1
$\begingroup$

This property is rather elementary, and not at all specific to $\zeta$, so I am wondering if it has any value in studying the zeros of the Riemann zeta function in the critical strip. It is a well known result? I can provide a proof sketch if you are interested, and it has been checked numerically.

If $\zeta(s)=0$, with $s=\sigma +it$ and $0<\sigma<1$ then for all real $\theta$, we have

$$\sum_{n=1}^{\infty}(-1)^{n+1}\frac{\cos(\theta+t\log n)}{n^\sigma}=0.$$

This would be true even if by chance, one of the zeroes is outside the critical line $\sigma=\frac{1}{2}$. In particular let $t_0$ be the imaginary part of a zero of $\zeta(s)$. Then in order to find all $\sigma$'s such that $\zeta(\sigma +it_0)=0$, we only need to focus on the $\sigma$'s that satisfy the following equation for every $\theta$:

$$\sum_{n=1}^{\infty}(-1)^{n+1}\frac{\cos(\theta+t_0\log n)}{n^\sigma}=0.$$

Of course the Riemann Hypothesis (RH) implies that $\sigma$ must be equal to $\frac{1}{2}$, but my assertion is true even if RH is not true. So if you prove that my equality can only happen if $\sigma=\frac{1}{2}$, then you would have proved RH. I have several strong reasons to believe that my equality does not lead to a proof of RH, yet I am wondering if it has any value.

$\endgroup$
4
  • 1
    $\begingroup$ How do you show convergence of this sum? $\endgroup$ – Zero Jan 7 at 7:07
  • 1
    $\begingroup$ For convergence of the sum, there is a result on Dirichlet series that if the series conditionally converges can at some complex $s_0$, then it also converges on the half-plane $\Re(s-s_0)>0$. The alternating zeta function converges at some positive $s$, hence converges inside the critical strip. The function in question is the real part of $exp(i\theta)$ times the alternating zeta, hence also converges. $\endgroup$ – Ralph Furman Jan 7 at 10:16
  • 3
    $\begingroup$ Isn't mathoverflow supposed to be for research level questions ? $\endgroup$ – reuns Jan 7 at 13:49
  • $\begingroup$ why use $\cos \theta, \sin \theta$ beyond the fact that they have some formulas; if you fix any $f,g$ real functions of any variable(s) independent of $t, \sigma$ and any real functions $A,B(\sigma, t)$ the same follows, namely $f(\theta)A(\sigma, t)+g(\theta)B(\sigma, t)=0$ whenever $A(\sigma, t)+iB(\sigma, t)=0$; not sure what this has to do with anything $\endgroup$ – Conrad Jan 7 at 15:34
3
$\begingroup$

That looks like just one of the two components of a "rotated" Dirichlet Eta function (sometimes called Alternate Zeta function): $$e^{i\theta} \; \eta (s)= e^{i\theta} \; \sum _{n=1}^{\infty}{\frac {(-1)^{n+1}}{n^{s}}}$$ It cannot hence help, as it is always possible to find a rotation angle $\theta$ that will bring to zero one of the two components. True that the non trivial zeros of $\eta$ coincide with those of the Riemann Zeta function $\zeta$. But the difficulty is of course that zeros of both said functions require by definition that both components be zero.

Sure enough, if $\eta(s) = 0$ then it would be so also under any rotation $\theta$.

Conversely, if any one of $\eta(s)$ two components remains $0$ under any rotation $\theta$, then $\eta(s)=0$.

$\endgroup$
4
  • $\begingroup$ Sure, this series converges for real part >1. How do you show convergence for real part >0? $\endgroup$ – Zero Jan 7 at 9:50
  • 2
    $\begingroup$ @Zero If you are referring to my answer, concerning the series for the $\eta$ function most books on the subject would lay out such a proof. See for example Prof. Stoppel's "A Primer of Analytic Number Theory". $\endgroup$ – Luca Jan 7 at 9:56
  • $\begingroup$ I will add a proof sketch, but obviously the status of convergence of that series is a tough nut to crack. $\endgroup$ – Vincent Granville Jan 7 at 11:09
  • 2
    $\begingroup$ @Vincent Do you really need to do that? I mean, it is a well known fact that the series for the $\eta$ function converges also inside the Critical Strip, and so it must also any of its two components. Whatever the rotation applied to the sum of the series. $\endgroup$ – Luca Jan 7 at 11:16

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.