0
$\begingroup$

I've found an article that essentially states that an integral of the form

$$ I[f] = \int_{\mathbb{R}^N} f(x_1,\ldots,x_N) dx_1 \ldots dx_N = \int_{\mathbb{R}^N} f(r) dx_1 \ldots dx_N $$

where $r = \sqrt{x_1^2 + \ldots x_N^2}$, can always be split as product of two integrals, where the first integral is the surface of the $n$ dimensional sphere, say $A_n$ and the second integral is the radial integral

$$ R[f] = \int_0^1 f(r) r^{N-1}dr $$

The article I've found only states the result, but it doesn't actually prove it, but assuming this results holds it proves that

$$ A_N = \frac{2 \pi^{N/2}}{\Gamma\left( \frac{N}{2}\right)} $$

Is there a way to prove the statement which doesn't rely on generalized polar coordinates, which are a bit hard to remember and manipulate, if there's no other way is there an easy way to derive to generalized polar coordinates and again derive the result?

I would assume the proof is a calculus fact.

$\endgroup$
0
$\begingroup$

Since $f$ is radially symmetric, $f(\vec{x})d^N\vec{x}=r^{N-1}f(r)drd\Omega$, with $d\Omega$ an $(N-1)$-dimensional infinitesimal element over the angle coordinate(s). We don't need to work out how $d\Omega$ looks at all, so don't get out your $\sin\theta$s here. The claimed factorisation is immediate.

If you prefer a different proof that an $(N-1)$-sphere of radius $R$ has measure $\frac{2\pi^{N/2}}{\Gamma(N/2)}R^{N-1}$, let's prove instead that an $N$-ball of radius $R$ has volume $\int_0^R\frac{2\pi^{N/2}}{\Gamma(N/2)}r^{N-1}dr=\frac{\pi^{N/2}}{\Gamma(N/2+1)}R^N$. (To clarify, an $n$-ball is $n$-dimensional, but an $n$-sphere is the $n$-dimensional surface of an $(n+1)$-ball.) Equivalently, the unit $N$-ball has measure $V_N:=\frac{\pi^{N/2}}{\Gamma(N/2+1)}$. We'll proceed by induction.

The result is correct for $N=0$; a "$0$-sphere" is the single point in $0$-dimensional space, and the definition of measure is counting that point. (If that argument seems to strange, take $N=1$ as your base step, for which measure is a line segment's width, or failing that $N=2$, for which you just want a circle's area.) By slicing an $N$-ball into hypercylinders of infinitesimal thickness with $(N-1)$-ball cross sections of radius $\sqrt{1-x^2}$,$$V_N=\int_{-1}^1V_{N-1}(1-x^2)^{(N-1)/2}dx=2V_{N-1}\int_0^1(1-x^2)^{(N-1)/2}dx\\=V_{N-1}\int_0^1y^{-1/2}(1-y)^{(N-1)/2}dx=V_{N-1}\operatorname{B}\left(\frac12,\,\frac{N+1}{2}\right).$$(The intuition of this is easiest with slicing a sphere into cylinders, since you can visualise that.) So the inductive step is$$V_{N-1}=\frac{\pi^{(N-1)/2}}{\Gamma((N+1)/2)}\implies V_N=\frac{\pi^{(N-1)/2}}{\Gamma((N+1)/2)}\frac{\sqrt{\pi}\Gamma((N+1)/2)}{\Gamma(N/2)+1},$$which is what we needed.

$\endgroup$
  • $\begingroup$ How did you get your very first equality? $\endgroup$ – user8469759 Oct 19 '19 at 18:22
  • $\begingroup$ @user8469759 Well, that's just the polar form of $d^N\vec{x}$. It needs to be proportional to $d(r^N)$, because an $N$-ball of radius $R$ wouldn't otherwise have measure $\propto R^N$. $\endgroup$ – J.G. Oct 19 '19 at 18:24
  • $\begingroup$ So you implicity applied a coordinate transformation am I right? $\endgroup$ – user8469759 Oct 19 '19 at 18:28
  • $\begingroup$ @user8469759 Yes, like when you write $f(x)dx=f(x)g^\prime(u)du$ if $x=g(u)$. $\endgroup$ – J.G. Oct 19 '19 at 18:30
  • $\begingroup$ Ok, but to prove that equality you want to know the coordinate transformation. Or proving the existance maybe. I wanted to avoid an explicit computation if possible. $\endgroup$ – user8469759 Oct 19 '19 at 21:38

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.