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: Show that transformation is Canonical

  1. May 19, 2012 #1
    1. The problem statement, all variables and given/known data
    Given the transformation [tex]Q=qe^{\gamma t}[/tex] and [tex]P=pe^{-\gamma t}[/tex] with the Hamiltonean [tex]H=\frac{p^2e^{-2\gamma t}}{2m}+\frac{m\omega^2q^2e^{2\gamma t}}{2}[/tex] show that the transformation is Canonical

    2. Relevant equations

    I know that the condition for a transformation to be canonical is [tex]\sum_i p_i\dot{q_i}-P_i\dot{Q_i}=H-K+\frac{dF_2}{dt}[/tex] but don't know what to do with it...

    I've managed to prove from [tex]\frac{dF_2}{dq_i}=p_i[/tex] and [tex]\frac{dF_2}{dP_i}=Q_i[/tex] that [tex]F_2=qPe^{\gamma t}[/tex].
    Last edited: May 19, 2012
  2. jcsd
  3. May 19, 2012 #2
    I have another question related to this problem. Does having the generating function guarantee that the coordinate transformation associated with it is canonical? Or is it a necessary but not sufficient condition?
  4. May 20, 2012 #3


    User Avatar
    Gold Member

    I know a way to check out whether the transformation is canonical but it doesn't involve the Hamiltonian.
    If you know Poisson's brackets, if you can show that [itex][Q,P]_{q,p}=1[/itex] then you'd be done.
    To answer your question "Does having the generating function guarantee that the coordinate transformation associated with it is canonical? Or is it a necessary but not sufficient condition?", I'm not 100% sure but I'd say yes (99% sure).
  5. May 20, 2012 #4
    I've been messing around with the coordinates and got this:

    If we have a general invertible transformation of the type [tex]Q=Q(q,p,t)\Leftrightarrow q=q(Q,P,t)[/tex] and [tex]P=P(q,p,t)\Leftrightarrow p=p(P,Q,t)[/tex] then the following is true:

    \dot{q}=\frac{\partial q}{\partial Q}\dot{Q}+\frac{\partial q}{\partial P}\dot{P}\\
    \dot{p}=\frac{\partial p}{\partial Q}\dot{Q}+\frac{\partial p}{\partial P}\dot{P}

    From Hamilton's eqs we have (inserting the given coordinate transformations):
    \dot{q}=\frac{\partial H}{\partial p}=\frac{\partial H}{\partial P}e^{-\gamma t}=\frac{\partial q}{\partial Q}\dot{Q}\Rightarrow \frac{\partial H}{\partial P}=\dot{Q}\\
    \dot{p}=-\frac{\partial H}{\partial q}= -\frac{\partial H}{\partial Q}e^{\gamma t}=\frac{\partial p}{\partial P}\dot{P}\Rightarrow -\frac{\partial H}{\partial Q}=\dot{P}

    The form of Hamilton's equations is the same but from what I've seen in Goldstein, when one makes a canonical transformation the form is preserved not for the initial Hamiltonian but for the "Kamiltonian" [tex]K=H+\frac{dF_2}{dt}[/tex] i.e.

    \frac{\partial K}{\partial P}=\dot{Q}\\
    -\frac{\partial K}{\partial Q}=\dot{P}

    and because I've already proved that there exists a function [itex]F_2[/itex] I get the feeling that the proof is not correct. What am I doing wrong here?

    Thanks for the advice! Actually, calculating with the Poisson brackets is another part of this problem, but instead it is asked to evaluate [itex][Q,Q],[P,P],[Q,P],[q,H],[p,H][/itex] without explicit calculation.

    For the first two the answer is obviously zero because of the crossed derivatives, the third one, being this a suposedly canonical transformation is 1 and for the last two I'm having a little bit of a problem because if I use [itex]\frac{du}{dt}=[u,H]+\frac{\partial u}{\partial t}[/itex] on q and p I am not sure if it is [itex]\frac{dq}{dt}[/itex] or [itex]\frac{\partial q}{\partial t}[/itex] (for p too) that is zero. I know, this is a dumb question, but when a function depends only on t does the partial derivative looses its meaning?
  6. May 20, 2012 #5


    User Avatar
    Gold Member

    Ok, since you found out [itex]F_2[/itex] then I think you are done. Let me give you a useful link: http://books.google.com.ar/books?id...a transformation is canonical poisson&f=false.
    About the Poisson's bracket question, you are right for the 3 first. As for the last 2 I must say I'm not sure; I'd consult Landau's book on classical mechanics first and then other books.
    About your question about partial derivatives; when you have a function of only 1 variable a partial derivative is the same as a total derivative.
    But I doubt q only depends on t.
  7. May 20, 2012 #6
    Thanks for the book! I tried to see if there was anything in Landau that could help but didn't find it yet. Anyhow, I found this http://solar.physics.montana.edu/dana/ph411/p_brack.pdf and I can't figure out why, in page 2 after equation (3) [itex]\frac{\partial q_1}{\partial t}=0[/itex]. Isn't [itex]q_1[/itex] also a function of t? Why isn't [itex]\frac{dq_1}{dt}[/itex] also zero?
  8. May 20, 2012 #7


    User Avatar
    Gold Member

    I'm in the same boat as you. I do not understand why it is so.
    For sure, [itex]q_1[/itex] isn't a constant of motion, otherwise we'd have [itex]\{ q_1 , H \}=0[/itex]. So there's indeed a dependency of [itex]q_1[/itex] with respect to time. Thus, I do not understand how it's possible that [itex]\frac{\partial q _1 }{\partial t } =0[/itex].
    I wish someone could enlighten us!
  9. May 20, 2012 #8
    I'm just posting this so that the thread doesn't seem closed.
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook