{"id":"6b7ee1be-cd5d-4b0b-9621-426a7444be63","arxiv_id":"2509.01921","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Under controllability and noise-growth conditions, the KdVB equation with localized or multiplicative white noise has a unique invariant measure, with exponential mixing in the localized case.","lead":"Proves exponential ergodicity for the Korteweg-de Vries-Burgers equation driven by space-time localized noise, and ergodicity with asymptotic stability when the noise is multiplicative. Introduces a new Carleman estimate for the linear KdVB operator, the main new tool that makes the controllability-based coupling argument work for this equation.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Verification of Hypothesis (LS) in Theorem 2.1 is unsupported: Prop. 4.2 only proves contraction relative to the deterministic solution, not for two paths driven by the same noise η as required by abstract coupling.","rationale":"The paper's central contribution is the proof of exponential ergodicity via coupling. The most load-bearing point is the correct verification of the abstract coupling hypotheses, not the scope of Condition (AC). The reader identified (AC) as the weakest premise, but (AC) is an explicit hypothesis and the paper provides a nontrivial example (Prop. 4.3). A more serious internal issue is the passage from Prop. 4.2 to Hypothesis (LS): the contraction property is stated for the deterministic reference, while the abstract theorem requires contraction for two paths driven by the same random noise. Because the noise cancels in the difference, the correct reference should depend on η, and the manuscript neither states nor proves that the control design works in that setting. This is a clear gap in the proof of Theorem 2.1. The paper has many independent supporting elements—new Carleman estimates, Foias-Prodi estimates, and the overall structure is coherent—so no fatal flaw is evident. The concern is a gap in the argument that a careful revision could fix, which supports the existing CONDITIONAL verdict rather than rejection or acceptance.","tokens_in":28003,"tokens_out":25902,"duration_ms":275409,"concrete_test":"Re-derive the contraction property for a common noise η: replace h by h+η in Proposition 4.2 and check that Lemma 4.1 and Theorem 3.2 hold uniformly over η∈K, i.e., verify that the truncated observability inequality still applies with coefficients from the noisy solution û and that the nonlinear error bound in Step 3 of the proof is independent of η. If the contraction holds for every η∈K, the concern is resolved; if it only holds for η=0, the verification of (LS) fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The abstract criterion Proposition 4.1 requires (LS): for every current noise η∈K, the controlled transition S(u, η+Φ(u,u′,η)) must contract toward S(u′, η). The manuscript's Prop. 4.2 instead proves contraction between S(û0, h) (deterministic forcing h, no noise) and S(u0, h+Υ(h,û0)(u0−û0)). If both copies are driven by the same η, the noise cancels in the difference v = u−û, and the correct reference û is the solution with forcing h+η, not h. The paper does not establish contraction for this noisy reference; it even defines Φ(u0,û0,η) := Υ(η,û0)(u0−û0), passing η into an argument that Proposition 4.2 requires to be a deterministic forcing in H^2(D_T). Since η is only in H^1_0(D_T), this is out of domain. Without a valid (LS), Proposition 4.1 cannot be applied, and Theorem 2.1's exponential ergodicity is not proven. This is a more direct gap than Condition (AC), which is at least a stated hypothesis with one verified special case.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the long-time statistical behavior of the randomly forced Korteweg-de Vries-Burgers (KdVB) equation on the one-dimensional torus. In the first part, for a space-time localized bounded noise of the form (2.2) satisfying Condition (DN) and an approximate controllability assumption (AC), Theorem 2.1 claims exponential mixing in the dual-Lipschitz metric. The proof combines a new global Carleman estimate (Theorem 3.1), an observability inequality (Proposition 3.1), a truncated observability inequality (Theorem 3.2), a squeezing property (Proposition 4.2), and the abstract coupling criterion of Shirikyan (Proposition 4.1). In the second part, for multiplicative white noise satisfying (g1)-(g3), Theorem 2.2 claims uniqueness of the invariant measure when L3<1, and Theorem 2.3 adds convergence to the invariant measure when L3<1/sqrt(5). These results rely on a Foias-Prodi estimate in expectation (Theorem 5.1), moment estimates, estimates in probability, and the asymptotic coupling method of Glatt-Holtz-Mattingly-Richards and Kulik-Scheutzow.","tokens_in":28249,"tokens_out":18866,"duration_ms":205129,"significance":"If correct, the results are significant: they would provide the first exponential ergodicity result for KdVB with degenerate space-time localized noise and the first ergodicity results for KdVB with multiplicative white noise. The new Carleman estimate and the Foias-Prodi estimate are potentially useful beyond this paper. The manuscript is largely self-contained on the analytic side, and the abstract probabilistic criteria are applied from the literature. The hypotheses are explicit, and the thresholds L3<1 and L3<1/sqrt(5) are concrete and verifiable. The main chain of reasoning from the PDE estimates to the abstract criteria is clearly laid out, except for the load-bearing gap discussed below.","major_comments":[{"comment":"The verification of Hypothesis (LS) is not valid as written. The Markov transition in Theorem 2.1 is u_k = S(u_{k-1}, h + η_k), where h is a fixed deterministic force. Proposition 4.2 proves the contraction (4.9) between S(û0, h_ref) and S(u0, h_ref + Υ(h_ref, û0)(u0 − û0)). To satisfy (LS) of Proposition 4.1, the reference force h_ref must be the total force on the first path, namely h + η_k (or h + ζ for a control ζ), not η_k alone. The paper instead defines Φ(u0,û0,η) := Υ(η,û0)(u0 − û0), dropping the deterministic term h. Thus the controlled transition used in (LS) is not the one required for the actual noise-driven system. This is a genuine mismatch and leaves Theorem 2.1 unproved.","section":"§4.3, definition of Φ"},{"comment":"Even after replacing η by the total force h+η, Proposition 4.2 is not directly applicable. Proposition 4.2 is stated for h ∈ H^2(D_T), while in Theorem 2.1 the deterministic force is only assumed to be in H^1_loc(R+×T), and the noise support K is only shown to be compact in H^1_0(D_T). Hence h+η is in general only in H^1(D_T), not H^2(D_T). No extension of Proposition 4.2 to H^1(D_T) is proved. Since (LS) is a necessary hypothesis of the abstract criterion Proposition 4.1, this domain mismatch is load-bearing. It may be repairable by strengthening Proposition 4.2, but as written the proof of Theorem 2.1 is incomplete.","section":"§4.2 and §4.3, domain of Proposition 4.2"}],"minor_comments":[{"comment":"In the Lipschitz continuity statement, the displayed formula reads '∥Υ(h1, ˆu1) − Ψ(h2, ˆu2)∥'; the second term should be Υ(h2, ˆu2), not Ψ.","section":"§4.2, Proposition 4.2"},{"comment":"The statement says 'ξu0,v0 ∈ ˆC(Pu0, Pz0)'; the second marginal should be Pv0. Also, in the definition of D and D_n^ε, 'u(n)' should presumably be 'y(n)'.","section":"§6.1, Theorem 6.1"},{"comment":"Step 2 proves convergence of the coupling using P(∥u(n) − ˜v(n)∥ > ε), but Theorem 5.1 controls E∥u(n) − v(n)∥^2 for v solving the nudged equation (5.2), not for the stopped-equation solution ˜v from (6.2). If ˜v is a typo for v, please correct it; otherwise the argument needs clarification.","section":"§6.3, Step 2 of Theorem 2.3"},{"comment":"The condition '|ψ′| > 0' cannot hold on the entire torus for a C∞ periodic function; it should be formulated as '|ψ′| > 0 on T\\ω' (or an analogous condition). The constructed example indicates this is intended, but the text should be unambiguous.","section":"§3.1, condition (3.2)"},{"comment":"The phrase 'η_l = ζ_l (1 ≤ l ≤ k)' is confusing; it should likely be '1 ≤ i ≤ l' or similar. Also, in Condition (DN), 'assumptations' is a typo.","section":"§2.2, notation after (2.4)"},{"comment":"References [14] and [26] appear to be the same paper (Glatt-Holtz, Martinez, Richards); please remove the duplicate or cite different versions appropriately.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The principal issue is the verification of Hypothesis (LS) in Theorem 2.1. It is a substantive but likely fixable gap: one needs to apply Proposition 4.2 with the total force h+η and prove the necessary estimates for H^1 forcing. The rest of the paper, including the Carleman estimate and the Foias-Prodi estimates, appears valuable, but the advertised Theorem 2.1 is not supported as written."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper is worth taking seriously. The genuinely new pieces are the global Carleman estimate for the linear KdVB operator (Theorem 3.1), the truncated observability inequality it feeds, and Foias–Prodi estimates in expectation for multiplicative noise. Those are real analytic contributions, not just repackaged machinery. The ergodicity theorems for localized and multiplicative noise go beyond the additive-noise KdV results in [14], and the abstract coupling framework is applied in a structurally coherent way. If the main theorems are correct, this is the first result of its kind for KdVB with degenerate non-additive noise.\n\nThe proofs are for the most part consistent: the Carleman estimate drives the observability and squeezing arguments, and the thresholds (L3<1, L3<1/√5) line up with the stopping-time estimates. I don't see circularity or hidden fitting. The author is honest about using standard coupling criteria from Shirikyan and Glatt-Holtz–Mattingly–Richards/Kulik–Scheutzow.\n\nNow the soft spots, in order of seriousness.\n\nFirst, the stress-test note lands: in Theorem 2.1 the verification of Hypothesis (LS) applies Proposition 4.2 with h = η, but Proposition 4.2 is stated for h ∈ H^2(D_T) while the noise η is only in H^1_0(D_T). The reference solution û in the squeezing argument is also the solution with deterministic forcing h, not with forcing h+η as the abstract (LS) condition requires. This is a genuine domain/statement mismatch. It looks fixable—extend Proposition 4.2 to H^1 forcing or adjust the abstract criterion—but as written Theorem 2.1's proof has a gap at that point.\n\nSecond, the paper leans on several deferred items: Lemma 4.1(1) is outsourced to [32], the implication (DN) ⇒ (D) is asserted, and Step 1 of Theorem 3.2 invokes backward uniqueness for (3.1) without proof. Each is plausible, but for a paper whose central claim depends on them, they should be stated as assumptions or proved.\n\nThird, there is a norm-power typo in Proposition 6.2: E_{R,m} is defined with ∫∥PN(u−v)∥ ds > R, but the subsequent Markov bound and Step 3 use ∫∥PN(u−v)∥^2 ds. Clearly the definition should use the squared norm; minor, but it will confuse readers.\n\nWho this is for: people working on ergodicity of randomly forced dispersive equations or on degenerate-noise SPDEs. They will get a lot from the Carleman and Foias–Prodi parts even if the final theorems need patching. The paper deserves a serious referee; I'd send it to review, expecting revisions on the (LS) verification and the deferred lemmas. A reader who needs only the analytic estimates can use them now, but I would not rely on Theorem 2.1's proof in its current form.","headline":"A serious, mostly sound paper that extends ergodicity theory to KdVB under localized and multiplicative noise; the main new estimates are credible, but Theorem 2.1 has a domain gap in the (LS) verification and a few deferred steps that need clean-up.","tokens_in":28789,"tokens_out":2895,"would_cite":true,"duration_ms":33432,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60H15","35R60","37A25"],"pacs":[],"model":"deepseek-v4-flash","headline":"Randomly forced KdVB equations are ergodic: their statistical state eventually forgets the initial condition and is unique.","keywords":["ergodicity","Korteweg-de Vries-Burgers equation","Carleman estimate","Foias-Prodi estimate","degenerate noise","coupling method","asymptotic coupling","invariant measure"],"falsifier":"A direct test of Condition (AC): pick a specific noise support K, a target u-bar, and a ball BH(R), then compute the reachable set {S_l(v, zeta_1,...,zeta_l) : zeta_j in K}. If for some R there is a gap of size epsilon > 0 separating the reachable set from u-bar, then (AC) fails and Theorem 2.1 cannot be invoked. A concrete instance would be a non-small deterministic forcing h for which the KdVB equation has two attracting periodic responses; trajectories near the two responses could not be steered close to one common point by controls near zero, so no unique stationary measure should be expec","tokens_in":27825,"feed_emoji":"🌊","tokens_out":7917,"duration_ms":90549,"temperature":0.7,"pith_summary":"The paper asks what happens, over long times, to solutions of the Korteweg-de Vries-Burgers equation when it is randomly forced by noise that is far from uniformly distributed: either concentrated in a small region of space-time, or multiplicative, meaning noise scaled by the solution value. It claims that in both cases the randomness eventually settles into a unique probability distribution on solutions, independent of where the solution started. For the space-time localized noise the convergence to this stationary measure is exponential in the number of noise periods; for multiplicative white noise the paper proves uniqueness of the invariant measure, and under a stronger growth condition, convergence of the law from any initial state. If correct, this means the long-term statistical behavior of randomly forced KdVB waves is fully determined by the forcing alone. The proof relies on new quantitative estimates for the underlying deterministic equation, in particular a Carleman estimate and a Foias-Prodi estimate.","feed_headline":"KdVB equation settles into one statistical state under random forcing","feed_subtitle":"Localized or multiplicative noise drives the dispersive wave equation to a unique invariant measure.","key_machinery":"The engine of the localized-noise proof is a new global Carleman estimate for the linear complex KdVB equation (Theorem 3.1). It gives weighted L2 bounds on a solution over the whole torus in terms of the equation's right-hand side and of the solution on an arbitrarily small subdomain omega; from it the paper derives an observability inequality and a truncated observability inequality (Theorem 3.2) saying that low modes of the initial data can be recovered from finitely many projected observations. This observability is fed into an optimal-control problem whose solution produces a 'squeezing' map: if two solutions start close, one can add a small control supported on finitely many modes so t","core_discovery":"In the author's own terms, the contribution is three theorems. Theorem 2.1: for the KdVB equation on the circle with a T-periodic deterministic forcing h, driven by an i.i.d. space-time localized noise satisfying structural condition (DN) and approximate controllability condition (AC), there exists a unique stationary measure mu and positive constants C, sigma such that every initial condition u0 satisfies ||P_k(u0,.) - mu||*_L <= C(1 + ||u0||^2)e^{-sigma k}. Theorem 2.2: for multiplicative white noise g(u)dW with Lipschitz coefficient satisfying (g1)-(g3) and linear-growth constant L3 < 1, provided the noise touches at least M >= N0 modes, the Markov semigroup has a unique ergodic invariant","pith_inferences":["The same Carleman machinery is likely to yield exact controllability and quantitative decay results for KdVB-type equations, as the author hints; if so, the mixing theorem would follow from controllability alone in a wider range of forcings.","Condition (AC) is probably not necessary: ergodicity might persist for larger h even if approximate controllability to a single point fails, but the current proof would need a different route, such as controllability to a set or partial controllability on low modes.","The thresholds L3 < 1 and L3 < 1/sqrt(5) are likely not sharp; numerical experiments with multiplicative noise could map the actual boundary for loss of uniqueness, for example with a diffusion coefficient of the form g(u) = alpha u + c near alpha = 1.","The space-time-localized noise result can be read as evidence that deterministic KdVB transport amplifies finite-dimensional randomness into full ergodicity, suggesting similar results for other third-order dispersive SPDEs with Burgers-type dissipation."],"forward_implications":["If Theorem 2.1 holds for a given h and noise, then for any two initial data the transition laws approach the same stationary distribution exponentially, so prediction of long-run statistics such as means, correlations, and probabilities of large-amplitude waves does not require knowing the initial condition.","For small T-periodic deterministic forcing h, Proposition 4.3 verifies Condition (AC), so exponential ergodicity applies to the physically relevant case of small periodic pumping plus localized random shaking.","Under (g1)-(g3) with L3 < 1 and enough active noise modes, there is a unique ergodic invariant measure, so time averages of observables converge almost surely to that measure's expectation.","With the stricter growth bound L3 < 1/sqrt(5), even a single initial distribution converges to the invariant measure, making the model asymptotically stable in law.","The Foias-Prodi estimate gives explicit moment controls on the difference between true and nudged trajectories, quantifying how quickly information in high Fourier modes is forgotten."],"supporting_citations":[{"why":"Introduces the control/coupling scheme, including the optimal-control problem and squeezing property, that the paper adapts to KdVB.","marker":"[32]"},{"why":"States Proposition 4.1, the abstract criterion turning approximate controllability, local stabilisability and noise decomposability into exponential ergodicity.","marker":"[33]"},{"why":"Supplies the fixed-point analysis used to show existence of a stable T-periodic solution and hence verify Condition (AC) for small h.","marker":"[12]"},{"why":"Provides the preceding ergodicity result for KdV-type equations with additive noise and a Foias-Prodi estimate for the KdV family.","marker":"[14]"},{"why":"Original Foias-Prodi estimate whose mechanism of controlling large modes via finite-dimensional projection is extended to KdVB.","marker":"[20]"},{"why":"Supplies the asymptotic-coupling uniqueness criterion, Theorem 6.1, used in the proof of Theorem 2.2.","marker":"[23]"},{"why":"Supplies the generalized-coupling convergence criterion, Theorem 6.2, used in the proof of Theorem 2.3.","marker":"[24]"},{"why":"Provides the well-posedness and existence-of-invariant-measure framework for PDEs with multiplicative noise, and the analogue for 2D Navier-Stokes equations.","marker":"[25]"},{"why":"Supplies the infinite-dimensional stochastic-analysis foundations, including cylindrical Wiener processes and invariant-measure existence used in the multiplicative-noise setting.","marker":"[8]"}],"fun_headline_variants":["Localized or multiplicative noise forces KdVB to a unique invariant measure","KdVB reaches unique steady state under two types of random noise","Exponential ergodicity for KdVB under random forcing","Unique invariant measure proven for KdVB with random noise","Randomly forced KdVB still has one unique steady state"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"For the localized-noise theorem, everything rests on Condition (AC): the noise must be able to steer any state in a large ball arbitrarily close to a fixed state using finitely many allowed random functions, and the paper proves this only for small deterministic forcing, so if the forcing is not small this precondition could fail and the exponential ergodicity proof would have no basis.","fun_headline_variants_meta":{"raw":{"variants":["Localized or multiplicative noise forces KdVB to a unique invariant measure","KdVB reaches unique steady state under two types of random noise","Exponential ergodicity for KdVB under random forcing","Unique invariant measure proven for KdVB with random noise","Randomly forced KdVB still has one unique steady state"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00077,"raw_usage":{"total_tokens":3212,"prompt_tokens":672,"completion_tokens":2540,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":416,"completion_tokens_details":{"reasoning_tokens":2452}},"tokens_in":416,"tokens_out":2540,"duration_ms":20995,"temperature":1.0,"reasoning_tokens":2452,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T12:05:59.697093+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct test of Condition (AC): pick a specific noise support K, a target u-bar, and a ball BH(R), then compute the reachable set {S_l(v, zeta_1,...,zeta_l) : zeta_j in K}. If for some R there is a gap of size epsilon > 0 separating the reachable set from u-bar, then (AC) fails and Theorem 2.1 cannot be invoked. A concrete instance would be a non-small deterministic forcing h for which the KdVB equation has two attracting periodic responses; trajectories near the two responses could not be steered close to one common point by controls near zero, so no unique stationary measure should be expec","supporting_citations":[{"cited_title":"Control and mixing for 2D Navier-Stokes equations with space-time localised noise[J]","cited_arxiv_id":null,"evidence_quote":"Introduces the control/coupling scheme, including the optimal-control problem and squeezing property, that the paper adapts to KdVB."},{"cited_title":"Controllability implies mixing II","cited_arxiv_id":null,"evidence_quote":"States Proposition 4.1, the abstract criterion turning approximate controllability, local stabilisability and noise decomposability into exponential ergodicity."},{"cited_title":"Recurrent solutions of the derivative Ginzburg-Landau equation with boundary forces[J]","cited_arxiv_id":null,"evidence_quote":"Supplies the fixed-point analysis used to show existence of a stable T-periodic solution and hence verify Condition (AC) for small h."},{"cited_title":"Sur le comportement global des solutions non-stationnaires des ´ equations de Navier-Stokes en dimension 2[J]","cited_arxiv_id":null,"evidence_quote":"Original Foias-Prodi estimate whose mechanism of controlling large modes via finite-dimensional projection is extended to KdVB."},{"cited_title":"On unique ergodicity in nonlinear stochastic partial differential equations[J]","cited_arxiv_id":null,"evidence_quote":"Supplies the asymptotic-coupling uniqueness criterion, Theorem 6.1, used in the proof of Theorem 2.2."},{"cited_title":"Generalized couplings and convergence of transition probabilities[J]","cited_arxiv_id":null,"evidence_quote":"Supplies the generalized-coupling convergence criterion, Theorem 6.2, used in the proof of Theorem 2.3."},{"cited_title":"Uniqueness of the invariant measure and asymptotic stability for the 2D Navier-Stokes equations with multiplicative noise[J]","cited_arxiv_id":null,"evidence_quote":"Provides the well-posedness and existence-of-invariant-measure framework for PDEs with multiplicative noise, and the analogue for 2D Navier-Stokes equations."},{"cited_title":"Stochastic Equations in Infinite Dimensions, Cambridge University Press, 2014","cited_arxiv_id":null,"evidence_quote":"Supplies the infinite-dimensional stochastic-analysis foundations, including cylindrical Wiener processes and invariant-measure existence used in the multiplicative-noise setting."}],"review_version":1}