🔢 mathematics

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

This paper extends the Green–Sawhney method to general intersective polynomials, establishing a quasipolynomial upper bound for the largest subset of $\{1, \dots, X\}$ avoiding nonzero differences of the form $h(n)$ by proving that the arithmetic level- $d$ inequality remains effective uniformly across the varying polynomials encountered in the density increment iteration.

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

Published 2026-05-18

📖 5 min read🧠 Deep dive

CC BY 4.0

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

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are hosting a massive party with guests numbered from 1 to $X$ . You want to invite as many people as possible, but you have a strict rule: no two guests can have a "difference" that matches a specific pattern.

For example, in the classic version of this problem, the rule is: "No two guests can have an age difference that is a perfect square (like 1, 4, 9, 16...)." The famous Furstenberg–Sárközy theorem proved that if you follow this rule, you can't invite everyone. As the party gets bigger, the percentage of people you can invite must get smaller and smaller, eventually approaching zero.

This paper takes that idea and makes it much more flexible. Instead of just "perfect squares," the forbidden differences can be the result of any complex polynomial formula (like $n^2 + 3n + 5$ , or other shapes), as long as that formula can produce numbers that fit into any modular arithmetic system (a property the authors call "intersective").

Here is the breakdown of what the authors did, using simple analogies:

1. The Problem: Finding the "Forbidden" Shapes

The authors are trying to find the maximum size of a group of numbers that avoids these specific polynomial differences.

The Old Way: Previous mathematicians had good estimates, but they were like using a sledgehammer to crack a nut. The estimates were "polynomial" in nature, meaning the group size shrank slowly as the party got bigger.
The New Goal: They wanted to prove that the group size shrinks much faster—so fast that it's "quasipolynomial." Think of it as switching from a slow leak in a boat to a gaping hole; the group of allowed numbers disappears much more quickly.

2. The Tool: The "Arithmetic Level-d" Inequality

To solve this, the authors used a powerful new mathematical tool recently invented by Green and Sawhney.

The Analogy: Imagine you are trying to find a hidden pattern in a noisy room. You have a "super-sensor" (the inequality) that can detect if the noise is actually a hidden signal.
How it works: If your group of numbers is too big, this sensor will scream, "Hey! There's a hidden structure here!" This structure tells you that the numbers aren't randomly scattered; they are clumped together in a specific way.
The Result: Once you find this clump, you can zoom in on it. Inside this smaller, denser clump, the rules still apply, but now you have a "density increment." You have found a smaller room where the guests are packed even tighter than before.

3. The Twist: The Shape Changes Every Time

This is the hardest part of the paper and their main innovation.

The Square Case (Old Method): When the forbidden difference was just a square ( $n^2$ ), the shape of the problem stayed the same every time you zoomed in. It was like looking at a picture of a square, then zooming in, and seeing a smaller square. The rules were stable.
The General Case (New Method): When the forbidden difference is a complex polynomial (like $n^3 + n$ $n^{3} + n$ ), the shape changes every time you zoom in.
- The Metaphor: Imagine you are looking at a fractal (like a snowflake). When you zoom in on one part, it doesn't look like the whole snowflake; it looks like a slightly different, distorted version of it.
- The Challenge: Every time the authors zoomed in to find a denser group, the "forbidden formula" they had to avoid changed. They had to prove that their "super-sensor" (the inequality) still worked perfectly, even though the shape of the problem was morphing at every single step.

4. The Solution: A Uniform Shield

The authors proved that their "super-sensor" is robust enough to handle these changing shapes.

They showed that no matter how the polynomial morphs during the zooming process, the sensor can still detect the hidden structure.
They also developed a new way to "smooth out" the data (using what they call "smoothly weighted exponential sums"). Imagine trying to count grains of sand on a beach. If you just count them one by one, you might miss some or count the same one twice. By "smoothing" the beach with a gentle brush, you get a much more accurate count of the total volume. This allowed them to make their estimates much sharper.

5. The Conclusion: A Quasipolynomial Bound

By repeatedly zooming in and finding denser and denser clusters of numbers, they proved that the maximum size of a group avoiding these polynomial differences is incredibly small.

The Result: They established a bound that looks like $X \times e^{-(\log X)^d}$ .
In Plain English: If you have a party with $X$ guests, the number of people you can invite without breaking the rule is roughly $X$ divided by a number that grows faster than any power of $\log X$ . It's a massive reduction.

Summary

The authors took a famous theorem about avoiding square differences and generalized it to any polynomial difference. The difficulty was that the "rules of the game" changed every time they tried to find a denser group of numbers. They overcame this by proving their detection tool works uniformly across all these changing rules, resulting in the best possible mathematical estimate for how small these groups must be.

Note on Limitations: The paper is purely theoretical mathematics. It does not discuss applications to computer science, cryptography, physics, or any real-world clinical uses. It is a proof about the fundamental structure of numbers.

Technical Summary: Extensions of the Furstenberg–Sárközy Theorem via the Arithmetic Level-d Inequality

Problem Statement
The paper addresses the quantitative bounds for the Furstenberg–Sárközy theorem, specifically concerning sets of integers that lack non-zero differences of the form $h(n)$ , where $h \in \mathbb{Z}[x]$ is an intersective polynomial. Let $D(F, X)$ denote the maximum size of a subset of $\{1, \dots, X\}$ whose difference set contains no non-zero element of a forbidden set $F$ . The classical theorem establishes that $D(\{n^2\}, X) = o(X)$ .

While recent work by Green and Sawhney [9, 10] achieved a quasipolynomial bound for the square case ( $h(n)=n^2$ ) using an "arithmetic level- $d$ " inequality, extending these results to general intersective polynomials presents significant technical hurdles. The primary challenge lies in the fact that, unlike the square case where the forbidden difference set is stable under dilation, general intersective polynomials require an iteration scheme where the underlying polynomial changes at each step. This necessitates uniform control over a family of auxiliary polynomials whose coefficients may grow during the iteration.

Methodology
The authors adapt the density increment strategy of Green and Sawhney to the setting of general intersective polynomials. The proof proceeds through the following key methodological components:

Auxiliary Polynomials and Iteration: The authors utilize the framework of auxiliary polynomials $h_\ell$ (defined via $p$ -adic roots of $h$ ) to reformulate the problem on rescaled subsets. A critical technical contribution is demonstrating that the coefficients of these auxiliary polynomials remain controlled throughout the density increment iteration, ensuring that circle method estimates remain valid at every stage.
Arithmetic Level- $d$ Inequality: The core engine of the proof is the application of the arithmetic level- $d$ inequality (Theorem A) of Green and Sawhney. This inequality allows for the localization of Fourier mass to much smaller denominators than classical methods. The authors apply this inequality twice in a case analysis to refine the range of moduli $q$ where the Fourier coefficients of the characteristic function of the set are large.
Smoothly Weighted Exponential Sums: To achieve the necessary "square root cancellation" for local Gauss sums required by the double iteration strategy, the authors develop smoothly weighted versions of Rice's exponential sum estimates [21]. This involves constructing a smooth weight function $w$ and analyzing the exponential sums of the auxiliary polynomials over sifted sets $W_\ell(U)$ .
Density Increment Scheme: The proof establishes a density increment proposition (Proposition 5.1) which guarantees that if a set $A$ lacks $h_\ell$ -differences, it either has a very small density (quasipolynomially small) or admits a density increment on a subprogression. The iteration of this proposition, combined with careful bookkeeping of the auxiliary polynomial coefficients, yields the final bound.

Key Contributions

Generalization to Intersective Polynomials: The paper extends the quasipolynomial bound from the square case to all intersective polynomials $h \in \mathbb{Z}[x]$ of degree at least 2.
Uniform Control of Auxiliary Polynomials: The authors provide a rigorous argument showing that the arithmetic level- $d$ inequality remains effective uniformly across the family of auxiliary polynomials generated during the iteration, overcoming the issue of coefficient growth.
Improved Exponential Sum Estimates: The development of smoothly weighted versions of Rice's estimates allows for the necessary narrowing of major arcs, which is essential for the application of the arithmetic level- $d$ inequality in the general polynomial setting.
Quantitative Improvement: The paper improves upon previous bounds for intersective polynomials (such as those by Lucier [17], Rice [21], and Arala [1]), which were of the form $X(\log X)^{-c \log \log \log X}$ .

Main Results
The central result is Theorem 1, which states that for any intersective polynomial $h \in \mathbb{Z}[x]$ of degree at least 2 with a positive leading term, there exist constants $c_0, X_{\min} > 0$ (depending on $h$ and a parameter $d$ ) such that for $X > X_{\min}$ :
$D(h(\mathbb{Z}), X) \leq X e^{-c_0 (\log X)^d}$
where $0 < d < 1/4$ .

The authors note that the exponent $d$ can be improved to $d < 1/2$ using ideas from the second version of Green and Sawhney's argument [10], a refinement they intend to incorporate in a future update.

Significance and Claims
The paper claims to provide the best quantitative upper bound presently known for sets lacking intersective polynomial differences. The authors emphasize that their method confirms the observation by Green and Sawhney that the arithmetic level- $d$ inequality is not intrinsically tied to the monomial $n^2$ but can be adapted to general polynomials.

The paper acknowledges a substantial gap between this upper bound and the best known lower bounds (which are of the form $X^c$ for $c < 1$ ). The authors do not claim to close this gap but rather to significantly improve the upper bound landscape for this class of problems. The work relies heavily on the interplay between the circle method, sieve theory, and the specific combinatorial structure of the arithmetic level- $d$ inequality.