Homepage › Solution manuals › Terence Tao › An Introduction to Measure Theory › Exercise 1.3.23 (Lusin's theorem II)
Exercise 1.3.23 (Lusin's theorem II)
Let be a
- (i)
- locally absolutely integrable
- (ii)
- Lebesgue measurable (almost everywhere finite)
Show that Lusin’s theorem still holds for .
Answers
- (i)
- We repeat the proof of Lusin’s theorem, this time adjusting for locality. Pick an arbitrary . For all consider the restricted function . Then converges pointwise to (we will later use Egorov’s theorem to translate this into the uniform convergence). Since each is absolutely integrable by theorem assumption, we can find a set such that and is continuous on by Theorem 1.3.28. Let with . Furthermore, since is a sequence of Lebesgue measurable functions converging pointwise to , we can find a set such that converges uniformly to on . Thus, set with . Then our sequence of continuous functions converges uniormly to by Egorov’s theorem, and the uniform limit of continuous functions is again continuous.
- (ii)
- Now suppose that is Lebesgue measurable (and a.e. finite for the convergence to make sense). Define a sequence of Lebesgue measurable functions . This implies an almost everywhere “binary” pointwise convergence of against (which we will convert to uniform convergence later). Each is a bounded Lebesgue measurable function; by definition we can find a sequence of simple functions which converges to pointwise. Notice that are (eventually) bounded by the boundedness of their limit (that is where the becomes useful); thus are absolutely integrable by their simpleness. By Theorem 1.3.28 for each our simple functions may be converted to continuous ones by exluding a set from the domain. Set and for each we have are continuous. By Egorov’s theorem (Theorem 1.3.26) we can find a set such that the convergence becomes uniform , which implies the continuity of each . Repeating this step, we can find a set such that the sequence converges to uniformly. Alas, set . We then have, outside of the set the uniform convergence of the continuous functions to the , which then also must be continuous by Analysis II, Theorem 3.3.1.
2020-07-10 00:00
Comments
-
A technicality for the second part): I think, each $E \cup A$ for the $(s_{m,n})$ is in reality an $E_n \cup A_n$ so that for the last step one has to build $\bigcup_{n=1}^\infty (E_n \cup A_n$. This is just a problem of housekeeping, cause then one would start each $E_{n,m}$ with $m(E_{n,m} \leq \frac{\epsilon}{4^n}$ and $m(A_n) \leq \frac{\epsilon}{2^n}$.Shinkenjoe • 2025-01-30