3
$\begingroup$

Short version:

Can anyone do $$ \int dxe^{-x^{2}}\sqrt{2x+a} $$ ?


Long Version:

In an answer to Integrating a product of exponentials and error functions the integral below is evaluated using a differentiation technique:

Let $\mathcal{I}(\gamma)= \int_0^\infty x^2\exp(-\delta x^2)\operatorname{erf}(\gamma x)\, \mathrm{d} x$, then $I^\prime(\gamma) = \frac{2}{\sqrt{\pi}} \int_0^\infty x^3 \exp(-x^2 \left( \delta + \gamma^2 \right) ) \mathrm{d} x = \frac{1}{(\gamma^2 + \delta)^2} \frac{1}{\sqrt{\pi}}$. Thus: $$ I(\gamma) = \int_0^\gamma \frac{1}{\sqrt{\pi}} \frac{1}{(\gamma^2+\delta)^2} \mathrm{d} \gamma = \frac{\gamma }{2 \sqrt{\pi } \delta \left(\gamma ^2+\delta \right)}+\frac{1}{2 \sqrt{\pi } \delta ^{3/2}} \arctan\left(\frac{\gamma }{\sqrt{\delta }}\right) $$

I have used this method in order to evaluate:

$$ \int_{0}^{\infty}\int_{0}^{\epsilon+\frac{1}{2r}x^{2}}e^{-\beta\left(x^{2}+y^{2}\right)}dydx $$

and my final step involves:

$$ \int dxe^{-x^{2}}\sqrt{2x+a} $$

$a$ is just some constant.

1) Do I need limits on my final integral like in Integrating a product of exponentials and error functions, i.e. $$ \int_{0}^{x}dxe^{-x^{2}}\sqrt{2x+a} $$

2) Can anyone help me evaluate my final integral, whether its indefinite or not? If you can't answer this question completely, does anyone know how to do the final integral $\int dxe^{-x^{2}}\sqrt{2x+a}$?

$\endgroup$
  • $\begingroup$ Just to be clear, are you looking for a primitive, or do you wish to evaluate that definite-integral on some interval? $\endgroup$ – Jonathan Y. Feb 14 '14 at 23:00
  • 1
    $\begingroup$ hopefully a primitive, though the long version indicates that I might need to integrate over an interval (the referenced question integrates over 0 to gamma). If you can get a primitive that would be good. $\endgroup$ – Alexander Kartun-Giles Feb 14 '14 at 23:08
1
$\begingroup$

I doubt there's any kind of closed form primitive in terms of common special functions, but we can try to find a series solution for $t < \infty$:

$$\int_0^t e^{-x^2}\sqrt{x+b}~dx=\sum_{k=0}^\infty \frac{(-1)^k}{k!} \int_0^t x^{2k} \sqrt{x+b}~dx$$

The integral can be transformed to yield a hypergeometric function:

$$\int_0^t x^{2k} \sqrt{x+b}~dx=\sqrt{b} ~t^{2k+1} \int_0^1 y^{2k} \left(1+\frac{t}{b} y\right)^{1/2}~dy= \\ =\frac{\sqrt{b} ~t^{2k+1}}{2k+1} {_2F_1} \left(-\frac{1}{2};2k+1;2k+2;-\frac{t}{b} \right)$$

Which gives us for the original integral:

$$\int_0^t e^{-x^2}\sqrt{x+b}~dx=\sqrt{b} ~t~\sum_{k=0}^\infty \frac{(-1)^k~t^{2k}}{k!(2k+1)}{_2F_1} \left(-\frac{1}{2};2k+1;2k+2;-\frac{t}{b} \right) \tag{1} $$

The hypergeometric function is elementary for integer $k$, but I haven't been able to find a complete closed form. The first few are:

$$\frac{1}{2}~{_2F_1} \left(-\frac{1}{2};1;2;-p \right)=\frac{(1+p)^{3/2}-1}{3p}$$

$$\frac{1}{2}~{_2F_1} \left(-\frac{1}{2};3;4;-p \right)=\frac{(1+p)^{3/2}(8-12p+15p^2)-8}{35p^3}$$

$$\frac{1}{2}~{_2F_1} \left(-\frac{1}{2};5;6;-p \right)=\frac{(1+p)^{3/2}(128-192p+240p^2-280p^3-315p^4)-128}{693p^5}$$

Some of the coefficients can be found in http://oeis.org/A061549 and http://oeis.org/A001803.


Using a simple susbsitution $y=1-z$ we obtain another form of the integral, resulting in another hypergeometric function:

$$\int_0^t e^{-x^2}\sqrt{x+b}~dx=\sqrt{t+b}~ ~t~\sum_{k=0}^\infty \frac{(-1)^k~t^{2k}}{k!(2k+1)}{_2F_1} \left(-\frac{1}{2};1;2k+2;\frac{t}{t+b} \right) \tag{2}$$

Numerically, this series converges slower than the first one. However, by looking at the denominators, it's clear that for exotic case $t=-b$ we should use $(1)$, while for the case $b=0$ we should use $(2)$.

$\endgroup$
  • $\begingroup$ Yes interesting, thank you. I will try this now. $\endgroup$ – Alexander Kartun-Giles Apr 14 '18 at 11:36

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.