1
$\begingroup$

Super dumb question but I can't seem to figure out what I'm doing wrong. Let $E$ be some set and $E_r= \{ x \mid rx \in E\}$ for $r>0$. Then if $\phi \in C^1_c(\mathbb{R}^n;\mathbb{R}^n)$ $$\int_{E_r} \operatorname{div}(\phi(x))\,dx = \frac{1}{r^{n-1}}\int_E \operatorname{div}(\phi(y/r)) \, dy.$$ I applied the change of variables $x = y/r$ so then $dx = dy/r^n$ and the new domain of integration becomes $E$, but I don't know how to get this $n-1$ term in place of $n$.

$\endgroup$
3
  • $\begingroup$ What's the subscript $c$ in $C^1_c$? $\endgroup$ – Jackozee Hakkiuz Jun 5 at 21:20
  • $\begingroup$ compact support $\endgroup$ – asuuuka Jun 5 at 21:21
  • $\begingroup$ I edited my answer, I hope this helps! $\endgroup$ – Jackozee Hakkiuz Jun 5 at 22:10
3
$\begingroup$

$\renewcommand\div{\mathrm{div}}$ Remember that the change of variables says $$\int_{g(E)} f = \int_E (f\circ g) \cdot \;|\det g'|$$

Now let $g(y)=y/r$, so that $$E_r=\{x\mid rx\in E\}=\{g(y)\mid y\in E\}=g(E)$$ Then $$\begin{align*} \int_{g(E)} \div\phi &= \int_{E} ((\div\phi)\circ g)\cdot |\det g'| \end{align*}$$ and you can calculate that $$\begin{align*} \div(\phi\circ g) &= \frac{1}{r}(\div\phi)\circ g \\[1mm] |\det g'| &= \frac 1 {r^n} \end{align*}$$ Solving from the first formula, you get $(\div\phi)\circ g = r\cdot \div(\phi\circ g)$, so that $$\begin{align*} \int_{E_r} \div\phi &= \int_{E} r\cdot ((\div\phi)\circ g)\cdot \frac 1{r^n} \\ &= \frac{1}{r^{n-1}}\int_{E} (\div\phi)\circ g \end{align*}$$ as you wanted. Or, in the notation $\int_E f(x)\,dx=\int_E f$ $$\begin{align*} \int_{E_r} (\div\phi)(x)\,dx &= \frac{1}{r^{n-1}}\int_{E} (\div\phi)(g(y))\,dy. \end{align*}$$

I think a possible source of your confusion is the notation $\div(\phi(x))$. Let me explain.

What we do is that we first apply the divergence operator $\div$ to the function $\phi:\mathbb R^n\to\mathbb R^n$, which yields another function $\div\phi:\mathbb R^n\to\mathbb R$. Then you evaluate $\div\phi$ at $x\in E_r$, giving you a real number $(\div\phi)(x)$. I'm of the opinion that you should not denote $(\div\phi)(x)$ as $\div(\phi(x))$.

First: if you take notation seriously, then $\div(\phi(x))$ should be zero, since $\phi(x)$ is the value of $\phi$ at $x$, so that it is already a point in $\mathbb R^n$ and hence its derivative is zero.

Second, and this the worst: if you write $(\div\phi)(x)$ as $\div(\phi(x))$, now the notation $\div(\phi(y/r))$ is confusing, because some people may interpret it as $(\div(\phi\circ g))(y)$, where $g(y)=y/r$ (as before) and some other may interpret it as $(\div\phi)(g(y))$. My hypothesis is that this made you think that $\div(\phi\circ g)$ is the same as $(\div\phi)\circ g$, which is not true: as you have seen, they differ by a factor of $r$, which is precisely the missing $r$.

$\endgroup$

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.