Skip to content

Erdős Problem #315

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.

The question in plain language

An Egyptian fraction writes a number as a sum of reciprocals of positive integers. Problem 315 starts with the greedy decomposition of \(1\), whose denominators are

\[ 2,3,7,43,1807,3263443,…. \]

The shifted sequence in the live statement is \(u_1=1\) and \(u_{n+1}=u_n(u_n+1)\), so the actual denominators are \(u_n+1\). Its growth is roughly doubly exponential, and the associated scale is the Vardi constant

\[ c_0=lim_{n oinfty}u_n^{1/2^n}=1.264085…. \]

The question asks whether every other strictly increasing sequence of positive integers \(a_1<a_2<cdots\) satisfying

\[ \sum_{k=1}^{\infty}\frac{1}{a_k}=1 \]

must have

\[ \liminf_{n\to\infty}a_n^{1/2^n}<c_0. \]

“Other” means that the sequence is not exactly the Sylvester denominator sequence. The strict inequality and the quantifiers matter: this is not merely a comparison of a few initial denominators, and it does not ask for a uniform numerical gap below \(c_0\).

A small example

The recurrence gives \(u_1=1\), \(u_2=2\), \(u_3=6\), \(u_4=42\), and \(u_5=1806\). Direct exact-arithmetic verification gives

\[ \frac12=\frac12, \]
\[ \frac12+\frac13=\frac56, \]
\[ \frac12+\frac13+\frac17=\frac{41}{42}, \]
\[ \frac12+\frac13+\frac17+\frac1{43}=\frac{1805}{1806}. \]

Thus, for each displayed partial sum, the residual is respectively \(1/u_2\), \(1/u_3\), \(1/u_4\), and \(1/u_5\). Equivalently, the checked identity is

\[ \sum_{k=1}^{n}\frac{1}{u_k+1}+\frac{1}{u_{n+1}}=1 \]

for \(n=1,2,3,4\). This is a finite telescoping check: from \(u_{k+1}=u_k(u_k+1)\), one has \(1/(u_k+1)=1/u_k-1/u_{k+1}\). It explains why the Sylvester denominators sum to \(1\), but it does not establish the asymptotic liminf comparison for an arbitrary sequence.

What is known

The live Erdős Problems entry for #315 gives the strict-increasing formulation, says that it is true, and cites Yuhi Kamio and Zheng Li with Quanyu Tang. It also records the earlier indexing convention and identifies \(c_0\) as the Vardi constant.

The principal primary source checked here is Li and Tang, On a conjecture of Erdős and Graham about the Sylvester's sequence, arXiv:2503.12277v4 (21 March 2025), published in Acta Mathematica Hungarica 177 (2025), 41–63, DOI 10.1007/s10474-025-01566-8. Their Conjecture 1.3, p. 3, matches the strict sequence and the liminf inequality. Theorem 1.5 and Theorem 1.6, pp. 3–4, establish the two constructive ingredients; Corollary 1.7, p. 4, states the conclusion, and the constructive proof is developed in Sections 3.1–3.3, pp. 13–19. The paper presents this as a proof of the original conjecture.

Kamio’s independent Asymptotic Analysis of Infinite Decompositions of a Unit Fraction into Unit Fractions, arXiv:2503.02317v1 (4 March 2025), states a weakly increasing version in Problem 1, p. 1, and proves the generalized result as Theorem 8, pp. 3–4. Since strict increase implies weak increase, this is relevant evidence, but the two papers should not be treated as byte-identical statements. Both correct the apparent indexing/recurrence error inherited from the 1980 source Erdős and Graham, Old and New Problems and Results in Combinatorial Number Theory, p. 41.

GCL's current state

The assignment issue is MATH-PROGRAMME #1585, assignment ERDOS-CATALOGUE-0315-S01. Its pinned catalogue source is teorth/erdosproblems commit 536408716724a2402bdfa7c1d764bf83d64a8c89, data/problems.yaml, entry blob 4e60263a80af6207aca494d2e1388e44fa372889. At that snapshot the informal status is proved, last updated 2026-01-31; formal status is Lean; and statement-formalized metadata is yes. Those are catalogue facts, not GCL mathematical certification.

The issue packet says that no #315 statement file was present in the admitted FC-GDM-85F86371 provider snapshot. An exact upstream file is nevertheless available at FormalConjectures commit 0cfdea64bb89881ce06f41c303e2009ec6c66ea5, blob e37c941524417f0312bd2480dc374494d41b75bc. It encodes positivity, StrictMono, equality of the infinite reciprocal sum to \(1\), and deviation from the Sylvester sequence, but its theorem ends with by sorry. Its linked hosted proof is a separate source at plby commit 1d7b3f00780b85ed0462e79a1cd5650ee9055655; no Lean replay was performed here.

At worker preflight the original issue had no source return. The original SOURCE RESULT/1 citation is therefore authenticated original RESULT/1; no future comment link is asserted here. V2 ledger state, not this article or catalogue metadata, decides accepted coverage.

What remains

The external proof provenance is strong enough for a source-locked documentary account, but the mathematical certification boundary remains separate. A useful bounded successor would compare the live strict statement, Li–Tang’s Conjecture 1.3, Kamio’s weak statement, and the FC theorem line by line, including the index shift from \(u_1=1\) to Sylvester denominators beginning at \(2\). It should also record whether each hosted formal proof uses StrictMono or only Monotone and whether that difference is an implication or an unintended statement change.

That comparison should not silently promote the broader generalization in Li–Tang, whose non-constructive route assumes an unproved claim about eventually greedy best Egyptian underapproximations. Nor should the finite telescoping calculation be presented as evidence for the limiting theorem. A source mismatch, an unverified dependency, or failure to align the sum and indexing conventions is a valid stopping point.

The most useful output of that successor is a reproducible normalization record: each source should name the same objects, quantify over the same class of sequences, and expose every assumption needed by its limiting expression.

Sources and claim boundaries

The checked Li–Tang v4 PDF has SHA-256 1db323d34b3ca4039c3d805e812e1aa181d3229e6c12ab4594472a71dca85f6a; the Kamio v1 PDF has SHA-256 13a7d126c043f2f2e73fbd74242d5c3416ea5aa3f2877cac21688ed16f110739. These hashes bind the primary documents actually read. The catalogue commit and Git blob, the exact FC commit and blob, and the original issue URL above bind the documentary trail.

This article reports an external published/preprint proof claim, a live catalogue status, a sorry-containing formal statement, and an independently checked finite identity. It does not certify either paper, replay Lean, decide the V2 ledger, or supply GCL mathematical certification. The original source-result citation remains authenticated original RESULT/1.

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1585 — 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