A complete solution of the Erdős-Kleitman matching problem for $n\le 3s$
Andrey Kupavskii, Georgy Sokolov
TL;DR
The paper resolves the Erdős–Kleitman non-uniform matching problem e(n,s) for all n ≤ 3s by identifying four extremal shifted up-set families and proving that the maximum size equals the largest among their complements, e(n,s)=max{|al P(s,ℓ)|,|al P'(s,ℓ)|,|al Q(s,ℓ)|,|al W(s,ℓ)|} for n=2s+c=3s−ℓ with suitable parameter ranges. The authors develop a unified framework around the d(F) parameter and bounds on the 2- and 3-element layers y(2), y(3), using an averaging argument over carefully chosen matchings to constrain the structure of extremal families. A detailed case analysis handles large, moderate, and small d, with c ≥ 5 treated in the main text and c ≤ 4 in an appendix, culminating in a complete classification of extremals in the specified regime. This work advances understanding of the EMC landscape, provides exact values in a broad new range, and demonstrates the pivotal role of four structured extremal families in the non-uniform setting.
Abstract
Given integers $n\ge s\ge 2$, let $e(n,s)$ stand for the maximum size of a family of subsets of an $n$-element set that contains no $s$ pairwise disjoint members. The study of this quantity goes back to the 1960s, when Kleitman determined $e(sm-1,s)$ and $e(sm,s)$ for all integer $m,s\ge 1$. The question of determining $e(n,s)$ is closely connected to its uniform counterpart, the subject of the famous Erdős Matching Conjecture. The problem of determining $e(n,s)$ has proven to be very hard and, in spite of some progress during these years, even a general conjecture concerning the value of $e(n,s)$ is missing. In this paper, we completely solve the problem for $n\le 3s$. In this regime, the average size of a set in an $s$-matching is at most $3$, and it is a delicate interplay between the `missing' $2$- and $3$-element sets that plays a key role here. Four types of extremal families appear in the characterization. Our result sheds light on how the extremal function $e(n,s)$ may behave in general.
