Neural Network Based Fermionic to Bosonic Potential Energy Surface Mapping

Potential energy surfaces can be generated through a variety of nuclear physics models. In this work, we utilize a mixed micro-macro model to generate fermionic potential energy surfaces as a function of the quadrupole and triaxial degrees of freedom for a few isotopic chains. Mapping techniques allow these fermionic potentials to be compared with bosonic potentials such that a best fit of bosonic Hamiltonian parameters is determined. Our choice of bosonic Hamiltonian is the Interacting Boson Model (IBM) 1 Consistent Q formalism. One output of the bosonic potential fit is the experimental low-lying spectra for even-even nuclei allowing for comparisons with experimental results. The innovation discussed in this work is the development and training of a neural network for the bosonic potentials. The trained network is then applied to our fermionic potentials determining the corresponding IBM parameters and spectra. This talk will focus on a comparison of results from the trained neural network mapping with previous mapping results based on purely statistical metrics.