On partial parallel classes in partial Steiner triple systems
Introduction
Suppose is a positive integer. A partial Steiner triple system of order , denoted PSTS, is a pair , where is a set of points and is a set of 3-subsets of , called blocks, such that every pair of points occurs in at most one block. The number of blocks in is usually denoted by . If every pair of points occurs in exactly one block, the PSTS is a Steiner triple system of order , denoted STS. It is clear that a PSTS is an STS if and only if . An STS exists if and only if . Colbourn and Rosa [8] is the definitive reference for Steiner triple systems and related structures.
Let denote the maximum number of blocks in a PSTS. The following well-known theorem gives the values of for all positive integers . (For a discussion of this result, see [8].)
Theorem 1.1 Suppose . Then
Suppose is a PSTS. For any point , the degree of , denoted , is the number of blocks in that contain . A PSTS is -regular if for all . It is clear that for all . Also, an STS is a -regular PSTS with .
Suppose is a PSTS. A parallel class in is a set of disjoint blocks in . A partial parallel class (or PPC) in is any set of disjoint blocks in . The size of a PPC is the number of blocks in the PPC. A PPC in of size is maximum if there does not exist any PPC in of size . Note that a maximum partial parallel class in is not necessarily unique, but the size of a maximum PPC is well-defined.
There has been considerable study concerning the sizes of maximum partial parallel classes in STS. It seems to be quite difficult to find STS with that do not have parallel classes. It was only in 2015 that the first infinite classes were found, by Bryant and Horsley [7]. For STS with that do not have partial parallel classes of size , there are two infinite classes known, one of which was found by Wilson (see [13]) and one discovered by Bryant and Horsley [6].
There are also lower bounds on the sizes of maximum partial parallel classes in STS. The first nontrivial bound is due to Lindner and Phelps [11], who proved in 1978 that every STS with has a partial parallel class of size at least . Improvements have been found by Woolbright [14] and Brouwer [5]. The result proven in [5] is that every STS has a PPC of size at least . In 1997, Alon, Kim and Spencer [1] used probabilistic methods to prove that any STS contains a PPC of size at least . Also, a recent preprint by Keevash, Pokrovskiy, Sudakov and Yepremyan [9] improves this lower bound to .
We should also mention Brouwer’s conjecture, made in [5], that any STS has a PPC of size at least .
The problem of determining lower bounds on the sizes of maximum partial parallel classes in partial Steiner triple systems has received less study. It was shown in [1] that a -regular PSTS has a PPC of size at least .
In this paper, we address the following problem. For an integer such that , define to be the maximum number of blocks in any PSTS in which the maximum partial parallel class has size . We obtain lower bounds on by giving explicit constructions, and upper bounds result from counting arguments.
Before continuing, we observe that, when , the values follow easily from known results concerning STS.
Theorem 1.2 For all , , it holds that . Also, .
Proof For all , it is known that there exists an STS containing a parallel class. (For example, we can use a Kirkman triple system of order ; see [8].) For all , , there exists an STS containing a PPC of size (For , we can use a nearly Kirkman triple system of order ; see [8]. For , it is noted in [8] that both of the two nonisomorphic STS have a PPC of size .) Finally, for , the PSTS consisting of blocks has a maximum PPC of size , and there is no PSTS having six blocks that has a maximum PPC of size . □
As mentioned above, there are STS in which the maximum partial parallel class has size less than . In this situation, we would have , where is the size of the maximum PPC in the given STS. For example, there is an STS having a maximum PPC of size , as well as an STS having a maximum PPC of size . Therefore, and .
We summarize the main contributions of this paper. In Section 2, we present constructions for partial Steiner triple systems that have maximum PPCs of a prespecified size, thus obtaining lower bounds on . In Section 3, we prove an upper bound on using a counting argument. In Section 4, we show that if is a constant, and if , where is a constant. When is a constant, our upper and lower bounds on differ by a constant that depends on . In Section 5, we apply our results to the problem of finding sequenceable PSTS. In particular, the constructions we describe in Section 2 provide infinite classes of sequenceable PSTS. Finally, Section 6 is a brief summary.
Section snippets
PSTS with small maximum PPCs
We present a construction for partial Steiner triple systems that have maximum PPCs of a prespecified size. This construction utilizes Room squares, which we define now. Suppose that is even and let be a set of size . A Room square of side on symbol set is an by array that satisfies the following conditions:
- 1.
every cell of either is empty or contains an edge from the complete graph on vertex set ,
- 2.
every edge of occurs in exactly one cell of ,
- 3.
the filled cells in every row
An upper bound on
In this section, we prove an upper bound on using a simple counting argument. The proof of this bound depends on a very useful observation due to Lindner and Phelps [11]. Suppose is a PSTS in which the maximum partial parallel class has size ; hence . Let be a set of disjoint blocks and let .
For , let denote the set of blocks that contain and two points in . Then induces a partial one-factor of . Define ; then .
Analyzing the bounds
First, we observe that we can compute the exact value of for any .
Theorem 4.1 Suppose . Then
Proof For any , we have from (1). When , (3) simplifies to . Applying Theorem 2.4, we see that for all . Therefore for . For , (3) yields . For , if we take the seven blocks of an STS, we see that . Hence for . The cases can be analyzed
Sequencings of PSTS
A PSTS, say , is sequenceable if there a permutation of such that no consecutive points in is the union of blocks in , for all such that . The question of determining which PSTS are sequenceable was introduced by Alspach [2]. This problem has been further studied in [3], [4], [10]. It has been shown that a PSTS is sequenceable if the size of the maximum PPC satisfies certain conditions. The known results are summarized in the following theorem.
Theorem 5.1 Suppose
Summary
It would be of interest to obtain tighter bounds on the values . However, our upper and lower bounds on are already relatively close. It is not difficult to see why this is the case. Our main construction, Theorem 2.2, produces an STS in which condition 2. of Lemma 3.1 is met with equality, for all blocks in a maximum partial parallel class . Therefore, our upper and lower bounds differ only because of blocks that contain two or three points from . Our upper bound
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
I would like to thank Simon Blackburn and Daniel Horsley for their helpful comments. I would also like to acknowledge the useful suggestions provided by the referees.
References (14)
- et al.
A second infinite family of Steiner triple systems without almost parallel classes
J. Combin. Theory Ser. A
(2013) - et al.
A note on partial parallel classes in Steiner systems
Discrete Math.
(1978) On the size of partial parallel classes in Steiner systems
- et al.
Nearly perfect matchings in regular simple hypergraphs
Israel J. Math.
(1997) Variations on the sequenceable theme
- et al.
Sequencing partial Steiner triple systems
J. Combin. Des.
(2020) - S.R. Blackburn, T. Etzion, Block-avoiding point sequencings....
Cited by (3)
Maximum parallel classes in 3-packings
2022, Discrete MathematicsOn maximum parallel classes in packing designs
2023, Scientia Sinica MathematicaOn maximum parallel classes in packings
2022, arXiv
- 1
D.R. Stinson’s research is supported by NSERC discovery, Canada grant RGPIN-03882.