Bayesian optimization and nonlocal effects method for α \alpha decay of superheavy nuclei based on CPPM

Xuanpeng Xiao 1, Panpan Qi 1, Gongming Yu 1, Haitao Yang 2, Qiang Hu 3 · 2025

Read Bayesian optimization and nonlocal effects method for α \alpha decay of superheavy nuclei based on CPPM by Xuanpeng Xiao 1, Panpan Qi 1, Gongming Yu 1, Haitao Yang 2, Qiang Hu 3 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-19091.txt.

Introduction

Xuanpeng Xiao and co-workers foretold alpha decay lifetimes of superheavy nuclei with a Bayesian neural network in a 2025 paper. Alpha decay half-lives guide the hunt for new elements, and the coupled-channels model with a nonlocal twist supplies the physics usual test. Adding a Bayesian network on top, the authors find nonlocal effects shift the math work materially. The network sharpens accuracy and extrapolates confidently beyond known nuclei.

An account tool ranks the inputs: the decay energy dominates the half-life, as physics expects. Applied to element chains beyond the current table's edge. The model flags a strong shell effect at a neutron count long foretold to be magic. Its forecasts obey the classic Geiger-Nuttall relation, a consistency check the authors highlight.

Catalog record

CREATORXuanpeng Xiao 1, Panpan Qi 1, Gongming Yu 1, Haitao Yang 2, Qiang Hu 3
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE32 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-19091

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Bayesian optimization and nonlocal effects method for α \alpha decay of superheavy nuclei based on CPPM Preprint: APS/123-QED Xuanpeng Xiao 1, Panpan Qi 1, Gongming Yu 1, Haitao Yang 2, Qiang Hu 3 Abstract We have integrated nonlocal effects with the Bayesian neural network (BNN) approach to enhance the prediction accuracy of α \alpha decay half-lives. The results indicate that incorporating nonlocal effects significantly impacts half-life calculations, while the BNN method substantially improves prediction accuracy and demonstrates strong extrapolation capabilities. Using Shapley Additive Explanations (SHAP), we conducted a quantitative comparison among the four input feature parameters in the BNN, revealing that the α \alpha decay energy Q α Q_{\alpha} is the predominant contributor to T 1 / 2 T_{1/2} . Leveraging the BNN’s robust extrapolation capability, we predicted α \alpha decay half-lives for the isotopic chains ( Z = 118,120 ) (Z=118,120) , revealing a significant shell effect at N = 184 N=184 of neutrons. Keywords: α \alpha decay, half-lives, nonlocal effects, Bayesian neural network, Coulomb and proximity potential model. I Introduction The synthesis and properties of superheavy nuclei represent a frontier in contemporary nuclear physics, playing a crucial role in exploring the limits of the nuclear chart and understanding fundamental nuclear forces. \alpha decay is one of the primary decay modes of superheavy nuclei, accurate prediction of \alpha decay half-lives is essential for identification and synthesis of new elements. \alpha decay is fundamentally a quantum tunneling phenomenon, independently proposed by Gamow 1 and Gurney and Condon 2 ; 3 in 1928. Various theoretical models have been developed to calculate \alpha decay half-lives of superheavy nuclei, including the generalized liquid drop model(GLDM) 4 ; 5 , effective liquid drop model(ELDM) 6 , modified generalized liquid drop model(MGLDM) 7 ; 8 ; 9 , Coulomb and proximity potential model(CPPM) 10 , and preformed cluster model(PCM) 11 . Concurrently, empirical formulas based on experimental data fitting have been extensively developed, such as the universal decay law(UDL) by Qi 12 , modified UDL(MUDL) incorporating angular momentum effects 13 , and the unified formula by Ni et al. 14 . While these models provide important theoretical foundations for understanding \alpha decay mechanisms, significant discrepancies persist between theoretical predictions and experimental data, particularly in the superheavy region. To improve prediction accuracy, researchers have focused on nonlocal effects in particle-nucleus interactions. The concept of nonlocal potentials was introduced by Feshbach 15 and Frahn et al. 16 in the late 1950s. Recent studies have further confirmed the importance of nonlocal effects: in 2022, Medeiros et al. 17 conducted a systematic study of the \alpha decay half-lives of \alpha particles with effective mass for 52\leq Z\leq 103 ; Hu et al. 18 systematically investigated nonlocal effects in \alpha decay half-lives for even-even nuclei within a two potential approach. These studies consistently demonstrate that incorporating nonlocal effects significantly improves theoretical predictions, offering a new avenue for reducing systematic deviations between theory and experiment. Traditional theoretical models face challenges in handling complex many-body interactions and nonlinear effects. Recent advances in machine learning have provided new tools for nuclear physics research. Machine learning exhibits unique advantages in treating high-dimensional, strongly correlated complex systems 19 . Various machine learning methods have been successfully applied in nuclear physics, including Gaussian processes(GP) 20 ; 21 , naive Bayes classifiers(NBC) 22 ; 23 , deep learning(DL) 24 ; 25 , restricted Boltzmann machines(RBM) 26 ; 27 , radial basis function(RBF) networks, Bayesian neural networks (BNN) 29 ; 30 , and light gradient boosting machines(LightGBM) 28 . Among these, BNN have attracted particular attention due to their advantages in uncertainty quantification and overfitting prevention. By incorporating the prior distribution of parameters, and Bayesian inference, BNN not only provides accurate predictions but also quantifies prediction uncertainties 19 . This capability has enabled BNN applications across multiple nuclear physics domains 31 ; 32 , including predictions of atomic radii 33 ; 34 , nuclear masses 35 ; 36 , and \beta decay half-lives 37 . In the present work, we systematically investigated theoretical calculations and model optimization for \alpha decay half-lives of superheavy nuclei. First, we calculate \alpha decay half-lives for 501 superheavy nuclides within the CPPM framework. To account for nonlocal dynamical effects, we introduce coordinate-dependent effective mass for \alpha particles and recalculate half-lives. The results indicate that the incorporation of nonlocal effects significantly impacts the \alpha decay half-lives in the model, consistent with previous studies 17 ; 18 . Subsequently, we construct a BNN model using mass number A , proton number Z , decay energy Q_{\alpha} , and orbital angular momentum l as input parameters to optimize CPPM and three other models (UDL, MUDL, and Ni formula). Comparing root-mean-square(RMSE) deviations before and after optimization reveals that BNN reduces RMSE by over 21\% for all models, with UDL showing 43.79\% improvement. Using Shapley additive explanations (SHAP) for feature importance analysis, the results reveal that the decay energy Q_{\alpha} plays a dominant role among the four input parameters. Finally, we predicted the alpha decay half-lives of the isotopic chains for Z=118 and Z=120 . This paper is organized as follows: Section 2 introduces \alpha decay theoretical models and BNN methodology; Section 3 presents results and analysis; Section 4 is a summary. II General Formalism II.1 Coulomb and proximity potential model framework The partial half-life is related to the decay constant \lambda by 38 \displaystyle T_{1/2}=\frac{\ln 2}{\lambda}=\frac{\ln 2}{\nu S_{c}P}, (1) where \lambda is the decay constant, \nu=1.0\times 10^{22}s^{-1} is the frequency of the assault on the barrier 39 ; 40 ; 41 . S_{c} is the preformation factor 38 , which can be expressed as 42 \displaystyle S_{c}=S_{\alpha}^{(A_{c}-1)/3}, (2) In this study, we choose S_{\alpha}=0.02897 for even-even parent nuclei and S_{\alpha}=0.0214 for odd-A parent nuclei 43 . The barrier penetration probability P can be calculated within the semi-classical Wentzel-Kramers-Brillouin (WKB) approximation. \displaystyle P=\exp(-\frac{2}{\hbar}\int_{R_{in}}^{R_{out}}\sqrt{2\mu|V_{r}-Q_{c}|}\mathrm{d}r), (3) where the reduced mass is \mu=\frac{mM}{m+M} . Here, with M and m being the daughter nucleus and the emitted cluster mass, respectively. Q_{c} represents the released energy 44 , which can be expressed as \displaystyle Q_{c}=B(A_{c},Z_{c})+B(A_{d},Z_{d})-B(A,Z), (4) where B(A_{c},Z_{c}) , B(A_{d},Z_{d}) and B(A,Z) are the binding energies of the emitted cluster, daughter nucleus, and parent nucleus, respectively 21 . Z_{c} , Z_{d} and Z are the proton numbers of the emitted cluster, daughter nucleus, and parent nucleus, respectively. The classical turning points R_{in}=A_{c}^{1/3}+A_{d}^{1/3} and R_{out}=(\sqrt{(\frac{Z_{c}Z_{d}e^{2}}{2Q_{c}})^{2}+\frac{\hbar^{2}(l+1/2)^{2}}{2\mu Q_{c}}}+\frac{Z_{c}Z_{d}e^{2}}{2Q_{c}}) 45 . Here, A_{c} , A_{d} , and A are the mass numbers of the emitted cluster, the daughter nucleus, and the parent nucleus, respectively. The total interaction potential V(r) between the emitted cluster and the daughter nucleus consists of the nuclear potential V_{N}(r) , Coulomb potential V_{C}(r) , and centrifugal potential V_{l}(r) . It can be expressed as \displaystyle V(r)=V_{N}(r)+V_{C}(r)+V_{l}(r), (5) In the present work, we adopt the proximity potential formalism in place of the nuclear potential.In the 1970s, Blocki et al. 46 proposed the original form of the proximity potential formalism for two interacting spherical nuclei, which is expressed as \displaystyle V_{N}(r)=4\pi\gamma b\bar{R}\Phi(\xi), (6) where b\approx 1 fm. Surface energy coefficient \gamma has the following form 47 : \displaystyle\gamma=\gamma_{0}(1-k_{s}I^{2}), (7) where I=\frac{N_{p}-Z_{p}}{A{p}} is the asymmetry parameter quantifying the neutron–proton excess of the parent nucleus; Here, \gamma_{0} and k_{s} denote the surface-energy coefficient and the surface-asymmetry coefficient,

Read the complete text