Graphs with girth and without longer odd holes that contain an odd -subdivision
Abstract
We say that a graph has an odd -subdivision if some subgraph of is isomorphic to a -subdivision which if embedded in the plane the boundary of each of its faces has odd length and is an induced cycle of . For a number , let denote the family of graphs which have girth and have no odd hole with length greater than . Wu, Xu and Xu conjectured that every graph in is 3-colorable. Recently, Chudnovsky et al. and Wu et al., respectively, proved that every graph in and is 3-colorable. In this paper, we prove that no -vertex-critical graph in has an odd -subdivision. Using this result, Chen proved that all graphs in are 3-colorable.
Key Words: chromatic number; odd holes.
1 Introduction
All graphs considered in this paper are finite, simple, and undirected. A proper coloring of a graph is an assignment of colors to the vertices of such that no two adjacent vertices receive the same color. A graph is -colorable if it has a proper coloring using at most colors. The chromatic number of , denoted by , is the minimum number such that is -colorable.
The girth of a graph , denoted by , is the minimum length of cycles in . A hole in a graph is an induced cycle of length at least four. An odd hole means a hole of odd length. For any integer , let be the family of graphs that have girth and have no odd holes of length at least . Robertson conjectured in [4] that the Petersen graph is the only graph in that is 3-connected and internally 4-connected. Plummer and Zha [5] disproved Robertson’s conjecture and proposed the conjecture that all 3-connected and internally 4-connected graphs in have bounded chromatic numbers, and proposed the strong conjecture that such graphs are 3-colorable. The first was proved by Xu, Yu, and Zha [9], who proved that all graphs in are 4-colorable. The strong conjecture proposed by Plummer and Zha in [5] was solved by Chudnovsky and Seymour [2]. Wu, Xu, and Xu [7] showed that graphs in are 4-colorable and conjectured
Conjecture 1.1.
([7], Conjecture 6.1.) For each integer , every graph in is -colorable.
We say that a graph has an odd -subdivision if some subgraph of is isomorphic to a -subdivision which if embedded in the plane the boundary of each of its faces has odd length and is an induced cycle of . Note that an odd -subdivision of maybe not induced. However, when for each integer , all odd -subdivisions of are induced by Lemma 2.6 (2). In this paper, we prove the following theorem.
Theorem 1.2.
No -vertex-critical graph in has an odd -subdivision.
2 Preliminaries
A cycle is a connected -regular graph. Let be a graph. A vertex is called a degree- vertex if it has exactly k neighbours. For any , let be the subgraph of induced on . For subgraphs and of , set and . Let denote the subgraph of whose vertex set is and edge set is . Let denote the subgraph of with edge set and without isolated vertex. Let be the set of vertices in that have a neighbour in . Set .
Let be an -path and be a -path. When and are internally disjoint, let denote the -path . Evidently, is a path when , and is a cycle when . Let denote the set of internal vertices of . When , let denote the subpath of with ends . For simplicity, we will let denote .
A graph is -vertex-critical if but for each . Dirac in [3] proved that every -vertex-critical graph is -edge-connected. Hence, we have
Lemma 2.1.
For each integer , each -vertex-critical graph has no -edge-cut.
A theta graph is a graph that consists of a pair of distinct vertices joined by three internally disjoint paths. Let be a hole of a graph . A path of is a chordal path of if and is an induced theta-subgraph of . Lemma 2.2 will be frequently used.
Lemma 2.2.
Let be an integer and be an odd hole of a graph . Let be a chordal path of , and be the internally disjoint paths of that have the same ends as . Assume that and have the same parity. If , then and all chordal paths of with the same ends as have length .
Proof.
Since , and and have the same parity, is an odd hole. Moreover, since and all odd holes in have length , we have and , so and all chordal paths of with the same ends as have length . ∎
Let be a path with vertices. If is disconnected, then we say that is a -cut. Usually, a -cut is also called a -cut. Evidently, every -vertex-critical graph has no -cut. Chudnovsky and Seymour in [2] proved that every -vertex-critical graph in has no -cut. Using the same argument as [2], Wu et al. [8] extend this result to graphs in . Since the paper [8] does not include a proof of Lemma 2.3, we give a proof here for completeness.
Lemma 2.3.
([8]) For any number , every -vertex-critical graph in has neither a -cut nor a -cut.
Proof.
It is well-known that every -vertex-critical graph has no clique as a cut. Hence, it suffice to show that every -vertex-critical graph in has no -cut. Let be a -vertex-critical graph. Assume to the contrary that is a path such that is disconnected. Since has no as its cut, . Let be the a component of , and let be the union of all other components. Set for . Since is -vertex-critical, both and are 3-colorable. Let be a 3-coloring for . By symmetry we may assume that and for . Thus . If , then is 3-colorable, which is a contradiction. Thus by symmetry we may assume that and . Let be the subgraph of induced on the set of vertices with . If belong to different components of , then by exchanging colors in the component containing , we obtain another 3-coloring of that can be combined with to show that is 3-colorable. So belong to the same component of . Then there is an induced -path in having even length as . Similarly, there is an induced -path in having odd length as and . Moreover, since are cycles of and , we have and , so is an odd hole of of length at least , which is a contradiction as . ∎
Lemma 2.4.
Let be an integer and be non-adjacent vertices of a graph . Let be an induced -path of . If and all induced -paths have length , then no block of contains two non-adjacent vertices in . In particular, each vertex in is a cut-vertex of .
Proof.
Assume not. Then there is a block of containing two consecutive edges of . Let be an induced path in with ends in and with . Since every pair of edges in a 2-connected graph is contained in a cycle, such a exists. Without loss of generality we may further assume that is chosen with as small as possible. Let be the unique cycle in . Then is an -path. Since is induced, the ends of are not adjacent. Moreover, since is chosen with as small as possible, is an induced -path, so by the assumption of the lemma. Hence, , contrary to the fact . ∎
Lemma 2.5.
Let be an integer and be non-adjacent vertices of a graph . Let be a vertex cut of with , and be an induced subgraph of whose vertex set consists of and the vertex set of a component of . If all induced -paths in have length with , then has a degree- vertex, a -cut, or a -cut.
Proof.
Assume that has no degree-2 vertices. Let be an induced -path in . Let be a subpath of . Such exists as . By the definition of , we have , so , which implies . By applying Lemma 2.4 to , there is a block of such that either , or is not isomorphic to and is or . When the first case happens, since , and , we have , for otherwise or is contained in a cycle of , so the vertex is a cut-vertex of as is a vertex cut of . When the latter case happens, by symmetry we may assume that . Since is a block of , and is a subpath of , we have , so is a -cut of and . ∎

Let be a graph that is isomorphic to a subdivision of , and let be a path of whose ends are degree-3 vertices in . If contains no degree-3 vertex of , then we say that is an arris of . Evidently, there are exactly six arrises of . See Figure 1.
Lemma 2.6.
For any integer , if a graph has an an odd -subdivision , then the following statements hold.
-
(1)
Each pair of vertex disjoint arrises in have the same length and their lengths are at most .
-
(2)
is an induced subgraph of .
-
(3)
When , every vertex in has at most one neighbour in .
Proof.
Without loss of generality we may assume that is pictured as the graph in Figure 1. First, we prove that (1) is true. Assume that . Since and are odd holes, . Hence, , which is a contradiction to the fact that and are both odd holes. So . By symmetry each pair of vertex-disjoint arrises have the same length. Moreover, since is an even cycle with length at least , we have . By symmetry we have . So (1) holds.
Secondly, we prove that (2) is true. Suppose not. Since odd holes have no chord, by symmetry we may assume that there is an edge in with and . On one hand, since and are cycles, by (1) we have
On the other hand, since by (1), we have , implying that . Moreover, by the symmetry between and , we have , which is a contradiction as . So (2) holds.
Finally, we prove that (3) is true. Suppose to the contrary that some vertex has at least two neighbours in . Since a vertex not in an odd hole can not have two neighbours in the odd hole, has exactly two neighbours in . By symmetry we may further assume that and . Since and are cycles whose lengths have different parity,
Hence, and as by (1). This implies that is a chordal path of with length 3, which is a contradiction to Lemma 2.2 as . ∎
By Lemma 2.6 (1), all odd -subdivisions of a graph have exactly edges for each number .
3 Proof of Theorem 1.2
Let be vertex disjoint induced subgraphs of a graph . An induced -path is a direct connection linking and if is the only vertex in having a neighbour in for each . Evidently, and the set of internal vertices of each shortest path joining and induces a direct connection linking and .
For convenience, Theorem 1.2 is restated here in another way.
Theorem 3.1.
Let be an integer, and be a graph in . If is -vertex-critical, then has no odd -subdivisions.
Proof.
Suppose not. Let be a subgraph of that is isomorphic to an odd -subdivision and pictured as the graph in Figure 1. By Lemma 2.6 (2), is an induced subgraph of . By Lemma 2.6 (1), we have
Without loss of generality we may assume that are longest arrises in .
Let be the edges of incident with , respectively. Since is 4-vertex-critical, is not an edge-cut of by Lemma 2.1, so there is a direct connection in linking and . Let be the ends of with having a neighbour in and having a neighbour in . By Lemma 2.6 (3), both and have a unique neighbour in . Let be the neighbours of and in , respectively. That is, and . Set . Since is an induced subgraph of , so is .
3.1.1.
.
Subproof..
Suppose not. By symmetry we may assume that . Set . Since is an odd hole, by symmetry we may assume that is an even hole and is an odd hole. Since is a chordal path of , by (3.1) and Lemma 2.2, we have . So by (3.1) again. Since is a chordal path of and is an odd hole, by (3.1) and Lemma 2.2 again. Moreover, since and , we have , which is not possible. So . ∎
Set . We say that is the difference of . Without loss of generality we may assume that among all odd -subdivisions, is chosen with difference as small as possible.
3.1.2.
.
Subproof..
Suppose to the contrary that . Then by 3.1.1. Without loss of generality we may assume that . Set . Since is an odd hole, either or is an odd hole. Suppose that is an odd hole. Since is an odd -subdivision, by Lemma 2.6 (1) and (3.1), we have , , and . So is an even hole of length by (3.1) again, implying as . Then , contrary to the fact . So is an odd hole.
Since is an odd -subdivision, it follows from Lemma 2.6 (1) and (3.1) that
Then . Since is not an odd hole,
When , since and , we have by (3.2), implying and as are longest arrises in . Hence, . Then is an odd -subdivision with difference , which is a contradiction to the choice of . So . Assume that . Then by (3.1). Since , the graph is an odd -subdivision whose difference is at most , which is a contradiction to the choice of as . So . Then by (3.3), implying by (3.2). Hence, by (3.1) and (3.2), and so by (3.1) again. Since , we have and by (3.1), which is a contradiction as . ∎
3.1.3.
When , the vertices and are adjacent, that is, .
Subproof..
3.1.4.
When , we have that , , , and is an odd hole.
Subproof..
Set . We claim that is an odd hole. Assume to the contrary that is an even hole. Since , the subgraph is an induced theta subgraph. Hence, by (3.1) and Lemma 2.2. Similarly, . Since is a chordal path of , we get a contradiction to Lemma 2.2. So the claim holds, implying that is an even hole.
Since , the graph is an induced theta subgraph of . Moreover, since is an even hole, by (3.1) and Lemma 2.2. Hence, by (3.1) again. Assume that are not adjacent. Since is an induced theta subgraph of , we have , implying . Since is a chordal path of and is an odd hole, by Lemma 2.2, a contradiction. Hence, . By symmetry we have . This proves 3.1.4. ∎
3.1.5.
Assume that has the structure stated as in 3.1.4. Then no vertex in has two neighbours in .
Subproof..
Assume to the contrary that some vertex has two neighbours in . Since no vertex has two neighbours in an odd hole, it follows from Lemma 2.6 (3) and 3.1.4 that has exactly two neighbours in with and . When , let be the unique -path in . Since is a direct connection in linking and , by 3.1.2-3.1.4 and the symmetry between and , we have , contrary to the fact . So . Moreover, since and , we have . Let be the neighbour of in . When , since is a direct connection in linking and , which is not possible by 3.1.4 and the symmetry between and . So . Then is a chordal path of with length 3, contrary to Lemma 2.2. ∎
3.1.6.
and .
Subproof..
By 3.1.1-3.1.3, it suffices to show that . Assume not. By symmetry we may assume that . By 3.1.4, we have that
Since no 4-vertex-critical graph has a -cut by Lemma 2.3, to prove that 3.1.6 is true, it suffice to show that is a -cut of . Suppose not. Let be a shortest induced path in linking and . Let and be the ends of with . By 3.1.5, and no vertex in has a neighbour in .
We claim that . Suppose not. By symmetry we may assume that . Let be the induced -path in . When has no neighbour in , set and . When has a neighbour in , let be a neighbour of closest to and set and . Note that is a hole, although maybe not a hole. Since is an odd hole with length at least when is an odd cycle, to prove the claim, it suffices to show that is odd. When has a neighbour in , since by (3.1) and the fact that , the subgraph is an even hole, which implies that is an odd cycle. So we may assume that has no neighbour in . When is an end of , since is a chordal path of , it follows from Lemma 2.2 and (3.1) that is an odd hole. When is an end of and is a chordal path of , for the similar reason, is an odd hole. Hence, we may assume that is an end of and is not a chordal path of , implying and . Since , we have , so is an even hole, implying that is an odd cycle. Hence, this proves the claim.
By symmetry we may therefore assume that . Let be the induced -path in . By 3.1.1 and 3.1.2, either and has no neighbour in or some vertex in has a neighbour in . No matter which case happens, we have . That is, is a chordal path of with length at least , which is a contradiction to Lemma 2.2 as are non-adjacent. Hence, is a -cut of . ∎
By 3.1.6, there is a minimal vertex cut of with and . Let be the induced subgraph of whose vertex set consists of and the vertex set of the component of containing . Since , we have by (3.1). If all induced -paths in have length , by Lemma 2.5, has a degree-2 vertex, a -cut or a -cut, which is not possible as is 4-vertex-critical. Hence, to finish the proof of Theorem 3.1, it suffices to show that all induced -paths in have length .
Let be an arbitrary induced -path in . When , since is a chordal path of by Lemma 2.6 (3) and the definition of , we have by Lemma 2.2, so by (3.1). Hence, by (3.1) we may assume that and . Since is an induced -path of length , either or and has the same parity as . Assume that the latter case happens. Without loss of generality we may further assume that is chosen with length at least and with as small as possible. Since and have different parity, is not bipartite. Moreover, by the choice of , the subgraph contains a unique cycle and is odd. Since is an induced path of length at least , we have . So is an odd hole of length at least , which is not possible. ∎
4 Acknowledgments
The authors thank the two referees for their careful reading of this manuscript and pointing out an error in our original version.
References
- [1] R. Chen, Graphs with girth and without longer odd holes are -colorable, arXiv:2301.00112.
- [2] M. Chudnovsky, P. Seymour, Proof of a conjecture of Plummer and Zha, J. Graph Theory, 2023, 103: 437-450.
- [3] G. Dirac, The structure of k-chromatic graphs, Fund. Math., 1953, 40: #P42.55.
- [4] D. Nelson, M. Plummer, N. Robertson, X. Zha, On a conjecture concerning the Petersen graph. Electron. J. Combin., 2011, 18: #P20.
- [5] M. Plummer, X. Zha, On a conjecture concerning the Petersen Graph: Part II. Electron. J. Combin., 2014, 21: #P1.34.
- [6] Y. Wang, R. Wu, Graphs with girth 9 and without longer odd holes are 3-colorable, arXiv:2307.01460.
- [7] D. Wu, B. Xu, Y. Xu, On coloring of graphs of girth without longer odd holes (in Chinese), Sci Sin Math., 2023, 53: 103-120.
- [8] D. Wu, B. Xu, Y. Xu, The chromatic number of heptagraphs, 2022, arXiv: 2206.01400v1.
- [9] B. Xu, G. Yu, X. Zha, A note on chromatic number and induced odd cycles. Electron. J. Combin., 2017, 24(4): #P4.32.