Pith. sign in

REVIEW 2 major objections 5 minor 1 cited by

Winding number on 3D lattice

T0 review · 2 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A tree-level improved discretization plus an over-improved gradient flow reproduces integer winding numbers for maps from a 3D torus to SU(2) even on coarse lattices.

desk verdict A clean, honest numerical methods paper: tree-level improved winding number plus over-improved gradient flow works on the tested one-parameter family, but generality is asserted more than demonstrated. read the letter →

arxiv 2412.03888 v2 pith:GZGGVP43 submitted 2024-12-05 hep-lat cond-mat.mes-hall

Okuto Morikawa , Hiroshi Suzuki This is my paper · ORCID
classification hep-latcond-mat.mes-hall MSC 81T2581T13 PACS 11.15.Ha
keywords windingnumber3Dtoruslatticediscretizationgradientflowover-improvedactionSU(2)mapstopologicalchargeChern-Simonstheory
topics Yang-Mills Mass Gap
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 proposes a numerical method for computing the winding number of a map from the 3D torus to the unitary group when the map is known only on lattice points. The authors claim that a tree-level improved discretization of the standard continuum formula, combined with a gradient flow generated by an over-improved lattice action, gives accurate integer winding numbers even on coarse lattices and in the presence of random noise. They demonstrate this on a one-parameter family of maps from $T^{3}$ to SU(2) with known winding numbers 0, 1, and -2, on lattices as small as L=10. If correct, the method offers a practical way to extract topological data from discrete configurations without diagonalizing or interpolating group elements, and it extends to higher-dimensional tori.

What carries the argument

The central object is the combination $H(x,\mu)$, a 'tree-level improved' finite-difference replacement for $g^{-1}\partial_\mu g$ that includes a tunable parameter $\eta$; the improved discretization of the winding number is $W_3^{\rm lat} = \frac{1}{4^3\,24\pi^2}\sum_{x,\mu,\nu,\rho}\epsilon_{\mu\nu\rho}\mathrm{tr}[H(x,\mu)H(x,\nu)H(x,\rho)]$. The gradient flow is defined by $\partial_t g(t,x) = -g(t,x)\,\partial^a_x S_{\rm lat}T^a$ with $S_{\rm lat}=-\frac{1}{16}\sum_{x,\mu}\mathrm{tr}H(x,\mu)^2$, and is integrated by a Runge-Kutta scheme. The load-bearing mechanism is the dependence of $S_{\rm lat}$ on $\eta$: for negative, over-improved $\eta$ the action as a function of the parameter $m$ acquires local minima inside each topological sector (Fig. 5), so the flow settles into the correct sector instead of sliding to the trivial map, while for $\eta=0,1$ the action has no such minima and the flow erases the topological information.

What would settle it

Take a different family of maps from $T^3$ to $SU(2)$ with known winding numbers, add noise comparable to the paper's, and check whether the over-improved flow with $\eta=-20$ on an $L=10$ lattice produces a plateau of $W_3^{\rm lat}$ at the correct integer over flow times $t\simeq2$–$40$; if the plateau drifts, fails to reach an integer, or jumps between sectors, the claimed stabilization is not generic. A cheaper check is to compute $S_{\rm lat}$ along a two-parameter family and see whether sectors with nonzero winding number lack a local minimum.

Watch

Extended reading notes

Core claim

The paper claims that the continuum winding number $W_3=\frac{1}{24\pi^2}\int_{T^3}\mathrm{tr}(g^{-1}dg)^3$ can be approximated on a cubic lattice by replacing $g^{-1}\partial_\mu g$ with an improved difference operator $H(x,\mu)$ (Eq. (2.2)) containing a parameter $\eta$, and summing $\epsilon_{\mu\nu\rho}\mathrm{tr}(HHH)$ over the lattice (Eq. (2.3)). Setting $\eta=1$ removes the leading lattice-discretization error and already reproduces $W_3=0,1,-2$ accurately for $L\gtrsim20$. For coarse or noisy configurations, flowing $g(x)$ by the gradient equation $\partial_t g = -g\,\partial^a_x S_{\rm lat}T^a$ with the over-improved action $S_{\rm lat}=-\frac{1}{16}\sum_{x,\mu}\mathrm{tr}H(x,\mu)^2$ (using negative $\eta$ such as $-20$) stabilizes the lattice winding number around the correct integer; the paper shows that for negative $\eta$ the lattice action develops local minima in each topological sector, pinning the flow. The method requires neither diagonalization of $g(x)$ nor interpolation of the map.

Load-bearing premise

The method's reliability rests on the untested premise that the single one-parameter family tested is representative: for sufficiently negative $\eta$, the lattice action develops local minima in every topological sector for generic coarse or noisy configurations, not just for this family.

Editorial extensions

If this is right

  • With tree-level improvement ($\eta=1$), the discretized winding number alone reaches essentially exact integers already at $L\simeq20$ for the tested family.
  • For coarse lattices ($L=10$) with added random noise, an over-improved flow ($\eta=-10$ or $-20$) holds the lattice winding number near the correct integer over a range of flow times; for example, $W_3^{\rm lat}=-1.96797(4)$ at $t=2$ for $m=3$.
  • Standard or tree-level flow ($\eta=0,1$) drives configurations toward the trivial sector, so the over-improved choice is essential for stabilization.
  • The same construction works for $U(N)$ and extends trivially to higher-dimensional tori, with light computational cost.

Reading between the lines

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

  • The mechanism suggests a practical tuning rule: choose $\eta$ negative enough that the lattice action has local minima in every sector of interest for the given lattice size; the required $\eta$ likely depends on $L$ and on the roughness of the configurations.
  • Because the stabilization comes from discretization error, the method's over-improved flow is complementary to standard improvement: on fine lattices the improved discretization suffices, while on coarse lattices the flow provides the stability, potentially making the pair useful for Monte Carlo samples with noisy topological charge.
  • One testable extension, not pursued in the paper, is to apply the method to configurations with multiple spatially separated topological structures, where pinning to a single integer is harder and the action landscape would need local minima in each sector separately.
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

2 major / 5 minor

Summary. The paper proposes a numerical method for approximating the winding number of a map from T^3 to U(N) when the map is only known on a cubic lattice. The method combines a tree-level improved lattice discretization of the W3 functional, Eqs. (2.2)-(2.3), with gradient-flow smearing generated by an over-improved lattice action, Eqs. (3.4)-(3.9). Using a one-parameter family of SU(2) maps with known winding numbers, the authors show that the eta=1 discretization converges quickly (Tables 1-2), that gradient flow with negative eta stabilizes W_lat against random perturbations (Figs. 1-2), and that the lattice action restricted to the family develops local minima in the corresponding sectors for negative eta (Fig. 5). The paper also explains the stabilization mechanism in terms of discretization errors in the flow action and notes that the method avoids diagonalization or interpolation of g(x).

Significance. If the results generalize beyond the tested family, the method is a useful and computationally light tool for lattice computations of Chern-Simons-type invariants. The paper is transparent in an exemplary way: the benchmark uses analytically known winding numbers, the code and numerical data are publicly available on GitHub, and the authors explicitly disclose that the choice of eta and flow time is exploratory rather than derived from first principles. The tree-level improvement is shown to reduce discretization errors by large factors on the tested family, and the over-improved flow provides clear numerical stabilization in the shown cases, which are genuine assets of the manuscript.

major comments (2)
  1. [Sec. 3, Figs. 1-2 and 5; footnote 9] The load-bearing claim is that over-improved gradient flow stabilizes the lattice winding number for generic coarse or noisy configurations. The evidence, however, is restricted to the one-parameter family (2.4): Fig. 5 evaluates S_lat only along that slice, so it establishes local minima on a one-dimensional subspace of a high-dimensional configuration space, not the size of the basins of attraction. Footnote 9 explicitly concedes that a configuration could slip away from the family under flow. The noisy tests in Figs. 1 and 2 are also perturbations of the same family and cover only m=-1, 1, and 3. Because W_lat in Eq. (2.3) is not integer-valued on the lattice, the flow can cross between sectors, and stabilization is an empirical property of the chosen eta and t rather than a theorem. I would ask for at least one additional family of maps with known winding number, or a quantitative basin-size diagnostic, before the method is advertised as general; alternatively, the claims should be narrowed explicitly to the tested class.
  2. [Secs. 3-4] The method requires two tuning parameters, the over-improvement coefficient eta and the flow time t, and no concrete criterion for choosing them is provided. Figures 1 and 2 are shown for eta=-10 and eta=-20 and for t up to 40 or 80, with t=2 used in Fig. 6, but nowhere is there a rule such as a plateau-time selection or an automatic stopping condition. The conclusion states that real applications require exploratory tuning of eta. Since the central practical promise is a 'simple and versatile' method, a concrete protocol for choosing (eta, t), or at least a sharper statement of how sensitive the result is to these choices, is needed for the method to be reproducible in new settings.
minor comments (5)
  1. [Eq. (3.6)] The displayed equation d_t g(t,x) = -sum_x (d^a_x S_lat)(d^a_x S_lat) is not what is meant; the context shows that the derivative of the action should be d_t S_lat = -sum_x (d^a_x S_lat)^2 <= 0. Please correct this typo, since the monotonic decrease of S_lat is used in the subsequent explanation.
  2. [Abstract and Sec. 1] The abstract and introduction state that the method is 'simple and versatile' and can be generalized to higher-dimensional tori, but the numerical support is only for d=3 and for the SU(2) family (2.4). If no further tests are added, please qualify these statements, for example by saying 'for the class of maps studied here'.
  3. [Sec. 1, Refs. [5-8]] No comparison with the existing lattice winding-number algorithms of Refs. [5-8] is made. A brief numerical or conceptual comparison would help the reader assess the claimed advantage of the present method.
  4. [Figs. 1-2] The 15 trajectories per m value are plotted as separate lines, but no legend connects line style to m. A legend or distinctive line styles would improve readability.
  5. [Eq. (2.6)] The piecewise winding-number formula for the family (2.4) is stated without derivation or citation; please add a reference or a short justification.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the method is validated against analytically known external winding numbers, and the disclosed tuning of the improvement parameter is not disguised as a prediction.

full rationale

The paper's central claim is an empirical numerical proposal, not a derivation from first principles. The discretized winding number W_lat (Eq. 2.3) is a direct lattice transcription of the continuum integral (Eq. 1.1), with a tree-level improvement parameter eta in Eq. (2.2). The gradient flow (Eqs. 3.4-3.9) is a standard lattice transcription of the continuum flow (Eqs. 3.1-3.3). No equation is defined in terms of the target quantity in a way that makes the result true by construction: W_lat is not integer-valued, and the paper explicitly notes in footnote 9 that the configuration could in principle slip away from the tested one-parameter family, so the stabilization is an observed property rather than an imported theorem. The benchmark uses the family (2.4) whose exact winding numbers (2.6) are known analytically and are external to the paper's numerical scheme. The only adjustable inputs are the improvement parameter eta and the flow time t; their selection is disclosed, and the conclusion explicitly states that real applications require exploratory tuning of eta. This is a generality limitation, not circularity. Citations to prior work, including one article co-authored by H. Suzuki, are used for context or standard analytic results and do not form a self-citation chain that forces the paper's conclusion. The paper is therefore self-contained against an external benchmark and contains no circular step.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The method introduces no new physical entities. It does rely on a tuned improvement coefficient and flow time, plus several standard background results in topology and lattice field theory. The main ad hoc element is the assumption that strongly negative eta creates sector-stabilizing minima of the lattice action.

free parameters (2)
  • over-improvement coefficient eta in gradient flow action = -10 and -20 chosen
    Negative eta in Eq. (2.2)/(3.7) is selected because it stabilizes W_lat for the test mapping in Figs 1-2; no first-principles criterion is given. The authors state that real applications require exploratory determination of eta.
  • flow time t = e.g., t=2 or 10
    The stopping time for the gradient flow is chosen by visual inspection of Figs 1-2 and 6; no automated convergence criterion is provided.
assumptions (4)
  • standard math The continuum winding number W3 (1.1) is well defined and integer-valued for smooth maps g: T^3 -> U(N).
    Standard algebraic topology; used as the target quantity throughout.
  • domain assumption The known winding numbers (2.6) for the one-parameter family (2.4) are correct.
    Used as ground truth for the numerical tests; the paper cites Refs [15-17] for this family but does not re-derive (2.6).
  • domain assumption The lattice action Slat (3.7) monotonically decreases along the lattice flow, and the continuum flow preserves the topological sector.
    Invoked in Sec. 3 to justify the stabilization mechanism; the lattice version's sector preservation is assumed to hold sufficiently for negative eta.
  • ad hoc to paper For sufficiently negative eta, the lattice action has local minima in each topological sector.
    This is a numerical observation from Fig. 5, not proven; it is the mechanism behind the claimed stabilization and is assumed to generalize beyond the tested family.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Winding number on 3D lattice." pith.science (2026). https://pith.science/paper/GZGGVP43

@misc{pith2026241203888,
  author       = {Pith},
  title        = {Pith review of: Winding number on 3D lattice},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GZGGVP43}},
  note         = {Machine review of arXiv:2412.03888}
}
abstract

We propose a simple numerical method which computes an approximate value of the winding number of a mapping from 3D torus~$T^3$ to the unitary group~$U(N)$, when $T^3$ is approximated by discrete lattice points. Our method consists of a ``tree-level improved'' discretization of the winding number and the gradient flow associated with an ``over-improved'' lattice action. By employing a one-parameter family of mappings from $T^3$ to $SU(2)$ with known winding numbers, we demonstrate that the method works quite well even for coarse lattices, reproducing integer winding numbers in a good accuracy. Our method can trivially be generalized to the case of higher-dimensional tori.

Figures

Figures reproduced from arXiv: 2412.03888 by the authors.

Figure 1
Figure 1. Wlat 3 (2.3) as the function of the flow time t. The flow is defined by Eq. (3.8) with various values of η, η = 0 (no improvement), η = 1 (tree-level improvement), η = −10, and η = −20 (“over-improved”), respectively. We set the initial configuration g(t = 0, x) for the flow by Eq. (2.4) with ξA(θ) → ξA(θ) + εrA(θ), where ε and rA(θ) are uniform random numbers, ε ∈ [0, 1] and rA(θ) ∈ [−1/2, 1/2]. We take 15 random i… view at source ↗
Figure 2
Figure 2. Same as Fig. 1 but the lattice size is [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. 2D distribution of the action density −(1/16)P µ tr H(x, y, L/2, µ) 2 as the function of the flow time t. L = 10 and the initial configuration g(t = 0, x) is Eq. (2.4) with m = 1. The case of the gradient flow with the parameter η = 1 in Eq. (2.2) is shown. 10 [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Same as Fig. 3 but the case of the gradient flow with the parameter [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: The value of the lattice action (3.7) of the mapping (2.4) as a function of the param [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]
Figure 6
Figure 6. Figure 6: Wlat 3 obtained by the gradient flow as the function of the parameter m. The lattice size is L = 10. 4 Conclusion In this paper, we proposed a simple and versatile numerical method which computes an approximate winding number of a mapping from 3D torus T 3 to U(N), whe…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. A Discrete Formulation of Second Stiefel-Whitney Class for Band Theory

    cond-mat.mes-hall 2024-12 conditional novelty 6.0 of 10

    A discrete, gauge-fixing-free formula computes the Z2 second Stiefel-Whitney class from Bloch states on a mesh, verified on tight-binding models.

Reference graph

Works this paper leans on

20 extracted references · 5 canonical work pages · cited by 1 Pith paper

  1. [1]

    Friedan, Commun

    D. Friedan, Commun. Math. Phys. 85, 481-490 (1982) doi:10.1007/BF01403500

    doi:10.1007/bf01403500 1982

  2. [2]

    L¨ uscher, Commun

    M. L¨ uscher, Commun. Math. Phys. 85, 39 (1982) doi:10.1007/BF02029132

    doi:10.1007/bf02029132 1982

  3. [3]

    Seiberg, Phys

    N. Seiberg, Phys. Lett. B 148, 456-460 (1984) doi:10.1016/0370-2693(84)90737-8

    work page doi:10.1016/0370-2693(84)90737-8 1984

  4. [4]

    A. P. Schnyder, S. Ryu, A. Furusaki, and A. W. W. Ludwig Phys. Rev. B 78, 195125 (2008) doi:10.1103/PhysRevB.78.195125 [arXiv:0803.2786 [cond-mat.mes-hall]]

    arXiv 2008

  5. [5]

    Shiozaki, [arXiv:2403.05291 [cond-mat.mes-hall]]

    K. Shiozaki, [arXiv:2403.05291 [cond-mat.mes-hall]]

    arXiv

  6. [6]

    Hamano and T

    F. Hamano and T. Fukui, [arXiv:2405.12537 [nlin.CD]]

    arXiv

  7. [7]

    J. Y. Chen, [arXiv:2406.06673 [hep-lat]]

    arXiv

  8. [8]

    Zhang and J

    P. Zhang and J. Y. Chen, [arXiv:2411.07195 [hep-lat]]

    arXiv

Show all 20 references
  1. [9]

    L¨ uscher, Nucl

    M. L¨ uscher, Nucl. Phys. B 549, 295-334 (1999) doi:10.1016/S0550-3213(99)00115-7 [arXiv:hep-lat/9811032 [hep-lat]]

    1999 arXiv

  2. [10]

    Fukui, Y

    T. Fukui, Y. Hatsugai and H. Suzuki, J. Phys. Soc. Jap. 74, 1674-1677 (2005) doi:10.1143/JPSJ.74.1382 [arXiv:cond-mat/0503172 [cond-mat.mes-hall]]

    2005 arXiv

  3. [11]

    L¨ uscher, JHEP 08, 071 (2010) [erratum: JHEP 03, 092 (2014)] doi:10.1007/JHEP08(2010)071 [arXiv:1006.4518 [hep-lat]]

    M. L¨ uscher, JHEP 08, 071 (2010) [erratum: JHEP 03, 092 (2014)] doi:10.1007/JHEP08(2010)071 [arXiv:1006.4518 [hep-lat]]

    2010 arXiv

  4. [12]

    Tanizaki, A

    Y. Tanizaki, A. Tomiya and H. Watanabe, [arXiv:2411.14812 [hep-lat]]

    arXiv

  5. [13]

    Garc ´ ıa P´ erez, A

    M. Garc ´ ıa P´ erez, A. Gonz´ alez-Arroyo, J. R. Snippe and P. van Baal, Nucl. Phys. B 413, 535-552 (1994) doi:10.1016/0550-3213(94)90631-9 [arXiv:hep-lat/9309009 [hep-lat]]

    1994 arXiv

  6. [14]

    de Forcrand and S

    P. de Forcrand and S. Kim, Nucl. Phys. B Proc. Suppl. 47, 278-281 (1996) doi:10.1016/0920-5632(96)00056-4 [arXiv:hep-lat/9509081 [hep-lat]]

    1996 arXiv

  7. [15]

    So, Prog

    H. So, Prog. Theor. Phys. 74, 585 (1985) doi:10.1143/PTP.74.585

    1985 doi

  8. [16]

    Coste and M

    A. Coste and M. L¨ uscher, Nucl. Phys. B 323, 631-659 (1989) doi:10.1016/0550-3213(89)90127-2

    1989 doi

  9. [17]

    Fujiwara, K

    T. Fujiwara, K. Nagao and H. Suzuki, JHEP 09, 025 (2002) doi:10.1088/1126-6708/2002/09/025 [arXiv:hep- lat/0208057 [hep-lat]]

    2002

  10. [18]

    Nagai and A

    Y. Nagai and A. Tomiya, [arXiv:2409.03030 [hep-lat]]

    arXiv

  11. [19]

    Iwasaki and T

    Y. Iwasaki and T. Yoshi` e, Phys. Lett. B 131, 159-164 (1983) doi:10.1016/0370-2693(83)91111-5

    1983 doi

  12. [20]

    Witten, Commun

    E. Witten, Commun. Math. Phys. 92, 455-472 (1984) doi:10.1007/BF01215276 14

    1984 doi

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.