The Mellin transform for:

$$\displaystyle \Gamma(s)\,\zeta(s)= \int_0^\infty x^{s-1} \frac{1}{e^x-1}\,dx$$


$$\frac{1}{2\,\pi\,i}\int_{c-\infty\,i}^{c-\infty\,i} s^{-x}\,\Gamma(x)\,\zeta(x) \,dx =\frac{1}{e^s-1}$$

but is only valid for $c>1$, however if we take the Dirichlet $\eta$-function:

$$\displaystyle \Gamma(s)\,\eta(s)= \int_0^\infty x^{s-1} \frac{1}{e^x+1}\,dx$$

and transform it into:

$$\frac{1}{2\,\pi\,i}\int_{c-\infty\,i}^{c-\infty\,i} s^{-x}\,\Gamma(x)\,\eta(x) \,dx =\frac{1}{e^s+1}$$

then it is valid for all $c >0$ and now allows for integrating over all complex values with $c=\frac12$.

Assuming the RH, this would mean that the integrand becomes zero for all $x=\rho=\frac12 \pm \gamma\,i$ or in other words: any value of $x$ that could encode information about the distribution of the primes is "knocked out" of the integral to be able to correctly produce $\frac{1}{e^s+1}$.

Wild question: suppose we could prove that there is absolutely no information about the distribution of primes in $\frac{1}{e^s+1}$, would that imply that all non-trivial zeros need to be on the critical line for the Mellin transform to be correct?

Just to share an additional learning I got based on the answer below:


$$\displaystyle \dfrac{\Gamma(s)}{\zeta(s)}= \int_0^\infty x^{s-1} \sum_{n=1}^{\infty}\mu(n)\,e^{-n\,x}\,dx$$

the Mellin transform is ($\mu(n)$= the Mobius function):

$$\frac{1}{2\,\pi\,i}\int_{c-\infty\,i}^{c-\infty\,i} s^{-x}\,\dfrac{\Gamma(x)}{\zeta(x)}\,dx =\sum_{n=1}^{\infty}\mu(n)\,e^{-n\,s}$$

and this immediately shows that the integrand would diverge when a non-trivial zero would lie off the critical line. Hence full convergence in $\frac12 < c < 1$ implies the RH.


Not at all, because the localization of the zeros of $\zeta(s)$ have virtually no impact on $\zeta(s)$ and its related inverse Mellin/Laplace transforms and integral representation.

On the other hand, the localization of the zeros of $\zeta(s)$ have a great impact on $\frac{1}{\zeta(s)}, \frac{\zeta'(s)}{\zeta(s)}, \ln \zeta(s)$ and their related inverse Mellin transforms and integral representation.

That's why finding the abscissa of convergence of $(1-2^{1-s}) \zeta(s) = \sum_{n=1}^\infty (-1)^{n+1} n^{-s}$ is trivial, while finding the abscissa of convergence of $\frac{1}{\zeta(s)} = \sum_{n=1}^\infty \mu(n) n^{-s}$ is the Riemann hypothesis.

The conclusion is that you should look at the Hurwitz zeta function $\zeta(s,a) = \sum_{n=0}^\infty (n+a)^{-s}$ (i.e. $\Gamma(s) \zeta(s,a)= \int_0^\infty \frac{x^{s-1} }{e^x-1}e^{(1-a)x}dx$ !!) having many zeros off the critical line (for $a \not \in \{ 1, 1/2\}$), because it has no Euler product (though it has a functional equation), so that the Dirichlet series for $\frac{1}{\zeta(s,a)}$ is much more complicated (and unrelated with the prime numbers).

  • $\begingroup$ Great answer. Many thanks for these deep insights. Much appreciated. $\endgroup$ – Agno Sep 11 '16 at 17:36

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.