The Density Finite Sums Theorem
By Bryna Kra and Joel Moreira and Florian K. Richter · 2025
Read The Density Finite Sums Theorem by By Bryna Kra and Joel Moreira and Florian K. Richter (2025) 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-2504-06424.txt.
Introduction
Bryna Kra and three co-authors prove a new theorem in this 2025 arXiv paper at the meeting point of number theory and dynamics. Hindman's classic result says that any large set of whole numbers, split any way, contains a rich pattern of sums. This paper asks for a density version: take any set that fills a positive share of the numbers in a strong sense.
Inside a shifted copy of that set, they find all sums of up to any fixed count of distinct elements drawn from one infinite pool. In other words, finite sum patterns of every bounded size hide inside dense sets, once a single shift is allowed. The proof weaves together sumset theory, the dynamics behind arithmetic progressions, and the geometry of nilpotent Lie groups. The result strengthens a central theme of Ramsey-style mathematics: density alone forces deep structure.
Catalog record
| CREATOR | By Bryna Kra and Joel Moreira and Florian K. Richter |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 74 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-06424
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
The Density Finite Sums Theorem By Bryna Kra and Joel Moreira and Florian K. Richter Abstract For any set A A of natural numbers with positive upper Banach density and any k ⩾ 1 k\geqslant 1 , we show the existence of an infinite set B ⊂ ℕ B\subset\mathbb{N} and a shift t ⩾ 0 t\geqslant 0 such that A − t A-t contains all sums of m m distinct elements from B B for all m ∈ { 1 , … , k } m\in\{1,\ldots,k\} . This can be viewed as a density analog of Hindman’s finite sums theorem. Our proof reveals the natural relationships among infinite sumsets, the dynamics underpinning arithmetic progressions, and homogeneous spaces of nilpotent Lie groups. Contents 1. Introduction 1.1. Infinite sumsets In 1927, van der Waerden [ 32 ] proved a classic result in Ramsey theory: no matter how one partitions the natural numbers \mathbb{N}=\{1,2,3,\ldots\} into finitely many pieces, some piece contains arbitrarily long arithmetic progressions. Settling a long standing conjecture of Erdős and Turán, Szemerédi [ 30 ] proved in 1975 a density version of van der Waerden’s theorem. Namely, he showed that any set A\subset\mathbb{N} whose upper Banach density , meaning \mathsf{d}^{*}(A)\coloneqq\lim_{N\to\infty}\sup\left\{\dfrac{|A\cap\{M+1,\dots,M+N\}|}{N}:M\in\mathbb{N}\right\}, is positive contains arbitrarily long arithmetic progressions. Around the same time, in a major breakthrough, Hindman [ 13 ] settled a conjecture of Graham and Rothschild showing that no matter how one partitions the natural numbers into finitely many pieces, some piece always contains an IP-set (also called a finite sums set) \displaystyle\big\{b(i_{1})+\dots+b(i_{m}):m\in\mathbb{N},i_{1}<\cdots<i_{m}\in\mathbb{N}\big\} (1.1) for some strictly increasing sequence b\colon\mathbb{N}\to\mathbb{N} . Erdős [ 5 ] then tried to find the underlying behavior connecting these two results, writing “I have tried to formulate a conjecture which would be in the same relation to Hindman’s theorem as Szemerédi’s theorem is to van der Waerden’s.” We believe that our main theorem comes as close as is possible to giving a density version. Theorem 1.1 . If A\subset\mathbb{N} has positive upper Banach density then for every k\in\mathbb{N} there is a strictly increasing sequence b\colon\mathbb{N}\to\mathbb{N} and an integer t\geq 0 such that \big\{b(i_{1})+\dots+b(i_{m}):1\leq m\leq k,\penalty\ i_{1}<\cdots<i_{m}\in\mathbb{N}\big\} (1.2) is a subset of A-t . The shift is in general necessary: for example, the odd numbers do not contain a configuration of the form ( 1.2 ) when k\geq 2 . Furthermore, an example of Straus (published by others, see for example [ 14 ] ) exhibits a set with density arbitrarily close to 1 that contains no shift of an IP-set, explaining the need to curtail m . The k=2 case of Theorem 1.1 was proved in our previous work [ 24 ] . The methods used there do not generalize to the setting of Theorem 1.1 , which was conjectured in [ 24 , Conjecture 1.5] and [ 25 , Conjecture 2.4] . Apart from the initial translation of Theorem 1.1 into a dynamical statement ( Theorem 1.2 ), our proof of Theorem 1.1 is new even in the case k=2 and gives, in particular, a simpler proof of the main result in [ 24 ] . Theorem 1.1 follows a long history of results on infinite sumset configurations in large subsets of the integers, initiated by various conjectures of Erdős from the late 1970s and early 1980s (see [ 8 , p. 85] , [ 5 , p. 305] , [ 6 , pp. 57–58] , and [ 7 , p. 105] ). Hindman [ 14 , 15 ] gave various examples and provided reformulations and refinements of some of these conjectures. Early results include the existence [ 29 ] of sumsets B+C with B\subset\mathbb{N} infinite and C\subset\mathbb{N} finite but arbitrarily large, and the existence [ 1 ] of restricted sumset configurations \{b+c:b\in B,\penalty\ c\in C,\penalty\ b<c\} where B and C are infinite. In [ 4 ] , it was proved that any set of density strictly greater than 1/2 contains B+C with both B and C infinite. In [ 28 ] , the weakest of Erdős’s conjectures was resolved. A more streamlined proof [ 17 ] simplified many of the technical arguments in [ 28 ] . The generalization to k -fold sumsets B_{1}+\cdots+B_{k} with B_{1},\dots,B_{k}\subset\mathbb{N} infinite was proved by the authors in [ 23 ] . Recent related results include work on analogs of Erdős’s conjectures in more general groups [ 3 ] , work on unrestricted sumsets [ 22 , 21 ] , and work on sumsets in the primes [ 31 ] . We refer to our survey [ 25 ] for further references and variations. Our proof of Theorem 1.1 relies on ergodic theory: the first step is to reduce the combinatorial problem to a dynamical statement, following Furstenberg’s framework for recasting combinatorial questions in terms of properties of orbits in a measure preserving system. Our main dynamical theorem, from which we derive Theorem 1.1 in Section 2 , states the following. Theorem 1.2 . Let T be a homeomorphism of a compact metric space X and let \mu be a T -invariant Borel probability measure on X . For every point a\in X with \mu(\overline{\{T^{n}a:n\in\mathbb{N}\}})=1, for every k\in\mathbb{N} , and for every open set E\subset X with \mu(E)>0 , there exist an integer t\geq 0 , points x_{1},\dots,x_{k}\in X , and a strictly increasing sequence c\colon\mathbb{N}\to\mathbb{N} such that \lim_{n\to\infty}T^{c(n)}a=x_{1},\hskip 10.00002pt\lim_{n\to\infty}T^{c(n)}x_{1}=x_{2},\hskip 10.00002pt\ldots,\hskip 10.00002pt\lim_{n\to\infty}T^{c(n)}x_{k-1}=x_{k}, (1.3) all hold, and x_{i}\in T^{-t}E for all i=1,\dots,k . The proof that Theorem 1.2 implies Theorem 1.1 is contained in Section 2 and uses by now well-understood techniques. The proof of Theorem 1.2 then occupies the rest of the paper. 1.2. Outline of the proof of Theorem 1.2 We describe the strategy used to prove Theorem 1.2 , referring as needed to terminology and notation introduced later in the paper. We start by introducing a new class of measures on X^{k+1} defined by a recurrence property which we call progressive ( Definition 3.1 ). We show that if a measure on X^{k+1} is progressive, then any open set U\subset X^{k+1} with positive measure contains a point (a,x_{1},\dots,x_{k}) satisfying ( 1.3 ) – we call such points Erdős progressions ( Definition 2.1 ). Thus every progressive measure is supported on the closure of the set of Erdős progressions. Our goal, then, is to associate to each system (X,\mu,T) a measure \sigma on X^{k+1} that is progressive. In the extreme case when the system is a rotation on a compact abelian group, there is a concise algebraic description of the set of Erdős progressions, which allows for a simple and explicit construction of a progressive measure (see Example 4.1 ). More generally, if X is a nilmanifold and (X,\mu,T) is a nilsystem (see Definition 4.3 ), one can also give an explicit description of a progressive measure as the Haar measure on a suitable sub-nilmanifold. In the opposite case of weakly mixing systems, we can easily show that the product measure is progressive due to the randomness inherent in such a system (see Example 4.2 ). Motivated by the insights of these two contrasting examples, we construct in Section 4 for any system (X,T) a measure \sigma on X^{k+1} (see Definition 4.5 ) consisting of a structured component arising from the maximal pronilfactor of (X,T) and a