A while back I was able to prove the following identity, $$\sum_{k=1}^{\infty}\frac{\Gamma(k+r)}{\Gamma(k)(k+r)^s}=\sum_{k=0}^{r}s(r+1,r+1-k)\zeta(s-r+k)$$ where $s(k,n)$ are the Sterling numbers of the first kind. To my satisfaction (and dismay) I was able to type this infinite series into WolframAlpha and it gave me my exact same finite summation to the point, meaning one I was correct but two it had already been done. After some googling I was seeing quite a few papers that tied finite sums involving Sterlings numbers and the Riemann Zeta function to infinite sums involving the Pochhammer symbol, $(x)_k$, $$(x)_k=\frac{\Gamma(x+k)}{\Gamma(x)}$$ I was wondering if anyone knew whether these results are well known and/or what central theorem/idea they stem from.

  • $\begingroup$ Are you using the gamma function because you expect $r\in\mathbb{R}$? If so, I will point out that defining the Stirling numbers for fractional values is a pain, and the RHS summation has issues with its upper bound as well. If not, why not use the factorial? $\endgroup$ – Simply Beautiful Art Jul 2 '16 at 12:38

Suppose we seek to simplify the sum

$$\sum_{k=1}^\infty \frac{\Gamma(k+r)}{\Gamma(k) (k+r)^s}.$$

with $r$ a positive integer. This is

$$\sum_{k=1}^\infty \frac{(k+r-1)!}{(k-1)! (k+r)^s} = \sum_{k=1}^\infty \frac{1}{(k+r)^s} (k+r-1)^{\underline{r}} \\ = \sum_{k=1}^\infty \frac{1}{(k+r)^{s+1}} (k+r)^{\underline{r+1}}.$$

Now recall that

$$x^{\underline{r}} = \sum_{q=0}^r (-1)^{r-q} \left[r\atop q\right] x^q.$$

We thus obtain

$$\sum_{k=1}^\infty \frac{1}{(k+r)^{s+1}} \sum_{q=0}^{r+1} (-1)^{r+1-q} \left[r+1\atop q\right] (k+r)^q \\ = \sum_{q=0}^{r+1} (-1)^{r+1-q} \left[r+1\atop q\right] \left(\zeta(s+1-q) - \sum_{p=1}^r \frac{1}{p^{s+1-q}}\right).$$

Observe however that

$$ \sum_{q=0}^{r+1} (-1)^{r+1-q} \left[r+1\atop q\right] \sum_{p=1}^r \frac{1}{p^{s+1-q}} \\ = \sum_{p=1}^r \frac{1}{p^{s+1}} \sum_{q=0}^{r+1} (-1)^{r+1-q} \left[r+1\atop q\right] p^q \\ = \sum_{p=1}^r \frac{1}{p^{s+1}} p^{\underline{r+1}}.$$

However with $1\le p\le r$ the term $p^{\underline{r+1}}$ includes a zero factor and hence the latter sum vanishes, leaving just

$$\sum_{q=0}^{r+1} (-1)^{r+1-q} \left[r+1\atop q\right] \zeta(s+1-q).$$

To ensure convergence we must have $s-r\gt 1.$

Wikipedia lists the Stirling number identity.

  • $\begingroup$ Is that the rising factorial you have going over there? ($r$ in second line) $\endgroup$ – Simply Beautiful Art Jul 2 '16 at 12:36

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.