2
$\begingroup$

Let $n$ be a positive integer.

Let $S \subseteq \mathbb{R}^n$. Is the Hausdorff dimension of the boundary of $S$ always smaller than the Hausdorff dimension of $S$?

I have not found anything concerning those questions in some looked up books, I was not able to prove one of the statements, and I failed finding a counterexample. Does anybody know something about that?

$\endgroup$
2
  • 1
    $\begingroup$ If $S$ is closed and nowhere dense, then $S$ is its own boundary, so it can't be strictly less in such a case. $\endgroup$ Aug 9, 2010 at 19:21
  • $\begingroup$ Maybe in your original question you meant "frontier" rather than boundary (the frontier is cl(S)-S). The answer is still negative, by the way (e.g: topologist's sine curve). $\endgroup$ Aug 9, 2010 at 22:12

3 Answers 3

6
$\begingroup$

"Smaller" in the sense of $\le$ ... If $S$ is closed and has Hausdorff dimension $< n$, then $S$ has empty interior, so (as noted by Joel) $S$ is its own boundary, and thus we have equality for the two dimensions. And of course if (perhaps not closed) set $S$ has dimension $n$, then the boundary could have any dimension from $0$ to $n$, inclusive. If $S$ is closed and has dimension $n$, then the boundary is either empty or has dimension $\ge n-1$.

$\endgroup$
10
$\begingroup$

The set of rational numbers has Hausdorff dimension 0, while its boundary is the set of real numbers, with Hausdorff dimension 1.

$\endgroup$
7
$\begingroup$

For closed sets, yes, since Hausdorff dimension comes from Hausdorff measure, which is an outer measure.

For non-closed sets, no. Take the rationals in the reals. They have Hausdorf dimension 0. But their boundary is the reals, which has dimension 1.

$\endgroup$
5
  • $\begingroup$ I don't understand your "outer measure" comment. Not all outer measures have this property? $\endgroup$ Aug 9, 2010 at 19:48
  • $\begingroup$ @Gerald, I think he just referring to the fact that if $H^d$ denotes d-dimensional Hausdorff outer measure and $A\subseteq B\subseteq \mathbb{R}^n$, then $H^d(B)=0$ implies $H^d(A)=0$, so $\inf\{d\geq0:H^d(A)=0\}\leq\inf\{d\geq0:H^d(B)=0\}$, along with the fact that $\partial B\subseteq B$ if $B$ is closed. $\endgroup$ Aug 9, 2010 at 19:56
  • $\begingroup$ @Gerald, I'm not understanding your concern. Jonas supplied more details. What property are you referring to? $\endgroup$ Aug 9, 2010 at 21:45
  • 1
    $\begingroup$ For the record, this answer seems to have been written independently of mine (even though technically it was posted a couple of minutes later) and it contains my answer as a proper subset. $\endgroup$ Aug 9, 2010 at 21:56
  • $\begingroup$ @Ryan: Jonas explained it. $\endgroup$ Aug 11, 2010 at 12:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.