{"id":"e0a84c65-4a16-41d2-9858-51138093cf36","arxiv_id":"1908.08146","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The splitting principle for cohomological invariants of finite orthogonal reflection groups with values in cycle modules, and for Witt and Milnor-Witt invariants, is proved over fields of characteristic coprime to the group order.","lead":"The paper proves Serre's splitting principle for finite orthogonal reflection groups: a cohomological invariant with values in a Rost cycle module is trivial once it is trivial on every elementary abelian 2-subgroup generated by reflections. The proof is extended to Witt and Milnor-Witt K-theory invariants, under the condition that the base field's characteristic is coprime to the group order.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3.5's splitting step fails for inseparable residue fields, so the unramifiedness argument on which Theorem 4.6 rests is not established for imperfect base fields.","rationale":"The central theorem, Theorem 4.6, is proved by showing that the value of an invariant on a versal torsor is unramified at all height-one primes of the quotient affine space, using the specialization theorem 3.5 at primes in U/W and the induction argument at primes outside. The proof of Theorem 3.5 is therefore load-bearing. The reader identified the excellence/approximation step as the weakest assumption, but conceded that it could be cited to EGA and Bosch-Lütkebohmert-Raynaud. My concern is stronger: even granting excellence, the specific factorization of a coefficient-field splitting through the henselization is false when the residue field is inseparable, because the henselization's fraction field is built from separable extensions only. This is not a mere presentation gap; it affects the proof for imperfect base fields with odd characteristic, which are explicitly allowed by the theorem's hypothesis that char(k0) is coprime to |W|. The theorem may be salvageable by adding a perfectness hypothesis or by finding a different specialization argument, but as written the proof does not establish the claimed generality. The proposed test settles the concern directly by exhibiting a DVR of the allowed type where the asserted factorization fails. No machine-checked verification is present, and the cited deep results do not repair this particular implication.","tokens_in":17955,"tokens_out":37867,"duration_ms":417231,"concrete_test":"Take k0 = F_p(s) with p odd, and R = k0[x]_{(x^p - s)}. Verify that (i) the completion R-hat has a coefficient field, so a splitting k → R-hat exists by Cohen's structure theorem; and (ii) the henselization R^h has no splitting k → R^h, because the fraction field of R^h is a filtered union of finite separable extensions of k0(x), hence cannot contain a root of T^p - s. If both hold, the factorization assertion in the proof of Theorem 3.5 is false for an allowed DVR. As a control, check the same factorization over a perfect base field or in characteristic zero, where the step is valid.","verdict_should_be":"REJECT","load_bearing_attack":"In the proof of Theorem 3.5 the authors need a section j: k → R^h of the residue map. They claim that the Cohen splitting j: k → R-hat factors through R^h because R^h is excellent and satisfies the approximation property. This implication is not valid in general. Let k0 = F_p(s) with p an odd prime not dividing |W|, and R = k0[x]_{(x^p - s)}. This is a DVR essentially of finite type over k0, with residue field k = k0(s^{1/p}), purely inseparable over k0. By Cohen's structure theorem R-hat has a coefficient field, hence a splitting k → R-hat. But R^h has no such splitting: its fraction field is a filtered union of finite separable extensions of K = k0(x), so it cannot contain a root of T^p - s, which any splitting would require. Thus the sentence \"This splitting factors via R^h\" fails for a DVR of the kind allowed in the theorem. In Section 4.9 the proof of the Claim sends all height-one primes Q in U/W to Theorem 3.5; primes such as (T_1^p - s) in the polynomial ring S(V^∨)^W lie in U/W for suitable W and have exactly this inseparability. The induction therefore rests on an invalid factorization at those primes.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a splitting principle for cohomological invariants with values in Rost cycle modules for finite orthogonal reflection groups W over a field k0 of characteristic coprime to |W|: an invariant is trivial if and only if its restrictions to all elementary abelian 2-subgroups generated by reflections are trivial. The proof combines an explicit versal W-torsor, a specialization/unramifiedness theorem for cycle-module invariants, and induction on |W|. In the final section the same principle is claimed for Witt invariants and Milnor-Witt K-theory invariants.","tokens_in":18094,"tokens_out":36976,"duration_ms":353447,"significance":"If correct, this is a valuable result. It extends Serre's splitting principle from Weyl groups and Galois cohomology to arbitrary orthogonal reflection groups and arbitrary cycle modules, and it provides a clean induction framework. The versal-torsor description and the reduction to elementary abelian subgroups are elegant. The paper also gives explicit transfer of the method to Witt and Milnor-Witt invariants. Both strengths and weaknesses are present, however: the specialization theorem (Theorem 3.5) is not established for imperfect base fields as written, and the induction in Theorem 4.6 uses a stronger subgroup hypothesis than the one stated.","major_comments":[{"comment":"The proof of Theorem 3.5 asserts that the Cohen splitting j: k -> R^h of the residue map exists and is compatible with the k0-structure. This is not valid in general. The Cohen structure theorem gives a coefficient field of the completion R-hat, i.e., an isomorphism of abstract fields k -> K ⊂ R-hat, but this isomorphism need not be a k0-algebra homomorphism. For example, let k0 = F_p(s), p odd, and R = k0[x]_(x^p - s). The residue field k = k0(s^{1/p}) is purely inseparable over k0, and Frac(R^h) is a direct limit of separable extensions of Frac(R), so no k0-algebra section k -> R^h exists. The subsequent use of r_j on H^1(-,G) for an arbitrary linear algebraic group G, and the claim that the fields k_i are finitely generated over k0, require exactly that k0-linearity. Since Section 4.9 invokes Theorem 3.5 for all height-one primes Q in U/W, including primes of the form (T_0^p - s) when W has a trivial summand, this gap affects the proof of Theorem 4.6 as stated.","section":"Section 3.5"},{"comment":"The induction in the Claim uses the assumption of triviality on elementary abelian 2-subgroups, but the proof repeatedly refers to arbitrary '2-subgroups generated by reflections.' In case (a), the subgroup H' = <s_alpha>.H is a 2-subgroup generated by reflections, but it is elementary abelian only if H is elementary abelian; for general H the assumption does not imply Res^{H'}_W(a) = 0. The same issue occurs in case (b), where 'H is a 2-subgroup of W_{±alpha} generated by reflections' is used to conclude Res^H_W(a) = 0. The argument is repairable by restricting H to elementary abelian 2-subgroups; then H' is elementary abelian because s_alpha commutes with W_alpha (Lemma 4.5(i)). As written, however, the induction step does not follow from the stated hypothesis.","section":"Section 4.9"},{"comment":"The abstract and the theorem in the introduction state the condition as triviality on 'all 2-subgroups of W generated by reflections,' while Theorem 4.6 states 'elementary abelian 2-subgroups.' This is not a cosmetic discrepancy: the proof only supports the elementary abelian version, and the stronger statement is not established. The authors should align the statements and indicate which formulation is intended.","section":"Introduction and Abstract"}],"minor_comments":[{"comment":"In case (a), the sentence beginning 'Let H subset W_alpha be a 2-subgroup generated by reflections' should read 'elementary abelian 2-subgroup' to match the induction hypothesis.","section":"Section 4.9"},{"comment":"There is a typo: 'orthgonal' should be 'orthogonal'.","section":"Section 4.10"},{"comment":"The notation T_K is used for the generic fiber of a torsor T over X, but it is not defined in the text; a short definition would improve readability.","section":"Section 3.5"}],"recommendation":"major_revision","confidential_remarks":"The main theorem is plausible and the overall strategy is sound, but the proof of Theorem 3.5 has a genuine gap for imperfect base fields, and the induction in Theorem 4.9 uses a stronger assumption than stated. The first issue may be curable either by proving a version of Theorem 3.5 for finite constant groups using non-k0-linear field sections, or by restricting the main theorem to perfect base fields. The second issue is a straightforward correction. I would not recommend rejection, because the flaws are localized and the central claim is likely salvageable, but the current manuscript does not fully prove the stated theorem."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nHere's my read. The paper proves Serre's splitting principle for cohomological invariants of finite orthogonal reflection groups with values in arbitrary cycle modules, and then extends it to Witt and Milnor-Witt invariants. That is a real result: the proof is self-contained, generalizes Serre's announced Weyl-group case, and handles all reflection groups uniformly. The induction on |W| with reduction to W±α is clean, and the authors are honest about the prior flawed thesis and Ducoat's unpublished preprint.\n\nThe bad news is that the proof has a load-bearing gap in Theorem 3.5. The authors need a section j: k -> R^h of the residue map. They claim a Cohen splitting k -> \\hat{R} factors through R^h because R^h is excellent and satisfies the approximation property. That is not generally true. Take k0 = F_p(s) and R = k0[x]_{(x^p - s)}. The residue field k = k0(s^{1/p}) is purely inseparable, and R^h admits no splitting at all: its fraction field is a filtered union of separable extensions of k0(x), so it cannot contain a root of T^p - s. The approximation property can lift solutions of polynomial equations with coefficients in R^h, but a coefficient field is not given by such equations unless the residue field is separable. This is not an edge case: after CSTB, the invariant ring is a polynomial algebra, and primes like (U^p - s) lie in U/W and have exactly this inseparable residue field. So the unramifiedness argument in Section 4.9 is not established for imperfect base fields of characteristic p. The main theorem may be true, but this proof does not cover it.\n\nThere are also two smaller issues. The abstract and the theorem statement in the introduction say '2-subgroups' where the actual theorem and the proof require 'elementary abelian 2-subgroups'. This is a presentation slip, not a mathematical one—the induction steps work once you add 'elementary abelian' at each spot—but it should be fixed. The Witt extension in Section 5.2 is terse; it silently uses homotopy invariance for unramified I^n on affine space. That is standard but should be stated.\n\nWho is this for? Specialists in cohomological invariants and quadratic forms. It deserves a serious referee. I would send it to peer review, with instructions to scrutinize Theorem 3.5. If the authors can fix the inseparable-residue-field issue—or restrict the main theorem to fields where the relevant residue fields are separable—the paper would be a solid contribution. As written, I would not cite it for the theorem over arbitrary base fields.\n\nBest.","headline":"A mostly solid proof of a significant splitting principle, but the specialization theorem has a real gap for inseparable residue fields that currently leaves the main theorem unproved over imperfect base fields.","tokens_in":18736,"tokens_out":13255,"would_cite":false,"duration_ms":129493,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["19D45","20G10"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that a cohomological invariant of a finite orthogonal reflection group with values in a cycle module is trivial if and only if its restrictions to all elementary abelian 2-subgroups generated by reflections are trivial…","keywords":["cohomological invariants","reflection groups","cycle modules","splitting principle","Witt invariants","Milnor-Witt K-theory","versal torsors","orthogonal reflection groups"],"falsifier":"A counterexample would be a finite orthogonal reflection group $W$ over a field $k_0$ with $\\operatorname{char}(k_0)\\nmid |W|$, a cycle module $M^*$, and a nonconstant invariant $a$ whose restriction to every elementary abelian 2-subgroup generated by reflections is zero. A concrete candidate to test computationally is the dihedral reflection group $I_2(4)$ over $\\mathbb{Q}(i)$ with $M^*=H^*(-,\\mathbb{Z}/2)$, using the paper's explicit versal torsor to compute the invariant group.","tokens_in":17615,"feed_emoji":"🔀","tokens_out":12703,"duration_ms":328356,"temperature":0.7,"pith_summary":"Finite orthogonal reflection groups are generated by linear reflections, and their first Galois cohomology classifies objects such as Galois algebras and torsors. This paper establishes a splitting principle for them: under the hypothesis that the base field has characteristic coprime to the group order, a cohomological invariant taking values in any cycle module is zero if and only if it is zero on every elementary abelian 2-subgroup of the group that is generated by reflections. The same statement is proved for Witt invariants and for Milnor-Witt K-theory invariants. This matters because it turns the usually hard computation of invariants of a large reflection group into the much easier computation on small 2-torsion subgroups, and it supplies the technical foundation for computing invariants of Weyl groups with values in mod-2 cohomology theories.","feed_headline":"Reflection-group invariants vanish if their 2-subgroups do","feed_subtitle":"Proof reduces cycle-module, Witt, and Milnor-Witt K-theory invariants to small cases.","key_machinery":"The central machinery is the unramified-cohomology formalism for cycle modules: a cycle module assigns a graded abelian group to every finitely generated field extension and carries second residue and specialization maps satisfying axioms (R3a), (R3c), and (R3d). The proof combines three components. First, an explicit versal $W$-torsor is described as the $W$-Galois algebra $E/K$, where $K$ is the invariant subfield of the polynomial ring $S(V^\\vee)$ under the reflection group $W$ acting on the dual of a faithful orthogonal representation $V$. Second, a specialization theorem shows that the value of any invariant on this versal torsor is unramified at every codimension-one point of the quotient $A(V)/W$. Third, the classical invariant-theory theorem that the invariant ring of a finite reflection group is a polynomial ring identifies $A(V)/W$ with affine space, and homotopy invariance of cycle modules forces the unramified value to be constant. The induction step uses the isotropy group $W_{\\pm\\alpha}=\\langle s_\\alpha\\rangle W_\\alpha\\simeq \\mathbb{Z}/2\\times W_\\alpha$ of a height-one prime and the root-system identity $\\bigcup_{\\alpha\\in\\Delta}\\ker(\\alpha^\\vee)=\\bigcup_{w\\neq 1}\\ker(w-\\mathrm{id})$, which together control the ramification along the divisor where the versal torsor degenerates.","core_discovery":"Under the assumption that the base field has characteristic coprime to $|W|$, the paper proves that an invariant $a: H^1(-,W)\to M^n(-)$ with values in a cycle module $M^*$ is trivial if and only if its restriction to every elementary abelian 2-subgroup of $W$ generated by reflections is trivial. The proof constructs an explicit versal $W$-torsor: the generic fiber of the quotient map from the complement of the reflection hyperplanes to the quotient affine space, whose function field is the invariant subfield of a polynomial ring. Using a specialization theorem for cycle modules, it shows that the value of any invariant on this versal torsor is unramified on the quotient affine space, hence constant by homotopy invariance; the vanishing on elementary abelian 2-reflection subgroups then forces that constant to be zero. The same argument, with the second residue map replaced by the first residue map, yields the analogous statement for Witt invariants, and via the pull-back description of Milnor-Witt K-theory, also for Milnor-Witt K-theory invariants.","pith_inferences":["The proof's reliance on a faithful orthogonal representation suggests the splitting principle might be stated more generally for any finite group equipped with a representation in an orthogonal group; the paper itself only treats orthogonal reflection groups, but the versal-torsor construction would adapt if such a representation is supplied.","Because the specialization theorem works through the henselization rather than the completion, the same argument should carry over to invariants with values in any cohomology theory that satisfies the three residue axioms and homotopy invariance, not just cycle modules, Witt groups, and Milnor-Witt K-theory.","The theorem leaves open the pseudo-reflection case raised in the paper's final remarks; if the splitting principle holds there, the root-system identity would need a replacement, since the kernel of a pseudo-reflection need not be a reflection hyperplane.","For practical computation, the result suggests an algorithm: compute the invariants of a reflection group by first computing invariants of its maximal elementary abelian 2-reflection subgroups and then checking which combinations extend to the whole group; the proof shows the extension is unique when it exists."],"forward_implications":["For every finite orthogonal reflection group $W$ with $\\operatorname{char}(k_0)\\nmid |W|$, the invariant group $\\operatorname{Inv}_{k_0}(W,M^*)$ embeds into a direct sum of invariant groups of maximal elementary abelian 2-subgroups generated by reflections.","Checking whether an invariant is zero therefore requires only knowledge of the invariants of groups of the form $(\\mathbb{Z}/2)^r$, which are far easier to compute than the original group.","The same reduction is valid for Witt invariants and for Milnor-Witt K-theory invariants, so computations of these invariants for reflection groups can be organized around elementary abelian 2-subgroups as well.","The explicit versal torsor description and the unramifiedness argument give a concrete route to computing invariants of Weyl groups with values in cycle modules annihilated by 2, the intended sequel of this work.","If the splitting principle is true, then invariants of reflection groups are detected on a very small family of subgroups, which also implies a strong rigidity property: a nontrivial invariant must already be nontrivial on some elementary abelian 2-reflection subgroup."],"supporting_citations":[{"why":"Defines cohomological invariants and versal torsors and contains the detection principle that the paper extends from Galois cohomology to cycle modules.","marker":"[9]"},{"why":"Introduces cycle modules and supplies the residue axioms, specialization maps, and homotopy invariance used in the proof.","marker":"[20]"},{"why":"Shows that henselizations of discrete valuation rings essentially of finite type are excellent, the key local-ring fact in the specialization theorem.","marker":"[10]"},{"why":"Provides the approximation property for excellent henselian local rings, which lets the residue-field splitting of the completion factor through the henselization.","marker":"[4]"},{"why":"Gives the structure theory of reflection groups, including root systems and the polynomial invariant ring theorem.","marker":"[6]"},{"why":"Identifies $H^1$ of filtered limits of fields and rings with limits of $H^1$ sets, used to pass from the henselization to the étale local rings.","marker":"[2]"},{"why":"Supplies the isomorphism between $H^1_{\\mathrm{et}}$ on a henselian local ring and $H^1$ of its residue field, used to identify the pulled-back torsor.","marker":"[7]"},{"why":"Records how second residue maps act on powers of the fundamental ideal of Witt groups, which makes the argument carry over to Witt invariants.","marker":"[1]"}],"fun_headline_variants":["Invariant trivial iff reflection 2-subgroups are trivial","Splitting principle for reflection group invariants","New proof: invariants reduce to elementary abelian 2-groups","Serre's splitting principle for reflection group invariants","Cycle, Witt, and Milnor-Witt invariants split by 2-groups"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the henselization of every discrete valuation ring built from a finitely generated field extension of the base field is excellent, so that a splitting of the residue field into the completion factors back through the henselization; the specialization theorem and the induction on $|W|$ both depend on this factorization.","fun_headline_variants_meta":{"raw":{"variants":["Invariant trivial iff reflection 2-subgroups are trivial","Splitting principle for reflection group invariants","New proof: invariants reduce to elementary abelian 2-groups","Serre's splitting principle for reflection group invariants","Cycle, Witt, and Milnor-Witt invariants split by 2-groups"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000652,"raw_usage":{"total_tokens":2939,"prompt_tokens":848,"completion_tokens":2091,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":464,"completion_tokens_details":{"reasoning_tokens":2004}},"tokens_in":464,"tokens_out":2091,"duration_ms":16095,"temperature":1.0,"reasoning_tokens":2004,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:49:52.899136+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A counterexample would be a finite orthogonal reflection group $W$ over a field $k_0$ with $\\operatorname{char}(k_0)\\nmid |W|$, a cycle module $M^*$, and a nonconstant invariant $a$ whose restriction to every elementary abelian 2-subgroup generated by reflections is zero. A concrete candidate to test computationally is the dihedral reflection group $I_2(4)$ over $\\mathbb{Q}(i)$ with $M^*=H^*(-,\\mathbb{Z}/2)$, using the paper's explicit versal torsor to compute the invariant group.","supporting_citations":[{"cited_title":"Garibaldi, A","cited_arxiv_id":null,"evidence_quote":"Defines cohomological invariants and versal torsors and contains the detection principle that the paper extends from Galois cohomology to cycle modules."},{"cited_title":"Rost, Chow groups with coefficients , Doc","cited_arxiv_id":null,"evidence_quote":"Introduces cycle modules and supplies the residue axioms, specialization maps, and homotopy invariance used in the proof."},{"cited_title":"Grothendieck, J","cited_arxiv_id":null,"evidence_quote":"Shows that henselizations of discrete valuation rings essentially of finite type are excellent, the key local-ring fact in the specialization theorem."},{"cited_title":"Bosch, W","cited_arxiv_id":null,"evidence_quote":"Provides the approximation property for excellent henselian local rings, which lets the residue-field splitting of the completion factor through the henselization."},{"cited_title":"Bourbaki, ´El´ ements de math´ ematique","cited_arxiv_id":null,"evidence_quote":"Gives the structure theory of reflection groups, including root systems and the polynomial invariant ring theorem."},{"cited_title":"Artin, A","cited_arxiv_id":null,"evidence_quote":"Identifies $H^1$ of filtered limits of fields and rings with limits of $H^1$ sets, used to pass from the henselization to the étale local rings."},{"cited_title":"Demazure, A","cited_arxiv_id":null,"evidence_quote":"Supplies the isomorphism between $H^1_{\\mathrm{et}}$ on a henselian local ring and $H^1$ of its residue field, used to identify the pulled-back torsor."},{"cited_title":"Arason, Kohomologische Invarianten quadratischer Formen , J","cited_arxiv_id":null,"evidence_quote":"Records how second residue maps act on powers of the fundamental ideal of Witt groups, which makes the argument carry over to Witt invariants."}],"review_version":1}