Sharp Onset and Unbounded Growth of Norm-Optimal Self-Commutator Rank

Strengths

• Complete resolution of the low-dimensional compatibility question (d <= 7) that explicitly certifies all singular strata (z > 0) rather than relying on an invalid zero-padding reduction.
• Explicit, hand-verifiable primal and sparse dual Horn certificates for the two-chamber dimension-eight cone (Theorem 4.1) and the two-valued dimension-nine witness (Proposition 5.1).
• Elegant symbolic coarse-grid restriction lemma for triangular hives (Lemma 6.1), where coarse rhombus slacks telescope as nonnegative t-by-t sums of fine rhombus slacks, transferring whole rank-face lower bounds across dimensions.
• Careful treatment of the affine hive normalization H'(I,L) = H(I,L) + P_n(I-L) needed to restore the zero terminal eigenvalue condition in unrestricted coarse-graining.
• Exemplary separation between floating-point discovery and exact rational certification (Fraction and GMP arithmetic) across multiple independent verification routes in the supplement.

Weaknesses

• The proofs of Theorem 4.1 and Proposition 5.1 in the main PDF omit the explicit tables of sparse Horn triples and dual multipliers, deferring those finite lists to the verification scripts and absorbed supplement.
• The exact value of r_*(G_t) in Theorem 7.3 is narrowed to the interval [17t + 1, 18t] rather than pinned down to 18t, and the conjectured intermediate rank-cost curve 30k - j for general k, j remains open.
• Theorem 3.1 and Proposition 7.2 rely on computer-assisted polyhedral/LP certificates rather than a purely conceptual or proof-assistant-verified proof, though the rational Farkas/dual certificates are compact and independently audited.

Potential errors

• In the excluded inventory metadata, math/results/d_le_7_epigraph_certificate.json is listed with SHA-256 146af6cd... (likely due to line-ending normalization in the outer repository), whereas Section 8 of the PDF, manifest.json, and run_scientific_replay.py record 0103a643...; this is a minor packaging metadata discrepancy that does not affect the mathematical content or script execution.

Strong novelty candidates

• The sharp dimension threshold (Theorem 3.1 and Corollary 4.3) showing that d = 8 is the least dimension in which Hilbert-Schmidt norm optimality forces self-commutator factor rank strictly above the inertia bound.
• The hive coarse-graining amplification mechanism (Lemma 6.1 and Theorem 7.3) proving that the additive excess r_*(F) - r_0(F) of norm-optimal self-commutator rank over inertia is unbounded.

Unresolved material objections

None reported.

Strengths

• Proposition 2.2 has consistent normalization: gamma=2lambda and the objective is half the common positive trace. Subtracting the least common eigenvalue preserves the difference, and C=P^(1/2)U correctly realizes isospectral positive P,Q with Q=U*PU.
• The low-dimensional verifier enumerates the stated 33 sign-zero cases at the literal inertia rank. Exact checks of discarded inequalities prevent floating redundancy proposals from silently changing the cone. The companion checker preserves lin_set and requests both Farkas orientations for projected equalities.
• The dimension-eight branches have the stated ordering, traces and endpoint ranks. The displayed sparse dual supports give the claimed objective coefficients and affine lower bounds. Affine residual checks at chamber endpoints genuinely cover continuous parameter intervals, rather than merely sampled targets.
• Lemma 6.1's three telescoping formulas have valid triangular index ranges, including boundary cases. The affine normalization H'=H+Pn(I-L) preserves rhombi and the target boundary while reducing cost by nPn/2. Consequently the amplification lower bounds apply to arbitrary feasible hives, without assuming block-diagonal optimizers.
• Direct inspection of the order-27 gamma multipliers gives dual right-hand sides 89 and 88. The rank-17 primal singular-square list has 17 positive entries and total 176, agreeing with cost 88. These are supporting spot checks, not a substitute for complete internal-vertex stationarity verification.

Weaknesses

• The universal classification and order-27 seed depend on substantial finite computation. Static inspection supports the verifier design but cannot establish that every regenerated face, Farkas identity and hive residual passes. An equipped independent reviewer should execute the exact release; packet omissions alone do not show that public proof objects are missing.
• Section 6 does not explicitly assign its three boundary increments. State h(0,j)-h(0,j-1)=p_j, h(j,N)-h(j-1,N)=-p_(N-j+1), and h(j,j)-h(j-1,j-1)=2lambda_j, as implemented in ExactHive. This would make the spectral interpretation and normalization signs directly checkable from the PDF.
• Sections 4 and 5 could include the short sparse dual tables already present in the code. Currently even the relatively small sharpness lower bounds require leaving the canonical manuscript, although their coefficient arithmetic is suitable for a concise appendix.
• Novelty remains qualified by the author's bounded search. The supplied historical audit narratives do not independently establish the contents of adjacent literature or priority, and I did not retrieve those sources.

Potential errors

• Certificate identity mismatch: the supplement's excluded-file inventory assigns the threshold JSON 137284 bytes and SHA-256 146af6cd02ec..., whereas §8, the manifest and run_scientific_replay.py require 144794 bytes and 0103a6436449.... If the inventoried bytes are used, check_hashes stops before scientific verification. Reconcile repository/release bytes, including possible newline normalization. This does not establish a mathematical error or absence of the public artifact.
• The optional --with-wsl-route requires a fixed summary_sha256, but verify_low_dimension_minimality_exact.py includes measured elapsed_seconds in the hashed case summaries. That digest is runtime-dependent and cannot serve as a stable acceptance invariant. Exclude timing from the digest or compare mathematical fields. The default proof route does not require this optional check.
• The abstract and Table 1 call all 272 projected H rows 'facets', while §8 identifies 46 as equalities. Proposition 3.2 should explicitly require both orientations of every projected equality and cite the companion equality audit there. The supplied companion code addresses this issue algorithmically, so it is not an unresolved mathematical gap.

Strong novelty candidates

• Universal inertia-rank attainment for every traceless Hermitian target through dimension seven, including singular spectra, together with an exact dimension-eight obstruction establishing the sharp onset.
• The explicit two-ray dimension-eight phase with rank-four penalty min{b,c-b}, interior minimum optimal rank five and the within-family sharp cost ratio 14/13.
• Transfer of finite rank-face obstructions through hive coarse-graining to obtain exact costs on the 15t, 16t and 17t faces and unbounded additive optimal-rank excess at least 2t+1.

Unresolved material objections

None reported.

Strengths

• Proposition 3.2/Theorem 3.1: the canonical verifier and an independently assembled native-GMP audit each cover all 33 strata, 813 face generators and 272 projected H rows. Both signs of all 46 affine-hull equalities were checked exactly, closing the singular-stratum and equality-flag risks.
• Theorem 4.1: recursive Horn and direct LR replay agree on 8,752 Horn triples and all three affine primal/dual branches. Endpoint residual checks extend to every real b/c in the respective interval by affinity and homogeneity. The strict interior gap min(b,c-b) proves r*=5, with rank four at both nonzero boundary rays.
• Propositions 5.1 and 7.2: exact replay gives dimension-nine costs 29/30 and order-27 rank-16/17 costs 89/88. The latter verifier checks every boundary and rhombus, dual sign, stationarity equation and fixed rank coordinate over Fraction; the certificates were executed, not merely read as stored PASS claims.
• Lemma 6.1/Theorem 7.3: rectangle telescoping works for every positive integer t, including I=L boundary rhombi. The affine correction H+P_n(I-L) correctly normalizes the unrestricted coarse spectrum. Literal rank-face transfer and direct sums give costs 87t,90t,89t,88t and excess at least 2t+1.
• Remark 3.3 is a useful nontrivial warning: replay confirms that zero padding changes cost from 74/7 to 73/7 while the padded rank-four cost stays 74/7. The manuscript therefore correctly treats singular spectra in their original ambient dimension instead of inferring them from support restriction.
• All 80 public-code manifest entries and the locked canonical PDF digest verify. The regenerated threshold and both family JSON streams are byte-identical to the supplied files. All 86 original ZIP-entry byte streams were unchanged after review.

Weaknesses

• The default replay promise on p.10 is false for the supplied publication bytes: even with the exact pinned versions, run_scientific_replay.py aborts because it expects threshold SHA-256 0103a643644977400052a68738102a7633d6371db8717383ddefa687da8a18a0, while both the supplied file and my regenerated output have 146af6cd02ec2975aa360bf3f8ebf7fb7cdd64a388ab2f6538201e539be8267d. Correct the published metadata; no hash check was changed here.
• The original source manifest also expects a different paper.txt hash and size. Its omitted narrative entries are permitted by INPUT-SCOPE.md and are not evidence of missing public files. Reconcile the actual release manifest with the deposited artifacts while preserving the independently verified mathematical content.
• The canonical threshold JSON omits lin_set and opposite-sign equality certificates. Consequently its 272 rows alone are not a self-contained equality-aware projection proof. The companion audit and referee reconstruction resolve the mathematical concern, but the canonical format should encode the 46 equalities explicitly rather than require this repair in a separate route.
• The README gives Python package pins but omits the cdd/GMP development headers and shared-library setup needed to build pycddlib in this environment. Several installation failures preceded successful local setup. The optional WSL script fails on POSIX path parsing after dependencies are available; it should remain explicitly platform-specific.
• The coarse-grid programs check finite ranges (scales 1..6, theorem regressions 1..5, endpoint algebra 1..32). Their theorem-labelled output is not a machine proof for all parameters. The displayed telescoping argument supplies that quantifier, and documentation should consistently maintain this distinction.
• Novelty and impact remain bounded judgments. The inspected primary literature supports a different-objective comparison, but Fong 1986 and Maher 2006 were not fully retrieved. No exhaustive priority claim or broad application follows. The universal Horn/hive machinery and puzzle inflation are established background; the principal advance is their exact rank-optimization application.

Potential errors

• The pp.4/9 census calls 272 projected rows facets, although 46 are affine-hull equalities. Interpreting the serialized orientation alone as the whole H representation would invalidate the containment proof. This potential gap was resolved by actual equality-preserving audits; distinguish 226 inequality rows from 46 equality rows in the exposition.
• The p.7 citation to KTW Lemma 8 for uniform puzzle inflation is abbreviated: that lemma stretches one label, whereas uniform two-label inflation also uses the dual operation described in KTW section 6.2 (https://arxiv.org/pdf/math/0107011). This is a citation-precision issue, not a defect in the independently proved coarse-grid lemma.

Strong novelty candidates

• A universal, singular-inclusive equality r*(F)=max(n+(F),n-(F)) through dimension seven, paired with a verified dimension-eight failure, yields an exact ambient-dimension threshold rather than a list of examples.
• The complete stated two-ray dimension-eight cone has explicit affine unrestricted and rank-four costs, a sharp 14/13 relative tax within that cone, and rank-five necessity at every projective interior point. This is not a classification of all dimension-eight targets.
• An exact order-27 rank-face certificate combined with coarse-grid restriction gives the explicit family in dimension 27t with 17t+1<=r*<=18t and inertia 15t. The resulting unbounded additive excess is the strongest qualitative advance; no exact formula r*=18t for general t is claimed.

Unresolved material objections

None reported.

Strengths

• Exact verification now passes all six chambers.
• The corrected package is complete and fully hash-checked against its manifest.
• The complete default replay passed.
• The canonical 11-page PDF was visually clean.

Weaknesses

• The correction does not add evidence warranting a higher assessment of mathematical novelty or significance.
• The evaluation record supports solid acceptance rather than an exceptional Millennium-anchor rating.

Potential errors

None reported.

Strong novelty candidates

None reported.

Unresolved material objections

None reported.