Use of Bayesian Optimization to Understand the Structure of Nuclei

TL;DR

This paper introduces a Bayesian optimization method to analyze nuclear reaction data, enabling precise extraction of nuclear state energies even with limited data and high uncertainties.

Contribution

The paper presents a novel Bayesian approach for fitting nuclear reaction data, improving the accuracy of energy level determination in complex experimental conditions.

Findings

Achieved excitation energy measurements with better than 90 keV precision.

Demonstrated effectiveness of Bayesian optimization in low-statistics, high-resolution broadening scenarios.

Provided a pathway for studying poorly known nuclei like $^{13}$Be.

Abstract

Monte Carlo simulations are widely used in nuclear physics to model experimental systems. In cases where there are significant unknown quantities, such as energies of states, an iterative process of simulating and fitting is often required to describe experimental data. We describe a Bayesian approach to fitting experimental data, designed for data from a $^{12}$Be(d,p) reaction measurement, using simulations made with GEANT4. Q-values from the $^{12}$C(d,p) reaction to well-known states in $^{13}$C are compared with simulations using BayesOpt. The energies of the states were not included in the simulation to reproduce the situation for $^{13}$Be where the states are poorly known. Both cases had low statistics and significant resolution broadening owing to large proton energy losses in the solid deuterium target. Excitation energies of the lowest three excited states in $^{13}$C were extracted to better than 90 keV, paving a way for extracting information on $^{13}$Be.