Erdős Problem #297
Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: solved (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 ledger determines editorial acceptance and publication status. This documentary explanation does not constitute mathematical certification.
The question in plain language
For a positive integer \(N\), look at every subset \(A\) of the first \(N\) positive integers and keep the ones whose distinct reciprocal sum is exactly \(1\). Write their number as
The denominators are distinct because \(A\) is a set, and the equality is an exact rational equality. The question is about the growth of the sequence \((a_N)\) as \(N\) tends to infinity. It is not asking whether one particular finite search finds a representation of \(1\).
The sharp asymptotic reported in the inspected literature is
In this notation, \(o(1)\) means a real error term depending on \(N\) that tends to zero. Equivalently, for every \(\varepsilon>0\), all sufficiently large \(N\) satisfy
Thus the literature status is proved or resolved in this asymptotic sense. It does not give an exact formula for every finite \(N\), nor an explicit error bound in the displayed statement. The maintained Erdős Problems entry for #297 and the authenticated source audit identify Erdős and Graham’s 1980 book, p. 36, as the original reference. The audit did not access a scan of that page, so the historical pointer is retained without pretending that the original wording was independently reread here.
A small example
At \(N=6\), the two solutions are
because
The following exhaustive check uses exact Fraction arithmetic. It examines all \(2^6=64\) subsets, so it does not rely on decimal rounding.
from fractions import Fraction
solutions = []
for mask in range(1 << 6):
A = tuple(n for n in range(1, 7) if mask & (1 << (n - 1)))
total = sum((Fraction(1, n) for n in A), Fraction(0))
if total == 1:
solutions.append(A)
print(solutions)
print(len(solutions))
The output is
[(1,), (2, 3, 6)]
2
This is an independently reproducible finite fact: \(a_6=2\). The same exact calculation is recorded in the authenticated RESULT/1 source audit. It is useful for checking the statement and notation, but a finite value cannot prove an asymptotic formula.
What is known
Stefan Steinerberger’s arXiv:2403.17041v5 gives an earlier exponential upper bound. Its theorem, on PDF p. 1, proves that for sufficiently large \(N\),
This is an upper bound for a larger family, since “at most \(1\)” includes subsets whose sum is less than \(1\). It therefore bounds \(a_N\), but it is not the sharp answer to #297.
Yang P. Liu and Mehtaab Sawhney’s arXiv:2404.07113v1 states the sharp rate in Theorem 1.2 on PDF p. 1. They define the unique positive \(\lambda^\ast\) satisfying
and then set
Their conclusion is
with \(\gamma^\ast\approx0.631573\) and \(e^{\gamma^\ast}\approx1.88057\).
Conlon, Fox, He, Mubayi, Pham, Suk, and Verstraëte give the more general Theorem 1 in arXiv:2404.16016v2, published as Discrete Analysis 2025:28. For each fixed rational \(x>0\), the number of subsets with reciprocal sum \(x\) is
where
and \(\lambda\) is the unique real number satisfying
For \(x=1\), they report \(c_1\approx0.91117\). These primary papers support “proved asymptotically”; this article does not claim to have independently reconstructed their analytic or combinatorial proofs.
GCL's current state
The authenticated GitHub CLI identity used for the audit is the active account fyremael. Original MATH-PROGRAMME issue #1567 is open and binds Problem 297 to the source-audit assignment ERDOS-CATALOGUE-0297-S01. Its exact authenticated return is RESULT/1, with disposition SOURCE_LOCK_READY. That return records the issue-body SHA-256 f893eafa722aea60645081816e38235cc677f6cbe70f42502fb2d0b12a450b74 and preserves the source and formalization boundaries.
The fixed catalogue packet is teorth/erdosproblems commit 536408716724a2402bdfa7c1d764bf83d64a8c89, file data/problems.yaml, Git blob 4e60263a80af6207aca494d2e1388e44fa372889, and source SHA-256 be65b0db56fe13becec3178f2937685ee7e6a2f5df9325712474722c279c4182. Its metadata calls the problem solved, with formal status unformalized and statement metadata formalized: no. Those are catalogue fields, not GCL certification.
At the fixed Formal Conjectures commit 85f863718beeec7b58a3a1926ee92e3472bc2020, the recursive tree contains neighboring files #296 (blob a043a9185efc8703ed8325d730f412a3e8f3b079) and #298 (blob d5c9653d076dd3c493260fed888a0586fcd65a04), but no FormalConjectures/ErdosProblems/297.lean. There is consequently no #297 file hash to report. A missing pinned formal statement, an external proof repository, Lean code containing sorry, or self-reporting CI cannot be treated as mathematical certification. The protected V2 ledger, separately, determines editorial acceptance and publication status.
What remains
The mathematical resolution reported by the primary papers is asymptotic, not finite and explicit. A reader wanting exact values of \(a_N\) for particular moderate \(N\) needs a separate finite computation; the theorem does not provide those values or a usable error term for finite \(N\). A reader wanting formal verification needs first to state the exact predicate for finite subsets and then formalize the asymptotic quantifiers, including the meaning of the constant \(c_1\) and the \(o(1)\) term.
The main documentary task is provenance reconciliation: preserve the unavailable 1980 page as an unavailable source, retain the inspected theorem locations and version identifiers, and avoid upgrading catalogue metadata into a formal result. The two sharp papers use different logarithm bases and constants, but their statements describe the same exponential rate after the conversion \(c_1=\gamma^\ast/\log 2\); this comparison is a normalization, not a new proof.
No claim is made here that the Programme has replayed either primary proof, compiled a complete formal development, or certified the asymptotic result. The finite certificate above validates only the \(N=6\) instance.
Sources and claim boundaries
- Maintained Erdős #297 entry — live statement and current catalogue context; access and status are reported with the source audit rather than treated as GCL authority.
- Erdős and Graham, Old and New Problems and Results in Combinatorial Number Theory (1980), p. 36 — original bibliographic pointer. The scan was unavailable to the authenticated source audit, so no page-level quotation is claimed.
- Steinerberger, On a problem involving unit fractions, arXiv:2403.17041v5, theorem on PDF p. 1, DOI 10.48550/arXiv.2403.17041 — inspected upper bound for sums at most \(1\).
- Liu–Sawhney, On further questions regarding unit fractions, arXiv:2404.07113v1, Theorem 1.2 on PDF p. 1, DOI 10.48550/arXiv.2404.07113 — inspected sharp asymptotic in natural-exponential form.
- Conlon–Fox–He–Mubayi–Pham–Suk–Verstraëte, A question of Erdős and Graham on Egyptian fractions, arXiv:2404.16016v2, Theorem 1 on PDF p. 1, DOI 10.48550/arXiv.2404.16016 — inspected general rational \(x\) asymptotic and \(c_1\) value.
- Original issue #1567 and its exact authenticated RESULT/1 comment — source lock, finite check, catalogue/formalization hashes, and the
SOURCE_LOCK_READYboundary. - Pinned catalogue source and pinned Formal Conjectures tree — fixed provenance snapshots, not mathematical certification.
The status “proved asymptotically” belongs to the inspected primary literature and catalogue context. It is not a GCL proof certificate, editorial admission, protected merge, or claim of independent verification.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1567 — 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.