Freezing In, Heating Up, and Freezing Out: Predictive Nonthermal Dark Matter and Low-Mass Direct Detection

Freeze-in dark matter (DM) mediated by a light ($\ll$ keV) weakly-coupled dark-photon is an important benchmark for the emerging low-mass direct detection program. Since this is one of the only predictive, detectable freeze-in models, we investigate how robustly such testability extends to other scenarios. For concreteness, we perform a detailed study of models in which DM couples to a light scalar mediator and acquires a freeze-in abundance through Higgs-mediator mixing. Unlike dark-photons, whose thermal properties weaken stellar cooling bounds, the scalar coupling to Standard Model (SM) particles is subject to strong astrophysical constraints, which severely limit the fraction of DM that can be produced via freeze-in. While it seems naively possible to compensate for this reduction by increasing the mediator-DM coupling, sufficiently large values eventually thermalize the dark sector with itself and yield efficient DM annihilation to mediators, which depletes the freeze-in population; only a small window of DM candidate masses near the $\sim$ GeV scale can accommodate the total observed abundance. Since many qualitatively similar issues arise for other light mediators, we find it generically difficult to realize a viable freeze-in scenario in which production arises only from renormalizable interactions with SM particles. We also comment on several model variations that may evade these conclusions.