pith. sign in

arxiv: 2403.16819 · v2 · submitted 2024-03-25 · ⚛️ nucl-th · physics.comp-ph· quant-ph

A neural network approach for two-body systems with spin and isospin degrees of freedom

Chuanxin Wang , Tomoya Naito , Jian Li , Haozhao Liang This is my paper

Pith reviewed 2026-05-24 03:25 UTC · model grok-4.3

classification ⚛️ nucl-th physics.comp-phquant-ph
keywords neural networkunsupervised learningtwo-body systemsspinisospindeuteronnuclear ground state
0
0 comments X

The pith

A non-fully connected deep neural network with unsupervised learning calculates the ground state of two-body systems that include spin and isospin.

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

The paper extends an existing machine learning technique for two-body ground states to handle the additional spin and isospin quantum numbers that appear in nuclear physics. It does so by replacing a fully connected network with a non-fully connected architecture and by training the network in an unsupervised manner. The resulting procedure is tested on the deuteron, the only bound two-nucleon system, and is shown to recover its known ground-state energy. A sympathetic reader would care because traditional numerical methods become more cumbersome once spin and isospin degrees of freedom must be tracked explicitly.

Core claim

The present method enables consideration of the spin and isospin degrees of freedom by employing a non-fully connected deep neural network and the unsupervised machine learning technique. The validity of this method is verified by calculating the unique bound state of the deuteron.

What carries the argument

A non-fully connected deep neural network trained unsupervised to represent the spin-isospin dependent nucleon-nucleon interaction and to yield the correct ground-state energy.

If this is right

  • Spin and isospin degrees of freedom can be incorporated into neural-network solvers for two-body nuclear systems without requiring full connectivity between layers.
  • Unsupervised training suffices to obtain the correct bound-state energy once the network architecture respects the relevant symmetries.
  • The same procedure can be applied to any two-body system whose interaction depends on spin and isospin.
  • The deuteron remains the only bound two-nucleon state under the interactions considered.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The architecture could be tested on two-body scattering states or on systems with different interaction strengths to check whether the unsupervised procedure still converges.
  • If the method scales, it might be combined with larger networks to treat three- or four-body nuclei while retaining explicit spin-isospin dependence.
  • Comparison against conventional variational calculations on the same potentials would quantify any hidden bias introduced by the network topology.

Load-bearing premise

The non-fully connected network architecture and unsupervised training procedure can faithfully encode the spin and isospin dependence of the nucleon-nucleon force without introducing systematic biases.

What would settle it

A numerical result in which the network produces a deuteron ground-state energy measurably different from the experimental value of approximately 2.224 MeV would falsify the claim that the method is valid.

Figures

Figures reproduced from arXiv: 2403.16819 by Chuanxin Wang, Haozhao Liang, Jian Li, Tomoya Naito.

Figure 1
Figure 1. Figure 1: FIG. 1. Schematic figure of the non-fully-connected deep [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Deuteron wave functions with the AV18 and AV8 [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Relative error of [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Relative error of DNN deuteron wave functions to [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
read the original abstract

We propose an enhanced machine learning method to calculate the ground state of two-body systems. By extending the original method [Naito, Naito, and Hashimoto, Phys. Rev. Research 5, 033189 (2023)], the present method enables consideration of the spin and isospin degrees of freedom by employing a non-fully connected deep neural network and the unsupervised machine learning technique. The validity of this method is verified by calculating the unique bound state of the deuteron.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript extends the authors' prior neural-network method for two-body ground states to systems with spin and isospin by using a non-fully-connected deep neural network trained via unsupervised machine learning. Validity is asserted through reproduction of the deuteron's unique bound state.

Significance. If the numerical results hold, the work would demonstrate that an unsupervised, architecture-constrained DNN can incorporate the full spin-isospin structure of the nucleon-nucleon interaction without supervised labels, offering a potential route to few-body calculations that avoids explicit antisymmetrization or basis expansions. The unsupervised aspect and the non-fully-connected design are the principal technical novelties relative to the cited 2023 predecessor.

major comments (2)
  1. [Abstract] The abstract states that validity is verified by calculating the deuteron bound state, yet supplies neither the computed energy, its deviation from the known value (-2.224 MeV), nor any error estimate or hyperparameter list. Without these quantities the central claim cannot be assessed.
  2. [Method description (implicit)] Because the architecture and loss function are defined only by reference to the 2023 paper, any hidden bias introduced by the non-fully-connected layers or the unsupervised training procedure remains untested against an independent benchmark in the present manuscript.
minor comments (1)
  1. Notation for the spin-isospin operators and the precise form of the non-fully-connected connections should be defined explicitly rather than assumed from the prior reference.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading of our manuscript and the constructive comments. We respond to each major comment below.

read point-by-point responses

  1. Referee: [Abstract] The abstract states that validity is verified by calculating the deuteron bound state, yet supplies neither the computed energy, its deviation from the known value (-2.224 MeV), nor any error estimate or hyperparameter list. Without these quantities the central claim cannot be assessed.

    Authors: We agree that the abstract would be strengthened by the inclusion of these quantitative details. In the revised version we will report the computed deuteron binding energy, its deviation from -2.224 MeV, an associated error estimate, and a pointer to the hyperparameter list. revision: yes

  2. Referee: [Method description (implicit)] Because the architecture and loss function are defined only by reference to the 2023 paper, any hidden bias introduced by the non-fully-connected layers or the unsupervised training procedure remains untested against an independent benchmark in the present manuscript.

    Authors: The manuscript focuses on the extension of the 2023 method to spin-isospin degrees of freedom through the non-fully-connected architecture; the deuteron calculation constitutes the benchmark for that extension. To address the concern we will insert a concise recapitulation of the key architectural and loss-function elements in the revised text. revision: partial

Circularity Check

1 steps flagged

Minor self-citation to base method; new extension verified independently

specific steps
  1. self citation load bearing [Abstract]
    "By extending the original method [Naito, Naito, and Hashimoto, Phys. Rev. Research 5, 033189 (2023)], the present method enables consideration of the spin and isospin degrees of freedom by employing a non-fully connected deep neural network and the unsupervised machine learning technique. The validity of this method is verified by calculating the unique bound state of the deuteron."

    The foundational architecture and unsupervised training technique are justified solely by citation to prior work by overlapping authors; the new spin-isospin handling inherits its grounding from that self-citation even though the deuteron test is performed here.

full rationale

The paper extends a prior method by the same research group (overlapping author Naito) but explicitly verifies the spin-isospin extension by reproducing the known deuteron bound state. This supplies independent content outside the self-citation. No self-definitional reductions, fitted inputs renamed as predictions, or load-bearing uniqueness theorems appear in the provided text. The self-citation is normal and not the sole justification for the central claim.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; no explicit free parameters, axioms, or invented entities are stated.

pith-pipeline@v0.9.0 · 5614 in / 958 out tokens · 59596 ms · 2026-05-24T03:25:31.182354+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Finite-nuclear-size effect for hydrogenlike ions under high external pressure

    physics.atom-ph 2026-03 unverdicted novelty 5.0

    High pressure markedly increases finite-nuclear-size corrections and electron-capture decay rates in confined hydrogenlike ions while removing level degeneracies.

Reference graph

Works this paper leans on

39 extracted references · 39 canonical work pages · cited by 1 Pith paper · 1 internal anchor

  1. [1]

    —” in the column “# of unit in layers

    quite nicely. The only region where small differences are visible is near the origin (r ≲ 0.05 fm). The rela- tive error uDNN (r) − ubenchmark (r) /ubenchmark (r) and wDNN (r) − wbenchmark (r) /wbenchmark (r) are shown in Fig. 3, whereu and w are the wave functions of3S1 and 3D1 states, respectively. The main origin of the error is due to the3D1 state bec...

    work page 2063

  2. [2]

    Hohenberg and W

    P. Hohenberg and W. Kohn, Inhomogeneous Electron Gas, Phys. Rev.136, B864 (1964)

    work page 1964

  3. [3]

    Kohn and L

    W. Kohn and L. J. Sham, Self-Consistent Equations In- cluding Exchange and Correlation Effects, Phys. Rev. 140, A1133 (1965)

    work page 1965

  4. [4]

    Kohn, Nobel Lecture: Electronic structure of matter—wave functions and density functionals, Rev

    W. Kohn, Nobel Lecture: Electronic structure of matter—wave functions and density functionals, Rev. Mod. Phys. 71, 1253 (1999)

    work page 1999

  5. [5]

    S. R. White, Density matrix formulation for quantum renormalizationgroups,Phys.Rev.Lett. 69,2863(1992)

    work page 1992

  6. [6]

    Shinaoka, D

    H. Shinaoka, D. Geffroy, M. Wallerberger, J. Otsuki, K. Yoshimi, E. Gull, and J. Kuneš, Sparse sampling andtensornetworkrepresentationoftwo-particleGreen’s functions, SciPost Phys.8, 012 (2020)

    work page 2020

  7. [7]

    W. L. McMillan, Ground State of LiquidHe4, Phys. Rev. 138, A442 (1965)

    work page 1965

  8. [8]

    Ceperley, G

    D. Ceperley, G. V. Chester, and M. H. Kalos, Monte Carlo simulation of a many-fermion study, Phys. Rev. B 16, 3081 (1977)

    work page 1977

  9. [9]

    D. M. Ceperley and B. J. Alder, Ground State of the Electron Gas by a Stochastic Method, Phys. Rev. Lett. 45, 566 (1980)

    work page 1980

  10. [10]

    Carleo and M

    G. Carleo and M. Troyer, Solving the quantum many- body problem with artificial neural networks, Science 355, 602 (2017)

    work page 2017

  11. [11]

    Nomura, A

    Y. Nomura, A. S. Darmawan, Y. Yamaji, and M. Imada, Restricted Boltzmann machine learning for solving strongly correlated quantum systems, Phys. Rev. B96, 205152 (2017)

    work page 2017

  12. [12]

    Carleo, Y

    G. Carleo, Y. Nomura, and M. Imada, Constructing ex- act representations of quantum many-body systems with deep neural networks, Nat. Commun.9, 5322 (2018)

    work page 2018

  13. [13]

    Stokes, J

    J. Stokes, J. R. Moreno, E. A. Pnevmatikakis, and G. Carleo, Phases of two-dimensional spinless lattice fermions with first-quantized deep neural-network quan- tum states, Phys. Rev. B102, 205122 (2020)

    work page 2020

  14. [14]

    K. Choo, G. Carleo, N. Regnault, and T. Neupert, Symmetries and Many-Body Excitations with Neural- Network Quantum States, Phys. Rev. Lett.121, 167204 (2018)

    work page 2018

  15. [15]

    Nomura, Machine Learning Quantum States — Extensions to Fermion–Boson Coupled Systems and Excited-State Calculations, J

    Y. Nomura, Machine Learning Quantum States — Extensions to Fermion–Boson Coupled Systems and Excited-State Calculations, J. Phys. Soc. Jpn.89, 054706 (2020)

    work page 2020

  16. [16]

    Yoshino, Spatially heterogeneous learning by a deep student machine, Phys

    H. Yoshino, Spatially heterogeneous learning by a deep student machine, Phys. Rev. Research5, 033068 (2023)

    work page 2023

  17. [17]

    J. W. T. Keeble and A. Rios, Machine learning the deuteron, Phys. Lett. B809, 135743 (2020)

    work page 2020

  18. [18]

    Saito, Method to Solve Quantum Few-Body Problems with Artificial Neural Networks, J

    H. Saito, Method to Solve Quantum Few-Body Problems with Artificial Neural Networks, J. Phys. Soc. Jpn.87, 074002 (2018)

    work page 2018

  19. [19]

    Pescia, J

    G. Pescia, J. Han, A. Lovato, J. Lu, and G. Carleo, Neural-network quantum states for periodic systems in continuous space, Phys. Rev. Research4, 023138 (2022)

    work page 2022

  20. [20]

    D. Pfau, J. S. Spencer, A. G. D. G. Matthews, and W. M. C. Foulkes,Ab initio solution of the many-electron Schrödinger equation with deep neural networks, Phys. Rev. Research 2, 033429 (2020)

    work page 2020

  21. [21]

    Hermann, Z

    J. Hermann, Z. Schätzle, and F. Noé, Deep-neural- network solution of the electronic Schrödinger equation, Nat. Chem. 12, 891 (2020)

    work page 2020

  22. [22]

    Adams, G

    C. Adams, G. Carleo, A. Lovato, and N. Rocco, Varia- tional Monte Carlo Calculations ofA ≤ 4 Nuclei with an Artificial Neural-Network Correlator Ansatz, Phys. Rev. Lett. 127, 022502 (2021)

    work page 2021

  23. [23]

    Gnech, C

    A. Gnech, C. Adams, N. Brawand, G. Carleo, A. Lovato, andN.Rocco,NucleiwithUpto A = 6 NucleonswithAr- tificial Neural Network Wave Functions, Few-Body Syst. 63, 7 (2022)

    work page 2022

  24. [24]

    Y. L. Yang and P. W. Zhao, Deep-neural-network ap- proach to solving theab initio nuclear structure problem, Phys. Rev. C107, 034320 (2023)

    work page 2023

  25. [25]

    J. R. Moreno, G. Carleo, A. Georges, and J. Stokes, Fermionic wave functions from neural-network con- strained hidden states, Proc. Natl. Acad. Sci. USA119, e2122059119 (2022)

    work page 2022

  26. [26]

    Lovato, C

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

    work page 2022

  27. [27]

    K. Choo, A. Mezzacapo, and G. Carleo, Fermionic neural-network states for ab-initio electronic structure, Nat. Commun. 11, 2368 (2020)

    work page 2020

  28. [28]

    Hermann, J

    J. Hermann, J. Spencer, K. Choo, A. Mezzacapo, W. M. C. Foulkes, D. Pfau, G. Carleo, and F. Noé, Ab initio quantum chemistry with neural-network wavefunc- tions, Nat. Rev. Chem.7, 692 (2023)

    work page 2023

  29. [29]

    Naito, H

    T. Naito, H. Naito, and K. Hashimoto, Multi-body wave function of ground and low-lying excited states using un- ornamented deep neural networks, Phys. Rev. Research 5, 033189 (2023)

    work page 2023

  30. [30]

    Hornik, M

    K. Hornik, M. Stinchcombe, and H. White, Multilayer feedforward networks are universal approximators, Neu- ral Netw. 2, 359 (1989). 8

    work page 1989

  31. [31]

    Hornik, Approximation capabilities of multilayer feed- forward networks, Neural Netw.4, 251 (1991)

    K. Hornik, Approximation capabilities of multilayer feed- forward networks, Neural Netw.4, 251 (1991)

    work page 1991

  32. [32]

    R. B. Wiringa, V. G. J. Stoks, and R. Schiavilla, Accu- rate nucleon-nucleon potential with charge-independence breaking, Phys. Rev. C51, 38 (1995)

    work page 1995

  33. [33]

    B. S. Pudliner, V. R. Pandharipande, J. Carlson, S. C. Pieper, and R. B. Wiringa, Quantum Monte Carlo cal- culations of nuclei with A ≤ 7, Phys. Rev. C 56, 1720 (1997)

    work page 1997

  34. [34]

    R. B. Wiringa and S. C. Pieper, Evolution of Nuclear Spectra with Nuclear Forces, Phys. Rev. Lett.89, 182501 (2002)

    work page 2002

  35. [35]

    D. P. Kingma and J. Ba, Adam: A Method for Stochastic Optimization, in 3rd International Conference on Learn- ing Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings , edited by Y. Bengio and Y. LeCun (2015) arXiv:1412.6980 [cs.LG]

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  36. [36]

    Abadi, A

    M. Abadi, A. Agarwal, P. Barham, E. Brevdo, Z. Chen, C. Citro, G. S. Corrado, A. Davis, J. Dean, M. Devin, S. Ghemawat, I. Goodfellow, A. Harp, G. Irving, M. Is- ard, Y. Jia, R. Jozefowicz, L. Kaiser, M. Kudlur, J. Lev- enberg, D. Mané, R. Monga, S. Moore, D. Murray, C. Olah, M. Schuster, J. Shlens, B. Steiner, I. Sutskever, K. Talwar, P. Tucker, V. Vanho...

    work page 2015

  37. [37]

    C.A.Bertulani, Nuclear Physics in a Nutshell (Princeton University Press, Princeton, New Jersey, 2007)

    work page 2007

  38. [38]

    C. Iung, F. Gatti, A. Viel, and X. Chapuisat, Vector parametrization of theN-atom problem in quantum me- chanics with non-orthogonal coordinates, Phys. Chem. Chem. Phys. 1, 3377 (1999)

    work page 1999

  39. [39]

    Machleidt, High-precision, charge-dependent Bonn nucleon-nucleon potential, Phys

    R. Machleidt, High-precision, charge-dependent Bonn nucleon-nucleon potential, Phys. Rev. C 63, 024001 (2001)

    work page 2001