Skip to content

Erdős Problem #335

Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: open (not automatically the live result).
Subject areas: number theory, additive combinatorics.

Read this status correctly

The status above is reproduced from GCL's immutable external-catalogue snapshot. It is not a new claim of current mathematical resolution, a source-semantic review, an independent proof check, or GCL certification. A published proof and a GCL replay are separate events.

The question in plain language

For sets of positive integers \(A\) and \(B\), write

\[ A+B=\{a+b:a\in A,\ b\in B\}. \]

When the limit exists, their natural density is

\[ d(A)=\lim_{N\to\infty}\frac{|A\cap\{1,\ldots,N\}|}{N}. \]

Erdős Problem 335 asks for a characterization of all pairs with \(d(A)>0\), \(d(B)>0\), and

\[ d(A+B)=d(A)+d(B). \]

The equality says that adding the two sets creates exactly the amount of large-scale coverage suggested by their separate densities. The problem is an inverse question: instead of estimating a sumset from information about \(A\) and \(B\), it asks what the sets must look like when the estimate is sharp. The quantifiers matter. This is a question about subsets of the positive integers, ordinary natural density, and the unrestricted pair of sets; it is not a question about finite sets, Schnirelmann density, or merely lower density.

A small example

A completely checkable equality comes from periodic residue classes. Let \(A\) contain precisely the positive integers congruent to \(0\) or \(3\) modulo \(6\), and let \(B\) contain precisely those congruent to \(0\) or \(1\) modulo \(6\). In one period the supports are

\[ P_A=\{0,3\},\qquad P_B=\{0,1\}. \]

Their residue sums are

\[ P_A+P_B=\{0+0,0+1,3+0,3+1\}\pmod 6=\{0,1,3,4\}. \]

Thus \(A\) and \(B\) have densities \(2/6=1/3\) each, while \(A+B\) is eventually, in fact periodically apart from the harmless positive-integer boundary, supported on four residue classes and has density \(4/6=2/3\). The exact identity is

\[ \frac{4}{6}=\frac{2}{6}+\frac{2}{6}. \]

This was verified by enumerating the four residue sums, with no floating-point approximation. It illustrates why finite cyclic groups and modular obstructions naturally enter the subject. It does not classify all equality cases: periodic examples are only one family, and a finite residue calculation says nothing about arbitrary nonperiodic sets.

What is known

The frozen catalogue entry at teorth/erdosproblems commit 536408716724a2402bdfa7c1d764bf83d64a8c89, whose data/problems.yaml entry has blob 4e60263a80af6207aca494d2e1388e44fa372889, records Problem 335 as open, unformalized, and last updated 2025-08-31. That snapshot is the programme's source lock; its status is not a proof that no later mathematics exists.

A major partial result is the primary paper by Ethan Ackelsberg and Florian K. Richter, An inverse theorem for sumsets of sets of positive density in the integers, arXiv:2604.12864v1 (submitted 14 April 2026). Their Theorem 1.4 assumes \(d(A)>0\), \(d(A)+d(B)<1\), that \(B\) meets every residue class in the sense that \(B\) intersects every set of the form \(a\mathbb N+b\), and equality along a sequence tending to infinity. It concludes that, after a finite-modulus decomposition, either the pair is described up to zero density by parallel Bohr intervals, or a second modular and translation-invariant alternative occurs. The paper explicitly identifies this as a partial answer to the Erdős–Graham question, not a solution of the unrestricted problem.

The maintained live Problem 335 page gives the same core question and an open status in the accessible indexed record. It also records the rotation or Bohr-interval construction and the obstruction from sets supported on even integers. The live site itself returned an access error during this audit, so live-page wording and comment history are reported as external context rather than silently promoted over the pinned catalogue record.

GCL's current state

The source-audit issue is MATH-PROGRAMME #1605, assignment ERDOS-CATALOGUE-0335-S01. Its required labels identify an available direct-editorial cooperative source task. The issue had zero comments and no existing RESULT/1 when checked, so there was no prior result to reuse or duplicate. Its packet classifies the existing GCL triage as STATEMENT_CAPTURE_REQUIRED and says that no admitted FormalConjectures statement is present in the FC-GDM-85F86371 provider snapshot.

Accordingly, GCL has no pinned Lean theorem for this problem, no replay receipt, and no mathematical certification to report. The 2026 paper is external proof evidence, not a GCL-verified proof. The original source-result citation is https://github.com/grandchallenge/MATH-PROGRAMME/issues/1605#issuecomment-6138462544; no future issue-comment URL is invented here. V2 ledger state alone determines compendium acceptance. Separately, the frozen catalogue records the mathematical question as open and its formalization status as unformalized.

What remains

The missing mathematical object is an unrestricted inverse theorem. Ackelsberg–Richter remove the difficulty when one summand meets every residue class, but Problem 335 permits both summands to have modular obstructions. A complete result would need to control equality after decomposing the sets into residue sections, including overlaps between different section-sums and sparse or exceptional sections whose translates can still affect a positive-density sumset.

A useful bounded successor would first capture the exact statement and conventions in a source-locked record, then verify a finite-modulus subclass: enumerate nonempty residue supports \(P,Q\subseteq\mathbb Z/q\mathbb Z\), retain those with \(|P+Q|=|P|+|Q|\), and prove the corresponding ultimately periodic density identity. That would be a genuine finite lemma and a regression test for notation, but not evidence for the arbitrary-set classification. Any broader claim should wait for specialist mathematical review of the primary theorem and the unrestricted modular case.

Sources and claim boundaries

The historical source is P. Erdős and R. L. Graham, Old and New Problems and Results in Combinatorial Number Theory, Monographie No. 28, 1980, p. 51; the publicly retrievable scan is the UCSD source PDF, SHA-256 0cbf0c32f0ab1e1c71db5121a88bac905bf976c4a6ab6bb6d7d9cf9ddd184ed3. The inspected primary modern source is Ackelsberg–Richter, arXiv:2604.12864v1, PDF SHA-256 c00f13c4f39029db1afd12995f2c92dd9fdf11ca0b9066359d815118abf8d1f6.

The catalogue commit and blob above pin the audited metadata; the live page and paper document later context. None of these materials supplies GCL mathematical certification. The article records a source audit and a verified finite example, not a solution, publication decision, novelty claim, or award claim.

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1605 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
  • GCL triage: STATEMENT_CAPTURE_REQUIRED.
  • Formal-statement coverage: No statement was confirmed in GCL's pinned formal-provider snapshot. Presence does not mean the statement or proof has been checked.
  • External mathematical status: consult current problem page and its dated bibliography.

Back to catalogue · Compendium introduction