Let $X$ be a locally compact metric space which is $\sigma$-compact, and let $\mu$ is an unsigned Borel measure which is finite on every compact set. Show that $\mu$ is a Radon measure.

I know that every unsigned Borel measure on a compact metric space which is $\sigma$-compact is a Radon measure. From the assumption, we only need to verify outer regularity and inner regularity. Inner regularity is easier, since we can write $X$ as a countable union and compacts sets $K_n$, and a closed set in each $K_n$ is also a closed set in $X$. I have difficulty verifying outer regularity, open set in $K_n$ may not be open in $X$.

  • $\begingroup$ My answer here math.stackexchange.com/questions/1574021/… also answers your question. $\endgroup$ – PhoemueX Dec 19 '15 at 14:07
  • $\begingroup$ @PhoemueX I have seen it. I don't want to invoke the Reisz representation theorem, can we prove it if we have known the compact case? $\endgroup$ – Xiang Yu Dec 19 '15 at 14:12


The regularity in the compact case ( see this post) implies the regularity in the $\sigma$-compact case. Indeed:

Take an exhaustion with compacts $K_n \subset \overset{\circ} K_{n+1}$, $\cup_n K_n = X$. Take a Borel set $A$ in $X$. Then $A \cap K_n$ is Borel in $K_{n+1}$. Take $L_n\subset A \cap K_n \subset U_n$, with $U_n$ open in $K_{n+1}$, $L_n$ compact and $\mu(U_n \backslash L_n) < \epsilon/2^n$. Modify $U_n$ to $U'_n = U_n \cap \overset{\circ} K_{n+1}$, open in $X$. We still have $$L_n\subset A \cap K_n \subset U'_n$$ and $\mu(U'_n \backslash L_n) < \epsilon/2^n$. Now take $U = \cup_n U'_n$ open and $L = \cup_n L_n$ countable union of compacts. We have $L \subset A \subset U$ and $$\mu( U \backslash L ) \le \sum \mu(U'_n \backslash L_n) < \epsilon$$

  • $\begingroup$ Nice answer. The key point is that every compact set in a LCH space has an open neighborhood whose closure is compact. $\endgroup$ – Xiang Yu Dec 20 '15 at 1:54
  • $\begingroup$ @XiangYu: Thanks. Yes, we managed to approximate pieces of $A$ of the form $A \cap K$ by using what you noticed. and then put together the approximations, $\endgroup$ – Orest Bucicovschi Dec 20 '15 at 13:59
  • $\begingroup$ Is this true in the non-metric case? $\endgroup$ – Cronus Feb 9 '18 at 2:23

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.