Let $n$ be a positive real number. Can the equality

$$\dfrac{d^{n}}{ds^{n}}\Big[s^{n-1}\ln\Big(\pi^{-s/2}\Gamma\Big(1+\frac{s}{2}\Big)\Big)\Big]\Bigg|_{s=1} = - \dfrac{d^{n}}{ds^{n}}\Big[s^{n-1}\ln\Big(\ln(s-1)\zeta(s)\Big)\Big]\Bigg|_{s=1}$$

be possible for any positive real $n$, where $\zeta(s)$ is the Riemann zeta function and $\Gamma(s)$ is the usual gamma function in number theory ?

My approach was by fractional calculus (to accommodate all reals), but I did not complete it since it appeared terribly malicious too me. I'm wondering if there can be some shorter and more intuitive way? Even a long complete proof by fractional calculus will still be very much appreciated.

  • $\begingroup$ Did you use the functional equation of $\zeta$? $\endgroup$ Sep 29 '16 at 14:43
  • $\begingroup$ Maybe an expansion of all the functions involved in Taylor series could help. $\endgroup$ Sep 29 '16 at 14:50
  • $\begingroup$ @SylvainJULIEN, wouldn't that make things extra messy ? $\endgroup$
    – SPD
    Sep 29 '16 at 14:51
  • 1
    $\begingroup$ The "usual gamma function in number theory" is usually denoted by $\Gamma(s)$, no? $\endgroup$ Sep 29 '16 at 23:08
  • 1
    $\begingroup$ There are different approaches to fractional differentiation; could you indicate in your post which one you have in mind? $\endgroup$
    – Todd Trimble
    Sep 30 '16 at 4:42

Many basic complex-analyis misunderstandings here.

  • First of all, $\frac{d^n}{ds^n}$ means $n \in \mathbb{N}$ (otherwise you have to define it). Then you probably wanted to write the Laurent series at $s=1$ of $\log \zeta(s)$. Unfortunately, $s=1$ is a branch point of $\log \zeta(s)$ not an isolated singularity, so it doesn't have a Laurent series there, and the coefficients you wrote on the RHS don't exist.

  • Then I can guess you wanted instead to consider $\dfrac{d^{n}}{ds^{n}}\Big[s^{n-1}\log\Big((s-1)\zeta(s)\Big)\Big]\Bigg|_{s=1}$ since $\log\Big((s-1)\zeta(s)\Big)$ is analytic at $s=1$, but it also means $\log\Big((s-1)\zeta(s)\Big)= \sum_{k=0}^\infty c_k (s-1)^k$ (for $|s-1| < 3$) and hence $\dfrac{d^{n}}{ds^{n}}\Big[s^{n-1}\log\Big((s-1)\zeta(s)\Big)\Big]\Bigg|_{s=1} = \dfrac{d^{n}}{ds^{n}}\Big[\sum_{k=0}^\infty c_k(s-1)^{n+k-1}\Big]\Bigg|_{s=1} $ $=\sum_{k=1}^\infty c_k\frac{(n+k-1)!}{n!}(s-1)^{k-1}\Bigg|_{s=1} = c_1$

    where $c_1 = \frac{d}{ds}\Big[\log\Big((s-1)\zeta(s)\Big)\Big]\Bigg|_{s=1} = \lim_{s \to 1} \frac{\zeta'(s)}{\zeta(s)}+\frac{1}{s-1} = \gamma$ (the Euler-Mascheroni constant)

  • In the same way $\log\Big(\pi^{-s/2}\Gamma\Big(1+\frac{s}{2}\Big)\Big) = \sum_{n=0}^\infty a_n (s-1)^n$ and $\dfrac{d^{n}}{ds^{n}}\Big[(s-1)^{n-1}\log\Big(\pi^{-s/2}\Gamma\Big(1+\frac{s}{2}\Big)\Big)\Big]\Bigg|_{s=1} = a_{1} = -\frac{\log(\pi)}{2}+ \frac{\Gamma'(3/2)}{2\Gamma(3/2)}$ $ -\frac{\log(\pi)}{2}+1+\frac{\Gamma'(1/2)}{2\Gamma(1/2)}=-\frac{\log(\pi)}{2}+1- \frac{2\log(2)+\gamma}{2} \ne -\gamma$

  • Finally, note that the coefficients $c_n$ of the Taylor series of $\log\Big((s-1)\zeta(s)\Big)$ around $s=1$ are given by $$c_n = \frac{1}{n!}\dfrac{d^{n}}{ds^{n}}\Big[\log\Big((s-1)\zeta(s)\Big)\Big]\Bigg|_{s=1}$$

  • $\begingroup$ actually, the coefficients on the right hand side are known to exist, hence i'm not sure what you mean. Anyway, does your answer suggest that the equality is impossible ? $\endgroup$
    – SPD
    Sep 30 '16 at 7:29
  • $\begingroup$ And please check that i clearly stated that $n$ is a positive real number, hence not necessarily any integer. $\endgroup$
    – SPD
    Sep 30 '16 at 7:35
  • $\begingroup$ @SPD you need to define $d^n/dn$ for $n \not \in \mathbb{N}$, and there is almost no interesting reason for looking at the fractional derivative of functions related to $\zeta(s)$, secondly I told you your formula is unclear, and one as to guess what you meant. Now I see that a possiblity is that you meant $\log(\log((s-1)\zeta(s))) = F(s)$ that is holomorphic at $s=1$. Hence (for $n \in \mathbb{N}$) your RHS is $F'(1)$. can you compute it ? $\endgroup$
    – reuns
    Sep 30 '16 at 7:45
  • $\begingroup$ it also seems that you're considering only the case when $n$ is a positive integer, which doesn't completely address the question. $\endgroup$
    – SPD
    Sep 30 '16 at 7:45
  • $\begingroup$ ''There is almost no interesting for looking at the fractional derivative involving $\zeta(s)$''. To me this sounds like your own perception which you shouldn't necessarily assume to hold for everyone. $\endgroup$
    – SPD
    Sep 30 '16 at 7:49

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.