{"id":"cceab7f1-060c-4fc9-b787-f864aa56a577","arxiv_id":"2505.13051","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Toroidal cycles of periodic cell complexes embed into canonical persistent local systems, enabling a proposed classification and polynomial-time computation for arbitrary periodicity.","lead":"This paper applies bisheaf theory and persistent local systems to tell apart true cycles of an infinite periodic space from fake 'toroidal' cycles created by wrapping the space into a finite box. It claims to classify these artifacts in all homology degrees for arbitrary periodicity and offers polynomial-time algorithms, but the main theorems currently prove only an embedding, not a full classification.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 1's reduction to one periodic direction is not established; the unproved 'quotient maps are injective on toroidal cycles' step is load-bearing for Theorem 2.","rationale":"The reader's weakest_assumption is exactly the concern identified here: Proposition 1 assumes that a toroidal class for the full quotient is toroidal for some intermediate 1-periodic quotient, and the proof relies on an unjustified injectivity statement. The paper's central claim, Theorem 2, is formally conditional on this proposition; without it, the product of 1-periodic PLSs need not capture every toroidal class. This is a genuine, specific, addressable gap in the argument rather than a disagreement with a consensus or a stylistic issue. The reader's CONDITIONAL verdict is therefore appropriate: the authors should either prove Proposition 1 with a rigorous lifting argument (or via the Cartan-Leray spectral sequence), weaken Theorem 2 and the abstract's 'completely classified' claim, or exhibit a counterexample. No additional fundamental objection beyond this was found; the algorithmic complexity claims in Theorem 3 appear plausible, and the examples support the 1-periodic theory, but they do not independently establish the d-periodic reduction. Thus the verdict should remain CONDITIONAL, unchanged from the reader's assessment.","tokens_in":27719,"tokens_out":10081,"duration_ms":114408,"concrete_test":"Construct a small 2-periodic CW complex whose quotient G is a 2-torus and whose universal cover is not simply connected, e.g., take the standard 2-skeleton of T^2 and add one 2-cell whose boundary is the diagonal 1-cycle of the two basis circles, repeated periodically; compute H_2(G), H_2(K̂_1), H_2(K̂_2), and H_2(K) via finite windows with boundary identifications. Check whether im(H_2(K̂_1)) ∩ im(H_2(K̂_2)) equals im(H_2(K)). If the intersection is strictly larger, Proposition 1 is false and Theorem 2 collapses. If no counterexample appears in a broad enumeration of small periodic complexes, the remaining risk is the unproved analytic step; settle it by deriving the equality im(H_•(K)) = ∩_i im(H_•(K̂_i)) from the Cartan-Leray spectral sequence of the fibre product K = K̂_1 ×_G ... ×_G K̂_d, or by exhibiting the missing injectivity argument for quotient maps on toroidal cycles.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 2 asserts that every non-zero class of H_{•+1}(G)/I_{•+1} embeds into the product of the d one-dimensional persistent local systems. The proof is 'immediate' from Theorem 1 and Proposition 1, so Proposition 1 is the sole bridge from the d-periodic setting to the 1-periodic machinery. Proposition 1 claims: if γ is toroidal for the full quotient K ↠ G, then γ is toroidal for some intermediate quotient K̂_i ↠ G obtained by killing a single translation direction. The proof's second paragraph is the critical step: after subtracting non-toroidal cycles, the author asserts that 'the quotient maps are injective on toroidal cycles' to conclude that two lifts that are homologous in G must be homologous in the lifts K̂_i. This assertion is not proven and is not a standard property of regular covering maps; deck transformations typically identify distinct homology classes in the cover that map to the same class in the base. If this assertion fails, there could exist a class [γ] ∈ H_{•+1}(G)/I_{•+1} that lies in the image of H_{•+1}(K̂_i) → H_{•+1}(G) for every i, yet does not lift to K. Such a class would be invisible to each individual Li(G), so the product embedding would map it to zero and Theorem 2 would be false. The surrounding argument about the lift supported on [0,1]^d and the linear combination ∂γ = Σ λ_i q_i does not resolve this: it treats only representatives whose boundary is a combination of the d periodicity relations and does not address the general case of arbitrary lifts or the injectivity claim. The assertion is load-bearing because without Proposition 1 there is no mechanism in the paper for showing that a toroidal class requiring simultaneous periodicity in multiple directions is detected by any one 1-periodic PLS.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a bisheaf-theoretic framework for classifying toroidal cycles in finite quotients of d-periodic cell complexes. The main claims are: (i) Theorem 1 gives a canonical embedding of the toroidal quotient H_{•+1}(G)/I_{•+1} into the persistent local system L(G) for 1-periodic complexes; (ii) Proposition 1 asserts that every toroidal class in a d-periodic quotient is already toroidal with respect to one of the d intermediate 1-periodic quotients; (iii) Theorem 2 combines these to embed H_{•+1}(G)/I_{•+1} into the product of d persistent local systems; and (iv) Section 4 gives polynomial-time algorithms for epification, monofication, and extraction of toroidal cycles from the monodromy matrices of the persistent local systems. The paper also presents several worked examples and two conjectures about the finer structure of the monodromy.","tokens_in":27968,"tokens_out":10903,"duration_ms":114126,"significance":"If the main theorems are correct, the paper would provide a genuinely useful computational classification of toroidal homology in all degrees for periodic complexes of arbitrary periodicity, extending earlier work on degree 0/1 and 1-periodic cases. The explicit algorithms, the worked examples in Section 5.2, and the use of persistent local systems as the invariant are appealing and likely to be of interest to the computational topology and topological data analysis communities. However, the central theorems are not yet rigorously established: the proof of Proposition 1 contains an unproved and nonstandard assertion, and the proof of Theorem 1 has a load-bearing gap when passing from relative to absolute homology. These issues directly affect the validity of Theorem 2 and the extraction pipeline in Section 5.1.","major_comments":[{"comment":"The proof of Proposition 1 is not rigorous and contains a load-bearing unproved assertion: in the second paragraph, the authors state that after subtracting non-toroidal cycles, homologous chains in the quotient 'must be homologous in the lifts (as the quotient maps are injective on toroidal cycles).' This is not a standard property of covering-space quotient maps, and no proof or reference is provided. The first case (lifts supported on [0,1]^d) only treats chains whose boundary is a linear combination of the period relations q_i, which is not the general situation. The later case of arbitrary lifts is dismissed in a few sentences that do not establish the claimed reduction. Since Theorem 2 is stated to follow immediately from Theorem 1 and Proposition 1, this gap undermines the central d-periodic claim. The authors should either supply a complete proof of Proposition 1 or replace Theorem 2 with a conditional statement.","section":"§3.2, Proposition 1"},{"comment":"In the first part of the proof of Theorem 1, the step 'Applying the homology functor to the embedding st(σ~),→R then implies [ζ~]=[ξ~]=0 in K' is not justified. The vanishing [ξ~]=0 is in the relative group F~(σ~)=H(K,K\\st(σ~)); a relative boundary is not necessarily zero in absolute homology. A cycle supported in st(σ~) can be a relative boundary while remaining nonzero in H(K), if it bounds through the complement. The argument as written therefore does not establish that ζ~ is a boundary in K, which is the key step in showing that γ is non-toroidal. This needs a careful homology long-exact-sequence argument or a different approach.","section":"§3.1, Theorem 1 proof"},{"comment":"The induction over the monofication kernels in the proof of Theorem 1 is asserted rather than proved. The claim that if ˙F~_σ[γ~_σ] lies in K~_j(σ) then it can be written as a sum of cycles α~_τ that are cycles in st(σ~) but homologous to a boundary in st(τ~), and that this implies γ is non-toroidal, is not derived from the definitions of Algorithm 2. Since this induction is the mechanism by which the proof passes from the cap-product vanishing to the full PLS vanishing, this is another load-bearing gap. The authors should give a precise statement of the invariant maintained by the monofication algorithm and prove the induction step.","section":"§3.1, Theorem 1 proof, induction step"},{"comment":"The complexity bound in Theorem 3 appears to be off by a factor of n. The proof says the second loop iterates at most nD times and that each cell computation costs O((c^ω+C^ω)D^ω), but each iteration of the while loop in Algorithms 1 and 2 passes over all cells of the complex. The total cost should therefore be O(n^2(c^ω+C^ω)D^{1+ω}) (plus the O(nd) first loop), not O(n(c^ω+C^ω)D^{1+ω}). Moreover, the assertion that 'each iteration of the second loop must reduce the dimension of at least one vector space' is not proved and is not an immediate consequence of the pseudocode. The polynomial-time claim is probably salvageable, but the stated bound and its proof need correction.","section":"§4.2, Theorem 3"},{"comment":"The sentence 'The 1-eigenvectors of M_i correspond exactly to the toroidal cycles of G' overstates what Theorem 1 establishes. Theorem 1 gives only an embedding ι(H_{•+1}(G)/I_{•+1}) ↪ L(G)_•; it does not assert surjectivity onto the 1-eigenspaces (or onto L(G) at all). The later Conjecture 1 explicitly leaves open the possibility that not every class of L(G) is realized by a toroidal cycle, and the anticipation that the embedding becomes an isomorphism only for some finite cover G_{i,k} is also conjectural. The extraction pipeline in Section 5.1 therefore needs a separate justification, or it should be phrased conditionally on Conjecture 1.","section":"§5.1, Extraction of toroidal cycles"}],"minor_comments":[{"comment":"The phrase 'Here, build on the work' is missing a subject; it should read 'Here, we build on the work...'.","section":"Abstract"},{"comment":"The word '1-peoridic' is a typo for '1-periodic'.","section":"Section 3 heading"},{"comment":"The notation for the lifted bisheaf is typeset inconsistently (for example 'eF e' appears in several places); this should be unified.","section":"Notation in Section 3.1"},{"comment":"The use of d both for the dimension of the cell complex and for the periodicity of the translation action is confusing; for example, 'the d-coordinate projections π_d : G→T^d' is hard to parse. A different index, such as p for periodicity, would improve readability.","section":"§4.1"},{"comment":"For a free and properly discontinuous Z^d action on R^d, the quotient is homeomorphic to T^d, not merely 'homologically equivalent to a torus'; the remark can be made more precise.","section":"Remark 6"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know about arXiv:2505.13051. It is the first paper I have seen that applies bisheaves and persistent local systems to periodic complexes, and the one-periodic core is a real advance: toroidal cycles in arbitrary homology degree are shown to embed into the PLS of the canonical quotient map, with polynomial-time epification/monofication algorithms and well-chosen examples. But the bridge from 1-periodic to d-periodic, Proposition 1, is not actually proved. The stress-test you sent me lands exactly where it should: the step asserting that 'the quotient maps are injective on toroidal cycles' is not a standard fact about regular covering maps and is not justified. If it fails, a toroidal class could be invisible to every individual 1-periodic PLS, and Theorem 2 would collapse.\n\nWhat is genuinely new is the framework: extending the degree-0/1 and 1-periodic results of [14,15] to all degrees and arbitrary Zd actions via PLSs from [10]. The examples are instructive, especially the Schwarz P surface, which shows why one needs the 1-periodic PLSs rather than just the torus-level PLS. The algorithms are concrete, the complexity bound is honestly flagged as a proof of concept, and the authors separate their conjectures from their theorems. The citation pattern is fine; building on their own prior work is expected here, and MacPherson–Patel is properly credited.\n\nSoft spots, in proportion: Proposition 1 is the main one. The proof splits into three regimes, and each subsequent regime relies on the previous plus the unproved injectivity claim; the final paragraph about arbitrary lifts is especially compressed. If the claim is true, it needs a real proof; if false, the d-periodic embedding theorem should be weakened or replaced. Theorem 1 also has a 'changing bases' step and an induction over kernels that are sketched rather than fully demonstrated. Section 5.1 states that the 1-eigenvectors of the monodromy matrix correspond exactly to toroidal cycles, but the correspondence is asserted, not proven, and the later Conjecture 1 hedges the same territory. Minor: the abstract promises a 'complete classification' while the theorems deliver embeddings; that overreach should be fixed or softened.\n\nNone of this makes me think the approach is wrong. The 1-periodic theorem is coherent, the examples are consistent, and the d-periodic claim is plausible. This is the kind of paper a serious referee should evaluate: the central idea is important enough that the missing proof is worth demanding. I would send it to peer review, with instructions to focus on Proposition 1, the eigenvector-to-cycle correspondence in Section 5.1, and the exact scope of the abstract. If those get resolved, this will be a useful paper for applied topologists working on crystals, periodic point clouds, and cosmology data.","headline":"A genuinely new bridge from bisheaves/persistent local systems to periodic complexes, with a solid 1-periodic core but a load-bearing d-periodic reduction (Proposition 1) that is asserted rather than proved.","tokens_in":28596,"tokens_out":2978,"would_cite":true,"duration_ms":35598,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["55N31","55N30"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper establishes that toroidal cycles of a finite quotient of a periodic cell complex are classified by persistent local systems built from bisheaves.","keywords":["persistent homology","periodic cell complexes","toroidal cycles","bisheaves","persistent local systems","epification","monofication","covering spaces"],"falsifier":"Search among 2-periodic cubical complexes for a toroidal class in $H_2(G)/I_2$ whose cap product with the orientation class of each coordinate circle is zero; such a class would not be visible in any coordinate persistent local system, contradicting Theorem 2.","tokens_in":27430,"feed_emoji":"🔄","tokens_out":11082,"duration_ms":99716,"temperature":0.7,"pith_summary":"Studying an infinite periodic complex through a finite quotient with periodic boundary conditions creates artificial homology classes, which the authors call toroidal cycles. The paper sets out to show that these classes, in every degree and for arbitrary-dimensional periodicity, are completely classified by persistent local systems coming from bisheaves, and that the classification is computable. The central results are an embedding of $H_{\\bullet+1}(G)/I_{\\bullet+1}$ into the persistent local system for one-periodic quotients, and an embedding into a product of $d$ such systems for $d$-periodic quotients. The accompanying algorithms compute epification and monofication, and hence the canonical persistent local system, in polynomial time. If the arguments hold, detecting toroidal cycles reduces to finite linear algebra on a single quotient space.","feed_headline":"Toroidal cycles in periodic spaces are classified by local systems","feed_subtitle":"Separates periodic-boundary artifacts from true cycles using bisheaf invariants computed in polynomial time.","key_machinery":"The machinery is the cellular bisheaf: a pair consisting of a sheaf (contravariant data on open stars) and a cosheaf (covariant data on preimages), linked cell-wise by cap products with a fixed orientation class of the target torus. Isobisheafification—taking the maximal sub-episheaf and the minimal quotient-monocosheaf—produces a canonical persistent local system, whose cell-wise maps form a locally constant cosheaf. For one-periodic quotients, covering space theory makes the bisheaf over the circle locally isomorphic to a lift over the real line, so the persistent local system records exactly which homology classes survive lifting. The $d$-periodic case runs the same construction for each coordinate projection $G \\to S^1$ and takes the product.","core_discovery":"The paper claims that for a $d$-periodic cell complex $K$ with quotient $G$, every toroidal homology class of $G$—one created by periodic boundary conditions and not lifting to a cycle in $K$—appears canonically in a persistent local system built from the bisheaf of the quotient map. In the one-periodic case this is an embedding of $H_{\\bullet+1}(G)/I_{\\bullet+1}$ into $L(G)_{\\bullet}$ (Theorem 1); in the $d$-periodic case it is an embedding into a product of $d$ persistent local systems, one for each independent translation direction (Theorem 2), using the reduction that it suffices to check one periodic direction at a time (Proposition 1). The result is stated over any field and covers all homology degrees.","pith_inferences":["If the embeddings hold, toroidal-cycle detection becomes a monodromy computation: the 1-eigenspaces of the persistent local system monodromy matrices enumerate classes that wrap around each translation direction, and the paper's Conjecture 1 would then imply that all eigenvalues are roots of unity, giving toroidal information for all finite covers from one small quotient.","The one-coordinate-at-a-time reduction suggests a general principle: high-dimensional periodicity behaves as a product of one-dimensional monodromies, and the same construction might extend to free actions of other abelian groups, with the target torus replaced by the group's classifying space.","The local quotient-independence seen in one-periodic bisheaves weakens in higher periodicity, as the paper's triply periodic minimal surface example shows; whether a canonical choice of quotient can make the persistent local systems fully quotient-independent is a question the paper leaves open."],"forward_implications":["For a one-periodic complex, every nonzero class in $H_{\\bullet+1}(G)/I_{\\bullet+1}$ appears as a nontrivial class of the persistent local system $L(G)_{\\bullet}$, so no toroidal cycle is lost.","For a $d$-periodic complex, every toroidal class is detected by at least one of the $d$ coordinate persistent local systems, so no infinite computation over $K$ is needed to separate toroidal from true cycles.","Algorithms 1 and 2 terminate and compute epification and monofication in polynomial time, making isobisheafification and the persistent local system computable for finite quotients.","The 1-eigenvectors of the monodromy matrix $M_i$ enumerate toroidal cycles of $G$, and the 1-eigenvectors of $M_i^k$ enumerate toroidal cycles of the $k$-fold cover $G_{i,k}$.","Non-toroidal cycles are represented trivially in the persistent local system, so the persistent local system separates the two cycle types."],"supporting_citations":[{"why":"Supplies the bisheaf machinery, isobisheafification, and the definition of persistent local systems into which toroidal classes are embedded.","marker":"[10]"},{"why":"Provides the cellular bisheaf conventions and canonical stratifications used throughout the paper.","marker":"[13]"},{"why":"Earlier work on degree-0 and degree-1 homology of periodic cell complexes that this paper extends to all degrees.","marker":"[14]"},{"why":"Earlier study of one-dimensional periodicity with finite windows that motivates the one-periodic case and is built on here.","marker":"[15]"},{"why":"Provides the abelian-category facts used to prove that the image of an isobisheaf is a persistent local system and that epification and monofication are well-defined.","marker":"[7]"}],"fun_headline_variants":["Toroidal cycles classified via persistent local systems","Periodic boundary artifacts caught by bisheaf invariants","All toroidal cycles revealed by persistent local systems","Polynomial-time classification of periodic-space cycles","Bisheaves separate true cycles from boundary artifacts"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The d-periodic classification depends on Proposition 1, which assumes that a toroidal cycle of the full quotient is toroidal in at least one intermediate 1-periodic quotient; the proof uses an unproven injectivity assertion about quotient maps on toroidal cycles, and if that fails the product embedding of Theorem 2 loses its foundation.","fun_headline_variants_meta":{"raw":{"variants":["Toroidal cycles classified via persistent local systems","Periodic boundary artifacts caught by bisheaf invariants","All toroidal cycles revealed by persistent local systems","Polynomial-time classification of periodic-space cycles","Bisheaves separate true cycles from boundary artifacts"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000304,"raw_usage":{"total_tokens":1743,"prompt_tokens":936,"completion_tokens":807,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":552,"completion_tokens_details":{"reasoning_tokens":735}},"tokens_in":552,"tokens_out":807,"duration_ms":7354,"temperature":1.0,"reasoning_tokens":735,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T20:20:56.689103+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Search among 2-periodic cubical complexes for a toroidal class in $H_2(G)/I_2$ whose cap product with the orientation class of each coordinate circle is zero; such a class would not be visible in any coordinate persistent local system, contradicting Theorem 2.","supporting_citations":[{"cited_title":"Canonical stratifications along bisheaves","cited_arxiv_id":null,"evidence_quote":"Provides the cellular bisheaf conventions and canonical stratifications used throughout the paper."},{"cited_title":"Abelian categories, volume 1964","cited_arxiv_id":null,"evidence_quote":"Provides the abelian-category facts used to prove that the image of an isobisheaf is a persistent local system and that epification and monofication are well-defined."}],"review_version":1}