{"id":"ac99a413-2322-48b7-bdad-b947f983a925","arxiv_id":"2508.19058","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For any adjoint reductive group at rational Kac-Moody level, cuspidal twisted D-modules with nilpotent singular support have at least one nonzero quantum Whittaker coefficient, proved microlocally.","lead":"This paper proves that cuspidal automorphic sheaves in the quantum geometric Langlands program, for any adjoint reductive group at rational level, have at least one nonzero Whittaker coefficient. The proof gives a new microlocal interpretation of these coefficients, replacing the Hecke-action argument that fails in the quantum setting.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The load-bearing step is the microlocal bridge: Theorem 5.3.6 depends on a twisted Nadler–Taylor theorem (Thm 5.1.2) that is only sketched, and Prop. 5.3.2 verifies the intersection is the single point λD but not transversality/cleanness. If this fails, the characteristic-cycle argument collapses.","rationale":"The reader's weakest assumption is also the one I would stress: Theorem 5.3.6 is the engine of the non-vanishing theorem, and it rests on the twisted microstalk identification. The proof of Prop. 5.3.2 concludes transversality from an isolated-point statement; this does not logically imply clean intersection, and the import of [NT23] via Thm 5.1.2 is only sketched. The concrete check of tangent spaces at λD would settle this. I also note a separate gap: 'cuspidal' is not defined for sheaves, and the implication needed to pass from Theorem 6.1.1 to Theorem 1.2.2 is absent; however, this is more a definitional/expository gap than the deeper geometric assumption. With the microlocal concern unresolved, the verdict should remain CONDITIONAL rather than ACCEPT, but no demonstrated error warrants REJECT.","tokens_in":27683,"tokens_out":11425,"duration_ms":106899,"concrete_test":"Verify transversality at λD directly: take the closed embedding ιD of Prop. 4.3.3, restrict to λD, and compute the tangent map dιD. Check that T_{λD}Kos_D ∩ T_{λD}Nilp^λ_red = 0 and that the intersection is clean of the stated dimension (equivalently, that (5.15) in Prop. 5.3.3 is an isomorphism of tangent spaces). If this fails, re-derive Cor. 5.3.4; if it cannot be derived, Theorem 5.3.6 is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The non-vanishing proof for Whittaker-tempered sheaves rests on Theorem 5.3.6, which identifies coeff_D(F) with a twisted microstalk at λD and concludes χ(coeff_D(F)) = c_{Nilp^D,F}. This identification requires the conditions of Theorem 5.1.2: the Cartesian property of Lemma 5.3.1, and a clean/transverse intersection of the shifted conormal with Λ (Cor. 5.3.4). Theorem 5.1.2 is stated as a twisted version of [NT23] and proved only by sketch; the key step (5.5) already assumes the isomorphism (5.2) and a reduction to the trivial gerbe. More concretely, the proof of Prop. 5.3.2 shows that Kos_D ∩ Nilp^λ_red is the single point λD, but it does not compute tangent spaces or prove clean/transverse intersection. An isolated intersection point does not imply cleanness (e.g. y=x^2 and y=0 in R^2), so the conclusion 'intersects transversely' is not justified as written. If T_{λD}Kos_D ∩ T_{λD}Nilp^λ_red ≠ 0, then Cor. 5.3.4 and Theorem 5.3.6 do not follow, and the non-vanishing conclusion has no basis.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves that for a reductive group G of adjoint type, any cuspidal twisted D-module on Bun_G with nilpotent singular support has a non-zero quantum/geometric Whittaker coefficient. The main theorem, Theorem 1.2.2, is deduced from a stronger conservativity statement, Theorem 6.1.1, for the Whittaker-tempered subcategory of Shv_{κ,Nilp}(Bun_G); a further conservativity statement for all of D_κ(Bun_G)^{Wh-temp}, Theorem 6.3.1, is made conditional on Conjecture 2. The proof introduces a D-Kostant slice Kos_D inside T^*Bun_G, embeds its intersection with the regular nilpotent locus into a Zastava space, and identifies the Whittaker coefficient functor coeff_D with a twisted microstalk at a single smooth point λ_D. This yields a characteristic-cycle formula χ(coeff_D(F)) = c_{Nilp^D,F}, from which non-vanishing is extracted by choosing a minimal component of the singular support. The paper is written in the style of the contemporary geometric Langlands literature and relies on a substantial amount of imported technology, especially a twisted version of the Nadler–Taylor theorem.","tokens_in":27958,"tokens_out":15644,"duration_ms":143224,"significance":"If the result and its proof are correct, this is a significant step toward the quantum/metaplectic geometric Langlands program: it extends the Faergeman–Raskin non-vanishing theorem from the critical level to rational Kac–Moody levels, produces an unconditional conservative functor out of the Whittaker-tempered nilpotent category, and gives a new microlocal understanding of Whittaker coefficients. The geometric core of the paper is genuinely new and original: the D-Kostant slice, the closed embedding into the Zastava space, and the identification of the single intersection point λ_D are substantial contributions. The proof is also unconditional in the main nilpotent case, and the dependence on Conjecture 2 is clearly flagged. However, the central microlocal bridge is not yet documented at the level of detail demanded by the argument: the twisted Nadler–Taylor theorem is only sketched, and transversality of a key intersection is asserted rather than proved. These points are load-bearing for the main theorem, so the manuscript needs a major revision before its claims can be regarded as established.","major_comments":[{"comment":"Theorem 5.1.2 is load-bearing: Theorem 5.3.6 identifies coeff_D(F) with a twisted microstalk and derives the characteristic-cycle formula (5.16) from it. The theorem is imported from [NT23] 'up to some modifications,' but the proof given here is only a sketch. In particular, the claimed isomorphism (5.2), (eY)^{>0}_{ey0} ≅ Y^{>0}_{y0}, is justified by asserting that s^*L is a trivial Gm-equivariant line bundle on A^1; this is plausible under the Cartesian hypothesis but not demonstrated. The second statement is then reduced to the trivial-gerbe case of [NT23, Prop. 2.3.1] without showing that the twisting can be trivialized compatibly with the clean-intersection hypothesis. Since this is the central microlocal bridge, the manuscript needs either a complete proof of the twisted theorem or a precise statement of the [NT23] result being imported and a proof that all hypotheses are satisfied","section":"§5.1, Theorem 5.1.2"},{"comment":"Proposition 5.3.2 proves only that the intersection of the shifted conormal (Kos_D) with Λ is the single point λ_D and that λ_D is a smooth point of Λ. The word 'transversely' in the statement is not justified: no tangent-space computation shows T_{λ_D}Kos_D ∩ T_{λ_D}Λ = 0, and an isolated intersection point need not be transverse (e.g. y=x^2 and y=0 in R^2). Corollary 5.3.4 and Theorem 5.3.6 require the clean-intersection version of this assertion. The proof of Prop. 5.3.3, cited for Cor. 5.3.4, ends with equation (5.15), which asserts the needed equality of the conormal to Bun_N with an intersection of tangent spaces without a proof; the displayed exact sequences (5.12)–(5.14) do not imply the desired spanning/dimension count. Without clean transversality, the formula χ(coeff_D(F)) = c_{Nilp^D,F} need not follow.","section":"§5.3, Prop. 5.3.2 / Cor. 5.3.4"},{"comment":"Theorem 1.2.2 is stated for cuspidal F ∈ Shv_{κ,Nilp}(Bun_G), but the proof supplied in §6.1 (and outlined in §1.3.6) proves non-vanishing only for objects in the Whittaker-tempered subcategory. The paper does not state or prove that a cuspidal object of Shv_{κ,Nilp}(Bun_G) is Whittaker-tempered; Proposition 3.3.5 only shows that objects with irregular nilpotent singular support are Whittaker-anti-tempered. Unless this implication is a known theorem, with a precise reference, Theorems 1.2.2 and 1.2.8 as stated do not follow from Theorem 6.1.1.","section":"§1.2 / §6.1"}],"minor_comments":[{"comment":"Typo: 'Prposition' should be 'Proposition'.","section":"§2.1, Prop. 2.1.4"},{"comment":"The expression 'Shv_{La,Λ}(W)' appears to be a typo; it should likely be 'Shv_{G,Λ}(W)' or a defined shorthand.","section":"§5.3, proof of Thm. 5.3.6"},{"comment":"The phrase 'not less than ˇλ' is ambiguous and the definition of Nilp^{reg,<ˇλ}_red is confusingly stated. Please rewrite with explicit inequalities (≤, <, ≥) and verify that the complement in Notation 4.3.11 is the intended one.","section":"§4.3.9"},{"comment":"Several typos: 'Whitttaker' in the section heading, 'folliwing' in §6.3, 'manifactures' in §6.4.","section":"§6"},{"comment":"Theorem 1.2.8 is stated in the Introduction without noting that its proof in §6.3 is conditional on Conjecture 2. Please state the hypothesis in the Introduction as well.","section":"§1.2.8 / §6.3"},{"comment":"The notation W_κ(ˇρ,−) for the integral Weyl group is used without definition; please define it or cite a reference.","section":"§3.4"}],"recommendation":"major_revision","confidential_remarks":"Please ensure that a second referee with expertise in [NT23] checks §5 carefully. The main gap is concentrated there: the twisted Nadler–Taylor theorem is only sketched, and Prop. 5.3.2 does not prove transversality. The rest of the geometric construction, especially the D-Kostant slice and the Zastava embedding, appears promising. Also, the cuspidal-to-Whittaker-tempered implication used implicitly in Theorem 1.2.2 needs an explicit reference or proof."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper proves a genuinely new result: at rational levels, cuspidal twisted D-modules with nilpotent singular support have nonvanishing quantum Whittaker coefficients. The method is the right one—replacing the Hecke action (unavailable at rational levels) with a microlocal description of coefficient functors via twisted microstalks. I read the proof in detail and the main architecture holds up. The D-Kostant slice Kos_D, the closed embedding into Zastava space, and the matching of components in Lemma 4.3.6 are worked out concretely, and the claim that the intersection Kos_D ∩ Nilp^{reg,λ}_red is the single reduced point λ_D is convincing. Proposition 3.3.5 (nilpotent irregular -> anti-tempered) is also coherent.\n\nThree soft spots, in increasing order of concern. First, Theorem 5.1.2 is imported from [NT23] and the twisted generalization is only sketched; the proof on pp. 21-22 reduces to the trivial gerbe but the reduction is not fully written. Second, the proof of Proposition 5.3.2 shows the intersection is a single point but never computes tangent spaces; for the microstalk identification you need clean (or transverse) intersection, and an isolated point alone does not give that. The same issue carries into Corollary 5.3.4. This is probably fixable—the setup is symmetric and the expected transversality is plausible—but as written it is a gap in the load-bearing step. Third, the term \"cuspidal\" is used in Theorem 1.2.2 without a definition for sheaves, and the reduction to Whittaker-tempered objects is implicit. Also, Theorem 1.2.8 is stated in the intro without the caveat that the proof uses Conjecture 2; the surrounding text says results in that subsection assume it, but a reader skimming the theorem statement could be misled.\n\nThe paper deserves a serious referee. The new content is real, the geometric construction is novel, and the conditional conservativity is a clean payoff. The gaps are in presentation and in one verification of transversality, not in a demonstrated error. With the clean-intersection step filled and the terminology pinned down, I would expect this to be an important reference for quantum geometric Langlands.","headline":"Genuinely new nonvanishing theorem for quantum Whittaker coefficients at rational levels, built on a microlocal method that avoids the failed Hecke action; the architecture is sound but the proof has a real gap in the clean-intersection verification and a few presentation issues.","tokens_in":28553,"tokens_out":2544,"would_cite":true,"duration_ms":22656,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14D24","14F10","22E57"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that every cuspidal twisted D-module with nilpotent singular support has at least one nonzero quantum Whittaker coefficient, via a microlocal interpretation of the coefficient functors.","keywords":["quantum geometric Langlands","metaplectic sheaves","Whittaker coefficients","twisted D-modules","nilpotent singular support","microstalks","Kostant slice","Zastava spaces"],"falsifier":"For G = PGL_2 on P^1 and D a single point, write out all points of Kos_D ∩ Nilp^reg_red using the explicit description in Example 4.1.9; if the intersection is not exactly the reduced point λ_D, or if it is transverse there but the microstalk formula is wrong, Theorem 5.3.6 collapses. A complementary check: compute χ(coeff_D(δ_E)) for a skyscraper sheaf at a generic bundle and compare it with the characteristic-cycle multiplicity c_{Nilp^D,δ_E}.","tokens_in":27449,"feed_emoji":"📐","tokens_out":11990,"duration_ms":96218,"temperature":0.7,"pith_summary":"Quantum Whittaker coefficients are the metaplectic analogue of Fourier coefficients of automorphic forms, and the paper proves that no cuspidal (zero-constant-term, cusp-form-like) object is killed by all of them. For any adjoint reductive group and rational level, every cuspidal twisted D-module with nilpotent singular support has a nonzero Whittaker coefficient for some divisor D. The proof is microlocal: each coefficient functor is shown to compute a twisted microstalk at a distinguished point, so nonvanishing follows from a characteristic-cycle multiplicity. The same argument makes the assembled Whittaker coefficient functor conservative on the Whittaker-tempered nilpotent category, and on all tempered D-modules conditional on a conjecture about the metaplectic spectral action.","feed_headline":"Every cuspidal sheaf has a nonzero Whittaker coefficient","feed_subtitle":"A microlocal proof in metaplectic geometric Langlands: no cuspidal object dies under all Whittaker coefficient functors.","key_machinery":"The D-Kostant slice Kos_D is the Lagrangian in T*Bun_G formed from the graph of the Whittaker function ψ_D on the twisted unipotent-bundle space Bun_N^{ω(-D)}. Via the twisted version of the imported microstalk theorem, the paper shows that when Kos_D meets the relevant regular-nilpotent component transversely at the single smooth point λ_D, the functor coeff_D is the twisted microstalk at λ_D; then the Euler characteristic of coeff_D(F) equals the coefficient of that component in the characteristic cycle of F. The intersection is identified with a Zastava space, which supplies the transversality and the point-counting needed for the proof.","core_discovery":"The central claim is that for an adjoint reductive group G and a rational level κ, every cuspidal object F in the category of twisted sheaves on Bun_G with nilpotent singular support has coeff_D(F) ≠ 0 for some Λˇ^+-valued divisor D. More generally, the coefficient functors assembled together are conservative on the Whittaker-tempered nilpotent category, and—assuming a conjectural factorization of the metaplectic Hecke action—on all Whittaker-tempered D-modules. The argument identifies the Whittaker coefficient functor for divisor D with the twisted microstalk at the point λ_D where the D-Kostant slice meets the regular-nilpotent component, giving χ(coeff_D(F)) = c_{Nilp^D,F}. This is the qu","pith_inferences":["A natural next step is to compute c_{Nilp^D,F} on Zastava spaces for explicit F; that would turn the nonvanishing theorem into an effective recipe for which divisors D work.","Because the argument bypasses the Hecke action, it may apply at irrational levels, where the metaplectic dual group is trivial and Hecke-eigensheaf information degenerates, leaving Whittaker coefficients as the only nontrivial invariants.","Conservativity of coeff_loc suggests that the Whittaker-anti-tempered part is exactly the joint kernel of all coefficient functors; the conditional theorem would make that kernel description unconditional.","For small groups such as PGL_2 on P^1, the explicit description of the D-Kostant slice allows direct coordinate checks of the transversality claim and of the microstalk formula for chosen divisors."],"forward_implications":["Every cuspidal twisted sheaf with nilpotent singular support has at least one nonzero quantum Whittaker coefficient.","The assembled functor coeff_loc is conservative on the Whittaker-tempered nilpotent category, so objects there are detected by their Whittaker invariants.","Assuming the metaplectic spectral action conjecture, conservativity extends to all Whittaker-tempered D-modules, making Whittaker coefficients a substitute for Hecke-eigensheaf decomposition.","For most rational levels the Whittaker-temperedness condition is automatic, so the nonvanishing result covers the whole nilpotent-singular-support category.","The adjoint assumption is expected to be dropped in a forthcoming companion result for general reductive groups."],"supporting_citations":[{"why":"Supplies the unramified nonvanishing theorem and the Whittaker-temperedness machinery that this paper adapts to the metaplectic setting.","marker":"[FR22]"},{"why":"Provides the theorem that a Whittaker functional is a shifted microstalk; the paper needs its twisted version and modifies it in Section 5.","marker":"[NT23]"},{"why":"Gives the chiral principal series and divisor-parameterization input used to construct the D-Kostant slice and connect it to Zastava spaces.","marker":"[Ras21a]"},{"why":"Supplies the local Whittaker averaging functor and its limit-preservation property used in the recollement of Whittaker-tempered categories.","marker":"[Ras21b]"},{"why":"Provides the coefficient-functor normalization and the Kac-Moody localization context used for coeff_loc and the comparison with coeff_D.","marker":"[ABC+24a]"},{"why":"Defines the metaplectic Langlands dual datum and the conjectural spectral action (Conjecture 2) that the conditional conservativity theorem depends on.","marker":"[GL18]"},{"why":"Supplies the nilpotent-singular-support formalism and the spectral-projector argument used in the proof of the conditional theorem.","marker":"[AGK+22]"},{"why":"Provides the microlocal foundations: microstalks, characteristic cycles, and the identities used to equate χ(coeff_D(F)) with a characteristic-cycle multiplicity.","marker":"[KS94]"},{"why":"Contributes the Zastava spaces and their connected components, which identify the intersection of the D-Kostant slice with regular-nilpotent loci.","marker":"[BFGM02]"},{"why":"Constructs the Whittaker function ψ_D on twisted unipotent bundles, which is used to define the coefficient functors.","marker":"[FGV01]"}],"fun_headline_variants":["Cuspidal sheaves always have a nonzero Whittaker coefficient","Whittaker coefficients detect all cuspidal twisted D-modules","Microlocal proof shows cuspidal Whittaker coefficients don't vanish"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The load-bearing premise is the microlocal bridge: for each divisor D, the D-Kostant slice meets the relevant regular-nilpotent component transversely at a single smooth point, so the Whittaker coefficient is a twisted microstalk whose Euler characteristic is a characteristic-cycle multiplicity; if the transversality or the imported twisted microstalk theorem fails, nonvanishing does not follow.","fun_headline_variants_meta":{"raw":{"variants":["Cuspidal sheaves always have a nonzero Whittaker coefficient","Whittaker coefficients detect all cuspidal twisted D-modules","Microlocal proof shows cuspidal Whittaker coefficients don't vanish"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00033,"raw_usage":{"total_tokens":1602,"prompt_tokens":599,"completion_tokens":1003,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":343,"completion_tokens_details":{"reasoning_tokens":941}},"tokens_in":343,"tokens_out":1003,"duration_ms":8910,"temperature":1.0,"reasoning_tokens":941,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T16:00:40.209195+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For G = PGL_2 on P^1 and D a single point, write out all points of Kos_D ∩ Nilp^reg_red using the explicit description in Example 4.1.9; if the intersection is not exactly the reduced point λ_D, or if it is transverse there but the microstalk formula is wrong, Theorem 5.3.6 collapses. A complementary check: compute χ(coeff_D(δ_E)) for a skyscraper sheaf at a generic bundle and compare it with the characteristic-cycle multiplicity c_{Nilp^D,δ_E}.","supporting_citations":[],"review_version":1}