Quantified Nuclear Mass Model for Nuclear Astrophysics Simulations*

Nuclear mass is a fundamental property of the nucleus. Nuclear binding energies determine the particle drip lines where the nuclear landscape ends as well as the Q-values of nuclear reactions. As a result, they are a key ingredient in astrophysical models and simulations. We are, however, limited in our ability to experimentally measure the complete mass table and have to rely on theory to carry out extrapolations. Since global mass models are fitted to known experimental data and then used to predict masses over the whole nuclear chart, assessing their uncertainties in regions far from stability is a non-trivial task. This makes it difficult to quantify the impact of the uncertainties of nuclear masses in astrophysical simulations involving exotic nuclei. To overcome this limitation, we present a model-based mass extrapolation technique where we model the binding energy residuals with fully Bayesian Gaussian Process Regression. The statistically corrected predictions obtained from 11 different global nuclear models are then combined via Bayesian model averaging, according to their experimental evidence. This leads us to a quantified mass model providing uncertainties and covariances that can be used for astrophysical modeling and sensitivity studies.