Skip to content

Erdős Problem #405

Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: proved (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 405 asks about an exponential Diophantine equation mixing a factorial and a power. In the original positive-integer formulation, choose positive integers \(a\) and \(k\), and let \(p\) be an odd prime. The question is whether

\[ (p-1)! + a^{p-1} = p^k \]

can hold for only finitely many triples \((a,k,p)\). The factorial is the product of all positive integers below \(p\); the right side is a pure power of the same prime \(p\). The restriction to odd primes matters. If \(p=2\), the equation becomes \(1+a=2^k\), and the choice \(a=2^k-1\) gives infinitely many solutions.

The question was recorded by Erdős and Graham in Old and new problems and results in combinatorial number theory (1980, p. 80). The original issue for this audit is MATH-PROGRAMME issue #1675, assignment ERDOS-CATALOGUE-0405-S01.

In other words, the issue is not merely whether examples can be found: it asks for a global finiteness statement across every odd prime and every permitted pair of exponents.

A small example

The three solutions reported by the modern problem entry can be checked with ordinary integer arithmetic:

\[ 2!+1^2=2+1=3=3^1, \]
\[ 2!+5^2=2+25=27=3^3, \]

and

\[ 4!+1^4=24+1=25=5^2. \]

I reproduced these calculations exactly, and an exhaustive finite search over odd primes \(p\leq 11\), \(1\leq a\leq20\), and \(1\leq k\leq10\) returned precisely [(1, 1, 3), (5, 3, 3), (1, 2, 5)]. This is a useful sanity check on ordering and arithmetic, not a proof that no larger solution exists. It also does not check the analytic arguments in the papers.

What is known

The readable primary proof is B. Brindza and P. Erdős, “On some diophantine problems involving powers and factorials,” Journal of the Australian Mathematical Society 51 (1991), 1–7, doi:10.1017/S1446788700033255. Their equation (6), printed p. 3, is the same equation with positive integers \(a,k\) and prime \(p>2\). Theorem 2, printed p. 4, states that an effectively computable absolute constant \(C\) bounds every solution: \(\max\{p,a,k\}<C\). That is a direct proof of finiteness, although it does not itself list the solutions.

The live Erdős Problems #405 entry, accessed 2026-10-13, reports the subsequent complete list and cites K. Yu and D. Liu, “A complete resolution of a problem of Erdős and Graham,” Rocky Mountain Journal of Mathematics 26 (1996), 1235–1244, doi:10.1216/rmjm/1181072047. The authors’ institutional abstract states the same three triples. The Project Euclid full-text endpoint was captcha-blocked during this audit, so the Yu–Liu proof pages were not independently read here; the exact-list attribution is therefore reported as an external source claim, not as a new certification.

There is a small but important normalization issue. The papers say positive integers, while the formal Lean set uses natural numbers for \(a\) and \(k\) and omits explicit positivity. For odd prime \(p\), the equation should force the zero cases away, but that implication is a separate lemma and has not been certified in this audit. Any formal reuse should state whether it is proving the natural-number encoding as written or first proving equivalence with the positive-integer statement.

GCL's current state

The frozen catalogue file is teorth/erdosproblems data/problems.yaml, commit 536408716724a2402bdfa7c1d764bf83d64a8c89; the exact file blob verified for this audit is 4e60263a80af6207aca494d2e1388e44fa372889. Its entry records informal status proved (last update 2025-08-31), formal status Lean (2026-08-24), overall status proved (Lean) (2026-08-24), and formalized yes (2026-09-18). Those are dated catalogue fields, not mathematical certification; V2 ledger policy alone decides accepted coverage.

The pinned FormalConjectures snapshot named by the issue is commit 85f863718beeec7b58a3a1926ee92e3472bc2020 (2026-07-29). It contains no Problem 405 file, so it cannot be cited as a formal statement source. A later commit, FormalConjectures commit a528dea8cb8d8df1987cc5eb16c81b82435f4209, adds the file with blob cb03dcd2c5122117ae92d6ad75821dab52f1bf51. Its two research declarations, erdos_405 and erdos_405.variants.yu_liu, both end in by sorry; the natural-number encoding and the stated three-element set are therefore formal theorem sources with unfilled proofs, not Lean replay evidence.

What remains

A bounded successor should obtain a readable copy of Yu–Liu (1996), record its theorem and page locations, and compare every quantifier and variable convention against Brindza–Erdős equation (6), the live page, and the later Lean file. It should separately prove or document the zero-parameter exclusion needed to align positive integers with the natural-number set. A lightweight formalization target is the finite arithmetic lemma for the three displayed triples plus the implication that any solution has \(a>0\) and \(k>0\); it must not be presented as the global theorem.

The successor stops after the source text, page-level comparison, and bounded lemmas are recorded, or earlier if the paper remains inaccessible or a quantifier mismatch changes the target. No exhaustive Lean build, publication claim, or mathematical certification is implied.

Sources and claim boundaries

The source chain is: original issue #1675; frozen catalogue commit and blob above; live problem page; Brindza–Erdős (1991), whose downloaded primary PDF had SHA-256 0330b2f0be44e394ae80fcb43129f0b7053bf975bbde81231ae684cec885b488; and Yu–Liu (1996), whose bibliographic record and abstract were read but whose full text was inaccessible in this timebox. The original SOURCE RESULT/1 citation is authenticated original RESULT/1.

This article distinguishes an external published proof of finiteness, an externally reported exact solution list, finite arithmetic checks, and a Lean source containing sorry. None is GCL mathematical certification. The current status language preserves those boundaries; acceptance or coverage remains governed by the V2 ledger rather than by this article alone.

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1675 — 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.

Back to catalogue · Compendium introduction