1
$\begingroup$

Let $X,Y$ be Banach spaces and $T : X \rightarrow Y$ a bounded operator that is weakly compact.

How to prove that there exists $\epsilon > 0$ and a sequence $(x_n)_n \subset X$ such that $\|x_n\| = 1, \|T(x_n)\| \geq \epsilon$ forall $n \in \mathbb N$, and the sequence $(T(x_n))_n$ is congerging weakly to $0$ ?

All what I can say is that I can extract a subsequence from $(Tx_n)_n$ which converges to a $y \in Y$.

Thanks for any help

$\endgroup$
  • $\begingroup$ I'm not sure you always can. In $Y=\ell_1$, a sequence is weakly null if and only if it is norm null. $\endgroup$ – David Mitra Aug 20 '18 at 13:09
  • $\begingroup$ Thank you David. This assersion was used in a paper wrote by William B. Johnson : eudml.org/doc/266092 (just under the first theorem in the first page). I ask my question here because I don't understand the reasons why this can be true. $\endgroup$ – Kébir J Aug 20 '18 at 13:50
  • $\begingroup$ There, it's assumed that $T$ is not compact. Can you deduce your result with that in mind? $\endgroup$ – David Mitra Aug 20 '18 at 13:59
  • $\begingroup$ I suppose I can extract a subsequence for which $\|(T(x_n))_n$ is decreasing. Then I extract a sequence weakly-convergent. $\endgroup$ – Kébir J Aug 20 '18 at 14:23
1
$\begingroup$

As in the paper you referenced in the comments, we'll assume $T$ is not compact.

Then, we may (and do) choose a sequence $(x_n)_n$ in $S_X$ such that no subsequence of $(Tx_n)_n$ is norm convergent. Now choose a subsequence $(x_{n_k})_k$ of $(x_n)_n$ so that $(T x_{n_k})_k$ is weakly convergent.

Now, $(T x_{n_k})_k$ is not norm-Cauchy. So we find $\epsilon>0$ and subsequences $(p_k)$ and $(q_k)$ of $(n_k)$ with

$$\Vert T( x_{p_k}- x_{q_k})\Vert>\epsilon,\quad k=1,2,\ldots.\tag{1}$$
Set $y_k= x_{p_k}- x_{q_k}$.

Then $(Ty_k)$ is weakly null (since $(T(x_{n_k}))_k$ is weakly convergent).

Since $T$ is bounded and since (1) holds, it follows that $(\Vert y_k\Vert)_k$ is bounded away from $0$. From this, it follows that $\bigl(T( y_k/\Vert y_k\Vert)\bigr)_k$ is weakly null. Appealing to (1) again, $\bigl\Vert T(y_k/\Vert y_k \Vert)\bigr\Vert>\epsilon/2 $ for each $k$.

So, $(y_k/\Vert y_k\Vert)$ is the sought after sequence.

$\endgroup$
  • $\begingroup$ Thank you. I was very very far from this. $\endgroup$ – Kébir J Aug 20 '18 at 16:48

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.