2
$\begingroup$

Consider real variables $x, y$ and a function $f(x, y) \in H^s(\mathbb{R}^2)$, say for some $s \in (0, 1)$. I am trying to get an understanding of mixed Sobolev spaces of the form $H^s_x(H^s_y)$, where the mixed Sobolev norm is defined by $$ \Vert f\Vert_{H^s_x(H^s_y)} = \Vert \Vert f\Vert_{H^s_y}\Vert_{H^s_x}.$$ In other words, $H^s_x$ functions taking values in $H^s_y$. My question is, does $f \in H^s(\mathbb{R}^2)$ automatically imply $f \in H^s_x(H^s_y)$?

It is obvious that $L^2_x(L^2_y) = L^2(\mathbb{R}^2)$, but in the case of Sobolev spaces, it is not so clear. Understanding this (relatively simple ?) case will help me understand the general theory better.

Edit: From Lions and Magenes, and also from Michael Renardy's answer below, I think that the correct thing to try to infer is $f \in H^s(\mathbb{R}^2) \Rightarrow f \in H^s_x(L^2_y)$. But it is not clear to me how to prove this.

$\endgroup$
2
  • 3
    $\begingroup$ These are called anisotropic Sobolev spaces. You can find a nice presentation in Chapter 4 (vol 2) of Lions & Magenes book. $\endgroup$ May 10, 2015 at 14:46
  • $\begingroup$ @LiviuNicolaescu Thanks for the comment. I am looking up Lions and Magenes and have edited the question accordingly. $\endgroup$
    – anonymous
    May 10, 2015 at 16:34

1 Answer 1

3
$\begingroup$

The answer to your question is no. If a function $f$ is in $H^1(R^2)$, it means that $f_x$ and $f_y$ are in $L^2$, but if $f$ is in $H^1(H^1)$, it means that $f_{xy}$ is in $L^2$.

$\endgroup$
5
  • $\begingroup$ I looked at Lions and Magenes, and I think that the correct inference should be $f \in H^s_x(L^2_y)$. Is this correct? I am running into trouble trying to prove this directly from the definition. Could you please give me a hint how to prove this? $\endgroup$
    – anonymous
    May 10, 2015 at 17:16
  • 2
    $\begingroup$ The correct inference is that $f\in H^s(L^2)\cap L^2(H^s)$. You should look at the equivalent statement for Fourier transforms. $\endgroup$ May 10, 2015 at 19:18
  • $\begingroup$ I think: $f \in H^s_x(L^2_y) \Leftrightarrow \Vert (1 + \xi^2)^{s/2}\hat{f}^x(\xi)\Vert_{L^2_y} \in L^2_\xi$, where $\hat{f}^x$ means Fourier transform with respect to the variable $x$ only. This holds iff $\Vert \Vert (1 + \xi^2)^{s/2}\hat{f}^x(\xi)\Vert_{L^2_y} \Vert_{L^2_\xi} < \infty$, which would follow from $f \in H^s$. I would really appreciate your opinion on whether I understand this correctly. $\endgroup$
    – anonymous
    May 10, 2015 at 20:21
  • $\begingroup$ The other inference $f \in L^2_x(H^s_y)$ would also follow similarly. $\endgroup$
    – anonymous
    May 10, 2015 at 20:23
  • $\begingroup$ @MichaelRenardy I read somewhere that the correct interpretation might be $f \in H^{s/2}_x(H^{s/2}_y)$ by interpolation. If this is right, could you point to a reference? Is this interpolation result in Lions and Magenes? $\endgroup$ Jun 9, 2015 at 1:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.