Equidistribution of subset sums
Péter Pál Pach
TL;DR
The paper resolves questions of Katona and Makar-Limanov on the equidistribution of $h$-element subset sums in abelian groups by proving asymptotic equidistribution for all abelian groups when the group order $n$ satisfies $n\ge 2h$, and by establishing a growing-$h$ version with an explicit error term. The core method uses an inclusion-exclusion decomposition via partitions of the index set to count distinct coordinates, combined with a gcd-based analysis to isolate uniformly distributed contributions. The main result gives a precise main term $(1/n)\cdot(n!)/(n-h)!$ for the count of representations of each $a\in G$, with a geometric error bound $\le (3/4)^h$ times the main term, and extends to the regime where $h$ grows with $n$, yielding equidistribution for the important case $n=2h$. Collectively, these findings resolve several open problems posed by KML08 and have potential implications for related coding-theoretic questions and combinatorial design.
Abstract
We answer a question of Katona and Makar-Limanov, by showing that in an abelian group of order $2h$ the $h$-element subset sums are asymptotically (as $h\to \infty$) equidistributed. In fact we prove a more general result where the order of the group can be arbitrary, also providing a bound for the ``error term''.
