Notes on k-rainbow independent domination in graphs
Enqiang Zhu
Chanjuan Liu***corresponding author
Institute of Computing Science and Technology, Guangzhou University, Guangzhou 510006, China
School of Computer Science and Technology, Dalian University of Technology, Dalian 116024, China
Abstract
The k-rainbow independent domination number of a graph G, denoted γrik(G), is the cardinality of a smallest set consisting of two vertex-disjoint independent sets V1 and V2 for which every vertex in V(G)∖(V1∪V2) has neighbors in both V1 and V2. This domination invariant was proposed by Šumenjak, Rall and Tepeh in (Applied Mathematics and Computation 333(15), 2018: 353-361), which allows to reduce the problem of computing the independent domination number of the generalized prism G□Kk to an integer labeling problem on G. They proved a Nordhaus-Gaddum-type theorem: 5≤γrik(G)+γrik(G)≤n+3 for every graph G of order n≥3, where G is the complement of G. In this paper, we improve this result by showing that if G is not isomorphic to the 5-cycle, then 5≤γrik(G)+γrik(G)≤n+2. Moreover, we show that the problem of deciding whether a graph has a k-rainbow independent dominating function of a given weight is NP-complete. Our results respond some open questions proposed by Šumenjak, et al.
keywords:
Domination, k-rainbow independent domination , Nordhaus-Gaddum , NP-complete
1 Introduction
All graphs considered in this paper are simple and for notation and terminology not defined here we follow the book [1]. Let G be a graph with vertex set V(G) and edge set E(G). Two vertices are adjacent in G if they are the endpoints of an edge of G. We say that a vertex u∈V(G) is adjacent to a set U⊆G in G if U contains a vertex adjacent to u in G.
For any v∈V(G), NG(v)={u∣uv∈E(G)} is called the open neighborhood of v in G and NG[v]=NG(v)∪{v} is the closed neighborhood of v in G.
Let dG(v)=∣NG(v)∣ denote the degree of v in G and Δ(G)=max{dG(v)∣v∈V(G)}. A vertex of degree k and at least k is called a k-vertex and k+-vertex, respectively. For any S⊆V(G), let NG(S)=⋃v∈SNG(v)∖S and NG[S]=NG(S)∪S. We say that S dominates a set S′ if S′⊆NG[S]. Moreover, we use the notation G−S to denote the subgraph of G obtained by deleting vertices of S and their incident edges in G, and G[S]=G−(V(G)∖S) subgraph of G induced by S. The complete graph with n vertices and the cycle of length n are denoted by Kn and Cn, respectively.
For two integers i,j, i<j, we will make use the notation [i,j] to denote the set {i,i+1,…,j}.
Given a graph G and a subset D⊆V(G), we call D a dominating set of G if D dominates V(G). An independent set of a graph is a set of vertices no two of which are adjacent in the graph. If a dominating set D of G is an independent set, then D is called an independent dominating set (IDS for short) of G. The independent domination number of G, denoted by i(G), is the cardinality of a smallest independent dominating set of G.
Domination and independent domination in graphs have always attracted extensive attention [2, 3], and many variants of domination [2] have been introduced increasingly, for the applications in diverse fields, such as electrical networks, computational biology, land surveying, etc. Recent studies on these variations include strong roman domination [4], sum of domination number [4], semitotal domination[5, 6], relating domination [7], just to name a few.
Let G□H be the cartesian product of G and H. To reduce the problem of determining i(G□Kk) to an integer labeling problem on G itself, Šumenjak et al. [8] recently proposed a new variation of domination, called k-rainbow independent dominating function of a graph G (kRiDF for short), which is a function f: V(G)→[0,k] such that Vi is an independent set and every vertex v with f(v)=0 is adjacent to a vertex u with f(u)=i, for all i∈[1,k]. Alternatively, a kRiDF f of a graph G may be viewed as an ordered partition (V0,V1,…,Vk) such that Vj is an independent set for j=0,1,…,k and NG(x)∩Vi=∅ for every x∈V0 and each i∈[1,k], where Vj denotes the set of vertices assigned value j under f. The weight w(f) of a kRiDF f is defined as the number of nonzero vertices, i.e., w(f)=∣V(G)∣−∣V0∣.
The k-rainbow independent domination number of G, denoted by γrik(G), is the minimum weight of a kRiDF of G.
From the definition, we have γri1(G)=i(G). A γrik(G)-function is a kRiDF of G with weight γrik(G).
Let G be a graph and H a subgraph of G. Suppose that g is a kRiDF of H. We say that a kRiDF f of G is extended from g if f(v)=g(v) for every v∈V(H). In what follows, to prove that a graph G has a kRiDF, we will first find a k′RiDF g of a subgraph G′ of G, k′≤k and then extend g to a kRiDF f of G. As for the remaining part of this paper, Section 2 is dedicated to characterizing graphs G with γri2(G)=∣V(G)∣−1, based on which we in Section 3 show an improved Nordhaus-Gaddum-type theorem on the sum of 2-rainbow independent domination number of G and its complement. In Section 4 we are devoted to the proof of NP-completeness of the k-rainbow independent domination problem, and in the last section we give a conclusion of this paper.
2 Graphs G with γri2(G)=∣V(G)∣−1
To get the improved Nordhaus-Gaddum-type theorem on the sum of 2-rainbow independent domination number of G and of G, we have to characterize the graphs G such that γri2(G)=∣V(G)∣−1. For this, we need the following special graphs.
A star Sn, n≥1, is a complete bipartite graph G[X,Y] with ∣X∣=1 and ∣Y∣=n, where the vertex in X is called the center of Sn and the vertices in Y are leaves of Sn.
The graph obtained from Sn by adding a single edge is denoted Sn+.
A double star [9] is defined as the union of two vertex-disjoint stars with an edge connecting their centers. Specifically, for two integers n,m such that n≥m≥0 the double star, denoted by S(n,m), is the graph with vertex set {u0,u1,…,un,v0,v1,…,vm}
and edge set {u0v0,u0ui,v0vj∣1≤i≤n,1≤j≤m}, where u0v0 is called the bridge of S(n,m) and the subgraphs induced by {ui∣i=0,1,…,n} and {vj∣j=0,1,…,m} are called the n-star at u0 and m-star at v0. Observe that S(n,m) is defined on the premise of n≥m. For mathematical convenience, we denote a double star S(n,m) as a vertex-sequence vmvm−1…v0u0u1…un.
We start with a known result which characterizes graphs G with γri2(G)=n.
Lemma 2.1**.**
[8]**
For any graph G of order n, γri2(G)=n if and only if every connected component of G is isomorphic either to K1 or K2. In addition, if γri2(G)=n, then γri2(G)=2, where G is the complement of G.
The following conclusion is simple but will be used throughout this paper.
Lemma 2.2**.**
Let G be a graph and H a subgraph of G. Suppose that g=(V0,V1,…,Vk) is a γrik(H)-function. Then g can be extended to a kRiDF of G with weight at most ∣V(G)∣−∣V0∣.
Proof
Let V(G)∖V(H)={x1,…,xℓ}. We will deal with these vertices in the order of x1,…,xℓ by the following rule: for each xi, i∈[1,ℓ], let j∈[1,k] be the smallest one such that xi is not adjacent to Vj in G. If such j does not exist, we update V0 by V0∪{xi}; otherwise we update Vj by Vj∪{xi}. After the last one, i.e., xℓ is handled, we obtain a kRiDF of G. Obviously, the weight of the resulting kRiDF of G is at most ∣V(G)∣−∣V0∣. ∎
The following theorem clarifies the structure of connected graphs G with γri2(G)=∣V(G)∣−1.
Theorem 2.3**.**
Let G be a connected graph of order n≥3. Then, γri2(G)=n−1 if and only if G is isomorphic to one among Sn−1, Sn−1+, S(n−3,1) (n≥4) and C5.
Proof
Let f=(V0,V1,V2) be an arbitrary γri2(G)-function. Observe that V0 does not contain any 1-vertex; one can readily derive that γri2(G)=n−1 when G is isomorphic to one of Sn−1,Sn−1+, S(n−3,1) and C5. Conversely, suppose that γri2(G)=n−1, i.e. ∣V0∣=1. By Lemma 2.2, G contains no subgraph H that has a 2RiDF of weight at most ∣V(H)∣−2.
Since γri2(C4)=2=∣V(C4)∣−2, G contains no subgraph isomorphic to C4. This also shows that every two vertices of G share at most one neighbor in G.
Observation 1. If G contains a 3+-vertex x, then every 2+-vertex of G belongs to NG(x). Suppose to the contrary that G contains a 2+-vertex y such that y∈/NG(x). Let {x1,x2,x3}⊆NG(x) and {y1,y2}⊆NG(y). Observe that ∣{x1,x2,x3}∩{y1,y2}∣≤1 and ∣NG(yi)∩{x1,x2,x3}∣≤1 for i=1,2; we without loss of generality assume that y2∈/{x1,x2,x3}, y2x2∈/E(G) and y2x3∈/E(G). Let f be: f(x)=f(y)=0,f(x2)=1,f(x3)=2. Notice that either y1=xj or y1xj∈/E(G) for some j∈[2,3]; we further let f(y1)=f(xj) and f(y2)=[1,2]∖{f(y1)}. Clearly, f is a 2RiDF of G[{x,x2,x3,y,y1,y2}] of weight ∣{x,x2,x3,y,y1,y2}∣−2, a contradiction.
Observation 2. G contains at most one 3+-vertex. Suppose to the contrary that G has two distinct 3+-vertices, say x and y.
By Observation 1, xy∈E(G). Let {y,x1,x2}⊆NG(x) and {x,y1,y2}⊆NG(y). Since G contains no subgraph isomorphic to C4, ∣{x1,x2}∩{y1,y2}∣≤1 and there are no edges between {x1,x2} and {y1,y2}. We assume that x2∈/{y1,y2} and y2∈/{x1,x2}. Then, the function f: {x,x1,x2,y,y1,y2}→{0,1,2} such that f(x)=f(y)=[math], f(x2)=f(y2)=2 and f(x1)=f(y1)=1,
is a 2RiDF of G[{x,y,x1,x2,y1,y2}] of weight ∣{x,y,x1,x2,y1,y2}∣−2, a contradiction.
Observation 3. If G contains a 3+-vertex x, then NG(x) contains at most two 2-vertices; in particular, when NG(x) contains two 2-vertices, these two 2-vertices are adjacent in G. If not, suppose that NG(x) contains three 2-vertices, say x1,x2,x3. Without loss of generality, we assume that x3∈/NG({x1,x2}) and let NG(x3)={x,y3}. Let NG(x1)={x,y1} (possibly y1=x2, but y1=y3). By Observation 1, dG(y3)=1, i.e., y1y3∈/E(G). Let f be: f(x1)=f(x3)=0,f(x)=1,f(y1)=f(y3)=2. Obviously, f is a 2RiDF of G[{x,x1,y1,x3,y3}] of weight ∣{x,x1,y1,x3,y3}∣−2, a contradiction.
Now, suppose that NG(x) contains two 2-vertices, say x1,x2. If x1x2∈/E(G), let NG(xi)={x,yi},i=1,2. Clearly, y1=y2 and y1y2∈/E(G). Let f be: f(x1)=f(x2)=0,f(x)=1,f(y1)=f(y2)=2. Then, f is a 2RiDF of G[{x,x1,y1,x2,y2}] of weight ∣{x,x1,x2,y1,y2}∣−2, a contradiction.
By the above three observations and the assumption that G is connected, we see that if G contains a 3+-vertex x, then V(G)∖{x} contains either only 1-vertices (G≅Sn−1), or one 2-vertex and n−2 1-vertices (G≅S(n−3,1)), or two adjacent 2-vertices and n−3 1-vertices (G≅Sn−1+); if Δ(G)=2, then it is easy to see that G is isomorphic to one of S2+,S2,S(1,1) and C5. ∎
According to Lemma 2.1, Theorem 2.3 and the fact that γri2(G)=∑i=1kγri2(Gi), where G1,…,Gk are the components of G, we have the following theorem.
Theorem 2.4**.**
Let G be a graph of order n≥3. Then, γri2(G)=n−1 if and only if G has one component G1 isomorphic to one among Sn1−1 (n1≥3), Sn1−1+ (n1≥3), S(n1−3,1) (n1≥4) and C5, and other components are isomorphic to K1 or K2, where n1=∣V(G1)∣.
3 An improved Nordhaus-Gaddum type theorem for γri2(G)
This section is devoted to achieve an improved Nordhaus-Gaddum type theorem by showing that γri2(G)+γri2(G)≤n+2 for every graph G≅C5 of order n≥2, which improves a result obtained by Šumenjak, et al [8]. Before doing so, we need to establish some simple lemmas.
Lemma 3.5**.**
Let G be a graph of order n≥3. If G is isomorphic to Sn−1,Sn−1+ or S(n−3,1), then γri2(G)≤3.
Proof
If G≅Sn−1 or G≅Sn−1+, let V(G)={v0,v1,…,vn−1} where v0 is the center and v1v2∈E(G) when G≅Sn−1+. Define a function f such that f(v1)=1,f(v0)=f(v2)=2 and f(v)=0 for every v∈V(G)∖{v0,v1,v2}. Since every vertex in V(G)∖{v0,v1,v2} is adjacent to both v1 and v2 in G, it follows that f is a 2RiDF of G of weight 3.
If G≅S(n−3,1), then n≥4. Let G=v1v0u0u1…un−3. If n=4, then both G and G are isomorphic to P4, the path of length 3, and the conclusion holds. If n≥5, then the function f: V(G)→{0,1,2} such that f(u1)=f(u0)=1,f(u2)=2 and f(v)=0 for every v∈V(G)∖{v0,v1,v2} is a 2RiDF of G with weight 3.∎
Lemma 3.6**.**
Let G be a graph of order n. If G≅C5 and γri2(G)=4, then γri2(G)≤n−2.
Proof
Clearly, n≥4. When n=4 and n=5, the conclusion is easy to prove and we assume that n≥6.
Suppose, to the contrary, that γri2(G)≥n−1. If γri2(G)=n, then γri2(G)=2 by Lemma 2.1, a contradiction. Therefore, γri2(G)=n−1. By Theorem 2.4 G has one component isomorphic to Sn1,Sn1+, S(n2,1) or C5 where n1≥2,n2≥1, and all of the other components of G are isomorphic to K1 or K2.
If G contains two vertices u,v such that NG{u,v}=∅, then each of u and v is adjacent to all vertices of V(G)∖{u,v} in G. We can obtain a 2RiDF of G by assigning 1 to u, 2 to v, and 0 to the remained vertices of G. This indicates that γri2(G)≤2 and a contradiction. Therefore, G contains no K2 components and contains at most one K1 component, which implies that G has at most two components. If G contains only one component, then G is isomorphic to Sn−1,Sn−1+ or S(n−3,1) (since G≅C5). By Lemma 3.5 γri2(G)≤3 and a contradiction.
Therefore, G has two components, denoted by G1 and G2, where G1≅K1 and G2 is isomorphic to Sn−2,Sn−2+, S(n−4,1) or C5. Let V(G1)={u} and define a function f as follows:
let f(u)=1; f(v0)=f(v′)=2 when G2≅Sn−2 or G2≅Sn−2+ (where v0 is the center of G2 and v′ is a 1-vertex of G2. Since n≥6 such v′ does exist), f(v0)=f(u0)=2 when G2≅S(n−4,1) (where v0u0 is the bridge of G2), or f(u1)=f(u2)=2 when G2≅C5 (where C5=u1u2u3u4u5u1); and all of the other remained vertices are assigned value 0. Clearly, every vertex assigned value 0 is adjacent to u and a vertex assigned value 2. Hence, f is a 2RiDF of G with weight 3, and a contradiction. ∎
Lemma 3.7**.**
Suppose that G is a graph of order n such that γri2(G)≥4 and γri2(G)+γri2(G)=n+3. Let f=(V0,V1,V2) be an arbitrary γri2(G)-function. Then,
- (1)
If ∣V0∣≥2, then for any two vertices u,v∈V0, there are no vertices u1,u2,v1,v2 such that {u1,u2}∈NG(u), {v1,v2}∈NG(v) and uivi∈/E(G) for i=1,2, where u1=u2,v1=v2 but possibly ui=vi;
2. (2)
Suppose that u,v be two arbitrary vertices of V0. Then, ∣NG({u,v})∣≥3.
3. (3)
∣Vi∣≥2* for i=0,1,2.*
Proof
For (1), if the conclusion were false, let g be: g(ui)=g(vi)=i for i=1,2 and g(u)=g(v)=0. Then, g is a 2RiDF of G[{u,v,u1,v1,u2,v2}] with weight ∣{u,v,u1,v1,u2,v2}∣−2. Since V1 and V2 are cliques in G, Vi, for i=1,2, contains at most two vertices not assigned 0 under every 2RiDF of G. Hence, by Lemma 2.2 g can be extended to a 2RiDF of G with weight at most ∣V0∣−2+4=∣V0∣+2. This shows that γri2(G)≤∣V0∣+2 and γri2(G)+γri2(G)≤∣V1∣+∣V2∣+∣V0∣+2=n+2, a contradiction.
For (2), if ∣NG({u,v})∣≤2, let f be: f(u)=1,f(v)=2 and f(x)=0 for x∈V(G)∖NG[{u,v}]. Clearly, f is a 2RiDF of G[V(G)∖NG({u,v})] with weight 2. By Lemma 2.2, f can be extended to a 2RiDF of G with weight at most 4, since ∣NG({u,v})∣≤2. Thus, γri2(G)=4 and by Lemma 3.6 γri2(G)≤n−2, a contradiction.
For (3), if ∣V0∣=1, then γri2(G)=n−1. By an analogous argument as that in Lemma 3.6, we can derive that γri2(G)+γri2(G)≤n+2, a contradiction. In the following, we prove that ∣V1∣≥2 (the proof of ∣V2∣≥2 is similar to that of ∣V2∣≥2). Suppose that ∣V1∣=1 and let V1={u}. Then, every vertex of V0 is adjacent to u in G, i.e., u is not adjacent to V0 in G. By Lemma 3.6 we assume that ∣V1∣+∣V2∣≥5.
If V0 contains a vertex v with two neighbors v1,v2 in G, then u∈/{v1,v2}.
Let g be: g(v)=0,g(v1)=1,g(v2)=2. Since V2 is a clique in G, by Lemma 2.2 g can be extended to a 2RiDF of G with weight at most ∣V0∣−1+3=∣V0∣+2. This shows that γri2(G)≤∣V0∣+2 and hence γri2(G)+γri2(G)≤n+2, a contradiction. Therefore, every vertex in V0 has degree at most 1 in G, which implies that ∣NG({x,y})∣≤2 for any two vertices x∈V0,y∈V0 (observe that ∣V0∣≥2). This contradicts (2).
Lemma 3.8**.**
Let G be a graph of order n≥4 and u∈V(G). If H=G−u, the resulting graph obtained from G by deleting u and its incident edges, is connected and γri2(H)=∣V(H)∣−1, then G has a 2RiDF f such that f(u)=1 and f(v)=0 for some v∈V(H).
Proof
Clearly, ∣V(H)∣≥3. If u is not adjacent to V(H), then let f be: f(u)=1 and f(v)=g(v) for every v∈V(H) where g is a γri2(H)-function of H. Since γri2(H)=∣V(H)∣−1, there is a v∈V(H) satisfying f(v)=g(v)=0. If u is adjacent to a vertex u1∈V(H), then there is a vertex u2∈V(H) adjacent to u1 since H is connected. Let f be: f(u1)=0,f(u)=1,f(u2)=2. Then, by Lemma 2.2 f can be extended to a desired 2RiDF of G. ∎
Now, we turn to the proof of our main result.
Theorem 3.9**.**
Let G be a graph of order n(≥2). If G≅C5, then γri2(G)+γri2(G)≤n+2.
Proof
We may assume that n≥5 as the statement holds trivially when n=2,3,4.
Let f0=(V0,V1,V2) be a γri2(G)-function such that G[V0] contains the maximum number of components isomorphic to K2.
Suppose to the contrary that γri2(G)+γri2(G)>n+2. Then, γri2(G)+γri2(G)=n+3 since γri2(G)+γri2(G)≤n+3 [8], that is,
[TABLE]
Formula (1) indicates that every 2RiDF of G has weight at least ∣V0∣+3. We will complete our proof by constructing a 2RiDF of G of weight at most ∣V0∣+2 or a 2RiDF of G of weight less than ∣V1∣+∣V2∣.
If ∣V1∪V2∣=2, then γri2(G)+γri2(G)≤2+n, a contradiction; if ∣V1∪V2∣=3, then γri2(G)=n and by Lemma 2.1 γri2(G)=2, also a contradiction. Therefore, by Lemma 3.6,
[TABLE]
Then, by Lemma 3.7 (3) we have ∣Vi∣≥2 for i=0,1,2.
In addition, because, by definition, G[Vi]
is a clique, i=1,2, it follows that for every 2RiDF g0=(V0′,V1′,V2′) of G,
[TABLE]
Therefore, by Lemma 2.2 every γri2(G[V0])-function can be extended to a 2RiDF of G with weight at most γri2(G[V0])+4, i.e., γri2(G[V0])≥∣V0∣−1 by Formula (1).
Claim 1. Let ℓ be the number of vertices in V1∪V2 which have degree ∣V1∣+∣V2∣−1 in G[V1∪V2]. Then, ℓ≤1−ℓ′ where ℓ′=∣V0∣−γri2(G[V0]) ≤1. If not, either ℓ≥2 or ℓ=ℓ′=1.
Suppose that ℓ≥2 and let v1 and v2 be two vertices of V1∪V2 that are adjacent to all vertices of (V1∪V2)∖{u,v} in G. Let g′ be: g′(v1)=1,g′(v2)=2,g′(x)=0 for x∈V1∪V2∖{v1,v2}. Clearly, g′ is a 2RiDF of G[V1∪V2] and by Lemma 2.2 g′ can be extended to a 2RiDF of G with weight at most ∣V0∣+2, a contradiction. Now, suppose that ℓ=ℓ′=1. Then, γri2(G[V0])=∣V0∣−1, which indicates that G[V0] has a component H′ such that γri2(H′)=∣V(H′)∣−1. Since ℓ=1, there is a vertex v, say v∈V1, which is adjacent to all vertices of V2 in G. By Lemma 3.8 G[V(H′)∪{v}] has a 2RiDF g′ for which g′(v)=1 and g′(x)=0 for some x∈V(H′). Observe that every vertex in (V1∪V2)∖{v} is adjacent to v in G;
by the rule of Lemma 2.2 g′ can be extended to a 2RiDF g of G under which there is at most one vertex in V1∖{v} (and V2) not assigned value 0. Thus, w(g)≤∣V0∣−1+3=∣V0∣+2, a contradiction. This completes the proof of Claim 1. ■
In the following, without loss of generality we assume ∣V1∣≥∣V2∣. Then, ∣V1∣≥3 by Formula (2).
Claim 2. G[V0] contains no isolated vertex v such that NG(v)∩V1=∅. Otherwise, let f′ be: f′(v)=1,f′(x)=2 for x∈V2. By Claim 1, V1 has at most one vertex adjacent to all vertices of V2 in G; say v′ if such a vertex exists. We further let f(y)=0 for y∈V1∪(V0∖{v}) (or for y∈(V1∖{v′})∪(V0∖{v}) if v′ exists). Since every vertex in V1∪V0 (except for v′) is adjacent to both v and V2 in G, f is a 2RiDF of G of weight at most ∣V2∣+2, a contradiction. ■
We proceed by distinguishing two cases: γri2(G[V0])=∣V0∣−1 and γri2(G[V0])=∣V0∣.
Case 1. γri2(G[V0])=∣V0∣−1. In this case, by Claim 1 every vertex in Vi has a neighbor in Vj in G where {i,j}=[1,2]; by Theorem 2.4, G[V0] has one component H isomorphic to one of S∣V(H)∣−1 (∣V(H)∣≥3), S∣V(H)∣−1+ (∣V(H)∣≥3), S(∣V(H)∣−3,1) (∣V(H)∣≥4) and C5, and other components of G[V0] are isomorphic to K1 or K2. Let u0∈V(H) be a vertex with dH(u0)=Δ(H). Clearly, dH(u0)≥2. Let u1∈NH(u0) and u2∈NH(u0) be two vertices such that every vertex in V(H)∖{u0,u1,u2} has degree in H not exceeding min{dH(u1),dH(u2)}.
By the structure of H, for i=1,2, we have that dH(ui)≤2 and if ui has a neighbor ui′(=u0) in H, then u0ui′∈/E(H).
Moreover, by Lemma 3.7 (1), (NG(u1)∩NG(u2))∖{u0}=∅, which implies that every vertex in V1∪V2 is adjacent to u1 or u2 in G.
Claim 3. ∣V0∖V(H)∣≤1. Otherwise, let {v1,v2}⊆(V0∖V(H)). Then, dG[V0](v1)≤1 and dG[V0](v2)≤1. Suppose that dG[V0](v1)=1 (the case of dG[V0](v2)=1 can be similarly discussed). Let v1v1′∈E(G[V0]) and clearly dG[V0](v1′)=1.
By Lemma 3.7 (2), there exists a vertex v0∈(V1∪V2) that is adjacent to {v1,v1′} in G. Without loss of generality, we assume that v1v0∈E(G). By Lemma 3.8, G[V(H)∪{v0}] has a 2RiDF g′ such that g′(v0)=1 and g′(x)=0 for some x∈V(H). Further, let g′(v1)=0 and g′(v1′)=2. Then, g′ is a 2RiDF of G[V(H)∪{v0,v1,v1′}], and by Lemma 2.2 and Formula (3) g′ can be extended to a 2RiDF of G with weight at most ∣V0∣−2+4=∣V0∣+2 (since g′(v1)=g′(x)=0), a contradiction. We therefore assume that dG[V0](v1)=dG[V0](v2)=0. By Lemma 3.7 (2) we have ∣NG{v1,v2}∩(V1∪V2)∣≥3. Without loss of generality, we may assume that v1 is adjacent to two vertices of V1∪V2 in G, say v11 and v12. By Lemma 3.7 (1), ui is not adjacent to both v11 and v12, and v1j is not adjacent to both u1 and u2 in G, where i∈[1,2] and j∈[1,2]. Thus, it follows that u1v11∈/E(G) and u2v12∈/E(G), or u1v12∈/E(G) and u2v11∈/E(G), which contradicts to Lemma 3.7 (1) again. ■
By Claim 3, we see that G[V0] contains no component isomorphic to K2 and contains at most one K1 component.
Claim 4. G[V0] contains a K1 component.
If not, we have G[V0]=H.
Claim 4.1. (NG(u1)∪NG(u2))∩(V1∪V2)=∅. Otherwise, both u1 and u2 are adjacent to all vertices of V1∪V2 in G, and by Lemma 3.7 (2)
dH(ui)=2 for i=1,2 and u1u2∈/E(G). Let {ui′}=NH(ui)∖{u0},i=1,2; then, u0ui′∈/E(G). Let f be: f(u1)=f(u1′)=1, f(u2)=f(u2′)=2 and f(x)=0 for x∈V(G)∖{u1,u1′,u2,u2′}. Then, f is a 2RiDF of G with weight 4, a contradiction. ■
Claim 4.2. ∣V1∣=3. Observe that ∣V1∣≥3; it is enough to show that G has a 2RiDF f with w(f)≤∣V2∣+3.
When u1u2∈E(G), let f be: f(u0)=f(u1)=f(u2)=1,f(x)=0 for x∈(V1∪V0)∖{u0,u1,u2} and f(y)=2 for y∈V2. By Lemma 3.7 (1), V1∪V0 contains no vertex adjacent to both u1 and u2 in G. Therefore, f is a 2RiDF of G of weight ∣V2∣+3. Now, suppose that u1u2∈/E(G).
By Lemma 3.7 (1), V1 contains at most one vertex adjacent to both u0 and u1 in G; say u if such a vertex exists. Let f be: f(u0)=f(u1)=1 (or f(u0)=f(u1)=f(u)=1 if u exists), f(x)=0 for x∈(V1∪(V0∖{u0,u1})) (or x∈(V1∪V0)∖{u0,u1,u}) and f(y)=2 for y∈V2.
Notice that by Claim 1 every vertex in V0∪V1 is adjacent to V2 in G, and by the structure of H and the selection of u1 and u2, every vertex of (V0∪V1)∖{u0,u1,u} is adjacent to {u0,u1} in G; f is a 2RiDF of G of weight at most ∣V2∣+3.
■
By Claim 4.2, we have 2≤∣V2∣≤3. Let V1={w1,w2,w3} in the following.
Claim 4.3. Every vertex of Vi is adjacent to at most one vertex of Vj in G for {i,j}=[1,2]. If not, suppose that V2 contains a vertex v adjacent to two vertices of V1 in G, say w1,w2.
By Lemma 3.7 (1) v is not adjacent to u1 or u2 in G, say u1v∈/E(G).
If u2w3∈/E(G), let g′ be: g′(ui)=i for i=0,1,2, g′(w1)=g′(w2)=0,g′(w3)=2, g′(v)=1. If u2w3∈E(G), then u1w3∈/E(G) and let g′ be: g′(u1)=g′(w3)=1,g′(w1)=g′(w2)=0,g′(v)=2; further, let g′(u2)=0 when u2v∈E(G), or let g′(u2)=2 and g′(u0)=0 when u2v∈/E(G). By Lemma 2.2 in either case we can extended the g′ defined above to a 2RiDF g of G under which
g(w1)=g(w2)=0 and g(u0)=0 or g(u2)=0. Therefore, by Formula (3) w(g)≤∣V0∣−1+3=∣V0∣+2, a contradiction. With a similar argument, we can also get a contradiction if we assume V1 contains a vertex adjacent to two vertices of V2 in G. ■
Now, we consider the value of ∣V2∣. Suppose that ∣V2∣=3 and let V2={w4,w5,w6}. By Claim 4.1, we may assume, without loss of generality, that u1w1∈E(G). This indicates that u2w1∈/E(G) by Lemma 3.7 (1). If u2 is adjacent to V2, say u2w4∈E(G), then by Lemma 3.7 (1), u1w4∈/E(G), w1w4∈E(G), and u1 (resp. u2) is not adjacent to {w2,w3} (resp. {w5,w6}) in G (otherwise w4 or w1 is adjacent to two vertices of V1 or V2 in G, respectively. This contradicts to Claim 4.3). Let f be: f(u1)=f(w1)=1,f(u2)=f(w4)=2 and f(x)=0 for x∈V(G)∖{u1,u2,w1,w4}.
Observe that w1 (resp. w4) is not adjacent to {w5,w6} (resp. {w2,w3}) in G and by Lemma 3.7 (1) V0∖{u0,u1,u2} contains no vertex adjacent to both ui and wi for some i∈[1,2]. Hence, f is a 2RiDF of G[V(G)∖{u0}] of weight 4 and by Lemma 2.2 f can be extended to a 2RiDF of G with weight at most 5<∣V1∣+∣V2∣, a contradiction.
Therefore, we may assume that NG(u2)∩V2=∅. In this case, when NG(u2)∩V1=∅, let f be: f(u2)=2, f(u0)=f(u1)=1. By Lemma 3.7 (1) V1∪V2 contains at most one vertex w′ adjacent to both u0 and u1 in G and V0∖{u0} contains at most one vertex u′ adjacent to u2 in G; we further let f(x)=0 for x∈V(G)∖{u0,u1,u2,u′,w′}. Then, f is a 2RiDF of G[V(G)∖{u′,w′}] of weight 3 and by Lemma 2.2 f can be extended to a 2RiDF of G of weight at most 5<∣V1∣+∣V2∣, a contradiction. We therefore suppose that u2 is adjacent to V1 in G, say u2w2∈E(G). Then, with the same argument as NG(u2)∩V2=∅, we can show that NG(u1)∩V2=∅ as well.
Then, if w3u1∈/E(G) and w3u2∈/E(G), the function f: f(u1)=f(w1)=1,f(u2)=f(w4)=2 and f(x)=0 for x∈V(G)∖{u1,u2,w1,w4,u0} is a 2RiDF of G[V(G)∖{u0}] with weight 4, and by Lemma 2.2 f can be extended to a 2RiDF of G with weight at most 5<∣V1∣+∣V2∣, a contradiction. Therefore, we suppose that w3u1∈E(G) by the symmetry. By Lemma 3.7 (1), it has that w3u2∈/E(G), and u0w1∈/E(G) or u0w3∈/E(G), say u0w1∈/E(G) by the symmetry. Let f be: f(u0)=f(u1)=1,f(u2)=f(w2)=2 and f(x)=0 for x∈V(G)∖{u1,u2,u0,w2,w3}. Since every vertex in V(G)∖{u1,u2,u0,w2,w3} is adjacent to both {u0,u1} and {u2,w2} in G, f is a 2RiDF of G[V(G)∖{w3}] of weight 4 and by Lemma 2.2 f can be extended to a 2RiDF of G of weight at most 5<∣V1∣+∣V2∣, and a contradiction.
A similar line of thought leads to a contradiction if we assume that ∣V2∣=2 and proves Claim 4. ■
By Claim 4, we see that G[V0] contains one component isomorphic to K1. Let s be the vertex of the K1 component. We first show that ∣NG(s)∩(V1∪V2)∣≤1. If not, we assume that s is adjacent to two vertices of V1∪V2 in G, say s1,s2. By Lemma 3.7 (1)
si (resp. uj) is not adjacent to both u1 and u2 (resp. s1 and s2) in G for every i,j∈[1,2]. This implies that either siui∈/E(G) for i=1,2 or s1u2∈/E(G) and s2u1∈/E(G), which contradicts to Lemma 3.7 (1) as well. Thus, by Claim 2 ∣NG(s)∩(V1∪V2)∣=1
and the vertex s′ adjacent to s in G belongs to V1.
Let f be: f(s)=2, f(x)=1 for x∈V1, f(y)=0 for y∈V2∪V(H)). Observe that by Claim 1 every vertex in V2 is adjacent to V1 in G and hence every vertex in V2∪V(H) is adjacent to both V1 and s in G; f is a 2RiDF of G with weight ∣V1∣+1<∣V1∣+∣V2∣ (since ∣V2∣≥2), a contradiction.
The foregoing discussion shows that there exists a contradiction if we assume that γri2(G[V0])=∣V0∣−1. In the following, we consider the case of γri2(G[V0])=∣V0∣.
Case 2. γri2(G[V0])=∣V0∣. Then by Lemma 2.1 each component of G[V0] is isomorphic to K1 or K2. Recall that ∣Vi∣≥2 for i=0,1,2. Let u,v be two vertices of V0 such that uv∈E(G) if G[V0] contains a K2 component and u,v are isolated vertices in G[V0] otherwise. By Lemma 3.7 (1), we have
[TABLE]
We deal with two subcases in terms of the adjacency property of u and v.
Case 2.1. uv∈E(G). Then every vertex in V0∖{u,v} is not adjacent to {u,v} in G.
Claim 5. Every vertex in V1∪V2 has degree at most ∣V1∣+∣V2∣−2 in G[V1∪V2]. Suppose that V1 contains a vertex w adjacent to all vertices of V2 in G. If uw∈E(G) (or vw∈E(G)), then by Lemma 2.2 the 2RiDF g′ of G[{u,v,w}] such that g′(u)=0 (or g′(v)=0),g′(w)=1 and g′(v)=2 (g′(u)=2) can be extended to a 2RiDF of G, under which (V1∪V2)∖{w} contains at most two vertices not assigned 0. Thus, w(g)≤∣V0∣−1+3=∣V0∣+2, a contradiction. We therefore assume that uw∈/E(G) and vw∈/E(G). By Lemma 3.7 (2), V1∪V2 contains at least three vertices adjacent to u or v. Without loss of generality, we may suppose that there is a vertex u′∈V1∪V2 that is adjacent to u in G. Define a 2RiDF g′ of G[{u,v,u′,w}] as follows: g′(u′)=2,g′(u)=0 and g′(v)=g′(w)=1. Then, by Lemma 2.2 g′ can be extended to a 2RiDF g of G, under which (V1∪V2)∖{w,u′} contains at most one vertex not assigned value 0. Therefore, w(g)≤∣V0∣−1+3=∣V0∣+2, a contradiction. With a similar argument, we can also obtain a contradiction if we assume that V2 contains a vertex adjacent to all vertices of V1. This completes the proof of Claim 5. ■
By Claim 5, every vertex in Vi has a neighbor in Vj in G for {i,j}=[1,2]. If V1∩(NG(u)∩NG(v))=∅, then every vertex in V1 is adjacent to u or v in G. Let f be: f(u)=f(v)=1,f(x)=2 for x∈V2 and f(y)=0 for y∈V1∪(V0∖{u,v}). Clearly, f is a 2RiDF of G with weight ∣V2∣+2<∣V1∣+∣V2∣, a contradiction. We therefore assume that V1 contains a vertex s such that su∈E(G) and sv∈E(G). Then, by Lemma 3.7 (1) V2∪(V1∖{s}) contains no vertex adjacent to both u and v in G. Analogously, the function f such that f(u)=f(v)=1,f(x)=2 for x∈V1 and f(y)=0 for y∈V2∪(V0∖{u,v}) (and f(u)=f(v)=f(s)=1,f(x)=2 for x∈V2 and f(y)=0 for y∈(V1∖{s})∪(V0∖{u,v})) is a 2RiDF of G with weight ∣V1∣+2 (and ∣V2∣+3). This implies that ∣V1∣=3 and ∣V2∣=2.
Let V1={s,s1,s2} and V2={s3,s4}. Then, {u,v} contains no vertex adjacent to both s1 and s2 in G; otherwise, we, by the symmetry, suppose that us1∈E(G) and us2∈E(G). Then, the function g′(u)=1,g′(s)=2,g′(v)=g′(s1)=g′(s2)=0 is a 2RiDF of G[{u,v,s,s1,s2}] with weight 2, and by Lemma 2.2 g′ can be extended to a 2RiDF of G with weight at most ∣V0∣−1+∣V2∣+1=∣V0∣+2, a contradiction. In addition, by Lemma 3.7 (1) si,i=1,2, is not adjacent to both u and v in G.
Therefore, we may assume, by the symmetry, that s1v∈/E(G) and s2u∈/E(G).
Suppose that there are no edges between {u,v} and V2 in G. By Lemmas 3.7 (2), us1∈E(G) and vs2∈E(G). Then, the function g′ such that g′(u)=1,g′(s2)=2,g′(s)=g′(s1)=g′(v)=0 is a 2RiDF of G[{u,v,s,s1,s2}] with weight 2. By Lemma 2.2 g′ can be extended to a 2RiDF of G with weight at most ∣V2∣+1+∣V0∣−1=∣V0∣+2, a contradiction. We therefore assume that there is an edge between {u,v} and V2 in G, say vs3∈E(G) by the symmetry.
If s4s∈E(G), then the function g′ such that g′(s3)=2,g′(s4)=0,g′(s)=1,g′(v)=0 is a 2RiDF of G[{s,v,s3,s4}] with weight 2, and by Lemma 2.2 and Formula 3 g′ can be extended to a 2RiDF of G of weight at most ∣V0∣−1+3=∣V0∣+2, a contradiction. Consequently, we have s4s∈/E(G). Then, the function g′ such that g′(s3)=0,g′(s4)=g′(s)=2,g′(v)=1,g′(u)=0 is a 2RiDF of G[{s,u,v,s3,s4}] with weight 3, and by Lemma 2.2 and Formula 3 g′ can be extended to a 2RiDF of G with weight at most ∣V0∣−1+3=∣V0∣+2, a contradiction.
Case 2.2. uv∈/E(G). Then, by the selection of u,v and f0, G[V0] contains only isolated vertices and G contains no γri2(G)-function for which the subgraph of G induced by the set of vertices assigned value 0 contains K2 components.
For every x∈V0, let Uix=NG(x)∩Vi for i=1,2. Let f′ be: f′(u)=1,f′(v)=2 and f′(x)=0 for x∈((V1∪V2)∖(U1u∪U2u∪U1v∪U2v))∪(V0∖{u,v}). Obviously, f′ is a 2RiDF of G−(U1u∪U2u∪U1v∪U2v)) with weight 2.
By Lemma 2.2 f′ can be extended to a 2RiDF of G with weight at most ∣(U1u∪U2u∪U1v∪U2v))∣+2. To ensure ∣(U1u∪U2u∪U1v∪U2v))∣+2≥∣V1∣+∣V2∣, we have
[TABLE]
Claim 6. ∣(V1∪V2)∖(U1u∪U2u∪U1v∪U2v)∣=2 and the two vertices in (V1∪V2)∖(U1u∪U2u∪U1v∪U2v) are adjacent in G.
Let g′ be a 2RiDF of G[V0] such that g′(u)=g′(v)=1.
Suppose that ∣(V1∪V2)∖(U1u∪U2u∪U1v∪U2v)∣≤1.
Since V1 and V2 are cliques in G and every vertex in U1u∪U2u∪U1v∪U2v is adjacent to u or v in G, by Lemma 2.2 g′ can be extended to a 2RiDF g of G under which at most one vertex in Vi,i=1,2, is not assigned value 0 (here if (V1∪V2)∖(U1u∪U2u∪U1v∪U2v) contains a vertex, say w, then let g(w)=2). Clearly, w(g)=w(g′)+2≤∣V0∣+2, a contradiction. Moreover, if (V1∪V2)∖(U1u∪U2u∪U1v∪U2v) contains two nonadjacent vertices in G, say w1,w2, then w1 and w2 are not in the same set Vi for some i∈[1,2]. Therefore, we can extend g′ to a 2RiDF g of G by letting g′(w1)=g′(w2)=2 and g′(x)=0 for x∈(V1∪V2)∖{w1,w2}. But w(g)=w(g′)+2≤∣V0∣+2, a contradiction. ■
By Claim 6, (V1∪V2)∖(U1u∪U2u∪U1v∪U2v) contains two adjacent vertices in G, say w1,w2. If V0∖{u,v} contains a vertex z that is adjacent to w1 (or w2) in G, then let g′ be: g′(u)=g′(v)=g′(z)=1, g′(w1)=0 (or g′(w2)=0), g′(w2)=2 (or g′(w1)=2). Since every vertex (V1∪V2)∖{w2} is adjacent to {z,u,v} in G and every vertex in V′∖{w2} is adjacent to w2 where w2∈V′ for some V′∈{V1,V2}, by Lemma 2.2 g′ can be extended to a 2RiDF g of G under which every vertex in V′∖{w2} is assigned value 0 and at most one vertex in {V1,V2}∖V′ is not assigned value 0. Therefore, w(g)≤∣V0∣+2, a contradiction. This shows that every vertex in V0 is not adjacent to {w1,w2} in G.
Furthermore, if there exists a vertex z∈V0∖{u,v}, then by Claim 6 we have (V1∪V2)∖(U1u∪U2u∪U1z∪U2z)={w1,w2} and (V1∪V2)∖(U1v∪U2v∪U1z∪U2z)={w1,w2}, which implies that NG(z)=U1u∪U2u∪U1v∪U2v. Then, the function g′ such that g′(z)=1,g′(u)=g′(v)=2 and g′(x)=0 for x∈U1u∪U2u∪U1v∪U2v is a 2RiDF of G−({w1,w2}∪(V0∖{u,v,z})) with weight 3, and by Lemma 2.2 g′ can be extended to a 2RiDF of G with weight at most (∣V0∣+2−3)+3=∣V0∣+2, a contradiction. So far, we have shown that V0={u,v}, i.e., γri2(G)=n−2.
Now, let f′ be: f′(u)=1,f′(v)=2 and f′(w1)=f′(w2)=0. Clearly, f′ is a 2RiDF of G[{u,v,w1,w2}]. Then, by Lemma 2.2 f′ can be extended to a 2RiDF f of G with weight at most n−2. To ensure w(f)≥γri2(G)=n−2, we have w(f)=n−2, i.e., f is a γri2(G)-function.
We see that the subgraph of G induced by {w1w2} is isomorphic to K2. But this contradicts the selection of f0. Eventually, we complete the proof of Theorem 3.9. ∎
4 The NP-completeness
In this section, we study the NP-completeness of the k-rainbow independent domination problem. To prove a given problem P to be NP-complete, we have to show that P∈NP and find a known NP-complete problem that can be reduced to P in polynomial time. Here, by establishing an equivalence relation between the domination problem and k-rainbow independent domination problem, we can show that the k-rainbow independent domination problem is NP-complete when restricted to bipartite graphs. Three problems involved in our proof are described as follows:
The independent domination problem (IDP) [10].
Input: A graph G and a positive integer k;
Property: G has an IDS with at most k vertices.
The domination problem (DP) [11].
Input: A graph G and a positive integer k;
Property: G has an dominating set with at most k vertices.
The k-rainbow independent domination problem (kRiDP).
Input: A graph G and two positive integers k and k′;
Property: G has a kRiDF with weight at most k′.
The operation of identifying two vertices x and y of a graph G is to replace these vertices
by a single vertex incident to all the edges which were incident in G to either x or
y.
Theorem 4.10**.**
The k-rainbow independent domination problem is NP-complete for bipartite graphs.
Proof
The kRiDP is a member of NP, since we can check in polynomial time that a function from vertex set to {0,1,…,k} has weight at most k′ and is a kRiDF.
When k=1, the kRiDP is equivalent to the IDP which is NP-complete when G is restricted to bipartite graphs [3]. Therefore, we assume that k≥2. To show NP-hardness, we give a reduction from the domination problem (DP) for bipartite graphs, which is NP-complete [12]. Given a bipartite G with a bipartition (X,Y) where X={x1,x2,…,xm} and Y={y1,y2,…,yn}, we construct a new graph G′ by adding m+n copies of star Sk−1, denoted by Sk−1(xi) and Sk−1(yj) for i∈[1,m] and j∈[1,n], and identifying w and the center of Sk−1(w) for all w∈{xi,yj∣i=1,…,m,j=1,…,n} (see Figure 1, in which we omit the edges between X and Y). Clearly, G′ is also a bipartite graph. We claim that G′ has a kRiDF with weight (k−1)(m+n)+ℓ if and only if G has a dominating set of size ℓ.
Given a kRiDF f=(V0,V1,…,Vk) of G′ with weight (k−1)(m+n)+ℓ, let D=(D′=V1∪…∪Vk)∩(X∪Y). Observe that all leaves of Sk−1(xi) and Sk−1(yj) for i∈[1,m] and j∈[1,n] belong to D′; therefore, ∣D∣=ℓ. Since f is a kRiDF, it follows that every vertex in V0 is adjacent to at least one vertex in D. Notice that X∪Y=V0∪D; we see that D is a dominating set of G.
Now, we assume that G has a dominating set D where ∣D∣=ℓ. Let D1=D∩X and D2=D∩Y. We define a function f: V(G′)→[0,k] as follows: f(v)=1 for every v∈D1, f(v)=2 for every v∈D2 and f(v)=0 for every v∈(X∪Y)∖D. Since G is bipartite and D is a dominating set of G, every vertex v∈(X∪Y)∖D is adjacent to either D1 or D2 in G.
If v is adjacent to Di for some i∈[1,2] in G, then we assign [1,k]∖{i} to the k−1 leaves of Sk−1(v) such that every leaf receives an unique number of [1,k]∖{i}.
Finally, for every u∈Di for i=1,2, we assign [1,k]∖{i} to the k−1 leaves of Sk−1(u) such that every leaf receives an unique number of [1,k]∖{i}.
Clearly, w(f)=∣D∣+(m+n)(k−1), Vi is an independent set and every vertex in V0 is adjacent to a vertex in Vi for all i∈[1,k]. Therefore, f is a kRiDF with weight ∣ℓ∣+(m+n)(k−1). ∎
5 Conclusion
In this paper, we respond some questions proposed by Šumenjak et al. [8], by proving an improved Nordhaus-Gaddum type inequality on k-rainbow independent domination number and showing that the problem of deciding whether a graph has a k-rainbow independent dominating function of a given weight is NP-complete. In the study, we proved that when G satisfies γri2(G)=∣V(G)∣−1 and G≅C5, it follows that G is isomorphic to Sn(n≥2), Sn+(n≥2) or S(n,1)(n≥1), and γri2(G)+γri2(G)=∣V(G)∣+2. Additionally, we observe that γri2(S(n,m)+γri2(S(n,m))=∣V(S(n,m))∣+1 when m≥2. Therefore, a question that arises is whether Sn(n≥2), Sn+(n≥2) and S(n,1)(n≥1) are enough for determining graphs G with the property of γri2(G)+γri2(G)=∣V(G)∣+2. We formulate this more generally as follows:
Question 5.1**.**
How to characterize graphs G with γri2(G)+γri2(G)=∣V(G)∣+2?
Acknowledgments
This work was supported by the National Natural Science Foundation of China (61872101, 61672051, 61309015, 61702075), the China Postdoctoral Science Foundation under grant (2017M611223).
References
J. A. Bondy, U. S. R. Murty, Graph theory, Springer, New York, 2008.
T. W. Haynes, S. T. Hedetniemi, P. J. Slater, Fundamentals of Domination in
Graphs, Marcel Dekker, New York, 1998.
W. Goddarda, M. A. Henning, Independent domination in graphs: A survey and
recent results 313 (2013) 839–854.
T. P. Àvarez-Ruiz, T. Mediavilla-Gradolph, S. M. Sheikholeslami, J. C.
Valenzuela-Tripodoro, I. G. Yero, On the strong roman domination number of
graphs, Discrete Applied Mathematics 231 (2017) 54–59.
W. Goddard, M. A. Henning, C. A. McPillan, Semitotal domination in graphs 94
(2014) 67–81.
E. Zhu, C. Liu, On the semitotal domination number of line graphs (2018)
https://doi.org/10.1016/j.dam.2018.06.010.
M. A. Henning, S. Jäger, D. Rautenbach, Relating domination, exponential
domination, and porous exponential domination, Discrete Optimization 23
(2017) 81–92.
T. K. Šumenjak, D. F. Rall, A. Tepeh, On k-rainbow independent
domination in graphs, Applied Mathematics and Computation 333 (2018)
353–361.
J. W. Grossman, F. Harary, M. Klawe, Generalized ramsey theory for graphs,
x:double stars, Discrete Mathematics 28 (1979) 247–254.
D. F. Manlove, On the algorithmic complexity of twelve covering and
independence parameters of graphs, Discrete Applied Mathematics 91 (1999)
155–175.
K. S. Booth, J. H. Johnson, Dominating sets in chordal graphs 11 (1) (1982)
191–199.
G. J. Chang, G. L. Nemhauser, The k-domination and k-stability problems on
sun-free chordal graphs, SIAM journal of Algebraic Discrete Methods 5 (3)
(1984) 332–345.