I am looking for an example of an incomplete measure space with a measure that is not sigma-finite.

All the measures which are not sigma-finite which I have come across so far are the following:

  • counting measure on a set that is not countable (e.g. on the measurable space $(\mathbb{R},\mathcal{P}(\mathbb{R}))$
  • or the measure $\mu$ on the trivial sigma algebra $\Sigma = \{\emptyset,X \}$ with $\mu(X) = \infty$

were complete.

Suppose $(\Omega, \Sigma, \mu)$ would be such a measure space. Let $\eta$ denote the induced outer measure and $\Sigma_\eta$ the $\sigma$-algebra of the $\eta$-measurable sets. Furthermore let $(\Omega,\tilde{\Sigma},\tilde{\mu})$ denote the completion of $(\Omega,\Sigma,\mu)$. Will in this case the inclusion $\tilde{\Sigma} \subset \Sigma_\eta$ be strict?

  • 1
    $\begingroup$ Take an uncountable direct sum of $([0,1], {\rm Borel}([0,1]), \lambda)$ where $\lambda$ is the Lebesgue measure. $\endgroup$ – Tomek Kania Aug 3 '15 at 22:09
  • $\begingroup$ Another example would be the long line with its natural Borel measure structure. en.wikipedia.org/wiki/Long_line_%28topology%29 $\endgroup$ – user24142 Aug 3 '15 at 22:18

A very simple example: \begin{align*} X\equiv&\,\{a,b,c\},\\ \Sigma\equiv&\,\{\varnothing,\{a,b\},\{c\},\{a,b,c\}\}, \end{align*} with \begin{align*} \mu(\varnothing)\equiv&\,0,\\ \mu(\{a,b\})\equiv&\,0,\\ \mu(\{c\})\equiv&\,\infty,\\ \mu(\{a,b,c\})\equiv&\,\infty. \end{align*} Then, $\Sigma$ is a $\sigma$-algebra, and $\mu$ is a measure on it. However, $\mu$ is

  • not $\sigma$-finite: no measurable set containing $c$ has finite measure; and
  • not complete: $\{a\}\subset\{a,b\}$ and $\mu(\{a,b\})=0$, but $\{a\}\notin\Sigma$.

The induced outer measure $$\eta(A)\equiv\inf\left\{\sum_{n=1}^{\infty}\mu(S_n)\,\Bigg|\,S_1,S_2,\ldots\in\Sigma\text{ and }A\subseteq\bigcup_{n=1}^{\infty}S_n\right\}\quad\forall A\subseteq X$$ is as follows: \begin{align*} \begin{array}{rclcrclcrclcrcl} \eta(\varnothing)&=&0&&\eta(\{a\})&=&0&&\eta(\{a,b\})&=&0&&\eta(\{a,b,c\})&=&\infty\\ &&&&\eta(\{b\})&=&0&&\eta(\{a,c\})&=&\infty&&&&\\ &&&&\eta(\{c\})&=&\infty&&\eta(\{b,c\})&=&\infty \end{array} \end{align*} It can be shown that any subset of $X$ is $\eta$-measurable, so $\Sigma_{\eta}=2^X$. In fact, $2^X$ is also the completion $\tilde{\Sigma}$ of $\Sigma$ with respect to $\mu$. Therefore, $\tilde\Sigma=\Sigma_{\eta}$.

I don't know if, in general, one can have $\tilde{\Sigma}\subset\Sigma_{\eta}$ strictly.


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.