3
$\begingroup$

I'm trying to evaluate the integral

$$ \int_0^\infty r^2 e^{-r^2/2\sigma^2} dr$$

Wolfram Alpha gives the answer

$$ \frac{\sqrt{\frac{\pi}{2}}}{(\frac{1}{\sigma^2})^{3/2}} $$

However, when I evaluated the integral it's undefined.

Let $u = r^2$ and $dv = e^{-r^2/2\sigma^2} dr$. Then $du = 2r \ dr$ and $v = - \frac{\sigma^2}{r} e^{-r^2/2\sigma^2}$.

$$ \int_0^\infty r^2 e^{-r^2/2\sigma^2} dr = \int_0^\infty u \ dv = u v\Big|^\infty_0 - \int_0^\infty v \ du $$

$$= \Big{[}- \sigma^2 r e^{-r^2/2\sigma^2} \Big{]}^\infty_0 + \int_0^\infty 2 \sigma^2 e^{-r^2/2\sigma^2} dr $$

For the first term,

$$ \lim_{r \to \infty} - \sigma^2 r e^{-r^2/2\sigma^2} = \lim_{r \to \infty} \frac{- \sigma^2 r }{e^{r^2/2\sigma^2}}$$

Since the numerator and denominator both tend to infinity, by l'Hospital's Rule,

$$ \stackrel{H}{=} \lim_{r \to \infty} \frac{- \sigma^2 }{\frac{r}{\sigma^2}e^{r^2/2\sigma^2}} = 0$$

For the other limit of the first term,

$$ \lim_{r \to 0} - \sigma^2 r e^{-r^2/2\sigma^2} = - \sigma^2 (0) (1) = 0 $$

So the first term is zero which means

$$ \int_0^\infty r^2 e^{-r^2/2\sigma^2} dr = \int_0^\infty 2 \sigma^2 e^{-r^2/2\sigma^2} dr = 2 \sigma^2 \Big{[}\frac{-\sigma^2}{r}e^{-r^2/2\sigma^2}\Big{]}^\infty_0$$

$$ = 2 \sigma^2 \Big{[}\lim_{r \to \infty}\frac{-\sigma^2}{r e^{r^2/2\sigma^2}} + \lim_{r \to 0}\frac{\sigma^2}{r e^{r^2/2\sigma^2}} \Big{]} = 2 \sigma^2 [0 \pm \infty]. $$

Any help is appreciated.

$\endgroup$
  • 5
    $\begingroup$ The integral of $\exp(-r^2/2\sigma^2)$ is not $-\sigma^2/r \cdot \exp(-r^2/2\sigma^2)$ $\endgroup$ – πr8 Dec 21 '16 at 19:02
6
$\begingroup$

Your error is to consider that $$ dv = e^{-r^2/2\sigma^2} dr \implies v = - \frac{\sigma^2}{r} e^{-r^2/2\sigma^2} $$ which is wrong as one may see by differentiating the latter expression.

If you want to make an integration by parts, you can rather write $$ u=r,\quad dv = re^{-r^2/2\sigma^2} dr, \quad du=dr,\quad v = -\frac1{\sigma^2}e^{-r^2/2\sigma^2}. $$

$\endgroup$
  • $\begingroup$ Ahh that makes sense. Thanks! $\endgroup$ – Y. Sargis Dec 21 '16 at 19:41
  • $\begingroup$ @Y.Sargis You are welcome. $\endgroup$ – Olivier Oloa Dec 21 '16 at 19:42
4
$\begingroup$

HINT

The way to fix it is to pick $dv = re^{-r^2/2\sigma^2}dr$ and let $u = r$

$\endgroup$
2
$\begingroup$

You dont choose the correct term in the integral by parts . You should pick $u=r$ and $v'=r e^{-r^2/2\sigma^2}$

Also, the term in the last integrals is wrong $- \sigma^2 r e^{-r^2/2\sigma^2} $.

One easy way to look at the results is to write $$\int_0^\infty r^2 e^{-r^2/2\sigma^2} dr=\frac{\sqrt{2\pi}}{\sqrt{2\pi}}\int_0^\infty r^2 e^{-r^2/2\sigma^2} dr$$ and use the change of variable $u=\frac{r}{\sigma}$

$$\int_0^\infty r^2 e^{-r^2/2\sigma^2} dr=\sigma^3\sqrt{2\pi}\frac{1}{\sqrt{2\pi}}\int_0^\infty u^2 e^{-u^2/2} dr$$

We know that if we define $X$ a standard normal variable(variance equals $1$), $E(X^2 1_{X>0})=\frac{1}{\sqrt{2\pi}}\int_0^\infty u^2 e^{-u^2/2} $ , which is also equal to $\frac{1}{2}$, by symmetry.

The final result is then $$\frac{\sigma^3\sqrt{2\pi}}{2}$$

$\endgroup$
1
$\begingroup$

$$\int_0^\infty r^2 e^{-r^2/2\sigma^2} dr$$

Let $r=\sigma\cdot x$, then we have:

$$\int_0^\infty \sigma^2x^2 e^{-x^2/2} \sigma dx=\sigma^3 \int_0^\infty x^2 e^{-x^2/2}dx$$

Now, we appeal to integration by parts:

$$\int_0^\infty x^2 e^{-x^2/2}dx=\int_0^\infty \left(-e^{-x^2/2}\right)'(x)dx$$

$$=-xe^{-x^2/2}|_0^\infty+\int_0^\infty e^{-x^2/2}\,dx=(0-0)+\sqrt{\frac{\pi}{2}}$$

by the celebrated Gaussian integral.

This gives the final solution as $\sigma^3 \sqrt{\frac{\pi}{2}}$.

$\endgroup$

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.