3
$\begingroup$

Let $\mu$ and $\nu$ be finite measures on a measure space $(X,\mathcal{A})$. Show that there is a nonnegative measurable function $f$ on $X$ such that for all $E \in \mathcal{A}$, $$\int_E(1-f)d\mu=\int_Efd\nu$$

I have no clue how to show the above:

I started using below but I stuck

$$\mu(E)=\int_Efd\nu + \int_Efd\mu = \int_Efd(\nu+\mu)$$

defining $\psi=\nu+\mu$ , I have to show that $\mu\ll\psi$?

$\endgroup$
2
  • $\begingroup$ Yes, that will do. Then you can apply the Radon-Nikodym theorem. $\endgroup$
    – drhab
    Oct 30, 2019 at 16:26
  • $\begingroup$ yes, for positive measure, (as Will suggested below) was trivially correct. But I was thinking the measures are signed measure. $\endgroup$
    – domath
    Oct 30, 2019 at 16:35

1 Answer 1

1
$\begingroup$

You want to show $\mu = f \cdot (\mu + \nu)$ for some $f.$

If the result were true for positive $\mu$ and $\nu,$ then $\mu^+ = f \cdot (\mu^+ + \nu^+)$ and $\mu^- = g \cdot (\mu^- + \nu^-),$ then $\mu = \mu^+ - \mu^- = f \cdot \mu^+ - g \cdot \mu^- + f \cdot \nu^+ - g \cdot \nu^-.$ Since $\mu^+ \perp \mu^-,$ it turn outs $fg = 0$ (one of them is always zero). Thus, $h = f - g$ satisfies $f \cdot \mu^+ = h \cdot \mu^+$ and similarly for $\mu^-, \nu^\pm.$ Hence, $\mu = h \cdot (\mu + \nu).$ Thus, suffices to show the result for positive measures.

The result is trivial for positive measures for you can do what you did already and the relation $\mu(N) \leq \psi(N)$ shows that $\mu$ is absolutely continuous relative to $\psi.$ Q.E.D

$\endgroup$
4
  • $\begingroup$ I just realised that I did not showed that $h \geq 0.$ In fact, that restriction may be hard to prove for signed measures (I believe it should be false, take $\nu$ a negative measure and $\mu$ positive). $\endgroup$
    – William M.
    Oct 30, 2019 at 16:38
  • $\begingroup$ is this $f$ in $\mu^+ = f \cdot (\mu^+ + \nu^+)$ the same as $f$ in $\mu = f \cdot (\mu + \nu)$? I saw as you did this: $f \cdot (\mu + \nu)=f\cdot. [(\mu^+-\mu^-)+((\nu^+-\nu^-))] = f\cdot. [(\mu^++\nu^+)-f\cdot.((\mu^-+\nu^-))]$ $\endgroup$
    – domath
    Oct 30, 2019 at 16:40
  • $\begingroup$ I do not understand what you are asking. Certainly my $f$ and my $h$ need not be equal. $\endgroup$
    – William M.
    Oct 30, 2019 at 16:49
  • 1
    $\begingroup$ My $f$ and your $f$ are also not equal, but you need to understand $f$ us just a symbol that represents a function, I used $f,$ $g$ and $h$ in place of $f$ depending on what measure was under consideration. $\endgroup$
    – William M.
    Oct 30, 2019 at 16:49

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.