2
$\begingroup$

Let $\mu$ be a Borel probability measure on $\mathbb{R}^n$ such that there are $c,C,d,D>0$ satisfying: for every $x \in \mathbb{R}^n$ and every $r>0$ $$ c r^d \leq \mu(B(x,r)) \leq Cr^D. $$ Let's call such a measure almost Ahlors Regular.

Note: when $d=D$ then we say that the measure is Ahlfors regular and, if $c$ were allowed to be $=0$ then the measure would be called upper Ahlfors regular.

What are examples of Ahlors regular measures?

$\endgroup$
1
  • 1
    $\begingroup$ Usually one asks that the inequality holds for every $x$ in the support of the measure (and for every radius up to the diameter), and not for every point of the space. In this way the concept is meaningful, and indeed there exist Ahlfors regular measures of intermediate dimension (e.g. Hausdorff measure restricted to a subspace). Otherwise, as pointed out by the current answer, there are only absolutely continuous measures $\endgroup$
    – Del
    Commented Jun 28, 2022 at 21:17

1 Answer 1

1
$\begingroup$

$\newcommand\R{\mathbb R}\newcommand\la{\lambda}$A measure $\mu$ is Ahlfors regular (according to your definition) iff it has a density (with respect to the Lebesgue measure $\la$) bounded away from $0$ and $\infty$ (if $B(x,r)$ denotes the ball centered at $x$ and of radius $r$).

Indeed, by inscribing and circumscribing, we can replace $B(x,r)$ in your definition by the (left-open) $n$-cube $b(x,r):=\prod_1^n(x_i,x_i+r]$ (say of size $r$) for $x=(x_1,\dots,x_n)\in\R^n$ (at the same time also appropriately changing the values of $c$ and $C$), so that $$ c r^d\le\mu(b(x,r))\le Cr^d \tag{1}\label{1} $$ for all $x\in\R^n$ and $r>0$.

Partitioning, in the natural manner, the unit $n$-cube $b(0,1)$ into $N^n$ $\ n$-cubes each of size $1/N$, from \eqref{1} we get $$ c N^{n-d}\le\mu(b(0,1))\le CN^{n-d} \tag{2}\label{2} $$ for all $N$, which is possible only when $d=n$ -- otherwise, $N^{n-d}$ will go either to $0$ or to $\infty$ as $N\to\infty$, which would contradict one of the two inequalities in \eqref{2}, since $\mu(b(0,1))\in(0,\infty)$.

Therefore and by partitioning and approximation, we get $$ c \prod_1^n r_i\le\mu\Big(\prod_1^n(x_i,x_i+r_i]\Big)\le C \prod_1^n r_i $$ for any real $x_1,\dots,x_n$ and any positive real $r_1,\dots,r_n$. That is, $c\la\le\mu\le C\la$ on the semiring of all left-open $n$-boxes.

We conclude that $c\la\le\mu\le C\la$ on the Lebesgue $\sigma$-algebra, as claimed.

$\endgroup$
4
  • $\begingroup$ Right, and my definition were on a Riemannian manifold, then the same (argument...but patched locally) would work to conclude that $\mu$ must have a density with respect to the Riemannian Volume measure; no? $\endgroup$
    – ABIM
    Commented Feb 24, 2022 at 15:19
  • $\begingroup$ If $d\ne D$, then the inequality $cr^d\le Cr^D$ cannot hold for all $r>0$. (My previous comment was incorrect.) $\endgroup$ Commented Feb 24, 2022 at 15:26
  • $\begingroup$ So if insead only required $\mu(b(x,r))\lesssim n^D$ (and not the lower-inequality) then this would only imply that $\mu$ has a density with respect to $\lambda$ but not that this density is bounded away from $0$ and only bounded away from $\infty$? $\endgroup$
    – ABIM
    Commented Feb 28, 2022 at 9:14
  • 1
    $\begingroup$ @A_K : Yes, except that I think you meant $r^D$ in place of $n^D$. $\endgroup$ Commented Feb 28, 2022 at 13:36

You must log in to answer this question.

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