3
$\begingroup$

On page 188.

Lemma 10.2.3 is $\sum_{\substack{\delta \leq x \\ 2 \nmid \delta}}\frac{\mu^{2}(\delta)}{\phi_{1}(\delta)} = A_{1}log(x) +A_{2} + O(\frac{1}{x^{1/4}})$ for positive $A_{1}$, $A_{2}$.

And $I = \int_{3}^{x} (\sum_{\substack{\delta \leq t \\ 2 \nmid \delta}}\frac{\mu^{2}(\delta)}{\phi_{1}(\delta)}) \frac{1}{t (1+log(\frac{x}{t}))} dt$

Now my question is how did they get the following equation by lemma 10.2.3?

$I = -A_{1} \int_{3}^{x} \frac{dt}{t} + (A_{1}log(x)+A_{1}+A_{2})\int_{3}^{x} \frac{dt}{t(1+log(x)-log(t)} + O(\frac{1}{x^{1/4}})$

I have been stuck on this for a long while. Any help is appreciated. :)

Edited

$\endgroup$
3
  • 2
    $\begingroup$ You should definitely replace the '$x$' appearing as an upper bound for the variable $\delta$ in the expression of the integral $I$ with '$t$'. $\endgroup$ Apr 30, 2014 at 14:03
  • $\begingroup$ Plugging in the lemma, you'd get an $A_1 \log t$ term. This is the same as $-A_1(1+\log(x/t))+A_1\log x + A_1$. Is that your question? $\endgroup$
    – rlo
    May 1, 2014 at 17:32
  • 1
    $\begingroup$ Actually, wait. Unless I'm missing something, their claimed error term is wrong, and it should be $O(1/\log x)$ rather than $O(1/x^{1/4})$ (this is fine for their purposes, as the error on the next line is $O(1/\log x)$ anyway). To see it has to be at least this big, just look at the integral of the error from 3 to 4. To see that it's no bigger, split the integral into two parts, integrating from 3 to $x^{1/2}$, say, and from $x^{1/2}$ to $x$. The latter decays like $x^{-1/8}$, while the former can be bounded by $O(1/\log x)$. $\endgroup$
    – rlo
    May 1, 2014 at 18:22

1 Answer 1

1
$\begingroup$

I am not really understanding the question, probably, but why can't you just substitute the estimate into the integral? if you omit the error term, you get:

$$ \log \left(\log \left(\frac{x}{3}\right)+1\right) (A_1 \log (x)+A_2). $$ Doing the error term separately, you get:

$$ \log \left(\log \left(\frac{x}{3}\right)+1\right) O(1/x^{1/4}). $$

Are you saying this is different from what the authors claim?

$\endgroup$
1
  • $\begingroup$ Sorry, I had a typo. The sum inside the integral should be a function with respects to t. Thanks. $\endgroup$
    – Doorbell
    May 1, 2014 at 8:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.