{"id":"e626c5e7-2d55-46e4-88e1-b6f072c68fec","arxiv_id":"2507.18453","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A new class of Weyl group elements, geometric Coxeter type, is shown to decompose affine Deligne-Lusztig varieties into classical Deligne-Lusztig varieties times affine spaces and tori.","lead":"The authors prove a general criterion showing that certain affine Deligne-Lusztig varieties, which appear in the geometry of Shimura varieties, decompose into simple building blocks. They define a new class of Weyl group elements, geometric Coxeter type, and show that the decomposition holds for this class.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The trivialization in Propositions 3.5/3.6 relies on an unproved algebraicity assertion for the coordinate map φ; pointwise uniqueness does not yet give a morphism of F_q-schemes, so Theorems 4.1 and 5.7 inherit the gap.","rationale":"The reader's weakest assumption identifies exactly the point where the paper's main engine is least secure. The proof of Theorem 4.1 is only the sentence 'Using induction, the theorem follows from Proposition 3.4, 3.5 and 3.6.' Those propositions are where the trivialization is constructed, and the only verification offered for the key coordinate φ is the phrase 'straightforward to check'. Pointwise well-definedness of a coordinate does not ensure that it is a scheme morphism; without that, the claimed universal homeomorphism to a product is not established. Proposition 3.6 has the same structure, and the same issue affects the root-coordinate maps in Proposition 3.4. I do not see a contradiction with existing results, and the missing step is likely repairable by a standard Bruhat-decomposition argument, so the correct status remains conditional rather than rejection. I also noted the assertion in Section 4 that the strong multiplicity one property and the numbers ℓ_I(p), ℓ_II(p) are independent of the reduction tree is deferred to 'one can prove'; this is a secondary gap, but the algebraicity of the trivialization maps is the primary load-bearing concern. Since the reader already assigned CONDITIONAL for essentially this reason, my read does not change the verdict.","tokens_in":19187,"tokens_out":9108,"duration_ms":99301,"concrete_test":"Re-derive Proposition 3.5 with an explicit lemma: for x ∈ I s˙ w˙ I, the unique t with x u_{σ(a)}(t)σ(s˙) ∈ I s w σ(s) I is given by an algebraic map I s˙ w˙ I → A^1, and then prove φ is the composition of this map with the morphism h ↦ ψ(h)^{-1}bσ(ψ(h)). Concretely, work out the smallest nontrivial case, split GL_2 (or SL_2) with Iwahori I, b = 1, and w = s_0 s_1: compute φ in explicit Iwahori coordinates on X_{s_0 s_1}(1) and check directly that γ and γ′ are inverse morphisms of F_q-schemes. If the coordinate map is algebraic, add it as a lemma; if a counterexample appears, Theorem 4.1 fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Proposition 3.5 (Section 3.3) defines φ(ψ(h)) as the unique element of A^1 satisfying ψ(h)^{-1}bσ(ψ(h))u_{σ(a)}(φ(ψ(h)))σ(s˙) ∈ I s w σ(s) I. The inverse map γ′ then uses x + φ(ψ(h)), so the entire Gm-trivialization is built from φ. The proof only says it is 'straightforward to check' that γ and γ′ are well-defined and inverse; it never proves that φ is a morphism of F_q-schemes. Pointwise uniqueness for each F_q-point is a set-theoretic statement, not an algebraic one. The same gap appears in Proposition 3.6 for the A^1-coordinate, and in the analogous root-subgroup coordinates in Proposition 3.4. Since Theorem 4.1 is proved by concatenating Propositions 3.4–3.6, and Theorem 5.7 depends on Theorem 4.1, the product decomposition is not fully established unless these coordinate maps are algebraic. This is a missing verification in a load-bearing step, not a known contradiction; it may well be repairable, but as written the proof is incomplete.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies affine Deligne-Lusztig varieties X_w(b) in affine flag varieties and aims to decompose them geometrically as products of classical Deligne-Lusztig varieties with multiplicative groups and affine spaces. To this end, the authors introduce a class of Iwahori-Weyl group elements called \"geometric Coxeter type,\" which strictly contains the previously studied positive Coxeter type and finite Coxeter type elements. The main result, Theorem 4.1, asserts that if a reduction tree of w has a unique path ending in [b] and the corresponding endpoint affine Deligne-Lusztig variety has a lift, then X_w(b) is universally homeomorphic to the trivial fiber bundle X_end(p)(b) × (G_m)^{ℓ_I(p)} × (A^1)^{ℓ_II(p)}. Theorem 5.7 then shows that for w of geometric Coxeter type, all irreducible components of X_w(b) lie in one J_b(F)-orbit and each component is universally homeomorphic to such a product with a classical Deligne-Lusztig variety X'. Section 6 establishes purity and saturation of Newton stratifications, dimension formulas, and explicit counts of type I and type II reduction steps for geometric Coxeter type elements.","tokens_in":19451,"tokens_out":14568,"duration_ms":139176,"significance":"If the missing algebraicity verifications are supplied, the paper gives a significant uniform framework that unifies and extends earlier work on positive Coxeter type and elements of the form t^μ c. The main theorems are powerful: Theorem 5.7(2) is new even for elements of the form t^μ c, and the class of geometric Coxeter type elements is demonstrably broader than previously considered classes. The paper also provides concrete dimension formulas and path-length counts, giving the statements concrete, checkable content. The systematic use of liftings in the Deligne-Lusztig reduction is an elegant and promising method. However, several load-bearing arguments are only sketched at the level of pointwise bijections and do not establish the required scheme-theoretic algebraicity; these gaps affect the central product decomposition and must be fixed before the main claims are fully proven.","major_comments":[{"comment":"The trivializations in Propositions 3.5 and 3.6 are only verified at the level of F_q-points. The element φ(\\tilde h) in the displayed equation before Proposition 3.5 is defined by a uniqueness condition on the value in A^1; uniqueness for each F_q-point does not imply that φ, or the coordinates y and x used in the maps γ and γ′, are morphisms of F_q-schemes. The text states that it is 'straightforward to check' that γ and γ′ are well-defined and inverse, but this does not address algebraicity. Since Theorem 4.1 is proved by concatenating Propositions 3.4–3.6, the product decomposition in Theorem 4.1 and hence in Theorem 5.7 depends on this missing verification.","section":"Section 3.3, Propositions 3.5 and 3.6"},{"comment":"The same scheme-theoretic gap appears earlier in Proposition 3.4 and in Lemma 3.3. In Lemma 3.3, the map p_a sending g to g_1 in the decomposition I \\dot w I/I ≅ U_a × (I \\dot v I/I) is asserted to be algebraic without proof. In Proposition 3.4, the element z in the root subgroup U_a, used to define the lift Ψ, is defined by a uniqueness condition; its algebraicity as a function on X_w(b) is not established. These are not merely presentation issues, because a lift must be a morphism of (perfect) schemes to be used in Theorem 4.1.","section":"Section 3.2 and Lemma 3.3"},{"comment":"The construction of the lift ψ for minimal Coxeter type elements relies on the bijection in (5.1), cited from the proof of [11, Theorem 4.8], but the text only states it as a bijection on F_q-points. The subsequent definition of ψ by choosing representatives for the cosets J_b(F)/J_b(F)∩K produces a set-theoretic lift; it is not shown that the resulting map is a morphism of perfect schemes, as required by Definition 3.1 and condition (2) of Theorem 4.1. This is another load-bearing algebraicity gap.","section":"Section 5.2"}],"minor_comments":[{"comment":"After Theorem 4.1, the sentence 'the number ℓ_I(p) and ℓ_I(p) in Theorem 4.1 is independent' should read 'ℓ_I(p) and ℓ_II(p)'.","section":"Section 4"},{"comment":"Reference [27] lists the author as 'Takamstsu'; the correct spelling is 'Takamatsu'.","section":"References"},{"comment":"The abstract contains a typo: 'varities' should be 'varieties'.","section":"Abstract"},{"comment":"The parenthetical 'some (or any)' asserts independence of the reduction tree for both conditions without proof; please either prove the independence or explicitly state which direction the definition uses.","section":"Definition 5.5"},{"comment":"The notation is inconsistent: 'ν_b' and 'b_max' appear where 'ν(b)' and 'b_{w,max}' are used elsewhere; the notation should be unified.","section":"Section 6.1"}],"recommendation":"major_revision","confidential_remarks":"The paper's novelty relative to the authors' previous work is substantial, and the overlap is appropriately acknowledged. The main gap is a missing algebraicity check that appears to be repairable with standard techniques from the theory of ind-schemes and Iwahori decompositions. I would not recommend rejection, but the missing verifications must be supplied before the central claims can be considered proven."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know: this paper is a real step forward on the affine Deligne-Lusztig side of the story. It introduces a class of Weyl group elements (geometric Coxeter type) that provably contains the earlier positive and finite Coxeter type classes, and gives a uniform criterion (Theorem 4.1) under which an ADLV becomes a trivial bundle over a classical Deligne-Lusztig variety with affine-space and torus fibers. The consequences in Theorem 5.7—single J_b(F)-orbit, explicit product decompositions—are new even for t^μ c elements, and Section 6 gives clean formulas for path lengths and dimensions. That is substantial.\n\nThe main engine is the lifting construction in Section 3. The authors write down explicit formulas for the lifts and for the trivializations. That is the right way to do it, and the induction in Section 4 is natural.\n\nThe soft spot is where the stress-test lands. In Propositions 3.5 and 3.6, the map φ (and y) is defined as the unique element of A^1 or G_m satisfying a certain containment. The text says it is 'straightforward to check' that γ and γ′ are well-defined and inverse. What is missing is the proof that φ is a morphism of F_q-schemes, not just a set-theoretic assignment on points. The whole trivial-bundle conclusion depends on that algebraicity. Without it, Theorem 4.1 and hence Theorem 5.7 inherit an unproved step. I do not see a contradiction or a known counterexample; it looks repairable, but as written the proof is incomplete.\n\nThere is a second, smaller gap: independence of the reduction tree for both the strong multiplicity one property and the numbers ℓ_I, ℓ_II is asserted with a citation to [14, Theorem 6.7], not shown here. That is probably fine, but it is load-bearing for the definition of geometric Coxeter type and for the dimension formulas in Section 6.\n\nThe paper is honest about what is new and what is imported. The citation pattern to prior work by the same authors (especially [15] and [24]) is appropriate; the new class genuinely extends those results, and the example in type C2 shows the classical part need not even be a σ-conjugate of a partial Coxeter element. The treatment of equal and mixed characteristic in parallel is careful.\n\nWho is this for? Anyone working on affine Deligne-Lusztig varieties, Rapoport-Zink spaces, or the EO stratification. It deserves a serious referee who can check the algebraicity of the coordinate maps and the reduction-tree independence. My recommendation: send it out, but require the authors to fill the gap in Propositions 3.5/3.6 before acceptance.","headline":"Worth serious refereeing: the geometric Coxeter type class and the lifting criterion are real contributions, but the main proof has a repairable gap around algebraicity of the trivializing maps.","tokens_in":19945,"tokens_out":4354,"would_cite":true,"duration_ms":41995,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["20G25","14G35","14M15","14L30"],"pacs":[],"model":"deepseek-v4-flash","headline":"For Weyl-group elements of geometric Coxeter type, every nonempty affine Deligne-Lusztig variety is a single orbit of components, each a product of a classical Deligne-Lusztig variety with affine and pointed-affine spaces.","keywords":["affine Deligne-Lusztig varieties","Deligne-Lusztig reduction","geometric Coxeter type","lifting","loop groups","Newton stratification","Shimura varieties","affine flag varieties"],"falsifier":"In the type $C_2$ example $w = s_1\\tau_2$ from Example 5.4(4), write the maps $\\gamma$ and $\\gamma'$ of Propositions 3.5 and 3.6 in explicit coordinates on the affine Schubert cell and check whether $\\varphi(\\tilde h)$ is given by regular functions over $\\mathbb{F}_q$; if it is not a morphism of schemes, the asserted trivial-bundle decomposition fails in that case.","tokens_in":19011,"feed_emoji":"🧩","tokens_out":17203,"duration_ms":157737,"temperature":0.7,"pith_summary":"This paper gives a uniform way to dissect affine Deligne-Lusztig varieties—schemes that encode relative positions in loop groups and appear in the reduction of Shimura varieties—into a product of a classical Deligne-Lusztig variety with copies of the affine line $\\mathbb{A}^1$ and of the multiplicative group $\\mathbb{G}_m$, the affine line with the origin removed. It introduces a class of elements, called geometric Coxeter type, and shows that for these elements every nonempty $X_w(b)$ has a single $J_b(F)$-orbit of irreducible components, each component being such a product. This covers and extends previously studied cases, including elements of positive Coxeter type and elements of the form $t^\\mu c$. The method works by lifting the variety into the loop group and showing that the $\\mathbb{G}_m$- and $\\mathbb{A}^1$-fibrations produced by the Deligne-Lusztig reduction are actually trivial.","feed_headline":"Geometric Coxeter type decomposes affine Deligne-Lusztig varieties","feed_subtitle":"For elements of geometric Coxeter type, every nonempty affine Deligne-Lusztig variety is a classical Deligne-Lusztig variety times affine…","key_machinery":"The technical engine is the lift: a morphism $\\psi$ from a subvariety $Y$ of the affine flag variety $G(L)/I$ to the loop group $G(L)$ whose composition with the projection is the identity on $Y$. Propositions 3.4–3.6 show that lifts propagate across the steps of the Deligne-Lusztig reduction, and that when the smaller variety is liftable, the $\\mathbb{G}_m$- and $\\mathbb{A}^1$-bundles appearing in the reduction are trivial. The second ingredient is the reduction tree of an affine Weyl group element, whose edges record the two ways length can drop: a type I edge cuts length by one and contributes a $\\mathbb{G}_m$-factor, while a type II edge cuts length by two and contributes an $\\mathbb{A}^1$-factor. Geometric Coxeter type is the class of elements for which every end vertex of the tree is a cyclic shift of a product $ux$ with $x$ $\\sigma$-straight and $u$ a twisted Coxeter element of the corresponding finite Weyl group; this form is what makes the endpoints liftable.","core_discovery":"The central claim is Theorem 5.7: if $w$ has geometric Coxeter type and $X_w(b)$ is nonempty, then all irreducible components of $X_w(b)$ lie in a single $J_b(F)$-orbit, and each component is universally homeomorphic to $X' \\times \\mathbb{G}_m^{\\ell_I(p)} \\times \\mathbb{A}^{1,\\ell_{II}(p)}$, where $X'$ is a classical Deligne-Lusztig variety of Coxeter type and $\\ell_I(p)$, $\\ell_{II}(p)$ count the two types of reduction steps along the unique reduction path attached to $[b]$. The engine is Theorem 4.1: whenever a reduction tree of $w$ has exactly one path ending in the $\\sigma$-conjugacy class of $b$ and the end variety $X_{\\mathrm{end}(p)}(b)$ admits a lift, $X_w(b)$ is a trivial fiber bundle over $X_{\\mathrm{end}(p)}(b)$ with $\\mathbb{G}_m$- and $\\mathbb{A}^1$-fibres. Geometric Coxeter type is designed so that both hypotheses hold: strong multiplicity one, meaning each $[b]$ is reached by a unique path, gives the uniqueness, and endpoints of minimal Coxeter type are liftable because their irreducible components sit inside affine Schubert cells. For these $w$ the paper also proves that the Newton stratification inside the double coset $I\\dot{w}I$ is saturated, gives explicit formulas for $\\ell_I$ and $\\ell_{II}$ in terms of $w$ and $b$, and derives a closed dimension formula.","pith_inferences":["The hypotheses actually used in Theorem 4.1 are only strong multiplicity one and liftability of the end variety, so the same trivial-bundle conclusion should hold for any class of elements satisfying those two conditions, even when the endpoints are not of Coxeter type.","Because universal homeomorphisms do not change étale cohomology, the theorem reduces the cohomology of $X_w(b)$ for geometric Coxeter type elements to that of classical Deligne-Lusztig varieties, whose cohomology is already controlled by the representation theory of finite groups of Lie type.","The explicit formulas for $\\ell_I(p)$ and $\\ell_{II}(p)$ make a point-counting check available: in small-rank examples, counting $\\mathbb{F}_{q^r}$-points on the product decomposition should reproduce the point count of $X_w(b)$ and could be verified computationally.","The lifting technique is also a natural tool for semi-infinite Deligne-Lusztig varieties, whose deep-level truncations are already known to be such products in special cases; extending Theorem 5.7 through inverse limits is a plausible next step."],"forward_implications":["For every geometric Coxeter type element, the full component-level geometry of $X_w(b)$ is now known: one orbit of irreducible components and an explicit product decomposition, going beyond the earlier dimension formulas and orbit counts.","For the previously studied elements of the form $t^\\mu c$, the component-level statement is new: each irreducible component is a product of a classical Deligne-Lusztig variety with affine and pointed-affine spaces, even though the class itself was already understood coarsely.","Since every Shimura datum contributes at least one Ekedahl-Oort stratum of the form $t^\\mu c$, the theorem gives an explicit geometric description of that stratum in the special fibre of the associated Rapoport-Zink space.","For these $w$, the Newton stratification of $I\\dot{w}I$ is saturated, the closure of each Newton stratum is a union of lower strata, and the dimension of $X_w(b)$ has the closed formula $\\frac12(\\ell(w)+\\ell_{R,\\sigma}(\\mathrm{cl}\\,w)-\\langle\\nu(b),2\\rho\\rangle-\\mathrm{def}(b))$."],"supporting_citations":[{"why":"It constructs the classical Deligne-Lusztig varieties that appear as the base factor $X'$ in the product decomposition.","marker":"[4]"},{"why":"It supplies the Deligne-Lusztig reduction that subdivides $X_w(b)$ into $\\mathbb{G}_m$- and $\\mathbb{A}^1$-fibrations over smaller varieties.","marker":"[6]"},{"why":"It provides the reduction-tree machinery and the $\\sigma$-straight decomposition theorem used to define paths and endpoints.","marker":"[14]"},{"why":"It studies elements of the form $t^\\mu c$ and proves they have geometric Coxeter type, a class the new component-level description extends.","marker":"[15]"},{"why":"It introduces positive Coxeter type elements and their lifting construction, which geometric Coxeter type generalizes.","marker":"[24]"},{"why":"It gives the isomorphism between $X_w^K(b)$ and a classical Deligne-Lusztig variety for minimal Coxeter type endpoints, producing the needed lifts.","marker":"[11]"},{"why":"It shows the classical Deligne-Lusztig variety of Coxeter type lies inside an affine Schubert cell, so it inherits a lift.","marker":"[18]"},{"why":"It gives the essential-gap dimension inequalities and purity criteria used in Section 6 for the Newton stratification.","marker":"[16]"}],"fun_headline_variants":["Geometric Coxeter type splits affine Deligne-Lusztig varieties","Geometric Coxeter elements yield classical times affine Deligne-Lusztig","New Coxeter class gives product decomposition for affine Deligne-Lusztig","Lifting reduction gives geometric Coxeter type product structure","Geometric Coxeter type makes affine Deligne-Lusztig classical times affine"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument depends on the unproved claim that the parameter $\\varphi(\\tilde h)$, defined pointwise by a coset equation in the loop group, is an algebraic (regular) function on the relevant scheme; the paper says this is \"straightforward to check,\" but if $\\varphi$ were only a map on $\\mathbb{F}_q$-points and not algebraic, the trivial-bundle conclusion of Propositions 3.5 and 3.6, and hence of Theorem 4.1, would not follow.","fun_headline_variants_meta":{"raw":{"variants":["Geometric Coxeter type splits affine Deligne-Lusztig varieties","Geometric Coxeter elements yield classical times affine Deligne-Lusztig","New Coxeter class gives product decomposition for affine Deligne-Lusztig","Lifting reduction gives geometric Coxeter type product structure","Geometric Coxeter type makes affine Deligne-Lusztig classical times affine"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001118,"raw_usage":{"total_tokens":4736,"prompt_tokens":1113,"completion_tokens":3623,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":729,"completion_tokens_details":{"reasoning_tokens":3531}},"tokens_in":729,"tokens_out":3623,"duration_ms":28411,"temperature":1.0,"reasoning_tokens":3531,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T18:11:40.595273+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"In the type $C_2$ example $w = s_1\\tau_2$ from Example 5.4(4), write the maps $\\gamma$ and $\\gamma'$ of Propositions 3.5 and 3.6 in explicit coordinates on the affine Schubert cell and check whether $\\varphi(\\tilde h)$ is given by regular functions over $\\mathbb{F}_q$; if it is not a morphism of schemes, the asserted trivial-bundle decomposition fails in that case.","supporting_citations":[{"cited_title":"Representations of Reductive Groups Over Finite Fields","cited_arxiv_id":null,"evidence_quote":"It constructs the classical Deligne-Lusztig varieties that appear as the base factor $X'$ in the product decomposition."},{"cited_title":"Dimension of affine Deligne-Lusztig varieties in affine flag varieties","cited_arxiv_id":null,"evidence_quote":"It supplies the Deligne-Lusztig reduction that subdivides $X_w(b)$ into $\\mathbb{G}_m$- and $\\mathbb{A}^1$-fibrations over smaller varieties."},{"cited_title":"Minimal length elements of extended affine Weyl groups","cited_arxiv_id":null,"evidence_quote":"It provides the reduction-tree machinery and the $\\sigma$-straight decomposition theorem used to define paths and endpoints."},{"cited_title":"Schremmer, R","cited_arxiv_id":null,"evidence_quote":"It introduces positive Coxeter type elements and their lifting construction, which geometric Coxeter type generalizes."},{"cited_title":"Geometric and homological properties of affine Deligne-Lusztig vari- eties","cited_arxiv_id":null,"evidence_quote":"It gives the isomorphism between $X_w^K(b)$ and a classical Deligne-Lusztig variety for minimal Coxeter type endpoints, producing the needed lifts."},{"cited_title":"Coxeter orbits and eigenspaces of Frobenius","cited_arxiv_id":null,"evidence_quote":"It shows the classical Deligne-Lusztig variety of Coxeter type lies inside an affine Schubert cell, so it inherits a lift."},{"cited_title":"Zero-dimensional affine Deligne--Lusztig varieties","cited_arxiv_id":"2402.15310","evidence_quote":"It gives the essential-gap dimension inequalities and purity criteria used in Section 6 for the Newton stratification."}],"review_version":2}