Erdős Problem #393
Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: open (not automatically the live result).
Subject areas: number theory, factorials.
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 \(n\geq2\), take the factorial \(n!=1\cdot2\cdots n\). Erdős asks how close together the factors can be when \(n!\) is written as a product of distinct increasing integers. In the maintained wording, \(f(n)\) is the least integer \(m\geq1\) for which
The natural working convention is that the \(a_i\) are positive integers. The displayed live statement does not spell out the sign domain or say separately that \(t\geq2\); the condition \(m\geq1\) already rules out a one-term representation, while positivity is a normalization that should be checked against the printed source. For fixed \(m\), every factor lies in the short interval from \(a_1\) to \(a_1+m\), and the endpoints must both occur. The problem is to understand the resulting function as \(n\) grows, not merely to find a few values.
The original source attribution is P. Erdős and R. L. Graham, Old and new problems and results in combinatorial number theory (1980), p. 76, recorded as [ErGr80] by the maintained page. The exact source-audit issue is MATH-PROGRAMME issue #1663; its original SOURCE RESULT/1 citation is authenticated original RESULT/1.
A small example
For \(n=4\), exact arithmetic gives \(4!=24\). Two valid factorizations with endpoint gap \(2\) are
There is no gap \(1\) factorization: with strictly increasing integer factors and largest factor one more than the smallest, the only possible tuple is \((a,a+1)\), so one would need \(a(a+1)=24\). The quadratic \(a^2+a-24=0\) has discriminant \(97\), which is not a square. An exhaustive exact enumeration of all increasing positive tuples whose product is \(24\) returned minimum gap \(2\), gap \(2\) witnesses \((4,6)\) and \((2,3,4)\), and no gap \(1\) witness. Thus \(f(4)=2\) under the positive-factor normalization.
This is a finite certificate only. It checks one small input and says nothing about whether \(f(n)\) is bounded, tends to infinity, or follows a regular asymptotic law. It also does not settle whether allowing negative factors would change the intended definition; that sign convention belongs to the source-normalization record.
What is known
The maintained entry says that Erdős and Graham did not even know whether \(f(n)=1\) for infinitely many \(n\). The condition \(f(n)=1\) means that \(n!\) is a product of two consecutive positive integers, because there is no integer strictly between the endpoints. This remains an open subquestion, not a numerical pattern that finite computation can settle.
There are rigorous sparsity results for each fixed gap. Let \(F_m(N)\) count the integers \(n\leq N\) with \(f(n)=m\). Once \(m\) is fixed, choose which of the internal points \(a_1+1,\ldots,a_1+m-1\) occur. Each choice gives a fixed polynomial in \(a_1\), namely the product of the selected terms from \(a_1\) through \(a_1+m\), and \(f(n)=m\) is contained in the union of the corresponding equations \(P(a_1)=n!\). Berend and Osgood proved that this kind of polynomial-factorial equation occurs for a density-zero set of \(n\); consequently the maintained page records \(F_m(N)=o(N)\) for every fixed \(m\). Their paper is Daniel Berend and Charles F. Osgood, “On the equation \(P(x)=n!\) and a question of Erdős,” Journal of Number Theory 42 (1992), 189–193, doi:10.1016/0022-314X(92)90020-P.
Bui, Pratt, and Zaharescu give the stronger fixed-polynomial estimate that leads here to \(F_m(N)\ll_m N^{33/34}\). Their primary paper is “Power savings for counting solutions to polynomial-factorial equations,” Advances in Mathematics 422 (2023), article 109021, doi:10.1016/j.aim.2023.109021. These results control every fixed gap separately; they do not provide a uniform bound in \(m\) and therefore do not determine the behavior of \(f(n)\) itself.
Finally, Luca showed that the ABC conjecture implies \(f(n)\to\infty\). The implication uses the same finite family of polynomial-factorial equations for each fixed \(m\): ABC gives finiteness for each such family, and hence no fixed gap can occur infinitely often. This is conditional, not an unconditional solution. See Florian Luca, “The Diophantine equation \(P(x)=n!\) and a result of M. Overholt,” Glasnik Matematički 37 (2002), 269–273, open primary PDF.
GCL's current state
The frozen catalogue packet identifies teorth/erdosproblems commit 536408716724a2402bdfa7c1d764bf83d64a8c89 and the data/problems.yaml blob 4e60263a80af6207aca494d2e1388e44fa372889. Its entry for number: "393" says informal status open, formal status unformalized, and formalized: no, with all relevant catalogue updates dated 2025-08-31. The maintained live page, accessed 2026-10-13, gives the same definition and open status, and reports no proof exposition or proof claim. Its page history contains a 2025-10-20 revision, so the live page and frozen catalogue are different dated observations, not interchangeable hashes.
The packet also says that no statement file for #393 is present in the admitted FC-GDM-85F86371 provider snapshot. The local tree contains no exact FormalConjectures source for this problem, so there is no pinned Lean theorem, sorry declaration, or replay result to report. The relevant primary contextual paper, Erdős and Graham, “On products of factorials,” Bulletin of the Institute of Mathematics, Academia Sinica 4 (1976), 337–355, is available as an author-hosted scan; its downloaded SHA-256 was bed77d559ca9ec1e4d028382e590acc198014bd5e14607fe078808a4da8576c7. It studies products of factorials, but is not being presented as a proof of Problem 393.
Any accepted coverage is decided by the protected V2 ledger alone. This article records source-audit evidence and boundaries; it supplies no mathematical certification, publication decision, or claim that the problem has been solved.
What remains
The main unresolved question is the global behavior of \(f(n)\): in particular, whether fixed gaps recur infinitely often, whether \(f(n)\) tends to infinity unconditionally, and whether a sharper asymptotic description exists. The estimates for a fixed \(m\) leave a genuine quantifier gap: they say what happens after \(m\) is held constant, while the minimizing \(m=f(n)\) may grow with \(n\).
A bounded successor task is to construct an exact-arithmetic certificate for a chosen limit \(N\): enumerate all strictly increasing positive tuples with product \(n!\) for every \(2\leq n\leq N\), record \(f(n)\) and one witness, and independently verify the enumeration by prime-exponent vectors. Acceptance requires equal products, strict ordering, endpoint gap, and a proof that every smaller gap has been excluded for each tested \(n\). Falsification is a single missed tuple, an incorrect minimum, or a discrepancy between integer products and prime exponents. Stop at the declared \(N\) or at resource exhaustion; do not infer an infinite theorem from the table. A separate theorem task would need to preserve the distinction between fixed and growing gaps and explicitly resolve the sign convention from [ErGr80, p. 76].
Sources and claim boundaries
The exact statement and status here are taken from the live Problem 393 page and compared with the frozen catalogue commit and blob named above. The original historical reference is the Erdős–Graham 1980 scan, cited by the packet as p. 76; page-level source extraction and the positivity wording remain a documented normalization boundary. The fixed-gap results are attributed only to the primary Berend–Osgood, Bui–Pratt–Zaharescu, and Luca papers linked above. The 1976 paper is relevant background, not a substitute for the original problem page.
The \(4!\) calculation is an independently reproducible finite check. It establishes \(f(4)=2\) under the stated convention and nothing asymptotic. No unverified web claim, future issue URL, Lean replay, formal theorem with sorry, or GCL mathematical certification is being promoted as an established solution. The pending coordinator-posted SOURCE RESULT/1 citation is intentionally not fabricated.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1663 — 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.