{"id":"a456f724-6211-4f9e-94de-e9c9eb43b835","arxiv_id":"2412.02648","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Most small Gevrey initial data for a 1D NLS with random convolution potential give global almost-periodic solutions whose Gevrey norm stays bounded for all time.","lead":"For a nonlinear Schrödinger equation on a circle with a small quintic nonlinearity and a convolution potential, the authors prove that for almost all potentials and almost all small Gevrey-regular initial data, the solution exists for all time, is almost-periodic, and keeps its Gevrey norm nearly constant. This is a step toward showing that the typical KAM picture, where most trajectories lie on invariant tori, survives in infinite-dimensional Hamiltonian PDEs.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 4.5's Lipschitz extension of the initial-data shift u is the least secure step: the proof for U is only sketched, relies on [33, Cor. 10.8], and is needed to solve (5.1) and to justify the measure estimates in Lemma 6.2.","rationale":"The paper's main theorem is conditional on the existence of quantitative Lipschitz extensions of the counterterm and of the correction to the initial datum. The extension of a and r is delegated to [33, Cor. 10.8], but the extension of u is asserted by an analogous McShane argument without a proof. This is the single most load-bearing assumption because the fixed point equation (5.1) is solved on all of W_N using these extensions, and the frequency map's Lipschitz properties, which drive the measure estimates in Lemma 6.2 and Proposition 6.4, are derived from the same bounds. I found no internal contradiction elsewhere: the measure sum in Proposition 6.4 is justified by the standard Bourgain estimate for sum-zero frequency vectors, the Fubini argument for the set of potentials and initial data is valid, and the scaling argument in Section 7 is coherent. The overclaim about being the first such KAM result and the slightly ambiguous 'unique solution' wording do not affect the mathematical validity of the construction. An independent verification of Proposition 4.5, or access to a fully detailed proof in [33], would resolve the concern; absent that, the present paper is best regarded as conditional.","tokens_in":41017,"tokens_out":43099,"duration_ms":449369,"concrete_test":"Provide a complete proof of Proposition 4.5 for the function u: take a finite truncation of the tree series for u_j(c,zeta,epsilon), extend each coefficient U_{j,nu}^{(a,b)}(omega(zeta),epsilon) from K_N(gamma) to W_N by the McShane formula, and verify (i) the extension satisfies (4.11e) with the same constant C1, (ii) the series converges uniformly in zeta in W_N, and (iii) the bound (4.12) holds for the extended coefficients. If any of these fail on a truncation, the step is invalid; if they hold, the gap is presentational and the claim stands.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim of Theorem 2.3 requires solving the implicit function problem (5.1) for all (V,W) in V, and then measuring the set where the frequency is non-resonant. Both steps depend on Proposition 4.5, which asserts Lipschitz extensions A, R, U of a, r, u from the Bryuno set K_N(γ) to the whole parameter space W_N, with bounds of order ε uniform in γ. The proof of Proposition 4.5 is only a sketch: for a and r it cites [33, Corollary 10.8], while for u it says the McShane Theorem is applied to the coefficients u_{j,nu}^{(k)}(0;c,omega(zeta)) and 'reasoning in the same way'. The function u is not covered by the cited corollary; its extension with the g(s,alpha)-valued Lipschitz bound (4.11e) is a new ingredient of this paper. If U fails to be Lipschitz on W_N with constant C1|epsilon|, or if the extension does not preserve separate analyticity in c (needed for Corollary 4.9 and Lemma 5.4), then the fixed point problem (5.1) cannot be solved for all parameters, and Lemma 6.2's derivative estimate has no basis. The paper does not supply a self-contained verification of this critical extension step.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the one-dimensional NLS equation with a smooth convolution potential, a small quintic nonlinearity, and Gevrey initial data. The central claim, Theorem 2.3, is that for a set of potentials of asymptotically full measure as the nonlinearity parameter tends to zero, and for each such potential a set of initial data of asymptotically full measure, every initial datum in that set produces a unique global almost-periodic solution whose Gevrey norm stays within a factor of two of its initial value. The proof combines a Moser-type counterterm theorem imported from the companion preprint [33] with a new implicit function problem that treats both the potential and the initial datum as parameters, then performs measure estimates on the resonant sets via a weak Diophantine condition. The paper also proves a full-measure statement for potentials on a sequence of shrinking balls (Theorem 2.9), Lyapunov statistical stability (Theorem 2.11), and the existence of a Cantor foliation by invariant tori in regions where all Fourier modes of the initial datum are bounded below (Theorem 2.13).","tokens_in":41133,"tokens_out":10191,"duration_ms":101611,"significance":"If the proof is complete, this is a striking extension of finite-dimensional KAM measure statements to an infinite-dimensional Hamiltonian PDE: it does not merely construct isolated almost-periodic solutions, but shows that they occupy asymptotically full measure sets in both the parameter space of potentials and the phase space of initial data. The strategy of parametrizing invariant tori by initial data rather than actions or linear amplitudes is original and well matched to the infinite-dimensional setting, where the image of an immersed torus need not be a submanifold. The paper is also commendably quantitative: the use of product measures is explicit, the constants are tracked, the scaling covariance is exploited systematically, and the geometric statements about embedded tori are separated from statements about merely immersed tori. The main risk is not circularity or hidden fitting parameters; it is that a load-bearing Lipschitz-extension step, Proposition 4.5, is only sketched and rests on the companion preprint [33].","major_comments":[{"comment":"Proposition 4.5 is the load-bearing step I cannot verify as written. The proof for the extension U(c,ζ,ε) of u(c,ζ,ε) is one sentence: 'applying the McShane Theorem also to u_{j,ν}^{(k)}(c,ω(·)) and reasoning in the same way.' This is not covered by [33, Corollary 10.8], which concerns the extensions A and R. The McShane theorem gives a real-valued Lipschitz extension of each coefficient individually, but the paper needs a g(s,α)-valued extension satisfying the uniform bound (4.11e) on all of W_N, continuity with respect to the product topology, and, crucially, preservation of separate analyticity in c and \\bar c, which is used in Proposition 4.7, Corollary 4.9, and Lemma 5.4. These properties enter the fixed point equation (5.1), the Lipschitz bound on Δ in Lemma 5.4, and the derivative estimate in Lemma 6.2. If the extension does not preserve separate analyticity, the implicit function problem cannot be solved for all (V,W), and the measure estimates for the resonant sets lose their basis. I request a complete proof of this extension lemma, not a reference to an analogous argument.","section":"Section 4.2, Proposition 4.5"},{"comment":"The paper relies on the companion preprint [33] for the core existence and regularity machinery: Theorem 3.5 is [33, Theorem 2.13], Proposition 3.10 is [33, Proposition 2.21], and the Lipschitz bounds (4.11a)–(4.11d) are imported from [33, Proposition 2.27 and Corollary 10.8]. These results carry the small-divisor estimates, the tree-sum bounds, and the uniform radius of convergence in ε on the Bryuno sets K_N(γ). Since [33] is an unreviewed preprint, the present manuscript is not self-contained at the level needed to certify Theorem 2.3. I am not asserting that these statements are false; I am asserting that the referee cannot verify the central theorem without checking them. Please either provide complete proofs or appendices, or state explicitly which results in [33] are accepted and reproduce their exact hypotheses.","section":"Sections 3 and 4, imported estimates"},{"comment":"The Fubini step at the end of the proof is too terse. From Proposition 6.4 one has μ_{1/4,1/2}(V \\setminus Γ_N(γ)) ≤ C_* γ. To conclude that μ_{1,1/4}(G) → 1 and μ_{2,1/2}(T_V) → 1, one must choose the set G explicitly, for instance G = {V : μ_{2,1/2}(T_V^c) ≤ √γ}; Markov's inequality then gives μ_{1,1/4}(G^c) ≤ C_* √γ, which is asymptotically small. The sentence 'a direct application of Fubini's theorem ensures... measure proportional to 1−O(γ)' does not by itself control the measure of G, because Fubini only controls the average over V. This is a small but central quantification, and it should be written out.","section":"Section 6.2, proof of Theorem 2.3, part 2"},{"comment":"The last paragraph of the proof states that there exists a measurable set of potentials G(γ) 'of measure less than 1 − √(γ/γ_*)', but the claim in Theorem 2.11, item 2, is that μ_{1,1/4}((G(γ))^c) ≤ √(γ/γ_*), which means G(γ) has measure at least 1 − √(γ/γ_*). As written, the proof contradicts the theorem statement. This is likely a typo, but it should be corrected because it appears exactly at the point where the measure-theoretic conclusion is drawn.","section":"Section 7.2, proof of Theorem 2.11"}],"minor_comments":[{"comment":"The section title contains a typo: 'Reuslts' should be 'Results'.","section":"Section 7 title"},{"comment":"There is a typo in 'impliy'; it should read 'imply'.","section":"Remark 5.7"},{"comment":"The word 'arbritrarily' should be 'arbitrarily'.","section":"Remark 2.14"},{"comment":"The exponent notation ρ^{1/N+1} is ambiguous; it should be ρ^{1/(N+1)} to match the summability condition ρ^{1/2(N+1)} used a few lines later.","section":"Section 7.1, equation (7.6)"},{"comment":"The theorem statement says G has 'positive measure' and later says its measure is asymptotically full; this is fine, but the wording could be tightened to make clear that the positive-measure assertion is strengthened in the final sentence.","section":"Section 2.1, Theorem 2.3"}],"recommendation":"major_revision","confidential_remarks":"The central concern is verifiability rather than plausibility. Proposition 4.5 is the one point where the argument is genuinely new relative to [33] and also genuinely load-bearing, and its proof is compressed to a McShane-extension remark. I would like to see a complete proof of that proposition, including the preservation of separate analyticity in c and of the product-topology continuity, before recommending acceptance. In addition, the manuscript's dependence on the unpublished preprint [33] for the existence theory and for the Lipschitz bounds (4.11a)–(4.11d) should be resolved, either by publication of [33] or by inclusion of the necessary statements and proofs. These are reparable issues, so I do not recommend rejection, but they are too central for minor revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper claims a first-of-its-kind KAM-type theorem for a PDE: asymptotically full-measure sets of almost-periodic solutions for 1D NLS with convolution potential, plus Lyapunov statistical stability and a Cantor foliation. The core idea is neat: parametrize initial data by solutions of the linearized equation through a bi-Lipschitz map, then run Bourgain-style measure estimates on the good frequencies. What is genuinely new here is that the construction yields positive measure in the space of initial data and potentials, not just uncountably many special solutions. The proof of the new parts—bi-Lipschitz parametrization, implicit function problem, measure estimates, and torus embedding—is detailed and internally consistent. The authors also handle the full-measure-set version for potentials and state the stability consequence cleanly.\n\nThe main soft spot is the weight placed on the authors' unpublished preprint [33]. The existence theory and many quantitative bounds are quoted from there as black boxes. That is acceptable in a research announcement, but this is a full paper making strong claims; a referee needs to verify that the imported estimates do what the authors need. In particular, Proposition 4.5, which extends the initial-data shift u to a Lipschitz map U on all of W_N, is only sketched. The stress-test note is right that this extension is load-bearing for the measure estimates. The paper says it follows by McShane and 'reasoning in the same way' as [33], but u is not covered by the cited corollary. This is a real gap in exposition: it may be fixable, but it needs a detailed proof.\n\nTwo smaller overclaims: the abstract's 'first result' priority statement is too broad given [5] exists, and the 'unique solution' in Theorem 2.3 is only uniqueness within the constructed family, not unconditional uniqueness for those initial data. Both should be softened.\n\nIf the extension in Prop 4.5 checks out and the [33] dependencies are solid, this is a significant advance. As written, the paper is not self-contained and the sketched step is important enough that I would want a referee to chase it down before accepting.\n\nRecommendation: send to a serious referee, but ask specifically to verify Prop 4.5 and the Fubini applications in Theorem 2.9. The paper deserves referee time; it just needs tightening. I would bring it to reading group, though the technical depth is high.","headline":"Full-measure KAM result for NLS with a bi-Lipschitz trick, but the proof leans hard on an unpublished preprint and one key extension is only sketched.","tokens_in":41853,"tokens_out":2220,"would_cite":true,"duration_ms":22501,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["37K55","35B15","35Q55","35B20"],"pacs":[],"model":"deepseek-v4-flash","headline":"Most small NLS waves are almost-periodic for all time","keywords":["nonlinear Schrödinger equation","convolution potentials","almost-periodic solutions","infinite-dimensional Bryuno condition","KAM theory","Gevrey regularity","Cantor foliation","Lyapunov statistical stability"],"falsifier":"Fix a small $\\varepsilon$ and a finite-support amplitude $c$, and compute the quotient $\\|U(c,\\zeta_1,\\varepsilon)-U(c,\\zeta_2,\\varepsilon)\\|_{s,\\alpha}/\\|\\zeta_1-\\zeta_2\\|$ along a segment whose endpoints straddle a resonant divisor $\\omega\\cdot\\nu=0$. If the supremum of these quotients fails to remain $O(\\varepsilon)$ as $\\varepsilon\\to 0$, Proposition 4.5 is false and the measure bound of Lemma 6.2 breaks down; the same check applies to the extended counterterm components $A$ and $R$.","tokens_in":40643,"feed_emoji":"🌊","tokens_out":9269,"duration_ms":93922,"temperature":0.7,"pith_summary":"This paper studies the one-dimensional nonlinear Schrödinger equation with a smooth convolution potential and a small quintic nonlinearity, starting from Gevrey-regular data. It proves that, as the nonlinearity parameter tends to zero, the set of potentials and the set of initial data that give global almost-periodic solutions both have measure tending to one: for every such solution the Gevrey norm stays within a factor of two of its initial value for all time. The solutions are dense on invariant tori; where every Fourier mode of the initial datum is bounded away from zero, those tori are genuine submanifolds and form a Cantor foliation of the phase space. The result gives an infinite-dimensional counterpart to the classical finite-dimensional KAM picture in which most of phase space is filled with invariant tori.","feed_headline":"Most small NLS waves are almost-periodic for all time","feed_subtitle":"Invariant tori fill asymptotically full measure sets of potentials and Gevrey initial data as ε → 0.","key_machinery":"The proof runs through a Moser counterterm theorem: one solves the modified equation $iu_t + (\\omega_j+\\eta_j)u_j + \\varepsilon(|u|^4u)_j=0$ for a frequency $\\omega$ in a weak Bryuno class, then chooses $\\omega$ and the counterterm $\\eta$ so that $\\omega_j+\\eta_j=j^2+V_j$, matching the original equation, and finally matches the initial datum. The frequency parameter space is written as $\\zeta=(\\kappa,\\xi)\\in W_N$, with $\\omega_j(\\zeta)=j^2+\\kappa_0+\\sum_{q=2}^{N-1}\\kappa_q/j^q+\\xi_j$; Proposition 3.10 gives the counterterm the same asymptotic form in $j$, so the compatibility condition can be solved. The load-bearing device is Proposition 4.5, which extends the counterterm, its finite-dimensional coefficients, and the initial-data shift from the non-resonant set $K_N(\\gamma)$ to the whole parameter space $W_N$ as Lipschitz functions of order $\\varepsilon$. With this extension, the implicit function problem $c+U=W$, $\\kappa+A=0$, $\\xi+R=V$ is solved by contraction on Lipschitz functions, producing a bi-Lipschitz parametrization $(V,W)\\leftrightarrow(\\xi,c)$. The measure estimates then show that, for fixed $W$, the resonant set $|\\omega(V,W)\\cdot\\nu|\\le \\delta$ has measure at most $C\\langle i_0(\\nu)\\rangle^N\\delta$, using the Lipschitz constant below $1/2$, and sum these bounds over all resonance vectors $\\nu$.","core_discovery":"The central claim is Theorem 2.3: for the NLS equation with potential $V\\in \\ell^{N,\\infty}(\\mathbb{R})$ and initial datum $W$ in the Gevrey space $g(s,\\alpha)$, there is a set of potentials $G$ and, for each $V\\in G$, a set of initial data $\\mathcal{T}_V$ such that $\\mu_{1,1/4}(G)\\to 1$ and $\\mu_{2,1/2}(\\mathcal{T}_V)\\to 1$ as $\\varepsilon\\to 0$, and every $W\\in \\mathcal{T}_V$ has a unique global solution $u(x,t)$ that is almost-periodic in time and Gevrey in both variables. The frequency $\\omega(V,W)$ of the solution lies in a weak Bryuno class, and its hull is an invariant torus immersed in $g(s,\\alpha)$. When $W$ lies in the region $A_{V,\\delta}$, where every Fourier coefficient has weight at least $\\delta$, the torus is an embedded submanifold analytically homeomorphic to $\\mathbb{T}^{\\mathbb{Z}}$ and the family forms a Cantor foliation; the remaining tori are still immersed. A companion result, Theorem 2.9, shows that the set of potentials for which the good initial data have asymptotically full measure along a prescribed sequence is actually of full measure. The proof's key novelty is to parametrize the tori by initial data instead of actions, through a bi-Lipschitz map from linear solutions to data, which replaces the missing twist condition.","pith_inferences":["Reading Theorem 2.3 probabilistically, randomizing the Fourier coefficients of the potential and of the initial datum independently and uniformly makes almost-periodic solutions occur with probability tending to one; the paper notes but does not explore other distributions such as Gaussians.","The bi-Lipschitz parametrization by initial data is a twist-free substitute for the classical action-angle twist condition; this device may carry over to other semilinear PDEs on the circle with the same counterterm structure, provided an analogous Lipschitz-extension lemma holds.","The paper leaves open whether the full-measure-potential statement can be upgraded to a genuine limit as $\\rho\\to 0$ rather than a limit along a sequence, explicitly because of measurability issues; a positive answer would make statistical Lyapunov stability a statement about almost every small datum.","The picture suggests that in Gevrey regularity most trajectories are trapped on tori, so any energy transfer to high frequencies must be sought in spaces of lower regularity, a direction the paper itself points toward."],"forward_implications":["For each potential in the good set, the initial data with eternal almost-periodic dynamics occupy a fraction of the Gevrey ball tending to one as $\\varepsilon\\to 0$.","Every such solution satisfies $\\|\\mathcal{F}(u(\\cdot,t))\\|_{s,\\alpha}\\le 2\\|W\\|_{s,\\alpha}$ for all real $t$, so the zero solution is statistically Lyapunov stable in the Gevrey norm.","Initial data whose Fourier modes are all bounded away from zero sit on maximal invariant tori that are embedded submanifolds; those tori give a Cantor foliation of a large region of phase space, while the remaining tori are only immersed.","The frequencies obtained are weak Bryuno rather than Diophantine, which is what allows the surviving set to have asymptotically full measure rather than being a thin Diophantine slice.","By the scaling symmetry $(u,W,\\varepsilon)\\mapsto(\\lambda u,\\lambda W,\\lambda^{-4}\\varepsilon)$, the same conclusion transfers to any small-amplitude ball, with the good sets filling the ball in measure as its radius shrinks."],"supporting_citations":[{"why":"Supplies the Moser counterterm theorem, the tree expansion for the solution and counterterm, and most of the Lipschitz and extension bounds reused here.","marker":"[33]"},{"why":"Provides local well-posedness for the Cauchy problem and the measure-estimate template that Lemma 6.2 follows.","marker":"[8]"},{"why":"Gives the measure-sum bound used to control the resonant sets and hence the measure of the complement of the good parameters.","marker":"[21]"},{"why":"Defines the quintic NLS with external potential setting whose full-dimensional tori form the benchmark being extended.","marker":"[16]"},{"why":"Recent construction of almost-periodic solutions for NLS without external parameters, whose typicality in measure is established here.","marker":"[5]"},{"why":"Supplies the fixed-point lemma for Lipschitz-continuous functions used to solve the implicit function problem (5.1).","marker":"[10]"},{"why":"Provides the Cantor manifold and foliation framework adopted for the structure theorem on invariant tori.","marker":"[69]"},{"why":"An earlier almost-periodic construction whose lack of injectivity motivates the new bi-Lipschitz parametrization by initial data.","marker":"[77]"}],"fun_headline_variants":["Almost all small NLS waves are almost-periodic","Full-measure torus sets for almost-periodic NLS solutions","NLS phase space gets Cantor foliation from invariant tori","KAM theory for PDEs with persistence of many tori","Statistical stability of NLS origin via KAM tori"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The construction depends on extending the counterterm and the initial-data shift from the set of good (non-resonant) parameters to all parameters while preserving small Lipschitz bounds; this extension is only sketched and rests on a previous paper.","fun_headline_variants_meta":{"raw":{"variants":["Almost all small NLS waves are almost-periodic","Full-measure torus sets for almost-periodic NLS solutions","NLS phase space gets Cantor foliation from invariant tori","KAM theory for PDEs with persistence of many tori","Statistical stability of NLS origin via KAM tori"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000391,"raw_usage":{"total_tokens":2102,"prompt_tokens":1035,"completion_tokens":1067,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":651,"completion_tokens_details":{"reasoning_tokens":982}},"tokens_in":651,"tokens_out":1067,"duration_ms":10004,"temperature":1.0,"reasoning_tokens":982,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T23:14:10.436983+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Fix a small $\\varepsilon$ and a finite-support amplitude $c$, and compute the quotient $\\|U(c,\\zeta_1,\\varepsilon)-U(c,\\zeta_2,\\varepsilon)\\|_{s,\\alpha}/\\|\\zeta_1-\\zeta_2\\|$ along a segment whose endpoints straddle a resonant divisor $\\omega\\cdot\\nu=0$. If the supremum of these quotients fails to remain $O(\\varepsilon)$ as $\\varepsilon\\to 0$, Proposition 4.5 is false and the measure bound of Lemma 6.2 breaks down; the same check applies to the extended counterterm components $A$ and $R$.","supporting_citations":[{"cited_title":"Biasco, J.E","cited_arxiv_id":null,"evidence_quote":"Provides local well-posedness for the Cauchy problem and the measure-estimate template that Lemma 6.2 follows."},{"cited_title":"Bourgain, On invariant tori of full dimension for 1D periodic NLS , J","cited_arxiv_id":null,"evidence_quote":"Gives the measure-sum bound used to control the resonant sets and hence the measure of the complement of the good parameters."},{"cited_title":"Bourgain, On invariant tori of full dimension for 1D periodic NLS , J","cited_arxiv_id":null,"evidence_quote":"Defines the quintic NLS with external potential setting whose full-dimensional tori form the benchmark being extended."},{"cited_title":"Biasco, J.E","cited_arxiv_id":null,"evidence_quote":"Supplies the fixed-point lemma for Lipschitz-continuous functions used to solve the implicit function problem (5.1)."},{"cited_title":"Kuksin, J","cited_arxiv_id":null,"evidence_quote":"Provides the Cantor manifold and foliation framework adopted for the structure theorem on invariant tori."},{"cited_title":"P¨ oschel,On the construction of almost-periodic solutions for a nonl inear Schr¨ odinger equation, Ergodic Theory Dynam","cited_arxiv_id":null,"evidence_quote":"An earlier almost-periodic construction whose lack of injectivity motivates the new bi-Lipschitz parametrization by initial data."}],"review_version":1}