{"id":"6628d331-1796-49e3-9fae-9ee08f9b7668","arxiv_id":"2608.06533","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For smooth positive initial vorticity, the random batch vortex blob method converges to the regularized 2D Navier-Stokes solution with relative entropy error O(epsilon^-4 tau^2 + N^-1).","lead":"This paper proves a quantitative error bound for a fast particle algorithm that simulates the vorticity form of the two-dimensional Navier-Stokes equations. The bound degrades only polynomially when the smoothing blob shrinks, which is what makes the acceleration method practical.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The proof of Proposition 4.1 is the load-bearing point: ε-uniform W^{2,1}∩W^{2,∞}, Gaussian, score and log-Hessian bounds are asserted from [3,14] with only contractivity of φε; the algebraic ε^{-4}τ² rate in Theorem 1.1 collapses if these constants degrade as ε→0.","rationale":"The reader identified Proposition 4.1 as the weakest assumption, and I agree: the entire ε-uniformity of the mean-field fluctuation argument rests on it. My independent check of the surrounding argument found no flaws in the local coupling construction, the triangular-discrimination bound, the Fisher-information telescoping, or the cancellation structure of Φε in Lemma 4.3. The concrete weakness is precisely that Proposition 4.1 is not proved in the manuscript; it is imported from [3,14] with an argument that the mollifier contractivity transfers the estimates. This is a standard kind of regularity theorem, and I would not call it fraudulent or even likely false, which is why I do not recommend REJECT. But because the main theorem's novelty is the algebraic ε^{-4}τ² rate, and that rate is contingent on constants independent of ε, the paper should not be accepted unconditionally without a complete proof of Proposition 4.1. A CONDITIONAL acceptance, with the condition being a full derivation of the ε-uniform bounds (4.3)–(4.6), accurately reflects the state of the manuscript.","tokens_in":23483,"tokens_out":40072,"duration_ms":346733,"concrete_test":"Independently re-derive Proposition 4.1 for the mollified equation (4.1) by writing out the Feng–Wang/Ben-Artzi Sobolev argument with explicit constants: track every occurrence of ∥φε∗∇^mωε∥_{L1∩L∞} and ∥K∗(φε∗∇^mωε)∥_{L∞} for m=0,1,2, and verify that the Grönwall factors in the resulting W^{2,1}∩W^{2,∞} bound (4.6) can be bounded independently of ε. If any of these factors diverges as ε→0, or if the comparison functions in the Gaussian-bound step produce an ε-dependent exponent, then the algebraic ε-dependence of Theorem 1.1 is unsupported.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim, Theorem 1.1, splits into a random-batch error (Proposition 3.3) and a mean-field fluctuation (Lemma 4.3). The random-batch part is carefully handled with the local coupling and Fisher-information dissipation. The mean-field part, however, is only as strong as Proposition 4.1: it supplies the Gaussian upper bound, the score bound |∇logωε|≤C_T(1+|x|), and the log-Hessian bound |∇²logωε|≤C_T(1+|x|²), all uniformly in 0<ε≤1. Lemma 4.3 uses these bounds to make the Jabin–Wang Orlicz norm Λ_T independent of ε; if any of them develops a 1/ε factor, the mean-field estimate R_mf≤C_T(H_N+N^{-1}) fails and the ε^{-4}τ² algebraic rate in Theorem 1.1 is lost. The manuscript itself says: 'The only point that requires verification is that all constants remain uniform for 0<ε≤1.' What follows is a proof sketch: it invokes the finite-time Sobolev argument of Feng–Wang [14, Lemma 2.2] and Ben-Artzi [3], and argues that L1/L∞-contractivity of φε transfers the estimates. The transfer is plausible, but it is not fully demonstrated: the Sobolev argument controls ∇ωε, ∇²ωε, ∇uε, and ∇²uε through a coupled system, and the ε-uniformity of that system is precisely the claim being made. In particular, the replacement of K∗∇^mω by K∗(φε∗∇^mωε) requires the same Biot–Savart estimate ∥K∗f∥∞≤C(∥f∥1+∥f∥∞) at every level, and the paper does not show how the constants in [14] evolve under this replacement. As written, Proposition 4.1 is a stated regularity result with a proof-sketch rather than a self-contained proof, and it is the single most load-bearing assumption of the paper.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes the random batch vortex blob method for the two-dimensional Navier–Stokes equation in vorticity form. The dynamics are an N-particle system with the regularized Biot–Savart kernel and viscosity, and the batches are refreshed randomly at intervals of length tau. The main result, Theorem 1.1, is a finite-time normalized relative-entropy bound between the joint law of the random batch particles and the tensorized solution of the regularized vorticity equation, of order h_N^0 + epsilon^{-4} tau^2 (1 + h_N^0) + N^{-1}. The proof separates the random batch error, controlled through a locally coupled auxiliary partition and Fisher-information dissipation, from the mean field fluctuation, controlled through the oddness and divergence-free structure of the Biot–Savart kernel together with the Jabin–Wang exponential estimate. The paper also states L1 propagation of chaos (Corollary 1.2) and convergence to the Navier–Stokes vorticity solution in the combined vanishing-blob and mean-field limit (Corollary 1.3).","tokens_in":23856,"tokens_out":8005,"duration_ms":72716,"significance":"If the main theorem is correct, the paper makes a substantial contribution: it gives an algebraic dependence on the blob radius, epsilon^{-4}, instead of an exponential dependence exp(C_T epsilon^{-2}) that would follow from treating the mollified kernel as a generic Lipschitz kernel. The random-batch analysis in Section 3 is careful and self-contained, and the symmetrization and cancellation structure in Section 4 is genuinely used. The paper also makes good use of the existing literature, building on Jabin–Wang, Feng–Wang, and Ben-Artzi rather than redeveloping those tools. The main caveat is that the epsilon-uniform regularity statement, Proposition 4.1, is load-bearing and is presented as a proof sketch; this needs to be completed or precisely cited before the main theorem can be considered fully proven.","major_comments":[{"comment":"The uniform-in-epsilon bound W^{2,1} cap W^{2,infty} for omega_epsilon and u_epsilon is the load-bearing input to Lemma 4.3, because it makes the constants in (4.3)-(4.5) and hence the Orlicz norm Lambda_T in (4.21) independent of epsilon. The proof given is a sketch: it asserts that the finite-time Sobolev argument of [14, Lemma 2.2] and Ben-Artzi [3] applies with K * nabla^m omega replaced by K * (phi_epsilon * nabla^m omega_epsilon), and that L1/Linfty-contractivity of phi_epsilon transfers the estimates. This transfer is not demonstrated. In particular, the coupled system controlling nabla omega_epsilon, nabla^2 omega_epsilon, nabla u_epsilon, and nabla^2 u_epsilon involves constants that must be tracked under this replacement, and (4.6) itself is used later in the same proof to define U_T and to justify the quotient bounds. As written, Proposition 4.1 is a stated regularity result with a proof sketch. Because the algebraic epsilon^{-4} tau^2 rate in Theorem 1.1 collapses if any of these constants degrades as epsilon goes to 0, the proof must be completed or an explicit reference with the uniform-in-epsilon statement must be supplied.","section":"Section 4.1, Proposition 4.1, Eq. (4.6)"},{"comment":"The cancellation conditions in (4.16) are central to the application of the Jabin-Wang exponential estimate, since they are exactly the hypotheses of Lemma 4.2. The verification is dismissed as 'straightforward,' but it is delicate because, by oddness, integral K_epsilon(x-y) omega_epsilon(x) dx equals -(K_epsilon * omega_epsilon)(y), and the sign matters. The cancellation does work out, but the computation should be written out explicitly so the reader can verify it without rederiving the signs. This is a presentational point, but it concerns the core mechanism of the mean-field estimate.","section":"Section 4.2, Lemma 4.3, Eq. (4.16)"}],"minor_comments":[{"comment":"The abstract contains a rendering glitch: 'and $varepsilon$' should be 'and $\\varepsilon$'.","section":"Abstract"},{"comment":"The passage from (3.18) to (3.20) hides the dependence on p and on the initial-entropy bound; it would help to state explicitly that the final constant in (3.20) absorbs p, sigma, phi, T, and omega_0, so the reader does not have to reconstruct the absorption.","section":"Section 3.2, Eq. (3.18)-(3.20)"},{"comment":"The notation sup_y |Phi_epsilon(t, dot, y)| is used before the L^q norm is written; please define the normed quantity explicitly as a function of x for each fixed y.","section":"Section 4.2, Eq. (4.21)"},{"comment":"Reference [14] is cited as 'Peking Mathematical Journal, 2026' without volume or article number; please update the citation if a final version is available.","section":"References"},{"comment":"The statement says 'for every tau > 0,' but the meaningful regime is epsilon^{-2} tau -> 0; a short comment noting this would clarify the order of the limits in Corollary 1.3.","section":"Section 1.1, Theorem 1.1"}],"recommendation":"major_revision","confidential_remarks":"The main reservation is the completeness of the proof of Proposition 4.1. The rest of the paper is careful and the algebraic epsilon-dependence is a genuine advance. If the author can supply the missing epsilon-uniform regularity proof, or point to a precise statement in the literature with the same uniformity, the paper should be publishable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe thing to know: this paper gets the random-batch error right and separates it from the mean-field fluctuation in a way that avoids the exp(CT/ε²) disaster. The decomposition in Proposition 2.1, the local coupling in Lemma 3.1, and the triangular-discrimination/Fisher-information bound in Proposition 3.3 all check out on inspection. The symmetrization of the Biot–Savart fluctuation (4.12)–(4.13) and the cancellation (4.16) also work, and the use of the Jabin–Wang exponential estimate is legitimate. That is a solid piece of work, and the ε⁻⁴τ² + N⁻¹ rate is a genuine extension of the generic Lipschitz-kernel framework in the author's earlier paper [26].\n\nThe soft spot is exactly what the stress-test note says. Proposition 4.1 is load-bearing: it supplies the Gaussian, score, and log-Hessian bounds uniformly in 0<ε≤1, and Lemma 4.3 needs those to keep the Orlicz norm independent of ε. But Proposition 4.1 is a proof sketch. The text says 'the only point that requires verification is that all constants remain uniform,' and then asserts (4.6) by combining [3] and [14] with L1/L∞-contractivity of the mollifier. That is plausible, but the ε-uniformity of the coupled Sobolev estimate is not actually demonstrated. If the constants degrade, the algebraic rate collapses. This is a gap in rigor as written, not necessarily a false claim—these are standard regularity inputs—but it needs to be filled, either with a self-contained proof or a precise statement of the imported estimates with epsilon-uniformity verified.\n\nMinor point: the bound |K_ε(z)| ≤ C/(|z|+ε) in Lemma 4.3 is assumed rather than derived. With φ compactly supported it should follow from the homogeneity K(εz)=ε⁻¹K(z), but the derivation is absent. That one is easy.\n\nBottom line: this deserves a serious referee. I would push for revision rather than desk rejection, with the main request being to expand Proposition 4.1. For people working on RBM or quantitative propagation of chaos with singular kernels, this is a useful paper.","headline":"The random-batch error analysis is the real contribution and it is careful; the algebraic-ε claim rests on a sketched ε-uniform regularity bootstrap (Proposition 4.1) that should be filled in before the theorem is taken as fully rigorous.","tokens_in":24469,"tokens_out":2401,"would_cite":true,"duration_ms":20998,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65M75","76M23","35Q30","60K35","65C35"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that the random batch vortex blob method approximates the two-dimensional Navier-Stokes vorticity equation with an error that is algebraic, not exponential, in the blob radius.","keywords":["random batch method","vortex blob method","2D Navier-Stokes vorticity equation","Biot-Savart kernel","mean-field limit","propagation of chaos","relative entropy","Fisher information dissipation"],"falsifier":"Solve the regularized vorticity equation numerically for a Gaussian initial vorticity at several small blob radii and measure $\\sup_{0\\le t\\le T}\\|\\nabla^2\\log\\omega_\\varepsilon(t,\\cdot)\\|_{L^\\infty}$; if this quantity grows like $\\varepsilon^{-1}$ or worse as $\\varepsilon\\to 0$, the uniform regularity assumption fails and the algebraic $\\varepsilon^{-4}\\tau^2$ rate cannot hold. A direct check is the asserted uniform bound $\\|\\omega_\\varepsilon\\|_{W^{2,1}\\cap W^{2,\\infty}}\\le C_T$ for $0<\\varepsilon\\le 1$: an admissible initial datum for which this norm diverges as $\\varepsilon\\to 0$ would falsify Proposition 4.1 and with it the main theorem.","tokens_in":23208,"feed_emoji":"🌀","tokens_out":12693,"duration_ms":99945,"temperature":0.7,"pith_summary":"This paper establishes a quantitative mean-field error estimate for the random batch vortex blob method, a particle scheme for the two-dimensional Navier-Stokes equations in vorticity form. The method replaces the singular Biot-Savart kernel by a mollified blob kernel and randomly regroups particles at each time step, lowering the per-step cost from $O(N^2)$ to $O(N)$. The main theorem bounds the normalized relative entropy between the $N$-particle law and the tensorized regularized vorticity solution by $C_T(h_0^N+\\varepsilon^{-4}\\tau^2(1+h_0^N)+N^{-1})$, with constants independent of $N$, $\\tau$, and $\\varepsilon$. The central point is that the blob radius enters only algebraically, unlike the exponential dependence one gets from treating the mollified kernel as a generic Lipschitz force; this keeps the method practical and permits a vanishing-blob limit back to the Navier-Stokes solution. A sympathetic reader should care because the result turns a heuristic acceleration into a provable approximation with explicit rates.","feed_headline":"Random batch vortex method keeps blob error algebraic, not exponential","feed_subtitle":"It cuts particle cost to O(N) while keeping provable convergence to the 2D Navier-Stokes vorticity equation.","key_machinery":"The load-bearing device is the exact relative-entropy evolution identity $\\frac{d}{dt}H_N=-\\sigma I_N+R_{\\mathrm{RBM}}+R_{\\mathrm{mf}}$, where $I_N$ is the normalized Fisher information. For the random-batch term, the paper constructs, for each particle label, a locally coupled auxiliary batch partition obtained by exchanging two batches; conditional unbiasedness and exchangeability turn the batch error into a symmetric difference of two fixed-partition laws whose drift difference is supported on at most $2p$ coordinates, and Boltzmann entropy dissipation telescopes over time steps to yield $O(\\varepsilon^{-4}\\tau^2)$. For the mean-field term, the oddness of the mollified Biot-Savart kernel pairs it with the score difference $\\nabla\\log\\omega_\\varepsilon(x)-\\nabla\\log\\omega_\\varepsilon(y)$, and uniform Gaussian, score, and log-Hessian estimates for the regularized vorticity solution allow an exponential mean-field fluctuation estimate with a constant independent of $\\varepsilon$, giving $R_{\\mathrm{mf}}\\le C_T(H_N+N^{-1})$.","core_discovery":"The paper's central claim is that, for smooth strictly positive Gaussian-tailed initial vorticity, the random batch vortex blob particle system is quantitatively close at the level of laws to the solution of the regularized 2D Navier-Stokes vorticity equation, uniformly on finite time intervals. The normalized relative entropy $H_N(\\tilde F^N_\\varepsilon(t)|\\omega_\\varepsilon(t)^{\\otimes N})$ stays below $C_T(h_0^N+\\varepsilon^{-4}\\tau^2(1+h_0^N)+N^{-1})$. This is achieved by separating the two error sources: the random-batch error is controlled through a locally coupled auxiliary partition, a symmetric law comparison, and Fisher-information dissipation, while the mean-field fluctuation is controlled by exploiting the oddness and divergence-free structure of the Biot-Savart kernel together with score-difference cancellation. The estimate implies $L^1$ propagation of chaos for fixed-particle marginals and, after letting $\\varepsilon\\to 0$, convergence of the particle approximation to the true vorticity solution of the Navier-Stokes equations.","pith_inferences":["The same separation of a structural kernel estimate from a random-batch combinatorial estimate likely applies to other divergence-free antisymmetric interaction kernels, such as vortex stretching in three dimensions, where generic Lipschitz bounds would be prohibitive.","Because the $\\varepsilon^{-4}$ factor comes only from $\\|K_\\varepsilon\\|_{L^\\infty}^4$, mollifiers with additional vanishing moments may lower the exponent; this is a testable numerical improvement.","The $\\varepsilon$-uniform regularity of the regularized vorticity solution is likely the true boundary of the method: for initial data outside the Gaussian-tailed smooth class, one should expect either slower rates or the need for weighted relative entropies.","The proof's mean-field step uses only oddness, divergence-freeness, and the score-difference structure, so the extension to unequal circulations, which the paper defers to future work, should be achievable by the same mechanism."],"forward_implications":["For fixed $\\varepsilon>0$, as $N\\to\\infty$ and $\\varepsilon^{-2}\\tau\\to0$, every fixed-$k$ marginal converges in $L^1$ to $\\omega_\\varepsilon^{\\otimes k}$ at rate $(h_0^N)^{1/2}+\\varepsilon^{-2}\\tau(1+h_0^N)^{1/2}+N^{-1/2}$.","The time step needs only to be polynomially small in the blob radius ($\\tau\\ll\\varepsilon^2$), rather than exponentially small as a generic Lipschitz-kernel analysis would demand.","Passing to the vanishing-blob limit $\\varepsilon\\to0$ recovers the vorticity solution of the two-dimensional Navier-Stokes equation, so the particle system approximates the true fluid equations.","The bounds hold uniformly in $N$, $\\tau$, and $\\varepsilon$, so the estimate is quantitative for every finite system and does not hide constants that blow up as the blob shrinks.","The proof also controls the time-integrated normalized Fisher information, giving quantitative smoothness of the particle law."],"supporting_citations":[{"why":"introduces the random batch partition of interacting particles that the algorithm implements.","marker":"[31]"},{"why":"supplies the exponential law-of-large-numbers estimate used to control the mean-field fluctuation term.","marker":"[29]"},{"why":"provides the quantitative relative-entropy route for the whole-space viscous vortex model, including the Gaussian and logarithmic estimates adapted in Proposition 4.1.","marker":"[14]"},{"why":"is the generic random-batch relative-entropy bound whose Lipschitz-kernel constant would give an exponential blob dependence, the baseline this paper improves.","marker":"[26]"},{"why":"introduces the locally coupled auxiliary partition construction and combinatorial Fisher-information averaging used for the random-batch error.","marker":"[41]"},{"why":"supplies the global Sobolev regularity theory for the two-dimensional Navier-Stokes equation used to make the regularity constants uniform in the blob radius.","marker":"[3]"},{"why":"introduces the vortex blob regularization of the singular Biot-Savart kernel that the method builds on.","marker":"[10]"}],"fun_headline_variants":["Random batch vortex blobs: algebraic error, not exponential","2D Navier-Stokes: random batch method cuts cost, keeps error algebraic","Blob radius error stays algebraic in random batch vortex method","Propagation of chaos proved for random batch vortex blob method","O(N) cost with algebraic blob error for 2D vorticity equations"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole estimate leans on the blob-regularized vorticity staying well behaved—Gaussian decay and controlled slope and curvature—uniformly as the blob shrinks; if that uniformity fails, the algebraic error rate collapses.","fun_headline_variants_meta":{"raw":{"variants":["Random batch vortex blobs: algebraic error, not exponential","2D Navier-Stokes: random batch method cuts cost, keeps error algebraic","Blob radius error stays algebraic in random batch vortex method","Propagation of chaos proved for random batch vortex blob method","O(N) cost with algebraic blob error for 2D vorticity equations"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000195,"raw_usage":{"total_tokens":1396,"prompt_tokens":1025,"completion_tokens":371,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":641,"completion_tokens_details":{"reasoning_tokens":281}},"tokens_in":641,"tokens_out":371,"duration_ms":3754,"temperature":1.0,"reasoning_tokens":281,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T14:34:09.164826+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Solve the regularized vorticity equation numerically for a Gaussian initial vorticity at several small blob radii and measure $\\sup_{0\\le t\\le T}\\|\\nabla^2\\log\\omega_\\varepsilon(t,\\cdot)\\|_{L^\\infty}$; if this quantity grows like $\\varepsilon^{-1}$ or worse as $\\varepsilon\\to 0$, the uniform regularity assumption fails and the algebraic $\\varepsilon^{-4}\\tau^2$ rate cannot hold. A direct check is the asserted uniform bound $\\|\\omega_\\varepsilon\\|_{W^{2,1}\\cap W^{2,\\infty}}\\le C_T$ for $0<\\varepsilon\\le 1$: an admissible initial datum for which this norm diverges as $\\varepsilon\\to 0$ would falsify Proposition 4.1 and with it the main theorem.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"introduces the vortex blob regularization of the singular Biot-Savart kernel that the method builds on."},{"cited_title":"Random batch methods (RBM) for interacting particle systems.Journal of Computational Physics, 400:108877, 2020","cited_arxiv_id":null,"evidence_quote":"introduces the random batch partition of interacting particles that the algorithm implements."},{"cited_title":"Quantitative estimates of propagation of chaos for stochastic systems withW −1,∞ kernels.Inventiones Mathematicae, 214(1):523–591, 2018","cited_arxiv_id":null,"evidence_quote":"supplies the exponential law-of-large-numbers estimate used to control the mean-field fluctuation term."},{"cited_title":"Quantitative propagation of chaos for 2D viscous vortex model on the whole space.Peking Mathematical Journal, 2026","cited_arxiv_id":null,"evidence_quote":"provides the quantitative relative-entropy route for the whole-space viscous vortex model, including the Gaussian and logarithmic estimates adapted in Proposition 4.1."},{"cited_title":"Mean field error estimate of the random batch method for large interacting particle system.ESAIM: Mathematical Modelling and Numerical Analysis, 59(1):265–289, 2025","cited_arxiv_id":null,"evidence_quote":"is the generic random-batch relative-entropy bound whose Lipschitz-kernel constant would give an exponential blob dependence, the baseline this paper improves."},{"cited_title":"Propagation of chaos and approximation error of random batch particle system in the mean field regime","cited_arxiv_id":"2505.12172","evidence_quote":"introduces the locally coupled auxiliary partition construction and combinatorial Fisher-information averaging used for the random-batch error."},{"cited_title":"Global solutions of two-dimensional Navier-Stokes and Euler equations","cited_arxiv_id":null,"evidence_quote":"supplies the global Sobolev regularity theory for the two-dimensional Navier-Stokes equation used to make the regularity constants uniform in the blob radius."}],"review_version":2}