{"id":"70fc9e71-f16f-4e8a-b26d-923b563d05c0","arxiv_id":"1908.00961","paper_version":1,"verdict":"CONDITIONAL","confidence":"LOW","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For the Guo-Jacquet relative trace formula, the unipotent contribution is shown to be a finite sum over nilpotent orbits, each term expressed through zeta integrals with a homogeneity scaling law.","lead":"This paper proves a fine expansion for the unipotent part of the Guo-Jacquet relative trace formula, expressing each nilpotent orbit term through zeta integrals and a homogeneity law. The result gives number theorists a computable handle on a key geometric contribution to a trace formula used to study automorphic periods and distinction.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Descent lemma 3.2.5.1 drops the cutoff ζ0: the proof that f is weakly cuspidal appears to require a weighted vanishing that (3.2.2.1) does not supply.","rationale":"The paper's advertised result is Theorem 3.2.4.1, whose proof is a reduction to the infinitesimal statements of §2 via the descent in §3.2.5, with Lemma 3.2.5.1 supplying the necessary weak cuspidality. I checked this reduction carefully. The displayed equality in the proof of Lemma 3.2.5.1 is not justified: the cutoff ζ_0 is dropped, and the domain of integration is effectively replaced by the full quotient N(F_0)/N_H(F_0). This is not cosmetic: f_0 is defined precisely so that it is compactly supported, and the cutoff is essential. The unweighted vanishing (3.2.2.1) is a consequence of very cuspidality at the level of G, but the proof needs the weighted integral to vanish for every smooth compactly supported weight arising from ζ_0; that is a stronger property and is neither stated nor proved. The reader's weakest_assumption identified Theorem 2.6.4.1, which is indeed external and important, but I find the descent gap more immediate because it is internal to the written proof and affects even the fine expansion (assertions 1 and 2), not only the zeta-integral identification (assertion 3). I therefore recommend CONDITIONAL: the manuscript should not be accepted until Lemma 3.2.5.1 is proved with the cutoff retained, or the class of very cuspidal functions is strengthened so that the weighted vanishing holds.","tokens_in":16961,"tokens_out":17168,"duration_ms":166629,"concrete_test":"Recompute the constant term in §3.2.5 retaining the cutoff ζ_0. Explicitly, verify whether f_{P,x}(X) equals c∫_{N(F_0)/N_H(F_0)} ζ_0(C(Ad(x)(X+U(n)))) Φ_{0,S}(ρ(x exp(X/2)n)) dn with a nonconstant ζ_0' arising from the change of variables. A decisive analytic check is to take X = 0 and a nontrivial x, so that the integrand is ζ_0(C(Ad(x)U)) Φ_{0,S}(exp(Ad(x)U)) on s_N(F_0); if ζ_0 is chosen as a smooth bump on c(F_0) that is not identically 1 on the image of the compact support of Φ_{0,S}, and if this weighted integral can be nonzero while the unweighted integral over N(F_0)/N_H(F_0) vanishes, then the appeal to (3.2.2.1) in Lemma 3.2.5.1 is invalid. The simpler algebraic test is to check the displayed equality f_0(Ad(x)(X+U)) = Φ_{0,S}(x exp(X+U)x^{-1}) for U outside the ζ_0 = 1 region and show such U contribute to the constant term.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Lemma 3.2.5.1 is the bridge from the infinitesimal theorems to Theorem 3.2.4.1, and its proof is flawed as written. In §3.2.5, f_i is defined with a cutoff ζ_i on the categorical quotient c(F_v), supported in ω^♭_v and equal to 1 only near c(0). In the proof of the lemma, for X ∈ s_M(F_0) ∩ N(F_0), the displayed equality f_0(Ad(x)(X+U)) = Φ_{0,S}(x exp(X+U)x^{-1}) holds only for U such that Ad(x)(X+U) lies in the region where ζ_0∘C ≡ 1. The integral defining the constant term runs over all s_N(F_0); f_0 is zero outside ω_0, but on the support of Φ_{0,S}, the value of ζ_0(C(Ad(x)(X+U))) need not be 1, since x is arbitrary and the compact support in U grows with x. After the change of variables U ↦ exp(-X/2)exp(X+U)exp(-X/2), the integral becomes c∫_{N(F_0)/N_H(F_0)} ζ_0'(n) Φ_{0,S}(ρ(x exp(X/2)n)) dn with ζ_0' a nonconstant, compactly supported weight. The cited vanishing (3.2.2.1) is the unweighted integral over the full quotient N(F_0)/N_H(F_0); it does not imply the weighted integral is zero for arbitrary smooth compactly supported weights. Hence weak cuspidality of f is not established. Since Theorem 2.3.4.1 and Corollary 2.3.4.2, and therefore assertions 1 and 2 of Theorem 3.2.4.1, presuppose weak cuspidality, this is a load-bearing gap in the central proof.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the unipotent contribution to the Guo–Jacquet relative trace formula for the pair (GL_n(D), GL_n(E)), where E/F is a quadratic extension of number fields and D is a quaternion algebra containing E. For a class of test functions called very cuspidal, the author establishes: (1) a fine expansion of the unipotent contribution as a finite sum of absolutely convergent integrals over nilpotent H(F)-orbits; (2) an expression of each nilpotent integral as the residue at s=0 of a product of a universal holomorphic function θ_X(s) and a zeta integral Z_X(f,s); and (3) a homogeneity law under dilation of the orbit parameter. The proof proceeds by reducing to an infinitesimal setting (functions on the tangent space s), proving analogous statements there using a new truncation method borrowed from the author's previous work, and then descending back via an exponential map and a descent lemma (Lemma 3.2.5.1). The main theorem is Theorem 3.2.4.1.","tokens_in":17385,"tokens_out":7322,"duration_ms":73329,"significance":"If the proof is completed, the result would be a significant contribution to the relative trace formula program: it gives a genuinely computable fine expansion of the unipotent contribution, with explicit zeta integrals and a clean homogeneity law, and it has concrete applications to distinction problems (as noted in Corollary 3.2.4.2 and Remark 1.1.4.5). The paper offers a new truncation-based approach that avoids some unknown global constants, which is a conceptual advance over earlier Arthur-style treatments. However, the current manuscript leaves several load-bearing steps as references to prior work or sketches, and one key descent lemma appears to have a gap in its proof. These issues must be addressed before the central claims can be considered established.","major_comments":[{"comment":"The proof of Lemma 3.2.5.1 is not valid as written. The displayed equality f0(Ad(x)(X+U)) = Φ_{0,S}(x exp(X+U)x^{-1}) omits the cutoff factor ζ0. According to the definition of f0 in §3.2.5, one has f0(Y) = ζ0(C(Y)) Φ_{0,S}(exp(Y)) for Y ∈ ω0, and 0 otherwise. The equality used in the proof holds only when ζ0(C(Ad(x)(X+U))) = 1. This is not guaranteed for all U in the domain of integration, since ζ0 is only assumed to be 1 in a neighborhood of c(0) and the set where the integrand does not vanish involves U for which C(Ad(x)(X+U)) varies over the support of ζ0. After the change of variables U ↦ exp(-X/2) exp(X+U) exp(-X/2) and the identification with N(F_0)/N_H(F_0), the integral becomes a weighted integral ∫ ζ0'(n) Φ_{0,S}(ρ(x exp(X/2)n)) dn with ζ0' a nonconstant, compactly supported smooth weight. The cited vanishing condition (3.2.2.1) applies to the unweighted integral over the full quotient; it does not imply the vanishing of the weighted integral for arbitrary smooth weights. Consequently, the weak cuspidality of f is not established, and the descent to the infinitesimal results, on which Theorem 3.2.4.1 relies, is not justified. This is a load-bearing gap in the proof of the main theorem.","section":"§3.2.5, Lemma 3.2.5.1"},{"comment":"Theorem 2.6.4.1 is a central step: it identifies the truncation-based integral J_O(f_t,s) with θ_X(s) Z_X(f_t,s), from which the homogeneity property (Theorem 2.3.5.1) and assertions 3–4 of Theorem 3.2.4.1 follow. The proof given in the text is only a reference to the analogy with Theorem 9.1.1 of [Cha18b], supplemented by the observation E_G(g)=E_G(σ(g)). This leaves substantial room for error in adapting the lengthy arguments of [Cha18b] to the relative context, especially regarding the definition of the functions E_G and the treatment of the σ-conjugation. The manuscript should provide a complete proof or a precise, verifiable dictionary between the notation and hypotheses of Theorem 9.1.1 of [Cha18b] and the present setting. As it stands, the reader cannot check that the identification is valid, and this weakens the derivation of the homogeneity law.","section":"§2.6.4, Theorem 2.6.4.1"},{"comment":"The convergence theorem for the truncated orbital sums, Theorem 2.4.1.1, is foundational: it is used to prove Theorem 2.3.4.1 and hence the absolute convergence asserted in parts 1–2 of the main theorem. Its proof is only sketched, with references to \"the same kind of majorization\" and to analogous lemmas in [Cha18a] (§3.8, §3.12). Given that the underlying arguments involve delicate estimates for weighted sums over nilpotent orbits in the relative setting, the paper should spell out at least the key majorization steps (e.g., Lemma 2.4.2.2 and the treatment of contributions (2.4.3.1)–(2.4.3.3)) rather than deferring to the prior article, since the current text is not self-contained on a load-bearing point.","section":"§2.4.1, Theorem 2.4.1.1"}],"minor_comments":[{"comment":"The proof of Proposition 2.2.2.1 contains an appeal to a dimension computation that is said to be \"not difficult\" and is only sketched. To make the paper more self-contained and the claim verifiable, the author should either provide the explicit computation of dim(B_Y) and dim(H_Y) in terms of the Jordan type or give a precise reference to the applicable result in [Spa82].","section":"§2.2.2, Proposition 2.2.2.1"},{"comment":"The observation \"E_G(g)=E_G(σ(g))\" is introduced without explanation of which objects are involved and why the identity holds. Since this is the main new ingredient claimed in the adaptation, a short justification or an explicit formula for E_G would improve readability.","section":"§2.6.4, proof of Theorem 2.6.4.1"},{"comment":"The statement \"the zeta function Z_X(f_t,s) is defined in §2.6.3 relatively to any function f_t ∈ C_c^∞(s(A)) such that f_t(Y)=Φ_S(exp(t^{-1}Y))\" is imprecise: for a fixed Φ_S, such a function f_t is not uniquely determined by this identity on the whole s(A), since exp is only a local diffeomorphism. The author should clarify how f_t is chosen (e.g., by the construction of §3.2.5) and why Z_X(f_t,s) is independent of the choice.","section":"§3.2.4, Theorem 3.2.4.1(3)"},{"comment":"The notation [H] is used for H(F)\\H(A), while earlier [H]^1 is the kernel of |det|. In Theorem 2.5.2.1 the integral is over [H], but the limit value J_O(f) is defined via an integral over [H]^1. The role of the extra |det(h)|^s factor is not fully explained; this is standard but a sentence of clarification would help.","section":"§2.5.2, Theorem 2.5.2.1"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is heavily dependent on the author's previous papers [Cha18a] and [Cha18b], and the review would benefit from an explicit verification that the cited theorems indeed apply in the present relative setting, particularly Theorem 9.1.1 of [Cha18b] used in Theorem 2.6.4.1. The gap in Lemma 3.2.5.1 is serious and appears to be more than a presentation issue; if it cannot be repaired, the main theorem may fail as stated, though it is plausible that a more careful choice of the cutoff ζ0 or a modified descent argument can resolve it. The paper is otherwise well motivated and the infinitesimal results are promising."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This is a serious paper and probably an important step, but it has a concrete gap in the descent lemma that blocks the main theorem as stated. The infinitesimal part (Section 2) is well organized and the author is explicit about what depends on prior work, but the bridge to the group setting is where I have a real problem.\n\nWhat is new and good: the paper gives the first fine expansion for the Guo-Jacquet unipotent contribution, expresses each nilpotent orbit term through explicit zeta integrals, and proves a homogeneity law. That is genuinely new and extends Arthur's fine expansion to a relative setting. The structure is clear, and the author explicitly notes where the broader quaternion setting is not written out (remark 1.1.4.3). I see no sign of post-hoc fitting or circularity; the main claims do not reduce to input assumptions.\n\nThe soft spots are real, though. The deferrals to [Cha18b] and [Cha18a] are heavy: theorem 2.6.4.1 is justified by analogy, and convergence is imported. In this field that is common, though it makes the paper hard to read standalone. More seriously, lemma 3.2.5.1—the descent from G to s—has a flawed proof. The function f0 is defined with a cutoff ζ0 which is 1 only near c(0), not on the whole support of ω0. In the proof of the lemma, the equality f0(Ad(x)(X+U)) = Φ0,S(x exp(X+U)x^{-1}) is asserted for all U, but that equality holds only when ζ0∘C equals 1. When you integrate over all of sN(F0), you will in general hit the region where ζ0 is nonconstant, and after the change of variables you get an integral with a nonconstant compactly supported weight. The cited vanishing condition (3.2.2.1) is for the unweighted integral, and it does not imply vanishing of the weighted one. So weak cuspidality of f is not established, and theorem 3.2.4.1 as written lacks support. This is a load-bearing gap, not a minor omission.\n\nThe stress-test note lands. The rest of the argument seems coherent—I see no other red flags—and the reader's low confidence is appropriate. The paper deserves a serious referee; it is too significant to desk reject, but I would not accept the current version. The author can probably fix the lemma with a more careful choice of cutoff or a support argument, but it needs correction.","headline":"Likely an important step, but the descent lemma 3.2.5.1 has a real gap: the cutoff is dropped, so the main theorem is not fully proved as written.","tokens_in":17901,"tokens_out":2698,"would_cite":false,"duration_ms":28146,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11F70","11F72","22E55"],"pacs":[],"model":"deepseek-v4-flash","headline":"The unipotent contribution of the Guo-Jacquet trace formula, for very cuspidal test functions, decomposes into absolutely convergent nilpotent-orbit integrals, each expressed by a zeta integral and obeying a homogeneity law.","keywords":["Guo-Jacquet trace formula","relative trace formula","unipotent contribution","nilpotent orbits","zeta integrals","homogeneity","very cuspidal functions","truncation"],"falsifier":"Compute both sides of theorem 2.6.4.1 explicitly for a small case, for instance n=2, F=Q, E=Q(i), D=M_2(Q), with a weakly cuspidal compactly supported f and a finite place v, comparing J_O(f_t,s) with theta_X(s)Z_X(f_t,s) for the regular nilpotent orbit as |t|_v tends to 0; a disagreement for any sequence t would falsify the theorem. Alternatively, a direct check of the partition-of-unity identity (2.4.4.4) for a non-standard maximal compact subgroup would test the convergence argument underlying the fine expansion.","tokens_in":16758,"feed_emoji":"🧮","tokens_out":8126,"duration_ms":76395,"temperature":0.7,"pith_summary":"For the Guo-Jacquet relative trace formula attached to the pair (GL_n(D), GL_n(E)), where E/F is a quadratic extension of number fields and D is a quaternion algebra containing E, the paper proves a fine expansion of the unipotent contribution on the geometric side. For any very cuspidal test function Phi, the integral over [H]^1 of the sum of Phi_S over nilpotent elements X is absolutely convergent and equals a finite sum over nilpotent orbits O of integrals J_O(Phi). Each such orbit integral is shown to be the residue at s=0 of a product of a fixed holomorphic factor and an Eulerian zeta integral, and to satisfy the homogeneity law $J_O^{{t t0}}$(Phi)=|t|$_v^{{dim(O)/2}}$$J_O^{{t0}}$(Phi) for small |t0|_v and |t|_v <= 1. These results make the unipotent side of the formula tractable for period problems and for proving existence of distinguished automorphic representations.","feed_headline":"Trace formula's unipotent part splits into orbit integrals","feed_subtitle":"A zeta-integral formula makes each term explicit and yields the scaling law periods depend on.","key_machinery":"The argument runs on an infinitesimal reduction combined with a truncation technique. One first replaces the group G by the tangent space s of the symmetric space S=G/H and Phi_S by a weakly cuspidal function f, with very cuspidality ensuring that all relevant constant terms vanish. On s, the key identity is theorem 2.6.4.1, J_O(f_t,s)=theta_X(s) Z_X(f_t,s) for |t|_v sufficiently small, where Z_X(f,s) is the Eulerian zeta integral obtained by integrating f_X^K(A) delta(A,s) over the Levi component L(A), and theta_X is a fixed holomorphic factor. The homogeneity of this zeta integral under scaling, established in lemma 2.6.3.1, has exponent dim(O)/2, and passing to residues through the truncation of [Cha18b] yields the homogeneity law first in the infinitesimal setting and then globally.","core_discovery":"The central result, theorem 3.2.4.1, states that for every very cuspidal function Phi in C_c^infty(G(A)) and every place v, each nilpotent orbit integral J_O^t(Phi)= integral over [H]^1 of the sum over X in O of Phi_S($h^{{-1}}$ exp($t^{{-1}}$X)h) dh is absolutely convergent for all t in F_v^times, the full unipotent integral is the finite sum over O of J_O(Phi), and for sufficiently small |t0|_v one has $J_O^{{t t0}}$(Phi)=|t|$_v^{{dim(O)/2}}$$J_O^{{t0}}$(Phi) whenever |t|_v <= 1. Moreover, J_O^t(Phi) is the limit as s tends to 0+ of s theta_X(s) Z_X(f_t,s), where theta_X is a fixed holomorphic function independent of Phi and Z_X is an Eulerian zeta integral. The proof descends to the infinitesimal symmetric space s, where the analogous statements are proved for weakly cuspidal compactly supported functions.","pith_inferences":["If the same infinitesimal machinery works for G the multiplicative group of any F-simple central algebra containing E, as the paper suggests, the fine expansion and homogeneity law would extend to a wider family of relative trace formulas beyond the quaternionic case.","The homogeneity exponent dim(O)/2, exactly half the F-dimension of the orbit, matches the heuristic reading of these integrals as volumes of tubes around orbits, a geometric interpretation the paper does not develop explicitly.","Because Z_X is Eulerian, the residue formula opens a concrete path toward explicit local computations of the unipotent side; testing it on a small-rank example, such as n=2, would make the abstract expansion quantitative.","The proof uses only the vanishing of constant terms at one place, so the class of test functions should be robust to replacing supercuspidal matrix coefficients by other functions satisfying the same local vanishing condition, potentially widening the applicability."],"forward_implications":["For any very cuspidal Phi, the unipotent contribution is not merely conditionally convergent: each nilpotent orbit contribution converges absolutely, so the sum can be studied term by term.","Each orbit contribution J_O^t(Phi) is recovered as the residue at s=0 of s theta_X(s) Z_X(f_t,s), and since Z_X is Eulerian, the local factors can in principle be computed explicitly.","The homogeneity law J_O^{t t0}(Phi)=|t|_v^{dim(O)/2}J_O^{t0}(Phi) determines the asymptotic behaviour of the unipotent side as the scaling parameter goes to zero.","As a direct corollary, when t tends to 0 the whole unipotent integral is asymptotically vol([H]^1) times the integral of Phi over H(A), isolating the zero orbit as the dominant contribution.","The theorem supplies the analytic foundation needed for applications such as proving the existence of H-distinguished cuspidal automorphic representations with prescribed local supercuspidal components."],"supporting_citations":[{"why":"Supplies the truncation technique, the partition-of-unity identity, the function theta_X, and the holomorphy of the zeta integral Z_X.","marker":"[Cha18b]"},{"why":"Provides the convergence majorizations and the orbit-dimension lemma adapted in the proof of theorem 2.4.1.1.","marker":"[Cha18a]"},{"why":"Gives the theory of induced unipotent classes used to prove the single P-orbit property for nilpotent orbits.","marker":"[LS79]"},{"why":"Provides dimension formulas for flag varieties and centralizers used to compute induced nilpotent orbit dimensions.","marker":"[Spa82]"},{"why":"Classifies nilpotent orbits by Jordan normal form and gives the identification of S' with G/H'.","marker":"[Guo97]"},{"why":"Yields the isomorphism n maps to nXn^{-1}-X used to rewrite the orbit integral as a zeta integral.","marker":"[Cha17]"}],"fun_headline_variants":["Unipotent trace term: orbit integrals are zeta integrals with scaling","Fine unipotent expansion: each orbit term is a zeta integral","Guo-Jacquet unipotent term: nilpotent integrals scale by dimension","Unipotent trace: zeta integrals give scaling law for orbit terms","Orbit integrals in unipotent term: explicit zeta form and scaling"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the local identity J_O(f_t,s)=theta_X(s)Z_X(f_t,s) for small |t|_v, whose proof is only sketched as analogous to an earlier theorem; if this identification fails, the zeta-integral expression and the homogeneity law do not follow.","fun_headline_variants_meta":{"raw":{"variants":["Unipotent trace term: orbit integrals are zeta integrals with scaling","Fine unipotent expansion: each orbit term is a zeta integral","Guo-Jacquet unipotent term: nilpotent integrals scale by dimension","Unipotent trace: zeta integrals give scaling law for orbit terms","Orbit integrals in unipotent term: explicit zeta form and scaling"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001042,"raw_usage":{"total_tokens":4342,"prompt_tokens":866,"completion_tokens":3476,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":482,"completion_tokens_details":{"reasoning_tokens":3382}},"tokens_in":482,"tokens_out":3476,"duration_ms":23563,"temperature":1.0,"reasoning_tokens":3382,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:26:58.473298+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute both sides of theorem 2.6.4.1 explicitly for a small case, for instance n=2, F=Q, E=Q(i), D=M_2(Q), with a weakly cuspidal compactly supported f and a finite place v, comparing J_O(f_t,s) with theta_X(s)Z_X(f_t,s) for the regular nilpotent orbit as |t|_v tends to 0; a disagreement for any sequence t would falsify the theorem. Alternatively, a direct check of the partition-of-unity identity (2.4.4.4) for a non-standard maximal compact subgroup would test the convergence argument underlying the fine expansion.","supporting_citations":[],"review_version":1}