{"id":"3fe45eb2-12ca-4613-999f-9d5d8ff8646c","arxiv_id":"2508.11595","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For infinite-level EL Rapoport-Zink diamonds, the structure morphism is cohomologically smooth over the non-very-special locus, and a general geometric criterion for that locus is established, with the smoothness conjecture proved in the EL case.","lead":"This paper describes, for moduli spaces of mixed-characteristic local shtukas, exactly where tangent spaces stay connected and where the spaces are conjecturally smooth for cohomology. The author proves this cohomological smoothness for all EL Rapoport-Zink spaces, extending a prior result that covered only the basic case.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 3.2.3's asserted identification of inscribed structures is 'almost formal'/'immediate' but unproved; Corollary 3.2.4 collapses if the two inscription structures differ.","rationale":"The reader's weakest assumption is exactly this inscription compatibility, and the manuscript text itself flags it by using 'almost formal' and 'immediate'. I agree this is the load-bearing point. I do not see an internal contradiction in the rest of the proof: given Prop 3.2.3, Lemma 2.2.1, Lemma 2.3.2, and Proposition 2.3.5, the Jacobian-criterion argument in Cor 3.2.4 is coherent, including the full-faithfulness/rank step. The concern is thus not a discovered error but an unverified compatibility between two separately developed formalisms; the appropriate disposition remains conditional acceptance pending an independent check. Hence no verdict change relative to the reader.","tokens_in":9996,"tokens_out":10564,"duration_ms":129655,"concrete_test":"Work out a non-basic EL example (e.g. G=GL_2/Q_p, b ordinary, with fixed determinant tau) and compare the two bundles at the corresponding section z: from [6, Sec 5.6 and Sec 6] compute s^*T_{Z/X^alg_{C_p^flat}}, and from [2, Cor 9.2.3] compute z^*E^o_max via the lattice g^{o,+}_{max}. Check whether their Harder–Narasimhan slope filtrations coincide and, more strongly, whether the associated modifications of g^o ⊗ O_X are identical. A positive result for all non-basic EL data, together with a direct functorial comparison of the two inscription functors on affinoid perfectoid test objects, would settle the concern; any mismatch in a single non-basic case would falsify Prop 3.2.3.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The single load-bearing step is Proposition 3.2.3. Theorem 1.1.5/Cor 3.2.4 requires the inscribed moduli-of-sections structure attached to the Ivanov–Weinstein scheme Z ([6, Sec 5.6] via [2, Sec 4.4]) to coincide with the inscribed structure on M^tau_{b,[mu]} used in Lemma 2.2.1. This identification is asserted, not proved: Section 1.2 calls it 'almost formal' and the proof of Proposition 3.2.3 says 'immediate from the construction in [6, Sec 5.6]'. It is not a routine consequence of an isomorphism of underlying diamonds: inscription is extra structure, and the two objects live in different presentations (scheme-theoretic sections vs. inscribed modifications). The equality is precisely what turns the tangent computation TM^tau_{b,[mu]} = BC(E^o_max) into the Jacobian-criterion identity BC(s^*T_Z) = BC(z^*E^o_max); without it, the non-very-special locus need not be Msm_Z and the conclusion of Cor 3.2.4 does not follow. Since the cited tangent computation [2, Cor 9.2.3] is also from an unreviewed preprint, the theorem rests on two unverified identifications, but Prop 3.2.3 is the one the text itself signals as omitted.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies infinite-level moduli of mixed-characteristic local shtukas with one leg and fixed determinant, viewed as inscribed v-sheaves in the sense of the author's preprint [2]. It characterizes a closed 'very special' locus in terms of the Harder–Narasimhan slopes of the vector bundle E^○_max obtained from the inscribed tangent bundle: a rank-one point is very special exactly when zero is an HN slope (Prop. 2.3.5, Cor. 2.3.7). It then conjectures that the structure morphism to Spd C_p is cohomologically smooth on the open non-very-special locus (Conj. 1.1.4), and proves this in the EL Rapoport–Zink case (Thm. 1.1.5, Cor. 3.2.4) by identifying the inscribed moduli space with the moduli of sections of the Ivanov–Weinstein scheme Z and applying the Fargues–Scholze Jacobian criterion. The proof also gives an endomorphism characterization of the very special locus in this case (Lemma 3.3.1).","tokens_in":10178,"tokens_out":5537,"duration_ms":70915,"significance":"If fully substantiated, the result is a meaningful advance: it removes the basic hypothesis from the Ivanov–Weinstein cohomological smoothness theorem and, more importantly, proposes a clean geometric characterization of the smooth locus that is conjecturally valid for all reductive groups. The paper is honest about its architectural dependence on the author's own unpublished inscription theory [2], and the equivalence chain in Prop. 2.3.5 is a useful conceptual contribution in its own right. The main risk is not internal inconsistency but whether the omitted comparison of inscribed structures in Prop. 3.2.3 can be supplied. The paper would be publishable once that comparison and the vector-bundle step in Cor. 3.2.4 are proved in detail.","major_comments":[{"comment":"The load-bearing step is the asserted identification between (a) the inscribed structure on M^τ_{b,[μ]} coming from [2, §9.4] and (b) the inscribed moduli-of-sections structure attached to the Ivanov–Weinstein scheme Z. The text calls this 'almost formal' and the proof of Prop. 3.2.3 says it is 'immediate from the construction in [6, §5.6]'. This is not a routine consequence of an isomorphism of underlying v-sheaves: inscription is extra structure, and the two objects are presented differently (moduli of exact sequences vs. moduli of sections of a scheme). The equality is exactly what converts the tangent computation TM^τ_{b,[μ]} = BC(E^○_max) from Lemma 2.2.1 into the Jacobian-criterion identity BC(s^*T_Z) = BC(z^*E^○_max) in Cor. 3.2.4. Without a proof, the identification could fail and the theorem would not follow. Please provide a complete argument, not a citation to [6] and [2] alon","section":"§1.2 and Prop. 3.2.3"},{"comment":"The deduction from equality of Banach–Colmez spaces to equality of vector bundles is too compressed. The text says BC(s^*T_Z) = BC(z^*E^○_max), then uses full faithfulness on the non-negative-slope part and a dimension equality to conclude z^*E^○_max = s^*T_Z. But for this to work one must know that s^*T_Z has no negative-slope summands (or otherwise justify that equality of BC spaces upgrades to equality of vector bundles after passing to the non-negative part). If s^*T_Z has a negative-slope summand, BC(s^*T_Z) is typically insensitive to it, and the claimed equality with the non-very-special locus M^sm_Z is not established. This step is necessary for the identification of M^τ,non-vsp with the locus where the Jacobian criterion applies.","section":"Cor. 3.2.4, final paragraph"},{"comment":"The paper's central computation of the tangent bundle is inherited from [2, Cor. 9.2.3 and §9.4], and the comparison in Prop. 3.2.3 relies on [2, §4.4]. Since [2] is an unpublished preprint by the same author, the present manuscript is conditional on an unreviewed body of work in a way that is not merely cosmetic. I am not objecting to the use of the author's own prior work, but the referee cannot fully verify the main theorem without access to the details of the inscription formalism. The text should either state these computations explicitly enough to be checked, or clearly frame the theorem as conditional on [2] and provide the cited statements in an appendix.","section":"Lemma 2.2.1 and [2]"}],"minor_comments":[{"comment":"Typo: 'such a morphism does does not exist' should be 'does not exist'.","section":"Lemma 2.2.2, proof"},{"comment":"The displayed commutative diagram contains garbled arrow notation ('/leftr⫯g⊸tl⫯ne'), which should be cleaned up.","section":"Lemma 2.3.3, proof"},{"comment":"The symbol M^τ_{b,[μ]} is used both for the inscribed v-sheaf and for its underlying diamond; the difference is essential to the proof (e.g. in §2.1 and Prop. 3.2.3). Please use visibly distinct notation or explicitly say when the distinction is being suppressed.","section":"General notation"}],"recommendation":"major_revision","confidential_remarks":"The paper is readable and the geometric characterization is attractive. The main issue is the unproved identification of inscribed structures in Prop. 3.2.3; if that is supplied, the theorem has a good chance of being correct. The reliance on the author's unpublished [2] is a review risk, but not a reason for rejection in itself. I recommend major revision rather than rejection: the missing arguments are local and can, in principle, be written up."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What Howe does here is concrete: Theorem 1.1.5 removes the basic hypothesis from Ivanov–Weinstein's cohomological smoothness result, covering the full EL Rapoport–Zink case. The new criterion—very special locus equals the locus where the inscribed tangent space is disconnected—is genuinely useful, and Conjecture 1.1.4 is a sensible target. The proof structure is coherent: compute the tangent bundle as BC(E_max), use duality and p-adic Lie group arguments to identify very special points with the zero-slope condition, then invoke the Fargues–Scholze Jacobian criterion. Nothing is fitted, and the conjecture is not assumed; the circularity burden is low.\n\nThe soft spot is Proposition 3.2.3. The theorem's final step needs the inscribed moduli-of-sections structure on the Ivanov–Weinstein scheme Z to match the inscription formalism from [2] used in the tangent computation. The text calls the comparison 'almost formal' and 'immediate from the construction in [6, §5.6]', but no proof is given. Inscription is extra structure, not something an isomorphism of underlying diamonds carries automatically. If the two structures disagreed, the equality z*E_max = s*T_Z would fail and the Jacobian criterion would not apply on the non-very-special locus. The reader and the stress-test both flag this, and I think the worry lands. That said, the paper is transparent about the omission—it does not pretend to have written the identification out—and the surrounding argument is clean enough that the gap looks fillable rather than fatal.\n\nI also note the heavy reliance on the author's unpublished preprint [2]. The tangent bundle computation is quoted from there. Self-citation is not a flaw here since [2] is independent work, but it does mean the full verification of Theorem 1.1.5 depends on two checks the present paper does not contain: the inscribed-structure identification and the correctness of [2, Cor. 9.2.3]. A careful referee should look at both.\n\nThis is for p-adic geometers working on local Shimura varieties and the geometrization program. I would send it to peer review with a referee who can cross-check the inscription theory. The main theorem is important if true, and the paper is honest about the deferred details.\n\nRecommendation: accept a serious referee; the referee should spend time on Proposition 3.2.3.","headline":"Howe proves a genuine extension of Ivanov–Weinstein to all EL Rapoport–Zink spaces, but the proof's last step rests on an identification of inscribed structures that the paper calls 'almost formal' and leaves unproved.","tokens_in":10845,"tokens_out":1847,"would_cite":true,"duration_ms":22161,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11G18","14G22","14L05"],"pacs":[],"model":"deepseek-v4-flash","headline":"The very special locus in shtuka moduli is exactly the zero Harder–Narasimhan slope locus, and cohomological smoothness holds on its complement in the EL Rapoport–Zink case.","keywords":["local shtukas","Rapoport–Zink spaces","cohomological smoothness","Banach–Colmez spaces","Fargues–Fontaine curve","Harder–Narasimhan slopes","inscribed v-sheaves","diamonds"],"falsifier":"Exhibit a geometric point in a non-basic EL Rapoport–Zink space whose intersection of the two period fibers is non-discrete while $z^*E^{\\circ}_{\\max}$ has no zero Harder–Narasimhan slope; that mismatch would refute the characterization in Proposition 2.3.5 and the theorem built on it.","tokens_in":2124,"feed_emoji":"","tokens_out":4931,"duration_ms":110959,"temperature":0.7,"pith_summary":"This paper isolates a simple numerical condition on the inscribed tangent bundle of moduli of mixed-characteristic local shtukas with one leg and fixed determinant: a geometric point is very special precisely when the pullback of the bundle $E^{\\circ}_{\\max}$ admits zero as a Harder–Narasimhan slope, equivalently when the corresponding inscribed tangent space is disconnected. The paper conjectures that the structure morphism to $\\operatorname{Spd} C_p$ is cohomologically smooth over the open complement of the very special locus, and proves the conjecture in the EL Rapoport–Zink case, generalizing earlier work from the basic case to arbitrary $b$. This matters because cohomological smoothness is the finiteness property needed for $\\ell$-adic cohomology on these infinite-level diamonds, so the result turns a structural question about period spaces into a checkable slope computation.","feed_headline":"Zero-slope test locates the smooth locus of shtuka moduli","feed_subtitle":"Conjecture proved for all EL Rapoport–Zink spaces; slope criterion defines the smooth locus.","key_machinery":"The central object is the inscribed Banach–Colmez tangent bundle. The paper works with inscribed v-sheaves, a differential enrichment of diamonds in which each space carries a tangent bundle valued in Banach–Colmez spaces. For the local shtuka moduli space, this tangent bundle is the space of global sections of a vector bundle $E^{\\circ}_{\\max}$ on the relative Fargues–Fontaine curve, formed from the adjoint action of $b$ on the derived Lie algebra. The very-special/non-very-special dichotomy is exactly whether zero appears among the Harder–Narasimhan slopes of $z^*E^{\\circ}_{\\max}$. In the EL case, comparing this inscribed structure with the Ivanov–Weinstein moduli-of-sections scheme shows","core_discovery":"For a connected reductive group $G$, a conjugacy class $[\\mu]$ of cocharacters, and a Kottwitz class $b$, the paper shows that the very special locus of the fixed-determinant infinite-level diamond $M^{\\tau}_{b,[\\mu]}$ over $\\operatorname{Spd} C_p$ is exactly the locus of rank-one geometric points $z$ where the vector bundle $z^*E^{\\circ}_{\\max}$—built from the adjoint action of $b$ on the derived Lie algebra—admits zero as a Harder–Narasimhan slope (Corollary 2.3.7). Equivalently, those are the points where the intersection of the Hodge and Hodge–Tate period fibers is non-discrete. The paper conjectures that the structure morphism is cohomologically smooth on the open complement of the very","pith_inferences":["If Conjecture 1.1.4 holds beyond the EL case, cohomological smoothness would reduce to a slope computation on the Fargues–Fontaine curve, turning a difficult geometric property into linear algebra of isocrystals; explicit Hodge–Newton reducible cases could serve as a test.","Proposition 2.3.5 suggests a possible strengthening: the cohomologically smooth locus might in general coincide with the locus where the inscribed tangent space is connected, not merely contain it, though the paper only establishes this equivalence pointwise and proves smoothness in the EL case.","The comparison of inscribed structures in Proposition 3.2.3 may extend to other minuscule cocharacter data, since the construction of $Z$ via exact sequences is not obviously limited to EL data; this offers a route toward the conjecture for Hodge-type or abelian-type local Shimura varieties.","The endomorphism characterization in Lemma 3.3.1 is computable in examples: for explicit $p$-divisible groups one can check whether the center of $B$ equals the full endomorphism ring, giving a practical test of the conjecture's prediction about the smooth locus."],"forward_implications":["Conjecture 1.1.4 now holds for all EL infinite-level Rapoport–Zink spaces, not just the basic case: the non-very-special locus is cohomologically smooth over $\\operatorname{Spd} C_p$.","The very special locus has a purely geometric description as the locus where the intersection of the two period fibers is non-discrete, and equivalently where the associated modification of $G$-bundles has extra infinitesimal automorphisms.","In the EL case, very special points are exactly those whose associated $p$-divisible group has extra $B$-linear endomorphisms beyond the center of $B$, matching the earlier definition of the special locus in the basic case.","The Harder–Narasimhan slope condition is semicontinuous, so the very special locus is a closed subdiamond and its complement is an open subdiamond, giving a natural candidate for the cohomologically smooth locus in all cases."],"supporting_citations":[{"why":"Supplies the basic-case theorem, the smooth quasi-projective scheme $Z$ whose moduli of sections models $M^{\\tau}_{b,[\\mu]}$, and the construction that the paper generalizes beyond the basic case.","marker":"[6]"},{"why":"Supplies the inscribed v-sheaf formalism and the computation $TM^{\\tau}_{b,[\\mu]} = BC(E^{\\circ}_{\\max})$ (Corollary 9.2.3) that is the starting point for the slope analysis.","marker":"[2]"},{"why":"Supplies the Fargues–Scholze Jacobian criterion (Theorem IV.4.2) and the semicontinuity of the Harder–Narasimhan polygon, used to conclude cohomological smoothness.","marker":"[1]"},{"why":"Provides the Scholze–Weinstein classification of p-divisible groups, used to translate the very special condition into the existence of extra endomorphisms in the EL case.","marker":"[10]"},{"why":"Gives full faithfulness of the functor from vector bundles of non-negative slope to Banach–Colmez spaces, used in Corollary 3.2.4 to identify $z^*E^{\\circ}_{\\max}$ with the pullback of the relative tangent bundle of $Z$.","marker":"[7]"},{"why":"Used to justify the existence of a determinant point $\\tau$ in the image of the determinant map, a standing assumption in the construction of $M^{\\tau}_{b,[\\mu]}$.","marker":"[5]"}],"fun_headline_variants":["Zero-slope condition pinpoints smooth locus in shtuka moduli","Slope-zero test defines smooth locus in shtuka moduli","Cohomological smoothness conjecture proved for EL Rapoport-Zink","Zero Harder-Narasimhan slope marks smooth shtuka locus","Shtuka moduli: zero-slope points define the smooth locus"],"cache_read_input_tokens":12416,"weakest_assumption_plain":"The whole argument hinges on the claim that the differential structure attached to the moduli space by the Ivanov–Weinstein construction is the same as the differential structure computed by the general inscription formalism; if these two enrichments disagreed at even one point, the proof of the theorem would fail.","fun_headline_variants_meta":{"raw":{"variants":["Zero-slope condition pinpoints smooth locus in shtuka moduli","Slope-zero test defines smooth locus in shtuka moduli","Cohomological smoothness conjecture proved for EL Rapoport-Zink","Zero Harder-Narasimhan slope marks smooth shtuka locus","Shtuka moduli: zero-slope points define the smooth locus"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000471,"raw_usage":{"total_tokens":2135,"prompt_tokens":658,"completion_tokens":1477,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":402,"completion_tokens_details":{"reasoning_tokens":1382}},"tokens_in":402,"tokens_out":1477,"duration_ms":12463,"temperature":1.0,"reasoning_tokens":1382,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T19:49:53.650841+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibit a geometric point in a non-basic EL Rapoport–Zink space whose intersection of the two period fibers is non-discrete while $z^*E^{\\circ}_{\\max}$ has no zero Harder–Narasimhan slope; that mismatch would refute the characterization in Proposition 2.3.5 and the theorem built on it.","supporting_citations":[{"cited_title":"Ivanov and Jared Weinstein","cited_arxiv_id":null,"evidence_quote":"Supplies the basic-case theorem, the smooth quasi-projective scheme $Z$ whose moduli of sections models $M^{\\tau}_{b,[\\mu]}$, and the construction that the paper generalizes beyond the basic case."},{"cited_title":"Inscription, twistors, and p-adic periods","cited_arxiv_id":null,"evidence_quote":"Supplies the inscribed v-sheaf formalism and the computation $TM^{\\tau}_{b,[\\mu]} = BC(E^{\\circ}_{\\max})$ (Corollary 9.2.3) that is the starting point for the slope analysis."},{"cited_title":"Moduli of p-divisible groups","cited_arxiv_id":null,"evidence_quote":"Provides the Scholze–Weinstein classification of p-divisible groups, used to translate the very special condition into the existence of extra endomorphisms in the EL case."},{"cited_title":"Espaces de Banach-Colmez et faisceaux coh´ erents sur la courbe de Fargues- Fontaine","cited_arxiv_id":null,"evidence_quote":"Gives full faithfulness of the functor from vector bundles of non-negative slope to Banach–Colmez spaces, used in Corollary 3.2.4 to identify $z^*E^{\\circ}_{\\max}$ with the pullback of the relative tangent bundle of $Z$."},{"cited_title":"Admissible pairs and $p$-adic Hodge structures II: The bi-analytic Ax-Lindemann theorem","cited_arxiv_id":"2308.11064","evidence_quote":"Used to justify the existence of a determinant point $\\tau$ in the image of the determinant map, a standing assumption in the construction of $M^{\\tau}_{b,[\\mu]}$."}],"review_version":1}