Decision Theory for the Mass Measurements at the Facility for Rare Isotope Beams

Abstract

Nuclear physics facilities, like the Facility for Rare Isotope Beams (FRIB), can potentially perform many nuclear mass measurements of exotic isotopes. Each measurement comes with a particular cost, both in time and money, and thus it is important to establish which mass measurements are the most informative. In this article, we show that one can use the Kullback-Leibler divergence to determine the information gained by a mass measurement. We model the information gain obtained by nuclear mass measurements from two perspectives: first from the perspective of theoretical nuclear mass models, and the second from the perspective of r-process nucleosynthesis. While this work specifically analyzes the abilities of FRIB, other facilities worldwide could benefit from a similar use of information gain in order to decide which experiments are optimal.

Model reduction methods for nuclear emulators

“…where φ is an arbitrary test function. It can be shown that if equation ( 12) holds for all φ, then equation (2) must hold as well: if (H − E)|ψ has any non-zero elements, then one can immediately find a φ such that equation (12) does not hold.…”

“…To obtain a reduced-order model from equation (12) we would ideally like to map ψ → ψ and find the β that satisfy equation (12) for all φ, hence ensuring the satisfaction of equation (2). Unfortunately this would result in an over-determined system because an arbitrary φ has many more degrees of freedom than ψ. Alternatively, we can derive the Ritz subspace method by imposing that the error made by the trial eigenvector is orthogonal to the N b -dimensional subspace X spanned by X:…”

“…Nuclear physics calculations often need to be repeated many times for different values of some model parameters, for example when sampling the model space for Bayesian uncertainty quantification [1][2][3][4][5][6][7][8][9] and experimental design [10][11][12]. The computational burden can be alleviated by using emulators, or surrogate models, which accurately approximate the response of the original (i.e., high-fidelity) model but are much cheaper to evaluate.…”

Model reduction methods for nuclear emulators

“…Nuclear physics calculations often need to be repeated many times for different values of some model parameters, for example when sampling the model space for Bayesian uncertainty quantification [1][2][3][4][5][6][7][8][9] and experimental design [10][11][12]. The computational burden can be alleviated by using emulators, or surrogate models, which accurately approximate the response of the original (i.e., high-fidelity) model but are much cheaper to evaluate.…”