Estimating the sum reciprocal of the Totient Function

asymptoticselementary-number-theorynumber theorysummationtotient-function

I'm going through "Additive Number Theory: The Classical Bases" at the moment, and I'm having difficulty with a certain set of inequalities. Namely,
$$\sum_{q>Q}\frac{1}{\phi(q)^2} \leq \sum_{q>Q}\frac{1}{q^{2-\epsilon}} = O\bigg( \frac{1}{Q^{1-\epsilon}}\bigg) \ .$$
Here, $f=O(g)$ means that there exists a constant $C>0$ such that $\lvert f(x) \lvert \leq Cg(x)$ where $g$ is some positive function.

The first inequality comes from the fact that for all $\epsilon>0$, we have $q^{1-\epsilon} < \phi(q)<q$. I'm not so sure how that last equality holds however. At first I was thinking that
$$\sum_{q>Q}\frac{1}{q^{2-\epsilon}} = \sum_{q>Q} \frac{1}{Q^{2-\epsilon}} = \frac{1}{Q^{2-\epsilon}} \sum_{q>Q}1 = O\bigg( \frac{1}{Q^{1-\epsilon}}\bigg) \ ,$$ but this can't be right as that last sum diverges. I was also tempted to use $\sum_{n>x}\frac{1}{\phi(n)^2}=O(\frac{1}{x})$, but this would not give the desired power of $1-\epsilon$. Pretty sure I'm missing something obvious here, but any insight or help would be greatly appreciated!

EDIT: Can we write $\sum_{q>Q}1=O(1)$? I don't think so, but even if we could the power of the denominator is still an issue.

Best Answer

Compare the sum with the integral $$\int_Q^\infty \frac1{x^{2-\epsilon}}dx$$

Related Question