Matricial characterization of tournaments with maximum number of diamonds
Wiam Belkouche, Abderrahim Boussa\"iri, Soufiane Lakhlifi, Mohamed, Zaidi

TL;DR
This paper provides a complete matrix-based characterization of n-tournaments with the maximum number of diamonds for certain sizes, assuming skew-conference matrices, and offers bounds for other sizes.
Contribution
It introduces a matricial framework to characterize maximum diamond tournaments for specific sizes, extending understanding of their structure.
Findings
Complete characterization for n ≡ 0 mod 4 and n ≡ 3 mod 4
Upper bounds and matricial descriptions for n ≡ 2 mod 4
Assumption of skew-conference matrices
Abstract
A diamond is a -tournament which consists of a vertex dominating or dominated by a -cycle. Assuming the existence of skew-conference matrices, we give a complete characterization of -tournaments with the maximum number of diamonds when and . For , we obtain an upper bound on the number of diamonds in an -tournament and we give a matricial characterization of tournaments achieving this bound.
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.
Matricial characterization of tournaments with maximum number of diamonds
Wiam Belkouche
Abderrahim Boussaïri
Soufiane Lakhlifi
Mohamed Zaidi
Faculté des Sciences Aïn Chock, Département de Mathématiques et Informatique, Laboratoire de Topologie, Algèbre, Géométrie et Mathématiques Discrètes, Université Hassan II Km 8 route d’El Jadida, BP 5366 Maarif, Casablanca, Maroc
Abstract
A diamond is a -tournament which consists of a vertex dominating or dominated by a -cycle. Assuming the existence of skew-conference matrices, we give a complete characterization of -tournaments with the maximum number of diamonds when and . For , we obtain an upper bound on the number of diamonds in an -tournament and we give a matricial characterization of tournaments achieving this bound.
keywords:
Tournaments, Diamonds, Skew-conference matrices, EW-matrices, Spectrum.
1 Introduction
One of the most important problems in Extremal Combinatorics is to determine the largest or the smallest possible number of copies of a given object in a finite combinatorial structure. We address this problem in the case of tournaments. Throughout this paper, we mean by an -tournament, a tournament with vertices. It is easy to see that, up to isomorphy, there are four distinct -tournaments. The two that contain a single -cycle are called diamonds [3, 4, 8]. They consist of a vertex dominating or dominated by a -cycle. The class of tournaments without diamonds was characterized by Moon [15]. These tournaments appear in the literature under the names local orders [5],* locally transitive tournaments* [12] or vortex-free tournaments [11]. For , Bouchaala [3] proved that the number of diamonds in an -tournament is either [math], , or at least . In another side, motivated by geometric considerations, Leader and Tan [13] proved that is at most . Moreover, by a probabilistic method, they showed that there is an -tournament with at least diamonds. To find the Turán density of a particular -uniform hypergraphs, Baber [10] associates with each tournament , the -uniform hypergraph on whose hyperedges correspond to subsets of which induce a diamond in . In this hypergraph, every -subset contains either [math] or hyperedges. Recently, using a combinatorial argument due to de Caen [6], Gunderson and Semeraro [10] proved that an -uniform hypergraph in which every -subset contains at most hyperedges has at most hyperedges, in particular, an -tournament contains at most diamonds. Moreover, using Paley tournaments, they showed that this bound is reached if for some prime power .
In this paper, we study the tournaments with the maximum number of diamonds. Our work is closely related to the existence of D-optimal designs. More precisely, assuming the existence of skew-conference matrices, we give a complete characterization of -tournaments with the maximum number of diamonds when and . For , we obtain an upper bound on the number in an -tournament. Moreover, we give a matricial characterization of tournaments achieving this bound.
2 Number of diamonds and -cycles in tournaments
Throughout this paper, all matrices are from the set , unless otherwise noted. The identity matrix of order is denoted by and the all ones matrix is denoted by . The absolute value , of a matrix , is obtained from by replacing each entry of by its absolute value. Two matrices and are -diagonally similar if for some -diagonal matrix.
Let be an -tournament. With respect to a labelling, the adjacency matrix of is the matrix in which is if dominates and [math] otherwise. The Seidel adjacency matrix of is where is the transpose of . Remark that, with respect to different labellings, the Seidel adjacency matrices of a tournament are permutationally similar.
Remark 1**.**
The determinant of the Seidel adjacency matrix of a -tournament is if it is a diamond and otherwise.
The following lemma is crucial in our study.
Lemma 2**.**
Let be an -tournament and let be its Seidel adjacency matrix. Then the sum of all principal minors of is .
Proof.
The number of principal minors of is . It follows from Remark 1 that the sum of all principal minors of is . ∎
Let be an complex matrix and let be its characteristic polynomial, then
[TABLE]
When is a real skew-symmetric matrix, its nonzero eigenvalues are all purely imaginary and come in complex conjugate pairs , where are real positive numbers. Equivalently, the characteristic polynomial of has the form
[TABLE]
where .
Assume now that is skew-symmetric and all its off-diagonal entries are from the set . Such matrix is sometimes known as a skew-symmetric Seidel matrix. By using [14, Proposition 1], if and only if is odd. Then, if is even, and
[TABLE]
If is odd, then by using [14, Proposition 1] again, any -principal minor is nonzero and thus, the multiplicity of the eigenvalue [math] is . It follows that
[TABLE]
A useful formula of the number of diamonds is given in the following proposition.
Proposition 3**.**
Let be an -tournament and let be its Seidel adjacency matrix. Then,
[TABLE]
where are the entries of .
Proof.
Let be the integer part of and the nonzero eigenvalues of . As we have seen above
[TABLE]
The nonzero eigenvalues of are , each of them appears two times. Hence, we can write in the following form.
[TABLE]
Let and let . By expanding expressions (2) and (3), we get
[TABLE]
[TABLE]
and
[TABLE]
It follows that . By Equality (1), we have , and hence .
Since is symmetric and all its diagonal entries are , by Equality (1), we have
[TABLE]
Applying Lemma (2) and Equality (1) again, we get
[TABLE]
It follows that
[TABLE]
∎
Let be a tournament, the switching of , according to a subset of , consists of reversing all the arcs between and , we denote the resulting tournament by . We say that two tournaments and on a vertex set are switching equivalent, if there exists such that .
Remark 4**.**
It is well-known that two tournaments are switching equivalent iff their Seidel adjacency matrices are -diagonally similar [16]. Since similarity by a diagonal matrix preserves the principal minors, by Remark 1, switching equivalent tournaments have the same diamonds.
Let be a vertex of a tournament , the out-neighbourhood of is the set of all vertices of dominated by . The in-neighbourhood of is the set of all vertices of which dominate . In the switching of according to , the vertex dominates . Hence, by Remark 4, to study the number of diamonds, we can assume that there is a vertex dominating . We obtain then the following lemma connecting the number of diamonds and the number of -cycles.
Lemma 5**.**
Let be a tournament containing a vertex that dominates . Then,
[TABLE]
where is the number of -cycles in .
Similarly to , the number of -cycles in can also be expressed in terms of the entries of .
Proposition 6**.**
Let be an -tournament and let be its Seidel adjacency matrix. Then,
[TABLE]
where are the entries of .
Before proving this proposition, we need some notions and basic results about tournaments. For more details, the reader is referred to [15].
Let be an -tournament. Without loss of generality, we can assume that the vertex set of is . The out-degree (resp. in-degree ) of a vertex is (resp. ). The out-degree of (resp. in-degree of ) is (resp. ).
The tournament is regular, if there is a constant such that for every ; it is doubly regular if there is a constant such that for every . A doubly regular tournament is also regular. If is regular, then is odd and for . If is doubly regular, then and for .
Recall the well-known equalities
[TABLE]
[TABLE]
[TABLE]
Remark 7**.**
It follows from Equality (6) that . Moreover equality holds iff is odd and is regular.
For , let . Then, we have
[TABLE]
[TABLE]
[TABLE]
Combining Equalities (8), (9) and (4), we get
[TABLE]
By double-counting principle, we obtain
[TABLE]
Using Equalities (4) and (11), we get
[TABLE]
It follows from Equalities (5) and (6) that
[TABLE]
Equality (7) allows to complete the proof of Proposition 5.
3 Characterization of -tournaments with maximum number of diamonds for
Throughout this section, denotes an -tournament and denotes its Seidel adjacency matrix.
It follows from Proposition 3 that
[TABLE]
Equality holds if and only if for every . Since , then if and only if , or equivalently . Skew-symmetric Seidel matrices that satisfy this equality are called skew-conference matrices, and exist only if is divisible by .
Example 8**.**
For any prime power , the Paley tournament is the tournament whose vertices are elements of where the vertex dominates the vertex iff is a square in . Let be the tournament on vertices obtained by adding to a new vertex which dominates all vertices of . It is well-known that the Seidel adjacency matrix of is a skew conference matrix. Then, the number of diamonds in is .
For odd, we obtain the following refinement of Equality (12).
Proposition 9**.**
If is odd, then
[TABLE]
Moreover, equality holds if and only if .
Proof.
Let . Since is odd, by Equality (7), is also odd. Hence .
By Proposition 3,
[TABLE]
Equality holds if and only if for , or equivalently , because . ∎
Theorem 10**.**
If , then the following assertions are equivalent
* has diamonds.* 2. 2.
There exists a diagonal -matrix D such that . 3. 3.
* is switching equivalent to a doubly regular tournament.*
Proof.
Assume that has diamonds. By Proposition 9, we have for every . Using Equalities (7) and (10), we get:
- i.
If , then , and . Hence . 2. ii.
If , then , and . Hence .
Let be the diagonal matrix such that if is even and otherwise. It is easy to check that for all . Then, .
The converse is trivial.
Suppose that there exists a diagonal -matrix D such that . The tournament whose Seidel adjacency matrix is is switching equivalent to . We will prove that is doubly regular.
Let . Thus, for , we have
[TABLE]
By Proposition 6, we get
[TABLE]
It follows from Remark 7 that the tournament is regular.
By Identities (7) and (10), the in-degree of each pair in is . Hence, by definition, is doubly regular.
Conversely, assume that is switching equivalent to a doubly regular tournament . The out-degree and the in-degree of each pair in is . Let be the Seidel adjacency matrix of and let . By Equalities (7) and (10), we get for and hence . Since and are switching equivalent, for some -diagonal matrix and then .
∎
Theorem 11**.**
If , then
[TABLE]
Moreover, equality holds if and only if
[TABLE]
for some permutation matrix .
Proof.
We label the vertices of so that the first vertices have an even out-degree and the remaining vertices have an odd out-degree. With respect to this labelling, the Seidel adjacency matrix of is where is a permutation matrix. Let . Let such that . By Equality (10), , and by Equality (7), . Therefore, .
By Proposition 3, we have
[TABLE]
Equality holds iff , if or and otherwise. ∎
We give two classes of tournaments that satisfy the conditions of Theorem 11.
Recall that an EW matrix of order is a -matrix verifying where . Ehlich [7] and Wojtas [17] independently proved that EW matrices have the maximum determinant among -matrices of order . EW matrices exist only if is the sum of two squares. An EW matrix is said to be of skew type if . Such matrix exist only if is a square, hence, there are no EW matrices of skew type with order .
Consider the matrix , where is an EW-matrix of skew type. Clearly, this matrix is skew-symmetric, moreover, it has the maximum determinant among skew-symmetric Seidel matrices of order [2]. By simple computation, we have
[TABLE]
Hence, by Theorem 11, if is an EW matrix, then has diamonds. Moreover, by [9, Lemma 3.3], the characteristic polynomial of is where . 2. 2.
Let be a doubly regular tournament on vertices and let be the tournament obtained by removing any vertex of . It is easy to see that for every vertices of , we have
[TABLE]
Let be the Seidel adjacency matrix of . Using Identities (7) and (9), we find that up to permutation
[TABLE]
Hence, again by Theorem 11, the tournament has diamonds. Moreover, by [9, Lemma 4.2.iii], .
Remark 12**.**
Let be a tournament with vertices and let be its Seidel adjacency matrix. It follows from [9, Lemmata 3.3 and 3.7] that the characteristic polynomial of is iff there is a -diagonal matrix D such that is a skew-symmetric EW matrix.
Remark 13**.**
Let be a tournament and let be its Seidel adjacency matrix. It follows from Lemma 2 and Equality (1) that if or , then has the maximum number of diamonds.
Up to switching, there are two 6-tournaments with the maximum number of diamonds , one of them is obtained by removing a vertex from a doubly regular tournament, and the other consists of two -cycles one dominating the other, its Seidel adjacency matrix is a skew symmetric EW matrix.
As for -tournaments, using SageMath, we found two switching classes of tournaments with the maximum number of diamonds . The characteristic polynomial of the tournaments in the first class is , we identified one as a tournament obtained by removing a vertex from a doubly regular tournament. The characteristic polynomial of the tournaments in the second class is . The tournament with the following Seidel adjacency matrix belongs in the second class.
[TABLE]
Curiously, has the maximum determinant among matrices, but is not an EW matrix. This leads to the following questions.
Question 14**.**
Let be a skew Seidel matrix with the maximum determinant, does its corresponding tournament have the maximum number of diamonds ?
The answer to this question is positive in the following two cases:
There exists a skew conference matrix of order . 2. 2.
There exists a skew EW matrix of order .
Question 15**.**
Let . Is there an infinite family of -tournaments with diamonds, such that their Seidel adjacency matrices has a characteristic polynomial that is neither nor .
4 The case of
We start with following lemma.
Lemma 16**.**
Let be a tournament on vertices with diamonds. Then
[TABLE]
for every .
Proof.
Let be a tournament on vertices containing diamonds. Fix a vertex of and let be the tournament switching equivalent to in which dominates the other vertices.
By adapting the proof of [1, Lemma 2.1], we show easily that the the score vector of is , and , each appearing , and times, respectively.
By Equality (6), the number of -cycles in is
[TABLE]
That is,
[TABLE]
Since dominates every vertex in in the tournament , by Lemma 5
[TABLE]
Hence,
[TABLE]
It follows that
[TABLE]
because is switching equivalent to . ∎
The previous lemma leads to the following conjecture.
Conjecture 17**.**
Let be a tournament on vertices. Then, the number of diamonds in is at most
[TABLE]
The following lemma gives another way to obtain -tournaments with diamonds.
Lemma 18**.**
Let be a tournament on vertices such that its Seidel adjacency matrix is a skew-conference matrix . Let be a tournament obtained from by adding a vertex that dominates all vertices in , then
[TABLE]
Proof.
Let be a tournament obtained from by adding a vertex that dominates all vertices in , the number of diamonds in the tournament is
[TABLE]
Let be the Seidel adjacency matrix of . By Propositions 3 and 6, we have
[TABLE]
Since the non-diagonal entries of are equal to zero, then
[TABLE]
∎
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] José Andrés Armario. On (- 1, 1)-matrices of skew type with the maximal determinant and tournaments. In Algebraic Design Theory and Hadamard Matrices , pages 1–11. Springer, 2015.
- 2[2] José Andrés Armario and María Dolores Frau. On skew e–w matrices. Journal of Combinatorial Designs , 24(10):461–472, 2016.
- 3[3] Houcine Bouchaala. Sur la répartition des diamants dans un tournoi. Comptes Rendus Mathématiques , 338(2):109–112, 2004.
- 4[4] Binh-Minh Bui-Xuan, Michel Habib, Vincent Limouzy, and Fabien De Montgolfier. Unifying two graph decompositions with modular decomposition. In International Symposium on Algorithms and Computation , pages 52–64. Springer, 2007.
- 5[5] Peter J Cameron. Orbits of permutation groups on unordered sets, ii. Journal of the London Mathematical Society , 2(2):249–264, 1981.
- 6[6] Dominique De Caen. Extension of a theorem of moon and moser on complete subgraphs. Ars Combinatoria , 16:5–10, 1983.
- 7[7] Hartmut Ehlich. Determinantenabschätzungen für binäre matrizen. Mathematische Zeitschrift , 83(2):123–132, 1964.
- 8[8] Cyprien Gnanvo and Pierre Ille. La reconstruction des tournois sans diamant. Mathematical Logic Quarterly , 38(1):283–291, 1992.
