8
$\begingroup$

Suppose $U$ is an open bounded set with $C^1$ boundary. It is a well-known result in the theory of Sobolev spaces $W^{1,p}$ that there is a continuous linear operator $T:W^{1,p}(U)\rightarrow L^p(\partial U )$ that equals ordinary restriction on continuous functions. Wikipedia tells me that this operator is in general not surjective. I know that one way to prove this is to show that functions in the image are actually more regular than your general $L^p$-function, and I have somewhat understood the proof for that. However, I feel that there should be some elementary counterexample. In fact, any function that is a trace of some $u\in W^{1,p}(U)$ is a limit of a Cauchy sequence of functions in $C^{\infty}(\bar{U})$ (and vice versa), and that could be exploited to exhibit some example where $T$ is not surjective. Does anyone have a good example?

$\endgroup$
5
$\begingroup$

At least in the best circumstances, it is easy to prove that the trace/restriction map loses ${\ell\over p}+\epsilon$ in "Sobolev units" (for arbitrarily small $\epsilon>0$) where codimension is $\ell$. And an easy extension of Sobolev imbedding theorems shows that (for example, for $L^2$ Sobolev spaces so I don't mess up the indexing shift...) $H^{k+{n\over 2}+\epsilon}\subset C^{k,\epsilon}$, the latter being $C^k$ functions with an $\epsilon$ Lipschitz condition.

Thus, with $p=2$ for example, $H^1(\Omega)$ maps to $C^{0,{1\over 2}-\epsilon}(\partial \Omega)$ for every $0<\epsilon<{1\over 2}$.

$\endgroup$
2
  • $\begingroup$ Hi. Do you know any reference I can check for the case of codimension greater than 1? $\endgroup$ – yess Apr 21 '14 at 5:07
  • 1
    $\begingroup$ Iterating the codimension-one case should cover the general-codimension case. $\endgroup$ – paul garrett Apr 21 '14 at 12:24
1
$\begingroup$

The image of $T(W^{1,p}(\Omega))$ is $L^{1}(\partial\Omega)$ if $p=1$ and the fractional Sobolev space $W^{1-1/p}(\partial\Omega)$ if $1<p<\infty$. You can find the theorems in Adams and Fournier "Sobolev spaces", Leoni "A First Course in Sobolev Spaces", Necas "Direct Methods in the Theory of Elliptic Equations". The case $p=1$ is in the last two books.

$\endgroup$
5
  • $\begingroup$ Are you able to specify which results, exactly, in the last two cases detail the $p=1$ case, please? Do you know if any of these texts provide a direct proof - or counterexample - that when $p\ge2$ the trace is not surjective on $L^{1}(\partial\Omega)$ (I'm not so concerned atm with the fractional case)? $\endgroup$ – Sarah Paulson's Scream Jan 26 at 12:53
  • $\begingroup$ I don't understand your question. When $p\ge 2$ The trace of a function in $W^{1,p}(\Omega)$ belongs to the fractional Sobolev space $W^{1-1/p,p}(\partial \Omega)$. There are many functions in $L^1(\Omega)$ which do not belong to this space. $\endgroup$ – Gio67 Jan 27 at 0:37
  • $\begingroup$ Sorry, let me phrase them another way; (i) can you indicate which results from Leoni "A First Course in Sobolev Spaces" and Necas "Direct Methods in the Theory of Elliptic Equations" detail equality for the $p=1$ case, please? (ii) Is there a reference (perhaps one of the books you noted) for a standard proof or counterexample that this trace is not surjective which doesn't rely on invoking the fractional Sobolev Spaces? I understand these might sound moronic, but this is not an area with which I am very well versed - perhaps apparent. $\endgroup$ – Sarah Paulson's Scream Feb 1 at 16:39
  • 1
    $\begingroup$ Secon edition in Leoni: Theorem 18.18. First edition: Theorem 15.6 for a half space. Necas 2012 edition there is only a remark for $p=1$. Remark 5.5. No proof. As for (ii) perhaps the best way is to do $p=2$ and use Fourier transforms. See Theorem 5.1 in Necas page 87. It’s giving conditions for the trace space in terms of the Fourier transform. $\endgroup$ – Gio67 Feb 2 at 11:16
  • $\begingroup$ I'll check this out - thanks for getting back to me on it! $\endgroup$ – Sarah Paulson's Scream Feb 2 at 13:11

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.