Bayesian optimization approach to model-based description of <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:mi>α</mml:mi></mml:math> decay
Zisheng Jin, Mingshuai Yan, Hao Zhou, An Cheng, Zhongzhou Ren, Jian Liu
Abstract
$\ensuremath{\alpha}$ decay serves as an important probe for the studies of unstable nuclei. This paper proposes an approach combining the sophisticated $\ensuremath{\alpha}$-decay model and Bayesian neural network (BNN) to improve the prediction accuracy of $\ensuremath{\alpha}$-decay half-lives. The global and extrapolated analyses show that the BNN method can improve the description of model-based predictions of $\ensuremath{\alpha}$ decay. In our calculation, the experimental decay energies ${Q}_{\ensuremath{\alpha}}$ are used to obtain the accurate $\ensuremath{\alpha}$-decay penetration probability, which indicates that the improvements come from the corrections of $\ensuremath{\alpha}$-cluster preformation factors. Further analyzing $\ensuremath{\alpha}$-decay half-lives of nuclide chains shows that the shell structure effect can be well introduced into estimations of $\ensuremath{\alpha}$-cluster preformation factors by utilizing the BNN. The studies of this paper provide an effective way to predict the $\ensuremath{\alpha}$-decay half-lives of unknown nuclei.