I'm following the proof of this result on Leoni, A First Course in Sobolev Spaces (theorem 7.13). Let $I\subset \textbf R$ be and interval and take $u \in W^{1,p}(I)$ where $1 \leq p < \infty$. Take $x_0 \in I$ a Lebesgue point of $u$ and define $$ \bar u(x)=u(x_0)+\int_{x_0}^x u'(t) \;\mathrm dt\,,\;\; x \in I $$ Since $u'\in L^p(I)\subset L^1_{\mathrm{loc}}(I)$, it follows that $\bar u$ is locally absolutely continuous. More precisely (using other kind results that give sufficient conditions in order to have absolute continuity like continuity and differentiability a.e. and the so called condition N of Lusin), $\bar u$ is absolutely continuous. Moreover $\bar u'=u'$ a.e and integrating against every test function $\varphi$ and using integration by parts, one has $$ \int_I (\bar u - u) \;\varphi ' \mathrm d x=0 $$ so that there exits a constant $c \in \textbf R$ such that $u - \bar u=c$ a.e.

Here there is the point that I can't get: now the text says that since $x_0$ is a Lebesgue point of $u$ and $\bar u(x_0)=u(x_0)$ then $c$ have to be zero.

What I know is that there exists a (measurable) null set $A$ such that for every $x \in I \setminus A$ the equality $u(x) - \bar u(x)=c$ holds. Then the fact that $c=0$ is proved if I know that $x_0 \notin A$. I know that there exists at least one Lebesgue point outside $A$ (the set of Lebesgue point of a locally integrable functions is such that its complement is a null set)

I know I could take the function $\bar u + c$ as continuous representative of $u$, by I'd like to understand the last step.


Let $L$ be the set of all Lebesgue points of $u$. You have proved that, given $x_0\in L$, there exist an absolutely continuous function $\overline{u}$, a constant $c$ and a null set $A$ such that $$u(x)-\overline{u}(x)=c,\quad\forall \ x\in I \setminus A.\tag{1}$$

It remains to show that $c=0$, which should be a consequence of the following two facts: (i) $\boldsymbol{x_0}$ is a Lebesgue point of $\boldsymbol{u}$ and (ii) $\boldsymbol{\overline{u}(x_0)=u(x_0)}$.

You said "there exists at least one $x\in L\cap (I\setminus A)$". Actually, almost all $x\in I$ belongs to $L\cap ( I\setminus A)$. Therefore there exists a sequence $(x_n)$ in $L\cap (I\setminus A)$ such that $$x_n\overset{n\to\infty}{\longrightarrow} x_0$$ and thus, by $(1)$, $$u(x_n)-\overline{u}(x_n)=c,\quad\forall \ n\in\mathbb N.\tag{2}$$ Since $\boldsymbol{x_0}$ is a Lebesgue point of $\boldsymbol{u}$, $u$ is approximately continuous at $x_0$ and thus $$u(x_n)\overset{n\to\infty}{\longrightarrow} u(x_0).$$ Taking the limit in $(2)$, we obtain $$u(x_0)-\overline{u}(x_0)=c.$$ Since $\boldsymbol{\overline{u}(x_0)=u(x_0)}$ we conclude that $c=0$.

  • $\begingroup$ Thanks a lot for your clear explanation! I've never encountered the concept of approximate continuity before, but I'm sure that this kind of arguments will be helpfull in my studies $\endgroup$ – Louis Oct 30 '17 at 12:30

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.