Refining Nuclear Mass Predictions in the Ni-Sn Region via Physics-Informed Gaussian Processes

Email contact : multiversou@gmail.com Accurate prediction of nuclear binding energies far from stability is crucial for r-process nucleosynthesis simulations. While Covariant Density Functional Theory (CDFT) provides a robust global description, it often exhibits systematic residuals near shell closures due to incomplete treatment of complex correlations. In this work, we implement a Physics-Informed Gaussian Process (GP) regression to correct the binding energy residuals of the DD-PC1 functional specifically for the Nickel ($Z=28$) and Tin ($Z=50$) isotopic chains. Unlike standard "black-box" machine learning approaches, our model integrates explicit physical features to ensure robust extrapolation. A key novelty is the inclusion of experimental quadrupole deformation parameters ($|\beta_2|$) derived from $B(E2)\uparrow$ transition probabilities, allowing the model to account for dynamical shape fluctuations often missed by static mean-field approaches.