6
$\begingroup$

Let $K \subset \mathbb{R}^{2}$ be a compact connected set such that $\mathcal{H}^{1}(\partial K)<+\infty$. Is $K$ arcwise connected?

$\endgroup$
2
  • $\begingroup$ By results of Gruber the generic copy of the pseudoarc in the plane has Hausdorff dimension 1 and infinite Hausdorff measure, but maybe it is possible to build one with finite Hausdorff measure? Just a wild guess $\endgroup$ Commented Mar 16, 2024 at 17:25
  • $\begingroup$ If you know that $\partial K$ is a continuous image of an interval then you are done. But how do you know this? $\endgroup$ Commented Mar 16, 2024 at 18:28

1 Answer 1

2
$\begingroup$

The answer is positive. First suppose that $K=\partial K$. I will show that in this case $K$ is the image of a curve. By definition of Hausdorff measure, for every $\epsilon>0$ we can cover $K$ with open disks of radii $\leq\epsilon$ so that the sum of the radii is at most $A$, where $A$ is independent of $\epsilon$. Since $K$ is compact, we can choose a finite subcovering with the same properties. The union of the disks of this subcovering is connected (otherwise $K$ will be disconnected). So by discarding if necessary some redundant disks we arrange that the union of the circles is connected.

Now consider this union of the circles as a graph (the vertices are intersection points of circles, the edges are the arcs into which these intersection points break the union of the circles). Notice that the degree of each vertex is even, since any number of circles with a common point will give even number of half-edges meeting at this point. Therefore, by Euler's theorem there is a closed path $\gamma$ passing through every edge exactly once. This path has length $\leq 2\pi A$.

Setting $\epsilon=1/n$, and parameterizing $\gamma$ by the the arclength, we obtain continuous functions $\gamma_n:[0,2\pi A]\to R^2$. This family is equicontinuous, since all functions satisfy the Lipschitz condition with constant $1$. Therefore by Arzela'a theorem there is a subsequence which converges uniformly to a curve $\gamma_\infty$. Since every point of $K$ is at a distance $\leq 1/n$ from $\gamma_n$, the image of $\gamma_\infty$ equals $K$, so we parameterized $K$ as a curve.

Now let us prove the general case. We know that every component of $\partial K$ is a rectifiable curve, and the sum of the lengths of these curves is finite. Let $x$ and $y$ be two points of $K$. Connect them with a segment $[x,y]$. If this segment is in $K$, we are done. If not, the intersection $[x,y]\cap cK$ consists of open intervals. We order these intervals into a sequence $(I_n)$ so that the lengths are non-increasing. Let $(a_1,b_1)$ be the longest of these intervals It belongs to a component $D$ of $cK$, and $\partial D$ is a closed connected set which belongs to $\partial K$. Therefore it belongs to some component of $\partial K$ which is a rectifiable curve. Replace this interval $(a_1,b_1)$ by a piece $\gamma_1$ of this curve. It is possible that $\gamma_1$ serves also some further intervals of our sequence $(I_n)$. Let $(a_2,b_2)$ be the first interval in the sequence $I_n$ whose endpoints are not on $\gamma_1$. Then repeat the procedure. In this way we obtain a sequence of curves $\gamma_k$ and the remainder of the original interval $[x,y]$ which together make a curve from $x$ to $y$, (even a rectifiable one).

$\endgroup$
3
  • $\begingroup$ I think you misread something. Obviously, we cannot cover a $K$ with non-empty interior by small disks with $\sum r_n \le A$. I think you are discussing the special case $K=\partial K$, but that was already done by the OP in a comment. $\endgroup$ Commented Mar 17, 2024 at 18:16
  • $\begingroup$ @Christian Remling: thanks for your remark. I amended my answer. $\endgroup$ Commented Mar 17, 2024 at 23:53
  • $\begingroup$ @Christian Remling: I do not see any problem if $[x,y]\cap K$ is Cantor. In this case, the curve that I constructed consists of this Cantor set itself plus countably many arcs of curves which are pars of $\partial K$, one for each gap in the Cantor set. $\endgroup$ Commented Mar 18, 2024 at 3:37

You must log in to answer this question.