Cosmology of atomic dark matter

While, to ensure successful cosmology, dark matter (DM) must kinematically decouple from the standard model plasma very early in the history of the Universe, it can remain coupled to a bath of ``dark radiation'' until a relatively late epoch. One minimal theory that realizes such a scenario is the atomic dark matter model, in which two fermions oppositely charged under a new U(1) dark force are initially coupled to a thermal bath of ``dark photons'' but eventually recombine into neutral atomlike bound states and begin forming gravitationally bound structures. As dark atoms have (dark) atom-sized geometric cross sections, this model also provides an example of self-interacting DM with a velocity-dependent cross section. Delayed kinetic decoupling in this scenario predicts novel DM properties on small scales but retains the success of cold DM on larger scales. We calculate the atomic physics necessary to capture the thermal history of this dark sector and show significant improvements over the standard atomic hydrogen calculation are needed. We solve the Boltzmann equations that govern the evolution of cosmological fluctuations in this model and find in detail the impact of the atomic DM scenario on the matter power spectrum and the cosmic microwave background (CMB). This scenario imprints a new length scale, the dark-acoustic-oscillation scale, on the matter density field. This dark-acoustic-oscillation scale shapes the small-scale matter power spectrum and determines the minimal DM halo mass at late times, which may be many orders of magnitude larger than in a typical weakly interacting-massive-particle scenario. This model necessarily includes an extra dark radiation component, which may be favored by current CMB experiments, and we quantify CMB signatures that distinguish an atomic DM scenario from a standard $\ensuremath{\Lambda}\mathrm{CDM}$ model containing extra free-streaming particles. We finally discuss the impacts of atomic DM on galactic dynamics and show that these provide the strongest constraints on the model.