1
$\begingroup$

Let $\Omega\subset\Bbb R^n$ be a domain (i.e. some open subset with smooth boundary). Recall that we can define the space of distributions $\mathcal{D}'(\Omega)$ as the topological dual of $C^\infty_{\mathrm c}(\Omega)$. Furthermore, for any distribution $u\in\mathcal{D}'(\Omega)$, and for $1\le p\le\infty$, we can define the $L^p(\Omega)$ norm $$\lVert u\rVert_{L^p(\Omega)}=\sup_{\lVert\varphi\rVert_{p'}=1}\lvert\langle u,\varphi\rangle\rvert$$ which corresponds with the usual $L^p$ norm whenever $u=f\in L^p$.

Now, if $\lVert u\rVert_{L^p(\Omega)}<\infty$, then $u$ can be extended (using Hahn-Banach) to some bounded linear functional $\ell\in L^{p'}(\Omega)'$. If $p>1$, then we know that $\ell=\langle f,\cdot\rangle$ for some $f\in L^p(\Omega)$, and since $C_{\mathrm c}^\infty(\Omega)\subset L^{p'}(\Omega)$, we have that $u=f$ as distributions.

But what if $\lVert u\rVert_{L^1(\Omega)}<\infty$? Then must we have $u\in L^1$?

$\endgroup$
  • 1
    $\begingroup$ If $u=\delta_x\notin L^1$ with $x\in\Omega$, then $\|u\|_{L^1(\Omega)}=1$. $\endgroup$ – user254433 Jul 8 '17 at 23:02
  • $\begingroup$ Is the issue just that my definition of $\lVert u\rVert_{L^1(\Omega)}$ was bad, or is there really no (reasonable) way of assigning a sensible $L^1$ norm to distributions? $\endgroup$ – Monstrous Moonshine Jul 8 '17 at 23:07
  • $\begingroup$ @MonstrousMoonshine Suppose $\text{supp}\,\varphi\subset K$ and $|\langle u,\varphi\rangle|\le C(K)\|\varphi\|_{L^\infty(\Omega)}$. If $C(K)\to 0$ as $dx(K)\to 0$, then $u\in L^1$. $\endgroup$ – user254433 Jul 9 '17 at 0:04
  • $\begingroup$ What do you mean by $dx(K)$? Is $K\subset\Bbb R^n$ a compact set here? $\endgroup$ – Monstrous Moonshine Jul 9 '17 at 8:32
2
$\begingroup$

As user254433 pointed out, having finite $L^1$ norm in the distributional sense does not imply we have an $L^1$ function -- it could be a singular finite measure instead, such as the Dirac delta.

It is difficult to characterize $L^1$ functions through their actions on some other things, because $L^1$ is not the dual space of any normed space.

One can describe $L^1$ functions as measures that are absolutely continuous with respect to the Lebesgue measure $dx$. This leads to the characterization that user254433 suggested: given an open set $\Omega$, define $$ \|u\|_{1,\Omega} = \sup \{|\langle u,\varphi\rangle|\colon |\varphi|\le 1, \ \operatorname{supp} \varphi\subset \Omega \} $$ be the "restricted $L^1$ norm of $u$". Then $u\in L^1$ if and only if for every $\epsilon>0$ there exists $\delta>0$ such that $\|u\|_{1,\Omega}<\epsilon$ whenever the Lebesgue measure of $\Omega$ is less than $\delta$.

This isn't really saying much; it's just the absolute continuity of a measure expressed in terms of a distribution.

$\endgroup$

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.