I am struggling to solve an exercise in my measure theory book and any help for solving it would be appreciated:

Let $(\Omega,\mathcal{A},\mu)$ be a measure space and let $f:\Omega \to \mathbb{R}$ be measurable. Find a function $g:\Omega \to \mathbb{R}$ which is equal to $f$ almost everywhere but is not measurable.

  • $\begingroup$ Please don't feel that I am cheating or so. It is a self-study. Thanks. $\endgroup$ – Cupitor Mar 28 '13 at 15:07
  • 1
    $\begingroup$ Are you familiar with the concept of complete measures? This doesn't sound true for arbitrary measures (under counting measure, "almost everywhere" is the same as "everywhere"). $\endgroup$ – Erick Wong Mar 28 '13 at 15:14
  • $\begingroup$ I just checked it on wiki but I don't see the relation? Well, thats why counting measure cannot generate a counterexample here! $\endgroup$ – Cupitor Mar 28 '13 at 15:16
  • $\begingroup$ If the measure space is complete then any function almost everywhere equal to a measurable function is also measurable. $\endgroup$ – Kavi Rama Murthy Oct 27 '17 at 9:21

Two cases:

  • There is a set $N$ contained in a measurable set of zero measure which is not measurable.

    In this case, take $g:=f+\mathbf 1_{N}$, where $\mathbf 1_N$ is the indicator function of $N$. It's not a measurable function, because otherwise so would be $\mathbf 1_{N}$.

  • Each set contained in a measurable set of zero measure is an element of $\mathcal A$ (the measure space is called complete).

    In this case, there is no such function $g$. Indeed, assume that $f$ is a measurable function and that $g\colon\Omega \to\mathbb R$ is such that $f(x)=g(x)$ for almost every $x$. We shall show that $g$ is $\mathcal A$-measurable.

    The set $D=\left\{x\in\Omega\mid f(x)\neq g(x) \right\}$ is contained in a set of measure zero hence is measurable. Since $$g(x)=f(x)\mathbf 1_{\Omega\setminus D}(x) +g(x)\mathbf 1_D (x),$$ it suffices show that the function $h\colon x\mapsto g(x)\mathbf 1_D (x)$ is $\mathcal A$ measurable. To this aim, let $A_t :=\left\{x\in\Omega\mid h(x)\lt t \right\}$. If $t\leqslant 0$, then $A_t\subset D$ hence $A_t$ is contained in a set of measure zero and is $\mathcal A$-measurable. If $t\gt 0$ then $A_t=\left(\Omega\setminus D\right) \cup\left(D\cap \left\{x\in\Omega\mid g(x)\lt t \right\}\right)$, which is the union of two measurable sets (the second one because it is contained in $D$ hence contained in a set of zero measure).

  • $\begingroup$ your second case is still a little bit vague for me. So you are saying that if we are in the second case $g$ will be measurable as well? $\endgroup$ – Cupitor Mar 28 '13 at 15:52
  • 1
    $\begingroup$ I've edited in order to clarify. $\endgroup$ – Davide Giraudo Mar 28 '13 at 16:01
  • 1
    $\begingroup$ Since $f \chi_{f = g} = g \chi_{f = g}$, it is evident that $g \chi_{f = g}$ is $\mathcal A$ measurable. But why does this imply that $g \chi_{f \neq g}$ is a pointwise limit of simple functions? $\endgroup$ – el_tenedor Oct 27 '17 at 5:05
  • $\begingroup$ @el_tenedor You are right, it deserved a clarification. There is no need to use simple functions actually. $\endgroup$ – Davide Giraudo Oct 27 '17 at 8:36
  • $\begingroup$ In the first case, how can you say that $g$ and $f$ coincide almost everywhere if the set on which they disagree can't be measured? $\endgroup$ – mattecapu Jan 11 '18 at 11:53

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.