Take the closed form of this Hadamard/Weierstrass product:

$$\displaystyle \frac{\sinh(\pi s)}{\pi s} := \prod_{n=1}^\infty \left(1- \frac{s}{0 + n i} \right) \left(1- \frac{s}{{0 - n i}} \right)$$

Easy to see that the complex zeros occur in conjugated pairs at $ni$ and $-ni$ and, similar to the prime-counting function, these can be plugged into an infinite series:

$$\displaystyle F(x) := \sum_{n=1}^\infty \left(\frac{x^{ni}}{n i} + \frac{x^{-ni}}{{- n i}} \right)$$

Note that this series only converges for $x>0, x \in \mathbb{R}$. It is easy to see that $F(x)=-F(\frac{1}{x})$.

I then put $F(x)$ into the following 'counting' function:

$$\displaystyle N(x) := x + F(x)$$

and I like to conjecture that: $\displaystyle\frac{N(x)}{x} \sim 1$, but how could I prove this?


Writing $x = e^\xi$, we see that

$$f(\xi) = F(e^\xi) = \sum_{n=1}^\infty \left(\frac{e^{in\xi}-e^{-in\xi}}{ni} \right) = 2\sum_{n=1}^\infty \frac{\sin (n\xi)}{n}$$

is periodic, with period $2\pi$. It is the Fourier series of a bounded, piecewise continuous function (the $2\pi$-periodic extension of $\pi-\xi$ on $[0,2\pi)$; thanks to Marko Riedel for remembering that), so the sum is bounded too.

Hence $F$ is bounded, and $\dfrac{N(x)}{x} \sim 1$.

  • $\begingroup$ Many thanks, Daniel. Transforming it into a sine via $x=e$ is a very nice trick! $\endgroup$ – Agno Oct 27 '13 at 23:38
  • 1
    $\begingroup$ We can identify the series in question, it is $\pi-\xi$ with period $2\pi$. This is not difficult to prove. $\endgroup$ – Marko Riedel Oct 27 '13 at 23:40
  • $\begingroup$ Thanks, @MarkoRiedel. I knew it was something like that, but couldn't remember what exactly. $\endgroup$ – Daniel Fischer Oct 27 '13 at 23:42

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.