Erdős Problem #399
Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: disproved (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
Erdős Problem 399 asks whether a factorial can ever be made from two like powers. In the notation of the problem, are there natural numbers \(n,x,y,k\) with \(xy>1\) and \(k>2\) such that either
or
The second form can be written with a nonnegative right-hand side as \(n!+y^k=x^k\). The condition \(xy>1\) excludes the uninteresting cases in which one of the two bases is zero or the product is too small. The exponent is at least \(3\). Thus the question is a universal nonexistence claim: one single valid quadruple \((n,x,y,k)\) would disprove it.
The exact formal encoding pinned for this audit uses natural numbers and the disjunction \(( n! = x^k+y^k ) \mathbin{\lor} ( n!+y^k=x^k )\), together with \(1<x y\) and \(2<k\). This is the right normalization for the reported example, but it should not be confused with a classification of every possible solution.
A small example
The reported counterexample is small enough to verify without approximation:
and
Here \(xy=48\cdot36=1,728>1\), \(k=4>2\), and the two bases are not coprime: \(gcd(48,36)=12\). The same identity may be written as \(10!+36^4=48^4\). I checked each displayed integer calculation with exact integer arithmetic; no floating-point or search heuristic is involved.
This calculation is decisive for the yes-or-no question as normalized: it supplies an admissible witness to the existential negation of “there are no solutions.” It does not show that this is the only witness, does not solve the coprime subproblems, and does not establish any general theorem about factorials and powers. In particular, the common divisor \(12\) is mathematically important because several older results impose \(gcd(x,y)=1\).
What is known
The primary source inspected here is P. Erdős and R. Obláth, “Über diophantische Gleichungen der Form \(n!=x^p+y^p\) und \(n!\pm m!=x^p\),” Acta litterarum ac scientiarum Regiae Universitatis Hungaricae Francisco-Josephinae, volume 8 (1937), pages 241–255. The authors explicitly restrict the main investigation to coprime \(x\) and \(y\). Their Theorem 1, on page 249, rules out factorial sums or differences of powers with exponent \(p\) in the stated non-power-of-two range, apart from the elementary exceptional solution for the square-sum case. Their fourth-power difference argument is different: Theorem 3, on page 254, gives nonexistence only for sufficiently large \(n\) when the two integers are coprime. These are substantial partial theorems, not a proof of the unrestricted Problem 399.
The maintained problem page also records two narrower results: Erdős and Obláth’s coprime result for \(k\ne4\), and Pollack–Shapiro’s result that \(n!=x^4-1\) has no solutions. It further records a congruence observation for the coprime equation \(n!=x^4+y^4\). Those statements concern restricted families. They do not invalidate the non-coprime witness above.
The exact source of the Barfield attribution remains less clear. The live page says that Jonas Barfield found the displayed solution, while the formal source says that the example is discussed in problem D2 of Guy’s 2004 collection. I found no primary publication by Barfield in the source packet, and therefore treat the attribution as a reported provenance claim. The identity itself is independently verifiable, so the mathematical disproof does not depend on trusting a numerical transcription; the historical credit does.
GCL's current state
The original source-audit request is MATH-PROGRAMME issue #1669. Its frozen catalogue packet points to teorth/erdosproblems commit 536408716724a2402bdfa7c1d764bf83d64a8c89, file blob 4e60263a80af6207aca494d2e1388e44fa372889. That entry records informal status “disproved,” with last update 2026-02-04, formal status Lean, and a formalized statement. The live Problem 399 page, accessed 2026-10-13, agrees that the displayed solution exists and currently lists no proof exposition.
The pinned FormalConjectures file has commit 85f863718beeec7b58a3a1926ee92e3472bc2020 and blob 22c49edfdd834fa9534cf9a14e9761f9ea53026d. Its main theorem records the negative answer and supplies the witness \((10,48,36,4)\) with by decide. Four narrower variant declarations remain sorry: the Erdős–Obláth coprime theorem, Pollack–Shapiro’s theorem, Cambie’s mod-8 statement, and the sum-of-two-squares classification. File presence and a proof term are evidence about the formal source, not GCL mathematical certification.
What remains
The useful next step is a bounded provenance-and-interface check, not an exhaustive Lean build. Its input should be the two frozen blobs above, the live Problem 399 page, the 1937 scan, and the bibliographic trail for the Barfield example. Its target should be a one-page ledger that states whether a primary document containing the exact \(10!,48,36,4\) counterexample can be identified, and separately records the hypotheses of each reusable coprime lemma.
Acceptance should require an author, date, stable document link, page or theorem location, and a line-by-line comparison with the normalized quantifiers. Falsification would be either a source showing a different equation or a hidden coprimality/positivity assumption that makes the cited result inapplicable. Stop after one bounded literature pass, or immediately when the primary provenance is located or confirmed unavailable; do not infer a stronger status from the catalogue label. A later formalization may target the finite witness and the elementary coprime interface, but any theorem still containing sorry must remain visibly incomplete.
Sources and claim boundaries
The historical primary evidence used here is the 1937 Erdős–Obláth article and scan, especially its pages 241–254. The maintained status and wording are from erdosproblems.com/399; the frozen catalogue and formal hashes are linked above. The issue-specific original SOURCE RESULT/1 citation is authenticated original RESULT/1.
This article distinguishes four layers: the displayed integer identity is an exact finite check; the 1937 paper is established external mathematics under coprimality and its stated ranges; the Barfield attribution is an unverified external provenance report; and the remaining unrestricted classification question is not supplied here. V2 ledger policy, rather than this article, decides accepted coverage. No claim of independent mathematical certification, publication, completeness of the literature search, or Lean replay is made.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1669 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
- GCL triage: OUTSIDE_STRICT_OPEN_TRIAGE.
- 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.