While reading the following paragraph from Real Analysis by Stein (I hope this does not breach any copyright; if so, I have to type it out), two questions occurred to me.

  1. In the proof of Fauto's lemma, it reads $\int g_n \leq \int f_n$, for all $n$, and therefore, $$\int g = \lim_n \int g_n \leq \liminf_n \int f_n.(+)$$ I understand that the reason why $\lim_n g_n$ exists is that $g_n$ is bounded and measurable on a finite support. And in addition, if $\lim_n g_n$ exists, then one has $$\lim_n g_n = \liminf_n g_n = \limsup_n g_n. (*)$$ However, I do NOT understand why one can take limit on the left hand side of $(+)$ and take liminf on the right hand side of $(+)$. I do NOT think one can use $(*)$ to adjustify $(+)$. If so, one might replace $\liminf_n f_n$ by $\limsup_n f_n$. That is, the theorem becomes $$\int g = \lim_n \int g_n \leq \limsup_n \int f_n. (++)$$ Is $(++)$ still true? Could anyone clarify this, please?
  2. Corollary $1.8$ seems to me stronger than the usual Monotone Convergence Theorem (DCT) since all it needs is to have $f_n \leq f$ in this corollary (plus pointwise convergence of course). There is no requirement on how $f_n$ should converge to $f$. To be more specific, $f_n$ certainly does NOT need to increasing at all. My question is then why one always works with DCT. Isn't it natural to work with a theorem which imposes looser constraint, please? Thank you!


  • $\begingroup$ 1. If $a_n\leq b_n$, then $\liminf a_n \leq \liminf b_n$ 2. It's a (version of) dominated convergence theorem, not MCT. Different theorem. In my (admittedly, limited) experience, indeed more easily applicable and used more often. $\endgroup$ Jun 18 '14 at 8:05
  1. For two real sequences $a_n \leq b_n$, there always results $\liminf_n a_n \leq \liminf_n b_n$. In your case, $\liminf_n a_n = \lim_n a_n$.
  2. I don't really understand your question. Who "always works with MCT"? The MCT requires the monotonicity assumption to be satisfied, while Corollary 1.8 requires a dominating function. These theorems have different assumptions, and are not comparable in a general framework. We use theorems whenever they can be used, you know.
  • $\begingroup$ MCT is a mis-type. I meant DCT as the words suggested. $\endgroup$
    – LaTeXFan
    Jun 18 '14 at 8:44
  • $\begingroup$ As for 1, if what you said is true, then can I write the theorem as $\int g = \limsup \int g_n \leq \limsup \int f_n$, please? $\endgroup$
    – LaTeXFan
    Jun 18 '14 at 8:46
  • $\begingroup$ @20824 Note that $$\liminf \int f_n \leq \limsup \int f_n$$ holds in any case. $\endgroup$
    – saz
    Jun 18 '14 at 8:57
  • $\begingroup$ Since $\inf \leq \sup$, it is wiser to use the best inequality, since $\sup = +\infty$ is allowed. And I can't understand question 2, again. $\endgroup$
    – Siminore
    Jun 18 '14 at 9:05

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.