Bourbaki

§ 4. Groups and groups with operators

1. Groups

Recall the following definition (§ 2, no. 3, Definition 6).

Definition 1

A set with an associative law of composition, possessing an identity element and under which every element is invertible, is called a group.

In other words, a group is a monoid (§ 2, no. 1, Definition 1) in which every element is invertible. A law of composition on a set which determines a group structure on it is called a group law. If G and H are two groups, a magma homomorphism of G into H is also called a group homomorphism. Such a homomorphism ff maps identity element to identity element; for, let ee (resp. e′e') be the identity element of G (resp. H); writing the group laws of G and H multiplicatively, e⋅e=ee \cdot e = e, whence f(e)⋅f(e)=f(e)f(e) \cdot f(e) = f(e) and, multiplying by f(e)−1,f(e)=e′f(e)^{-1}, f(e) = e'. Hence ff is unital. It then follows from no. 3 of § 2 that f(x−1)=f(x)−1f(x^{-1}) = f(x)^{-1} for all x∈Gx \in G.

Example

In any monoid E the set of invertible elements with the structure induced by that on E is a group. In particular, the set of bijective mappings of a set F onto itself (or set of permutations of F) is a group under the law (f,g)↦f∘g(f, g) \mapsto f \circ g, called the symmetric group of the set F and denoted by SF\mathfrak{S}_F.

In this paragraph, unless otherwise indicated, the law of composition of a group will always be written multiplicatively and ee will denote the identity element of such a group law.

A group G is called finite if the underlying set of G is finite; otherwise it is called infinite; the cardinal of a group is called the order of the group.

If a law of composition on G determines a group structure on G, so does the opposite law. The mapping of a group G onto itself which associates with each x∈Gx \in G the inverse of xx is an isomorphism of G onto the opposite group (§ 2, no. 3, Proposition 4).

Following our general conventions (Set Theory, II, § 3, no. 1), we shall denote by A−1A^{-1} the image of a subset A of G under the mapping x↦x−1x \mapsto x^{-1}. But it is important to note that, in spite of the analogy of notation, A−1A^{-1} is definitely not the inverse element of A under the law of composition (X,Y)↦XY(X, Y) \mapsto XY between subsets of G (recall that XY is the set of xyxy with x∈X,y∈Yx \in X, y \in Y): the identity element under this law is {e}\{e\} and the only invertible elements of P(G)\mathscr{P}(G) under this law are the sets A consisting of a single element (such an A, moreover, certainly has inverse A−1A^{-1}). The identity

(AB)−1=B−1A−1(AB)^{-1}=B^{-1}A^{-1} holds for A⊂GA\subset G, B⊂GB\subset G. AA is called a symmetric subset of GG if A=A−1A=A^{-1}. For all A⊂GA\subset G, A∪A−1A\cup A^{-1}, A∩A−1A\cap A^{-1} and AA−1AA^{-1} are symmetric.

2. Groups with operators

Definition 2

Let Ω\Omega be a set. A group GG together with an action of Ω\Omega on GG which is distributive with respect to the group law, is called a group with operators in Ω\Omega.

In what follows xαx^\alpha will denote the composition of α∈Ω\alpha\in\Omega and x∈Gx\in G. Distributivity is then expressed by the identity (xy)α=xαyα(xy)^\alpha=x^\alpha y^\alpha.

In a group with operators GG, each operator defines an endomorphism of the underlying group structure; these endomorphisms will sometimes be called the homotheties of the group with operators GG.

A group with operators GG is called commutative (or Abelian) if its group law is commutative.

In what follows a group GG will be identified with the group with operators in ∅\varnothing obtained by giving GG the unique action of ∅\varnothing on GG. This allows us to consider groups as special cases of groups with operators and to apply to them the definitions and results relating to the latter which we shall state.

Example

In a commutative group GG, written multiplicatively, (xy)n=xnyn(xy)^n=x^ny^n for all n∈Zn\in\mathbf{Z} (§ 2, no. 8, equation (1)); the action n↦(x↦xn)n\mapsto(x\mapsto x^n) of Z\mathbf{Z} on GG therefore defines, together with the group law, the structure of a group with operators on GG.

Definition 3

Let GG and G′G' be groups with operators in Ω\Omega. A homomorphism of groups with operators of GG into G′G' is a homomorphism of the group GG into the group G′G' such that

f(xα)=(f(x))α f(x^\alpha)=(f(x))^\alpha

for all α∈Ω\alpha\in\Omega and all x∈Gx\in G.

An endomorphism of the group with operators GG is an endomorphism of the group GG which is permutable with all the homotheties of GG.

As two homotheties of a group with operators GG are not necessarily permutable, a homothety of GG is not in general an endomorphism of the group with operators GG.

The identity mapping of a group with operators is a homomorphism of groups with operators; the composition of two homomorphisms of groups with operators is also one. For a mapping to be an isomorphism of groups with operators, it is necessary and sufficient that it be a bijective homomorphism of groups with operators and the inverse mapping is then an isomorphism of groups with operators.

More generally, let G (resp. G') be a group with operators in Ω\Omega (resp. Ω′\Omega'). Let ϕ\phi be a mapping of Ω\Omega into Ω′\Omega'. A ϕ\phi-homomorphism of G into G' is a homomorphism of the group G into the group G' such that

f(xα)=(f(x))(ϕ(α)) f(x^\alpha) = (f(x))^{(\phi(\alpha))}

for all α∈Ω\alpha \in \Omega and all x∈Gx \in G.

In the rest of this paragraph we shall be given a set Ω\Omega. Unless otherwise mentioned the groups with operators considered will admit Ω\Omega as set of operators.

3. Subgroups

Definition 4

Let G be a group with operators. A stable subgroup of G is a subset H of G with the following properties*:

(i) e∈He \in H; (ii) x,y∈H  ⟹  xy∈Hx, y \in H \implies xy \in H; (iii) x∈H  ⟹  x−1∈Hx \in H \implies x^{-1} \in H; (iv) x∈Hx \in H and α∈Ω\alpha \in \Omega imply xα∈Hx^\alpha \in H.

If H is a stable subgroup of G, the structure induced on H by the structure of a group with operators on G is the structure of a group with operators and the canonical injection of H into G is a homomorphism of groups with operators.

Let G be a group. A stable subgroup of G with the action of ∅\varnothing (no. 2), which is a subset of G satisfying conditions (i), (ii), (iii) of Definition 4, is called a subgroup of G. When we speak of a subgroup of a group of operators we shall always mean a subgroup of the underlying group of G. A subgroup of a group with operators G is not necessarily a stable subgroup of G.

Example (1). Let Σ\Sigma be a species of structure (Set Theory, IV, § 1, no. 4) and S a structure of species Σ\Sigma on a set E (loc. cit.). The set of automorphisms of S is a subgroup of GE\mathscr{G}_E.

Proposition 1

Let G be a group with operators and H a subset of G which is stable under the homotheties of G. The following conditions are equivalent*:

(a) H is a stable subgroup of G. (b) H is non-empty and the relations x∈H,y∈H  ⟹  xy∈Hx \in H, y \in H \implies xy \in H and x−1∈Hx^{-1} \in H. (c) H is non-empty and the relations x∈H,y∈H  ⟹  xy−1∈Hx \in H, y \in H \implies xy^{-1} \in H. (d) H is stable under the law on G and the law of composition induced on H by the law of composition on G is a group law.

Clearly (a) implies (b). We show that (b) implies (a). It suffices to show that H contains the identity element of G. As the subset H is non-empty, let x∈Hx \in H. Then x−1∈Hx^{-1} \in H and e=xx−1∈He = xx^{-1} \in H. Clearly (b) implies (c). We show that (c) implies (b). First of all since H is non-empty it contains an element x. Hence xx^{-1} = e is an element of H. For every element x of H, x−1=ex−1x^{-1} = ex^{-1} belongs to H; hence the relations x∈H,y∈Hx \in H, y \in H imply x(y−1)−1=xy∈Hx(y^{-1})^{-1} = xy \in H. Clearly (a) implies (d). We show that (d) implies (a): the canonical injection of H into G is a group homomorphism; hence e∈He \in H and the relation x∈Hx \in H implies x−1∈Hx^{-1} \in H (no. 1).

Remark

(1) Similarly it can be shown that condition (b) is equivalent to the condition (c') H≠∅H \neq \varnothing and the relations x∈Hx \in H and y∈Hy \in H imply y−1x∈Hy^{-1}x \in H. (2) For every subgroup H of G there are the following relations

(1)(1)
H.H=HandH−1=H.H.H = H \quad \text{and} \quad H^{-1} = H.
For H.H⊂HH.H \subset H and H−1⊂HH^{-1} \subset H by (b). As e∈H,H.H⊃e.H=He \in H, H.H \supset e.H = H and taking inverses transforms the inclusion H−1⊂HH^{-1} \subset H into H⊂H−1H \subset H^{-1}, whence formulae (1).

If H is a stable subgroup of G and K is a stable subgroup of H, clearly K is a stable subgroup of G.

The set {e}\{e\} is the smallest stable subgroup of G. The intersection of a family of stable subgroups of G is a stable subgroup. There is therefore a smallest stable subgroup H of G containing a given subset X of G; it is called the stable subgroup generated by X and X is called a generating system (or generating set) of H.

Proposition 2

Let X be a non-empty subset of a group with operators G and X^\hat{X} the stable subset under the action of Ω\Omega on G generated by X. The stable subgroup generated by X is the stable subset under the law on G generated by the set Y=X^∪X^−1Y = \hat{X} \cup \hat{X}^{-1}.

The latter subset Z is the set of compositions of finite sequences all of whose terms are elements of X^\hat{X} or inverse of elements of X^\hat{X}: the inverse of such a composition is a composition of the same form (§ 2, no. 3, Corollary 1 to Proposition 5) and Z is stable under the action of Ω\Omega, as is seen by applying § 3, no. 4, Proposition 1 to the homotheties of G, hence (Proposition 1) Z is a stable subgroup of G. Conversely, every stable subgroup containing X obviously contains Y and hence Z.

Corollary 1

Let G be a group with operators and X a subset of G which is stable under the action of Ω\Omega. The subgroup generated by X and the stable subgroup generated by X coincide.

Corollary 2

Let G be a group and X a subset of G consisting of pairwise permutable elements. The subgroup of G generated by X is commutative.

The set Y=X∪X−1Y = X \cup X^{-1} consists of pairwise permutable elements (§ 2, no. 3, Proposition 5) and the law induced on the stable subset generated by Y is commutative (§ 1, no. 5, Corollary 2).

If G is a group with operators, the stable subgroup generated by a subset of G consisting of pairwise permutable elements is not necessarily commutative.

Corollary 3

Let f:G→G′f : G \to G' be a homomorphism of groups with operators and X a subset of G. The image under f of the stable subgroup of G generated by X is the stable subgroup of G' generated by f(X)f(X).

Let X′=f(X)X' = f(X). Then X^′=f(X^)\hat{X}' = f(\hat{X}) and X′−1=f(X−1){X'}^{-1} = f(X^{-1}). Hence

f(X^∪X^−1)=X^′∪X′^−1. f(\hat{X} \cup \hat{X}^{-1}) = \hat{X}' \cup \hat{{X}'}^{-1}.
The corollary then follows from § 1, no. 4, Proposition 1.

Example (2). Let G be a group and x an element of G. The subgroup generated by {x}\{x\} (called more simply the subgroup generated by x) is the set of xn,n∈Zx^n, n \in \mathbf{Z}. The stable subset (under the law on G) generated by {x}\{x\} is the set of xnx^n where n∈N∗n \in \mathbf{N}^*. These two sets are in general distinct.

Thus, in the additive group Z\mathbf{Z}, the subgroup generated by an element x is the set x.Zx.\mathbf{Z} of xn,n∈Zxn, n \in \mathbf{Z}, and the stable subset generated by x is the set of xn,n∈N∗xn, n \in \mathbf{N}^*. These two sets are always distinct if x≠0x \neq 0.

The union of a right directed family of stable subgroups of G is obviously a stable subgroup. It follows that, if P is a subset of G and H a stable subgroup of G not meeting P, the set of stable subgroups of G containing H and not meeting P, ordered by inclusion, is inductive (Set Theory, III, § 2, no. 4). Applying Zorn's Lemma (Set Theory, III, § 2, no. 4), we obtain the following result:

Proposition 3

Let G be a group with operators, P a subset of G and H a stable subgroup G not meeting P. The set of stable subgroups of G containing H and not meeting P has a maximal element.

4. Quotient groups

Theorem 1

Let R be an equivalence relation on a group with operators G; if R is left (resp. right) compatible (§ 3, no. 3) with the group law on G and compatible with the action of Ω\Omega, the equivalence class of e is a stable subgroup H of G and the relation R is equivalent to x−1y∈Hx^{-1}y \in H (resp. yx−1∈Hyx^{-1} \in H). Conversely, if H is a stable subgroup of G, the relation x−1y∈Hx^{-1}y \in H (resp. yx−1∈Hyx^{-1} \in H) is an equivalence relation which is left (resp. right) compatible with the group law on G and compatible with the action of Ω\Omega and under which H is the equivalence class of e.

We restrict our attention to the case where the relation R is left compatible with the law on G (the case of a right compatible relation follows by replacing the law on G by the opposite law). The relation y≡x(modR)y \equiv x \pmod{R} is equivalent to x−1y≡e(modR)x^{-1}y \equiv e \pmod{R}, for y≡xy \equiv x implies x−1y≡x−1x=ex^{-1}y \equiv x^{-1}x = e and conversely x−1y≡ex^{-1}y \equiv e implies y=x(x−1y)≡xy = x(x^{-1}y) \equiv x. If H denotes the equivalence class of e, the relation R is then equivalent to x−1y∈Hx^{-1}y \in H. We show that H is a stable subgroup of G. For every operator α\alpha, the relation x≡ex \equiv e implies xα≡eα=ex^\alpha \equiv e^\alpha = e, hence Hα⊂HH^\alpha \subset H and H is stable under the action of Ω\Omega. It suffices to establish (Proposition 1) that x∈Hx \in H and y∈Hy \in H imply x−1y∈Hx^{-1}y \in H, that is x≡ex \equiv e and y≡ey \equiv e imply x≡yx \equiv y, which is a consequence of the transitivity of R.

Conversely, let H be a stable subgroup of G; the relation x−1y∈Hx^{-1}y \in H is reflexive since x−1x=e∈Hx^{-1}x = e \in H; it is symmetric since x−1y∈Hx^{-1}y \in H implies y−1x=(x−1y)−1∈Hy^{-1}x = (x^{-1}y)^{-1} \in H; it is transitive, for x−1y∈Hx^{-1}y \in H and y−1z∈Hy^{-1}z \in H imply x−1z=(x−1y)(y−1z)∈Hx^{-1}z = (x^{-1}y)(y^{-1}z) \in H; it is left compatible with the law of composition on G, for x−1y=(zx)−1(zy)x^{-1}y = (zx)^{-1}(zy) for all z∈Gz \in G; finally, for every operator α\alpha, the relation y∈xHy \in xH implies yα∈xαHα⊂xαHy^\alpha \in x^\alpha H^\alpha \subset x^\alpha H and hence the equivalence relation x−1y∈Hx^{-1}y \in H is compatible with the action of Ω\Omega on G.

Let G be a group and H a subgroup of G; the relation x−1y∈Hx^{-1}y \in H (resp. yx−1∈Hyx^{-1} \in H) is also written in the equivalent form y∈xHy \in xH (resp. y∈Hxy \in Hx). Every subgroup H of G thus defines two equivalent relations on G, namely y∈xHy \in xH and y∈Hxy \in Hx: the equivalent classes under these relations are respectively the sets xHxH, which are called left cosets of H (or modulo H), and the sets HxHx, which are called right cosets of H (or modulo H). By saturating a subset A⊂GA \subset G with respect to these relations (Set Theory, II, § 6, no. 4), we obtain respectively the sets AH and HA. The mapping x↦x−1x \mapsto x^{-1} transforms left cosets modulo H into right cosets modulo H and conversely.

The cardinal of the set of left cosets (mod. H) is called the index of the subgroup H with respect to G and is denoted by (G:H)(G:H); it is also equal to the cardinal of the set of right cosets.

If a subgroup K of G contains H, it is a union of left (or right) cosets of H. Since a left coset of K is obtained from K by left translation, the set of left cosets of H contained in a left coset of K has cardinal independent of the latter. Hence (Set Theory, III, § 5, no. 8, Proposition 9):

Proposition 4

Let H and K be two subgroups of a group G such that H ⊂\subset K. Then*

(G:H)=(G:K)(K:H). (G:H) = (G:K)(K:H).

Corollary

If G is a finite group of order g and H is a subgroup of G of order h, then*

h⋅(G:H)=g h \cdot (G:H) = g
(in particular, the order and index of H are divisors of the order of G).

Theorem 1 allows us to determine the equivalence relations compatible with the laws on a group with operators G: if R is such a relation, it is both left and right compatible with the group law on G and with the action of Ω\Omega. Hence, if H is the class of ee (mod. R), H is a stable subgroup such that the relations y∈xHy \in xH and y∈Hxy \in Hx are equivalent (since both are equivalent to R); hence xH=HxxH = Hx for all x∈Gx \in G. Conversely, if this is so, one or other of the equivalent relations y∈xH,y∈Hxy \in xH, y \in Hx is compatible with the group law, since it is both left and right compatible with this law (§\S 3, no. 4) and is compatible with the action of Ω\Omega. Since the equation xH=HxxH = Hx is equivalent to xHx−1=HxHx^{-1} = H, we make the following definition:

Definition 5

Let G be a group with operators. A stable subgroup H of G is called a normal (or invariant) stable subgroup of G if xHx−1=HxHx^{-1} = H for all x∈Gx \in G.

If Ω=∅\Omega = \varnothing, a normal stable subgroup of G is called a normal (or invariant) subgroup of G. In a commutative group every subgroup is normal.

To verify that a stable subgroup H is normal, it suffices to show that xHx−1⊂HxHx^{-1} \subset H for all x∈Gx \in G; for if so then x−1Hx⊂Hx^{-1}Hx \subset H for all x∈Gx \in G, that is H⊂xHx−1H \subset xHx^{-1}, and hence H=xHx−1H = xHx^{-1}.

Let H be a normal stable subgroup of G and R the equivalence relation y∈xHy \in xH defined by H; on the quotient set G/RG/R, the internal law, the quotient by R of the law of the group G, is associative; the class of e is the identity element of this quotient law; the classes of two inverse elements in G are inverses under the quotient law and the action of Ω\Omega, the quotient by R of the action of Ω\Omega on G, is distributive with respect to the internal law on G/RG/R (§\S 3, no. 5). Hence, summarizing the results obtained:

Theorem 2

Let G be a group with operators. For an equivalence relation R on G to be compatible with the group law and the action of Ω\Omega, it is necessary and sufficient that it be of the form x−1y∈Hx^{-1}y \in H, where H is a normal stable subgroup of G (the relation x−1y∈Hx^{-1}y \in H being moreover equivalent to yx−1∈Hyx^{-1} \in H for such a subgroup). The law of composition on G/RG/R the quotient of that on G and the action of Ω\Omega on G/RG/R the quotient of that of Ω\Omega on G by such a relation R give G/RG/R the structure of a group with operators, called the quotient structure, and the canonical mapping of the passage to the quotient is a homomorphism of groups with operators.

Definition 6

The quotient of a group with operators G by the equivalence relation defined by a normal subgroup H of G, with the quotient structure, is called the quotient group with operators of G by H and is denoted by G/HG/H. The canonical mapping G→G/HG \to G/H is called a canonical homomorphism

Let G be a group and H a normal subgroup of G. The quotient G/HG/H, with its group structure, is called the quotient group of G by H. For a mapping from G/HG/H to a group with operators to be a homomorphism of groups with operators, it is necessary and sufficient that its composition with the canonical mapping of G onto G/HG/H be one: this justifies the name "quotient group" (Set Theory, IV, §\S 2, no. 6).

The equivalence relation defined by a normal stable subgroup of G is denoted by x≡y(modH)x \equiv y \pmod{H} or x≡y(H)x \equiv y(H).

Proposition 5

Let f:G→G′f : G \to G' be a homomorphism of groups with operators and HH and H′H' normal stable subgroups of GG and G′G' respectively such that f(H)⊂H′f(H) \subset H'. The mapping ff is compatible with the equivalence relations defined by HH and H′H'. Let π:G→G/H\pi : G \to G/H and π′:G′→G′/H′\pi' : G' \to G'/H' be the canonical homomorphisms. The mapping fˉ:G/H→G′/H′\bar{f} : G/H \to G'/H' derived from ff by passing to the quotients is a homomorphism.

If x≡y(modH)x \equiv y \pmod{H}, then x−1y∈Hx^{-1}y \in H, whence

f(x)−1f(y)=f(x−1)f(y)=f(x−1y)∈f(H)⊂H′ f(x)^{-1}f(y) = f(x^{-1})f(y) = f(x^{-1}y) \in f(H) \subset H'
and hence f(x)≡f(y)(modH′)f(x) \equiv f(y) \pmod{H'}. The second assertion follows from the universal property of quotient laws (§1§ 1, no. 6).

Remark

(1) If AA is any subset of a group GG and HH is a normal subgroup of GG, then AH=HAAH = HA; this set is obtained by saturating AA with respect to the relation x≡y(modH)x \equiv y \pmod{H}.

(2) If HH is a normal subgroup of GG of finite index, the quotient group G/HG/H is a finite group of order (G:H)(G:H).

Note that if HH is a normal subgroup of a group GG and KK is a normal subgroup of HH, KK is not necessarily a normal subgroup of GG (I, §5§ 5, Exercise 10).

Let GG be a group with operators. The intersection of every family of normal stable subgroups of GG is a normal stable subgroup. Hence, for every subset XX of GG, there exists a smallest normal stable subgroup containing XX, called the normal stable subgroup generated by XX.

In a group with operators GG, the stable subgroups GG and {e}\{e\} are normal.

Definition 7

A group with operators* GG is called simple if G≠{e}G \neq \{e\} and there exists no normal stable subgroup of GG other than GG and {e}\{e\}.

5. Decomposition of a homomorphism

Proposition 6

Let* GG be a group with operators and G′G' a magma with an action by Ω\Omega, written exponentially. Let f:G→G′f : G \to G' be a homomorphism of the magna GG into the magma G′G' such that, for all α∈Ω\alpha \in \Omega and all x∈Gx \in G, f(xα)=f(x)αf(x^\alpha) = f(x)^\alpha. Then f(G)f(G) is a stable subset of G′G' under the law on G′G' and the action of Ω\Omega; the set f(G)f(G) with the induced laws is a group with operators and the mapping x↦f(x)x \mapsto f(x) of GG into f(G)f(G) is a homomorphism of groups with operators.

By virtue of §1§ 1, no. 4, Proposition 1, f(G)f(G) is a stable subset of G′G' under the internal law on G′G'. For every element x∈Gx \in G and for every operator α\alpha, f(x)α=f(xα)∈f(G)f(x)^\alpha = f(x^\alpha) \in f(G) and therefore f(G)f(G) is stable under the action of Ω\Omega on G′G'. Writing the internal law of G′G' multiplicatively,

(f(x)f(y))f(z)=f(xy)f(z)=f((xy)z)=f(x(yz))=f(x)f(yz)=f(x)(f(y)f(z)) (f(x)f(y))f(z) = f(xy)f(z) = f((xy)z) = f(x(yz)) = f(x)f(yz) = f(x)(f(y)f(z))
for all elements x,y,zx, y, z in GG; therefore the induced law on f(G)f(G) is associative.

Let ee be the identity element of GG. Its image f(e)f(e) is an identity element of f(G)f(G) (§2§ 2, no. 1). Every element of f(G)f(G) is invertible in f(G)f(G) (§2§ 2, no. 3). Therefore the law induced on f(G)f(G) by the internal law on G′G' is a group law. For all elements xx and yy in GG and every operator α\alpha,

(f(x)f(y))α=(f(xy))α=f((xy)α)=f(xαyα)=f(xα)f(yα)=(f(x))α(f(y))α (f(x)f(y))^\alpha = (f(xy))^\alpha = f((xy)^\alpha) = f(x^\alpha y^\alpha) = f(x^\alpha)f(y^\alpha) = (f(x))^\alpha(f(y))^\alpha

which shows that the action of Ω\Omega is distributive with respect to the group law on f(G)f(G). Therefore f(G)f(G) with the induced laws is a group with operators and clearly the mapping x↦f(x)x \mapsto f(x) is a homomorphism of groups with operators.

Definition 8

Let f:G→G′f : G \to G' be a homomorphism of groups with operators. The inverse image of the identity element of G′G' is called the kernel of ff.

The kernel of ff is often denoted by Ker(f)\mathrm{Ker}(f) and the image f(G)f(G) of ff is sometimes denoted by Im(f)\mathrm{Im}(f).

Theorem 3

Let f:G→G′f : G \to G' be a homomorphism of groups with operators.* (a) Ker(f)\mathrm{Ker}(f) is a normal stable subgroup of GG; (b) Im(f)\mathrm{Im}(f) is a stable subgroup of G′G'; (c) the mapping ff is compatible with the equivalence relation defined on GG by Ker(f)\mathrm{Ker}(f); (d) the mapping f~:G/Ker(f)→Im(f)\tilde{f} : G/\mathrm{Ker}(f) \to \mathrm{Im}(f) derived from ff by passing to the quotient is an isomorphism of groups with operators; (e) *f=i∘f~∘πf = i \circ \tilde{f} \circ \pi, where ii is the canonical injection of Im(f)\mathrm{Im}(f) into G′G' and π\pi is the canonical homomorphism of GG onto G/Ker(f)G/\mathrm{Ker}(f).

Assertion (b) follows from Proposition 6. The equivalence relation f(x)=f(y)f(x) = f(y) on GG is compatible with the group with operators structure on GG. By Theorem 2 (no. 4), it is therefore of the form y∈xHy \in xH, where HH is a normal stable subgroup of GG and HH is the class of the identity element, whence H=Ker(f)H = \mathrm{Ker}(f). Assertions (a), (c) and (d) then follow. Assertion (e) is obvious (Set Theory, II, § 6, no. 5).

6. Subgroups of a quotient group

Proposition 7

Let GG and HH be two groups with operators, ff a homomorphism of GG into HH and NN the kernel of ff.* (a) Let H′H' be a stable subgroup of HH. The inverse image G′=f−1(H′)G' = f^{-1}(H') is a stable subgroup of GG and G′G' is normal in GG if H′H' is normal in HH. Further, NN is a normal subgroup of G′G'. If ff is surjective, then H′=f(G′)H' = f(G') and ff defines an isomorphism of G′/NG'/N onto H′H' on passing to the quotient. (b) *Let G′G' be a stable subgroup of GG. The image H′=f(G′)H' = f(G') is a stable subgroup of HH and f−1(H′)=G′N=NG′f^{-1}(H') = G'N = NG'. In particular, f−1(H′)=G′f^{-1}(H') = G' if and only if N⊂G′N \subset G'. If ff is surjective and G′G' is normal in GG, then H′H' is normal in HH.

(a) Let xx and yy be in G′G' and α∈Ω\alpha \in \Omega; then f(x)∈H′f(x) \in H' and f(y)∈H′f(y) \in H', whence f(xy−1)=f(x)f(y)−1∈H′f(xy^{-1}) = f(x)f(y)^{-1} \in H', that is xy−1∈G′xy^{-1} \in G'; hence G′G' is a subgroup of GG. Now f(xα)=f(x)α∈H′f(x^\alpha) = f(x)^\alpha \in H', whence xα∈G′x^\alpha \in G' and therefore G′G' is stable. Suppose H′H' is normal in HH and let x∈G′,y∈Gx \in G', y \in G; then f(x)∈H′f(x) \in H' and

f(yxy−1)=f(y)f(x)f(y)−1∈H′ f(yxy^{-1}) = f(y)f(x)f(y)^{-1} \in H'
whence yxy−1∈G′yxy^{-1} \in G'; hence G′G' is normal in GG. For all n∈N,f(n)=e∈H′n \in N, f(n) = e \in H', whence N⊂G′N \subset G'; as NN is normal in GG, it is normal in G′G'. Finally, if ff is surjective,
f(f−1(A))=A for every subset A of H, f(f^{-1}(A)) = A \text{ for every subset } A \text{ of } H,
whence H′=f(G′)H' = f(G'); the restriction of ff to G′G' is a homomorphism f′f' of G′G' onto H′H' of kernel NN, hence f′f' defines on passing to the quotient an isomorphism of G′/NG'/N onto H′H'.

(b) Let aa and bb be in H′H' and α\alpha in Ω\Omega; there exist x,yx, y in G′G' with a=f(x)a = f(x) and b=f(y)b = f(y), whence ab−1=f(xy−1)∈H′ab^{-1} = f(xy^{-1}) \in H', hence H′H' is a subgroup of HH which is stable, for aα=f(xα)∈H′a^\alpha = f(x^\alpha) \in H'. Let x∈Gx \in G; then x∈f−1(H′)x \in f^{-1}(H') if and only if f(x)∈H′=f(G′)f(x) \in H' = f(G'), that is if and only if there exists yy in G′G' with f(x)=f(y)f(x) = f(y); the relation f(x)=f(y)f(x) = f(y) is equivalent to the existence of n∈Nn \in N with x=ynx = yn; finally, x∈f−1(H′)x \in f^{-1}(H') is equivalent to x∈G′N=NG′x \in G'N = NG'. Clearly the relation G′=G′NG' = G'N is equivalent to G′⊃NG' \supset N. Suppose finally that ff is surjective and G′G' is normal in GG; let a∈H′a \in H' and b∈Hb \in H; there exist x∈G′x \in G' and y∈Gy \in G with a=f(x)a = f(x) and b=f(y)b = f(y), whence bab−1=f(yxy−1)∈f(G′)=H′bab^{-1} = f(yxy^{-1}) \in f(G') = H'. Hence H′H' is normal in HH.

Corollary 1

Suppose that ff is surjective. Let G\mathscr{G} (resp. G′\mathscr{G}') be the set of stable (resp. normal stable) subgroups of GG containing NN and H\mathfrak{H} (resp. H′\mathfrak{H}') the set of stable (resp. normal stable) subgroups of HH, these sets being ordered by inclusion. The mapping G′↦f(G′)G' \mapsto f(G') is an ordered set isomorphism Φ:G→H\Phi : \mathscr{G} \to \mathfrak{H}; the inverse isomorphism Ψ:H→G\Psi : \mathfrak{H} \to \mathscr{G} is the mapping H′↦f−1(H′)H' \mapsto f^{-1}(H'). Further Φ\Phi and Ψ\Psi induce isomorphisms Φ′:G′→H′\Phi' : \mathscr{G}' \to \mathfrak{H}' and Ψ′:H′→G′\Psi' : \mathfrak{H}' \to \mathscr{G}'.

Corollary 2

Let f:G→Hf : G \to H be a homomorphism of groups with operators, NN the kernel of ff, G′G' a stable subgroup of GG and LL a normal stable subgroup of G′G'. Then LN,L.(G′∩N)LN, L.(G' \cap N) and f(L)f(L) are normal stable subgroups of G′N,G′G'N, G' and f(G′)f(G') respectively and the three quotient groups with operators G′N/LN,G′/L.(G′∩N)G'N/LN, G'/L.(G' \cap N) and f(G′)/f(L)f(G')/f(L) are isomorphic.

Let H′=f(G′)H' = f(G') and f′f' denote the homomorphism of G′G' onto H′H' which coincides with ff on G′G'; the kernel of f′f' is G∩NG \cap N and f′(L)=f(L)f'(L) = f(L); by Proposition 7, f′(L)f'(L) is a normal stable subgroup of H′H' and

f′−1(f′(L))=L.(G′∩N) {f'}^{-1}(f'(L)) = L.(G' \cap N)
is a normal stable subgroup of G′G'. Let λ\lambda be the canonical homomorphism of H′H' onto H′/f′(L)=f(G′)/f(L)H'/f'(L) = f(G')/f(L): as λ∘f′\lambda \circ f' is surjective with kernel
f′−1(f′(L))=L.(G′∩N), {f'}^{-1}(f'(L)) = L.(G' \cap N),

it defines an isomorphism of G′/L⋅(G′∩N)G'/L\cdot(G'\cap N) onto f(G′)/f(L)f(G')/f(L). By Proposition 7, (b), f−1(H′)=G′Nf^{-1}(H')=G'N; if f′′f'' is the homomorphism of G′NG'N onto H′H' which coincides with ff on G′NG'N, the homomorphism λ∘f′′\lambda\circ f'' of G′NG'N onto f(G′)/f(L)f(G')/f(L) is surjective with kernel f−1(f(L))=LNf^{-1}(f(L))=LN; this proves that LNLN is a normal stable subgroup of G′NG'N and that λ∘f′′\lambda\circ f'' defines an isomorphism of G′N/LNG'N/LN onto f(G′)/f(L)f(G')/f(L).

Corollary 3

Let f:G→Hf:G\to H be a homomorphism of groups with operators, NN its kernel, XX a subset of GG such that f(X)f(X) generates HH and YY a subset of NN which generates NN. Then X∪YX\cup Y generates NN.

Let G′G' be the stable subgroup of GG generated by X∪YX\cup Y. As Y⊂G′Y\subset G', N⊂G′N\subset G'.

As f(X)⊂f(G′)f(X)\subset f(G'), f(G′)=Hf(G')=H, whence G′=f−1(H)=GG'=f^{-1}(H)=G.

Remark

In the notation of Proposition 7, the fact that the inverse image of a subgroup of HH is a subgroup of GG follows from the following more general fact.

If AA and BB are subsets of HH and ff is surjective, then

f−1(A⋅B)=f−1(A)⋅f−1(B),f−1(A−1)=f−1(A)−1. f^{-1}(A\cdot B)=f^{-1}(A)\cdot f^{-1}(B),\qquad f^{-1}(A^{-1})=f^{-1}(A)^{-1}.

Obviously f−1(A)⋅f−1(B)⊂f−1(A⋅B)f^{-1}(A)\cdot f^{-1}(B)\subset f^{-1}(A\cdot B); on the other hand, if z∈f−1(A⋅B)z\in f^{-1}(A\cdot B), there exists a∈Aa\in A and b∈Bb\in B such that f(z)=abf(z)=ab; as ff is surjective, there exists x∈Gx\in G such that f(x)=af(x)=a; writing y=x−1zy=x^{-1}z, f(y)=a−1f(z)=bf(y)=a^{-1}f(z)=b and z=xyz=xy, whence z∈f−1(A)⋅f−1(B)z\in f^{-1}(A)\cdot f^{-1}(B). The relation x∈f−1(A−1)x\in f^{-1}(A^{-1}) is equivalent to f(x)∈A−1f(x)\in A^{-1}, hence to f(x−1)∈Af(x^{-1})\in A, that is to x−1∈f−1(A)x^{-1}\in f^{-1}(A) and finally to x∈f−1(A)−1x\in f^{-1}(A)^{-1}.

Proposition 8

Let GG be a group with operators and AA and BB two stable subgroups of GG. Suppose that the relations a∈Aa\in A and b∈Bb\in B imply aba−1∈B∗aba^{-1}\in B^{*} (in other words, AA normalizes BB)∗^{*}. Then AB=BAAB=BA is a stable subgroup of GG, A∩BA\cap B is a normal stable subgroup of AA and BB is a normal stable subgroup of ABAB. The canonical injection of AA into ABAB defines on passing to the quotient an isomorphism of A/(A∩B)A/(A\cap B) onto AB/BAB/B.

The formulae

(ab)(a′b′)=aa′(a′−1ba′⋅b′) (ab)(a'b')=aa'({a'}^{-1}ba'\cdot b')
(ab)−1=a−1(ab−1a−1) (ab)^{-1}=a^{-1}(ab^{-1}a^{-1})
(ab)α=aαbα (ab)^\alpha=a^\alpha b^\alpha

for a,a′∈Aa,a'\in A, b,b′∈Bb,b'\in B and every operator α\alpha on GG, show that ABAB is a stable subgroup of GG. Let a∈Aa\in A and x∈A∩Bx\in A\cap B; then axa−1∈Baxa^{-1}\in B by the hypotheses made on AA and BB and clearly axa−1axa^{-1} belongs to AA, hence A∩BA\cap B is normal in AA. Let a∈Aa\in A and b,b′b,b' be in BB; the formula (ab)b′(ab)−1=a(bb′b−1)a−1(ab)b'(ab)^{-1}=a(bb'b^{-1})a^{-1} shows that BB is normal in ABAB. Let ϕ\phi be the restriction to AA of the canonical homomorphism of AB onto AB/B; then ϕ(a)=aB\phi(a) = aB and hence the kernel of ϕ\phi is equal to A∩BA \cap B. Clearly ϕ\phi is surjective and hence defines an isomorphism of A/(A∩B)A/(A \cap B) onto AB/BAB/B.

Theorem 4

Let G be a group with operators and N a normal stable subgroup of G.

(a) The mapping G′↦G′/NG' \mapsto G'/N is a bijection of the set of stable subgroups of G containing N onto the set of stable subgroups of G/N.

(b) Let G′G' be a stable subgroup of G containing N. For G′/NG'/N to be normal in G/NG/N, it is necessary and sufficient that G′G' be normal in G and the groups G/G′G/G' and (G/N)/(G′/N)(G/N)/(G'/N) are then isomorphic.

(c) Let G′G' be a stable subgroup of G. Then G′NG'N is a stable subgroup of G and N is normal in G′NG'N. Further G′∩NG' \cap N is normal in G′G' and the groups G′/(G′∩N)G'/(G' \cap N) and G′N/NG'N/N are isomorphic.

Let ff denote the canonical homomorphism of G onto G/NG/N. For all x∈Gx \in G, f(x)∈xNf(x) \in xN; therefore, f(G′)=G′/Nf(G') = G'/N for every subgroup G′G' of G containing N. As ff is surjective, assertion (a) follows from Corollary 1 to Proposition 7; similarly for the equivalence "G' normal" ⇔\Leftrightarrow "G'/N normal". Suppose that G′G' is a normal stable subgroup of G containing N. By no. 4, Proposition 5 applied to IdG\mathrm{Id}_G, there exists a homomorphism uu of G/NG/N into G/G′G/G' defined by u(xN)=xG′u(xN) = xG' for all x∈Gx \in G. It is immediate that uu is surjective with kernel G′/NG'/N whence the desired isomorphism of (G/N)/(G′/N)(G/N)/(G'/N) onto G/G′G/G'. Finally, (c) follows immediately from Proposition 8.

7. The jordan-hölder theorem

Definition 9

A composition series of a group with operators G is a finite sequence (Gi)0≤i≤n(G_i)_{0 \leq i \leq n} of stable subgroups of G, with G0=GG_0 = G and Gn={e}G_n = \{e\} and such that Gi+1G_{i+1} is a normal subgroup of GiG_i for 0≤i≤n−10 \leq i \leq n-1. The quotients Gi/Gi+1G_i/G_{i+1} are called the quotients of the series. A composition series Σ′\Sigma' is said to be finer than a composition series Σ\Sigma if Σ\Sigma is a series taken from Σ′\Sigma'.

If (Gi)0≤i≤n(G_i)_{0 \leq i \leq n} and (Hj)0≤j≤m(H_j)_{0 \leq j \leq m} are respectively composition series of two groups with operators G and H, they are said to be equivalent if m=nm = n and there exists a permutation ϕ\phi of the interval {0,n−1}\{0, n-1\} of N\mathbf{N}, such that the groups with operators Gi/Gi+1G_i/G_{i+1} and Hϕ(i)/Hϕ(i)+1H_{\phi(i)}/H_{\phi(i)+1} are isomorphic for all ii.

Note that in general a series taken from a composition series (Gi)(G_i) is not a composition series, since for j>i+1j > i + 1, GjG_j is not in general a normal subgroup of GiG_i.

Theorem 5 (Schreier). *Given two composition series Σ1,Σ2\Sigma_1, \Sigma_2 of a group with operators G, there exist two equivalent composition series Σ1′,Σ2′\Sigma'_1, \Sigma'_2, finer respectively than Σ1\Sigma_1 and Σ2\Sigma_2.

Let Σ1=(Hi)0≤i≤n\Sigma_1 = (H_i)_{0 \leq i \leq n} and Σ2=(Kj)0≤j≤p\Sigma_2 = (K_j)_{0 \leq j \leq p} be the two given composition series with respectively n+1n + 1 and p+1p + 1 terms; we shall see that the composition series Σ1′\Sigma'_1 can be formed by inserting p−1p - 1 subgroups Hi,j′H'_{i,j} (1 \leq j \leq p - 1) between H_i and H_{i+1} for 0 \leq i \leq n - 1 and the series \Sigma_2' by inserting n - 1 subgroups K_{j,i}' (1 \leq i \leq n - 1) between K_j and K_{j+1} for 0 \leq j \leq p - 1; thus two series of pn + 1 stable subgroups of G will be obtained; by choosing suitably the inserted stable subgroups, we shall show that these series are equivalent composition series.

To this end note that H_i \cap K_j is a stable subgroup of H_i and of K_j, and hence (Theorem 4) H_{i+1}.(H_i \cap K_j) is a stable subgroup of H_i containing H_{i+1} and K_{j+1}.(H_i \cap K_j) is a stable subgroup of K_j containing K_{j+1}. If we write H_{i,j}' = H_{i+1}.(H_i \cap K_j) and K_{j,i}' = K_{j+1}.(H_i \cap K_j), H_{i,j+1}' is a stable subgroup of H_{i,j}' (0 \leq j \leq p - 1) and K_{j,i+1}' is a stable subgroup of K_{j,i}' (0 \leq i \leq n - 1). Moreover H_{i,0}' = H_i, H_{i,p}' = H_{i+1}, K_{j,0}' = K_j and K_{j,p}' = K_{j+1}. To show the theorem, it suffices to show that H_{i,j+1}' (resp. K_{j,i+1}') is a normal stable subgroup of H_{i,j}' (resp. K_{j,i}') and that the quotient groups H_{i,j}'/H_{i,j+1}' and K_{j,i}'/K_{j,i+1}' are isomorphic (0 \leq i \leq n - 1, 0 \leq j \leq p - 1). This follows from the following lemma by taking H = H_i, H' = H_{i+1}, K = K_j, K' = K_{j+1}.

Lemma 1 (Zassenhaus). Let H and K be two stable subgroups of a group with operators G and H' and K' normal stable subgroups of H and K respectively; then H'.(H \cap K') is a normal stable subgroup of H'.(H \cap K), K'.(K \cap H') is a normal stable subgroup of K'.(K \cap H) and the quotient groups with operators (H'.(H \cap K))/(H'.(H \cap K')) and (K'.(K \cap H))/(K'.(K \cap H')) are isomorphic.

By Theorem 4, H' \cap K = H' \cap (H \cap K) is a normal stable subgroup of H \cap K; similarly K' \cap H is a normal stable subgroup of K \cap H; hence (no. 6, Corollary 2) (H' \cap K)(K' \cap H) is a normal stable subgroup of H \cap K. By Theorem 4 applied to the group H,

H′.(H′∩K).(K′∩H)=H′.(H∩K′) H'.(H' \cap K).(K' \cap H) = H'.(H \cap K')

is a normal stable subgroup of H'.(H \cap K) and the quotient group

(H′.(H∩K))/(H′.(H∩K)) (H'.(H \cap K))/(H'.(H \cap K))

is isomorphic to

(H∩K)/((H′∩K).(K′∩H)). (H \cap K)/((H' \cap K).(K' \cap H)).

In the last quotient, H and H' on the one hand and K and K' on the other appear symmetrically; permuting them the stated result is obtained.

Definition 10

A Jordan-Hölder series of a group with operators G is a strictly decreasing decomposition series \Sigma such that there exists no strictly decreasing decomposition series distinct from \Sigma and finer than \Sigma.

Proposition 9

For a decomposition series of G to be a Jordan-Hölder series it is necessary and sufficient that all the quotients of the series be simple.

A decomposition series is strictly decreasing if and only if none of its successive quotients is reduced to the identity element. If a strictly decreasing composition series Σ\Sigma is not a Jordan-Hölder series, there exists a strictly decreasing composition series Σ′\Sigma' which is finer than Σ\Sigma and distinct from Σ\Sigma. There are therefore two consecutive terms Gi,Gi+1G_i, G_{i+1} of Σ\Sigma which are not consecutive in Σ′\Sigma'; let HH be the first term which follows GiG_i in Σ′\Sigma'; HH is a normal stable subgroup of GiG_i, containing Gi+1G_{i+1} and distinct from the latter; hence H/Gi+1H/G_{i+1} is a normal stable subgroup of Gi/Gi+1G_i/G_{i+1}, distinct from the latter and from the identity element; therefore Gi/Gi+1G_i/G_{i+1} is not simple. Conversely, if Σ\Sigma is a strictly decreasing composition series one of whose quotients Gi/Gi+1G_i/G_{i+1} is not simple, this quotient contains a normal stable subgroup other than itself and {e}\{e\}, whose inverse image in GiG_i is a normal stable subgroup HH of GiG_i, distinct from GiG_i and Gi+1G_{i+1} (Theorem 4); it suffices to insert HH between GiG_i and Gi+1G_{i+1} to obtain a strictly decreasing composition series distinct from Σ\Sigma and finer than Σ\Sigma.

Theorem 6 (Jordan-Hölder). Two Jordan-Hölder series of a group with operators are equivalent.

Let Σ1,Σ2\Sigma_1, \Sigma_2 be two Jordan-Hölder series of a group with operators GG; by applying Theorem 5 two equivalent composition series Σ1′,Σ2′\Sigma'_1, \Sigma'_2 are obtained which are respectively finer than Σ1\Sigma_1 and Σ2\Sigma_2; since the latter are Jordan-Hölder series, Σ1′\Sigma'_1 is identical with Σ1\Sigma_1 or is derived from it by repeating certain terms; the series of quotients of Σ1′\Sigma'_1 is derived from that for Σ1\Sigma_1 by inserting a number of terms isomorphic to the group {e}\{e\}; since Σ1\Sigma_1 is strictly decreasing, the series of quotients of Σ1\Sigma_1 is derived from that of Σ1′\Sigma'_1 by suppressing in the latter all the terms isomorphic to {e}\{e\}. Similarly for Σ2\Sigma_2 and Σ2′\Sigma'_2. As the series of quotients of Σ1′\Sigma'_1 and Σ2′\Sigma'_2 differ (up to isomorphism) only in the order of the terms, the same is true of Σ1\Sigma_1 and Σ2\Sigma_2; the theorem is proved.

Corollary

Let GG be a group with operators in which there exists a Jordan-Hölder series. If Σ\Sigma is any strictly decreasing composition series of GG, there exists a Jordan-Hölder series finer than Σ\Sigma.

Let Σ0\Sigma_0 be a Jordan-Hölder series of GG; by Theorem 5, there exist two equivalent composition series, Σ′\Sigma' and Σ0′\Sigma'_0 respectively finer than Σ\Sigma and Σ0\Sigma_0; the argument of Theorem 6 shows that, by suppressing from Σ′\Sigma' the repetitions, a sequence Σ′′\Sigma'' is obtained which is equivalent to Σ0\Sigma_0 and hence a Jordan-Hölder series, since all its quotients are simple (Proposition 9). As Σ\Sigma is strictly decreasing, Σ′′\Sigma'' is finer than Σ\Sigma, whence the corollary.

Remark

A group with operators does not always possess a Jordan-Hölder series; an example is given by the additive group Z\mathbf{Z} of rational integers: the sequence (2n⋅Z)n≥0(2^n \cdot \mathbf{Z})_{n \geq 0} is a strictly decreasing infinite sequence of (normal) subgroups of Z\mathbf{Z}; for all pp, the first pp terms of this sequence form with the group {0}\{0\} a strictly decreasing composition series; if there existed a Jordan-Hölder series for Z\mathbf{Z}, it would have at least p+1p+1 terms, by the Corollary to Theorem 6; absurd, since pp is arbitrary.

On the other hand, there exists a Jordan-Hölder series in every finite group with operators G: if G≠{e}G \ne \{e\}, among the normal stable subgroups of G distinct from G, let H1H_1 be a maximal subgroup; similarly define Hn+1H_{n+1} by induction as a maximal element in the set of normal subgroups of HnH_n, distinct from HnH_n, when Hn≠{e}H_n \ne \{e\}; the sequence of orders of the HnH_n is strictly decreasing, hence there exists nn such that Hn={e}H_n = \{e\} and the sequence consisting of G and the HiH_i (1≤i≤n1 \leq i \leq n) is by its formation a Jordan-Hölder series.

Definition 11

Let G be a group with operators; the length of G is the upper bound of the integers nn such that there exists a strictly decreasing composition series of G (Gi)0≤i≤n(G_i)_{0\leq i\leq n}.

If G admits a Jordan-Hölder series, the length of G is the number of successive quotients of this series, as follows from the Corollary to Theorem 6. If G does not admit a Jordan-Hölder series, its length is infinite; by Proposition 9, for every strictly decreasing series of G there exists a strictly finer strictly decreasing series. The group consisting of the identity element is the only group with operators of length zero. A group with operators is simple if and only if it is of length 1.

Let G be a group with operators, H a normal stable subgroup of G, K the quotient G/H and π:G→K\pi:G\to K the canonical homomorphism. Let

Σ′=(Hi)0≤i≤n \Sigma'=(H_i)_{0\leq i\leq n}

be a decomposition series of H and

Σ′′=(Kj)0≤j≤p \Sigma''=(K_j)_{0\leq j\leq p}

be a composition series of K. Writing Gi=π−1(Ki)G_i=\pi^{-1}(K_i) for 0≤i≤p0\leq i\leq p and Gi=Hi−pG_i=H_{i-p} for p≤i≤n+pp\leq i\leq n+p, a composition series

Σ=(Gi)0≤i≤n+p \Sigma=(G_i)_{0\leq i\leq n+p}

of G is obtained. The sequence of quotients of Σ\Sigma is obtained by juxtaposing the sequence of quotients of Σ′′\Sigma'' and the sequence of quotients of Σ′\Sigma'. If Σ′\Sigma' and Σ′′\Sigma'' are Jordan-Hölder series, Σ\Sigma is a Jordan-Hölder series of G, by Proposition 9. If H or K admits composition series of arbitrary length, so does G. We have proved:

Proposition 10

Let G be a group with operators and H a normal stable subgroup of G. The length of G is the sum of the lengths of H and G/H.

Corollary

Let G be a group with operators and (Gi)0≤i≤n(G_i)_{0\leq i\leq n} a composition series of G. The length of G is the sum of the lengths of the Gi/Gi+1G_i/G_{i+1}, 0≤i≤n−10\leq i\leq n-1.

If G and G′G' are isomorphic groups with operators and G admits a Jordan-Hölder series, so does G′G' and the Jordan-Hölder series of G and G′G' are equivalent. However, non-isomorphic groups can have equivalent Jordan-Hölder series; such is true for Z/4Z\mathbf{Z}/4\mathbf{Z} and (Z/2Z)×(Z/2Z)(\mathbf{Z}/2\mathbf{Z})\times(\mathbf{Z}/2\mathbf{Z}), cf. no. 10.

8. Products and fibre products

Let (Gi)i∈I(G_i)_{i \in I} be a family of groups with operators. Let GG be the product monoid of the GiG_i. Consider the action of Ω\Omega on GG defined by

((xi)i∈I)α=(xiα)i∈I(α∈Ω,xi∈Gi). ((x_i)_{i \in I})^\alpha = (x_i^\alpha)_{i \in I} \quad (\alpha \in \Omega, x_i \in G_i).

With this structure GG is a group with operators. For all i∈Ii \in I, the projection mapping pri:G→Gi\mathrm{pr}_i : G \to G_i is a homomorphism of groups with operators.

Definition 12

The group with operators* G=∏i∈IGiG = \prod_{i \in I} G_i defined above is the product group with operators of the GiG_i. The mappings pri:G→Gi\mathrm{pr}_i : G \to G_i are called the projection homomorphisms.

A particular case of the product of groups with operators is the group GEG^E consisting of the mappings of a set EE into a group with operators GG, the laws being defined by:

(fg)(x)=f(x)g(x)(f,g∈GE,x∈E) (fg)(x) = f(x)g(x) \quad (f, g \in G^E, x \in E)
fα(x)=f(x)α(f∈GE,α∈Ω,x∈E). f^\alpha(x) = f(x)^\alpha \quad (f \in G^E, \alpha \in \Omega, x \in E).

Let (ϕi:H→Gi)i∈I(\phi_i : H \to G_i)_{i \in I} be a family of homomorphisms of groups with operators. The mapping h↦(ϕi(h))i∈Ih \mapsto (\phi_i(h))_{i \in I} of HH into ∏i∈IGi\prod_{i \in I} G_i is a homomorphism of groups with operators. It is the only homomorphism Φ:H→∏i∈IGi\Phi : H \to \prod_{i \in I} G_i satisfying pri∘Φ=ϕi\mathrm{pr}_i \circ \Phi = \phi_i for all ii. This justifies the name "product group with operators" (Set Theory, IV, § 2, no. 4).

Let (ϕi:Hi→Gi)i∈I(\phi_i : H_i \to G_i)_{i \in I} be a family of homomorphisms of groups with operators. The mapping ∏i∈Iϕi:(hi)i∈I↦(ϕi(hi))i∈I\prod_{i \in I} \phi_i : (h_i)_{i \in I} \mapsto (\phi_i(h_i))_{i \in I} of ∏i∈IHi\prod_{i \in I} H_i into ∏i∈IGi\prod_{i \in I} G_i is a homomorphism of groups with operators.

Proposition 11

Let* (ϕi:Hi→Gi)i∈I(\phi_i : H_i \to G_i)_{i \in I} be a family of homomorphisms of groups with operators and let Φ=∏i∈Iϕi\Phi = \prod_{i \in I} \phi_i. *Then:

(a) Ker(Φ)=∏i∈IKer(ϕi)\mathrm{Ker}(\Phi) = \prod_{i \in I} \mathrm{Ker}(\phi_i); in particular, if all the ϕi\phi_i *are injective, Φ\Phi is injective.

(b) Im(Φ)=∏i∈IIm(ϕi)\mathrm{Im}(\Phi) = \prod_{i \in I} \mathrm{Im}(\phi_i); in particular, if all the ϕi\phi_i *are surjective, Φ\Phi is surjective.

This is immediate.

In particular, let (Gi)i∈I(G_i)_{i \in I} be a family of groups with operators and, for all ii, let HiH_i be a stable (resp. normal stable) subgroup of GiG_i. The product ∏i∈IHi\prod_{i \in I} H_i is a stable (resp. normal stable) subgroup of ∏i∈IGi\prod_{i \in I} G_i and the canonical mapping of ∏i∈IGi\prod_{i \in I} G_i onto ∏i∈I(Gi/Hi)\prod_{i \in I} (G_i / H_i) defines when passing to the quotient an isomorphism of

(∏i∈IGi)/(∏i∈IHi) \left( \prod_{i \in I} G_i \right) / \left( \prod_{i \in I} H_i \right)
onto ∏i∈I(Gi/Hi)\prod_{i \in I} (G_i / H_i). For example, let JJ be a subset of II. The subgroup GJG_J of ∏i∈IGi\prod_{i \in I} G_i consisting of the (xi)i∈I(x_i)_{i \in I} such that xi=eix_i = e_i for i∉Ji \notin J is a normal stable subgroup. The mapping ιJ\iota_J which associates with x=(xj)j∈Jx = (x_j)_{j \in J} the element y=(yi)i∈Iy = (y_i)_{i \in I} such that yi=eiy_i = e_i for i∉Ji \notin J and yi=xiy_i = x_i for i∈Ji \in J, is an isomorphism of ∏j∈JGj\prod_{j \in J} G_j onto GJG_J. The mapping prI−J\mathrm{pr}_{I-J} defines when passing to the quotient an isomorphism θJ\theta_J of the quotient group G/GJG / G_J onto ∏i∈I−JGi\prod_{i \in I-J} G_i. The composition prJ∘ιJ\mathrm{pr}_J \circ \iota_J is the identity mapping of ∏j∈JGj\prod_{j \in J} G_j. G/GJG / G_J is often identified with ∏i∈I−JGi\prod_{i \in I-J} G_i because of θJ\theta_J and ∏i∈JGi\prod_{i \in J} G_i with GJG_J because of ιJ\iota_J.

If J1J_1 and J2J_2 are disjoint subsets of II, it follows from the definitions that every element of GJ1G_{J_1} commutes with every element of GJ2G_{J_2}.

Definition 13

Let GG be a group with operators and (Hi)i∈I(H_i)_{i \in I} a family of normal stable subgroups of GG. Let pi:G→G/Hip_i : G \to G / H_i be the canonical homomorphism. GG is called the internal product (or product) of the family of quotient groups (G/Hi)(G / H_i) if the homomorphism g↦(pi(g))i∈Ig \mapsto (p_i(g))_{i \in I} is an isomorphism of GG onto ∏i∈IG/Hi\prod_{i \in I} G / H_i.

Let GG and HH be groups with operators and let ϕ\phi and ψ\psi be two homomorphisms of GG into HH. The set of elements xx in GG such that ϕ(x)=ψ(x)\phi(x) = \psi(x) is a stable subgroup of GG, called the coincidence group of ϕ\phi and ψ\psi. In particular, let ϕ1:G1→H\phi_1 : G_1 \to H and ϕ2:G2→H\phi_2 : G_2 \to H be homomorphisms of groups with operators; the coincidence group of the homomorphisms ϕ1∘pr1\phi_1 \circ \mathrm{pr}_1 and ϕ2∘pr2\phi_2 \circ \mathrm{pr}_2 of G1×G2G_1 \times G_2 into HH is called the fibre product of G1G_1 and G2G_2 over HH relative to ϕ1\phi_1 and ϕ2\phi_2. It is denoted by G1×HG2G_1 \times_H G_2 when there is no ambiguity about ϕ1\phi_1 and ϕ2\phi_2 and the restrictions p1p_1 and p2p_2 of pr1\mathrm{pr}_1 and pr2\mathrm{pr}_2 to G1×HG2G_1 \times_H G_2 are also called projection homomorphisms. Then ϕ1∘p1=ϕ2∘p2\phi_1 \circ p_1 = \phi_2 \circ p_2. The elements of G1×HG2G_1 \times_H G_2 are the ordered pairs (g1,g2)∈G1×G2(g_1, g_2) \in G_1 \times G_2 such that ϕ1(g1)=ϕ2(g2)\phi_1(g_1) = \phi_2(g_2). If fif_i is a homomorphism of a group with operators KK into GiG_i (i=1,2i = 1, 2) and ϕ1∘f1=ϕ2∘f2\phi_1 \circ f_1 = \phi_2 \circ f_2, there exist one and only one homomorphism ff of KK into G1×HG2G_1 \times_H G_2 such that fi=pi∘ff_i = p_i \circ f for i=1,2i = 1, 2.

9. Restricted sums

Let (Gi)i∈I(G_i)_{i \in I} be a family of groups with operators and, for i∈Ii \in I, let HiH_i be a stable subgroup of GiG_i. The subset of ∏i∈IGi\prod_{i \in I} G_i consisting of the (xi)i∈I(x_i)_{i \in I} such that the set of i∈Ii \in I with xi∉Hix_i \notin H_i is finite is a stable subgroup of ∏i∈IGi\prod_{i \in I} G_i equal to ∏i∈IGi\prod_{i \in I} G_i if II is finite. It is called the restricted sum of the GiG_i with respect to the HiH_i. When, for all ii except for a finite number, HiH_i is a normal stable subgroup of GiG_i, the restricted sum is a normal stable subgroup of the product. When, for all ii, the subgroup HiH_i is reduced to the identity element of GiG_i, the direct sum of the GiG_i with respect to the HiH_i is called simply the restricted sum of the GiG_i and is sometimes denoted by ∏i∈IGi\prod_{i \in I} G_i. For all i0∈Ii_0 \in I, the mapping ιi0:Gi0→∏i∈IGi\iota_{i_0}: G_{i_0} \to \prod_{i \in I} G_i defined by ιi0(x)=(xi)i∈I\iota_{i_0}(x) = (x_i)_{i \in I}, where xi0=xx_{i_0} = x and xi=eix_i = e_i if i≠i0i \neq i_0, is an injective homomorphism of groups with operators called the canonical injection. GiG_i is identified with the stable subgroup Im⁡(ιi)\operatorname{Im}(\iota_i). The subgroups GiG_i are normal. For i≠ji \neq j, the elements of GiG_i and GjG_j commute and Gi∩Gj={e}G_i \cap G_j = \{e\}. The group ∏i∈IGi\prod_{i \in I} G_i is generated by the set ⋃i∈IGi\bigcup_{i \in I} G_i.

Proposition 12

Let (ϕi:Gi→K)i∈I(\phi_i: G_i \to K)_{i \in I} be a family of homomorphisms of groups with operators such that, for all i∈Ii \in I and j∈Ij \in I with i≠ji \neq j, x∈Gix \in G_i, y∈Gjy \in G_j, the elements ϕi(x)\phi_i(x) and ϕj(y)\phi_j(y) of KK commute; there exists one and only one homomorphism of groups with operators Φ\Phi of ∏i∈IGi\prod_{i \in I} G_i into KK such that ϕi=Φ∘ιi\phi_i = \Phi \circ \iota_i for all i∈Ii \in I. For every element x=(xi)i∈Ix = (x_i)_{i \in I} of ∏i∈IGi\prod_{i \in I} G_i, Φ(x)=∏i∈Iϕi(xi)\Phi(x) = \prod_{i \in I} \phi_i(x_i).

If Φ\Phi and Φ\Phi are solutions to the problem, they coincide on ⋃i∈IGi\bigcup_{i \in I} G_i and hence on ∏i∈IGi\prod_{i \in I} G_i, whence the uniqueness of Φ\Phi. We now show the existence of Φ\Phi: for every element x=(xi)i∈Ix = (x_i)_{i \in I} of ∏i∈IGi\prod_{i \in I} G_i, let Φ(x)=∏i∈Iϕi(xi)\Phi(x) = \prod_{i \in I} \phi_i(x_i) (§ 1, no. 5). Clearly Φ∘ιi=ϕi\Phi \circ \iota_i = \phi_i for all ii and Φ\Phi commutes with homotheties; the formula Φ(xy)=Φ(x)Φ(y)\Phi(xy) = \Phi(x)\Phi(y) follows from § 1, no. 5, formula (9).

Definition 14

Let GG be a group with operators and (Hi)i∈I(H_i)_{i \in I} a family of stable subgroups of GG. GG is called the internal restricted sum (or restricted sum) of the family of subgroups (Hi)(H_i) if every element of HiH_i is permutable with every element of HjH_j for j≠ij \neq i and the unique homomorphism of ∏i∈IHi\prod_{i \in I} H_i into GG whose restriction to each of the HiH_i is the canonical injection is an isomorphism.

When II is finite, we also say, by an abuse of language, internal direct product (or direct product, or product) instead of internal restricted sum. Every stable subgroup H of G for which there exists a stable subgroup H' of G such that G is the direct product of H and H' is called a direct factor of G.

Proposition 13

Let G be a group with operators and (Hi)i∈I(H_i)_{i \in I} a family of stable subgroups of G such that every element of HiH_i is permutable with every element of HjH_j for j≠ij \neq i. For G to be the restricted sum of the family of subgroups (Hi)i∈I(H_i)_{i \in I}, it is necessary and sufficient that every element x of G be expressible uniquely in the form ∏i∈Iyi\prod_{i \in I} y_i, where (yi)i∈I(y_i)_{i \in I} is a family with finite support of elements of G with yi∈Hiy_i \in H_i for all i.

Obvious.

Proposition 14

Let G be a group with operators and (Hi)i∈I(H_i)_{i \in I} a finite family of stable subgroups of G. For G to be the restricted sum of the family of subgroups (Hi)(H_i), it is necessary and sufficient that each HiH_i be normal and that G be the product of the quotient groups (G/Hi)(G/H^i), where HiH^i is the subgroup generated by the HjH_j for j≠ij \neq i.

The condition is obviously necessary. Conversely, suppose G is the product of the Ki=G/HiK_i = G/H^i and let G be identified with the product of the KiK_i. Then HiH_i is identified with a subgroup of KiK_i, so that, for i≠ji \neq j, every element of HiH_i is permutable with every element of HjH_j; on the other hand, HiH^i is identified with the product of the KjK_j for j≠ij \neq i, hence Hi=KiH_i = K_i for all ii and G is the direct product of the HiH_i.

Proposition 15

Let G be a group with operators and (Hi)1≤i≤n(H_i)_{1 \leq i \leq n} a sequence of normal stable subgroups of G such that*

(H1H2…Hi)∩Hi+1={e}for 1≤i≤n−1, (H_1 H_2 \ldots H_i) \cap H_{i+1} = \{e\} \quad \text{for } 1 \leq i \leq n-1,
*the set H1H2…HnH_1 H_2 \ldots H_n is a normal stable subgroup of G, the restricted sum of the HiH_i.

By induction on nn, this is immediately reduced to proving the proposition for n=2n = 2. We show first that, if x∈H1x \in H_1 and y∈H2y \in H_2, x and y are permutable; for xyx−1y−1=(xyx−1)y−1=x(yx−1y−1)xyx^{-1}y^{-1} = (xyx^{-1})y^{-1} = x(yx^{-1}y^{-1}) and hence (since H1H_1 and H2H_2 are normal) xyx−1y−1∈H1∩H2xyx^{-1}y^{-1} \in H_1 \cap H_2, that is xyx−1y−1=exyx^{-1}y^{-1} = e, by the hypothesis. Moreover H1H2H_1 H_2 is a subset of G which is stable under the homotheties of G. It follows (by no. 3, Proposition 1) that H1H2H_1 H_2 is a stable subgroup of G and it is immediately verified that this subgroup is normal. Suppose finally that xy=x′y′xy = x'y', with x∈H1,x′∈H1,y∈H2,y′∈H2x \in H_1, x' \in H_1, y \in H_2, y' \in H_2; then x′−1x=y′y−1{x'}^{-1}x = y'y^{-1}, hence x′−1∈H1∩H2={e}{x'}^{-1} \in H_1 \cap H_2 = \{e\}, x′=xx' = x and similarly y′=yy' = y; H1H2H_1 H_2 is thus the direct product of H1H_1 and H2H_2.

When the group considered are commutative, the term direct sum is used instead of direct product.

10. Monogenous groups

Let a∈Za \in \mathbf{Z}; since aZa\mathbf{Z} is a subgroup of Z\mathbf{Z}, the relation between elements x,yx, y of Z\mathbf{Z} which states "there exists z∈Zz \in \mathbf{Z} such that x−y=azx - y = az" is an equivalence relation, which we agree, once and for all, to write as x≡y(moda)x \equiv y \pmod{a} or x≡y(a)x \equiv y(a) and which is called congruence modulo aa. Replacing aa by −a-a an equivalent relation is obtained, hence it may be supposed that a⩾0a \geqslant 0; for a=0a = 0, x≡y(0)x \equiv y(0) means x=yx = y, hence a relation distinct from equality is obtained only if a≠0a \neq 0: we shall therefore suppose in what follows that a>0a > 0 unless otherwise indicated.

For a>0a > 0, the quotient of Z\mathbf{Z} by the congruence x≡y(a)x \equiv y(a), that is the group Z/aZ\mathbf{Z}/a\mathbf{Z}, is called the group of rational integers modulo aa.

Proposition 16

Let aa be an integer >0> 0. The integers rr such that 0⩽r<a0 \leqslant r < a form a system of representatives of the equivalence relation x≡y(moda)x \equiv y \pmod{a} on Z\mathbf{Z}.

If xx is an integer ⩾0\geqslant 0, there exist (Set Theory, III, § 5, no. 6) integers qq and rr such that x=aq+rx = aq + r and 0⩽r<a0 \leqslant r < a and x≡r(moda)x \equiv r \pmod{a}. If xx is an integer ⩽0\leqslant 0, the integer −x-x is ⩾0\geqslant 0 and by the above there exists an integer rr such that 0⩽r<a0 \leqslant r < a and −x≡r(moda)-x \equiv r \pmod{a}. Writing r′=0r' = 0 if r=0r = 0 and r′=a−rr' = a - r if r>0r > 0, then

x≡−r≡r′(moda) x \equiv -r \equiv r' \pmod{a}

and 0⩽r′<a0 \leqslant r' < a. We now show that if 0⩽r<r′<a0 \leqslant r < r' < a, then r≢r′(moda)r \not\equiv r' \pmod{a}. Now r′−r<nar' - r < na for n⩾1n \geqslant 1 and r′−r>nar' - r > na for n⩽0n \leqslant 0, whence r′−r∉aZr' - r \notin a\mathbf{Z}.

Corollary

Let aa be an integer >0> 0. The group Z/aZ\mathbf{Z}/a\mathbf{Z} of rational integers modulo aa is a group of order aa.

Proposition 17

Let H\mathrm{H} be a subgroup of Z\mathbf{Z}. There exists one and only one integer a⩾0a \geqslant 0 such that H=aZ\mathrm{H} = a\mathbf{Z}.

If H={0}\mathrm{H} = \{0\}, then H=0Z\mathrm{H} = 0\mathbf{Z}. Suppose that H≠{0}\mathrm{H} \neq \{0\}. The subgroup H\mathrm{H} has an element x≠0x \neq 0. Then x>0x > 0 or −x>0-x > 0, and therefore H\mathrm{H} has elements >0> 0. Let aa be the smallest element >0> 0 in H\mathrm{H}. The subgroup aZa\mathbf{Z} generated by aa is contained in H\mathrm{H}; we show that H⊂aZ\mathrm{H} \subset a\mathbf{Z}. Let y∈Hy \in \mathrm{H}. By Proposition 16, there exists an integer rr such that y≡r(moda)y \equiv r \pmod{a} and 0⩽r<a0 \leqslant r < a. A fortiori y≡r(modH)y \equiv r \pmod{\mathrm{H}}, whence r∈Hr \in \mathrm{H}. But this is only possible if r=0r = 0 and therefore y∈aZy \in a\mathbf{Z}. The integer aa is unique: if H={0}\mathrm{H} = \{0\}, then necessarily a=0a = 0, and if H≠{0}\mathrm{H} \neq \{0\}, the integer aa is the order of Z/H\mathbf{Z}/\mathrm{H}.

Definition 15

A group is called monogenous if it admits a system of generators consisting of a single element. A finite monogenous group is called cyclic.

Every monogenous group is commutative (no. 3, Corollary 2 to Proposition 2). Every quotient group of a monogenous group is monogenous (no. 3, Corollary 3 to Proposition 2).

The additive group Z\mathbf{Z} is monogenous: it is generated by {1}\{1\}. For every positive integer aa, the group Z/aZ\mathbf{Z}/a\mathbf{Z} is monogenous, for it is a quotient of Z\mathbf{Z}.

Proposition 18

A finite monogenous group of order a is isomorphic to Z/aZ\mathbf{Z}/a\mathbf{Z}. An infinite monogenous group is isomorphic to Z\mathbf{Z}.

Let G be a monogenous group (written multiplicatively) and x a generator of G. The identity xmxn=xm+nx^m x^n = x^{m+n} (§ 1, no. 3, formula (1)) shows that the mapping n↦xnn \mapsto x^n is a homomorphism of Z\mathbf{Z} into G. Its image is a subgroup of G containing x and hence it is G. By no. 5, Theorem 3, the group G is isomorphic to the quotient of Z\mathbf{Z} by a subgroup, which is necessarily of the form aZa\mathbf{Z} (Proposition 17). If a>0a > 0, the group G is finite of order a and if a=0a = 0, the group G is isomorphic to Z\mathbf{Z}.

Proposition 19

Let a be an integer > 0. Let H be a subgroup of Z/aZ\mathbf{Z}/a\mathbf{Z}, b the order of H and c its index in Z/aZ\mathbf{Z}/a\mathbf{Z}. Then a=bca = bc, H=cZ/aZH = c\mathbf{Z}/a\mathbf{Z} and H is isomorphic to Z/bZ\mathbf{Z}/b\mathbf{Z}.

Conversely, let b and c be two integers > 0 such that a=bca = bc. Then aZ⊂cZa\mathbf{Z} \subset c\mathbf{Z} and cZ/aZc\mathbf{Z}/a\mathbf{Z} is a subgroup of Z/aZ\mathbf{Z}/a\mathbf{Z}, of order b and index c.

a=bca = bc by no. 4, Corollary to Proposition 4. By no. 7, Theorem 4, H is of the form H′/aZH'/a\mathbf{Z}, where H′H' is a subgroup of Z\mathbf{Z} and Z/H′\mathbf{Z}/H' is isomorphic to (Z/aZ)/H(\mathbf{Z}/a\mathbf{Z})/H and hence of order c. By Proposition 17 and the Corollary to Proposition 16, H′=cZH' = c\mathbf{Z} and hence H is monogenous. Finally, H is isomorphic to Z/bZ\mathbf{Z}/b\mathbf{Z} by Proposition 18. Conversely, if a=bca = bc, then aZ⊂cZa\mathbf{Z} \subset c\mathbf{Z} for a∈cZa \in c\mathbf{Z}: the quotient group (Z/aZ)/(cZ/aZ)(\mathbf{Z}/a\mathbf{Z})/(c\mathbf{Z}/a\mathbf{Z}) is isomorphic to Z/cZ\mathbf{Z}/c\mathbf{Z} (no. 7, Theorem 4) and hence of order c (no. 4, Corollary to Proposition 4) and index b (no. 4, Corollary to Proposition 4).

Corollary

Every subgroup of a monogenous group is monogenous.

Let a and b be two integers ≠0\neq 0. The relation b∈aZb \in a\mathbf{Z} is also written: b is a multiple of a, and also a divides b or a is a divisor of b.

Definition 16

An integer p>0p > 0 is called prime if p≠1p \neq 1 and it admits no divisor > 1 other than p.

Proposition 20

An integer p>0p > 0 is prime if and only if the group Z/pZ\mathbf{Z}/p\mathbf{Z} is a simple group.

This follows from Proposition 19.

Corollary

Every commutative simple group is cyclic of prime order.

Let G be such a group. Then G≠{e}G \neq \{e\}; let a≠ea \neq e be an element of G. The subgroup generated by a is normal since G is commutative, it is not reduced to {e}\{e\} and hence is equal to G. Therefore G is monogenous and hence isomorphic to a group of the form Z/pZ\mathbf{Z}/p\mathbf{Z} with p>0p > 0, since Z\mathbf{Z} is not simple, and pp is necessarily prime.

Remark

A finite group G of prime order is necessarily cyclic. G admits no subgroup other than G and {e}\{e\} and hence it is generated by every element ≠e\neq e.

Lemma 2

Let aa be an integer >0> 0. By associating with every composition series (Hi)0≤i≤n(\mathbf{H}_i)_{0 \leq i \leq n} of the group Z/aZ\mathbf{Z}/a\mathbf{Z} the sequence (si)1≤i≤n(s_i)_{1 \leq i \leq n}, where sis_i is the order of Hi−1/Hi\mathbf{H}_{i-1}/\mathbf{H}_i, a one-to-one correspondence is obtained between the composition series of Z/aZ\mathbf{Z}/a\mathbf{Z} and the finite sequences (si)(s_i) of integers >0> 0 such that a=s1…sna = s_1 \ldots s_n. The composition series (Hi)0≤i≤n(\mathbf{H}_i)_{0 \leq i \leq n} is a Jordan-Hölder series if and only if the sis_i are prime.

If (Hi)0≤i≤n(\mathbf{H}_i)_{0 \leq i \leq n} is a composition series of Z/aZ\mathbf{Z}/a\mathbf{Z}, it follows, by induction on nn, from no. 4, Proposition 4, that a=∏i=1n(Hi−1:Hi)a = \prod_{i=1}^n (\mathbf{H}_{i-1} : \mathbf{H}_i).

Conversely, let (si)1≤i≤n(s_i)_{1 \leq i \leq n} be a sequence of integers >0> 0 such that a=s1…sna = s_1 \ldots s_n. If (Hi)0≤i≤n(\mathbf{H}_i)_{0 \leq i \leq n} is a composition series of Z/aZ\mathbf{Z}/a\mathbf{Z} such that (Hi−1:Hi)=si(\mathbf{H}_{i-1} : \mathbf{H}_i) = s_i for 1≤i≤n1 \leq i \leq n, then necessarily ((Z/aZ):Hi)=∏1≤j≤isj((\mathbf{Z}/a\mathbf{Z}) : \mathbf{H}_i) = \prod_{1 \leq j \leq i} s_j as is seen by induction on ii, whence Hi=(∏j=1isj)Z/aZ\mathbf{H}_i = \left( \prod_{j=1}^i s_j \right) \mathbf{Z}/a\mathbf{Z} (Proposition 19), which shows the injectivity of the mapping in question. We now show its surjectivity. Writing Hi=(∏j=1isj)Z/aZ\mathbf{H}_i = \left( \prod_{j=1}^i s_j \right) \mathbf{Z}/a\mathbf{Z} for 0≤i≤n0 \leq i \leq n, a composition series of Z/aZ\mathbf{Z}/a\mathbf{Z} is obtained such that (Hi−1:Hi)=si(\mathbf{H}_{i-1} : \mathbf{H}_i) = s_i (Proposition 19). The second assertion follows from Proposition 20 and no. 7, Proposition 9.

Let P\mathfrak{P} denote the set of prime numbers.

Theorem 7 (decomposition into prime factors)

Let aa be a strictly positive integer. There exists one and only one family (vp(a))p∈P(v_p(a))_{p \in \mathfrak{P}} of integers >0> 0 such that the set of p∈Pp \in \mathfrak{P} with vp(a)≠0v_p(a) \neq 0 is finite and

a=∏p∈Ppvp(a). a = \prod_{p \in \mathfrak{P}} p^{v_p(a)}.

As the group Z/aZ\mathbf{Z}/a\mathbf{Z} is finite, it admits a Jordan-Hölder series. Lemma 2 then implies that aa is a product of prime integers, whence the existence of the family (vp(a))(v_p(a)); further, for every family (vp(a))p∈P(v_p(a))_{p \in \mathfrak{P}} satisfying the conditions of Theorem 7, the integer vp(a)v_p(a) is, for all p∈Pp \in \mathfrak{P}, equal to the number of factors of a Jordan-Hölder series of Z/aZ\mathbf{Z}/a\mathbf{Z} isomorphic to Z/pZ\mathbf{Z}/p\mathbf{Z} (Lemma 2). The uniqueness of the family (vp(a))(v_p(a)) therefore follows from the Jordan-Hölder theorem (no. 7, Theorem 6).

Corollary

Let aa and bb be two integers >0> 0. Then vp(ab)=vp(a)+vp(b)v_p(ab) = v_p(a) + v_p(b). For aa to divide bb, it is necessary and sufficient that vp(a)≤vp(b)v_p(a) \leq v_p(b) for every prime number pp.

In any group GG, if the (monogenous) subgroup generated by an element x∈Gx \in G is of finite order dd, xx is called an element of order dd; the number dd is therefore the least integer >0> 0 such that xd=ex^d = e; if the subgroup generated by xx is infinite, xx is said to be of infinite order. These definitions, together with Proposition 4 (no. 4), imply in particular that in a finite group G the order of every element of G is a divisor of the order of G.

Proposition 21

In a finite group G of order n, xn=ex^n = e for all x∈Gx \in G.

If pp is the order of xx, then n=pqn = pq, with qq an integer, and hence xn=(xp)q=ex^n = (x^p)^q = e.

Exercises

See the exercises for § 4.

Transcribed from the printed edition, 1998, Springer. Not translated.