Independence Number and Connectivity of Maximal Connected Domination Vertex Critical Graphs
Norah Almalki1 and Pawaton Kaemawicharnurat2,3
1Department of Mathematics and Statistics
College of Science, Taif University, Saudi Arabia
Email: [email protected]
2Department of Mathematics, Faculty of Science
King Mongkut’s University of Technology Thonburi, Thailand
3Mathematics and Statistics with Applications (MaSA)
Bangkok, Thailand
Email: [email protected]
Abstract
A graph G is said to be k-γc-edge critical if the connected domination number γc(G)=k and γc(G+uv)<k for every uv∈E(G). A 2-connected graph G is said to be k-γc-vertex critical if γc(G)=k and γc(G−v)<k for any v∈V(G). A maximal k-γc-vertex critical graph is a graph which are both k-γc-edge critical and k-γc-vertex critical. Let δ,κ and α be respectively the minimum degree, the connectivity and the independence number. In this paper, we prove that if G is a maximal 3-γc-vertex critical graph satisfying α=κ, then the equation κ=δ still holds. We provide a class of maximal 3-γc-vertex critical graphs with α<κ and κ<δ. We finally show that every two vertices of any 3-connected maximal 3-γc-vertex critical graph are joined by hamiltonian path when κ<δ.
**Keywords: connected domination, independence number, connectivity.
AMS subject classification:05C69, 05C40**
1 Introduction
Our basic graph theoretic notation and terminology follows that of Bondy and Murty[3]. Thus G denotes a finite graph with vertex set V(G) and edge set E(G). For S⊆V(G), G[S] denotes the subgraph of G induced by S. Throughout this paper all graphs are simple and connected. The open neighborhood NG(v) of a vertex v in G is {u∈V(G):uv∈E(G)}. Further, the closed neighborhood NG[v] of a vertex v in G is NG(v)∪{v}. The degree degG(v) of a vertex v in G is ∣NG(v)∣. The minimum degree of a graph G is denoted by δ(G). NS(v) denotes NG(v)∩S where S is a vertex subset of G. A tree is a connected graph contains no cycle. A star K1,n is a tree containing n vertices of degree 1 and exactly one vertex of degree n. An independent set is a set of pairwise non-adjacent vertices. The independence number α(G) is the maximum cardinality of an independent set. For a connected graph G, a cut set is a vertex subset S⊆V(G) such that G−S is connected. The minimum cardinality of a vertex cut set of a graph G is called the connectivity and is denoted by κ(G). If G has S={a} as a minimum cut set, then a is a cut vertex of G and κ(G)=1. A graph G is s-connected if κ(G)≥s. When ambiguity occur, we abbreviate δ(G),α(G) and κ(G) to δ,α and κ, respectively. A hamiltonian path is a path containing every vertex of a graph. A graph G is hamiltonian connected if every two vertices of G are joined by a hamiltonian path. Obviously, a graph of connectivity one is not hamiltonian connected. Moreover, a graph of connectivity two does not contain a hamiltonian path between the two vertices in a minimum cut set. Thus we always focus on 3-connected graphs when we study on the hamiltonian connected property of graphs. For a graph G, the Mycielskian μ(G) of G is the graph with vertex set V(G)∪V′∪{x} where V′={u′∣u∈V(G)} and edge set E(G)∪{uv′∣uv∈E(G)}∪{v′x∣v′∈V′}.
For subsets D,X⊆V(G), D dominates X if every vertex in X is either in D or adjacent to a vertex in D. If D dominates X, then we write D≻X, further, we write a≻X when D={a}. If X=V(G), then D is a dominating set of G and we write D≻G instead of D≻V(G). A connected dominating set of a graph G is a dominating set D of G such that G[D] is connected. If D is a connected dominating set of G, we then write D≻cG. A smallest connected dominating set is call a γc-set. The cardinality of a γc-set of G is called the connected domination number of G and is denoted by γc(G). A total dominating set of a graph G is a subset D of V(G) such that every vertex in G is adjacent to a vertex in D. The minimum cardinality of a total dominating set of G is called the total domination number of G and is denoted by γt(G).
A graph G is k-γc-edge critical if γc(G)=k and γc(G+uv)<k for each pair of non-adjacent vertices u and v of G. A graph G is k-γc-vertex critical if γc(G)=k and γc(G−v)<k for any vertex v of G. A graph G is a maximal k-γc-vertex critical if G is both k-γc-edge critical and k-γc-vertex critical. It is easy to see that a disconnected graph cannot contain a connected dominating set. Thus if a graph G contains a cut vertex c, then G−c has no connected dominating set. Hence, in the following, we always assume that k-γc-vertex critical graphs are 2-connected. A k-γt-edge critical graph is similarly defined.
This paper focuses on the relationship of the connectivity and the independence number of 3-γc-edge critical graphs. For related results on 3-γ-edge critical graphs, Zhang and Tian[11] showed that the independence number of every 3-γ-edge critical graph does not exceed κ+2 and, moreover if α=κ+2, then κ=δ. They used this result to established, the short proof, that every 3-γ-edge critical graph contains a hamiltonian path. Further, Kaemawichanurat and Caccetta [6] showed that every 3-γc-edge critical graph satisfies α≤κ+2. Moreover, if G is a 3-γc-edge critical graph such that κ+1≤α≤κ+2, then κ=δ with only one exception. They also showed that if κ≥3, then α=κ+p if and only if α=δ+p for all p∈{1,2}. The condition κ+1≤α≤κ+2 is best possible to establish that κ=δ.
In this paper, we prove that if a 3-γc-edge critical graph G is 3-γc-vertex critical (G is maximal 3-γc-vertex critical) satisfying α=κ, then κ=δ. We provide a class of maximal 3-γc-vertex critical graphs with α<κ and κ<δ. Hence, the condition α=κ is needed to prove that κ=δ in maximal 3-γc-vertex critical graphs. We finish this work by showing that every 3-connected maximal 3-γc-vertex critical graph are hamiltonian connected if κ<δ.
2 Preliminaries
In this section, we state a number of results that we make use of in establishing our theorems. We begin with a result of Chvátal and Erdös[5] which gives hamiltonian properties of graphs according to the connectivity and the independence number.
Theorem 1
[5]**
Let G be an s-connected graph containing no independent set of s vertices. Then G is hamiltonian connected.
On 3-γc-edge critical graphs, Ananchuen[1] characterized such graphs with a cut vertex. For positive integers ni and r≥2, let H=∪i=1rK1,ni. For 1≤j≤r, let cj be the center of K1,nj in H and w1j,w2j,...,wnjj the end vertices of K1,nj in H. The graphs G1 and G2 are defined as follows. Set V(G1)=V(H)∪{x,y} and E(G1)=E(H)∪{xy}∪{xwij:1≤i≤nj and 1≤j≤r}. Set V(G2)=V(H)∪{x,y}∪U where ∣U∣≥1 and E(G2)=E(H)∪{xy}∪{xwij:1≤i≤nj and 1≤j≤r}∪{uz:u∈U and z∈V(G1)−{x,y,u}}.
The following two lemmas, observed by Chen et al.[4], give fundamental properties of 3-γc-edge critical graphs.
Lemma 1
[4]**
Let G be a 3-γc-edge critical graph and, for a pair of non-adjacent vertices u and v of G, let Duv be a γc-set of G+uv. Then
- (1)
∣Duv∣=2,
2. (2)
Duv∩{u,v}=∅* and*
3. (3)
if u∈Duv and v∈/Duv, then NG(v)∩Duv=∅.
Lemma 2
[4]**
Let G be a 3-γc-edge critical graph and I an independent set with ∣I∣=p≥3. Then the vertices in I can be ordered as a1,a2,...,ap and there exists a path x1,x2,...,xp−1 in G−I with {ai,xi}≻cG+aiai+1 for 1≤i≤p−1.
On 3-γc-vertex critical graphs, Ananchuen et al.[2] provided fundamental properties of such graphs.
Lemma 3
[2]**
Let G be a 3-γc-vertex critical graph and, for a vertex v of G, let Dv be a γc-set of G−v. Then
- (1)
∣Dv∣=2* and*
2. (2)
Dv∩NG[v]=∅.
In the study of 3-γt-edge critical graphs, Simmons[10] proved that such graphs satisfy α≤δ+2. It was pointed out in Ananchuen[1] that a 3-γc-edge critical graph is also 3-γt-edge critical and vice versa. Then every 3-γc-edge critical graph satisfies α≤δ+2. In [7], we established the result of 3-γc-edge critical graphs when α=δ+2.
Theorem 2
([10], [7])
Let G be a 3-γc-edge critical graph with δ≥2. Then α≤δ+2. Moreover if α=δ+2, then G contains exactly one vertex x of degree δ and G[N[x]] is a clique.
In [8], we established results of maximal 3-γc-vertex critical graphs. Let G be a maximal 3-γc-vertex critical graph with a vertex cut set S and C1,C2,...,Cm be the components of G−S.
Lemma 4
[8]**
For all v∈V(G), if m≥3 or v∈S∪V(Ci) where ∣V(Ci)∣>1, then G satisfies these following properties.
- (1)
Dv∩S=∅.
2. (2)
v* does not dominate S.*
Lemma 5
[8]**
Let a∈V(Ci) for some i∈{1,2,...,m}. Then G has these following properties.
- (1)
Let b∈V(Cj) for some j∈{1,2,...,m} such that {a,b} does not dominate G, if m≥3 or ∣V(Ci)∣,∣V(Cj)∣>1, then ∣Dab∩{a,b}∣=1 and ∣Dab∩S∣=1.
2. (2)
If c∈Da where c is an isolated vertex in S, then m=2 and {w}=V(Cj) for some j∈{1,2} where {w}=Da−{c}.
We characterized all maximal 3-γc-vertex critical graphs of connectivity s≥2 with a smallest cut set contains no edge
Theorem 3
[8]**
Let G be a maximal 3-γc-vertex critical graph of connectivity s≥2 with a smallest cut set contains no edge. Then G is isomorphic to G3=μ(Ks).
vFigure 1 : A graph G3=μ(Ks)
In [9], we established the upper bound of the independence number of maximal 3-γc-vertex critical graphs in term of the minimum degree.
Theorem 4
[9]**
Let G be a maximal 3-γc-vertex critical graph. Then α≤δ.
3 Maximal 3-γc-Vertex Critical Graphs
In this section we show, by using Theorem 4, that every maximal 3-γc-vertex critical graph satisfies α≤κ. We also study the such graphs when α=κ. Since we characterized all maximal 3-γc-vertex critical graphs with κ≤3 in [8], in the following, we focus on ∣S∣=κ≥4. We define α1,α2,p,S,H1 and H2 by the same as in the previous section.
Theorem 5
The independence number of any maximal 3-γc-vertex critical graph does not exceed the connectivity.
**Proof. **Suppose to the contrary that κ+1≤α. So ∣S∣+1≤α1+α2+∣S∩I∣. Then
[TABLE]
Claim 1 : ∣V(Ci)∣>1 for all i∈{1,2,...,m}, in particular, ∣Hi∣>1.
Suppose V(Ci)={c} for some i∈{1,2,...,m}. Thus NG(c)⊆S. By Theorem 4,
δ≤degG(c)≤∣S∣<∣S∣+1=κ+1≤α≤δ,
a contradiction and thus establishing Claim 1.
Let p=α1+α2 and {a1,a2,...,ap}=∪i=12Ii. If p=1, then, by (1), ∣S−I∣=0. It follows that S=S∩I. Thus S is an independent set. By Theorem 3, G is G3. Hence, NG3(x) in the graph G3 is a minimum cut set such that G3−NG3(x) contains x as a singleton component, contradicting Claim 1. Therefore p≥2.
Claim 2 : For a,b∈∪i=12Ii, ∣Dab∩{a,b}∣=1 and ∣Dab∩(S−I)∣=1.
Since ∣S∣≥4 and 2≤p=α1+α2. If p≥3, then ∪i=12Ii−{a,b}=∅. If p=2, then, by Equation 1, ∣S∣−∣S∩I∣+1≤2. As ∣S∣≥4, we must have ∣S∩I∣≥3, in particular, S∩I=∅. Hence (∪i=12Ii−{a,b})∪(S∩I)=∅. So {a,b} does not dominate G. By Claim 1 and Lemma 5(1), ∣Dab∩{a,b}∣=1 and ∣Dab∩S∣=1. Without loss of generality, let a∈Dab and {a′}=Dab∩S. By the connectedness of (G+ab)[Dab], a′∈S−I and thus establishing Claim 2.
Suppose that p=2. Consider G+a1a2. By Claim 2, ∣Da1a2∩(S−I)∣=1. Since Da1a2∩(S−I)⊆S−I, by Equation 1,
1≤∣S−I∣≤α1+α2−1=p−1=1.
Therefore, Da1a2∩(S−I)=S−I. If p≥3, then, by Lemma 2, the vertices a1,a2,...,ap can be ordered as x1,x2,...,xp and there exists a path y1,y2,...,yp−1 such that {xi,yi}≻cG+xixi+1 for i∈{1,2,...,p−1}. Since {x1,x2,...,xp}⊆∪i=12Ii, it follows by Claim 2 that {y1,y2,...,yp−1}⊆S−I. Hence, by (1), p−1≤∣S−I∣≤α1+α2−1=p−1. In both cases p=2 and p≥3, we have that {y1,y2,...,yp−1}=S−I.
If p=2, then clearly G[{y1}] is a clique. Suppose p≥3. Consider G+xixj for 2≤i=j≤p. By Claim 2, ∣Dxixj∩{xi,xj}∣=1 and ∣Dxixj∩(S−I)∣=1. Without loss of generality, let xi∈Dxixj. Because S−I={y1,y2,...,yp−1}, by Lemma 1(3), Dxixj∩(S−I)={yj−1}. Since xiyi−1∈/E(G), yi−1yj−1∈E(G). Therefore, G[{y1,y2,...,yp−1}] is a clique. Since {x1,x2,...,xp}⊆I, yi≻(S∩I) for i∈{1,2,...,p−1}. Therefore yi≻S, contradicting Lemma 4(2). Hence, α≤κ.
□
We have by Theorem 3 that the graph G3=μ(Ks) has NG3(x) as a minimum cut set and also a maximum independent set. Hence α(G3)=κ(G3). Therefor, the bound in Theorem 5 is best possible. We, further, focus on maximal 3-γc-vertex critical graphs which satisfy α=κ. We then have ∣S−I∣+∣S∩I∣=∣S∣=α1+α2+∣S∩I∣. So
[TABLE]
We may assume without loss of generality that α1≤α2. We would like to show that if a maximal 3-γc-vertex critical graph G satisfies α=κ, then G−S contains at least one singleton component for every minimum cut set S. We then suppose to the contrary that there is no singleton component of G−S, in particular, ∣Hi∣>1 for i=1,2. We next establish the following lemmas.
Lemma 6
Let G be a maximal 3-γc-vertex critical graph. If α=κ and ∣V(Ci)∣>1 for all i∈{1,2,...,m}, then p≥3.
**Proof. **By the assumption, ∣Hi∣>1 for i=1,2. Suppose first that p=0. Thus S=S∩I. By Theorem 3, G is G3. We have NG3(x) is a minimum cut set of G3 and x is a singleton component of G−NG3(x), contradicting the assumption. We distinguish 2 cases.
Case 1 : p=1.
By Equation 2, ∣S−I∣=1. Let {v}=S−I,{a1}=∪i=12Ii and {a2,a3,...,aα}=S∩I. Therefore α1=0,α2=1. Therefore, a1∈H2. Since ∣S∣≥4, ∣S∩I∣≥3. Lemma 2 yields that the vertices in {a2,a3,...,aα} can be ordered as x1,x2,...,xα−1 and there exists a path P=y1,y2,...,yα−2 with {xi,yi}≻cG+xixi+1 for i∈{1,..,α−2}. Since each yi is adjacent to at least one vertex of I for i=1,2,...,α−2, yi=a1. To dominate a1, yi∈H2∪{v}.
Subcase 1.1 : v∈/V(P).
So V(P)⊆H2. Thus xi≻H1 for i=1,2,...,α−2. Since S is a minimum cut set, NH1(v)=∅. Let u∈NH1(v). So u≻{x1,x2,...,xα−2,v}. Lemma 4(2) implies that uxα−1∈/E(G). Consider G+uyα−2. Since uxα−1,yα−2xα−1∈/E(G), by Lemma 5(1), ∣Duyα−2∩{u,yα−2}∣=1 and ∣Duyα−2∩S∣=1. So either yα−2∈Duyα−2 or u∈Duyα−2. In the first case, by Lemma 1(3), {x1,x2,...,xα−2,v}∩Duyα−2=∅. Hence xα−1∈Duyα−2. But G[Duyα−2] is not connected. Thus u∈Duyα−2. By the connectedness of (G+uyα−2)[Duyα−2], xα−1∈/Duyα−2. If xi∈Duyα−2 for i∈{1,2,...,α−2}, then Duyα−2 does not dominate xα−1. Therefore v∈Duyα−2. Thus va1∈E(G). Consider G+ua1. Since uxα−1,a1xα−1∈/E(G), by Lemma 5(1), ∣Dua1∩{u,a1}∣=1 and ∣Dua1∩S∣=1. Hence either u∈Dua1 or a1∈Dua1. In the first case, v∈/Dua1 by Lemma 1(3). By the connectedness of (G+ua1)[Dua1], xα−1∈/Dua1. To dominate xα−1, Dua1∩{x1,x2,...,xα−2}=∅. So Dua1∩S=∅, a contradiction. Hence, a1∈Dua1. Lemma 1(3) gives that v∈/Dua1. By the connectedness of (G+ua1)[Dua1], {x1,x2,...,xα−1}∩Dua1=∅. Clearly Dua1∩S=∅, a contradiction and Case 1.1 cannot occur.
Subcase 1.2 : v∈V(P).
Thus there exists j∈{1,2,...,α−2} such that yj=v. Thus xi≻H1 for i=j,α−1 and va1∈E(G). Since a1,xα−1∈I, a1xα−1∈/E(G). If xα−1 is not adjacent to w∈H1, then consider G+wa1. Lemma 5(1) implies that ∣Dwa1∩{w,a1}∣=1 and ∣Dwa1∩S∣=1. Thus either w∈Dwa1 or a1∈Dwa1. In both case, xα−1∈/Dwa1 because (G+wa1)[Dwa1] is connected. If w∈Dwa1, then, by Lemma 1(3), v∈/Dwa1. To dominate xα−1, {x1,x2...,xα−2}∩Dwa1=∅. So Dwa1∩S=∅, a contradiction. Hence, a1∈Dwa1.By the connectedness of (G+wa1)[Dwa1], Dwa1∩{x1,x2,...,xα−1}=∅. To dominate xj+1, v∈/Dwa1. We then have Dwa1∩S=∅, a contradiction. Thus xα−1≻H1. Clearly xi≻H1 for i=j. Since S is a minimum cut set, NH1(v)=∅. Let u′∈NH1(v). Lemma 4(2) yields that u′≻S−{xj}. Consider G+u′a1. By the same arguments as considering G+ua1, we have a contradiction. Thus Case 1 cannot occur.
Case 2 : p=2.
Let {a1,a2}=∪i=12Ii. By Equation 2, ∣S−I∣=p=2. Since ∣S∣≥4, it follows that ∣S∩I∣≥2, in particular, S∩I=∅ and {a1,a2} does not dominate G. Consider G+a1a2. Lemma 5(1) gives that ∣Da1a2∩{a1,a2}∣=1 and ∣Da1a2∩S∣=1. Without loss of generality, let a1∈Da1a2. By the connectedness of (G+a1a2)[Da1a2], ∣(S−I)∩Da1a2∣=1. Let {u}=(S−I)∩Da1a2. Thus ua1∈E(G),ua2∈/E(G) and u≻S∩I. Let v∈S−(I∪{u}). Lemma 4(2) implies that uv∈/E(G). Therefore a1v∈E(G)
Subcase 2.1 : α1=1 and α2=1.
Without loss of generality, let a1∈I1 and a2∈I2. Since ∣S∩I∣≥2, there exist a3,a4∈S∩I. Consider G+a3a4. Lemma 1(2) gives that Da3a4∩{a3,a4}=∅. To dominate a1, Da3a4={a3,a4}. Without loss of generality, let a3∈Da3a4. Lemma 1(1) implies that ∣Da3a4−{a3}∣=1. Let y∈Da3a4−{a3}. To dominate {a1,a2}, y∈/∪i=12Hi. By the connectedness of (G+a3a4)[Da3a4], y∈{v,u}. Since uv∈/E(G), a3u,a3v∈E(G). Consider G−a3 . Lemma 3(2) then implies that Da3∩{u,v}=∅. Lemma 4(1) implies also that Da3∩S=∅. So there exists z∈Da3∩(S∩I). Lemma 3(1) yields that ∣Da3−{z}∣=1. Let Da3−{z}={z′}. Since z∈S∩I, za1∈/E(G). Thus z′∈H1 to dominate a1. So Da3 does not dominate a2, a contradiction and Subcase 2.1 cannot occur.
Subcase 2.2 : α1=0 and α2=2.
Thus u≻H1. Let b1∈H1. Clearly {a1,b1} does not dominate G. Consider G+a1b1. Lemma 5(1) gives that ∣Da1b1∩S∣=1 and either b1∈Da1b1 or a1∈Da1b1. In the first case, {u,v}∩Da1b1=∅ by Lemma 1(3). To dominate a2, Da1b1∩(S∩I)=∅. Hence, Da1b1∩S=∅, a contradiction. Therefore, a1∈Da1b1. To dominate H1−b1 and by the connectedness of (G+a1b1)[Da1b1], (Da1b1−{a1})⊆{u,v}. Lemma 1(3) implies that v∈Da1b1. Thus v≻H1−b1. Let b2∈H1−{b1}. Therefore b2≻{u,v}. Consider G+a1b2. Lemma 5(1) implies that we have ∣Da1b1∩S∣=1 and either a1∈Da1b2 or b2∈Da1b2. In the first case, {u,v}∩Da1b2=∅ by Lemma 1(3). By the connectedness of (G+a1b2)[Da1b2], (S∩I)∩Da1b2=∅. Thus Da1b2∩S=∅, a contradiction. Therefore, b2∈Da1b2. To dominate a2, (S∩I)∩Da1b2=∅. Lemma 1(3) yields that Da1b2∩{u,v}=∅. Therefore Da1b2∩S=∅, a contradiction and Case 2 cannot occur. So p≥3 and this completes the proof.
□
In view of Lemma 6, p≥3. Lemma 2 then implies that the vertices in ∪i=12Ii can be ordered as x1,x2,...,xp and there exists a path y1,y2,...,yp−1 with {xi,yi}≻cG+xixi+1 for i=1,2,...,p−1.
Lemma 7
yi≻S∩I* and yi∈S−I for i=1,2,...,p−1.*
**Proof. **Since {xi,yi}≻cG+xixi+1 for i=1,2,...,p−1 and xi∈I, yi≻S∩I. By the connectedness of (G+xixi+1)[Dxixi+1] and Lemma 5(1), yi∈S−I and this completes the proof.
□
Lemma 7 implies that {y1,y2,...,yp−1}⊆S−I. By Equation 2, ∣(S−I)−{y1,y2, ...,yp−1}∣=1. Let {yp}=(S−I)−{y1,y2,...,yp−1}.
Lemma 8
For i,j∈{2,3,...,p}, if ypxi,ypxj∈E(G), then yi−1yj−1∈E(G).
**Proof. **Consider G+xixj. Lemma 5(1) yields that ∣Dxixj∩{xi,xj}∣=1 and ∣Dxixj∩S∣=1. Without loss of generality, let xi∈Dxixj and {a}=Dxixj∩S. By the connectedness of (G+xixj)[Dxixj], a∈S−I. Since xj≻(S−I)−{yj−1}, it follows by Lemma 1(3) that a=yj−1. Since yi−1xi∈/E(G), yj−1yi−1∈E(G) and this completes the proof.
□
Lemma 9
α1,α2>0.
**Proof. **By the assumption that α1≤α2, we suppose that α1=0. Clearly {x1,x2,...,xp}⊆H2 and yi≻H1 for i=1,2,...,p−1. Since S is a minimum cut set, NH1(yp)=∅. Let b∈NH1(yp). Therefore b≻S−I. Consider G+x1b. Lemma 5(1) implies that ∣Dx1b∩S∣=1 and either b∈Dx1b or x1∈Dx1b. Suppose that b∈Dx1b. To dominate x2, Dx1b∩(S−I)=∅. Lemmas 2 and 1(3) then imply that Dx1b∩(S−I)={yp}. So yp≻{x2,x3,...,xp}. Lemma 8 gives, further, that G[y1,y2,...,yp−1] is a clique. Lemma 7 then implies that yi≻S∩I for i=1,2,...,p−1. By Lemma 4(2), yiyp∈/E(G) for i=1,2,...,p−1. Therefore y1yp∈/E(G). Because {x1,y1}≻cG+x1x2, x1yp∈E(G), contradicting Lemma 1(3). Therefore, x1∈Dx1b. By the connectedness of (G+x1b)[Dx1b], Dx1b∩(S∩I)=∅. Lemma 1(3) implies that Dx1b∩(S−I)=∅. Thus Dx1b∩S=∅ contradicting Lemma 5(1) and this completes the proof.
□
Theorem 6
Let G be a maximal 3-γc-vertex critical graph. If α=κ, then G−S contains at least one singleton component for all minimum cut set S.
**Proof. **Let a graph G be a maximal 3-γc-vertex critical graph with α=κ. By Equation 2, ∣S−I∣=α1+α2. Suppose that there is no singleton component of G−S, in particular, ∣Hi∣>1 for i=1,2. Let α1+α2=p. Lemma 6 implies that p≥3. Lemma 9 gives also that 0<α1≤α2. We also define x1,x2,...,xp a path y1,y2,...,yp−1 and a vertex yp by the same as in the previous lemmas.
Suppose there exist xi,xj for i,j∈{2,3,...,p} such that yp∈Dxixj. Lemma 1(1) and (2) then implies that Dxixj={xi,yp} or Dxixj={xj,yp}. Without loss of generality, let Dxixj={xj,yp}. Thus yp≻{x1,x2,...,xp}−{xi}. Since {xi,yi}≻cG+xixi+1, yiyp∈E(G). Lemma 8 yields that G[{y1,y2,...,yp−1}−{yi−1}] is a clique. Since yiyi−1∈E(G), yi≻S−I. Lemma 7 implies that yi≻S∩I. Therefore yi≻S, contradicting Lemma 4(2). Hence, yp∈/Dxixj for any i,j∈{2,3,...,p}. By the same arguments as in the proof of Lemma 8, G[{y1,y2,...,yp−1}] is a clique. As yi≻S∩I, by Lemma 4(2), we must have yiyp∈/E(G) for i∈{1,2,...,p−1}. Since {xi,yi}≻cG+xixi+1 for i∈{1,2,...,p−1}, xiyp∈E(G). So x1≻S−I. By Lemma 4(2), S∩I=∅ as otherwise x1≻S. Let x1∈Hi for some i∈{1,2}. Consider G−x1. Since ∣Hj∣>1 for j=1,2, neither Dx1⊆H1 nor Dx1⊆H2. Lemma 4(1) gives, further, that Dx1∩S=∅. Lemma 3(2) implies that Dx1∩(S−I)=∅. Thus Dx1∩(S∩I)=∅. Let u1∈Dx1∩(S∩I). Lemma 3(1) implies also that ∣Dx1−{u1}∣=1. Let {w}=Dx1−{u1}. If w∈Hi, then u1≻H3−i. Since u1∈I, α3−i=0 contradicting Lemma 9. So w∈H3−i and u1≻Hi−x1. Since u1∈I, Ii={x1}. It follows that {x2,x3,...,xp}⊆H3−i.
Claim 1 : There is no u∈S∩I such that u≻S−I.
Suppose u≻S−I. Consider G−u. Lemma 4(1) implies that Du∩S=∅. Lemma 3(2) implies also that Du∩(S−I)=∅. Thus there exists u′∈Du∩(S∩I). Lemma 3(1) gives that ∣Du−{u′}∣=1. Let {z}=Du−{u′}. To dominate x1, z∈Hi. So Du does not dominate I3−i, a contradiction and thus establishing Claim 1.
Claim 1 together with Lemma 7 yield that yp is not adjacent to any vertex in S∩I. We now have yp an isolated vertex in S.
Claim 2 : y1≻Hi.
Suppose y1 is not adjacent to b1∈Hi. Consider G+b1x2. We see that b1y1,x2y1∈/E(G). Lemma 5(1) gives that ∣Db1x2∩S∣=1and either b1∈Db1x2 or x2∈Db1x2. If b1∈Db1x2, then (S−{y1,yp})∩Db1x2=∅ to dominate I3−i. Since ypx2∈E(G), by Lemma1(3), yp∈/Db1x2. By the connectedness of (G+b1x2)[Db1x2], y1∈/Db1x2. Therefore Db1x2∩S=∅, a contradiction. Hence, x2∈Db1x2. To dominate I3−i∪(S∩I), Db1x2∩{y2,y3,...,yp}=∅. By the connectedness of (G+b1x2)[Db1x2], ((S∩I)∪{y1})∩Db1x2=∅. Therefore, Db1x2∩S=∅, a contradiction and we settle Claim 2.
Let b1∈Hi−{x1}. Recall that u1≻Hi−x1. Clearly b1u1∈E(G). By Claim 2 and Lemma 2, b1≻{y1,y2,...,yp−1}∪{u1}. Consider G−b1. Lemma 4(1) implies that Db1∩S=∅. Lemma 3(2) gives that Db1∩({y1,y2,...,yp−1}∪{u1})=∅. If there is u2∈Db1∩((S∩I)−{u1}), then, by Lemma 3(1), let {y′}=Db1−{u2}. To dominate x1, y′∈Hi. Thus Db1 does not dominate x2, a contradiction. Therefore, {yp}=Db1∩S. Since yp is an isolated vertex in S, by Lemma 5(2), at least one of Ci is a singleton component contradicting the assumption. We then finish the proof.
□
By Theorem 6, we obtain the following corollary.
Corollary 1
Let G be a maximal 3-γc-vertex critical graph. If α=κ, then κ=δ.
**Proof. **By Theorem 6, we have at least one singleton component Ci of G−S. Let {c}=V(Ci). Thus NG(c)⊆S. Hence, δ≤degG(c)≤∣S∣=κ≤δ.
□
We next construct the class G4(s) of maximal 3-γc-vertex critical graphs which α<κ and κ<δ in order to show that the condition α=κ is necessary to prove Corollary 1. Let R,T,W and Z be non-empty disjoint vertex sets where R={r1,r2,...,rs},T={t1,t2,...,ts},W={w1,w2,...,ws},Z={z1,z2,...,zs} and s≥3. A graph G in the class G4(s) can be constructed from R,T,W and Z by adding edges according the join operations :
for 1≤i≤s, ri∨R∪T∪W−{ri,ti},
ti∨R∪W∪Z−{wi,ri},
wi∨R∪T∪Z−{ti,zi},
zi∨Z∪T∪W={zi,wi} and
adding edges so the the vertices in R and Z form cliques.
Observe that, for 1≤i≤s, NG(ri)=R∪T∪W−{ri,ti}, NG(ti)=R∪W∪Z−{wi,ri}, NG(wi)=R∪T∪Z−{ti,zi} and NG(zi)=Z∪T∪W={zi,wi}. Observe also that T and W are independent sets. A graph G is illustrated in Figure 4 where double lines between two sets means joining each vertex in one set to every vertex in the other set.
\begin{picture}(0.0,9.0)\par\put(3.2,6.5){\circle*{0.2}}
\put(4.2,6.5){\circle*{0.2}}
\put(6.8,6.5){\circle*{0.2}}
\par\put(3.2,2.05){\circle*{0.2}}
\put(4.2,1.45){\circle*{0.2}}
\put(6.8,0.5){\circle*{0.2}}
\par\put(10.2,0.5){\circle*{0.2}}
\put(12.8,1.45){\circle*{0.2}}
\put(13.8,2.05){\circle*{0.2}}
\par\put(10.2,6.5){\circle*{0.2}}
\put(12.8,6.5){\circle*{0.2}}
\put(13.8,6.5){\circle*{0.2}}
\par\par\put(3.2,2.05){\line(0,1){0.4}}\put(3.2,2.7){\line(0,1){0.4}}\put(3.2,3.35){\line(0,1){0.4}}\put(3.2,4.0){\line(0,1){0.4}}\put(3.2,4.65){\line(0,1){0.4}}\put(3.2,5.3){\line(0,1){0.4}}\put(3.2,5.95){\line(0,1){0.4}}
\put(4.2,1.45){\line(0,1){0.4}}\put(4.2,2.1){\line(0,1){0.4}}\put(4.2,2.75){\line(0,1){0.4}}\put(4.2,3.4){\line(0,1){0.4}}\put(4.2,4.05){\line(0,1){0.4}}\put(4.2,4.7){\line(0,1){0.4}}\put(4.2,5.35){\line(0,1){0.4}}\put(4.2,6.0){\line(0,1){0.4}}
\put(6.8,0.5){\line(0,1){0.4}}\put(6.8,1.1){\line(0,1){0.4}}\put(6.8,1.7){\line(0,1){0.4}}\put(6.8,2.3){\line(0,1){0.4}}\put(6.8,2.9){\line(0,1){0.4}}\put(6.8,3.5){\line(0,1){0.4}}\put(6.8,4.1){\line(0,1){0.4}}\put(6.8,4.7){\line(0,1){0.4}}\put(6.8,5.3){\line(0,1){0.4}}\put(6.8,5.9){\line(0,1){0.4}}
\par\put(10.2,0.5){\line(0,1){0.4}}\put(10.2,1.1){\line(0,1){0.4}}\put(10.2,1.7){\line(0,1){0.4}}\put(10.2,2.3){\line(0,1){0.4}}\put(10.2,2.9){\line(0,1){0.4}}\put(10.2,3.5){\line(0,1){0.4}}\put(10.2,4.1){\line(0,1){0.4}}\put(10.2,4.7){\line(0,1){0.4}}\put(10.2,5.3){\line(0,1){0.4}}\put(10.2,5.9){\line(0,1){0.4}}
\put(12.8,1.45){\line(0,1){0.4}}\put(12.8,2.1){\line(0,1){0.4}}\put(12.8,2.75){\line(0,1){0.4}}\put(12.8,3.4){\line(0,1){0.4}}\put(12.8,4.05){\line(0,1){0.4}}\put(12.8,4.7){\line(0,1){0.4}}\put(12.8,5.35){\line(0,1){0.4}}\put(12.8,6.0){\line(0,1){0.4}}
\put(13.8,2.05){\line(0,1){0.4}}\put(13.8,2.7){\line(0,1){0.4}}\put(13.8,3.35){\line(0,1){0.4}}\put(13.8,4.0){\line(0,1){0.4}}\put(13.8,4.65){\line(0,1){0.4}}\put(13.8,5.3){\line(0,1){0.4}}\put(13.8,5.95){\line(0,1){0.4}}
\par\put(3.2,2.05){\line(1,0){0.4}}\put(3.8,2.05){\line(1,0){0.4}}\put(4.4,2.05){\line(1,0){0.4}}\put(5.0,2.05){\line(1,0){0.4}}\put(5.6,2.05){\line(1,0){0.4}}\put(6.2,2.05){\line(1,0){0.4}}\put(6.8,2.05){\line(1,0){0.4}}\put(7.4,2.05){\line(1,0){0.4}}\put(8.0,2.05){\line(1,0){0.4}}\put(8.6,2.05){\line(1,0){0.4}}\put(9.2,2.05){\line(1,0){0.4}}\put(9.8,2.05){\line(1,0){0.4}}\put(10.4,2.05){\line(1,0){0.4}}\put(11.0,2.05){\line(1,0){0.4}}\put(11.6,2.05){\line(1,0){0.4}}\put(12.2,2.05){\line(1,0){0.4}}\put(12.8,2.05){\line(1,0){0.4}}\put(13.4,2.05){\line(1,0){0.4}}
\put(4.4,1.45){\line(1,0){0.4}}\put(5.0,1.45){\line(1,0){0.4}}\put(5.6,1.45){\line(1,0){0.4}}\put(6.2,1.45){\line(1,0){0.4}}\put(6.8,1.45){\line(1,0){0.4}}\put(7.4,1.45){\line(1,0){0.4}}\put(8.0,1.45){\line(1,0){0.4}}\put(8.6,1.45){\line(1,0){0.4}}\put(9.2,1.45){\line(1,0){0.4}}\put(9.8,1.45){\line(1,0){0.4}}\put(10.4,1.45){\line(1,0){0.4}}\put(11.0,1.45){\line(1,0){0.4}}\put(11.6,1.45){\line(1,0){0.4}}\put(12.2,1.45){\line(1,0){0.4}}
\put(6.8,0.5){\line(1,0){0.4}}\put(7.4,0.5){\line(1,0){0.4}}\put(8.0,0.5){\line(1,0){0.4}}\put(8.6,0.5){\line(1,0){0.4}}\put(9.2,0.5){\line(1,0){0.4}}\put(9.8,0.5){\line(1,0){0.4}}
\par\put(5.0,6.5){\oval(5.0,1.7)}
\put(12.0,6.5){\oval(5.0,1.7)}
\par\put(2.5,0.0){\dashbox(0.2)(5.0,2.5)[s]{}}
\put(9.5,0.0){\dashbox(0.2)(5.0,2.5)[s]{}}
\par\put(5.0,2.5){\line(0,1){3.15}}
\put(5.1,2.5){\line(0,1){3.15}}
\par\put(12.0,2.5){\line(0,1){3.15}}
\put(11.9,2.5){\line(0,1){3.15}}
\par\put(7.5,0.8){\line(1,0){2.0}}
\put(7.5,0.9){\line(1,0){2.0}}
\par\put(7.5,2.5){\line(1,2){2.0}}
\put(7.5,2.35){\line(1,2){2.0}}
\par\put(9.5,2.5){\line(-1,2){2.0}}
\put(9.5,2.35){\line(-1,2){2.0}}
\put(11.2,1.0){...}
\put(11.2,6.5){...}
\put(5.3,1.0){...}
\put(5.3,6.5){...}
\par\put(3.2,6.7){r1}
\put(4.2,6.7){r2}
\put(6.8,6.7){rs}
\par\put(3.0,1.55){t1}
\put(4.0,0.95){t2}
\put(6.6,0.0){ts}
\par\put(14.0,1.55){w1}
\put(13.0,0.95){w2}
\put(10.4,0.0){ws}
\par\put(13.8,6.7){z1}
\put(12.8,6.7){z2}
\put(10.2,6.7){zs}
\par\par\put(5.0,7.5){R}
\put(5.0,-0.5){T}
\put(12.0,-0.5){W}
\put(12.0,7.5){Z}
\put(5.0,-1.9){\footnotesize{Figure 4} : A graph G in the class G4(s)}
\end{picture}
...$$...$$...$$...
r_{1}$$r_{2}$$r_{s}
t_{1}$$t_{2}$$t_{s}
w_{1}$$w_{2}$$w_{s}
z_{1}$$z_{2}$$z_{s}
R$$T$$W$$ZFigure 4 : A graph G in the class G4(s)
Lemma 10
If G∈G4(s), then G is a maximal 3-γc-vertex critical graph.
**Proof. **We see that {r1,t2,w2}≻cG. Thus γc(G)≤3. Suppose there exist u,v∈V(G) such that {u,v}≻cG. Let i∈{1,..,s}. Suppose that u=ri. To dominate Z, v∈/R. If v∈T, then, by connected, v=ti. Thus {u,v} does not dominate ti. To dominate Z, v∈/W. Clearly v∈Z but G[{u,v}] is not connected, a contradiction. Hence, {u,v}∩R=∅. By symmetric, {u,v}∩Z=∅. Therefore {u,v}⊆T∪W. Without loss of generality, u=ti. By connected, v∈W−{wi}. Thus {u,v} does not dominate wi. Therefore γc(G)=3.
To establish the criticality, let u,v be a pair of non-adjacent vertices of G. For 1≤i≤s, if {u,v}={ri,ti}, then Duv={ri,ti}. If {u,v}={ti,wi}, then Duv={ti,wi}. If {u,v}={wi,zi}, then Duv={wi,zi}. For 1≤i=j≤s, if {u,v}={ti,tj}, then Duv={ti,rj}. If {u,v}={wi,wj}, then Duv={wi,zj}. If {u,v}={ri,zl} where l∈{1,2,...,s}, then Duv={ri,zl}. Thus G is a 3-γc-edge critical graph. Let v∈V(G). For 1≤i=j≤s, if u=ri, then Dv={ti,zj}. If v=ti, then Dv={tj,ri}. If v=wi, then Dv={zi,wj}. Finally, if v=zi, then Dv={wi,rj}. Therefore G is a maximal 3-γc-vertex critical graph and this completes the proof.
□
We see that G has T a maximum independent set and T∪W a mimimum cut set. Thus α=s<2s=κ. We, further, have that G is a regular graph with degG(v)=3s−2 for all v∈V(G). Since s≥3, δ=3s−2>2s=κ. Hence, the condition α=κ is necessary to prove Corollary 1.
We conclude this paper by establishing hamiltonian property of maximal 3-γc-vertex critical graphs. We prove the following result by using Theorem 1.
Corollary 2
For any 3-connected maximal 3-γc-vertex critical graph G. If κ<δ, then G is hamiltonian connected.
**Proof. **Suppose that κ<δ. Theorem 5 and Corollary 1 then imply that α<κ. By Theorem 1, G is hamiltonian connected.
□
However, we are sure that every 3-connected maximal 3-γc-vertex critical graph is hamiltonian connected. To prove this, by Theorem 1 and Corollary 2,we need only prove the following conjecture.
Conjecture 1
Let G be a 3-connected maximal 3-γc-vertex critical graph with α=κ=δ. Then G is hamiltonian connected.