Skip to content

Erdős Problem #292

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: Publication status comes exclusively from the protected V2 ledger. This explanation does not constitute mathematical certification.

The question in plain language

An Egyptian fraction is a finite sum of reciprocals of distinct positive integers. Erdős and Graham ask which integers can occur as the largest denominator in a representation of \(1\). In their printed p. 35, the question is phrased for an increasing list \(x_1<\cdots<x_k\) whose reciprocal sum is \(1\), and asks whether the largest entry assumes almost all integer values.

The source-faithful set is

\[ A=\left\{n\in\mathbb N:\exists S\subseteq\{1,\ldots,n\},\ n\in S,\ \sum_{m\in S}\frac{1}{m}=1\right\}. \]

The quantifiers matter. For a particular \(n\), membership in \(A\) asks for one finite witness \(S\): every denominator lies between \(1\) and \(n\), the denominators are distinct because \(S\) is a set, and \(n\) is present, hence is the largest denominator. The question is not whether every set of denominators works, nor whether one fixed representation works for infinitely many \(n\).

“Almost all” means natural density one:

\[ \lim_{x\to\infty}\frac{|A\cap[1,x]|}{x}=1. \]

Equivalently, the exceptional set \(B=\mathbb N\setminus A\) has density zero. This does not mean that every sufficiently large integer belongs to \(A\); a density-zero exceptional set may still be infinite. On the evidence inspected here, the mathematical problem is proved by a published theorem of Greg Martin. The GCL formal/documentary state remains a STATUS_CONFLICT, described below.

A small example

The integer \(6\) is in \(A\), witnessed by the exact identity

\[ \frac12+\frac13+\frac16 =\frac6{12}+\frac4{12}+\frac2{12}=1. \]

The largest denominator is \(6\), and the denominators are distinct. Splitting the term \(1/3\) is not needed here; another direct witness is \(12\):

\[ \frac12+\frac14+\frac16+\frac1{12} =\frac6{12}+\frac3{12}+\frac2{12}+\frac1{12}=1. \]

These calculations can be independently reproduced with exact rational arithmetic. The following finite search tests increasing tuples and never uses floating point:

from fractions import Fraction
from itertools import combinations

def first_by_max(k, bound):
    for largest in range(k, bound + 1):
        for earlier in combinations(range(1, largest), k - 1):
            ns = earlier + (largest,)
            if sum((Fraction(1, n) for n in ns), Fraction(0)) == 1:
                return ns
    return None

assert first_by_max(3, 20) == (2, 3, 6)
assert first_by_max(4, 100) == (2, 4, 6, 12)
print(first_by_max(3, 20), first_by_max(4, 100))

The output is (2, 3, 6) (2, 4, 6, 12). This verifies finite membership, and minimality for those fixed term counts within the stated search. It is not evidence for the density theorem. The same two identities and the finite-check requirement are recorded in the authenticated source audit RESULT/1.

What is known

Greg Martin’s “Denser Egyptian Fractions,” Acta Arithmetica 95 (2000), 231–260, DOI 10.4064/aa-95-3-231-260, directly addresses the Erdős–Graham question. I inspected the accessible arXiv paper, which identifies the journal reference and gives the definitions and theorem statements. In the introduction, Martin defines \(L_1(r)\) to be the integers that cannot be the largest denominator in an Egyptian-fraction representation of a positive rational \(r\). His Theorem 4, equation (6), says that \(L_1(r)\) has density zero and, for \(x\geq3\), has order of growth \(x\log\log x/\log x\), with constants depending on \(r\).

For \(r=1\), the complement of \(A\) differs from Martin’s \(L_1(1)\) only by the harmless initial convention at \(1\). Thus Martin’s theorem implies

\[ \lim_{x\to\infty}\frac{|A\cap[1,x]|}{x}=1. \]

The result is stronger than density one: the exceptional count is bounded above and below by constant multiples of \(x\log\log x/\log x\). It is not an asymptotic equivalence with a certified leading constant. Martin’s paper also discusses exceptional integers related to prime powers, but that discussion must not be inflated into a complete classification of \(B\).

This is a published external proof of the mathematical statement, not a proof newly established by this manuscript. The original Erdős–Graham scan is historical source evidence for the question; Martin’s theorem is the resolving primary result inspected for the status claim.

GCL's current state

The authenticated GitHub identity was verified as fyremael. Issue #1562 is open and its exact issue-body SHA-256 is 14b6c1e7cf910d75fe009a1b7de0b86d40a699badd812a3acda6b4b5377497fc. The pinned catalogue packet names teorth/erdosproblems commit 536408716724a2402bdfa7c1d764bf83d64a8c89, data/problems.yaml blob 4e60263a80af6207aca494d2e1388e44fa372889, and catalogue status proved. The fixed Programme registry entry in governance/erdos_catalogue/REGISTRY.json records the same problem as proved, with triage class OUTSIDE_STRICT_OPEN_TRIAGE and no protected FC file.

The exact authenticated source return is RESULT/1, disposition STATUS_CONFLICT. It reports that the maintained Erdős entry agrees with the normalized largest-denominator question and reports Martin’s result, while also reporting statement formalization without a proof exposition or proof claim. Those website fields are maintained-entry evidence, not certification. The return’s “zero comments” observation was a pre-posting readback; the linked comment is now the visible audit comment.

The admitted FormalConjectures snapshot google-deepmind/formal-conjectures@85f863718beeec7b58a3a1926ee92e3472bc2020 contains no FormalConjectures/ErdosProblems/292.lean. A later upstream file at commit 573d06b08947dbff577e453ab102fa2b39d8f759 has blob SHA afc210ed25485fe34e09ed2bbe48c04600c5ae31, but its relevant theorem declarations use by sorry. The external plby/lean-proofs source at commit 8822f7ddef30fadbd92e1c6ab4ed897af356af5e has blob SHA 414e803794220fe39d1952b05c17f6e75724098f and no visible sorry, but the audit records it as UNREPLAYED. A commit message, external proof repository, pinned Lean by sorry, or self-reporting CI is not mathematical certification.

What remains

The published mathematics is proved, but the GCL formalization boundary is not closed. The exact successor task recorded by the audit is to replay the pinned external source with its exact dependency graph, record the Lean version and #print axioms, and prove the semantic bridge between its largestDenominators definition and the finite-set definition of \(A\). The replay must also check the finite guards \(6\in A\) and \(12\in A\), including distinctness and the largest-denominator condition.

Acceptance would require an actual kernel replay of the exact source and a checked semantic bridge, with dependency identities recorded. Falsification hooks include a missing import, dependency drift, hidden sorry, a mismatch in density conventions, or failure of the finite-set equivalence. The task stops at that bounded source and semantics audit; it does not turn Martin’s analytic proof into a Lean proof.

There is no unavailable original statement to conceal: the 1980 scan was retrieved and visually checked at printed p. 35. The unavailable or missing-source exception concerns the admitted FC snapshot’s absent #292 file, not the historical problem. The semantic exception is the conflict among catalogue proved, separate statement-formalized metadata, absent pinned FC source, later by sorry source, and an un-replayed external artifact. Those facts require preserving STATUS_CONFLICT, not upgrading it to GCL certification.

Sources and claim boundaries

The article therefore makes one status claim only: Problem 292 is proved in the cited published literature, while its GCL formal/source record remains a documentary STATUS_CONFLICT. No claim is made of independent re-proof, Lean certification, novelty, publication authority, or mathematical certification by GCL.

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1562 — 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