1. Limited time only! Sign up for a free 30min personal tutor trial with Chegg Tutors
    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!

Homework Help: Difference equation

  1. Feb 14, 2008 #1
    1. The problem statement, all variables and given/known data
    Find a closed-form expression for [tex]c_n[/tex].

    [tex]c_{n+1} = \frac{c_0 (3c_n + c_0)}{2c_n + c_0}[/tex]

    2. Relevant equations

    3. The attempt at a solution
    Besides finding [tex]c_1, c_2, c_3, \ldots[/tex] and looking for a pattern, I have absolutely no idea.
  2. jcsd
  3. Feb 14, 2008 #2
    Every recurrence equation of the form
    [tex]c_{n+1}=\frac{\kappa\,c_n+\lambda}{\mu\,c_n+\nu},\quad \kappa\,\nu-\mu\,\lambda\neq0[/tex]
    can be brought into the linear form [itex]x_{n+1}=x_n+b_n[/itex] with [itex]b_n[/itex] known.
    The transformation that does that, is
    Plug this to your equation and choose [itex]\alpha,\beta[/itex] in order to arrive to [itex]x_{n+1}=x_n+b_n[/itex]
  4. Feb 14, 2008 #3
    How do I get that to simplify into the required form? I tried plugging it in like this:

    [tex]\frac{ \alpha ^{-n-1} }{x_{n+1} }+ \beta =\left( \frac{1}{x_{0} }+ \beta \right)\left[ \frac{3\left( \frac{ \alpha ^{-n} }{x_{n} }+ \beta \right)+\frac{1}{x_{0} }+ \beta }{2\left( \frac{ \alpha ^{-n} }{x_{n} }+ \beta \right)+\frac{1}{x_{0} }+ \beta } \right][/tex]

    but I couldn't figure out how to simplify it without being left with a [tex]x_n x_{n+1} [/tex] term.
  5. Feb 15, 2008 #4
    You don't change [itex]c_0[/itex] into [itex]\frac{1}{x_0}+\beta[/itex] because it is a constant. Just write

    [tex]\frac{ \alpha ^{-n-1} }{x_{n+1} }+ \beta =c_0 \frac{3\left( \frac{ \alpha ^{-n} }{x_{n} }+ \beta \right)+c_0}{2\left( \frac{ \alpha ^{-n} }{x_{n} }+ \beta \right)+c_0}[/tex]

    and calculate [itex]\alpha,\beta[/itex]. No term [itex]x_n\,x_{n+1}[/itex] must survive.
  6. Feb 15, 2008 #5
    Here's what I have so far:

    After making the substitution you suggested, I solved for [tex]\beta[/tex] by noticing that in order for the [tex]x_n x_{n+1}[/tex] term to disappear, either [tex]\alpha=0[/tex] or [tex]c_0^2 + 2c_0\beta -2\beta^2 = 0[/tex]. Solving, I got [tex]\beta=\pm \frac{c_0}{2} (1+\sqrt{3}).[/tex] For convenience, I only used [tex]\beta = \frac{c_0}{2} (1+\sqrt{3}).[/tex] Am I allowed to do that?

    So, substituting in, I got
    [tex]x_{n+1} = \left ( \frac{2+\sqrt{3}}{2-\sqrt{3}} \right ) \frac{x_n}{\alpha} + \frac{2}{c_0\alpha^{n+1} (2-\sqrt{3})}.[/tex]​

    This is where I get "stuck". I figured out a way to solve for [tex]x_n[/tex] that's analagous to finding an integrating factor for a linear ODE, but it involves really nasty algebra. Is there a better way?
    Last edited: Feb 15, 2008
  7. Feb 15, 2008 #6
    You are almost there! :smile:
    Choose for
    in order to eliminate the coefficient of [itex]x_n[/itex]. Can you solve the resulting equation?
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook