Curve-free $1$-set has lower density $\leq \frac{3}{4}$

67 Views Asked by At

Def : In $\mathbb{R}^2$, $1$-set is a set that has 1-dimensional Hausdorff dimension and positive measure.

Def : Lower density and upper density of $F$ are $$\lim\ \inf_r\ H^1 (F\cap B(x,r))/2r ,\ \lim\ \sup_r\ H^1(F\cap B(x,r))/2r $$ where $H^1$ is a Hausdorff measure.

Def : $F$ is curve-free if $H^1 (F\cap c)=0$ where $c: [0,1]\rightarrow \mathbb{R}^2$ is an injective continuous map of finite length.

EXE : Let $F$ be a curve-free $1$-set. Then it has a lower density at $H^1$-almost all $x\in F$ which is $\leq 3/4$.

Proof : In general, upper density of $1$-set at almost all points is between $\frac{1}{2}$ and $1$.

How can we prove this ?

1

There are 1 best solutions below

0
On BEST ANSWER

This is Theorem 3.23 in The Geometry of Fractal Sets by Falconer (the result is originally due to Besicovitch). Based on the length of the proof, it doesn't look like a good exercise to me. The proof relies on the following Lemma, in which $R(a,b) = B(a, r)\cap B(b, r)$ is the intersection of disks of radius $r=|a-b|$ centered at $a$ and $b$.

Lemma 3.22 Let $E$ be a $1$-set in $\mathbb{R}^2$ and suppose that $\alpha>0$. Let $E_0$ be a compact subset of $E$ with $H^1(E_0)>0$ such that $$H^1(E\cap R(a,b))\ge \alpha |a-b|\quad \forall \ a,b\in E_0$$ Then $E$ is not curve-free.

The proof of this lemma, including preliminary results, is about 3 pages (pages 40-43 in the book).

Assuming the lemma, the proof goes as follows. By inclusion-exclusion, $$ H^1(E\cap R(a,b)) + H^1(E\cap (B(a,r)\cup B(b, r))) = H^1(E\cap B(a,r)) + H^1(E\cap B(b,r)) $$ where $r=|a-b|$. If the lower density is greater than $3/4$ on a set of positive 1-measure, then by picking a small $\alpha>0$ and restricting $a,b$ to a suitable subset of $E$ we can get $H^1(E\cap B(a,r)) > (\frac34+\alpha)2 r$ and $H^1(E\cap B(b,r)) > (\frac34+\alpha)2 r$. On the other hand, the upper density is $\le 1$ a.e., and this is true even for the more general version of upper density, defined by $$ D_c^*(E, x) = \lim_{r\to 0} \sup_{x\in U, \ \operatorname{diam}U\le r} \frac{H^1(E\cap U)}{\operatorname{diam} U} $$ (It's called "upper convex density" in the book, but $U$ doesn't have to be convex here.) Applying this with $U=B(a,r)\cup B(b, r)$ yields $$ H^1(E\cap (B(a,r)\cup B(b, r))) \le (1+\alpha) \operatorname{diam}U = (1+\alpha) 3r $$ Putting it all together, $$ H^1(E\cap R(a,b)) \ge 4 \left(\frac34+\alpha\right) r - 3 (1+\alpha) r = \alpha r $$ allowing us to apply Lemma 3.22.

Aside: "purely unrectifiable" is a term often used for such sets.