Let $\{s_n\}$ be an increasing sequence of step functions which converges pointwise on an interval $I$ to a limit function $f$. If $I$ is unbounded and if $f(x) \geq 1$ everywhere on $I$ except on a set of measure zero, prove that the sequence $\{\int_I s_n\}$ diverges. (Here $\int_I s_n$ refers to the Lebesgue Integral.)

Conceptually, this makes sense. Clearly, if $f(x) \geq 1$ almost everywhere on an unbounded interval, then the sequence of Lebesgue integrals should diverge. However, in proving this, I am running into issues with the convergence of $\{s_n\}$ being pointwise. If it were uniform, I could easily show this.

How do I show that (even for pointwise convergence) I can choose $n$ large enough to make $\int_I s_n$ as big as I want?

  • $\begingroup$ you can use MCT on this. $\endgroup$ – Matematleta Mar 17 '16 at 20:18
  • $\begingroup$ @Chilango: don't you need $s_n\ge 0$ for MCT? $\endgroup$ – user251257 Mar 17 '16 at 20:34
  • $\begingroup$ yes, and I misread the problem. I thought they were. $\endgroup$ – Matematleta Mar 17 '16 at 20:46
  • $\begingroup$ can you consider $f_n =s_n\chi_{s_n^{-1}((0,\infty)))}$ ? $\endgroup$ – Matematleta Mar 17 '16 at 21:08
  • $\begingroup$ @user251257 If $f>0$ a.e. and $s_n\to f$ pointwise then $s_n\geqslant0$ a.e. for all but finitely many $n$. $\endgroup$ – Math1000 Mar 17 '16 at 22:27

I take step functions mean at least that the image of $s_n$ is finite and $\{s_n \ne 0\}$ is finite measured. In particular, $s_n$ is integrable.

Assume $s_1^- := \max(-s_1,0)$ has positive integral. Otherwise we can apply monotone convergence theorem directly.

Notice that $\int_I s_1^-$ is finite, as $s_1$ is a step function and thus integrable. Then, $s_n + s_1^- \ge s_1 + s_1^- = s_1^+ = \max(s_1,0)$ is non-negative and increases pointwise to $f - s_1^-$. Thus, MCT applies and yields $$ \lim_{n\to\infty} \int_I s_n + s_1^- = \int_I f + s_1^- \ge \int_I f = \infty. $$ Thus, it follows $$ \int_I s_n = \int_I s_n + s_1^- - \underbrace{\int_I s_1^-}_{< \infty} \to \infty. $$

Aside: Notice that $s_n$ is a step function is not crucial. We can replace it with "$s_n$ is integrable".


Does this work?:

We have that $s_n$ converges from below to $f$ almost everywhere on I.

Assume $\lim_{n\rightarrow \infty} \int_I s_n$ is finite. Then $f$ is an upper function, denoted $f \in U(I)$.

Then we have that $\lim_{n\rightarrow \infty} \int_I s_n = \int_I f$.

However, $f \geq 1$ a.e. on $I$. So $\int_I f\geq \int_I 1 = 1 \cdot \mu(I) = \infty$.

Therefore $f \notin U(I)$ and $\int_I s_n$ diverges.

  • $\begingroup$ You need to prove that $\lim_{n\rightarrow \infty} \int_I s_n = \int_I f$ $\endgroup$ – Matematleta Mar 17 '16 at 22:39
  • $\begingroup$ what def are you using? $\endgroup$ – Matematleta Mar 17 '16 at 23:33

Let $f_n =s_n\chi_{s_n^{-1}((0,\infty)))}$. Then, $f_n\geq 0$ are measureable, and increase to $f$ a.e.:

Let $x\in I$. Since $s_n(x)\nearrow f(x)\geq 1$ we can find an $N\in \mathbb N$ s.t. $n>\mathbb N\Rightarrow s_n(x)>0$. But then, $s_n(x)=f_n(x)$.

Now use MCT.


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.