{"id":"61af49db-d19e-4f50-9234-426c9ff36f64","arxiv_id":"2501.14920","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For the complex-valued periodic mKdV, the authors construct infinitely many invariant weighted Gaussian measures and prove unconditional well-posedness at H^s for s>4/3.","lead":"This paper constructs a sequence of random probability measures that stay unchanged as the complex modified KdV equation evolves in time, one measure for each of the equation's infinitely many conservation laws. It also proves a new uniqueness statement for the equation's strong solutions, and shows that almost every random initial condition grows at most logarithmically in time.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"General-n extension of Proposition 4.1 is asserted by \"mutatis mutandis\" in Section 6; the key coefficient (6.9) needs explicit verification, and its denominator appears to contain a typo.","rationale":"I read the paper in good faith. The deterministic part (Theorem 1.2) appears carefully proved, and the n = 2 case of the probabilistic construction is given in detail with explicit multilinear estimates. The central claim, however, is the full sequence of invariant measures for every n ≥ 2, and that claim depends on Proposition 4.1 for all n. The paper explicitly says the proof will be given only for n = 2 and that the general case is sketched in Section 6. The sketch reduces the problem to the coefficient (6.9) and states without proof that it has the same symmetric structure as (4.15). This is precisely the load-bearing point: the n = 2 proof relies on a delicate imaginary-part cancellation and on a case-by-case analysis of pairings, so an assertion of sameness is not a substitute for the actual estimates. The notational mismatch in the denominator of (6.9) reinforces the concern, since the displayed expression is not literally the coefficient that would arise from the measure mu_n defined in (1.9). The reader's weakest_assumption identifies exactly this gap, and I agree with the CONDITIONAL verdict. I do not see a separate fatal flaw in the deterministic theory or in the n = 2 construction; the appropriate action is to require completion of the general-n proof before full acceptance, which matches the reader's conditional recommendation.","tokens_in":47287,"tokens_out":22891,"duration_ms":180587,"concrete_test":"Derive (6.9) from (6.6)-(6.7) by substituting the random vector (1.9) for general n, and correct the denominator to product <j_i^n>. Then, for n = 3, explicitly reproduce the proof of (4.15): decompose the frequency set into 0-pairing and all nine 1-pairing classes, prove the cancellation lemma for the analogues of \\tilde I^{j2,j6} and \\tilde I^{j3,j6}, and verify that every residual sum is O(N^{-\\eta}) for some \\eta > 0. If every class closes with the corrected denominator, the Section 6 claim is substantiated; if any class fails because the denominator correction weakens the decay, Proposition 4.1 for general n is false as written and Theorem 1.5 needs revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 1.5 for every n ≥ 2 rests on Proposition 4.1, the almost-invariance of rho_{n,R,N} along the truncated flows. Proposition 4.1 is proved in detail only for n = 2. For n > 2, Section 6 isolates the leading dangerous term and asserts that (6.9) \"has the same symmetric structure as (4.15) and hence can be treated in the same way mutatis mutandis\". No lemmas, pairing decompositions, or convergence rates are supplied for general n, and even in the n = 2 proof the estimate (4.14) is skipped as \"similar\". Moreover, (6.9) as written has a notational inconsistency: the random vector (1.9) has Fourier coefficients with denominator sqrt(1+|j|^{2n}) = <j^n>, so substituting it into (6.6) should produce denominator product <j_i^n>, not product <j_i^{2n}>. If the denominator is corrected to the actual Gaussian weight, the delicate cancellations and summability bounds from Section 4 must be redone for general n; the paper does not do this. Since the existence of the full sequence of invariant measures with increasing regularity is the headline result, this unverified general-n step is the load-bearing risk. The deterministic well-posedness result and the n = 2 measure construction are not affected.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the periodic complex-valued modified KdV equation ∂_t u + ∂_x^3 u = 6|u|^2 ∂_x u. The deterministic result, Theorem 1.2, establishes unconditional local well-posedness of strong solutions in H^s for s > 4/3, together with uniform estimates for the frequency-truncated flows Φ_N; Corollary 1.3 globalizes the statement for s ≥ 2. The main probabilistic result, Theorem 1.5, asserts that for every integer n ≥ 2 and every s ∈ (4/3, (2n−1)/2) there is a Borel set of full µ_n-measure on which the flow is global, grows at most logarithmically in H^s, and leaves the weighted Gaussian measure ρ_{n,R} invariant. The construction is carried out in detail for n = 2 in Sections 4 and 5; Section 6 sketches the extension to general n by exploiting the structure of the higher conservation laws E_{2n+1} and reducing the problem to a single multilinear estimate, Eq. (6.9).","tokens_in":47579,"tokens_out":11211,"duration_ms":99162,"significance":"If fully established, the result would be a significant advance. It constructs invariant weighted Gaussian measures for complex-valued mKdV, a case not accessible to the straightforward Zhidkov argument because the quadratic part of the Hamiltonian is indefinite; it also provides a sequence of measures supported on increasingly regular Sobolev spaces. The deterministic part is essentially self-contained and includes unconditional uniqueness at H^s for s > 4/3, which appears to be below the previously known regularity threshold. The n = 2 proof demonstrates a genuine cancellation mechanism specific to complex-valued data, and no parameters are fitted; the main probabilistic input is cited from published work. At present, however, the full statement of Theorem 1.5 for every n ≥ 2 rests on a sketched extension that is not proved, so the verified core of the paper is the deterministic theorem together with the n = 2 measure construction.","major_comments":[{"comment":"Theorem 1.5 is stated for every n ≥ 2, but the proof of Proposition 4.1, which is the almost-invariance statement needed in Section 5, is given in detail only for n = 2. Section 6 reduces the general case to the assertion that (6.9) 'has the same symmetric structure as (4.15) and hence can be treated in the same way mutatis mutandis'. No general-n analogues of Lemmas 4.5–4.12 are provided, and the cancellation for the 1-pairing cases j2 = j6 and j3 = j6, which is delicate already for n = 2, is not verified for the exponents n−1. Since the existence of the infinite sequence of invariant measures is the headline claim, this is a load-bearing gap. Please provide a complete proof for all n, or reformulate Theorem 1.5 as a statement for n = 2.","section":"Section 6; Eq. (6.9)"},{"comment":"There is a consistency error in the displayed estimate (6.9). The Gaussian vector (1.9) has Fourier coefficients with denominator sqrt(1 + |j|^{2n}) = ⟨j^n⟩, so substitution of (1.9) into (6.6) gives denominator product ∏_{i=1}^6 ⟨j_i^n⟩, not ∏_{i=1}^6 ⟨j_i^{2n}⟩. If the displayed exponent is kept, one is estimating a different and easier Gaussian model; if the denominator is corrected, the summability and cancellation arguments of Section 4 must be redone for general n. This needs to be fixed and explained.","section":"Section 6; Eq. (6.9)"},{"comment":"The invariance of the measure on the invariant set Σ is justified only by the sentence 'the invariance of the measure on Σ follows by exactly the same computations done for the Benjamin-Ono equation in [68]'. Since the truncated measures are not exactly invariant and one must pass through the almost-invariance of Proposition 4.1, the approximation argument should be written out, including the convergence of the densities and the handling of sets that depend on the truncation parameter.","section":"Section 5.2"}],"minor_comments":[{"comment":"The estimate (4.14) is asserted to follow 'similarly' and is not proved. Since it is one of the three reductions in the proof of Proposition 4.3, please provide at least a concise completion of the argument.","section":"Section 4.1, after Eq. (4.14)"},{"comment":"In the paragraph after Corollary 1.3, the text refers to 'Corollary 1.2' when describing deterministic global well-posedness; the intended reference appears to be Corollary 1.3.","section":"Section 1.3"},{"comment":"The notation ρ_{n,j} is used for the radius parameter, whereas the measures were defined as ρ_{n,R}; please align the notation.","section":"Section 3, Proposition 3.4"},{"comment":"The almost-invariance estimate is invoked on the interval [−2^j, 2^j], while Proposition 4.1 is stated for t ∈ [0, T]; the extension to negative times should be stated explicitly.","section":"Section 5.1, Eq. (5.10)"},{"comment":"References [34] and [36] appear to be the same arXiv preprint and should be merged or one removed.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The main issue is structural rather than stylistic: the paper's headline theorem is not proved as stated. If the authors can supply a complete proof of the general-n estimate (6.9) with the corrected denominator, the paper would be strong; otherwise the theorem should be restricted to n = 2. I do not see circularity or parameter fitting; the deterministic part and the n = 2 construction are sound enough to be the basis for a publishable paper."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper has two genuinely new things: unconditional uniqueness of strong solutions for complex periodic mKdV at s>4/3, and a sequence of invariant weighted Gaussian measures associated to higher conservation laws. The deterministic theorem is proved cleanly and self-contained, using Kato-Ponce estimates and a careful truncated-flow argument. The n=2 measure construction is also convincing: the almost-invariance along the truncated flows is a real technical obstacle, and the cancellation via the imaginary part in (4.15) is clever. The paper honestly marks what is proved in detail and what is only sketched.\n\nThe soft spot is exactly where the reader and stress-test put it: Section 6 extends the argument to all n by saying the dangerous term (6.9) has the same symmetric structure as (4.15) and can be treated mutatis mutandis. That is not a proof. Propositions 4.5–4.12 for n=2 are long and delicate; the general-n version needs at least statements of the analogous lemmas and the convergence-rate bounds. The stress-test is also right that (6.9) appears to have a typo: the random vector (1.9) has denominator sqrt(1+|j|^{2n}) = <j^n>, so the product in (6.9) should be <j_i^n>, not <j_i^{2n}>. If that is just a typesetting slip, it still needs correcting; if it reflects the actual computation, then the claimed analogy with (4.15) is off. Either way, Theorem 1.5 for all n is not yet supported. The n=2 case and the deterministic theorem stand.\n\nWho is this for? Researchers working on probabilistic well-posedness and invariant measures for integrable dispersive equations. The deterministic result and the n=2 measure are worth citing now. The full ladder of measures is a plausible conjecture, but it is not established in this version.\n\nRecommendation: send it to peer review, but require the authors to either complete the general-n argument or explicitly restrict the invariant-measure construction to n=2. A serious referee should not desk-reject it; the reviewer time is warranted by the new deterministic result and the substantial n=2 proof.","headline":"New unconditional uniqueness for complex mKdV and a detailed n=2 invariant measure, but the advertised general-n ladder is asserted on a sketch that needs real work.","tokens_in":48107,"tokens_out":3485,"would_cite":true,"duration_ms":31924,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q53","35A01","35A02","37K10","37K05"],"pacs":[],"model":"deepseek-v4-flash","headline":"Complex-valued mKdV admits an infinite family of invariant measures, at higher and higher regularity, even though its Gibbs measure does not exist.","keywords":["complex-valued mKdV","invariant measures","weighted Gaussian measures","unconditional well-posedness","Sobolev spaces","conservation laws","pairing estimates","periodic torus"],"falsifier":"For a fixed n > 2, for example n = 3, compute the L2(omega) norm of the finite sum in (6.9) for increasing values of the truncation N and check whether it decays at least at a power rate in N; failure to decay would show that the cancellation structure does not survive to higher conservation laws. A complementary check is to verify whether the analogue of Lemma 4.5, the symmetry identity that zeroes the problematic pairing sets, holds for the general exponent patterns of Section 6.","tokens_in":47120,"feed_emoji":"🌊","tokens_out":4850,"duration_ms":43172,"temperature":0.7,"pith_summary":"The paper studies the complex-valued periodic modified Korteweg–de Vries equation and proves two results. First, it establishes unconditional local well-posedness for strong solutions with initial data in H^s for s > 4/3, a threshold below the previously known s > 3/2. Second, and central, it constructs a sequence of weighted Gaussian measures rho_{n,R} supported on Sobolev spaces of increasing regularity, and proves that almost every initial datum yields a global flow whose Sobolev norm grows at most logarithmically in time. The invariance proof is written in full for n = 2, while for general n the paper gives a structural reduction and asserts that the same cancellation mechanism applies. If correct, this provides an infinite family of invariant measures for a Hamiltonian system whose Gibbs measure is not available, extending Zhidkov's program to the complex setting.","feed_headline":"Infinite invariant measures built for complex mKdV","feed_subtitle":"New proof works even though the Gibbs measure fails, and almost surely Sobolev norms grow only logarithmically.","key_machinery":"The central mechanism is a two-layer approximation: the finite-dimensional truncated flow Phi_N(t) is Liouville and preserves the Gaussian part of the measure rho_{n,R,N}, while the density factor built from lower conservation laws is almost invariant because the time derivative of E_{2n+1} along Phi_N(t) is shown to vanish in L2(mu) as N tends to infinity. The vanishing is proved with two tools: Kato–Ponce type commutator estimates for the deterministic energy bounds, and a pairing combinatorics (Definition 3.5, Proposition 3.6) that controls the second moment of multilinear Gaussian sums. The critical cancellation for n = 2 is the identity that the imaginary part of the six-frequency sum in (4.15) is zero on certain pairing sets (Lemma 4.5), which removes the would-be divergences.","core_discovery":"The main discovery is that, despite the complex-valued mKdV having no Gibbs measure (because the quadratic part of its Hamiltonian energy E4 is indefinite), there exist weighted Gaussian measures built from the higher conservation laws E_{2n+1} that are invariant under the flow. The key obstacle in the complex case is a term that vanishes identically for real-valued solutions by integration by parts; the paper shows that in the complex case this term converges to zero in L2(mu_2) via a Fourier-side imaginary-part cancellation, proved by decomposing the frequency set according to pairings and exploiting symmetries. The full result for every n >= 2 is stated as Theorem 1.5, with Section 6 arguing that the dangerous high-order term (6.9) has the same symmetric structure as the n = 2 case.","pith_inferences":["If the Section 6 sketch can be completed for general n, the method shows that the absence of a Gibbs measure is not an obstruction to constructing invariant measures from higher conservation laws, suggesting the technique may transfer to other integrable hierarchies with indefinite quadratic energies.","The proof of the invariance of the set Sigma_R uses a constant D that must be uniform in s in [4/3, bar s]; verifying that the Gaussian bounds are indeed uniform is a checkable step that the paper asserts but does not fully spell out.","A natural testable extension is the focusing case, which the paper states follows with minor changes, but the sign of the nonlinearity could affect the delicate cancellation structure and would deserve a separate check.","The cancellation mechanism may be rephrased as a statement about random tensor decoupling, suggesting that the pairing estimates from the probabilistic toolbox might yield even sharper almost-sure bounds, such as sub-logarithmic growth, under additional assumptions."],"forward_implications":["The flow of complex mKdV is almost surely global for initial data in H^s with s > 4/3, whereas the deterministic global theory in the paper is proved only for s >= 2.","Almost surely, the H^s norm of the solution grows at most logarithmically in time, improving the exponential growth bound that follows from iterating the local Cauchy theory.","For every integer n >= 2 there is a weighted Gaussian measure rho_{n,R} invariant under the flow, giving an infinite sequence of invariant measures at increasing regularity.","The construction opens the way to Poincaré recurrence and statistical-mechanics interpretations for complex mKdV, paralleling the Gibbs-measure theory for real-valued equations."],"supporting_citations":[{"why":"Introduces invariant measures built from cutoffs of conservation laws for NLS and KdV, the approach extended here to complex mKdV.","marker":"[73]"},{"why":"Provides the almost-invariance plus full-measure globalization scheme that Section 5 adapts to the present setting.","marker":"[10]"},{"why":"Supplies the Fourier-restriction bound (8.37) on which the local Cauchy theory and the key time-space estimates rely.","marker":"[9]"},{"why":"Gives the Kato–Ponce commutator estimates used in the energy estimate (2.5) and in Lemma 2.4.","marker":"[35]"},{"why":"Provides the energy-method strategy and short-time bounds used in Proposition 2.1 for the truncated flow.","marker":"[53]"},{"why":"Supplies the pairing large-deviation bounds for multilinear Gaussian variables that underlie Proposition 3.6.","marker":"[22]"},{"why":"Companion random-tensor propagation estimates used in the same probabilistic machinery.","marker":"[23]"},{"why":"Proves the invariance of the measure once an invariant full-measure set has been constructed, used at the end of Section 5.","marker":"[68]"}],"fun_headline_variants":["Invariant measures for complex mKdV despite Gibbs failure","Complex mKdV gets invariant measures from higher laws","Weighted Gaussian measures invariant for complex mKdV","Sobolev norms grow slowly under new invariant measures","No Gibbs measure, but complex mKdV has invariant ones"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof that the sequence of invariant measures exists for every n >= 2 depends on the assertion in Section 6 that the general high-order term (6.9) 'has the same symmetric structure as (4.15) and hence can be treated in the same way mutatis mutandis'; no lemmas or convergence rates are proved for n > 2, and Proposition 4.1 is demonstrated only for n = 2.","fun_headline_variants_meta":{"raw":{"variants":["Invariant measures for complex mKdV despite Gibbs failure","Complex mKdV gets invariant measures from higher laws","Weighted Gaussian measures invariant for complex mKdV","Sobolev norms grow slowly under new invariant measures","No Gibbs measure, but complex mKdV has invariant ones"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001189,"raw_usage":{"total_tokens":4822,"prompt_tokens":772,"completion_tokens":4050,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":388,"completion_tokens_details":{"reasoning_tokens":3970}},"tokens_in":388,"tokens_out":4050,"duration_ms":25103,"temperature":1.0,"reasoning_tokens":3970,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T14:46:26.558373+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a fixed n > 2, for example n = 3, compute the L2(omega) norm of the finite sum in (6.9) for increasing values of the truncation N and check whether it decays at least at a power rate in N; failure to decay would show that the cancellation structure does not survive to higher conservation laws. A complementary check is to verify whether the analogue of Lemma 4.5, the symmetry identity that zeroes the problematic pairing sets, holds for the general exponent patterns of Section 6.","supporting_citations":[{"cited_title":"Zhidkov, Korteweg-de Vries and Nonlinear Schrödi nger Equations: Qualitative Theory","cited_arxiv_id":null,"evidence_quote":"Introduces invariant measures built from cutoffs of conservation laws for NLS and KdV, the approach extended here to complex mKdV."},{"cited_title":"Bourgain, Periodic nonlinear Schrödinger equation and invariant measures, Comm","cited_arxiv_id":null,"evidence_quote":"Provides the almost-invariance plus full-measure globalization scheme that Section 5 adapts to the present setting."},{"cited_title":"Bourgain, Fourier transform restriction phenomena f or certain lattice subsets and applications to nonlinear ev olution equations II","cited_arxiv_id":null,"evidence_quote":"Supplies the Fourier-restriction bound (8.37) on which the local Cauchy theory and the key time-space estimates rely."},{"cited_title":"Ionescu and C","cited_arxiv_id":null,"evidence_quote":"Gives the Kato–Ponce commutator estimates used in the energy estimate (2.5) and in Lemma 2.4."},{"cited_title":"Koch and N","cited_arxiv_id":null,"evidence_quote":"Provides the energy-method strategy and short-time bounds used in Proposition 2.1 for the truncated flow."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the pairing large-deviation bounds for multilinear Gaussian variables that underlie Proposition 3.6."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Companion random-tensor propagation estimates used in the same probabilistic machinery."},{"cited_title":"Tzvetkov, N","cited_arxiv_id":null,"evidence_quote":"Proves the invariance of the measure once an invariant full-measure set has been constructed, used at the end of Section 5."}],"review_version":1}