{"id":"40c154e9-2734-4f8d-9501-87eddbf7a03b","arxiv_id":"1908.06424","paper_version":3,"verdict":"REJECT","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"For a CM elliptic curve E over a number field, the quotient of the n-torsion of the image of the flat pullback on codimension-two cycles by the n-torsion of the model's Chow group is claimed to be pro-finite.","lead":"This paper studies a Chow group quotient associated to the self-product of a CM elliptic curve and a smooth integral model, claiming a pro-finiteness statement for its n-torsion. The result is a technical contribution to arithmetic geometry, but the proof contains gaps that call the stated theorem into question.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The proof of Theorem 2.2 applies Lemma 2.3 to the base-changed kernel Σ_{\\bar K}, but Mildenhall's theorem only gives finiteness of the kernel over the base field K; profiniteness of Σ_{\\bar K} is asserted without support, so the central conclusion is not established.","rationale":"The reader's weakest_assumption identifies exactly the load-bearing gap I find in the proof. Section 2 first defines the finite kernel Σ_K via Mildenhall's theorem, then silently introduces the base-changed kernel Σ_{\\bar K} and uses it in the Galois cohomology sequence. The final application of Lemma 2.3 says Σ_{\\bar K} is pro-finite by Mildenhall's theorem, but the cited theorem's finiteness statement concerns the base-field kernel, not the kernel after scalar extension to \\bar K. One cannot deduce profiniteness of the base-changed kernel from finiteness over K: Chow groups can grow under scalar extension, and the finite group over K does not control the group over \\bar K. This is not a presentational issue; it is the only bridge from Mildenhall's finiteness to the pro-finiteness of S_n, and it is missing. The additional step from S_n to the quotient is also not fully justified, but it is secondary. Because the central theorem inherits this gap, the reader's REJECT verdict stands unchanged.","tokens_in":5208,"tokens_out":12301,"duration_ms":126656,"concrete_test":"Read Mildenhall's theorem as stated in the cited Duke 1992 paper and verify exactly which kernel it bounds. If it only proves Σ_K finite for the model over O_K[1/6N], then test the base-changed kernel by computing the direct limit of the kernels over finite extensions L/K: Σ_{\\bar K} = colim_L ker(A^2(𝒳_{O_L[1/6N]}) → A^2(E_L×E_L)). For the CM elliptic curve E, try to construct, for infinitely many L, nontrivial elements in these kernels using correspondences from endomorphisms of E defined over L, and check whether they yield an infinite direct sum inside Σ_{\\bar K}. A strictly increasing direct limit of finite abelian groups is not pro-finite unless it stabilizes, so exhibiting such a sequence would settle that Lemma 2.3 cannot be applied. If, instead, all such classes are already defined over K and the kernel is finite, the profiniteness hypothesis would be satisfied.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is Theorem 2.2: im(j*)[n](K)/A2(X)[n](K) is pro-finite. The proof's final step invokes Lemma 2.3 with M = Σ_{\\bar K}, the kernel of the flat pullback after base change to the algebraic closure, and says this is pro-finite by Mildenhall's theorem. But Mildenhall's theorem, as cited in §2, states only that the kernel Σ_K of the restriction map for the model over O_K[1/6N] is finite. It says nothing about Σ_{\\bar K} = ker(A^2(𝒳_{\\bar O[1/6N]}) → A^2(E_{\\bar K}×E_{\\bar K})). These are different groups. In the Galois-invariant exact sequence at the start of §2, the cohomology group that receives the quotient im(j*)(K)/A2(X)(K) is H^1(G, Σ_{\\bar K}), not H^1(G, Σ_K); replacing Σ_{\\bar K} by the finite Σ_K would require a G-stable identification that is not provided and is generally false, since Chow groups can acquire new kernel classes after scalar extension. Without profiniteness of Σ_{\\bar K}, the hypothesis of Lemma 2.3 is unmet, and the pro-finiteness of S_n(Σ_{\\bar K}) does not follow. The theorem therefore rests on an unsupported assumption. A secondary gap is that the paper moves from pro-finiteness of S_n to pro-finiteness of the quotient without showing the quotient embeds as a closed subgroup of S_n; but the primary failure is already the profiniteness of Σ_{\\bar K}.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the flat pullback map j*: A^2(X) → A^2(X), where X is a smooth spread of the self-product E×E of a CM elliptic curve over the ring of integers localized away from 6N, and A^2 denotes the group of algebraically trivial codimension-two cycles modulo rational equivalence. It defines an n-Selmer group S_n(Σ_{\\bar K}) as the kernel of H^1(G,Σ_{\\bar K}[n]) → ∏_v H^1(G_v,A^2(X_{\\bar K_v}))[n], and attempts to prove that S_n(Σ_{\\bar K}) is profinite, and hence that im(j*)[n](K)/A^2(X)[n](K) is profinite. The proof follows the pattern of finiteness proofs for elliptic curve Selmer groups: it invokes Mildenhall's finiteness result for the kernel of restriction, Roitman's isomorphism between n-torsion of A^2 and Alb, and Raskind's injectivity of specialization to show that Selmer elements are unramified outside a finite set, then cites a profinite version of Silverman's Lemma X.4.3.","tokens_in":5542,"tokens_out":16900,"duration_ms":168514,"significance":"If the argument were correct, the theorem would be a nontrivial contribution: it would provide a Selmer-type profiniteness statement for the cokernel of flat pullback on algebraically trivial zero-cycles for self-products of CM elliptic curves, going beyond Mildenhall's finiteness of the kernel. The paper is concise and the analogy with elliptic curve Selmer groups is appealing. The construction of the Selmer group is a standard one and is not circular, and the main cited theorems (Mildenhall, Roitman, Raskind) are appropriate tools. However, as detailed below, the central proof rests on an unsupported profiniteness assertion, and there are additional gaps in the passage from the Selmer group to the desired quotient. The intended result is therefore not established by the manuscript.","major_comments":[{"comment":"The claim that Σ_{\\bar K} is profinite by Mildenhall's theorem is unsupported. Mildenhall's cited theorem gives finiteness of the kernel Σ_K for the spread over O_K[1/6N], not of the base-changed kernel Σ_{\\bar K} over the integral closure of O_K in \\bar K. These are different groups, and Chow groups may acquire new kernel classes after scalar extension. Even if Mildenhall's finiteness held over every finite extension L/K, the group Σ_{\\bar K} would be a filtered colimit of finite groups, which need not be profinite. Since Lemma 2.3 requires M to be a profinite G-module, its application to M=Σ_{\\bar K} is unjustified.","section":"Section 2, exact sequence after Mildenhall's finiteness; final paragraph of proof of Theorem 2.2"},{"comment":"The proof applies Lemma 2.3 to M=Σ_{\\bar K}, but S_n(Σ_{\\bar K}) is defined as a subgroup of H^1(G,Σ_{\\bar K}[n]), not of H^1(G,Σ_{\\bar K}). Even if Lemma 2.3 were applicable to Σ_{\\bar K}, it would yield profiniteness of H^1(G,Σ_{\\bar K};S), not of S_n(Σ_{\\bar K}). To prove S_n profinite one would need the lemma for the module Σ_{\\bar K}[n] (or a separate argument), and this is not what the manuscript does.","section":"Definition 2.1 and proof of Theorem 2.2 (use of Lemma 2.3)"},{"comment":"The step 'S_n(Σ_{\\bar K}) is pro-finite, hence im(j*)[n](K)/A^2(X)[n](K) is pro-finite' is not justified. The quotient embeds into H^1(G,Σ_{\\bar K}[n]) and, under the intended argument, into S_n, but an abstract subgroup of a profinite group need not itself be profinite unless it is closed (or otherwise known to be finite). No closedness or finiteness argument is supplied, so the final implication does not follow from the profiniteness of S_n alone.","section":"Proof of Theorem 2.2, final sentence"}],"minor_comments":[{"comment":"The notation is inconsistent: the paper writes both S_n(Σ_{\\bar K}) and S_n(Σ_K), and it defines S_n using H^1(G,Σ_{\\bar K}[n]) while later invoking Lemma 2.3 with M=Σ_{\\bar K}; these should be reconciled.","section":"Throughout, especially Definition 2.1 and final paragraph"},{"comment":"The sentence 'Σ^G_{\\bar K} is a subgroup of Σ_K. For our notational convenience we continue to denote Σ^G_{\\bar K} as Σ_K' is confusing, because the immediately following cohomology group is H^1(G,Σ_{\\bar K}), not H^1(G,Σ^G_{\\bar K}); this notational shift should be clarified or removed.","section":"Section 2, paragraph after the exact sequence"},{"comment":"Lemma 2.3 is stated as a 'variance' of Silverman's Lemma X.4.3 without proof or a precise topology on H^1(G,M;S); the cited lemma concerns finiteness of Selmer groups for elliptic curves and does not directly justify the stated profiniteness assertion for arbitrary profinite G-modules.","section":"Lemma 2.3"},{"comment":"There are several typographical errors, including 'flat-pullbcak' in the introduction, 'th ere' in the proof of Theorem 2.2, and 'variance of lemma' instead of 'variant of Lemma'.","section":"Introduction and proof"}],"recommendation":"reject","confidential_remarks":"The central gap is not a local fix: the paper needs profiniteness of Σ_{\\bar K}, which Mildenhall's theorem does not provide and which is unlikely to hold in the natural colimit description. The additional conflation of the Selmer group with H^1(G,Σ_{\\bar K};S) and the unsupported final implication would require substantial reworking. I therefore recommend rejection, though the underlying Selmer-group analogy may be worth pursuing in future work."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nHere is my take on arXiv:1908.06424. The paper proves — or aims to prove — that for the self-product of a CM elliptic curve over a number field, the quotient im(j^*)[n](K)/A^2(X)[n](K) is pro-finite. That statement is new; Mildenhall proved finiteness of the kernel of the flat pullback over the base ring, not pro-finiteness of this quotient.\n\nWhat the paper does well: the strategy is sensible. It sets up the Galois exact sequence involving the kernel Σ, defines a Selmer group S_n(Σ) by slicing with n-torsion, and tries to show the relevant cohomology classes are unramified away from a finite set of places, using Roitman's theorem and Raskind's injectivity result. The authors engage with the right literature and the proof structure is recognizable from standard Selmer group arguments.\n\nThe soft spot, and it is load-bearing: the proof applies Lemma 2.3 to M = Σ_{\\bar K}, the kernel over the algebraic closure, and says this is pro-finite by Mildenhall's theorem. Mildenhall's theorem only gives finiteness of the kernel over the base number field K. The kernel over \\bar K is a colimit of the finite kernels over finite extensions; there is no reason it is pro-finite. The lemma's hypothesis is therefore unmet. This is not a minor gap: without pro-finiteness of Σ_{\\bar K}, the conclusion that S_n is pro-finite does not follow.\n\nA secondary issue: even granting S_n pro-finite, the paper does not fully justify that the quotient injects into S_n as a closed subgroup, though that step is likely repairable.\n\nThe notation and typos also make the paper harder to trust (regular/overline fonts are confused throughout).\n\nOverall: the theorem may be true, and a fix is likely possible — apply Lemma 2.3 to the finite n-torsion module Σ_{\\bar K}[n] instead of Σ_{\\bar K}, then check the local conditions. But as written, the central proof is invalid. I would not accept the paper in this form. I'd send it back for a serious revision, but if I had to decide desk reject vs. referee, I'd lean toward sending to a referee who knows Mildenhall's paper, because the error is specific and the intended result is plausible.\n\nFor your reading group: skip it unless someone wants to work through the fix.\n\nBest.","headline":"A plausible extension of Mildenhall's theorem is undermined by a misapplication of finiteness to the algebraic closure, leaving the central claim unproven.","tokens_in":6124,"tokens_out":8308,"would_cite":false,"duration_ms":81601,"reading_group":"no","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11G05","11G15","14C25","14K22"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that for the self-product of a CM elliptic curve over a number field, the Galois-fixed n-torsion cokernel of flat pullback on algebraically trivial codimension-two cycles is a profinite group.","keywords":["complex multiplication","elliptic curve","Selmer group","Tate-Shafarevich group","Chow group","Abelian variety","codimension two cycles","profinite group"],"falsifier":"Take a concrete CM elliptic curve $E/K$, fix an integer $n$, and look for a class $\\eta\\in H^1(G,\\Sigma_{\\bar K}[n])$ whose image in every local group $H^1(G_v,A^2(\\mathscr{X}_{\\bar K_v}))[n]$ is zero but whose restriction to the inertia group $I_v$ is nonzero for infinitely many good primes $v\\nmid n$. The theorem predicts no such class exists, so exhibiting one would disprove the profiniteness conclusion.","tokens_in":4962,"feed_emoji":"🧮","tokens_out":9124,"duration_ms":89848,"temperature":0.7,"pith_summary":"For a surface obtained as the self-product of an elliptic curve with complex multiplication over a number field, the paper proves a finiteness principle for codimension-two cycles. The object studied is the cokernel, at the level of n-torsion and Galois-fixed cycles, of the flat pullback from a smooth integral spread to the generic fiber: this cokernel is shown to be a profinite group, an inverse limit of finite groups. The proof constructs an n-Selmer group attached to the kernel of that pullback and shows all its elements are unramified outside finitely many places. If correct, the result says that the arithmetic of algebraically trivial 2-cycles in this family is controlled by cohomological data of the same flavor as the Selmer groups bounding abelian varieties.","feed_headline":"CM elliptic curve squares have profinite cycle cokernels","feed_subtitle":"For each n, the Galois-fixed n-torsion image of flat pullback modulo the spread's cycles is an inverse limit of finite groups.","key_machinery":"The central object is the Galois module $\\Sigma_{\\bar K}$, the kernel of $A^2(\\mathscr{X}_{\\bar K})\\to A^2(X_{\\bar K})$, together with its associated $n$-Selmer group $S_n(\\Sigma_{\\bar K})$, defined as the kernel of the global-to-local map $H^1(G,\\Sigma_{\\bar K}[n])\\to \\prod_v H^1(G_v,A^2(\\mathscr{X}_{\\bar K_v}))[n]$. The argument shows that every class in this Selmer group is unramified at all but finitely many places: at good primes $v\\nmid n$, the class maps into inertia cohomology, and Raskind's specialization injectivity forces the image to vanish. Then, assuming $\\Sigma_{\\bar K}$ is profinite, Silverman's profinite unramified-cohomology lemma makes $S_n(\\Sigma_{\\bar K})$ profinite. Roitman's isomorphism $A^2(X)[n]\\cong \\operatorname{Alb}(X)[n]$ as Galois modules is the bridge that lets standard abelian-variety Selmer techniques be applied to the Chow group.","core_discovery":"Let $E$ be an elliptic curve over a number field $K$ with complex multiplication by the ring of integers, discriminant $N$, and let $X=E\\times E$. Choose a smooth spread $\\mathscr{X}$ over $\\mathcal{O}_K[1/6N]$, and let $A^2$ denote the group of algebraically trivial codimension-two cycles modulo rational equivalence. The paper's main theorem, Theorem 2.2, states that for the flat pullback $j^*:A^2(\\mathscr{X})\\to A^2(X)$, the quotient $\\operatorname{im}(j^*)[n](K)/A^2(\\mathscr{X})[n](K)$ is a profinite group. The proof introduces the $n$-Selmer group $S_n(\\Sigma_{\\bar K})$ attached to the kernel $\\Sigma_{\\bar K}$ of the restriction map after base change to the algebraic closure, proves that elements of this Selmer group are unramified outside a finite set of places, and then applies a profinite cohomology lemma. The conclusion is an arithmetic finiteness statement: the cokernel is not merely a torsion or divisible group but is built from finite layers in a compact inverse limit.","pith_inferences":["Editorial inference: the proof scheme should extend to any smooth projective surface over a number field for which the restriction kernel on $A^2$ is profinite after base change and $A^2[n]$ is Galois-isomorphic to the n-torsion of an abelian variety; Remark 2.4 already exercises one such case.","Editorial inference: the theorem establishes profiniteness, not finiteness; computing $S_n(\\Sigma_{\\bar K})$ for small $n$ on a concrete CM elliptic curve could reveal whether the Chow-group Selmer group is finite like an abelian-variety Selmer group or has infinite profinite components.","Editorial inference: the Selmer-group construction is not obviously limited to $E\\times E$; applying the same global-to-local map to higher Chow groups or motivic cohomology of codimension two would test whether the unramified-class mechanism is a general phenomenon for algebraic cycles.","Editorial inference: if a future strengthening replaced profiniteness of $\\Sigma_{\\bar K}$ by finiteness, the same proof would promote the conclusion from profinite to finite."],"forward_implications":["The same profiniteness conclusion holds for the Fermat quartic surface, since the analogue of Mildenhall's kernel-finiteness was proved there, as noted in Remark 2.4.","The quotient $\\operatorname{im}(j^*)[n](K)/A^2(\\mathscr{X})[n](K)$ embeds into $H^1(G,\\Sigma_{\\bar K}[n])$ and its image is contained in classes unramified outside a finite set of places, so the cokernel is governed by finitely many local conditions.","The construction associates to each $n$ an $n$-Selmer group for the restriction map on $A^2$, giving a Chow-group analogue of the Selmer group of an abelian variety.","Because the quotient is profinite, all of its arithmetic information is recoverable from its finite quotients, so the cokernel cannot contain an infinite divisible direct summand.","The theorem upgrades the known finiteness of the kernel of flat pullback to structural control of the cokernel on n-torsion Galois-fixed cycles."],"supporting_citations":[{"why":"Supplies Mildenhall's theorem that the kernel of the flat-pullback restriction map over the number field is finite, the departing point invoked for profiniteness of $\\Sigma_{\\bar K}$.","marker":"[Mil]"},{"why":"Roitman's theorem identifies $A^2(X)[n]$ with $\\operatorname{Alb}(X)[n]$ as Galois modules, enabling the transfer of abelian-variety Selmer techniques.","marker":"[R2]"},{"why":"Raskind's theorem gives injectivity of specialization on torsion algebraic cycles over local fields, used to force inertia-triviality of Selmer classes.","marker":"[Ras]"},{"why":"Silverman's lemma 4.3 of Chapter X gives profiniteness of unramified Galois cohomology $H^1(G,M;S)$ for a profinite module $M$.","marker":"[Sil]"},{"why":"Otsubo's analogue of kernel-finiteness for Fermat quartic surfaces is cited in Remark 2.4 to extend the main theorem to that surface.","marker":"[Ot]"}],"fun_headline_variants":["CM elliptic squares have profinite cycle cokernels","Profinite cokernels for CM elliptic curve squares","CM square cycles: Galois-fixed cokernel is profinite","CM squares: flat pullback cycles form inverse limits","Profinite bound on CM square cycle quotients"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof requires that the kernel of the restriction map after base change to the algebraic closure, the Galois module the paper calls $\\Sigma_{\\bar K}$, is profinite; the cited theorem invoked for this gives finiteness of the analogous kernel over the base field $K$, not over the algebraic closure.","fun_headline_variants_meta":{"raw":{"variants":["CM elliptic squares have profinite cycle cokernels","Profinite cokernels for CM elliptic curve squares","CM square cycles: Galois-fixed cokernel is profinite","CM squares: flat pullback cycles form inverse limits","Profinite bound on CM square cycle quotients"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000642,"raw_usage":{"total_tokens":2951,"prompt_tokens":943,"completion_tokens":2008,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":559,"completion_tokens_details":{"reasoning_tokens":1927}},"tokens_in":559,"tokens_out":2008,"duration_ms":15589,"temperature":1.0,"reasoning_tokens":1927,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:46:48.451523+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a concrete CM elliptic curve $E/K$, fix an integer $n$, and look for a class $\\eta\\in H^1(G,\\Sigma_{\\bar K}[n])$ whose image in every local group $H^1(G_v,A^2(\\mathscr{X}_{\\bar K_v}))[n]$ is zero but whose restriction to the inertia group $I_v$ is nonzero for infinitely many good primes $v\\nmid n$. The theorem predicts no such class exists, so exhibiting one would disprove the profiniteness conclusion.","supporting_citations":[],"review_version":1}