Wiki
Wiki

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

Updated

On partitioning Sidon sets with quasi-independent sets

../


K. J. Harrison and L. Thomas Ramsey, “On partitioning Sidon sets with quasi-independent sets,” Colloquium Mathematicum 69 (1996), no. 1, 117--131, DOI 10.4064/cm-69-1-117-131; the issue is dated 1995 in the print, while Crossref gives 1996. The copy read for this card is the held 15-page publisher PDF, and page numbers below are its PDF pages (printed pp. 117--131). The file's text layer carries no copyright or license line; the publisher's record offers the PDF under the link "Pobierz zgodnie z CC-BY" ("Free download under CC-BY license" on the English site) and names no Creative Commons version or URL (https://www.impan.pl/get/doi/10.4064/cm-69-1-117-131, read 2026-10-02), so the term is the Creative Commons Attribution license with its version unstated; the site footer "Copyright © 2026 by IMPAN. All rights reserved." speaks for the site, not the article.

Terminology

The paper calls a set NN-independent if it has no nonzero integer relation whose coefficients lie in [−N,N][-N,N]. Its 11-independent sets are called quasi-independent, while it reserves dissociate for 22-independence (definition, pp. 1--2). Thus quasi-independence here is exactly dissociation in Problem 774. The introduction states the integer Sidon decomposition problem in this equivalent language and leaves it open.

Random positive examples

Theorem 1 chooses pj=O(log⁡j)p_j=O(\log j) points independently from rapidly dilated progressions

Qj=Mj{1,…,j2},Q_j=M_j\{1,\ldots,j^2\},

where the MjM_j grow fast enough that earlier blocks cannot cancel a nonzero contribution from the latest block. For each fixed NN, an explicit entropy condition W(N,K,λ)<1/2W(N,K,\lambda)<1/2 implies that almost every resulting sequence is a finite union of NN-independent sets, apart from a finite exceptional set which can be split into singletons (Theorem 1 and Remark 1, pp. 2--3).

The proof supplies three useful ingredients.

  • Scale separation. Lemma 2 proves that a subset of the tail union is NN-independent exactly when its intersection with every block is NN-independent.
  • Uniform local control. Lemmas 3--4 count short bounded-coefficient relations in a random block. A union bound and Borel--Cantelli show that, eventually, every subfamily below a fixed proportion of a block is NN-independent.
  • Capacity benchmark. Proposition 6 proves that the largest NN-independent subset of a dilation of {1,…,j}\{1,\ldots,j\} has size asymptotic to log⁡j/log⁡(N+1)\log j/\log(N+1).

For N=1N=1, these constructions are proportionately dissociated sets for which the desired finite partition exists. They are evidence for the positive side, not counterexamples.

Finite determination and block assembly

Let μ(E,m)\mu(E,m) be the least number of mm-independent classes covering EE, or infinity when no finite cover exists.

  • Lemma 11 (pp. 10--11) gives
μ(E,m)=sup⁡{μ(F,m):F⊂E finite}.\mu(E,m)=\sup\{\mu(F,m):F\subset E\text{ finite}\}.

Its compactness argument turns uniformly bounded colorings of finite truncations into a coloring of the whole set.

  • Theorem 7 (pp. 9--12) says that if each mm-independent subset of Z\mathbb Z splits into finitely many nn-independent sets, then the number of classes needed has one bound valid for all such subsets. The contrapositive assembles finite examples of unbounded cover number at rapidly increasing scales.
  • Theorem 8 (pp. 10--13) gives the Sidon analogue: if every Sidon subset of Z∖{0}\mathbb Z\setminus\{0\} has a finite mm-independent cover, then μ(E,m)≤φ(r)\mu(E,m)\le\varphi(r) for every such Sidon set EE with Sidon constant at most rr, for some increasing φ:[1,∞)→Z+\varphi:[1,\infty)\to\mathbb Z^+. Its contrapositive uses a rapidly dilated sup-norm partition to preserve a common Sidon bound while retaining unbounded finite cover numbers.

These results sharpen the construction target for E0774. It is enough to find finite positive sets FtF_t with one uniform proportional quasi-independent extraction constant, uniformly bounded Sidon constants, and μ(Ft,1)→∞\mu(F_t,1)\to\infty. Theorem 8 then assembles them into an infinite Sidon set with no finite quasi-independent cover. The paper does not construct such blocks: its random blocks have a uniformly bounded cover by design.

Bears on. #774