3
$\begingroup$

Motivation: As a consequence of this question, every function $u$ in $W^{1,2}(\mathbb{R}^n,\mathbb{R})$ has a representative $u^*$ such that there is a null set $E\subset \mathbb{R}^n$ with the property that $u^*:\mathbb{R}^n\setminus E\to \mathbb{R}$ has path-connected graph. This would be obvious if for all Sobolev functions $u$ there were a null set $E$ such that $\mathbb{R}^n\setminus E$ is path-connected and $u$ restricted to $\mathbb{R}^n\setminus E$ is continuous. But I have never heard of this latter fact being true, even without the requirement that $\mathbb{R}^n\setminus E$ is path-connected, so I suppose it is false in general. Here I would like to see an explicit counterexample.

Question: I am looking for an explicit function $u\in W^{1,2}(\mathbb{R}^n,\mathbb{R})$ for which there is no null-set $E$ such that $u:\mathbb{R}^n\setminus E\to \mathbb{R}$ is continuous.

$\endgroup$
0

1 Answer 1

3
$\begingroup$

Here is an explicit example for $n \geq 5$.

Fix an enumeration $\{x_i\}_{i \in \mathbb N}$ of some dense set in $\mathbb R^n$. Let $\phi$ be a smooth, nonnegative cutoff function that is $1$ on an open neighbourhood of the origin and supported on the closed unit ball. Set $\psi(x) = \frac{\phi(x)}{|x|}$ for nonzero $x$. For $x = 0$ we define it arbitrarily to be $0$.

Now define $u(x) := \sum_{i=0}^\infty \frac{1}{2^i} \psi(x - x_i)$.

Since $|\psi(x)|^2 \sim 1/|x|^2$ and $|\nabla \psi(x)|^2 \sim 1/|x|^4$, each term and hence each partial sum is in $W^{1,2}(\mathbb R^n)$ as soon as $n \geq 5$. The infinite sum can be seen to converge absolutely in $W^{1, 2}$ norm, so $u$ too is in $W^{1,2}(\mathbb R^n)$.

The key property of $u$ is that its essential supremum is $+\infty$ on any open set, so no matter whether you remove a null set, $u$ remains discontinuous everywhere.

$\endgroup$
2
  • $\begingroup$ Thank you for the nice counterexample in the case $n\geq 5$. The counterexample for $n=2$ doesn't seem to work though, it needs $n\geq 3$ to get $L^2$ integrability of the derivative, doesn't it? $\endgroup$ Commented Oct 17 at 18:55
  • $\begingroup$ @No-one Hm that might be true. I'll remove it for now. $\endgroup$ Commented Oct 18 at 6:56

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.