Isfahan University of Technology, Isfahan, [email protected]://orcid.org/0000-0003-4401-2110Institute for Research in Fundamental Sciences (IPM), Tehran, [email protected] research was supported by a grant from IPM. \CopyrightRamin Javadi and Meysam Miralaei\hideLIPIcs
A Conjecture on Rainbow Hamiltonian Cycle Decomposition
Abstract
Wu in 1999 conjectured that if is a subgraph of the complete graph with edges, then there is a Hamiltonian cycle decomposition of such that each edge of is in a separate Hamiltonian cycle. The conjecture was partially settled by Liu and Chen (2023) in cases that , is a linear forest, or . In this paper, we settle the conjecture completely. This result can be viewed as a complete graph analogous of Evans conjecture and has some applications in linear arboricity conjecture and restricted size Ramsey numbers.
keywords:
Hamiltonian Cycle Decomposition, Evans Conjecture, Linear Arboricity, Size Ramsey Numbercategory:
\supplement1 Introduction
An (edge) decomposition of a graph is a partition of the edge set into disjoint subsets ’s. By we mean the number of edges in . Given a decomposition and a subgraph of , we say that is rainbow in if no two edges of are contained in the same set , for some . A Hamiltonian cycle (resp. path) decomposition of a graph is a decomposition of into subsets such that each induces a Hamiltonian cycle (resp. path) in .
It is well-known that the complete graph has a decomposition into Hamiltonian cycles. In 1982, Hilton [8], in a seminal work, proposed a procedure for generating all Hamiltonian cycle decompositions of and found necessary and sufficient conditions for extending a decomposition of (for some ) to a Hamiltonian cycle decomposition of .
Wu [14], while studying linear arboricity conjecture, raised the question that for any subgraph of with edges, if the complete graph has a Hamiltonian cycle decomposition in which is rainbow?
Conjecture 1.
[14] Let be a subgraph of with edges. Then, admits a Hamiltonian cycle decomposition such that is rainbow in .
In [11], Liu and Chen partially proved Conjecture 1 when , or has at most vertices, or is a linear forest, where a linear forest is a disjoint union of paths.
Theorem 1.1.
[11] Suppose that is a subgraph of with edges such that either
-
•
has at most vertices, or
-
•
is a linear forest, or
-
•
.
Then, there is a Hamiltonian cycle decomposition of such that is rainbow in .
There are also some partial results regarding Conjecture 1 in [9]. In this paper, we settle Conjecture 1 completely. The conjecture has some applications in linear arboricity conjecture as well as Ramsey theory.
The linear arboricity of a graph denoted by is the minimum number such that the edge-set of can be decomposed into linear forest. Linear arboricity conjecture [1] asserts that if has maximum degree , then (see e.g. [7, 10] for the literature). A graph is called critical if and for any edge , . One can check that if Conjecture 1 is true, then removing any edges from yields a critical graph. It can also be seen that if the conjecture is true, then for every graph with vertices and at most edges, (see [9]).
Conjecture 1 can also be viewed as a complete graph similar result of Evans conjecture [6]. In fact, Evans conjecture (proved by Smetianuk [13]) equivalently asserts that a proper edge coloring of any subgraph of with edges, can be extended to a proper edge coloring of . Andersen and Hilton [2] proved analogous results regarding proper edge coloring of complete graphs. In particular, they proved that
Theorem 1.2.
[2]
-
•
Let be a subgraph of with edges. Any proper edge coloring of can be extended to a proper edge coloring of with colors.
-
•
Let be a subgraph of with edges. Any proper edge coloring of can be extended to a proper edge coloring of with colors.
One can view Conjecture 1 as an analogous result of Theorem 1.2 where proper edge coloring is replaced with Hamiltonian cycle decomposition and is colored with different colors. Therefore, finding necessary and sufficient conditions for extending an arbitrary coloring of to a Hamiltonian cycle decomposition of is an interesting problem and can be viewed as a generalization of Conjecture 1 (see Concluding Remarks and Conjecture 3).
Another application of Conjecture 1 is in Ramsey theory. Given graphs , the Ramsey number is the minimum number such that in every edge coloring of with colors, there is a copy of with color , for some . The restricted size Ramsey number denoted by is the minimum number such that there is a graph with vertices and edges, such that in every edge coloring of with colors, there is a copy of with color for some . Miralaei and Shahsiah in [12] proved some upper bounds for the restricted size Ramsey numbers of stars versus cliques and conjectured that equality holds for these upper bounds. They mentioned that if Conjecture 1 is correct, then their conjecture is proved for the case that the number of stars of size even is positive and even. So, combining our result with their proof can yield the following result.
Theorem 1.3.
Let be positive integers and be the number of even integers in the set . If is positive and even, then
where .
In order to prove Conjecture 1, we need the following result from [8] which deals with the extension of an edge decomposition of to a Hamiltonian cycle decomposition of .
Theorem 1.4.
[8] Let be an integer. A decomposition of can be extended to a Hamiltonian cycle decomposition of where , for each , if and only if the induced subgraph of on each is a linear forest with at most disjoint paths counting a vertex of with no edge in as a path of lengths .
Note that any linear forest on vertices with edges has exactly connected components. Therefore, Theorem 1.4 can be reformulated as follows.
Corollary 1.5.
[8] Let be an integer. An edge decomposition of can be extended to a Hamiltonian cycle decomposition of where , for each , if and only if
-
1.
for each , the induced subgraph of on is a linear forest, and
-
2.
for each , .
To prove Conjecture 1, we take a minimal counterexample to the conjecture with edges, in the sense that is the smallest integer for which the conjecture does not hold and among all counterexamples with edges, is the graph with the smallest number of vertices. We give rise to a contradiction in two phases. First, in Section 2, we embed the components of which are non-isomorphic to into an edge decomposition of for some with desired properties and then in Section 3, we extend this decomposition to a Hamiltonian cycle decomposition of in which is rainbow. This comes to a contradiction and proves Conjecture 1.
1.1 Notations and Terminologies
For positive integers , , the sets and are respectively denoted by and . Given a graph , and respectively stand for and . For a subset , the induced subgraph of on is denoted by . For any vertex , the set of neighbors of in and are respectively denoted by and . Also, and stand for the number of edges incident with in and , respectively. For two disjoint subsets , stands for the number of edges in with one endpoint in and one endpoint in and we define . Also, we define . If is a subgraph of , then is the graph obtained from by removing all edges of .
2 Embedding Components Non-isomorphic to
Let be a minimal counterexample to Conjecture 1 with . Also, let be the subgraph of consisting of all connected components of which are non-isomorphic to . In this section, we prove that there exists an edge decomposition of the complete graph on vertices with some desired properties in which is rainbow. Then, in the next section, we extend such decomposition to a Hamiltonian cycle decomposition of in which is rainbow.
Theorem 2.1.
Let be a minimal counterexample to Conjecture 1 with edges. Also let be the subgraph of containing all its connected components which are not isomorphic to and suppose that has vertices and edges. Then, the complete graph admits an edge decomposition such that
-
(1)
For each , induces a linear forest in .
-
(2)
is rainbow in .
-
(3)
For each , .
-
(4)
For each , .
Proof 2.2.
Due to Theorem 1.1, is not a linear forest and . Also, every connected component of has at least two edges. Thus, has at most connected components and so . Also, we can assume that . To see this, suppose that . Then, we can extend to a graph such that . Thus, by Theorem 1.1, admits a Hamiltonian cycle decomposition such that is rainbow in . Without loss of generality, we can assume that is rainbow in . Also, since , we can remove vertices from to obtain an edge decomposition of where is rainbow in . Conditions (3) and (4) trivially hold as . Therefore, we can assume that .
Now, let and be a subgraph of with . Since is a minimal counterexample and has fewer edges than , there is a Hamiltonian cycle decomposition for such that is rainbow in . Since , we can remove from one vertex whenever is even and two vertices whenever is odd to obtain an edge decomposition of in which is rainbow and each contains edges when is even and at least edges when is odd. Since , we can assume that is a subgraph of . The subgraph has exactly edges which are distributed in the subsets ’s. Now, define new empty subsets and move each edge of from to a separate set , . Also, define as empty subsets. By abuse of notation, let be the obtained decomposition of where is rainbow in . Also, without loss of generality, assume that . We are going to move some edges from large ’s to small ’s to ensure that each induces a linear forest such that
(1) |
Now, we consider the following two cases.
Case 1. .
First, we claim that . To prove the claim, by the contrary, suppose that . Therefore, , for all . Since before moving edges of , each has at least edges, we have
Thus, . On the other hand, when , the function attains its minimum on or . However, as and . This is a contradiction which proves the claim. Therefore, .
Now, for each , , do the following. Let be the edge of in . Consider the set and let be the unique edge of in . Also, let and be two edges in incident with and , respectively (if exist) such that in , and are in different connected components. Now, move edges from to . Since , this can be done. Since has already contained the edge and we excluded the edges and , so after moving these edges, is a linear forest. Also, the remaining edges in are at least .
Now, for each , , do the following. Consider the set and let be the unique edge of in . Now, move arbitrary edges from to . Since , this can be done. Also, the remaining edges in are at least . On the other hand, for each , , , as .
Hence, every satisfies Condition (1), is a linear forest and we are done.
Case 2. .
It is clear that in this case . First, we claim that , where . To prove the claim, by the contrary, suppose that . Therefore, , for all . Now, define when is even and when is odd. Since before moving edges of , each has at least edges, so
Thus, . On the other hand, when , the function attains its minimum on or . However, , as whenever and whenever . Also, . Note that if , then and . So, , then , and the function attains its minimum on or . However, as and as . This leads to a contradiction and proves the claim. Therefore, . Now, for each , , do the following process.
Suppose that , where is an edge of . We are going to move edges from some sets and to to guarantee Condition (1). First, consider which has at least edges. Also, let be the unique edge of in . Also, we can choose at most two edges in incident with , respectively, such that in , vertices and are in different connected components. Now, let be a set of arbitrary edges in (it exists since ). Move all edges in from to . Then, now induces a linear forest in with edges. Let and be respectively the set of all inner vertices and endpoints of the paths in .
Now, consider and let be the unique edge of in . Now, let be a subset of containing all edges incident with along with at most one edge incident with each vertex in such that if are two endpoints of a path in then and are in different connected components of . Since induces a linear forest, we have . Therefore, since , we have . Finally, move arbitrary edges from to . Choosing such edges guarantees that induces a linear forest with edges.
Finally, for each , , do the following process. First, note that since , we have . First, consider . Also, let be the unique edge of in . We choose an arbitrary subset such that . This can be done since . Now, move all edges in to . Then, is now a linear forest in with edges. Let and be respectively the set of all inner vertices and endpoints of the paths in .
Now, consider and let be the unique edge of in . Now, let be a subset of containing all edges incident with along with at most one edge incident with each vertex in such that if are two endpoints of a path in then and are in different connected components of . Since induces a linear forest, we have . Therefore, since , we have . Finally, move arbitrary edges from to . Choosing such edges guarantees that induces a linear forest with edges.
At the end of this process, for each , and for each , . Also, for each , we have moved from at most edges whenever and at most edges whenever . Therefore, at most edges are moved from . Thus, . Finally, for each , as and . Hence, satisfies the conditions of the theorem. This completes the proof.
3 Embedding Components Isomorphic to
Let be a minimal counterexample to Conjecture 1 with edges. Suppose that , where is the union of all connected components of non-isomorphic to and let and be respectively the number of edges and vertices of . In Theorem 2.1, we proved that can be embedded into an edge decomposition of with certain properties. In this section, we are going to prove that such decomposition can be extended to a Hamiltonian cycle decomposition of in which is rainbow and so Conjecture 1 is settled completely. In fact, we prove the following theorem.
Theorem 3.1.
Let be a graph with edges and vertices, where and also let be an integer. Suppose that admits an edge decomposition such that
-
•
for each , induces a linear forest in ,
-
•
is rainbow in and
-
•
for each , and for each , .
Then, can be extended to a Hamiltonian cycle decomposition of in which the graph is rainbow.
In order to prove Theorem 3.1, we begin with the given decomposition of and apply an idea by Hilton [8] to add vertices of one by one and we try to construct a rainbow matching in the remaining classes , . For this, we need a couple of lemmas. First, let us review some definitions.
For an edge coloring of a multigraph with colors , let be the set of edges of color incident with and let be the set of edges of color with endpoints . An edge coloring of is called a balanced edge coloring if for every vertex ,
and for every pair of vertices ,
We need the following result due to de Werra.
Lemma 3.2.
Let be a bipartite multigraph with bipartition . By a pairing of on , we mean a collection such that each is a family of disjoint two-subsets of the edges incident with . If , then we say and form an -pair. We have also an additional condition that if there are more than one edge between two vertices and , then at most one of these parallel edges is paired with an edge , . We also need the following lemma which can be derived from the proof of Theorem 1 in [8]. For the completeness, we provide a proof of the following lemma in Appendix A.
Lemma 3.3.
[8] Let be a bipartite multigraph with bipartition such that the degree of each vertex in is even. Also, consider a pairing of on . Then, has a balanced edge-coloring with two colors such that if form an -pair for some , then and have different colors.
Finally, we need the following lemma.
Lemma 3.4.
Suppose that is a bipartite multigraph with bipartition whose edge set is partitioned into two sets such that for every , and for every , for some integer . Also, suppose that for some vertex , we have either
-
•
, or
-
•
is odd, is even and for every , is even.
Then, there is a subset such that
-
•
, for all ,
-
•
, and
-
•
, for all .
Proof 3.5.
Let be the set of all vertices such that there is an alternating path from to with edges in and alternately starting from an edge in (note that ).
If for some , , then we set which satisfies all the conditions. Now, suppose that for all , . Let and so . On the other hand, it is clear that and since for every , , we have . If , then we have which is a contradiction. Now, suppose that is odd, is even and for every , is even. Then, is odd and is even which is again a contradiction. This contradiction completes the proof.
Now, we are ready to prove Theorem 3.1.
Proof 3.6 (Proof of Theorem 3.1.).
We start with the edge decomposition of with the described properties and for each , we are going to prove that can be extended to a decomposition of such that is rainbow in and
(2) |
For we have . Suppose that it has been done until some and we do it for .
Set and suppose that the vertex set of is . We define an auxiliary bipartite multigraph with a bipartition and such that is adjacent to with an edge if is an endpoint of a path in and is adjacent to with two parallel edges if is incident with no edge in . Then, we have
(3) | |||||
(4) |
To see (3), for each , let be the number of sets where . Therefore, . On the other hand, is incident with edges in . So, and we have .
To see (4), note that if is empty, then . Now, adding any edge to reduces the degree of in by two (because of its endpoints). Thus, .
Therefore, by (2), for each , we have and for each , . Now, we prove the following claim.
Claim 2.
There exist two disjoint subgraphs of such that for ,
-
i.
for each , ,
-
ii.
for each , and ,
-
iii.
for each , if , for some integer , then
-
iv.
for each , if is adjacent to some vertices and in , then and are not the endpoints of a path in ,
-
v.
if is adjacent to some vertex in and in , then and are not the endpoints of a path in . Also, has no parallel edges.
Proof 3.7 (Proof of Claim 2).
First, let be obtained from by duplicating each edge of . Also, we add a new vertex and join it to all vertices each one with 4 parallel edges. Therefore, for all , and . Also, for each , we have .
By Lemma 3.2, there is a balanced -edge coloring of with colors. Fix a color class . Since and , we have and . Also, for each , if , then (since as ). Moreover, for each , if , then (as and ).
Since , there is a color class say , such that is incident with exactly one edge in . Without loss of generality, suppose that four parallel edges between and are in the color classes and . Now, for each , let be the edge-set obtained from by removing all edges incident with .
We know that . If , then we apply Lemma 3.4 by setting , , , and to obtain a set and define . Also, if , then define . It is clear that and , for all and , for all . With a similar argument, using the color classes and , we can find a color class such that and , for all and , for all . Let . Then, we have
-
•
for all , ,
-
•
for each , if , for some , then ,
-
•
if , for some , then ,
-
•
for each , if , for some , then except for possibly one , say , for which .
Now, we apply Lemma 3.3 for the graph and the following pairing: if there are two parallel edges joining to , then and form an -pair. If is adjacent to and , where and are the endpoints of one path in , then and form an -pair. Also, if there is more than one edge incident with still not paired off, then such edges are paired arbitrarily (one edge may possibly be remained unpaired). Therefore, there is a balanced two-coloring for , called , such that if is adjacent to some vertices and in , , then and are not the endpoints of a path in . Now, since , one of the sets , say , satisfies . Finally, applying Lemma 3.2 to the graph , we find a balanced two-edge-coloring . Setting and , we obtain the desired subgraphs.
Now, we are ready to extend the decomposition of to a decomposition of by adding two new vertices and as follows (see Figure 1). For each , if is adjacent to in , then we add the edge to . Also, we add the edge to . So, by abuse of notation, we obtain a decomposition for . It is clear that is rainbow in . Moreover, because of Conditions (ii), (iv) and (v) of Claim 2, each induces a linear forest in . Now, we check Condition (2). First, for each , if for integer , then, by (4) and Condition (iii) of Claim 2, we have
Note that when , we have included the new edge into and so we have at least new edges therein. Hence, inductively we can construct an edge decomposition for such that
-
•
for each , induces a linear forest in ,
-
•
is rainbow in and
-
•
for each , .
Now, we apply Corollary 1.5 to extend to a Hamiltonian cycle decomposition of in which is rainbow. This completes the proof.
Combining Theorems 2.1 and 3.1 settles Conjecture 1 as follows. Let be a minimal counterexample to Conjecture 1, where and has no component isomorphic to . Then, by Theorem 2.1, we can construct an edge decomposition with the desired properties and by Theorem 3.1, we can extend it to a Hamiltonian cycle decomposition of in which is rainbow. This contradicts the fact that is a counterexample. So, Conjecture 1 is settled.
4 Concluding Remarks
In this paper, we proved a conjecture by Wu asserting that for any subgraph of with edges, there is a Hamiltonian cycle decomposition for in which is rainbow. As it was mentioned in the introduction, this result is, in some point of view, a complete graph counterpart of Evans conjecture. In other words, the conjecture asserts that if the edges of are colored by different colors, then we can extend such coloring to a coloring of such that each color class induces a Hamiltonian cycle. So, one may ask if its generalization to an arbitrary coloring of is correct. We state this assertion as the following conjecture.
Conjecture 3.
Let be a subgraph of with edges which are colored such that each color class induces a linear forest. Then, the coloring can be extended to a coloring of with colors such that each color class induces a Hamiltonian cycle.
Note that Conjecture 1 is a special case of Conjecture 3 when the edges of are colored with different colors. Also, it is noteworthy that in Conjecture 3, the number for the number of edges of is optimal, because if we consider as a subgraph of and color the edges of with color and the edge of with color , then such a coloring cannot be extended to a Hamiltonian cycle decomposition of .
References
- [1] J. Akiyama, G. Exoo, and F. Harary. Covering and packing in graphs IV: Linear arboricity. Networks, 11(1):69–72, 1981.
- [2] L. D. Andersen and A. J. W. Hilton. Symmetric Latin square and complete graph analogues of the evans conjecture. Journal of Combinatorial Designs, 2(4):197–252, 1994.
- [3] D. de Werra. Balanced schedules. INFOR: Information Systems and Operational Research, 9(3):230–237, 1971.
- [4] D. de Werra. A few remarks on chromatic scheduling. In Combinatorial Programming: Methods and Applications: Proceedings of the NATO Advanced Study Institute held at the Palais des Congrès, Versailles, France, 2–13 September, 1974, pages 337–342. Springer, 1975.
- [5] D. de Werra. On a particular conference scheduling problem. INFOR: Information Systems and Operational Research, 13(3):308–315, 1975.
- [6] T. Evans. Embedding incomplete Latin squares. The American Mathematical Monthly, 67(10):958–961, 1960.
- [7] A. Ferber, J. Fox, and V. Jain. Towards the linear arboricity conjecture. Journal of Combinatorial Theory, Series B, 142:56–79, 2020.
- [8] A. J. W. Hilton. Hamiltonian decompositions of complete graphs. Journal of Combinatorial Theory, Series B, 36(2):125–134, 1984.
- [9] T. Khoplyklang. Hamiltonian decompositions and linear arboricity of graphs. PhD thesis, Queen Mary University of London, 2023.
- [10] R. Lang and L. Postle. An improved bound for the linear arboricity conjecture. Combinatorica, pages 1–23, 2023.
- [11] Y. Liu and Y. Chen. Rainbow subgraphs in Hamiltonian cycle decompositions of complete graphs. Discrete Mathematics, 346(8):113479, 2023.
- [12] M. Miralaei and M. Shahsiah. On the multicolor size Ramsey number of stars and cliques. Discrete Mathematics, 343(8):111899, 2020.
- [13] B. Smetianuk. A new construction on Latin squares I. A proof of the Evans conjecture, Ars Combinatorica, 11:155–172, 1981.
- [14] J. Wu. The linear arboricity theory of graphs. PhD thesis, Shandong University, 1999.
Appendix A Proof of Lemma 3.3
Here we give a proof of Lemma 3.3. Let be a bipartite multigraph with bipartition such that and the degree of each vertex in is even. Also, consider a pairing of on and extend it as follows. If there are two edges incident with a vertex which are unpaired, then form a -pair and do this until there is at most one unpaired edge on each vertex of .
We proceed by induction on . There is nothing to prove for . So, assume that . Now, we claim that there is an even cycle or an even path (a path with even edges) in , say , satisfying the following properties:
-
1.
if is a path, its endpoints are in ,
-
2.
for every vertex , if , then the edges of incident with form an -pair,
-
3.
there is a balanced 2-edge coloring on with the color classes and such that the edges of are alternately in and .
In order to prove the claim, select an arbitrary edge and place it in . If has an -pair, say , place in . Since the degree of is even, it has still at least one unassigned edge, let it be and place in . Continue in this way until one of the following occurs.
-
•
An edge is found (and placed in ), where for some . In this case we have an even cycle and the claim is proved.
-
•
An edge is found (and placed in ) and has no -pair; then let be the endpoint of this path. Now, there is at least one unassigned edge on . Let this edge be and place it in . If has no -pair, let be the other endpoint of the path . If has an -pair, let it be and place it in . The vertex still has an unassigned edge on it, say . Place in . Continue in this way until an edge is found (and placed in ) such that the edge has no -pair. Then, let be the other endpoint of the path constructed.
Let be the spanning subgraph of obtained by deleting the edges of . By the induction hypothesis, has a balanced -edge coloring with the color classes and satisfying the assertion. Clearly, the edge coloring on with the color classes and with and is the desired balanced -edge coloring on .