I have tried to prove the following result, but it looks a bit messy in my opinion. How can it be improved?
Proposition. Let $a_n$ be a bounded sequence with values in $\mathbf{R}$. Let $A=\{l\in\mathbf{R} \mid \text{ there exists a subsequence of $a_n$ >converging to }l\}$. Then $A$ is closed.
Lemma: For each subsequence of $a_n$ which converges to $l$ it is true that $\forall {\varepsilon}>0$ $a_n\in B(l,{\varepsilon})$ for arbitrarily large $n$ -- that is, for each $m\in\mathbf{N} \; \exists M\geq m \mid a_M\in B(l,{\varepsilon})$.
Let $a_{k_n}$\,be such a subsequence. By definition, for each $\varepsilon>0 \; \exists N$ such that $\forall n\ge N \; a_{k_n}\in B(l,\varepsilon)$. Let $m\in\mathbf{N}$. Then $k_m\ge m$, since $a_{k_n}$ is a subsequence of $a_n$.
- if $m\ge N$, take $M:=k_m$;
- if $m<N$, take $M:=k_N$.
Proof. Let $x$ be a closure point of $A$. By definition, $\forall \varepsilon>0 \; B(x,\varepsilon)\cap A\neq\varnothing$, which means that $\forall \varepsilon>0$ there exists a subsequence $a_{k_n}$ which converges to ${l}_\varepsilon\in B(x,\varepsilon)$.
This means in particular that for all $m\in\mathbf{N}$ there exists a subsequence $a_{k^{(m)}_n}$ that converges to $l_m\in B(x,1/2^{m+1})$ .
Define the sequence $x_n$ inductively:
- $x_0=0$;
- $x_n=\min\{m>x_{n-1} \mid a_m \in B(l_m,1/2^{m+1}) \}$.
The sequence is well-defined, because there exists a subsequence converging to $l_m$, and by the lemma there are infinite indices for which the sequence lies in the neighborhood. So the set is not empty and by well-ordering there exists a minimum element.
It follows that for each $n\in\mathbf{N}$ $$x-1/2^{m}<l_m-1/2^{m+1}<a_{x_n}<l_m+1/2^{m+1}<x+1/2^{m},$$
so by squeeze theorem (strict inequalities imply non-strict inequalities) $a_{x_n}\to x$.