Suppose we have $\alpha \in \mathbb{R}$. Then we know that

$$\sum_{1 \leq n \leq X} e(n \alpha) \ll \min \{ X, \|\alpha\|^{-1} \}$$

where $\| \cdot \|$ is the distance to the nearest integer. I was wondering if we put von Mangoldt function as a weight and consider the exponential sum

$$S(\alpha) = \sum_{1 \leq n \leq X} \Lambda(n) e(n \alpha)$$

can we obtain an estimate similar to above in terms of $\| \alpha \|$ and $X$?

Thank you very much!

PS Here $\Lambda$ is the von Mangoldt function, $e(m) = e^{2 \pi i m}$.


It's a classical sum going back to Vinogradov, but the estimate is more complicated. Usually, it is phrased in terms of an approximation $a/q$ of $\alpha$, with $a$ coprime to $q$, such that $|\alpha-a/q|\leq 1/q^2$ as $$S(\alpha)\ll ((Xq)^{1/2}+Xq^{-1/2}+X^{4/5})$$ (up to powers of $\log X$). This is explained in many places, e.g. in Vaughan's "Hardy--Littlewood method" if I remember right.

| cite | improve this answer | |
  • 1
    $\begingroup$ Yes. And the proof relies on "Vaughan's identity". $\endgroup$ – GH from MO Jan 19 '17 at 0:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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