Frames and doubly resolvable group divisible designs with block size three and index two (Q6063287)

A \(2\)-\([u, k, \lambda]\) block design \((X, \mathcal{B})\) is a combinatorial structure formed by a \(u\)-set \(X\) of points together with a family \(\mathcal{B}\) of \(k\)-subsets, called blocks, such that any \(2\) distinct points are contained in exactly \(\lambda\) blocks. A \(2\)-\([u, k, \lambda]\) block design \((X, \mathcal{B})\) is resolvable, if there exists a partition of \(\mathcal{B}\), called resolution, into classes of pairwise disjoint (or parallel) blocks, such that any class of parallelism is a partition of the whole \(X\). If \((X, \mathcal{B})\) has two different resolutions such that any class of the first resolution has in common at most one block with any class of the second resolution, the block design is said to be doubly resolvable. Resolvable designs are part of the history of Combinatorics, whose birth is often dated on Kirkman's schoolgirl problem, technically the query for a resolvable \(2\)-\([15, 3, 1]\) block design with \(35\) blocks in \(7\) classes of \(5\) parallel blocks of \(3\) points. A \(2\)-\([u, k, \lambda]\) block design is a group divisible design, if there exists a partition \(\mathcal{G}\) of \(X\) into subsets, called groups, such that any block and any group have at most one point in common, and the group divisible design is of type \(h^m\), if all the groups have the same cardinality \(h=u/m\), where \(m\) is the cardinality of the partition \(\mathcal{G}\). The existence of a doubly resolvable \(2\)-\([u, 3, 1]\) group divisable block design of type \(1^u\) has been settled in [\textit{R. J. R. Abel} et al., J. Comb. Des. 21, No. 7--8, 342--358 (2013; Zbl 1269.05015); Discrete Math. 308, No. 7, 1102--1123 (2008; Zbl 1134.05010); \textit{C. J. Colbourn} et al., Des. Codes Cryptography 26, No. 1--3, 169--196 (2002; Zbl 0995.05021)], and that of type \(2^{u/2}\) in [Abel et al., 2013, loc. cit.], for \(u\) even. Doubly resolvable \(2\)-\([u, 3, 2]\) group divisable block design of type \(1^m\), \(3^m\), and \(6^m\) have been investigated in [Abel et al., 2008. loc. cit.; \textit{E. R. Lamken}, J. Comb. Theory, Ser. A 72, No. 1, 50--76 (1995; Zbl 0836.05008); \textit{J. Shi} and \textit{J. Wang}, Graphs Comb. 36, No. 5, 1525--1543 (2020; Zbl 1459.05026)]. In the present paper, the existence of doubly resolvable \(2\)-\([u, 3, 2]\) group divisible block design of type \(h^m\) is settled under the necessary assumptions \(m\geq 3\) and \(hm \equiv 0\mod 3\) and a list of possible exceptions. The construction method makes use of uniform \((1, 2; 3)\)-frames of the \(2\)-\([u, 3, 2]\) group divisible block design \((X, \mathcal{B})\), that is, a square \(u \times u\) array of cells, such that each cell is either empty or contains a block, all blocks appear in the array, the \(m\) disjoint \(h \times h\) squares in the main diagonal are empty, and each point of \(X\), not in the \(i\)-th group of \(\mathcal{G}\), appears exactly once in each row (resp. column) from \((i-1)h\) to \(ih-1\). The existence of \((1, 2; 3)\)-frames of type \(1^u\) for all \(u \equiv 1 \mod 3\) and greater than 9, with the possible exception of \(u=13\), was proved in [Abel et al., 2008, loc. cit.]. The authors give sufficient conditions for the existence of \((1, 2; 3)\)-frames of type \(h^m\) in almost all cases.