{"id":"e7ce36fe-7361-48a2-9096-8358ad52844b","arxiv_id":"2607.07042","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"The algebraic stratification of Hitchin fibers via local smoothening types is proven equivalent to the geometric stratification by partial normalizations for GL_n in the Bass case, yielding an explicit cohomology formula for compactified Jacobians of spectral curves with double singularities.","lead":"This paper connects a local matrix-based method for computing orbital integrals with the global geometric framework of Hitchin fibrations, proving the two approaches yield the same decomposition. It derives an explicit closed formula for the cohomology of compactified Jacobians of certain singular curves over finite fields.","discovery_kind":"unclear","skeptic_critique":{"model":"glm-5.2","headline":"No significant objection identified","rationale":"The reader correctly identifies the dependency on local smoothness results from [CKL] and [CHL] as the foundational assumption. Having examined the paper's internal logic, I find the argument chain is internally consistent: the global smoothness (Theorem 3.24) properly reduces to local formal smoothness via Proposition 3.22 and Lemma 3.23, the stratification equivalence (Theorem 4.16) is supported by a concrete matrix-theoretic argument connecting types to overorders via nilpotency indices, and the cohomology formula (Theorem 5.3) follows from the affine paving established in Lemma 5.2. The key observation is that the nilpotency index computation is correct: for J_{l_x+1, n-l_x-1} with l_x ≤ (n-3)/2, the index is n-l_x-1, matching the overorder parameter r_x from Proposition 4.15. The restriction to Bass case (double singularities, integral domain local rings) and P^1 normalization is clearly stated. The factorization of the Poincaré polynomial via Künneth (Remark 5.5) provides additional structural confirmation. The confidence should remain MODERATE because full verification requires checking [CHL, Proposition 4.7] and the matrix computations in Section 3.5, but no internal inconsistency or logical gap is apparent in the present paper's arguments. The reader's verdict of ACCEPT is appropriate.","tokens_in":35249,"tokens_out":1203,"duration_ms":543651,"concrete_test":"Independently verify the matrix computation underlying Theorem 4.16 for a concrete case: take n=5, d_x=1, ord_x(a_5)=0 (so n·d_x+ord_x(a_n)=5>2, which violates the Bass condition—instead take n=5, d_x=0, ord_x(a_5)=2 so n·d_x+ord_x(a_n)=2). Choose l_x=1 (so 0≤1≤3/2). Verify that the nilpotency index of J_{2,3} is 3 = n-l_x-1 = 5-1-1, and that the overorder R_{x,r_x} with r_x=3 matches the overorder determined by the type T with l_x=1. Then check that the dimension formula from Lemma 5.2 gives dim = (n-3)/2 - l_x = 1 - 1 = 0 for this stratum, consistent with the point count q^{d(r_x - (n+1)/2)} = q^{d(3-3)} = 1.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim (Theorem 5.3 / Corollary 5.4) depends on a chain: (1) local smoothness from [CKL]/[CHL] (Theorems 2.2, 2.4), (2) global smoothness of the restricted Hitchin fibration Φ_T (Theorem 3.24), (3) equivalence of algebraic and geometric stratifications (Theorem 4.16), and (4) the affine paving yielding the cohomology decomposition. The reader correctly identifies (1) as the foundational dependency. Having examined the argument, the most load-bearing step is actually (3)—Theorem 4.16—because it is the bridge converting the algebraic stratification (built from matrix-theoretic types) into the geometric stratification (partial normalizations of Y_χ). If this equivalence fails, the strata (Φ_T)^{-1}(χ) need not be affine spaces, and the cohomology formula collapses. The proof of Theorem 4.16 proceeds by showing that the type T determines the tuple of overorders (g_x O_x^n : g_x O_x^n)_{x∈C} via Corollary 4.14 and Proposition 4.15, and that this tuple in turn determines the partial normalization. The key step is the case-by-case analysis: for x ∈ S with n·d_x + ord_x(a_n) = 2 and n≥3, the nilpotency index of g_x^{-1}γ_x g_x mod π_x determines the overorder R_{x,r_x}, and this index is claimed to equal n - l_x - 1 when x ∈ S_F^1 (from Remark 3.12) and n when x ∈ S'∖S_F^1. This is a concrete, checkable matrix computation. The argument appears sound: the Jordan form J_{l_x+1, n-l_x-1} has nilpotency index max(l_x+1, n-l_x-1) = n-l_x-1 (since l_x ≤ n/2-1 implies n-l_x-1 ≥ l_x+1), and rank n-2 matrices have nilpotency index n. The connection to overorders via Proposition 4.15 (where r_x is the minimal integer with γ_x^{r_x}/π_x ∈ R') matches: r_x = n-l_x-1 for x ∈ S_F^1 and r_x = n for x ∈ S'∖S_F^1. The global smoothness (Theorem 3.24) relies on Proposition 3.22 (existence of global lift), which in turn uses Lemma 3.23 (lifting centralizers) and the local formal smoothness from Corollaries 2.3/2.5. The argument in Lemma 3.23 reduces to a linear equation βθ_0 - θ_0β","agreement_with_reader":"agree"},"referee_report":{"model":"glm-5.2","summary":"This paper geometrizes a local smoothening method (developed in the authors' prior works [CKL] and [CHL] for computing orbital integrals of GL_n in the Bass case) within the global framework of the Hitchin fibration. The main results are: (1) an algebraic stratification of the Hitchin fiber Φ^{-1}(χ) into smooth locally closed strata indexed by 'types' T (Theorem 3.24, Corollary 4.5); (2) a proof that this algebraic stratification coincides with the geometric stratification of the compactified Jacobian Pic^0(Y_χ) by partial normalizations (Theorem 4.16); and (3) under the assumption that the normalization of Y_χ is P^1_k, a closed formula for the ℓ-adic cohomology H^i_c(Pic^0(Y_χ)_k̄, Q_ℓ) (Theorem 5.3, Corollary 5.4). The cohomology is shown to vanish in odd degrees, with even-degree pieces being direct sums of Q_ℓ(-i) whose multiplicities are encoded by an explicit Poincaré polynomial factoring as a product over singularities.","tokens_in":35949,"tokens_out":1322,"duration_ms":730620,"significance":"The paper bridges local orbital integral computations and global Hitchin fibration geometry, which is a natural and potentially influential direction. The explicit cohomology formula (Corollary 5.4) is a concrete, falsifiable prediction that provides new geometric information about compactified Jacobians of spectral curves with double singularities. The factorization of the Poincaré polynomial via a Künneth-type formula (Remark 5.5) is a nice structural observation. The equivalence of algebraic and geometric stratifications (Theorem 4.16) is the conceptual heart of the paper and is a significant result connecting matrix-theoretic types to partial normalizations.","major_comments":[{"comment":"In case (3) of the proof, for x ∈ S∖S_F^1 with n·d_x + ord_x(a_n) = 2 and n≥3, the claim is that the nilpotency index of g_x^{-1}γ_x g_x mod π_x equals n. The justification is that 'g_x^{-1}γ_x g_x has rank n−1' and thus the nilpotency index is n. However, a rank n−1 nilpotent matrix does not necessarily have nilpotency index n; for example, a matrix similar to J_{1,n-1} has rank n-1 and nilpotency index n-1, not n. The claim that the nilpotency index is exactly n (rather than at most n) needs a more precise argument. This is load-bearing because the nilpotency index determines the overorder R_{x,r_x} via Proposition 4.15, which in turn determines the partial normalization and hence the stratum. If the index could be smaller, the bijection between types and overorders would break. The authors should clarify why the nilpotency index is exactly n in this case, not merely at most n.","section":"Theorem 4.16 (proof, p.26-27)"}],"minor_comments":[{"comment":"The title in the manuscript header contains spacing artifacts: 'GEOMETRIZA TION' and 'FIBRA TION'. These should be corrected to 'GEOMETRIZATION' and 'FIBRATION'.","section":null},{"comment":"In the proof of Theorem 4.16, case (3): the statement 'g_x^{-1}γ_x g_x has rank n−1' should specify that this is the reduction modulo π_x. The nilpotency index is computed on the reduction, and making this explicit would avoid confusion.","section":null},{"comment":"In Diagram (3.4) (p.17), the caption states 'the middle is not' Cartesian, but the visual layout could be clearer. A brief textual clarification of which square is non-Cartesian would aid the reader.","section":null},{"comment":"The notation S_F^1 (Definition 3.14) is somewhat opaque. A brief remark explaining the superscript '1' (presumably related to the Jordan block structure) would improve readability.","section":null},{"comment":"In Lemma 5.2, the dimension formula for n≥3 involves the sum over x∈S_F^1 of (n-3)/2 - l_x plus a term from S'∖S_F^1. It would help to explicitly state that l_x := -1 for x ∈ S'∖S_F^1 (as done in Theorem 5.3) so the formula reads as a single sum over S'.","section":null},{"comment":"The reference [Che] in the text (Remark 5.5) appears to be an arXiv preprint (arXiv:2411.08403). A complete bibliographic entry should be provided in the references list.","section":null},{"comment":"In Setting 3.4.(2), condition (a) states that χ_x(X) is separable over k for x∉S. Since S is defined as the set where χ_x(X) is irreducible (Definition 3.2), it would be clearer to state that for x∉S, the reduction χ̄_x(X) is separable, emphasizing the distinction between χ_x and its reduction.","section":null},{"comment":"On p.29, the argument that O_{Y'_χ} = Hom_{O_{Y_χ,k'}}(L,L) yields a disjoint union relies on [Vas68, Theorem 3.1]. A brief sentence explaining how this theorem applies (L being rank-1 torsion-free, Y'_χ being Gorenstein) would strengthen the citation.","section":null}],"recommendation":"minor_revision","confidential_remarks":"The reader's assessment of the paper as fundamentally sound is accurate. The main concern I raise (nilpotency index in Theorem 4.16, case 3) is likely a gap in exposition rather than a genuine error—the matrix structure probably forces the maximal nilpotency index, but the authors need to state why. The dependency on [CKL]/[CHL] for local smoothness is acceptable as these are stated theorems with proofs. The paper fits the journal's scope well."},"author_rebuttal":null,"desk_editor":{"model":"glm-5.2","letter":"The main thing to know: this paper takes the authors' local smoothening machinery (from [CKL] and [CHL]) and globalizes it inside the Hitchin fibration for GL_n in the Bass case. The payoff is a closed formula for the ℓ-adic cohomology of compactified Jacobians of spectral curves with double singularities, assuming the normalization is P^1. The formula is clean: odd cohomology vanishes, even cohomology is a direct sum of Q_ℓ(-i) with multiplicities given by a factored Poincaré polynomial (P_m(X^2))^t for n≥3 odd, or a product of Q_j(X^2) factors for n=2, where t is the number of singularities. The factorization reflects a Künneth-type product structure on affine Springer fibers, which is a nice structural observation (Remark 5.5). The central new result is Theorem 4.16, which shows that the algebraic stratification (built from matrix-theoretic types indexing local smoothening strata) agrees with the geometric stratification from partial normalizations of the spectral curve (Gagne 1997). This bridge is what makes the cohomology formula work: it converts smooth algebraic strata into affine spaces, yielding an affine paving. The proof of Theorem 4.16 is a case-by-case analysis connecting the nilpotency index of g_x^{-1}γ_x g_x mod π_x to overorders of the local Bass order R_x, which is concrete and checkable. The matrix computation is correct: the Jordan form J_{l_x+1, n-l_x-1} has nilpotency index n-l_x-1 (since l_x ≤ n/2-1), and this matches the overorder parameter r_x from Proposition 4.15. The global smoothness argument (Theorem 3.24) rests on Proposition 3.22 and Lemma 3.23, which lifts centralizers over Artinian rings using McDonald's Jordan decomposition. The reduction in Lemma 3.23 to a linear equation βθ_0 - θ_0β = ... is standard and sound. The soft spot is the dependency chain: everything rests on the local smoothness results from [CKL, Theorem 4.6] and [CHL, Proposition 4.7], stated here as Theorems 2.2 and 2.4. If those prior results have gaps, the global smoothness of the restricted Hitchin fibration fails and the stratification doesn't produce smooth strata. But these are theorems with stated proofs, not fitted parameters, and the global geometrization here is independent of their internal details. The restriction to the Bass case and P^1 normalization is clearly stated as a hypothesis throughout. The authors note (Remark 1.6) that the geometric stratification fails outside the Bass case but expect an algebraic stratification to survive, which is a reasonable direction for future work. This paper is for specialists in the Hitchin fibration / orbital integral program who want explicit cohomological computations. It deserves a serious referee who can check the matrix-theoretic arguments in Section 3.5 and the overorder classification in Section 4.3 against [CHL].","headline":"Solid paper: geometrizes local smoothening in the Hitchin fibration, gives a clean closed cohomology formula for compactified Jacobians with double singularities.","tokens_in":36250,"tokens_out":781,"would_cite":true,"duration_ms":106929,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"glm-5.2","headline":"Closed formula for cohomology of compactified Jacobians with double points","keywords":[],"falsifier":"A counterexample to the smoothness of the local morphisms φ_M (Theorem 2.2) or φ_{l_1} (Theorem 2.4) at some point over χ_γ would break the formal smoothness lifting in Proposition 3.22, collapsing the proof that the global strata are smooth. Alternatively, a spectral curve with double Bass singularities and normalization P^1 whose compactified Jacobian cohomology does not match the predicted Poincaré polynomial would directly falsify Theorem 5.3.","tokens_in":35365,"feed_emoji":"🧮","tokens_out":1315,"duration_ms":136542,"temperature":0.7,"pith_summary":"The paper bridges two previously separate approaches to studying orbital integrals for GL_n. On one side is a local, matrix-theoretic method called smoothening, developed in the authors' prior work, which stratifies the space of matrices sharing a characteristic polynomial into smooth pieces indexed by combinatorial data called types. On the other side is the global Hitchin fibration, where the fiber over a point encodes the same orbital integral as the compactified Jacobian of a spectral curve. The paper's central achievement is showing that the local smoothening stratification globalizes: the type-indexed strata from the local theory correspond exactly to strata of the compactified Jacobian cut out by partial normalizations of the spectral curve. This equivalence between an algebraic stratification (from matrix types) and a geometric stratification (from curve normalizations) holds when the singularities are double points whose local rings are Bass orders. When the normalization of the spectral curve is P^1, each stratum is an affine space, yielding an affine paving. The cohomology then decomposes as a direct sum of one-dimensional pieces, and the paper gives an explicit closed formula for the multiplicities as coefficients of certain polynomials raised to the number of singularities.","feed_headline":"Closed formula for cohomology of compactified Jacobians with double points","feed_subtitle":"Local smoothening of matrix spaces globalizes to the Hitchin fibration, matching partial normalizations of spectral curves and yielding an P","key_machinery":"The restricted Hitchin fibration Φ_T: M_T → A_T (Diagram 3.4), which cuts out a smooth locally closed subvariety of the Hitchin fiber for each global type T. The proof of equivalence between stratifications runs through the classification of overorders of Bass orders (Proposition 4.15) and the product formula relating the Hitchin fiber to a product of affine Springer fibers (Proposition 4.12). The affine paving and dimension formula (Lemma 5.2) come from identifying each stratum with a product of quotients of unit groups (O_{E_x})^× / (R_{x,r_x})^×, which are split tori of computable rank.","core_discovery":"The algebraic stratification of the Hitchin fiber induced by the local smoothening process (Theorem 1.3/Corollary 4.5) agrees scheme-theoretically with the geometric stratification of the compactified Jacobian by partial normalizations (Theorem 4.16). In the Bass case with normalization P^1, this yields an affine paving whose dimensions are explicitly computable, producing the closed cohomology formula of Theorem 5.3/Corollary 5.4: odd-degree cohomology vanishes, and even-degree cohomology is a direct sum of Q_ℓ(-i) with multiplicities given by the coefficient of X^i in (P_m(X))^t for n odd ≥ 3, or in a product of polynomials Q_j(X) for n = 2, where t is the number of singularities.","pith_inferences":["If the local smoothness results from [CKL] and [CHL] that underpin Theorems 2.2 and 2.4 turn out to have gaps for certain n or certain residue characteristics, the global smoothness of the strata (Theorem 3.24) would fail, and the affine paving argument would break. The cohomology formula would then need modification to account for non-smooth strata.","The restriction to normalization P^1 is essential for the strata being affine spaces rather than general abelian varieties. For higher-genus normalizations, each stratum would be a torsor under a generalized Jacobian, and the cohomology would acquire additional non-trivial contributions from the Jacobian of the normalization, complicating the closed formula.","The Bass condition (double points) is the threshold where both the overorder classification and the geometric stratification by partial normalizations remain tractable. Beyond Bass — for instance, triple points — the geometric stratification by Gagne fails, and one would need a different geometric description of the strata even if the algebraic one survives."],"forward_implications":["The closed cohomology formula gives explicit Betti numbers for compactified Jacobians of curves with double points over finite fields, which can be compared against point-counting data from the trace formula.","The equivalence of algebraic and geometric stratifications provides a new tool for studying Hitchin fibers in cases where the spectral curve has worse-than-nodal singularities, if the local smoothening process is extended.","The factorization of the Poincaré polynomial as a product over singularities reflects a Künneth-type decomposition of the Hitchin fiber into a product of local affine Springer fibers, making the local-to-global structure explicit.","The framework may extend to other reductive groups beyond GL_n where analogues of the smoothening construction and Bass-type singularity conditions can be formulated."],"fun_headline_variants":["Local-to-global smoothening yields closed cohomology formula for compactified Jacobians","Odd cohomology vanishes for compactified Jacobians of spectral curves with double points","Affine paving of compactified Jacobians matches local smoothening stratification","Cohomology multiplicities computed via polynomial formula in the Bass case"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The entire argument depends on local smoothness results from the authors' prior work, which assert that certain characteristic-polynomial maps on carefully chosen matrix spaces are smooth at the relevant points. If those local smoothness claims contain gaps, the global smoothness of the stratification fails and the cohomology formula does not follow.","fun_headline_variants_meta":{"raw":{"variants":["Local-to-global smoothening yields closed cohomology formula for compactified Jacobians","Odd cohomology vanishes for compactified Jacobians of spectral curves with double points","Affine paving of compactified Jacobians matches local smoothening stratification","Cohomology multiplicities computed via polynomial formula in the Bass case"]},"model":"glm-5.2","effort":"low","cost_usd":0.0,"raw_usage":{"total_tokens":631,"prompt_tokens":558,"completion_tokens":73,"prompt_tokens_details":null},"tokens_in":558,"tokens_out":73,"duration_ms":18628,"temperature":1.0,"reasoning_tokens":null,"cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-09T21:04:58.442807+00:00","model_set":{"reader":"glm-5.2"},"falsifier":"A counterexample to the smoothness of the local morphisms φ_M (Theorem 2.2) or φ_{l_1} (Theorem 2.4) at some point over χ_γ would break the formal smoothness lifting in Proposition 3.22, collapsing the proof that the global strata are smooth. Alternatively, a spectral curve with double Bass singularities and normalization P^1 whose compactified Jacobian cohomology does not match the predicted Poincaré polynomial would directly falsify Theorem 5.3.","supporting_citations":[],"review_version":1}