Intersective polynomials and polynomial Szemeredi theorem

TL;DR

This paper characterizes when polynomial families have the polynomial Szemerédi property, linking it to joint intersectivity, and provides a new ergodic proof of the polynomial Szemerédi theorem with a related combinatorial corollary.

Contribution

It establishes a precise criterion for polynomial families to have the PSZ property, connecting it to joint intersectivity, and introduces a novel ergodic approach using nilmanifolds.

Findings

Polynomial families have PSZ property iff they are jointly intersective.

New ergodic proof of the polynomial Szemerédi theorem based on nilmanifolds.

Generalization of polynomial van der Waerden theorem for jointly intersective polynomials.

Abstract

Let $P=\{p_{1},\ld,p_{r}\}\subset\Q[n_{1},\ld,n_{m}]$ be a family of polynomials such that $p_{i}(\Z^{m})\sle\Z$, $i=1,\ld,r$. We say that the family $P$ has {\it PSZ property} if for any set $E\sle\Z$ with $d^{*}(E)=\limsup_{N-M\ras\infty}\frac{|E\cap[M,N-1]|}{N-M}>0$ there exist infinitely many $n\in\Z^{m}$ such that $E$ contains a polynomial progression of the form \hbox{$\{a,a+p_{1}(n),\ld,a+p_{r}(n)\}$}. We prove that a polynomial family $P=\{p_{1},\ld,p_{r}\}$ has PSZ property if and only if the polynomials $p_{1},\ld,p_{r}$ are {\it jointly intersective}, meaning that for any $k\in\N$ there exists $n\in\Z^{m}$ such that the integers $p_{1}(n),\ld,p_{r}(n)$ are all divisible by $k$. To obtain this result we give a new ergodic proof of the polynomial Szemer\'{e}di theorem, based on the fact that the key to the phenomenon of polynomial multiple recurrence lies with the dynamical systems defined by translations on nilmanifolds. We also obtain, as a corollary, the following generalization of the polynomial van der Waerden theorem: If $p_{1},\ld,p_{r}\in\Q[n]$ are jointly intersective integral polynomials, then for any finite partition of $\Z$, $\Z=\bigcup_{i=1}^{k}E_{i}$, there exist $i\in\{1,\ld,k\}$ and $a,n\in E_{i}$ such that $\{a,a+p_{1}(n),\ld,a+p_{r}(n)\}\sln E_{i}$.