Coarse information and finite rank obstructions for flat lattice Hodge forms

Strengths

• The central construction is explicit and checkable: a real Fourier subspace of dimension 2(2R+1)^d on which the unit-spacing Hodge energy is at most 4 pi^2 d R^2/n^2 times the squared norm. The energy identity ||GA||^2 = <A,K A> and the single-direction witness identity (2) are correct as stated.
• The rank-nullity argument in Theorem 5.1 and Proposition 5.2 is clean and correctly allows an arbitrary zero-vanishing penalty Phi and a non-symmetric, non-positive V, since only the kernel vector of (C,V) restricted to the band is used.
• The paper is honest about necessity versus sufficiency: Proposition 10.1 and the diag(0,2) example correctly show that a dimension count is not a sufficiency criterion, and the paper repeatedly disclaims any interacting or continuum Yang-Mills conclusion.
• Hypothesis boundaries are tested with explicit counterexamples (nonlinear observation C=||A||^2, the C^1 pair with composition exactly quadratic, the curved sector with z=||u||^2, the C^1-energy variant, the critical constraint P=||x||^2), which materially improves trust in the hypotheses of Theorems 7.1, 8.1 and 9.1.
• The curve construction in Theorem 8.1 correctly notes that the implicit curve leaves S and uses the quadratic form’s expansion to repair this; the order of choosing L before shrinking the curve (Proposition 8.2) is correct.
• Theorem 9.1’s fixed-R trace argument is a neat way to get output-dimension divergence without uniform neighborhoods, and the large-single-site-weight example correctly illustrates why a directional mean rather than a maximum is the right condition.

Weaknesses

• Scope is narrow. The obstruction concerns an unscaled flat form at unit spacing with a fixed-dimensional observation; Section 11 itself shows that replacing K0 by l^2 K0 moves the band threshold to a constant, so the result does not bear on any scaled or physical operator and does not constrain the programme’s actual estimates. Practical significance for a Yang-Mills gap is essentially nil beyond ruling out one naive route.
• The mechanism (low-energy Fourier modes versus finite rank) is textbook, as the paper concedes. The incremental value is mainly bookkeeping for this setting, and no priority search was done by the author.
• Most evidence for rigor of the later sections is external and unverifiable from the PDF (Lean modules, pinned SHAs, Colab runs, declaration map). The text contains an extremely large number of anchors, one of which (the SHA for the uniform endpoints, 1217ca39...e46, 42 hex characters before the final digits) does not have SHA-1 length, so the pinning cannot be audited as printed.
• Section 12 and the closing review-context section devote substantial space to provenance, audit procedure and self-referential claims about reviews, which belong in a package rather than a mathematical paper and obscure the mathematical content.
• Section 11 ‘source inspection’ claims that the repository’s flat covariance endpoint matches Theorem 6.1 with Phi_L(y)=a_L||y||^2 and no fine repair. This is neither proved nor formally checked here and is the paper’s only connection to its stated practical consumer.
• Several proofs are compressed: the path-comparison inequality ||A - lift mu(A)||^2 <= d n^2 ||GA||^2 is only sketched with a ‘deliberately non-optimal’ constant, and the Section 3 spectral multiplicity argument depends on the cross-term vanishing for distinct frequencies under a root-of-unity Gram sum that is asserted rather than shown.

Potential errors

• Theorem 9.1 states the bound for every integer 0 <= R < n, but the construction needs 1 <= s <= R distinct nonzero frequencies with e^{2 pi i s/n} giving linearly independent real quadratures; for R >= n/2 the characters s and n-s coincide up to conjugation so orthonormality of 2R fields can fail. The proof’s ‘also covers even n’ remark does not address R near or above n/2. The stated range should be 2R < n (or R < n/2), and the application only needs fixed R, so the conclusion is likely unaffected.
• Theorem 9.1 asserts |e^{2 pi i s/n}-1| <= 2 pi s/n <= 2 pi R/n and then bounds the sum of s from 1 to R of |...|^2 by 8 pi^2 R^3/n^2 (with the factor 2(1+b)). The summed bound is 4 pi^2 R(R+1)(2R+1)/(6 n^2) times 2(1+b), which is at most 8 pi^2 R^3/n^2 only for the stated middle-to-right inequality if constants are checked carefully (the sum of s^2 is about R^3/3, so the bound holds but with slack); no error seen, but the arithmetic is not displayed.
• In Theorem 9.1 the penalty bound Phi_n(y) <= K_n ||y||^2 is required only near zero, and the derivative-kernel argument requires v in D_n intersect ker L_n. The rank-nullity step uses U = S_{R,n} intersect ker L_n with dim U >= 2R - dim F_n. This is correct, but the final ‘dim F_n >= 2R eventually’ conclusion uses integrality of dim F_n and the strict inequality c(2R-dim F_n) > right side; the case dim F_n = 2R - 1 and equality needs the hypothesis to be strict as written, which it is only when the right side is below c. Fine for large n, but as a general statement (10) is safe only after that observation.
• Section 3: the claim that the sum of the spaces iota(c e^{2 pi i r x_q/l}) lies in ker Q_l for 1 <= r <= R <= floor((l-1)/2) relies on the root-of-unity sum over each block vanishing for frequency r with r not a multiple of l. This holds, but the quotient ‘at most 4 sin^2(pi R/l)’ for the Rayleigh quotient is stated for the Hodge part only; the added (T Q)^ (T Q) vanishes there. Correct, but the statement of mu_{2R}(H_l) <= 4 pi^2 R^2/l^2 also requires that the divergence of these fields be zero, which holds since they are independent of x_p; this is asserted rather than shown.
• Section 9.1: the claim that the example with w = 2 d n^2 and Phi_n(y) = 2 n^d ||y||^2 gives ||A||^2 <= <A, G^* W G A> + Phi(mu(A)) uses ||A||^2 <= 2||A - lift mu||^2 + 2||lift mu||^2 <= 4 d n^2 ||GA||^2... The stated constants (2 d n^2 weights, 2 n^d penalty) are consistent only if the inequality is ||A - lift||^2 <= d n^2 ||GA||^2 as written, which gives ||A||^2 <= 2 d n^2 ||GA||^2 + 2 n^d ||mu||^2; this matches. But b_ij(n) = 2 d n^2 is reported while the same weight is also applied in all components, so the repair G^*WG energy is 2dn^2 times ||GA||^2, i.e. the mean weight in the (i,j) slot is 2dn^2, matching the claim. No error found, but the path-comparison constant d n^2 is unproved in the text.
• The mean-weight statement in the abstract speaks of ‘subquadratic directional spatial mean’, but Theorem 9.1 requires (1+b_ij(n))/n^2 -> 0 only for one fixed pair (i,j); the abstract and Section 1 wording may overstate that the obstruction holds for every weight family that is only small in the mean in other directions. This is a presentation precision point.
• Proposition 8.2 allows P_L, C_L, Phi_L, F_L, G_L to be chosen per L but requires the strict dimension budget dim S > dim F + dim G to hold with a single fixed R. It also requires, for each L, hypotheses of Theorem 8.1 including differentiability of C_L at zero and P_L submersive at zero. The statement says ‘satisfying the hypotheses’, so this is consistent, but ‘no uniform implicit-function radius is required’ means the conclusion is only about each fixed L, i.e. a pointwise-in-L statement, which is weaker than the paper’s phrasing ‘no positive uniform coercivity constant exists’ might suggest if the constant is meant to be independent of the radii r_L; it is, since c is fixed first.
• Section 6 Theorem 6.1: ‘Arbitrary rescalings of C_L preserve its fixed codomain’ is correct for linear C_L, but if the rescaling is L-dependent and unbounded, the theorem still applies because Phi_L is arbitrary with Phi_L(0)=0; no error, but the later use for scaled Poincare in Section 11 should note that the lack of uniform c is for the unscaled K0, which Section 11 does.
• The dimension-four factorized no-go and original Section 3 statements are described as ‘preserved’ from an earlier version and a different source anchor than the new material; their exact content is not reproduced, so claims such as ‘leakage o(l^{-1}) cannot even repair this line’ cannot be checked from this PDF alone (though the displayed arithmetic eps^2 = l^{d-2} eta^2 supports it).

Strong novelty candidates

None reported.

Unresolved material objections

• Practical and scientific significance is not established: the obstruction applies to the unscaled unit-spacing flat Hodge form with fixed-dimensional observations, and Section 11 shows it disappears under natural l^2 rescaling. The paper offers no argument that any proposed repair route in the actual programme is excluded, and the consumer specialization is asserted by source inspection only.
• The formal-verification claims are central to the paper’s credibility but are unverifiable from the PDF; I could not inspect any Lean source, commit, declaration map or log, and the paper itself states the author-side preparation has no independent review (including the prior assessor’s retained objection O6 about missing independent complete Lean rebuild and dependency-closure audit). One printed source SHA is malformed, so at least one pin cannot be resolved as written.
• The range of R in Theorem 9.1 (0 <= R < n) is inconsistent with the construction of 2R orthonormal real fields, which needs distinct frequencies up to R with R < n/2 (or an explicit treatment of the Nyquist and conjugate pairs). The theorem should be restated with 2R < n or the proof extended; the main fixed-R conclusion is probably unaffected but the stated inequality (10) is not proved as written for all R.