Bourbaki

6. Topological groups with operators; topological rings, division rings and fields

1. Topological groups with operators

On a set GG, a structure of a group with operators and a topology are said to be compatible if the topology and the group structure of GG are compatible (§ 1, no. 1) and if in addition the endomorphisms of GG produced by the operators are continuous. The set GG, together with the given topology and structure of group with operators, is then said to be a topological group with operators.

If HH is a stable subgroup of a topological group GG with operators, then the topology induced on HH by the topology of GG is compatible with the structure of group with operators on HH. Furthermore:

Proposition 1

If HH is a stable subgroup of a topological group GG with operators, then the closure H‾\overline{H} of HH in GG is a stable subgroup of GG.

We know already (§ 2, no. 1, Proposition 1) that H‾\overline{H} is a subgroup of GG. Also if α\alpha is any operator on GG, the image of HH under the continuous mapping x→xαx \to x^\alpha is contained in HH, and therefore the image of H‾\overline{H} is contained in H‾\overline{H} (Chapter I, § 2, no. 1, Theorem 1).

Let HH be a stable normal subgroup of a topological group GG with operators. Then for each operator α\alpha on GG, the mapping of G/HG/H into itself induced by x→xαx \to x^\alpha is continuous (§ 2, no. 8, Remark 3), and the structure of group with operators on G/HG/H is therefore compatible with the quotient topology on G/HG/H.

Let (Gi)i∈I(G_i)_{i \in I} be a family of topological groups with operators, where each GiG_i is assumed to have the same operator set Ω\Omega. For each α∈Ω\alpha \in \Omega, the mapping x→((prix)α)x \to ((\mathrm{pr}_i x)^\alpha) of G=∏i∈IGiG = \prod_{i \in I} G_i into itself is continuous (Chapter I, § 4, no. 1, Proposition 1), and the structure of groups with operators on GG is therefore compatible with the product topology on GG.

If GG is a Hausdorff topological group with operators and if GG has a completion G^\hat{G} (§ 3, no. 4), then every endomorphism x→xαx \to x^\alpha of GG defined by an operator on GG can be extended by continuity to an endomorphism of G^\hat{G} (§ 3, no. 3, Proposition 5). Hence G^\hat{G} has the structure of a topological group with operators, and the operator set is the same for G^\hat{G} as for GG.

2. Topological direct sum of stable subgroups

Since the study of commutative groups with operators is equivalent to the study of modules we shall sometimes allow ourselves to use the terminology proper to modules for arbitrary commutative groups with operators; thus we may speak of linear mappings instead of homomorphisms of commutative groups with operators, and we may use the word projector to denote an idempotent endomorphism of a commutative group with operators.

If a commutative topological group EE with operators (written additively) is the direct sum of a finite family (Mi)1≤i≤n(M_i)_{1 \leq i \leq n} of stable subgroups, then the canonical bijection (xi)→∑i=1nxi(x_i) \to \sum_{i=1}^n x_i of the product group ∏i=1nMi\prod_{i=1}^n M_i onto EE is continuous, but is not necessarily a homeomorphism.

Definition 1

Let EE be a commutative topological group with operators, and let (Mi)1≤i≤n(M_i)_{1 \leq i \leq n} be a finite family of stable subgroups of EE, such that EE is the direct sum of the MiM_i. Then EE is said to be the topological direct sum of the MiM_i if the canonical mapping (xi)→∑i=1nxi(x_i) \to \sum_{i=1}^n x_i of the product group ∏i=1nMi\prod_{i=1}^n M_i onto EE is a homeomorphism* (and therefore an isomorphism of topological groups with operators).

Proposition 2

Let EE be a commutative topological group with operators which is the direct sum of stable subgroups MiM_i (1≤i≤n1 \leq i \leq n). Let (pi)1≤i≤n(p_i)_{1 \leq i \leq n} be the family of projectors associated with the decomposition E=∑i=1nMiE = \sum_{i=1}^n M_i. Then EE is the topological direct sum of the MiM_i if and only if the pip_i are continuous.

For the mapping x→(pi(x))x \to (p_i(x)) is the inverse of the mapping (xi)→∑i=1nxi(x_i) \to \sum_{i=1}^n x_i.

Since 1=∑i=1nPi1 = \sum_{i=1}^n P_i (where 11 denotes the identity mapping of EE) it is sufficient for n−1n - 1 of the projectors pip_i to be continuous in order that the nn th should also be continuous.

If EE is the topological direct sum of two stable subgroups M,NM, N, then NN is said to be a topological complement of MM in EE; this is the case if and only if the canonical mapping of E/ME/M onto is an isomorphism of topological groups with operators.

Corollary

Let EE be a commutative topological group with operators, and let MM be a stable subgroup of EE. Then the following conditions are equivalent:

a) MM has a topological complement in EE.

b) There is a continuous projector pp of EE into EE such that p(E)=Mp(E) = M.

c) The identity mapping of MM can be extended to a continuous linear mapping of EE onto MM.

It follows from Proposition 2 that a) implies b), and it is clear that b) implies c). Finally, if pp is a continuous linear mapping of EE onto MM which extends the identity mapping of MM, then pp is a continuous projector, and the projectors pp and 1−p1 - p are associated with the direct sum decomposition E=M+NE = M + N, where N=p‾(0)N = \overline{p}(0).

Remark 1

To avoid confusion, we shall sometimes say that a stable subgroup of EE which is a complement of MM (in the sense of the structure of group with operators, without a topology) is an algebraic complement of MM.

Remark 2

If a Hausdorff commutative topological group with operators is the topological direct sum of a family (Mi)1≤i≤n(M_i)_{1 \leq i \leq n} of stable subgroups, then each of the subgroups MiM_i is closed in EE, for MiM_i is the set of all x∈Ex \in E such that pi(x)=xp_i(x) = x (Chapter I, § 8, no. 1, Proposition 2).

Proposition 3

Let E,FE, F be two commutative topological groups with operators, and let uu be a continuous linear mapping of EE into FF. In order that there should exist a continuous linear mapping vv of FF into EE such that u∘vu \circ v is the identity mapping of FF (in which case uu is said to be right invertible and vv is said to be a right inverse of uu) it is necessary and sufficient that uu is a strict morphism (§ 2, no. 8) of EE onto FF and that u‾(0)\overline{u}(0) has a topological complement in EE.

The conditions are necessary. For we have then u(v(F))=Fu(v(F)) = F and a fortiori u(E)=Fu(E) = F; furthermore, if p=v∘up = v \circ u, then pp is a continuous linear mapping of EE into itself such that p2=pp^2 = p; therefore (Corollary to Proposition 2) p(E)=v(u(E))=v(F)p(E) = v(u(E)) = v(F) has a topological complement p‾(0)\overline{p}(0) in EE; but since u(p(x)))=u(x)u(p(x))) = u(x) by hypothesis, we have u‾(0)=p‾(0)\overline{u}(0) = \overline{p}(0). Finally, the bijective mapping of E/u‾(0)E/\overline{u}(0) onto FF, associated with uu, is the composition of the bijective mapping of E/u‾(0)E/\overline{u}(0) onto v(F)v(F), associated with pp, and the restriction of uu to v(F)v(F); since vv is continuous, both these mappings are isomorphisms, and therefore uu is a strict morphism of EE onto FF.

The conditions are sufficient. For if φ\varphi is the canonical homomorphism of E onto E/u‾1(o)E/\overline{u}^1(o), to say that u‾1(o)\overline{u}^1(o) has a topological complement M in E is to say that the restriction of φ\varphi to M is an isomorphism of M onto E/u‾1(o)E/\overline{u}^1(o). Since on the other hand u=w∘φu = w \circ \varphi, where w is an isomorphism of E/u‾1(o)E/\overline{u}^1(o) onto F, we see that the restriction of u to M is an isomorphism of M onto F, and the inverse isomorphism v is therefore such that u∘vu \circ v is the identity mapping of F onto itself.

Proposition 4

Let E, F be two commutative topological groups with operators, and let u be a continuous linear mapping of E into F. In order that there should exist a continuous linear mapping v of F into E such that v∘uv \circ u is the identity mapping of E onto itself* (in which case u is said to be left invertible and v is said to be a left inverse of u) *it is necessary and sufficient that u is a (topological) isomorphism of E onto u(E), and that u(E) has a topological complement in F.

The conditions are sufficient, for if they are satisfied we obtain a left inverse v of u by taking the composition of the isomorphism of u(E) onto E which is the inverse of u, with a continuous projector of F onto u(E).

The conditions are necessary. For the relation v(u(x))=xv(u(x)) = x shows that u‾1(o)={o}\overline{u}^1(o) = \{ o \}; u is therefore a bijection of E onto u(E), and since the restriction of v to u(E) is continuous, it follows that u is an isomorphism of E onto u(E). On the other hand, if we put q=u∘vq = u \circ v, then q is a continuous linear mapping of F onto u(E) such that q2=qq^2 = q, which proves (Corollary to Proposition 2) that u(E) has a topological complement in F.

3. Topological rings

Definition 2

A topological ring is a set A which carries a ring structure and a topology satisfying the following axioms:

(ATI_\mathrm{I}). The mapping (x,y)→x+y(x, y) \to x + y of A×AA \times A into A is continuous. (ATII_\mathrm{II}). The mapping x→−xx \to -x of A into A is continuous. (ATIII_\mathrm{III}). The mapping (x,y)→xy(x, y) \to xy of A×AA \times A into A is continuous.

The first two axioms express that the topology of A is compatible with its additive group structure (§ 1, no. 1).

If a ring structure and a topology are given on a set A, they are said to be compatible if they satisfy axioms (ATI_\mathrm{I}), (ATII_\mathrm{II}) and (ATIII_\mathrm{III}).

Example 1

On any ring A, the discrete topology is compatible with the ring structure. A topological ring whose topology is discrete is called a discrete ring.

Example 2

We shall see in Chapter IV that the topology of the rational line Q (resp. the real line R) is compatible with the ring structure of Q (resp. R).

In a topological ring every left homothety x→axx \to ax (resp. every right homothety x→xax \to xa) is continuous (and is a homeomorphism if aa is a unit of AA).

Since we can write identically

xy−x0y0=(x−x0)(y−y0)+(x−x0)y0+x0(y−y0) xy - x_0 y_0 = (x - x_0)(y - y_0) + (x - x_0)y_0 + x_0(y - y_0)
the axiom (AT_{III}) [in view of (AT_I) and (AT_{II})] is equivalent to the conjunction of the following two axioms: (AT_{III a}). Given any x0∈Ax_0 \in A, the mappings x→x0xx \to x_0 x and x→xx0x \to xx_0 are continuous at the point x=0x = 0. (AT_{III b}). The mapping (x,y)→xy(x, y) \to xy of A×AA \times A into AA is continuous at the point (0,0)(0, 0).

From this we can deduce a necessary and sufficient set of conditions which the filter B\mathfrak{B} of neighbourhoods of 00 in a ring AA must satisfy in order to define a topology on AA compatible with its ring structure: B\mathfrak{B} must satisfy axioms (GA_I) and (GA_{II}) of § 1, and also the following two axioms: (AV_I). For all x0∈Ax_0 \in A and all V∈BV \in \mathfrak{B}, there exists W∈BW \in \mathfrak{B} such that x0W⊂Vx_0 W \subset V and Wx0⊂VW x_0 \subset V. (AV_{II}). For all V∈BV \in \mathfrak{B}, there exists W∈BW \in \mathfrak{B} such that WW⊂VWW \subset V.

Remark

In analysis one fairly often meets rings which satisfy axioms (AT_I), (AT_{II}) and (AT_{III a}), but not (AT_{III b}). *An example is the ring of measures on a compact group, where multiplication is convolution and the topology is the vague topology.

Example 3).* Let B\mathfrak{B} be a filter base on a ring AA, consisting of two-sided ideals. B\mathfrak{B} is a fundamental system of neighbourhoods of 00 for a topology compatible with the additive group structure of AA, and it follows immediately from (AV_I) and (AV_{II}) that this topology is compatible with the ring structure of AA.

Let XX be a topological space, and let ff and gg be two mappings of XX into a topological ring AA. If ff and gg are continuous at a point x0∈Xx_0 \in X, then f+gf + g, −f-f and fgfg are continuous at this point. It follows that the continuous mappings of XX into AA form a subring of the ring AXA^X of all mappings of XX into AA. We see also that, if AA is commutative, then every polynomial in n variables, with coefficients in AA and defined on AnA^n, is continuous on AnA^n. Again, let ff and gg be two mappings of a set XX, filtered by a filter F\mathfrak{F}, into a Hausdorff topological ring AA; if lim⁡Ff\lim_{\mathfrak{F}} f and lim⁡Fg\lim_{\mathfrak{F}} g exist, then so do lim⁡F(f+g)\lim_{\mathfrak{F}} (f + g), lim⁡F(−f)\lim_{\mathfrak{F}} (-f) and lim⁡F(fg)\lim_{\mathfrak{F}} (fg), and we have (Chapter I, § 7, no. 4, Proposition 9, Corollary 1, and § 8, no. 1, Proposition 1)

(1)

lim⁡F(f+g)=lim⁡Ff+lim⁡Fg, \lim_{\mathfrak{F}} (f + g) = \lim_{\mathfrak{F}} f + \lim_{\mathfrak{F}} g,
(2)
lim⁡F(−f)=−lim⁡Ff, \lim_{\mathfrak{F}} (-f) = -\lim_{\mathfrak{F}} f,
(3)
lim⁡F(fg)=(lim⁡Ff)(lim⁡Fg). \lim_{\mathfrak{F}} (fg) = (\lim_{\mathfrak{F}} f) (\lim_{\mathfrak{F}} g).

4. Subrings; ideals; quotient rings; products of rings

If HH is a subring of a topological ring AA, then the topology induced on HH by that of AA is compatible with the ring structure of HH. The topological ring structure thus defined on HH is said to be induced by that of AA.

Proposition 5

Let HH be a dense subring of a topological ring AA, and let KK be a subring (resp. left ideal, right ideal, two-sided ideal) of HH. Then the closure K‾\overline{K} of KK in AA is a subring (resp. left ideal, right ideal, two-sided ideal) of AA.

The proof is the same as for Proposition 8 of § 2, n. 3 : if for example KK is a left ideal in HH, then the mapping (z,x)→zx(z, x) \to zx is continuous on A×AA \times A and maps H×KH \times K into KK; hence it maps A×K=H‾×KA \times K = \overline{H} \times K into H‾\overline{H}.

Let HH be a two-sided ideal in a topological ring AA. By the same argument as for quotient groups, we see that the quotient of the topology of AA by the relation x−y∈Hx - y \in H is compatible with the ring structure of A/HA/H. In particular, if AA is not Hausdorff, then the closure NN of {0}\{0\} in AA is a closed two-sided ideal, by Proposition 5; the quotient ring, which is Hausdorff (§\S 2, no. 5, Proposition 13) is called the Hausdorff ring associated with AA.

Let (Ai)i∈I(A_i)_{i \in I} be a family of topological rings. On the set A=∏i∈IAiA = \prod_{i \in I} A_i, the product of the topologies of the AiA_i is compatible with the product of the ring structures of the AiA_i (same proof as for product groups); the topological ring AA thus defined is called the product of the topological rings AiA_i.

5. Completion of a topological ring

When we speak of the uniformity of a topological ring, we always mean the uniformity of its additive group, unless the contrary is expressly stated; in particular, AA is said to be a complete ring if the additive group of AA is complete.

Let AA be a Hausdorff topological ring; as an additive group AA can be considered as a dense subgroup of a complete Hausdorff commutative group A^\hat{A}, which is determined up to isomorphism (§\S 3, no. 5, Theorem 2). In order that we should be able to consider AA as a subring of a complete ring, it is necessary to be able to extend the function xyxy by continuity to the space A^×A^\hat{A} \times \hat{A}. The possibility of this extension will follow from the following more general theorem :

Theorem 1

Let E, F, G be three complete Hausdorff commutative groups; let A be a dense subgroup of E and let B be a dense subgroup of F. If f is a Z\mathbf{Z} continuous Z\mathbf{Z}-bilinear (*) mapping of A×BA \times B into G, then f can be extended by continuity to a continuous Z\mathbf{Z}-bilinear mapping of E×FE \times F into G.

Let (x0,y0)(x_0, y_0) be an arbitrary point of E×FE \times F, and let U,B\mathfrak{U}, \mathfrak{B} be the traces on A, B respectively of the neighbourhood filters of x0,y0x_0, y_0 (U,B\mathfrak{U}, \mathfrak{B} are filters, by hypothesis). To show that f can be extended by continuity, it is enough to show that f(U×B)f(\mathfrak{U} \times \mathfrak{B}) is a Cauchy filter-base on G (Chapter II, § 3, no. 6, Proposition 11). Consider the identity

f(x′,y′)−f(x,y)=f(x′−x,y1)+f(x1,y′−y)+f(x′−x,y′−y1). f(x', y') - f(x, y) = f(x' - x, y_1) + f(x_1, y' - y) + f(x' - x, y' - y_1).

We shall show that by taking (x,y)(x, y) and (x′,y′)(x', y') in a sufficiently small set of U×B\mathfrak{U} \times \mathfrak{B}, and by choosing x1x_1 and y1y_1 suitably, we can make each of the terms on the right-hand side very small. Let W be any neighbourhood of 0 in G; since f is continuous at (0,0)∈A×B(0, 0) \in A \times B, there is a set U∈BU \in \mathfrak{B} and a set V∈BV \in \mathfrak{B} such that f(x′−x,y′−y)∈Wf(x' - x, y' - y) \in W whenever x∈U,x′∈U,y∈V,y′∈Vx \in U, x' \in U, y \in V, y' \in V. Take a point x1∈Ux_1 \in U and a point y1∈Vy_1 \in V; then for all x,x′x, x' in U and all y,y′y, y' in V we shall have

f(x′−x,y′−y1)+f(x−x1,y′−y)∈W+W. f(x' - x, y' - y_1) + f(x - x_1, y' - y) \in W + W.

On the other hand, the partial mapping x→f(x,y1)x \to f(x, y_1) is continuous on A; hence there is a set U′⊂UU' \subset U, belonging to U\mathfrak{U}, and such that, whenever xx and x′x' belong to U′U', we have f(x′−x,y1)∈Wf(x' - x, y_1) \in W. Likewise there exists V′⊂VV' \subset V belonging to B\mathfrak{B} such that, whenever yy and y′y' belong to V′V', we have f(x1,y′−y)∈Wf(x_1, y' - y) \in W. Consequently, if (x,y)(x, y) and (x′,y′)(x', y') are any two points of U′×V′U' \times V', then

f(x′,y′)−f(x,y)∈W+W+W+W; f(x', y') - f(x, y) \in W + W + W + W;

this proves the existence of the extension f‾\overline{f} of f. The fact that f‾\overline{f} is Z\mathbf{Z}-bilinear is an immediate consequence of the principle of extension of identities (Chapter I, § 8, no. 1, Corollary 1 to Proposition 2).

Q.E.D.

In the application of this theorem to a Hausdorff topological ring A, we take E, F and G to be A^\hat{A}, A and B to be the ring A, and f to be

(*) ff is said to be Z\mathbf{Z}-bilinear if, for all elements x,x′x, x' of A and all elements y,y′∈By, y' \in B we have

f(x+x′,y)=f(x,y)+f(x′,y) f(x + x', y) = f(x, y) + f(x', y)
and
f(x,y+y′)=f(x,y)+f(x,y′). f(x, y + y') = f(x, y) + f(x, y').

the Z\mathbf{Z}-bilinear mapping (x,y)→xy(x, y) \to xy, which by hypothesis is continuous. We denote again by xyxy the value of the extended function on A^×A^\hat{A} \times \hat{A}; this function is a law of composition on A^\hat{A}, and to say that it is Z\mathbf{Z}-bilinear means that it is distributive on both sides with respect to addition; and it is also associative, by the principle of extension of identities. Consequently:

Proposition 6

A Hausdorff topological ring* AA is isomorphic to a dense subring of a complete Hausdorff ring A^\hat{A}, which is determined up to isomorphism (and is called the completion of AA).

If AA is commutative (resp. has an identity element) then the same is true of A^\hat{A} (principle of extension of identities).

Let AA be a topological ring, not necessarily Hausdorff; let NN be the closure of {0}\{0\} in AA, and let A′=A/NA' = A/N be the Hausdorff ring associated with AA. Then A^′\hat{A}', the completion of A′A', is called the Hausdorff completion of AA and is also denoted by A^\hat{A}. One shows as in § 3, no. 4, Proposition 8 that every continuous ring homomorphism uu of AA with a complete Hausdorff topological ring CC can be uniquely factorized as u=v∘φu = v \circ \varphi, where vv is a continuous homomorphism of A^\hat{A} into CC and φ\varphi is the canonical mapping of AA into A^\hat{A}. If A,BA, B are two topological rings, and u:A→Bu : A \to B is a continuous homomorphism, there is therefore a unique continuous homomorphism u^:A^→B^\hat{u} : \hat{A} \to \hat{B} such that the diagram

A→uBφ↓ψ↓A^→u^B^ \begin{array}{ccc} A & \xrightarrow{u} & B \\ \varphi \downarrow & & \psi \downarrow \\ \hat{A} & \xrightarrow{\hat{u}} & \hat{B} \end{array}

is commutative (φ\varphi and ψ\psi being the canonical mappings); we have only to apply the preceding result to ψ∘u\psi \circ u.

6. Topological modules

Definition 3

Given a topological ring* AA with an identity element, a topological left AA*-module is a set* EE, together with:

1) a left AA*-module structure*; 2) a topology compatible with the additive group structure of EE, and satisfying the following axiom: (MT) The mapping (λ,x)→λx(\lambda, x) \to \lambda x of A×EA \times E into EE is continuous.

We define similarly the notion of a topological right AA-module; since every right AA-module can be considered as a left A0A^0-module, where A0A^0 is the opposite ring of AA, and since the topology of AA is compatible with the ring structure of A0A^0, there is no need to distinguish between topological right AA-modules and topological left A0A^0-modules.

Example 1

A topological vector space over R\mathbf{R} (resp. C\mathbf{C}) is a topological R\mathbf{R} (resp. C\mathbf{C}) module.

Example 2

Let AA be a ring and let B\mathfrak{B} be a filter base on AA consisting of two-sided ideals of AA; let EE be a left AA-module. If we give AA the topology (compatible with its ring structure) for which B\mathfrak{B} is a fundamental system of neighbourhoods of 00 (no. 3, Example 3), and EE the topology (compatible with its additive group structure) in which the sets aEaE, as aa runs through B\mathfrak{B}, form a fundamental system of neighbourhoods of 00 (§1§ 1, no. 2, Example), it is immediately verified that EE is a topological AA-module.

Remark

Given a topological ring AA, consider, on a left AA-module EE, a topology compatible with the additive group structure of EE. By virtue of the identity

λx−λ0x0=(λ−λ0)x0+λ0(x−x0)+(λ−λ0)(x−x0) \lambda x - \lambda_0 x_0 = (\lambda - \lambda_0)x_0 + \lambda_0(x - x_0) + (\lambda - \lambda_0)(x - x_0)
the axiom (MT) is equivalent to the conjunction of the following three axioms: (MTI′_I'). For each x0∈Ex_0 \in E, the mapping λ→λx0\lambda \to \lambda x_0 is continuous at the point λ=0\lambda = 0. (MTII′_{II}'). For each λ0∈A\lambda_0 \in A, the mapping x→λ0xx \to \lambda_0 x is continuous at the point x=0x = 0. (MTIII′_{III}'). The mapping (λ,x)→λx(\lambda, x) \to \lambda x is continuous at the point (0,0)(0, 0).

We deduce from this a necessary and sufficient set of conditions that the filter B\mathfrak{B} of neighbourhoods of 00 in an AA*-module* EE must satisfy in order to define a topology on EE compatible with its module structure; B\mathfrak{B} must satisfy the axioms (GAI_I) and (GAII_{II}) of §1§ 1, no. 2, and in addition must satisfy the following three axioms: (MVI_I). For each x0∈Ex_0 \in E and V∈BV \in \mathfrak{B}, there is a neighbourhood SS of 00 in AA such that S.x0⊂VS.x_0 \subset V. (MVII_{II}). For each λ0∈A\lambda_0 \in A and V∈BV \in \mathfrak{B}, there exists W∈BW \in \mathfrak{B} such that λ0W⊂V\lambda_0 W \subset V. (MVIII_{III}). For each V∈BV \in \mathfrak{B} there exists U∈BU \in \mathfrak{B} and a neighbourhood TT of 00 in AA such that T.U⊂VT.U \subset V.

Every commutative topological group is a topological Z\mathbf{Z}-module when the ring Z\mathbf{Z} is given the discrete topology.

If MM is a submodule of a topological AA-module EE, it is clear that the topology induced on MM by the topology of EE is compatible with the module structure of MM. Moreover, on the quotient AA-module E/ME/M, the topology which is the quotient by MM of the topology of EE is compatible with the AA-module structure. To see this it is enough to show that the mapping (λ,x)→λx(\lambda, x) \to \lambda x of A×(E/M)A \times (E/M) onto E/ME/M is continuous (where x→x′x \to x' denotes the canonical mapping of EE onto E/ME/M). Now, since we may identify the additive topological groups A×(E/M)A \times (E/M) and (A×E)/({0}×M)(A \times E)/(\{0\} \times M) (§ 2, no. 9, Corollary to Proposition 26), it is enough to show that the mapping (λ,x)→λx′(\lambda, x) \to \lambda x' of A×EA \times E into E/ME/M is continuous; and this is immediate, since the mapping in question is the composition of x→x′x \to x' and (λ,x)→λx(\lambda, x) \to \lambda x.

Let (Ei)i∈I(E_i)_{i \in I} be an arbitrary family of topological AA-modules, and let E=∏i∈IEiE = \prod_{i \in I} E_i be the AA-module which is the product of the EiE_i. Then the product topology on EE is compatible with the AA-module structure of EE. To prove this it is enough to show that the mapping (λ,x)→(λ⋅prix)i∈I(\lambda, x) \to (\lambda \cdot \mathrm{pr}_i x)_{i \in I} of A×EA \times E into EE is continuous, or (by Proposition 1 of Chapter I, § 4, no. 1) that for each index x∈Ix \in I the mapping (λ,x)→λ⋅prxx(\lambda, x) \to \lambda \cdot \mathrm{pr}_x x is a continuous mapping of A×EA \times E into ExE_x; but this mapping is the composition of (λ,xx)→λxx(\lambda, x_x) \to \lambda x_x and (λ,x)→(λ,prxx)(\lambda, x) \to (\lambda, \mathrm{pr}_x x), both of which are continuous.

Let AA be a Hausdorff topological ring and EE a Hausdorff topological AA-module. Let E^\hat{E} be the additive group which is the completion of the commutative topological group EE (§ 3, no. 5, Theorem 2). The Z\mathbf{Z}-bilinear mapping (λ,x)→λx(\lambda, x) \to \lambda x of the product A×EA \times E of the additive groups A,EA, E into the additive group EE can be extended by continuity to a Z\mathbf{Z}-bilinear mapping of A^×E^\hat{A} \times \hat{E} into E^\hat{E} (no. 5, Theorem 1), and this mapping we continue to denote by (λ,x)→λx(\lambda, x) \to \lambda x. By virtue of the principle of extension of identities, we have λ(μx)=(λμ)x\lambda (\mu x) = (\lambda \mu) x for λ∈A^,μ∈A^\lambda \in \hat{A}, \mu \in \hat{A} and x∈E^x \in \hat{E}, and 1⋅x=x1 \cdot x = x for all x∈E^x \in \hat{E}; the external law (λ,x)→λx(\lambda, x) \to \lambda x therefore defines an A^\hat{A}-module structure on E^\hat{E} compatible with its topology. The topological A^\hat{A}-module E^\hat{E} thus defined is called the completion of the topological AA-module EE.

Let EE be a topological module over a topological ring AA, where neither AA nor EE is necessarily Hausdorff. Let NN (resp. FF) be the closure of {0}\{0\} in AA (resp. EE). NN is a two-sided ideal of AA (no. 4, Proposition 5) and FF is a sub-AA-module of EE (no. 1, Proposition 1); furthermore, by continuity we have λx∈F\lambda x \in F whenever λ∈N\lambda \in N or x∈Fx \in F. We can therefore define, by passing to the quotients, a mapping (λ˙,x˙)→λ˙x˙(\dot{\lambda}, \dot{x}) \to \dot{\lambda} \dot{x} of (A/N)×(E/F)(A/N) \times (E/F) into E/FE/F; it is easily verified (by use of the Corollary to Proposition 26 of § 2, no. 9) that this mapping is continuous, and therefore defines a structure of a topological (A/N)(A/N)-module on E/FE/F. If we put B=A/NB = A/N and L=E/FL = E/F, then the BB-module LL is called the Hausdorff module associated with EE; its completion L^\hat{L} is a topological module over the Hausdorff completion A^\hat{A} (equal by definition to B^\hat{B}) of AA (no. 5), and this module L^\hat{L} is called the Hausdorff completion of EE and is denoted by E^\hat{E}. We see as in , no. 4, Proposition 8 that every continuous homomorphism u:E→Gu : E \to G of EE into a complete Hausdorff A^\hat{A}-module GG factorizes uniquely into u=v∘φu = v \circ \varphi, where vv is a continuous homomorphism of E^\hat{E} into GG and φ\varphi is the canonical mapping of EE into E^\hat{E}. We conclude that if E,E′E, E' are two topological AA-modules and u:E→E′u : E \to E' is a continuous homomorphism, then there is a unique continuous homomorphism u^:E^→E^′\hat{u} : \hat{E} \to \hat{E}' such that the diagram

E→uE′↓φ↓φ′E^→u^E^′ \begin{array}{ccc} E & \xrightarrow{u} & E' \\ \downarrow \varphi & & \downarrow \varphi' \\ \hat{E} & \xrightarrow{\hat{u}} & \hat{E}' \end{array}

is commutative, φ\varphi and φ′\varphi' being the canonical mappings.

The definitions and results of this section apply equally well to pseudo-modules over an arbitrary topological ring, by deleting all mention of the identity element of the ring.

7. Topological division rings and fields

In what follows, and in Chapters IV and V, if KK is a division ring we shall denote by K∗K^* the multiplicative group of non-zero elements of KK.

Definition 4

A topological division ring is a set* KK carrying a division ring structure and a topology compatible with the ring structure of KK, *and satisfying in addition the following axiom:

(KT) The mapping x→x−1x \to x^{-1} of K∗K^* into K∗K^* is continuous.

A commutative topological division ring is called a topological field.

A division ring structure and a topology on a set KK are said to be compatible if the corresponding ring structure and the topology are compatible and if in addition axiom (KT) is satisfied.

Example 1

On any division ring KK, the discrete topology is compatible with the division ring structure. A topological division ring whose topology is discrete is called a discrete division ring.

Example 2

\* The topology of the rational line QQ (resp. the real line RR) is compatible with the field structure of RR (resp. RR) (see Chapter IV, § 3). \

Definition 4 shows that, if KK is a topological division ring, then the topology induced by that of KK on the multiplicative group K∗K^* is compatible with the group structure of K∗K^*.

If a≠0a \neq 0, the homotheties x→axx \to ax and x→xax \to xa are homeomorphisms of KK onto itself; and so is the mapping x→ax+bx \to ax + b for all b∈Kb \in K. Note that the homotheties x→axx \to ax and x→xax \to xa are automorphisms of the (topological) additive group of KK if a≠0a \neq 0. If VV is any neighbourhood of oo in KK, it follows therefore that aVaV and VaVa are neighbourhoods of oo for all a≠0a \neq 0.

Let XX be a topological space, and let ff be a mapping of XX into a topological division ring KK. If ff is continuous at a point x0∈Xx_0 \in X and if f(x0)≠0f(x_0) \neq 0, then f−1f^{-1} is continuous at x0x_0. In particular, if KK is a topological field, then every rational function in nn variables with coefficients in KK is continuous at every point of KnK^n where its denominator does not vanish.

Likewise, if ff is a mapping of a set XX, filtered by a filter F\mathfrak{F}, into a Hausdorff topological division ring KK, and if lim⁡Ff\lim_{\mathfrak{F}} f exists and is not 00, then lim⁡Ff−1\lim_{\mathfrak{F}} f^{-1} exists and we have

lim⁡Ff−1=(lim⁡Ff)−1. \lim_{\mathfrak{F}} f^{-1} = (\lim_{\mathfrak{F}} f)^{-1}.

If HH is a division subring of a topological division ring KK, then the topology induced on HH by the topology of KK is compatible with the division ring structure of HH. The structure of a topological division ring thus defined on HH is said to be induced by that of KK. Furthermore, H‾\overline{H} is also a division subring of KK (the proof is analogous to that of Proposition 5).

In a topological division ring KK, the closure of the set {0}\{0\} is a two-sided ideal, by Proposition 5, and hence must be either {0}\{0\} or KK. In other words, if the topology of KK is not the coarsest topology (Chapter I, § 2, no. 2) then it is Hausdorff (§ 1, no. 2, Proposition 2).

8. Uniformities on a topological division ring

If KK is a topological division ring it is necessary to distinguish between: 1) the uniformity of the additive group of KK, which is defined on KK and is called the additive uniformity of KK; and 2) the left and right uniformities of the multiplicative group K∗K^*, which are defined on K∗K^* and are called (by abuse of language) the multiplicative uniformities of KK.

The uniformity induced on K∗K^* by the additive uniformity of KK is in general distinct from the multiplicative uniformities of KK (see Exercise 17).

By Proposition 6, a Hausdorff topological division ring KK can be considered as a dense subring of a complete Hausdorff ring K^\hat{K}. In order that K^\hat{K} should be a topological division ring it is necessary that the mapping x \to x^{-1} be extendable by continuity to (\hat{K})^*; and this necessary condition is also sufficient, for the functions xx^{-1}, x^{-1}x and 1 are then equal on (\hat{K})^* by reason of the principle of extension of identities, and therefore the value of the extended function is the inverse of x for each x≠0x \neq 0 in K^\hat{K}. In other words (cf. Chapter II, , no. 6, Proposition 11)

Proposition 7

The completion K^\hat{K} of a Hausdorff topological division ring K is a topological division ring if and only if the image under the mapping x→x−1x \to x^{-1} of every Cauchy filter (with respect to the additive structure) which does not have a cluster point at 0, is a Cauchy filter (with respect to the additive structure).

There are topological division rings in which this condition is not satisfied and in which the ring K^\hat{K} has zero divisors (see Exercise 26). Moreover, even if the completion K^\hat{K} is a topological division ring, there is no a priori reason to assume that the multiplicative structures of K^\hat{K} are structures of a complete space. However, this will be the case for division rings K such that K^\hat{K} is locally compact (see Chapter I, § 9, no. 7, Proposition 13, and Chapter III, § 3, no. 3, Proposition 4) and for topological fields; for the latter, we have the following proposition:

Proposition 8

If the additive uniform structure of a topological field K is a structure of a complete Hausdorff space, then the multiplicative structure on K* is a structure of a complete space.

We shall show that if F\mathfrak{F} is a Cauchy filter with respect to the multiplicative structure on K*, then F\mathfrak{F} is a Cauchy filter with respect to the additive structure on K, and does not converge to 0; this will establish the result. Let U be any neighbourhood of 0 in K, let V be a closed neighbourhood of 0 such that V⊂U,VV⊂UV \subset U, VV \subset U [axiom (AVII_\text{II})] and −1∉V-1 \notin V; then by hypothesis there is a set A∈FA \in \mathfrak{F} such that, for all x∈Ax \in A and y∈Ay \in A, we have x−1y∈1+Vx^{-1}y \in 1 + V. Let a∈Aa \in A, then A⊂a+aVA \subset a + aV, and a+aVa + aV is a closed set which does not contain 0; hence 0 is not in the closure of A and is therefore not a cluster point of F\mathfrak{F}. Let W be a neighbourhood of 0 such that aW⊂VaW \subset V [axiom (AVI_\text{I})]; then by hypothesis there is a set B∈FB \in \mathfrak{F} such that B⊂AB \subset A and, for all x∈Bx \in B and y∈By \in B, we have x−1y∈1+Wx^{-1}y \in 1 + W; hence y−x∈xW⊂AW⊂aW+aVWy - x \in xW \subset AW \subset aW + aVW; but K is commutative, and therefore aVW=aWV⊂VW⊂UaVW = aWV \subset VW \subset U; consequently y−x∈U+Uy - x \in U + U and the Proposition is proved.

The same proof shows that Proposition 8 can be extended to the case in which every Cauchy filter with respect to one of the multiplicative structures of K is also a Cauchy filter with respect to the other multiplicative structure.

Exercises

See the exercises for § 6.

Transcribed from the printed volume. Not translated.