Wiki
Wiki

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

Updated


Statement

Setting (pp. 1-2). A subset SS of an abelian semigroup is a basis (of order two) if every element is s1+s2s_1+s_2 with s1,s2∈Ss_1,s_2\in S. The basis is perfect if each element has exactly one such representation, up to the order of the summands. The representation function counts ordered representations. For a subset CC of an abelian group and an integer n≥1n\ge1, nC={nc:c∈C}nC=\{nc:c\in C\}, so 2G2G is the set of doubles of elements of GG.

Theorem 1 (p. 2, quoted). "Let GG be an infinite abelian group with ∣2G∣=∣G∣|2G|=|G|. (i) If GG is not the direct sum of a group of exponent 3 and the group of order 2, then GG has a perfect basis. (ii) If GG is the direct sum of a group of exponent 3 and the group of order 2, then GG does not have a perfect basis, but has a basis such that every element of GG has at most two representations (distinct under permuting the summands) as a sum of two elements of the basis."

The hypothesis ∣2G∣=∣G∣|2G|=|G| is needed: the paper observes (p. 2) that when GG is infinite with ∣2G∣<∣G∣|2G|<|G|, every basis SS (indeed every subset with ∣S∣=∣G∣|S|=|G|) gives some element ∣G∣|G| representations of the form 2s2s with s∈Ss\in S; this covers every infinite group of exponent 2. The abstract calls the theorem a complete solution of the Erdős-Turán problem for infinite groups. The proofs assume the axiom of choice (p. 3).

Source. Sergei V. Konyagin and Vsevolod F. Lev, The Erdős-Turán problem in infinite groups, arXiv:0901.1649v1 (2009); published in Additive Number Theory, Springer, New York, 2010, 195--202. Labels and pages here are those of arXiv v1: the definitions on p. 1, Theorem 1 on p. 2, Lemmas 1 and 2 on pp. 3-4, the proof on pp. 5-6. The edition read is identified on the source card.

Read depth. Claims checked: the definitions and the statement were read clause by clause on the printed pages. The proof was read but not checked step by step. Nothing here is independently reviewed.

Proof pointer

Pages 5-6, in three parts. For GG of exponent 3, Lemma 1 (p. 3: an infinite abelian group of prime exponent pp is isomorphic to F×F\mathbb F\times\mathbb F for an algebraically closed field F\mathbb F of characteristic pp) reduces to F×F\mathbb F\times\mathbb F, where the parabola {(x,x2):x∈F}\{(x,x^2):x\in\mathbb F\} is a perfect basis, since a representation of (u,v)(u,v) solves a quadratic in xx with one or two roots. For G=F⊕{0,h}G=F\oplus\{0,h\} with FF of exponent 3 and hh of order 2, a perfect basis SS of FF together with its translate h+Sh+S gives at most two representations, and a perfect basis of GG is ruled out by taking the unique representation of hh and doubling. In general the perfect basis is built by transfinite recursion along a well-ordering of GG indexed by the initial ordinal of ∣G∣|G|, adding, at each successor step whose element is not yet represented, a pair s,ts,t with that element as their sum, subject to conditions (a)-(e) on p. 6, of which (b)-(e) keep representations unique. Lemma 2 (p. 4: if 2G2G is infinite and max⁡{∣A∣,∣B∣}<min⁡{∣2G∣,∣3G∣}\max\{|A|,|B|\}<\min\{|2G|,|3G|\}, some s∈Gs\in G has 2s∉A2s\notin A and 3s∉B3s\notin B) supplies the pair when ∣3G∣=∣G∣|3G|=|G|; a coset argument handles 3≤∣3G∣<∣G∣3\le|3G|<|G|; ∣3G∣=1|3G|=1 means GG has exponent 3 (the first part), and ∣3G∣=2|3G|=2 makes GG the direct sum of a group of exponent 3 and the group of order 2, the case excluded in (i).

Dependencies

Lemma 1 (p. 3) and Lemma 2 (p. 4) of the same paper, and the standard facts of linear algebra and set theory listed on p. 3.

Bears on

  • Problem 1192: the problem asks, for each r≥2r\ge2, for a basis A⊂NA\subset\mathbb N of order rr with ∑n≤xfr(n)2≪x\sum_{n\le x}f_r(n)^2\ll x. Theorem 1 concerns infinite abelian groups with ∣2G∣=∣G∣|2G|=|G| and order two only; it is a group analogue of the case r=2r=2 and says nothing about bases of N\mathbb N.