Erdős Problem #372
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
For an integer \(n>1\), let \(P(n)\) be its largest prime factor. Erdős Problem #372 asks whether the prime factors at the tops of three consecutive integers can fall strictly at every step:
The question is existential and infinite: are there infinitely many starting values \(n\) with this pattern? It is not asking that every triple behave this way, nor that the three integers themselves be prime. For example, a composite integer can have a very large prime factor, while its neighbors may have only smaller prime factors. Strict inequalities matter: ties do not count.
The maintained problem page says that Erdős and Pomerance conjectured this descending pattern and that Balog later proved it. It also reports a stronger quantitative statement, namely that the number of such \(n\le x\) is at least a constant multiple of \(\sqrt{x}\) for all sufficiently large \(x\). That stronger assertion is useful context, but it is kept separate here from the exact infinitude statement because the primary Balog paper could not be opened during this audit.
A small example
The first simple witness is \(n=13\). Exact factorization gives
Therefore
This calculation can be reproduced by trial division: divide each of \(13\), \(14\), and \(15\) by primes up to its square root, retain the largest prime divisor, and compare the three retained values. An exact enumeration using the same integer factorization procedure found 72 witnesses among \(2\le n\le500\), beginning with \(13,14,34,37,38,43,61,62,73,79,86,94\). The adjacent starts \(13\) and \(14\) are not a mistake: they refer to overlapping triples.
The finite calculation establishes only that the pattern occurs at these checked values. It cannot establish infinitude, the asymptotic lower bound reported for Balog's result, or any density such as the later heuristic value \(1/6\). A computation also does not substitute for a proof that the factorization routine was applied to every integer beyond the tested range.
What is known
The live Erdős Problems entry for #372, accessed 2026-10-13, gives the exact statement and attributes the resolution to Antal Balog, “On triplets with descending largest prime factors,” Studia Scientiarum Mathematicarum Hungarica 38 (2001), 45–50, DOI 10.1556/SScMath.38.2001.1-4.3. The public abstract says that infinitely many consecutive triples have descending largest prime factors. The maintained page additionally reports the \(\gg\sqrt{x}\) lower bound for starts \(n\le x\).
The primary paper is the relevant proof source, but its publisher page returned an access denial in this investigation and no stable full-text copy was located within the bounded timebox. Consequently, author, journal, year, volume, pages, DOI, and the abstract-level conclusion are recorded; a theorem number, exact proof hypotheses, and the precise location of the quantitative lower bound remain unverified from the paper itself. This is a source-access limitation, not evidence against the published result.
The earlier primary paper by Paul Erdős and Carl Pomerance, “On the largest prime factors of \(n\) and \(n+1\),” Aequationes Mathematicae 17 (1978), 311–321, is available from Carl Pomerance's copy. It studies comparisons of largest prime factors of consecutive integers and discusses the difficulty of extending such questions to three or more consecutive values. It is historical and contextual evidence for #372, not the proof of the descending triple claim.
GCL's current state
The frozen catalogue packet is teorth/erdosproblems at commit 536408716724a2402bdfa7c1d764bf83d64a8c89, entry number: "372", file blob 4e60263a80af6207aca494d2e1388e44fa372889, source SHA-256 be65b0db56fe13becec3178f2937685ee7e6a2f5df9325712474722c279c4182. At that revision, the catalogue records proved, with status date 2025-08-31, formal status unformalized, and statement metadata formalized: yes dated 2026-08-26. Those are snapshot fields, not GCL mathematical certification.
The current catalogue file is still the same entry at commit 503d0a58ff3b6a12d7ff1665aa37fb953416d85f, with file blob 97ae243a7fc056bb9c6413ce665b575eadd27ff3. This is a documented source update, not a reason to collapse catalogue metadata into proof authority. The original GCL issue #1642 is the source-audit record for this assignment and had no comments when checked. The original SOURCE RESULT/1 citation is authenticated original RESULT/1.
The admitted FC-GDM-85F86371 provider snapshot contains no #372 statement file. A later upstream commit, FormalConjectures bfea74a5521e83c60360b10c7af6c5d890df4553, adds a Lean statement with file blob 22a14565e651fd8627c423d79e482cbd0122eedc; its theorem ends in by sorry. It formalizes infinitude over natural numbers, but does not formalize Balog's reported \(\gg\sqrt{x}\) refinement. A statement containing sorry is not a checked proof and does not certify the mathematics.
What remains
The next bounded research step is to obtain a readable copy of Balog's 2001 paper and inspect pp. 45–50 line by line. The reviewer should record the exact theorem or proposition, its quantifiers, the convention for the centered triple, and whether the paper proves only infinitely many starts or the stronger count \(gg\sqrt{x}\). The comparison must identify explicitly that Balog's notation for a centered triple, such as \(P(m-1)>P(m)>P(m+1)\), is equivalent to the registered formulation after setting \(n=m-1\).
Acceptance for that successor task would be a source hash plus a page-level statement matching the registered inequalities and a clear separation between the primary theorem, the maintained-page summary, and the later Lean statement. A falsification or status conflict would be recorded if the primary text proves a different ordering, imposes an unreported hypothesis, or gives only a weaker claim. The task should stop after the bounded source search if the paper remains inaccessible; it should not infer theorem details from a citation, a catalogue label, or a sorry body.
Sources and claim boundaries
- Original GCL issue #1642: assignment, labels, pinned source packet, and authority boundary. Its original SOURCE RESULT/1 citation is authenticated original RESULT/1.
- Maintained Erdős Problems #372 page, accessed 2026-10-13: exact statement, dated live presentation, Balog attribution, quantitative report, and bibliography.
- Balog, “On triplets with descending largest prime factors”, Studia Sci. Math. Hungar. 38 (2001), 45–50: primary proof reference; full text was unavailable in this audit, so theorem-level details remain bounded and explicit.
- Erdős and Pomerance, “On the largest prime factors of \(n\) and \(n+1\)”, Aequationes Math. 17 (1978), 311–321: accessible primary context, not the descending-triple proof.
- Pinned catalogue bytes: commit and blob identities above; catalogue metadata remains distinct from acceptance or certification.
- Later FormalConjectures statement: exact upstream commit and blob above; statement-only, with
sorry, and outside the frozen GCL provider snapshot.
This article explains the registered question and its evidence boundaries. It does not claim a GCL proof, Lean replay, source certification, publication decision, or accepted-coverage status; the V2 ledger alone decides accepted coverage.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1642 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
- GCL triage: OUTSIDE_STRICT_OPEN_TRIAGE.
- Formal-statement coverage: No statement was confirmed in GCL's pinned formal-provider snapshot. Presence does not mean the statement or proof has been checked.
- External mathematical status: consult current problem page and its dated bibliography.