1
$\begingroup$

Let $\Omega \subset \mathbb{R}^n$ be some open set. If, for all $\psi \in L^2(\Omega)$ and some fixed integral kernel $k \in L^2(\Omega\times \Omega)$ and $\ell>0$, it is true that both $\int_{\Omega} k(\cdot,y) \psi(y)\,dy$ and $\int_{\Omega} k(y,\cdot) \psi(y)\,dy$ are $H^\ell$-functions.

Does it follow hat $k \in H^\ell(\Omega \times \Omega)?$ I doubt that this is true, because componentwise differentiability should probably not include the common differentiability. However, I am looking for the best regularity we get for $k$ with this property. (Maybe also classical properties like continuity or existence of partial derivatives could be established).

$\endgroup$

1 Answer 1

1
$\begingroup$

The assumptions imply that $k\in H^\ell(\Omega)\hat\otimes L^2(\Omega) \cap L^2(\Omega) \hat\otimes H^\ell(\Omega)$, where $\hat\otimes$ is the (completed) Hilbert space tensor product. This space equals $H^\ell(\Omega_x;L^2(\Omega_y))\cap L^2(\Omega_x;H^\ell(\Omega_y))$ which, by interpolation, is contained in $H^{\ell-t}(\Omega_x;H^t(\Omega_y))$, for any $0\leq t\leq \ell$. So, indeed, your kernel belongs to $H^\ell(\Omega\times \Omega)$.

$\endgroup$
1
  • $\begingroup$ many thanks, I did not think of this view at all. $\endgroup$
    – Kinzlin
    Commented Feb 15, 2017 at 9:53

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .