Extensions of the Furstenberg–Sárközy theorem via the arithmetic level- d d inequality

Carlo Francisco E. Adajar, Rishika Agrawal, Mukul Rai Choudhuri, Chian Yeong Chuah, Steve Fan, Swaroop Hegde, Andrew Lott, Krishnamohan Nandakumar, Nagendar Reddy Ponagandla · 2026

Read Extensions of the Furstenberg–Sárközy theorem via the arithmetic level- d d inequality by Carlo Francisco E. Adajar, Rishika Agrawal, Mukul Rai Choudhuri, Chian Yeong Chuah, Steve Fan, Swaroop Hegde, Andrew Lott, Krishnamohan Nandakumar, Nagendar Reddy Ponagandla (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-16216.txt.

Introduction

Adajar and eight colleagues extend the Furstenberg-Sárközy theorem to general intersective polynomials, building on Green and Sawhney's arithmetic level-d inequality. For sets lacking intersective polynomial differences, they prove a bound that decays as X times an exponential in log X to the power mu, for every mu below one half. This is the best quantitative bound presently known for such sets, improving on Arala's work. Novel exponential sum estimates via sieve methods and random sparsification make the full exponent range admissible.

Catalog record

CREATORCarlo Francisco E. Adajar, Rishika Agrawal, Mukul Rai Choudhuri, Chian Yeong Chuah, Steve Fan, Swaroop Hegde, Andrew Lott, Krishnamohan Nandakumar, Nagendar Reddy Ponagandla
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE105046
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.16216

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Extensions of the Furstenberg–Sárközy theorem via the arithmetic level- d d inequality Carlo Francisco E. Adajar, Rishika Agrawal, Mukul Rai Choudhuri, Chian Yeong Chuah, Steve Fan, Swaroop Hegde, Andrew Lott, Krishnamohan Nandakumar, Nagendar Reddy Ponagandla Abstract Very recently, Green and Sawhney obtained a quasipolynomial bound in the Furstenberg–Sárközy theorem for square differences by proving an “arithmetic level- d d ” inequality, thereby yielding a greatly improved density increment scheme. We adapt their method to general intersective polynomials h ∈ ℤ ⁡ [ x ] h\in\mathbb{Z}[x] and obtain an analogous quasipolynomial upper bound for the largest subset of { 1 , 2 , … , X } \{1,2,\dots,X\} whose difference set contains no nonzero element of the form h ⁡ ( n ) h(n) with n ∈ ℤ n\in\mathbb{Z} . This is the best quantitative upper bound presently known for sets lacking intersective polynomial differences. In contrast to the square case, extending the method to general intersective polynomials requires performing a density increment iteration in which the underlying polynomial changes at each step; a key contribution of this paper is to show that the arithmetic level- d d inequality remains effective uniformly across all auxiliary polynomials arising in the iteration. We also develop smoothly weighted versions of the exponential sum estimates of Rice. 1 University of Georgia 2 HUN-REN Alfréd Rényi Institute of Mathematics (Erdős Center) 1 Introduction For a set F\subseteq\mathbb{Z} and X>1 , write D(F,X)\ :=\ \max\bigl\{|A|:\ A\subseteq\{n\in\mathbb{N}:n\leq X\}\ \text{and}\ (A-A)\cap F\subseteq\{0\}\bigr\}. Here, we refer to F as the set of forbidden differences . The Furstenberg–Sárközy theorem asserts that D(\{n^{2}:n\in\mathbb{N}\},X)=o(X) , answering a question of Lovász. This was proved independently by Furstenberg [ 8 ] using ergodic theory and by Sárközy [ 23 , 24 ] using the Hardy–Littlewood circle method and the density increment method of Roth; the latter gives the quantitative bound D(\{n^{2}:n\in\mathbb{N}\},X)\ll X(\log\log X)^{2/3}(\log X)^{-1/3}. Pintz–Steiger–Szemerédi [ 19 ] introduced a more elaborate double iteration strategy to obtain a bound of the form D(\{n^{2}:n\in\mathbb{N}\},X)\ll X(\log X)^{-c_{1}\log\log\log\log X} with c_{1}=1/12 , and Bloom–Maynard [ 3 ] later strengthened this to D(\{n^{2}:n\in\mathbb{N}\},X)\ll X(\log X)^{-c_{2}\log\log\log X} where c_{2}>0 is an absolute constant. Very recently, Green and Sawhney [ 9 ] developed a new “arithmetic level- d ” inequality (building on hypercontractive ideas of Keller–Lifshitz–Marcus [ 14 ] and Keevash–Lifshitz–Long–Minzer [ 13 ] ) and pushed the density increment strategy to obtain a quasipolynomial bound D(\{n^{2}:n\in\mathbb{N}\},X)\ll X\exp\bigl(-c_{3}(\log X)^{1/4}(\log\log X)^{-1/2}\bigr) (1.1) for some constant c_{3}>0 . Later, Green and Sawhney posted a second version of their argument [ 10 ] , improving ( 1.1 ) to D(\{n^{2}:n\in\mathbb{N}\},X)\ll X\exp\bigl(-c_{4}(\log X)^{1/2}\bigr) (1.2) with some absolute constant c_{4}>0 . We record the arithmetic level- d inequality. Theorem A . Set C_{0}:=2^{13} . Let \alpha\in(0,1/2) and let X\geq 1 be parameters with \alpha>2X^{-1/2} . Let \mathcal{Q} be a set of pairwise coprime positive integers such that \max_{q\in\mathcal{Q}}q\leq X^{1/32\log(1/\alpha)} . Let 1\leq d\leq 2^{-7}\log(1/\alpha) . Let f:[X]\rightarrow\mathbb{C} be a function with such that |f(x)|\leq 1 for all x . Then either \sum_{\begin{subarray}{c}S\subseteq\mathcal{Q}\\ |S|=d\end{subarray}}\sum_{\begin{subarray}{c}a\,(\operatorname{mod}\,\prod_{q\in S}q)\\ q\in S\Rightarrow q\nmid a\end{subarray}}\Big|\widehat{f}\Big(\frac{a}{\prod_{q\in S}q}\Big)\Big|^{2}\leq\alpha^{2}X^{2}\Big(\frac{C_{0}\log(1/\alpha)}{d}\Big)^{d}, (1.3) or else for some set S\subseteq\mathcal{Q} , 1\leq|S|\leq 2\log(1/\alpha) , and for some r\in\mathbb{Z} , the average of |f(x)| on the progression P=\{x\in[X]:x\equiv r\,(\operatorname{mod}\,\prod_{q\in S}q)\} is greater than 2^{|S|}\alpha . A natural generalization of the Furstenberg–Sárközy theorem is to replace squares by the image of a polynomial. If h\in\mathbb{Z}[x] is nonconstant, then a necessary condition for any decay D(h(\mathbb{Z}),X)=o(X) is that h(\mathbb{Z}) contain a nonzero multiple of every q\in\mathbb{N} , in which case we say that h is intersective 1 1 1 This definition makes sense for integer-valued polynomials in \mathbb{Q}[x] as well. Given a nonconstant integer-valued polynomial h\in\mathbb{Q}[x] , there exists m\in\mathbb{N} such that f:=mh\in\mathbb{Z}[x] . It is clear that if h is intersective, then so is f . Moreover, we have D(\{h(n)\colon n\in\mathbb{N}\},X)\leq D(\{f(n)\colon n\in\mathbb{N}\},mX) . . Kamae and Mendès France [ 12 ] showed that intersectivity is also sufficient for the qualitative conclusion D(h(\mathbb{Z}),X)=o(X) . On the quantitative side, Lucier [ 17 ] proved the first general density-decay bound for intersective polynomials: if h has degree \deg h=k\geq 2 , then D(h(\mathbb{Z}),X)\ll_{h}X\frac{(\log\log X)^{\mu/(k-1)}}{(\log X)^{1/(k-1)}}, where \mu=\begin{cases}3,&k=2,\\ 2,&k\geq 3.\end{cases} Stronger decay of Pintz–Steiger–Szemerédi type was previously available in special cases (e.g. for monomials [ 2 ] and for intersective quadratics [ 11 ] ), and Rice [ 21 ] fully extended this to all intersective polynomials, obtaining bounds of the form D(h(\mathbb{Z}),X)\ll_{h,c_{5}}X(\log X)^{-c_{5}\log\log\log\log X}. for some constant c_{5}>0 depending on k . In order to prove this bound, Rice developed a method of achieving square root cancellation for local Gauss sums, which was necessary to perform the double iteration strategy of Pintz–Steiger–Szemerédi. More recently, Arala [ 1 ] used the exponential sum estimates of Rice to generalize the Bloom–Maynard proof to the full class of intersective polynomials, yielding the bound D(h(\mathbb{Z}),X)\ll_{h}X(\log X)^{-c_{6}\log\log\log X}, for some constant c_{6}>0 depending on k . The objective of this paper is to prove the following generalization of ( 1.1 ). Theorem 1 . Let 0<d<1/4 . Given any intersective polynomial h\in\mathbb{Z}[x] of degree at least 2 with positive leading term, then there are constants c_{0},X_{\text{min}}>0 depending on h and d , such that D(h(\mathbb{Z}),X)\leq Xe^{-c_{0}(\log X)^{d}} (1.4) whenever X>X_{\text{min}} . In [ 9 ] , Green and Sawhney remark that their method should extend beyond the square case to more general polynomials. We confirm this observation: the arithmetic level- d inequality yields a density increment scheme once sufficient Fourier mass is detected on appropriate major arcs, and this technique is not intrinsically tied to the monomial n^{2} . However, implementing this strategy for general intersective polynomials presents additional technical difficulties. In contrast to the square case, where the forbidden difference set is stable under dilation so that the relevant polynomial remains fixed throughout the iteration, the natural iteration for intersective polynomials necessarily introduces a family of auxiliary polynomials whose coefficients evolve from step to step. As a result, one must establish circle method estimates uniformly across all auxiliary polynomials that arise, while simultaneously controlling the growth of the auxiliary polynomial coefficients in order to ensure that these estimates remain valid at every stage of the iteration. In an earlier draft of the project, we used classical polynomial exponential sum estimates to prove a weaker version of theorem 1 which required 0<d<1/(k(k-1)+2)) , where k is the degree of the intersective polynomial h . We improved this to 0<d<1/4 by developing smoothly weighted versions of Rice’s exponential sum estimates [ 21 ] . As in [ 9 ] , we need the smooth weight to make the major arcs sufficiently narrow for the applications of the arithmetic level- d inequality. There remains a substantial gap between our upper bound in Theorem 1 and the best available lower bounds for D(h(\mathbb{Z}),X) when h\in\mathbb{Z}[x] is intersective (see [ 5 ] ). Indeed, all known lower bounds in this setting are of the form X^{c} for some c<1 . The greedy construction yields c=1-\frac{1}{\deg(h)} , and improvements beyond this are currently only known in special cases, such as monomials and certain polynomials h\in\mathbb{Z}[x] which are divisible by x^{2} (see [ 22 , 16 , 26 , 25 ] ). Remark . We believe that Theorem 1 should hold for 0<d<1/2 using the

Read the complete text