English

Iterated logarithm approximations to the distribution of the largest prime divisor

Number Theory 2009-03-17 v1

Abstract

The paper is concerned with estimating the number of integers smaller than xx whose largest prime divisor is smaller than yy, denoted ψ(x,y)\psi (x,y). Much of the related literature is concerned with approximating ψ(x,y)\psi (x,y) by Dickman's function ρ(u)\rho (u), where u=lnx/lnyu=\ln x/\ln y. A typical such result is that ψ(x,y)=xρ(u)(1+o(1))\eqno(1) \psi (x,y)=x\rho (u)(1+o(1)) \eqno (1) in a certain domain of the parameters xx and yy. In this paper a different type of approximation of ψ(x,y)\psi (x,y), using iterated logarithms of xx and yy, is presented. We establish that ln(ψx)=u[ln(2)xln(2)y+ln(3)xln(3)y+ln(4)xa]\eqno(2) \ln (\frac {\psi}{x})=-u [\ln ^{(2)}x-\ln ^{(2)}y+\ln ^{(3)}x-\ln ^{(3)}y+\ln ^{(4)}x-a] \eqno (2) where a<a<aˉ\underbar{a}<a<\bar{a} for some constants a\underbar{a} and aˉ\bar{a} (denoting by ln(k)x=ln...lnx\ln ^{(k)}x=\ln ...\ln x the kk-fold iterated logarithm). The approximation (2) holds in a domain which is complementary to the one on which the approximation (1) is known to be valid. One consequence of (2) is an asymptotic expression for Dickman's function, which is of the form lnρ(u)=u[lnu+ln(2)u](1+o(1))\ln \rho (u)=-u[\ln u+\ln ^{(2)}u](1+o(1)), improving known asymptotic approximations of this type. We employ (2) to establish a version of Bertrand's Conjecture, and indicate how this method may be used to sharpen the result.

Keywords

Cite

@article{arxiv.0903.2712,
  title  = {Iterated logarithm approximations to the distribution of the largest prime divisor},
  author = {Arie Leizarowitz},
  journal= {arXiv preprint arXiv:0903.2712},
  year   = {2009}
}

Comments

44 pages