This is problem 4T in Bartle's The elements of integration and Lebesgue measure.

Suppose $f_n$ are non-negative measurable function such that $(f_n)$ converges to $f$, and that $$\int fd\mu = \lim \int f_n d\mu<\infty.$$ Prove that for all measurable set $E$, we have $$\int_E fd\mu = \lim \int_E f_n d\mu.$$

I found a solution here.

The solution used the reverse Fatou's lemma, which needs the hypothesis of dominated boundedness, that is there exists a measurable function $g$ such that $f_n\le g$ for all $n$ and $\displaystyle \int g <\infty$. I cannot see how can we get this function $g$.

Thank you very much.

  • $\begingroup$ such function exists by convergence. $\endgroup$ – Henricus V. Feb 14 '16 at 17:04
  • $\begingroup$ Hint: It's enough to show that $\lim_n\int|f-f_n|\,d\mu=0$. To that end write $|f-f_n|$ as $2(f-f_n)^+-f+f_n$, and observe that $(f-f_n)^+\le f$ which is integrable. $\endgroup$ – John Dawkins Feb 14 '16 at 17:10
  • $\begingroup$ @JohnDawkins If we find $g$ integrable such as $|f-f_n|\le g$ for any $n$ large enaugh, we can indeed conclude using the dominated convergence theorem. I understand that $(f-f_n)^+\le f$. But we need to dominate $|f-f_n|=(f-f_n)^++(f-f_n)^-$, and I couldn't do it. Written in the form you gave, we get $|f-f_n|\le 2f-f+f_n =f+f_n$. I also tried to dominate $(f-f_n)^-$ and finding that is it smaller than $f_n$, but we get to the same point. Would you please help me? $\endgroup$ – Scientifica Nov 14 '16 at 12:46
  • $\begingroup$ @HenryW. Would you please give me a hint on how to find such function? I explained what I tried in the comment above. $\endgroup$ – Scientifica Nov 15 '16 at 14:38

The proof you mention seems wrong to me, since it uses the inequality $$ \limsup\int_Ef_n\le\int_E\limsup f_n $$ You may complete the proof applying Fatou's lemma to $E^c=\mathbb{R}\setminus E$. $$ \begin{multline} \int_Ef=\int_{\mathbb R}f-\int_{E^c}f\ge\int_{\mathbb R}f-\liminf\int_{E^c}f_n\\=\int_{\mathbb R}f-\liminf\Bigl(\int_{\mathbb R}f_n-\int_{E}f_n\Bigr)=\limsup\int_Ef. \end{multline}$$

  • $\begingroup$ I mean the inequality $\lim \sup \int f_n \le \int \lim \sup f_n$. $\endgroup$ – Tien Kha Pham Feb 14 '16 at 17:24
  • $\begingroup$ Yes, I realized that. I have edited the answer. $\endgroup$ – Julián Aguirre Feb 14 '16 at 17:32
  • $\begingroup$ This an old(ish) answer, but would you mind clearing up: a) why the use of the inequality you mention is wrong; and b) whether you meant to write $f_n$ in place of $f$ at the very end of the answer? Thank you! $\endgroup$ – mlaci Oct 10 '17 at 3:52
  • $\begingroup$ The inequality is incorrect; take $f_n$ the characteristic function of $[n,n+1]$. And yes, I meant $f_n$. $\endgroup$ – Julián Aguirre Oct 12 '17 at 13:38
  • $\begingroup$ You're saying that, in that case, $lim\ sup\ 1 \leq \int_{E} lim\ sup\ f_n = \int_{E} 0$, correct? But wouldn't this example function break the requirement in the hypothesis that $\int fd\mu = \lim \int f_n d\mu<\infty$, since $\int 0 d\mu \neq lim\ 1$? Thanks again! $\endgroup$ – mlaci Oct 15 '17 at 20:31

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.