Prediction on α-decay Energy of Heavy and Superheavy Nuclei Using Machine Learning

Abstract

α-decay is one of the most dominant decay modes of unstable nuclei. It is helpful for the investigation of exotic nuclear structures. Moreover, α-decay is a powerful tool for identifying the newly synthesized nuclides for heavy and superheavy nuclei in the experiment. It is of vital importance to propose more theoretical models to provide more accurate predictions on α-decay properties. Recently, machine learning is widely applied to investigate various essential problems in nuclear physics. As one of the popular machine learning algorithms, the Gaussian process is successfully used in the studies of the α-decay properties of some neutron-deficient actinide nuclei. It is of great interest to probe the feasibility of the Gaussian process in the studies of α-decay properties for nuclei in more nuclear regions. In this work, the Gaussian process was extended to study the α-decay energy of heavy and superheavy nuclei. Two common kernel functions named the Matérn 3/2 kernel function and the Matérn 5/2 kernel function, were used in the calculations. First, the α-decay energy of the new nuclide 207Th was calculated using the Gaussian process. The deviations between the theoretical α-decay energy calculated using the Gaussian process with two different kernel functions and the experimental one are 0.025 MeV and 0.060 MeV, respectively. The small deviations show that the calculated α-decay energy is in good accord with the experimental data, which demonstrates that the Gaussian process is reliable for the studies of α-decay energy when combined with the good cross-validation results. Besides, it is found that there would be slightly better results when calculating α-decay energy using the Gaussian process with the Matérn 3/2 kernel function. The reason may be that the Matérn 3/2 kernel function is smoother than the Matérn 5/2 kernel function. Due to the small change of α-decay energy for heavy and superheavy nuclei, the Matérn 3/2 kernel function is more suitable for the calculation of them. Then, the unknown α-decay energy for some heavy and superheavy nuclei with 89≤Z≤118 was predicted using the Gaussian process, and the predicted α-decay energy was compared with that calculated using the traditional models. It is found that the theoretical results calculated using the Gaussian process show good agreement with those calculated using the traditional models. Furthermore, the predicted α-decay energy for 251Bk, 270Sg, and 272Hs shows the influence of the deformed shell at N=154 and the deformed subshell at N=164. The predicted results can be new references for future experiments on heavy and superheavy nuclei.

Keywords

• α-decay
• machine learning
• gaussian process