{V})⩽\leqslant Card(W\mathscr{W})2=^2= Card(W\mathscr{W})

by the corollary of E, III, p. 50 and Corollary 1 of E, III, p. 49. The theorem follows from relations (1) to (5) above.

Remark 01NR

From the foregoing, if the topology of B admits a countable basis, the same is true of that of E and the fibres of E are countable (cf. TG, I, p. 88, Poincaré-Volterra theorem).

Exercises

See the exercises for § 2.

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