Does a limit that has the value of infinite exist or not?
I've recently come across certain sources that say that if the value of a limit is infinite, then that limit does not exist. This contradicts what my calculus teachers and lecturers taught me however, that a limit doesn't exist if the right hand limit and left hand limit differ.
So which one is it?
Obviously it depends on the definition of "exists". Some authors explicitly work over the extended real line with $\pm\infty$ adjoined, so that such infinite limits do explicitly "exist" as first-class values. But there is no consensus. One needs to pay attention to the author's definitions and conventions.
Perhaps it is worth mention - even though this case is rather trivial - that adjoining points at infinity is a special case of various constructions that attempt to simplify matters by some type of existential closure. Below I append an excerpt from an old 1996 post where I give a little further discussion:
This thread originated in a query as to whether infinity or $1/0$ could be admitted as a "value", and soon drifted into discussion of the Riemann sphere and other topological manifestations of infinity via compactification. Below I point out a couple of marvelous references on these topics; further I would like to bring to your attention a much wider perspective on such topics, namely that of existential closure as studied in model theory.
There is a beautiful exposition of points at infinity, projective closure, compactifications, modifications, etc. in [FM][1] Chapter 7, Points at Infinity, by H. Behnke and H. Grauert. This is volume III in the excellent "Fundamentals of Mathematics" series, which deserves to be on the bookshelf of every budding mathematician.
A much deeper appreciation of the methodology behind these constructions can be had by studying them from a model-theoretic perspective, in particular from the standpoint of existential closure and model completion. Kenneth Manders has written a series of thought provoking papers [2],[3] from this perspective. I close with an excerpt from the introduction to [2]:
"The systematic adjunction of roots, or solutions to other simple conditions, as in formation of the complex numbers by adjoining imaginaries, or in adjunction of points "at infinity" in traditional geometry, may be analysed as existential closure and model completion. 'Existential closure' refers to a class of processes which attempt to round off a domain and simplify its theory by adjoining elements -- more properly, it refers to the formal relationship that obtains in such a process. 'Model completion' is one of the terms employed when this process is successful. The formation of the complex numbers, and the move from affine to projective geometry, are successes of this kind. Thus, the theory of existential closure gives a theoretical basis of Hilbert's "method of ideal elements." I now sketch the theory of existential closure, to bring out when, how, and in what sense existential closure gives conceptual simplification."
[FM] Fundamentals of mathematics. Vol. III. Analysis.
Edited by H. Behnke, F. Bachmann, K. Fladt and W. Suss.
Translated from the second German edition by S. H. Gould.
Reprint of the 1974 edition. MIT Press,
Cambridge, Mass.-London, 1983. xiii+541 pp. ISBN: 0-262-52095-8 00A05
[2] Manders, Kenneth
Domain extension and the philosophy of mathematics.
J. Philos. 86 (1989), no. 10, 553--562.
http://www.jstor.org/stable/2026666
[3] Manders, Kenneth L.
Logic and conceptual relationships in mathematics.
Logic colloquium '85 (Orsay, 1985), 193--211,
Stud. Logic Found. Math., 122,
North-Holland, Amsterdam-New York, 1987.
http://dx.doi.org/10.1016/S0049-237X(09)70554-3
It is important to consider the context in which you are taking the limit. When taking the limit of a sequence $\{ x_{n}\}$, we must consider what set the elements of that sequence are coming from. For example, if $\{ x_{n}\} \subset \mathbb{R}$, then for the limit $L$ to exist (in $\mathbb{R}$) we must have that $L \in \mathbb{R}$. If $\{ x_{n}\} \to \infty$ as $n \to \infty$, then the limit doesn't exist (in $\mathbb{R}$).
But, there is another notion of limit called $\text{lim sup}$. According to Baby Rudin (slightly modified):
3.16 Definition Let $\{s_{n}\}$ be a sequence of real numbers. Let $E$ be the set of numbers $x$ (in the extended real number system [which includes $\pm \infty$]) such that $s_{n_{k}} \to x$ for some subsequence $\{s_{n_k}\}$. This set $E$ contains all subsequential limits* plus possibly the numbers $+ \infty$, $-\infty$.
Now define $$\begin{align}\text{lim sup} \;\; s_{n} &= \sup E, \;\;\;\;\text{and} \\\ \text{lim inf} \;\; s_{n} &= \inf E.\end{align}$$ * A subsequential limit is just the limit of the subsequence, if that subsequence converges.
With this definition we can discuss infinite limits (superior or inferior).
EDIT: Removed my erroneous claim that finite $\text{lim sup}$'s and $\text{lim inf}$'s collapse to the regular old $\lim$. (Counterexample: $\text{lim sup}_{x \to \infty} \sin(x) = 1$ but $\lim _{x \to \infty}\sin(x) \neq 1$.
While I'll defer to other's answers (e.g. Bill Dubuque's) that in the most general sense "it depends" is the correct answer, it may be worth noting that in the context of AP Calculus and across numerous common AP Calculus texts (at both the AB and BC level), a limit exists if and only if it has a real value, so saying that $\underset{x\to c}\lim f(c)=\infty$ or $\underset{x\to c}\lim f(c)=-\infty$ is saying that the limit as $x$ goes to $c$ of $f(x)$ does not exist and simultaneously giving more information about how the limit does not exist.
The language use in my calculus class was that this is an improper limit. So that strictly speaking there is no limit in the proper sense.
The criterion with right hand limits and left hand limits works at points really belonging to the real line (and then first of all both one-sided limits have to exist for themselves).
There is no ambiguity about infinite limits. For instance, when we define that a sequence or real numbers $(x_n)$ "tends to infinity", $(x_n) \longrightarrow +\infty$, we just mean that, for each real number $N$ there is some $n_0 \in \mathbb{N}$ such that, for all $n\geq n_0$, we have $x_n > N$. That is, the sequence $(x_n)$ grows indifinitely. There is no implication here that $+\infty$ "exists", or doesn't, nor any problem with the fact that, certainly, $+\infty$ is not a real number. In fact, there is no $\infty$ on the right hand side of the definition, but just real numbers. And symbols on the left hand side of a definition mean just you want them to mean according with what you put on the right hand side.
It looks that infinity is a bone of contention among mathematicians. In my view the notation + or - infinity is very useful to highlight the behavior of the graph of a function as the variable x tends to a finite value, remarking the function values are going to very large values impossible to bound with the biggest one. One solution is to accept there are two kind of limits in Mathematics: a finite limit, represented by a real number if it exists, and an infinite limit whose quantity can not be numbered as usually we do when finite limits, and to be happy that + or - infinity is applicable in reliable situations to sort or identify a value incommensurable large. Also I think that philosophic convictions and religious standings are playing strongly in this analysis. There will be light at the end of the tunnel! sandor
It depends on what you are working with. For example, one definition of the derivative of a function that the limit: $\lim_{h \rightarrow 0} \frac{f(x+h)-f(x)}{h}$ exists and it is finite. On the other hand, if you're talking about the limit of a sequence ${ x_n }$, then you may want to distinguish between cases when the limit is infinite, or it doesn't exist, which means, like you said, that taken two subsequences you get different values for the limit -in this case you say that the sequence oscillates-.
Definitions, of course, solve a lot of problems, you define in $\mathbb{R}$ $\lim_{x \rightarrow \infty} f(x) = \infty$ as follows: $\forall N >0, \; \exists M >0: \; x > M \Rightarrow f(x) > N$, then you don't really have to call in the extended number line.
