On The Double Roman bondage numbers of Graphs
,Β
N. Jafari Rad
N. Jafari Rad, Department of Mathematics,
Shahed University, Tehran, Iran
[email protected]
,Β
H.R. Maimani
H.R. Maimani, Mathematics Section, Department of Basic Sciences,
Shahid Rajaee Teacher Training University, P.O. Box 16785-163,
Tehran, Iran
[email protected]
,Β
M. Momeni
M. Momeni, Mathematics Section, Department of Basic Sciences,
Shahid Rajaee Teacher Training University, P.O. Box 16785-163,
Tehran, Iran
[email protected]
Β andΒ
F. Rahimi Mahid
F. Rahimi Mahid, Mathematics Section, Department of Basic Sciences,
Shahid Rajaee Teacher Training University, P.O. Box 16785-163,
Tehran, Iran
[email protected]
Abstract.
For a graph G=(V,E), a double roman dominating function (DRDF)
is a function f:VβΆ{0,1,2,3} having the
property that if f(v)=0 for some vertex v, then v has at
least two neighbors assigned 2 under f or one neighbor w
with f(w)=3, and if f(v)=1 then v has at least one neighbor
w with f(w)β₯2. The weight of a DRDF f is the sum f(V)=βuβVβf(u). The minimum weight of a DRDF on a
graph G is the double Roman domination number of G and is
denoted by Ξ³dRβ(G). The double roman bondage number of
G, denoted by bdRβ(G), is the minimum cardinality among all
edge subsets BβE(G) such that Ξ³dRβ(GβB)>Ξ³dRβ(G). In this paper we study the double roman
bondage number in graphs. We determine the double roman
bondage number in several families of graphs, and present several bounds for the double roman
bondage number. We also study the complexity issue of the double roman
bondage number and prove that the decision problem for the double roman
bondage number is NP-hard even when restricted to bipartite graphs.
Key words and phrases:
Domination, Double Roman
Domination, Bondage number, Planar graph, 3-SAT, NP-hard.
2010 Mathematics Subject Classification:
05C69
1. Introduction
Throughout this paper all graphs are finite, simple and
undirected. We denote the vertex set and the edge set of a graph
G by V=V(G) and E=E(G), respectively.
Let G=(V,E) be
a graph of order n. The open neighborhood of a vertex
vβV is the set N(v)={uβ£uvβE}, and the
closed neighborhood of v is N[v]=N(v)βͺ{v}.
The degree of a vertex v is degGβ(v)=β£N(v)β£. The
maximum (respectively, minimum) degree among
the vertices of G is denoted by Ξ(G) (respectively,
Ξ΄(G)). The open neighborhood of a set SβV is
N(S)=βͺvβSβN(v), and the closed neighborhood of S
is N[S]=N(S)βͺS=βͺvβSβN[v]. A vertex with
exactly one neighbor is called a leaf and its unique
neighbor is a support vertex. A strong support
vertex is a support vertex adjacent to at least two leaves. The
distance dGβ(x,y) (or briefly d(x,y)) between vertices x
and y of a graph G is the length of shortest path connecting
them. The girth g(G)=g of G is the length of a
shortest cycle in G, and g(G)=β when G is a forest. A
k-partite graph is a graph which its vertex set can be
partitioned into k sets V1β,V2β,β―,Vkβ
such that every edge of the graph has an end point in Viβ
and an end point in Vjβ for some 1β€iξ =jβ€k. A
complete k-partite graph is a k-partite graph that
every vertex of each partite set is adjacent to all vertices of
the other partite sets. We denote by Kn1β,n2β,β―nkββ the
complete k-partite graph where β£Viββ£=niβ for 1β€iβ€k.
In the case k=2, the k-partite and complete k-partite graph
are called bipartite and complete bipartite graphs. We denote by
Pnβ,Cnβ,Knβ and Knββ, the path, the cycle, the
complete graph and the empty graph of order n, respectively.
For a graph G and a nonempty subset SβV(G), the
vertex-induced subgraph, denoted by G[S],
is the subgraph of G with vertex-set S and edges incident to
members of S. For a subset S of vertices, we refer to GβS
as the subgraph of G induced by V(G)βS. If
S={v}, then the subgraph GβS is denoted by Gβv. For a
nonempty subset XβE(G), we denote by GβX the
spanning subgraph of G obtained by deleting the edges of X
from G. If X={e}, then we denote GβX by Gβe. A
planar graph is a graph that can be drawn on the plane
in such a way that its edges intersect only at their endpoints. A
connected graph G is called 2-connected, if for every vertex
xβV(G), the graph Gβx is connected. The join of
two graphs G and H, Gβ¨H, is the graph with vertex-set
V(Gβ¨H)=V(G)βͺV(H) and edge set E(Gβ¨H)=E(G)βͺE(H)βͺ{uv:uβV(G),vβV(H)}. The graph K1ββ¨Cnβ1β is called a wheel and is denoted by Wnβ.
A set SβV is called a dominating set if
N[S]=V. The domination number, Ξ³(G) of G,
is the minimum cardinality of a dominating set in G. A
dominating set of G of cardinality Ξ³(G) is called a
Ξ³-set of G or just a Ξ³(G)-set. For other graph
theory notation and terminology not given here we refer to
[10].
Let f:Vβ{0,1,2} be a function having the
property that for every vertex vβV with f(v)=0, there
exists a neighbor uβN(v) with f(u)=2. Such a function
is called a Roman dominating function. The weight of a
Roman dominating function is the sum f(V)=βvβVβf(v). The minimum weight of a Roman dominating function on G
is called the Roman domination number of G and is
denoted by Ξ³Rβ(G). A Roman dominating function on G of
weight Ξ³Rβ(G) is called a Ξ³Rβ-function of G
or just a Ξ³Rβ(G)-function. A Roman dominating function f
can be represented as a triple f=(V0β,V1β,V2β) (or
f=(V0fβ,V1fβ,V2fβ)), where Viβ={v:f(v)=i} for i=0,1,2.
The mathematical concept of Roman domination, defined and
discussed by Stewart [15], and ReVelle and Rosing [14],
and subsequently developed by Cockayne et al. [6]. Several
variants of Roman domination already have been defined and
studied. For a graph G=(V,E), a double roman dominating
function (DRDF) is a function f:VβΆ{0,1,2,3} having the property that if f(v)=0, then v has at
least two neighbors assigned 2 under f or one neighbor w
with f(w)=3, and if f(v)=1 then v has at least one neighbor
w with f(w)β₯2. The weight of a double Roman dominating
function f is the sum f(V)=βuβVβf(u). The
minimum weight of a DRDF is called double Roman
domination number of G and is denoted by Ξ³dRβ(G). The
concept of double roman domination is defined by Beeler, Haynes
and Hedetniem [5], and further studied in, for example
[1, 12, 17]. Beeler et al. [5] observed that in a
DRDF of minimum weight no vertex needs to be assigned the value
1. In fact for every DRDF f:VβΆ{0,1,2,3}, there is a DRDF fβ²:VβΆ{0,2,3} with
w(fβ²)β€w(f). Thus, since Ξ³dRβ(G) is the minimum
weight among all double Roman dominating functions on G,
without loss of generality, we only consider double Roman
dominating functions with no vertex assigned 1. We use the
notation f=(V0fβ,V2fβ,V3fβ) for a DRDF f:VβΆ{0,2,3}, where Vifβ={v:f(v)=i}, for
i=0,2,3.
Bauer, Harary, Nieminen and Suffel [4] introduced the
concept of bondage number in graphs. The bondage number
b(G) of a nonempty graph G is the minimum cardinality among
all sets of edges Eβ²βE(G) for which
Ξ³(GβEβ²)>Ξ³(G). This concept has been further studied
for various domination variants, see for example, [2, 3, 7, 8, 11, 13, 16].
In this paper we consider the concept of bondage number for the
double roman domination number. The double roman bondage
number of a graph G, denoted by bdRβ(G), is the minimum
cardinality among all edge subsets BβE(G) such that
Ξ³dRβ(GβB)>Ξ³dRβ(G). The organization of the
paper is as follows. In Section 2, we present some preliminary
results and determine the double Roman bondage number in some
families of graphs. In Section 3, we present various bounds for
the double Roman bondage number. In Section 4, we study
complexity issue of the double Roman bondage number, and show
that the decision problem of the double Roman bondage number is
NP-hard even when restricted to bipartite graphs. We make use of
the following.
Theorem 1.1**.**
[1]**
Let G be a connected graph of order nβ₯3. Then
Ξ³dRβ(G)=3* if and only if Ξ(G)=nβ1.*
Ξ³dRβ(G)=4* if and only if G=K2βββ¨H, where H is a graph with Ξ(H)β€β£V(H)β£β2.*
Ξ³dRβ(G)=5* if and only if Ξ(G)=nβ2 and Gξ =K2βββ¨H for any graph H of order nβ2.*
Theorem 1.2**.**
[1]**
Let n be a positive integer. Then
[TABLE]
[TABLE]
Lemma 1.3**.**
[17]**
Let 1β€n1ββ€n2ββ€β―β€nrβ be integers. Then
[TABLE]
Theorem 1.4**.**
[18]**
If G is a planar graph of girth gβ€β, then
[TABLE]
Corollary 1.5**.**
[18]**
If G is a planar graph of order n(G)β₯3. then
[TABLE]
Corollary 1.6**.**
[18]**
If G is a planar graph, then Ξ΄(G)β€5. If G is a
planar graph of girth gβ₯4, then Ξ΄(G)β€3. If G
is a planar graph of girth gβ₯6, then Ξ΄(G)β€2.
Lemma 1.7**.**
[18]**
Let v be a vertex of a planar graph G with degGβ(v)β₯3,
and let Evβ={xyβ£x,yβN(v),xyβ/E(G)}. Then there
exists a subset SβEvβ such that H=G+S is still a
planar graph and H[N(v)] is 2-connected.
2. Preliminaries and Exact values
In this section we present some preliminary results, and determine
the double Roman bondage number in several families of graphs
including paths, cycles, complete graphs and complete bipartite
graphs. We first determine the double bondage number of paths
Pnβ.
Theorem 2.1**.**
For any nβ₯1, bdRβ(Pnβ)=1.
Proof.
Let V(Pnβ)={v1β,v2ββ¦vnβ}. If n=3k, then by Theorem
1.2, Ξ³dRβ(Pnββv2βv3β)=Ξ³dRβ(P2ββͺPnβ2β)=n+2>Ξ³dRβ(Pnβ) and so bdRβ(Pnβ)=1. If
n=3k+1, then by Theorem 1.2,
Ξ³dRβ(Pnββv2βv3β)=Ξ³dRβ(P2ββͺPnβ2β)=3+nβ2+1=n+2>Ξ³dRβ(Pnβ) and so bdRβ(Pnβ)=1.
Now assume that n=3k+2. Then by Theorem 1.2,
Ξ³dRβ(Pnββv1βv2β)=Ξ³dRβ(P1ββͺPnβ1β)=2+nβ1+1=n+2>Ξ³dRβ(Pnβ) and so bdRβ(Pnβ)=1.
β
Theorem 2.2**.**
*For any nβ₯3 we have:
[TABLE]
Proof.
Let V(Cnβ)={v1β,v2ββ¦vnβ}. Clearly removing any vertex
of Cnβ leaves a Pnβ. If nβ‘2 or 4 mod 6, then
according to Theorem 1.2, Ξ³dRβ(Cnβ)=n, while
Ξ³dRβ(Pnβ)=n+1, and so bdRβ(Cnβ)=1. Next assume that
nξ β‘2 or 4 mod 6. Then Theorem 1.2 leads
to Ξ³dRβ(Cnβ)=Ξ³dRβ(Pnβ). We thus obtain that
bdRβ(Cnβ)β₯2. On the other hand because of bdRβ(Cnβ)β₯2, for each edge e1ββE(Cnβ) we have
Ξ³dRβ(Cnβ)=Ξ³dRβ(Cnββe1β)=Ξ³dRβ(Pnβ). By
Theorem 2.1, there exists an edge e2ββE(Cnβ) such that
Ξ³dRβ(Cnββ{e1β,e2β})=Ξ³dRβ(Pnββe2β)>Ξ³dRβ(Pnβ)=Ξ³dRβ(Cnβ),
thus bdRβ(Cnβ)β€2. Consequently, bdRβ(Cnβ)=2.
β
Corollary 2.3**.**
If G is a graph of order nβ₯3 with exactly kβ₯1 vertices of degree nβ1, then bdRβ(G)=βk/2β.
Proof.
Since kβ₯1, we have Ξ³dRβ(G)=3 by Theorem 1.1.
Let S={vβG,degGβ(v)=nβ1}. Then S is a clique of G.
Consider a minimum edge cover Eβ² of S. We know that
β£Eβ²β£=β2kββ. Now GβEβ² has
no vertex of degree nβ1, and so Ξ³dRβ(GβEβ²)>3=Ξ³dRβ(G). Thus we conclude that bdRβ(G)β€β2kββ.
Now suppose that bdRβ(G)=l, and let S be an edge set of
size l such that Ξ³dRβ(GβS)>Ξ³dRβ(G).
Note that S covers at most 2l vertices of G. If 2l<k,
then GβS has at least one vertex of degree nβ1, and
hence Ξ³dRβ(GβS)=3=Ξ³dRβ(G), a
contradiction. Therefore 2lβ₯k, and we conclude that
bdRβ(G)β₯β2kββ.
β
As a consequence of Corollary 2.3, we have the following.
Corollary 2.4**.**
If nβ₯3, then bdRβ(Knβ)=βn/2β, bdRβ(Wnβ)=1.
We next determine the double bondage number of complete multi partite graphs.
Lemma 2.5**.**
Let 1β€n1ββ€n2ββ€β―β€nrβ be integers. Then
[TABLE]
Proof.
Let G=Kn1β,n2β,β―,nrββ and V1β,V2β,β―,Vrβ be
partite sets of V(G), where β£Vjββ£=njβ for 1β€jβ€r.
Also suppose that XβE(G) such that Ξ³dRβ(G)<Ξ³dRβ(GβX). Let H be the subgraph of G induced by X and let
T=V(H).
If β£V1ββ£=β£V2ββ£=β―=β£Vlββ£=1 and β£Vl+1ββ£β₯2, then bdRβ(Kn1β,n2β,β―,nrββ)=β2lββ
by Corollary 2.3. Now suppose that β£V1ββ£=β£V2ββ£=β―=β£Vlββ£=2 and β£Vl+1ββ£β₯3. If Tβ©Viβ=β
for some 1β€iβ€l, then GβXβ
K2βββ¨K for some graph K such that Ξ(K)β€β£V(K)β£β2. Hence
Ξ³dRβ(G)=Ξ³dRβ(GβX)=4 by Theorem 1.1, which is
a contradiction. Hence Tβ©Viβξ =β
for each
1β€iβ€l. Then β£Tβ£β₯l. Therefore β£Xβ£β₯β2lββ and we conclude that bdRβ(G)β₯β2lββ. For 1β€iβ€l, choose viββViβ and consider Y={v1βv2β,v3βv4β,β―vlβ1βvlβ} if
l is even and Y={v1βv2β,v3βv4β,β―vlβ2βvlβ1β,v1βvlβ} if l is odd. In both cases GβYβK2βββ¨K for each graph K. Hence
Ξ³dRβ(GβY)β₯5 by Theorem 1.1 and therefore
bdRβ(G)β€β2lββ.
Now suppose that niββ₯3, for any 1β€iβ€r. In this
case Ξ³dRβ(G)=6 by Lemma 1.3. Therefore
ViββT for each i, except probably one i, since if
vtββVtββT for tβ{t1β,t2β}, then {Vt1ββ,Vt2ββ} dominate all vertices of GβX and hence
Ξ³dRβ(GβX)β€6, which is a contradiction. We consider
two cases.
Case 1. For each 1β€iβ€r, niβ=3.
Suppose that βi=1rβ1βViββT. Then β£Vrββ©Tβ£β₯2, since otherwise Ξ³dRβ(GβX)=6. First suppose that β£Vrββ©Tβ£=2. Assume that Vrβ={y,yβ²,yβ²β²}
and yβVrββT. If there exists zββi=1rβ1βViβ such that z is adjacent to yβ² and yβ²β² in GβX, then (Vβ{z,y},β
,{z,y}) is a DRDF, and hence
Ξ³dRβ(GβX)β€6, which is a contradiction. So for any zββi=1rβ1βViβ, there exists an edge zyβ² or zyβ²β² which belongs to X. On the other hand there exists
xββi=1rβ1βViβ, which is not adjacent to yβ²,yβ²β² in GβX, since otherwise (Vβ{y,yβ²,yβ²β²},{y,yβ²,yβ²β²},β
) is a DRDF for GβX, which is a contradiction. Hence
β£Xβ£β₯3(rβ1)+1. Now suppose that VrββT. Hence H is a spanning subgraph of G. If for any vertex xβV(G) we have degHβ(x)β₯2, then
[TABLE]
and so β£Xβ£β₯3r>3(rβ1)+1. Suppose that there exists a vertex xβV(G), with degHβ(x)=1. Without lose of generality, suppose that xβV1β={x,xβ²,xβ²β²} and x is
adjacent to yβVrβ. If y is adjacent to xβ² and xβ²β² in GβX, then (Vβ{x,y},β
,{x,y}) is a DRDF for GβX, which is a contradiction. Hence yxβ² or yxβ²β²
belongs to X. Also there are two edges yβ²t1β,yβ²β²t2ββX for some vertices t1β,t2β of G. If there exists a vertex zββi=2rβ1βViβ, such that z is adjacent to all vertices y,xβ²,xβ²β²
in GβX, then (Vβ{z,x},β
,{z,x}) is a DRDF for GβX, which is impossible. Hence for any zββi=2rβ1βViβ, Xβ©{zxβ²,zxβ²β²,zy}ξ =β
. Therefore
[TABLE]
Hence in this case bdRβ(G)β₯3(rβ1)+1. On the other hand for x,xβ²βV1β,yβV2β consider the set X={xz;zββi=2rβViβ,xβ²y}. Clearly GβX=K1ββͺK, and
KβK2βββ¨L for any graph L. This means that Ξ³dRβ(GβX)β₯2+5=7, and hence bdRβ(G)β€3(rβ1)+1. We conclude that bdRβ(G)=3(rβ1)+1.
Case 2. For at least one i, niββ₯4. The argument of this case is similar to case 1.
β
3. Bounds for the double Roman bondage number
In this section we present various bounds for the double Roman bondage number. We begin with the following.
Theorem 3.1**.**
*If G is a graph, and xyz a path of length 2 in G, then
(1) bdRβ(G)β€degGβ(x)+degGβ(y)+degGβ(z)β3ββ£N(x)β©N(y)β£.
If x and z are adjacent, then
(2) bdRβ(G)β€degGβ(x)+degGβ(y)+degGβ(z)β4ββ£N(x)β©N(y)β£.
Proof.
Let H be the graph obtained from G by removing the edges
incident to x,y and z with exception of yz and all edges
between y and N(x)β©N(y). In H, the vertex x is
isolated, z is leaf, y is adjacent to z, and all neighbors
of y in H, if any, lie in NGβ(x).
Let f=(V0β,V2β,V3β) be a Ξ³dRβ(H)-function. Then xβV2β and, without loss of generality, assume that zβV0ββͺV2β. If zβV0β, then yβV3β and therefore (V0ββͺ{x},V2ββ{x},V3β) is a DRDF on G of
weight less than f, and (1) as well as (2) are proved.
Now assume that zβV2β. Obviously yξ βV3β. If yβV2β, then (V0ββͺ{z},V2ββ{y,z},V3ββͺ{y}) is a DRDF on H of weight less than f, that is a
contradiction. However, if yβV0β, then ((V0ββͺ{x,z})β{y},V2ββ{x,z},V3ββͺ{y}),
is a DRDF of G of weight less than w(f), and again (1) and
(2) are proved.
β
Applying Theorem 3.1 on the path xyz such that one of the vertices x,y or z has minimum degree, we obtain the next result immediately.
Corollary 3.2**.**
If G is a connected graph of order nβ₯3, then
[TABLE]
Corollary 3.3**.**
If a graph G has a support vertex v of degree at least three such that all of its neighbors except one are leaves, then bdRβ(G)β€2.
Proof.
Let N(v)={v1β,v2β,...,vkβ} such that degGβ(vkβ)β₯2. Applying Theorem 3.1 (1) on the path v1βvv2β in the
case degGβ(v)=k=3, we obtain bdRβ(G)β€2 immediately.
Assume now that degGβ(v)=kβ₯4. Let f=(V0β,V2β,V3β) be a
Ξ³dRβ-function of Gβvv1β. It follows that v1ββV2β
and, without loss of generality, assume that vβV3β.
Therefore (V0ββͺ{v1β},V2ββ{v1β},V3β) is a
DRDF on G of weight Ξ³dRβ(Gβvv1β)β2, and thus
bdRβ(G)=1.
β
Corollary 3.4**.**
For any tree T with at least three vertices, bdRβ(T)β€2.
Proof.
If T has a support vertex v of degree at least three such
that all of its neighbors except one is leaf, then bdRβ(T)β€2 by Corollary 3.3. So assume that for any support
vertex v either degTβ(v)=2 or v has at least two neighbors
which are not leaves. Let P=v1β,v2β,...,vkβ be a longest path
of T. By the assumption, degTβ(v2β)=2. If g is a
Ξ³dRβ(Tβ{v1βv2β,v2βv3β})-function, then
g(v1β)=g(v2β)=2. Now if we let h(v1β)=0,h(v2β)=3 and h=g
for other vertices of T, then h is a DRDF on T of weight
less than w(g) and so Ξ³dRβ(Tβ{v1βv2β,v2βv3β})>Ξ³dRβ(T). Thus the proof is complete.
β
Problem: Characterize trees T with bdRβ(T)=1 or bdRβ(T)=2.
We next improve Theorem 3.1.
Theorem 3.5**.**
If G is a connected graph of order nβ₯2 and uvβE(G) then
[TABLE]
Proof.
It is not hard to see that Ξ³dRβ(GβU)>Ξ³dRβ(G),
where U=({tuβ£tβN(u)}βͺ{svβ£sβN(v)}β{rvβ£rβN(u)β©N(v)}). Note that β£Uβ£=degGβ(u)+degGβ(v)β1ββ£N(u)β©N(v)β£.
β
Corollary 3.6**.**
If G is a connected graph, then bdRβ(G)β€Ξ(G)+Ξ΄(G)β1.
Note that Corollary 3.6 improves Corollary 3.2. By Corollary 1.6 we obtain the following improvement of Corollary 3.6 if the graph is planar.
Theorem 3.7**.**
If G is a connected graph and uwv is a path of G, then
bdRβ(G)β€degGβ(u)+degGβ(v)β1.
Proof.
It is not hard to see that Ξ³dRβ(GβU)>Ξ³dRβ(G),
where U={tuβ£tβN(u)}βͺ{svβ£sβ(N(v)β{w})}.
Since β£Uβ£=degGβ(u)+degGβ(v)β1, the proof is complete.
β
Remark 3.8**.**
If G is a planar graph and gβ₯4, then Ξ΄(G)β€3. Also
bdRβ(G)β€Ξ(G)+Ξ΄(G)β1β€Ξ(G)+2. If G is a
planar graph and gβ₯6, then Ξ΄(G)β€2. Also
bdRβ(G)β€Ξ(G)+Ξ΄(G)β1β€Ξ(G)+1.
Theorem 3.9**.**
If G is a connected planar graph of order nβ₯2 without
vertices of degree five, then bdRβ(G)β€7.
Proof.
If A={vβV(G)β£degGβ(v)β€4}={v1β,v2β,β¦,vkβ},
then Corollary 1.6 and the hypothesis imply that Aξ =β
. Suppose on the contrary that bdRβ(G)β₯8. In
view of Theorem 3.7 and the assumption, we deduce that
dGβ(x,y)β₯3 for any two distinct vertices x,yβA. Using
Lemma 1.7, we define H0β=G and Hiβ=Hiβ1β+Siβ for
1β€iβ€k, where Siβ is a subset of Eviββ={xyβ£x,yβN(viβ),xyβ/E(Hiβ1β)} such that Hiβ1β+Siβ is still a
planar graph and Hiβ[N(viβ)] is 2-connected when
degGβ(viβ)β₯3.
Now let xβA and yβNGβ(x). If degGβ(x)β€2, then it
follows from the assumption and Theorem 3.7 that
degGβ(y)β₯7 and so degHkββ(y)β₯7. Assume next that
degGβ(x)=3. By the assumption and Theorem 3.5, we obtain
[TABLE]
If β£NGβ(x)β©NGβ(y)β£β₯1, then we deduce that
degHkββ(y)β₯degGβ(y)β₯7. In the remaining case
NGβ(x)β©NGβ(y)=β
, inequality chain leads to
degGβ(y)β₯6 and thus degHkββ(y)β₯8. Finally assume
that degGβ(x)=4. By the assumption and Theorem 3.5, we
obtain
[TABLE]
If β£NGβ(x)β©NGβ(y)β£β₯2, then we deduce that degHkββ(y)β₯degGβ(y)β₯7. If β£NGβ(x)β©NGβ(y)β£=1, then degGβ(y)β₯6 and so degHkββ(y)β₯7. In the remaining case NGβ(x)β©NGβ(y)=β
, we observe that degGβ(y)β₯5 and thus degHkββ(y)β₯7.
According to Lemma 1.7, the graph Hkβ is planar. However, since dGβ(x,y)β₯3 for any two distinct vertices x,yβA, we observe that HkββA is a planar graph with minimum degree at
least 6. This is a contradiction to Corollary 1.6, and the proof is complete.
β
Theorem 3.10**.**
If G is a connected planar graph of order nβ₯2, then
bdRβ(G)β€8.
Proof.
Let
[TABLE]
If A5β=β
, then Theorem 3.9 implies the desired
result. Thus we can assume A5βξ =β
. Suppose on the
contrary that bdRβ(G)β₯9. In view of Theorem 3.7 and
the assumption, if xβA3ββͺA4β and yβA3ββͺA4ββͺA5β, then dGβ(x,y)β₯3. In addition, if xβA3β, then
deg(y)β₯7 for all yβNGβ(x), by theorem 3.5.
Let I be a maximum independent subset of A5β. Then
A5ββIβͺN(I) and N(A4β)β©N(I)=β
. Now
let A4ββͺI={v1β,v2β,β¦,vkβ} and H=GβA3β.
Applying Lemma 1.7, we define H0β=H and Hiβ=Hiβ1β+Siβ
for 1β€iβ€k, where Siβ is a subset of Eviββ={xyβ£x,yβN(viβ),xyβ/E(Hiβ1β)} such that Hiβ1β+Siβ is
still a planar graph and Hiβ[N(viβ)] is 2-connected. We
proceed with the following claims.
Claim 1. If A4βξ =β
, then degHkββ(y)β₯7
for each vertex yβNGβ(A4β).
To see this, assume that A4βξ =β
. Let xβA4β and
yβNGβ(x). Then Theorem 3.5 and the assumption imply
that
[TABLE]
If β£NGβ(x)β©NGβ(y)β£β₯1, then we deduce that
degHkββ(y)β₯degGβ(y)β₯7. If NGβ(x)β©NGβ(y)=β
, then the inequality chain leads to degGβ(y)β₯6
and thus degHkββ(y)β₯8.
Claim 2. degHkββ(y)β₯7 for each vertex yβNGβ(I).
To see this, let xβI and yβNGβ(x). Then Theorem
3.5 and the assumption imply that
[TABLE]
If β£NGβ(x)β©NGβ(y)β£β₯2, then we deduce that degHkββ(y)β₯degGβ(y)β₯7. If β£NGβ(x)β©NGβ(y)β£=1, then degGβ(y)β₯6 and so degHkββ(y)β₯7. In the remaining case NGβ(x)β©NGβ(y)=β
, we observe that degGβ(y)β₯5 and thus degHkββ(y)β₯7.
Combining Claims 1 and 2, we find that Gβ=HkββA4β is a planar graph with the following properties. The minimum degree of Gβ is 5, I={vβV(Gβ)β£degGββ(v)=5} is an independent
set in Gβ and degGββ(v)β₯7 for each vertex vβNGββ(I)=NGβ(I). Let B be the bipartite graph with the bipartition I and N(I) and the edge set {uvβE(Gβ)β£uβI,vβN(I)}. Then B is a bipartite planar graph with exactly 5β£Iβ£ edges. Applying
Theorem 1.4 with girth gβ₯4, we obtain 5β£Iβ£β€2β£Iβ£+2β£N(I)β£β4(note that this bound remains valid if g=β, that means that B is a forest) and therefore β£N(I)β£β₯23ββ£Iβ£+2.
Altogether we find
[TABLE]
This is a contradiction to Corollary 1.5, and the proof is complete.
β
4. Complexity of double Roman bondage number
In this section, we study the NP-hardness of the double Roman bondage number problem. The decision problem of the double Roman bondage number problem is stated as follows.
Double Roman bondage number problem (DRBN):
Instance: A nonempty graph G and a positive integer k.
Question: Is bdRβ(G)β€k?
We will prove the NP-hardness of the double Roman bondage number problem by transforming from a known NP-complete problem, namely 3-satisfiability problem that is known to be NP-complete [9].
At first we recall some terms for the 3SAT problem. Let U be
a set of Boolean variables. If u is a variable in U, then u
and u are literals over U. A
truth assignment for U is a mapping t:Uβ{T,F}. We call u true under t if t(u)=T, otherwise
it is called false under t. The literal u is true under t
if and only if the variable u is true under t; the literal
u is true if and only if the variable u is false.
A clause over U is a set of literals over U. A
clause represents the disjunction of literals and is satisfied by
a truth assignment if and only if at least one of its members is
true under that assignment. A collection ΞΎ of clauses over
U is satisfiable if and only if there exists some
truth assignment for U that simultaneously satisfies all the
clauses in ΞΎ. Such a truth assignment is called a
satisfying truth assignment for ΞΎ. The 3-SAT is
specified as follows.
3-satisfiablity problem (3-SAT):
Instance: A collection ΞΎ={D1β,D2β,β¦,Dmβ} of clauses over a finite set
U of variables such that β£Djββ£=3 for j=1,2,β¦,m.
Question: Is there a truth assignment for U that satisfies all the clauses
in ΞΎ?
Theorem 4.1**.**
DRBN is NP-hard even for bipartite graphs.
Proof.
Let U={u1β,u2β,β¦,unβ} and ΞΎ={D1β,D2β,β¦,Dmβ}
be an arbitrary instance of 3-SAT. We construct a bipartite
graph G and a positive integer k such that ΞΎ is
satisfiable if and only if bdRβ(G)β€k. The graph G is
constructed as follows. For each i=1,2,β¦,n, corresponding
to the variable uiββU, we associate a graph Hiβ with
vertex set V(Hiβ)={uiβ,uiβ,wiβ,viβ,viβ²β,xiβ,yiβ,ziβ} and edge set
[TABLE]
For each j=1,2,β¦,m, corresponding to the clause Djβ={pjβ,qjβ,rjβ}βΞΎ, associate a single vertex cjβ and add edge set Ejβ={cjβpjβ,cjβqjβ,cjβrjβ},1β€jβ€m. Next, add a cycle
C8β=l1βl2βl3βl4βl5βl6βl7βl8βl1β, and join l2β and l4β to each vertex cjβ with 1β€jβ€m. Finally add a new vertex l9β and join it to both l1β and l5β, and set k=1.
Figure 1 shows an example of the graph G when U={u1β,u2β,u3β,u4β} and ΞΎ={D1β,D2β,D3β}, where D1β={u1β,u2ββ,u4β}, D2β={u1ββ,u2ββ,u4β}
and D3β={u2β,u3β,u4ββ}.
[TABLE]
It is easy to see that G is a bipartite graph, and the
construction can be accomplished in polynomial time. Let k=1.
We will prove that ΞΎ is satisfiable if and only if
bdRβ(G)=1. We proceed with a series of claims namely Claim 1,
Claim 2, Claim 3 and Claim 4.
Claim 1. Ξ³dRβ(G)β₯6n+8. If equality hold, then for any
Ξ³dRβ-function f on G,
f(Hiβ)=6, for any i=1,β¦,n,
β£{uiβ,uiβ}β©V3ββ£β€1, for any i=1,β¦,n,
{uiβ,uiβ}β©V2β=β
, for any i=1,β¦,n,
f(l1β)=f(l3β)=f(l5β)=f(l7β)=2 and {cjβ:j=1,2,β¦,m}βV0β.
Proof of Claim 1. Let f be a Ξ³dRβ-function of G, and iβ{1,2,...,n}.
If f(uiβ)β₯2 and f(uiβ)β₯2, then βvβN[xiβ]βf(v)β₯2 and so βvβV(Hiβ)βf(v)β₯6. If f(uiβ)=0 or f(uiβ)=0, then by considering Hiββuiββuiβ and Hiββuiβ or Hiββuiβ and Theorem 1.1, we obtain that βvβV(Hiβ)βf(v)β₯6.
If f(uiβ)=2 and f(uiβ)=0 or f(uiβ)=3 and f(uiβ)=0, then similarly we observe that βvβV(Hiβ)βf(v)β₯6. On the other hand βi=19βf(liβ)β₯8.
Consequently, Ξ³dRβ(G)β₯6n+8.
Assume that Ξ³dRβ(G)=6n+8. Then βi=19βf(liβ)=8 and f(Hiβ)=βvβV(Hiβ)βf(v)=6 for i=1,2,...,n. Suppose that there is an integer jβ{1,2...,n} such that
f(ujβ)=f(ujβ)=3. Since βvβN[xjβ]βf(v)β₯2, and βvβN[wjβ]βf(v)β₯2, we find that f(Hjβ)>6, a contradiction. Thus β£{uiβ,uiβ}β©V3ββ£β€1, for
any i=1,β¦,n. Parts (iii) and (iv) are proved similarly. Thus the proof of Claim 1 is complete. β’
Claim 2. Ξ³dRβ(G)=6n+8 if and only if ΞΎ is satisfiable.
Proof of Claim 2. Assume that Ξ³dRβ(G)=6n+8 and
let f be a Ξ³dRβ(G)-function. By Claim 1, at most one
of f(uiβ) and f(uiβ) is equal to 3 for each
i=1,2,...,n. Define a mapping t:Uβ{T,F} by
[TABLE]
We now show that t is a satisfying truth assignment for ΞΎ.
It is sufficient to show that every clause in ΞΎ is satisfied
by t. To this end, we arbitrarily choose a clause DjββΞΎ
with 1β€jβ€m. By Claim 1, f(cjβ)=f(l2β)=f(l4β)=0.
Thus there exists some i with 1β€iβ€n such that
f(uiβ)=3 or f(uiβ)=3, where cjβ is adjacent
to uiβ or uiβ. Assume that cjβ is adjacent to
uiβ, where f(uiβ)=3. Since uiβ is adjacent to cjβ in
G, the literal uiβ is in the clause Djβ by the construction
of G. Since f(uiβ)=3, it follows that t(uiβ)=T which
implies that the clause Djβ is satisfied by t. Next assume
that cjβ is adjacent to uiβ, where
f(uiβ)=3. Since uiβ is adjacent to
cjβ in G, the literal uiβ is in the clause
Djβ. Since f(uiβ)=3, it follows that t(uiβ)=F. Thus, t assigns uiβ the truth value T, that
is, t satisfies the clause Djβ. Since j is chosen
arbitrarily, thus t satisfies all the clauses in ΞΎ.
Consequently, ΞΎ is satisfiable.
Conversely, assume that ΞΎ is satisfiable, and let t:Uβ{T,F} be a satisfying truth assignment for ΞΎ.
Create a function f on V(G) as follows: if t(uiβ)=T,
then let f(uiβ)=f(vβ²)=3, and if t(uiβ)=F, then let
f(uiβ)=f(viβ)=3. Let f(l1β)=f(l3β)=f(l5β)=f(l7β)=2. Clearly, f(V(G))=6n+8. Since t is a satisfying
truth assignment for ΞΎ, at least one of literals in Djβ is
true under the assignment t, for each j=1,2,β¦,m. It
follows that the corresponding vertex cjβ in G is adjacent to
at least one vertex w with f(w)=3, since cjβ is adjacent
to each literal in Djβ by the construction of G. Thus f is
a DRDF of G, and so Ξ³dRβ(G)β€f(G)=6n+8. Since by
Claim 1, Ξ³dRβ(G)β₯6n+8, we conclude that
Ξ³dRβ(G)=6n+8. Thus the proof of Claim 2 is complete.
β’
Claim 3. Ξ³dRβ(Gβe)β€6n+9 for any eβE(G).
Proof of Claim 3. For any edge eβE(G), it is
sufficient to construct a DRDF f on Gβe with weight 4n+9.
We first assume that eβ{l1βl2β,l1βl8β,l1βl9β,l3βl4β,l6βl7β} or e=cjβl4β, cjβuiβ
or e=cjβuiβ, for some j=1,2,β¦,m, and
i=1,2,β¦,n. We define a function f by f(l2β)=f(l5β)=f(l8β)=3, f(uiβ)=f(viβ²β)=3 for each i=1,2,β¦,n, and f(v)=0 otherwise. Then f is a DRDF of
Gβe with f(Gβe)=6n+9. If eβ{l5βl4β,l5βl6β,l5βl9β,l2βl3β,l7βl8β}, then we define f by
f(l1β)=f(l4β)=f(l6β)=3, f(uiβ)=f(viβ²β)=3 for
each i=1,2,β¦,n, and f(v)=0 otherwise. Then f is
a DRDF of Gβe with f(Gβe)=6n+9. If e is not incident
with uiβ or viβ²β for each i, then we define f
f(l1β)=f(l4β)=f(l6β)=3 and f(uiβ)=f(viβ²β)=3,
and f(v)=0 otherwise. If e is not incident with
uiβ or viβ, then we define f by f(l1β)=f(l4β)=f(l6β)=3, f(uiβ)=f(viβ)=3, and f(v)=0
otherwise. If e=uiβviβ or uiβviβ²β, for
some i, then we define f by f(l1β)=f(l4β)=f(l6β)=3,
f(xiβ)=f(ziβ)=3 for each i=1,2,β¦,n, and
f(v)=0 otherwise. Then f is a DRDF of Gβe with f(Gβe)=6n+9 and thus Ξ³dRβ(Gβe)β€6n+9. β’
Claim 4.
Ξ³dRβ(G)=6n+8 if and only if bdRβ(G)=1.
Proof of Claim 4.
Assume Ξ³dRβ(G)=6n+8 and consider the edge e=l1βl2β. Suppose
Ξ³dRβ(G)=Ξ³dRβ(Gβe). Let fβ² be a
Ξ³dRβ-function of Gβe. It is clear that fβ² is also
a Ξ³dRβ-function on G. By Claim 1, we have fβ²(cjβ)=0 for each j=1,2,β¦,m
and f(l2β)=f(l4β)=f(l6β)=f(l8β)=f(l9β)=0. But then fβ²(N[l2β])=2, a contradiction. Hence,
Ξ³dRβ(G)<Ξ³dRβ(Gβe), and so bdRβ(G)=1.
Conversely, assume that bdRβ(G)=1. By Claim 1, we have
Ξ³dRβ(G)β₯6n+8. Let eβ² be an edge such that
Ξ³dRβ(G)<Ξ³dRβ(Gβeβ²). By Claim 3, we have
Ξ³dRβ(Gβeβ²)β€6n+8. Thus, 6n+8β€Ξ³dRβ(G)<Ξ³dRβ(Gβeβ²)β€6n+9, which yields
Ξ³dRβ(G)=6n+8. β’
By Claims 2 and 4, bdRβ(G)=1 if and only if there is a truth
assignment for U that satisfies all clauses in ΞΎ. Since the
construction of the double Roman bondage number instance is
straightforward from a 3-satisfiability instance, the size of
the double Roman bondage number instance is bounded above by a
polynomial function of the size of 3-satisfiability instance. It
follows that this is a polynomial reduction and the proof is
complete.
β