Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Existence of limit question

  1. Aug 28, 2013 #1
    Hi all,

    I have a quick question about limits. This is something I should know but shame on me I forgot.

    If a function is bounded both above and below but isn't monotonic and is not necessarily continuous at all points, how do I go about proving its limit exists?

    In particular I am thinking about something along the lines of this function:
    where $$F_{n}$$ is the nth Fibonacci number. I know that the limit exists and I know what that limit is but say I didn't. Say I wanted to prove that it did exist. How do I do that without using the closed form expression for Fibonacci numbers? I know that it is bounded but it isn't monotonic.

    This wasn't a homework question but a question that came to my mind after proving that this limit was equal to the golden ratio using the closed form expression of the Fibonacci numbers.
  2. jcsd
  3. Aug 28, 2013 #2
    Your question is stated in terms of functions, but your example problem is stated in terms of sequences. I'll answer for sequences.

    Obviously there isn't one single method that can work in general, because the limit of such a sequence doesn't always exist.

    Here are a few options to try:

    1. If the differences of a sequence are alternating in sign and have limit zero, then the sequence has a limit. This is the easiest method to apply in this case. It doesn't tell you what the limit is, but you should be able to figure it out using some other method (for example, it must satisfy L = 1 + 1/L).

    2. "Squeeze theorem": If you can construct an upper bound and a lower bound for the sequence such that both the upper bound and lower bound have limit L, then the sequence has limit L.

    3. Prove the sequence is Cauchy.

    4. Guess the limit and prove that the sequence approaches it using the definition of limit.
  4. Aug 28, 2013 #3


    User Avatar
    Staff Emeritus
    Science Advisor
    Gold Member

    You know that Fn+1 = Fn+Fn-1, so if the limit exists:

    [tex] \lim_{n\to \infty} F_{n+1}/F_n = \lim_{n\to \infty} (F_n+F_{n-1})/F_n = \lim_{n\to \infty} 1 + F_{n-1}/F_n [/tex]

    If [itex] \lim_{n\to \infty} F_{n+1}/F_n = \alpha[/itex] then the rightmost guy above is equal to [itex] 1 + 1/\alpha[/itex] so if the limit exists we get
    [tex] \alpha = 1+1/\alpha[/tex]

    From which you can solve for what the limit is equal to. Ironically calculating the limit is much more difficult than showing the limit exists. In this case if I remember correctly for
    [tex] a_n = F_{n+1}/F_n[/tex]
    the a2n form a monotonic sequence and the a2n+1 form a monotonic sequence. The argument above with some tweaking shows that the numbers they converge to are the same, and therefore the sequence [itex] a_n[/itex] converges.
  5. Aug 28, 2013 #4
    Yes that is exactly what I wanted to do. I have a proof that the limit is equal to the golden ratio using the closed form expression but I had to do a lot of footwork to show that the Fibonacci numbers could in fact be expressed as that closed form (I did this using induction).

    I was also able to show the limit was equal to the golden ratio by using the assumption that it was equal to a. What I had forgotten was how to show functions/sequences converged if I only had boundedness. Thank you and eigenperson for reminding me.

    I guess it was a bad idea to leave home without Spivak...
  6. Aug 29, 2013 #5


    User Avatar
    Science Advisor
    Homework Helper

    have you thought about sin(x) as x --> infinity?
  7. Aug 30, 2013 #6
    That is definitely an interesting one. I have always found limits of trig functions to be kind of funky. I know that the limit doesn't exist. The only proof I know of to prove that is the epsilon delta proof that derives a contradiction.
  8. Aug 31, 2013 #7
    Is that true? What about the sequence [tex]S_n = (S_{n-1}+\frac{4}{n} \text{ where n is even, } S_{n-1} - \frac{2}{n-1} \text{ where n is odd}), S_0 = S_1 = 0 [/tex]Note that this generates the series ## \frac{4}{2} - \frac{2}{2} + \frac{4}{4} - \frac{2}{4} + \frac{4}{6} - \frac{2}{6} \ldots ##
  9. Aug 31, 2013 #8
    Yeah, I should have said "decreasing" instead of "has limit zero".
  10. Aug 31, 2013 #9
    Leibniz Criterion
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook