{"id":"e175b8a1-dbad-427a-9fb1-8c5463d4e704","arxiv_id":"1908.01701","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The authors prove the local and global Kudla-Rapoport conjectures and, combining with results of Liu and Garcia-Sankaran, the arithmetic Siegel-Weil formula in arbitrary dimension.","lead":"This paper proves the local Kudla-Rapoport conjecture, an exact identity between arithmetic intersection numbers of special cycles on unitary Rapoport-Zink spaces and derivatives of local representation densities. It also derives the global Kudla-Rapoport conjecture and new cases of the arithmetic Siegel-Weil formula in arbitrary dimension.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 6.4.6 is the linchpin of the geometric Fourier transform, but its proof treats only isotropic vectors in Λ∨∖Λ; the anisotropic cases used in Lemma 6.4.7 are not justified.","rationale":"The reader's weakest assumption identifies the Bruhat–Tits stratification and the Tate-class computation as the fragile input to the geometric side; I agree that this is the most delicate region, but I want to localize the concern further. The specific step that actually carries the Fourier-invariance for arbitrary vertex lattices is Lemma 6.4.6, and its proof as written omits the anisotropic vectors in Λ∨∖Λ. Lemma 6.4.7 then relies on those values, and its counting formula for anisotropic x does not have an obvious justification. If this gap is real, the central local theorem is affected, not merely the global applications. On the other hand, the rest of the proof is substantial and mostly explicit: the uncertainty principle, the analytic Fourier computation, and the induction on val(L♭) are coherent, and the horizontal parts are matched by Theorems 4.2.1 and 6.1.3. I therefore do not see grounds for rejection; the appropriate disposition is to keep the paper conditional pending a direct verification of Lemma 6.4.6 for anisotropic vectors, and also to keep the reader's earlier conditions about Conjecture 10.4.1 and the sketched semi-global deduction. The concrete test proposed above would settle the anisotropic question in the first nontrivial case and should be enough to indicate whether the gap is cosmetic or structural.","tokens_in":82159,"tokens_out":19177,"duration_ms":200484,"concrete_test":"Test Lemma 6.4.6 in the smallest nontrivial case d = 2 (Λ type 5, V(Λ) a Deligne–Lusztig surface) over F_{q^2}: choose an anisotropic vector x ∈ Λ∨∖Λ with val(x) = 0 and compute c_{V(Λ)}(x) from the two-term complex [O(−Z(x))|V(Λ) → O] and the Chern-character formula of Definition 6.4.1, using the explicit Bruhat–Tits decomposition into curves. If the resulting value is not c(2) = 1 − q^2, Lemma 6.4.6 is false and the geometric side of (1.4.0.1) fails for such Λ. An independent check is to recompute the count in Lemma 6.4.7 case (2) for anisotropic x: the number of type-3 vertex lattices Λ′ ⊇ Λ with x ∈ Λ′∨∖Λ′ should be evaluated directly for d = 2, q = 3 and compared with S_{2d−1,d−1} q^{2d−2}.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Section 6.4 extends the Fourier-invariance of vertical cycles from Deligne–Lusztig curves to all vertex lattices. This extension is needed to prove (1.4.0.1) for arbitrary L♭ and hence Theorem 3.4.1 via the uncertainty principle. Its key numerical content is Lemma 6.4.6, which asserts c_{V(Λ)}(x) = c(d) for every x ∈ Λ∨∖Λ with val(x) ≥ 0. The proof, however, says that after the support observation it remains only to verify (6.4.6.1) for x ∈ VΛ∖{0} with (x,x) = 0. It then verifies the cases x ∈ Λ and x isotropic outside Λ. Nonzero vectors in Λ∨∖Λ whose reduction in the finite hermitian space VΛ = Λ∨/Λ is anisotropic have val(x) = 0 and are not covered by that argument. Such vectors exist in nonsplit hermitian spaces of odd dimension. Lemma 6.4.7 case (2) then uses the missing value to evaluate the right-hand side for all x ∈ Λ∨∖Λ, and its count of type-3 sublattices Λ′ with x ∈ Λ′∨∖Λ′ is written as S_{2d−1,d−1} q^{2d−2}; for anisotropic x the ambient space x⊥ is nondegenerate of dimension 2d, so the number of (d−1)-dimensional totally isotropic subspaces of x⊥ is S_{2d,d−1}, which is not obviously equal to the displayed count. If the anisotropic value is not c(d), Corollary 6.4.8 and Theorem 6.4.9 fail, so the vertical cycle is not Fourier-invariant and the uncertainty-principle induction that proves Theorem 3.4.1 would not close.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves the local Kudla–Rapoport conjecture for unitary Rapoport–Zink spaces: for every full-rank O_F-lattice L in the nonsplit hermitian space V over an unramified extension F/F0 with p>2, the arithmetic intersection number Int(L) equals the central derivative of the local Siegel series, ∂Den(L). The proof is by induction on the dimension and on lattice valuation, decomposing the special cycles into horizontal and vertical parts, establishing Fourier invariance (local modularity) for both the geometric and analytic vertical cycles, and closing the induction with an uncertainty principle. As applications, the paper proves the global Kudla–Rapoport conjecture and, in combination with results of Liu and of Garcia–Sankaran, cases of the arithmetic Siegel–Weil formula in any dimension. Part 2 treats the almost self-dual level case, explicitly flagging that some comparison statements there depend on Conjecture 10.4.1, which is left to a companion paper.","tokens_in":82550,"tokens_out":55797,"duration_ms":559460,"significance":"If the proof is accepted, this is a major breakthrough: it settles a central conjecture of Kudla–Rapoport in the unramified local case and yields the first higher-dimensional cases of the arithmetic Siegel–Weil formula. The argument is original and highly nontrivial, combining the Bruhat–Tits stratification of Vollaard–Wedhorn with Deligne–Lusztig varieties, a reduction of the needed Tate conjecture to Lusztig's eigenvalue computations, the Cho–Yamauchi local density formula, and a Fourier-analytic uncertainty principle. The paper is honest about its external inputs and about the conditional status of Conjecture 10.4.1, which is not used for the main local theorem or for the stated global applications. I specifically examined the potential gap raised during stress-testing concerning anisotropic vectors in the finite hermitian space VΛ in §6.4; that concern does not survive close reading. Under the standard normalization of the reduced form as p(·,·) mod p, every vector of nonnegative norm valuation has isotropic reduction, while anisotropic reduction classes have negative norm valuation and are covered by the zero-support case.","major_comments":[],"minor_comments":[{"comment":"The proof is correct but terse about why it suffices to verify the formula for isotropic classes in VΛ. Since the reduced hermitian form on VΛ is normalized as p(·,·) mod p, a class represented by a vector x∈Λ∨ with val(x)≥0 is automatically isotropic, whereas anisotropic classes correspond to vectors with val(x)<0 and are already in the zero-support case. Adding this one-sentence clarification would prevent the reader from worrying about a missing anisotropic case.","section":"§6.4, Lemma 6.4.6"},{"comment":"The reference 'Lemmas 6.4.8 and 6.3.1' should be 'Lemmas 6.4.7 and 6.3.1'; Lemma 6.4.8 does not exist.","section":"§6.4, Corollary 6.4.8"},{"comment":"The displayed count S_{2d−1,d−1}q^{2d−2} is correct, but the dimension of Λ′/Λ (namely d−1 over F_{q^2}) is not stated. A short parenthetical identifying these dimensions would improve readability and make the fiber-size computation q^{2d−2} less opaque.","section":"§6.4, Lemma 6.4.7, case (2)"},{"comment":"The paper clearly states that Theorems 10.4.3 and 10.5.1 are conditional on Conjecture 10.4.1. Since the central local theorem and the global applications use the unconditional Theorem 10.3.1 rather than these conditional comparison statements, I recommend adding a sentence in the introduction explicitly listing which results are conditional on Conjecture 10.4.1.","section":"§10.4, Conjecture 10.4.1"},{"comment":"The phrase 'asp>2' should read 'as p>2'.","section":"§13.6, proof of Theorem 13.6.1"}],"recommendation":"minor_revision","confidential_remarks":"This is an important paper with a long and dense proof. The referee found no load-bearing mathematical error. The specific stress-test concern about anisotropic vectors in §6.4 does not land: the normalization of the reduced hermitian form on VΛ implies that all classes in the support of Lemma 6.4.6 are isotropic, and anisotropic classes have negative norm valuation and are in the zero case. The main caveat is the reliance on Conjecture 10.4.1 for the almost self-dual comparison theorems in Part 2, but this is explicitly acknowledged and does not affect the main local theorem or the global applications. The paper should be accepted after minor revisions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague—\n\nYou should know: this is the long-awaited proof of the local Kudla–Rapoport conjecture in full generality, and the main architectural idea is sound. The paper also gets the global KR conjecture and new cases of the arithmetic Siegel–Weil formula. It deserves a serious referee. But there is a real, narrowly repairable error in the geometric Fourier-transform section you should know about before sending it out.\n\nWhat is genuinely new: the proof by induction on n, decomposing the intersection number into horizontal and vertical parts, then using the uncertainty principle and Fourier invariance rather than explicit computations. The proof of the self-dual local theorem is detailed, and the paper is candid about relying on Cho–Yamauchi's local density formula, Vollaard–Wedhorn's stratification, and Lusztig's eigenvalue computations.\n\nThe soft spot: Lemma 6.4.6 (and behind it the type-3 base case, Lemma 6.2.1) states that the function c_{V(Λ)} takes the value c(d) on all x∈Λ^∨\\Λ with val(x)≥0. The proof only treats the isotropic case, i.e. val(x)>0. Nonzero vectors with anisotropic reduction in Λ^∨/Λ have val(x)=0 and are skipped. The stress-test is right that these cases are not covered, and the formulas as written are false for them. The correct value is 0, not c(d): for such x the special divisor meets V(Λ) only at the type-1 boundary, which is empty because no type-1 lattice between Λ and Λ^∨ contains x. Lemma 6.3.1's own Fourier computation (its case (iii)) gives 0 there, which is consistent with the corrected value. So the repair is to change 'val(x)≥0' to 'val(x)>0' in those statements. That is not a load-bearing flaw; it would not sink the induction. But it is a mathematical error in the text and a referee should ask for it to be fixed.\n\nTwo more things, both flagged by the authors: Theorem 10.5.1 is conditional on Conjecture 10.4.1 (deferred to [KRSZ19]), and the semi-global identity Theorem 13.6.1 is only sketched. These are honest caveats but mean the advertised applications are not fully unconditional without the companion paper.\n\nWho this is for: anyone in arithmetic geometry or automorphic forms working on special cycles; this will be a standard cite. Should it go to referees? Yes—with a request to correct the val≥0/val>0 issue and to expand the sketched deductions before acceptance.","headline":"Important proof of the local Kudla–Rapoport conjecture; the induction works, but the geometric Fourier lemmas need a small, essential correction (val>0, not val≥0) and the paper is explicitly conditional in parts.","tokens_in":83059,"tokens_out":20097,"would_cite":true,"duration_ms":195238,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11G18","11F27","11F70","14G35","14G40"],"pacs":[],"model":"deepseek-v4-flash","headline":"The local Kudla–Rapoport conjecture holds in every dimension.","keywords":["Kudla-Rapoport cycles","local Siegel series","arithmetic intersection numbers","Rapoport-Zink spaces","arithmetic Siegel-Weil formula","local representation densities","unitary Shimura varieties","uncertainty principle"],"falsifier":"Compute both sides for a small lattice beyond the previously known cases, for instance take n=4 and a lattice L in the nonsplit hermitian space with fundamental invariants (1,1,1,1). On the analytic side ∂Den(L) is a finite explicit sum over intermediate lattices from the weighted lattice-counting formula; on the geometric side Int(L) is the Euler characteristic of an intersection that, by the special-fiber stratification, breaks into products of intersections with algebraic curves. If the two integers differ for any such lattice, the main theorem fails; a direct computation in one such case would settle the identity independently of the paper's induction.","tokens_in":81926,"feed_emoji":"🧮","tokens_out":9160,"duration_ms":88057,"temperature":0.7,"pith_summary":"The paper proves the local Kudla–Rapoport conjecture: for every full-rank lattice in the nonsplit hermitian space over a p-adic field, the arithmetic intersection number of the associated special cycle on a unitary Rapoport–Zink space is equal to the central derivative of the local Siegel series. This is a precise equality between a geometric quantity, the Euler characteristic of a derived intersection of divisors, and an analytic quantity, the derivative at the central point of a representation density. The authors then apply this local identity to prove the global Kudla–Rapoport conjecture and, with previously known archimedean results, cases of the arithmetic Siegel–Weil formula, which relate arithmetic degrees of special cycles on unitary Shimura varieties to central derivatives of incoherent Eisenstein series. If the theorem is right, it supplies the missing nonarchimedean input for a higher-dimensional analogue of Gross–Zagier-type height formulas.","feed_headline":"Local Kudla–Rapoport conjecture proved in all dimensions","feed_subtitle":"Arithmetic intersections now match local density derivatives, unlocking Siegel–Weil cases.","key_machinery":"The load-bearing identity is the equality of the two sides of the local Kudla–Rapoport conjecture, but the mechanism that carries the proof is a Fourier-theoretic recursion on the dimension n of the hermitian space. For a fixed lattice L♭ of rank n−1, the relevant cycles are decomposed into a horizontal part (quasi-canonical lifting cycles, counted explicitly by a local-density formula) and a vertical part (supported on the special fiber). The vertical part gives a compactly supported function on V that is an eigenfunction of the Fourier transform with eigenvalue −1; this 'local modularity' is checked first for the curves in the special fiber and then extended to higher-codimension Deligne–Lusztig strata via the Chern character and cycle-class maps. The uncertainty principle, a simple statement about the local Weil representation, says that no nonzero compactly supported function can vanish together with its Fourier transform on the set of vectors of nonpositive valuation; applying it to the difference Int_{L♭} − ∂Den_{L♭} completes the induction. The analytic side also uses the explicit formula for the local Siegel series as a weighted sum over intermediate lattices together with its functional equation X ↔ 1/X, which makes the central value vanish and the central derivative well defined.","core_discovery":"The central discovery is the identity Int(L) = ∂Den(L) for all lattices L of full rank n in the nonsplit F/F0-hermitian space V. The integer Int(L) is defined by choosing an O_F-basis x1,...,xn of L and taking the Euler characteristic of the derived tensor product of the structure sheaves of the Kudla–Rapoport divisors Z(xi) on the unitary Rapoport–Zink space Nn; the integer ∂Den(L) is the derivative at X=1 of the normalized local Siegel series Den(X,L), the polynomial whose values at X = (−q)^{−k} are normalized local densities of integral representations of the self-dual lattice of rank n+k by L. The proof is an induction on n. Fixing a codimension-one lattice L♭, both sides become functions of x ∈ V \\ L♭_F; the horizontal contributions are matched explicitly using quasi-canonical lifting cycles, and the vertical contributions are shown to be compactly supported and to satisfy the Fourier symmetry f̂ = −f. An uncertainty principle for the local Weil representation then forces the difference to vanish. The vertical Fourier symmetry itself rests on the stratification of the special fiber of Nn into Deligne–Lusztig varieties and on a Tate-conjecture statement for those varieties, proven in the paper by eigenvalue computations. Completed with the almost-self-dual level variant and the semi-global identities, this yields the global Kudla–Rapoport conjecture and the arithmetic Siegel–Weil formula under the stated hypotheses.","pith_inferences":["The same uncertainty-principle induction should adapt to the orthogonal case announced by the authors, and more generally to any setting where the relevant cycles admit a horizontal/vertical split with Fourier-self-dual vertical parts.","One can test the local modularity directly for ramified quadratic extensions: the paper leaves that case open, and a version of the Fourier self-duality f̂ = −f for the vertical cycle would be a concrete necessary condition for an extension of the main theorem.","The Fourier-self-duality of the vertical cycle is essentially a finite-field identity on the special-fiber strata; interpreting it through the Grothendieck–Lefschetz fixed-point formula may yield a shorter proof of the Tate-class input.","The equality of the two cancellation laws, geometric and analytic, suggests that the identity is stable under orthogonal sums with self-dual lattices; this stability may be the right way to formulate a ramified or even characteristic-2 analogue."],"forward_implications":["The local Kudla–Rapoport conjecture holds for every n, not only the non-degenerate cases and the n=3 case that were known before.","The global Kudla–Rapoport conjecture follows: when the set of places where the hermitian space fails to represent T is a single inert prime, the arithmetic degree of the global special cycle equals c_K times the central derivative of the nonsingular Fourier coefficient of the incoherent Eisenstein series.","Cases of the arithmetic Siegel–Weil formula in every dimension follow: under the hypotheses that F/F0 is unramified at finite places and split above 2, and that ϕK is nonsingular at two split places, the generating series of arithmetic degrees equals c_K ∂Eis(z,ϕK), hence is a nonholomorphic hermitian modular form of genus n.","At an almost self-dual level, the corrected intersection number Int(L) equals 1/(q+1) ∂Den_Λ(L); the uncorrected Int′(L) equals 1/(q+1)(∂Den_Λ(L) − Den(L)), conditional on the blow-up description of the auxiliary Rapoport–Zink space.","The proof yields the identity linking arithmetic intersections to local Whittaker derivatives, Int(L) = W′_T(1,0,ϕ0)/(log q^2 · ∏(1−(−q)^{−i})), giving a direct bridge to the Fourier expansion of the Eisenstein series."],"supporting_citations":[{"why":"formulated the local Kudla–Rapoport conjecture and defined the cycles and intersection numbers that are the subject of the paper.","marker":"[KR11]"},{"why":"formulated the global conjecture and the interpolation of local densities to Whittaker functions used in the applications.","marker":"[KR14]"},{"why":"gave the n=3 case and the independence of the intersection number from the choice of basis, used as input in the induction.","marker":"[Ter13a]"},{"why":"supplies the explicit formula for the local Siegel series as a weighted sum over intermediate lattices, the analytic backbone of the derivative.","marker":"[CY18]"},{"why":"provides the stratification of the special fiber used to compute the vertical part of the cycles.","marker":"[VW11]"},{"why":"contains the eigenvalue computation for the cohomology of the Deligne–Lusztig varieties that yields the needed Tate-conjecture statement.","marker":"[Lus76]"},{"why":"supplies the K-theoretic intersection machinery, including filtered Grothendieck groups and projection formulas, used to define and manipulate the derived intersections.","marker":"[Zha19]"},{"why":"gives the archimedean identity that, combined with the local theorem, produces the arithmetic Siegel–Weil formula.","marker":"[Liu11a]"},{"why":"constructs the semi-global integral models of unitary Shimura varieties that carry the global cycles.","marker":"[RSZ17b]"}],"fun_headline_variants":["Kudla–Rapoport conjecture proven for all lattices","Intersections equal density derivatives in every dimension","Local proof unlocks global Siegel–Weil cases","Kudla–Rapoport: local identity, global Siegel–Weil payoff","All-dimension proof of Kudla–Rapoport conjecture"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument depends on the special fiber of the Rapoport–Zink space having the expected stratification into the known algebraic pieces and on the Tate conjecture holding on those pieces; if either failed in the singular cases used for the Fourier transform, the geometric side of the identity would not be established.","fun_headline_variants_meta":{"raw":{"variants":["Kudla–Rapoport conjecture proven for all lattices","Intersections equal density derivatives in every dimension","Local proof unlocks global Siegel–Weil cases","Kudla–Rapoport: local identity, global Siegel–Weil payoff","All-dimension proof of Kudla–Rapoport conjecture"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000999,"raw_usage":{"total_tokens":4229,"prompt_tokens":948,"completion_tokens":3281,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":564,"completion_tokens_details":{"reasoning_tokens":3196}},"tokens_in":564,"tokens_out":3281,"duration_ms":25609,"temperature":1.0,"reasoning_tokens":3196,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:05:24.412780+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute both sides for a small lattice beyond the previously known cases, for instance take n=4 and a lattice L in the nonsplit hermitian space with fundamental invariants (1,1,1,1). On the analytic side ∂Den(L) is a finite explicit sum over intermediate lattices from the weighted lattice-counting formula; on the geometric side Int(L) is the Euler characteristic of an intersection that, by the special-fiber stratification, breaks into products of intersections with algebraic curves. If the two integers differ for any such lattice, the main theorem fails; a direct computation in one such case would settle the identity independently of the paper's induction.","supporting_citations":[],"review_version":1}