GRAPH THEORY

STUDY MATERIAL

B.SC. MATHEMATICS

VI SEMESTER

ELECTIVE COURSE

(2011 A DMISSION)

UNIVERSITY OF CALICUT
SCHOOL OF DISTANCE EDUCATION
THENJIPALAM, CALICUT UNIVERSITY P.O., MALAPPURAM, KERALA - 673 635

361
UNIVERSITY OF CALICUT
SCHOOL OF DISTANCE EDUCATION
Study Material

B.Sc.Mathematics
VI SEMESTER

ELECTIVE COURSE
GRAPH THEORY

Prepared and Scrutinised by :

Dr. Anilkumar V.
Reader,
Department of Mathematics,
University of Calicut

Type settings and Lay out :

Computer Section, SDE
©
Reserved
Contents

1 Graphs and Subgraphs 3

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 What is a graph ? . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Representation of a graph . . . . . . . . . . . . . . . . . . . . . 4
1.4 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 Degrees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.6 Solved Problems . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.7 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.8 Subgraphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.9 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.10 Isomorphism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.11 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.12 Ramsey Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.13 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.14 Indepedent Sets and Coverings . . . . . . . . . . . . . . . . . . . 19
1.15 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.16 Intersection graphs and line graphs . . . . . . . . . . . . . . . . 22
1.17 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.18 Operations on graphs . . . . . . . . . . . . . . . . . . . . . . . . 24
1.19 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.20 Walks, Trails and Paths . . . . . . . . . . . . . . . . . . . . . . 27
1.21 Connectness and components . . . . . . . . . . . . . . . . . . . 29
1.22 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.23 Blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
1.24 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
1.25 Connectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

1
 1.26 Solved Problems . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.27 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

2 Eulerian graphs, Hamiltonian graphs and Trees 44

2.1 Eulerian graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.1.1 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.2 Hamiltonian Graphs . . . . . . . . . . . . . . . . . . . . . . . . 47
2.3 Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.3.1 Characterization of Trees . . . . . . . . . . . . . . . . . . 49
2.3.2 Centre of a Tree . . . . . . . . . . . . . . . . . . . . . . . 52
2.3.3 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

3 Matchings and Planarity 54

3.1 Matchings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.2 Worked Problems . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.3 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.4 Matchings in Bipartite Graphs . . . . . . . . . . . . . . . . . . . 58
3.4.1 Personnel Assignment Problem . . . . . . . . . . . . . . 58
3.4.2 The marriage Problem . . . . . . . . . . . . . . . . . . . 58
3.5 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.6 Planarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.6.1 Characterization of Planar Graphs . . . . . . . . . . . . 66

4 Colourability 68
4.1 Chromatic Number and Chromatic Index . . . . . . . . . . . . . 68
4.2 The Five Colour Theorem . . . . . . . . . . . . . . . . . . . . . 73
4.3 Chromatic Polynomials . . . . . . . . . . . . . . . . . . . . . . . 74
4.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.5 Directed graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.6 Path and Connectedness . . . . . . . . . . . . . . . . . . . . . . 83
4.7 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

2
Module 1

Graphs and Subgraphs

1.1 Introduction
Graph theory is a branch of mathematics which deals the problems, with the
help of diagrams. There are may applications of graph theory to a wide variety
of subjects which include operations research, physics, chemistry, computer
science and other branches of science. In this chapter we introduce some basic
concepts of graph theory and provide variety of examples. We also obtain
some elementary results.

1.2 What is a graph ?

Definition 1.2.1. A graph G consists of a pair (V (G), X(G)) where V (G)
is a non empty finite set whose elements are called points or vertices and
X(G) is a set of unordered pairs of distinct elements of V (G). The elements
of X(G) are called lines or edges of the graph G. If x = {u, v} ∈ X(G), the
line x is said to join u and v. We write x = uv and we say that the points u
and v are adjacent. We also say that the point u and the line x are incident
with each other. If two lines x and y are incident with a common point then
they are called adjacent lines. A graph with p points and q lines is called a
(p, q) graph. When there is no possibility of confusion we write V (G) = V and
X(G) = X.

3
 a

b c d

Figure 1.1: A an example of a (4, 3) graph

1.3 Representation of a graph

It is customary to represent a graph by a diagram and refer to the diagram
itself as the graph. Each point is represented by a small dot and each line
is represented by a line segment joining the two points with which the line
is incident. Thus a diagram of graph depicts the incidence relation holding
between its points and lines. In drawing a graph it is immaterial whether the
lines are drawn straight or curved, long or short and what is important is the
incidence relation between its points and lines.

Example 1.3.1.

1. Let V = {a, b, c, d} and X = {{a, b}, {a, c}{a, d}}, G = (V, X) is a (4, 3)
graph. This graph can be represented by the diagram given in figure
1.1. In this graph the points a and b are adjacent whereas b and c are
nonadjacent.

2. Let V = {1, 2, 3, 4} and X = {{1, 2}, {1, 3}, {1, 4}, {2, 3}, {2, 4}, {3, 4}}.
Then G = (V, X) is a (4, 6) graph. This graph is represented by the
diagram given in figure 1.2 Although the lines {1, 2} and {2, 4} intersect
in the diagram, their intersection is not a point of the graph. Figure 1.3
is another diagram for the graph given in figure 1.2.

3. The (10, 15) graph given in figure 1.4 is called the Petersen graph.

Remark 1.3.1. The definition of a graph does not allow more than one line
joining two points. It also does not allow any line joining a point to itself.
Such a line joining a point to itself is called a loop.

4
 4 3

1 2

Figure 1.2: An example of a (4, 6) graph

2 3

Figure 1.3: Another representation of graph shown in figure 1.1

a
2 e
b 5
d
c
3 4

Figure 1.4: Peterson graph

5
 Figure 1.5: A multiple graph

Figure 1.6: A pseudograph

Definition 1.3.1. If more than one line joining two vertices are allowed, the
resulting object is called a multigraph. Line joining the same points are
called multi lines. If further loops are also allowed, the resulting object is
called Pseudo graph.

Example 1.3.2. Figure1.5 is a multigraph and figure 1.6 is a pseudo graph.

! !
p p
Remark 1.3.2. Let G be a (p, q) graph. Then q 6 and q = iff
2 2
any two distinct points are adjacent.

Definition 1.3.2. A Graph in which any two distinct points are adjacent is
called a complete graph. The complete graph with p points is denoted by
Kp . K3 is called a triangle. The graph given Fig. 1.3 is K4 and K5 is shown
in Fig.1.7

6
 Figure 1.7: K5

Definition 1.3.3. A graph whose edge set is empty is called a null graph or
a totally disconnected graph.
Definition 1.3.4. A graph G is called labeled if its p points are distinguished
from one another by names such as v1 , v2 · · · vp .
The graphs given in Fig. 1.1 and Fig. 1.3 are labelled graphs and the graph
in Fig. 1.7 is an unlabelled graph.
Definition 1.3.5. A graph G is called a bigraph or bipartite graph if V
can be partitioned into two disjoint subsets V1 and V2 such that every line of
G joins a point of V1 to a point of V2 . (V1 , V2 )is called a bipartition of G. If
further G contains every line joining the points of V1 to the points of V2 then
G is called a complete bigraph. If V1 contains m points and V2 contains n
points then the complete bigraph G is denoted by Km,n . The graph given in
Fig. 1.1 is K1,3 . The graph given in Fig. 1.8 is K3,3 . K1,m is called a star for
m ≥ 1.

1.4 Exercise
1. Draw all graphs with 1, 2, 3 and 4 points.

2. Find the number of points and lines in Km,n .

3. Let V = {1, 2, 3, · · · , n}. Let X = { {i, j}| i, j ∈ V and are relatievly

prime}. The resulting graph (V, X) is denoted by Gn . Draw G4 and G5 .

1.5 Degrees
Definition 1.5.1. The degree of a point vi in a graph G is the number of
lines incident with vi .The degree of vi is denoted by dG (vi ) or deg vi or d(vi ).

7
 Figure 1.8: bigraph

A point v of degree 0 is called an isolated point. A point v of degree 1 is

called an endpoint.

Theorem 1.5.1. The sum of the degrees of the points of a graph G is twice
P
the number of lines. That is, i degvi = 2q.

Proof. Every line of G is incident with two points. Hence every line contribute
P
2 to the sum of the degrees of the points. Hence i degvi = 2q.

Corollary 1.5.1. In any graph G the number of points of odd degree is even.

Proof. Let v1 , v2 , · · · , vk denote the point of odd degree and w1 , w2 · · · , wm

Pk
denote the points of even degree in G. By theorem 1.5.1, i=1 deg(vi ) +
Pw Pm Pm
i=1 degwi = 2q which is even. Further i=1 degwi is even. Hence i=1 degvi
is also even. But degvi is odd for each i. Hence k must be even.

Definition 1.5.2. For any graph G,we define

δ(G) = min{degv/v ∈ V (G)} and

∆(G) = max{degv/v ∈ V (G)}.

It all the points of G have the same degree r, then δ(G) = ∆(G) = r and this
case G is called a regular graph of degree r. A regular graph of degree 3 is
called a cubic graph. For example, the complete graph Kp is regular of degree
p − 1.

Theorem 1.5.2. Every cubic graph has an even number of points.

P
Proof. Let G be a cubic graph with p points, then degv = 3p which is even
by theorem 1.5.1. Hence p is even.

8
1.6 Solved Problems
Problem 1. Let G be a (p, q) graph all of whose points have degree k or k + 1.
If G has t > 0 points of degree k, show that t = p(k + 1) − 2q.
Solution
Since G has t points of degree k, the remaining p − t points have degree k + 1.
P
Hence v∈V d(v) = tk + (p − t)(k + 1).

∴ tk + (p − t)(k + 1) = 2q
∴ t = p(k + 1) − 2q.

Problem 2. Show that in any group of two or more people, there are always
two with exactly the same number of friends inside the group.

Solution. We construct a graph G by taking the group of people as the set of

points and joining two of them if they are friends, then degv is equal to number
of friends of v and hence we need only to prove that at least two points of G
have the same degree. Let V (G) = {v1 , v2 , · · · , vp .}. Clearly 0 ≤ degvi ≤ p − 1
for each i. Suppose no two points of G have the same degree. Then the degrees
of v1 , v2 , · · · , vp . are the integers 0, 1, 2, · · · , p − 1 in some order. However a
point of degree p − 1 is joined to every other point of G and hence no point
can have degree zero which is a contradiction. Hence there exist two points of
G with equal degree.

Problem 3. Prove that δ ≤ 2q/p ≤ ∆

Solution
Let V (G) = {v1 , v2 , · · · , vp }. We have δ ≤ degvi ≤ ∆. for all i. Hence
p
X
pδ ≤ degvi ≤ p∆.
i=1
∴ pδ ≤ 2q ≤ p∆ (by theorem2.1)
2q
∴δ ≤ ≤∆
p

Problem 4. Let G be a k-regular bibgraph with bipartion (V1 , V2 ) and k > 0.

Prove that |V1 | = |V2 | .

Solution
Since every line of G has one end in V1 and other end in V2 it follows that

9
P P
d(v) = v∈V2 d(v) = q. Also d(v) = k for all v ∈ V = V1 ∪ V2 . Hence
v∈V1
P P
v∈V1 d(v) = k|V 1 | and v∈V2 d(v) = k|V2 | so that k |V1 | = k|V2 |. Since k > 0,
we have |V1 | = |V2 |.

1.7 Exercise
1. Given an example of a regular graph of degree 0

2. Give three examples for a regular graph of degree 1

3. Give three examples for a regular graph of degree 2

4. What is the maximum degree of any point in a graph with p points?

5. Show that a graph with p points is regular of degree p − 1 if and only if

it is complete

6. Let G be a graph with at least two points show that G contains two
vertices of the same degree

7. A (p, q) graph has t points of degree m and all other points are of degree
n. Show that (m − n)t + pn = 2q.

1.8 Subgraphs
Definition 1.8.1. A graph H = (V1 , X1 ) is called subgraph of G = (V, X).
V1 ⊆ V and X1 ⊆ X. If H is a subgraph of G we say that G is a supergraph
of H. H is called a spanning subgraph of G if H is the maximal subgraph
of G with point set V1 . Thus, if H is an induced subgraph of G, two points are
adjacent in H they are adjacent in G. If V2 ⊆ V , then the induced subgraph
of G induced by V2 and is denoted by G[X]. If X2 ⊆ X, then the sub graph
of G with line set X2 and is denoted by G[X2 ]

Examples. Consider the petersen graph G given in Fig. 1.4. The graph
given in Fig.1.9 is a subgraph of G. The graph given in Fig.1.10 is an induced
subgraph of G. The graph given in Fig.1.10 is an induced subgraph of G. The
graph given in Fig1.11 is a spanning subgraph of G.

10
 1

2 a 5

b e
c

Figure 1.9: Subgraph

a

b e

c d

Figure 1.10: Induced subgraph

1

2 a 5
b d

c e
3 4

Figure 1.11: Spanning subgraph

11
 u u
v1 v1
v v u v

v1
x1 X1
u0 v
G G-v1 G-x1 G+uv

Figure 1.12:

Definition 1.8.2. Let G = (V, X) be a graph.Let vi ∈ V . The subgraph

of G obtained by removing the point vi and all the lines incident with vi is
called the subgraph obtained by the removal of the point vi and is
denoted by G-vi . Thus if G − vi = (Vi , Xi ) then Vi = V − vi and Xi = {x/x ∈
X and x is not incident with vi }. Clearly G − vi is an induced subgraph of G.
Let xi ∈ X. Then G − xi = (V, X − xj ) is called the subgraph of G obtained
by the removal of the line xj . Clearly G − xj is a spanning subgraph of G
which contains all the lines of G except xj . The removal of a set of points or
lines from G is defined to be the removal of single elements in succession.

Definition 1.8.3. Let G = (V, X) be a graph. Let vi , vj be two points which

S
are not adjacent in G. Then G + vi vj = (V, X {vi , vj }) is called the graph
obtained by the addition of the line vi vj to G

Clearly G+vi vj is the smallest super graph of G containing the line vi vj .We
listed these concepts in Fig1.12. The proof given in the following theorem is
typical of several proofs in theory.

Theorem 1.8.1.
h 2 i The maximum number of lines among all p point graph no
triangles is p4 . ([x] denotes the greatest integer not exceeding the the real
number x).

Proof. The result can be easily verified for p ≤ 4. For p > 4, we will prove by
induction separately for odd p and for every p.
Part 1. For odd p.
Suppose the result is true for all odd p ≤ 2n + 1. Now hlet iG be a (p, q) graph
2
with p = 2n + 3 and no triangles. Ifq = 0, then q ≤ p4 . Hence let q > 0.
Let u and v be a pair of adjacent points. The subgraph G′ = G − {u, v} has

12
2n + 1 points and no triangles. Hence induction hypothesis,

(2n + 1)2
   2 
′ 4n + 4n + 1
q(G ) ≤ =
4 4
 
1
= n2 + n + = n2 + n
4

Since G has no triangles, no point of G′ can be adjacent to both u and G.

Now, lines in G are of three types.
1. Lines of G′ (≤ n2 + n in number by(1))

2. Lines between G′ and {u, v}(≤ 2n + 1 innumberby(2))

3. Line uv

Hence

q ≤ (n2 + n) + (2n + 1) + 1 = n2 + 3n + 2
1
= (4n2 + 12n + 8)
4
4n2 + 12n + 9 1
 
= −
4 4
2
   2
(2n + 3) p
= =
4 4

Also for p = 2n + 3, the graph Kn+1,n+2 has no triangles and has (n +

2
1)(n + 2) = n2 + 3n + 2 = [ p4 ] lines. Hence this maximum q is attained.
Part 2. For even p.
Suppose the result is true for all even p ≤ 2n. Now let G be a (p, q) graph
with p = 2n + 2 and no triangles. As before, let u and v be a pair of adjacent
points in G and let G′ = G − {u, v}.
Now G′ has 2n points and no triangles. Hence by hypothesis,

(2n)2
 
′
q(G ) ≤ = n2
4

Lines in G are of three types.

13
 (i) Lines of G′

(ii) Lines between G′ and {u, v}

(iii) line uv.

(2n+2)2
Hence q ≤ n2 + 2n + 1 = (n + 1)2 = 4
= [p2 /4]. Hence the result holds
2
for even p also. We see that for p = 2n + 2. Kn+1,n+1 is a (p, [ p4 ] graph without
triangles.

1.9 Exercise
1. Show that Kp − v = Kp−1 for any point v of Kp .

2. Show that an induced subgraph of a complete graph is complete.

3. Let G = (V, X) be a (p, q) graph. Let v ∈ V and x ∈ X. Find the

number of points and lines in G − v and G − x.

4. If every induced proper subgraph of a graph G is complete and p > 2

then show that G is complete.

5. If every induced proper subgraph of a graph G is totally disconnected,

then show that G is totally disconnected.

6. Show that in a graph G every induced graph is complete iff every induced
graph with two points is complete.

1.10 Isomorphism
Definition 1.10.1. Two graphs G1 = (V1 , X1 ) and G2 = (V2 , X2 ) are said
to be isomorphic if there exists a bijection f : V1 → V2 such that u, v are
adjacent in G1 if and only if f (u), f (v) are adjacent in G2 . If G1 is isomorphic
to G2 , we write G1 ∼
= G2 . The map f is called an isomorphism from G1 to G2 .

Example 1.10.1. 1. The graph given in Fig. 2.2 and Fig. 2.3 are isomor-
phic.

2. The two graphs given in Fig.1.13 are isomorphic. f (ui ) = vi is an iso-

morphism between these two graphs.

14
 v2
u1 u2 v1

u5

u4 u3 v4 v3

v5

Figure 1.13:

Figure 1.14:

3. The three graphs given in Fig.1.14 are isomorphic with each other.

Theorem 1.10.1. Let f be an isomorphism of the graph G1 = (V1 , X1 ) to the

graph G2 = (V2 , X2 ). Let v ∈ V1 . Then deg v = deg f (v). i.e., isomorphism
preserves the degree of vertices.

Proof. A point u ∈ V1 is adjacent to v in G1 iff f (u) is adjacent to f (v) in G2 .

Also f is bijection. Hence the number of points in V1 which are adjacent to v
is equal to the number of points in V2 which are adjacent to f (v). Hence deg
v = deg f (v).

Remark 1.10.1. Two isomorphic graphs have the same number of points
and the same number of lines. Also it follows from Theorem 1.10.1that two
isomorphic graphs have equal number of points with a given degree. However
these conditions are not sufficient to ensure that two graphs are isomorphic.
For example consider the two graphs given in figure 1.15. By theorem 1.10.1,
under any isomorphism w4 must correspond to v3 ; w1 , w5 , w6 must correspond
to v1 , v5 , v6 in some order. The remaining two points w2 , w3 are adjacent
whereas v2 , v4 are not adjacent. Hence there does not exist an isomorphism

15
 w6 v6

w1 w2 w3 w4 v1 v2 v3 v v
4 5
w5

Figure 1.15:

Figure 1.16:

between these two graphs. However both graphs have exactly one vertex of
degree 3, three vertices of degree 1 and two vertices of degree 2.

Definition 1.10.2. An isomorphism of a graph G onto itself is called an

automorphism of G.

Remark 1.10.2. Let Γ(G) denote the set of all automorphism of G. Clearly
the identity map i : V → V defined by i(v) = v is an automorphism of G
so that i ∈ Γ(G). Further if α and β are automorphisms of G then α.β and
α−1 are also automorphism of G. Hence Γ(G) is a group and is called the
automorphism group of G.

Definition 1.10.3. Let G = (V, X) be a graph. The complement G of G

is defined to be the graph which has V as its set of points and two points
are adjacent in G iff they are not adjacent in G. G is said to be a self
complementary graph if G is isomorphic to G.

For example the graphs given in Fig.1.16 are self complementary graphs.
It has been conjectured by Ulam that the collection of vertex deleted sub-
graphs G − v determines G upto isomorphism.
Solved Problems

Problem 5. Prove that any self complementary graphs has 4n or 4n+1 points

16
Solution. Let G = (V (G), X(G)) be a self complementary graph with p
points.
Since G is self complementary, G is isomorphic to G.
∴ |X(G)| = |X(G)|. Also
!
p p(p−1)
|X(G)| + |X(G)| = = 2
2
p(p − 1)
∴ 2|X(G)| =
2
p(p − 1)
∴ |X(G)| = is an integer.
4

Further one of p or p − 1 is odd. Hence p or p − 1 is a multiple of 4. ∴ p is of

the the form 4n or 4n + 1.

Problem 6. Prove that Γ(G) = Γ(G).

Solution. Let f ∈ Γ(G) and let u, v ∈ V (G).

Then u, v are adjacent in G ⇔ u, v are not adjacent in G.

⇔ f (u), f (v) are not adjacent in G
(since f is an automorphism of G)
⇔ f (u), f (v) are adjacent in G.

Hence f is an automorphism of G.

∴ f ∈ Γ(G) and hence Γ(G) ⊆ Γ(G).

Similarly Γ(G) ⊆ Γ(G) so that Γ(G) = Γ(G).

1.11 Exercise
1. Prove that any graph with p points is isomorphic to a subgraph of Kp .

2. Show that isomorphism is an equivalence relation among graphs.

3. Show that the two graphs given in Fig. 2.17 are not isomorphic.

17
 4. Show that upto isomorphism there are exactly four graphs on three ver-
tices.

5. Prove that a graph G is complete iff G is totally disconnected.

6. Let G be (p, q) graph degG (v) = p − 1 − degG (v).

7. Prove that Γ(Kn ) ∼

= Sn , the symmetric group of degree n.

1.12 Ramsey Numbers

We start by considering the following puzzle. In any set of six people there
will always be either a subset of three who are mutually acquainted, or a
subset of three who are mutually strangers. This situation may be represented
by a graph G with six points representing the six people in which adjacency
indicates acquaintances. The above puzzle then asserts that G contains three
mutually adjacent points or three mutually non-adjacent points. Equivalently
G or G contains a triangle.

Theorem 1.12.1. For any graph G with 6 points, G or G contains a triangle.

Proof. Let v be a point of G. Since G contains 5 points other than v, v must

be either adjacent to three points in G or non-adjacent to three points in
G.Hence v must be adjacent to three points either in G or in G Without loss
of generality, let us assume that v is adjacent to three points u1 , u2 , u3 in G. If
two of these three points are adjacent, G contains a triangle. Otherwise these
three points from a triangle in G. Hence G or G contains a triangle.

It is easy to see that the above theorem is not true for graphs with less
than 6 points and we have this as an exercise to the reader. Thus 6 is the
smallest positive integer such that any graph G on 6 points contains K3 or K3 .
This suggests the following general question. What is the least positive integer
r(m, n) such that for any graph G with r(m, n) points, G contains Km or Kn .
For example r(3, 3) = 6 . The numbers r(m, n) are called Ramsey numbers
after F. Ramsey who proved the existence of r(m, n). The determination of
the Ramsey numbers is difficult unsolved problem. Solved Problems

18
Problem 7. Prove that r(m, n) = r(n, m).

Solution Let r(m, n) = s. LetG be any graph on s points. Then G also has
s points. Since r(m, n) = s,G has either Km or Kn as an induced subgraph.
Hence G has Kn or Km as an induced subgraph. Thus an arbitrary graph
on s points contains Kn or Km as an induced subgraph. ∴ r(n, m) ≤ s. i.e,
r(n, m) ≤ r(m, n). Interchanging m and n we get r(m, n) ≤ r(n, m). Hence
r(m, n) = r(n, m).

Problem 8. Prove that r(2, 2) = 2

Solution Let G be a graph on 2 points. Let V (G) = {u, v}. Then u and v
are either adjacent in G or adjacent in G. Hence G or G contains K2 . Thus
if G is any graph on two points, then G or G contains K2 and clearly 2 is the
least positive integer with this property. Hence r(2, 2) = 2.

1.13 Exercise
1. Prove, by suitable examples, that theorem 1.12.1 is not true graphs with
less than 6 points.

2. Find r(1, 1).

3. Find r(k, 1) for any positive integer k.

4. Find r(2, 3).

5. Find r(2, k) for any positive integer k.

1.14 Indepedent Sets and Coverings

Definition 1.14.1. A covering of a graph G = (V, X) is a subset K of V
such that every line of G is incident with a vertex in K. A covering K is called
a minimum covering if G has no covering K ′ with |K ′ | < |K|. The number
of vertices in a minimum covering of G is called the covering number of G
and is denoted by β.
A subset S of V is called an independent set of G if no two vertices S
are adjacent in G. An independent set S is said to be maximum if G has

19
no independent set S ′ with |S ′ | > |S|. The number of vertices in a maximum
independent set is called independence number of G and is denoted α.

Example
Consider the graph given in Fig. 1.18 {v6 } is an independent set. {v1 , v3 } is a
maximum independent set. {v1 , v2 , v3 , v4 , v5 } is a covering and {v2 , v3 , v4 , v5 }
is a minimum covering.

Theorem 1.14.1. A set S ⊆ V is an independent set of G if and only if V is

a covering of G.

Proof. By definition, S is independent iff no two vertices of S are adjacent.That

is, iff every line of S is incident with at least one point of V − S. That is, iff
V − S. is a covering of G.

Corollary 1.14.1. α + β = p

Proof. Let S be a maxium independent set of G and K be a minimum covering

of G.
∴ |S| = α and |K| = β.
Now V − S is a covering of G and K is a minimum covering of G. Hence
|K| ≤ |V − S| so that β ≤ p − α

∴β+α≤p (1.1)

Also V − K is an independent set and S is a maximum independent set

Hence |S| ≤ |V − K| so that α ≥ p − β.

α+β ≥p (1.2)

From 1.1 and (1.2) , we get α + β = p.

In the following definition we give the line analogue of coverings indepen-

dence.

Definition 1.14.2. A line covering of G is a subset L of X such that

every vertex is incident with a line of L. The number of line in a minimum
line covering of G is called the line covering number of G and is denoted

20
by β ′ . A set of lines is called independent if no two of them are adjacent.
The number of lines in a maximum independent set of lines is called the edge
independence number and is denoted by α′ . Gallai has proved that for any
non-trivial graph, α′ + β ′ = p, though it is not true that the complement of an
independent set of lines is a line covering.
Result α′ + β ′ = p.

Proof. Let S be a maximum independent set of lines of G so that |S| = α′ .

Let M be a set of lines, one incident for each of the p − 2α′ points of G not
S
covered by any line of S. Clearly S M is a line covering of G.

∴ |S ∪ M| ≥ β ′
∴ α′ + P − 2α′ ≥ β ′
∴ p ≥ α′ + β ′ (1.3)

Now, let T be a minimum line cover of G, so that |T | = β ′ . T cannot have a

line x both of whose ends are also incident with lines of T other than x (since,
otherwise T − {x} will become a line covering of G). Hence G|T |, the spanning
subgraph of G induced by T , is the union of stars. Hence each line of T is in-
cident with at least one endpoint of G[T ]. Let W be a set of endpoints of G[T ]
consisting of exactly one end point for each line of T . Hence |W | = |T | = β ′
and each star has exactly one point not in W . Hence

p = |W | + (number of stars in G[T ]) (1.4)

∴ p = β ′ + (number of stars in G[T ]) (1.5)

By choosing one line from each star of G[T ], we get set of independent lines of
G. Hence
α′ ≥ (number of stars in G[T ])
Hence (1.5) gives p ≤ β ′ + α′ .
Therefore by ((1.3)), α′ + β ′ = p. This complete the proof.

21
1.15 Exercise
1. Find α, β, α′ and β ′ for the complete graph Kp .

2. Prove or disprove. Every covering of a graph contains a minimum cover.

3. Prove or disprove. Every independent set of lines is contained in a max-

imum independent set of lines.

4. Give an example to show that the complement of an independent set of

lines need not be a line covering.

5. Give an example to show that the complement of a line covering need be

an independent set of lines.

1.16 Intersection graphs and line graphs

Definition 1.16.1. Let F = {S1 , S2 , · · · , Sp }be a non- empty family of dis-
tinct non empty subsets of a given set S. The intersection graph of F,
denoted Ω(F ) is defined as follows:
The set of points V of Ω(F ) is F itself and two points Si , Sj are adjacent if
T
i 6= j and Si Sj 6= ∅. A graph G is called an intersection graph on S if there
exist a family F of subsets of S such that G is isomorphic to Ω(F ).

Theorem 1.16.1. Every graph is an intersection graph.

Proof. Let G = (V, X) be a graph. Let V = {v1 , v2 , · · · , vp }. Let S = V ∪ X

For each vi ∈ V , let Si = {vi } ∪ {x ∈ X|vi ∈ x}.

Cleary F = {S1 , S2 , · · · , Sp } is a family of distinct non-empty subsets of S

Further if vi , vj are adjacent in V then vi vj ∈ Si∩Sj and hence Si∩Sj 6= ∅.

Conversly if Si ∩ Sj 6= ∅ then the element common to Si ∩ Sj is the line joining
vi and vj so that vi , vj are adjacent in G. Thus f : V → F defined by f (vi ) = Si
is an isomorphism of G to Ω(F ). Hence G is an intersection graph.

Definition 1.16.2. Let G = (V, X) be a graph with X 6= ∅. Then X can be

thought of a family of 2 element subsets of V. The intersection graph Ω(X) is

22
called the line graph of G and is denoted by L(G). Thus the points of L(G)
are lines of G and two points in L(G) are adjacent iff the corresponding lines
are adjacent in G.

A example of a graph and line graph are given in Fig.1.19.

Theorem 1.16.2. Let G be a (p, q) graph. L(G) is a (q, qL ) graph where

qL = 21 ( pi=1 di 2 ) − q.
P

Proof. By definition, number of points in L(G) is q. To find the number of

lines in L(G). Any two of the di lines incident with vi are adjacent in L(G)
di (di −1)
and hence we get 2
lines in L(G).

p
X di (di − 1)
Hence qL =
i=1
2
p p
1 X 2 1 X
= ( di ) − ( di )
2 i=1 2 i=1
p
1 X 2 1
= ( di ) − (2q)
2 i=1 2
p
1 X 2
= ( di ) − q
2 i=1

1.17 Exercise
Show that the line graphs of the two graphs given in Fig.1.20 are isomorphic.
The two graphs given in figure 2.20 constitute the only pair of non-isomorphic
connected graphs having isomorphic line graphs. In all other cases, L(G) ∼ =
L(G′ ) implies G ∼
= G′ as claimed in the following theorem.

Theorem 1.17.1. (Whitney.) Let G and G′ be connected graphs with iso-

morphic line graphs. Then G and G′ are isomorphic unless one is K3 and the
other K1,3 .

Definition 1.17.1. A Graph G is called a line graph if G ∼

= L(H) for some
graph H.

23
Example K4 − x is a line graph as seen in Fig.1.19. The following theorem is
called Beineke’s forbidden subgraph characteristics of line graphs.

Theorem 1.17.2. (Beineke.) G is a line graph iff none of the nine graphs
of Fig. 2.20 is an induced subgraph of G.

1.18 Operations on graphs

Definition 1.18.1. Let G1 = (V1 , X1 ) and G2 = (V2 , X2 ) be two graphs with
V1 ∩ V2 = Φ. We define:

• The union G1 ∪ G2 to be (V, X) where

V = V1 ∪ V2 and X = X1 ∪ X2

• The sum G1 + G2 as G1∪G2 together with all the lines joining points of
V1 to points of V2 .

• The product G1 × G2 having V = V1 × V2 and u = (u1, u2 ) and v =

(v1 , v2 ) are adjacent to v1 in G1 and u2 = v2 .

• The composition G1 [G2 ] as having V = V1 × V2 and u = (u1 , u2) and

v = (v1 , v2 ) are adjacent if u1 is adjacent to v1 in G1 or (u1 = v1 and u2
is adjacent to v2 in G2 ).

We note that Km + Kn = Km,n .

Theorem 1.18.1. Let G1 be a (p1 , q1 ) and G2 a (p2 , q2 ) graph.

1. G1 ∪ G2 is a (p1 + p2 , q1 + q2 ) graph.

2. G1 + G2 is a (p1 + p2 , q1 + q2 + p1 p2 ) graph.

3. G1 × G2 is a (p1 p2 , q1 p2 + q2 p1 ) graph.

4. G1 [G2 ] is (p1 p2 , p1 q2 + p22 q1 ) graph.

Proof.

24
 1. is obvious.

2.

number of lines in G1 + G2 = number of lines in G1 + number of lines inG2

+ number of lines joining points of V1 of points of V2 .
= q1 + q2 + p1 p2 . Hence we get (2)

3. Clearly number of points in G1 × G2 is p1 p2 .

Now, let (u1 , u2) ∈ V1 × V2 . The points adjacent to (u1 , u2 ) are (u1 , v2 )
where u2 is adjacent to v2 (v1 , u2) where adjacent to u1 .

∴ deg(u1, u2 ) = degu1 + degu2

The total number of lines in G1 × G2

1X
= [ deg(ui ) + deg(vj )]
2 i,j
p1 p2
1 XX
= (degui + degvj ) where ui ∈ V1 , vj ∈ V2
2 i=1 j=1
p1 p2
1X X
= (p2 degui + degvj )
2 i=1 j=1
p1
1X
= (p2 degui + 2q2 )
2 i=1
1
= (2p2 q1 + 2p1 q2 )
2
= p2 q1 + p1 q2

The proof of (4) is left to the reader.

1.19 Exercise
1. Prove (4) of Theorem1.17.1.

2. If G1 and G2 are regular, determine whether G1 + G2 , G1 × G2 and G1

are regular.

25
 3. What is Km + Kn ?

4. Express K4 − x in terms of K2 and K2 .

5. Express the graph in Fig. 2.21 in terms of K3 and K2 .

6. Express the graph G of Fig. 2.19 in terms of K1 and K3 .

7. Define two more binary operations on graphs in your own way.

Revision Questions Determine which of the following statements are true

and which are false.
!
p
1. If G is a (p, q) graph q ≤
2
!
p
2. If G is a (p, q) graph and q = then G is complete.
2

3. A subgraph of a complete graph is complete.

4. An induced subgraph of a complete graph is complete.

5. A subgraph of a bipartite graph is bipartite.

6. In any graph G the number of points of odd degree is even.

7. Any complete graph is regular.

8. Any complete bigraph is regular.

9. A regular graph of degree 0 is totally disconnected.

10. The only regular graph of degree 1 is K2 .

11. The only connected regular graph of degree i is K2 .

12. A graph G is regular iff δ = ∆.

13. An induced subgraph of regular graph is regular.

14. If G is regular, then G − V is regular.

15. If G is complete, then G − V is complete.

26
 16. Any two isomorphic graphs have the same number of points and same
number of lines.

17. Any two graphs having the same number of points and same number of
lines are isomorphic.

18. Isomorphism preserves the degree of vertices.

19. If G1 and G2 are regular, G1 + G2 is regular.

20. If G1 and G2 are regular G1 [G2 ] is regular.

Answers
1, 2, 4, 5, 6, 7, 9, 11, 12, 15, 16, 18 and 20 are true.

1.20 Walks, Trails and Paths

Definition 1.20.1. A walk of a graph G is an alternating sequence of points
and lines v0 , x1 , v1 , x2 , v2 , · · · , vn−1 , xn , vn beginning and ending with points
such that each line xi is incident with vi−1 and vi .

We say that the walks join v0 and vn and it is called a v0 -vn walk. v0 is
called the initial point and v1 is called the terminal point of the walk. The
above walk is also denoted by v0 , v1 , · · · , vn the lines of the walks being self
evident. n, the number of lines in the walk, is called the length of this walk.
A single point is considered as a walk of length 0. A walk is called a trail if
all its lines are distinct and is called a path if all its points are distinct.

Example 1.20.1. For the graph given in 1.23 v1 , v2 , v3 , v4 , v2 , v1 , v2 , v5 is a

walk. v1 , v2 , v4 , v3 , v2 , v5 is a trail but not a path. v1 , v2 , v4 , v5 is a path. Obliv-
iously, every path is a trail and a trail need not be a path.
The graph consisting of a path with n points is denoted by Pn .

Definition 1.20.2. A v0 − vn walk is called closed if v0 = vn . A closed walk

v0 , v1 , · · · , vn = v0 in which n > 3 and v0 , v1 , · · · , vn−1 are is distinct is called
of length n. A graph consisting of a cycle of length n is denoted by Cn .
C3 is called a triangle.

Theorem 1.20.1. In a graph G, any u − v walk contains a u − v path.

27
Proof. We prove the result by induction on the length of the walk. Any walk
of length 0 or 1 is obviously a path. Now, assume the result for all walks of
length less than n. If u = u0 , u1, · · · , un = v be a u − v walk of length n. If all
the points of the walk are distinct it is already a path. If not, there exists i and
j such that 0 ≤ i < j ≤ n and ui = uj . Now u = u0 , · · · , ui , uj+1, · · · , un = v
is a u − v walk of length less than n which by induction hypothesis contains a
u − v path.

Theorem 1.20.2. If δ ≥ k, then G has a path of length k.

Proof. Let v1 be an arbitrary point.Choose v2 adjacent to v1 . Since δ ≥ k, there

exists at least k − 1 vertices other than v1 which are adjacent to v2 . Choose
v1 6= v1 such that v3 is adjacent to v2 . In general having chosen v1 , v2 , · · · , vi
where 1 < i ≤ δ there exist a point vi+1 6= v0 , v1 , · · · , vn such that vi+1 is
adjacent to vi . This process yields a path of length k in G.
Aliter.Let P = (v0 , v1 , · · · , vn ) be the longest path in G. Then every vertex
adjacent to v0 lies on P . Sinced(v0 ) ≥ δ it follows that length of P ≥ δ ≥ k.
Hence P1 = (v0 , v1 , · · · , vk ) is a path of length k in G.

Theorem 1.20.3. A closed walk of odd length contains a cycle.

Proof. Let v = v0 , v1 , · · · , vn = v be a closed walk of odd length. Hence n ≥ 3.

If n = 3 this walk is itself the cycle C3 and hence the result is trivial. Now
assume the result for all walks of length less than n. If the given walk of length
n is itself is a cycle there is nothing to prove.If not there exists two positive
integers i and j such that i < j, {i, j} =
6 {0, n} and vi = vj . Now vi , vi+1 , · · · , vj
and v = v0 , v1 , · · · , vi , vj+1, · · · , vn = v are closed walks contained in the given
walk and the sum of their lengths is n. Sin ce n is odd at least one of these
walks is of odd length which by induction hypothesis contains a cycle.

Solved Problem

Problem 9. If A is the adjacency matrix of a graph with V = {v1 , v2 , · · · , vp },prove

that for any n ≥ 1 the (i, j)th entry of An is the number of vi − vj walks of
length n in G.

Solution We prove the result by induction on n. The number of vi − vj walks

of length 1

28
 (
1, if vi and vj are adjacent;
=
0, otherwise.
= aij .

Hence the result is true for n = 1.

(n−1)
We now assume that the result is true for n − 1. Let An−1 = (aij ) so
(n−1)
that aij is number of vi − vj walks of length n − 1 in G. Now An−1 A =
(n−1)
(aij )aij .Hence (i, j)t h entry of

p
(n−1)
X
n
A = aik akj (1.6)
k=1

Also every vi − vj walk of length n inG consists of a vi − vj walk of length

n − 1 followed by a vertex vj which is adjacent to vk . Hence vj is adjacent to
(n−1)
vk then akj = 1 and aij represents the number of vi − vj walks of length n
whose last edge is vi vj . Hence the right hand side of equation (1.6) gives the
number of vi − vj walks of length n in G. This completes the induction and
the proof.

1.21 Connectness and components

Definition 1.21.1. Two points u and v of a graph G are said to be connected
if there exists a u − v path in G.

Definition 1.21.2. A graph G is said to be connected if every pair of its

points are connected. A graph which is not connected is said to be discon-
nected.
For example, for n > 1 the graph Kn consisting of n points and no lines is
disconnected. The union of two graphs is disconnected.

It is an easy exercise to verify that connectedness of points is an equivalence

relation on the set of points V . Hence v is partitioned into nonempty subsets
V1 , V2 , · · · , Vn such that two vertices u and v are connected iff both u and v
belongs to the same set Vi .Let Gi denote the induced subgraph of G with vertex

29
set Vi . Clearly the subgraphs G1 , G2 , · · · , Gn are connected and are called the
Components of G.
Clearly a graph G is connected iff it has exactly one component. 1.24 gives
a disconnected graph with 5 components.
p−1
Theorem 1.21.1. A graph G with p points and δ ≥ 2
is connected.

Proof. Suppose G is not connected. Then G has more than one component.
Consider any component G1 = (V1 , X1 ) of G. Let v1 ∈ V1 . Since δ ≥ p−1
2
there
p−1
exist at least 2 points in G1 adjacent to v1 and hence V1 contains at least
p−1
2
+ 1 = p+12
points. Thus each component of G contains at least p+1
2
points
and G has at least two components. Hence number of points in G ≥ p + 1
which is a contradiction. Hence G is connected.

Theorem 1.21.2. A graph G is connected iff for any partition of V into

subsets V1 and V2 there is a line of G joining a point of V1 to a point of V2 .

Proof. Suppose G is connected.Let V = V1 ∪ V2 be a partition of a V into two

subset. Let u ∈ V1 and v ∈ V2 . Since G is connected, there exists a u − v
path in G, say, u = v0 , v1 , v2 , · · · , vn = v. Let i be the least positive integer
such that vi ∈ V2 .(Such an i exists since vn = v ∈ V2 ). Then vi−1 ∈ V1 and
vi−1 , vi are adjacent. Thus there is a line joining vi−1 ∈ V1 and vi ∈ V2 . To
prove the converse, suppose G is not connected. Then G contains at least two
components. Let V1 denote the set of all vertices of one component and V2 the
remaining vertices of G. Clearly V = V1 ∪ V2 is a partition of V and there is
no line joining any point of V1 to any point of V2 . Hence the theorem.

Theorem 1.21.3. If G is not connected then G is connected.

Proof. Since G is not connected, G has more than one component.Let u, v be

any two points of G. We will prove that there is a u−v path in G. If u, v belong
to different components in G, they are not adjacent in G and hence they are
adjacent in G.If u, v lie in the same component of G, choose w in a different
component. Then u, w, v is a u − v path in G. Hence G is connected.

Definition 1.21.3. For any ( two points u, v of a graph we define the distance
the length of the shortest u − v path , if such a path exists;
between u and v by d(u, v) =
∞, otherwise.
If G is a connected Graph, d(u, v) is always a non-negative integer. In this
case d is actually a metric on the set of points V (See problem 2).

30
Theorem 1.21.4. A graph G with at least two points is bipartite iff all its
cycles are of even length.
Proof. Suppose G is a bipartite. Then V can be partitioned into two subsets
V1 and V2 such that every line joins a point of V1 to a point of V2 . Now con-
sider any cycle v0 , v1 , v2 , · · · , vn = v0 of length n. Suppose v0 ∈ V1 . Then
v2 , v4 , v6 · · · ∈ V1 and v1 , v3 , v5 · · · ∈ V2 .Further vn = v0 ∈ V1 and hence n
is even. Conversely,suppose all cycles in G are of even length. We may as-
sume without loss of generality that G is connected.(If not we consider the
components of G separately). Let v1 ∈ V . Define

V1 = {v ∈ V |d(v, v1) is even}

V2 = {v ∈ V |d(v, v1) is odd}.

Clearly, V1 ∩ V2 = ∅ and V1 ∪ V2 = V . We claim that every line of G joins a

point of V1 to a point of V2 . Suppose two points u, v ∈ V1 are adjacent. Let
p be a shortest v1 − u path of length m and let Q be a shortest v1 − v path
of length n. Since u, v ∈ V1 both m and n are even. Now, let u1 be the last
point common to P and Q. Then the v1 − u1 path along P and the v1 − u1
path along Q are both shortest path and hence have the same length, say i.
Now the u1 − u path along P , the line uv followed by the v − u1 path along
Q form a cycle of length (m − i) + 1 + (n − i) = m + n − 2i + 1 which is odd
and this is a contradiction. Thus no two points of V1 are adjacent. Similarly
no two points of V2 are adjacent and hence G is bipartite. Hence the theorem.

To study the measure of connectedness of a graph G we consider the min-

imum number of points or lines to be removed from the graph in order to
disconnect it.
Definition 1.21.4. A cut point of a graph G is a point whose removal
increases the number of components.A bridge of a graph G is a line whose
removal increases the number of components.
Clearly if v is a cut point of a connected graph, G − v is disconnected.
For the graph given in Fig.1.25,1,2, and 3 are cut points. The lines {1, 2} and
{3, 4} are bridges. 5 is non-cut point.

31
Theorem 1.21.5. Let v be a point of a connected graph G. The following
statements are equivalent.

1. v is a cut-point of G.

2. There exists a partition of V − {v} into subsets U and W such that for
each u ∈ U and w ∈ W , the point v is on every u − w path.

3. There exists two points u and w distinct from v such that v is on every
u − w path.

Proof. (1) ⇒ (2). Since v is a cut-point of G, G − v is disconnected. Hence

G − v has at least two components. Let U consist of the points of one of the
components of G−v and W consist of the points of the remaining components.
Clearly V − {v} = U ∪ W is a partition of V − {v}. Let u ∈ U and w ∈ W .
Then u and w lie in different components of G − v. Hence there is no u − w
path in G − v.
Therefore every u − w path in G contains in v.
(2) ⇒ (3). This is trivial.
(3) ⇒ (1). Since v is on every u − w path in G there is no u − w path in G − v.
Hence G − v is not connected so that v is a cut point of G.

Theorem 1.21.6. Let x be a line of a connected graph G. The following

statements are equivalent.

1. x is bridge of G.

2. There exists a partition of V into two subsets U and W such that for
every point u ∈ U and w ∈ W , the line x is on every u − w path.

3. There exists two points u, w such that the line x is on every u − w path.

Proof. The proof is analogous to that of theorem 1.21.5 and is left as an exer-
cise.

Theorem 1.21.7. A line x of a connected graph G is a bridge iff x is not on

any cycle of G.

32
Proof. Let x be a bridge of G. Suppose x lies on a cycle C of G. Let w1 and
w2 be any two points in G. Since G is connected, there exists a w1 − w2 path
P in G. If x is not on P , then P is a path in G − x. If x is on P , replacing x
by C − x, we obtain a w1 − w2 walk in G − x. Walk contains a w1 − w2 path in
G − x.Hence G − x is connected which is contradiction to (1). Hence x is not
on any cycle on G. Conversely, let x = uv be not on any cycle of G. Suppose
x is not a bridge. Hence G − x is connected.
∴ There is a u − v path in G − x. This path together with the line x = uv
forms a cycle containing x and contradicts (2).Hence x is a bridge.

Theorem 1.21.8. Every non-trivial connected graphs has at least two points
which are not cut points.

Proof. Choose two points u and v such that d(u, v) is maximum. We claim
that u and v are not cut points. Suppose v is a cut point.Hence G − v has
more than one component. Choose a point w in a component that does not
contain u.Then v lies on every u − w path and hence d(u, w) > d(u, v) which
is impossible.Hence v is not a cut point. Similarly u is not a cut point. Hence
the theorem.

1.22 Exercise
1. Prove that connectedness of points is an equivalence relation on the
points of G.

2. Prove that for a connected graph G the distance function d(u, v) is ac-
tually a metric on G. i.e, d(u, v) ≥ 0 and d(u, v) = 0 iff u = v, d(u, v) =
d(v, u) and d(u, w) ≤ d(u, v) + d(v, w) for all u, v, w ∈ V.

3. Prove theorem 4.9.

4. If x = uv is a bridge for a connected graph G 6= K2 , show that either u

or v is a cut point of G.

5. Prove that if x is a bridge of a connected graph G, then G−x has exactly

two components. Give an example to show that a similar result is not
true for a cut point.

33
 6. The girth of a graph is defined to be the length of its shortest cycle. Find
the girths of (i) Km (ii)Km,n (iii)Cn (iv) The Peterson graph.

7. The circumference of a graph is defined to be the length of its longest

cycle. Find the circumference of the graphs given in problem 6.

8. Prove that if G is connected then its line graph is also connected.

9. Prove that any graph G with δ ≥ r ≥ 2 contains a cycle of length at

least r + 1.

10. Prove that if there exists two distinct cycles each containing a line x,
then there exists a cycle not containing x.

11. Prove that if a graph G has exactly two points of odd degree there must
be a path joining these two points.

12. Give an example of a connected graph in which every line is a bridge.

13. Prove that any graph with p points satisfying the conditions of problem
12 must have exactly p − 1 lines.

14. Give an example of a graph which has a cut point but does not have a
bridge.

15. Prove that if v is a cut point of G, then v is not a cut point of G

1.23 Blocks
Definition 1.23.1. A connected non-trivial graph having no cut point is a
block. A block of a graph is a subgraph that is a block and is maximal with
respect to this property.

A graph and its blocks are given in 1.26. In the following theorem we give
several equivalent conditions for a given block.

Theorem 1.23.1. Let G be a connected graph with at least three point,following

statements are equivalent.

1. G is a block.

34
 2. Any two points of G lie on a common cycle.

3. Any point and any line of G lie on a common cycle.

4. Any two lines of G lie on a common cycle.

Proof. (1) ⇒ (2) Suppose G is a block. We shall prove by induction on the

distance d(u, v) between u and v any two vertices u and v lie on a common
cycle. Suppose d(u, v) = 1. Hence u and v are adjacent. By hypothesis,
G 6= K2 and G has no cut points. Hence the line x = uv is not a bridge and
Theorem 1.21.7 x is on a cycle of G. Hence the points u and v lie on a common
cycle of G. Now assume that the result is true for any two vertices at distance
k and let d(u, v) = k ≥ 2. Consider a u − v path of length k. Let w be the
vertex that precedes v on this path. Then d(u, v) = k − 1. Hence by induction
hypothesis there exists a cycle C that contains u and w. Now since G is a
block, w is not a cut point of G and so G − w is. Hence there exists a u − v
path P not containing w. Let v ′ be the last point common to P and C. (See
Fig.1.27). Sinceu is common to P and C, such a v ′ exists. Now, let Q denote
the u − v ′ path along the cycle C not containing the point w.Then, Q followed
by the v ′ − v path along P , the line vw and the w − v path along the cycle
that contains both u and v. This completes the induction.
Thus any two points of G lie on a common cycle of G.
(2) ⇒ (1).Suppose any two points of G lie on a common cycle of G. Suppose v
is a cut point of G. Then there exists two points u and w distinct from v such
that every u − w path contains v.(Refer Theorem 4.8). Now, by hypothesis
uand w lie on a common cycle and this cycle determines two u − w paths and
at least one of these paths does not contain v which is a contradiction. Hence
G has no cut points so that G is a block.
(2) ⇒ (3). Let u be a point and vw a line of G. By hypothesis u and v lie on
a common cycle C. If w lies on C, then the line uw together with the v − w
path of C containing u is the required cycle containing u and the line vw. If w
is not on C, let C ′ be a cycle containing u and w. This cycles determines two
w − u paths and at least one of these paths does not contain v. Denote this
path by P . Let u′ be the first point common to P and C.(u′ may be u itself).
Then the line vw followed by the w − u′ sub path of P and the u′ − v path in
C containing u form a cycle containing u and the line vw. (3) ⇒ (2) is trivial.

35
(3) ⇒ (4). The proof is analogous to the proof of (2) ⇒ (3) and is left as an
exercise. (4) ⇒ (3) is trivial.

1.24 Exercise
1. Prove that each line of a graph lies in exactly one of its blocks.

2. Prove that the lines of any cycle of G lie entirely in a single block of G

3. Prove that if a point v is common to two distinct block ofG, then v is a

cut point of G.

4. Prove that a graph G is a block iff for any three distinct points of G,
there is a path joining any two of them which does not contain the third.

5. Prove that a graph G is a block iff for any three distinct points of G,
there is a path joining any two of them which contains the third.

1.25 Connectivity
We define two parameters of a graph, its connectivity and edge connectivity
which measures the extend to which it is connected.

Definition 1.25.1. The connectivity κ = κ(G) of a graph G is the minimum

number of points whose removal results in a disconnected or trivial graph. The
connectivity λ = λ(G) of G is the minimum number of lines whose removal
results in a disconnected or trivial graph.

Example 1.25.1.

1. The connectivity and line connectivity of a disconnected graph is 0.

2. The connectivity of a connected graph with a cut point is 1.

3. The line connectivity of a connected graph with a bridge is 1.

4. The complete graph Kp cannot be disconnected by removing any number

of points, but the removal of p − 1 points results in a trivial graph.Hence
κ(Kp ) = p − 1

36
Theorem 1.25.1. For any graph G ,κ ≤ λ ≤ δ.
Proof. We first prove λ ≤ δ. If G has no lines, λ = δ = 0. Other wise
removal of all the lines incident with a point of minimum degree results in a
disconnected graph . Hence λ ≤ δ. Now to prove κ ≤ λ, we consider the
following cases.
Case(i)G is disconnected or trivial.Then κ = λ = 0
Case(ii)G is a connected graph with a bridge x. Then λ = 1.Further in case
G = K2 or one of the points incident withx is a cut point. Hence κ = 1 so
that κ = λ = 1.
Case(iii)λ ≥ 2.Then there exist λ lines the removal of which disconnects
graph. hence the removal of λ − 1 of lines results in a graph G with bridge
x = uv. For each of these λ − 1 line select an incident point different from u or
v .The removal of these λ − 1 points removes all the λ − 1 lines. If the resulting
graph is disconnected, then κ ≤ λ − 1.If not x is a bridge of this subgraph and
hence the removal of u or v results in a disconnected or trivial graph. Hence
κ ≤ λ and this completes the proof.
Remark 1.25.1. The inequalities in theorem 1.25.1 are often strict. For the
graph given in fig 1.28 κ = 2, λ = 3 and δ = 4.

Definition 1.25.2. A graph G is said to be n-connected if κ(G) ≥ n and

n-line connected if λ(G) ≥ n.
Thus a non trivial graph is 1− connected iff it is connected. A non trivial
graph is 2− connected iff it is block having more than one line. Hence K2 is
the only block which is not 2− connected.

1.26 Solved Problems

Problem 10. Prove that G is k− connected graph then q ≥ pk 2
.
Solution.Since G is k− connected, k ≤ δ(by theorem 1.25.1).

1
∴q =
2
1
≥ pδ ( since d(v) ≥ δ for all v
2
pk
≥ .
2

37
Problem 11. Prove that there is no 3− connected graph with 7 edges.
Solution Suppose G is a 3− connected graph with 7 edges.G has 7 edges
⇒ p ≥ 5. Now q ≥ 3p2
. Therefore q ≥ 15
2
. Hence q ≥ 8 which is a contradiction.
Hence there is no 3− connected graph with 7 edges.

1.27 Exercise
1. Find the connectivity of Km,n .

2. Show that if G is n− line connected andE is a set of n lines,the the

number of components in the graph G − E is either 1 or 2.

3. give an example to show that the analogue of the above result is not true
for a n− connected graph.

4. Give an example of a closed walk of even length which does not contain
a cycle.

5. Give an example to show that the union of two distinct u − v walks need
not contain a cycle.

6. Prove that the union of two distinct u − v paths contain a cycle.

7. Show that if a line is in a closed trail of G then it is in a cycle of G.

8. Determine which of the following statements are true and which are false.

(a) Any u − v walk contains a u − v path.

(b) The union of any two distinct u − v walks contains a cycle.
(c) The union of any two distinct u − v paths contains a cycle.
(d) A graph is connected iff it has only one component.
(e) The complement of a connected graph is connected
(f) Any subgraph of a connected graph is connected
(g) An induced subgraph of a connected graph is connected
(h) If a graph has a cut point ,then it has a bridge.
(i) If a graph has a bridge ,then it has a cut point.

38
 (j) If v is a cut point of a G then ω(G − v) = ω(G) + 1
(k) If x is a bridge of G,then ω(G − x) = ω(G) + 1
(l) In a connected graph every line can be a bridge.
(m) In a connected graph every point can be a cut point.
(n) A point common to two distinct blocks of a graph G is a cut point
of G.
(o) Every line of a graph G lies in exactly one block of G.
(p) If a graph is n− connected then it is n− line connected.
(q) Every block is 2− connected.
Answers
1,3,4,11,12,14,15 and 16 are true.

39
 Figure 1.17:

v1

v2 v5
v6

v3 v4

Figure 1.18:

x1 x1
x2 x3
x2 x3
x4 x4
G L(G)

Figure 1.19:

Figure 1.20:

40
Figure 1.21:

41
 v2
v1
v2
v1
G1 v2
G2 G1 U G2
G1+ G2

(v1,v2) (v1,v2) (v1,v2) (v1,v2)(v1,v2) (v1,v2)

(v1,v2) (v1,v2) (v1,v2) (v1,v2)(v1,v2) (v1,v2)

G1*G2 G1[G2]

Figure 1.22:

v1
v
5 v
2

v v3
4

Figure 1.23:

Figure 1.24:

42
 1 2
5
3

Figure 1.25:

Figure 1.26:

V' P
U V
W
P

Figure 1.27:

Figure 1.28:

43
Module 2

Eulerian graphs, Hamiltonian

graphs and Trees

2.1 Eulerian graphs

Definition 2.1.1. A closed trail containing all the points and lines is called
an eulerian trail. A graph having an eulerian trail is called an eulerian graph.

Remark 2.1.1. In an eulerian graph, for every pair of points u and v there
exists at least two edge disjoint u − v trails and consequently there are at least
two edge disjoint u − v paths. The graph shown in figure 2.1 is eulerian.

Theorem 2.1.1. If G is a graph in which the degree of every vertex is at least

two then G contains a cycle.

Proof. First, we construct a sequence of verices v1 , v2 , v3 , . . . as follows. Choose

any vertex v. Let v1 be any vertex adjacent to v. Let v2 be any vertex adjacent

Figure 2.1: A Eulerian graph

44
to v1 other than v. At any stage, if the vertex vi , i ≥ 2 is already chosen, then
choose vi+1 to be any vertex adjacent to vi other than vi−1 . Since degree of
each vertex is at least 2, the existence of vi+1 is always guaranteed. G has only
finite number of vertices, at some stage we have to choose a vertex which has
been chosen before. Let vk be the first such vertex and let vk = vi where i < k.
Then vi vi+1 . . . vk is a cycle.

Theorem 2.1.2. Let G be a connected graph. Then the following statements

are equivalent.

(1) G is eulerian.

(2) every point has even degree.

(3) the set of edges of G can be partitioned into cycles.

Proof.

(1) ⇒ (2) Assume that G is eulerian. Let T be an eulerian trail in G, with

origin and terminus u. Each time a vertex v occurs in T in a place
other than the origin and terminus, two of the edges incident with v are
accounted for. Since an eulerian trail contains every edges of G, d(v) is
even for v 6= u. For u, one of the edges incident with u is accounted for
by the origin of T , another by the terminus of T and others are accounted
for in pairs. Hence d(u) is also even.

(2) ⇒ (3) Since G is connected and nontrivial every vertex of G has degree
at least 2. Hence G contains a cycle Z. The removal of the lines of Z
results in a spanning subgraph G1 in which again vertex has even degree.
If G1 has no edges, then all the lines of G form one cycle and hence (3)
holds. Otherwise, G1 has a cycle Z1 . Removal of the lines of Z1 from G1
results in spanning subgraph G2 in which every vertex has even degree.
Continuing the above process, when a graph Gn with no edge is obtained,
we obtain a partition of the edges of G into n cycles.

(3) ⇒ (1) If the partition has only one cycle, then G is obviously eulerian,
since it is connected. Otherwise let z1 , z2 , . . . , zn be the cycles forming
a partition of the lines of G. Since G is connected there exists a cycle
zi 6= z1 having a common point v1 with z1 . Without loss of generality,

45
 let it be z2 . The walk beginning at v1 and consisting of the cycles z1 and
z2 in succession is a closed trail containing the edges of these two cycles.
Continuing this process, we can construct a closed trail containing all
the edges of G. Hence G is eulerian.

Corollary 2.1.1. Let G be a connected graph with exactly 2n(n ≥ 1), odd
vertices. Then the edge set of G can be partitioned into n open trails.

Proof. Let the odd vertices of G be labelled v1 , v2 , . . . , vn ; w1 , w2 , . . . , wn in any

arbitrary order. Add n edges to G between the vertex pairs (v1 , w1 ), (v2 , w2 ), . . . , (vn , wn )
to form a new graph G′ . No two of these n edges are incident with the same
vertex. Further every vertex of G′ is of even degree and hence G′ has an eule-
rian trail T . If the n edges that we added to G are now removed from T , it
will split into n open trails. These are open trails in G and form a partition
of the edges of G.

Corollary 2.1.2. Let G be a connected graph with exactly two odd vertices.
Then G has an open trail containing all the vertices and edges of G.

Corollary 2.1.2 answers the question: Which diagrams can be drawn with-
out lifting one’s pen from the paper not covering any line segment more than
once?

Definition 2.1.2. A graph is said to be arbitrarily traversable(traceable)from

a vertex v if the following procedure always results in an eulerian trail. Start at
v by traversing any incident edge. On arriving at a vertex u, depart through
any incident edge not yet traversed and continue until all the lines are tra-
versed.
If a graph is arbitrary traversable from a vertex then it obviously eulerian.

The graph shown in figure 2.1 is arbitrarily traversable from v. From no

other point it is arbitrarily traversable.

Theorem 2.1.3. An eulerian graph G is arbitrarily traversable from a vertex

v in G iff every cycle in G contains v.

46
 Figure 2.2: A theta graph

2.1.1 Exercise
1. For what values of n, is Kn eulerian?

2. For what values of m and n is Kn,m is eulerian?

3. Show that if G has no vertices of odd degree, then there are edge disjoint
cycles C1 , C2 , . . . , Cn such that

E(G) = E(C1 ) ∪ E(C2 ) ∪ . . . ∪ E(Cm )

4. Show that every block of a connected graph G is eulerian then G is

eulerian.

2.2 Hamiltonian Graphs

Definition 2.2.1. A spanning cycle in a graph is called a hamiltonian cycle.
A graph having a hamiltonian cycle is called a hamiltonian graph.

Definition 2.2.2. A block with two adjacent vertices of degree 3 and all other
vertices of degree 2 is called a theta graph.

Example 2.2.1. The graph shown in figure 2.2is a theta graph. A theta graph
is obviously nonhamiltonian and every nonhamiltonian 2-connected graph has
a theta subgraph.

Theorem 2.2.1. Every hamiltonian graph is 2-connected.

Proof. Let G be a hamiltonian graph and let Z be a hamiltonian cycle in G.

For any vertex v of G, Z − v is connected and hence G − v is also connected.
Hence G has no cutpoints and thus G is 2-connected.

47
Theorem 2.2.2. If G is hamiltonian, then for every nonempty proper subset
S of V (G), ω(G − S) ≤ |S| where ω(H) denote the number of components in
any graph H.

Proof. Let Z be a hamiltonian cycle of G. Let S be any nonempty proper

subset of V (G). Now, ω(Z − S) ≤ |S|. Also Z − S is a spanning subgraph of
G − S and hence ω(G − S) ≤ ω(Z − S). Hence ω(G − S) ≤ |S|.

Theorem 2.2.3. The bipartite graph Km,n is nonhamiltonian.

Proof. Let (V1 , V2 ) be a bipartition of the graph with |V1 | = m and |V2 | = n.
The graph Km,n − V1 is the totally disconnected graph with n points. Hence
ω(Km,n − V1 ) = n > m = |V1 |. Therefore Km,n is non hamiltonian.

Remark 2.2.1. The converse of theorem 2.2.2 is not true. For example,
Petersen graph satisfies the condition of the theorem but is nonhamiltonian.

Theorem 2.2.4. If G is a graph with p ≥ 3 vertices and δ ≥ p/2, then G is

hamiltonian.

Proof. Suppose the theorem is false. Let G be a maximal nonhamiltonian

graph with p vertices and δ ≥ p/2. Since p ≥ 3, G can not be complete.
Let u and v be nonadjacent vertices in G. By the choice of G, G + uv is
hamiltonian. Moreover, since G is nonhamiltonian, each hamiltonian cycle of
G + uv must contain the line uv. Thus G has a spanning path v1 , v2 , . . . , vp
with origin u = v1 and terminus v = vp . Let S = {vi : uvi+1 ∈ E} and
T = {vi : i < p and vi v ∈ E} where E is the edge set of G. Clearly vp ∈
/ S∪T
and hence
|S ∪ T | < p (2.1)

Again if vi ∈ S ∩ T , then v1 v2 . . . vi vp vp−1 . . . vi+1 vi is a hamiltonian cycle in G,

contrary to the assumption. Hence S ∩ T = ∅ so that

|S ∩ T | = 0. (2.2)

Also by the definition of S and T , d(u) = |S| and d(v) = |T |. Hence by

equations (2.1) and (2.2) , d(u) + d(v) = |S| + |T | = |S ∪ T | < p. Thus
d(u) + d(v) < p. But since δ ≥ p/2, we have d(u) + d(v) ≥ p which gives a
contradiction.

48
 Figure 2.3: A tree(left) and a forest(right)

Lemma 1. Let G be a graph with p points and let u and v be nonadjacent

points in G such that d(u) + d(v) ≥ p. Then G is hamiltonian if and only if
G + uv is hamiltonian.

Proof. First, assume that G is hamiltonian. Then obviously G + uv is hamil-

tonian. Conversely, assume that G + uv is hamiltonian, but G is not. Then,
as in the proof of theorem 2.2.4, we obtain d(u) + d(v) < p. This contradicts
the hypothesis that d(u) + d(v) ≥ p. Thus G + uv is hamiltonian implies G is
hamiltonian.

2.3 Trees
2.3.1 Characterization of Trees
Definition 2.3.1. A graph that contains no cycles is called a an acyclic graph.
A connected acyclic graph is called a tree.A graph without cycles is also called
a forest so that the components of a forest are trees.

Example 2.3.1. An example of a tree and a forest is shown in figure 2.3.

Theorem 2.3.1. Let G be a (p, q) graph. The following statements are equiv-
alent.

(1) G is a tree.

(2) every two points of G are joined by a unique path.

(3) G is connected and p = q + 1

49
 (4) G is acyclic and p = q + 1

Proof.

(1) ⇒ (2) Assume that G is a tree. Let u and v be any two points of G. Since
G is connected there exists a u − v path in G. Now suppose that there
exists two distinct u − v paths, say:

P1 : u = v0 , v1 , v2 , . . . , vn = v and P2 : u = w0 , w1 , . . . , wm = v

Let i be the least positive integer such that 1 ≤ i < m and wi ∈

/ P1 (such
an i exists since P1 and P2 are distinct). Hence wi−1 ∈ P1 ∩ P2 . Let j
be the least positive integers such that i < j ≤ m and wj ∈ P1 . Then
the wi−1 − wj path along P2 followed by the wj wi−1 path along P1 form
a cycle which is a contradiction. Hence there exists a unique u − v path
in G.

(2) ⇒ (3) Assume that every two points of G are joined by a unique path. This
implies that G is connected. We will show that p = q + 1 by induction on
p. The result is trivial for connected graphs with 1 or 2 points. Assume
that the result is true for all graphs with fewer than p points. Let G be
a graph with p points. Let x = uv be any line in G. Since there exists a
unique u − v path in G, G − x is a disconnected graph with exactly two
components G1 and G2 . Let G1 be a (p1 , q1 ) graph and G2 be a (p2 , q2 )
graph. Then p1 + p2 = p and q1 + q2 = q − 1. Further by induction
hypothesis p1 = q1 + 1 and p2 = q2 + 1. Hence

p = p1 + p2 = q1 + q2 + 2 = q − 1 + 2 = q + 1

(3) ⇒ (4) Assume that G is connected and p = q + 1. We will show that G is

acyclic. Suppose G contains a cycle of length n. There are n points and
n lines on this cycle. Fix a point u on the cycle. Consider any one the
remaining p − n points not on the cycle, say v. Since G is connected we
can find a shortest u − v path in G. Consider the line on this shortest
path incident with v. The p − n lines thus obtained are all distinct.
Hence q ≥ (p − n) + n = p which is a contradiction since q + 1 = p. Thus
G is acyclic.

50
(4) ⇒ (1) Assume thatG is acyclic and p = q + 1. We will prove that G is a
tree. Since G is acyclic to prove that G is a tree we need only prove that
G is connected. Suppose G is not connected. Let G1 , G2 , . . . , Gk (k ≥ 2)
be the components of G. Since G is acyclic each of these components
is a tree. Thus qi + 1 = pi where Gi is a (pi , qi ) graph. This implies
that ki=1 qi + 1) = ki=1 pi . That is, q + k = p and k ≥ 2, which is a
P P

contradiction. Hence G is connected.

Corollary 2.3.1. Every non trivial tree G has at least two vertices of degree
one.
P
Proof. Since G is non trivial, d(v) ≥ 1 for all points v. Also d(v) = 2q =
2(p − 1) = 2p − 2. Hence d(v) = 1 for at least two vertices.

Theorem 2.3.2. Every connected graph has a spanning tree.

Proof. Let G be a connected graph. Let T be a minimal connected spanning

subgraph of G. Then for any line x of T , T − x is disconnected and hence x
is a bridge of T . Hence T is acyclic. Further T is connected and hence is a
tree.

Corollary 2.3.2. Let G be a (p, q) connected graph. Then q ≥ p − 1.

Proof. Let T be a spanning tree of G. Then the number of lines in T is p − 1.

Hence q ≥ p − 1.

Theorem 2.3.3. Let T be a spanning tree of a connected graph G. Let x = uv

be an edge of G not in T . Then T + x contains a unique cycle.

Proof. Since T is acyclic every cycle in T + x must contain x. Hence there

exists a one to one correspondence between cycles in T + x and u − v paths
in T . As there is a unique u − v path in tree T , there is a unique cycle in
T + x.

51
2.3.2 Centre of a Tree
Definition 2.3.2. Let v be a point in a connected graph G. The eccentricity
e(v) of v is defined by e(v) = max{d(u, v) : u ∈ V (G)}. The radius r(G) is
defined by r(G) = min{e(v) : v ∈ V (G)}. The point v is called the central
point if e(v) = r(G) and the set of central points is called the centre of G.

Theorem 2.3.4. Every tree has a centre consisting of either one point or two
adjacent points.

Proof. The result is trivial if G = K1 or K2 . So assume that let T be any

tree with p ≥ 2 points. T has at least two end points and maximum distance
from a given point u to any other point v occurs only when v is an end point.
Now delete all the end points from T . The resulting graph T ′ is also a tree
and eccentricity of each point in T ′ is exactly one less than the eccentricity of
the same point in T . Hence T and T ′ have the same centre. If the process
of removing the end points is repeated, we obtain successive trees having the
same centres as T and we eventually obtain a tree which is either K1 or K2 .
Hence the centre of T consists of either one point or two adjacent points.

2.3.3 Exercise
1. Show that there does not exists a nonhamiltonian graph with arbitrarily
high eccentricity.

2. Prove that a graph G is tree iff G is connected and every line of G is a

bridge.

3. Prove that if G is a forest with p points and k components then G has

p − k lines.

4. Prove that the origin and terminus of a longest path in a tree have degree
one.

5. Show that every tree with exactly 2 vertices of degree one is a path.

6. Show that every tree is a bipartite graph. Which trees are complete
bipartite graphs.

52
 7. Prove that every block of a tree is K2 .

8. Draw all trees with 4 and 5 vertices.

9. Prove that any edge of a connected graph G one of whose end point is
of degree one is contained in every spanning tree of G.

10. Prove that a line x of a connected graph is in every spanning tree of G

iff x is a bridge.

53
Module 3

Matchings and Planarity

3.1 Matchings
Definition 3.1.1. Any set M of independent lines of a graph G is called a
matching of G. If uv ∈ M, we say that u and v are matched under M. We
also say that the points u and v are M-saturated. A matching M is called a
perfect matching if every point of G is M-saturated. M is called a maximum
matching if there is no matching M ′ in G such that |M ′ | > |M|.
Example 3.1.1. Consider the graph G1 shown in figure 3.1. Let M1 =
{v1 v2 , v6 v3 , v5 v4 } is a perfect matching in G1 . Also M2 = {v1 v3 , v6 v5 } is a
matching in G1 . However M2 is not a perfect matching. The points v2 and
v4 are not M2 saturated. For the graph G2 , M2 = {v8 v4 , v1 v2 } is a maximum
matching but not a perfect matching.

Definition 3.1.2. Let M be a matching in G = (V, E). A path in G is called

an M-alternating path if its lines are alternatively in E − M and M. An M-
alternating path whose origin and terminus are both M-unsaturated is called
an M-augmenting path.

Example 3.1.2. Consider the graph G1 shown in figure in 3.1. P1 = {v6 , v5 , v4 , v3 }

is an M1 alternating path. Also P2 = {v2 , v1 , v3 , v6 , v5 , v4 } is an M2 augmented
path. In the graph G2 , (v7 , v9 , v4 ) is an M-alternating path.

Remark 3.1.1. If a graph G has a perfect matching M, then p = 2|M| and

hence p is even. However the converse is not true. The graph G2 shown in
figure 3.1 has an even number of vertices but has no perfect matching.

54
 v v2
2

v v v v
1 3 1 3

v v v
7
4 4

v v
v v 6 5
6 5

Figure 3.1: Graphs G1 (left) and G2 (right)

Theorem 3.1.1. Let M1 and M2 be two matchings in a graph G. Let M1 △M2

be the symmetric difference of M1 and M2 . Let H = G[M1 △M2 ] be the
subgraph of G induced by M1 △M2 . Then each component of H is either an
even cycle with edges alternatively in M1 and M2 or a path P with edges
alternatively in M1 and M2 such that the origin and the terminus of P are
unsaturated in M1 or M2 .

Proof. Let v be any point in H. Since M1 and M2 are matchings in G, at

most one line of M1 and at most one line of M2 are incident with v. Hence
the degree of v in H is either 1 or 2. Hence it follows that the components of
H must be as described in the theorem.

Example 3.1.3. Consider the graph G1 shown in figure 3.1. Note that

M1 △M2 = {v1 v2 , v6 v3 , v5 v4 , v1 v3 , v6 v5 }

The graph H1 = G1 [M1 △M2 ] is shown in figure 3.2.

Clearly H1 is a path whose edges are alternatively in M1 or M2 . The origin

v2 and the terminus v4 are both M2 - unsaturated. The following theorem due
to Berge gives a characterization of maximum matching.

Theorem 3.1.2. A matching M in a graph G is a maximum matching if and

only if G contains no M-augmented path.

Proof. Let M be a maximum matching in G. Suppose G contains an M-

augmented path P = (v0 , v1 , . . . , v2k+1 ). By the definition of M-augmenting

55
 v2

v1

v3 v4

v6 v5

Figure 3.2: The graph H1

path the lines v0 v1 , v2 v3 , . . . , v2k v2k+1 are not in M and the lines v1 v2 , v3 v4 , . . . , v2k−1 v2k
are in M. Hence

M ′ = M − {v1 v2 , v3 v4 , . . . , v2k−1 v2k } ∪ {v0 v1 , v2 v3 , . . . , v2k v2k+1 }

is a matching in G and |M ′ | = |M| + 1, which is a contradiction, since M is a

maximum matching. Hence G also has no M-augmenting path.
Conversely, suppose G has no M-augmenting path. If M is a not a maxi-
mum matching in G then there is exits a matching M ′ of G such that |M ′ | >
|M|. Let H = G[M△M ′ ]. By theorem 3.1.1, each component of H is either
an even cycle with edges alternatively in M and M ′ or a path P with edges
alternatively in M and M ′ such that the origin and terminus of P are unsat-
urated in M or M ′ . Clearly any component of H which is a cycle contains
equal number of edges from M and M ′ . Since |M ′ | > |M| there exists at least
one component of H which is a path whose first and last edges are from M ′ .
Thus the origin and terminus of P are M ′ - unsaturated in H and hence they
are M-saturated in G. Thus P is an M-augmenting path in G, which is a
contradiction. Hence M is maximum matching in G.

3.2 Worked Problems

Problem 12. For what values of n does the complete graph Kn have perfect
matching.

Clearly any graph with p odd has no perfect matching. Also the complete
graph Kn has a perfect matching if n is even. For example, if V (Kn ) =
{1, 2, . . . , n} then {12, 34, . . . , (n − 1)n} is a perfect matching of Kn . Thus Kn

56
has a perfect matching if and only if n is even.

Problem 13. Show that a tree has at most one perfect matching.

Let T be a tree. Suppose T has two perfect matchings say M1 and M2 .

Then degree of every vertex in H = T [M1 △M2 ] is 2. Hence every compo-
nent of H is an even cycle with edges alternatively in M1 and M2 . This is
a contradiction, since T has no cycles. Therefore T has at most one perfect
matching.

Problem 14. Find the number of perfect matching in the complete bipartite
graph Kn,n .

Let A = {x1 , x2 , . . . , xn } and B = {y1 , y2 , . . . , yn } be a bi-partition of Kn,n .

We observe that any matching of Kn,n that saturates every vertex of A is a
perfect matching. Now the vertex x1 can be saturated in n ways by choosing
any one of the edges x1 y1 , x1 y2 , . . . , x1 yn . Having saturated x1 the vertex x2
can be saturated in n − 1 ways. In general having saturated x1 , x2 , . . . , xi the
next vertex xi+1 can be saturated in n − i ways. Hence the number of perfect
matchings in Kn,n is n.(n − 1) . . . 2.1 = n!.

Problem 15. Find the number of perfect matchings in the complete graph
K2n .

Let V (K2n ) = {v1 , v2 , . . . , v2n }. The vertex v1 can be saturated in 2n − 1

way by choosing any line e1 incident at v1 . In general having chosen e1 , e2 . . . , ek
can be saturated in 2n − (2k + 1) ways. We obtain a perfect matching after the
choice of n lines in the above process. Hence the number of perfect matching
in K2n is equal to 1.3.5 . . . (2n − 1). Note that

1.2.3.4.5. . . . (2n − 1)(2n)

1.3.5 . . . (2n − 1) =
2.4.6 . . . 2n
(2n)!
= n
2 n!

3.3 Exercise
1. Find maximum matching in the tree shown in figure 3.3.

57
 Figure 3.3: A Tree

2. Prove that a 2-regular graph G has a perfect matching if and only if

every component of G is an even cycle.

3. Give an example of a 3- regular graph which has no perfect matching.

3.4 Matchings in Bipartite Graphs

3.4.1 Personnel Assignment Problem
In a company, n workers x1 , x2 , . . . , xn and m jobs j1 , j2 , . . . , jm are available.
Each worker is qualified for at least one of the jobs. Is it possible to assign
one job for each worker for which he is qualified? This problem is known as
personnel assignment problem. We construct a bipartite graph G with bi-
partition A = {x1 , x2 , . . . , xn } and B = {j1 , j2 , . . . , jm }. xi being joined to jk
if and only if xi is qualified for the job jk . The personnel assignment problem
reduce to the following question. Does G have a matching that saturates every
vertex in A?

3.4.2 The marriage Problem

Let A = {x1 , x2 , . . . , xn } be a set of n boys and B = {y1 , y2 , . . . , ym } be a set
of m girls in a village. Each boy has one or more girl friends. Under what
conditions can we arrange marriage in such a way that each boy marries one
of his girl friends? This problem is known as the marriage problem.
We now obtain a graph theoretical formulation of the above problem. Let

58
G be the bipartite graph with the bi-partition {A, B} such that xi is joined to
yj if and only if yi is a girl friend of xi . The marriage is equivalent to finding
the conditions under which G has a matching that saturates every vertex of
A.

Definition 3.4.1. For a subset S of V the neighbor set N(S) is the set of all
points each of which is adjacent to at least one vertex in S.

Theorem 3.4.1 (Halli’s Marriage Theorem). Let G be a bipartite graph with

bi-partition (A, B). Then G has a matching that saturates all the vertices of
A if and only if |N(S)| ≥ |S|, for every subset S of A.

Proof. Suppose G has a matching M that saturates all the vertices in A.Let
S ⊆ A. Then every vertices in S is matched under M to a vertex in N(S) and
two distinct vertices of N(S). Hence it follows that |N(S)| ≥ |S|.
Conversely, suppose |N(S)| ≥ |S| for all S ⊆ A. We wish to show that
G contains a matching which saturates every vertex in A. Suppose G has no
such matching. Let M ∗ be a maximum matching in G. By assumption there
exists a vertex x0 ∈ A which is M ∗ unsaturated. Let

Z = {v ∈ V (G) : there exists a M ∗ alternating path conecting x0 and v}

Since M ∗ is a maximum matching, by Berge’s theorem, G has no no M ∗

augmenting path and hence x0 is the only M ∗ unsaturated vertex in Z. Let
S = Z ∩ A and T = z ∩ B. Clearly x0 in S and every vertex of S − {x0 } is
matched under M ∗ with a vertex of T . Thus

|T | = |S| − 1. (3.1)

We now claim that N(S) = T . Clearly, from the definition of T , we have

T ⊆ N(S). (3.2)

Now let v ∈ N(S). Hence there exists u ∈ S such that v is adjacent

to u. Since S = Z ∩ A it follows that u ∈ Z. Hence there exits an M ∗
alternating path P = (x0 , y1, x1 , y2 , . . . , xk−1 , yk , u). If v lies on P , then clearly
v ∈ Z ∩ B = T . Suppose v does not lie on P . Now the edge yk u ∈ M ∗ . Hence

59
the edge uv is not in M ∗ . Hence the path P1 consisting of P followed by the
edge uv is an M ∗ -alternating path. Hence v ∈ Z ∩ B = T . Thus

N(S) ⊆ T. (3.3)

From equations (3.2) and (3.4.2) we have

N(S) = T (3.4)

From equations (3.1) and (3.4) we have

|N(S)| = |T | = |S| − 1 < |S|

which is a contradiction.
Remark 3.4.1. Hall’s theorem answers the marriage problem. The marriage
problem with n boys has a solution if and only if for every k with 1 ≤ k ≤ n,
every set of k boys has collectively at least k girl friends.
The following is an important consequence of Hall’s marriage theorem.
Theorem 3.4.2. Let G be a k regular bipartite graph with k > 0. Then G
has a perfect matching.
Proof. Let (V1 , V2 ) be a bi-partition of G. Since each edge of G has one end
in V1 and there are k edges incident with each vertex of V1 , we have q = k|V1 |.
By a similar argument q = k|V2 |, so that k|V1 | = k|V2 |. Since k > 0, we get
|V1 | = |V2 |. Now let S ≤ V1 . Let E1 denote the set of all edges incident with
vertices in N(S). Since G is k- regular, |E1 | = k|E2 | and |E2 | = k|N(S)|.
Also by definition of N(S), we have E1 ⊆ E2 , and hence it follows that k|S| ≤
k|N(S)|. Thus |N(S)| ≥ |S|. Hence by Hall’s theorem, G, has a matching M
that saturates every vertex in V1 . Since |V1 | = |V2 |, M also saturates all the
vertices of V2 . Thus M is a perfect matching.

3.5 Exercise
1. For any graph G, let O(G) denote the number of odd components of G.
Let G = (V, X) be any graph. Prove that if G has a perfect matching

60
 s

s
w t
w t

v u v u

Figure 3.4: A planar graph (left) and its embbeding (right)

M, then O(G − S) ≤ |S| for all S ⊆ V .

2. Using the above problem show that the following graph has no perfect
matching.

3.6 Planarity
Definition 3.6.1. A graph is said to be embedded in a surface S when it is
drawn on S such that no two edges intersect(meetins of edges at a vertex is
not considered an intersection). A graph is called planar if it can be drawn
on a plane without intersecting edges. A graph is called non planar if it is
not planar. A graph that is drawn on the plane without intersecting edges is
called a plane graph.

Example 3.6.1. The graph shown in figure (3.4) is planar.

Theorem 3.6.1. The complete graph K5 is non planar.

Proof. If possible, let K5 be planar. Then K5 contains a cycle of length 5 say

(s, t, u, v, w, s). Hence, without loss of generality, any plane embedding of K5
can be assumed to contain this cycle drawn in the form of a regular pentagon.
Hence the edge wt must lie either wholly inside the pentagon or wholly outside
it.
Suppose that wt is wholly inside the pentagon( the argument when it lies
wholly outside the pentagon is quite similar). Since the edge sv and su do not
cross the edge wt, they must be both lie outside the pentagon. The edge vt
cannot cross the edge su. Hence vt must be inside the pentagon. But now, the

61
edge uw crosses one of the edges already drawn, giving a contradiction. Hence
K5 is not planar.

Definition 3.6.2. Let G be a graph embedded on a plane π. Then π − G is

the union of disjoint regions. Such regions are called faces of G. each plane
graph has exactly one unbounded face and it is called the exterior face. Let F
be a face of plane graph G and e be an edge of G. Let P be a point in F . e
is said to be in the boundary of F if for every point Q of π on e there exists a
curve joining P and Q which lies entirely in F .

Theorem 3.6.2. A graph can be embedded in the surface of a sphere iff it

can be embedded in a plane.

Proof. Let G be a graph embedded on a sphere. Place the sphere on the plane
L and call the point of contact S(south pole). At point S, draw a normal to
the plane and let N (North pole) be the point where this normal intersects the
surface of the sphere.
Assume that the sphere is placed in such a way that N is disjoint from
G. For each point P on the sphere, let P ′ be the unique point on the plane
where the line NP intersects the surface of the plane. There is a one to one
correspondence between the points of the sphere other than N and the points
on the plane. In this way, the vertices and the edges of G can be projected on
the plane L, which gives an embedding of G in L.
The reverse process obviously gives an embedding in the sphere for any
graph that is embedded in the plane L. This completes the proof.

Theorem 3.6.3. Every planar graph can be embedded in a plane such that
all edges are straight line segments

Definition 3.6.3. A graph is ployhedral if its vertices and edges may be

identified with the vertices and edges of a convex polyhedron in the three
dimensional space.

Theorem 3.6.4. A graph is polyhedral if and only if it is planar and 3 con-

nected.

Theorem 3.6.5. Every polyhedron that has at last two faces with the same
number of edges on the boundary.

62
Proof. The corresponding graph G is 3 connected. Hence δ(G) ≥ 3 and the
number of faces adjacent to any chosen face f is equal to the number of edges
in the boundary of the face f ( if two faces have the edges u and vw with r 6= w
in common, then G − {r, w} is disconnected contradicting 3 connectedness).
Let f1 , f2 , . . . , fm be the faces of the polyhedron and ei be the number of edges
on the boundary of the ith face. Let the faces be labelled so that ei ≤ ei+1 for
every i. If no two faces have the same number of edges in their boundaries,
then ei+1 − ei ≥ 1 for every i. Hence em − e1 = m−1
P
i=1 (ei+1 − ei ) ≥ m − 1 so
that em ≥ e1 + m − 1. Since e1 ≥ 3, this implies that em ≥ m + 2 so that the
mth face is adjacent to at least m + 2 faces. This gives a contradiction as there
are only m faces. This proves the theorem.

Theorem 3.6.6 (Euler Theorem). If G is a connected plane graph having V ,

E, and F as the set of vertices, edges and faces respectively, then |V | − |E| +
|F | = 2.

Proof. The proof is by induction on the number of edges of G. Let |E| = 0.

Since G is connected, it is K1 so that |V | = 1, |F | = 1 and hence |V | − |E| +
|F | = 2. Now let G be a graph as in theorem and suppose that the theorem is
true for all connected plane graphs with at most |E| − 1 edges.
If G is a tree, then |E| = |V |−1 and |F | = 1 and hence |V |−|E|+|F | = 2. If
G is not a tree, let x be an edge contained in some cycle of G. Then G′ = G−x
is a connected plane graph such that |V (G′ )| = |V |, |E(G′ )| = |E| − 1 and
|F (G′ )| = |F |−1. Hence by induction hypothesis |V (G′ )|−|E(G′)|+|F (G′)| =
2 so that |V | − (|E| − 1) + |F | − 1 = 2. Hence |V | − |E| + |F | = 2.

Theorem 3.6.7. If G is a plane (p, q) graph with r faces and k components

then p − q + r = k + 1.

Proof. Consider a plane embedding of G such that the exterior face of each
component contains all other components. Now let the ith component be a
(pi , qi ) graph with ri faces for each i. By the theorem pi − qi + ri = 2. Hence
X X X
pi − qi + ri = 2k (3.5)
P P P
But pi = p, qi = q and ri = r + (k − 1). Since the infinite face is

63
 P
counted k times in ri , hence equation (3.5) gives p − q + r + k − 1 = 2k so
that p − q + r = k + 1.

Corollary 3.6.1. If G is a (p, q) plane graph in which every face is an n cycle

then q = n(p − 2)/(n − 2).

Proof. Every face is an n-cycle. Hence each edge lies on the boundary of
exactly two faces. Let f1 , f2 , . . . , fr be the faces of G. Therefore
r
X
2q = (number of edges in the boundary of the face fi ) = nr
i=1

This implies that r = 2q/n. By Eulers formula p − q + r = 2. That is

p − q + 2q/n = 2
q(2/n − 1) = 2 − pq = n(p − 2)/(n − 2)

Corollary 3.6.2. In any connected plane (p, q) graph (p ≥ 3) with r faces

q ≥ 3r/2 and q ≤ 3p − 6.

Proof.

Case 1 Let G be a tree. Then r = 1, q = p − 1 and p ≥ 3. Hence q ≥ 3r/2

and q ≤ 3p − 6 since p − 1 ≤ 3p − 6 (as p ≥ 3).

Case 2 Let G have a cycle. let fi i = 1, 2, . . . , r be the faces of G. Since each

edge lies on the boundary of almost two faces,
r
X
2q ≥ (number of edges in the boundary of face fi )
i=1

That is,
2q ≥ 3r

That is
q ≤ 3r/2 (3.6)

By Euler’s formula, p − q + r = 2. Substituting for r in equation (3.6),

we get q ≥ 3/2(2 + q − p). After simplification we get, q ≤ 3p − 6.

64
Definition 3.6.4. A graph is called maximal planar if no line can be added to
it without losing planarity. In a maximal planar graph, each face is a triangle
and such a graph is sometimes called a triangulated graph.

Corollary 3.6.3. If G is a maximal planar (p, q) graph then q = 3p − 6.

Corollary 3.6.4. If G is a plane connected (p, q) graph without triangles and

p ≥ 3, then q ≤ 2p − 4.

Proof. If G is a tree, then q = p − 1. Hence we have p − 1 = q ≤ 2p − 4.

Now let G have a cycle. Since G has no triangles, the boundary of each face
has at least four edges. Since each edge lies on at most two faces we have,
2q ≥ ri=1 (number of edges in the boundary of the ith face). That is,
P

2q ≥ 4r. (3.7)

By Euler’s formula, we have p − q + r = 2. Substituting for r in equation (3.7),

we get 2q ≥ 4(2 + q − p). Hence 4p − 8 ≥ 2q so that q ≤ 2p − 4.

Corollary 3.6.5. The graphs K5 and K3,3 are not planar.

Proof. Note that K5 is a (5, 10) graph. For any planar (p, q) graph, q ≤ 3p − 6.
But q = 10 and p = 5 do not satisfy this inequality. Hence K5 is not planar.
Also note that K3,3 is a (6, 9) bipartite graph and hence has no triangles. If
such a graph is planar, then by Corollary refq12, q ≤ 2p − 4. But p = 6 and
q = 9 do not satisfy this inequality. Hence K3,3 is not planar.

Corollary 3.6.6. Every planar graph G with p ≥ 3 points has at least three
points of degree less than 6.
P
By Corolary 3.6.2, q ≤ 3p−6. That is, 2q ≤ 6p−12. That is, di ≤ 6p−12
where di are the degrees of the vertices of G. Since G is connected, di ≤ 1 for
P
every i. If at most two di are less than 6, then di ≥ 1 + 1 + 6 + . . . + (p −2) =
6p − 10 which is a contradiction. Hence di < 6 for at least three values of i.

Theorem 3.6.8. Every planar graph G with at least 3 points is a subgraph

of a triangulated graph with the same number of points.

65
Proof. Let G have p vertices. If p ≤ 4, then G must be a subgraph of Kp
which is a triangulated graph. Hence let p ≥ 5.
We construct a triangulated graph G′ which contains G as a subgraph as
follows:
Consider a plane embedding of G. If R is a face of G and v1 and v2 are two
vertices on the boundary of R without a connecting edge we connect v1 and v2
with an edge lying entirely in R. This yields a new plane graph. This yields a
new plane graph. This operation is continued until every pair of vertices on the
boundary of the same face are connected by an edge. The number of vertices
remains the same under these operation. Hence the process terminates after
some time yielding a plane triangulated graph G′ . G is obviously a subgraph
of G′ .

3.6.1 Characterization of Planar Graphs

Definition 3.6.5. Let x = uv be an edge of a graph G. Line x is said to be
subdivided when a new point w is adjoined to G and the line x is replaced
by the lines uw and wv. This process is also called an elementary subdivision
of the edge x. Two graphs are called homeomorphic if both can be obtained
from the same graph by a sequence of subdivisions of the lines.

Example 3.6.2. Any two cycles are homeomorphic.

Theorem 3.6.9 (Kuratowski Theorem). A graph is planar if and only if it

has no subgraph homeomorphic to K5 or K3,3 .

Remark 3.6.1. The graphs K5 and K3,3 are called Kuratowski’s graphs.

Definition 3.6.6. Let u and v be two adjacent points in a graph G. The

graph obtained from G by the removal of u and v and the addition of a new
point w adjacent to those points to which u or v was adjacent is called an
elementary contraction of G. A graph G is contractible to a graph H if H can
be obtained from G by a sequence of elementary contractions.

Example 3.6.3. The Petersen graph given in figure 3.5 is contractible to K5

by contracting the lines 1a, 2b, 3c, 4d and 5e.

66
 1

2 5

b e

c d

3 4

Figure 3.5: Petersen Graph

Theorem 3.6.10. A graph is planar if and only if it does no have a subgraph

contractible to K5 or K3,3 .

Since the Petersen graph is contractible to K5 , it is not planar according

to the theorem 3.6.10.

Definition 3.6.7. Given a plane graph G, its geometrical dual G∗ is con-

structed as follows: Place a vertex in each face of G(including the exterior
face). For each edge x of G, draw an edge x∗ joining the vertices representing
the faces on both sides x such that x∗ crosses only the edge x. The result is
always a plane graph G∗ (possibly with loops and multiple edges).

67
Module 4

Colourability

4.1 Chromatic Number and Chromatic Index

Definition 4.1.1. An assignment of colours to the vertices of a graph so that
no two adjacent vertices get the same colour is called a colouring of the graph.
For each colour, the set of all points which get that colour is independent and is
called a colour class. A colouring of a graph G using at most n colours is called
an n colouring. The chromatic number χ(G) of a graph Gis the minimum
number of colours needed to colour G. A graph G is called n-colourable if
χ(G) ≤ n.

Example 4.1.1. The chromatic numbers of some well known graphs are given
below:

Graph Kp Kp − x Kp Km,n C2n C2n+1

χ(G) p p−1 1 2 2 3
Example 4.1.2. If T is a tree with at least two points, then χ(T ) = 2.

Example 4.1.3. Let W be a wheel. Then χ(W ) is 3 or 4 according as it has

an odd or even number of points.

Definition 4.1.2. Each n-colouring of G partitions V (G) into independent

sets called colour classes. Such a partitioning induced by a χ(G) colouring of
G is called a chromatic partitioning. In other words, a partition of V (G) into
smallest possible number of independent sets is called a chromatic partitioning
of G.

68
 2

1 3
4
5

8 7 6

Figure 4.1: A graph with χ(G) = 3

Example 4.1.4. Consider the graph shown in figure 4.1. Note that chi(G) =
3. {1, 4, 8}, {3, 6, 7}, {2, 5} is a chromatic partitioning of this graph.

Theorem 4.1.1. Let G be any graph. Then the following statements are
equivalent.

1. G is 2-colourable.

2. G is bipartite.

3. every cycle of G has even length.

Proof.

(1) ⇒ (2) Assume that G is 2-colourable. Then V (G) can be partitioned into
two colour classes. These colour classes are independent sets and hence
a partition of G. Hence G is bipartite.

(2) ⇒ (1) Assume that G is bipartite. Then V (G) can be partitioned into two
sets V1 and V2 such that V1 and V2 are independent sets. A 2-colouring
of G can be obtained by colouring all the points of V1 white and all the
points of V2 blue. Hence G is 2-colourable.

(2) ⇔ (3) (See Theorem 4.7, page )

Remark 4.1.1. G is bipartite does not imply that χ(G) = 2. Consider the
graph K2 . Note that K2 is bipartite and χ(K2 ) = 1.

69
Definition 4.1.3. A graph G is called critical if χ(H) < χ(G) for every proper
subgraph H of G. A k-chromatic graph that is critical is called k- critical. It
is obvious that every k-chromatic graph has a k- critical subgraph.

Theorem 4.1.2. If G is k-critical, then δ(G) ≥ k − 1.

Proof. Since G is k-critical, for any vertex v of G, χ(G − v) = k − 1. If

deg(v) < k − 1, then a (k − 1) colouring of G − v can be extended to a
k − 1 colouring of Gby assigning to v, a colour which is assigned to none of its
neighbours in G. Hence deg(v) ≥ k − 1, so that δ(G) ≥ k − 1.

Corollary 4.1.1. Every k-chromatic graph has at least k vertices of degree

at least k − 1.

Proof. Let G be a k-chromatic graph and H be a k-critical subgraph of G. By

theorem 4.1.2, δ(H) ≥ k − 1. Also since χ(H) = k, H has at least k vertices.
Hence H has at least k vertices of degree at least k − 1. Since H is a subgraph
of G, the result follows.

Corollary 4.1.2. For any graph G, χ ≤ ∆ + 1.

Proof. Let G have chromatic number χ. Let H be a χ- critical subgraph of G.

By theorem 4.1.2, δ(H) ≥ χ − 1. Hence χ ≤ δ(H) + 1. Since δ(H) ≤ ∆(G),
this implies that χ ≤ ∆(G) + 1.

Theorem 4.1.3. For any graph G, χ(G) ≤ 1+maxδ(G′ ) where the summation
is taken over all induced subgraphs G′ of G.

Proof. The theorem is obvious for totally disconnected graphs. Now let G
be an arbitrary n- chromatic graph, n ≥ 2. Let H be any smallest induced
subgraph of G such that χ(H) = n. Hence χ(H − v) = n − 1 for every point
v of H. If degH v < n − 1, then a (n − 1) colouring of H − v can be extended
to a n − 1 colouring of H by assigning to v, a colour which is assigned to none
of its neighbours in H. Hence degH v ≥ n − 1. Since v is an arbitrary vertex
of H, this implies that δ(H) ≥ n − 1 = χ(G) − 1.
Hence χ(G) ≤ 1 + δ(H) ≤ 1 + maxδ(H ′ ) where the maximum is taken over
the set B of induced subgraphs G′ of G.

70
Definition 4.1.4. Ifχ(G) = n and every n-colouring of G induces the same
partition on V (G) then G is called uniquely n-colourable or uniquely colourable.

Example 4.1.5. K3 and K4 − x are uniquely 3-colourable. Kn is uniquely

n-colourable. Kn − x is uniquely (n − 1) colourable. Any connected bipartitle
graph is uniquely 2-colourable.

Theorem 4.1.4. If G is uniquely n-colourable, then δ(G) ≥ n − 1.

Proof. Let v be any point of v. In any n-colouring, v must be adjacent with at

least one point of every colour different from that assigned to v. Otherwise, by
reclouring v with a colour which none of its neighbours is having, a different
n-colouring can be achieved. Hence degree of v is at least (n − 1) so that
δ(G) ≥ n − 1.

Theorem 4.1.5. Let G be a uniquely n-colourable graph. Then in any n-

colouirng of G, the subgraph induced by the union of any two colour class is
connected.

Proof. If possible, let C1 and C2 be two classes in a n-colouring of G such that

the subgraph induced by C1 ∪C2 is disconnected. Let H be a component of the
subgraph induced by C1 ∪ C2 . Obviously, no point of H is adjacent to a point
in V (G) − V (H) that is coloured C1 or C2 . Hence interchanging the colours of
the points in H and retaining the original colours for all other points, we get
a different n-colouring for G. This gives a contradiction.

Theorem 4.1.6. Every uniquely n-colourable graph is (n − 1)- connected.

Proof. LetG be a uniquely n-colourable graph. Consider an n-colouring of G.

If possible, let G be not (n − 1) connected. Hence there exits a set S of at
most n − 2 points such that G − S is either trivial or disconnected. If G − S is
trivial, then G has at most n − 1 points so that G is not uniquely n-colourable.
G − S has at least two components. In the considered n-colouring, there are
at least two colours say c1 and c2 that are not assigned to any point of S.
If every point in a component of G − S has colour different from c1 and
c2 , then by assigning colour c1 to a point of this component, we get a different
n-colouring of G. Otherwise, by interchanging the colours c1 and c2 in a

71
component of G − S, a different n-colouring of G is obtained. In any case,
G is not uniquely n-colourable, giving a contradiction. Hence G is (n − 1)
connected.

Corollary 4.1.3. In any n-colouring of a uniquely n-colourable graph G, the

subgraph induced by the union of any k colour classes, 2 ≤ k ≤ n, is (k − 1)
connected.

Proof. If the subgraph H induced by the union of any k colour classes, 2 ≤

k ≤ n, had different k-colourings then these k-colourings will induce different
n-colourings for G giving a contradiction. Hence H is uniquely k-colourable.
Hence by theorem 4.1.2 H is (k − 1) connected.

Definition 4.1.5. An assignment of colours to the edges of a graph G so that

no two adjacent edges get the same colour is called an edge colouring or line
coluring of G. An edge colouring of G using n colours is called a n-edge coluring
( or n- line colouring. The edge chromatic number( also called line chromatic
number or chromatic index) χ′ (G) is the minimum number of colours needed
to edge colour G. A graph G is called n-edge colourable if χ′ (G) ≤ n.

Theorem 4.1.7. For any graph G, the edge chromatic number is either ∆ or
∆ + 1.

Theorem 4.1.8. χ′ (Kn ) = n if n is odd (n 6= 1) and χ′ (Kn ) = n − 1 if n is

even.

Proof. If n = 2, the result is obvious. Hence let n > 2. Let n be odd. Now
the edges of Kn can be n-coloured as follows.
Place the vertices of Kn in the form of a regular n-gon. Colour the edges
around the boundary using a different colour for each edge.
Let x be any one of the remaining edges. x divides the boundary into two
segments, one say B1 containing an odd number of edges and other containing
an even number of edges. Colour x with the same colours as the edge that
occurs in the middle of B1 . Note that these two edges are parallel. The result
is a n-edge colouring of Kn since any two edges having the same colour are
parallel and hence are not adjacent. Hence

χ′ (Kn ) ≤ n. (4.1)

72
Since Kn has n points and n is odd, it can have at most (n − 1)/2 mutually
independent edges. Hence each colour class can have at most (n − 1)/2 edges,
so that the number of colour classes is at least n2 21 (n − 1) = n so that

χ′ (Kn ) ≥ n (4.2)

Equations (4.1) and (4.2) imply χ′ (Kn ) = n.

Let n(≥ 4) be even. Let Kn have vertices v1 , v2 , . . . , vn . Colour the edges of
the subgraph Kn−1 induced by the first n−1 points using the method described
above. In this colouring, at each vertex, one colour( the colour assigned to the
edge opposite to this vertex on the boundary) will be missing. Also, these
missing colours are different. This edge colouring of Kn−1 can be extended to
an edge colouring of Kn by assigning the colour that is missing at vi to edge
vi vn for every i, i < n. Hence χ′ (Kn ) ≤ n − 1. Also χ′ (Kn ) ≥ ∆(Kn ) = n − 1.
Hence χ′ (Kn ) = n − 1.

Exercises
1. Give an example of a graph with ∆ = χ′ and a graph with ∆ < χ′ .

2. Show that every outplanar graph is 3-colourable.

3. What is the smallest uniquely 3 colourable graph ?

4. What is the smallest uniquely 3 colourable graph which is not complete

?

5. Show that for any independent set S of points of a critical graph G,

χ(G − S) = χ(G) − 1.

6. Show that the petersen graph has chromatic index 4.

4.2 The Five Colour Theorem

Heawood (1890) showed that one can always colour the vertices of a planar
graph with at most five colours. This is known as the five colour theorem.

Theorem 4.2.1. Every planar graph is 5- colourable.

73
Proof. We will prove the theorem by induction on the number p of points. For
any planar graph having p ≤ 5 points, the result is obvious since the graph is
p-colourable.
Now assume that all planar graphs with p points is 5- colourable for some
p ≥ 5. Let G be a planar graph with p + 1 points. Then G has a vertex
v of degree 5 or less. By induction hypothesis the plane graph G − v is 5-
colourable. Consider a 5-colouring of G − v where ci , 1 ≤ i ≤ 5, are the
colours used. If some colour, say cj is not used in colouring vertices adjacent
to v, then by assigning the colour cj to v the 5 colouring of G − v can be
extended to 5-colouring of G.
Hence we have to consider only the case in which degv = 5 and all the five
colours are used for colouring the vertices adjacent to v. Let v1 , v2 , v3 , v4 , v5 be
the vertices adjacent to v coloured c1 , c2 , c3 , c4 and c5 respectively.
Let G13 denote the subgraph of G − v induced by those vertices coloured c1
or c3 . If v1 and v3 belong to different components of G13 , then a 5 colouring
of G − v can be obtained by interchanging the coloures of vertices in the
component of G13 containing v1 (Since no point of this component is adjacent
to a point with colour c1 or c3 outside this component, this interchange of
colours results in a colouring of G − v. In this 5 colouring no vertex adjacent
to v is coloured c1 , and hence by colouring v with c1 , a 5-coloring of G obtained.
If v1 and v3 are in the component of G13 , then in G there exits a v1 − v3
path of all of whose points are coloured c1 or c3 . Hence there is no v2 − v4 path
all whose points are colouredc2 , c4 .
Hence if G24 denotes the subgraph of G − v induced by the points coloured
c2 or c4 , then v2 and v4 belong to different components of G24 . Hence if we
interchange the colours of the points in the component of G24 containing v2 ,
a new colouring G − v results and in this, no point adjacent to v is coloured
c2 . Hence by assigning colour c2 to v, we can get a 5-colouring of G. This
completes the induction and the proof.

4.3 Chromatic Polynomials

Birkhoff(1912) introduced chromatic polynomials as a possible means of at-
tacking the four colour conjecture. This concept considers the number of ways

74
 v1 c3

v5 c1
v
v2 c3
v4
c1
v3 c1 c

Figure 4.2: A graph explaining 5 colour theorem

of colouring a graph with a given number of colours.

Let G be a labeled graph. A colouring of G from λ colours is a colouring
of G which uses λ or fewer colours. Two colourings of G from λ colours will
be considered different if at least one of the labeled points is assigned different
colours. Let f (G, λ) denote the number of different colourings of G from λ
colours. For example f (K1 , λ) = λ and f (K2 , λ) = λ2 .

Theorem 4.3.1. f (Kn , λ) = λ(λ − 1)(λ − 2) · · · (λ − n + 1)

Proof. The first vertex in Kn can be coloured in λ different ways(as there

are λ colours.) For each colouring of the first vertex, the second vertex can
be coloured in λ − 1 ways ( as there are λ − 1 colours remaining). For each
colouring of the first two verties, the third can be coloured in λ − 2 ways and
so on. Hence f (Kn , λ) = λ(λ − 1)(λ − 2) · · · (λ − n + 1).

Remark 4.3.1. f (Kn , λ) = λn , since each of the n points of Kn may be

coloured independently in λ ways.

Theorem 4.3.2. If G is a graph with k components G1 , G2 , . . . , Gk , then

f (G, λ) = ni=1 f (Gi , λ)
Q

Proof. Number of ways of colouring Gi with λ colours is f (Gi , λ). Since any
choice of λ colouring for G1 , G2 , . . . , Gk can be combined to give a λ colouring
for G, f (G, λ) = ni=1 f (Gi , λ).
Q

Definition 4.3.1. Let u and v be two nonadjacent points in a graph G. The

graph obtained from G by removal of u and v and the addition of a new point
w adjacent to those points to which u or v was adjacent is called an elementary
homomorphism.

75
Theorem 4.3.3. If u and v are nonadjacent points in a graph G and hG
denotes the elementary homomorphism of G which identifies u and v, then
f (G, λ) = f (G + uv, λ) + f (hG, λ) where G + uv denotes the graph obtained
from G by adding the line uv.

Proof.

f (G, λ) = number of colourings of G from λ colours

= (number of colourings G from λ colours in which u and v get diffrent colours)+
(number of colurings of G from λ colurs in which u and v get the same colour)
= number of colurings of G + uv from λ colurs+
number of colurings of hG from λ colours
= f (G, λ) = f (G + uv, λ) + f (hG, λ)

Corollary 4.3.1. Let G be a graph. Then

1. f (G, λ) is a polynomial in λ.

2. f (G, λ) has degree |V (G)|.

3. the constant term in f (G, λ) is 0.

Proof. Theorem 4.3.3, states that f (G, λ) can be written as the sum of f (G1 , λ)
and f (G2 , λ) where G1 has the same number of points as G with one more edge
and G2 has one point less than G. Doing this process repeatedly, f (G, λ) can
P
be written as f (Gi , λ) where each Gi is a complete graph and max|V (Gi )| =
|V (G)|.
Since f (Kn , λ) is a polynomial of degree n, it follows that f (G, λ) is a
polynomial of degree |V (G)|. Since f (Kn , λ) has constant term 0, the constant
P
term in f (Gi , λ) is 0 so that (3) holds.

Note 1. Because of the above corollary f (G, λ) is called the chromatic poly-
nomial of G.

The chromatic polynomial of a graph can be determined using theorem

4.3.1 as illustrated in the following example.

76
 = + = + + K3 + K4

Figure 4.3: An example illusstrating the chromatic polynomial of a graph

Example 4.3.1. Find the chromatic polynomial of the graph G given in figure
A diagram of the graph is used to denote the chromatic polynomial. The
nonadjacent points considered at each step are indicated by u and v. Then

f (G, λ) = [(K5 + K4 ) + (K4 + K3 )] + (K4 + K3 )

= K5 + 3K4 + 2K3
= f (K5 , λ) + 3f (K4, λ) + 2f (K3 , λ)
= λ(λ − 1)(λ − 2)(λ − 3)(λ − 4) + 3λ(λ − 1)(λ − 2)(λ − 3) + 2λ(λ − 1)(λ − 2)
= λ5 − 7λ4 + 19λ3 − 23λ2 + 10λ.

Theorem 4.3.4. If G is a tree with n points, n ≥ 2, then f (G, λ) = λ(λ −

1)n−1 .

Proof. We prove the result by induction on n. For n = 2, G = K2 and hence

f (G, λ) = f (K2 , λ) = λ(λ − 1) so that the theorem holds. Assume that the
chromatic polynomial of any tree with n − 1 points is λ(λ − 1)n−2 . Let G be
a tree with n points. Let v be an end point of G and let u be the unique
point of G adjacent to v. By hypothesis, the tree G − v has λ(λ − 1)n−2 for its
chromatic. The point v can be assigned any colour different that assigned to
u. Hence v may be coloured in λ − 1 ways for each colouring of G − v. Thus

f (G, λ) = (λ − 1)f (G − v, λ) = (λ − 1)(λ − 1)n−2 = λ(λ − 1)n−1

This complete the induction and the proof.

The converse of the above theorem is given below:

77
Theorem 4.3.5. A graph G with n points and f (G, λ) = λ(λ − 1)n−1 is a
tree.

Worked Examples
Problem 16. Prove that the coefficients of f (G, λ) are alternate in sign.
We prove the result by induction on the number of lines q. When q = 0,
f (G, λ) = λp where p is the number of points of G. In this case the polynomial
has just one non-zero coefficient and hence the result is trivially true.
Now assume that the result is true for all graphs with less than q lines. Let
G be a (p, q) graph with q > 0. Let e = uv be an edge of G. Let G1 = G − uv.
Clearly, u and v are nonadjacent in G1 . Hence

f (G1 , λ) = f (G1 + uv, λ) + f (hG1 , λ)

= f (G, λ) + f (hG1 , λ)

Hence
f (G1 , λ) = f (G, λ) + f (hG1 , λ) (4.3)

Now G1 is a (p, q − 1) graph and hG1 is a (p − 1, q1 ) graph where q1 < q. Hence

by induction hypothesis

f (G1 , λ) = λp − α1 λp−1 + αλp−2 − · · · + (−1)p−1 αp−1 λ

and
f (hG1 , λ) = λp−1 − β1 λp−2 + αλp−2 − · · · + (−1)p−1 βp−2 λ

where αi and βi are non negative integers. Hence by equation (4.3), we have

f (G, λ) = λp − (α1 + 1)λp−1 + (α1 + β1 )λp−2 − · · · + (−1)p−1 (αp−1 + βp−2 )λ

This is a polynomial in which the coefficients are alternate in sign.

Problem 17. Prove that if G is a (p, q) graph, the coefficient of λp−1 in f (G, λ)
is −q.
We prove the result by induction on q. If q = 0 then f (G, λ) = λp . Hence
the coefficient of λp−1 is −q. Now assume that the result is true for all graphs

78
with less than q edges. As in the previous problem,

f (G, λ) = f (G1 , λ) − f (hG1 , λ)

Since G1 is a (p, q − 1) graph by induction hypothesis coefficient of λp−1 in

f (G1 , λ) = −(q − 1). Also, the coefficient of λp−1 in f (hG1 , λ) = 1. Hence
the coefficient of λp−1 in f (G, λ) = −(q − 1) − 1 = −q. This complete the
induction and the proof.

Problem 18. Prove that λ4 − 3λ3 + 3λ2 cannot be the chromatic polynomial
of any graph.

Suppose there exits a graph G such that

f (G, λ) = λ4 − 3λ3 + 3λ2

Therefore the number of points in G is 4. Also the number of lines in G is 3.

Case 1 Suppose G is connected. Since q = 3 = p − 1, G is a tree. Hence

f (G, λ) = λ(λ − 1)3 = λ4 − 3λ3 + 3λ2 − λ

which is a contradiction.

Case 2 Suppose G is not connected. Then G = K3 ∪ K1 . Therefore,

f (G, λ) = f (K3 , λ)f (K1 , λ) = λ(λ − 1)(λ − 2)λ = λ4 − 3λ3 + 3λ2

which is again a contradiction.

4.4 Exercises
1. Find the chromatic polynomial of K4 − x where x is a line.

2. Show that λ4 − 3λ3 + 5λ2 − 1 can not be the chromatic polynomial of a

graph.

3. Prove that a graph G is connected iff the coefficients of λ in f (G, λ) is

not zero.

79
 2

1 3

Figure 4.4: A digraph

4. Prove that if k is the least positive integer such that λk has non zero
coefficients in f (G, λ) than G is a graph with k components.

4.5 Directed graphs

Definition 4.5.1. A directed graph (digraph) D is a pair (V, A) where V is a
finite non empty set and A is a subset of V ×V −{(x, x) : x ∈ V }. The elements
of V and A are respectively called vertices(points) and arcs. If (u, v) ∈ A then
the arc (u, v) is said to have u as its initial vertex(tail) and v as its terminal
vertex (head). Also the arc (u, v) is said to join u to v.

Just as graphs, digraphs can also be represented by means of diagrams. In

these diagrams, vertices are denoted by points and arc (u, v) is represented by
means of arrow from u to v. We shall often refer to the diagram of a digraph
as the digraph itself.

Example 4.5.1. Let V = {1, 2, 3} and A = {(1, 2), (2, 3), (1, 3), (3, 1)}. Then
(V, A) is a digraph. The diagrammatic representation of this digraph is shown
in figure 4.8.

Definition 4.5.2. The indegree d− (v) of a vertex v in a digraph D is the

number of arcs having v as its terminal vertex. The outdegree d+ (v) of v is the
number of arcs having v as its initial vertex. The ordered pair (d+ (v), d− (v))
is called the degree pair of v.

Consider the digraph shown in figure 4.8. The degree pairs of the points
1, 2, 3 and 4 are (2, 1), (1, 1), (1, 2) and (0, 0) respectively.

80
 2 b

1 4

a
c
3

Figure 4.5: Two isomorphic digraphs

Theorem 4.5.1. In a digraph D, sum of all the indegrees of all the vertices
is equal to the sum of their out degrees, each sum being equal to the number
of arcs in D.

Proof. Let q denote the number of arcs in D = (V, A). Let B = v∈V d+ (v)
P
P
and C = v∈V d− (v). An arc (u, w) contributes one to the out-degree of u
and one to the in degree of w. Hence each arc contributes 1 to the sum B and
1 to the sum C. Hence B = C = q.

Definition 4.5.3. A digraph D ′ = (V ′ , A′ ) is called a subdigraph of D =

(V, A) if V ′ ⊆ V and A′ ⊆ A. The definition of induced subdigraph is analogous
to that of induced subgraph. The underlying graph G of a digraph D is a graph
having the same vertex set as D and two vertices u and v are adjacent in G
whenever (u, w) or (w, u) is in A.

For example, consider the digraph (V, A) where V = {1, 2, 3, 4} and A =

{(1, 2), (3, 4), (4, 3), (3, 2), (1, 4), (4, 1), (2, 4)} has as its underlying graph. Sim-
ilarly if we are given a graph G we can obtain a digraph from G by giving
orientation to each edge of G. A digraph thus obtained from G is called an
orientation of G.

Definition 4.5.4. Two digraphs D1 = (V1 , A1 ) and D2 = (V2 , A2 ) are said

to be isomorphic (D1 ≃ D2 ) if there exits a bijection f : V1 → V2 such that
(u, w) ∈ A1 iff (f (u), f (w)) ∈ A2 . The function f is called an isomorphism
from D1 to D2 .

Example 4.5.2. Consider the digraphs shown in figure. These graphs are
isomorphic. The isomorphism being f (1) = a, f (2) = b, f (3) = c, f (4) = d.

81
Theorem 4.5.2. If two digraphs are isomorphic then the corresponding points
have the same degree pair.

Proof. Let D1 = (V1 , A1 ) and D2 = (V2 , A2 ) be isomorphic under an isomor-

phism f . Let v ∈ V1 . Let

N(v) = {w : w ∈ V1 and (v, w) ∈ A1 }

N(f (v)) = {w : w ∈ V2 and (f (v), f (w)) ∈ A1 }

Now

w ∈ N(v) ⇔ (v, w) ∈ A1
⇔ (f (v), f (w)) ∈ A2
f (w) ∈ N(f (v)).

Hence |N(v)| = |N(f (v))|. This implies that v and f (v) have the same out
degree pair. Similarly, we can prove that v and f (v) have the same in- degree
pair.

From theorems 4.5.1 and 4.5.2, it is obvious that two isomorphic digraphs
have the same number of vertices and same number of arcs.

Definition 4.5.5. The converse digraph D ′ of a digraph D is obtained from

D by reversing the direction of each arc.

Obviously D and D ′ have the same number of points and arcs. Moreover,
the in-degrees of a point v in D is equal to its out-degree in D ′ and vice versa.

Definition 4.5.6. A digraph D = (V, A) is called complete if for every pair

of distinct points v and w in V , both (v, w) and (w, v) are in A.

Thus if a complete digraph has n vertices then it has n(n − 1) arcs.

Definition 4.5.7. A digraph is called functional if every point has out-degree

one.

If a functional digraph has n vertices then the sum of the out-degrees of

the points is n. Hence by theorem 4.5.1 the number of arcs in the digraph is
n.

82
4.6 Path and Connectedness
Definition 4.6.1. A walk ( directed walk) in a digraph is a finite alternating
sequence W = v0 x1 v1 . . . xn vn of vertices and arcs in which xi = (vi−1 , vi ) for
every arc xi . W is called a walk from v0 to vn or a v0 −vn walk. The vertices v0
and vn are called origin and terminus of W respectively and v1 , v2 , . . . vn−1 are
called its internal vertices. The length of a walk is the number of occurrence of
arcs in it. A walk in which the origin and terminus coincide is called a closed
walk. A path (directed path) is a walk in which all the vertices are distinct.
A cycle (directed cycle or circuit) is a nontrivial closed walk whose origin and
internal vertices are distinct.

If there is a path from u to v then v is said to be reachable from u.

Definition 4.6.2. A digraph is called strongly connected if every pair of points

are mutually reachable. A digraph is called unilaterally connected or unilateral
if for every pair of points, at least one is reachable from the other. A digraph
is called weakly connected or weak if the underlying graph is connected. A
digraph is called disconnected if the underlying graph is disconnected.

The trivial digraph consisting just one point is strong since it does not
contain two distinct points. Obviously

Strongly Connected ⇒ Unilaterally connected ⇒ weakly connected

But the converse is not true.

Theorem 4.6.1. The edges of a connected graph G = (V, E) can be ori-

ented so that the resulting digraph is strongly connected iff every edge of G is
contained in at least one cycle.

Proof. Suppose the edges of G can be oriented so that the resulting digraph
becomes strongly connected.
If possible, let e = vw be an edge of G not lying on any cycle. Now as
soon as e is oriented, one of the vertices u and w becomes non reachable from
the other. Hence an orientation of the required type is not possible, giving
contradiction. Hence every edge of G lies on a cycle.
Conversely, let every edge of G lie on a cycle.

83
 2 b

1 4

a
c
3

Figure 4.6: A digraph

Let S = v1 , v2 , . . . , vn , v1 be a cycle in G. Orient the edges of S so that S

becomes a directed cycle and hence becomes a strongly connected subdigraph.
If V = {v1 , v2 , . . . , vn } then we are through. Otherwise, let w be a vertex of G
not in S such that w adjacent to a vertex vi of S. Let e = vi w. By hypothesis,
e lies on some cycle C. We choose a direction of C and give the orientation
determined by this direction to the edges of C which are not already oriented.
The resulting enlarged oriented graph is also strongly connected as it can be
got from S by a sequence of additionss of simple directed paths( For example,
if v ∈ S and u is a point on a simple directed vi − vj path P added to S then
the enlarged oriented graph the u − vj subpath of P followed by the vj − v
subpath of S give a directed u − v path. Also, the v − vi subpath of S followed
by the vi − u subpath of P give a directed v − u path. This type of argument
can be repeated for each addition of simple directed paths ) This process can
be repeated till we get a strongly connected oriented spanning subgraph of G.
The remaining edges now be oriented in any way. The resulting oriented graph
is strongly connected. This completes the proof.

There are three diffrent kinds of components of a digraph.

Definition 4.6.3. Let D = (V, A) be a digraph.

(a) Let W1 be a maximal subset of V such that for every pair of points
u, v ∈ W1 , u is reachable from v and v is reachable from u. Then the
subdigraph of D induced by W1 is called a strong component of D

(b) Let W2 be a maximal subset of V such that for every pair of points
u, v ∈ W2 , either u reachable from v or v is reachable from u. Then the
subdigraph of D induced by W2 is called a unilateral component of D.

84
 vi w
s

c
vj

Figure 4.7: A digraph

(c) Let W3 be a maximal subset of V such that for every pair of points
u, v ∈ W3 , u and v are joined by a pat in the underlying graph of D.
Then the subdigraph of D induced by W3 is called a weak component of
D.

Note 2. Let D be a digraph. Then each point of D is in exactly one strong

component of D. An arc x lies in exactly one strong component if it lies on
a cycle. There is no strong component containing an arc that does not lie on
any cycle.

Example 4.6.1. Consider the digraph D shown in figure. The strong com-
ponents are those subdigraphs induced by the sets of points A = {1, 2},B =
{3}, C = {4}, D = {5}, E = {6, 7}, F = {8} and G = {9, 10, 11, 12}. The
unilateral components are those induced by the sets points

{1, 2}, {3, 4, 6, 7, 8}, {4, 5}, {5, 6, 7}, {4, 6, 7, 9, 10, 11, 12}, {6, 7, 8, 9, 10, 11, 12}

weak components are those induced by the sets of points

{1, 2}, {3, 4, 5, 6, 7, 8, 9, 10, 11, 12}

Definition 4.6.4. The condensation D ∗ of a digraph D has the strong com-

ponents S1 , S2 , . . . , Sn of D as its points with an arc from Si to Sj whenever
there is at least one arc from Si to Sj in D.

85
 1 4 5

6 7 8
2
9 12

10 11

Figure 4.8: A digraph

Remark 4.6.1. If the condensation of a digraph has a cycle C then the strong
components corresponding to points of C together form a strong component.
This contradicts the maximality of strong components and hence the strong
condensation has no cycles.

Definition 4.6.5. In a digraph D, a closed spanning walk in which each arc

of D occurs exactly once is called an Eulerian trail(Euler Tour). A digraph is
called Eulerian if it has an Eulerian trail.

Theorem 4.6.2. A weak digraph D is an Eulerian iff every point of D has

equal in-degree and out-degree.

Proof. Let D be Eulerian and T be an Eulerian trail in D. Each occur-

rence(occurrence at origin and terminus of T together is to be considered
as a single occurrence) of a given point in T contributes one to d− (v) and
one to d+ (v). Since each arc of D occurs exactly once in T , the contribution
of each arc of D to d− (v) and d+ (v) can be accounted in this way. Hence
d− (v) = d+ (v) for every point v of D.
Conversely let d+ (v) = d− (v) for every point v of D. Since the trivial
digraph is vacuously eulerian, let D have at least two points. Hence every
point of D has positive in-degree and out-degree.
Hence D contains a cycle Z ( Since if you reach a point for the first time,
you can always move out). The removal of the lines of Z results in a spanning

86
subdigraph D1 in which again d− (v) = d+ (v) for every point v. If D1 has
no arcs, then Z is an eulerian trail in D. Otherwise, D1 has a cycle Z1 .
Continuing the above process when a digraph Dn with no arcs is obtained, we
have a partition of the arcs of D into n cycles, n ≥ 2. Among these n cycles,
take two cycles Zi and Zj having a point in common. The walk beginning at v
and consisting of the cycles Zi and Zj in succession is a closed trail containing
the lines of these two cycles. Continuing this process, we construct a closed
trail containing all the arcs of D. Hence D is eulerian.

4.7 Exercise
1. Show that every eulerian digraph is strongly connected. Give an example
to show that the converse is not true.

2. Show that a weak digraph D is eulerian iff the set of arcs of D can be
partitioned into cycles.

3. Show that no strictly weak digraph contains a point whose removal re-
sults in a strong digraph.

87