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

Can anyone help with eigenfunction?

  1. May 31, 2010 #1
    if y''-2y'+y=ky k=eigenvalue, y(0,pi)=0, 0<x<pi
    find corresponding eigenvalues and eigenfunctions.

    thx a lot!
  2. jcsd
  3. May 31, 2010 #2
    How would you solve the ode


    Regarding k as a parameter for the moment
  4. May 31, 2010 #3
    I got this

    sqrt{k} = m
    A(1-m)2 -2A(1-m)+A(1-k)=0
    B(1+m)2 -2B(1+m)+B(1-k)=0

    and what then !??

    Please help! thank you very much!
  5. May 31, 2010 #4
    can anyone help!?
  6. May 31, 2010 #5
    It's a 2nd order linear ODE with constant coefficients.
    Trying the solution y=e^mx (as I think you've done) gets your solutions (unless...etc, look it up).
    Solve for m, then you have one function that's an eigenfunction, with eigenvalue corresponding to m.
  7. Jun 1, 2010 #6


    User Avatar
    Science Advisor

    The characteristic equation for y"- 2y'+ (1- k)y= 0 is [itex]r^2- 2r+ 1-k= 0[/itex] which is the same as [itex]r^2- 2r+ 1= (r- 1)^2= k[/itex] and has roots [itex]r= 1\pm \sqrt{k}= 1\pm m[/itex] with your choice of m as [itex]\sqrt{k}[/itex].

    The general solution is [itex]y= Ae^{(1+m)t}+ Be^{(1-m)t}[/itex]
    Setting that equal to 0 at x= 0 and [itex]\pi[/itex], we find that as long as 1+ m and 1- m are real, A and B must be 0.

    In order for k to be an eigenvalue, k will have to be negative so that m is imaginary. Given that, and writing m= ni, [itex]y= Ae^{(1+ni)t}+ Be^{(1-nit)}= e^t(Ae^{nit}+ Be^{-nit})[/itex]. We can write that as [itex]y= e^t(C cos(nt)+ D sin(nt))[/itex].

    Now we have [itex]y(0)= e^0(Ccos(0)+ D sin(0))= C= 0[/itex] and [itex]y(\pi)= e^{\pi}(Ccos(n\pi)+ Bsin(n\pi)= Be^{\pi}sin(n\pi)[/itex] (because C= 0) and that must be equal to 0. That will be true either for B= 0 or for [itex]sin(n\pi)= 0[/itex] which will be the case as long as n is an integer.

    Can you find the eigenvalues and eigenvectors from there?
  8. Jun 3, 2010 #7

    k have to be negative and n have to be integer, do u mean k=0 ?!
    Please answer !! Thanks !!
  9. Jun 3, 2010 #8

    Therefore, any k that is a square of an integer (with a minus in front) is an eigenvalue to this problem.
Know someone interested in this topic? Share this thread via Reddit, Google+, Twitter, or Facebook