On sums of distinct integers belonging to certain sequences (Q1054783): Difference between revisions

The present paper deals with the classical problem of additive number theory of representing integers as sums of distinct terms taken from a fixed sequence. Here combinatorial methods are used to prove the following result: Let \(\mathcal A\) be a sequence of natural numbers such that \(1\in\mathcal A\), and for every large \(x\), \((x, 2x) \cap\mathcal A \ne \emptyset\). Then there exists a number \(L=L(\mathcal A)\) with the following property: if \(\mathcal B\) is a sequence of natural numbers such that (i) if \(a, a' \in \mathcal A\), \(b,b' \in\mathcal B\) and \(ab = a'b'\) then \(a = a'\) and \(b = b'\); (ii) \(\vert \{b\in\mathcal B, b\le y\}\vert > L \log^2y\) for some \(y > 10\), then \(S = \{m, m = a_1+\cdots+ a_r,\ a_1 > \cdots > a_r,\ a_j\in \mathcal{AB}\}\) (here we use \(\mathcal{AB}= \{ab, a\in \mathcal A, b\in \mathcal B\})\) contains an arithmetical progression. From this result one obtains that, if the G.C.D. of all large terms of \(\mathcal A\) is \(1\), then \(S\) contains all large integers. The above theorem is perhaps not optimal: in particular cases one obtains stronger results [cf. \textit{B. J. Birch}, Proc. Camb. Philos. Soc. 55, 370--373 (1959; Zbl 0093.05003)], but it covers situations which escape from theorems proved by sophisticated analytical methods [cf. \textit{J. W. S. Cassels}, Acta Sci. Math. 21, 111--124 (1960; Zbl 0217.32102)].