{"id":"0bfea49a-fe05-40db-9afb-5da5dc6012a5","arxiv_id":"2507.22319","paper_version":4,"verdict":"REJECT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"For smooth projective curves over global fields, the mod-l kernel of the boundary map is governed by local reduction data and the Galois coinvariants J[l]_{G_F}.","lead":"This paper derives an exact sequence for the mod-l kernel of the boundary map on higher Chow groups of curves over global fields, using a Hasse principle in Galois cohomology. The result extends the author's earlier number-field elliptic curve work to all global fields and includes explicit elliptic curve computations.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The trivial-image case of Theorem 1.1 is internally inconsistent with the paper's own Theorem 4.6: for the constant elliptic curve y^2+y=x^3 over F_4(t) at l=3, V(E,3)=0 forces Coker(∂^good_{E,3})≅E[3]_{G_F}, of dimension 2, so the boundary map cannot be bijective.","rationale":"The reader's verdict of REJECT is supported. I read the paper in good faith: the central contribution is the exact sequence (4.6) describing Ker(∂^good) and Coker(∂^good) via local terms and J[l]_{G_F}, under a Hasse-principle injectivity assumption. The advertised Theorem 1.1, however, contains a trivial-image clause that cannot survive the paper's own exact sequence. The constant elliptic curve y^2+y=x^3 over F_4(t) at l=3 realizes exactly that clause: E[3] is rational, reduction is good everywhere, V(E,3)=0, and E[3]_{G_F} is a 2-dimensional vector space. Plugging into (4.6) gives Coker(∂^good) of dimension 2, so the boundary map is not bijective, contradicting Theorem 1.1; Proposition 4.11's surjectivity claim for E[l]⊂E(F) is also false in this example. This is an internal inconsistency, not a disagreement with external consensus: the contradiction uses only Theorem 4.6 together with standard facts about a concrete curve. The reader's formal weakest_assumption focuses on injectivity of loc_l, but in this counterexample loc_l is injective; the load-bearing failure is that the trivial-image case leaves a nonzero coinvariant term J[l]_{G_F} in the exact sequence. Thus I partially agree with the reader: the rationale identifies the same contradiction, while the stated weakest assumption does not. The concrete check above would settle the matter decisively by making the contradiction explicit; if the computation is accepted, the paper as stated requires substantial repair. No ad hominem is intended; the issue is a mathematical inconsistency in the stated theorems.","tokens_in":23641,"tokens_out":35857,"duration_ms":435878,"concrete_test":"Run the paper's own Theorem 4.6 on the explicit constant curve E: y^2+y=x^3 over F=F_4(t) with l=3. Verify three hypotheses: (1) E has good reduction at all finite places because the Weierstrass discriminant is a nonzero constant; (2) ρ_{E,3} is trivial because E0[3]⊂E0(F_4)⊂E(F); (3) loc_3 is injective by Lemma 4.2 and Proposition 4.3(ii). Then V(E,3)=0 and E[3]_{G_F}≅F_3^2, so the exact sequence (4.6) reads 0→0→0→F_3^2→Coker(∂^good_{E,3})→0, forcing dim Coker(∂^good_{E,3})=2. If a defender believes Theorem 4.6 is not applicable, they should identify the exact step in the constant-curve specialization of diagram (4.8) and the snake lemma that fails; if Theorem 4.6 is applicable, Theorem 1.1 and Proposition 4.11 are false as stated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 1.1 advertises that if the mod-l image of ρ_{J,l} is either sufficiently large or trivial, then ∂^good_{X,l} is bijective. The trivial-image clause is the exposed part. Theorem 4.6, which the paper presents as the precise refinement, gives the exact sequence 0→Ker(∂^good_{X,l})→V(X,l)→J[l]_{G_F}→Coker(∂^good_{X,l})→0 whenever loc_l is injective, and the hypotheses guaranteeing loc_l injection are satisfied when the representation is trivial (Lemma 4.2, Proposition 4.3(ii)). Now take E0: y^2+y=x^3 over F_4; one checks E0(F_4)≅(Z/3)^2, so E0[3]⊂E0(F_4). Let F=F_4(t), l=3, and E=E0⊗_{F_4}F. The equation has unit discriminant, so E has good reduction at every finite place; hence the function-field definition in Theorem 4.6 gives V(E,3)=Σ_{v bad} V(E_v)/3=0. Since E[3]=E0[3]⊂E(F), G_F acts trivially on E[3], so E[3]_{G_F}=E[3] is a 2-dimensional F_3-vector space. Substituting into (4.6) yields 0→Ker(∂^good)→0→E[3]→Coker(∂^good)→0, so Coker(∂^good_{E,3})≅E[3], which has dimension 2. This contradicts the bijectivity claimed in Theorem 1.1 and also contradicts Proposition 4.11, which asserts surjectivity under the E[l]⊂E(F) clause. The problem is structural: for trivial image, the coinvariant term J[l]_{G_F} is nonzero, so the exact sequence cannot collapse to an isomorphism unless V(X,l) is large enough to map onto it. In the constant-curve example V(X,l) is zero, so the advertised bijection fails. The reader's stated weakest assumption, injectivity of loc_l, is not the issue here: loc_l is injective in this example. The inconsistency is between Theorem 1.1 and Theorem 4.6, and it is load-bearing because the paper's headline claims the bijection in the trivial-image case.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the kernel V(X) of the push-forward map CH^2(X,1) -> F^× for a smooth projective curve X over a global field F. Combining Bloch's exact sequence, local class field theory for curves, and a cohomological Hasse principle for the mod-l Galois representation on J[l], the author derives an exact sequence (Theorem 4.6) relating the kernel and cokernel of the mod-l boundary map ∂^good_{X,l} to local data V(X,l) and the coinvariants J[l]_{G_F}, under an injectivity assumption loc_l. It then claims (Theorem 1.1) that ∂^good is bijective when the image of ρ_{J,l} is either sufficiently large or trivial, and it gives local computations for elliptic curves, including split and non-split multiplicative reduction and worked examples over Q.","tokens_in":24126,"tokens_out":12745,"duration_ms":151936,"significance":"If the exact sequence in Theorem 4.6 is correct, it provides a genuine Hasse-principle framework for the mod-l structure of V(X): the failure of ∂^good to be an isomorphism is measured by bad-reduction local terms and by J[l]_{G_F}. The derivation from Bloch's exact sequence and standard duality theorems is transparent, and the local computations for Tate curves are useful. However, the advertised trivial-image clause of Theorem 1.1 is false, and Proposition 4.11's E[l]⊂E(F) surjectivity claim is false; these must be corrected. The core exact sequence appears salvageable and, after removing the false claims, the paper would be a worthwhile contribution.","major_comments":[{"comment":"The trivial-image case is internally inconsistent with the paper's own refinement. Let E0 be y^2+y=x^3 over F_4; one checks #E0(F_4)=9, so E0[3]⊂E0(F_4). Put F=F_4(t), E=E0⊗_{F_4}F, and l=3. The discriminant of this model is 1, so E has good reduction at every finite place; hence the function-field definition in Theorem 4.6 gives V(E,3)=0. Since G_F acts trivially on E[3], E[3]_{G_F} is a 2-dimensional F_3-vector space. The image of ρ_E,3 is trivial, so loc_3 is injective by Lemma 4.2 and Proposition 4.3(ii). Substituting into (4.6) yields 0→0→0→E[3]→Coker(∂^good_{E,3})→0, forcing Coker(∂^good_{E,3})≅E[3], which is nonzero. This contradicts the bijectivity claimed in Theorem 1.1. The flaw is structural: with trivial image the coinvariant quotient J[l]_{G_F} need not vanish, and it must be hit by V(X,l) before an isomorphism can occur.","section":"§1, Theorem 1.1; §4.2, Theorem 4.6"},{"comment":"The assertion that E[l]⊂E(F) implies ∂^good_{E,l} is surjective is false; the same constant elliptic curve satisfies E[3]⊂E(F) but has Coker(∂^good_{E,3})≅E[3] by the exact sequence. The local-approximation proof cannot be valid globally, since it would conflict with the global reciprocity constraint encoded in (4.6). Please remove this case or replace it with the exact statement supplied by Theorem 4.6.","section":"§4.3, Proposition 4.11"},{"comment":"Theorem 4.6 is stated under the assumption that loc_l: V(X)/l → ⊕_v V(X_v)/l is injective. This hypothesis is not mentioned in the abstract or in the statements of Theorems 1.1 and 1.2 that depend on Theorem 4.6. Even after removing the false trivial-image clause, the remaining statements should explicitly include this hypothesis or prove it under their hypotheses; as written, the reader cannot tell when the advertised Hasse principle is conditional.","section":"§4.2, Theorem 4.6 and hypothesis loc_l"}],"minor_comments":[{"comment":"The commutative diagram used for the snake lemma is not fully typeset; the vertical maps from the top row to the bottom row and the source of the arrow to J[l]_{G_F} are missing. Please redraw the diagram so that the application of the snake lemma is checkable.","section":"§4.2, diagram (4.8)"},{"comment":"Lemma 4.7 is cited from the unpublished preprint [Hir], although a proof is given in the text. Please state explicitly that the proof is self-contained, so the main theorem does not depend on unpublished work.","section":"§4.3, Lemma 4.7"},{"comment":"The examples invoke [Hir, Lemma 4.1] for surjectivity in the (SC_l) case over Q; since [Hir] is unpublished, please either include a proof of the needed number-field statement or clearly label those example conclusions as conditional on [Hir].","section":"Section 5, Examples 5.1 and 5.2"},{"comment":"The phrase 'or is trivial' in Theorem 1.1 and the matching claim in Proposition 4.11 should be deleted or replaced by the correct cokernel statement from Theorem 4.6, and the abstract should be amended accordingly.","section":"Abstract and Introduction"}],"recommendation":"major_revision","confidential_remarks":"The referee's main concern is that the false trivial-image clause may indicate the authors did not test the result against constant curves; the constant elliptic curve counterexample is elementary and should have been caught. If the authors can fix the statements by removing the false clause and correcting Proposition 4.11, the remaining core of the paper, especially Theorem 4.6, appears defensible. The reliance on [Hir] for the worked examples should also be addressed before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You probably already saw this from the reader's report, but let me confirm: the trivial-image clause in Theorem 1.1 is wrong, and the paper's own Theorem 4.6 proves it. Take E0: y^2+y=x^3 over F_4, E = E0 ⊗ F_4(t), l=3. E has good reduction everywhere, so V(E,3)=0 by the definition in Theorem 4.6. E[3] lies in E(F) so G_F acts trivially, making E[3]_{G_F} a 2-dimensional F_3-space. Lemma 4.2 gives loc_l injectivity (trivial representation kills X^1). Plug into (4.6): 0→Ker→0→E[3]→Coker→0, so Coker ≅ E[3] has dimension 2 and the good boundary map is not surjective. That directly contradicts Theorem 1.1 and also Proposition 4.11. The flaw is not the loc_l assumption; it's the overclaim that trivial image forces bijectivity. The coinvariant term J[l]_{G_F} cannot vanish in the trivial-image case, so the exact sequence only collapses to an isomorphism if V(X,l) is big enough to map onto it—which fails for constant curves.\n\nWhat's genuinely good: Theorem 4.6 itself appears correctly derived from Bloch's exact sequence and the snake lemma. The function-field case is new, and dropping the condition E[l]^{G_F}≠0 from the elliptic curve statement is real progress over [Hir]. The use of the cohomological Hasse principle X^1(G, J[l]^∨)=0 to get loc_l injective is a nice input, and the local computations (Lemmas 3.6, Prop 3.10) are solid and useful. The examples for conductor 11 and 14 are worked carefully.\n\nThe other substantive issue is modest: the paper cites [Hir] for Lemma 4.7 and for the number-field surjectivity result, and [Hir] is unpublished. Not a fatal problem, but the referee should be able to see those arguments.\n\nMy take: the advertised big-monodromy consequence (semistable toric dimension one) avoids the trivial-image case and may well survive, but Theorem 1.1 and Prop 4.11 need to be corrected before the paper is publishable. The exact sequence is worth saving. This deserves a serious referee and a major revision, not a desk reject and not an accept as-is.","headline":"Theorem 1.1's trivial-image clause is false, contradicted by the paper's own Theorem 4.6 via a constant elliptic curve, but the exact sequence is new and worth salvaging.","tokens_in":24759,"tokens_out":3323,"would_cite":false,"duration_ms":35670,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11G05","14C15","19D45"],"pacs":[],"model":"deepseek-v4-flash","headline":"The mod-l structure of the kernel of the higher Chow group boundary map of a curve is determined by the l-torsion of its Jacobian and local reduction data.","keywords":["higher Chow groups","global fields","elliptic curves","Jacobian varieties","mod-l Galois representations","Somekawa K-groups","boundary map","Bloch's conjecture"],"falsifier":"For the elliptic curve 14a1 at $l=3$, compute the four $\\mathbb{F}_3$-vector spaces in the exact sequence (4.6) — $\\mathrm{Ker}(\\partial^{\\mathrm{good}}_{E,3})$, $V(E,3)$, $E[3]_{G_\\mathbb{Q}}$, and $\\mathrm{Coker}(\\partial^{\\mathrm{good}}_{E,3})$ — from the paper's local formulas and independently from a direct computation of the Somekawa $K$-group $V(E)/3$; exactness forces the alternating sum of their dimensions to be zero, so any non-zero value would disprove the central claim.","tokens_in":23356,"feed_emoji":"🧮","tokens_out":16180,"duration_ms":163108,"temperature":0.7,"pith_summary":"The paper proves a Hasse principle for the higher Chow group $\\mathrm{CH}^2(X,1)$ of a smooth projective curve $X$ over a global field: the mod-$l$ quotient of the kernel $V(X)$ of the push-forward to $F^\\times$ is governed by the mod-$l$ Galois representation on the $l$-torsion of the Jacobian $J$ of $X$ and by local reduction data. The main result is an exact sequence $0\\to\\mathrm{Ker}(\\partial^{\\mathrm{good}}_{X,l})\\to V(X,l)\\to J[l]_{G_F}\\to\\mathrm{Coker}(\\partial^{\\mathrm{good}}_{X,l})\\to 0$ for odd primes $l\\neq\\mathrm{char}(F)$, which says that the boundary map's failure to be an isomorphism is measured exactly by special local terms and by the Galois coinvariant quotient of $J[l]$. When the mod-$l$ Galois image contains $\\mathrm{Sp}_{2g}(\\mathbb{F}_l)$ or is trivial, the boundary map is bijective; for elliptic curves the sequence is unconditional and gives explicit computations of the kernel and cokernel. The result makes the mod-$l$ structure of these torsion groups an explicit function of computable arithmetic data, placing higher Chow groups of curves on the same footing as the classical Milnor $K$-group $K_2$ of a number field.","feed_headline":"Jacobian l-torsion controls mod-l higher Chow groups of curves","feed_subtitle":"The boundary map's failures reduce to local data plus the Galois action on Jacobian torsion.","key_machinery":"The machinery is the boundary map $\\partial_X\\colon V(X)\\to\\bigoplus_v A_0(X_v)$ supplied by the class field theory of arithmetic surfaces, together with the global-to-local exact sequence of Proposition 4.1(ii) (the 'Bloch sequence') $V(X)/m\\to\\bigoplus_v V(X_v)/m\\to T(X)_{G_F}\\to 0$, which reduces mod $l$ to identify the final term with the Galois coinvariants $J[l]_{G_F}$. The other load-bearing ingredient is a Hasse principle in Galois cohomology: when $X^1(G,J[l]^\\vee)=0$, the localization map $\\mathrm{loc}_l$ is injective, and for elliptic curves this vanishing is unconditional via Ramakrishnan's cohomological Hasse principle and Chebotarev. A snake-lemma chase over the commutative diagram relating the global sequence to the sum of local boundary maps then yields the four-term exact sequence, with the vanishing of kernels of local boundary maps at good places (Lemma 3.5) and the explicit computations for Tate curves at bad places (Lemma 3.6, Proposition 3.10) supplying the local terms.","core_discovery":"The central claim is the exact sequence (4.6): for an odd prime $l\\neq\\mathrm{char}(F)$, assuming the localization map $\\mathrm{loc}_l$ is injective, there is an exact sequence of finite-dimensional $\\mathbb{F}_l$-vector spaces $0\\to\\mathrm{Ker}(\\partial^{\\mathrm{good}}_{X,l})\\to V(X,l)\\to J[l]_{G_F}\\to\\mathrm{Coker}(\\partial^{\\mathrm{good}}_{X,l})\\to 0$, where $\\partial^{\\mathrm{good}}_{X,l}$ is the mod-$l$ boundary map at places of good reduction, $V(X,l)$ is built from kernels of local boundary maps at places dividing $l$ and from $V(X_v)/l$ at bad places, and $J[l]_{G_F}$ is the Galois coinvariant quotient of the $l$-torsion of the Jacobian. In particular, the kernel and cokernel of the boundary map are governed by local reduction data and the representation on $J[l]$. Two consequences the paper draws are that the boundary map is bijective whenever the image of $\\rho_{J,l}$ contains $\\mathrm{Sp}_{2g}(\\mathbb{F}_l)$ or is trivial, and that for elliptic curves the sequence is unconditional, with explicit computations of the mod-$l$ kernel for the conductor-11 and conductor-14 curves in Section 5.","pith_inferences":["An implicit corollary of the four-term exact sequence is an Euler-characteristic identity $\\dim\\mathrm{Ker}(\\partial^{\\mathrm{good}}_{X,l})-\\dim V(X,l)+\\dim J[l]_{G_F}-\\dim\\mathrm{Coker}(\\partial^{\\mathrm{good}}_{X,l})=0$, which a reader could verify numerically for any explicit curve with a computable mod-$l$ representation; this is the cleanest testable footprint of the theorem.","Because the only global hypothesis is injectivity of $\\mathrm{loc}_l$, the theorem suggests that any failure of the Hasse principle for higher Chow groups of curves of genus $\\ge 2$ would be purely Galois-cohomological in origin, not geometric; searching for a genus-2 Jacobian with reducible mod-$l$ image and non-vanishing $X^1(G,J[l]^\\vee)$ would probe that boundary.","The parallel with $K_2$ of a global field, whose tame-symbol kernel is finite and connected to special $L$-values, suggests that the mod-$l$ boundary map may lead to analytic formulas for the size of its kernel in terms of $L$-functions of $J$, especially when combined with the cited function-field relation between $V(E)/l$ and $L(E,0)$."],"forward_implications":["For an elliptic curve $E$ over a global field, whenever the mod-$l$ Galois representation contains $\\mathrm{SL}_2(\\mathbb{F}_l)$, one has $E[l]_{G_F}=0$, so the mod-$l$ boundary map is surjective and its kernel is exactly the explicit local group $V(E,l)$.","If $\\mathrm{End}_{\\overline F}(J)=\\mathbb{Z}$ and $J$ has semistable reduction of toric dimension one at some place, then $\\partial^{\\mathrm{good}}_{X,l}\\colon V(X)/l\\to\\bigoplus_{v\\text{ good}}J_v(F_v)/l$ is an isomorphism for all but finitely many primes $l\\neq\\mathrm{char}(F)$.","Computing the mod-$l$ structure of $V(X)$ reduces to finitely many local calculations (Tate periods, reduction types, residue fields) plus the dimension of the Galois coinvariant space $J[l]_{G_F}$; the Section 5 examples exhibit the resulting isomorphisms for conductor-11 and conductor-14 elliptic curves.","In the function-field case, if $E[l]\\subset E(F)$ or the isogeny condition $(SC_l)$ of the paper holds, the boundary map $\\partial^{\\mathrm{good}}_{E,l}$ is surjective, giving a positive-characteristic analogue of the number-field theorem.","The exact sequence determines the boundary map's behaviour at bad and archimedean places: a good place $v$ with $v\\mid l$ contributes nothing under mild ramification assumptions (Lemma 3.5), and real places contribute nothing for odd $l$."],"supporting_citations":[{"why":"provides the global-to-local exact sequence (Prop 4.1(ii)) and the boundary map $\\partial_X$ used to define the local terms.","marker":"[KS83]"},{"why":"proves the injectivity of the Galois symbol map $V(X)/m\\to H^2(F,J[m](1))$, the bridge from higher Chow groups to Galois cohomology.","marker":"[Yam05]"},{"why":"supplies the global Tate duality that identifies the obstruction with the dual of $X^1(G,J[l]^\\vee)$, giving the Hasse-principle hypothesis.","marker":"[Mil06]"},{"why":"is the cohomological Hasse principle used to prove $X^1(G,E[l]^\\vee)=0$ for elliptic curves, making $\\mathrm{loc}_l$ injective.","marker":"[Ram]"},{"why":"establishes $V(X)\\simeq K(F;J,\\mathbb{G}_m)$, the Somekawa $K$-group identification that underlies the boundary-map formula.","marker":"[Som90]"},{"why":"is the source of local class field theory for curves over local fields, giving finiteness of $V(X)_{\\mathrm{red}}$ and the structure of the boundary map.","marker":"[Sai85]"},{"why":"computes the cokernel of the reciprocity map and the toric rank, fixing the local terms in the exact sequence.","marker":"[Yos03]"},{"why":"gives the Tate uniformization and the $j$-invariant--$q(E)$ relation used to compute $V(E)/l$ for split multiplicative reduction.","marker":"[Sil13]"},{"why":"describes Milnor $K_2$ of a local field as $\\mu_M\\oplus$ divisible, the key to Lemma 3.6's dimension formula.","marker":"[FV02]"}],"fun_headline_variants":["Jacobian torsion coinvariants dictate higher Chow group boundary maps","Hasse principle yields exact sequence for higher Chow groups of curves","Boundary map failures reduce to local data and Jacobian torsion coinvariants","Exact sequence links higher Chow kernel and cokernel to Jacobian torsion"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the localization map $\\mathrm{loc}_l\\colon V(X)/l\\to\\bigoplus_v V(X_v)/l$ is injective, a Hasse-principle condition the paper proves for elliptic curves and for Jacobians with large or trivial Galois image, but which is not established for arbitrary curves of genus at least two; if it fails, the exact sequence (4.6) is not known to hold.","fun_headline_variants_meta":{"raw":{"variants":["Jacobian torsion coinvariants dictate higher Chow group boundary maps","Hasse principle yields exact sequence for higher Chow groups of curves","Boundary map failures reduce to local data and Jacobian torsion coinvariants","Exact sequence links higher Chow kernel and cokernel to Jacobian torsion"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00135,"raw_usage":{"total_tokens":5512,"prompt_tokens":1008,"completion_tokens":4504,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":624,"completion_tokens_details":{"reasoning_tokens":4427}},"tokens_in":624,"tokens_out":4504,"duration_ms":35405,"temperature":1.0,"reasoning_tokens":4427,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T11:54:40.624207+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For the elliptic curve 14a1 at $l=3$, compute the four $\\mathbb{F}_3$-vector spaces in the exact sequence (4.6) — $\\mathrm{Ker}(\\partial^{\\mathrm{good}}_{E,3})$, $V(E,3)$, $E[3]_{G_\\mathbb{Q}}$, and $\\mathrm{Coker}(\\partial^{\\mathrm{good}}_{E,3})$ — from the paper's local formulas and independently from a direct computation of the Somekawa $K$-group $V(E)/3$; exactness forces the alternating sum of their dimensions to be zero, so any non-zero value would disprove the central claim.","supporting_citations":[],"review_version":1}