Wiki
Wiki

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

Updated

Tikhomirov 2024 remark ramsey number hypercube

../

corollary_1_2: Tikhomirov's upper bound on the Ramsey number of the hypercube: for all large n, r(Q_n) is at most 2^{2n-cn+1} + 2 with the universal constant c of Theorem 1.1, for which 0.03656 is admissible.

theorem_1_1: The embedding theorem behind the hypercube Ramsey bound: for large n, every bipartite graph with both parts of size at least 2^{2n-cn} and at least half of all cross pairs as edges contains the n-cube; c = 0.03656 is admissible.


K. Tikhomirov, A remark on the Ramsey number of the hypercube, European J. Combin. 120 (2024), 103954, doi:10.1016/j.ejc.2024.103954 (Elsevier; the Crossref record, dates the issue August 2024 and the record's creation 25 March 2024). Preprint arXiv:2208.14568 (v1 30 August 2022, v3 2 March 2024; the arXiv listing carries no journal reference).

The copy read for this card is arXiv:2208.14568v3 [math.CO] 2 Mar 2024, 24 pages with a text layer; its page numbers are the preprint's, and the journal text was not compared. Pages 1--3 were read on the page images. The arXiv record names arXiv's non-exclusive distribution license (arXiv:2208.14568), every other right reserved.

Read status: claims checked for Theorem 1.1, the Remark after it, Corollary 1.2 and Remark 1.3 (pp. 2--3), read clause by clause on the page images, and for the introduction's restatement of the prior bound (p. 1); the proof (Sections 2--6 and the appendix) was not read.

The paper improves the exponent in the upper bound for the Ramsey number of the hypercube QnQ_n (the graph on {−1,1}n\{-1,1\}^n whose edges are the geometric edges). The introduction (p. 1) recalls the Burr--Erdős conjecture that r(Qn)=O(2n)r(Q_n)=O(2^n), the improvements of the trivial bound r(Qn)≤r(K2n)r(Q_n)\le r(K_{2^n}) by Beck, Graham--Rödl--Ruciński, Shi, Fox--Sudakov, Conlon--Fox--Sudakov and Lee, and quotes the previous best bound as "[4, Theorem 4.1]", its [4] being the same authors' 2016 Short proofs of some extremal results II: for every bipartite graph HH on mm vertices with maximum degree dd, r(H)≤2d+6mr(H)\le2^{d+6}m, which gives r(Qn)=O(22n)r(Q_n)=O(2^{2n}); it also notes (from its [8]) that for every d≥2d\ge2 and m≥d+1m\ge d+1 some bipartite graph HH on mm vertices of maximum degree at most dd has r(H)≥2c′dmr(H)\ge2^{c'd}m, with c′>0c'>0 a constant, so any proof of the conjecture must use more of the cube than its degree and order. Pages 2--3 explain why the dependent random choice scheme (I)--(II) alone stalls around 22n2^{2n} (a random ambient graph Γ\Gamma whose common neighborhoods of nn-tuples concentrate on small sets) and how a randomized "block" embedding of the facets of the cube gets past it. The main result, Theorem 1.1, embeds QnQ_n, for n≥n0n\ge n_0, into every bipartite graph with both parts of size at least 22n−cn2^{2n-cn} and edge density at least 1/21/2, for universal constants n0,c>0n_0,c>0; the Remark after it says the proof allows c=0.03656c=0.03656 for large nn, and Corollary 1.2 deduces r(Qn)≤22n−cn+1+2r(Q_n)\le2^{2n-cn+1}+2 for n≥n0n\ge n_0 by the standard majority-color equipartition (Remark 1.3). The paper does not print the exponent 2−c2-c numerically; 2−0.03656=1.963442-0.03656=1.96344 is a subtraction made here, so the corollary reads r(Qn)≤21.96344n+1+2r(Q_n)\le2^{1.96344n+1}+2 for large nn. Corollary 3.2 (p. 6, page image) is the embedding statement that the scheme (I)--(II) alone gives, for n≥n3.2(ε)n\ge n_{3.2}(\varepsilon) and bipartite graphs of density α≥ε\alpha\ge\varepsilon with parts of sizes at least 2n+εn2^{n+\varepsilon n} and 2n+εn/αn2^{n+\varepsilon n}/\alpha^n; with Remark 1.3 it yields the weaker r(Qn)≤22n+o(n)r(Q_n)\le2^{2n+o(n)} (p. 7).

Contents

  • Introduction (p. 1): the definitions; the Burr--Erdős conjecture r(Qn)=O(2n)r(Q_n)=O(2^n); the prior bound restated as Theorem [4, Theorem 4.1], r(H)≤2d+6mr(H)\le2^{d+6}m for bipartite HH on mm vertices with maximum degree dd, and Theorem [4, Theorem 4.7], the embedding statement behind it; the lower bound r(H)≥2c′dmr(H)\ge2^{c'd}m for some bipartite graphs of maximum degree dd (from [8]).
  • Pages 2--3: the barrier example Γ\Gamma on 22n−εn2^{2n-\varepsilon n} vertices per part, and the block-embedding idea; the trichotomy (a)--(c) behind the proof and its structural part, Proposition 6.3.
  • Theorem 1.1 (p. 2): universal constants n0,c>0n_0,c>0 such that for every n≥n0n\ge n_0 and every bipartite graph G=(VGup,VGdown,EG)G=(V_G^{up},V_G^{down},E_G) with ∣VGup∣,∣VGdown∣≥22n−cn|V_G^{up}|,|V_G^{down}|\ge2^{2n-cn} and ∣EG∣≥12∣VGup∣∣VGdown∣|E_G|\ge\frac12|V_G^{up}||V_G^{down}|, the hypercube QnQ_n can be embedded into GG; Remark: c=0.03656c=0.03656 is admissible for n0n_0 large.
  • Corollary 1.2 (p. 2): for n≥n0n\ge n_0, r(Qn)≤22n−cn+1+2r(Q_n)\le2^{2n-cn+1}+2; Remark 1.3 (p. 3) gives the deduction from Theorem 1.1.
  • Corollary 3.2 (p. 6, page image): for every ε∈(0,1)\varepsilon\in(0,1) there is n3.2(ε)n_{3.2}(\varepsilon) such that for n≥n3.2(ε)n\ge n_{3.2}(\varepsilon), QnQ_n embeds into every bipartite graph of density at least α∈[ε,1]\alpha\in[\varepsilon,1] with one part of size at least 2n+εn2^{n+\varepsilon n} and the other of size at least 2n+εn/αn2^{n+\varepsilon n}/\alpha^n; with Remark 1.3 it gives r(Qn)≤22n+o(n)r(Q_n)\le2^{2n+o(n)} (p. 7), the bound of the scheme (I)--(II) alone.

Compiled scope

Pages 1--3 were read on the page images; of pp. 4--24 (the proof, Proposition 6.3 and Appendix A), only the notation of p. 5, the section headings, the statement of Corollary 3.2 with the sentence after it (pp. 6--7) and the reference list (p. 22) were read, on the page images. No proof was checked and nothing here is independently reviewed.

Source: https://arxiv.org/abs/2208.14568.

Bears on. #181: Corollary 1.2 with the Remark's c=0.03656c=0.03656 is an upper bound on r(Qn)r(Q_n) of order 2(2−c)n2^{(2-c)n} for all large nn, short of the linear bound r(Qn)≤C 2nr(Q_n)\le C\,2^n the problem asks for; the paper presents it as improving the previous best bound O(22n)O(2^{2n}) (abstract, p. 1), and the site's commentary cites it as "R(Qn)≪2(2−c)nR(Q_n)\ll2^{(2-c)n} ... c≈0.03656c\approx0.03656 is permissible"; the introduction's quotation of [4, Theorem 4.1] is the page's second-hand record of that prior bound, and its [4] is Conlon, Fox and Sudakov's Short proofs of some extremal results II (J. Combin. Theory Ser. B 121 (2016), 173--196), not their 2012 On two problems in graph Ramsey theory: the reference list (p. 22, page image) names the 2016 paper, and Theorem 4.1 with Corollary 4.2, r(Qd)≤22d+6r(Q_d)\le2^{2d+6}, stands on p. 7 of the arXiv preprint filed as conlon_2016_short_proofs_extremal_results_ii (page image); that paper's journal text is not held.

No file of this source is held: no license on record permits its redistribution, and the card cites the edition it names above.