{"id":"437eac66-4e82-4f81-a140-6867994ad37c","arxiv_id":"2501.10418","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A twisted first moment of triple product L-functions satisfies a spectral reciprocity formula, yielding a level-aspect subconvexity bound with saving 225/2624 unconditionally.","lead":"This paper derives a spectral reciprocity formula for a twisted first moment of triple product L-functions over number fields. It then uses this formula with the amplification method to prove subconvexity bounds in the level aspect, saving about 1/11.7 of the convexity exponent.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The subconvex bound in Theorem 1.2 rests on the cited local period lower bound H(pi3,q) >> 1/q for the exact test vector in Section 5.3; if [Hu17, Thm 4.1] does not apply to this shifted vector and basis, the amplified-lower-bound step in Section 6.1 fails.","rationale":"The reader's weakest assumption correctly identifies the local period factors in Section 5.3 as the linchpin of the amplification argument. I agree that the unproved lower bound H(pi3,q) >> 1/q is load-bearing, and a mismatch between the test vector used in this paper and the one covered by [Hu17, Theorem 4.1] would invalidate the subconvexity claim. However, I would sharpen the failure mode: ell_v is defined as a sum over an orthonormal basis of local integrals, so it is nonnegative by construction; the real risk is vanishing or an incorrect size (q^{-1-delta} instead of q^{-1}), not a sign change. The archimedean lower bound from Proposition 5.1 is secondary, since pi_infty is fixed in Theorem 1.2 and any positive lower bound would be absorbed into the implied constant; the principal-series hypothesis only restricts the class of pi3 covered. The rest of the paper's reciprocity calculation is internally consistent, and the moment bound in Theorem 1.1 is deferred to [Zac20] rather than derived, which is an omission but not a detected contradiction. A direct local computation for the simplest ramified case would settle whether the cited lower bound applies, and the verdict should remain conditional pending that verification.","tokens_in":18415,"tokens_out":30404,"duration_ms":300123,"concrete_test":"Take F=Q, q=p (m=1), pi1 and pi2 unramified principal series at p, and pi3 the principal series of trivial central character and conductor p. Evaluate the local factor ell_v defined in Section 5.3 directly: compute the local integral in (3.1)-(3.3) with psi_v the unique (up to phase) K0(p)-invariant vector in pi3,v, with phi_{2,v}^q = a(vartheta_p) phi_{2,v}, and sum over the orthonormal basis B(pi,p). Verify that this ell_v is positive and equals c/p with c>0, as asserted by [Hu17, Theorem 4.1]. If instead it vanishes or has size p^{-1-delta}, the q^{-1} lower bound in Section 6.1 fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 6.1 isolates the target L-value through the inequality C q^{-1} A(pi3) L(...)/Lambda(1,pi3,Ad) f(pi3,infty) <= M_A. The factor q^{-1} comes from H(pi,q) >> 1/q for a representation of conductor q, where H is defined in Section 5.3 via ell(pi,q) = product_{v|uvq} ell_v. This lower bound is not proved in the paper; it is quoted from [Hu17, Theorem 4.1] for v|q and [BJN24] for v|uv. The paper asserts that for c(pi_v)=m >= 1 one has product_{v|q} ell_v ≈ 1/q, but it does not verify that the hypotheses of [Hu17, Theorem 4.1] match the local data of the period P_q in Sections 5.1-5.3: the vector in the pi2-factor at v|q is not the unramified newvector but the translate phi_2^q = a(vartheta^m) phi_2, and the basis B(pi,uvq) runs over all K0(uvq)-invariant vectors, including old vectors when c(pi_v)<m. Positivity of ell_v is automatic from its definition as a sum of squared local integrals, so the delicate point is the exact size and nonvanishing of the target local factor. If the target factor vanishes or has size q^{-1-delta} instead of q^{-1}, the lower bound yields L << q^{1-delta0+delta}, which is no longer subconvex. A second cited input, Proposition 5.1's archimedean lower bound, is less critical for the stated theorem because pi_infty is fixed and any positive lower bound depending on pi_infty would be absorbed into the implied constant.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims a spectral reciprocity formula for the twisted first moment of triple product L-functions over a number field, and derives level-aspect subconvexity bounds for L(1/2, π1⊗π2⊗π3) and for L(1/2, π1⊗π2⊗χ). The central objects are a symmetric period P_q(l,Φ,Φ) built from fixed cuspidal representations π1, π2 and a varying representation π3 of conductor q, expanded according to the spectral decomposition of level uvq. The reciprocity relation (5.8) re-expresses this period in terms of lower-level periods, leading through Ichino's formula to the moment bound in Theorem 1.1. Amplification then gives the subconvex bounds in Theorems 1.2 and 1.3. The proof relies on several external inputs: the local period lower bounds from [Hu17] and [BJN24], the archimedean lower bound of [MV10], and the technical treatment of Eisenstein contributions following [Zac20] and [Blo12].","tokens_in":18802,"tokens_out":22812,"duration_ms":212068,"significance":"If the gaps identified below are repaired, the paper would provide a new level-aspect subconvex exponent for triple product L-functions over number fields: an unconditional saving exponent 225/2624 and a saving 1/6 under the Ramanujan–Petersson conjecture. The reciprocity formula itself is a useful extension of Zacharias's framework and could be of independent interest. The manuscript is explicit about its exponents, has no fitted parameters, and credits the main inputs clearly. However, the central moment bound and the key local lower bound are not proved in the text but quoted or deferred, so the significance cannot be assessed without additional verification.","major_comments":[{"comment":"Theorem 1.1 is the central moment estimate on which the two subconvexity theorems rest, yet its proof is not contained in the manuscript. After inequality (5.18) the text states that the estimation of Theorem 1.1 can be achieved from the discussion in Sections 4 and 5 (see also [Zac20, Section 4.5]), but the passage from (5.18) to (1.4) is not carried out: one must show that the generic part G_q(l,Φ,Φ) in (5.4) equals the moment M(π1,π2,q,l) up to the explicit constants using identity (5.20) and the definition H(π,q)=ℓ(π,q)/(2Δ_F^{1/2}), and the Eisenstein contribution in (1.2) must be bounded by the same method. This is a load-bearing step, and a reference to another paper cannot replace the proof.","section":"End of §5 / Theorem 1.1"},{"comment":"The lower bound H(π,q) ≫ 1/q for a representation of conductor q is quoted from [Hu17, Theorem 4.1] (for v|q) and [BJN24] (for v|uv), but the manuscript does not verify that the hypotheses of those theorems match the local data used here. In the period P_q(l,Φ,Φ), the vector in the π2-factor at v|q is the translate φ_2^q = a(ϖ_v^m)φ_2, not the unramified newvector, and the basis B(π,uvq) runs over all K0(uvq)-invariant vectors, including old vectors when c(π_v)<m. The amplified lower bound in §6.1 uses exactly the factor C q^{-1} A(π3) L(1/2,π1⊗π2⊗π3)/Λ(1,π3,Ad) f(π3,∞) ≤ M_A; if the product of local factors were q^{-1-δ} instead of q^{-1}, the resulting exponent would be 1 - δ0 + δ, which is no longer subconvex. A precise local statement and a verification of its hypotheses are required.","section":"§5.3 and §6.1"},{"comment":"The symmetric relation (5.8) is the key mechanism behind the reciprocity formula, but it is asserted without derivation. The text says only that the relation is obtained by grouping the Hecke translation in two different ways. Since this identity is not proved and is presented as a generalization of [Zac20, Equation 4.9], the author should provide a full derivation or a precise reference together with a verification that the additional q-translation and the q1^{-1} correction term are consistent with the definitions of Ψ1 and Ψ2 in (5.9).","section":"§5.2, Eq. (5.8)"},{"comment":"The proof of Theorem 1.3 is only sketched. The text acknowledges that the quadratic-character case t=0 creates a zero of order two in the quotient L(1/2+it,π1⊗π2⊗ω)L(1/2−it,π1⊗π2⊗ω)/Λ*(1,πω(it),Ad) and states that 'one can overcome this obstacle by an application of Holder's inequality, as in [Blo12, Section 4]'. This is not a proof of Theorem 1.3; the relevant argument from [Blo12] must be reproduced or stated as a lemma with its hypotheses verified. The same applies to the treatment of the continuous contribution in Theorem 1.1.","section":"§6, proof of Theorem 1.3"},{"comment":"The implied constant in (1.5) depends on π∞, which includes the archimedean component of π3. Since q ranges over integral ideals while π3 varies among representations of conductor q, the archimedean type is not fixed; without an upper bound on c(π∞) in terms of q, the final use of Proposition 5.1 and (2.11) yields a bound only for each fixed archimedean type. The statement should explicitly quantify over π∞ or prove a polynomial dependence in c(π∞).","section":"Theorem 1.2, uniformity in the archimedean type"}],"minor_comments":[{"comment":"There are several typos and minor errors: 'APPLICA TIONS' in the title, 'Propostition' in Proposition 5.1, 'intergers' in Section 2, 'Cauchy-Schwartz' for Cauchy-Schwarz, and 'Holder' should be 'Hölder'. Also, 'θi-tempered' should be 'θ_i-tempered'.","section":"Throughout"},{"comment":"In the proof of Proposition 4.2, the equality ||φ2^l||_{L4}=||φ2||_{L4} follows from right-translation invariance of the quotient measure; this should be stated explicitly, since as written the proof appears to drop the l without justification.","section":"§4.1, Proposition 4.2"},{"comment":"The sentence 'If 0 ≤ c(π_v) ≤ m−1, by definition, it is known that ℓ_v ≥ 0' is confusing: nonnegativity follows from ℓ_v being a sum of squared moduli, but the lower bound H(π,q)≫1/q additionally requires nonvanishing, which is not addressed for old vectors.","section":"§5.3"},{"comment":"The notation u and v is used both for integral ideals and for their norms. Since this is a common source of confusion, the ideals should be typeset in a distinct font (e.g., fraktur) throughout.","section":"Abstract and §1"},{"comment":"The chain of inequalities '1/6 > 1/7.5 > ...' is difficult to parse. It would be clearer to state that 225/2624 ≈ 0.0857, 625/5696 ≈ 0.1097, and 25/192 ≈ 0.130, so the claimed savings are ordered accordingly.","section":"Remark 1.4"},{"comment":"After the choice L = q^{(1/2−θ)/(3−2θ1−2θ2)}, the text claims that q^{1/7} < L ⩽ q^{1/6}. This is not correct for all admissible parameters (e.g., θ=7/64 and θ1=θ2=0 gives L = q^{25/192} < q^{1/7}). The argument only needs L ≍ q^α with 0 < α < 1 and the stated bounds q^{1/100} < L < q.","section":"§6.1, choice of L"}],"recommendation":"major_revision","confidential_remarks":"The manuscript relies heavily on [Zac20, Section 4.5] for the proof of its main moment estimate, and on [Hu17] and [BJN24] for the local period lower bounds. The relationship with the author's companion paper [Miao24] should also be clarified, since both papers treat subconvexity for triple product L-functions in the level aspect and there is overlap in the methods and results. I recommend that acceptance be conditional on the author providing the missing details of the proof of Theorem 1.1, the verification of the local lower bound, and a self-contained treatment of Theorem 1.3."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: a real but incremental piece of work. The author generalizes Zacharias's spectral reciprocity to a twisted first moment of triple product L-functions over number fields and derives a level-aspect subconvexity bound for L(1/2, π1⊗π2⊗π3) with saving 225/2624 (or 1/6 under Ramanujan). The reciprocity formula (5.15) is the genuine new object, and the amplification section is standard.\n\nThe paper is clearly organized and the spectral-theoretic setup is sound. The constant terms and Eisenstein contributions follow the expected pattern, and the author is explicit about where he quotes rather than proves. The bound for Hecke-character twists (Theorem 1.3) is a useful addition, and the comparison with Ghosh and Sun is reasonable.\n\nThe soft spots are real. First, Theorem 1.1, the engine of the paper, is not proved; after Section 5.4 the text says the bound 'can be achieved' from Sections 4–5 and refers to Zacharias's Section 4.5. A referee will want to see that argument spelled out. Second, the crucial lower bound H(π,q) ≫ 1/q for the exact test vector is quoted from Hu and Blomer–Jana–Nelson. The question is whether Hu's Theorem 4.1 applies to the translated vector φ2^q and to a basis that includes old vectors. That is a fair challenge; the paper does not check the hypotheses, and if the bound were weaker by a q^{-δ}, the amplified lower bound would lose subconvexity. I have no counterexample, but this needs a concrete verification. Third, the paper does not compare its result with [HMN23] and [BJN23], which prove stronger bounds in related settings. A subconvexity paper should position itself against those.\n\nWho is this for? People working on subconvexity via period integrals, especially those using Zacharias's reciprocity. It is not a breakthrough, but it is a legitimate contribution. It deserves a serious referee, though the referee should press for the deferred proof and the local lower bound. I might bring it to a reading group; I would not cite it as established until the gaps are filled.","headline":"A competent Zacharias-style reciprocity argument for the triple product first moment over number fields, yielding a modest subconvexity saving; the result is plausible but the proof leans heavily on cited inputs and lacks a positioning against stronger recent bounds.","tokens_in":19377,"tokens_out":8007,"would_cite":false,"duration_ms":73443,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11F70","11M41","11F72"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves a subconvexity bound for triple product L-functions in the level aspect over any number field, with an unconditional saving of 225/2624 in the exponent.","keywords":["subconvexity","triple product L-functions","spectral reciprocity","level aspect","period integrals","automorphic forms","PGL(2)","amplification method"],"falsifier":"For a finite place $v\\mid\\mathfrak{q}$, compute the normalized local triple product period factor $\\ell_v$ from (3.3) over an orthonormal basis of $K_{0}(\\mathfrak{p}_v^{m})$-invariant vectors for a representation of conductor $\\mathfrak{p}_v^{m}$, for instance a Steinberg or principal-series representation; a zero or negative value would falsify the lower bound $H(\\pi,\\mathfrak{q})\\gg 1/q$ on which the proof relies. A coarser check is to test Theorem 1.1 numerically for small $\\mathfrak{q}$ and $\\mathfrak{l}$: any violation of the stated moment bound would invalidate the amplification argument that produces Theorem 1.2.","tokens_in":18170,"feed_emoji":"📐","tokens_out":9028,"duration_ms":70215,"temperature":0.7,"pith_summary":"The paper studies the first moment of triple product L-functions $L(\\tfrac12, \\pi \\otimes \\pi_1 \\otimes \\pi_2)$ over automorphic representations $\\pi$ of $\\mathrm{PGL}_2$ whose conductor divides a growing ideal $\\mathfrak{q}$, twisted by the Hecke eigenvalue $\\lambda_\\pi(\\mathfrak{l})$. Following a method due to Zacharias, it derives a spectral reciprocity formula expressing this moment in terms of spectral expansions of triple product periods over representations of conductor dividing the twisting ideal $\\mathfrak{l}$, instead of $\\mathfrak{q}$. The formula, combined with the amplification method, yields a subconvex bound for an individual central value $L(\\tfrac12, \\pi_1 \\otimes \\pi_2 \\otimes \\pi_3)$ in the level aspect: $q^{1 - (\\tfrac12-\\theta)(1-2\\theta_1-2\\theta_2)/(3-2\\theta_1-2\\theta_2)+\\varepsilon}$. With the current exponent toward the Ramanujan-Petersson conjecture ($\\theta=\\theta_1=\\theta_2=7/64$), the saving is $225/2624$ unconditionally; under the Ramanujan-Petersson conjecture it becomes $1/6$. The result is the first level-aspect subconvexity of this strength for triple product L-functions over number fields, and it also covers twists by unitary Hecke characters.","feed_headline":"Triple product L-functions get a subconvexity saving of 225/2624","feed_subtitle":"A reciprocity formula ties the level-q moment to level-l data and breaks the convexity barrier.","key_machinery":"The machinery is the symmetric period $P_{\\mathfrak{q}}(\\mathfrak{l},\\Phi,\\Phi)=\\langle T_{\\mathfrak{l}}(\\Phi),\\Phi\\rangle$ with $\\Phi=\\varphi_1\\varphi_2^{\\mathfrak{q}}$, expanded in two ways: first by Plancherel over forms of level $\\mathfrak{u}\\mathfrak{v}\\mathfrak{q}$, where the Hecke operator produces the eigenvalue $\\lambda_\\pi(\\mathfrak{l})$ inside the moment; second by using the Hecke relation to expand $T_{\\mathfrak{l}}$ and then reversing the roles of $\\mathfrak{q}$ and $\\mathfrak{l}$ to obtain a spectral expansion over level-$\\mathfrak{l}$ representations. The resulting identity, equation (5.15)/(5.18), bounds $q^{1/2}\\,\\zeta_{\\mathfrak{q}}(1)/\\zeta_{\\mathfrak{q}}(2)\\,|G_{\\mathfrak{q}}(\\mathfrak{l},\\Phi,\\Phi)|$ by $(\\ell q)^\\varepsilon\\,(q^{1/2}/\\ell^{1/2-\\theta_1-\\theta_2}+\\ell^{1/2}q^{\\theta})$. The link between these periods and triple product L-functions is Proposition 3.1, the Ichino formula and its Eisenstein analogue, and the positivity of the local period factors $\\ell_v$ supplies the nonnegativity needed for amplification.","core_discovery":"The central claim is Theorem 1.2: for fixed $\\theta_i$-tempered cuspidal representations $\\pi_1,\\pi_2$ and a cuspidal $\\pi_3$ of conductor $\\mathfrak{q}$, with $(\\mathfrak{q},\\mathfrak{u}\\mathfrak{v})=1$ and an archimedean principal-series condition, one has $L(\\tfrac12,\\pi_1\\otimes\\pi_2\\otimes\\pi_3) \\ll q^{1 - (\\tfrac12-\\theta)(1-2\\theta_1-2\\theta_2)/(3-2\\theta_1-2\\theta_2)+\\varepsilon}$. The same bound is proved in Theorem 1.3 when $\\pi_3$ is replaced by a unitary Hecke character of conductor $\\mathfrak{q}$. The engine is a reciprocity formula for the twisted first moment: the moment over representations of conductor dividing $\\mathfrak{u}\\mathfrak{v}\\mathfrak{q}$, weighted by $\\lambda_\\pi(\\mathfrak{l})$, is bounded by a spectral expansion over representations of conductor dividing $\\mathfrak{l}$, up to acceptable errors. This swaps the spectral length of the family and makes the moment accessible to the amplifier, which then isolates the single value at $\\pi_3$ and converts the moment bound into a subconvex bound. The paper describes the resulting exponent as the limit of the amplification method in this setting.","pith_inferences":["One step beyond the paper, the same spectral reciprocity identity could be iterated or fed into higher moments, potentially yielding non-vanishing results or sharper averaged bounds for triple product L-functions in the level aspect.","The saving formula shows the method's boundary: as $\\theta_1+\\theta_2$ approaches $1/2$ the exponent saving tends to zero, so any improvement in the Ramanujan bound for $\\mathrm{GL}(2)$ over number fields would directly strengthen Theorem 1.2.","The archimedean principal-series hypothesis appears to be a removable barrier in the Eisenstein case (Theorem 1.3) because the Eisenstein component is itself principal series; a testable extension is to relax this hypothesis in the cuspidal case along the lines of Remark 5.2.","A numerical experiment on a fixed number field with small conductors could check whether the reciprocity identity (5.18) holds as an approximate equality rather than only as the bound used here, which would indicate whether the method is lossless."],"forward_implications":["An individual triple product value $L(\\tfrac12,\\pi_1\\otimes\\pi_2\\otimes\\pi_3)$ in the level aspect is subconvex over any number field, with exponent saving $225/2624$ unconditionally and $1/6$ under the Ramanujan-Petersson conjecture.","The same saving holds for $\\mathrm{GL}(1)$ twists $L(\\tfrac12,\\pi_1\\otimes\\pi_2\\otimes\\chi)$, improving known depth-aspect bounds for twists of $\\mathrm{GL}(2)\\times\\mathrm{GL}(2)$ Rankin-Selberg L-functions.","The reciprocity formula is a reusable identity: the first moment twisted by $\\lambda_\\pi(\\mathfrak{l})$ with conductor bound $\\mathfrak{u}\\mathfrak{v}\\mathfrak{q}$ is controlled by spectral data at level $\\mathfrak{l}$, so the spectral length can be traded between the two ideals.","The fixed-two-vary-one normalization gives a stronger saving than hybrid bounds that vary all three representations, and the paper identifies the final exponent as the limit of the amplification method."],"supporting_citations":[{"why":"Supplies the period-expansion method and the symmetric relation that the reciprocity formula generalizes.","marker":"[Zac20]"},{"why":"Provides the triple product integral representation, the canonical norm comparison, and Proposition 5.1's archimedean lower bound.","marker":"[MV10]"},{"why":"Gives the cuspidal case of the trilinear form formula that connects periods to central L-values.","marker":"[Ich08]"},{"why":"Supplies the nonnegativity and lower bound $\\prod_{v\\mid\\mathfrak{q}}\\ell_v \\asymp 1/q$ for local triple product period factors.","marker":"[Hu17]"},{"why":"Provides nonnegativity of local triple product factors and the regularization of the Eisenstein adjoint quotient used in the continuous contribution.","marker":"[BJN24]"},{"why":"Supplies the amplifier construction and expansion used to isolate the target L-value in Section 6.1.","marker":"[BK19]"},{"why":"Gives the asymptotic $\\Lambda^*(1,\\pi,\\mathrm{Ad})=C(\\pi)^{o(1)}$ needed to pass from the amplified moment bound to the individual L-value.","marker":"[HL94]"}],"fun_headline_variants":["Reciprocity breaks convexity for triple product L-functions","Spectral reciprocity yields subconvexity for triple L-functions","Triple L-functions: convexity broken via reciprocity","Subconvex bound for triple L-functions via reciprocity"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof depends on quoted results that the local triple product period factors are nonnegative and that, for a target representation of conductor $\\mathfrak{q}$, their product over $v\\mid\\mathfrak{q}$ is about $1/q$; the archimedean lower bound $f(\\pi_\\infty) \\ge c(\\pi_\\infty)^{-1-\\varepsilon}$ is a second quoted input. If any of these local factors could vanish or change sign, or if the archimedean bound failed, the amplification step would lose its lower bound.","fun_headline_variants_meta":{"raw":{"variants":["Reciprocity breaks convexity for triple product L-functions","Spectral reciprocity yields subconvexity for triple L-functions","Triple L-functions: convexity broken via reciprocity","Subconvex bound for triple L-functions via reciprocity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001567,"raw_usage":{"total_tokens":6329,"prompt_tokens":1090,"completion_tokens":5239,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":706,"completion_tokens_details":{"reasoning_tokens":5180}},"tokens_in":706,"tokens_out":5239,"duration_ms":39649,"temperature":1.0,"reasoning_tokens":5180,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T23:50:46.529830+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a finite place $v\\mid\\mathfrak{q}$, compute the normalized local triple product period factor $\\ell_v$ from (3.3) over an orthonormal basis of $K_{0}(\\mathfrak{p}_v^{m})$-invariant vectors for a representation of conductor $\\mathfrak{p}_v^{m}$, for instance a Steinberg or principal-series representation; a zero or negative value would falsify the lower bound $H(\\pi,\\mathfrak{q})\\gg 1/q$ on which the proof relies. A coarser check is to test Theorem 1.1 numerically for small $\\mathfrak{q}$ and $\\mathfrak{l}$: any violation of the stated moment bound would invalidate the amplification argument that produces Theorem 1.2.","supporting_citations":[],"review_version":1}