Erdős Problem #265
Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: open (not automatically the live result).
Subject areas: irrationality.
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.
Documentary explanation. Editorial acceptance is determined by the protected V2 admissions ledger and the exact manuscript source hash; this text does not confer mathematical certification.
The question in plain language
Start with a strictly increasing sequence of positive integers
Assume that both reciprocal series converge and have rational sums:
The question is how quickly such a sequence can grow. The interesting scale is double exponential. Erdős conjectured that growth with \(a_n^{1/n}\to\infty\) should be possible, but that the critical endpoint should satisfy
The latter says that \(\log a_n=o(2^n)\): the sequence may still grow much faster than exponentially, but not at a persistent rate comparable with \(c^{2^n}\) for any fixed \(c>1\).
There is an important normalization warning. The 1980 source prints \(1+a_n\) in the second denominator, while the 1988 source's prose and the current page use \(a_n-1\). A constant shift of every term converts one problem class into the other, but it changes which displayed example represents the sequence. The sign discrepancy is therefore a genuine source conflict, not a harmless reason to quote the sources as though their formulas were identical.
A small example
For the current minus-one normalization, take the Cantor sequence
The first sum is \(1\), not \(2\). Indeed,
so the partial sums telescope and
For the shifted denominator,
Three terms survive in the second telescoping sum, giving
Thus both infinite sums are rational, namely \(1\) and \(11/9\). The example grows quadratically, so it demonstrates simultaneous rationality without answering the growth question. An exact finite check is reproducible with:
from fractions import Fraction
for N in (3, 4, 10, 100):
s = sum((Fraction(1, k*(k-1)//2) for k in range(3, N+1)), Fraction())
t = sum((Fraction(1, k*(k-1)//2-1) for k in range(3, N+1)), Fraction())
print(N, s, t)
This returns \((1/3,1/2)\) at \(N=3\), \((1/2,7/10)\) at \(N=4\), \((4/5,1516/1485)\) at \(N=10\), and \((49/50,1803151/1499850)\) at \(N=100\).
What is known
Erdős and Graham recorded the problem in Old and New Problems and Results in Combinatorial Number Theory (1980, p.64), with the plus-one denominator. Erdős's 1985 note on Straus (p.334) says that they could not decide whether the sequence could grow exponentially or faster. In the 1988 paper “On the irrationality of certain series: problems and results” (printed p.104), the prose uses \(1/(n_k-1)\) but the source packet records a displayed Cantor example with \(\binom{k}{2}+1\). Kovač and Tao explicitly choose the minus-one formulation and display the Cantor witness with \(\binom{k}{2}-1\).
The strongest published interface identified in the source audit is Kovač–Tao, On several irrationality problems for Ahmes series. In §2.2.1 (pp.8–9), Theorem 2.8 and Corollary 2.9 show, for every finite number \(d\) of consecutive shifted reciprocal series, that there is a strictly increasing integer sequence for which all \(d\) sums are rational and
for some \(\beta>1\). Their proof permits, when \(d=2\), any \(1<\beta<\sqrt{6/5}\) through equation (7.9). After a constant shift, this applies to the minus-one pair as well. It proves that simultaneous rationality can coexist with genuinely double-exponential growth at some base.
The proof interface matters. Kovač and Tao obtain a nonempty interior for a set of possible sum vectors in dimension \(d\), then use the density of \(\mathbb{Q}^d\) to obtain rational targets. This is an existence argument, not an explicit formula for the rapidly growing sequence. It also does not decide whether the critical base-two quantity can stay above \(1\).
The same paper cites the classical one-series obstruction that \(a_k^{1/2^k}\to\infty\) forces \(\sum 1/a_k\) to be irrational. That rules out one extreme, but it does not rule out \(\limsup a_k^{1/2^k}>1\), nor does it prove the conjectured limit \(1\).
GCL's current state
The pinned catalogue snapshot at teorth/erdosproblems commit 536408716724a2402bdfa7c1d764bf83d64a8c89 marks 265 open, unformalized, and ambiguously stated; its entry is blob 4e60263a80af6207aca494d2e1388e44fa372889 and the packet hash is be65b0db5bf94a072de7e094e6d42864ba2d1f639e7cb951c02b4cf1d3d52b57. The authenticated latest source audit on original issue #1535, RESULT/1 therefore correctly returns DISPOSITION: STATUS_CONFLICT. The live page was recorded as OPEN on 2026-10-13 UTC, while also displaying an unreviewed public proof claim; that is evidence drift, not a solved-status readback.
That public claim points to KitaKen1/erdos-265-lean at commit 16db9f29ad68a861552f74b8c2fe2148433c3e80. Its declaration of erdos265_negative_answer denies an admissible sequence with \(c^{2^n}\leq a_n\) infinitely often for some \(c>1\); the statement file spells out strict monotonicity, \(a_0\geq2\), real summability, and rational values of both sums. This is a formal-provider claim separated from accepted mathematics: the audit found no obvious sorry or top-level axiom, but the attempted local build stopped at Read-only file system (os error 30). No GCL mathematical certification has been issued.
The neighboring Problem 260 and its reported 2026 arXiv preprint are outside this article; any such preprint remains an external claim requiring mathematical review and is not imported here as proof.
What remains
First, an editorially authoritative statement must resolve the plus-one/minus-one source history while recording the harmless shift equivalence and preserving the correct witness for the selected normalization. The catalogue's “open” label remains the official status in the pinned evidence, and the current page's public Lean claim cannot silently replace it.
Second, the exact Lean commit needs a clean kernel build in a writable, pinned Lean/Mathlib environment, followed by #print axioms and manual semantic review. The crucial checks are the translation from one-based source indices to a 0, the strength of the real summability hypotheses, the eventual quadratic recurrence for the tail envelope, the regularization argument, and the final transfer to \(\log(a_n)/2^n\to0\). A theorem can be syntactically complete yet address a stronger or differently normalized problem.
Until those checks and mathematical review are complete, the accepted conclusion is only that the problem is open with a documented status conflict. The published Kovač–Tao construction is a genuine positive growth result, the Cantor calculation is a verified example, and the Lean endpoint is an unresolved external claim. None of these should be described as a completed solution to Problem 265.
Sources and claim boundaries
- Maintained catalogue packet: Problem 265 at the pinned
problems.yamlcommit, and the live problem page. - P. Erdős and R. L. Graham, Old and New Problems and Results in Combinatorial Number Theory, Monographies de l'Enseignement Mathématique 28 (1980), p.64, original scan.
- P. Erdős, “E. Straus (1921–1983),” Rocky Mountain Journal of Mathematics 15 (1985), p.334, original scan.
- P. Erdős, “On the irrationality of certain series: problems and results,” in New Advances in Transcendence Theory (1988), pp.102–109, especially p.104, Cambridge DOI record and author scan.
- V. Kovač and T. Tao, “On several irrationality problems for Ahmes series,” Acta Mathematica Hungarica 175 (2025), 572–608; arXiv record and exact source, especially §2.2.1, pp.8–9, Theorem 2.8, Corollary 2.9, and equation (7.9).
- GCL evidence boundary: the authenticated original issue #1535 and its exact latest RESULT/1 comment. These are source-audit records, not mathematical certification.
The article reports the cited claims and the exact rational calculation but does not certify the external Lean repository, settle the primary-source sign conflict, promote the live page's proof claim, or alter the official open status. Its editorial acceptance is governed by the protected V2 ledger, separately from all mathematical certification controls.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1535 — 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.