Generalized Latin squares and their defining sets (Q2427495)

A generalized Latin square of type \((n, k)\) is an \(n\times n\) array whose entries consist of \(1,2,\dots,k\) such that each symbol occurs at most once in each row and column. The defining number, \(d(n, k)\), is the size of the minimal set of entries that uniquely extends to a generalized Latin square of type \((n, k)\). Clearly \(d(n, k)= n^2\) for \(k> 2n- 1\), but for \(n\leq k< 2n- 1\), \(d(n, k)\) is unknown. \textit{M. Mahdian} and \textit{E. S. Mahmoodian} [Bull. Inst. Comb. Appl. 28, 48--54 (2000; Zbl 0943.05037)] established that \(d(n,2n- 1)= n^2- n\) for \(n\) even, and for odd \(n> 1\), \(d(n,2n- 1)= n^2- n+ 1\). The author provides an alternative proof for \(d(n, 2n- 1)\) for even \(n\) which is exploited in establishing \(d(n,2n- 2)\geq n^2- \lfloor{8n\over 5}\rfloor\), with \(d(n,2n- 2)= n^2- \lfloor{8n\over 5}\rfloor\) for \(n= 5\) and \(n\) divisible by 10.