A Data-Driven Density Functional Model for Nuclear Systems

Abstract

Through ensemble learning with multitasking and complex connection neural networks, we aggregated nuclear properties, including ground state charge radii, binding energies, and single-particle state information obtained from the Kohn-Sham auxiliary single-particle systems. Compared to traditional density functional theory, our model can more accurately characterize nuclear ground state information. Aiming at binding energy, the root mean square error is reduced to 450 keV. Although the complexity involving the nuclear interaction is skipped, the model has not completely devolved into a black box. Leveraging the correlation between densities and binding energies, we calculate the neutron skin thickness of $^{208}$Pb to be 0.223 fm. This model will advance our understanding of nuclear properties and accelerate the integration of machine learning into modern nuclear physics.