{"id":"162f1476-dd5e-4a16-8449-7a580b663623","arxiv_id":"2412.13868","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For bosons on a lattice, the mean-field approximation error far from the initial condensate is bounded by any inverse power of the distance for times up to a distance-dependent light cone.","lead":"This paper proves that, for a lattice Bose-Einstein condensate in the mean-field regime, the usual mean-field approximation is much more accurate in regions far from the initial condensate, at least for short times. The error decays like an arbitrary power of the distance to the condensate, a new local refinement of global convergence results.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The main theorem inherits its ballistic Hartree-flow bound from [Zha24b] without proof; if that bound requires extra hypotheses, Theorem 1.1 is not established as stated.","rationale":"The reader identified (DE) as the most fragile assumption and Theorem 5.2 as a second dependency. I agree that Theorem 5.2 is load-bearing, but I rate it as the single most critical gap because it is not proved anywhere in the manuscript, whereas (DE) is at least partially verified in Appendix A under explicit small-coupling conditions. The proof of Theorem 1.1 needs the Hartree flow to remain away from the observation region with polynomial precision; that is exactly the content of Theorem 5.2. If that bound fails, the suppression in (5.3) and the estimates for the second and third terms in Lemma 6.1 collapse, so the central claim would not be supported. I found no internal contradiction in the main proof, and the issue is not that the paper is wrong but that it is conditional on an unverified external result. A conditional acceptance, pending verification of the imported ballistic bound, matches the evidence. The reader's ACCEPT is defensible given usual citation practice, but the unpublished nature of [Zha24b] and the centrality of Theorem 5.2 make a conditional verdict more precise.","tokens_in":36765,"tokens_out":16022,"duration_ms":148792,"concrete_test":"Extract the proof of [Zha24b, Thm. 1.4] and check, line by line, that it applies to the Hartree nonlinearity lambda|phi|^2 phi on Z^d for d >= 3 and to the geometric choice Y = supp phi_0, with B_r subset Y^c_rho. Confirm that the constant C(n, M, v) is independent of rho and of the solution, and that no extra smallness of ||phi_0||_1, no additional decay, and no re-striction on |lambda| beyond the stated assumptions is needed. If the adaptation requires extra hypotheses, either add them to Theorem 5.1 and Theorem 1.1 or include a self-contained proof in an appendix.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim rests on two structural assumptions: the dispersive estimate (DE), which is verified in Appendix A for small |lambda|, and the ballistic upper bound for the nonlinear Schroedinger flow, stated as Theorem 5.2 and imported from [Zha24b] with no proof in this manuscript. Theorem 5.2 is used in the proof of Theorem 5.1 (step 2) to show that the Hartree condensate remains supported away from the observation ball, so that the quantities MR(t) and ER,r(t) in Theorem 3.1 are small, and again in the proof of Theorem 1.1 to control the second and third terms in Lemma 6.1. Without this bound, the local fluctuation estimate (5.3) and hence Theorem 1.2 and Theorem 1.1 do not follow. The paper only states that the exact same statement is proved in [Zha24b, Thm. 1.4] and that a straightforward adaptation extends it to general subsets Y; no details, constants, or hypotheses are given. Because [Zha24b] is an unpublished preprint by one of the authors, this is a load-bearing missing proof rather than a mere citation. The concern is not that the bound is false, but that the central theorem is conditional on an unverified external ingredient.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript studies the mean-field dynamics of N bosons on the lattice Z^d with Hamiltonian (1.3), starting from an approximate condensate. Its central claim, Theorem 1.1, is a local enhancement of the mean-field approximation: if the initial condensate is supported outside B_{r+rho} and the observable is supported in B_r, then for times t <= rho/v the error |Tr((gamma_{psi_N,t} - N|phi_t><phi_t|)O)| is bounded by C ||O||_op rho^{-n} for every integer n >= 1, under a dispersive assumption (DE). The proof mechanism is a ballistic propagation bound for fluctuations, Theorem 1.2, obtained by adapting the ASTLO method to the particle non-conserving generator of the fluctuation dynamics. The paper also contains a global fluctuation estimate (Theorem 4.1) and a verification of (DE) for small |lambda| (Appendix A).","tokens_in":36987,"tokens_out":12942,"duration_ms":111763,"significance":"If the result is fully established, it is a genuine new local improvement over the standard global 1/N mean-field error bound, and the ASTLO adaptation to the fluctuation generator is an interesting methodological contribution. I credit the paper for deriving most of the technical estimates in the text, including the commutator bounds of Lemma 2.1, the local fluctuation estimate of Theorem 3.1, and the dispersive verification in Appendix A; there are no fitted parameters or post-hoc exclusions. However, the central theorem inherits one load-bearing ingredient, the ballistic upper bound for the nonlinear Schroedinger flow, from an unpublished preprint without proof, and this makes the current version conditional.","major_comments":[{"comment":"In the paragraph after (5.4), the proof says 'Inserting the global fluctuation bound (4.7) into (5.3)'. Equation (4.7) is Corollary 4.3, a local fluctuation estimate, not a global bound; the argument only needs Theorem 4.1 to control sup_tau <N+1>_tau. This is a misreference rather than a mathematical gap, but it should be corrected.","section":"Section 5, proof of Theorem 1.2"}],"minor_comments":[{"comment":"Assumption (1.16) is missing a closing angle bracket: it should read <psi_N,0, N^+_{B_{r+rho}}(0) psi_N,0> = 0.","section":"Equation (1.16)"},{"comment":"The displayed dispersive estimate has the norm indices swapped: it should be ||e^{itDelta} f||_infty <= C1 <t>^{-d/3} ||f||_1, which is the form used later in (A.10).","section":"Appendix A, equation (A.2)"},{"comment":"The reference to 'global fluctuation bound (4.7)' should be to Theorem 4.1; equation (4.7) is a local bound from Corollary 4.3.","section":"Section 5, proof of Theorem 1.2"},{"comment":"'Not that' should be 'Note that'.","section":"Remark 3.1"},{"comment":"In (2.23), 'as on operator inequality' should read 'as an operator inequality'.","section":"Lemma 2.1"},{"comment":"The notation i\\phi_-(h) is introduced twice; define it once, earlier in the proof.","section":"Section 6.1, equation (6.31)"}],"recommendation":"major_revision","confidential_remarks":"For the editor: the core argument of the paper appears coherent and the ASTLO adaptation is promising, but I cannot accept the paper in its current form because Theorem 5.2 is an unproved, load-bearing external result from an unpublished preprint by one of the authors. If the authors supply a full proof of Theorem 5.2 (or replace it by a published reference with verified hypotheses), I would be inclined to accept. The reader's report is more optimistic than mine on this specific point."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper delivers what it promises: a rigorous local enhancement of the mean-field approximation for lattice bosons, with the error at distance ρ improved to ρ^{-n} for any n up to time t ≤ ρ/v. The core idea—adapting the ASTLO machinery to the particle non-conserving generator of the fluctuation dynamics—is a real novelty, and the local fluctuation bounds in Theorem 1.2 are not consequences of the known global N^{-1} results. The proof is careful, well organized, and the assumptions are stated plainly. The paper also does the reader the courtesy of explaining when the dispersive estimate (DE) is actually available (small coupling, d ≥ 4 globally, d = 3 up to an exponentially large time), with a sensible remark that the short-time regime is the relevant one.\n\nThe main soft spot is exactly what the stress-test flags: Theorem 5.2, the ballistic upper bound for the nonlinear Schrödinger flow, is imported from the unpublished preprint [Zha24b] and not proved here. This bound is load-bearing—it controls the moving condensate support in the proofs of Theorem 5.1 and Theorem 1.1. Without it, the claimed ρ^{-n} enhancement does not follow. The citation is honest and the statement is explicit, but for a self-contained paper this is a gap. A referee would reasonably ask either for a proof of Theorem 5.2 in an appendix or for a precise statement of the hypotheses and constants from [Zha24b] so the dependency is fully visible. This is a revision request, not a reason to reject—the bound itself is plausible and the rest of the argument hangs together.\n\nThere are also minor issues: the typo in assumption (1.16) (missing angle bracket) and the fact that (DE) in d = 3 only holds up to T = exp(1/√|λ|) − 1, which limits the time window for the largest ρ values. Neither changes the main picture.\n\nWho is this for? Mathematical physicists working on mean-field limits, Lieb-Robinson bounds, or effective equations for Bose gases. It deserves a serious referee. I would ask the authors to either integrate the ballistic lemma or clearly state it as a black box with full hypotheses, then accept after that check.","headline":"Genuinely new local refinement of the mean-field program with a clean ASTLO adaptation, but the main theorem leans on an unproved ballistic bound imported from an author's preprint.","tokens_in":37565,"tokens_out":1946,"would_cite":true,"duration_ms":20197,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q55","81V70","82C10"],"pacs":[],"model":"deepseek-v4-flash","headline":"For bosons on a lattice, the mean-field approximation error in a local observable decays faster than any power of the distance from the initial condensate, for times up to that distance divided by a finite velocity.","keywords":["mean-field approximation","Bose-Einstein condensate","lattice bosons","propagation bound","Lieb-Robinson","Hartree equation","fluctuation dynamics","adiabatic spacetime localization observable"],"falsifier":"A direct numerical simulation of the lattice nonlinear Schrödinger equation (1.5) in d = 3 at small λ, measuring ∫₀ᵗ ‖φ_s‖_∞ ds for times up to the claimed T = exp(1/√|λ|) − 1, would verify whether the dispersive assumption actually holds; a violation for t < T would falsify the assumption. Alternatively, compute the mean-field error for a local observable in a ball at distance ρ for times t slightly exceeding ρ/v; if the error grows as a power of N rather than remaining O($ρ^{{-n}}$), the local enhancement bound fails.","tokens_in":36537,"feed_emoji":"🧊","tokens_out":6517,"duration_ms":57419,"temperature":0.7,"pith_summary":"This paper asks how accurately a Bose-Einstein condensate evolving under the many-body Schrödinger equation on a lattice is described by the single-particle Hartree equation. The global mean-field error is known to be of order 1/N, but the paper shows that the error is much smaller in spatial regions far from the initial condensate: for every integer n, the error in a local observable at distance ρ is O($ρ^{{-n}}$) for times t ≤ ρ/v, where v is an order-1 velocity. This local enhancement is derived from new ballistic propagation bounds on the fluctuations around the condensate, which say that the number of excitations inside a ball at distance ρ stays O($ρ^{{-n}}$) over those times. The method is a variant of the adiabatic spacetime localization observable (ASTLO) technique adapted to the particle non-conserving generator of the fluctuation dynamics. If correct, the result shows that information about the condensate shape travels at finite speed and that local measurements far from the condensate are described by Hartree dynamics to extreme precision.","feed_headline":"Boson mean-field error vanishes faster than any power of distance","feed_subtitle":"For times up to distance divided by a velocity, local observables far from the condensate see Hartree dynamics to O(ρ^{-n}) precision.","key_machinery":"The key object is the fluctuation dynamics W_N(t;s) on the truncated Fock space of excitations orthogonal to the Hartree state, whose generator L_N(t) contains non-particle-conserving quadratic terms of the form b^*b^* and bb. The paper adapts the adiabatic spacetime localization observable (ASTLO) method to this generator: it constructs smoothed spatial cutoffs f((R−v′t−|x|)/s) that move inward at a velocity v′ and proves a differential inequality showing that these observables decay except for small remainders. A Grönwall bootstrap and an induction in the parameter n convert the differential inequality into a local estimate on the excitation number in a ball, and the dispersive estimate (DE), which states that ∫₀ᵗ ‖φ_s‖_∞ ds ≤ c‖φ₀‖₁, controls the time integrals of the Hartree solution. A ballistic upper bound for the nonlinear Schrödinger flow, taken as a black box, keeps the condensate away from the probed ball over the relevant times.","core_discovery":"The central claim is that the mean-field approximation is locally enhanced by arbitrary polynomial factors in the distance from the initial condensate. Concretely, if the initial condensate wave function φ₀ vanishes on a ball of radius r+ρ and the observable O is supported in the ball of radius r, then for every integer n ≥ 1 the error |Tr((γ_{ψ_{N,t}}/N − |φ_t⟩⟨φ_t|)O)| is bounded by C‖O‖_op $ρ^{{-n}}$ for all times t ≤ ρ/v, where v > 4d is a state-independent velocity and C depends only on n, v, d, |λ|, the dispersive constant c, and the ℓ¹ norm of φ₀. The proof shows the same for the local fluctuation number: ⟨ψ_{N,t}, N⁺_{B_r}(t)ψ_{N,t}⟩ ≤ C $ρ^{{-n}}$ for t ≤ ρ/v, assuming the initial state has no fluctuations near B_r. These bounds are the first of their kind for the particle non-conserving generator of fluctuations around a Hartree state.","pith_inferences":["A natural testable extension is to check numerically on a lattice whether the O(ρ^{-n}) local enhancement persists for times longer than ρ/v or for interactions beyond the small-|λ| regime; the proof suggests the finite-speed propagation should break down at a scale set by the discrete Schrödinger group velocity.","The result implies that local entanglement generation between distant regions is suppressed for times up to ρ/v, connecting to Lieb-Robinson-type bounds for bosonic systems in the mean-field scaling.","Since the error depends on ‖φ₀‖_{ℓ¹}, condensates that are more spread out (larger ℓ¹ norm for fixed ℓ² norm) have weaker enhancement; one could investigate whether an ℓ²-based assumption suffices.","The ASTLO adaptation to non-particle-conserving generators may apply to other quadratic fluctuation generators, such as Bogoliubov dynamics on lattices, giving ballistic propagation bounds for quasiparticles."],"forward_implications":["For t ≤ ρ/v, any bounded local observable supported at distance ρ from the initial condensate is described by the Hartree evolution up to an error O(ρ^{-n}) for every n, which is far better than the global O(1/N) bound.","Fluctuations around the condensate propagate at most at a finite speed v; the local number of excitations in a ball at distance ρ stays O(ρ^{-n}) for times up to ρ/v.","Because the bound holds for every n simultaneously, the local error is rapidly decreasing in ρ, not merely suppressed by one fixed power.","Without the dispersive assumption (DE), the same argument yields a bound Ce^{C|λ|t}ρ^{-n}, so spatial locality persists but with exponential time growth.","The method extends to initial states that are not pure condensates, provided the initial fluctuations are localized away from the probed region."],"supporting_citations":[{"why":"Supplies the original ASTLO method for Bose-Hubbard Hamiltonians that this paper adapts to the fluctuation generator.","marker":"[FLS22b]"},{"why":"Provides the maximal-speed propagation estimates for Bose-Hubbard models that motivate the ballistic picture.","marker":"[FLS22a]"},{"why":"Develops ASTLO for long-range systems and supplies the commutator estimate for the discrete Laplacian used in the differential inequality.","marker":"[LRZ23]"},{"why":"Proves the ballistic upper bound for the nonlinear Schrödinger flow that keeps the condensate support away from the probe region.","marker":"[Zha24b]"},{"why":"Establishes the ℓ¹ to ℓ∞ dispersive estimate for the discrete free propagator that underlies the dispersive assumption (DE).","marker":"[SK05]"},{"why":"Gives the uniform-in-time global fluctuation bound under dispersion, which the local estimate refines.","marker":"[DL23]"},{"why":"Provides the standard global mean-field convergence rate of order 1/N that the new bound locally enhances.","marker":"[RS09]"},{"why":"Introduces the fluctuation dynamics framework and the excitation number operator that the paper controls locally.","marker":"[LNS15]"}],"fun_headline_variants":["Boson mean-field error decays faster than any polynomial with distance","Local boson mean-field error vanishes as any power of distance","Mean-field approximation for bosons improves locally with distance","Far from boson condensate, mean-field error drops super-polynomially","Local boson approximation: error suppressed by any power of distance"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole result depends on the dispersive estimate (DE): the Hartree solution φ_t must satisfy ∫₀ᵗ ‖φ_s‖_∞ ds ≤ c ‖φ_0‖₁ for all times up to the ones considered, and this is only proven here for small interaction strength |λ| (and in dimension 3 only up to a large but finite time).","fun_headline_variants_meta":{"raw":{"variants":["Boson mean-field error decays faster than any polynomial with distance","Local boson mean-field error vanishes as any power of distance","Mean-field approximation for bosons improves locally with distance","Far from boson condensate, mean-field error drops super-polynomially","Local boson approximation: error suppressed by any power of distance"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001432,"raw_usage":{"total_tokens":5751,"prompt_tokens":894,"completion_tokens":4857,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":510,"completion_tokens_details":{"reasoning_tokens":4769}},"tokens_in":510,"tokens_out":4857,"duration_ms":31648,"temperature":1.0,"reasoning_tokens":4769,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T12:42:31.810367+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct numerical simulation of the lattice nonlinear Schrödinger equation (1.5) in d = 3 at small λ, measuring ∫₀ᵗ ‖φ_s‖_∞ ds for times up to the claimed T = exp(1/√|λ|) − 1, would verify whether the dispersive assumption actually holds; a violation for t < T would falsify the assumption. Alternatively, compute the mean-field error for a local observable in a ball at distance ρ for times t slightly exceeding ρ/v; if the error grows as a power of N rather than remaining O($ρ^{{-n}}$), the local enhancement bound fails.","supporting_citations":[],"review_version":1}