{"id":"0307c4ea-39b3-4f12-a3e4-f9392baff0ff","arxiv_id":"2602.19846","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"Uniqueness of mild Navier–Stokes solutions in critical Besov spaces holds exactly for p<n (any q) or p=n (q≤2), and fails — even for zero initial data — for p=n, q>2 and for p>n.","lead":"This paper claims to settle exactly when mild Navier–Stokes solutions are unique across the full family of critical Besov spaces: uniqueness in L^n-type classes is sharp, and every slightly larger critical space admits non-unique solutions, even from zero initial state. The mechanism builds explicit non-trivial steady flows and attaches global non-decaying solutions to them, the first 'non-dissipative' critical solutions of the unforced equations.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Smallness bound (4.1) is not supported by Lemma 4.9 under the printed tetration schedule and ψ_j cutoff; the j=1 block alone is astronomically too large, so the perturbation argument loses its premise.","rationale":"The reader's weakest_assumption correctly identifies the magnitude control (4.1) as the load-bearing step. Under the printed definitions of λ_j and ψ_j, Lemma 4.9's lower bound for the first block already violates the claimed smallness, so the whole perturbative construction lacks its starting premise. This is an internal inconsistency visible in the text, not a speculative disagreement with consensus. I nevertheless do not move the verdict to REJECT because the inconsistency may stem from a correctable typo (e.g., the intended spatial cutoff or the index on λ in (4.7)), as the reader suggests; the surrounding strategy is coherent and the remaining estimates are standard. Therefore a conditional verdict, pending clarification of the spatial-cutoff scale and exponents, is appropriate. The paper contains independent positives—clear statement of the classification, honest discussion of the 2D obstruction and endpoint cases—but the central smallness estimate must be fixed before the main theorem can be accepted.","tokens_in":30145,"tokens_out":22611,"duration_ms":179629,"concrete_test":"Recompute the j=1 term in Lemma 4.9 with the printed definitions: set λ_0=λ, λ_1=λ^{2^{λ^{4μ}}}, ψ_1(x)=ψ(|x|−λ_1); verify the lower bound ‖V_1^(p)‖_{B^{n/p−1}_{p,q}} ≥ c λ_1^{n/p} λ^{4−2μ} σ_1(n/p−1,q) and compare it to λ^{−α(p)}. If the ratio (λ_1^{n/p} λ^{4−2μ} σ_1)/λ^{−α(p)} is unbounded as λ→∞, then (4.1) fails. As a concrete numerical example, take n=3, p=4, q=1, μ=17; the log of the ratio is of order 2^{λ^{68}}, which is unbounded and astronomically large.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 4.1(i) asserts ‖V_λ‖_{B^{n/p−1}_{p,q}} ≤ Cλ^{−α(p)}. This is obtained by summing the Lemma 4.9(4.7) bound ‖V_j^(p)‖ ≤ C λ_j^{n/p} λ_{j−1}^{4−2μ} σ_j(n/p−1,q). With the printed definitions λ_0=λ, λ_j=λ^{2^{λ_{j−1}^{4μ}}} (4.4), and ψ_j(x)=ψ(|x|−λ_j), the envelope Γ_{j,k}[V_{j−1}]ψ_j is O(1) on a ball of radius λ_j, so its L^p norm is ~λ_j^{n/p}. For p>n, s=n/p−1<0 and σ_j(s,q) ≥ c λ_{j−1}^{sμ}(log λ_{j−1})^{-1/2}; at j=1 this yields ‖V_1^(p)‖ ≥ c λ_1^{n/p} λ^{4−2μ} λ^{(n/p−1)μ}(log λ)^{-1/2} = c exp( (n/p) 2^{λ^{4μ}} log λ ) times a power of λ. This is not ≤ Cλ^{−α(p)}. The same problem occurs for p=n, q>2 because of the λ_j^{n/p} factor. Consequently the series defining V^(p) does not converge in the advertised Besov space; the smallness of F in (4.2), the Step 1 contraction in §5, and the smallness of w0 in Step 2 all depend on (4.1). As printed, the construction is internally inconsistent.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies uniqueness of mild solutions to the incompressible Navier–Stokes equations in critical homogeneous Besov spaces \\dot B^{n/p-1}_{p,q}(R^n). It claims a complete classification: uniqueness for 1≤p<n with any q and for p=n with q≤2; non-uniqueness for p=n, q>2 and for n<p≤∞, any q. The non-uniqueness construction first produces a small stationary profile V_λ as an infinite sum of frequency-localized building blocks, with residual force F_λ belonging to a better Besov space and much smaller than V_λ; a contraction argument then yields a stationary solution U_λ, and a second fixed point around U_λ produces global non-stationary solutions, including from zero initial data. The paper also claims that these are the first non-dissipative unforced Navier–Stokes flows with critical regularity and that the L^n uniqueness theorem of Lions–Masmoudi and Furioli–Rieusset–Terraneo is sharp.","tokens_in":30452,"tokens_out":19911,"duration_ms":167289,"significance":"If the claims were fully established, the paper would be a major advance: it would give a complete Besov-space classification of uniqueness for mild Navier–Stokes solutions, show sharpness of the L^n endpoint, and provide the first critical-regularity non-dissipative unforced flows. The strategy is explicit and self-contained, combining Besov calculus with a Nash-type geometric lemma and a clear fixed-point architecture; these are genuine strengths. However, the central magnitude estimate for the principal profile is not justified as printed, and the significance of the paper is therefore conditional on a substantial correction.","major_comments":[{"comment":"The bound (4.1) is not derivable from Lemma 4.9 with the printed definitions. With λ_1 = λ^{2^{λ^{4μ}}} and ψ_j(x)=ψ(|x|-λ_j), Lemma 4.9 gives at j=1: ‖V_1^{(p)}‖_{B^{n/p-1}_{p,q}} ≥ c λ_1^{n/p} λ^{4-2μ} σ_1(n/p-1,q). Since σ_1(n/p-1,q) ≥ c λ^{(n/p-1)μ} for p>n and σ_1(0,q) ≥ c for p=n, q>2, and λ_1^{n/p} = exp((n/p)2^{λ^{4μ}} log λ), the first block alone is exponentially large, not ≤ Cλ^{-α(p)}. Consequently the series defining V^{(p)} does not converge in the advertised Besov space, and the smallness of F in (4.2) and of w_0 in §5 lose their premise. The likely source is the radius λ_j in the definition of ψ_j; if the intended radius is λ_{j-1}, all subsequent estimates must be redone and the proof re-checked. As printed, Theorem 4.1(i) is not established.","section":"§4.1–4.2, Lemma 4.9 and Theorem 4.1(i), Eqs. (4.7), (4.1), (4.4)"},{"comment":"The reduction at the beginning of Step 2 is not justified as written. The quoted embedding \\dot B^{s_1}_{p_1,q_1} ↪ \\dot B^{s_2}_{p_2,q_2} for p_1≤p_2, q_1≤q_2 embeds a smaller space into a larger one, whereas the constructed U_λ is only shown to lie in the target space. For endpoint cases p≥2n or q=∞ the theorem only needs the zero initial datum, so a repair may be possible by constructing the solution first in a smaller admissible critical space and then embedding, but this is not what is written and needs to be supplied.","section":"§5, Step 2 (reduction to p<2n, q<∞)"}],"minor_comments":[{"comment":"In the estimate of F^{(r)}_{λ,m,L}, the phrase 'the similar strategy as for the estimate of F^{(r)}_{λ,m,L}' appears self-referential; it should refer to the earlier remainder estimate. Several sums also use j where j' or ℓ is intended.","section":"§3, proof of Proposition 3.1"},{"comment":"There is a typo: 'a constant 1 η=η(n,p,q)⩾0' contains a stray '1'.","section":"Theorem 1.2"},{"comment":"In the line defining F_{1,2;j}, the sum over ℓ_1,ℓ_2∈Λ_j writes Φ_{j,k_1,ℓ}⊗Φ_{j,k_2,ℓ}; the index ℓ should be ℓ_1,ℓ_2.","section":"§4.2, decomposition of F_1"},{"comment":"The formula for h_j has an unbalanced parenthesis and is hard to parse; it should be written as h_j ∼ log_2( (log_λ λ_j)/(μ log_λ λ_{j-1}) ), which simplifies to λ_{j-1}^{4μ} up to a logarithmic correction.","section":"§4.1, notation for h_j"}],"recommendation":"major_revision","confidential_remarks":"The central magnitude estimate (4.1) is the linchpin of the paper, and as printed it is wrong: the j=1 block is astronomically too large under the tetration schedule and the λ_j-radius cutoff. I would only support eventual publication if the author corrects the cutoff scale (or otherwise repairs the estimate) and verifies all dependent bounds in Lemmas 4.6, 4.9 and in the proof of Theorem 4.1. The embedding gap in Step 2 is secondary and likely repairable. I do not see a novelty or attribution concern; the issue is technical correctness."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a serious referee, but the manuscript as printed has a load-bearing typo in Section 4. The headline result is a complete (p,q) classification of uniqueness in critical Besov spaces for mild Navier-Stokes, with non-unique solutions from zero data converging to nontrivial stationary states. If correct, that closes a long-standing problem, and the construction (multi-scale frequency shells, Nash-lemma cancellation) is genuinely novel. The author is also honest about the 2D obstruction and the η=0 cases.\n\nThe stress-test note is right: Theorem 4.1(i) does not follow from Lemma 4.9 under the stated definitions. With ψ_j(x)=ψ(|x|-λ_j), the amplitude of the j-th building block has L^p norm about λ_j^{n/p}, and the j=1 block alone is astronomically large under the tetration schedule (4.4). The paper's own lower bound for V_1 uses λ^{n/p}, not λ_1^{n/p}, which is inconsistent. The natural fix is to take the spatial cutoff at scale λ_{j-1}, not λ_j — that makes the L^p normalization λ_{j-1}^{n/p} and the series sums to the advertised λ^{-α(p)}. This is probably a typo rather than broken idea: the rest of the exponent bookkeeping (α(p)<β, the force estimates, the fixed-point steps) is internally coherent, and the cancellation identity requires only ψ_j=1 on supp V_{j-1}, which a cutoff at scale λ_{j-1} can provide.\n\nThat said, a typo that changes the magnitude of the principal term by a factor of exp(2^{λ^{4μ}}) is exactly the kind of thing a referee should check. I would not desk-reject this. Send it to a specialist who is comfortable with Besov calculus and convex integration, with specific instructions to verify (4.1), (4.5), and (4.7) after the scale correction. The classification claim, if it holds, is a major contribution; the current manuscript is not yet at the point where I would trust it without that verification.","headline":"Big claim with a fixable-looking typo in the key bound; worth referee time after the author sorts out the spatial cutoff scale.","tokens_in":31165,"tokens_out":9099,"would_cite":true,"duration_ms":72481,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q30","35A02","76D03"],"pacs":[],"model":"deepseek-v4-flash","headline":"Uniqueness of mild Navier–Stokes solutions in critical Besov spaces holds exactly for p<n (any q) or p=n with q≤2; for p=n with q>2 or p>n, zero initial data already generates infinitely many global solutions converging to distinct stationa","keywords":["Navier-Stokes equations","non-uniqueness","mild solutions","scaling critical Besov spaces","stationary solutions","global solutions","sharp uniqueness","multi-scale construction"],"falsifier":"Directly compute (or numerically estimate) the Besov norm of the first building block V_1^{(p)} at j=1 with the tetration schedule λ_1 = λ^{2^{λ^{4μ}}} and cutoff ψ_1 localized at radius λ_1, and compare it with the claimed bound Cλ^{-α(p)}; if ‖V_1^{(p)}‖_{B^{n/p-1}_{p,q}} ≳ λ^{n/p} λ^{4-2μ} σ_1(n/p-1,q) cannot be reconciled with Cλ^{-α(p)} because of the additional factor coming from the spatial scale of ψ_1, the main theorem fails.","tokens_in":29850,"feed_emoji":"🌊","tokens_out":5773,"duration_ms":49775,"temperature":0.7,"pith_summary":"This paper asks exactly when uniqueness of mild solutions to the incompressible Navier–Stokes equations survives in the smallest natural spaces, the scaling-critical Besov spaces indexed by two parameters (p,q). The author proves a complete dichotomy: uniqueness holds precisely for 1≤p<n with any q, or p=n with q≤2; for p=n with q>2, or p>n with any q, uniqueness fails. In the failure regime, even zero initial data produces infinitely many small global mild solutions, and each one tends as time goes to infinity toward a distinct nontrivial stationary flow that is itself a non-unique solution of the stationary equations. This proves that the classical uniqueness result in the critical Lebesgue space is sharp: any slightly larger critical space made from the same scale can already contain non-unique flows. If correct, the classification is complete and the constructed solutions are the first examples of non-dissipative, unforced Navier–Stokes flows with critical regularity.","feed_headline":"Navier–Stokes uniqueness fails in larger critical spaces","feed_subtitle":"Zero initial data already yields infinitely many global flows converging to distinct stationary states.","key_machinery":"The load-bearing object is the principal profile V = Σ_j V_j^{(p)}, a sum of almost frequency-localized, divergence-free building blocks. Each V_j lives on a tetration-like sequence of frequency scales λ_j (so the frequency gaps grow super-exponentially), and its amplitude is written as a finite superposition of plane waves with directions chosen through a geometric lemma that decomposes the identity matrix into rank-one tensors. The key identity is the cancellation −ΔV_{j-1} + P div(λ^8 ψ_j^2 Id − D V_{j-1}) = 0, which removes the worst low-frequency part of the nonlinear interaction of V_j with itself, leaving a residual forcing F that belongs to a better-regularity Besov space and is much","core_discovery":"The central discovery is that the boundary of uniqueness for Navier–Stokes mild solutions in the critical Besov scale is exactly the line p<n (any q) together with the point p=n, q≤2; beyond that line — p=n with q>2, or p>n with any q — uniqueness collapses. The mechanism is a new multi-scale construction of stationary building blocks: each block is spread over many frequency shells within a dyadic band, with amplitudes decreasing like 1/√ℓ, and the blocks are arranged so that the low-frequency part of the self-interaction cancels the linear diffusion of the previous block. The remaining forcing is smooth enough and small enough to be absorbed by a contractive fixed point in a better-regular","pith_inferences":["The mechanism points to a testable numerical check: assemble the first few building blocks on the torus and measure their critical Besov norm and the residual forcing, to see whether the cancellation identity holds at the level of computed magnitudes.","If the classification is right, then any proof of well-posedness in critical Besov spaces for p>n must necessarily exclude some solutions that are mild in the classical sense, or must add an auxiliary dissipative condition; this would explain why existing well-posedness arguments all use a Chemin–Lerner auxiliary norm.","The same multi-scale cancellation may apply to other equations with a quadratic nonlinearity and a diffusion operator, such as the magnetohydrodynamics system or the surface quasi-geostrophic equation, wherever one can find a geometric decomposition of the identity into rank-one tensors that survives the projection to divergence-free fields."],"forward_implications":["If the theorem is correct, uniqueness of mild solutions in C([0,T); L^n) cannot be extended to any critical space that is strictly larger but still of the same scaling; in particular, enlarging L^n to the Besov or Triebel–Lizorkin scale with p=n and q>2 already destroys uniqueness.","The result yields a complete yes/no classification of uniqueness in the critical Besov scale for every pair (p,q), leaving no gap.","The constructed stationary solutions are the first examples of non-uniqueness for steady Navier–Stokes flow in scaling-critical spaces, including the three-dimensional case.","Non-dissipative global solutions that do not decay to zero exist in critical spaces where the heat semigroup still decays, so viscous dissipation alone does not force decay for small critical initial data.","For the endpoint case BMO^{-1} (p=∞, q=2), unconditional uniqueness of small mild solutions fails, although well-posedness for small data in that space is known."],"fun_headline_variants":["Navier–Stokes uniqueness fails beyond p=n, q≤2","Zero initial data yields infinitely many Navier–Stokes flows","Sharp non-uniqueness for Navier–Stokes at criticality","Critical Besov spaces where Navier–Stokes uniqueness collapses","p=n, q>2 breaks Navier–Stokes uniqueness"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The whole construction rests on the claim that the infinite sum of building blocks V = Σ_j V_j^{(p)} is genuinely small in the critical Besov norm (of order λ^{-α(p)}), and specifically that the first building block, built on the enormous scale λ_1 = λ^{2^{λ^{4μ}}} with the radius-λ_1 cutoff, obeys the same size estimate as the rest of the series; if that exponent bookkeeping fails, the stationary state is not small, the forcing is not small, and the fixed-point argument coll","fun_headline_variants_meta":{"raw":{"variants":["Navier–Stokes uniqueness fails beyond p=n, q≤2","Zero initial data yields infinitely many Navier–Stokes flows","Sharp non-uniqueness for Navier–Stokes at criticality","Critical Besov spaces where Navier–Stokes uniqueness collapses","p=n, q>2 breaks Navier–Stokes uniqueness"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000775,"raw_usage":{"total_tokens":3249,"prompt_tokens":711,"completion_tokens":2538,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":455,"completion_tokens_details":{"reasoning_tokens":2461}},"tokens_in":455,"tokens_out":2538,"duration_ms":18153,"temperature":1.0,"reasoning_tokens":2461,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T21:34:12.095825+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Directly compute (or numerically estimate) the Besov norm of the first building block V_1^{(p)} at j=1 with the tetration schedule λ_1 = λ^{2^{λ^{4μ}}} and cutoff ψ_1 localized at radius λ_1, and compare it with the claimed bound Cλ^{-α(p)}; if ‖V_1^{(p)}‖_{B^{n/p-1}_{p,q}} ≳ λ^{n/p} λ^{4-2μ} σ_1(n/p-1,q) cannot be reconciled with Cλ^{-α(p)} because of the additional factor coming from the spatial scale of ψ_1, the main theorem fails.","supporting_citations":[],"review_version":1}