2
$\begingroup$

I'm studying Functional Analysis by myself. For a counterexample of every Banach space is not weakly sequentially complete, I was suggested to check $C([0,1])$ is not weakly sequentially complete. For this, let $f_n(t)=1-nt$ if $0\leq t\leq \frac{1}{n}$ and $f_n(t)=0$ if $\frac{1}{n}\leq t \leq 1$. But I do not know how to use monotone convergence theorem to show $\{f_n\}$ is weakly Cauchy and also why it is not weakly sequentially complete. Please help me. Thanks in advance.

$\endgroup$
1
$\begingroup$

Let $x^*\in C([0,1])^*$ be any continuous functional. Then there is a finite measure $\mu$ on $[0,1]$, such that $$ \int_{[0,1]} f\, d\mu = x^*(f), \qquad f \in C[0,1] $$ Now, by dominated convergence, we have $$ \int_{[0,1]} (f_n - f_m) \, d\mu \to 0 ,\qquad n,m \to \infty $$ Edit: By looking at $\mu = \delta_x$, $x \in [0,1]$, show that $(f_n)$ cannot converge weakly to a continuous function.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ I thought $f_n \rightharpoonup 0$ too for a moment, but consider $\delta_0 \in C([0,1])^\ast$. $\endgroup$ – Daniel Fischer Jul 14 '14 at 9:42
  • $\begingroup$ @DanielFischer Thanks. $\endgroup$ – martini Jul 14 '14 at 10:00
4
$\begingroup$

Fact 1. Note that $c_0$ is not weakly sequentially complete. Just consider sequence $\left(\sum_{k=1}^n e_k\right)_{n\in\mathbb{N}}$.

Fact 2. Any $C(K)$ for infinite Hausdorff compact $K$ contains a copy of $c_0$.

Fact 3. The property of being weakly sequentially complete is preserved by subspaces. Indeed, let $Y$ be a closed subspace of weakly sequentially complete space $X$. Consider weakly Cauchy sequence $(y_n)_{n\in\mathbb{N}}$ in $Y$, then obviously it is weakly Cauchy in $X$. Since $X$ is weakly sequentially complete then $(y_n)_{n\in\mathbb{N}}$ weakly converges to some $x\in X$. Thus $x$ lies in the weak closure of $Y$. Since $Y$ is closed and convex, then by Mazur's theorem weak closure coincides with $Y$. Thus weak limit of $(y_n)_{n\in\mathbb{N}}$ $\in Y$, and $Y$ is weakly sequentially complete.

Conclusion. Any $C(K)$ is not weakly sequentially complete.

It is standard trick to show that a Banach space contains a copy of $c_0$ and then conclude that this space is not weakly sequentially complete.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ I think C(K) is a subspace of $L^\infty(K)$ and $L^\infty(K)$ is weakly sequentially complete while C(K) is not. it's a counter example for fact3. isn't it? $\endgroup$ – niki Jul 14 '14 at 11:18
  • $\begingroup$ Why is $L_\infty(K)$ is weakly sequentially complete? I think it is not. $\endgroup$ – Norbert Jul 14 '14 at 11:22
  • $\begingroup$ Oh, In the old version of Conway's Functional Analysis it's an exercise. but checking in new version, I found $L^\infty$ is weak* sequentially complete. $\endgroup$ – niki Jul 14 '14 at 11:40
1
$\begingroup$

The monotone convergence theorem also works for decreasing sequences $\{f_n\}$ provided $f_1\in L^1$ (cf. Rudin's Real and Complex Analysis, exercises to chap.1), which is the case in your problem. For any regular Borel measure $\mu$ on $[0,1]$, $\int_0^1 f_n d\mu \to \mu(\{0\})$, so the sequence is weakly convergent and hence weakly Cauchy. Here we are using the basic property of finite measures that $\lim_{n\to\infty}\mu([0,1/n])= \mu(\cap_{n=1}^\infty[0,1/n])$. Check that $f_n$ converges weakly to the (discontinuous) characteristic function $\chi_{\{0\}}$.

| cite | improve this answer | |
$\endgroup$

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.