How does dynamic casimir effect generate correlated photons? There is a recent paper on arxiv receiving lot of acclaim http://arxiv.org/abs/1105.4714
The authors experimentally show that moving a mirror of a cavity at high speeds produces light from high vacuum. The usual doubts about the experimental techniques seem to be very clearly addressed and reviewed (as per Fred Capasso's comments) http://www.nature.com/news/2011/110603/full/news.2011.346.html)
My question is:
Can someone explain how correlated/squeezed photons are generated in this process? I can get a feel for how a moving mirror can generate real photons by imparting energy to the vacuum (correct me if this is not consistent with the detailed theory). But, I don't see how photons are generated in pairs. Could someone describe the parametric process happening here? 
 A: Ref [19] in the arXiv paper, C. M. Caves and B. L. Schumaker, Phys Rev A 31, 3068 (1985), gives a clean description of a parametric amplifier as the prototype of a two-photon device, at the bottom of its second page:

In a [parametric amplifier], an intense laser beam at frequency
  $2\Omega$ —the pump beam— illuminates a suitable nonlinear medium. The
  nonlinearity couples the pump beam to other modes of the electromagnetic
  field in such a way that a pump photon at frequency $2\Omega$ can be
  annihilated to create "signal" and "idler" photons at frequencies
  $\Omega\pm\epsilon$ and, conversely, signal and idler photons can be
  annihilated to create a pump photon.

One way to think of the present situation would be as a dual of this description. That is, the medium, the Josephson junction, oscillates at a pump frequency $2\Omega$, and interacts nonlinearly with the vacuum state.
From the arXiv paper itself, there is a clear parallel,

Quantum theory allows us to make more detailed predictions than just that
  photons will simply be produced. If the boundary is driven sinusoidally at
  an angular frequency $\omega_d = 2\pi f_d$, then it is predicted that
  photons will be produced in pairs such that their frequencies, $\omega_1$ and
  $\omega_2$, sum to the drive frequency, i.e., we expect $\omega_d = \omega_1 + \omega_2$.

One of the comments on the Nature News page, Edward Schaefer at 2011-06-06 12:39:04 PM, makes a point that I think is worth highlighting, that "The mirror transfers some of its own energy to the virtual photons to make them real."
