Wiki
Wiki

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

Updated


Source. Theorem 4, p. 7, of J. A. Dias da Silva and Melvyn B. Nathanson, Maximal Sidon sets and matroids, arXiv:math/0504226v1 (2005), as identified on the source card.

Statement

BhB_h-sets and Bh,kB_{h,k}-sets are as defined on pp. 1--2 (see Theorem 3). A matroid M=M(X,I)M=M(X,\mathcal I) (p. 7) is a finite set XX with a collection I\mathcal I of subsets of XX such that ∅∈I\emptyset\in\mathcal I, every subset of a member of I\mathcal I is in I\mathcal I, and whenever A,B∈IA,B\in\mathcal I with ∣A∣<∣B∣|A|<|B| there is b∈B∖Ab\in B\setminus A with A∪{b}∈IA\cup\{b\}\in\mathcal I.

Theorem 4 (p. 7, quoted). "Let h≥2h\geq 2, and let XX be a finite B2h−1,h−1B_{2h-1,h-1}-subset of an abelian group. Let I\mathcal{I} be the collection of BhB_h-sets contained in XX. Then M=M(X,I)M=M(X,\mathcal{I}) is a matroid."

Since all bases (maximal independent sets) of a matroid have the same size, Theorem 4 contains Theorem 3; the paper proves Theorem 3 first and derives Theorem 4 from it. Two consequences follow in Section 4, for XX a B2h−1,h−1B_{2h-1,h-1}-set in an abelian group as those statements put it. Theorem 5 (p. 8): if ℓ\ell is the BhB_h-covering number of XX (the least number of BhB_h-sets whose union is XX), then for every positive integer k≤ℓk\le\ell there is a number nX(k)n_X(k) such that every maximal subset SS of XX with BhB_h-covering number kk has ∣S∣=nX(k)|S|=n_X(k). Theorem 6 (p. 9): let kk be the BhB_h-covering number of XX, let ρj\rho_j, j=1,…,kj=1,\ldots,k, be the largest size of a union of jj BhB_h-subsets of XX, and let μ=(μ1,…,μr)\mu=(\mu_1,\ldots,\mu_r) be a partition of ∣X∣|X| with μ1≥⋯≥μr\mu_1\ge\cdots\ge\mu_r; then XX is the union of pairwise disjoint BhB_h-sets I1,…,IrI_1,\ldots,I_r with ∣Ij∣=μj|I_j|=\mu_j for j=1,…,rj=1,\ldots,r if and only if r≥kr\ge k and ρj≥μ1+⋯+μj\rho_j\ge\mu_1+\cdots+\mu_j for j=1,…,kj=1,\ldots,k.

Proof pointer

Page 8. Subsets of BhB_h-sets and the empty set are BhB_h-sets. For the exchange property, given BhB_h-subsets A,BA,B of XX with ∣A∣<∣B∣|A|<|B|, the set X′=A∪BX'=A\cup B is again a finite B2h−1,h−1B_{2h-1,h-1}-set, so by Theorem 3 its maximal BhB_h-subsets have a common size m≥∣B∣m\ge|B|. A maximal BhB_h-subset A∗A^* of X′X' containing AA then has an element b∈A∗∖Ab\in A^*\setminus A, which lies in B∖AB\setminus A, and A∪{b}⊆A∗A\cup\{b\}\subseteq A^* is a BhB_h-set. For Theorem 5, the unions of kk independent sets of MM are the independent sets of a matroid M(k)M^{(k)} (the paper cites Welsh, Matroid theory, Section 8.3), the maximal subsets with BhB_h-covering number kk are its bases, and nX(k)n_X(k) is its rank; Theorem 6 follows from Dias da Silva's theorem on μ\mu-colorings of a matroid (Linear and Multilinear Algebra 27 (1990), 25--32), as stated on p. 9.

Dependencies

Theorem 3. Read depth: claims checked; the statements of Theorems 4, 5 and 6 and the matroid definitions were read clause by clause on pp. 7--9, and the proofs were read but not checked step by step.

Bears on

  • Problem 156: background only. When the hypothesis holds, Theorem 4 makes the maximal Sidon subsets (h=2h=2) the bases of a matroid, all of one size, so no maximal Sidon subset of such XX is smaller than the largest Sidon subset. The hypothesis fails for {1,…,N}\{1,\ldots,N\} when N≥4N\ge4 (see Theorem 3), so the theorem does not address the problem.