{"id":"81cbfa27-13e0-4811-8595-c6aeb8c63c56","arxiv_id":"2608.01595","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For every affine algebraic group G over an algebraically closed field, every finite abelian subgroup A with |A| not divisible by the characteristic lies in some maximal torus up to index dividing the Grothendieck torsion index t(G).","lead":"This mathematics paper proves new structural facts about the finite abelian subgroups that can sit inside a large class of algebraic groups, and uses them to settle several open questions about when certain geometric spaces with group actions, called torsors, split over special fields. The results sharpen a decades-old theorem of Borel and give a positive answer to a question of Totaro and Wang.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 5.2 relies on an unverified reduction for [SS70, Thm 5.16] from reductive to semisimple; if the center-quotient lifting fails, Theorem 1.1 collapses.","rationale":"The reader identified Proposition 5.2 and the [SS70, Theorem 5.16] reduction as the weakest assumption, and I agree that this is the most delicate point in the proof of Theorem 1.1. However, I do not see a concrete failure: the center-quotient lifting appears to be standard, and the paper points to an alternative reference [CGP25, Proposition 5.10]. The concern is real but localized and likely repairable, so it does not change the accept verdict. I checked the rest of the main chain: Proposition 3.1's properness and valuative-criterion arguments are sound; the use of Hilbert 90 for tori is justified because k is algebraically closed and every k-torus is split; Lemma 4.3's index/depth equality is coherent; and the E8 computation in Section 7 is supported by explicit trace tables and code arguments. The only aspect that would benefit from a referee request is a precise statement of the exact version of [SS70, Theorem 5.16] being used and a written verification of the center-quotient reduction.","tokens_in":47577,"tokens_out":42251,"duration_ms":375378,"concrete_test":"Verify the original statement of [SS70, Theorem 5.16] and test the center-quotient reduction: for H = C_G(A1)^0 and Z = Z(H), let \\bar T be an A2-stable maximal torus of H/Z and check that \\pi^{-1}(\\bar T) is an A2-stable maximal torus of H. If [SS70, 5.16] treats only a single automorphism, confirm that iterating over the abelian group A2 preserves the reductivity of the fixed-point group at each step and that the resulting torus is maximal in H. If either check fails, Proposition 5.2 needs an additional argument before Theorem 1.1 can be regarded as fully proved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The decisive step in the proof of Theorem 1.1 is Proposition 5.2, where [SS70, Theorem 5.16] is applied to the reductive group H = C_G(A1)^0, although the cited theorem is stated for semisimple groups. The footnote proposes to reduce to the semisimple case by modding out by Z(H), asserting that an A2-stable maximal torus of H/Z lifts to one of H. This reduction is plausible: Z(H) is a torus, H/Z is semisimple, and the preimage of a maximal torus of H/Z is a torus of H. However, the paper gives no proof of the lifting, and it does not state whether [SS70, Theorem 5.16] covers a finite abelian group of automorphisms A2 or only a single semisimple automorphism. If only the latter, one must argue by induction over the elements of A2, and each step requires that the fixed-point group of a semisimple automorphism remains reductive and that an invariant maximal torus of that fixed group is maximal in H; this is not automatic and is not discussed. No other step in the chain to Theorem 1.1 appears to carry a comparable hidden assumption: Proposition 3.1, Lemma 4.3, and Corollary 5.3 are otherwise self-contained after standard references, and the E8 depth computations in Section 7 are explicit and case-checked.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies finite abelian k-subgroups A of affine algebraic groups over an algebraically closed field k, with char(k) not dividing |A|. Its main theorem (Theorem 1.1) asserts that for every such A there exists a maximal torus T of G such that the index [A : A∩T] divides the Grothendieck torsion index t(G). The proof introduces the depth invariant t3(G), shows t2(G)=t3(G) under mild characteristic assumptions, and proves a structural result (Proposition 5.2) saying that products of toral subgroups of pairwise coprime orders are toral. The applications are: a positive answer to Totaro's question for torsors over iterated Laurent series fields (Theorem 1.5), a variant of Tits' optimistic hypothesis for E8 with t2(E8)=60 (Theorem 1.3), and obstructions to splitting torsors by genus-one curves (Theorem 1.8). Section 7 contains detailed computations of depths of maximal finite abelian subgroups of E8, including a code-theoretic construction via Wood's C16 code and an explicit trace computation for the μ4^3×μ2^2 case.","tokens_in":47860,"tokens_out":23845,"duration_ms":216399,"significance":"If Theorem 1.1 is correct, it answers a question of Totaro and Wang and provides a uniform bound, depending only on G, on how far a finite abelian prime-to-p subgroup is from being toral. The applications are substantial, and the paper is unusually careful: the eight maximal Levi subgroups of E8 are checked, the trace computations in Claim 7.3 are exhibited, and the construction of the toral subgroup in Proposition 6.2(e) via the 8-by-16 generator matrix for the self-dual code C16 is explicit and verifiable. The main risk is a cited-theorem gap in Proposition 5.2; provided that gap is repaired, the paper is a strong contribution.","major_comments":[{"comment":"The proof of Proposition 5.2 invokes [SS70, Theorem 5.16] to obtain a maximal torus of H = C_G(A1)^0 that is invariant under the action of the finite abelian group A2. The cited theorem is stated for semisimple groups, while H is merely reductive. The footnote asserts that one can reduce to the semisimple case by modding out by the center, and cites [CGP25, Proposition 5.10] as an alternative, but the paper gives no proof that an A2-stable maximal torus of H/Z(H) lifts to an A2-stable maximal torus of H, and it does not state whether [SS70, Theorem 5.16] covers a finite abelian group of automorphisms or only a single semisimple automorphism. If iteration over elements of A2 is needed, each step requires a statement about invariant maximal tori in fixed-point subgroups of semisimple automorphisms, which is not automatic. Since Proposition 5.2 feeds Corollary 5.3 and hence Theorem 1.1, this is load-bearing; please supply a complete argument or a precise statement and reference covering the reductive case with a finite abelian automorphism group.","section":"§5, Proposition 5.2 and footnote 1"}],"minor_comments":[{"comment":"When citing [RY01b, Proposition 5.3] for a self-centralizing subgroup isomorphic to μ2^8 of depth 4, the different convention for depth in [RY01b] should be spelled out immediately in the text, since the footnote already notes that the present paper uses the index rather than the exponent.","section":"§7.2 and footnote 3"},{"comment":"Please ensure that the notation for the non-elementary maximal finite abelian subgroups of E8 is unambiguous: the group written as μ_3^6 in the introduction and in Proposition 6.2(c) appears to be μ_6^3 (three copies of cyclic group of order 6), and the proof in §7.5 only makes sense with that reading.","section":"§1.5 and §7.5"},{"comment":"The assertion that the finite extension F/k_r is isomorphic to k_r over k, cited as [GR09, Corollary 5.4], is surprising and is used to apply Lemma 2.8 over F; a brief explanation or a direct quotation of the cited statement would improve readability.","section":"§3, proof of Theorem 1.2"},{"comment":"The displayed conclusion of Lemma 7.1(2) is visually garbled in the typeset text; it should state that the fixed-point subalgebra of the adjoint action of A_C on so_{2n} is zero.","section":"§7.4, Lemma 7.1(2)"}],"recommendation":"major_revision","confidential_remarks":"The paper is strong and the applications are convincing. The only blocking issue is the use of [SS70, Theorem 5.16] in Proposition 5.2: the reduction from reductive to semisimple and the handling of a finite abelian group of automorphisms must be written out or replaced by a precise reference. If the authors can do that, I would support acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth your time. The paper answers the Totaro–Wang question (Theorem 1.1): for an affine k-group over an algebraically closed field, every finite abelian prime-to-p subgroup is close to toral, with the defect dividing the Grothendieck torsion index. That alone justifies publication. It also computes t2(E8)=60, saving a version of Tits' “hypothèse optimiste,” and extends the genus-one splitting obstruction from [RS26] to a general setting.\n\nThe proofs are serious rather than merely sketched. The E8 section checks all eight Levi subgroups and does the depth computations with explicit traces; the spin/code case is concrete. The chain from Proposition 3.1 through Lemma 4.3 and Corollary 5.3 is mostly self-contained and does not reduce to the authors' previous work. The reuse of [RS26] and [RY01b] is legitimate; the new theorems are genuinely new.\n\nThe one point I would flag is in Proposition 5.2. The proof invokes [SS70, Theorem 5.16] to get a maximal torus of C_G(A1)^0 stable under the action of the finite group A2. The theorem as usually cited is stated for semisimple groups, and the footnote's reduction to the semisimple case by quotienting the center is plausible: the preimage of a maximal torus is a torus. But the paper doesn't verify that [SS70, Thm 5.16] covers a finite group of automorphisms, as opposed to a single automorphism. If it covers only a single automorphism, one needs a short extra argument to get the whole group A2 stabilized. This is a local presentation gap, not evidence that the theorem is false; the authors also cite [CGP25, Prop 5.10] as an alternative. I'd want a referee to check that passage before publication.\n\nNo other soft spot matches it. The circularity burden is low: Theorem 1.1 is proved via the new depth/cohomology machinery, not by assuming what it proves.\n\nThis is for specialists in algebraic groups, torsors, and essential dimension. It deserves a serious referee and will be read closely. I recommend sending it to peer review with a request to spell out the [SS70] step.\n\nBest,","headline":"Strong paper that answers the Totaro–Wang question and computes the E8 variant t2=60; the only real soft spot is a terse citation in Proposition 5.2 that should be checked by a referee.","tokens_in":48469,"tokens_out":12437,"would_cite":true,"duration_ms":106147,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["20G07","20G10","20G41","12J10"],"pacs":[],"model":"deepseek-v4-flash","headline":"For any affine algebraic group G over an algebraically closed field, every finite abelian subgroup A whose order is not divisible by the characteristic meets some maximal torus T in a subgroup of index dividing the Grothendieck torsion…","keywords":["finite abelian subgroups","maximal torus","Grothendieck torsion index","torsion primes","E8","torsors","iterated Laurent series fields","genus one curves"],"falsifier":"To refute the main theorem, exhibit an algebraically closed field $k$, an affine $k$-group $G$, and a finite abelian subgroup $A$ with $\\operatorname{char}(k)\\nmid |A|$ such that every maximal torus $T$ of $G$ satisfies $[A : A\\cap T]\\nmid t(G)$. A targeted check is to test the semisimple reduction directly: find a finite group acting on a reductive group whose invariant maximal torus in the quotient by the center does not lift to an invariant maximal torus of the original group.","tokens_in":47352,"feed_emoji":"🎯","tokens_out":11084,"duration_ms":91611,"temperature":0.7,"pith_summary":"The paper proves a uniform structural bound on finite abelian subgroups of algebraic groups. Over an algebraically closed field, for any affine group G and any finite abelian subgroup A whose order is not divisible by the field characteristic, some maximal torus T of G satisfies $[A : A\\cap T] \\mid t(G)$, where $t(G)$ is the Grothendieck torsion index. Since $t(G)$ is a fixed integer computed from the degrees of fields needed to split $G$-torsors, the theorem reduces a question about arbitrary finite abelian subgroups to a computation of one invariant. This is the engine behind the paper's applications: a positive answer to a question about $G$-torsors over iterated Laurent series fields, the value $t_2(E_8)=60$ for an $E_8$-group in characteristic zero, and obstructions to splitting certain torsors by genus one curves.","feed_headline":"Maximal tori capture finite abelian subgroups up to one fixed index","feed_subtitle":"The Grothendieck torsion index $t(G)$ bounds how far any finite abelian subgroup is from being toral.","key_machinery":"The main machinery is the depth of $A$, the greatest common divisor of the indices $[A : A\\cap T]$ as $T$ ranges over maximal tori, together with the Grothendieck torsion index $t(G)$, which is the least common multiple of indices of all $G$-torsors over all field extensions and can be computed as the index of a single versal torsor. The pivotal structural proposition shows that if $A$ is a product of pairwise-coprime-order subgroups and each factor is toral, then $A$ is toral; its proof needs, for a finite group acting on the connected centralizer of one factor, a maximal torus invariant under the action. Around this core, the paper connects torsors over iterated Laurent series fields with Galois groups of their splitting fields, and uses a classification of maximal finite abelian subgroups of $E_8$ together with a correspondence between such subgroups and self-dual binary codes to compute depths.","core_discovery":"The central claim is that non-toral parts of finite abelian subgroups are controlled by torsion. Concretely, if $A\\subset G$ is finite abelian and $\\operatorname{char}(k)\\nmid |A|$, then $A$ has a toral subgroup $A\\cap T$ whose index divides the Grothendieck torsion index $t(G)$; moreover, one can choose $T$ so that this index is exactly the depth of $A$, the greatest common divisor of all such indices. The paper obtains this by extending an earlier characteristic-zero result, which treated only $p$-subgroups, to all prime-to-characteristic finite abelian subgroups, and by working without resolution of singularities so that positive characteristic is covered. Under the mild 'good characteristic' assumption, the paper also identifies the torsion-index variant $t_2(G)$, defined through torsors over iterated Laurent series fields, with the depth invariant $t_3(G)$.","pith_inferences":["The bound $[A : A\\cap T]\\mid t(G)$ is probably sharp only for $p$-primary subgroups at torsion primes; for composite-order $A$ the realized indices may be much smaller, and a prime-by-prime refinement of the depth invariant could sharpen the applications to torsor splitting.","The equality $t_2(G)=t_3(G)$ in good characteristic suggests that the full Grothendieck torsion index may sometimes be recovered by looking only at torsors over iterated Laurent series fields; testing whether $t_2(G)=t(G)$ for reductive groups with no small torsion primes would clarify how much of the classical invariant this new viewpoint captures.","The use of self-dual binary codes in the $E_8$ depth computation indicates that the same path could compute $t_2(\\operatorname{Spin}_n)$ and relate splitting obstructions for $\\operatorname{Spin}_n$-torsors to coding-theoretic invariants such as minimum weight.","Because the main theorem excludes finite abelian subgroups whose order is divisible by the field characteristic, any analogue in positive characteristic would require a separate invariant in place of $t(G)$; the paper leaves that boundary unexplored."],"forward_implications":["For a prime $p$ that does not divide $t(G)$, every finite abelian $p$-subgroup of $G$ is toral; this makes the classical characterization of torsion primes by non-toral elementary abelian subgroups quantitative.","Over $k((t_1))\\cdots((t_r))$, every $G$-torsor with a zero-cycle of degree $d$ has a closed point of degree dividing $d$ whenever $G$ is smooth affine and the characteristic is good for $G$.","For a group of type $E_8$ in characteristic zero, the iterated-Laurent-series torsion index is $t_2(E_8)=60$, so the 'optimistic hypothesis' on splitting fields of $E_8$-torsors holds for this variant even though the full Grothendieck torsion index is $26325$.","A torsor induced from an elementary abelian $p$-subgroup of sufficiently large non-toral rank cannot be split by any torsor under a $d$-dimensional abelian variety, and in particular not by a genus one curve; this yields explicit $E_8$-torsors that cannot be split by a genus one curve."],"supporting_citations":[{"why":"Defines the Grothendieck torsion index and proves it is the index of a versal torsor; this is the invariant Theorem 1.1 bounds.","marker":"[Gro58]"},{"why":"Proves the characteristic-zero predecessor for p-subgroups that Theorem 1.1 extends, and supplies depth computations used in the E8 analysis.","marker":"[RY01b]"},{"why":"Provides the theory of torsion primes, existence of non-toral elementary abelian p-subgroups, and structural facts about centralizers used throughout.","marker":"[Ste75]"},{"why":"Supplies the invariant-maximal-torus theorem in the semisimple case that Proposition 5.2 reduces to; this is the load-bearing step flagged in the weakest assumption.","marker":"[SS70]"},{"why":"Classifies maximal finite abelian subgroups of E8 up to conjugacy, the input for the t2(E8)=60 computation.","marker":"[DE17]"},{"why":"Establishes the prior genus-one splitting obstruction and the lemmas on abelian-variety torsors that Theorem 1.8 generalizes.","marker":"[RS26]"},{"why":"Computes the full torsion index of E8 and its simple subgroups, providing the numerical contrast with t2(E8)=60 and the torsion indices of the Levi factors.","marker":"[Tot05a]"}],"fun_headline_variants":["Finite abelian subgroups sit inside a torus up to a fixed index","Torsion index bounds how far abelian subgroups stray from tori","Maximal torus captures every abelian subgroup after bounded quotient","Abelian subgroups almost lie in a torus: index divides t(G)","One torus tames all finite abelian subgroups, index ≤ t(G)"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing step is the reduction, in the proof of the key proposition, from a reductive connected centralizer to its semisimple quotient when seeking a maximal torus invariant under the action of a complementary finite subgroup; if that reduction fails to produce an invariant maximal torus, the proof of Theorem 1.1 collapses.","fun_headline_variants_meta":{"raw":{"variants":["Finite abelian subgroups sit inside a torus up to a fixed index","Torsion index bounds how far abelian subgroups stray from tori","Maximal torus captures every abelian subgroup after bounded quotient","Abelian subgroups almost lie in a torus: index divides t(G)","One torus tames all finite abelian subgroups, index ≤ t(G)"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000447,"raw_usage":{"total_tokens":2284,"prompt_tokens":1000,"completion_tokens":1284,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":616,"completion_tokens_details":{"reasoning_tokens":1188}},"tokens_in":616,"tokens_out":1284,"duration_ms":9508,"temperature":1.0,"reasoning_tokens":1188,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T15:07:14.757623+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"To refute the main theorem, exhibit an algebraically closed field $k$, an affine $k$-group $G$, and a finite abelian subgroup $A$ with $\\operatorname{char}(k)\\nmid |A|$ such that every maximal torus $T$ of $G$ satisfies $[A : A\\cap T]\\nmid t(G)$. A targeted check is to test the semisimple reduction directly: find a finite group acting on a reductive group whose invariant maximal torus in the quotient by the center does not lift to an invariant maximal torus of the original group.","supporting_citations":[],"review_version":2}