Permutations and Orbits
Permutations and Orbits
In the earliest conceptions of group theory, all groups were considered permutation groups. Essentially
a group was a collection of ways in which one could rearrange some set or object: for example, each
‘rearrangement’ of an equilateral triangle corresponds to one of its symmetries. It was Arthur Cayley,
of Cayley-table fame, who formulated the group axioms we now use. Importantly, he also proved
what is now known as Cayley’s Theorem: that the old definition and the new are identical.
Theorem 5.2. If A is any set, then the set of permutations S A of A forms a group under composition.
Inverse If ϕ is a permutation then it is a bijection, whence the function ϕ−1 exists and is also a
bijection. X
Definition 5.3. The symmetric group on n-letters Sn is the group of permutations of any1 set A of n
elements. Typically we choose A = {1, 2, . . . , n}.
To describe a group as a permutation group simply means that each element of the group is being
viewed as a permutation of some set. As the following result shows, all groups are permutation
groups, although sadly not in a particularly useful way!
Proof. For each element a ∈ G, let ρ a : G → G be the function ρ a : g 7→ ag (i.e. left multiplication by
a). We make two claims:
1. Each ρ a is a permutation of G.
2. {ρ a : a ∈ G }, ◦ forms a group isomorphic to G.
1 S is the explicit group of permutations of {1, 2, . . . , n } or the abstract group of permutations of any set with n elements.
n
2 In contrast to C or Z where the subscript is the order of the group.
n n
1
The first claim is straightforward. ρ a−1 is the inverse function to ρ a :
∀ g ∈ G, (ρ a−1 ◦ ρ a )( g) = a−1 ag = g = idG ( g)
whence each ρ a is a bijection.
Now define a map φ : G → {ρ a } by φ( a) = ρ a . We claim this is an isomorphism:
1–1 φ( a) = φ(b) =⇒ ρ a = ρb =⇒ ∀ g ∈ G, ag = bg =⇒ a = b. X
Onto Certainly every permutation ρ a is in the image of φ. X
Homomorphism If a, b ∈ G, then
φ( a) ◦ φ(b) : g 7→ ρ a (ρb ( g)) = abg = ρ ab ( g)
from which φ( ab) = φ( a) ◦ φ(b).
Note that Cayley’s Theorem is not saying that every group is isomorphic to some Sn . It is saying that
every group G is isomorphic to some subgroup of SG .
4 2
We could then write
1 2 3 4
σ=
3 1 4 2
where you read down columns to find where σ maps an element in the top row. Composition is read
in the usual way for functions, do the right permutation first. Thus if
1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4
ρ= then σρ = = (∗)
1 4 3 2 3 1 4 2 1 4 3 2 3 2 4 1
Matrix Notation σ can also be viewed as acting on the vector (1, 2, 3, 4)T , which suggests a matrix
method for encoding elements of Sn . For example, our permutation σ may be written
0 0 1 0
1 0 0 0
σ= 0 0 0 1
0 1 0 0
as a permutation matrix: multiplying on the left by such a matrix permutes the entries of vectors.
Definition 5.6. A permutation matrix is an n × n matrix with exactly one entry of 1 in each row and
column and the remaining entries 0.
Indeed we may conclude:
Theorem 5.7. The set of n × n permutation matrices forms a group under multiplication which is isomorphic
to Sn . By Cayley’s Theorem, every finite group of permutations is isomorphic to a group of matrices.
2
Cycle notation Our example permutation can be more compactly written as σ = (1 3 4 2). We read
from left to right, looping back to 1 at the end, each entry telling us where the previous is mapped to.
Thus (1 3 4 2) maps
1 7→ 3 7→ 4 7→ 2 7→ 1
We have shorter cycles if some of the elements are fixed; for example in our two notations
1 2 3 4
(1 3) = ∈ S4
3 2 1 4
Juxtaposition is used for composition:
(1 3 4 2)(2 4) = (1 3 4)
(compare with (∗) above).
Remember that multiplication of cycles is really composition of functions: although each cycle is read
from left to right when determining how it acts on elements of {1, 2, . . . , n}, we multiply cycles by
considering the rightmost cycle first. For example, in S5 ,
1 7→ 3
2 7 → 3 7 → 5
(1 3 5 4)(2 3 4) : 3 7→ 4 7→ 1 =⇒ (1 3 5 4)(2 3 4) = (1 3)(2 5 4) (†)
4 7→ 2
5 7→ 4
This notation can be used to describe non-symmetric groups: e.g. if we label the corners of a rhombus,
the Klein 4-group V can be written in terms of cycles as a subgroup of S4 .
2
4
Definition 5.8. Suppose that k ≤ n. A k-cycle in Sn is an element ( a1 a2 · · · ak ).
Cycles ( a1 · · · ak ) and (b1 · · · bl ) are disjoint if no element appears in both cycles: that is, if
{ a1 , . . . , ak } ∩ {b1 , . . . , bl } = ∅
The identity element is the only 0-cycle. It is sometimes written (), if not otherwise denoted by e.
As the example (†) illustrates, when computing the product of several cycles, the result will typically
be a product of disjoint cycles. This will prove very useful in the next section when we discuss orbits.
3
5.2 Dihedral groups
Definition 5.9. The dihedral group Dn is the group of symmetries of the regular n-gon (polygon with
n sides).
Group? Now that we have defined permutation groups it is very easy to see that the dihedral
groups are indeed groups. Observe that a symmetry of an n-gon can be viewed as a permutation of
the corners of the n-gon for which ‘neighborliness’ is preserved. This is similar to how we viewed
the Klein 4-group above.
For example, if we label the corners of the regular hexagon 1 through 6 then 3 2
we see that D6 is the set of σ ∈ S6 such that σ(1) is always next to σ (6) and
σ(2), etc.
Clearly this says that D6 ⊆ S6 . 4 1
To see that D6 is a subgroup of S6 , we need only note that the composition
of two neighbor-preserving transforms must also preserves neighbors, as
does the inverse of such a map. 5 6
Elements of Dn The regular n-gon has 2n distinct symmetries and so | Dn | = 2n. These consist of:
2πj
n rotations For each j = 0, . . . , n − 1, let ρ j be rotation counter-clockwise by n radians.
n reflections Let µ j be reflection across the line making angle with the positive x-axis (make sure
πj
n
you put one of the corners of the n-gon on the x-axis!).3
Remarks Some authors write D2n instead of Dn precisely because | Dn | = 2n: in this course, Dn will
always mean the symmetries of the n-gon.
Every dihedral group Dn is a subgroup of the orthogonal group O2 (R). The correspondence is:
2πj
− sin 2πj 2πj 2πj
cos n cos sin
ρj = n µ j =
n n
2πj 2πj 2πj
− cos 2πj
sin n cos n sin n n
It is a good exercise to convince yourself that these matrices really do correspond to the rotations and
reflections claimed. In particular multiply any two of them together and see what you get. . .
4
Explicit descriptions of D3 and D4
2
D3 is the group of symmetries of an equilateral triangle. If we label the
corners as in the picture, we can easily define the elements of the group.
ρ0 the identity
2π 1
ρ1 rotate counter-clockwise by 3 radians
2π
ρ2 rotate clockwise by 3 radians
µi reflect in the altitude through i
3
Here we write all the elements in permutation notation ad give the Cayley table.
µ2 µ2 µ3 µ1 ρ2 ρ0 ρ1 1 2 3
µ2 (13)
µ3 µ3 µ1 µ2 ρ1 ρ2 ρ0 3 2 1
1 2 3
µ3 (12)
2 1 3
It should be immediately obvious that all the permutations of {1, 2, 3} are elements of D3 , and so
D3 ∼= S3 . Now we consider all the subgroups of D3 and its subgroup diagram.
How can we be certain that there are no other subgroups of D3 ? A careful consideration of generating
sets should convince you. For example, suppose that a subgroup contains two reflections: WLOG
suppose these are µ1 , µ2 . We compute the subgroup generated by {µ1 , µ2 }. It must include
µ1 µ2 = ρ1 , ρ21 = ρ2 µ1 ρ1 = µ3
5
D4 is the group of symmetries of the square. It consists of four rotations
2
and four reflections: the notation δj for reflection across a diagonal is
used here, rather than labelling all reflections µ j .
ρ0 the identity
ρ1 rotate counter-clockwise by π
2 radians 3 1
ρ2 rotate counter-clockwise by π radians
3π
ρ3 rotate counter-clockwise by 2 radians
µi reflect across midpoints of sides
δi reflect across diagonals 4
Element Standard notation Cycle notation
1 2 3 4
◦ ρ0 ρ1 ρ2 ρ3 µ1 µ2 δ1 δ2 ρ0 e = ()
1 2 3 4
ρ0 ρ0 ρ1 ρ2 ρ3 µ1 µ2 δ1 δ2 1 2 3 4
(1234)
Rotations
ρ1
ρ1 ρ1 ρ2 ρ3 ρ0 δ2 δ1 µ1 µ2 2 3 4 1
1 2 3 4
ρ2 (13)(24)
3 4 1 2
ρ2 ρ2 ρ3 ρ0 ρ1 µ2 µ1 δ2 δ1
1 2 3 4
ρ3 ρ3 ρ0 ρ1 ρ2 δ1 δ2 µ2 µ1 ρ3 (1432)
4 1 2 3
µ1 µ1 δ1 µ2 δ2 ρ0 ρ2 ρ1 ρ3 1 2 3 4
µ1 (12)(34)
µ2 µ2 δ2 µ1 δ1 ρ2 ρ0 ρ3 ρ1 2 1 4 3
1 2 3 4
Reflections
µ2 (14)(23)
δ1 δ1 µ2 δ2 µ1 ρ3 ρ1 ρ0 ρ2 4 3 2 1
1 2 3 4
δ2 δ2 µ1 δ1 µ2 ρ1 ρ3 ρ2 ρ0 δ1 (24)
1 4 3 2
1 2 3 4
δ2 (13)
3 2 1 4
All the subgroups are summarised in the following table. In particular, note that D4 S4 : the latter
has many more elements!
D4
Subgroup Isomorph Generating sets
{ ρ0 } C1 { ρ0 }
{ ρ0 , µ i } C2 {µi } for each i
{ρ0 , δi } C2 {δi } for each i V C4 V
{ ρ0 , ρ2 } C2 { ρ2 }
{ ρ0 , ρ1 , ρ2 , ρ3 } C4 {ρ1 } or {ρ3 }
{ ρ0 , µ1 , µ2 , ρ2 } V {µ1 , µ2 }, {µ1 , ρ2 } or {µ2 , ρ2 }
{ρ0 , δ1 , δ2 , ρ2 } V {δ1 , δ2 }, {δ1 , ρ2 } or {δ2 , ρ2 } C2 C2 C2 C2 C2
any pair {ρi , µ j } or {ρi , δj }
D4 D4 where i = 1, 3 and j = 1, 2 or
any pair {µk , δl } where k, l = 1, 2
C1
In the subgroup diagram, the middle C2 is {ρ0 , ρ2 } while the two copies on each side contain either
the reflections δi or µi .
6
5.3 Orbits
In this section we continue the idea of a group being a set of permutations. In particular, we will see
how any element σ ∈ Sn partitions the set {1, 2, . . . , n}. This concept will be generalized later when
we consider group actions.
Examples If σ ∈ Sn is written in cycle notation (recall Definition 5.8) using disjoint cycles, then the
cycles are the orbits! For example, in S5 ,
The same does not hold if the cycles are not disjoint. For example, σ = (13)(234) ∈ S4 maps
1 7→ 3 7→ 4 7→ 2 7→ 1
so there is only one orbit: orb j (σ) = {1, 2, 3, 4} for any j. In fact, σ = (1234), from which the orbit is
obvious.
Given that disjoint cycle notation is so useful for reading orbits, it is a natural question to ask if any
permutation σ can be written as a product of disjoint cycles. The answer, of course, is yes, with the
disjoint cycles turning out to be precisely the orbits of σ!
x ∼ y ⇐⇒ y ∈ orbx (σ)
Reflexivity x ∼ x since x = σ0 ( x ). X
Transitivity Suppose that x ∼ y and y ∼ z. Then y = σk ( x ) and z = σl (y) for some k, l ∈ Z. But
then z = σk+l ( x ) and so x ∼ z. X
The equivalence classes of ∼ are clearly the orbits of σ, which therefore partition X.
7
Proof. Write out each of the orbits of σ ∈ Sn in order, placing each orbit in parentheses ( ). Since the
orbits of σ partition X = {1, 2 . . . , n} the cycles obtained are disjoint.
More concretely,
orbx1 (σ) = { x1 , σ( x1 ), σ2 ( x1 ), . . .}
This orbit must be disjoint with orb1 (σ). Now repeat. It is immediate from the construction that
σ = 1 σ (1) σ 2 (1) · · · x1 σ ( x1 ) σ 2 ( x1 ) · · · · · · · · ·
Examples If you mechanically follow the algorithm for multiplying cycles (see the previous section)
you will automatically end up with a product of disjoint cycles:
1. (13)(234)(1432) = (123)
2. (13)(24)(12)(34) = (14)(23)
Now that we are able to write any permutation as a product of disjoint cycles, we are able to compute
much more easily. For example:
Theorem 5.13. The order of a permutation σ is the least common multiple of the lengths of its disjoint cycles.
Proof. Write σ as a product of disjoint cycles σ = σ1 · · · σm . Since disjoint cycles commute, it it imme-
diate that
σn = σ1n · · · σmn
Moreover, since the terms σjn permute disjoint sets, it follows that
σn = e ⇐⇒ ∀ j, σjn = e
A k-cycle clearly has order k (the least positive integer l such that ( a1 · · · ak )l = e). If the orbits of σ
have lengths α j ∈ N respectively, it follows that
σjn = e ⇐⇒ α j | n
Thus n must be a multiple of α j for all j. The least such n is clearly lcm{α j }.
8
Example The order of σ = (145)(3627)(89) ∈ S9 is lcm(3, 4, 2) = 12.
We can easily calculate σ3465 for the above σ. Since 3465 = 12 · 288 + 9 we have
5.4 Transpositions
Instead of breaking a permutation σ into disjoint cycles, we can consider a permutation as being
constructed from only the simplest bijections.
Definition 5.14. A 2-cycle ( a1 a2 ) is also known as a transposition, since it swaps two elements of
{1, 2, . . . , n} and leaves the rest untouched.
Theorem 5.15. Every σ ∈ Sn is the product of transpositions.
Proof. There are many, many ways to write out a single permutation as a product of transpositions.
One method is to first write σ as a product of disjoint cycles, then write each cycle as follows:
( a1 · · · ak ) = ( a1 ak )( a1 ak−1 ) · · · ( a1 a2 )
Theorem 5.17. The concepts of even/odd are well-defined: every permutation is either even or odd, and not
both.
Proof. Recall that any permutation σ ∈ Sn can be written as an n × n permutation matrix (Definition
5.6). A 2-cycle is a permutation matrix which swaps two rows: it therefore differs from the n × n
identity matrix only in that two of its rows are swapped. For example
1 0 0 0
0 0 0 1
(24) =
0
∈ S4
0 1 0
0 1 0 0
From linear algebra we have that swapping two rows of a matrix changes the sign of its determinant.
Hence det(2-cycle) = −1. It follows that
(
1 if σ is the product of an even number of 2-cycles,
det(σ) =
−1 if σ is the product of an odd number of 2-cycles.
9
5.5 The Alternating Groups
The concept of even permutations leads us to the definition of a new collection of groups.
Definition 5.18. The alternating group An (n ≥ 2) is the group of even permutations in Sn .
n!
Theorem 5.19. An has exactly half the elements of Sn : that is | An | = 2.
Proof. Since n ≥ 2, we have (12) ∈ Sn . Define φ : An → {odd permutations} by φ(σ) = (12)σ. We
claim that this is a bijection
1–1 φ(σ) = φ(τ ) =⇒ (12)σ = (12)τ =⇒ στ. X
Onto If ρ is an odd permutation, then (12)ρ is even and so in An . Therefore ρ = φ((12)ρ). X
Since φ is a bijection, it follows that there are exactly the same number of even and odd permutations
in Sn . Exactly half of them are therefore even.
10