Simulation-based inference on warm dark matter from HERA forecasts

The redshifted 21cm signal from Cosmic Dawn promises to open a new window into the early history of our universe and enable the probing of an unprecedented comoving survey volume. In this work, we revisit the imprint of Warm Dark Matter (WDM) on the 21cm signal power spectrum using an updated implementation of the WDM effect in the public code $\texttt{21cmFast}$ and considering a single population of cosmic dawn galaxies. By focusing on inferring the WDM mass, we analyze the degeneracies between the latter and the astrophysics parameters characterizing star formation and X-ray heating and we emphasize the role of the threshold mass for star-forming galaxies, $M_{\rm turn}$. We study the capability of the recently built HERA telescope to reconstruct the WDM mass by adopting the statistical approach of simulation-based inference. We include a comparison of the per-parameter reconstruction quality for different number of simulations used in the training of the algorithm. Our results indicate that HERA could surpass current Lyman-$\alpha$ forest constraints if Cosmic Dawn galaxies exhibit a threshold mass $M_{\rm turn}\lesssim 10^{8}\, M_\odot$. The X-ray source properties considered in this study may also influence the strength of the WDM constraint for lower threshold masses.