Let $A_n$ be a decreasing sequence of Borel sets with finite but with measure $\geq \epsilon > 0$ for all $n$. Then there exists a compact set $B_n \subset A_n$ for all $n$ such that $\mu(A_n \setminus B_n) < \frac{\epsilon}{2^{n+1}}$ for some $\epsilon >0$. Then define $C_n := \cap_{k=1}^n B_k$. It is clear that $C_n \subset B_n \subset A_n$. Then how come we have $\mu(A_n \setminus C_n) < \frac{\epsilon}{2}$ and $\mu(C_n) > \frac{\epsilon}{2}$, please? I tried the following argument.

$$ \mu(A_n \setminus C_n) = \mu\left( A \setminus \cap_{k=1}^n B_k \right) = \mu \left( A_n \cap \left( \cap_{k=1}^n B_k \right)^C \right) = \mu \left( A_n \cap \left( \cup_{k=1}^n B_k^C \right) \right) = \mu \left( \cup_{k=1}^n \left( A_n \cap B_k^C \right) \right) = \mu \left( \cup_{k=1}^n \left( A_n \setminus B_k \right) \right) \leq \sum_{k=1}^n \mu(A_n \setminus B_k) $$

I cannot see how to proceed from the above to show the two results.

  • $\begingroup$ This is false unless you assume $\mu(A_n)<\infty$, there is no such compact set for, say $A_n=B(0,n)^c$. $\endgroup$ – Adam Hughes Oct 13 '14 at 11:20
  • $\begingroup$ @AdamHughes Thank you for pointing that out. I forgot to mention it. How to show the results after I assume what you said? $\endgroup$ – LaTeXFan Oct 13 '14 at 11:31
  • $\begingroup$ Also, the condition $\mu(C_n)>{\epsilon\over 2}$ is false. If all the $A_n$ have measure $0$, for example, then sub-additivity says all the $C_n$ have measure $0$. $\endgroup$ – Adam Hughes Oct 13 '14 at 11:36
  • $\begingroup$ @AdamHughes I guess I should add $A_n$ has non-zero measure. And the whole point is to show that $C_n$ is not empty. $\endgroup$ – LaTeXFan Oct 13 '14 at 11:41
  • $\begingroup$ @AdamHughes Thanks. What about the second result? BTW, it is allowed to comment on your answer. $\endgroup$ – LaTeXFan Oct 13 '14 at 11:47

We use that

$$A_n\setminus C_n=A_n\cap\left(\bigcup_{k=1}^n B_k^c\right)$$ $$=\bigcup_{k=1}^n (A_n\cap B_k^c)\;\subseteq \;\bigcup_{k=1}^n (A_k\cap B_k^c)$$ $$=\bigcup_{k=1}^n (A_k\setminus B_k)$$


$$\mu(A_n\setminus C_n)\le\sum_{k=1}^n\mu(A_k\setminus B_k)=\sum_{k=1}^n {\epsilon\over 2^{n+1}}<{\epsilon\over 2}.$$

Edit to reflect edited question

Now note that

$$\mu(B_1\cup (A_1\setminus B_1))=\epsilon\le \mu(A_1)= \mu(B_1)+\mu(A_1\setminus B_1)\le\mu(B_1)+{\epsilon\over 4}$$

so that

$${3\epsilon\over 4}\le\mu(B_1)$$

Now continuing on like this, we see that

$$\epsilon\le \mu(A_n)=\mu(C_n)+\mu(A_n\setminus C_n)< \mu(C_n)+{\epsilon\over 2}$$


$${\epsilon\over 2}<\mu(C_n).$$


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.