Formally speaking, the Lebesgue-measure $\lambda$ on $\mathbb{R}$ is the restriction of the outer Lebesgue-measure $$ \lambda^*(A)=\inf\left\{\sum_{n=1}^{\infty}p(C_n): A\subset\bigcup_{n=1}^{\infty}C_n, \forall n\in\mathbb{N}: C_n\in\mathcal{S}\right\} $$ where $p$ is the Lebesgue pre-measure on the semiring $\mathcal{S}$ of all intervalls of the form $(a,b]$ defined by $p((a,b])=b-a$ and $p(\emptyset)=0$ on the $\sigma$-algebra $\mathcal{M}(\lambda^*)$ of all sets that are $\lambda^*$-measurable. One says that $\lambda$ is the unique continuation of $p$ and especially $\mathcal{B}(\mathbb{R})\subset\mathcal{M}(\lambda^*)$, i.e. all Borel sets are Lebesgue-measurable. So far so good.

Now I have one (maybe very stupid) problem.

If one has f.e. the intervall $(a,b)$ (which is in the Borel-$\sigma$-algebra and therefore formally it is Lebesgue-measurable) it is $$\lambda((a,b))=\lambda^*_{|\mathcal{B}(\mathbb{R})}((a,b))=b-a. $$

  1. Why?

  2. Is it right that calculating the Lebesgue-measure of a Borel-set is calculating the outer Lebesgue-measure in fact? So that writing $\lambda$ is in fact not very precisely? Instead one should write $\lambda^*_{|\mathcal{B}(\mathbb{R})}$.

  3. In which way is $\lambda=\lambda^*_{|\mathcal{B}(\mathbb{R})}$ a continuation of $p$, because $\lambda=p$ on $\mathcal{S}$?

  1. Because that is the definition of restriction: if we have a function $f\colon A\to B$ and a $C\subseteq A$, then $f|_C$ is the function $C\to B$ such that for all $c\in C$ we have $f|_C(c)=f(c)$. Unless you're asking why $\lambda^*((a,b))=b-a$ -- that follows from continuity of $p$ on $\emptyset$, for example.
  2. Why would that be imprecise? If you define $f(x)=x^5+3x^2+6x$, would you regard writing $f(1)=10$ (instead of $1^5+3\cdot 1^2+6\cdot 1=10$) as imprecise?
  3. I'm not sure what is meant by continuation, but I think it's most likely meant in a sense synonymous to "extension". In that case, yes, the point is, $p=\lambda|_{\mathcal S}$.
  • $\begingroup$ 1. Yes, my question is, because $\lambda^*((a,b))=b-a$, but now I think it is clear: $(a,b)=\bigcap_{n\geqslant 1}(a,b+\frac{1}{n}]$ and $\lambda^*(a,b+1]=b+1-a<\infty$ so it is $\lambda^*((a,b)=\lim_{n\to\infty}b+\frac{1}{n}-a=b-a$. 2. I do not see what your example has in common with that to be honest. I think it would be more clear not to write $\lambda$ but to write it as the restriction of an outer measure to make clear how the measure is constructed. 3. Yes, I mean extension. $\endgroup$ – mathfemi Sep 16 '14 at 12:24
  • 1
    $\begingroup$ @mathfemi: It is not really important how the measure is constructed: its important properties are that: it is a (regular, countably additive, positive) measure, it attains on intervals the values we want it to attain, and it is translation invariant. The rest is only really a construction to show that it exists. $\endgroup$ – tomasz Sep 16 '14 at 12:31

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.