{"id":"de939172-fed5-42e3-94fe-f8bf581b5721","arxiv_id":"2505.01173","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Every affine embedding of a symmetric space admits a dual canonical basis, and the canonical embedding admits an integral model over Z.","lead":"This paper proves that every affine embedding of a symmetric space carries a dual canonical basis and that the canonical embedding has an integral model over the integers. It is significant because it gives uniform, characteristic-independent constructions for compactifications of symmetric spaces.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Integral ring structure of R(L) in Definition 3.18 is asserted, not proved; if the Z-level filtration is not multiplicative, Theorem 2's basis construction collapses.","rationale":"The reader's weakest_assumption was Proposition 2.7, but I find a more immediate threat to the headline claim at Definition 3.18. Theorem 2's R(L) is a union of filtration pieces of O(G/K), and a union of submodules is not automatically a subring; the missing integral filtration property is exactly what would supply the ring structure. The property is stated only after specialization to a field in Theorem 4(4), so the paper as written does not establish the ring structure needed for Spec R(L). This is not a disagreement with the literature—the property is very likely true and may already be in [3]—but it is a load-bearing assertion that should be stated and proved or cited. The same applies to finite generation of R(L) over Z, without which V(L) is not an integral model. I therefore keep the CONDITIONAL status: the construction is plausible and the gap is likely repairable, but the central claim should not be accepted until the integral filtration/multiplicativity is supplied. This is a partial agreement with the reader: the reader's weakest_assumption concerned the valuation cone, but their rationale already flagged the integral ring structure as a main weakness.","tokens_in":23001,"tokens_out":25310,"duration_ms":259928,"concrete_test":"Locate in [3] (or prove directly from the Z[q,q^{-1}]-filtration of O_q(G/K)) the integral filtration inequality O(G/K)_{≤μ'} · O(G/K)_{≤μ''} ⊂ O(G/K)_{≤μ'+μ''} for all μ',μ'' ∈ ˘X+. If it is absent, compute in a minimal example such as (G,K) = (SL_2, SO_2) with L generated by the fundamental spherical weight: take two dual canonical basis elements in O(G/K)_{≤1}; if their product contains any basis element outside B_{≤2}, then R(L) is not a ring and Definition 3.18 fails. A positive verification of the inequality would settle that the concern does not land.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central construction of Theorem 2 is Definition 3.18: R(L) = ∪_{μ∈L} O(G/K)_{≤μ} is declared to be a commutative Z-algebra, V(L) = Spec R(L) is declared an integral model, and B(L) = B(G/K) ∩ R(L) is declared a Z-basis. This requires that the union is closed under multiplication. The only filtration property stated in the paper is Theorem 4(4), and it is formulated after base change to an algebraically closed field: k[G_k/K_k]_{≤μ'} · k[G_k/K_k]_{≤μ''} ⊂ k[G_k/K_k]_{≤μ'+μ''}. The integral version O(G/K)_{≤μ'} · O(G/K)_{≤μ''} ⊂ O(G/K)_{≤μ'+μ''} is never stated or proved. Without it, R(L) is just a Z-submodule spanned by a subset of the dual canonical basis; it is not known to be a subring, so Spec R(L) is not defined and Theorem 3.19 has no content. The text also does not show that R(L) is finitely generated over Z, which is needed for an integral model. This gap is more load-bearing than Proposition 2.7: even if the affine-embedding classification were granted, Theorem 2 would still not supply a dual canonical basis for an arbitrary affine embedding unless the integral algebra structure of R(L) is established. The authors may intend this to follow from [3], but the necessary statement is not included among the properties in Theorem 4.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies affine embeddings of symmetric spaces G_k/K_k over algebraically closed fields of characteristic not 2. Theorem 1 (Theorem 3.3) classifies affine embeddings by closed saturated submonoids L of the spherical dominant weight lattice, via V_k(L)=Spec R_k(L) with R_k(L)=∪_{μ∈L} k[G_k/K_k]_{≤ μ}. Theorem 2 (Definition 3.18, Theorem 3.19, Definition 3.20) claims an integral model R(L) over Z, a dual canonical basis B(L), and a quantum analog R_q(L) for every such embedding. Theorem 3 (Theorem 4.11, Definition 4.13) constructs a projective scheme over Z whose geometric fibers are the canonical embedding, and in the adjoint case the wonderful compactification, together with a local structure theorem and a smoothness criterion. The proofs combine the dual canonical basis formalism of the authors' earlier paper [3] with spherical embedding theory, including a valuation-cone computation in arbitrary characteristic (Proposition 2.7).","tokens_in":23275,"tokens_out":11655,"duration_ms":131479,"significance":"If the central construction is sound, the paper is significant: it extends dual canonical bases from symmetric spaces to all their affine embeddings, provides characteristic-independent integral models, and gives a unified construction of the canonical and wonderful compactifications. The affine-embedding classification via saturated monoids is a natural and useful statement, and the orbit parametrization in Theorem 3.16 and the smoothness criterion in Corollary 4.16 are concrete contributions. The proof of the valuation cone in characteristic not 2 via the De Concini-Springer local structure theorem (Proposition 2.7) is also a genuine step beyond existing references. However, the central ring-theoretic construction in Section 3.3 is not fully justified as written, and Theorem 2 is conditional on that missing justification.","major_comments":[{"comment":"The definition of R(L) as a commutative Z-algebra is not justified. The paper defines R(L)=∪_{μ∈L} O(G/K)_{≤ μ} and then V(L)=Spec R(L), but this requires O(G/K)_{≤ μ'}·O(G/K)_{≤ μ''}⊂O(G/K)_{≤ μ'+μ''}. The only multiplicativity statement available, Theorem 4(4), is formulated after base change to an algebraically closed field k. The integral statement over Z is neither stated nor proved. Without it, R(L) is only a Z-submodule spanned by a subset of the dual canonical basis; equation (3.2) is not an isomorphism of k-algebras, and Theorem 3.19(1) does not follow. Since Theorem 2 is built entirely on this construction, this is a load-bearing gap.","section":"Section 3.3, Definition 3.18 and Theorem 4(4)"},{"comment":"Even if R(L) were known to be a subring, the paper does not prove that R(L) is finitely generated over Z, which is required for V(L) to be an integral model in the usual sense of a finite-type Z-scheme. Lemma 3.2 proves finite generation only after base change to an algebraically closed field, and the Z-basis B(L) is infinite; no argument shows that the filtered union stabilizes in finitely many degrees. The authors should either prove finite generation of R(L) over Z or give an explicit statement in [3] that implies it.","section":"Section 3.3, Definition 3.18 and Theorem 3.19"},{"comment":"The assertion 'It follows from Lemma 3.12 (by taking k=C) that R(J) is a prime ideal of R(L)' is not a consequence of Lemma 3.12 as written. Lemma 3.12 concerns prime ideals in the C-algebra R_C(J)⊂R_C(L); descending to Z requires at least that R(L) is a ring, that R(J) is an ideal, and a flatness or purity argument for primality. As written, the construction of V(L\\J) and hence Theorem 3.19(2) is not justified. This needs either a direct proof or a precise reference.","section":"Section 3.3, sentence after (3.2)"}],"minor_comments":[{"comment":"There are typos in the text: 'the discussion afetr Remark 2.1.1' should be 'after', and 'which is call the spherical root system' should be 'called'.","section":"Section 2.3"},{"comment":"The section heading 'Abelianzation' and the later use of 'abelization' should be corrected to 'Abelianization' for consistency.","section":"Section 3.4"},{"comment":"The notation 'χ1∼χ2' is used without definition; from the context it should mean χ1−χ2∈M0, but this should be stated explicitly.","section":"Proposition 3.25(3)"},{"comment":"The definition of P_λ(G/K) depends on a choice of λ, while Theorem 3 asserts existence of a projective scheme P over Z. Since independence of λ is only conjectured in Remark 4.14, the theorem should either fix one λ or state clearly that the construction is independent over Z; otherwise the object P is not uniquely defined.","section":"Definition 4.13 and Remark 4.14"},{"comment":"The displayed definition of eL_{J1,J2} is hard to read and appears to have a formatting issue in the second line; it should be rewritten so that the conditions on µ and γ are unambiguous.","section":"Proposition 4.9"}],"recommendation":"major_revision","confidential_remarks":"The main gap in Section 3.3 is central but appears repairable: the authors should either prove the integral filtration multiplicativity directly or extract it as a numbered property of the quantum algebra from [3]. I would not reject the paper on this basis, but the current Theorem 2 is unsubstantiated until this is supplied. The heavy reliance on [3] for unstated structural properties should also be made explicit in the revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"I read this through carefully. The genuine new content is real: extending the construction of dual canonical bases and integral models from the symmetric space itself to every affine embedding, plus an integral model for the canonical compactification. That is a substantial step forward for the theory of spherical varieties, and the authors get credit for the positive-characteristic classification of affine embeddings and for the orbit-closure results. The main proofs are detailed and the reliance on the prior quantization paper is mostly explicit.\n\nThe soft spot is exactly the one the stress-test note flags. Definition 3.18 declares R(L) to be a commutative ring, but the paper never proves the integral version of the filtration multiplication: O(G/K)_{≤μ} · O(G/K)_{≤ν} ⊂ O(G/K)_{≤μ+ν}. Theorem 4(4) only states it after base change to an algebraically closed field. Without that, R(L) is just a Z-submodule spanned by a subset of the dual canonical basis, not a subring, so Spec R(L) and the integral model in Theorem 2 have no content. Finite generation over Z is also not shown. I checked the surrounding text; there is no hidden proof. This is not a fatal mathematical flaw — the statement probably follows from the quantum filtration in [3], and any expert can supply it — but it is load-bearing and must be fixed before the paper is published.\n\nI am less worried about the other concern mentioned, Proposition 2.7 relying on De Concini–Springer. The restriction to the adjoint case and the valuation computation look sound; that part holds together. Corollary 3.8 and the orbit-closure bijections are nice.\n\nBottom line: this paper deserves a serious referee. The gap should be written up as a request for a proof of the integral filtration property, or a precise reference to where it is proved in [3], plus a finite-generation argument for R(L). Anyone working on symmetric spaces or spherical embeddings gets real value from this, even as a preprint.","headline":"A serious paper with a real gap in the integral ring structure that is likely fixable; deserves peer review.","tokens_in":23795,"tokens_out":2548,"would_cite":true,"duration_ms":28432,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14M27","17B37","14L30","20G15"],"pacs":[],"model":"deepseek-v4-flash","headline":"The coordinate ring of any affine embedding of a symmetric space admits a dual canonical basis.","keywords":["symmetric spaces","affine embeddings","dual canonical basis","integral models","spherical varieties","wonderful compactification","quantum symmetric pairs","valuation cone"],"falsifier":"Exhibit an affine embedding over a field of characteristic $p\\neq 2$ whose associated monoid $L(V_k)$ is not closed and saturated, or construct a $G_k$-invariant valuation $v$ on $k(G_k/K_k)$ with $v(\\chi_{\\alpha_i})>0$ for some spherical root $\\alpha_i$; either would contradict Proposition 2.7 and hence Theorem 3.3.","tokens_in":22773,"feed_emoji":"🧮","tokens_out":5518,"duration_ms":54595,"temperature":0.7,"pith_summary":"This paper proves that every affine embedding of a symmetric space $G_k/K_k$ over an algebraically closed field of characteristic not 2 is controlled by a finitely generated saturated submonoid $L$ of spherical dominant weights, and that the coordinate ring of the embedding has a dual canonical basis. It constructs a commutative ring $R(L)$ over $\\mathbb{Z}$ whose geometric fibres are precisely the given affine embeddings, together with a quantum deformation $R_q(L)$ over $\\mathbb{Z}[q,q^{-1}]$. It then builds an integral model for the canonical embedding of the symmetric space, and in the adjoint case for the wonderful compactification. If the construction is correct, the classification of affine embeddings and their canonical-basis structure are independent of the base field.","feed_headline":"Every affine embedding of a symmetric space has a canonical basis","feed_subtitle":"A uniform Z-model and quantum deformation cover all such embeddings, over any algebraically closed field of characteristic not 2.","key_machinery":"The central objects are saturated submonoids $L$ of the spherical weight lattice $\\breve{X}^+$: finitely generated submonoids that generate $\\breve{X}$ as a group and are closed under taking roots. They cut out subspaces $R_k(L)=\\bigcup_{\\mu\\in L} k[G_k/K_k]_{\\le \\mu}$ using the filtration by dual Weyl modules. The argument combines this with the integral model $O(G/K)$ and its dual canonical basis from the quantization $O_q(G/K)$, and with the enveloping variety $\\widetilde{V}_k$ defined by the monoid $\\widetilde{L}=\\{(\\mu,\\lambda):\\mu\\preceq\\lambda\\}$. The canonical embedding is obtained as a GIT quotient of the enveloping variety by a torus. The local structure theorem for the wonderful compactification supplies the valuation-cone description that underlies the proof that the monoid associated to any affine embedding is closed and saturated.","core_discovery":"For a connected reductive group $G_k$ with an involution $\\theta_k$ and fixed-point subgroup $K_k$, the paper's central claim is that the affine embedding theory of $G_k/K_k$ is captured by a monoid $L$ of spherical dominant weights. The subspace $R_k(L)=\\bigcup_{\\mu\\in L} k[G_k/K_k]_{\\le \\mu}$ is a normal $G_k$-subalgebra, and the assignment $L\\mapsto \\operatorname{Spec} R_k(L)$ is a bijection onto all affine embeddings up to isomorphism. Moreover, the coordinate ring of every such embedding carries a basis $B(L)$ that specializes to a basis of $k[V_k(L)]$ over any algebraically closed field of characteristic not 2, and this basis arises from a non-commutative $\\mathbb{Z}[q,q^{-1}]$-algebra $R_q(L)$. For the canonical embedding, a projective scheme over $\\mathbb{Z}$ is constructed whose geometric fibres are the canonical embeddings; when $G_k$ is of adjoint type, these fibres are the wonderful compactifications.","pith_inferences":["The same monoidal and filtration machinery could plausibly yield canonical bases for all spherical embeddings, not only affine ones; the paper establishes the affine case, and the toroidal/local-structure ingredients suggest a path for the general case.","The smoothness criterion for the canonical embedding, expressed in terms of spherical roots, could be tested computationally on small-rank symmetric spaces in positive characteristic to see exactly when smooth compactifications exist over $\\mathbb{Z}$.","The quantization $R_q(L)$ invites a categorification through quantum symmetric pair canonical bases; the paper does not claim this, but the structure it builds is the natural setting for such a construction.","Because the affine-embedding classification is base-field independent, one could check the normality and orbit-closure behavior of $V_k(L)$ over $\\mathbb{F}_p$ for small examples; the paper predicts the same monoid governs all characteristics."],"forward_implications":["Every affine embedding of a symmetric space is obtained as $\\operatorname{Spec} R_k(L)$ for a unique closed saturated submonoid $L$ of $\\breve{X}^+$.","The coordinate ring of every affine embedding admits a $\\mathbb{Z}$-basis that specializes to a dual canonical basis in any characteristic not 2, and admits a quantum deformation over $\\mathbb{Z}[q,q^{-1}]$.","Orbit closures on affine embeddings are normal, have good filtrations as $G_k$-modules, and are defined over $\\mathbb{Z}$.","For semisimple $G_k$, all affine embeddings of $G_k/K_k$ are trivial, so interesting affine embeddings arise only from reductive groups with a nontrivial center.","The canonical embedding of a symmetric space, and the wonderful compactification in the adjoint case, have integral models over $\\mathbb{Z}$."],"supporting_citations":[{"why":"Supplies the good filtration of $k[G_k/K_k]$, the dual canonical basis $B_q(G/K)$, and the base-change identification of $O(G/K)$ with coordinate rings.","marker":"[3]"},{"why":"Provides the local structure theorem for the wonderful compactification that identifies the valuation cone in arbitrary characteristic.","marker":"[8]"},{"why":"Supplies the spherical-embedding framework: colored cones, the valuation-cone containment, and uniqueness of the canonical embedding.","marker":"[14]"},{"why":"Gives the classification of involutions and symmetric spaces independently of the base field for characteristic not 2.","marker":"[20]"},{"why":"Gives the criterion that a good filtration together with normal unipotent-invariants implies normality of the algebra.","marker":"[10]"},{"why":"Used to show that saturated monoid algebras are integrally closed.","marker":"[5]"},{"why":"Provides the abelianization and enveloping-variety framework in the group case that the present construction generalizes.","marker":"[24]"},{"why":"Defines the canonical embedding as the unique complete simple toroidal embedding and records its existence.","marker":"[9]"}],"fun_headline_variants":["Dual canonical bases for every affine symmetric space embedding","Canonical basis universal for affine embeddings of symmetric spaces","Quantum deformation yields canonical bases for symmetric embeddings","Integral models cover all affine symmetric space embeddings","Wonderful compactification fits into dual canonical basis framework"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole classification rests on the claim that the cone of $G_k$-invariant valuations is exactly $\\{t\\in \\mathbb{Q}\\breve{X}^* : t(\\alpha_i)\\le 0 \\text{ for } i\\in I'_\\circ\\}$ in every characteristic not 2; this is proved using the local structure theorem for the wonderful compactification, so if that theorem fails or the reduction to the adjoint case breaks, the closure argument for $L(V_k)$ collapses.","fun_headline_variants_meta":{"raw":{"variants":["Dual canonical bases for every affine symmetric space embedding","Canonical basis universal for affine embeddings of symmetric spaces","Quantum deformation yields canonical bases for symmetric embeddings","Integral models cover all affine symmetric space embeddings","Wonderful compactification fits into dual canonical basis framework"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000366,"raw_usage":{"total_tokens":1939,"prompt_tokens":888,"completion_tokens":1051,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":504,"completion_tokens_details":{"reasoning_tokens":977}},"tokens_in":504,"tokens_out":1051,"duration_ms":10350,"temperature":1.0,"reasoning_tokens":977,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T04:24:34.928742+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibit an affine embedding over a field of characteristic $p\\neq 2$ whose associated monoid $L(V_k)$ is not closed and saturated, or construct a $G_k$-invariant valuation $v$ on $k(G_k/K_k)$ with $v(\\chi_{\\alpha_i})>0$ for some spherical root $\\alpha_i$; either would contradict Proposition 2.7 and hence Theorem 3.3.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the spherical-embedding framework: colored cones, the valuation-cone containment, and uniqueness of the canonical embedding."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the classification of involutions and symmetric spaces independently of the base field for characteristic not 2."},{"cited_title":"Grosshans, Contractions of the actions of reductive algebraic groups in arbi- trary characteristic, Invent","cited_arxiv_id":null,"evidence_quote":"Gives the criterion that a good filtration together with normal unipotent-invariants implies normality of the algebra."},{"cited_title":"Cox, John B","cited_arxiv_id":null,"evidence_quote":"Used to show that saturated monoid algebras are integrally closed."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the abelianization and enveloping-variety framework in the group case that the present construction generalizes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the canonical embedding as the unique complete simple toroidal embedding and records its existence."}],"review_version":1}