International Journal of Scientific and Innovative Mathematical Research (IJSIMR)

Volume 2, Issue 10, October 2014, PP 837-843

ISSN 2347-307X (Print) & ISSN 2347-3142 (Online)
www.arcjournals.org

Characterization of Uniquely Colorable and Perfect Graphs

B.R. Srinivas A. Sri Krishna Chaitanya
Associate Professor of Mathematics, Associate Professor of Mathematics,
St. Marys Group of Institutions Guntur, Chebrolu Engineering College, Guntur
A.P, India A.P, India
[email protected] [email protected]

Abstract: This paper studies the concepts of uniquely colorable graphs & Perfect graphs. The main
results are
1) Every uniquely k-colorable graph is (k - 1)-connected.
2) If G is a uniquely k-colorable graph, then  (G) ≥ k - l.
3) A maximal planar graph G of order 3 or more has chromatic number 3 if and only if G is Eulerian.
4) Every interval graph is perfect.
5) A graph G is chordal if and only if G can be obtained by identifying two complete. Sub graphs if the
same order in two chordal graphs.
Mathematics Subject Classification 2000: 05CXX, 05C15, 05C20, 37E25.
Keywords: complete graph, bipartite graph,k-chromatic graph, planar graph, maximal planar
graph, chromatic number, clique number, chordal graph, replication graph.

1. UNIQUELY COLORABLE GRAPHS

1.1 Definition
Suppose that G is a k-chromatic graph. Then every k-coloring of G produces a partition
of V(G) into k independent subsets (color classes). If every two k-colorings of G result in
the same partition of V(G) into color classes, then G is called uniquely k-colorable or
simply uniquely colorable. Trivially, the complete graph Kn is uniquely colorable. In
fact, every complete k-partite graph, k ≥ 2, is uniquely colorable.
1.2 Theorem
In every k-coloring of a uniquely k-colorable graph G, where k  2, the sub graph of G induced by
the union of every two color classes of G is connected.
Proof. Assume, to the contrary, that there exist two color classes V1 and V2 in some k-
coloring of G such that H = G [V1V2] is disconnected. We may assume that the vertices
in V1 are colored 1 and those in V2 are colored 2. Let H1 and H2 be two components of H.
Interchanging the colors 1 and 2 of the vertices in H1 produces a new partition of V(G)
into color classes, producing a contradiction.
1.3 Note
As a consequence of Theorem1.2, every uniquely k-colorable graph, k ≥ 2, is connected.
In fact, Gary Chartrand, and Dennis Paul Geller [1] showed that more can be said.
1.4 Theorem
Every uniquely k-colorable graph is (k-1)-connected.
Proof. The result is trivial for k=1 and, by Theorem 1.2, the result follows for k=2 as
well. Hence we may assume that k≥3. Let G be a uniquely k-colorable graph, where k ≥
3. If G=Kk, then G is (k - 1)-connected; so we may assume, that G is not complete.
Assume, to the contrary, that G is not (k - 1)-connected. Hence there exists a vertex cut W
©ARC Page | 837
B.R. Srinivas & A. Sri Krishna Chaitanya

of G with W = k - 2.

Let there be given a k-coloring of G. Consequently, there are at least two colors, say 1
and 2, not used to color any vertices of W. Let V1 be the color class consisting of the
vertices colored 1 and V2 the set of the vertices colored 2. By Theorem1.2, H = G[Vl 
V2] is connected, Hence H is a subgraph of some component G1 of G-W. Let G2 be
another component of G-W. Assigning some vertex of G2 the color 1 produces a new k-
coloring of G that results in a new partition of V(G) into color classes, contradicting our
assumption that G is uniquely k-colorable.
1.5 Corollary
If G is a uniquely k-colorable graph, then  (G) ≥ k - l.
Proof: Much of the interest in uniquely colorable graphs has been directed towards
planar graphs. Since every complete graph is uniquely colorable, each complete graph
Kn, 1 ≤ n ≤ 4, is a uniquely colorable planar graph. Indeed, each complete graph Kn, 1 ≤
n ≤ 4, is a uniquely colorable maximal planar graph. Since the complete 3 partite graph
K2,2,2 (the graph of the octahedron) is also uniquely colorable, K2,2,2 is a uniquely 3-
colorable maximal planar graph.(seeFigures1(a)).

Figure (a). uniquely 3-colorable maximal planar graphs

The graph G in Figures 1(b) is also a uniquely 3-colorable maximal planar graph. The
fact that the 3-colorable maximal planar graphs shown in Figure 1 are also uniquely
colorable is not surprising, as Chartrand and Geller [1] observed.
1.6 Note
The two 3-colorable maximal planar graphs in Figure 1 have another property in
common. There are both Eulerian. That this is a characteristic of all maximal planar 3 -
chromatic graphs was first observed by Percy John Heawood [5] in 1898.
1.7 Theorem
A maximal planar graph G of order 3 or more has chromatic number 3 if and only if G is
Eulerian.
Proof. Let there be given a planar embedding of G. suppose first that G is not Eulerian.
Then G contains a vertex v of odd degree k ≥3. Let
N(v) = {v1,v2, ... ,vk,},
Where C = (v1, v2... vk, v1) is an odd cycle in G. Because v is adjacent to every vertex of C,
it follows that (G) = 4.
We verify the converse by induction on the order of maximal planar Eulerian graphs. If
the order of G is 3, then G = K3 and (G) = 3. Assume that every maximal planar Eulerian
graph of order k has chromatic number 3 for an integer k ≥ 3 and let G be a maximal
planar Eulerian graph of order k + 1. Let there be given a planar embedding of G and let
uw be an edge of G. Then uw is on the boundary of two (triangular) regions of G. Let x be
the third vertex on the boundary of one of these regions and y the third vertex on the
boundary of the other region. Suppose that
N(x) = {u=x1,x2,….xk = w} and N(y) ={u= y1,y2,….yl =w},
Where k and  are even, such that C = (x1,x2,….xk,x1) and C  = (y1,y2,….yl,y1) are even
cycles. Let G be the graph obtained from G by (1) deleting x, y, and uw from G and (2)
International Journal of Scientific and Innovative Mathematical Research (IJSIMR) Page 838
Characterization of Uniquely Colorable and Perfect Graphs

adding a new vertex z and joining z to every vertex of C and C  . Then G is a maximal
planar Eulerian graph of order k. By the induction hypothesis, ( G ) = 3. According to
Theorem1.7, G is uniquely colorable. Since z is adjacent to every vertex of C and C  we
may assume that z is colored 1 and that the vertices of C and C  alternate in the colors 2
and 3. From the 3-coloring of G , a 3-coloring of G can be given where every vertex of
V(G) - {x, y} is assigned the same color as in G and x and y are colored 1.
On the other hand, Chartrand and Geller [1] showed that every uniquely 4-colorable
planar graph must be maximal planar.
2. PERFECT GRAPHS
2.1 Definition
For any graph G, if  (G) =  (G), then g is called perfect graph. While there are many
examples of graphs G for which  (G) = (G), such as complete graphs and bipartite graphs,
there are also many graphs whose chromatic number exceeds its clique number such as the
Petersen graph and the odd cycles of length 5 or more. As we are about to see, the chromatic
number of a graph can be considerably larger than its clique number. The fact that a graph
can be triangle-free and yet have a large chromatic number has been established by a
number of mathematicians, including Blanche Descartes [2] John Kelly and Leroy Kelly [6],
and Alexander Zykov [10], Jan Mycielski [9].
2.2 Theorem
Every bipartite graph is perfect.
Proof. Let G be a bipartite graph and let H be an induced sub graph of G. If H is
nonempty, then  (H) =  (H) = 2; while if H is empty, then  (H) =  (H)=1. In either
case,  (H) =  (H) and so G is perfect.
2.3 The Perfect Graph Conjecture
A graph is perfect if and only if its complement is perfect. In 1972, Laszlo Lovasz [7] showed
that this conjecture is, in fact, true.
2.4 Theorem
Every interval graph is perfect.
Proof. Let G be an interval graph with V(G) = {v1 ,v2 …,vn }. Since every induced sub graph
of an interval graph is also an interval graph, it suffices to show that  (G) = (G). Because
G is an interval graph, there exist n closed intervals Ii = [ai, bi], 1 ≤ i ≤ n, such that vi
is adjacent to vj (i ≠ j) if and only if Ii  Ij ≠  . We may assume that the intervals (and
consequently, the vertices of G) have been labeled so that al ≤ a2 ≤ ... ≤ an.
We now define a vertex coloring of G. First, assign v1 the color 1. If v1 and v2 are not
adjacent (that is, if II and I2 are disjoint), then assign v2 the color 1 as well; otherwise,
assign v2 the color 2. Proceeding inductively, suppose that we have assigned colors to v1,
v2, ... , vr where 1 ≤ r < n; We now assign vr+ 1 the smallest color (positive integer) that has
not been assigned to any neighbor of vr+1 in the set {v1, v2, ... , vr }. Thus if vr+ 1 is adjacent to
no vertex in {v1, v2, ..., vr}, then vr+1 is assigned the color 1. This gives a k-coloring of G for
some positive integer k and so  (G) ≤ k. If k = 1, then G = K n and  (G) = (G) = 1.
Hence we may assume that k ≥ 2.
Suppose that the vertex vt has been assigned the color k. Since it was not possible to assign
vt any of the colors 1,2,...k-1, this means that the interval It = [at,bt] must have a
nonempty intersection with k-l intervals Ij1, Ij2..... , Ijk-1 where say 1 ≤ j1 < j2 < ... < jk-1 < t.
Thus ajl ≤ aj2≤ ….,≤ ajk-1≤at. Since Iji  It ≠  for 1 ≤ i ≤ k - 1, it follows that

at  Ij1  Ij2 …. Ijk-1It

Thus for U = {vj1, vj2, ...., vjk-1,vt},
International Journal of Scientific and Innovative Mathematical Research (IJSIMR) Page 839
B.R. Srinivas & A. Sri Krishna Chaitanya

G[U] = Kk
and so  (G) ≤ k ≤ (G). Since  (G) ≥  (G), we have  (G) =  (G), as desired.
2.5 Note
We now consider a more general class of graphs. Recall that a chord of a cycle C in a graph
is an edge that joins two non-consecutive vertices of C. For example, wz and xz are chords in
the cycle C = (u, v, w, x, y, z, u) in the graph G of Figure2; while in the cycle C' = (w, x, y, z,
w) in G, the edge xz is a chord and wz is not. The cycle C" = (u, v, w, z, u) has no chords.
Obviously no triangle contains a chord.

Figure (b). Chords in cycles

2.6 Definition
A graph G is a chordal graph if every cycle of length 4 or more in G has a chord. Since the
cycle C" = (u, v, w, z, u) in the graph G of Figure2 contains no chords, the graph G is not a
chordal graph.
While every complete graph is a chordal graph, no complete bipartite graph Ks,t, where s,t ≥
2, is chordal, for if u1 and v1 belong to one partite set and u2 and v2 belong to the other partite
set, then the cycle (u1, u2,,v1, v2, u1) contains no chord. Indeed, no graph having girth 4 or more
is chordal. The graphs G1 and G2 of Figure 2.1 are chordal graphs. For the subset S1 = {u1, v1,
x1} of V(G1) and, the subset S2 = {u2, w2, x2 } of V(G2), let the graph G3 be obtained by
identifying the vertices in the complete sub graph GI [S1] with the vertices in the complete
sub graph G2[S2], where, say, u1 and u2 are identified, v1 and x2 are identified, and x1 and w2
are identified. The graph G3 shown in Figure 2.1 is also a chordal graph.

Figure 2.1. Chordal graphs

More generally, suppose that G1 and G2 are two graphs containing complete sub graphs H1
and H2, respectively, of the same order and G3 is the graph obtained by identifying the
vertices of H1 with the vertices of H2 (in a one-to-one manner). If G3 contains a cycle of
length 4 or more having no chord, then C must belong to G1 or G2. That is, if G1 and G2 are
chordal, then G3 is chordal. Furthermore, if G3 is chordal, then both G1 and G2 are chordal.
We have now observed that every graph obtained by identifying two complete sub graphs of
the same order in two chordal graphs is also chordal. These are not only sufficient conditions
for a graph to be chordal. They are necessary conditions as well. The following
characterization of chordal graphs is due to Andras Hajnal and Janos Suranyi [4] and Gabriel

International Journal of Scientific and Innovative Mathematical Research (IJSIMR) Page 840
Characterization of Uniquely Colorable and Perfect Graphs

Dirac [3].
2.7 Theorem
A graph G is chordal if and only if G can be obtained by identifying two complete sub graphs
of the same order in two chordal graphs.
Proof. From our earlier observations, we need only show that every chordal graph can be
obtained from two chordal graphs by identifying two complete sub graphs of the same order
in these two graphs. If G is complete, say G = Kn , then G is chordal and can trivially be
obtained by identifying the vertices of G1 = Kn and the vertices of G2 = Kn in any one-to-one
manner. Hence we may assume that G is a connected chordal graph that is not complete.
Let S be a minimum vertex-cut, of G. Now let V1 be the vertex set of one component of G - S
and let V2 = V(G) - (V1  S). Consider the two S-branches. G1 = G[V1  S] and G2 = G[V2
 S]of G. Consequently, G is obtained by identifying the vertices of S in G1 and G2 . We now
show that G[S] is complete. Since this is certainly true if S = 1, we may assume that S ≥
2.
Each vertex v in S is adjacent to at least one vertex in each component of G - S, for otherwise
S- {v} is a vertex-cut of G, which is impossible. Let u, w  S. Hence there are u - w paths in
G1, where every vertex except u and w belongs to V1. Among all such paths, let P = (u, x1,
x2 ….. xs, w) be one of minimum length. Similarly, let P = (u, y1, y2 ..., yt, w) be a u - w path
of minimum length where every vertex except u and w belongs to V2.
Hence
C = (u, x1, x2 ….. xS, w, yt, yt-l,… y1, u)
is a cycle of length 4 or more in G. Since G is chordal, C contains a chord. No vertex xi
(1 ≤ i ≤ s) can be adjacent to a vertex yj (1 ≤ j ≤ t) since S is a vertex-cut of G.
Furthermore, no non-consecutive vertices of P or of P' can be adjacent due to the
manner in which P and P' are defined. Thus uw  E (G), implying that G[S] is complete.
G1 and G2 are chordal.
2.8 Corollary
Every chordal graph is perfect.
Proof. Since every induced sub graph of a chordal graph is also a chordal graph, it
suffices to show that if G is a connected chordal graph, then  (G) =(G). We proceed by
induction on the order n of G. If n = 1, then G = K1 and  (G) =  (G) = 1. Assume
therefore that  (H) =  (H) for every chordal graph H of order less than n, where n ≥ 2
and let G be a chordal graph of order n ≥ 2.
If G is a complete graph, then  (G) = (G) = n. Hence we may assume that G is not
complete. G can be obtained from two chordal graphs G1 and G2 by identifying two
complete sub graphs of the same order in G1 and G2. Observe that
 (G)≤ max{x(Gl),  (G2)} = k.
By the induction hypothesis,  (G1)=(G1} and  (G2) = (G2).

Thus  (G) ≤ max { (G1), (G2)} =k. On the other hand, let S denote the set of vertices in
G that belong to G1 and G2. Thus G[S] is complete and no vertex in V(G1) - S is adjacent to
a vertex in V(G2) - S. Hence (G) = max{(G1),(G2)} = k.
Thus  (G) ≥ k.
Therefore,  (G) = k =  (G).
2.9 Definition
We now consider a class of perfect graphs that can be obtained from a given perfect
International Journal of Scientific and Innovative Mathematical Research (IJSIMR) Page 841
B.R. Srinivas & A. Sri Krishna Chaitanya

graph. Let G be a graph where v  V(G). Then the replication graph Rv(G) of G (with
respect to v) is that graph obtained from G by adding a new vertex v' to G and joining v' to
the vertices in the closed neighborhood N[v] of v. In 1972 Laszlo Lovasz [8] obtained few
results.
2.10 Theorem
Let G be a graph where v  V(G). If G is perfect, then Rv(G) is perfect.

Figure 2.2. Non-chordal graphs

Proof. Let G = Rv(G). First, we show that  ( G ) = ( G ). We consider two cases,
depending on whether v belongs to a maximum clique of G.
Case 1. V belongs to a maximum clique of G. Then ( G ) = (G) + 1. Since
 ( G ) ≤  (G) + 1 = (G) +1= ( G ),
it follows that  ( G ) = ( G ).
Case 2. V does not belong to any maximum clique of G. Suppose that  (G) = (G) = k. Let
there be given a k-coloring of G using the colors 1, 2...k. We may assume that v is
assigned the color 1. Let V1 be the color class consisting of the vertices of G that are
colored 1. Thus v  Vl. Since (G) = k, every maximum clique of G must contain a vertex
of each color. Since v does not belong to a maximum clique, it follows that V1 ≥ 2. Let
U1 = V1 - {v}. Because every maximum clique of G contains a vertex of U1, it follows
that (G-U1) = (G)-1 = k - 1. Since G is perfect,  (G–U1 ) = k -1. Let a (k - 1)-coloring of
G - U1 be given, using the colors 1, 2... k - 1. Since Vl is an independent set of vertices, so
is U1 {v1 }. Assigning the vertices of U1 {v'} the color k produces a k-coloring of G .
Therefore,
k =  (G) ≤ ( G ) ≤  ( G ) ≤ k

and so  ( G ) = ( G ).
It remains to show that  (H) = (H) for every induced sub graph H of G'.
This is certainly the case if H is a sub graph of G. If H contains v' but not v, then H 
G[(V(H)-{v'}){v}] and so  (H) =(H). If H contains both v and v' but H  G , then H is
the replication graph of G[V(H) - {v'}] and the argument used to show that  (G') = (G')
can be applied to show that  (H) = (H).

REFERENCES
[1] Chartrand.G and Geller. D.P., on uniquely colorable planar graphs. Combin.J.
Theory 6(1969) 271-278.
[2] Descartes.B, A three color problem. Eureka (1947) 21.
[3] Dirac.G.D, On rigid circuit graphs. Abh. Math. Sem. Univ. Hamburg 25(1961)71-76.
[4] Hajnal.A and Suranyi.J, Uber die Auflosung Von Graphen in vollstandige

International Journal of Scientific and Innovative Mathematical Research (IJSIMR) Page 842
Characterization of Uniquely Colorable and Perfect Graphs

Teilgraphen. Ann. Univ. Sci. Budapest, Eotvos. Sect. Math. 1 (1958) 113-121.
[5] Heawood.P.J,On the four color map theorem. Quarterly J. Math. 29 (1898) 270-285.
[6] Kelly.J and L.Kelly, Path and circuits in critical graphs. Amer. J. Math.76 (1954)
786-792.
[7] Lovasz.L, A characterization of perfect graphs. J. Combin. Theory Ser. B 13(1972)
95-98.
[8] Lovasz.L, Normal hyper graphs and perfect graphs conjecture. Discrete Math.2
(1972)253-267.
[9] Mycielski.J, Sur le coloriage des graphes. Colloq. Math. 3 (1955) 161-162.
[10] Zykov.A.A, on some properties of linear complexes (Russian). Mat. Sbornik 24
(1949) 163-188.
AUTHORS’ BIOGRAPHY
Mr. B.R. Srinivas , completed his M.Sc., AO Mathematics from Acharya
Nagarjuna university in the year 1995, M.Phil Mathematics from Madurai
Kamaraj University in the year 2005and M.Tech computer science
engineering from Vinayaka Mission University in the year 2007. He
attended for five international conferences, six faculty development
programs and twenty national workshops. At present he is working as
Associate Professor, St. Marys Group of Institutions Guntur, A.P, INDIA.
His area of interest is graph theory and theoretical computer science.

Mr. A. Sri Krishna Chaitanya, completed his M.Sc., Mathematics from

Acharya Nagarjuna university in the year 2005 and M.Phil Mathematics
from Alagappa University in the year 2007. At present he is working as
Associate Professor, Chebrolu Engineering College, Chebrolu, Guntur Dist,
A.P, INDIA.
He is interested to work in the areas of Graph Theory, Boolean algebra,
Lattice Theory and the Related Fields of Algebra.

International Journal of Scientific and Innovative Mathematical Research (IJSIMR) Page 843