1
$\begingroup$

I claim that

$$ \| u - \int_{(0,1)} u(x) \, dx \|_{L^{\infty} (0,1)} \leq \| u' \|_{L^1 (0,1)} $$

for every $u \in W^{1,1}(0,1)$ (Here, $(0,1) \subseteq \mathbb{R}$.)

I have checked that

\begin{equation} \| u - \int_{(0,1)} u(x) \, dx \|_{L^{\infty} (0,1)} \leq C \| u' \|_{L^1 (0,1)} \end{equation}

for some constant $C$ by using the Poincaré-Wirtinger inequality.

How can I show that this inequality holds for $C = 1$?

Any help will be appreciated!

$\endgroup$

1 Answer 1

5
$\begingroup$

We first get the following result formally. $$ \| u - \int_{(0,1)} u(x) \, dx \|_{L^{\infty} (0,1)} =\|\int_{0}^1 u(y)-u(x) dx\|_\infty =\|\int_{0}^1 \int_x^y u'(t) dt dx\|_\infty, $$ which implies $$ \| u - \int_{(0,1)} u(x) \, dx \|_{L^{\infty} (0,1)} \leq \int_0^1\int_0^1 |u'(t)|dt dx=\|u'\|_{L^1(0,1)} $$ The formal computation above also woks for functions in $W^{1,1}$. First, we can use the density argument: We first prove the inequality in $C^1[0,1]$ (by assuming $u\in C^1$) and obtain the general inequality by passing limit, since $C^1[0,1]$ is dense in $W^{1,1}(0,1)$. That is, we choose $\{u_n\}\subset C^1[0,1]$ such that $u_n\to u$ in $W^{1,1}$, and prove that $$ \|u_n-\int_0^1 u'_n(y)dy\|_\infty\leq \|u'_n\|_{L^1(0,1)}, $$ and then passing limit as $n\to\infty$ in both sides, which is valid because $W^{1,1}$ is embedded into $L^\infty$. This argument is standard in the theory of Sobolev space.

The second method is more elementary. We observe that $u\in W^{1,1}(0,1)$ if and only if $u$ is AC and $\frac{d u}{dt}= u'\in L^1$ a.e. More precisely, we have the following result: enter image description here

By using this result, we first prove that $$ \|\widetilde{u}-\int_0^1 \widetilde{u}(x)dx\|_\infty\leq \|u'\|_{L^{1}}. $$ Then we replace $\widetilde{u}$ by $u$ above, since they are identical almost everywhere.

$\endgroup$
8
  • $\begingroup$ Thanks! How can you know that $u(y) - u(x) = \int_{x}^{y} u' (t) \, dt$? $\endgroup$
    – cdamle
    Commented Sep 24, 2017 at 7:42
  • $\begingroup$ I presume that in the qustion, $u'$ refers to the weak derivative since we are working in $W^{1,1}$. I'm don't think your method works if that is the case. $\endgroup$ Commented Sep 24, 2017 at 7:42
  • $\begingroup$ @Zestylemonzi Yes. $u'$ is the weak derivative of $u$. $\endgroup$
    – cdamle
    Commented Sep 24, 2017 at 7:43
  • $\begingroup$ @Zestylemonzi It works, i will update why. $\endgroup$
    – Ice sea
    Commented Sep 24, 2017 at 7:44
  • 1
    $\begingroup$ @cdamle H.Brezis's "Functional analysis, Sobolev space and PDEs" $\endgroup$
    – Ice sea
    Commented Sep 24, 2017 at 8:50

You must log in to answer this question.

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