Wiki
Wiki

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

Updated


Claim. For every partition N=A⊔B\mathbb N=A\sqcup B,

max⁡(δ‾(S(A)),δ‾(S(B)))≥2+34≈0.93301,\max\bigl(\overline{\delta}(S(A)),\overline{\delta}(S(B))\bigr)\ge\frac{2+\sqrt3}4\approx0.93301,

where S(X)S(X) is the set of sums of finitely many distinct elements of XX and δ‾\overline{\delta} the upper logarithmic density, and the partition into the nn with ⌊log⁡blog⁡n⌋\lfloor\log_b\log n\rfloor even and the nn with it odd, b=2+3b=2+\sqrt3, has maximum equal to (2+3)/4(2+\sqrt3)/4. So the least possible value of the maximum, the quantity Problem 1211 asks for, is c2=(2+3)/4c_2=(2+\sqrt3)/4. This is the case r=2r=2 of Theorem 1 of the library's source card, which bounds the analogous constant crc_r for every number r≥2r\ge2 of classes and proves the bound tight for r=2r=2. The upper bound generalizes Erdős's block coloring; the lower bound rests on the paper's Lemma 2, a statement about subset sums of a class of a partition of [N,eN)[N,eN) filling an interval [CN,C′N2][CN,C'N^2], turned into the density bound through an auxiliary coloring and the Brouwer fixed-point theorem. The site's question asks how large the maximum must be, and the theorem answers it with the exact value, so the claim settles the whole question; the tightness of the bound for r≥3r\ge3 is the paper's Conjecture 10 and is not part of this problem.

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

Acceptance. Reviewed: the site's curator, T. F. Bloom, labels the problem SOLVED and credits Conlon, Fox and Pham in the problem's commentary with the value c=(2+3)/4c=(2+\sqrt3)/4 and with the coloring that attains it (page last edited 8 April 2026); the curator is independent of the authors, and the community database records the problem solved (last updated 4 April 2026). Refereed: Mathematika 68 (2022), no. 4, 1292--1301, published online 10 October 2022 (per the Crossref record). The text cited is arXiv:2105.15195v3 (22 September 2022); the journal text was not compared, so every locator is a preprint page. The search (arXiv, Crossref, OpenAlex) found no dispute of the theorem and no second determination of c2c_2.

Read depth. Theorem 1, the definitions and the upper-bound argument on pp. 1--2 of the preprint were checked, as were the statements of Lemma 2, Theorem 3, Remark 5 and Conjecture 10; the lower-bound proof (Section 3, pp. 4--8) was not checked, and nothing is independently reviewed in this corpus. The problem page checks by an elementary block count that Erdős's own example has both densities equal to 14/1514/15, as the paper states. The page is dated by the first arXiv posting, v1 of 31 May 2021.