Compatible relations and normal subgroups
Let G a group and ρ an equivalence relation defined on G, compatible with the operations of the group, i.e. such that
g1 ρ g'1, g2 ρ g2' ⇒ g1g2 ρ g'1g'2
Remembering the definitions of ρr, right and ρl, left congruence modulo a subgroup, we introduce the following proposition:
7.12.1 Proposition. Let G a group and ρ an equivalence compatible relation defined on G. Let also
H = {x ∈ G | x ρ e}
H is a subgroup of G
Indicating with ρr and ρl, the right and left congruence relations modulo H, it results
ρ = ρr = ρl
That is
x ρ y ⇐⇒ xy−1 ∈ H ⇐⇒ y−1x ∈ H
Proof.
We have
e ρ e ⇒ e ∈ H
Moreover ∀x,y ∈ H
x ρ e, y ρ e ⇒ xy ρ ee = e ⇒ xy ∈ H
And
x ρ e, x−1 ρ x−1 ⇒ e ρ x−1 ⇒ x−1 ∈ H
x ρ y and y−1 ρ y−1 imply owing to the compatibility of ρ: xy−1 ρ e, that is x ρr y. Conversely, if x ρr y then xy−1 ∈ H, that is xy−1 ρ e, from which for the compatibility of ρ: x ρ y. Thus x ρ y ⇒ x ρr y. The same argument can be applied to ρl. □
The previous proposition states that given in G an equivalence relation compatible with the operations of the group, the partition in equivalence classes coincides with the partition in right (left) cosets modulo the subgroup H with this latter representing the class of the nuetral element of the group. Moreover left and right cosets coincide i.e. Hx = xH, ∀x ∈ G.
We now ask the converse given a subgroup H of G, when the relations ρr, ρl modulo H are compatible? It is not generally the case that the two relation are compatible as the following examples shows off:
7.12.2 Example. Let G = 𝓢3 and H the subgroup H = ⟨(23)⟩.
For ρr we have: (13) ρr (123), (12) ρr (132), (e) ρr (23)
For ρl we have: (13) ρl (132), (12) ρl (123), (e) ρl (23).
Neither of the two relations are compatible; We've to verify that for example
(13) ρr (123), (12) ρr (132) ⇒ (13)(12) ρr (123)(132)
But this is not the case since:
(123)(132) = id ∈ H but (13)(12) = (123) ∉ H, so it cannot be true that (13)(12)id−1 ∈ H.
Observe also that ρr ≠ ρl. Indeed there exist elements x,y such that xy−1 ∈ H but yx−1 ∉ H. It suffices to take for example x = (132) and y = (12). We have xy−1 = (23) ∈ H, but y−1x = (13) ∉ H. ■
The following proposition clarify under which additional condition ρr and ρl are compatible.
7.12.2 Proposition. Let H a subgroup of G, and let ρr and ρl the right and left congruence relation modulo H. The necessary and sufficient condition for ρr (or ρl) to be compatile is
ρr = ρl, Hx = xH, ∀x ∈ G
Proof. We observe first that by multiplying to the right (or left) by a same element of G two elements still congruent modulo ρr (ρl). Let ρr = ρl = ρ and let x ρ x' and y ρ y'. Then we have xy ρ x'y, from which
xy ρ x'y
that is ρr (= ρl) is compatible.
Conversely suppo that ρr is compatible. Then x ρr y ⇐⇒ xy−1 ρr e. But then, owing to the compatibility of ρr, we have y−1 ρr x−1, that is y−1(x−1)−1 = y−1x ∈ H, thus x ρr y ⇐⇒ x ρl y and ρr = ρl. □
7.12.3 Corollary. Let ρ an equivalence relation compatible defined on G. Then the subset H = {x ∈ G | x ρ e} is a subgroup of G and it holds that ρ = ρr = ρl modulo H. Conversely, if H is a subgroup of G such that ρr = ρl, then ρr = ρl = ρ is a compatible relation. □
Quotient groups and Normal subgroups
With the equivalence relations defyining right and left cosets 7.8.1 and 7.8.2, we denote the totality of all equivalence classes modulo H by G/H, known as the quotient set of the group G over its subgroup H. This set is the set of cosets G/H = {gH| g ∈ G}. Analogously, the set of right cosets G/H = {Hg | g ∈ G}.
An interesting question is under what conditions on the subgroup H the quotient set G/H becomes a quotient group.
The subgroup H of G such that ρr (and thus also ρl) is compatible, and for which the right and left cosets coincide are known as normal.
7.12.4 Defintion. A subgroup H of G is said normal in G if ∀x ∈ G
Hx = xH
We write H ⊲̲ G. In group theory, (big group)/(subgroup) nearly always means the quotient group, meaning that the subgroup is normal.
Every subgroup of an abelian group is normal, hence all congruence relations modulo a certain subgroup are compatible. In addition every group has two trivial normal subgroups itself and the subgroup reduced to the identity element.
For example, in S3 the alternating subgroup is a normal subgroup whilst the subgroup {id, (12)} is not.
7.12.5 Example. Let H = A3 = {(e), (1, 2, 3), (1, 3, 2)} = ⟨(1, 2, 3)⟩
and
G = S3 = {(e), (1, 2, 3), (1, 3, 2), (1, 2), (1, 3), (2, 3)}.
For x = (1, 2) we have
xH= {(1, 2)(1), (1, 2)(1, 2, 3), (1, 2)(1, 3, 2)} = {(1, 2), (2, 3), (1, 3)}
and
Hx = {(1)(1, 2), (1, 2, 3)(1, 2), (1, 3, 2)(1, 2)} = {(1, 2), (1, 3), (2, 3)}.
Note that the condition xH = Hx is an equality of sets, and it does not require that xh ≠ hx for all h ∈ H. We have xH = Hx, but xh ≠ hx when h = (1, 2, 3) ∈ H. Similar computations show that
(e)H = (1, 2, 3)H = (1, 3, 2)H = {(1), (1, 2, 3), (1, 3, 2)} = H
H(e) = H(1, 2, 3) = H(1, 3, 2) = {(1), (1, 2, 3), (1, 3, 2)} = H
(1, 2)H = (1, 3)H = (2, 3)H = {(1, 2), (1, 3), (2, 3)}
H(1, 2) = H(1, 3) = H(2, 3) = {(1, 2), (1, 3), (2, 3)}.
Thus H is a normal subgroup of G. Additionally we note that G can be expressed as G = H ∪ (1, 2)H. ■
Remark. In the definition of normal subgroup we have specified normal in G. Indeed a subgroup N may be normal in a subgroup H but not in the whole G. The previous corollary states that all and the only compatible relations defined in a group G are the relations congruent modulo a normal subgroup. If we set
𝓡 := {equivalence relations defined in G, compatible with the operations in G}
𝓝 := {normal subgroups in G}
The mapping
Ψ: | 𝓡 | ⟶ | 𝓝 |
ρ | ⟼ | 𝓝 = {x ∈ G | x ρ e} |
is a bijective function between 𝓡 and 𝓝, whose inverse is
Ψ*: | 𝓝 | ⟶ | 𝓡 |
N | ⟼ | ρ = ρr = ρl |
where x ρ y ⇐⇒ xy−1 ∈ N.
The role that ideals play in ring theory is played by normal subroups in group theory. ■
As we did for the case of Sn we introduce the definition of conjiugate elements.
7.12.6 Defintion. Let G a group. Two elements x,y ∈ G are said conjugates if there exists an element g ∈ G such that
y = gxg−1
Every element of the set gH is equal to some element of Hg, i.e. ghi = hjg or hi = g−1hjg.
Conjugacy is an equivalence relation, since it is
Reflexive: x = e−1xe for any x ∈ G (x is conjugate to itself).
Symmetric: if y = g−1xg, then x = (g−1)−1yg−1.
Transitive: if y = g−1xg and z = h−1 yh, then z = (gh)−1 x(gh).
Hence, G is the disjoint union of the equivalence classes, which are called conjugacy classes.
Given a subgroup H of G we indicate by xHx−1, the set
Hx = xHx−1 := {xHx−1 | h ∈ H}
It is easy to show that it is still a subgroup of G, known as Conjugated subgroup of H, and may be also indicated with Hx. It coincides with H if and only if H is a normal subroup of G.
We now give a list of equivalence definition of normal subgroup
Proposition 7.12.7. Let N a subgroup of a group G. Then the following propositions are equivalent
N is normal in G
Nx = N ∀x∈G
xnx−1 ∈ N, ∀x ∈ G and ∀n ∈ N
N is the uniion of conjugate classes of G.
Proof. b) xNx−1 = N if and only if xN = Nx, for one direction, we multiply on the right by x, and, for the converse we multiply on the right by x−1.
The condition for normality x−1Nx = N for all x ∈ G, says that x−1nx ∈ N for all n ∈ N and all x ∈ G; that is, for every element of N, its entire conjugacy class is contained in N. □
Definition 7.12.8. Let N be a normal subgroup in the group G. The factor group, or quotient group, G/N
G/N := {xN = Nx | x ∈ G}
is the set of all left (or right) cosets of N ∈ G. Our first goal is to define an operation on right cosets so that G/N becomes a group. We define it as follows
Nx * Ny := Nxy
Example. If G = {e,a,a2} = C3, and N = {e} then G/{e} = {{e}, {a}, {a2}}.
7.12.9 Proposition. The quotient group G/N, as defined above, is indeed a group.
Proof. First we have to verify that the definition is well-defined. Let Nx = Nx' (that is x' = nx), Ny = Ny' (that is y' = n'y), where n,n' ∈ N, x,y ∈ G: we must prove that Nxy = Nx'y'. Indeed x'y' = nxn'y = nn̅xy. Where the last equality holds because xn' ∈ xN = Nx, and so xn' is equal to n̅x for some n̅ ∈ N. But then Nx'y' = Nxy, as required.
Group axioms, are routine and are left to the reader to verify. The closure law needs no proof, since the operation is well defined. For the associative law, we have
((Nx)(Ny)) (Nz) = (Nxy)Nz = N(xy)z,
(Nx) ((Ny)(Nz)) = Nx(Nyz) = Nx(yz),
and the right-hand sides are equal, by the associative law for G. Because (aN)(eN) = (ae)N = aN = (ea)N = (eN)(aN), we see that eN = N is the identity element in G/N. Finally the inverse, (a−1N)(aN) = (a−1a)N = eN = (aa−1)N = (a N)(a−1N) shows that a−1N = (aN)−1. ■
The quotient group comes as the image of a natural homomorphism, which (as in the case of rings) is called the canonical homomorphism. Recall that the elements of G/N are the cosets of N in G.
Proposition 7.12.10 If φ: G → G' is any group homomorphism, then ker φ is a normal subgroup of G.
Proof. We already know that ker φ is a sugroup of G; to verify it is normal notice that
φ(gng−1) = φ(g)φ(n)φ(g−1) = φ(g) eG'φ(g)−1 = eG' ∀g ∈ G, ∀n ∈ ker φ
proving that gng−1 ∈ ker φ. □
We are now going that not only kernel ⇒ Normal, but that the converse is also true leading to kernel ⇐⇒ Normal subgroup. Define a map
φ: G → G/N
g ⟼ Ng
Checking that π is a homomorphism is straightforward from the definition of the operation in G/N.
π(g1g2) = (Ng1)(Ng2) = Ng1g2 = π(g1g2)
The image of π is G/N, since every coset has the form Ng for some g ∈ G.
For the kernel of π, since the identity element of G/N is the coset eGN, that is, N itself, we have
Ker(π) = {g ∈ G : Ng = N} = N.
Before we had noticed that every kernel (of a group homomorphism) is a normal subgroup; and now we have verified that every normal subgroup is in fact a kernel (of some group homomorphism). We conclude that
kernel ⇐⇒ Normal subgroup
in group theory, 'kernel' and 'normal subgroup' are equivalent concepts.
Furthermore, every subgroup in an abelian group is the kernel of some homomorphism: yet another indication that life is simpler in Abelian groups.
7.12.11 Example, Normal subgroups of D4 Here is the Cayley Table for D4
∘ | R0 | R1 | R2 | R3 | FA | FB | FC | FD |
---|---|---|---|---|---|---|---|---|
R0 | R0 | R1 | R2 | R3 | FA | FB | FC | FD |
R1 | R1 | R2 | R3 | R0 | FC | FD | FB | FA |
R2 | R2 | R3 | R0 | R1 | FB | FA | FD | FC |
R3 | R3 | R0 | R1 | R2 | FD | FC | FA | FB |
FA | FA | FD | FB | FC | R0 | R2 | R3 | R1 |
FB | FB | FC | FA | FD | R2 | R0 | R1 | R3 |
FC | FC | FA | FD | FB | R1 | R3 | R0 | R2 |
FD | FD | FB | FC | FA | R3 | R1 | R2 | R0 |
There are 10 distinct subgroups of D4. They are
{R0} which is normal
{R0, R2} which is normal: gR0g−1 ∈ {R0,R2} for all g ∈ D4 and n ∈ {R0, R2}.
{R0, FA} which is not normal.
{R0, FB} which is not normal.
{R0, FC} which is not normal.
{R0, FD} which is not normal.
{R0, R1, R2, R3} which is not normal.
{R0, R2, FA, FB} which is normal.
{R0, R2, FC, FD} which is normal.
D4 which is normal
7.12.12 Proposition. Every subgroup of index 2 of a group is a normal subgroup
Proof. To see this, suppose H is a subgroup of G such that [G : H] = 2. One coset is H itself: if x happens to be in H, then xH = H and Hx = H ⇒ xH = Hx. Now let a and element of G that is not contained in H. Then the left and right coset decompositions of G by H yield
G = H + aH = H + Ha
Accordingly, aH = Ha or aHq−1 = H, i.e. H is a normal subgroup of G. □