Let $Q(\Bbb R)^\uparrow$ be the set of every countable union of half-open intervals, $\lambda: Q(\Bbb R)^\uparrow \rightarrow [0, \infty]$ the extensional Lebesgue-Pre-measure and $\lambda^*: \mathscr P(\Bbb R) \rightarrow [0, \infty]$, $\lambda^*(X)$ $=$ $\inf\{\lambda(B) : X \subset B, B \in Q(\Bbb R)^\uparrow\}$ be the outer Lebesgue-measure. We define:

$v: \mathscr P(\Bbb R) \rightarrow [0, \infty]$,

$v(X) := \inf \{\lambda(A) : A \subset \Bbb R$ open, $X \subset A\}$.

Show that

$\lambda^*(X) = v(X)$.

My attempt:

Since $A$ is an open set, and since every open set $A \subset \Bbb R$ can be written as a countable union of disjoint, half-open intervals with rational endpoints, it follows that $A \in Q(\Bbb R)^\uparrow$.

First, we prove that

$\lambda^*(X) \le v(X)$:

We are allowed to assume that $\lambda^*(X) = \lambda(M)$ for some $M \in Q(\Bbb R)^\uparrow$. This means that there is no $B \in Q(\Bbb R)^\uparrow$ such that $\lambda(B) \lt \lambda(M)$ with $B \supset X$. Therefore, it follows that for every open set $A \in Q(\Bbb R)$, we have that $\lambda(M) \le \lambda(A)$, and therefore, by defintion, $\lambda^*(X) \le v(X)$.

$\lambda^*(X) \ge v(X)$ can be proven analogous, which proves the identity.

Now, since this seems way too easy for me, I guess I made a mistake at some point. Yet, I used the fact that $A$ is open only in order to conclude that it is a part of $Q(\Bbb R)^\uparrow$, but there should be more into it.


Asked other people, they couldn't find a mistake, and there doesn't seem to be such a stackexchange-user either. But the more I think about it, the worse does the proof look for me. I simply don't want to believe that the core of the proof is to use that $A \in Q(\Bbb R)^\uparrow$.


First thing to note : As you said, every open set (except maybe the empty set, but this is not important) is in $Q(\mathbf R)^\uparrow$ so

$$\{\lambda(A): A \text{ is open}, X\subset A\} \subset \{\lambda(B): B\in Q(\mathbf R)^\uparrow , X\subset B\}$$

And taking the inf on both sides gives you automatically $\lambda^*\geq v$, this is the easy part.

The error in your proof is right at the beginning, you cannot assume $\lambda^*(X)=\lambda(M)$ for some $M\in Q(\mathbf R)^\uparrow$ because this is not true. For example if $X=[a;b]$ then for every $M\in Q(\mathbf R)^\uparrow$ such that $X\subset M$ one have $\lambda(M)>\lambda(X)$ (prove it).

Now we want to show that $\lambda^* \leq v$, take a non-empty $X\subset \mathbf R$ and a set $B\in Q(\mathbf R)^\uparrow$ such that $X\subset B$. Now $B= \bigcup_{i\in \mathbf N} [a_i;b_i[$, take $\varepsilon>0$ and define the open set $A$ by

$$A=\bigcup_{i\in \mathbf N}]a_i-\frac{\varepsilon}{2^i}; b_i[.$$

$B$ is included in $A$ so obviously $X\subset A$. Moreover $\lambda(A)\leq \lambda (B)+\varepsilon \sum_{i=0}^\infty 2^{-i}=\lambda(B)+2\varepsilon$. This show that $\lambda^*(X)\leq v(X)+2\varepsilon$ (prove it). Since $\varepsilon$ was arbitrary we get that $\lambda^*(X)\leq v(X)$ for every $X$ and thus $\lambda^*\leq v$.

The proof i gave for $\lambda^*\leq v$ is classic in measure theory.

  • $\begingroup$ Thanks, glad someone finally pointed out the mistake! :-) $\endgroup$ – Julian Nov 20 '16 at 13:09

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.