pith. sign in

arxiv: 2508.00173 · v1 · submitted 2025-07-31 · ✦ hep-ph · nucl-th

Variational Neural Network Approach to QFT in the Field Basis

Kevin Braga , Nobuo Sato , Adam P. Szczepaniak This is my paper

Pith reviewed 2026-05-19 01:21 UTC · model grok-4.3

classification ✦ hep-ph nucl-th
keywords variational neural networkquantum field theoryKlein-Gordon modelfield basismomentum spaceground-state wavefunctionalvariational methodbenchmark observables
0
0 comments X

The pith

A neural network variational ansatz reproduces the exact ground-state energy, correlators, and field values of the free Klein-Gordon theory in the momentum-space field basis.

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

The paper develops a variational method that represents the ground-state wavefunctional of a quantum field theory as a neural network defined directly on field configurations. For the free Klein-Gordon model in momentum space, the network is trained by minimizing the expectation value of the Hamiltonian on a finite set of discretized modes. When compared to known analytic results, it recovers the ground-state energy, two-point correlators, and field expectation value to high accuracy. This benchmark demonstrates that the momentum-space field basis is practical for testing neural-network approaches to quantum field theory and supplies a controlled starting point for extensions to interacting cases.

Core claim

We represent the ground-state wavefunctional as a neural network on a discretized set of momentum-space field configurations and train it by minimizing the Hamiltonian expectation value. For the free Klein-Gordon theory this ansatz reproduces the exact ground-state energy, two-point correlators, and field expectation value to high numerical accuracy while also recovering the structure of the learned wavefunctional itself.

What carries the argument

The neural-network variational ansatz for the wavefunctional, trained by direct minimization of the Hamiltonian expectation value on discretized field configurations.

If this is right

  • The approach supplies quantitative accuracy diagnostics for neural-network methods applied to exactly solvable free theories.
  • Momentum space is shown to be a suitable setting for benchmarking these methods before moving to position space.
  • The same training procedure provides a concrete foundation for extending the ansatz to interacting scalar theories.
  • Direct comparison to analytic observables becomes possible once the wavefunctional is represented explicitly in the field basis.

Where Pith is reading between the lines

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

  • Success on the free theory suggests the method could be tested on models where no exact solution exists, such as phi^4 theory in strong coupling.
  • The discretization scale in momentum space may need systematic extrapolation to the continuum limit in future applications.
  • The learned network weights could be inspected to extract approximate functional forms for the wavefunctional beyond simple Gaussian cases.

Load-bearing premise

A finite set of momentum modes with discretized field values is enough to capture the essential physics of the continuum theory without uncontrolled truncation errors.

What would settle it

Compute the neural-network ground-state energy on the same discretized momentum grid and observe a deviation from the exact analytic value that persists or grows when the number of modes or network capacity is increased.

Figures

Figures reproduced from arXiv: 2508.00173 by Adam P. Szczepaniak, Kevin Braga, Nobuo Sato.

Figure 1
Figure 1. Figure 1: FIG. 1 [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Training history of the neural network energy es [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Mean field configurations (left) and the Two-point correlation matrix [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. MC field configurations sampled from the trained NN-based wavefunctional (left), variance of the field configurations [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
read the original abstract

We present a variational neural network approach for solving quantum field theories in the field basis, focusing on the free Klein-Gordon model formulated in momentum space. While recent studies have explored neural-network-based variational methods for scalar field theory in position space, a systematic benchmark of the analytically solvable Klein-Gordon ground state -- particularly in the momentum-space field basis -- has been lacking. In this work, we represent the ground-state wavefunctional as a neural network defined on a discretized set of field configurations and train it by minimizing the Hamiltonian expectation value. This framework enables direct comparison to exact analytic results for a range of key observables, including the ground-state energy, two-point correlators, expectation value of the field, and the structure of the learned wavefunctional itself. Our results provide quantitative diagnostics of accuracy and demonstrate the suitability of momentum space for benchmarking neural network approaches, while establishing a foundation for future extensions to interacting models and position-space formulations.

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

1 major / 3 minor

Summary. The manuscript introduces a variational neural network ansatz for the ground-state wavefunctional of the free Klein-Gordon theory in the momentum-space field basis. The wavefunctional is represented by a neural network on a discretized set of field configurations at finite momentum modes; the network is trained by minimizing the expectation value of the Hamiltonian, and results for the ground-state energy, two-point correlators, field expectation value, and wavefunctional structure are compared directly to the corresponding closed-form analytic expressions on the same finite discretization.

Significance. If the reported agreement holds under the stated discretization, the work supplies a clean, controlled benchmark for neural-network variational methods in QFT. Momentum-space formulation simplifies the free theory (diagonal Hamiltonian, product Gaussian ground state), enabling unambiguous validation against analytics; this is a useful foundation for later extensions to interacting theories where such benchmarks are unavailable.

major comments (1)
  1. [§4] §4 (results section): the central claim of quantitative agreement with analytic results is load-bearing; the manuscript must explicitly confirm that every reported observable is evaluated on exactly the same finite set of momentum modes and field discretization used in the analytic expressions, and should include a brief convergence check with respect to the number of retained modes.
minor comments (3)
  1. [Abstract] Abstract: replace the generic phrase 'quantitative diagnostics of accuracy' with concrete figures (e.g., relative error on ground-state energy and correlator values) so readers can immediately gauge the achieved precision.
  2. [§2] Notation: define the precise form of the neural-network ansatz (number of layers, activation functions, input/output dimensions) in a dedicated subsection rather than only in the methods paragraph.
  3. [Figures] Figure captions: ensure every figure caption states the number of momentum modes, the field discretization grid, and the training hyperparameters used for that panel.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the careful reading of the manuscript and the positive assessment that this work provides a clean benchmark for neural-network variational methods in QFT. We address the single major comment below.

read point-by-point responses

  1. Referee: [§4] §4 (results section): the central claim of quantitative agreement with analytic results is load-bearing; the manuscript must explicitly confirm that every reported observable is evaluated on exactly the same finite set of momentum modes and field discretization used in the analytic expressions, and should include a brief convergence check with respect to the number of retained modes.

    Authors: We agree that this explicit confirmation strengthens the central claim. In the revised manuscript we will add a clear statement that the ground-state energy, two-point correlators, field expectation value, and wavefunctional structure are all evaluated on precisely the same finite set of momentum modes and field discretization employed for the analytic expressions. We will also include a brief convergence check (e.g., a table or plot) demonstrating stability of the reported observables as the number of retained modes is increased. revision: yes

Circularity Check

0 steps flagged

No significant circularity: variational benchmark against independent analytic results

full rationale

The paper defines a neural-network ansatz for the wavefunctional on a discretized momentum-space field basis, minimizes the Hamiltonian expectation value via variational optimization, and then evaluates observables (ground-state energy, two-point correlators, field expectation) by direct comparison to closed-form analytic expressions for the identical finite-mode free Klein-Gordon theory. This comparison uses the exact solvability of the discretized Hamiltonian (product of independent Gaussians) as an external benchmark, not as a fitted input or self-referential definition. No load-bearing self-citations, ansatzes smuggled via prior work, or predictions that reduce to the training data by construction are present. The derivation chain is self-contained and externally falsifiable against the analytic solution on the same truncation.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The approach rests on the assumption that the variational principle applied to a neural-network ansatz can recover the true ground state when the Hamiltonian is known exactly, plus standard assumptions of quantum mechanics for the free scalar field.

axioms (2)
  • standard math The ground state of the free Klein-Gordon Hamiltonian is the unique minimum of the energy functional.
    Invoked implicitly when the training objective is defined as minimization of the Hamiltonian expectation value.
  • domain assumption A finite set of momentum modes and discretized field values adequately approximates the continuum theory for the purpose of benchmarking.
    Required for the numerical representation of the wavefunctional on a computer.

pith-pipeline@v0.9.0 · 5690 in / 1458 out tokens · 26780 ms · 2026-05-19T01:21:29.684194+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

19 extracted references · 19 canonical work pages · 6 internal anchors

  1. [1]

    R. P. Feynman, in Variational Calculations in Quantum Field Theory: Proceedings of the International Work- shop, edited by L. Polley and D. E. L. Pottinger (World Scientific Publishing Co. Pte. Ltd., 1988) pp. 28–40

    work page 1988

  2. [2]

    Jackiw and A

    R. Jackiw and A. Kerman, Phys. Lett. A 71, 158 (1979)

    work page 1979

  3. [3]

    Coulomb string tension, asymptotic string tension, and the gluon chain

    J. Greensite and A. P. Szczepaniak, Phys. Rev. D 91, 034503 (2015), arXiv:1410.3525 [hep-lat]

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  4. [4]

    Testing Proposals for the Yang-Mills Vacuum Wavefunctional by Measurement of the Vacuum

    J. Greensite, H. Matevosyan, S. Olejnik, M. Quandt, H. Reinhardt, and A. P. Szczepaniak, Phys. Rev. D 83, 114509 (2011), arXiv:1102.3941 [hep-lat]

    work page internal anchor Pith review Pith/arXiv arXiv 2011

  5. [5]

    I. I. Kogan and A. Kovner, Phys. Rev. D 52, 3719 (1995), arXiv:hep-th/9408081

    work page internal anchor Pith review Pith/arXiv arXiv 1995

  6. [6]

    A. P. Szczepaniak, Phys. Rev. D 69, 074031 (2004), arXiv:hep-ph/0306030

    work page internal anchor Pith review Pith/arXiv arXiv 2004

  7. [7]

    Variational solution of the Yang-Mills Schr\"odinger equation in Coulomb gauge

    C. Feuchter and H. Reinhardt, Phys. Rev. D 70, 105021 (2004), arXiv:hep-th/0408236

    work page internal anchor Pith review Pith/arXiv arXiv 2004

  8. [8]

    On the Yang-Mills wave functional in Coulomb gauge

    H. Reinhardt and C. Feuchter, Phys. Rev. D 71, 105002 (2005), arXiv:hep-th/0408237

    work page internal anchor Pith review Pith/arXiv arXiv 2005

  9. [9]

    Carleo and M

    G. Carleo and M. Troyer, Science 355, 602 (2017)

    work page 2017

  10. [10]

    D. P. et al., Phys. Rev. Research 2, 033429 (2020)

    work page 2020

  11. [11]

    Hermann, Z

    J. Hermann, Z. Sch¨ atzle, and F. No´ e, Nat. Chem. 12, 891 (2020)

    work page 2020

  12. [12]

    J. Kim, G. Pescia, B. Fore, C. H. Greene, and B. K. Clark, Commun. Phys. 7, 148 (2024)

    work page 2024

  13. [13]

    Z. Chen, D. Luo, K. Hu, and B. K. Clark, (2022), arXiv:2212.06835 [hep-lat], arXiv:2212.06835 [hep-lat]

    work page arXiv 2022

  14. [14]

    J. M. Martyn, K. Najafi, and D. Luo, arXiv preprint arXiv:2212.00782 (2022), arXiv:2212.00782

    work page arXiv 2022

  15. [15]

    Rovira, A

    M. Rovira, A. Parre˜ no, and R. J. Perry, (2024), arXiv:2409.17887 [hep-lat], arXiv:2409.17887 [hep-lat]

    work page arXiv 2024

  16. [16]

    Verstraete and J

    F. Verstraete and J. I. Cirac, Phys. Rev. Lett. 104, 190405 (2010)

    work page 2010

  17. [17]

    J. H. et al., Phys. Rev. B 88, 085118 (2013)

    work page 2013

  18. [18]

    Shachar and E

    T. Shachar and E. Zohar, Phys. Rev. D 105, 045016 (2022)

    work page 2022

  19. [19]

    Hatfield, Quantum Field Theory of Point Particles and Strings, Frontiers in Physics (Westview Press, CRC Press, 1992)

    B. Hatfield, Quantum Field Theory of Point Particles and Strings, Frontiers in Physics (Westview Press, CRC Press, 1992)

    work page 1992