Suppose $\Omega\subset \mathbb{R}^3$ is a smooth and bounded domain, and $f:\Omega\to[0,\infty]$ is a given function which is finite almost everywhere and satisfies

  • Assumption A: For all $g\in C_0^1(\Omega)$ we have the product $fg\in L^1(\Omega)$. (Here $C_0^1(\Omega)$ refers to functions which are continuously differentiable in $\Omega$ and extend continuously to $0$ on $\partial\Omega$).

Question 1: Can we show that $f\in L^1(\Omega)$?

Question 2: Does the answer to Question 1 change if we include some or all of the following assumptions:

  • Assumption B: $f$ possesses a weak derivative which is finite almost everywhere in $\Omega$;

  • Assumption C: There exists a nonnegative function $f_0 \in H^2(\Omega)\cap C(\bar{\Omega})$ such that $f-f_0=0$ (in the sense of trace) on $\partial\Omega$;

  • Assumption D: There exists a nonnegative function $h\in H^1(\Omega)$ such that $h$ is nonzero almost everywhere in $\Omega$ and $f=-\ln h$ in $\Omega$.

Note: Assumption D more or less implies Assumption B. I wrote them separately in the hopes of formulating the problem as simply as possible.

Notation: Here $H^k$ is the standard Sobolev space notation for $W^{k,2}$.

9/14/20 Edit:
Question 1 has been answered in the affirmative. I additionally pose the following

Question 3: Answer Questions 1 and 2 in the case that Assumption A is replaced by

  • Assumption A': $f\in L^1_{\text{loc}}(\Omega)$.
  • 2
    $\begingroup$ In the definition of $C^1_0$, do you mean: (i) continuously differentiable on $\Omega$, continuous on $\overline \Omega$, vanishing on $\partial \Omega$, or (ii) : continuously differentiable on $\overline \Omega$, and vanishing on $\partial \Omega$. For instance : if $\Omega$ is the unit ball, and $f(x)=\sqrt{1-\|x\|^2}$, does $f\in C^1_0$ ? $\endgroup$ Sep 12, 2020 at 19:57
  • $\begingroup$ @PietroMajer Your first definition. Thank you for pointing that out. I have edited the post to make it more clear. $\endgroup$
    – Ben Ciotti
    Sep 12, 2020 at 21:02

1 Answer 1


Let $(g_k)_{k\ge0}$ be a sequence of smooth functions such that $g_k(x)=1$ if $\text{dist}(x,\partial\Omega)\ge 2^{-k}$, $g_k (x)=0$ if $\text{dist}(x,\partial\Omega)\le 2^{-k-1}$ and $0\le g_k\le 1$ everywhere.

To prove the affirmative answer to Question 1 by contrapositive, let $f\not\in L^1(\Omega)$ be given: we want to find $g\in C^1_0(\Omega)$ such that $fg\not\in L^1(\Omega)$. We can assume $fg_k\in L^1(\Omega)$ for all $k$, otherwise we are done with $g=g_k$ for some $k$. Then $ \int_\Omega fg_k$ is an increasing sequence of positive real numbers, that diverges to $+\infty$, for if it were bounded, $f\in L^1(\Omega)$ by Beppo Levi's theorem. So for some subsequence $(g_{k_j})_j$ we have $ \int_\Omega fg_{k_{j+1}}\ge \int_\Omega fg_{k_j}+1 $, that is $ \int_\Omega f(g_{k_{j+1}}-g_{k_j})\ge 1$. For all $j$ the function $g_{k_{j+1}}-g_{k_j}$ is bounded between $0$ and $1$, and supported in the set $\big\{ 2^{-k_{j+1}-1}\le \text {dist}(x,\partial\Omega)\le 2^{-k_j}\big\}$.

But then $g:=\sum _{j\ge1} \frac{g_{k_{j+1}}-g_{k_j} }j$ is a locally finite sum of smooth functions, hence smooth in $\Omega$; clearly $g(x)\to0$ for $x\to\partial\Omega$, and, again by Beppo Levi's theorem, $\int_\Omega fg\ge \sum_{j\ge1}\frac1j=+\infty$.

  • 2
    $\begingroup$ Note that no assumption on $\partial\Omega$ are needed $\endgroup$ Sep 12, 2020 at 23:23
  • $\begingroup$ Very nice Pietro, thank you! May I ask a followup (which essentially comes down to the distinction you raised in your first comment)? Suppose I remove the assumption that $fg\in L^1$ for all $g\in C_0^1(\Omega)$, and replace it by the assumption that $fg\in L^1$ for all $g\in C_c^1(\Omega)$. Would the result still hold? Here $C^1_c(\Omega)$ is the set of continuously differentiable functions with compact support in $\Omega$. $\endgroup$
    – Ben Ciotti
    Sep 13, 2020 at 0:44
  • 2
    $\begingroup$ Isn't this the same as saying $f$ is $ L^1_{loc}(\Omega)$? $\endgroup$ Sep 13, 2020 at 6:06
  • $\begingroup$ @BenCiotti, as a matter of fact, substituting $C^1_0(\Omega)$ with $C^1_c(\Omega)$ is equivalent to say that $f$ is a locally integrable function $\endgroup$ Sep 13, 2020 at 8:52
  • $\begingroup$ Ok, so in that case, could the result be shown? Perhaps using the additional assumptions? $\endgroup$
    – Ben Ciotti
    Sep 14, 2020 at 0:17

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.