Factoring complete graphs and hypergraphs into factors with few maximal cliques
Abstract
For integers and let be the minimum, over all factorizations of the complete -uniform hypergraph of order into factors , of where is the number of maximal cliques in . It is known that ; in fact, if is a graph of order , then with equality iff where is the clique number and the independence number. In this paper we investigate when or . We also characterize graphs of order with .
1 Introduction
Let denote the number of maximal cliques and the number of maximal anticliques in a graph . In an earlier paper [2] two of the present authors showed that, if is a graph of order , then with equality just in case where is the clique number and the independence number. (Proofs of these facts will be given here; see Theorem 28 and Theorem 34(a).) In this paper we generalize the problem in two ways: by considering factorizations of the complete graph into factors instead of just the two factors and , and by considering uniform hypergraphs instead of graphs. We also characterize the graphs of order with .
An -uniform hypergraph (in this paper ) is a structure comprising a nonempty finite set of vertices and a set of edges; is the order and is the size of . (A -uniform hypergraph is a graph; a -uniform hypergraph is a triple system.) A set is a clique if , an anticlique if . The complement of is the hypergraph . We write for the number of maximal cliques and for the number of maximal anticliques in . We write for the number of maximal cliques and for the number of maximal anticliques containing a vertex . We define and . The complete -uniform hypergraph of order is where . Note that and . An edge of a graph may be written as if no confusion will result. The neighborhood of a vertex in a graph is the set , and we write for . The clique number of a graph is the maximum number of vertices in a clique of ; the independence number is the maximum number of vertices in an anticlique of . See West [4] for graph-theoretic notation and terminology not defined here.
For integers and we define as the minimum of over all factorizations of into edge-disjoint spanning subgraphs . In this notation, part of the aforementioned result of [2] may be expressed as follows. (A slightly stronger result will be proved here as Theorem 28.)
Theorem 1.
.
Sometimes we find it convenient to use the language of colorings: is the least possible number of maximal monochromatic cliques in an edge coloring of with colors, it being understood that a set which is a maximal monochromatic clique for more than one color is counted once for each color; e.g., . Most of this paper is a study of the function . We have determined exact values only in some very special cases; mostly we have upper and lower bounds which are far apart.
In Β§2 we establish some general facts about the function , most notably an asymptotic lower bound: for fixed integers and any we show that for all sufficiently large (Theorem 7).
In Β§3 we find upper bounds for , the minimum number of maximal cliques in a triple system of order and its complement. The main result here is that . For this is definitely not the best possible, e.g., is attained by the Fano plane (Theorem 11).
In Β§4 we find upper bounds and a few exact values for when . We show that, if there is a projective plane of order , then for , and for (Theorem 19). For we observe that if and only if a projective plane of order exists (Theorem 20). We obtain an upper bound of the form whenever a projective plane of order or exists (Theorem 21); in particular see Theorem 23 for the case , Theorem 25 for , Theorems 26 and 27 for . We have no counterexample to the conjecture that for and otherwise (Question 24).
In Β§5 we characterize the graphs with (Theorem 34(b)), thereby showing that for all graphs outside of a well-defined class of exceptions. In particular, we show that if (Lemma 33, Corollary 35(c)).
Some of our results (Theorems 6, 7, 8, 17, 20, and 23, perhaps in a less general formulation) were stated without proof in the earlier paper [2] or in the abstract [1]. Some were presented by the third author in invited addresses to the Mid-Atlantic Mathematical Logic Seminar, Dartmouth College, Hanover, New Hampshire, October 2003, and to the CombinaTexas Conference, Houston, Texas, April 2009.
2 Generalities
In this section we establish some basic properties of the function such as monotonicity (Theorem 3), the trivial bounds (Theorem 6), and an asymptotic lower bound (Theorem 7). We begin by disposing of the trivial cases where . Recall that the maximum size of a -partite graph of order is attained by the TurΓ‘n graph .
Theorem 2.
For any integers we have
Proof.
Only the case is in need of a proof. Consider a factorization of . Let and let , so that .
First suppose for all . Then for the set is a maximal clique only for the unique factor having as an edge, while for the set is a maximal clique of just in case and . Hence the factors have a total of maximal cliques of size and maximal cliques of size , so that in this case
(1) |
On the other hand, if , then
the same number of maximal cliques as when . Hence we may assume that for all . The quantity (1) is minimized when is maximized, i.e., when . This shows that . Finally, setting , we have
β
Theorem 3.
.
Proof.
If is an induced subhypergraph of then . β
The inequality is not necessarily strict, e.g., if is odd then by Corollary 15(b). See Theorem 19 for more examples.
Question 4.
Does the inequality hold for all integers and ?
An obvious way to get a -coloring from a -coloring is by merging two colors. This can increase the number of maximal monochromatic cliques, i.e., if and are edge-disjoint graphs on the same vertex set , we can have . For example, let be the set of vertices of a unit cube, and let
Then , , and , so that , , and .
Lemma 5.
Consider a coloring of . For each color , let denote the number of maximal -cliques. Then, for any two colors , we have .
Proof.
Let be the vertex set. For each color let be the set of all maximal -cliques. Each set is an -clique; let be some maximal -clique containing . If , then the map is an injection from to . β
Theorem 6.
For all integers and ,
.
Proof.
For the upper bound, give all edges the same color.
For the lower bound, let be -colored so that where is the number of maximal -cliques. By Lemma 5 we have for . Hence:
β
Theorem 7.
For any integers and , and for any , we have for all sufficiently large .
Proof.
Choose so that . Let and . Define so that
By Ramseyβs theorem there is an integer such that any -coloring of the edges of the complete hypergraph contains a monochromatic clique of size . Let .
Let and consider a -coloring of the edges of . Let be the number of maximal -cliques. Let be the size of the largest monochromatic clique.
If there is a -clique of size , then for each color there are at least maximal -cliques. If then
Let be the number of monochromatic cliques of size . Since every monochromatic clique is contained in a maximal monochromatic clique, . On the other hand, if , then and we have
Thus for sufficiently large . β
3 A triple system and its complement
In this section we find upper bounds for . For our best result is (Theorem 12). We know the exact value of only for . By tedious case analysis (which we omit) we have verified that Theorem 8 is optimal for , so that in those cases, and that Theorem 11 is optimal for , so that . See Theorem 2 for the trivial cases .
Theorem 8.
for .
Proof.
Let where . Let , and consider the triple system where . Then ; the maximal cliques are the sets , , and the -element subsets of . Taking and , we see that
Alternatively, let where and . For integers of the same parity, let be the set consisting of and and all integers of the opposite parity between and ; e.g., . Then the maximal cliques (anticliques) of are just the sets where and are even (odd) integers, . It follows that
β
Lemma 9.
A triple system of order can be extended to a triple system of order such that , , , and .
Proof.
Let be a triple system of order . Choose a vertex with . Choose a new vertex and let where and .
The maximal anticliques of are the maximal anticliques of not containing and the sets where is a maximal anticlique of H containing . Plainly and .
The maximal cliques of are the maximal cliques of , and the sets where is a maximal clique of containing , and the set . Now it can be seen that , and that , while for , whence . β
Lemma 10.
If there is a triple system of order with and and , then there is a triple system of order with and and .
Proof.
By Lemma 9 there is a triple system of order such that and and . Let be the complement of . β
Theorem 11.
for .
Proof.
We prove by induction that for each there is a triple system of order with and and . The inductive step follows from Lemma 10 since
For the base case , consider the Fano plane as a hypergraph H with the points as vertices and the lines as edges. Then the maximal cliques are the lines, and the maximal anticliques are the complements of lines. Thus , and and ; each vertex is in exactly three maximal cliques and four maximal anticliques. β
In fact ; the proof of is a tedious case analysis which we omit. We do not know if equality holds in Theorem 11 when . For a better bound is given by Theorem 12.
Theorem 12.
.
Proof.
The cases with , while easily verified, are of no interest. For we prove by induction that there is a triple system of order with and and . The inductive step follows from Lemma 10 since
Now consider where and .
We can visualize this as the vertices and of the faces of a regular octahedron, the two missing faces meeting at one point. There are maximal cliques, namely, the edges and the antipodal pairs
and there are maximal anticliques, namely, the missing faces
and the complements of the antipodal pairs,
Thus
while
and
The complement of satisfies the requirements of the base case . β
4 Edge-colored graphs and projective planes
The main results of this section are upper bounds for from recursive constructions using projective planes. For example, we use projective planes of orders and to show that for (Theorem 27). From Theorem 20 we see that determining the exact value of in all cases will be at least as hard as determining all possible orders of finite projective planes.
Theorem 13.
For all integers and ,
.
Proof.
Theorem 14.
, with equality if and only if .
Proof.
Let the edges of be colored with colors. Consider the graph where are vertex-disjoint graphs such that is isomorphic to the spanning subgraph of formed by the edges of color . Let be the number of components of . Then
(2) |
For equality to hold in (2) we must have , meaning that each component of is a clique, and also , meaning that is acyclic. So equality holds just in case each component of is or , which means that the given coloring is a proper edge-coloring of . Of course, is -edge-colorable if and only if . β
Corollary 15.
Lemma 16.
Let be a -colored complete graph of order , and let where is the number of maximal -cliques in . If some vertex of is in a unique maximal -clique for each color , then for all .
Proof.
Let be an optimally -colored complete graph of order , so that where is the number of maximal -cliques in . Replace the vertex of with a copy of . In the resulting -colored complete graph of order , the number of maximal -cliques meeting is , while the number of maximal -cliques disjoint from is . β
Theorem 17.
If there is a projective plane of order , then for all .
Proof.
Let be the point set of a projective plane of order . Choose a line and let be the points on . Let be the complete graph of order with vertex set . Color the edges of with colors by assigning color to an edge if the points are collinear. For each color there are maximal -cliques corresponding to the lines other than through the point , and each vertex of is in just one of them. We obtain the desired inequality by setting , , in Lemma 16. β
The following unpublished observation by Graham and Van Lint is included by permission.
Theorem 18.
(Ronald L. Graham and Jack van Lint) If there is a projective plane of order , then for all .
Proof.
Let be the point set of a projective plane of order . Choose a line ; let be the points on , and let be the other lines through . Let be the complete graph of order with vertex set . For any edge of , let be the line through and . Color the edges of with colors by assigning color to an edge if either or else meets at the point . For each color there are maximal -cliques; of them correspond to the lines other than through ; the remaining one, corresponding to the line , is the only maximal -clique containing . We obtain the desired inequality by setting , , , and in Lemma 16. β
Theorem 19.
If there is a projective plane of order , then:
Proof.
On the other hand, if there is a projective plane of order , and if , then
by Theorem 17. This completes the proof of .
Take a projective plane of order . Choose a line with points
and another line through . Let be the complete graph of order whose vertices are the points not on or . Assign the color to an edge of if the line through and meets at . Now maximal -cliques correspond to lines through other than and , so the number of maximal -cliques is if and if , for a total of . Hence, for , we have
This completes the proof of . β
Theorem 20.
For the following statements are equivalent:
Proof.
We have by Theorem 19, and is trivial; we have to show .
Assume that and . Consider a factorization with , where . Since for , the geometric mean of is at least , that is,
whence .
For each , there are exactly maximal -cliques, which are pairwise disjoint and contain vertices each. If is a maximal -clique and a maximal -clique, , then . Hence there is a projective plane of order ; the points are the vertices of and the numbers ; the lines are the set and the sets where is a maximal -clique. β
In view of Theorem 20, a computation by Lam, Thiel, and Swiercz [3] shows that . On the other hand,
by Theorem 18.
Theorem 21.
(Due to Ronald L. Graham and Jack van Lint in the case of a projective plane of order .) If there is a projective plane of order or , then there is a constant such that for all .
Proof.
Question 22.
For each integer , is there a constant such that for all ?
If exists then since, when , we have by Corollary 15. In fact the inequality holds for (Theorems 1, 23, 25, 26). On the other hand it fails for and , since and ; we omit the details.
Theorem 23.
with equality at least for .
Proof.
By Theorem 17 with , to verify the inequality it suffices to observe that it holds for . To verify equality for all , it suffices to show that and ; we omit the details. β
Question 24.
Does equality hold in Theorem 23 for all ?
Theorem 25.
with equality at least for .
Proof.
Use in Theorem 17. β
Theorem 26.
for all .
Proof.
Theorem 27.
for . In fact, if , then
5 When
For a graph we define . In this section we characterize the graphs with . In doing so we also prove the characterization of graphs with which was proved more simply in [2]. We need the following improved version of the result of [2] that for every graph .
Theorem 28.
If is a graph of order , then
.
Proof.
We use induction on . Let be a graph of order ; let , , , . We may assume that there is a vertex such that ; otherwise or is a complete graph and the result is clear. Let and ; Let , , , , . By the inductive hypothesis, .
Choose a vertex of which is in exactly maximal cliques of . The number of maximal cliques of containing is equal to ; the number of maximal cliques containing is at least ; the number of maximal cliques containing neither nor is at least . Since no clique contains both and , Similarly, .
Adding these two inequalities we get
β
We define four classes of graphs, unimaginatively named after their smallest members. As motivation note that is the smallest graph with , while , , and are the smallest graphs with . Recall that is a split graph if is the union of a clique and an anticlique. We begin by defining two classes of split graphs.
A graph is -like if where is a clique, is an anticlique, and ; equivalently, if .
A graph is -like if where is a maximal clique, is a maximal anticlique, and .
Lemma 29.
If is a split graph then just one of the following statements holds:
A graph is -like if there is a -element set such that while , and where is a clique and is an anticlique, each vertex in is joined to and and at least one more vertex in , and no vertex in is joined to any vertex in .
A graph is -like if is -like.
Lemma 30.
If a graph is -like or -like or -like, then and .
Lemma 31.
Let be a graph. If and , then is -like or -like or a split graph.
Proof.
Let and . Let be a clique and an anticlique with . We may assume that ; let , , , . Since the statement of the lemma is invariant under complementation, we may assume that is joined to . We may assume that some vertex is not joined to , and some vertex is joined to , as otherwise is a split graph. Then , since and are contained in different maximal cliques.
Then . From this and we have
(5) |
Since equality holds in (5), we have for each , for each , and every maximal anticlique meets .
If some vertex is not joined to , then, according as is joined to or not, either or , neither of which is possible. Hence is a clique.
Now suppose is joined to some vertex . If is joined to , then , which is impossible; otherwise is contained in a maximal anticlique disjoint from , also impossible. Hence is joined to no vertex in , so is an anticlique and is a split graph.
Then . From this and we have
(6) |
Since equality holds in (6), we have for each , for each , and every maximal anticlique meets .
Each vertex is joined to or to , as otherwise we would have .
If were joined to some vertex , then we would have or according as is joined to or not. Hence is not joined to any vertex in .
Now the anticlique is contained in a maximal anticlique, which must meet but cannot contain . Hence there is a vertex which is not joined to or to any vertex in . Of course is joined to , since .
It is easy to check that is -like, with playing the roles of in the definition. β
Lemma 32.
If is a graph and , where is a vertex such that is a clique and , then .
Proof.
Let and . Then
β
Lemma 33.
If then .
Proof.
We use induction on . Let be a graph, , , so . By Theorem 28, , so or . We may assume that , i.e., some vertex is in a unique maximal clique; i.e., is a clique. We may assume that . Let , ; then . Let ; Then by Lemma 32. By the inductive hypothesis, . By Lemmas 31 and 29, is -like or -like or -like or -like. Since is a clique which meets every maximal anticlique, and since , we have
(7) |
So where is a clique, is an anticlique, and . Then and , so
(8) |
Let and let be the number of maximal cliques disjoint from . Since is an anticlique,
(9) |
Since , it follows that
(10) |
Since , we have to show that , i.e., that where . If is a clique weβre done, so we may assume that some vertex is not joined to some vertex . Then either (if is joined to ) or (otherwise), so , i.e., every maximal clique meets . Hence it will suffice to show that is a clique, since, being disjoint from , itβs not a maximal clique.
So , and for all . Then is joined to every vertex in , since is not joined to , and . Likewise, if and is not joined to , then is joined to every vertex in , since . Finally, suppose there are vertices and such that is joined to and not to . Then the cliques , , and extend to three different maximal cliques, contradicting .
So , for all , and . Suppose ; if is joined to then is joined to every vertex in because ; if is not joined to , then is joined to every vertex in because . Finally, is joined to every vertex in because otherwise we would have .
In Cases 2 β 4 we have by Lemma 30, and we have to show that . Since , we have to show that .
So where is a maximal clique in , is a maximal anticlique in , , and . Then , so (7) becomes
(11) |
The set is an anticlique. Let be the number of maximal cliques disjoint from ; , since is contained in a maximal clique which is disjoint from . Then we have
(12) |
Since , it follows that for all , and for all . We have to show that , i.e., . It will suffice to show that is a clique. Assume for a contradiction that some vertex is not joined to some vertex . Since is a maximal anticlique in , is joined to some vertex . But then either (if is joined to ) or else (otherwise); either way we have a contradiction.
So there is a -element set such that while ; and where is a clique and is an anticlique, each vertex in is joined to and and at least one more vertex in , and no vertex in is joined to any vertex in . Then , , and (7) becomes
(13) |
Since is an anticlique and ,
(14) |
Since , we must have and for all . A similar argument, using the anticlique instead of , shows that . We have to show that , i.e., .
Proof.
Suppose , , not joined to . Then must be joined to and , since . But then the cliques , , extend to three different maximal cliques, contradicting . β
Proof.
If neither nor is a clique, then there are vertices (not necessarily distinct) such that is not joined to and is not joined to . Then and are joined to , because every vertex in is joined to , and if then must be joined to because . But then , , and are cliques extending to three different maximal cliques, contradicting . β
Proof.
Similar to Claim 2, using . β
It follows from Claims 2 and 3 that there are two adjacent vertices such that is a clique, so .
So there is a -element set such that while ; and where is an anticlique and is a clique, no vertex of is joined to or or to more than one vertex in , and every vertex in is joined to every vertex in . Then , and so (7) becomes
(15) |
We have to show that , i.e., that . The set is an anticlique. Let be the number of maximal cliques disjoint from . Then
(16) |
But , so , and for all , and , i.e., every maximal clique meets .
Now is a clique, since and are cliques, and if some vertex were not joined to some vertex , then, since is joined to , we would have either (if is joined to ) or else (otherwise). So is a clique of size , but it canβt be a maximal clique since itβs disjoint from , so . β
Theorem 34.
Let be a graph of order .
Proof.
Corollary 35.
Let be a graph of order .
Acknowledgement
References
- [1] P. ErdΕs, F. Galvin, and M. Krieger, Factoring complete hypergraphs into factors with few cliques, Abstract 750-A23, Notices. Amer. Math. Soc. 24 (1977), A-629.
- [2] F. Galvin and M. Krieger, The minimum number of cliques in a graph and its complement, Proc. Second Louisiana Conf. on Combinatorics, Graph Theory, and Computing (Louisiana State University, Baton Rouge, La., 1971), pp. 345β352.
- [3] C. W. H. Lam, L. Thiel, and S. Swiercz, The non-existence of finite projective planes of order 10, Canad. J. Math. 41 (1989), 1117β1123.
- [4] D. B. West, Introduction to Graph Theory, 2nd edition, PrenticeβHall, Inc., Upper Saddle River, NJ, 2001.