5.

Permutation groups
Definition 5.1. Let S be a set. A permutation of S is simply a
bijection f : S −→ S.
Lemma 5.2. Let S be a set.
(1) Let f and g be two permutations of S. Then the composition of
f and g is a permutation of S.
(2) Let f be a permutation of S. Then the inverse of f is a permu-
tation of S.
Proof. Well-known. 
Lemma 5.3. Let S be a set. The set of all permutations, under the
operation of composition of permutations, forms a group A(S).
Proof. (5.2) implies that the set of permutations is closed under com-
position of functions. We check the three axioms for a group.
We already proved that composition of functions is associative.
Let i : S −→ S be the identity function from S to S. Let f be a
permutation of S. Clearly f ◦ i = i ◦ f = f . Thus i acts as an identity.
Let f be a permutation of S. Then the inverse g of f is a permutation
of S by (5.2) and f ◦ g = g ◦ f = i, by definition.
Thus inverses exist and G is a group. 
Lemma 5.4. Let S be a finite set with n elements.
Then A(S) has n! elements.
Proof. Well-known. 
Definition 5.5. The group Sn is the set of permutations of the first n
natural numbers.
We want a convenient way to represent an element of Sn . The first
way, is to write an element σ of Sn as a matrix.
 
1 2 3 4 5
∈ S5 .
3 1 5 4 2
Thus, for example, σ(3) = 5. With this notation it is easy to write
down products and inverses. For example suppose that
   
1 2 3 4 5 1 2 3 4 5
σ= τ= .
3 1 5 4 2 4 3 1 2 5
Then  
1 2 3 4 5
τσ = .
1 4 5 2 3
1
On the other hand
 
1 2 3 4 5
στ = .
4 5 3 1 2
In particular S5 is not abelian.
The problem with this way of representing elements of Sn is that we
don’t see much of the structure of τ this way. For example, it is very
hard to figure out the order of τ from this representation.
Definition 5.6. Let τ be an element of Sn .
We say that τ is a k-cycle if there are integers a1 , a2 , . . . , ak such
that τ (a1 ) = a2 , τ (a2 ) = a3 , . . . , and τ (ak ) = a1 and τ fixes every
other integer.
More compactly

ai+1 i < k

τ (ai ) = a1 i=k

a
i otherwise.
For example  
1 2 3 4
2 3 4 1
is a 4-cycle in S4 and  
1 2 3 4 5
.
1 5 3 2 4
is a 3-cycle in S5 . Now given a k-cycle τ , there is an obvious way to
represent it, which is much more compact than the first notation.
τ = (a1 , a2 , a3 , . . . , ak ).
Thus the two examples above become,
(1, 2, 3, 4)
and
(2, 5, 4).
Note that there is some redundancy. For example, obviously
(2, 5, 4) = (5, 4, 2) = (4, 2, 5).
Note that a k-cycle has order k.
Definition-Lemma 5.7. Let σ be any element of Sn .
Then σ may be expressed as a product of disjoint cycles. This fac-
torisation is unique, ignoring 1-cycles, up to order. The cycle type
of σ is the lengths of the corresponding cycles.
2
Proof. We first prove the existence of such a decomposition. Let a1 = 1
and define ak recursively by the formula
ai+1 = σ(ai ).
Consider the set
{ ai | i ∈ N }.
As there are only finitely many integers between 1 and n, we must
have some repetitions, so that ai = aj , for some i < j. Pick the
smallest i and j for which this happens. Suppose that i 6= 1. Then
σ(ai−1 ) = ai = σ(aj−1 ). As σ is injective, ai−1 = aj−1 . But this
contradicts our choice of i and j. Let τ be the k-cycle (a1 , a2 , . . . , aj ).
Then ρ = στ −1 fixes each element of the set
{ ai | i ≤ j }.
Thus by an obvious induction, we may assume that ρ is a product
of k − 1 disjoint cycles τ1 , τ2 , . . . , τk−1 which fix this set.
But then
σ = ρτ = τ1 τ2 . . . τk ,
where τ = τk .
Now we prove uniqueness. Suppose that σ = σ1 σ2 . . . σk and σ =
τ1 τ2 . . . τl are two factorisations of σ into disjoint cycles. Suppose that
σ1 (i) = j. Then for some p, τp (i) 6= i. By disjointness, in fact τp (i) = j.
Now consider σ1 (j). By the same reasoning, τp (j) = σ1 (j). Continuing
in this way, we get σ1 = τp . But then just cancel these terms from both
sides and continue by induction. 
Example 5.8. Let  
1 2 3 4 5
σ= .
3 4 1 5 2
Look at 1. 1 is sent to 3. But 3 is sent back to 1. Thus part of the
cycle decomposition is given by the transposition (1, 3). Now look at
what is left {2, 4, 5}. Look at 2. Then 2 is sent to 4. Now 4 is sent to
5. Finally 5 is sent to 2. So another part of the cycle type is given by
the 3-cycle (2, 4, 5).
I claim then that
σ = (1, 3)(2, 4, 5) = (2, 4, 5)(1, 3).
This is easy to check. The cycle type is (2, 3).
As promised, it is easy to compute the order of a permutation, given
its cycle type.
Lemma 5.9. Let σ ∈ Sn be a permutation, with cycle type (k1 , k2 , . . . , kl ).
The the order of σ is the least common multiple of k1 , k2 , . . . , kl .
3
Proof. Let k be the order of σ and let σ = τ1 τ2 . . . τl be the decompo-
sition of σ into disjoint cycles of lengths k1 , k2 , . . . , kl .
Pick any integer h. As τ1 , τ2 , . . . , τl are disjoint, it follows that
σ h = τ1h τ2h . . . τlh .
Moreover the RHS is equal to the identity, iff each individual term is
equal to the identity.
It follows that
τik = e.
In particular ki divides k. Thus the least common multiple, m of
k1 , k2 , . . . , kl divides k. But σ m = τ1m τ2m τ3m . . . τlm = e. Thus m divides
k and so k = m. 
Note that (5.7) implies that the cycles generate Sn . It is a natural
question to ask if there is a smaller subset which generates Sn . In fact
the 2-cycles generate.
Lemma 5.10. The transpositions generate Sn .
Proof. It suffices to prove that every permutation is a product of trans-
positions. We give two proofs of this fact.
Here is the first proof. As every permutation σ is a product of
cycles, it suffices to check that every cycle is a product of transpositions.
Consider the k-cycle σ = (a1 , a2 , . . . , ak ). I claim that this is equal to
σ = (a1 , ak )(a1 , ak−1 )(a1 , ak−2 ) · · · (a1 , a2 ).
It suffices to check that they have the same effect on every integer j
between 1 and n. Now if j is not equal to any of the ai , there is nothing
to check as both sides fix j. Suppose that j = ai . Then σ(j) = ai+1 .
On the other hand the transposition (a1 , ai ) sends j to a1 , the ones
befores this do nothing to j, and the next transposition then sends a1
to ai+1 . No other of the remaining transpositions have any effect on
ai+1 . Therefore the RHS also sends j = ai to ai+1 . As both sides have
the same effect on j, they are equal. This completes the first proof.
To see how the second proof goes, think of a permutation as just be-
ing a rearrangement of the n numbers (like a deck of cards). If we can
find a product of transpositions, that sends this rearrangement back to
the trivial one, then we have shown that the inverse of the correspond-
ing permutation is a product of transpositions. Since a transposition
is its own inverse, it follows that the original permutation is a product
of transpositions (in fact the same product, but in the opposite order).
In other words if
τk · · · · · τ3 · τ2 · τ1 · σ = e,
4
then multiplying on the right by τi , in the opposite order, we get
σ = τ1 · τ2 · τ3 · · · · · τk .
The idea is to put back the cards into the right position, one at
a time. Suppose that the first i − 1 cards are in the right position.
Suppose that the ith card is in position j. As the first i − 1 cards are in
the right position, j ≥ i. We may assume that j > i, otherwise there is
nothing to do. Now look at the transposition (i, j). This puts the ith
card into the right position. Thus we are done by induction on i.