Despite the advancements of research on $\alpha$-decay half-lives $T_{1/2}$, accurate studies of the theoretical calculation of unfavored $\alpha$ transitions remain an open problem in nuclear physics. This paper performs a separate physical interpretation of both favored and unfavored $\alpha$ transitions by combining the deformed density-dependent cluster model (DDCM) with a Bayesian neural network (BNN). The global and extrapolated analyses demonstrate the effectiveness of the BNN in refining predictions of $T_{1/2}$, particularly for the intricate unfavored transitions. Furthermore, quantitative comparisons using the maximum information coefficient (MIC) across nine characteristic parameters within BNN inputs systematically uncover hidden nuclear structure effects inherent to both favored and unfavored $\alpha$ decays. The comparisons show that the contribution of $\alpha$-decay energy $Q_{\alpha }$ to $T_{1/2}$ is dominant in nine characteristic parameters, and the influence of deformation is conspicuous in the open-shell region. Compared to favored $\alpha$ transitions, the correlations reflected by MIC between each characteristic parameter and $T_{1/2}$ are stronger in unfavored $\alpha$ transitions, indicating its more complex internal physical mechanism. This study advances the exploration of physical interpretation of $\alpha$ transitions and provides actionable insights to guide future investigations in nuclear physics.

Close

PrevNext

View figure in article

• Alpha decay
• Lifetimes & widths
• Bayesian methods
• Machine learning