Discrete Applied Mathematics 157 (2009) 1634–1639

Contents lists available at ScienceDirect

Discrete Applied Mathematics

journal homepage: www.elsevier.com/locate/dam

Upper bounds on the k-domination number and the k-Roman

domination number
Adriana Hansberg ∗ , Lutz Volkmann
Lehrstuhl II für Mathematik, RWTH Aachen University, 52056 Aachen, Germany

article info a b s t r a c t

Article history: Let G be a graph and let k be a positive integer. A subset D of the vertex set of G is a k-
Received 3 June 2008 dominating set if every vertex not in D has at least k neighbors in D. The k-domination
Received in revised form 9 October 2008 number γk (G) is the minimum cardinality among the k-dominating sets of G. A Roman k-
Accepted 19 October 2008
dominating function on G is a function f : V (G) −→ {0, 1, 2} such that every vertex u
Available online 6 December 2008
for which f (u) = 0 is adjacent to at least k vertices v1 , v2 , . . . , vk with f (vi ) = 2 for
iP= 1, 2, . . . , k. The weight of a Roman k-dominating function is the value f (V (G)) =
Keywords:
Domination v∈V (G) f (v). The minimum weight of a Roman k-dominating function on a graph G is
k-domination called the Roman k-domination number γkR (G).
k-Roman domination In 2007, Rautenbach and Volkmann [D. Rautenbach, L. Volkmann, New bounds on
the k-domination number and the k-tuple domination number, Appl. Math. Lett. 20
(2007) 98–102] gave an upper bound for the k-domination number γk (G). Using again the
probabilistic method, we achieve better bounds for this parameter and prove new bounds
for the k-Roman domination number γkR (G). Moreover, we generalize known inequalities
for the case k = 1 [V.I. Arnautov, Estimations of the external stability number of a graph by
means of the minimal degree of vertices, Prikl. Mat. Programm. 11 (1974) 3–8 (in Russian);
C. Payan, Sur le nombre d’absorption d’un graphe simple, Cahiers Centre Études Recherche
Opér. 17 (1975) 307–317; E.J. Cockayne, P.A. Dreyer Jr., S.M. Hedetniemi, S.T. Hedetniemi,
Roman domination in graphs, Discrete Math. 278 (2004) 11–22].
© 2008 Elsevier B.V. All rights reserved.

1. Terminology and introduction

We consider finite, undirected and simple graphs G with vertex set V (G) and edge set E (G). The number of vertices |V (G)|
of a graph G is called the order of G and is denoted by n = n(G).
The open neighborhood N (v) = NG (v) of a vertex v consists of the vertices adjacent to v and d(v) = dG (v) = |N (v)| is
the degree of v . We identify with δ(G) the minimum of all degrees of the vertices of G. The closed neighborhood of a vertex v
is defined by N [v] = NG [v] = N (v) ∪ {v}.
Let k be a positive integer. A subset D ⊆ V (G) is a k-dominating set of the graph G, if |N (v)∩ D| ≥ k for every v ∈ V (G)− D.
The k-domination number γk = γk (G) is the minimum cardinality among the k-dominating sets of G. In [6,7], Fink and
Jacobson introduced the concept of k-domination. Note that the 1-domination number γ1 coincides with the definition of
the usual domination number γ . For a comprehensive treatment of domination in graphs, see the monographs by Haynes,
Hedetniemi and Slater [8,9].
A Roman k-dominating function on G is a function f : V (G) −→ {0, 1, 2} such that every vertex u for which f (u) = 0
is adjacent to at least k vertices v1 , v2 , . . . , vk with f (vi ) = 2 for i = 1, 2, . . . , k. The weight of a Roman k-dominating

∗ Corresponding author. Fax: +49 241 80 92136.

E-mail addresses: hansberg@math2.rwth-aachen.de (A. Hansberg), volkm@math2.rwth-aachen.de (L. Volkmann).

0166-218X/$ – see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.dam.2008.10.011
 A. Hansberg, L. Volkmann / Discrete Applied Mathematics 157 (2009) 1634–1639 1635

function is the value f (V (G)) = v∈V (G) f (v). The minimum weight of a Roman k-dominating function on a graph G is
P
called the Roman k-domination number γkR = γkR (G). If f : V (G) −→ {0, 1, 2} is a Roman k-dominating function on G, then
let (V0 , V1 , V2 ) be the ordered partition of V (G) induced by f , where Vi = {v ∈ V (G) | f (v) = i} for i = 0, 1, 2. Note that
there is a one-to-one correspondence between the functions f : V (G) → {0, 1, 2} and the ordered partitions (V0 , V1 , V2 ) of
V (G). The Roman 1-domination number γ1R corresponds to the well-known Roman domination number γR , which was given
implicitly by Steward in [13] and ReVelle and Rosing in [12].
In 1985, Cockayne, Gamble and Shepherd [5] proved that, if G is a graph with minimum √ degree δ and k is an integer with
δ ≥ k, then γk (G) ≤ kkn +1
. Under the stronger assumption on the minimum degree that ln δ > k, Caro and Yuster [3] proved
δ
an upper bound for γk (G) for graphs G of order n of the form of (1 + oδ (1)) n ln
δ
. Weakening considerably the condition on
the minimum degree, Rautenbach and Volkmann presented the following upper bound on the k-domination number γk .
Due to the weaker conditions, this bound is as expected not as strong as the previous mentioned bounds.

Theorem 1 (Rautenbach, Volkmann [11] 2007). Let G be a graph on n vertices with minimum degree δ and let k ∈ N. If
δ+1
ln(δ+1)
≥ 2k, then
!
k−1
n X 1
γk (G) ≤ k ln(δ + 1) + .
δ+1 i =0
i! (δ + 1)k−1−i

Following the same probabilistic method as in [11] (see also [1]), we will present two new bounds for the k-domination
number γk . The first one is better than the bound of Theorem 1, even though it preserves the same assumptions. The second
one weakens a little more the assumption on the minimum degree δ and, for k ≥ 3, it is even better than the previous one.
In addition, we present upper bounds for the k-Roman domination number.

2. Results

Using the same method as in the proof of Theorem 1, changing only the construction of the k-dominating set, we achieve
a better bound for γk .
δ+1
Theorem 2. Let G be a graph on n vertices with minimum degree δ ≥ 1 and let k ∈ N. If ln(δ+1)
≥ 2k, then
!
n
k−1
X δi
γk (G) ≤ k ln(δ + 1) + .
δ+1 i =0
i! (δ + 1)k−1

δ+1 k ln(δ+1)
Proof. Let 2k ≤ ln(δ+1)
and define p = δ+1
. The condition on δ implies that p ≤ 1
2
. Also, it implies that δ ≥ k. This can
be shown assuming by contradiction that δ ≤ k − 1. Since the function h(x) = is monotonically increasing for x > e
x
ln x
and h(2) = h(4), we obtain that h(x) ≤ h(y) for integers y ≥ x ≥ 2, except for the case that x = 2 and y = 3. Thus, when
δ 6= 1 or k 6= 3, it follows that
δ+1 k
2k ≤ ≤ ,
ln(δ + 1) ln k
√
which leads to the contradiction k ≤ e. If δ = 1 and k = 3, evidently 2k ≤ lnδ+ 1
(δ+1)
is not fulfilled and we are done.
Now we select a set of vertices A ⊆ V (G) independently at random with P (v ∈ A) = p. Let B be the set of vertices of
V (G) − A with less than k neighbors in A. Then A ∪ B is a k-dominating set of G. We will now determine P (v ∈ B).
P (v ∈ B) = P (|N (v) ∩ A| ≤ k − 1, v 6∈ A)
k−1
X
= P (|N (v) ∩ A| = i)(1 − p)
i =0

d(v)
k−1  
pi (1 − p)d(v)−i+1
X
=
i =0
i

d(v)
k−1   i
p
(1 − p)d(v)+1
X
=
i =0
i 1 − p

d(v)
k−1  
(1 − p)d(v)+1
X
≤
i =0
i

d(v)i
k−1
(1 − p)d(v)+1 .
X
≤
i =0
i!
1636 A. Hansberg, L. Volkmann / Discrete Applied Mathematics 157 (2009) 1634–1639

Let fi (d(v)) = −p(d(v) + 1) + i ln(d(v)). Then

∂ fi i k−1
(−p(d(v) + 1) + i ln(d(v))) = −p + ≤ −p + .
∂ d(v) d(v) δ
Since δ ≥ k, it is a simple matter to check that δ+k 1 ≥ k−δ 1 . Thus, if δ ≥ 2, we obtain ln(δ + 1) δ+k 1 ≥ k−δ 1 . In the case that
δ = 1, it follows that k = 1 and the inequality ln(δ + 1) δ+k 1 ≥ k−δ 1 is trivial. Hence,
∂ fi k−1 k ln(δ + 1) k−1
(−p(d(v) + 1) + i ln(d(v))) ≤ −p + =− + ≤0
∂ d(v) δ δ+1 δ
and thus fi (d(v)) is monotonically decreasing. Together with 1 − x ≤ e−x for x ∈ [0, 1], this implies
k−1
X 1 −p(d(v)+1)+i ln(d(v))
P (v ∈ B) ≤ e
i=0
i !
k−1
X 1 −p(δ+1)+i ln(δ)
≤ e
i=0
i !
k−1
X 1 −k ln(δ+1)+i ln(δ)
= e
i=0
i !
k−1
X 1 δi
= .
i=0
i! (δ + 1)k

Hence we obtain finally

γk (G) ≤ E [|A ∪ B|] = E [|A|] + E [|B|]
k ln(δ + 1) δi
Xk−1
≤n +n
δ+1 i =0
i! (δ + 1)k
!
n
k−1
X δi
= k ln(δ + 1) + . 
δ+1 i =0
i! (δ + 1)k−1

δ+1
Corollary 1. Let G be a graph on n vertices with minimum degree δ and let k ∈ N. If ln(δ+1)
≥ 2k, then
n
γk (G) ≤ (k ln(δ + 1) + 1).
δ+1

δ i ≤ (δ + 1)k−1 , it follows that

Pk−1
Proof. Using i =0

k−1
X 1 δi 1
k−1
X
≤ δi ≤ 1
i=0
i! (δ + 1 )k− 1 (δ + 1 )k− 1
i =0

and hence Theorem 2 implies

!
n
k−1
X δi
γk (G) ≤ k ln(δ + 1) +
δ+1 i=0
i! (δ + 1)k−1
n
≤ (k ln(δ + 1) + 1). 
δ+1

For the case k = 1, we obtain directly the well-known bound for the usual domination number γ , given by Arnautov in [2]
and Payan in [10].

Corollary 2 (Arnautov [2] 1974, Payan [10] 1975). If G is a graph on n vertices with minimum degree δ ≥ 1, then
n
γ (G) ≤ (ln(δ + 1) + 1).
δ+1

In the next theorem, we use a slightly different probability p as in Theorem 2 and we obtain new results for the k-
domination and the k-Roman domination numbers.
 A. Hansberg, L. Volkmann / Discrete Applied Mathematics 157 (2009) 1634–1639 1637

δ+1+2 ln(2)
Theorem 3. Let G be a graph on n vertices with minimum degree δ ≥ k, where k ∈ N. If ln(δ+1)
≥ 2k then
!
n
k−1
X δi
γk (G) ≤ k ln(δ + 1) − ln(2) + 2 and
δ+1 i=0
i! (δ + 1)k−1
!
2n
k−1
X 1 δi
γkR (G) ≤ k ln(δ + 1) − ln(2) + .
δ+1 i =0
i! (δ + 1)k−1

δ+1+2 ln(2) k ln(δ+1)−ln(2)

Proof. Let 2k ≤ ln(δ+1) and define p = δ+1
. The condition on δ implies that p ≤ 21 . We select a set of vertices
A ⊆ V (G) independently at random with P (v ∈ A) = p. Let B be the set of vertices of V (G) − A with less than k neighbors in
A. Then A ∪ B is a k-dominating set of G and f = (V (G) − (A ∪ B), B, A) is a k-Roman domination function for G. It is easy to
see that −p + k−δ 1 ≤ 0 when k = 1. Since p ≤ 12 , the case δ = k = 2 has not to be considered. For the case that δ ≥ 3 and
k ≥ 2, note that the inequality −p + k−δ 1 ≤ 0 is equivalent to

δ+1 δ+1
 
k − ln(δ + 1) ≤ − ln(2),
δ δ
which is obviously true, since the expression on the left is negative and the one on the right positive for δ ≥ 3 and k ≥ 2.
Altogether it follows for any δ ≥ k ≥ 1 that −p + k−δ 1 ≤ 0 and hence, as in the proof of Theorem 2, we have

k−1
X 1 −p(δ+1)+i ln(δ)
P (v ∈ B) ≤ e .
i=0
i!

Thus, we obtain in this case

2 δi
k−1 k−1
X 1 −k ln(δ+1)+ln(2)+i ln(δ) X
P (v ∈ B) ≤ e = .
i=0
i! i=0
i! (δ + 1)k

This implies

γk (G) ≤ E [|A ∪ B|] = E [|A|] + E [|B|]

k ln(δ + 1) − ln(2) 2 δi
  k−1
X
≤ n +n
δ+1 i=0
i! (δ + 1)k
!
n Xk−1
δi
= k ln(δ + 1) − ln(2) + 2 and
δ+1 i=0
i! (δ + 1)k−1
γkR (G) ≤ E [f (V (G))] = E [|A|] + 2E [|B|]
k ln(δ + 1) − ln(2) 2 δi
  k−1
X
≤ 2n +n
δ+1 i =0
i! (δ + 1)k
!
2n
k
X − 1
δi
= k ln(δ + 1) − ln(2) + . 
δ+1 i =0
i! (δ + 1)k−1

Although the condition on δ in this theorem is weaker as in Theorem 2, for k ≥ 3, the bound for γk given here is even better
than the former one. This can be seen in the following way. We want to show that for k ≥ 3
k−1
X δi k−1
X δi
2 − ln (2) ≤ .
i=0
i! (δ + 1)k−1 i=0
i! (δ + 1)k−1

This is equivalent to proving that

δ
k−1 i k−1  
X X k−1
≤ ln(2)(δ + 1)k−1 = ln(2) δi
i =0
i! i=0
i

or rather that
k−1    
X k−1 1
1 − ln(2) ≤ ln(2) − δi . (1)
i =1
i i!
1638 A. Hansberg, L. Volkmann / Discrete Applied Mathematics 157 (2009) 1634–1639

Note that, considering k ≥ 3, the coefficients of the sum given in (1) are always positive. Thus, using δ ≥ 1, it is easy to see
that
k−1     k −1    
X k−1 1 X k−1 1
ln(2) − δ ≥
i
ln(2) −
i=1
i i! i =1
i i!
≥ ln(2)(k − 1) − 1
≥ 1 − ln(2)
and therefore (1) is proved. On the other side, if we compute a bound for γkR in the way like in Theorem 2, we obtain that
!
n
k−1
X δi
γkR (G) ≤ 2k ln(δ + 1) + .
δ+1 i =0
i! (δ + 1)k−1

In this case, since

k−1
X δi
≤ 1 < 2 ln(2),
i=0
i!(δ + 1)k−1

it is not difficult to see that the bound for γkR given in Theorem 3 is better than this one for any k ≥ 1.

δ+1+2 ln(2)
Corollary 3. Let G be a graph on n vertices with minimum degree δ ≥ k, where k ∈ N. If ln(δ+1)
≥ 2k, then
2n
γkR (G) ≤ (k ln(δ + 1) − ln(2) + 1).
δ+1

δ i ≤ (δ + 1)k−1 imply
Pk−1
Proof. As in the proof of Corollary 1, Theorem 3 and the fact that i=0
!
2n
k−1
X 1 δi
γkR (G) ≤ k ln(δ + 1) − ln(2) +
δ+1 i=0
i! (δ + 1)k−1
2n
≤ (k ln(δ + 1) − ln(2) + 1). 
δ+1

Setting k = 1, we obtain the following bound for the usual Roman domination number γR . Note that this should be the
same bound computed by Cockayne, Dreyer, Hedetniemi and Hedetniemi in [4] if we consider the error made in that paper,
forgetting a factor 2 in their computation. Moreover, the condition δ ≥ 1 cannot be avoided, otherwise the chosen probability
p of the above theorem, which is also the same for k = 1 as in [4], would be negative. Hence, the theorem given by the former
authors, should be as follows.

Corollary 4. Let G be a graph on n vertices with minimum degree δ ≥ 1. Then

2n
γR (G) ≤ (ln(δ + 1) − ln(2) + 1).
δ+1

The following observation shows that, for k ≥ 4, Corollaries 1 and 3 can be improved.

Observation 1. Let k ≥ 4 be an integer and G a graph of minimum degree δ ≥ k.

δ+1
(i) If 2k ≤ ln(δ+1)
, then
 
n k−1
γk (G) ≤ k ln(δ + 1) + 1 − .
δ+1 δ
δ+1+2 ln(2)
(ii) If ln(δ+1)
≥ 2k, then
 
n k−1
γk (G) ≤ k ln(δ + 1) − ln(2) + 2 − 2 and
δ+1 δ
 
2n k−1
γkR (G) ≤ k ln(δ + 1) − ln(2) + 1 − .
δ+1 δ
 A. Hansberg, L. Volkmann / Discrete Applied Mathematics 157 (2009) 1634–1639 1639

Proof. Due to Theorems 2 and 3, we only have to show that, for k ≥ 4,

k−1
X δi k−1
≤1− .
i =0
i! (δ + 1)k−1 δ
If δ ≥ k = 4, it is easy to check this property. If δ ≥ k ≥ 5, it follows by the induction hypothesis that
!
k
X δi 1 δk k−1
X δi
= +
i =0
i! (δ + 1)k δ+1 (δ + 1)k−1 k! i =0
i! (δ + 1)k−1

δk
 
1 k−1
≤ +1− .
δ + 1 (δ + 1)k−1 k! δ
Thus, it remains to show that
δk
 
1 k−1 k−1
+ 1 − ≤1− .
(δ + 1) k! δ + 1
k δ δ

Since (δ+δ 1)k ≤ δ+δ 1 , it is enough to show that

k

δ
 
1 k−1 k−1
+ 1− ≤1− ,
(δ + 1) k! δ + 1 δ δ
which is equivalent to
δ ≤ (δ − k + 1)k!.
From the fact that δ ≤ (δ − k + 1)k, which is straightforward to prove, this last inequality follows immediately and the
proof is complete. 
Note that for k = 3, instead of the term k−δ 1 of previous observation, we can set everywhere the term k−δ 2 and we obtain
for this case a better result as in Corollaries 1 and 3, too.

Acknowledgments

We would like to thank both the anonymous referees for some valuable comments on the former version of this
manuscript.

References

[1] N. Alon, J. Spencer, The Probabilistic Method, in: Wiley-Interscience Series in Discrete Mathematics and Optimization, Wiley, Chichester, 2000.
[2] V.I. Arnautov, Estimations of the external stability number of a graph by means of the minimal degree of vertices, Prikl. Mat. Programm. 11 (1974)
3–8 (in Russian).
[3] Y. Caro, R. Yuster, Dominating a family of graphs with small connected subgraphs, Combin. Probab. Comput. 9 (2000) 309–313.
[4] E.J. Cockayne, P.A. Dreyer Jr., S.M. Hedetniemi, S.T. Hedetniemi, Roman domination in graphs, Discrete Math. 278 (2004) 11–22.
[5] E.J. Cockayne, B. Gamble, B. Shepherd, An upper bound for the k-domination number of a graph, J. Graph Theory 9 (1985) 533–534.
[6] J.F. Fink, M.S. Jacobson, n-domination in graphs, in: Graph Theory with Applications to Algorithms and Computer Science, John Wiley and Sons, New
York, 1985, pp. 282–300.
[7] J.F. Fink, M.S. Jacobson, On n-domination, n-dependence and forbidden subgraphs, in: Graph Theory with Applications to Algorithms and Computer
Science, John Wiley and Sons, New York, 1985, pp. 301–311.
[8] T.W. Haynes, S.T. Hedetniemi, P.J. Slater, Fundamentals of Domination in Graphs, Marcel Dekker, New York, 1998.
[9] T.W. Haynes, S.T. Hedetniemi, P.J. Slater (Eds.), Domination in Graphs: Advanced Topics, Marcel Dekker, New York, 1998.
[10] C. Payan, Sur le nombre d’absorption d’un graphe simple, Cahiers Centre Études Recherche Opér. 17 (1975) 307–317.
[11] D. Rautenbach, L. Volkmann, New bounds on the k-domination number and the k-tuple domination number, Appl. Math. Lett. 20 (2007) 98–102.
[12] C.S. ReVelle, K.E. Rosing, Defendens imperium romanum: A classical problem in military strategy, Amer. Math. Monthly 107 (2000) 585–594.
[13] I. Steward, Defend the Roman Empire! Sci. Amer. 281 (1999) 136–139.