Erdős Problem #383
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
Erdős Problem 383 asks whether a striking pattern can persist for arbitrarily long finite stretches. Let \(P^+(n)\) mean the largest prime factor of a positive integer \(n\). For every natural number \(k\), are there infinitely many primes \(p\) for which
The factor with index \(0\) is \(p^2\), so the product already has \(p\) as a prime factor. The equality therefore says that every prime factor of every later number \(p^2+1,p^2+2,\ldots,p^2+k\) is at most \(p\). The difficulty is not finding an isolated example for one value of \(k\); it is proving that, for each fixed length, infinitely many prime starting points work.
The quantifier order matters. The prime \(p\) may depend on \(k\), but one needs infinitely many choices of \(p\) after \(k\) has been fixed. The question does not ask for one prime that works for all \(k\), nor does a finite search establish the claim. The live problem page also records a heuristic suggesting a positive answer, and notes a connection with the second part of Problem 382. Those are motivation, not proof.
A small example
A finite calculation gives a useful scale without pretending to settle the infinitary question. The live discussion reports the prime \(p=9,188,057\). I checked its displayed factorizations by exact integer multiplication and deterministic primality tests. The largest prime factor of each term \(p^2+i\) for \(0\leq i\leq 10\) is as follows:
| \(i\) | largest prime factor of \(p^2+i\) |
|---|---|
| 0 | 9,188,057 |
| 1 | 43,261 |
| 2 | 166,273 |
| 3 | 574,003 |
| 4 | 496,229 |
| 5 | 4,027 |
| 6 | 166,643 |
| 7 | 228,299 |
| 8 | 349,913 |
| 9 | 305,441 |
| 10 | 2,185,369 |
For example, the last entry is verified by the exact equality
and \(2,185,369<9,188,057\). The analogous factorizations for the preceding ten values have all prime factors no larger than \(p\), so this one prime witnesses the finite conditions through \(k=10\). It fails immediately at the next term: the displayed factorization gives
whose final factor exceeds \(p\). This verifies a finite example, not infinitely many primes for any fixed \(k\); it also says nothing about whether other primes work beyond \(10\).
What is known
The primary source is P. Erdős and R. L. Graham, Old and New Problems and Results in Combinatorial Number Theory (1980). On printed page 73, in the discussion of greatest prime factors of products of consecutive integers, they ask precisely whether, for every \(k\), infinitely many values of \(p\) satisfy \(P^+(\prod_{i=0}^{k}(p^2+i))=p\). The source poses the question; the audited source passage does not supply a proof or a theorem answering it.
The current Erdős Problems page, accessed 2026-10-13, lists two ordinary discussion comments and one proof claim. The proof-claim page explicitly warns that appearance is no guarantee of correctness. Its summary, submitted 2026-07-25, reports an alleged partial bound for fixed \(k\geq1\) and every \(\varepsilon>0\) of the form \(P^+(\prod_{i=0}^{k}(p^2+i))\leq p^{2-1/(2k)+\varepsilon}\) for infinitely many primes, based on a theorem of Dartyge, Martin, and Tenenbaum. Even if that report is later verified with all hypotheses and the index convention aligned, its exponent is not \(1\) and so it does not establish Problem 383.
The page also records a separate finite search up to \(10^8\) and a larger reported witness. Those reports are useful leads, but they are not external proof and were not treated as certified mathematical results here.
GCL's current state
The frozen catalogue evidence is teorth/erdosproblems at commit 536408716724a2402bdfa7c1d764bf83d64a8c89, file data/problems.yaml, entry number: "383", blob 4e60263a80af6207aca494d2e1388e44fa372889. At that revision the informal status is open, last updated 2025-08-31; the formal status is unformalized, while statement-formalized metadata is yes.
The pinned FormalConjectures file is at commit 85f863718beeec7b58a3a1926ee92e3472bc2020, blob 323b6098804bbad09bdf2a27f9fb7a689f53a2d4. It states an equivalence involving the same universal quantifier, prime condition, interval product, and Nat.maxPrimeFac, but its theorem is by sorry. Thus it is a formal statement interface, not a formal proof or semantic certification. V2 ledger adjudication, not this article or the catalogue flag alone, determines accepted coverage.
What remains
The smallest substantive gap is an unconditional infinitude theorem at the exact threshold \(p\): for each fixed natural \(k\), prove that infinitely many primes have no prime divisor larger than \(p\) anywhere in the product. A bounded successor should first normalize the original product against the partial-result convention \(\prod_{i=1}^{k}(p^2+i)\), then recover the exact hypotheses and conclusion of the cited Dartyge–Martin–Tenenbaum theorem. It should check the restriction to primes near the size parameter, the dependence on \(k\) and \(\varepsilon\), and whether the claimed bound is genuinely uniform over infinitely many primes.
Acceptance should be either a source-quoted theorem with page or theorem identifiers and a line-by-line specialization proving the exact Problem 383 conclusion, or a rigorous obstruction showing why the proposed specialization cannot reach exponent \(1\). A finite factorization may be accepted as a reproducible computational certificate only for its stated range. Stop when the source text, parameter translation, or prime-selection step is unavailable; do not turn an unverified proof claim into a solved status.
Sources and claim boundaries
The primary reference is the 1980 Erdős–Graham PDF, p. 73, bibliographically identified on the live page as Old and New Problems and Results in Combinatorial Number Theory, Monographies de L'Enseignement Mathématique (1980), MR 0592420. The original MATH-PROGRAMME issue supplies the frozen hashes, task boundary, and protected triage context. The live page and its discussion supply dated but unverified reports; the proof-claim page is not an adjudication record.
Accordingly, this article separates the established external statement, an independently checked finite calculation, an unverified external partial-result report, and a Lean theorem source containing sorry. It does not supply mathematical certification, claim that the problem is solved, or infer current accepted coverage from a V2 ledger alone. Original SOURCE RESULT/1 citation: authenticated original RESULT/1.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1653 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
- GCL triage: PROTECTED_READY_BACKLOG.
- 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.