An inverse and a stability result for Ruzsa's inequality on triple sumsets
Swaroop Hegde
TL;DR
This paper investigates inverse and stability phenomena for Ruzsa's inequality on triple sumsets in abelian groups. It develops a quantitatively sharp inverse theorem: if |A+A+A| is close to the universal upper bound given |A+A|, then A contains a large almost-3-dissociated subset Y whose triple sum captures most of A+A+A, aligning with Ruzsa's spline construction. The results extend to higher sumsets |(h+1)A| ≤ |hA|^{(h+1)/h} and yield a 99%-stability version of Ruzsa's inequality, plus explicit extremal constructions demonstrating sharpness and limitations of structural descriptions. The paper also provides a generalized inverse theorem for arbitrary h≥2 and develops a robust toolkit combining Loomis-Whitney, Kruskal-Katona, and Clements-Lindström/Macaulay perspectives, along with Plünnecke-type large-subset inequalities. These contributions deepen the understanding of near-extremal sumset behavior and offer quantitative, broadly applicable structure results in additive combinatorics.
Abstract
Ruzsa's inequality states that $|A+A+A| \leq |A+A|^{3/2}$ for any finite set $A$ in a commutative group. Ruzsa has constructed examples showing that this inequality is sharp asymptotically, up to a constant factor. We prove an inverse result which says that if $|A+A+A| \geq \frac{1}{M} |A+A|^{3/2}$ for some parameter $M,$ then the set $A$ resembles the sets in Ruzsa's construction. We then construct more families of examples which suggest that our inverse result is likely best possible qualitatively. The method extends to give an inverse result for a higher sumset analogue of Ruzsa's inequality, namely $|(h+1)A| \leq |hA|^{\frac{h+1}{h}}$ for any $h\geq 2.$ We also provide a "99%-stability" version of Ruzsa's inequality, which describes near optimal structures when $M$ is very close to $1.$
