I'm having some trouble with what I think should be an easy calculation. On qutip.org/docs/2.2.0/examples/me/ex-25.html they calculate the steady state for a master equation, more specifically they take an example from nanomechanics: they c a low-frequency nanomechanical resonator (with frequency $\omega_m$ lower than temperature) coupled to a high-frequency (optical) resonator (with frequency $\omega_r$ higher than temperature).

By applying a driving field to the high-frequency resonator with a frequency matching the frequency difference between the resonator ($\omega_d = \omega_r - \omega_m$), the two oscilators can effectively be brought into resonance in a rotating frame, allowing for excitation transfer from the low-frequency mechanical resonator to the high-frequency resonator (i.e., sideband cooling of the mechanical resonator).

They then go on to write that the Hamiltonian of this system is given by

$H = \omega_r a^\dagger a + \omega_m b^\dagger b + g a^\dagger a (b + b^\dagger) + A \sin(\omega_d t) (a + a^\dagger)$

which I fully accept. However, they then write that using the rotating wave approximation this can be rewritten as

$H = (\omega_r-\omega_d) a^\dagger a + \omega_m b^\dagger b + g a^\dagger a (b + b^\dagger) + \frac{1}{2} A (a + a^\dagger)$

This I do not see at all. The rotating wave approximation I know I have seen in the Jaynes Cummings model, where there are some phase factors that you recombine in a clever way such that some of them are much faster than the others, and you subsequently drop them. However, I see no such phase factors in the original question here, and I am honestly not sure what kind of steps are being taken in the above. Any help is much appreciated!

Some specific confusion comes from the fact that we get a term $\omega_r - \omega_d$, but no $\omega_m - \omega_d$.


The rotating-wave approximation requires a transformation to a rotating frame. In particular, use the transformation $U(t) = {\rm e}^{{\rm i} \omega_d t a^\dagger a}$, i.e. transform your states as $\lvert \psi(t)\rangle \to \lvert \tilde{\psi}(t)\rangle = U(t) \lvert \psi(t)\rangle $. Now, in the original frame, the evolution is given by the Schroedinger equation ($\hbar = 1$) $$ {\rm i}\lvert \dot{\psi}(t)\rangle = H\lvert \psi(t)\rangle. $$ As an easy exercise, show that in the rotating frame the Schroedinger equation reads as $${\rm i}\lvert \dot{\tilde{\psi}}(t)\rangle = \tilde{H}(t)\lvert \tilde{\psi}(t)\rangle, $$ where $$\tilde{H}(t) = U(t) H U^\dagger(t) - \omega_d a^\dagger a. $$ You should now be able to prove the desired result using the rotating-wave approximation, which consists of neglecting "counter-rotating" terms oscillating at frequency $2\omega_d$.

  • $\begingroup$ That is what I had in mind indeed. But what that does not tell me is why $w_r$ is special, compared to $\omega_m$, in the sense that we get the $\omega_r-\omega_d$ term but not the $\omega_m - \omega_d$ term. Is there some implicit assumption regarding the low/high frequency part of the story? $\endgroup$ – user129412 May 17 '16 at 17:19
  • $\begingroup$ There is nothing special about any of the frequencies. This transformation is just one possible choice out of (infinitely) many. It was apparently chosen to remove the tricky $\sin(\omega_d t)$ term, leading to an approximately time-independent Hamiltonian. This driving term only couples to the $a$ oscillator, and not the $b$ one. It is easiest to see why it comes out this way if you just work through it yourself. If you want a $\omega_m - \omega_d$ term, use a transformation like ${\rm e}^{{\rm i}\omega_d b^\dagger b}$. But this will make the radiation pressure term time-dependent. $\endgroup$ – Mark Mitchison May 17 '16 at 17:24
  • $\begingroup$ Well, that is incredibly stupid, somehow I missed the fact that the other one only had b terms. I guess that clarifies the situation. $\endgroup$ – user129412 May 17 '16 at 17:27
  • $\begingroup$ A short follow up. Say I also add a drive to the b part, so that $H = \omega_r a^\dagger a + \omega_m b^\dagger b + g a^\dagger a (b + b^\dagger) + A \sin(\omega_d t) (a + a^\dagger) + + A \sin(\omega_d t) (b + b^\dagger)$. How would this change my transformation? Do I simply use $e^{i \omega_d (a^\dagger a + b^\dagger b)}$? $\endgroup$ – user129412 May 18 '16 at 13:43
  • $\begingroup$ @user129412 you should try it and see. But as I said, the radiation pressure term (coupling between a and b) will make your life difficult since it will not transform nicely. $\endgroup$ – Mark Mitchison May 18 '16 at 16:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.