1
$\begingroup$

Let $\mu$ be a finite, compactly supported, non-zero measure on $\mathbb{R}^d$ for an integer $d$. Let $B(x,r)$ denote the ball of radius $r>0$ centered at $x \in \mathbb{R}^d$. For $\delta \in [0,d]$, we define two conditions: $\mu$ satisfies (C1)-$\delta$ if

$$ \mu(B(x,r)) \leq r^\delta $$

for all $x \in \mathbb{R}^d$ and $r >0$; and $\mu$ satisfies (C2)-$\delta$ if

$$ \mu(B(x,r)) \geq r^\delta$$

for all $r \in (0,1]$ and $x \in \text{supp}(\mu)$, the closed support of $\mu$.

Frostman's Lemma characterizes the sets which support measures satisfying (C1)-$\delta$.

Frostman's Lemma. A Borel set $A \subset \mathbb{R}^d$ supports a non-zero measure satisfying (C1)-$\delta$ if and only if $A$ has positive $\delta$-dimensional Hausdorff measure.

I am interested in a complimentary result concerning measures which satisfy (C2)-$\delta$. Observe that any non-empty set supports a point measure, which satisfies (C2)-$0$ and hence satisfies (C2)-$\delta$ for any $\delta > 0$. So any non-empty set trivially supports a measure satisfying (C2)-$\delta$. The question is rather when a given set is equal to the closed support of such a measure. In this case we only need consider closed sets. We will also restrict to $A$ being compact.

It can be shown using a covering argument that if $A$ is equal to the support of a (C2)-$\delta$ measure, then its Hausdorff dimension is at most $\delta$. This gives a necessary condition. The problem is then as follows.

Problem. Give a sufficient (or necessary and sufficient) condition (e.g. in terms of Hausdorff measure) on a compact set $A \subset \mathbb{R}^d$ which implies the existence of a finite measure $\mu$ satisfying (C2)-$\delta$ such that the closed support of $\mu$ equals $A$.

$\endgroup$
7
  • $\begingroup$ Technically, the condition should be something like: one cannot find disjoint balls centered at $K$ of radii less than $1$ with arbitrarily large sum of radii to the power $\delta$. It is, clearly, necessary, but I do not see what use any such condition would be because how is one possibly going to check that besides constructing a measure with (C2)-$\delta$ property? $\endgroup$
    – fedja
    Commented Jan 4, 2021 at 23:02
  • $\begingroup$ Yes, this seems equivalent (or very close to equivalent) to the existence of a measure satisfying (C2)-$\delta$. I am also interested in a potentially sub-optimal condition in terms of Hausdorff dimension. It is plausible that if $\delta < \text{dim}(A)$, then there exists a (C2)-$\delta$ measure whose closed support equals $A$. I am unaware of any counter-examples to this. However, I have never seen a statement like this written anywhere. $\endgroup$
    – Thomas
    Commented Jan 5, 2021 at 16:53
  • $\begingroup$ No, you cannot do it in terms of the Hausdorff dimension: that is controlled by just a sequence of very efficient covers, but in between you may have very inefficient ones that do not prevent you from getting a bad packing. $\endgroup$
    – fedja
    Commented Jan 5, 2021 at 18:03
  • $\begingroup$ My understanding is that box / Minkowski dimension does not permit this efficiency -- what about a criterion in terms of that dimension? $\endgroup$
    – Thomas
    Commented Jan 5, 2021 at 18:39
  • $\begingroup$ That one is trivial: just take a maximal $2^{-k}$ separated set and put the mass $2^{-k\delta}$ at each point. Then take all $k\in\mathbb Z_+$ and add the resulting measures up. If the box dimension is less than $\delta$, the norms will decay geometrically, so the series will converge. The "packing criterion" actually holds (the proof is pretty much the same as (one of those) for the Frostman lemma; if you are really interested, I can post it later), but it doesn't look so great... $\endgroup$
    – fedja
    Commented Jan 6, 2021 at 11:40

1 Answer 1

2
$\begingroup$

The equivalence to packing mentioned by Fedja is due to Tricot [1]. See also [2] Sec 5 and [3] sec. 3 for variations and extensions. The early proofs and most textbooks used dyadic cubes in Euclidean space. But Howroyd [4], [5] gave the argument in compact metric spaces and this is included in Mattila's textbook [6]. The packing measure version is in the thesis [7], see also [8].

[1] C. TRICOT. TWO definitions of fractional dimension. Math. Proc. Camb. Philos. Soc. 91 (1982), 57-74.

[2] S. J. TAYLOR and C. TRICOT. Packing measure, and its evaluation for a Brownian path. Trans. Amer. Math. Soc. 288 (1985), 679-699.

[3] Cutler, Colleen D. "Strong and weak duality principles for fractal dimension in Euclidean space." In Mathematical Proceedings of the Cambridge Philosophical Society, vol. 118, no. 3, pp. 393-410. Cambridge University Press, 1995.

[4] Howroyd, John D. "On dimension and on the existence of sets of finite positive Hausdorff measure." Proceedings of the London Mathematical Society 3, no. 3 (1995): 581-604.

[5] Howroyd, John David. "On the theory of Hausdorff measures in metric spaces." PhD diss., UCL (University College London), 1995.

[6] Mattila, Pertti. Geometry of sets and measures in Euclidean spaces: fractals and rectifiability. No. 44. Cambridge university press, 1999.

[7] Joyce, Helen Janeith. "Packing measures, packing dimensions, and the existence of sets of positive finite measure." PhD diss., UCL (University College London), 1995.

[8] Joyce, Helen, and David Preiss. "On the existence of subsets of finite positive packing measure." Mathematika 42, no. 1 (1995): 15-24.

$\endgroup$
3
  • $\begingroup$ Indeed. The full texts of the papers are behind the paywall, so let me ask: do they give proofs for an arbitrary compact metric space (as it should be) or they rely on the Euclidean structure/finite dimension (like it is, to my surprise, for the Frostman lemma in many analysis textbooks)? $\endgroup$
    – fedja
    Commented Jan 9, 2021 at 3:37
  • $\begingroup$ I addressed this in my answer now. $\endgroup$ Commented Jan 9, 2021 at 17:01
  • $\begingroup$ Thank you for providing these references. $\endgroup$
    – Thomas
    Commented Jan 14, 2021 at 16:01

You must log in to answer this question.

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