I've been trying to answer another question, and as a part of the solution encountered the following series:

$$S_q(x)=\sum_{k=0}^\infty \frac{x^{2 k+1}}{(2k+1) \Gamma (\frac{2k+1}{q})}$$

Where $q=1,2,3,4,\ldots$.

For $q=1$ and $q=2$ the closed form is trivial, however, for $q \geq 3$ it becomes harder and harder to find.

I would like to know the general form, which should look something like this:

$$S_q(x)= \sum_{n=1}^q F_n(x)$$ Where $F_n(x)$ is a hypergeometric function, possibly multiplied by some factors.



We have:

$$\sum_{m=0}^\infty \frac{x^{qm+n}}{(qm+n) \Gamma (m+\frac{n}{q})}= \frac{x^n}{n \Gamma(\frac{n}{q})} {_1 F_1} \left(1; 1+ \frac{n}{q}; x^q \right)$$

And the series in general should reduce to a finite sum of such terms, but I don't know how to correctly describe all the possible choices of $n$ for a particular $q$.

  • $\begingroup$ $S_q(x)=\frac{1}{q}\Im E_{\frac{1}{q}}(ix)$ please have a look at e.g. here . Maybe there is some literature about the Mittag-Leffler function. $\endgroup$ – user90369 Aug 8 '19 at 11:05
  • $\begingroup$ @user90369, thank you for the reference, I didn't know about this function $\endgroup$ – Yuriy S Aug 8 '19 at 11:26

First, multiply by $q$ to simplify the denominator:


If all you want is hypergeometric functions, then just look at the ratio between terms. Let $k=qm+n$ and $a_{m+1}/a_m$:


Hence we have


In the case that $q$ is even, less terms are needed:


  • $\begingroup$ Oh, this is more simple than I thought. Somehow I didn't think that ratio of terms will help in this case, but it clearly does. Thank you $\endgroup$ – Yuriy S Aug 8 '19 at 16:31
  • $\begingroup$ Made some fixes. $\endgroup$ – Simply Beautiful Art Aug 8 '19 at 16:38
  • $\begingroup$ So, it is a sum of hypergeometric functions after all $\endgroup$ – Yuriy S Aug 8 '19 at 16:43
  • $\begingroup$ Yeah $\hphantom{blank text here}$ $\endgroup$ – Simply Beautiful Art Aug 8 '19 at 16:43

Unfortunately, I don't understand the sense to make a sum of series out of one series. At least these sub-series should have something particular, not only another structure. Perhaps it's clearer to build a closed form using the incomplete gamma function.

$x!:=\Gamma(1+x) ~ ; ~~z>0$

$\displaystyle E_z(x):=\sum\limits_{k=0}^\infty\frac{x^k}{(kz)!}\hspace{2cm}$ Mittag-Leffler function

$\displaystyle f_z(x):=\sum\limits_{k=0}^\infty\frac{x^{k+z}}{(k+z)!}=e^x\left(1-\frac{\Gamma(z,x)}{\Gamma(z)}\right)$

$\displaystyle z:=\frac{1}{q} , ~ q\in\mathbb{N}$

$\displaystyle E_{1/q}(x) = e^{x^q} + \sum\limits_{j=1}^{q-1}f_{j/q}(x^q) = e^{x^q}\left(q - \sum\limits_{j=1}^{q-1}\frac{\Gamma(j/q,x^q)}{\Gamma(j/q)}\right) $

$\displaystyle S_q(x)=\frac{1}{q}\Im E_{1/q}(ix)$


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.