2
$\begingroup$

Can anyone give small hint for me to solve the following:

Suppose $f_n:[0,1]\rightarrow \mathbb{R}$ are $1$-Lipschitz. If $f_n\rightarrow 0$ weakly in $L^3[0,1]$, then $f_n\rightarrow 0$ strongly in $L^3[0,1]$.

My attempt: Let $x_0\in[0,1]$. We have $|f_n(x)-f_n(x_0)|\leq |x_0-x|\leq 1$. Then

$\int_0^1 |f_n|^3=\int_0^1 |f_n-f_n(x_0)+f_n(x_0)|^3\leq 8(\int_0^1 |f_n-f_n(x_0)|^3+\int_0^1 |f_n(x_0)|^3)$.

Now $f_n\rightarrow 0$ weakly in $L^3[0,1]$ means that $\int_0^1 f_ng\rightarrow 0$ for any $g\in L^{3/2}[0,1]$.

I do not know what $x_0$ and what $g$ I should choose. I am completely lost.

$\endgroup$
4
$\begingroup$

Suppose $|f_n(x) | \ge 2$ for some $x,n$, then $|f_n(x')| \ge 1$ for all $x'$ and furthermore $f_n(x), f_n(x')$ have the same sign. Hence $|\int f_n| \ge 1$. Since $\int f_n \to 0$, we can assume (by renumbering if necessary) that $|f_n(x)| \le B$ for all $x$, and for all $n$.

If we can show that $\lim_n f_n(x) = 0$ for all $x$ then the dominated convergence theorem shows that $\int |f_n|^3 \to 0$.

Suppose there is some $x'$ such that $\delta=\limsup_n |f_n(x')| > 0$. By considering $-f_n$ if necessary we can assume that there is some subsequence $n_k$ such that $f_{n_k}(x') \ge {\delta \over 2}$. Hence $f_{n_k}(x) \ge f_{n_k}(x')-|x-x'|\ge {\delta \over 2} - |x-x'|$ and so $\int_{|x-x'| \le {\delta \over 2}} f_{n_k} \ge ({\delta \over 2})^2$. By choosing $g=1_{[x'-{\delta \over 2},x'+{\delta \over 2}]}$ we get $\int g f_{n_k} \ge ({\delta \over 2})^2$, a contradiction. Hence $\lim_n f_n(x) = 0$.

(Note that this proof holds for any $p \in [1,\infty)$, not just $p=3$.)

$\endgroup$
  • 1
    $\begingroup$ Nice idea. Since $(f_{n})_{n \in \mathbb{N}}$ is uniformly bounded in $L^{\infty}([0,1])$ by the first paragraph of your response, one can apply the Arzela-Ascoli Theorem and obtain a slightly simpler proof that way. $\endgroup$ – fourierwho Jul 16 '17 at 22:00
  • $\begingroup$ @fourierwho: Thanks! I think the compactness argument ends up being about the same length, so I opted for the less technical approach. $\endgroup$ – copper.hat Jul 17 '17 at 1:05
0
$\begingroup$

The space $\text{Lip}([0,1])$ of Lipschitz continuous functions on $[0,1]$ equals the Sobolev space $W^{1,\infty}([0,1])$. Since $[0,1]$ is a finite measure space (with Lebesgue measure), $W^{1,\infty}([0,1]) \subseteq W^{1,3}([0,1])$. One can show that since the sequence of Sobolev derivatives $(f_{n}')_{n \in \mathbb{N}}$ is bounded in $L^{\infty}([0,1])$, the derivatives $(f_{n}')_{n \in \mathbb{N}}$ converge weakly to zero in $L^{3}([0,1])$. In particular, $(f_{n},f_{n}')_{n \in \mathbb{N}}$ converges weakly to zero in $W^{1,3}([0,1])$. By the Rellich-Kondrachov Theorem, it follows that $f_{n} \to 0$ strongly in $L^{3}([0,1])$.

There may be a simpler proof, particularly since we're working in one dimension, but I don't see it yet.

$\endgroup$
  • $\begingroup$ Why does $(f_n')$ bounded in $L^\infty$ imply $f_n'\rightharpoonup0$ weakly in $L^3$? Take $f_n(x)=x$ for a counterexample. $\endgroup$ – Jason Jul 16 '17 at 21:21
  • 1
    $\begingroup$ In your example $(f_{n})_{n \in \mathbb{N}}$ does not converge weakly to zero. $\endgroup$ – fourierwho Jul 16 '17 at 21:56
  • $\begingroup$ Ah, okay. So you meant $f_n\rightharpoonup0$ weakly and $(f_n')$ bounded in $L^\infty$ implies $f_n'\rightharpoonup0$ weakly? How would one show that? $\endgroup$ – Jason Jul 16 '17 at 23:17
  • $\begingroup$ Intuitively, one tests against smooth functions and then concludes. Put more slowly, one doesn't need the bound on $(f_{n}')_{n \in \mathbb{N}}$ to see that $\int_{0}^{1} f_{n}'(x) g(x) \, dx \to 0$ when $g \in C^{\infty}_{c}([0,1])$. The bound is necessary to pass from this (by density) to arbitrary functions in $L^{3}([0,1])^{*}$. Note that it's necessary and sufficient for $\|f_{n}'\|_{L^{3}([0,1])}$ to be bounded independently of $n$. $\endgroup$ – fourierwho Jul 16 '17 at 23:33

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.