Generalized Cayley’s Theorem

July 2023
Tim Bates

Everyone who has taken an introductory course in algebra has likely seen Cayley’s theorem: that every
group G of order n can be embedded into Sn . It is one of the first applications that we see of group actions
which yields a rather pretty result. It is also likely the first representation theorem that we encounter. We
will remind ourselves of the proof:

Proof. Let G act on itself via left translation. Fix a ∈ G, then g 7→ ag is a bijection on G (since we can
compose with the action by a−1 and obtain identity). Now let us view the left action by a as a permutation
calling it La . La ∈ SG is clearly a symmetry of G by the previous discussion, now we simply need to
show that the following map φ : G → SG defined by a 7→ La is an injective homomorphism. To show it is a
homomorphism let a, b ∈ G, then Lab (x) = (ab)x = a(bx) = La Lb (x) is a homomorphism by the associativity
of group actions. Injectivity follows since if a, b ∈ G with a ̸= b then La (e) = a ̸= b = Lb (e).

Notice that this proof does not suppose the order of G to be finite; however, we will be interested in cases
where |G| < ∞. Another thing to note is that the embedding of G (with |G| = n) in Sn is a simply transitive
group acting on G (which we identified with {1, . . . , n}). A simply transitive group action is one where given
a, b ∈ G there is a unique σ such that σ(a) = σ(b). This follows immediately from the cancellation property
of groups (ax = bx ⇒ a = b). Now we want to show what Rotman [1] calls a generalization of Cayley’s
theorem.

Theorem 1. If H ≤ G is a subgroup of G with [G : H] = n, then there exists a homomorphism

It is a generalization since if G is finite and we take H = ⟨e⟩, then we can find a homomorphism
φ : G → Sn with ker φ = ⟨e⟩, thus forcing φ to be injective. Now we prove this statement.

Proof. Let H denote the collection of all left cosets of H in G. We will let G act on H and then find our
homomorphism. Let a ∈ G and define φa : H → H by gH 7→ agH. This is a well defined function on
H since if gH and g ′ H represent the same coset then (ag ′ )−1 ag = g ′−1 a−1 ag = g ′−1 g ∈ H. It also again
represents a bijection on H due to the existence of inverses which compose to form the identity permutation
for any a ∈ G. Now let φ : G → SH ∼ = Sn by g 7→ φa . This clearly is a homomorphism by the same
argument as in the proof of Cayley’s theorem. Now let a ∈ ker φ, then it follows that aH = H which in turn
implies a ∈ H, thus ker φ ≤ H and we are done.

Now a couple of nice results follow from this.

Proof. We know from Theorem 1 that there is a homomorphism φ : G → Sn with ker φ ≤ H, but since G is
normal, it follows that ker φ = ⟨e⟩ thus φ is injective and constitutes an embedding.

This corollary can be quite useful when analyzing the subgroup structure of simple groups.
Corollary 2. A6 has no subgroup of prime index.

Proof. First, |A6 | = 360 = 23 · 32 · 5. By Lagrange’s theorem, the index of any subgroup must divide the
order of A6 . Thus if A6 were to have a subgroup of prime index, then the index would need to be either
2, 3, or 5. Since A6 is simple, it follows from the previous corollary to the generalized Cayley theorem that
if there were to be subgroups of index 2, 3, or 5, then A6 could be embedded into S2 , S3 , or S5 . However,
|S2 | = 2 < |A6 |, |S3 | = 6 < |A6 |, and |S5 | = 120 < |A6 |, thus no embedding can exist because not even an
injective map could exist from A6 to S2,3,5 . We conclude that A6 admits no prime index subgroups.

1
 This means for example that A6 does not admit a subgroup of order 64. We can potentially generalize
this to other cases. Recall that An is simple for n > 4.
Proposition 1. Let k > 3, A2k has no subgroup of prime index.

Proof. By Lagrange’s theorem, the index must divide the order of A2k so the largest potential prime index
subgroup has index [A2k : H] = 2k − 1. Suppose that such a subgroup exists, then there is by Generalized
Cayley’s thereom a homomorphism φ : A2k → S2k−1 , however

(2k)!
|S2k−1 | = (2k − 1)! <
2
for k > 3. Since such a map cannot be injective, we reach a contradiction since A2k is simple.

This argument also holds for odd composite An ; however, fails for Ap when p is prime. This also clearly
fails for A4 since there is an index 3 subgroup isomorphic to V (the Klein-4 Group). Let’s explore another
example.
Example 1. Let D2n = a, b|an = e, b2 = e, bab = a−1 denote the dihedral groups (which have order 2n).
Note that for n at least 3, ⟨b⟩ (a subgroup of index n) is not a normal subgroup since aba−1 = a2 b ̸= e. By
Theorem 1 we see then that there is an embedding of D2n in Sn , which is a significant improvement over
Cayley’s theorem.

Let’s continue to think about these types of situations. We have a group G with a subgroup H which
contains no normal subgroups and is itself not normal. We can potentially utilize such an H to find better
embedding results.
Corollary 3. Let H < G be a proper subgroup containing no nontrivial normal subgroup of G with [G :
H] = n, then G ,→ Sn .

Proof. By Theorem 1, there is a homomorphism φ : G → Sn with ker φ ≤ H but since H contains no

Proof. The only normal subgroups of Sn (with n > 4) are ⟨e⟩ , An and Sn . This means that any index k
subgroup with k ∈ {3, . . . , n − 1} is not normal and does not contain An , thus ker φ = ⟨e⟩, but this implies
that there is an embedding Sn → Sk with k < n, a contradiction.

Proposition 3. There is no embedding of Zp into Sn with n < p.

Proof. The only subgroups of Zp are the whole group and the trivial subgroup, thus any embedding will be
into Sp or Sn with n ≥ p.

Notice that we did not need the generalization of Cayley’s theorem for the previous proposition as it is
obvious that there is no element of order p in Sn with n < p and thus no embedding can exist. Nevertheless
we bring it up to demonstrate the versatility of Theorem 1. The utility comes mostly in the fact that we
now have some bounds on the index of subgroups of G.
Proposition 4. A group of order 2n with n odd has a subgroup of index two.

Proof. Here we utilize Cayley’s theorem. Let G act on itself by left translation, then we have an embedding
φ of G into S2n which we call imφ. Since there are elements of order 2 in G we know that these elements
in imφ will consist of disjoint transpositions. Since this embedding is simply transitive, it follows that there
must be n transpositions. Since n is odd, we know that the elements of order two have a negative sign. Thus
the map Sgn : im(φ) → {±1} is surjective and ker Sgn ∼ = Im(φ)/{±} is an index two subgroup.

2
 Now I want to end with a discussion of representation theory.
Definition 1. Let V be a vector space over a field of characteristic 0. A representation of a group G on V
is a map
ρ : G → GL(V )

This definition can be generalized to R-modules rather than vector spaces, but for the moment a vector
space over C will be sufficient.
Theorem 2. Sn is isomorphic to the group of n × n nonsingular matrices with only 1’s (the permutation
matrices).
 
1 2... n
We can see this isomorphism by defining the map which takes Sn ∋ σ = to
σ(1) σ(2) . . . σ(n)
the matrix which has a 1 in the (σ(i), i)th entry. Now this collection of permutation matrices is obviously a
subgroup of GLn (C) and thus a group homomorphism φ : G → Sn ,→ GLn (C) constitutes a representation
of G in GLn (C). The representation of G obtained via our proof of Cayley’s theorem is often called the left
regular representation of G. At the very beginning we could have just as well taken the right action of G on
itself and obtained the right regular representation of G.

References
[1] Joseph Rotman. An introduction to the Theory of Groups. 1995.