Erdős Problem #282
Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: open (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.
Editorial authority: The protected GCL V2 admissions ledger alone determines publication status. This source-bounded explanation is not mathematical certification.
The question in plain language
An Egyptian fraction writes a rational number as a sum of unit fractions, fractions whose numerator is \(1\). Problem 282 asks about a very specific greedy way to make such a sum. Start with a reduced fraction \(x=a/b\) satisfying \(0<a<b\) and with odd denominator \(b\). At a stage where the positive remainder is \(r\), choose the smallest odd positive integer \(n\) such that
but impose the crucial rule that a denominator already used may not be chosen again. Subtract \(1/n\) and repeat. The question is whether this process always reaches remainder \(0\) after finitely many steps. In symbols, the proposed conclusion is that for every reduced \(a/b\) with \(0<a<b\) and \(b\) odd, the odd greedy expansion terminates.
The word “greedy” matters twice. The next denominator is the least eligible one, not merely any odd denominator that works, and “eligible” means unused as well as large enough. A finite sum of distinct odd-denominator unit fractions is therefore not automatically the output of this algorithm. The problem concerns the behavior of one deterministic procedure for every allowed starting fraction.
This distinction also explains why the problem is subtle. The remainder decreases whenever a positive unit fraction is subtracted, but a decreasing sequence of positive rational remainders need not hit zero in finitely many steps. The denominators can grow, and the question is about exact finite termination rather than convergence toward zero.
A small example
Take \(x=2/3\). The first eligible odd denominator is \(3\), giving a remainder of \(1/3\). The denominator \(3\) is now unavailable, so the next choice is \(5\), not \(3\) again. The successive choices are \(3\), \(5\), \(9\), and \(45\):
Thus
Every denominator is odd and every denominator occurs once. This example is also a warning about the wording. If reuse were allowed, one could write \(2/3=1/3+1/3\), but that is not the odd greedy algorithm in the source problem. The exact choice at each line can be checked by comparing the reciprocal with the current remainder: after \(1/3\), the unused odd denominator \(5\) is the first one with reciprocal at most \(1/3\); after \(1/5\), the first such unused odd denominator is \(9\); and after \(1/9\), it is \(45\).
What is known
Erdős and Graham’s Old and New Problems and Results in Combinatorial Number Theory gives the source formulation on p. 30, including the “not yet used” convention, attribution to Stein, and related questions about arithmetic progressions and square denominators. The broad subject has many positive representation theorems, but those theorems do not settle the greedy process.
For example, Graham’s On Finite Sums of Unit Fractions gives a criterion for representing a rational number as a finite sum of distinct unit fractions whose denominators lie in an arithmetic progression; the relevant concluding result is on pp. 206–207. Graham’s On Finite Sums of Reciprocals of Distinct nth Powers gives the general criterion in Theorem 4 and the square-denominator consequence in Corollary 1, p. 91. These results say when a representation exists. They do not say that repeatedly taking the least available denominator will find one.
The ordinary greedy algorithm with all positive denominators is much better behaved: the classical Fibonacci result gives finite termination in that unrestricted setting. Restricting the allowed denominators to the odd integers changes the problem. Pihko’s “Remarks on the ‘Greedy Odd’ Egyptian Fraction Algorithm” defines the odd procedure in (1.1), gives its expansion in (1.2), and records the corresponding uniform termination question as Open Problem 1.1 in The Fibonacci Quarterly 39 (2001), 221–227. Pihko’s 2010 sequel, p. 202 and §3, including Corollary 3.6, gives structural families with prescribed increasing numerator runs; that is meaningful progress about the process, not a proof of universal termination. Louwsma and Martino, Rational Numbers with Two-Term Odd Greedy Expansion, Integers 25 (2025), A46, classify the two-term case while leaving general termination open.
The authenticated RESULT/1 also reports a finite exact computation: 2,006 reduced fractions with odd denominator less than \(100\) were tested, and all terminated in that tested range. The reported maximum was 18 terms, for \(84/85\), with a largest denominator having 3,122 decimal digits. Such a computation is useful evidence and a source of examples, but no finite search settles a statement quantified over all odd denominators.
GCL's current state
The pinned catalogue packet records Problem 282 as open, with formal-solution status unformalized and statement metadata marked formalized. The protected GCL triage record places it in PROTECTED_READY_BACKLOG; those are workflow and source-status labels, not a mathematical solution. The original issue is #1552, and its authenticated original SOURCE RESULT/1 has disposition PARTIAL. It preserves the unused-denominator normalization, the primary-source map, and a bounded exact-computation design, but reports no proof or disproof.
The pinned Formal Conjectures file is FormalConjectures/ErdosProblems/282.lean at commit 85f863718beeec7b58a3a1926ee92e3472bc2020, file blob 76f3cd2d6a83fe744bbfd2ef9b1b931e6eaf1796. Its principal declaration quantifies over rational \(x\) in \((0,1)\) with odd reduced denominator and states
In the pinned file’s Lean syntax, this target is greedyUnitFractionRem { n | Odd n } x =ᶠ[atTop] 0.
This is the right broad interface for the main question: x.den is the reduced denominator, and the recursive definition chooses the least odd denominator meeting the reciprocal bound. In the positive-remainder regime, the next remainder is smaller than the reciprocal just used, so a later eligible denominator must be larger; that is why the definition is compatible with distinct choices. Nevertheless, the declaration ends in by sorry. Its variants also contain answer(sorry). A formal file with sorry is not a Lean proof, and eventual-zero notation in this file is not independent certification of a finite representation.
What remains
The central unresolved task is a theorem proving or refuting finite termination for every reduced \(a/b\) with \(b\) odd under the exact unused-odd greedy rule. A bounded successor computation can still be valuable: enumerate all such fractions with \(b\le199\), record each exact denominator sequence and remainder, and independently check minimality, oddness, distinctness, and the final identity. It should stop at the first counterexample or at the prescribed bound. Its acceptance test must be exact, and its conclusion must remain “verified through the bound,” never “proved for all inputs.”
The next mathematical step would need a mechanism that controls the changing remainder and the rapidly growing unused denominators, or a rigorously certified counterexample. Existing representation criteria do not supply that mechanism. Any stronger status would require source-semantic review and a genuine proof record, not an extrapolation from the computation or from the formal interface.
For orientation only, Problem 279 remains OPEN despite an unreviewed GPT-generated 2026 proof claim; that claim is not treated as a proof and must not be imported into the status of 282. A separate source issue also remains material: Problem 286 has an EXACT_BLOCKER involving a width coefficient stated as \((e-1)k\) in Erdős–Graham versus a different width formulation in Croot. That source-semantic conflict must be investigated, not silently erased or reused as evidence for 282.
Sources and claim boundaries
The documentary identity is the GCL registry entry for problem 282 in governance/erdos_catalogue/REGISTRY.json: issue-body SHA-256 52f5540272eb9f48450039581f32b9e1e5f6f7d48d6db266abf321629fa8a457; catalogue source teorth/erdosproblems at commit 536408716724a2402bdfa7c1d764bf83d64a8c89, blob 4e60263a80af6207aca494d2e1388e44fa372889, source SHA-256 be65b0db56fe13becec3178f2937685ee7e6a2f5df9325712474722c279c4182.
The primary references are Erdős and Graham, Old and New Problems and Results in Combinatorial Number Theory (1980), p. 30, scan; R. L. Graham, On Finite Sums of Unit Fractions, Proc. London Math. Soc. 14 (1964), 193–207, especially the arithmetic-progression result on pp. 206–207, scan; and R. L. Graham, On Finite Sums of Reciprocals of Distinct nth Powers, Pacific J. Math. 14 (1964), 85–92, Theorem 4 and Corollary 1, p. 91, paper. The odd-greedy status references are Pihko (2001), Definition (1.1), (1.2), and Open Problem 1.1, pp. 221–227, paper; Pihko (2010), p. 202, §3, Corollary 3.6, paper; and Louwsma–Martino (2025), A46, paper.
These sources support the stated historical results, the open-status report, the worked calculation, and the distinction between representability and greedy termination. They do not certify a solution. The catalogue, issue, RESULT/1, and Lean file are evidence records with explicit scope; the finite computation is bounded evidence; and this manuscript is an editorial explanation only. It carries no mathematical independence, certification, publication, or acceptance authority.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1552 — 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.