Normal Subgroups and Quotient Groups
Cosets
chopped a group G into equal-sized, non-overlapping tiles: the left
cosets gH of a subgroup H. That felt tantalisingly
close to something bigger — if the tiles are all the same size and perfectly partition the group, could
we treat each whole tile as a single new "super-element" and build a smaller group out of them?
Sometimes yes, sometimes no. The tiles form a group exactly when the subgroup is normal
— the one extra condition that makes coset multiplication behave. When it holds, the resulting group of
tiles is called the quotient group (or factor group) G/N,
and it is one of the most powerful constructions in all of algebra: it is how we "divide" one group by
another, how we collapse detail we don't care about, and — via
homomorphisms —
how every structure-preserving map secretly encodes a quotient.
This one page is about that single idea: what makes a subgroup normal, and why exactly that
condition is the key that unlocks the quotient group.
The definition of a normal subgroup
A subgroup N \le G is normal, written
N \triangleleft G, if any one of these equivalent conditions holds for
every g \in G:
-
gNg^{-1} = N — conjugating N by any element
gives back N (as a set);
-
gN = Ng — the left coset equals the right coset;
-
gng^{-1} \in N for all n \in N — conjugation
never escapes N.
The slogan to memorise is "left cosets equal right cosets." Ordinarily
gH and Hg are different sets — normality is
precisely the demand that they coincide, so there is only one family of cosets to talk about,
not two competing ones.
Why normality is exactly what makes the cosets a group
We want to multiply cosets by the natural rule
(aN)(bN) = (ab)N.
This looks obvious, but there is a hidden danger: a coset can be named many different ways. The coset
aN is the same set as a'N whenever
a' = an for some n \in N. For the rule to define a
genuine operation on tiles, the answer (ab)N must not depend on
which representatives a, b we happened to grab. This is called being
well-defined, and it is the whole ball game.
Take another pair of names a' = an_1 and
b' = bn_2 with n_1, n_2 \in N. Then
a'b' = a n_1 \, b\, n_2 = a b\,(b^{-1} n_1 b)\, n_2.
For this to land in the same coset (ab)N, we need the middle piece
b^{-1} n_1 b to sit inside N — that is exactly
b^{-1} N b \subseteq N, the normality condition! If
N is normal, b^{-1} n_1 b \in N, so
a'b' \in abN and the product is unambiguous. If N
is not normal, you can find representatives that give genuinely different answers, and the
"operation" isn't an operation at all.
-
If N \triangleleft G, the cosets form a group under
(aN)(bN) = abN, called the quotient group
G/N.
-
Its identity is N = eN, and the inverse of
aN is a^{-1}N.
-
Its size is the index: |G/N| = [G : N] = |G| / |N|.
See it: collapsing \mathbb{Z}_6 onto \mathbb{Z}_3
Let G = \mathbb{Z}_6 = \{0,1,2,3,4,5\} under addition mod
6, and take N = \langle 3 \rangle = \{0, 3\}.
Because \mathbb{Z}_6 is abelian, N is
automatically normal. Its cosets are the three tiles
\{0,3\}, \qquad 1+N = \{1,4\}, \qquad 2+N = \{2,5\}.
The figure below reveals the partition, then shows how adding tiles — e.g.
(1+N) + (2+N) = 3+N = N — makes the three cosets behave exactly
like \mathbb{Z}_3 = \{0,1,2\}. Step through it and watch the six-element group
collapse into a three-element one.
This is the archetype of the whole subject: \mathbb{Z}/n\mathbb{Z} \cong \mathbb{Z}_n.
The infinite group \mathbb{Z} modulo the subgroup
n\mathbb{Z} of multiples of n is precisely
"clock arithmetic" with n hours — the quotient is the mod-
n world.
More worked quotients
Even and odd: S_n / A_n. The alternating group
A_n (the even permutations) sits inside the symmetric group
S_n with index 2. Any subgroup of index
2 is automatically normal (each of its two left cosets, and each of its two
right cosets, is either the subgroup itself or "everything else"). So
A_n \triangleleft S_n, and the quotient has just two elements:
S_n / A_n = \{\, A_n,\ \tau A_n \,\} \cong \{+1, -1\} \cong \mathbb{Z}_2,
where \tau is any single transposition. The two cosets are exactly "even"
and "odd," and multiplying them follows the sign rule
(-1)(-1) = +1 — odd composed with odd is even.
The centre is always normal. The centre
Z(G) = \{ z : zg = gz \text{ for all } g\} is normal, because if
z commutes with everything then gzg^{-1} = z g g^{-1} = z \in Z(G).
So G / Z(G) always makes sense — a quotient that, remarkably, controls
whether G itself is abelian.
Everything in an abelian group is normal. If G is abelian
then gN = Ng trivially for every subgroup, so every subgroup is
normal and every subgroup gives a quotient. Non-abelian groups are where normality becomes a
real, discriminating condition.
Kernels: where normal subgroups come from
There is a beautiful two-way street between normal subgroups and
homomorphisms.
-
Every kernel is normal. If \varphi : G \to H is a
homomorphism, then \ker\varphi = \{g : \varphi(g) = e_H\} is a normal
subgroup of G.
-
Every normal subgroup is a kernel. Given N \triangleleft G,
the quotient map \pi : G \to G/N,\ \pi(g) = gN is a surjective
homomorphism whose kernel is exactly N.
Why is a kernel normal? For n \in \ker\varphi and any
g,
\varphi(gng^{-1}) = \varphi(g)\,\varphi(n)\,\varphi(g)^{-1} = \varphi(g)\, e_H\, \varphi(g)^{-1} = e_H,
so gng^{-1} \in \ker\varphi — the kernel survives conjugation, which is
exactly normality. Together these two facts say: normal subgroups are precisely the subgroups
you can quotient by, and equivalently precisely the subgroups that arise as kernels. This is
the doorway to the First Isomorphism Theorem, G/\ker\varphi \cong \operatorname{im}\varphi.
Trap 1 — "setwise" is not "elementwise." The condition
gNg^{-1} = N is an equality of sets. It does not say
gng^{-1} = n for each individual n (that stronger
condition would make every element of N commute with
g — that's being central, not merely normal). Normality only asks
that conjugation permutes the elements of N among themselves; it may
shuffle them around, as long as nothing leaves the set.
Trap 2 — normality is not transitive. It is tempting to think
N \triangleleft H and H \triangleleft G force
N \triangleleft G. They do not! The smallest counterexample lives in the
symmetries of a square, D_4: you can have
N \triangleleft H \triangleleft D_4 with N
not normal in D_4, because an element of the big group can conjugate
N out of place even though elements of H cannot.
Normality is a relationship between a subgroup and one specific parent group — never assume it
transfers up the chain.
A group with no normal subgroups except the two trivial ones (\{e\}
and the whole group) is called simple. Simple groups are the atoms of group theory:
because you can't quotient a simple group down into anything smaller, they play the role that
prime numbers play in arithmetic. Just as every whole number factors into primes, every finite
group can be broken apart, layer by layer via quotients, into a stack of simple pieces (its
composition factors) — a result called the Jordan–Hölder theorem.
That analogy drove one of the great collective achievements of twentieth-century mathematics: the
Classification of Finite Simple Groups, a proof spanning tens of thousands of pages by
hundreds of authors. It says the finite simple groups fall into a few infinite families
(cyclic groups of prime order, the alternating groups A_n for
n \ge 5, and the groups of Lie type) plus exactly 26 sporadic
exceptions — the largest of which, the "Monster," has more elements than there are atoms in Jupiter.
And the humble alternating group A_5, of order 60,
is the very smallest non-abelian simple group — the reason there's no formula for the roots of a general
quintic.