3

$$\int_{-\infty}^\infty e^{-x²-\frac{1}{x²}}dx$$ if I do the substitution $u=x²+\frac{1}{x²}$ then as $x\to\infty$, $u\to \infty$ and as $x\to -\infty$, $u\to\infty$ since there is x squared. What do I do? I know the function is even, and I can split the integral in $2\int^\infty_0$ but that seems to get the same bounds as well. What do I do?

Please do not tell me a solution/method, I still want to try this integral on my own without hints or anything

Acyex
  • 487
  • The integral is improper at $x=0$ as it isn't defined there anyways, so you will need to break it there. Then it will be improper at both $0$ and $\infty$, so you will need to break it at some point $c$ in between, so do $0$ to $c$ then $c$ to $infty$ – Alan Aug 03 '21 at 21:14
  • Check out Glasser's master theorem. – Cameron Williams Aug 03 '21 at 21:18
  • @Alan wait so, I have to split it into $\int_{-\infty}^0+\int_0^\infty$ since it is undefined at 0 but why do I have to split it again into the c thing? – Acyex Aug 03 '21 at 21:34
  • @Cameron Williams I already have, and I'm pretty sure that's a suggestion to solve the integral, correct me if I'm wrong on that, but I've specifically mentioned not to suggest anything in the post to help me solve the problem. – Acyex Aug 03 '21 at 21:36
  • 1
    ... Then what are you even trying to accomplish here? – Cameron Williams Aug 03 '21 at 21:43
  • @Cameron Williams read the post, I'm wondering about the substitution I, myself, wrote. Is it simply not a valid substitution to make if the bounds end up being the same? – Acyex Aug 03 '21 at 21:56
  • It is a valid substitution if the integral is divided multiple integrals. – Trobeli Aug 03 '21 at 22:57
  • @Acyex If an integral is improper on both upper and lower bounds you need to split it up into two at some arbitrary cut point in between and test each separately – Alan Aug 03 '21 at 23:39
  • @Acyex: your integral is discussed in the original paper of M.L. Glasser. You can still, with pedestrian methods show that the integral converges. – Mittens Aug 04 '21 at 00:05
  • @Acyex Why are you so upset about the bounds being the same? – Arjun Vyavaharkar Aug 04 '21 at 01:48
  • @Alan The function is clearly defined at $x=0$ and is equal to $0$ there. – Arjun Vyavaharkar Aug 04 '21 at 01:53

2 Answers2

1

If your domain is $]-\infty, \infty[$, the substitution $u=x^2 + \frac{1}{x^2}$ isn't injective, hence, it gives inconsistency when computing the integral. You can solve this problem by splitting the integral.

Trobeli
  • 3,242
-1

$J=\int_{-\infty}^{\infty}e^{-x^2-\frac{1}{x^2}}dx.$ since $e^{-x^2-\frac{1}{x^2}}$ is an even function

$J=2\int_0^{\infty}e^{-x^2-\frac{1}{x^2}}dx$

substituting $x=\frac{1}{u}$ :we find $J=2\int_0^{\infty}\frac{1}{u^2}e^{-u^2-\frac{1}{u^2}}du$

from $(1) $ and $(2)$ we have :

$2J=2\int_0^{\infty }(1+\frac{1}{x^2})e^{-x^2-\frac{1}{x^2}}dx\implies J=\int_0^{\infty}(1+\frac{1}{x^2})e^{-((x-\frac{1}{x})^2+2)}dx=e^{-2}\int_0^{\infty}(1+\frac{1}{x^2})e^{-(x-\frac{1}{x^2})^2}dx$

substituting $v=x-\frac{1}{x} \implies dv =(1+\frac{1}{x^2})$ we find : $J=e^{-2}\int_{-\infty}^{\infty}e^{-v}dv=\sqrt{\pi}e^{-2}$

  • But what is wrong with my answer? Using Wolfram Alpha as well, I found the same value as I found in an answer – Bacha mohamed Aug 04 '21 at 00:02
  • (+1) There is nothing wrong with your answer in my opinion. In fact, I think you can provide s proof of Glasser's master formula using your idea. – Mittens Aug 04 '21 at 00:11
  • He clearly said that he doesn't want a solution because he wants to try to solve it by its own, so the best thing you can do is to delete your answer. – Trobeli Aug 05 '21 at 22:19