Define a sequence $(a_n)_{n = 1}^{\infty}$ of real numbers by $a_n = \sin(n)$. This sequence is bounded (by $\pm1$), and so by the Bolzano-Weierstrass Theorem, there exists a convergent subsequence.

My question is this: Is there any way to construct an explicit convergent subsequence?

Naïvely, I tried considering the subsequence $$\sin(3), \,\sin(31), \,\sin(314), \,\sin(3141),\, \sin(31415),\, \dots$$ hoping it would converge to $0$, but I was unable to prove this (and I think it's probably not even true).

  • 3
    $\begingroup$ You can come up with a subsequence that converges to $1$ (or any other value you choose in $[-1,1]$): Let $a_1=\sin(1)$. Then let $a_2=\sin(k_2)$ where $k_2$ is the least integer for which $\sin(k_2)\geq \sin(1)$. In general, let $a_n = \sin(k_n)$ where $k_n$ is the least integer for which $\sin(k_n) \geq \sin(k_{n-1})$. A similar process gives a subsequence that converges to any value in $[-1,1]$ you choose. The proof shouldn't be too taxing. However, it would be quite a push to call this an explicit subsequence. $\endgroup$ May 21 '14 at 7:27

This isn't a very analytic way of doing it but consider the continued fraction expansion $\pi = [a_0; a_1, a_2, \ldots]$. Generally we know that, if $p_n/q_n$ are the convergents to $\pi$ then $|q_n\pi - p_n| \leq 1/q$ so the $p_n$ are the closest integers to being multiples of $\pi$ so (with slight abuse of notation) $\sin(p_n) \rightarrow \sin m\pi = 0$ with $n \rightarrow \infty$ meaning that this is an explicit construction of the sort that you're after.

  • $\begingroup$ I've just realised that I've written this in a really weird way, you want are the numerators of the convergents $p_n$ so the sequence starts $3, 22, 333, 355, 103993, \ldots$ and then $\sin$ of this sequence tends to $0$ pretty darn quickly. $\endgroup$ May 21 '14 at 7:47
  • $\begingroup$ then either edit your answer to clarify, or even delete it and write a new, clear o one $\endgroup$
    – vonbrand
    May 21 '14 at 8:24
  • $\begingroup$ I have but I forgot to delete my comment. $\endgroup$ May 21 '14 at 12:31

I am not very confident that it converges, because this is how well you can approximate $10^k \cdot \pi$ with an integer. But as the difference will be infinite many times greater than $0.9$ and as you get bounds you see that this sequence is divergent.

I think this is more a question what is explicit for you, because you need to know how well you can approximate a multiple of $\pi$ with an Integer.


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.