{"id":"68e166c7-9480-41d0-8f87-34ded0c2ad9f","arxiv_id":"2502.07300","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Every Arthur packet of U(p,q) contains at most one unitary lowest weight representation, with explicit conditions for existence and a formula for the lowest K-type.","lead":"This paper classifies which Arthur packets of the real unitary group U(p,q) contain an irreducible unitary lowest weight representation, and gives an explicit formula for the lowest K-type when one exists. It shows each Arthur packet has at most one such representation, extending earlier scalar-case results to the full non-scalar setting.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 5.1's reduction to d0 is incompletely proved: the stated two-mixed-block condition does not cover non-holomorphic data with a single mixed block, so the uniqueness claim and Theorems 4.7/4.8 rest on an unproven case.","rationale":"The reader's weakest assumption is exactly Lemma 5.1, and my reading confirms that this is the most load-bearing insecure step. The central claim of the paper—the classification of Arthur packets containing a unitary lowest weight representation—depends on the reduction of any such packet to the holomorphic datum d0. That reduction is Lemma 4.6, whose proof is deferred entirely to Lemma 5.1. The proof of Lemma 5.1 is a two-sentence combinatorial argument: if two distinct blocks are both mixed, the signed tableau has a forbidden row. But non-holomorphic data need not have two mixed blocks: a single mixed block followed by a plus-only block already violates holomorphicity, and the stated hypothesis p_k q_k p_ℓ q_ℓ ≠ 0 does not apply. I worked through the minimal example U(2,1), ψ = χ_3⊗S_2 ⊕ χ_0⊗S_1, d = {(1,1),(1,0)}. The datum is non-holomorphic, the resulting signed tableau has shape (2,1) with rows [+,−] and [+], which is compatible with Corollary 3.3, and the ν-antitableau is row1 [2,1], row2 [0]. This shows the proof of Lemma 5.1 omits a real case; the lemma may still be true, but the argument as written does not establish it. There is a second gap, Lemma 3.2(3), whose proof assumes the isomorphism it is meant to prove and whose statement overclaims nonzero low-weight property for mediocre-range inductions; this lemma is used in Lemma 5.2(8) and Lemma 5.5. I focus on Lemma 5.1 because a failure there directly threatens the main theorem's iff characterization and uniqueness, not just a subcase of the proof. Since the reader already identified this same point and assigned CONDITIONAL, my stress-test does not change the verdict; it does sharpen the requested repair: the omitted single-mixed-block case must be analyzed explicitly, and Lemma 3.2(3) should be either proved correctly or reformulated to state only the conditional inequality that the applications actually use.","tokens_in":27597,"tokens_out":28026,"duration_ms":236404,"concrete_test":"Compute Trapa's invariants for the explicit U(2,1) example: ψ = χ_3⊗S_2 ⊕ χ_0⊗S_1 and d = {(1,1),(1,0)}, with λ_d = (1,1,1) and infinitesimal character {2,1,0}. Use the algorithm in §3.3 to build the signed tableau and the ν-antitableau for A_d(ψ), and determine whether A_d(ψ) has a vector annihilated by b− (equiv., whether it is isomorphic to the unique unitary lowest weight module π_{(2,2;-1)} in this packet). If it is not lowest weight, identify which tableau property rules it out and add that property to Lemma 5.1; if it is lowest weight, Lemma 5.1 and the reduction to d0 are false.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Lemma 5.1 is the sole justification for Lemma 4.6, which reduces the packet to the holomorphic datum d0; Theorem 4.8 then gives necessary and sufficient conditions based only on d0. The proof of Lemma 5.1 argues that if two distinct blocks k,ℓ satisfy p_k q_k p_ℓ q_ℓ ≠ 0, the signed tableau of Ad(ψ) has a row with three boxes or a −,+ row, contradicting Corollary 3.3. This argument does not cover non-holomorphic data with exactly one mixed block. For example, in U(2,1) with ψ = χ_3⊗S_2 ⊕ χ_0⊗S_1 and d = {(1,1),(1,0)}, the datum is non-holomorphic (q_1>0 and p_2>0), yet p_2 q_2 = 0, so the stated hypothesis is false. The signed tableau construction of §3.3 for this d yields a shape (2,1) whose two-box row has signs [+,−] and whose second row is [+], which is allowed by Corollary 3.3. Hence the proof does not exclude this d, and the reduction to d0 is not established for a class of data that can arise in D(ψ). If some such Ad(ψ) were nonzero and lowest weight, the uniqueness assertion and the classification in Theorem 4.8 would need revision. The lemma may be true, but the combinatorial case analysis is missing and the stated sufficient condition is not equivalent to non-holomorphicity.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies Arthur packets of the real unitary group G = U(p, q).  For a good A-parameter ψ, the packet Π(ψ) is described via Mœglin-Renard as the set of cohomologically induced representations A_d(ψ) indexed by data d = {(p_i, q_i)} with p_i + q_i = a_i.  The author determines, in Theorems 4.7 and 4.8, necessary and sufficient conditions for an irreducible unitary lowest weight representation π_λ to lie in Π(ψ), and more generally for Π(ψ) to contain a nonzero unitary lowest weight representation.  The main structural claim is Lemma 4.6 (proved as Lemma 5.1): if A_d(ψ) is nonzero and lowest weight, then the parabolic is automatically holomorphic, i.e., d = d_0, so that in particular a packet contains at most one such representation.  The proof uses Barbasch-Vogan invariants (annihilator and asymptotic support), Trapa's algorithm for cohomological induction, and nonvanishing criteria due to Huang and Du.","tokens_in":27941,"tokens_out":14898,"duration_ms":122173,"significance":"If the results are correct, they give a complete and explicit answer to a natural question with direct applications to automorphic forms and Shimura varieties; the author mentions the application [HMY25].  The main theorem is concrete and falsifiable: it produces explicit lowest K-type formulas for the unique unitary lowest weight representation in a packet, and it gives clean necessary and sufficient conditions in terms of the segments ν_i.  The paper makes good use of external theorems (Mœglin-Renard, Trapa, Huang, Du, EHW) and does not appear to presuppose the classification it proves.  However, two gaps in the proof — the incomplete reduction to holomorphic parabolics in Lemma 5.1 and the circular proof of Lemma 3.2(3) — currently prevent the results from being fully established.","major_comments":[{"comment":"The proof of Lemma 5.1 does not cover the case of a non-holomorphic datum with exactly one mixed block.  The argument only rules out two distinct blocks k, ℓ with p_k q_k p_ℓ q_ℓ ≠ 0.  For a datum with a single mixed block, such as d = {(1,1),(1,0)} for U(2,1) with ψ = χ_3⊗S_2 ⊕ χ_0⊗S_1, the condition p_k q_k p_ℓ q_ℓ ≠ 0 is false, so the proof gives no reason to exclude this d.  The signed tableau constructed in §3.3 for this example does not visibly violate Corollary 3.3, and the author does not provide a separate argument showing that A_d(ψ) cannot be a nonzero lowest weight representation.  Since Lemma 4.6 and the classification in Theorems 4.7 and 4.8 rely on the reduction to d = d_0, the main theorem is not fully proved without an additional argument for the single-mixed-block case.","section":"§5.1, Lemma 5.1"},{"comment":"Lemma 3.2(3) is stated as an existence and nonvanishing statement: for the specified θ-stable parabolic q and weight µ satisfying the displayed conditions, A_q(µ) is supposed to be a nonzero lowest weight representation with lowest K-type λ satisfying the inequality.  The proof, however, begins with \"Suppose A_q(µ) ≅ π_λ\", which assumes the isomorphism that the lemma is meant to establish.  Consequently the nonvanishing of A_q(µ) and the existence of its lowest K-type are not proved.  This lemma is cited in the proof of Lemma 5.2(8) and again in Theorem 4.8(4), so the gap affects the derivation of the main results.","section":"§3.4, Lemma 3.2(3)"}],"minor_comments":[{"comment":"In the statement of Lemma 3.2(3), the integers p′ and q′ appear in the definition of µ before they are defined.  The author should clarify whether p′ and q′ are the multiplicities attached to µ or to the resulting lowest K-type λ.","section":"§3.4, Lemma 3.2(3)"},{"comment":"The construction of the initial signed tableau S_1 is described as \"the Young tableau of size 1 + · · · + 1\" with p_1 + q_1 boxes; this wording is ambiguous.  Please state explicitly whether S_1 is a single row, a single column, or something else, since subsequent tableau comparisons depend on the convention.","section":"§3.3"},{"comment":"In the \"converse\" parts of Theorem 4.7, the paper invokes Lemma 3.2, Corollary 3.6, and Corollary 3.7 to assert A_{d_0}(ψ) ≅ π_λ, but it does not explicitly verify the mediocre-range hypotheses of those corollaries in each of the four cases.  Adding these routine checks would make the argument easier to follow.","section":"§5.2–§5.5"},{"comment":"There are several typographical inconsistencies (e.g., \"Mœglin\" vs. \"Mogelin\", and the spelling \"Théor`eme\"), which are harmless but should be corrected in revision.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper treats a natural and consequential question, and the main formulas are explicit and plausible.  The two gaps noted above are significant but appear reparable within the manuscript's scope: the single-mixed-block case of Lemma 5.1 requires a dedicated argument (for instance, using the K-type formula or Adams's theorem on holomorphic parabolics), and Lemma 3.2(3) needs a genuine nonvanishing proof.  I would recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new thing here is a complete-looking classification of when an Arthur packet of U(p,q) contains a unitary lowest weight representation, including non-scalar cases, plus explicit formulas for the unique lowest K-type. That goes beyond Mœglin–Renard's scalar Sp(2n,R) case, and Huang's and Du's nonvanishing work, which never addressed packet membership. If correct, this is a real advance and is already being used in a Shimura variety application.\n\nWhat the paper does well: the Barbasch–Vogan/Trapa strategy is the right tool, most lemmas are proved in detail, and the external dependencies are clearly cited. I see no circular dependence on the target classification and no invented entities.\n\nThe soft spots are real. Lemma 3.2(3) asserts nonvanishing and existence, but the proof starts by assuming Aq(mu) is already isomorphic to pi_lambda; that does not prove the existence part. More seriously, Lemma 5.1 is the sole support for the reduction to d0 in Lemma 4.6, and its proof only handles data with two mixed blocks. The single-mixed-block case is simply not addressed. The stress-test example in U(2,1) with d = {(1,1),(1,0)} is legal data, has one mixed block, and its signed tableau is permitted by Corollary 3.3. That is not a counterexample to the theorem, but it is a counterexample to the proof as written: the stated hypothesis does not imply the conclusion, so the reduction to d0 that Theorem 4.8 rests on is not established for that class of data. This is load-bearing, not cosmetic. The tableau comparisons in Lemmas 5.3–5.5 are also sketched; they look credible but need expansion under referee pressure.\n\nWho this is for: representation theorists and automorphic forms people working with Arthur packets of unitary groups and Shimura varieties. The paper deserves a serious referee, not a desk reject. I would send it to review and ask the author to prove Lemma 5.1 in full, including the single-mixed-block case, and to fix the proof of Lemma 3.2(3). The classification is probably right, but the proof as written does not yet show it.","headline":"A genuinely new classification of Arthur packets of U(p,q) containing unitary lowest weight representations, with a load-bearing proof gap in the reduction to holomorphic data.","tokens_in":28479,"tokens_out":4957,"would_cite":true,"duration_ms":45244,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["22E46","22E47","11F70"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper determines, for every good A-parameter of U(p,q), exactly when its Arthur packet contains a nonzero irreducible unitary lowest weight representation, proves that at most one such representation can occur, and gives explicit…","keywords":["Arthur packets","A-parameters","unitary lowest weight representations","cohomological induction","lowest K-type","nonvanishing of A_q(lambda)","segments and overlap inequalities","U(p,q)"],"falsifier":"Enumerate two-block sign partitions with $p_1q_1p_2q_2\\ne 0$ for a small example such as $\\mathrm{U}(3,2)$ and apply the tableau reduction algorithm to every resulting $A_d(\\psi)$: if any reduced signed tableau has no row of length at least three and no row with signs $-,+$, then the reduction to $d_0$ fails, and the classification would need revision.","tokens_in":2272,"feed_emoji":"📐","tokens_out":3061,"duration_ms":104862,"temperature":0.7,"pith_summary":"Arthur's multiplicity formula describes automorphic representations place-by-place through Arthur packets, but deciding which packet contains a given archimedean representation is often hard. This paper solves that problem for the unitary lowest weight representations of the real group U(p,q), including non-scalar cases: it determines, for every good A-parameter $\\psi$, whether the packet $\\Pi(\\psi)$ contains a nonzero irreducible unitary lowest weight representation, and proves that it can contain at most one. The answer is a two-part test on the parameter's segments: the segments before and after the 'middle' segment $\\nu_j$ must be multiplicity-free, and the middle segment's overlap with the right-hand segments must not exceed $p_j$ while its overlap with the left-hand segments must not exceed $q_j$. When the test passes, the paper gives explicit formulas for the lowest $K$-type of the unique representation, split into four cases. This matters because it turns the counting of holomorphic lowest weight automorphic forms into a checkable combinatorial condition on the local parameter.","feed_headline":"Arthur packets hold at most one unitary lowest weight rep","feed_subtitle":"For U(p,q), two overlap inequalities decide existence; the unique lowest K-type is then explicit.","key_machinery":"The load-bearing objects are the segments $\\nu_i$ attached to the summands of $\\psi$, which together form the Harish-Chandra parameter $\\nu$. The reduction that makes the problem tractable is the assertion that a nonzero member $A_d(\\psi)$ of the packet can be a unitary lowest weight representation only when the sign partition $d$ is the holomorphic one $d_0$, in which the plus boxes occur before the minus boxes; otherwise the signed tableau of the annihilator would contain a row with three boxes or a row with signs $-,+$, which cannot happen for a lowest weight module. With $d=d_0$, the nonvanishing criterion for the cohomologically induced module $A_{d_0}(\\psi)$ is computed by an algorithm that passes through adjacent segments and corrects the associated pair of invariants, an antitableau encoding the annihilator and a signed tableau encoding the asymptotic support; the resulting criterion is exactly the two overlap inequalities and the multiplicity-free condition. The proof of the main theorem, in the only-if direction, compares these computed tableaux with the tableaux of $\\pi_\\lambda$, while the if direction uses the known realization of $\\pi_\\lambda$ as a cohomological induction from a holomorphic parabolic subalgebra.","core_discovery":"The paper's central result is a complete criterion for a good A-parameter $\\psi = \\bigoplus_{i=1}^r \\chi_{t_i}\\otimes S_{a_i}$ of $\\mathrm{U}(p,q)$ to have a nonzero unitary lowest weight representation in its Arthur packet $\\Pi(\\psi)$. With $j$ the first index for which $a_1+\\cdots+a_j\\ge p$, write $\\nu_i=[(t_i-a_i+1)/2,(t_i+a_i-1)/2]$, $\\nu_{<j}=\\bigsqcup_{k<j}\\nu_k$, $\\nu_{>j}=\\bigsqcup_{k>j}\\nu_k$, $p_j=p-\\sum_{k<j}a_k$, and $q_j=q-\\sum_{k>j}a_k$. The packet contains a nonzero unitary lowest weight representation if and only if $\\nu_{<j}$ and $\\nu_{>j}$ are multiplicity free, $\\#(\\nu_j\\cap\\nu_{>j})\\le p_j$, and $\\#(\\nu_j\\cap\\nu_{<j})\\le q_j$; when this holds, the representation is unique. Its lowest $K$-type $\\lambda$ is then given explicitly: in the four subcases $q_j=0$, $p_j=\\#(\\nu_j\\cap\\nu_{>j})\\ne 0$, $q_j=\\#(\\nu_j\\cap\\nu_{<j})\\ne 0$, and the residual case, Theorem 4.8 describes $P(\\lambda)$ and $Q(\\lambda)$, or the coordinates $\\lambda_i$ through an auxiliary index $i_0$. Theorem 4.7 is the equivalent criterion from the opposite direction: for a fixed lowest weight module $\\pi_\\lambda$, it states, according to four ranges of $\\lambda_p-\\lambda_{p+1}$, exactly which A-parameters $\\psi$ have $\\pi_\\lambda\\in\\Pi(\\psi)$. The paper also proves packet-level uniqueness directly: because a lowest weight member forces the sign partition to be the holomorphic one $d_0$, no other member of the same packet is a unitary lowest weight representation.","pith_inferences":["The criterion can be read as a packing condition: the left and right tails of the parameter must be distinct (multiplicity free) and the middle segment may borrow from each side only up to the number of plus and minus boxes available; this suggests a purely combinatorial reformulation of A-packet membership for holomorphic representations.","Because the proof relies only on the injective parametrization of irreducible representations by annihilator and asymptotic support, the 'at most one lowest weight member' conclusion likely persists for other real reductive groups where that parametrization and signed-tableau models of nilpotent orbits are available.","One natural testable extension is to carry out the same overlap-inequality test for other Hermitian symmetric real groups, such as $\\mathrm{SO}^*(2n)$ or $\\mathrm{Sp}(2n,\\mathbb{R})$, whose unitary lowest weight modules are known but whose non-scalar packet membership has not been classified uniformly.","The explicit formulas in Theorem 4.8 could be verified computationally for small $p,q$: enumerate good A-parameters satisfying the two inequalities, run the cohomological induction tableau algorithm, and compare the output with the stated lowest $K$-type in each of the four cases."],"forward_implications":["No Arthur packet of $\\mathrm{U}(p,q)$ can contain two distinct irreducible unitary lowest weight representations; the holomorphic member, when present, is unique.","The nonvanishing of the holomorphic cohomological induction $A_{d_0}(\\psi)$ is settled by two inequalities together with the multiplicity-free conditions, giving a closed-form criterion in the range where the induction is weakly fair.","For any fixed unitary lowest weight module $\\pi_\\lambda$, Theorem 4.7 lists all A-parameters whose packet contains $\\pi_\\lambda$, organized by the four ranges of $\\lambda_p-\\lambda_{p+1}$.","The explicit lowest $K$-type formula in Theorem 4.8 makes the unique holomorphic member of a packet algorithmically constructible from $\\psi$ without computing full $K$-type multiplicities.","Combined with Arthur's multiplicity formula at the archimedean place, the criterion gives concrete counting conditions for square-integrable holomorphic automorphic forms of $\\mathrm{U}(p,q)$."],"supporting_citations":[{"why":"Supplies the description of the Arthur packet as the set of cohomologically induced modules $A_d(\\psi)$ indexed by sign partitions $d$, on which the whole paper is built.","marker":"[MR19]"},{"why":"Provides the algorithm that computes the annihilator and asymptotic-support tableaux for $A_q(\\lambda)$; the main proof uses these tableaux to compare $A_d(\\psi)$ with $\\pi_\\lambda$.","marker":"[Tra01]"},{"why":"Establishes the injective parametrization of irreducible representations by annihilator and asymptotic support, which is the identification tool used throughout.","marker":"[BV83]"},{"why":"Gives the unitarity criterion $\\lambda_p-\\lambda_{p+1}\\ge N-p'-q'$ that characterizes which lowest weight modules are unitary.","marker":"[EHW83]"},{"why":"Shows that cohomological induction from a holomorphic $\\theta$-stable parabolic subalgebra is a lowest weight module, used to identify the candidate representations.","marker":"[Ada87]"},{"why":"Supplies recent nonvanishing results and closed-form descriptions for $A_q(\\lambda)$ used in the overlap and segmentation arguments.","marker":"[Hua24]"},{"why":"Provides an independent treatment of the nonvanishing of $A_q(\\lambda)$ in the mediocre range and explicit overlap formulas used in the tableau computations.","marker":"[Du24]"},{"why":"Provides the general theory of cohomological induction, including the $K$-type formula in the mediocre range that underlies the converse direction.","marker":"[KV95]"}],"fun_headline_variants":["U(p,q) Arthur packets: at most one unitary lowest weight rep","Explicit K-type for the unique unitary lowest weight rep in U(p,q)","Criteria for Arthur packets of U(p,q) to contain unitary lowest weight reps","Unitary lowest weight reps in Arthur packets: uniqueness and K-types","Complete answer: which U(p,q) Arthur packets contain unitary lowest weight reps"],"cache_read_input_tokens":30592,"weakest_assumption_plain":"The reduction to the holomorphic sign partition $d_0$ rests on the combinatorial assertion that any nonzero cohomologically induced module whose sign partition mixes plus and minus boxes in two different blocks has a signed tableau with a row of at least three boxes or a row with signs $-,+$; if that assertion fails, the classification in Theorems 4.7 and 4.8 would need revision.","fun_headline_variants_meta":{"raw":{"variants":["U(p,q) Arthur packets: at most one unitary lowest weight rep","Explicit K-type for the unique unitary lowest weight rep in U(p,q)","Criteria for Arthur packets of U(p,q) to contain unitary lowest weight reps","Unitary lowest weight reps in Arthur packets: uniqueness and K-types","Complete answer: which U(p,q) Arthur packets contain unitary lowest weight reps"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00035,"raw_usage":{"total_tokens":1982,"prompt_tokens":1091,"completion_tokens":891,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":707,"completion_tokens_details":{"reasoning_tokens":793}},"tokens_in":707,"tokens_out":891,"duration_ms":7508,"temperature":1.0,"reasoning_tokens":793,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T13:11:03.006964+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Enumerate two-block sign partitions with $p_1q_1p_2q_2\\ne 0$ for a small example such as $\\mathrm{U}(3,2)$ and apply the tableau reduction algorithm to every resulting $A_d(\\psi)$: if any reduced signed tableau has no row of length at least three and no row with signs $-,+$, then the reduction to $d_0$ fails, and the classification would need revision.","supporting_citations":[],"review_version":1}