Since we only need to check that $B^*$ is complete, we should prove Cauchy sequence $\lbrace l_n \rbrace$ converges.

In general idea, $$|(l-l_n)(f)| \leq |(l-l_m)(f)| + |(l_m-l_n)(f)| \leq |(l-l_m)(f)| + \epsilon/2||f||$$ then let m be large enough. We have $$|(l-l_n)(f)| \leq \epsilon ||f||$$ And we finish the proof.

(This proof is in Stein's book Functional Analysis)

However, can I prove it in this way? I mean prove Cauchy sequence $\lbrace l_n \rbrace$ converges.

Can I use the definition of norm that $$||l-l_n||=\underset{||f||=1}{sup}|(l-l_n)(f)|$$ and due to $|(l-l_n)(f)|$ converges to $0$ when $n$ goes to infinity, we have $l_n$ converges to $l$.

It seems to be wrong, but I don't know why.

  • $\begingroup$ The implication you're looking for is true but a bit strange to write down. If $X$ and $Y$ are normed vector spaces then the space of bounded, linear operators from $X$ to $Y$, $L(X,Y)$ is complete (and hence a Banach space) if and only if $Y$ is complete. $X^* = L(B, \mathbb{R})$ and so is a Banach space since $\mathbb{R}$ is complete. This doesn't require $X$ to be complete. $\endgroup$ – Rhys Steele Jan 14 '19 at 22:37
  • $\begingroup$ Thank you for your reply! It's correct? But why does the book not use this simpler idea? I think the idea is probably wrong. $\endgroup$ – Busy Zhu Jan 14 '19 at 23:01
  • $\begingroup$ If I'm honest I find your attempt confusing. Given a Cauchy sequence $\{l_n\}_{n \geq 0}$ in $B^*$ you don't a priori have a good candidate for its limit $l$, so the $l$ you write down isn't defined here. Further, writing a bound like $|(l-l_n)(f)| \leq |(l-l_m)(f)| + \frac{\varepsilon}{2} \|f\|$ isn't particularly helpful because you'll still be left to bound $|(l-l_m)(f)|$ in terms of $\|f\|$. For this proof to work you'd need to be able to pick $m$ independent of $f$ and if you could do that you'd just bound $|(l-l_n)(f)|$ for large enough $n$ straight away rather than doing this first. $\endgroup$ – Rhys Steele Jan 14 '19 at 23:05
  • $\begingroup$ The usual technique is to check that for $f \in B$, $(l_n(f))$ is a Cauchy sequence in $\mathbb{R}$ and so converges to some number $L(f) \in \mathbb{R}$. You can then define a map $l: B \to \mathbb{R}$ by $l(f) = L(f)$. It's then not to hard to check that $l$ is linear by properties of limits. Then you check that $l$ in fact must be a bounded linear map and finally you show that $l_n \to l$ in $B^*$ by sending $m \to \infty$ in $|l_n(f) - l_m(f)| \leq \varepsilon \|f\|$ for suitably large $m,n$. $\endgroup$ – Rhys Steele Jan 14 '19 at 23:10

Hint :

The space $B^*$ is the dual space of $B$, which is the space that contains all linear functionals such that $f : B \to \mathbb R$.

But, note that $\mathbb R$ is a complete space with respect to its metric (norm), the absolute value. That means that $\mathbb R$ is Banach.

  • $\begingroup$ Thank you for your reply. But it seems to not match with my question. $\endgroup$ – Busy Zhu Jan 14 '19 at 22:34
  • $\begingroup$ @BusyZhu You want to prove that $B^*$ is a Banach space. All you need to take into account is that $\mathbb R$ is complete, thus Banach. That means that if $x \in B^*$ and is Cauchy, then it converges and the proof is really straightforward. $\endgroup$ – Rebellos Jan 14 '19 at 22:38
  • $\begingroup$ thanks again! But actually what I want to know is the way I prove it is correct or not. The idea of using definition is straightforward but seems to be weird. $\endgroup$ – Busy Zhu Jan 14 '19 at 22:59

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.