2
$\begingroup$

Let $\mathcal{V}$ be an outer measure on $X$, $(A_\alpha)_{\alpha\in I}$ be a chain of increasing subsets of $X$.

  1. Is it true that $\mathcal{V}(\bigcup_{\alpha\in I}A_\alpha)=\sup_{\alpha\in I}\mathcal{V}(A_\alpha)$?
  2. If this is not true in general, are there classes of spaces $X$ and outer measures $\mathcal{V}$ (for example Hausdorff measures on metric spaces) for which this holds?
$\endgroup$

2 Answers 2

3
$\begingroup$

One can see a counterexample easily for the reals $\mathbb{R}$ if the Continuum Hypothesis holds, for in this case the reals $\mathbb{R}$ are the union of a chain of countable sets. Simply well-order the reals in order type $\omega_1$ and for countable ordinals $\alpha$ let $X_\alpha$ be the first $\alpha$ many points in this enumeration. So every $X_\alpha$ is countable, but the union is all of $\mathbb{R}$.

More generally, avoiding the CH assumption, let $\kappa$ be the cardinality of the smallest non-measure $0$ set $X$ of reals. We can enumerate $X$ in order type $\kappa$, and the initial segments of this enumeration all have measure $0$, but the union is $X$, which is non-measure $0$.

The cardinal $\kappa$ used above is known as the uniformity number for Lebesgue measure in the theory of cardinal characteristics; for example, see this MO answer or Andreas Blass' handbook article.

$\endgroup$
2
  • $\begingroup$ A variant of your idea works without $CH$. Well-order the reals and let $X_{\alpha}$ be the first $\alpha$ points in the well-ordering. Then there is a least $\lambda$ such that $X_{\lambda}$ doesn't have measure 0. Clearly $\lambda$ is a limit ordinal and $X_{\lambda} = \bigcup_{\alpha < \lambda}X_{\alpha}$ is a counterexample. $\endgroup$ Sep 28, 2010 at 12:15
  • $\begingroup$ Simon, you and I hit on that idea simultaneously! I had added it just before you submitted your comment. $\endgroup$ Sep 28, 2010 at 12:30
3
$\begingroup$

This is true if $X$ is a locally compact Hausdorff space, $\mathcal{V}$ is a Radon measure on $X$ and the sets $A_\alpha$ are open. See Folland's "Real Analysis" where he proves a slightly more general result.

$\endgroup$
1
  • $\begingroup$ Thanks a lot! If anyone else is interested, this can be found in Theorem 7.12. $\endgroup$ Sep 30, 2010 at 7:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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