0
$\begingroup$

Let $X$ and $Y$ be compact regions of $n$- and $m$-dimensional Euclidean spaces respectively.

For any $p,q \in [1,\infty)$, define $L^{p,q}(X \times Y)$ be the space of real valued functions $f :X \times Y \to \mathbb{R}$ such that $\int_X \Bigl(\int_Y \lvert f(x,y) \rvert^q dy\Bigr)^{p/q} dx < \infty$.

Let $\phi^n : \mathbb{R}^n \to \mathbb{R}$ be the standard mollifier defined by \begin{equation} \phi_n(x):=C\exp\Bigl(\frac{1}{\lvert x \rvert^2-1}\Bigr) \chi_{\lvert x \rvert \leq 1}(x) \end{equation} where $C>0$ is chosen so that $\int_{\mathbb{R}^n} \phi^n=1$ Note that $\phi^n$ is a smooth function compactly supported on the closed unit ball.

Let $\phi^n_{\epsilon}(x):=\frac{1}{\epsilon^n}\phi_n(x/\epsilon)$ for any small $\epsilon>0$.

Similarly, we think of $\phi^m_{\epsilon}(y) : \mathbb{R}^m \to \mathbb{R}$.

For any $f \in L^{p,q}(X \times Y)$, define the convolution with $\phi^n_\epsilon \otimes \phi^m_\epsilon$ by \begin{equation} f*[\phi^n_\epsilon \otimes \phi^m_\epsilon](x,y):=\int_{X \times Y} f(x',y')\phi^n_\epsilon(x-x')\phi^m_\epsilon(y-y')dx'dy' \end{equation}

I wondder if $f * [\phi^n_\epsilon \otimes \phi^m_\epsilon]$ converges to $f$ as $\epsilon \to 0^+$ in the above $L^{p,q}$. That is, \begin{equation} \int_X \Bigl(\int_Y \bigl\lvert f*[\phi^n_\epsilon \otimes \phi^m_\epsilon](x,y) -f(x,y)\bigr\rvert^q dy\Bigr)^{p/q} dx \to 0^+ \end{equation} holds as $\epsilon \to 0^+$?

Could anyone please clarify for me?

$\endgroup$
9
  • 1
    $\begingroup$ (i) What is the "$L^{p'q}$ space"? (ii) "converges to $f$" or to $0$? What if $f=1$ on $X\times Y$? (iii) How is the convolution of $f$ with the mollifier defined, given that $f$ is only defined on $X\times Y$? (iv) What is "the standard mollifier"? $\endgroup$ Commented Jul 27, 2023 at 13:52
  • $\begingroup$ OK, I will specify all the details. Thank you. $\endgroup$
    – Isaac
    Commented Jul 27, 2023 at 14:16
  • $\begingroup$ In your definition of the convolution, you probably wanted $f(x',y')$ instead of $f(x,y)$. Also, my point (iii) is still not addressed. $\endgroup$ Commented Jul 28, 2023 at 14:52
  • $\begingroup$ Where did you see the $L^{p,q}$ spaces? Also, do you have a proof that $f * [\phi^n_\epsilon \otimes \phi^m_\epsilon]\in L^{p,q}$ if $f\in L^{p,q}$? $\endgroup$ Commented Jul 28, 2023 at 18:19
  • $\begingroup$ Ok, I think I corrected everything... I ran into $L^{p,q}$ spaces in standard real analysis textbooks. Also, I believe at least the convolution is jointly continuous, so that it belongs to $L^{p,q}$ for the compact underlying regions $X$ and $Y$. $\endgroup$
    – Isaac
    Commented Jul 28, 2023 at 19:00

0

You must log in to answer this question.

Browse other questions tagged .