{"id":"6e5651b2-6831-4812-bead-86fd8d835b9e","arxiv_id":"2509.01927","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Generic potentials on any connected periodic graph produce no flat bands.","lead":"This paper proves that on any connected periodic graph, a generic choice of on-site potentials eliminates flat bands from the spectrum of a discrete Schrödinger-type operator. The result answers an open problem in spectral theory and clarifies when infinite-multiplicity eigenvalues can appear in periodic lattice models.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3.1 Step 6 fails for ℓ=1, leaving the combinatorial core of the proof incomplete.","rationale":"The reader correctly identifies Theorem 3.1 as the load-bearing combinatorial step, but points mainly to Lemma 3.2. My reading finds a more concrete and earlier gap in Step 6. For ℓ=1 the replacement path has the same length and the same footprint as Q, so the claimed extremality contradiction does not follow. Because L is even in this case, the second alternative in Theorem 3.1 is parity-impossible for a symmetric extremal loop of length L+1, so the first alternative must cover it. Since Step 6 is where multi-edge ambiguities are ruled out, the proof of Theorem 3.1 is incomplete as written. I do not claim the main theorem is false; rather, the paper should remain conditional until this case is either repaired or excluded by additional argument. The reader's verdict of CONDITIONAL is therefore unchanged, but for a different reason than the one emphasized in the reader's weakest_assumption.","tokens_in":18790,"tokens_out":33365,"duration_ms":376601,"concrete_test":"Perform a brute-force verification of Theorem 3.1 for small connected periodic graphs: enumerate all connected d=1 graphs with N≤4 and quasimomentum sets A⊂{0,1,2} satisfying (p1)-(p4), compute the minimal extremal j-loop length L, and for every extremal loop of form (3.2) with ℓ=1 and two quasimomenta β_1≠β'_1 between n_s and m_1, list all loop configurations with the same footprint and quasimomentum. If any such footprint/quasimomentum class contains more than one configuration and no symmetric extremal loop of length L+1 exists, Theorem 3.1 has a concrete counterexample. Alternatively, test whether the original extremal loop can be canceled by the replacement loop while all other extremal loops also cancel.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central reduction (Corollary 2.12) depends entirely on Theorem 3.1 producing a non-cancelable extremal loop or a non-cancelable symmetric extremal loop of length L+1. In Step 6 of the proof, the authors consider a multi-edge in the non-repeated part Q = n_s -> m_1 -> ... -> m_ℓ -> n_s and claim that if β_i ≠ β'_i are two possible quasimomenta on an edge with i < ℓ/2 + 1, then the replacement path is either shorter than Q or has the same length and more repetitions. For ℓ = 1 and i = 1, this is false: the replacement path n_s --β_1--> m_1 --(-β'_1)--> n_s has exactly the same length as Q (2 edges) and its footprint is still {m_1}, so it creates no extra repetition. Adding back the symmetric part gives another extremal loop of the same length and footprint, not a contradiction. This is not cosmetic: when ℓ=1 the extremal loop in (3.2) has even length L=2s+2, while any symmetric extremal loop has even length, so the second case of Theorem 3.1 cannot supply a length-L+1 replacement. Thus the proof must show all such extremal loops are non-cancelable, and Step 6 is the only argument offered for that; it does not cover this case. The main theorem may still be true, but the combinatorial route given in the paper is incomplete.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"Using the Floquet transform, the paper reduces the absence of flat bands for generic potentials on a connected Z^d-periodic graph to a perturbative statement: for each eigenvalue branch λ_j, after rescaling the hopping by ε, one must find a non-cancelable loop configuration with nonzero quasimomentum (§2). The analytic part is careful: the Rayleigh–Schrödinger series is absolutely convergent in a suitable region, constant dependence on z is detected coefficientwise, and elimination theory upgrades an open set of good potentials to a proper algebraic variety bound. The remaining work is the combinatorial Theorem 3.1, which asserts that extremal loop configurations either are non-cancelable or produce a non-cancelable symmetric extremal loop one order longer. The proof of that theorem is the main technical content.","tokens_in":19168,"tokens_out":12577,"duration_ms":146307,"significance":"If Theorem 3.1 is repaired, the paper resolves [33, Problem 2] for fixed hopping and generic potential, and gives Zariski-open genericity rather than merely measure-zero. The analytic reduction via perturbation series and elimination theory is elegant; the explicit next-order construction of symmetric extremal loops is potentially useful. The paper does not rely on fitted quantities or circular input; the perturbation-series formula from [13] is a published, independent technique.","major_comments":[{"comment":"The proof of Theorem 3.1 is incomplete in the case ℓ=1. For ℓ=1 and i=1, the replacement path n_s --β_1--> m_1 --(-β'_1)--> n_s has exactly the same length as Q and the same footprint {m_1}; it creates no extra repetition, so the stated contradiction with extremality does not arise. Moreover, when ℓ=1 the original extremal loop has length L=2s+2, which is even, while any symmetric extremal loop has even length (one vertex of multiplicity one and all others of multiplicity two give |P|=2r+2). Hence the second alternative of Theorem 3.1 cannot supply a non-cancelable loop of length L+1. This is a load-bearing gap because Corollary 2.12 depends entirely on Theorem 3.1.","section":"§3.2, Step 6"},{"comment":"Lemma 3.2 is a structural statement about permutations preserving inversion distances, but its proof is only a two-sentence sketch. Step 8 relies on this lemma to conclude that every alternative arrangement of the non-repeated part is obtained by mirror flips of non-overlapping segments; without a complete proof of Lemma 3.2 that conclusion is unsupported. The lemma is probably true, but the proof needs to be written in full (e.g., by induction on the first element moved).","section":"§3.2, Steps 7–8 and Lemma 3.2"}],"minor_comments":[{"comment":"The statement reads 'H it has a flat band'; this should be 'H has a flat band'.","section":"Theorem 1.4"},{"comment":"The set is denoted B_r, but the following sentence says 'Clearly, V_r is an open subset of C^N'. The symbol should be B_r.","section":"Eq. (2.3)"},{"comment":"'polynominals' should be 'polynomials'.","section":"Remark 1.5"},{"comment":"The notation P^{-1}_1 is used without definition; it should be defined explicitly as the reverse path.","section":"§3.2, Step 3"},{"comment":"References [26] and [28] appear to be the same article, as do [25] and [30]. Please consolidate or distinguish them.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The main gap is serious but localized to the combinatorial proof of Theorem 3.1, specifically the ℓ=1 case. I see no novelty or scope problems; a repair of Step 6 for that case would make the argument plausible. The rest of the reduction appears sound."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe short version: this paper probably has the right theorem—generic potentials kill flat bands on connected periodic graphs—and the analytic reductions in Section 2 are clean and convincing. But the combinatorial core, Theorem 3.1, is not fully proved as written. The stress-test note found a concrete gap in Step 6 that I checked and think is correct.\n\nWhat's good: the reduction from flat bands to non-constant Laurent coefficients in Rayleigh-Schrödinger series, and then to the vanishing of totalcont, is elegant. The extremal loop idea is genuinely new, and the step from an open dense set of potentials to a Zariski-closed bad set via elimination theory is neat. This answers Sabri–Youssef's Problem 2 and is not a special case of the companion paper [5], since edge weights are fixed. The analytic reductions—Floquet theory, convergence of the series, treating W-factors as formal variables—look solid. I don't see circularity beyond the legitimate use of the published perturbation-series formalism from [13].\n\nWhere it falls short: the proof of Theorem 3.1. In Step 6, the case ℓ=1 (a single non-repeated vertex in the non-repeating part) is not covered by the claimed 'shorter or more repetitions' dichotomy. If the edge between n_s and m_1 has two quasimomenta β_1 and β'_1, the replacement n_s -> m_1 -> n_s has the same length and the same footprint, so a contradiction does not follow. And since any symmetric extremal loop in the paper's sense has even length (ℓ=1 forces length 2s+2), the second arm of Theorem 3.1—a non-cancelable symmetric extremal loop of length L+1—cannot apply when L is even. So the proof leaves open the possibility of complete cancellation in exactly this case. I didn't find a fix in the text; the Section 3.3 parallelogram/max-norm argument is applied only to the symmetric loop constructed in Step 6, not to this ℓ=1 extremal loop itself. It's likely the argument can be repaired by applying that trick directly, but as written the proof is incomplete.\n\nThe sketch of Lemma 3.2 is also thin, but that's a smaller concern.\n\nBottom line: I'd send this to a serious referee, with the instruction to push hard on Theorem 3.1, especially the ℓ=1 case. If the proof is fixable, this is a strong paper.\n\nBest.","headline":"The paper likely proves the right theorem, but the combinatorial core (Theorem 3.1) has a real gap for ℓ=1 that the stress-test note correctly identifies.","tokens_in":19589,"tokens_out":9175,"would_cite":false,"duration_ms":91292,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["47B39","05C50"],"pacs":[],"model":"deepseek-v4-flash","headline":"For Schrödinger-type operators on any discrete connected periodic graph, potentials that produce a flat band at some energy form a proper algebraic subvariety of potential space; hence generic potentials have no flat bands.","keywords":["flat bands","periodic graphs","generic potentials","perturbation series","loop configurations","Bloch variety","Floquet theory","algebraic subvariety"],"falsifier":"Search for the smallest N and permutation σ of {1,...,N} with |σ(i)−σ(j)|=|i−j| for every inversion i<j, σ(i)>σ(j), where σ is not a union of interval identities and reflections; Lemma 3.2 says none exists, so any example would refute the combinatorial core. Alternatively, numerically compute the full perturbation coefficients for a specific connected periodic graph and find a potential outside the algebraic exceptional set for which a band function is identically constant.","tokens_in":18737,"feed_emoji":"📐","tokens_out":5251,"duration_ms":60329,"temperature":0.7,"pith_summary":"This paper proves that flat bands—spectral bands that collapse to a single energy and produce eigenvalues of infinite multiplicity—are exceptional for Schrödinger-type operators on discrete connected periodic graphs. With the graph structure and edge weights fixed and only the on-site potential allowed to vary, the theorem says the potentials that produce a flat band at any energy form a thin algebraic set; both Zariski-open and Euclidean-open-dense sets of complex potentials, and an open-dense set of real potentials, are free of flat bands. Previous examples showed flat bands occur for special potentials, and the paper resolves an open problem by showing that such examples are non-generic. The proof reduces the question to perturbation theory in the small-hooping regime, then to a combinatorial statement about loop configurations in the quotient graph, and finally to an elimination-theory step that converts an open set of good potentials into genericity.","feed_headline":"Almost every potential kills flat bands on periodic graphs","feed_subtitle":"A perturbation-series proof shows only a thin algebraic set of potentials can produce constant spectral bands.","key_machinery":"Rayleigh–Schrödinger perturbation series expanded in small hopping, encoded as loop configurations—closed paths in the quotient graph with quasimomentum labels, possibly with attached loops. The argument isolates extremal loop configurations (minimal length and footprint among loops with nonzero quasimomentum) and proves Theorem 3.1: each extremal loop is either itself non-cancelable (the only configuration with its footprint and quasimomentum), or can be modified into a non-cancelable symmetric extremal loop of length L+1. That theorem, together with the algebraic-geometric closure theorem, converts a single nonvanishing perturbative term into generic absence of flat bands.","core_discovery":"The central claim is Theorem 1.4: fix a Zd-periodic graph satisfying natural finiteness conditions (finite fundamental domain, finite degree, weak symmetry of edge weights) with connected quotient/infinite graph. Then the set of potential tuples (V1,...,VN) ∈ C^N for which the operator has a flat band at some energy E is contained in a proper affine algebraic subvariety of C^N. Equivalently, absence of flat bands holds on a Zariski-open, hence Euclidean open dense, set of potentials; the same holds for real potentials. The key mechanism is perturbation theory: after rescaling so hopping is small, the claim that an eigenvalue band is constant in quasimomentum is equivalent to a nonvanishing o","pith_inferences":["A natural testable extension is quantitative: estimate the degree and number of defining equations of the exceptional algebraic variety in terms of N, d, and edge-range, which would yield a measure of how rare flat-band potentials are.","The combinatorial core suggests a purely graph-theoretic characterization: flat bands persist on a codimension-one set only when the quotient graph has symmetric multi-edge structures (like the Lieb lattice's V2 = V3 condition); one could investigate whether all extremal cancellations arise from such symmetries.","The perturbation argument may extend to other matrix families with Laurent-polynomial entries beyond graph Schrödinger operators, such as tight-binding models with magnetic phases, where the same loop-configuration cancellation question determines whether nondegenerate bands are generic.","A reader could test the key Lemma 3.2 independently: search for permutations of {1,...,N} satisfying |σ(i)-σ(j)| = |i-j| for all inversions but not decomposing into interval reflections; finding one would pinpoint the exact fragility of the proof."],"forward_implications":["If the theorem is correct, any flat band in a discrete periodic graph is a codimension-at-least-one phenomenon in potential space, not a structural necessity of the graph.","For self-adjoint operators, the theorem removes the only mechanism that produces eigenvalues of infinite multiplicity; every spectral band is a genuine interval for a generic potential.","It answers the question posed in [33, Problem 2] affirmatively: fixing edge weights, a generic potential alone suffices to destroy flat bands.","The connectedness assumption can be relaxed: the conclusion holds for any graph with no finite connected components, since finite components are handled separately and trivially.","Because the exceptional set is described by finitely many polynomial equations, any small perturbation of a generic potential also has no flat bands."],"supporting_citations":[{"why":"Supplies the Rayleigh–Schrödinger perturbation series and the loop-configuration formalism that the cancellation analysis uses.","marker":"[13]"},{"why":"Poses the problem (Problem 2) that Theorem 1.4 answers and provides examples and background on flat bands in periodic graphs.","marker":"[33]"},{"why":"Provides the algebraic-geometry closure theorem used to pass from an open set of good potentials to a proper algebraic subvariety.","marker":"[4]"},{"why":"Supplies the Floquet-theory and direct-integral framework used to define flat bands and link them to constant band functions.","marker":"[20]"},{"why":"Proves the related genericity statement when both edge weights and potentials vary, setting the context for the fixed-weights result.","marker":"[5]"}],"fun_headline_variants":["Generic potentials abolish flat bands on periodic graphs","Flat bands vanish for almost all potentials on periodic graphs","Almost all potentials eliminate flat bands in periodic graphs","Generic potentials rule out flat bands on periodic graphs","Flat bands are rare: generic potentials forbid them on periodic graphs"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The load-bearing premise is the combinatorial claim in Section 3 that every shortest non-trivial loop path either is unique for its footprint and quasimomentum or can be lengthened by one step into such a unique path, and this is the step that must be trusted.","fun_headline_variants_meta":{"raw":{"variants":["Generic potentials abolish flat bands on periodic graphs","Flat bands vanish for almost all potentials on periodic graphs","Almost all potentials eliminate flat bands in periodic graphs","Generic potentials rule out flat bands on periodic graphs","Flat bands are rare: generic potentials forbid them on periodic graphs"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000456,"raw_usage":{"total_tokens":2018,"prompt_tokens":528,"completion_tokens":1490,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":272,"completion_tokens_details":{"reasoning_tokens":1416}},"tokens_in":272,"tokens_out":1490,"duration_ms":13009,"temperature":1.0,"reasoning_tokens":1416,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T12:05:05.286573+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Search for the smallest N and permutation σ of {1,...,N} with |σ(i)−σ(j)|=|i−j| for every inversion i<j, σ(i)>σ(j), where σ is not a union of interval identities and reflections; Lemma 3.2 says none exists, so any example would refute the combinatorial core. Alternatively, numerically compute the full perturbation coefficients for a specific connected periodic graph and find a potential outside the algebraic exceptional set for which a band function is identically constant.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Rayleigh–Schrödinger perturbation series and the loop-configuration formalism that the cancellation analysis uses."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Poses the problem (Problem 2) that Theorem 1.4 answers and provides examples and background on flat bands in periodic graphs."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the algebraic-geometry closure theorem used to pass from an open set of good potentials to a proper algebraic subvariety."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Floquet-theory and direct-integral framework used to define flat bands and link them to constant band functions."},{"cited_title":"Rare Flat Bands for Periodic Graph Operators","cited_arxiv_id":"2503.03632","evidence_quote":"Proves the related genericity statement when both edge weights and potentials vary, setting the context for the fixed-weights result."}],"review_version":1}