Theorem: Let $V$ be a Banach Space and $W$ be a normed linear space. Let $I$ be an arbitrary indexing set and for each $i\in I$, let $T_i\in L(V,W)$. Then either,

(1)there exists $M>0$ such that $\|T_i\|\le M$ for all $i\in I$ or

(2) $\sup_{i\in I}$ $\|T_i(x)\|=\infty$, for all $x$ belonging to some dense $G_{\delta}$ set in $V$.

This is how the proof goes by,

For each $x$, define $\phi(x)=\sup_{i\in I}\|T_i(x)\|$

Define $V_n=\{x\in V| \;\phi(x)>n\}$.

$V_n$ is open for each $n$.

Assume now that there exists $N$ such that $V_N$ fails to be dense in $V$.

Then there exists $x_0\in V$ and $r>0$ such that $x+x_0\notin V_N$ if $\|x\|<r.$

This implies that $\phi(x+x_0)\le N$ for all such $x$ and so, for all $i\in I$ $\|T_i(x+x_0)\|\le N$.

Thus if $\|x\|\le r/2$, we have , for all $i\in I$, $\|T_i(x)\|\le \|T_i(x+x_0)\|+\|T_i(x_0)\|\le 2N$

It follows from this that, for all $i\in I$, $\|T_i\|\le \frac{4N}{r}$

I do not understand how the above highlighted argument holds. Can someone please explain it to me.

  • 1
    $\begingroup$ When you wrote the inequality "$\|T_i(x)\|\le \|T_i(x+x_0)\|+\|T_i(x_0)\|\le N$", did you actually mean to write $2N$ instead of $N$ on the RHS. (That would explain why it is $\frac{2N}{r/2}$ and not $\frac{N}{r/2}$.) $\endgroup$ – Martin Sleziak Oct 22 '17 at 5:31
  • $\begingroup$ Oops! It was a typo, thanks for pointing it out @MartinSleziak $\endgroup$ – Naive Oct 22 '17 at 5:33
  • $\begingroup$ "$\sup_{i\in I}$ $\|T_i\|=\infty$, for all $x$" that sentence does not make sense because there is no $x$ in this sentence. $\endgroup$ – miracle173 Oct 22 '17 at 11:37

Note that \begin{align*} \|T_i\| = \sup_{\|x\|\le 1}\|T_i(x)\| = \sup_{\|x\|\le 1}\frac{2}{r}\left\|T_i\left(\frac{rx}{2}\right)\right\| \le \frac{2}{r}2N = \frac{4N}{r}, \end{align*} where the inequality follows from the previous line and the fact that $\|rx/2\|\le r/2$.


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.