MATH1081 – Discrete Mathematics

Topic 5 – Graph theory

Lecture 5.01 – Graph and multigraph terminology

Lecturer: Dr Sean Gardiner – [email protected]

Introduction to graph theory
In very simple terms, graphs are mathematical constructions represented by
dots and connecting lines.

These constructions can represent many real-world situations, such as social

networks, transportation options, or linked pages of the internet.
For this topic we will be investigating the abstract notion of a graph and the
many properties different graphs can have.
Much of the topic will be spent establishing precise definitions and proving
useful facts about graphs, but in almost all cases it is usually useful to
accompany your learning with plenty of diagrams.
In lectures, we will sometimes prove statements loosely using diagrammatic
arguments, while properly-written proofs will be provided in the completed
versions of the slides.

MATH1081 Topic 5 – Graph theory Lecture 5.01 – Graph and multigraph terminology 1/17
Graphs
Definition. A graph G is a mathematical object consisting of two related
sets:
• The vertex set of G , written as V (G ) or just V , is a set of vertices (often
labelled v1 , v2 , v3 , ... ).
• The edge set of G , written as E (G ) or just E , is a set of edges
(sometimes labelled e1 , e2 , e3 , ... ). Each edge is itself a subset of V (G )
with cardinality 2, representing the two vertices it connects, which are
known as the edge’s endpoints.
Notice that since each edge is a set of cardinality 2, no edge can have a
single vertex as both of its endpoints in a graph.
Notice also that since E (G ) is a set, there can be at most one edge between
any two distinct vertices in a graph.
Notation. An edge with endpoints v and w is written as {v , w }, but is
more often represented by the shorthand vw , or equivalently, wv .

MATH1081 Topic 5 – Graph theory Lecture 5.01 – Graph and multigraph terminology 2/17
Representing graphs
Graphs can be represented diagrammatically as dots (vertices) joined by
connecting line or curve segments (edges). The relative positions of the
vertices and edges do not matter.
For example, the graph G with vertex set V (G ) = {a, b, c, d} and edge set
E (G ) = {ab, ac, ad, bc, cd} can be represented by any of these diagrams:
a b b d b

d c a c c a

Definition. Given a graph G , if a vertex v ∈ V (G ) is an endpoint of an

edge e ∈ E (G ), then we say v is incident with e and e is incident with v .
For example, ac is incident with a but not with b above.
Definition. Given a graph G , two vertices v , w ∈ V (G ) are said to be
adjacent or neighbours if they are connected by an edge in E (G ), that is, if
{v , w } ∈ E (G ). For example, b is adjacent to a above, but not to d.
MATH1081 Topic 5 – Graph theory Lecture 5.01 – Graph and multigraph terminology 3/17
Multigraphs
Definition. A multigraph G is a mathematical object consisting of two sets
and a function:
• The vertex set V (G ) is simply a set of vertices.

• The edge set E (G ) is simply a set of edges.

• The endpoint function f : E (G ) → P(V (G )) is a function mapping
edges of G to subsets of V (G ) that have cardinality 2 or 1.
Notice that for multigraphs, an edge’s endpoints may be a set of cardinality
1, meaning an edge may start and end at the same vertex.
Definition. A loop is an edge in a multigraph with exactly one endpoint.
Notice also that for multigraphs, the endpoint function might map two
distinct edges to the same pair of endpoints, meaning there can be more
than one edge between any two vertices.
Definition. If two edges in a multigraph have identical endpoints, they are
called parallel edges or multiple edges. If a multigraph has exactly k edges
with endpoints v and w , we say {v , w } is an edge of multiplicity k.

MATH1081 Topic 5 – Graph theory Lecture 5.01 – Graph and multigraph terminology 4/17
Representing multigraphs
Multigraphs can be represented diagrammatically in the same way as graphs.
For example, the multigraph G with vertex set V (G ) = {a, b, c}, edge set
E (G ) = {e1 , e2 , e3 , e4 }, and endpoint function

f = { (e1 , {a, b}), (e2 , {b, c}), (e3 , {b, c}), (e4 , {a}) }

can be represented by either of these diagrams:

e1
e4 a b e2
e2
e1
e3 b a e4 c

c e3

Notice that all graphs are multigraphs, with the restrictions that the
endpoint function cannot map any edge to a single vertex, nor map any two
distinct edges to the same pair of vertices.
Definition. A simple multigraph is a multigraph without any loops or
parallel edges. That is, a simple multigraph is a graph.
MATH1081 Topic 5 – Graph theory Lecture 5.01 – Graph and multigraph terminology 5/17
Directed graphs
Definition. A directed graph G is a mathematical object consisting of two
related sets:
• The vertex set V (G ) is a set of vertices.
• The edge set E (G ), for which each edge is an ordered pair of distinct
vertices of G , representing the edge’s start and end vertices in order.
Definition. A directed multigraph G is a mathematical object consisting of
two sets and a function:
• The vertex set V (G ) is a set of vertices.
• The edge set E (G ) is a set of edges.
• The endpoint function f : E (G ) → V (G ) × V (G ) is a function mapping
edges of G to ordered pairs of vertices.
When representing a directed (multi)graph diagrammatically, each edge is
represented by an arrow pointing from its start vertex to its end vertex.
For example, the arrow diagram for a function f : X → Y is a directed
graph, and the arrow diagram for a relation R on X is a directed multigraph.
The Hasse diagram for any poset is also a directed graph, since all its edges
are technically directed upwards.
MATH1081 Topic 5 – Graph theory Lecture 5.01 – Graph and multigraph terminology 6/17
Special graphs: Path graphs
Most of the Topic 5 content is focused specifically on graphs, as opposed to
multigraphs or directed (multi)graphs.
Some graphs are common or useful enough that they have special names.
We will define several such named graphs in terms of labelled vertices,
though they are often also interpreted as being unlabelled.
Definition. The path graph on n ≥ 1 vertices, written as Pn , is the graph
with vertex set V (Pn ) = {v1 , v2 , ... , vn } and edge set

E (Pn ) = {v1 v2 , v2 v3 , ... , vn−1 vn }.

For example,

v1 v2 v1 v2 v3 v1 v2 v3 v4
P2 P3 P4

Exercise. How many vertices and edges does Pn have?

Solution. The number of vertices is |V (Pn )| = n, and the number of edges
is |E (Pn )| = n − 1.
MATH1081 Topic 5 – Graph theory Lecture 5.01 – Graph and multigraph terminology 7/17
Special graphs: Cycle graphs
Definition. The cycle graph (or cyclic graph) on n ≥ 3 vertices, written as
Cn , is the graph with vertex set V (Cn ) = {v1 , v2 , ... , vn } and edge set

E (Cn ) = {v1 v2 , v2 v3 , ... , vn−1 vn , vn v1 }.

For example,
v2
v1 v1 v2
v1 v3

v3 v2 v4 v3 v5 v4
C3 C4 C5

Exercise. How many vertices and edges does Cn have?

Solution. The number of vertices is |V (Cn )| = n, and the number of edges
is also |E (Cn )| = n.

MATH1081 Topic 5 – Graph theory Lecture 5.01 – Graph and multigraph terminology 8/17
Special graphs: Complete graphs
Definition. The complete graph on n ≥ 1 vertices, written as Kn , is the
graph with vertex set V (Kn ) = {v1 , v2 , ... , vn } and edge set
E (Kn ) = { {v , w } ⊆ V (Kn ) : v ̸= w } = {vi vj : 1 ≤ i < j ≤ n}.
That is, the complete graph Kn is the graph with n vertices and all possible
edges between these vertices. For example,
v2
v1 v1 v2
v1 v3

v3 v2 v4 v3 v5 v4
K3 K4 K5

Exercise. How many vertices and edges does Kn have?

Solution. The number of vertices is |V (Kn )| = n, and the number of edges
is given by the

number of ways to choose two distinct unordered vertices, so
|E (Kn )| = n2 .
MATH1081 Topic 5 – Graph theory Lecture 5.01 – Graph and multigraph terminology 9/17
Bipartite graphs
Definition. A (multi)graph G is called bipartite if its vertex set V (G ) can
be partitioned into two sets V1 and V2 such that each edge in G has one
endpoint in V1 and one endpoint in V2 .
That is, G is bipartite if and only if its edge set E (G ) is a subset of
{ {v , w } ⊆ V (G ) : v ∈ V1 , w ∈ V2 } where V (G ) = V1 ∪ V2 and
V1 ∩ V2 = ∅.
It is standard to represent bipartite graphs by separating the two partitions
of V (G ). Some examples of standard bipartite graph representations are:
a b
V1 a d
V1 b e V2
V2 c f
c d e

Notice that the arrow diagram for a function f : X → Y is a bipartite

directed graph whose vertex set can be partitioned into the sets X and Y .
To decide whether a graph is bipartite, we must construct a valid partition
of its vertex set, or show that no such partition exists.
MATH1081 Topic 5 – Graph theory Lecture 5.01 – Graph and multigraph terminology 10/17
Example – Bipartite graphs
Example. Is the cycle graph C4 bipartite?
Solution. The graph C4 is bipartite, since using the standard labelling, its
vertex set can be partitioned into the sets V1 = {v1 , v3 } and V2 = {v2 , v4 }
such that every edge in C4 has one endpoint in V1 and one endpoint in V2 .
v1 v2 v1 v3
V1

V2
v4 v3 v2 v4

Example. Is the cycle graph C5 bipartite?

Solution. Suppose by way of contradiction that C5 is bipartite, so that its
vertex set can be partitioned into sets V1 and V2 where each edge in C5 has
one endpoint in V1 and one endpoint in V2 . Using the standard labelling, let
v1 be an element of V1 . Then we must have v2 ∈ V2 since v1 v2 ∈ E (C5 ), so
v3 ∈ V1 since v2 v3 ∈ E (C5 ), so v4 ∈ V2 since v3 v4 ∈ E (C5 ), so v5 ∈ V1 since
v4 v5 ∈ E (C5 ). But then the edge v5 v1 in C5 has both its endpoints in V1 ,
which is a contradiction. Thus C5 is not bipartite.
We will see another way to identify non-bipartite graphs in Lecture 5.03.
MATH1081 Topic 5 – Graph theory Lecture 5.01 – Graph and multigraph terminology 11/17
Special graphs: Complete bipartite graphs
Definition. The complete bipartite graph on m ≥ 1 and n ≥ 1 vertices,
written as Km,n , is the graph with vertex set V (Km,n ) = A ∪ B where
A = {v1 , v2 , ... , vm } and B = {w1 , w2 , ... , wn } with A ∩ B = ∅, and edge set
E (Km,n ) = {vw : v ∈ A, w ∈ B}.
That is, the complete bipartite graph Km,n is the graph with a vertex set
partitioned into sets of size m and n, where every vertex in one partition is
adjacent to every vertex in the other partition (and no others). For example,
v1 v2 v1
v1 w1

v2 w2

v3 w3
w1 w2 w3 w1 w2 w3 w4
K2,3 K3,3 K1,4

Exercise. How many vertices and edges does Km,n have?

Solution. The number of vertices is |V (Km,n )| = m + n, and the number of
edges is |E (Km,n )| = mn.
MATH1081 Topic 5 – Graph theory Lecture 5.01 – Graph and multigraph terminology 12/17
Special graphs: Cube graphs
Definition. The cube graph (or hypercube graph) of dimension n, written
as Qn , is a graph that uses a special labelling of its vertices as binary strings.
Its vertex set is given by V (Qn ) = {binary strings of length n}, and its edge
set is
E (Qn ) = { {v , w } ⊆ V (Qn ) : v and w differ by exactly one digit }.
The cube graph Qn is a representation of an n-dimensional hypercube. For
example,
110 111
10 11 010 011

100 101
0 1 00 01 000 001
Q1 Q2 Q3

Exercise. How many vertices and edges does Qn have?

Solution. The number of vertices is |V (Qn )| = 2n , and the number of edges
is |E (Qn )| = n2n−1 , which will be explained in Lecture 5.02.
MATH1081 Topic 5 – Graph theory Lecture 5.01 – Graph and multigraph terminology 13/17
Subgraphs
Definition. Given a graph G , a subgraph of G is a graph H composed of
some subset of the vertices and edges of G . That is, V (H) ⊆ V (G ) and
E (H) ⊆ E (G ), and the endpoints of each edge in E (H) belong to V (H).
The same definition also applies for multigraphs.
For example, G ′ and G ′′ below are both subgraphs of the given graph G ,
but G ′′ is not a subgraph of G ′ .
a b a b a b

d c d c c
G G′ G ′′
Notice that:
• For every integer n ≥ 3, we have that Pn is a subgraph of Cn .
• For any n ∈ Z+ , the (unlabelled) cube graph Qn is a subgraph of the
cube graph Qm for all integers m ≥ n.
• Every (unlabelled) graph with n vertices is a subgraph of Kn , and indeed
a subgraph of Km for all integers m ≥ n.
MATH1081 Topic 5 – Graph theory Lecture 5.01 – Graph and multigraph terminology 14/17
Example – Finding subgraphs
Example. How many subgraphs are there of the graph G with vertex set
V (G ) = {a, b, c} and edge set E (G ) = {ab, ac, bc}?
Solution. Let H be a subgraph of G . We consider different cases for
different sizes of V (H).
• If V (H) = {a, b, c}, then E (H) ⊆ {ab, ac, bc}, so there are 23 = 8
possible values for E (H).
• If V (H) = {a, b}, then E (H) ⊆ {ab}, so there are 21 = 2 possible values
for E (H). Similarly, there are two subgraphs with V (H) = {a, c} and
with V (H) = {b, c}.
• If V (H) = {a}, then E (H) ⊆ {}, so there is 20 = 1 possible value for
E (H). Similarly, there is one subgraph with V (H) = {b} and with
V (H) = {c}.
• If V (H) = {}, then E (H) ⊆ {}, so there is 1 possible value for E (H).

So altogether, there are 8 + 3 × 2 + 3 × 1 + 1 = 18 different subgraphs of G .

Notice this includes the empty graph and the whole graph G .

MATH1081 Topic 5 – Graph theory Lecture 5.01 – Graph and multigraph terminology 15/17
Example – Finding unlabelled subgraphs
Example. How many subgraphs are there of the unlabelled cyclic graph C3 ?
Solution. In this case, it is easier to simply list the subgraphs. They are:
• Subgraphs with exactly 3 vertices:

, , ,
• Subgraphs with exactly 2 vertices:

,
• Subgraphs with exactly 1 vertex:

• Subgraphs with exactly 0 vertices:

(empty graph)
So there are 4 + 2 + 1 + 1 = 8 subgraphs of the unlabelled cyclic graph C3 .

MATH1081 Topic 5 – Graph theory Lecture 5.01 – Graph and multigraph terminology 16/17
Graph complement
Definition. The complement of a graph G , written as G , is the graph with
the same vertex set as G , but with an edge set whose edges have endpoints
v and w if and only if vw ̸∈ E (G ). That is, V (G ) = V (G ) and
E (G ) = E (Kn ) − E (G ) where n = |V (G )|.
Note the graph complement is so-called because if a graph G has n vertices,
then setting the universal set of edges to be U = E (Kn ), we have that
E (G ) = (E (G ))c .
For example, a graph G and its complement G are:
a b a b

d c d c

Example. If G is a graph with n vertices and m edges, how many vertices

and edges does its complement graph G have?
Solution. We have |V (G )| = |V (G )| = n, and since G is a subgraph of Kn ,
|E (G )| = |E (Kn ) − E (G )| = |E (Kn )| − |E (G )| = n2 − m.

MATH1081 Topic 5 – Graph theory Lecture 5.01 – Graph and multigraph terminology 17/17
MATH1081 – Discrete Mathematics
Topic 5 – Graph theory
Lecture 5.02 – Vertex degree and isomorphisms

Lecturer: Dr Sean Gardiner – [email protected]

Vertex degrees
Definition. Given a (multi)graph G , the degree of any vertex v in G ,
written as deg(v ), is the number of times it is incident with an edge in G .
Notice that in a graph G , the degree of a vertex v is just the number of
edges incident with it. For a multigraph G , the degree of a vertex v is the
number of edges incident with it, except that loops must be counted twice.
Definition. A vertex of degree 0 is called an isolated vertex. A vertex of
degree 1 is called a pendant vertex.
Example. Describe the degrees of the vertices in the following graphs.
• Pn has 2 vertices of degree 1, and n − 2 vertices of degree 2.
• Cn has n vertices of degree 2.
• Kn has n vertices of degree n − 1.
• Km,n has m vertices of degree n, and n vertices of degree m.
• Qn has 2n vertices of degree n.

Definition. A regular (multi)graph is a (multi)graph whose vertices all have

the same degree. The graphs Cn , Kn , Kn,n , and Qn (for appropriate values
of n) are all examples of regular graphs.
MATH1081 Topic 5 – Graph theory Lecture 5.02 – Vertex degree and isomorphisms 1/11
Handshaking lemma
Theorem. (Handshaking lemma)
For any (multi)graph G , the sum of the degrees of all the vertices in G
equals twice the number of edges in G . That is, for any multigraph G ,
X
deg(v ) = 2 |E (G )|.
v ∈V (G )

Proof. Each edge in a multigraph G is either incident with exactly two

vertices, or in the case of a loop, exactly one vertex twice. Thus summing
the degrees of every vertex in G counts each edge in G exactly twice.
For example, check the vertex degree sums of the following multigraphs:

3 + 2 + 3 + 2 = 2 × 5. 2 + 3 + 2 + 2 + 1 = 2 × 5. 3 + 3 + 2 = 2 × 4.
Example. How many edges does Qn have?
Solution. We have seen that |V (Qn )| = 2n and that the degree of each
vertex in Qn is n. So by the handshaking lemma, we have
1 X 1
|E (Qn )| = deg(v ) = (n2n ) = n2n−1 .
2 2
v ∈V (Qn )
MATH1081 Topic 5 – Graph theory Lecture 5.02 – Vertex degree and isomorphisms 2/11
Degree sequences
Definition. A vertex degree sequence of a (multi)graph is a sequence of the
degrees of all vertices of the graph. It is usually written in increasing or
decreasing order.
Example. Does a graph or multigraph with the vertex degree sequence
3, 3, 2, 2, 2, 1 exist?
Solution. If such a multigraph existed, its vertex degree sum would be
3 + 3 + 2 + 2 + 2 + 1 = 13, which is odd. But by the handshaking lemma,
this total must be even, since it should equal twice the number of edges in
the multigraph. So no such multigraph exists.
Fact. No multigraph has a vertex degree sequence whose sum is odd.
Proof. This is a direct result of the handshaking lemma.
Fact. Every multigraph has an even number of vertices of odd degree.
Proof. If a multigraph had an odd number of vertices of odd degree, its
vertex degree sum would be odd, which is impossible.

MATH1081 Topic 5 – Graph theory Lecture 5.02 – Vertex degree and isomorphisms 3/11
Example – Degree sequences
Example. Does a graph or multigraph with the vertex degree sequence
4, 3, 3, 3, 3 exist?
Solution. One such graph (which is also a multigraph) is given below.

Example. Does a graph or multigraph with the vertex degree sequence

4, 4, 4, 2, 2 exist?
Solution. Label the vertices with degrees 4, 4, 4, 2, 2 as v1 , v2 , v3 , v4 , v5
respectively. If a graph with these vertices were to exist, we would require
each of v1 , v2 , and v3 to be adjacent to each of the other 4 vertices. This
would mean v4 and v5 would each have degree at least 3, which is greater
than their expected degree of 2. So no such graph exists.
However, a multigraph with the degree sequence does exist, for example:

MATH1081 Topic 5 – Graph theory Lecture 5.02 – Vertex degree and isomorphisms 4/11
Degree sequences for graphs and multigraphs
Fact. Not every vertex degree sequence whose sum is even has a
corresponding graph.
Fact. Every vertex degree sequence whose sum is even has a corresponding
multigraph.
Proof. Since the vertex degree sum is even, there must be an even number
of vertices of odd degree. Partition the vertices of odd degree into pairs, and
construct a single edge between each pair. All vertices now have degree that
differs from their expected degree by an even number. If a vertex has degree
differing from its expected degree by 2k for some integer k, construct k
loops at that vertex. Once completed, we have constructed a multigraph
with the required vertex degree sequence.
Example. Find multigraphs with the following degree sequences.
4, 4, 4, 2, 2. 5, 3, 3, 2, 1.

MATH1081 Topic 5 – Graph theory Lecture 5.02 – Vertex degree and isomorphisms 5/11
Graph isomorphism
We have seen that graphs can be represented diagrammatically in many
different ways, but are always considered the same graph. We might also
wish to treat two graphs that have the same shape but different vertex or
edge labellings as equivalent in some way, like the following examples:
a b p q w
x

d c s r y z
Definition. Two graphs G and H are isomorphic, written G ≃ H, if and
only if there exists a bijective function f : V (G ) → V (H) such that vertex
adjacency is preserved, that is, for all v , w ∈ V (G ), we have that
{v , w } ∈ E (G ) if and only if {f (v ), f (w )} ∈ E (H).
Equivalently, G ≃ H if and only if there exists a bijective function
f : V (G ) → V (H) such that V (H) = f (V (G )) and
E (H) = { {f (v ), f (w )} : {v , w } ∈ E (G ) } .
If two graphs are isomorphic, we call the function f a graph isomorphism.
Graph isomorphism is an equivalence relation on the set of all graphs.
MATH1081 Topic 5 – Graph theory Lecture 5.02 – Vertex degree and isomorphisms 6/11
Example – Isomorphic graphs
To prove that two graphs are isomorphic, we must provide a graph
isomorphism between them.
Example. Prove that the following two graphs are isomorphic.
a b w
x

d c y z
G1 G2

Solution. Consider the function f : V (G1 ) → V (G2 ) defined as follows.

vi ∈ V (G1 ) a b c d
f (vi ) y w z x
The edges of G1 under this mapping f are given by the following table.
vi vj ∈ E (G1 ) ab ac ad bc cd
f (vi )f (vj ) yw yz yx wz zx
Since the second row is precisely E (G2 ), we can conclude that f is a graph
isomorphism, and so G1 ≃ G2 .
MATH1081 Topic 5 – Graph theory Lecture 5.02 – Vertex degree and isomorphisms 7/11
Example – Isomorphic graphs
Example. Are the following two graphs isomorphic?
a b u v w

f c

e d x y z

G1 G2

Solution. Consider the function f : V (G1 ) → V (G2 ) defined as follows.

vi ∈ V (G1 ) a b c d e f
f (vi ) u x v y w z
The edges of G1 under this mapping f are given by the following table.
vi vj ∈ E (G1 ) ab bc cd de ef fa ad be cf
f (vi )f (vj ) ux xv vy yw wz zu uy xw vz
Since the second row is precisely E (G2 ), we can conclude that f is a graph
isomorphism, and so G1 ≃ G2 .
MATH1081 Topic 5 – Graph theory Lecture 5.02 – Vertex degree and isomorphisms 8/11
Graph invariants
To decide whether two graphs are isomorphic, we can try identifying
properties or features that are common to both graphs.
Definition. Any property of a graph G that is also a property of any graph
isomorphic to G is called a graph invariant.
Graph invariants include:
• Number of vertices; number of edges.
• Degree sequence; regularity; number of vertices of a particular degree.
• Degrees of adjacent vertices.
• Properties of subgraphs.
• Bipartiteness.
• Connectedness (see Lecture 5.03).
• Number of paths or cycles of a particular length (see Lecture 5.03).
• Existence of Euler or Hamilton walks (see Lecture 5.03).
• Planarity (see Lecture 5.04).

MATH1081 Topic 5 – Graph theory Lecture 5.02 – Vertex degree and isomorphisms 9/11
Example – Non-isomorphic graphs
To prove that two graphs are not isomorphic, we can identify an invariant
property that one graph has and the other does not.
Example. Prove that the following two graphs are not isomorphic.
b w

a c v x

e d z y
G1 G2

Solution. In G1 , the vertex b has degree 4, while in G2 , there is no vertex of

degree 4. Since vertex degrees are a graph invariant, we can conclude that
G1 and G2 are not isomorphic.
We could also have used the fact that G1 has two vertices of degree 2 while
G2 only has one, or that G1 has two vertices of degree 3 while G2 has four.

MATH1081 Topic 5 – Graph theory Lecture 5.02 – Vertex degree and isomorphisms 10/11
Example – Non-isomorphic graphs
Example. Are the following two graphs isomorphic?
b v w

a c

e x y z
d
G1 G2

Solution. In G1 , the only two vertices of degree 3 (a and c) are adjacent,

while in G2 , the only two vertices of degree 3 (v and w ) are not adjacent.
Since the degrees of adjacent vertices are a graph invariant, we can conclude
that G1 and G2 are not isomorphic.
Alternate solution. Notice that G2 is clearly bipartite (with vertex
partitions {v , w } and {x, y , z}). However, if G1 were bipartite with vertex
partitions V1 and V2 , supposing vertex b belonged to V1 , then both a and c
would have to belong to V2 since they are both adjacent to b, which is a
contradiction since a and c are adjacent. Since bipartiteness is a graph
invariant, we can conclude that G1 and G2 are not isomorphic.
MATH1081 Topic 5 – Graph theory Lecture 5.02 – Vertex degree and isomorphisms 11/11
MATH1081 – Discrete Mathematics
Topic 5 – Graph theory
Lecture 5.03 – Standard, Euler, and Hamilton walks

Lecturer: Dr Sean Gardiner – [email protected]

Walks
Definition. A walk of length n in a (multi)graph G is any alternating
sequence of vertices vi ∈ V (G ) and edges ei ∈ E (G ) of the form
v0 , e1 , v1 , e2 , v2 , ... , vn−1 , en , vn such that every pair of consecutive terms in
the sequence are incident (that is, every edge ei has endpoints vi−1 and vi ).
We say such a walk starts at v0 and ends at vn , or it travels from v0 to vn .
Definition. A walk in a (multi)graph G is closed if it starts and ends at the
same vertex (that is, if it travels from v0 to v0 ).
Notation. In a graph G , since there cannot be any parallel edges, instead of
representing a walk by the sequence v0 , e1 , v1 , e2 , v2 , ... , vn−1 , en , vn we can
use the shorthand v0 v1 v2 ... vn−1 vn . We can also denote a walk by
W = v0 v1 ... vn , and denote the reverse walk by W −1 = vn vn−1 ... v0 .
Example. Which sequences below represent walks in the given graph?
a b • abcda is a closed walk.
• abcdcda is a closed walk.
• abcd is a non-closed walk.
• abcadc is a non-closed walk.
d c • abcdba is not a walk.
MATH1081 Topic 5 – Graph theory Lecture 5.03 – Standard, Euler, and Hamilton walks 1/18
Connected graphs
Definition. A (multi)graph G is connected if and only if for every pair of
vertices vi , vj ∈ V (G ), there exists a walk from vi to vj .
Definition. The maximal connected subgraphs of any (multi)graph G are
called its connected components.
For example, the first graph below is connected, while the second is not.
They have 1 and 2 connected components respectively.

Theorem. For all (multi)graphs G , the relation ∼ on V (G ) defined by

vi ∼ vj if and only if there exists a walk from vi to vj is an equivalence
relation, and its equivalence classes are the connected components of G .
Proof. For all v ∈ V (G ), there exists a walk of length 0 from v to itself, so
v ∼ v and ∼ is reflexive. For all u, v ∈ V (G ), if u ∼ v then there is a walk
W from u to v , in which case W −1 is a walk from v to u, so v ∼ u and ∼
is symmetric. For all u, v , w ∈ V (G ), if u ∼ v and v ∼ w then there is a
walk W1 from u to v and a walk W2 from v to w , in which case W1 W2 is a
walk from u to w , so u ∼ w and ∼ is transitive.
MATH1081 Topic 5 – Graph theory Lecture 5.03 – Standard, Euler, and Hamilton walks 2/18
Adjacency matrices
Recall that a matrix is a rectangular array of entries (usually numbers), and
that the entry in the ith row and jth column is called the (i, j) entry of the
matrix. If the matrix is denoted by A = (ai,j ), then its (i, j) entry is ai,j .
Definition. For any finite (multi)graph G and particular ordering of its
vertices v1 , v2 , ... , vn , an adjacency matrix for G is the n × n matrix A for
which each (i, j) entry is the number of edges in G with endpoints vi and vj .

v1 v2  
1 0 1 1
0 0 0 0
For example, has adjacency matrix 
1
.
0 0 2
1 0 2 0
v4 v3
Notice that any adjacency matrix for a graph can only have entries that are
0 or 1 (since there are no parallel edges), and every entry on the main
diagonal must be 0 (since there are no loops).
Notice that an adjacency matrix A = (ai,j ) for any (multi)graph must be
symmetric, that is, ai,j = aj,i for all i, j. (An adjacency matrix for a directed
(multi)graph might not be symmetric.)
MATH1081 Topic 5 – Graph theory Lecture 5.03 – Standard, Euler, and Hamilton walks 3/18
Counting walks
Recall that the product of the n × n matrices A = (ai,j ) and B = (bi,j ) is the
matrix C = (ci,j ) whose entries are given by
n
X
ci,j = ai,k bk,j = ai,1 b1,j + ai,2 b2,j + · · · + ai,n bn,j .
k=1

Theorem. Given a (multi)graph G with adjacency matrix A = (ai,j ) for

vertex ordering v1 , v2 , ... , vn , the number of walks of length m from vi to vj
(m)
is the (i, j) entry of Am (sometimes written as ai,j ).
a b
Example. How many walks of length 3 are there from
a to a, and from a to b, in the following graph?
d c
Solution. Ordering
 the vertices
 as a, b, c, d, a corresponding
 adjacency

0 1 1 1 4 5 5 5
1 0 1 0
5 2 5 2
3

1 1 0 1, so we can find A = 5 5
matrix is A =   .
4 5
1 0 1 0 5 2 5 2
The (1, 1) entry of A3 is 4, so there are 4 different walks of length 3 from a
(3)
to a, and since a1,2 = 5, there are 5 different walks of length 3 from a to b.
MATH1081 Topic 5 – Graph theory Lecture 5.03 – Standard, Euler, and Hamilton walks 4/18
Counting walks – Proof
Theorem. Given a (multi)graph G with adjacency matrix A = (ai,j ) for
vertex ordering v1 , v2 , ... , vn , the number of walks of length m from vi to vj
(m)
is the (i, j) entry of Am (sometimes written as ai,j ).
Proof. Let P(m) be the statement “The (i, j) entry of Am is the number of
walks of length m from vi to vj ” for any m ∈ N.
When m = 0, we have Am = A0 = I , the identity matrix whose main
diagonal entries are each 1 and whose all other entries are 0. Since there is
exactly one walk of length 0 from a vertex to itself, and no walks of length 0
between two distinct vertices, we can conclude that P(0) is true.
Next assume that P(k) is true, so that the (i, j) entry of Ak is the number
of walks of length k from vi to vj . Consider the matrix Ak+1 = (Ak )A. Then
(k+1) (k) (k) (k)
ai,j = ai,1 a1,j + ai,2 a2,j + · · · + ai,n an,j .
(k)
By the inductive hypothesis, each term ai,p ap,j counts the number of walks
of length k from vi to vp times the number of walks of length 1 from vp to
vj . This is precisely the number of walks of length k + 1 from vi to vj where
the penultimate vertex is vp , so taking the sum of these terms for all
possible vp must give the total number of walks of length k + 1 from vi to
vj . Thus P(k) ⇒ P(k + 1), and so P(m) is true for all m ∈ N by induction.
MATH1081 Topic 5 – Graph theory Lecture 5.03 – Standard, Euler, and Hamilton walks 5/18
Trails, tours, paths, cycles
Definition. In any (multi)graph:
• A trail is any non-closed walk with no repeated edges.
• A tour is any closed walk with no repeated edges.
• A path is any trail with no repeated vertices, that is, a non-closed walk
with no repeated (edges or) vertices.
• A cycle is any tour with no repeated vertices other than its first and last
vertices, that is, a closed walk with no repeated edges or vertices (except
for its first/last vertex).
We can think of a trail or tour as a walk that can only intersect with itself at
vertices, and a path or cycle as a walk that cannot intersect with itself
anywhere (except at its start and end vertex in the case of a cycle).
Example. Classify each of the below walks in the given graph.
a b • abcde is a path (and a trail).
• aecbed is a trail (but not a path).
• abcdea is a cycle (and a tour).
e
• aecbeda is a tour (but not a cycle).
d c • abedceb is a walk (but not a trail or tour).
MATH1081 Topic 5 – Graph theory Lecture 5.03 – Standard, Euler, and Hamilton walks 6/18
Proofs involving paths/cycles
Theorem. In a (multi)graph G , if there is a non-closed walk from vertices
v0 to vn , then there is a path from vertices v0 to vn .
Proof. Suppose there is a walk v0 , e1 , v1 , e2 , v2 , ... , vn−1 , en , vn in G . If none
of the vertices are repeated, then the walk is already a path. Otherwise,
suppose vi = vj for some i < j. Then by removing all steps in the walk
between vi and vj , we obtain a smaller walk v0 , ... , vi−1 , ei , vi , ej , vj+1 , ... , vn ,
which is valid since vi = vj is an endpoint of ej .
This process can be repeated until there are no repeated vertices remaining
in the walk, reducing the walk to a path from v0 to vn .
Corollary. In a (multi)graph G with n vertices, if there is a non-closed walk
from vertices vi to vj , then there is a walk of length at most n − 1 from vi
to vj .
Proof. If there is a non-closed walk from vi to vj , then there is a path from
vi to vj , which has no repeated vertices and therefore at most n vertices and
n − 1 edges.

MATH1081 Topic 5 – Graph theory Lecture 5.03 – Standard, Euler, and Hamilton walks 7/18
Cycles in bipartite graphs
Theorem. A (multi)graph G is not bipartite if and only if it contains a cycle
of odd length.
Equivalently, a (multi)graph G is bipartite if and only if every cycle in G has
even length.
Proof of forward implication. First suppose G is a bipartite multigraph
with vertex set partitions V1 and V2 . Then any walk in G must alternate
vertices between V1 and V2 . So for any cycle in G , since it must start and
end in the same vertex partition, its length must be even.

MATH1081 Topic 5 – Graph theory Lecture 5.03 – Standard, Euler, and Hamilton walks 8/18
Cycles in bipartite graphs
Proof of backward implication. Now suppose that G is a multigraph for
which every cycle in G has even length. For simplicity, we shall assume G is
connected – if it is not, a similar approach can be applied to each connected
component of G .
Choose some vertex w in G , and construct the sets
V1 = {v ∈ V (G ) : there is a path in G from v to w of even length},
V2 = {v ∈ V (G ) : there is a path in G from v to w of odd length}.
Certainly V1 ∪ V2 = V (G ), since every vertex must be either an odd or even
distance from a particular vertex in a connected graph. We also have that
V1 ∩ V2 = ∅, since if there were some vertex v ∈ V1 ∩ V2 , then there would
exist a path P1 from v to w of even length, and a path P2 from v to w of
odd length, meaning the walk P1 P2−1 would be a cycle of odd length, which
is not possible in G . So V1 and V2 partition V (G ).
Furthermore, if two vertices u, v ∈ V1 or u, v ∈ V2 were adjacent, the walk
w ... uv ... w would be a cycle of odd length, which again is not possible in
G . So no two vertices both in V1 or both in V2 are adjacent, meaning G is
bipartite.

MATH1081 Topic 5 – Graph theory Lecture 5.03 – Standard, Euler, and Hamilton walks 9/18
Euler walks
Definition. In any connected (multi)graph G , an Euler trail is a trail that
visits each edge of G exactly once, and an Euler tour is a tour that visits
each edge of G exactly once.
For example, consider the following graphs:
b a b a b
a c
e e
e d d c d c
G1 G2 G3
• The graph G1 has an Euler tour, for example abcdeacebda.
• The graph G2 has an Euler trail, for example baedcebc.
• The graph G3 does not have an Euler walk.

To prove the existence of an Euler walk, we can simply construct one. To

prove the non-existence of an Euler walk, we shall establish some necessary
and sufficient conditions.
MATH1081 Topic 5 – Graph theory Lecture 5.03 – Standard, Euler, and Hamilton walks 10/18
Existence of Euler tours
Theorem. A connected (multi)graph G contains an Euler tour if and only if
every vertex has even degree.
Proof of forward implication. First suppose G contains an Euler tour.
Then the tour starts and ends at some vertex v0 , and for each vertex vi that
the tour passes through, it must enter that vertex the same number of times
it exits it. So the number of times the tour enters or exits each vertex must
be an even number, and since the tour passes over every edge of G , this
number must also be the vertex’s degree. So every vertex has even degree.
Proof of backward implication. First note that it is sufficient to prove the
statement only for multigraphs that contain no loops, since if a multigraph
G becomes the multigraph G ′ upon removal of all loop edges, G has n
vertices all of even degree if and only if G ′ does (since each loop edge adds
2 to its endpoint vertex’s degree), and G has an Euler tour if and only if G ′
does (since each loop e can be trivially included in a tour at its endpoint
vertex v by inserting the walk vev ).
Let P(n) be the statement “Any connected multigraph with n vertices all of
even degree contains an Euler tour” for any n ∈ Z+ . The only multigraph
with 1 vertex and no loops has no edges, so it vacuously satisfies the
conditions for containing an Euler tour. So P(1) is true.
MATH1081 Topic 5 – Graph theory Lecture 5.03 – Standard, Euler, and Hamilton walks 11/18
Existence of Euler tours (continued)
Proof of backward implication (cont’d).
Now suppose that P(1), P(2), P(3), ... , P(k) are all true for some k ∈ Z+ .
Consider an arbitrary connected multigraph G with k + 1 vertices all of even
degree (with no loops). Choose a vertex w in G and consider all the edges
incident with it. Since w has even degree, it must have an even number of
′
incident edges, so they can be arbitrarily labelled e1 , e1′ , e2 , e2′ , ... , em , em for
some m ∈ N. For each edge ei , we label its other endpoint vi , and similarly
label the other endpoint of each ei′ as vi′ .
Now consider the graph H formed by removing the vertex w and replacing
each pair of edges ei and ei′ with a new edge ei′′ whose endpoints are vi and
vi′ . Constructing this new graph H does not change the degrees of any of
the other vertices in G , so H is a graph with k vertices all of even degree.
By the inductive hypotheses, H must contain an Euler tour in each of its
connected components. We can convert each Euler tour in H into a tour in
G by replacing each instance of vi ei′′ vi′ (or vi′ ei′′ vi ) in the tour with vi ei wei′ vi′
(or vi′ ei′ wei vi ). So we can construct an Euler tour in G by starting at w and
tracing out the Euler tour of each connected component of H, revisiting w
between each tour.
Thus P(k + 1) is true, and so P(n) is true for all n ∈ Z+ by strong induction.
MATH1081 Topic 5 – Graph theory Lecture 5.03 – Standard, Euler, and Hamilton walks 12/18
Existence of Euler trails
Theorem. A connected (multi)graph G contains an Euler trail if and only if
exactly two vertices have odd degree.
Proof. Suppose G contains an Euler trail with start vertex u and end
vertex v . Then by adding an edge between u and v , we can create a new
graph G ′ which has an Euler tour. By the previous theorem, every vertex of
G ′ has even degree, so in G every vertex must have had even degree except
for u and v (since their degrees are each increased by 1 when the extra edge
is added).
Next suppose G has exactly two vertices of odd degree, labelled u and v .
Then by adding an edge between u and v , we can create a new graph G ′ in
which every vertex has even degree (since the degrees of u and v are each
increased by 1 when the extra edge is added). By the previous theorem, G ′
must have an Euler tour, and so G must have an Euler trail that starts at u
and ends at v .

MATH1081 Topic 5 – Graph theory Lecture 5.03 – Standard, Euler, and Hamilton walks 13/18
Example – Euler walks
Example. Decide whether each graph below has an Euler trail or tour.
a b a b a b

f c f c f c
g g g

e d e d e d
G1 G2 G3

Solution. Since every vertex of G1 has even degree, an Euler tour exists in
G1 . For example, the tour abgcdgefgacea.
Since exactly two vertices of G2 have odd degree (d and e), an Euler trail
exists in G2 , and its start and end points must be at d and e in some order.
For example, the trail degdcgbagfe.
Since more than two vertices of G3 have odd degree (a, b, c, and f ), no
Euler path exists in G3 .

MATH1081 Topic 5 – Graph theory Lecture 5.03 – Standard, Euler, and Hamilton walks 14/18
Hamilton walks
Definition. In any connected (multi)graph G , a Hamilton path is a path
that visits each vertex of G exactly once, and a Hamilton cycle is a cycle
that visits each vertex of G exactly once (except for its first and last vertex).
For example, consider the following graphs:
b a b a b
a c
f c f c
g g
e d e d e d
G1 G2 G3
• The graph G1 has a Hamilton path, for example abcde, and a Hamilton
cycle, for example abcdea.
• The graph G2 has a Hamilton path, for example bacefgd, but no
Hamilton cycle.
• The graph G3 does not have a Hamilton walk.
To prove the existence of a Hamilton walk, we can simply construct one. It
is generally difficult to prove non-existence of a Hamilton walk, but we can
still establish some useful partial results.
MATH1081 Topic 5 – Graph theory Lecture 5.03 – Standard, Euler, and Hamilton walks 15/18
Properties of Hamilton walks
Unlike for Euler walks, there is no known necessary and sufficient condition
for a graph to contain a Hamilton walk. The following partial results are
provided without proof.
Fact. If any vertex v in a connected (multi)graph G has degree 1, there is
no Hamilton cycle, and if there is a Hamilton path, it must start or end at v .
Fact. If any vertex v in a connected (multi)graph G has degree 2, then
both edges incident with v must be part of a Hamilton cycle if it exists.
Fact. If a Hamilton cycle exists for a (multi)graph G , every vertex of G
must be incident with exactly 2 edges that are used in the walk. If a
Hamilton path exists for a (multi)graph G , exactly two vertices of G must be
incident with exactly 1 edge that is used in the walk, and all other vertices
of G must be incident with exactly 2 edges that are used in the walk.
Theorem. (Dirac) If a graph G with n ≥ 3 vertices has the property that
every vertex has degree at least n2 , then G contains a Hamilton cycle.
Theorem. (Ore) If a graph G with n ≥ 3 vertices has the property that
every pair of non-adjacent vertices has degree sum at least n, then G
contains a Hamilton cycle.
MATH1081 Topic 5 – Graph theory Lecture 5.03 – Standard, Euler, and Hamilton walks 16/18
Example – Hamilton walks
Example. Prove that G2 cannot have a Hamilton cycle.
a b

f c
g
e d
G2

Solution. Suppose a Hamilton cycle exists for G2 . Any Hamilton cycle must
pass through both edges incident with each vertex of degree 2. Since b has
degree 2, edges ab and bg must be included in the Hamilton cycle. Since d
has degree 2, edges dc and dg must be included in the Hamilton cycle.
Since f has degree 2, edges fe and fg must be included in the Hamilton
cycle. But this means g is incident with more than 2 edges in the Hamilton
cycle, which is impossible. So no Hamilton cycle exists for G2 .

MATH1081 Topic 5 – Graph theory Lecture 5.03 – Standard, Euler, and Hamilton walks 17/18
Example – Hamilton walks
Example. Prove that G3 cannot have a Hamilton walk.
a b

f c
g
e d
G3

Solution. Suppose a Hamilton walk exists for G3 . There are 6 vertices of

degree 2, and any Hamilton walk must pass through both edges incident
with at least 4 of these vertices. By the pigeonhole principle (with the 3
pigeonholes {a, b}, {c, d}, and {e, f }), at least 2 of these vertices must be
adjacent, for example a and b. But then the walk must include edges ab,
ag , and bg , which forms a cycle that does not visit every vertex of G3 . So
there cannot exist a Hamilton path nor a Hamilton cycle.

MATH1081 Topic 5 – Graph theory Lecture 5.03 – Standard, Euler, and Hamilton walks 18/18
MATH1081 – Discrete Mathematics
Topic 5 – Graph theory
Lecture 5.04 – Planar graphs and maps

Lecturer: Dr Sean Gardiner – [email protected]

Planar graphs
Definition. A (multi)graph G is called planar if it has a diagrammatic
representation in the 2-dimensional plane in which no edges cross.
Definition. If a (multi)graph G is planar, a representation of G with no
edges crossing is called a planar map or planar representation of G .
For example, all graphs below are representations of K4 , but only the second
and third representations are planar maps of K4 (since in the first
representation, two of the edges cross each other).

Example. Is K2,3 planar?

Solution. The graph K2,3 is planar, as demonstrated by
the planar map provided on the right (each of the 2 red
vertices is adjacent to each of the 3 blue vertices, and no
edges cross).

MATH1081 Topic 5 – Graph theory Lecture 5.04 – Planar graphs and maps 1/12
Regions
Definition. In any planar map of a planar (multi)graph G , the plane is
divided into regions (often labelled R1 , R2 , R3 , ... ) whose boundaries are the
edges of G . Exactly one of these regions is unbounded, in the sense that it
extends infinitely outside of the planar map.
For example, each of the planar multigraphs below has 3 regions labelled R,
S, and T , and in both cases T is the unbounded region.

R
S
T T S
R

Definition. Given a planar (multi)graph G and a planar map of G , if an

edge e ∈ E (G ) lies on the boundary for a region R in the planar map, then
we say e is incident with R and R is incident with e.
Definition. Given a planar (multi)graph G and a planar map of G , two
regions R and S in the planar map are said to be adjacent or neighbours if
there is an edge in G that they are both incident with. For example, in the
second graph, T is adjacent to R and to S, but R and S are not adjacent.
MATH1081 Topic 5 – Graph theory Lecture 5.04 – Planar graphs and maps 2/12
Region degrees
Definition. Given a planar (multi)graph G and a planar map of G , the
degree of any region R in the planar map, written as deg(R), is the number
of times it is incident with an edge in the planar map.
One way to evaluate the degree of a region is to count the number of times
an edge is passed when traversing its boundary.
Example. Find the degrees of the regions in the following multigraphs.
S1
R2
R3 S3 S2
R1

Solution. We have deg(R1 ) = deg(R2 ) = 3, deg(R3 ) = 4, deg(S1 ) = 1,

deg(S2 ) = 2, and deg(S3 ) = 5.
Example. Describe the degrees of the regions in the following graphs.
• Cn has 2 regions of degree n.
• Pn has 1 region of degree 2(n − 1).
• K4 has 4 regions of degree 3.
• K5 is not planar (to be proven!), so it does not have well-defined regions.
MATH1081 Topic 5 – Graph theory Lecture 5.04 – Planar graphs and maps 3/12
Handshaking lemma
Notice that the region degree sequence for a graph might be different
depending on the planar map chosen. For example, the two graphs below
are different planar maps of the same planar graph, but the planar maps
have region degree sequences 3, 7 and 5, 5 respectively.

However, regardless of the planar map chosen, we have an analogue for the
handshaking lemma on degrees of regions.
Theorem. (Handshaking lemma for regions)
For any planar (multi)graph G and planar map of G , the sum of the degrees
of all the regions in the planar map equals twice the number of edges in G .
That is, for any planar multigraph G and planar map of G ,
X
deg(R) = 2 |E (G )|.
regions R

Proof. Each edge in a multigraph G is either incident with exactly two

regions, or in the case of an edge that contributes twice to a region’s
boundary, exactly one region twice. Thus summing the degrees of every
region in a planar map of G counts each edge in G exactly twice.
MATH1081 Topic 5 – Graph theory Lecture 5.04 – Planar graphs and maps 4/12
Euler’s formula
Lemma. Every connected (multi)graph with at least one edge either has a
pendant vertex or contains a cycle (of positive length).
Proof. See Lecture 5.05.
We shall use this lemma to help prove an important result relating the
number of regions in a planar map to the number of vertices and edges.
Theorem. (Euler’s formula)
Suppose a (multi)graph G is connected and planar, with v ≥ 1 vertices, e
edges, and r regions in a planar map of G . Then r − e + v = 2.
Proof. We shall induct on the number of edges e. Let P(e) be the
statement “For any connected planar multigraph G with e edges, v vertices,
and r regions in a planar map of G , we have r − e + v = 2” for any e ∈ N.
When e = 0, since G is connected, it can only have one vertex, so v = 1
and r = 1. Then r − e + v = 1 − 0 + 1 = 2, so P(0) is true.
Now suppose P(k) is true for some k ∈ N, so that Euler’s formula holds for
all connected planar graphs with k edges. Consider an arbitrary connected
planar multigraph G with eG = k + 1 edges, vG vertices, and rG regions in a
planar map of G . We wish to show that P(k + 1) holds for G , that is, that
rG − eG + vG = 2.
MATH1081 Topic 5 – Graph theory Lecture 5.04 – Planar graphs and maps 5/12
Euler’s formula
Theorem. (Euler’s formula)
Suppose a (multi)graph G is connected and planar, with v ≥ 1 vertices, e
edges, and r regions in a planar map of G . Then r − e + v = 2.
Proof (cont’d). As per the earlier lemma, first suppose G has a pendant
vertex. Then we can remove this pendant vertex and its incident edge to
create a new connected planar graph G ′ with eG − 1 = k edges, rG regions,
and vG − 1 vertices. Then since P(k) is true for G ′ , we know
2 = rG − (eG − 1) + (vG − 1) = rG − eG + vG .
So Euler’s formula holds for G , meaning P(k + 1) is true in this case.
Next suppose G has no pendant vertex, so that by the earlier lemma it must
contain a non-trivial cycle. Then we can remove one of the edges from this
cycle to create a new connected planar graph G ′ with eG − 1 = k edges,
rG − 1 regions, and vG vertices. Then since P(k) is true for G ′ , we know
2 = (rG − 1) − (eG − 1) + vG = rG − eG + vG .
So again Euler’s formula holds for G , meaning P(k + 1) is true in this case.
Thus in all possible cases we have P(k) ⇒ P(k + 1), and so P(e) is true for
all e ∈ N by induction.
MATH1081 Topic 5 – Graph theory Lecture 5.04 – Planar graphs and maps 6/12
Planarity conditions
Theorem. For any connected planar graph G with v ≥ 3 vertices and e
edges, we have e ≤ 3v − 6.
Proof. Let G be a connected planar graph with v ≥ 3 vertices, e edges, and
r regions in a planar map of G . Since G is a graph, it contains no loops,
which means in a planar map of G , there can be no regions of degree 1.
Similarly, since G is a graph, it contains no parallel edges, which means in a
planar map of G , there can be no regions of degree 2. Hence every region R
in a planar map of G has degree at least 3, so the sum of all region degrees
in a planar map of G is Pat least 3r . Using the handshaking lemma for
regions, we find 2e = R deg(R) ≥ 3r = 3(e − v + 2) by Euler’s formula.
This rearranges to give e ≤ 3v − 6 as required.
Theorem. For any connected planar graph G with v ≥ 3 vertices and e
edges which contains no cycles of length 3, we have e ≤ 2v − 4.
Proof. Again since G is a graph, a planar map of G contains no regions of
degree 1 or 2, nor 3 in this case sinceP
it contains no cycles of length 3. So
by a similar argument, we have 2e = R deg(R) ≥ 4r = 4(e − v + 2) by
Euler’s formula, which rearranges to give e ≤ 2v − 4 as required.

MATH1081 Topic 5 – Graph theory Lecture 5.04 – Planar graphs and maps 7/12
Example – Planarity conditions
Example. Show that the graph below is not planar.
a b

f c
g
e d
Solution. The graph is connected, so if it were planar we would require that
e ≤ 3v − 6. However here the graph has v = 7 vertices and e = 16 edges,
so we have 3v − 6 = 15 ̸≥ 16 = e, meaning the graph cannot be planar.
Example. Show that K5 is not planar.

We know that K5 is a connected graph with v = 5 vertices and
Solution.
e = 52 = 10 edges. Since 3v − 6 = 9 ̸≥ 10 = e, we can conclude that K5 is
not planar.
Example. Show that K3,3 is not planar.
Solution. We know that K3,3 is a connected graph with 3 + 3 = 6 vertices
and 3 × 3 = 9 edges. Because K3,3 is bipartite, it has no cycles of odd
length, and so in particular has no cycles of length 3. So since
2v − 4 = 8 ̸≥ 9 = e, we can conclude that K3,3 is not planar.
MATH1081 Topic 5 – Graph theory Lecture 5.04 – Planar graphs and maps 8/12
Kuratowski’s theorem
Definition. A subdivision of a (multi)graph G is a new multigraph G ′ that
has had additional vertices inserted by replacing some edges with
non-intersecting paths of arbitrary length. For example, G1 and G2 below are
subdivisions of G , but G3 is not a subdivision of G .

G G1 G2 G3
We can create any subdivision of a graph G by repeating the following
process: remove an edge uw from E (G ), add a new vertex v to V (G ), and
add new edges uv and vw to E (G ).
Clearly a (multi)graph G is planar iff any subdivision of G is planar.
Theorem. (Kuratowski)
A (multi)graph is planar if and only if it does not contain a subgraph which
is a subdivision of K5 or K3,3 .
While the forward implication is obvious, the backward implication is very
difficult to prove, and so for this course its proof is not provided.
MATH1081 Topic 5 – Graph theory Lecture 5.04 – Planar graphs and maps 9/12
Example – Kuratowski’s theorem
Example. Decide whether the graph below is planar.
a b a b
c h
h
has subgraph g c
g d
d
f e f e
Solution. Consider the subgraph with the edge dh removed. This is a
subdivision of K3,3 , with additional vertices d and h. So the graph is not
planar by Kuratowski’s theorem.
Example. Decide whether the graph below is planar.
a b g
a b
has subgraph f c
f c
g
e d e d
Solution. Consider the subgraph with the edges ag and bg removed. This
is a subdivision of K5 , with additional vertices a and b. So the graph is not
planar by Kuratowski’s theorem.
MATH1081 Topic 5 – Graph theory Lecture 5.04 – Planar graphs and maps 10/12
Dual maps
Definition. Given a planar (multi)graph G and a planar map of G , we can
construct its dual planar map of a multigraph G ∗ by doing the following:
• For each region R in the planar map of G , create a corresponding vertex
vR in the planar map of G ∗ .
• For each vertex v in the planar map of G , create a corresponding region
Rv in the planar map of G ∗ .
• For each edge e in the planar map of G that is incident with regions R
and S and incident with vertices v and w , create a corresponding edge
e ∗ in the planar map of G ∗ that is incident with vertices vR and vS and
incident with regions Rv and Rw .

a b Ra Rb vT
vS
S
T vR vT Rb Rd Ra
R Rc
d c Rd Rc vS vR
∗
G G

MATH1081 Topic 5 – Graph theory Lecture 5.04 – Planar graphs and maps 11/12
Properties of dual maps
Example. Find the dual planar map of the planar map below.
a b vR R a Rb
R Rb
vS vR vT vS
T S Ra Rc
vT

c Rc
G G∗
Fact. Given a planar (multi)graph G and a planar map of G with v vertices,
e edges, and r regions, its dual planar map has r vertices, e edges, and v
regions.
Fact. Given a planar (multi)graph G and a planar map of G , two vertices v
and w are adjacent if and only if their corresponding regions Rv and Rw in
the dual planar map are adjacent.
Fact. Given a planar (multi)graph G , the vertex degree sequence of G is the
same as the region degree sequence for any dual planar map of G ∗ , and the
vertex degree sequence of any dual graph G ∗ is the same as the region
degree sequence for some planar map of G .
MATH1081 Topic 5 – Graph theory Lecture 5.04 – Planar graphs and maps 12/12
MATH1081 – Discrete Mathematics
Topic 5 – Graph theory
Lecture 5.05 – Trees and weighted graphs

Lecturer: Dr Sean Gardiner – [email protected]

Trees
Definition. A tree is a connected graph which contains no cycles (of
positive length).
For example, G1 and G2 below are trees, while G3 and G4 are not.

G1 G2 G3 G4
Theorem. Any tree with at least one edge contains at least 2 pendant
vertices.
Proof. Suppose G is a tree with at least one edge. Since the edge cannot
be a loop, G must contain a path between two distinct vertices. Find a path
P = v0 v1 ... vn of maximal length in G . Clearly v0 has degree at least 1.
Suppose v0 has degree greater than 1. Then v0 is adjacent to some other
vertex u in G . If u is not part of P, then we can extend P to make a longer
path uv0 v1 ... vn , which contradicts the maximality of P. If u is part of P,
then v0 v1 ... uv0 is a cycle, which contradicts that G is a tree and must
contain no non-trivial cycles. So deg(v0 ) = 1, and by a similar argument we
also have that deg(vn ) = 1, meaning G contains at least 2 pendant vertices.
MATH1081 Topic 5 – Graph theory Lecture 5.05 – Trees and weighted graphs 1/11
Number of edges in a tree
Theorem. A connected nonempty multigraph G is a tree if and only if we
have |E (G )| = |V (G )| − 1.
Proof of forward implication. Let P(n) be the statement “Any tree with n
vertices has n − 1 edges”.
The only tree with 1 vertex is the graph with 1 vertex and 0 edges, since it
cannot contain any loops (which are cycles), so P(1) is true.
Next suppose P(k) is true for some positive integer k, and consider a tree G
with k + 1 vertices. Because G is a tree, it contains at least two pendant
vertices, so consider the new graph G ′ formed by removing a pendant vertex
and its incident edge from G . Then G ′ has k vertices, and is also a tree
since it is still connected and contains no cycles. So by the inductive
hypothesis, G ′ has k − 1 edges, meaning the original graph G has k edges.
Thus we have shown P(k) ⇒ P(k + 1), meaning P(n) is true for all n ∈ Z+
by induction.

MATH1081 Topic 5 – Graph theory Lecture 5.05 – Trees and weighted graphs 2/11
Number of edges in a tree
Theorem. A connected nonempty multigraph G is a tree if and only if we
have |E (G )| = |V (G )| − 1.
Proof of backward implication. Suppose G is a connected multigraph
with |E (G )| = |V (G )| − 1. Further suppose that G contains a cycle. Then
we may remove this cycle by deleting an edge in the cycle, which reduces the
number of edges by 1 but does not change the number of vertices and does
not disconnect the graph. Repeating this procedure for all cycles in G , we
eventually reach a new graph G ′ that is connected and has no cycles, and is
therefore a tree. By the forward implication proof, we must have
|E (G ′ )| = |V (G ′ )| − 1. But then we have

|E (G )| > |E (G ′ )| = |V (G ′ )| − 1 = |V (G )| − 1,

which contradicts the fact that |E (G )| = |V (G )| − 1. So G cannot contain

a cycle, and is therefore a tree.

MATH1081 Topic 5 – Graph theory Lecture 5.05 – Trees and weighted graphs 3/11
Spanning trees
Definition. A spanning tree of a (multi)graph G is a tree which is subgraph
of G and which contains all vertices of G . That is, a spanning tree T of a
multigraph G is a tree for which V (T ) = V (G ) and E (T ) ⊆ E (G ).
For example, the following blue graphs are all spanning trees of K4 .

Theorem. Every connected multigraph G has a spanning tree.

Proof. Suppose G is a connected multigraph. For each cycle in G , we can
remove it by deleting an edge from the cycle, which does not disconnect the
graph. Once all cycles are removed, we are left with a new graph G ′ that
contains all vertices of G , is connected, and has no cycles. So G ′ is a
spanning tree of G .

MATH1081 Topic 5 – Graph theory Lecture 5.05 – Trees and weighted graphs 4/11
Weighted graphs
Definition. A weighted (multi)graph is a multigraph whose edges are
associated with numbers called weights. Each edge is assigned a particular
number, and typically these numbers are positive and rational or natural.
In general, we can say a weighted multigraph G is a multigraph equipped
with a weight function w : E (G ) → R.
Notation. The weight of an edge e is denoted w (e).
Definition. The weight of a (multi)graph G , sometimes denoted w (G ), is
the sum of all its edge weights. That is,
X
w (G ) = w (e).
e∈E (G )

In a diagrammatic representation of a weighted multigraph, the weight of

each edge is indicated by writing a number near the edge. For example, each
multigraph below has weight 15.
1 4
3
4 5 2 3
5
3
MATH1081 Topic 5 – Graph theory Lecture 5.05 – Trees and weighted graphs 5/11
Minimal spanning trees
Definition. A minimal spanning tree of a weighted (multi)graph G is a
spanning tree T of G with minimal weight w (T ).
Note that for any weighted (multi)graph G , there may not be a unique
minimal spanning tree.

2 3 1
3 5
4
3 2

For example, the graph above has two minimal spanning trees, each with
weight 8:

2 1 2 3 1
3

2 2

MATH1081 Topic 5 – Graph theory Lecture 5.05 – Trees and weighted graphs 6/11
Kruskal’s algorithm
A procedure for finding a minimal spanning tree for any connected weighted
(multi)graph G is Kruskal’s algorithm:
• Begin with a subgraph T with V (T ) = V (G ) and E (T ) = {}.
• Sort the edges in G by increasing weight order.
• For each edge in the ordered sequence, add the edge to E (T ) if its
inclusion will not introduce a cycle in T .
• Once |E (T )| = |V (G )| − 1, the process ends and T is a minimal
spanning tree of G .
Kruskal’s algorithm can be carried out efficiently by filling out a table like
the following, with the edges ordered by increasing weight:

edge ei weight w (ei ) included?

.. .. ..
. . .

MATH1081 Topic 5 – Graph theory Lecture 5.05 – Trees and weighted graphs 7/11
Example – Kruskal’s algorithm
Example. Find a minimal spanning tree for G . a 6 b 3 c
Solution. Applying Kruskal’s algorithm gives: 6 14
5 2
edge ei weight w (ei ) included? 1 5
d f
be 1 yes
3 2 e 2
de 1 yes 4 4
hi 1 yes
cf 2 yes g 5 h 1 i
eh 2 yes
ei 2 no (ehie is a cycle)
G
bc 3 yes
eg 3 yes
ce 4 no (bceb is a cycle)
dg 4 no (degd is a cycle) a b 3 c
fi 4 no (bcfiheb is a cycle)
ad 5 yes 5 1 2
ef 5 no (8 edges have been selected) 1 e
d f
gh 5 no (8 edges have been selected)
ab 6 no (8 edges have been selected) 3 2
ae 6 no (8 edges have been selected)
g h 1 i
Thus a minimal spanning tree (with weight 18) is:
Note that if we had ordered the edges by weight differently, we might have
included ei instead of eh and found a different minimal spanning tree.
MATH1081 Topic 5 – Graph theory Lecture 5.05 – Trees and weighted graphs 8/11
Minimal v -path spanning trees
Definition. Given a weighted (multi)graph G and vertices u, v ∈ V (G ), a
shortest path from u to v is a path P from u to v in G with minimal weight
w (P). We call the weight of this shortest path the distance from u to v ,
and write d(u, v ) = w (P).
Definition. Given a weighted (multi)graph G and a vertex v ∈ V (G ), a
minimal v -path spanning tree is a spanning tree T of G in which for every
vertex u ∈ V (G ), the unique path from v to u in T is the shortest path
from v to u in G .
Note that for any weighted (multi)graph G and vertex v in G , a minimal
v -path spanning tree of G is usually not a minimal spanning tree of G . For
example, the graph G on the left has a minimal z-path spanning tree shown
on the right. w w
2 3 1 3 1
u 3 5 x u 3 x
z z
4 4
3 2
y y
So in G , we know d(z, u) = 3, d(z, w ) = 3, d(z, x) = 4, and d(z, y ) = 4.
MATH1081 Topic 5 – Graph theory Lecture 5.05 – Trees and weighted graphs 9/11
Dijkstra’s algorithm
A procedure for finding a minimal v -path spanning tree for any connected
weighted (multi)graph G and vertex v ∈ V (G ) is Dijkstra’s algorithm:
• Begin with a tree T with V (T ) = {v } and E (T ) = {}.
• Consider all edges ei ∈ E (G ) that have one endpoint ui ∈ V (T ) and one
endpoint wi ̸∈ V (T ). Out of all these edges, choose the one for which
d(v , ui ) + w (ei ) is minimal, and add ei to E (T ) and wi to V (T ).
• Repeat the previous step until V (T ) = V (G ), at which point the process
ends and T is a minimal v -path spanning tree of G .
Dijkstra’s algorithm can be carried out efficiently by filling out a table like
the following:

possible edges ui wi included included distance

(weight d(v , ui ) + w (ui wi )) edge ui wi vertex wi d(v , wi )
v 0
.. .. .. ..
. . . .

MATH1081 Topic 5 – Graph theory Lecture 5.05 – Trees and weighted graphs 10/11
Example – Dijkstra’s algorithm
Example. Find a minimal a-path spanning tree for G . a 2 b 4 c
Solution. Applying Dijkstra’s algorithm gives: 4 14
3 1
possible edges ui wi included included distance 1 3
d f
(weight d(a, ui ) + w (ui wi )) edge ui wi vertex wi d(a, wi )
3 4 e 2
a 0 2 2
ab (2), ad (3), ae (4) ab b 2
ad (3), ae (4), bc (6), be (3) ad d 3 g 2 h 3 i
ae (4), bc (6), be (3), de (4), be e 3 G
dg (5)
bc (6), dg (5), ec (7), ef (6), dg g 5
eg (6), eh (7), ei (5)
bc (6), ec (7), ef (6), eh (7), ei i 5
ei (5), gh (7)
a 2 b 4 c
bc (6), ec (7), ef (6), eh (7), bc c 6
gh (7), if (7), ih (8) 3 1
e 3
cf (7), ef (6), eh (7), gh (7), ef f 6 d f
if (7), ih (8) 2
eh (7), gh (7), ih (8) eh h 7
2 4

Thus a minimal a-path spanning tree for the graph is: g h i

Note that if we had included gh instead of eh in the last step, we would
have found a different minimal a-path spanning tree.
MATH1081 Topic 5 – Graph theory Lecture 5.05 – Trees and weighted graphs 11/11