Optimal Polynomial Recurrence

TL;DR

This paper proves that for polynomial shifts, large sets contain near-expected intersections, and the set of near-optimal return times is syndetic and dense in long intervals, using Fourier analysis.

Contribution

It establishes quantitative bounds for recurrence and intersection properties involving polynomial shifts, extending classical results with explicit bounds and density properties.

Findings

Existence of n with near-expected intersection proportion for large N.

Set of extit{VE}-optimal return times is syndetic.

Set of extit{VE}-optimal return times is dense in long intervals.

Abstract

Let $P\in\Z[n]$ with $P(0)=0$ and $\VE>0$. We show, using Fourier analytic techniques, that if $N\geq \exp\exp(C\VE^{-1}\log\VE^{-1})$ and $A\subseteq\{1,\...,N\}$, then there must exist $n\in\N$ such that \[\frac{|A\cap (A+P(n))|}{N}>(\frac{|A|}{N})^2-\VE.\] In addition to this we also show, using the same Fourier analytic methods, that if $A\subseteq\N$, then the set of \emph{$\VE$-optimal return times} \[R(A,P,\VE)=\{n\in \N \,:\,\D(A\cap(A+P(n)))>\D(A)^2-\VE\}\] is syndetic for every $\VE>0$. Moreover, we show that $R(A,P,\VE)$ is \emph{dense} in every sufficiently long interval, in the sense that there exists an $L=L(\VE,P,A)$ such that \[|R(A,P,\VE)\cap I| \geq c(\VE,P)|I|\] for all intervals $I$ of natural numbers with $|I|\geq L$ and $c(\VE,P)=\exp\exp(-C\,\VE^{-1}\log\VE^{-1})$.