Suppose we have a measure $\mu$ on $(a,b]$ such that $\mu(a,b]=F(b)-F(a)$ where $F$ is non-decreasing, continuous function from the right,

Definition: A function $F$ is said to be absolutely continuous if for every $\epsilon$ there exist a $\delta$ such that for each finite collection $[a_i,b_i], i=1,...,k$, of nonoverlapping intervals, $\sum (b_i-a_i)<\delta$ implies $\sum |F(b_i)-F (a_i)|<\epsilon$

I'd like to prove that:

  • $F$ is absolutely continuous if and only if for each Borel set $A$, $\mu(A+x)$ is continuous in $x$.

where $A+x=\{a+x:a\in A\}$

Thank you

  • $\begingroup$ What have you tried? For the first question, you can start by writing down the definitions of absolute continuity. The answer will follow. For the second question, try assuming that $\mu(A)=0$ for all lebesgue measurable $A$. $\endgroup$ – Alex R. Jun 13 '16 at 20:36
  • $\begingroup$ Thanks @AlexR. There are lots of definitions for a measure to be absolutely continuous (with respect to Lebesgue measure), and unfortunately I don't know how to deal with continuity of $\mu(A+x)$ at $x$. $\endgroup$ – Stat95 Jun 13 '16 at 20:58
  • $\begingroup$ Presumably the notation $A+x$ implies translating $A$ by $x$? $\endgroup$ – Alex R. Jun 13 '16 at 21:01
  • $\begingroup$ Can you proof the assertion for all $A$ of the form $(a,b]$? If yes then you can approximate any $A$ as a union of these sets. $\endgroup$ – Alex R. Jun 13 '16 at 21:13
  • $\begingroup$ Actually, I'm not good at measure theory and I need an elementary proof, if it is possible. $\endgroup$ – Stat95 Jun 13 '16 at 21:32

One Direction:

I didn't see for a while exactly how @AlexR 's hint "Can you [prove] the assertion for all $A$ of the form $(a,b]$? If yes then you can approximate any $A$ as a union of these sets." led to a solution. So a much more detailed version, one of those things that's a proof for some readers and a hint for others:

For anyone who wants just an actual hint of a hint: The aha for me was realizing that we should do said approximation with respect to Lebesgue measure $m$, not $\mu$.

First, it's not true without some additional finiteness assumption somewhere. Let $A$ be the union of the $(n,n+1)$ for even $n$, let $B$ be the union of the $[n,n+1]$ for odd $n$, and define $\mu(E)=m(E\cap B)$. The corresponding $F$ is $Lip_1$, hence absolutely continuous. But $\mu(0+A)=0$ while $\mu(x+A)=\infty$ for $0<x<1$.

Exactly what the extra hypothesis really is nobody knows. I'm going to show it's true if $m(A)<\infty$; then below we see it's true for all $A$ if $\mu(\Bbb R)<\infty$.

Lemma 0. If $F$ is continuous then $\mu(x+(a,b])$ depends continuously on $x$. Also $\mu(\{a\})=0$.

Lemma 1. If $F$ is absolutely continuous then for every $\epsilon>0$ there exists $\delta>0$ such that $m(A)<\delta$ implies $\mu(A)<\epsilon$.

(Hint: If $m(A)<\delta$ then $A\subset\bigcup(a_j,b_j)$ with $\sum(b_j-a_j)<\delta$.)

Now suppose $F$ is absolutely continuous and $A$ is a Borel set. Assume that $m(A)<\infty$. For each $n$ choose $\delta_n$ as in Lemma 1, with $\epsilon=1/n$. Let $A_n$ be a finite union of intervals with $$m(A\Delta A_n)<\delta_n.$$

Now $\mu(x+A_n)$ depends continuously on $x$, and $$|\mu(x+A)-\mu(x+A_n)|\le\mu((x+A)\Delta(x+A_n))<1/n,$$so $\mu(x+A)$ depends continuously on $x$, by uniform convergence.

Below: Now assume that $\mu(\Bbb R)<\infty$. Define $$\mu_n(E)=\mu(E\cap[-n,n]).$$Then $\mu_n(x+A)\to\mu(x+A)$ uniformly. So we can assume that $\mu$ is supported in some compact set.

Suppose $A$ is a Borel set and $x_0\in\Bbb R$. There exists a bounded set $B$ such that $$\mu(x+A)=\mu(x+B)\quad(|x-x_0|<1).$$So the previous case shows that $\mu(x+A)$ is continuous at $x_0$.

The Other Direction

Now suppose that $\mu(x+A)$ depends continuously on $x$ for every Borel set $A$. Suppose that $m(A)=0$. Fubini's theorem shows that $$\int\mu(x+A)\,dx=0.$$Since the integral depends continuously on $x$ this shows that $\mu(A)=0$.

So $m(A)=0$ implies $\mu(A)=0$. It's not hard to deduce from this that for every $\epsilon>0$ there exists $\delta>0$ such that $m(A)<\delta$ implies $\mu(A)<\epsilon$, and it's not hard to show that that implies $F$ is absolutely continuous.

| cite | improve this answer | |
  • $\begingroup$ Thanks, but it needs extra lemmas (lemma 0) which is hard for me to understand. I've added a definition to my post, could you please prove it using this definition. $\endgroup$ – Stat95 Jun 14 '16 at 10:40
  • $\begingroup$ @EhsanTavassoli Not sure which lemma you're talking about, since you mention lemma 0, which has nothing to do with absolute continuity. In any case, I want to leave something for you to do. Lemma 0 is trivial, and Lemma 1 is very easy (note there's a hint for Lemma 1). $\endgroup$ – David C. Ullrich Jun 14 '16 at 13:18

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.