Erdős Problem #374
Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: open (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
Problem 374 concerns when a product of different factorials becomes a square. For a positive integer \(m\), define \(F(m)\) to be the smallest number of distinct factorials needed in a product whose largest factorial is \(m!\) and whose product is a perfect square. More explicitly, we look for integers
such that \(a_1!a_2!\cdots a_k!\) is a square. If no such product exists, \(F(m)\) is left undefined for the counting question. Let \(D_k\) be the integers with \(F(m)=k\). The problem asks how quickly the number of elements of \(D_k\) up to \(n\) grows, for \(3\leq k\leq 6\). One highlighted case is whether six-factor examples have positive density: roughly, whether a fixed positive proportion of integers require exactly six factorials.
This is a question about asymptotic size, not about finding one especially large example. It also matters that “exactly \(k\)” excludes integers already represented with fewer factors.
A small example
For \(m=6\), direct integer arithmetic gives
The five possible two-factor products with largest argument 6 are \(1!6!=720\), \(2!6!=1440\), \(3!6!=4320\), \(4!6!=17280\), and \(5!6!=86400\); none is a square. Thus this finite check verifies \(F(6)=3\). It illustrates both parts of the definition: producing a square is not enough; minimality must also be checked.
A second exact identity explains why four-factor examples are plentiful. If \(m=v^2u\) with \(v\geq2\) and \(u\geq2\), then
For instance, with \(m=12\), \(v=2\), and \(u=3\), both sides equal \(229442532802560000\). The four arguments are distinct and ordered. This proves a useful finite-family inclusion, but it does not determine whether the four-factor set has a limiting density, and it says nothing by itself about six factors.
What is known
The frozen catalogue packet records Problem 374 as open and unformalized at the pinned teorth/erdosproblems commit 536408716724a2402bdfa7c1d764bf83d64a8c89; the entry’s Git blob is 4e60263a80af6207aca494d2e1388e44fa372889, with last update 2025-08-31. Its four listed OEIS links are metadata, not proofs. The maintained page at erdosproblems.com/374 likewise presents the open formulation and summarizes earlier facts: primes have no representation, \(D_2\) is the square set, the six-factor set begins at 527, and no \(D_k\) exists for \(k>6\).
A materially newer primary source is Fedir Yudin’s Square products of factorials and a conjecture of Erdős and Graham, arXiv:2610.01899v1, submitted 1 October 2026. Theorem 1.1 on p. 1 states \(D_3(X)=\kappa_3\sqrt X+O_\varepsilon(X^{2/5+\varepsilon})\), with explicit \(\kappa_3=2.709751\ldots\). Theorem 1.2 on p. 1 gives a positive-proportion lower bound for exactly five and exactly six factors. Combining these results with classical identities, Corollary 1.3 on p. 2 states \(D_3(X)\sim\kappa_3\sqrt X\) and \(D_4(X)\asymp D_5(X)\asymp D_6(X)\asymp X\), while \(D_2(X)\) has square-root order and \(D_k(X)=0\) for \(k\geq7\).
The same paper is careful about scope: it says the orders of growth asked for in Problem 374 are settled, but asks separately whether \(D_k(X)/X\) has a limit for \(k=4,5,6\). Thus “positive density in the order-of-growth sense” and “a limiting density exists” are different claims. The older website status and the newer primary preprint have not been reconciled in the frozen ledger.
GCL's current state
The original GCL issue #1644 has zero comments at this audit, so no source result is being duplicated. Its protected triage class is STATEMENT_CAPTURE_REQUIRED, and its source packet explicitly says that no statement file is present in the admitted FC-GDM-85F86371 provider snapshot. The local triage record agrees: protected_fc.present is false, with no path or blob hash.
The current upstream google-deepmind/formal-conjectures tree also has no 374.lean entry at that exact tree commit. Therefore there is no pinned Lean statement to inspect, no theorem marked by sorry for this problem, and no Lean replay evidence. The external paper is mathematical evidence, not GCL certification. V2 ledger alone decides accepted coverage.
What remains
The first bounded successor should capture a source-locked statement that explicitly separates the counting functions from the stronger density-limit questions. It should record the normalization \(D_k(X)=\#\{m\leq X:F(m)=k\}\), explain the harmless difference between starting at 1 or 2, and attach the exact theorem locations in Yudin v1 and the historical Erdős--Graham source.
The next mathematical task can then be a semantic comparison, not a fresh attempt to solve the whole problem: verify line by line that the primary theorem’s \(F(m)\), distinctness condition, and minimality convention match the catalogue’s \(D_k\). A falsification would be a counterexample to equivalence, a missing hypothesis, or a theorem that only counts representations rather than exact minima. If equivalence holds, a later formalization may target the finite definitions and the elementary four-factor identity first; it must not promote the asymptotic theorem without a separately pinned formal source and proof.
Sources and claim boundaries
- Original assignment and issue-specific instructions: MATH-PROGRAMME issue #1644. The original SOURCE RESULT/1 citation is authenticated original RESULT/1.
- Frozen catalogue:
problems.yamlat commit536408716724a2402bdfa7c1d764bf83d64a8c89, entry blob4e60263a80af6207aca494d2e1388e44fa372889. - Historical primary source: Erdős and Graham, “On Products of Factorials,” Bull. Inst. Math. Acad. Sinica 4 (1976), 337–355, especially the quoted six-factor conjectural discussion on p. 354, archived PDF, locally retrieved SHA-256
bed77d559ca9ec1e4d028382e590acc198014bd5e14607fe078808a4da8576c7. - Current primary report: Yudin v1, arXiv:2610.01899, locally retrieved PDF SHA-256
fa2d536863e9b70034086847d39c13ebb3b09ede989ec10b32841c75bca84297.
The catalogue and live page are dated status records; Yudin v1 is an external preprint report; the displayed finite calculations are independently checked exact arithmetic; and no FormalConjectures or GCL mathematical certification is present. Those categories must remain separate until an accountable coordinator reconciles the status conflict.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1644 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
- GCL triage: STATEMENT_CAPTURE_REQUIRED.
- 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.