Skip to content

Erdős Problem #247

Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: open (not automatically the live result).
Subject areas: number theory, 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 article does not confer mathematical certification.

The question in plain language

Start with a strictly increasing sequence of positive integers

\[ 1\leq a_1<a_2<a_3<\cdots. \]

Its terms tell us which powers of \(1/2\) to keep. Erdős Problem #247 asks whether the resulting number

\[ \sum_{n=1}^{\infty}2^{-a_n} \]

must be transcendental whenever the exponents sometimes run arbitrarily far ahead of their indices. The exact growth condition is

\[ \limsup_{n\to\infty}\frac{a_n}{n}=\infty. \]

In quantifier form, for every real \(C>0\) and every starting index \(N\), there is some \(n\geq N\) with \(a_n/n>C\). Thus occasional very large excursions of the exponents are enough; the ratio need not tend to infinity at every index. The question is about transcendence, which is stronger than irrationality: a transcendental number is not a root of any nonzero polynomial with rational coefficients.

The apparent simplicity is deceptive. The series converges extremely quickly, but a rapidly convergent series can still have an algebraic or even rational sum if its exponents have special structure. The problem asks whether the particular amount of irregular growth above rules out every algebraic possibility.

A small example

Take \(a_n=n^2\). It is strictly increasing and

\[ \frac{a_n}{n}=n\longrightarrow\infty, \]

so it satisfies the problem's hypothesis. The first five terms can be checked with exact integer arithmetic:

\[ \sum_{n=1}^{5}2^{-n^2} =2^{-1}+2^{-4}+2^{-9}+2^{-16}+2^{-25} =\frac{18940417}{33554432} \approx0.5644684136. \]

This is a finite partial sum, not the value of the infinite series and not a proof of transcendence. The remaining tail is already tiny. For \(n\geq6\), \(n^2\geq6n\), so

\[ 0<\sum_{n=6}^{\infty}2^{-n^2} \leq\sum_{n=6}^{\infty}2^{-6n} =\frac{2^{-36}}{1-2^{-6}}<2^{-35}. \]

This gives an independently reproducible finite check of the scale of the number while preserving the exact infinite question. It also shows why numerical approximation cannot settle the problem: an enormous amount of precision would still not identify whether the limiting value is algebraic.

For contrast, \(a_n=2n\) gives the geometric series

\[ \sum_{n=1}^{\infty}2^{-2n}=\frac13, \]

but \(a_n/n=2\), so this sequence does not satisfy the required growth condition. The contrast is useful: the hypothesis is doing real work, but the finite examples do not say whether it does enough.

What is known

The target with only \(\limsup a_n/n=\infty\) remains unresolved in the audited source record. The public problem page labels it as an open question, has no proof exposition and no proof claim, and was last edited on 20 January 2026. The pinned catalogue snapshot likewise records open, with the solution marked unformalized.

A stronger result is classically proved and reported in Erdős's literature. The 1980 Erdős–Graham book states that transcendence follows when

\[ \limsup_{n\to\infty}\frac{a_n}{n^t}=\infty \qquad\text{for every positive integer }t. \]

That family includes much faster-growing sequences, but not every sequence satisfying the problem's linear-limsup condition. In particular, \(a_n=n^2\) satisfies the open condition but does not satisfy the displayed requirement for \(t>2\). The 1975 paper cited by the audit contains the underlying Liouville-type results for rapidly growing reciprocal-series denominators; it should not be silently identified with a direct proof of the weaker Problem #247 statement.

The 1988 paper gives a related, stronger reported interface: it says that the series is transcendental if \(a_n/n^t\to\infty\) for every positive integer power \(t\), and then asks whether the weaker limsup condition might suffice. Its discussion also records the unresolved direction of proving that \(a_n>c n^2\) forces the series not to be a root of a quadratic polynomial. These are external mathematical reports, not a GCL replay. No externally reported proof of the exact linear-limsup question was found in the audited packet.

GCL's current state

The authenticated source audit on MATH-PROGRAMME issue #1516 returned RESULT/1 with disposition PARTIAL. It normalizes the source statement, records the live and pinned status, checks the finite calculation above, and distinguishes the stronger 1980 and 1988 conditions from the open target. Its proposed next step is only a bounded indexing-and-quantifier normalization task; it expressly does not launch a solver campaign or attempt the open transcendence implication.

The pinned Formal Conjectures file is useful evidence about a formal statement, not proof evidence. It quantifies over a function on \(\mathbb N\), assumes StrictMono, uses \(k+1\) in the denominator, and does not require the first Lean value to be positive. Both its open theorem and its stronger variant contain sorry. Therefore the file neither certifies the theorem nor establishes literal equivalence with the source indexing \(1\leq a_1<a_2<\cdots\).

GCL has no recorded independent theorem certification, formal proof replay, or source acceptance for this problem. The catalogue's protected status is operational metadata, and the editorial manuscript is documentary work only.

What remains

The mathematical target is still the exact implication

\[ \left(1\leq a_1<a_2<\cdots\ \land\ \limsup_{n\to\infty}\frac{a_n}{n}=\infty\right) \Longrightarrow \sum_{n=1}^{\infty}2^{-a_n}\text{ is transcendental}. \]

A positive solution must handle every strictly increasing positive-integer sequence with the stated occasional superlinear excursions. A negative solution would require an explicit sequence satisfying that condition whose series is algebraic, not merely a sequence with bounded ratio such as \(2n\).

The immediate bounded GCL task is narrower and safer: formalize the shift from a source-indexed sequence \(a_1,a_2,\ldots\) to a Lean function indexed from zero, preserve positivity, and prove exact equivalence of the limsup hypotheses for the stronger polynomial family. It should be tested on concrete sequences such as \(n^2\) and a factorial-growth sequence, while keeping the weak linear-limsup condition separate. That task can repair statement semantics; it cannot solve the transcendence problem. Any attempt to promote it into a proof of the open implication would cross the evidence boundary.

Sources and claim boundaries

  • Erdős Problems #247 — live statement, status, bibliography and edit date. It is a catalogue page, not proof authority.
  • Pinned catalogue evidence — the protected 2025-08-31 snapshot records open, unformalized, and statement-formalized metadata. The live file has drifted, so the pinned packet is retained.
  • Erdős, “Some problems and results on the irrationality of the sum of infinite series,” Journal of Mathematical Sciences 10 (1975), printed pp. 1–2, source PDF. The audit records SHA-256 f5377859f73a9dd84901ae3c5188eb5eb5ab44c1bb86ce0d2933fafda569afe0; that retained hash is provenance, not a GCL certification.
  • Erdős and Graham, Old and New Problems and Results in Combinatorial Number Theory (1980), printed p. 61, scan, SHA-256 0cbf0c32f0ab1e1c71db5121a88bac905bf976c4a6ab6bb6d7d9cf9ddd184ed3. This is the primary source for the stronger limsup family reported above; its proof was not independently reconstructed here.
  • Erdős, “On the irrationality of certain series: problems and results” (1988), printed p. 106, source PDF, SHA-256 b2bfc375d04b65332d6b8817633ff3968283a3f33c1f1ace366b03ac9fab8c88. The cited scan matched this audit-recorded hash in the present read-only check; the hash is provenance, not a GCL certification.
  • RESULT/1 audit and pinned Lean statement — GCL evidence and formal syntax, not independent proof certification.

The article makes no claim of a new theorem, a proof of the open case, independent verification of the cited deep arguments, formal certification, editorial acceptance, or publication authority. Editorial acceptance, if granted, is separately recorded in the protected admissions ledger.

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1516 — 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.

Back to catalogue · Compendium introduction