4
$\begingroup$

Contour integration is a very powerful tool.

But what if a function has no poles or zero's ?

For instance :

How to find $\int_{-\infty}^\infty \exp(-x^2) \, dx$ with contour integration?

$\endgroup$
  • 1
    $\begingroup$ You can integrate $e^{-\pi x^2}$ using contour integration. See math.stackexchange.com/questions/1266856/… $\endgroup$ – Ron Gordon Feb 15 '16 at 22:28
  • $\begingroup$ what reason do you have to think you can use Contour Integration to find this integral? $\endgroup$ – user247327 Feb 15 '16 at 22:29
  • 1
    $\begingroup$ @Jonas: Wrong. Not a duplicate - question here explicitly asks about using contour integration, no solution in your link uses it. $\endgroup$ – Ron Gordon Feb 15 '16 at 22:32
  • 1
    $\begingroup$ @Omnomnomnom: actually, the approach is to integrate $$\oint_C dz \frac{e^{-\pi z^2}}{\sin{\pi z}} $$ where $C$ is a $45^{\circ}$ parallelogram. $\endgroup$ – Ron Gordon Feb 15 '16 at 22:53
  • 1
    $\begingroup$ @Ragnar: Although the question in that link is the same, no solution actually uses contour integration. The only solution offered is quite nonsensical and useless. I would thus not mark this question as a duplicate of the previous one. $\endgroup$ – Ron Gordon Feb 15 '16 at 22:54
1
$\begingroup$

It can be found by using a rectangular contour and the function $$ \frac{e^{-z^2/2}}{1-e^{-\sqrt{\pi}(1+i)z}}. $$

See the eighth proof in here for the motivation and analysis.

$\endgroup$
1
$\begingroup$

You can perform an integration of certain functions that have no poles by performing a contour integration of another function that may have poles (or a branch cut) in order to transform the contour integral into the sought-after integral. In this case, one may consider

$$\oint_C dz \frac{e^{-\pi z^2}}{\sin{\pi z}} $$

where $C$ is a parallelogram in the complex plane, where $C= C_1+C_2+C_3+C_4$:

Along $C_1$, $z=-1/2 + e^{i \pi/4} t$, $t \in [R,-R]$.

Along $C_2$, $z=x-i (R/\sqrt{2})$, $x \in \left [-1/2-R/\sqrt{2},1/2-R/\sqrt{2} \right ]$.

Along $C_3$, $z=1/2 + e^{i \pi/4} t$, $t \in [-R,R]$.

Along $C_4$, $z=x+i (R/\sqrt{2})$, $x \in \left [-1/2+R/\sqrt{2},1/2+R/\sqrt{2} \right ]$.

One then completes the evaluation by taking the limit as $R \to \infty$. By adding the integrals along $C_1$ and $C_3$, the gaussian integral is reproduced. One then shows that the integrals about $C_2$ and $C_4$ vanish in the limit as $R \to \infty$. For details of the evaluation, see this solution.

$\endgroup$
  • $\begingroup$ I downvoted because you didn't say if in your opinion it is only a (counter)example or if there is a general method behind it, and you didn't explain how to think to that transformation of $\int_{-\infty}^\infty f(x) dx $ to $\lim_{R \to \infty} \oint_{C(R)} f(z) g(z) dz$ and how to choose $g(z)$ or "why" it is related to the former : so please tell us more or give a reference ! $\endgroup$ – reuns Feb 16 '16 at 20:00
  • $\begingroup$ @user1952009: thanks for posting that. There are other examples, I just chose to tackle the one the OP put out there. That's why contour integration techniques are at the same time so powerful and yet so difficult to master: there is so much freedom in choosing a contour and/or integrand to produce an integral. But there really isn't anything I am aware of in the way of systematic techniques. If I encounter a nonstandard example, I experiment and play and poke until I get what I want. It is a messy process few people have the patience for. If you want, I will post another example. $\endgroup$ – Ron Gordon Feb 16 '16 at 20:05
  • $\begingroup$ so you found it nearly "by chance" ? oh I just saw that Ragnar in his related answer posted a link on that trick : it has been discovered only in ~1910 or 1940 ! $\endgroup$ – reuns Feb 16 '16 at 20:06
  • $\begingroup$ @user1952009: This one? No, I learned how to do this from someone else. However, I have other examples of, say, rectangular contours and really bizarre integrands that lead to specified integrals that I did discover myself. Again, time permitting, I will get a link to you. As for the gaussian...I guess, not sure what you mean. $\endgroup$ – Ron Gordon Feb 16 '16 at 20:08

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.