Nuclear Mass Models with Deformation Optimized and Diagnosed by Bayesian Analysis with Markov Chain Monte Carlo

arXiv:2606.30519 · nucl-th, nucl-ex · Submitted 2026-06-29 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Nuclear Theory. The day in nuclear theory: structure, reactions and the models behind them.

Kai: We're looking at the paper: "Nuclear Mass Models with Deformation Optimized and Diagnosed by Bayesian Analysis with Markov Chain Monte Carlo".

Mira: Here's the gist of the paper. Nuclear Mass Models with Deformation Optimized and Diagnosed by Bayesian Analysis with Markov Chain Monte Carlo reveals that using full Bayesian analysis with adaptive Metropolis-Hastings Markov chain Monte Carlo sampling to systematically study the posterior probability distributions of the strengths of energy terms in optimized nuclear mass models of Bethe-Weizs¨acker variants provides a robust method for inferring parameter degeneracies and proposing new mass models like BWL.

Kai: Model Refinement and Optimization The authors employ a full Bayesian analysis with adaptive Metropolis-Hastings Markov chain Monte Carlo (BA-MCMC) sampling to systematically study the posterior probability distributions of the strengths of energy terms in optimized nuclear mass models of Bethe-Weizs¨acker variants.

Mira: They analyze selected refined models to determine parameter degeneracies while proposing a new macroscopic-microscopic mass model, BWL, which considers quadrupole and highmultipole deformation and shell corrections.

Kai: All mass models in this work are analyzed and optimized through the BA-MCMC method.

Title and authors: Kai: So, let's talk about who wrote this: Xiangnan Lee, Yi Hua Lam, Zi-Ao Zhang, and Jayke Ren. They are from Zhejiang Sci-Tech University and RIKEN in Japan.

Mira: They have a solid background in both the theoretical side of condensed matter and the experimental side of nuclear physics, which is important when you're trying to link those two approaches.

Lev: I wonder what their specific expertise is with this particular Bayesian analysis method, since MCMC sampling can be tricky to implement correctly for complex parameter spaces.

Kai: Their work focuses on using this full Bayesian analysis with adaptive Metropolis-Hastings Markov chain Monte Carlo sampling to systematically study the posterior probability distributions of the strengths of energy terms in optimized nuclear mass models of Bethe-Weizs¨acker variants.

Mira: That method allows them to analyze selected refined models to determine parameter degeneracies while proposing a new macroscopic-microscopic mass model, BWL, which considers quadrupole and highmultipole deformation and shell corrections.

Lev: The title itself suggests the core contribution is not just the model itself, but the rigorous diagnostic tool—the Bayesian analysis coupled with MCMC—for optimizing those models.

Kai: It's about using this full Bayesian analysis to systematically study those energy term strengths and then proposing BWL as a new model.

Mira: The implication here is that they are moving beyond just finding one good set of parameters and instead providing a probability distribution over all the possible good sets.

The paper's summary: Kai: So, the summary says they use this framework to analyze selected refined models to determine parameter degeneracies while proposing a new macroscopic-microscopic mass model, BWL.

Mira: What that means practically is they are finding which parameters are correlated with each other in a complex way within the fitting matrix.

Lev: And how do we know if these correlations matter? That's where the next part of the paper should show its work, because if those correlations aren't physical, you haven't really improved anything.

Kai: They analyze all mass models using this method and optimize them through the BA-MCMC method.

Mira: And when they propose BWL, they are incorporating deformation effects and a phenomenological shell correction into the macroscopic energy by using a multiplicative enhancement factor applied to the macroscopic energy.

Lev: So, the core idea is that you get a model that accounts for shape explicitly, not just implicitly through some fixed parameters in an older formula.

Kai: And when they compare BWL with two thousand two hundred forty-two precise experimental binding energies from AME2020, they report a root-mean-square deviation of seven hundred fifty-nine keV.

Mira: That seven hundred fifty-nine keV deviation is particularly good for describing masses in the light nuclei and the actinide regions, which are often where these models struggle most.

The paper's improvements: Kai: So, let's focus on what they claim as specific improvements. They show that the inclusion of deformation corrections in BWL alleviates multi-modality observed in several posterior distributions of the BWN model, resulting in more unimodal and symmetric posterior distributions.

Mira: Multi-modality means the analysis was getting confused, with several distinct possible solutions for the parameters, which is a big problem when you're trying to pin down a single physical reality.

Lev: If it becomes unimodal and symmetric, that means the model has settled on one clear physical picture instead of bouncing around between several bad ones. That makes inference much more reliable.

Kai: They also show that using the Spearman rank correlation matrix is more appropriate than the Pearson correlation coefficient for analyzing those non-linear correlations among parameters of a mass model.

Mira: That’s interesting because Pearson measures linear relationships, but Spearman looks at the ranks, which is better suited for the kind of complex, non-linear dependencies you see in these models.

Lev: If they can diagnose these degeneracies using that rank correlation matrix—say when K(RS) is greater than ten to the power of three—that gives us a concrete way to say "this parameter combination is probably unphysical <ref:2606.30519#pg1>."

Conclusion: Kai: So, to wrap up, the paper on "Nuclear Mass Models with Deformation Optimized and Diagnosed by Bayesian Analysis with Markov Chain Monte Carlo" shows that BA-MCMC is a robust tool for diagnosing parameter degeneracies and guiding the refinement of nuclear mass models.

Mira: The new model, BWL, which has its clearer physical picture and more regular parameter space, provides a solid foundation for further improvements on macroscopic-microscopic nuclear mass models.

Lev: From my side, what I see is that this approach gives us a statistically sound path forward because it moves away from just point estimates and gives us that full posterior distribution quantifying all the uncertainty.

Kai: And they’ve managed to reduce the global root-mean-square deviation of the BWN model from eight hundred sixty-three keV down to seven hundred fifty-nine keV with the BWL model.

Mira: It’s a tangible improvement, especially in how accurately we describe masses for isotopes in light nuclei and actinides.

Lev: So, this study confirms that using this full Bayesian analysis framework is a reliable way to refine these models for future work on nuclear structure.

Xiangnan Lee, Yi Hua Lam, Zi-Ao Zhang, Jayke Ren

Zhejiang Key Laboratory of Quantum State Control and Optical Field Manipulation · Astrophysical Big Bang Laboratory, Pioneering Research Institute, RIKEN

nucl-th, nucl-ex

Submitted: 2026-06-29

Updated: 2026-09-10

Comments: Accepted by Chinese Physics C on 7 Sep 2026 with minor revision, finalized version, 24 pages, 7 main figures (colorblind-friendly colors), 12 tables, data availability: this https URL{https://zenodo.org/records/21988436}

Journal ref: Chinese Physics C 51, 014102 (2026); the published version is unexpectedly and slightly edited by the production department of Chin. Phys. C. for further clarity

DOI: 10.1088/1674-1137/aea548

Project page: https://lamyihua.github.io

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 81/100

The gist: Nuclear Mass Models with Deformation Optimized and Diagnosed by Bayesian Analysis with Markov Chain Monte Carlo reveals that using full Bayesian analysis with adaptive Metropolis-Hastings Markov…

Key concepts

Full Bayesian Analysis with MCMC
This is a statistical framework where unknown parameters in a nuclear model are treated as random variables. It combines prior knowledge with experimental data to produce a full posterior distribution, which shows the probability of different parameter values, giving a complete picture of uncertainty.
Parameter Degeneracy Diagnosis
This involves using tools like the Spearman rank correlation coefficient to measure how strongly different parameters are correlated or redundant within the model. High correlation indicates that multiple parameters can produce similar results, which helps identify physical degeneracies in the model's parameter space.
Deformation Corrections
These terms are added to the energy calculations in models like BWL to account for how a nucleus deforms (changes shape). By including these corrections, the model better describes nuclear properties, leading to a more physically realistic and less ambiguous mass prediction.
Root-Mean-Square Deviation (RMSD)
This is a measure of the average difference between the theoretical binding energies calculated by a mass model and the experimentally measured binding energies. A lower RMSD indicates that the model provides more accurate predictions for nuclear masses across various isotopes.

Terminology

Summary

Nuclear Mass Models with Deformation Optimized and Diagnosed by Bayesian Analysis with Markov Chain Monte Carlo reveals that using full Bayesian analysis with adaptive Metropolis-Hastings Markov chain Monte Carlo sampling to systematically study the posterior probability distributions of the strengths of energy terms in optimized nuclear mass models of Bethe-Weizs¨acker variants provides a robust method for inferring parameter degeneracies and proposing new mass models like BWL. p. 0

Model Refinement and Optimization

The authors employ a full Bayesian analysis with adaptive Metropolis-Hastings Markov chain Monte Carlo (BA-MCMC) sampling to systematically study the posterior probability distributions of the strengths of energy terms in optimized nuclear mass models of Bethe-Weizs¨acker variants. They analyze selected refined models to determine parameter degeneracies while proposing a new macroscopic-microscopic mass model, BWL, which considers quadrupole and highmultipole deformation and shell corrections. p. 0 All mass models in this work are analyzed and optimized through the BA-MCMC method. p. 0 Compared with 2242 precise experimental binding energies of AME2020, BWL produces a root-mean-square deviation of 759 keV, particularly improving the description of masses in the light-nuclei and actinide regions. p. 0

Bayesian Inference Framework

Bayesian inference offers an alternative to exploring the parameter space by treating unknown parameters as random variables and combining prior information with the likelihood of observed data, giving the full posterior distribution that quantifies all sources of uncertainty. The Markov chain Monte Carlo (MCMC) sampling method generates approximately independent samples from the target posterior distribution to support subsequent statistical inference, avoiding linear approximations and thus preventing the introduction of unknown errors. p. 1

Parameter Correlation Diagnosis

The authors employ the Spearman rank correlation coefficient to diagnose and quantify parameter correlations and degeneracies of the multi-dimensional parameter spaces evaluated from the MCMC posterior samples. Parameter degeneracy within the entire fitting matrix is monitored by the condition number of a given Spearman rank correlation matrix, K(RS) = λmax λmin, typically indicating a high degree of parameter degeneracy when K(RS) > 103. p. 6 They show that using the Spearman rank correlation matrix is more appropriate than the Pearson correlation coefficient for analyzing the non-linear correlations among parameters of a mass model. p. 1

Deformation Effects and Model Improvement

The BWL model incorporates deformation effects and a phenomenological shell correction into the LDM energy, using a multiplicative enhancement factor applied to the macroscopic energy. The analysis shows that the inclusion of deformation corrections in BWL alleviates the multi-modality observed in several posterior distributions of the BWN model, resulting in more unimodal and symmetric posterior distributions. p. 9 The BWL† model reduces the global RMS deviation of the BWN† model from 863 keV to 759 keV for the. p. 0

Conclusion and Advantage

The current study indicates that BA-MCMC is a robust tool for diagnosing parameter degeneracies and guiding the refinement of presently available (and future) nuclear mass models. p. 0 The new mass model, BWL, with its clearer physical picture and more regular parameter space, provides a solid foundation for further improvements on macroscopic-microscopic nuclear mass models. p. 0 p. 0

Pruning redundant reasoning tokens

The paper systematically investigates the nuclear mass models of the Bethe-Weizs¨acker variants, i.e., BW, BWK, BWN, and BWL, using full Bayesian analysis with Markov chain Monte Carlo sampling (BAMCMC). p. 3

How the training signal is built

The likelihood function assumes that the model residuals (the differences or deviations between experimental and theoretical binding energies) are independent and identically distributed according to Gaussian uncertainty. The log-likelihood reads-n/2 log(2πσ 2) - 1/2σ squared X i=1 h B Exp i − B Th i(θfull) i squared. p. 4

Where the gains come from

The introduction of the deformation correction terms with the parameters g1, g2, and γ (Eq. (5)) in BWL alleviates the multi-modality observed in several posterior distributions of the BWN model. p. 9 The consideration of deformation smooths the likelihood surface, making posterior distributions more unimodal. p. 7 The overall RMS deviation for 2242 nuclei is reduced from 863 keV to 759 keV, corresponding to an improvement of approximately 12 %. p. 0

The authors employ a full Bayesian analysis with adaptive Metropolis-Hastings Markov chain Monte Carlo (BA-MCMC) sampling to systematically study the posterior probability distributions of the strengths of energy terms in optimized nuclear mass models of Bethe-Weizs¨acker variants.

The posterior distribution for a new set of binding energies B new given by the observed data D is defined as p(B new D) = Z p(B new θfull) p(θfull D) dθfull.

The analysis of posterior distributions reveals mild multi-modality in the posterior distributions of several BWN parameters, i.e., αv, αs, αc, and αr. p. 9 The condition number of Pearson correlation matrix for BWN drops from 1. p. 642×10 6 to 1.13×10 6 (BWL), demonstrating a moderate reduction in parameter degeneracies. p. 0

The authors employ the Spearman rank correlation coefficient to diagnose and quantify parameter correlations and degeneracies of the multi-dimensional parameter spaces evaluated from the MCMC posterior samples.

The Spearman rank correlation coefficient, ρ ij = 1 − 6 P diag k=1 d 2 k ndiag(n squared diag − 1), where d k = R[θi] k − R[θj] k represents the difference between the R[θi] and R[θj] ranks of the k-th MCMC sample for the two parameters.

The new mass model, BWL, with its clearer physical picture and more regular parameter space, provides a solid foundation for further improvements on macroscopic-microscopic nuclear mass models. The paper demonstrates that using the Spearman rank correlation matrix is more appropriate than the Pearson correlation coefficient for analyzing the non-linear correlations among parameters of a mass model. p. 1

The posterior distribution for all model parameters exhibits a symmetric and unimodal Gaussian shape, with the posterior means concentrated within narrow ranges. The symmetrical parameter space is mainly due to (i) the four parameters (αv, αs, αc, αsym) of the BW model are linearly correlated. p. 7

The authors show that using the Spearman rank correlation matrix is more appropriate than the Pearson correlation coefficient for analyzing the non-linear correlations among parameters of a mass model. The current study indicates that BA-MCMC is a robust tool for diagnosing parameter degeneracies and guiding the refinement of presently available (and future) nuclear mass models.

The detailed correlation strengths among the model parameters are presented in Table X, which shows that for BWL, the condition number of the Pearson correlation matrix increases to 1.09×10 cubed.

The authors show that using the Spearman rank correlation matrix is more appropriate than the Pearson correlation coefficient for analyzing the non-linear correlations among parameters of a mass model. The new mass model, BWL, with its clearer physical picture and more regular parameter space, provides a solid foundation for further improvements on macroscopic-microscopic nuclear mass models.

The posterior distributions for each parameter of BWK model slightly deviate from Gaussian shapes and show skewness with longer tails, shifting the overall joint-posterior distributions from an elliptical shape.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Accurate Nuclear Mass Prediction for Exotic Nuclei: The improved system can now predict binding energies for isotopes far from stability with a root-mean-square deviation of 759 keV, particularly improving the description of masses in the light-nuclei and actinide regions compared to previous models.

  2. Robust Parameter Degeneracy Diagnosis: By using Spearman rank correlation matrix analysis, the system can diagnose parameter degeneracies in mass models, revealing strong correlations of some energy terms for some mass models, which is crucial for understanding the physical meaning of the model.

  3. Uncertainty Quantification via Full Posterior Distributions: The system moves beyond point estimates to provide full posterior distributions that quantify all sources of uncertainty, including correlations among parameters and the intrinsic noise level sigma, offering a complete description of parameter uncertainty.

  4. Model Refinement through Deformation Incorporation: The system can now implicitly account for deformation specifically for each isotope by incorporating deformation corrections (Eq. 5) and a deformation quenching factor (Eq. 9), leading to a reduction of the global RMS deviation from BWN to BWL from 863 keV to 759 keV.

  5. Improved AI Training Database Generation: The system produces a high precision training database for developing artificial intelligence (AI) mass models, which enables the development of more accurate and interpretable physics features for physics-informed AI approaches.

Sources

Related papers