{"id":"bab53e16-bdc0-4956-bcdb-5dadebea1add","arxiv_id":"2507.13473","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For every order r, the r-th central derivative of a corank one Fourier coefficient of a unitary Eisenstein series equals the degree of a corank one special 0-cycle on Hermitian shtukas.","lead":"This paper proves a formula connecting counting problems on special geometric objects called shtukas to derivatives of number-theoretic series called Eisenstein series, for the first singular 'corank one' case. It matters because it extends a central tool in arithmetic geometry to a harder class of inputs, with applications to derivatives of L-functions.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 9.3.1 is asserted without proof in the function-field generality needed for Theorem 11.2.2; a missing q-power or term in that corank-one identity would make the analytic side of the headline formula incorrect even if all geometric computations are right.","rationale":"The reader's CONDITIONAL verdict with MODERATE confidence is the right assessment. I read §§2–10 as a coherent geometric proof; the support theorem and the Hecke-action computations are long but do not exhibit an obvious internal contradiction. The paper's own text flags the two missing proofs: Proposition 9.3.1 and Lemma 11.2.1. Of these, Proposition 9.3.1 is the single load-bearing unverified step for the statement of the main theorem: it is used verbatim in the final deduction, and its normalization is exactly what determines the r-th derivative. The cited m=2 function-field case and the number-field general case do not by themselves certify the general function-field identity. Since the stated formula is exact and involves q-powers depending on d(E0), E_{(E♭,a♭)}, and χ, a small error there would invalidate Theorem 11.2.2 even if every sheaf-theoretic statement is correct. I therefore keep the conditional verdict rather than raising it to accept or lowering it to reject; the concern is about completeness and verification, not a detected contradiction.","tokens_in":74727,"tokens_out":12152,"duration_ms":142771,"concrete_test":"Compute both sides of Proposition 9.3.1 directly from the local Whittaker functions of [FYZ24, (2.15)] for the first unproved case, m=3, over a function field. Concretely, take X=P^1_k, an étale double cover X'→X, choose a degree-one inert point v, set E0=O_{X'}, E♭=O_{X'}^2 with an injective Hermitian a♭ whose v-adic lattice is the standard one, and let E=E0⊕E♭ with the induced corank-one a. Evaluate E_{(E,a)}(s,χ)_3 and E_{(E♭,a♭)}(s+1/2,χ)_2 by the product formula of Proposition 9.2.1 at v and compare the two sides of Proposition 9.3.1 as rational functions in q^{-s} to order 5. If any q-power, sign in s+1/2, or χ(E0) weight differs, the analytic side of Theorem 11.2.2 is not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 11.2.2 is deduced by combining Theorem 11.1.2 (the geometric side, proved through §§2–10) with Proposition 9.3.1. Proposition 9.3.1 is the only bridge from corank-one Fourier coefficients of E(g,s,χ)_n to nonsingular coefficients of E(g,s,χ)_{n-1} at shifted arguments. Its proof in §9.3 consists of a reference to [GS19, §5.2.2], a number-field version in [Che24d, §2.4], and the m=2 function-field case in [CH25, §2.5], followed by the statement that the general calculation 'requires no new ideas' and a deferral to the reader. The exact q-powers q^{(m/2±s) deg E0} q^{±ms deg ω_X} L_m(±s,χ0) in the identity are precisely the delicate normalizations in the function-field setting; a single missing factor, a wrong sign in the second shift, or an omitted χ(E0) weight would change the r-th derivative at s=0 and break the equality with the independently computed geometric side (Theorem 8.3.1, Proposition 11.1.1). Lemma 11.2.1 is also cited rather than proved, but it is a more routine virtual-class comparison; the analytic identity is the point at which an error would be invisible to the rest of the paper's machinery. The central claim is therefore not yet self-contained at its analytic hinge.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a corank-one case of the higher Siegel–Weil formula for unitary groups over function fields. For a rank n bundle E with a Hermitian morphism a of rank n−1, Theorem 11.2.2 identifies the degree of the virtual 0-cycle [Z^r_E(a)]^{vir} with the r-th central derivative of the corank-one Fourier coefficient of an unramified Siegel–Eisenstein series on U(n,n), up to explicit q-powers, an L-factor, and a character value. The proof has two independent sides: the geometric side (Sections 2–8), based on an enhanced Hitchin fibration, a support theorem, Hecke correspondences, and the Grothendieck–Lefschetz trace formula, and the analytic side (Sections 9–11), based on density polynomials, twisted density polynomials, and a reduction of corank-one Fourier coefficients to nonsingular coefficients on a lower-rank group. The paper also derives applications to intersection multiplicities on Sht^r_{U(2)} and to the modularity conjecture for special cycles.","tokens_in":74992,"tokens_out":5314,"duration_ms":67250,"significance":"If the main theorem is correct, it is a substantial extension of the non-singular higher Siegel–Weil formula of Feng–Yun–Zhang to the first singular case, and it holds for all derivative orders r rather than only for the leading term. The paper contains a great deal of original and intricate geometry: the complementary line trick, the support theorem for the enhanced Hitchin fibration, the explicit computation of three Hecke actions, and the geometrization of twisted density polynomials in terms of Springer sheaves. The final comparison is explicit and the two sides are computed independently, so the result is genuinely falsifiable. The main weakness is that two load-bearing analytic/geometric reductions are deferred to external references or to arguments described as routine, with Proposition 9.3.1 being the most serious gap because the exact q-power normalizations there are essential for the final equality.","major_comments":[{"comment":"This proposition is the only bridge from corank-one Fourier coefficients of E(g,s,χ)_n to nonsingular coefficients of E(g,s,χ)_{n−1}, and it is used directly in the proof of Theorem 11.2.2 after equation (11.2.3). The text defers the proof: it cites [GS19, §5.2.2], [Che24d, §2.4], and the m=2 case [CH25, §2.5], and then states that the calculation 'requires no new ideas' and leaves it to the reader. The identity involves delicate normalizations q^{(m/2±s) deg E_0} q^{±ms deg ω_X} L_m(±s, χ_0) and a χ(E_0) weight; even a single missing q-power or a wrong sign in the second shift would change the r-th derivative at s=0 and break the equality with the independently computed geometric side of Theorem 8.3.1. I therefore ask that a complete proof of Proposition 9.3.1 in the exact function-field generality used here be included in the revision.","section":"§9.3, Proposition 9.3.1"},{"comment":"The reduction of the corank-one cycle class to a nonsingular cycle class, equation (11.2.1), is cited from [CH25, Lemma 3.2.5], with the comment that the proof of that reference works 'essentially verbatim' under the more general hypotheses of this paper. This lemma determines the Chern-class factor ∏ c_1(p_i^*σ^*E_0^{-1}⊗ℓ_i) multiplying [Z^r_{E^♭}(a^♭)]^{vir}, and that factor is precisely the geometric side of Theorem 11.2.2. Since the hypotheses here do not include the orthogonal splitting assumed in [CH25], the revision should include a proof of (11.2.1), or an explicit verification that the cited lemma applies verbatim in the present setting.","section":"§11.2, Lemma 11.2.1"},{"comment":"The final paragraph of the proof of Theorem 10.1.1 reduces the general case to a multiplicativity claim for the right-hand side of (10.4.1) with respect to support decompositions, and then says that the proof 'involves no new ideas' and omits it. This multiplicativity is needed to obtain the global twisted density polynomial that enters Proposition 11.1.1 and hence the final comparison. Please supply the argument, or a precise reference with hypotheses matching the situation here, rather than a deferred sketch.","section":"§10.4, proof of Theorem 10.1.1"}],"minor_comments":[{"comment":"The abstract contains a typo: 'unit ary groups' should read 'unitary groups'.","section":"Abstract and title page"},{"comment":"The discussion of Ryan Chen's work would be clearer if the papers [Che24a–d] were identified by title or by the specific corank-one result being used; currently the reader must guess which of the four references contains Proposition 9.3.1's number-field version.","section":"§1.3"},{"comment":"The informal 'vector cross product' remark is entertaining, but the phrase 'The vector cross product of freshman physics' should be reworded for a mathematical journal.","section":"§2.3, Remark 2.3.3"},{"comment":"Corollary 11.3.4 is stated as a consequence of Theorem 1.1.1 but its proof is essentially a reference to [CH25] and it is conditional on Conjecture 11.3.2; the displayed corollary should be labeled as conditional in the statement, not only in the preceding paragraph.","section":"§11.3"}],"recommendation":"major_revision","confidential_remarks":"The paper is a serious and substantial contribution if the main theorem is correct, but in its current form it is not self-contained at the analytic hinge: Proposition 9.3.1 is used to prove the headline formula and is explicitly deferred. I would not recommend acceptance until that proposition is proved in the text or matched exactly to a published statement in the function-field setting. The extensive reliance on author-overlapping references ([FYZ24], [FYZ25], [FYZa], [FYZb], [CH25]) is understandable in a series, but for a referee report the missing proofs in this manuscript are the relevant issue."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a genuine advance, but as written it is a conditional one. The corank-one higher Siegel–Weil formula is new—[FYZ24] only treated nonsingular coefficients, and the proof strategy is genuinely different from Chen's limiting argument. The enhanced Hitchin fibration, the complementary line trick, the support theorem, and the geometrization of the twisted density polynomials are substantial pieces of work, and the geometric side is developed in serious detail. The final comparison is explicit, and there are no fitted parameters: the two sides are defined independently.\n\nThe soft spot is exactly Proposition 9.3.1. It is the only bridge from corank-one Fourier coefficients of E(g,s,chi)_n to nonsingular coefficients of E(g,s,chi)_{n-1}, and it is not proved in the text. The paper cites [GS19, §5.2.2], [Che24d, §2.4], and [CH25, §2.5], then says the general calculation requires no new ideas. That is a load-bearing normalization; a wrong q-power or sign in the shifted arguments would break Theorem 11.2.2 even if every geometric computation is correct. The stress-test note is right to flag this. Lemma 11.2.1 is also cited to [CH25] rather than proved, though it is a more routine virtual-class comparison. There is a smaller gap inside Theorem 10.1.1, where multiplicativity over the support decomposition is asserted with details omitted; that looks repairable but is still a gap.\n\nThe applications are honestly labelled as conditional on Conjecture 11.3.2 and forthcoming work [FYZb], so that is not a flaw in the main theorem. I do not see circularity: the geometric degrees and the analytic derivatives are computed separately. The heavy overlap with [FYZ24] is natural and not a problem.\n\nWho is this for? Arithmetic geometers and number theorists working on Siegel–Weil formulas, shtukas, and higher-derivative analogues of Kudla's program. It deserves a serious referee. I would send it out, but the report should be conditional in the same way as my verdict: the referee should require either a proof of Proposition 9.3.1 or a precise statement with all q-powers, chi(E_0) factors, and L-factor normalizations written out. If that identity checks out, the main theorem stands.","headline":"Real corank-one advance with an unproved analytic hinge; referee should demand the details of Proposition 9.3.1.","tokens_in":75549,"tokens_out":2537,"would_cite":true,"duration_ms":30365,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11F46","11F70","11G09","14C17"],"pacs":[],"model":"deepseek-v4-flash","headline":"The corank-one higher Siegel–Weil formula is proved for unitary groups over function fields: for every r, the degree of a corank-one virtual special 0-cycle on Hermitian shtukas equals the rth central derivative of the associated…","keywords":["higher Siegel–Weil formula","Hermitian shtukas","Siegel–Eisenstein series","corank one","special cycles","Hitchin fibration","Springer sheaves","twisted density polynomials"],"falsifier":"Compute both sides of Theorem 1.1.1 in the smallest explicit case — for example $n=2$, $r=2$, with a chosen double cover $X' \\to X$ and a corank-one Hermitian pair $(E,a)$ — where the left side is the finite degree of a virtual 0-cycle and the right side is an explicit derivative of a product of a Dirichlet $L$-function and a twisted density polynomial; the claim is then a rational-number equality that can be checked directly, and any mismatch would localize to the unproved identity of Proposition 9.3.1, the only step the paper defers.","tokens_in":74486,"feed_emoji":"🧮","tokens_out":20187,"duration_ms":208734,"temperature":0.7,"pith_summary":"This paper proves the corank-one case of the higher Siegel–Weil formula for unitary groups over function fields. For every order $r \\ge 0$ of differentiation, it equates the degree of a virtual special 0-cycle $[Z^r_E(a)]^{\\mathrm{vir}}$, attached to a rank $n$ bundle $E$ with a Hermitian map $a$ of rank $n-1$, with the $r$th central derivative of the $(E,a)$-Fourier coefficient of an unramified Siegel–Eisenstein series on the quasi-split unitary group $U(n,n)$. The identity holds uniformly for all $r$, even when the Eisenstein series vanishes to order greater than $r$, so it is a genuine equality of functions rather than a leading-term comparison. Earlier work proved only the non-singular (corank zero) case, and corank one is the first singular case in which the special cycles are proper enough for degrees to exist; the paper also records applications to intersection numbers of special cycles on $\\mathrm{Sht}^r_{U(2)}$ and, conditionally on a modularity conjecture, to $r$th derivatives of twisted base-change $L$-functions.","feed_headline":"For every r, corank-one cycles match Eisenstein derivatives","feed_subtitle":"The first singular case of the higher Siegel–Weil formula now holds for every order r, past the non-singular case.","key_machinery":"The object that carries the argument is the enhanced Hitchin fibration $f: M \\to A \\times \\mathrm{Bun}_{U^\\dagger(1)}$, whose base $A$ parametrizes pairs $(E,a)$ (vector bundle plus injective Hermitian map) and whose source parametrizes embeddings of $E$ into a Hermitian bundle $F$, with $\\mathrm{Bun}_{U^\\dagger(1)}$ the twisted moduli stack of rank-one Hermitian bundles. The complementary line trick upgrades an embedding $E \\to F$ to an isometric embedding $E \\oplus D \\hookrightarrow F$, where the line bundle $D = \\det(\\sigma^*E) \\otimes \\det^\\dagger(F)$ is built from the twisted determinant; this makes the regular-semisimple fibers explicit $(\\mathbb{P}^1)^d$-bundles. The support theorem (Theorem 3.1.2) then identifies $Rf_*\\mathbb{Q}_\\ell$ with $\\bigoplus_{i=0}^d K^i_d \\langle -i \\rangle \\boxtimes \\mathbb{Q}_\\ell$, where each $K^i_d$ is a full-support Springer-theoretic intersection complex on $A$ (a $W_i \\times W_{d-i}$-invariant part of a Hermitian Springer sheaf); full support is the key structural fact that replaces the small-morphism argument available in the non-singular case. On top of this, the Hecke operator is decomposed into three pieces pulled back from the two factors and from their product, whose semisimplified actions are scalar multiples of a single involution $w$ on the cohomology of $\\mathrm{Bun}_{U^\\dagger(1)}$ (Theorems 4.5.2–4.5.4). Finally, Theorem 10.1.1 geometrizes the twisted density polynomials used on the analytic side: the polynomial $\\mathrm{Den}_\\eta(T,E)$ is the Frobenius-trace generating function of the very same sheaves $K^i_d$, which is what makes the geometric and analytic trace computations match.","core_discovery":"The paper's central result, Theorem 11.2.2 (Theorem 1.1.1), states that for any unramified Hecke character $\\chi$ with $\\chi_0 = \\eta^n$, the stack $Z^r_E(a)$ is proper over $k$ and\n$$\\deg [Z^r_E(a)]^{\\mathrm{vir}} = \\frac{1}{(\\log q)^r}\\, $q^{{\\frac{n}}${2}d(E)} \\chi(\\det E)^{-1} \\frac{d^r}{ds^r}\\bigg|_{s=0} \\left( $q^{{ns \\deg_X \\omega_X}}$ L_n(s,\\chi_0)\\, E_{(E,a)}(s,\\chi)_n \\right).$$\nThe path to it runs through a second theorem (Theorem 1.1.4, in the more general form Theorem 11.1.2 with an auxiliary line bundle $E_0$): for a rank $n-1$ bundle with injective Hermitian map, the degree of the cycle capped by the Chern classes of the $r$ tautological bundles is an off-center $r$th derivative, at shift $1/2$, of the Eisenstein series on the lower-rank group $U(n-1,n-1)$, vanishing for odd $r$. The corank-one statement is then assembled from Lemma 11.2.1, which factors a corank-one virtual class as tautological Chern classes times a non-singular class, and Proposition 9.3.1, which expresses corank-one Fourier coefficients of the rank-$n$ Eisenstein series through non-singular coefficients of the rank-$(n-1)$ series at the shifted arguments $s \\pm 1/2$.","pith_inferences":["The corank-one theorem is assembled structurally — a geometric factorization plus an analytic two-term identity — so a reader who wants corank two can already write down the expected shape: a longer alternating sum of non-singular coefficients at several shifted arguments, constrained by the functional equation $s \\mapsto -s$, before any geometric theorem exists.","Because the equality holds for every $r$ at once, it is a statement about the full Taylor expansion and not its first nonzero term; this suggests the underlying perverse-sheaf isomorphisms (support theorem, Hecke actions, twisted density geometrization) should hold canonically on the nose, rather than only up to the semisimplifications used in the trace computations.","The paper's geometry is described as resembling the non-singular terms of the symplectic/orthogonal case; if the support theorem is as insensitive to the underlying group as the reduction to the split case suggests, the same enhanced-Hitchin method should yield corank-one formulas, with the corresponding twisted density polynomials, for orthogonal and symplectic groups as well."],"forward_implications":["For every fixed $r$, the corank-one special-cycle degree is given by the $r$th central derivative formula, uniformly in $r$ and independently of the order of vanishing of the Eisenstein series.","Both sides vanish for odd $r$; nontrivial odd-$r$ identities require similitude twists, as outlined in Section 11.4 of the paper.","Intersections of a corank-one cycle with a fixed non-singular cycle on $\\mathrm{Sht}^r_{U(2)}$ are computed by the corresponding doubled-kernel derivative (Corollary 11.3.1), and the resulting numbers are Fourier coefficients of an unramified automorphic form on $U(1,1)$ (Corollary 11.3.3).","Assuming the Modularity Conjecture of the companion paper, the pairing of the arithmetic theta lift $\\vartheta_{r,\\chi}(f)$ with a special cycle equals an $r$th central derivative of the twisted base-change $L$-function $L(s+\\tfrac12, BC(\\pi)\\otimes\\chi)$ (Corollary 1.2.1): a higher-derivative, function-field version of the arithmetic Rallis inner product formula.","The methods — complementary line trick, support theorem, three-part Hecke decomposition, twisted density geometrization — are presented as the template for a general corank formula; the paper isolates the properness of higher-corank cycles as the obstruction to even formulating it (Remark 1.1.2)."],"supporting_citations":[{"why":"Predecessor that proves the non-singular (corank zero) higher Siegel–Weil formula; supplies the Hitchin-space and Springer-sheaf constructions, the trace formula, the density-polynomial geometrization, and the case $\\chi_0 = \\eta^m$ of Proposition 9.2.1 that this paper extends.","marker":"[FYZ24]"},{"why":"Companion paper that constructs the special cycles $Z^r_E(a)$, their virtual fundamental classes, and the Modularity Conjecture used in the applications; also the source of the Hermitian stacks and tautological bundles used throughout.","marker":"[FYZ25]"},{"why":"Provides Lemma 11.2.1 (the factorization of the corank-one virtual class into tautological Chern classes times a non-singular class), the $m=2$ function-field case of Proposition 9.3.1, and the intersection formulas behind Corollaries 11.3.1 and 11.3.3.","marker":"[CH25]"},{"why":"Number-field corank-one calculation: supplies the $\\chi_0 = \\eta^{m+1}$ local Whittaker formula used in Proposition 9.2.1 and the corank-one Fourier-coefficient analysis that Proposition 9.3.1 cites.","marker":"[Che24d]"},{"why":"Gives, in Section 5.2.2, the general method for expanding corank-$r$ Fourier coefficients of Eisenstein series in terms of non-singular coefficients of lower-rank Eisenstein series, the template for Proposition 9.3.1.","marker":"[GS19]"},{"why":"Number-field arithmetic Siegel–Weil formula for non-singular terms; its local geometry (cited in Section 1.4.2 for the $\\mathbb{P}^1$-bundle phenomena) is the stated analogue against which the present geometric setup is compared.","marker":"[LZ22]"}],"fun_headline_variants":["Corank-one Siegel–Weil formula holds for every r","All r: corank-one cycles match Eisenstein derivatives","Corank-one Eisenstein derivatives are cycle degrees for all r","Beyond corank zero: Siegel–Weil corank-one for every order"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is Proposition 9.3.1 (Section 9.3), the identity that rewrites a corank-one Fourier coefficient of the rank-$n$ Eisenstein series as a sum of two non-singular coefficients of the rank-$(n-1)$ series evaluated at $s \\pm 1/2$; the paper states it and defers the proof, saying the calculation needs no new ideas and citing the number-field treatment and the $m=2$ function-field case. If that identity carries a wrong $q$-power or a missing term, the analytic side of Theorem 1.1.1 would not match the geometric side even if every sheaf-theoretic computation is correct.","fun_headline_variants_meta":{"raw":{"variants":["Corank-one Siegel–Weil formula holds for every r","All r: corank-one cycles match Eisenstein derivatives","Corank-one Eisenstein derivatives are cycle degrees for all r","Beyond corank zero: Siegel–Weil corank-one for every order"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000833,"raw_usage":{"total_tokens":3672,"prompt_tokens":1018,"completion_tokens":2654,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":634,"completion_tokens_details":{"reasoning_tokens":2581}},"tokens_in":634,"tokens_out":2654,"duration_ms":21650,"temperature":1.0,"reasoning_tokens":2581,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T16:24:07.603614+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute both sides of Theorem 1.1.1 in the smallest explicit case — for example $n=2$, $r=2$, with a chosen double cover $X' \\to X$ and a corank-one Hermitian pair $(E,a)$ — where the left side is the finite degree of a virtual 0-cycle and the right side is an explicit derivative of a product of a Dirichlet $L$-function and a twisted density polynomial; the claim is then a rational-number equality that can be checked directly, and any mismatch would localize to the unproved identity of Proposition 9.3.1, the only step the paper defers.","supporting_citations":[],"review_version":1}