Wiki
Wiki

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

Updated


Source. Theorem 1.7, p. 5, with Theorems 3.1 (p. 7) and 4.2 (p. 10), of Javier Cilleruelo, Imre Z. Ruzsa and Carlos Vinuesa, Generalized Sidon sets, Advances in Mathematics 225 (2010), 2786--2807, arXiv:0909.5024. Labels and pages are those of arXiv:0909.5024v1 (28 Sep 2009), the edition named on the source card.

Read depth. Claims checked: the three statements and the definitions they use were read clause by clause on the page images; the proofs (Sections 3--4, pp. 7--11) were read for structure only. Nothing here is independently reviewed.

Statement

Setting (pp. 1--4). A set AA in a commutative group is a gg-Sidon set if every xx has at most gg representations x=a1+a2x=a_1+a_2 as an ordered pair (a1,a2)∈A2(a_1,a_2)\in A^2 (Definitions 1.1 and 1.2, pp. 1--2). For a finite commutative group GG, αg(G)\alpha_g(G) is the largest size of a gg-Sidon set A⊂GA\subset G, αg(q)=αg(Zq)\alpha_g(q)=\alpha_g(\mathbb Z_q) (Definition 1.6, p. 4), and

αg=lim sup⁡q→∞αg(q)q(p. 4).\alpha_g=\limsup_{q\to\infty}\frac{\alpha_g(q)}{\sqrt q}\qquad\text{(p. 4)}.

The paper notes the obvious bound αg(q)≤gq\alpha_g(q)\le\sqrt{gq} (p. 4).

Theorem 1.7 (p. 5). We have αg=g+O ⁣(g3/10)\alpha_g=\sqrt g+O\!\left(g^{3/10}\right); in particular lim⁡g→∞αg/g=1\lim_{g\to\infty}\alpha_g/\sqrt g=1.

Theorem 3.1 (p. 7). For each kk and every sufficiently large prime p≥p0(k)p\ge p_0(k), there is a set A⊆Zp2A\subseteq\mathbb Z_p^2 with kp−k+1kp-k+1 elements that is a gg-Sidon set for g=⌊k2+2k3/2⌋g=\lfloor k^2+2k^{3/2}\rfloor.

Theorem 4.1 (p. 10). If A⊆Zp2A\subseteq\mathbb Z_p^2 is a gg-Sidon set with ∣A∣=m\lvert A\rvert=m and q=p2sq=p^2s with ss a positive integer, there is a g′g'-Sidon set A′⊆ZqA'\subseteq\mathbb Z_q with ∣A′∣=ms\lvert A'\rvert=ms and g′=g(s+1)g'=g(s+1).

Theorem 4.2 (p. 10). For any positive integers k,sk,s and every sufficiently large prime pp, there is a set A⊆Zp2sA\subseteq\mathbb Z_{p^2s} with (kp−k+1)s(kp-k+1)s elements that is a ⌊k2+2k3/2⌋(s+1)\lfloor k^2+2k^{3/2}\rfloor(s+1)-Sidon set.

Proof pointer

Theorem 3.1 (pp. 7--9) takes the union of the kk parabolas {(x,x2/u):x∈Zp}⊂Zp2\{(x,x^2/u):x\in\mathbb Z_p\}\subset\mathbb Z_p^2 for u=t+1,…,t+ku=t+1,\ldots,t+k. Lemma 3.2 (p. 7), a Legendre-symbol identity for the representation counts of two parabolas, bounds r(x)r(x) by k2k^2 plus a character sum in tt, and an average over tt finds a tt for which that sum is small. Theorem 4.1 (p. 10) maps (a,b)(a,b) to the integers a+cp+bspa+cp+bsp, 0≤c≤s−10\le c\le s-1, modulo p2sp^2s; Theorem 4.2 combines the two. On p. 11 the paper takes g=⌊k2+2k3/2⌋(s+1)g=\lfloor k^2+2k^{3/2}\rfloor(s+1) with k=4s2k=4s^2, so s=Θ(g1/5)s=\Theta(g^{1/5}), and the prime number theorem gives αg/g≥1+O(g−1/5)\alpha_g/\sqrt g\ge1+O(g^{-1/5}) for these gg. The matching upper bound αg≤g\alpha_g\le\sqrt g is the obvious estimate of p. 4, which p. 11 does not restate, and the page does not write out the passage from these values of gg to all gg.

Dependencies

None from the corpus; Weil's bound for character sums (p. 9) and the prime number theorem (p. 11).

Bears on

The theorem concerns finite cyclic groups and bears on no Erdős problem directly. Its construction (Theorem 4.2) feeds the lower bound of Theorem 1.5.