Pith. sign in

REVIEW 5 major objections 3 minor 76 references

Machine learning orbital-free density functional theory: taming quantum shell effects in deformed nuclei

T0 review · 5 major / 3 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A kernel-ridge-regression orbital-free functional reproduces the deformed ground state and potential energy curve of 20Ne.

desk verdict A credible proof-of-principle that an ML-trained orbital-free EDF can reproduce deformed minima in light nuclei; the main caveats are the narrow test set and the untested derivative. read the letter →

arxiv 2412.20739 v1 pith:6MPYEZID submitted 2024-12-30 nucl-th nucl-exquant-ph

X. H. Wu , Z. X. Ren , P. W. Zhao This is my paper · ORCID
classification nucl-thnucl-exquant-ph
keywords orbital-freedensityfunctionaltheorymachinelearningkernelridgeregressionnucleardeformationshelleffectsneon-20oxygen-16Hohenberg-Kohntheorem
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper aims to show that a fully orbital-free nuclear energy density functional—one that depends only on the density, with no auxiliary single-particle orbitals—can reproduce quantum shell and deformation effects in deformed nuclei. The authors train a kernel ridge regression model to map nucleon density onto the combined kinetic and spin-orbit energy, then add a Skyrme interaction term and minimize the total functional self-consistently. For deformed 20Ne the result matches Kohn-Sham deformation ($\beta_2=0.49$ vs 0.48), radius (3.05 vs 3.02 fm), and total energy (−156.02 vs −156.58 MeV), and the potential energy curve keeps both oblate and prolate minima. This is presented as the first orbital-free energy density functional to capture complex shell effects in deformed nuclei, turning the Hohenberg-Kohn theorem from a formal statement into a practical tool.

What carries the argument

The engine of the construction is the kernel ridge regression map from density to kinetic-plus-spin-orbit energy, $E^\mathrm{ML}_{\mathrm{kin+so}}[\rho]=\sum_i \omega_i K(\rho_i,\rho)$, where $K$ is a Gaussian kernel with width $\sigma$ and the weights solve $\boldsymbol{\omega}=(K+\lambda I)^{-1}\mathbf{E}$. Its functional derivative with respect to the density provides the gradient for self-consistent minimization, so the Kohn-Sham orbitals are bypassed entirely. The density is represented on an axial grid of 1128 discrete points, and the training densities are generated by solving Schrödinger equations with random spherical and quadrupole external potentials plus a spin-orbit potential built from a preliminary density. Together with the density-dependent Skyrme interaction (SkP), the learned term forms the total orbital-free functional, and constrained runs add an augmented-Lagrangian penalty to hold the quadrupole moment while tracing potential energy curves.

What would settle it

Train or reuse the same construction on a deformed nucleus outside the training pair, such as 24Mg, and compare the self-consistent $\beta_2$ and potential energy curve with Kohn-Sham results. A disagreement in the deformed minimum would show that the random-potential training family does not cover the densities that real nuclei produce.

Watch

Extended reading notes

Core claim

The central claim is that shell and deformation effects in nuclei can be described without ever constructing single-particle orbitals. The authors implement this by learning the density-to-energy map for the kinetic plus spin-orbit part, $E^\mathrm{ML}_{\mathrm{kin+so}}[\rho]=\sum_i \omega_i K(\rho_i,\rho)$, with a Gaussian kernel over discretized axial densities, while the interaction part is taken from the Skyrme functional SkP. The training data come from 24,000 Schrödinger equation solutions under randomly parameterized spherical and quadrupole mean fields, split into 20,000 training, 2,000 validation, and 2,000 test samples. Self-consistent minimization of the resulting functional for 20Ne yields a quadrupole deformation $\beta_2=0.49$ against the Kohn-Sham value 0.48, a root-mean-square radius 3.05 fm against 3.02 fm, and a total energy −156.02 MeV against −156.58 MeV; the constrained calculation reproduces the Kohn-Sham potential energy curve including the deformed minima, and the density profiles capture the spatial fluctuations associated with shell effects.

Load-bearing premise

The load-bearing premise is that densities produced by the randomly parameterized external potentials cover the density manifold of real self-consistent nuclei, so the learned kinetic-plus-spin-orbit map remains accurate when the functional is minimized with the Skyrme interaction; only 16O and 20Ne are tested.

Editorial extensions

If this is right

  • Nuclear orbital-free DFT can describe deformed ground states, not just spherical ones, without solving Kohn-Sham equations.
  • Potential energy curves and shape-isomeric minima become accessible from a density-only functional.
  • The learned functional captures the spatial density fluctuations that correspond to quantum shell effects, which Thomas-Fermi-type functionals miss.
  • The Hohenberg-Kohn theorem gains a practical nuclear implementation, opening quantitative orbital-free studies beyond spherical nuclei.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the functional transfers to neighboring nuclei, the same protocol could generate orbital-free mass or radius tables without Kohn-Sham iterations; testing that would require training over a larger mass range.
  • The result weakens the common argument that shell effects are intrinsically single-particle effects, since the learned functional captures them from density fluctuations alone.
  • A natural stress test is the region around $\beta_2=0$ in 20Ne, where the authors note Kohn-Sham struggles without pairing; an orbital-free description of that region could reveal whether the learned functional implicitly encodes pairing-like correlations.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 3 minor

Summary. This paper proposes to construct an orbital-free nuclear energy density functional by using kernel ridge regression (KRR) to learn the sum of kinetic and spin-orbit energies as a functional of the nucleon density. The KRR model is trained on 24,000 densities and energies generated by solving Schrödinger equations with randomly parameterized external potentials for A=16 and A=20, and is then minimized self-consistently to obtain ground states and constrained potential energy curves for 16O and 20Ne. The authors report that the resulting orbital-free approach reproduces the Kohn-Sham quadrupole deformations, rms radii, total energies, and density profiles, including the deformed minimum of 20Ne, and they claim in the abstract and summary that this is the first fully orbital-free EDF to account for shell effects in deformed nuclei.

Significance. The result, if confirmed by broader tests, is a significant proof-of-principle: it shows that an explicit orbital-free functional can describe a deformed nucleus without introducing Kohn-Sham orbitals, which has been a longstanding challenge. The paper's main strengths are the clear construction of the KRR map with a density-space kernel, the use of a large training set generated from random potentials, and the implementation of self-consistent gradient-descent minimization including constrained calculations with an augmented Lagrangian. The reported agreement with Kohn-Sham for the two test nuclei is encouraging. However, the present validation covers only two N=Z light nuclei, both with mass numbers contained in the training set, and the functional derivative is not independently benchmarked; these are important caveats that limit the strength of the central claim. The paper does not release code or data, only availability upon request, which further hampers independent verification.

major comments (5)
  1. [Eq. (S11), Fig. 1, Table 1] The self-consistent minimization in Section S3 uses the functional derivative of the KRR model, Eq. (S11), but the reported validation in Fig. 1 concerns rms errors of predicted energies on the validation and test sets, not errors of the derivative. The derivative is the quantity that drives the gradient descent in Eq. (S12) and determines the stationary density and the constrained potential energy curves. A small systematic bias in the derivative can displace the energy minimum even when the energy surface at the training densities is accurately reproduced. The standard deviations in Table 1 reflect sensitivity to 100 random initial densities, not the model error of the derivative at the self-consistent densities. I request an explicit benchmark of the functional derivative against the exact Kohn-Sham potential for the test-set densities, or an equivalent demonstration that derivative errors are small enough not to shift the ground-state density and PEC beyond the reported statistical uncertainties.
  2. [Methods S1, Table 1] The KRR functional is trained exclusively on densities generated for A=16 and A=20 (Methods S1), and the paper reports ground-state and PEC results only for these same two nuclei. Thus the statement in the abstract and summary that this is the first fully orbital-free EDF to tame shell effects in deformed nuclei is supported by only one deformed nucleus, 20Ne, and by interpolation within the trained mass window. To substantiate the claim of a general orbital-free functional, at least one deformed nucleus outside the training set (e.g., 24Mg or 28Si) should be tested without retraining, or the training set should be shown to include densities that are sufficiently diverse to make the A=16/A=20 restriction inconsequential. If the method requires retraining for each nucleus, the scope of the claim should be narrowed accordingly.
  3. [Paragraph after Fig. 2] The authors state that Kohn-Sham calculations cannot converge near β2=0 for 20Ne due to degeneracies, and that this phenomenon is 'accurately captured' by the ML orbital-free approach. Since no Kohn-Sham reference is available in this region, there is no benchmark for the ML PEC at β2=0. The agreement at the deformed minima is good, but the barrier height and the shape of the curve near β2=0 are predictions of the learned functional, not validated results. The paper should either provide a benchmark from a different method (e.g., a Hartree-Fock-Bogoliubov calculation with pairing) or present this region as a prediction with a clear caveat, rather than as part of the demonstrated accuracy.
  4. [Methods S1, Eq. (5)] The training densities are generated by random external potentials with rms radii restricted to an empirical band [0.8 A^{1/3}, 1.2 A^{1/3}] (Methods S1). The self-consistent densities of 16O and 20Ne under the Skyrme interaction are likely inside this band, but the paper does not quantify the proximity of the target densities to the training distribution. KRR is an interpolation method, and the reported accuracy may reflect that the target densities lie inside the training hull rather than a physical generalization. I ask for a quantitative analysis of the distance, in the kernel metric of Eq. (5), between the self-consistent densities and the training set, and for a discussion of how the results depend on the random potential parameter ranges.
  5. [Eq. (2), Eq. (3)] The learned functional EML_{kin+so}[ρ] represents the kinetic plus spin-orbit energy as a functional of the total density alone. In the Skyrme EDF, the spin-orbit energy depends on the spin-orbit density J (second line of Eq. (2)), which is not generally a unique functional of the total density for arbitrary spin-orbit potentials. The paper does not address this formal issue; if the ρ → (Ekin+Eso) map is not universal, the trained functional may not transfer to nuclei or densities with different spin-orbit structure. At minimum, a discussion of the conditions under which this map is well-defined (e.g., time-reversal invariant, even-even N=Z systems) is needed.
minor comments (3)
  1. [Eq. (5)] The distance ||ρ(r)-ρ'(r)|| in Eq. (5) is not defined explicitly; please state that it is the L2 norm over the discrete mesh and specify how the density vectorization is performed, as this is needed for reproducibility.
  2. [Section S3.2] The text mentions that c.m. correction energies and Coulomb energies are included, but does not specify the formulas or parameter values used; provide these details for reproducibility.
  3. [Abstract and Summary] The phrase 'inaugural instance' is used twice in prominent places; consider using a more conventional phrase such as 'first demonstration' for clarity and style.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the learned kinetic+spin-orbit functional is fitted to random-potential densities and tested against self-consistent Skyrme/Kohn-Sham densities, so the benchmark is external to the training labels.

full rationale

The central construction is a kernel ridge regression map from density to kinetic+spin-orbit energy, Eq. (3), with weights fixed by Eq. (4) using energies obtained from Schrödinger equations with randomly generated external potentials (Methods S1). The resulting orbital-free functional, Eq. (6), is then minimized self-consistently, and the ground-state properties and potential energy curves are compared with Kohn-Sham results. No claim in the paper reduces, by construction, to the training labels: the target densities are the minimizers of the fitted functional, not training densities, and the reported Kohn-Sham benchmark energies are not used as fitting targets. The two self-citations (Ref. 70, for the combined-Gaussian potential ansatz and the adaptive functional derivative stabilization method) supply numerical machinery, not the physical result, and the central comparison with Kohn-Sham is externally defined. The main limitations—two test nuclei, same A as training, no released code or data—are generalization concerns rather than circularity. Therefore no step satisfies the quota: there is no equation that is equivalent to its input by definition, no fitted parameter renamed as a prediction, and no load-bearing self-citation chain.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the existence of a density-only functional (HK theorem), on the choice of the Skyrme SkP interaction for the density-dependent part, and on the assumption that the KRR map learned from random external potentials generalizes to self-consistent nuclear densities. The only fitted numbers are the KRR hyperparameters λ and σ and the hand-chosen training potential ranges. No new physical entities are introduced.

free parameters (3)
  • KRR regularizer λ = 2.19e-6
    Chosen by optimizing validation set performance (main text).
  • KRR kernel width σ = 1.86 fm^-2
    Chosen by optimizing validation set performance (main text).
  • Mean-potential random parameter ranges = see Methods S1 (12 ranges)
    Hand-chosen ranges for the 12 Gaussian potential parameters; the learned functional's accuracy depends on the training density distribution (Methods S1).
assumptions (4)
  • domain assumption Hohenberg-Kohn theorem guarantees the total energy is a functional of density alone
    Invoked in the abstract and introduction to justify the orbital-free approach.
  • domain assumption The Skyrme SkP density-only interaction energy E'_int[ρ] can be combined with the learned kinetic+spin-orbit term to form the exact functional
    Used in Eq. (2) and Eq. (6); the interaction is taken from prior literature.
  • domain assumption The KRR model trained on Schrödinger solutions in random external potentials approximates the universal kinetic+spin-orbit density functional at the self-consistent nuclear densities
    Load-bearing generalization assumption; Methods S1 and Eqs. (3)-(5).
  • domain assumption Densities are axially symmetric, ρ(r⊥,z)
    Restricts the approach to axially deformed nuclei; stated in the Methods.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Machine learning orbital-free density functional theory: taming quantum shell effects in deformed nuclei." pith.science (2026). https://pith.science/paper/6MPYEZID

@misc{pith2026241220739,
  author       = {Pith},
  title        = {Pith review of: Machine learning orbital-free density functional theory: taming quantum shell effects in deformed nuclei},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6MPYEZID}},
  note         = {Machine review of arXiv:2412.20739}
}
abstract

Accurate description of deformed atomic nuclei by the orbital-free density functional theory has been a longstanding textbook challenge, due to the difficulty in accounting for the intricate quantum shell effects that are present in such systems. Orbital-free density functional theory is, in principle, capable of describing all effects of nuclear systems, as guaranteed by the Hohenberg-Kohn theorem. However, from a microscopic perspective, shell and deformation effects are believed to be intrinsically connected to single-orbital structures, posing a significant challenge for orbital-free approaches. Here, we develop a machine learning approach to the orbital-free density functional theory, which is capable of achieving a high level of accuracy in describing the ground-state properties and potential energy curves for both spherical $^{16}$O and deformed $^{20}$Ne nuclei. This is the inaugural instance where a fully orbital-free energy density functional has succeeded in taming the complex shell effects in deformed nuclei. It demonstrates that the orbital-free energy density functional, which is directly based on the Hohenberg-Kohn theorem, is not only a theoretical concept but also a practical one for nuclear systems.

Figures

Figures reproduced from arXiv: 2412.20739 by the authors.

Figure 1
Figure 1. The root-mean-square (rms) deviations ∆rms of the kinetic and spin-orbit energies between the KRR predicted results and the exact values in the validation and test sets. The quadrupole deformations β2 are in different slots, such as −0.45 < β2 < −0.35, −0.35 < β2 < −0.25, etc. Once the orbital-free energy density functional E ML tot [ρ] has been obtained, self-consistent procedures can be performed to find the densi… view at source ↗
Figure 2
Figure 2. Potential energy curves of 16O (a) and 20Ne (b) obtained with the machine learning orbital-free DFT, in comparison with the Kohn-Sham and (extended) Thomas-Fermi results. The stars denote the energy minima obtained by the machine learning orbital-free DFT. enhance the results, yet it consistently yields spherical energy minima due to the absence of deformed quantum shell effects. This provides the major challenge of… view at source ↗
Figure 3
Figure 3. Density profiles of the ground state (prolate) and the shape-isomer state (oblate) for 20Ne. The results given by the Kohn-Sham DFT and the machine learning orbital-free DFT are shown in the left and right panels, respectively. 6/14 [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

76 extracted references · 76 canonical work pages

  1. [2]

    & Kohn, W

    Hohenberg, P . & Kohn, W . Inhomogeneous electron gas. Phys. Rev. 136, B864–B871 (1964)

    work page 1964

  2. [3]

    & Sham, L

    Kohn, W . & Sham, L. J. Self-consistent equations including e xchange and correlation effects. Phys. Rev. 140, A1133–A1138 (1965)

    work page 1965

  3. [4]

    & Hakansson, H.-B

    Brack, M., Guet, C. & Hakansson, H.-B. Selfconsistent semiclassical description of average nuclear properties—a link between microscopic and macroscopic models. Phys. Rep. 123, 275–364 (1985)

    work page 1985

  4. [5]

    & Vi ˜ nas, X

    Centelles, M., Schuck, P . & Vi ˜ nas, X. Thomas–Fermi theory f or atomic nuclei revisited. Ann. Phys. 322, 363 (2007)

    work page 2007

  5. [6]

    & Hagino, K

    Col ` o, G. & Hagino, K. Orbital-free density functional theo ry: Differences and similarities between electronic and nuclear systems. Prog. Theor. Exp. Phys. 2023, 103D01 (2023)

    work page 2023

  6. [7]

    & Barranco, M

    Centelles, M., Pi, M., Vinas, X., Garcias, F . & Barranco, M. S elf-consistent extended thomas-fermi calculations in nuclei. Nucl. Phys. A 510, 397–416 (1990)

    work page 1990

  7. [8]

    & T onde ur, F

    Dutta, A., Arcoragi, J.-P ., Pearson, J., Behrman, R. & T onde ur, F . Thomas-fermi approach to nuclear mass formula: (i). spherical nuclei. Nucl. Phys. A 458, 77–94 (1986)

    work page 1986

  8. [9]

    & T ondeur, F

    Aboussir, Y ., Pearson, J., Dutta, A. & T ondeur, F . Nuclear ma ss formula via an approximation to the hartree—fock method. Atom. Data Nucl. Data T ables 61, 127 – 176 (1995)

    work page 1995

Show all 76 references
  1. [10]

    Brack, M. et al. Funny hills: The shell-correction approach to nuclear shel l effects and its applications to the fission process. Rev. Mod. Phys. 44, 320–405 (1972)

    1972

  2. [11]

    & Pauli, H

    Brack, M. & Pauli, H. On strutinsky’s averaging method. Nucl. Phys. A 207, 401–424 (1973)

    1973

  3. [12]

    Shell effects in nuclear masses and deformation energies

    Strutinsky , V . Shell effects in nuclear masses and deformation energies. Nucl. Phys. A 95, 420–442 (1967)

    1967

  4. [13]

    “shells” in deformed nuclei

    Strutinsky , V . “shells” in deformed nuclei. Nucl. Phys. A 122, 1–33 (1968)

    1968

  5. [14]

    & Quentin, P

    Brack, M. & Quentin, P . Self-consistent average density matrices and the strutinsky energy theorem. Phys. Lett. B 56, 421–423 (1975)

    1975

  6. [15]

    & Treiner, J

    Bohigas, O., Campi, X., Krivine, H. & Treiner, J. Extensions of the thomas-fermi approximation for finite nuclei. Phys. Lett. B 64, 381–385 (1976)

    1976

  7. [16]

    & Brack, M

    Chu, Y ., Jennings, B. & Brack, M. Nuclear binding energies and liquid drop parameters in the extended thomas- fermi approximation. Phys. Lett. B 68, 407–411 (1977)

    1977

  8. [17]

    Carleo, G. et al. Machine learning and the physical sciences. Rev. Mod. Phys. 91, 045002 (2019)

    2019

  9. [18]

    Boehnlein, A. et al. Colloquium: Machine learning in nuclear physics. Rev. Mod. Phys. 94, 031003 (2022)

    2022

  10. [19]

    He, W . et al. Machine learning in nuclear physics at low and intermediate energies. Sci. China-Phys. Mech. Astron. 66, 282001 (2023)

    2023

  11. [20]

    Niu, Z. M. et al. Radial basis function approach in nuclear mass predictions . Phys. Rev. C 88, 024325 (2013)

    2013

  12. [21]

    & Prosper, H

    Utama, R., Piekarewicz, J. & Prosper, H. B. Nuclear mass pred ictions for the crustal composition of neutron stars: A bayesian neural network approach. Phys. Rev. C 93, 014311 (2016)

    2016

  13. [22]

    Niu, Z. M. & Liang, H. Z. Nuclear mass predictions based on bay esian neural network approach with pairing and shell effects. Phys. Lett. B 778, 48 – 53 (2018)

    2018

  14. [23]

    C., Nazarewicz, W

    Neufcourt, L., Cao, Y . C., Nazarewicz, W . & Viens, F . Bayesia n approach to model-based extrapolation of nuclear observables. Phys. Rev. C 98, 034318 (2018)

    2018

  15. [24]

    & Barton, C

    Pastore, A., Neill, D., Powell, H., Medler, K. & Barton, C. Im pact of statistical uncertainties on the composition of the outer crust of a neutron star. Phys. Rev. C 101, 035804 (2020)

    2020

  16. [25]

    Wu, X. H. & Zhao, P . W . Predicting nuclear masses with the kern el ridge regression. Phys. Rev. C 101, 051301 (R) (2020)

    2020

  17. [26]

    H., Guo, L

    Wu, X. H., Guo, L. H. & Zhao, P . W . Nuclear masses in extended kernel ridge regression with odd-even effects. Phys. Lett. B 819, 136387 (2021). 11/14

    2021

  18. [27]

    H., Lu, Y

    Wu, X. H., Lu, Y . Y . & Zhao, P . W . Multi-task learning on nuclea r masses and separation energies with the kernel ridge regression. Phys. Lett. B 834, 137394 (2022)

    2022

  19. [28]

    Niu, Z. M. & Liang, H. Z. Nuclear mass predictions with machin e learning reaching the accuracy required by r-process studies. Phys. Rev. C 106, L021303 (2022)

    2022

  20. [29]

    Akkoyun, S., Bayram, T., Kara, S. O. & Sinan, A. An artificial n eural network application on nuclear charge radii. J. Phys. G-Nucl. Part. Phys. 40, 055106 (2013)

    2013

  21. [30]

    & Piekarewicz, J

    Utama, R., Chen, W .-C. & Piekarewicz, J. Nuclear charge radi i: density functional theory meets bayesian neural networks. J. Phys. G-Nucl. Part. Phys. 43 (2016)

    2016

  22. [31]

    Ma, Y . et al. Predictions of nuclear charge radii and physical interpret ations based on the naive bayesian probability classifier. Phys. Rev. C 101, 014304 (2020)

    2020

  23. [32]

    L., Sagawa, H

    Wu, D., Bai, C. L., Sagawa, H. & Zhang, H. Q. Calculation of nuc lear charge radii with a trained feed-forward neural network. Phys. Rev. C 102, 054323 (2020)

    2020

  24. [33]

    & Zhang, Z.-H

    Ma, J.-Q. & Zhang, Z.-H. Improved phenomenological nuclear charge radius formulae with kernel ridge regression. Chin. Phys. C 46, 074105 (2022)

    2022

  25. [34]

    & Geng, L.-S

    Dong, X.-X., An, R., Lu, J.-X. & Geng, L.-S. Nuclear charge ra dii in bayesian neural networks revisited. Phys. Lett. B 838, 137726 (2023)

    2023

  26. [35]

    M., Liang, H

    Niu, Z. M., Liang, H. Z., Sun, B. H., Long, W . H. & Niu, Y . F . Pred ictions of nuclear β-decay half-lives with machine learning and their impact on r-process nucleosynthesis. Phys. Rev. C 99, 064307 (2019)

    2019

  27. [36]

    E., Nunes, F

    Lovell, A. E., Nunes, F . M., Catacora-Rios, M. & King, G. B. Re cent advances in the quantification of uncertain- ties in reaction theory . J. Phys. G-Nucl. Part. Phys. 48, 014001 (2020)

    2020

  28. [37]

    L., Sagawa, H., Nishimura, S

    Wu, D., Bai, C. L., Sagawa, H., Nishimura, S. & Zhang, H. Q. β-delayed one-neutron emission probabilities within a neural network model. Phys. Rev. C 104, 054303 (2021)

    2021

  29. [38]

    Saxena, G., Sharma, P . K. & Saxena, P . Modified empirical form ulas and machine learning for al pha-decay systematics. J. Phys. G-Nucl. Part. Phys. 48, 055103 (2021)

    2021

  30. [39]

    Neudecker, D. et al. Informing nuclear physics via machine learning methods wit h differential and integral experiments. Phys. Rev. C 104, 034611 (2021)

    2021

  31. [40]

    W ang, X., Zhu, L. & Su, J. Modeling complex networks of nuclea r reaction data for probing their discovery processes. Chin. Phys. C 45, 124103 (2021)

    2021

  32. [41]

    X., Wu, X

    Huang, T. X., Wu, X. H. & Zhao, P . W . Application of kernel ridg e regression in predicting neutron-capture reaction cross-sections. Commun. Theor. Phys. 74, 095302 (2022)

    2022

  33. [42]

    G., Hagen, G

    Jiang, W . G., Hagen, G. & Papenbrock, T. Extrapolation of nuc lear structure observables with artificial neural networks. Phys. Rev. C 100, 054326 (2019)

    2019

  34. [43]

    & Penon, A

    Lasseri, R.-D., Regnier, D., Ebran, J.-P . & Penon, A. T amingnuclear complexity with a committee of multilayer neural networks. Phys. Rev. Lett. 124, 162502 (2020)

    2020

  35. [44]

    Nonparametric bayesian approach to extrapolation problems in configuration interaction methods

    Y oshida, S. Nonparametric bayesian approach to extrapolation problems in configuration interaction methods. Phys. Rev. C 102, 024305 (2020)

    2020

  36. [45]

    W ang, X., Zhu, L. & Su, J. Providing physics guidance in bayes ian neural networks from the input layer: The case of giant dipole resonance predictions. Phys. Rev. C 104, 034317 (2021)

    2021

  37. [46]

    & Niu, Y

    Bai, J., Niu, Z., Sun, B. & Niu, Y . The description of giant dip ole resonance key parameters with multitask neural networks. Phys. Lett. B 815, 136147 (2021)

    2021

  38. [47]

    & Viens, F

    Neufcourt, L., Cao, Y ., Nazarewicz, W ., Olsen, E. & Viens, F . Neutron drip line in the ca region from bayesian model averaging. Phys. Rev. Lett. 122, 062502 (2019)

    2019

  39. [48]

    Neufcourt, L. et al. Quantified limits of the nuclear landscape. Phys. Rev. C 101, 044307 (2020)

    2020

  40. [49]

    & Qiang, Y

    W ang, Z.-A., Pei, J., Liu, Y . & Qiang, Y . Bayesian evaluation of incomplete fission yields. Phys. Rev. Lett. 123, 122501 (2019)

    2019

  41. [50]

    E., Mohan, A

    Lovell, A. E., Mohan, A. T. & T alou, P . Quantifying uncertainties on fission fragment mass yields with mixture density networks. J. Phys. G-Nucl. Part. Phys. 47, 114001 (2020). 12/14

    2020

  42. [51]

    Qiao, C. Y . et al. Bayesian evaluation of charge yields of fission fragments of 239U. Phys. Rev. C 103, 034621 (2021)

    2021

  43. [52]

    & Rios, A

    Keeble, J. & Rios, A. Machine learning the deuteron. Phys. Lett. B 809, 135743 (2020)

    2020

  44. [53]

    & Rocco, N

    Adams, C., Carleo, G., Lovato, A. & Rocco, N. V ariational mon te carlo calculations of a ≤ 4 nuclei with an artificial neural-network correlator ansatz. Phys. Rev. Lett. 127, 022502 (2021)

    2021

  45. [54]

    & Rocco, N

    Lovato, A., Adams, C., Carleo, G. & Rocco, N. Hidden-nucleons neural-network quantum states for the nuclear many-body problem. Phys. Rev. Res. 4, 043178 (2022)

    2022

  46. [55]

    & Zhao, P

    Y ang, Y . & Zhao, P . A consistent description of the relativis tic effects and three-body interactions in atomic nuclei. Phys. Lett. B 835, 137587 (2022)

    2022

  47. [56]

    Y ang, Y . L. & Zhao, P . W . Deep-neural-network approach to sol ving the ab initio nuclear structure problem. Phys. Rev. C 107, 034320 (2023)

    2023

  48. [57]

    & Pederiva , F

    Rigo, M., Hall, B., Hjorth-Jensen, M., Lovato, A. & Pederiva , F . Solving the nuclear pairing model with neural network quantum states. Phys. Rev. E 107, 025310 (2023)

    2023

  49. [58]

    & Burke, K

    Pederson, R., Kalita, B. & Burke, K. Machine learning and den sity functional theory . Nature Reviews Physics 4, 357–358 (2022)

    2022

  50. [59]

    Huang, B., von Rudorff, G. F . & von Lilienfeld, O. A. The centr al role of density functional theory in the ai age. Science 381, 170–175 (2023)

    2023

  51. [60]

    C., Rupp, M., Hansen, K., M ¨ uller, K.-R

    Snyder, J. C., Rupp, M., Hansen, K., M ¨ uller, K.-R. & Burke, K . Finding density functionals with machine learning. Phys. Rev. Lett. 108, 253002 (2012)

    2012

  52. [61]

    By-passing the kohn-sham equations with machine learning

    Brockherde, F .et al. By-passing the kohn-sham equations with machine learning. Nat. Commun. 8, 872 (2017)

    2017

  53. [62]

    & Sugino, O

    Nagai, R., Akashi, R. & Sugino, O. Completing density functional theory by machine learning hidden messages from molecules. npj Computational Materials 6, 43 (2020)

    2020

  54. [63]

    E., M ¨ uller, K.-R

    Bogojeski, M., V ogt-Maranto, L., Tuckerman, M. E., M ¨ uller, K.-R. & Burke, K. Quantum chemical accuracy from density functional approximations via machine learni ng. Nat. Commun. 11, 5223 (2020)

    2020

  55. [64]

    R., Carleo, G

    Moreno, J. R., Carleo, G. & Georges, A. Deep learning the hohe nberg-kohn maps of density functional theory . Phys. Rev. Lett. 125, 076402 (2020)

    2020

  56. [65]

    Li, L. et al. Kohn-sham equations as regularizer: Building prior knowle dge into machine-learned physics. Phys. Rev. Lett. 126, 036401 (2021)

    2021

  57. [66]

    Kirkpatrick, J. et al. Pushing the frontiers of density functionals by solving the fractional electron problem. Science 374, 1385–1389 (2021)

    2021

  58. [67]

    Margraf, J. T. & Reuter, K. Pure non-local machine-learned d ensity functional theory for electron correlation. Nat. Commun. 12, 1–7 (2021)

    2021

  59. [68]

    Ma, H., Narayanaswamy , A., Riley , P . & Li, L. Evolving symbol ic density functionals. Science Advances 8, eabq0279 (2022)

    2022

  60. [69]

    Bai, Y ., V ogt-Maranto, L., Tuckerman, M. E. & Glover, W . J. Ma chine learning the hohenberg-kohn map for molecular excited states. Nat. commun. 13, 7044 (2022)

    2022

  61. [70]

    H., Ren, Z

    Wu, X. H., Ren, Z. X. & Zhao, P . W . Nuclear energy density funct ionals from machine learning. Phys. Rev. C 105, L031303 (2022)

    2022

  62. [71]

    & Y oshida, K

    Hizawa, N., Hagino, K. & Y oshida, K. Analysis of a skyrme ener gy density functional with deep learning. Phys. Rev. C 108, 034311 (2023)

    2023

  63. [72]

    Hartree-fock calculations with skyrme’s int eraction

    V autherin, D. Hartree-fock calculations with skyrme’s int eraction. ii. axially deformed nuclei. Phys. Rev. C 7, 296–316 (1973)

    1973

  64. [73]

    Lee, S.-J. et al. Relativistic hartree calculations for axially deformed nu clei. Phys. Rev. Lett. 57, 2916–2919 (1986). URL https://link.aps.org/doi/10.1103/PhysRevLett.57.2916

    1986 doi

  65. [74]

    & Boguta, J

    Pannert, W ., Ring, P . & Boguta, J. Relativistic mean-field th eory and nuclear deformation. Phys. Rev. Lett. 59, 2420–2422 (1987). URL https://link.aps.org/doi/10.1103/PhysRevLett.59.2420

    1987 doi

  66. [75]

    & Treiner, J

    Dobaczewski, J., Flocard, H. & Treiner, J. Hartree-fock-bo golyubov description of nuclei near the neutron-drip line. Nucl. Phys. A 422, 103–139 (1984). 13/14

    1984

  67. [76]

    National Nuclear Data Center (NNDC), https://www.nndc.bn l.gov/ URL https://www.nndc.bnl.gov/

  68. [77]

    & Nazarewicz, W

    Staszczak, A., Stoitsov , M., Baran, A. & Nazarewicz, W . Augmented lagrangian method for constrained nuclear density functional theory . Eur. Phys. J. A 46, 85–90 (2010). 14/14

    2010

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.