0
$\begingroup$

I'm trying to arrive at estimate 1.17 (page 21) of Koukoulopoulos lecture notes [https://dms.umontreal.ca/~koukoulo/documents/notes/sievemethods.pdf]

$$\#\{n \leq x : p|n \Rightarrow p > \sqrt{x}\} = \frac{(2e^{- \gamma}+ o(1))x}{\log x} + O(4^{(1+o(1))\frac{\sqrt{x}}{\log x}})$$

In the error term I can see that we are using the Prime Number Theorem, but i don't see how we get the $1+o(1)$ or how to get the $o(1)$ in the main term

We start with $\#\{n \leq x: p|n \Rightarrow p > \sqrt{x}\} = \lfloor{x}\rfloor - \sum_{i=1}^{r}\lfloor{\frac{x}{p_{i}}\rfloor}\pm\dots$ by an inclusion-exclusion argument.

We have the estimate that $\lfloor{x}\rfloor = x + O(1).$ Therefore we obtain $\#\{n \leq x: p|n \Rightarrow p > \sqrt{x}\} = x \prod_{p \leq \sqrt{x}}\Big(1 - \frac{1}{p}\Big).$

Using Merten's theorem that for $z \geq 2$ we have $\prod_{p \leq z}\Big(1 - \frac{1}{p}\Big) = \frac{e^{- \gamma}}{\log z}\Big(1 + O\Big(\frac{1}{\log z}\Big)\Big),$ gives

$$ \#\{n \leq x: p|n \Rightarrow p > \sqrt{x}\} = \frac{xe^{- \gamma}}{\log{\sqrt{x}}}\Big(1 + O\Big(\frac{1}{\log{\sqrt{x}}}\Big)\Big) = \frac{2xe^{-\gamma}}{\log x}\Big(1 + O\Big(\frac{2}{\log x}\Big)\Big).$$

I can rewrite the inclusion-exclusion argument in terms of the Mobius function as is done in equation 1.1.6 (page 20), but then I don't see how we get the overall result.

It's mostly the asymptotic notation that I'm stuck on as I'm not too familiar with it yet, particularly using "little oh" notation inside of "big oh" notation.

Thanks.

$\endgroup$
3
  • $\begingroup$ You seem to have dropped the error term from approximating $\lfloor x \rfloor$ by $x$ in the equation after `therefore we obtain'. I won't vote to close, but I think this question might be better suited to MathStackexchange. $\endgroup$ Commented Jun 11, 2019 at 18:06
  • $\begingroup$ @MarkWildon Thanks. I'm not too familiar with how to manipulate these error terms, so we have a O(1) contribution from approximating the $\lfloor{x}\rfloor$ but I don't see how this will give us the o(1) terms instead... Also thanks for not closing, I'm not sure what the differences are between MathStackexchange and MathOverflow... $\endgroup$ Commented Jun 11, 2019 at 18:14
  • $\begingroup$ Roughly, learners should start with math.stackexchange. While we don't card at the door, we hope people asking here on MathOverflow have more experience with manipulating O notation than appears in this post. The exponential term (4^stuff) comes from collecting the error terms of the product. (You might check out MathOveflow question 37679 for an older sieve argument.) Gerhard "Big Errors And Small Errors" Paseman, 2019.06.11. $\endgroup$ Commented Jun 11, 2019 at 18:42

0

You must log in to answer this question.