Wiki
Wiki

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

Updated

Ruzsa 1985 note additive bases integers

../

conjecture_1: Ruzsa and Turjányi's modified form of the Erdős-Graham conjecture, in which the twofold sumset is counted up to 2x against the basis counted up to x; the paper proves the threefold analogue and leaves this open.

theorem_1: Ruzsa and Turjányi's construction, for every order h at least 3, of a basis of density zero whose (h-1)-fold sumset has counting function within a constant factor of the basis's own along a sequence tending to infinity; the case h = 3 answers Problem 337 in the negative.

theorem_2: Ruzsa and Turjányi's theorem that for every basis of density zero the number of threefold sums below 3x is eventually larger than any constant multiple of the number of elements below x; it is deduced from Theorem 3.

theorem_3: Ruzsa and Turjányi's bound on the iterated sumsets of a finite set of integers in terms of the doubling-type constant of its threefold sumset, proved from Ruzsa's 1976 difference-set inequality.


I. Z. Ruzsa and S. Turjányi, A note on additive bases of integers, Publ. Math. Debrecen 32 (1985), 101--104. Received November 28, 1983.

The copy read for this card is an image-only scan of the four printed pages (physical PDF p. nn is printed p. 100+n100+n; the first page carries no page number, the others carry the journal's running heads). Its metadata title is Pub_Mat_1985_32__1_2_13. It has no text layer; the statements below were read on the page images of pp. 101--102, with an OCR pass used only to locate them. Provenance: downloaded in September 2026 from a URL that was not recorded; 861,413 bytes. No notice is printed on the scanned pages; the journal's site (https://publi.math.unideb.hu/, read 2026-10-02) states on its page for authors that "the authors agree to transfer the copyright to the publisher" and that "the version published in PMD cannot be uploaded to any repository", every other right reserved.

Read status: claims checked. Theorems 1, 2 and 3 and Conjectures 1 and 2 were read clause by clause on the page images; none of the proofs (pp. 101--103) was checked.

Contents

Notation (p. 101): A±B={a±b}A\pm B=\{a\pm b\}, kAkA is the kk-fold sumset, A(x)A(x) and Ak(x)A_k(x) count the elements of AA and of kAkA below xx. A basis of order hh is a set of natural numbers whose sums of at most hh elements include every sufficiently large integer.

  • Introduction (p. 101): records the conjecture of Erdős and Graham (1980) that A2(x)/A(x)→∞A_2(x)/A(x)\to\infty for every basis AA with A(x)=o(x)A(x)=o(x), and Turjányi's (1981) counterexamples: for each k≥4k\ge4, a basis of order kk with lim inf⁡A2(x)/A(x)<∞\liminf A_2(x)/A(x)<\infty.
  • Theorem 1 (p. 101): for each h≥3h\ge3 some basis AA of order hh has density zero, A(x)=o(x)A(x)=o(x), and Ah−1(x)≤CA(x)A_{h-1}(x)\le CA(x) for a constant CC and arbitrarily large xx, that is, lim inf⁡Ah−1(x)/A(x)<∞\liminf A_{h-1}(x)/A(x)<\infty. The construction (pp. 101--102) adds to a basis BB of order hh with B(x)=O(x1/h)B(x)=O(x^{1/h}) (printed o(x1/h)o(x^{1/h}), a misprint: a basis of order hh has B(x)≫x1/hB(x)\gg x^{1/h}, and the proof needs only the OO bound) the integer intervals [dn−dn r,dn][d_n-d_n^{\,r},d_n] for a rapidly increasing sequence dnd_n and an exponent r∈(1−1/h,1)r\in(1-1/h,1).
  • Conjecture 1 (p. 102), as posed: "If AA is a basis and A(x)=o(x)A(x)=o(x), then A2(2x)/A(x)→∞A_2(2x)/A(x)\to\infty." The authors motivate it by the example: A(x)A(x) jumps in a short interval, but sums of two numbers near xx lie near 2x2x.
  • Theorem 2 (p. 102): every basis AA with A(x)=o(x)A(x)=o(x) has A3(3x)/A(x)→∞A_3(3x)/A(x)\to\infty. It is deduced from Theorem 3 on p. 103.
  • Theorem 3 (p. 102): a finite set XX of nn integers with ∣3X∣=sn|3X|=sn satisfies ∣kX∣≤skn|kX|\le s^k n for every kk. The proof (p. 103) uses the inequality ∣X∣ ∣Y−Z∣≤∣X−Y∣ ∣X−Z∣|X|\,|Y-Z|\le|X-Y|\,|X-Z| of Ruzsa (1976).
  • Conjecture 2 (p. 102; recorded on the page of Conjecture 1): ∣X∣=n|X|=n and ∣2X∣=sn|2X|=sn imply ∣kX∣≤f(s,k)n|kX|\le f(s,k)n for a function of ss and kk alone; the authors expect it to follow from Freiman's theorem with f(s,k)=exp⁡(cks)f(s,k)=\exp(cks) and guess the true order scks^{ck}. It would imply Conjecture 1 in the same way.

Compiled scope

The five statements above were checked on the page images; the proofs of Theorems 1, 2 and 3 were not read beyond the pointers given. Nothing here is independently reviewed.

Bears on. #337: the problem asks whether every basis AA with A(x)=o(x)A(x)=o(x) has ∣(A+A)∩[1,N]∣/∣A∩[1,N]∣→∞|(A+A)\cap[1,N]|/|A\cap[1,N]|\to\infty; Theorem 1 with h=3h=3 gives a basis of order 33 with A(x)=o(x)A(x)=o(x) and lim inf⁡A2(x)/A(x)<∞\liminf A_2(x)/A(x)<\infty, so the ratio does not tend to infinity for it, and the introduction records Turjányi's earlier counterexamples of every order k≥4k\ge4. Conjecture 1, A2(2x)/A(x)→∞A_2(2x)/A(x)\to\infty, is the paper's "modified form" (p. 101) of the conjecture, which the paper leaves open; Theorem 2, A3(3x)/A(x)→∞A_3(3x)/A(x)\to\infty, is the threefold variant it proves, deduced from Theorem 3. Neither decides the problem as stated.

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