Skip to content

Erdős Problem #259

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

Let \(\mu(n)\) be the Möbius function. Its square is especially simple: \(\mu(n)^2\) is \(1\) when \(n\) is squarefree and \(0\) when \(n\) has a squared prime factor. Problem 259 asks whether the positive-term series

\[ S=\sum_{n\ge 1}\mu(n)^2\frac{n}{2^n} \]

is irrational. In other words, after throwing away every integer containing a factor \(p^2\) with \(p\) prime, weight each remaining integer \(n\) by \(n/2^n\), and ask whether the resulting real number can equal a ratio of two integers.

The original problem page writes the sum without an explicit lower limit, as is usual in this catalogue. The pinned Lean declaration sums over natural numbers, including \(n=0\); its \(n=0\) term is \(0\), so this convention does not change the value. The question is therefore about one fixed number, not about an unspecified family of bases or coefficients.

The maintained problem page records the classical status as proved by Chen and Ruzsa. That status is mathematically distinct from the GCL status reported below: this manuscript is a documentary draft, and neither a catalogue label nor an external formalization is itself a GCL-certified proof.

A small example

The squarefree positive integers up to \(20\) are

\[ 1,2,3,5,6,7,10,11,13,14,15,17,19. \]

Thus the corresponding finite approximation is

\[ \frac{1}{2}+\frac{2}{2^2}+\frac{3}{2^3}+\frac{5}{2^5}+\cdots+\frac{19}{2^{19}}. \]

Writing \(Q\) for the set of positive squarefree integers, reducing the terms exactly gives

\[ \sum_{n\in Q,\ n\le20}\frac{n}{2^n}=\frac{890183}{524288}=1.6978893280029296875. \]

This is a reproducible check of the statement's normalization, not a test for irrationality. Since the squarefree indicator is at most \(1\), the omitted positive tail is bounded by the full geometric-derivative tail:

\[ 0<S-\frac{890183}{524288}\le\sum_{n>20}\frac{n}{2^n}=\frac{20+2}{2^{20}}=\frac{11}{524288}<2.1\times10^{-5}. \]

The finite calculation and the bound can be checked using exact integer fractions; they do not establish that the limiting value is irrational. They only show how quickly the sparse squarefree series begins to settle numerically.

What is known

The useful change of viewpoint is to list the squarefree numbers in increasing order, say \(a_1<a_2<\cdots\). Reindexing the nonzero terms gives

\[ S=\sum_{j\ge1}\frac{a_j}{2^{a_j}}. \]

Erdős's 1981 paper proves an earlier gap criterion: if a sequence of positive integers satisfies \(a_{j+1}-a_j\to\infty\), then the corresponding series \(\sum_j a_j2^{-a_j}\) is irrational. That interface does not directly settle Problem 259, because the squarefree sequence is not a sequence whose successive gaps tend to infinity. Erdős later stated a stronger squarefree-subseries conjecture, cited on the problem page at page 103 of his 1988 chapter.

Chen and Ruzsa's 1999 result is the decisive strengthening. In the theorem interface identified by the source audit, let \(r\ge2\) and \(\ell\ge2\). If \((a_j)\) is an injective sequence of positive integers and, for every \(j\), \(a_j\) is either r-free or squarefull, then

\[ \sum_{j\ge1}a_j\ell^{-a_j} \]

is irrational. Here “r-free” means that no rth power of an integer greater than \(1\) divides the term; “squarefull” means that every prime divisor occurs with exponent at least two. For Problem 259 take \(r=2\), \(\ell=2\), and let \((a_j)\) enumerate the squarefree integers. Every \(a_j\) is 2-free, so the theorem's disjunction is satisfied and the displayed series is exactly \(S\).

The proof interface is a rationality contradiction. Assuming the series is a rational number \(u/v\), one forms carefully shifted and rescaled tails. Rationality forces a certain correction expression, after multiplication by the fixed denominator \(v\), to be an integer. The number-theoretic construction supplies forbidden intervals containing no terms of the allowed type; those gaps make the correction expression positive but drive it to \(0\) by geometric-tail estimates. An integer that is eventually strictly between \(0\) and \(1\) cannot exist. This explains why the theorem needs both an injective sequence and structural restrictions on its values: they control the finite correction and the tail separately.

GCL's current state

EXCEPTION — PARTIAL, with a source-verification gap. The authenticated latest audit on original issue #1529 is the single RESULT/1 comment, posted by fyremael. It normalizes the Möbius statement, records the Chen–Ruzsa criterion, and gives the exact finite check above, but it does not claim source-lock completion. The older cohort snapshot saying NO_RESULT_AT_READBACK is stale relative to that live comment and must not be used in place of it.

The pinned maintained catalogue is teorth/erdosproblems at commit 5364087, entry 259; it says proved, with reported Lean formalization metadata dated 2026-04-21. The pinned Formal Conjectures declaration states the target but ends in by sorry. It is a statement and proof-interface record, not a completed proof.

The linked external Lean source contains a much fuller development: a rational-tail lemma, the generalized Chen–Ruzsa theorem, the squarefree reindexing, and the specialization to \(S\). The audit reports no sorry or axiom in that source, but no clean build under its exact toolchain was established. It is therefore an external, unreviewed formalization claim, clearly separate from accepted classical proof attribution and from GCL certification.

What remains

The central unresolved documentary claim is the exact printed location and text of the decisive Chen–Ruzsa theorem. The published article is Chen and Ruzsa, “On the irrationality of certain series,” Periodica Mathematica Hungarica 38 (1999), 31–37, DOI 10.1023/A:1004742930674. The source audit identifies the relevant result as “Theorem 4” through the external formal development, but the article's exact theorem page has not been verified from an accessible copy. The article is closed access in the current audit environment; that limitation must remain visible rather than being filled by an inferred page number.

A bounded follow-up is consequently clear: obtain a licensed or author copy, hash it, record the printed theorem page, and compare every quantifier with the generalized interface above. Then compile the linked Lean source under its declared Lean and mathlib versions and check the two semantic bridges: Möbius square equals the squarefree indicator, and increasing enumeration preserves the sum. A failed retrieval, failed exact-toolchain build, or mismatch in positivity, injectivity, r-free, squarefull, or base assumptions should be recorded as a blocker. None of these tasks changes the classical published attribution; they determine only whether the external formal claim can be reviewed.

Sources and claim boundaries

The original source packet is the Erdős Problems page for 259, which cites Erdős and Graham, Old and New Problems and Results in Combinatorial Number Theory (1980), p. 62; Erdős, “Some Problems and Results on Additive and Multiplicative Number Theory,” in Analytic Number Theory, LNM 899 (1981), p. 180; Erdős, “Sur l'irrationalité d'une certaine série,” C. R. Acad. Sci. Paris 292 (1981), 765–768, primary PDF; and Erdős, “On the Irrationality of Certain Series: Problems and Results,” in New Advances in Transcendence Theory (1988), p. 103, audit-linked PDF. The catalogue page's bibliography also gives the exact Chen–Ruzsa publication above.

The 1981 paper is primary evidence for the earlier gap-condition interface, not a direct proof of the squarefree case. The 1988 citation is retained as source evidence for Erdős's stronger conjecture, while its exact scan was not independently re-read in this environment. The Chen–Ruzsa publication is the classical proof attribution, but its theorem page and proof text remain unresolved in this source audit. The maintained catalogue and both Lean artifacts document status or formal interfaces only. This draft makes no claim of a new proof, no GCL mathematical certification, no accepted Lean formalization, and no completion of the open documentary obligations.

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1529 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
  • GCL triage: OUTSIDE_STRICT_OPEN_TRIAGE.
  • 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