{"id":"7718d302-c54d-453e-9002-79c5789d3aa5","arxiv_id":"1908.08541","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A band is a band representation exactly when it splits into analytic, topologically trivial unit-rank bands permuted by symmetry, which is used to prove that specific photonic crystals are fragile topological with removable boundary states.","lead":"This paper proves a theorem that reduces complex band representations into simple single-band pieces, then uses it to show certain photonic crystals are 'fragile topological' with removable surface states. The result reframes how to detect symmetric Wannier functions and challenges the common assumption that gapped topological photonic crystals have robust surface states.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The boundary-stability proof relies on an unproved band-representable complement Q'; without a completion lemma, the general incompatibility theorem is not established.","rationale":"The reader's conditional verdict is appropriate. The crystallographic splitting theorem itself is proven in App. C with a detailed induction argument, and I do not see an internal inconsistency in that proof. The weakest point is indeed the Boundary Stability Criterion in Sec. IX B: the construction of the model Hamiltonian Hb (Eq. 17) depends on a finite-rank, analytic, band-representable complement Q', and the one-sentence justification in Sec. IX B 1 is an unsupported completeness assertion. A continuum description supplies unbounded rank, but it does not supply band-representability of an arbitrary finite-rank high-energy subspace; the paper needs a positive completion theorem for G-equivariant bundles that is not stated or proven. This is load-bearing because the general claim that band representations are incompatible with robust boundary states is used to strengthen the specific photonic-crystal conclusions. However, the photonic fragility classification (Sec. VIII) and the explicit domain-wall manipulations (Figs. 8(d), 9(c-e)) provide independent numerical evidence for the specific cases, so the correct response is to keep the paper conditional rather than reject it. A formal completion lemma, or a counterexample to it, would settle the concern either way.","tokens_in":56219,"tokens_out":13418,"duration_ms":170754,"concrete_test":"Prove or disprove the completion lemma: for every finite-rank G-equivariant bundle P that is a band representation, there exists a finite-rank band representation Q' such that P ⊕ Q' is isomorphic to a tightly-bound band representation of sufficiently large rank. One concrete route is to search the equivariant K-theory data for a counterexample: take a finite-rank tight-binding model whose occupied subspace P is a band representation and whose first N high-energy bands form an obstructed representation, then check whether adjoining any symmetry-allowed band representation outside the fixed Hilbert space makes the total bundle isomorphic to a tightly-bound band representation. If such a completion can always be constructed, the Q' assumption is vindicated; if a counterexample exists, the Boundary Stability Criterion must be weakened or proven by a different argument.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The contrapositive proof of the Boundary Stability Criterion (Sec. IX B 1, Eq. 17) requires a finite-rank, analytic projector Q' with I = P + Q + Q' that transforms as a band representation of G. The text asserts that such bands 'always exist because we are approximating a continuum description of crystals,' but this is not a proof. A continuum description supplies infinitely many high-energy bands, hence unbounded rank, but it does not by itself guarantee that any finite subset of them can be chosen to form a band representation; in class AII or in class A with magnetic translation structure, high-energy bands in a fixed energy window can themselves be obstructed representations. What is needed is a completion lemma: for every finite-rank G-equivariant bundle P that is a band representation, there exists a finite-rank band-representation Q' such that P ⊕ Q' is isomorphic to a tightly-bound band representation of sufficiently large rank. The paper's symmetric tight-binding limit theorem appears to assume this lemma rather than prove it, and the universal G-bundle theorem only provides a homotopy once the ambient bundle is already band-representable. Because this step underlies the claimed general incompatibility of band representations with spectrally robust boundary states, the photonic corollary's 'removable boundary states' statement is not fully supported as a theorem, although the specific numerical demonstrations in Figs. 8(d) and 9(d-e) may still be valid.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a 'crystallographic splitting theorem': a rank-N band representation of a (monomial) space group is equivalent to a splitting into N analytic, unit-rank bands with trivial first Chern class, on which the space group acts by permutations. The theorem is proved in Appendix C for integer-spin band representations, with stated exceptions for half-integer-spin cubic double space groups. The authors use the theorem to give proof methods for band representability, to prove fragility of certain rotation-invariant class-AI TCIs, to relate Zak-phase winding to Wannier obstructions, and to reclassify existing photonic crystals as fragile topological, including Yang et al.'s hexagonal crystal. They also state a general criterion that band-representable low-energy subspaces cannot host spectrally robust boundary or domain-wall states.","tokens_in":56517,"tokens_out":3027,"duration_ms":34221,"significance":"If the splitting theorem stands, it is a significant structural result: it reduces band-representability questions to unit-rank bundle computations and gives a rigorous bridge between Zak's real-space definition of band representations and modern topological band theory. The proof in Appendix C is detailed and appears internally consistent, and the use of Huppert's theorem and the unit-rank result of Ref. 78 gives the argument independent anchors. The applications are broad and falsifiable: the photonic classification of Yang et al.'s crystal is a concrete, checkable claim, and the numerical constructions of symmetric Wannier functions are reproducible from the stated parameters. The main weakness is that the general boundary-stability criterion in Sec. IX relies on an unproved existence lemma for a finite-rank band-representable complement Q', so the headline statement 'band representations are incompatible with robust boundary states' is not fully established as a theorem, although the specific numerical demonstrations in Secs. VIII may still be correct.","major_comments":[{"comment":"The proof of the boundary stability criterion assumes a finite-rank, analytic projector Q' with I = P + Q + Q' that transforms as a band representation of G. The text asserts such bands 'always exist because we are approximating a continuum description of crystals,' but this is not a proof. A continuum Hamiltonian provides infinitely many high-energy bands, hence unbounded rank, and does not by itself guarantee that a finite subset in an energy window can be chosen to form a band representation; those bands could themselves be obstructed. What is needed is a completion lemma: for every finite-rank G-equivariant, band-representable P, there exists a finite-rank band-representable Q' such that P ⊕ Q' is isomorphic to a tightly-bound band representation. The symmetric tight-binding limit theorem and the universal G-bundle theorem supply a homotopy only when the ambient bundle is already band-representable. Because this step is load-bearing for the claimed incompatibility of band representations with spectrally robust boundary states, the general theorem as stated is not fully supported, although the specific numerical demonstrations in Figs. 8(d) and 9(d-e) are unaffected.","section":"Sec. IX B 1, Eq. (17)"},{"comment":"The no-go theorem for Wigner-Dyson class AI ('there exists no obstructed representation of G3 = ZT2 × T2 in d = 2') is presented with a sketch rather than a proof. The argument that no symmetry-enforced degeneracy of the Zak phase exists, and that all winding numbers are therefore reducible to zero, is stated as 'applying this analysis,' but the underlying symmetry analysis of the Wilson loop is not given. Since this theorem is used to support the contrast between class AI and class AII and is a strong claim in its own right, it should either be proved in detail or explicitly labelled as a conjecture with the evidence summarized.","section":"Sec. V C 1"}],"minor_comments":[{"comment":"The proof refers to 'Condition (iii)' twice, but the lemma states only conditions (i) and (ii); the intended reference appears to be condition (ii). Please correct the numbering.","section":"Sec. V B, proof of symmetric splitting lemma"},{"comment":"The sentence 'Then Φ1(0)≡Φ1(π) vs Φ1(0)⁄≡Φ1(kz) correspond respectively to the trivial vs nontrivial Z2 class' appears to contain a typo: the second comparison should presumably be Φ1(0) ≠ Φ1(π), not Φ1(0) ≠ Φ1(kz).","section":"Sec. VIII A"},{"comment":"There is a typo in 'obstruction against time-reversal-symmetric Wannief functions'; 'Wannief' should be 'Wannier'. Similar minor typos occur elsewhere, such as 'elabroated' in Sec. IV D and 'an tight-binding' in Sec. IX A.","section":"Sec. II"},{"comment":"The discussion of the two mechanisms for symmetric tight-binding obstruction is useful, but the Hopf-insulator example is deferred to 'a later publication,' which makes the completeness of the taxonomy hard to assess. Please either provide the supporting calculation or state explicitly that this is a preview.","section":"Sec. IX B 1"}],"recommendation":"major_revision","confidential_remarks":"The splitting theorem itself is a genuine mathematical contribution and the photonic classification is likely to be influential. The main issue is the unproved complement lemma in Sec. IX B 1: the reviewer's stress-test concern is valid, and the paper should either prove the completion lemma, restrict the boundary-stability claim to models where Q' is explicitly constructed, or downgrade the claim from a theorem to a conditional statement. This is fixable within the manuscript's scope, so I do not recommend rejection. The no-go theorem in Sec. V C 1 should also be upgraded from a sketch to a proof or explicitly qualified."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here is the short version. The crystallographic splitting theorem in Sec. IV is a genuine and useful contribution, and the proof in App. C is detailed and internally consistent. I did not find a circular step. App. F's proof that essentially all crystallographic and grey magnetic point groups are monomial is also substantial. The applications—fragile TCIs in class AI and the reclassification of the Yang et al. photonic crystal as fragile with removable surface states—are concrete, with explicit tight-binding parameters and Chern/Zak diagnostics. The paper earns its length.\n\nThe main soft spot is exactly what the stress-test note flags. The boundary stability criterion in Sec. IX B requires a finite-rank, analytic, band-representable complement Q' above the gap. The text asserts such bands 'always exist because we are approximating a continuum description of crystals.' That is not a proof. A continuum supplies infinitely many high-energy bands, but in a fixed energy window those bands could themselves be obstructed. What is needed is a completion lemma: every finite-rank BR P should be extendable by a finite-rank BR complement so that P plus that complement sits inside a tightly-bound BR of large rank. The symmetric tight-binding limit theorem appears to assume rather than prove that embedding. Without it, the 'band representations are incompatible with spectrally robust boundary states' theorem is not established at the level the paper claims.\n\nThis does not sink the paper. The splitting theorem is independent, and the specific photonic demonstrations—the alternate surface termination in Fig. 8(d) and the domain-wall thickness sweep in Figs. 9(d–e)—stand as numerical evidence even if the general criterion is weakened. But the abstract's claim to 'disprove a widespread perception' goes through Sec. IX, so that conclusion should be trimmed until the Q' question is settled.\n\nMinor point: no code or data archive is provided, which makes the numerical tight-binding results harder to check. The parameters are reported, which is acceptable, but a release would help. The citation pattern is fine; self-citations are used as tools, not as the conclusion.\n\nBottom line: I would send this to a serious referee. The central theorem and the monomiality results deserve a careful reading and likely publication. The referee should be asked to focus on the Q' existence/completion lemma and to separate the general boundary-stability claim from the specific photonic demonstrations.","headline":"The crystallographic splitting theorem is a solid, publishable result, but the general boundary-stability theorem rests on an unproved Q' completion step and should not anchor the broadest claims.","tokens_in":57002,"tokens_out":2849,"would_cite":true,"duration_ms":31800,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["20C35"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves a crystallographic splitting theorem: a rank-$N$ band representation is exactly a sum of $N$ unit-rank, topologically trivial bands that space-group symmetries permute.","keywords":["crystallographic splitting theorem","band representations","fragile topological insulators","photonic crystals","Wannier functions","Zak phase","boundary states","space-group symmetry"],"falsifier":"For a continuum periodic Schrödinger or photonic Hamiltonian whose lowest gapped two-band subspace is a band representation, compute the surface spectrum with all symmetry-allowed high-energy bands included; if any in-gap boundary state cannot be removed by a symmetry-preserving band deformation, the boundary stability criterion fails. The specific missing check is whether the required finite-rank, band-representable $Q'$ exists in such a continuum model.","tokens_in":56044,"feed_emoji":"🌀","tokens_out":10051,"duration_ms":96752,"temperature":0.7,"pith_summary":"Band representations are the proposed building blocks of band theory: bands whose Wannier functions are generated by a space group from finitely many symmetric Wannier centers. This paper proves a crystallographic splitting theorem that recasts them topologically: a rank-$N$ band representation is equivalent to a splitting into $N$ unit-rank bands that are analytic, have trivial first Chern class, and are permuted by every space-group symmetry. This makes band-representability decidable by diagonalizing projected symmetry or position operators, without guessing Wannier functions, and it turns fragility into a checkable property: an obstructed band is fragile when adding a band representation removes the obstruction. Applying the theorem, the paper proves that rotation-symmetric topological crystalline insulators in symmetry class AI are fragile, and that an existing hexagonal photonic crystal is a fragile topological system whose domain-wall states can be removed by symmetry-allowed couplings. If right, this undermines the widespread reading of those photonic surface states as topologically protected and gives a general criterion: band representations cannot host spectrally robust boundary or domain-wall states.","feed_headline":"Band representations are exactly permuted stacks of trivial bands","feed_subtitle":"New theorem makes band-representability checkable by symmetry and Chern class; photonic surface states become removable.","key_machinery":"The load-bearing object is the symmetric Wannier splitting: a decomposition $P=\\oplus_{j=1}^N P_j$ into unit-rank projectors that are analytic in $k$, have trivial first Chern class, and are permuted by $G$. The crystallographic splitting theorem states that such a splitting exists iff $P$ is a monomial band representation, i.e., a band representation induced from a monomial representation of a site stabilizer, equivalently a representation with a basis of complex permutation matrices. The proof organizes the unit-rank bands into space-group orbits, shows the number of Wannier centers divides the rank, and induces a Wannier basis from a single representative. Computationally, the paper constructs symmetric splittings by diagonalizing a projected symmetry operator or a projected position operator; the Zak phase of each unit-rank band then certifies the trivial Chern condition.","core_discovery":"The central claim is the equivalence, for monomial band representations of crystallographic space groups (and their grey magnetic and double extensions), between being a rank-$N$ band representation and admitting a symmetric Wannier splitting. Explicitly, $P$ is a monomial band representation of $G$ iff $P=\\oplus_{j=1}^N P_j$ with each $P_j$ an analytic projector with trivial first Chern class and with every $g\\in G$ acting as a permutation on $\\{P_j\\}$. The theorem is proven for all integer-spin band representations of all space groups; in half-integer spin, the equivalence holds in two spatial dimensions and for all three-dimensional non-cubic double point groups, with double cubic point groups as the stated exceptions. From this, the paper derives that rotation-invariant topological crystalline insulators in class AI are fragile obstructed representations, that the filled bands of certain photonic crystals realize such fragile representations with removable boundary and domain-wall states, and that for space groups generated by time reversal and/or spatial inversion, the symmetric Wannier obstruction is equivalent to a nonzero Zak-phase winding.","pith_inferences":["If the boundary stability criterion transfers to real crystals, the practical search for topological surface states should focus on low-energy subspaces that remain obstructed after adding all occupied elementary band representations; the paper notes that k-space symmetry representations alone are not an exhaustive diagnostic.","The theorem suggests a computational pipeline for materials: compute the projector of a target band set, diagonalize projected symmetry operators, and read fragility directly from the Chern numbers of the resulting unit-rank bands, without Wannier interpolation.","For the exceptional double cubic point groups, a natural extension is that the splitting should allow rank-two irreducible blocks rather than unit-rank bands; the paper leaves this as speculation."],"forward_implications":["Band representability can be decided algorithmically: diagonalize a projected symmetry operator or projected position operator and check eigenvalue nondegeneracy and trivial first Chern class; no trial Wannier functions are needed.","Rotation-symmetric topological crystalline insulators in Wigner-Dyson class AI are fragile: adding a single unit-rank band representation converts the obstructed filled band into a band representation.","The hexagonal photonic crystal built from split-ring resonators is a fragile obstructed representation in class AI, not a Kane-Mele-type AII insulator; its domain-wall Dirac states can be removed from the gap by symmetry-preserving deformations.","Band representations are incompatible with spectrally robust boundary or domain-wall states: a band-representable low-energy subspace cannot protect in-gap boundary states against all symmetry-allowed couplings.","For space groups generated by time reversal and/or spatial inversion, an obstructed representation is equivalent to a nontrivial Zak-phase winding; in particular, Kane-Mele $\\mathbb{Z}_2$ order in class AII is proven equivalent to being an obstructed representation of $T^2\\times \\mathbb{Z}_4^T$."],"supporting_citations":[{"why":"Defines band representations as Wannier sets generated from finitely many symmetric Wannier centers; the object the theorem recasts in topological language.","marker":"[17,18]"},{"why":"Supplies the tetragonal tight-binding model with a rank-two gapped band and surface states covering the gap; the main fragile case study.","marker":"[53]"},{"why":"Supplies the experimentally realized hexagonal photonic crystal whose bulk and domain-wall states are reclassified as fragile.","marker":"[56]"},{"why":"Supplies the tetragonal photonic crystal realizing the same fragile obstructed representation.","marker":"[55]"},{"why":"Proposes the hexagonal rotation-invariant topological crystalline insulator whose fragility is proven by the same argument.","marker":"[36]"},{"why":"Defines the halved-mirror chirality invariant and a Zak-phase reformulation used to diagnose the photonic bands.","marker":"[67]"},{"why":"Proves the unit-rank band-representability criterion and locally-symmetric Wannier basis construction on which the higher-rank theorem builds.","marker":"[78]"},{"why":"Establishes the relation between symmetry-protected Zak phases and Wilson-loop spectra, used to certify trivial Chern class and derive the Zak winding theorem.","marker":"[20]"}],"fun_headline_variants":["Splitting theorem: band reps are permuted trivial stacks","Fragile photonic crystals: surface states are removable","Band reps split into trivial bands; photonic fragility proven","Splitting theorem simplifies band reps; photonic states removable","Band reps are permuted trivial bands; photonic fragility shown"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof that band-representable low-energy subspaces cannot have spectrally robust boundary states assumes that a finite-rank, analytic high-energy subspace $Q'$ transforming as a band representation can always be added to a continuum crystal; the paper asserts this because continuum descriptions have infinitely many bands, but does not prove it for a given Hamiltonian.","fun_headline_variants_meta":{"raw":{"variants":["Splitting theorem: band reps are permuted trivial stacks","Fragile photonic crystals: surface states are removable","Band reps split into trivial bands; photonic fragility proven","Splitting theorem simplifies band reps; photonic states removable","Band reps are permuted trivial bands; photonic fragility shown"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000476,"raw_usage":{"total_tokens":2451,"prompt_tokens":1129,"completion_tokens":1322,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":745,"completion_tokens_details":{"reasoning_tokens":1238}},"tokens_in":745,"tokens_out":1322,"duration_ms":10017,"temperature":1.0,"reasoning_tokens":1238,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:36:45.123036+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a continuum periodic Schrödinger or photonic Hamiltonian whose lowest gapped two-band subspace is a band representation, compute the surface spectrum with all symmetry-allowed high-energy bands included; if any in-gap boundary state cannot be removed by a symmetry-preserving band deformation, the boundary stability criterion fails. The specific missing check is whether the required finite-rank, band-representable $Q'$ exists in such a continuum model.","supporting_citations":[{"cited_title":"For simplicity, let us consider a rank-N BR(G, ϖ,D )","cited_arxiv_id":null,"evidence_quote":"Establishes the relation between symmetry-protected Zak phases and Wilson-loop spectra, used to certify trivial Chern class and derive the Zak winding theorem."}],"review_version":1}