Let $(a_n)$ be a bounded sequence of real numbers, and define $$\beta_n = \sup \{ a_k : k \geq n \}. $$ This sequence converges to a limit, $$\lim_{n \to \infty} \beta_n = \limsup a_n.$$ I'm interested in proving the existence of a subsequence of $(a_n)$ that converges to $\limsup a_n.$ I've read over the other posts concerning this and gave a satisfactory proof similar to these arguments, but I'm wondering if the following argument could be used as well. Note that in my study of analysis I've already proven the Bolzano-Weirstass theorem in $\mathbb{R}$. Here goes.

If $\beta_n$ is a bounded real sequence, Bolzano-Weierstrass implies that it has a convergent subsequence $(\beta_{n_j})$. Since $\lim \beta_n$ exists and is equal to $\limsup a_n,$ then $(\beta_{n_j})$ converges to $\limsup a_n.$

I'm not certain if this argument works because I want a subsequence of $(a_n).$ Is there a way to argue that $(\beta_{n_j})$ is a subsequence of $(a_n)$?

  • 3
    $\begingroup$ No because $(\beta _{n_j})$ is not a subsequence of $(a_n)$. Moreover, why taking a subsequence of $(\beta _n)$ since it already converge ? $\endgroup$ – Surb Dec 16 '18 at 21:47
  • $\begingroup$ math.stackexchange.com/questions/581128/… $\endgroup$ – parsiad Dec 16 '18 at 22:09

No, it doesn't work because $\beta_n$ needs not be an element of $(a_k)$.
We need a slight modification: for $n$, by the definition of supremum, there's an element $a_{k_n}>\beta_n-\frac1n$.
Then we get $L\leftarrow \ \beta_n-\frac1n<a_{k_n}\le\beta_n\ \to L$, where $L=\lim\beta_n$.

  • $\begingroup$ Great, thanks, I thought so! Appreciate the speedy response. $\endgroup$ – dove Dec 17 '18 at 0:23

In context:

$b_n= \sup (a_k| k \ge n)$

$A:= \lim_{n \rightarrow \infty} b_n=\lim \sup a_n.$

Show that $A$ is a limit point of $(a_n)_{n \in \mathbb{N}}$.

1)Let $\epsilon >0$ be given.

There is a $N$ such that for $n \ge N$

$|b_n -A| \lt \epsilon/2.$

2) By definition of $b_n = \sup (a_k|k \ge n)$ there is a

$a_{n_k}$ such that for $k \ge k_0$

$|a_{n_k}-b_n| \lt \epsilon/2.$

For $n \ge N$, and $k \ge k_0$

3) $|a_{n_k}-A| \le$

$ |a_{n_k}-b_n| +|b_n-A| \lt \epsilon$, i.e.

$A$ is a limit point of $(a_n)_{n \in \mathbb{N}}$, and $a_{n_k}$ converges to.$A$.

  • $\begingroup$ Thank you! I haven't been formally introduced to the notion of limit points yet, but this makes things clear. I assume this follows from a proof that if $A$ is a limit point of $(a_n)$ then there is a subsequence of $(a_n)$ converging to $A$? $\endgroup$ – dove Dec 17 '18 at 18:26
  • $\begingroup$ dove.Welcome. Yes.Hopefully adds a bit of context.I myself am not very much at ease with limsup and liminf. Bounded a_n has limit points, a subsequence converges to every limit point.Shown above: limsup is a imit point A of (a_n), hence a subsequence converges to A.Greetings. $\endgroup$ – Peter Szilas Dec 17 '18 at 19:22

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.