Constraints on the fuzzy dark matter mass window from high-redshift observables

We use a combination of high-redshift observables to extract the strongest constraints to date on the fraction of axion fuzzy dark matter (FDM) in the mass window ${10}^{\ensuremath{-}26}\text{ }\text{ }\mathrm{eV}\ensuremath{\lesssim}{m}_{\mathrm{FDM}}\ensuremath{\lesssim}{10}^{\ensuremath{-}23}\text{ }\text{ }\mathrm{eV}$. These observables include ultraviolet luminosity functions at redshifts 4--10 measured by the Hubble Space Telescope, a constraint on the neutral hydrogen fraction from high-redshift quasar spectroscopy, the cosmic microwave background optical depth to reionization measurement from Planck, and upper bounds on the 21 cm power spectrum from HERA. In order to calculate these signals for FDM cosmology, we use the 21 cmfirstclass code to interface between axiclass and 21 cmfast and consistently account for the full cosmic history from recombination to reionization. To facilitate a full Bayesian likelihood analysis, we developed a machine-learning-based pipeline, which is both accurate and enables a swift statistical inference, orders of magnitude faster than a brute-force approach. We find that FDM of mass ${m}_{\mathrm{FDM}}={10}^{\ensuremath{-}23}\text{ }\text{ }\mathrm{eV}$ is bound to less than 16% of the total dark matter, where the constrains strengthen toward smaller masses, reaching down to 1% for ${m}_{\mathrm{FDM}}={10}^{\ensuremath{-}26}\text{ }\text{ }\mathrm{eV}$, both at 95% confidence level. In addition, we forecast that a future detection of the 21 cm power spectrum with HERA will lower the upper bound at ${m}_{\mathrm{FDM}}={10}^{\ensuremath{-}23}\text{ }\text{ }\mathrm{eV}$ to $\ensuremath{\lesssim}1%$.