I've been playing with series involving odd values of the zeta function. Some time ago I found the following closed form $$ \sum_{k=1}^{\infty}\frac{\eta(2k+1)}{2^{2k+1}}=\frac{1}{2}-\ln(2) $$ and recently I stumbled upon this one $$ \sum_{n=1}^{\infty}(-1)^{n-1}\frac{\zeta(2n+1)}{2^{2n+1}} $$ but had no luck, yet, finding a nice closed form. When I say "nice", I mean in terms of $ln$, $\pi$, etc.

So, maybe someone here can find and share the hypothetical nice closed form.


  • 2
    $\begingroup$ By switching the order of summation I get $\gamma+\Re \psi(1+i/2)$. $\endgroup$ – Olivier Oloa May 24 '16 at 22:07
  • 1
    $\begingroup$ how do you find the result for the first series ? $\endgroup$ – reuns May 24 '16 at 22:10
  • $\begingroup$ I'm trying to run away from $\psi$... $\endgroup$ – Neves May 24 '16 at 22:11
  • $\begingroup$ If it weren't alternating, I think $\eta(n)=(1-2^{1-n})\zeta(n)$ would've taken care of most of the problem. $\endgroup$ – Simply Beautiful Art May 24 '16 at 22:32
  • $\begingroup$ @user1952009, see this previous question of mine and from the answer accepted you can get to "the result for the first series". $\endgroup$ – Neves May 26 '16 at 15:42

This is not a complete answer.








According to Wolframalpha

$$\sum_{n=1}^\infty(-1)^{n+1}\frac{\zeta(2n+1)}{2^{2n+1}}=\frac12\left(2\gamma+\psi^{(0)}(1-\frac i2)+\psi^{(0)}(1+\frac i2)\right)$$

And looking at the integral representation,

$$=\gamma+\Re\psi^{(0)}(1+\frac i2)$$


We can conclude the Simple Art's answer observing that $$\sum_{k\geq1}\frac{1}{k\left(4k^{2}+1\right)}=\sum_{k\geq1}\frac{1}{k\left(2k+i\right)\left(2k-i\right)} $$ $$\frac{i}{4}\sum_{k\geq1}\frac{1}{k\left(k+\frac{i}{2}\right)}-\frac{i}{4}\sum_{k\geq1}\frac{1}{k\left(k-\frac{i}{2}\right)} $$ $$=\frac{1}{2}\left(2\gamma+\psi^{\left(0\right)}\left(1+\frac{i}{2}\right)+\psi^{\left(0\right)}\left(1-\frac{i}{2}\right)\right) $$ hence $$\sum_{n\geq1}\left(-1\right)^{n+1}\frac{\zeta\left(2n+2\right)}{2^{2n+1}}=-\frac{1}{4}\left(2\gamma+\psi^{\left(0\right)}\left(1+\frac{i}{2}\right)+\psi^{\left(0\right)}\left(1-\frac{i}{2}\right)\right). $$

  • 1
    $\begingroup$ Hmmm... you ended up just at my starting place. I asked for something in terms of $\ln(2)$, $\pi$, etc. See my question. $\endgroup$ – Neves May 26 '16 at 6:33
  • $\begingroup$ @Neves Sorry, I didn't see the "nice" request in your question. Personally I doubt that it is possible to what do you want, and Mathematica seems agree with me. $\endgroup$ – Marco Cantarini May 26 '16 at 7:37
  • 1
    $\begingroup$ @Neves : you didn't answer to my comment above, but you are criticizing the answers ? $\endgroup$ – reuns May 26 '16 at 8:32
  • $\begingroup$ @ user1952009, I think that Marco wonderstood, my comment, because his answer was not in terms of $\ln$, etc... which I asked for explicitly. $\endgroup$ – Neves May 26 '16 at 15:46
  • 1
    $\begingroup$ the trick is to compute $\displaystyle F(a) = \sum_{k=1}^\infty \frac{\eta(k)}{a^k}$ then $\displaystyle \sum_{k=1}^\infty \frac{\eta(2k)}{a^{2k}} = \frac{F(a)+F(-a)}{2}$. but here $a$ is imaginary, and this case is not treated in your link... @Neves $\endgroup$ – reuns May 26 '16 at 15:56

In order to clarify the four cases - and possibly correction in the question - we have:

$$ \sum_{n=1}^{\infty}\frac{\eta(2n+1)}{2^{2n+1}} \quad = \frac{1}{2}\int_{0}^{\infty}\frac{\cosh(x/2)-1}{e^{x}+1}\,dx = \frac{1}{2}\left(1-\log2\right) \tag{1} \\[6mm] $$ $$ \sum_{n=1}^{\infty}\frac{\zeta(2n+1)}{2^{2n+1}} \quad = \frac{1}{2}\int_{0}^{\infty}\frac{\cosh(x/2)-1}{e^{x}-1}\,dx = \frac{1}{2}\left(2\log2-1\right) \tag{2} \\[6mm] $$ $$ \sum_{n=1}^{\infty}(-1)^{n-1}\frac{\eta(2n+1)}{2^{2n+1}} = \frac{1}{2}\int_{0}^{\infty}\frac{1-\cos(x/2)}{e^{x}+1}\,dx = \frac{1}{2}\log2\,+ \qquad\quad\qquad \frac{1}{8}\left(H_{-1/2+i/4}+H_{-1/2-i/4}\right)-\frac{1}{8}\left(H_{+i/4}+H_{-i/4}\right) \tag{3} \\[6mm] $$ $$ \sum_{n=1}^{\infty}(-1)^{n-1}\frac{\zeta(2n+1)}{2^{2n+1}} = \frac{1}{2}\int_{0}^{\infty}\frac{1-\cos(x/2)}{e^{x}-1}\,dx = \frac{1}{4}\left(H_{+i/2}+H_{-i/2}\right) \tag{4} \\[8mm] $$ $$ \small \color{blue}{H_s=\gamma+\psi(s+1)=\int_{0}^{1}\frac{1-x^s}{1-x}\,dx}\quad\text{"Generalized Harmonic Number"} $$

$$ \#\space\sum_{n=1}^{\infty}\frac{\eta(2n+1)}{2^{2n+1}} = \frac{1}{2}\sum_{n=1}^{\infty}\frac{\Gamma(2n+1)\,\eta(2n+1)}{\Gamma(2n+1)\,2^{2n}} = \frac{1}{2}\sum_{n=1}^{\infty}\frac{1}{2^{2n}\,(2n)!}\int_{0}^{\infty}\frac{x^{2n}}{e^x+1}\,dx = \\ \frac{1}{2}\int_{0}^{\infty}\frac{1}{e^x+1}\left(\sum_{n=1}^{\infty}\frac{(x/2)^{2n}}{(2n)!}\right)\,dx = \frac{1}{2}\int_{0}^{\infty}\frac{\cosh(x/2)-1}{e^{x}+1}\,dx = \color{red}{\frac{1}{2}\left(1-\log2\right)} \\[8mm] $$

$$ \#\space\sum_{n=1}^{\infty}(-1)^{n-1}\frac{\zeta(2n+1)}{2^{2n+1}} = \frac{1}{2}\sum_{n=1}^{\infty}(-1)^{n-1}\frac{\Gamma(2n+1)\,\zeta(2n+1)}{\Gamma(2n+1)\,2^{2n}} = \\ \frac{1}{2}\sum_{n=1}^{\infty}(-1)^{n-1}\frac{1}{2^{2n}\,(2n)!}\int_{0}^{\infty}\frac{x^{2n}}{e^x-1}\,dx = \frac{1}{2}\int_{0}^{\infty}\frac{1}{e^x-1}\left(\sum_{n=1}^{\infty}(-1)^{n-1}\frac{(x/2)^{2n}}{(2n)!}\right)\,dx = \\ \frac{1}{2}\int_{0}^{\infty}\frac{1-\cos(x/2)}{e^{x}-1}\,dx = \color{red}{\frac{1}{4}\left(H_{+i/2}+H_{-i/2}\right)} = \frac{1}{2}\,\Re(H_{i/2}) = \frac{1}{2}\left[\gamma+\Re\left(\psi(1+i/2)\right)\right] \\[8mm] $$ And, unfortunately, neither $Re\left\{H_{i/2}\right\}$ nor $Re\left\{\psi(1+i/2)\right\}$ have known closed form in term of log, exp or similar functions.


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.