Skip to content

Erdős Problem #286

Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: proved (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 GCL V2 admissions ledger alone determines publication status. This source-bounded explanation is not mathematical certification.

The question in plain language

An Egyptian-fraction representation of \(1\) writes one as a sum of distinct unit fractions, meaning reciprocals of positive integers. Erdős Problem #286 asks how tightly such a representation can be packed when it has exactly \(k\) terms. If

\[ 1=\frac{1}{n_1}+\frac{1}{n_2}+\cdots+\frac{1}{n_k}, \]

with \(n_1<\cdots<n_k\), its width is the distance from the smallest to the largest denominator, \(n_k-n_1\). The original Erdős–Graham source asks, in the notation of its set of representations with \(n\) terms, whether the smallest possible width is

\[ (e-1)n+o(n) \]

as \(n\) tends to infinity. Here \(o(n)\) means an error whose ratio to \(n\) tends to zero; it is an asymptotic assertion, not an exact formula for small values.

The modern wording packages the same-looking idea as an interval question: for every sufficiently large \(k\), can one find an interval of width \((e-1+o(1))k\) containing \(k\) distinct positive integers whose reciprocals sum to \(1\)? The quantifier matters. “For every sufficiently large \(k\)” is stronger than “for infinitely many \(k\),” and a change in the coefficient or in what the parameter measures changes the problem.

A small example

For \(k=3\), take the distinct denominators \(2\), \(3\), and \(6\). Direct arithmetic gives

\[ \frac12+\frac13+\frac16=\frac36+\frac26+\frac16=1. \]

They lie in the interval \([2,6]\), whose width is \(6-2=4\). A finite exact-fraction enumeration of triples drawn from \(\{2,3,\ldots,200\}\) also returned width \(4\) with the triple \((2,3,6)\). This is a checked illustration of the definition, not evidence for an asymptotic coefficient. In particular, comparing \(4\) with \((e-1)\cdot3\) would say nothing reliable about the limit as \(k\) grows.

The example also shows why “distinct” must be stated. Repeating a denominator would turn the problem into a different additive question, while the source definition uses a set of distinct denominators. Likewise, an interval may contain many integers, but the assertion only asks that the selected \(k\) denominators be distinct members of it.

What is known

The pinned source audit checked P. Erdős and R. L. Graham, Old and new problems and results in combinatorial number theory (1980), printed p. 33. That source defines the family of distinct-denominator representations and records the minimum-width expression with coefficient \(e-1\). The scan is the UCSD copy with SHA-256 0cbf0c32f0ab1e1c71db5121a88bac905bf976c4a6ab6bb6d7d9cf9ddd184ed3.

Croot’s primary paper, E. S. Croot III, “On unit fractions with denominators in short intervals,” Acta Arithmetica 99 (2001), 99–114, defines the same kind of set on p. 99. Its introduction states two Erdős–Graham questions: one about the largest possible smallest denominator, and one about minimum width asymptotic to \(k\); it also says that the earlier questions were misstated. Those are not merely typographical details: the paper’s displayed parameterization is materially different from the printed Erdős–Graham coefficient recorded above.

Croot’s Main Theorem, pp. 100–101, says that for every fixed rational \(r>0\) and every \(N>1\), there are distinct integers \(N<x_1<\cdots<x_k\) whose reciprocal sum is \(r\) and whose largest denominator is at most

\[ \left(e^r+O_r\left(\frac{\log\log N}{\log N}\right)\right)N. \]

The paper proves this using Propositions 1 and 2 on pp. 103–104, with the construction completed on p. 105. It explicitly describes the result as solving the Erdős–Graham questions for infinitely many \(k\). That is a genuine published theorem, but it does not by itself establish the current formulation for all sufficiently large \(k\), and the source audit did not resolve whether its corrected width statement is semantically identical to the original \((e-1)k\) statement.

GCL's current state

The pinned catalogue entry at teorth/erdosproblems commit 536408716724a2402bdfa7c1d764bf83d64a8c89, file data/problems.yaml (blob 4e60263a80af6207aca494d2e1388e44fa372889, source SHA-256 be65b0db56fe13becec3178f2937685ee7e6a2f5df9325712474722c279c4182), labels #286 as proved, last updated 2025-08-31. The same entry says the statement is formalized but the formal status is unformalized. The registry row binds it to original issue #1556 by issue-body SHA-256 3703d7d69446e8d47a03f75972a224f4bfb51da687838f0ef07ac413bfa30e16 and classifies it OUTSIDE_STRICT_OPEN_TRIAGE. These are provenance and snapshot fields, not GCL certification.

The live Erdős Problems page for #286 says that the answer is yes and attributes it to Croot. That live summary is important evidence to reconcile, but it cannot by itself decide whether Croot’s theorem matches the original coefficient and quantifier.

The authenticated original issue #1556 contains the exact source packet and the sole SOURCE RESULT/1. That return gives the disposition EXACT_BLOCKER: it preserves the Erdős–Graham \((e-1)n+o(n)\) wording, Croot’s different formulation with width measured against \(k\), and the quantifier mismatch as an unresolved source-semantic conflict.

The current upstream FormalConjectures file at commit 817d8a74e6bcd505ac8407388cc12ac3cfda3d0d, blob 3f114a23e1c4db2fcf45afe80a27c89956a8847e, encodes a proposition for all sufficiently large \(k\): there exists a real sequence \(o(k)\) tending to zero, and for each such natural \(k\) there are real endpoints and a finite set \(S\) of cardinality \(k\), with \(0\) excluded from \(S\), exact reciprocal sum \(1\), and every element of \(S\) inside the interval. Its theorem body ends in by sorry. The admitted GCL provider snapshot contains no #286 statement file. A formal statement, especially one containing sorry, is not a Lean proof.

What remains

The first remaining task is source adjudication, not a new slogan about the problem being solved. A careful comparison must align Erdős–Graham’s \(n\), Croot’s \(k\) and \(N\), the smallest and largest denominators, and the definition of interval width. It must decide whether Croot’s “misstated” correction is a correction of the coefficient, of the parameterization, or of both. Until that comparison is settled from the primary text, the catalogue’s proved label cannot be used to erase the exact blocker.

The second task is a bounded formalization review. If the intended target is the original theorem, the formal statement must be shown to match its source-level quantifiers and coefficient. If the intended target is Croot’s corrected result, it should be encoded as that result, with the “infinitely many \(k\)” consequence kept distinct from a claim for all sufficiently large \(k\). Either route needs a checked semantic bridge before it can support a proof-status statement.

The published Croot theorem remains valuable evidence and may support a reusable lemma about representing a fixed rational in a short multiplicative interval. It is not, without the missing comparison, a certificate for the exact registered #286 statement. The stopping condition for this documentary pass is therefore precise: resolve the coefficient and quantifier comparison from primary sources, or retain EXACT_BLOCKER and identify the missing source argument.

Sources and claim boundaries

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1556 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
  • GCL triage: OUTSIDE_STRICT_OPEN_TRIAGE.
  • 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