{"id":"99912dcd-d3e2-4bf8-9e4c-6618134a3926","arxiv_id":"2412.18796","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A discrete, gauge-fixing-free formula computes the Z2 second Stiefel-Whitney class from Bloch states on a mesh, verified on tight-binding models.","lead":"This paper gives a recipe for calculating a quantum mechanical invariant called the second Stiefel-Whitney class directly from electron wave functions sampled on a discrete momentum grid. The recipe avoids the smooth gauge choices that earlier methods required and is suited to first-principles band-structure codes.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The gauge-invariance proof is incomplete for the perturbations used to avoid -1 eigenvalues in the SO(r) overlap matrices, so the algorithm's well-definedness is not fully established.","rationale":"I read the paper's central claim as: for any sufficiently fine BZ mesh and any frame gauge with no -1 eigenvalues in the SO parts of the normalized overlaps, the product over plaquettes of the signs defined in Eq. (15) equals (-1)^{w2}. The load-bearing step is the passage from the continuum Wilson line description to the lattice algorithm, and within that, the most fragile point is the treatment of the -1 eigenvalue. The paper's own text acknowledges the issue: 'applying random gauge transformations... but as we shall see soon, the SW class will stay gauge invariant.' The subsequent proof (Eqs. (18)-(20)) is a sketch: Eq. (18) is stated, not derived, and the argument that z changes by an exact cocycle is only shown for a gauge transformation between configurations where all lifts are defined. The perturbation that removes a -1 eigenvalue is precisely such a transformation, but at the crossing point the lift is undefined, so the proof does not directly apply. The complement of the singular locus is connected for r>=2, so I expect the invariance to hold; nevertheless, this is an unproven assumption on which the algorithm's well-definedness rests. The numerical tests in Sec. III provide some support (Eqs. (28) and (31)), but they omit the mesh size and a perturbation-independence check. My proposed test would settle the issue directly. If it passes, the algorithm is a plausible and valuable discrete formula; if it fails, the central claim is invalid. Therefore I do not change the reader's conditional verdict, but I have flagged the specific check that should be run before accepting it.","tokens_in":10985,"tokens_out":21039,"duration_ms":203498,"concrete_test":"Implement the algorithm on the 4x4 model (27) for m=1 (known nu=1) and m=3 (known nu=0) on a fixed LxL mesh. In one run, choose frames so that one overlap matrix has SO(r) part exactly -I (a -1 eigenvalue); in a separate run, start from a generic gauge. Apply the paper's small-random-gauge perturbation 100 times with independent seeds, recording nu each time. Verify that all 100 values agree with the known W2 and with the generic-gauge control. Additionally, sample 10^4 random O(r) matrices w, V0, V1 and confirm that u(V0^T w V1) = +/- u(V0)^dagger u(w) u(V1) always holds, testing the asserted Eq. (18).","verdict_should_be":"UNCHANGED","load_bearing_attack":"The algorithm's definition requires that the SO(r) part of every SVD-normalized overlap matrix have no -1 eigenvalue, because the logarithm in Eq. (7) is otherwise undefined. The paper's prescription (Sec. II.B) is to apply random gauge transformations to lift this degeneracy, and it asserts that the resulting Z2 value is unchanged via the gauge-invariance argument Eqs. (17)-(20). However, that argument rests on the asserted identity Eq. (18), which is not derived, and it assumes that all lifts u(w), u(V) are defined. It does not analyze a perturbation path that crosses the -1 eigenvalue locus, where the lift is undefined and the sign eta can jump. Since the computed nu is only defined after an arbitrary perturbation, a failure of invariance under different perturbations would make the algorithm ill-defined as a function of the Bloch states. The numerical tests in Sec. III do not report the mesh size or a systematic check over many independent perturbations, so they do not settle this. If the invariance fails, the central claim of a manifestly quantized gauge-fixing-free formula collapses.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a fully discrete, gauge-fixing-free algorithm for computing the second Stiefel-Whitney class w2 of a real vector bundle over a closed manifold, motivated by band-theoretic applications in PT-symmetric systems. The method uses Bloch states sampled on a triangular or cubic lattice mesh, forms SVD-normalized overlap matrices on each link, lifts the resulting O(r) Wilson lines to Pin+(r) via a logarithmic protocol, and defines a Z2-valued 2-cocycle from the Pin+ Wilson loop on each plaquette. The authors claim the resulting Z2 number is manifestly quantized and independent of the gauge choices, and they demonstrate the algorithm on a known 4x4 tight-binding model, an embedded 8x8 model, and a Whitney-sum test. The paper also draws an analogy to lattice field theory through spinon decomposition and Villainization. The central claim is that the method provides a practical and rigorous numerical tool for w2 without global gauge fixing or continuum smoothness assumptions.","tokens_in":11183,"tokens_out":6093,"duration_ms":70463,"significance":"If the missing technical steps are supplied, this would be a valuable contribution: it offers a direct, parameter-free, and explicit algorithm for a topological invariant that is central to classifying higher-order topological insulators and nodal-line semimetals, and it avoids the spectral-crossing counting that can be numerically fragile. The construction is concrete and implementable, and the numerical checks against known values, including the Whitney sum formula, are appropriate and encouraging. The connection to lattice field theory is suggestive and may help unify viewpoints, although it is not the main result. The main strengths are that the algorithm has no fitted parameters, the quantization is built in by construction, and the tests are independent of the method's own output.","major_comments":[{"comment":"The gauge-invariance of the lattice Z2 value rests on the identity u(w'_{01}) = (-1)^{eta_{01}} u(V_0)^dagger u(w_{01}) u(V_1), which is asserted without proof. This identity is not automatic for an arbitrary section of Pin+(r); it holds because both sides are lifts of the same O(r) element and Pin+(r) -> O(r) is a double cover, but the paper should state and prove this property, including its domain of validity, since the lift protocol fails when the SO part has a -1 eigenvalue. Without a derivation, the central invariance claim of the algorithm is not fully established.","section":"Sec. II.B, Eq. (18)"},{"comment":"The handling of overlap matrices whose SO part has a -1 eigenvalue is a load-bearing point for the algorithm's well-definedness. The paper states that such an occurrence is nongeneric and can be removed by random gauge transformations, but it does not prove that the set of gauge choices for which any link matrix has a -1 eigenvalue has measure zero, nor does it explicitly argue that two different good perturbations produce the same Z2 value. The latter would follow from Eq. (18) once proved, but the logical connection is not made. Since the algorithm is undefined on the bad set, the 'manifestly quantized' claim requires a precise statement of the measure-zero condition and the invariance on the good set.","section":"Sec. II.B, after Eq. (7)"},{"comment":"The numerical tests reported in Eqs. (28) and (31) confirm the expected values, but essential reproducibility parameters are missing: the mesh size L, the number of random gauge samples, and the threshold used in the proximity tests such as 'W_box ~ 1' versus 'W_box ~ -1' are not reported. Without these, the reader cannot assess the robustness of the quantization or reproduce the results. For a paper whose central deliverable is a numerical algorithm, this is a necessary detail.","section":"Sec. III"}],"minor_comments":[{"comment":"The symbol W012 is used for both the O(r) Wilson loop in Eq. (13) and its Pin+(r) lift in Eq. (14); please use distinct notations (for example, W012 and tilde{W}_{012}) to avoid ambiguity.","section":"Eq. (14) and neighboring text"},{"comment":"The phrase 'randomness of the overlap matrix can be ensured by applying random gauge transformations' is vague; a precise probability statement about the codimension of the -1 eigenvalue locus in the gauge group would be more useful.","section":"Sec. II.B, paragraph after Eq. (7)"},{"comment":"The phrase 'see the Supplementary Material of Ref. [21]' is not specific; please provide a section number or a more detailed pointer to the review of Wilson-loop approaches.","section":"Ref. [21]"},{"comment":"The abstract's claim of a 'fully discrete' method is slightly overstated because the algorithm requires a sufficiently fine mesh and a generic gauge choice; consider adding qualifiers such as 'for sufficiently fine meshes'.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":"The paper is likely to be a useful contribution after revision. The main technical gaps are the missing proof of Eq. (18) and the incomplete treatment of the -1 eigenvalue locus; both are fixable with a short lemma and a genericity argument. The numerical section should be expanded with concrete parameters (mesh size, number of samples, thresholds). I see no grounds for rejection, but the current version does not yet fully justify the 'manifestly quantized, gauge-fixing-free' claim as a mathematical statement."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe short version: this paper gives a discrete, gauge-fixing-free formula for the second Stiefel-Whitney class w2, computed directly from Bloch states on a BZ mesh. That is genuinely useful, and the construction is new; it is not just a repackaging of the Wilson-loop spectral-crossing method of Ref. 21. The SVD-normalized overlaps plus Pin+ lift yields a per-plaquette Z2 value, and the authors test it on a known 4x4 model and check the Whitney sum formula on a sum of 1D models. Those checks are the right ones, and they pass. The lattice field theory discussion in Sec. IV is a nice bonus for people who like that language.\n\nThe soft spots are in the rigor around the -1 eigenvalue issue. The algorithm requires that the SO(r) part of every overlap matrix have no -1 eigenvalue, because the logarithm in Eq. (7) is otherwise undefined. The paper's fix is to apply random gauge transformations to push away from that locus, and it asserts that the resulting Z2 value is gauge invariant via Eqs. (17)-(20). But Eq. (18) is stated without derivation, and more importantly, the argument assumes all lifts u(w), u(V) are defined. It never analyzes what happens when a perturbation path crosses the -1 eigenvalue locus. At that crossing the lift is undefined, and the sign eta in Eq. (18) can jump. So the well-definedness of the algorithm as a function of the Bloch states is not fully established. The numerical tests don't settle it either: the paper reports \"confirmed\" results but gives no mesh size, no number of random gauges, and no systematic scan over perturbations. That is a detail gap, not necessarily a fatal flaw, but it makes the \"manifestly quantized\" claim stronger than the evidence.\n\nThe core construction is plausible and likely correct. The missing pieces are a proper proof of Eq. (18) and a continuity argument that handles paths crossing the -1 locus. Both look like revision-level problems, not deep conceptual obstacles.\n\nWho is this for? Anyone who needs to compute w2 in first-principles band structure, and anyone working on discrete topological invariants. The paper deserves a serious referee; it should be sent out, with the expectation that the authors tighten the gauge-invariance argument and report numerical parameters. If those get fixed, this will be a citable algorithm.","headline":"A genuinely discrete, gauge-fixing-free w2 formula with real numerical checks, but the gauge-invariance proof is sketched and the -1 eigenvalue perturbation is not rigorously handled.","tokens_in":11690,"tokens_out":4153,"would_cite":true,"duration_ms":33333,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper presents a fully discrete, gauge-fixing-free algorithm for computing the second Stiefel-Whitney class $w_2$ directly from Bloch states on a Brillouin-zone mesh, and verifies it on tight-binding models.","keywords":["second Stiefel-Whitney class","band theory","PT symmetry","discrete Brillouin zone","Wilson loop","Pin+ group","gauge-fixing-free","lattice field theory"],"falsifier":"Run the algorithm on the 4x4 model at $m=1$ with a mesh chosen so coarse that some plaquette Wilson loop is not close to $\\pm1$, or so that some overlap matrix has a $-1$ eigenvalue in its $SO(r)$ part; if repeated random gauge transformations produce different values of $\\prod_\\square (-1)^{z_\\square}$, the claimed gauge-invariant quantization is false. A cleaner test is to find any PT-symmetric configuration where two different infinitesimal gauge perturbations of the same overlap matrix yield different lifted signs, which would make the cocycle ill-defined.","tokens_in":10803,"feed_emoji":"🧮","tokens_out":10033,"duration_ms":86195,"temperature":0.7,"pith_summary":"The paper claims that the second Stiefel-Whitney class $w_2$, a $\\mathbb{Z}_2$ invariant that characterizes PT-symmetric higher-order topological insulators and nodal-line semimetals, can be computed directly from Bloch states sampled on a discrete Brillouin-zone mesh, with no smooth interpolation and no global gauge fixing. The proposed algorithm forms overlap matrices between neighboring mesh points, normalizes them by singular value decomposition, lifts the resulting $O(r)$ matrices into the Pin$_+$ group, and reads off a $\\pm1$ value per plaquette whose product over the torus is $w_2$. The authors prove the gauge invariance of this plaquette product, derive the Whitney sum formula at the lattice level, and verify the result on a 4x4 tight-binding model and on sums of 2x2 models. They also connect the construction to Villainized lattice field theory, interpreting the discrete $w_2$ as the saddle point of a gauged $\\sigma$ model over interpolations of the projection-valued fields. If the claim is right, first-principles band-structure codes can obtain $w_2$ without the fragile spectral-crossing counting used in Wilson-loop methods.","feed_headline":"Mesh-only recipe computes the second Stiefel-Whitney class","feed_subtitle":"Overlap matrices lifted to Pin+ yield a Z2 invariant without smooth interpolation or gauge fixing.","key_machinery":"The central object is the SVD-normalized overlap matrix $w_{01} = L R^\\top$ obtained from $\\tilde w_{01} = \\Phi_0^\\top \\Phi_1$, together with its explicit lift to the Pin$_+$ group. The lift is defined by decomposing $w = q_w R_1^{s_w}$, taking the principal logarithm of the $SO(r)$ part $q_w$ (assuming no eigenvalue $-1$), exponentiating the rotation parameters with the Clifford generators $\\Sigma_{ij} = [\\gamma_i,\\gamma_j]/(-4)$, and multiplying by $\\gamma_1$ when $s_w = 1$. This lift satisfies $u(w^\\top) = u(w)^\\dagger$, which makes the plaquette Wilson loop orientation-independent; the $\\mathbb{Z}_2$ value $z_\\square$ is read from whether the lifted loop is near $+1$ or $-1$. Gauge invariance follows because a gauge change shifts $z$ by a coboundary $(\\delta\\eta)_{012}$, which cancels when plaquette values are multiplied, and the Whitney sum formula follows from the anticommutation of lifts of odd orthogonal matrices.","core_discovery":"In a PT-symmetric band structure, the occupied Bloch states can be chosen real, giving a frame $\\Phi_v$ at each mesh vertex $v$. For each link the paper forms $\\tilde w_{01} = \\Phi_0^\\top \\Phi_1$ and SVD-normalizes it to $w_{01} = L R^\\top \\in O(r)$, the discrete Wilson line. Each $w$ is lifted to $u(w) \\in \\mathrm{Pin}_+(r)$ by writing $w = q_w R_1^{s_w}$, diagonalizing the $SO(r)$ part $q_w$, exponentiating its logarithm with Clifford-algebra generators, and including a $\\gamma_1^{s_w}$ factor when $\\det w = -1$. On a plaquette $(0123)$ the lifted product $u(w_{01})u(w_{12})u(w_{23})u(w_{30})$ is close to either $+1$ or $-1$, and the bit $z_\\square$ records which. The product $\\prod_\\square (-1)^{z_\\square}$ over the entire torus is shown to be invariant under local $O(r)$ gauge transformations and under changes of lift, so it defines the $\\mathbb{Z}_2$ number $\\nu$ that pairs $w_2$ with the Brillouin-zone torus. The paper verifies this on the Hamiltonian $H_k = \\sin k_x \\, \\sigma_x\\otimes\\sigma_0 + \\sin k_y \\, \\sigma_y\\otimes\\sigma_y + (m-\\cos k_x-\\cos k_y)\\,\\sigma_z\\otimes\\sigma_0$, obtaining $\\nu=1$ for $0<|m|<2$ and $\\nu=0$ otherwise, and checks the Whitney sum formula on $H^{(x)}_k \\oplus H^{(y)}_k$.","pith_inferences":["The paper does not pursue this, but a numerical study that coarsens the mesh near a gap-closing point would turn \"sufficiently fine\" into a quantitative criterion for when SVD normalization and the $-1$-eigenvalue removal are reliable.","The paper does not state this explicitly, but because the input is only the projection $P_v$ of the occupied bands, the same algorithm should apply to Wannier-interpolated or other projection-valued data sets without modification.","The paper does not demonstrate it, but the lift protocol could be reused to compute $\\mathbb{Z}_2$ invariants in magnetic space groups or superconducting classes where reality of the occupied frame is enforced by a different antiunitary symmetry.","The paper does not carry this out, but a three-dimensional extension could pair $w_2$ on $k$-slices with $W_3$ on cubes to give a fully discrete $\\mathbb{Z}_2$-monopole charge for nodal-line semimetals."],"forward_implications":["A band-structure code that outputs Bloch states on a finite $k$-mesh can evaluate $w_2$ directly, without constructing smooth gauges or counting Wilson-loop spectral crossings.","The same plaquette construction works on any closed two-dimensional base manifold, so $w_2$ can be assigned in geometries beyond the torus whenever a real frame exists.","The formula is manifestly $\\mathbb{Z}_2$-quantized: each plaquette contributes exactly $0$ or $1$, so the result cannot drift continuously under small Hamiltonian perturbations.","The lattice-level Whitney sum formula provides a practical consistency check and a way to compute $w_2$ of a direct-sum band group from its factors.","The Villainized-field-theory interpretation gives a common language for the discrete invariant and for lattice constructions of the third integral Stiefel-Whitney class $W_3$ as a Pinc obstruction."],"supporting_citations":[{"why":"Supplies the discrete Chern-number algorithm whose U(1) Wilson-loop logic the paper generalizes to O(r) Wilson lines and Pin+ lifts.","marker":"[10]"},{"why":"The author's prior fully discrete formula for 3D winding numbers, used as precedent for gauge-fixing-free lattice invariants.","marker":"[13]"},{"why":"The Wilson-loop spectral-crossing method for $w_2$ that this paper positions itself against as a complement.","marker":"[21]"},{"why":"Provides the interpolation and Villainization framework that justifies the lattice formula as a saddle-point approximation.","marker":"[22]"},{"why":"The S2 sigma model with spinon decomposition and dynamical U(1) gauge field that the paper adapts to the Grassmannian target.","marker":"[23]"},{"why":"Supplies the Villainized flux construction used to reinterpret the discrete SW class as a lattice gauge flux.","marker":"[26]"},{"why":"Supports the assumption that the total O(N) bundle over the BZ is flat and topologically trivial, so the occupied projection carries all topology.","marker":"[28]"},{"why":"Provides the analogous Villainization of PSU(N) connections that motivates the Pin+ flux step for the O(r) Berry connection.","marker":"[29]"}],"fun_headline_variants":["Discrete points alone nail the Stiefel-Whitney class","Gauge-free Z2 invariant from mesh overlaps","No smooth Bloch waves needed for PT-protected Z2 invariant","Discrete w2 from overlap matrices on a mesh","Second Stiefel-Whitney class without gauge fixing or interpolation"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the mesh is fine enough that every SVD-normalized overlap matrix has an $SO(r)$ part with no eigenvalue $-1$ and every plaquette Wilson loop sits clearly near $+1$ or $-1$; the paper removes the $-1$ cases by random gauge perturbations but does not rigorously prove the resulting $\\mathbb{Z}_2$ class is unchanged by that removal.","fun_headline_variants_meta":{"raw":{"variants":["Discrete points alone nail the Stiefel-Whitney class","Gauge-free Z2 invariant from mesh overlaps","No smooth Bloch waves needed for PT-protected Z2 invariant","Discrete w2 from overlap matrices on a mesh","Second Stiefel-Whitney class without gauge fixing or interpolation"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001393,"raw_usage":{"total_tokens":5696,"prompt_tokens":1063,"completion_tokens":4633,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":679,"completion_tokens_details":{"reasoning_tokens":4550}},"tokens_in":679,"tokens_out":4633,"duration_ms":28428,"temperature":1.0,"reasoning_tokens":4550,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T04:27:05.495416+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the algorithm on the 4x4 model at $m=1$ with a mesh chosen so coarse that some plaquette Wilson loop is not close to $\\pm1$, or so that some overlap matrix has a $-1$ eigenvalue in its $SO(r)$ part; if repeated random gauge transformations produce different values of $\\prod_\\square (-1)^{z_\\square}$, the claimed gauge-invariant quantization is false. A cleaner test is to find any PT-symmetric configuration where two different infinitesimal gauge perturbations of the same overlap matrix yield different lifted signs, which would make the cocycle ill-defined.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the interpolation and Villainization framework that justifies the lattice formula as a saddle-point approximation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The S2 sigma model with spinon decomposition and dynamical U(1) gauge field that the paper adapts to the Grassmannian target."},{"cited_title":"Rabinovici and S","cited_arxiv_id":null,"evidence_quote":"Supports the assumption that the total O(N) bundle over the BZ is flat and topologically trivial, so the occupied projection carries all topology."},{"cited_title":"Di Vecchia, A","cited_arxiv_id":null,"evidence_quote":"Provides the analogous Villainization of PSU(N) connections that motivates the Pin+ flux step for the O(r) Berry connection."}],"review_version":1}