Erdős Problem #431
Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: open (not automatically the live result).
Subject areas: number theory, primes.
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
Could all sufficiently large prime numbers be expressed as sums from two fixed infinite sets, and could those sums avoid all sufficiently large composite numbers? Erdős Problem 431 asks whether there exist infinite sets \(A,B\subseteq\mathbb N\) whose sumset
agrees with the set of primes up to finitely many exceptions. This means that beyond some threshold every prime is in \(A+B\) and every integer in \(A+B\) is prime. The formulation permits a finite number of missing primes and unwanted composites, but no infinite collection of either.
The question is known as Ostmann's inverse Goldbach problem. Ordinary Goldbach-type questions ask whether each even integer can be represented by primes. Ostmann asks the opposite kind of structural question: can a set with the irregular pattern of the primes itself be decomposed into the complete sumset of two infinite component sets?
One must not confuse this with constructing infinitely many individual prime sums or a restricted triangular collection of prime sums. A full Cartesian sumset \(A+B\) includes every pair, not only selected pairs.
A small example
Take \(A=\{1,2\}\) and \(B=\{2,3\}\). Their sumset is
The element \(4\) is composite, so this finite example is not a set consisting only of primes. However, a finite defective example does not disprove Ostmann's conjecture, because the original question concerns infinite sets and explicitly tolerates finitely many exceptional values.
There is an elementary parity obstruction to any proposed full decomposition. Suppose \(A\) contains one even and one odd number. For any sufficiently large \(b\in B\), the two sums formed with those two elements have opposite parity. One is even and greater than \(2\), so it is composite. Since \(B\) is infinite, this gives infinitely many distinct composite members of \(A+B\), contradicting agreement with the primes outside a finite set. Thus all members of \(A\) must have one parity. The same argument applies to \(B\).
Since all sufficiently large primes are odd, the two fixed parity classes must be opposite: one summand set contains only even integers, the other only odd integers. This constraint is necessary, but far from sufficient. Removing parity obstructions does not control composite values arising from other small prime divisors.
What is known
The problem was proposed by Alexander Ostmann and discussed by Erdős in 1961 and 1977, in his 1980 survey, and with Graham in Old and New Problems and Results in Combinatorial Number Theory (1980), printed p. 85 of the historical monograph.
Christian Elsholtz and Adam Harper, Additive decompositions of sets with restricted prime factors, Transactions of the American Mathematical Society (2015), Theorem 2.6, proved strong restrictions on any hypothetical representation. If infinite \(A,B\) existed as described, each counting function would satisfy, for large \(x\),
and the same for \(B\). These are necessary size bounds, not a construction of the desired sets or a proof of impossibility.
A related theorem of Elsholtz (2001) rules out a decomposition into three nontrivial summand sets. Tao and Ziegler's 2023 Infinite partial sumsets in the primes constructs infinite sequences with certain triangular cross-sums prime, which does not imply that all pairs form primes. These distinctions are part of the established mathematical landscape.
The maintained Erdős Problems #431 page, last edited April 8, 2026, labels the full two-set problem OPEN. But there has been a material development after that page's last edit. On October 6, 2026, OpenAI published a mathematical research collection. Its overview, family 013 claims that no finite modification of the primes is a sumset \(A+B\) even when both summands merely contain at least two nonnegative integers. The linked manuscript is titled The additive indecomposability of the primes (dated September 24, 2026). If mathematically correct, this stronger theorem would answer Ostmann's original question negatively.
An announced proof and its independent verification are different evidence classes. The repository explicitly states that its manuscript families have different levels of formal verification and that some unverified results may contain errors. An inspected Ostmann comparator file defines both the strong inverse-Goldbach statement and the original infinite-summand impossibility statement, but the displayed proof bodies are by sorry. Other proof modules exist in the repository; that observation alone neither invalidates them nor certifies a complete Lean replay. Consequently the October announcement should be treated as a new externally claimed resolution requiring mathematical and formal-source review, not silently substituted for a verified theorem.
GCL's current state
The original GCL source issue #1701 pins teorth/erdosproblems at commit 536408716724a2402bdfa7c1d764bf83d64a8c89, data/problems.yaml blob 4e60263a80af6207aca494d2e1388e44fa372889. That frozen entry records OPEN, consistent with the maintained page's April position but predating the new October release.
The source issue records no 431.lean statement file in GCL's admitted FormalConjectures provider snapshot 85f863718beeec7b58a3a1926ee92e3472bc2020. Later upstream formal statements and the separate OpenAI proof library have independent provenance; they are not automatically part of this source lock. The precise remaining editorial issue is therefore a dated source-status conflict between an older maintained OPEN label and a newer publicly claimed negative theorem.
The authenticated source audit is original issue #1701, RESULT/1 comment #6157465402. It records the October claim as a claim, not as a certified mathematical resolution. V2 admission remains a separate protected editorial decision.
What remains
The next bounded task is a critical audit of the actual The additive indecomposability of the primes manuscript and its full proof dependencies. Establish whether the theorem really concerns every finite modification of the primes and all summand sets with at least two elements, without an unmentioned hypothesis. Check that such a statement implies the original positive-integer infinite-set version.
For any purported Lean formal verification, inspect the exact manuscript-to-theorem alignment, the full imports and axioms, the build toolchain and the terminal proof—not only an overview or comparator definition. In particular, the two theorem declarations in the comparator source presently contain sorry and cannot themselves serve as completed proof evidence. The bounded review succeeds with a fully checkable proof or complete exact-head Lean replay and a source-concordance certificate; it fails upon a mathematical gap, wrong exceptional-set convention, hidden unproved assumption or unsupported formal-verification claim. Stop after the critical source review rather than inferring admission from an announcement.
Sources and claim boundaries
Historical problem: Erdős–Graham (1980), p. 85. Earlier necessary conditions: Elsholtz–Harper (2015), Additive decompositions of sets with restricted prime factors, Theorem 2.6. Maintained pre-announcement status: Erdős Problem #431. New public research claim: OpenAI math overview, family 013 and associated comparator definitions. GCL custody: original issue #1701, frozen catalogue and authenticated source report.
The parity argument is an elementary verified obstruction. The 2015 counting restrictions are cited external theorems; the October 2026 stronger proof is an external claim awaiting independent verification. No original GCL theorem, certified Lean replay or protected editorial admission is asserted.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1701 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
- GCL triage: FORMAL_SOURCE_REFRESH_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.