On degree sum conditions for 2-factors with a prescribed number of cycles
Shuya Chiba

TL;DR
This paper establishes new degree sum conditions in highly connected graphs that guarantee the existence of a 2-factor with a specified number of cycles, generalizing previous results in the field.
Contribution
It introduces a generalized degree condition involving maximum degree sums of independent sets that ensures the existence of 2-factors with a given number of cycles, extending prior work.
Findings
Proves a degree sum condition for 2-factors with exactly k cycles
Generalizes previous results by Brandt et al. and Yamashita
Applicable to m-connected graphs of order n ≥ 5k - 2
Abstract
For a vertex subset of a graph , let be the maximum value of the degree sums of the subsets of of size . In this paper, we prove the following result: Let be a positive integer, and let be an -connected graph of order . If for every independent set of size in , then has a 2-factor with exactly cycles. This is a common generalization of the results obtained by Brandt et al. [Degree conditions for 2-factors, J. Graph Theory 24 (1997) 165-173] and Yamashita [On degree sum conditions for long cycles and cycles through specified vertices, Discrete Math. 308 (2008) 6584-6587], respectively.
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
TopicsLimits and Structures in Graph Theory · Finite Group Theory Research · graph theory and CDMA systems
On degree sum conditions
for 2-factors with a prescribed number of cycles
Shuya Chiba
Applied Mathematics
Faculty of Advanced Science and Technology
Kumamoto University
2-39-1 Kurokami, Kumamoto 860-8555, Japan E-mail address: [email protected]; This work was supported by JSPS KAKENHI Grant Number 17K05347.
Abstract
For a vertex subset of a graph , let be the maximum value of the degree sums of the subsets of of size . In this paper, we prove the following result: Let be a positive integer, and let be an -connected graph of order . If for every independent set of size in , then has a 2-factor with exactly cycles. This is a common generalization of the results obtained by Brandt et al. [Degree conditions for 2-factors, J. Graph Theory 24 (1997) 165–173] and Yamashita [On degree sum conditions for long cycles and cycles through specified vertices, Discrete Math. 308 (2008) 6584–6587], respectively.
Keywords: Hamilton cycles, 2-factors, Vertex-disjoint cycles, Degree sum conditions
AMS Subject Classification: 05C70, 05C45, 05C38
1 Introduction
In this paper, we consider finite simple graphs, which have neither loops nor multiple edges. For terminology and notation not defined in this paper, we refer the readers to [4]. The independence number and the connectivity of a graph are denoted by and , respectively. For a vertex of a graph , we denote by and the degree and the neighborhood of in . Let be the minimum degree sum of an independent set of vertices in a graph , i.e., if , then
[TABLE]
otherwise, . If the graph is clear from the context, we often omit the graph parameter in the graph invariant. In this paper, “disjoint” always means “vertex-disjoint”.
A graph having a hamilton cycle, i.e., a cycle containing all the vertices of the graph, is said to be hamiltonian. The hamiltonian problem has long been fundamental in graph theory. But, it is NP-complete, and so no easily verifiable necessary and sufficient condition seems to exist. Therefore, many researchers have focused on “better” sufficient conditions for graphs to be hamiltonian (see a survey [14]). In particular, the following degree sum condition, due to Ore (1960), is classical and well known.
Theorem A** **(Ore [15])
Let be a graph of order . If , then is hamiltonian.
Chvátal and Erdős (1972) discovered the relationship between the connectivity, the independence number and the hamiltonicity.
Theorem B** **(Chvátal, Erdős [8])
Let be a graph of order at least . If , then is hamiltonian.
Bondy [2] pointed out that the graph satisfying the Ore condition also satisfies the Chvátal-Erdős condition, that is, Theorem B implies Theorem A.
By Theorem B, we should consider the degree condition for the existence of a hamilton cycle in graphs with . In fact, Bondy (1980) gave the following degree sum condition by extending Theorem B.
Theorem C** **(Bondy [3])
Let be an -connected graph of order . If , then is hamiltonian.
In 2008, Yamashita [17] introduced a new graph invariant and further generalized Theorem C as follows. For a vertex subset of a graph with , we define
[TABLE]
Let , and if , then let
[TABLE]
otherwise, . Note that .
Theorem D** **(Yamashita [17])
Let be an -connected graph of order . If , then is hamiltonian.
This result suggests that the degree sum of non-adjacent “two” vertices is important for hamilton cycles.
On the other hand, it is known that a 2-factor is one of the important generalizations of a hamilton cycle. A 2-factor of a graph is a spanning subgraph in which every component is a cycle, and thus a hamilton cycle is a 2-factor with “exactly 1 cycle”. As one of the studies concerning the difference between hamilton cycles and 2-factors, in this paper, we focus on 2-factors with “exactly cycles”. Similar to the situation for hamilton cycles, deciding whether a graph has a 2-factor with cycles is also NP-complete. Therefore, the sufficient conditions for the existence of such a 2-factor also have been extensively studied in graph theory (see a survey [11]). In particular, the following theorem, due to Brandt, Chen, Faudree, Gould and Lesniak (1997), is interesting. (In the paper [5], the order condition is not “” but “”. However, by using a theorem of Enomoto [9] and Wang [16] (“every graph of order at least with contains disjoint cycles”) for the cycles packing problem, we can obtain the following. See the proof in [5, Lemma 1].)
Theorem E** **(Brandt et al. [5])
Let be a positive integer, and let be a graph of order . If , then has a 2-factor with exactly cycles.
This theorem shows that the Ore condition guarantees the existence of a hamilton cycle but also the existence of a 2-factor with a prescribed number of cycles.
By considering the relation between Theorem A and Theorem E, Chen, Gould, Kawarabayashi, Ota, Saito and Schiermeyer [6] conjectured that the Chvátal-Erdős condition in Theorem B also guarantees the existence of a 2-factor with exactly cycles (see [6, Conjecture 1]). Chen et al. also proved that if the order of a 2-connected graph with is sufficiently large compared with and with the Ramsey number , then the graph has a 2-factor with cycles. In [12], Kaneko and Yoshimoto “almost” solved the above conjecture for (see the comment after Theorem E in Chen et al. [6] for more details). Another related result can be found in [7]. But, the above conjecture is still open in general. In this sense, there is a big gap between hamilton cycles and 2-factors with exactly cycles.
In this paper, by combining the techniques of the proof for hamiltonicity and the proof for 2-factors with a prescribed number of cycles, we give the following Yamashita-type condition for 2-factors with cycles.
Theorem 1
Let be a positive integer, and let be an -connected graph of order . If , then has a 2-factor with exactly cycles.
This theorem implies the following:
Remark 2
- •
Theorem 1 is a generalization of Theorem D.
- •
Theorem 1 leads to the Bondy-type condition: If is an -connected graph of order with , then has a 2-factor with exactly cycles. Therefore, Theorem 1 is also a generalization of Theorem E for sufficiently large graphs. (Recall that and for .)
- •
Theorem 1 leads to the Chvátal-Erdős-type condition: If is a graph of order at least with , then has a 2-factor with exactly cycles.
The complete bipartite graph ( is odd) does not contain a 2-factor, and hence the degree condition in Theorem 1 is best possible in this sense. The order condition in Theorem 1 comes from our proof techniques. Similar to the situation for the proof of Theorem E, we will use the order condition only for the cycles packing problem (see Lemma 5 and the proof of Theorem 1 in Section 3). The complete bipartite graph shows that is necessary. In the last section (Section 4), we note that “” can be replaced with “” for the Bondy-type condition (and the Chvátal-Erdős-type condition) in Remark 2.
Table 1 summarizes the conditions mentioned in the above.
To prove Theorem 1, in the next section, we extend the concept of insertible vertices which was introduced by Ainouche [1], and we prove Theorem 1 in Section 3 by using it.
2 The concept of insertible vertices
In this section, we prepare terminology and notations and give some lemmas.
Let be a graph. For and , we let and . For , let . For , we denote by the subgraph of induced by . An -path in is a path from a vertex to a vertex in . We write a cycle (or a path) with a given orientation by . If there exists no fear of confusion, we abbreviate by . Let be an oriented cycle (or path). We denote by the cycle with a reverse orientation. For , we denote the successor and the predecessor of on by and . For , we denote by the -path on . The reverse sequence of is denoted by . In the rest of this paper, we consider that every cycle (path) has a fixed orientation, unless stated otherwise, and we often identify a subgraph of with its vertex set .
The following lemma is obtained by using the standard crossing argument, and so we omit the proof.
Lemma 1
Let be a graph of order , and let be an -path of order at least in . If , then contains a cycle of order at least .
In [1], Ainouche introduced the concept of insertible vertices, which has been used for the proofs of the results on hamilton cycles. In this paper, we modify it for 2-factors with cycles, and it also plays a crucial role in our proof. Let be a graph, and let be the set of cycles and paths in which are pairwise disjoint. For a vertex in , the vertex is insertible for if there is an edge in such that for some with . In the following lemma, “partition” of a graph means a partition of the vertex set.
Lemma 2
Let be a graph, and let be the set of cycles and paths in which are pairwise disjoint, and be a path in . If every vertex of is insertible for , then G\big{[}\bigcup_{1\leq p\leq r+s}V(D_{p})\cup V(P)\big{]} can be partitioned into cycles and paths.
Proof of Lemma 2. By choosing the following two vertices and the edge inductively, we can get the desired partition of G\big{[}\bigcup_{1\leq p\leq r+s}V(D_{p})\cup V(P)\big{]}. Let be the first vertex along , and take an edge in E(D_{i})\ \big{(}\subseteq\bigcup_{1\leq p\leq r+s}E(D_{p})\big{)} such that for some with (since is insertible for , we can take such an edge). We let be the last vertex along such that (may be ). Then, we can insert all vertices of into . In fact, by replacing the edge by the path , we can obtain a spanning subgraph of such that is a cycle if is a cycle; otherwise, is a path. By the choice of and , we have or for each vertex of , and hence every vertex of is insertible for . Thus, we can repeat this argument for the path and the set , and we get then the desired partition. ∎
In the rest of this section, we fix the following. Let be disjoint cycles in a graph , and let . Choose so that
[TABLE]
Suppose that does not form a 2-factor of . Let , and let be a component of and . Let
[TABLE]
We assume that appear in this order on , and let . Note that by the maximality of , for . We denote by and a -path in and a -path passing through a vertex of in , respectively.
Lemma 3
For , contains a non-insertible vertex for .
Proof of Lemma 3. Suppose that every vertex of is insertible for . Then, by Lemma 2, G\big{[}\bigcup_{2\leq p\leq k}V(C_{p})\cup V(u_{i}^{+}\overrightarrow{C_{1}}u_{i+1}^{-})\big{]} has a 2-factor with exactly cycles. With the cycle , we can get disjoint cycles in such that the sum of the orders is larger than , a contradiction. ∎
For , let be the first non-insertible vertex for in on , i.e., every vertex of is insertible for , but is not insertible (Lemma 3 guarantees the existence of such a vertex ).
Lemma 4
Let be integers with and . If and , then (i) , and (ii) .
Proof of Lemma 4. Consider the path
[TABLE]
See Figure 1. Then, is a path in passing through all vertices of and a vertex of . Recall that every vertex of is insertible for , and hence G\big{[}\bigcup_{2\leq p\leq k}V(C_{p})\cup V(u_{i}^{+}\overrightarrow{C_{1}}x^{-})\big{]} has a 2-factor with exactly cycles (by Lemma 2). Hence, the maximality of and Lemma 1 yield that
[TABLE]
In particular, (i) holds. Then, by applying (i) for each vertex in and the vertex , we have . Combining this with the above inequality, we get,
[TABLE]
Thus (ii) also holds. ∎
3 Proof of Theorem 1
Before proving Theorem 1, we will give the following lemma for the cycles packing problem.
Lemma 5
Let and be the same ones as in Theorem 1. Under the same degree sum condition as Theorem 1, contains disjoint cycles.
Proof of Lemma 5. If , then it is easy to check that contains a cycle. If or , then by a theorem of Enomoto [9], contains disjoint cycles (note that if , then is -connected, that is, the minimum degree is at least ). Thus, we may assume that and . Then, we have and . Note that, by the definition of and , . Note also that because . Hence, by a theorem of Fujita et al. [10] (“every graph of order at least with contains disjoint cycles”), we can get the desired conclusion. ∎
Now we are ready to prove Theorem 1.
Proof of Theorem 1. Let be an -connected graph of order such that . We show that has a 2-factor with exactly cycles. By Theorem E, we may assume that . By Lemma 5, contains disjoint cycles. Let (), , , , and () be the same graphs and vertices as the ones described in the paragraph preceding Lemma 3 in Section 2. In particular, we may assume that . Because, since is -connected, it follows that (note that by the maximality of , ), and hence, without loss of generality, we may assume that .
We first consider the set
[TABLE]
Then, Lemma 4 implies the following:
- (1)
is an independent set of size . 2. (2)
for ().
On the other hand, by the maximality of and Lemma 2, is non-insertible for . This implies the following:
- (3)
for , and hence .
Since , it follows from (1) that there exist two distinct vertices and in such that . Then, by (2), we get
[TABLE]
Combining this with (3) and the definition of , we may assume that
- (4)
.
Next, let be the first non-insertible vertex for in the path on (we can take such a vertex by Lemma 3 and the symmetry of and ), and we consider the set
[TABLE]
Then, by the symmetry of and , Lemma 4, and since is non-insertible for , we have the following:
- (5)
is an independent set of size . 2. (6)
for (). 3. (7)
for , and hence .
Since , it follows from (5) that there exist two distinct vertices and in such that . Then, by (6), we get
[TABLE]
Combining this with (3), (7) and the definition of , we have the following:
- (8)
for some with .
[TABLE]
Hence, there exists a cycle (), say , such that
[TABLE]
This implies that there exists an edge in such that . By changing the orientation of if necessary, we may assume that . Note that , and consider two cycles
[TABLE]
Then, are disjoint cycles such that the sum of the orders is larger than , a contradiction.
This completes the proof of Theorem 1. ∎
4 Notes on the order condition
As shown in the argument of the previous section, in the proof of Theorem 1, the order condition “” is required only to show the existence of disjoint cycles in a graph (recall that the order condition in Theorem E is also). Therefore, the proof of Theorem 1 actually implies the following.
Theorem 3
Let be a positive integer, and let be an -connected graph of order . Suppose that contains disjoint cycles. If , then has a 2-factor with exactly cycles.
From this theorem, if we can obtain better results on the cycles packing problem, then the order conditions in Theorem 1 and Remark 2 can be improved. In fact, by using the result of Kierstead, Kostochka and Yeager (2017) and modifying the proof of Lemma 5, we can obtain a sharp order condition for the result in Remark 2 (see Corollary 4).
Theorem F** **(Kierstead et al. [13])
Let be an integer with , and let be a graph of order with . Then contains disjoint cycles if and only if (i) , and (ii) if is odd and , then and if , then is not a wheel.
Lemma 6
Let be a positive integer, and let be an -connected graph of order . If , then contains disjoint cycles.
Proof of Lemma 6. By a similar argument as in the proof of Lemma 5, we have the following: If , then we can easily find a cycle; If or , then by a theorem of Enomoto [9], contains disjoint cycles; If , and or , then by a theorem of Fujita et al. [10], contains disjoint cycles. Thus, we may assume that , and . Then, and . Since and , it follows that and is not a wheel. Hence, by Theorem F, contains two disjoint cycles. Thus, the lemma follows. ∎
Recall that for , and hence Theorem 3 and Lemma 6 lead to the following.
Corollary 4
Let be a positive integer, and let be an -connected graph of order . If , then has a 2-factor with exactly cycles.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] A. Ainouche, An improvement of Fraisse’s sufficient condition for hamiltonian graphs , J. Graph Theory 16 (1992) 529–543.
- 2[2] J.A. Bondy, A remark on two sufficient conditions for hamilton cycles , Discrete Math. 22 (1978) 191–193.
- 3[3] J.A. Bondy, Longest paths and cycles in graphs with high degree , Research Report CORR 80-16, Department of Combinatorics and Optimization, University of Waterloo, Waterloo, Ontario, Canada (1980).
- 4[4] J.A. Bondy, U.S.R. Murty, “Graph Theory,” Springer-Verlag, London, 2008.
- 5[5] S. Brandt, G. Chen, R. Faudree, R.J. Gould, L. Lesniak, Degree conditions for 2-factors , J. Graph Theory 24 (1997) 165–173.
- 6[6] G. Chen, R.J. Gould, K. Kawarabayashi, K. Ota, A. Saito, I. Schiermeyer, The Chvátal-Erdős condition and 2-factors with a specified number of components , Discuss. Math. Graph Theory 27 (2007) 401–407.
- 7[7] G. Chen, A. Saito, S. Shan, The existence of a 2-factor in a graph satisfying the local Chvátal-Erdős condition , SIAM J. Discrete Math. 27 (2013) 1788–1799.
- 8[8] V. Chvátal, P. Erdős, A note on hamiltonian circuits , Discrete Math. 2 (1972) 111–113.
