On a continuous analog of Sperner's problem (Q1073035): Difference between revisions

Let the set of partitions of an n-element set S be partially ordered by refinement, and define the rank of a partition of S to be the number of nonempty subsets in it, so that the number of partitions of rank k is the Stirling number S(n,k) of the second kind. It was conjectured by Rota that the size of the largest antichain of partitions of S is \(\max_{k}S(n,k)\). However, \textit{R. Canfield} [Advances Math. 29, 1-10 (1978; Zbl 0378.05003)] showed that if n is sufficiently large then the conjecture is false: larger antichains exist. This raised the following question: is the ratio of the size of the largest antichain to \(\max_{k}S(n,k)\) asymptotic to 1? By using arguments based on the assumption that the lattice of partitions may be approximated by a Gaussian process, the author shows that this asymptotic conjecture is likely to be false, there existing an antichain whose size is at least 1.69 times the largest Stirling number.