{"id":"f195e532-632c-4abd-b0ae-cdb244fabd83","arxiv_id":"2506.02638","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For every fan supported in the negative Weyl chamber, universal equivariant toroidal embeddings of split reductive group schemes over Z exist and specialize to the classical embeddings over every algebraically closed field.","lead":"A mathematician has constructed, over the integers, a single family of toroidal embeddings of reductive groups that specializes to the classical compactifications over every algebraically closed field at once. The result supplies uniform integral models for all equivariant toroidal compactifications of Chevalley groups, objects used to study how algebraic groups and their reductions modulo primes can be completed.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Construction 4.6's representability of X_σ as an algebraic space is imported from unproved theorems in [Li25]; their hypotheses are not verified for the non-group toric big cell Ω_σ.","rationale":"Good-faith reading: the paper's own contribution is the careful construction of the rational action A_σ (Theorem 4.4) and the gluing argument; if Section 3's black boxes are valid, the proof is coherent. I found no algebraic contradiction in the explicit formulas, and the paper gives real evidence: the big-cell computations in Lemma 4.1 and Lemma 4.3, the use of flat descent and fibral criteria, and the reduction to [Li23]'s wonderful compactification. However, the central claim depends on a representability theorem that is quoted, not proved, and the hypotheses for the specific Y=Ω_σ are not checked. This is more than a stylistic issue: Construction 4.6 defines X_σ as a sheaf; all later statements (scheme structure, fibral identification with classical toroidal embeddings, quasi-projectivity, smoothness/properness criteria) presuppose that this sheaf is an algebraic space. The reader's weakest_assumption identifies exactly this point, and I agree. My verdict remains conditional: the author must verify [Li25]'s hypotheses for Ω_σ or include the necessary proofs. No scientific-integrity flags; the concern is about unsupported import of a companion result, not about any claim's being dishonest.","tokens_in":16473,"tokens_out":10893,"duration_ms":110831,"concrete_test":"Independently re-prove Theorem 3.7 for Y=Ω_σ: compute the graph Γ of φ(g1,g2,ω)=A_σ(g1^{-1}, A_σ(g2,ω)) inside (G×G)×Ω_σ×Ω_σ using the explicit formula for A_σ from Theorem 4.4 and the description of Dom(A_σ), and verify that the two projections Γ⇉(G×G)×Ω_σ are fppf covers. If this verification cannot be completed without additional hypotheses (for example, Y being a group scheme), then X_σ in Construction 4.6 is not known to be an algebraic space, and the fibral identification in Lemma 4.8 has no defined target.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3 quotes Theorem 3.6 ([Li25, Cor. 5.9, Prop. 5.11]) and Theorem 3.7 ([Li25, Thm. 5.6]) to assert that the fppf quotient sheaf (G×_S Ω_σ×_S G)/∼_{A_σ} of Construction 4.6 is an algebraic space. This is the single step that turns the rational action A_σ of Theorem 4.4 into an actual G×G-scheme X_σ; without it, Lemma 4.8, Theorem 4.9, Proposition 4.10, and the gluing in Section 5 have no representable object to act on. The present text does not prove these theorems or verify their hypotheses for Y=Ω_σ. Ω_σ is flat and finitely presented, but it is not a group scheme, and Theorem 4.4 establishes only e×_S Ω_σ×_S e ⊂ Dom(A_σ), not the full conditions on the graph Γ of φ(g1,g2,ω)=A_σ(g1^{-1}, A_σ(g2,ω)) that Theorem 3.7 requires (for instance, that the two projections Γ⇉G×_S Ω_σ form an fppf equivalence relation). If [Li25]'s results apply only to group-valued Y, or require a stronger definition-domain condition, then Construction 4.6 yields only an fppf sheaf, and the central theorem has an unsupported step. This is a structural risk: a central representability result is black-boxed from an unpublished companion paper and should be supplied or verified before the claim is accepted.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a theory of toroidal embeddings for Chevalley group schemes over Z, in analogy with the classical classification of equivariant toroidal embeddings of reductive groups over algebraically closed fields. For a split reductive group G over a scheme S, a maximal split torus T contained in a Borel B, and a fan Σ in the negative Weyl chamber, the main theorem (Theorem 5.4) asserts the existence of a scheme X_Σ over Z with a G×G action extending the left and right translations of G, whose base change to any algebraically closed field is the classical toroidal embedding associated with Σ. The construction proceeds by first defining a rational action A_σ of G×G on the big cell Ω_σ = U^- × T_σ × U^+ (Theorem 4.4), then forming an fppf quotient sheaf X_σ = (G×Ω_σ×G)/∼ (Construction 4.6), which is claimed to be an algebraic space by results quoted from the companion paper [Li25]. The single-cone case is then glued along face inclusions to obtain X_Σ. The paper also states quasi-projectivity of the single-cone pieces (Theorem 4.9) and combinatorial criteria for smoothness and properness (Proposition 5.6).","tokens_in":16627,"tokens_out":12234,"duration_ms":117106,"significance":"If the gaps identified below are resolved, this paper would provide a uniform integral model for all toroidal embeddings of reductive groups over Z, going beyond earlier constructions for affine and projective cases and for the wonderful compactification. The explicit formulas for the rational action on the big cell (Lemma 4.1, Lemma 4.3, Theorem 4.4) are a concrete and useful contribution, and the sheaf-theoretic approach is natural. The paper is also careful to check its construction against the classical classification over algebraically closed fields (Lemma 4.8, Theorem 5.4), which is the right benchmark. However, the central representability step is imported from an unpublished companion paper without verification of its hypotheses, and the properness argument in Proposition 5.6 is not convincing. These issues are load-bearing, so the main theorem is not established within the present text.","major_comments":[{"comment":"The assertion that X_σ is an algebraic space rests entirely on Theorem 3.6 ([Li25, Cor. 5.9, Prop. 5.11]) and Theorem 3.7 ([Li25, Thm. 5.6]), neither of which is proved in this manuscript. For the specific scheme Y = Ω_σ, the paper verifies only the condition e×Ω_σ×e ⊂ Dom(A_σ) (Theorem 4.4). It does not verify the hypotheses needed to apply [Li25, Thm. 5.6], for instance that the two projections from the graph Γ of φ(g_1,g_2,ω) = A_σ(g_1^{-1},A_σ(g_2,ω)) to G×Ω_σ form an fppf equivalence relation, nor that the quotient by this graph relation is isomorphic to the quotient sheaf defined by ∼_A. Since the algebraic-space structure is the foundation for Lemma 4.8, Theorem 4.9, and the gluing in Section 5, the central theorem is unsupported unless the results of [Li25] are supplied or their hypotheses are explicitly checked for this Y.","section":"§3, Construction 4.6"},{"comment":"The proof of the properness criterion is not valid as written. It states that the criterion holds over an algebraically closed field by [BK05, Prop. 6.2.3(iv)] and then concludes that X_Σ is proper over Z because, by Lemma 4.8, all geometric fibers are proper and geometrically connected, citing [EGA IV3, Cor. 15.7.11]. Properness is not a fibral property: a separated finite-type morphism with proper geometric fibers need not be proper, and the cited corollary does not provide such a criterion. A direct argument, e.g., via the valuative criterion, is required. This gap affects the 'if' direction of the claimed combinatorial characterization of properness.","section":"§5, Proposition 5.6(2)"},{"comment":"The proof that X_σ is quasi-projective is too terse and depends on unstated assumptions. It shows that π: X_σ → X is affine by reducing, after étale descent to a strictly Henselian base, to showing that χ: Ω_σ → π^{-1}(Ω) is an isomorphism, using the fibral criterion [EGA IV4, Cor. 17.9.5]. The hypotheses of that fibral criterion (such as finite presentation and properness of the relevant morphism) are not checked, and the equality of the open immersion with the classical big-cell identification is asserted via Lemma 4.8 and [BK05, Prop. 6.2.3(i)] without spelling out the compatibility. This needs a more detailed proof, especially because the quasi-projectivity of X_σ is what turns the algebraic space of Construction 4.6 into a scheme.","section":"§4.3, Theorem 4.9"}],"minor_comments":[{"comment":"The notation f_1 in the proof of Lemma 4.2 is confusing: the text writes \"we define f_1 in a similar way\" but the displayed definition then uses f_j for the induced automorphism of V_j. Please use distinct symbols for the composed rational map and the individual simple-reflection maps.","section":"§4.2, Lemma 4.2"},{"comment":"In the proof of Lemma 4.8, the map ξ sending (g_1,ω,g_2) to g_1·ω·g_2 is called a monomorphism, but the well-definedness of ξ with respect to the equivalence relation ∼_A is not explicitly checked. It follows from the compatibility of A_σ with the classical action on each geometric fiber, but this should be stated.","section":"§4.3, Lemma 4.8"},{"comment":"The phrase \"the only way I can imagine\" is informal for a journal article; the remark also asserts without proof that the direct fppf quotient sheaf for a general fan Σ is an algebraic space, which again depends on the unverified results from [Li25].","section":"§5, Remark 5.5"},{"comment":"The arXiv text contains numerous transcription artifacts (e.g., '−/∫hortrightarrow', 'Ş', 'Gk−/∫hortrightarrowpXΣqk'), which should be corrected in the final version. Some displayed equations in the proof of Lemma 4.1 have potential typographical issues in the coordinates; please proofread the formulas against a standard Chevalley-system reference.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The main theorem depends on results from the author's companion preprint [Li25] (arXiv:2505.12777), which is not yet published and whose hypotheses are not checked for the specific non-group toric scheme Ω_σ. I suggest the editor ask the author to either include a self-contained proof of the representability of the fppf quotient sheaf for the rational action A_σ, or clearly state and verify the exact hypotheses of [Li25] used here. The properness argument in Proposition 5.6(2) appears to be incorrect as written and needs a genuine proof. These are fixable but substantial revisions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear [Colleague],\n\nYou should know this about arXiv:2506.02638: it proves a genuinely new existence theorem—universal equivariant toroidal embeddings over Spec Z for split reductive group schemes, for arbitrary fans supported in the negative Weyl chamber. The statement is new: prior work covered only the wonderful compactification ([Li23]) and affine/projective cases ([BS25]), while the classical BK05 classification was over algebraically closed fields. The construction is uniform and sheaf-theoretic: for each cone, X_σ is an fppf quotient sheaf (G×Ω_σ×G)/∼, with Ω_σ the toric big cell, and the paper shows this is a quasi-projective scheme over Z and glues to X_Σ. Theorem 4.4, where the rational action is built step-by-step with explicit formulas, is the technical heart, and it looks right.\n\nThe main soft spot is structural: the step that turns the quotient sheaf into an algebraic space is imported wholesale from the author's companion paper [Li25] (Theorems 3.6 and 3.7 here). The present text does not verify that the hypotheses of those theorems hold for Ω_σ, which is flat and finitely presented but emphatically not a group scheme. The stated theorems are general enough to cover it, but a referee cannot check this without reading [Li25] side-by-side. This is addressable—the author should either state the hypotheses and check them, or include the necessary arguments. It is not a fatal gap, but it is a real dependency.\n\nThe stress-test's specific worry—that only the identity-section domain condition is shown—is not quite right: Theorem 4.4 constructs the rational action on a dense open and verifies it agrees with the group law, which gives associativity. The real issue is simply that the representability result is black-boxed.\n\nTwo smaller things. Lemma 4.8, the fibral identification with the classical embedding, is compressed: the monomorphism to G_{k,σ} is clear from the rational action, but the surjectivity argument needs a sentence or two. Proposition 5.6's smoothness/properness criteria are stated with only brief sketches; if they're meant as results, they deserve fuller proofs.\n\nIf [Li25]'s results are as general as claimed, the paper is correct. The proof structure is sound, the citations to SGA3/BLR/EGA are appropriate, and I see no circularity or overreach. I'd send this to a serious referee: the theorem is important enough to merit the time, and the known gaps are all fillable. It's also a good reading-group paper for people working on integral compactifications.","headline":"A serious, likely correct construction of universal toroidal embeddings over Z for arbitrary fans, with the main caveat being a black-boxed representability theorem from a companion paper.","tokens_in":17334,"tokens_out":5433,"would_cite":true,"duration_ms":50101,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14L15","14M25","14M27","14L30","20G35"],"pacs":[],"model":"deepseek-v4-flash","headline":"For any fan supported in the negative Weyl chamber, this paper constructs a scheme X_Σ over Spec(Z) with a G×_Z G action whose base change to every algebraically closed field is the classical equivariant toroidal embedding of G_k.","keywords":["toroidal embeddings","Chevalley group schemes","split reductive group schemes","rational actions","fans in the negative Weyl chamber","toric varieties","integral models","algebraic spaces"],"falsifier":"Compute the definition domain of A_σ for a non-smooth cone σ over Spec(Z), for instance the cone generated by (1,1) in a rank-two cocharacter lattice; if e × Ω_σ × e is not contained in it, or if the equivalence relation on G × Ω_σ × G is not an fppf equivalence relation with representable diagonal, then X_σ is not an algebraic space and Theorem 1.1 fails.","tokens_in":16067,"feed_emoji":"🌐","tokens_out":7135,"duration_ms":61051,"temperature":0.7,"pith_summary":"This paper establishes an integral, or 'universal', version of equivariant toroidal embeddings for split reductive group schemes over the integers. For every fan in the negative Weyl chamber, it constructs a scheme X_Σ over Spec(Z) equipped with an action of G×_Z G that extends left and right translation of G, and whose base change to any algebraically closed field is the classical toroidal embedding of G_k determined by the same fan. Because the classical classification is combinatorial and characteristic-independent, such integral models are expected to specialize fiberwise; the paper proves they exist. The construction works uniformly for arbitrary fans, including non-affine and non-projective ones, and gives quasi-projectivity for single cones plus combinatorial smoothness and properness criteria.","feed_headline":"Toroidal embeddings for reductive groups now live over Z","feed_subtitle":"New schemes base-change to classical toroidal embeddings over every algebraically closed field.","key_machinery":"The load-bearing object is the fppf quotient sheaf X_σ = (G ×_S Ω_σ ×_S G)/∼_{A_σ}, built from the rational action A_σ: G ×_S Ω_σ ×_S G ⇢ Ω_σ that extends the two-sided translation of G on its open cell Ω_G ≅ U^- × T × U^+. The rational action is reconstructed step by step from explicit root-subgroup formulas, using the morphisms f_i and f to swap positive and negative root subgroups through T_σ, and its definition domain is forced to contain e × Ω_σ × e. Quoted theorems on rational actions then promote the quotient sheaf to an algebraic space with a G×G action, and the sheaf-theoretic description is what makes the special-fiber statement transparent: base change commutes with the quotient, so (X_σ)_k is literally the quotient that the classical embedding satisfies. Gluing via open immersions X_τ ↪ X_σ for faces τ ⊂ σ assembles X_Σ.","core_discovery":"The central discovery is that the classical toroidal embedding of a reductive group can be lifted to a single flat scheme over Spec(Z) that remembers the whole family of special fibers. For a single cone σ, the paper constructs X_σ as the fppf quotient sheaf (G ×_S Ω_σ ×_S G)/∼_{A_σ}, where Ω_σ = U^- ×_S T_σ ×_S U^+ is the big-cell toric scheme and A_σ is the rational action of G×G on Ω_σ extending the group law on the open cell G. Quoting rational-action theorems, the quotient is shown to be a quasi-projective algebraic space over S and, étale locally on the base, a scheme; gluing these along face inclusions yields X_Σ. The geometric fiber over an algebraically closed field k is identified with the classical toroidal embedding G_{k,σ} by matching the big-cell open subschemes and the G×G orbits. The paper also proves that X_Σ is smooth exactly when every cone is generated by a subset of a basis of the cocharacter lattice, and proper exactly when the Weyl-group saturation of the fan is complete.","pith_inferences":["If the quoted rational-action theorems hold in the generality assumed, the same quotient-sheaf recipe should produce integral models for other equivariant embeddings with a big-cell structure, such as spherical varieties or symmetric-space compactifications.","The sheaf-theoretic construction may still work even when T_σ is not smooth over Z, since it never uses a group scheme structure on Ω_σ; concrete rank-two examples with non-smooth cones would test this directly.","One can compare these Z-models with the canonical-basis models from a recent preprint for affine and projective embeddings; where both exist, they should agree on the big cell, which would give an independent check of the fibral identification.","A direct calculation of the definition domain of A_σ in the SL_2 case would give an explicit local picture of X_σ and clarify how the rational-action axioms from the companion paper are used."],"forward_implications":["For every fan Σ in the negative Weyl chamber there is now a single flat-scheme model over Spec(Z) whose geometric fibers are the classical toroidal embeddings, so the combinatorial classification holds uniformly across all characteristics.","Each single-cone embedding X_σ is quasi-projective over the base, so the universal models are scheme-theoretically reasonable, not merely algebraic spaces.","Smoothness and properness of X_Σ can be read off the fan: smooth iff every cone is generated by a subset of a basis of X_*(T), proper iff W·Σ is complete.","The construction is uniform for arbitrary fans and does not require the embedding to be affine or projective, unlike previous integral models built from canonical bases.","The quotient-sheaf description supplies a functoriality X_{σ1} → X_σ for cone inclusions, so the whole fan is glued from compatible local models."],"supporting_citations":[{"why":"Provides the rational-action theorems (quoted here as Theorems 3.6 and 3.7) that make the fppf quotient sheaf an algebraic space.","marker":"[Li25]"},{"why":"Supplies the classical classification of toroidal embeddings and the big-cell structure used to identify the geometric fibers.","marker":"[BK05]"},{"why":"Supplies the Chevalley group scheme, root subgroups, and Chevalley system used to define the rational action A_σ.","marker":"[SGA 3III]"},{"why":"Supplies the toric-variety facts about dual cones, closed orbits, and face open immersions used throughout the paper.","marker":"[KKMS73]"},{"why":"Provides the flatness and section-density lemmas used to control definition domains and quasi-projectivity.","marker":"[BLR90]"},{"why":"Gives the smoothness and completeness criteria for toric varieties used in Proposition 5.6.","marker":"[Oda88]"},{"why":"Contains the earlier wonderful-compactification integral model used to deduce quasi-projectivity of X_σ.","marker":"[Li23]"},{"why":"Gives the formalism of rational morphisms and the quotient representability criterion behind Construction 4.6.","marker":"[SGA 3II]"}],"fun_headline_variants":["Universal toroidal embeddings of Chevalley groups over Z","Toroidal embeddings for reductive schemes, now over Z","Chevalley groups get Z-relative toroidal embeddings","Lifting toroidal embeddings to a single scheme over Z","Base-change complete: toroidal embeddings over Z"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The construction depends on the quoted rational-action theorems applying to the toric big-cell scheme Ω_σ, which is not a group scheme, and that application is not verified in this paper.","fun_headline_variants_meta":{"raw":{"variants":["Universal toroidal embeddings of Chevalley groups over Z","Toroidal embeddings for reductive schemes, now over Z","Chevalley groups get Z-relative toroidal embeddings","Lifting toroidal embeddings to a single scheme over Z","Base-change complete: toroidal embeddings over Z"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000628,"raw_usage":{"total_tokens":2862,"prompt_tokens":864,"completion_tokens":1998,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":480,"completion_tokens_details":{"reasoning_tokens":1919}},"tokens_in":480,"tokens_out":1998,"duration_ms":12748,"temperature":1.0,"reasoning_tokens":1919,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T11:22:57.008422+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the definition domain of A_σ for a non-smooth cone σ over Spec(Z), for instance the cone generated by (1,1) in a rank-two cocharacter lattice; if e × Ω_σ × e is not contained in it, or if the equivalence relation on G × Ω_σ × G is not an fppf equivalence relation with representable diagonal, then X_σ is not an algebraic space and Theorem 1.1 fails.","supporting_citations":[],"review_version":1}