Research of Single-Nucleon Separation Energies in the Liquid Drop Model Based on Artificial Neural Networks

This research is based on the Liquid Drop Model (LDM). First, we introduce the artificial neural network (ANN) method to calculate the theoretical predictions of single-nucleon separation energies within the LDM. Subsequently, neural networks are utilized to optimize the theoretical values of single-nucleon separation energies, reducing the root mean square deviation (RMSD) between the theoretical predictions and experimental values. Three partitioning optimization schemes are then implemented: proton magic-number partitioning, neutron magic-number partitioning and proton/neutron parity partitioning. After partitioned optimization training, the RMSD is further reduced. The partitioned training of single-nucleon separation energies yields better results than direct global training, particularly in significantly lowering the RMSD in the light nuclei region. Specifically, single-neutron separation energies exhibit improved accuracy with neutron magic-number partitioning, while single-proton separation energies show better performance with proton magic-number partitioning.