Necessary and sufficient conditions for three and four variable orthogonal designs in order 36 (Q698021)

An orthogoal design of order \(n\) and type \((s_1,s_2,\dots, s_u)\) \((s_i> 0)\), denoted \(\text{OD}(n; s_1,s_2,\dots, s_u)\), on the commuting variables \(x_1,\dots, x_u\) is an \(n\times n\) matrix \(A\) with entries from \(\{0,\pm x_1,\dots,\pm x_u\}\) such that \[ AA^{\text{T}}= \Biggl(\sum^u_{i=1} s_i x^2_i\Biggr) I_n. \] The method of constructing orthogonal designs was used in \textit{C. Koukouvinos} (1997, 1998), \textit{C. Koukouvinos} et al. (1996) and \textit{C. Koukouvinos} and \textit{J. Seberry} (2000). Furthermore in \textit{C. Koukouvinos} et al. (1996) a complete study for two variable orthogonal designs in order 36 was given, and some new four variable orthogonal designs of the same order 36 were constructed. More results on three and four variable orthogonal designs in order 36 were presented in \textit{C. Koukouvinos} (1997, 1998). In this paper the authors used a new algorithm to find new sets of sequences with entries from \(\{0,\pm a,\pm b,\pm c,\pm d\}\), on the commuting variables with zero autocorrelation function. Then they used these sequences to construct a series of new three and four variable orthogonal designs in order 36. They also showed that the necessary conditions plus \((s_1,s_2,s_3,s_4)\) not equal to \[ \begin{matrix} 1\;2\;8\;16 & 1\;8\;8\;16 & 2\;2\;13\;13 & 2\;6\;7\;21 & 3\;6\;8\;16 & 4\;8\;8\;9\\ 1\;2\;8\;25 & 1\;9\;13\;13 & 2\;3\;4\;24 & 2\;8\;9\;9 & 3\;8\;10\;15 & 8\;8\;9\;9\\ 1\;4\;4\;25 & 2\;2\;9\;16\end{matrix} \] are sufficient for the existence of an \(\text{OD}(36;s_1,s_2,s_3,s_4)\), which was constructed by using four circulant matrices in the Goethals-Seidel array. Out of the 154 theoretically possible cases 133 are known. They further showed that the necessary conditions plus \((s_1, s_2, s_3)\neq (2,8,25)\), \((6,7,21)\), \((8,9,17)\) or \((9,13,13)\) are sufficient for the existence of an \(\text{OD}(36; s_1,s_2,s_3)\), which was constructed by using four circulant matrices in the Goethals-Seidel array. Then out of the 433 theoretically possible cases 429 designs are known to exist. Again the authors showed that the necesary conditions are sufficient for the existence of an \(\text{OD}(36; s_1,s_2, 36-s_1- s_2)\) in each of the 54 theoretically possible cases. Further, of the 27 theoretically possible cases \(\text{OD}(36; s_1,s_2,s_3, 36- s_1-s_2-s_3)\), 23 designs are known to exist, and four designs, namely \((1, 2, 8, 25)\), \((1, 9, 13, 13)\), \((2, 6, 7, 21)\), and \((3, 8, 10, 15)\), could not be constructed by using four circulant matrices. By suitably replacing the variables by \(\pm 1\) these lead to more than 200 potentially inequivalent Hadamard matrices of order 36. By considering 12 \(\text{OD}(36; 1,s_1,35- s_1)\) and suitably replacing the variables by \(\pm 1\) they obtained 48 potentially inequivalent skew-Hadamard matrices of order 36. Finally all known results in order 36 (sequences with zero periodic autocorrelation function) are presented in tables at the end of the paper.