{"id":"c9d2fb7b-bb7e-44b0-be65-888bd29f0c66","arxiv_id":"2508.06333","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Arbitrarily small supercritical H^s vorticities for 3D Euler can have a unique classical solution whose Sobolev regularity drops continuously from s at the sharp rate (s-ct)/(1+ct).","lead":"This paper constructs solutions of the 3D Euler equations whose vorticity starts in a moderately regular space H^s (any s below 3/2) but instantly and continuously loses regularity: at each time t the solution sits exactly in H^{(s-ct)/(1+ct)} and in no smoother Sobolev space. The solutions are unique in a natural class, so the loss is a property of the true flow, and it gives a sharp, explicit regularity curve.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 13 Step 4 never proves the high-order W estimates needed for the full H^β range of (3.68); the H^β error bound for β near 3/2 is therefore unsubstantiated, and Theorem 1's growth in (1.4) rests on an unverified induction.","rationale":"Read in good faith, the construction is coherent: a stable high-frequency background, hyperbolic stretching, two growth mechanisms, an infinite gluing argument, and a uniqueness proof. I found no contradiction in the exponent bookkeeping, and the parameter hierarchy (a>(3-2s)/s, B large, γ small, T small) plausibly closes. The display errors noted by the reader are repairable. The single load-bearing concern is not the construction itself but the unproved high-derivative closure of the perturbation error. Because Theorem 3's threshold is obtained by summing infinitely many blocks whose individual H^β growth is taken from (1.4)/(3.58), any hidden extra eλ factor in the error term for β close to s would shift the per-block growth and could close the bootstrap before T. This is exactly the kind of technical premise that can be checked by a finite computation. I therefore keep the reader's conditional verdict: if the high-order estimates close, the result likely stands; until then it is not fully verified.","tokens_in":37324,"tokens_out":16393,"duration_ms":171732,"concrete_test":"Carry out the missing induction for ∥D^3W∥_{L^2} in Step 4 of Prop. 13: differentiate (3.80) three times and bound each resulting term using (3.57), Lemma 12, Steps 1–3, and the elliptic equation for Q, tracking all powers of eλ, eN, λ, N, logN. Verify that the worst term is at most C(eλ eN)^{β-s} λ^{-c*} N^{ct}(logN)^c for β near s, so no extra positive power of eλ or eN enters. If a term such as D^3(v[ψ]·∇v[ψ]) forces a factor eλ^{5/2-s}eN^{3-s}λ^c, the H^β error bound (3.68) fails and Theorem 1's norm-inflation exponent is not justified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing gap is the derivation of (3.68) for β∈[0,s], s<3/2. In Step 4 of Prop. 13, only ∥W∥_{L^2} and ∥∇W∥_{L^2} are estimated; the text then says 'Estimating higher integer order L2 Sobolev norms are analogous' and that general H^β for W follows by interpolation. But Ψ=curl W, so ∥Ψ∥_{H^β}≤C∥W∥_{H^{β+1}}. For β arbitrarily close to 3/2 this requires W∈H^{5/2-}, in particular control of ∥D^3W∥_{L^2}. The Cauchy estimates (3.57) show ∥D^k v[ψ]∥_∞ grows like eλ^{3/2+k-s}eN^{k-s}N^{ckt}logN; for k=2,3 these are positive powers of eλ (e.g. for s=1, k=3: λ^{B(5/2+2a)}), not the small factor eλ^{3/2-s}eN^{-s} used to close the L^∞/C^γ bootstrap. The missing induction must show that after three derivatives of (3.80), every term—especially D^3(v[ψ]·∇v[ψ]), D^3(v[ψ]·∇W), D^3(W·∇v[ψ]), and D^3∇Q—is controlled by the claimed RHS of (3.68) with only harmless powers N^{ct} and λ^c. No such calculation is given. The same pattern appears in Prop. 5's H^{β'} estimate ('can be done along similar lines'). This is not a disagreement with the result; it is a precise, checkable technical premise on which Theorem 1 and hence the infinite-block threshold (4.18) depend.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs divergence-free initial vorticities in 3D incompressible Euler with arbitrarily small H^s norm, for any s<3/2, whose unique classical solution loses Sobolev regularity continuously in time: at time t it lies in H^{(s-ct)/(1+ct)} and in no higher H^β. The proof has two ingredients. First, a norm-inflation building block (Theorem 1) decomposes the solution into a near-steady, oscillatory background φ and a perturbation ψ; a hyperbolic flow generated by the background stretches ψ so that ∥ψ∥_{H^β} grows like N^{t(1+β)/2}(eλ eN)^{β-s}. Error estimates (Propositions 5 and 13) show that the exact solution remains close to this pseudosolution on a fixed time interval. Second, an infinite disjoint sum of such blocks with rapidly increasing λ_j and suitable K_j yields the continuous loss curve, while a gluing and H^2 uniqueness argument selects the unique classical solution.","tokens_in":37797,"tokens_out":14196,"duration_ms":143372,"significance":"If the technical estimates close, this is a landmark result: it gives instantaneous and continuous-in-time loss of supercritical Sobolev regularity for the 3D Euler equations, with uniqueness in a natural classical class. The construction is highly nontrivial: the stable oscillatory background, the hyperbolic stretching mechanism, the Riemann-Lebesgue-type gains of N^{-1} and λ^{-c_*}, and the final threshold derivation from exponent bookkeeping are original and plausible. The paper also provides a clear conceptual explanation of the two mechanisms (transport and stretching) that produce the numerator and denominator in (s-ct)/(1+ct). The main weakness is that several load-bearing error estimates are only sketched, especially the high-order Sobolev estimates needed to reach β close to 3/2. Those gaps are technical in nature, but they are essential for the proof as written.","major_comments":[{"comment":"The H^β bound on Ψ is not proved for the full range β∈[0,s]. The proof estimates only ∥W∥_{L^2} and ∥∇W∥_{L^2}, and then says that higher integer order L^2 Sobolev norms are 'analogous' and that interpolation gives general H^β. Since Ψ=curl W, ∥Ψ∥_{H^β}≤C∥W∥_{H^{β+1}}; for β close to 3/2 this requires control of ∥D^3W∥_{L^2}. From (3.57), ∥ψ∥_{C^3} ~ eλ^{9/2-s}eN^{3-s}N^{ct}, which is a positive power of eλ (for s=1, eλ^{7/2}eN^2), and D^3(v[ψ]·∇v[ψ]) in (3.80) is not obviously small. No explicit induction is given to show that after three derivatives every term is controlled by the claimed right-hand side of (3.68). This gap is load-bearing because (3.68) is used to obtain the ψ-growth in Theorem 1 and hence the threshold in (4.18).","section":"§3.2.2, Proposition 13, Step 4 (Eq. (3.68))"},{"comment":"The H^{β'} estimate for the background error Φ is only sketched. After an L^2 estimate, the text says the H^1 and H^2 estimates follow by differentiating and 'repeat similar estimates', with a brief comment about the difficult term D^2v[Φ]·∇Φ and a Gagliardo-Nirenberg bound. The other terms arising from two derivatives of (3.15) are not listed; in particular the terms involving D^2v[φ]·∇Φ and D^2(v[φ]-v[φ])·∇Φ need a careful comparison with the claimed factor (λN)^{β'-s}N^{-1+ε}. Since ∥Φ∥_{H^{β'}} enters directly into the background bound ∥φ∥_{H^β}∼K^{-1}λ^{c_2(β-s)} in Theorem 1, an unchecked positive power of λN here would invalidate the background estimate. The proof must be supplied in full, not just indicated.","section":"§3.1, Proposition 5 (Eq. (3.25))"},{"comment":"The displayed K-scaling in the ψ-H^β estimate is inconsistent with the construction. Eq. (1.4) states ∥ψ(·,t)∥_{H^β}∼K^{-t(1+β)/s}λ^{...}, which at t=0 and β=s gives an O(1) quantity independent of K. However the initial data (3.42) carry an explicit factor K^{-1}, and (3.58) (written for K=1) together with (3.3) implies ∥ψ(0)∥_{H^s}∼K^{-1}. The same omission appears in Eq. (4.18). The missing factor K^{-1} does not change the λ-based convergence criterion, but it is part of the statement of Theorem 1 and is needed for the displayed initial-smallness mechanism in Theorem 3. The formula should read (up to logs) K^{-1-t(1+β)/s} or an equivalent corrected expression.","section":"Theorem 1, Eq. (1.4) vs. §3.2, (3.42), (3.58)"},{"comment":"The bound ∥ω∥_{C^k}≲λ^{ck} is asserted to follow from an induction 'similar' to Lemma 9, with the claim that it is simpler. This estimate is used to obtain the H^5 bound in Theorem 1 and in the gluing bootstrap (4.13). Since ω=φ+ψ and the flow is nonlinear, the induction requires controlling composition with the full particle map of v[ω], not just v[φ]. No details are provided. Please give the induction and state explicitly which constants are independent of λ and K.","section":"§3.2.3, Eq. (3.88)"}],"minor_comments":[{"comment":"In the proof, 'det ∇η = 0, due to incompressibility' should be 'det ∇η = 1'. The subsequent formula uses the change of variables, so this is a typo.","section":"§3.2, Lemma 12, proof"},{"comment":"The condition '(2+a)γ<1 <0' contains a typo; presumably it should be '(2+a)γ<1'.","section":"Eq. (3.65)"},{"comment":"The definition of K_j is ambiguous in the text 'Kj := 2j ε'. Please write K_j=2^{-j}ε or K_j=2^j ε explicitly, and check consistency with the requirement that ∥ω0∥_{H^s}≤ε.","section":"§4, after Eq. (4.2)"},{"comment":"The log factors are displayed inconsistently: Theorem 15 uses (log N)^{-t(1+β)/s}, while (1.4) and (4.18) use (log λ)^{-t(1+β)/s}. Since N and λ are related by (3.3), the precise relation should be stated once and used consistently.","section":"Theorem 15 vs. Eq. (1.4)"}],"recommendation":"major_revision","confidential_remarks":"The central construction is original and very likely correct in spirit, and the exponent bookkeeping is coherent. The major obstacles are omitted high-order estimates, not a detected contradiction in the mechanism. I therefore recommend major revision rather than rejection: the authors should supply complete proofs for Proposition 13 Step 4, Proposition 5's H^{β'} bound, and (3.88), and fix the K-scaling in Theorem 1. If those estimates close, the paper would be a strong addition to the literature."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The new thing is Theorem 3: for every s in (0,3/2) you get a divergence-free H^s vorticity, arbitrarily small, whose unique classical solution lies in H^{(s-ct)/(1+ct)} at each time and in no higher space. That is genuinely stronger than Luo's norm inflation and Bourgain–Li's endpoint result, and it caps the program from the 2D and SQG papers. The mechanism is coherent: an oscillatory background creates a hyperbolic stretching, the perturbation is tuned to align with the stretching and its frequency beats any power of the background, and the loss exponent comes from balancing exponents, not from fitting parameters. I checked the bookkeeping; the threshold formula is internally consistent.\n\nThe soft spot is real. Prop. 13, Step 4 estimates W in L^2 and H^1, then says higher integer norms are analogous and interpolates to H^β for β up to s. But Ψ=curl W, so the claimed H^β bound on Ψ for β near 3/2 requires W in H^{5/2-}, which means controlling ∥D^3 W∥_{L^2}. That is precisely where the small factors from the Riemann-Lebesgue cancellations have to survive three derivatives of the nonlinear terms in (3.80). The stress-test note names the dangerous terms—D^3(v[ψ]·∇v[ψ]), D^3(W·∇v[ψ]), and the pressure—and by dimensional counting they look like they could carry extra powers of λ or N. No calculation is given to show they don't. Same pattern in Prop. 5's H^β' bound. I could not verify or refute this in the time I gave it; it is a checkable technical premise, not a conceptual objection.\n\nThere are also a few display errors: (3.90) asserts a false inclusion, Lemma 11 has an exponent typo, and (3.65) contains '1<0'. The uniqueness class in the abstract differs slightly from the one in Section 4. All repairable, but they tell you the manuscript is not final.\n\nThe citation pattern is fine. The SQG adaptation is legitimate; the 3D problem has genuinely new difficulties (stretching, divergence constraint, non-scalar transport). The parameters a, B, γ, α, c* are chosen to satisfy constraints, not to hit the target exponent.\n\nWho gets value: anyone working on Euler ill-posedness. I would send this to a serious referee, with the instruction that the referee must actually verify the missing estimates in Prop. 13 Steps 3–4 and Prop. 5's H^β' claim. If they close, this is a major result. If they don't, the paper needs a real fix and cannot be published as is. My honest verdict: conditional, leaning positive, with the burden on the authors to supply the missing computation.","headline":"If the sketched estimates close, this is a major result: the first sharp continuous loss curve for 3D Euler. But the missing H^β closure for the perturbation error is load-bearing, so the paper is conditional, not proven.","tokens_in":38421,"tokens_out":3339,"would_cite":true,"duration_ms":31984,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q35"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper constructs 3D Euler vorticity that is arbitrarily small in $H^s$ yet loses Sobolev regularity continuously from time zero, with regularity exactly $H^{(s-ct)/(1+ct)}$ at time $t$.","keywords":["3D Euler equation","Sobolev regularity","ill-posedness","norm inflation","instantaneous loss of regularity","continuous loss of regularity","vorticity","pseudosolution"],"falsifier":"Perform the $H^2$ error estimate for the background error $\\Phi$ in Proposition 5 term by term, especially $D^2v[\\Phi]\\cdot\\nabla\\Phi$; if its $L^2$ bound exceeds $N^{-1+\\varepsilon+ct}(\\lambda N)^{\\beta'-s}(\\log N)^2$ by any power of $\\lambda N$, the bootstrap interval cannot reach $T$ and the norm-inflation growth $N^{t(1+\\beta)/2}$ fails. This is a direct calculation a reader can carry out on the skipped terms.","tokens_in":37072,"feed_emoji":"🌀","tokens_out":11288,"duration_ms":118358,"temperature":0.7,"pith_summary":"The paper proves that supercritical Sobolev regularity is lost, not just inflated, for the 3D incompressible Euler equations. For every $s\\in(0,3/2)$ and every $\\varepsilon>0$, it produces a divergence-free initial vorticity $\\omega_0$ with $\\|\\omega_0\\|_{H^s}\\le\\varepsilon$ whose unique classical solution satisfies $\\omega(\\cdot,t)\\in H^{(s-ct)/(1+ct)}$ and $\\omega(\\cdot,t)\\notin H^{\\beta'}$ for every $\\beta'>(s-ct)/(1+ct)$ at each time $t\\in[0,T_0]$. The construction builds each solution from a near-steady background vorticity and a much smaller, rapidly oscillating perturbation; a hyperbolic stagnation point stretches the perturbation so that its $H^\\beta$ norm grows like $N^{t(1+\\beta)/2}$. Summing infinitely many such building blocks at separated scales makes the total $H^\\beta$ norm finite exactly when $\\beta\\le(s-ct)/(1+ct)$, which is the stated threshold. This settles the supercritical range below the $H^{3/2}$ borderline where earlier critical ill-posedness results stopped.","feed_headline":"3D Euler flows can lose Sobolev regularity at every positive time","feed_subtitle":"For any tiny H^s initial vorticity, a unique classical solution loses smoothness along the curve (s-ct)/(1+ct).","key_machinery":"The engine is a pseudosolution pair $(\\varphi,\\psi)$: $\\varphi$ is a near-steady, highly oscillatory background vorticity, and $\\psi$ is a much smaller, even more oscillatory perturbation. The background's Biot-Savart velocity $v[\\varphi]$ is, up to an error $N^{-1}$ smaller, the hyperbolic stagnation flow $\\nabla v[\\varphi]\\approx\\operatorname{diag}(0,\\log N,-\\log N)$ in the $x_2$-$x_3$ plane. $\\psi$ is defined as the solution of the linear transport-stretching equation $\\partial_t\\psi+v[\\varphi]\\cdot\\nabla\\psi=\\psi\\cdot\\nabla v[\\varphi]$, so the Cauchy formula $\\psi(x,t)=\\nabla\\eta(y,t)\\psi(y,0)$ applies. The flow squeezes in $x_3$ and stretches in $x_2$ at rate $N^{t/2}$: advection of the","core_discovery":"On the paper's own terms, the central discovery is a construction of smooth, compactly supported 'building blocks' in which a background vorticity $\\varphi$ is an almost-steady, highly oscillatory state near the origin, and a perturbation $\\psi$ solves the linear transport-stretching equation $\\partial_t\\psi+v[\\varphi]\\cdot\\nabla\\psi=\\psi\\cdot\\nabla v[\\varphi]$. Because the Biot-Savart velocity $v[\\varphi]$ is, up to a small error, the hyperbolic stagnation flow $\\operatorname{diag}(0,\\log N,-\\log N)$ in the $x_2$-$x_3$ plane, the Cauchy formula $\\psi(x,t)=\\nabla\\eta(y,t)\\psi(y,0)$ shows that $\\psi$ gains a factor $N^{t/2}$ from vortex stretching and a further factor $N^{\\beta t/2}$ from adv","pith_inferences":["The authors leave implicit that the threshold formula may be read as an ODE for the maximal regularity exponent $\\beta(t)=(s-ct)/(1+ct)$, whose derivative $d\\beta/dt=-c(1+\\beta)$ combines stretching and transport rates; analogous 'regularity ODEs' might hold for other hyperbolic-point constructions.","One testable extension is numerical: simulating a single block with the parameter relation $\\lambda^{3/2-s}N^{-s}=K\\log N$ and measuring $\\|\\psi(t)\\|_{H^\\beta}$ directly would check whether the predicted exponent $t(1+\\beta)/2$ is robust to small perturbations of the exact construction.","Because the background vorticity sits near a steady Euler solution, the same short-time stretching mechanism could plausibly survive for the Navier-Stokes equations with small viscosity before dissipation dominates; the present proof provides a no-dissipation baseline for such an extension."],"forward_implications":["For every $s\\in(0,3/2)$, the loss is instantaneous: no matter how small $t>0$ is, the solution misses every Sobolev space above $H^{(s-ct)/(1+ct)}$.","The regularity threshold decreases continuously in time, with rate governed by the same constants that control norm growth, so the construction ties the rate of loss to the Lyapunov exponent of the background hyperbolic flow.","Uniqueness holds in a natural class of classical solutions, so the loss is a property of the Euler evolution itself rather than a selection effect of weak solutions.","The gluing argument shows that finite-time norm inflation for a single smooth block can be promoted to continuous-in-time loss whenever infinitely many summable blocks can be placed at separated scales."],"supporting_citations":[{"why":"Supplies the hyperbolic-point growth mechanism and the instantaneous-continuous-loss template that the paper adapts from SQG to 3D Euler.","marker":"[5]"},{"why":"Provides the 2D Euler gap-loss construction and gluing strategy that this paper extends to continuous-in-time loss in 3D.","marker":"[6]"},{"why":"Established norm inflation for supercritical $H^s$ vorticity in 3D; this paper proves the stronger continuous loss by a different mechanism.","marker":"[16]"},{"why":"Established local well-posedness of 3D Euler for velocities in $H^s$, $s>5/2$, the classical baseline this ill-posedness result contrasts with.","marker":"[13]"},{"why":"Gives the commutator estimates used in Sobolev regularity estimates for Euler and Navier-Stokes.","marker":"[14]"},{"why":"Settled strong ill-posedness at the borderline $H^{5/2}$ velocity space, the critical endpoint of the supercritical result.","marker":"[3]"},{"why":"Refines the borderline ill-posedness theory and frames the $H^{3/2}$ vorticity threshold the new result sits below.","marker":"[4]"},{"why":"Provides the Sobolev-Slobodeckij characterization used in the fractional Sobolev estimates, including the disjoint-support decomposition (2.6).","marker":"[7]"}],"fun_headline_variants":["3D Euler: instantaneous continuous Sobolev loss","Small vorticity, instant regularity drop in 3D Euler","3D Euler sheds regularity continuously from t=0","Even tiny vorticity loses smoothness instantly in 3D Euler","3D Euler: regularity decays along (s-ct)/(1+ct)"],"cache_read_input_tokens":2816,"weakest_assumption_plain":"The proof stands only if every error estimate left to 'similar lines' or 'bootstrapping' is no larger than the displayed $N^{-1}$- and $\\lambda^{-c_*}$-small bounds; a single skipped term with one extra power of the frequency parameters would close the control window before the time at which the growth is needed.","fun_headline_variants_meta":{"raw":{"variants":["3D Euler: instantaneous continuous Sobolev loss","Small vorticity, instant regularity drop in 3D Euler","3D Euler sheds regularity continuously from t=0","Even tiny vorticity loses smoothness instantly in 3D Euler","3D Euler: regularity decays along (s-ct)/(1+ct)"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000196,"raw_usage":{"total_tokens":1222,"prompt_tokens":793,"completion_tokens":429,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":537,"completion_tokens_details":{"reasoning_tokens":343}},"tokens_in":537,"tokens_out":429,"duration_ms":4665,"temperature":1.0,"reasoning_tokens":343,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T22:53:37.412915+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Perform the $H^2$ error estimate for the background error $\\Phi$ in Proposition 5 term by term, especially $D^2v[\\Phi]\\cdot\\nabla\\Phi$; if its $L^2$ bound exceeds $N^{-1+\\varepsilon+ct}(\\lambda N)^{\\beta'-s}(\\log N)^2$ by any power of $\\lambda N$, the bootstrap interval cannot reach $T$ and the norm-inflation growth $N^{t(1+\\beta)/2}$ fails. This is a direct calculation a reader can carry out on the skipped terms.","supporting_citations":[{"cited_title":"Córdoba, L","cited_arxiv_id":null,"evidence_quote":"Supplies the hyperbolic-point growth mechanism and the instantaneous-continuous-loss template that the paper adapts from SQG to 3D Euler."},{"cited_title":"J., 173 (2024), pp","cited_arxiv_id":null,"evidence_quote":"Provides the 2D Euler gap-loss construction and gluing strategy that this paper extends to continuous-in-time loss in 3D."},{"cited_title":"Kato, Nonstationary flows of viscous and ideal fluids inR3, J","cited_arxiv_id":null,"evidence_quote":"Established local well-posedness of 3D Euler for velocities in $H^s$, $s>5/2$, the classical baseline this ill-posedness result contrasts with."},{"cited_title":"Kato and G","cited_arxiv_id":null,"evidence_quote":"Gives the commutator estimates used in Sobolev regularity estimates for Euler and Navier-Stokes."},{"cited_title":"Bourgain and D","cited_arxiv_id":null,"evidence_quote":"Settled strong ill-posedness at the borderline $H^{5/2}$ velocity space, the critical endpoint of the supercritical result."},{"cited_title":"Bourgain and D","cited_arxiv_id":null,"evidence_quote":"Refines the borderline ill-posedness theory and frames the $H^{3/2}$ vorticity threshold the new result sits below."},{"cited_title":"Di Nezza, G","cited_arxiv_id":null,"evidence_quote":"Provides the Sobolev-Slobodeckij characterization used in the fractional Sobolev estimates, including the disjoint-support decomposition (2.6)."}],"review_version":1}