{"id":"ff293d84-b207-42f9-b891-2b1d94c039cc","arxiv_id":"2506.08685","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Sheaves of modules are exactly the J-saturated presheaves, every Grothendieck topology on a noetherian EI directed category is rigid, and all topologies on type N/Z categories are classified.","lead":"This paper shows that sheaves of modules on a ringed site can be described purely in terms of a torsion functor: a presheaf is a sheaf exactly when it is saturated, or perpendicular, to torsion presheaves. It also classifies when all Grothendieck topologies on a directed category are subcategory topologies, with applications to infinite subcategories of FI and VI.","discovery_kind":"unification","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified.","rationale":"The paper's main homological characterization of sheaves as J-saturated modules is proved in detail and the argument is coherent: the torsion radical is left exact, the hereditary torsion pair is established from the Grothendieck topology axioms, and the injective-hull argument correctly reduces the right-perpendicular condition to vanishing of R^1T_J. The subsequent identification Sh(C^op,O) ≃ O-Mod/T(J) is a standard consequence once T(J) is localizing. I therefore do not see a load-bearing defect in the central claim. The reader's weakest assumption about finite endomorphism monoids is a genuine structural limitation of the rigidity classification, but it is explicit, isolated to the only-if direction of Theorem 1.4, and not a correctness failure of the paper's stated theorem. The more concrete issues I found are in peripheral parts: the false 'more explicitly' shape claim in Proposition 6.3 and the misapplication of Theorem 5.7 in Example 7.14 to an artinian category instead of a noetherian one. These do not overturn the classification parameterization or the main torsion-theoretic results, so the reader's conditional verdict remains appropriate.","tokens_in":36864,"tokens_out":36862,"duration_ms":455602,"concrete_test":"Instantiate C=N as a type-N category with d(2k)=1 and d(2k+1)=0, and define J(n) to consist of the sieves S(n,r) for r≤d(n). Verify the maximal, stability, and transitivity axioms directly; if J is a Grothendieck topology, Proposition 6.3's 'more explicitly' decomposition is false, confirming the concern.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim Theorem 1.1/4.2 appears sound: the proof that T(J) is a hereditary torsion class, the injective-hull argument for the equivalence of saturation with right perpendicularity, and the sheafification argument closing the loop are internally consistent, and the Serre quotient equivalence follows once T(J) is recognized as localizing. I found no load-bearing flaw in this part. The reader's flagged finiteness assumption limits only Theorem 1.4, and the paper explicitly acknowledges that limitation. Two non-central issues should still be corrected: (a) Proposition 6.3's 'more explicitly' decomposition is false — the function d(2k)=1, d(2k+1)=0 on N satisfies the stated recurrence, and the rule J(n)={S(n,r) | r≤d(n)} is a valid generic Grothendieck topology, yet it is not a single finite block followed by zeros or an infinity tail; the parameterization by functions d remains correct, so this is an error in the stated shape, not in the classification. (b) Example 7.14 applies Theorem 5.7 to an artinian EI category, namely the orbit category of the Prufer p-group, but Theorem 5.7 requires noetherian; the asserted rigidity of every topology on that orbit category is unsupported as written.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a torsion-theoretic framework for sheaves of modules on a ringed site (C^op, O). For a Grothendieck topology J, the authors define a J-torsion subfunctor T_J on O-Mod and prove that (T(J), F(J)) is a hereditary torsion pair; they then characterize sheaves of modules as J-saturated O-modules, equivalently as modules right perpendicular to all J-torsion modules (Theorem 1.1). This yields the Serre-quotient equivalence Sh(C^op, O) ≃ O-Mod/T(J) (Corollary 1.2). In the second part, the paper studies Grothendieck topologies on directed categories: Theorem 1.4 asserts that, when all endomorphism monoids are finite, every topology on C^op is rigid if and only if C is a noetherian EI category. The paper also classifies all topologies on EI categories of type N and Z (Theorem 1.7) and derives applications to infinite full subcategories of FI and VI_q, including noetherianity, local self-injectivity, and Serre-quotient descriptions (Theorem 1.8).","tokens_in":37048,"tokens_out":6794,"duration_ms":73507,"significance":"The homological characterization in Theorem 1.1 is a substantial and useful bridge between sheaf theory and representation theory; it extends the authors' earlier atomic-site result [8] to arbitrary Grothendieck topologies and gives an explicit description of the localizing subcategory in the sheafification localization. The proofs of the central theorem are detailed and internally consistent, and the paper is careful to state the finiteness assumption in Theorem 1.4 and to note in Remark 5.14 where it is used. The classifications and applications to FI/VI_q are interesting and give falsifiable, checkable statements. The main reservation is that two non-central claims—the 'more explicitly' decomposition in Proposition 6.3 and the application in Example 7.14—are incorrect as stated and need correction.","major_comments":[{"comment":"The example asserts that for the orbit category of an artinian group such as the Prüfer p-group, 'by Theorem 5.7 every Grothendieck topology J on it is rigid.' Theorem 5.7, however, gives this conclusion only for noetherian EI categories (under the finite-endomorphism hypothesis); its necessary direction even shows that if every topology is rigid then C is noetherian. An artinian EI category that is not noetherian (the Prüfer p-group orbit category is such) need not have all topologies rigid, and the dense topology is a natural counterexample. This application should be re-examined and either restricted to a noetherian setting or proved directly.","section":"§7.2, Example 7.14"},{"comment":"The 'More explicitly' description of the sequences d is false. The function d(2k)=1, d(2k+1)=0 satisfies the condition 'if d(n)≠0 then d(n+1)=d(n)−1' and therefore defines a generic Grothendieck topology, but the sequence 1,0,1,0,... cannot be written as a block of zeros followed by at most one finite descending block (r,r−1,...,1,0) and then an optional ∞ tail. The parameterization by functions d is correct, but the claimed normal form and its analogue in Proposition 6.8 must be corrected or removed.","section":"§6, Proposition 6.3"}],"minor_comments":[{"comment":"There are several typographical errors, including 'Noth that Sy = C(y, −)' in the proof of Corollary 5.17, 'T ransitivity' in the proof of Proposition 6.3, 'wight' for 'weight' near the end of the paper, and 'Pr¨ uferp-group' missing a space in Example 7.14.","section":"Throughout"},{"comment":"The abstract and introduction describe Theorem 1.8 as extending properties of 'F and VI', but the theorem itself states FI and VI_q; please make the notation consistent.","section":"Abstract and §1"},{"comment":"The term 'generic Grothendieck topology' is used informally in Section 6; a formal definition would improve clarity, since the term appears in Proposition 6.3 and Corollary 6.7.","section":"§6"},{"comment":"The proof of Proposition 6.6 is omitted with the note that it is similar to Proposition 6.3; for the non-generic case, the transitivity axiom requires a case analysis involving the empty sieve, so at least a sketch of that check would be helpful.","section":"§6, Proposition 6.6"},{"comment":"The dense topology is denoted Jd in Section 2 and also appears as Jd in Section 7.1 after Lemma 7.1; these two uses could be confused, and a different symbol for one of them would be preferable.","section":"§7.1"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear [colleague],\n\nThe headline: this paper delivers a clean torsion-theoretic characterization of sheaves of modules on ringed sites—an O-module is a sheaf iff it is J-saturated, i.e., T_J(V)=0=R^1T_J(V)—and consequently Sh(C^op,O) ≃ O-Mod/T(J). That result is new for arbitrary Grothendieck topologies, extends the authors' atomic-site work, and the proof looks solid. The paper also proves a rigidity theorem for directed categories: with finite endomorphism monoids, every Grothendieck topology is rigid iff the category is noetherian EI. That generalizes earlier poset/quiver results and is a genuine step forward. The applications to infinite full subcategories of FI and VI—noetherianity, local self-injectivity, finite injective dimension—are useful and non-obvious, especially because the shift functor does not exist for arbitrary infinite subcategories.\n\nNow the soft spots, in increasing order of seriousness.\n\nFirst, the abstract says 'artinian EI category' where the theorem says 'noetherian EI category.' Clear typo, but it will confuse readers.\n\nSecond, Proposition 6.3's 'more explicitly' decomposition of functions d is false as stated. The condition 'if d(n)≠0 then d(n+1)=d(n)−1' permits d = 1,0,1,0,1,0,...: at every even n, d(n)=1 and d(n+1)=0; at odd n, d(n)=0 imposes nothing. Such a sequence is not a finite block followed by zeros or an ∞ tail. The main parameterization by functions d remains correct; only the claimed shape is wrong.\n\nThird, Example 7.14 applies Theorem 5.7 to the orbit category of the Prüfer p-group, calling it artinian. But Theorem 5.7 requires noetherian. The Prüfer p-group's subgroup lattice contains an infinite ascending chain, so its orbit category is not noetherian. The asserted rigidity of every topology on that site is therefore unsupported as written.\n\nFourth, the type Z classification (Remark 6.10) is stated without proof; the authors say 'the reader can mimic our approach.' That is acceptable for a remark, but for a theorem it needs more.\n\nThe finiteness of C(x,x) in Theorem 1.4 is a real limitation, but the authors flag it, and the if direction works without it.\n\nWho is this for: people working in topos-theoretic representation theory, torsion theories, and representation stability. The core torsion-theoretic result is worth having, and the rigidity theorem is the right kind of classification. I would send this to a serious referee. The referee should not desk-reject it; they should ask for fixes to the above issues and a proof or explicit caveat for the type Z case.\n\nMy own verdict is that the central claims are correct and important, but the current version has enough loose ends—especially the false decomposition and the misapplied theorem—that I would not yet treat it as definitive.","headline":"A solid, important torsion-theoretic characterization of sheaves and a noetherian rigidity theorem, marred by a few genuine but fixable errors (abstract typo, wrong shape claim in Prop 6.3, and a misapplied theorem in Example 7.14).","tokens_in":37597,"tokens_out":5093,"would_cite":true,"duration_ms":53707,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["18F10","18E40","18E35","18G10"],"pacs":[],"model":"deepseek-v4-flash","headline":"Sheaves of modules are exactly the torsion-saturated presheaves.","keywords":["Grothendieck topologies","sheaves of modules","torsion pairs","Serre quotients","EI categories","rigid topologies","representation stability","directed categories"],"falsifier":"On a one-object directed category whose endomorphism monoid is the monoid of all surjections of an infinite set, check directly whether the dense topology and the maximal topology are the only two Grothendieck topologies; if a third topology exists, the finite-monoid dichotomy in Lemma 5.9 breaks down and the necessity direction of the rigidity theorem would fail without the finiteness condition.","tokens_in":36631,"feed_emoji":"🕸️","tokens_out":8277,"duration_ms":103072,"temperature":0.7,"pith_summary":"The paper proves that sheaf theory on a ringed site can be recast as torsion theory in the category of presheaves of modules. For any Grothendieck topology J, a presheaf of modules is a sheaf exactly when it is J-saturated: it has no J-torsion submodule and its first derived torsion functor vanishes. Equivalently, a sheaf is a module right perpendicular to every J-torsion module, meaning both Hom and $Ext^{1}$ to torsion modules vanish. The sheaf category is therefore the Serre quotient of the category of all O-modules by the J-torsion modules. The paper also shows that on directed categories with finite endomorphism monoids, all Grothendieck topologies are rigid whenever the category is noetherian and EI, which makes every sheaf category equivalent to a presheaf category over a full subcategory; it then classifies all topologies on a large family of such categories and transfers finiteness and injectivity results from FI and VI to their infinite full subcategories.","feed_headline":"Sheaves are exactly the torsion-saturated presheaves","feed_subtitle":"A torsion-theoretic test replaces matching families and makes sheaf categories Serre quotients.","key_machinery":"The load-bearing object is the torsion functor T_J, which sends an O-module V to the submodule generated by all elements killed by some covering sieve in J; an element v in V_x is J-torsion when a covering sieve S ∈ J(x) sends v to 0. The saturation condition T_J(V)=0 and $R^{1}$T_J(V)=0 is what characterizes sheaves, and it is shown to be equivalent to being right perpendicular to all J-torsion modules. For the rigidity results, the carrying mechanism is the family of minimal covering sieves S_x: a Grothendieck topology on a noetherian EI category is determined by a consistent family satisfying S_x = ⋃_{y≠x} S_y ∘ C(x,y), and rigidity means each S_x is generated by morphisms to J-irreducible objects.","core_discovery":"The central claim is a homological characterization of sheaves of modules on a ringed site: an O-module V is a sheaf if and only if it is J-saturated, meaning T_J(V)=0 and $R^{1}$T_J(V)=0, where T_J is the left exact endofunctor sending V to its maximal J-torsion submodule. This is equivalent to V being right perpendicular to every J-torsion module, and it yields the equivalence Sh(C^op,O) ≃ O-Mod / T(J). The paper further claims that for a directed category C whose endomorphism monoids are all finite, every Grothendieck topology on C^op is rigid if and only if C is a noetherian EI category, so sheaf categories reduce to presheaf categories over the full subcategory of J-irreducible objects. For EI categories of type N or Z, all Grothendieck topologies are explicitly classified by sequences d satisfying d(n) ≠ 0 implies d(n+1) = d(n) - 1, and non-rigid topologies are almost atomic.","pith_inferences":["The Serre-quotient description gives a practical recipe the paper leaves implicit: to study sheaves on a ringed site, one can work entirely inside O-Mod and compute R^1T_J directly rather than constructing injective resolutions in the sheaf category.","The classification by sequences d suggests a testable invariant: two topologies on a type-N or type-Z category might give equivalent sheaf categories exactly when their d-functions agree on a cofinite tail; the paper does not state this, but it is consistent with the almost-atomic description.","One could use the torsion-theoretic characterization to decide whether the irreducible sheaves constructed from weights in group representations are inequivalent; if they are, the paper's closing questions give a direct route toward a bijective formulation of Alperin's weight conjecture."],"forward_implications":["Injective objects in the sheaf category are exactly the J-torsion-free injective O-modules.","Sheaf cohomology groups can be computed as R^iΓ_x ≅ Γ^p_x ∘ R^{i+1}T_J, so cohomology of sheaves is controlled by the derived functors of the torsion functor.","Sheafification has an elementary two-step description: take the torsion-free quotient, embed it into an injective hull, and pull back the torsion part of the cokernel.","For noetherian EI categories, every sheaf category is equivalent to a presheaf category over the full subcategory of J-irreducible objects, and J-torsion modules are precisely modules supported outside that subcategory.","For EI categories of type N or Z, all Grothendieck topologies are classified by stepping sequences, and every non-rigid topology is almost atomic, reducing sheaf questions to atomic-site questions.","Every finitely generated module over an infinite full subcategory of FI or VI_q is noetherian, and over characteristic-zero fields the projective-injective, finite-injective-dimension, and Serre-quotient properties hold."],"supporting_citations":[{"why":"Supplies the right-perpendicular and Serre-quotient machinery used to prove the sheaf characterization and Corollary 1.2.","marker":"[15]"},{"why":"Provides the standard definitions of Grothendieck topologies, sheaves, sheafification, and the orbit-category bridge to group representations.","marker":"[23]"},{"why":"Supplies the Comparison Lemma and the rigid/subcategory-topology framework that reduces sheaf categories to presheaf categories over full subcategories.","marker":"[17]"},{"why":"Gives the classical observation that finite Karoubi categories have only rigid topologies, which the paper generalizes to noetherian EI categories.","marker":"[2]"},{"why":"The authors' earlier atomic-site torsion theory whose proofs and sheafification strategy are extended to all Grothendieck topologies.","marker":"[8]"},{"why":"Provides the analogous classification of Grothendieck topologies on posets that the type-N and type-Z classifications extend.","marker":"[22]"},{"why":"Establishes local self-injectivity for FI and VI representations, which Theorem 1.8 transfers to infinite full subcategories.","marker":"[12]"},{"why":"Proves noetherianity of FI-modules, used in Proposition 7.4 to show every finitely generated D-module is noetherian.","marker":"[7]"},{"why":"Provides VI-module injectivity and saturation results used for the sheaf-theoretic arguments in the applications section.","marker":"[25]"}],"fun_headline_variants":["Sheaf iff torsion-saturated: a new test without matching families","Torsion-saturated presheaves are exactly the sheaves","Sheaf categories are Serre quotients of presheaf modules","Grothendieck topologies on EI categories fully classified","Saturation condition simplifies sheaf theory on ringed sites"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The necessity direction of the rigidity classification assumes every object has finitely many endomorphisms, because the proof needs the fact that a finite monoid with exactly two Grothendieck topologies is a group; without that finiteness the characterization can fail, though the sheaf characterization itself does not depend on it.","fun_headline_variants_meta":{"raw":{"variants":["Sheaf iff torsion-saturated: a new test without matching families","Torsion-saturated presheaves are exactly the sheaves","Sheaf categories are Serre quotients of presheaf modules","Grothendieck topologies on EI categories fully classified","Saturation condition simplifies sheaf theory on ringed sites"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000315,"raw_usage":{"total_tokens":1807,"prompt_tokens":989,"completion_tokens":818,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":605,"completion_tokens_details":{"reasoning_tokens":731}},"tokens_in":605,"tokens_out":818,"duration_ms":8281,"temperature":1.0,"reasoning_tokens":731,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T05:05:32.411497+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"On a one-object directed category whose endomorphism monoid is the monoid of all surjections of an infinite set, check directly whether the dense topology and the maximal topology are the only two Grothendieck topologies; if a third topology exists, the finite-monoid dichotomy in Lemma 5.9 breaks down and the necessity direction of the rigidity theorem would fail without the finiteness condition.","supporting_citations":[{"cited_title":"Geigle and H","cited_arxiv_id":null,"evidence_quote":"Supplies the right-perpendicular and Serre-quotient machinery used to prove the sheaf characterization and Corollary 1.2."},{"cited_title":"Mac Lane and I","cited_arxiv_id":null,"evidence_quote":"Provides the standard definitions of Grothendieck topologies, sheaves, sheafification, and the orbit-category bridge to group representations."},{"cited_title":"Johnstone","cited_arxiv_id":null,"evidence_quote":"Supplies the Comparison Lemma and the rigid/subcategory-topology framework that reduces sheaf categories to presheaf categories over full subcategories."},{"cited_title":"Artin, A","cited_arxiv_id":null,"evidence_quote":"Gives the classical observation that finite Karoubi categories have only rigid topologies, which the paper generalizes to noetherian EI categories."},{"cited_title":"Sheaves of modules on atomic sites and discrete representations of topological groups","cited_arxiv_id":"2108.13600","evidence_quote":"The authors' earlier atomic-site torsion theory whose proofs and sheafification strategy are extended to all Grothendieck topologies."},{"cited_title":"Grothendieck topologies on a poset","cited_arxiv_id":"1405.4408","evidence_quote":"Provides the analogous classification of Grothendieck topologies on posets that the type-N and type-Z classifications extend."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes local self-injectivity for FI and VI representations, which Theorem 1.8 transfers to infinite full subcategories."}],"review_version":1}