Skip to content

Erdős Problem #269

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

The question in plain language

Choose a finite set of primes \(P\) with at least two members. List, in increasing order, every positive integer whose prime divisors all belong to \(P\):

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

For each \(n\), form the least common multiple of the first \(n\) listed numbers, and ask whether the resulting reciprocal series is irrational:

\[ S(P)=\sum_{n=1}^{\infty}\frac{1}{\mathrm{lcm}(a_1,\ldots,a_n)}. \]

The quantifier is important. The problem asks whether, for every finite prime set \(P\) satisfying \(|P|\ge 2\), the corresponding number \(S(P)\) is irrational. It does not ask whether one particular choice such as \(P=\{2,3\}\) works, nor whether a finite partial sum is irrational. The same \(P\) determines the whole infinite list and every later least common multiple.

The convention \(a_1=1\) matters: \(1\) has no prime divisors, so it is included for every \(P\). Also, successive denominators may repeat. For example, adjoining \(8\) after \(6\) changes the LCM, but adjoining \(9\) after \(8\) changes it again; other terms can leave it unchanged. “Removing duplicate summands,” mentioned in the historical source, is a different series and must not be silently substituted for the one above.

A small example

Take the set \(P=\{2,3\}\). The first twelve numbers smooth with respect to \(P\) and their running LCMs are

\[ (a_1,\ldots,a_{12})=(1,2,3,4,6,8,9,12,16,18,24,27), \]

and

\[ (L_1,\ldots,L_{12})=(1,2,6,12,12,24,72,72,144,144,144,432), \]

where \(L_n=\mathrm{lcm}(a_1,\ldots,a_n)\). Exact integer arithmetic gives

\[ \sum_{n=1}^{12}\frac1{L_n} =1+\frac12+\frac16+\frac1{12}+\frac1{12}+\frac1{24}+\frac1{72}+\frac1{72}+\frac1{144}+\frac1{144}+\frac1{144}+\frac1{432} =\frac{52}{27}. \]

This is independently reproducible with Python’s exact Fraction type:

from fractions import Fraction
from math import gcd
a = [1, 2, 3, 4, 6, 8, 9, 12, 16, 18, 24, 27]
L, current = [], 1
for n in a:
    current = current * n // gcd(current, n)
    L.append(current)
assert sum((Fraction(1, x) for x in L), Fraction()) == Fraction(52, 27)

The calculation checks the indexing and repeated denominators, not the infinite claim. No finite computation can decide irrationality of the limit merely by producing rational partial sums.

What is known

Erdős and Graham state in Old and New Problems and Results in Combinatorial Number Theory (1980), p. 65, that if the prime set is infinite, the analogous LCM-reciprocal series is irrational, and then ask what happens for a finite set with more than one prime. Erdős’s 1988 survey, p. 106, calls the infinite-prime assertion a simple exercise and says that the finite version “probably remains true.” Those passages establish the historical shape of Problem 269: the finite case is the unresolved target, not a consequence of the infinite case.

For an infinite set of primes, the smooth numbers have enough new prime-power structure for the elementary irrationality argument recorded by Erdős to apply. That result does not provide the needed conclusion for a fixed finite \(P\). In the finite case, every number is built from a fixed collection of prime powers, and the LCM growth has a much more rigid multi-dimensional pattern.

The latest source audit also records a public reduction for the two-prime case. With \(P=\{p,q\}\), it rewrites the series through a floor-sum and a Hecke–Mahler value, and reports a transcendence route using cited external theorems. The audit explicitly says that this route is not formalized or independently peer-reviewed there. It is therefore useful evidence about a possible resolution of the two-prime subfamily, but it is not a certified proof in this article and it does not settle all finite sets, beginning with three primes.

The live maintained entry labels the problem OPEN and says that finite computation cannot resolve it. Its note that the infinite-prime case is known is consistent with the two original papers. The historical Fibonacci Quarterly letter, including its statement about removing duplicate summands, was not inspected in this source audit; that unavailable source remains an exception rather than a basis for a stronger claim.

GCL's current state

The authenticated GitHub identity used for the source audit is fyremael. The original issue is #1539, and the exact authenticated RESULT/1 source-audit comment has disposition STATUS_CONFLICT. It reports the finite catalogue entry as open, while preserving the public two-prime argument as an unreviewed status conflict. That disposition is documentary evidence, not mathematical certification.

The pinned catalogue packet is data/problems.yaml at commit 536408716724a2402bdfa7c1d764bf83d64a8c89, with entry blob 4e60263a80af6207aca494d2e1388e44fa372889. Entry 269 records informal and formal status open, formal status unformalized, and statement-formalized metadata yes. The live Erdős Problem 269 entry remains the maintained status reference and currently presents the finite problem as open.

The fixed GCL catalogue registry binds problem 269 to issue #1539, records catalogue status open and triage class PROTECTED_READY_BACKLOG, and records issue-body SHA-256 03987e7cc5089b86b8dd4666dcf8ed71b6905b354d505d774a6af89bcacf780d. Its pinned source SHA-256 is be65b0db56fe13becec3178f2937685ee7e6a2f5df9325712474722c279c4182; the registry file itself has SHA-256 9e3e8c8f22a1ef59c5c9fefb7d409d5404d018cb64ff13f4b8cb5baf28920609.

The pinned FormalConjectures declaration is at commit 85f863718beeec7b58a3a1926ee92e3472bc2020, blob f721d3ed515ed92b01dc157e09cd5dc172dd22b3. Its theorem statements contain sorry, so the file records formal interfaces only. It also starts its sequence at \(a_0=1\) and includes the empty-prefix term in its series. Consequently its series is \(1+S(P)\), not literally \(S(P)\); adding \(1\) preserves rationality or irrationality, but the semantic translation must be stated.

The protected V2 ledger determines editorial acceptance. Neither this manuscript, the catalogue, the issue result, a sorry-bearing Lean file, an external proof repository, nor self-reporting CI supplies mathematical certification.

What remains

The original finite-prime question remains open at the programme’s documentary boundary. A complete resolution would need a proof for every finite \(P\) with \(|P|\ge2\), or one rigorously verified counterexample. Establishing the two-prime route would still leave the three-prime and higher-prime cases. Conversely, a failed finite search is not evidence that the universal statement is false: the missing quantifier ranges over an infinite series and all admissible finite prime sets.

The most sharply bounded next task is the first unresolved case \(P=\{2,3,5\}\). The audit proposes proving or refuting the exact cofinal residue-escape condition used by the three-prime carry reduction. Acceptance would require every quantifier, denominator, starting index, and tail estimate, or a rigorous counterexample showing that the condition fails infinitely often. Computation can test initial windows, but it cannot discharge a cofinal assertion.

Any formal or computational result must first reconcile the \(a_0\) versus \(a_1\) indexing and the additive \(1\) in the formal series. A successful build, an external repository, or a theorem tagged as solved is evidence about an artifact, not by itself a proof of the mathematical statement. The source audit also identifies the exact three-prime artifact hashes and preserves their own boundary: structural LCM identities do not imply irrationality, and the carry-escape premise remains unproved.

Sources and claim boundaries

The status reported here is open, with an explicitly preserved source-status conflict around the unreviewed two-prime route. No claim of a completed proof, disproof, formal certification, editorial acceptance, or external-claim promotion is made.

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1539 — 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