Normal Subgroups, Quotient Groups, and Automorphisms

December 10, 2025 · Luciano Muratore

So far we have seen how groups can be generated and how they can be combined. Now we turn to a more structural question: what subgroups of a group are well-behaved enough to let us form a new group out of their cosets?

Normal Subgroups

A subgroup HGH \leq G is called normal if it is invariant under conjugation by any element of GG, that is, if gHg1=HgHg^{-1} = H for all gGg \in G, or equivalently, if left and right cosets coincide: gH=HggH = Hg for all gGg \in G.

There are several ways to recognize normality. One is that HH is a union of conjugacy classes. Another is that xHx1HxHx^{-1} \subseteq H for all xGx \in G — which, combined with the finite or symmetric argument, gives equality. And perhaps the most important one: HH is normal if and only if the cosets form a group G/HG/H, the quotient group.

A simple but useful criterion is the following: if [G:H]=2[G : H] = 2, then HH is normal. The reason is purely combinatorial — there are only two left cosets (HH and gHgH) and only two right cosets (HH and HgHg), so for gHg \notin H, both gHgH and HgHg must equal GHG \setminus H, forcing gH=HggH = Hg. The canonical example is AnSnA_n \trianglelefteq S_n, since [Sn:An]=2[S_n : A_n] = 2.

Normal Subgroups of S3S_3

Let us work through a concrete case. The symmetric group S3={e,(12),(13),(23),(123),(132)}S_3 = \{e, (12), (13), (23), (123), (132)\} has subgroups {e}\{e\}, (12)\langle(12)\rangle, (13)\langle(13)\rangle, (23)\langle(23)\rangle, A3A_3, and S3S_3 itself. Of these, only {e}\{e\}, A3A_3, and S3S_3 are normal.

Why are the order-2 subgroups not normal? Because conjugation sends one transposition to another — for example, (123)(12)(132)=(23)(123)(12)(132) = (23) — so (12)\langle(12)\rangle is not preserved by all conjugations. On the other hand, A3A_3 has index 2 and is therefore normal by the criterion above, and S3/A3Z2S_3/A_3 \cong \mathbb{Z}_2.

The Quaternion Group Q8Q_8

The quaternion group Q8={1,1,i,i,j,j,k,k}Q_8 = \{1, -1, i, -i, j, -j, k, -k\} with relations i2=j2=k2=1i^2 = j^2 = k^2 = -1, ij=kij = k, ji=kji = -k is another instructive example. Its center is Z(Q8)={1,1}Z(Q_8) = \{1, -1\}, and its subgroups are {1}\{1\}, {1,1}\{1,-1\}, i\langle i \rangle, j\langle j \rangle, k\langle k \rangle, and Q8Q_8.

Remarkably, every subgroup of Q8Q_8 is normal. The reason is that conjugation sends any generator to itself or its inverse: for example, jij1=ji(j)=jij1j \cdot i \cdot j^{-1} = j \cdot i \cdot (-j) = -ji \cdot j^{-1}… and one can check directly that each subgroup is preserved. The quotient groups are also illuminating: the index-2 quotients Q8/iQ_8/\langle i \rangle, Q8/jQ_8/\langle j \rangle, Q8/kQ_8/\langle k \rangle are all isomorphic to Z2\mathbb{Z}_2, and the index-4 quotient Q8/{1,1}Z2×Z2Q_8/\{1,-1\} \cong \mathbb{Z}_2 \times \mathbb{Z}_2.

Products of Normal Subgroups

If H,KGH, K \trianglelefteq G, then their intersection HKH \cap K is normal in GG, and their product HK={hkhH,kK}HK = \{hk \mid h \in H, k \in K\} equals HK\langle H \cup K \rangle and is also normal in GG. The key idea is that normality ensures conjugation preserves both HH and KK, so it preserves their product as well.

Homomorphisms

A map f:GHf : G \to H is a homomorphism if f(xy)=f(x)f(y)f(xy) = f(x)f(y) for all x,yGx, y \in G. The kernel ker(f)={xf(x)=eH}\ker(f) = \{x \mid f(x) = e_H\} is always a normal subgroup of GG, and the image Im(f)=f(G)\text{Im}(f) = f(G) is a subgroup of HH.

Some examples worth keeping in mind: f(x)=exf(x) = e^x is a homomorphism from (R,+)(\mathbb{R}, +) to (R>0,)(\mathbb{R}_{>0}, \cdot); f(x)=eixf(x) = e^{ix} maps R\mathbb{R} to the unit circle; and f(x,y)=xf(x,y) = x is a projection from Z2\mathbb{Z}^2 to Z\mathbb{Z}.

Endomorphisms and Automorphisms of Z\mathbb{Z}

Every endomorphism φ:ZZ\varphi : \mathbb{Z} \to \mathbb{Z} is determined by φ(1)=k\varphi(1) = k, giving φ(n)=kn\varphi(n) = kn. For φ\varphi to be an automorphism it must be bijective, which forces k=1|k| = 1. So Aut(Z)={φ1,φ1}Z2\text{Aut}(\mathbb{Z}) = \{\varphi_1, \varphi_{-1}\} \cong \mathbb{Z}_2: the only automorphisms of Z\mathbb{Z} are the identity and the inversion map nnn \mapsto -n.

Inner Automorphisms

For any gGg \in G, conjugation by gg defines an automorphism σg:GG\sigma_g : G \to G by σg(x)=gxg1\sigma_g(x) = gxg^{-1}. These satisfy σgσh=σgh\sigma_g \circ \sigma_h = \sigma_{gh} and (σg)1=σg1(\sigma_g)^{-1} = \sigma_{g^{-1}}, so the collection Int(G)={σggG}\text{Int}(G) = \{\sigma_g \mid g \in G\} forms a subgroup of Aut(G)\text{Aut}(G), and in fact a normal one.

Now consider the map f:GInt(G)f : G \to \text{Int}(G) defined by f(g)=σgf(g) = \sigma_g. It is a surjective homomorphism, and its kernel is precisely Z(G)Z(G), the center: gker(f)g \in \ker(f) if and only if gxg1=xgxg^{-1} = x for all xx, that is, if and only if gg commutes with everything. By the First Isomorphism Theorem,

G/Z(G)Int(G).G/Z(G) \cong \text{Int}(G).

This is a neat result: the group of inner automorphisms is completely determined by how far GG is from being abelian.

The Inversion Map

Let me close with a small observation that ties back to what we saw about groups of exponent 2. Consider the map f(x)=x1f(x) = x^{-1}. When is this a homomorphism?

For ff to satisfy f(xy)=f(x)f(y)f(xy) = f(x)f(y), we would need (xy)1=x1y1(xy)^{-1} = x^{-1}y^{-1}. But (xy)1=y1x1(xy)^{-1} = y^{-1}x^{-1}, so the condition becomes y1x1=x1y1y^{-1}x^{-1} = x^{-1}y^{-1}, which is exactly the statement that GG is abelian. So ff is a homomorphism if and only if GG is abelian — a compact way of saying that inversion reverses order, and the group only forgives that if order does not matter.