Skip to content

Erdős Problem #313

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.

The question in plain language

Erdős Problem 313 asks whether a particular kind of Egyptian fraction can be built in infinitely many ways. Choose distinct primes \(p_1<\cdots<p_k\) and an integer \(m\geq 2\). The question is whether there are infinitely many choices for which

\[ \frac{1}{p_1}+\cdots+\frac{1}{p_k}=1-\frac{1}{m}. \]

The original source is Erdős and Graham, Old and New Problems and Results in Combinatorial Number Theory (1980), p. 40, where this is recorded as an open problem. The formulation matters: the primes must be distinct, \(m\) is an integer, and the question concerns infinitely many solutions, not merely finding another example.

There is a useful normalization. If \(P=p_1\cdots p_k\), clearing denominators shows that every \(p_i\) divides \(m\); writing \(m=Pt\) then forces \(t=1\). Thus every solution has \(m=P\). Equivalently, one asks for infinitely many squarefree integers \(m\) satisfying

\[ \frac{1}{m}+\sum_{p\mid m}\frac{1}{p}=1. \]

These \(m\) are called primary pseudoperfect numbers.

A small example

Take the primes \(2\), \(3\), and \(7\). Exact integer arithmetic gives

\[ \frac{1}{2}+\frac{1}{3}+\frac{1}{7}=\frac{21+14+6}{42}=\frac{41}{42}=1-\frac{1}{42}. \]

So \(m=42=2\cdot3\cdot7\) is one solution. The same calculation verifies the first terms \(m=6\) and \(m=1806\), using \((2,3)\) and \((2,3,7,43)\) respectively; the denominators are products of the displayed distinct primes. This is a reproducible finite check, not evidence of infinitude. A pattern that works for several steps can stop when the next candidate is composite, and a finite list cannot settle the unbounded question.

The normalization also explains why the two languages are equivalent. Multiplying the reciprocal identity by \(m\) gives \(1+\sum_{p\mid m}m/p=m\), which is exactly the primary-pseudoperfect equation. Conversely, a squarefree \(m\) satisfying that equation supplies the prime set of an Erdős solution.

This reduction is elementary but important for auditing claims. A statement about infinitely many admissible prime sets is the same statement as one about infinitely many admissible squarefree products; it is not a change to a weaker counting problem. It also explains why primality of every displayed factor must be checked before a numerical candidate is called a solution.

What is known

The frozen catalogue packet records Problem 313 as open, with informal and formal status last updated on 2025-08-31. It records eight known examples, ending with \(8490421583559688410706771261086\), and tags the problem as number theory and unit fractions. The maintained catalogue at commit 76848eddd25687dfc69bb26bec740c35c4a9c6b4 still gives the same open status and dates, so no catalogue status drift was found; its current data/problems.yaml blob is 97ae243a7fc056bb9c6413ce665b575eadd27ff3.

The relevant primary source is the Erdős–Graham volume cited above, linked through the stable PDF copy here. Later literature gives finite progress rather than a resolution. Butske, Jaje, and Mayernik's paper “On the equation ...” is the standard computational source for the first layers; the 2026 preprint Port Fillings for Primary Pseudoperfect Numbers, version 1 dated 18 May 2026, reports two further examples, including \(5998279018951962402\) and its inherited successor \(35979351189199316534587473905773572006\). Its stated unconditional results are finite constructions and a port formalism. Its infinitude statement is conditional on a multivariable prime-points hypothesis, not a proof of Problem 313.

The live Erdős Problems page has three public comments. One says that the chain \(6,42,1806\) continues infinitely, but another comment records that the next suggested term fails because \(1807=13\cdot139\) is not prime. The third points to the 2026 preprint and explicitly calls it partial. These comments are useful leads, not verified mathematical authority; the page itself says comments are not checked for correctness.

GCL's current state

The source packet is bound to original issue #1583, assignment ERDOS-CATALOGUE-0313-S01. The issue is a direct-editorial, cooperative source-audit assignment, and its protected triage class is PROTECTED_READY_BACKLOG. At audit time it had no comments and no existing RESULT/1, so there was no duplicate result to preserve. The requested article is documentary evidence only: accepted coverage is decided by the V2 ledger and its protected controls, not by this manuscript.

The pinned formal source is FormalConjectures/ErdosProblems/313.lean at commit 85f863718beeec7b58a3a1926ee92e3472bc2020, blob f66fde9147bba84a217750bbcfc504168a8703f4. It defines solutions using a nonempty finite set of natural-number primes and proves finite example lemmas in the source. The central research theorem is declared as answer(sorry) ↔ ...Infinite and ends by sorry. The file therefore records a formal statement and some finite certificates, not a completed formal proof; no Lean replay or certification is claimed here.

What remains

The central unresolved task is an unconditional proof or disproof that the set of primary pseudoperfect numbers is infinite. The current preprint offers a conditional recursive route and new finite examples, but neither closes that gap. A safe next investigation is a bounded audit of one fixed successor layer: take the exact key port and identities from the preprint, verify its displayed factorizations and primality certificates with independently reproducible integer arithmetic, and enumerate only the stated finite parameter interval. The acceptance test should be exact equality plus primality and squarefreeness checks; a failed equality, composite candidate, or incomplete certificate is a falsification of that finite candidate, not of the original conjecture. Stop when the prescribed interval is exhausted or a source identity cannot be reproduced.

Sources and claim boundaries

The source lock consists of the original issue; the frozen catalogue commit 536408716724a2402bdfa7c1d764bf83d64a8c89, catalogue blob 4e60263a80af6207aca494d2e1388e44fa372889; the live catalogue commit and blob named above; the formal-conjectures commit and blob named above; Erdős–Graham (1980), p. 40; Butske–Jaje–Mayernik; and Wang's arXiv:2605.21518v1. Dates and statuses are reported as snapshots, not permanent declarations. External forum claims remain unverified reports. This article supplies no mathematical certification, publication decision, or claim that the open problem has been solved. Original SOURCE RESULT/1 citation: authenticated original RESULT/1.

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1583 — 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