New sufficient condition for Hamiltonian graphs

Similar documents
Every 3-connected, essentially 11-connected line graph is hamiltonian

Eulerian Subgraphs and Hamilton-Connected Line Graphs

Every line graph of a 4-edge-connected graph is Z 3 -connected

Every 3-connected, essentially 11-connected line graph is Hamiltonian

Hamilton-Connected Indices of Graphs

Small Cycle Cover of 2-Connected Cubic Graphs

Degree Sequence and Supereulerian Graphs

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and

Regular matroids without disjoint circuits

Disjoint Hamiltonian Cycles in Bipartite Graphs

Supereulerian planar graphs

Every 4-connected line graph of a quasi claw-free graph is hamiltonian connected

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and

HAMILTONIAN CYCLES AVOIDING SETS OF EDGES IN A GRAPH

arxiv: v1 [cs.dm] 12 Jun 2016

(This is a sample cover image for this issue. The actual cover is not yet available at this time.)

SPANNING TREES WITH A BOUNDED NUMBER OF LEAVES. Junqing Cai, Evelyne Flandrin, Hao Li, and Qiang Sun

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and

Graph homomorphism into an odd cycle

A Fan-Type Condition for Hamiltonian Graphs *

An Ore-type Condition for Cyclability

Hamiltonian claw-free graphs

Solution to a problem on hamiltonicity of graphs under Oreand Fan-type heavy subgraph conditions

Upper Bounds of Dynamic Chromatic Number

Eulerian Subgraphs in Graphs with Short Cycles

Discrete Mathematics

Graphs with large maximum degree containing no odd cycles of a given length

Hamilton cycles and closed trails in iterated line graphs

Cycles of length 1 modulo 3 in graph

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and

Antoni Marczyk A NOTE ON ARBITRARILY VERTEX DECOMPOSABLE GRAPHS

The Chvátal-Erdős condition for supereulerian graphs and the hamiltonian index

be a path in L G ; we can associated to P the following alternating sequence of vertices and edges in G:

Group connectivity of certain graphs

Neighborhood intersections and Hamiltonicity in almost claw-free graphs

and critical partial Latin squares.

Conditional colorings of graphs

Cycles in 4-Connected Planar Graphs

Observation 4.1 G has a proper separation of order 0 if and only if G is disconnected.

Improved degree conditions for 2-factors with k cycles in hamiltonian graphs

Jian Cheng, Cun-Quan Zhang & Bao- Xuan Zhu

Spanning Paths in Infinite Planar Graphs

CYCLE STRUCTURES IN GRAPHS. Angela K. Harris. A thesis submitted to the. University of Colorado Denver. in partial fulfillment

ARTICLE IN PRESS Theoretical Computer Science ( )

arxiv: v2 [math.co] 7 Jan 2016

4 CONNECTED PROJECTIVE-PLANAR GRAPHS ARE HAMILTONIAN. Robin Thomas* Xingxing Yu**

Nowhere-zero 3-flows in triangularly connected graphs

Applied Mathematics Letters

Line Graphs and Forbidden Induced Subgraphs

On Brooks Coloring Theorem

Connectivity of graphs with given girth pair

The Reduction of Graph Families Closed under Contraction

Distance between two k-sets and Path-Systems Extendibility

Domination and Total Domination Contraction Numbers of Graphs

Long cycles in 3-cyclable graphs

Spanning cycles in regular matroids without M (K 5 ) minors

Discrete Mathematics. The average degree of a multigraph critical with respect to edge or total choosability

Observation 4.1 G has a proper separation of order 0 if and only if G is disconnected.

Degree Sum Conditions and Vertex-Disjoint Cycles in a Graph

On the Turán number of forests

Group Colorability of Graphs

On Hamiltonian cycles and Hamiltonian paths

Degree Conditions for Spanning Brooms

Combined degree and connectivity conditions for H-linked graphs

Set-orderedness as a generalization of k-orderedness and cyclability

The Turán number of sparse spanning graphs

arxiv: v1 [math.co] 3 Mar 2015 University of West Bohemia

arxiv: v1 [math.co] 13 May 2016

A characterization of diameter-2-critical graphs with no antihole of length four

Proper connection number and 2-proper connection number of a graph

ON THE NUMBERS OF CUT-VERTICES AND END-BLOCKS IN 4-REGULAR GRAPHS

HAMILTONICITY AND FORBIDDEN SUBGRAPHS IN 4-CONNECTED GRAPHS

A characterisation of eccentric sequences of maximal outerplanar graphs

Graphs with Integer Matching Polynomial Roots

Disjoint triangles and pentagons in a graph

Discrete Mathematics. The edge spectrum of the saturation number for small paths

Some Nordhaus-Gaddum-type Results

Every 3-Connected Claw-Free Z 8 -Free Graph Is Hamiltonian

Extendability of Contractible Configurations for Nowhere-Zero Flows and Modulo Orientations

1 Hamiltonian properties

Double domination edge removal critical graphs

Average degrees of edge-chromatic critical graphs

Cographs; chordal graphs and tree decompositions

arxiv: v1 [math.co] 29 Apr 2016

THE COMPLEXITY OF DISSOCIATION SET PROBLEMS IN GRAPHS. 1. Introduction

Graph Theory. Thomas Bloom. February 6, 2015

Path decompositions and Gallai s conjecture

arxiv: v1 [math.co] 12 Jul 2017

AALBORG UNIVERSITY. Total domination in partitioned graphs. Allan Frendrup, Preben Dahl Vestergaard and Anders Yeo

Spanning 2-trails from degree sum conditions

arxiv: v1 [math.co] 20 Oct 2018

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and

Rao s degree sequence conjecture

Generalized Pigeonhole Properties of Graphs and Oriented Graphs

Graphs and Combinatorics

Subdivisions of Large Complete Bipartite Graphs and Long Induced Paths in k-connected Graphs Thomas Bohme Institut fur Mathematik Technische Universit

Closing the gap on path-kipas Ramsey numbers

Stability in the Erdős Gallai Theorems on cycles and paths

Erdös-Ko-Rado theorems for chordal and bipartite graphs

Transcription:

Applied Mathematics Letters ( ) www.elsevier.com/locate/aml New sufficient condition for Hamiltonian graphs Kewen Zhao a,b,,hong-jian Lai c,yehong Shao d a Department of Mathematics, Qiongzhou University, Wuzhishan, Hainan, 572200, China b Department of Mathematics, Hainan Normal University, Haikou, Hainan, 571100, China c Department of Mathematics, West Virginia University, Morgantown, WV 26506-6310, United States d Ohio University Southern, Ironton, OH 45638, United States Received 26 December 2003; accepted 17 October 2005 Abstract Let G be a graph, and δ(g) and α(g) be the minimum degree and the independence number of G, respectively. For a vertex v V (G), d(v) and N(v) represent the degree of v and the neighborhood of v in G, respectively. A number of sufficient conditions for a connected simple graph G of order n to be Hamiltonian have been proved. Among them are the well known Dirac condition (1952) (δ(g) n 2 )and Ore condition (1960) (for any pair of independent vertices u and v, d(u) + d(v) n). In 1984 Fan generalized these two conditions and proved that if G is a 2-connected graph of order n and max{d(x), d(y)} n/2for each pair of nonadjacent vertices x, y with distance 2 in G, theng is Hamiltonian. In 1993, Chen proved that if G is a 2-connected graph of order n, andifmax{d(x),d(y)} n/2 for each pair of nonadjacent vertices x, y with 1 N(x) N(y) α(g) 1, then G is Hamiltonian. In 1996, Chen, Egawa, Liu and Saito further showed that if G is a k-connected graph of order n, andif max{d(v) : v S} n/2 foreveryindependent set S of G with S =k which has two distinct vertices x, y S such that the distance between x and y is 2, then G is Hamiltonian. In this paper, we generalize all the above conditions and prove that if G is a k-connected graph of order n, andifmax{d(v) : v S} n/2 foreveryindependent set S of G with S =k which has two distinct vertices x, y S satisfying 1 N(x) N(y) α(g) 1, then G is Hamiltonian. c 2006 Published by Elsevier Ltd Keywords: Hamiltonian graphs; Dirac condition; Ore condition; Fan condition; Chen condition 1. Introduction We consider finite and simple graphs in this paper; undefined notations and terminology can be found in [1]. In particular, we use V (G), E(G), κ(g), δ(g) and α(g) to denote the vertex set, edge set, connectivity, minimum degree and independence number of G,respectively. If G is a graph and u,v V (G),thenapath in G from u to v is called a (u,v)-path of G.Ifv V (G) and H is asubgraph of G,thenN H (v) denotes the set of vertices in H that are adjacent to v in G. Thus, d H (v), thedegree of v relative to H,is N H (v). Wealsowrite d(v) = d G (v) and N(v) = N G (v). If C and H are subgraphs of G, thenn C (H ) = u V (H) N C (u), andg C denotes the subgraph of G induced by V (G) V (C). Forvertices u,v V (G), thedistance between u and v, denoted d(u,v),isthelength of a shortest (u,v)-path in G, or if no such path exists. Corresponding author at: Department of Mathematics, Qiongzhou University, Wuzhishan, Hainan, 572200, China. 0893-9659/$ - see front matter c 2006 Published by Elsevier Ltd doi:10.1016/j.aml.2005.10.024

2 K. Zhao et al. / Applied Mathematics Letters ( ) Let C m = x 1 x 2 x m x 1 denote a cycle of order m. Define N + C m (u) ={x i+1 : x i N Cm (u)}, N C m (u) ={x i 1 : x i N Cm (u)}, and define N ± C m (u) = N + C m (u) N C m (u), where subscripts are taken modulo m. Asubset S V (G) is said to be an essential independent set if S is an independent set in G and there exist two distinct vertices x, y S with d(x, y) = 2. The following sufficient conditions to assure the existence of a Hamiltonian cycle in a simple graph G of order n 3arewellknown. Theorem 1.1 (Dirac [4]). If δ(g) n/2,then G is Hamiltonian. Theorem 1.2 (Ore [6]). If d(u)+d(v) n for each pair of nonadjacent vertices u,v V (G),then G is Hamiltonian. Theorem 1.3 (Fan [5]). If G is a 2-connected graph and if max{d(u), d(v)} n/2 for each pair of vertices u,v V (G) with d(u,v) = 2, then G is Hamiltonian. Theorem 1.4 (Chen [2]). If G is a 2-connected graph and if max{d(u), d(v)} n/2 for each pair of vertices u,v V (G) with 1 N(u) N(v) α(g) 1,thenGisHamiltonian. Theorem 1.5 (Chen et al. [3]). If G is a k-connected (k 2) graph and if max{d(v) : v S} n/2 for every independent set S of order k such that S has two distinct vertices x, ywith d(x, y) = 2,thenGisHamiltonian. The purpose of this paper is to unify and extend the theorems above. We shall prove the following result. Theorem 1.6. If G is a k-connected (k 2) graph of order n, and if max{d(v) : v S} n/2 for every independent set S of order k, such that S has two distinct vertices x, ywith 1 N(x) N(y) α(g) 1,thenGisHamiltonian. The proof of Theorem1.6 will be given in Section 2. It is straightforward to verify that if a graph G satisfies the hypothesis of any one of Theorems 1.1 1.5, then it will also satisfy the hypothesis of Theorem 1.6. InSection 3, we shall show that there exist Hamiltonian graphs satisfying the hypothesis of Theorem 1.6, butwhose Hamiltonicity cannot be assured by any one of Theorems 1.1 1.5. InthissenseTheorem 1.6 extends Theorems 1.1 1.5. 2. Proof of Theorem 1.6 For a cycle C m = x 1 x 2 x m x 1,wewrite [x i, x j ] to denote the section x i x i+1 x j of the cycle C m,where subscripts are taken modulo m. Fornotational convenience, [x i, x j ] will denote the (x i, x j )-path x i x i+1 x j of C m,aswellasthevertexsetofthispath. If C 1 and C 2 are cycles of a graph G such that V (C 1 ) V (C 2 ) and V (C 2 ) > V (C 1 ), thenwesay that C 2 extends C 1.Suppose that P 1 is an (x, y)-path of G and P 2 is a (y, z)- path of G such that V (P 1 ) V (P 2 ) ={y}, thenweusep 1 P 2 to denote the (x, z)-path of G induced by the edges E(P 1 ) E(P 2 ).IfV(P 1 ) V (P 2 ) ={x, y} and x = z, thenp 1 P 2 denotes the cycle of G induced by the edges E(P 1 ) E(P 2 ).Weneed to establish some lemmas. Lemma 2.1. Suppose that C m = x 1 x 2 x m x 1 is a cycle of a graph G, H is a component of G V (C m ), and x i, x j are distinct vertices in N Cm (H ).IfGdoes not have a cycle extending C m,theneach of the following holds. (i) {x i 1, x i+1, x j 1, x j+1 } N Cm (H ) =. (ii) x i+1 x j+1 E(G) and x i 1 x j 1 E(G). (iii) If x t x j+1 E(G) for some vertex x t [x j+2, x i ],thenx t 1 x i+1 E(G) and x t 1 N Cm (H ). (iv) If x t x j+1 E(G) for some vertex x t [x i+1, x j ],thenx t+1 x i+1 E(G). (v) No vertex of G (V (C m ) V (H )) is adjacent to both x i+1 and x j+1. (vi) If x V (H ) such that xx i E(G), then{x} N (H ) must be an independent set. Proof. (i), (ii) and (v) follow immediately from the assumption that G does not have a cycle extending C m.itremains to show that (iii), (iv) and (vi) must also hold. Since x i, x j N Cm (H ), x i, x j V (H ) such that x i x i, x j x j E(G). Let P denote an (x i, x j )-path in H.

K. Zhao et al. / Applied Mathematics Letters ( ) 3 Suppose that (iii) fails. Then there exists a vertex x t {x j+2, x j+3,...,x i } satisfying x t x j+1 E(G). If x t 1 x i+1 E(G), thenx i P x j [x j 1, x t 1 ][x t 2, x t ][x t+1, x i ] is alonger cycle than C m,contrary to the assumption that C m is longest. Hence x t x j+1 E(G).Nextweassume that x t 1 is adjacent to some vertex x t 1 V (H ).LetP denote an (x t 1, x j )-path in H.Thenx j P x t 1 [x t 2, x t ][x t+1, x j ] is a cycle extending C m,acontradiction. Hence (iii) must hold. The proof for (iv) is similar, and so it is omitted. To prove (vi), assume to the contrary, that G[{x} N (H )] has an edge e. ByLemma 2.1(i), N (H ) is an independent set. Hence e = xx j+1 for some x j N Cm (H ),andso x j V (H ) such that x j x j E(G).LetP denote an (x j, x)-path in H.Thenx[x j+1, x j ]P is a cycle extending C m.thiscompletes the proof of the lemma. Lemma 2.2. Let G be a 2-connected graph of order n, C m = x 1 x 2 x m x 1 be a cycle of G, and H be a component of G V (C m ).Ifone of the following holds, (i) there exist two distinct vertices x i, x j V (C m ) with x i+1, x j+1 in N (H ) such that d(x i+1 ) n/2 and d(x j+1 ) n/2,or (ii) there exists a vertex x i+1 N (H ) and a vertex y V (H ) such that d(x i+1 ) n/2 and d(y) n/2, then there exists a cycle C extending C m. Proof. Suppose, to the contrary, that G does not have a cycle C extending C m. (1) First we assume that (i) holds. Then x i, x j V (H ) such that x i x i, x j x j E(G). SinceH is connected, H has an (x i, x j )-path P. We define a map f : N G (x j+1 ) {x j } V (G) as follows: v N G (x j+1 ) {x j }, v if v V (C m ) f (v) = v + = x t+1 if v = x t [x i+1, x j 1 ] v = x t 1 if v = x t [x j+2, x i ]. Claim 1. f is an injection, and v N G (x j+1 ) {x j }, f (v) N G (x i+1 ) {x x+1 }. It is straightforward to verify that f is an injection. Firstly, by (1), v V (H ), x i+1 f (v) = x i+1 v E(G). If v = x t [x i+1, x j 1 ] with x t x j+1 E(G), thenbylemma 2.1(iv), x t+1 x i+1 E(G). If x t [x j+2, x i 1 ] with x t x j+1 E(G), thenbylemma 2.1(iii), x t 1 x i+1 E(G). Finally, the definition of f shows that v N G (x j+1 ) {x j }, f (v) x i+1.thisproves Claim 1. By Claim 1, f (N G (x j+1 ) {x j }) N G (x i+1 ) =.ByLemma 2.1(i), f (N G (x j+1 ) {x j }) V (H ) =. Therefore, N G (x i+1 ) ( f (N G (x j+1 ) {x j }) V (H ) {x i+1 }) =.Since f is an injection, f (N G (x j+1 ) {x j }) = N G (x j+1 ) 1 = d(x j+1 ) 1. It follows that d(x i+1 ) = N G (x i+1 ) n f (N G (x j+1 )) V(H) {x i+1 } n d(x j+1 ) V (H ), and so d(x i+1 ) + d(x j+1 ) n 1, contrary to the assumption that both d(x i+1 ) n/2 andd(x j+1 ) n/2. Thus if (i) holds, there exists a cycle C extending C m. The proof for the case when (ii) holds is similar, and so it is omitted. Lemma 2.3. Let G be a k-connected (k 2), C m = x 1 x 2 x m x 1 be a cycle of G with V (C m ) < V (G) such that G does not have a cycle extending C m, and let H be a component of G V (C m ).Letx V (H ) be a vertex satisfying N Cm (x) 1 and d(x) <n/2.then (i) N Cm (H ) k. Moreover, if x i N Cm (x),theneach of the following holds. (ii) 1 N(x) N(x i+1 ) α(g) 1. Proof. Since C m is alongest cycle with V (C m ) < V (G), V (C m ) N Cm (H ),andson Cm (H ) separates V (H ) and V (C m ) N Cm (H ).SinceG is k-connected, N Cm (H ) κ(g) k.thisproves Lemma 2.3(i). By Lemma 2.1(vi), {x} N + C m (H ) must be an independent set. Thus N Cm (H ) = N + C m (H ) α(g) 1, and so 1 N(x) N(x i+1 ) α(g) 1. This proves Lemma 2.3(ii).

4 K. Zhao et al. / Applied Mathematics Letters ( ) Lemma 2.4. Let G be a k-connected (k 2), C m = x 1 x 2 x m x 1 be a cycle of G with V (C m ) < V (G) such that G does not have a cycle extending C m, and H be a component of G V (C m ).If Gsatisfies the hypothesis of Theorem 1.6, thend(x) n/2 for every x V (H ) with N Cm (x) 1. Proof. By contradiction, we assume that there exists an x V (H ) satisfying N Cm (x) 1andd(x) <n/2. We assume that x i N Cm (x). ByLemma 2.3(i), V N (H ) with x i+1 V and with V = k 1. By Lemma 2.1(vi), {x} N (H ) is an independent set. It follows by Lemma 2.3(ii) that V {x} is a k-element set satisfying the hypothesis of Theorem 1.6. Sinced(x) <n/2, by the hypothesis of Theorem 1.6, theremust be avertex(sayx h+1 )inv with d(x h+1 ) n/2. By Lemma 2.2(i), every vertex of N (H ) {x h+1 } must have degree less than n/2. Since G is k-connected (k 2), there must be some vertex y V (H ) and x j V (C m ) with x j+1 N (y) {x h+1 } (possibly y = x). By Lemma 2.3 (with y replacing x in Lemma 2.3), we have 1 N(y) N(x j+1 ) α(g) 1. Hence pick a subset X N (H ) {x h+1, x j+1 } with X =k 2andletV = X {y, x j+1 }.ThenV also satisfies the hypothesis of Theorem 1.6. Hence there must be a vertex u V such that d(u) n/2. Since d(x h+1 ) n/2 and d(u) n/2, it follows by Lemma 2.2(i) that G has a cycle C extending C m,contrary to the assumption that C m has no extension in G. Lemma 2.5. Let G be a k-connected (k 2) graph, C m = x 1 x 2 x m x 1 be a cycle of G with V (C m ) < V (G) such that G has no cycleextending C m, and let H be a component of G V (C m ).If d(x) n/2 for every x V (H ) with N Cm (x) 1,thenH = G V (C m ). Proof. By contradiction, we assume that G V (C m ) has at least two distinct components H and H. Since G is connected, there must exist a vertex y V (H ) adjacent to some vertex of C m.bylemma 2.4, d(y) n/2. It follows that d(x)+d(y) n.ontheother hand, if a vertex x V (H ) is adjacent to some x i V (C m ), then x is not adjacent to x i+1 and x i 1.Hence we have d(x) V (C m ) /2 + V (H ) {x}. Similarly, we have d(y) V (C m ) /2 + V (H ) {y}. Itfollows that d(x) + d(y) [ V (C m ) /2 + V (H ) {x} ] + [ V (C m ) /2 + V (H ) {y} ] n 2, a contradiction. Hence we must have H = G V (C m ). Proof of Theorem 1.6. Suppose, to the contrary, that G is not Hamiltonian. Let C m = x 1 x 2 x m x 1 be a longest cycle of G, andleth be a component of subgraph G V (C m ).ByLemma 2.4,wemayassume that d(x) n/2, x V (H ) with N Cm (x) 1. (2) By Lemma 2.5, H = G V (C m ).Without loss of generality, assume that for some x V (H ), thereexists an x i N Cm (x). By Lemma 2.3(i) and since k 2, N Cm (H ) 2. (3) Choose x i, x j N Cm (H ) such that {x i+1, x i+2,...x j 1 } N Cm (H ) =. (4) If V (H ) =1, then C m = C n 1 and N Cm (H ) =d(x) n/2. Then there must exist x i, x i+1 N G (x) V (C n 1 ), and so G is Hamiltonian, a contradiction. Hence we must have V (H ) 2. (5) x V (H ), N G (x) {x} V (H ) N Cm (H ),andso by (2), V (H ) + N Cm (H ) N G (x) {x} d(x) + 1 n/2 + 1. (6) It follows by (4) (6) that {x i+1, x i+2,...,x j 1 } (n V (H ) N Cm (H ) )/ N Cm (H ) [n ( V (H ) + N Cm (H ) )]/ N Cm (H ) [n (1 + n/2)]/ N Cm (H ) (n/2 1)/ N Cm (H ) ( V (H ) + N Cm (H ) 2)/ N Cm (H ) V (H ) / N Cm (H ) +1 2/ N Cm (H ) < V (H ).

K. Zhao et al. / Applied Mathematics Letters ( ) 5 In other words, {x i+1, x i+2,...,x j 1 } < V (H ). Recall that we have chosen x i, x j N Cm (H ) satisfying (3).Letx i, x j V (H ) with x i x i, x j x j E(G),andletP be an (x i, x j )-path of H.ThenL = x j [x j+1, x i ]Px j is a cycle of G.Choose a cycle C of G such that V (L) V (C ) and V (C ) is maximized. (8) By (8) and by applying Lemmas 2.3 2.5 to the cycle C,weobtain the following Claim 2. Claim 2. Let H be a component of G V (C ). (i) Then x V (H ) with N C (x) and with d(x ) n/2. (ii) G V (C ) = H has only one component. Claim 3. V (H ) {x i+1, x i+2,...,x j 1 }=,orv (H ) V (H ) =. By Lemma 2.5, H = G V (C m ),andsov (H ) {x i+1, x i+2,...x j 1 } V (H ). Ifforsomex, x V (H ) such that x {x i+1, x i+2,...x j 1 } and x V (H ), thensinceh is connected, H has an (x, x )-path Q. Since V (H ) {x i+1, x i+2,...x j 1 } V (H ),thereisan edge in Q joining a vertex in {x i+1, x i+2,...,x j 1 } and a vertex in V (H ),contrary to (3).Thisproves Claim 3. If V (H ) {x i+1, x i+2,...,x j 1 }=,thenbyclaim 2(ii), V (C m ) V (C ).By(8), V (H ) V (C ),and so C extends C m,contrary to the assumption that C m is alongest cycle. Thus by Claim 3, V (H ) V (H ) =, and so V (H ) {x i+1, x i+2,...,x j 1 }.By (7), V (H ) > {x i+1, x i+2,...,x j 1 } V (H ). ByLemma 2.5 and Claim 2(ii), we have V (C ) =n V (H ) > n V (H ) = V (C m ), (7) contrary to the assumption that C m is alongest cycle. Therefore, the proof of Theorem 1.6 is complete. 3. Examples The purpose of this section is to show that there exist Hamiltonian graphs satisfying the hypothesis of Theorem 1.6, but whose Hamiltonicity cannot be assured by any one of Theorems 1.1 1.5. Let H and K be two vertex disjoint graphs. As in [1], H K denotes the disjoint union of H and K,andH K denotes the graph obtained from the disjoint union of H and K by adding all the edges in {uv : u V (H ) and v V (K )}. Similarly, for two disjoint vertex subsets Z 1 and Z 2,wedefineZ 1 Z 2 to be the graph whose vertex set is Z 1 Z 2 and whose edge set is {v 1 v 2 : v 1 Z 1 and v 2 Z 2 } (E(Z 1 ) E(Z 2 )). IfZ denotes a vertex subset, let K (Z) denote the complete graph whose vertex set is Z. IfZ E(H ) and Z is an edge set not in H but the two ends of each edge in Z are in H,thenH Z + Z denotes the graph obtained from H by deleting the edges in Z and adding the edges in Z. Let H 1, H 2,...,H 6 be vertex disjoint graphs each of which is isomorphic to a K 3.For each i with 1 i 6, pick avertexx i V (H i ).LetY ={y 1, y 2, y 3, y 4, y 5, y 6 } denote a vertex set disjoint from 6 i=1 V (H i), andletk6 c denote the edgeless graph with V (K6 c ) = Y.Defineedge subsets as follows: W 1 = E({x 6} (Y {y 6 })) W 2 = E({y 6} ( 3 j=1 (V (H j) {x j }))) W 3 = E(K ({x 1, x 2, x 3, x 4, x 5 })) W 4 = E(K ( 6 j=4 V (H j) {x 6 })) W 5 = E((V (H 5) {x 5 }) ( 3 j=1 (V (H j) {x j }))). Define L = (K c 6 ( 6 i=1 H i)) (W 1 W 2 ) + (W 3 + W 4 + W 5 ).

6 K. Zhao et al. / Applied Mathematics Letters ( ) Table 1 Degrees of all the vertices in L The vertex v N L (v) d(v) Remark x i (V (H i ) {x i }) Y (X {x i, x 6 }) 12 1 i 3 x i, x i (V (H i ) {v}) (Y {y 6 }) {x 5, x 5 } 9 1 i 3 x 4 ( 6 j=4 V (H j ) Y X) {x 4, x 6 } 16 x 4, x 4 ( 6 j=4 V (H j ) Y ) {v, x 6 } 13 x 5 ( 6 j=4 V (H j ) Y X) {x 5, x 6 } 16 x 5, x 5 ( 6 j=1 V (H j ) Y ) ((X {x 5 }) {v}) 19 x 6 {x 6, x 6, y 6} 3 x 6, x 6 ( 6 j=4 V (H j )) Y {v} 14 y i 6 j=1 V (H j ) {x 6 } 17 1 i 5 y 6 ( 6 i=4 V (H i)) {x 1, x 2, x 3 } 12 For each i with 1 i 6, let V (H i ) ={x i, x i, x i }.Then, for each vertex v V (L), the neighborhood of v can be expressed as in Table 1. We have the following observations. Proposition 3.1. Each of the following holds. (i) V (L) =24 and κ(l) = 3, and α(l) = 6. (ii) Lsatisfies the hypothesis of Theorem 1.6 with k = 3. (iii) L does not satisfy the hypothesis of any one of Theorems 1.1 1.5. Proof. (i) We only need to show that α(l) = 6. Since Y is an independent set in L, wehaveα(l) 6. Let S V (L) be an independent set of L. Note that for each i with 1 i 5, N L (y i ) = 6 i=1 V (H i) {x 6 },and N L (y 6 ) = 6 i=4 V (H i) {x 1, x 2, x 3 }.Therefore, if for some y i S with 1lei 5, then S Y {x 6 },andsoas y 6 x 6 E(L), S 6. Hence we assume that Y S {y 6 }.Ify 6 S, then since each H i is a complete graph, S V (H i ) 1, and so S 6. Hence y 6 S. Hence in this case, ((Y {y 6 }) N L (y 6 )) S =.Butfor i = 1, 2, 3, since H i is complete, S V (H i ) 1, and so S 4. This implies that α(l) = 6, and soproves(i). (ii) As shown in Table 1, ( 3 i=1 V (H i) X) {x 6 } is the set of all vertices of degree less than V (L) /2 = 12 in L.For any two distinct, nonadjacent vertices u,v ( 3 i=1 V (H i) X), wehave,bytable 1, N L (u) N L (v) Y {y 6 } + {x 5, x 5 } 7. Moreover, v ( 3 i=1 V (H i) X) {x 6 },bytable 1, N L (x 6 ) N L (v) =. It follows that if S V (L) is an independent with S = 3suchthatforsomeu,v S, itholds that 1 N L (u) N L (v) α(l) 1 = 5, then one vertex in S must have degree at least 12 = V (L) /2. Hence L satisfies the condition of Theorem 1.6. (iii) By Table 1,itisclear that the hypotheses of Theorems 1.1 1.3 cannot be satisfied. Let S 1 ={x 1, x 6, y 6 }.As N L (x 1 ) N L(y 6 ) ={x 1, x 5, x 5 }, and N L(x 6 ) 3, S 1 is a vertex set of L such that u,v S 1, N L (u) N L (v) α(l) 1. As d L (x 6 ) = 3andd L (x 1 ) = 8, L does not satisfy the condition of Theorem 1.4. Let S 2 ={z 1, z 2, z 3 } ( 3 i=1 V (H i) X) be an independent set of L. Thenasz 1, z 2, z 3 N L (y 1 ),thedistance between z 1, z 2 in L is 2. By Table 1,max{d L (v) : v S 2 } < 11, and so L does not satisfy the condition of Theorem 1.5. The same construction also works for an arbitrary integer k 3. Let H 1, H 2,...,H 2k be vertex disjoint graphs, each of which is isomorphic to a K k.for each i with 1 i 2k 1, pick two distinct vertices x i, x i V (H i ),and

K. Zhao et al. / Applied Mathematics Letters ( ) 7 Table 2 Degrees of all the vertices in L k The vertex v N Lk (v) d(v) Remark v V (H i ) {x i, x i } x i, x i v V (H i ) {x i, x i } x i, x i v V (H 2k 1 ) {x 2k 1, x 2k 1 } x 2k 1, x 2k 1 ( 2k 1 j=1 (V (H j ) {x j, x j })) Y (V (H i ) {v}) 2k(k 1) 1 i k (V (H i ) {v}) (Y {y 6 }) {x 2k 1, x 2k 1 } 3k 1 i k ( k j=1 (V (H j ) {x j, x j })) ( 2k j=k+1 V (H j )) Y {v, x 2k } 2(k 2 1) k + 1 i 2k 2 2k j=k+1 V (H j ) Y {v, x 2k } k 2 +2k 2 k + 1 i 2k 2 ( k j=1 (V (H j ) {x j, x j })) ( 2k j=k+1 V (H j )) Y {v, x 2k } 2(k 2 1) ( k j=1 {x j, x j } ( 2k j=k+1 V (H j ) Y ) {v, x 2k }) x 2k {y 6 } V(H 2k ) {x 2k } k v V (H 2k ) {x 2k } y i y 2k k 2 +4k 2 ( 2k j=k+1 V (H j )) Y {v} k 2 +2k 1 2k j=1 V (H j ) {x 2k } 2k 2 1 1 i 2k 1 ( 2k j=k+1 (V (H j ))) ( k j=1 (V (H j ) {x j, x j })) 2k2 2k let x 2k V (H 2k ).LetY ={y 1, y 2,...,y 2k } denote a vertex set disjoint from 2k i=1 V (H i), andletk2k c denote the edgeless graph with V (K2k c ) = Y.Defineedge subsets as follows: W 1 = E({x 2k } (Y {y 2k })), W 2 = E({y 2k } ( k j=1 ({x j, x j }))), W 3 = E(K ( 2k 1 j=1 (V (H j) {x j, x j }))), W 4 = E(K ( 2k j=k+1 V (H j) {x 2k })), W 5 = E(({x 2k 1, x 2k 1 }) ( k j=1 {x j, x j })). Define L k = (K2k c ( 2k i=1 H i)) (W 1 W 2 ) + (W 3 + W 4 + W 5 ). Note that L 3 = L above. For each vertex v V (L k ),the neighborhood of v can be expressed as in Table 2. Imitating the proof for Proposition3.1, wehavethefollowing Proposition 3.2. Proposition 3.2. Each of the following holds. (i) V (L k ) =2k(k + 1) and κ(l k ) = k, and α(l k ) = 2k. (ii) L k satisfies the hypothesis of Theorem 1.6 with κ(l k ) = k. (iii) L k does not satisfy the hypothesis of any one of Theorems 1.1 1.5. Acknowledgments The first author s research was partially supported under NFS of Hainan Province (No.10301) and (No.10501). References [1] A.J. Bondy, U.S.R. Murty, Graph Theory with Applications, American Elsevier, New York, 1976. [2] G. Chen, Hamiltonian graphs involving neighborhood intersections, Discrete Math. 112 (1993) 253 258. [3] G. Chen, Y. Egawa, X. Liu, A Saito, Essential independent setand Hamiltonian cycles, J. Graph Theory 21 (1996) 243 250. [4] G.A. Dirac, Some theorems on abstract graphs, Proc. London Math. Soc. 2 (1952) 69 81. [5] G. Fan, New sufficient conditions for cycles ingraphs, J. Combin. Theory Ser. B 37 (1984) 221 227. [6] O. Ore, Note on Hamiltonian circuits, Amer. Math. Monthly 67 (1960) 55.