6
$\begingroup$

With $z=\sigma + x \, i;\,\, \sigma,x \in \mathbb{R}$, numerical evidence strongly suggests that:

$$\displaystyle \int_{0}^{\infty} \,\zeta(z)\,\zeta(\overline{z})\,\Gamma(z)\,\Gamma(\overline{z}) \,dx =\pi\,\Gamma(2\,\sigma)\,\big(\zeta(2\,\sigma-1)-\zeta(2\,\sigma)\big)$$

for all $\sigma > 1$.

Could this be proven?


Just to share that the equation can be extended towards $0<\sigma<1$ by starting from: $$\Gamma(s)\, \zeta(s) = \int_0^\infty x^{s-1}\left(\frac{1}{e^{x}-1}-\frac{1}{x}\right)\,dx$$ and following the same logic as in the answer below. This gives the closed form:

$$\displaystyle \int_{0}^{\infty} \,\zeta(z)\,\zeta(\overline{z})\,\Gamma(z)\,\Gamma(\overline{z}) \,dx =\pi\,\Gamma(2\,\sigma)\,\left(\frac{(2\,\sigma-3)}{(2\,\sigma-1)} \,\zeta(2\,\sigma-1)-\zeta(2\,\sigma)\right)$$

Note that since $\sigma$ could also be $\in \mathbb{C}$, this implies that when $\rho=2\,\sigma$ or $\rho=2\,\sigma-1$ ($\rho$ = a non-trivial zero of $\zeta(s)$), then the RHS becomes fully multiplicative.

The process can be extended indefinitely by for instance starting from $$\Gamma(s)\, \zeta(s) = \int_0^\infty x^{s-1}\left(\frac{1}{e^{x}-1}-\frac{1}{x} +\frac12\right)\,dx$$ which gives for the domain $-1<\sigma<0$ the following closed form:

$$\displaystyle \int_{0}^{\infty} \,\zeta(z)\,\zeta(\overline{z})\,\Gamma(z)\,\Gamma(\overline{z}) \,dx =\pi\,\left(2\,\sigma-3)\,\Gamma(2\,\sigma-1)\,\zeta(2\,\sigma-1\right)$$

And subsequently for the domain $-2<\sigma<-1$ we start from: $$\Gamma(s)\, \zeta(s) = \int_0^\infty x^{s-1}\left(\frac{1}{e^{x}-1}-\frac{1}{x} +\frac12-\frac{x}{12}\right)\,dx$$

which gives the closed form:

$$\displaystyle \int_{0}^{\infty} \,\zeta(z)\,\zeta(\overline{z})\,\Gamma(z)\,\Gamma(\overline{z}) \,dx =\pi\,\Gamma(2\,\sigma)\,\left(\frac{(2\,\sigma-3)}{(2\,\sigma-1)} \,\zeta(2\,\sigma-1)-\frac{\sigma}{3}\,\zeta(2\,\sigma+1)\right)$$

Don't think there is an easy pattern or generic formula since Bernoulli numbers are involved.

$\endgroup$
11
  • $\begingroup$ What did you try? As a start you may think about substituting the definition of $\Gamma$ and $\zeta$ (and c.c.). $\endgroup$ – Turing Sep 26 '16 at 22:41
  • $\begingroup$ By the way: "numerical evidences". Is that your hypothesis or you found it somewhere? $\endgroup$ – Turing Sep 26 '16 at 22:44
  • 1
    $\begingroup$ This can be rewritten $$\int_{0}^{\infty} |\zeta(\sigma +xi)\Gamma(\sigma+xi)|^2\,dx = ...$$ $\endgroup$ – Thomas Andrews Sep 26 '16 at 22:47
  • $\begingroup$ Purely my conjecture. But pretty sure it works at high precision. I was experimenting with a Mellin transform that had $\zeta(z)\,\Gamma(z)+\zeta(\overline{z})\,\Gamma(\overline{z})$ as the integrand. Then just for fun tried to multiply instead of add and found a closed form. Not rigorous math, I know... $\endgroup$ – Agno Sep 26 '16 at 22:49
  • $\begingroup$ Do you have a short proof that $\Gamma(s)\zeta(s) = \int_0^\infty x^{s-1}(\frac{1}{e^x-1}-1/x)dx$ for $Re(s) \in (0,1)$ ? $\endgroup$ – reuns Oct 1 '16 at 17:34
4
$\begingroup$

Again, you have to look at the Fourier/Laplace transform. For $Re(s) > 1$ : $$\Gamma(s) \zeta(s) = \int_0^\infty \frac{x^{s-1}}{e^{x}-1}dx \overset{(x \, = \,e^u)}= \int_{-\infty}^\infty \frac{e^{su}}{e^{e^u}-1}du = \int_{-\infty}^\infty \frac{e^{-su}}{e^{e^{-u}}-1}du$$

i.e. $\Gamma(s)\zeta(s)$ is the Laplace transform of $\frac{1}{e^{e^{-u}}-1}$ and $F_\sigma(\xi) = \Gamma(\sigma+2i \pi \xi)\zeta(\sigma+2i \pi \xi)$ is the Fourier transform of $f_\sigma(u) = \frac{e^{-\sigma u}}{e^{e^{-u}}-1}$.

Now we can apply the Parseval theorem and get $$\int_{-\infty}^\infty |F_\sigma(\xi)|^2d\xi = \int_{-\infty}^\infty |f_\sigma(u)|^2du = \int_{-\infty}^\infty \frac{e^{-2\sigma u}}{(e^{e^{-u}}-1)^2}du $$

$$=\int_{-\infty}^\infty \frac{e^{2\sigma u}}{(e^{e^{u}}-1)^2}du = \int_0^\infty \frac{x^{2\sigma-1}}{(e^{x}-1)^2}dx = \int_0^\infty x^{2\sigma-1}\sum_{n=2}^\infty (n-1) e^{-nx}dx = (\zeta(2\sigma-1)-\zeta(2\sigma))\Gamma(2\sigma)$$


And note that $\zeta(s)$ is the Laplace transform of the distribution $\sum_{n=1}^\infty \delta(u-\ln n)$, while (under the Riemann hypothesis) the non-trivial zeros define the spectrum of the distribution $\sum_{p^k} \frac{\ln p}{\sqrt{p^k}} \ (\delta(u-\ln p^k)+\delta(u+\ln p^k))$

$\endgroup$
3
  • $\begingroup$ Thanks (again), user1952009. I am deeply impressed by how easy you make it look playing with all these various transforms. $\endgroup$ – Agno Sep 27 '16 at 8:42
  • $\begingroup$ I believe there is a small typo in the last two integrals. Shouldn't the $-\infty$ be $0$ in both? $\endgroup$ – Agno Sep 30 '16 at 22:24
  • $\begingroup$ @Agno yep ${}{}$ $\endgroup$ – reuns Oct 1 '16 at 17:29

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.