Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am stuck on what I think may be the very last line of the proof I am seeking.

Let $(X, \mathcal{B})$ be a measurable space which has associated with it the finite measures $\mu$ and $\nu$ s.t. $\nu \ll \mu$. I aim to show that $\forall \epsilon > 0$, $\exists \delta > 0$ s.t. $\forall A \in \mathcal{B}$,

$$\mu(A) < \delta \implies \nu(A) < \epsilon$$

  1. Fix $\epsilon > 0$.

  2. For all $n \in \mathbb{N}$, let $\delta_n = \frac{1}{n^2}$.

  3. For all $n \in \mathbb{N}$, let $A_n \in \mathcal{B}$ s.t. if $\exists E \in \mathcal{B}$ s.t. $\mu(E) < \delta_n$ and $\nu(E) > \epsilon$, then set $A_n = E$. Otherwise, set $A_n = \emptyset \in \mathcal{B}$.

  4. Suppose for sake of contradiction that $|\{A_n\}| = \infty$, so that no matter the $\delta > 0$, we could find a $\delta_n = \frac{1}{n^2} < \delta$ which has associated with it a measurable $A_n \ne \emptyset$ with $\mu(A_n) < \delta_n < \delta$ and $\nu(A_n) > \epsilon$.

  5. Now if we let $\underset{n \rightarrow \infty}{\text{limsup}}$ $A_n = S$, we have (from a prior problem) that $\mu(S) = 0$ since $\mu$ is a finite measure and $\sum_{n=1}^\infty \mu(A_n) \le \sum_{n=1}^\infty \delta_n = \sum_{n=1}^\infty \frac{1}{n^2} < \infty$. Since $\nu \ll \mu$ we therefore have $\nu(S) = 0$ as well.

  6. Yet $\nu(S) = \nu(\bigcap_{n=1}^\infty \bigcup_{n=m}^\infty A_m) \ge \epsilon > 0$ since...

and it's here where I'm stuck in the proof.

share|cite|improve this question
What do you mean by "finite measurable space"? I only know a definition for the space $(X, \mathcal B, \mu)$ to be finite, but not for $(X,\mathcal B)$ alone. – Patrick Da Silva Dec 18 '12 at 23:52
Should have read "measurable space" and not "finite measurable space". Fixed above. – user1770201 Dec 18 '12 at 23:54
Yes, that is more comprehensible now. Thank you – Patrick Da Silva Dec 18 '12 at 23:55
What is you definition of $\nu \ll \mu$? – leo Dec 19 '12 at 20:41
up vote 3 down vote accepted

You are almost there. Notice that since $\nu(S) = 0$, $$ 0 = \nu(S) = \nu \left( \bigcap_{n=1}^{\infty} \bigcup_{m=n}^{\infty} A_m \right) = \lim_{n \to \infty} \nu \left( \bigcup_{m=n}^{\infty} A_m \right). $$ Therefore, there exists $n$ such that $$ \nu(A_n) \le \nu \left( \bigcup_{m=n}^{\infty} A_m \right) < \varepsilon, $$ a contradiction of the choice of $A_n$.

I must say though that the right way to formulate this proof would be this : Suppose by sake of contradiction that the result is false, i.e. that there exists an $\varepsilon > 0$ such that for all $\delta > 0$, there is $C_{\delta} \in \mathcal B$ with $$ \mu(C_{\delta}) < \delta, \quad \nu(C_{\delta}) \ge \varepsilon. $$ Then you can choose $\delta_n = \frac 1{n^2}$ and $A_n = C_{\delta_n}$ and then continue by following our steps.

Hope that helps,

share|cite|improve this answer
Thanks! -- By this do you mean that the proof -- as currently formulated -- is merely awkward, or that it is actually wrong in its current form (even given your help above with the last step)? – user1770201 Dec 19 '12 at 3:00
@user1770201 : I am just saying that you look like you're not comfortable with the $\varepsilon-\delta$ proofs because you feel like putting a lot of unnecessary comments. Usually, when we do a proof by contradiction with those kind of statements, it's better to suppose the whole thing is wrong instead of starting the contradiction hypothesis in the middle. – Patrick Da Silva Dec 19 '12 at 3:37

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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