Erdős Problem #293
Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: open (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 sum of reciprocals of positive integers. For a positive integer \(k\), consider every identity
The strict inequalities are part of the question: denominators must be pairwise distinct. Let \(D_k\) be the set of all positive integers that occur as one of the denominators in at least one identity with \(k\) summands. Erdős Problem #293 asks about
Thus the quantifiers are important. For each fixed \(k\), a number belongs to \(D_k\) if there exists one suitable decomposition containing it; \(v(k)\) is the first integer greater than \(1\) for which no such decomposition exists. The problem asks how \(v(k)\) grows as \(k\) grows. It does not ask whether one particular decomposition is unique, nor whether every denominator can occur in every decomposition.
The original Erdős–Graham scan explicitly includes the condition \(v(n)>1\). The maintained Erdős Problems entry states the modern version but flags the original source as ambiguous. This article uses the pairwise-distinct normalization above because it is the formulation in the inspected modern paper; the notation-to-source mapping remains a documented semantic exception, not a silently certified equivalence.
A small example
The denominator \(42\) occurs in a four-term decomposition:
The exact check has common denominator \(42\):
The denominators \(2,3,7,42\) are positive, strictly increasing, and distinct. Consequently \(42\in D_4\). This proves only membership in \(D_4\); it says nothing about whether \(42\) is the first missing denominator, and it cannot establish a value of \(v(4)\) by itself.
An independently reproducible exact-arithmetic check is:
from fractions import Fraction
denominators = (2, 3, 7, 42)
assert len(set(denominators)) == len(denominators)
assert all(a < b for a, b in zip(denominators, denominators[1:]))
assert sum((Fraction(1, n) for n in denominators), Fraction(0)) == 1
This uses integer rational arithmetic, not rounded decimal arithmetic. The authenticated source audit also records a much larger exact nine-term certificate, but the associated claim that a nearby integer is missing was not promoted from a reported computation to a theorem.
What is known
The historical source is Erdős and Graham, Old and New Problems and Results in Combinatorial Number Theory, printed p. 35. It records that results of Bleicher and Erdős imply a factorial-scale lower bound and speculates about substantially faster growth. The directly inspected Bleicher–Erdős paper, The Number of Distinct Subsums of \(\sum_{i=1}^{N}1/i\), Mathematics of Computation 29 (1975), 29–42, proves lower bounds for the number of distinct harmonic subsums. It is relevant primary evidence, but this manuscript does not reconstruct every step needed to turn that subsum theorem into the claimed bound for \(v(k)\).
The strongest directly checked modern theorem is Theorem 1.1 of Wouter van Doorn and Quanyu Tang, The smallest denominator not contained in a unit fraction decomposition of 1 with fixed length:
This is a genuine published-source theorem statement inspected in the authors’ arXiv text, with pairwise-distinct denominators and the same \(D_k\) interface. The paper also says that extracting the older \(v(k)\gg k!\) claim from the earlier literature is not straightforward. That creates a historical evidence conflict about the ease and status of the factorial implication; it does not change the present open status of #293.
For the other direction, the maintained entry records an elementary upper bound of the form
where \(c_0=1.26408\ldots\) is the Vardi constant. These lower and upper bounds leave a very wide gap. The maintained page also reports a conditional route to doubly exponential growth through a stronger estimate for the related rational-representation quantity \(N(b)\); that conditional suggestion is not a theorem about #293 as written.
GCL's current state
The authenticated original issue #1563 is the source-audit work item. Its exact RESULT/1 source comment is marked DISPOSITION: PARTIAL. It reports no blocker to reading the original scan, but preserves the semantic normalization issue, the historical factorial-evidence conflict, and the boundary that the larger computational claim was not replayed.
The fixed catalogue packet is teorth/erdosproblems@536408716724a2402bdfa7c1d764bf83d64a8c89, data/problems.yaml, Git blob 4e60263a80af6207aca494d2e1388e44fa372889, full-file SHA-256 be65b0db56fe13becec3178f2937685ee7e6a2f5df9325712474722c279c4182. Its #293 record says informal status open, formal status unformalized, statement formalized no, and last update 2025-08-31. The local fixed GCL registry retains issue-body SHA-256 8c8e4ca77ccf3979ba9659440c1b81e0930c29b3e96c364a7d91f021be872c30 and triage class STATEMENT_CAPTURE_REQUIRED.
The pinned FormalConjectures provider is google-deepmind/formal-conjectures@85f863718beeec7b58a3a1926ee92e3472bc2020. The audited provider snapshot contains no FormalConjectures/ErdosProblems/293.lean; the current tree was likewise reported to contain no such file. The source audit records snapshot hash values 33d1929f630431fe731cf0880cd01312cee3c7fa043ed5f1d4499fa28c0533fd and b9284333f076b0b4b702f08d977c8e0bdede815599e22750e0c70c5ee214c54f. This absence is an unavailable formal-source record, not evidence that the mathematics is false.
No pinned Lean file, external proof repository, self-reporting CI result, or catalogue label is treated as mathematical certification. GCL has issued no certification for this problem.
What remains
The status is open, with the caution that this is the reconciled documentary status, not a new mathematical adjudication. A useful bounded successor task is to reproduce exact rational certificates for a finite interval of candidate denominators, beginning with a specified fixed \(k\), while checking denominator distinctness, the identity sum, duplicate records, and interval coverage. A separate exact branch-and-bound search can test a reported small missing-value case. Each search must state its denominator bounds and termination rule.
The acceptance test must distinguish positive and negative evidence. Finding a valid decomposition proves membership in \(D_k\). Finding no decomposition proves non-membership only when the search is exhaustive under a stated finite bound and the bound is known to contain every relevant branch. A finite list of successful certificates never proves that the next integer is \(v(k)\) unless all smaller candidates have also been excluded.
The full reported search around a large value was not executed here: the audit did not replay its C search, checker, certificate release, or 2000-shard computation. The reported website comments are unverified discussion evidence. The smallest missing lemma for stronger exposition is therefore not merely a coding detail; it is a source-faithful, independently checkable argument that connects the complete bounded computation, its exact inputs, and the claimed absence statement.
Sources and claim boundaries
- Erdős Problem #293 — live statement, status context, citations, and the site’s ambiguity note. Its comments and computational reports are not primary certification.
- Erdős–Graham, Old and New Problems and Results in Combinatorial Number Theory, printed p. 35; inspected scan SHA-256
0cbf0c32f0ab1e1c71db5121a88bac905bf976c4a6ab6bb6d7d9cf9ddd184ed3. - Bleicher–Erdős, The Number of Distinct Subsums of \(\sum_{i=1}^{N}1/i\), pp. 29–42, DOI 10.1090/S0025-5718-1975-0366795-4; inspected PDF SHA-256
b9b8292315a6c0cd3533632e84a22b6669e563577fca2863e77867d1bc7545ed. - van Doorn–Tang, arXiv:2512.22083, Theorem 1.1; inspected PDF SHA-256
5b82e94df622f6b161ac7d1479e331621f44ef073e3eeb889a7e151c46ebb666. - Pinned catalogue file and fixed GCL registry — identity and status metadata, not proof authority.
- Pinned FormalConjectures tree and the authenticated source audit — absence and evidence boundaries, not certification.
The source return remains PARTIAL; no STATUS_CONFLICT was silently upgraded to resolution, and the formal-source absence remains an explicit blocker to formal-source claims. This manuscript itself is an unaccepted documentary draft and carries no authority to certify, publish, merge, or promote the problem’s mathematical status.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1563 — 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.