{"id":"df817bf2-f507-49ba-80a0-cc4671cb5bd6","arxiv_id":"2504.21272","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For abelian varieties over global fields with multiplication by an order, the Selmer group is the intersection of two maximal isotropic subspaces in an orthogonal, symplectic, unitary, or split unitary quadratic space.","lead":"This paper attaches quadratic forms to the cohomology spaces containing Selmer groups of abelian varieties with extra multiplication, extending the Poonen-Rains random subspace model. The result gives new predictions for the sizes of Selmer groups in real- and complex-multiplication families.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 3.11(1) relies on a Galois-equivariance claim that is false, so the p=2 orthogonal Selmer theorem is not proven as written.","rationale":"The reader identified the p=2 quadratic conditions as the weakest assumption; this stress test finds a more specific and load-bearing problem: the proof that those conditions hold in the key orthogonal case is flawed. The explicit Galois-equivariance error in Proposition 3.11(1) is concrete and checkable, and it affects Theorem 3.12, Proposition 4.7(2), and hence Theorem 5.7 for p=2. The conclusion may well be true, and the proof may be repairable by choosing a different quadratic refinement (as the elliptic curve example shows), so the appropriate verdict is CONDITIONAL rather than REJECT. The concern is not with the theorem's plausibility or the framework, but with the validity of a central proof step for a nontrivial class of abelian varieties.","tokens_in":48061,"tokens_out":47662,"duration_ms":510630,"concrete_test":"On E: y^2=x^3+x+1 over Q, take µ=λ and choose a basis with θ(e1,e2)=1⊗−1. Since Gal(E[2])≅S3, use the transvection σ(e1)=e1+e2, σ(e2)=e2. Then q(σ(e2))=q(e1+e2)=1 while q(e2)=0 in F2, directly contradicting the claimed Galois equivariance in Proposition 3.11(1). To check the theorem's conclusion in this same case, verify c_λ=0 via the rational symmetric line bundle O(−O), which exists, confirming the proof needs repair rather than the claim being false.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"In Proposition 3.11(1), for K=K0, [K:Q]=dim A, and p=2, the proof chooses a k-basis e1,e2 of A[p] with θ(e1,e2)=1⊗−1 and defines q(ae1+be2)=ab⊗−1. It then claims q is Galois equivariant because θ is Galois equivariant. This is incorrect: unless the basis is Galois-stable, the coordinates of σx are not simply σ(a), σ(b). Over k=F2, the transvection σ(e1)=e1+e2, σ(e2)=e2 lies in Sp2(F2)=S3; for x=e2 (a=0,b=1), σx=e1+e2, so q(σx)=1 while q(x)=0. Thus q is not a morphism over F. This invalidates the proof of c_µ=0 in the orthogonal case. That proposition is the sole input to Theorem 3.12 in this case, and Theorem 3.12 is used in Proposition 4.7(2) and Corollary 1.4 to establish hypothesis (L) for p=2. Consequently Theorem 5.7 for p=2 orthogonal spaces is not established for RM abelian varieties as written, even if the statement may be true (for elliptic curves a different q works).","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper extends the Poonen–Rains quadratic-space framework from elliptic curves to abelian varieties carrying an order O in a number field inside their endomorphism algebra, for a symmetric isogeny λ whose Rosati involution stabilizes O. It constructs metabolic orthogonal, symplectic, unitary, or split-unitary structures on the adelic cohomology group H^1(A_F, A[p]), proves under certain p=2 hypotheses that the Selmer group is the intersection of two maximal isotropic subspaces, and derives consequences for the structure of Shafarevich–Tate groups. Section 2 develops a self-contained combinatorial and topological theory of quadratic spaces over finite fields, including new split-unitary distributions and a Rogers–Ramanujan identity. The main arithmetic results are Theorems 3.12, 5.7, and 6.2, with conditional hypotheses needed in the p=2 orthogonal and symplectic cases.","tokens_in":48300,"tokens_out":15474,"duration_ms":167950,"significance":"If the main theorems are correct, this is a substantial extension of the Poonen–Rains Selmer-group model to abelian varieties with real and complex multiplication, and it gives new conjectural distributions (Conjecture 1.6) for Selmer ranks in RM families. The paper also contributes interesting structural results for Shafarevich–Tate groups, and Section 2 contains useful and apparently correct combinatorial material, including explicit formulas for split-unitary maximal-isotropic intersections. The proof architecture is coherent: local duality gives local quadratic structures, the restricted product gives the adelic space, and the Cassels–Tate argument gives the Shafarevich–Tate structure. However, one load-bearing proof step in the p=2 orthogonal case is invalid as written, and another step in the proof of Theorem 3.12 is incomplete. These issues are local and appear repairable, but they must be fixed before the stated theorems can be accepted.","major_comments":[{"comment":"The proof that the quadratic map q is Galois equivariant is incorrect. After choosing a basis e1, e2 with θ(e1, e2) = 1⊗−1, the paper defines q(ae1+be2) = ab⊗−1 and asserts that q is Galois equivariant because θ is. This does not follow: for a Galois element σ with σ(e1)=e1+e2 and σ(e2)=e2, which is a transvection in Sp_2(F_2), one has q(σe2)=q(e1+e2)=1 while σ(q(e2))=0. The equality q(σx)=σ(q(x)) fails whenever the chosen basis is not Galois-stable, and the coordinates of σx are not simply σ(a), σ(b). This claim is the sole input to Theorem 3.12 in the case †=1, and through Theorem 3.12 it is used in Proposition 4.7(2), Corollary 1.4, and Theorem 5.7 for p=2 orthogonal spaces. The proposition is likely salvageable: the quadratic form q(v)=1 for every nonzero v in A[p] has Arf invariant one and is invariant under all of Sp_2(F_2), so it is a valid Galois-equivariant refinement of the given alternating form. The proof as written, however, does not establish the claim and must be repaired.","section":"§3.3, Proposition 3.11(1)"},{"comment":"In the proof of Theorem 3.12, after the decomposition 2O = p_1⋯p_r q_1 q_1^†⋯q_s q_s^†, the paper states that e^µ_qj + e^µ_qj† has the quadratic refinement (x,y) ↦ e^µ_qj(x,y). This is not a valid definition of a quadratic refinement: a quadratic refinement is a function of one variable, not of two variables, and the identity q(x+y)=q(x)q(y)β(x,y) is not verified for the proposed map. This step is used to prove c_µ=0 for the split prime ideals above 2 in the case †≠1, and therefore it is load-bearing for Theorem 3.12. The intended construction may be standard, but the present text does not supply the needed argument; please replace this sentence with a complete verification or a precise reference.","section":"§3.3, Theorem 3.12 (split primes)"}],"minor_comments":[{"comment":"The word 'Primiary' in the Mathematics Subject Classification line should be 'Primary'.","section":"§1, MSC line"},{"comment":"The displayed equality in the proof of Proposition 4.7(1) is garbled: the text appears to claim e^λ_p = e^{2λ}|_{A[p]×A[p]} = e^λ_2|_{A[p]×A[p]}, which is not meaningful as printed. Please restate the precise intended relationship between e^λ_p and e^λ_2.","section":"§4.3, Proposition 4.7(1)"},{"comment":"The word 'Theoreom' should be 'Theorem' in the sentence citing Theorem 2.13.","section":"§2.5, paragraph before Proposition 2.22"},{"comment":"The companion paper [59] is cited for the distribution of Selmer ranks in CM twist families; since that paper is not yet published, please state explicitly which results from [59] are assumed and which are merely anticipated.","section":"§1.3.2"}],"recommendation":"major_revision","confidential_remarks":"The main risk is the p=2 orthogonal case. The transvection example in Proposition 3.11(1) shows that the printed proof is wrong, but the Arf-invariant-one quadratic form appears to give a quick fix, so I do not regard this as fatal. The split-prime step in Theorem 3.12 also needs a careful written argument; if that step fails, the p=2 part of the main theorem in the †≠1 case is also at risk. The paper has substantial merit and the overall framework is convincing, but the revised version should contain a rigorous proof of these two points."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nQuick take: this is a real extension of Poonen-Rains to RM and CM abelian varieties, and the first half is strong. But the p=2 orthogonal case has a load-bearing proof gap that needs fixing before the main theorem can be trusted.\n\nWhat is genuinely new: the construction of symplectic, unitary, and split-unitary quadratic structures on H^1(A_F,A[p]) and the Selmer intersection theorem in those cases, plus the Cassels-Tate structure theorem. The combinatorial treatment of split-unitary spaces, including the Rogers-Ramanujan identity, is careful and goes beyond [50]. The conditional setup (hypotheses (L) and quadratic-everywhere) is explicit, and the reader can see where the hypotheses enter.\n\nThe problem is Proposition 3.11(1). The proof defines a quadratic refinement q on A[p] by choosing a k-basis e1,e2 with θ(e1,e2)=1⊗−1 and setting q(ae1+be2)=ab⊗−1. It then asserts q is Galois-equivariant because θ is. That assertion is false: Galois need not preserve the basis. For example, over k=F_2, a transvection σ(e1)=e1, σ(e2)=e1+e2 gives q(σ(e2))=q(e1+e2)=1 but σ(q(e2))=0. So q is not an F-morphism. This invalidates the proof of c_µ=0 in the †=1, [K:Q]=dim A case. That result feeds Theorem 3.12, then Proposition 4.7(2) (hypothesis (L)), and finally Theorem 5.7 for p=2 orthogonal RM varieties. The statement may still be true—for elliptic curves there is a different construction—but the present proof doesn't cover it.\n\nWhere this lands: the gap is important, but it is confined to one case. The p odd, symplectic, unitary, and split-unitary parts stand, and the combinatorial sections are fine. The Shafarevich-Tate results for p≠2 are fine; the p=2 orthogonal part inherits the same gap.\n\nI'd send this to a serious referee. The right result is major revision: fix Proposition 3.11(1) (or restrict the statement and make the conditions explicit), and then the paper would be a solid contribution. If you work on Selmer distributions, it's worth reading, but don't quote the p=2 orthogonal theorem as it stands.\n\nRecommendation: peer review, leaning major revision.","headline":"Genuine extension of Poonen-Rains to RM/CM abelian varieties, but the p=2 orthogonal case has a false Galois-equivariance claim in Proposition 3.11(1) that leaves Theorem 5.7 unproven in that case.","tokens_in":48842,"tokens_out":5449,"would_cite":true,"duration_ms":53135,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11G10","11E04","11G05","14K15","14K22"],"pacs":[],"model":"deepseek-v4-flash","headline":"For abelian varieties with multiplication, the adelic p-torsion cohomology carries a metabolic orthogonal, symplectic, or unitary quadratic structure, and the Selmer group is the intersection of two maximal isotropic subspaces.","keywords":["Selmer groups","quadratic forms","abelian varieties","real multiplication","complex multiplication","Weil pairing","Shafarevich-Tate groups","isotropic subspaces"],"falsifier":"Take an abelian variety A over a global field F with real multiplication by an order O of totally real degree equal to dim A, with p=2 prime to the discriminant of O, and compute dim Sel_2(A) and dim(L∩W) in $H^{1}$(A_F,A[2]). The theorem predicts these are equal; any example where they differ breaks the central isomorphism.","tokens_in":47836,"feed_emoji":"🧮","tokens_out":8463,"duration_ms":86234,"temperature":0.7,"pith_summary":"This paper establishes the arithmetic quadratic-space picture of Selmer groups for abelian varieties that carry extra endomorphisms, not just for elliptic curves and hyperelliptic Jacobians. For an abelian variety A over a global field F with a symmetric isogeny λ and an order O in its endomorphism algebra stable under the Rosati involution, the adelic cohomology $H^{1}$(A_F,A[p]) is shown to be a metabolic quadratic space of orthogonal, symplectic, or unitary type. The p-Selmer group, modulo the p-primary Tate-Shafarevich obstruction, is then the intersection of two maximal isotropic subspaces: the Kummer image and the image of global cohomology. If this holds, random maximal isotropic subspace models for Selmer ranks extend to families with real or complex multiplication, and the primary parts of Shafarevich-Tate groups acquire a self-dual M⊕M shape.","feed_headline":"Multiplication puts a quadratic structure on Selmer groups","feed_subtitle":"With real or complex multiplication, p-Selmer groups are intersections of maximal isotropic subspaces.","key_machinery":"The load-bearing object is the form parameter Λ of the finite residue field k together with the pairings that it makes quadratic. The Rosati-involution-compatible multiplication by O on the p-adic Tate module gives an extended Weil pairing Θ^λ_{p_0} valued in O_{p_0}(1), whose reduction modulo p produces the local pairing h_v on $H^{1}$(F_v,A[p]). Quadratic refinements, supplied by $\\theta$ groups and, in the p=2 orthogonal case, by line bundles L(a) with φ_{L(a)}=λa, turn these even hermitian forms into genuine quadratic spaces, and trace compatibility lifts the F_p-quadratic structure to k. Local Tate duality makes each local Kummer image self-orthogonal, the restricted product over all places is metabolic, and global Poitou-Tate duality makes the image of global cohomology self-orthogonal, so Sel_p(A)/$X^{1}$(F,A[p]) becomes L∩W.","core_discovery":"The central theorem states that, under suitable hypotheses, the adelic space $H^{1}$(A_F,A[p]) is a metabolic orthogonal, symplectic, or unitary k-space depending on whether the Rosati involution is trivial on K, p is ramified, or p is inert in K/K_0, and that Sel_p(A)/$X^{1}$(F,A[p]) ≅ L∩W, where L is the product of local Kummer images and W is the image of $H^{1}$(F,A[p]). In the split case p_0 splits in K/K_0, the same conclusion holds for $H^{1}$(A_F,A[p_0]) with a split unitary structure. For p=2, the orthogonal case requires a rational symmetric line bundle inducing λ, and the symplectic case requires the pair (A,λ) to be quadratic at every place; the paper proves these conditions in several arithmetic situations, including when [K:Q] equals dim A or 2 dim A. The corresponding structural result for Shafarevich-Tate groups is that X(A)/div[p^∞] is isomorphic to M⊕M for a finite O_0-module M in the covered cases.","pith_inferences":["Because both L and W are O-stable, a faithful random model for these families should sample only maximal isotropic subspaces that are O-submodules; for O strictly larger than Z this sample space is much smaller than the full Grassmannian, so observed Selmer-rank distributions should deviate from the plain orthogonal distribution as [K:Q] grows.","A computable test of the theory is to work out the local obstruction c_{λ,v} at bad-reduction places for p=2 in the symplectic case; the theorem predicts vanishing everywhere, so any nonzero value would force a different global mechanism.","In the split unitary case, the Rogers-Ramanujan-type identity suggests that Selmer ranks in twist families of CM abelian varieties are governed by a random-subspace model rather than a classical random-matrix model, which could be probed by comparing the generating functions for explicit families."],"forward_implications":["For every covered triple (A,λ,O), the rank of Sel_p(A) is the dimension of L∩W, and when X^1(F,A[p])=0 the Selmer group itself is described this way.","Families of abelian varieties with real multiplication by a fixed order O admit a conjectural distribution of dim Sel_p given by the orthogonal distribution D^Ort_q, while CM families in the ramified and inert cases correspond to symplectic and unitary distributions.","In the split unitary case, maximal isotropic subspaces are parameterized by ordinary subspaces of a k_0-space, so the relevant limit distributions are the co-rank distributions of random matrices over k_0, connected to Rogers-Ramanujan-type identities.","The p-primary component X(A)/div[p^∞] is a self-dual module M⊕M in the stated conditions, giving parity constraints and even dimensions over O_0/p_0.","For p=2 with [K:Q]=dim A and odd discriminant, the results imply the existence of rational theta characteristics in the branched-covering examples, and finiteness of X(A) forces X(A) to have the form M⊕M."],"supporting_citations":[{"why":"Supplies the metabolic-quadratic-space framework and the Selmer-as-intersection theorem that this paper generalizes.","marker":"[50]"},{"why":"Constructs theta-group quadratic refinements and the self-cup-product obstruction used for quadratic maps and for the p=2 orthogonal case.","marker":"[49]"},{"why":"Provides the Cassels-Tate pairing properties used for local self-orthogonality and for the Shafarevich-Tate structure.","marker":"[51]"},{"why":"Supplies Tate duality over local and global fields, making Kummer images and the global image self-dual.","marker":"[64]"},{"why":"Gives the generalized Cassels-Tate pairing result used for the p-primary M⊕M structure.","marker":"[25]"},{"why":"Produces the line bundles L(a) with φ_{L(a)}=λa needed in the p=2 orthogonal construction.","marker":"[47]"},{"why":"Provides the hermitian-category and quadratic-space background, including form parameters and metabolic spaces.","marker":"[55]"},{"why":"Classifies nondegenerate quadratic spaces by form parameters, fixing the orthogonal, symplectic, and unitary types.","marker":"[5]"}],"fun_headline_variants":["Multiplication turns Selmer groups into isotropic intersections","With multiplication, p-Selmer groups are intersections of maximal isotropic subspaces","Multiplication makes Selmer groups intersections of maximal isotropic subspaces","Selmer groups as isotropic intersections under multiplication","Multiplication yields Selmer groups as isotropic intersections"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The construction needs quadratic refinements of λ at every local place, which for p=2 is imposed as a rational symmetric line bundle in the orthogonal case and as vanishing of all local obstructions c_{λ,v} in the symplectic case; if these conditions fail, the adelic quadratic structure and the Selmer intersection statement are not established.","fun_headline_variants_meta":{"raw":{"variants":["Multiplication turns Selmer groups into isotropic intersections","With multiplication, p-Selmer groups are intersections of maximal isotropic subspaces","Multiplication makes Selmer groups intersections of maximal isotropic subspaces","Selmer groups as isotropic intersections under multiplication","Multiplication yields Selmer groups as isotropic intersections"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001035,"raw_usage":{"total_tokens":4341,"prompt_tokens":910,"completion_tokens":3431,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":526,"completion_tokens_details":{"reasoning_tokens":3355}},"tokens_in":526,"tokens_out":3431,"duration_ms":24483,"temperature":1.0,"reasoning_tokens":3355,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T05:08:09.253624+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take an abelian variety A over a global field F with real multiplication by an order O of totally real degree equal to dim A, with p=2 prime to the discriminant of O, and compute dim Sel_2(A) and dim(L∩W) in $H^{1}$(A_F,A[2]). The theorem predicts these are equal; any example where they differ breaks the central isomorphism.","supporting_citations":[{"cited_title":"Poonen and E","cited_arxiv_id":null,"evidence_quote":"Supplies the metabolic-quadratic-space framework and the Selmer-as-intersection theorem that this paper generalizes."},{"cited_title":"Poonen and E","cited_arxiv_id":null,"evidence_quote":"Constructs theta-group quadratic refinements and the self-cup-product obstruction used for quadratic maps and for the p=2 orthogonal case."},{"cited_title":"Poonen and M","cited_arxiv_id":null,"evidence_quote":"Provides the Cassels-Tate pairing properties used for local self-orthogonality and for the Shafarevich-Tate structure."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies Tate duality over local and global fields, making Kummer images and the global image self-dual."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the generalized Cassels-Tate pairing result used for the p-primary M⊕M structure."},{"cited_title":"Polishchuk","cited_arxiv_id":null,"evidence_quote":"Produces the line bundles L(a) with φ_{L(a)}=λa needed in the p=2 orthogonal construction."},{"cited_title":"Scharlau.Quadratic and Hermitian forms, volume 270 ofGrundlehren der Mathematischen Wis- senschaften","cited_arxiv_id":null,"evidence_quote":"Provides the hermitian-category and quadratic-space background, including form parameters and metabolic spaces."},{"cited_title":"Bak.K-theory of forms, volume 98 ofAnnals of Mathematics Studies","cited_arxiv_id":null,"evidence_quote":"Classifies nondegenerate quadratic spaces by form parameters, fixing the orthogonal, symplectic, and unitary types."}],"review_version":1}