1,535
Views
6
CrossRef citations to date
0
Altmetric
Articles

Certified domination

, , , &

Abstract

Imagine that we are given a set D of officials and a set W of civils. For each civil xW, there must be an official vD that can serve x, and whenever any such v is serving x, there must also be another civil wW that observes v, that is, w may act as a kind of witness, to avoid any abuse from v. What is the minimum number of officials to guarantee such a service, assuming a given social network?

In this paper, we introduce the concept of certified domination that models the aforementioned problem. Specifically, a dominating set D of a graph G=(VG,EG) is said to be certified if every vertex in D has either zero or at least two neighbours in VGD. The cardinality of a minimum certified dominating set in G is called the certified domination number of G. Herein, we present the exact values of the certified domination number for some classes of graphs as well as provide some upper bounds on this parameter for arbitrary graphs. We then characterise a wide class of graphs with equal domination and certified domination numbers and characterise graphs with large values of certified domination numbers. Next, we examine the effects on the certified domination number when the graph is modified by deleting/adding an edge or a vertex. We also provide Nordhaus–Gaddum type inequalities for the certified domination number.

1 Introduction

Imagine that we are given a set D of officials and a set W of civils. For each civil xW, there must be an official vD that can serve x, and whenever any such v is serving x, there must also be another civil wW that observes v, that is, w may act as a kind of witness, to avoid any abuse from v. What is the minimum number of officials to guarantee such a service, assuming a given social network? This problem motivates us introducing the concept of certified domination. Specifically, let D be a subset of the vertex set of a graph G=(VG,EG). We say that D dominates G (or is a dominating set of G) if each vertex in the set VGD has a neighbour in D. The cardinality of a minimum dominating set in G is called the domination number of G and denoted by γ(G), and any minimum dominating set of G is called a γ-set. A dominating set D of G is called certified if every vertex vD has either zero or at least two neighbours in VGD. The cardinality of a minimum certified dominating set in G is called the certified domination number of G and denoted by γcer(G). A minimum certified dominating set of G is called a γcer-set. Notice that, by the definition, VG is a certified dominating set of G, and certainly 1γcer(G)|VG|. Furthermore, one can observe that γcer(G)|VG|1.

There is a wealth of literature about domination and its variations in graphs; we refer to the excellent books of Haynes, Hedetniemi, and Slater [Citation1,2]. The domination concept we introduce perfectly fits into that area where, for a given graph G, domination parameters are defined by imposing additional constraints on a dominating set D or its complement VGD. This area includes, to mention but a few, multiple domination, distance domination, or global domination. In particular, the problem of certified domination is closely related to the problem of existence a DD2-pair in a graph, introduced by Henning and Rall in [Citation3]. Recall, a set XVG of vertices is 2-dominating in G if every vertex in VGX has at least two neighbours in X. A DD2-pair of G is a pair (D,D2) of disjoint sets of vertices of G such that D is a dominating set of G and D2 is a 2-dominating set of G; a graph that has a DD2-pair is called a DD2-graph. One can observe that if G has a DD2-pair (D,D2), then the set D is a certified dominating set. However, there are graphs G with γcer(G)<|D| for any (D,D2)-pair in G (if any), see for an illustration.

Fig. 1 The family of graphs Gi. (a) Black vertices form a certified dominating set Dc with |Dc|=i+3, i2. (b) Black and grey vertices form a (D,D2)-pair, respectively, with |D|=2i+1. Observe that if i3, then Gi has no (D,D2)-pair with |D|i+3.

Fig. 1 The family of graphs Gi. (a) Black vertices form a certified dominating set Dc with |Dc|=i+3, i≥2. (b) Black and grey vertices form a (D,D2)-pair, respectively, with |D|=2i+1. Observe that if i≥3, then Gi has no (D,D2)-pair with |D|≤i+3.

In Section 2, we present the exact values of the certified domination number for some elementary classes of graphs. Some upper bounds on this new parameter for an arbitrary graph are presented in Section 3. Then, in Sections 4 and 5, respectively, we characterise a wide class of graphs with equal domination and certified domination numbers and characterise graphs with large values of certified domination numbers. Next, in Section 6, we examine the effects on the certified domination number when the graph is modified by deleting/adding an edge or a vertex. Finally, Section 7 is devoted to Nordhaus–Gaddum type inequalities for the certified domination number.

1.1 Definitions and notation

For general graph theory terminology, we follow [Citation4]. In particular, for a vertex v of a graph G=(VG,EG), its neighbourhood, denoted by NG(v), is the set of all vertices adjacent to v, and the cardinality of NG(v), denoted by degG(v), is called the degree of v. The closed neighbourhood of v, denoted by NG[v], is the set NG(v){v}. In general, for a subset XVG of vertices, the neighbourhood of X, denoted by NG(X), is defined to be vXNG(v), and the closed neighbourhood of X, denoted by NG[X], is the set NG(X)X. The minimum (maximum, resp.) degree of a vertex in G is denoted by δ(G) (Δ(G), resp.). A vertex of degree |VG|1 is called a universal vertex of G. A vertex of degree one is called a leaf, and the only neighbour of a leaf is called its support vertex (or simply, its support). If a support vertex has at least two leaves as neighbours, we call it a strong support, otherwise it is a weak support. The set of leaves of G is denoted by LG. For a leaf vLG, its support vertex is denoted by sG(v), and for a weak support v, the unique leaf adjacent to v is denoted by lG(v). The set of weak supports of G is denoted by S1(G), while the set of strong supports of G is denoted by S2(G).

2 Elementary graph classes

We begin by presenting the exact values of the certified domination number for some elementary classes of graphs.

Observation 2.1

If Pn is an n-vertex path, then γcer(Pn)=1if n=1 or n=3;2if n=2;4if n=4;n3otherwise.

Observation 2.2

If Cn is an n-vertex cycle, n3, then γcer(Cn)=n3.

Observation 2.3

If Kn is an n-vertex complete graph, then γcer(Kn)=1if n=1 or n3;2if n=2.

Observation 2.4

If Km,n is a complete bipartite graph with 1mn, then γcer(Km,n)=1if m=1 and n>1;2otherwise.

Observation 2.5

If Wn is an n-vertex wheel, then γcer(Wn)=1.

In addition, we have the following two general observations on the certified domination number of a graph.

Observation 2.6

If G is a graph of order at least three, then γcer(G)=1 if and only if G has a universal vertex.

Observation 2.7

If G1,,Gk are the connected components of a graph G, then γcer(G)=i=1kγcer(Gi).

3 Upper bounds on the certified domination number

In this section we focus on upper bounds on the certified domination number. We start with two simple observations and then present our main result of this section: an upper bound on γcer(G) with respect to the domination number γ(G) and the number |S1(G)| of weak supports in G.

Observation 3.1

Every support vertex of a graph G belongs to every certified dominating set of G.

Proof

Let Dc be a certified dominating set of G, let s be a support vertex of G, and let l be a leaf adjacent to s. If s were not in Dc, then l should be in Dc. But then l would have only one neighbour in VGDc, and Dc would not be a certified dominating set. □

Observation 3.2

Let G be a graph of order n. If the strong supports of G are adjacent to k leaves in total, then γcer(G)nk. In particular, γcer(G)n2|S2(G)|.

Proof

Let L be the set of all leaf-neighbours of strong supports of G. Then |L|=k and the set VGL is a certified dominating set of G. Thus γcer(G)|VGL|=nkn2|S2(G)| as |L|2|S2(G)|. □

Before we present our main result, let us introduce some useful terminology. Let D be a dominating set of a graph G. An element of D that has all neighbours in D is said to be shadowed with respect to D (shadowed for short), an element of D that has exactly one neighbour in VGD is said to be half-shadowed with respect to D (half-shadowed for short), while an element of D having at least two neighbours in VGD is said to be illuminated with respect to D (illuminated for short). It is easy to observe that if D is a minimum dominating set of a graph with no isolated vertices, then D has no shadowed element, and if D is a certified dominating set, then D has no half-shadowed element.

Theorem 3.3

If G is a connected graph, then γcer(G)γ(G)+|S1(G)|.

Proof

If G is a graph of order at most two, then the inequality is obvious. Thus assume that G has at least three vertices. Let D be a γ-set of G that minimises the number of half-shadowed vertices and such that D does not contain any leaf of G. (Notice that such D always exists as G is connected and |VG|3.) Let DhsD be the set of all half-shadowed vertices of D. If Dhs=, then γcer(G)=γ(G)γ(G)+|S1(G)|. Thus assume that Dhs.

Claim 1

If vDhs, then degG(v)2 and vS2(G).

The inequality degG(v)2 follows from the choice of D, that is, from the assumption that DLG=. To argue the second property, suppose on the contrary that v is a strong support. Again, since v has at least two neighbours in LG and LGVGD, v would not be half-shadowed, a contradiction.

Next we show that all half-shadowed vertices are weak supports. Suppose on the contrary that there is a half-shadowed vertex vDhsS1(G) and let u be the unique neighbour of v in VGD. Since v is neither a weak nor strong support (by assumption and Claim 1, respectively), it implies that u is not a leaf. Furthermore, we have the following claim.

Claim 2

The set NG(u){v} is a subset of VGD.

Otherwise the set D{v} would be a smaller (than D) dominating set of G.

Finally, we have the following claim.

Claim 3

No vertex belonging to the set NG(v){u} is shadowed.

If a vertex wNG(v){u} was shadowed, then the set D{w} would be a smaller (than D) dominating set of G.

Consequently, keeping in mind the fact that none of neighbours of v is a leaf (see Claim 1), and combining Claims 2 and 3, we conclude that the set (D{v}){u} would be a γ-set of G with a smaller number of half-shadowed vertices, a contradiction. This proves that the set Dhs of half-shadowed vertices consists of weak supports of G only.

Observe now that adding to D all leaves adjacent to half-shadowed weak supports results in a dominating set D of G with no half-shadowed vertices, that is, D is a certified dominating set of G. Therefore γcer(G)|D|=|D|+|Dhs|=γ(G)+|Dhs|γ(G)+|S1(G)|. □

From Observation 2.7 and Theorem 3.3, we immediately obtain the following corollary.

Corollary 3.4

If G is a graph, then γcer(G)γ(G)+|S1(G)|.

4 Graphs with γcer=γ

We continue our study on the certified domination number by focusing now on the class of graphs with γcer=γ. When trying to characterise this class, one may expect that the main problem lies in leaves of a graph. In fact, from Corollary 3.4 we immediately have the first result.

Corollary 4.1

If G is a graph with no weak support, then γcer(G)=γ(G).

The above corollary also follows from the next more general lemma.

Lemma 4.2

If a connected graph G has at least three vertices, then γcer(G)=γ(G) if and only if there exists a minimum dominating set D of G such that NG(s)LGD for every sS1(G).

Proof

Assume first that γcer(G)=γ(G). Let Dc be a minimum certified dominating set of G. Then Dc is a minimum dominating set of G. Now, if sS1(G), then Dc{s,lG(s)} (as Dc is dominating in G), |Dc{s,lG(s)}|2 (otherwise Dc{lG(s)} would be a smaller dominating set of G), and Dc{s,lG(s)}{lG(s)} (otherwise lG(s) would be half-shadowed). Thus Dc{s,lG(s)}={s} and (NG(s)LG)(VGDc)=(NG(s){lG(s)})(VGDc) (otherwise s would be half-shadowed), and so NG(s)LGDc.

Assume now that in G there exists a γ-set D such that NG(s)LGD for every sS1(G). Of all such sets, choose one, say D, that does not contain any leaf of G (such D exists in every connected graph of order at least three) and minimises the number of its half-shadowed vertices. We claim that such D is a certified dominating set of G (and therefore γ(G)=|D|=γcer(G)). Suppose, on the contrary, that some element v of D is half-shadowed. Let v be the unique element of NG(v)D. Since v is half-shadowed, vS2(G), and vS1(G) (as every element of S1(G) is illuminated by the adjacent leaf and, by the assumption, by at least one non-leaf). Finally, since DLG= (by the choice of D) and vD, we have vLG and dG(v)2. Now, if it were NG(v)(D{v}), then D{v} would be a dominating set of G smaller than D, a contradiction. Thus NG(v){v} must be a nonempty subset of VGD and, then, D=(D{v}){v} is a minimum dominating set of G and it has less half-shadowed vertices than D, a final contradiction which proves that γ(G)=γcer(G). □

Observe that if G=Kn¯, then γcer(G)=n=γ(G). Next, if G=lK2, then γcer(G)=2ll=γ(G). In the latter case, S1(G)=VG=LG and G has no minimum dominating set D of G such that NG(s)LGD for every sS1(G). Therefore, taking into account Observation 2.7 and Lemma 4.2, we obtain the following corollary for graphs which are not necessarily connected.

Corollary 4.3

If G is a graph, then γcer(G)=γ(G) if and only if there exists a minimum dominating set D in G such that NG(s)LGD for every sS1(G).

Furthermore, we have the following relation between graphs each of which has a unique minimum dominating set and those for which γcer and γ are equal.

Corollary 4.4

If a graph G has a unique minimum dominating set, then γcer(G)=γ(G).

Proof

If S1(G)=, then γcer(G)=γ(G) by Corollary 4.1. Thus assume that S1(G). Let D be the minimum dominating set of G. From the uniqueness and minimality of D it follows that S1(G)D and LGVGD. Now, if it were γcer(G)γ(G), then, by Lemma 4.2, we could find sS1(G) such that NG(s){lG(s)}D, and then the set (D{s}){lG(s)} would be another minimum dominating set of G, which is impossible. □

Remarks

From Corollary 4.1 and the fact that the problem of determining the domination number in bipartite planar subcubic graphs with no leaves is NP-hard (as it was observed in [Citation5,6]), we immediately obtain the following: The problem of determining the certified domination number is NP-hard even in bipartite planar subcubic graphs with no leaves. Next, let G be a graph with no isolated vertex. If G has a minimal dominating set D which is also a certified dominating set, then its complement VGD is a 2-dominating set of G, and, therefore, G is a DD2-graph. Thus, from Corollaries 4.1 and 4.4, we have the following generalisation of Theorem 3 in [Citation3]: Let G be a graph with no isolated vertex. If G has no weak support or G has a unique dominating set, then G is a DD2-graph and it has a (D,D2)-pair in which |D|=γ(G).

5 Graphs with large values of γcer

As we have already observed, for any graph G of order n, γcer(G)n, γcer(G)n1, and there are graphs G with γcer(G)=n, for example, the complement of a complete graph Kn or a 4-vertex path P4. Thus it is natural to try to characterise all graphs with γcer=n and γcer=n2, respectively, which is carried out in this section. In particular, we prove that γcer(G)=n if and only if G is the complement of a complete graph, the corona of a graph, or the union of both of them. Recall, the corona product (or simply, the corona) of two graphs H and F is the graph G=HF resulting from the disjoint union of H and |VH| copies of F in which the ith vertex of H is joined to all vertices of the ith copy of F. If F is a 1-vertex graph, F=K1, then the corona HK1 is simply called the corona of H.

Lemma 5.1

Let G be a connected graph of order n. If G is the corona of some graph, then γcer(G)=n.

Proof

Let Dc be a smallest certified dominating set of G. It suffices to prove that Dc=VG. This is obvious if n=2. Thus assume n>2. In this case, since G is the corona of some graph, every vertex of G either is a leaf of G or is adjacent to exactly one leaf of G. From this and from Observation 3.1 it follows that VGLGDc. Moreover, every leaf l of G also belongs to Dc (as otherwise its only neighbour sG(l) would be half-shadowed). Consequently, LGVG and therefore Dc=VG. □

Lemma 5.2

Let G be a connected graph of order n2. If γcer(G)=n, then G is the corona of some graph.

Proof

The statement is obvious for connected graphs of order at most 4. Thus assume that G is a connected graph of order n5 and γcer(G)=n. Now, since γ(G)n2 for every graph with no isolated vertex, so by Theorem 3.3 we have γcer(G)γ(G)+|S1(G)|2γ(G)n=γcer(G). Thus γ(G)=n2 and so G is the corona of some graph (as it was proved in [Citation7,8]). □

From the above lemmas, we immediately conclude with the following theorem.

Theorem 5.3

If G is a graph of order n, then γcer(G)=n if and only if G is either the complement of a complete graph, or the corona of a graph, or the union of both of them.

Remark

We incidentally observe that the above result implies the sharpness of the upper bound in the inequality γcer(G)γ(G)+|S1(G)| (see Theorem 3.3 and Corollary 3.4) as well as in the inequality γcer(G)2γ(G), since for the corona G of any graph without an isolated vertex, we have |S1(G)|=γ(G) and γcer(G)=2γ(G).

5.1 Graphs with γcer=n2

A diadem graph of a graph H is a graph obtained from the corona HK1 by adding a new vertex, say v, and joining v to one of support vertices of HK1 (see ).

Fig. 2 The diadem graph resulting from the corona G=(K3K2)K1 by adding a leaf to the support vertex s of G.

Fig. 2 The diadem graph resulting from the corona G=(K3∪K2)∘K1 by adding a leaf to the support vertex s of G.

Lemma 5.4

If G is a diadem graph of order n, then γcer(G)=n2.

Proof

Let s be the unique strong support of G, and let l1,l2 be the two leaves of G adjacent to s in G. It is obvious that VG{l1,l2} is a certified dominating set of G. Let Dc be a smallest certified dominating set of G. Then VGLGDc (by Observation 3.1) and {l1,l2}Dc=. Moreover, every leaf l different from l1 and l2 belongs to Dc (otherwise sG(l) would be half-shadowed). Consequently Dc=VG{l1,l2} and therefore γcer(G)=n2. □

Lemma 5.5

Let G be a connected graph of order n. If γcer(G)=n2, then G=C3, G=C4, or G is a diadem graph ( of a connected graph).

Proof

If G is a connected graph of order at most n4 and γcer(G)=n2, then G=K1,2, G=C3 or G=C4. Thus assume that n5. In this case δ(G)=1, as otherwise, since γcer(G)=γ(G) (by Corollary 4.1), γcer(G)=n2, and γ(G)n2, we would have n2=γcer(G)=γ(G)n2, which is impossible. We now claim that G is a diadem graph.

By way of contradiction, suppose that the claim is false. Let G be a smallest counterexample, say of order n (n5), such that γcer(G)=n2 and G is not a diadem graph. Let Dc be a γcer-set of G, and let v and u be the only elements of VGDc. From the fact that Dc=VG{v,u} is a certified dominating set of G it follows that if xDc, then either xNG(v)NG(u) or xNG(v)NG(u). This proves that NG(v)Dc=NG(u)Dc. In addition, the set VGNG[{v,u}] is nonempty, as otherwise {v,u} would be a certified dominating set of G and we would have n2=γcer(G)|{v,u}|=2, which is impossible.

Let G denote the subgraph GNG[{v,u}] of G. From the assumption γcer(G)=n2 it easily follows that γcer(G)=|VG|. Thus, by Theorem 5.3, every connected component of G is an isolated vertex or the corona of a graph.

Let H be a connected component of G. From the fact that Dc=VG{v,u} is a minimum certified dominating set of G it follows that at least one vertex of H is not adjacent to any vertex belonging to NG[{v,u}]{v,u} as otherwise DcVH would be a certified dominating set of G, which is impossible as γcer(G)|DcVH|<|Dc|=γcer(G). From this we conclude that G has no isolated vertex. Consequently, every connected component of G is the corona of a graph.

We now claim that K2 is not a connected component of G. Suppose on the contrary that K2 on vertices a and b is a connected component of G. Then one of the vertices a and b is a leaf in G and the latter one is adjacent to a vertex in NG[{v,u}]{v,u}, say aLG and b is adjacent to a vertex wNG[{v,u}]{v,u}. Let G˜ denote the graph G{a,b} (of order n2). For this graph either γcer(G˜)<n4, or γcer(G˜)=n4, or γcer(G˜)>n4. Assume first that γcer(G˜)<n4. Let D˜c be a smallest certified dominating set of G˜. Then D˜c{b} (if (NG(b){a})D˜c) or D˜c{a,b} (if NG(b){a}D˜c) is a certified dominating set of G and γcer(G)|D˜c{a,b}|=γcer(G˜)+2<n2, a contradiction. Assume now that γcer(G˜)>n4. Then γcer(G˜)=n2=|VG˜| and, by Theorem 5.3, G˜ is the corona of a graph. But this is impossible as no vertex of NG[{v,u}]{v,u} is a leaf or a neighbour of exactly one leaf. Finally, assume that γcer(G˜)=n4=|VG˜|2. In this case the choice of G implies that G˜ is the diadem graph in which v and u are leaves and w is their only common neighbour. Now, it is obvious that the graph G (obtained from G˜ by the addition of the vertices a and b, and the edges ab and bw) is a diadem graph, a contradiction.

Now, to complete the proof, it suffices to show that this smallest counterexample is not a counterexample, that is, it suffices to show that G is a diadem graph. It is enough to prove that: (1) no vertex belonging to NG[{v,u}] is adjacent to a leaf of a connected component of G of order at least four, (2) v and u have exactly one common neighbour, and (3) v and u are not adjacent in G.

(1)

Suppose on the contrary that there is a vertex in NG[{v,u}] adjacent to a leaf l of a connected component H (of order at least four) of G. Let L be the set of leaves of H within the distance at most 2 from sH(l). Then D=Dc(L{sH(l)}) is a certified dominating set of G and |D|<|Dc|, a contradiction.

(2)

Suppose on the contrary that |NG[{v,u}]{v,u}|2. Let us consider the set S={xVG:NG(x)NG({v,u})}. By (1), S is a subset of VGLG. In addition, since G is the corona of a graph, every vertex of S is adjacent to a vertex of LG. From the supposition |NG[{v,u}]{v,u}|2 and from properties of elements of S it follows that D={v,u}(VGLG)(LGNG(S)) (={v,u}(VG(NG(S)LG))) is a certified dominating set of G and |D|<|Dc|, a contradiction.

(3)

Suppose on the contrary that vuEG, and consider the graph G=Gvu of order n, in which, by (2), v and u are leaves, and they have exactly one common neighbour, say w. In this graph we have either γcer(G)>n2 (and therefore γcer(G)=n), or γcer(G)=n2, or γcer(G)<n2. Assume first that γcer(G)=n. Then, by Theorem 5.3, G is the corona of a graph, but this is impossible as leaves v and v share the same neighbour w. Assume now that γcer(G)=n2. Then, by the choice of G, G is a diadem graph. Let L be the set of leaves of G within the distance at most 3 from v (and u). Then D=(Dc(L{w})){v} is a certified dominating set of G and |D|<|Dc|, a contradiction. Finally, assume that γcer(G)<n2. Let Dc be a smallest certified dominating of G. Since w is a strong support of G, wDc by Observation 3.1, and v,uDc by minimality of Dc. But then, Dc is also a certified dominating set of G and so γcer(G)<n2, a final contradiction. □

From Theorem 5.3, Lemmas 5.4 and 5.5, we have the final characterisation of graphs of order n with γcer=n2.

Theorem 5.6

Let G be a graph of order n3. Then γcer(G)=n2 if and only if G is C3,C4, or a diadem graph, or G is one of these three graphs with possible number of isolated vertices, or G is the union of one of these three graphs with the corona of some graph, with possible number of isolated vertices.

6 Influence of deleting/adding edge/vertex

In this section, following [Citation9–12], to mention but a recent few, we examine the effects on the certified domination number when the graph is modified by deleting/adding an edge or a vertex. We observe that deleting an edge or a vertex may arbitrarily increase the certified domination number. For example, for the graph Gi of order 2i+4 illustrated in (a) we have γcer(Gi)=i+1 and γcer(Gie)=2i+4. To argue a similar influence of deleting a vertex, consider a wheel graph Wn with the hub v. We have γcer(Wn)=1 and γcer(Wnv)=(n1)3.

Fig. 3 Adding or deleting an edge may arbitrarily increase the certified domination number.

Fig. 3 Adding or deleting an edge may arbitrarily increase the certified domination number.

Adding an edge to a graph may also arbitrarily increase the certified domination number. Namely, consider the disconnected graph Hi of order 2i+4 illustrated in (b). We have γcer(Hi)=i+2 and γcer(Hi+e)=2i+4. However, adding an edge to a connected graph does not increase the certified domination number, that is, γcer(G+e)γcer(G) for any connected graph G. To argue this property, we use the following lemma.

Lemma 6.1

Let Dc be a γcer-set of a connected graph G of order n2. Then:

(a)

Every shadowed vertex in Dc is a weak support or a leaf.

(b)

Every non-leaf neighbour of a shadowed weak support is either an illuminated vertex or a shadowed weak support.

Proof

(a) Consider a shadowed vertex vDc. Suppose on the contrary that v is neither a weak support nor a leaf in G. By minimality of Dc, there are no shadowed strong supports in Dc, in particular, v is not a strong support, and thus all neighbours of v are of degree at least two. Let XDc be a maximal subset of shadowed vertices such that (i) vX, (ii) the induced subgraph G[X] is connected, and (iii) none of elements of X is an illuminated vertex or a shadowed weak support. Next, for a vertex xX, define the set BG(x)=NG(x)X. Analogously, define the set BG(X)=xXBG(x).

Observe that by minimality of Dc, each vertex xX is a non-support vertex, and by the choice of X, and every element in BG(X) is either an illuminated vertex or a shadowed weak support.

Case 1: G[X] is a 1-vertex graph. Let L be the set of shadowed leaves within the distance 2 from v. Then the set D=Dc(L{v}) would be a certified dominating set of G and |D|<|Dc|, a contradiction.

Case 2: |X|2 and γcer(G[X])=|X|. By Theorem 5.3, G[X] is the corona of some connected graph. Observe that by the choice of X and minimality of Dc, if xX is a leaf of G[X], then the set BG(x) is non-empty.

Consider now a weak support s in G[X]. Let L1 be the set of leaves of G[X] within the distance at most 2 from s and let L2 be the set of shadowed leaves of G within the distance 2 from L1{s}. Then the set D=Dc(L1L2{s}) is a certified dominating set of G and |D|<|Dc|, a contradiction.

Case 3: |X|3 and γcer(G[X])|X|2 (as the case γcer(G[X])=|X|1 is impossible). Let DX be a γcer-set of G[X] and let D¯X=XDX. Let L3 be the set of shadowed leaves within the distance 2 from D¯X. Then the set D=Dc(D¯XL3) is a certified dominating set of G and |D|<|Dc|, a contradiction.

(b) A non-leaf neighbour of a shadowed weak support sDc is either illuminated or shadowed. If s is shadowed, then, since it is not a leaf, it must be a weak support by (a). □

Theorem 6.2

If G is a connected graph of order n2, then γcer(G+e)γcer(G).

Proof

One can verify the validity of the theorem for graphs of order at most n4. So assume n5 and let Dc be a γcer-set of G.

Let e=vw, v,wVG, be the added edge to G. If both v,wDc, then Dc is also a certified dominating set of the graph G+e. Similarly, if either both v,wDc, or vDc and wDc is illuminated, or vDc is illuminated and wDc, then Dc is a certified dominating set of G+e as well. Therefore, in all aforementioned cases, we have γcer(G+e)|D|=γcer(G) as required.

Without loss of generality assume now that vDc and wDc is shadowed (the case wDc and vDc is shadowed can be analysed in a similar way). By Lemma 6.1 (a), w is either a weak support or a leaf of G.

Case 1: w is a weak support of G. Then the set D=Dc{lG(w)} is a certified dominating set of G+e, and thus, γcer(G+e)|D|<|Dc|=γcer(G).

Case 2: w is a leaf of G. By the choice of Dc, it follows that the support vertex sG(w) is weak and shadowed. Therefore, by Lemma 6.1 (b), every non-leaf neighbour of the weak support sG(w) in G is either an illuminated vertex or a shadowed weak support of G. Let L be the set of shadowed leaves within the distance 2 from sG(w) in G. Then, the set D=Dc(L{sG(w)}) is a certified dominating set in G+e, and hence γcer(G+e)|D||Dc|1<γcer(G). □

Adding a vertex can arbitrarily increase the certified domination number, which is not the case as in the model of classic domination. Indeed, for the graph Gi of order 2i+1 depicted in , we have γcer(Gi)=i, while γcer(Gi+v)=2i+2. However, bearing in mind Corollary 4.1, one can expect that the clue of the above construction lies in adding a leaf. Indeed, this is the case since one can prove that adding a non-leaf vertex does not effect the certified domination number significantly (the being added vertex v is called a non-leaf vertex if v is not a leaf in the resulting graph). Namely, we have the following theorem.

Fig. 4 Graph Gi has 2i+1 vertices, and γcer(Gi)=i, while γcer(Gi+v)=2i+2.

Fig. 4 Graph Gi has 2i+1 vertices, and γcer(Gi)=i, while γcer(Gi+v)=2i+2.

Theorem 6.3

If we add a non-leaf vertex v to a graph G, then γcer(G+v)γcer(G)+1.

Proof

Let Dc be a γcer-set of a graph G and let v be a new added vertex.

Case 1: degG+v(v)=2. Let u and w be the two neighbours of v in G+v. If either u,wDc or both u,wDc, then the set Dc{v} is a certified dominating set of G+v, and thus γcer(G+v)γcer(G)+1. Otherwise, without loss of generality, we consider two subcases.

Subcase 1.a: uDc, wDc, and w is illuminated. Then the set Dc remains a certified dominating set of G+v, and in this case, γcer(G+v)γcer(G) holds.

Subcase 1.b: uDc, wDc, and w is shadowed. If the vertex w constitutes a 1-vertex component of G, then the set (Dc{w}){v} is a certified dominating set in G+v, thus getting γcer(G+v)γcer(G)+1. Otherwise, by Lemma 6.1 (a), w is either a weak support or a leaf of G. Now, similarly as in the proof of Theorem 6.2, we consider two subcases.

Subcase 1.b.1: w is a weak support of G. (We emphasise that this subcase includes the case when w and the lG(w) constitute a 2-vertex component of G.) Then the set Dc{lG(w)} is a certified dominating set in G+v. In this case, γcer(G+v)γcer(G)1 holds.

Subcase 1.b.2: w is a leaf of G, and w together with the support vertex sG(w) does not constitute a 2-vertex component of G. By the choice of Dc, the support vertex sG(w) is weak and shadowed. By Lemma 6.1 (b), every non-leaf neighbour of sG(w) in G is either an illuminated vertex or a shadowed weak support of G. Again, let L be the set of shadowed leaves within the distance 2 from sG(w) in G. Then, the set D=Dc(L{sG(w)}) is a certified dominating set in G+v. In this case, γcer(G+v)γcer(G)1.

Case 2: degG+v(v)3. Then, when adding v to G, we first add only two edges, thus obtaining a temporary graph G, where degG(v)=2. Now, taking into account Case 1, we conclude that γcer(G)γcer(G)+1. Next, when adding all the remaining edges to G, sequentially, to obtain the final graph G+v, we apply Theorem 6.2, sequentially, for each of added edge, thus getting γcer(G+v)γcer(G)γcer(G)+1 as required. □

7 Nordhaus–Gaddum type results

Following the precursory paper of Nordhaus and Gaddum [Citation13], the literature has became abundant in inequalities of a similar type for many graph invariants, see a recent survey by Aouchiche and Hansen [Citation14]. In particular, the following result is known for the domination number.

Theorem 7.1

[15–17] If G¯ is the complement of a graph G of order n, then:

(a)

γ(G)γ(G¯)n;

(b)

γ(G)+γ(G¯)n+1 with equality if and only if G=Kn or G¯=Kn.

Further sharpening of bounds was done for the case when, for example, both G and G¯ are connected [Citation18] or for graphs with specified minimum degree [Citation19], to mention but a few. In particular, the following theorem was proved in [Citation20].

Theorem 7.2

[Citation20] If G is a graph of order n and neither G nor G¯ has an isolated vertex, that is, 1δ(G)n2, then γ(G)+γ(G¯)n2+2.Moreover, if n9, the bound is attained if and only if {γ(G),γ(G¯)}={n2,2}.

In this section we provide some Nordhaus–Gaddumtype inequalities for the certified domination number. First, taking into account Corollary 4.1, Theorems 7.1 and 7.2, we obtain the following corollary.

Corollary 7.3

If G is a graph of order n and min{δ(G),δ(G¯)}2, then γcer(G)+γcer(G¯)n2+2andγcer(G)γcer(G¯)n.

By enumerating all graphs of order at most 4, we obtain the following observation.

Observation 7.4

Let G be a graph of order n. Then:

(a)

γcer(G)+γcer(G¯)=γcer(G)γcer(G¯)=4 if n=2;

(b)

γcer(G)+γcer(G¯)=4 and γcer(G)γcer(G¯)=3 if n=3;

(c)

(γcer(G)+γcer(G¯),γcer(G)γcer(G¯)){(3,2),(5,4),(6,6),(8,16)} if n=4.

Next, we have the following theorem.

Theorem 7.5

If G is a graph of order n3 and min{δ(G),δ(G¯)}=0, then γcer(G)+γcer(G¯)n+1andγcer(G)γcer(G¯)n.In addition, if min{δ(G),δ(G¯)}=0, then each of the above upper bounds is attainable, and the following statements are equivalent:

(a)

γcer(G)+γcer(G¯)=n+1;

(b)

γcer(G)γcer(G¯)=n;

(c)

G or G¯ is the complement of Kn or the union of the corona of some graph and a positive number of isolated vertices.

Proof

From the assumption min{δ(G),δ(G¯)}=0, it follows that max{Δ(G¯),Δ(G)}=n1 and, therefore, γcer(G¯)=1 or γcer(G)=1. Now, since γcer(G)n and γcer(G¯)n, we get γcer(G)+γcer(G¯)n+1 and γcer(G)γcer(G¯)n.

Assume now that min{δ(G),δ(G¯)}=0, say δ(G)=0. Then Δ(G¯)=n1, and so γcer(G¯)=1. Now, since γcer(G¯)=1, it follows from each of the equalities γcer(G)+γcer(G¯)=n+1 and γcer(G)γcer(G¯)=n that γcer(G)=n. Finally, since δ(G)=0 and γcer(G)=n, we conclude from Theorem 5.3 that G is the complement of K¯n or the union of the corona of some graph and a positive number of isolated vertices. This proves the implications (a)(c) and (b)(c). Opposite implications are straightforward. □

Finally, we have the following theorem.

Theorem 7.6

If G is a graph of order n5, then γcer(G)+γcer(G¯)n+2andγcer(G)γcer(G¯)2n.In addition, each of the above upper bounds is attainable, and the following statements are equivalent:

(a)

γcer(G)+γcer(G¯)=n+2;

(b)

γcer(G)γcer(G¯)=2n;

(c)

G or G¯ is the corona of some graph.

Proof

If min{δ(G),δ(G¯)}2, then γcer(G)+γcer(G¯)n2+2n+2 and γcer(G)γcer(G¯) n2n by Corollary 7.3. If min{δ(G),δ(G¯)}=0, then γcer(G)+γcer(G¯)n+1n+2 and γcer(G)γcer(G¯)n2n by Theorem 7.5.

Thus assume min{δ(G),δ(G¯)}=1. Then max{Δ(G),Δ(G¯)}=n2. This also implies that γcer(G)>1 and γcer(G¯)>1. Thus, since γcer(G)n and γcer(G¯)n, it suffices to show that γcer(G)=2 or γcer(G¯)=2. Without loss of generality assume that δ(G)=1. Let l be a leaf of G and let s be the only element of NG(l). We consider two cases: degG(s)=n2, and degG(s)n3.

Case 1: degG(s)=n2. Let t be the only element of VGNG[s]. Assume first that dG(t)2. Let u and w be two neighbours of t (and s). Now, because NG[{s,t}]=NG[s]NG[t]=(VG{t})NG[t]=VG, {u,w}NG(s)(VG{s,t}), {u,w}NG(t)(VG{s,t}), and γcer(G)>1, we conclude that {s,t} is a minimum certified dominating set of G, and γcer(G)=2. Assume now that dG(t)=1. In this case, let u and w be vertices such that NG(t)={u} and wNG(s){l,u}. Since NG¯[{l,t}]=NG¯[l]NG¯[t]=(VG{s})(VG{u})=VG, the set {l,t} is dominating in G¯. In addition, since {u,w}NG¯(l)(VG{l,t}) and {s,w}NG¯(t)(VG{l,t}), the set {l,t} is certified dominating in G¯. From this and from the fact that γcer(G¯)>1 it follows that γcer(G¯)=2.

Case 2: degG(s)n3. Let t and u be two elements of the set VGNG[s]. In this case, {l,s} is a certified dominating set of G¯, since NG¯[{l,s}]=VG, {t,u}NG¯(l)(VG{l,s}), and {t,u}NG¯(s)(VG{l,s}). From this it again follows that γcer(G¯)=2.

We now prove the equivalence of (a), (b), and (c). Let G be a graph of order n5 such that γcer(G)+γcer(G¯)=n+2 (γcer(G)γcer(G¯)=2n, respectively). From this assumption, from Corollary 7.3 and Theorem 7.5 it follows that min{δ(G),δ(G¯)}=1. Then, as we have already proved, γcer(G)=2 or γcer(G¯)=2, and therefore γcer(G¯)=n or γcer(G)=n, respectively. From this and from Theorem 5.3 it follows that G¯ or G is the corona of some graph. Thus, we have proved the implications (a)(c) and (b)(c). Finally, assume that G is the corona of some graph and G is of order n5. Then γcer(G)=n by Theorem 5.3. From the fact that the corona has no isolated vertex, it follows that γcer(G¯)>1. Now, since δ(G)=1, as in Case 2, we get γcer(G¯)=2. Consequently, γcer(G)+γcer(G¯)=n+2 and γcer(G)γcer(G¯)=2n. This proves the implications (c)(a) and (c)(b). □

8 Concluding remarks

Since over the years researchers have published thousands of papers on the topic of domination in graphs, our paper cannot claim the right to cover the new model even partially, it should only be thought of as a very beginning, a small contribution to. In this section, we present three exemplary open problems that we find interesting and which research on we feel worth of being continued.

It is natural to characterise the class of critical graphs where the certified domination number increases on the removal of any edge/vertex as well as the class of stable graphs where the certified domination number remains unchanged on the removal of any edge/vertex. We point out that by Corollary 4.1, the class of critical (resp. stable) (with respect to the certified domination number) graphs with minimum degree δ3 is the same as the class of critical (resp. stable) graphs with respect to domination number, see for example [21–23]. Therefore, we are left with characterising critical (resp. stable) graphs with minimum degree δ2. This is an open problem.

The problem of constructive characterisations of trees with equal domination parameters has received attention in the literature, see for example [24,3,25,26], to mention but a few recent. Following this concept, we leave as an open problem to provide a constructive characterisation of (γ,γcer)-trees, that is, the class of trees with γcer=γ.

Finally, let G be a graph with no isolated vertex. Then no minimal dominating set of G has a shadowed vertex. Consequently, if γcer(G)=γ(G), then none of γcer-sets of G has a shadowed vertex. (In particular, it follows from Corollary 4.1 that if G has no weak support or δ(G)2, then none of γcer-sets of G has a shadowed vertex.) A natural question then is whether the existence of a γcer-set with no shadowed vertex implies the equality of the numbers γcer and γ. The answer to this question is not positive in general. For example, if i is a positive integer and Ti is the tree of order 8i+3 obtained from the corona P2i+1K1 by subdividing each of its edges exactly once, that is, Ti=S(P2i+1K1), then it is a routine exercise to check that γ(Ti)=3i+1, γcer(Ti)=4i+1, and Ti has a γcer-set with no shadowed vertex, see for an illustration. Therefore, we conclude our paper with the problem of characterising all graphs having a minimum certified dominating set with no shadowed vertex.

Fig. 5 Tree Ti=S(P2i+1K1) in which black vertices form a γ-set and a γcer-set, respectively.

Fig. 5 Tree Ti=S(P2i+1∘K1) in which black vertices form a γ-set and a γcer-set, respectively.

Acknowledgement

We would like to thank the referee for the remarkable comments and suggestions that improved the presentation of our results. R. Ziemann was supported by the grant BW 538-5300-B865-15.

References

  • HaynesT.W., HedetniemiS.T., SlaterP.J., Fundamentals of Domination in Graphs 1998Marcel Dekker New York
  • HaynesT.W., HedetniemiS.T., SlaterP.J., Domination in Graphs: Advanced Topics 1998Marcel Dekker New York
  • HenningM.A., RallD.F., On graphs with disjoint dominating and 2-dominating sets, Discuss. Math. Graph Theory, 33 2013 139–146
  • DiestelR., Graph Theory 2012Springer Heidelberg
  • KikunoT., YoshidaN., KakudaY., The NP-completeness of the dominating set problem in cubic planar graphs, IEICE Trans., E63 1980 443–444
  • J. Suomela, Is the dominating set problem restricted to planar bipartite graphs of maximum degree 3 NP-complete? (Posted by Florent Foucaud) Theoretical Computer Science Stack Exchange:http://cstheory.stackexchange.com/questions/2505 [2016-01-08].
  • FinkJ.F., JacobsonM.S., KinchL.F., RobertsJ., On graphs having domination number half their order, Period. Math. Hungar., 16 1985 287–293
  • ToppJ., VestergaardP.D., Well irredundant graphs, Discrete Appl. Math., 63 1995 267–276
  • BalbuenaC., HansbergA., HaynesT.W., HenningM.A., Total domination edge critical graphs with total domination number three and many dominating pairs, Graphs Combin., 31 2015 1163–1176
  • BurgerA.P., de VilliersA.P., van VuurenJ.H., Edge stability in secure graph domination, Discrete Math. Theor. Comput. Sci., 17 2015 103–122
  • ChenX.-G., FujitaS., FuruyaM., SohnM.Y., Constructing connected bicritical graphs with edge-connectivity 2, Discrete Appl. Math., 160 2012 488–493
  • EdwardsM., MacGillivrayG., The diameter of total domination and independent domination vertex-critical graphs, Australas. J. Combin., 52 2012 33–39
  • NordhausE.A., GaddumJ., On complementary graphs, Amer. Math. Monthly, 63 1956 175–177
  • AouchicheM., HansenP., A survey of Nordhaus-Gaddum type relations, Discrete Appl. Math., 161 2013 466–546
  • BorowieckiM., On the external stability number of a graph and its complement, Prace Nauk. Inst. Mat. Politech. Wrocaw., 12 1976 39–43
  • CockayneE.J., HedetniemiS.T., Toward a theory of domination in graphs, Networks, 7 1977 247–261
  • JaegarF., PayanC., Relations du type Nordhaus-Gaddum pour le nombre d’absorption d’un graphe simple, C. R. Acad. Sci. Paris A, 274 1972 728–730
  • LaskarR., PetersK., Vertex and edge domination parameters in graphs, Congr. Numer., 48 1985 291–305
  • DunbarJ.E., HaynesT.W., HedetniemiS.T., Nordhaus-Gaddum bounds for domination sums in graphs with specified minimum degree, Util. Math., 67 2005 97–105
  • JosephJ.P., ArumugamS., Domination in graphs, Int. J. Manage. Syst., 11 1995 177–182
  • BrighamR.C., ChinnP.Z., DuttonR.D., Vertex domination-critical graphs, Networks, 18 1988 173–179
  • FulmanJ., HansonD., MacGillivrayG., Vertex domination-critical graphs, Networks, 25 1995 41–43
  • ManimuthuY., KumarasamyK., Domination Stable Graphs 2012Lambert Academic Publishing Saarbrücken
  • CymanJ., Total outer-connected domination in trees, Discuss. Math. Graph Theory, 30 2010 377–383
  • HouX., A characterization of (2γ,γp)-trees, Discrete Math., 308 2008 3420–3426
  • LuY., HouX., XuJ.-M., LiN., A characterization of (γt,γ2)-trees, Discuss. Math. Graph Theory, 30 2010 425–435