Erdős Problem #266
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.
Editorial authority: The protected V2 ledger determines acceptance. This explanation is documentary synthesis, not a mathematical proof or formal certification.
The question in plain language
Problem 266 concerns an infinite list of positive integers \(a_1,a_2,\ldots\) whose reciprocal sum converges. The Stolarsky conjecture asks whether at least one positive integer shift must make the shifted sum irrational. Written with the quantifiers made explicit, the claim is
The live maintained entry phrases this as “there exists some integer \(t\geq1\)” after introducing an arbitrary admissible sequence. It attributes the question to Stolarsky and labels it DISPROVED. The positive shift matters: because \(a_n+t>0\), no denominator can vanish. The original sources also use the equivalent sign convention \(1/(a_n-t)\) for integers \(t\) avoiding the poles \(t=a_n\); replacing \(t\) by \(-t\) explains that convention, but it should not be silently conflated with the positive-shift display.
The quantifier order is the heart of the problem. It does not say that one fixed \(t\) works for every sequence. It says that each admissible sequence should have at least one successful shift. To disprove it, one sequence must defeat every positive integer \(t\) simultaneously.
A small example
A finite calculation helps separate rational arithmetic from the infinite assertion. Take the increasing list \((2,3,7,43)\) and form the four-term sum
Exact integer arithmetic gives
For example, \(S_2=1/4+1/5+1/9+1/45=45/180+36/180+20/180+4/180=105/180=7/12\). Every finite sum of rational terms is rational, so this is only a sanity check; it is not a counterexample to an assertion about an infinite series.
The calculation is independently reproducible with Python’s standard-library Fraction implementation:
from fractions import Fraction
a = (2, 3, 7, 43)
for t in range(4):
print(t, sum((Fraction(1, x + t) for x in a), Fraction()))
The output is the four displayed fractions. This exact finite check is also recorded in the authenticated source-audit return RESULT/1, whose scope is evidence reporting rather than certification.
What is known
The historical formulation is cited to Erdős and Graham, Old and New Problems and Results in Combinatorial Number Theory (1980), p. 64, and to Paul Erdős, “On the irrationality of certain series: problems and results” (1988), p. 104. The 1980 book page was not directly available for inspection in this audit; its wording is therefore retained as a cited historical source, not presented as a fresh transcription. The 1988 author scan was inspected directly. On printed p. 104 Erdős records Stolarsky’s question in the \(1/(n_k-t)\) convention, with \(t\) excluded from the sequence, and says the “all \(n\)” version remained open at that time.
Kovač and Tao’s arXiv version 4, dated 14 July 2025, resolves the question negatively. Their Theorem 2.11 states that there exists a strictly increasing sequence of positive integers \((a_n)\) such that
converges to a rational number for every \(t\in\mathbb{Q}\setminus\{-a_n:n\in\mathbb{N}\}\). This is stronger than required: every positive integer is among those allowed rational shifts, and for positive shifts the pole exclusion is automatic. Hence the existential conclusion in Problem 266 fails for this one sequence, so the status is disproved, not open or merely partially settled.
Section 8 of the same paper gives the construction. It enumerates the rationals, builds increasingly large blocks, and uses rational target values plus approximation and perturbation steps so that all shifted sums are rational. The paper’s displayed scale has eventual size comparable to \(2^{2^{\sqrt{k}}}\); the proof is an existence construction, not a short closed-form list of the \(a_n\).
GCL's current state
The authenticated GitHub identity used for this audit is fyremael. The authoritative Programme issue is #1536, and the exact source return is RESULT/1. The batch record classifies the return as RESULT_PRESENT_UNADJUDICATED; that label records documentary evidence, not a mathematical verdict issued by GCL.
The fixed catalogue entry is number: "266" at teorth/erdosproblems commit 536408716724a2402bdfa7c1d764bf83d64a8c89, Git blob 4e60263a80af6207aca494d2e1388e44fa372889, with inspected full-file SHA-256 be65b0db56fe13becec3178f2937685ee7e6a2f5df9325712474722c279c4182. The local fixed registry is governance/erdos_catalogue/REGISTRY.json, Git blob 90f6499cfbaca8c8a9e1ac4c6beb468ca09127a5; its Problem 266 row says disproved and OUTSIDE_STRICT_OPEN_TRIAGE.
The protected formal-conjectures source is FormalConjectures/ErdosProblems/266.lean at commit 85f863718beeec7b58a3a1926ee92e3472bc2020, Git blob 7576d18de3197c75c3553969926924398d6c410b, inspected SHA-256 9e6dd0888fb70e3dc62d38359b547400db58ee42d4928e719416f403263e2044. Its main theorem negates the universal claim over functions \(a:\mathbb{N}\to\mathbb{N}\) with positive values and real reciprocal summability; its variant adds strict increase and all rational non-pole shifts. Both declarations contain sorry. Thus “formalized: yes” or “formal status: Lean” is metadata about a statement, not a checked formal proof.
What remains
No mathematical question remains at the level of Problem 266’s truth value once Kovač–Tao’s Theorem 2.11 is accepted as the primary proof: the conjecture is disproved. Documentary work does remain. The 1980 p. 64 book page should be inspected if a direct quotation or exact historical typography is needed. The sign and indexing conventions should continue to be reported explicitly: the 1988 source uses \(n_k-t\), while the live entry and Kovač–Tao use \(a_n+t\); the stronger rational-shift theorem bridges them, but this is a semantic comparison, not permission to rewrite the sources.
A separate formalization project could prove the elementary bridge from a strictly increasing positive sequence to convergence of each shifted tail, and then prove the logical implication from an all-rational counterexample to the negation of the original quantifier statement. That would not formalize Kovač–Tao’s infinite-dimensional construction. Any such work must use a pinned environment and a proof term without sorry.
Neither an external proof repository, self-reporting CI, a statement file, nor a Lean declaration containing sorry is mathematical certification. This manuscript therefore reports the published result and its boundaries without promoting formalization metadata into a proof claim.
Sources and claim boundaries
- Live maintained entry: Erdős Problem #266, read back over HTTPS on 13 October 2026; it displays DISPROVED, the positive-integer statement, and the Kovač–Tao attribution.
- Primary resolution: V. Kovač and T. Tao, “On several irrationality problems for Ahmes series,” arXiv:2406.17593v4, dated 14 July 2025; inspected HTML, Theorem 2.11 and Section 8. No DOI is asserted here because none was needed or inspected for this source.
- Historical primary source inspected: P. Erdős, “On the irrationality of certain series: problems and results,” New Advances in Transcendence Theory (1988), pp. 102–109, especially printed p. 104, author scan; the chapter DOI record is 10.1017/CBO9780511897184.009.
- Historical source cited but not directly inspected: P. Erdős and R. L. Graham, Old and New Problems and Results in Combinatorial Number Theory (1980), p. 64. This unavailable-source exception is deliberate.
- Fixed catalogue and formal sources: pinned
problems.yamlentry and pinned formal file. Their hashes and limitations are recorded above. - Exact Programme evidence: original issue #1536 and its exact source-comment link. The return is source-audit evidence only; it does not grant certification, publication, or admission authority.
Publication and editorial acceptance require separate protected evidence and rendering controls; manuscript presence alone grants neither.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1536 — 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.