2

It is a homework problem.

Function $f:\mathbb{R}^2\rightarrow\mathbb{R}$,$\Omega$ is a disk at origin with radius as $\dfrac{1}{2}$.

$f(x) = \log(\log(\dfrac{1}{|x|}))$, where $x\in\Omega$(i,e. $|x|\in (0,1/2)$),

it is easy to show that $f\in H^1(\Omega)$, the problem says, find a optimal $s\ge 1$, such that

$f(x)\in H^s(\Omega)$.

Just some hints will be ok.

The proof for it be to in $H^1(\Omega)$ is here

Thanks!

Yimin
  • 3,311
  • 17
  • 32

2 Answers2

0

I think there is a problem at $0$ that you should look at. Also, first check your assertion that this function belongs in $H^1$. A general suggestion is: use polar coordinates, substitute $r = e^{-e^t}$, so that $f(r) = t$, check Sobolev space inclusions for integer points and use the Sobolev interpolation results to conclude for numbers in between integers.

MBM
  • 571
  • There is no problem for it to be in $H^1 = W^{1,2}$. see http://books.google.com/books?id=Od5BxTEN0VsC&pg=PA173&lpg=PA173&dq=loglog+r+Sobolev+space&source=bl&ots=o5DR7-AhT-&sig=Gji1jdpRcFD0SGmdy4X27M-1KDs&hl=en&sa=X&ei=yg02UZWwGKTS2AW9rIG4AQ&ved=0CFEQ6AEwBA#v=onepage&q=loglog%20r%20Sobolev%20space&f=false – Yimin Mar 06 '13 at 00:15
  • What I need is the existence of an $s>1$ for it to be in $H^s$. – Yimin Mar 06 '13 at 00:17
0

I think $s=1$ is the optimal solution.

If $s>1$, since we can also manage to show that $f$ does not belong to $H^2$, then $1<s<2$.

$$\|f\|_{H^s} = \|f\|_{H^1}+(\int_D\int_D \dfrac{|Df(x)-Df(y)|^2}{|x-y|^{2+2(s-1)}}dxdy)^{1/2}=\|f\|_{H^1}+(\int_D\int_D \dfrac{|Df(x)-Df(y)|^2}{|x-y|^{2s}}dxdy)^{1/2}$$

where $x\in D,y\in D$.

$f(x) = \log\log(\dfrac{1}{r})$, $Df = -\dfrac{\mathbf{r}}{r^2\log r}$, we can see that even we choose $(s-1)$ very small, the integral will give you something as

$$\int_{0}^{1/2} \dfrac{1}{r^4(\log r)^2}dr$$

which is not bounded.

Thus $f\in H^1(\Omega)$, which is the optimal sobolev space.

Yimin
  • 3,311
  • 17
  • 32