{"id":"fcbbe23c-1d70-4087-810b-d40104ad722b","arxiv_id":"2509.10318","paper_version":1,"verdict":"CONDITIONAL","confidence":"LOW","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Existence of small-amplitude, linearly stable, time quasi-periodic traveling solutions for 3D pure gravity water waves in finite depth on tori, for generic lattices and most depths.","lead":"This paper proves that three-dimensional gravity water waves in a periodic domain can travel as quasi-periodic waves with several different speeds at once, without ever becoming stationary in any moving frame. Mathematicians can now build on this to understand long-time behavior of water waves and other dispersive equations in higher dimensions.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Prop. 6.4 transversality for the depth parameter is unverified; the exponentially small h-dependence of the dispersion may make the bad-depth set non-small, so the asymptotically-full-measure claim lacks support.","rationale":"The reader's weakest assumption identifies exactly the depth-parameter transversality behind the measure estimates. I agree that this is the load-bearing point, because Theorem 1.4's 'large measure set of depths' is the only place the parameter h is used to overcome the degeneracy, and the provided text does not include Section 6/Proposition 6.4. The manuscript itself flags the exponential smallness of ∂_h√(|j|tanh(h|j|)) and asserts that momentum conservation plus Hypothesis 1.2 compensate, but the compensation is not demonstrated in the excerpt. The standard measure estimate for Melnikov conditions requires a lower bound on the derivative in h. For the unperturbed second Melnikov function N(h)=ω(h)·ℓ+ω_j(h)-ω_{j'}(h), the derivative is ℓ·ω'(h) + exponentially small terms. The momentum condition does not obviously restrict ℓ to a bounded set, so |ℓ·ω'(h)| may be very small for infinitely many ℓ unless an additional Diophantine/transversality property of the Gram matrix is proved. If the best available lower bound is exponential in |j|, the bad-set measure grows with |j| and the asymptotic-full-measure conclusion fails. This is not an accusation of error; it is a concrete unverified step that a specialist must check. The proposed test—isolating Proposition 6.4 and verifying the polynomial lower bound—would settle the issue. Given the reader already assigned CONDITIONAL at low confidence, my read does not change that verdict: the paper should remain conditional pending verification of this step. I also note the paper contains substantial independent technical content (pseudo-differential expansions, Egorov estimates, tame estimates) and no red flags beyond this unexhibited transversality lemma.","tokens_in":85881,"tokens_out":10204,"duration_ms":121932,"concrete_test":"Extract Proposition 6.4 and check whether it proves a polynomial lower bound of the form |∂_h(ω(h)·ℓ + ω_j(h) - ω_{j'}(h))| ≥ c⟨ℓ,j'⟩^{-a}, uniformly over all ℓ≠0 and j,j' satisfying V^Tℓ + j - j' = 0, with c>0, a>0 independent of |j|. Then re-run the measure estimate in §13.2 with this bound. If the bound is only O(e^{-c'|j|}), the asymptotic-full-measure statement in Theorem 1.4 is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 1.4 asserts existence for a Cantor-like set of depths Gζ with asymptotically full measure. This measure statement rests on the non-resonance/Melnikov conditions of Section 6, specifically Proposition 6.4, which is not reproduced in the provided text. The introduction (after (1.17)) acknowledges that the frequencies depend on h only through exponentially small corrections: |∂_h^k r(j,h)| ≤ C_k e^{-h|j|}. The standard degenerate-KAM measure estimate for the bad set of h where |ω(h)·ℓ + ω_j(h) - ω_{j'}(h)| < γ⟨ℓ,j'⟩^{-τ} requires a lower bound on the derivative in h of this small-divisor function. That derivative equals ℓ·ω'(h) + ∂_h(ω_j(h)-ω_{j'}(h)). Here ∂_h(ω_j(h)-ω_{j'}(h)) is O(⟨j⟩^{3/2} e^{-2h|j|}), exponentially small for large |j|, while ℓ·ω'(h) is a linear form in the integer vector ℓ with coefficients c_i(h)=∂_h ω_{ȷ_i}(h) > 0. Under the momentum constraint V^Tℓ + j - j' = 0, ℓ can grow with |j|, and |ℓ·c(h)| has no evident uniform lower bound: it can be as small as |ℓ|^{-A} if c(h) is Diophantine, or much smaller for Liouville-type ratios. If Proposition 6.4 only yields an exponentially small lower bound for the derivative, then the measure of the bad h set is O(γ e^{c|j|}) rather than O(γ⟨ℓ,j'⟩^{-τ}); summing over j,j' would diverge, and the claim that Gζ has asymptotically full measure would not follow. Since the entire parameter-selection mechanism of the theorem depends on this transversality, this is the single most load-bearing unverified piece.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims Theorem 1.4: under a generic non-resonance condition on the dual lattice (Hypothesis 1.2), for any finite ordered set of tangential wave vectors and for a Cantor-like set of depths with asymptotically full measure, the 3D pure-gravity water-wave equations on a flat torus admit small-amplitude, time quasi-periodic traveling solutions with an arbitrary number of rationally independent speeds. These solutions are asserted to be global in time, not stationary in any moving frame, and linearly stable. The strategy is a Nash-Moser/KAM scheme built on a detailed pseudo-differential expansion of the Dirichlet-Neumann operator, a reduction of the linearized operator to constant coefficients up to smoothing remainders, and a KAM diagonalization controlled by second Melnikov conditions and conservation of momentum. The version under review contains substantial parts of the functional calculus and the elliptic estimates in Sections 2-4, but the decisive Sections 5-13, including Proposition 6.4 and the measure estimates, are not reproduced and are only summarized in the introduction.","tokens_in":86352,"tokens_out":9656,"duration_ms":117807,"significance":"If the proof is correct, this is a major advance: it would be the first KAM result for an autonomous, quasi-linear, dispersive PDE in dimension greater than one, and the first construction of global time-quasi-periodic solutions of 3D water waves on compact domains that are not steady in any moving frame. The paper has clear strengths: it formulates an explicit genericity condition on the lattice, develops a substantial space-time tame pseudo-differential calculus for the Dirichlet-Neumann operator, and uses momentum conservation in a genuinely non-trivial way to remove exact resonances. I found no circularity in the strategy: the non-resonance conditions are imposed on parameter sets whose measure is claimed to be large, not assumed to equal the conclusion. However, the central measure statement rests on Proposition 6.4, which is not available in the submitted text; the proof of the theorem therefore cannot currently be certified.","major_comments":[{"comment":"The theorem's 'asymptotically full measure' claim for the good set G_ζ depends on transversality properties of the unperturbed Melnikov functions. The introduction, after (1.17), states that |∂_h^k r(j,h)| ≤ C_k e^{-h|j|}, and says the required transversality is proved in Proposition 6.4 using momentum conservation and Hypothesis 1.2. In the version under review, Section 6 and Proposition 6.4 are not reproduced. This is load-bearing: for F_{ℓ,j,j'}(h)=ω(h)·ℓ+ω_j(h)-ω_{j'}(h) with V^Tℓ+j-j'=0, the derivative is ℓ·ω'(h)+∂_h(ω_j-ω_j'). The second term is O(⟨j⟩^{3/2} e^{-2h|j|}), while ℓ·ω'(h) is a linear form in ℓ with fixed positive coefficients; for unfavorable h it can be superpolynomially small even when |ℓ|≍|j|. A merely exponential lower bound on |∂_h F| would give bad-set measure O(γ e^{c|j|}), whose sum over j,j' diverges. The proof must supply a polynomial, or at least summable, lo","section":"§6 / Prop. 6.4"},{"comment":"The set G_∞ is defined in (13.37) as the intersection over n of Λ_{γ_n}^∞(i~_n), i.e. infinitely many 0th-, 1st-, and 2nd-order Melnikov conditions at every Nash-Moser step. The measure estimate for such an intersection is not a formality. The eigenvalues μ_∞(j;λ,i) are explicitly said not to have an asymptotic expansion in powers of 1/|j|, and the only quantitative handle stated is |∇m_7(j)| ≲ ⟨j⟩^{-1/2} in Lemma 13.5. In Section 12, the second Melnikov condition (1.54) has an extra factor |j'|^τ in the denominator, and Lemma 12.4 is asserted to control the resulting derivative loss using the smoothing structure of the remainder and momentum conservation. These sections are not present in the version under review. Since the Nash-Moser inversion (Theorem 8.5) depends on the invertibility estimate (8.19), this is a second load-bearing gap in the proof as submitted.","section":"§12 and §13.2"},{"comment":"The entire normal-form reduction starts from the pseudo-differential expansion of G(η), Theorem 4.2. The proof is deferred to Section 5, whose content is not included in the version under review. The lemmas in Section 4 provide elliptic estimates but do not by themselves establish the symbol expansion (1.30) with remainders satisfying the unbalanced tame estimates (1.32). These remainders are later used to compensate for the derivative losses from the weak Melnikov conditions. As the submitted text stands, this foundational step is unverified, and the subsequent reducibility argument cannot be checked from the material provided.","section":"§5 / Theorem 4.2"}],"minor_comments":[{"comment":"Typo: 'bounded iopen set' should read 'bounded open set'.","section":"Hypothesis 4.1"},{"comment":"The quantifier over the tangential sites S is not precise. It is not stated whether ε0(s) and the measure estimates are allowed to depend on |ȷ_i|; if they do, the phrase 'for any choice of wave vectors' should specify that dependence, since the h-derivatives of the linear frequencies decay exponentially in |ȷ_i|.","section":"Theorem 1.4"},{"comment":"The composition estimate (2.35) has different derivative losses in the two terms (s+k+1 and s0+k+2). If intentional, this should be explained; otherwise it may be a typo.","section":"Lemma 2.8"},{"comment":"The symbol χ is used both for the frequency cut-off in (2.89) and for the cut-off in (2.39). This is confusing and should be renamed.","section":"§2.4 / (2.89)-(2.91)"},{"comment":"The discussion around (1.42)-(1.43) refers to |Re(a)| ≳ |ξ| to make y^M e^{y a} a symbol of order -M, but the sign convention on the symbol a is only described in words. Please state the hypothesis explicitly before the lemma that uses it.","section":"§5 / Lemma 5.8"}],"recommendation":"major_revision","confidential_remarks":"To the editor: the manuscript is ambitious and the presented portions are technically serious, with a coherent overall strategy and no apparent circularity. However, the version supplied to me omits the sections where the two most load-bearing arguments live: Proposition 6.4 (transversality and measure estimates for the depth parameter) and the detailed reducibility/measure arguments in Sections 12-13. Consequently I cannot certify Theorem 1.4. I recommend requesting the complete manuscript and, in the next round, a self-contained statement and proof of Proposition 6.4 with explicit lower bounds on the h-derivatives, together with the measure estimates in Section 13.2. If those parts are correct, the result is likely a significant advance; I do not see grounds for rejection based on the material presented. This is a completeness concern rather than a demonstrated technical error."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nPunchline: this is a serious, likely landmark paper. It is the first KAM theorem for an autonomous, dispersive, quasi-linear PDE in dimension greater than one, and the first construction of global in time, quasi-periodic, 3D pure-gravity water waves that are not stationary in any moving frame. The introduction and the first four sections make a convincing case that the strategy is genuinely new, going beyond Craig–Nicholls, Iooss–Plotnikov, and the earlier 2D quasi-periodic traveling-wave results.\n\nWhat is actually new: an arbitrary number of tangential speeds, solutions that are non-steady in any frame, finite depth with a Cantor-like set of depth parameters of asymptotically full measure, and linear stability. The normal form strategy is interesting: the sublinear dispersion makes the time derivative dominate, and the momentum conservation is used to kill dangerous resonances, combined with a pseudo-differential reduction of the Dirichlet–Neumann operator. The quantitative tame estimates in Sections 4–5 look carefully derived, with explicit derivative losses that do not grow with the Sobolev index. That is substantial work.\n\nWhat I could not do: verify Sections 5–13. The text provided to me stops partway through Section 4, so the heart of the proof—the KAM reducibility, the iterative scheme, and especially Proposition 6.4 with its transversality and the asymptotically full measure claim for the good-depth set Gζ—is not in front of me. The stress-test concern is not a manufactured flaw: the h-derivative of the dispersion is exponentially small in |j|, so a naive measure estimate would fail, and the whole parameter-selection mechanism depends on momentum conservation and the Diophantine lattice hypothesis to get the needed lower bounds. The authors say they prove this in Proposition 6.4, and I cannot check it. That is the single most load-bearing unverified piece. It is also the most likely place for a subtle gap.\n\nI want to stress what the paper does well. It is not circular: non-resonance conditions are imposed on a parameter set whose measure is proven large, and the counterterm method is standard. The paper is honest about its limitations: it explicitly notes the exponential smallness, the derivative losses, and the restriction to sublinear dispersion (no gravity–capillary case). That is the mark of careful, serious engagement, not a defect.\n\nWho this is for: PDE/KAM specialists, the water waves community, and anyone tracking high-dimensional KAM. It is not a quick read. A serious referee must go through Sections 6–13 in detail, with particular attention to Prop. 6.4 and the iterative reducibility in Section 12.\n\nRecommendation: this deserves a serious referee. The novelty and the evident care of the visible parts justify referee time even if the final verdict is conditional. I would send it to peer review.","headline":"A genuine first-result KAM paper for 3D gravity water waves; the proof is coherent and detailed in the visible sections, but the load-bearing measure-theoretic core (Prop. 6.4) is unverified in the text I saw, and that is where a specialist referee must dig.","tokens_in":86829,"tokens_out":2354,"would_cite":true,"duration_ms":28172,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q35","35B40","35S05","76B15"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper constructs small-amplitude, linearly stable, time quasi-periodic traveling wave solutions to the 3D pure-gravity water wave equations in finite depth, for generic lattices and almost every depth.","keywords":["quasi-periodic traveling waves","pure gravity water waves","finite depth","KAM theory","small divisors","Dirichlet-Neumann operator","normal forms","Melnikov conditions"],"falsifier":"At the unperturbed level, for a fixed finite set of tangential sites, compute the zeros in h of the second-Melnikov functions ω(h)·ℓ + Ω(j;h) ± Ω(j';h) under the momentum constraint; if any non-zero triple (ℓ,j,j') yields a zero set containing an interval, or if h ↦ that combination has zero derivative at a zero, then the transversality behind Proposition 6.4 breaks and the full-measure depth estimate would fail.","tokens_in":85797,"feed_emoji":"🌊","tokens_out":4593,"duration_ms":55014,"temperature":0.7,"pith_summary":"The paper aims to prove that the three-dimensional pure-gravity water wave equations on a flat torus, in finite depth, admit global-in-time solutions whose motion is genuinely quasi-periodic in time and traveling in space, not reducible to stationary Stokes waves in any moving frame. The existence holds for a generic choice of the horizontal lattice, for any finite number of propagation speeds, and for an asymptotically full-measure set of depths as the amplitude shrinks. The solutions are small, arbitrarily smooth, and linearly stable. If the proof is right, this supplies the first non-steady global solutions for 3D water waves on compact domains and the first KAM-type result for an autonomous quasi-linear dispersive PDE in dimension greater than one.","feed_headline":"Quasi-periodic traveling waves exist in 3D pure-gravity water","feed_subtitle":"A KAM construction yields global, linearly stable solutions for almost every depth on generic tori.","key_machinery":"The key machinery is a normal-form reduction of the linearized water wave operator at a quasi-periodic traveling wave. It combines a quantitative Egorov theorem and sharp tame pseudo-differential estimates for the Dirichlet-Neumann operator, the use of the sublinear dispersion relation to let the time derivative ω·∂_φ dominate the order-1/2 spatial operator, and conservation of momentum, which enforces that momentum-preserving time-independent operators are Fourier multipliers. This first stage reduces the linearized operator to a bounded Fourier multiplier up to smoothing remainders; a second KAM stage diagonalizes it using weak Melnikov conditions, with losses controlled by the smoothness","core_discovery":"On the paper's own terms, Theorem 1.4 states that, under Hypothesis 1.2 on the rational independence of the dual-lattice Gram matrix, for any finite set of tangential wave vectors with distinct lengths and for a large-measure set of depths h, the pure-gravity water wave system has a small-amplitude quasi-periodic traveling wave solution with a Diophantine frequency vector ω=ω(h,ζ) close to the linear frequencies, whose remainder is o(√ζ) in high Sobolev regularity. The central discovery is that this can be achieved despite three compounding difficulties: the weak sublinear dispersion relation √(|j|tanh(h|j|)), the quasi-linear (derivative-order) nonlinearity, and the absence of a sharp asymp","pith_inferences":["If the transversality mechanism fails for lattices not satisfying Hypothesis 1.2—for example, a square lattice with rational Gram-matrix components—then the asymptotically full-measure depth set may collapse; this is a natural numerical and theoretical boundary to probe.","The same normal-form scheme might carry over to other fluid models with sublinear dispersion and conserved momenta, such as internal waves or rotating shallow water, provided the linearized operator admits the same pseudo-differential structure.","Because the proof avoids reversibility assumptions, imposing reversibility as an additional symmetry would likely yield reversible (standing-type) quasi-periodic waves in 3D as well, a case the paper does not explicitly state."],"forward_implications":["The constructed quasi-periodic tori are global-in-time and linearly stable: the linearized flow near each torus is conjugate to a diagonal operator with purely imaginary spectrum in the normal directions.","The solutions are not stationary in any moving reference frame, so they enlarge the known building blocks of 3D periodic water waves beyond traveling Stokes waves.","The tori exist for an asymptotically full-measure set of depths as the amplitude tends to zero, with Diophantine frequencies that adjust with the amplitude.","The solutions are arbitrarily Sobolev regular in both time and space, so they are classical smooth solutions for high enough regularity indices.","The two-stage reducibility strategy, built on sublinear dispersion and momentum conservation, is presented as a template for other translation-invariant quasi-linear dispersive PDEs in dimension at least two."],"fun_headline_variants":["3D gravity waves: quasi-periodic traveling solutions proven","First KAM result for quasi-linear PDE in higher dimension","Global quasi-periodic waves exist for 3D water equations","Quasi-periodic waves break steady-state barrier in 3D"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The construction needs the depth parameter h to separate the resonances: although the linear frequencies depend on h only through exponentially small corrections, those corrections—together with the rational independence of the lattice Gram matrix and momentum conservation—must make every relevant Melnikov combination transverse in h; if that fails, the set of usable depths need not have full measure and the iteration cannot run.","fun_headline_variants_meta":{"raw":{"variants":["3D gravity waves: quasi-periodic traveling solutions proven","First KAM result for quasi-linear PDE in higher dimension","Global quasi-periodic waves exist for 3D water equations","Quasi-periodic waves break steady-state barrier in 3D"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000504,"raw_usage":{"total_tokens":2301,"prompt_tokens":751,"completion_tokens":1550,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":495,"completion_tokens_details":{"reasoning_tokens":1489}},"tokens_in":495,"tokens_out":1550,"duration_ms":10982,"temperature":1.0,"reasoning_tokens":1489,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T17:51:23.386576+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"At the unperturbed level, for a fixed finite set of tangential sites, compute the zeros in h of the second-Melnikov functions ω(h)·ℓ + Ω(j;h) ± Ω(j';h) under the momentum constraint; if any non-zero triple (ℓ,j,j') yields a zero set containing an interval, or if h ↦ that combination has zero derivative at a zero, then the transversality behind Proposition 6.4 breaks and the full-measure depth estimate would fail.","supporting_citations":[],"review_version":1}