2
$\begingroup$

Let $C$ be a compact convex set in $\mathbb{R}^d$. Let the origin $O$ by in the internal of $C$. The gauge function $\gamma_C(.) : \mathbb{R}^d \to [0, \infty]$ of $C$ is defined as $$ \gamma_C(x) = \inf\{t : x \in t \cdot C\}. $$ The radial function is defined as a reciprocal of gauge function or, equivalently, $$ 1 / \gamma_C(x) = \sup\{x : t \cdot x \in C\} $$ I'm considering the following integral: $$ I(C) = \int_{S^{d - 1}} |1 / \gamma_C(u)| du. $$

I have two following notions:

  1. Intuitively, it looks like a volume of $C$: we integrate over all directions and summarize all distances from the origin $O$ to the boundary of $C$.

  2. However, if we consider the body $a \cdot C$ for some constant $a > 0$, we will have $1 / \gamma_{a \cdot C}(u) = a \cdot 1 / \gamma_C(u)$, and: $$ I(a \cdot C) = \int_{S^{d - 1}} |a \cdot 1 / \gamma_C(u)| du = a \int_{S^{d - 1}} |1 / \gamma_C(u)| du = a \cdot I(C). $$ From this equality, we have that $I(C)$ is not a volume of $C$ because $Vol(a \cdot C) = a^3 \cdot Vol(C)$.

So, which of (1) and (2) is wrong?

And, if (1) is wrong, the next question: can we construct a volume of $C$ from the integral of its radial function?

$\endgroup$
1
$\begingroup$

1 is wrong. For example consider the unit disk in $\mathbb{R}^2$. Then the integral is $\int_0^{2\pi} 1= 2\pi$. The volume, however, is of course $\pi$. The problem is, like you point out in 2, that the volume is not actually linear with respect to scaling. The integral treats the area of a disc like a rectangle on the circle, but it’s actually much more like a triangle. (With both sides depending on the scaling factor.)

The problem is that the infinitesimal volume element in spherical coordinates is not $d\rho du$ where $du = vol_{S^{d-1}}$ is the volume form of $S^{d-1}$ and $\rho$ is the radius. You need to scale with the Jacobi determinant when you change coordinates.

$\endgroup$
7
  • $\begingroup$ Thanks! What do you mean by this? "The integral treats the area of a disc like a rectangle on the circle, but it’s actually much more like a triangle. (With both sides depending on the scaling factor.)". Can you clarify this statement? What is a rectangle on the circle? $\endgroup$ Aug 6 '20 at 9:09
  • 1
    $\begingroup$ Fair enough, that was pretty unclear. Let $C$ be the disc of radius $a$. Then $\int_{S^1}| 1/\gamma_C(U)|du= 2\pi a$ I.e the area of a rectangle with sides $a$ and $2\pi$. The correct area/volume is however $\pi a^2=\frac{1}{2}(2\pi a) a$. $\endgroup$
    – Knaus
    Aug 6 '20 at 9:20
  • $\begingroup$ The infinitesimal volume element is as follows, right? $dV = \rho^2 sin \theta d\rho d\phi d\theta = \rho^2 d\rho du$. Then we can say that volume is $\int_{S^{d - 1}} \int_0^\rho \rho^2 d\rho du = \frac{1}{3} \int_{S^{d - 1}} \rho^3 du = \frac{1}{3} \int_{S^{d - 1}} |1 / \gamma_C(u)|^3 du$. Is this correct? $\endgroup$ Aug 6 '20 at 9:54
  • 1
    $\begingroup$ I think this is correct if $d=3$. $\endgroup$
    – Knaus
    Aug 6 '20 at 9:59
  • 1
    $\begingroup$ Yes, the volume element will have $\rho^{d-1}$ as a factor. $\endgroup$
    – Knaus
    Aug 6 '20 at 10:12

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.