This work combines the generalized liquid drop model (GLDM) with the Bayesian neural network (BNN) to systematically investigate the $\alpha$-decay half-lives of nuclei in the superheavy region. By training the BNN to learn the nonlinear relation between the residuals of theoretical results and experimental data and the input characteristic parameters, the GLDM results are optimized. Using the GLDM-BNN, the standard deviation of the $\alpha$-decay half-lives for all nuclei decreases from 1.107 to 0.480, corresponding to a relative difference of $56.6\%$, indicating that the BNN effectively captures hidden physical information. Furthermore, the maximal information coefficient (MIC) is employed to analyze the correlation between each input feature and the $\alpha$-decay half-life. Among them, the $\alpha$-decay energy exhibits the highest MIC value of 0.490, highlighting its dominant role in determining the $\alpha$-decay half-life. Finally, the GLDM-BNN is used to predict the $\alpha$-decay half-lives of experimentally new superheavy nuclei with atomic numbers $Z=113–118$. The predicted results provide strong evidence supporting $N=184$ as the next neutron magic number after $N=126$.

Close

PrevNext

View figure in article

• Alpha decay
• Lifetimes & widths
• A ≥ 220
• Artificial neural networks
• Bayesian networks
• Machine learning