{"id":"d91c3713-2e3c-43ee-93b2-17395c9ba430","arxiv_id":"2505.05368","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The authors introduce a Lorentz-invariant hard/soft decomposition of supermomenta and use it to give explicit realizations, branching rules, and a memory-carrying example for generic unitary irreducible representations of the 4D BMS group.","lead":"This paper gives explicit wavefunctions and transformation rules for the particle states (unitary irreducible representations) of the BMS group, the symmetry group of gravity at large distances, and introduces a new way to split each state's 'supermomentum' into a hard part that behaves like ordinary momentum and a soft part tied to gravitational infrared effects.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 3.1 as stated fails for distributional soft charges: the soft part ∂_z^2δ^(2) is not the Paneitz image of a smooth shift N, so the claimed equivalence to pairs (p,N) with N ∈ Er1s/R^3,1 needs repair.","rationale":"The reader’s weakest assumption concerned the completeness of McCarthy’s classification in the nuclear topology. That is a legitimate structural caveat, openly acknowledged by the authors, and it does not challenge the internal validity of Proposition 3.1. My concern is different and more basic: the central hard/soft decomposition, Proposition 3.1, is stated for pairs (p,N) with N a smooth shift of vacua, but the proof and subsequent examples require N to be distributional for the derivative-of-delta soft charges that appear in the massless orbits. This is an internal inconsistency, not a disagreement with external consensus. The paper’s Appendix A.2 actually proves a distributional version with N in (Ann(R^3,1))′, so the gap is visible within the paper itself. Because the decomposition underpins the explicit wavefunctions, the branching rules, and the memory example, the claim as stated is not fully supported. The most likely repair is to replace Er1s/R^3,1 by a suitable distributional completion and then re-derive the orbit parametrization and inner products accordingly. That is a substantive correction but not a rejection of the overall programme, so the appropriate verdict is conditional acceptance rather than rejection or unconditional acceptance.","tokens_in":68339,"tokens_out":11638,"duration_ms":131084,"concrete_test":"Take the reference supermomentum K = δ^(2)(z,zbar) + σ(∂_z^2 + ∂_zbar^2)δ^(2)(z,zbar) from Eq. (6.20). Compute the spherical-harmonic coefficients of the soft part Σ = σ(∂_z^2 + ∂_zbar^2)δ^(2). If Proposition 3.1 is correct with N ∈ Er1s/R^3,1, then Σ must equal ∂_z^2∂_zbar^2 N for a smooth N, so the coefficients of Σ divided by ℓ(ℓ^2−1)(ℓ+2) must decay faster than any power of ℓ. An explicit evaluation shows the coefficients of ∂_z^2δ^(2) grow like ℓ^{5/2}, giving N-coefficients ∼ ℓ^{−3/2}, which are not rapidly decreasing. Thus no smooth N exists. Alternatively, attempt to solve ∂_zbar^2 N = δ^(2) in local coordinates and observe that the solution has a logarithmic singularity, violating the smoothness required by Definition 2.6.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Proposition 3.1 asserts that every supermomentum decomposes uniquely as P = P_hard + B_z^2 B_zbar^2 N with N ∈ Er1s/R^3,1, i.e. N a smooth conformal density modulo translations. But the massless hard piece is ωδ^(2), so the difference P − P_hard is generically distributional. A concrete counterexample is P = δ^(2)(z,zbar) + ∂_z^2δ^(2)(z,zbar). Its soft part Σ = ∂_z^2δ^(2) annihilates all translations, so Σ is a soft supermomentum, yet there is no smooth N with Σ = ∂_z^2∂_zbar^2 N. In spherical harmonics, the Paneitz operator on smooth densities has eigenvalue ℓ(ℓ^2−1)(ℓ+2) and image with no ℓ=0,1 modes; ∂_z^2δ^(2) has zero ℓ=0,1 moments but its higher coefficients grow like ℓ^{5/2}, forcing the prospective N to have coefficients decaying only as ℓ^{−3/2}. Such an N is not smooth and does not lie in Er1s/R^3,1; it is a distribution. This is not a minor technicality: the paper’s own massless examples (6.20) and (6.21), and the wavefunctions and branching rules built from them, use exactly such derivative-of-delta soft charges. Appendix A.2 proves only a dual statement with N in (Ann(R^3,1))′, a strict enlargement of Er1s/R^3,1. Thus the central decomposition as stated is internally inconsistent for distributional supermomenta, and the paper’s proof does not close the gap because it invokes Proposition 2.14, whose smooth and Hilbert-topology versions do not cover this case. The construction is likely repairable by allowing N to be a suitable distribution modulo translations, but that repair changes the target space in Proposition 3.1 and must be propagated through the orbit and wavefunction formulas.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper revisits McCarthy's classification of unitary irreducible representations (UIRs) of the four-dimensional BMS group and aims to make the representations as explicit as possible. The main new ingredient is a Lorentz-invariant decomposition of supermomenta into a hard part, which is a nonlinear function of an ordinary four-momentum, and a soft part written as the Paneitz operator acting on a shift of vacuum. Using this decomposition, the authors give explicit wavefunctions, inner products, group actions, and branching rules for massless and massive BMS representations, including cases with non-trivial little groups. They argue that generic BMS particles branch to Poincaré subgroups into infinite towers of usual particles plus continuous-spin representations, that the wavefunctions reproduce the dependence on gravity vacua previously proposed in related work, and that a particular BMS state can carry an average supermomentum with a memory-like soft component. The paper is largely self-contained, with distributional identities and proofs collected in appendices, and it explicitly notes the open status of the full classification in the nuclear topology.","tokens_in":68758,"tokens_out":7554,"duration_ms":84936,"significance":"If the central decomposition and the explicit realizations are correct, the paper provides a valuable toolkit for BMS representation theory: it makes the hard/soft split of supermomenta concrete, supplies explicit formulas for wavefunctions and inner products in the massless and massive cases, and derives branching rules that are new or only implicit in McCarthy's work. The detailed appendices with Paneitz identities and distributional checks are a genuine strength, as is the careful treatment of boundary terms in the distributional identities for hard supermomenta. The connection between generic BMS particles, continuous-spin representations, gravity vacua, and the memory effect is physically suggestive and gives concrete calculable examples. However, the central decomposition as stated does not cover the distributional supermomenta that the paper itself uses in the examples, so the main technical claim needs repair before the rest of the construction can be regarded as established.","major_comments":[{"comment":"","section":"Section 3.1, Proposition 3.1; Section 2.2.2; Appendix A.2"},{"comment":"","section":"Section 4.2, Theorem 4.3 and footnote 16"}],"minor_comments":[{"comment":"","section":"Section 3.1, equations (3.1)-(3.4)"},{"comment":"","section":"Proposition 2.14"},{"comment":"","section":"Section 6.4.2, Proposition 6.11"},{"comment":"","section":"Footnote 16"}],"recommendation":"major_revision","confidential_remarks":"The central issue is the distributional gap in Proposition 3.1: the decomposition is false as stated for the derivative-of-delta supermomenta used in the examples. This is not a reason to reject the manuscript, because the construction seems repairable by working in the appropriate distributional dual space, but it must be fixed and the orbit/wavefunction formulas re-examined for such N. The paper is otherwise careful and contains many explicit and useful computations; I would be happy to consider a revised version."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know before you open it. First, the paper actually delivers on its abstract: explicit wavefunctions, branching rules, and a memory example for BMS UIRs. Second, the load-bearing decomposition (Prop. 3.1) is stated too broadly; as written, it fails for distributional supermomenta, and the authors' own example soft charges are distributional.\n\nThe genuinely new parts are the Lorentz-invariant hard/soft split, the completion of the massless branching analysis (Propositions 6.10, 6.11), and the normalizable BMS state with memory in Section 9. The appendices contain careful distributional identities and a real proof of the Paneitz exact sequence. That is reproducible math and should be credited.\n\nNow the soft spot, and it is load-bearing. Proposition 3.1 claims every supermomentum decomposes uniquely as hard plus ∂²_z∂²_z̄ N with N a smooth conformal density of weight one modulo translations. But the massless hard piece is a δ, so for a supermomentum like δ + ∂²_z δ the soft part is ∂²_z δ. That annihilates translations, so it is soft, but it is not the Paneitz image of any smooth N; in spherical harmonics it has infinitely many nonvanishing high-ℓ modes, and the prospective N would have coefficients decaying like ℓ^{−3/2}, i.e., not smooth. The authors' own examples (6.20) and (6.21) are precisely of this type, and Appendix A.2 proves only a dual statement with N in a larger distributional dual space. So the proposition as stated, and the proof invoking Proposition 2.14, does not close the gap. The repair is likely straightforward—allow N to be a distribution modulo translations—but it changes the target space in Proposition 3.1 and will ripple through the orbit parametrization and wavefunction formulas. This is a major revision point, not a typo.\n\nOther caveats are minor by comparison. The classification in the nuclear topology is still incomplete (footnote 16), and Sections 8 and 9 are informal, as the authors admit.\n\nWho is this for? People working on asymptotic symmetries, infrared structure, or continuous-spin representations will get real value from the explicit branching rules and the memory construction, once the decomposition is fixed.\n\nRecommendation: send it to peer review, but the referee should force a careful statement of Proposition 3.1 in the distributional setting before acceptance.","headline":"Serious paper with a repairable but load-bearing gap: Proposition 3.1 as stated fails for distributional soft charges, and the paper's own examples use them.","tokens_in":69317,"tokens_out":4091,"would_cite":false,"duration_ms":42093,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that every BMS supermomentum splits uniquely into a hard momentum piece and a soft charge, yielding explicit wavefunctions for generic BMS particles with infinite Poincaré content and memory-carrying states.","keywords":["BMS group","supermomentum","hard/soft decomposition","unitary irreducible representations","continuous-spin representations","memory effect","gravity vacua","soft charges"],"falsifier":"Exhibit a supermomentum with two distinct Lorentz-invariant hard/soft decompositions (or a soft distribution outside the image of the Paneitz operator): Proposition 3.1 would fail and the classification built on it would collapse. Alternatively, find a unitary irreducible representation of BMS4 in the nuclear topology that is strictly ergodic — not induced from a supermomentum orbit — which would show the explicit wavefunctions miss some generic particles; a reader could also check whether the Hilbert-topology exhaustion proof extends, since the paper explicitly leaves that open. A lighter test targets the memory example: compute the full $L^2$ norm of the Section 9 state including subleading corrections in $\\|B_z^2C\\|_\\infty/\\omega$; if that norm diverges, the advertised memory state is not normalizable.","tokens_in":68145,"feed_emoji":"⚛️","tokens_out":16572,"duration_ms":141685,"temperature":0.7,"pith_summary":"The paper sets out to show that the unitary irreducible representations of the BMS group — the asymptotic symmetry group of flat spacetime — can be built explicitly and uniformly, one particle at a time, from a single canonical fact: every supermomentum $P(z,\\bar z)$ splits uniquely and Lorentz-invariantly into a hard piece, a fixed nonlinear function of an ordinary four-momentum $p_\\mu$, plus a soft charge $B_z^2B_{\\bar z}^2N$ that measures how far the state departs from an ordinary Poincaré particle. From that split the authors write down explicit wavefunctions, inner products, and group actions for generic BMS particles, and show that restricting such a particle to any Poincaré subgroup always produces an infinite multiplet of ordinary particles — including, in the massless case, continuous-spin representations. They also display a normalizable BMS state whose average supermomentum carries a soft, memory-like piece that hard states cannot carry. If the construction is right, it matters because it gives representation theory a concrete route into gravitational infrared physics: supermomentum conservation forces soft charges into scattering, and two observers in different gravity vacua will disagree about which ordinary particles a given BMS state contains.","feed_headline":"Every supermomentum splits uniquely into hard and soft parts","feed_subtitle":"The split yields explicit wavefunctions for generic BMS particles and ties soft charges to gravitational memory.","key_machinery":"The central object is the hard/soft decomposition of supermomenta, carried by the two-dimensional Paneitz operator $\\eth^2\\bar\\eth^2 = B_z^2B_{\\bar z}^2$ and the exact sequence $0\\to R^{3,1}\\hookrightarrow E[1]\\xrightarrow{B_z^2B_{\\bar z}^2} E[-3]\\xrightarrow{\\pi} (R^{3,1})^*\\to 0$, which identifies soft supermomenta — those annihilating every translation — with the image of the Paneitz operator on the quotient $E[1]/R^{3,1}$ of supertranslations by translations. Proposition 3.1 converts this into a canonical pair $(p_\\mu, N)$: the hard piece is fixed by the momentum, the soft charge $B_z^2B_{\\bar z}^2N$ is fixed up to that momentum, and the whole decomposition is Lorentz-invariant but not linear. The non-linearity is the striking feature: adding two hard supermomenta produces a soft remainder built from the Weinberg soft factor $S=(q\\cdot p_1)\\ln|q\\cdot p_1|+(q\\cdot p_2)\\ln|q\\cdot p_2|-(q\\cdot p_3)\\ln|q\\cdot p_3|$, which is why hard particles alone cannot conserve supermomentum. Feeding the decomposition into the Wigner-Mackey induction recipe then yields everything else: supermomentum orbits with little groups (stabilizers) given by the intersection of the Poincaré little group and the soft little group, explicit wavefunctions in the coordinates $(\\omega,\\zeta,\\bar\\zeta,\\alpha,\\beta,\\bar\\beta)$ of $SL(2,\\mathbb C)$, and branching rules obtained by Fourier decomposing over the Euclidean little group, with the Fourier label $|\\vec\\pi|$ becoming the continuous-spin parameter.","core_discovery":"The paper's load-bearing claim is Proposition 3.1: every supermomentum $P(z,\\bar z)\\in E[-3]$ admits a unique, $SL(2,\\mathbb C)$-invariant decomposition $P = \\mathbb P + B_z^2B_{\\bar z}^2 N$, where the hard part $\\mathbb P$ is the specific nonlinear function of the momentum $p_\\mu=\\pi_\\mu(P)$ written as $-(m^4/\\pi)(p\\cdot q(z,\\bar z))^{-3}$ for massive momenta and $\\omega\\,\\delta^{(2)}(z-\\zeta,\\bar z-\\bar\\zeta)$ for massless ones, and where the soft charge $B_z^2B_{\\bar z}^2N$ is built from a shift of vacua $N\\in E[1]/R^{3,1}$. Reading McCarthy's classification through this lens, the paper organizes BMS particles into hard (vanishing soft charge), soft (vanishing momentum), and generic (both nonzero) representations, and insists that generic ones — whose little group is typically trivial — are the physically relevant case. For those it provides explicit wavefunctions on $SL(2,\\mathbb C)$, the action of supertranslations, and branching rules: a generic massless BMS particle becomes, in any Poincaré subgroup, one massless particle of every helicity plus a direct integral of continuous-spin representations, while a generic massive one becomes a tower of all spins with multiplicity $2j+1$. Finally the paper constructs a finite-norm BMS state built from a scalar field superposed with its own supertranslated copy in another gravity vacuum, and shows its average supermomentum acquires a soft part $B_{\\bar z}^2\\langle N\\rangle$ — a memory effect that hard/Poincaré states cannot produce. The upshot, stated on the paper's own terms, is that BMS particles with nonzero soft charge are exactly as legitimate as ordinary hard particles, and that the gravitational infrared structure is encoded in the representation theory rather than added by hand.","pith_inferences":["Reading the soft charge as the representation-theoretic counterpart of the usual dressed-state construction suggests a cleaner route to infrared-finite scattering: each BMS particle carries its own soft charge, and the Fock space built from such particles is automatically separable; the paper leaves the connection to explicit dressed amplitudes implicit.","The Lorentz-invariant distance between gravity vacua introduced in Corollary 8.3 is a natural control parameter for memory: one could test whether the memory amplitude of generic superpositions scales with that distance, an extension the paper only sketches at the end of Section 9.","Because every generic massless BMS particle branches to continuous-spin representations, any BMS-invariant S-matrix would evade the Coleman-Mandula theorem, whose finiteness assumptions the infinite Poincaré multiplet manifestly violates; whether a nontrivial S-matrix built on these multiplets actually exists is an open, testable question.","The Gaussian-regularized states of Section 6.5, which soften the $L^2$ norm enough to admit energy poles, offer a concrete mechanism for constructing infrared-finite amplitudes: checking whether their Fock-space amplitudes reproduce the standard soft theorems in the limit where the continuous-spin cutoff is removed would be a direct test of the whole picture."],"forward_implications":["Hard BMS representations coincide with Poincaré representations: their BMS and Poincaré little groups are equal, they do not branch under restriction to any Poincaré subgroup, and observers in different gravity vacua agree on their particle content.","Hard supermomenta do not add linearly, so hard particles alone can never conserve supermomentum in a scattering process; any BMS-invariant process must involve states with nonzero soft charge, and the soft charge produced by adding two hard particles is exactly the Weinberg soft factor.","A generic massless BMS particle restricts to any Poincaré subgroup as one massless particle of every helicity (all integer or all half-integer, multiplicity one) plus a direct integral of continuous-spin representations over all $\\mu>0$; a generic massive one restricts to a tower of all spins $j$, each with multiplicity $2j+1$, all of the same mass.","The interpretation of a BMS state depends on the gravity vacuum: a state that is a single scalar in one vacuum is an infinite superposition of particles of all spins, including continuous spin, in another.","Generic BMS particles can encode gravitational memory: a specific normalizable state formed by superposing a scalar field with its supertranslated copy has average supermomentum with a soft, memory-like part that hard/Poincaré states cannot carry, because their $L^2$ norm forbids the required energy pole."],"supporting_citations":[{"why":"McCarthy's series that classified the UIRs of BMS4 and determined their little groups; this paper revisits and makes those representations explicit.","marker":"[8–17]"},{"why":"The nuclear-topology classification paper whose little-group tables (Theorem 5.1) and branching results the present paper uses and completes for the massless case.","marker":"[14]"},{"why":"Crampin-McCarthy branching rules for Poincaré spin multiplicities, which the paper re-derives explicitly in Sections 6 and 7.","marker":"[12]"},{"why":"Gelfand-Graev-Vilenkin's SL(2,C) representation theory, which underlies the conformal-density spaces and the Paneitz exact sequence.","marker":"[45]"},{"why":"The authors' earlier construction of BMS wavefunctions as superpositions of Poincaré particles over gravity vacua, which this paper proves in detail.","marker":"[31]"},{"why":"Ashtekar's asymptotic quantization: the notion of gravity vacua and the no-memory result for hard representations that Section 9 addresses.","marker":"[5]"},{"why":"The realization of BMS as a global symmetry of gravitational scattering and the derivation of supermomentum conservation from soft theorems.","marker":"[6, 7]"},{"why":"Piard's proof, in the Hilbert topology, that the induced representations exhaust the UIRs — the completeness result whose nuclear-topology version is still missing (footnote 16).","marker":"[54]"},{"why":"The Longhi-Materassi massive hard BMS realization, identified here as the massive hard case of the decomposition.","marker":"[30]"}],"fun_headline_variants":["Supermomenta split uniquely into hard and soft","Unique hard-soft split for all BMS supermomenta","Generic BMS particles encode memory via soft charge","BMS representations: hard, soft, and generic particles","Every supermomentum: one hard part, one soft part"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The construction assumes that every unitary irreducible representation of BMS4 is obtained by the standard Wigner-Mackey induction recipe — choose an $SL(2,\\mathbb C)$-orbit of supermomenta, choose a unitary representation of the little group, build square-integrable sections — and the paper itself flags (Section 4.2, footnote 16) that this exhaustion is proven only in the Hilbert topology, not in the nuclear topology that is physically preferred; if additional, ergodic representations exist that escape the recipe, the explicit wavefunctions would not cover 'generic' BMS particles.","fun_headline_variants_meta":{"raw":{"variants":["Supermomenta split uniquely into hard and soft","Unique hard-soft split for all BMS supermomenta","Generic BMS particles encode memory via soft charge","BMS representations: hard, soft, and generic particles","Every supermomentum: one hard part, one soft part"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000195,"raw_usage":{"total_tokens":1453,"prompt_tokens":1136,"completion_tokens":317,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":752,"completion_tokens_details":{"reasoning_tokens":237}},"tokens_in":752,"tokens_out":317,"duration_ms":3317,"temperature":1.0,"reasoning_tokens":237,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T23:05:42.051396+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibit a supermomentum with two distinct Lorentz-invariant hard/soft decompositions (or a soft distribution outside the image of the Paneitz operator): Proposition 3.1 would fail and the classification built on it would collapse. Alternatively, find a unitary irreducible representation of BMS4 in the nuclear topology that is strictly ergodic — not induced from a supermomentum orbit — which would show the explicit wavefunctions miss some generic particles; a reader could also check whether the Hilbert-topology exhaustion proof extends, since the paper explicitly leaves that open. A lighter test targets the memory example: compute the full $L^2$ norm of the Section 9 state including subleading corrections in $\\|B_z^2C\\|_\\infty/\\omega$; if that norm diverges, the advertised memory state is not normalizable.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gelfand-Graev-Vilenkin's SL(2,C) representation theory, which underlies the conformal-density spaces and the Paneitz exact sequence."},{"cited_title":"Representations of the bondi-metzner-sachs group with the hilbert topology,","cited_arxiv_id":null,"evidence_quote":"Piard's proof, in the Hilbert topology, that the induced representations exhaust the UIRs — the completeness result whose nuclear-topology version is still missing (footnote 16)."}],"review_version":1}