4

How can I find the limit superior and inferior of given sequence: $x_n = (1 + \frac{1}{2n})\cos{\frac{n\pi }{3}}$ as $ n \in \mathbb N $

I did the following: since $\lim_{n\to\infty}(1 + \frac{1}{2n}) = 1$ and the second term' limit oscillates between -1 and 1, I decided that the supremum should be $1$ and infimum should be $-1$. But this is an incorrect answer.

What I do wrong? Thank you.

UPD: It is a task from the online-courses site and there is an automatic answer checking system. So, as I said in my question, answer $-1$ and $1$ not passed and there is no explanation why.

I believe there should be used Bolzano–Weierstrass theorem to find all possible convergent subsequences, and then find limits for each of them...

Alex
  • 157
  • 1
    It does not look wrong to me; all values are close to one of $-1,-\frac12,\frac12,1$, and get arbitrarily close as $n$ increases – Henry Apr 11 '19 at 22:56
  • Henry, unfortunately, the answer $-1 \frac{-1}{2} \frac12 1$ is wrong as well as the answer $-1 1$. I edited my question so there could be new helpful information. – Alex Apr 12 '19 at 06:12

1 Answers1

1

We have $\lim_{n\to\infty}(1+\frac{1}{2n}) = 1$ and $0 \le |\cos(x)| \le 1$ for all $x \in \Bbb R$, so $$\limsup_n \,(1+\frac{1}{2n}) = 1 \text{ and } \limsup_n \, \left|\cos{\frac{n\pi }{3}} \right| \le 1.$$ Using $\limsup ( a_n b_n ) \leqslant \limsup a_n \limsup b_n$ (with $a_n = 1 + \frac{1}{2n}$ and $b_n = |\cos\frac{n\pi}{3}|$), we have $$\limsup_n \, (1 + \frac{1}{2n}) \left|\cos{\frac{n\pi }{3}} \right| \le 1,$$ so $$-1 \le -\limsup_n \, (1 + \frac{1}{2n}) \left|\cos{\frac{n\pi }{3}} \right| \le \liminf_n\, (1 + \frac{1}{2n}) \cos{\frac{n\pi }{3}}.$$

(Recall that $\liminf_n a_n$ can be equivalently defined as the minimum possible limit of subsequence $(a_{n_k})_k$, which is bounded below by $(-|a_{n_k}|)_k$.)

  • $n = 6k + 3$: $x_n = (1 + \frac{1}{12k + 6})(-1) \to -1$ as $k \to \infty$.
  • $n = 6k$: $x_n = (1 + \frac{1}{12k})(1) \to 1$ as $k \to \infty$.

This shows that $\limsup_n x_n = 1$ and $\liminf_n x_n = -1$.


(Edit in response to OP's update)

OP has appealed to his online system that we have no access to to conclude that my answer is wrong.

Here, I'm going to prove that why $$\limsup_n \,x_n < 1+\epsilon \quad \forall \epsilon > 0.$$ In other words, wny number $1 + \epsilon$ strictly greater than $1$ can't be a limit of any subsequence of $(x_n)_n$. With the example subsequence $n = 6k$ in the last part of my original answer, this would show that $\limsup_n x_n = 1$.

The idea is simple: $\require{cancel}$ $$x_n = \underbrace{\left( \color{red}{1+\frac{1}{2n}} \right)}_{\to 1} \; \xcancel{\color{blue}{\cos{\frac{n\pi }{3}}}}.$$

Wiki's picture Source: Wikipedia's page on $\limsup$ in French

To make it rigorous, for any $\epsilon > 0$, take $N = \left\lceil \dfrac{1}{\epsilon} \right\rceil$. For all $n \ge N$, $$\left( 1 + \dfrac{1}{2n} \right) \cos{\frac{n\pi }{3}} \le \left( 1 + \dfrac{1}{2n} \right) \, (1) \le 1 + \dfrac12\:\epsilon < 1 + \epsilon,$$ which violates the later part of (1) in the proposition in an older question: the above inequality shows that $x_n \le 1 + \dfrac\epsilon2$ eventually, instead of $x_n > (1 + \epsilon) - \dfrac\epsilon2$ frequently.


A shorter alternative solution: use the following computation rule (Proposition 2.4 from this note).

Let $(a_n)_n$ and $(b_n)_n$ be two bounded real sequences such that $b_n$ converges to $b \ge 0$. Then \begin{align} \limsup_n\, a_nb_n &= b \limsup_n\, a_n \text{ and } \\ \liminf_n\, a_nb_n &= b \liminf_n\, a_n. \end{align}

Apply this rule with $a_n = \cos\dfrac{n\pi}{3}$ and $b_n = 1 + \dfrac{1}{2n}$ to get the same conclusion.

  • Thank you, but the answer $-1$ and $1$ is wrong. I edited my question, can you look at this again, maybe it will be helpful? – Alex Apr 12 '19 at 06:10
  • @Alex I'm sorry to say that your edit doesn't make any difference to the community. This is Math.SE, and we exchange maths. We deal with maths issues, not technical ones. You may question my arguments here, but nobody can help with your system. Since we have no access to your system, we can't debug your system. If you bother to visit GitHub, where many softwares are hosted, you'll be astonished by the amount of issues/errors and pull requests of any popular project. Btw, a math book may contain errors, as one can see from the errata. – GNUSupporter 8964民主女神 地下教會 Apr 12 '19 at 09:09
  • Sorry, you probably misunderstood me. The platform I was talking about is stepik site. I don't think we should try to find bugs in their platform. I just wanted to say that this answer is incorrect there and I only have some hint (I wrote it at the end of my edit): "I believe there should be used Bolzano–Weierstrass theorem to find all possible convergent subsequences, and then find limits for each of them" – Alex Apr 12 '19 at 10:56
  • @Alex Why's the mathematical reason that you think my answer is incorrect? – GNUSupporter 8964民主女神 地下教會 Apr 12 '19 at 14:54
  • if I'd knew it, I wouldn't ask my question here. Again and again: my update contains 2 parts. In the first one, I said that this answer is incorrect and explained why (!) I don't know why this answer is incorrect. Please, notice the second part of my edit: there are a hint from course teacher that Bolzano-Weierstrass theorem should be used to find all existing convergent subsequences of the sequence given, and then find limits for each of them. Thank you. – Alex Apr 13 '19 at 12:14
  • What can be seen in your "explanation" is just an appeal to an online system that we don't have access to. The Bolzano-Weiertrass theorem isn't needed here. Just reason from the basics will do: 1. limsup as the maximum possible limit of a subsequent 2. ordinary limit, if it's defined, is just limsup/liminf 3. – GNUSupporter 8964民主女神 地下教會 Apr 13 '19 at 18:27
  • 1
    @Alex I've [edit]ed my answer in response to your update. You may invoke the Bolzano-Weierstrass Theorem to say something about the liminf and limsup of $\cos(\cdots)$ in the alternative approach, but it's fine to draw the conclusion without it. – GNUSupporter 8964民主女神 地下教會 Apr 19 '19 at 11:59