Transfer learning and neural networks in predicting quadrupole deformation

Abstract

Abstract Accurately determining the quadrupole deformation parameters of atomic nuclei is crucial for understanding their structural and dynamical properties. This study introduces an innovative approach that combines transfer learning techniques with neural networks to predict the quadrupole deformation parameters of even-even nuclei. With the application of this innovative technique, the quadrupole deformation parameters of 2331 even-even nuclei are successfully predicted within the nuclear region defined by proton numbers \(8 \leq Z \leq 134\) and neutron numbers \(N \geq 8\). Additionally, the study discusses the impact of nuclear quadrupole deformation parameters on the capture cross-sections in heavy-ion fusion reactions, reconstructing the capture cross-sections for the reactions \(^{48}\text{Ca} + ^{244}\text{Pu}\) and \(^{48}\text{Ca} + ^{248}\text{Cm}\). This research offers new insights into the application of neural networks in nuclear physics and underscores the potential of merging advanced machine learning techniques with both theoretical and experimental data, particularly in fields where experimental data is limited.