Wiki
Wiki

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

Updated


Claim. Theorem 2.1 of the note "An improved lower bound on Erdős Problem #962": for every fixed cc with 0<c<1/20<c<1/\sqrt2 there is n0(c)n_0(c) such that for all n≥n0(c)n\ge n_0(c), k(n)≥⌊exp⁡(clog⁡nlog⁡log⁡n)⌋k(n)\ge\lfloor\exp(c\sqrt{\log n\log\log n})\rfloor. Its Theorem 1.2 is the form k(n)≥exp⁡((1/2−o(1))log⁡nlog⁡log⁡n)k(n)\ge\exp((1/\sqrt2-o(1))\sqrt{\log n\log\log n}). The proof is a pigeonhole lemma (if fewer than ⌊n/y⌋\lfloor n/y\rfloor integers up to nn are yy-smooth, some block of yy consecutive integers below nn has no yy-smooth member) with the de Bruijn--Hildebrand estimate Ψ(n,y)=nρ(u)(1+o(1))\Psi(n,y)=n\rho(u)(1+o(1)) and ρ(u)=u−u+o(u)\rho(u)=u^{-u+o(u)}, cited from Granville's 2008 survey. The same bound follows from Erdős's 1976 display (6) through the inverse k(n)=max⁡{k:nk≤n}k(n)=\max\{k:n_k\le n\}, as Erdős 1976 records; the site credits Tang separately with the explicit constant.

Covers. The lower bound on k(n)k(n) only; neither question of Problem 962 is settled by it.

Claimant and postings. Quanyu Tang posted the note, a PDF in the repository QuanyuTang/erdos-problem-962, to the site's thread on 28 December 2025; the PDF's metadata dates it 28 December 2025 (UTC). The formal-conjectures file for the problem states the bound as the variant tang_lower_bound, with a sorry body.

Depends on. No page of this wiki.

Standing. Claimed. The note has no refereed version, and the site's commentary on a problem it labels OPEN is not acceptance.