An additive property of almost periodic sets (Q1410411)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | An additive property of almost periodic sets |
scientific article |
Statements
An additive property of almost periodic sets (English)
0 references
14 October 2003
0 references
A subset \(\mathcal N\) of the positive integers is called extremal if the density \[ f(q,a) = \lim_{x\to\infty} {1\over x} \cdot \# \{ n\leq x, n\in{\mathcal N}, n \equiv a \bmod q\} \] exists for any \(a\), \(q\), if it has positive density \(\rho = f(1,1)\), and if the equation \[ {1\over\rho} = \sum_{q=1}^\infty \sum_{(a,q)=1} \left|\sum_{b=1}^q {{f(q,a)}\over \rho} \exp \left( 2\pi i\cdot {{ab}\over q}\right) \right|^2 \] holds. For example, as the author shows, subsets \(\mathcal N\) with positive density and multiplicative characteristic function are extremal. The motivation for the author's paper comes from a conjecture of J. Brüdern (from a still unpublished paper on Binary additive problems and the circle method, multiplicative sequences and convergent sieves): The intersection of extremal sets is extremal. The author shows that this conjecture is true. (In the meantime, J. Brüdern solved this conjecture too, by another method). He characterizes \(\alpha\)-extremal sets by the following result. If \(\mathcal N\), with characteristic function \(f\), has positive density, and if \(\alpha=(\alpha_n)_n\) is some sequence, where \(\alpha_n \in [0,1]\), then \(\mathcal N\) is \(\alpha\)-extremal if and only if \(f\in {\mathcal A}^2\) and spec\((f)\) (the at most countable set of the non-vanishing Fourier coefficients of \(f\)) is contained in \(\alpha\). Here, \({\mathcal A}^2\) is the Banach space of functions arbitrarily near to finite trigonometric sums \(\sum a_\nu e(\beta_\nu), \beta\) real, with respect to the [semi]norm \(\displaystyle\|f\|_2 = \sqrt{\limsup_{x\to\infty} {1\over x} \sum_{n\leq x} |f(n)|^2}\). Major and minor arcs are given by \[ M_\alpha(x,Q)= \bigcup_{n\leq Q} \left[\alpha_n - {Q\over x} , \alpha_n + {Q\over x} \right], \quad m_\alpha(x,Q) = [0,1] \setminus M_\alpha(x,Q). \] \(\mathcal N\) is called \(\alpha\)-extremal if its density is positive, if all Fourier coefficients \(\hat{f}_\gamma = M(f\cdot e_{-\gamma})\) of its characteristic function \(f\) exist, and if \[ \int_{m_\alpha(x,Q)} |S(\theta)|^2 d\theta = o(x) \] for any \(Q=Q(x)\) tending to infinity as \(x\to\infty\). This theorem implies that \(\mathcal N\) is extremal if and only if its characteristic function \(f\) is \({\mathcal A}^2\)-almost periodic and has its spectrum contained in the set of rationals (so that \(\hat{f}(\alpha)\not=0\) implies that \(\alpha\) is rational). Another corollary is: If \(r(n)\) is the number of positive integer solutions of the Diophantine equation \( n= [\sqrt{2a }] + [\sqrt{3b }]\), then \(r(n) \sim {n \over {\sqrt{6 }}}\), as \(n\to\infty\).
0 references
almost periodic arithmetical functions
0 references
circle method
0 references
minor arcs
0 references
extremal sets of integers
0 references
conjecture of Brüdern
0 references
trigonometric polynomials
0 references
Fourier series of arithmetical functions
0 references
Fourier coefficients
0 references
theorem of Elliott-Daboussi
0 references
multiplicative arithmetical function
0 references
binary additive problems
0 references
mean-value problems
0 references