Weighing neutrinos with 21cm intensity mapping at the SKAO

Abstract

We explore the constraining power of future 21cm intensity mapping (IM) observations at the SKAO, focusing primarily on the sum of neutrino masses, Σmν . We forecast observations of the 21cm IM auto-power spectrum as well as the 21cm IM and galaxy surveys cross-correlation power spectrum. We construct different synthetic data sets of observations for the 21cm IM observables in the redshift range 0 < z < 3. For galaxy clustering, we consider two stage-IV surveys to mimic a DESI-like and Euclid-like cross-correlation signal. We study the impact of assuming three different fiducial values for the sum of neutrino masses, i.e. Σmν = 0.06, 0.1, 0.4 eV, in the synthetic data sets. To investigate the constraining power of the forecasted 21cm observations, we build a likelihood code that will be made publicly available upon publication. The results of the analysis, obtained through Markov Chain Monte Carlo techniques, are promising. We find that the 21cm auto-power spectrum alone could provide an upper limit on the sum of neutrino masses of Σmν < 0.287 eV, at 95% confidence level, for the case of the lowest fiducial value of Σmν . This result is comparable to the upper limits provided by cosmic microwave background (CMB) observations alone. When combining the 21cm auto-power spectrum synthetic data set with Planck 2018 CMB measurements, we find a tighter upper limit of Σmν < 0.105 eV, which improves on the constraints from Planck alone. We obtain a similar result already at the level of 21cm and galaxy clustering cross-correlation power spectrum, whose detection is more easily achieved as they are less affected by systematic effects. Combining synthetic data sets with Planck 2018 data, we find the upper limits of Σmν < 0.116 eV and Σmν < 0.117 eV for the 21cm signal in cross-correlation with the DESI-like and Euclid-like surveys, respectively. These constraints are comparable to those obtained by combining Planck data with the 21cm auto-power spectrum synthetic data sets, thus supporting the case for 21cm cross-correlation detections.