UNIVERZA NA PRIMORSKEM

Fakulteta za matematiko, naravoslovje in informacijske tehnologije

Algebraic Graph Theory

doc. dr. Ademir Hujdurović

DRUGO UČNO GRADIVO

41 strani

Matematika, dodiplomski študijski program

PRVA IZDAJA

Koper, 2018
Contents
1 Preliminaries 4
1.1 Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Cyclic and dihedral groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Direct and semidirect product of groups . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Symmetric group Sn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Alternating group An . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.6 Group actions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.7 Orbits and stabilizers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.8 Transitive, semiregular and regular group actions . . . . . . . . . . . . . . . . . 15
1.9 Counting orbits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.10 Cycle index of permutation group . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.11 Polya enumeration theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2 Graphs 19
2.1 Number of non-isomorphic graphs . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.2 Number of non-equivalent (proper) colourings . . . . . . . . . . . . . . . . . . . 21

3 Vertex-transitive graphs 23
3.1 Cayley graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2 Normal Cayley graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3 Graph products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.4 Hamiltonicity of vertex-transitive graphs . . . . . . . . . . . . . . . . . . . . . . 30

4 Edge-transitive and arc-transitive graphs 33

5 s-arc-transitive graphs 35
5.1 Cubic-arc-transitive graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

6 Distance-transitive graphs 37
6.1 Eigenvalues of strongly regular graphs . . . . . . . . . . . . . . . . . . . . . . . 38

2
Introduction
This is set of lecture notes on undergraduate course ”Algebraic Graph Theory” at Faculty
of Mathematics, Natural Sciences and Information Technologies of University of Primorska,
Slovenia. As this is an undergraduate subject, the material gives only introduction to Alge-
braic Graph Theory. The material presented in these lecture notes is collected from different
sources, books, lecture notes etc.

3
1 Preliminaries
1.1 Groups
In this section we review the background on group theory.
Definition 1.1. A group is a set G together with a binary operation ∗ such that:
(i) for every a, b ∈ G, a ∗ b ∈ G;

(ii) (∀a, b, c ∈ G) a ∗ (b ∗ c) = (a ∗ b) ∗ c, (∗ is associative,);

(iii) there exists an identity element e ∈ G, such that x ∗ e = e ∗ x = x, ∀x ∈ G;

(iv) given identity element e ∈ G, for every element x ∈ G, there exists its inverse x−1 ∈ G,
such that x ∗ x−1 = x−1 ∗ x = e, for all x ∈ G.
If a subset H ⊆ G is also a group with the same binary operation ∗ then we say that H is a
subgroup of G, and we write H ≤ G.
Identity element in G will sometimes be denoted with 1G or simply by 1. To simplify the
notation, we usually write gh instead of g ∗ h. We refer to the number |G| of elements in G
as the order of the group G.
If x∗y = y ∗x then we say that x and y commute. If every pair of elements in G commutes,
then G is said to be abelian.
Exercise 1.2. Let G be a group and H ⊆ G. Prove that H ≤ G if and only if the following
three conditions hold:
(i) 1G ∈ H;

(ii) ∀h1 , h2 ∈ H, h1 h2 ∈ H;

(iii) ∀h ∈ H, h−1 ∈ H.
Exercise 1.3. If H and K are subgroups of G, then also H ∩ K is a subgroup of G.
Definition 1.4. If H ≤ G, then the right coset of H in G is the set of the form Hg = {hg |
h ∈ H}. Similarly, left coset of H in G is the set of the form gH = {gh | h ∈ H}.
Exercise 1.5. If Hg1 and Hg2 are cosets, then either Hg1 = Hg2 or Hg1 ∩ Hg2 = ∅.
Exercise 1.5 implies that G is a disjoint union of right cosets of H.
Definition 1.6. If H ≤ G, then the index [G : H] of H in G is the number of cosets of H in
G.
It is now easy to see that for a finite group G, |G| = [G : H]|H|, and in particular the
order of H divides the order of G. This is known as Lagrange’s theorem.
Note that the possible converse of the Lagrange’s whould be the following: If n is a divisor
of |G|, then there exists H ≤ G such that |H| = n. We will see latter that this does not hold.
If g is an element of a group G, the order of g is the least integer n ≥ 1 such that g n = e,
if such n exists, otherwise we say that g is of infinite order. In finite group all elements have
finite orders.

4
 If G is any group and S is a subset of G, then with hSi we denote the smallest subgroup
of G that contains S, and call it the subgroup generated by S. If S = {g1 , g2 , . . . , gn } then
instead of h{g1 , g2 , . . . , gn i we simply write hg1 , g2 , . . . , gn i.
If a group G contains an element g such that G = hgi, then we say that G is a cyclic
group.

Exercise 1.7. Let p be a prime, and let G be a group of order p. Prove that G is cyclic
group.

Definition 1.8. Let G be a group and let N ≤ G. If g −1 N g = N for every g ∈ G, then N is

said to be a normal subgroup of G, and we denote this with N
G.

Exercise 1.9. Prove that the following claims are equivalent:

(i) N is a normal subgroup of G;

(ii) gN = N g for every g ∈ G;

(iii) the set of all left cosets of N in G coincides with the set of all right cosets of N in G.

Exercise 1.10. Prove that in an abelian group every subgroup is normal.

Exercise 1.11. Let G be a group and H an index 2 subgroup of G. Prove that H is a normal
subgroup of G.

Exercise 1.12. Let G be a group and N1 and N2 be two normal subgroups of G. Prove that
N1 ∩ N2 is a normal subgroup of G.

Let G be a group and H, K ≤ G. Let HK = {hk | h ∈ H, k ∈ K}. Set HK is not in

general a subgroup of G.

Exercise 1.13. If H and K are subgroups of G then we have that

|H||K|
|HK| = .
|H ∩ K|

Exercise 1.14. Let G be a group and H, K ≤ G. If one of H or K is normal in G, then HK

is a subgroup of G.

Definition 1.15. If G is a group and N
G, then the factor group G/N is the group with
elements {gN : g ∈ G} and the group operation (g1 N ) ∗ (g2 N ) = (g1 g2 )N .

If G is a finite group, then |G/N | = |G|/|N |.

Exercise 1.16. Prove that the factor group defined in Definition 1.15 is indeed a group.

Definition 1.17. If G and H are groups, a homomorphism ϕ : G → H is a map ϕ from G

to H such that ∀g1 , g2 ∈ G:

ϕ(g1 ∗G g2 ) = ϕ(g1 ) ∗H ϕ(g2 ), (1)

where ∗G is the group operation in G and ∗H is the group operation in H.

5
Exercise 1.18. Let ϕ : G → H be a homomorphism and let K ≤ G. Prove that ϕ(K) ≤ H.

Definition 1.19. Let ϕ : G → H be a homomorphism and let eH denote the identity in H.

Kernel Ker(ϕ) of ϕ is defined with Ker(ϕ) = {g | g ∈ G, ϕ(g) = eH }.

Exercise 1.20. Let ϕ : G → H be a homomorphism. Prove that Ker(ϕ)
G.

Definition 1.21. A bijective homomorphism ϕ : G → H is called isomorphism. If there

exists an isomorphism between G and H we say that G and H are isomorphic and write
G∼= H.

1.2 Cyclic and dihedral groups

Let n be a positive integer. With Cn we will denote a cyclic group of order n, that is
Cn = hai, and an = 1. It can be shown that any two cyclic groups of order n are isomorphic.
The following construction of cyclic group of order n will be useful later on.
Let Zn = {0, 1, . . . , n − 1} be the set of residues modulo n. Then it is not difficult to see
that Zn is a group with operation +n , that is addition modulo n. Identity element in this
group is 0, and x−1 = −x (mod n). This group is cyclic and of order n.
For a positive integer n, dihedral group D2n is generated by two elements ρ and τ , such
that ρn = 1 and τ 2 = 1. The remaining elements in D2n are given with

D2n = {1, ρ, . . . , ρn−1 , τ, ρτ, . . . , ρn−1 τ }.

The operation in D2n is given with

ρi ∗ ρj = ρi+j ;
ρi ∗ ρj τ = ρi+j τ ;
ρi τ ∗ ρj = ρi−j τ ;
ρi τ ∗ ρj τ = ρi−j .

Dihedral group D2n is isomorphic to the group of all symmetries of a regular n-gone. One
step rotation of the n-gon corresponds to the element ρ, and τ corresponds to a reflection of
the n-gon.

1.3 Direct and semidirect product of groups

In this section we briefly present methods of constructing new groups from old ones. The first
and the simplest method is the direct product of groups.

Definition 1.22. Given two group G1 and G2 , the direct product G1 × G2 of G1 and G2 is
the group with elements {(g1 , g2 ) : g1 ∈ G1 , g2 ∈ G2 } and the group operation is given with
(x1 , x2 ) ∗ (y1 , y2 ) = (x1 y1 , x2 y2 ).

Exercise 1.23. Prove that G1 × G2 as defined above is indeed a group.

Theorem 1.24. Let G be a group and let H and K be a subgroups of G. Then G is

isomorphic to the direct product of H and K if and only if the following three conditions
hold:

6
 (i) H and K are normal in G;

(ii) G = HK;

(iii) H ∩ K = {1G }.
Exercise 1.25. Determine which of the following groups are isomorphic: Z2 ×Z3 , Z6 , Z4 ×Z2 ,
Z8 .
A generalization of the notion of direct product is the notion of semidirect product.
Definition 1.26. Let G be a group and let N and K be a subgroups of G. Then G is
semidirect product of N and K if and only if the following three conditions hold:
(i) N is a normal subgroup of G;

(ii) G = N K;

(iii) N ∩ K = {1G }.
We let N o K denote the semidirect product of N by K.
It is clear that every direct product is also a semidirect product, but semidirect product
does not need to be a direct product. To get a better understanding of the structure of a
semidirect product we first need to introduce some more terminology.
Definition 1.27. If G is a group then an automorphism of G is an isomorphism from G to
G. The set of all automorphisms of a group G is denoted with Aut(G).
Exercise 1.28. The set Aut(G) together with the operation of composition of functions is a
group.
Definition 1.29. Two elements x and y in a group G are said to be conjugate if there exists
an element g ∈ G such that gxg −1 = y. We often write gxg −1 = xg .
We can extend the notion of conjugate elements to the notion of conjugate subgroups. We
say that H and K are conjugate subgroups of G if there exists g ∈ G such that gHg −1 = K.
Again we will usually write H g = gHg −1 . It is not difficult to see that the relation ”is
conjugate” is an equivalence relation in G. The equivalence classes of this relation are called
conjugacy classes.
Exercise 1.30. If G is a group and g ∈ G, then the mapping ϕ : G → G defined with
ϕ(x) = gxg −1 is an automorphism of G.
The automorphisms given in previous exercise are called inner automorphisms.
Exercise 1.31. The set Inn(G) of all inner automorphisms of G is a normal subgroup of
Aut(G).
Exercise 1.32. Prove that N is normal subgroup of G if and only if ϕ(N ) = N for every
ϕ ∈ Inn(G).
Definition 1.33. A subgroup N of a group G is called characteristic if ϕ(N ) = N , for every
ϕ ∈ Aut(G). If N is characteristic subgroup of G we write N char G.

7
Exercise 1.34. If N char G then N
G.

An important property of characteristic subgroup is the following:

Exercise 1.35. If G is a group and C and N are subgroups of G with C ≤ N ≤ G such that
C char N and N
G then C
G.

We now go back to the semidirect product of two groups.

Theorem 1.36. Let N and H be a groups and ϕ : H → Aut(N ) be a homomorphism. Let

G be the set N × H with operation

(n1 , h1 ) ∗ (n2 , h2 ) = (n1 · ϕ(h−1

1 )(n2 ), h1 h2 ).

Then G is isomorphic to a semidirect product of N by H.

Proof. We identify N with the set {(n, 1H ) | n ∈ N } and H with the set {(1N , h) | h ∈ H}.
It is now clear that N ∩ H = {1}. We leave as an exercise to the reader to verify that
the operation ∗ in G is associative, that (1N , 1H ) is the identity element in G, and that
(n, h)−1 = (ϕ(h)(n−1 ), h−1 ).
To see that N is normal in G we observe that

(n, h)−1 (n0 , 1H )(n, h) = (ϕ(h)(n−1 ), h−1 )(n0 , 1H )(n, h)

= (ϕ(h)(n−1 )ϕ(h)(n0 )), h−1 )(n, h)
= (ϕ(h)(n−1 )ϕ(h)(n0 )ϕ(h)(n), 1)
= (ϕ(h)(n−1 n0 n), 1) ∈ N

It remains to prove that G = N H. Let g ∈ G be arbitrary. Then by the definition of G,

it follows that g = (n, h) for some n ∈ N and h ∈ H. It is now easy to see that g = nh, where
n = (n, 1) and h = (1, h).

We are now going to show how the dihedral group D2n can be constructed as the semidirect
product of groups Cn and C2 . Let N = hρ | ρn = 1i and H = hτ | τ 2 = 1i. In order to
construct a semidirect product of groups N and H, we need a homomorphism ϕ : H →
Aut(N ). Let ϕ be defined with ϕ(1H ) = idN and ϕ(τ ) = ιN , where idN stands for identity
map in N , that is idN (x) = x (∀x ∈ N ) and ιN stands for the inversion map in N , that is
ιN (x) = x−1 (∀x ∈ N ).

Exercise 1.37. Prove that ϕ defined as above is indeed a homomorphism that maps H to
Aut(N ).

Now using Theorem 1.36 we construct the group N o H ∼ = Cn o C2 . We have that the
elements in the group N o H are given with N o H = {(ρi , τ k ) : i ∈ {0, . . . , n − 1}, k ∈ 0, 1}.
The group operation is given with

(ρi , 1H ) ∗ (ρj , 1H ) = (ρi · ϕ(1H )(ρj ), 1H ) = (ρi+j , 1H )

(ρi , 1H ) ∗ (ρj , τ ) = (ρi · ϕ(1H )(ρj ), τ ) = (ρi+j , τ )
(ρi , τ ) ∗ (ρj , 1H ) = (ρi · ϕ(τ )(ρj ), 1H ) = (ρi−j , τ )
(ρi , τ ) ∗ (ρj , τ ) = (ρi · ϕ(τ )(ρj ), 1H ) = (ρi−j , 1H )

8
Exercise 1.38. Prove that the mapping φ : N o H → D2n defined with

φ((ρi , 1H )) = ρi ;
φ((ρi , τ )) = ρi τ ;

is isomorphism.

1.4 Symmetric group Sn

Let X be a nonempty set. A bijective mapping g from X to X is also called a permutation
of X. Let Sym(X) denote the set of all permutations of the set X. If X has size n, then
|Sym(X)| = n!.
Exercise 1.39. Let ◦ denote the symbol for composition of functions, that is

∀g, h ∈ Sym(X) (g ◦ h)(x) = g(h(x)),

Prove that Sym(X) together with operation ◦ forms a group.

The definition of the operation ◦ in Sym(X) means that for g1 , g2 ∈ Sym(X) the product
g1 g2 is carried out by first applying g2 to x ∈ X , and then g1 to the image g2 (x), that is we
are multiplying permutations from right to left.
We denote the identity element of Sym(X) with idX . Let x ∈ X and g ∈ Sym(X). We
say that x is a fix element of g, or x is fixed by g if g(x) = x. Otherwise, we say x is moved
by g.
Definition 1.40. A permutation group is any subgroup G of Sym(X) for some nonempty
set X.
We say G is a finite permutation group when X is a finite set. In this note every set is
meant to be finite, unless it is said otherwise.
Notation. We denote by Sn the permutation group Sym({1, . . . , n}). We write idn (or id
when n is understood from the context) for id{1,...,n} .
There are two common ways in which permutation groups can be written. First is the
two line notation for permutations, that was introduced by Cauchy. A permutation g of a set
X = {x1 , x2 , . . . , xn } is written in the following way:
 
x1 x2 x3 ··· xn−1 xn
g=
g(x1 ) g(x2 ) g(x3 ) · · · g(xn−1 ) g(xn )

The other notation is to write g ∈ Sym(X) as a product of disjoint cycles.

Definition 1.41. A permutation g ∈ Sym(X) is called a k-cycle if there exists distinct
elements x1 , . . . , xk in X such that g(xk ) = x1 , and g(xi ) = xi+1 for all i ∈ {1, . . . , k − 1},
and g(x) = x for all x ∈ X \ {x1 , . . . , xk }.

Notation. We denote by (x1 x2 . . . xk ) the k-cycle g in Definition 1.41.

A 2-cycle is also called a transposition. Two cycles g1 = (x1 . . . xk ) and g2 = (x01 . . . x0l )
are disjoint if {x1 , . . . , xk } ∩ {x01 , . . . , x0l } = ∅. Note that, this implies that g1 g2 = g2 g1 .

9
Example 1.42. Let X = {0, 1, 2, 3, 4, 5, 6} and let g ∈ Sym(X) be given with g(x) = 4x + 1
(mod 7). Then the permutation g can be written as
 
0 1 2 3 4 5 6
g=
1 5 2 6 3 0 4

or as a product of disjoint cycles g = (0 1 5)(2)(6 4 3) = (6 4 3)(0 1 5) = . . . = (0 1 5)(3 6 4).

Observe that the order in which the disjoint cycles are written is not important.

Theorem 1.43. (Cyclic Decomposition) Every permutation g ∈ Sym(X) can be written

as the product of disjoint cycles. Factorization is unique up to the order of factors.

Corollary 1.44. Every permutation can be written as the product of transpositions.

Proof. It is clear that

(a1 a2 . . . ar ) = (a1 ar )(a1 ar−1 ) . . . (a1 a2 ).

By Theorem 1.43 it follows that every permutation can be written as a product of disjoint
cycles. On the other hand, every cycle can be written as the product of transposition, hence
the claim holds.

Exercise 1.45. Let g, h ∈ Sym(X), such that g has a cyclic decomposition

g = (a1 . . . ar ) . . . (as . . . an ).

Then hgh−1 has a cylic decomposition

hgh−1 = (h(a1 ) . . . h(ar )) · · · (h(as ) . . . h(an )).

We call a cyclic decomposition of g to be maximal if it includes all possible 1-cycles. A

permutation g ∈ Sym(X) has cyclic structure [c1 , . . . , cn ] if a maximal cyclic decomposition
of g consists of ck number of k-cycles, where k ∈ {1, . . . , n}. Note that, the equality c1 + 2c2 +
· · · + ncn = n always holds. The following corollary of Theorem 1.45 is useful when studying
the conjugacy of two permutations.

Corollary 1.46. The permutations g1 and g2 are conjugate in Sym(X) if and only if g1 and
g2 have the same cyclic structure.

Exercise 1.47. Let g1 , g2 ∈ S12 be given with their cyclic decompositions:

g1 = (1 3 5)(2 4 6)(7)(8 12)(9 10 11)

g2 = (1)(3 2 6)(7 4 5)(9 8 11)(10 12)

Find permutation h ∈ S12 such that hg1 h−1 = g2 .

10
1.5 Alternating group An
In this part we construct a subgroup of Sn of index 2. We saw that any permutation can
be written as the product of transpositions. However, the number of transpositions is not
uniquely determine. On the other hand, we are now going to prove that the parity of the
number of transpositions for a given permutation is fixed.

Theorem 1.48. If a permutation g ∈ Sn can be written as a product of r and s transpositions,

then r ≡ s (mod 2).

Definition 1.49. A permutation g ∈ Sn is called even (odd) if g is the product of even

(odd) number of transpositions.

Exercise 1.50. Prove that a cycle of length k is even permutation if and only if k is odd.

Notation. We denote by Alt(X) the subgroup of even permuations in Sym(X). In the case
when |X| = n, then we denote by An the set all even permuations in Sn .
We show below that An , n ≥ 2 is in fact a subgroup of Sn of index 2 . The group An is
called the alternating group of degree n.

Theorem 1.51. The set An , n ≥ 2 of all even permutations is a subgroup of Sn of index 2.

Proof. It is clear that id ∈ An .

Let g, g 0 ∈ An . The group Sn is generated by the set of all transpositions. Thus g = t1 · · · tr
and g 0 = t01 · · · t0s , where r and s are even. Then gg 0 = t1 · · · tr t01 · · · t0s , r + s is even, and so
gg 0 ∈ An . Also, g −1 = tr · · · t1 , and so g −1 ∈ An . We obtain that An is indeed a subgroup of
Sn . Then (1, 2) ∈ / An . It is not difficult to see that the mapping g 7→ (1, 2)g, where g ∈ An ,
is a bijective mapping from An to Sn \ An . Thus |An | = n!/2.

Note that, since An has index 2 in Sn , An / Sn .

Exercise 1.52. Let n ≥ 2, and let H = {id, (1 2)} be a cyclic subgroup of Sn of order 2.
Prove that Sn ∼
= An o H.

Exercise 1.53. Let g ∈ Sn be a permutation and let [c1 , c2 , . . . , cn ] be the cyclic structure of
g. Prove that g is odd permutation if and only if
n/2
X
c2i ≡ 1 (mod 2).
i=1

Exercise 1.54. Prove that the group An (n ≥ 3) is generated by the set of all cycles of length
3.

Exercise 1.55. Prove that A4 has no subgroup of order 6. (Hint: Use Exercises 1.11 and
1.13).

11
Proof. Suppose that A4 has a subgroup N of order 6. Then N is an index 2-subgroup of A4
and by Exercise 1.11 N is normal in A4 . Let V = {id, (1 2)(3 4), (1 3)(2 4), (1 4)(2 3)}. Then
V is a subgroup of A4 of order 4. Since 4 does not divide 6 and V is not subgroup of N , and
so N is a proper subgroup of V N . This implies that N V = A4 . Now, using Exercise 1.13 we
obtain
|N ||V | 6·4
|N ∩ V | = = = 2.
|N V | 12
Since both N and V are normal in A4 , it follows that N ∩ V is also a normal subgroup of A4 .
However, the only subgroups of A4 of order 2 are of the form {id, (a b)(c d)}. Let g = (1 2 3).
Then (1 2)(3 4)g = (2 3)(1 4), (1 3)(2 4)g = (2 1)(3 4) and (1 4)(2 3)g = (2 4)(3 1), hence no
subgroup of order 2 in A4 is normal.

1.6 Group actions

In this part we introduce the concept of an action of a group. The motivation behind this
concept is to study groups by representing them as permutation groups.

Definition 1.56. An action of a group G on a set X is a mapping f : G × X → X such

that, writing the image f ((x, g)) as g · x, the following properties hold.

A1. 1G · x = x for every x ∈ X.

A2. (g1 g2 ) · x = g1 · (g2 · x) for every x ∈ X and for every g1 , g2 ∈ G.

Given an action of a group G on a set X, we will also say that G acts on X. The cardinality
|X| is called the degree of the action. Instead of g · x we will sometimes write gx or g(x).

Example 1.57. Let G = R \ {0} be the group of nonzero real numbers under the multipli-
cation, and let X be the set of all vectors in R3 . Then G acts on X via scalar multiplication,
that is g · (x1 , x2 , x3 ) = (gx1 , gx2 , gx3 ), for (x1 , x2 , x3 ) ∈ R3 and g ∈ R \ {0}.

Example 1.58. Let G ≤ Sym(X). Then G acts naturally on X with g · x = g(x), for x ∈ X
and g ∈ G.

Example 1.59. Let G be the group acting on a set X. Then G acts on the set X × X with
g · (x1 , x2 ) = (g · x1 , g · x2 ).

Theorem 1.60. Let G act on X and fix g ∈ G.

(1) The mapping πg : X → X, πg : x 7→ g(x) is in Sym(X).

(2) The mapping ρ : G → Sym(X), ρ : g 7→ πg is a homomorphism.

(3) Let Φ : G → Sym(X) be a homomorphism. Then this homomorphism defines action of a

group G on the set X with g · x := Φ(g)(x).

Definition 1.61. For an action of a group G on a set X we call the homomorphism ρ in

Theorem 1.60 the corresponding permutation representation of G on X. The degree of
an action (or a premutation representation) is the size of X.

12
Definition 1.62. By the kernel and image of an action of G on a set X we mean the kernel
and image, respectively, of the corresponding permutation representation ρ. We say that the
action is faithful if its kernel is trivial.

We illustrate the above definitions through the following example.

Example 1.63. Let G = D12 be the group of symmetries of a regular hexagon acting on its
points {1, 2, 3, 4, 5, 6}. Let d1 = 14, d2 = 25 and d3 = 36 be the three main diagonals of the
hexagon (order in which the points are written is not important, for example d1 = 14 = 41).
Let X = {d1 , d2 , d3 }. Prove that G acts on the set X via g · x1 x2 = g(x1 )g(x2 ). Determine
the image and the kernel of this action. Is this action faithful?

Now we jump to a more theoretical example.

Theorem 1.64. Let G be a group, and let us take X to be G. Then G acts on X with

g · x = gx, x, g ∈ G.

This action is faithful.

Proof. We show first that we indeed have an action, i.e., the mapping defined above satisfies
Axioms A1 and A2.
A1: For every x ∈ X, 1G · x = 1G x = x.
A2: For every x ∈ X and for every g1 , g2 ∈ G, (g1 g2 ) · x = g1 g2 x = (g1 )(g2 x) = g1 · (g2 · x).
Let g ∈ G be in the kernel of the action. This means g · x = x for all x ∈ X. But this just
means gx = x for all x ∈ X, hence g = 1G , and the kernel is the trivial group. By definition,
the action is faithful.

The permutation representation corresponding to the action of G in the previous theorem

is called the left regular representation of G. We illustrate the usage of the left regular
representation in proving that every group is isomorphic to a permutation group.

Theorem 1.65 (Cayley). Every group is isomorphic to a permutation group.

Proof. Let G be a group. Let G act on X = G via left multiplication. By Theorem 1.60,
it follows that ρ : G → Sym(X) is a group homomorphism (recall that ρ(g) = πg , and
πg (x) = gx for x ∈ G). Then the image of G under this homomorphism ρ(G) is a subgroup
of Sym(X). By the first isomorphism theorem, it follows that G/Ker(ρ) ∼
= ρ(G). Since by
Theorem 1.65 it follows that Ker(ρ) = {1G } it follows that G ∼
= ρ(G).

1.7 Orbits and stabilizers

In this part we introduce the two most basic concepts in connection with an action – the
orbits and the stabilizers.

Definition 1.66. Let G act on X, and let x ∈ X. The orbit of G induced by x is the set
{g · x | g ∈ G}.

13
Notation. We denote by OrbG (x) the orbit of G induced by x.

Definition 1.67. Let G act on X, and let x ∈ X. The stabilizer of x in G is the set
{g ∈ G | g · x = x}.

Notation. We denote by Gx the stabilizer of x in G.

Example 1.68. Let X = Z10 = {0, 1, . . . , 9}, and let G = Aut(Z10 ) = Z∗10 acting naturally
on X. Calculate the stabilizers G0 , G1 , and orbits 0G , 1G , 2G .

We have the following properties of orbits and stabilizers.

Theorem 1.69. Let G act on X and let x ∈ X.

(1) The set of all orbits OrbG (x) form a partition of X.

(2) The stabilizer Gx is a subgroup of G.

Proposition 1.70. Let G act on a set X, let g ∈ G, x ∈ X and let y = g · x. Then

Gy = gGx g −1 .

Proof.

Gy = {h ∈ G | h · y = y}
= {h ∈ G | h · (g · x) = g · x}
= {h ∈ G | (g −1 hg) · x = x}
= {h ∈ G | g −1 hg ∈ Gx }
= {h ∈ G | h ∈ gGx g −1 }
= gGx g −1 .

Theorem 1.71 (Orbit-stabilizer). Let G be a group acting on a set X. Then for every x ∈ X
it holds |G| = |OrbG (x)| · |Gx |.

Proof. Let x ∈ X be fixed. Define the set Ω = {(g, y) ∈ G × OrbG (x) | gx = y}. We calculate
the cardinality of Ω with “double counting”. First, we obtain that
X  X X
|Ω| =  {(g, y) ∈ G × OrbG (x) | y = gx}  = |{y ∈ OrbG (x) | y = gx}| = 1 = |G|.
g∈G g∈G g∈G

Second, we obtain that

X   X
|Ω| =  {(g, y) ∈ G × OrbG (x) | y = gx}  = |{g ∈ G | y = gx}|.
y∈OrbG (x) y∈OrbG (x)

Let y ∈ OrbG (x) be fixed. Then there exists h ∈ G such that y = hx. Now we have

{g ∈ G | y = gx} = {g ∈ G | hx = gx} = {g ∈ G | h−1 gx = x} = {g ∈ G | h−1 g ∈ Gx } = h(Gx ).

14
Since |hGx | = |Gx |, it follows that
X
|Ω| = |Gx | = |OrbG (x)| · |Gx |.
y∈OrbG (x)

We conclude that |Ω| equals |G| = |OrbG (x)| · |Gx |, as required.

We are now going to present two more important types of group actions. First is the
so-called conjugation action of G on G defined with

xg = gxg −1 , x, g ∈ G.

Proposition 1.72. The conjugation action of a group G on G is well-defined group action.

For g ∈ G, the orbit ORbG (g) is the conjugacy class of g in G. The stabilizer Gg is the
centralizer CG (g) of g in G. The kernel of this action is the center Z(G) of G.
Exercise 1.73. Does group G act on G with

xg = g −1 xg, x, g ∈ G?

Next important example of group action is the so called coset action. Let G be a group
and H ≤ G. The coset action is the action of G on the set of left H-cosets in G defined
with
g · (xH) = gxH, x, g ∈ G.
Proposition 1.74. The coset action is indeed an action. The stabilizer of xH is GxH =
xHx−1 . The kernel is ∩x∈G xHx−1 .
The kernel of the coset action is also called the core of H in G (notation: coreG (H)).
Exercise 1.75. Prove that coreG (H) is the largest normal subgroup of G which is contained
in H.
Exercise 1.76. Let G be a p-group, that is a group of order pk , where p is a prime. Using
the conjugation action, prove that Z(G) the center of G, is non-trivial.

1.8 Transitive, semiregular and regular group actions

Definition 1.77. Let G act on X. The action is transitive if it has only one orbits, and it
is semiregular if |Gx | = 1 for all x ∈ X. An action which is both transitive and semiregular
is called regular.
Example 1.78. Left regular representation of G, that is the action of G on G given with
g · x = gx is regular action.
Proposition 1.79. Let G act transitively on X. Then cardinality |X| is a divisor of |G|.
Proposition 1.80. Let G act transitively on X. Then G is regular if and only if |G| = |X|.
Proposition 1.81. Let G be an abelian group which is transitive and faithful on the set X.
Then G is regular on X.

15
Proof. Since G is transitive, it remains to prove that G is semiregular. Suppose that G is not
semiregular, that is, there exists x ∈ X, such that Gx 6= {1}. Since G acts transitively on X,
then for any x ∈ X we have OrbG (x) = X. Using Proposition 1.70 it follows that any two
stabilizers are conjugate. However, since G is abelian, it follows that all stabilizers are the
same. Then the kernel of this action is equal to Gx , which contradicts the assumption that
G is faithful.
Exercise 1.82. Let p be a prime, and let G ≤ Sp be transitive. Prove that there exists a
cyclic subgroup H of G, such that H is regular.
Exercise 1.83. Let G be a group. Prove that the direct product G × G acts on G with
(g1 , g2 ) · x = g1 xg2−1 . Prove that this action is transitive, and calculate the vertex stabilizer
(G × G)1 . When is this action faithful?
Exercise 1.84. Let G act transitively on X and let H ≤ G. Prove that G = HGx if and
only if H is transitive.

1.9 Counting orbits

Definition 1.85. Let G be a group acting on a set X, and let g ∈ G. Then with f ixX (g) we
denote the set of elements of X fixed by g, that is f ixX (g) = {x ∈ X | g · x = x}. When the
set X is clear from the context we simply write f ix(g).
There is a simple relationship between the number of orbits of a finite group acting on
a finite set and the number of fixed points of its elements. A wide range of applications
in counting problems and combinatorics is based on elaborations of this relationship. The
theorem itself has a long history and is often referred to (inaccurately) as the ”Burnside
Lemma”; the simplest version is the following result.
Theorem 1.86 (Cauchy-Frobenius or Burnside). Let G be a finite group acting on a finite
set X, and let m be the number of orbits in this action. Then
1 X
m= |f ix(g)|.
|G|
g∈G

Proof. Define the set Ω = {(x, g) ∈ X × G | gx = x}. We are going to count the number of
elements of Ω in two different ways. First suppose that the orbits of G are X1 , . . . , Xm . Then
we have
m
X
|Ω| = |{(x, g) ∈ Xi × G | gx = x}|. (2)
i=1
Suppose that x ∈ Xi is arbitrary. Then {(x, g) ∈ Xi × G | gx = x} = {xi } × Gx . Recall that
by orbit-stabilizer property, we have |Gx ||Xi | = |G|. Hence we have
m X m X m
X X |G| X
|Ω| = |Gx | = = |G| = m|G|. (3)
|Xi |
i=1 x∈Xi i=1 x∈Xi i=1

On the other hand we have

X X
|Ω| = |{(x, g) ∈ X × G | gx = x}| = |f ix(g)|. (4)
g∈G g∈G

Combining (3) and (4) the result follows.

16
Corollary 1.87. If G is a finite transitive permutation group of degree n > 1 then G contains
an element with no fixed points.

Proof. Since G acts transitively, we have P that the number of orbits is m = 1. Then by
Cauchy-Frobenius theorem, we have |G| = g∈G |f ix(g)|. Since for 1G we have f ix(1G ) = X
and |X| ≥ 2, then it follows that there exists g ∈ G, such that f ix(g) = ∅.

1.10 Cycle index of permutation group

Definition 1.88. Let G ≤ Sym(X), where X is a finite set of size n, and let g ∈ G. Let
[c1 (g), . . . , cn (g)] be the cyclic structure of permutation g, that is, ck (g) denotes the number
of cycles of length k in the cyclic decomposition of g. The cycle index ZG of permutation
group G is the polynomial
1 X c1 (g) c2 (g)
ZG = x1 · x2 · . . . · xcnn (g)
|G|
g∈G

Exercise 1.89. Calculate ZS3 .

Exercise 1.90. Calculate ZS4 .

1 P n/d
Exercise 1.91. Prove that ZCn = n d|n ϕ(d)xd .
(1 (n−1)/2
1 2x1 x2 , n odd,
Exercise 1.92. Prove that ZDn = 2 ZCn + 1 2 (n−2)/2 n/2

4 x1 x2 + x2 , n even.

Exercise 1.93. Prove that

n
X 1 Y
ZSn = Qn ck c !
xckk ,
k=1 k k
(c) k=1

where the summation is over all integer partitions (c) of n.

1.11 Polya enumeration theorem

The result we are going to present now, and is usually referred to as Polya enumerations
theorem, was first published by John Howard Redfield in 1927. In 1937 it was independently
rediscovered by George Polya, who then greatly popularized the result by applying it to many
counting problems, in particular to the enumeration of chemical compounds.
Let X be a finite set and let G ≤ Sym(X) be a group of permutations of X. The set
X may represent a finite set of beads, and G may be a chosen group of permutations of the
beads. For example, if X is a necklace of n beads in a circle, rotations and reflections are
relevant so G is the dihedral group D2n of order 2n.

Definition 1.94. Let C be a finite set of colors. Let C X be the set of functions X → C, that
is C X is the set of all colorings of the set X with colors from C.

17
 For f ∈ C X , and for g ∈ G, we define an induced action of G on the set C X , with
g · f = f ◦ g −1 . Two functions f1 , f2 ∈ C X are called G-equivalent, if they belong to the
same orbit under the induced action of G on C X , or more precisely, if there exists g ∈ G such
that f1 (g(x)) = f2 (x), for every x ∈ X. Polya’s enumeration theorem given below counts the
number of orbits in this action, or equivalently the number of non-equivalent colorings of the
set X with colors from C.

Theorem 1.95 (Polya enumeration theorem). Let X and C be finite sets, and let G ≤
Sym(X). The number of orbits in the induced action of G on the set C X is equal to ZG with
x1 = . . . = xi = . . . x|X| = |C|.

Proof. By Cauchy-Frobenius theorem, it follows that the number of orbits is

1 X
|f ixC X (g)|.
|G|
g∈G

It remains to determine the size of f ixC X (g) = {f : X → C | f ◦ g −1 = f }. Let f ∈ f ixC X (g).

Suppose that (x1 x2 . . . xk ) is one cycle in the cyclic decomposition of g. Since f ◦ g = f , it
follows that f (x2 ) = (f ◦ g −1 )(x2 ) = f (g −1 (x2 )) = f (x1 ). Similarly, we conclude that f (x1 ) =
f (x2 ) = . . . f (xk ). Therefore, all vertices in the same orbit of g have the same image under
f . Since the number of possible images under f is |C|, it follows that |f ixC X (g)| = |C|c(g) ,
where c(g) is the number of cycles in the cyclic decomposition of g.

18
2 Graphs
A graph Γ consist of vertex set V (Γ) and edge set E(Γ), where an edge is an unordered pair
of distinct vertices {u, v}. We will sometimes write uv instead of {u, v}. If {u, v} is an edge,
we say that u and v are adjacent and that v is a neighbour of u, and we will denote this by
u ∼ v. A set Γ(v) denotes the set of all neighbours of v, that is Γ(v) = {u ∈ V (Γ) : u ∼ v}.
Degree of a vertex v ∈ V (Γ) is defined as the number of its neighbours, and is denoted by
dΓ (v), that is dΓ (v) = |Γ(v)|. If the graph Γ is clear from context, then we simply write d(v).
Order of a graph is the cardinality of its vertex-set.
Two graphs X1 = (V1 , E1 ) and X2 = (V2 , E2 ) are equal if V1 = V2 and E1 = E2 . Although
this is a reasonable definition, for most purposes, the model of relationship is not changed if
we rename the vertices of a graph. This naturally leads to the definition of isomorphic graphs.

Definition 2.1. For graphs X and Y , a function ϕ : V (X) → V (Y ) is called isomorphism

from X to Y , if

1. ϕ is bijection;

2. ∀u, v ∈ V (X) {u, v} ∈ E(X) ⇔ {ϕ(u), ϕ(v)} ∈ E(Y ).

Graphs X and Y are called isomorphic if there exist an isomorphism from X to Y , and we
denote this by X ∼
=Y.
For a function ϕ : A → B and for a subset S ⊆ A we define ϕ(S) = {ϕ(s) : s ∈ S}.

Lemma 2.2. Let ϕ be an isomorphism from X to Y . Then for every v ∈ V (X), we have
ϕ(X(v)) = Y (ϕ(v)).

Corollary 2.3. Isomorphisms preserve degrees of vertices, that is for an isomorphism ϕ from
X to Y we have dX (v) = dY (ϕ(v)), ∀v ∈ V .

One can also consider isomorphism from a graph Γ to itself. In this case, instead of ”iso-
morphism” we use word ”automorphism”. We repeat the definition to stress its importance.

Definition 2.4. Let Γ be a graph. An automorphism of Γ is a bijective function ϕ : V (X) →

V (X) such that ∀u, v ∈ V (X) we have {u, v} ∈ E(Γ) ⇔ {ϕ(u), ϕ(v)} ∈ E(Γ). The set of all
automorphisms of a graph Γ is denoted by Aut(Γ).

Exercise 2.5. Determine all the automorphisms of the following graph.

1 2

3 5 6
4

7 8 9

Theorem 2.6. For a graph Γ, the set Aut(Γ) is group under the operation of composition
of functions, that is for f, g ∈ Aut(Γ), f g = f ◦ g, or equivalently (f g)(v) = f (g(v)) for every
v ∈ V (Γ).

19
Definition 2.7. The complement Γ of a graph Γ is the graph with V (Γ) = V (Γ) and for
every two distinct vertices u, v ∈ V (Γ) we have {u, v} ∈ E(Γ) ⇔ {u, v} 6∈ E(Γ).

Lemma 2.8. For every graph Γ we have Aut(Γ) = Aut(Γ).

A graph Γ is called asymmetric if Aut(Γ) = {id}.

Exercise 2.9. For every positive integer n ≥ 6 there exists an asymmetric graph with n
vertices.

2.1 Number of non-isomorphic graphs

In this section we will show how Polya’s enumeration theorem can be used for determining
the number of non-isomorphic graph of given order. We will first demonstrate this on graphs
of order 4.

Example 2.10. Prove that there are 11 non-isomorphic graphs with 4 vertices.

Proof. Let G(4) be the set of all graphs with vertex set V = {1, 2, 3, 4}. Let X = {{i, j} |
1 ≤ i < j ≤ 4}. We can think of X as the edge set of complete graph K4 . Let C = {w, b}.
Consider the set C X , that is set of all functions f : X → C. For any such function f , one can
define a graph Γf , with E(Γf ) = {x ∈ X : f (x) = b}, or in other words f , is colors edges of
K4 with two colors (w for white, and b for black), the graph Γf is the subgraph with all black
edges. Conversely, for any graph Γ with vertex set V , one can define a function fΓ : X → C,
where fΓ ({u, v}) = 1 if and only if {u, v} ∈ E(Γ).
For a permutation g ∈ S4 , with g (2) we denote the permutation of X induced by g, and
(2)
with Sn the corresponding permutation group of X. For example, if g = (1 2)(3 4) then

g (2) = ({1, 2})({3, 4})({1, 3} {2, 4})({1, 4} {2, 3}).

It is now easy to see that two graphs with vertex set V are isomorphic if and only if the
(2)
corresponding collorings are equivalent under the action of S4 .
By Polya’s theorem, the number of non-equivalent collorings is ZS (2) (x1 = x2 = x3 = x4 =
4
(2)
2). It is left as exercise to calculate the cyclic index of S4 .

Exercise 2.11. Let g ∈ Sn be a permutation with ck cycles of length k. The number of

cycles in the cyclic decomposition of g (2) is
n
X X X kck (ck − 1) X
γ(g) = kc2k+1 + kc2k + + (r, t)cr ct .
2 r<t
k=1 k k

Exercise 2.12. Let V = {1, . . . , n}, let X = {{i, j} | 1 ≤ i < j ≤ n}. For g ∈ Sn let
g (2) denote the induced permutation of X and let γ(g) denote the number of cycles in cyclic
decomposition of g (2) . Prove that the number of non-isomorphic graphs of order n is
1 X γ(g)
2 .
n!
g∈Sn

20
Definition 2.13. Let V be a finite set, and let P(V ) denote the power set of V . Hypergraph
H = (V, E) is given with the vertex set V and the hyperedge set E ⊆ P(V ) \ {∅}. Two
hypergraphs H1 = (V1 , E1 ) and H2 = (V2 , E2 ) are said to be isomorphic if there exists bijective
function f : V1 → V2 such that f (E1 ) = E2 .
Exercise 2.14. Determine the number of non-isomorphic hypergraphs with n vertices, for
n ∈ {3, 4}.

2.2 Number of non-equivalent (proper) colourings

Suppose that we are given a graph Γ and a set of colors C, and we want to color the vertices
of Γ with colors from C. For two such colourings f1 and f2 we will say that are equivalent, if
there exist an automorphism ϕ ∈ Aut(Γ) such that f2 = f1 ◦ϕ. In other words, two colourings
are equivalent, if one can be obtained from the other after applying an automorphism of the
graph Γ. Note that here we are not talking about proper colourings, that is adjacent vertices
can be coloured in the same way.
Theorem 2.15. The number of non-equivalent colourings of a graph Γ of order n with m
colors is given by ZAut(G) (x1 = . . . = xn = m)
Example 2.16. Determine the number of non-equivalent colourings of C8 (cycle of length 8)
with 3 colors.
Definition 2.17. A proper colouring of a graph Γ with a set of colors S is a function f :
V (Γ) → C such that adjacent vertices have different colors, that is ∀u, v ∈ V (Γ), u ∼ v ⇒
f (u) 6= f (v). A chromatic polynomial χ(Γ, k) of a graph Γ counts the number of its proper
vertex-colourings with k colors, that is χ(Γ, k) is the number of proper colourings of Γ with k
colors.
Example 2.18. Let Pn be a path on n vertices. Then χ(Pn , k) = k(k − 1)n−1 .
Example 2.19. Let Cn be a cycle of length n (C1 is a loop with single vertex, and C2 ∼
= K2 .
Then χ(Cn , k) = (k − 1)n + (−1)n (k − 1).
Proof. By deletion-contraction formula, it follows that χ(Cn , k) = χ(Cn − e, k) − χ(Cn /e, k),
where Cn /e is the graph obtained by contracting edge e of Cn . Observe that Cn − e ∼ = Pn
∼
and Cn /e = Cn−1 . Now use induction, and previous exercise.

Definition 2.20. Let Γ be a graph, and let P be a partition of V (Γ). We defined a quotient
graph of Γ with respect to P to be the graph with vertex set P and with two elements
P1 , P2 ∈ P being adjacent if and only if there exist u ∈ P1 and v ∈ P2 such that {u, v} ∈ E(Γ).
In this definition we allow loops as well, that is P1 = P2 is allowed. The obtained
graph is denoted by ΓP . For an automorphism g ∈ Aut(Γ), let Pg be the set of orbits of hgi.
The corresponding quotient graph is denoted by Γg .
Remark 2.21. If a graph Γ contains a loop, then there is no proper colouring of Γ.
Theorem 2.22. The number of non-equivalent proper colourings of a graph Γ with k colors
is equal to
1 X
χ(Γg , k).
| Aut(Γ)|
g∈Aut(Γ)

21
 In the case when Γ ∼ = Cn , then we have the following result about the number of non-
equivalent proper colourings.

Theorem 2.23. Let Cn be the cycle of length n. The number of non-equivalent proper
colourings of Cn with k colors is equal to
1 X
ϕ(d)χ(C nd , k) if n is odd ,
2n
d|n
 
1  X n
ϕ(d)χ(C nd , k) + k(k − 1)n/2  if n is even
2n 2
d|n

Exercise 2.24. Let X be the graph shown below.

• Find all automorphisms of X.

• Determine the number of non-equivalent proper colouring of X with k colors.

1 2 3 4 5 6

Exercise 2.25. Let X be the graph shown below.

7 8 9

1 2 3 4 5 6

• Find all automorphisms of X.

• Determine the number of non-equivalent proper colouring of X with k colors.

22
3 Vertex-transitive graphs
Definition 3.1. A graph Γ is said to be vertex-transitive if Aut(Γ) acts transitively on V (Γ),
that is ∀u, v ∈ V (Γ), there exists ϕ ∈ Aut(Γ) such that ϕ(u) = v.

Example 3.2. The complete graph Kn of order n, is vertex-transitive, for every n ≥ 1.

Example 3.3. The graph Cn , cycle of length n is vertex-transitive graph, for every n ≥ 3.

We will now present some elementary properties of vertex-transitive graphs. Recall that a
graph Γ is said to be d-regular if all the vertices of Γ have degree equal to d. If Γ is d-regular
for some d, then we say that Γ is regular, and d is called the valency of Γ.

Proposition 3.4. A graph Γ is vertex-transitive if and only if its complement Γ is vertex-

transitive.

Proof. The result follows from the fact that Aut(Γ) = Aut(Γ).

Proposition 3.5. Let Γ be a vertex-transitive graph. Then Γ is regular.

Proof. Let u, v ∈ V (Γ). Since Γ is vertex-transitive, it follows that there exists ϕ ∈ Aut(Γ)
such that ϕ(u) = v. Then by Corollary 2.3 it follows that dΓ (u) = dΓ (v), hence the result
follows.

We saw that every vertex-transitive graph is regular. The converse is not true, a graph
which is regular does not need to be vertex-transitive.

Example 3.6. The graph shown on Figure 3.6 is not vertex-transitive. In fact it is the
smallest (with respect to the number of vertices) connected regular graph which is not vertex-
transitive. To see that this graph is not vertex-transitive, observe that its complement is
disjoint union of C4 and C3 .

Figure 1: Example of a regular graph which is not vertex-transitive

3.1 Cayley graphs

Definition 3.7. Let G be a finite group, S ⊂ G such that S does not contain 1G , and
S −1 = S (here S −1 = {x−1 : x ∈ S}. The Cayley graph of G with respect to S denoted
by Cay(G, S) is the graph with vertex-set G, and two vertices x, y ∈ G are adjacent if and
only if x−1 y ∈ S. In other words E(Cay(G, S)) = {{x, xs} : x ∈ G, s ∈ S}. Set S is called
the connections set of Cay(G, S).

23
Example 3.8. Let n ≥ 3, and let G = Zn be the cyclic group of order n. Let S = {1, n − 1}.
Then Cay(G, S) ∼
= Cn .
Theorem 3.9. Every Cayley graph is vertex-transitive.
Proof. Let Γ = Cay(G, S). For g ∈ G, let gL : G → G be the mapping defined by gL (x) = gx.
It is not difficult to see that for every g ∈ G, the mapping gL is an automorphism of Cay(G, S)
(verify this!). Now for any x, y ∈ G, take g = yx−1 . Observe that gL (x) = y. Hence Cay(G, S)
is vertex-transitive graph.

In fact we proved something stronger in previous theorem. For a group G, let GL =

{gL : g ∈ G}. GL is called left regular representation of G. Observe that GL is a group,
GL ≤ Sym(G), and GL ∼ = G. In previous theorem we proved that GL ≤ Aut(Cay(G, S)).
Exercise 3.10. Prove that for every Cayley graph Cay(G, S), the group GL is a regular
subgroup of Aut(Cay(G, S)).
Exercise 3.11. Prove that valency of Cay(G, S) is equal to |S|.
Exercise 3.12. Let X = Cay(Z10 , {±2, 5}) and Y = Cay(D10 , {ρ, ρ4 , τ }), where D10 is
dihedral group of order 10, that is, D10 = hρ, τ | ρ5 = τ 2 = 1, ρτ = τ ρ−1 i. Represent
graphically graphs X and Y and check if X and Y are isomorphic.
Lemma 3.13. Cayley graph Cay(G, S) is connected if and only if hSi = G (recall that hSi
denotes the group generated by S).
Proof. Suppose that hSi = G. Let x, y ∈ G, and g = x−1 y. Then there exists s1 , . . . , sk ∈ S
such that s1 · s2 . . . · sk = g. For 1 ≤ i ≤ k, let vi = x · s1 · . . . si . Observe that vk =
x · s1 . . . sk = xg = x(x−1 y) = y. This constructs a path from x to y, (x, v1 , . . . , vk−1 , vk = y)
hence Cay(G, S) is connected.
Suppose now that Cay(G, S) is connected. Then for any g ∈ G, there exists a path
1, v1 , . . . , vk = g. Since vi ∼ vi+1 , it follows that vi+1 = vi si for some si ∈ S. Therefore,
g = vk = vk−1 sk−1 = . . . = s1 . . . sk−1 , hence hSi = G.

For a graph Γ with minimum degree at least 2, girth of Γ is defined as the minimum
number of vertices of a cycle in Γ.
Example 3.14. Let Γ = Cay(G, S), where G is abelian group and |S| ≥ 3. Then girth of Γ
is at most 4.
Proof. Let a, b, c be three distinct elements from S. If a + b = 6 0, then we obtain a 4-cycle,
0, a, a + b, b. If a + b = 0, then a + c 6= 0, and hence we again obtain 4-cycle 0, a, a + c, c.

Recall that an automorphism of a group G is a bijective function α : G → G such that

α(g1 · g2 ) = α(g1 ) · α(g2 ), for every g1 , g2 ∈ G.
Lemma 3.15. Let G be a finite group, S an inverse closed subset of G, and α an automor-
phism of the group G. Then Cay(G, S) ∼
= Cay(G, α(S)).
Proof. Since α is automorphism of the group G, it is bijective mapping of the vertex set of
Cay(G, S) into the vertex set of Cay(G, α(S)). We leave the details for the reader to verify
that α is also an isomorphism between these two graphs.

24
 Figure 2: Petersen graph

Corollary 3.16. Let G be a group, S an inverse closed subset of G and α ∈ Aut(G). Then
α ∈ Aut(Cay(G, S)) if and only if α(S) = S.
We denote with Aut(G, S) = {α ∈ Aut(G) | α(S) = S}. The previous corollary shows
that Aut(G, S) is a subgroup of Aut(Cay(G, S). Hence we have the following.
Proposition 3.17. Let Γ = Cay(G, S). Then hGL , Aut(G, S)i ∼
= GL o Aut(G, S) ≤ Aut(Γ).
We have seen that every Cayley graph is vertex-transitive. But it is not true that every
vertex-transitive graph is Cayley graph.
Example 3.18. Petersen graph shown on Figure 2 is vertex-transitive, but it is not a Cayley
graph.
Proof. Suppose that there exists a group of G order 10 such that Cay(G, S) is isomorphic
to the Petersen graph. Since Petersen graph is 3-valent, then |S| = 3. There are two non-
isomorphic groups of order 10, Z10 and D10 . Observe that girth of the Petersen graph is 5.
This shows that G cannot be abelian, hence it remains to consider the case when G = D10 .
Recall that D10 = {1, ρ, ρ2 , ρ3 , ρ4 , τ, τ ρ, τ ρ2 , τ ρ3 , τ ρ4 }. Since S is inverse closed, it follows
that it contains one or three involutions (elements of order 2). Suppose S contains exactly
one involution. Then without loss of generality, we may assume that S = {ρ, ρ−1 , τ } (verify
this). Observe that the obtained graph is not isomorphic to the Petersen graph. The case
when S contains three involutions is left for exercise.

Exercise 3.19. Prove that the automorphism group of the Petersen graph is isomorphic to
S5 .
Proof. We will represent Petersen graph in the following way. Let I = {1, 2, 3, 4, 5} and let
V = {A : A ⊂ I and |A| = 2}. For A, B ∈ V we define that A ∼ B ⇔ A ∩ B = ∅. It is left for
reader to verify that the graph obtained this way is isomorphic to the Petersen graph.

25
 Let A be the automorphism group of the Petersen graph. For g ∈ S5 , let g (2) denote the
corresponding permutation of V . It is easy to see that g (2) is automorphism of the Petersen
(2)
graph for every g ∈ S5 . Hence S5 ≤ A. We claim that we have equality. We will use orbit
(2)
stabilizer lemma several times to prove that |A| = 120, which will establish A = S5 ∼ = S5 .
By the orbit stabilizer lemma, it follows that

|A| = |A{1,2} ||OrbA ({1, 2})| = 10 · |A{1,2} |, (5)

since Petersen graph is vertex-transitive. Let G = A{1,2} . We now use orbit-stabilizer lemma
for G to obtain:
|G| = |G{3,4} ||OrbG ({3, 4})| = 3 · |G{3,4} |, (6)
(2)
since |OrbG ({3, 4})| = 3 (To see this we can use elements of S5 ). Let H = G{3,4} . Use
orbit-stabilizer lemma for H to obtain:

|H| = |H{4,5} ||OrbH ({4, 5})| = 2 · |H{4,5} |, (7)

since OrbH ({4, 5}) = {{4, 5}, {3, 5}}. To see this one can use automorphism g (2) where
g = (3 4) ∈ S5 . Let K = H{4,5} . Use orbit-stabilizer lemma for K to obtain

|K| = |K{1,3} ||OrbK ({1, 3})| = 2 · |K{1,3} |, (8)

since OrbK ({1, 3}) = {{1, 3}, {2, 3}}. To see this one can use automorphism g (2) where
g = (1 2) ∈ S5 . Finally, we claim that K{1,3} = {1A }. Let α ∈ K{1,3} . Then α fixes
vertices {1, 2}, {3, 4}, {4, 5} and {1, 3}. Observe that {2, 4} is the unique common neighbour
of {3, 4} and {1, 3}. Hence it must be fixed by α. Now it easy to see that α fixes every vertex.
Combining the above equalities implies that |A| = 120.

In the following theorem, we give necessary and sufficient conditions for a given graph to
be a Cayley graph. Recall that a permutation group G ≤ Sym(V ) is called regular if G is
transitive, and Gv = {1G } for every v ∈ V .

Theorem 3.20 (Sabidussi). Let Γ be a graph. Then Γ is isomorphic to Cayley graph on

group G if and only if there exists a regular subgroup H ≤ Aut(Γ) such that H ∼
= G.

Proof. We already proved that for a Cayley graph Cay(G, S), the group GL ≤ Aut(Cay(G, S))
is regular.
Suppose now that Γ is a graph, and that there exists a regular subgroup H ≤ Aut(Γ),
such that G ∼= H. Suppose that V (Γ) = {v1 , . . . , vn }. Then for each i ∈ {1, . . . , n}, there
exists a unique element hi ∈ H such that hi (v1 ) = vi (uniqueness follows from the fact
that H is regular). Observe that h1 = 1H . So we have H = {h1 = 1H , h2 , . . . , hn }. Let
α : H → G be an isomorphism, and let gi := α(hi ). So we have G = {g1 = 1G , . . . , gn }. Let
D = {hi ∈ H : v1 ∼Γ hi (v1 )}, and let S = {α(hi ) : hi ∈ D}. We claim that Γ is isomorphic
to the graph Cay(G, S). Let ϕ : V → G be defined with ϕ(vi ) = gi . It is clear that ϕ is
bijection.

26
 Let vi , vj ∈ V (Γ). Observe that

vi ∼Γ vj ⇔ h−1 −1
i (vi ) ∼Γ hi (vj )
⇔ v1 ∼Γ (h−1
i ◦ hj )(v1 )
⇔ h−1
i hj ∈ D
⇔ α(h−1
i hj ) ∈ S
⇔ α(h−1
i )α(hj ) ∈ S
⇔ gi−1 gj ∈ S
⇔ gi ∼Cay(G,S) gj
⇔ α(vi ) ∼Cay(G,S) α(vj )

This shows that α is indeed an isomorphism between Γ and Cay(G, S).

We have seen that Petersen graph is not Cayley graph. The followin result shows that all
vertex-transitive graphs of prime order are Cayley graphs.
Lemma 3.21. Let Γ be a vertex-transitive graph of order p, where p is a prime. Then Γ is
a Cayley graph on cyclic group Zp .
Proof. Let Γ be a vertex-transitive graph of order p, and let A = Aut(Γ). By orbit-stabilizer
theorem, we have |A| = p|Av | for some vertex v. Hence, the order of group A is divisible
by p. By Cayley theorem, there exists an element g of order p in A. Let G = hgi. Then
G is cyclic group of order p. Then using the orbit-stabilizer theorem for G, it follows that
p = |G| = |Gv ||OrbG (v)|. Since p is prime, it follows that |Gv | = p, or |Gv | = 1. It is now easy
to see that G is a regular subgroup of A, hence by Sabidussi theorem Γ is a Cayley graph on
G.

Can it happen that Cay(G, S1 ) ∼

= Cay(H, S2 ), where G and H are non-isomorphic groups?
Example 3.22. Let G = Z32 , H = Z4 × Z2 , S1 = {(1, 0, 0), (0, 1, 0), (0, 0, 1)}, and S2 =
{(1, 0), (3, 0), (0, 1)}. Observe that both Cay(G, S1 ) and Cay(H, S2 ) are isomorphic to the
3-dimensional cube Q3 .
Example 3.23. Prove that 3-dimensional cube has 48 automorphisms.
Proof. We will represent cube as Cay(Z32 , {(1, 0, 0), (0, 1, 0), (0, 0, 1)}). Let A be the automor-
phism group of the cube. Then by orbit-stabilizer lemma, it follows that

|A| = 8|A(0,0,0) |. (9)

Now apply orbit-stabilizer lemma for A(0,0,0) . We have

|A(0,0,0) | = |A(0,0,0),(1,0,0) ||Orb|A(0,0,0) | (0, 1, 0)|. (10)

For a permutation g ∈ S3 define function g : Z32 → Z32 with g(x1 , x2 , x3 ) = (xg(1),xg(2) ,xg(3) .
It is left for reader to prove that g is automorphism of the cube. Using these permutations,
it can be easily seen that |Orb|A(0,0,0) | (0, 1, 0)| = 3. Now again use orbit-stabilizer lemma to
obtain:
|A(0,0,0),(1,0,0) | = |A(0,0,0),(1,0,0),(0,1,0) ||OrbA(0,0,0),(1,0,0) (0, 1, 0)|. (11)

27
Again using permutations g it is easy to see that |OrbA(0,0,0),(1,0,0) (0, 1, 0)| = 2. Furthermore,
it is not difficult to verify that |A(0,0,0),(1,0,0),(0,1,0) | = 1. Hence

|A| = 8|A(0,0,0) | = 8·3·|A(0,0,0),(1,0,0) | = 24·2·|A(0,0,0),(1,0,0),(0,1,0) ||OrbA(0,0,0),(1,0,0) (0, 1, 0)| = 48.

(12)

Example 3.24. Let Cn be the cycle of length n. Find all groups G, such that Cn is cayley
graph on G.

Proof. Let Γ = Cn . We know that Aut(Γ) ∼ = D2n = hρ, τ | ρn = τ 2 = 1, ρτ = τ ρ−1 i. If Γ

is Cayley graph on G, then by Sabidussi theorem, G is regular subgroup of D2n . Hence G is
of order n. It is easy to see that the subgroup generated by ρ (which is cyclic of order n) is
regular. Denote H = hρi. Let G 6= H be another regular subgroup of D2n . Then GH = D2n .
We have
|G||H| n2
|G ∩ H| = = = n/2.
|GH| 2n
This means that n must be even, and G ∩ H is an index 2 subgroup of H, hence G ∩ H = hρ2 i.
So H contains n/2 rotations of form ρ2i , and n/2 reflections without fixed point. It can be
seen that G is isomorphic to the dihedral group of order n.

3.2 Normal Cayley graphs

Let Γ = Cay(G, S). Recall that we proved that GL is regular subgroup of Aut(Γ), and
also that Aut(G, S) ≤ Aut(Γ). For such automorphisms we might say that they are obvious
automorphisms of Cayley graphs. In some cases, it turns out that these generate the full
automorphism group of a given Cayley graph.

Definition 3.25. Let Γ = Cay(G, S). If GL is normal subgroup of Aut(G) then Cay(G, S)
is said to be normal cayley graph on G with respect to S.

Theorem 3.26. Let Γ = Cay(G, S) and A = Aut(X). The following claims are equaivalent:

1. Cay(G, S) is normal cayley graph.

2. A1G = Aut(G, S);

3. A = hGL , Aut(G, S)i ∼

= GL o Aut(G, S);

4. |A| = |G| · |Aut(G, S)|.

Remark: If a graph Γ can be represented as Cay(G, S1 ) and Cay(H, S2 ) then it is possible

that with respect to one representation it is normal Cayley graph, and with respect to the
other one it is not normal Cayley graph.

Example 3.27. Let Γ be the cube. Then it can be represented as Cayley graph on Z32 and
Z4 × Z2 , as we saw in Example 3.22. It can be seen that the first representation is normal,
while the second is not.

28
Theorem 3.28. Let G be finite abelian group, and let S be an inverse closed generating set,
such that 0 6∈ S. If S satisfies

∀s1 , s2 , s3 , s4 ∈ S s1 + s2 = s3 + s4 6= 0 ⇒ {s1 , s2 } = {s3 , s4 },

then Cay(G, S) is normal Cayley graph.

Example 3.29. Let Γ = Cay(Z12 , {1, 6, 11}). Calculate Aut(Γ).

Proof. By Theorem 3.28 it follows that Γ is normal Cayley graph. Let us determine Aut(Z12 , {1, 6, 11}).
We know that Aut(Z12 ) = {ϕ1 , ϕ5 , ϕ7 , ϕ11 }, where ϕk (x) = k · x. It is now easy to see that
Aut(Z12 , {1, 6, 11}) = {ϕ1 , ϕ11 }. Observe that ϕ1 = id. Therefore, | Aut(Γ)| = 12 · 2 = 24
and Aut(Γ) = h(Z12 )L , ϕ11 , ϕ11 i. Observe that elements of (Z12 )L represent rotations, and
ϕ11 : x 7→ −x is reflection. Hence Aut(Γ) ∼ = D24 .

Example 3.30. Let Γ = Cay(Z29 , {1, 13, 16, 28}). Determine all automorphisms of Γ.

Example 3.31. Let G = Zn2 and let S = {(1, 0 . . . , 0), . . . , (0, 0, . . . , 1)}. Prove that Cay(G, S)
is normal, and determine Aut(Cay(G, S)). (Observe that Cay(G, S) is n-dimensional cube.)

Exercise 3.32. Let X = Cay(Z25 , {±1, ±7}.

• Prove that X is normal Cayley graph.

• Calculate Aut(Z25 , {±1, ±7}).

• Calculate |Aut(X)|.

3.3 Graph products

In this section we introduce notions of Cartesian, direct, lexicographic and strong product of
graphs.

Definition 3.33. Let X and Y be graphs. Cartesian product XY of graphs X and Y is the
graph with vertex set V (X) × V (Y ) and (x1 , y1 ) is adjacent to (x2 , y2 ) if and only if x1 = x2
and {y1 , y2 } ∈ E(Y ) or y1 = y2 and {x1 , x2 } ∈ E(X).

Definition 3.34. Let X and Y be graphs. Direct product X × Y of graphs X and Y is

the graph with vertex set V (X) × V (Y ) and (x1 , y1 ) is adjacent to (x2 , y2 ) if and only if
{x1 , x2 } ∈ E(X) and {y1 , y2 } ∈ E(Y ).

Definition 3.35. Let X and Y be graphs. Strong product X  Y of graphs X and Y is the
graph with vertex set V (X) × V (Y ) and (x1 , y1 ) is adjacent to (x2 , y2 ) if and only if one of
the following holds

1. x1 = x2 and {y1 , y2 } ∈ E(Y );

2. y1 = y2 and {x1 , x2 } ∈ E(X);

3. {x1 , x2 } ∈ E(X) and {y1 , y2 } ∈ E(Y ).

Observe that the edge set of X  Y is disjoint union of edge sets of XY and X × Y .

29
Definition 3.36. Let X and Y be graphs. Lexicographic product X[Y ] of X by Y is the
graph with vertex set V (X) × V (Y ) and (x1 , y1 ) is adjacent to (x2 , y2 ) if and only if x1 = x2
and {y1 , y2 } ∈ E(Y ) or {x1 , x2 } ∈ E(X).

The Cartesian, direct, strong and lexicographic products are called standard graph prod-
ucts. If G ≤ Sym(A) and H ≤ Sym(B), then G × H acts naturally on A × B with
(g, h) : (x, y) 7→ (g(x), h(y), for all g ∈ G and h ∈ H.

Proposition 3.37. Let G ≤ Aut(X) and H ≤ Aut(Y ). Then:

1. G × H ≤ Aut(XY );

2. G × H ≤ Aut(X × Y );

3. G × H ≤ Aut(X  Y );

4. G × H ≤ Aut(X[Y ]).

Proof. It is clear that (g, h) is permutation of vertex set of any of the four standard graph
products. Case by case analysis leads to the proof that (g, h) is automorphism of graph
obtained as one of the four products. Details are left to the reader.

Proposition 3.38. Let X, Y be two graphs and Z is obtained as one of the four standard
products of X and Y . If X and Y are vertex-transitive then Z is vertex-transitive. If X and
Y are Cayley graphs then Z is Cayley graph.

Proof. Follows by Proposition 3.37 and Theorem 3.20.

3.4 Hamiltonicity of vertex-transitive graphs

Definition 3.39. Hamilton path in a graph Γ is a path that visits each vertex exactly once.
A Hamilton cycle (or Hamilton circuit) is a Hamilton path that is a cycle. A graph that
admits a Hamilton cycle is called Hamiltonian graph.

The problem of determining whether or not a given connected vertex-transitive graph

has a Hamilton cycle or Hamilton path is one of the oldest and most well-known problem
concerning vertex-transitive graphs, and as we shall see many additional problems have arisen
via attempts to solve these problems. The problem began with what is usually referred to as
a conjecture, posed by Lovasz.

Example 3.40. Let Γ = Cay(Z10 , {2, 5, 8}). Prove that Γ has a Hamiltonian cycle.

Example 3.41. Find a Hamilton cycle in graph Cay(Z12 , {±3, ±4}).

Example 3.42. Petersen graph has a Hamiltonian path, but it has no Hamiltonian cycle.

Proof. It is easy to construct a Hamiltonian path in Petersen graph. To see that Petersen
graph does not have a Hamilton cycle we will use the fact that the girth of Petersen graph
is 5. Now suppose contrary, that there is a cycle v0 . . . v9 of length 10 in the Petersen graph.
There are 5 more edges that are not lying on this cycle, and each vertex vi is incident with
1 such edge. Since girth of the Petersen graph is 5, it follows that vi can be adjacent only

30
with vi+4 , vi+5 or vi+6 . If each vertex vi is adjacent with vi+5 then we obtain cycle of length
4, for example v0 , v1 , v6 , v5 . So without loss of generality, we may assume that v0 is adjacent
with one of v4 or v6 , and furthermore, without loss of generality we may assume that v0 is
adjacent with v6 . By the girth condition, we have that v1 can only be adjacent with v5 , v6 or
v7 . Moreover, since v0 ∼ v6 , it follows that v1 is not adjacent with v6 . If v1 ∼ v5 , we obtain
4-cycle v0 , v1 , v5 , v6 and if v1 ∼ v7 we obtain 4 cycle v0 , v1 , v7 , v6 , a contradiction with the fact
that girth of Petersen graph is 5. The obtained contradiction shows that Petersen graph does
not admit a Hamilton cycle.

The only known vertex-transitive graphs that don’t admit a Hamilton cycle are K1 , K2 ,
Petersen graph, truncation of Petersen graph, Coxeter graph (cubic graph of order 28) and
truncation of Coxeter graph. Of course graphs K1 and K2 can be considered as trivial exam-
ples, as these graphs have no cycles at all. It turns out that none of the 4 non-trivial examples
is a Cayley graph. This has led to the following ”Folklore” conjecture.
Conjecture 3.43. Every connected Cayley graph of order at least 3 has a Hamilton cycle.
This conjecture has drawn a lot of attention of researchers working in the algebraic graph
theory.
We will prove that this conjecture is true for Cayley graphs on Abelian groups. Before
presenting the proof we need the following lemma.
Lemma 3.44. For every n ≥ 1 and m ≥ 3 graph Cm Pn admits a Hamilton cycle.
Proof. Proof is left for reader as exercise.

Theorem 3.45. A connected Cayley graph of an abelian group of order at least 3 is Hamil-
tonian.
Proof. Let Γ = Cay(G, S) be connected, where G is abelian and S = S −1 . Since Γ is
connected, it follows that hSi = G. We proceed by induction on |S|. If |S| = 1, then S = {a},
a ∈ G, a is self-inverse and G is generated by a. This implies that G = {1, a}, which is of
order 2, a contradiction. If |S| = 2, then either S = {a, a−1 } with a a generator of G, or
S = {a, b}, where both a and b are self-inverse. In the former case, we have the Hamilton
cycle 1, a, a2 , . . . , an−1 , where an = 1 and n = |G| = |hai|. In the later case, as G is abelian
and G = ha, bi, we see that G = Z2 × Z2 . The only connected Cayley graphs of Z2 × Z2 are
C4 and K4 , which are both Hamiltonian. Assume the result holds for every generating set of
size at most m − 1 ≥ 2, and let S be a self-inverse generating set of G of size m.
If m = 3, then S contains a self-inverse element a, and we define T := S \ {a}. If m > 3,
then we take a to be arbitrary element of S, and define T := S \ {a, a−1 }. Observe that in
both cases we have that 2 ≤ |T | < m. Let H = hT i. Then H is abelian group of order at
least 3. Let X = Cay(H, T ). Then we can apply induction hypothesis on X, since it is a
connected Cayley graph on abelian group of order at least 3 and its generating set has size
less than m. Therefore, there exists a Hamilton cycle in the graph X. Suppose that this
Hamilton cycle is h1 , h2 . . . hk , where H = {h1 , . . . , hk }. Let n be the smallest positive integer
such that an ∈ H. Then G is disjoint union of H,aH,a2 H,. . . , an−1 H. Observe now that
ah1 , ah2 . . . ahk is also a cycle, and that each x is adjacent with ax. Hence we obtain that
Γ contains a spanning subgraph isomorphic to Ck Pn . By previous Lemma, it follows that
there is a Hamilton cycle in Ck Pn , hence there is also a Hamilton cycle in Γ.

31
Example 3.46. Let G = Z6 × Z4 in S = {(3, 1), (3, 3), (2, 1), (4, 3)}. Find a Hamilton cycle
in graph Cay(G, S).

32
4 Edge-transitive and arc-transitive graphs
Automorphism of a graph is defined as permutation of vertex-set, that preserves edges. Every
automorphism of a graph can be also considered as a permutation of edges, that is we can
consider natural action of Aut(Γ) on E(Γ). For an automorphism ϕ ∈ Aut(Γ) and an edge
e = {x, y} we define ϕ(e) = {ϕ(x), ϕ(y)}.

Definition 4.1. A graph Γ = (V, E) is said to be edge-transitive if Aut(Γ) acts transitively

on the edge set, that is for any two edges {u, v}, {x, y} ∈ E, there exists ϕ ∈ Aut(Γ) such
that ϕ{x, y} = {u, v}.

Example 4.2. Complete bipartite graph Km,n is edge-transitive. Furthermore, it is vertex-

transitive if and only if m = n. Hence there exist edge-transitive graphs which are not
vertex-transitive.

We have seen that every Cayley graph is vertex-transitive. Cayley graphs don’t need to
be edge-transitive.

Example 4.3. Cay(Z6 , {±2, 3}) is not edge-transitive.

Example 4.4. For every odd positive integer m ≥ 3 graph Cay(Zm × Z2 , {(±1, 0), (0, 1)}) is
not edge-transitive.

Theorem 4.5. Let Γ be an edge-transitive graph without isolated vertices. If Γ is not vertex-
transitive, then Γ is bipartite. Moreover, Aut(G) has two orbits on vertices and these two
orbits form a bipartition of Γ.

Proof. Let {u, v} be an edge of Γ. Let U = {ϕ(u) : ϕ ∈ Aut(Γ)} and V = {ϕ(v) : ϕ ∈ Aut(Γ)}.
Since Γ is not vertex-transitive, it follows that U ∩ V = ∅. Let x be an arbitrary vertex of Γ.
Since Γ has no isolated vertices it follows that there exist y ∈ V (Γ) such that x ∼ y. Since Γ
is edge-transitive, there exists ϕ ∈ Aut(Γ) such that ϕ{u, v} = {x, y}. This shows that x ∈ U
or x ∈ V . Hence V (Γ) is disjoint union of U and V . Suppose now that there is an edge {x, y}
where both x and y belong to U . Then there exists ϕ ∈ Aut(Γ) such that ϕ{u, v} = {x, y}.
This would imply that ϕ(v) = x or ϕ(v) = y and therefore x ∈ V or y ∈ V . However, this is
impossible, since U ∩ V = ∅. This shows that sets U and V form a bipartition of Γ.

Definition 4.6. Let Γ be a graph. Arc in Γ is a pair (u, v), where u and v are adjacent
vertices of Γ.

Observe that for any graph Γ, the number of arcs is twice the number of edges, that is for
every edge {u, v} there are two arcs (u, v) and (v, u).

Definition 4.7. A graph Γ is said to be arc-transitive if Aut(Γ) acts transitively on the arc
set of Γ, that is for any two pairs of adjacent vertices u ∼ v and x ∼ y, there exists an
automorphism ϕ ∈ Aut(Γ) such that ϕ(u) = x and ϕ(v) = y.

Proposition 4.8. Let Γ be a graph. If Γ is arc-transitive then Γ is edge-transitive. Moreover,

if Γ is without isolated vertices, then Γ is vertex-transitive.

Proof is straightforward and is omitted.

33
Remark 4.9. There exist graphs which are vertex-transitive and edge-transitive but not
arc-transitive. Such graphs are called half-arc-transitive. The smallest example is a 4-valent
graph of order 27 and it so called Doyle-Holt graph.

Proposition 4.10. Let Γ be a vertex-transitive graph. Then Γ is arc-transitive if and only if

for an arbitrary vertex v, the stabilizer Aut(G)v acts transitively on Γ(v), the set of neighbour
of v in Γ.

Proof. Suppose first that Γ is arc-transitive. Let v ∈ V (Γ) and let a, b ∈ Γ(v) be arbitrary.
Since Γ is arc-transitive, we can map arc (v, a) to arc (v, b), hence there exists automorphism
ϕ such that ϕ(v) = v and ϕ(a) = b. Hence ϕ ∈ Aut(Γ)v and maps a to b. This shows that
Aut(G)v acts transitively on Γ(v).
Suppose now that Aut(G)v acts transitively on Γ(v). Let (x, y) and (a, b) be two arbitrary
arc in Γ. Then there exist automorphism ϕ1 and ϕ2 such that ϕ1 (x) = v and ϕ2 (y) = v. Let
c = ϕ1 (y and d = ϕ2 (b). Observe that c, d ∈ Γ(v). Then there exists ϕ3 ∈ Aut(Γ)v such that
ϕ3 (c) = d. It is now easy to verify that ϕ−1
2 ϕ3 ϕ1 maps arc (x, y) to arc (a, b). This proves
that Γ is arc-transitive.

Example 4.11. Let Γ = Cay(Z17 , {1, 2, 4, 8, 16, 15, 13, 9}. Then Γ is arc-transitive.

Proof. Let S = {1, 2, 4, 8, 16, 15, 13, 9}. Observe that Aut(Z17 , S) = {ϕ1 , ϕ2 , ϕ4 , ϕ8 , ϕ9 , ϕ13 , ϕ15 , ϕ16 }
and that it acts transitively on S. Observe also that Γ(0) = S. Recall that Aut(Z17 , S) ≤
Aut(Γ)0 . This shows that Aut(Γ)0 acts transitively on neighbours of 0, and hence Γ is arc-
transitive.

Proposition 4.12. A Cayley graph on Abelian group is arc-transitive if and only if it is

edge-transitive.

Proposition 4.13. Let Γ be a vertex-transitive and edge-transitive graph which is not arc-
transitive. Then Γ has even valency.

A Cayley graph on cyclic group is called circulant

Exercise 4.14. Find up to isomorphism all cubic edge-transitive circulants.

34
5 s-arc-transitive graphs
Definition 5.1. An s-arc in a graph is a sequence of vertices (v0 , . . . , vs ) such that consecutive
vertices are adjacent and vi−1 6= vi when 0 < i < s. Note that an s-arc is permitted to use
the same vertex more than once. A graph Γ is s-arc transitive if its automorphism group
is transitive on s-arcs, that is for any two s-arcs (v0 , . . . , vs ) and (u0 , . . . , us ) there exists
α ∈ Aut(Γ) such that α(vi ) = ui for i = 0, . . . , s

Observe that 0-arcs are just the vertices and 1-arcs are the usual arcs.

Example 5.2. Cn is s-arc-transitive for every s

Example 5.3. Kn is s-arc-transitive only for s = 0, 1, 2.

Example 5.4. Cube Q3 is 2-arc-transitive but not 3-arc-transitive.

Theorem 5.5. Let Γ be a graph with minimum degree at least 2. Then Γ is (s + 1)-
arc-transitive if and only if Γ is s-arc-transitive and the stabilizer in Aut(Γ) of any s-arc
(v0 , . . . , vs−1 , vs ) acts transitively on the Γ(vs ) \ {vs−1 }.

Example 5.6. Petersen graph is 3-arc-transitive but not 4-arc-transitive.

Proposition 5.7. If Γ is an s-arc-transitive graph with valency at least three and girth g,
then g ≥ 2s − 2.

Proof. We may assume that s ≥ 3, since the condition on the girth is otherwise meaningless.
It is easy to see that Γ contains a cycle of length g and a path of length g whose end-vertices
are not adjacent. Therefore Γ contains a g-arc with adjacent end-vertices and a g-arc with
nonadjacent end-vertices; clearly, no automorphism can map one to the other, and so s < g.
Since Γ contains cycles of length g, and since these contain s- arcs, it follows that any s-arc
must lie in a cycle of length g. Suppose that u0 , . . . , us is an s-arc. Denote it by A. Since us−1
has valency at least three, it is adjacent to a vertex w other than us−2 and us , and since the
girth of Γ is at least s, this vertex cannot lie in A. Hence we may replace us by w, obtaining
a second s-arc B that intersects A in an (s − 1)-arc. Since B must lie in a circuit of length g,
we thus obtain a pair of circuits of length g that have at least s − 1 edges in common. If we
delete these s − 1 edges from the graph formed by the edges of the two circuits of length g,
the resulting graph still contains a cycle of length at most 2g − 2s + 2. Hence 2g − 2s + 2 > g,
and the result follows.

5.1 Cubic-arc-transitive graphs

A graph is said to be cubic if it is 3-regular, that is if each vertex has degree 3. In 1947 Tutte
showed that for any s-arc transitive cubic graph, s ≤ 5. This was, eventually, the stimulus for
a lot of work. One outcome of this was a proof, by Richard Weiss, that for any s-arc transitive
graph, s ≤ 7. This is a very deep result, the proof of which depends on the classification of
the finite simple groups.

Definition 5.8. A graph Γ is s-arc regular if it is s-arc-transitive and stabilizer in Aut(Γ) of

any s-arc is trivial, or equivalently, the automorphism group Aut(Γ) acts regularly on the set
of all s-arcs of Γ.

35
Lemma 5.9. Let Γ be a connected cubic graph that is s-arc transitive, but not (s + 1)-arc
transitive. Then Γ is s-arc regular.

Theorem 5.10 (Tutte). If Γ is an s-arc regular cubic graph, then s ≤ 5.

Corollary 5.11. If Γ is cubic arc-transitive graph then |Aut(Γ)| = |V (Γ)| · 3 · 2s−1 for some
s ∈ {1, 2, 3, 4, 5}.

Example 5.12. Petersen graph is 3-arc-regular.

36
6 Distance-transitive graphs
Definition 6.1. A connected graph Γ is distance-transitive if given any two ordered pairs of
vertices (u, u0 ) and (v, v 0 ) such that d(u, u0 ) = d(v, v 0 ), there is an auto- morphism α of Γ such
that α(v, v 0 ) = (u, u0 ).

Example 6.2. Complete graph Kn is distance-transitive.

For a connected graph Γ and a vertex v ∈ V (Γ) let Γi (v) be the set of vertices at distance
i from v, that is Γi (v) = {u ∈ V (Γ) | d(u, v) = i}. Observe that for any connected graph
Γ with diameter d we have that V (Γ) is partitioned into sets Γ0 (v), Γ1 (v), . . ., Γd (v). The
following theorem gives a characterization of distance-transitive graphs, based on the action
of automorphism group on sets Γi (v). This partition is called distance-partition of Γ.

Theorem 6.3. Let Γ be a connected graph. Then Γ is distance transitive if and only if the
following conditions hold:

1. Γ is vertex-transitive;

2. Aut(Γ)v acts transitively on each of sets Γi (v) (i = 1, . . . , diam(Γ)), for any vertex
v ∈ V (Γ).

Corollary 6.4. Let Γ be a connected distance-transitive graph. Then Γ is arc-transitive.

Example 6.5. Petersen graph is distance-transitive.

We would like to note that distance-transitive graphs are not necessarily s-arc-transitive
for higher values of s.

Example 6.6. A cube graph Q3 is distance-transitive, it has diameter 3, but it is not 3-arc-
transitive.

Example 6.7. Let G = Z17 and S = {1, 2, 4, 8, 9, 13, 15, 16} and let Γ = Cay(G, S). Then Γ
is distance-transitive with diameter 2.

Example 6.8. Let p ≡ 1 (mod 4) be a prime, and let G = Zp . Let S = {x2 : x ∈ 1, . . . , p − 1.

Observe that S = −S and that |S| = (p − 1)/2. Let Γ = Cay(G, S). Then Γ is distance-
transitive with diameter 2.

Example 6.9. Let p be a prime, and q = pn such that q ≡ 1 (mod 4). Let Fq be the field of
order q. Let S be the set of all non-zero squares in Fq , that is S = {x2 | x ∈ Fq \ {0}}. Let G
be the additive group of the field Fq , that is G = (Fq , +). The graph Γ = Cay(G, S) is called
Paley graph. Prove that Paley graphs are distance-transitive.

Suppose that Γ is a connected distance-transitive graph and u ∈ V (Γ). Since the cells of
the distance partition Γi (u) are orbits of Aut(Γ)u , every vertex in Γi (u) is adjacent to the
same number of other vertices, say ai , in Γi (u). Similarly, every vertex in Γi (u) is adjacent
to the same number, say bi , of vertices in Γi+1 (u) and the same number, say ci , of vertices in
Γi−1 (u). The graph X is regular, and its valency is given by b0 , so if the diameter of Γ is d,
we have
ci + ai + bi = b0 , i = 0, 1, ..., d.

37
These numbers are called the parameters of the distance-transitive graph, and determine
many of its properties.
Distance transitivity is a symmetry property in that it is defined in terms of the existence
of certain automorphisms of a graph. These automorphisms impose regularity properties on
the graph, namely that the numbers ai , bi , and ci are well-defined. There is an important
combinatorial analogue to distance transitivity, which simply asks that the numerical regu-
larity properties hold, whether or not the automorphisms exist. Given any graph Γ we can
compute the distance partition from any vertex u, and it may occur ”by accident” that every
vertex in Γi (u) is adjacent to a constant number of vertices in Γi−1 (u),Γi (u) and Γi+1 (u),
regardless of whether there are any automorphisms that force this to occur. Such graphs are
called distance-regular graphs.

Proposition 6.10. Every distance-transitive graph is distance-regular.

A special case are distance-regular graphs of diameter 2. Such graphs are called strongly
regular graphs.

Exercise 6.11. Let Γ be a distance-regular graph with diameter 2. Prove that there exist
parameters (n, k, λ, µ) such that Γ has order n, it is k-regular, every edge lies on λ triangles
(in other words every pair of adjacent vertices have λ common neighbours), and every pair of
distinct non-adjacent vertices have µ common neighbours. A graph Γ is said to be strongly
regular with parameters (n, k, λ, µ).

Example 6.12. Prove that the complement of a strongly regular graph is strongly regular.

Example 6.13. Let Γ be a connected vertex-transitive graph, and suppose that Aut(G)v
have exactly three orbits in its action on V (Γ). Prove that Γ is distance-transitive.

Example 6.14. The graph L(Kn ) (that is the line graph of Kn ) is strongly regular with
parameters (n(n − 1)/2, 2n − 4, n − 2, 4).

Example 6.15. The graph L(Kn,n ) is strongly regular with parameters (n2 , 2n − 2, n − 2, 2).

Exercise 6.16. Let G = Z4 × Z4 and let S = {±(1, 0), ±(0, 1), ±(1, 1)}. The graph Γ =
Cay(G, S) is called Shrikhande graph. Prove that it is strongly regular, but not distance-
transitive.

6.1 Eigenvalues of strongly regular graphs

Let Γ be a graph of order n with vertex set {v1 , . . . , vn }. Adjacency matrix A = A(Γ) is n-
by-n matrix with value Ai,j equal to 1 if and only if {vi , vj } ∈ E(Γ). Observe that adjacency
matrix is a symmetric matrix, that is AT = A. Such matrices have several nice properties.

Lemma 6.17. All eigenvalues of a real symmetric matrix are real.

Proof. Suppose that α ∈ C is an eigenvalue of a symmetric matrix A. Let x 6= 0 be a

corresponding eigenvector. Then Ax = αx. Then x∗ Ax = x∗ αx = αx∗ x. On the other hand
αx∗ x = x∗ αx = x∗ Ax = (x∗ Ax)∗ = αx∗ x. Therefore, α = α, hence α is real.

Lemma 6.18. Let A be a symmetric real matrix. Then A is diagonalizable.

38
Exercise 6.19. Let Γ be a connected k-regular graph. Then k is an eigenvalue of A(Γ) with
multiplicity one.

Proof. Let Γ be a connected k-regular graph of order n, and let {v1 , . . . , vn } be the vertex set
of Γ. Let A be its adjacency matrix. Then it is easy to see that A · 1 = k1, hence k is an
eigenvalue. Suppose that x is an eigenvector of A corresponding to k. Let x = (x1 , . . . , xn )T .
Since x is eigenvector corresponding to eigenvalue k, it follows that Ax = kx. Let xj be the
maximum of x1 , . . . , xn , that is xi ≤ xj for every i ∈ {1, . . . , n}. Then
X X
kxj = (kx)j = (Ax)j = Aj,i xi = xi ≤ kxj .
vi ∈Γi (vj )

Therefore, we conclude that xi = kj for every i such that vi ∈ Γ(vj ). The connectedness of Γ
now implies that all xi must be equal to xj , hence x = xj 1. This shows that the multiplicity
of k as an eigenvalue if 1.

Suppose A is the adjacency matrix of the (n, k, λ, µ) strongly regular graph Γ. We can
determine the eigenvalues of the matrix A from the parameters of Γ and thereby obtain some
strong feasibility conditions. The uv-entry of the matrix A2 is the number of walks of length
two from the vertex u to the vertex v. In a strongly regular graph this number is determined
only by whether u and v are equal, adjacent, or distinct and nonadjacent. Therefore, we get
the equation
A2 = kI + λA + µ(J − I − A)
Where J denotes all 1 matrix. The above equation can be rewritten as

A2 − (λ − µ)A − (k − µ)I = µJ.

We can use this equation to determine the eigenvalues of A. Since Γ is regular with valency
k, it follows that k is an eigenvalue of A with eigenvector (1, . . . , 1). Any other eigenvector of
A is orthogonal to (1, . . . , 1) (this follows since A is symmetric matrix, that is AT = A). Let
z be an eigenvector for A with eigenvalue θ 6= k. Then

A2 z − (λ − µ)Az − (k − µ)Iz = µJz = 0,

so
θ2 − (λ − µ)θ − (k − µ) = 0.
Therefore, the eigenvalues of A different from k must be zeros of the quadratic

x2 − (λ − µ)x − (k − µ).

If we set ∆ = (λ − µ)2 + 4(k − µ) (the discriminant of the quadratic) and denote the two zeros
of this polynomial by τ and θ, we get
√
λ−µ− ∆
τ=
2
and √
λ−µ+ ∆
θ=
2

39
Now, τ θ = (µ−k), and so, provided that µ < k, we get that τ and θ are nonzero with opposite
signs. We see that the eigenvalues of a strongly regular graph are determined by its parameters
(although strongly regular graphs with the same parameters need not be isomorphic). The
multiplicities of the eigenvalues are also determined by the parameters.
Exercise 6.20. Determine the multiplicities of eigenvalues of an (n, k, λ, µ) strongly regular
graph.
Theorem 6.21. A connected regular graph with exactly three distinct eigenvalues is strongly
regular.
Proof. Suppose that Γ is connected and regular with eigenvalues k, θ, and τ , where k is the
valency of Γ, and let n be the order of Γ. Let A be the adjacency matrix of G. Since
A is symmetric, the sum of multiplicities of its eigenvalues equals n. Moreover, since Γ is
connected, eigenvalue k has multiplicity 1. Define matrix M with
1
M := (A − θI)(A − τ I)
(k − θ)(k − τ )
Observe that kernel of M consists precisely of eigenvectors of A corresponding to θ or τ , hence
kernel of M has dimension n − 1. Moreover, we have
1
M · [1, . . . , 1]T = (A − θI)(A − τ I) · [1, . . . , 1]T
(k − θ)(k − τ )
1
= (A − θI)((A · [1, . . . , 1]T − τ · [1, . . . , 1]T ))
(k − θ)(k − τ )
1
= (A − θI)((k − τ ) · [1, . . . , 1]T ))
(k − θ)(k − τ )
1
= (A − θI) · [1, . . . , 1]T = [1, . . . , 1]T .
(k − θ)
This implies that M = n1 J (explain why).
We have shown that J is a quadratic polynomial in A, and thus A2 is a linear combination
of I, J, and A. Accordingly, Γ is strongly regular.

Exercise 6.22. Let Γ be a Moore graph with valency k and diameter 2. Prove that k ∈
{2, 3, 7, 57}.
Proof. Observe that a Moore graph with valency k and diameter 2 is (k 2 + 1, k, 0, 1) strongly
regular. Calculate the eigenvalues and the corresponding multiplicities. Use the fact that
multiplicities of eigenvalues are integers.

Exercise 6.23. Let Γ be a connected graph of order 25 and let A be its adjacency matrix. If
A2 + A − 6I = 6J, prove that Γ is strongly regular, and determine its parameters (n, k, λ, µ).
Proof. Observe that we can rewrite the given equality as:
A2 = 12I + 5A + 6(J − I − A).
This means that for vertex v ∈ V (Γ), the number of walks of length 2 between v and v is 12.
Hence Γ is regular with valency 12. Similarly we see that for two adjacent vertices u and v
the number of common neighbours is 5 and for two non-adjacent vertices, the number of their
common neighbours is 6. Therefore, Γ is (25, 12, 5, 6)-strongly regular graph.

40
Exercise 6.24. Let Γ be a connected graph of order 21 and let A be its adjacency matrix
such that A2 − A − 6I = 4J.

(a) Prove that Γ is strongly regular graph;

(b) Determine parameters (n, k, λ, µ) of Γ;

(c) Determine eigenvalues of A and their multiplicities.

41