Skip to content

Erdős Problem #262

Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: solved (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

Choose a strictly increasing sequence of positive integers

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

Call it an irrationality sequence if, for every choice of positive integers \(t_1,t_2,\ldots\), the series

\[ \sum_{n=1}^{\infty}\frac{1}{t_n a_n} \]

is irrational. The multipliers are positive: \(t_n\geq1\). They are not allowed to be zero or negative, and the requirement is universal rather than existential. Erdős asks how slowly an irrationality sequence can grow.

The question is subtler than asking whether one rapidly convergent series is irrational. The sequence \(a_n=2^{2^n}\) works for every positive multiplier sequence, while merely requiring \(a_n\to\infty\) does not suffice. For instance, the factorial sequence has a positive multiplier choice that makes the sum rational. The target is therefore a growth boundary for a whole family of adversarial series.

A small example

For \(a_n=n!\), choose \(t_n=n+2\). Every multiplier is positive, but

\[ \frac{1}{(n+2)n!}=\frac{1}{(n+1)!}-\frac{1}{(n+2)!}. \]

Consequently the first \(N\) terms sum to

\[ \sum_{n=1}^{N}\frac{1}{(n+2)n!}=\frac12-\frac{1}{(N+2)!}, \]

and the infinite sum is \(1/2\). This is a concrete disproof of the claim that factorial growth automatically gives the required universal irrationality.

For the positive example \(a_n=2^{2^n}\), the first three denominators are \(4,16,256\), so with \(t_n=1\) the first three terms are

\[ \frac14+\frac1{16}+\frac1{256}=\frac{81}{256}. \]

The remaining terms are smaller than the geometric tail beginning with \(2^{-16}\), hence are less than \(1/65535\). The following independent calculation checks both exact identities and the tail bound:

from fractions import Fraction
from math import factorial

for N in (1, 2, 5, 10):
    s = sum((Fraction(1, (n + 2) * factorial(n))
             for n in range(1, N + 1)), Fraction())
    assert s == Fraction(1, 2) - Fraction(1, factorial(N + 2))

a = [2 ** (2 ** n) for n in range(1, 4)]
assert a == [4, 16, 256]
assert sum((Fraction(1, x) for x in a), Fraction()) == Fraction(81, 256)
assert Fraction(1, 2**16) / (1 - Fraction(1, 2**16)) == Fraction(1, 65535)

This verifies finite arithmetic and a convergence estimate only. It does not prove irrationality.

What is known

Erdős’s primary proof interface is a divisibility theorem, not a characterization of the slowest possible sequence. In “Some problems and results on the irrationality of the sum of infinite series,” printed p. 3 defines property \(P\): whenever \(m_k>0\) and \(m_k\) is divisible by \(n_k\), the sum of \(1/m_k\) is irrational. Theorem 3 on printed p. 6 proves that \(n_k=2^{2^k}\) has this property. Substituting \(m_k=t_k2^{2^k}\) gives the Problem #262 example for every positive integer sequence \((t_k)\).

The proof’s useful interface is exact: it accepts arbitrary positive multiples of the prescribed \(n_k\). Its limitations are equally important. Erdős uses denominator and least-common-multiple estimates after rearranging a sequence of denominators; the theorem proves one explicit construction, not that its double-exponential scale is optimal. The same paper records uncertainty about slower sequences having property \(P\).

Hančl’s 1991 paper supplies the principal obstruction. In Theorem 2 on p. 100, a sequence containing a subsequence \((c_n)\) with a bound of the form

\[ c_n<K2^{2^{n-F(n)}},\qquad F(n)<n,\qquad \sum_{n=1}^{\infty}2^{-F(n)}<\infty, \]

admits positive integer multipliers producing a prescribed sufficiently small positive target. Taking a rational target makes the sequence fail the irrationality-sequence property. Corollary 2 on p. 102 gives the simpler consequence

\[ \limsup_{n\to\infty}\frac{\log_2\log_2 a_n}{n}<1 \quad\Longrightarrow\quad (a_n)\text{ is not an irrationality sequence}. \]

Thus the classical record places a sharp-looking double-exponential scale between a proved example and a broad impossibility region, but it does not establish the exact answer to “how slowly.”

GCL's current state

The authenticated latest audit is RESULT/1 on original issue #1532, assignment ERDOS-CATALOGUE-0262-S01. Its disposition is conspicuously STATUS_CONFLICT. It confirms the positive-multiplier normalization, the factorial counterexample, Erdős’s Theorem 3, and Hančl’s obstruction. It also records the retrieved source hashes and the proposed reusable checks without presenting them as a proof certificate.

The pinned maintained catalogue entry calls #262 “solved (Lean)” and gives formal status Lean, while simultaneously recording formalized: no. The live problem page, last edited 28 September 2025, repeats the classical account and says “Formalised statement? No.” The audit reports that the current google-deepmind/formal-conjectures tree contains no ErdosProblems/262.lean declaration; the pinned admitted provider snapshot likewise has no statement file.

This is not a minor label difference. No Lean declaration, external formal proof, or GCL replay was available to inspect for #262. Therefore the catalogue’s Lean label is retained as metadata under dispute, not promoted here to an accepted formal proof. The classical papers establish the stated mathematical interfaces, while this article makes no GCL certification.

What remains

The central mathematical remainder is a characterization, or a sharper lower and upper growth boundary, for irrationality sequences. Erdős’s example and Hančl’s obstruction do not decide whether a substantially slower sequence exists, nor whether the displayed limsup threshold is sufficient in an appropriate sense. The live page’s phrase “essentially solved” should therefore be read as a report of strong partial results, not completion of the original optimization question.

The documentary and formal remainder is to reconcile the contradictory catalogue metadata. A future bounded successor could identify the source of the “solved (Lean)” label, formalize the exact predicate with \(t_n\geq1\), and replay any claimed development in its actual toolchain with its axiom and sorry report. The elementary factorial identity and finite geometric estimate are suitable interface lemmas; neither discharges Erdős’s or Hančl’s analytic arguments. Until that evidence exists, claims of Lean formalization remain external or unresolved.

Sources and claim boundaries

  • Original statement and authenticated source audit: MATH-PROGRAMME issue #1532, especially the exact source packet, and its latest authenticated RESULT/1 comment.
  • Maintained catalogue: teorth/erdosproblems, pinned problems.yaml, entry number: "262"; the audit identifies Git blob 4e60263a80af6207aca494d2e1388e44fa372889.
  • Primary positive construction: P. Erdős, “Some problems and results on the irrationality of the sum of infinite series,” Journal of Mathematical Sciences 10 (1975), 1–7 (1976), Theorem 3, printed p. 6, exact author-hosted PDF. The retrieved PDF hash recorded by the audit is f5377859f73a9dd84901ae3c5188eb5eb5ab44c1bb86ce0d2933fafda569afe0.
  • Historical restatement and factorial example: P. Erdős, “On the irrationality of certain series: problems and results,” New Advances in Transcendence Theory (1988), 102–109, definition and discussion on printed pp. 105–106, exact author-hosted PDF.
  • Obstruction: J. Hančl, “Expression of real numbers with the help of infinite series,” Acta Arithmetica 59 (1991), 97–104, Theorem 2 p. 100 and Corollary 2 p. 102, DOI and maintained DML record.
  • Formal-source boundary: the pinned formal-conjectures tree and current tree contain no admitted #262 declaration. Absence of a declaration is evidence of the conflict, not evidence that the classical theorem is false.

The article separates published mathematical results, finite recomputation, maintained-site metadata, and any future formal claim. It does not claim a complete solution, originality, certification, publication authority, or resolution of the status conflict.

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1532 — 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