Journal of Combinatorial Optimization

https://doi.org/10.1007/s10878-021-00779-1

On the complexity of minimum q-domination partization

problems

Sayani Das1 · Sounaka Mishra1

Accepted: 8 July 2021

© The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2021

Abstract
A domination coloring of a graph G is a proper vertex coloring with an additional
condition that each vertex dominates a color class and each color class is dominated
by a vertex. The minimum number of colors used in a domination coloring of G is
denoted as χdd (G) and it is called domination chromatic number of G. In this paper, we
give a polynomial-time characterization of graphs with domination chromatic number
at most 3 and consider the approximability of a node deletion problem called minimum
q-domination partization. Given a graph G, in the Minimum q-Domination Partization
problem (in short Min-q-Domination-Partization), the objective is to find
a vertex set S of minimum size such that χdd (G[V \S]) ≤ q. For q = 2, we prove
that it is APX-complete and is best approximable
√ within a factor of 2. For q = 3, it is
approximable within a factor of O( log n) and it is equivalent to minimum odd cycle
transversal problem.

Keywords Domination coloring · q-domination partization · Node deletion

problem · Approximation algorithm

1 Introduction

A dominating set in G is a vertex set D such that D ∩ N [v] ̸= ∅, for each vertex
v ∈ V . In minimum dominating set problem, given a graph G, it is required to compute
a dominating set D of minimum size in G. γ (G) denotes the size of a minimum
dominating set and it is called domination number of G. A proper vertex coloring of
a graph G = (V , E) is a vertex coloring such that no two adjacent vertices have the
same color and the set of vertices having the same color is called a color class. The

B Sounaka Mishra
[email protected]
Sayani Das
[email protected]

1 Department of Mathematics, IIT Madras, Chennai 600036, India

123
 Journal of Combinatorial Optimization

color classes under a proper vertex coloring are independent sets in G. The minimum
chromatic number problem is to compute a proper vertex coloring of a given graph
G with minimum number of colors. The minimum number of colors used in a proper
vertex coloring of G is called chromatic number of G and it is denoted as χ (G).
These two problems are known to be NP-complete. It is known that, unless P = NP,
minimum chromatic number cannot be approximated within a factor of n 1−ε , for any
ε > 0 (Zuckerman 2006). Whereas γ (G) can be approximated within a factor of
O(log n) (Chvatal 1979) and it is the best possible approximation factor (Dinur and
Steurer 2014).
A set of vertices S in G is dominated by a vertex v ∈ V if S ⊆ N [v]. If G
has a vertex v with d(v) = 0 then {v} is dominated by the vertex v itself. If S is
an independent set with at least 2 vertices and S is dominated by v then v ∈ / S.
By imposing domination property on the color classes of a proper vertex coloring,
different types of vertex coloring problems are defined such as dominator coloring
(Gera et al. 2006), domination coloring (Zhou and Zhao 2019) and dominated coloring
(Swaminathan and Sundareswaran 2010; Merouane et al. 2015; Shalu et al. 2020).
Imposing domination property on coloring, Gera et al. (2006) introduced dominator
coloring problem. A dominator coloring of a graph G is a proper vertex coloring in
which every vertex of G dominates some color class. The minimum number of colors
used in a dominator coloring of G is denoted as χd (G) and it is called dominator
chromatic number of G. G is said to be q-dominator colorable if χd (G) ≤ q. Gera et
al. proved that it is NP-complete by a reduction from minimum chromatic number and
also studied this problem for different graph classes and compared χd (G) with χ (G)
and γ (G) (Gera et al. 2006; Gera 2007a, b). Chellali and Maffray (2012) showed For
P4 -free graphs χd (G) can be computed in polynomial time and gave a polynomial-time
characterization of graphs with χd (G) ≤ 3. This problem is known to be NP-complete
for bipartite, split or planar graphs and it has a FPT algorithm for chordal graphs
(Arumugam et al. 2011).
Swaminathan and Sundareswaran (2010) modified the domination property in dom-
inator coloring problem and introduced a new coloring problem called color class
domination. Later, Merouane et al. (2015) referred this problem as minimum dom-
inated coloring problem. This problem is also known as class domination coloring
or CD coloring problem. A CD coloring of a graph G is a proper vertex coloring in
which every color class is dominated by some vertex v ∈ G. The minimum number of
colors used in a CD coloring of G is denoted as χcd (G) and it is called CD chromatic
number of G. G is said to be q-cd colorable if χcd (G) ≤ q. Merouane et al. (2015)
proved that minimum CD coloring is NP-complete for graphs with χcd (G) ≥ 4 and
designed polynomial time algorithm for recognizing graphs with χcd (G) ≤ 3. Sub-
sequently, Chen (2014) proved that minimum CD coloring is NP-hard to approximate
1−ε
within a factor of n 1.5+1 , for any ε > 0. Chen also gave the first nontrivial exact
exponential time algorithm for minimum cd-coloring problem and a polynomial time
O(log n) factor approximation for bipartite graphs. Shalu et al. (2020) gave a lower
bound on χcd (G) using maximum subclique problem. They also proved that minimum
CD coloring is NP-complete for chordal graphs. In Das and Mishra (2020) authors
proved that for bipartite graphs minimum CD coloring is hard to approximate within

123
Journal of Combinatorial Optimization

a factor of ( 21 − ε) ln n, for any ε > 0. This result proves that Chen’s O(log n) factor
approximation algorithm for bipartite graphs is an optimal algorithm.
Domination Coloring was introduced by Zhou and Zhao (2019). This problem is
a combination of the above two problems. A domination coloring of a graph G is a
proper vertex coloring in which every color class is dominated by some vertex v ∈ G as
well as every vertex in G dominates some color class. The minimum number of colors
used in a domination coloring of G is denoted as χdd (G) and it is called domination
chromatic number of G. G is said to be q-domination colorable if χdd (G) ≤ q. They
proved that it is NP-complete for general graphs and present some results on graphs
with χdd (G) = χ (G). They have given general bounds and characterizations on
several classes of graphs. We also found a characterization of 2-domination colorable
graphs in their work.
If a graph G is not q-cd colorable, a natural question arises, if we can delete
a minimum number of vertices from G such that the cd-chromatic number of the
resultant graph is at most q? Krithika et al. (2017) defined such a node deletion
problem associated with minimum cd-coloring problem as minimum q-cd partization
problem. In this problem, given a graph G = (V , E) and a positive integer q, aim is
to find a vertex set S of minimum size such that χcd (G[V \S]) ≤ q. They proved that
it is NP-complete for q ∈ {2, 3}, and NP-hard for q ≥ 4. Latter, we proved in one of
our communicated manuscripts that minimum 2-cd partization is APX-complete and
is approximable within a factor of 2. We also prove that minimum 3-cd partization is
equivalent
√ to minimum odd cycle transversal and is approximable within a factor of
O( log n).
We also consider similar problems associated with dominator coloring. In minimum
q-dominator partization problem, given a graph G = (V , E) and a positive integer q,
aim is to find a vertex set S of minimum size such that χd (G[V \S]) ≤ q. We have
proved that minimum 2-dominator partization is APX-complete and is approximable
within a factor of 2, and minimum 3-dominator partization is equivalent √ to minimum
odd cycle transversal and can be approximated within a factor of O( log n).
With a theoretical interest, we define similar partization problems associated
with domination coloring as minimum q-domination partization problem (in short
Min-q-Domination-Partization). Given a graph G and a positive integer
q, in Min-q-Domination-Partization it is required to find a vertex set S of
minimum size such that χdd (G) ≤ q. In Zhou and Zhao (2019), authors proved that
for k ≥ 4 deciding whether χdd (G) ≤ k is NP-hard. In this paper, we consider the
approximability of Min-q-Domination-Partization for q ∈ {2, 3} and they
are defined formally as follows.
• Min-2-Domination-Partization: Given an undirected simple graph G =
(V , E), find a subset S ⊆ V of minimum size such that χdd (G[V \S]) ≤ 2.
• Min-3-Domination-Partization: Given an undirected simple graph G =
(V , E), find a subset S ⊆ V of minimum size such that χdd (G[V \S]) ≤ 3.
We prove that Min-2-Domination-Partization is APX-complete. It is
approximable within a factor of 2 and this approximation factor is the best possible
approximation factor. Then, we have given the characterizations for 3-domination
√
colorable graphs. By using this characterization, we design an O( ln n)-factor

123
 Journal of Combinatorial Optimization

approximation algorithm for Min-3-Domination-Partization. Finally, we

prove that Min-3-Domination-Partization is as hard as minimum odd cycle
transversal problem.

2 Notations and preliminaries

Throughout this paper, we shall use the following notations. G = (V , E) is a graph

with vertex set V and edge set E. We assume that G is a simple undirected graph.
N G (v) = {u | (u, v) ∈ E} is the set of neighbors of v in G. We will use N G [v] =
N G (v)∪{v} as the closed neighborhood of v in G. If there is no confusion, then we will
use N (v) and N [v] as neighborhood of v and closed neighborhood of v, respectively, in
the graph G. Given two graphs G 1 = (V1 , E 1 ) and G 2 = (V2 , E 2 ) with V1 ∩ V2 = ∅,
the union of G 1 and G 2 is the graph G 1 ∪ G 2 = (V1 ∪ V2 , E 1 ∪ E 2 ). A biclique is
a complete bipartite graph K r ,s with r ≥ 1 and s ≥ 1. If G = (V , E) is a biclique
then the vertex set V has a unique bipartization V = X ∪Y ˙ such that all the edges of
G are across the vertex sets X and Y . For a given set of vertices A in G, we define
E[A] as the set of edges of G with both end vertices in A. Given a pair of disjoint
vertex sets X and Y in G, we define c(X , Y ) = {(u, v) ∈ E | u ∈ X and v ∈ Y } and
c(X , Y ) = {(u, v) ∈/ E | u ∈ X and v ∈ Y }.
A subset O ⊆ V is an odd cycle transversal in G if O has a nonempty intersection
with every odd cycle in G. In minimum odd cycle transversal (Min-OCT) it is required
to find an odd cycle transversal of minimum size in a given simple graph G. A subset
I ⊆ V is an independent set in G if no two vertices in G[I ] are adjacent. A subset
S ⊆ V is a vertex cover in G if S ∩ {u, v} ̸= ∅, for all (u, v) ∈ E. It is easy to observe
that S is a vertex cover in G if and only if V \S is an independent set in G. In minimum
vertex cover problem (Min-VC) it is required to find a vertex cover S of minimum
size in a given simple graph G. Given a graph G = (V , E) and a vertex weight
function w : V → N+ , in minimum weight vertex cover problem ! (Min-W-VC) it is
required to find a vertex cover S of minimum weight w(S) = v∈S w(v). A graph
G is q-domination colorable if χdd (G) ≤ q. Given a graph G = (V , E), a vertex set
S ⊆ V is called a q-domination partization set in G if χdd (G[V \S]) ≤ q. Given a
graph G = (V , E), a set S ⊆ V is called biclique deletion set if G[V \S] is a complete
bipartite graph (or a biclique).
In order to prove the APX-hardness of the mentioned problems, we will use
L-reduction from a known APX-hard problem and it is defined as follows.

Definition 1 (Papadimitriou and Yannakakis 1991) Given two optimization problems

π and π ′ , we say that π L-reduces to π ′ (π ≤ L π ′ ) if there exist polynomial time
computable algorithms f and g and positive constants α and β such that for every
instance I of π , the following properties are satisfied:
1. Algorithm f produces an instance I ′ = f (I) of π ′ such that O P T (I ′ ) ≤ α ·
O P T (I), where O P T (I) and O P T (I ′ ) are the costs of optimal solutions of I
and I ′ , respectively.
2. Given any solution of I ′ with cost c′ , algorithm g produces a solution of I with
cost c such that |c − O P T (I)| ≤ β · |c′ − O P T (I ′ )|.

123
Journal of Combinatorial Optimization

The tuple ( f , g, α, β) is called an L-reduction from π to π ′ .

Proposition 1 (Papadimitriou and Yannakakis 1991) If π ≤ L π ′ and π is APX-hard

then π ′ is also APX-hard.

Next, we mention a few known results about lower and upper bound on the approx-
imability of Min-VC and Min-OCT that we will use in this paper.

Proposition 2 (Khot and Regev 2008) If the Unique Games Conjecture is true then
Min-VC cannot be approximated within (2 − ε), for any ε > 0.
Proposition 3 (Nemhauser and Trotter 1975) Min-W-VC can be approximated within
a factor of 2.

Proposition 4 (Khot 2002) If Unique Games Conjecture is true then Min-OCT does
not admit a constant factor approximation algorithm.
Proposition 5 (Chlebík and Chlebíková 2004)√ Unless P = NP, Min-OCT can not be
approximated within a factor smaller than 8 5 − 15.

Proposition
√ 6 (Agarwal et al. 2005) Min-OCT is approximated within a factor of
O( log n).

3 Minimum 2-domination partization

In this section, we will consider the approximability of Min-2-Domination-

Partization. Our results are based on the following characterization of
2-domination colorable graphs.

Proposition 7 (Zhou and Zhao 2019) If G = (V , E) is a graph with χdd (G) ≤ 2 then
G satisfies one of the following properties.
(i) G is isomorphic to K 1 ,
(ii) G is isomorphic to 2K 1 ,
(iii) G is isomorphic to K r ,s with r ≥ 1 and s ≥ 1.

Lemma 1 Min-2-Domination-Partization is APX-hard.

Proof We prove this by an L-reduction from Min-VC. Given an instance G = (V , E)
of Min-VC, in polynomial time, we construct a graph G ′ = (V ′ , E ′ ) an instance of
Min-2-Domination-Partization as follows.
First, we make two copies G 1 = (V1 , E 1 ) and G 2 = (V2 , E 2 ) of G. We introduce
|V |2 edges such that each vertex in V1 becomes adjacent to each vertex in V2 . We
also, introduce complete bipartite graph K n,n with vertex sets X 1 and X 2 as partites.
Finally, we introduced the edges such that each vertex in X 1 becomes adjacent to each
vertex in V2 , also each vertex in X 2 becomes adjacent to each vertex in V1 . For an
illustration of G ′ we refer to Fig. 1.
If S is a minimal vertex cover in G then F = S1 ∪ S2 is a minimal 2-domination
partization set in G ′ , where S1 and S2 are vertex sets in G 1 and G 2 corresponding to the

123
 Journal of Combinatorial Optimization

X1 X2

G = (V, E)
G1 G2

G′ = (V ′ , E ′ )

Fig. 1 An illustration of construction of G ′ from G

set S, respectively. F is a minimal 2-domination partization set in G ′ because V1 \S1

and V2 \S2 are maximal independent sets in G 1 and G 2 , respectively, also G ′ [V ′ \F] is
a maximal complete bipartite subgraph of G ′ . Therefore, G ′ [V ′ \F] is 2-domination
colorable and F is a 2-domination partization set in G ′ with |F| = 2|S|.
Before proving the converse part, we first claim that if F ∗ is a minimum 2-
domination partization set in G ′ then (X 1 ∪ V1 )\F ∗ and (X 2 ∪ V2 )\F ∗ are nonempty
independent sets in G ′ and F ∗ ⊆ (V1 ∪ V2 ). Without loss of generality, let us assume
that (X 1 ∪ V1 )\F ∗ is not an independent set in G ′ . This implies that G 1 [V1 \F ∗ ] con-
tains at least one edge (such a case can happen when G ′ [V ′ \F ∗ ] is a complete bipartite
subgraph of G 1 ). Then, (X 2 ∪ V2 ) ⊆ F ∗ otherwise G ′ [V ′ \F ∗ ] will contain a triangle.
Also, X 1 ⊆ F ∗ as G ′ [V ′ \F ∗ ] is a connected graph. This implies that |F ∗ | ≥ 3n. This
is a contradiction to the fact that size of a minimum 2-domination partization set in
G ′ can be at most 2n − 2, because (V1 \{x}) ∪ (V2 \{y}) is a 2-domination partization
set for any x ∈ V1 and y ∈ V2 .
Based on these properties of a minimum 2-domination partization set in G ′ , we will
assume that if F is a minimal 2-domination partization set in G ′ then F ⊆ (V1 ∪ V2 ).
Also, V1 \F and V2 \F are independent sets in G 1 and G 2 , respectively. Let F = S1 ∪S2
where S1 = V1 ∩ F and S2 = V2 ∩ F. Without loss of generality, we shall assume
that |S1 | = |S2 |. Suppose |S1 | < |S2 |, then it is easy to prove that F ′ = S1 ∪ S2′ is
a minimal 2-domination partization set in G ′ , where S2′ is the set of vertices in V2
corresponding to the vertices in S1 . Also |F ′ | < |F|. Now, from F we construct a
vertex set S ⊆ V as the set of vertices in V corresponding to the vertices in S1 . It is
easy to observe that |F| = 2|S|.
Now, it is easy to observe that if S ∗ is a minimum vertex cover in G and F ∗ is
a minimum 2-domination partization set in G ′ then |F ∗ | = 2|S ∗ |. Also, from every
minimal 2-domination partization set F in G ′ , in polynomial time, we can construct a
minimal vertex cover S in G with 2|S| = |F|. From these two observations, we have
for every minimal 2-domiantion partization set F in G ′ , (|S| − |S ∗ |) ≤ 21 (|F| − |F ∗ |).
Therefore, it is a L-reduction with α = 2 and β = 21 . /
⊓

123
Journal of Combinatorial Optimization

Since the L-reduction in Lemma 1 is a ratio preserving reduction, by using the

lower bound result in Proposition 2, we have the following lower bound result for
Min-2-Domination-Partization.
Theorem 1 If the Unique Games Conjecture is true then Min-2-Domination-
Partization cannot be approximated within (2 − ε), for any ε > 0.
Next, we prove that Min-2-Domination-Partization is approximable
within a factor of 2 and it is the best possible approximation factor because of Theo-
rem 1. From Proposition 7, it follows that in polynomial time we can verify whether
χdd (G) ≤ 2. Given a graph G, we are interested in computing a vertex set S such
that χdd (G[V \S]) ≤ 2 and |S| is of minimum size. If χdd (G[V \S]) = 1 then a min-
imum size S can be computed in polynomial time. Therefore, it remains to compute
a minimal 2-domination partization set S such that χdd (G[V \S]) = 2, equivalently
G[V \S] is a biclique. In such a case, Min-2-Domination-Partization is
equivalent to minimum biclique deletion problem (Min-Biclique-Deletion).
In Min-Biclique-Deletion, given a graph G, it is required to compute a vertex
set S such that G[V \S] is a biclique. It is a known NP-complete problem. The maxi-
δ
mum biclique subgraph problem is hard to approximate within a factor of 2(log n) , for
3/4+ε
some δ > 0 under the assumption that 3S AT ∈ / DTIME(2n ) for some ε > 0 (Feige
and Kogan 2004). In Hochbaum (1998) Hochbaum designed a 2 factor approxima-
tion algorithm for Min-Biclique-Deletion based on an integer linear program
formulation.
Proposition 8 (Hochbaum 1998) Min-Biclique-Deletion can be approxi-
mated within a factor of 2.
Since Min-2-Domination-Partization and Min-Biclique-
Deletion are equivalent problems, we have the following result.
Theorem 2 Min-2-Domination-Partization can be approximated within a
factor of 2.

4 Minimum 3-domination partization

In this section, we will characterize 3-domination colorable graphs and show that
Min-3-Domination-Partization is APX-hard.
Theorem 3 A graph G satisfies χdd (G) ≤ 3 if and only if G is one of the following
types of graph classes.
– (Type 0) D0 = {G = (V , E) | |V | ≤ 3, |E| = 0}.
– (Type 1) D1 = {K r ,s | r ≥ 1, s ≥ 1}.
– (Type 2) D2 = {K r ,s ∪ K 1 | r ≥ 1, s ≥ 1}.
– (Type 3) A graph G is in the class D3 if the vertex set V is partitioned into three
independent sets X 1 , X 2 , X 3 such that
• every vertex in X 1 is adjacent to every vertex in X 2 .

123
 Journal of Combinatorial Optimization

• every vertex in X 2 is adjacent to every vertex in X 3 .

• G[X 1 ∪ X 3 ] is a bipartite graph with at least one edge.
– (Type 4) A graph G is in the class D4 if its vertex set can be partitioned into four
independent sets X 1 , X 2 , X 3 , X 4 such that
• every vertex in X 2 is adjacent to every vertex in X 3 ,
• every vertex in X 1 is adjacent to every vertex in X 2 ,
• every vertex in X 4 is adjacent to every vertex in X 3 ,
• X 1 ∪ X 4 is an independent set,
• there exists y ∈ X 2 ∪ X 3 such that y is adjacent to every vertex in X 1 ∪ X 4 .
– (Type 5) A graph G is in the class D5 if its vertex set can be partitioned into six
nonempty sets A1 , A2 , A3 , B1 , B2 , B3 such that for i = 1, 2, 3,
• Ai ∪ Bi is an independent set,
• every vertex in Ai is adjacent to all of Ai+1 ∪ Bi+1 (modulo 3),
• every vertex in Bi is adjacent to every vertex in Ai+2 ∪ Bi+2 (modulo 3),
• every vertex in Ai has a non-neighbor in Bi+2 (modulo 3).
Proof First we check that every graph G in D0 ∪ D1 ∪ D2 ∪ D3 ∪ D4 ∪ D5 satisfies
χdd (G) ≤ 3. To see that χdd (G) ≤ 3, we exhibit a coloring of G with at most 3 colors
and we leave it to the reader to check the (obvious) fact that it is a domination coloring.
If G is in D0 , give a different color to every vertex of G as G has no edge.
If G is in D1 , then G is a complete bipartite graph. Therefore, we assign color 1 to
one partite and color 2 to other partite. It is easy to verify that such a vertex coloring
is a domination coloring of G with χdd (G) = 2.
If G is in D2 , Then G has exactly two components, one is an isolated vertex and
the other one is a complete bipartite graph. We assign color 1 to its isolated vertex and
colors 2 and 3 to the two partites of the vertex bipartition of the other component.
If G is in D3 , then V = X 1 ∪X ˙ 2 ∪X
˙ 3 . The color assignment of color 1 to vertices
in X 1 , color 2 to vertices in X 2 and color 3 to vertices in X 3 is a domination coloring
of G.
If G is in D4 , then the color assignment of color 1 to vertices in X 1 ∪ X 4 , color 2
to vertices in X 2 and color 3 to vertices in X 3 is a domination coloring of G, where
V = X 1 ∪X ˙ 2 ∪X
˙ 3 ∪X
˙ 4.
If G is in D5 , then the color assignment of color i to Ai ∪ Bi (i = 1, 2, 3) is a
domination coloring of G.
Now we prove the converse part. If χdd (G) = 1 then G ∼ = K 1 , so G ∈ D0 . If
∼ ∼
χdd (G) = 2 then G = 2K 1 or G = K r ,s with r ≥ 1, s ≥ 1. So, G ∈ D0 or G ∈ D1 .
If χdd (G) = 3 then there will be 3 color classes S1 , S2 and S3 such that S1 ̸= φ,
S2 ̸= φ and S3 ̸= φ. Now we consider following 4 cases:
– Case 1: |S1 | = |S2 | = |S3 | = 1.
It is a trivial case with |V | = 3 and |E| ∈ {0, 1, 2, 3}. If |E| = 2 then χdd (G) = 2
otherwise χdd (G) = 3. If |E| = 0 then G ∈ D0 . If |E| = 1 then G ∈ D2 . If
|E| = 2 then G ∈ D1 and if |E| = 3 then G ∈ D3 .
– Case 2: |S1 | = |S2 | = 1 and |S3 | ≥ 2.
Let S1 = {x} and S2 = {y}. Here, we consider two subcases based on whether x

123
Journal of Combinatorial Optimization

and y are adjacent or not.

Subcase 2.1: (x, y) ∈ / E. Since the vertex x has to dominate a color class and
(x, y) ∈ / E, x has to dominate the color class S3 . Therefore, x is adjacent to every
vertex in S3 . If y is not adjacent to any vertex in S3 , y is dominating itself. So,
S3 ∪ S1 is a complete bipartite graph and hence G ∈ D2 .
Subcase 2.2: (x, y) ∈ E. Then x dominates S2 and y dominates S1 . Since it is
a domination coloring, S3 must be dominated by either x or y. Without loss of
generality, let us assume S3 is dominated by x. Then S3 ⊂ N (x). Since χdd (G) =
3, y must have at least one neighbor in S3 . So, G ∈ D3 . If S3 = N (x) ∩ N (y),
then G is a complete tripartite graph. So, G ∈ D3 .
For an illustration, we refer to Fig. 2.
– Case 3: |S1 | = 1, |S2 | ≥ 2, |S3 | ≥ 2.
Let S1 = {x}. Now, x must dominate at least one of the color classes. We consider
three subcases accordingly.
Subcase 3.1: If x does not dominate S2 or S3 , then x must dominate itself, S1 .
Also every vertex in S2 must be adjacent to every vertex in S3 . So, G ∈ D2 .
Subcase 3.2: Next, we consider the case that x dominates either S2 or S3 . Without
loss of generality, we assume x dominates S2 . Then each vertex in S2 is dominating
the color class S1 . Now, S3 has to be dominated by a vertex v ∈ S1 ∪ S2 and each
vertex in S3 must dominate either of the color classes S1 or S2 . We examine these
domination properties by considering three subcases.
3.2(a): x is adjacent to every vertex in S3 (S3 ⊂ N (x)), then the vertices of S3 is
dominating the color class S1 . Since χdd (G) = 3, there must be at least one edge
between S2 and S3 . So, G ∈ D3 .
3.2(b): x is not adjacent to every vertex in S3 . Then there must exist a y ∈ S2 such
that y is adjacent to every vertex in S3 (y dominates S3 ), so that y dominates S3 .
Now every vertex in S3 must dominate a color class. So the vertices in S3 which
are not adjacent to x must be adjacent to all vertices in S2 . So, G ∈ D4 .
3.2(c): x is not adjacent to any vertex in S3 . Then every vertex in S3 must be
adjacent to every vertex in S2 . This implies G is a complete bipartite graph and
χdd (G) = 2. So G ∈ D1 . Since, we are assuming that χdd (G) = 3 such a case
does not arise.
For an illustration refer to Fig. 3.
– Case 4: |S1 | ≥ 2, |S2 | ≥ 2, |S3 | ≥ 2. Here we consider two subcases.
Subcase 4.1: There is i, j ∈ {1, 2, 3} with i ̸= j such that G[Si ∪ S j ] ∼
= K |Si |,|S j | .
Without loss of generality we assume, every vertex in S2 is adjacent to every vertex
in S3 . Then S2 and S3 are dominated by the vertices of S3 and S2 respectively. Also,
each vertex x ∈ S2 ∪ S3 dominates either S2 or S3 . Now the vertices of S1 must
dominate at least one of the color classes S2 and S3 , and there must be a vertex v
in S2 ∪ S3 that dominates the color class S1 . Depending on the adjacency within
the vertex sets S1 and S2 ∪ S3 , we consider three subcases. For an illustration refer
to Fig. 4.
4.1(a): Every vertex in S1 is adjacent to every vertex in S2 ∪ S3 . Then G is a
complete tripartite graph. So, G ∈ D3 .
4.1(b): Every vertex in S1 is adjacent to every vertex in S2 but not adjacent to
every vertex in S3 . So there is some vertex v ∈ S1 such that there is a u ∈ S3 and

123
 Journal of Combinatorial Optimization

S1 S2 S1 S2 S1 S2

x y x y x y

S3 S3 S3

Fig. 2 An illustration of 3-domination colorable graphs in Case 2

S1 S1 S1

x y S2 x y S2 x y S2

S3 S3 S3

Fig. 3 An illustration of 3-domination colorable graphs in Case 3

S2 S3 S2 S3 S2 S3
y

A B

S1 S1 S1

Fig. 4 An illustration of 3-domination colorable graphs in Subcase 4.1

(v, u) ∈/ E. Since χdd (G) = 3, there is an edge between S1 and S3 . So, G ∈ D3 .

4.1(c): Every vertex in S1 is not adjacent to every vertex in S2 ∪ S3 . But the vertices
of S1 must dominate one of the color class S2 or S3 . So we can partition S1 in two
disjoint set of vertices, A and B such that, every vertex in A is adjacent to every
vertex in S2 and every vertex in B is adjacent to every vertex in S3 . Since S1 must
be dominated by some vertex, so there exists y ∈ S2 ∪ S3 such that y is adjacent
to every vertex in S1 . So, G ∈ D4 .
Subcase 4.2: Here we assume that for each pair i, j ∈ {1, 2, 3} with i ̸= j
the set Si contains at least one vertex that has a non-neighbor in S j ; note that
consequently every such vertex must dominate the third set Sk (k ̸= i, j) because
it is a domination coloring. So, for each i ∈ {1, 2, 3}, Si can be partitioned into
two non-empty sets Ai , Bi such that every vertex of Ai is adjacent to all of Si+1
and every vertex of Bi is adjacent to all of Si+2 (modulo 3), any vertex of Si that
is adjacent to all of Si+1 ∪ Si+2 is placed in Bi . So, G ∈ D5 . For an illustration
refer to Fig. 5. /
⊓

123
Journal of Combinatorial Optimization

A1 A2 A3

B2 B3 B1

Fig. 5 An illustration of 3-domination colorable graphs in Subcase 4.2. The black edges from one set to
another set represents complete connection between the two sets. The red dashed edges from one set to
another represents that there is no edge between those two sets

Theorem 4 There exists a polynomial time algorithm that tests whether a graph G
satisfies χdd (G) ≤ 3.
Proof We will prove this statement using the characterization of graphs with χdd (G) ≤
3. From Theorem 3, it follows that if χdd (G) ≤ 3 then G must belong to one of the
graph classes D0 , D1 , . . . , D5 .
Testing membership in D0 is trivial, because a graph in this class has at most 3
vertices and no edge.
If G is in D1 , then G is a complete bipartite graph and we can test it in polynomial
time. For this we need to check (i) G is odd cycle free graph and (ii) the vertex set can
be partitioned into two disjoint subsets A and B such that for each v ∈ A, N (v) = B
and for each u ∈ B, N (u) = A.
Testing membership of a graph G in D2 can be done, in polynomial time, as follows:
look for an isolated vertex u, and, if u exists, check whether G[V \{u}] is a complete
bipartite graph.
Testing membership of a graph G in D3 can be done as follows: for each component
K of the complementary graph G, test whether K is a clique and G[V \V (K )] is a
bipartite graph with at least one edge.
For testing membership of a graph in D4 , observe that if a graph G is in D4 , then,
with the same notation as in the definition of D4 , there exists a path < x1 , x2 , x3 , x4 >
with (x2 , x4 ) ∈ E obtained by choosing any x1 ∈ X 1 , x2 ∈ X 2 , x3 ∈ X 3 and x4 ∈ X 4 .
Therefore, for each set of 4 vertices x1 , x2 , x3 , x4 ∈ V , we test the following steps in
polynomial time.
(i) G[{x1 , . . . , x4 }] has 4 edges with exactly one vertex of degree 1. Without loss of
generality, we assume that x1 has degree 1, x2 has degree 3 in G[{x1 , . . . , x4 }].

123
 Journal of Combinatorial Optimization

y1
x y2
y3

yn+2
G
Fig. 6 An illustration of G ′

(ii) Construct the vertex sets X 2 = N (x1 ) ∩ N (x3 ) and X 3 = N (x2 ) ∩ N (x4 );
X 4 = {x | x ∈ [N (x2 ) ∩ N (x3 )] and x is adjacent to all vertices in X 3 };
X 1 = {x | x ∈ N (x2 ) and x is adjacent to all vertices in X 2 };
(iii) Verify, whether the vertex sets X 1 , . . . , X 4 satisfy the properties defined in D4 .
Testing membership of a graph G in D5 can be done as follows: if a graph G is
in D5 , then, with the same notation as in the definition of D5 , there exist six vertices
a1 , a2 , a3 , b1 , b2 , b3 of G, obtained by taking a pair of non-adjacent vertices ai ∈
Ai , bi+2 ∈ Bi+2 for each i = 1, 2, 3, with edges (a1 , a2 ), (a2 , a3 ), (a3 , a1 ), (b1 , b2 ),
(b2 , b3 ), (b3 , b1 ), (a1 , b2 ), (a2 , b3 ), (a3 , b1 ) (these vertices induce a C 6 ); also, if we
set Si = N (ai+1 ) ∩ N (ai+2 ) ∩ N (bi+1 ), then Ai is the set of vertices of Si that have a
non-neighbor in Si+2 , and Bi = Si \Ai for i = 1, 2, 3. So we can test if a graph G is
in D5 as follows. For every 6-tuple a1 , a2 , a3 , b1 , b1 , b3 of G that induce a C 6 , define
sets A1 , A2 , A3 , B1 , B2 , B3 as above, and test whether these sets satisfy the desired
properties (which is easy to observe that these tests can be done in polynomial time).
/
⊓

Theorem 5 Min-3-Domination-Partization is APX-hard, for graphs with

diameter 2.

Proof Given an instance G = (V , E) of Min-OCT, in polynomial time, we con-

struct a graph G ′ = (V ′ , E ′ ) an instance of Min-3-Domination-Partization
of diameter 2 as follows. We assume that G has n vertices and define V ′ =
V ∪ {x, y1 , . . . , yn+2 } and E ′ = E ∪ {(v, x) | v ∈ V } ∪ {(x, yi ) | 1 ≤ i ≤ n + 2}.
So, |V ′ | = 2n + 3 and |E ′ | = |E| + 2n + 2. The diameter of G ′ is 2 as the vertex x is
adjacent to all other vertices and there are at least two nonadjacent vertices in G ′ . For
an illustration of G ′ we refer to Fig. 6. /
⊓

Claim G has a minimal odd cycle transversal of size k if and only if G ′ has a minimal
set of k vertices whose deletion results in a 3-domination colorable graph.

Proof Let S ⊆ V be a minimal odd cycle transversal in G with |S| = k. Then G[V \S]
is a bipartite graph with vertex bipartition A and B. Now in the graph G ′ [V ′ \S], the
vertex sets {x}, A and B ∪ {y1 , . . . , yn+2 } are independent sets and each vertex except
x, dominates {x} and x dominates A. x is dominated by rest of the vertices and other
two color classes are dominated by x. Therefore, G ′ [V ′ \S] is 3-domination colorable
graph and S is a 3-domination partization set in G.

123
Journal of Combinatorial Optimization

Conversely, let S ′ be a minimal 2-domination partization set in G ′ . If x ∈ S ′ , then

G ′ \{x} will have n +2 isolated vertices. Since χdd (G ′ [V ′ \S ′ ]) ≤ 3, S ′ must contain at
least n + 1 vertices from the vertex set {y1 , . . . , yn+2 }. This implies that |S ′ | ≥ n + 2.
But a minimum size 3-domination partization set in G ′ has size at most n − 1, because,
for any v ∈ V , T = V \{v} is a 3-domination partization set in G ′ as G ′ [V ′ \T ] is
a complete bipartite graph. Therefore, x ∈ / S ′ and S ′ is a proper subset of V . Since,
χdd (G ′ [V ′ \S ′ ]) ≤ 3 and x is adjacent to all the vertices in G ′ [V ′ \S ′ ], the vertex x
must be in one color class, say C1 . As {y1 , y2 , . . . , yn+2 } is an independent set, it will
be in another color class, say C2 . As G ′ [V ′ \S ′ ] is 3-domination colorable, the vertex
set V \S ′ must be divided into two color classes, one will be in C2 and other is C3 .
Therefore, G[V \S ′ ] is a bipartite graph and S ′ is a minimal odd cycle transversal in
G. /
⊓

Let O ∗ and S ∗ be the optimal solutions of Min-OCT and Min-3-Domination-

Partization for the graphs G and G ′ , respectively. From the proof of above
claim, it follows that |O ∗ | = |S ∗ |. Now, it is easy to observe that, Min-OCT ≤ L
Min-3-Domination-Partization with α = 1, β = 1. /
⊓
From Proposition 4, Proposition 5 and the reduction given in Theorem 5, we have
the following lower bound results for Min-3-Domination-Partization.

Theorem 6 If Unique Games Conjecture is true then Min-3-Domination-

Partization does not admit a constant factor approximation algorithm. It is NP-
hard to√
approximate Min-3-Domination-Partization within a factor smaller
than 8 5 − 15. Moreover, these lower bound results for Min-3-Domination-
Partization hold even when instances are restricted to graphs with diameter 2.

5 Algorithm for minimum 3-domination partization

Our aim is to compute a minimal 3-domination partization set S of size as small as

possible. If S is a minimal 3-domination partization set in G then the graph G[V \S]
can be one of the 6 types mentioned in Theorem 3. We compute five 3-domination
partization (approximate) sets S0 , S2 , . . . , S5 corresponding to 6 types of 3-domination
colorable graphs in polynomial time and then choose the one with minimum size. Due
to some technical issues, we combine both Type 1 and Type 3 together while designing
algorithms for different graph classes.
If G has a 3-domination partization set S such that G[V \S] ∈ D0 then computation
of S0 is trivial. In fact, an optimal deletion set S0 will satisfy the property that |S0 | = 0
if |V | ≤ 3 and |E| = 0. If |V | ≥ 4 then |S0 | = |V | − 3 and V \S0 is an independent
set in G.
For other types, we are designing separate algorithm for each type. In each of these
algorithms, we compute either approximate vertex cover or odd cycle transversal in
some related graphs.

123
 Journal of Combinatorial Optimization

5.1 Algorithm for graph class D 2

We assume that G is a graph such that it has a minimum 3-domination partization set
S2∗ such that G[V \S2∗ ] ∈ D2 . In this case, in polynomial time we compute a 2 factor
approximate 3-domination partization set S2 of Type 2. From the description of D2 it
is required that G[V \S2 ] consists of two components with one is a complete bipartite
graph and the other one is an isolated vertex. Because of this structural property, we
take a vertex v ∈ V then we compute a minimal 2-domination partization set Dv in the
subgraph G[V \N [v]] and return Sv = Dv ∪ N (v) as a 3-domination partization set
with respect to the vertex v. We will try for each vertex in G and take the deletion set
of minimum size. For a formal description of this algorithm, we refer to Algorithm 1.

Algorithm 1: Deletion-Type-2(G)
Input: G = (V , E) such that there exists a vertex set S with G[V \ S] ∈ D2
Output: S2 ⊆ V such that G[V \ S2 ] ∈ D2
S2 = ∅; temp = n;
for each vertex v ∈ V do
Construct the subgraph G v = G[V \ N [v]];
Compute a 2 factor biclique deletion set Dv in G[V \ N [v]];
Sv = Dv ∪ N (v);
if |Sv | < temp then
S2 = Sv ;
temp = |S2 |;
end
end
Return S2 ;

Lemma 2 Let G be a graph such that it has a minimum size 3-domination partization
set S2∗ with G[V \S2∗ ] ∈ D2 . Then S2 =Deletion-Type-2(G) is a 2 factor approximate
solution.
Proof Let S2∗ be a minimum size 3-domination partization set in G such that
G[V \S2∗ ] ∈ D2 . Let v ∗ be the vertex that forms the component with only one vertex
and let Dv∗∗ be the minimum biclique deletion set in G[V \N [v ∗ ]]. S2 = N (v) ∪ Dv
be the deletion set given by Algorithm 1 and v be the corresponding isolated vertex.
Now,

|S2 | |N (v)| + |Dv | |N (v ∗ )| + |Dv ∗ |

= ≤
|S2∗ | |N (v ∗ )| + |Dv∗∗ | |N (v ∗ )| + |Dv∗∗ |
(since |N (v)| + |Dv | ≤ |N (v ∗ )| + |Dv ∗ |)
|Dv ∗ |
≤ (since |Dv∗∗ | ≤ |Dv ∗ |)
|Dv∗∗ |
≤ 2.

The last inequality follows from the fact that Dv ∗ is a 2 factor approximate biclique
deletion set in the graph G[V \N [v ∗ ]]. /
⊓

123
Journal of Combinatorial Optimization

5.2 Algorithm for graph class D1 ∪ D3

Next, we will design approximation algorithm for both Type 1 and Type 3. Let G be
a graph such that it has a 3-domination partization set S with G[V \S] ∈ D1 ∪ D3 .
Following the structural properties of the graph class D1 ∪ D3 , we design an algorithm
as follows. For each ordered pair < x, y > of adjacent vertices x, y ∈ V , we compute
a 3-domination partization set Sx y and then choose the one with minimum size as S3 .
Now, it remains to describe the construction of the set Sx y for an ordered pair < x, y >
of adjacent vertices x, y ∈ V . For such an ordered pair < x, y > , construct the sets
Tx y = V \(N [x] ∪ N [y]), B = N (x) and A = N (y)\N (x). For an edge ( p, q) ∈ E,
we define a gadget J pq = (V J pq , E J pq ) with

V J pq = { p, q, pq1 , pq2 , pq3 }

and E J pq = {( p, q), ( p, pq1 ), ( p, pq2 ), ( p, pq3 ), ( pq1 , q), ( pq2 , q), ( pq3 , q)}.

With respect to the pair of vertices x and y, construct a graph Hx y = (Vx y , E x y ) with

Vx y = B ∪ [∪( p,q)∈[E[A]∪c(A,B)] V J pq ]
and E x y = E[B] ∪ [∪( p,q)∈[E[A]∪c(A,B)] E J pq ].
√
Let Ox y be a O( log n) factor odd cycle transversal in Hx y . We claim that Sx y =
Tx y ∪Ox y is 3-domination partization set in G with respect to the ordered pair < x, y > .
From the structure of J pq , it is easy to observe that a minimal odd cycle transversal must
contain at least one of the vertices p and q. Therefore, we assume that Ox y ⊆ (A ∪ B).
Since, for each edge ( p, q) in E[A] ∪ c(A, B), Hx y contains 3 odd cycles with a
common edge ( p, q) and Ox y contains either p, q or both, A\Ox y is an independent
set and each vertex in A\Ox y is adjacent to each vertex in B\Ox y in G. Observe
that, G[B\Ox y ] is a bipartite graph. If G[B\Ox y ] contains at least one edge then
G[V \Sx y ] ∈ D3 , otherwise G[V \Sx y ] ∈ D1 . Formally, we write this algorithm in
Algorithm 2.

Algorithm 2: Deletion-Type-1&3(G)
Input: G = (V , E) such that there exists a vertex set S with G[V \ S] ∈ D1 ∪ D3
Output: S3 ⊆ V such that G[V \ S3 ] ∈ D1 ∪ D3
S3 = ∅; temp = n;
for each ordered pair < x, y > of adjacent vertices x, y ∈ V do
Tx y = V \ [N [x] ∪ N [y]]; √
Construct the graph Hx y and compute a O( log n) factor odd cycle transversal O x y in it;
Sx y = Tx y ∪ O x y ;
if |Sx y | < temp then
S3 = Sx y ;
temp = |S3 |;
end
end
Return S3 ;

123
 Journal of Combinatorial Optimization

Lemma 3 Let G be a graph such that it has a minimum size 3-domination partization
√
set S3∗ with G[V \S3∗ ] ∈ D1 ∪ D3 . Then S3 =Deletion-Type-1&3(G) is a O( log n)
factor approximate solution.

Proof Let S3∗ = Tx ∗ y ∗ ∪ Ox∗∗ y ∗ be a minimum 3-domination partization set and the cor-
responding order pair < x ∗ , y ∗ > of adjacent pair of vertices be x ∗ and y ∗ . Here, Ox∗∗ y ∗
√
and Ox ∗ y ∗ are a minimum size odd cycle transversal and a O( log n)-approximate
odd cycle transversal in the subgraph G[N [x ∗ ] ∪ N [y ∗ ]], respectively. Now

|S3 | |Tx y | + |Ox y | |Tx ∗ y ∗ | + |Ox ∗ y ∗ |

= ≤
|S3∗ | |Tx ∗ y ∗ | + |Ox∗∗ y ∗ | |Tx ∗ y ∗ | + |Ox∗∗ y ∗ |
(since |Tx y | + |Ox y | ≤ |Tx ∗ y ∗ | + |Ox ∗ y ∗ |)
|Ox ∗ y ∗ |
≤ (since |Ox∗∗ y ∗ | ≤ |Ox ∗ y ∗ |)
|Ox∗∗ y ∗ |
"
≤ O( log n)
"
(since Ox ∗ y ∗ is a O( log n) factor approximate solution).

/
⊓

5.3 Algorithm for graph class D4

Next, we design algorithm for the graph class D4 . Let P E 4 =< x1 , x2 , x3 , x4 > is
an ordered set of 4 distinct vertices such that G[{x1 , x2 , x3 , x4 }] contains the path
P4 =< x1 , x2 , x3 , x4 > along with the edge (x2 , x4 ) (path with a particular edge).
From the description of D4 it follows that G has a 3-domination partization set S with
G[V \S] ∈ D4 if and only if G has P E 4 . We compute an approximate 3-domination
partization set S P E 4 in G with respect to each P E 4 in G, and finally return the one
with minimum size.
For a P E 4 , first we construct the following five sets.

X̂ 2 = N (x1 ) ∩ N (x3 )
X̂ 3 = N (x2 ) ∩ N (x4 )
X̂ 4 = [N (x2 ) ∩ N (x3 )]\[ X̂ 2 ∪ X̂ 3 ]
X̂ 1 = N (x2 )\[ X̂ 2 ∪ X̂ 3 ∪ X̂ 4 ]
TP E 4 = V \[ X̂ 1 ∪ X̂ 2 ∪ X̂ 3 ∪ X̂ 4 ]

Based on these sets, we construct a vertex weighted graph H P E 4 = (V P E 4 , E P E 4 ) with

V P E 4 = ∪i=1 4 X̂ and E 4
i P E 4 = ∪i=1 E[ X̂ i ] ∪ c( X̂ 1 , X̂ 4 ) ∪ c( X̂ 1 , X̂ 2 ) ∪ c( X̂ 2 , X̂ 3 ) ∪
c( X̂ 3 , X̂ 4 ). The vertex weight function is defined as w(v) = n, if v ∈ {x1 , x2 , x3 , x4 }
and w(v) = 1, if v ∈ [V P E 4 \{x1 , x2 , x3 , x4 }]. Let C P E 4 be a 2 factor approximate
vertex cover in the vertex weighted graph H P E 4 . Finally, S P E 4 = TP E 4 ∪ C P E 4 is
a 3-domination partization set with G[V \S P E 4 ] ∈ D4 . We formally describe this
algorithm in Algorithm 3.

123
Journal of Combinatorial Optimization

Algorithm 3: Deletion-Type-4(G)
Input: G = (V , E) such that there exists a vertex set S with G[V \ S] ∈ D4
Output: S4 ⊆ V such that G[V \ S4 ] ∈ D4
S4 = ∅; temp = n;
for each P E 4 =< x1 , x2 , x3 , x4 > do
Construct the vertex weighted graph H P E 4 associated with P E 4 ;
Compute 2 factor vertex cover C P E 4 in the graph H P E 4 ;
S P E4 = TP E4 ∪ C P E4 ;
if |S P E 4 | < temp then
S4 = S P E 4 ;
temp = |S4 |;
end
end
Return S4 ;

Now, we claim that S P E 4 is a 3-domination partization set in G with G[V \S P E 4 ] ∈

D4 . We define the sets X i = X̂ i \C P E 4 , for i = 1, . . . , 4. Since the vertices
x1 , x2 , x3 , x4 are assigned with a larger weight, these vertices are not in the vertex cover
C P E 4 . As a result of which xi ∈ X i , for i = 1, . . . , 4. Since C P E 4 contains at least one
end vertices of the edges in ∪i=1 4 E[ X̂ ], each X is an independent set. Inclusion of
i i
the edge set c( X̂ 1 , X̂ 4 ) in H P E 4 implies that there is no cross edge between the vertex
sets X 1 and X 4 . As H P E 4 contains the edge set c( X̂ 1 , X̂ 2 ) ∪ c( X̂ 2 , X̂ 3 ) ∪ c( X̂ 3 , X̂ 4 ),
the induced subgraphs G[X 1 ∪ X 2 ], G[X 2 ∪ X 3 ] and G[X 3 ∪ X 4 ] are bicliques.

Lemma 4 Let G be a graph such that it has a minimum size 3-domination partization
set S4∗ with G[V \S4∗ ] ∈ D4 . Then S4 =Deletion-Type-4(G) is a 2 factor approximate
solution.

Proof We have mentioned that G has a 3-domination partization set S with G[V \S] ∈
D4 if and only if G has an induced P E 4 . Let S4∗ be a minimum size 3-domination
partization set in G and let P E 4∗ be an ordered set of 4 vertices associated with S4∗ .
Then S4∗ = TP E 4∗ ∪ C ∗P E ∗ with C ∗P E ∗ is a minimum weight vertex cover in the vertex
4 4
weighted graph H P E 4∗ . Let S P E 4 be the set computed by Algorithm 3 and P E 4 be the
associated ordered set. Now

|S4 | |TP E 4 | + |C P E 4 | |TP E 4∗ | + |C P E 4∗ |

= ≤
|S4∗ | |TP E 4∗ | + |C ∗P E ∗ | |TP E 4∗ | + |C ∗P E ∗ |
4 4
(since |TP E 4 | + |C P E 4 | ≤ |TP E 4∗ | + |C P E 4∗ |)
|C P E 4∗ |
≤ (since |C ∗P E ∗ | ≤ |C P E 4∗ |)
|C ∗P E ∗ | 4
4
≤ 2.

The last inequality follows from the fact that C P E 4∗ is a 2 factor approximate vertex
cover in the graph H P E 4∗ . /
⊓

123
 Journal of Combinatorial Optimization

5.4 Algorithm for graph class D5

Finally, we will design a 2 factor approximation algorithm for the graph class D5 . From
the definition of the graph class D5 it follows that G ∈ D5 if and only if G contains an
ordered set J6 =< a1 , a2 , a3 , b1 , b2 , b3 > of 6 vertices such that G[J6 ] has exactly
9 edges and this edge set is {(a1 , a2 ), (a2 , a3 ), (a1 , a3 ), (b1 , b2 ), (b2 , b3 ), (b1 , b3 ),
(a1 , b2 ), (a2 , b3 ), (a3 , b1 )}. We compute an approximate 3-domination partization set
S J6 in G with respect to each J6 in G, and finally return the one with minimum size.
For a J6 we compute the following sets.

B̂1 = [N (b2 ) ∩ N (b3 ) ∩ N (a3 )]\N (a1 )

B̂2 = [N (b1 ) ∩ N (b3 ) ∩ N (a1 )]\N (a2 )
B̂3 = [N (b1 ) ∩ N (b2 ) ∩ N (a2 )]\N (a3 )
Â1 = [N (a2 ) ∩ N (a3 ) ∩ N (b2 )]\N (b1 )
Â2 = [N (a1 ) ∩ N (a3 ) ∩ N (b3 )]\N (b2 )
Â3 = [N (a1 ) ∩ N (a2 ) ∩ N (b1 )]\N (b3 )
T J6 = V \[ Â1 ∪ Â2 ∪ Â3 ∪ B̂1 ∪ B̂2 ∪ B̂3 ]

Based on these vertex sets, we construct a vertex weighted graph H J6 = (VH , E H )

with VH = Â1 ∪ Â2 ∪ Â3 ∪ B̂1 ∪ B̂2 ∪ B̂3 and E H = E 1 ∪ E 2 ∪ E 3 , where

3 3
E 1 = {∪i=1 E[ Âi ]} ∪ {∪i=1 E[ B̂i ]},
E 2 = c( Â1 , B̂1 ) ∪ c( Â2 , B̂2 ) ∪ c( Â3 , B̂3 ) and
E 3 = c( Â1 , Â2 ) ∪ c( Â2 , Â3 ) ∪ c( Â3 , Â1 ) ∪ c( B̂1 , B̂2 ) ∪ c( B̂2 , B̂3 ) ∪ c( B̂3 , B̂1 ) ∪
c( Â1 , B̂2 ) ∪ c( Â2 , B̂3 ) ∪ c( Â3 , B̂1 ).

The vertex weight function is defined as w(v) = n, if v ∈ J6 , and w(v) = 1, if

v ∈ VH \J6 . Let C J6 be a 2 factor vertex cover in H J6 . We set S J6 = T J6 ∪ C J6 as
the 3-domination partization set in G with respect to J6 . We express this algorithm
formally in Algorithm 4.
Next, we claim that G[V \S J6 ] ∈ D5 . We define the sets Ai′ = Âi \C J6 , for i =
1, 2, 3 and Bi′ = B̂i \C J6 , for i = 1, 2, 3. Since the vertices in J6 are assigned with
a larger weight these vertices are not in the vertex cover C J6 . As a result of which
ai ∈ Ai′ and bi ∈ Bi′ , for i = 1, 2, 3. Since C J6 contains at least one end vertices
of the edges in E 1 , the sets A′1 , A′2 , A′3 , B1′ , B2′ , B3′ are independent sets. Inclusion
of the edge set E 2 in H J6 implies that the sets A′1 ∪ B1′ , A′2 ∪ B2′ and A′3 ∪ B3′ are
independent sets. At the same time, inclusion of the edge set E 3 implies that the
induced subgraphs G[A′1 ∪ A′2 ], G[A′2 ∪ A′3 ], G[A′1 ∪ A′3 ], G[B1′ ∪ B2′ ], G[B2′ ∪ B3′ ],
G[B1′ ∪ B3′ ], G[A′1 ∪ B2′ ], G[A′2 ∪ B3′ ] and G[A′3 ∪ B1′ ] are bicliques. Now, if there
exists a vertex a1′ ∈ A′1 such that it is adjacent to all the vertices in B3′ , then we exclude
a1′ from A′1 and include it in B1′ . After shifting all such vertices from A′1 to B1′ , we
will denote the resulting vertex sets as A1 and B1 . Similarly, we construct the vertex
sets A2 , B2 , A3 , B3 . Now, it is easy to verify that G[{∪i=1 3 A } ∪ {∪3 B }] ∈ D .
i i=1 i 5

123
Journal of Combinatorial Optimization

Algorithm 4: Deletion-Type-5(G)
Input: G = (V , E) such that there exists a vertex set S with G[V \ S] ∈ D5
Output: S5 ⊆ V such that G[V \ S5 ] ∈ D5
S7 = ∅; temp = n;
for each ordered set J6 =< a1 , a2 , a3 , b1 , b2 , b3 > do
Construct the vertex weighted graph H J6 associated with J6 ;
Compute 2 factor vertex cover C J6 in the graph H J6 ;
S J6 = T J6 ∪ C J6 ;
if |S J6 | < temp then
S5 = S J6 ;
temp = |S5 |;
end
end
Return S5 ;

Lemma 5 Let G be a graph such that it has a minimum size 3-domination partization
set S5∗ with G[V \S5∗ ] ∈ D5 . Then S5 =Deletion-Type-5(G) is a 2 factor approximate
solution.

Proof We have mentioned that G has a 3-domination partization set S with G[V \S] ∈
D5 if and only if G has an induced J6 . Let S5∗ be a minimum size 3-domination
partization set in G and let J6∗ be an ordered set of 6 vertices associated with S5∗ . Then
S5∗ = T J6∗ ∪ C ∗J ∗ with C ∗J ∗ is a minimum weight vertex cover in the vertex weighted
6 6
graph H J6∗ . Let S J6 be the set computed by Algorithm 4 and J6 be the associated
ordered set of 6 vertices. Now

|S5 | |T J6 | + |C J6 | |T J6∗ | + |C J6∗ |

= ≤ (since |T J6 | + |C J6 | ≤ |T J6∗ | + |C J6∗ |)
|S5∗ | |T J6∗ | + |C ∗J ∗ | |T J6∗ | + |C ∗J ∗ |
6 6
|C J6∗ |
≤ (since |C ∗J ∗ | ≤ |C J6∗ |)
|C ∗J ∗ | 6
6
≤ 2.

The last inequality follows from the fact that C J6∗ is a 2 factor approximate vertex
cover in the graph H J6∗ . /
⊓

5.5 Algorithm for Min-3-Domination-Partization

Now, we describe our approximation algorithm for Min-3-Domination-

Partization.

123
 Journal of Combinatorial Optimization

Algorithm 5: Algorithm for Min-3-Domination-Partization

Input: A graph G = (V , E)
Output: S ⊆ V such that G[V \ S] is 3-domination colorable.
if |V | ≤ 3 and |E| = 0 then
S0 = ∅;
end
else
S0 = V \ {x, y, z};
/* where {x, y, z} forms an independent set in G */
end
Compute S2 using Algorithm 1;
Compute S3 using Algorithm 2;
Compute S4 using Algorithm 3;
Compute S5 using Algorithm 4;
S = argmin{|S0 |, |S2 |, |S3 |, |S4 |, |S5 |};
Return S;

Theorem 7 Algorithm√ 5 approximates Min-3-Domination-Partization

within a factor of O( log n).

Proof Let S0 , S2 , . . . , S5 be the approximate deletion sets computed by the algorithms

described in this section. Let S2∗ , S3∗ , S4∗ , S5∗ be the minimum size 3-domination parti-
zation sets in respective graph classes.

|Si | |S j |
∗ ≤ ∗ (as |Si | ≤ |S j |)
|S j | |S j |
"
≤ O( log n).

The last inequality is because of Lemmas 2, 3, 4 and 5. /

⊓

6 Conclusion

We proved that Min-2-Domination-Partization is approximable within a

factor of 2 and it is the best possible approximation factor. We also prove that
Min-3-Domina- tion-Partization √ is equivalent to Min-OCT and is approx-
imable within a factor of O( log n). We expect that Min-3-Domination-
Partization cannot be approximated within a factor better than O(n 1−ε ), for
any ε > 0, unless P = NP.

References
√
Agarwal A, Charikar M, Makarychev K, Makarychev Y (2005) O( log n) approximation algorithms for
min uncut, min 2cnf deletion, and directed cut problems. In: Proceedings of the thirty-seventh annual
ACM symposium on theory of computing, pp 573–581
Arumugam S, Chandrasekar KR, Misra N, Philip G, Saurabh S (2011) Algorithmic aspects of dominator
colorings in graphs. In: International workshop on combinatorial algorithms. Springer, pp 19–30

123
Journal of Combinatorial Optimization

Chellali M, Maffray F (2012) Dominator colorings in some classes of graphs. Graphs Comb 28(1):97–107
Chen YH (2014) The dominated coloring problem and its application. In: International conference on
computational science and its applications. Springer, pp 132–145
Chlebík M, Chlebíková J (2004) On approximation hardness of the minimum 2sat-deletion problem. In:
International symposium on mathematical foundations of computer science. Springer, pp 263–273
Chvatal V (1979) A greedy heuristic for the set-covering problem. Math Oper Res 4(3):233–235
Das S, Mishra S (2020) Lower bounds on approximating some variations of vertex coloring problem over
restricted graph classes. Discrete Math Algorithms Appl 12(06):2050086
Dinur I, Steurer D (2014) Analytical approach to parallel repetition. In: Proceedings of the forty-sixth annual
ACM symposium on theory of computing, pp 624–633
Feige U, Kogan S (2004) Hardness of approximation of the balanced complete bipartite subgraph problem.
Dept. Comput. Sci. Appl. Math., Weizmann Inst. Sci., Rehovot, Israel, Tech. Rep. MCS04-04
Gera R (2007) On dominator colorings in graphs. Graph Theory Notes N Y 52:25–30
Gera R (2007b) On the dominator colorings in bipartite graphs. In: Fourth international conference on
information technology (ITNG’07). IEEE, pp 947–952
Gera R, Rasmussen CW, Horton S (2006) Dominator colorings and safe clique partitions. Congress 181:19–
32
Hochbaum DS (1998) Approximating clique and biclique problems. J Algorithms 29(1):174–200
Khot S (2002) On the power of unique 2-prover 1-round games. In: Proceedings of the thiry-fourth annual
ACM symposium on theory of computing, pp 767–775
Khot S, Regev O (2008) Vertex cover might be hard to approximate to within 2 − ϵ. J Comput Syst Sci
74(3):335–349
Krithika R, Rai A, Saurabh S, Tale P (2017) Parameterized and exact algorithms for class domination
coloring. In: International conference on current trends in theory and practice of informatics. Springer,
pp 336–349
Merouane HB, Haddad M, Chellali M, Kheddouci H (2015) Dominated colorings of graphs. Graphs Comb
31(3):713–727
Nemhauser GL, Trotter LE (1975) Vertex packings: structural properties and algorithms. Math Program
8(1):232–248
Papadimitriou CH, Yannakakis M (1991) Optimization, approximation, and complexity classes. J Comput
Syst Sci 43(3):425–440
Shalu M, Vijayakumar S, Sandhya T (2020) On the complexity of cd-coloring of graphs. Discrete Appl
Math 280:171–185
Swaminathan V, Sundareswaran R (2010) Color class domination in graphs
Zhou Y, Zhao D (2019) On domination coloring in graphs. arXiv preprint arXiv:1909.03715
Zuckerman D (2006) Linear degree extractors and the inapproximability of max clique and chromatic
number. In: Proceedings of the thirty-eighth annual ACM symposium on theory of computing, pp
681–690

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps
and institutional affiliations.

123