6
$\begingroup$

Let $\mu\lll\nu$ be $\sigma$-finite Borel measures, which are not finite, on a topological space $X$. Under what conditions is $0<\operatorname{ess-supp}(\frac{d\mu}{d\nu}I_K)<\infty$ for every compact subset $\emptyset\subset K\subseteq X$.

In other words when is $\frac{d\mu}{d\nu} \in L^{\infty}_{\nu,\mathrm{loc}}(X)$?

$\endgroup$
  • $\begingroup$ My google ninja's skill got me the following: math.stackexchange.com/questions/3119941/… is this helpful? $\endgroup$ – Alan Apr 30 at 18:26
  • $\begingroup$ @Alan It is pretty interesting, but I don't want to assume any such geometric structure. Btw, I like the ninja skills terminology I'll have to borrow it :) $\endgroup$ – AnnieLeKatsu Apr 30 at 18:29
  • 1
    $\begingroup$ I got this terminology from someone else in physicsforums, can't remember from whom though. :-D $\endgroup$ – Alan Apr 30 at 18:33
  • $\begingroup$ What type of conditions do you need? $\endgroup$ – Fedor Petrov Apr 30 at 19:04
  • 1
    $\begingroup$ in fact $\| \frac{d\mu}{d\nu}\|_{\infty,K}=\sup_{H\subset K} \frac{\mu(H)}{\nu(H)}$ $\endgroup$ – Pietro Majer Apr 30 at 19:33
4
$\begingroup$

Let $$f:=\frac{d\mu}{d\nu}.$$ Then $$f\in L^{\infty}_{\nu,loc}(X)\iff\text{$\forall$ compact $K\subseteq X$ $\exists$ $c_K\in(0,\infty)$ $\forall$ Borel $A$ we have $\mu(A\cap K)\le c_K\nu(A\cap K)$.}$$

Indeed, for the $\Rightarrow$ implication, take any compact $K\subseteq X$. Then $\exists$ $c_K\in(0,\infty)$ such that $f\le c_K$ $\nu$-a.e. on $K$. So, for any Borel $A$ we have $$\mu(A\cap K)=\int_{A\cap K}f\,d\nu\le c_K\nu(A\cap K),$$ as desired.

Vice versa, for the $\Leftarrow$ implication, take any compact $K\subseteq X$ and suppose that $\mu(A\cap K)\le c_K\nu(A\cap K)$ for some $c_K\in(0,\infty)$ and all Borel $A$. Let now $A:=f^{-1}((c_K,\infty))$, so that $f>c_K$ on $A$. Then $$\mu(A\cap K)=\int_{A\cap K}f\,d\nu\ge c_K\nu(A\cap K),$$ and the latter inequality is strict (and hence contradicts condition $\mu(A\cap K)\le c_K\nu(A\cap K)$) if $\nu(A\cap K)>0$. So, $\nu(A\cap K)=0$, that is, $f\le c_K$ $\nu$-a.e. on $K$, as desired.


Similarly, for any compact $K\subseteq X$, $$\operatorname{esssup}_Kf>0\iff \text{ $\exists$ $b_K\in(0,\infty)$ $\exists$ Borel $A$ such that $\mu(A\cap K)\ge b_K\nu(A\cap K)$.}$$

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ As a general question of interest, is this relation used/ does it have a name or is it a bit arbitrary? $\endgroup$ – AnnieLeKatsu Apr 30 at 19:40
  • 1
    $\begingroup$ @AnnieTheKatsu : I don't know of a name for this relation. It should be folklore, rather transparent from a standard proof of the Radon--Nikodym theorem; cf. the comment by Pietro Majer. $\endgroup$ – Iosif Pinelis Apr 30 at 19:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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