{"id":"d6a11e0e-44e8-49ca-baa5-0f9e125c881d","arxiv_id":"2505.20008","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"At the critical point of a Glauber-Kawasaki reaction-diffusion process, the only slow observable is the total density, whose fluctuations converge to a cubic nonlinear SDE, while all other modes are Gaussian and fast.","lead":"This paper proves that at a special critical setting, a one-dimensional particle system slows down and its only slow observable is the global density, which follows a non-linear random equation. The result gives a rigorous microscopic explanation of critical slowdown in a reaction-diffusion model, with all other observables remaining fast and Gaussian.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"General-a branch of Theorems 2.3 and 2.7 rests entirely on Assumption B.5, a log-Sobolev inequality the text states is conjectured and currently unknown; only the small-a branch is an unconditional theorem.","rationale":"The reader's weakest-assumption analysis correctly identifies Assumption B.5 as the single load-bearing unproved input. Reading the manuscript in good faith, the small-a branch is a substantial unconditional theorem with a real proof, and the general-a branch is honestly presented in the main text as conditional. The concern is not circularity or a hidden inconsistency: it is that the abstract and the framing of the central claim describe the general-a results without the condition, while the proof of those results requires a log-Sobolev inequality that the authors state is conjectured and currently not known. The impact is precise: Theorem 2.10, the free-energy bound, and all estimates in Section 6 that use it depend on Assumption B.5; without it the non-perturbative control of the fast modes and the resulting convergence for all a > 0 is unsupported. I therefore agree with the reader's verdict and recommend no change. The only adjustment worth making is editorial: the abstract should either carry the small-a qualification or explicitly state that the general-a results are conditional on the unproved Assumption B.5.","tokens_in":79656,"tokens_out":6301,"duration_ms":82638,"concrete_test":"The decisive check is to prove Assumption B.5 for \\nu_{n,m}^g with the kernel g of Proposition 2.9 in the regime a \\ge a_0; no finite computation can fully replace a proof. A concrete first milestone is to estimate the nearest-neighbour spectral gap on fixed-magnetisation slices for n = 16, 32, 64, m = n/2, and a \\in \\{1,10,100\\}, using exact diagonalization for small n and Monte Carlo gap estimates for larger n, and test whether the gap is bounded below by c/n^2. If the gap is not \\Theta(n^{-2}), Assumption B.5 is false and the general-a branch fails. If the gap scales correctly, the Bernoulli-Laplace component still requires proof; a natural route is to extend the comparison in [3] and check that the extra log n factor is removable. Until one of these checks succeeds, Theorems 2.3(ii) and 2.7 should be stated as conditional.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Assumption B.5 is used exactly where the non-perturbative argument would otherwise fail: Lemma 6.10 and (6.84) apply the entropy inequality plus a slicewise log-Sobolev inequality to convert controls on W^{J,\\phi}_p into a multiple of \\delta n^2 \\Gamma, and Proposition B.7 uses the Bernoulli-Laplace log-Sobolev inequality to get the needed concentration. These estimates feed into Theorem 2.10, the free-energy bound, which is the only route to Theorem 2.3(ii) and Theorem 2.7 for a \\ge a_0. The text itself flags the gap: Section 6 says the log-Sobolev inequality is conjectured to be true and not known at the moment, and Remark B.6 notes that for g \\neq 0 only a modified Dirichlet-form version with an additional log n prefactor is currently known. That extra log n is not harmless: in (6.84) the entropy term must be absorbed into \\delta n^{5/2}\\Gamma with a parameter \\lambda constrained by the concentration scale; a log n prefactor would leave an unbounded remainder and destroy the Gronwall argument for the free energy. Thus the full 'all a' claim is not presently a theorem; it is a conditional statement pending a proof of B.5. This is not an internal inconsistency, but it is the load-bearing unproved input of the central claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a one-dimensional Glauber+Kawasaki reaction-diffusion process at the critical point γ = (1/2)(1 − θ/√n). With the magnetisation rescaled by n^{3/4} and time accelerated by √n, the authors claim that the magnetisation converges in L^p path space to the solution of the non-linear stochastic ODE dY_t = −2aθY_t dt − 2aY_t^3 dt + √a dW_t, and that the density fluctuation field projects onto this magnetisation. A second theorem asserts that mean-zero Gaussian-scaled density fluctuations converge to a time-white, space-coloured Gaussian field with explicit covariance. The proof develops a relative-entropy/free-energy method with tilted reference measures: first a magnetisation-tilted product measure ν_U for small a, then a two-point-tilted measure ν_g for all a. The all-a branch is explicitly made conditional on Assumption B.5, a conjectured log-Sobolev inequality for the tilted canonical measures.","tokens_in":79895,"tokens_out":9027,"duration_ms":98471,"significance":"If the small-a results are correct, they constitute a substantial rigorous advance: they give quantitative critical slowdown, identify the single slow observable, and derive non-Gaussian magnetisation fluctuations and explicit Gaussian fast-mode covariances for a short-range microscopic model. The proof strategy, especially the decoupling of slow and fast modes and the construction of the kernel g in Proposition 2.9, is ingenious and technically demanding. The paper is commendably honest in flagging the conjectural nature of its key log-Sobolev input. However, the advertised general-a results are not currently theorems: they rest on a conjecture that the text itself states is not known, and the known weaker form with an extra log n factor would break the main Gronwall/free-energy argument. The unconditional small-a branch remains a solid, publishable contribution.","major_comments":[{"comment":"The general-a branch of the paper is conditional on Assumption B.5, which is a conjectured log-Sobolev inequality for the tilted canonical measures ν_{n,m}^g. Section 6 states that this inequality is 'conjectured to be true' and 'not known at the moment', and Remark B.6 explains that for g ≠ 0 only a modified Dirichlet-form version with an additional log n prefactor is currently known. This is load-bearing: Lemma 6.10 and Eq. (6.84) use the slicewise log-Sobolev inequality to convert entropy controls into a multiple of δ n^{5/2} Γ, and the log n prefactor would leave an unbounded remainder, destroying the Gronwall derivation of Theorem 2.10. Consequently Theorems 2.3(ii) and 2.7 for all a > 0 are conditional statements, not established theorems. The authors should either prove Assumption B.5, or explicitly restructure the paper so that the abstract and introduction present the all-a statements as conditional on a conjecture and reserve the unconditional claims for a ≤ a0.","section":"Assumption B.5; Remark B.6; Theorems 2.3(ii), 2.10; Section 6, Eq. (6.84)"},{"comment":"Theorem 2.7 asserts convergence in finite-dimensional distributions of the fast-mode field, but the proof establishes Proposition 8.1 only for a single time t and for asymptotic independence from the initial σ-algebra F_0. The text asserts that linearity of y^n and the Markov property are enough to extend this to several times, but the required induction is not given. One needs to apply the Markov property at intermediate times and check that the estimates in Proposition 8.1 hold uniformly in the random initial law at those times. Please supply this induction or state Theorem 2.7 only for one-time marginals.","section":"Section 8, Proposition 8.1 and the paragraph following Eq. (8.1)"}],"minor_comments":[{"comment":"In the statement of Theorem 2.3(i), the expression 'YN_s(H)' appears to be a typo; it should be the fluctuation field, presumably Y^n_s(H), to match Eq. (7.3).","section":"Theorem 2.3(i)"},{"comment":"The scalar magnetisation and the density fluctuation field are both denoted by Y^n_t / Yn_t in Eq. (2.24), which is confusing; please use distinct notation for the process and the field throughout.","section":"Eq. (2.24)"},{"comment":"The results that depend on Assumption B.5 are labelled as theorems without the qualifier 'conditional' in their displayed statements; please mark them explicitly (e.g. 'Conditional Theorem') so that the dependence on a conjectured log-Sobolev inequality is visible without reading the proof.","section":"Theorem 2.10 and Theorems 2.3(ii), 2.7"},{"comment":"Remark B.6 states that a modified Dirichlet-form version of (B.25) is known for a range of a, but gives no theorem number or precise range; please add a precise reference and statement so the reader can verify what is actually known.","section":"Remark B.6"},{"comment":"The abstract says 'We prove' and the introduction describes the main results without prominently stating that the general-a versions are conditional on Assumption B.5; even if the main text is clear, the front matter should carry this caveat.","section":"Abstract and Introduction"}],"recommendation":"major_revision","confidential_remarks":"The paper's core small-a results appear sound and technically impressive. The all-a branch, however, depends on a log-Sobolev inequality that the authors themselves state is not known, so the manuscript's main advertised claim is only conditional. The editor may wish to ensure that the published abstract and theorem statements clearly separate conditional from unconditional results. The proof also leans heavily on the authors' previous work [3,6,10,11] for structural estimates; this is legitimate, but the novelty of the present contribution relative to those papers deserves scrutiny."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my honest take. The paper is a substantial rigorous-probability contribution, but the advertised generality is not yet earned. The small-a branch (Theorem 2.3(i), Theorem 2.7 for a <= a0) is an unconditional theorem and looks sound: it derives critical slowdown, projection onto the magnetization, convergence to the cubic SDE, and an explicit white-in-time Gaussian covariance for the fast modes. For a short-range model where local equilibrium fails, that is genuinely new. The technical machinery — tilted reference measures, free-energy estimates, canonical-measure concentration — is serious and the main-line arguments in Sections 4–5 and 7 are detailed and coherent.\n\nThe soft spot is exactly where the stress-test note lands. The general-a versions of Theorems 2.3 and 2.7 rest entirely on Assumption B.5, a log-Sobolev inequality for the tilted canonical measures that the text itself says is conjectured and not known. The note's point about the log n prefactor is correct: Remark B.6 says that for g != 0 only a modified Dirichlet-form version with an extra log n is currently known, and in the derivation of the free-energy estimate (6.84) that extra log n would destroy the Gronwall absorption. So the 'all a' claim is not a theorem; it is a conditional statement pending a proof of B.5. The main text is honest about this, but the abstract overstates by presenting the result without the condition, which is a real presentational flaw.\n\nThe construction is not circular: g is chosen to cancel a quadratic adjoint term, but the limiting SDE and covariance are derived rather than fitted. The self-citations are used where relevant (e.g., log-Sobolev results), not as padding.\n\nBottom line: the small-a contribution alone is publishable and valuable, and the conditional framework gives a credible route to the full result. A referee should demand that the conditional status be flagged clearly in the abstract and the conjecture isolated as a separate open problem. This deserves serious peer review.\n\nSend it to a serious referee, with the instruction that the general-a claims must be stated as conditional until B.5 is resolved.","headline":"A real small-a theorem plus a conditional general-a branch that depends on an unproved log-Sobolev inequality.","tokens_in":80478,"tokens_out":2017,"would_cite":true,"duration_ms":26084,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60K35","82C22","60F17","82C31"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that at the critical point of a one-dimensional reaction-diffusion model, the rescaled magnetisation converges to an explicit cubic stochastic differential equation while all other density fluctuations stay Gaussian with…","keywords":["critical slowdown","reaction-diffusion equation","Glauber-Kawasaki dynamics","magnetisation fluctuations","cubic stochastic differential equation","relative entropy method","log-Sobolev inequality","density fluctuation field"],"falsifier":"Estimate numerically the log-Sobolev constant of the fixed-magnetisation tilted measure $\\nu^{n,m}_g$ for the nearest-neighbour exchange Dirichlet form, say at $a=1$, $\\theta=0$ and $n$ up to a few hundred: if the constant grows faster than order $n^2$ for some $a>0$, the free-energy bound (Theorem 2.10) and with it the large-$a$ branch of Theorems 2.3 and 2.7 fail, settling the conjecture behind Assumption B.5. A second, direct check: simulate the rescaled magnetisation at $a > a_0$ and test the cubic drift $-2a\\theta y - 2ay^3$ with quartic-moment bounds; a systematic mismatch would pinpoint the same failure.","tokens_in":79365,"feed_emoji":"🧲","tokens_out":16188,"duration_ms":139048,"temperature":0.7,"pith_summary":"This paper studies a one-dimensional interacting particle system built from two dynamics — a particle-conserving exclusion process and a spin-flip (Glauber) reaction — tuned to sit exactly at a dynamical phase transition. Its central claim is that the critical slowdown of the dynamics at this point is caused by a single observable: the global density, or magnetisation, of the system. The paper proves that, on the critical time-scale and with the critical scaling $n^{3/4}$, the rescaled magnetisation converges to the solution of an explicit non-linear stochastic differential equation with cubic drift, so its fluctuations are non-Gaussian; simultaneously, every other observable equilibrates so quickly that the density field on mean-zero test functions converges to a Gaussian field, white in time and coloured in space, with a covariance the paper computes in closed form. A reader should care because this is a proof from the microscopic dynamics that the slow/fast decoupling assumed by critical-dynamics heuristics can hold in a short-range model, precisely where local equilibrium fails because the non-linearity at the transition is global, involving powers of the density itself.","feed_headline":"One scalar observable drives critical slowdown in reaction-diffusion","feed_subtitle":"Magnetisation fluctuations follow a cubic stochastic ODE; all other modes stay Gaussian with closed-form covariance.","key_machinery":"The argument is carried by a relative-entropy estimate in which the reference measure is chosen to encode the slow mode. The plain product measure fails: the magnetisation term $\\sqrt n\\int f (Y^n)^2\\,d\\nu^n_{1/2}$ cannot be absorbed by the energy (carré du champ), so the paper first tilts by the magnetisation, $\\nu^n_U \\propto e^{nU(m/n)}$, which cancels that term and yields the unconditional small-$a$ result. The second, more important object is the kernel $g = g_{\\gamma,a}$, the unique smooth solution of the boundary-value problem $$g''(x) - c_{\\delta,b}g(x) - \\tfrac14\\int_{\\mathbb{T}} g'(x-z)g'(z)\\,dz + \\tfrac{b}{2}\\int_{\\mathbb{T}} g(x-z)g(z)\\,dz = 0, \\quad g'(0+)-g'(1-) = -16\\delta b,$$ with Fourier coefficients $\\lambda^-_\\ell = [4\\pi^2\\ell^2 + 2b(1+2\\delta) - |4\\pi^2\\ell^2 - 2b(1-2\\delta)|]/(b+2\\pi^2\\ell^2)$. Tilting the product measure by the quadratic form $\\exp[\\tfrac{1}{2n}\\sum_{i\\ne j} g_{i,j}\\bar\\eta_i\\bar\\eta_j]$ makes the adjoint $L^*_n 1$ contain no leading-order two-point term — the boundary condition exactly cancels $16\\gamma a\\sqrt n\\sum_i\\bar\\eta_i\\bar\\eta_{i+1}$ — leaving only four-point and higher correlations, which are then controlled by energy estimates, concentration inequalities, and large-deviation bounds restricting the magnetisation. Fast modes are handled through log-Sobolev inequalities for the dynamics at fixed magnetisation.","core_discovery":"On the paper's own terms, the central discovery is Theorem 2.3: with the reaction coefficient $\\gamma = \\tfrac12(1-\\theta/\\sqrt n)$ and the generator accelerated by $\\sqrt n$, starting from any initial distribution whose relative entropy against the uniform measure is $O(\\sqrt n)$ and whose magnetisation converges weakly, the rescaled magnetisation $Y^n_t = n^{-3/4}\\sum_{i\\in\\mathbb{T}_n}(\\eta_i(t)-\\tfrac12)$ converges in $L^p$ path space, $1<p<\\tfrac43$, to the unique solution of $$dY_t = -2a\\$\\theta$ Y_t\\,dt - $2aY_t^{3}$\\,dt + \\sqrt a\\,dW_t,$$ and the density fluctuation field acts by projection: for every smooth $H$, the integral $\\int_0^t |Y^n_s(H) - \\langle H\\rangle Y^n_s|\\,ds$ vanishes in expectation as $n\\to\\infty$. Theorem 2.7 completes the picture: on mean-zero test functions the Gaussian-scaled field $n^{-1/2}\\sum_i \\bar\\eta_i(t)H(i/n)$ converges in finite-dimensional distributions to a Gaussian field with covariance $\\mathbf{1}_{t=s}\\{\\tfrac14(H,G) + \\tfrac{a}{2}(H,(-\\Delta)^{-1}G)\\}$, independent of the initial condition. The small-reaction regime $a\\le a_0$ is proven unconditionally; the full range of $a$ is conditional on a conjectured log-Sobolev inequality that the paper flags as not yet known.","pith_inferences":["If Assumption B.5 is eventually proved, the same slow/fast recipe — identify the conserved slow observables, tilt the reference measure to cancel two-point correlations, then prove log-Sobolev bounds on fixed-conserved-quantity slices — should apply to any conservative critical dynamics with finitely many slow modes.","The covariance relation $\\tfrac14 \\mathrm{id} + \\tfrac{a}{2}(-\\Delta)^{-1} = (4\\,\\mathrm{id} - g^0)^{-1}$ suggests that in dimension two or higher the fast-mode Gaussian field ceases to be a function because $(-\\Delta)^{-1}$ diverges, exactly the obstruction the paper identifies for generalising beyond one dimension.","A numerical estimation of the log-Sobolev constant of $\\nu^{n,m}_g$ at moderate $n$ across a range of $a$ would directly test the paper's conditional branch: quadratic growth in $n$ would confirm the conjecture, while faster growth would pinpoint where the free-energy estimate breaks."],"forward_implications":["At criticality the density field collapses onto one scalar: for every smooth $H$, $\\int_0^t |Y^n_s(H) - \\langle H\\rangle Y^n_s|\\,ds$ vanishes in expectation, so the magnetisation carries all slow fluctuation information.","Magnetisation fluctuations are non-Gaussian at scale $n^{3/4}$: they follow $dY_t = -2a\\theta Y_t\\,dt - 2aY_t^3\\,dt + \\sqrt a\\,dW_t$, the cubic drift arising from the degenerate quartic term in the reaction potential $V$ at $\\gamma = 1/2$.","All fast observables remain Gaussian: the covariance $\\mathbf{1}_{t=s}\\{\\tfrac14(H,G) + \\tfrac{a}{2}(H,(-\\Delta)^{-1}G)\\}$ is white in time and coloured in space, and the limiting field does not depend on the initial condition.","The slowdown is quantified: critical fluctuations evolve on the $\\sqrt n$ time-scale, and the special kernel $g$ with explicit Fourier coefficients makes the two-point correlations cancel so the slow/fast decoupling works for arbitrary $a$."],"supporting_citations":[{"why":"Supplies the quantitative relative-entropy method — entropy inequality, flow lemma, and sub-Gaussian estimates — that the paper adapts and sharpens at the critical point.","marker":"[33]"},{"why":"Establishes the hydrodynamic reaction-diffusion equation for this Glauber + Kawasaki model, from which the critical tuning and the cubic drift are read off.","marker":"[13]"},{"why":"Provides the entropy method for deriving closed macroscopic equations, the ancestor of the relative-entropy bounds used here.","marker":"[25]"},{"why":"Gives the log-Sobolev inequality for generalized simple exclusion processes, used to control the relaxation of fast modes at fixed magnetisation.","marker":"[51]"},{"why":"Provides the Lee-Yau log-Sobolev inequalities for uniform canonical (Bernoulli-Laplace) slices, the model that Assumption B.5 generalises.","marker":"[39]"},{"why":"Supplies the dynamical large-deviation principle for this reaction-diffusion model, used in Appendix A to restrict the magnetisation to its typical values.","marker":"[38]"},{"why":"Provides concentration and integration-by-parts estimates reused in the replacement lemmas and in the bounds on fast modes.","marker":"[31]"},{"why":"Gives the well-posedness theorem for the limiting non-linear SDE, so convergence to its law characterises the magnetisation limit.","marker":"[35]"}],"fun_headline_variants":["Critical slowdown driven by magnetisation alone in reaction-diffusion","Non-Gaussian magnetisation, Gaussian fast modes at criticality","One scalar controls critical fluctuations in reaction-diffusion","Magnetisation follows cubic SDE; field modes stay Gaussian","Critical dynamics: slow magnetisation, fast Gaussian field"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The results for every strength $a$ of the reaction term rest on Assumption B.5, a log-Sobolev inequality for the tilted fixed-magnetisation measures $\\nu^{n,m}_g$ with constant of order $n^2$, which the paper itself states is \"conjectured to be true\" and \"not known at the moment\"; only the small-$a$ regime is proven unconditionally.","fun_headline_variants_meta":{"raw":{"variants":["Critical slowdown driven by magnetisation alone in reaction-diffusion","Non-Gaussian magnetisation, Gaussian fast modes at criticality","One scalar controls critical fluctuations in reaction-diffusion","Magnetisation follows cubic SDE; field modes stay Gaussian","Critical dynamics: slow magnetisation, fast Gaussian field"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000544,"raw_usage":{"total_tokens":2670,"prompt_tokens":1077,"completion_tokens":1593,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":693,"completion_tokens_details":{"reasoning_tokens":1525}},"tokens_in":693,"tokens_out":1593,"duration_ms":11982,"temperature":1.0,"reasoning_tokens":1525,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T14:01:58.526324+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Estimate numerically the log-Sobolev constant of the fixed-magnetisation tilted measure $\\nu^{n,m}_g$ for the nearest-neighbour exchange Dirichlet form, say at $a=1$, $\\theta=0$ and $n$ up to a few hundred: if the constant grows faster than order $n^2$ for some $a>0$, the free-energy bound (Theorem 2.10) and with it the large-$a$ branch of Theorems 2.3 and 2.7 fail, settling the conjecture behind Assumption B.5. A second, direct check: simulate the rescaled magnetisation at $a > a_0$ and test the cubic drift $-2a\\theta y - 2ay^3$ with quartic-moment bounds; a systematic mismatch would pinpoint the same failure.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the quantitative relative-entropy method — entropy inequality, flow lemma, and sub-Gaussian estimates — that the paper adapts and sharpens at the critical point."},{"cited_title":"De Masi, P","cited_arxiv_id":null,"evidence_quote":"Establishes the hydrodynamic reaction-diffusion equation for this Glauber + Kawasaki model, from which the critical tuning and the cubic drift are read off."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the entropy method for deriving closed macroscopic equations, the ancestor of the relative-entropy bounds used here."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the log-Sobolev inequality for generalized simple exclusion processes, used to control the relaxation of fast modes at fixed magnetisation."},{"cited_title":"Landim, K","cited_arxiv_id":null,"evidence_quote":"Provides the Lee-Yau log-Sobolev inequalities for uniform canonical (Bernoulli-Laplace) slices, the model that Assumption B.5 generalises."},{"cited_title":"Landim, G","cited_arxiv_id":null,"evidence_quote":"Supplies the dynamical large-deviation principle for this reaction-diffusion model, used in Appendix A to restrict the magnetisation to its typical values."},{"cited_title":"Jacod, A","cited_arxiv_id":null,"evidence_quote":"Provides concentration and integration-by-parts estimates reused in the replacement lemmas and in the bounds on fast modes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the well-posedness theorem for the limiting non-linear SDE, so convergence to its law characterises the magnetisation limit."}],"review_version":1}