Wiki
Wiki

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

Updated


Source. Remark 1.4, p. 3, with its proof under "Optimizing the error term" at the end of Section 3, pp. 14-16, including Proposition 3.8 (pp. 14-15), of A. F. Holmsen, H. N. Mojarrad, J. Pach and G. Tardos, Two extensions of the Erdős-Szekeres problem, J. Eur. Math. Soc. 22 (2020), 3981-3995, arXiv:1710.11415; read in arXiv:1710.11415v3 (3 August 2020), the edition named on the source card.

Read depth. Claims checked: the statement was read clause by clause on the printed page; the supporting calculation (pp. 14-16) was read for structure only. Nothing here is independently reviewed.

Statement

With b(n)b(n) the pseudo-configuration function of Theorem 1.3, Remark 1.4 (p. 3) states that a less wasteful calculation, given at the end of the paper, yields

b(n)≤2n+(823+o(1))nlog⁡n.b(n)\le 2^{n+\left(\frac{8\sqrt2}{3}+o(1)\right)\sqrt{n\log n}}.

The paper states the bound for b(n)b(n); since b(n)≥e(n)b(n)\ge e(n) (p. 3), it bounds the Erdős-Szekeres function e(n)e(n) of point sets in general position in the plane in the same way. The paper writes log⁡\log without a base; the proof of Theorem 1.3 bounds products of binomial coefficients by powers of 22 with exponent 2nlog⁡n2n\log n (p. 14), which reads as base 22.

Proof pointer

Proposition 3.8 (pp. 14-15) refines Theorem 2.4: for an integer k≥3k\ge3 and a pseudo-configuration with ∣P∣=N≥2(1+o(1))4k|P|=N\ge 2^{(1+o(1))4k}, either some kk-subset in convex position has spike sizes with product at least 2−83k2Nk2^{-\frac83k^2}N^k, or some 2k2k-subset in convex position has spike sizes with product at least 2−403k2−o(k2)N2k2^{-\frac{40}{3}k^2-o(k^2)}N^{2k}. With the sharper binomial estimate (3.9), ∏i(ci+di−2ci−1)<(ek)2n\prod_i\binom{c_i+d_i-2}{c_i-1}<(ek)^{2n} (p. 15), the argument of Theorem 1.3 is run in each case, and kk is taken to be the least even integer at least nlog⁡n/(22)\sqrt{n\log n}/(2\sqrt2) (p. 16).

Dependencies

Theorem 1.3, Theorem 2.4 and Proposition 3.8 of the same paper.

Bears on

  • Problem 107: through b(n)≥e(n)b(n)\ge e(n) this makes the constant in the error term of the upper bound on the problem's f(n)=e(n)f(n)=e(n) explicit. It is an upper bound only and settles no value of f(n)f(n).