1
$\begingroup$

This is very simple and I'm probably missing something obvious but I can't seem to spot the mistake.

We know that $C^1$ functions on compact sets are Lipschitz.

Let $U \subset \mathbb R^n$ be open, $S \subset U$ be compact, and $f: U \to \mathbb R^m$ be $C^1$.

$f$ is Lipschitz,so $\exists M \in \mathbb R$ so that $|f(x)-f(y)| \leq M|x-y|$ for all $x,y \in S$.

But this also means that it's true for all $x,y$ in the interior of $S$. So $f$ is also Lipschitz with constant $M$ in the interior of a compact set. Compacts are closed and bounded, so only criterion for being the interior of a compact set is being bounded.

The conclusion is that $f$ is Lipschitz on bounded sets, as we can close it to get a compact set, there it will be Lipschitz with some constant, so it also must be Lipschitz with that same constant inside.

Is this correct? I don't see anything in the literature or online about it. Everything is just about compactness. Where is the mistake here?

Very important edit

I meant that $f$ is $C^1$ on an open set $U$ which contains compact set $S$, then $f$ is Lipschitz on interior of $S$.

$\endgroup$
  • $\begingroup$ Sounds true to me. Why are you so sure it's wrong $\endgroup$ – Dalamar Apr 18 '18 at 7:45
  • $\begingroup$ Lack of references and google hits, "too good to be true" feel to it. $\endgroup$ – Oria Gruber Apr 18 '18 at 7:48
  • $\begingroup$ The Lipschitz constant $M$ depends in general on the compact set $S \subset U$. $\endgroup$ – user539887 Apr 18 '18 at 7:49
  • $\begingroup$ After your edit, yes, it is true. $\endgroup$ – user539887 Apr 18 '18 at 8:04
0
$\begingroup$

Note that $\sqrt{x}$ is $C^1$ in $(0,1]$ (which is not compact) but this function is not Lipschitz in any bounded set $(0,a]$ with $0<a\leq 1$: if $(x_n)_n$ and $(y_n)_n$ are two sequences which go to $0^+$ with $x_n>y_n$ then $$\lim_{n\to +\infty} \frac{\sqrt{x_n}-\sqrt{y_n}}{x_n-y_n}= \lim_{n\to +\infty} \frac{1}{\sqrt{x_n}+\sqrt{y_n}}=+\infty.$$

On the other hand, if $f$ is $C^1$ in an open set $U$ which contains a compact set $S$, then $|f'|$ is continuous in the compact set $S$ and in $S$ it attains a maximum value $M$. Hence $f$ is Lipschitz in $S$ (and therefore also in its interior) with constant $M$.

$\endgroup$
  • $\begingroup$ Yes but notice that I required that $f$ to be $C^1$ on $U$ which contains the compact set $S$. In your case, $\sqrt{x}$ is not $C^1$ on an open set that contains $[0,1]$ so it fails this property. $\endgroup$ – Oria Gruber Apr 18 '18 at 7:47
  • $\begingroup$ Do you mind to rephrase your question clearly please. $\endgroup$ – Robert Z Apr 18 '18 at 7:52
  • $\begingroup$ I'm terribly sorry for the confusion. What I meant is: $f$ is $C^1$ in open set $U$ which contains compact set $S$, then $f$ is lipschitz in the interior of $S$. Infact, $f$ is lipschitz in any bounded set $A \subset U$. So for instance if $f$ is $C^1$ in $\mathbb R^n$, then it will be Lipschitz in every bounded subset of $\mathbb R^n$, even those which are not closed. $\endgroup$ – Oria Gruber Apr 18 '18 at 7:53
  • $\begingroup$ Please see my P.S. $\endgroup$ – Robert Z Apr 18 '18 at 7:56
  • $\begingroup$ @OriaGruber Even the two statements in that comment of yours are not equivalent, since if $A \subseteq U$ is bounded, it doesn't mean that there is a compact set $S$ such that $A \subseteq S \subseteq U.$ $\endgroup$ – Adayah Apr 18 '18 at 7:58
0
$\begingroup$

Let $U=(0,1)$ and consider $f(x)=x^{-1}$ on $U$? We can choose by Mean Value Theorem that $\cdots<x_{2}<\eta_{2}<y_{2}<x_{1}<\eta_{1}<y_{1}<1$ such that $\left|-\dfrac{1}{\eta_{i}^{2}}\right|\leq M$ for $i=1,2,...$, and $\eta_{i}\rightarrow 0$ as $i\rightarrow\infty$.

$\endgroup$
  • $\begingroup$ So the inequality blows up. $\endgroup$ – user284331 Apr 18 '18 at 7:51
  • $\begingroup$ I think I understand, thank you. $\endgroup$ – Oria Gruber Apr 18 '18 at 7:52

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.