2

I'm trying to do the Gamma function of 3/2, so $$\int_0^{\infty } e^{-x} \sqrt{x} \, dx$$ So far I have this u substitution $$u=\sqrt{x}$$ $$du=\frac{dx}{2 \sqrt{x}}$$ $$\int_0^{\infty } e^{-u^2} u2u \, du$$ $$2\int_0^{\infty } u^2e^{-u^2}\, du$$ and then integration by parts (I'm going to abuse some notation because I don't understand LaTeX, please forgive me) $$2\int_0^{\infty } u^2e^{-u^2}\, du$$ $$2((\frac{-\infty e^{-\infty ^2}}{2}-\frac{-0 e^{-0^2}}{2})-\int_0^{\infty } \frac{-e^{-u ^2}}{2} \, du)$$ $$\int_0^{\infty } e^{-u^2}\, du$$ After all that we are left with the familiar Gaussian integral. From what I've read on the internet, first you $$\sqrt{(\int_0^{\infty } e^{-u^2}\, du)^2}$$ $$\sqrt{(\int_0^{\infty } e^{-u^2}\, du)(\int_0^{\infty } e^{-u^2}\, du)}$$ $$\sqrt{(\int_0^{\infty } e^{-u^2}\, du)(\int_0^{\infty } e^{-v^2}\, dv)}$$ $$\sqrt{\int_0^{\infty }\int_0^{\infty } e^{-u^2}e^{-v^2}\, du\, dv}$$ $$\sqrt{\int_0^{\infty }\int_0^{\infty } e^{-(u^2+v^2)}\, du\, dv}$$ So far so good. However, I'm having trouble switching to polar coordinates here. The internet just kind of says $$r^2=u^2+v^2$$ $$dxdy=rdrdt$$ as a given, after which point it becomes literally free. I'm trying to figure out how to come up with that but I'm struggling. I've tried doing the following substitutions $$r=u\sec(t),dr=du\sec(t)$$ $$t=\arcsin(v/r),dt=\frac{dv}{r\sqrt{1-\frac{v}{r}^2}}$$ which transforms the integrand to $$e^{-r^2} r \cos ^2(t)$$ which can be integrated, but now we need to find the limits.

The integral with respect to r needs to be from 0 to infinity in order for both u and v to range from 0 to infinity. However, I can't figure out how to come up with the right limit for t. I would presume that it would be from 0 to Pi/2, since u and v both range from 0 to infinity. I presume that should mean that we are integrating just the 1st quadrant. But unfortunately, $$\int_0^{\frac{\pi }{2}} \left(\int_0^{\infty } e^{-r^2} r \cos ^2(t) \, dr\right) \, dt=\frac{\pi }{8}\neq\frac{\pi }{4}$$ Setting the upper limit to Pi gets us the right answer of Pi/4, but the only reason I know that is because I already know what the answer is supposed to be. I'm still in the dark as to why that's the limit. Would anyone know what I'm doing wrong with the substitutions or the limits?

  • Possible duplicate of https://math.stackexchange.com/questions/9286/proving-int-0-infty-mathrme-x2-dx-frac-sqrt-pi2?lq=1 – Brevan Ellefsen Sep 12 '18 at 07:43
  • Well, if, as you write yourself, $$r^2=u^2+v^2\qquad dudv=rdrdt$$ then $$\iint e^{-(u^2+v^2)}dudv=\iint e^{-r^2}rdrdt$$ on suitable domains, and not what you wrote. – Did Sep 12 '18 at 07:48
  • I know but like where does the $dudv=rdrdt$ come from? Like I said earlier, every source I've found just kind of takes that as well as the new bounds as a given and I have no idea how to come up with that. – Nevan Lowe Sep 12 '18 at 07:52

1 Answers1

0

Use the coordinate transformation (cartesian to polar): $[u,v] = [r\sin(\phi),r\cos(\phi)]$. Then the Jacobian reads: $$J = \left[\matrix{\sin(\phi)&&r\cos(\phi) \\ \cos(\phi)&&-r\sin(\phi)}\right]$$ with the determinant $|J| = r$. From the measure-theory we know $du\wedge dv = |J|dr\wedge d\phi$.

If you are not familiar with measure-theory: Notice, the Jacobian determinant plays the role of a volume element distortion factor due to coordinate transformation. Also check out https://en.wikipedia.org/wiki/Jacobian_matrix_and_determinant.

denklo
  • 781
  • While one can certainly derive this using measure theory, I'm not sure viewing this from the perspective is measure theory is the best approach for a beginner. Any good introductory level multivariable-calculus text will prove the role of the Jacobian in coordinate transformation without appealing to measure theory. That being said, having the perspective of measure theory doesn't hurt :) (I would also hesitate to use differential forms, as that isn't even a measure theory concept - more of a differential geometry concept, though that's a bit hard to define) – Brevan Ellefsen Sep 12 '18 at 07:51