{"id":"1e4ed630-0444-4f22-85a1-41442548e291","arxiv_id":"2506.21210","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Over rings of integers in number fields, finite flat linearly reductive subgroups of SL2 are exactly fppf local forms of μ_n, with finitely many GL2-conjugacy classes and controlled quotient singularities.","lead":"The paper gives an arithmetic counterpart of Klein's classification: over the ring of integers of a number field, every finite flat linearly reductive subgroup scheme of SL2 is locally, in the fppf topology, a conjugate of the standard cyclic group scheme μ_n. It proves finiteness of the classification, describes the quotient singularities, and shows these singularities are of type A_{n-1} except on half of the primes, where they become type B_β(n).","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The split exact sequence (2)/(6)/(13) is not exact on the left: the swap matrix acts on H^1(G_m^2) by interchanging line-bundle factors, so H^1(G_m^2) fails to inject into H^1(N_GL2(μ_n)) whenever Pic(O_K) is nontrivial. This breaks the lower bound |Klein(n,O_K)| ≥ h_K.","rationale":"The paper's central classification is built on the rigidity theorem and the cohomological description of twisted forms, both of which appear sound. The reader identified the most load-bearing weakness: the asserted short exact sequence of pointed cohomology sets in (2)/(6)/(13) ignores the nontrivial action of the Weyl group on the line-bundle pair. I verified this directly: the normalizer contains a global swap matrix, so for any 1-cocycle (a_ij, b_ij) in G_m^2, conjugating by the constant 0-cochain w gives the swapped cocycle, forcing (L1,L2) and (L2,L1) to have the same image in H^1(N). Hence H^1(G_m^2) does not inject into H^1(N) when Pic is nontrivial. This invalidates the identification of the trivial fiber with Cl^2 and the count of h_K classes from the kernel of Cl^2 → Cl. The correct count of the trivial-fiber contribution is the number of orbits of the inversion action on Cl, namely (h_K + |Cl_K[2]|)/2. For K = Q(√-23), this gives |Klein(n,O_K)| = 2, while the paper claims at least 3, so Theorem 1.3(2)(b) is false. The error is localized: finiteness, independence of n, the description as twisted forms of μ_n, and the singleton criterion all appear repairable by replacing the erroneous exactness with the action-quotient statement. Therefore the reader's CONDITIONAL verdict should stand, requiring correction of the lower bound and the cohomological sequence. I see no further objection of comparable weight.","tokens_in":33636,"tokens_out":14474,"duration_ms":143742,"concrete_test":"Take K = Q(√-23), so Cl_K ≅ Z/3, h_K = 3, and H^1(O_K,Z/2) = 0 (since Cl^1_K has odd order). Let L be a non-trivial element of Cl_K of order 3; by Steinitz there is an isomorphism L ⊕ L^{-1} ≅ O^2. Form the subgroup scheme G ⊂ SL_{2,O_K} obtained by twisting the standard μ_n embedding by this line-bundle pair via the construction of Section 7.3/Remark 7.8. Check that G is not GL_2(O_K)-conjugate to the standard μ_n embedding (e.g., reduce modulo a split prime and verify no integral conjugating matrix exists, as in the argument of Proposition 8.1 or Example 8.3). Since H^1(O_K,Z/2) = 0, these two classes are all of Klein(n,O_K), so |Klein(n,O_K)| = 2 < 3 = h_K, settling the falsehood of Theorem 1.3(2)(b).","verdict_should_be":"UNCHANGED","load_bearing_attack":"For n ≥ 3, the normalizer split exact sequence is 1 → G_m^2 → N_GL2(μ_n) → Z/2 → 1, with the nontrivial element of Z/2 acting by the swap matrix w = [[0,1],[-1,0]], which conjugates diag(a,b) to diag(b,a). In nonabelian cohomology, the fiber of H^1(N) → H^1(Z/2) over the trivial class is H^1(G_m^2)/H^0(Z/2), not H^1(G_m^2): the global section w makes the cocycle (a_ij,1) cohomologous to (1,a_ij) in H^1(N). Thus, for any scheme with nontrivial Pic, the map H^1(G_m^2) → H^1(N) is not injective, so (2) and (6) are not short exact sequences of pointed sets. In the proof of Theorem 9.5(2), the fiber over the trivial class is identified with Cl^2, and its intersection with the kernel of H^1(N) → H^1(GL_2) is identified with pairs (L,L^{-1}), giving h_K distinct classes. After correctly quotienting by the swap action, (L,L^{-1}) and (L^{-1},L) are identified, so the number of classes contributed by the trivial fiber is (h_K + |Cl_K[2]|)/2, not h_K. A concrete counterexample to the stated lower bound: take K = Q(√-23), where Cl_K ≅ Z/3 and Cl^1_K ≅ Z/3, so H^1(O_K,Z/2) = 0. The corrected count gives |Klein(n,O_K)| = (3+1)/2 = 2, while h_K = 3, contradicting Theorem 1.3(2)(b) and Theorem 9.5(2). The finiteness and n-independence statements likely survive with the quotient correction, and the singleton criterion can still be proved without the false lower bound, but the asserted inequality is false as stated.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes an arithmetic analogue of Klein's classification. For a number field K, the authors study the set Klein(n,O_K) of finite flat linearly reductive subgroup schemes of SL_{2,O_K} of length n modulo GL_2-conjugacy. Their main theorem asserts that every such subgroup scheme is fppf locally conjugate to the standard embedding μ_n, that Klein(n,O_K) is finite and independent of n for n≥3, that its cardinality is at least the class number h_K, and that it is a singleton exactly when the ray class group Cl^1_K is trivial. The proof combines rigidity of finite flat group schemes over Dedekind schemes, a torsor-theoretic description of locally conjugate subgroups, and an analysis of the normalizer N_{GL_2}(μ_n) and its flat cohomology.","tokens_in":34111,"tokens_out":10434,"duration_ms":124526,"significance":"The result is significant if the counting statements are corrected: it gives a new arithmetic classification theorem in which the Klein set is governed by the class group and by unramified quadratic twists, and it yields concrete density statements for the associated families of rational double point singularities. The paper has genuine strengths: the rigidity theorem (Theorem 4.7), the cohomological dictionary (Theorem 5.2), the explicit examples of non-standard embeddings in Section 8, and the divisor-density results in Section 9 are all valuable contributions. However, the proof of the lower bound |Klein(n,O_K)| ≥ h_K relies on a left-exactness assertion that is false in general, so the main numerical claim is not established as stated.","major_comments":[{"comment":"The displayed 'short exact sequence of pointed cohomology sets' is not exact on the left. In the split extension 1 → G_m^2 → N_{GL_2}(μ_n) → Z/2 → 1, the nontrivial element of Z/2 acts on G_m^2 by interchanging the two diagonal entries, i.e., by the swap matrix w = [[0,1],[1,0]] (or its SL_2 analogue). Consequently H^1(S,G_m^2) need not inject into H^1(S,N_{GL_2}(μ_n)): a class (L,M) and the class (M,L), which are distinct in H^1(S,G_m^2) when Pic(S) is nontrivial, become cohomologous after the pushforward to the normalizer. The correct description of the fiber over the trivial class of H^1(S,Z/2) is a quotient of H^1(S,G_m^2) by this twisting action, not H^1(S,G_m^2) itself. This affects the literal exactness claims in Eqs. (2), (6), and (13).","section":"Section 7.2, Eq. (6); also Eqs. (2) and (13)"},{"comment":"The lower bound |Klein(n,O_K)| ≥ h_K is obtained by identifying the contribution of the trivial Z/2-fiber with Cl_K via the injectivity that fails. After taking the swap action into account, the contribution of pairs (L,L^{-1}) with L ∈ Cl_K is the number of orbits under L ↔ L^{-1}, namely (h_K + |Cl_K[2]|)/2, not h_K. For K = Q(√-23) one has Cl_K ≅ Z/3, so Cl_K[2] is trivial and Cl^1_K has odd order; hence H^1(O_K,Z/2) = 0 and the only contribution is from the trivial fiber. The corrected count gives (3+1)/2 = 2 classes, whereas h_K = 3. Thus the inequality asserted in Theorem 1.3(2)(b) and Theorem 9.5(2) is false as stated. Finiteness and n-independence may survive, but the proof of the singleton criterion in Theorem 9.5(3) also uses the false lower bound to conclude h_K = 1, so that part needs a different argument.","section":"Section 9.1, proof of Theorem 9.5(2); Theorem 1.3(2)(b)"},{"comment":"The proof of Theorem 4.7(3), which is the rigidity statement for conjugate closed subgroup schemes of GL_{n,S}, is deferred with 'we leave it to the reader.' This statement is load-bearing for Theorem 6.1, the reduction of all Klein(n,S) to twisted forms of μ_n, and hence for the whole paper. A full proof, or a precise reference to a result that contains it, should be supplied.","section":"Section 4.2, Theorem 4.7(3)"}],"minor_comments":[{"comment":"There is a typo: 'such such that' should read 'such that'.","section":"Theorem 1.4(2)"},{"comment":"In the chain H^1(O_K,G_m^2) → H^1(O_K,N_{GL_2}(μ_n)) → H^1(O_K,GL_{1,O_K}), the final displayed target should be GL_{2,O_K}, not GL_{1,O_K}.","section":"Proof of Theorem 9.5(2)"},{"comment":"The matrix (5), namely [[0,1],[-1,0]], has square -I and therefore does not give a group-theoretic splitting of the exact sequence for the SL_2 normalizer; for the GL_2 normalizer one can use [[0,1],[1,0]], which has order 2. The statement in the introduction that both sequences in (1) are split should be adjusted accordingly.","section":"Proposition 7.1 and Eq. (5)"},{"comment":"The treatment of the infinite places is summarized with 'we leave this to the reader.' Since this material is part of Theorem 1.4, a brief argument or a reference for the three listed cases would be desirable.","section":"Section 9.3"}],"recommendation":"major_revision","confidential_remarks":"The main classification thesis is likely correct and the paper is well within the journal's scope, but the lower bound claim is a load-bearing numerical assertion that appears to be false as stated. I would encourage the authors to correct the quotient by the swap action, revise the statements using the corrected count, and re-evaluate the singleton criterion. This is fixable within the scope of the manuscript, which is why I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, what you should know: this is a real step beyond [LS25]. The authors move the classification of finite linearly reductive subgroup schemes of SL_2 from fields to rings of integers, show that over O_K every such scheme is fppf-locally conjugate to the standard μ_n, and give a cohomological classification. Finiteness and density results are new, and the paper is carefully organized.\n\nThe proof architecture is sound in the main. The rigidity theorem over Dedekind schemes, the reduction to μ_n using characteristic 2, the normalizer computation, and the Steinitz-based analysis of Zariski-local forms all work. Section 8's explicit non-standard embedding and invariant ring computation is careful and informative.\n\nThe soft spot is real and affects one theorem. The paper claims (6)/(13) is a short exact sequence of pointed sets 1→H^1(G_m^2)→H^1(N_GL2(μ_n))→H^1(Z/2)→1. It is not exact on the left. The splitting matrix w=[[0,1],[-1,0]] acts on H^1(G_m^2) by swapping the two line-bundle factors, and in nonabelian cohomology this identifies (L1,L2) with (L2,L1) in the fiber over the trivial class. So H^1(G_m^2) does not inject when Pic(O_K) is nontrivial.\n\nThis undermines the proof of Theorem 9.5(2). The lower bound |Klein(n,O_K)| ≥ h_K comes from identifying a fiber with Cl_K via pairs (L,L^{-1}); after quotienting by the swap action, the contribution is (h_K+|Cl_K[2]|)/2. For K=Q(√-23), h_K=3 but the corrected count is 2, so the inequality is false as stated.\n\nThe rest survives. Finiteness, n-independence, the singleton criterion, and the density statements do not rely on that lower bound; the singleton proof can be reworked using h_K=1 once the correct count is in place. This is a conditionally acceptable paper: the structure is right, the quantitative statement needs repair.\n\nI would send it to a serious referee. The flaw is subtle and the authors are well positioned to fix it. Readers interested in arithmetic quotient singularities should use the main classification and examples, but should not quote the lower bound from the current version.\n\nRecommendation: accept for peer review, expect a substantive revision.","headline":"A genuine extension of Klein's classification to rings of integers, but the quantitative lower bound rests on a false left-exactness in a split nonabelian cohomology sequence.","tokens_in":34600,"tokens_out":4786,"would_cite":false,"duration_ms":45137,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11E57","11E72","14L15","14L30","14J17"],"pacs":[],"model":"deepseek-v4-flash","headline":"Over the ring of integers of a number field, every finite flat linearly reductive subgroup scheme of SL2 is, up to a flat cover, conjugate to the standard embedding of the n-th roots of unity.","keywords":["finite group schemes","linear algebraic groups","arithmetic base schemes","quotient singularities","rational double points","flat cohomology","class group","quadratic twists"],"falsifier":"For a number field with class number greater than 1, such as $K = \\mathbb{Q}(\\sqrt{-5})$, compute the flat cohomology of the normalizer and check whether the two pairs $([\\mathfrak{p}],[\\mathcal{O}_K])$ and $([\\mathcal{O}_K],[\\mathfrak{p}])$ in $\\mathrm{Cl}_K^2$ become equal in $H^1_{\\mathrm{fppf}}(\\mathcal{O}_K,N_{\\mathrm{GL}_2}(\\mu_n))$; equality would reduce the size of the conjugacy set below the class number and refute the lower bound.","tokens_in":33436,"feed_emoji":"🔢","tokens_out":8683,"duration_ms":80478,"temperature":0.7,"pith_summary":"The paper proves that over the ring of integers of a number field, every finite flat linearly reductive subgroup scheme of SL2 of length n is, after a flat cover, conjugate to the standard diagonal embedding of the n-th roots of unity. It then turns this geometric statement into a cohomological count: the set of such subgroup schemes up to GL2-conjugacy is finite, has size independent of n for n at least 3, and is at least as large as the class number of the field. The associated quotient singularities form arithmetic families of rational double points, whose fiber types split evenly between type $A_{n-1}$ and type $B_{\\beta(n)}$ according to an unramified quadratic twist. This gives an arithmetic counterpart to the classical complex classification, with only one infinite family surviving because residue characteristic 2 forces the special fiber to be cyclic.","feed_headline":"All finite SL2 subgroup schemes over number rings are cyclic twists","feed_subtitle":"Every such group is locally the standard μ_n; the count ties to class groups.","key_machinery":"The argument is carried by a rigidity theorem for finite flat linearly reductive group schemes over an excellent Dedekind scheme: if two such subgroup schemes of $\\mathrm{GL}_{n,S}$ are conjugate at one fiber, then they are conjugate after an fppf cover of the base. Applied to a point of residue characteristic 2, where the classification of linearly reductive subgroup schemes of $\\mathrm{SL}_2$ over an algebraically closed field contains only the cyclic $\\mu_n$, rigidity forces every $G$ to be an fppf twisted form of the standard embedding. The twisted forms are then classified by the kernel of $H^1_{\\mathrm{fppf}}(\\mathcal{O}_K,N_{\\mathrm{GL}_2}(\\mu_n)) \\to H^1_{\\mathrm{fppf}}(\\mathcal{O}_K,\\mathrm{GL}_2)$, with the normalizer sitting in a split exact sequence $1 \\to \\mathbb{G}_m^2 \\to N \\to \\mathbb{Z}/2\\mathbb{Z} \\to 1$. Zariski-local forms correspond to pairs of line bundles on $\\mathrm{Spec}\\,\\mathcal{O}_K$, controlled by a classical structure theorem for Dedekind domains, while the $\\mathbb{Z}/2\\mathbb{Z}$ part corresponds to unramified quadratic extensions, controlled by the ray class group of modulus 1.","core_discovery":"The central claim is that a finite flat linearly reductive subgroup scheme $G \\subset \\mathrm{SL}_{2,\\mathcal{O}_K}$ of length $n$ is an fppf locally trivial conjugate of the standard embedding $\\mu_n \\subset \\mathrm{SL}_{2,\\mathcal{O}_K}$. In particular, the set of such subgroup schemes up to $\\mathrm{GL}_2$-conjugacy is finite for every $n$; for $n \\ge 3$ its cardinality is independent of $n$ and at least the class number $h_K$, and it is a singleton exactly when the ray class group of modulus 1 is trivial. When $G$ is not the standard $\\mu_n$, it is the quadratic twist of $\\mu_n$ with respect to inversion over an unramified quadratic extension of $K$; fibers over primes split or inert in that extension give rational double points of type $A_{n-1}$ or $B_{\\beta(n)}$, respectively, each set of primes having Dirichlet density $1/2$.","pith_inferences":["The count $|\\mathrm{Klein}(n,\\mathcal{O}_K)|$ likely admits a formula in terms of the class numbers of $K$ and of each unramified quadratic extension $L/K$, with the split or inert distinction entering through a quotient by the inversion action; the paper only sketches this formula.","The rigidity mechanism suggests a broader principle: over any base with a fiber of characteristic 2, finite linearly reductive subgroups of $\\mathrm{SL}_2$ collapse to twisted $\\mu_n$, so similar finiteness should hold for more general arithmetic bases than rings of integers.","The worked examples show that the invariant ring can need more than three generators even when a degree bound holds; a natural testable extension is to determine the minimal number of generators for the quadratic-twist families.","The predicted density $1/2$ for inert versus split primes could be checked computationally for a fixed number field and a fixed non-standard $G$, by counting primes below a large bound."],"forward_implications":["The set for length 2 is a singleton, so the only length-2 finite flat linearly reductive subgroup scheme is the standard $\\mu_2$ up to conjugacy.","For $n \\ge 3$, the size of the conjugacy set is independent of $n$ and is at least the class number, so it is unbounded as the number field varies.","The conjugacy set is a singleton exactly when the ray class group of modulus 1 is trivial; this includes the rational numbers and imaginary quadratic fields of class number 1.","A non-standard subgroup scheme yields an arithmetic family of rational double points over $\\mathcal{O}_K$ in which half the closed fibers are of type $A_{n-1}$ and half of type $B_{\\beta(n)}$, with both sets of primes infinite.","Removing a finite set of primes while keeping at least one prime above 2 preserves finiteness and $n$-independence; removing all primes above 2 brings additional group-scheme types such as binary dihedral ones into play."],"supporting_citations":[{"why":"Supplies the definition and permanence properties of linearly reductive group schemes, including the well-split covers used in the rigidity proof.","marker":"[AOV08]"},{"why":"Provides the representability of normalizer schemes used to turn conjugacy into a cohomology classification.","marker":"[Co14]"},{"why":"Supplies the classification of finite linearly reductive subgroup schemes of SL2 over algebraically closed fields, from which only $\\mu_n$ survives in characteristic 2.","marker":"[Ha15]"},{"why":"Gives the classical complex classification of finite subgroups of SL2(C) that this paper arithmetizes.","marker":"[Kl84]"},{"why":"Extends the twisted-form classification to nonclosed fields and identifies the type $B_m$ quotient singularities used for fibers over split primes.","marker":"[LS25]"},{"why":"Supplies the etale-cohomology and torsor representability facts behind the flat cohomology description of twisted forms.","marker":"[Mi80]"},{"why":"Provides the class field theory facts: class group, ray class group of modulus 1, and Dirichlet density results used for finiteness and density statements.","marker":"[Ne99]"},{"why":"The Dedekind-domain module structure theorem controls when pairs of line bundles sum to a free rank-2 module, governing the Zariski-local forms.","marker":"[St12]"}],"fun_headline_variants":["Every finite SL2 subgroup scheme over a number ring is a cyclic twist","SL2 over number rings: finite subgroups are cyclic twists","Klein's classification over number rings: only cyclic twists","Number rings tame SL2: finite subgroups are μ_n twists"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The counting argument assumes that the two line-bundle parameters that describe a class in the normalizer cohomology are not identified with each other by the swap symmetry of the normalizer; if they are identified, the lower bound would become a smaller quotient of the class group.","fun_headline_variants_meta":{"raw":{"variants":["Every finite SL2 subgroup scheme over a number ring is a cyclic twist","SL2 over number rings: finite subgroups are cyclic twists","Klein's classification over number rings: only cyclic twists","Number rings tame SL2: finite subgroups are μ_n twists"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00075,"raw_usage":{"total_tokens":3272,"prompt_tokens":809,"completion_tokens":2463,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":425,"completion_tokens_details":{"reasoning_tokens":2392}},"tokens_in":425,"tokens_out":2463,"duration_ms":18775,"temperature":1.0,"reasoning_tokens":2392,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T22:49:46.856707+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a number field with class number greater than 1, such as $K = \\mathbb{Q}(\\sqrt{-5})$, compute the flat cohomology of the normalizer and check whether the two pairs $([\\mathfrak{p}],[\\mathcal{O}_K])$ and $([\\mathcal{O}_K],[\\mathfrak{p}])$ in $\\mathrm{Cl}_K^2$ become equal in $H^1_{\\mathrm{fppf}}(\\mathcal{O}_K,N_{\\mathrm{GL}_2}(\\mu_n))$; equality would reduce the size of the conjugacy set below the class number and refute the lower bound.","supporting_citations":[],"review_version":1}