Speaker
Description
ABSTRACT
Nuclear decay properties of heavy and neutron-rich nuclei strongly influence the stability and abundance evolution of nuclei produced in rapid neutron-capture (r-process) nucleosynthesis. In this work, we develop a transfer learning-driven machine learning framework to derive interpretable analytical expressions for α-decay observables. A Random Forest model is first trained to predict α-decay energies (Qα) using nuclear structure parameters, including proton number (Z), neutron number (N), pairing factor (δ), and promiscuity factor (P). The predicted (Qα) values are subsequently incorporated as an informative feature in a symbolic regression framework to derive compact analytical formulae for both (Qα) and the α-particle preformation factor (Pα). To quantify the physical relevance of the input variables, feature contributions are investigated using SHapley Additive exPlanations (SHAP), revealing the dominant role of (Qα), shell structure, and pairing effects in determining α-decay characteristics. The symbolic regression models reproduce experimental trends with high accuracy while providing transparent mathematical relations between nuclear properties and α-decay observables. The derived analytical expressions provide physically interpretable estimates of (Qα) and α-particle preformation factors, enabling systematic investigations of α-decay systematics in heavy and neutron-rich nuclei relevant to nucleosynthesis pathways.
Reference
1. W. Huang et al., Chin. Phys. C 45, 030002 (2021).
2. K. Jyothish, Govardhan Manangode, A.K. Rhine Kumar, Phys. Rev. C 112 (2025) , 064309
3. K. Jyothish, V. Parvathi, A. K. Rhine Kumar, Pramana- J. Phys. 99, 53 (2025).
4. J E Perez Velasquez et al. J. Phys. G: Nucl. Part. Phys. 50 (2023).