Let $f \in L^1(\mathbb{R})$. Find $$ \lim_{n \rightarrow \infty} \int_{-\infty}^\infty f(x)\sin(nx) dx \,. $$

Here's an attempt:

If we use a simple function approximation, we end up with something of the form $$\int_\mathbb{R} \sum_{i=1}^{k}c_i \chi_{E_i}(x)\sin(nx)dx = \sum_{i=1}^k c_i \int_{E_i} \sin(nx)dx$$ which goes to $0$ as $n \to \infty$ (The $E_i$'s can be chosen to be disjoint). My issue is that at least one of these $E_i$'s would necessarily be infinite. How do I fix this?

Does it then suffice to take a sequence of simple functions $s_m$ such that $$||s_m - f ||_{L^1} < \epsilon$$ for $n \geq N$ and each $s_m$ agrees with $s_{m-1}$ in it's sum except for the infinite term in which it splits the infinite $E_i$ further? That is, suppose $$s_{m-1} = \sum_{i=1}^{k} c_i \chi_{E_i}$$ with $||s_{m-1} - f||_{L^1} < \epsilon$. Wlog, suppose $E_k$ has infinite Lebesgue measure. Then surely there exists $s_m$ with $$ s_m = s_{m-1} + c_k'\chi_{E_k'} + c_{k+1}\chi_{E_{k+1}}$$ where $E_k' \cup E_{k+1} = E_k$ and $||s_m - f||_{L^1} < \epsilon$. Does this suffice? Can I then proceed to claim that

$$\lim_{n \to \infty} \int_\mathbb{R} (\lim_{m \to \infty} s_m) \sin(nx) \, dx = 0$$ by swapping the sum and the limits to get $$\sum_{i=1}^{\infty} \lim_{n \to \infty}\int_{E_i} c_i \sin (nx) \, dx$$ with $\mu (E_i) < \infty$? Or is this swapping not justified?

The previous post about this question is here.
Summary of prev post: Use the uniform continuity of the linear functional $$\mathcal{F}(f) = \lim_{n \rightarrow \infty} \int_{-\infty}^\infty f(x)\sin(nx) dx$$ to extend to all of $L^1$ using the density of something.

  • $\begingroup$ Why would you want an infinite $E_i$? That would give you a function not in $L^1$. $\endgroup$ – Robert Israel Oct 23 '15 at 5:15
  • $\begingroup$ @RobertIsrael Poor wording. By "need" I meant that if I split $\mathbb{R}$ into finitely many disjoint intervals, one $E_i$ would necessarily be infinite. I want to avoid this. $\endgroup$ – Anthony Peter Oct 23 '15 at 5:19

Hint: there is a finite interval $E$ such that $\int_{E^c} |f(x)| \; dx < \epsilon$. Then approximate the restriction of $f$ to $E$ by simple functions...

  • $\begingroup$ Is it not enough to say that since $s \in L^1$, either $\mu(E_i) < \infty$ forall $i$, or $c_i = 0$ for $\mu(E_i) = \infty$? Or is this essentially saying the same thing since $f$ can be approximated by $s$? I think I'm trying to say this is reverse and less elegantly... I believe what I've said in this comment implies that such an $E$ exists implicitly $\endgroup$ – Anthony Peter Oct 23 '15 at 6:09
  • $\begingroup$ Namely, if $||f-s||_{L^1} < \epsilon$, then if $$E = \bigcup_{i=1}^{k} E_i$$ $$\int_{E} |f| dx < \epsilon.$$ $\endgroup$ – Anthony Peter Oct 23 '15 at 6:15
  • $\begingroup$ This is part of how you approximate an $L^1$ function by a simple function. $\endgroup$ – Robert Israel Oct 23 '15 at 6:17
  • $\begingroup$ So then I believe on this $E$ the integral is easily found to be $0$ and we can extend properly to all of $L^1$. Is there anything else that I need take into consideration? $\endgroup$ – Anthony Peter Oct 23 '15 at 6:19

I'd first prove the limit statement for $f\in L^1$ that is also a step function: $f(x) = \sum_{k=1}^n 1_{(a_k,b_k]}(x)\cdot c_k$, where $a_1<b_1\le a_2<b_2<\cdots\le a_n<b_n$, and the $c_k$ are real. Then show that a general $f\in L^1$ can be approximated in $L^1$ by such step functions. For this second step it might be helpful to first approximate $f\in L^1$ by continuous functions of compact support.


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.