August 31, 2014 18:46 WSPC/Book Trim Size for 9in x 6in bgtbook˙saidur

Chapter 5

Matching and Covering

5.1 Matching

A matching in a graph is a set of non-loop edges with no common endpoints.

The set {(a, e), (b, c), (d.f )} of edges is a matching in graphs in Figs. 5.1(a)
and (b). The vertices incident to the edges of a matching M are saturated by
M ; the others are unsaturated. In the graph in Fig 5.1(a) vertices a, b, c, d, e
and f are saturated wheras the vertex g is unsaturated.
c c
b b

f f

a g
a

d e d
e
(a) (b)

Fig. 5.1 (a) Illustration of matching.

5.1.1 Perfect Matching

A perfect matching in a graph is a matching that saturates every vertex.

The matching {(a, e), (b, c), (d.f )} is a perfect matching in the graph in

71
August 31, 2014 18:46 WSPC/Book Trim Size for 9in x 6in bgtbook˙saidur

72 Basic Graph Theory

Fig 5.1(b), since all vertices are saturated. A complete graph of odd vertices
does not have a perfect matching, but a complete graph of even vertices
always has a perfect matching. A graph may have many perfect matchings.
For example, Kn,n has n! perfect matchings. The bipartite graph in Fig. 5.2
has unique perfect matching.

Fig. 5.2 A bipartite graph with unique perfect matching.

5.1.2 Maximum Matching

A maximal matching in a graph is a matching that cannot be enlarged by

adding an edge. A maximum matching is a matching of maximum size
among all matchings in the graph. The matching {(b, c)} in Fig 5.3 is a
maximal matching but not a maximum matching since there is a larger
matching {(a, b), (c, d)} of this graph as shown in Fig. 5.3(b).
d d
a a

b c b c
(a) (b)

Fig. 5.3 (a) Illustration of maximal and maximum matching.

Given a matching M in a graph G, an M -alternating path is a path that

alternates between edges in M and edges not in M . An M -alternating path
is an M -augmenting path if the two endpoints of the path is unsaturated
by M . If G has an M -augmenting path P , one can obtain a new matching
M ′ with one more edge by replacing the edges of M in P with the other
August 31, 2014 18:46 WSPC/Book Trim Size for 9in x 6in bgtbook˙saidur

Matching and Covering 73

edges of P . Thus the following fact holds.

Fact 5.1.1 M is not a maximum matching in G if G has an M -

augmenting path.

On the other hand, it can be proved that if M is not a maximum

matching in G then G has an M -augmenting path.
Let M and M ′ be two matchings in G = (V, E). The symmetric differ-
ence of two matchings M and M ′ is the graph whith the vertex set V whose
edges are all those edges appearing exactly one of M and M ′ . The symmet-
ric difference of two matchings M and M ′ is denoted by M △M ′ . Let S be
the set of edges which is contained in M but not in M ′ and let S ′ be the set
of edges which is contained in M ′ but not in M . Then M △M ′ = S ∪ S ′ . In
Fig. 5.4 M = {(a, b), (c.f ), (d, e), (i, h)}, M ′ = {(a, e), (b, d), (c, f ), (g, h)}
and M △M ′ = {(a, b), (b, d), (d, e), (a, e).(i, h), (g, h)}. Since the edge (c, f )
is contained in both M and M ′ , it does not appear in M △M ′ . Observe
that M △M ′ has two connected components; one is a cycle and the other is
a path. In fact a connected component of the symmetric difference of two
matchings is either a cycle or a path as in the following lemma.

Lemma 5.1.2 Every connected component of the symmetric difference

of two matchings is a path or an even cycle.

Proof. Let M and M ′ be two matchings in a graph G and let H =

M △M ′ .
We first claim that Δ(H) ≤ 2. At most one edge from a matching is
incident to a vertex of G. Since M and M ′ are both matchings, at most
two edges can be incident to a vertex of H. Hence Δ(H) ≤ 2.
Since Δ(H) ≤ 2, every connected component of H is either a path or a
cycle. One can observe that every path or cycle in H alternates between
edges of S or S ′ , where S is the set of edges which is contained in M but
not in M ′ , and S ′ is the set of edges which is contained in M ′ but not in
M . To close a cycle C in H, C must have equal number of edges from M
August 31, 2014 18:46 WSPC/Book Trim Size for 9in x 6in bgtbook˙saidur

74 Basic Graph Theory

and M ′ , and hence C is an even cycle. �

b c

i
g
d
a

f h
e

Fig. 5.4 Symmetric difference of two matchings.

We now ready to prove the following lemma.

Lemma 5.1.3 A matching M in a graph G has an M -augmenting path

if M is not a maximum matching in G.

Proof. Let M ′ be a matching in G larger than M . We will prove that

G has an M -augmenting path. We give a constructive proof. Let H =
M △M ′ . By Lemma 5.1.2, every connected component of H is either a
path or a cycle. If every component of H is a cycle then |M | = |M ′ |. Since
M ′ | > M , H has a connected component which is a path P containing
more edges of M ′ than of M . Then P starts with an edge of M ′ and also
ends with an edge of M ′ . Since P contains alternating edges from M and
M ′ , P is an M -augmenting path. �

Fact 5.1.1 and Lemma 5.1.3 immediately prove the following theorem
due to Berge [Ber57].

Theorem 5.1.4 A matching M in a graph G is a maximum matching in

G if and only if G has no M -augmenting path.
August 31, 2014 18:46 WSPC/Book Trim Size for 9in x 6in bgtbook˙saidur

Matching and Covering 75

5.1.3 Hall’s Matching Condition

5.2 Independent Set

A set S of vertices of a graph G is independent if no two of its vertices

are adjacent in G. The subgraph of G induced by the vertices in S is a
null graph. Figure 5.5(a) and (b) illustrate independent sets of 3 and 5
vertices, respectively, where vertices in the independen sets are drawn by
while circles. A set containing a single vertex trivially an independent set.
Finding a larger independent set needs careful checking of the adjacency of
vertices. An independent set S is maximal if S is not a proper subset of any
other independent set of G. An independent set S is maximum if no other
independent set has more vertices than S. The independent set shown in
Figure 5.5(a) is a maximal independent set whereas the independent set
shown in Figure 5.5(b) is a maximum independent set. The independence
number of G, denoted by α(G), the the number of vertices in a maximum
independent set. For the graph in Figure 5.5, α(G) = 5.

(a) (b)

Fig. 5.5 Illustration for independent set.

5.3 Covers

A vertex cover of a graph G = (V, E) is a set Q ⊆ V that contains at least

one endpoint of every edge.
The government plans to establish police check post with sophisticated
August 31, 2014 18:46 WSPC/Book Trim Size for 9in x 6in bgtbook˙saidur

76 Basic Graph Theory

equipment in road crossings of a city in such a way that every road has
a check post. In a graph model of the city, where each vertex represents
a road crossings and each edge represents a road, a vertex cover gives a
feasible solution for the locations of police check posts. If the government
wishes to minimize the number of police checkposts for budget constraint,
a minimum vertex cover gives a feasible solution.

Lemma 5.3.1 Let G = (V, E) be a graph. Then a set S ⊆ V is an

independent set of G if and only if V − S is a vertex cover of G.

Proof. Necessity. Let S be an independent set of G. We show that

V − S is a vertex cover. Let (u, v) be an arbitrary edge of G. Since S is
an independent set, S cannot contain both u and v; one of u and v will
be contained in V − S. This implies that every edge has at least one end
vertex in V − S, and hence V − S is a vertex cover.
Sufficiency. Assume that V − S is a vertex cover. We show that S is an
independent set. Assume for a contradiction that S is not an independent
set. Then there is a pair of adjacent vertices u and v in S. That means
there is an edge (u, v) such that none of its end vetices belongs to V − S,
and hence V − S is not a vertex cover, a contradiction. �

An edge cover of G is a set L of edges such that every vertex of G is

incident to some edge of L.

5.4 Dominating Set

For a graph G = (V, E), a set D ⊆ V (G) of vertices is a dominating set of G

if every vertex in V is either in D or adjacent to a vertex of D. A dominating
set D of G is minimal if D does not properly contain a dominating set of
D. A dominating set D of G is minimum if no other dominating set has
fewer vertices than D. The cardinality of a minimum dominating set of
G is called the domonation number of G and denoted by γ(G). We say a
August 31, 2014 18:46 WSPC/Book Trim Size for 9in x 6in bgtbook˙saidur

Matching and Covering 77

vertex in a dominating set dominates itself and all of its neighbors.

b e
c
f
a
g
d i
h

Fig. 5.6 illustration for domination set.

The government plans to establish fire stations in a new city in such a

way that a locality or one of its neigbor locality will have a fire station.
In a graph model of the city, where each vertex represents a locality and
each edge represents the neighborhood of two localities, a dominating set
gives a feasible solution for the locations of fire stations. If the govern-
ment wishes to minimize the number of fire stations for budget constraint,
a minimum dominating set gives a feasible solution. Domination of ver-
tices has been studied extensively due its practical applications in scenarios
described above [HHS98].
Domination number has a relation with diameter of a graph as we see
in the following lemma.

Lemma 5.4.1 Let G be a connected simple graph, γ(G) be the domination

number of G and diam(G) be the diameter of G. Then
� �
diam(G) + 1
γ(G) ≥ .
3

Proof. Let x and y be two vertices of G such that d(x, y) = diam(G) = k,

and let P = u0 (= x), u1 , · · · , uk (= y) be a path of length k in G from x
to y. Let D be a domination set of G. We now prove that each vertex in
D can dominate at most three vertices on P . Let u be a vertex in D. If
u is on P , u can dominate at most three vertices on P : v itself and its (at
most) two neighbors. If u is not on P , u can also dominates at most three
August 31, 2014 18:46 WSPC/Book Trim Size for 9in x 6in bgtbook˙saidur

78 Basic Graph Theory

vertices on P and those vertices must be consecutive on P ; otherwise, there

would exists a path between x and y shorter than P , a contradiction to the
definition of diam(G). Therefore, each vertex in D dominates at most three
vertices on P .
Since the number of vertices on P is k + 1 = diam(G) + 1, γ(G) ≥
⌈ diam(G)+1
3 ⌉. �

We now describe an algorithm to compute domination number. If the

graph has a few number of vertices we can compute domination number by
simple observation. Thus the idea is to make the graph into smaller pieces
recursively by deleting edges, compute domination numbers of smaller
pieces and compute the domination number of a larger graph from the
domination number of its smaller pieces.
Let e = (u, v) be an edge of a simple graph G = (V, E). Let Ev (u) be
the set of edges incident to u not including the edge (u, v). Similarly Eu (v)
be the set of edges incident to v not including the edge (u, v). Then it is
known that the domination number γ(G) satisfies

γ(G) = min({γ(G − e), γ(G − Ev (u)), γ(G − Eu (v))}). (5.1)

Using Eq. 5.1 we recursively compute γ(G). Of course, the algorithm

is not efficient. In fact the decision version of the problem is known to be
NP-Complete and hence it is unlikely to have a polynomial time algorithm
to compute domination number.
In recent years, a variation of dominating set called connected dominat-
ing set has been studied extensive due to its application in wireless sensor
networks [DW13, YWWY13]. A dominating set D of a graph G is a con-
nected dominating set if the vertices in D induces a connected subgraph of
G. If D is a connected dominating set, one can form a spanning tree of G
in which D forms the set of non-leaf vertices of the tree; conversely, if T
is any spanning tree in a graph with more than two vertices, the non-leaf
August 31, 2014 18:46 WSPC/Book Trim Size for 9in x 6in bgtbook˙saidur

Matching and Covering 79

vertices of T form a connected dominating set. Therefore, finding minimum

connected dominating sets is equivalent to finding spanning trees with the
maximum possible number of leaves.

5.5 Factor of a graph

A factor of a graph G is a spanning subgraph of G. A k-factor is a spanning

k-regular subgraph. Clearly, a 1-factor is a perfect matching and exists only
for graphs with an even number of vertices. A 2-factor of G is a disjoint
union of cycles of G if the 2-factor is not connected; a connected 2-factor
is a Hamiltonian cycle.
We now read Tutte’s condition for 1-factor theorem. A connected com-
ponent H of graph is an odd component if H has odd number of vertices.
We denote by oc(G) the number of odd components in a graph G. The
following theorem is from Tutte.

Theorem 5.5.1 A graph G has a 1-factor if and only if oc(G − S) ≤ |S|

for every S ⊆ V (G).

If we delete the vertex x from the graph in Figure 5.8(a) then we get
two odd components. Taking S = {x}, the graph violates the condition
in Theorem 5.5.1, and hence it has a 1-factor. However the graph in Fig-
ure 5.8(b) satisfies the condition in Theorem 5.5.1 and it has a 1-factor as
shown by thick edges.
We now see some applications of Theorem 5.5.1. The next theorem
gives a necessary and sufficient condition for a tree to have a 1-factor, and
the proof presented here is due to Amahashi [AK11, Ama85].

Theorem 5.5.2 (Chungphaisan ). A tree T of even order has a 1-factor

if and only if oc(T − v) = 1 for every vertex v of T .

Proof. Assume that T has a 1-factor F . Then for every vertex v of T ,

let w be the vertex of T joined to v by an edge of F . It follows that the
August 31, 2014 18:46 WSPC/Book Trim Size for 9in x 6in bgtbook˙saidur

80 Basic Graph Theory

component of T − v containing w is odd, and all the other components of

T −v are even (Fig. 2.25). Hence oc(T −v) = 1. Suppose that oc(T −v) = 1
for every v ∈ V (T ). It is obvious that for each edge e of T , T −e has exactly
two components, and both of them are simultaneously odd or even. Define
a set F of edges of T as follows: F = {e ∈ E(T ) : oc(T − e) = 2}. For every
vertex v of T , there exists exactly one edge e that is incident with v and
satisfies oc(T − e) = 2 since T − v has exactly one odd component, where e
is the edge joining v to this odd component (Fig. 2.25). Therefore e is an
edge of F , and thus F is a 1-factor of G. �

Exercise

1. Compute the number of perfect matchings in Kn .

2. Write a formal proof of Theorem 5.1.4.
3. Show that every k-regular bipartite graph has a perfect matching.
4. How many maximum independent sets of vertices can a path of n vertices
have?
5. Let T be a tree in which every leaf is of distance 2 from another leaf.
Show that every maximum independent set of T must contain all the
leaves of T .
6. Show that a tree in which every maximal path has an even length has a
unique maximum independent set. Can you describe the vertices of this
unique set?
7. Show that a graph with n vertices, m edges and n − m components is
acyclic.
8. A dominating set D of a graph G is an independent dominating set if D
is an independent set of G. Show that D is an independent dominating
set if and only if D is a maximal independent set.
9. Prove or disprove: if a tree has a 1-factor, then the 1-factor is uniqe.
August 31, 2014 18:46 WSPC/Book Trim Size for 9in x 6in bgtbook˙saidur

Matching and Covering 81

f
c
b i g

a d h
j
e
k
l

f f
c c
b i g b i g
a d h a d h
j j
e e
k k
l l

f
c
b i g

a d h
j
e
k
l

f
c
b i g

a d h
j
f f
e
c c k
b b l
i g i g
f
a d h a d h c
j j
b i g
e e
k k
l l a d h
j
e
k
l
f
c
b i g

a d h
j
e
k
l

f
c
b i g

a d h
j
e
k
l

f f
c c
b b i g
i g

a h a d h
d j
j
e e
k k
l l

f
c
b i g

a d h
j
e
k
l

Fig. 5.7 Computation of domination number.

August 31, 2014 18:46 WSPC/Book Trim Size for 9in x 6in bgtbook˙saidur

82 Basic Graph Theory

(a)

(b)

Fig. 5.8 (a) A graph violating the condition in Theorem 5.5.1 and (b) a graph violating
the condition in Theorem 5.5.1