Pith. sign in

REVIEW 4 major objections 4 minor 47 references

A radial-basis-function expansion turns interacting Euclidean path integrals into closed-form per-mode products and reproduces the φ⁴ phase transition line in 1+1 dimensions.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-04 15:43 UTC pith:UHU4THGM

load-bearing objection A genuinely new route to factorizing Euclidean path integrals, but the load-bearing approximation is tested only in a regime far from the one used for the phi^4 claims; worth refereeing, not yet believable as is. the 4 major comments →

arxiv 2509.18785 v3 pith:UHU4THGM submitted 2025-09-23 hep-ph hep-lat

Neural network expansion of Euclidean path integrals and its application to interacting scalar fields

Gabor Balassa This is my paper
classification hep-ph hep-lat MSC 81T2781T8068T07
keywords Euclidean path integralradial basis function networkphi-four theoryphase transitionlattice field theoryneural network approximation1+1 dimensionseffective potential
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper proposes a way to make interacting Euclidean path integrals analytically tractable: approximate the nonlinear part of the discretized action at each lattice site by a sum of Gaussian (radial basis function) kernels. After diagonalizing the kinetic term by a discrete Fourier transform, the author argues that with kernel centers symmetric about zero the enormous sum over all kernel combinations collapses into a product of independent per-momentum-mode Gaussian integrals, with only a few percent error in the log-partition function. This factorized form yields closed-form expressions for the partition function, propagators, fluctuations, and vacuum expectation values, turning lattice-scale calculations that normally take hours into computations that take seconds. Applied to the 1+1-dimensional φ⁴ theory, the method produces renormalized masses that grow monotonically with coupling and a phase transition line extracted from the zero of the effective-potential fit that agrees well with previous lattice Monte Carlo results.

Core claim

The central discovery is the factorized momentum-space path integral of Eq. 35: by choosing symmetric centers and a common width for the RBF kernels, the full sum over K^(N_t N_x) kernel combinations can be replaced by a product over momentum modes of sums of K Gaussians, reducing the computational complexity from exponential to O(K N) with a few-percent error in ln Z. From this factorized form, all standard observables—the two-point correlator, total fluctuations, and field expectation value—are available in closed form, and the paper shows that the free-field correlator is reproduced almost exactly. For the interacting theory, the renormalized mass increases monotonically with bare couplin

What carries the argument

Radial basis function (RBF) expansion of the local interaction factor F[φ_ij] = exp(-a² L_I) as Σ_k a_k exp(-A(φ - c_k)²) with a common width A. After diagonalizing the kinetic matrix M by discrete Fourier transform, the key step is replacing the Fourier-transformed center vector U^T c_k with the untransformed c_k, which makes the full sum over kernel combinations factorize into a product over momentum modes of K one-dimensional Gaussian integrals (Eqs. 28, 34, 35). This reduces the complexity from K^(N_t N_x) to O(K N) and yields closed-form expressions for correlators, fluctuations, and vacuum expectation values.

Load-bearing premise

The entire calculation rests on replacing the Fourier-transformed RBF center vector U^T c_k by the untransformed c_k so that the sum over kernel combinations factorizes; this is justified only numerically, for ln Z on small lattices with equal kernel weights, and would invalidate all results if it fails for fitted weights or for observables.

What would settle it

Compute the full non-factorized sum over kernel combinations for a φ⁴ lattice using the actual fitted, sign-alternating weights used in Sec. 4 (or even on a smaller lattice with those weights), and compare ln Z, the momentum-space two-point correlator, and the derived phase-transition crossing against the factorized results; a relative error in ln Z above the reported few percent, or a visible mismatch in the correlator or mass, would falsify the central claim.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • The partition function, two-point correlators, total fluctuations ⟨φ²⟩, and vacuum expectation value ⟨φ⟩ all have closed-form per-mode expressions, so lattice-size calculations that take hours or days in Monte Carlo run in seconds.
  • The free scalar field correlator matches the exact modified Bessel function result for several RBF parameterizations (Fig. 8), supporting the internal consistency of the factorized approximation.
  • The renormalized mass in the unbroken phase increases monotonically with coupling for bare masses (am₀)² = 1 and 4, matching the expected behavior from previous lattice and perturbative studies (Fig. 9).
  • The b = 0 crossing of the effective-potential fit J(⟨φ⟩) = a⟨φ⟩³ + b⟨φ⟩ gives a phase transition line in the (λ, m²) plane that agrees well with lattice Monte Carlo data over a wide coupling range (Fig. 12).
  • If the factorization holds, the method extends in principle to higher dimensions, finite temperature, and finite density (through complex actions) without the critical slowing down that limits standard Monte Carlo.

Where Pith is reading between the lines

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

  • Editorial extension: The paper's numerical justification for the factorization is limited to ln Z on 10×10 lattices with unit kernel weights; the extrapolation to 100×100 lattices with fitted, sign-alternating weights, and to observables such as correlators and masses, is an unproven step that a direct comparison against the full non-factorized sum would settle.
  • Editorial extension: If the factorization survives with fitted weights, the same closed-form machinery should reproduce the full momentum-space propagator at all momenta, not just the p² ≈ 0 fit used here; a direct test would be to compare the RBF propagator shape against lattice Monte Carlo propagators at intermediate momenta.
  • Editorial extension: Extending the method to finite density, gauge fields, or fermions would require new treatment of kinetic terms and complex actions; the paper's speculation about QCD-like theories at finite density is forward-looking rather than demonstrated.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper proposes an RBF-neural-network expansion of the non-quadratic part of a Euclidean lattice path integral. In 1+1 dimensions the interaction factor F[φ] is approximated by a sum of Gaussian kernels with common width A, and the resulting sum over K^{N} quadratic subsystems is approximated by a factorized product over momentum modes (Eqs. 27–28, 34). This yields closed-form expressions for the partition function, two-point correlators, fluctuations, and vacuum expectation values. The method is applied to the lattice φ^4 theory on a 100×100 lattice: renormalized masses in the unbroken phase are extracted from momentum-space propagators (Sec. 4.2), and the phase transition line is read off from the sign change of the coefficient b in a cubic-linear fit to J(⟨φ⟩) (Eq. 65, Fig. 12). The free-field propagator is checked against the exact 1/(2π)K₁(mr) correlator (Fig. 8), and the phase line is compared to two Monte Carlo determinations. The central technical claim is that the factorization error is only a few percent in ln Z for symmetric RBF centers.

Significance. If the central factorization step were rigorously established, the paper would offer a strikingly fast approximate method for nonperturbative scalar field theory, with closed-form propagators and order parameters that run in seconds on a laptop. The paper has two genuinely creditable checks: the free-field correlator matches the exact Bessel-function result, and the extracted phase boundary agrees visually with independent Monte Carlo data. However, the paper does not provide a derivation of the key factorization; its numerical support covers only unit RBF weights, a 10×10 lattice, and the partition function. The applications use fitted sign-alternating weights, a 100×100 lattice, and J-dependent observables, which are outside the tested regime. The favorable benchmark results make the approach worth further scrutiny, but the central claim is not yet supported at the level required for publication.

major comments (4)
  1. [Sec. 3.1, Eqs. (27)–(28)/(34)] The factorization replaces Uᵀ ĉ_k by ĉ_k. The numerical justification (Figs. 3–6, Eq. (31)) is restricted to a_k=1, symmetric centers, N_t=N_x=10, and the relative error in ln Z only. The φ^4 applications use fitted a_k that are sign-alternating and O(10–100) (Fig. 7), N_t=N_x=100, field scaling S_c, and observables obtained as derivatives of ln Z with respect to J (Eqs. 62–66) or as ratios of one- and two-point integrals (Eqs. 39–46). A few-percent error in ln Z does not control the error in ∂ln Z/∂J or in the fitted coefficient b of Eq. (65), especially near the b=0 crossing. Please test Z_0 versus Z_1 directly for the J-dependent generating function and for the extracted ⟨φ⟩_J and correlators using the actual fitted weights and the 100×100 lattice, and quantify the resulting shift in the phase boundary.
  2. [Sec. 4.3, after Eq. (63)] The treatment of the constant source is not derived. The text states that if J is not included in F(φ,J), one may make the replacement (2Ac_k)φ̃ → (2Ac_k+J)φ̃ in Eq. (34). But a spatially constant source in coordinate space couples to the zero momentum mode, not to every momentum mode, after Fourier transformation. As written, this step changes the theory being solved. A derivation (or at least a numerical demonstration that the b=0 crossing is invariant under the implied rescaling of J) is required before the J(⟨φ⟩) curves can be used for the effective potential.
  3. [Eqs. (17), (25), (26), (34), (35)] The sign conventions in the Gaussian exponent are inconsistent. Eq. (17) writes Z = ∫Dφ exp(1/2 φᵀMφ)···, and Eq. (24) defines λ_{ij} ≥ 0. Eq. (25) then has exp(1/2 φ̃ᵀ[λ−2A]φ̃ + ···), while Eq. (26) and the final factorized form Eq. (35) use denominators (λ_{ij}+2A), which correspond to a convergent integral with exponent −1/2(λ+2A)φ̃² + ···. Taken literally, the printed quadratic form is not negative-definite when λ>2A, and the Gaussian integrals do not converge. Please correct the signs consistently throughout the derivation so that the intermediate expressions match the closed-form results.
  4. [Sec. 4.3, Fig. 12] The phase transition line is the main quantitative output, but the comparison with Refs. [44,45] is visual only. The text states that uncertainties are estimated from the covariance of the b fit, yet no error bars or residual measures appear in Fig. 12. Please show uncertainty bands on the RBF phase line and provide a numerical goodness-of-fit or residual comparison with the Monte Carlo data points, so that 'very good agreement' can be assessed quantitatively.
minor comments (4)
  1. [Eq. (41)] The first moment integral omits the factor 2Ac_k/(λ+2A) that appears in Eq. (46). As written, Eq. (41) does not vanish for symmetric centers with c_k=0 and is inconsistent with the expression used for ⟨φ̃⟩.
  2. [Sec. 3.1, Fig. 2] The text says the test uses A=3, while the caption of Fig. 2 says A=5. Please reconcile.
  3. [References] Some reference names appear garbled: [10] 'Z. Hanada et al.' and [14] 'U. Schmidhuber' are likely incorrect; please verify all citation data against the original sources.
  4. [Title/abstract] The arXiv metadata title uses 'Neural network expansion', while the manuscript title is 'Neural network approximation of Euclidean path integrals...'. Please make these consistent.

Circularity Check

0 steps flagged

No significant circularity: RBF weights are fitted to the input interaction factor, and the phase line and free-field correlator are checked against external benchmarks.

full rationale

The paper's main computation is not circular. The RBF weights a_k are fitted by least squares to F(phi), which is the known interaction factor of the input Lagrangian, not to the target observables; the correlators, masses, and phase transition line are then computed forward from the resulting closed-form expressions. The phase boundary is determined from the b=0 crossing of a fit to J(<phi>) data generated by the model, and is compared to independent lattice Monte Carlo results from Refs. [44,45]; the free-field correlator is compared to the exact Bessel-function result. These are external checks, not equivalences to inputs. The central approximation—replacing the transformed RBF centers U^T c_k by c_k in passing from Eq. 27 to the factorized form Eq. 28/34—is explicitly presented as a numerical approximation and is validated only for ln Z with unit weights on small lattices. That is a genuine validation-scope limitation and a correctness concern, but not a circular reduction: the factorized form is not defined in terms of the later observables, and the observables are not fitted to the validation data. Self-citations [16] and [23] provide background and are not load-bearing for the new results. The paper also clearly states its acknowledged limitations (no continuum limit, no finite-size scaling), which do not constitute circularity. Overall, no step reduces by construction or by self-citation to the quantity it claims to predict, so the circularity score is 0.

Axiom & Free-Parameter Ledger

5 free parameters · 5 axioms · 0 invented entities

The method is not parameter-free: it fits RBF weights, hand-tunes widths, centers, kernel counts, and field scales, and fits an effective-potential Ansatz whose zero crossing defines the phase boundary. The costliest assumption is the 'good parametrization' axiom (dropping U^T c_k), which is the actual content of the method and is bought with a small, idealized numerical experiment. No new physical entities are postulated; the Gaussian kernels are mathematical basis functions, not new physics.

free parameters (5)
  • RBF output weights a_k = not tabulated for phi^4 runs
    Least-squares fit to F(phi) at every (m0, lambda0) point; results are averaged over several fits, so the weights are data-dependent degrees of freedom.
  • RBF width A (common b_k) = 0.8, 3, 4 in examples; O(1) guidance
    Hand-chosen; enters every closed-form observable (Eqs. 35, 39) and the error estimate of Figs. 3-6.
  • RBF centers c_k and kernel number K = 8 kernels on [-0.4,0.4] etc.; unspecified for phi^4
    Hand-chosen intervals constrained to be symmetric around zero; K and spacing are part of the parametrization ensemble.
  • Field scaling S_c = 1, 3, 4 in free-field tests; per-point values not reported
    Chosen to remap F onto a favorable approximation interval; rescales eigenvalues and the measure (Eq. 36).
  • Effective potential fit coefficients a and b (Eq. 65) = fitted from J(<phi>) data; transition at b=0
    The phase boundary is defined by the sign change of the fitted b coefficient, so the fit parameters determine the central result.
axioms (5)
  • domain assumption RBF networks with fixed width A and symmetric centers can approximate the relevant interaction factors F(phi) accurately enough that path-integral results converge across parametrizations.
    Shown for one oscillatory test function (Fig. 2) and for the free-field Gaussian (Fig. 7); asserted as general for the phi^4 runs and for higher dimensions (Sec. 3.1).
  • ad hoc to paper The Fourier mixing of the RBF centers, U^T c_k, can be replaced by c_k for symmetric centers, with a few percent error in ln Z.
    Load-bearing: it is the step that turns the intractable K^(N_t N_x) sum into the factorized product (Eqs. 28, 34). Validated only numerically, with a_k = 1, on a 10x10 lattice, for ln Z (Figs. 3-6), not for the fitted weights, N=100 lattices, or the observables used in Sec. 4.
  • domain assumption The effective potential of the lattice phi^4 theory has quartic form V_eff = A<phi>^2 + B<phi>^4, so J(<phi>) = a<phi>^3 + b<phi>, with the phase transition at b = 0.
    Standard minimal ansatz (Eqs. 64-65); the paper neglects higher-order and logarithmic terms and relies on the fit to locate the transition.
  • standard math The discrete Laplace matrix is circulant and diagonalized by the discrete Fourier transform with eigenvalues lambda_ij = 4 sin^2(pi n_i/N_t) + 4 sin^2(pi n_j/N_x).
    Invoked at Eq. 24 under periodic boundary conditions; standard lattice spectral theory.
  • domain assumption The cited Monte Carlo phase lines ([44], [45]) are valid direct benchmarks for the a=1, N=100 regularization used here.
    The paper compares its finite-a, finite-N results directly to these MC results without finite-size scaling or continuum extrapolation, and without stating the lattice parameters of the references.

pith-pipeline@v1.3.0-alltime-deepseek · 25531 in / 37366 out tokens · 262088 ms · 2026-08-04T15:43:21.487643+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of Neural network expansion of Euclidean path integrals and its application to interacting scalar fields." pith.science (2026). https://pith.science/paper/UHU4THGM

@misc{pith2026250918785,
  author       = {Pith},
  title        = {Pith review of: Neural network expansion of Euclidean path integrals and its application to interacting scalar fields},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UHU4THGM}},
  note         = {Machine review of arXiv:2509.18785}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Studying phase transitions in interacting quantum field theories generally requires the numerical study of the dynamical system on a large lattice, which is, in most cases, computationally very challenging. In this work an alternative method is proposed to solve Euclidean path integrals in quantum field theories, using radial basis function-type neural networks. The method allows us to approximate observables in a very efficient manner, taking only seconds to do calculations that would otherwise take hours or even days with other existing methods. The model is used to describe phase transitions in the scalar $\phi^4$ theory for a wide range of coupling strength. The obtained phase transition line is compared to previous lattice results, giving very good agreement between them.

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

47 extracted references · 1 canonical work pages

  1. [1]

    Wilczek, Phys

    F. Wilczek, Phys. Rev. Lett.,30, 1343–1346 (1973), arXiv:hep-th/0307346

    arXiv 1973

  2. [2]

    Stevenson, Phys

    M. Stevenson, Phys. Rev. D,109, 014005 (2024), arXiv:2409.01228

    Pith/arXiv arXiv 2024

  3. [3]

    Aoki et al., Eur

    S. Aoki et al., Eur. Phys. J. A,49, 123 (2013), arXiv:1309.4200

    Pith/arXiv arXiv 2013

  4. [4]

    Bazavov et al., Phys

    A. Bazavov et al., Phys. Rev. D,87, 054505 (2013), arXiv:1212.4768

    Pith/arXiv arXiv 2013

  5. [5]

    Aoki et al., Nature,443, 675–678 (2006), arXiv:hep-lat/0611014

    Y. Aoki et al., Nature,443, 675–678 (2006), arXiv:hep-lat/0611014

    Pith/arXiv arXiv 2006

  6. [6]

    Bazavov et al., Phys

    A. Bazavov et al., Phys. Rev. D,98, 074512 (2018), arXiv:1712.09262

    Pith/arXiv arXiv 2018

  7. [7]

    Iritani et al., Phys

    T. Iritani et al., Phys. Rev. D,99, 014514 (2019), arXiv:1805.04150

    Pith/arXiv arXiv 2019

  8. [8]

    Lepage, Phys

    P. Lepage, Phys. Rev. D,59, 074502 (1999), arXiv:hep-lat/9809157

    Pith/arXiv arXiv 1999

  9. [9]

    Nagata, Prog

    K. Nagata, Prog. Part. Nucl. Phys.,123, 103991 (2022), arXiv:2108.12423

    Pith/arXiv arXiv 2022

  10. [10]

    Hanada et al., Phys

    Z. Hanada et al., Phys. Rev. D,86, 074510 (2012), arXiv:1207.1915

    Pith/arXiv arXiv 2012

  11. [11]

    Alexandru et al., Rev

    AA. Alexandru et al., Rev. Mod. Phys.,94, 015006 (2022), arXiv:2106.05934

    Pith/arXiv arXiv 2022

  12. [12]

    V. A. Goy et al., Prog. Theor. Exp. Phys., 2017(3), 031D01 (2017)

    2017

  13. [13]

    LeCun, Y

    Y. LeCun, Y. Bengio, and G. Hinton, Nature,521, 436–444 (2015)

    2015

  14. [14]

    Schmidhuber, Neural Networks,61, 85–117 (2015)

    U. Schmidhuber, Neural Networks,61, 85–117 (2015)

    2015

  15. [15]

    Y. Che, C. Gneiting, F. Nori, Phys. Rev. B.105, 21 (2022). http://dx.doi.org/10.1103/PhysRevB.105.214205

    work page doi:10.1103/physrevb.105.214205 2022

  16. [16]

    Balassa, ’On the solution of Euclidean path integrals with neural networks’ (2025), arXiv:2509.16953 [hep-ph]

    G. Balassa, ’On the solution of Euclidean path integrals with neural networks’ (2025), arXiv:2509.16953 [hep-ph]

    arXiv 2025

  17. [17]

    Zinn-Justin, Quantum Field Theory and Critical Phenomena, Int

    J. Zinn-Justin, Quantum Field Theory and Critical Phenomena, Int. Ser. Monogr. Phys., Oxford University Press, 4th ed. (2002)

    2002

  18. [18]

    M. J. D. Powell, Comput. Appl. Math.,123, 145 (2000)

    2000

  19. [19]

    R. P. Feynman, Rev. Mod. Phys.,20, 367 (1948)

    1948

  20. [20]

    Mosel, Path Integrals in Field Theory: An Introduction, Springer-Verlag, Berlin Heidelberg, Advanced Texts in Physics, (2004), ISBN: 978-3-540-21056-9

    U. Mosel, Path Integrals in Field Theory: An Introduction, Springer-Verlag, Berlin Heidelberg, Advanced Texts in Physics, (2004), ISBN: 978-3-540-21056-9

    2004

  21. [21]

    N. V. Quan, M. T. Long, Electrical Engineering,111, 1123 (2025)

    2025

  22. [22]

    Liao, S.-C

    Y. Liao, S.-C. Fang, and H. L. Chen, IEEE/ACM Trans. Comput. Biol. Bioinform.,8, 187 (2011)

    2011

  23. [23]

    Balassa, Prog

    G. Balassa, Prog. Theor. Exp. Phys.,2023, 113A01 (2023)

    2023

  24. [24]

    M. H. Hassoun and M. A. Ali, IEEE Trans. Neural Netw.,10, 524 (1999)

    1999

  25. [25]

    Schaefer and E

    T. Schaefer and E. V. Shuryak, Rev. Mod. Phys.,70, 323 (1998)

    1998

  26. [26]

    M. E. Peskin and D. V. Schroeder, An Introduction to Quantum Field Theory, Addison-Wesley, Reading, MA, (1995), ISBN: 978-0201503975

    1995

  27. [27]

    J. M. Pawlowski, I.-O. Stamatescu, F. P. G. Ziegler, Phys. Rev. D,96, 114505 (2017)

    2017

  28. [28]

    Itzykson, J.-B

    C. Itzykson, J.-B. Zuber, Quantum Field Theory, McGraw-Hill, New York, (1980), ISBN: 978-0070320710

    1980

  29. [29]

    Srednicki, Quantum Field Theory, Cambridge University Press, 2007, ISBN: 978-0521864497

    M. Srednicki, Quantum Field Theory, Cambridge University Press, 2007, ISBN: 978-0521864497

    2007

  30. [30]

    S.-K. Jian, E. Barnes, S. D. Sarma, Phys. Rev. Res.,2, 023310 (2020), 38

    2020

  31. [31]

    J. R. Klauder, C. B. Lang, P. Salomonson, B.-S. Skagerstam, Z. Phys. C,26, 149 (1984)

    1984

  32. [32]

    W. D. McComb, Renormalization Methods: A Guide for Beginners, Oxford University Press, 1st Edition (2008), ISBN: 9780199236527

    2008

  33. [33]

    A. K. De, A. Harindranath, J. Maiti, T. Sinha, Phys. Rev. D,72, 094503 (2005)

    2005

  34. [34]

    F. K. Diakonos, Y. F. Contoyiannis, S. M. Potirakis, Prog. Theor. Exp. Phys.,2021, 043A01 (2021)

    2021

  35. [35]

    Goldenfeld, Lectures on Phase Transitions and the Renormalization Group, CRC Press (1992)

    N. Goldenfeld, Lectures on Phase Transitions and the Renormalization Group, CRC Press (1992)

    1992

  36. [36]

    Akiyama, Y

    S. Akiyama, Y. Kuramashi, Y. Yoshimura, Phys. Rev. D,104, 034507 (2021)

    2021

  37. [37]

    Inagaki, S

    T. Inagaki, S. Nojiri, S. D. Odintsov, JCAP, 0506, 010 (2005)

    2005

  38. [38]

    Alexandre, Phys

    J. Alexandre, Phys. Rev. D,86, 025028 (2012)

    2012

  39. [39]

    R. W. Haymaker, J. Perez-Mercader, Phys. Rev. D,27, 1948 (1983)

    1948

  40. [40]

    Tabata, I

    K. Tabata, I. Umemura, Prog. Theor. Phys.,74, 1360 (1985)

    1985

  41. [41]

    D. J. E. Callaway, D. J. Maloof, Phys. Rev. D,27, 406 (1983)

    1983

  42. [42]

    Straumann, Cosmological Phase Transitions, arXiv:astro-ph/0409042 (2004)

    N. Straumann, Cosmological Phase Transitions, arXiv:astro-ph/0409042 (2004)

    Pith/arXiv arXiv 2004

  43. [43]

    Ryder, Quantum Field Theory, Cambridge University Press, Cambridge, 1985

    Lewis H. Ryder, Quantum Field Theory, Cambridge University Press, Cambridge, 1985. ISBN: 9780521237642

    1985

  44. [44]

    Loinaz, R

    W. Loinaz, R. S. Willey, Phys. Rev. D,58, 076003 (1998)

    1998

  45. [45]

    Ardekani, A

    A. Ardekani, A. G. Williams, Austral. J. Phys.52, 929 (1999)

    1999

  46. [46]

    Binder, Z

    K. Binder, Z. Phys. B,43, 119 (1981)

    1981

  47. [47]

    Milchev, D

    A. Milchev, D. W. Heermann, K. Binder, J. Stat. Phys.,44, 749 (1986). 39

    1986