Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated

Problem 383

../

claims/: The 0 claim pages of Problem 383, one per claimant's result; the problem's standing derives from them.


Statement. Is it true that for every kk there are infinitely many primes pp such that the largest prime divisor of

∏0≤i≤k(p2+i)\prod_{0\leq i\leq k}(p^2+i)

is pp?

Status. Open, the site's label (OPEN; the site marks the problem as not decidable by a finite computation). Its proof-claims tab carries one partial proof claim, submitted 2026-07-25 and without response which settles no case of the question and gets no claim page; the Current assessment records it with the reason.

Source. erdosproblems.com/383, accessed 2026-09-04. Cite as: T. F. Bloom, Erdős Problem #383, https://www.erdosproblems.com/383.

Formalization. Statement in formal-conjectures.

Current assessment

Open; no result decides a single case. The site formulation above asks, for every kk, for infinitely many primes pp such that pp is the largest prime factor of ∏0≤i≤k(p2+i)\prod_{0\le i\le k}(p^2+i), that is, every p2+ip^2+i with 1≤i≤k1\le i\le k has all its prime factors below pp. The site's commentary notes that a positive answer would answer the second part of Problem 382, and that the heuristic probability 1−log⁡21-\log2 that an integer nn has no prime factor of size at least n1/2n^{1/2} predicts the answer yes. No source proves the statement for any k≥1k\ge1, and the site labels the problem OPEN.

Proof claim without a page. The site's proof-claims tab lists a partial proof claim by Rafik Zeraoulia (using OpenAI GPT-5.6 Thinking, as the tab names the system), submitted 2026-07-25, with the write-up A positive-proportion smoothness bound and computational results for Erdős Problem 383 (Zenodo record 21544693, 2026-07-25, CC BY 4.0). It asserts that for every fixed k≥1k\ge1 and ε>0\varepsilon>0 a positive proportion of primes pp satisfy P+(∏i=0k(p2+i))≤p2−1/(2k)+εP^+(\prod_{i=0}^k(p^2+i))\le p^{2-1/(2k)+\varepsilon}, by applying a theorem of Dartyge, Martin and Tenenbaum on smooth values of polynomials to ∏i=1k(X2+i)\prod_{i=1}^k(X^2+i) and restricting to primes near xx, and it reports an exhaustive computation over the primes p≤108p\le10^8: the prime p=93 609 881p=93\,609\,881 is the largest prime factor of the product for k=12k=12, the only such prime below 10810^8, and no prime in that range works for k=13k=13. The claim gets no page because it settles no instance of the problem: the exponent 2−1/(2k)+ε2-1/(2k)+\varepsilon is at least 3/23/2 for every k≥1k\ge1, where the question needs exponent 11, and the computation exhibits single primes, not infinitely many; the write-up itself says that the bound does not resolve the exponent-11 problem and that the computation proves no infinitude.

Forum remarks. In the problem's discussion thread (two comments as of 2026-10-07), a comment of 2026-06-21 reports that the prime p=9 188 057p=9\,188\,057 is the largest prime factor of the product for every k≤10k\le10 and fails at k=11k=11, where p2+11p^2+11 has the prime factor 1 407 006 523 921>p1\,407\,006\,523\,921>p, and gives the counts of primes p≤107p\le10^7 satisfying the condition through each kk (181 281181\,281 for k=1k=1 down to one for k=10k=10 and none for $11\le k\le15$); the commenter discloses that the computation and comment were prepared with assistance from Codex 5.5 and ChatGPT 5.5 Pro. A comment of 2025-09-08 observes that the first nontrivial case, k=1k=1, would follow from the conjecture that there are infinitely many Newman–Shanks–Williams primes, and asks for an unconditional proof; none is recorded. Neither remark settles an instance.

Formalization and search scope. The formal-conjectures statement file ErdosProblems/383.lean, as of its last change on 18 September 2026, states erdos_383 as answer(sorry) ↔ ∀ k, {p : ℕ | p.Prime ∧ Nat.maxPrimeFac (∏ i ∈ Finset.Icc 0 k, (p ^ 2 + i)) = p}.Infinite, category research open, with no formal_proof attribute. Search scope: the site's page, discussion thread and proof-claims tab, the Zenodo record of the proof claim, the formal-conjectures file and its commit history, and the community database, which lists the problem as open and formalized as of its last update on 2025-08-31, without dating when either state was set.