I have been reviewing measure theory and am a bit confused over which integral definition to use:

If we let $(X, A, \mu)$ be a measure space then

Firstly we have an integral of a non-negative simple function: $\int_{X}f d\mu := \sum_{i=1}^{N}a_{i}\mu(A_{i})$ where $f:= \sum_{i=1}^{N}a_{i}\chi_{A_{i}}$

Then we progress to the integral of a non-negative measurable function. This is the sup over all non-negative simple functions.

Finally, we progress to an integral of a general measurable function which we split into the positive and negative part.

I understand all of the above.

When it comes to apply this integration theory to any problems I always get confused what definition to use. If we are given a non-negative measurable function or a general measurable function that we want to integrate, is the general approach to rewrite it, to be a simple function. Then simply apply that definition?

As a concrete example of what I mean I found the following exercise online

Let $(X, A, \mu)$ be a measure space and $f: X \rightarrow [0, \infty]$ be a measurable function such that $0 < c:= \int_{X}f d\mu < \infty$. Prove that

$$\lim_{n \to \infty} \int_{X} n \log \Big(1 + \frac{f^{\alpha}}{n^{\alpha}}\Big)d \mu = c $$ (if alpha = 1)

Now I know the different theorems, MCT, DCT, fatou etc. I understand the theory but I have no idea how to apply it to the above problem since the integral doesn't contain any characteristic functions and/or simple functions...

Please let me know if things aren't clear as I really want to be able to apply this theory to problems!

  • $\begingroup$ the general definition of being integrable is that $f$ is measurable and $\int|f|$ is finite, you can achieve this conclusion using various theorems. For you case note that $n\log\left(1+\frac{f^\alpha }{n^\alpha }\right)$ is non-negative and increasing so you can use the monotone convergence theorem $\endgroup$
    – Masacroso
    Mar 14 '20 at 10:05
  • $\begingroup$ @Masacroso the non-negative is obvious. To show it is increasing can I just say that if we write $f_{n} = n\log\left(1+\frac{f^\alpha }{n^\alpha }\right)$ then we will have that $f_{n} \leq f_{n+1}$? Applying MCT lets us bring the limit inside the integral so I can rewrite it as $\int_{X} \lim_{n \to \infty} \log(1+\frac{f}{n})^{n} = \lim_{n \to \infty} \log(e^{f}) = f$ (assuming alpha = 1). If alpha isn't 1 then the limit becomes $\lim_{n \to \infty} \log(e^{\alpha f})$?? $\endgroup$
    Mar 14 '20 at 11:15
  • $\begingroup$ yes, you can do this. For $\alpha >1$ just write $n=n^{1-\alpha }n^{\alpha }$ and apply the dominated convergence theorem $\endgroup$
    – Masacroso
    Mar 14 '20 at 11:18
  • $\begingroup$ @Masacroso I will try and apply the DCT however i always get stuck on proving that a function meets the necessary conditions to use it. $\endgroup$
    Mar 14 '20 at 11:25

Suppose that $\alpha >1$, then for some non-negative function $f$ we have that $$ n^{1-\alpha }\ln \left(1+\frac{f(x)^{\alpha }}{n^{\alpha }}\right)^{n^\alpha }\leqslant \ln \left(1+\frac{f(x)^{\alpha }}{n^{\alpha }}\right)^{n^\alpha }\leqslant f(x)^{\alpha } $$ for all $x$ and all $n\in \Bbb N\setminus\{0\} $, then you can apply the dominated convergence theorem whenever $\int f(x)^{\alpha }\mathop{}\!d x<\infty $. When $\alpha =1$ you can use the monotone convergence theorem to finish.

  • $\begingroup$ sorry I'm confused. To apply DCT don't I need to show the following two conditions: 1): The sequence of functions converge pointwise a.e. and 2): There is some other integrable function $g$ such that for all $i \in \mathbb{N}$ $|f_{i}| \leq g$ a.e.? I also don't fully understand why the middle term in your inequality is to the power $n^{\alpha}$ and not just $n$? Apologies, measure theory really confuses me when it comes to applying it to actual problems. $\endgroup$
    Mar 14 '20 at 11:38
  • $\begingroup$ @VBACODER if $f(x)\neq\infty$ then the sequence for such $x$ of $$n^{1-\alpha }\ln \left(1+\frac{f(x)^{\alpha }}{n^{\alpha }}\right)^{n^\alpha }$$ converges to zero $\endgroup$
    – Masacroso
    Mar 14 '20 at 11:40
  • $\begingroup$ I think I need to review the basics as I'm not seeing it... My analysis knowledge is awful :( $\endgroup$
    Mar 14 '20 at 11:42
  • $\begingroup$ for 2): this is just because $n=n^{1-\alpha }n^{\alpha }$ $\endgroup$
    – Masacroso
    Mar 14 '20 at 11:42
  • $\begingroup$ This is where I get confused. Because we also need to show that $g$ is an integrable function. Also, how do we then know that $n=n^{1- \alpha}n^{alpha} \geq |f|$ a.e. [assuming you're talking about (2) that I said]. $\endgroup$
    Mar 14 '20 at 11:47

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.