{"id":"608c75b2-6fc8-41fa-9b15-23ac1378318c","arxiv_id":"2506.13346","paper_version":1,"verdict":"CONDITIONAL","confidence":"LOW","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For Noetherian F-finite semi-separated F_q-schemes, W_n-Cartier crystals are dual to constructible perverse sheaves with W_n(F_q) coefficients.","lead":"This paper proves a duality between W_n-Cartier crystals and constructible perverse etale sheaves with Witt vector coefficients on F_q-schemes, generalizing the n=1 Cartier crystal duality. It gives a Witt-vector Riemann-Hilbert style bridge between coherent sheaf theory and perverse sheaf theory, with potential applications to vanishing theorems in positive characteristic.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 5.2.7 rests entirely on the imported Riemann–Hilbert equivalence (Theorem 3.2.3, from [BL19, Thm 9.6.1]) and on the n=1 base case [Bau23, Thm 5.2.7]; the paper does not re-prove either, so a mismatch of hypotheses would invalidate the central duality.","rationale":"The reader's weakest_assumption identifies exactly the point that would invalidate Theorem 5.2.7 if it failed: the imported Riemann–Hilbert correspondence (Theorem 3.2.3 from [BL19, Thm 9.6.1]) and the n=1 base [Bau23, Thm 5.2.7]. My reading of §5.2 confirms that the proof is a reduction to these imports; no independent construction or verification is supplied. This is not a criticism of the paper's internal logic, which appears coherent, but a statement about the load-bearing status of an external result. In good faith, the paper's own Theorem 5.2.7 is clearly stated over W_n(F_q), and the proof strategy is plausible. The main additional observation is that the title and abstract overstate the coefficient ring, claiming Z/p^nZ-perverse sheaves when the theorem is about W_n(F_q)-sheaves; for r>1 these are different categories. This does not change the mathematical theorem as stated, but it reinforces the need for a conditional verdict requiring correction of the abstract and introduction. Since the reader already gave CONDITIONAL, my analysis does not move the verdict; it sharpens the reason for the condition.","tokens_in":26741,"tokens_out":26633,"duration_ms":234360,"concrete_test":"Check the exact statement and hypotheses of [BL19, Theorem 9.6.1] (and the gluing used from Theorem 10.2.7): determine whether it is proved for all n≥1 and for F_q-coefficients on arbitrary Noetherian F-finite F_q-schemes, or only for n=1 and F_p-coefficients. If it is narrower, test the affine reduction in the proof of Theorem 3.2.3: (1) verify that Lemma 3.2.6 identifies the categories exactly as needed; (2) verify that the étale gluing argument of [BL19, Thm 10.2.7] applies unchanged to W_n-Frobenius modules; (3) check whether the passage from compact objects detects constructibility. Also independently verify [Bau23, Thm 5.2.7] for semi-separated F-finite F_q-schemes. A minimal concrete case: take X = Spec(k) with k F-finite, n=2 and q=p^2, and manually compute Sol∘D to see whether it hits every constructible W_2(F_q)-sheaf and nothing else.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 5.2.7 (§5.2) is an induction that reduces the t-structure statement to two imports: the n=1 case [Bau23, Thm 5.2.7] and the equivalence Sol: Crys^Fr_{W_nX} ≅ Sh_c(X_et, W_n(F_q)) stated as Theorem 3.2.3, attributed to [BL19, Thm 9.6.1]. Nothing in the paper verifies that the hypotheses of [BL19, Thm 9.6.1] match the stated generality: Noetherian, F-finite, semi-separated F_q-schemes, for arbitrary n≥1 and q=p^r. The proof of Theorem 3.2.3 says only to reduce to the affine case 'as in [BL19, Thm 10.2.7]', apply Lemma 3.2.6 and [BL19, Thm 9.6.1], and then pass to compact objects. If [BL19, Thm 9.6.1] is actually stated for n=1 only, or for finite-type schemes only, then the equivalence used as the topological side of the duality is not established by the text, and Sol∘D in Theorem 5.2.7 may fail to be essentially surjective or may not be defined at the claimed level of generality. The induction base [Bau23, Thm 5.2.7] is likewise taken from an unpublished preprint. This is load-bearing because no step in §5.2 independently constructs or verifies the Riemann–Hilbert correspondence for W_n coefficients. Separately, the abstract and title advertise 'Z/p^nZ-perverse sheaves', while Theorem 5.2.7 is over W_n(F_q); these coefficient rings differ for r>1, so the advertised statement is also broader than the proved one.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"Fix a prime p, integers n, r ≥ 1, and q = p^r. The paper develops a theory of W_n-Cartier modules and crystals on Noetherian F-finite F_p-schemes, proves several finiteness and duality results for these categories, and states as its main theorem (Theorem 5.2.7) that for a Noetherian, F-finite, semi-separated F_q-scheme X equipped with a W_n-unit dualizing complex, the functor Sol∘D induces an equivalence D^b(Crys^Cr_{W_nX})^op ≅ D^b_c(X_ét, W_n(F_q)) that sends the canonical t-structure to the perverse t-structure, hence (Crys^Cr_{W_nX})^op ≅ Perv_c(X_ét, W_n(F_q)). The proof is an induction on n: Theorem 5.1.12 establishes the duality D between W_n-Cartier and W_n-Frobenius crystals, and the Riemann-Hilbert equivalence Sol between W_n-Frobenius crystals and constructible W_n(F_q)-sheaves is imported as Theorem 3.2.3 from [BL19, Theorem 9.6.1]; the base case n = 1 is the author's earlier theorem [Bau23, Theorem 5.2.7].","tokens_in":27190,"tokens_out":13468,"duration_ms":115360,"significance":"If the hypotheses on the imported results are verified, this is a substantial extension of the Cartier-crystal/perverse-sheaf duality of [Bau23] to truncated Witt-vector coefficients. The paper's own contributions—the finiteness theory of W_n-Cartier modules (Propositions 4.4.4 and 4.4.8), the duality Theorem 5.1.12, and the devissage induction in Lemmas 5.2.5–5.2.6—are well structured and appear correct. The main risk is that Theorem 5.2.7 depends entirely on the imported Riemann-Hilbert correspondence and on the unpublished base case [Bau23, Theorem 5.2.7], and the manuscript does not provide enough detail to confirm that the cited results apply in the stated generality. The abstract and introduction further state the result with Z/p^nZ-sheaves and F_p-schemes, while the theorem is over W_n(F_q)-sheaves and F_q-schemes; this ambiguity must be resolved.","major_comments":[{"comment":"The proof of Theorem 3.2.3 reduces the statement to the affine case by citing [BL19, Theorem 10.2.7], then invokes Lemma 3.2.6 and [BL19, Theorem 9.6.1], but it does not state the precise hypotheses of [BL19, Theorem 9.6.1] nor verify that they hold for every Noetherian F-finite F_q-scheme, for all n ≥ 1 and r ≥ 1. Since Theorem 5.2.7 uses this equivalence as the entire bridge from W_n-Frobenius crystals to constructible W_n(F_q)-sheaves, an unstated hypothesis in the imported theorem would invalidate the main claim. Please provide the exact statement of the cited theorem and a complete verification of the hypotheses, or restrict the main theorem to the generality for which the Riemann-Hilbert correspondence is established.","section":"Section 3.2, Theorem 3.2.3"},{"comment":"The abstract and the theorem displayed in the Introduction claim an equivalence with perverse Z/p^nZ-sheaves (and, in the Introduction, for X an F_p-scheme), whereas Theorem 5.2.7 is stated for F_q-schemes and W_n(F_q)-sheaves. Since W_n(F_q) is different from Z/p^nZ when r > 1, and an arbitrary F_p-scheme need not be an F_q-scheme for r > 1, the precise relationship between these statements is unclear. The paper should consistently state the main theorem in the full W_n(F_q)-form and explain that the Z/p^nZ version is the special case r = 1, or prove the Z/p^nZ version as advertised.","section":"Abstract and Introduction, §1.1"},{"comment":"The induction in Theorem 5.2.7 has as its base case [Bau23, Theorem 5.2.7], which is an unpublished preprint. The proof of the n-level statement reduces entirely to this base and to Theorem 3.2.3, so the paper should either include the precise statement of [Bau23, Theorem 5.2.7] with its hypotheses, or reproduce enough of its proof to make the induction self-contained. As written, the central equivalence is not fully verifiable from the manuscript alone.","section":"Section 5.2, proof of Theorem 5.2.7"}],"minor_comments":[{"comment":"The sentence 'The link between W_n–Frobenius crystals and étale Z/p^nZ–sheaves was worked out in [BL19, Section 9]' should read 'W_n(F_q)-sheaves' to be consistent with Theorem 3.2.3.","section":"Section 1.1, first paragraph"},{"comment":"There are small typographical errors: 'essential image consist of objects' should be 'essential image consists of objects', and in the proof 'CrysF r WnX' appears where 'CrysF r WmX' or 'CrysF r WnX' is intended depending on context; please proofread this paragraph.","section":"Lemma 3.1.6 and its proof"},{"comment":"The displayed exact sequences use periods in place of slashes or arrows (for example 'M[p]∩p^iM. M[p]∩p^{i+1}M'); these should be typeset properly as quotients in the exact sequences.","section":"Lemma 4.2.2"},{"comment":"The word 'asusme' should be 'assume'.","section":"Proof of Proposition 4.4.16"},{"comment":"The second author's name appears as 'T. Wei β'; if this is a rendering issue with a German character, it should be corrected (likely 'Weiß' or 'Weiss').","section":"References, [MW24]"}],"recommendation":"major_revision","confidential_remarks":"The main concern for the editor is the dependence of the central theorem on two results external to the manuscript: the Riemann-Hilbert correspondence of Bhatt-Lurie, whose hypotheses are not checked in the text, and the author's own unpublished [Bau23] for the base case. The abstract/theorem mismatch (Z/p^nZ versus W_n(F_q)) should be resolved in revision; note that the advertised Z/p^nZ statement is best read as the r = 1 specialization of the W_n(F_q) theorem, and the wording should make that explicit. I do not see circularity: the n-level statement goes beyond its inputs, and the n = 1 base is a different, previously established theorem."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is what it says on the tin: a generalization of Baudin's n=1 Cartier-perverse duality to W_n-Cartier crystals and W_n(F_q)-perverse sheaves. That is real new content, not a renaming. The finiteness theorems in Section 4 (pushforwards stay ind-coherent, crystals are Noetherian and Artinian, exactness of the solution functor on ind-coherent Cartier modules) are the strongest part of the paper and are genuinely new at the W_n level. The duality Theorem 5.1.12 also appears to be proven by a sound reduction to the n=1 case, using the equivalence of derived categories established earlier. I read carefully through the reduction arguments and found no obvious circularity or algebraic error.\n\nTwo soft spots, one mostly cosmetic and one load-bearing. The cosmetic one: the abstract and introduction say the topological side is perverse Z/p^nZ-sheaves, but Theorem 5.2.7 is over W_n(F_q). For r>1 these are different coefficient rings. That should be fixed in a revision; it matters for potential applications and for anyone citing the advertised statement.\n\nThe load-bearing spot is the Riemann-Hilbert import. Theorem 3.2.3 restates [BL19, Thm 9.6.1] as an equivalence for all Noetherian F-finite F_q-schemes and all n, with constructible sheaves on the étale site. The proof given here only reduces to the affine case and invokes [BL19]. Nothing in the paper verifies that the hypotheses of [BL19, Thm 9.6.1] match this generality. If that theorem is actually proved only for finite-type schemes or only for n=1, then the equivalence Sol in Theorem 5.2.7 is not established by the text. The same holds for the n=1 base case, which is taken from an unpublished preprint [Bau23]. This is not a fatal flaw on its face, but it is exactly what a referee should check before the paper is accepted. The author should either quote the precise theorem with its hypotheses or prove the needed generality.\n\nWho is this for? Anyone working on the positive-characteristic Riemann-Hilbert program, especially on Witt-vector coefficients, and anyone who wants to use W_n-canonical sheaves for vanishing theorems. It deserves a serious referee: the main construction is plausible, the n-level extension is significant, and the gaps I found are fixable rather than fundamental. I would send it to review, with a specific request to verify the BL19 hypotheses and to correct the coefficient ring statement in the abstract.","headline":"A genuine Witt-vector extension of the author's n=1 Cartier-perverse duality, with new finiteness results, but the advertised Z/p^nZ coefficients are broader than the W_n(F_q) theorem actually proved, and the central proof leans on imports from BL19 and Bau23 that a referee should check.","tokens_in":27702,"tokens_out":2497,"would_cite":true,"duration_ms":26369,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14G17","13A35","13F35"],"pacs":[],"model":"deepseek-v4-flash","headline":"On Noetherian F-finite schemes, W_n-Cartier crystals and constructible perverse Z/p^nZ-sheaves are equivalent up to opposite categories.","keywords":["perverse sheaves","Cartier crystals","Witt vectors","Riemann-Hilbert correspondence","Frobenius modules","constructible sheaves","t-structures","dualizing complexes"],"falsifier":"Take an $F$-finite imperfect field $k$, set $X = \\mathbb{A}^1_k$ and $n=2$, and explicitly compute $\\mathrm{Sol}(D(M))$ for the $W_2$-Frobenius modules corresponding to the constant, skyscraper, and rank-one sheaves; the theorem predicts their perverse cohomology is concentrated in the predicted degree with the predicted $W_2(\\mathbb{F}_q)$-ranks. A mismatch in any one of these explicit computations would disprove the $t$-structure part of the main theorem.","tokens_in":26547,"feed_emoji":"🔁","tokens_out":10879,"duration_ms":89293,"temperature":0.7,"pith_summary":"This paper proves that a coherent algebraic object—a $W_n$-Cartier crystal, roughly a coherent module over the $n$-truncated Witt vectors equipped with a Frobenius-linear Cartier operator, considered up to nilpotence—is the same data as a constructible perverse sheaf of $\\mathbb{Z}/p^n\\mathbb{Z}$-modules on the \\'etale site. The duality is an equivalence of derived categories, compatible with proper pushforwards, and it matches the canonical $t$-structure on the crystal side with the middle perverse $t$-structure on the sheaf side. The result extends the known $n=1$ duality to every truncation level $n$, which matters because it lets coherent techniques prove statements about perverse sheaves with $W_n(\\mathbb{F}_q)$ coefficients and, conversely, topological constructibility results about Witt-vector canonical sheaves. The proof goes through a new duality between $W_n$-Frobenius modules and $W_n$-Cartier modules, built from dualizing-complex pairings against a $W_n$-unit dualizing complex.","feed_headline":"Witt-vector crystals match perverse Z/p^nZ-sheaves","feed_subtitle":"A duality theorem pairs Frobenius and Cartier data through a dualizing complex and shows the two hearts agree.","key_machinery":"The load-bearing mechanism is a pair of Hom pairings between $W_n$-Frobenius modules and $W_n$-Cartier modules, both given by $\\mathrm{Hom}_{W_n\\mathcal{O}_X}(-,-)$ with the Frobenius action on one side and the Cartier operator on the other. Pairing against a $W_n$-unit dualizing complex $W_n\\omega_X^\\bullet$ yields two duality functors $D$, shown to be mutual inverses; through derived equivalences between coherent and quasi-coherent variants, this gives the Frobenius–Cartier duality. On the topological side, the functor $\\mathrm{Sol}$ extracts the kernel of $\\kappa-1$ and, by the imported Riemann–Hilbert correspondence, identifies $W_n$-Frobenius crystals with constructible $W_n(\\mathbb{F}_q)$-sheaves. Composing $\\mathrm{Sol}\\circ D$ is what converts the coherent $t$-structure into the middle perverse $t$-structure; the inductive lemmas decompose a complex into its $p$-torsion and $W_{n-1}$ pieces to reduce to $n=1$.","core_discovery":"The central theorem (Theorem 5.2.7) states: let $X$ be a Noetherian, $F$-finite, semi-separated $\\mathbb{F}_q$-scheme admitting a $W_n$-unit dualizing complex $W_n\\omega_X^\\bullet$. After shifting this dualizing complex on each connected component, the functor $\\mathrm{Sol}\\circ D$ sends the canonical $t$-structure on $D^b(\\mathrm{Crys}^C_{W_nX})^{\\mathrm{op}}$ to the perverse $t$-structure on $D^b_c(X_{\\acute{e}t},W_n(\\mathbb{F}_q))$, and therefore induces an equivalence $(\\mathrm{Crys}^C_{W_nX})^{\\mathrm{op}} \\cong \\mathrm{Perv}_c(X_{\\acute{e}t},W_n(\\mathbb{F}_q))$. Here $D$ is the new duality between $W_n$-Frobenius and $W_n$-Cartier modules, and $\\mathrm{Sol}$ is the solution functor taking a Cartier module $(M,\\kappa_M)$ to $\\ker(\\kappa_M-1)$ as an \\'etale sheaf. The proof reduces to the $n=1$ case by an induction that decomposes any $W_n$ object into its mod-$p$ and $W_{n-1}$ parts, using the imported Riemann–Hilbert correspondence for $W_n$-Frobenius modules as the bridge to the topological side.","pith_inferences":["If the duality holds, any perverse-sheaf theorem over $\\mathbb{Z}/p^n\\mathbb{Z}$ should have a purely coherent counterpart; an immediate test case is to translate known generic vanishing statements across the equivalence for all $n$.","The induction on $n$ suggests a general d\\'evissage principle: many properties of perverse $W_n(\\mathbb{F}_q)$-sheaves can be proven by reducing first to $\\mathbb{F}_p$-sheaves and then to $W_{n-1}$-sheaves, possibly simplifying future constructibility arguments.","The duality may extend to derived $\\infty$-categories of Cartier modules, where the $t$-structure comparison could be obtained formally rather than by induction; the paper notes this as a plausible but unimplemented route.","A consequence the author leaves implicit: the equivalence should induce a bijection between support conditions on perverse $W_n(\\mathbb{F}_q)$-sheaves and support-theoretic strata of $W_n$-Cartier crystals, so classification results on one side can be read off from the other."],"forward_implications":["Every theorem about constructible perverse $\\mathbb{Z}/p^n\\mathbb{Z}$-sheaves translates into a theorem about $W_n$-Cartier crystals, so topological proofs can settle coherent questions about $W_n$-canonical sheaves.","Vanishing theorems and finiteness results for the Witt canonical sheaf $W_n\\omega_X$ can be approached through perverse sheaves with $W_n(\\mathbb{F}_q)$ coefficients, extending the $n=1$ applications.","Objects of $\\mathrm{Crys}^C_{W_nX}$ are both Noetherian and Artinian, and direct images of ind-coherent $W_n$-Cartier modules satisfy the corresponding finiteness and stability properties.","The equivalence is compatible with proper pushforwards and with the inclusions from level $m$ to level $n$, so perverse cohomology of direct images matches Cartier-crystal cohomology in a way that is stable under truncation."],"supporting_citations":[{"why":"supplies the Riemann-Hilbert correspondence identifying W_n-Frobenius crystals with constructible W_n(F_q)-sheaves, the bridge to the topological side","marker":"[BL19, Theorem 9.6.1]"},{"why":"is the n=1 base case of the duality between Cartier crystals and perverse F_p-sheaves that the induction reduces to","marker":"[Bau23, Theorem 5.2.7]"},{"why":"defines the middle perverse t-structure on the etale derived category used for the target category","marker":"[Gab04]"},{"why":"provides the Frobenius-module and Cartier-module techniques, including localisation and derived-category equivalences adapted here to Witt vectors","marker":"[BP09]"},{"why":"supplies the finiteness theorems for Cartier modules that the paper generalizes to W_n-Cartier crystals","marker":"[BB11]"},{"why":"supplies the Witt-vector scheme-theoretic facts, including finiteness of Frobenius and Cartesian diagrams used throughout","marker":"[LZ04]"},{"why":"gives existence of a W_n-unit dualizing complex on finite-type schemes over an F-finite algebra, used to justify the hypotheses of the main theorem","marker":"[KTT+24, Theorem 9.1]"}],"fun_headline_variants":["Cartier crystals dual to perverse p^n sheaves","Duality: Cartier crystals ↔ perverse p^n sheaves","Frobenius-Cartier duality yields perverse equivalence","Perverse sheaves from Cartier crystals via duality","W_n-Cartier crystals match perverse p^n sheaves"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof inherits the imported Riemann–Hilbert correspondence for $W_n$-Frobenius crystals and the $n=1$ duality result; if either of those has hidden hypotheses outside the stated framework, the induction proving the full duality collapses.","fun_headline_variants_meta":{"raw":{"variants":["Cartier crystals dual to perverse p^n sheaves","Duality: Cartier crystals ↔ perverse p^n sheaves","Frobenius-Cartier duality yields perverse equivalence","Perverse sheaves from Cartier crystals via duality","W_n-Cartier crystals match perverse p^n sheaves"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000205,"raw_usage":{"total_tokens":1360,"prompt_tokens":880,"completion_tokens":480,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":496,"completion_tokens_details":{"reasoning_tokens":394}},"tokens_in":496,"tokens_out":480,"duration_ms":4095,"temperature":1.0,"reasoning_tokens":394,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T20:05:25.279080+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take an $F$-finite imperfect field $k$, set $X = \\mathbb{A}^1_k$ and $n=2$, and explicitly compute $\\mathrm{Sol}(D(M))$ for the $W_2$-Frobenius modules corresponding to the constant, skyscraper, and rank-one sheaves; the theorem predicts their perverse cohomology is concentrated in the predicted degree with the predicted $W_2(\\mathbb{F}_q)$-ranks. A mismatch in any one of these explicit computations would disprove the $t$-structure part of the main theorem.","supporting_citations":[],"review_version":1}