1. PF Contest - Win "Conquering the Physics GRE" book! Click Here to Enter
    Dismiss Notice
Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Very difficult limit

  1. Sep 15, 2009 #1
    1. The problem statement, all variables and given/known data

    http://www.sosmath.com/CBB/latexrender/pictures/f45d5e5b84cb5aecef3fd2f568739e32.gif [Broken]

    2. Relevant equations


    http://www.sosmath.com/CBB/latexrender/pictures/fee1863dab606636f986e63c261b565f.gif [Broken]
    http://www.sosmath.com/CBB/latexrender/pictures/61d6e1cc712fcb6ac6f19b40851e85be.gif [Broken]
    http://www.sosmath.com/CBB/latexrender/pictures/ae169a7ff6b1d2c09a977e5aad223650.gif [Broken]
    http://www.sosmath.com/CBB/latexrender/pictures/cd14347d968bdb5be67f72e832558702.gif [Broken]

    3. The attempt at a solution
    I don't understand how the hint helps in this limit. Could someone point me in the right direction?
    Last edited by a moderator: May 4, 2017
  2. jcsd
  3. Sep 15, 2009 #2
    I didn't finish it, but here's my possible hint. If L=your limit, consider log L.

    Edited to add: f(x) will be log(1+x). I "feel" like this is right.
  4. Sep 16, 2009 #3
    Yes, that's helps a little bit but what do you mean by f(x) will be 1+x? How is that helpful?
  5. Sep 16, 2009 #4
    It will turn out that you will want to use f(x)=log(1+x).

    Anyway, I did finish it now, and it works out. Use the log of the limit idea as I first suggested and apply properties of logs. At some step you will get terms with log(n+k). Rewrite as log[n(1+k/n)]=log n + log(1+k/n). Eventually you'll have to decide what to use for x, and it's not too hard to guess. Write out what you can, and I bet you will get it. If not, post your work and I'll give you more hints. I predict from time to time you'll say things to yourself like "wow, I can't believe those terms canceled out exactly, this must be right."
  6. Sep 16, 2009 #5
    [tex] \ln(L) = \sum_{k=0}^n \frac{1}{2^n} \left( \stackrel{n}{k} \right) \left( \ln(n) + \ln \left(1+\frac{k}{n} \right) \right) - \ln(n) [/tex]

    Now if [tex]f(x) = \ln(1+x)[/tex] then B_n(1) is what I have, I only need to add an extra (1/2^n)* ln(2) and (subtract it later) to use the hint but for B_n(1)

    [tex] \ \sum_{k=0}^n \left( \stackrel{n}{k} \right) \ln \left(1+\frac{k}{n} \right) - \ln(2) \leq \frac{1}{2 \sqrt{n}} \leq \frac{1}{2}[/tex]

    [tex] \left( \frac{1}{2} \right) ^n \sum_{k=0}^n \left( \stackrel{n}{k} \right) \ln(n) = \ln(n) [/tex]

    so this cancels the last term leaving only that annoying thing from above plus that term which I needed to substract:

    [tex]\left( \frac{1}{2} \right) ^n \left[ \frac{1}{2} + \ln(2) \right] [/tex]

    So the limit is 1? Am I going in the right direction?
    Last edited: Sep 16, 2009
  7. Sep 16, 2009 #6
    Looking good, but B_n(1) would be 0. I used a different value for x.

    (Also, leave [tex]\frac{1}{2\sqrt{n}}[/tex] as is. Don't replace it by 1/2.)
  8. Sep 16, 2009 #7
    x=0? then

    \ \sum_{k=0}^n \left( \stackrel{n}{k} \right) \ln \left(1+\frac{k}{n} \right) - \ln(1) \leq \frac{1}{2 \sqrt{n}}

    This results in:

    \left( \frac{1}{2} \right) ^n \left[ \frac{1}{2\sqrt{n}}\right] = \frac{ \frac{1}{2}^{n+1} }{\sqrt{n}} \rightarrow 0

    as [itex] n \rightarrow \infty [/itex]

    Conclusion: the limit is 1.
    Last edited: Sep 16, 2009
  9. Sep 16, 2009 #8

    [tex]B_n(x)=\sum_{k=0}^n \binom{n}{k} x^k (1-x)^{n-k} f(\frac{k}{n})[/tex]

    so [tex]B_n(0)=\sum_{k=0}^n \binom{n}{k} 0^k 1^{n-k} f(\frac{k}{n})[/tex]

    so [tex]B_n(0)=\binom{n}{0} 0^0 f(0)[/tex], which can't be helpful.

    Also, [tex]B_n(1)=\sum_{k=0}^n \binom{n}{k} 1^k 0^{n-k} f(\frac{k}{n})[/tex]

    so [tex]B_n(1)=\sum_{k=0}^n \binom{n}{k} 1^k 0^{n-k} f(\frac{k}{n})[/tex]

    so [tex]B_n(1)=\binom{n}{n} 0^0 f(1)[/tex] which may not be 0 as I claimed earlier, but it still can't be helpful.

    Oh well, average the two.....
  10. Sep 17, 2009 #9
    Oops :biggrin:

    I see, if you take x =1/2 it matches my term exactly:

    B_n \left( \frac{1}{2} \right) =\sum_{k=0}^n \binom{n}{k} \left( \frac{1}{2} \right) ^n f \left(\frac{k}{n} \right)


    [tex] | B_n \left( \frac{1}{2} \right) - f \left(\frac{1}{2} \right)| \leq \frac{1}{2 \sqrt{n} } [/tex]

    \frac{1}{2 \sqrt{n}} + \ln \left( \frac{3}{2} \right) \longrightarrow \ln \left( \frac{3}{2}\ \right) \mbox{as}\ n \rightarrow \infty

    So the limit is 3/2!
    Last edited: Sep 17, 2009
  11. Sep 17, 2009 #10
    Hooray! Isn't that nice!

    I wrote [tex]\ln \left( \frac{3}{2} \right)-\frac{1}{2 \sqrt{n} }\le B_n \left( \frac{1}{2} \right) \le \ln \left( \frac{3}{2} \right)+ \frac{1}{2 \sqrt{n} } [/tex] to make the final steps crystal clear.

    Don't forget to verify [tex]|f(x)-f(y)|\le|x-y|[/tex] so that the hint applies.
  12. Sep 17, 2009 #11
    Yes it is :)

    Right, so the squeeze theorem applies.

    I'm currently thinking on the proof for the hint but CS (after triangle inequality) gives me some nasty squares...
Know someone interested in this topic? Share this thread via Reddit, Google+, Twitter, or Facebook