0
$\begingroup$

Is anything useful known about the function defined by \[ f(s, \alpha) = \prod_{p} (1 - e^{-2 \pi i p \alpha}p^{-s})^{-1} \quad ? \] Here, $\alpha$ is real. When $\alpha = 1$, this is certainly the Riemann zeta-function.

I want to find an analytic continuation of this into $\sigma > 0$ without success.

If $\alpha$ is rational, may it help?

Thanks in advance.

$\endgroup$

1 Answer 1

5
$\begingroup$

Suppose $\alpha = a/b$ is rational. All but finitely many primes are relatively prime to $b$. For these primes, we have $$ e^{ 2\pi i ap/b} = \sum_{\chi: (\mathbb Z/b)^\times \to \mathbb C^\times} \chi(p) \frac{ \sum_{x \in \mathbb Z/b^\times} \overline{\chi}(x) e^{ 2\pi i ax/b} }{ \phi(b)}$$

so $$\frac{1}{ 1- e^{ 2\pi i ap/b}p^{-s} } = \prod_{\chi: (\mathbb Z/b)^\times \to \mathbb C^\times} \left( \frac{1}{1 - \chi(p) p^{-s} }\right)^{\frac{ \sum_{x \in \mathbb Z/b^\times} \overline{\chi}(x) e^{ 2\pi i x/b} }{ \phi(b)}} + O(p^{-2s})$$

and thus your Euler product is $$ \prod_{\chi: (\mathbb Z/b)^\times \to \mathbb C^\times} L (\chi,s) ^{\frac{ \sum_{x \in \mathbb Z/b^\times} \overline{\chi}(x) e^{ 2\pi i ax/b} }{ \phi(b)}} $$

where we take $L(1,s)$ to be the Riemann zeta function, times something holomorphic on $\sigma>1/2$. In particular, it has a pole at $s=1$ of order $ (1/\phi (b))\sum_{x \in \mathbb Z/b^\times}e^{2\pi i a/b}$, which for $a$ relatively prime to $b$ is $\mu(b)/\phi(b)$ (Ramanujan sum).

In particular, if $b$ is squarefree so $\mu(b) \neq 0$, this has a pole or zero at $s=1$ of non-integer order, hence isn't holomorphic.

If $b$ is squarefree, its holomorphicity in the half-space $\sigma > 1/2$ depends on the Riemann hypothesis for these Dirichlet $L$-functions and the Riemann zeta function.

$\endgroup$
3
  • $\begingroup$ Thanks. I could learn a lot from your answer. What if $\alpha$ is irrational? With only some basic knowledges on Dirichlet series, I can't even show that there exists an A > 0 such that $f(\sigma + it, \alpha) \ll t^{A} $ as $t \to \infty$ for $\sigma > 1/2$; if we don't know about the behavior of $f(\sigma_{0} + it, \alpha)$ for some $\sigma_{0} < 1$ (its analyticity alone doesn't seem to help), then we can't know anything about $f(\sigma + it, \alpha)$ for $\sigma > \sigma_{0}$ after all, can we? $\endgroup$ May 20, 2018 at 23:59
  • $\begingroup$ @Mr.SnowRemover It likely depends on $\alpha$. We can take $\alpha$ a Liouville-type number which is a close approximation to infinitely many rational numbers. By combining this with what I wrote, we should be able to demonstrate that these $\alpha$ have some weird singularity at $s=1$. I suspect for generic $\alpha$ (say, a set of measure $1$) there is analytic continuation to the line $\sigma =1/2$, but don't know if this can be proved. $\endgroup$
    – Will Sawin
    May 21, 2018 at 4:47
  • $\begingroup$ Okay. As I am interested in the vertical growth of $f(\sigma + it, \alpha)$ for $\sigma > 1/2$, your answer seems to resolve one of my own problems whether it grows at most polynomially when $\alpha$ is rational or not. $\endgroup$ May 22, 2018 at 4:22

Your Answer

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