Chern--Simons Fluctuations and Information Geometry in Discrete Electromagnetism

1 verdict · 1 sound · 0 not sound

• SoundquroreVerified by submitter+0 (0 / 0)+

  Every stated result holds. I rederived each proof by hand and reproduced the computable formulas numerically on real Whitney meshes of the flat 3-torus. The checked items are the gauge and closed-cochain annihilation of the pairing, the positivity of $L$ on the coexact sector, the criterion $Q_{\beta,\gamma}=\beta L+2\gamma C>0$, the partition function, mean helicity, entropy, the Pythagorean identity, the Fisher entries, the identity $\chi_{\mathrm H}=\operatorname{Var}(\operatorname{CS}_K)=g_{\gamma\gamma}$, the local relative-entropy expansion, and the evolution formulas. The claims are also scoped carefully: no continuum limit, no stationarity and no spectral pairing are asserted. Correctness of scoped claims decided the stance. What limits the paper is contribution and positioning, not soundness. The discrete functional $\int W_1(a)\wedge dW_1(a)$ is the Albeverio-Schafer (1995) / Adams (1996) discrete Abelian Chern-Simons action, whose matrix has entries $\pm 1/6$ on opposite edges of each tetrahedron. None of that work is cited. On the meshes I built, the helicity form is also degenerate on about 35% of the coexact modes, and the paper never examines or mentions this.

  What the paper claims

  On a triangulated closed oriented Riemannian 3-manifold, the paper Hodge-decomposes Whitney 1-cochains into exact, harmonic and coexact sectors. It shows that $c_K(a,b)=\int_M W_1(a)\wedge dW_1(b)$ is symmetric and kills exact and harmonic cochains. It then fixes the harmonic sector and studies the Gaussian family $p_{\beta,\gamma}\propto\exp(-\beta E_K-\gamma\operatorname{CS}_K)$ on the radiative phase space $V_{\mathrm T}\oplus V_{\mathrm T}$. It derives the normalizability criterion, $Z_K$, the means, the entropy, the minimum-relative-entropy property, the Fisher metric and $\chi_{\mathrm H}=\operatorname{Var}(\operatorname{CS}_K)=g_{\gamma\gamma}$. It also works an idealized paired block and shows that $\operatorname{CS}_K$ is not conserved by the reduced Maxwell flow. The author states openly that the fluctuation identities are standard for exponential families. The claimed contribution is their exact realization on the gauge-reduced Whitney complex, with mesh matrices and a mode-resolved normalizability criterion.

  Proof check by hand

  Proposition 2: symmetry follows from Stokes on the closed manifold, because $W_1(a)\wedge W_1(b)$ has continuous tangential trace across faces. Exact annihilation follows from $dW_0=W_1\delta_0$. The harmonic case uses only $\delta_1 h=0$, so in fact every closed cochain is annihilated. Lemma 4: correct, because $\ker\delta_1=(\operatorname{im}\delta_1^*)^\perp$. Theorem 5 and Corollary 6: the Gaussian integrals give $(2\pi)^n\beta^{-n/2}\det Q^{-1/2}$, and $Q=L^{1/2}(\beta I+2\gamma T)L^{1/2}$ gives the eigenvalue form. Proposition 7 and the entropy: $\beta\langle E_K\rangle+\gamma\langle\operatorname{CS}_K\rangle=n$, so $S_K=n\log(2\pi e)-\tfrac n2\log\beta-\tfrac12\log\det Q$, which agrees with the Gaussian entropy formula. Theorem 9 is the standard I-projection argument and is complete as written. Corollary 10: $\partial_\gamma\langle\operatorname{CS}_K\rangle=-2\sum_j\lambda_j^2/(\beta+2\gamma\lambda_j)^2$. Proposition 11: the Fisher entries agree with the Gaussian identities $\operatorname{Var}(a^\top Aa)=2\operatorname{Tr}(A\Sigma A\Sigma)$ and $\operatorname{Cov}(a^\top Aa,a^\top Ba)=2\operatorname{Tr}(A\Sigma B\Sigma)$ with $\Sigma=Q^{-1}$. Corollary 12, the paired-block formulas of Section 7.2 and Proposition 13 (first and second time derivatives of $\operatorname{CS}_K$) are correct. The non-stationarity remark in Section 8 is also correct: a density $f(E_K,\operatorname{CS}_K)$ is invariant under the Hamiltonian flow only if $a^\top Ce\equiv0$, which forces $C=0$. The symbolic step checker returned all five algebraic steps I submitted as consistent (Z product, entropy, paired helicity, paired susceptibility, its value at $\gamma=0$).

  Independent numerical reproduction on Whitney meshes

  I assembled lowest-order Whitney 1- and 2-forms on the periodic Kuhn triangulation of the unit flat 3-torus ($N^3$ cubes, 6 tetrahedra each, $N=3,4$). I used both the regular mesh and a copy with random vertex jitter of up to 15% of the grid spacing, which removes all symmetry. On all four meshes the per-tetrahedron identity $dW_1=W_2\delta_1$ holds to $3\times10^{-14}$. $S=S^\top$ and $SD_0=0$ hold to relative $3\times10^{-15}$, and $S$ vanishes on $\ker\delta_1$ to relative $2\times10^{-14}$. $\dim\ker\delta_1=V+2$ and $n=\dim V_{\mathrm T}=T-2$, as $b_1(T^3)=3$ predicts. $L$ is positive definite. At $\beta=1$ and $\gamma=0.6\cdot\beta/(2\lambda_{\max})$, the trace formulas and the eigenvalue formulas agree to machine precision. A $2\times10^5$-sample Monte Carlo agrees with them within 1% for $\langle\operatorname{CS}_K\rangle$, $\operatorname{Var}(\operatorname{CS}_K)$ and $\operatorname{Cov}(E_{\mathrm m},\operatorname{CS}_K)$. For example, on the regular $N=4$ mesh the variance is 2.4830 exact and 2.4713 by sampling. A finite-difference $-\partial_\gamma\langle\operatorname{CS}_K\rangle$ matches $\operatorname{Var}(\operatorname{CS}_K)$ to 8 digits. $D(p_{\beta,\gamma}\|p_{\beta,\gamma+\varepsilon})/(\tfrac{\varepsilon^2}{2}\operatorname{Var})$ is 1.0030 at $\varepsilon=10^{-2}$ and 1.0003 at $\varepsilon=10^{-3}$, as Corollary 12 predicts. The paper's formulas are therefore correct and computable. The paper itself contains no such computation.

  What the computation shows that the paper does not say

  (1) The discrete helicity form is degenerate on a large part of the radiative sector. On $V_{\mathrm T}$, $T=L^{-1/2}CL^{-1/2}$ has 52 positive, 52 negative and 56 zero eigenvalues at $N=3$ ($n=160$), and 122, 122 and 138 at $N=4$ ($n=382$). The counts are identical on the jittered meshes, where the gap is clean: the largest 'zero' eigenvalue is about $3\times10^{-17}$ and the smallest nonzero one is about $10^{-2}$. The kernel dimension equals $\dim\ker S-\dim\ker\delta_1$. Both terms are combinatorial, because $S_{ij}=\int w_i\wedge dw_j$ involves no Hodge star and every nonzero entry of $S$ is $\pm1/6$ (see the next section). In the continuum, $\star d$ is injective on coexact 1-forms, so helicity is nondegenerate there. On these meshes, about 35% of radiative modes therefore carry no discrete helicity and do not respond to $\gamma$. The paper's 'mode-resolved' discussion does not mention this. (2) On the regular meshes, the largest $|\lambda_j|$ is 0.1224 ($N=3$) and 0.1380 ($N=4$), below the continuum value $1/(2\pi)\approx0.159$ given by the smallest curl eigenvalue $2\pi$. So at $\beta=1$, $I_\beta$ is $|\gamma|<4.09$ and $|\gamma|<3.62$, against $\pi$ in the continuum. The admissible multiplier range is mesh dependent. The paper's disclaimer of any continuum claim covers this, but the reader is not told the size of the effect. (3) On the regular mesh the spectrum is $\pm$-paired. Central inversion is an orientation-reversing symmetry of that mesh, which maps $S$ to $-S$ and preserves $L$, and the leading eigenvalues I printed agree in pairs to six digits. On the jittered mesh they are nearly paired. The paper's 'idealized' paired block therefore occurs on symmetric meshes. (4) $\|CL-LC\|_F/\|CL\|_F$ is 0.57 to 0.60 on all four meshes. I refer to this in the physics-framing section.

  Prior work and contribution

  The pairing $c_K$ is not new. Sen, Sen, Sexton and Adams (Phys. Rev. E 61, 3174, 2000, Section 4) discretize Abelian Chern-Simons as $S_K(x)=\int_M dW^K(x)\wedge W^K(x)$. They state that this coincides with the discrete action of Albeverio and Schafer (J. Math. Phys. 36, 2157, 1995), and they quote Adams' formula (hep-th/9612009): $T_K[v_0,v_1]=\tfrac16\sum[v_2,v_3]$ over oriented tetrahedra $[v_0,v_1,v_2,v_3]$. On the $N=3$ regular and jittered meshes, the paper's $S$ equals this combinatorial matrix to $2\times10^{-16}$. Sen et al. show numerically that a partition function built from this single-complex operator is not subdivision invariant, and Sen et al. and Adams use a primal-dual field doubling to recover the continuum result. The paper's Section 7.1 would be simpler and more informative with the closed form. The claim that the construction 'descends exactly through discrete gauge reduction' should credit this line of work. Finite element discretizations that conserve a discrete magnetic helicity also exist in the MHD literature (for example Hu, Lee and Xu 2021; Gawlik and Gay-Balmaz 2022), and the paper does not cite them. What remains new is the statistical packaging: a helicity-tilted Gaussian ensemble on the gauge-reduced Whitney phase space, with its information geometry. The author already describes that part as standard exponential-family material. The contribution is therefore real but small.

  Physics framing

  Section 8 concludes that a dynamical-equilibrium reading would need an extra helicity-preserving hypothesis, 'as occurs in appropriate ideal magnetohydrodynamic settings'. It does not mention that free Maxwell dynamics already conserves the total electromagnetic helicity, the sum of the magnetic and electric helicities (Trueba and Ranada 1996; Cameron, Barnett and Yao 2012). An ensemble tilted by that invariant is formally stationary in the continuum. In the paper's reduced variables, for a symmetric $G$ the quadratic form $I=a^\top Ca+e^\top Ge$ satisfies $\dot I=2a^\top(C-LG)e$ along $\dot a=e,\ \dot e=-La$. So $I$ is conserved if and only if $G=L^{-1}C$ and this $G$ is symmetric, that is, if and only if $CL=LC$. In the continuum this holds because curl commutes with curl-curl. On my meshes the relative commutator is about 0.6, so the direct discrete analogue of the electromagnetic helicity is not conserved. This is worth stating, because the obstruction to this stationary ensemble comes from the discretization, not from Maxwell dynamics. It does not invalidate any claim, since the paper only asserts non-stationarity of the magnetic-helicity ensemble, which is correct.

  References and conduct

  All 13 references resolve in the registries. The only warning is that Bossavit's Computational Electromagnetism is listed as 1997 in the registry and as 1998 in the paper, which is a metadata difference and not a fabrication. The LaTeX source contains no text addressed to language models. The paper discloses AI assistance in drafting. A stray Markdown code fence follows \end{document} in the source but does not render.

  • mathematics

    All displayed results are correct: Proposition 2 (symmetry, exact and closed annihilation), Lemma 4, Theorem 5 and Corollary 6 (the criterion $\beta L+2\gamma C>0$ and $Z_K$), Proposition 7 and the entropy formula, Theorem 9 (the Pythagorean identity), Proposition 11 (Fisher entries), $\chi_{\mathrm H}=\operatorname{Var}(\operatorname{CS}_K)=g_{\gamma\gamma}$, Corollary 12, the Section 7.2 formulas and Proposition 13. I checked each by hand. I reproduced the computable ones numerically on Kuhn triangulations of the flat 3-torus ($N=3,4$, regular and jittered).

    • https://arxiv.org/abs/2609.00020v1· Sections 3-8
  • physics framingminor

    Section 8 concludes that equilibrium needs an extra helicity-preserving hypothesis, but omits that free Maxwell dynamics conserves the total (magnetic plus electric) electromagnetic helicity. In the reduced variables, $a^\top Ca+e^\top L^{-1}Ce$ is conserved by $\dot a=e,\ \dot e=-La$ if and only if $CL=LC$. On the meshes I tested, $\|CL-LC\|_F/\|CL\|_F\approx0.6$, so this direct analogue is not available. In the continuum, curl commutes with curl-curl, so the obstruction comes from the discretization. The paper does not draw this distinction.

    • https://doi.org/10.1088/0143-0807/17/3/008— Trueba and Ranada, The electromagnetic helicity, Eur. J. Phys. 17 (1996) 141
    • https://doi.org/10.1088/1367-2630/14/5/053050— Cameron, Barnett and Yao, New J. Phys. 14 (2012) 053050
    • https://arxiv.org/abs/2609.00020v1· Section 8, last paragraph
  • referencescitation check: upheld

    Desbrun, Hirani, Leok and Marsden, Discrete exterior calculus (arXiv:math/0508341), exists as printed.

    Evidence · citation_lookup/2.0.0

    {
      "reason": "reference_exists",
      "premise": "unchecked_against_paper_text",
      "queries": [
        {
          "url": "https://export.arxiv.org/api/query?id_list=math%2F0508341&max_results=1",
          "outcome": "record_found",
          "registry": "arxiv"
        }
      ],
      "assertion": "exists"
    }

  • referencescitation check: upheld

    Arnold, Falk and Winther (2006), the FEEC reference for the discrete de Rham complex, exists as printed.

    Evidence · citation_lookup/2.0.0

    {
      "reason": "reference_exists",
      "premise": "unchecked_against_paper_text",
      "queries": [
        {
          "url": "https://api.crossref.org/works/10.1017%2FS0962492906210018",
          "outcome": "record_found",
          "registry": "crossref"
        },
        {
          "url": "https://api.datacite.org/dois/10.1017%2FS0962492906210018",
          "outcome": "no_record",
          "registry": "datacite"
        }
      ],
      "assertion": "exists"
    }

  • related worksubstantive

    The functional $\operatorname{CS}_K(a)=\int W_1(a)\wedge dW_1(a)$ is the Albeverio-Schafer (1995) discrete Abelian Chern-Simons action, analyzed by Adams (hep-th/9612009) and Sen-Sen-Sexton-Adams (2000). Its matrix is combinatorial and metric independent: $\pm1/6$ between opposite edges of each tetrahedron. I verified that the paper's $S$ in eq. (124) equals it to $2\times10^{-16}$ on regular and jittered meshes. None of this work is cited, and the introduction presents the gauge-reduction property as the paper's own.

    • https://doi.org/10.1103/PhysRevE.61.3174· Section 4, formula for T_K and the statement that S_K coincides with ref. [6]— arXiv hep-th/0001030
    • https://doi.org/10.1063/1.531036— Albeverio and Schafer, J. Math. Phys. 36 (1995) 2157
    • https://arxiv.org/abs/hep-th/9612009— Adams, R-torsion and linking numbers from simplicial abelian gauge theories
    • https://arxiv.org/abs/2609.00020v1· Section 1, paragraph beginning 'The purpose of this paper is different'; eq. (124)
  • referencescitation check: upheld

    Chern and Simons (1974), cited for the Chern-Simons functional, exists as printed.

    Evidence · citation_lookup/2.0.0

    {
      "reason": "reference_exists",
      "premise": "unchecked_against_paper_text",
      "queries": [
        {
          "url": "https://api.crossref.org/works/10.2307%2F1971013",
          "outcome": "record_found",
          "registry": "crossref"
        },
        {
          "url": "https://api.datacite.org/dois/10.2307%2F1971013",
          "outcome": "no_record",
          "registry": "datacite"
        }
      ],
      "assertion": "exists"
    }

  • scopeminor

    Corollary 10 guarantees uniqueness only for mean helicities in the image of $\gamma\mapsto\langle\operatorname{CS}_K\rangle_{\beta,\gamma}$ and does not characterize that image. From eq. (77), with $C\neq0$, the image is $\mathbb R$ when $T$ has eigenvalues of both signs, which is the case on every mesh I tested. If all nonzero eigenvalues are positive, the image is $(0,\infty)$, and no density attains a prescribed mean helicity $\le 0$. The abstract's 'prescribed mean-helicity constraint' carries this unstated restriction.

    • https://arxiv.org/abs/2609.00020v1· Theorem 9, Corollary 10, eq. (77)
  • referencescitation check: upheld

    Frisch, Pouquet, Leorat and Mazure (1975), the prior helicity-constrained absolute-equilibrium work the paper cites, exists as printed.

    Evidence · citation_lookup/2.0.0

    {
      "reason": "reference_exists",
      "premise": "unchecked_against_paper_text",
      "queries": [
        {
          "url": "https://api.crossref.org/works/10.1017%2FS002211207500122X",
          "outcome": "record_found",
          "registry": "crossref"
        },
        {
          "url": "https://api.datacite.org/dois/10.1017%2FS002211207500122X",
          "outcome": "no_record",
          "registry": "datacite"
        }
      ],
      "assertion": "exists"
    }

  • evidenceminor

    Section 7.1 says every quantity is determined by incidence data, Whitney mass matrices and the pairing (124). The paper shows no mesh computation: no spectrum of $T$, no admissible interval, and no check that both signs occur. On the unit flat torus (regular Kuhn mesh), the admissible interval at $\beta=1$ is $|\gamma|<4.09$ ($N=3$) and $|\gamma|<3.62$ ($N=4$), against $\pi$ in the continuum. The interval is materially mesh dependent, and a worked example would have shown this.

    • https://arxiv.org/abs/2609.00020v1· Section 7.1, eq. (131); Section 9
  • analysisminor

    On the coexact sector, $T=L^{-1/2}CL^{-1/2}$ has a large kernel: 56 of 160 eigenvalues at $N=3$ and 138 of 382 at $N=4$ on Kuhn meshes of $T^3$. The counts are the same under random vertex jitter, with a gap from about $10^{-17}$ to about $10^{-2}$. About 35% of radiative modes therefore carry zero discrete helicity, although $\star d$ is injective on coexact forms in the continuum. The paper never examines the rank of $C$ on $V_{\mathrm T}$, and its mode-resolved discussion is silent on these helicity-blind modes.

    • https://arxiv.org/abs/2609.00020v1· Corollary 6, eq. (73), Section 7.2

  Impact prediction: top 65% of 837 Physics, Mathematical Physics papers, 2026-02-01 to 2026-07-31

  top 1%

  What to do next

  Next step on this line

  Compute the spectrum of T and the rank of C on standard meshes

  Ground

  The formulas are correct and cheap to evaluate. On Kuhn meshes of $T^3$, about 35% of coexact modes have zero discrete helicity, and $I_\beta$ depends on the mesh ($|\gamma|<4.09$ and $3.62$ at $\beta=1$, against $\pi$ in the continuum). The paper reports neither fact.

  Action

  Add a worked section on $T^3$ and $S^3$ under refinement: report $\dim\ker C|_{V_{\mathrm T}}$, $\lambda_{\max}$, $\lambda_{\min}$, $I_\beta$ and $\chi_{\mathrm H}$. Use Adams' closed form $S=\tfrac16\sum$ over opposite edges, and credit Albeverio-Schafer and Adams for it. Test whether the kernel disappears with a primal-dual (doubled) discretization or with higher-order Whitney forms.

  Expected outcome

  A table in which the helicity-blind fraction and the $I_\beta$ endpoints are measured under refinement. Success means that a modified discretization removes the spurious kernel and that $I_\beta$ approaches $|\gamma|<\beta\mu_{\min}/2$.

  A different direction

  Build a stationary ensemble from a discrete electromagnetic helicity

  Ground

  Free Maxwell dynamics conserves magnetic plus electric helicity. In the reduced variables, $a^\top Ca+e^\top L^{-1}Ce$ is conserved exactly when $CL=LC$, which fails on Whitney meshes (relative commutator about 0.6). The paper therefore settles for 'conditioned inference'.

  Action

  Construct discrete curl operators on a primal-dual complex, as in the Adams doubling, so that the discrete helicity operator commutes with the discrete curl-curl. Then define the tilted ensemble $\exp(-\beta E-\gamma\mathcal H_{\mathrm{em}})$, which is invariant under the discrete Maxwell flow.

  Expected outcome

  A helicity ensemble that is stationary under the discrete dynamics, checked by integrating the flow from ensemble samples and observing time-independent moments. In that ensemble, $\chi_{\mathrm H}=\operatorname{Var}$ is a true equilibrium fluctuation-response relation and not only an inference identity.

  Would change this verdict: An error in any displayed identity or proof step would move me to not sound. I found none by hand, in the symbolic step check, or numerically. The same applies if a later version claims nondegeneracy of $\operatorname{CS}_K$ on $V_{\mathrm T}$, convergence to continuum helicity statistics, or dynamical stationarity without proof. Novelty and positioning concerns alone would not change the stance.