On sum sets of Sidon sets. I (Q1328370)

From MaRDI portal
scientific article
Language Label Description Also known as
English
On sum sets of Sidon sets. I
scientific article

    Statements

    On sum sets of Sidon sets. I (English)
    0 references
    0 references
    0 references
    0 references
    2 May 1995
    0 references
    For a finite or infinite set \(A\subseteq \mathbb{N}=\{ 1,2,\dots\}\) let \(A(n) = |A\cap [1,n]|\) and \(2A=\{a+a'\mid a,a'\in A\}\). \(A\) is called a Sidon set if all sums \(a+a'\) in \(2A\), \(a\leq a'\) are distinct. Sum sets \(2A\) of Sidon sets \(A\) cannot consist of ``few'' generalized arithmetic progressions of the same difference. To be more precise let \(B_d= \{a\in 2A\mid a-d\not\in 2A\}\) for \(d\in\mathbb{N}\). There are absolute constants \(c_1, c_2>0\) such that for all \(d\in\mathbb{N}\) we have \(|B_d|> c_1|A|^2\) if \(A\) is a finite Sidon set and \((*)\) \(\limsup_{N\to +\infty} B_d(N) (A(N))^{-2} >c_2\) if \(A\) is an infinite Sidon set. For the proof in the case of infinite \(A\) the generating function \(f(z)= \sum_{a\in A} z^a\), where \(z= e^{-1/N} e^{2\pi i\alpha}\) for large \(N\in\mathbb{N}\) and real \(\alpha\) is considered. Assuming the contrary of the proposition, ingenious estimates of \(I:= \int_0^1 |(1-z^d)f^2 (z)|^2 d\alpha\) lead to contradicting lower and upper bounds for \(I\). By example it is shown that \((A(N))^{-2}\) in \((*)\) cannot be replaced by \((A(N) )^{-2} \log^{-1} N\). While these results in the case \(d=1\) deal with blocks of consecutive elements in \(2A\) for Sidon sets \(A\), the next theorems give information about gaps between consecutive elements of \(2A\). Let \(2A= \{s_1,s_2,\dots\}\), \(s_1<s_2< \dots\;\). For \(n\in\mathbb{N}\), \(n>n_0\) there exists a Sidon set \(A\subseteq \{1,2,\dots, n\}\) such that \(s_{i+1}-s_i < 3\sqrt{n}\) for all \(s_{i+1}\in 2A\setminus \{s_1\}\). The prime number theorem is used for constructing such sets \(A\). For infinite Sidon sets the probabilistic method of Erdős and Rényi is adapted to prove the following result: For \(\varepsilon>0\) there is a Sidon set \(A\) such that \[ s_{i+1}-s_i < \sqrt{s_i} (\log s_i)^{(3/2)+ \varepsilon} \] for all \(i>i_0 (\varepsilon)\) and \(s_i\in 2A\). Also given are lower estimates for \(s_{i+1}- s_i\). A catalog of unsolved problems concerning Sidon sets and \(B_2[g]\) sets closes this part I.
    0 references
    0 references
    addtive bases
    0 references
    \(B_2\)-sequences
    0 references
    sum sets of Sidon sets
    0 references
    infinite Sidon sets
    0 references
    0 references
    0 references