Most Latin squares have many subsquares (Q1291124)

This (in my opinion excellent and important) paper contains a wealth of results on subsquares of Latin squares. The attention is focused on subsquares of order 2 (intercalates). The methods and results are too technical to be reproduced here but some of the more easily stated corollaries include: (1) For arbitrary \(\varepsilon> 0\), with probability approaching \(1\) as \(n\to\infty\), a random Latin squares of order \(n\) contains at least \(n^{(3/2-\varepsilon)}\) intercalates. (2) For arbitrary \(\varepsilon> 0\), the probability of a random Latin square of order \(n\) not containing any intercalates is \(O(\exp(-n^{(2-\varepsilon)})\) as \(n\to\infty\). The results of computer enumeration for small order Latin square as well as a brief consideration of larger order subsquares are also included. As the authors suggest, a possible subtitle could be ``\dots however, almost all of those subsquares are of order 2''.