Wiki
Wiki

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

Updated


Claim. For a fixed translation-invariant linear relation, the theorem of J. Komlós, M. Sulyok and E. Szemerédi, Linear problems in combinatorial number theory, Acta Math. Acad. Sci. Hungar. 26 (1975), no. 1--2, 113--121, compares the largest relation-free subset guaranteed in every nn-element set of integers, g(n)g(n), with the largest relation-free subset of {1,…,n}\{1,\ldots,n\}, f(n)f(n): for all sufficiently large nn, g(n)≥f(n)/(8α6)g(n)\ge f(n)/(8\alpha^6), where α\alpha is the largest row ℓ1\ell^1-norm of the relation's coefficients. Applied to the Sidon relation a+b=c+da+b=c+d, with α=4\alpha=4 and f(n)=(1+o(1))n1/2f(n)=(1+o(1))n^{1/2}, it shows that every set of NN integers contains a Sidon subset of size at least cN1/2cN^{1/2}. Points of the line have pairwise distinct distances exactly when they form a Sidon set, so in the notation of Problem 1088 f1(n)≤Cn2f_1(n)\le Cn^2 for an absolute constant CC. The paper states its theorem for integers; the transfer to finite sets of reals is recorded on [[problems/additive_bases/E0530/claims/1975_01_01_komlos_sulyok_szemeredi|the same theorem's claim page for Problem 530]], the problem the site names for d=1d=1.

Covers. The upper bound f1(n)≪n2f_1(n)\ll n^2. With the Erdős–Turán bound that a Sidon subset of {1,…,N}\{1,\ldots,N\} has at most (1+o(1))N1/2(1+o(1))N^{1/2} elements, which gives f1(n)≥(1+o(1))n2f_1(n)\ge(1+o(1))n^2, it fixes the order n2n^2 of the instance d=1d=1. The constant is not determined.

Depends on. [[problems/additive_bases/E0530/claims/1975_01_01_komlos_sulyok_szemeredi|The claim page for Problem 530]], for the transfer of the integer theorem to finite sets of reals.

Acceptance. Refereed: the paper is a journal publication in Acta Mathematica Academiae Scientiarum Hungaricae, volume 26, issue 1--2 (1975). The site's remarks state f1(n)≍n2f_1(n)\asymp n^2 through Problem 530, but the site labels the problem OPEN, so the remark is not reviewed evidence. The page is dated to the first day of the publication year, the citation giving no month or day.