3
$\begingroup$

By Rademacher’s theorem, Lipschitz functions are differentiable almost everywhere. I am wondering how badly this pointwise differentiability fails for functions that “just barely” fail to be Lipschitz.

On $\mathbb R^n$, the space of bounded Lipschitz functions coincides with the Sobolev space $W^{1, \infty}$. Thus, we ask the following:

Question: Does there exist a bounded $f$ on $\mathbb R^n$ that is in $W^{1, p}$ for all $1 \leq p < \infty$ but is not differentiable at any point in $\mathbb R^n$?

$\endgroup$
3
  • $\begingroup$ Isn't there a kind of generalization of Rademacher saying that a function $f\in W^{1,p}(\mathbb R^n)$ is differentiable a.e. (and it's differential equals the weak differential)? See Thm 5 in Chapter 5.8 in Evans's Partial Differential Equations $\endgroup$
    – Teri
    Commented Nov 14 at 8:18
  • $\begingroup$ Sorry I forgot to specify $p>n$ (otherwise the claim isn't true). $\endgroup$
    – Teri
    Commented Nov 14 at 8:26
  • $\begingroup$ @Teri Oh I was aware that $f$ is Holder continuous if $p > n$, but not of this generalisation... $\endgroup$
    – Nate River
    Commented Nov 14 at 11:22

2 Answers 2

1
$\begingroup$

As pointed out, this is possible only for $p\le n$. Let $u$ be an unbounded function in $W_0^{1,n}(B_{1/2})$, for instance $$ u(x)=\log |\log |x||-\log \log 2 $$ and let ${q_i}_i$ be a countable dense set. Then the function $$ v(x)=\sum_i 2^{-i} u(x-q_i) $$ is in $W^{1,n}$ and unbounded in the neighborhood of every point, and thus everywhere not differentiable.

$\endgroup$
1
  • $\begingroup$ Thanks! I learnt a new Rademachers theorem today apparently. $\endgroup$
    – Nate River
    Commented Nov 15 at 11:04
-3
$\begingroup$

No, such a function does not exist. If a bounded function $f$ on $\mathbb{R}^n$ belongs to the Sobolev space $W^{1,p}(\mathbb{R}^n)$ for all $1 \leq p < \infty$, then its weak derivative $\nabla f$ is in $L^p(\mathbb{R}^n)$ for all $p$. This implies that $\nabla f$ is essentially bounded, meaning $\nabla f \in L^\infty(\mathbb{R}^n)$. Therefore, $f$ is Lipschitz continuous, and by Rademacher's theorem, it is differentiable almost everywhere. Consequently, there cannot be a bounded function in all $W^{1,p}$ spaces that is nowhere differentiable.

$\endgroup$
2
  • 7
    $\begingroup$ Sorry, how did you deduce the essential boundedness of $\nabla f$? $\endgroup$
    – Nate River
    Commented Nov 14 at 7:43
  • $\begingroup$ because $f \in W^{1,p}(\mathbb{R}^n))$ for all $1 \leq p < \infty$, its gradient $\nabla f$ must be essentially bounded, making $f$ Lipschitz continuous and differentiable almost everywhere. Thus, there cannot exist a bounded function in all $ W^{1,p}(\mathbb{R}^n))$ spaces that is nowhere differentiable. $\endgroup$ Commented Nov 14 at 12:42

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .