So I am a bit stuck on where to begin with this one...

Show that $C^1[0,1]$ with the norm defined as $\|f\|=\|f\|_\infty + \|f^\prime\|_\infty$ is a Banach space.

I started with an arbitrary Cauchy sequence, $f_n \to f$, so we know that for every $\epsilon > 0, \exists M \in \mathbb{N}$ such that $\|f_n - f\| \leq \epsilon , \forall n \geq M.$

Where I am getting stuck, is how do we then show that $f \in C^1[0,1]$, and that therefore $C^1[0,1]$ with the norm defined, is complete (and therefore a Banach space).

So here is my attempted solution, based on the comments below:

Assume $f_n$ is a Cauchy sequence in $C^1[0,1]$, then for every $\epsilon > 0, \exists M \in \mathbb{N}$ such that

$$\|f_n - f_m\|=\|f_n - f_m\|_\infty + \|f_n^\prime - f_m^\prime \|_\infty \leq \epsilon, \forall m,n\geq M.$$

So for any fixed $x \in C^1[0,1], \|f_n - f_m\|_\infty \leq \epsilon$ and $\|f_n^\prime - f_m^\prime \|_\infty \leq \epsilon$ and thus for this fixed $x$ $f_n$ and $f_n^\prime$ are Cauchy sequences, and therefore converge: $f_n \to f$ and $f_n^\prime \to f^\prime.$

So let $m \to \infty$ then $\|f_n - f\|\leq \epsilon, \forall n\geq M, \forall x\in [0,1]$. Then:

\begin{align}\|f_n - f\|&= \|f_n - f\|_\infty + \|f_n^\prime - f^\prime \|_\infty\\ &=\sup_{x\in [0,1]}|f_n(x)-f(x)| + \sup_{x\in [0,1]}|f_n^\prime (x)-f^\prime (x)|\\&\leq \epsilon\quad \forall n\geq M. \end{align}

Therefore $\lim_{n\to \infty} \|f_n - f\| = 0 \implies$ $C^1[0,1]$ is complete.

How is this looking?

  • $\begingroup$ Where'd you get $f$ from? You can use the fact that $C[0,1]$ is complete to show $f\in C^1[0,1]$ (and if you haven't proven that $C[0,1]$ is a Banach space already you should start there). $\endgroup$ Sep 5 '14 at 4:29
  • $\begingroup$ You want to assume that you have a sequence of functions $f_1, f_2,..$. Such that as $n,m$ get large $$||f_n-f_m|| = ||f_n-f_m||_{\infty}+||f'_n-f'_m||_{\infty} \to 0$$ So the point here is that individual norms go to zero, too (and uniformly so). $\endgroup$
    – David P
    Sep 5 '14 at 4:33
  • $\begingroup$ Well, you start with a Cauchy $f_n$ and the standard route gives $f_n \to f$ and $f'_n \to g$ for some $f,g$. You need to show that $g = f'$. $\endgroup$
    – copper.hat
    Sep 5 '14 at 4:38
  • $\begingroup$ I have updated the original post with my new attempt at the proof... how is it looking? $\endgroup$ Sep 5 '14 at 5:28
  • $\begingroup$ Anyone able to take a look at the update, and let me know how it is looking? $\endgroup$ Sep 5 '14 at 6:47

Define $S:C[0,1] \times C[0,1] \to C[0,1]$ by $S(f,g)(t) = f(0)+\int_0^t g(\tau) d \tau$. It is straightforward to verify that $S$ is continuous in the product norm on $C[0,1] \times C[0,1]$.

If $f_n$ is $ \|\cdot \|$ Cauchy, then $f_n $ and $f'_n$ are $\|\cdot\|_\infty$ Cauchy and so we have $f_n \to f$ and $f'_n \to g$ (in the $\|\cdot\|_\infty$ norm).

We have $S(f_n,f'_n) = f_n$ for all $n$. Since $S$ is continuous, we have $S(f,g) = f$, from which it follows that $f$ is differentiable and $f'=g$.

  • $\begingroup$ Wouldn't you need to show that $f$ is differentiable first, so that $Df$ is well-defined? $\endgroup$ Sep 5 '14 at 11:13
  • $\begingroup$ @VincentBoelens: Since $D$ is continuous and $f_n \to f$ (in $\|\cdot\|$), we have $Df_n \to Df$. $\endgroup$
    – copper.hat
    Sep 5 '14 at 13:33
  • $\begingroup$ @VincentBoelens: You are right, I think my argument is circular. $\endgroup$
    – copper.hat
    Sep 5 '14 at 14:47
  • $\begingroup$ @VincentBoelens: I fixed my rather egregious mistake. Thanks for catching that. $\endgroup$
    – copper.hat
    Sep 5 '14 at 18:24
  • 1
    $\begingroup$ @VincentBoelens: Sad part is that I have answered the same question somewhere else in the side after correcting someone for making the error I just made :-). $\endgroup$
    – copper.hat
    Sep 5 '14 at 19:10

Since $f_n,f'_n\in C([0,1])$ and both sequences converge in the sup-norm, there exist $f,g\in C([0,1])$ such that $$\|f_n-f\|_\infty \rightarrow 0 \text{ and } \|f'_n-g\|\rightarrow0.$$ It needs to be shown that $f$ is differentiable in $f'=g$. This means showing that $$\lim_{h\to 0}\frac{f(x+h)-f(x)-g(x)h}{h}=0 \ \forall x\in [0,1].$$ Fix $x_0\in [0,1]$. Let $\epsilon>0$ be given. Since $g$ is continuous, there exists $\delta>0$ such that $|x_0-x|<\delta$ implies $|g(x_0)-g(x)|<\frac\epsilon4$. Let $h$ be arbitrary but fixed, such that $|h|<\delta$. Choose $n$ such that $\|f_n-f\|_\infty<\frac{|h|\epsilon}{4}$ and $\|f'_n-g\|<\frac\epsilon4$. By the mean value theorem, there exists $\xi$ such that $|x_0-\xi|<|h|$ and $\frac{f_n(x_0+h)-f_n(x_0)}{h}=f'_n(\xi).$ It follows that \begin{align} \left|\frac{f(x_0+h)-f(x_0)-g(x_0)h}{h}\right|&\le \left| \frac{f(x_0+h)-f_n(x_0+h)}{h}\right|+\\ &+\left|\frac{f_n(x_0+h)-f_n(x_0)}{h}-g(x_0)\right|+\left|\frac{f_n(x_0)-f(x_0)}{h}\right|\le\\ &\le\frac{\epsilon}{2}+\left|\frac{f_n(x_0+h)- f_n(x_0)}{h}-g(x_0)\right|\le\\ &\le\frac\epsilon2+\left|\frac{f_n(x_0+h)-f_n(x_0)}{h}-g(\xi)\right|+\left|g(\xi)-g(x_0)\right|=\\ &=\frac\epsilon2+|f'_n(\xi)-g(\xi)|+|g(\xi)-g(x_0)|<\epsilon. \end{align} Since $h,x_0$ and $\epsilon$ were arbitrary, the conclusion follows.


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.