{"id":"ce0264c0-accc-4919-9b36-010f3de66993","arxiv_id":"2607.28963","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"As ε→0, solutions of the Klein-Gordon–Schrödinger system stay within O(ε²) of a pair of decoupled linear Schrödinger equations on a long time interval.","lead":"This paper proves that the Klein-Gordon–Schrödinger system converges to two decoupled linear Schrödinger equations in the nonrelativistic limit, with error of order ε² over long time. It rigorously explains why numerical simulations see this rate and gives a sharper uniform-in-time error for the Schrödinger component.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The advertised O(ε²) optimality is not established: the proof gives upper bounds only, and the WKB-heuristic in §1.3 does not rule out cancellations that could make the exact error o(ε²).","rationale":"The reader's stated weakest assumption is the W^{σ,1} dispersive-decay condition in Lemma 1.1. This is a strong regularity/decay assumption, but it is explicitly stated in the theorems and appears necessary for the proof; I see no internal inconsistency there. However, the reader's rationale also flags the unsupported optimality claim as the condition for the verdict, and that is the load-bearing issue I independently identify. The proof of the upper bounds appears plausible and internally consistent—modulo a secondary misstatement in (2.32), where ∥ψ_c(t)∥_{Hσ} ≤ C(1+t)^{−d+1} is false (the integral ∫_0^t (1+s)^{−d} ds is bounded but does not decay); replacing it by the correct uniform bound only improves the bootstrap constants, so it is not load-bearing. The absence of a lower bound for the distance to the leading profile means the word 'optimal' in the title/abstract is not justified by the arguments in the paper. The numerical agreement cited from [6] is not a proof. Therefore the appropriate outcome is a conditional acceptance pending either a lower-bound theorem or a qualified formulation of optimality, matching the reader's CONDITIONAL verdict; hence I record UNCHANGED.","tokens_in":24650,"tokens_out":18643,"duration_ms":153741,"concrete_test":"Analyze the linearized/decoupled case: set ψ0=0 so the KGS system reduces to the linear Klein-Gordon equation. With Gaussian (or Schwartz) initial data in d=3, compute the exact solution via Fourier. For fixed t (e.g., t=1), expand ∥µϕ(t)−(e^{iµt/ε²}ϕℓ(t)+c.c.)∥_{Hσ} in powers of ε and evaluate the ε² coefficient. If this coefficient is nonzero, the O(ε²) rate is genuinely optimal in this setting; if it vanishes (so the error is o(ε²)), the claimed optimality would be refuted and would indicate cancellations are possible.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The title and abstract claim the O(ε²) convergence rates are 'optimal.' Theorems 1.3 and 1.4 prove only upper bounds. The only support for optimality is the final paragraph of §1.3: the WKB expansion contains O(ε²) components (Φ2, ψ2), so increasing the expansion order cannot improve the final rate. This shows that the WKB approximate solution (Φa,ψa) is at distance O(ε²) from the leading profile, but it does not show that the exact solution (Φ,ψ) is. The exact solution could, in principle, have a different O(ε²) correction that cancels these terms, yielding a faster rate. No lower bound (e.g., ∥µϕ−(e^{iµt/ε²}ϕℓ+c.c.)∥_{Hσ} ≥ c ε² for some t) is provided. Remark 2.1 even observes that the initial error can be reduced to O(ε³) by choosing nonzero initial data for (ϕc,ψc), so 'the order of initial error' is not an intrinsic obstruction. The optimality assertion is thus heuristic, not a theorem; the convergence results stand, but the 'optimal' claim requires proof or a qualified statement.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the nonrelativistic limit ε→0 of the Klein-Gordon–Schrödinger system (1.1)–(1.2) in dimensions d≥3. It constructs WKB approximate solutions whose leading profiles solve two free Schrödinger equations (1.5), and proves by Duhamel/bootstrap estimates that the exact solution stays within C(1+t)ε² in H^σ on time scales of order ε^{-1} (Theorem 1.3). Under additional regularity it proves the longer time scale ε^{-4/3} with bounds C(1+εt+ε²t²)ε², yielding a uniform O(ε²) bound for the Schrödinger component on the ε^{-1} interval (Theorem 1.4 and Corollary 1.5). The title and abstract further claim that the O(ε²) rates are optimal.","tokens_in":24944,"tokens_out":10986,"duration_ms":89591,"significance":"If read as an upper-bound paper, this is a solid contribution. The WKB construction in §2 and the bootstrap argument in §3 are coherent and the stated bounds follow from them; the proof is self-contained modulo standard dispersive estimates and the Kato–Ponce inequality, with no fitted parameters and no circular dependence on the numerical results of [6], which serve only as motivation. The results rigorously justify the numerically observed rates and improve the existing long-time convergence literature for this system. The central weakness is the advertised optimality: the paper proves only upper bounds, and the argument offered in §1.3 shows only that the WKB approximate solution is not closer to the leading profile; it does not show that the exact solution cannot be closer. This is a load-bearing overstatement in the title and abstract, but it is local and fixable by either proving a lower bound or qualifying the claim.","major_comments":[{"comment":"The paper claims the O(ε²) rates are 'optimal,' but Theorems 1.3 and 1.4 establish upper bounds only. The argument in the final paragraph of §1.3 — that the WKB expansion contains O(ε²) components and hence higher-order WKB cannot improve the rate — concerns the distance between the WKB approximate solution and the leading profile, not the distance between the exact solution and the leading profile. The exact solution could in principle have a different O(ε²) correction that cancels these terms. No lower bound such as ‖error‖ ≥ c ε² for some t is proved. Moreover, Remark 2.1 explicitly states that the initial error can be reduced to O(ε³), so the abstract's inference 'coincide with the order of initial error, and thus are optimal' is not valid. I recommend either proving a lower bound for a natural class of data or replacing 'optimal' by a qualified statement such as 'optimal among WKB l","section":"Abstract; §1.3, final paragraph; Remark 2.1"}],"minor_comments":[{"comment":"The symbol T1 is used both for the dimensionless constant T1 = 1/(2M1) and for the time scale T1/ε. This makes the bootstrap definition and the statement of Theorem 1.3 confusing. Suggest using, e.g., τ1 for the constant and T1/ε for the time horizon.","section":"§1.2, notation"},{"comment":"The displayed formula for rΦ reads rΦ = −ε²(0, µ∂tϕℓ, λ/µ |ψℓ|²)^T. This appears dimensionally inconsistent: from the ansatz (2.27), the ε² term in the second component is ε²/µ(∂tϕℓ e^{iθ} + c.c.), so the formula should contain (∂tϕℓ + ∂tϕℓ)/µ (or the real part) rather than µ∂tϕℓ. The estimate ‖rΦ‖ ≤ Cε² is unaffected, but the displayed formula should be corrected.","section":"§2.3, initial error formula"},{"comment":"The notation [F(ψa,ψa)]_2^{(2)} is ambiguous: the displayed computation contains e^{±iθ} terms but then writes '= 2λ Re(ψℓψc)', silently projecting onto the zero mode. Please clarify that the p=0 mode is being selected here and that the nonzero modes are treated separately.","section":"§2.2, order n=2"},{"comment":"In the display following (3.22), the term '1/2 T2 ε^{4/3}' should read '1/2 T2^2 ε^{4/3}'. The subsequent bound is unaffected, but the typo makes the verification harder.","section":"§3.3, bootstrap exponent"},{"comment":"The sentence 'The proof is based on the energy inequality energy-expan' contains a broken cross-reference; it should refer to (3.6). Also, 'bootstarp' in the same section is a typo.","section":"§3.3, first line"},{"comment":"Minor typographical issues: 'the the approximate system' in §2.4, 'compoents' in §2.3, and in the bibliography [35] 'Ann. Mat. Pura Appl.4198' should be 'Ann. Mat. Pura Appl. 198'. These do not affect the mathematics.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The upper-bound results are strong and, in my view, publishable after revision. The only serious obstacle is the unsupported optimality claim in the title and abstract; Remark 2.1 further shows that the stated justification for optimality is not valid. If the authors prove a genuine lower bound, or alternatively rephrase the claim as sharpness relative to the WKB approximation rather than sharpness of the exact convergence rate, I would be satisfied. The comparison with [6] and with the authors' earlier work is appropriate, and I see no circularity or attribution problem."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague —\n\nThe main thing to know: this paper proves upper bounds that were previously only numerical observations. Theorems 1.3 and 1.4 establish O((1+t)ε²) convergence on the ε⁻¹ time scale for the coupled Klein-Gordon–Schrödinger system, and the uniform-in-time O(ε²) rate for the Schrödinger component ψ on that scale is genuinely new for this system. Prior long-time results were for scalar KG or Klein-Gordon–Zakharov, not this coupled system. The WKB construction is detailed, the profile estimates are checked, and the bootstrap argument for the perturbation system is internally consistent. I did not find a fatal gap in the upper bounds.\n\nThe soft spot is exactly what the stress-test flags: the advertised optimality. The abstract and the end of Section 1.3 claim the O(ε²) rates are optimal, but the paper only proves upper bounds. The WKB argument shows that any WKB approximate solution contains O(ε²) components, so increasing the expansion order does not bring the WKB solution closer than O(ε²) to the leading profile. That says nothing about the exact solution, which could in principle have cancellations. Remark 2.1 even acknowledges the initial error can be reduced to O(ε³) by choosing nonzero corrector data, so the 'order of initial error' is not an intrinsic obstruction. A genuine lower bound — say, ∥µϕ − (e^{iµt/ε²}ϕ_ℓ + c.c.)∥ ≥ c ε² for some time — is absent. That is a real overclaim, but it does not undermine the convergence theorems. The fix is easy: either prove a lower bound or replace 'optimal' with 'consistent with the WKB expansion.'\n\nOne more limitation worth stating plainly: the proof leans on W^{σ,1} initial regularity and d≥3 because the dispersive decay (1+t)^{-d/2} must be integrable. That is a strong assumption, but it is explicit and standard for this technique. Not a flaw; just scope.\n\nThe citation pattern is fine. The earlier papers by the same authors provide the machinery, and the target theorem for this coupled system is not already in the literature.\n\nWho is this for: anyone working on singular limits of dispersive PDEs, and numerical analysts who want rigorous backing for observed rates. It deserves a serious referee. My recommendation is to send it to peer review with a clear request to correct or qualify the optimality claims before acceptance. I would cite the convergence results once that language is fixed.","headline":"New long-time O(ε²) bounds for the Klein-Gordon–Schrödinger system are real; the 'optimal' claim is a heuristic overreach that should be qualified or proven.","tokens_in":25464,"tokens_out":1567,"would_cite":true,"duration_ms":15372,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35B40","35Q55"],"pacs":[],"model":"deepseek-v4-flash","headline":"The Klein-Gordon–Schrödinger system converges to two decoupled Schrödinger equations at the optimal O(ε²) rate.","keywords":["Klein-Gordon-Schrödinger system","nonrelativistic limit","optimal convergence rate","WKB expansion","geometric optics","dispersive decay","long time estimates","Schrödinger approximation"],"falsifier":"Run a numerical simulation of the Klein-Gordon–Schrödinger system in dimension d≥3 with initial data in H^{σ+4}∩W^{σ,1}, and measure ∥ψ(t)−ψℓ(t)∥_{H^σ} at t=T₁/ε; if the error is not bounded by C ε² or grows faster than (1+t)ε², the theorem's bound is false. Conversely, a simulation in d=1 or d=2 showing the same O(ε²) rate would indicate the dispersive-decay hypothesis is not necessary.","tokens_in":24530,"feed_emoji":"⚛️","tokens_out":5615,"duration_ms":43916,"temperature":0.7,"pith_summary":"The paper proves that, in the nonrelativistic limit ε→0, the coupled Klein-Gordon–Schrödinger system is approximated, on a long time interval of order 1/ε, by two decoupled linear Schrödinger equations, with errors proportional to ε² (growing at most linearly in time). With more regular initial data the validity extends to times of order ε^{-4/3}, and the error in the Schrödinger component becomes (1+εt+ε²t²)ε². These rates are shown to be optimal: the ε² order coincides with the unavoidable initial error, and adding higher-order WKB terms cannot improve the final rate. The result rigorously justifies numerical observations that had suggested exactly these error forms.","feed_headline":"Optimal ε² convergence proved for Klein-Gordon–Schrödinger","feed_subtitle":"Rigorous proof matches numerics: decoupled Schrödinger profiles stay accurate on time scales up to 1/ε.","key_machinery":"The argument runs through a WKB (geometric-optics) expansion of the symmetric-hyperbolic reformulation of the system, with phase θ=µt/ε². The load-bearing components are the Schrödinger profiles ϕℓ and ψℓ and, at higher order, the explicit correctors ϕc(t)=−it/(8µ³)Δ²ϕℓ(t) and ψc(t)=∫₀ᵗ e^{i(t−s)Δ}(iλ²/µ²)|ψℓ|²ψℓ(s) ds. The proof exploits the dispersive decay estimate ∥f(t)∥_{W^{σ,∞}} ≤ C t^{-d/2}∥f₀∥_{W^{σ,1}} for the linear Schrödinger evolution, which controls the nonlinear differences, and then closes a bootstrap/Grönwall estimate for the perturbation ˙u=(Φ−Φa, ψ−ψa). The WKB approximate solution is constructed so that its residual and initial error are O(ε²) (or O(ε³+ε⁴t) at higher orde","core_discovery":"The central claim is that the exact solution (ϕ,ψ) of the Klein-Gordon–Schrödinger system with well-prepared initial data can be written, up to an H^σ error of order ε², as the leading WKB profiles e^{iµt/ε²}ϕℓ(t)+c.c. for µϕ and ψℓ(t) for ψ, where ϕℓ and ψℓ solve the free Schrödinger equations in (1.5). Theorem 1.3 establishes that, for d≥3 and initial data in H^{σ+4}∩W^{σ,1}, ∥µϕ(t)−(e^{iµt/ε²}ϕℓ(t)+e^{−iµt/ε²}ϕℓ(t))∥_{H^σ} + ∥ψ(t)−ψℓ(t)∥_{H^σ} ≤ C(1+t)ε² for 0≤t≤T₁/ε. Under higher regularity, Theorem 1.4 gives the refined bounds ∥µϕ(t)−(… )∥_{H^σ} ≤ C(1+t+ε²t²)ε² and ∥ψ(t)−ψℓ(t)∥_{H^σ} ≤ C(1+εt+ε²t²)ε² on the longer interval 0≤t≤T₂/ε^{4/3}, and on the 1/ε interval the ψ error is uniformly","pith_inferences":["The proof's dependence on the L¹-based dispersive-decay estimate ties the optimal-rate statement to dimension d≥3; in d=1 or d=2 the same bootstrap would not close, so the optimality should not be assumed to carry over without additional hypotheses or a different mechanism.","A natural testable extension is to replace the free Schrödinger limit profiles by nonlinear Schrödinger profiles (adding the cubic term to ψℓ); the WKB construction suggests the same ε² barrier would appear, meaning the optimality argument would have to be revisited for a nonlinear limit system.","The explicit correctors ϕc and ψc identify the leading sub-ε² corrections; numerical experiments could directly check that the dominant error after subtracting the leading profiles is described by these correctors, providing a sharper verification than the H^σ norm bound alone.","The same WKB-plus-bootstrap framework, with a suitable phase and corrected profiles, likely yields analogous optimal rates for related systems such as Klein-Gordon-Zakharov or Maxwell-Klein-Gordon in their nonrelativistic limits, though each system would need its own phase analysis."],"forward_implications":["For initial data in the stated H^σ∩W^{σ,1} spaces in dimension d≥3, the nonrelativistic limit is rigorously justified on the natural 1/ε time scale, with the exact solution staying within O(ε²) of the two decoupled Schrödinger profiles.","Over the longer ε^{-4/3} time scale allowed by higher regularity, the Schrödinger component ψ attains a uniform O(ε²) bound on the 1/ε subinterval, matching the numerical observation that ψ is better approximated than the Klein-Gordon component ϕ.","Because the ε² order is fixed by the initial error and by the unavoidable presence of ε²-order WKB terms, no higher-order expansion can beat the O(ε²) rate; the result is optimal.","The specific error forms (1+t)ε² for µϕ and ε² for ψ explain the numerically observed asymmetry between the two components on the long time scale.","The reformulation as stability of WKB approximate solutions in nonlinear geometric optics provides a template for proving optimal nonrelativistic-limit rates in other singularly perturbed dispersive systems."],"fun_headline_variants":["KG-Schrödinger converges with optimal ε² error in nonrelativistic limit","Proof: ε² optimality for KG-Schrödinger, matches numerics on 1/ε scale","Nonrelativistic limit: KG-Schrödinger error is ε², uniform for ψ","Optimal ε² convergence proven for KG-Schrödinger, matching numerics"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The proof rests on the dispersive-decay bound for the limiting Schrödinger profiles, which requires the initial data to lie in a W^{σ,1}-based space and the spatial dimension to be at least three; if that decay is unavailable, the bootstrap interval and the ε² rate would not follow.","fun_headline_variants_meta":{"raw":{"variants":["KG-Schrödinger converges with optimal ε² error in nonrelativistic limit","Proof: ε² optimality for KG-Schrödinger, matches numerics on 1/ε scale","Nonrelativistic limit: KG-Schrödinger error is ε², uniform for ψ","Optimal ε² convergence proven for KG-Schrödinger, matching numerics"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000433,"raw_usage":{"total_tokens":2094,"prompt_tokens":846,"completion_tokens":1248,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":590,"completion_tokens_details":{"reasoning_tokens":1151}},"tokens_in":590,"tokens_out":1248,"duration_ms":10090,"temperature":1.0,"reasoning_tokens":1151,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T16:23:41.264744+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a numerical simulation of the Klein-Gordon–Schrödinger system in dimension d≥3 with initial data in H^{σ+4}∩W^{σ,1}, and measure ∥ψ(t)−ψℓ(t)∥_{H^σ} at t=T₁/ε; if the error is not bounded by C ε² or grows faster than (1+t)ε², the theorem's bound is false. Conversely, a simulation in d=1 or d=2 showing the same O(ε²) rate would indicate the dispersive-decay hypothesis is not necessary.","supporting_citations":[],"review_version":1}