I'm reading Elementary Analysis by Ross and working through the proof of Theorem 11.7 (see below). I don't understand the claim in the proof that "... ($s_n$) is bounded above, so that lim sup $s_n$ is finite." For example, consider the sequence {$-n^2$} for all $n \in \mathbb{N}$. This is bounded above by $0$, but isn't lim sup $s_n$ = $-\infty$ in this case, and therefore not finite? What am I missing?

See various theorem references and proof below from the text:

Theorem 11.7:

Let ($s_n$) be any sequence. There exists a monotonic subsequence whose limit is $\limsup s_n$ and there exists a monotonic subsequence whose limit is $\lim inf s_n$.


If ($s_n$) is not bounded above, then by Theorem 11.2(ii), a monotonic subsequence of ($s_n$) has limit $+\infty = \limsup s_n$. Similarly, if ($s_n$) is not bounded below, a monotonic subsequence has limit $-\infty = \liminf s_n$.

The remaining cases are that ($s_n$) is bounded above or is bounded below. These cases are similar, so we only consider the case that ($s_n$) is bounded above, so that lim sup $s_n$ is finite. Let $t = \limsup s_n$, and consider $\epsilon > 0$. There exists $N_0$ so that

$\sup\{s_n : n > N\} < t + \epsilon$ for $ N \geq N_0$.

In particular, $s_n$ $ < t + \epsilon $ for all $n > N_0$. We now claim

$\{n \in \mathbb{N} : t - \epsilon $ < $s_n$ < $t + \epsilon\}$ is infinite.

Otherwise, there exists $N_1 > N_0$ so that $s_n$ $\leq t - \epsilon$ for $n > N_1$. Then $\sup \{s_n : n > N\} \leq t - \epsilon$ for $N \geq N_1$, so that $\limsup s_n < t$, a contradiction. Since (1) holds for each $\epsilon > 0$, Theorem 11.2(i) shows that a monotonic subsequence of $(s_n)$ converges to $t = \limsup s_n$.

Theorem 11.2(i):

If $t$ is in $\mathbb{R}$, then there is a subsequence of $(s_n)$ converging to $t$ if and only if the set $\{n \in \mathbb{N} : |s_n - t| < \epsilon\}$ is infinite for all $\epsilon > 0$.

Theorem 11.2 (ii):

If the sequence $(s_n)$ is unbounded above, it has a subsequence with limit $+\infty$.

  • $\begingroup$ Alternatively, and this might be obvious, but I thought it was worth mentioning, if $(s_n)_{n\in\mathbb{N}}$ is a sequence, then $\limsup(s_n)$ and $\liminf(s_n)$ are finite if and only if $(s_n)_{n\in\mathbb{N}}$ is bounded above and below. $\endgroup$
    – C Squared
    Jun 4 at 9:36

You are correct that there is a slight error here. The proof should probably say "so that $\limsup s_n<\infty$; if $\limsup s_n=-\infty$ then also $\liminf s_n=-\infty$ and we already have a suitable sequence, so we may assume $\limsup s_n$ is finite."

An alternative would be to start with the observation that if $\limsup s_n=\liminf s_n$ then any monotonic subsequence works for both, so we may assume they are different.


Let me first state a lemma:

Lemma: If $\limsup (x_n)$ is finite then $(x_n)$ is bounded above.

Proof: Let $s=\limsup (x_n)$ and so there exists $N\in \mathbb N$ such that we have $x_n\lt s+1$ for all $n\gt N$. It follows that $x_n\lt\max\{x_1,x_2,\cdots,x_N,s+1\}$. This proves our lemma.

Now, coming back to your question. You are right about observing that $\limsup (-n^2)=-\infty$ and this shows that converse to the lemma stated above is not true, which you probably thought should be true.

  • $\begingroup$ Could you please explain why $\limsup(-n^2) = -\infty$? $\endgroup$
    – fwd
    Jun 4 at 11:25
  • $\begingroup$ @fwd: Use definition of limsup. Every real no. bounds the sequence $(-n^2)$ from above when n is large enough so what's the lowest upper bound? $\endgroup$
    – Koro
    Jun 4 at 12:09
  • $\begingroup$ You're right, thanks. For some reason I was stupid and kept assuming only non-negative reals bound the sequence above. $\endgroup$
    – fwd
    Jun 4 at 12:15
  • $\begingroup$ @fwd: It's absolutely no problem. It's part of learning. Glad that you understand it now :) $\endgroup$
    – Koro
    Jun 4 at 12:17

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.