{"id":"828c10b9-d027-4cb9-ae43-beb7b192d34c","arxiv_id":"2507.01395","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For ramified quadratic extensions, the paper introduces arithmetic transfer conjectures, proves them for n=1, and establishes structural results on exceptional divisors and correspondences.","lead":"This paper formulates new arithmetic transfer conjectures for unitary Rapoport-Zink spaces in the ramified quadratic case, and proves them in the lowest dimensional setting, n=1. It also constructs regular splitting models and characterizes exceptional special divisors on these moduli spaces.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Open flatness of large-correspondence cycles (Conj. 9.6.1, 10.1.2) is the load-bearing gap: the ATC intersection numbers in rows 5 and 7 are defined via flat closures whose equality with the intended cycles is unproved, though the n=1 theorem bypasses this.","rationale":"The reader's weakest assumption identifies exactly the flatness of the large-correspondence cycles as the load-bearing premise for the general ATC framework. My review confirms this: the paper's own Conjectures 9.6.1 and 10.1.2 explicitly leave open the flatness of fM[t]_n and eN[t]_n, and without flatness the flat closures on which the intersection numbers are computed can differ from the intended cycles outside the worst points, changing the geometric content of the conjectures. For the central n=1 theorem, the relevant fM[0]_1 is a base change of the finite flat level-changing morphism, so flatness is automatic and the theorem is not threatened by this gap; however, the paper does not spell out this check, and the general conjectural statements in rows 5 and 7 of the table remain conditional. The verdict CONDITIONAL is therefore appropriate: the headline n=1 result is credible and the structural framework is valuable, but the unresolved flatness for the large correspondences is a genuine soft spot that should be settled before the general conjectures are accepted as fully formulated. I also note the dangling 'see footnote ??' in the proof of Theorem 12.2.5, but that theorem concerns the local-model analogue and is not needed for the n=1 proof, so it does not change the assessment.","tokens_in":82956,"tokens_out":20719,"duration_ms":210626,"concrete_test":"For the first open case (n=2, t=0, Y-divisor side), compute the π-power torsion of fM[0]_2 and eN[0]_2 by a local calculation at each worst point: use the local model M[2,0]_3 and the Y(u)[2]_2 identification to check whether the fiber of fM[0]_2 over the worst point has any embedded components. If the π-power torsion vanishes, flatness holds and Conjectures 9.6.1 and 10.1.2 are verified in this case; if not, exhibit the extra component and recompute the intersection number in Conjecture 9.10.1 for n=2 to measure the discrepancy.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's new AT conjectures for the large correspondences (Conjecture 9.10.1 for fM[t]_n and Conjecture 10.6.1 for eN[t]_n) define the intersection numbers on the flat closure of the base change to the splitting model (§9.8, §10.5). The geometric interpretation of these intersections as counting the intended Z- or Y-cycles requires the original cycles to be flat over Spf O_ŝF, so that the flat closure coincides with the strict transform away from the exceptional divisors. The paper states this as Conjecture 9.6.1 and Conjecture 10.1.2 and proves it only in the cases where the correspondence is isomorphic to an RZ space (Theorems 8.1.2, 9.1.2, 11.1.2). For the central n=1 theorem (Conjecture 9.10.1, type (0,0)), the relevant fM[0]_1 is the base change of the finite flat morphism N[2,0]_2 → N[2]_2 along the closed immersion N[0],◦_1 → N[2]_2, so it is finite flat; however, the paper does not state this verification, and for the general framework the equality of the flat closure with the intended correspondence is open. In particular, the n=2 cases (ii)-(v) listed in §1.3, which the paper explicitly leaves open, already depend on this flatness.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs regular formal moduli spaces for unitary groups attached to a ramified quadratic extension of a p-adic field, using splitting models and correspondences of atomic type. It formulates several arithmetic transfer (AT) conjectures, organized in a table of types (n,t), (n-1,t), (t,n), and (t,n+1), and proves the lowest-dimensional cases. For n=1 the proof relies on prior results of [31] and [32] and on a new computation in Section 13 for type (n-1,t)=(0,0), which is the only genuinely new instance. The paper also proves structural results on exceptional special divisors and on the coincidence of small and large correspondences in some cases (Theorems 9.1.2, 11.1.2).","tokens_in":83313,"tokens_out":4970,"duration_ms":56430,"significance":"If the new Section 13 computation is complete, the paper is a substantial contribution: it gives a systematic framework for arithmetic transfer in the ramified quadratic case, introduces the splitting-model ambient spaces that make the intersection numbers well-defined, identifies the correct test functions via lattice models, and proves the first non-trivial cases. The exceptional divisor isomorphisms (Theorems 1.5.1, 6.1.3-6.1.5) are valuable in their own right, and the comparison of small and large correspondences via flatness results (Theorems 8.1.2, 9.1.2, 11.1.2 and Corollary 11.1.3) is an important structural step. The main caveat is that the n=1 theorem, the central new assertion, is not verifiable from the review copy because Section 13 breaks off before the actual intersection computation and analytic comparison.","major_comments":[{"comment":"The proof of the new n=1 case, Conjecture 9.10.1 for type (n-1,t)=(0,0), is incomplete in the supplied manuscript. The section reduces to the inhomogeneous setting, recalls the Iwahori-level Lubin-Tate isomorphism, and begins to describe the cycles fM[0],±_1, but the text is truncated before the intersection multiplicity computation in §13.4 and the comparison with ∂Orb in §13.5. Since this is the only case not already covered by [31] or [32], the central claim of Theorem 1.3.1 cannot be assessed from the manuscript as presented. The computation must be supplied in full.","section":"Section 13, Theorem 1.3.1(iii)"},{"comment":"The proof of Theorem 12.2.5 is only sketched and contains an unresolved footnote marker ('see footnote ??'). The key steps (2) and (3), showing that the constructed filtration is a direct summand of rank n-1 and satisfies the strengthened spin condition, are deferred to [45, Prop. 5.14], but that reference is for a bijection between RZ spaces, not for the local model statement. Since Theorem 12.2.5 underlies Theorem 11.1.2 and Corollary 11.1.3, the missing details are load-bearing for the flatness claims in Section 11.","section":"§12.2, Theorem 12.2.5"},{"comment":"The intersection numbers in Conjectures 9.10.1 and 10.6.1 are defined using the flat closures fM[t],spl_n and eN[t],spl_n. The paper proves flatness of the underlying correspondences only in cases where they are isomorphic to RZ spaces (Theorems 8.1.2, 9.1.2, 11.1.2), and leaves Conjectures 9.6.1 and 10.1.2 open in general. If flatness fails, the flat closure can differ from the intended correspondence away from the worst points, so the geometric interpretation of the intersection number as counting the Z- or Y-cycles is not established for the general framework. For the n=1 theorem this issue is bypassed because fM[0]_1 is finite flat, but the paper does not state this verification explicitly.","section":"§9.8, §10.5, Conjectures 9.6.1 and 10.1.2"},{"comment":"The relation between Theorem 1.3.1 and the table in §1.2 should be stated more precisely. In particular, Theorem 1.3.1 says it proves Conjecture 1.1.1 for n=1, but the new Section 13 computation is only for Conjecture 9.10.1, type (0,0), which is the fifth row of the table. The paper should explain explicitly how the remaining rows for n=1 follow from [31] and [32], and whether any of them require an argument beyond the cited theorems.","section":"§1.3 and §9.10"}],"minor_comments":[{"comment":"The notation 'F = ¯k for a fixed algebraic closure' is confusing: likely \\(\\bar{k}\\) is intended for the residue field algebraic closure, not the field F itself.","section":"§1.12.1"},{"comment":"The phrase 'defined defined by ht(g)' contains a duplicated word and should be corrected.","section":"§5.3, Proposition 5.3.5"},{"comment":"The word 'embedddings' is a typo for 'embeddings'.","section":"Lemma 9.6.2"},{"comment":"In the proof, references to 'case (ii) of Theorem 7.2.4' and 'case (i) of Theorem 7.2.4' appear to refer to cases in Theorem 7.2.3; the numbering should be corrected.","section":"§7.2, proof of Theorem 7.2.4"},{"comment":"The notation 'fM[0]_0' appears where the dimension should be n=1; the subscript seems to be a typo.","section":"§13.3"}],"recommendation":"major_revision","confidential_remarks":"The supplied review copy truncates Section 13 at the start of the cycle description. If the full version contains the missing computation of the intersection number and the analytic comparison, the main technical objection would be resolved locally. I would be willing to review the complete version. The unresolved footnote in §12.2 is another point that the authors must fix before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The systematic treatment of the ramified quadratic case is the real news here. Replacing embeddings by small and large correspondences between RZ spaces, introducing splitting models as ambient spaces, and proving the exceptional divisor isomorphisms (Theorem 1.5.1) plus the spin-condition comparison (Theorem 6.1.2) gives later workers a toolkit that the earlier papers [31,32,21] lack. The authors are also honest about what is proved versus conjectured, which is rare and welcome.\n\nThe n=1 theorem has two parts from prior literature and one new computation in Section 13. The new computation is truncated in the review copy, so I cannot fully check it; the surrounding reduction to the inhomogeneous setting and the use of germ expansions look standard. The other two n=1 cases follow from [32], so the main theorem is as solid as its cited base. That is modest, but the structural theorems are the load-bearing contribution, not the n=1 case.\n\nNow the soft spots. The stress-test note is on target: Conjectures 9.6.1 and 10.1.2 assert flatness of the large-correspondence cycles fM[t]_n and eN[t]_n, and the intersection numbers in Conjectures 9.10.1 and 10.6.1 are defined via flat closures. If that flatness fails, the geometric interpretation of those intersection numbers changes. The proofs only establish flatness in the cases where the correspondence is isomorphic to an RZ space (Theorems 8.1.2, 9.1.2, 11.1.2). For the n=1 theorem this is not fatal: fM[0]_1 is finite flat, though the paper does not explicitly say so. But for the n=2 test cases listed in Section 1.3, which the authors leave open, this flatness is already needed. So the gap is real but clearly identified, and it does not affect the proved n=1 statement.\n\nThere is also an unresolved footnote in the proof of Theorem 12.2.5 (the proof of the local model isomorphism). The argument is sketched via reduction to Yao's computation, but the footnote marker is left as '??'. That needs to be fixed before publication; it is likely minor but it is not merely cosmetic.\n\nOverall: this is a careful, valuable paper. It deserves a serious referee. I would cite the structural results even if the full conjectures fall. For the next test case, the authors should either prove Conjecture 9.6.1 for t=0 or state explicitly where it is used in the n=2 cases.","headline":"The structural framework for ramified arithmetic transfer is a real contribution, and the n=1 theorem is credible, but the proof of a key local model isomorphism and the flatness of large-correspondence cycles are not fully settled in this version.","tokens_in":83798,"tokens_out":1339,"would_cite":true,"duration_ms":17312,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11G18","11S37","14G35","14L05"],"pacs":[],"model":"deepseek-v4-flash","headline":"Ramified quadratic arithmetic transfer proved in n=1 case","keywords":["arithmetic transfer conjecture","arithmetic fundamental lemma","Rapoport–Zink spaces","splitting models","ramified quadratic extension","special cycles","formal moduli spaces","unitary groups"],"falsifier":"For n=3, t=0, choose a regular semisimple matched pair (γ,g) and compute both sides of Conjecture 9.10.1; a mismatch for any such pair would disprove the conjecture. Alternatively, show that the larger-correspondence cycle fM or eN is not flat by finding a geometric point where its flat closure has extra components over a worst point, which would falsify the geometric interpretation of the intersection number.","tokens_in":1960,"feed_emoji":"🧮","tokens_out":3621,"duration_ms":207037,"temperature":0.7,"pith_summary":"The paper aims to extend the arithmetic fundamental lemma to the ramified quadratic setting of unitary groups. It constructs regular formal moduli spaces by replacing Rapoport–Zink spaces with splitting models, and it proves exceptional divisor isomorphisms that identify special cycles with lower-dimensional Rapoport–Zink spaces. These isomorphisms allow the formulation of arithmetic transfer conjectures connecting intersection numbers on regular ambient products to derivatives of weighted orbital integrals. The conjectures are proved when n=1, including the previously open case of type $(n-1,t)=(0,0)$, where the identity $\\langle f\\mathbb{M}^{[0],\\mathrm{spl}}_1, g f\\mathbb{M}^{[0],\\mathrm{spl}}_1\\rangle\\cdot\\log q = -\\partial\\mathrm{Orb}(\\gamma,\\varphi')$ is established for matched regular semisimple elements. A sympathetic reader should care because this is the ramified analogue of the arithmetic fundamental lemma that underpins the arithmetic Gan–Gross–Prasad approach to Gross–Zagier-type formulas.","feed_headline":"Ramified quadratic arithmetic transfer proved in n=1 case","feed_subtitle":"Splitting models yield regular ambient spaces, and the conjectured intersection–orbital identity is established in the lowest dimension.","key_machinery":"The load-bearing object is the splitting model $N^{[t],\\mathrm{spl}}_{n,\\varepsilon}$, defined as the flat closure of the naive splitting model and realized as the blow-up of the Rapoport–Zink space $N^{[t]}_{n,\\varepsilon}$ in its worst points; it is regular, indeed semi-stable, and therefore can serve as a factor in an ambient product for intersection theory. The second central mechanism is the strengthened spin condition in the moduli-theoretic description of these Rapoport–Zink spaces, which the paper uses to prove the exceptional special-divisor isomorphisms $Z(u)^{[t]}_{n,\\varepsilon} \\simeq N^{[t]}_{n-1,\\varepsilon'}$ and the Y-cycle analogue; these isomorphisms turn special cycles into lower-dimensional Rapoport–Zink spaces, making flatness and finiteness of intersections accessible. The correspondence framework of small and large correspondences, together with lattice models on the unitary side, identifies the test functions whose orbital integrals match the intersection numbers.","core_discovery":"The central claim, on the paper's own terms, is that the ramified quadratic case can be treated by the same global strategy as the unramified case, once the ambient spaces are made regular. The paper proves that the splitting model $N^{[t],\\mathrm{spl}}_{n,\\varepsilon}$—the flat closure of the naive splitting model, equivalently the blow-up of $N^{[t]}_{n,\\varepsilon}$ in its worst points—is always flat and semi-stable, and that products such as $N^{[n]}_n \\times N^{[t],\\mathrm{spl}}_{n+1}$ are regular. It then proves the exceptional divisor theorem: for a unit-length special vector $u$, the special cycle $Z(u)^{[t]}_{n,\\varepsilon}$ is isomorphic to $N^{[t]}_{n-1,\\varepsilon'}$ in all but a few listed cases, and the Y-cycle analogue holds through a double cover. These identifications let the paper define arithmetic intersection numbers and state arithmetic transfer conjectures of types $(n,t)$, $(n-1,t)$, $(t,n)$, and $(t,n+1)$. The lowest-dimensional case $n=1$ is proved completely: all rows of the summary table hold, including the previously open case of Conjecture 9.10.1 for $(n-1,t)=(0,0)$, where the identity $\\langle f\\mathbb{M}^{[0],\\mathrm{spl}}_1, g f\\mathbb{M}^{[0],\\mathrm{spl}}_1\\rangle\\cdot\\log q = -\\partial\\mathrm{Orb}(\\gamma,\\varphi')$ is established for matched regular semisimple elements.","pith_inferences":["A natural next step is to prove the flatness conjectures for the larger correspondences by giving a moduli-theoretic description of the spaces fM and eN that exhibits them as Rapoport–Zink spaces of mixed parahoric level, in the spirit of the isomorphism proved for type (t,n+1).","The n=2 cases advertised as future work may be accessible with the same explicit Lubin–Tate and Drinfeld-type models used for n=1, since the relevant Rapoport–Zink spaces in dimension two are known concretely.","Because the splitting model is a blow-up in the worst points, the difference between the naive and split intersection numbers is concentrated on exceptional divisors; this suggests that error-term formulas of the type conjectured for one extreme case may be a general phenomenon.","The lattice-model origin of the test functions indicates that the same functions, transported to a global Shimura variety, should reproduce the local Hecke action appearing in the global arithmetic Gan–Gross–Prasad conjecture; a computational check in small residue characteristic would be a cheap test."],"forward_implications":["If the arithmetic transfer conjectures in Sections 8–11 are correct, the arithmetic Gan–Gross–Prasad machinery extends to ramified quadratic extensions with parahoric level structures, not only hyperspecial ones.","The n=1 cases form the base of the expected induction: the n=2 cases listed in Section 1.3 are the next test cases standing between the current proof and higher dimensions.","The exceptional divisor isomorphisms imply that all Z-divisors attached to unit-length vectors are lower-dimensional Rapoport–Zink spaces, so the corresponding intersection numbers are finite and computable by lattice counting.","In the extreme types t=n (even n) and t=n+1 or t=n−1 (odd n), the new conjectures specialize to the earlier arithmetic transfer conjectures of the paper's predecessors, so the framework is consistent with known results."],"supporting_citations":[{"why":"Supplies the moduli-theoretic definition of all relevant Rapoport–Zink spaces through the strengthened spin condition, on which the exceptional divisor isomorphisms rest.","marker":"[22]"},{"why":"Constructs splitting models as blow-ups in the worst points and proves their flatness and semi-stability, the key to making ambient products regular.","marker":"[15]"},{"why":"Proves the n=1 cases for the unramified-level variants and supplies the t=n+1 and t=n−1 cases to which some of the new conjectures reduce.","marker":"[32]"},{"why":"Formulates the original arithmetic transfer and arithmetic fundamental lemma conjectures in the exotic smooth cases, recovered as limiting cases here.","marker":"[31]"},{"why":"Proves the isomorphism of Y-cycles with lower-dimensional Rapoport–Zink spaces used in the exceptional divisor theorem and in the proof of Theorem 11.1.2.","marker":"[45]"},{"why":"Introduces the small and large correspondence framework and the lattice-model test functions in the unramified case that the paper adapts to the ramified setting.","marker":"[21]"}],"fun_headline_variants":["Ramified quadratic transfer conjecture proved for n=1","Regular moduli spaces unlock ramified quadratic case","Splitting model proves arithmetic transfer in lowest dimension","n=1 arithmetic transfer holds for ramified quadratic units","Ramified case tamed: transfer conjecture verified at n=1"],"cache_read_input_tokens":85888,"weakest_assumption_plain":"The load-bearing premise is that the cycles used in the larger correspondences are flat over the base ring, so that taking their flat closure does not change the intended cycle; the paper proves this only when the cycle is isomorphic to a Rapoport–Zink space.","fun_headline_variants_meta":{"raw":{"variants":["Ramified quadratic transfer conjecture proved for n=1","Regular moduli spaces unlock ramified quadratic case","Splitting model proves arithmetic transfer in lowest dimension","n=1 arithmetic transfer holds for ramified quadratic units","Ramified case tamed: transfer conjecture verified at n=1"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000187,"raw_usage":{"total_tokens":1332,"prompt_tokens":952,"completion_tokens":380,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":568,"completion_tokens_details":{"reasoning_tokens":301}},"tokens_in":568,"tokens_out":380,"duration_ms":4621,"temperature":1.0,"reasoning_tokens":301,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T20:52:35.167955+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For n=3, t=0, choose a regular semisimple matched pair (γ,g) and compute both sides of Conjecture 9.10.1; a mismatch for any such pair would disprove the conjecture. Alternatively, show that the larger-correspondence cycle fM or eN is not flat by finding a geometric point where its flat closure has extra components over a worst point, which would falsify the geometric interpretation of the intersection number.","supporting_citations":[{"cited_title":"Luo, On the moduli description of ramified unitary local models of signature (n − 1, 1)","cited_arxiv_id":null,"evidence_quote":"Supplies the moduli-theoretic definition of all relevant Rapoport–Zink spaces through the strengthened spin condition, on which the exceptional divisor isomorphisms rest."},{"cited_title":"Rapoport, B","cited_arxiv_id":null,"evidence_quote":"Proves the n=1 cases for the unramified-level variants and supplies the t=n+1 and t=n−1 cases to which some of the new conjectures reduce."},{"cited_title":"Rapoport, B","cited_arxiv_id":null,"evidence_quote":"Formulates the original arithmetic transfer and arithmetic fundamental lemma conjectures in the exotic smooth cases, recovered as limiting cases here."},{"cited_title":"A Kudla-Rapoport Formula for Exotic Smooth Models of Odd Dimension","cited_arxiv_id":"2404.14431","evidence_quote":"Proves the isomorphism of Y-cycles with lower-dimensional Rapoport–Zink spaces used in the exceptional divisor theorem and in the proof of Theorem 11.1.2."}],"review_version":1}