Sumsets of dense sets and sparse sets (Q1760400): Difference between revisions

For two sets \(A, B\) of integers, we define their sumset by \(A+B := \{ a+b : a \in A, b \in B \}\). This paper concerns the structure of such sets, a fairly central topic in additive combinatorics, for infinite sets \(A, B\) satisfying certain density conditions. Under the condition that \(A\) and \(B\) have positive upper Banach density, so that there is some \(\delta > 0\) for which there are arbitrarily long intervals \(I\) with \(\frac{|A \cap I|}{|I|} \geq \delta\) (and similarly for \(B\)), \textit{R. Jin} [Proc. Am. Math. Soc. 130, No. 3, 855--861 (2002; Zbl 0985.03066)] proved that \(A+B\) is a so-called piecewise syndetic set: it is the intersection of syndetic set -- a set of integers with uniformly bounded gaps between consecutive elements -- and a thick set -- a set containing arbitrarily long intervals. Under the same hypotheses, \textit{V. Bergelson} et al. [in: Topics in discrete mathematics. Dedicated to Jarik Nešetřil on the occasion of his 60th birthday. Berlin: Springer, 13--37 (2006; Zbl 1114.37008)] proved the stronger conclusion that \(A+B\) is piecewise Bohr: it is the intersection of a thick set with a set containing the positive level set \(\{ n \in \mathbb Z: \text{Re}\, P(n) > 0 \}\) of a trigonometric polynomial \(P(t) = \sum_{k=1}^m c_k e^{i \lambda_k t}\) with real frequencies \(\lambda_k\). The paper under review generalises these results by weakening the density condition on one of the sets; one needs only that \(A\) has positive density according to a sequence of measures satisfying a certain pseudorandomness condition, and one still obtains the conclusion that \(A+B\) is piecewise Bohr. The author gives some examples of sets \(A\) satisfying this condition and some related results. In the finitary version of the above general setup, one instead looks at the structure of \(A+B\) for dense sets \(A, B \subset \{1,2,\ldots,N\}\). This study was initiated by \textit{J. Bourgain} [in: A tribute to Paul Erdős, 105--109 (1990; Zbl 0715.11006)], and the best bounds currently known can be found in [\textit{E. Croot} et al., Comb. Probab. Comput. 22, No. 3, 351--365 (2013; Zbl 1348.11011)]: one can there draw the similar conclusion that \(A+B\) contains translates of large subsets of Bohr sets.