Sets of multiples of finite sequences (Q1301645)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Sets of multiples of finite sequences
scientific article

    Statements

    Sets of multiples of finite sequences (English)
    0 references
    0 references
    0 references
    16 June 2003
    0 references
    For a finite set \(A\) of integers, let \(\mu (A)\) denote the density of integers divisible by at least one element of \(A\). We have \(\mu (A)<1\) unless \(1\in A\), and it is easy to exhibit examples with \(\mu (A)>1-\varepsilon \). The starting problem of the paper is to find many disjoint sets with this property. The authors denote by \(\beta (N, \varepsilon)\) the maximal number of disjoint subsets of \([1,N]\) such that each satisfies \(\mu (A)>1-\varepsilon \). For this quantity they give the sharp estimate \[ \beta (N, \varepsilon) = ( \log N)^{ \log 2} e^{O( \xi (N) \sqrt { \log \log N})} \] with any function \(\xi (N)\to \infty \). The upper estimate follows immediately from the classical results on the normal order of the number of divisors. The authors show the lower estimate by a random example, as well as by a natural construction consisting of certain residues modulo a suitably sized \(q\). To achieve this they define \(\tau (n,N,a,q)\) as the number of divisors \(d\) of \(n\) satisfying \(d\leq N\) and \(d\equiv a \pmod{ q}\). They prove that if the sizes of \(q\) and \(N\) are connected by the relation \[ q = ( \log N)^{ \log 2} 2^{(z+o(1)) \sqrt { \log \log N}}, \] then the density of integers \(n\) such that \[ \min _{(a,q)=1} \tau (n,N,a,q) \geq 1 \] is approximately \(\Phi (z)\), the normal distribution function. It is known that \(c_1<\mu ([N,N^{1+\alpha }])<1-c_2\) with positive constants \(c_1, c_2\) depending on \(\alpha \). The paper contains two improvements of this result. One asserts that many integers have even two very near divisors from such an interval. With \[ \varepsilon _N = ( \log N)^{1- \log 3} e^{ \xi (N) \sqrt { \log \log N}} \] and \[ A = \{ab: N<a<b<(1+\varepsilon _N)a<N^{1+\alpha }\} \] we have \(\mu (A)>c_3\) with some positive \(c_3\) depending on \(a\). The other says that for any polynomial \(F\) with integer coefficients the density of those values of \(n\) for which \(F(n)\) has a divisor in \((N, N^{1+\alpha })\) tends to a positive limit as \(N\to \infty \) for fixed \(\alpha >0\).
    0 references
    sets of multiples
    0 references
    maximal number of disjoint subsets
    0 references
    density of integers
    0 references
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references