Browse Teach Curate

Physics-Informed Neural Networks for Nuclear Atomic Mass Prediction

Nuclear masses are measured only near stability, so extrapolating to exotic nuclei needs models. This concept covers training fully connected neural networks on a macroscopic-microscopic decomposition, feeding a liquid-drop baseline and predicting the microscopic corrections as labels, to reproduce and extrapolate atomic-mass tables while recovering pairing and magic-number effects.

Data Text

Validation and extrapolation of atomic mass with physics-informed fully connected neural network

Atomic nuclear masses are measured only near the valley of stability, so predicting the masses of exotic neutron-rich nuclei relies on models. This paper trains fully connected neural networks to rep…