Statistical uncertainty quantification for multireference covariant density functional theory
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: I'm Kai.
Mira: And I'm Mira, here with Lev.
Kai: We're looking at the paper: "Statistical uncertainty quantification for multireference covariant density functional theory".
Mira: Here's the gist of the paper. Statistical uncertainty quantification for multireference covariant density functional theory presents a theoretical framework to quantify statistical uncertainties in covariant density functional theory (CDFT) for both nuclear matter and finite nuclei based on a relativistic point-coupling energy density functional (EDF) By sampling approximately one million parameter sets, with nine parameters varied around their values in the PC-PK1 functional, we construct a probability density function for nuclear matter properties Incorporating empirical values of nuclear matter at saturation density and those of predictions from chiral nuclear forces, and measured B(E2) values of finite nuclei, we infer posterior distributions for the model parameters within a Bayesian framework These posterior distributions are then propagated to the low-lying states of finite nuclei using the newly developed subspace-projected (SP)-CDFT approach We find that the observables of low-lying states in deformed nuclei Nd 150 and Sm 150 are well reproduced once statistical uncertainties are taken into account Theoretical Framework Development The paper introduces a theoretical framework to quantify statistical uncertainties in covariant density functional theory (CDFT) for both nuclear matter and finite nuclei based on a relativistic point-coupling energy density functional (EDF).
Kai: This framework involves constructing a probability density function for nuclear matter properties by sampling approximately one million parameter sets around the PC-PK1 functional.
Mira: Furthermore, it incorporates empirical values of nuclear matter at saturation density and predictions from chiral nuclear forces to infer posterior distributions for the model parameters within a Bayesian framework.
Kai: These distributions are then propagated to the low-lying states of finite nuclei using the subspace-projected (SP)-CDFT approach.
Title and authors: Kai: We're moving on to the improvements suggested by this paper regarding "Statistical uncertainty quantification for multireference covariant density functional theory" >
Mira: They introduce subspace projected covariant density functional theory or SP-CDFT as an emulator for multireference covariant density functional theory, which they call MR-CDFT >
Lev: They use the eigenvector continuation method or EC to build the wave function of a target Hamiltonian as a superposition of the first k max states >
Kai: That means they are trying to make these computationally expensive MR-CDFT calculations run much faster by using this subspace projection technique >
Mira: The efficiency gain comes from breaking down the interaction energy into nine terms, and using those to compute the Hamiltonian kernels of EC efficiently with the GCM kernels of their training EDFs >
Lev: They show that for certain nuclear mass numbers A less than N squared t, SP-CDFT is faster than MR-CDFT by a factor related to O(N 2t) compared to O(N 2qNC⊙∆T1) for the MR-CDFT calculation > <ref:2408.13209#pg1>
Kai: That speed up factor increases almost linearly up to ten to the fourth when you have one million sampling EDFs, which means you can run millions of SP-CDFT calculations with several orders of magnitude less time than doing MR-CDFT > <ref:2408.13209#pg3>
Mira: It's a big computational win, but they also show that for near-spherical nuclei, the low-lying states are still challenging to reproduce with this approach >
Lev: So, the method is very good for deformed systems like Nd one hundred fifty and Sm one hundred fifty but it clearly stops working as well on near spherical systems unless you do more work on that specific part >
Kai: It seems like the authors are admitting a limitation right there about where this particular SP-CDFT method is strongest >
The paper's summary: Mira: So, to summarize what the paper actually does, they create a framework to quantify statistical uncertainties in covariant density functional theory for both nuclear matter and finite nuclei using a relativistic point-coupling energy density functional or EDF >
Kai: They do this by sampling one million parameter sets around the PC-PK1 functional and then using those samples to build a probability density function for nuclear matter properties >
Lev: And they use empirical values from saturation density and predictions from chiral nuclear forces alongside measured B E2 values of finite nuclei to infer posterior distributions for the model parameters within a Bayesian framework >
Kai: Then they take those parameter distributions and propagate them to the low-lying states of finite nuclei using that new subspace projected covariant density functional theory, SP-CDFT approach >
Mira: The core idea is combining multireference CDFT with the eigenvector continuation method to get an emulator for nuclear low-lying states >
Lev: This approach allows them to quantify how reliable those predictions are by showing the statistical errors associated with the nine EDF parameters reach up to twenty one percent for excitation energies and twelve percent for E2 transition strengths >
Kai: It’s a detailed picture of how much uncertainty we have when we try to use this kind of relativistic framework to study nuclear structure >
The paper's improvements: Mira: The paper suggests that the main way they improve things is by introducing the SP-CDFT approach, which acts as an emulator for MR-CDFT >
Kai: This allows them to bypass the huge computational cost of calculating multireference DFT calculations for nuclear low-lying states >
Lev: They do this by using eigenvector continuation to represent the eigenvector of a target Hamiltonian within a low dimensional subspace spanned by the eigenvectors of sampling Hamiltonians >
Mira: And they also emphasize that this method can be efficiently coupled with various many-body methods, which have been tested in toy models and also in nuclear structure and reaction studies >
Kai: The efficiency gain is tied directly to how they break down the interaction energy into nine terms to compute the Hamiltonian kernels of EC using those GCM kernels from their training EDFs >
Lev: So, breaking it into nine terms is the key that allows them to execute millions of SP-CDFT calculations much faster than MR-CDFT, which is what they needed for real hardware constraints >
Mira: It’s essentially a practical way to make these complex multireference calculations feasible for studying actual nuclear structure data >
Conclusion: Kai: So to wrap up on this paper "Statistical uncertainty quantification for multireference covariant density functional theory" the main implication is that we can now predict low-lying states in deformed nuclei with quantified statistical errors >
Mira: Because they are well reproduced for Nd one hundred fifty and Sm one hundred fifty but they aren't as good for near spherical nuclei, so it sets expectations about what we should expect from this method moving forward >
Lev: From my side, the main thing is that we can use Bayesian analysis to constrain the parameters of these relativistic EDFs using nuclear matter properties to refine the parameter sets used in calculations for finite nuclei >
Kai: And they’ve shown that you get those constraints on parameters, and then you get predictions with statistical errors up to twenty one percent for excitation energies and twelve percent for E2 transition strengths >
Mira: It’s a solid result, but it highlights the intrinsic limitations of the underlying EDF framework because they say that even after accounting for statistical uncertainties, near-spherical nuclei remain challenging >
Lev: For real hardware, this means we have a robust way to estimate the uncertainty of our predictions when we use these tools to model things like exotic isotopes or even those in neutron stars >
Kai: So that's the idea. We can use this Bayesian approach to refine the parameters, and then use the SP-CDFT method for faster calculations, keeping in mind those twenty-one percent and twelve percent uncertainty bounds > <ref:2408.13209#pg3>
X. Zhang, C. C. Wang, C. R. Ding, J. M. Yao
School of Physics and Astronomy, Sun Yat-sen University · Department of Physics, Kyoto University · Graduate School of China Academy of Engineering Physics · Yukawa Institute for Theoretical Physics, Kyoto University
nucl-th, nucl-ex
Submitted: 2024-08-23
Updated: 2026-04-09
Comments: 12 pages with 9 figures and 3 tables
Journal ref: Chinese Physics C 50, 104107 (2026)
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 93/100
The gist: Statistical uncertainty quantification for multireference covariant density functional theory presents a theoretical framework to quantify statistical uncertainties in covariant density functional…
Key concepts
- Covariant Density Functional Theory (CDFT)
- A theoretical framework used to describe nuclear structure based on relativistic principles. It is employed here to calculate properties for both bulk nuclear matter and specific finite nuclei, providing a more rigorous description than non-relativistic methods.
- Subspace-Projected DFT (SP-CDFT)
- A new approach that acts as an emulator for computationally intensive multireference calculations. It constructs the wave function of a target state by superimposing a small number of basis states, making the calculation much faster than traditional methods.
- Bayesian Framework
- A statistical method used to update beliefs about model parameters based on experimental data. The authors use this to infer the most probable values for EDF parameters by combining prior knowledge (like nuclear matter data) with new information from measurements.
- Multireference (MR)-CDFT Emulator
- SP-CDFT is designed to mimic the results of complex multireference calculations efficiently. It achieves this by using eigenvector continuation, which builds the target state wave function from a superposition of a few initial states, saving significant computational time.
Terminology
Summary
Statistical uncertainty quantification for multireference covariant density functional theory presents a theoretical framework to quantify statistical uncertainties in covariant density functional theory (CDFT) for both nuclear matter and finite nuclei based on a relativistic point-coupling energy density functional (EDF) By sampling approximately one million parameter sets, with nine parameters varied around their values in the PC-PK1 functional, we construct a probability density function for nuclear matter properties Incorporating empirical values of nuclear matter at saturation density and those of predictions from chiral nuclear forces, and measured B(E2) values of finite nuclei, we infer posterior distributions for the model parameters within a Bayesian framework These posterior distributions are then propagated to the low-lying states of finite nuclei using the newly developed subspace-projected (SP)-CDFT approach We find that the observables of low-lying states in deformed nuclei Nd 150 and Sm 150 are well reproduced once statistical uncertainties are taken into account
Theoretical Framework Development
The paper introduces a theoretical framework to quantify statistical uncertainties in covariant density functional theory (CDFT) for both nuclear matter and finite nuclei based on a relativistic point-coupling energy density functional (EDF). This framework involves constructing a probability density function for nuclear matter properties by sampling approximately one million parameter sets around the PC-PK1 functional. p. 0 Furthermore, it incorporates empirical values of nuclear matter at saturation density and predictions from chiral nuclear forces to infer posterior distributions for the model parameters within a Bayesian framework. p. 0 These distributions are then propagated to the low-lying states of finite nuclei using the subspace-projected (SP)-CDFT approach. p. 0
Emulating Multireference Calculations
The paper introduces the subspace-projected covariant density functional theory (SP-CDFT) as an emulator of the multireference (MR)-CDFT for nuclear low-lying states based on the eigenvector continuation (EC) method. This approach constructs the wave function of the k-th state for a target EDF as a superposition of the first kmax states. p. 2 The mixing coefficient is determined by an equation involving the Hamiltonian kernel and norm kernel of the EC method. p. 2 The main ingredients of SP-CDFT are the norm kernels and Hamiltonian kernels which can be efficiently determined using the nine terms of the interaction energy. p. 3
Computational Efficiency Gains
The computational cost of MR-CDFT calculations is given by TMR−CDFT = O(N 2qNC⊙∆T1). In contrast, the time complexity of SP-CDFT(Nt, kmax) is composed of two parts: TSP−CDFT = O(N 2qN 2t∆T1 + O(N 2ECNC⊙∆T2)). p. 4 The paper finds that TSP−CDFT > TMR−CDFT for N⊙ < N 2t. p. 4 By breaking the interaction energy into nine terms, one can compute the Hamiltonian kernels of EC efficiently using the GCM kernels of the training EDFs. p. 4 Quantitatively, we find ∆T1/∆T2 ≃ 10 5 for Nd 150. p. 4
Quantification of Uncertainties
The paper uses a Bayesian method to derive the posterior distribution p(OD) for the quantity O given the data. The predictive distribution p(OC) is given by p(OC) = Z p(OC;σ(syst))p(σ(syst))dσ(syst), where the prior of the systematic error p(σ syst) is chosen as a delta function for observable with experimental data. p. 9 The covariance matrix Σij is obtained from the Pearson coefficient ρij by Σij = σiρijσj. p. 8
Key Findings on Nuclear Structure
The analysis shows that the B(E2: 0+1 → 2+1) values can exhibit either positive or negative correlations with the ground-state proton radius Rp depending on the nuclear structure. The propagated statistical uncertainties associated with the nine EDF parameters reach up to 21% for excitation energies and 12% for E2 transition strengths. p. 9 p. 9 After accounting for statistical uncertainties, the excitation energies and B(E2) values of deformed nuclei are well reproduced, whereas those of near-spherical nuclei remain challenging. p. 9
Future Directions
Future work will focus on improving the description of nearspherical systems within SP-CDFT and on further constraining nuclear EDFs using data on low-lying nuclear states.
Pruning redundant reasoning tokens
The paper presents a comprehensive formulation of the subspace-projected covariant density functional theory (SP-CDFT), a novel framework that combines the eigenvector continuation (EC) method with the quantumnumber projected generator coordinate method (PGCM) within covariant density functional theory (CDFT). We demonstrate that SP-CDFT provides an efficient and accurate emulator of multireference (MR)-CDFT for the description of nuclear low-lying states. p. 9
How the training signal is built
The training signal is built by sampling approximately one million parameter sets around the PC-PK1 functional. The parameter ranges for Latin hypercube sampling are selected based on the uncorrelated tolerance of parameters with χ squared ≤ χ squared min + 1. p. 4
Where the gains come from
The gains come from breaking the interaction energy (4) into nine terms to compute the Hamiltonian kernels of EC efficiently using the GCM kernels of the training EDFs. This method enables the execution of millions of SP-CDFT calculations for different parameter sets of EDFs with several orders of magnitude less computational time than MR-CDFT. p. 3 p. 3 The speed-up factor TMR−CDFT/TSP−CDFT increases almost linearly up to 10 4 when the number of sampling EDFs reaches to 10 6. p. 4
Summary of Results
The obtained p(DC) of the quantities D, including nuclear matter and B(E2), is used to constrain the parameter C with the Bayesian method. The main peaks of the posterior distribution p(CD) for most parameters derived from the B(E2) data of 136Xe are offset from their values in the PC-PK1. p. 9 p. 9 It is clear that the excitation energies of the deformed nuclei Nd 150 and Sm 150 are excellently reproduced. p. 9 p. 9
The training signal is built by sampling approximately one million parameter sets around the PC-PK1 functional.
The gains come from breaking the interaction energy (4) into nine terms to compute the Hamiltonian kernels of EC efficiently using the GCM kernels of the training EDFs. This method enables the execution of millions of SP-CDFT calculations for different parameter sets of EDFs with several orders of magnitude less computational time than MR-CDFT.
The obtained p(DC) of the quantities D, including nuclear matter and B(E2), is used to constrain the parameter C with the Bayesian method. The main peaks of the posterior distribution p(CD) for most parameters derived from the B(E2) data of 136Xe are offset from their values in the PC-PK1. It is clear that the excitation energies of the deformed nuclei Nd 150 and Sm 150 are excellently reproduced. The propagated statistical uncertainties associated with the nine EDF parameters reach up to 21% for excitation energies and 12% for E2 transition strengths. The results highlight the intrinsic limitations of the underlying EDF framework. p. 9
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, $5.
Improvements for AI systems
-
Bayesian uncertainty quantification for multireference covariant density functional theory allows AI systems to predict nuclear low-lying states with quantified statistical errors up to 21% for excitation energies and 12% for E2 transition strengths.
-
The subspace-projected covariant density functional theory approach SP-CDFT enables the emulation of computationally expensive multireference DFT calculations, such that "one would expect TSP-CDFT << TMR-CDFT when the number of target sets NC⊙ is sufficient large."
-
AI systems can be trained to constrain parameters of relativistic EDFs by using constraints from nuclear matter properties, as shown in the Bayesian analysis where
Constraints from nuclear-matter properties are employed to refine the parameter sets used in calculations for finite nuclei.
-
The system can perform correlation analysis on low-lying state observables, such as finding that "the B(E2: 0+1 -> 2+1) values can exhibit either positive or negative correlations with the ground-state proton radius Rp, depending on the nuclear structure."
Sources
- Toward ab initio density functional theory for nuclei
- The propagation of statistical errors in covariant density functional theory: ground state observables and single-particle properties
- Nuclear mass table in deformed relativistic Hartree-Bogoliubov theory in continuum, II: Even-$Z$ nuclei
- Recent Progress and New Challenges in Isospin Physics with Heavy-Ion Reactions
- Covariant energy density functionals: nuclear matter constraints and global ground state properties
- Error Estimates of Theoretical Models: a Guide
- Towards Density Functional Calculations from Nuclear Forces
- Toward a systematic strategy for defining power counting in the construction of the energy density functional theory
- Energy-density functionals inspired by effective-field theories: Applications to neutron drops
- Functional Renormalization Group and Kohn-Sham scheme in Density Functional Theory
- Application of an ab-initio-inspired energy density functional to nuclei: impact of the effective mass and the slope of the symmetry energy on bulk and surface properties
- Nuclear energy density functionals grounded in ab initio calculations
- Microscopically-based energy density functionals for nuclei using the density matrix expansion: Full optimization and validation
- Optimized nuclear energy density functionals including long-range pion contributions
- Towards an ab initio covariant density functional for nuclear structure
- Effective density functionals beyond mean field
- Turning the nuclear energy density functional method into a proper effective field theory: reflections
- Uncertainty Quantification for Nuclear Density Functional Theory and Information Content of New Measurements
- Global performance of covariant energy density functionals: ground state observables of even-even nuclei and the estimate of theoretical uncertainties
Related papers
- Constrained Hamiltonian Formulation of Kinetic Theory for Relativistic Spin- 1/2 Particles
- Active-subspace learning reveals dominant parameter directions in nuclear observables
- Long- and short-range longitudinal flow decorrelations in the quark-gluon plasma
- Recurrent Fourier Neural Operator for Event-by-Event Relativistic Hydrodynamics
- Bayesian Inference of the Landau Parameter G'0 from Joint Gamow-Teller Measurements
- UrQMD-Agent: An LLM-Assisted Framework for Microscopic Tracing of Proton Azimuthal Correlations