arXiv++ Combinatorics

Two matrices with elements taken from the set {-1,1} are Hadamard equivalent if one can be converted into the other by a sequence of permutations of rows and columns, and negations of rows and columns. In this paper we summarize what is known about the number of equivalence classes of matrices having maximal determinant. We establish that there are 7 equivalence classes for matrices of order 21 and that there are at least 9,884 equivalence classes for matrices of order 26. The latter result is obtained primarily using a switching technique for producing new designs from old.

We define several operations that switch substructures of Hadamard matrices thereby producing new, generally inequivalent, Hadamard matrices. These operations have application to the enumeration and classification of Hadamard matrices. To illustrate their power, we use them to greatly improve the lower bounds on the number of equivalence classes of Hadamard matrices in orders 32 and 36 to 3,578,006 and 18,292,717.

We study the question of finding the maximal determinant of matrices of odd order with entries {-1,1}. The most general upper bound on the maximal determinant, due to Barba, can only be achieved when the order is the sum of two consecutive squares. It is conjectured that the bound is always attained in such cases. Apart from these, only in orders 3, 7, 9, 11, 17 and 21 has the maximal value been established. In this paper we confirm the results for these orders, and add order 15 to the list. We follow previous authors in exhaustively searching for candidate Gram matrices having determinant greater than or equal to the square of a known lower bound on the maximum. We then attempt to decompose each candidate as the product of a {-1,1}-matrix and its transpose. For order 15 we find four candidates, all of Ehlich block form, two having determinant (105*3^5*2^14)^2 and the others determinant (108*3^5*2^14)^2. One of the former decomposes (in an essentially unique way) while the remaining three do not. This result proves a conjecture made independently by W. D. Smith and J. H. E. Cohn. We also use our method to compute improved upper bounds on the maximal determinant in orders 29, 33, and 37, and to establish the range of the determinant function of {-1,1}-matrices in orders 9 and 11.

The Hadamard maximal determinant problem asks for the largest n-by-n determinant with entries in {+1,-1}. When n is congruent to 1 (mod 4), the maximal excess construction of Farmakis & Kounias has been the most successful general method for constructing large (though seldom maximal) determinants. For certain small n, however, still larger determinants have been known; several new records were recently reported in ArXiv preprint math.CO/0304410 . Here, we define ``3-normalized'' n-by-n Hadamard matrices, and construct large-determinant matrices of order n+1 from them. Our constructions account for most of the previous ``small n'' records, and set new records when n=37, 49, 65, 73, 77, 85, 93, and 97, most of which are beyond the reach of the maximal excess technique. We conjecture that our n=37 determinant, 72 x 9^{17} x 2^{36}, achieves the global maximum.

We report new world records for the maximal determinant of an n-by-n matrix with entries +/-1. Using various techniques, we beat existing records for n=22, 23, 27, 29, 31, 33, 34, 35, 39, 45, 47, 53, 63, 69, 73, 77, 79, 93, and 95, and we present the record-breaking matrices here. We conjecture that our n=22 value attains the globally maximizing determinant in its dimension. We also tabulate new records for n=67, 75, 83, 87, 91 and 99, dimensions for which no previous claims have been made. The relevant matrices in all these dimensions, along with other pertinent information, are posted at http://www.indiana.edu/~maxdet \.