I've been working with proofs involving $\limsup$ and $\liminf$, and I'm a bit confused regarding their general methodology. More specifically, I'm unsure about whether my approach to the following problem makes sense.

Problem. Let $(s_n)$ and $(t_n$) be sequences and suppose that there exists $N_0$ such that $s_n \leq t_n$ for all $n > N_0$. Show that $\liminf s_n \leq \liminf t_n$ and that $\limsup s_n \leq \limsup t_n$.

The way I approached it was as follows:

Let $N > N_0$. Then $\limsup_{N \rightarrow \infty} \{ s_n : n > N \} \leq t_n$ as $s_n \leq t_n$, and $\limsup s_n $ is the largest possible limit of a subsequence of $s_n$. As $t_n : n > N$ is (by definition) less than $\limsup_{N \rightarrow \infty} \{ t_n : n > N \}$, the proof is complete.

I'm pretty sure this is incorrect, however, and I'm generally unclear about the method behind such a proof. Any help is appreciated!


Let $x_n = t_n - s_n$. Your problem is equivalent to the following: if $x_n \ge 0 \ \forall n \ge N$, then $\liminf x_n \ge 0$.

Remember that, for sequences of real numbers, $\liminf$ is the $\inf$ of the set of limit points of $\{x_n\} _{n \ge N}$. I find working with this description of $\liminf$ very visual, therefore very productive - give it a try! Therefore, let us focus on this set of limit points, first.

Let $x$ be a limit point of the set $\{x_n\} _{n \ge N}$. This means that there exist a subsequence of $(x_n) _{n \ge N}$ which tends to $x$. Since all the $x_n$ are positive, so will be the terms of the subsequence, hence so will be its limit: $x \ge 0$. We have discovered that the set of limit points is made up of only positive numbers.

Next, you have to take the $\inf$ of this set. But the $\inf$ of a set of positive numbers is itself positive, which is exactly what you needed to show.

A similar argument is for $\limsup$.

Notice that the essential ingredients that we have used here are that the limit of a sequence of positive numbers is itself positive, and that the $\inf$ of a set of positive numbers is itself positive.

  • $\begingroup$ Thanks! This definitely helps. I always thought of $\liminf$ as the limit of the set of infima (which we know exists as it is nondecreasing, and therefore monotone). Does this definition reconcile with the one you listed in your answer? I guess I'm still a bit confused regarding the meaning of the terms. $\endgroup$ – manan Jul 17 '16 at 15:41
  • $\begingroup$ @manan: Yes, the two are equivalent, and are in turn equivalent to several other definitions. The one that I chose to work with is the one that I have found most helpful in reasoning about problems. $\endgroup$ – Alex M. Jul 17 '16 at 15:54
  • $\begingroup$ Ok, great! One more thing -- would you mind telling me what was incorrect about my original idea? I understand your approach, but I also want to see where I went wrong the first time so I don't make the same mistake in the future. Thanks :) $\endgroup$ – manan Jul 17 '16 at 15:56
  • $\begingroup$ @manan: It is essentially correct, but with a caveat: it is not true that, in general, $x_n \le \limsup x_n$ - there might be a finite number of terms larger than $\limsup x_n$. Think of the sequence $1, 0, 0, \dots$ (there is a single $1$, on the first position): the $\limsup$ is not $1$ (because there is no sub-sequence tending to $1$), it is $0$ - so the first term of the sequence is greater than the $\limsup$. Similarly for $\liminf$. So your proof is correct, save for a finite number of terms, which are not important when talking about limits, anyway. $\endgroup$ – Alex M. Jul 17 '16 at 16:03
  • $\begingroup$ Awesome, thank you so much! $\endgroup$ – manan Jul 17 '16 at 16:13

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.