Consider the set $\mathcal{F}$ of continuous functions on $[0;1]$ with boundary values $$ f(0)=f(1)=0 \qquad \forall f \in \mathcal{F}. $$ Define the metric $d(f,g) = \lVert f-g \rVert_\infty = \sup_{0 \leq x \leq 1} |f(x)-g(x)|$ and the Lipschitz constant $$ L(f) = \sup_{0 \leq x_1 < x_2 \leq 1} \left|\frac{f(x_1)-f(x_2)}{x_1-x_2}\right|. $$ Define the subset of Lipschitz-continuous functions as $\mathcal{L}_M = \left\{ f \in \mathcal{F} \colon L(f) \leq M \right\}$.
I want to show that the metric space $(\mathcal{L}_1,d)$ is a compact subset of $(\mathcal{F},d)$. What are the steps to do this?
Here is my proof sketch:
- Prove that $(\mathcal{L}_1,d)$ is totally bounded (easy!)
- Prove that any sequence in $\mathcal{L}_1$ is equicontinuous.
- Infer from 1. and 2. by the Arzelà-Ascoli Theorem that $(\mathcal{L}_1,d)$ is relatively compact in the Banach space $(\mathcal{F},d)$
- Prove that any Cauchy sequence in $(\mathcal{L}_1,d)$ converges to an element from $\mathcal{L}_1$, i.e., show that $(\mathcal{L}_1,d)$ is complete.
- Since $(\mathcal{L}_1,d)$ is relatively compact and complete, it is therefore compact. QED.
Step 4. gives me headaches. That's because for any Cauchy sequence $(f_n)_n$, the sequence $(L(f_n))_n$ can be any sequence you like (in the unit interval). As a remedy, I could switch to a stronger metric $$ d(f,g) = \lVert f-g \rVert_\infty + \sup_{0 \leq x_1 < x_2 \leq 1} \left|\frac{f(x_1)-f(x_2)}{x_1-x_2}-\frac{g(x_1)-g(x_2)}{x_1-x_2}\right| $$ But then how do I eventually get back to my original metric?