980
Views
1
CrossRef citations to date
0
Altmetric
Articles

Total mixed domination in graphs

, &
Pages 229-237 | Received 28 Oct 2021, Accepted 02 Aug 2022, Published online: 29 Aug 2022

Abstract

For a graph G=(V,E), we call a subset SVE a total mixed dominating set of G if each element of VE is either adjacent or incident to an element of S, and the total mixed domination number of G is the minimum cardinality of a total mixed dominating set of G. In this paper, we initiate to study the total mixed domination number of a connected graph by giving some tight bounds in terms of some parameters such as order and total domination numbers of the graph and its line graph. Then we discuss on the relation between total mixed domination number of a graph and its diameter. Study of this number in trees and 2-corona graphs is our next work. Also we show that the total mixed domination number of a graph is equal to the total domination number of its total graph. Giving the total mixed domination numbers of the paths, cycles, complete bipartite graphs, complete graphs and wheels is our last work.

MSC (2010):

1. Introduction

All graphs considered here are non-empty, finite, undirected and simple. For standard graph theory terminology not given here we refer to [Citation5]. Let G=(V,E) be a graph with the vertex set V of order n and the edge set E of size m. NG(v) and NG[v] denote the open neighborhood and the closed neighborhood of a vertex v, respectively, while δ=δ(G) and Δ=Δ(G) denote the minimum and maximum degrees of G, respectively. Also we define NE(v)={eE | ve} for any vertex v, NG(e)={vV | ve} and NE(e)={eE | e and e are adjacent} for any edge e, and NT(G)(x)=NG(x)NE(x) for any element xV(G)E(G) (T(G) denotes the total graph of G which will define after). For any two vertices u and v in a connected graph G the distance between u and v is the minimum length of a shortest (u, v)-path in G and is denoted by d(u, v). The maximum distance among all pairs of vertices of G is the diameter of G, which is denoted by diam(G). A Hamiltonian path in a graph G is a path which contains every vertex of G.

We write Kn, Cn and Pn for a complete graph, a cycle and a path of order n, respectively, while G[S], Wn and Kn1,n2,,np denote the subgraph of G induced by a subset SV(G)E(G) of G, a wheel of order n + 1, and a complete p-partite graph, respectively. The complement of a graph G, denoted by G¯, is a graph with the vertex set V(G) and for every two vertices v and w, vwE(G¯) if and only if vwE(G). The line graph L(G) of G is a graph with the vertex set E(G) and two vertices of L(G) are adjacent when they are incident in G.

Domination in graphs is now well studied in graph theory and the literature on this subject has been surveyed and detailed in the two books by Haynes, Hedetniemi, and Slater [Citation2, Citation3]. A famous type of them is total domination. The literature on the subject Total domination in graphs has been surveyed and detailed in the recent book [Citation4] by Henning and Yeo.

Definition 1.1.

A subset SV of a graph G is called a total dominating set, briefly TDS, of G if each vertex of V is adjacent to a vertex in S, and the minimum cardinality of a total dominating set is called the total domination number γt(G) of G.

Y. Zhao, L. Kang, and M. Y. Sohn in [Citation6] introduced another domination number as follows.

Definition 1.2.

[Citation6] A subset SVE of a graph G is called a mixed dominating set, briefly MDS, of G if each element of (VE)S is either adjacent or incident to an element of S, and the mixed domination number γm(G) of G is the minimum cardinality of a mixed dominating set.

Here, we define the concepts of total mixed dominating set and total mixed domination number of a graph.

Definition 1.3.

A subset SVE of a graph G with δ(G)1 is a total mixed dominating set, briefly TMDS, of G if each element of VE is either adjacent or incident to an element of S, and the total mixed domination number γtm(G) of G is the minimum cardinality of a total mixed dominating set.

The goal of this paper is to initiate studying of total mixed domination number of a graph as follows. In Sec. 2, we give some tight lower and upper bounds for the total mixed domination number of a connected graph in terms of some parameters such as the order of the graph or the total domination numbers of the graph and its line graph. Also we discuss on the relation between the total mixed domination number of a graph with its diameter. In Sec. 3, study of total mixed domination number of trees and 2-corona graphs is our first work. Then, we show that the total mixed domination number of a graph is equal to the total domination number of its total graph. In Sec. 4, we will calculate the total mixed domination number of the paths, cycles, complete bipartite graphs, complete graphs and wheels. Finally, in the last section we present some open problems that can be as some motivations for next works on this topic.

Here, we fix a notation for the vertex set, the edge set and open neighborhood of a graph which are used thorough this paper. For a graph G with the vertex set V={vi|1in},E(G) or simply E denotes the edge set of G in which an edge vivj is denoted by eij. Then V(L(G))=E, and the edge set of L(G) is the set {eijeik | eij,eikE}. A min-TDS/min-TMDS of G denotes a TDS/TMDS of G with minimum cardinality. Also we agree that a vertex v dominates an edge e or an edge e dominates a vertex v mean ve. Similarly, we agree that an edge dominates another edge means they have a common vertex.

2. Some bounds

Here, we give some tight bounds for the total mixed domination number of a connected graph in terms of some parameters such as order of the graph or the total domination numbers of the graph and its line graph. Also we discuss on the relation between the total mixed domination number of a graph and its diameter. First an observation.

Observation 2.1.

Let G be a graph with the vertex set V={vi | 1in} and δ(G)1.

  • A subset SE is a TDS of L(G) if and only if {i,j | eijE}{i,j | eijS}.

  • A TDS S of G is a TMDS of G if and only if S¯ is independent in G.

Theorem 2.2.

Let G be a connected graph with δ(G)1. Then max{γt(G),γt(L(G))}γtm(G)γt(L(G))+γt(G), and the bounds are tight.

Proof.

Let G be a connected graph with the vertex set V={v1,,vn} and edge set E and let eij denote the edge vivj. Then V(L(G))=E and E(L(G))={eijeij | {i,j}{i,j}}. Since the union of a TDS of G and a TDS of L(G) is a TMDS of G, we have γtm(G)γt(L(G))+γt(G). To prove the lower bound, let S be a min-TMDS of G. If either every vertex of V is dominated by a vertex in SV, or every edge of E is dominated by an edge in SE, then SE or SV is a TDS of G or L(G), respectively, and there is nothing to prove. Otherwise, SG=(SE){vi | vi is adjacent to some vjS such that NT(G)(vj)SE}{vi,vj | eij,ejkS but vj,vkS} is a TDS of G with cardinality at most |S|. Since also by changing the roles of G and L(G) we may obtain a TDS SL of L(G) with cardinality at most |S|, we have proved max{γt(G),γt(L(G))}max{|SG|,|SL|}|S|=γtm(G).

The lower bound is tight for the complete graphs K3n by Propositions 4.6 and 4.7 when max{γt(G),γt(L(G))}=γt(L(G)). The case max{γt(G),γt(L(G))}=γt(G)=γtm(G) is discussed in Corollary 2.3. To show that the upper bound is tight, consider the graph G illustrated in with γt(G)=2 (because {v1,v5} is a min-TDS) and γt(L(G))=4 by Observation 2.1 (because {e12,e23,e56,e67} is a TDS of L(G) and for any set {eij,ejk,ek} we have {i,j,k,}{0,1,9}). Hence it is sufficient to prove γtm(G)=6. First since {v1,v5,e56,e67,e12,e23} is a TMDS of G, we have γtm(G)6. Let G1 and G2 be the subgraphs of G induced by {vi | 0i4} and {vi | 5i9}, respectively, which are isomorphic together (see ). And let also S be a TMDS of G such that |S(V(G2)E(G2))||S(V(G1)E(G1))|2. By the contrary, let |S(V(G1)E(G1))|=2. Since NT(G)(v0)S{v1,e01}, we have S(V(G1)E(G1))={e01,w} for some w{v1,v0,e1j | 2j4} or S(V(G1)E(G1))={v1,w} for some w{vj,e1j | 2j4}{v0,e01}. Hence S(V(G1)E(G1)) is one the sets {e01,vj} for some j = 0, 1, or {e01,e1j} for some j = 2, 3, 4, or {v1,vj} for some j=0,2,3,4, or {v1,e1j} for some j=0,2,3,4. Since in each case NT(G)(ek)S= for some k,{2,3,4}{j}, we conclude |S(V(G1)E(G1))|3 and so γtm(G)=6.

Figure 1. The illustration of the graphs G (left) and L(G) (right) discussed in Theorem 2.2.

Figure 1. The illustration of the graphs G (left) and L(G) (right) discussed in Theorem 2.2.

Figure 2. The set {v1,v5,e56,e67,e12,e23} is a min-TMDS of the graph G.

Figure 2. The set {v1,v5,e56,e67,e12,e23} is a min-TMDS of the graph G.

Figure 7. The total graph T(P12) with a min-TDS {v2,v3,e56,e67,v9,v10,v11}, and the connected components G1,G2,G3 that are arranged from left to right where V(G1)={v2,v3},V(G2)={e56,e67} and V(G3)={v9,v10,v11}.

Figure 7. The total graph T(P12) with a min-TDS {v2,v3,e56,e67,v9,v10,v11}, and the connected components G1, G2, G3 that are arranged from left to right where V(G1)={v2,v3}, V(G2)={e56,e67} and V(G3)={v9,v10,v11}.

Corollary 2.3.

For any graph G which has a min-TDS such that its complement is an independent set, γt(G)=γtm(G).

For an example, γt(G)=γtm(G) where G is the complete bipartite graph K1,n or its derivation S1,n,n (we recall that the derivation of a graph is the graph which is obtained by replacing every its edge by a path of length 2). In , the set of yellow vertices {v0,v1,v2,v3} is both a min-TDS and a min-TMDS of S1,3,3.

Figure 3. The set {v0,v1,v2,v3} is both a min-TDS and a min-TMDS of S1,3,3.

Figure 3. The set {v0,v1,v2,v3} is both a min-TDS and a min-TMDS of S1,3,3.

The next theorem improves the upper bound given in Theorem 2.2 when either the line graph L(G) has a min-TDS D such that there exist less than γt(G) disjoint maximal cliques in L(G)D, or G has a min-TDS S such that the minimum size of vertex cover of GS is less than γt(L(G)) (recall that a vertex cover of G is a subset S of V(G) such that each edge of G has a vertex in S and the β(G) denotes the minimum size of a vertex cover of G).

Theorem 2.4.

For any connected graph G with δ(G)1, let cD be the minimum number of disjoint maximal cliques in L(G)D where D is a min-TDS of L(G), and let β(GS) be the minimum size of a vertex cover of GS where S is a min-TDS of G. Then, by the assumptions cL(G) = min{cD | D is a min-TDS of L(G)} and βG = min{β(G\S) | S is a min-TDS of G}, γtm(G)min{γt(L(G))+cL(G),γt(G)+βG}.

Proof.

We show that the total mixed domination number of G is at most the minimum of the given set, when G=(V,E) is a connected graph with δ(G)1 and V={v1,,vn}, and eij denotes the edge vivj. So E(L(G))={eijeik | eij,eikE}. First we prove γtm(G)γt(L(G))+cL(G). Let D be a min-TDS of L(G) such that cL(G)=cD. Obviously D dominates all elements of E(L(G)){vi | eijD for some j}. Let C be a subset of E(L(G)) with cardinality cL(G) such that every maximal clique of L(G)D has exactly one vertex in C, and let vi be a vertex that does not dominated by D. Since NL(G)(vi)={eij|vivjE(G)} is a maximal clique in L(G)D, we are sure that vi is dominated by the unique vertex which is in CNL(G)(vi). Thus DC is a TMDS of G, and so γtm(G)|DC|=γt(L(G))+cL(G). In a similar way, the inequality γtm(G)γt(G)+βG can be proved and this completes our proof. □

As we show in below, our motivation to state Theorem 2.4 is existence of graphs that the upper bound in Theorem 2.4 is better for them than the upper bound in Theorem 2.2. Let Wn be a wheel of order n+14 with the vertex set V={vi | 0in} and the edge set E={e0i,ei(i+1) | for 1in}. Then, since S={v0,v1} is a min-TDS of Wn,γt(Wn)=2. On the other hand, WnSPn1 implies βWn=β(Pn1)=n12. Hence γt(Wn)+βWn=2+n12=n2+1. Since γt(L(Wn))=n2 by Lemma 2.5, we have min{γt(L(Wn))+cL(Wn),γt(Wn)+βWn}=min{n2+cL(Wn),n2+1}=n2+1=γtm(Wn)   (by Proposition 4.8)<γt(L(Wn))+γt(Wn).

Lemma 2.5.

For any wheel Wn of order n+14,γt(L(Wn))=n2.

Proof.

Let Wn be a wheel of order n+14 with the vertex set V={vi | 0in} and the edge set E={e0i,ei(i+1) | for 1in} (note: n + 1 is considered 1 to modulo n). Let S be a TDS of L(Wn) where V(L(Wn))=E. Then {i,i+1}{1,2,,n} for each 1in because of NWn(ei(i+1))S={vi,vi+1}S. Hence |S|n2. Now since {e0(2i1) | 1in2} is a TDS of L(Wn), we have γt(L(Wn))=n2.

Lemma 2.5 shows the upper bound in Theorem 2.4 is tight for any wheel. We know for any graph G with a non-empty edge set, γtm(G)2, and Corollary 2.3 characterizes graphs G satisfy γtm(G)=2. The next theorem gives a sufficient condition for that the total mixed domination number of a graph be at least 3.

Theorem 2.6.

For any connected graph G of order at least 2,γtm(G)=2 impliesdiam (G)3.

Proof.

Let G=(V,E) be a connected graph with diam(G)4 and δ(G)1 in which V={v1,,vn}. The condition diam(G)4 implies that G has a path P of length at least 4 as an induced subgraph of G. Let S be a min-TMDS of G of cardinality 2. If S is {vi,vj} or {vi,eij} for some i, j, then NT(G)(epq)S= for some vp,vqV(P), a contradiction. Also if S={eij,ejk} for some i, j, k, then NT(G)(v)S= for some vV(P) such that i,j,k, a contradiction. So γtm(G)2.

Now, we present another upper bound for γtm(G) in term of the order of the graph which is tight by Proposition 4.6.

Theorem 2.7.

For any connected graph G of order n2 which has a Hamiltonian path, γtm(G)5n3n.

Proof.

Let P:v1v2vn be a Hamiltonian path in a connected graph G of order n2. Since each of the sets S0={e(3i+1)(3i+2)e(3i+2)(3i+3) | 0in31}if n0(mod 3),S1=S0{e(n1)n}if n1(mod 3),S2=S0{e(n2)(n1),e(n1)n}if n2(mod 3), is a TMDS of G, the result holds. □

3. Trees, 2-corona graphs and total graphs

The facts that diam(C4)=2 and γtm(C4)=3 show that the converse of Theorem 2.6 is not true in general. But next theorem shows that it holds for trees.

Theorem 3.1.

For any tree T of order at least 2,γtm(T)=2 if and only if diam(T)3.

Proof.

By Theorem 2.6, it is sufficient to prove that diam(T)3 implies γtm(T)=2. If diam(T)=1, then TK2 and so γtm(T)=2. If diam(T)=2, then T is isomorphic to the complete bipartite graph K1,n1 and so γtm(T)=2 by Proposition 4.4. Now let diam(T)=3. Then T is a double star tree which is obtained by joining the central vertex v of a tree K1,p and the central vertex w of a tree K1,q where p+q=n2. Since {v, w} is a TMDS of T, we have γtm(T)=2.

Next theorem improves the upper bound given in Theorem 2.7 for trees.

Theorem 3.2.

For any tree T of order n3,γtm(T)2n3.

Proof.

Let T=(V,E) be a tree in which V={vi | 1in}. Choose a leaf v of T and label each vertex of T with its distance from v to modulo 3. This partitions V to the three independent sets A0,A1 and A2 where Ai={uV | dT(u,v)i(mod 3)} for 0i2. Then by the pigeonhole principle at least one of them, say A0, contains at least one third of the vertices of T, and so |A1A2|2n3. We see that every non-leaf and every leaf viV(T)A1A2 is adjacent to some vertex in A1A2. If needed, we replace every leaf viA1A2 by an its neighbour out of A1A2. The obtained set S by this way is a TMDS of T. Because obviously NT(vi)S for each viV(T), and {vi,vj}S for each eijE (because dT(v,vi)dT(v,vj)(mod 3)), and so every eijE is dominated by viS or vjS. Therefore γtm(T)|S|2n3.

In the next theorem, we show that for every graph G of order n the total mixed domination number of G°P2 is 2n, where G°P2 is the 2-corona of G which is obtained from G by adding a path of length 2 to each vertex of G. shows the graph P6°P2.

Figure 4. The set {vi | 1i12} is a min-TMDS of P6°P2.

Figure 4. The set {vi | 1≤i≤12} is a min-TMDS of P6°P2.

Theorem 3.3.

For any connected graph G of order n2,γtm(G°P2)=2n.

Proof.

Let G=(V,E) be a connected graph in which V={vi | 1in}. Then V(G°P2)={vi | 1i3n} andE(G°P2)=E{ei(n+i),e(n+i)(2n+i) | 1in}. Since {vi,vn+i | 1in} is a TMDS of G°P2, we have γtm(G°P2)2n.

Now let S be a min-TMDS of G°P2. Then {vn+i,e(n+i)(2n+i)}S contains an element wi (because NT(G°P2)(v2n+i)S) for each 1in. Since also every wi must be dominated by an element wiNT(G°P2)(wi)S, and all of the elements wi and wi are distinct, we conclude that S includes the set {wi,wi | 1in} of cardinality 2n, and so γtm(G°P2)2n, which completes our proof. For an example, shows P6°P2 with the min-TMDS {vi | 1i12} of it. □

By Theorem 3.3 the upper bound in Theorem 3.2 is tight for any 2-corona T°P2 in which T is a tree of order n3.

The total graph T(G) of a graph G=(V,E) is the graph whose vertex set is VE and two vertices are adjacent whenever they are either adjacent or incident in G [Citation1]. It is obvious that if G has order n and size m, then T(G) has order n + m and size 3m+|E(L(G))|, and also T(G) contains both G and L(G) as two induced subgraphs and it is the largest graph formed by adjacent and incidence relation between graph elements. See for an example.

Figure 5. The total graph of the graph in with the min-TDS {v1,v5,e12,e23,e56,e67}.

Figure 5. The total graph of the graph in Figure 1 with the min-TDS {v1,v5,e12,e23,e56,e67}.

It is clear that a total mixed dominating set of a graph G corresponds with a total dominating set of the total graph T(G) of G. Hence we have the next theorem, and so to find the total mixed domination number of a graph we may calculate the total domination number of the total graph of the graph.

Theorem 3.4.

For any graph G with δ(G)1,γtm(G)=γt(T(G)).

As we will see in Proposition 4.1, for an example of Theorem 3.4, while shows a path P12 with a min-TMDS, shows the total graph of the path with the same set as a min-TDS of it.

4. Some classes of graphs

In this section, we present formulas for the total mixed domination number of some known classes of graphs. The first two propositions are devoted to the paths and cycles.

Proposition 4.1.

For any path Pn of order n2, γtm(Pn)={4n73if n1(mod 7),4n72if n2,3,4(mod 7),4n71if n5(mod 7),4n7if n0,6(mod 7).

Proof.

By Theorem 3.4, we calculate the total domination number of T(Pn) when Pn=(V,E) is a path of order n2 in which V={v1,v2,,vn} and E={ei(i+1) | 1in1}. Then V(T(Pn))=VE and E(T(Pn))=EE(L(Pn)){ei(i+1)vi,ei(i+1)vi+1 | 1in1} in which E(L(Pn))={ei(i+1)e(i+1)(i+2) | 1in2}.

Claim: There exists a min-TDS S of T(Pn) with the following three properties.

P.1. V(Gi)V if and only if V(Gi+1)E for each i,

P.2. |V(Gi)|=2 for each i, perhaps except for i = w,

P.3. V(G1)V,

in which G1,,Gw are all connected components of the induced subgraph T(Pn)[S] that are arranged from left to right in T(Pn) (see for an example).

Figure 6. A path P12 with a min-TMDS {v2,v3,e56,e67,v9,v10,v11}.

Figure 6. A path P12 with a min-TMDS {v2,v3,e56,e67,v9,v10,v11}.

By proving the claim, each of the sets S0={v7i+2,v7i+3,e(7i+5)(7i+6),e(7i+6)(7i+7) | 0in71}if n0(mod 7),S=S0{e(n1)n}if n1(mod 7),S=S0{vn1,vn}if n2,3(mod 7),S=S0{vn2,vn1}if n4(mod 7)S=S0{vn3,vn2,vn1}if n5(mod 7),S=S0{vn4,vn3,e(n2)(n1),e(n1)n}if n6(mod 7) will be a min-TDS of T(Pn), and this completes our proof.

Proof

of the claim: Let S be a min-TDS of T(Pn). We may assume for every eE(T(Pn)[S]),e=vivi+1 or e=ei(i+1)e(i+1)(i+2) for some i. Because otherwise if e=vie(i1)i or e=viei(i+1) for some i, then we can replace S by (S{vi}){ei(i+1)} or (S{ei(i+1)}){vi+1}, respectively, that each of them is again a min-TDS of T(Pn). So we may assume that every connected component of T(Pn)[S] is a path of order at least 2 whose vertex set is either a subset of V or a subset of E. By the minimality of S, we have |V(Gi)|4 for each i. Our proof will be completed by showing that S satisfies the above three property.

P.1: Let V(Gj)={vi|ik} and V(Gj+1)={vi|k+2ik+r} for some j,,k,r. Then we can replace S by (SV(Gj+1)){ei(i+1)|k+2ik+r} which is again a min-TDS of T(Pn). There is a similar proof when both of V(Gi) and V(Gi+1) are subsets of E.

P.2: Since the case V(Gi)E has a similar proof, we assume V(Gi)V. If for some i and j,V(Gi)={vj,vj+1,vj+2,vj+3}, then we can replace V(Gi) by {vj+2,vj+3}{e(j1)j,ej(j+1)} or {vj,vj+1}{e(j+2)(j+3),e(j+3)(j+4)}, and find the min-TDSs (S{vj,vj+1}){e(j1)j,ej(j+1)} or (S{vj+2,vj+3}){e(j+2)(j+3),e(j+3)(j+4)}, respectively. Now let V(Gi)={vj+1,vj+2,vj+3} for some i and j. Then V(Gi+1)E by P.1, and j+4=min{m | em(m+1)V(Gi+1)} by the minimality of |S|. Then we can replace S by the min-TDS (S{vj+2}){e(j+3)(j+4)}ofT(Pn).

P.3: Since S is minimum, P.3 holds. □

Proposition 4.2.

For any cycle Cn of order n3, γtm(Cn)={4n73if n1(mod 7),4n72if n2,3(mod 7),4n71if n4(mod 7),4n7if n0,5,6(mod 7).

Proof.

Let Cn=(V,E) be a cycle of order n3 in which V={v1,v2,,vn} and E={ei(i+1) | 1in}. Then V(T(Cn))=VE and E(T(Cn))={ei(i+1)vi,ei(i+1)vi+1 | 1in}EE(L(Cn)) where E(L(Cn))={ei(i+1)e(i+1)(i+2) | 1in}. In a similar way to the proof of Proposition 4.1, it can be easily verified that the sets S0={v7i+2,v7i+3,e(7i+5)(7i+6),e(7i+6)(7i+7) | 0in71}if n0(mod 7),S=S0{e(n1)n}if n1(mod 7),S=S0{vn1,vn}if n2,3(mod 7),S=S0{vn2,vn1,vn}if n4(mod 7)S=S0{vn3,vn2,vn1,vn}if n5(mod 7),S=S0{vn4,vn3,e(n2)(n1),e(n1)n}if n6(mod 7). are min-TMDSs of Cn in each case, and this completes our proof. See as an example. □

Figure 8. The set {v2,v3,e56,e67,v9,v10,v11} is a min-TMDS of C11.

Figure 8. The set {v2,v3,e56,e67,v9,v10,v11} is a min-TMDS of C11.

Propositions 4.1 and 4.2 show that the total mixed domination numbers of a cycle and a path of the same order are roughly same in the following meaning.

Corollary 4.3.

For any integer n3, γtm(Cn)={γtm(Pn)+1if n4,5(mod 7),γtm(Pn)otherwise.

In the next step, we calculate the total mixed domination number of a complete bipartite graph.

Proposition 4.4.

For any integers nm1,γtm(Km,n)=m+1.

Proof.

Let VU be the partition of the vertex set of the complete bipartite graph Km,n to the independent sets V={vi | 1im} and U={uj | 1jn}. Since V{u1} is a TMDS of Km,n, we have γtm(Km,n)m+1.

Now, by the contrary, let S be a TMDS of Km,n with cardinality m. Since the subgraph of Km,n induced by V or U is isomorphic to the empty graphs Km¯ or Kn¯, respectively, we have SV and SU. For 1im and 1jn, we define Ri={eih | 1hn} and Cj={ehj | 1hm}. Let I={i | 1im, RiS} and J={j | 1jn, CjS}. Since SE implies I{1,,m} or J{1,,n}, and so for some iI or some jJ,vi or uj is not dominated by S, we have SE. Hence both of the sets IV={i | 1im, viS} and JU={j | 1jn, ujS} are nonempty. Because IV and JU= imply NT(Km,n)(vi)S= for some i, and IV= and JU imply NKm,n(uj)S= for some j, which are contradictions. Therefore RiS for each iIV or CjS for each jJU (beacause RiS= for some iIV and CjS= for some jJU imply NT(Km,n)(eij)S=). Hence |S(EEVU)|min{n|IV|,m|JU|} in which EVU={eij | iIV and jJU}, and so |S|=m|S(EEVU)|+|IV|+|JU|min{n|IV|,m|JU|}+|IV|+|JU|>m, a contradiction. Therefore γtm(Km,n)=m+1. See as an example. □

Figure 9. The set {v1,v2,v3,u1} is a min-TMDS of K3,3.

Figure 9. The set {v1,v2,v3,u1} is a min-TMDS of K3,3.

To find the total mixed domination number of a complete graph, we need next lemma.

Lemma 4.5.

Let S be a min-TMDS of a graph G=(V,E) in which V={v1,,vn} and E={eij | vivj is an edge}. If A={vi | viS, eijS for some j}, then |SE|{2|A|3if |A|0(mod 3),2|A|3+1if |A|0(mod 3).

Proof.

Let G=(V,E) be a graph in which V={v1,,vn} and E={eij | vivj is an edge}. Let S be a min-TMDS of G and let A={vi | viS, eijS for some j}. For any BA, we define EB={eijS | {vi,vj}B}. If NG(vi)A= for each viA, then EA and A have same cardinality, and so |SE||EA|=|A|2|A|3+1, as desired. Therefore, we assume that there exist two vertices vi,vjA while eijS, and continue our proof by induction on |A|. It can be easily verified that for any BA with cardinality at most 2, the following inequality Equation(4.0.1) holds, and we assume it holds for any set of cardinality less than |A|. Then we may assume eiNT(G)(eij)S for some j. By using the induction hypothesis for the set B=A{vi,vj,v}, which has cardinality m3 or m2, we have (4.0.1) |EB|{2|B|3if |B|0(mod 3),2|B|3+1if |B|0(mod 3).(4.0.1) Now by inequality Equation(4.0.1) and the fact that |EA||EB{eij,ei}|=|EB|+2, our proof will be completed. □

Proposition 4.6.

For any complete graph Kn of order n2,γtm(Kn)=5n3n.

Proof.

Let Kn be a complete graph with the vertex set V={v1,v2,,vn} and the edge set E. First we notice 5n3n={2n3if n0(mod 3),2n3+1if n1,2(mod 3). By Propositions 4.1 and 4.2, we may assume n4. For any arbitrary TMDS S of Kn we show (4.0.2) |S|{2n3if n0(mod 3),2n3+1if n1,2(mod 3).(4.0.2) If SE=, then |S|n12n3+1, and there is nothing to prove (because otherwise, for any two vertices vi and vj out of S, the edge eij can not be dominated by S). Also if SV=, then there exists an edge eikiS for dominating vi by S, and so inequality Equation(4.0.2) holds by Lemma 4.5. Therefore we assume SV and SE. Let |SV|=1. Then the set A={vi | viVS and eijS for some j} has cardinality at least n1 (because otherwise, for any two vertices vi,viVS, the edge eij does not dominate by S), and so |SE|{2(n1)3if n+1(mod 3),2(n1)3+1if n,+2(mod 3), by Lemma 4.5. Hence |S|=|SV|+|SE|{2n+23if n+1(mod 3),2n+23+1if n,+2(mod 3), which implies |S|{2n3if n0(mod 3),2n3+1if n1(mod 3),2n3+1if n2(mod 3) and 1.

Now we discuse on the only remained case n2(mod 3) and =1. Let SV={vn} and EA={eijS | {vi,vj}A}. Then n2|A|n1, and |EA|{2(n2)3if |A|=n20(mod 3),2(n1)3+1if |A|=n11(mod 3). Since |A|=n1 implies |S|=|SV|+|SE|1+|EA|2n3+1, as desired, we assume |A|=n2. For some pn, let epnS be an edge that dominates vn. Then vpA and epnEA, and so |A{vp}|=n32(mod 3). Hence |EA{vp}|2(n3)3+1 by (4.0.1). Now the facts epnEA{vp} and EA{vp}{epn,vn}S imply |S||EA{vp}|+22n3+1, as desired. On the other hand, since each of the sets S0={e(3i+1)(3i+2),e(3i+2)(3i+3) | 0ik1}if n0(mod 3),S1=S0{e(3k)(3k+1)}if n1(mod 3),S2=S0{e(3k)(3k+1)},e(3k+1)(3k+2)}if n2(mod 3), is a TMDS of Kn with the minimum cardinality when k=n3, we have proved γtm(Kn)={2n3if n0(mod 3),2n3+1if n1,2(mod 3). For an example, see . □

Figure 10. The set {e12,e23,e34} is a min-TMDS of K4.

Figure 10. The set {e12,e23,e34} is a min-TMDS of K4.

Before giving the total mixed domination number of a wheel, as we promise in the proof of Theorem 2.2, we show that the lower bound in Theorem 2.2 is tight by Propositions 4.6 and 4.7.

Proposition 4.7.

For any complete graph Kn of order n4,γt(L(Kn))=2n3.

Proof.

Let Kn=(V,E) be a complete graph of order n4 with the vertex set V={v1,,vn} and edge set E. Then V(L(Kn))=E. For any TDS S of L(Kn), let I be the set of all indices of the vertices of S. Obviously for any three indices 1i<j<kn,|{i,j,k}I|2 because for any i,jI, the vertex eij can not be dominated by S. Thus |S|2n3, and since the sets S0=S2{e(n2)(n1)}if n0(mod 3),S1={e(3i+1)(3i+2),e(3i+2)(3i+3), | 0in31}if n1(mod 3),S2=S1{e(n2)(n1)}if n2(mod 3), are TDSs of L(Kn) in each of the cases, the result holds. □

Proposition 4.8.

For any wheel Wn of order n+14,γtm(Wn)=n2+1.

Proof.

Let Wn=(V,E) be a wheel of order n+14 with the vertex set V={vi | 0in} and edge set E={e0i,ei(i+1) | 1in}. Since S={v0}{v2i1 | 1in2} is a TMDS of Wn, we have γtm(Wn)n2+1.

In the sequel, we show γtm(Wn)n2+1. Let S be an arbitrary TMDS of Wn. If SE=, then, since N(ei(i+1))S for each 1in,S{vi,vi+1}, and so |S|n2. Since we have nothing to prove when {v1,,vn}S, we assume viS for some 1in. This implies v0S (because S dominates e0i and SE=), and so |S|n2+1. Now let SV=. Then, for dominating every vertex viV by S, there exists an edge epiS for some pi. By knowing NWn(epi)={vp,vi}, we conclude S has cardinality at least n+12 which is n2+1 for even n and is n2 for odd n. Since the subgraph of Wn induced by S is connected and SE, we obtain |{vi | eijS for some j}|<n+1 if |S|=n+12 and n is odd, a contradiction. Thus |S|n+12+1=n2+1 for odd n.

Thus SV and SE. By assumption |SV|= it is sufficient to prove |SE|n2+1. Let E0={ei(i+1) | |{vi,vi+1}S|=0 for 1in}. Since every ei(i+1)E0 must be dominated by an edge epqS in which |{p,q}{i,i+1}|=1, the set E00={epqS | epqdominates an edge ei(i+1)E0} is not empty, and more |E00||E0|2 because every epqE00 is adjacent to at most two edges in E0. Let E1={ei(i+1) | |{vi,vi+1}S|0for 1in}. If v0S, then |E1|2(1), and so |E0|n2+2 which implies |SE||E0|2=n2+22=n2+1, as desired.

Therefore we may assume v0S. Then |E1|2 and so |E0|n2. Let |E0|n2+1 by the contrary. Hence 21|E1|2. Since by the assumption |E1|=2 we reach to this contradiction that the subgraph of Wn induced by SV contains  isolated vertices, we may assume |E1|=21. Again, since the subgraph of Wn induced by SV does not have isolated vertex, we must have =2, and so |E0|=n2+1=n3. Since obviousely |SE||E0|2=n2+1 for even n, let n be odd. Without loss of generality, we assume SV={v1,v2}, and so E0={ei(i+1) | 3in1}.

By the contrary let |SE|=n32. Since there is nothing to prove for n = 3 by Proposition 4.6, we assume n5. For n = 5, since v4 does not dominated by SE,|SE|=1 implies SE={ei4} for some i4, that is, S={v1,v2,ei4}. But since ei4 does not dominated by S, we reach contradiction. Thus |SE|>n32=1 and so γtm(W5)=52+1=4. Therefore, in the sequel, we assume n7. If SEE0, then SE=E0{e(2i)(2i+1) | 2in12} orSE={e(2i)(2i+1) | 2in32}{α} where  α{e(n2)(n1),e(n1)n}, which imply one of the edges e0n or e03 does not respectively dominated by S, a contradiction. Thus |S{e0i | 1in}|=m1. Let also V=VNG({v1,v2})={v4,,vn1},E0={ei(i+1)SE0 | {e0i,e0(i+1)}S} and E0=E0E0. So (4.0.3) |SE0|=n32m=|E0|+|E0|.(4.0.3) Since NWn(ei(i+1))={vi,vi+1} for ei(i+1)E0 and |N(e0j){vi,vi+1}|=1 when e0j{e0i,e0(i+1)}S, we have |NT(Wn)(SE0)V|m+|E0|. Since the subgraph Wn[E0] ofWn induced by E0 dominates the most number of vertices in V when it has as possible as the most number of the complete graphs K2 as induced subgraphs, we conclude that at least two edges of E0 are needed for dominating every three vertices of V by E0, and so |NT(Wn)(SE0)V|3|E0|2. Hence (4.0.4) |NT(Wn)(SE0)V|m+|E0|+3|E0|2.(4.0.4) Now by knowing VNT(Wn)(SE0) which implies |NT(Wn)(SE0)||V|=n4, and by the relations Equation(4.0.3) and Equation(4.0.4), we obtain |E0|n5, and so m = 1. Then |SE0|=n321=n52. Thus the number of vertices of V dominated by SE0 is at most 3(n5)4+1 which is less than n4=|V| when n7. Therefore |SE|n32+1=n12=n2+1, as desired. See for an example. □

Figure 11. The set {v0,v1,v3,v5} is a min-TMDS of W5.

Figure 11. The set {v0,v1,v3,v5} is a min-TMDS of W5.

5. Some open problems

Problem 5.1.

1. For any graph G, is it true that γtm(G)=γt(G) if and only if it has a min-TDS such that its complement is an independent set of G?

2. Find some families of non-complete graphs G satisfyγtm(G)=γt(L(G)).

3. Find some families of connected graphs G satisfyγtm(G)=γt(L(G))+γt(G).

It can be easily verified that Theorem 2.7 is true for any connected graph of order at most 5. So the existence of a Hamiltonian path in a graph is not a necessary condition in the theorem, and naturally the following problem arises.

Problem 5.2.

Find some other family of graphs G of order n satisfy γtm(G)5n3n.

We know γtm(G)>γt(G) for almost all graphs. As we saw in some graphs such as complete graphs and wheels, γtm(G)γt(G) when n for many graphs G. So, we end our paper with the following important problem.

Problem 5.3.

Find some real number α>1 such that for any graph G,γtm(G)αγt(G).

References

  • Behzad, M. (1967). A criterion for the planarity of a total graph. Math. Proc. Camb. Phil. Soc. 63(3): 679–681.
  • Haynes, T. W., Hedetniemi, S. T., Slater, P. J., eds. (1998). Fundamentals Domination in Graphs. New York: Marcel Dekker, Inc.
  • Haynes, T. W., Hedetniemi, S. T., Slater, P. J., eds. (1998). Domination in Graphs: Advanced Topics. New York: Marcel Dekker, Inc.
  • Henning, M. A., Yeo, A. (2013). Total Domination in Graphs (Springer Monographs in Mathematics).
  • West, D. B. (2001). Introduction to Graph Theory, 2nd ed. USA: Prentice Hall.
  • Zhao, Y., Kang, L, Sohn, M. Y. (2011). The algorithmic complexity of mixed domination in graphs. Theor. Comput. Sci. 412(22): 2387–2392.