Erdős Problem #264
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 \(a_1<a_2<\cdots\). Call it a Type-3 irrationality sequence if every bounded integer perturbation still has an irrational reciprocal sum. Precisely, the requirement is
implies
The boundedness quantifier is uniform: one single \(C\) must work for all indices. The perturbations may vary arbitrarily from term to term, but they may not be zero, and no denominator may vanish. For \(a_n=2^n\) and \(a_n=n!\), every permitted perturbed series converges absolutely, so the question is genuinely about rationality rather than convergence.
There is an important historical normalization issue. Erdős and Graham’s 1980 book and Erdős’s 1988 survey describe bounded integer shifts without explicitly writing \(b_n\ne0\). Kovač and Tao add that condition so that the \(2^n\) question is meaningful: if zero shifts were allowed, the rational identity \(\sum_{n\ge1}2^{-n}=1\) would immediately defeat the property. The modern website statement and the pinned Lean declaration use the added nonzero-shift convention. These are related formulations, but they must not be silently treated as identical.
A small example
For \(a_n=2^n\), the basic series is geometric:
The relevant Kovač–Tao test does not inspect this sum alone. It measures the squared tail by
Thus the lower-tail quantity is uniformly positive, exactly the shape required by their theorem. A finite, independently reproducible check is
for every \(n\). Exact rational Python arithmetic reproduced this value for \(n=1,3,8\); the omitted positive tail is what changes the finite value into \(1/3\).
The same calculation also shows why a finite experiment cannot settle the problem. A finite list of chosen shifts only produces a rational partial sum. The theorem must control an infinite continuation and show that some bounded continuation lands exactly on a rational number.
For comparison, factorials behave differently under this test. Writing \(k=n+j\) and using
gives
This is a check on a hypothesis, not a solution of the factorial problem.
What is known
Kovač and Tao prove the decisive negative result for the exponential branch. Their Theorem 2.5 in arXiv version 4, §2.1.3 (reported as Theorem 5 in the published numbering) says that if \(a_n\) is strictly increasing and positive, \(\sum_n1/a_n\) converges, and
then \(a\) is not a Type-3 irrationality sequence. Their Corollary 2.6 further covers every strictly increasing sequence with bounded successive ratio in the limsup sense. Since \(2^n\) has ratio \(2\) and the calculation above gives the required tail value \(1/3\), the theorem supplies a bounded nonzero integer perturbation whose reciprocal sum is rational. The source’s proof interface is existential, not an explicit formula for the perturbation.
The proof works by an interval-subsums lemma. After choosing a constant \(C\) from the lower-tail hypothesis, it allows positive shifts from a finite interval such as \(1,2,\ldots,4C+1\). The possible tail sums overlap enough that a rational target is hit. Positive shifts automatically satisfy the modern \(b_n\ne0\) condition, and the construction keeps denominators positive. This proves that the modern \(2^n\) branch is answered negatively.
The factorial branch is not answered by that theorem. The exact estimate above shows that its lower-tail hypothesis fails strongly, tending to zero. That only says that this proof interface cannot be reused directly. It does not produce an irrationality proof, a rational perturbation, or a counterexample. Kovač and Tao also show that Type-3 sequences can exist with prescribed sufficiently super-exponential growth, and explicitly note that the factorial variant is probably more difficult; growth rate alone therefore cannot settle the issue.
GCL's current state
STATUS EXCEPTION — mixed branch status and definition drift. The latest authenticated audit on original issue #1534, comment RESULT/1 has disposition SOURCE_LOCK_READY, not STATUS_CONFLICT or PARTIAL. Nevertheless, the mathematical content is split: modern external mathematics disposes of \(2^n\) negatively, while the \(n!\) question remains open. The older 1988 source also contains an unqualified Borwein attribution about \(2^n\) whose hypotheses do not match the modern Type-3 definition. It must not be quoted as a current proof of the opposite conclusion.
The protected catalogue snapshot teorth/erdosproblems at commit 536408716724a2402bdfa7c1d764bf83d64a8c89, entry number: "264", records informal status open, unformalized formal status, and statement-formalized metadata yes, all dated 2025-08-31. The live problem page erdosproblems.com/264 was reported in the audit as last edited 20 January 2026 and records the newer split: \(2^n\) is not Type 3, while \(n!\) remains unresolved.
The pinned FormalConjectures declaration defines bounded-above and bounded-below integer perturbations, excludes zero shifts and zero denominators, and indexes from \(\mathbb{N}\). Its IsIrrationalitySequence predicate does not itself require strict monotonicity; strict monotonicity is supplied separately in the general negative theorem. That is a semantic interface difference to review, not a cosmetic reindexing. The file declares the \(2^n\) negative result, the factorial question, and the Kovač–Tao interfaces, but every theorem body is by sorry. It is a useful statement and interface record, not a checked proof and not GCL certification.
What remains
The \(2^n\) subproblem needs no claim of a new GCL proof: the external Kovač–Tao theorem is the relevant primary proof source, and its hypotheses can be checked exactly. What remains is to preserve the definition boundary, reconcile historical wording, and avoid presenting the existence argument as an explicit computable perturbation.
The \(n!\) subproblem remains genuinely open in the source packet. A finite search over bounded perturbation prefixes can test particular rational targets, but failure to find a target proves nothing about all infinite perturbations. Conversely, one exact infinite perturbation with a rational sum would disprove the factorial Type-3 property. Any future formal work must first prove semantic equivalence between its indexing, boundedness, convergence, and nonzero-shift conventions and the modern statement. A Lean declaration, an external repository, or an unreviewed formalization claim cannot be promoted to an accepted proof merely because it compiles or is labelled solved.
Sources and claim boundaries
- Erdős and Graham, Old and New Problems and Results in Combinatorial Number Theory (1980), p. 63, the original book scan: the bounded-perturbation problem for \(2^n\) and \(n!\).
- P. Erdős, “On the irrationality of certain series: problems and results,” New Advances in Transcendence Theory (1988), pp. 102–109, especially p. 105, the exact PDF: historical variants and the older, differently worded attribution.
- V. Kovač and T. Tao, “On several irrationality problems for Ahmes series,” arXiv:2406.17593v4 (14 July 2025), §2.1.3, Theorem 2.5 and Corollary 2.6; published in Acta Mathematica Hungarica 175 (2025), pp. 572–608, arXiv record, HTML proof text, and journal DOI.
- Authenticated RESULT/1 source audit: exact source packet, status drift, finite checks, formal-definition comparison, and the
SOURCE_LOCK_READYdisposition. - Formal declaration at the pinned commit: statement metadata only; theorem bodies contain
sorry.
The catalogue, live page, primary papers, and formal declaration are evidence with different authority. This manuscript reports the external proof for \(2^n\), keeps \(n!\) open, and makes no claim of GCL mathematical certification or editorial acceptance.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1534 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
- GCL triage: PROTECTED_READY_BACKLOG.
- Formal-statement coverage: A provider statement was included in the pinned snapshot. Presence does not mean the statement or proof has been checked.
- External mathematical status: consult current problem page and its dated bibliography.