ALAGAPPA UNIVERSITY

Accredited with ’A+’ Grade by NAAC (CGPA: 3.64) in the Third Cycle and
Graded as Category-I University by MHRD-UGC
KARAIKUDI - 630 003

DIRECTORATE OF DISTANCE EDUCATION

M.Sc

31141

GRAPH THEORY

Copy Right Reserved For Private use only

Contents

1 GRAPHS 1

2 WALK & CYCLE 11

3 TREES, CUT EDGES & VERTICES’S 15

4 BLOCKS 21

5 PERFECT MATCHINGS 27

6 INDEPENDENT SETS, CLIQUES & RAMSEY’S NUMBERS 31

7 EDGE COLOURINGS 35

8 VERTEX COLORING 41

9 PLANER GRAPHS 47

10 BRIDGES 59

11 FIVE COLOR PROBLEM 67

12 DIRECTED GRAPHS & DIRECTED PATH 71

13 NETWORKS 75

14 MAX-FLOW MIN-CUT THEOREM 83

i
Chapter 1

GRAPHS

Unit- I
1.1 GRAPHS
Definition 1.1 A graph is an ordered triple G = (V (G), E(G), IG ), where V (G)
is a nonempty set, E(G) is a set disjoint from V (G), and IG is an ”incidence” map
that associates with each element of E(G), an unordered pair of elements(same or
distinct) of V (G).
Elements of V (G) are called the vertices(or nodes or points) ofG, and elements of
E(G) are called the edges(or lines) of G. If, for the edge e of G, IG (e) = {u, v},
we write IG (e) = uv.

Example 1.1 If V (G) = {v1 , v2 , v3 , v4 , v5 }, E(G) = {e1 , e2 , e3 , e4 , e5 , e6 } and IG is

given by IG (e1 ) = {v1 , v5 },IG (e2 ) = {v2 , v3 },IG (e4 ) = {v2 , v5 },IG (e5 ) = {v2 , v5 },IG (e6 ) =
{v3 , v3 }, then (V (G), E(G), IG ) is a graph.

Figure 1.1:

1
2 CHAPTER 1. GRAPHS

Definition 1.2 If IG (e) = {u, v}, then the vertices u and v are called the end
vertices or ends of the edge e. Each edge is said to join its ends; in this case we
say that e is incident with each one of its ends. Also, the vertices u and v are then
incident with e. A set of two or more edges of a graph G is called a set of multiple
or parallel edges if they have the same ends. If e is the only edge with end vertices
u and v, we write e = uv. An edge for which the two ends are the same is called a
loop at the common vertex.
A vertex u is a neighbor of v in G, if uv is an edge of G, and u 6= v. The set of
all neighbors of v is the open neighborhood of v or the neighbor set of v, and
is denoted by N (v); the set N [v] = N (v) ∪ {v} is the closed neighborhood of v
in G. When G must be explicit, these open and closed neighborhoods are denoted by
NG (v) and NG [v], respectively.
Vertices u and v are adjacent to each other in G if, and only if, there is an edge
of G with u and v as its ends. Two distinct edges e and f are said to be adjacent
if, and only if, they have a common end vertex.

Definition 1.3 A graph is simple if it has no loops and no multiple edges. Thus,
for a simple graph G, the incidence function IG is one-to-one. Hence, an edge of a
simple graph is identified with the pair of its ends. A simple graph therefore may be
considered as an unordered pair (V (G), E(G)), where V (G) is a non-empty set and
E(G) is a set of unordered pairs of elements of V (G)(each edge of the graph being
identified with the pair of its ends).

Figure 1.2:

Example 1.2 In the above graph(1.2) the edge e3 = v2 v4 , edges e4 and e5 form
multiple edges, e6 is a loop at v3 , N (v2 ) = {v3 , v4 , v5 }, N (v3 ) = {v2 }, N [v2 ] =
{v2 , v3 , v4 , v5 } and N [v2 ] = N (v2 ) ∪ {v2 }. Further, v2 and v5 are adjacent vertices
and e3 and e4 are adjacent edges.

Definition 1.4 A graph is called finite if both V (G) and E(G) are finite. A graph
that is not finite is called infinite.

Notation 1.1 We denote by n(G) and m(G) the number of vertices and edges of
the graph G, respectively. The number n(G) is called the order of G and m(G) is
called the size of G.
1.1. GRAPHS 3

Definition 1.5 A graph is said to be labeled, if its n vertices are distinguished

from one another by labels such as v1 , v2 , . . . , vn .

Figure 1.3: A labeled graph G and an unlabeled graph H

Definition 1.6 A simple graph G is said to be complete if every pair of distinct

vertices of G are adjacent in G. Any two complete graphs each on a set of n vertices
are isomorphic; each such graph is denoted by Kn .

Figure 1.4: Some complete graphs

Note 1.1 A simple graph with n vertices can have at most n2 = n(n−1)


2
edges. Kn
has the maximum number of edges among all simple graphs with n vertices. Thus,
for a simple graph G with n vertices, we have 0 ≤ m(G) ≤ n(n − 1)/2.

Definition 1.7 A graph is trivial if its vertex set is singleton and it contains no
edges.

Definition 1.8 Let G be a simple graph. Then the complement Gc of G is defined

by taking V (Gc ) = V (G) and making two vertices u and v adjacent in Gc if, and
only if, they are nonadjacent in G. It is clear that Gc is also a simple graph and
that (Gc )c = G.

Note 1.2 If |V (G)| = n, then clearly, |E(G)| + |E(Gc )| = |E(Kn )| = n(n − 1)/2.

Definition 1.9 A simple graph G is called self-complementary if G ∼

= Gc .
4 CHAPTER 1. GRAPHS

Figure 1.5: Two simple graphs and their complements

Figure 1.6: Self-complementary graphs

1.2 Subgraphs
Definition 1.10 A graph H is called a subgraph of G if V (H) ⊂ V (G), E(H) ⊂
E(G), and IH is the restriction of IG to E(H). If H is a subgraph of G, then G is
said to be a supergraph of H. A subgraph H of a graph G is a proper subgraph
of G if either V (H) 6= V (G) or E(H) 6= E(G).

Figure 1.7: A subgraph of G

Definition 1.11 A subgraph H of G is said to be an induced subgraph of G if

each edge of G having its ends in V (H) is also an edge of H. A subgraph H of G
is a spanning subgraph of G, if V (H) = V (G). The induced subgraph of G with
vertex set S ⊂ V (G) is called the subgraph of G induced by S and is denoted by
G[S].

Definition 1.12 A clique of G is a complete subgraph of G. A clique of G is a

maximal clique of G if it is not properly contained in another clique of G.
1.3. GRAPH ISOMORPHISM 5

Figure 1.8:

Figure 1.9:

Definition 1.13 Let G be a graph, S a proper subset of the vertex set V , and E 0 a
subset of E. The subgraph G[V − S] is said to be obtained from G by the deletion
of S. This subgraph is denoted by G − S.
The spanning subgraph of G with the edge set E/E 0 is the subgraph obtained from G
by deleting the edge subset E 0 . This subgraph is denoted by G − E 0 .

Note 1.3 When a vertex is deleted from G, all the edges incident to it are also
deleted from G, whereas the deletion of an edge from G does not affect the vertices
of G.

1.3 Graph Isomorphism

Definition 1.14 Let G = (V (G), E(G), IG ) and H = (V (H), E(H), IH ) be two
graphs. A graph isomorphism from G to H (written G ∼ = H) is a pair (φ, θ),
where φ : V (G) → V (H) and θ : E(G) → E(H) are bijections with the property that
IG (e) = {u, v} if, and only if, IH (θ(e)) = {φ(u), φ(v)}.

Note 1.4 If (φ, θ) is a graph isomorphism, the pair of the inverse mappings (φ−1 , θ−1 )
is also a graph isomorphism. Also note that the bijection φ satisfies the condition
6 CHAPTER 1. GRAPHS

Figure 1.10: Deletion of vertices and edges from a graph G

Figure 1.11: Isomorphic graphs

that u and v are end vertices of an edge e of G if, and only if, φ(u) and φ(v) are
end vertices of the edge θ(e) in H.

Definition 1.15 If graphs G and H are simple, a bijection φ : V (G) → V (H) such
that u and v are adjacent in G if, and only if, φ(u) and φ(v) are adjacent in H
induces a bijection θ : E(G) → E(H) satisfying the condition that IG (e) = {u, v} if,
and only if, IH (θ(e)) = {φ(u), φ(v)}.
Hence φ itself is referred to as an isomorphism in the case of simple graphs G and
H. Thus if G and H are simple graphs, an isomorphism from G to H is a bijection
φ : V (G) → V (H) such that u and v are adjacent in G if, and only if, φ(u) and
φ(v) are adjacent in H.

1.4 Incidence and adjacency matrices

Definition 1.16 Let G be a graph with n vertices, namely v1 , v2 , . . . , vn . The ad-
jacency matrix of G, with respect to these n vertices of G, is the n × n matrix
A(G) = (aij ) where the (i, j)th entry aij is the number of edges joining the vertex vi
to the vertex vj .

Definition 1.17 Suppose that G has n vertices, namely v1 , v2 , . . . , vn and t edges,

listed as e1 , e2 , . . . , et . The incidence matrix of G, with respect to these particular
1.5. VERTEX DEGREES 7

listing of the vertices and edges of G, is the n × t matrix M (G) = (mij ) where mij
is the number of times that the vertex vi is incident with the edge ej , i.e.,


0 if vi is not an end of ej

mij = 1 if vi is an end of the non-loop ej

2 if vi is an end of the loop ej .


1.5 Vertex degrees

Definition 1.18 Let G be a graph and v ∈ V . The number of edges incident at v
in G is called the degree(or valency) of the vertex v in G and is denoted by dG (v),
or simply d(v) when G requires no explicit reference. A loop at v is to be counted
twice in computing the degree of v.

Figure 1.12: Degrees of vertices of a graph G

Notation 1.2 The minimum(respectively, maximum) of the degrees of the vertices

of a graph G is denoted δ(G) or δ(respectively, ∆(G) or ∆).

Definition 1.19 A graph G is called k-regular, if every vertex of G has degree k.

A graph is said to be regular if it is k-regular for some nonnegative integer k. In
particular, a 3-regular graph is called cubic graph.

Definition 1.20 A spanning 1-regular subgraph of G is called a 1-factor or a per-

fect matching of G.

Definition 1.21 A vertex of degree 0 is known as an isolated vertex of G. A

vertex of degree 1 is called a pendant vertex of G, whereas the unique edge of G
incident to such a vertex of G is a pendant edge of G. A sequence formed by the
degrees of vertices of G is called a degree sequence of G.
8 CHAPTER 1. GRAPHS

Figure 1.13: Degrees of vertices of a graph G

Theorem 1.1 The sum of the degrees of the vertices of a graph is equal to twice
the number of its edges.

Proof 1.1 If e = uv is an edge of G, e is counted once while counting the degrees

of each of u and v(even when u = v). Hence each edge contributes 2 to the sum of
the degrees of the vertices. Thus the m edges of G contributes 2m to the degree sum.

Corollary 1.1 In any graph G, the number of vertices of odd degree is even.

Proof 1.2 Let V1 and V2 be the subsets of vertices of G with odd and even degrees,
respectively. By theorem7.1,
X X X
2m(G) = dG (v) = dG (v) + dG (v).
v∈V v∈V1 v∈V2
P P
As 2m(G) and v∈V2 dG (v) are even, v∈V1 dG (v) is even. Since for each v ∈ V1 ,
dG (v) is odd, |V1 | must be even.

Definition 1.22 A sequence of nonnegative integers d = (d1 , d2 , . . . , dn ) is called

graphical if there exists a simple graph whose degree sequence is d.

Example 1.3 The sequence d = (7, 6, 3, 3, 2, 1, 1, 1) is not graphical, even though

each term of d is a nonnegative integer and the sum of the terms is even. Indeed,
if d were graphical, there must exits a simple graph G with eight vertices whose
degree sequence is d. Let v0 and v1 be the vertices of G whose degrees are 7 and 6,
respectively. Since, G is simple, v0 is adjacent to another five vertices. This means
that in V − {v0 , v1 } there must be at least five vertices of degree at least 2. But this
not the case.

Exercise 1.1 (1) Let G and H be simple graphs and let φ : V (G) → V (H) be a
bijection such that uv ∈ E(G) implies that φ(u)φ(v) ∈ E(H). Show, by means
of an example, that φ need not be an isomorphism from G to H.
(2) Find the complement of the following simple graph.
(3) Show that if G and H are isomorphic graphs, then each pair of corresponding
vertices of G and H have the same degree.
1.5. VERTEX DEGREES 9

(4) Let d = (d1 , d2 , . . . , P

dn ) be the degree sequence of a graph, and r be any positive
integer. Show that ni=1 dri is even.

(5) Prove that in any group of n persons(n ≥ 2), there are at least two with the
same number of friends.

(6) Draw the graphs having the following matrices as their adjacency matrices.

1 1 0 1 0
1 0 1 0 0
(a) 0 1 0 1 0
1 0 1 0 0
0 0 0 0 1

Notes:
10 CHAPTER 1. GRAPHS
Chapter 2

WALK & CYCLE

Unit- II
2.1 Walk
Definition 2.1 A Walk in a graph G is an alternating sequence W : v0 e1 v1 e2 v2 . . . en vn
of vertices and edges beginning and ending with vertices in which vi−1 and vi are the
ends of ei ; v0 is the origin and vn is the terminus of W . The walk W is said to
join v0 and vn ; it is also referred to as a v0 − vn walk.
If the graph is simple, a walk is determined by the sequence of its vertices. The walk
is closed if v0 = vn and is open otherwise.

2.2 Path and Cycle

Definition 2.2 A walk is called a trial if all the edges appearing in the walk are
distinct. It is called a path if all the vertices are distinct. Thus a path in G is
automatically a trial in G.

Definition 2.3 A cycle is a closed trial in which the vertices are all distinct. The
length of a walk is the number of edges in it. A walk of length zero consists of just
a single vertex.

Definition 2.4 A graph that is a cycle of length n is denoted by Cn . Pn denotes

a path on n vertices. In particular, C3 is often referred to as a triangle, C4 as a
square, and C5 as a pentagon. If P = v0 e1 v1 e2 v2 . . . en vn is a path, then P −1 =
vn en vn−1 en−1 vn−2 . . . v1 e1 v0 is also a path and P −1 is called the inverse of the path
P . The subsequence vi ei+1 vi+1 . . . ej vj of P is called the vi − vj section of P .

11
12 CHAPTER 2. WALK & CYCLE

Figure 2.1: Graph illustrating walks,trails,paths and cycles

2.3 Bipartite graphs

Definition 2.5 A graph is bipartite if its vertex set can be partitioned into two
non-empty subsets X and Y such that each edge of G has one end in X and the
other in Y . The pair (X, Y ) is called a bipartition of the bipartite graph. The
bipartite graph G with bipartition (X, Y ) is denoted by G(X, Y ). A simple bipartite
graph G(X, Y ) is complete if each vertex of X is adjacent to all the vertices of Y .
If G(X, Y ) is complete with |X| = p and |Y | = q, then G(X, Y ) is denoted by Kp,q .
A complete bipartite graph of the form K1,q is called a star.

Figure 2.2: Bipartite graphs

Exercise 2.1 (1) Give an example of a nonsimple disconnected graph with δ ≥

n−1
2
.

(2) Show that if G is a self-complementary graph of order n, then n ≡ 0 or 1 (mod

4).

(3) Show that if a self-complementary graph contains a pendant vertex, then it

must have at least another pendant vertex.

(4) Prove that in a simple graph G, the union of two distinct paths joining two
distinct vertices contains a cycle.
2.3. BIPARTITE GRAPHS 13

(5) Show by means of an example that the union of two distinct walks joining two
distinct vertices of a simple graph G need not contain a cycle.

(6) Prove or disprove: If H is a subgraph of G, then

(a) δ(H) ≤ δ(G)

(b) ∆(H) ≤ ∆(G).

(7) In the following graph, find a closed trail of length 7 that is not a cycle:

(8) If δ ≥ 2, then show that G contains a cycle.

Notes:
14 CHAPTER 2. WALK & CYCLE
Chapter 3

TREES, CUT EDGES &

VERTICES’S

Unit- III
3.1 Trees
Definition 3.1 A connected acyclic graph is called a tree.

Theorem 3.1 A simple graph is a tree if, and only if, any two distinct vertices are
connected by a unique path.

Proof 3.1 Let T be a tree. Suppose that two distinct vertices u and v are connected
by two distinct u − v paths. Then their union contains a cycle in T , contradicting
that T is a tree.
Conversely, suppose that any two vertices of a graph G are connected by a unique

Figure 3.1: Examples of isomorphic trees

15
16 CHAPTER 3. TREES, CUT EDGES & VERTICES’S

Figure 3.2: Graph G and two of its spanning trees

path. Then G is obviously connected. Also G can not contain a cycle, since any two
distinct vertices of a cycle are connected by two distinct paths. Hence G is a tree.

Definition 3.2 A spanning subgraph of a graph, which is also a tree, is called a

spannning tree of the graph.

Theorem 3.2 Every connected graph contains a spanning tree.

Theorem 3.3 The number of edges in a tree with n vertices is n − 1. Conversely,

a connected graph with n vertices and n − 1 edges is a tree.

Theorem 3.4 A tree with at least two vertices contains at least two pendant ver-
tices.

Proof 3.2 Consider a longest path of a tree T . The end vertices of this path must
be pendant vertices of T ; otherwise, the path is extendable to a longer path or else
T contains a cycle, a contradiction.

Corollary 3.1 If δ(G) ≥ 2, then G contains a cycle.

Proof 3.3 If G has no cycles, then G is a forest and hence δ(G) ≤ 1 by theorem
(7.1).
3.2. CUT EDGES AND BONDS 17

Figure 3.3: Graph illustrating vertex cuts and edge cuts

3.2 Cut edges and bonds

Definition 3.3 Let G be a nontrivial connected graph with vertex set V and let S
be a nonempty subset of V . For S̄ = V §, let [S, S̄] denote the set of all edges of G
that have one end vertex in S and tha other in S̄. A set of edges of G of the form
[S, S̄] is called an edge cut of G. An edge e is a cut edge of G, if {e} is an edge
cut of G.

Example 3.1 For the above graph, {v2 } and {v3 , v4 } are vertex cuts. The edge
subsets {v3 v5 , v4 v5 }, {v1 v2 } and {v4 v6 } are all edge cuts. Of these, v2 is a cut vertex,
and {v1 v2 } and {v4 v6 } are both cut edges. For the edge cut {v3 v5 , v4 v5 }, we may
take S = {v5 } so that S̄ = {v1 , v2 , v3 , v4 , v6 }.

Theorem 3.5 An edge e = xy of a graph G is a cut edge of a connected graph G

if, and only if, e does not belong to any cycle of G.

Proof 3.4 Let e be a cut edge of G, and let [S, S̄] = {e} be the partition of V defined
by G − e so that x ∈ S and y ∈ S̄. If e belongs to a cycle of G, then [S, S̄] must
contain at least one more edge, contradicting that {e} = [S, S̄]. Hence e cannot
belong toa cycle.
Conversely, assume that e is not a cut edge of G. Then G − e is connected, and
hence there exists an x − y path P in G − e. Then P ∪ {e} is a cycle in G containing
e.

Theorem 3.6 An edge e = xy is a cut edge of a connected graph G if, and only if,
there exist vertices u and v such that e belongs to every u − v path in G.

Proof 3.5 Let e = xy be a cut edge of G. Then G − e has two components, G1 and
G2 . Let u ∈ V (G1 ) and v ∈ V (G2 ) Then clearly, every u − v path in G contains e.
Conversely, suppose that there exist vertices u and v satisfying the condition of the
theorem. THen, there exists no u − v path in G − e so that G − e is disconneccted.
Hence e is a cut edge of G.

Proposition 3.1 A simple cubic connected grasph G has a cut vertex if, and only
if, it has a cut edge.
18 CHAPTER 3. TREES, CUT EDGES & VERTICES’S

3.3 Cut vertex

Definition 3.4 A subset V 0 of the vertex set V (G) of a connected graph G is a
vertex cut of G, if G − V 0 is disconnected; it is a k − vertexcut if |V 0 | = k. V 0 is
then called a separating set of vertices of G. A vertex v of G is a cut vertex of
G, if {v} is a vertex cut of G.

Theorem 3.7 A vertex v of a connected graph G with at least three vertices is a

cut vertex of G if, and only if, there exist vertices u and w of G, distinct from v,
such that v is in every u − w path in G.

Proof 3.6 If v is a cut vertex of G, then G − v is disconnected and has at least two
components, G1 and G2 . Take u ∈ V (G1 ) and w ∈ V (G2 ). THen every u − w path
in G must contain v, as otherwise u and w would belong to the same component of
G − v.
Conversely, suppose that the condition of the theorem holds. Then the deletion of
v destroys every u − w path in G, and hence u and w lie in distinct components of
G − v. THerefore, G − v is disconnected and v is a cut vertex of G.

3.4 Cayley’s formula

Theorem 3.8 The number of spanning trees of a complete labeled graph G on n
vertices is τ (Kn ) = nn−2 where n ≥ 2.

Before we prove Theorem (7.2), we establish two lemmas.

Lemma 3.9 Let (d1 , . . . , dn )) be a sequence of positive integers with ni=1 di = 2(n−
P
1), then there exists a tree T with vertex set {v1 , . . . , vn } and d(vi ) = di ,1 ≤ i ≤ n.

Proof 3.7 It is easy to prove the result by induction on n.

Lemma 3.10 Let P {v1 , . . . , vn },n ≥ 2 be given and let {d1 , . . . , dn } be a set of positive
integers such that ni=1 di = 2(n − 1). Then the number of trees with {v1 , . . . , vn }
(n−2)!
as the vertex set in which vi has degree di ,1 ≤ i ≤ n, is (d1 −1)!...(d n −1)!
.

Proof 3.8 We prove this result by induction on n.

The total number of trees Tn with vertex setP{v1 , . . . , vn } is obtained by summing

over all possible sequences (d1 , . . . , dn ) with ni=1 di = 2n − 2. Hence,
n
X (n − 2)! X
τ (Kn ) = with di = 2n − 2
(d1 − 1)! . . . (dn − 1)!
d ≥1
i i=1
X (n − 2)! Xn
= with ki = n − 2, whereki = di − 1, 1 ≤ i ≤ n
k ≥0
k1 ! . . . kn ! i=1
i
3.4. CAYLEY’S FORMULA 19

Putting x1 = x2 = . . . = xn and m = n − 2 in the multinomial expansion (x1 + · · · +

k
x 1 ...xkn
xn )m = ki ≥0 k11 !...knn! m! with (k1 + k2 + . . . + kn ) = m, we get nn−2 = ki ≥0 k(n−2)!
P P
1 !...kn !
with (k1 + k2 + . . . + kn ) = n − 2. Thus τ (Kn ) = nn−2 .

Exercise 3.1 (1) IF {x, y} is a 2-edge cut of a graph G, show that every cycle of
G that contains x must also contain y.

(2) Prove or disprove: Let G be a simple connected graph with n(G) ≥ 3. Then G
has a cut edge iff G has a cur vertex.

(3) Show that in a graph, the number of edges common to a cycle and an edge cut
is even.

(4) Give an example of a graph with n vertices and n − 1 edges that is not a tree.

Notes:
20 CHAPTER 3. TREES, CUT EDGES & VERTICES’S
Chapter 4

BLOCKS

Unit- IV
4.1 CONNECTIVITY
Definition 4.1 For a nontrivial connected graph G having a pair of nonadjacent
vertices, the minimum k for which there exists a k-vertex cut is called the vertex
connectivity or simply the connectivity of G; it is denoted by κ(G) or simply κ.

Definition 4.2 A set of vertices or edges of a connected graph G is said to dis-

connect the graph if its deletion results in a disconnected graph.

Definition 4.3 The edge connectivity of a connected graph G is the smallest k

for which there exists a k-edge cut.

Definition 4.4 A graph G is r-connected if κ(G) ≥ r. G is r-edge connected

if λ(G) ≥ r.

Theorem 4.1 For any loopless connected graph G, κ(G) ≤ λ(G) ≤ δ(G)

Figure 4.1: 1-connected graph

21
22 CHAPTER 4. BLOCKS

Figure 4.2: A graph G and its blocks

Theorem 4.2 The connectivity and edge connectivity of a simple cubic graph G are
equal.

Definition 4.5 A family of two or more paths in a graph G is said to be internally

disjoint if no vertex of G i an internal vertex of more than one path in the family.

Theorem 4.3 A graph G with at least three vertices is 2-connected iff any two
vertices of G are connected by at least two internally disjoint paths.

4.2 Blocks
Definition 4.6 A graph G is nonseparable if it si nontrivial, connected and has
no cut vertices. A block of a graph G is a maximal nonseparable subgraph of G. If
G has no cut vertex, G itself is a block.

Theorem 4.4 If C is any cycle of a simple block G with at least three vertices, then
there exists a sequence of non-separable subgraphs c = B0 , B1 , . . . , Br = G such that
Bi+1 is an edge-disjoint union of Bi and a path Pi ,where the only vertices common
to Bi and Pi are the end vertices of Pi ,0 ≤ i ≤ r − 1.

Proof 4.1 Assume that we already determined Bi . If Bi 6= G, there exists an edge

e = uv not belonging to Bi , but with u in Bi . If v also belongs to Bi , take Pi = uv and
Bi+1 = Bi ∪Pi . Otherwise e = uv is an edge of G having only one of its ends, namely
u, in Bi . Let u0 be any other vertex of Bi . Then , since G is 2-connected, e and u0
belong to a common cycle Ci . Let ui be the first vertex of Bi in the u − u0 section
C 0 of Ci containing v, and let Pi be the u − ui section of C 0 . Define Bi+1 = Bi ∪ Pi .
Then Bi+1 is non-separable, and the proof follows by induction on i.
4.3. EULER TOURS 23

Figure 4.3: (a) Eulerian graph and (b) Non-Eulerian graph

Figure 4.4: (a) Hamiltonian graph, (b) Non-Hamiltonian graph

4.3 Euler tours

Definition 4.7 An Euler trail in a graph G is a spanning trail in G that contains
all the edges of G. An Euler tour of G is a closed Eulertrail of G. G is called
Eulerian if G has an Euler tour.

Theorem 4.5 For a connected graph G, the following statements are equivalent:
(i) G is Eulerian
(ii) The degree of each vertex of G is an even positive integer.
(iii) G is an edge-disjoint union of cycles.

Theorem 4.6 A graph is Eulerian if, and only if, each edge e of G belongs to an
odd number of cycles of G.

Corollary 4.1 A graph is Eulerian if, and only if, it has an odd number of cycle
decompositions

4.4 Hamiltonion cycles

Definition 4.8 A graph is called Hamiltonian if it has a spanning cycle. These
are often called Hamiltonian cycle of G.

Theorem 4.7 If G is Hamiltonian, then for every non-empty proper subset S of

V , ω(G − S) ≤ |S|.
24 CHAPTER 4. BLOCKS

Figure 4.5: Closure of a graph

Proof 4.2 Let C be a Hamiltonian cycle in G. Then, since C is a spanning subgraph

of G, ω(G − S) ≤ ω(C − S). If |S| = 1, C − S is a path, and therefore ω(C − S) =
1 = |S|. The removal of a vertex from a path P results in one or two components,
according to whether the removed vertex is an end vertex or an internal vertex of P ,
respectively. Hence, by induction, the number of components in C − S cannot exceed
|S|. This proves that ω(G − S) ≤ ω(C − S) ≤ |S|.

Theorem 4.8 Let G be a simple graph with n ≥ 3 vertices. If for every pair of
nonadjacent vertices u, v of G, d(u) + d(v) ≥ n, then G is Hamiltonian.

4.5 Closure of a graph

Definition 4.9 The closure of a graph G, denoted by cl(G) is defined to be the
supergraph of G obtained from G by recursively joining pairs of nonadjacent vertices
whose degree sum is at least n until no such pair exists.

Theorem 4.9 The closure cl(G) of a graph G is well-defined.

Theorem 4.10 If cl(G) is Hamiltonian, then G is Hamiltonian.

Corollary 4.2 If cl(G) is complete, then G is Hamiltonian.

4.6. CHAVATAL THEOREM FOR NON-HAMILTONIAN SIMPLE GRAPHS 25

Figure 4.6: Graphs for proof of the theorem (4.11)

4.6 Chavatal theorem for non-Hamiltonian sim-

ple graphs
Theorem 4.11 If for a simple 2-connected graph G, α ≤ κ, then G is Hamiltonian.

Proof 4.3 Suppose α ≤ κ but G is not Hamiltonian. Let C : v0 v1 . . . vp−1 be a

longest cycle of G. We fix this orientation on C. By Dirac’s theorem p ≥ κ. Let
v ∈ V (G) V (C). Then by Menger’s theorem there exist κ internally disjoint paths
P1 , . . . , Pκ from v to C. Let vi1 , vi2 , . . . , vik be the end vertices of the paths on C. No
two of the consecutive vertices vi1 , vi2 , . . . , vik , vi1 can be adjacent vertices of C, since
oterwise we get a cycle of G longer than C. Hence, between any two consecutive
vertices of {vi1 , vi2 , . . . , vik , vi1 }, there exists at least one vertex of G. Let uij be the
vertex next to vij in the vij − vij +1 path along C.
We claim that {ui1 , ui2 , . . . , uik } is an independent set of G. Suppose uij is adjacent
to uim , m ≥ j; then uij , . . . vij+1 . . . vim vvij−1 . . . uim uij is a cycle of G longer than C,
a contradiction.
Clearly, {v, ui1 , ui2 , . . . , vik } is an independant set of G.(Otherwise, vuim ∈ E(G)
for some m. Then vuim . . . im+1 . . . vik . . . vi1 . . . vim Pm−1 v is a cycle longer than C,
a contradiction) But than α ≥ κ, a contradiction to our assumption. Thus G is
Hamiltonian.

Exercise 4.1 (1) Determine the closure of the following graph.

(2) Does there exist an Eulerian graph with (i) An even number of vertices and an
odd number of edges? (ii) An odd number of vertices and an even number of edges?
Draw such a graph if it exists.
(3) Show that in a tree, any path of maximum length contains the center of the tree.
(4) Show that a simple graph with ω components is a forest if and only if m = n − ω.
26 CHAPTER 4. BLOCKS

(5) A vertex v of a tree T with at least three vertices is a cut vertex of T if and only
if v is not a pendant vertex.
(6) Prove or disprove: If H is a subgraph of G; then κ(H) ≤ κ(G)

Notes:
Chapter 5

PERFECT MATCHINGS

Unit- V
5.1 MATCHINGS
Definition 5.1 A subset M of E is called a matching in G if its elements are links
and no two are adjacent in G; the two ends of an edge in M are said to be matched
under M . A matching M saturates a vertex v, and v is said to be M − saturated, if
some edge of M is incident with v; otherwise, v is M − unsaturated. If every ver-
tex of G is M −saturated, the matching M is perfect. M is a maximum matching if
G has no matching M 0 with |M 0 | > |M |; clearly, every perfect matching is maximum.

Let M be a matching in G. An M −alternating path in G is a path whose edges

are alternately in E M and M . For example, the path v5 v8 v1 v7 v6 in the graph of
figure ?? is an M −alternating path. An M −augmenting path is an M −alternating
path whose origin and terminus are M −unsaturated.

Theorem 5.1 (Berge, 1957) A matching M in G is a maximum matching if and

only if G contains no M −augmenting path.

Proof: Let M be a matching in G, and suppose that G contains an M-

augmenting path v0 v1 ...v2m+1 . Define M 0 ⊆ E by

M 0 = (M \ v1 v2 , v3 v4 , ..., v2m−1 ) ∪ {v0 v1 , v2 v3 , ...v2m v2m+1 }

Then M 0 is a matching in G, and |M 0 | = |M | + 1. Thus M is not a maximum

matching.
Conversely, suppose that M is not a maximum matching, and let M 0 be a maximum
matching in G. Then

|M 0 | > |M | (5.1)

27
28 CHAPTER 5. PERFECT MATCHINGS

Set H = G[M 4 M 0 ], where M 4 M 0 denotes the symmetric difference of M and

M 0.
Each vertex of H has degree either one or two in H, since it can be incident with at
most one edge of M and one edge of M 0 Thus each component of H is either an even
cycle with edges alternately in M and M 0 , or else a path with edges alternately in M
and M 0 . By equation 5.1, H contains more edges of M 0 than of M , and therefore
some path component P of H must start and end with edges of M 0 The origin and
terminus of P , being M 0 −saturated in H, are M −unsaturated in G. Thus P is an
M −augmenting path in G.

Exercises
5.1.1 (a) Show that every k−cube has a perfect matching (k ≥ 2).
(b) Find the number of different perfect matchings in k2n and kn,n .
5.1.2 Show that a tree has at most one perfect matching.
5.1.3 For each k > 1, find an example of a k−regular simple graph that has no
perfect matching.

5.2 MATCHINGS AND COVERINGS IN BIPAR-

TITE GRAPHS
For any set S of vertices in G, we define the neighbour set of S in G to be the set of
all vertices adjacent to vertices in S; this set is denoted by NG (S). Suppose, now,
that G is a bipartite graph with bipartition (X, Y ). In many applications one wishes
to find a matching of G that saturates every vertex in X. Necessary and sufficient
conditions for the existence of such a matching were first given by Hall (1935).

Theorem 5.2 Let G be a bipartite graph with bipartition (X, Y ). Then G contains
a matching that saturates every vertex in X if and only if

|N (S)| ≥ |S| for all S ⊆ X (5.2)

Proof:
Suppose that G contains a matching M which saturates every vertex in X, and let S
be a subset of X. Since the vertices in S are matched under M with distinct vertices
in N (S), we clearly have |N (S)| ≥ |S|.

Conversely, suppose that G is a bipartite graph satisfying equation 13.2, but that
G contains no matching saturating all the vertices in X. We shall obtain a contra-
diction. Let M ∗ be a maximum matching in G. By our supposition, M ∗ does not
saturate all vertices in X. Let u be an M ∗ −unsaturated vertex in X, and let Z
denote the set of all vertices connected to u by M ∗ − alternating paths. Since M ∗ is
a maximum matching, it follows from theorem 1 that u is the only M ∗ −unsaturated
5.3. PERFECT MATCHINGS 29

vertex in Z. Set S = Z ∩ X and T = Z ∩ Y . Clearly, the vertices in S \ {u} are

matched under M ∗ with the vertices in T . Therefore

|T | = |S| − 1 (5.3)

and N (S) ⊇ T . In fact, we have

N (S) = T (5.4)

since every vertex in N (S) is connected to u by an M ∗ −alternating path. But

equation 5.3 and 5.4 imply that |N (S)| = |S| − 1 < |S| contradicting assumption
13.2.

Corollary 1:
If G is a k−regular bipartite graph with k > 0, then G has a perfect matching.

Exercises:
1. Show that it is impossible, using 1 × 2 rectangles, to exactly cover an 8 × 8 square
from which two opposite 1 × 1 corner squares have been removed.

2. (A)Show that a bipartite graph G has a perfect matching if and only if |N (S)| ≥
|S| for all S ⊆ V .

5.3 PERFECT MATCHINGS

A necessary and sufficient condition for a graph to have a perfect matching was ob-
tained by Tutte (1947). The proof given here is due to Lovasz (1973). A component
of a graph is odd or even according as it has an odd or even number of vertices. We
denote by o(G) the number of odd components of G.

Theorem 5.3 G has a perfect matching if and only if o(G−S) ≤ |S| for all S ⊂ V .

Theorem 5.4 Every 3-regular graph without cut edges has a perfect matching.

Notes:
30 CHAPTER 5. PERFECT MATCHINGS
Chapter 6

INDEPENDENT SETS,
CLIQUES & RAMSEY’S
NUMBERS

Unit - VI
6.1 INDEPENDENT SETS & CLIQUES
A subset S of V is called an independent set of G if no two vertices of S are adjacent
in G. An independent set is maximum if G has no independent set S 0 with |S 0 | > |S|.
Recall that a subset K of V such that every edge of G has at least one end in K is
called a covering of G.

Theorem 6.1 A set S is an independent set of G if and only if V \ S is a covering

of G.
Proof:
By definition, S is an independent set of G if and only if no edge of G has both ends
in S or, equivalently, if and only if each edge has at least one end in V \ S. But this
is so if and only if V \ S is a covering of G

The number of vertices in a maximum independent set of G is called the indepen-

dence number of G and is denoted by α(G); similarly, the number of vertices in a
minimum covering of G is the covering number of G and is denoted by β(G).

Theorem 6.2 α + β = ν.
Proof:
Let S be a maximum independent set of G, and let K be a minimum covering of
G. Then, by theorem 7.1, V \ K is an independent set and V \ S is a covering.
Therefore,
ν − β = |V \ K| ≤ α (6.1)

31
32 CHAPTER 6. INDEPENDENT SETS, CLIQUES & RAMSEY’S NUMBERS

ν − α = |V \ S| ≥ β (6.2)

combining equation 6.1 and 6.2, we have α + β = ν.

The edge analogue of an independent set is a set of links no two of which are adja-
cent, that is, a matching. The edge analogue of a covering is called an edge covering.
An edge covering of G is a subset L of E such that each vertex of G is an end of
some edge in L. Note that edge coverings do not always exist; a graph G has an
edge covering if and only if δ > 0. We denote the number of edges in a maximum
matching of G by α0 (G), and the number of edges in a minimum edge covering of
G by β 0 (G); the numbers α0 (G) and β 0 (G) are the edge independence number and
edge covering number of G, respectively.

Matchings and edge coverings are not related to one another as simply as are in-
dependent sets and coverings; the complement of a matching need not be an edge
covering, nor is the complement of an edge covering necessarily a matching. How-
ever, it so happens that the parameters α0 and β 0 are related in precisely the same
manner as are α and β.

Theorem 6.3 (Gallai, 1959) If δ > 0, then α0 + β 0 = ν.

Proof:
Let M be a maximum matching in G and let U be the set of M −unsaturated vertices.
Since δ > 0 and M is maximum, there exists a set E 0 of |U | edges, one incident
with each vertex in U . Clearly, M ∪ E 0 is an edge covering of G, and so

α0 + β 0 ≤ ν (6.3)

Now let L be a minimum edge covering of G, set H = G[L] and let M be a maximum
matching in H. Denote the set of M −unsaturated vertices in H by U . Since M is
maximum, H[U ] has no links and therefore

|L| − |M | = |L \ M | ≥ |U | = ν − 2|M |

Because H is a subgraph of G, M is a matching in G and so

α0 + β 0 ≥ |M | + |L| ≥ ν (6.4)

Combining equation 6.3 and 6.4, we have α0 + β 0 = ν

Theorem 6.4 In a bipartite graph G with δ > 0, the number of vertices in a maxi-
mum independent set is equal to the number of edges in a minimum edge covering.

6.2 RAMSEY’S THEOREM

In this section we deal only with simple graphs. A clique of a simple graph G is a
subset S of V such that G[S] is complete. Clearly, S is a clique of G if and only if
6.2. RAMSEY’S THEOREM 33

S is an independent set of Gc , and so the two concepts are complementary.

If G has no large cliques, then one might expect G to have a large independent set.
That this is indeed the case was first proved by Ramsey (1930). He showed that,
given any positive integers k and l , there exists a smallest integer r(k, l) such that
every graph on r(k, l) vertices contains either a clique of k vertices or an independent
set of l vertices. For example, it is easy to see that

r(1, l) = r(k, 1) = 1 (6.5)

and

r(2, l) = l, r(k, 2) = k (6.6)

The numbers r(k, l) are known as the Ramsey numbers.

Theorem 6.5 For any two integers k ≥ 2 and l ≥ 2

r(k, l) ≤ r(k, l − 1) + r(k − 1, l) (6.7)

Furthermore, if r(k, l − 1) and r(k − 1, l) are both even, then strict inequality holds
in equation 6.7.
Proof:

Proof Let G be a graph on r(k, l − 1) + r(k − 1, l) vertices, and let v ∈ V . We

distinguish two cases:
(i) v is nonadjacent to a set S of at least r(k, l − 1) vertices, or
(ii) v is adjacent to a set T of at least r(k − 1, l) vertices.
Note that either case (i) or case (ii) must hold because the number of vertices to
which v is nonadjacent plus the number of vertices to which t; is adjacent is equal
to r(k, l − 1) + r(k − 1, l) − 1.

In case (i), G[S] contains either a clique of k vertices or an independent set of l − 1

vertices, and therefore G[S ∪ {v}] contains either a clique of k vertices or an inde-
pendent set of l vertices. Similarly, in case (ii), G[T ∪ {v}] contains either a clique
of k vertices or an independent set of l vertices. Since one of case (i) and case (ii)
must hold, it follows that G contains either a clique of k vertices or an independent
set of l vertices. This proves equation 6.7.

Now suppose that r(k, l − 1) and r(k?1, l) are both even, and let G be a graph on
r(k, l?1) + r(k?1, l) − 1 vertices. Since G has an odd number of vertices, it follows
from corollary 1 that some vertex v is of even degree; in particular, v cannot be
adjacent to precisely r(k − l, l) − 1 vertices. Consequently, either case (i) or case (ii)
above holds, and therefore G contains either a clique of k vertices or an independent
set of l vertices. Thus, r(k, l) ≤ r(k, l − 1) + r(k − 1, l) − 1 as stated.

Theorem 6.6 r(k, k) > 2k/2

34 CHAPTER 6. INDEPENDENT SETS, CLIQUES & RAMSEY’S NUMBERS

Theorem 6.7 If m = mink, l then r(k, l) ≥ 2m/2

Notes:
Chapter 7

EDGE COLOURINGS

Unit - VII
7.1 EDGE CHROMATIC NUMBER

A k-edge colouring C of a loopless graph G is an assignment of k colours, 1, 2, . . . , k,

to the edges of G. The colouring C is proper if no two adjacent edges have the
same colour. Alternatively, a k-edge colouring can be thought of as a partition
(E1 , E2 , . . . , Ek ) of E, where B denotes the (possibly empty) subset of E assigned
colour i. A proper k-edge colouring is then a k-edge colouring (E1 , E2 , . . . , Ek ) in
which each subset Ei is a matching.
G is k-edge colourable if G has a proper k-edge-colouring- Trivially, every loopless
graph G is -edge-colourable; and if G is k-edge-colourable, then G is also l-edge-
colourable for every l > k. The edge chromatic number χ0 (G), of a loopless graph
G, is the minimum k for which G is k-edge- colourable. G is k-edge-chromatic if
χ0 (G) = k. It can be readily verified that the graph of figure 6.1 has no proper
3-edge colouring. This graph is therefore 4-edge-chromatic. Clearly, in any proper
edge colouring, the edges incident with any one vertex must be assigned different
colours. It follows that
χ0 ≥ k (7.1)

Lemma 7.1 Let G be a connected graph that is not an odd cycle. Then G has a
2-edge colouring in which both colours are represented at each vertex of degree at
least two.

Proof 7.1 We may clearly assume that G is nontrivial. Suppose, first, that G is

35
36 CHAPTER 7. EDGE COLOURINGS

Figure 7.1: 1

eulerian. If G is an even cycle, the proper 2-edge colouring of G has the required
property. Otherwise, G has a vertex v0 of degree at least four. Let be an Euler tour
of G, and set v0 e1 v1 . . . e v0

E1 = {ei |i odd}andE2 = {ei |i even} (7.2)

Then the 2-edge colouring (E1 , E2 ) of G has the required property, since each vertex
of G is an internal vertex of v0 e1 v1 . . . e v0 . If G is not eulerian, construct a new
graph G∗ by adding a new vertex v0 and joining it to each vertex of odd degree in G.
Clearly G∗ is eulerian. Let v0 e1 v1 . . . e v0 be an Euler tour of G∗ and define E1 and
E2 as in (7.1). It is then easily verified that the 2-edge colouring (E1 ∩ E, E2 ∩ E)
of G has the required property.

Lemma 7.2 Let C = (E1 , E2 , . . . , Ek ) be an optimal k-edge colouring of G. If there

is a vertex u in G and colours i and j such that i is not represented at u and j is
represented at least twice at u, then the component of G[Ei ∪ Ej ] that contains u is
an odd cycle.

Proof 7.2 Let u be a vertex that satisfies the hypothesis of the lemma, and denote
by H the component of G[Ei ∪ Ej ] containing u. Suppose that H is not an odd cycle.
Then, by lemma 7.1 H has a 2-edge colouring in which both colours are represented at
each vertex of degree at least two in H. When we recolour the edges of H with colours
i and j in this way, we obtain a new k-edge colouring C 0 = (E10 , E20 , . . . , Ek0 ) of G.
Denoting by c0 (v) the number of distinct colours at v in the colouring C 0 , we have
7.2. VIZING’S THEOREM 37

c0 (u) = c(u) + 1 since, now, both i and j are represented at u, and also c0 (v) ≥ c(v)
for u 6= v. Thus v∈V c0 (v) > v∈V c(v), contradicting the choice of C 0 . It follows
P P

that H is indeed an odd cycle .

Theorem 7.1 If G is bipartite, then χ0 = ∆ + 1.

Proof 7.1 Let G be a graph with χ0 > ∆ + 1, let C 0 = (E10 , E20 , . . . , E∆

0
) be an
optimal A-edge colouring of G, and let u be a vertex such that c(u) < d(u). Clearly,
u satisfies the hypothesis of lemma 7.2. Therefore G contains an odd cycle and so
is not bipartite. It follows from (7.1) that if G is bipartite, then χ0 = ∆ + 1.

Exercises:

1. Show that the Petersen graph is 4-edge-chromatic.

2. Describe a good algorithm for finding a proper A-edge colouring of a bipartite

graph G.

7.2 VIZING’S THEOREM

As has already been noted, if G is not bipartite then we cannot necessarily conclude
that χ0 = ∆. An important theorem due to Vizing A964) and, independently, Gupta
A966), asserts that, for any simple graph G, either χ0 = ∆ or χ0 = ∆ + 1. The proof
given here is by Fournier 1973).

Theorem 7.2 If G is simple, then either χ0 = ∆ or χ0 = ∆ + 1.

Proof 7.2 Let G be a simple graph. By virtue of (7.1) we need only show that
χ0 ≤ ∆ + 1 Suppose, then, that χ0 > ∆ + 1. Let C = (E1 , E2 , . . . , E∆ ) be an optimal
(∆ + 1)-edge colouring of G and let u be a vertex such that c(u) < d(u). Then there
exist colours i0 and i1 such that i0 is not represented at u, and i1 is represented at
least twice at u. Let uvi have colour i1 , as in figure 7.2a Since d(v1 ) < ∆ + 1, some
colour i2 is not represented at v1 . Now i2 must be represented at u since otherwise,
by recolouring uvx with i2 , we would obtain an improvement on C. Thus some edge
uv2 has colour i2 . Again, since d(v2 ) < ∆ + 1, some colour i3 is not represented at
38 CHAPTER 7. EDGE COLOURINGS

Figure 7.2: 2

v2 ; and i3 must be represented at u since otherwise, by recolouring uv1 with i2 and

uv2 with i3 , we would obtain an improved (∆ + 1)-edge colouring. Thus some edge
uv3 has colour i3 . Continuing this procedure we construct a sequence i1 , i2 , . . . of
vertices and a sequence i1 , i2 , . . . of colours, such that

(i) uvj has colour ij

(ii) ij+1 is not represented at vj

Since the degree of u is finite, there exists a smallest integer l such that, for
some k < l,

(iii) il+1 = ik .

The situation is depicted in figure 7.2a. We now recolour G as follows. For 1 ≤ j ≤

k − 1 recolour uvj with colour ij+i , yielding a new (∆ + 1)-edge colouring χ0 > ∆ + 1,
let C 0 = (E10 , E20 , . . . , E∆+1
0
). (7.2b) Clearly c0 (v) ≥ c(v) for all v ∈ V and therefore
C 0 is also an optimal (∆ + 1)-edge colouring of G. By lemma 7.2, the component
H 0 of G[Ei00 ∪ Ei0k ] that contains u is an odd cycle. Now, in addition, recolour uvj
with colour ij+i , k ≤ j ≤ l?1, and uvl with colour ik , to obtain a (∆ + 1)-edge
colouring C 0 = (E100 , E200 , . . . , E∆+1
00
) (7.2c). As above c00 (v) ≥ c(v) for all v ∈ V and
7.2. VIZING’S THEOREM 39

the component H 00 of G[Ei000 ∪ Ei00k ] that contains u is an odd cycle. But, since vk has
degree two in H 0 , uk clearly has degree one in H 00 This contradiction establishes the
theorem.
Notes:

Exercises:

1. Show that if G is loopless, then G has a A-regular loopless supergraph.

2. G is called uniquely k-edge-colourable if any two proper k-edge colourings of

G induce the same partition of E. Show that every uniquely 3-edge-colourable
3-regular graph is hamiltonian.
40 CHAPTER 7. EDGE COLOURINGS
Chapter 8

VERTEX COLORING

Unit - VIII

8.1 CHROMATIC NUMBER

A k-vertex colouring of G is an assignment of k colours, 1, 2, . . . , k, to the vertices of

G; the colouring is proper if no two distinct adjacent vertices have the same colour.
Thus a proper k-vertex colouring of a loopless graph G is a partition (V1 , V2 , . . . , Vk )
of V into k (possibly empty) independent sets. G is k-vertex-colourable if G has a
proper k-vertex colouring. It will be convenient to refer to a ’proper vertex colouring’
as, simply, a colouring and to a ’proper k-vertex colouring’ as a k-colouring; we shall
similarly abbreviate ’k-vertex-colourable’ to k-colourable. Clearly, a graph is k-
colourable if and only if its underlying simple graph is k-colourable. Therefore, in
discussing colourings, we shall restrict ourselves to simple graphs; a simple graph is
1-colourable if and only if it is empty, and 2-colourable if and only if it is bipartite.
The chromatic number, χ(G), of G is the minimum k for which G is k-colourable; if
χ(G) = k, G is said to be k-chromatic. A 3-chromatic graph is shown in figure 8.1.
It has the indicated 3-colouring, and is not 2-colourable since it is not bipartite. It
is helpful, when dealing with colourings, to study the properties of a special class
of graphs called critical graphs. We say that a graph G is critical if χ(H) < χ(G)
for every proper subgraph H of G. Such graphs were first investigated by Dirac
A952). A k-critical graph is one that is k-chromatic and critical; every k-chromatic
graph has a k-critical subgraph. A 4-critical graph, due to Grotzsch A958), is shown
in figure 8.2. An easy consequence of the definition is that every critical graph is
connected. The following theorems establish some of the basic properties of critical

41
42 CHAPTER 8. VERTEX COLORING

graphs.

Theorem 8.1 If G is k -critical, then δ ≥ k − 1

Proof 8.1 By contradiction. If possible, let G be a k-critical graph with δ < k − 1,

and let v be a vertex of degree δ in G. Since G is k-critical, G−v is (k−1)-colourable.
Let (V1 , V2 , . . . , Vk−1 ) be a (k − 1)-colouring of G − v. By definition, v is adjacent in
G to δ < k − 1 vertices, and therefore v must be nonadjacent in G to every vertex
of some Vj . But then (V1 , V2 , . . . , Vj ∪ v, . . . , Vk−1 ) is a (k − l)-colouring of G, a
contradiction. Thus δ < k − 1

Theorem 8.2 In a critical graph, no vertex cut is a clique.

Proof 8.2 By contradiction. Let G be a k-critical graph, and suppose that G has a
vertex cut S that is a clique. Denote the S -components of G G1 , G2, . . . , Gn . Since
G is k-critical, each Gi is (k − 1)-colourable. Furthermore, because S is a clique,
the vertices in S must receive distinct colours in any (k − 1)-colouring of Gi. It
follows that there are (k − 1)-colourings of G1 , G2 , . . . , Gn which agree on S. But
these colourings together yield a (k − 1)-colouring of G, a contradiction.

Corollary 8.1 Every critical graph is a block.

Proof 8.1 If t¿ is a cut vertex, then v is a vertex cut which is also, trivially, a clique.
It follows from theorem 8.2 that no critical graph has a cut vertex; equivalently, every
critical graph is a block

Corollary 8.2 Let G be a k-critical graph with a 2-vertex cut {u, v}. Then

1. G = G1 ∪ G2 , where G1 is a {u, v}-component of type i (i = 1, 2), and

2. both G1 + uv and G2 · uv are k-critical .

8.2 Brooks’ theorem

Theorem 8.3 If G is a connected simple graph and is neither an odd cycle nor a
complete graph, then χ ≤ ∆.
8.3. HAJO’S’ CONJECTURE 43

Proof 8.2 Let G be a k-chromatic graph which satisfies the hypothesis of the theo-
rem. Without loss of generality, we may assume that G is k-critical. By corollary 8.1,
G is a block. Also, since 1-critical and 2-critical graphs are complete and 3-critical
graphs are odd cycles, we have k ≥ 4 If G has a 2-vertex cut {u, v}, corollary8.2
gives 2∆ ≥ d(u) + d(c) ≥ 3k − 5 ≥ 2k − 1. This implies that χ = k ≤ ∆, since 2∆
is even.
Assume, then, that G is 3-connected. Since G is not complete, there are three vertices
u, v and w in G such that uv, vw ∈ E and uw ∈
/ E . Set u = V1 and w = v2 let
{v3 , v4 , . . . vv = v}be any ordering of the vertices of G − {u, w} such that each vi
is adjacent to some vj with j > i. Finally, since vv is adjacent to two vertices of
colour 1 (namely t)i and D2), it is adjacent to at most ∆ − 2 other colours and can
be assigned one of the colours 2, 3, . . . , ∆,

8.3 Hajo’s’ Conjecture

A subdivision of a graph G is a graph that can be obtained from G by a sequence

of edge subdivisions. A subdivision of K4 is shown in figure 8.5. Although no
necessary and sufficient condition for a graph to be k- chromatic is known when
k ≥ 3, a plausible necessary condition has been proposed by Hajos A961): if G
is k -chromatic, then G contains a subdivi- sion of Kk . This is known as Hajos’
conjecture. It should be noted that the condition is not sufficient; for example, a
4-cycle is a subdivision of K3 , but is not 3-chromatic.
For k = 1 and k = 2, the validity of Hajos’ conjecture is obvious. It is also easily
verified for k = 3, because a 3-chromatic graph necessarily contains an odd cycle,
and every odd cycle is a subdivision of K3. Dirac A952) settled the case k = 4.

Theorem 8.4 If G is 4-chromatic, then G contains a subdivision of K4 .

Proof 8.3 Let G be a 4-chromatic graph. Note that if some subgraph of G contains
a subdivision of K4 , then so, too, does G. Without loss of generality, therefore, we
may assume that G is critical, and hence that G is a block with ∆ ≥ 3. If v = 4,
then G is K4 and the theorem holds trivially. We proceed by induction on v.
Assume the theorem true for all 4-chromatic graphs with fewer than n vertices, and
let v(G) = n > 4. Suppose, first, that G has a 2-vertex cut {u, v}. By theorem
8.3, G has two {u, v}-components G1 and G2 , where G1 + uv is 4-critical. Since
44 CHAPTER 8. VERTEX COLORING

Figure 8.1: Subdivision of K4

v(G1 + uv) < v(G), we can apply the induction hypothesis and deduce that G1 + uv
contains a subdivision of K4 . It follows that, if P is a (u, v)-path in G2 , then G ∪ P
contains a subdivision of K4 . Hence so, too, does G, since G1 ∪P ⊂ G. Now suppose
that G is 3-connected. Since ∆ ≥ 3, G has a cycle C of length at least four. Let u
and v be nonconsecutive vertices on C. Since G − {u, v} is connected, there is a path
P in G − {u, v} connecting the two components of C − {u, v} we may assume that
the origin x and the terminus y are the only vertices of P on C. Similarly, there is
a path Q in G − {x, y}. If P and Q have no vertex in common, then C ∪ P ∪ Q is a
subdivision of K4 . Otherwise, let w be the first vertex of P on Q, and let P 0 denote
the (x, w)-section of P . Then C ∪ P 0 ∪ Q is a subdivision of K4 (figure 8.1). Hence,
in both cases, G contains a subdivision of K4

8.4 CHROMATIC POLYNOMIALS

In the study of colourings, some insight can be gained by considering not only the
existence of colourings but the number of such colourings; this approach was devel-
oped by Birkhoff 1912) as a possible means of attacking the four-colour conjecture.

Theorem 8.5 If G is simple, then πk (G) = πk (G − e) − πk (G · e) for any edge e of

G.
8.4. CHROMATIC POLYNOMIALS 45

Exercises:

• If a k-chromatic graph G has a colouring in which each colour is assigned to

at least two vertices, show that G has a k-colouring of this type.

• Use Brooks’ theorem to show that if G is loopless with ∆ = 3, then χ0 ≤ 4.

Notes:
46 CHAPTER 8. VERTEX COLORING
Chapter 9

PLANER GRAPHS

Unit - IX

9.1 PLANE AND PLANAR GRAPHS

A graph is said to be embeddable in the plane, or planar, if it can be drawn in

the plane so that its edges intersect only at their ends. Such a drawing of a planar
graph G is called a planar embedding of G. A planar embedding G of G can itself
be regarded as a graph isomorphic to G; the vertex set of G is the set of points
representing vertices of G, the edge set of G is the set of lines representing edges of
G, and a vertex of G is incident with all the edges of G that contain it. We therefore
sometimes refer to a planar embedding of a planar graph as a plane graph. Figure
2 (b) shows a planar embedding of the planar graph in figure 1 (a).
It is clear from the above definition that the study of planar graphs necessarily
involves the topology of the plane. However, we shall not a tempt here to be strictly
rigorous in topological matters, and will be content to adopt a naive point of view
toward them. This is done so as not to obscure the combinatorial aspect of the
theory, which is our main interest.
The results of topology that are especially relevant in the study of planar graphs
are those which deal with Jordan curves. (A Jordan curve is a continuous non-self-
intersecting curve whose origin and terminus coincide.) The union of the edges in a
cycle of a plane graph constitutes a Jordan curve; this is the reason why properties
of Jordan curves come into play in planar graph theory. We shall recall a well-known
theorem about Jordan curves and use it to demonstrate the nonplanarity of K5 .

47
48 CHAPTER 9. PLANER GRAPHS

Figure 9.1: A planar graph G

Figure 9.2: A planar embedding of G

Let J be a Jordan curve in the plane. Then the rest of the plane is partitioned
into two disjoint open sets called the interior and exterior of J. We shall denote the
interior and exterior of J, respectively, by int J and ext J, and their closures by Int J
and Ext J. Clearly Int J n Ext J = J. The Jordan curve theorem states that any line
joining a point in int J to a point in ext J must meet J in some point (see figure 3).
Although this theorem is intuitively obvious, a formal proof of it is quite difficult.

Theorem 9.1 Ks is nonplanar.

Proof 9.1 By contradiction. If possible let G be a plane graph corresponding to K5 .

Denote the vertices of G by v1 , v2 , v3 , v4 and v5 . Since G is complete, any two of its
vertices are joined by an edge. Now the cycle C = vl v2 v3 v1 is a Jordan curve in the
9.1. PLANE AND PLANAR GRAPHS 49

Figure 9.3:

Figure 9.4:

Figure 9.5: (a)An embedding of K5 on the torus; (b) An embedding of K3,3 on the
Möbius band
50 CHAPTER 9. PLANER GRAPHS

Figure 9.6: Stereographic Projection

plane, and the point v4 must lie either in int C or ext C. We shall suppose that v4 ∈
int C. (The case where v4 ∈ ext C can be dealt with in a similar manner.) Then the
edges v4 vl , v4 v2 and v4 v3 divide int C into the three regions int C1 , int C2 and int
C3 , where C1 = v1 v4 v2 v1 , C2 = v2 v4 v3 v2 and C3 = v3 v4 v1 v3 (see figure 4).
Now v5 must lie in one of the four regions ext C, int C1 , int C2 and int C3 . If v5 ∈
ext C then, since v4 ∈ int C, it follows from the Jordan curve theorem that’ the edge
v4 v5 must meet C in some point. But this contradicts the assumption that G is a
plane graph. The cases v5 ∈ int Ci , i = 1, 2, 3, can be disposed of in like manner.

A similar argument. can be used to establish that K3,3 , too, is nonplanar. We

shall see in section 9.5 that, on the other hand, every nonplanar graph contains a
subdivision of either K5 or K3,3 .
The notion of a planar embedding extends to other, surfaces. A graph G is said to
be embeddable on a surface S if it can be drawn in S so that its edges intersect only
at their ends; such a drawing (if one exists) is called an embedding of G on S. Figure
5 (a) shows an embedding of K5 on the torus, and figure 5 (b) an embedding of K3,3
on the Mobius band. The torus is represented as a rectangle in which opposite sides
are identified, and the Mobius band as a rectangle whose two ends are identified
after one half-twist.
We have seen that not all graphs can be embedded in the plane; this is also true
of other surfaces. It can be shown (see, for example, Freshet and Fan, 1967) that,
for every surface S, there exist graphs which are not embeddable on S. Every graph
can, however, be ’embedded’ in 3- dimensional space R3 (exercise 9.1.3).
9.2. DUAL GRAPHS 51

Planar graphs and graphs embeddable on the sphere are one and the same. To
show this we make use of a mapping known as stereographic projection. Consider
a sphere S resting on a plane P, and denote by z the point of S that is diagonally
opposite the point of contact of Sand P. The mapping π : S / {z} → P, defined by
π(S) = p if and only if the points z, s and p are collinear, is called stereographic
projection from z; it is illustrated in figure 9.5.

Theorem 9.2 A graph G is embeddable in the plane if and only if it is embeddable

on the sphere.

Proof 9.2 Suppose G has an embedding G̃ on the sphere. Choose a point z of the
sphere not in G̃. Then the image of G̃ under stereographic projection from z is an
embedding of G in the plane. The converse is proved similarly.

On many occasions it is advantageous to consider embeddings of planar graphs on

the sphere; one instance is provided by the proof of theorem 9.3 in the next section.
Exercises

1. Show that K3,3 is nonplanar.·

2. (a) Show that K5 - e is planar for any edge e of K5 .

(b) Show that K3,3 - e is planar for any edge e of K3,3 .

9.2 DUAL GRAPHS

A plane graph G partitions the rest of the plane into a number of connected regions;
the closures of these regions are called the faces of G. Figure 7 shows a plane graph
with six faces, f1 , f2 , f3 , f4 , f5 and f6 . The notion of a face applies also to embeddings
of graphs on other surfaces. We shall denote by F(G) and φ(G), respectively, the
set of faces and the number of faces of a plane graph G.
Each plane graph has exactly one unbounded face, called the exterior face; in the
plane graph of figure 7, f1 is the exterior face.

Theorem 9.3 Let v be a vertex of a planar graph G. Then G can be embedded in

the plane in such a way that v is on the exterior face of the embedding.
52 CHAPTER 9. PLANER GRAPHS

Figure 9.7: A plane graph with six faces

Proof 9.3 Consider an embedding G̃ of G on the sphere; such an embedding exists

by virtue of theorem 9.2. Let z b.e a point in the interior of some face containing v,
and let π(G̃) be the image of G̃ under stereographic projection from z. Clearly π(G̃)
is a planar embedding of G of the desired type.

We denote the boundary of a face f, of a plane graph G by b(f). If G is connected,

then b(f) can be regarded as a closed walk in which each cut edge of G in b(f) is
traversed twice; when b(f) contains no cut edges, it is a cycle of G. For example, in
the plane graph of figure 7,

b(f2 ) = vl e3 v2 e4 v3 e5 v4 e1 v1
and
b(f5 ) = v7 e10 v5 e11 v8 e12 v8 e11 v5 e8 v6 e9 v7

A face f is said to be incident with th,e vertices and edges in its boundary. If e is
a cut edge in a plane graph, just one face is incident with e; otherwise, there are
two faces incident with e. We say that an edge separates the faces incident with it.
The degree, dG (f), of a face f is the number of edges with which it is incident (that
is, the number of edges in b(f)), cut edges being counted twice. In figure 7, f1 is
incident with the vertices vt , v3 , v4 , v5 , v6 , v7 and the edges e1 , e2 , e5 , e6 , e7 , e9 , e10 ; e1
separates f1 from f2 and e11 separates f5 from f5 ; d(f2 )= 4 and d(f5 )=6.
Given a plane graph G, one can defile another graph G* as follows: corresponding
to each face f of G. there is a vertex f* of G *, and corresponding to each edge e of
G there is an edge e*of G*; two vertices f* and g* are joined by the edge e* in G* if
and only if their corresponding faces f and g. are separated by the edge e in G. The
graph G* is called the dual of G. A plane graph and its dual are shown in figures 8
(a) and 8 (b).
9.2. DUAL GRAPHS 53

Figure 9.8: A plane graph and its dual

54 CHAPTER 9. PLANER GRAPHS

Figure 9.9: Isomorphic plane graphs with nonisomorp

It is easy to see that the dual G* of a plane graph G is planar; in fact, there is a
natural way to embed G* in the plane. We place each vertex f* in the corresponding
face f of G,and then draw each edge e* in such a way that it crosses the correspond-
ing edge e of G exactly once (and crosses no other edge of G). This procedure is
illustrated in figure 9.7C, where it is indicated by heavy points and lines. It is intu-
itively clear that we call always draw the dual as a plane graph in this way, but we
shall float prove this fact. Note that if e is a loop of G, then e* is a cut edge of G*,
and vice versa.
Although defined abstractly, it is sometimes convenient to regard the dual G* of
a plane graph G as a plane graph (embedded as described above). One can then
consider the dual G** of G*, and it is not difficult to prove that, when G is connected,
G**∼= G; a glance at figure 8 (c) will indicate why this is so.
It should be noted that isomorphic plane graphs may have nonisomorphic duals.
For example, the plane graphs in figure 9 are isomorphic, but their duals are not-the
plane graph of figure 9 (a) has a face of degree five, whereas the plane graph of
figure 9 (b) has no such face. Thus the notion of a dual is meaningful only for plane
graphs, and cannot be extended to planar graphs in general.
The following relations are direct consequences of the definition of G*:

v(G∗ ) = φ(G)

ε(G∗) = ε(G)0 (9.1)

do.(f ∗ ) = do(f )

for all f ∈ F(G).

9.3. EULER’S FORMULA 55

Theorem 9.4 Theorem 9.4 If G is a plane graph, then

X
d(f ) = 2ε
f ∈F

Proof 9.4 Let G* be the dual of G. Then

X X
d(f ) = d(f ∗ )
f ∈F (G) f ∗ ∈V (G∗ )

= 2ε(G∗ )

= 2ε(G)

Exercises

1. (a) Show that a graph is planar if and only if each of its blocks is planar.
(b) Deduce that a minimal nonplanar graph is a simple block.

2. A plane triangulation is a plane graph in which each face has degree three.
Show that every simple plane graph is a spanning subgraph of some simple
plane triangulation (v ≥ 3).

3. Let G be a simple plane triangulation with v ≥ 4. Show that G* is a simple

2 - edge - connected 3 - regular planar graph.

4. Show that any plane triangulation Gcontains a bipartite subgraph with 2e

(G)/3 edges.
(F. Harary, D. Matula)

9.3 EULER’S FORMULA

There is a simple formula relating the numbers of vertices, edges and faces in a
connected plane graph. It is known as Euler’s formula because Euler established it
for those plane graphs defined by the vertices and edges of polyhedra.

Theorem 9.5 IfG is a connected plane graph, then v-B+cP=2

Proof 9.5 By induction on φ, the number of faces of G. If φ = 1, then each edge

of G is a cut edge and so G, being connected, is a tree. In this case ε = v -1, the
56 CHAPTER 9. PLANER GRAPHS

theorem clearly holds. Suppose that it is true for all connected plane graphs with
fewer than n faces, and let G be a connected plane graph with n ≥ 2 faces. Choose
an edge e of G that is not a cut edge. Then G - e is a connected plane graph and
has n - 1 faces, since the two faces of G separated by e combine to form one face of
G - e. By the induction hypothesis

v(G − e) − ε(G − e) + φ(G − e) = 2

and, using the relations

v(G − e) = v(G)

ε(G − e) = ε(G) − 1

φ(G − e) = φ(G) − 1

we obtain

v(G) − ε(G) + φ(G) = 2 (9.2)

The theorem follows by the principle of induction

Corollary 9.6 All planar embeddings of a given connected planar graph have the
same number of faces.

Proof 9.6 Let G and H be two planar embeddings of a given connected planar graph.
Since G ∼
= H, v(G) = v(H) and ε(G) = ε(H). Applying theorem9.5, we have

φ(G) = ε(G) − v(G) + 2 = ε(H) − v(H) + 2 = φ(H)

Corollary 9.7 If G is a simple planar graph with v ≥ 3, then ε ≤ 3v − 6.

Proof 9.7 It clearly suffices to prove this for connected graphs. Let G be a simple
connected graph with v ≥ 3. Then d(f )≥ 3 for all f ∈ F, and
X
d(f ) ≥ 3φ
f ∈F
9.3. EULER’S FORMULA 57

By theorem 9.4

2ε ≥ 3φ

Thus, from theorem 9.5

v − ε + 2ε/3 ≥ 2

or

ε ≤ 3v − 6

Corollary 9.8 If G is a simple planar graph, then δ ≤ 5.

Proof 9.8 This is trivial for v = 1, 2. If v ≥ 3, then,

X
δv ≤ d(v) = 2ε ≤ 6v − 12
v∈V

It follows that 8≤5.

We have already seen that K5 and K3,3 are nooplanar. Here, we shall derive these
two results as corollaries of theorem9.5.

Corollary 9.9 K5 is nonplanar.

Proof 9.9

If K5 were planar then” by corollary 9.7, we would have

10 = ε(K5 ) < 3v(K5 ) − 6 = 9

Thus Ks must be nonplanar

Corollary 9.10 K3,3 is nonplanar.

Proof 9.10
58 CHAPTER 9. PLANER GRAPHS

Suppose that K3,3 is planar and let G be a planar embedding of K3,3 . Since K3,3 has
no cycles of length less than four, every face of G must have degree at least four.
Therefore, by theorem 9.4, we have
X
4φ ≤ d(f ) = 2ε = 18
f ∈F

That is

φ≤4

Theorem 9.5 now implies that

2=v−ε+φ≤6−9+4=1

which is absurd.
Exercises

1. Show that every planar graph is 6 - vertex - colourable.

Notes:
Chapter 10

BRIDGES

Unit - X
10.1 Bridges

In the study of planar graphs, certain subgraphs, called bridges, play an important
role. We shall discuss properties of these subgraphs in this section.
Let H be a given subgraph of a graph G. We define a relation ∼ on E(G)\E(H) by
the condition that e1 ∼ e2 if there exists a walk W such that

1. the first and last edges of W are e1 and e2 , respectively, and

2. W is internally-disjoint from H (that is, no internal vertex of W is a vertex of

H).

It is easy to verify that ∼ is an equivalence relation on E(G)\E(H). A subgraph of

G - E (H) induced by an equivalence class under the relation ∼ is called a bridge of
H in G. It follows immediately from the definition that if B is a bridge of H, then
B is a connected graph and, moreover, that any two vertices of B are connected by
a path that is internally-disjoint from H. It is also easy to see that two bridges of
H have no vertices in common except, possibly, for vertices of H. For a bridge B of
H, we write V(B) ∩ V(H) = V(B, H), and call the vertices in this set the vertices of
attachment of B to H. Figure 10.1 shows a variety of bridges of a cycle in a graph;
edges of different bridges are represented by different kinds of lines.
In this section we are concerned with the study of bridges of a cycle C. Thus, to avoid

59
60 CHAPTER 10. BRIDGES

Figure 10.1: Bridges in a grap

repetition, we shall abbreviate ’bridge of C’ to ’bridge’ in the coming discussion; all

bridges will be understood to be bridges of a given cycle C.
In a connected graph every bridge has at least one vertex of attachment, and in
a block every bridge has at least two vertices of attachment. A bridge with k
vertices of attachment is called a k-bridge. Two k-bridges with the same vertices
of attachment are equivalent k-bridges; for example, in figure 10.1 B1 and B2 are
equivalent 3-bridges.
The vertices of attachment of a k-bridge B with k ≥ 2 effect a partition of C into
edge-disjoint paths, called the segments of B. Two bridge avoid one another if all
the vertices of attachment of one bridge lie in a single segment of the other bridge;
otherwise they overlap. In figure 10.1, B2 and B3 avoid one another, whereas B1
and B2 overlap. Two bridges B and B’ are skew if there are four distinct vertices u,
v, u’ and v’ of C such that u and v are vertices of attachment of B, u’ and v’ are
vertices of attachment of B’, and the four vertices appear in the cyclic order u, u’,
v, v’ on C. In figure 10.1, B3 and B4 are skew, but B1 and B2 are not.

Theorem 10.1 If two bridges overlap, then either they are skew or else they are
equivalent 3 - bridges.

Proof 10.1 Suppose that the bridges B and B 0 overlap. Clearly, each must have at
least two vertices of attachment. Now if either B or B 0 is a 2-bridge, it is easily
10.2. KURATOWSKI’S THEOREM 61

verified that they must be skew. We may therefore assume that both B and B 0 have
at least three vertices of attachment. There are two cases.

Case 1 B and B 0 are not equivalent bridges. Then B 0 has a vertex of attachment u0
between two consecutive vertices of attachment u and v of B. Since B and B 0 over-
lap, some vertex of attachment v 0 of B 0 does not lie in the segment of B connecting
u and v. It now follows that B and B 0 are skew.

Case 2 B and B 0 are equivalent k-bridges, k ≥ 3. If k ≥ 4, then B and B 0 are

clearly skew; if k = 3, they are equivalent 3 - bridges

Theorem 10.2 If a bridge B has three vertices of attachment v1 , v2 and v3 , then

there exists a vertex v0 in V(B)\V(C) and three paths P1 , P2 and P3 in B joining
v0 to v1 , v2 and v3 , respectively, such that, for i 6= j, Pi and Pj have only the vertex
v0 in common (see figure 10.2).

Proof 10.2 Let P be a (v1 , v2 )-path in B, internally-disjoint from C. P must have

an internal vertex v, since otherwise the bridge B would be just P, and would not
contain a third vertex v3 . Let Q be a (v3 , v) - path in B; internally disjoint from C,
and let v0 be the first vertex of Q on P. Denote by P1 the (v0 , v1 ) - section of p−1 ,
by P2 the (v0 , v2 ) - section of P, and by P3 the (v0 , v3 )-section of Q−1 . Clearly P1 , p2
and P3 satisfy the required conditions

We shall now consider bridges in plane graphs. Suppose that G is a plane graph
and that C is a cycle in G. Then C is a Jordan curve in the plane, and each edge of
E(G)\E(C) is contained in one of the two regions Int C and Ext C. It follows that
a bridge of C is contained entirely in Int C or Ext C. A bridge contained in Int C is
called an inner bridge, and a bridge contained in Ext C, an outer bridge. In figure
9.11 B1 and B2 are inner bridges, and B3 and B4 are outer bridges.

10.2 KURATOWSKI’S THEOREM

Since planarity is such a fundamental property, it is clearly of importance to know

which graphs are planar and which are not. We have already noted that, in partic-
ular, K5 and K3,3 are non-planar and that any proper subgraph of either of these
graphs is planar. A remarkably simple characterization of planar graphs was given
62 CHAPTER 10. BRIDGES

Figure 10.2:

by Kuratowski (1930). This section is devoted to a proof of Kuratowski’s theorem.

The following lemmas are simple observations, and we leave their proofs as an ex-
ercise (9.5.1).

Lemma 10.3 If G is non-planar, then every subdivision of G is non-planar.

Lemma 10.4 If G is planar, then every subgraph of G is planar.

Since K5 and K3,3 are non planar, we see from these two lemmas that if G is planar,
then G cannot contain a subdivision of K5 or of K3,3 . Kuratowski showed that this
necessary condition is also sufficient.
Before proving Kuratowski’s theorem, we need to establish two more simple lemmas.
Let G be a graph with a 2-vertex cut {u, v}. Then there exist edge-disjoint subgraphs
G1 and G2 such that V (G1 ) ∩ V (G2 ) = u, v and G1 ∪ G2 = G. Consider such a
separation of G into subgraphs. In both G1 and G2 join u and v by a new edge e
to obtain graphs H1 and H2 . Clearly G = (H1 U H2 ) − e. It is also easily seen that
ε(Hi ) < ε(G) for i = 1, 2.

Lemma 10.5 If G is non planar, then at least one of H1 and H2 is also non planar.

Proof 10.3 By contradiction. Suppose that both H1 and H2 are planar. Let H̃1 be
a planar embedding of H1 , and let f be a face of H̃1 incident with e. If H̃2 is an
10.2. KURATOWSKI’S THEOREM 63

embedding of H2 in f such that H̃1 and H̃2 have only the vertices u and v and the
edge e in common, then (H̃1 ∪ H̃2 ) - e is a planar embedding of G. This contradicts
the hypothesis that G is non planar.

Lemma 10.6 Let G be a non planar connected graph that contains no subdivision
of K5 or K3,3 and has as few edges as possible. Then G is simple and 3 - connected.

Proof 10.4 By contradiction. Let G satisfy the hypotheses of the lemma. Then G
is clearly a minimal non planar graph, and therefore must be a simple block. If G
is not 3-connected, let {u, v} be a 2 - vertex cut of G and let H1 and H2 be the
graphs obtained from this ut as described above. By lemma 10.5, at least one of H1
and H2 ,say H1 , is non planar. Since ε(H1 ) < ε(G), H1 must contain a subgraph K
which is a subdivision of K5 or K3,3 ; moreover K * G, and so the edge e is in K.
Let P be a (u, v)-path in H2 - e. Then G contains the subgraph (K ∪ P ) − e, which
is a subdivision of K and hence a subdivision of K5 or K3,3 . This contradiction
establishes the lemma

We shall find it convenient to adopt the following notation in the proof of Kura-
towski’s theorem. Suppose that C isa cycle in a plane graph. Then we can regard
the two possible orientations of C as ’clockwise’ and ’anticlockwise’. For any two
vertices, u and v of C, we shall denote by C[u, v] the (u, v)-path which follows
the clockwise orientation of C; similarly we shall use the symbols C(u, v], C[u, v)
and C(u, v )to denote the paths C[u, v] - u, C[u, v]-v and C[u, v]-{u, v}. We are
now ready to prove Kuratowski’s theorem. Our proof is based on that of Dirac and
Schuster (1954).

Theorem 10.7 A graph is planar if and only if it contains no subdivision of K5 or

K3,3 .

Proof 10.5 We have already noted that the necessity follows from lemmas 9.10.1
and 9.10.2. We shall prove the sufficiency by contradiction.
If possible, choose a nonplanar graph G that contains no subdivision of K5 or K3,3
and has as few edges as possible. From lemma 10.6 it follows that G is simple and
3-connected. Clearly G must also be a minimal nooplanar graph.
Let uv be an edge of G, and let H be a planar embedding of the planar graph G -
uv. Since G is 3-connected, H is 2-connected and, by corollary 3.2.1, u and v are
64 CHAPTER 10. BRIDGES

contained together in a cycle of H. Choose a cycle C of H that contains u and v and

i,s such that the number of edges in Int C is as large as possible.
Since H is simple and 2-connected, each bridge of C in H must have at least two
vertices of attachment. Now all outer bridges of C must be 2-bridges that overlap uv
because, if some outer bridge were a k - bridge for k ≥ 3 or a 2 - bridge that avoided
uv, then there would be a cycle C’ containing u and v with more edges in its interior
than C, contradicting the choice of C.
In fact, all outer bridges of C in H must be single edges. For if a 2 - bridge with
vertices of attachment x and y had a third vertex, the set x, y would be a 2 - vertex
cut of G, contradicting the fact that G is 3 - connected.
No two inner bridges overlap. Therefore some inner bridge skew to uv must overlap
some outer bridge. For otherwise, all such bridges could be transferred (one by one),
and then the edge uv could be drawn in Int C to obtain a .planar embedding of G;
since G is non planar, this is not possible. Therefore, there is an inner bridge B
that is both skew to uv and skew to some outer bridge xy.

Two cases now arise, depending on whether B has a vertex of attachment different
from u, v, x and y or not.

Case 1 B has a vertex of attachment different from u, v, x and y. We can choose

the notation so that B has a vertex of attachment v1 in C(x, u). We consider two
sub-cases, depending on whether B has a vertex of attachment in C(y, v) or not.
Case 1a B has a vertex of attachment v2 in C(y, v). In this case there is a (v1 , v2 )
- path P in B that is internally-disjoint from C. But then (C ∪ P) + {uv, xy} is a
subdivision of K3,3 in G, a contradiction.
Case 1b B has no vertex of attachment in C(y, v). Since B is skew to uv and to xy,
B must have vertices of attachment v2 in C(u, y] and v3 in C[v, x). Thus B has three
vertices of attachment v1 , v2 and v3 . Then, there exists a vertex v0 in V(B)\V(C)
and three paths P1 , P2 and P3 in B joining v0 to v1 , v2 and v3 , respectively, such that,
for i 6= j, Pi and Pj have only the vertex vo in common. But now (C ∪ P1 ∪ P2 ∪
P3 ) + {uv, xy} contains a subdivision of K3,3 , a contradiction. The subdivision of
K3,3 is indicated by, heavy lines.
Case 2 B has no vertex of attachment other than u, v, x and y. Since B is skew
10.3. THE TIMETABLE PROBLEM 65

to both uv and xy, it follows that u, v, x and y must all be vertices of attachment of
B. Therefore there exists a (u, v)-path P and an (x, y)-path Q in B such that (i) P
and Q are internally-disjoint from C, and (ii) |V (P ) ∩ V (Q)| ≥ 1. We consider two
sub-cases, depending on whether P and Q have one or more vertices in common.
Case 2a |V (P ) ∩ V (Q)| ≥ 1. In this case (C ∪ P ∪ Q) + {uv, xy} is a sub, division
of K5 in G, again a contradiction.
Case 2b |V (P ) ∩ V (Q)| ≥ 2. Let u0 and v 0 be the first and last vertices of P
on Q, and let P1 and P2 . denote the (u, u0 )− and (v 0 , v)− sections of P. Then
(C ∪ P1 ∪ P2 ∪ Q) + {uv, xy} contains a subdivision of K3,3 in G, once more a
contradiction.
Thus all the possible cases lead to contradictions, and the proof is complete.

10.3 The Timetable Problem

Suppose in a school there are r teachers, T1 , T2 ,..., Tr , and s classes, C1 , C2 , .., Cs .

Each teacher Ti is expected to teach the class Cj for pij periods. It is clear that
during any particular period, no more than one teacher can handle a particular class
and no more than one class can be engaged by any teacher. Our aim is to draw up
a timetable for the day that requires only the minimum number of periods. This
problem is known as the ”timetable problem”.
To convert this problem into a graph - theoretic one, we form the bipartite graph
G = g(T, C) with bipartition (T, C), where T represents the set of teachers Ti
and C represents the set of classes Cj . Further, Ti is made adjacent to Cj in G
with pij edges iff teachers Ti is to handle class Cj for pij periods. Now, color the
edges of G so that no two adjacent edges receive the same color. Then the edges
in a particular color class, that is, the edges in that color form a matching in G
and correspond to a schedule of work for a particular period. Hence, the minimum
number of periods required is the minimum number of colors in an edge - coloring
of G in which adjacent edges receive distinct colors; in other words, it is the edge -
chromatic number of G. We now present these notions as formal definition.
Definition
An edge - coloring of a loopless graph G is a function π : E(G) → S, where S is a set
of distinct colors; it is proper if no two adjacent edges receive the same color. Thus
a proper edge - coloring π of G is a function π : E(G) → S such that π(e) 6= π(e0 )
66 CHAPTER 10. BRIDGES

whenever edges e and e0 are adjacent in G.

Definition
The minimum k for which a loopless graph G has a proper k - edge - coloring is
called the edge chromatic number or chromatic index of G. It is denoted by χ0 (G).
G is k - edge - chromatic if χ0 (G) = k.

Theorem 10.8 If G is a loopless bipartite graph, χ0 (G) = ∆(G).

Proof 10.6 The proof is by induction on the size (i.e., number of edges) m of G.
The result is true for m = 1. Assume the result for bipartite graphs of size at most
m - 1. Let G have m edges. Let e = uv ∈ E(G). Then G-e has a proper ∆-
edge-coloring, say c. Out of these ∆ colors, suppose that, one particular color is
not represented at both u and v. Then edge uv can be colored with this color and a
proper ∆-edge-coloring of G is obtained.
In the other case (i.e., in the case for which each of the ∆ colors is represented
either at u or at v) , since the degrees of u and v in G-e are at most ∆ - 1, there
exists a color out of the ∆ colors that is not represented at u, and similarly there
exists a color not represented at v. Thus, if color j is not represented at u in c,
then j is represented at v in c, and if color i is not represented at v in c, then i is
represented at u in c. Since G is bipartite and u and v are not in the same parts of
the bipartition, there can exist no u-v path in G in which the colors alternate between
i and j.

Notes:
Chapter 11

FIVE COLOR PROBLEM

Unit - XI
11.1 The Five Colour Theorem

Definition 11.1 An assignment of colours to the vertices of a graph, there is no

two adjacent vertices get the same colour is called colouring of a graph.

Definition 11.2 The vertices of a planar graph with atmost five colours is known
as five colour theorem.

Theorem 11.1 Every planar graph is 5-colourable

Proof 11.1 We will prove the theorem by induction on the number of p points. For
any planar graph having p ≤ 5 points, the result is obvious since the graph is p-
colourable.
Now, let us 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 a G − v where ci , 1 ≤ i ≤ 5, are the colours are 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 a 5-colouring of G.
Hence, we have to consider only the case in which deg v = 5 and all the five
colours are used for colouring the vertices of G adjacent to v.
Let v1 , v2 , v3 , v4 , v5 be the vertices adjacent to v coloured c1 , c2 , c3 , c4 and c5 re-
spectively and G13 denote the subgraph of G − v induced by those vertices coloured c1

67
68 CHAPTER 11. FIVE COLOR PROBLEM

or c3 . If v1 and v3 belong to different components of G13 , then 5-colouring of G − r

can be obtained by interchanging the colours 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 the component. this interchange of colours results in a coiluring
of G − v). In this 5-colouring no vertex adjacent to v is coloured c1 , and hence by
colouring v with c1 , a colouring of G is obtained.
If v1 and v3 are the same component of G13 , then in G there exists a v1 − v3
path all of whose points are colored c1 or c3 . Hence there is no v2 − v4 path all whose
points are colored c2 , c4 .
Hence, if G24 denotes the subgraph of G − v induced by the points colored c2 or
c4 , then v2 and v4 belong to different components of G24 . Hence if we interchange the
colors of the points in the component of G24 containing v2 , a new 5-coloring G − v
results and this, no point adjacent to v is colored c2 . Hence, by assigning colour c2
to v, we can get a 5-coloring of G. This completes the induction and the proof.

11.2 Non-Hamiltonian Graph

Definition 11.3 A spanning cycle in a graph is called a Hamiltonian cycle. A

graph having a Hamiltonian cycle is called a Hamiltonian graph

Definition 11.4 The closure of a graph G with p points is the graph obtained from
G by repeatedly joining pairs of non adjacent vertices whose degree sum is at least p
until no such pair remains. The closure of G is denoted by c(G)

Theorem 11.2 A graph is Hamiltonian iff its closure is hamiltonian

Proof 11.2 Let x1 , x2 , · · · , xn be the sequence of edges added to G in obtaining c(G).

Let G1 , G2 , · · · , Gn = c(G) be the successive graphs obtained.

G is Hamiltonian ⇔ G1 is Hamiltonian
⇔ G2 is Hamiltonian
..
.
⇔ Gn = c(G) is Hamiltonian.

Problem 11.2.1 Show that the Petersen graph is Hamiltonian.

11.2. NON-HAMILTONIAN GRAPH 69

Solution 11.2.1 If the Petersen graph G has a Hamiltonian cycle C, then G-E(C)
must be regular spanning subgraph of degree 1.
Let us search for all 1-factors in G and show that none of them arise out a Hamil-
tonian cycle of G.
Case 1. Consider the subset A = {1a, 2b, 3c, 4d, 5e} of the edge set of G.
Clearly A is a 1-factor of G, but G-A is the union of two disjoint cycles and hence
is not a Hamiltonian cycle of G.
Case 2. If the 1-factor contains 4 edges from A, then the only line passing through
the remaining two points must also be included in the 1-factor, so that we again get
A.
Case 3. If a 1-factor contains just 3 edges from A, then two such choices can be
made.
Sub-case 3A. Let the one 1-factor contain 1a, 2b, and 3c. Now the subgraph in-
duced by the remaining four points is a P4 whose unique 1-factor is {4d, 5e}. Thus
the 1-factor of G considered becomes A.
Sub-case 3B. Let the 1-factor contain 1a, 2b and 4d. Here again the remaining
four points induce P4 , whose unique 1-factor is {3c, 5e }. Thus the 1-factor of G
considered becomes A.
Case 4. If a 1-factor contains just 2 edges from A, then again two such choices
are possible.
Sub-case 4A. Let the 1-factor contain 1a and 2b. In the subgraph induced by the
remaining 6 points, point d has degree one and hence any 1-factor of that subgraph
must contain edge 4d. Thus case 3 is repeated.
Sub-case 4B. Let the 1-factor contain 1a and 3b. In the subgraph induced by the
remaining 6 points, point 2 has degree one and hence any 1-factor of that subgraph
must contain edge 2b. Thus case 3 is repeated.
Case 5. Let a one factor contain just one edge of A, say 1a. If it a contains one
more edge from A, then one of the earlier cases will be repeated. Hence we have
choose the other four edges of this 1-factor from two paths, each of length 3. Hence
the 1-factor is B = {1a, ce, bd, 23, 45}. Now G-B is again union of two disjoint
cycles, and not a Hamiltonian cycle.
Case 6. Suppose there exists a 1-factor that does not contain any edge from A. It
can contain at most two edges from the cycle 123451 and at most two edges from
the cycle acebda. Hence it can contain at most four edges.
Hence there does not exist such a 1-factor.
Since the above 6 cases cover all possible types of 1-factors, we see that G has no
70 CHAPTER 11. FIVE COLOR PROBLEM

1-factor arising out of a Hamiltonian cycle.

Hence, G has no hamiltonian cycle.
Thus, G is non-Hamiltonian.

Exercise 11.2.1 Given an example of graph G such that c(G) is not complete.

Exercise 11.2.2 Show that if G is a bipartite graph with an odd number of points,
then G is non-Hamiltonian.

Notes:
Chapter 12

DIRECTED GRAPHS &

DIRECTED PATH

Unit - XII
12.1 Directed Graphs

Definition 12.1 A directed graph D is a pair (V, A) where V is a finite nonempty

set and A is a subset of V × V − {(x, x)/x ∈ V }. The elements of V and A are
respectively are called vertices and arcs. If (u, v) ∈ A then the arc (u, v) is said to
have u as its initial vertex and v as its terminal vertex. Also the arc (u,v) is said to
join u to v.

Theorem 12.1 In a graph D, sum of the in-degrees 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 12.1 Let q denote the number of arcs in D = (V, A).

+ −
P P
Let B = v∈V d (v) 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 12.2 A 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 the origin and
terminus of W respectively and v1 , v2 , · · · , vn−1 are called its internal vertices.

71
72 CHAPTER 12. DIRECTED GRAPHS & DIRECTED PATH

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.

12.2 Directed Paths and Cycles

Definition 12.3 A path is a walk in which all the vertices are distinct. A cycle
is a nontrivial closed walk whose origin and internal vertices are distinct.
If there is a path from u to v is said to be reachable from u. A digraph is called
strongly connected or connected or strong if every pair of points are mutu-
ally 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.

Theorem 12.2 The edges of a connected graph G = (V, E) can be oriented so that
the resulting digraph is strongly connected iff every edge of G is contained in at least
one cycle.

Proof 12.2 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.
Let S = v1 , v2 , · · · , vn , v1 be a cycle in G. Orient the edges of S so that S becomes a di-
rected cycle and hence becomes a strongly connected sub-digraph. If V = {v1 , · · · , vn }
then we are through. Otherwise, let w be a vertex of G not in S such that w is ad-
jacent 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 a strongly connected as it can be got from S by a sequence of additions 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 in the enlarged oriented graph the u − vj sub-path of
P followed by the vj − v sub-path of S give a directed u − v path. Also the v − vj
sub-path of S followed by the vi − u sub-path of P give a directed v − u path. This
12.2. DIRECTED PATHS AND CYCLES 73

type of argument can be repeated for each addition og simple, directed paths.)
This process can be repeated till we get a strongly connected oriented spanning sub-
graph of G. The remaining edges can now be oriented in any way. The resulting
oriented graph is strongly connected. This completes the proof.

Exercise 12.2.1 Show that every Eulerian graph is strongly connected and prove
its converse not true.

Notes:
74 CHAPTER 12. DIRECTED GRAPHS & DIRECTED PATH
Chapter 13

NETWORKS

Unit - XIII
13.1 Flows

A network N is a digraph D (the underlying digraph of N ) with two distinguished

subsets of vertices, X and Y and a non-negative integer valued function c defined
on its arc set A; the sets X and Y are assumed to be disjoint and nonempty.
We represent a network by drawing its underlying digraph and labeling each arc
with its capacity. Then the below digram (13.1) shows that the network with two
sources x1 and x2 , three sinks y1 , y2 andy3 and four intermediate vertices’s v1 , v2 , v3
and v4 .
If S ⊆ V , we denote V \S by S. If f is a real-valued function defined on the arc set
P
of A of N , and if K ⊆ A, we denote f (a) by f (K) . Furthermore, if K is a set
a∈K
of arcs of the form (S, S), we shall write f + (S) for f (S, S) and f − (S) for f (S, S).
A flow in a network N is an integer-valued f defined on A such that

0 ≤ f (a) ≤ c(a) for all a ∈ A (13.1)

and
f − (v) = f + (v) for all v ∈ I (13.2)
The value of f (a) of f on an arc a can be likened to the rate at which material is
transported along a under the flow f . The upper bound in condition (13.1) is called
the capacity constraint; it imposes the natural restriction that the rate of flow along
an arc cannot exceed the capacity of the arc. Condition (13.2) , is called conservation
condition, requires that, for any intermediate vertex v, the rate at which material

75
76 CHAPTER 13. NETWORKS

Figure 13.1: A Network

is transported into v is equal to the rate at which it is transported out of v. Note

that every network has at least one flow, since the function f defined by f (a) = 0,
for all a ∈ A, clearly satisfies both (13.1) and (13.2) and; it is called zero flow. A
less trivial example of a flow is given in figure (13.2). The flow along each arc is
indicated in bold type.
If S is a subset of vertices’s in a network N and f is a flow in N , then f + (S) − f ( S)
is called the resultant flow of out of S, and f − (S) − f + (S) the resultant flow into
S, relative to f . Since the conservation condition requires that the resultant flow
out of X is equal to the resultant flow into Y . This common quantity is called the
value of f , and is denoted by valf ; thus

valf = f + (X) − f − (X)

The value of the flow indicated in figure(13.2) is 6. A flow f in N is a maximum

0 0
flow if there is no flow f in N such that val f > valf . Such flows are of obvious
importance in the context of transportation networks. The problem of determining
a maximum flow in an arbitrary network can be reduced to the case of networks that
have just one source and one sink by means of a simple device. Given a network N ,
0
construct a new network N as follows:
13.1. FLOWS 77

Figure 13.2: A flow in network

(i) adjoin two new vertices’s x and y to N ;

(ii) join x to each vertex in X by an arc of capacity ∞;

(iii) join each vertex in Y to y by an arc of capacity ∞;

0
(iv) designate x as the source and y as the sink of N

Figure (13.3) illustrates this procedure as applied to the network N of figure (13.1).
0
Flows in N and N correspond to one another in a simple way. If f is a flow in N
such that the resultant flow out of each source and into each sink is non-negative (it
0
suffices to restrict our attention to such flows) then the function f defined by


 f (a), if a is an arc of N
0
+ −
f (a) = f (v) − f (v) if a = (x, v) (13.3)
 −
f (v) − f + (v) if a = (v, y)

0 0
is a flow N such that valf = valf . Conversely, the restriction to the arc set of N
0
of a flow in N is a flow in N having the same value. Therefore, throughout the next
three sections, we shall confirm our attention to networks that have a single source
x and a single sink y.
78 CHAPTER 13. NETWORKS

Figure 13.3:

Figure 13.4: Exercise: 1

Exercise

(1) For each of the following networks , determine all possible flows and the value
of a maximum flow.

(2) Show that, for any flow f in N and any S ⊆ V ,

X
(f + (v) − f − (v)) = f + (S) − f − (S)
X X
(Note that, in general, f + (v) 6= f + (S)and f − (v) 6= f − (S))

(3) Show that, relative to any flow f in N , the resultant flow out of X is equal to
the resultant flow into Y .

(4) Show that

13.2. CUTS 79

Figure 13.5: A cut in a network

0 0 0
(a) the function f given by Equ.(13.3)is a flow in N and that valf = valf ;
(b) the restriction to the arc set of N of a flow in N having the same value.

13.2 Cuts

Let N be a network with a single source x and a single sink y. A cut in N is a set
of arcs of the form (S, S), where x ∈ S and y ∈ S. In the network of figure(13.5), a
cut is indicated by heavy lines.
The capacity of a cut K is the sum of the capacities of its arcs. We denote the
P
capacity of K by cap K; thus capK = c(a) The cut indicated in figure(13.5) has
a∈K
capacity 16.

Lemma: For any flow f and any cut (S, s) in N

valf = f + (S) − f − (S)

Proof Let f be a flow and (S, S) a cut in N . From the definitions of flow and
value of a flow, we have

+ − valf if v=x
f (v) − f (v) =
0 if v ∈ S\{x}

Summing these equations over S and simplifying, we obtain

X
valf = (f + (v) − f − (v)) = f + (S) − f − (S)

It is convenient to call an arc a f-zero if f (a) = 0, f-positive if f (a) > 0, f-unsaturated

if f (a) < c(a) and f-saturated if f (a) = c(a).
80 CHAPTER 13. NETWORKS

Figure 13.6: Exercise: 1

Theorem: For any flow f and any cut K = (S, S) inN

valf ≤ capK

Furthermore, the above equality holds if each arc in (S, S) is f-saturated and each
arc in (S, S) is f-zero.

0
Note: A cut K in N is a maximum cut if there is no cut K in N such that
0
capK < capK.. If f ∗ is a maximum flow and K̃ is a minimum cut, we have, as a
special case of theorem, that
valf ∗ ≤ capK̃

Corollary: Let f be a flow and K be a cut such that valf = capK. Then f is a
maximum flow and K is a minimum cut.

Exercise

(1) In the above network,

(a) determine all cuts;

(b) find the capacity of a minimum cut;

(c) show that the flow indicated is a maximum flow.

13.2. CUTS 81

(2) Show that, if there exists no directed (x, y)− path in N , then the value of a
maximum flow and the capacity of a minimum cut are both zero.

Notes:
82 CHAPTER 13. NETWORKS
Chapter 14

MAX-FLOW MIN-CUT
THEOREM

Unit - XIV
14.1 The Max-Flow Min-Cut theorem

Let f be a flow in a network N . With each path P in N we associate a non-negative

integer υ(p) = minυ(a)
where

c(a) − f (a) if a is a forward arc of P
υ(a) =
f (a) if a is a reverse arc of P
As may be easily be seen, υ(P ) is the largest amount by which the flow along P
can be increased (relative to f ) without violating condition Equ(13.1). The path
P is said to be f-saturated if υ(P ) > 0(or, equivalently, if each forward arc of P is
f-unsaturated and each reverse arc of P is f-positive). Put simply, an f-unsaturated
path is one that is not being used to its full capacity. An f-incrementing path is an
f-unsaturated path from the source x to the sink y.
The existence of an f-incrementing path P in a network is significant since it implies
that f is not a maximum flow; in fact; by sending an additional flow of υ(P ) along
P , one obtains a new flow fˆ is defined by

 f (a) + υ(P ) if a is a forward arc of P
fˆ(a) = f (a) − υ(P ) if a is a reverse arc of P
f (a) otherwise.


for which valfˆ = valf + υ(P ). We shall refer to fˆ as the revised flow based on P .

83
84 CHAPTER 14. MAX-FLOW MIN-CUT THEOREM

Figure 14.1: An f -unsaturated tree

Theorem: A flow f in N is a maximum flow if and only if N contains no f-

incrementing path.

Theorem: (Max-flow min-cut theorem) In any network, the value of a maxi-

mum flow is equal to the capacity of minimum cut.

Proof: It is of central importance in graph theory. Many results on graphs turn

out to be easy consequences of this theorem as applied to suitably chosen networks.
We prove this theorem by finding algorithm for a maximum flow in a network. It
is also known as labeling method. Starting with a known flow, for instance the zero
flow, it recursively constructs a sequence of flows of increasing value, and terminates
with a maximum flow. After the construction of each new flow f , a subroutine called
the labeling procedure is used to find an f -incrementing path, if one exists. If such
a path P is found, then , the revised flow based on P , is constructed and taken as
the next flow in the sequence. If there is no such path, the algorithm terminates;
then by the above theorem f is a maximum flow.

To describe the labeling procedure we need the following definition. A tree T in N

is an f -unsaturated tree if (i) x ∈ V (T ), and (ii) for every vertex υ of T , the unique
(x, v) path in T is an f -unsaturated path. Such a tree is shown in figure.
The search of an f-incrementing path involves growing an f -unsaturated tree T in
N . Initially, T consists of just the source x. At any stage, there are two ways in
which the tree may grow:

• If there exists an f -unsaturated arc a in (S, S), where S = V (T ), then both a

and its head are adjoined to T .
14.1. THE MAX-FLOW MIN-CUT THEOREM 85

• If there exists an f -positive arc a in (S, S), then both a and its tail are adjoined
to T .

Clearly, each of the above procedures results in an enlarged f -unsaturated tree.

Now either T eventually reaches the sink y or it stops growing before reaching y.
The former case is referred to as breakthrough; in the event of breakthrough, the
(x, y)- path in T is our desired f -incrementing path. If, however, T stops growing
before reaching y, we deduce from theorem and corollary that f is a maximum flow.
In figure two iterations of this tree-growing procedure are illustrated. The first leads
to breakthrough;the second shows that the resulting revised flow is a maximum flow.

The labeling procedure is a systematic way of growing an f -unsaturated tree T . In

the process of growing T , it assigns to each vertex v of T the label l(v) = υ(Pv ),
where Pv is unique (x, v)-path in T . The advantage of this labeling is that, in the
event of breakthrough, we not only have the f -incrementing path Py , but also the
86 CHAPTER 14. MAX-FLOW MIN-CUT THEOREM
14.1. THE MAX-FLOW MIN-CUT THEOREM 87

quantity υ(Py ) with which to calculate the revised flow based on Py .The labeling
procedure begins by assigning to the source x the label l(x) = ∞. It continues
according to the following rules:

• If a ia an f -unsaturated arc whose tail u is already labeled but whose head v

is not, then v is labeled l(v) = min{l(u), c(a) − f (a)}

• If a is an ff -positive arc whose head u is already labeled but whose tail v is

not, then v is labeled l(v) = min{l(u), f (a)} .

In each of the above case, v is said to be labeled based on u. To scan a labeled

vertex u is to label all unlabeled vertices that can be labeled based on u. The
labeling procedure is continued until either the sink y is labeled(breakthrough) or
all labeled vertices have been scanned and no mire vertices can be labeled (implying
that f is a maximum flow).

Consider, for example, the network N in figure(14.1), Clearly, the value of a maxi-
mum flow is N is 2m. The labeling method will use the labeling procedure 2m + 1
times if it starts with the zero flow and alternate between selecting xpuvsy and
xrvuqy as an incrementing path; for, in each case, the flow value increases by ex-
actly one. Since m is arbitrary, the number of computational steps required to
implement the labeling method in this instance can be bounded by no function of
v ans ε. In other words, it is not good algorithm. The refinement suggested as
88 CHAPTER 14. MAX-FLOW MIN-CUT THEOREM

follows: in the labeling procedure, scan on a ’first-labeled first-scanned’ basis; that

is, before scanning a labeled vertex u, scan the vertices that were labeled before u.
It can be seen that this amounts to selecting a shortest incrementing path. With
this refinement clearly, the maximum flow in the network of figure(14.1) would be
found in just two iterations of the labeling procedure.
Exercise

1. Show that, in any network N (with integer capacities), there is a maximum

flow f such that f (a) is an integer for all a ∈ A.

2. Consider a network N such that with each arc a is associated an integer

b(a) ≤ c(a). Modify the labeling method to find a maximum flow f in N
subject to the constraint f (a) ≥ b(a) for all a ∈ A (assuming that there is an
initial flow satisfying this condition).

14.2 Applications

14.2.1 MENGER’S THEOREMS

Lemma: Let N be a network with the source x and sink y in which each arc has
unit capacity. Then

(a) the value of a maximum flow in N is equal to the maximum number m of

arc-disjoint directed (x, y)-paths in N ; and

(b) the capacity of a minimum cut in N is equal to the minimum number n of

arcs whose deletion destroys all directed (x, y)-paths in N.

Theorem: Let xand y be two vertices of a digraph D. Then the maximum number
of arc-disjoint directed (x, y)-paths in D is equal to the minimum number of arcs
whose deletion destroys all directed (x, y)-paths in D.

Theorem: Let x and y be two vertices of a graph G. Then the maximum number
of edge-disjoint (x, y)- paths in G is equal to the minimum number of edges whose
deletion destroys all (x, y)-paths in G.
14.2. APPLICATIONS 89

Corollary: A graphG is k− edge connected if and only if any two distinct vertices
of G are connected by at least k edges-disjoint paths.

Theorem: Let x and y be two vertices of a digraph D, such that x is not joined
to y. Then the maximum number of internally-disjoint directed (x, y) -paths in D
is equal to the minimum number of vertices whose deletion destroys all directed
(x, y)-paths in D.

0
Proof: Construct a new digraph D from D as follows:

0
• split each vertex v ∈ V \{x, y} into two new vertices v and v ” , and join them
0
by an arc (v, v );
0
• replace each arc of D with head v ∈ V \{x, y} by new arc with head v , and each
arc of D with tail v ∈ V \{x, y} by a new arc with tail v ” . This construction
is illustrated in figure.

0
Now to each directed (x, y) path in D there corresponds a directed (x, y)-paths in
0
D obtained by contracting all arcs of type (c , v ” ); and, conversely to each directed
0
(x, y)-path in D, there corresponds a directed (x, y)-path in D obtained by splitting
0
each internal vertex of the path.Furthermore, two directed (x, y)-paths in D are
arc-disjoint if and only if the corresponding paths in D are internally-disjoint. It
0
follows that the maximum number of arc-disjoint directed(x, y)-paths in D is equal
to the maximum number of internally -disjoint directed(x, y)-paths in D. Similarly,
0
the minimum number of arcs in D whose deletion destroys all directed (x, y)-paths
is equal to the minimum number of vertices’s in D whose deletion destroys all di-
rected (x, y)-paths.

Theorem: Letx and y be two non adjacent vertices’s of a graph G. Then the
maximum number of internally-disjoint (x, y)-paths in G is equal to the minimum
number of vertices’s whose deletion destroys all (x, y)-paths.

Corollary: A graph G with v ≥ k + 1 is k-connected if and only if any two distinct

vertices’s of G are connected by at least k internally disjoint paths.
90 CHAPTER 14. MAX-FLOW MIN-CUT THEOREM

Exercise

1. Let G be a graph and let S and T be two disjoint subsets of V , Show that
the maximum number of vertex-disjoint paths with one end in S and one end
in T is equal to the minimum number of vertices’s whose deletion separated
S from T (that is, after deletion no component contains a vertex of S and a
vertex of T )

2. Show that if G is k-connected with k ≥ 2, then any k vertices’s of G are

contained together in some cycle.

Notes:
14.2. APPLICATIONS 91

DISTANCE EDUCATION - CBCS

MODEL QUESTION PAPER
M.Sc., DEGREE EXAMINATION, NOVEMBER 2019
Mathematics
GRAPH THEORY
(2018-2019 onwards)

Time: 3 hours Maximum: 75 Marks

PART A (10×2=20)

Answer all questions.

1. Define simple graph.

2. Define tree.

3. What is block?

4. Explain shortly in Ramsay’s numbers.

5. If G is bipartite, then χ0 = ∆ + 1.

6. Prove that every critical graph is a block.

7. If G is planar, prove that every subgraph of G is planar.

8. Define coloring.

9. What is network?

10. Define cuts.

PART B (5×5=25)

Answer all questions choosing either (a) or (b).

11. (a) Prove that in any group of n persons(n ≥ 2), there are at least two with
the same number of friends.

(Or)
92 CHAPTER 14. MAX-FLOW MIN-CUT THEOREM

(b) If δ ≥ 2, then show that G contains a cycle.

12. (a) Show that in a tree, any path of maximum length contains the center of
the tree.

(Or)

(b) Prove that a set S is an independent set of G if and only if V \ S is a

covering of G.

13. (a) Let G be a connected graph that is not an odd cycle. Then G has a
2-edge colouring in which both colours are represented at each vertex of
degree at least two.

(Or)

(b) Prove that a graph G is embeddable in the plane if and only if it is

embeddable on the sphere.

14. (a) If G is a loopless bipartite graph, prove that χ0 (G) = ∆(G).

(Or)

(b) Prove that every planar graph is 5-colourable.

15. (a) For any flow f and any cut (S, s) in N, prove that

valf = f + (S) − f − (S)

(Or)

(b) Write the applications of Max-flow and min-cut theorem.

PART C (3×10=30)

Answer any three questions.

16. i) Show that in a graph, the number of edges common to a cycle and an edge
cut is even.
ii) Give an example of a graph with n vertices and n − 1 edges that is not a
tree.

17. Prove that a matching M in G is a maximum matching if and only if G contains

no M −augmenting path.
14.2. APPLICATIONS 93

18. State and Prove Brook’s theorem.

19. Prove that the edges of a connected graph G = (V, E) can be oriented so that
the resulting digraph is strongly connected iff every edge of G is contained in
at least one cycle.

20. In any network, prove that the value of a maximum flow is equal to the capacity
of minimum cut.

*******
94 CHAPTER 14. MAX-FLOW MIN-CUT THEOREM

REFERENCE BOOKS:
1. J. A. Bondy and U.S.R Murty, Graph Theory with Applications Macmillan, Lon-
don
2. A Text book of Graph Theory, Balakrishnan, R, Ranganathan .K Second Edition,
Springer.
3. Invitation to Graph Theory, S. Arumugam and S. Ramachandran, Sci-tech Pub-
lications India.