Erdős Problem #287
Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: falsifiable (not automatically the live result).
Subject areas: number theory, unit fractions.
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.
Editorial authority: The protected V2 admissions ledger alone determines publication status. This explanation is not independent mathematical certification.
The question in plain language
Take a finite list of distinct integers bigger than \(1\), written in increasing order,
and suppose their reciprocals add exactly to one:
Erdős Problem #287 asks whether every such list has a consecutive gap of at least \(3\); in symbols,
The words “every such list” are important. A gap is the difference between neighboring denominators. Since the denominators are strictly increasing, each gap is at least \(1\). The conjecture says that one cannot represent \(1\) by unit fractions while keeping every gap equal to either \(1\) or \(2\).
This is a finite, exact-arithmetic assertion. A counterexample would be one explicit tuple \((n_1,\ldots,n_k)\) whose reciprocal sum is exactly \(1\) and whose gaps are all at most \(2\). That is the sense in which the pinned catalogue labels the statement falsifiable. The label does not assert that the statement is false, undecidable, independent of a named theory, or already settled. The ambient objects here are ordinary positive integers and rational numbers; no stronger logical status is justified by the wording.
A small example
The tuple \((2,3,6)\) gives the exact identity
Its consecutive gaps are
so its largest gap is exactly \(3\). I independently checked this calculation with Python’s fractions.Fraction type: the computed sum was 1, the gap tuple was (1, 3), and the maximum was 3. This check uses exact rational arithmetic rather than floating-point rounding.
The example explains the sharpness language attached to the problem. If the universal lower bound \(3\) is true, it cannot be increased to \(4\), because this representation already attains maximum gap \(3\). But the example is not a proof of the universal claim: it exhibits one admissible representation, not all admissible representations. It also does not decide whether there is a different tuple whose gaps are all \(1\) or \(2\).
What is known
The primary historical source is P. Erdős and R. L. Graham, Old and New Problems and Results in Combinatorial Number Theory (1980), printed page 33. There they define the family of reciprocal representations of \(1\) by distinct denominators and ask whether the maximum neighboring gap must be at least \(3\). The same page records \((2,3,6)\) as the sharpness example and notes that the weaker lower bound \(2\) follows from the impossibility of a sum of reciprocals of consecutive integers equaling \(1\).
That weaker fact is supported by the primary paper P. Erdős, “Egy Kürschák-féle elemi számelméleti tétel általánosítása” (1932), printed page 6. In the special case \(d=1\), Erdős identifies the result with Kürschák’s theorem: a finite sum of reciprocals taken from one consecutive arithmetic progression cannot be an integer. A representation with every gap equal to \(1\) would be exactly such a consecutive block, so the theorem implies only
It does not rule out a mixture of gaps \(1\) and \(2\), which is precisely the extra difficulty in Problem #287.
The maintained problem page also records a conditional route involving primes: if every sufficiently large interval \([N,2N]\) contains a prime \(p\) for which \((p+1)/2\) is prime, then the gap of \(3\) conclusion would hold apart from finitely many exceptions. This is a conditional implication, not an unconditional proof of #287. The primary 1980 source itself presents the gap of \(3\) question as unresolved; the current source audit found no source-authoritative proof that removes that boundary.
GCL's current state
The exact original GCL issue is #1557, titled “Problem 287: special status and logical boundary audit.” Its pinned catalogue source is teorth/erdosproblems@536408716724a2402bdfa7c1d764bf83d64a8c89, file data/problems.yaml, entry blob 4e60263a80af6207aca494d2e1388e44fa372889, with source-registry SHA-256 be65b0db56fe13becec3178f2937685ee7e6a2f5df9325712474722c279c4182. The entry reports informal_status: falsifiable, formal_status: unformalized, and formalized: yes. The registry row binds the issue body with SHA-256 6224fa2593652e90b9faebcb293af5664821e061f3921814ed03abe452dc6981.
The authenticated source-audit return is the original issue’s exact RESULT/1 comment, posted by fyremael with ASSIGNMENT_ID: ERDOS-CATALOGUE-0287-S01 and DISPOSITION: PARTIAL. It reports that the source lock is substantially complete but that certification-level replay was not performed. That PARTIAL status must remain visible: it is not SOURCE_LOCK_READY, and it is not a proof disposition.
The pinned Formal Conjectures file is 287.lean at commit 85f863718beeec7b58a3a1926ee92e3472bc2020, blob 8df34cde0a0aef90bd69d1ff38b2be3327b2967f. Its main conjecture and the gap \(2\) variant have sorry bodies; its separate sharpness test proves the displayed example only. The audit also inspected external Lean and prime-chain repositories, but did not replay them. An external claimed Lean proof, author manuscript, theorem annotation, or successful CI run is not GCL mathematical certification. GCL has issued no such certification here.
What remains
The immediate mathematical task is statement capture: compare every formal binder with the primary wording, including \(k\geq2\), the condition \(1<n_1\), strict ordering, the exact reciprocal-sum type, and the index range of the maximum. A formal file’s presence proves only that a statement has been written down; it does not establish semantic equivalence with the historical question, and a sorry body is not a proof.
A bounded successor audit can then replay the exact external commits named in the RESULT/1 return and separately test the finite prime-chain verifier. Its acceptance hooks should distinguish three outcomes: a finite exclusion bound, a conditional implication from the prime hypothesis, and the infinite universal conjecture. They must not be merged. Any changed dependency, missing generated source, remaining sorry, failed axiom audit, or endpoint mismatch is a stopping condition for the corresponding claim.
The documentary stopping point is therefore precise: retain the weaker gap \(2\) theorem as established by the 1932 source, retain \((2,3,6)\) as the exact sharpness example, and leave the gap \(3\) assertion at the source-audit status PARTIAL until a source-faithful proof or a finite counterexample is independently adjudicated. Publication or admission of this article would be a separate editorial transaction and would not change that proof status.
Sources and claim boundaries
- Erdős and Graham, Old and New Problems and Results in Combinatorial Number Theory (1980), scan, printed p. 33. The source-audit record gives PDF SHA-256
0cbf0c32f0ab1e1c71db5121a88bac905bf976c4a6ab6bb6d7d9cf9ddd184ed3. No DOI is asserted for this monograph because none was confirmed in the cited record. - Erdős, “Egy Kürschák-féle elemi számelméleti tétel általánosítása” (1932), scan, printed p. 6 for the \(d=1\) special case. The source-audit record gives PDF SHA-256
1d08335e65da7cf40d1c4a7ba0088e0f39b9c5a4b2e42de95fc9ffa69fb96c7a; no DOI is asserted because none was confirmed in the cited record. - Maintained Erdős Problems entry #287, last edited 23 January 2026: statement, bibliography, the zero-proof-exposition indicator, and the conditional prime statement. It is a catalogue page, not GCL mathematical certification.
- Original GCL issue #1557, issue-body SHA-256
6224fa2593652e90b9faebcb293af5664821e061f3921814ed03abe452dc6981, and its authenticated source-auditRESULT/1. These are documentary evidence with dispositionPARTIAL, not a proof or admission record. - Pinned catalogue registry source, source-registry SHA-256
be65b0db56fe13becec3178f2937685ee7e6a2f5df9325712474722c279c4182, and entry blob4e60263a80af6207aca494d2e1388e44fa372889. - Pinned Formal Conjectures statement, commit
85f863718beeec7b58a3a1926ee92e3472bc2020, blob8df34cde0a0aef90bd69d1ff38b2be3327b2967f, and raw SHA-2569e90445fcbb14f334a23d1afbd7e46a69fdfd55794ecfa134e23754a21d8c402. Its formal statements and example check do not constitute GCL certification.
This is a source-qualified documentary explanation. It makes no claim of a proof, disproof, independence, novelty, publication, editorial admission, or successful protected CI result for Erdős #287.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1557 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
- GCL triage: OUTSIDE_STRICT_OPEN_TRIAGE.
- Formal-statement coverage: A provider statement was included in the pinned snapshot. Presence does not mean the statement or proof has been checked.
- External mathematical status: consult current problem page and its dated bibliography.