Forecasting neutrino masses from combining KATRIN and the CMB: Frequentist and Bayesian analyses

TL;DR

This paper compares Bayesian and frequentist bounds on neutrino masses from KATRIN and cosmological data, highlighting the importance of combining different data sources for improved detection sensitivity.

Contribution

It provides a detailed comparison of Bayesian and frequentist analyses for neutrino mass bounds and demonstrates the benefits of combining laboratory and cosmological data.

Findings

Bayesian upper bound on m_beta is 0.17 eV at 90% confidence.

Frequentist bound on m_beta is 0.20 eV.

Combining KATRIN and Planck data increases detection significance.

Abstract

We present a showcase for deriving bounds on the neutrino masses from laboratory experiments and cosmological observations. We compare the frequentist and Bayesian bounds on the effective electron neutrino mass m_beta which the KATRIN neutrino mass experiment is expected to obtain, using both an analytical likelihood function and Monte Carlo simulations of KATRIN. Assuming a uniform prior in m_beta, we find that a null result yields an upper bound of about 0.17 eV at 90% confidence in the Bayesian analysis, to be compared with the frequentist KATRIN reference value of 0.20 eV. This is a significant difference when judged relative to the systematic and statistical uncertainties of the experiment. On the other hand, an input m_beta=0.35 eV, which is the KATRIN 5sigma detection threshold, would be detected at virtually the same level. Finally, we combine the simulated KATRIN results with cosmological data in the form of present (post-WMAP) and future (simulated Planck) observations. If an input of m_beta=0.2 eV is assumed in our simulations, KATRIN alone excludes a zero neutrino mass at 2.2sigma. Adding Planck data increases the probability of detection to a median 2.7sigma. The analysis highlights the importance of combining cosmological and laboratory data on an equal footing.