Wiki
Wiki

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

Updated

Pohoata 2022 convex polytopes fewer points

../


Cosmin Pohoata, Dmitrii Zakharov, Convex polytopes from fewer points. arXiv:2208.04878 (2022); Duke Math. J. 174 (2025), no. 3, 449-471, doi:10.1215/00127094-2024-0034. The arXiv record (https://arxiv.org/abs/2208.04878, read 2026-10-02) names the Creative Commons Attribution 4.0 license. The folder-name PDF is arXiv v1 of 9 August 2022; the statements below are v1's, and the journal version was not compared.

Let ES_d(n) be the least N such that any N points in general position in R^d contain n in convex position. Theorem 1.1 proves that for every eps > 0 and n large, any general-position set in R^3 of size at least 2^{eps n} contains n points in convex position, i.e. ES_3(n) = 2^{o(n)}; by Valtr's projection chain ES_d(n) <= ES_{d-1}(n) this gives ES_d(n) = 2^{o(n)} for all d >= 3, in sharp contrast with the planar lower bound ES_2(n) >= 2^{n-2}+1 of Erdos and Szekeres. This disproves the Morris-Soltan prediction that ES_d(n) = Omega(2^{2n/d}) and ES_d(n) = 4 ES_d(n-d) - 3. Theorem 1.2 is a quantitative positive-fraction statement in R^3: for all n >= n_0, any general-position X with |X| >= ES_3(8n) contains a collection X_1,...,X_n in convex position, meaning that each conv(X_i) is disjoint from the convex hull of the union of the other X_j, with every |X_i| at least |X|/ES_3(8n)^8; Theorem 1.3 transfers this to all d >= 3 and n >= 3 with density eps_n = (1/2)^{o(n)}, so that 1/eps_n is subexponential in n, unlike the planar bound eps_n = n 2^{-32n} of Por and Valtr. The tools are the Erdos-Szekeres cups-vs-caps theorem plus higher-dimensional structural arguments. The paper bears on Erdos problem 651, which asks whether ES_k(n) > (1+c_k)^n for some constant c_k > 0; since ES_k(n) = 2^{o(n)} for every k >= 3, no such constant exists, and the answer is no in every dimension k >= 3.

Source: https://arxiv.org/abs/2208.04878.

Bears on. #651

Results to transcribe.

  • Theorem 1.1: ES_3(n) = 2^{o(n)}: for any eps > 0 and n large, any general-position X in R^3 with |X| >= 2^{eps n} contains n points in convex position.
  • Corollary of (1): Combined with ES_d(n) <= ES_{d-1}(n), Theorem 1.1 gives ES_d(n) = 2^{o(n)} for all d >= 3, disproving the Morris-Soltan conjecture ES_d(n) = Omega(2^{2n/d}).
  • Theorem 1.2: Quantitative positive-fraction Erdos-Szekeres in R^3: for all n >= n_0, any general-position X with |X| >= ES_3(8n) contains a collection X_1,...,X_n in convex position (each conv(X_i) disjoint from the convex hull of the union of the others) with each |X_i| >= |X|/ES_3(8n)^8.
  • Theorem 1.3: For d >= 3 and n >= 3 there is eps_n = (1/2)^{o(n)} such that any general-position X in R^d with |X| >= ES_3(8n) contains a collection of n subsets in convex position, in the sense of Theorem 1.2, each of size at least eps_n |X|.