Wiki
Wiki

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

Updated


Claim. Theorem 1.1 of Leng, Sah and Sawhney, Improved Bounds for Szemerédi's Theorem, states that for each fixed k≥5k\ge5 there is ck∈(0,1)c_k\in(0,1) with

rk(N)≪Nexp⁡(−(log⁡log⁡N)ck),r_k(N)\ll N\exp\bigl(-(\log\log N)^{c_k}\bigr),

where rk(N)r_k(N) is the largest size of a subset of {1,…,N}\{1,\ldots,N\} with no non-trivial kk-term arithmetic progression. Since the exponential factor tends to zero, the bound gives rk(N)=o(N)r_k(N)=o(N) for every k≥5k\ge5, those instances of Problem 139, with a rate the problem does not ask for. The paper improves Gowers's bound N(log⁡log⁡N)−2−2k+9N(\log\log N)^{-2^{-2^{k+9}}}, the only earlier bound for k≥5k\ge5, by feeding the authors' quasipolynomial inverse theorem for the Gowers Uk+1U^{k+1} norm into the density-increment strategy of Heath-Brown and Szemerédi in the form Green and Tao gave it. The library card is Leng, Sah and Sawhney 2024.

Covers. Every instance k≥5k\ge5 of the statement, rk(N)=o(N)r_k(N)=o(N), which Szemerédi's accepted full claim already settles; the page records the bound's rate, which no claim of this problem requires. Nothing about k=3k=3 or k=4k=4.

Depends on. Nothing in this wiki; the theorem is the paper's own.

Standing. Claimed. The paper is an arXiv preprint with no journal record on its arXiv listing, so refereed is not listed. The site's curator labels the problem proved on Szemerédi's theorem and cites this paper in the commentary only as the best known bound for k≥5k\ge5, which credits the bound and not a settlement of the problem, so no reviewed evidence is listed. The proof is not checked here.