Skip to content

Erdős Problem #303

Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: proved (not automatically the live result).
Subject areas: number theory, unit fractions.

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

Colour the integers with finitely many colours. Erdős Problem #303 asks whether one colour must contain three different nonzero integers \(a\), \(b\), and \(c\) satisfying

\[ \frac{1}{a}=\frac{1}{b}+\frac{1}{c}. \]

“Monochromatic” means that the same colour is assigned to all three numbers. The maintained Erdős Problems page uses the word “integers”, while the original Brown–Rödl paper states its result for the positive integers. The positive-integer theorem is the important primary formulation: a colouring of all integers restricts to a finite colouring of the positive integers, so a positive witness is also an integer witness. Distinctness matters: allowing \(b=c\), or allowing a zero denominator, would describe a different question.

The equation is a small unit-fraction equation. The problem is not asking for one especially convenient triple. It asks for a triple in every finite colouring, no matter how the colours have been assigned.

A small example

Take the two-colouring of the integers given by parity. The three distinct denominators \(4\), \(6\), and \(12\) all have the same colour, and exact fraction arithmetic gives

\[ \frac{1}{6}+\frac{1}{12}=\frac{2}{12}+\frac{1}{12}=\frac{3}{12}=\frac{1}{4}. \]

Thus \((a,b,c)=(4,6,12)\) is a monochromatic solution for this particular colouring. The calculation was checked with rational arithmetic, so it does not rely on decimal approximations. It establishes only one example for one colouring; it does not establish the universal assertion over all finite colourings. In particular, a finite search or a convenient colouring cannot replace the theorem’s quantifiers.

What is known

The primary source is T. C. Brown and V. Rödl, “Monochromatic solutions to equations with unit fractions”, Bulletin of the Australian Mathematical Society 43 (1991), 387–392. In the introduction on printed page 387, the authors describe the unit-fraction colouring question and then state a stronger result. Their Theorem 2.1, on printed page 389, converts a finite-colouring result for homogeneous integer equations into a finite-colouring result after replacing variables by reciprocals. Corollary 2.2, on printed page 390, says that for every finite colouring, every \(n\ge 2\), and \(1\le a\le n\), there are distinct monochromatic positive integers solving \(a/x_0=1/x_1+\cdots+1/x_n\). Taking \(a=1\) and \(n=2\) gives exactly the three-variable equation in #303.

This is an external published proof, not a conjectural report. The paper’s proof uses a compactness step and a least-common-multiple substitution; the source audit checked the five-page article, including the theorem and corollary pages. The maintained Erdős Problems entry also records the colouring version as true and cites Brown–Rödl. The live page was last edited 28 December 2025. The exact source semantics should still be kept visible: the paper says positive integers, whereas the website’s short wording says integers.

The quantifier “finite colouring” means that the number of colours is fixed and finite, but the colouring itself is arbitrary. Brown–Rödl’s statement is correspondingly stronger than checking a few named colourings: for every fixed number of colours, every assignment of those colours to the positive integers has a monochromatic witness. Their general reciprocal-substitution theorem explains why a homogeneous partition-regular equation can yield a unit-fraction equation. For #303, only the specialization \(n=2\) is needed; the paper actually supplies the whole family with \(n\ge2\).

GCL's current state

The original GCL issue #1573 supplies a frozen source packet. Its catalogue lock is teorth/erdosproblems commit 536408716724a2402bdfa7c1d764bf83d64a8c89, data/problems.yaml, Git blob 4e60263a80af6207aca494d2e1388e44fa372889. The entry says proved, with last update 2025-12-21, and reports Lean formalization metadata. The live catalogue file has since advanced to commit 503d0a58ff3b6a12d7ff1665aa37fb953416d85f; the #303 fields checked in that revision retain the same status lines. These catalogue fields are provenance evidence, not GCL mathematical certification.

The pinned FormalConjectures file is from commit 85f863718beeec7b58a3a1926ee92e3472bc2020, blob 609e2fcce3ad8e3523d2d885f88ff94201dded96. It quantifies over a finite-range colouring 𝓒 : ℤ → ℤ, then asks for integers \(a,b,c\) with [a,b,c,0].Nodup, the real equation, and a subsingleton colour image. That captures finite colours, nonzero values, distinctness, and monochromaticity, but it does not literally state positivity. Its theorem body is by sorry; the file is a formal interface, not a replayed proof or GCL certification.

The live upstream formal file has since been reorganized and has a different blob, so the pinned commit is essential for reproducibility. The source audit did not compile either version and did not treat the formal-status annotation as a proof result. At the worker preflight there was no source return. The subsequently posted authenticated original-issue RESULT/1 source return is the controlling evidence.

What remains

No unresolved mathematical proof gap was found in the primary source for the registered three-variable claim. The bounded documentary successor is a semantic comparison: record formally that Brown–Rödl’s positive-integer conclusion implies the integer-domain interface, and decide whether the Lean statement should instead preserve the source’s positive domain. The comparison should also check that answer(True) ↔ ... is being used as a statement of the answer rather than mistaken for proof evidence.

The distinction between external proof, formal statement, and GCL certification must remain explicit. The catalogue’s V2 ledger alone decides accepted coverage. This article does not certify the mathematics, validate Lean, or promote any catalogue label; it records the source evidence and its limits.

Sources and claim boundaries

The primary paper supports the mathematical status described here. The formal file supports only a source-level statement with sorry. No claim of independent certification, publication, or accepted V2 coverage is made.

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1573 — 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