Motivated by Gautschi double inequality, $$ \frac{n^{s}}{n^{\small1}}\ge\frac{\Gamma(n+s)}{\Gamma(n+1)}\ge\frac{(n+1)^{s}}{(n+1)^{\small1}}\ge\frac{\Gamma(n+1+s)}{\Gamma(n+1+1)}\ge\,\cdots \quad\colon\,0\lt{s}\lt1\tag{1} $$ From the main definition of zeta function, $$ \zeta(s)=\sum_{n=1}^{\infty}\frac{1}{n^s}\,\,\,\colon\,Re\{s\}\gt1 \space\Rightarrow\space \zeta(1-s)=\sum_{n=1}^{\infty}\frac{n^{s}}{n^{\small1}} \qquad\colon\,Re\{s\}\lt0\tag{2} $$ And the sum identity of gamma function, $$ \sum_{n=0}^{\infty}\frac{\Gamma(n+s)}{n!}=0 \quad\Rightarrow\quad \Gamma(s)=-\sum_{n=1}^{\infty}\frac{\Gamma(n+s)}{\Gamma(n+1)} \qquad\colon\,Re\{s\}\lt0\tag{3} $$

How to Prove, Disprove, or Justify: $$ \zeta(1-s)+\Gamma(s)=\sum_{n=1}^{\infty}\left[\,\frac{n^{s}}{n^{\small1}}-\frac{\Gamma(n+s)}{\Gamma(n+1)}\,\right] \qquad\qquad\colon\,0\lt{s}\lt1\tag{4} $$ Does it hold if extended to complex plan $\,s\in\mathbb C\,$ inside the critical strip $\small 0\lt Re\{s\}\lt1\,$ ?

By the monotonic decreasing behavior of the inequality, the subtracting result of the two divergent series converge one step forward, covering the critical strip!

  • $\begingroup$ Did you try the Stirling approximation for finding an $\sim$ of $n^{s-1}-\frac{\Gamma(s+n)}{n!}$ ? $\endgroup$ – reuns Feb 7 '17 at 14:26
  • $\begingroup$ (+1) Nice question. Anyway, it is well-known that the values of $\zeta(s)$ in the critical strip can be computed through suitable regularizations of divergent series or integrals. $\endgroup$ – Jack D'Aurizio Feb 7 '17 at 18:56
  • $\begingroup$ @JackD'Aurizio: Thanks. The term-wise regularization of zeta by gamma would shed light on the behavior of $\,\zeta’\,$ inside the critical strip from a totally different angle. Check this out: $$ \zeta'(1-s)=\sum_{n=1}^{\infty}\left[\,\frac{\Gamma(n-1+s)\,\psi(n-1+s)}{(n-1)!}-\frac{\log{n}}{n^{\small{1-s}}}\,\right] \quad\colon\,Re\{s\}\in\,(0,\,1) $$ $\endgroup$ – Hazem Orabi Feb 7 '17 at 20:23
  • $\begingroup$ @user1952009: A term-wise asymptotic approximation with a manageable error-term would be really interesting! I barely manage to justify the whole relation with an error of order $\,\mathcal{O}\left(N^{-\sigma}\right)\,$. $\endgroup$ – Hazem Orabi Feb 7 '17 at 20:24

My first thought is to give an integral representation for the general term of the series, $$\frac{1}{n^{1-s}}-\frac{B(n+s,1-s)}{\Gamma(1-s)}=\frac{1}{\Gamma(1-s)}\left(\int_{0}^{+\infty}x^{-s}e^{-n x}\,dx-\int_{0}^{1}x^{-s}(1-x)^{n+s-1}\,dx\right)$$

$$\frac{1}{n^{1-s}}-\frac{B(n+s,1-s)}{\Gamma(1-s)}=\frac{1}{\Gamma(1-s)}\left(\int_{0}^{+\infty}x^{-s}e^{-n x}\,dx-\int_{0}^{+\infty}(1-e^{-x})^{-s}e^{-(n+s)x}\,dx\right)$$ Summing over $n\geq 1$ we get:

$$\sum_{n\geq 1}\left(\frac{n^s}{n^1}-\frac{\Gamma(n+s)}{\Gamma(n+1)}\right)=\frac{1}{\Gamma(1-s)}\int_{0}^{+\infty}\left(\frac{1}{x^s}-\frac{1}{(e^x-1)^s}\right)\frac{dx}{e^x-1}$$ for every $s$ with real part $\in(0,1)$. The explicit computation of the last integral as $\,\frac{\pi}{\sin(\pi s)}+\Gamma(1-s)\,\zeta(1-s)\,$ proves OP's identity $(4)$. For the computation we may use, for instance, the classical application of Ramanujan's master theorem to Bernoulli polynomials.


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.