{"id":"d6e8554b-b7df-4fa9-a7ab-aa0b36ccdaec","arxiv_id":"2412.05950","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper derives explicit N^{-κ} error bounds for the mollified empirical measure of moderately interacting particles with common noise, and proves local strong well-posedness of the limiting stochastic Fokker-Planck equation.","lead":"An international group of mathematicians proves quantitative convergence rates for large systems of interacting particles with common noise and singular kernels to a nonlinear stochastic Fokker-Planck equation. The result covers classical models such as 2D Navier-Stokes, Keller-Segel chemotaxis, and Burgers turbulence, which matters for justifying particle simulations of stochastic PDEs.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central estimate (16) rests on an unjustified application of Krylov's Itô formula to the difference ρ − ρ^N; the empirical process has an idiosyncratic martingale term outside the cited SPDE framework.","rationale":"The reader's weakest_assumption identifies exactly the load-bearing point: equation (16) is the starting point of Theorem 2, and the empirical process ρ^N does not obviously belong to the solution class for which Krylov's Itô formula is stated. The paper's remaining estimates — commutator bounds, the uniform-in-N Lemma 4, the martingale bound Lemma 6, and the Grönwall argument — are coherent once (16) is granted, and I see no independent reason to think the theorem is false. The Keller-Segel (AK) citation and the ε in Corollary 1 are secondary; they do not affect the core argument. The correct outcome is therefore unchanged: conditional acceptance, with the required condition being a rigorous justification of the Itô formula for ρ − ρ^N, either by checking the hypotheses of [35] or by proving the formula directly through a spectral truncation argument.","tokens_in":29819,"tokens_out":15631,"duration_ms":164862,"concrete_test":"Re-derive (16) analytically for q = 2 in the simplified case ν = Id, σ ≡ 0, F = 0, K = 0, so that ρ solves (1/2)Δρ and ρ^N is the mollified empirical process of N independent Brownian motions. Use the standard Hilbert-space Itô formula directly on (13) and (15) to compute d∥ρ_t − ρ^N_t∥_2^2, keeping the idiosyncratic quadratic variation term. If the resulting identity does not match (16) specialized to q = 2 — same I2 and same sign of M^N_t — the cited Itô formula is not applicable. If it does match, repeat the check for q > 2 by verifying the hypotheses of [35, Thm 4.2] for δ, with particular attention to whether the idiosyncratic empirical martingale can be absorbed into the admissible class of noise coefficients.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 2 begins by 'Applying to (ρ − ρ^N) the Itô formula for the L^q-norm of a H^1_q-valued process in [35]' and then writes down (16). The difference process δ = ρ − ρ^N is not shown to fit the hypotheses of [35, Theorems 4.2/5.1]. The SPDE for ρ has a martingale term −∇ρ·σ dB, but the equation (13) for ρ^N contains the additional idiosyncratic martingale (1/N)Σ_i ∫ ∇V^N(x−X^i_s)·dW^i_s. Hence dδ has a martingale term that is not of the form σ^k(x)∇u dB^k nor a function of the state u = δ; it depends on the empirical particle positions X^i. The paper does not verify the hypotheses of Krylov's theorem for δ, nor does it provide an alternative Itô formula for q > d. Since (16) generates every subsequent estimate — the dissipation term, I1, I2, and the bound on M^N_t — an uncorrected extra term or a missing cross-variation term would break the Grönwall argument and the claimed N^{−κ} rate. This is a genuine gap at the beginning of Section 2.1, not merely a cosmetic issue.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript studies an N-particle system on the torus with moderately interacting singular kernels, idiosyncratic noise, and common noise. Its main result (Theorem 2) is a quantitative estimate in L^m(Omega) for the distance between the mollified empirical measure rho^N and the solution rho of a nonlinear stochastic Fokker-Planck equation with transport noise, of the form ||rho - rho^N||_{T,q} <= C ||rho_0 - rho_0^N||_q + C N^{-kappa}, with kappa = min(beta gamma/d, 1/2 - beta(1 + 1/d - 1/q)), under assumptions (AV), (Ac), (AF), (AK), with nu = Id and sigma spatially independent. Theorem 3 gives a one-dimensional Burgers-type variant with K = delta_0, and Corollaries 1 and 2 translate the result to the empirical measure and to cutoff approximations. The paper also proves local well-posedness of the limiting SPDE (Theorem 1) by a fixed-point argument in Krylov's L^q theory, and discusses applications to stochastic Navier-Stokes, Keller-Segel, and Burgers equations.","tokens_in":30061,"tokens_out":29364,"duration_ms":281805,"significance":"If correct, the paper is a valuable quantitative mean-field limit for singular kernels in the presence of common noise, with explicit polynomial rates and no fitted parameters. The common-noise setting is genuinely different from the earlier moderately interacting results on which the method builds, and the well-posedness theorem is a useful complement. The applications to Navier-Stokes, Keller-Segel, and Burgers equations are credible and give the results broader interest. The main caveat is that the central Itô formula application to the difference process is not verified in the manuscript; the significance is conditional on that step being supplied.","major_comments":[{"comment":"The proof of Theorem 2 invokes the Itô formula for the L^q norm of an H^1_q-valued process from [35] and applies it to delta = rho - rho^N, but the hypotheses of [35, Theorems 4.2 and 5.1] are not verified for delta. The equation for delta contains, in addition to the common-noise term -sigma^T nabla delta dB, the idiosyncratic martingale term (1/N) sum_i nabla V^N(· - X^i_s) dW^i_s, which depends on the particle positions and is not of the form sigma(x) nabla delta. If this term is intended to be treated as a free term g^k in Krylov's framework, the paper should state the corresponding version of the Itô formula and check the required predictability, integrability, and H^1_q regularity; otherwise Eq. (16) is not justified. Because Eq. (16) is the starting point for the dissipation estimate, the bounds on I1 and I2, and the estimate of M^N, this verification is load-bearing for the claimed N^{-kappa} rate. The same comment applies to Eq. (20) in Lemma 4, where the formula is applied to rho^N itself; for the smooth rho^N this can be justified by pointwise Itô calculus, but the manuscript should say so explicitly.","section":"Section 2.1, Eq. (16)"},{"comment":"Theorem 3 assumes K = delta_0 without (AK), so Theorem 1 does not provide the solution rho or the a.s. bounds used in the proof of Theorem 3. The proof implicitly relies on the Burgers equation well-posedness and maximum principle from [1], which is mentioned only in Section 1.5. The theorem statement should either include the external well-posedness/maximum-principle hypotheses explicitly or the paper should prove the needed regularity of rho, because the proof of Theorem 3 uses the existence of rho with sup_t ||rho_t||_2 and ||rho||_{L^2 H^1_2} bounded.","section":"Section 1.4, Theorem 3"}],"minor_comments":[{"comment":"The term C(N^{-beta gamma/d})^m E(||rho^N||_{T,q} ||rho^N||_{T,q})^m is malformed; if it is meant to be E||rho^N||_{T,q}^{2m}, it should be written that way, since this is the moment that requires the uniform bound from Lemma 4 for exponent 2m.","section":"Display before Eq. (17)"},{"comment":"The parameter epsilon in the exponent N^{-kappa + epsilon} is unexplained and appears unnecessary: Theorem 2 or 3 gives N^{-kappa}, and the deterministic regularization error N^{-beta/d} is no larger than N^{-kappa} because kappa <= beta gamma/d <= beta/d. Either remove epsilon or justify why a slightly worse rate is needed.","section":"Corollary 1"},{"comment":"The sign of the idiosyncratic stochastic integral in the displayed Itô formula is inconsistent with the definition of M^N_t: the formula writes -M^N_t, while M^N_t is defined as the positive stochastic integral. Since only |M^N_t| is used later, this sign inconsistency does not affect the estimates, but it should be corrected.","section":"Eq. (16) and definition of M^N_t"},{"comment":"In the statement of Corollary 2, the initial-error term should be ||~rho^N_0 - rho_0||_q (or ||~rho^N_0 - rho_0||_{0,q}), not ||~rho^N_0 - rho_0||_{T,q}, since no time interval is involved in the initial norm.","section":"Corollary 2"},{"comment":"The Young/Jensen step in the proof of Lemma 6 is abbreviated; in particular, the exponents for q > 2 and the separate q = 2 case should be spelled out, because the displayed estimates require the reader to reconstruct the cancellation of the factor 1/N.","section":"Appendix A, Lemma 6"}],"recommendation":"major_revision","confidential_remarks":"The main issue is exactly the one raised in the stress test: the Itô formula for the difference process is not verified. I believe this is repairable, since rho^N is smooth and the idiosyncratic term can be treated as a free source in Krylov's framework, but the verification must be written out in full. The other issues are local and presentation-level. I therefore recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThis paper extends quantitative moderately interacting particle approximation to SPDEs with common noise and singular kernels, covering Navier–Stokes, Burgers, and Keller–Segel. That combination is genuinely new: prior work by Olivera–Richard–Tomašević only had idiosyncratic noise, and common-noise results for singular kernels were limited. The well-posedness theorem for the SPDE with the nonlinear term K*ρ is also new and reasonably clean.\n\nThe proof follows the established semigroup/Krylov route: Itô formula for L^q norms of H^1_q processes, commutator estimates, and uniform bounds on the mollified empirical measure. The estimates are explicit and the rate is N^{-κ} with κ = min(βγ/d, 1/2 − β(1+1/d−1/q)). I believe the main theorem is likely correct, but the proof has a real gap that needs fixing before publication.\n\nThe gap is at the start of Section 2.1. Equation (16) applies an Itô formula to δ = ρ − ρ^N and writes down a formula that includes the idiosyncratic martingale term M^N_t and its quadratic variation. That is the right formula—the stress-test worry that those terms are missing is wrong, they are there. The actual problem is that δ does not satisfy the SPDE framework of Krylov [35] used for the citation: the drift and common-noise diffusion are functions of δ, but the W^i-term is an L^q-valued stochastic integral whose integrand depends on the particle positions X^i, not on δ. So [35, Theorems 4.2/5.1] does not directly apply, and the paper gives no alternative justification, e.g. a mollification/approximation argument. Since (16) underpins every later estimate, a referee should ask for this to be closed.\n\nTwo smaller things. The bound for M^N_t with δ in Theorem 2 is only sketched as “similar computations” to Lemma 6; it should be written out because the integrand differs (δ instead of ρ^N). And Corollary 1 states a rate N^{-κ+ε} but the proof appears to give N^{-κ} directly (since κ ≤ β/d, the N^{-β/d} from (5) is dominated); the ε is unexplained and probably unnecessary. Also, the Keller–Segel kernel bound is cited to an unpublished preprint [44]; the authors should verify it or include a proof.\n\nThese are addressable. The structure is sound, the constants are generic, and the paper is honest about what is borrowed. I would send it to a knowledgeable referee.","headline":"Solid extension of the moderate-interaction program to common noise; the main theorem is likely right but the Itô formula gap at (16) needs a proof.","tokens_in":30615,"tokens_out":6218,"would_cite":true,"duration_ms":61877,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["49N90","60H30","60K35"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper establishes an explicit $N^{-\\kappa}$ convergence rate for the mollified empirical measure of a moderately interacting particle system with common noise and singular kernels to the solution of a nonlinear stochastic…","keywords":["stochastic Fokker-Planck equation","mean-field limit","common noise","singular kernel","moderate interaction","empirical measure","propagation of chaos","Keller-Segel"],"falsifier":"Derive the Itô differential of $\\|\\rho_t-\\rho_t^N\\|_q^q$ directly from equation (13) and compare it with inequality (16); if any cross-variation term between the common and idiosyncratic Brownian motions appears, the displayed martingale $M_t^N$ is incomplete and the rate is not established. Alternatively, in the one-dimensional Burgers setting of Theorem 3, simulate the particle system and the SPDE solution for increasing $N$ and check that the empirical $L^2$ error decays with exponent at least $\\min(\\beta/2,\\frac12-\\frac32\\beta)$; a slower measured exponent would falsify the theorem.","tokens_in":29593,"feed_emoji":"🎲","tokens_out":11825,"duration_ms":103928,"temperature":0.7,"pith_summary":"This paper derives quantitative large-population limits for $N$ interacting diffusions on the torus, each driven by an independent Brownian motion and by one common environmental Brownian motion, with a singular interaction kernel. The main result, Theorem 2, states that the mollified empirical measure of the particle system converges in $L^m$ over the probability space, uniformly in time and in $L^q$ in space, to the unique solution of a nonlinear stochastic Fokker-Planck equation, with an explicit error of order $N^{-\\kappa}$, where $\\kappa=\\min(\\beta\\gamma/d,\\frac12-\\beta(1+1/d-1/q))$. If the paper is right, the mean-field limit is quantitative for rough kernels and common noise, covering the vorticity formulation of the 2D Navier-Stokes equation, the stochastic Burgers equation, and the parabolic-elliptic Keller-Segel model. The paper also proves that the limiting SPDE has a unique strong solution on a short time interval.","feed_headline":"Explicit particle rate for stochastic Fokker-Planck limits","feed_subtitle":"A quantitative mean-field limit for singular kernels and common noise, covering Navier-Stokes, Burgers, and Keller-Segel.","key_machinery":"The proof is carried by the Itô formula for the $L^q$ norm of an $H^1_q$-valued process taken from [35], applied to the difference $\\rho-\\rho^N$. That formula produces the dissipative term $-\\frac12 q(q-1)\\int|\\rho-\\rho^N|^{q-2}|\\nabla(\\rho-\\rho^N)|^2$, which absorbs the Lipschitz drift error, while the common-noise terms cancel because $\\sigma$ is spatially constant. The drift error is reduced by a commutator estimate bounding $\\langle S_s^N, V^N(x-\\cdot)(F(x,K*\\rho_s^N)-F(\\cdot,K*\\rho_s^N))\\rangle$ by $L(N^{-\\beta/d}+\\|\\rho_s^N\\|_q N^{-\\beta\\gamma/d})|\\rho_s^N(x)|$, using the Holder regularity of $K*\\rho^N$ and the support of $V$. The idiosyncratic noise contributes a martingale $M_t^N$ and a quadratic-variation term, both controlled by Burkholder-Davis-Gundy and the scaling identity $\\int|\\nabla V^N|^q\\,dx=N^{q\\beta(1+1/d)-\\beta}\\|\\nabla V\\|_q^q$; these two sources produce the two exponents inside $\\kappa$. A uniformity lemma bounds $\\sup_N\\|\\rho^N\\|_{T,q}$ in $L^m(\\Omega)$ and closes the Gronwall argument.","core_discovery":"The central object is the mollified empirical measure $\\rho_t^N=V^N*S_t^N$, where $S_t^N$ is the empirical measure of the particles and $V^N(x)=N^\\beta V(N^{\\beta/d}x)$ with $\\beta\\in(0,1)$ is the moderate-interaction mollifier. Theorem 2 asserts that, when the idiosyncratic diffusion coefficient is the identity and the common noise coefficient is spatially constant, the difference between the SPDE solution $\\rho$ and $\\rho^N$ satisfies $$\\big\\|\\|\\rho-\\rho^N\\|_{T,q}\\big\\|_{L^m(\\$\\Omega$)}\\le C\\big\\|\\|\\rho_0-\\rho_0^N\\|_q\\big\\|_{L^m(\\$\\Omega$)}+$CN^{{-\\kappa}}$,$$ with $\\kappa=\\min(\\beta\\gamma/d,\\frac12-\\beta(1+1/d-1/q))$, for every $m\\ge1$. Here $\\gamma$ is the Holder exponent supplied by the kernel assumption $\\|K*f\\|_{C^\\gamma}\\le C_K\\|f\\|_q$, and the parameter restrictions make both exponents positive. A one-dimensional Burgers case with $K=\\delta_0$ is treated separately in Theorem 3 with rate $\\min(\\beta/2,\\frac12-\\frac32\\beta)$. The same theorem, combined with a Kantorovich-Rubinstein estimate, gives propagation of chaos for the genuine empirical measure, and a cut-off argument extends the rate to particle systems without a truncated drift.","pith_inferences":["The two terms in $\\kappa$ have distinct origins: $N^{-\\beta\\gamma/d}$ is the cost of mollifying the empirical measure, while $N^{-(\\frac12-\\beta(1+1/d-1/q))}$ is the fluctuation cost of the idiosyncratic noise; equating them would select an optimal moderate-interaction exponent $\\beta$ for a given kernel, a choice the paper does not discuss.","The boundedness of $F$ in (AF) is likely removable: Corollary 2 shows the truncation level is controlled by $\\|K*\\rho\\|_{\\infty}$, so in models with a maximum principle, such as Burgers, the cut-off can be chosen deterministically and the rate should hold for the untruncated system.","Since the common Brownian motion drives both the particle system and the SPDE, environmental noise cancels from the leading error; this suggests the rate is driven by the idiosyncratic fluctuations, and one could look for conditional-on-common-noise or uniform-in-time versions under stronger regularity."],"forward_implications":["Theorem 2 turns the mean-field limit into an explicit algebraic rate for the mollified empirical measure in $L^m(\\Omega)$, uniformly in time, for singular kernels and common noise.","Corollary 1 transfers the rate to the un-mollified empirical measure in the Kantorovich-Rubinstein metric, which is a quantitative propagation of chaos for the marginals of the particle system.","The same estimates apply to the stochastic 2D Navier-Stokes vorticity equation, the stochastic Burgers equation, and the parabolic-elliptic Keller-Segel model in any dimension.","Corollary 2 shows the bounded-drift truncation can be removed: for the uncut particle system the same rate holds up to a probability bound, with the cut-off level chosen from $\\|K*\\rho\\|_{\\infty}$.","Theorem 1 supplies well-posedness of the limiting SPDE, so the particle approximation is approximating a well-defined unique strong solution."],"supporting_citations":[{"why":"Supplies the Itô formula for $L^q$ norms of $H^1_q$-valued processes and the $L^q$-theory of SPDEs used in the main error estimate and in Theorem 1.","marker":"[35]"},{"why":"Provides the quantitative particle-approximation framework for singular kernels, including the Corollary 2.3 cut-off argument used in Corollary 2.","marker":"[43]"},{"why":"Introduces moderately interacting particle systems and the scaling $V^N=N^\\beta V(N^{\\beta/d}\\cdot)$ on which the particle model is built.","marker":"[40]"},{"why":"Contributes the fixed-point strategy adapted in Appendix B to prove existence and uniqueness for the nonlinear limiting SPDE.","marker":"[31]"},{"why":"Supplies the kernel estimate $\\|K*f\\|_{C^{1-d/q}}\\le C\\|f\\|_{L^q}$ that verifies assumption (AK) for the Keller-Segel application.","marker":"[44]"},{"why":"Gives the maximum principle for the stochastic Burgers equation used to set the cut-off level in the Burgers application of Corollary 2.","marker":"[1]"}],"fun_headline_variants":["Quantitative rates for singular-kernel Fokker-Planck limits","Propagation of chaos with explicit speed for singular SPDEs","Moderate interaction: rates for nonlinear stochastic Fokker-Planck","Common noise and singular kernels: quantitative particle limits","Explicit mean-field rates for interacting particle systems"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof applies the Itô formula of [35] to the difference $\\rho_t-\\rho_t^N$, treating the mollified empirical process as if it satisfied the same SPDE hypotheses as $\\rho$; the paper does not verify this, and the extra idiosyncratic-martingale terms in equation (13) are exactly where the framework could fail.","fun_headline_variants_meta":{"raw":{"variants":["Quantitative rates for singular-kernel Fokker-Planck limits","Propagation of chaos with explicit speed for singular SPDEs","Moderate interaction: rates for nonlinear stochastic Fokker-Planck","Common noise and singular kernels: quantitative particle limits","Explicit mean-field rates for interacting particle systems"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000578,"raw_usage":{"total_tokens":2732,"prompt_tokens":962,"completion_tokens":1770,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":578,"completion_tokens_details":{"reasoning_tokens":1686}},"tokens_in":578,"tokens_out":1770,"duration_ms":13936,"temperature":1.0,"reasoning_tokens":1686,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T20:11:34.438187+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Derive the Itô differential of $\\|\\rho_t-\\rho_t^N\\|_q^q$ directly from equation (13) and compare it with inequality (16); if any cross-variation term between the common and idiosyncratic Brownian motions appears, the displayed martingale $M_t^N$ is incomplete and the rate is not established. Alternatively, in the one-dimensional Burgers setting of Theorem 3, simulate the particle system and the SPDE solution for increasing $N$ and check that the empirical $L^2$ error decays with exponent at least $\\min(\\beta/2,\\frac12-\\frac32\\beta)$; a slower measured exponent would falsify the theorem.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Itô formula for $L^q$ norms of $H^1_q$-valued processes and the $L^q$-theory of SPDEs used in the main error estimate and in Theorem 1."},{"cited_title":"Olivera, A","cited_arxiv_id":null,"evidence_quote":"Provides the quantitative particle-approximation framework for singular kernels, including the Corollary 2.3 cut-off argument used in Corollary 2."},{"cited_title":"Oelschl¨ ager, A martingale approach to the law of large numbers for weakly interacting stochastic processes , Ann","cited_arxiv_id":null,"evidence_quote":"Introduces moderately interacting particle systems and the scaling $V^N=N^\\beta V(N^{\\beta/d}\\cdot)$ on which the particle model is built."},{"cited_title":"Huang and J","cited_arxiv_id":null,"evidence_quote":"Contributes the fixed-point strategy adapted in Appendix B to prove existence and uniqueness for the nonlinear limiting SPDE."},{"cited_title":"Alonso-Oran, D., A","cited_arxiv_id":null,"evidence_quote":"Gives the maximum principle for the stochastic Burgers equation used to set the cut-off level in the Burgers application of Corollary 2."}],"review_version":1}