Measuring neutrino masses with a future galaxy survey

TL;DR

This paper forecasts how a future Euclid-like galaxy survey combined with CMB data can detect the total neutrino mass with high significance, breaking degeneracies between key cosmological parameters.

Contribution

It provides a detailed forecast of neutrino mass constraints from a Euclid-like survey, highlighting the importance of combined galaxy and cosmic shear measurements.

Findings

Potential to detect neutrino mass sum ~0.06 eV at 1.5-2.5 sigma significance.

Combined galaxy and shear data break degeneracies between neutrino mass, matter density, and Hubble parameter.

Neutrino mass spectrum details are not sensitive with Euclid-like data.

Abstract

We perform a detailed forecast on how well a Euclid-like photometric galaxy and cosmic shear survey will be able to constrain the absolute neutrino mass scale. Adopting conservative assumptions about the survey specifications and assuming complete ignorance of the galaxy bias, we estimate that the minimum mass sum of sum m_nu ~ 0.06 eV in the normal hierarchy can be detected at 1.5 sigma to 2.5 sigma significance, depending on the model complexity, using a combination of galaxy and cosmic shear power spectrum measurements in conjunction with CMB temperature and polarisation observations from Planck. With better knowledge of the galaxy bias, the significance of the detection could potentially reach 5.4 sigma. Interestingly, neither Planck+shear nor Planck+galaxy alone can achieve this level of sensitivity; it is the combined effect of galaxy and cosmic shear power spectrum measurements that breaks the persistent degeneracies between the neutrino mass, the physical matter density, and the Hubble parameter. Notwithstanding this remarkable sensitivity to sum m_nu, Euclid-like shear and galaxy data will not be sensitive to the exact mass spectrum of the neutrino sector; no significant bias (< 1 sigma) in the parameter estimation is induced by fitting inaccurate models of the neutrino mass splittings to the mock data, nor does the goodness-of-fit of these models suffer any significant degradation relative to the true one (Delta chi_eff ^2< 1).