A Block Negacyclic Bush-Type Hadamard Matrix and Two Strongly Regular Graphs

Similar documents
Hadamard matrices and strongly regular graphs with the 3-e.c. adjacency property

Divisible Design Graphs

Hadamard matrices of order 32

Tilburg University. Strongly Regular Graphs with Maximal Energy Haemers, W. H. Publication date: Link to publication

Difference Systems of Sets and Cyclotomy

3-Class Association Schemes and Hadamard Matrices of a Certain Block Form

Hadamard matrices of order 32

Some new constructions of orthogonal designs

Affine designs and linear orthogonal arrays

Some constructions of integral graphs

Transversal Designs in Classical Planes and Spaces. Aiden A. Bruen and Charles J. Colbourn. Computer Science. University of Vermont

SOME SYMMETRIC (47,23,11) DESIGNS. Dean Crnković and Sanja Rukavina Faculty of Philosophy, Rijeka, Croatia

Small Orders of Hadamard Matrices and Base Sequences

P. J. Cameron a,1 H. R. Maimani b,d G. R. Omidi b,c B. Tayfeh-Rezaie b

3-Designs from PSL(2, q)

A Design and a Code Invariant

#A30 INTEGERS 17 (2017) CONFERENCE MATRICES WITH MAXIMUM EXCESS AND TWO-INTERSECTION SETS

Linear Algebra and its Applications

THE MAXIMUM SIZE OF A PARTIAL 3-SPREAD IN A FINITE VECTOR SPACE OVER GF (2)

Some matrices of Williamson type

Generalized hyper-bent functions over GF(p)

On the Amicability of Orthogonal Designs

New quasi-symmetric designs constructed using mutually orthogonal Latin squares and Hadamard matrices

QUASIDOUBLES OF HADAMARD 2-(15,7,3) DESIGNS WITH AUTOMORPHISMS OF ORDER 3. Zlatka T. Mateva, Svetlana T. Topalova

Resolvable partially pairwise balanced designs and their applications in computer experiments

New Infinite Families of Orthogonal Designs Constructed from Complementary Sequences

Representations of disjoint unions of complete graphs

Supplementary difference sets and optimal designs

An algebraic proof of the Erdős-Ko-Rado theorem for intersecting families of perfect matchings

Two infinite families of symmetric Hadamard matrices. Jennifer Seberry

On unit weighing matrices with small weight

STEINER 2-DESIGNS S(2, 4, 28) WITH NONTRIVIAL AUTOMORPHISMS. Vedran Krčadinac Department of Mathematics, University of Zagreb, Croatia

Simple 3-(q + 1, 5, 3) designs admitting an automorphism group PSL(2, q) with q 1 (mod 4)

Graphs with prescribed star complement for the. eigenvalue.

Intriguing sets of vertices of regular graphs

Tactical Decompositions of Steiner Systems and Orbits of Projective Groups

Orthogonal arrays of strength three from regular 3-wise balanced designs

FRACTIONAL FACTORIAL DESIGNS OF STRENGTH 3 AND SMALL RUN SIZES

Some V(12,t) vectors and designs from difference and quasi-difference matrices

Mathematical Department, University of Bayreuth, D Bayreuth

On the Classification of Splitting (v, u c, ) BIBDs

Spectra of Semidirect Products of Cyclic Groups

Directed cyclic Hamiltonian cycle systems of the complete symmetric digraph

Existence of doubly near resolvable (v, 4, 3)-BIBDs

An infinite family of skew weighing matrices

Complex Hadamard matrices and 3-class association schemes

Finite geometry codes, generalized Hadamard matrices, and Hamada and Assmus conjectures p. 1/2

Hadamard and conference matrices

arxiv: v4 [math.co] 12 Feb 2017

DIRECTED CYCLIC HAMILTONIAN CYCLE SYSTEMS OF THE COMPLETE SYMMETRIC DIGRAPH

Integral trees of odd diameters

Zero-Sum Flows in Regular Graphs

Codes, Designs and Graphs from the Janko Groups J 1 and J 2

K. T. Arasu Jennifer Seberry y. Wright State University and University ofwollongong. Australia. In memory of Derek Breach.

Square 2-designs/1. 1 Definition

ORTHOGONAL ARRAYS OF STRENGTH 3 AND SMALL RUN SIZES

On A Special Case Of A Conjecture Of Ryser About Hadamard Circulant Matrices

A construction for orthogonal designs with three variables

Algorithmic Approach to Non-symmetric 3-class Association Schemes

TRIPLE FACTORIZATION OF NON-ABELIAN GROUPS BY TWO MAXIMAL SUBGROUPS

PARTITIONS OF FINITE VECTOR SPACES INTO SUBSPACES

A REPRESENTATION THEORETIC APPROACH TO SYNCHRONIZING AUTOMATA

arxiv: v2 [math.co] 6 Apr 2016

Strongly Regular Decompositions of the Complete Graph

Algebraic Methods in Combinatorics

Algebraic Methods in Combinatorics

On integer matrices obeying certain matrix equations

ON THE CONSTRUCTION OF ORTHOGONAL BALANCED INCOMPLETE BLOCK DESIGNS

Matrices and Vectors

On Good Matrices and Skew Hadamard Matrices

On the (v,5,)-family of Bhaskar Rao Designs. Ghulam R Chaudhry. School of IT & CS, University of Wollongong. Wollongong, NSW 2522, AUSTRALIA

Maximal Energy Graphs

A Series of Regular Hadamard Matrices

Trades in complex Hadamard matrices

Butson-Hadamard matrices in association schemes of class 6 on Galois rings of characteristic 4

Some New D-Optimal Designs

A new look at an old construction: constructing (simple) 3-designs from resolvable 2-designs

Hadamard ideals and Hadamard matrices with two circulant cores

Weighing matrices and their applications

Characterization of quasi-symmetric designs with eigenvalues of their block graphs

Cospectral graphs and the generalized adjacency matrix

The Witt designs, Golay codes and Mathieu groups

Butson-Hadamard matrices in association schemes of class 6 on Galois rings of characteristic 4

Quasimultiples of Geometric Designs

SYMMETRIC (66,26,10) DESIGNS HAVING F rob 55 AUTOMORPHISM GROUP AS AN. Dean Crnković and Sanja Rukavina Faculty of Philosophy in Rijeka, Croatia

Some results on the existence of t-all-or-nothing transforms over arbitrary alphabets

Sets of MOLSs generated from a single Latin square

On Good Matrices and Skew Hadamard Matrices

Math 3121, A Summary of Sections 0,1,2,4,5,6,7,8,9

On some incidence structures constructed from groups and related codes

Steiner Triple Systems Intersecting in Pairwise Disjoint Blocks

There is no 2-(22, 8, 4) Block Design

Primitive 2-factorizations of the complete graph

Graphs where every k-subset of vertices is an identifying set

International Mathematical Forum, Vol. 6, 2011, no. 4, Manjusri Basu

Linear Algebra and its Applications

Additional Constructions to Solve the Generalized Russian Cards Problem using Combinatorial Designs

Inner Product Vectors for skew-hadamard Matrices

On Construction of a Class of. Orthogonal Arrays

Hadamard and conference matrices

Transcription:

Journal of Combinatorial Theory, Series A 98, 118 126 (2002) doi:10.1006/jcta.2001.3231, available online at http://www.idealibrary.com on A Block Negacyclic Bush-Type Hadamard Matrix and Two Strongly Regular Graphs Zvonimir Janko Mathematical Institute, University of Heidelberg, Heidelberg, Germany and Hadi Kharaghani Department of Mathematics and Computer Science, University of Lethbridge, Lethbridge, Alberta T1K 3M4, Canada Received November 19, 2000; published online February 11, 2002 A block negacyclic Bush-type Hadamard matrix of order 36 is used in a symmetric BGW(26, 25, 24) with zero diagonal over a cyclic group of order 12 to construct a twin strongly regular graph with parameters v=936, k=375, l=m=150 whose points can be partitioned in 26 cocliques of size 36. The same Hadamard matrix then is used in a symmetric BGW(50, 49, 48) with zero diagonal over a cyclic group of order 12 to construct a Siamese twin strongly regular graph with parameters v=1800, k=1029, l=m=588. 2002 Elsevier Science (USA) Key Words: symmetric design; Bush-type Hadamard matrix; strongly regular graph; block negacyclic matrix; balanced generalized weighing matrix. 1. INTRODUCTION A symmetric 2-(v, k, l) design is a set X of v points together with a collection B of vk-subsets of X called blocks such that every two points are contained in exactly l blocks. A symmetric 2-(v, k, l) design can be described in terms of its incidence matrix, being a square v by v (0, 1)- matrix with constant row and column sum equal to k, and constant scalar product of pairs of rows equal to l. For more on designs see Beth et al. [1]. 0097-3165/02 $35.00 2002 Elsevier Science (USA) All rights reserved. 118

A BUSH-TYPE HADAMARD MATRIX 119 A Hadamard matrix of order n is a square n by n matrix with entries ±1 whose rows are pairwise orthogonal. A Hadamard matrix H of order 4n 2 is called regular if every row and column sum of H is 2n. A Bush-type Hadamard matrix [4] is a regular Hadamard matrix of order 4n 2 with the additional property of being a block matrix H=[H ij ], with blocks of size 2n such that H ii =J 2n and H ij J 2n =J 2n H ij =0, i ] j, 1 [ i [ 2n, 1 [ j [ 2n, where J m denotes the all-one m by m matrix. Let U=circ(0, 1, 0,..., 0) be the circulant matrix of order n with first row (010...0) and N=diag( 1, 1, 1,..., 1) be the diagonal matrix of order n with 1 at the (1, 1)-position and 1 elsewhere on the diagonal. Let W=UN. Matrices which are polynomials in W are called negacyclic; see [5, 2]. A block negacyclic Bush-type Hadamard matrix of order 4n 2 is a Bush-type Hadamard matrix which is block negacyclic (with block size 2n, of course) with all blocks being symmetric. Bush [4] showed that the existence of a projective plane of order 2n implies the existence of a symmetric Bush-type Hadamard matrix of order 4n 2. A result of Kharaghani [11] can be used to show that the existence of a Hadamard matrix of order 4n implies the existence of a negacyclic Bush-type Hadamard matrix of order 16n 2 =(4n) 2. No examples of block negacyclic Bush-type Hadamard matrices of order 4n 2, n>1 n odd, appeared to be known previously. A strongly regular graph with parameters (v, k, l, m), denoted SRG(v, k, l, m), is a regular graph with v vertices of degree k such that every two adjacent vertices have exactly l common neighbors, and every two nonadjacent vertices have exactly m common neighbors; see [3]. A balanced generalized weighing matrix BGW(n, o, l) over a multiplicative group G is a n by n matrix W=[g ij ] with entries from Ḡ=G 2 {0} such that each row of W contains exactly o nonzero entries, and for every a, b {1,..., n}, a ] b, the multi-set {g ai g 1 bi :1[ i [ n,g ai ] 0, g bi ] 0} contains exactly l/ G copies of each element of G. In a recent work [9], Kharaghani introduced a method in which from a single Bush-type Hadamard matrix of order 4n 2 one can generate an infinite class of twin symmetric designs provided that one of the numbers 2n 1 or 2n+1 is a prime power, and two infinite classes of twin symmetric designs if both 2n 1 and 2n+1 are prime powers. A twin design is a (0, ± 1)-matrix where both signs, 1 and 1, form symmetric designs with the same parameters. In this paper, for the first time, a block negacyclic Bush-type Hadamard matrix of order 36 is constructed and used in a symmetric BGW(26, 25, 24) with zero diagonal over the cyclic group of order 12 to generate a twin

120 JANKO AND KHARAGHANI SRG(936, 375, 150, 150) whose vertices can be partitioned into 26 cocliques of size 36. Using the same matrix, but in a symmetric BGW(50, 49, 48) with zero diagonal over the cyclic group of order 12, we are also able to generate a Siamese twin SRG(1800, 1029, 588, 588). A Siamese twin SRG is a symmetric (0, ± 1)-matrix with zero diagonal for which both signs, 1 and 1 to which a certain number of negated 1 entries are added, are the incidence matrices of strongly regular graphs with the same parameters sharing cliques of the same size; see [10] for the definition of Siamese twin designs. 2. A BLOCK NEGACYCLIC BUSH-TYPE HADAMARD MATRIX OF ORDER 36 In this section we first introduce the first block negacyclic Bush-type Hadamard matrix of order 36. This is a Hadamard matrix H of order 36 of the following form: A B C D E F F A B C D E E F A B C D H=R S, D E F A B C C D E F A B B C D E F A where the symmetric blocks are S 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 A=R1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 B=R S 1 1 1 1 1 1 1 1 1

A BUSH-TYPE HADAMARD MATRIX 121 1 1 1 1 1 1 1 1 1 C=R 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 D=R 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 E=R 1 1 1 1 1 1 1 1 1 S S S 1 1 1 1 1 1 1 1 1 F=R S. 1 1 1 1 1 1 1 1 1 To explain our method of search for this matrix briefly, we denote the rows of H by l1, l2,..., l36. The jth entry in the row li is denoted by l(i, j). At the beginning the following reduction is essential. Permuting the rows and columns of H passing through the first block B (in the first block-row) we may assume (without loss of generality) that l(1, 7)=1, l(1, 8)=1, l(1, 9)=1, l(1, 10)= 1, l(1, 11)= 1, l(1, 12)= 1. Note that in every block-matrix we have in each row (and column) 3 times the entry 1 and 3 times the entry 1. Therefore (for example) for the entries l(1, 13), (l1, 14), l(1, 15), l(1, 16), l(1, 17), l(1, 18) we have exactly 20 possibilities. Therefore for the first row l1 of H we have at the start 20 4 possibilities. It is clear that it is enough to construct only the first 6 rows l1, l2, l3, l4, l5, l6.

122 JANKO AND KHARAGHANI Now, when the row l1 is chosen, then the rows l7, l13, l19, l25 and l31 are uniquely determined. But all these rows have to be pair-wise orthogonal. This diminishes the number of possibilities for l1 and in fact we get exactly 13268 possibilities for l1. Similarly, for the row l2 we have at the start exactly 20 5 possibilities (no more reduction!). When l2 is chosen, then the rows l8, l14, l20, l26, l32 are uniquely determined and all these rows (and the previously constructed) should be pair-wise orthogonal. Also we should not forget that all block matrices must be symmetric and this gives additional conditions which diminish the number of possibilities for l1 and l2. As a result we get exactly 53268 possibilities for l1 and l2. We have arrived at the critical point of the construction because of the explosion of possibilities! Similarly, we start with the construction of l3. Here the computer computation takes 2 to 3 days and we get exactly 576 possibilities for l1, l2, l3. This is already half of the matrix H. From now on the construction goes quickly: for l1, l2, l3, l4 we get exactly 320 possibilities; for l1, l2, l3, l4, l5 we get exactly 288 possibilities and finally for l1, l2, l3, l4, l5, l6 we get exactly (also) 288 possibilities! The search is successful and the matrix H above is constructed. The symmetric (36, 15, 6)- design obtained from the matrix H is rigid and, unfortunately, a search of this sort seems not feasible for the next interesting negacyclic Bush-type Hadamard matrix which has to be of order 100. The matrix H being block negacyclic commutes with block negacyclic matrices of the same order (and the same block size, of course), if the blocks all commute and this is essential for our method of construction for strongly regular graphs in this paper. 3. SOME SYMMETRIC BALANCED GENERALIZED WEIGHING MATRICES WITH ZERO DIAGONAL Despite the fact that several method of construction is available for balanced generalized matrices, see, for example, [6, 8], not much is known about the existence of symmetric or skew balanced generalized weighing matrices with zero diagonal. Symmetric BGWs with zero diagonal, as we will see, are essential to our construction of SRGs. In this section we reproduce some known construction methods to obtain a symmetric BGW(26, 25, 24) and BGW(50, 49, 48) with zero diagonal over the cyclic group C 12 of order 12. Our main reference for the construction of BGWs is Gibbons and Mathon [6]. Lemma 1. There are symmetric BGW(26, 25, 24) and BGW(50, 49, 48) with zero diagonal over the cyclic group C 12 of order 12.

A BUSH-TYPE HADAMARD MATRIX 123 Proof. Let 25=(2)(12)+1, GF(25)={a 1,a 2,...,a 25 } and a a primitive root in GF(25). Note that if a k a j belongs to the cyclotomic class {a i, a 12+i } for some i, then a j a k also belongs to the same cyclotomic class. Now apply the cyclotomic method of construction given by Gibbons and Mathon in [6], page 8. The same argument applies to the prime power 7 2. Remark 2. Note that the cyclotomic method of construction for the BGW matrices given in [6] always provides a skew matrix for BGW(q+1, q, q 1) over the cyclic group C q 1 of order q 1,ifq is an odd prime power. 4. TWO TWIN STRONGLY REGULAR GRAPHS We now have the building blocks for the two twin strongly regular graphs SRG(936, 375, 150, 150) and SRG(1800, 1029, 588, 588). Let G 4n be the cyclic subgroup of order 4n of signed permutation matrices generated by the negacyclic matrix W, defined in Section 1, of order 2n. Then C 4n ={G 4n é I 2n } is a cyclic subgroup (of order 4n) of the signed permutation matrices of order 4n 2. The next lemma is crucial for our construction. Lemma 3. Let C be any element in C 4n, H a block negacyclic matrix of order 4n 2 and L=R 2n é I 2n, where I 2n is the identity matrix of order 2n and R 2n denotes the back diagonal identity matrix of order 2n (for simplicity we will drop the indices where there is no fear of confusion). Then the matrix HCL is symmetric. Proof. Let (A 1,A 2,...,A 2n ) be the first row (blocks) of H. So we can write H=I é A 1 +W é A 2 + +W 2n 1 é A 2n, where W is the negacyclic shift matrix of order 2n defined in Secion 1. Let C=W i é I, for some i, 0 [ i [ 4n. Now, (HCL) t =LC t H t =(R é I)((W i ) t é I)(I é A t 1+W t é A t 2+ +(W 2n 1 ) t é A t 2n 1) =(R(W i ) t é I)(I é A 1 +W t é A 2 + +(W 2n 1 ) t é A 2n 1 ) =((W i ) é I)(R é I)(I é A 1 +W t é A 2 + +(W 2n 1 ) t é A 2n 1 ) =((W i ) é I)(R é A 1 +RW t é A 2 + +R(W 2n 1 ) t é A 2n 1 ) =((W i ) é I)(I é A 1 +W é A 2 + +W 2n 1 é A 2n 1 )(R é I) =(I é A 1 +W é A 2 + +W 2n 1 é A 2n 1 )((W i ) é I)(R é I) =HCL.

124 JANKO AND KHARAGHANI Remark 4. An easier version of Lemma 3 is valid for circulant matrices. More precisely, the matrix HCL is symmetric, whenever, H and C are circulant matrices. Theorem 5. There is a twin strongly regular graph, SRG(936, 375, 150, 150) whose vertices can be partitioned into 26 disjoint cocliques of size 36. Proof. Let H be any block negacyclic Bush-type Hadamard matrix of order 36 and W=[w ij ] a symmetric BGW(26, 25, 24) with zero diagonal over the cyclic group C 12. It is shown in [9] that D=[Mw ij ], where M=H I 6 é J 6 is a twin 2-(936, 375, 150) design by splitting it into and D + = 1 2 [P w ij +Mw ij ] D = 1 2 [P w ij Mw ij ], where P=J 36 I 6 é J 6, and showing that D + and D are both symmetric 2-(936, 375, 150) designs. All we need to do now is to multiply each block of D from the right by the matrix L=R 6 é I 6 to get the matrix DŒ=[Mw ij L]. The matrix DŒ is: symmetric with zero diagonal blocks. This follows from Lemma 3 (note that the matrix M is negacyclic) and the fact that the diagonal entries of the matrix W are all zero. a twin SRG(936, 375, 150, 150). To see this, note that the symmetric matrix DŒ + = 1 [(P w 2 ij +Mw ij )L]is a 2-(936, 375, 150) design (we need the remark after Lemma 3 to show that the matrix [P w ij L] is symmetric here). The same is true for the symmetric matrix DŒ = 1 [(P w 2 ij Mw ij )L]. Theorem 6. There is a Siamese twin strongly regular graph, SRG(1800, 1029, 588, 588), whose vertices can be partitioned in 50 cocliques of size 36. The two graphs share only cliques of size 6. Proof. Let H be any block negacyclic Bush-type Hadamard matrix of order 36 and W=[w ij ] a symmetric BGW(50, 49, 48) with zero diagonal over the cyclic group C 12. Let M=H I 6 é J 6. It is proven in [10] that 1 [Mw 2 ij+((j 6 +I 6 ) é J 6 ) w ij ]

A BUSH-TYPE HADAMARD MATRIX 125 and 1 [ Mw 2 ij+((j 6 +I 6 ) é J 6 ) w ij ] are both symmetric 2-(1800, 1029, 588) designs. The fact that these two matrices have their diagonal blocks zero is a consequence of the fact that the BGW(50, 49, 48) has all its diagonal entries zero. We now need to make these two matrices symmetric and we can do this by multiplying all the blocks of each of the two matrices from the right by the matrix L=R 6 é I 6 of Lemma 3. This gives us a Siamese twin SRG(1800, 1029, 588, 588) whose vertices can be partitioned into 50 disjoint cocliques of size 36, sharing only the cliques of size 6. Remark 7. Clearly the construction method above works for any block negacyclic Bush-type Hadamard matrix and a symmetric BGW with zero diagonal of suitable parameters. It should be now clear for the reader the reason that a block negacyclic Bush-type is necessary in our construction above. Of course, it would be nice if one could use objects with less complexity in their structures than the block negacyclic Bush-type Hadamard matrices. But this may not be that easy. Indeed, there seems to be a balance in the complexity of the structure of the main ingredients for our construction. On one hand it is fairly easy to construct block negacyclic Bush-type Hadamard matrices for all orders 16n 2 for which there is a Hadamard matrix of order 4n. But then it is very hard to construct a proper symmetric BGW. On the other hand it is very hard to construct block negacyclic Bush-type Hadamard matrices of order 4n 2, n odd, n>1, whereas the required BGW is fairly simple to construct. ACKNOWLEDGMENTS The work was completed while the author was on sabbatical leave visiting the Institute for studies in theoretical Physics and Mathematics, IPM, in Tehran, Iran. Support and hospitality is appreciated. The research is supported by an NSERC operating grant. REFERENCES 1. T. Beth, D. Jungnickel, and H. Lenz, Design Theory, 2nd ed., Cambridge Univ. Press, Cambridge, UK, 1999. 2. G. Berman, Families of skew-circulant weighing matrices, Ars Combin. 4 (1977), 293 307. 3. A. E. Brouwer, Strongly regular graphs, in The CRC Handbook of Combinatorial Designs (C. J. Colbourn and J. H. Dinitz, Eds.), pp. 667 683, CRC Press, Boca Raton, FL, 1996.

126 JANKO AND KHARAGHANI 4. K. A. Bush, Unbalanced Hadamard matrices and finite projective planes of even order, J. Combin. Theory 11 (1971), 38 44. 5. P. Delsarte, J. M. Goethals, and J. J. Seidel, Orthogonal matrices with zero diagonal, II, Canad. J. Math. 23 (1971), 816 832. 6. P. Gibbons and R. Mathon, Construction methods for Bhaskar Rao and related designs, J. Austral. Math. Soc. Ser. A 42 (1987), 5 30. 7. D. R. Hughes and F. C. Piper, Design Theory, Cambridge Univ. Press, Cambridge, UK, 1985. 8. D. Jungnickel and V. Tonchev, Perfect codes and balanced weighing matrices, Finite Fields Appl. 5 (1999), 294 300. 9. H. Kharaghani, On the twin designs with the Ionin type parameters, Electron. J. Combin. 7 (2000), R1. 10. H. Kharaghani, On the Siamese twin designs, in Finite Fields and Applications (D. Jungnickel and H. Niederreiter, Eds.), pp. 301 312, Springer-Verlag, New York/ Berlin, 2001. 11. H. Kharaghani, New classes of weighing matrices, Ars Combin. 19 (1985), 69 72.