0
$\begingroup$

I am trying to calculate the expectation value of some quantity and it eventually boils down to the following sum (and another similar one): $$ \sum_{n=0}^{\infty} \frac{\beta^n (-\alpha)^{n-1/2}\Gamma(1/2-n,-\alpha)}{n!} $$

where $0<\beta<\alpha$ and $\Gamma$ is the partial Gamma function. Does anyone have any ideas of how to evaluate this in terms of known special functions? This would be the best case scenario. I would also be interested in numerical algorithms which approximate it. It seems like I am sampling the partial Gamma function in the most annoying parts of its domain...although the summand is related to $\gamma*$, the usual holomorphic extension function.

I have been trying to express the partial gamma function in terms of various integrals and sums, switch the order, and see what happens, with no success. I have been searching through function handbooks and partial gamma function papers trying to find useful identities, but with no luck.

$\endgroup$
1
$\begingroup$

$$\begin{eqnarray*}\sum_{n\geq 0}\frac{\beta^n}{n!}\cdot\left((-\alpha)^{n-1/2}\int_{-\alpha}^{+\infty}x^{-n-1/2}e^{-x}\,dx\right)&=&\int_{-1}^{+\infty}\sum_{n\geq 0}\frac{\beta^n}{n!}(-1)^{n-1/2}x^{-n-1/2}e^{-\alpha x}\,dx\\&=&\int_{-1}^{+\infty}-\frac{ i e^{-\frac{\beta}{x}-\alpha x}}{\sqrt{x}}\,dx\end{eqnarray*} $$ hence we have a peculiar situation: the real part of the series depends on the contribute given by $\int_{-1}^{0}$, while the imaginary part depends on the contribute given by $\int_{0}^{+\infty}$, that is the value of a Laplace transform. In particular, $$ \int_{0}^{+\infty}\exp\left(-\frac{\beta}{x}-\alpha x\right)\frac{dx}{\sqrt{x}}=\sqrt{\frac{\pi}{\alpha}}\exp\left(-2\sqrt{\alpha\beta}\right).$$ So our problem boils down to computing: $$ \int_{0}^{1}\exp\left(\frac{\beta}{x}+\alpha x\right)\frac{dx}{\sqrt{x}}=2\left(\frac{\beta}{\alpha}\right)^{1/4}\int_{0}^{\left(\frac{\alpha}{\beta}\right)^{1/4}}\exp\left[\sqrt{\alpha\beta}\left(x^2+\frac{1}{x^2}\right)\right]\,dx$$ that through the substitution $x-\frac{1}{x}=z$ is mapped into an integral of the following kind: $$ e^{2\sqrt{\alpha\beta}}\int_{-\infty}^{t}e^{z^2\sqrt{\alpha\beta}}\left(1+\frac{z}{\sqrt{4+z^2}}\right)\,dz$$ with $t>1$.

$\endgroup$
  • $\begingroup$ Thank you for your insight! I did not mention in my question because I did not know it was relevant, but I am only interested in the real part; I know the imaginary part must cancel out in another term that I have computed. $\endgroup$ – Ian Hincks Apr 19 '16 at 13:30
  • $\begingroup$ can you explain what $PV$ means? $\endgroup$ – Ian Hincks Apr 19 '16 at 13:48
  • $\begingroup$ Also I am unsure why the very first integral is $\int x^{n-3/2}e^{-x}dx$ rather than $\int x^{-n+1/2}e^{-x} dx$. $\endgroup$ – Ian Hincks Apr 19 '16 at 14:18
  • $\begingroup$ @IanHincks: you are right, I had to make some adjustements, but the outcome is pleasant anyway. $\endgroup$ – Jack D'Aurizio Apr 19 '16 at 17:17

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.