Prove that the following integral: $$\int_0^{\infty} \int_0^{\infty} e^{-(x^2+y^2+2xy \cos \theta)} \,dx dy = \frac{\theta}{2\sin\theta}$$

The hints written on the book are beta function and to suppose the slanted $xy$ plane. However, I cannot figure out a thing.

  • $\begingroup$ Have you tried polar coordinates ? $\endgroup$
    – Lucian
    Dec 28 '14 at 10:27
  • $\begingroup$ Suggestion: Complete the square in the argument of the exponential function: $x^2+y^2+2xy\cos\theta=(x+\cos\theta y)^2+(1-\cos^2\theta)y^2$. Then, it might be easier to do the change of variables... (note that $(1-\cos^2\theta)=\sin^2\theta$) $\endgroup$
    – mickep
    Dec 28 '14 at 10:31

Assume $\theta\in\left[0,\frac{\pi}{2}\right]$. Then, by setting $z=x+y\cos\theta$ we have: $$ I = \int_{0}^{+\infty}\int_{y\cos\theta}^{+\infty} e^{-(z^2+y^2\sin^2\theta)}\, dz \,dy = \frac{1}{\sin\theta}\int_{0}^{+\infty}\int_{w\cot\theta}^{+\infty}e^{-(z^2+w^2)}\,dz\,dw.$$ Switching now to polar coordinates we get $I=\frac{\theta}{2\sin\theta}$ as stated.

  • $\begingroup$ Where is beta function here, or was that a red herring? $\endgroup$ Dec 28 '14 at 11:01
  • $\begingroup$ @SuzuHirose: IMHO, it is not needed to use the Euler Beta function here, just a simple geometric argument. The last integral is the integral of $e^{-\rho^2}$ over an angle with vertex in the origin of $\mathbb{R}^2$, hence you just need to compute the amplitude of such angle. $\endgroup$ Dec 28 '14 at 12:18
  • $\begingroup$ @JackD'Aurizio why is it necessary for $\theta\in (0,\pi/2]$? $\endgroup$
    – take008
    Jul 20 '19 at 13:56
  • $\begingroup$ The stated identity does not hold unconditionally (for instance it does not hold if theta is positive but sin theta is negative), so I assumed a reasonable constraint. $\endgroup$ Jul 20 '19 at 15:59

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.