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

let $X$ be a finite measure space and $\{f_n\}$ be a sequence of nonnegative integrable functions, $f_n \rightarrow f\ a.e.$ on $ X$. We know that $\lim_{n \rightarrow \infty}\int_X f_n d\mu=\int_X fd\mu$ and on any measurable $E_i \subset X$

I should apply Egoroff's theorem to conclude that $\lim_{n \rightarrow \infty}\int_X |f_n-f|d\mu=0$.

My attempt:

I broke the set $X$ to two sets: $F_\sigma$ on which $f_n \rightarrow f$ uniformly based on Egoroff's theorem and $X\backslash F_\sigma$ which is a very small set, i.e. $\mu\{X\backslash F_\sigma\}=\sigma$ and $f_n \nrightarrow f$

I want to show that on each of these sets, the integral is less than $\frac{\epsilon}{2}\ \forall \epsilon$ to finish. How can I show this for the set $X \backslash F_\sigma$?

share|cite|improve this question
I am missing something here. Take $f_n = n\cdot1_{[0,\frac{1}{n}]}$. Then $f_n(x) \to f(x) = 0$ a.e., but $\int f_n = 1$, and $\int f = 0$. – copper.hat Oct 29 '12 at 6:06
@copper.hat The "We know" is an assumption, I presume. – Alex Becker Oct 29 '12 at 6:07
@AlexBecker: Thanks, I missed that interpretation completely... – copper.hat Oct 29 '12 at 6:08
+1 for the interesting problem. – Davide Giraudo Oct 29 '12 at 10:06

I assume $f$ integrable, otherwise it's not necessarily true, as copper.hat shows.

Fix an integer $k$, and $E_k$ measurable such that $\mu(X\setminus E_k)\leq k^{-1}$ and $\sup_{x\in E_k}|f_n(x)-f(x)|\to 0$. We have, using the hypothesis, \begin{align} \int_X|f-f_n|d\mu&=\int_{E_k}|f-f_n|d\mu+\int_{X\setminus E_k}|f-f_n|d\mu\\ &\leq \mu(E_k)\sup_{x\in E_k}|f(x)-f_n(x)| +\int_{X\setminus E_k}|f-f_n|d\mu\\ &\leq \mu(X)\sup_{x\in E_k}|f(x)-f_n(x)| +\int_{X\setminus E_k}|f-f_n|d\mu. \end{align} Taking $\limsup_{n\to +\infty}$, we get $$\limsup_{n\to +\infty}\int_X|f-f_n|d\mu\leq \limsup_{n\to +\infty}\int_{X\setminus E_k}\left(|f-f_n|-(|f_n|-|f|)\right)d\mu,$$ using the fact that $\int_{X\setminus E_k}f_nd\mu\to \int_{X\setminus E_k}fd\mu$.

As $0\leq |f-f_n|-(|f_n|-|f|)\leq 2f$, we have for each $k$, $$\limsup_{n\to +\infty}\int_X|f-f_n|d\mu\leq \int_{X\setminus E_k}fd\mu.$$ Let $s_l$ a simple function such that $0\leq s_l\leq f$ and $\int_X (f-s_l)d\mu\leq l^{—1}$. We can write $s_l(x)=\sum_{j=1}^{N_l}a_{j,l}\chi_{B_{j,l}}$, where $B_{j,l}$ are measurable sets and $a_{j,l}$ positive real numbers. Then for each $k,l$ positive integers, \begin{align} \limsup_{n\to +\infty}\int_X|f-f_n|d\mu&\leq\int_{X\setminus E_k}(f-s_l)d\mu+ \int_{X\setminus E_k}s_ld\mu\\ &\leq \int_X(f-s_l)d\mu+\sum_{j=1}^{N_l}a_{j,l}\mu(B_{j,l}\cap E_k^c)\\ &\leq l^{-1}+\sum_{j=1}^{N_l}a_{j,l}\mu(B_{j,l}\cap E_k^c)\\ &\leq l^{-1}+k^{-1}\sum_{j=1}^{N_l}a_{j,l}\mu(B_{j,l}). \end{align} Taking the $\limsup_{k\to +\infty}$ then $\limsup_{l\to +\infty}$, we get the result.

share|cite|improve this answer
Giraudo: Thanks, but I have some questions: 1- Why do you assume $f \geq 0$? since $f_n \geq 0, sup |f-f_n|$ can happen when $f \leq 0\ \& f_n \geq 0$ so is it correct to change $\limsup_{n\to +\infty}\int_X|f-f_n|d\mu\leq \int_{X\setminus E_k}fd\mu$ to $\limsup_{n\to +\infty}\int_X|f-f_n|d\mu\leq \int_{X\setminus E_k}2|f|d\mu$? – Anita Oct 29 '12 at 12:15
$f\geq 0$ as $f_n(x)\geq 0$ and $f_n\to f$ almost everywhere. – Davide Giraudo Oct 29 '12 at 12:16
The latter part of your proof can be simplified by noting that if $f$ is integrable (and $X$ has finite measure), then the measure $\nu A = \int_A f$ is absolutely continuous, from which it follows that $\lim_k \int_{E_k^C} f = 0$. – copper.hat Oct 29 '12 at 16:24
@copper.hat Yes, you are right. But I was in the spirit to give a proof which only uses more basic facts about Lebesgue integral. – Davide Giraudo Oct 29 '12 at 16:30
@DavideGiraudo: are you assuming $|f_n|\geq |f|$? – Anita Oct 30 '12 at 16:07

I do not think this is true unless you assume $f$ is integrable.

Take $X=[0,1]$, Lebesgue measure, $f(x) = \frac{1}{x}$, $f_n = f \cdot 1 _{[\frac{1}{n},1]}$. Then $f_n(x) \to f(x)$ on $(0,1]$, $\int f_n = \log n$, hence $f_n$ are integrable, and $\int_E f_n \to \int_E f$ for all $E$ measurable. However, $\int |f_n-f| = \infty$ for all $n$.

share|cite|improve this answer
I agree (for sake of completeness, you should say that $\int_Ef_n\to \int_Ef$ for all $E$ measurable). – Davide Giraudo Oct 29 '12 at 16:32
Good suggestion! – copper.hat Oct 29 '12 at 16:49

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.