lost in space, part 2

Could you, instead of using the temperature difference in the CMB anisotropies, use the fact that the photon temperature drops with time for generating mechanical work?

There is a second possibility compared to using the temperature anisotropies, using the momopole: One can “trap” radiation in a cavity to keep it at a given temperature and wait until the Hubble expansion has cooled down the outside CMB. Of course, the cavity shouldn’t participate in the Hubble expansions, otherwise the trapped photons would lower their temperature \(\propto 1/a\). Then, there are two baths of photons, the photons in the cavity and the colder photons outside, which can be used to drive a Carnot engine.