Erdős Problem #268
Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: proved (not automatically the live result).
Subject areas: number theory.
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.
The question in plain language
Let \(A\) be an infinite set of positive integers for which the harmonic subseries converges. Form the three numbers
Problem 268 asks whether the collection \(X_3\) of all such triples contains a nonempty open set in \(\mathbb{R}^3\). “Contains an open set” means that there is at least one point and a positive radius \(r\) such that every point of the Euclidean ball of radius \(r\) around it belongs to \(X_3\). It does not mean that every triple is attainable, or that one fixed set \(A\) works for all triples. The quantifiers are: there exists a whole open region, and for each target point in that region there exists an infinite set \(A\) that represents it.
The equivalent sequence language is to enumerate \(A\) increasingly as \(a_1<a_2<\cdots\). The positive-integer convention matters: the denominators are \(n,n+1,n+2\) with \(n\geq 1\), so no zero denominator is present. The convergence assumption on the first sum also makes the other two sums converge, since each shifted term is no larger than \(1/n\).
The status is proved at the published-paper level. The formal-status evidence is a separate unresolved matter, reported below rather than silently promoted to certification.
A small example
Take the finite set \(A_0=\{1,3,10\}\). Its three partial sums are exactly
This is a deliberately finite calculation: \(A_0\) is not an admissible infinite set and therefore does not solve the problem. It does show how one target triple is assembled and makes the indexing visible. The arithmetic is independently reproducible with exact rational arithmetic:
from fractions import Fraction
A = {1, 3, 10}
print(tuple(sum((Fraction(1, n + j) for n in A), Fraction()) for j in range(3)))
The output is (Fraction(43, 30), Fraction(37, 44), Fraction(37, 60)). The same three-coordinate normalization, including the positive-index convention, is the one used in Kovač’s Theorem 1 and in the exact audit calculation linked in the sources below.
What is known
Erdős and Graham’s 1980 book records the two-dimensional theorem of Erdős and Straus and then asks on printed p. 65 whether the analogous three-dimensional set, with denominators \(a_k\), \(a_k+1\), and \(a_k+2\), contains an open set. Erdős’s 1988 paper repeats that the planar set contains open sets, says the higher-dimensional analogue was probably true, and explains that their proof was not published because the higher-dimensional case had not been worked out.
Kovač’s paper, “On the set of points represented by harmonic subseries,” gives the affirmative answer. Its Theorem 1 states that
has nonempty interior, where the paper explicitly takes \(\mathbb{N}\) to mean the positive integers. The proof uses a linear change of coordinates and a convergence game to handle perturbation errors. Section 5 gives a concrete open ball of radius \(10^{-24}\) inside the set. This is a published mathematical proof of the registered question, not merely a numerical search.
Kovač and Tao’s later paper, “On several irrationality problems for Ahmes series,” gives a broader result. Their Theorem 2.8 proves nonempty interior for tuples of \(d\) consecutive reciprocal sums even while requiring a suitably double-exponential lower-growth condition on the increasing sequence. Corollary 2.10 removes that growth restriction in the set formulation and proves the analogous assertion for every positive integer \(d\), hence for \(d=3\). Their construction is existential and does not provide one explicit sequence for every chosen rational target; that distinction is important.
GCL's current state
The pinned catalogue packet at teorth/erdosproblems commit 536408716724a2402bdfa7c1d764bf83d64a8c89 records Problem 268 as proved (Lean), updated 2026-04-13. The packet hash is be65b0db56fe13becec3178f2937685ee7e6a2f5df9325712474722c279c4182, and the protected MATH-PROGRAMME registry entry binds issue #1538 to issue-body SHA-256 0d042d806debfdb8670bcbc4c226fee51a9aea388b96a440baabbe9b60736d18. The maintained Erdős entry currently says “yes” and attributes the answer to Kovač, while the exact authenticated source audit on issue #1538, RESULT/1 has disposition STATUS_CONFLICT.
The authenticated GitHub identity used for that readback was fyremael. The issue is evidence custody and source reconciliation, not a mathematical certification channel.
That conflict is about formal provenance, not the inspected published proof. The pinned FormalConjectures file, 268.lean at commit 85f863718beeec7b58a3a1926ee92e3472bc2020, has Git blob 1feb3909bbba8f37584e4addcf493521da4b5363 and theorem body by sorry. It quantifies over arbitrary sets in Set ℕ, permits \(0\), and uses generic dimension d; that is not literally the positive-integer three-dimensional statement without a semantic bridge. The linked external gist has audited hash c4bd3b940fbb99cedddf4aee9ef8180d3e16cf01f8d78b2fae73e25bf79a4d8c, but the audit did not compile it. native_decide, an external proof repository, or self-reporting CI is not treated as mathematical certification.
The live page is mutable evidence: the audit recorded raw HTML hash 0d5ae934a2e777deb1e5164a4881364415dc5555dbbcd3bb5333d486d8ada99f, while the current readback has changed hash b24330362534e35dfa3c8ce32ee7721218468ae7c7f67762aba5c1d1ea19cbf2. Both report the affirmative paper-level status; the hash drift is retained rather than hidden.
What remains
No mathematical question remains at the paper level for the exact three-dimensional problem: Kovač’s theorem settles it positively. What remains is documentary and formal reconciliation. A rigorous formal claim would need a pinned, reproducible Lean/Mathlib build of the external gist, a check of its dependencies and axioms, and an explicit proof that its positive-index three-dimensional set is included in—or otherwise semantically matches—the pinned generic declaration. The zero-index convention, subtype summability, indexing, and the meaning of the three coordinates must all be handled explicitly.
The original-source trail also has bounded exceptions. The 1980 scan was inspected at p. 65 and has SHA-256 0cbf0c32f0ab1e1c71db5121a88bac905bf976c4a6ab6bb6d7d9cf9ddd184ed3. The 1988 scan was retrieved and inspected at pp. 104–105 with SHA-256 b2bfc375d04b65332d6b8817633ff3968283a3f33c1f1ace366b03ac9fab8c88; the audit records that its hash required TLS verification to be disabled, so archival transport is an explicit exception. The Kovač v3 PDF and Kovač–Tao v4 PDF were inspected with hashes 975c2a5ce714bf289918c178180cda33eb2cb62e0c3085881182ab7e71357282 and 5ef233db5bf94a072de7e094e6d42864ba2d1f639e7cb951c02b4cf1d3d52b57.
The protected V2 ledger determines editorial acceptance. That ledger boundary is independent of the published proof and of any future formal replay.
Sources and claim boundaries
- Original statement: Erdős and Graham, Old and New Problems and Results in Combinatorial Number Theory, p. 65, scan. The scan is the inspected primary source for the displayed three-dimensional question.
- Historical discussion: Erdős, “On the irrationality of certain series: problems and results,” pp. 104–105, scan. Its transport exception is described above.
- Resolution: Kovač, “On the set of points represented by harmonic subseries,” Theorem 1 and §5, Amer. Math. Monthly 132 (2025), 895–911, DOI 10.1080/00029890.2025.2540753.
- Generalization: Kovač and Tao, “On several irrationality problems for Ahmes series,” Theorem 2.8 and Corollary 2.10, Acta Mathematica Hungarica 175 (2025), 572–608, DOI 10.1007/s10474-025-01528-0.
- Maintained status: Erdős Problem 268, read as mutable maintained-entry evidence, and the pinned catalogue packet.
- GCL provenance: original issue #1538 and its exact authenticated RESULT/1 source comment.
The article therefore claims: proved by the inspected published mathematics; affirmative on the maintained live entry; formal status unresolved and source-audit disposition STATUS_CONFLICT. It does not claim that the pinned by sorry declaration, the external gist, or any CI report certifies the theorem, and it does not upgrade a mutable page or catalogue metadata beyond the cited primary sources.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1538 — 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.