I have been reading Terry Tao's notes on Real Analysis and there's a part he just says, but does not really explain, so I am wondering if someone here would. The notes are http://terrytao.wordpress.com/2010/10/02/245a-notes-4-modes-of-convergence/ and my particular question is from Section 4, Corollary 3. It goes as follows,

Let $f_n \rightarrow f$ in $L^1$ then there exists a sub sequence $(f_{n_j}) \subset (f_n)$ such that $f_{n_j} \rightarrow f$ pointwise a.e. Moreover $(f_{n_j})$ converges almost uniformly to $f$.

The proof he gives is simply that since $\|f_n-f\|_1 \rightarrow 0$ as $n \rightarrow \infty$ we can pick a sub sequence such that $\|f_{n_j}-f\|_1<2^{-j}$ which is enough to show pointwise a.e and almost uniform convergence. But what allows you to pick such a sub sequence is it maybe some Cauchy property or is it some weird construction? Then how do you go from that to pointwise a.e and even almost uniform convergence. I am assuming that for almost uniform, you do something similar to Egorov's theorem without the assumption the domain of $f$ has finite measure. Also I am aware that if you get almost uniform, you immediately have pointwise a.e, but I'd like to see how to get to both. Thank you.


1 Answer 1


Choose $N_{k}$ such that $N_{1} \le N_{2} \le N_{3} \le \cdots$ and such that $\|f_{m}-f_{n}\| < 1/2^{k}$ whenever $m, n \ge N_{k}$. This is possible because $\{ f_{n}\}_{n=1}^{\infty}$ is a Cauchy sequence. Then $\{ f_{N_{k}}\}_{k=1}^{\infty}$ is a subsequence such that $\|f_{N_{l}}-f_{N_{m}}\| < 1/2^{k}$ whenever $l,m \ge k$. Then $$ f_{N_{m}}=f_{N_{1}}+\sum_{l=1}^{m}f_{N_{l+1}}-f_{N_{l}} $$ converges pointwise a.e. absolutely because $$ g_{m}=|f_{N_{1}}| + \sum_{l=1}^{m}|f_{N_{l+1}}-f_{N_{l}}| $$ converges pointwise a.e. to an extended real function $0 \le g \le \infty$ such that, by the monotone convergence theorem, $$ \int g\,d\mu = \int |f_{N_{1}}|d\mu+\sum_{l=1}^{\infty}\int |f_{N_{l+1}}-f_{N_{l}}|d\mu = \|f_{N_{1}}\|+\sum_{l=1}^{\infty}\|f_{N_{l+1}}-f_{N_{l}}\| < \infty. $$ So $g < \infty$ a.e., which means that $\lim_{l}f_{N_{l}}$ converges pointwise a.e.. to an $L^{1}$ function.

  • 2
    $\begingroup$ $\int|\lim f_{n_l}|=\int \lim|f_{n_l}|\leq \int\lim g_l=\int g<\infty$. $\endgroup$
    – KC.
    Jul 11, 2017 at 19:04
  • $\begingroup$ Uniqueness: Denote the $L^1$ limit of ${f_n}$ by $\hat{f}$. By dominated convergence theorem, $$\int | \lim f_{N_m} - \hat{f} | d\mu=\lim \int | f_{N_m} - \hat{f} | d\mu = \lim \| f_{N_m} - \hat{f} \|=0,$$ where the dominating $L^1$ function is $g+ | \hat{f} |.$ So $\lim f_{N_m} = \hat{f}$ a.e. $\endgroup$
    – Sam Wong
    Jul 25, 2021 at 10:28
  • $\begingroup$ The sum in $f_{N_m} = f_{N_1} + \sum_{i=1}^{m} f_{N_{i+1}} - f_{N_i}$ should only be taken up to $m-1$. $\endgroup$ Jan 1 at 2:25

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.