Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am not sure I am using the standard definitions so I will open by defining what I need:

  1. Let $X$ be a set, $\nu:\, \mathscr{P}(X)\to[0,\infty]$ will be called an external measure if $\nu(\emptyset)=0$ and for any $\{A_{i}\}_{i=1}^{\infty}\subseteq\mathscr{P}(X)$ (not neccesarily disjoint) it holds that $\nu(\cup_{i=1}^{\infty}A_{i})\leq\sum_{i=1}^{\infty}\nu(A_{i})$

  2. Let $\nu$ be an external measure on a set $X$ then we say that a set $A$ is $\nu$ measurable if for any $E\subseteq X$: $\nu(E)=\nu(E\cap A)+\nu(E\cap A^{c})$

Let $\mu$ be a $\sigma$-additive measure on an algebra $A\subseteq X$, let $\mu^*$ be the outer measure on $X$ that comes from $\mu$.

It was proven in another exercise that the set of $\nu:=\mu^{*}$ measurable sets, $M$, is a $\sigma$ algebra and that $\nu$ is $\sigma$-additive on $M$.

Since $\nu$ is also a measure on the algebra $M$ there is an outer measure $\nu^*=(\mu^*)^{*}$.

The exercise wishes to prove that $(\mu^{*})^{*}=\mu^*$.

I don't really know how to even start here, what I first wanted to figure out is what is $M$. my intuition is that $A=M$ but I tried to prove it and couldn't start (I wrote the definitions but I couldn't see why if $B\in A$ it is $\nu$-measurable, not to mention the other direction which seems even harder).

Can someone please help me get started on this problem ? maybe a hint or an observation that might help me figure out what to do?

ADDED: by an outer measure that comes from a measure I mean $$\forall E\subseteq X:\,\nu(E):=Inf\{\sum_{i=1}^{\infty}\mu(A_{i})\,|\, A_{i}\in A,E\subseteq\cup A_{i}\}$$

share|cite|improve this question
One question. How do you define an outer measure on $X$ coming from a measure? – T. Eskin Dec 16 '12 at 20:12
@ThomasE. - $\forall E\subseteq X:\,\nu(E):=Inf\{\sum_{i=1}^{\infty}\nu(A_{i})\,|\, A_{i}\in A,E\subseteq\cup A_{i}\}$ – Belgi Dec 16 '12 at 20:16
@ThomasE. - hope this makes things clear now – Belgi Dec 16 '12 at 20:16
Thanks, it does. Another question though. Did you mean to show that $(\mu^{\*})^{\*}=\mu^{\*}$? Or that $(\mu^{\*})^{\*}|_{A}=\mu$? As currently stated, the two set-functions are defined on different domains. – T. Eskin Dec 16 '12 at 20:41
@ThomasE. - yes, thanks for the correction. I updated the question – Belgi Dec 16 '12 at 20:45
up vote 3 down vote accepted

Let $\scr A$ be the $\sigma-$algebra on which $\mu$ is defined, and let $\scr M$ be the set of measurable sets of $\mu^*$.

Claim: $ \forall E\subseteq X[\mu^{**}(E)\leq\mu^*(E)]$

Proof: Note that $\scr A\subseteq\scr M$. Thus: $\{\sum_{i=1}^{\infty}\mu(A_{i})\,|\, E\subseteq\cup A_{i}, A_{i}\in \scr{A}\}\subseteq$$\{\sum_{i=1}^{\infty}\mu(A_{i})\,|\, E\subseteq\cup A_{i}, A_{i}\in \scr{M}\}$,


$\inf\{\sum_{i=1}^{\infty}\mu(A_{i})\,|\, E\subseteq\cup A_{i}, A_{i}\in \scr{A}\}\geq$$\inf\{\sum_{i=1}^{\infty}\mu(A_{i})\,|\, E\subseteq\cup A_{i}, A_{i}\in \scr{M}\}$

Now we proceed with the rest of the proof. Let $\epsilon>0$. Let $(A_i)_{i\in Z^+}$ be a sequence of sets in $\scr{M}$, such that $E\subseteq\cup_{i\in Z^+}A_i$ and $\sum_{i=1}^{\infty}\mu^*(A_i)<\mu^{**}(E)+\epsilon$. For each $A_i$, we choose a sequence of sets $(A_{i,j})_{j\in Z^+}$ in $\scr A$ such that $ A_i\subseteq \cup_{j\in Z^+}A_{i,j}$ and $\sum_{j\in Z^+}\mu(A_{i,j})<\mu^*(A_i)+\frac{\epsilon}{2^{i}}$. It follows that: $$\sum_{i\in Z^+}\sum_{j\in Z^+}\mu(A_{i,j})\leq\sum_{i\in Z^+}[\mu^*(A_i)+\frac{\epsilon}{2^{i}}]=\sum_{i\in Z^+}\mu^*(A_i)+\epsilon\leq (\mu^{**}(E)+\epsilon)+\epsilon$$

Note that $(A_{i,j})_{i,j\in Z^+}$ is a countable sequence of elements of $\scr A$ whose union contains $E$. Therefore, $\mu^*(E)\leq\sum_{i\in Z^+}\sum_{j\in Z^+}\mu(A_{i,j})$. Hence: $$\mu^*(E)\leq\mu^{**}(E)+2\epsilon$$ Since $\epsilon$ is an arbitrary positive number, we get $\mu^*(E)\leq\mu^{**}(E)$. We combine this with the previous claim to get $\mu^*(E)=\mu^{**}(E)$.

share|cite|improve this answer
Thanks for the answer, I will review it later today – Belgi Dec 25 '12 at 16:38

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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