{"id":"3f7f643f-ba3b-4e3e-830b-775a06d8bc69","arxiv_id":"1908.01601","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For sufficiently small additive noise, a 2D droplet in the stochastic Cahn-Hilliard equation remains near the deterministic slow manifold for polynomial times, and its center satisfies a derived SDE with noise essentially the projection of the Wiener process onto the translational modes.","lead":"This paper derives the random motion of a single droplet in the stochastic Cahn-Hilliard equation in two dimensions, when the noise is very small. It proves the droplet's center follows a stochastic differential equation driven by the noise, and that the droplet stays close to the deterministic slow manifold for long times with high probability.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Stability depends on the imported order-epsilon spectral gap lambda_3 >= C' epsilon (Thm 2.4); if the true d=2 gap were O(epsilon^2) as in Remark 2.6 for d=3, the damping a in (4.10) and the noise restrictions in Thms 4.7/4.12 would change.","rationale":"The reader's conditional verdict is appropriate. The main theorems 4.7 and 4.12 are conditional on an externally cited spectral gap, and this is the least secure link: the a = O(epsilon) term in (4.10) is exactly the product of the order-epsilon gap, and the paper itself notes that the gap changes in d=3. I found no fatal internal contradiction in the SDE derivation or in the stability estimates; the interpolation exponents in Lemma 4.2 check out, and the moment-estimate strategy from [7] is reasonable. I did notice a likely typo in Theorem 4.7's initial condition, which should be corrected to ||v(0)|| <= nu epsilon^m. Since the reader's weakest_assumption identifies the same load-bearing concern and the verdict is already CONDITIONAL, I recommend UNCHANGED.","tokens_in":23108,"tokens_out":12592,"duration_ms":115294,"concrete_test":"Independently re-derive the lower bound lambda_3 >= C' epsilon in Theorem 2.4 from [2,3], tracking the epsilon-dependence and the uniformity in xi in Omega_{rho+delta}; if the third eigenvalue of L_xi scales only as O(epsilon^2), recompute q/B^2 in the proof of Theorem 4.7 with a = O(epsilon^2) and check whether the noise condition must become eta_0 <= C epsilon^{2m+2+kappa}.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central stability estimate in Lemma 4.2 uses Theorem 2.4 to get, for v orthogonal to the two tangential eigenmodes, <L_xi v,v>_{H^{-1}} <= -c epsilon ||v||^2_{H^{-1}} (eq. 4.7), which is the a = O(epsilon) damping in (4.10). Theorem 2.4 is quoted from [2,3] and is not proved here; Remark 2.6 explicitly warns that in d=3 the gap is only O(epsilon^2). If the d=2 gap were O(epsilon^2), Lemma 4.2 would give a = O(epsilon^2), and the condition eta_0 <= C epsilon^{2m+1+kappa} in Theorem 4.7 (and the analogous eta_2 condition in Theorem 4.12) would need to be replaced by eta_0 <= C epsilon^{2m+2+kappa}, weakening the headline claim. The SDE derivation in Section 3 is algebraically exact and does not depend on this gap, but the long-time stability that justifies the SDE as the effective motion does. A separate, minor internal inconsistency: Theorem 4.7 states ||v(0)|| <= nu epsilon^4 instead of nu epsilon^m; read literally this makes the m>4 statement vacuous, though it is almost certainly a typo.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the stochastic Cahn-Hilliard equation in a bounded two-dimensional domain with small additive, spatially smooth noise. It builds on the deterministic slow manifold of single-droplet states constructed in earlier work, derives by an explicit Itô computation an exact SDE for the droplet center, and then proves long-time stochastic stability of the slow manifold in both H^{-1} and L^2 for polynomially long times under small-noise conditions. The central quantitative claim is that, starting within O(ε^m) of the manifold and with noise intensities η0 ≤ C ε^{2m+1+κ} and η2 ≤ C ε^{2k+4+κ}, the probability that the solution exits an ε^m-neighborhood in H^{-1} (and an ε^{k+1}-neighborhood in L^2) before time ε^{-N} is smaller than any power of ε; on the manifold the droplet center approximately satisfies the SDE dξ = f(ξ)dt + σ(ξ)dW with σ obtained by projecting the Wiener process onto the translational modes.","tokens_in":23427,"tokens_out":20780,"duration_ms":206164,"significance":"If the stability statements are correct, this is a valuable rigorous contribution: Section 3 gives a parameter-free derivation of the effective SDE, and Section 4 extends the stability framework of [7] to the stochastic Cahn-Hilliard equation with droplet motion. The explicit Itô computation and the careful identification of the noise projection are genuine strengths. However, the proof of the stability theorems has a quantitative gap concerning the relation between the initial condition and the small quantity q defined in (4.13); until that gap is closed, the headline super-polynomial exit-probability claims are not established as stated.","major_comments":[{"comment":"The moment bound (4.14) is derived under the standing assumption (4.13) that ||v(0)||² ≤ q, where q = (C_ε + ||Q||)/a. With the noise condition η0 ≤ C ε^{2m+1+κ} and the exit radius B = ε^m, one has q = O(ε^{2m+κ}). The theorems instead assume ||v(0)|| ≤ ν ε^m (or, in Theorem 4.5, ||v(0)|| ≤ ν ε^4), which gives ||v(0)||² of order ε^{2m}, much larger than q for small ε. Thus the induction leading to (4.14) does not close: the initial-data term ||v(0)||^{2p} cannot be absorbed into the q^p factor, and the claimed 'smaller than any power of ε' bound does not follow from the displayed estimates. The authors should either strengthen the initial-condition hypothesis to ||v(0)|| ≤ C ε^{m+κ/2} (so that ||v(0)||² ≤ q) or give a different argument controlling the ||v(0)||^{2p} terms.","section":"§4.1, Lemma 4.2, Eq. (4.7)"},{"comment":"The step 'absorbed the positive L2-term into its negative counterpart' is not justified as written. With γ3 ≈ ε², the positive term is C ε² ||v||²_{L2}; it can be absorbed only if the coefficient γ2 in the negative ε²-term is uniformly bounded away from zero and the constants are chosen compatibly with γ1 + γ2 + γ3 = 1. This is plausible, but the choice of γ1 and γ2 and the resulting constants should be displayed. Because this inequality provides the damping rate a = O(ε) used throughout Section 4, the argument should be completed explicitly.","section":"§2.3, Theorem 2.4(ii) and Remark 2.6"},{"comment":"The stability analysis relies on the order-ε spectral gap λ3 ≥ C'ε in d = 2, but this fact is imported from [2,3] and is not proved here. Remark 2.6 states that in d = 3 the gap is only O(ε²), so the ε-gap is a special d = 2 fact and is load-bearing: if the d = 2 gap were ε², the damping a in (4.10) would become O(ε²) and the noise restrictions in Theorems 4.5 and 4.12 would change by powers of ε. Please cite the precise theorem of [2,3] and, ideally, give a short proof sketch or at least an explicit statement of how the ε-gap enters the estimates.","section":"§4.2, Theorems 4.5 and 4.7"}],"minor_comments":[{"comment":"The hypothesis reads ||v(0)||_{H^{-1}} ≤ ν ε^4, which must be a typo for ||v(0)||_{H^{-1}} ≤ ν ε^m; as printed, the m > 4 statement is not what is used in the proof.","section":"Lemma 4.11 and Theorem 4.12"},{"comment":"Lemma 4.11 assumes k ≥ 2, while Theorem 4.12 allows k ∈ (0, m−4); the proof of Theorem 4.12 invokes Lemma 4.11, so the range should be restricted to k ≥ 2 (hence m > 6) or Lemma 4.11 should be extended.","section":"§5, Lemma 5.2"},{"comment":"The proof refers to 'Definition 2.9', but no Definition 2.9 appears in the paper; the reference should be to Theorem 2.4 or Remark 2.8.","section":"Eq. (3.18)"},{"comment":"The remainder O(1) in (3.18) should specify that it is an H^{-1}-norm bound; otherwise the order-of-magnitude statement is ambiguous.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The SDE derivation in Section 3 is a clear strength and is mostly self-contained. The main risk is the mismatch between the initial-condition hypotheses of Theorems 4.5/4.7 and assumption (4.13) in the moment estimate; this affects the central stability claim and needs to be corrected. The paper also depends heavily on [2,3,7,8], including works sharing an author; the spectral-gap fact in Theorem 2.4 should be checked carefully against [2,3] during revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read it if you work on stochastic interface motion. The main result is honest and mostly checks out, but the paper is less general than its title.\n\nWhat is new: for a single droplet in 2D stochastic Cahn-Hilliard, the authors split the solution into slow-manifold coordinate xi and orthogonal part v, and derive an exact SDE for xi. The formulas for drift and diffusion, (3.11)-(3.12), come out of a clean Ito calculation, and Lemma 3.2 verifies that the ansatz reproduces the SPDE. That is genuinely new: previous stochastic droplet results were 1D Cahn-Hilliard or mass-conserving Allen-Cahn. The diffusion is approximately the projection of the Wiener process onto translations, which is the physically right answer. Section 4's stability estimates follow a known framework, but they adapt it carefully and give explicit noise restrictions.\n\nWhat worries me. The load-bearing spectral input is Theorem 2.4 from Alikakos-Fusco: in 2D the third eigenvalue of the linearized Cahn-Hilliard operator is bounded below by C' epsilon. That is not proved here. If the true 2D gap were only O(epsilon^2), the damping coefficient in (4.10) and the noise restrictions in Theorems 4.7 and 4.12 would both change. This is a conditional risk, not a demonstrated error, but the paper should at least give a proof or a citable statement with the 2D computation. Remark 2.6 concedes the 3D gap is only O(epsilon^2), which is exactly why the title's 'higher space dimensions' is misleading: the paper treats d=2 only, and d=3 is deferred to a thesis.\n\nSecond, Lemma 4.2 has an 'absorbed' step in (4.7) where a positive L2 term is absorbed into negative terms by choosing gamma weights. I cannot verify the constant balance from what is written. This is fixable but it is the kind of spot a referee needs to see nailed down. Third, the L2 moment estimates in Section 4.4 are imported from [7] with details skimmed; acceptable for a paper in this tradition, but it makes the result hard to check independently. Minor: Theorem 4.7 states ||v(0)|| <= nu epsilon^4 instead of epsilon^m, a typo that makes the m>4 statement vacuous if read literally.\n\nCitation pattern is fine: [7] and [8] share authors and supply the general method, but the 2D Cahn-Hilliard SDE derivation and stability result are not in those papers. The paper does not assume its conclusion.\n\nVerdict: worth a serious referee. I would send it out, asking the referee to verify Lemma 4.2 and to pressure-test the imported epsilon-gap. In current form I would not bet on the stability theorem without checking those steps, but I would cite it after revision.","headline":"Credible 2D stochastic Cahn-Hilliard droplet result with an explicit SDE and long-time stability; the title oversells 'higher dimensions' and the stability rests on an imported spectral gap that deserves scrutiny.","tokens_in":23960,"tokens_out":2848,"would_cite":true,"duration_ms":29461,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35R60","60H15","37L25","35K55","35B25"],"pacs":[],"model":"deepseek-v4-flash","headline":"For sufficiently small noise, a single droplet in the two-dimensional stochastic Cahn-Hilliard equation stays close to its slow manifold for polynomial times, and its center solves an explicit stochastic differential equation.","keywords":["stochastic Cahn-Hilliard equation","slow manifold","droplet motion","spectral gap","stochastic stability","additive noise","Fermi coordinates","sharp interface limit"],"falsifier":"Compute the third eigenvalue λ3(ε) of the linearized Cahn-Hilliard operator around a centered droplet on a two-dimensional disk with Neumann boundary conditions; if λ3(ε)/ε → 0 as ε→0, the order-ε spectral gap assumed by Theorem 2.4 fails and the exit-probability bounds of Theorem 4.12 lose their stated noise exponents.","tokens_in":22910,"feed_emoji":"🔵","tokens_out":12513,"duration_ms":127711,"temperature":0.7,"pith_summary":"This paper aims to prove a stability-and-motion statement for the last stage of phase separation in the stochastic Cahn-Hilliard equation on a two-dimensional domain. If a solution starts near a single-droplet state, and the additive noise is small in the precise sense of the paper's assumptions, then the solution remains near the family of droplet states for times of order $ε^{{-N}}$, with probability higher than any fixed power of ε. While it stays near that family, the droplet's center is shown to move according to an explicit stochastic differential equation, whose noise term is, to leading order, just the projection of the driving noise onto the two translation directions of the droplet. The upshot is a rigorous random-walk description of droplet motion, replacing the exponentially slow deterministic motion of the noiseless equation.","feed_headline":"Bubbles wander: noisy Cahn-Hilliard droplets obey an explicit SDE","feed_subtitle":"Under small noise a single bubble keeps its shape for polynomial times and its center diffuses like a random walk.","key_machinery":"The construction rests on a two-dimensional slow manifold M̃_ε^ρ of droplet profiles ũ_ξ, each an exponentially small correction of a translated radial bubble, chosen so the deterministic residual lies in the tangent space. The spectral fact doing the heavy lifting is Theorem 2.4: around every droplet, the linearized Cahn-Hilliard operator has two eigenvalues exponentially close to zero, with eigenvectors almost tangent to translations, and a third eigenvalue bounded below by C'ε. That order-ε gap produces the damping coefficient a=O(ε) in the $H^{{-1}}$ inequality (4.9)–(4.10), which in turn fixes the admissible noise sizes. Fermi coordinates (ξ,v), with v orthogonal to the two critical eigenvectors, allow an exact stochastic projection: differentiating the constraint ⟨v,ψ^ξ_k⟩=0 and the equation itself yields the SDE for ξ, with the invertible matrix A_{kj}=⟨ψ^ξ_k, ũ^ξ_j⟩-⟨v,ψ^ξ_{k,j}⟩ governing the projection. The $L^{2}$ stability argument uses the mass-conserving Allen-Cahn operator, whose spectral gap is ε², and interpolation to control the nonlinearity.","core_discovery":"The central claim is that, in two dimensions, one-droplet solutions of the stochastic Cahn-Hilliard equation are stable against small additive noise for polynomial times, and the droplet center obeys an explicit SDE. More precisely, for m>4 and 0<k<m-4, if the initial distance from the slow manifold satisfies ‖v(0)‖_{$H^{{-1}}$} ≤ ν ε^m and ‖v(0)‖_{$L^{2}$} ≤ ν $ε^{{k+1}}$, and the noise intensities satisfy η_0 ≤ C $ε^{{2m+1+κ}}$ and η_2 ≤ C $ε^{{2k+4+κ}}$, then the probability of leaving the ε^m ($H^{{-1}}$) and $ε^{{k+1}}$ ($L^{2}$) tube before time $ε^{{-N}}$ is smaller than any power of ε. On the manifold, the center solves dξ = f(ξ)dt + σ(ξ)dW with f and σ given by (3.11)–(3.12); Lemma 3.4 shows σ is asymptotically the normalized tangent vector, so the noise enters only through the direction in which the droplet is free to move. Up to exponentially small corrections, the stochastic motion is exactly the projection of the driving noise onto the slow manifold.","pith_inferences":["The paper does not simulate the SDE, but a direct test is mean-square displacement: on times short compared with ε^{-N}, the droplet center should satisfy roughly E|ξ(t)-ξ(0)|² ≈ C η0 t, so tracking many realizations would test the projection formula for σ and estimate the noise amplitude.","Because the three-dimensional spectral gap is only ε², the same stability argument would have a weaker damping rate; quantifying how the stability radius and noise thresholds must shrink in three dimensions is a concrete open problem.","The Fermi-coordinate projection is not tied to the specific nonlinearity F(u)=¼(u²-1)², so the same derivation should produce SDEs for droplet motion in mass-conserving Allen-Cahn models or for droplets sliding along the boundary, where a slow manifold of the same type exists."],"forward_implications":["For any fixed large N, if the noise bounds hold, the exit probability from the ε^m tube before time ε^{-N} is smaller than every power of ε.","Close to the slow manifold, the center motion is dξ = f(ξ)dt + σ(ξ)dW with explicit f and σ, and σ is, up to O(1), the normalized translational mode, so the noise enters through the direction in which the droplet is free to move.","The H^{-1} and L^2 stability radii are coupled to two different noise measures, η0 and η2, so both the total noise strength and its spatial smoothness must be small.","The stability window covers times ε^{-N} for any N, long enough that the stochastic motion of the center, rather than noise-induced nucleation or shape destruction, is what happens before a boundary collision.","The theorem is two-dimensional; in three dimensions the order-ε spectral gap is replaced by ε², so the same proof would not go through unchanged."],"supporting_citations":[{"why":"supplies the spectral estimates for the mass-conserving Allen-Cahn operator used in Theorem 2.7 for the L^2 stability part.","marker":"[1]"},{"why":"supplies the eigenvalue gap theorem, Theorem 2.4, for the linearized Cahn-Hilliard operator around a droplet, which is the core damping fact.","marker":"[2]"},{"why":"constructs the slow manifold of droplet states and the residual estimates on which the Fermi-coordinate setup is built.","marker":"[3]"},{"why":"derives the deterministic droplet motion and the norm estimates used to bound tangent vectors and their derivatives in Section 5.","marker":"[6]"},{"why":"provides the martingale stopping-time argument that converts the differential inequality into the super-polynomial exit-probability bounds.","marker":"[7]"},{"why":"introduces the exact SDE derivation for a droplet center in the one-dimensional stochastic Cahn-Hilliard equation, adapted here to two dimensions.","marker":"[8]"},{"why":"fixes the representation of the noise as a series and the stochastic calculus conventions used throughout.","marker":"[20]"},{"why":"gives the abstract eigenvalue perturbation lemma used to identify the slow eigenvectors with the tangent space of the manifold.","marker":"[21]"}],"fun_headline_variants":["Noisy bubbles follow explicit SDE for droplet center","Droplet center obeys SDE under stochastic Cahn-Hilliard","Stochastic bubbles: explicit SDE for motion of center","Noise-driven droplet motion: explicit SDE derived","Bubble motion under noise: explicit SDE for center"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole proof depends on an imported spectral estimate: around each droplet, the linearized operator's first two eigenvalues are exponentially small and the third is at least of order ε, uniformly in the center; if the true gap were only ε², as the paper notes in three dimensions, the damping coefficient in the stability estimate would disappear and the stated noise bounds would be too large.","fun_headline_variants_meta":{"raw":{"variants":["Noisy bubbles follow explicit SDE for droplet center","Droplet center obeys SDE under stochastic Cahn-Hilliard","Stochastic bubbles: explicit SDE for motion of center","Noise-driven droplet motion: explicit SDE derived","Bubble motion under noise: explicit SDE for center"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00073,"raw_usage":{"total_tokens":3227,"prompt_tokens":862,"completion_tokens":2365,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":478,"completion_tokens_details":{"reasoning_tokens":2282}},"tokens_in":478,"tokens_out":2365,"duration_ms":16943,"temperature":1.0,"reasoning_tokens":2282,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:10:23.848825+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the third eigenvalue λ3(ε) of the linearized Cahn-Hilliard operator around a centered droplet on a two-dimensional disk with Neumann boundary conditions; if λ3(ε)/ε → 0 as ε→0, the order-ε spectral gap assumed by Theorem 2.4 fails and the exit-probability bounds of Theorem 4.12 lose their stated noise exponents.","supporting_citations":[{"cited_title":"Alikakos, L","cited_arxiv_id":null,"evidence_quote":"supplies the spectral estimates for the mass-conserving Allen-Cahn operator used in Theorem 2.7 for the L^2 stability part."},{"cited_title":"Alikakos and G","cited_arxiv_id":null,"evidence_quote":"supplies the eigenvalue gap theorem, Theorem 2.4, for the linearized Cahn-Hilliard operator around a droplet, which is the core damping fact."},{"cited_title":"Alikakos and G","cited_arxiv_id":null,"evidence_quote":"constructs the slow manifold of droplet states and the residual estimates on which the Fermi-coordinate setup is built."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"derives the deterministic droplet motion and the norm estimates used to bound tangent vectors and their derivatives in Section 5."},{"cited_title":"Antonopoulou, P","cited_arxiv_id":null,"evidence_quote":"provides the martingale stopping-time argument that converts the differential inequality into the super-polynomial exit-probability bounds."},{"cited_title":"Antonopoulou, D","cited_arxiv_id":null,"evidence_quote":"introduces the exact SDE derivation for a droplet center in the one-dimensional stochastic Cahn-Hilliard equation, adapted here to two dimensions."},{"cited_title":"Da Prato and J","cited_arxiv_id":null,"evidence_quote":"fixes the representation of the noise as a series and the stochastic calculus conventions used throughout."},{"cited_title":"Hellfer and S","cited_arxiv_id":null,"evidence_quote":"gives the abstract eigenvalue perturbation lemma used to identify the slow eigenvectors with the tangent space of the manifold."}],"review_version":1}