6
Discrete Applied Mathematics 160 (2012) 488–493 Contents lists available at SciVerse ScienceDirect Discrete Applied Mathematics journal homepage: www.elsevier.com/locate/dam Constructing connected bicritical graphs with edge-connectivity 2 Xue-gang Chen a,, Shinya Fujita b , Michitaka Furuya c , Moo Young Sohn d a Department of Mathematics, North China Electric Power University, Beijing 102206, China b Department of Mathematics, Gunma National College of Technology, Maebashi 371-8530, Japan c Department of Mathematical Information Science, Tokyo University of Science 1-3 Kagurazaka, Sinjuku-ku, Tokyo 162-8601, Japan d Department of Mathematics, Changwon National University, Changwon, 641-773, Republic of Korea article info Article history: Received 15 June 2010 Received in revised form 10 March 2011 Accepted 15 March 2011 Available online 22 April 2011 Keywords: Domination number Domination bicritical graph abstract A graph G is said to be bicritical if the removal of any pair of vertices decreases the domination number of G. For a bicritical graph G with the domination number t , we say that G is t -bicritical. Let λ(G) denote the edge-connectivity of G. In [2], Brigham et al. (2005) posed the following question: If G is a connected bicritical graph, is it true that λ(G) 3? In this paper, we give a negative answer toward this question; namely, we give a construction of infinitely many connected t -bicritical graphs with edge-connectivity 2 for every integer t 5. Furthermore, we give some sufficient conditions for a connected 5- bicritical graph to have λ(G) 3. © 2011 Elsevier B.V. All rights reserved. 1. Introduction Graph theory terminology not presented here can be found in [2]. Let G = (V , E ) be a graph with |V |= n. The degree, neighborhood and closed neighborhood of a vertex v in the graph G are denoted by d(v), N (v) and N [v]= N (v) ∪{v}, respectively. The minimum degree and maximum degree of the graph G are denoted by δ(G) and (G), respectively. The graph induced by S V is denoted by G[S ]. Let G denote the complement of G. For any two graphs G and H, let G + H denote the graph obtained from G and H by joining edges from every vertex of G to every vertex of H. A set S is called a dominating set if, for every vertex u V (G) S , there exists v S such that uv E . The domination number of G, denoted by γ(G), is the minimum cardinality of a dominating set of G. A minimum dominating set of a graph G is called a γ(G)-set. For many graph parameters, criticality is a fundamental issue. Brigham et al. [1], Sumner [5], and Sumner and Blitch [6] began the study of graphs where the domination number decreases on the removal of any vertex. Further properties of these graphs were explored in [3,4]. Note that removing a vertex can increase the domination number by more than one, but can decrease it by at most one. It is useful to write the vertex set of a graph as a disjoint union of three sets according to how their removal affects γ(G). Let V (G) = V 0 V + V where V 0 ={v V |γ(G v) = γ(G)}, V + ={v V |γ(G v) > γ (G)} and V ={v V |γ(G v) < γ (G)}. As defined in [1], a vertex v is said to be critical if v V , and a graph G is said to be domination critical if every vertex of G is critical. Brigham et al. [2] gave a generalization of this concept. A graph G is said to be domination (γ , k)-critical, if γ(G S )<γ(G) for any set S of k vertices. Obviously, a domination (γ , k)-critical G has γ(G) 2. In the special case of k = 2, we say that G is domination bicritical, or just call it bicritical. For a bicritical graph G with γ(G) = t , we say that G is t -bicritical. Brigham et al. [2] showed the following results concerning the edge-connectivity λ(G) of a bicritical graph G. Research supported by the Fundamental Research Funds for the Central Universities (10ML39) (to X.C), JSPS Grant 20740068 (to S.F). The fourth author was supported by Changwon National University in 2009–2010. Corresponding author. Fax: +86 010 51963871. E-mail address: [email protected] (X.-g. Chen). 0166-218X/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.dam.2011.03.012

Constructing connected bicritical graphs with edge-connectivity 2

Embed Size (px)

Citation preview

Page 1: Constructing connected bicritical graphs with edge-connectivity 2

Discrete Applied Mathematics 160 (2012) 488–493

Contents lists available at SciVerse ScienceDirect

Discrete Applied Mathematics

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

Constructing connected bicritical graphs with edge-connectivity 2✩

Xue-gang Chen a,∗, Shinya Fujita b, Michitaka Furuya c, Moo Young Sohn d

a Department of Mathematics, North China Electric Power University, Beijing 102206, Chinab Department of Mathematics, Gunma National College of Technology, Maebashi 371-8530, Japanc Department of Mathematical Information Science, Tokyo University of Science 1-3 Kagurazaka, Sinjuku-ku, Tokyo 162-8601, Japand Department of Mathematics, Changwon National University, Changwon, 641-773, Republic of Korea

a r t i c l e i n f o

Article history:Received 15 June 2010Received in revised form 10 March 2011Accepted 15 March 2011Available online 22 April 2011

Keywords:Domination numberDomination bicritical graph

a b s t r a c t

A graph G is said to be bicritical if the removal of any pair of vertices decreases thedomination number of G. For a bicritical graph G with the domination number t , we saythatG is t-bicritical. Let λ(G) denote the edge-connectivity ofG. In [2], Brigham et al. (2005)posed the following question: If G is a connected bicritical graph, is it true that λ(G) ≥ 3?

In this paper, we give a negative answer toward this question; namely, we give aconstruction of infinitely many connected t-bicritical graphs with edge-connectivity 2 forevery integer t ≥ 5. Furthermore, we give some sufficient conditions for a connected 5-bicritical graph to have λ(G) ≥ 3.

© 2011 Elsevier B.V. All rights reserved.

1. Introduction

Graph theory terminology not presented here can be found in [2]. Let G = (V , E) be a graph with |V | = n. The degree,neighborhood and closed neighborhood of a vertex v in the graph G are denoted by d(v),N(v) and N[v] = N(v) ∪ {v},respectively. The minimum degree and maximum degree of the graph G are denoted by δ(G) and ∆(G), respectively. Thegraph induced by S ⊆ V is denoted by G[S]. Let G denote the complement of G. For any two graphs G andH , let G+H denotethe graph obtained from G and H by joining edges from every vertex of G to every vertex of H . A set S is called a dominatingset if, for every vertex u ∈ V (G) − S, there exists v ∈ S such that uv ∈ E. The domination number of G, denoted by γ (G), isthe minimum cardinality of a dominating set of G. A minimum dominating set of a graph G is called a γ (G)-set.

For many graph parameters, criticality is a fundamental issue. Brigham et al. [1], Sumner [5], and Sumner andBlitch [6] began the study of graphs where the domination number decreases on the removal of any vertex. Furtherproperties of these graphs were explored in [3,4]. Note that removing a vertex can increase the domination number bymore than one, but can decrease it by at most one. It is useful to write the vertex set of a graph as a disjoint union of threesets according to how their removal affects γ (G). Let V (G) = V 0

∪ V+∪ V− where V 0

= {v ∈ V |γ (G − v) = γ (G)},V+

= {v ∈ V |γ (G − v) > γ (G)} and V−= {v ∈ V |γ (G − v) < γ (G)}.

As defined in [1], a vertex v is said to be critical if v ∈ V−, and a graph G is said to be domination critical if every vertexof G is critical. Brigham et al. [2] gave a generalization of this concept. A graph G is said to be domination (γ , k)-critical, ifγ (G − S) < γ (G) for any set S of k vertices. Obviously, a domination (γ , k)-critical G has γ (G) ≥ 2. In the special case ofk = 2, we say that G is domination bicritical, or just call it bicritical. For a bicritical graph G with γ (G) = t , we say that G ist-bicritical.

Brigham et al. [2] showed the following results concerning the edge-connectivity λ(G) of a bicritical graph G.

✩ Research supported by the Fundamental Research Funds for the Central Universities (10ML39) (to X.C), JSPS Grant 20740068 (to S.F). The fourth authorwas supported by Changwon National University in 2009–2010.∗ Corresponding author. Fax: +86 010 51963871.

E-mail address: [email protected] (X.-g. Chen).

0166-218X/$ – see front matter© 2011 Elsevier B.V. All rights reserved.doi:10.1016/j.dam.2011.03.012

Page 2: Constructing connected bicritical graphs with edge-connectivity 2

X.-g. Chen et al. / Discrete Applied Mathematics 160 (2012) 488–493 489

Fig. 1. Connected 5-bicritical graph Gwith λ(G) = 2.

Theorem 1. If G is a connected bicritical graph, then δ(G) ≥ 3 and λ(G) ≥ 2.

Theorem 2. If G is a connected graph that is 3-bicritical or 4-bicritical, then λ(G) ≥ 3.

Motivated by the above theorems, they proposed the following open question.

Question 3. If G is a connected bicritical graph, is it true that λ(G) ≥ 3? In particular, if G is a connected 5-bicritical graph, isit true that λ(G) ≥ 3?

In this paper, we give a negative answer toward this question. We give a construction of infinitely many connectedt-bicritical graph with edge-connectivity 2 for every integer t ≥ 5. Moreover, we give some sufficient conditions for aconnected 5-bicritical graph G to have λ(G) ≥ 3 in view of forbidden subgraph conditions andminimum degree conditions.

2. Main results

First,we construct a connected 5-bicritical graphwith edge-connectivity 2. LetX = K5 and setV (X) = {v5, u5, w5, s5, t5}.Let Y ′

1 = Y ′

2 = C10, say Y ′

1 = v1u1w1s1t1v2u2w2s2t2 and Y ′

2 = v3w3t3u3s3v4w4t4u4s4. Let Y1 = Y ′

1 and Y2 = Y ′

2.Let F = (Y1 + Y2) + X . Let Sv = {vi|i = 1, 2, . . . , 5}, Sw = {wi|i = 1, 2, . . . , 5}, Su = {ui|i = 1, 2, . . . , 5},Ss = {si|i = 1, 2, . . . , 5} and St = {ti|i = 1, 2, . . . , 5}. Let H = F \ (E(F [Sv]) ∪ E(F [Su]) ∪ E(F [Sw]) ∪ E(F [Ss]) ∪ E(F [St ])).Let G1 be the graph obtained from H and P2 = ac by joining a to every vertex of Y1 with edges and joining c to every vertexof Y2 with edges. Let G2 be a graph as in Fig. 1.

Let G be the graph obtained from G1 and G2 by joining edges ab and cd. Assume that these vertices are located as in Fig. 1.Here dashed lines in Fig. 1 denote non-edges in Yi for i = 1, 2.

By the construction of G1 and G2, we have the following lemmas.

Lemma 4. γ (G1) = 3, γ (G2) = 2 and γ (G2 − b) = γ (G2 − d) = 2.

Lemma 5. Let G be the graph as in Fig. 1. Then G is a connected 5-bicritical graph with λ(G) = 2.

Proof. By the construction of G, it is obvious that λ(G) = 2. We first prove that γ (G) = 5. Let S be a γ (G)-set. In order todominate V (X), |S ∩ (V (X) ∪ V (Y1) ∪ V (Y2))| ≥ 2. If a, c ∈ S, since S ∩ N[x] = ∅, it follows that |S| ≥ 5. If a ∈ S and c ∈ S,then S ∩ V (G2) is a dominating set of G2 − b. By Lemma 4, |S| ≥ 5. Suppose that a, c ∈ S. If b, d ∈ S, then |S| ≥ 5. If b ∈ Sand d ∈ S, then for any S ′

⊆ N[b], since γ (G2[V (G2) − S ′]) ≥ 2, it follows that |S| ≥ 5. If b, d ∈ S, then S ∩ V (G1) and

S ∩ V (G2) are dominating sets of G1 and G2, respectively. By Lemma 4, |S ∩ V (G1)| ≥ 3 and |S ∩ V (G2)| ≥ 2. This implies|S| ≥ 5, and hence γ (G) ≥ 5. Since {v1, u1, c, x, g} is a dominating set of G, γ (G) ≤ 5. Therefore, γ (G) = 5. In what follows,we will prove that G is a bicritical graph. To see this, for any u, v ∈ V (G), we divide the proof into four cases.

Case 1. {u, v} ∩ (V (Y1) ∪ V (Y2)) = ∅, say v = v1. If u ∈ {x, g}, then {t2, v3, x, g} or {u1, v4, x, g} is a dominating set ofG − {u, v}. If u = x, then {u1, t2, b, d} is a dominating set of G − {u, v}. If u = g , then {u1, t2, f , h} is a dominating set ofG − {u, v}.

Page 3: Constructing connected bicritical graphs with edge-connectivity 2

490 X.-g. Chen et al. / Discrete Applied Mathematics 160 (2012) 488–493

Case 2. {u, v} ∩ (V (Y1) ∪ V (Y2)) = ∅ and {u, v} ∩ V (X) = ∅, say v = v5. If u ∈ {a, c, x}, then {v1, a, c, x} is a dominatingset of G − {u, v}. If u ∈ {a, c}, say u = a, then {v3, w3, x, g} is a dominating set of G − {u, v}. If u = x, then {v1, u1, b, d} is adominating set of G − {u, v}.

Case 3. {u, v} ∩ (V (X) ∪ V (Y1) ∪ V (Y2)) = ∅ and {u, v} ∩ {a, c} = ∅, say v = a. If u ∈ {x, g}, then {v3, w3, x, g} is adominating set of G − {u, v}. If u = x, then {v3, w3, b, d} is a dominating set of G − {u, v}. If u = g , then {v3, w3, f , h} is adominating set of G − {u, v}.

Case 4. v, u ∈ V (G2). Suppose that {u, v} ∩ {b, d} = ∅, say v = d. Then {x, a, v3, w3} or {b, f , v3, w3} is a dominatingset of G − {u, v}. Without loss of generality, we can assume that {u, v} ∩ {b, d} = ∅. If x ∈ {u, v}, then {b, d, v1, u1} is adominating set of G − {u, v}. Hence, we can assume that u, v ∈ N(x). Suppose that {u, v} ∩ {e, t} = ∅, say v = e. If u = f ,then {f , d, v1, u1} is a dominating set of G − {u, v}. If u = f , then {g, d, v1, u1} is a dominating set of G − {u, v}. Hence, wecan assume that {u, v} ⊆ {g, h, f }. Then {e, d, v1, u1} or {t, b, v3, w3} is a dominating set of G − {u, v}.

By Cases 1–4, for any u, v ∈ V (G), there exists a dominating set of G − {u, v} with cardinality at most 4. Hence, G is abicritical graph. So, G is a connected 5-bicritical graph with λ(G) = 2. �

Next we investigate the structure of connected bicritical graphs with edge-connectivity 2. For a connected bicriticalgraph Gwith λ(G) = 2, there is an edge cut {ab, cd} in G. In this paper, we would like to characterize such graphs, althoughit seems difficult in general. As a first step, we are concernedwith such a graph Gwhose component of G−{ab, cd} has smalldomination number. We give the following lemmas.

Lemma 6. Let G be a graph with γ (G) ≥ 2. If V (G) = A ∪ B, A ∩ B = ∅ and |A| ≥ 4, then there exist A1 ⊆ A and B1 ⊆ B suchthat |A1| = |A| − 2 and γ (G[A1 ∪ B1]) ≥ 2.

Proof. Let v1 ∈ A. Since γ (G) ≥ 2, there exists a vertex u1 such that v1u1 ∈ E(G). Let S0 = ∅ and S1 = {v1, u1}.Let v2 ∈ A − (A ∩ S1). If γ (G[S1 ∪ {v2}]) ≥ 2, then we let S2 = S1 ∪ {v2}. If γ (G[S1 ∪ {v2}]) = 1, then there exists

u2 ∈ V (G) − S1 such that v2u2 ∈ E(G). Then we let S2 = S1 ∪ {v2, u2}. Choose v3 ∈ A − (A ∩ S2). If γ (G[S2 ∪ {v3}]) ≥ 2,then we let S3 = S2 ∪ {v3}. If γ (G[S2 ∪ {v3}]) = 1, then there exists u3 ∈ V (G) − S2 such that v3u3 ∈ E(G). Now letS3 = S2 ∪ {v3, u3}. Repeat this process to obtain a sequence {S1, S2, . . . , Sk} such that γ (G[Si]) ≥ 2 for i = 1, 2, . . . , k and|A−(A∩Sk)| = 2, 3. If |A−(A∩Sk)| = 2, thenwe let A1 = A∩Sk and B1 = B∩Sk, as desired. Suppose that |A−(A∩Sk)| = 3.Let u, v, w ∈ A− (A∩ Sk). If there exists a vertex, say w, such that w does not dominate Sk, then we let A1 = (A∩ Sk) ∪ {w}

and B1 = B ∩ Sk, as desired. If there exists a vertex, say w, such that w does not dominate a vertex of B − B ∩ Sk, say t , thenwe let A1 = (A∩Sk)∪{w} and B1 = (B∩Sk)∪{t}, as desired. Suppose that each vertex of A− (A∩Sk) dominates Sk ∪B. Thenγ (G[{u, v, w}]) ≥ 2, say uv ∈ E(G). If |A ∩ Sk| − |A ∩ Sk−1| = 2, then we let A1 = (A ∩ Sk−1) ∪ {v, w, u} and B1 = B ∩ Sk−1.Then the result holds.

Suppose that |A ∩ Sk| − |A ∩ Sk−1| = 1. Let A1 = (A ∩ Sk−1) ∪ {v, u} and B1 = B ∩ Sk−1. Then the result holds. �

By Lemma 6, we have the following corollary.

Corollary 7. Let G be a graph with γ (G) ≥ 2 and n ≥ 4. Then there exists a subset S ⊆ V (G) such that |S| = n − 2 andγ (G[S]) ≥ 2.

Lemma 8 (Brigham et al. [2]). Suppose that G is a connected bicritical graph with λ(G) = 2 and an edge cut {ab, cd}. Let G1 andG2 be the two components of G − ab − cd, with a, c ∈ V (G1), b, d ∈ V (G2) and a = c. Then all of the following must be true.

(i) γ (G) = γ (G1) + γ (G2).(ii) a, c ∈ V+(G1) and b, d ∈ V+(G2).(iii) b = d.(iv) Without loss of generality, a, c ∈ V−(G1) and b, d ∈ V 0(G2).(v) Neither b nor d is in a γ (G2)-set.(vi) γ (G2 − {b, d}) = γ (G2) − 1 and a γ (G2 − {b, d})-set dominates neither b nor d.(vii) There is a γ (G2 − d)-set containing b, and there is a γ (G2 − b)-set containing d.(viii) There is no γ (G1)-set containing both a and c.(ix) There is no γ (G1 − a)-set containing c, and there is no γ (G1 − c)-set containing a.(x) γ (G1) ≥ 3.

Let G21 = G2,G22 = G2 + gf ,G23 = G2 + ge, G24 = G22 + gt,G25 = G22 + gh and G26 = G23 + gt . LetG27 = G21 + et,G28 = G27 + gf and G29 = G28 + gt . Notice that by Lemma 8(vi), we have γ (G2) ≥ 2. Now we provethe following theorem.

Theorem 9. Let G be a connected t-bicritical graph with λ(G) = 2, and let {ab, cd} be an edge cut of G. Suppose that there existsa connected component G2 of G − ab − cd such that γ (G2) = 2. Then G2 is isomorphic to G2i for some i with 1 ≤ i ≤ 9.

Page 4: Constructing connected bicritical graphs with edge-connectivity 2

X.-g. Chen et al. / Discrete Applied Mathematics 160 (2012) 488–493 491

Proof. Suppose that G1 and G2 are the two components of G − ab − cd, with a, c ∈ V (G1), b, d ∈ V (G2) and γ (G2) = 2. ByLemma 8(vi) and (x), γ (G1) = t −2. Then γ (G2 −{b, d}) = 1. By Lemma 8(v) and (vi), bd ∈ E(G). Let {x} be a γ (G2 −{b, d})-set. Let A1 = {v|v ∈ N(x) ∩ N(b), v ∈ N(d)}, A2 = {v|v ∈ N(x) ∩ N(b) ∩ N(d)}, A3 = {v|v ∈ N(x) ∩ N(d), v ∈ N(b)}and A4 = {v|v ∈ N(x), v ∈ N(b) ∪ N(d)}. By Lemma 8(v), note that N(x) ∩ {b, d} = ∅. Hence, it follows thatV (G2) = {b, d, x} ∪ A1 ∪ A2 ∪ A3 ∪ A4.

Claim 1. For any u, v ∈ N(x), let D be a γ (G − {u, v})-set. Let D1 = D ∩ V (G1) and D2 = D ∩ V (G2 − {u, v}).

(1) If |D2| = 2, then D2 ∩ {b, d} = ∅ and D2 dominates both b and d.(2) If |D2| = 1, then D2 dominates at least one vertex of b and d.(3) If u, v ∈ N(x) ∩ N(b) and |D2| = 1, then D2 ∩ N(b) = ∅.(4) If u, v ∈ N(x) ∩ N(b) and |D2| = 2, then D2 = {w, d}, where w ∈ (N(b) ∩ N(x)) − {u, v}.

Proof. Since G is bicritical, |D| = γ (G − {u, v}) ≤ γ (G) − 1 = t − 1.

(1) If D2 ∩ {b, d} = ∅, then D1 is a dominating set of G1. Hence, |D1| ≥ γ (G1) = t − 2. So, |D| = |D1| + |D2| ≥ t , which isa contradiction. Hence, D2 ∩ {b, d} = ∅, say d ∈ D2. Suppose that b is not dominated by D2. Then b is dominated by D1.So, a ∈ D1. Then D1 ∪ {c} is a dominating set of G1. Since there is no γ (G1)-set containing both a and c , it follows that|D1| ≥ t − 2. Hence, |D| = |D1| + |D2| ≥ t , which is a contradiction. So, D2 dominates both b and d.

(2) Suppose that |D2| = 1. If D2 dominates neither b nor d, then D1 is a dominating set of G1, and D1 dominates bothb and d. Then a, c ∈ D1. Since there is no γ (G1)-set containing both a and c , it follows that |D1| > t − 2. Hence,|D| = |D1| + |D2| ≥ t , which is a contradiction. So, D2 dominates at least one vertex of b and d.

(3) Let D2 = {w}. By (2), it follows that w ∈ A1 ∪ A2 ∪ A3. If w ∈ A3, then w dominates A3 ∪ A4. So, {w, b} is a γ (G2)-setcontaining b, which contradicts Lemma 8(v). Hence, w ∈ A1 ∪ A2. This shows w ∈ N(b).

(4) Since bu, bv ∈ E(G), b ∈ D2. By (1), D2 = {w, d}, where w ∈ (N(b) ∩ N(x)) − {u, v}. �

By the symmetry of the roles of b and d, Claim 1(3), (4) holds if we interchange bwith d.

Claim 2. |A2| ≤ 1.

Proof. Suppose that |A2| ≥ 2. Let u, v ∈ A2. Let D be a γ (G − {u, v})-set and D2 = D ∩ V (G2 − {u, v}). Then |D| ≤ t − 1.Since u, v ∈ N(x) ∩ N(b) ∩ N(d), then x, b, d ∈ D2. By Claim 1, |D2| = 1,D2 ∩ N(b) = ∅ and D2 ∩ N(d) = ∅, say D2 = {w}.Then it follows that w ∈ A2 and w dominates N[x] − {u, v}. Hence {w, b} is a γ (G2)-set containing b, which contradictsLemma 8(v). �

Claim 3. If |A2| = 1, then A4 = ∅.

Proof. Suppose that A4 = ∅. Let u ∈ A2. Let D be a γ (G − {u, x})-set and D2 = D ∩ V (G2 − {u, x}). In order to dominateA4,D ∩ (N(x) − {u}) = ∅, say w ∈ D ∩ (N(x) − {u}). Since wx, bu, du ∈ E(G), b, d ∈ D. Otherwise, D is a dominating setof G, which is a contradiction. Then |D ∩ V (G1)| ≥ t − 2. Hence, |D2| = 1. This shows D2 = {w}. Since D dominates {b, d},by Lemma 8(viii), it follows that w ∈ (A1 ∪ A3). Then {w, d} or {w, b} is a γ (G2)-set, which contradicts Lemma 8(v). So wehave A4 = ∅. �

Claim 4. If |A2| = 1 and |A1| ≥ 3, then for any v ∈ A1, there exists a vertex w ∈ A1 − {v} such that w dominates A1 − {v} andvw ∈ E(G).

Proof. Letu ∈ A2, and letDbe aγ (G−{u, v})-set. Since xu, xv, bu, bv ∈ E(G), it follows that b, x ∈ D2 andD2∩(N(x)−{u}) =

∅. Suppose that |D2| = 1, say D2 = {w}. By Claim 1, w ∈ A1 − {v} and w dominates A1 − {v}. If vw ∈ E(G), then {w, d} is aγ (G2)-set containing d, which contradicts Lemma 8(v). Hence vw ∈ E(G). Suppose that |D2| = 2. By Claim 1, D2 = {w, d},where w ∈ A1 − {v}. Then w dominates A1 − {v}. If vw ∈ E(G), then {w, d} is a γ (G2)-set containing d, which contradictsLemma 8(v). Hence vw ∈ E(G). �

By Claims 2 and 3, we divide the proof into the following cases.Case 1. |A2| = 1. By Claim 3, A4 = ∅. Let A2 = {u}. If γ (G[A1]) = 1, then there exists a γ (G2)-set containing d, which is

a contradiction. Hence γ (G[A1]) ≥ 2. Similarly, γ (G[A3]) ≥ 2. That is |A1| ≥ 2 and |A3| ≥ 2. Let A1 = {v1, . . . , v|A1|} andA3 = {w1, . . . , w|A3|}. We further divide the proof into the following subcases.

Case1.1. |A1| ≥ 4 is even. By Claim4,we can assume that vivi+1 ∈ E(G) for i = 1, 3, . . . , |A1|−1. LetDbe aγ (G−{v1, v2})-set. Since vivi+1 ∈ E(G) for i = 3, . . . , |A1| − 1, by Claim 1, it follows that u ∈ D and u dominates v3 and v4. Let D′ be aγ (G − {v3, v4})-set. Arguing similarly, we have u ∈ D′. Then D′ is a dominating set of G, which is a contradiction.

Case 1.2. |A1| ≥ 3 is odd. By Claim 4, we can assume that vivi+1 ∈ E(G) for i = 1, 3, . . . , |A1| − 2. Let v|A1| ∈

A1 − {v1, . . . , v|A1|−1}. Let D be a γ (G − {u, v|A1|})-set. By Claim 1, there exists a vertex vi ∈ {v1, . . . , v|A1|−1} such thatvi dominates A1 − {v|A1|}, which is a contradiction.

By Cases 1.1 and 1.2, without loss of generality, we may assume that |A1| = 2 and |A3| = 2.Case 1.3. |A1| = 2 and |A3| = 2. Since γ (G[A1]) ≥ 2 and γ (G[A3]) ≥ 2, v1v2 ∈ E(G) and w1w2 ∈ E(G). Let D be a

γ (G − {b, x})-set. Since (A1 ∪ A2) ∩ D2 = ∅, we can assume that D2 ⊆ {w1, d}. Then w1v1, w1v2 ∈ E(G). Arguing similarly,

Page 5: Constructing connected bicritical graphs with edge-connectivity 2

492 X.-g. Chen et al. / Discrete Applied Mathematics 160 (2012) 488–493

we may assume that v1w1, v1w2 ∈ E(G). If u is adjacent to both v1 and v2, then {u, d} is a dominating set of G2 containingd, which is a contradiction. So, u is adjacent to at most one vertex in {v1, v2}. Similarly, u is adjacent to at most one vertexin {w1, w2}. Then there exists an i such that G2 is isomorphic to G2i.

Case 2. A2 = ∅ and A4 = ∅. By Theorem 1 and Claim 1, |A1| ≥ 3 and |A3| ≥ 3. Suppose that |A1| ≥ 4 or |A3| ≥ 4, say|A1| ≥ 4. Since no vertex of A1 dominates A1 ∪ A4, arguing similarly as in the proof of Lemma 6, there exists B1 ⊆ A1 andB4 ⊆ A4 such that γ (G[B1 ∪ B4]) ≥ 2 and |A1 − B1| = 2, say A1 − B1 = {u, v}. Let D be a γ (G− {u, v})-set. By Claim 1, thereexists a vertex w ∈ A1 such that w dominates (A1 ∪ A4) − {u, v}, which is a contradiction.

Suppose that |A1| = |A3| = 3. If there exists a vertex of A1 such that it does not dominate A4, arguing similarly as above,we get a contradiction. So we can assume that every vertex of A1 dominates A4. Then γ (G[A1]) ≥ 2 and γ (G[A3]) ≥ 2, sayAi = {ui, vi, wi} and uivi ∈ E(G) for i = 1, 3. Let D be a γ (G − {w1, w3})-set. By Claim 1, there exists w ∈ D such thatw ∈ (A1 ∪ A3) − {w1, w3} and w dominates (A1 ∪ A3) − {w1, w3}, which is a contradiction.

Case 3. A2 = A4 = ∅. By Theorem 1, Claim 1 and Lemma 8(v), |A1| ≥ 3, |A3| ≥ 3, γ (G[A1]) ≥ 2 and γ (G[A3]) ≥ 2.If |A1| ≥ 4 or |A3| ≥ 4, say |A1| ≥ 4, by Corollary 7, there exists B1 ⊆ A1 such that γ (G[B1)] ≥ 2 and |A1 − B1| = 2,say A1 − B1 = {u, v}. Let D be a γ (G − {u, v})-set. Arguing similarly as in Case 2, we get a contradiction. Suppose that|A1| = |A3| = 3, say Ai = {ui, vi, wi} and uivi ∈ E(G) for i = 1, 3. Let D be a γ (G − {w1, w3})-set. Arguing similarly as inCase 2, we get a contradiction. �

Utilizing Theorem 9, we can easily get the following sufficient condition for a connected 5-bicritical graph G to satisfyλ(G) ≥ 3 in view of forbidden subgraph conditions and minimum degree conditions.

Corollary 10. Let G be a connected 5-bicritical graph. If G satisfies one of the following conditions, we have λ(G) ≥ 3.(i) G does not contain G2i as an induced subgraph for each i with 1 ≤ i ≤ 9.(ii) δ(G) ≥ 5.

Proof. SupposeG is a connected5-bicritical graphwithλ(G) = 2. Let {ab, cd}be an edge cut ofG. LetG1,G2 be two connectedcomponents of G − ab − cd with γ (G1) ≥ γ (G2). Then in view of Lemma 8, we have γ (G1) = 3, γ (G2) = 2. However, if Gsatisfies one of the conditions (i) and (ii), it is easy to see that we cannot have a structure described as in Theorem 9 for G2.This is a contradiction. �

Also, with a slight change of G in Fig. 1, we can construct a connected 5-bicritical graph with arbitrary many vertices.To see this, for any integer z ≥ 1, let C25z−10 = v3w3t3u3s3v4w4t4u4s4, . . . , v5zw5z t5zu5zs5zv3. Let X ′

= K5 and setV (X ′) = {v5z+1, u5z+1, w5z+1, s5z+1, t5z+1}, and reset Sv = {vi|i = 1, 2, . . . , 5z}, Sw = {wi|i = 1, 2, . . . , 5z}, Su = {ui|i =

1, 2, . . . , 5z}, Ss = {si|i = 1, 2, . . . , 5z}, St = {ti|i = 1, 2, . . . , 5z}.Let G′ be the graph obtained from G by replacing Y2 and X by C25z−10 and X ′, respectively (also use new sets Sv, . . . , St

re-defined above). Then, arguing similarly as in the proof of Lemma 5, we have the following.

Lemma 11. The graph G′ is a connected 5-bicritical graph with λ(G) = 2.

In what follows, we construct a connected l-bicritical graph with edge-connectivity exactly 2 for any l with l > 6. Let Fbe the connected t-critical and t-bicritical graph where t ≥ 3. According to Brighams’ paper [2], such a graph is obtained(see Proposition 2 in [2]). Let y be a vertex in F . Let G be the graph as in Fig. 1 and let x be the vertex as in Fig. 1. LetG′′

= (F ◦ G)(y, x : w). (Here this graph means the graph obtained from F and G by identifying y and x. Let w be theidentified vertex in G′′.)

Lemma 12. For t ≥ 3, the graph G′′ is a connected (t + 4)-bicritical graph with λ(G′′) = 2.

Proof. It is obvious that λ(G′′) = 2. By the construction of G, x is a critical vertex of G. We first prove that γ (G) = t + 4. LetSG be a γ (G)-set containing x, and let SF be a γ (F)-set containing y. Then (SG ∪ SF ∪ {w}) \ {x, y} is a dominating set of G′′.Hence, γ (G′′) ≤ |(SG ∪ SF ∪ {w}) \ {x, y}| = γ (G) + γ (F) − 1 = t + 4.

Let S be a γ (G′′)-set. Ifw ∈ S, then (S∪{x}\{w})∩V (G) is a dominating set ofG and (S∪{y}\{w})∩V (F) is a dominating setof F . So,γ (G′′) = |S| = |(S∪{x}\{w})∩V (G)|+|(S∪{y}\{w})∩V (F)|−1 ≥ γ (G)+γ (F)−1. Suppose thatw ∈ S. Then S∩V (G)is a dominating set of G and S ∩V (F) is a dominating set of F . So, γ (G′′) = |S| = |S ∩V (G)|+ |S ∩V (F)| ≥ γ (G)+γ (F)−1.Hence, γ (G′′) = γ (G) + γ (F) − 1 = t + 4.

In what follows, we will prove that G′′ is a bicritical graph. For any u, v ∈ V (G′′), we divide the proof into the followingcases.

Case 1. v, u ∈ (V (F) \ {y})∪{w}. Let SF be a γ (F −v −u)-set and SG be a γ (G− x)-set. Then SF ∪ SG is a dominating set ofG′′

−u−v. Since F is a bicritical graph, γ (F −v−u) ≤ γ (F)−1. So, γ (G′′−u−v) ≤ |SF ∪SG| ≤ (γ (F)−1)+ (γ (G)−1) =

γ (G′′) − 1.Case 2. v, u ∈ (V (G)\ {x})∪{w}. Let SF be a γ (F − y)-set and SG be a γ (G−u−v)-set. Then SF ∪ SG is a dominating set of

G′′−u−v. Since F is a critical graph andG is a bicritical graph, it follows that γ (F−y) ≤ γ (F)−1 and γ (G−u−v) ≤ γ (G)−1.

So, γ (G′′− u − v) ≤ |SF ∪ SG| ≤ (γ (F) − 1) + (γ (G) − 1) = γ (G′′) − 1.

Case 3. v ∈ V (F)\{y} and u ∈ V (G)\{x}. Let SF be a γ (F−v)-set and SG be a γ (G−x−u)-set. Then SF∪SG is a dominating setofG′′

−u−v. Since F is a critical graph,γ (F−v) ≤ γ (F)−1. So,γ (G′′−u−v) ≤ |SF∪SG| ≤ (γ (F)−1)+(γ (G)−1) = γ (G′′)−1.

Hence, G′′ is a bicritical graph with λ(G′′) = 2. �

Page 6: Constructing connected bicritical graphs with edge-connectivity 2

X.-g. Chen et al. / Discrete Applied Mathematics 160 (2012) 488–493 493

Finally, we construct such an example for t = 6. Consider the graph G as in Fig. 1. Now we construct a connected 6-bicritical graph from G−V (X) by adding some vertices. Take three vertices p, q, r from Y1 ∪Y2 such that {p, q, r} dominatesV (Y1) ∪ V (Y2) ∪ {a, c}. For such a triple {p, q, r}, we introduce a new vertex v{p,q,r}. For the vertex v{p,q,r}, join v{p,q,r}to (V (Y1) ∪ V (Y2)) \ {p, q, r} completely with edges. Define those new vertices for all the triples satisfying the aboveproperties, and repeat this operation to all the new vertices. Let G∗ be the resulting graph. Let G′

1 = G∗− V (G2) and

X ′= {v{p,q,r}|p, q, r ∈ V (Y1 ∪ Y2), {p, q, r} dominates V (Y1 ∪ Y2) ∪ {a, c}}. By the construction, we can easily show the

following.

Lemma 13. For any two vertices y1, y2 ∈ V (Y1)∪V (Y2), there exist two vertex y3, y4 ∈ V (Y1)∪V (Y2) such that each {y1, y2, yj}with j = 3, 4 dominates V (Y1) ∪ V (Y2) ∪ {a, c}.

Lemma 14. The graph G′

1 satisfies γ (G′

1) = 4.

Proof. Since {v1, u1, w1, c} is a dominating set of G′

1, γ (G′

1) ≤ 4. Suppose γ (G′

1) ≤ 3, and let S be a γ (G′

1)-set. ByLemma 13 and |V (X ′)| ≥ 3, since there exists no set of two vertices in V (Y1) ∪ V (Y2) ∪ V (X ′) which dominatesV (X ′), S ∩ ({a, c} ∪ V (X ′)) = φ. S dominates V (Y1) ∪ V (Y2) ∪ {a, c}, and hence there exists a vertex in X ′ not dominated byS, which is a contradiction. �

Lemma 15. The graph G∗ is a connected 6-bicritical with λ(G∗) = 2.

Proof. By the construction of G∗, it is obvious that λ(G∗) = 2.We first prove that γ (G∗) = 6. Let S be a γ (G∗)-set. In order todominate V (X ′), |S∩ (V (X ′)∪V (Y1)∪V (Y2))| ≥ 3. If a, c ∈ S, then we see from S∩N[x] = ∅ that |S| ≥ 6. If |{a, c}∩S| = 1,then S∩V (G2) is a dominating set of G2 −b or G2 −d. By Lemma 4, it follows that |S| ≥ 6. Thus wemay assume that a, c ∈ S.If b, d ∈ S, then |S| ≥ 6. If b ∈ S and d ∈ S, then for any S ′

⊆ N[b], since γ (G2[V (G2) − S ′]) ≥ 2, it follows that |S| ≥ 6. If

b, d ∈ S, then S ∩ V (G′

1) and S ∩ V (G2) are dominating sets of G′

1 and G2, respectively. By Lemmas 4 and 14, |S ∩ V (G′

1)| ≥ 4and |S∩V (G2)| ≥ 2. Hence |S| ≥ 6. Thus we have γ (G∗) ≥ 6. Since {v1, u1, w1, c, x, g} is a dominating set of G∗, γ (G∗) ≤ 6.Therefore, γ (G∗) = 6. In the following, we will prove that G∗ is a bicritical graph. For any u, v ∈ V (G∗), we divide the proofinto the following cases.

Case 1. {u, v}∩(V (Y1)∪V (Y2)) = ∅, say v = v1. If u ∈ {x, g}, then {u1, t2, v3, x, g}, {u1, t2, v4, x, g}, or {v2, v3, v4, x, g} isa dominating set of G∗

−{u, v}. If u = x, then {u1, t2, v4, b, d} is a dominating set of G∗−{u, v}. If u = g , then {u1, t2, v4, f , h}

is a dominating set of G∗− {u, v}.

Case 2. {u, v}∩(V (Y1)∪V (Y2)) = ∅ and {u, v}∩V (X ′) = ∅, say v ∈ V (X ′). Then v = vp,q,r for some p, q, r ∈ V (Y1)∪V (Y2).If u ∈ {x, g}, then {p, q, r, x, g} is a dominating set ofG∗

−{u, v}. If u = x, then {p, q, r, b, d} is a dominating set ofG∗−{u, v}.

If u = g , then {p, q, r, f , h} is a dominating set of G∗− {u, v}.

Case 3. {u, v} ∩ (V (X ′) ∪ V (Y1) ∪ V (Y2)) = ∅ and {u, v} ∩ {a, c} = ∅, say v = a. If u ∈ {x, g}, then {v3, w3, t3, x, g} is adominating set of G∗

−{u, v}. If u = x, then {v3, w3, t3, b, d} is a dominating set of G∗−{u, v}. If u = g , then {v3, w3, t3, f , h}

is a dominating set of G∗− {u, v}.

Case 4. v, u ∈ V (G2). Suppose that {u, v}∩{b, d} = ∅, say v = d. Then {x, a, u3, w3, t3} or {b, f , u3, w3, t3} is a dominatingset of G∗

− {u, v}. Without loss of generality, we can assume that {u, v} ∩ {b, d} = ∅. If x ∈ {u, v}, then {b, d, v1, u1, w1} is adominating set of G∗

− {u, v}. Hence, we can assume that u, v ∈ N(x). Suppose that {u, v} ∩ {e, t} = ∅, say v = e. If u = f ,then {f , d, v1, u1, w1} is a dominating set of G∗

− {u, v}. If u = f , then {g, d, v1, u1, w1} is a dominating set of G∗− {u, v}.

Hence, we can assume that {u, v} ⊆ {g, h, f }. Then {e, d, v1, u1, w1} or {t, b, v3, w3, t3} is a dominating set of G∗− {u, v}.

By Cases 1–4, for any u, v ∈ V (G∗), there exists a dominating set of G∗− {u, v} with cardinality at most 5. Hence G∗ is a

bicritical graph. So G∗ is a connected 6-bicritical graph with λ(G∗) = 2. �

Notice that we can also construct such 6-bicritical graphs with arbitrary many vertices in a similar manner as observedin connected 5-bicritical graphs in the previous argument. Combining Lemmas 11, 12 and 15, we finally obtain the followingresult.

Theorem 16. Let t be an integer with t ≥ 5. There exist infinitely many connected t-bicritical graphs G with λ(G) = 2.

References

[1] R.C. Brigham, P.Z. Chinn, R.D. Dutton, Vertex domination-critical graphs, Networks 18 (1988) 173–179.[2] R.C. Brigham, T.W. Haynes, M.A. Henning, D.F. Rall, Bicritical domination, Discrete Math. 305 (2005) 18–32.[3] J. Carrington, F. Harary, T.W. Haynes, Changing and unchanging the domination number of a graph, J. Combin. Math. Combin. Comput. 9 (1991) 57–63.[4] O. Favaron, D. Sumner, E. Wojcicka, The diameter of domination-critical graphs, J. Graph Theory 18 (1994) 723–734.[5] D.P. Sumner, Critical concepts in domination, Discrete Math. 86 (1990) 33–46.[6] D.P. Sumner, P. Blitch, Domination critical graphs, J. Combin. Theory Ser. B 34 (1983) 65–76.