Let a sequence of nonnegative function $f_n$ on a measure space $(X,\beta, \mu)$ converge pointwise to $f$. Prove that for any nonnegative simple function $\rho (x)\leq f(x)$ we have $$\int\rho (x) d\mu \leq \liminf \int f_n (x)d\mu$$ assuming that the left hand side is finite.

I found this problem on a previous qualifying exam and I need help.

What I Think.
Before I proceed, let me ask this. Because $f_n$ is defined on a measure space $(X,\beta, \mu)$, can we say $f_n$ is measurable $\forall n$?

Presuming the above is true

If $f_n$ is measurable then the pointwise limit $f$ is also measurable. Since $f_n \ge 0$ and measurable, it is integrable and we invoke Fatou's lemma $$\int \liminf f_n d\mu \leq \liminf\int f_n d\mu \Rightarrow \int f d\mu \leq \liminf \int f_n d\mu \quad ........(1)$$

Let $\rho \ge 0$ be a simple function and $\rho (x)\leq f(x)$ , f is integrable [also nother presumption] and by definition $$\int f d\mu =sup\big\{\int \rho d \mu , \rho (x)\leq f(x)\big\}$$

Then from $(1)$ above, we have $$sup\big\{\int \rho d \mu , \rho (x)\leq f(x)\big\} \leq \liminf\int f_n d\mu$$ and hence $$\int \rho d \mu \le sup\int \rho d \mu \leq \liminf\int f_n d\mu$$

I believe I may have taken a lot of things for granted here. can someone correct me or write a proof it. thanks


This is a bit of a strange question, because if you're allowed to use Fatou's lemma then there's really no reason to consider only simple functions $\rho\leq f$. If you could find another way to prove the stated inequality, then it could be used to prove Fatou's lemma by taking the supremum over all simple functions $\rho\leq f$.

Anyway, to address your first concern, I think the problem implicitly assumes that the functions $f_n$ are measurable, for otherwise there's no way to make sense of $\int f_n\;d\mu$. If we can use Fatou's lemma, then we know that $$ \int f\;d\mu\leq\liminf_{n\to\infty}\int f_n\;d\mu$$ and then if $g$ is any measurable function (not necessarily a simple function) with $0\leq g\leq f$ it follows that $$ \int g\;d\mu\leq \int f\;d\mu\leq\liminf_{n\to\infty}\int f_n\;d\mu$$ by the monotonicity of the integral. Note that these inequalities hold even if some (or all) of these expressions are infinite.

  • $\begingroup$ Okay, so if I understand you well, the question assumes that $f_n$ is measurable that is why it goes aheard to write the integral for it. But my question is , whenever I see the statement '$f_n$ on a measure space $(X, \beta, \mu $)' , does it mean that $f_n$ are all measurable w.r.t $\mu$? $\endgroup$ – J. Kyei Dec 29 '16 at 23:35
  • $\begingroup$ There's no reason for an arbitrary non-negative function to be measurable, but if a problem considers the integral of a certain function, I think it's safe to assume that the function is measurable. It would of course be preferable for the question to carefully state the assumptions. $\endgroup$ – carmichael561 Dec 29 '16 at 23:41

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.