Generalizing $\sum_{n=1}^\infty \frac{\Gamma(n)\Gamma(x)}{\Gamma(n+x)} = \frac1{x-1}$ to double sum

beta functiongamma functionsequences-and-series

Using the definition of the beta function, it's easy to show that

$$\begin{align*}
\sum_{n=1}^\infty \frac{\Gamma(n)\Gamma(x)}{\Gamma(n+x)} &= \sum_{n=1}^\infty \operatorname{B}(x,n) \\[1ex]
&= \sum_{n=1}^\infty \int_0^1 t^{x-1} (1-t)^{n-1} \, dt \\[1ex]
&= \int_0^1 t^{x-1} \sum_{n=1}^\infty (1-t)^{n-1} \, dt \\[1ex]
&= \int_0^1 t^{x-2} \, dt \\[1ex]
&= \frac1{x-1}
\end{align*}$$

I was wondering if we could generalize this to the case of a double sum,

$$\sum_{m=1}^\infty \sum_{n=1}^\infty \frac{\Gamma(m) \Gamma(n) \Gamma(x)}{\Gamma(m+n+x)}$$

I've considered rewriting

$$\frac{\Gamma(m) \Gamma(n) \Gamma(x)}{\Gamma(m+n+x)} = \begin{cases} \operatorname{B}(m,x) \operatorname{B}(m+x,n) \\ \operatorname{B}(n,x) \operatorname{B}(n+x,m) \\ \operatorname{B}(m,n) \operatorname{B}(m+n,x) \end{cases}$$

The third form seems the most useful, and applying the method above leads me to

$$\begin{align*}
\sum_{m=1}^\infty \sum_{n=1}^\infty \operatorname{B}(m,n) \operatorname{B}(m+n,x) &= \int_0^1 (1-t)^{x-1} \sum_{m=1}^\infty \sum_{n=1}^\infty \operatorname{B}(m,n) t^{m+n-1} \, dt
\end{align*}$$

$\operatorname{B}$ is symmetric, so

$$\begin{align*}
\sum_{m=1}^\infty \sum_{n=1}^\infty \operatorname{B}(m,n) t^{m+n-1} &= \sum_{m=1}^\infty \operatorname{B}(m,m) t^{2m-1} + 2 \sum_{m=2}^\infty \sum_{1 \le n < m} \operatorname{B}(m,n) t^{m+n-1} \\[1ex]
\end{align*}$$

and we can rewrite the first sum as

$$\begin{align*}
\sum_{m=1}^\infty \operatorname{B}(m,m) t^{2m-1} &= \sum_{m=1}^\infty \frac{\Gamma(m)^2}{\Gamma(2m)} t^{2m-1} \\[1ex]
&= \sum_{m=1}^\infty \frac{t^{2m-1}}{\binom{2m}m} \\[1ex]
&= \frac{4\arcsin\left(\frac t2\right)}{\sqrt{4-t^2}}
\end{align*}$$

The subsequent integral over $t\in[0,1]$ seems doable if $x$ is a positive integer; not so sure otherwise. I also don't know what else could be done with the other sum.

Mathematica gives a result in terms of hypergeometric functions but is unable to simplify any further.

$$\sum_{m=1}^\infty \sum_{n=1}^\infty \frac{\Gamma(m)\Gamma(n)\Gamma(x)}{\Gamma(m+n+x)} = \frac1{x^2} {}_3F_2\left(\left.\begin{array}{c|c}1,1,x\\1+x,1+x\end{array}\right\vert 1\right)$$

Best Answer

You may find a simple solution in (Almost) Impossible Integrals, Sums, and Series, Sect. 6.60, page $533$, showing that $$\sum_{i=1}^\infty \sum_{j=1}^\infty \frac{\Gamma(i) \Gamma(j) \Gamma(x)}{\Gamma(i+j+x)}=\frac{1}{2}\left(\psi^{(1)}\left(\frac{x}{2}\right)-\psi^{(1)}\left(\frac{x+1}{2}\right)\right),$$ where $\psi^{(1)}(x)$ is Trigamma function.