Erdős Problem #270
Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: disproved (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.
The question in plain language
Let \(\mathbb N\) mean the positive integers. Given a function \(f:\mathbb N\to\mathbb N\), form
Erdős and Graham asked whether the condition
forces \(S(f)\) to be irrational. Read as a proposition, the question asks whether, for every positive-integer-valued \(f\) whose values eventually exceed every fixed bound, one has \(S(f)\notin\mathbb Q\). There is no monotonicity hypothesis in this original quantifier. The condition \(f(n)\to\infty\) permits the values to go up and down, provided that after some point they stay above each prescribed threshold.
The last point matters because the source also mentions a different question: what if \(f\) is nondecreasing? That is a stronger restriction and a different problem. A result for the unrestricted class does not automatically produce a result for the nondecreasing subclass. The original wording appears on printed page 66 of Erdős and Graham’s 1980 problem book; the scanned page is source-pinned below.
A small example
The series’ indexing is easy to shift accidentally, so here is an exact finite check of the denominator pattern. For integers \(N\geq 1\) and \(m\geq 1\), telescoping gives
Indeed,
Taking \(N=2\) and \(m=3\) therefore gives an exact value of \(1/40\). If the sum stops at \(L=10000\), the omitted tail is exactly
This can be reproduced without floating point:
from fractions import Fraction
partial = sum(
(Fraction(1, (n + 1) * (n + 2) * (n + 3)) for n in range(3, 10001)),
Fraction(0),
)
assert Fraction(1, 40) - partial == Fraction(1, 200100012)
The same calculation is recorded in the authenticated source-audit return RESULT/1. It checks a finite identity only; it is not itself a counterexample, because the displayed finite choice does not define a function on all \(n\) with \(f(n)\to\infty\).
What is known
The answer to the original unrestricted question is disproved, and in a particularly strong way. Crmarić and Kovač prove in Theorem 1 of their paper, printed arXiv PDF page 2, that
Thus for every prescribed \(\alpha>0\) there exists an admissible \(f\) with \(S(f)=\alpha\). Choosing the rational value \(\alpha=1\) supplies an \(f\) for which the sum is rational, so the universal irrationality claim is false. This is an existence conclusion: the theorem does not say that every such \(f\) has rational sum, nor does it give a simple closed formula for the witness \(f\) at \(\alpha=1\).
Their proof partitions the positive integers into the dyadic layers
and builds finite families of local choices on each layer. The generalized subsum principle is Lemma 4, printed PDF pages 4–6; the construction applying it is in Section 3, pages 6–9. The choices are arranged so that the layer sums fill intervals and the assembled function still satisfies \(f(n)\to\infty\). No monotonicity is imposed.
The same paper’s Theorem 2, printed PDF page 3, concerns the nondecreasing subclass. It says that the set of sums arising from increasing \(f\) has Lebesgue measure zero and empty interior. That statement does not identify whether this smaller set contains a rational number. Accordingly, the increasing version remains open on the evidence inspected here. The maintained Erdős Problems entry reports exactly this distinction: the unrestricted question has a strong negative answer, while the nondecreasing question may still have an affirmative irrationality answer.
GCL's current state
The authenticated original issue is #1540, and the exact authenticated source audit is comment 6138000286. Its disposition is PARTIAL, not a mathematical certificate. It records the original statement, the primary theorem, the nonmonotone/monotone distinction, and the formal-provenance limits. The authenticated GitHub identity used for that return is fyremael.
The fixed Programme registry entry in governance/erdos_catalogue/REGISTRY.json binds problem 270 to issue 1540, issue-body SHA-256 847f1cbdac784e89b73d4dc11dc893918179b9c0ea193ce77bd398859a82aa03, catalogue status disproved, and triage class OUTSIDE_STRICT_OPEN_TRIAGE. That registry identity is routing and source-tracking evidence, not an independent proof.
The pinned catalogue packet is the 270 entry in teorth/erdosproblems at commit 536408716724a2402bdfa7c1d764bf83d64a8c89. Its Git blob is 4e60263a80af6207aca494d2e1388e44fa372889, its file SHA-256 is be65b0db56fe13becec3178f2937685ee7e6a2f5df9325712474722c279c4182, and it records informal status disproved, formal status Lean, and formalized: yes. Those are catalogue metadata, not GCL mathematical certification. The maintained Erdős Problems #270 entry also reports the strong negative result and labels its formalised statement “Yes.”
Formal provenance has an important unavailable-source exception. The governed MATHFORGE snapshot is commit 638190868060e60e109197af602fda4fb80fd02b; its provider manifest has SHA-256 faa43f174717ef0126114b9bbd61bb9fdc827ce7eb9f64e56c1e63c7b15608db, its composition has SHA-256 d63066c67be1aa341466da4fa85d6d1eb4530fe435d2f359e634eebb6982e4d1, and its two FormalConjectures shards have SHA-256 33d1929f630431fe731cf0880cd01312cee3c7fa043ed5f1d4499fa28c0533fd and b9284333f076b0b4b702f08d977c8e0bdede815599e22750e0c70c5ee214c54f. The pinned provider revision is 85f863718beeec7b58a3a1926ee92e3472bc2020, but FormalConjectures/ErdosProblems/270.lean is absent there.
The later upstream file does not repair that gap as certification. The inspected file at commit fce15fcc952657d9865918b839912e9b45c1ee9c has SHA-256 29271c676bad4d08577e66256f4a2bb8f85e3fdc757db9292ab2b2d231384106 and contains sorry declarations. An external plby/lean-proofs file at commit 8822f7ddef30fadbd92e1c6ab4ed897af356af5e has SHA-256 770cb83813d0ed9d56fb03fdd88916a3c59b56c4e76a207f1aaea474bc379e97 and no visible sorry, but it was not independently replayed and has no check-run evidence in this audit. Neither artifact is treated as GCL mathematical certification.
What remains
For the original problem, the mathematical status is disproved: Theorem 1 supplies a rational value in the unrestricted value set. The remaining mathematical question is the semantically distinct nondecreasing variant. Theorem 2 narrows its possible values topologically and measure-theoretically, but “measure zero” does not mean “contains no rational,” so no upgrade to proved or disproved is justified there.
The formal task is narrower and still open as a documentary matter. A future replay would need to bind the exact external source, toolchain, theorem type, imports, and axioms, and compare its function and index conventions with the original expression. In particular, it must retain positivity of \(f(n)\) and the limit-to-infinity hypothesis. A successful project build or a theorem whose body contains sorry would not establish the mathematical result. The source audit therefore remains PARTIAL, and this manuscript preserves that exception rather than upgrading formal metadata into proof authority.
Sources and claim boundaries
- The original statement is in P. Erdős and R. L. Graham, Old and New Problems and Results in Combinatorial Number Theory (1980), printed p. 66, in the inspected scan
80_11_number_theory.pdf, SHA-2560cbf0c32f0ab1e1c71db5121a88bac905bf976c4a6ab6bb6d7d9cf9ddd184ed3. - The resolving primary paper is Tonći Crmarić and Vjekoslav Kovač, “On the irrationality of certain super-polynomially decaying series,” arXiv:2504.18712v1, submitted 25 April 2025. The inspected arXiv record and HTML text identify Theorem 1, Theorem 2, Lemmas 3–4, and Section 3; the inspected PDF has SHA-256
af4ffbfaff21442f542b70020d93b628002d2db8c3de3783c0f4a8aacc044b13. - The maintained status snapshot is Erdős Problems #270, which was inspected for the current explanation and its formalisation label. The fixed catalogue and FormalConjectures identities are given above; the absent pinned
270.lean, latersorryfile, and unreplayed external proof are deliberate claim boundaries. - The exact original issue and exact RESULT/1 source comment are documentary evidence. The protected V2 ledger determines editorial acceptance. This explanation asserts no GCL mathematical certification, no external CI success, and no claim beyond the inspected sources and stated exceptions.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1540 — 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.