Hamiltonian cycles in tough $(P_2\cup P_3)$-free graphs
Songling Shan

TL;DR
This paper proves that every 15-tough $(P_2rac12;P_3)$-free graph with at least three vertices contains a Hamiltonian cycle, advancing understanding of toughness conditions for Hamiltonicity.
Contribution
It establishes a new toughness threshold for Hamiltonicity specifically in $(P_2rac12;P_3)$-free graphs, a class previously not fully characterized.
Findings
15-tough $(P_2rac12;P_3)$-free graphs are Hamiltonian
Provides a specific toughness bound for Hamiltonicity in this graph class
Advances the toughness-Hamiltonicity conjecture in restricted graph classes
Abstract
Let be a real number and be a graph. We say is -tough if for every cutset of , the ratio of to the number of components of is at least . Determining toughness is an NP-hard problem for arbitrary graphs. The Toughness Conjecture of Chv\'atal, stating that there exists a constant such that every -tough graph with at least three vertices is hamiltonian, is still open in general. A graph is called -free if it does not contain any induced subgraph isomorphic to , the union of two vertex-disjoint paths of order 2 and 3, respectively. In this paper, we show that every 15-tough -free graph with at least three vertices is hamiltonian.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAdvanced Graph Theory Research · Limits and Structures in Graph Theory · graph theory and CDMA systems
Hamiltonian cycles in tough -free graphs
Songling Shan
Illinois State University, Normal, IL 61790
(8 January 2019)
Abstract. Let be a real number and be a graph. We say is -tough if for every cutset of , the ratio of to the number of components of is at least . Determining toughness is an NP-hard problem for arbitrary graphs. The Toughness Conjecture of Chvátal, stating that there exists a constant such that every -tough graph with at least three vertices is hamiltonian, is still open in general. A graph is called -free if it does not contain any induced subgraph isomorphic to , the union of two vertex-disjoint paths of order 2 and 3, respectively. In this paper, we show that every 15-tough -free graph with at least three vertices is hamiltonian.
Keywords. Toughness; Hamiltonian cycle; -free graph
1 Introduction
Graphs considered in this paper are simple, undirected, and finite. Let be a graph. Denote by and the vertex set and edge set of , respectively. For , denotes the set of neighbors of in . For and , define . If , we simply write for . We skip the subscript if the graph in consideration is clear from the context. Let . Then the subgraph induced on is denoted by . For notational simplicity, we write for . If is an edge, we write . Let be two disjoint vertex sets. Then is the set of edges of with one end in and the other end in .
The number of components of is denoted by . Let be a real number. The graph is said to be -tough if for each with . The toughness is the largest real number for which is -tough, or is if is complete. This concept, a measure of graph connectivity and “resilience” under removal of vertices, was introduced by Chvátal [7] in 1973. It is easy to see that if has a hamiltonian cycle then is 1-tough. Conversely, Chvátal [7] conjectured that there exists a constant such that every -tough graph is hamiltonian (Chvátal’s toughness conjecture). Bauer, Broersma and Veldman [2] have constructed -tough graphs that are not hamiltonian for all , so must be at least .
There are a number of papers on Chvátal’s toughness conjecture, and it has been verified when restricted to a number of graph classes [3], including planar graphs, claw-free graphs, co-comparability graphs, and chordal graphs. A graph is called -free if it does not contain two independent edges as an induced subgraph. Recently, Broersma, Patel and Pyatkin [5] proved that every 25-tough -free graph on at least three vertices is hamiltonian, and the author of this paper improved the required toughness in this result from 25 to 3 [13].
Let denote a path on -vertices. A graph is -free if it does not contain any induced copy of , the disjoint union of and . In this paper, we confirm Chvátal’s toughness conjecture for the class of -free graphs, a superclass of -free graphs.
Theorem 1**.**
Let be a -tough -free graph with at least three vertices. Then is hamiltonian.
In [10] it was shown that every 3/2-tough split graph on at least three vertices is hamiltonian. And the authors constructed a sequence of split graphs with no 2-factor and . So is the best possible toughness for split graphs to be hamiltonian. Since split graphs are -free, we cannot decrease the bound in Theorem 1 below 3/2. Although it is certain that 15-tough is not optimal, we are not sure about the “best possible” toughness for giving a hamiltonian cycle in a -free graph.
The class of -free graphs is well studied, for instance, see [5, 6, 8, 9, 11, 12]. It is a superclass of split graphs, where the vertices can be partitioned into a clique and an independent set. One can also easily check that every cochordal graph (i.e., a graph that is the complement of a chordal graph) is -free and so the class of -free graphs is at least as rich as the class of chordal graphs. By the definition, the class of -free graphs is a superclass of -free graphs but with much more complicated structures than graphs that are -free. The proof techniques used in [5] and [13] for showing that certain tough -free graphs are hamiltonian seem to be not applicable for -free graphs. The proof approach used in this paper for showing Theorem 1 is new and more general and reveals some structural properties of -free graphs.
2 Proof of Theorem 1
We start this section with some definitions. Let be a graph and a cutset of , and let be a component of . For a vertex , we say that * is adjacent to * if is adjacent in to a vertex of . We call a clique component of if is a clique in . We call a trivial component of if has only one vertex.
A star-matching is a set of vertex-disjoint copies of stars. The vertices of degree at least 2 in a star-matching are called the centers of the star-matching. In particular, if all the stars in a star-matching are isomorphic to , where is an integer, we call the star-matching a -matching. For a star-matching , we denote by the set of vertices covered by .
Let be an oriented cycle. For , denote the successor of on by and the predecessor of on by . For , denotes the portion of starting at , following in the orientation, and ending at . Likewise, is the opposite portion of with endpoints as and . We assume all cycles in consideration afterwards are oriented. A path connecting two vertices and is called a -path, and we write or in specifying the two endvertices of . Let and be two paths. If is an edge, we write as the concatenation of and through the edge .
Lemma 2.1** ([1], Theorem 2.10).**
Let be a bipartite graph with partite sets and , and let be a function from to the set of positive integers. If for every , it holds that , then has a subgraph such that , for every , and for every .
Lemma 2.2** (Bauer et al. [4]).**
Let be real and be a -tough -vertex graph () with . Then is hamiltonian.
Lemmas 2.3 and 2.4 below are consequences of -freeness.
Lemma 2.3**.**
Let be a -free graph and a cutset of . If has a component that is not a clique component, then all other components of are trivial. Consequently, if has at least two nontrivial components, then all components of are clique components.
Lemma 2.4**.**
Let be a -free graph and a cutset of , and let . Suppose that is adjacent to exactly one component of , and has a nontrivial component to which is not adjacent, then is adjacent in to all vertices of .
Lemma 2.5**.**
Let be a connected -free graph and a cutset of such that each vertex in is adjacent to at least two components of . Then each of the following statement holds.
- (i)
For every nontrivial clique component and for every vertex , is adjacent to . 2. (ii)
For every nontrivial clique component and for every vertex , if is adjacent to at least three components, then is adjacent in to at least vertices of . 3. (iii)
Let and be two nontrivial clique components of . Then for every vertex , either is adjacent in to at least vertices of each , or is adjacent in to all vertices of one of , .
Proof. For (i), let and be two neighbors of in respectively from two distinct components of . Then is an induced . Now for every edge , we must have that or is adjacent in to or , by the -freeness. Therefore, is adjacent to . For (ii), let and be a nontrivial clique component of . Since is adjacent to at least three components, there exists , respectively from two components of that are distinct from such that and in . Thus, is an induced in . Furthermore, since , . Thus, by the -freeness assumption, for every edge in , is adjacent to at least one endvertex of that edge. This, together with the fact that is a clique, we know that is adjacent at least vertices in . For (iii), assume to the contrary that the statement does not hold. By symmetry, we assume that there exists such that in , and there exists such that in . Let that exists by Lemma 2.5 (i). Then is an induced , giving a contradiction. ∎
Lemma 2.6**.**
Let and be an -vertex -tough graph, and let be a non-hamiltonian cycle of . If satisfies that , then has a cycle such that .
Proof. It is clear that if is adjacent to two consecutive vertices on , then
[TABLE]
is a cycle with the desired property. So we assume that for any , . Let be the set of the successors of the neighbors of on . Because there is a one-to-one correspondence between and , by the assumption that , we know that
[TABLE]
If there exist with such that in , then
[TABLE]
is a desired cycle. Therefore, we assume that is an independent set in . Let . Then , as is an independent set in . However, by (1),
[TABLE]
showing a contradiction to the toughness of . ∎
Lemma 2.7**.**
Let be an -vertex -tough -free graph, and let be a non-hamiltonian cycle of . Let be an -path. If both and are adjacent to more than vertices in , then has a cycle such that .
Proof. It is clear that if is adjacent to a vertex on and is adjacent to a vertex on such that , then
[TABLE]
is a cycle with the desired property. So we assume that
[TABLE]
Let
[TABLE]
Clearly,
[TABLE]
If there exist and with and such that in , then
[TABLE]
is a desired cycle. Therefore, we assume that
[TABLE]
We further claim that
[TABLE]
By symmetry, we only show that no two vertices in are consecutive on .
Assume to the contrary that there exists a path with such that for each with , , , and . Note that such vertices and exist by the assumption in (2). We assume that there exist such that in . For otherwise, letting gives that , and consequently, by (3).
Then is an induced in . Consider the edge . By the assumption in (2), in (otherwise, or with ), and by the assumption in (4), in . Thus, or in by the -freeness assumption. However, , showing a contradiction to (4).
Therefore, by (5),
[TABLE]
Let
[TABLE]
By the assumption in (4), is an independent set in . By the toughness of , we know that . Therefore, . These, together with (3) and (6), imply that
[TABLE]
showing a contradiction. ∎
Lemma 2.8**.**
Let be an -vertex -tough -free graph with , and let be a cutset of with . Assume that has at least two nontrivial clique components, and that for every edge , . Then has a hamiltonian cycle.
Proof. By Lemma 2.3, every component of is a clique component. If there exists such that is adjacent to exactly one component, say of , then we move from into . By Lemma 2.4, every component of is still a clique component. We move out all such vertex from iteratively and denote the remaining vertices in by . Note that , since is a connected graph and is a cutset of . Also, and has at least two nontrivial components. By Lemma 2.3, every component of is a clique component. Let
[TABLE]
Note that .
Since has a nontrivial component that has no edge going to , the -freeness of implies that consists of vertex-disjoint complete subgraphs of . Thus is a cutset of with components consisting those from and . Also, all components of are clique components in which at least two of them are nontrivial. By the toughness of , .
We will construct a hamiltonian cycle in through two steps: (1) combing spanning cycles from every clique component of that has at least three vertices into a single cycle , and (2) insterting remaining vertices in into to obtain a hamiltonian cycle of .
Suppose that has exactly clique components with , and that the first () of them are components that contain at least three vertices. Since has at least two nontrivial components, both and are nontrivial.
Claim 1**.**
The component contains at least 5 vertices.
Proof: Since , . Also, by . Therefore, a largest component of contains at least
[TABLE]
vertices.
Let
[TABLE]
By Lemma 2.5 (i) and the definition of , we know that if , then every vertex in is adjacent to at least three components of . By Lemma 2.5 (ii), we get the following claim.
Claim 2**.**
Suppose that . Then for every and for every nontrivial component of , is adjacent to at least vertices of .
Claim 3**.**
Suppose that . Then for every , is adjacent to all vertices of and is a clique in .
Proof: Note that both and are nontrivial components of . Since by Claim 1, is not adjacent to at least three vertices of by the definition of . Therefore, is adjacent to all vertices of by Lemma 2.5 (iii). For the second part, suppose to the contrary that there exist such that in . Let . Then and in by the first part of this claim. Thus, we find an induced . Since , the -freeness implies that for every edge in , at least one of and is adjacent to at least one endpoint of the edge. Since is complete, by Pigeonhole Principle, one of and is adjacent to at least vertices of . This gives a contradiction to the assumption that .
Similarly, we have the following result.
Claim 4**.**
Suppose that . Then for every , is adjacent to all vertices of and is a clique in .
By Claims 2 to 4, we have that
[TABLE]
Define
[TABLE]
If , we claim that there is a -matching between and such that every vertex in is the center of a -star. For otherwise, by Theorem 2.1, there exists such that . By the definition of , we then see that has at least components (vertices in form at least one component of that is disjoint from those containing vertices from ), implying that
[TABLE]
This gives a contradiction to the toughness.
[TABLE]
Claim 5**.**
If , then for every , . Consequently, for every nontrivial component of , is adjacent to at least vertices of .
Proof: If has at least three nontrivial components, then every vertex of is adjacent to all those nontrivial components by Lemma 2.5 (i). Therefore, by the definition of . In particularly, for . Hence, we assume that has exactly two nontrivial components, which are and . This assumption implies that . Consequently, since . Then for every , is adjacent to both and by Lemma 2.5 (i). Also is adjacent to a trivial component of . Thus . The second part of Claim 5 is a consequence of Claim 2.
Claim 6**.**
There is a cycle in with at least vertices such that contains all vertices from every , , and .
Proof: Suppose first that has at least three nontrivial components. Then by Lemma 2.5 (i), every vertex of is adjacent to all those nontrivial components of . Consequently, and . Therefore, for every and every , is adjacent to at least vertices of by Claim 2.
Let be distinct vertices in . (By the toughness of , . Since , we have enough vertices in to pick.) Let be a hamiltonian cycle of , and let with such that for , , and in . Then
[TABLE]
is a cycle that contains all vertices from each and the vertices from . Also trivially as .
So we assume that has exactly two nontrivial clique components, which are and , call this assumption (). Let
[TABLE]
By Claims 3 and 4, we know that both and are complete subgraphs of .
If and , we let be a hamiltonian cycle of . Clearly, .
Thus, we assume that or . Since has at least three vertices by Claim 1, contains at least three vertices. Note that by the assumption that has at least two nontrivial components and is one of them. Thus, contains at least three vertices either by or .
If there are two disjoint edges between and , then has a hamiltonian cycle . Thus, we may assume, without loss of generality, that there is either no edge between and or all edges between and are incident to only a single vertex, say in .
If , then by the definitions of and . Since has a cutvertex or is disconnected, the toughness of implies that . In addition, there are vertex-disjoint paths and connecting and in such that each only has exactly one of its endvertices in and . Let and , . Let be a hamiltonian cycle in such that , and be a hamiltonian cycle in such that . Then
[TABLE]
is a cycle that contains all vertices in clique components of that contain at least three vertices and the vertices from and . Also by the construction of .
So we assume that . Since and are the only nontrivial components of by assumption (), the assumption that implies that . Thus , since each trivial component of uses exactly two vertices from . Hence, we can find two vertices such that both and are adjacent to at least vertices of , and at least vertices of by Claim 5. We claim that is adjacent to at least two vertices of . This is clear if is adjacent to at least two vertices of . So we assume that is adjacent to only one vertex of . Let be the neighbor of from , and let , . By this choice, in . Note that is not adjacent to any vertex of , and is adjacent to less than vertices of . Therefore, we can find a vertex such that in and in . By the choice of , there is a vertex such that in . However, is an induced . This gives a contradiction. Since has at least 5 vertices, both and have at least four neighbors in . Thus we can select distinct vertices and such that and in .
Let be a hamiltonian cycle of such that , and let be a hamiltonian cycle of such that . Then
[TABLE]
is a cycle that contains all vertices in clique components of that contain at least three vertices and the vertices and . Furthermore, .
Since for each , , and , we have that
[TABLE]
Claim 7**.**
Let be the cycle defined in Claim 6. For any , has more than neighbors on .
Proof: Note that every vertex in is adjacent to at least two components of . If has at least three nontrivial clique components, then Lemma 2.5 (ii) implies that for every , and for every nontrivial clique component of , is adjacent to at least vertices of . By the toughness of and the assumption that , has at least
[TABLE]
neighbors in the union of the nontrivial clique components of that contain at least three vertices. Therefore, has more than neighbors on .
So we assume that has exactly two nontrivial clique components. Since (recall that ), we know that is adjacent to at least vertices of , and is adjacent to at least vertices of . Note that
[TABLE]
Since contains all vertices from , we conclude that is adjacent to at least (by ) neighbors on .
By Claim 7, and by applying Lemma 2.6 for and vertices in iteratively, we get a longer cycle such that . Note also that
[TABLE]
Recall that for every , is adjacent to at least vertices in each , by Claim 5. Therefore, if , then has in at least
[TABLE]
neighbors on . Then applying Lemma 2.7 for and every path in iteratively, we obtain a hamiltonian cycle in . Hence we
[TABLE]
Claim 8**.**
For any two -stars , if is a 2-vertex component of and , then at least one of and has more than neighbors on .
Proof: For otherwise, since is adjacent to exactly one vertex in , and ,
[TABLE]
show a contradiction to the assumption that for every edge , .
Let
[TABLE]
Take , note that and . By the definition of , . By Claim 7, and . Now applying Lemma 2.6 for and every path in iteratively, we get a longer cycle such that .
By the toughness of , has at most components in total. Particularly, has at most components that have at most two vertices in total. By Claim 8, we know that for every 2-vertex component of , at least one of or has more than neighbors on . Therefore, at least one of the two -stars centered, respectively, at and is contained in . In other words, there is at most one -star from that centers at a vertex from a same component of . Therefore,
[TABLE]
By the definition of and by the assumption that for any , , we know that for any path , where and , we have that . Therefore, the number of neighbors that has in on is at least
[TABLE]
Similarly, the vertex has in at least neighbors on . Now applying Lemma 2.7 for and every path in iteratively gives a hamiltonian cycle in . ∎
Proof of Theorem 1.
Since is 15-tough, it is 30-connected, and consequently, . By Lemma 2.2, we may assume that
[TABLE]
We consider two case to finish the proof.
**Case 1: For every edge , **.
Denote by
[TABLE]
By the assumption of Case 1, we know that is an independent set in . Therefore,
[TABLE]
by .
Since is 15-tough, Lemma 2.1 implies that has a -matching with all vertices in as the centers of the -matching. Let be the set of the vertices contained in . By (12), we have that
[TABLE]
Denote by . Then by the definitions of and (13), we get that
[TABLE]
We first assume that has a hamiltonian cycle . For every copy of , say , by (15),
[TABLE]
Let
[TABLE]
By (16), applying Lemma 2.6 with respect to and every vertex in iteratively, we get a longer cycle such that .
By the definition of and by the assumption that for any , , we know that for any path , where and , we have that . Therefore, the number of neighbors that has in on is at least
[TABLE]
Similarly, the vertex has in at least neighbors on . Now applying Lemma 2.7 for and every path in iteratively gives a hamiltonian cycle in .
Hence we assume that does not have a hamiltonian cycle. By Lemma 2.2 and (14), we know that . Therefore, there exists such that
[TABLE]
If , then we get that , and thus by (13),
[TABLE]
showing a contradiction to . So we assume that , thus . As by (14), we know that each component of contains at least
[TABLE]
vertices. By Lemma 2.3, we know that every component of is a clique component. Let . We then see that all components of are nontrivial. Also, since and by (13). Furthermore, by the assumption of Case 1, for every edge , . Now we can apply Lemma 2.8 on and to find a hamiltonian cycle in .
**Case 2: There exists an edge such that **.
Let
[TABLE]
such that is smallest among all the degree sums of two adjacent vertices in .
By the assumption of this case and the choice of , we know that
[TABLE]
By the definition of , and is one of the components of . Since , and , has a component with at least 5 vertices. This, together with the fact that is one of the components of , Lemma 2.3 implies that every component of is a clique component, and has at least two nontrivial components. Again Lemma 2.8 implies that has a hamiltonian cycle. ∎
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] Jin Akiyama and Mikio Kano. Factors and factorizations of graphs , volume 2031 of Lecture Notes in Mathematics . Springer, Heidelberg, 2011. Proof techniques in factor theory.
- 2[2] D. Bauer, H. J. Broersma, and H. J. Veldman. Not every 2-tough graph is Hamiltonian. In Proceedings of the 5th Twente Workshop on Graphs and Combinatorial Optimization (Enschede, 1997) , volume 99, pages 317–321, 2000.
- 3[3] D. Bauer, H.J. Broersma, and E. Schmeichel. Toughness in graphs – a survey. Graphs and Combinatorics , 22(1):1–35, 2006.
- 4[4] Douglas Bauer, H. J. Broersma, J. van den Heuvel, and H. J. Veldman. Long cycles in graphs with prescribed toughness and minimum degree. Discrete Math. , 141(1-3):1–10, 1995.
- 5[5] H. Broersma, V. Patel, and A. Pyatkin. On toughness and Hamiltonicity of 2 K 2 2 subscript 𝐾 2 2K_{2} -free graphs. J. Graph Theory , 75(3):244–255, 2014.
- 6[6] F. R. K. Chung, A. Gyárfás, Z. Tuza, and W. T. Trotter. The maximum number of edges in 2 K 2 2 subscript 𝐾 2 2K_{2} -free graphs of bounded degree. Discrete Math. , 81(2):129–135, 1990.
- 7[7] V. Chvátal. Tough graphs and Hamiltonian circuits. Discrete Math. , 5:215–228, 1973.
- 8[8] M. El-Zahar and P. Erdős. On the existence of two nonneighboring subgraphs in a graph. Combinatorica , 5(4):295–300, 1985.
