Existence of generalized Latin squares which are not embeddable in any group (Q1705831)

This paper is one of a series of recent papers (including several by the present authors) that look at the combinatorial structure of submatrices of the Cayley table of groups. A generalised Latin square (GLS) is a matrix of symbols in which no symbol is repeated within a row or within a column. A partial Latin square (PLS) is like a GLS, but some entries are allowed to be left vacant. Let \(S\) be a finite subset of some group \(G\). The multiplication table of \(S\) is a GLS, which the authors would say embeds in \(G\). They are interested in finding GLS which cannot be embedded in this way (that is, for which no such \(S\) and \(G\) exist). They prove two results. Firstly they show that for \(n\geq6\) and \(m\in\{n,n+1,\dots,n(n+1)/2-2\}\) there is a symmetric GLS of order \(n\) with exactly \(m\) distinct symbols in it, which is not embeddable in any group. They prove this by building into their GLS a partial Latin square that was found by \textit{I. M. Wanless} and \textit{B. S. Webb} [Australas. J. Comb. 67, 352--363 (2017; Zbl 1375.05041)] to be the equal smallest PLS which cannot be found inside any Cayley table. Their second result is that for \(n\geq 4\) and \(m\in\{n,n+1,\dots,n^2-2\}\) there is a non-symmetric GLS of order \(n\) with exactly \(m\) distinct symbols in it, which is not embeddable in any group. Again this achieved by finding a small obstacle to embedding and including it in GLS for all the required values of \(m\) and \(n\). To this reviewer's mind it makes more sense to focus on the PLS inside the GLS which is the true obstacle to embedding in a Cayley table. Once such an obstacle is identified, any other entries which are filled in around it are irrelevant. The aforementioned paper of Wanless and Webb finds the smallest PLS that do not embed in any group, and also the smallest PLS that do not embed in any abelian group. Perhaps even more interesting is a recent result of \textit{H. Dietrich} and \textit{I. M. Wanless} [J. Symb. Comput. 86, 142--152 (2018; Zbl 1380.05019)], who found the smallest PLS which can be embedded in an infinite group, but which cannot be embedded in any finite group. This was motivated by a question asked by \textit{R. Hirsch} and \textit{M. Jackson} [J. Symb. Log. 77, No. 4, 1211--1244 (2012; Zbl 1279.03084)] who first identified a PLS with the stated property.