Symmetric Groups

1. BASIC RESULTS
Most of us are familiar with the symmetric groups. The symmetric group on n letters,
Sn , is the set of all bijective maps from Xn = {1, 2, · · · , n} to itself. This set forms a
group under the composition of maps. The elements of Sn are known as permutations.
Thus, if α and β are two permutations in Sn , their product αβ is defined to be the compo-
sition map αoβ (the order is important - here β acts first and then α). Let us compute one
product before moving on. Consider S5 and the permutations α, β and γ given below.

     
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
α= ,β = ,γ = .
2 3 1 4 5 3 2 4 1 5 1 3 2 5 4
Thus, (do it yourself) the product αβγ is given by the permutation
 
1 2 3 4 5
αβγ = .
1 4 3 5 2
A permutation α ∈ Sn is said to move a letter x ∈ Xn if α(x) 6= x. Otherwise, we say
that α fixes x. Two permutations α and β are said to be disjoint if the elements moved by
one are fixed by the other. In other words, α and β are disjoint if
α(x) 6= x =⇒ β(x) = x, and β(y) 6= y =⇒ α(y) = y.
Note that the definition does not say that α(x) = x implies β(x) 6= x. In fact, it is possible
that two disjoint permutations fix an element of Xn .
   
1 2 3 4 5 6 1 2 3 4 5 6
α= and β= .
2 1 3 4 5 6 1 2 3 4 6 5
It is clear that α and β given above are disjoint, although both fix 3 and 4. Another quick
computation shows that αβ = βα in this case, and both being equal to
 
1 2 3 4 5 6
.
2 1 3 4 6 5
This is in fact true for any two disjoint permutations.
Proposition 1. Let α and β be mutually disjoint permutations in Sn . Then
(i) αβ = βα, and
(ii) if αβ = 1, then α = 1 = β.
Proof. This is easy, and is left as an exercise. 

Next, we define a cycle. These are particular kind of elementary permutations that all
permutations can be factored into. A permutation ρ ∈ Sn is called a cycle or an r-cycle if
a) There are distinct elements {i0 , i2 , · · · , ir−1 } ⊆ {1, 2, · · · , n} such that M (ρ) =
{i0 , i2 , · · · , ir−1 } and
b) ρ(i` ) = i`+1 for all 0 ≤ ` ≤ r − 2 and ρ(ir−1 ) = α1 .
1
2

The integer r is called the length of the cycle ρ.

Remark: Let Y ⊆ Xn = {1, 2, · · · , n}. A subgroup H ≤ Sn is said to act transitively on
Y if for every pair x, y in Y , there is a σ ∈ H such that σ(x) = y. Thus. α is a cycle if
and only if hαi acts transitively on M (α).

Some properties of a r-cycle are listed below.

Proposition 2. Let ρ = (i0 i1 i2 · · · ir−1 ) be a r-cycle. Then we have the following.
(i) (i0 i1 i2 · · · ir−1 ) = (i1 i2 · · · ir−1 i0 ) = (i2 i3 · · · i0 i1 ) = · · · = (ir−1 i0 i1 · · · ir−2 ).
Thus, there are exactly r many r-cycles equal to a given r-cycle ρ (picturing this
becomes easy once we view the elements i0 , i1 , · · · , ir−1 as being arranged on a
circle).


(ii) There are exactly n(n − 1)(n − 2) · · · (n − r + 1) /r distinct r-cycles in Sn . In
particular, the number of n-cycles in Sn is (n − 1)!
(iii) Let ` ∈ {0, 1, · · · , r − 1}. Then for any integer k, one has ρk (i` ) = ik+` , where
k + ` is the unique integer in {0, 1, · · · , r − 1} that represents the residue class of
k + ` (mod r).
(iv) o(ρ) = r.
(v) If γ is another cycle such that there is a i = i` satisfying γ k (i) = ρk (i) for all
positive integers k, then γ = ρ.
Proof. Proofs of parts (i) and (ii) are left as exercises. For (iii), observe that
i` = ρ(i`−1 ) = ρ2 (i`−2 ) = · · · = ρ` (i0 ).
Thus, ρk (i` ) = ρk+` (i0 ). By division algorithm, we can write k + ` = qr + s where
0 ≤ s < r, i.e., s ∈ {0, 1, · · · , r − 1} is the unique integer representing k + `. Therefore,
noting that ρr (i0 ) = i0 , we have
ρk+` (i0 ) = ρs ρqr (i0 ) = ρs ρ(q−1)r ρr (i0 ) = ρs ρ(q−1)r (i0 ) = · · · = ρs (i0 ) = is = ik+` .
For part (iv), observe that for any ell ∈ {0, 1, · · · , r − 1}, we have
ρr (i` ) = ir+` = i` .
Thus, ρr is the identity map, i.e., ρr = 1. Therefore, o(ρ)|r. On the other hand, if o(ρ) = k,
then for any ` ∈ {0, 1, · · · , r − 1}, one has that
i` = ρk (i` ) = ik+` .
This means k + ` ≡ ` (mod r), which implies that r|k.
For (v), note that since, γ(i) = ρ(i), and as ρ moves i, we deduce that γ moves i. let
γ be a t-cycle. Thus o(γ) = t by (iv). We assume wlog that i = i0 (by (i)). As we now
show, it is also true that if x ∈
/ I := {i0 , i1 , i2 · · · , ir−1 } then γ(x) = x. Else, x is moved
by γ. Hence, there is an integer m such that γ m (i0 ) = x (as x and i0 belong to the same
cycle γ). But then
im = ρm (i0 ) = γ m (i0 ) = x,
i.e., we get that x ∈ I, a contradiction. Now, for any i` ∈ I, we have
i` = ρr (i` ) = ρr+` (i0 ) = γ r+` (i0 ) = γ r γ ` (i0 ) = γ r ρ` (i0 ) = γ r (i` ).
Furthermore, γ r fixes every element of Xn \I (Why?). It now follows that γ r = 1, and as
such, t|r.
 3

Conversely, we find that

i` = γ t (i` ) = γ t ρ` (i0 ) = γ t+` (i0 ) = ρt+` (i0 ) = it+` .
But this implies that t + ` ≡ ` (mod r), i.e., r|t. Therefore, we may conclude that t = r,
and that γ = ρ. 
For the next result, first observe that if we consider a permutation, say
 
1 2 3 4 5 6
α= .
2 1 5 4 6 3
This permutation can be decomposed into disjoint cycles. An easy calculation shows that
α = (12)(356) = (356)(12) = (12)(356)(4)
The method applied here shows that this process can be carried out with any permutation
in Sn . The above example also shows that there are not too many ways of doing this. i.e.,
factorization (without the order of the factors in which they appear) is essentially unique.
This is the content of our next result. The last factorization containing the 1-cycle (4) just
emphasizes the fact that we are inside the symmetric group S5 . Such a factorization where
all the elements of Xn appear in some cycle (including the 1-cycles) is called a complete
factorization. Complete factorizations are sometimes very convenient to work with.
Theorem 1. Every permutation α in Sn can be uniquely (as described above) expressed
as a product of disjoint cycles.
Proof. We will proceed by induction on the number r of elements moved by α. For r = 1,
there is nothing to prove. So we suppose that r > 1 and in our induction hypothesis, we
assume that if a permutation α0 moves < r elements, then α0 can be uniquely expressed as
a product of disjoint cycles. Since α moves some element in Xn , we label this element as
i1 . Let i2 = α(i1 ). Thus, i2 6= i1 . Next, set i3 = α(i2 ). If i3 ∈ {i1 , i2 }, then we stop.
Else, we take i4 = α(i3 ). If i4 ∈ {i1 , i2 , i3 }, then we stop. We continue in this fashion
until after k steps the element ik+1 ∈ {i1 , i2 , · · · , ik } for the first time. That is, k is the
least positive integer with this particular property. We claim that ik+1 = i1 . Note that
ik+1 = α(ik ), and that the elements i1 , i2 , · · · , ik are all distinct. Thus,
α(is ) 6= α(it ) ∀ s 6= t ∈ {1, 2, · · · , k}.
Now, suppose that α(ik ) = i` for some ` ≥ 2. Note that since 2 ≤ ` ≤ k, we have that ` −
1 ∈ {1, 2, · · · , k − 1}. But then α(i`−1 ) = i` forces that α(ik ) = α(i`−1 ), a contradiction
as k and ` − 1 (note that ` − 1 cannot be k) are distinct elements of {1, 2, · · · , k}.
Thus, using α and an element moved by α, we have gotten hold of a cycle ρ =
(i1 i2 · · · ik ) which is a part of α. Let Y be the set Xn \{i1 , i2 , · · · , ik }, and let α0 de-
notes the restriction αY of α to Y . Then it is clear that α0 and ρ are disjoint permutations.
We claim that α = ρα0 . To see this consider β = ρ−1 α. Note that α = ρ on the set
{i1 , i2 , · · · , ik }. Thus,
β(i` ) = ρ−1 α(i` ) = ρ−1 ρ(i` ) = i` .
/ {i1 , i2 , · · · , ik }. Also, ρ−1 fixes the
Now, let y ∈ Xn \{i1 , i2 , · · · , ik }. Note that α(y) ∈
elements of Xn \{i1 , i2 , · · · , ik } (Why?). Thus,
β(y) = ρ−1 (α(y)) = α(y).
It now follows that ρ−1 α = α0 , and as such, α = ρα0 . Now, since α0 moves less number of
elements than α, by our induction hypothesis, we find that α0 can be expressed as a product
of disjoint cycles, and consequently, α can be expressed as a product of disjoint cycles.
4

For the uniqueness part, assume that there are two different factorizations of α into
product of disjoint cycles. Let
α = ρ1 ρ2 · · · ρs = γ1 γ2 · · · γt
be two different factorizations of α into disjoint cycles. We assume wlog that α 6= 1, and
let x ∈ Xn be moved by α. Then some ρi moves x. Since ρi ’s commute, we may assume
wlog that i = s. Note that if j 6= s, then ρj (x) = x and, in fact, if j 6= s, then ρkj (x) = x
for any integer k. Again, using the commutativity of ρ’s, we find that for any integer k
αk (x) = ρk1 ρk2 · · · ρks (x) = ρks ρks−1 · · · ρk1 (x) = ρks (x).
Similarly, we may assume that γt moves x, and working exactly as before lead us to con-
clude that αk (x) = γtk (x) for all integers k. Therefore, ρks (x) = γtk (x) for all integers
k. Now, appealing to (v) of Proposition 2, we deduce that ρs = γt . Continuing in this
fashion, we end up getting something like
β1 β2 · · · β|s−t| = 1,
where βi are being chosen from leftover ρj or γj depending on whether s > t or t > s.
But since βi are disjoint, a result proved earlier shows that β1 = β2 = · · · = β|s−t| = 1,
and the result follows. 

Now that we have factorization of a permutation into disjoint cycles, we have an easy
formulation for the order of a permutation in terms of the lengths of its disjoint cyclic
factors.
Proposition 3. Let α ∈ Sn and let α = ρ1 ρ2 · · · ρs be the decomposition of α into disjoint
cycles ρi of length ri . Then o(α) = L = lcm(r1 , r2 , · · · , rs ).
Proof. Let o(α) = u. Since ρi commute with ρj and o(ρi ) = ri , we have that
α L = ρL L L
1 ρ2 · · · ρs = 1.

Thus, u|L. On the other hand αu = 1 yields that

ρu1 ρu2 · · · ρus = 1.
Since, ρuj are disjoint ∗, we deduce by Proposition 1 that
ρu1 = 1, ρu2 = 1, · · · ρus = 1.
Therefore, rj |u for all j = 1, 2, · · · , s. Consequently, L|u, and the result follows.
We prove ∗ now. If ρuj (x) 6= x, then ρj (x) 6= x either. Consequently, ρ` (x) = x for
any ` 6= j, and as such, ρu` (x) = x. Therefore, it follows that ρuj are mutually disjoint. 

A crucial corollary of Theorem 1 is the following result which eventually leads to the
notion of even and odd permutations.
Corollary 1. Every permutation can be expressed as a product (not unique) of transposi-
tions (2-cycles).
Proof. This is going to be another one liner. The proof follows immediately from the
observation that any cycle (i1 i2 · · · ik ) can be expressed as
(i1 i2 · · · ik ) = (i1 ik )(i1 ik−1 ) · · · (i1 i2 ).

 5

Note that a given permutation can be expressed in several ways as a product of transpo-
sition. For example, consider (123)(47) ∈ S7 . An easy calculation shows that
(123)(47) = (13)(12)(47) = (23)(13)(23)(12)(57)(45)(57).
Clearly, the number of transposition factors is not unique either So, is there anything at all
unique about this factorization? Indeed there is. It is the parity (even or oddness) of the
number of transposition factors that is unique about the factorizations into transpositions.
In our example these numbers are 3 and 7, i.e., both are odd. This particular kind of
uniqueness leads to the notion of even and odd permutations. For a discussion on even
and odd permutations, refer to Herstein. An important consequence of Corollary 1 is the
following result on generating set for Sn .
Corollary 2. The 2-cycle α = (12) and the n-cycle β = (12 · · · n) generates Sn (i.e.,
every permutation in Sn can be expressed as αe1 β f1 αe2 β f2 · · · where ei ∈ {0, 1} and
fj ∈ {0, 1, · · · , n − 1}).
Proof. Let H ≤ Sn be the subgroup generated by α and β Since Sn is generated by
transpositions, it is enough to show that any transposition (ij) can be generated by α and
β. We assume wlog that i < j and write j = i + k. Observe that one has β −1 =
(1 n n − 1 n − 2 · · · 3 2). Thus
β(12)β −1 = (12 · · · n)(12)(1 n n − 1 n − 2 · · · 3 2) = (23)
So, (23) ∈ H. Now conjugating (23) by β, we have
β 2 (12)β −2 = β(23)β −1 = (34) ∈ H.
Continuing in this fashion for ` − 1 steps, we find that
β `−1 (12)β −(`−1) = (`` + 1) ∈ H.
Thus, all transpositions of the form (`` + 1) are in H. Now an easy calculation shows that
(i i + k) = (i + k − 1 i + k)(i + k − 2 i + k − 1) · · · (i + 1 i + 2)(i i + 1) ∈ H.
This finishes the proof. 
Up next, we pay a little closer attention to conjugation of a permutation. Let α ∈ Sn .
Let α = γ1 γ2 · · · γs be the factorization of α into disjoint cycles. Note that if β ∈ Sn , then
Y
βαβ −1 = β(γ1 γ2 · · · γs )β −1 = βγi β −1 .
i

Thus, it suffices to see what a conjugation does to a r-cycle, namely, we would like to
know whether a conjugation of a r-cycle is a r-cycle and if so, how could we describe
this cycle. We consider a r-cycle γ = {i1 i2 · · · ir }. Let β ∈ Sn and that β(i` ) = j` for
` ∈ {1, 2, · · · , r}. If ρ = βγβ −1 , then
ρ(j` ) = βγβ −1 (j` ) = βγ(i` ) = β(i`+1 ) = j`+1 ∀ ` ∈ {1, 2, · · · , r − 1}.
Moreover,
ρ(jr ) = βγβ −1 (jr ) = βγ(ir ) = β(i1 ) = j1 .
Thus, the r-cycle (j1 j2 · · · jr ) is a part of ρ. To finish our proof, we note that if x ∈ /
{j1 j2 · · · jr }, then by virtue of β being a bijection, we find that β −1 (x) ∈
/ {i1 i2 · · · ir }.
But this means that γ(β −1 (x)) = β −1 (x). Hence, we find that
ρ(x) = βγβ −1 (x) = x.
6

Thus ρ fixes everything outside {j1 j2 · · · jr }, and consequently, we deduce the ρ = βγβ −1
is indeed the r-cycle. (j1 j2 · · · jr ). Thus, in summary
If γ = (i1 i2 · · · ir ), then βγβ −1 = (β(i1 )β(i2 ) · · · β(ir )).
As an immediate corollary, we have
Corollary 3. Conjugation in Sn preserves the cycle structure, i.e., if α = γ1 γ2 · · · γs and
where γi is a ri -cycle, then for any β ∈ Sn ,
βαβ −1 = λi λ2 · · · λs ,
where λi is a ri -cycle.
It is really remarkable that the converse of Corollary 3 holds as well.
Proposition 4. If two permutations have the same cycle structure, then they are conjugate
of each other.
Proof. Consider α and β in Sn and assume that they have the same cycle structure. Here,
we will consider the complete factorization of α and β into product of disjoint cycles. That
is, every i ∈ Xn appears in some cycle (possibly a 1-cycle). This is going to get quite
difficult notationally. So bear with me. Let
α = (a1 a2 · · · as )(b1 b2 · · · bt ) · · · (c1 )(c2 ) · · · (d1 d2 · · · du )

β = (a01 a02 · · · as )(b01 b02 · · · b0t ) · · · (c01 )(c02 ) · · · (d01 d02 · · · d0u )
be the factorizations of α and β into product of disjoint cycles. Note that since we have
written down the complete factorizations, the sets
{a1 , · · · as , b1 , · · · bt , c1 , c2 , · · · , d1 , · · · , du } and {a01 , · · · a0s , b01 , · · · b0t , c01 , c02 , · · · , d01 , · · · , d0u }
are exactly equal to Xn . Thus, it makes sense to talk about the following permutation
 
a1 a2 · · · as b1 b2 · · · bt c1 c2 · · · d1 · · · du
π= .
a01 a02 · · · a0s b01 b02 · · · b0t c01 c02 · · · d01 · · · d0u
Now it is not too hard to check that β = παπ −1 , for from the description of conjugation,
one has
παπ −1 = (π(a1 )π(a2 ) · · · π(as ))(π(b1 )π(b2 ) · · · π(bt )) · · · (π(c1 ))π((c2 )) · · · (π(d1 )π(d2 ) · · · π(du ))

= (a01 a02 · · · as )(b01 b02 · · · b0t ) · · · (c01 )(c02 ) · · · (d01 d02 · · · d0u ) = β.

Thus, cycle-structures and conjugacy classes are synonymous in Sn .
Corollary 4. Let 1 ≤ r ≤ n. Then any two r-cycles are conjugate of one another.
For a positive integer n, let p(n) denote the number of ways of writing n as a sum of
positive integers including the expression n = n. The function p(·) is called the partition.
For example,
2 = 1 + 1, 3 = 1 + 2 = 1 + 1 + 1, 4 = 1 + 3 = 1 + 1 + 2 = 1 + 1 + 1 + 1 = 2 + 2.
Thus, p(1) = 1, p(2) = 2, p(3) = 3, p(4) = 5 and so on. Based on Corollary 3 and
Corollary 4, we may deduce that
Corollary 5. The number of distinct conjugacy classes in Sn is equal to p(n).
Proof. Exercise. 
 7

Problems.
1. Show that Sn is non-abelian for all n ≥ 3
2. Show that S3 ∼
= D3 .
3. Find the center Z(Sn ) for n ≥ 3.
4. Show, by an example, that the power of a cycle is not necessarily a circle. (Hint:
Consider a 6-cycle in S6 and look at its powers). In fact, make an attempt to show that if
d = gcd(n, k) and α is a n-cycle in Sn , then αk is a product of d many disjoint cycles,
each of length n/d.
5. Show that an r-cycle is an even permutation if and only if r is odd.
6. Let p be a prime number ≤ n. Is every element of order p in Sn a p-cycle? How about
the cases (n, p) ∈ {(5, 3), (6, 3), (7, 5), (8, 5), (9, 5)}.

2. S IMPLICITY OF An FOR n ≥ 5
A group is called simple if it has no normal subgroup other than {1} and itself. Trivial
examples of simple groups would be any cyclic group of prime order. In this section, we
establish the following key theorem on symmetric groups.
Theorem 2. The alternating group An is simple for all n ≥ 5.
Somewhat of similar nature is the following theorem.
Theorem 3. For n ≥ 5, the only nontrivial normal subgroup of Sn is An .
If you look at the literature, you will find various proofs of this result. I would like to
give a proof which uses Cauchy’s theorem, namely that
Theorem 4. Let G be a finite group and let p be a prime number dividing the order of G.
Then there is an element g ∈ G with o(g) = p.
I have given a proof of Theorem 4 in the case where G is abelian (Herstein’s version). If
you have not understood that, it is fine. Just take the statement of the theorem for granted.
Here, we will digress a little bit and prove the theorem in the non-abelian case. But we
will be using the class equation described in the p-group notes. If you don’t dig that, do
not bother to look at the proof of Cauchy’s theorem and skip straight to the simplicity of
An lemmas.
proof of Theorem 4. We assume G is non-abelian, i.e., G 6= Z(G). We write down the
class equation of G as
X
(1) |G| = |Z(G)| + |G|/|N (xα )|,
α∈I
xα∈Z(G)
/

where N (xα ) is the normalizer of xα . The proof is based on an induction argument. So

assume that if H is a group with |H| < |G|, and if p divides |H|, then H has an element
of order p. Note that in this particular case, we have |Z(G)| < |G|, and as such, we may
assume that p - |Z(G)| (else Z(G), and hence, G has an element of order p by induction
hypothesis). Next observe that since xα ∈ / Z(G), we have that N (xα ) < G (strictly
contained) (Explain!). Again by induction hypothesis, p - |N (xα )|. But it is given that
p | |G|. Thus, p divides the quantity |G|/|N (xα )| for all α ∈ I . Consequently, p divides
the second summand on the right-hand-side of (1). Also, p divides the left-hand-side |G|.
But this forces p | |Z(G)|, a contradiction, finishing the proof. 
8

The proof of the simplicity of An for n ≥ 5 follows from the following lemmas.
Lemma 1. For n ≥ 3, the alternating group An can be generated by all the 3-cycles.
Proof. Note that every 3-cycle is in An . Further observe that every permutation in An can
be expressed as a product of even number of transpositions. Thus, it suffices to show that
the product of any two transpositions can be expressed as a product of 3-cycles (why?).
Two adjacent transpositions can be in one of the following two forms
(ij)(jk) i, j, k are distinct or (ij)(kl) i, j, k, l all are distinct.
The first type can be expressed as (ij)(jk) = (ijk). For the second type, we note that
(jk)(jk) = 1 (transpositions have order 2). Thus, we have
(ij)(jk) = (ij)(jk)(jk)(kl) = (ijk)(jkl).
This completes the proof of the lemma. 
A permutation in Sn is said to be of type (2, 2) if it is product of two disjoint transpo-
sitions. An important consequence of Lemma 1 is the following result which we state as a
lemma.
Lemma 2. for n ≥ 5, An is generated by (2, 2) permutations.
Proof. Let H be the subgroup of Sn generated by all (2, 2) permutations. Note that any
(2, 2) permutation is in An . Consequently, H ≤ An . Now, we will show that any 3-cycle
can be generated by (2, 2) permutations. Since An is generated by 3-cycles, we may then
conclude that the (2, 2) permutations generate An . Consider a 3-cycle (ijk). Note that
(ijk) = (ij)(jk). Since n ≥ 5, we can pick two distinct symbols l and s, both outside
{i, j, k}. But then we have
z }| { z }| {
(ijk) = (ij)(jk) = (ij)(ls) (ls)(jk) .
where the permutations under the braces are of (2, 2) type. Thus, (ijk) ∈ H, and we are
done. 
Lemma 3. If H is a normal subgroup of An where n ≥ 5, that contains either a 3-cycle
or a (2, 2) permutation. Then H = An .
Proof. Let α denote the 3-cycle contained in H ≤ An . It suffices by Lemma 1, to show
that any other 3-cycle of Sn is also contained in H. Recall that any 3-cycle is conjugate to
any other 3-cycle in Sn . Thus, any 3-cycle is conjugate to α in Sn . But in order to achieve
our result, we would want any 3-cycle is conjugate to α in An . For then H being normal
in An and since α ∈ H, we may deduce that any 3-cycle must be in. Therefore, An ≤ H.
The lemma will then follow. So, we need to prove that any two 3-cycles are conjugate of
one another in An . Let β = (ijk) and γ = (lst) be 3-cycles, where the distinct letters
may represent the same object. We have already seen that (ijk) ans (lst) are conjugate in
Sn . Thus there is a σ ∈ Sn such that (ijk) = σ(lst)σ −1 . We assert that there is a λ ∈ An
such that λ(lst)λ−1 = (ijk). If σ is even, i.e., σ ∈ An , then we are done. If σ is odd, then
we pick symbols a, b in {1, 2, · · · , n} − {ijk} (we can do this as n ≥ 5). Now, consider
λ = σ(ab) ∈ An . We thus have
λ(lst)λ−1 = λ(lst)λ−1 = (ab)σ(lst)σ −1 (ab) = (ab)(ijk)(ab) = (ab)(ab)(ijk) = (ijk),
proving our assertion. Now, suppose α is a (2, 2) permutation, say α = (ij)(kl). The
arguments to be given here is somewhat hand waving. However, if you look at the proof of
simplicity of A5 , then you will find out that the total number of conjugates of α is 15, both
 9

in S5 as well as in A5 . This means that every (2, 2) permutation, being a conjugate to α in

An must be in H as H C An .
Now getting back to the proof, if n ≥ 6, then the argument goes exactly as in the case
where α was a 3-cycle. So we only need to verify the lemma in the case n = 5. We argue
that any other (2, 2) permutation can be obtained from α by conjugating by a 3-cycle.
Since 3-cycles are in A5 , we will be done in this case. Let m be the 5th element so that
{i, j, k, l, m} = {1, 2, 3, 4, 5}. Typically, another (2, 2) permutation can be obtained from
α by exchanging elements between the disjoint sets

{i, j}, {k, l}, {m} .
It needs just a little imagination to see that any such exchanges can be made via conju-
gating with 3-cycles. For example, suppose we would like to get (il)(jk), then we would
conjugate α by (ijk). If we wanted to exchange j with m, i.e., try to get (im)(kl), we
would conjugate α by (ijm). Now, you should be convinced that this process can indeed
be carried out. If you have an alternative easier argument, do let me know. This finishes
the proof of the lemma. 

Now we have developed enough material to prove Theorem 3.

proof of Theorem 3. Let {1} = 6 H C Sn . It is enough to show that An ≤ H (why?).

The proof essentially uses a commutator element. Observe that it suffices to show that H
contains a 3-cycle or a permutation of the type (2, 2). This is because if H contains one
of these, say a 3-cycle τ , then H being normal in Sn must contain all the conjugates of τ .
Since any 3-cycle is conjugate of τ in Sn , it follows that H contains all 3-cycles. Now by
Lemma 1, we may deduce that An ≤ H. If τ is (2, 2) permutation, then an exact similar
argument will lead to An ≤ H again. Thus it remains to show that H contains a 3-cycle
or a permutation of the type (2, 2).
Since H 6= {1}, there is a nontrivial element α in H. Thus there is an i that is being
moved by α, i.e., α(i) 6= i. We pick a j ∈ {1, 2, · · · , n} such that j ∈/ {i, α(i)}. Now,
construct the transposition β = (ij). Next, consider the element
γ = αβα−1 β −1 .
Observe that γ = α(βα−1 β −1 ) is in H as H is normal in Sn . On the other hand,
γ = αβα−1 β −1 = α(ij)α−1 (ij)−1 = (α(i)α(j))(ij)
Clearly, α(i) 6= α(j), so that (α(i)α(j)) is a transposition. Next, observe that γ 6= 1 as
α(i) ∈ / {i, j}. Thus, γ is a product of two transpositions. Now there are two possibilities.
In the first scenario, α(j) ∈ {i, j}. In that case γ becomes a 3-cycle (why?). In the second
scenario, α(j) ∈ / {i, j}, in which case we find that {α(i), α(j)} ∩ {i, j} = ∅. Thus, in the
latter scenario, γ becomes a permutation of the type (2, 2). In either case, since γ ∈ H, we
have achieved our result. 

Lemma 4. A5 is simple.
Proof. Let us record that |A5 | = 60. Let H C A5 . Note that if H contains a 3-cycle, then
we are through by Lemma 3. So assume that H does not contain a 3-cycle. We assert
that 3 - |H|. Else, by Cauchy’s theorem H contains an element of order 3. But what
are elements of order 3 in S5 ? If α is an element of order 3 in S5 , then we consider the
decomposition of α in to disjoint cycles, say α = β1 β2 · · · βk . Note that 3 = o(α) =
o(β1 β2 · · · βk ) is the lcm of lengths of βi . Thus, each βi has to be a 3-cycles. Since βi
10

are mutually disjoint and they must be formed from {1, 2, 3, 4, 5}, deduce that there can at
most be one βi . Consequently, α is a 3-cycle. This proves our assertion.
Since |H| | 60 and 3 - |H|, the possible orders of |H| are

2, 4, 5, 10, 20.

Now, we eliminate possibilities one by one. Firstly |H| 6= 2, else H contains a transposi-
tion or a (2, 2) permutation. The former is impossible as H ≤ A5 can only contain even
permutations. In the latter case, H being normal in A5 , must contain all the conjugates of
the (2, 2) permutation in A5 . Below, we have shown that this number is 12. But |H| = 2.
So we are done in this case. In case |H| = 4, then H can either be cyclic of order 4, in
which case H contains a 4-cycle (why?). But a 4-cycle is an odd permutation and as such
cannot be in H ≤ A5 . Else, H contains two elements of order 2 and none of which is a
transposition. This means both are (2, 2) permutations. But then as noted before |H| ≥ 12,
an impossibility. If |H| = 5, 10 or 20, then by Cauchy’s theorem, H contains an element
of order 5. That is H contains a 5-cycle. Let us call it α. We show that this much infor-
mation is enough to deduce that |H| ≥ 13. This eliminates the possibilities that |H| = 5
and |H| = 10. Note that any conjugate of α in A5 must be in H (why?). We show that
there are 12 conjugates of α in A5 . This, along with the identity element yields |H| ≥ 13.
Consider the conjugacy class [α] in S5 . Since every 5-cycle is a conjugate of α in S5 , we
deduce that |[α]| = 4! = 24. Thus, the normalizer, NS5 (α) has order |Sn |/|[α]| = 5.
Therefore, NS5 (α) is cyclic of order 5. Next, we consider the conjugacy class [α]0 in A5
and the normalizer NA5 (α). Note that x ∈ NA5 (α) if and only if xαx−1 = α and x is an
odd permutation. Thus, this in particular, means that α ∈ NA5 (α) (Note that o(α) = 5 so
that, α ∈ A5 ). Since, NA5 (α) ≤ NS5 (α), and that NS5 (α) is cyclic of order 5, it follows
that NA5 (α) is cyclic of order 5. Thus |[α]0 | = |A5 |/|NA5 (α)| = 12. Therefore, α has
12 conjugates in A5 , as desired. Thus, we are left to deal with the case |H| = 20. There
is a very easy argument using Sylow’s theory, namely, that H has to have a subgroup of
order 4 but, we have seen that a subgroup of order four contains odd permutations which
is not permissible here. Let us prove this without the Sylow theory. Since 2 divides |H|,
it has an element β of order 2. Since β cannot be a transposition, it has to be a (2, 2) type
permutation (why?). We take β = (ij)(kl) where distinct letters represent distinct objects
in X5 = {1, 2, 3, 4, 5}. We take m denote the fifth element of X5 , i.e.,

{i, j, k, l, m} = {1, 2, 3, 4, 5}.

Instead of trying to find the conjugacy class of β in A5 , it is easier to compute NA
5 (β).

In
5 3
fact let us determine the normalizer NS5 (β) of β in S5 . Observe that there are 2 2 /2 =
15 distinct (2, 2) type permutations in S5 and each of them are conjugate to β in S5 . Hence,
|NS5 (β)| = 120/15 = 8. An element α ∈ NS5 (β) fixes beta by conjugation. Thus, we
should have

(ij)(kl) = α(ij)(kl)α−1 = α(ij)α−1 α(kl)α−1 = (α(i)α(j))(α(k)α(l)).

Since factorization into disjoint cycles is unique up to the order of appearance, we have the
following.

(α(i)α(j)) = (ij), (α(l)α(k)) = (lk) =⇒ α = 1, (ij), (lk) or (ij)(lk)

(α(i)α(j)) = (lk), (α(l)α(k)) = (ij) =⇒ α = (il)(jk), (ik)(jl).

 11

Thus far we have accounted for 6 elements of NS5 (β). The remaining two can be computed
by considering
(ij)(il)(jk) = (iljk) and (ij)(ik)(jl) = (ikjl).
We have thus accounted for all the eight elements of NS5 (β). In order to compute |NA5 (β)|,
we look at the number of even permutations in NS5 (β). This number is 4. Thus, the conju-
gacy class of β in A5 has 60/4 elements, and all these elements must be in H as H C A5 .
Thus, H has, in total, at least 1 + 12 + 15 = 28 elements. Clearly, |H| = 6 20, and we have
covered all the cases. 
In order to prove Theorem 2, we need one last lemma.
Lemma 5. Let n ≥ 5 and α 6= 1 in An . Then there is a conjugate of α in An α0 and a
i ∈ {1, 2, · · · , n} such that α(i) = α0 (i).
Proof. Since α 6= 1, it can be factored into product of disjoint cycles. Let r be the length
of the largest cycle appearing in the factorization of α, and let us denote this cycle by
(i1 i2 · · · ir ). We consider two case, r = 2 and r ≥ 3. If r = 2, then α is a product of
disjoint transpositions. In case α is a product of three or more disjoint transpositions, then
α looks like
α = (i1 i2 )(j1 j2 )(k1 k2 )(· · · ),
we take τ = (i1 i2 )(j1 k1 ) ∈ An . Thus, α0 = τ ατ −1 looks like
α0 = τ ατ −1 = (i1 i2 )(k1 j2 )(j1 k2 )(· · · ).
Clearly, α0 6= α (why?) but α(i1 ) = i2 = α(i1 ). So, i1 plays the role of i in the
lemma, and we are done here. Next consider the case where α0 is a product of exactly
two transpositions α = (ij)(kl). Since, n ≥ 5, we can find a m ∈ / {i, j, k, l} and take
τ = (ij)(km) ∈ An . Then α0 = τ ατ −1 = (ij)(ml) 6= α and α0 (i) = j = α(i). Now
consider the case where r ≥ 3. Thus α = (i1 i2 i3 · · · ). Since n ≥ 5, we can pick j and k
outside {i1 , i2 }. Now take τ = (i3 jk), so that
α0 = τ ατ −1 = (i3 jk)(i1 i2 i3 · · · )(i3 kj) = (i1 i2 j · · · ) 6= (i1 i2 i3 · · · ).
Moreover, α0 (i1 ) = i2 = α(i1 ). Thus we have dealt with all the cases and the lemma is
proved. 
We end this section with the proof of Theorem 2.
proof of Theorem 2. We assume n ≥ 6, and as our inductive hypothesis, we assume that
An−1 is simple. Let X(i) = {1, 2, · · · , n} − {i}. Note that the group H(i) of even
permutations on X(i) is isomorphic to An−1 . Since, An−1 is simple by hypothesis, we
deduce that H(i) is simple for all 1 ≤ i ≤ n. Let N C An and assume that N 6= {1}.
We would like to show that N = An . Observe that because of Lemma 3, it is enough
to show that N contains a 3-cycle. Also, observe that H(i) being isomorphic to An−1
contains a 3-cycle (since n − 1 ≥ 5). Therefore, it suffices to establish that N ⊇ H(i)
for some i ∈ {1, 2, · · · , n}. To do this, we pick an element α 6= 1 in N . By Lemma 5,
there is a conjugate α0 of α in An such that α0 6= α and there is an i ∈ {1, 2, · · · , n}
such that α0 (i) = α(i). Since N C An , we deduce that α0 ∈ N . Consequently, the
element β = α0 α−1 6= 1 is inside N and has the property that β(i) = i (why?). Therefore
β ∈ H(i), and hence, N ∩ H(i) 6= {1}. On the other hand, we note that N ∩ H(i) C H(i).
As H(i) is simple and N ∩ H(i) 6= {1}, we deduce that N ⊇ H(i). Thus, indeed N
contains a 3-cycle (as H(i) contains one). The proof of the theorem is thus complete.