{"id":"badad901-adab-476e-b62d-560abe139e68","arxiv_id":"2607.25009","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Dual-scale hereditary Navier–Stokes is ill-posed for p < pc = N(1+α∞)/(1−α∞) by norm inflation, and globally well-posed for small high-frequency-adherent data in Ḃ^{−κ}_{∞,∞} with κ=(1−α∞)/(1+α∞).","lead":"A viscoelastic Navier-Stokes model with dual-scale memory has a sharp critical Lebesgue index and is ill-posed below it via norm inflation, yet globally well-posed for small data in a critical Besov space broader than the usual little-Besov closure. The memory's temporal smoothing is claimed to block the classical Bourgain–Pavlović cascade.","discovery_kind":"extension","skeptic_critique":{"model":"moonshotai/kimi-k3","headline":"Lemma 6.2's low-frequency (j→−∞) bounds are not uniform in time and omit the low–low paraproduct channel; the standard repair forces α₀≥α∞, which the flagship dual-scale kernels (Cole–Cole α₀=0, Prabhakar α₀<α∞) violate — Theorem 6.3 is unproven for its motivating kernel class.","rationale":"Good-faith reading: the high-frequency architecture of §6 checks out. I verified the scaling identities (1−2/(1+α∞)=−κ, 2γ/(1+α∞)=κ+ϵ), the paraproduct and residue bounds (6.3)–(6.4) with the single s^{−γ} allocation (ϵ>κ makes the residue sum converge), the Mittag-Leffler convolution F(τ′) bounded with (τ′)^{−γ} decay, the linear adherence argument in Theorem 6.3, and the uniqueness/localization argument. The p_c threshold and the norm-inflation proof are internally consistent. The soft spot is isolated and specific: the j→−∞ output of the bilinear form. The paper's own bound there (2^{j(1−κ)}t^{1−γ}) is visibly not uniform in t, contradicting the sup_{t>0} claim of Lemma 6.2, and the low–low paraproduct channel is simply absent from the proof. When one performs the omitted estimate with the (H3)-based resolvent lifetime, the j-exponent is 2/(1+α∞)−2/(1+α₀), which closes iff α₀≥α∞. Theorem 6.3 states no such restriction, and Examples 3–4 (the physically motivated \"terminal relaxation\" kernels) have α₀<α∞. Possible author responses: (i) add the hypothesis α₀≥α∞ and rewrite the low-frequency proof using the lifetime bound (pure fractional then closes with borderline exponent 0); (ii) supply a sharper low-frequency estimate that beats the lifetime scaling; (iii) if the bound is saturated, the theorem is false as stated for α₀<α∞. Caveat: I verified the estimate algebra, not its saturation — the concrete test distinguishes a lossy estimate from a genuine obstruction. Since the headline result (global well-posedness beyond little Besov) rests entirely on Lemma 6.2 and the gap is presently unfilled for the advertised kernel class, I move the reader's CONDITIONAL to REJECT as written, with a clear repair path that would restore a restricted (α₀≥α∞) version. I agree only partially with the reader's weakest_assumption: the imported Lemma 4.1(ii) is a real dependency, but the more central vulnerability is internal to §6.","tokens_in":28321,"tokens_out":29032,"duration_ms":1475374,"concrete_test":"Re-derive the j→−∞ bounds of Lemma 6.2 including the omitted low–low paraproduct channel: for data saturating ‖Δ_m u‖_∞=2^{mκ}, bound 2^{−jκ}‖Δ_j∫₀^t S(t−s)P∇·(S_{j−1}u Δ_jv)ds‖ using ∫₀^∞|E_{1+α₀}(−c2^{2j}τ^{1+α₀})|dτ ~ 2^{−2j/(1+α₀)}, and verify the j-exponent equals 2/(1+α∞)−2/(1+α₀). Run it on two kernels: pure fractional ĝ=λ^{−α} (α₀=α∞: exponent 0, closes) and Cole–Cole ĝ=(λ^{1/2}+γ)^{−1}, N=3 (α₀=0, κ=1/3: exponent 1+κ−2=−2/3, diverges). If divergence is confirmed, test saturation: take u₀=Σ_{m≤j}2^{mκ}ψ_m with aligned-sign dyadic bumps and evaluate ‖B(u₀,u₀)(t)‖_{Ḃ^{−κ}} at t~2^{−2j}; growth in j refutes Lemma 6.2 for α₀<α∞ and forces adding the hypothesis α₀≥α∞; absence of growth means the estimate is lossy and a sharper low-frequency argument must be supplied.","verdict_should_be":"REJECT","load_bearing_attack":"§6, proof of Lemma 6.2, low-frequency step (p. 22). Two problems. (a) For j<0 the paper estimates only the high-high-into-low channel R_j, uses Ŝ≈1, and obtains 2^{−jκ}‖Δ_jB(t)‖ ≲ 2^{j(1−κ)}t^{1−γ}. This grows like t^{1−γ}, so the claimed sup_{t>0}‖B(u,v)(t)‖_{Ḃ^{−κ}} ≤ C‖u‖_X‖v‖_X — needed for the global contraction in X — is not established. (b) More seriously, the low–low paraproduct channel Δ_j(T_uv) for j→−∞ is never estimated at all. The natural repair uses the (H3) low-frequency resolvent profile Ŝ(t,ξ)~E_{1+α₀}(−c|ξ|²t^{1+α₀}) (the one the paper itself invokes in Lemma 4.2 and in the K_low step of Theorem 6.3). With ‖S_{j−1}u‖≲2^{jκ}‖u‖, ‖Δ_jv‖≲2^{jκ}‖v‖ (no t^γ weight is available on low blocks), gradient factor 2^j, and resolvent lifetime ∫₀^∞|E_{1+α₀}(−c2^{2j}τ^{1+α₀})|dτ ~ 2^{−2j/(1+α₀)}, one gets 2^{−jκ}‖Δ_jB‖ ≲ 2^{j[1+κ−2/(1+α₀)]} = 2^{j[2/(1+α∞)−2/(1+α₀)]}, using 1+κ = 2/(1+α∞). This exponent is 0 exactly when α₀=α∞ (pure fractional, Ex. 1: borderline-bounded, salvageable), positive when α₀>α∞ (Ex. 2), but strictly negative whenever α₀<α∞ — i.e., precisely for the Cole–Cole retardation kernel (Ex. 3, α₀=0) and the multi-scale Prabhakar kernel (Ex. 4), the examples the introduction uses to motivate \"dual-scale\" memory with terminal Newtonian relaxation. For those kernels the bilinear low-frequency output is not controlled in Ḃ^{−κ}_{∞,∞}, so Lemma 6.2, and hence Theorem 6.3, is unproven for the general admissible class. The failure is structural: κ is tuned to the short-time smoothing α∞, while low-mode temporal decay is governed by α₀ — exactly where broken scale invariance bites. No choice of ϵ∈(κ,1) repairs this channel, since ϵ never enters it. This is independent of the reader's flagged dependence on [12]: even granting Lemmas 4.1–4.2 wholesale, the §6 low-frequency analysis has this gap.","agreement_with_reader":"partial"},"referee_report":{"model":"moonshotai/kimi-k3","summary":"The paper studies a Navier–Stokes system in which instantaneous viscosity is replaced by convolution against a 'dual-scale' memory kernel whose Laplace transform decays like |λ|^{-α∞} at high frequency and |λ|^{-α₀} at low frequency (hypotheses H1–H3). Three main results are claimed: (i) sharp short-time L^q–L^p estimates for the resolvent and its gradient, identifying a critical Lebesgue exponent p_c=N(1+α∞)/(1−α∞) (Lemma 4.1, Theorem 5.2); (ii) instantaneous norm inflation of the second Picard iterate, hence failure of uniform continuity of the flow map at the origin, for 1<p<p_c, proved by a frequency-modulation argument using compactly supported spectra and Bernstein inequalities in place of Hausdorff–Young (Theorem 5.3); and (iii) global Hadamard well-posedness for small divergence-free data in Ḃ^{−κ}_{∞,∞}, κ=(1−α∞)/(1+α∞), satisfying a one-sided high-frequency adherence condition — a class strictly broader than the little Besov closure — via an asymmetric paraproduct estimate in a temporally weighted path space (Lemma 6.2, Theorem 6.3).","tokens_in":28956,"tokens_out":15767,"duration_ms":872313,"significance":"If correct, the results cleanly delineate the well-posedness/ill-posedness boundary for a physically motivated class of hereditary-viscosity fluids. Strengths worth naming: the critical index p_c is derived parameter-free from the integrability condition ˜γ_{p/2,p}<1; the inflation argument bypasses Hausdorff–Young via compact spectra and Bernstein equivalence and makes a falsifiable prediction (inflation precisely for p<p_c=N/κ); the critical-regime Kato-space contraction (Thm 5.2(ii)) checks out by exact Beta-function bookkeeping; and the asymmetric paraproduct idea for taming the high-high cascade without non-integrable s^{-2γ} weights is a genuine technical contribution. However, the significance of the flagship Besov result is currently conditional: as written, Theorem 6.3 appears provable (by the natural repair) only under α₀≥α∞, which excludes the Cole–Cole and Prabhakar kernels used to motivate the entire 'dual-scale' program.","major_comments":[{"comment":"§6, proof of Lemma 6.2, low-frequency step (p. 22). Two defects. (a) For j<0 the paper estimates only the high-high-into-low channel R_j, uses Ŝ≈1, and obtains 2^{-jκ}‖Δ_jB(t)‖ ≲ 2^{j(1-κ)}t^{1-γ}; since t^{1-γ}→∞ as t→∞, the claimed sup_{t>0}‖B(t)‖_{Ḃ^{-κ}}≤C‖u‖_X‖v‖_X is not established (the lemma statement requires the sup over all t>0). (b) The low-low paraproduct channel Δ_j(T_uv) for j→−∞ is never estimated. Its natural bound — ‖S_{j-1}u‖≲2^{jκ}‖u‖, ‖Δ_jv‖≲2^{jκ}‖v‖ (no t^γ weight exists on low blocks), gradient 2^j, resolvent lifetime ∫₀^∞|E_{1+α₀}(-c2^{2j}τ^{1+α₀})|dτ∼2^{-2j/(1+α₀)} — gives 2^{-jκ}‖Δ_jB‖≲2^{j[2/(1+α∞)-2/(1+α₀)]}. This is borderline for α₀=α∞ (Ex. 1), decays for α₀>α∞ (Ex. 2), but diverges as j→−∞ whenever α₀<α∞ — precisely the Cole–Cole kernel (Ex. 3, α₀=0) and Prabhakar kernel (Ex. 4) that the introduction advances as the motivation for dual-scale memory. Theore","section":"§6, Lemma 6.2 / Theorem 6.3"},{"comment":"The load-bearing linear estimates are imported from the author's concurrent unrefereed preprint [12] (arXiv:2607.17430): Lemma 4.1(i), the S^{-2}_{1,0} membership of the scaled resolvent symbol [12, Lemma 3.1] used in the proof of Lemma 4.1(ii), the low-frequency dyadic partition [12, Lemma 3.4] used in Lemma 4.2, the Mittag-Leffler profile bound on annuli C_j used in Lemma 6.2 and (6.6), and the uniform L^q bounds for the modulated operators K(s,λ) [12, Theorem 4.2] used in Theorem 5.3. Since p_c, the contraction, and the inflation argument all collapse if the multiplier-class membership or the temporal penalties fail, the manuscript is not currently verifiable on its own. At minimum the imported results should be stated precisely (with hypotheses) in an appendix; ideally the key multiplier estimates should be proved here.","section":"§4, Lemma 4.1; dependence on [12]"}],"minor_comments":[{"comment":"Abstract and §5.2: 'confirms intrinsic ill-posedness' overstates Theorem 5.3, which proves failure of uniform continuity of the data-to-solution map at the origin via norm inflation of the second Picard iterate. Please align the wording with the statement.","section":"Abstract / Theorem 5.3"},{"comment":"p. 17: 'Recall the dynamic scaling invariance of the high-frequency system' is misleading — the dual-scale system has no exact scale invariance (this is the paper's premise); only the high-frequency asymptotic profile is self-similar. Please rephrase.","section":"§5.2"},{"comment":"Lemma 4.1(ii), p. 9: the claim that an S^{-1}_{1,0} symbol yields a convolution kernel with singularity O(|w|^{-(N-1)}) should carry a precise reference (e.g., Grafakos [15] or Hörmander) and an explicit statement that the kernel bounds are uniform in t∈(0,T] for the scaled family.","section":"§4, Lemma 4.1"},{"comment":"Notation: κ denotes the Besov index throughout but is reintroduced in §5.2 as 'the spatial scaling factor'; σ is used both for the temporal scaling modulation and for the symbol σ(P) of the Leray projector. Suggest renaming.","section":"§5.2, §3.5"},{"comment":"Remark 6.4 uses the space ḃ^{-κ,+}_{∞,∞}, which is never defined (only ḃ^{ε,+}_{∞,∞} is). Please define or correct.","section":"§6, Remark 6.4"},{"comment":"Theorem 6.3's statement should list which of (H1)-(H3) are assumed; as written it begins 'Let N≥2' with the hypotheses implicit, while the proof uses (H3) essentially in the K_low step.","section":"§6, Theorem 6.3"},{"comment":"To substantiate 'strictly broader than ˙b^{-κ}_{∞,∞}', an explicit example of u_0 satisfying (6.5) but failing the low-frequency adherence lim_{j→-∞}2^{-jκ}‖Δ_ju_0‖_∞=0 would help the reader.","section":"§6, Theorem 6.3"},{"comment":"p. 22, low-frequency step of Lemma 6.2: the assertion 'the resolvent lacks high-frequency decay within this regime (Ŝ≈1)' needs quantification via (H3) (regime of validity in |ξ|²t^{1+α₀}), especially if the t^{1-γ} bound is to be repaired.","section":"§6, Lemma 6.2"}],"recommendation":"major_revision","confidential_remarks":"The load-bearing linear theory is imported from the author's own concurrent, unrefereed preprint [12], and the reference list leans heavily on the author's recent output ([9]-[13]). This is not improper, but it means the paper's correctness cannot be fully assessed from the published literature; the editor may wish to have [12] refereed in tandem or to require an appendix. The manuscript also declares Gemini assistance for English; the prose is occasionally inflated (\"deeply counterintuitive\", \"elegantly\") but this does not affect the mathematics."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The two things worth knowing: de Andrade gets a clean critical Lebesgue threshold p_c = N(1+α_∞)/(1-α_∞) for dual-scale hereditary NS and a carefully written supercritical norm-inflation argument that works in Fourier space with Bernstein, avoiding the usual localization headaches. The bigger claim—global small-data Hadamard well-posedness in Ḃ^{-κ}_{∞,∞} under only high-frequency adherence, strictly larger than little Besov—is where the paper is trying to beat Bourgain–Pavlović via temporal memory, and that part does not currently close.\n\nWhat is new and done well: the dual-scale kernel that breaks global self-similarity, the explicit p_c(α_∞), and the inflation theorem (compact spectra, modulated convergence to the fractional resolvent). The short-time critical-index derivation from γ̃_{p/2,p}<1 is standard and clean. The asymmetric paraproduct idea in the high-frequency regime is a reasonable attempt to keep the Duhamel integral integrable. Linear L^q–L^p setup is organized, though the gradient estimates lean hard on the concurrent self-cited preprint [12].\n\nSoft spots in proportion. Dependence on unfinished [12] is real but secondary. The load-bearing issue is Lemma 6.2’s low-frequency step. For j<0 the paper only treats the high-high residue, uses Ŝ≈1, and obtains a factor t^{1-γ}. That grows in time, so the claimed uniform bound sup_t ‖B‖_{Ḃ^{-κ}} fails and the global contraction in X is not established. The low–low paraproduct channel is simply missing. Repairing it with the paper’s own low-frequency resolvent (H3) produces an exponent controlled by α_0 versus α_∞; it is non-positive precisely when α_0<α_∞—i.e., for the Cole–Cole and Prabhakar examples that motivate “dual-scale” in the introduction. So Theorem 6.3 is unproven for the flagship kernels. High-frequency analysis and the inflation theorem are not damaged by this.\n\nThis is for people who already work on hereditary or fractional NS and critical Besov spaces. It deserves a serious referee: the supercritical half is publishable material and the Besov claim is interesting enough to force a fix. I would not cite the global Besov statement until the low-frequency bilinear estimate is repaired; I would engage the p_c/inflation part. Send it out.","headline":"Sharp critical index and supercritical inflation look solid; the headline Besov well-posedness has a real low-frequency gap that hits exactly the dual-scale kernels the paper advertises.","tokens_in":29485,"tokens_out":637,"would_cite":false,"duration_ms":26209,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q30","35R09","76D05","42B25"],"pacs":[],"model":"grok-4.5","headline":"Dual-scale memory viscosity makes the Navier-Stokes equations well-posed in a critical Besov space larger than the classical little-Besov closure, while remaining ill-posed below a sharp Lebesgue threshold.","keywords":["Navier-Stokes equations","dual-scale memory","Besov spaces","norm inflation","para-differential calculus","anomalous diffusion","hereditary viscosity","critical spaces"],"falsifier":"Construct an explicit dual-scale kernel satisfying the stated high- and low-frequency asymptotics for which either the gradient resolvent fails the claimed L^q-L^p bound, or a sequence of high-frequency-adherent data of arbitrarily small Ẋ^{-κ}_{∞,∞} norm produces a mild solution that leaves every bounded set of that space in arbitrarily short time.","tokens_in":28847,"feed_emoji":"〜","tokens_out":1257,"duration_ms":19872,"temperature":0.7,"pith_summary":"Classical Navier-Stokes is known to be ill-posed in the largest scale-critical Besov space because high-frequency convective interactions can instantly dump energy into macroscopic modes. This paper replaces ordinary viscosity by a dual-scale hereditary memory kernel that encodes short-time elastic response and long-time fluid relaxation. The non-local dissipation breaks exact scaling, so the linear resolvent must be treated as a pseudo-differential operator whose spatial gradient exacts a fractional temporal penalty. The resulting critical Lebesgue index is p_c = N(1+α_∞)/(1−α_∞). Below that index the bilinear Picard iterate inflates instantly, proving ill-posedness. At the opposite extreme p\to∞ the same memory damps the high-high resonant cascade strongly enough that small divergence-free data with high-frequency adherence produce unique global mild solutions in the critical space Ẋ^{-κ}_{∞,∞}, a regime strictly larger than the separable little-Besov subspace. The result therefore maps the precise topological boundary between well-posedness and collapse for this class of viscoelastic fluids.","feed_headline":"Memory viscosity tames critical Besov collapse for fluids","feed_subtitle":"Dual-scale hereditary kernels restore global well-posedness beyond the little-Besov space while keeping a sharp ill-posedness threshold","key_machinery":"Asymmetric interpolation inside Bony’s paraproduct decomposition: one factor of the high-high residue is controlled by a fractional time weight t^γ while the other is controlled only by the base Besov norm. The resulting temporal singularity remains integrable, so the hereditary smoothing overpowers the convective cascade before it can inflate low-frequency modes.","core_discovery":"For the incompressible Navier-Stokes system driven by a dual-scale admissible memory kernel, the Cauchy problem is globally well-posed in the Hadamard sense for small divergence-free data in the critical Besov space Ẋ^{-κ}_{∞,∞}(ℝ^N) that satisfy the high-frequency adherence condition lim_{j\to+∞} 2^{-jκ}‖Δ_j u_0‖_{L^∞}=0, where κ=(1-α_∞)/(1+α_∞). This regime properly contains the little-Besov closure. At the same time, for every Lebesgue exponent 1<p<p_c with p_c=N(1+α_∞)/(1-α_∞) the data-to-solution map fails to be uniformly continuous at the origin by instantaneous norm inflation of the second Picard iterate.","pith_inferences":["The same asymmetric-paraproduct device may apply verbatim to other hereditary or fractional parabolic systems whose linear symbols lie in S^{-2}_{1,0}.","If the open question on the non-adherent complement is settled negatively, dual-scale memory would give a complete topological dichotomy for this class of viscoelastic models.","Rheological measurements that fix the short-time exponent α_∞ would immediately translate into a concrete numerical value of the critical integrability threshold p_c for laboratory fluids."],"forward_implications":["The classical Bourgain-Pavlović collapse is not universal for every non-local fluid model; dual-scale memory can suppress the low-frequency resonant cascade.","Well-posedness holds in a strictly larger set than the little-Besov space, so data whose high-frequency dyadic tail merely tends to zero (rather than belonging to the separable closure) are admissible.","The critical Lebesgue threshold p_c recovers the classical Kato space L^N when the short-time anomaly α_∞ vanishes and diverges as the fluid becomes strongly elastic.","Ill-posedness is confined to the non-separable high-frequency tail; any further extension of well-posedness must confront lacunary data that violate high-frequency adherence."],"fun_headline_variants":["Dual-scale memory restores global well-posedness beyond little Besov","Hereditary viscosity blocks critical Besov collapse for Navier-Stokes","Supercritical norm inflation meets broader Besov well-posedness","Dual-scale kernels confine ill-posedness to high-frequency Besov tail","Memory viscosity yields Hadamard well-posedness in critical Besov"],"cache_read_input_tokens":16512,"weakest_assumption_plain":"The short-time linear estimate that the gradient of the dual-scale resolvent maps L^q into L^p with a precise fractional time decay whenever 1/q-1/p is less than 1/N; if that decay rate is false, both the critical index and the contraction argument collapse.","fun_headline_variants_meta":{"raw":{"variants":["Dual-scale memory restores global well-posedness beyond little Besov","Hereditary viscosity blocks critical Besov collapse for Navier-Stokes","Supercritical norm inflation meets broader Besov well-posedness","Dual-scale kernels confine ill-posedness to high-frequency Besov tail","Memory viscosity yields Hadamard well-posedness in critical Besov"]},"model":"grok-4.5","effort":"low","cost_usd":0.006717,"raw_usage":{"total_tokens":1884,"prompt_tokens":1038,"num_sources_used":0,"completion_tokens":98,"cost_in_usd_ticks":67168000,"prompt_tokens_details":{"text_tokens":1038,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":748,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":1038,"tokens_out":98,"duration_ms":12608,"temperature":1.0,"reasoning_tokens":748,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-31T03:42:36.680105+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Construct an explicit dual-scale kernel satisfying the stated high- and low-frequency asymptotics for which either the gradient resolvent fails the claimed L^q-L^p bound, or a sequence of high-frequency-adherent data of arbitrarily small Ẋ^{-κ}_{∞,∞} norm produces a mild solution that leaves every bounded set of that space in arbitrarily short time.","supporting_citations":[],"review_version":1}