{"id":"162693ba-8403-403d-a1d6-e36c65db7951","arxiv_id":"2607.19071","paper_version":1,"verdict":"CONDITIONAL","confidence":"LOW","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"For small initial perturbations in H^k ∩ W^{3,1} with k ≥ 14, the 2D Boussinesq equations with only horizontal dissipation admit global classical solutions that decay at explicit anisotropic rates.","lead":"This paper proves that the two-dimensional Boussinesq equations with only horizontal dissipation remain stable near the hydrostatic balance, even though the dissipation alone is too weak. The stabilizing mechanism is the dispersive decay of internal gravity waves generated by the velocity-temperature coupling, which supplies the missing time integrability.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 3.9's L4 estimate is not supported by Lemma 3.7: the lemma states L1-L8, but the proof gives only L1-L∞, and the Hilbert transform splitting is not L1-bounded.","rationale":"The paper addresses a genuinely hard problem and the proposed stabilizing mechanism is plausible. Credit is due for stating explicit decay rates and carrying through a nonlinear bootstrap. However, the central linear estimate is not merely hand-wavy: the proof of Lemma 3.7 states L1-L8, yet the actual argument gives L1-L∞, which is what Proposition 3.9 needs but does not prove; and the Hilbert transform step cannot be justified in L1. Because Section 5's nonlinear estimates all call Proposition 3.9, this is load-bearing. The reader's CONDITIONAL verdict with LOW confidence is justified; the work needs a verified correction of Proposition 3.9 before acceptance. The concern does not in itself warrant rejection, since a corrected L1-L∞ estimate may be true and provable, so I keep the verdict UNCHANGED rather than moving to ACCEPT or REJECT.","tokens_in":1141,"tokens_out":4879,"duration_ms":216225,"concrete_test":"Check Proposition 3.9 analytically: (i) Prove or disprove the L1 to L∞ bound for the semigroup e^{t(∂1^2+R1)}Δ_j with phase |ξ1|/|ξ|, without splitting by sgn(ξ1); if the bound fails, identify the largest p for which L1 to Lp holds and recompute the interpolation. (ii) Verify the Riesz-Thorin step: apply Lemma 3.7 as stated (L1 to L8) to derive the endpoint at q=4 and compare with (3.12); if the input space and exponent differ from L4/3 and (1+t)^(-1/2), correct Proposition 3.9. (iii) Numerically test (3.4) for p=1: take a band-limited Schwartz function f with Fourier support in the annulus used by Δ_0 and compute ||H f||_1 versus ||f||_1 for several grid sizes; if the ratio grows, the Hilbert transform is not L1-bounded as claimed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper relies on the linear L4 estimate Proposition 3.9 to close the entire nonlinear bootstrap. Lemma 3.7 claims an L1 to L8 bound (3.7), but its proof estimates the inverse Fourier kernel in L∞ and applies Young's inequality, which yields L1 to L∞, not L1 to L8. These are not equivalent: interpolating (3.6) L2 to L2 with an L1 to L8 bound by Riesz-Thorin would give an L6/5 to L4 estimate with decay (1+t)^(-2/3), not the L4/3 to L4 estimate with decay (1+t)^(-1/2) stated in Proposition 3.9. The stated Proposition 3.9 (and the localized bound (3.12) from which it is derived) corresponds exactly to interpolating L2 to L2 with an L1 to L∞ bound. Thus the proof must be read as requiring an L1 to L∞ estimate, but no such estimate is proved. Moreover, the derivation of the L∞ estimate splits the phase |ξ1|/|ξ| using the sign of ξ1 and invokes boundedness of the Hilbert transform on L1 through (3.4). Riesz transforms are not bounded on L1; (3.4) is stated for all p, including p=1, which is false. Since the nonlinear Duhamel integrals in Section 5 are estimated with Proposition 3.9, this missing and misstated linear estimate is load-bearing: the decay exponents, for example m2≈1.00003 in Section 5.5, are extremely close to 1, so any loss in the linear rate or change in the norm used for the nonlinear data can break the time integrability needed to close the bootstrap.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the two-dimensional Boussinesq equations with only horizontal dissipation near the hydrostatic equilibrium (U,Θ)=(0,x_2). The perturbation system is rewritten as a coupled dispersive-dissipative system, symmetrized in Section 2, and the linearized operator is diagonalized through internal gravity wave modes. Section 3 develops linear estimates, including an L^4 dispersive estimate for the semigroup; Sections 4 and 5 then propose a bootstrap argument to prove global existence and componentwise decay rates for sufficiently small data in H^k(R^2)∩W^{3,1}(R^2), k≥14. The claimed decay rates include ||(u_1|θ)||_{L^2}≤C_0ε(1+t)^{-1/8-η}, ||u_2||_{L^2}≤C_0ε(1+t)^{-1/4}, and ||u_2||_{L^4}≤C_0ε(1+t)^{-7/8-η+δ} with η=1/120 and δ=10^{-5}.","tokens_in":28005,"tokens_out":22118,"duration_ms":191082,"significance":"If the proof were complete, the paper would be a significant contribution to the anisotropic Boussinesq stability problem: it identifies a concrete stabilizing mechanism (dispersive decay of internal gravity waves combined with horizontal dissipation) that compensates for the absence of vertical dissipation, and it provides explicit anisotropic and componentwise rates. The organization is clear, the bootstrap assumptions and improved estimates are stated explicitly, and the linear semigroup representation is worked out in detail. However, the central linear estimate on which the whole bootstrap rests is not proved as stated, and several exponent checks that are needed to close the convolution estimates are only verified numerically. The result is plausible, but the manuscript currently does not provide a complete proof.","major_comments":[{"comment":"Lemma 3.7 states (3.7) as an L^1(R^2)→L^8(R^2) estimate, but the proof bounds the kernel in L^∞ and uses Young's inequality with the L^1 norm of the data, which yields an L^1→L^∞ estimate. The subsequent Riesz–Thorin interpolation in Proposition 3.9 is algebraically the one associated with the L^∞ endpoint: interpolating (3.6) with an L^1→L^∞ bound at θ=1/2 gives L^{4/3}→L^4 with the dyadic factor 2^j and decay (1+t)^{-1/2} displayed in (3.12). Interpolating the stated L^1→L^8 bound would give L^{6/5}→L^4 and decay (1+t)^{-2/3}. Thus, as written, Proposition 3.9 is not proved. This is load-bearing because the Duhamel estimates in Sections 5.1–5.4 invoke (3.11) to close the bootstrap. The lemma and proposition must be reconciled: either correct (3.7) to an L^∞ estimate or provide a genuine L^8 estimate that still yields (3.12).","section":"§3.3, Lemma 3.7 and Proposition 3.9"},{"comment":"The passage from (3.12) to the W^{1+γ,4/3} bound contains an unshown dyadic summation. The estimate (3.12) has a factor 2^j, while the next display introduces 2^{j/2}+2^j without explanation, and the reduction of ∑_j 2^j‖Δ_j f_0‖_{L^{4/3}} to ‖f_0‖_{L^{4/3}}+‖Λ^{1+γ}f_0‖_{L^{4/3}} requires a separate treatment of the positive and negative frequency sums. Please provide the complete summation argument, or state the estimate with the appropriate Besov norm on the right-hand side.","section":"§3.3, proof of Proposition 3.9"},{"comment":"The text verifies that s_1, s_2, m_1, m_2, m_3 are larger than 1 only by approximate decimal evaluations for the chosen parameters (for example, m_2≈1.000030553 for η=1/120, k=14, δ=10^{-5}). No derivation or exact algebraic inequality is shown, and the preceding estimates have unspecified multiplicative constants. Since Lemma 3.3 requires the exponent β>1 to avoid a logarithmic factor, the closing of the bootstrap depends on these inequalities holding exactly. The numerical checks should be replaced by a rigorous verification, e.g., exact rational arithmetic or a clear monotonicity argument for the stated parameter values.","section":"§5.1 and §5.5, exponent verifications"}],"minor_comments":[{"comment":"The uniform L^p bound for R_jΔ_j is correct for all 1≤p≤∞ because the symbol is smooth on the dyadic annulus and the rescaled kernel has a uniformly integrable profile. The text should say this explicitly rather than invoking global Riesz-transform boundedness for p=1,∞, where the latter statement would be false.","section":"§3.3, equation (3.4)"},{"comment":"The factor 2^{j/2} appears in the dyadic sum after (3.12) but not in (3.12) itself. Please remove the typographical discrepancy or justify where the 2^{j/2} comes from.","section":"§3.3, equation (3.12) and following display"},{"comment":"The exponent (1+(1−δ)α)/2 in the nonlinear Duhamel estimate is stated without derivation, whereas the corresponding linear exponent is (1+α)/2. Please show explicitly how the horizontal heat semigroup and Proposition 3.9 combine to produce the δ-dependent exponent.","section":"§5.2"},{"comment":"In the estimates for G_11 and G_12, the replacement of B_1Λ^{-1}Δ_j by a bounded dyadic multiplier is not stated explicitly. Because ξ_1/|ξ| is uniformly bounded on each dyadic annulus, one has B_1Λ^{-1}Δ_j∼Δ_j in L^{4/3}; this step should be written out.","section":"§5.4, equation (5.1)"},{"comment":"The symbol L^8 appears in Lemma 3.7 and its proof; if it is a rendering of L^∞, it should be corrected throughout. The reader should not have to infer the intended exponent from the interpolation argument.","section":"Global"}],"recommendation":"major_revision","confidential_remarks":"This is a technically demanding paper by established authors, and the proposed mechanism is plausible. My main concern is purely technical: the linear L^4 estimate is the foundation of the nonlinear bootstrap, and the current statement/proof of Lemma 3.7–Proposition 3.9 is not consistent, while the closing exponent inequalities are only verified numerically. These issues are substantial but appear fixable within the manuscript's scope, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper takes on a genuinely hard problem: global stability near hydrostatic balance for the 2D Boussinesq equations with only horizontal dissipation on R^2. The mixed-domain result relied on Poincaré, unavailable here; the authors correctly identify dispersive decay of internal gravity waves as the stabilizing mechanism. That is the right idea, and the architecture is sensible: symmetrize, diagonalize, estimate the semigroup, close via bootstrap. If the technical steps check out, it is a real advance.\n\nThe soft spots are real but not all fatal. The stress-test's claim about Lemma 3.7 is partly wrong: interpolating L1 to L∞ with L2 to L2 gives exactly the L^{4/3} to L4 rate used in Proposition 3.9, not the L^{6/5} rate. The stated L1 to L8 is inconsistent with the proof, but that mismatch is a typo-level flaw. The deeper problem is that the proof of the L∞ bound splits the phase using sgn(ξ1) and invokes Hilbert transform boundedness on L1, which is false. Frequency localization might rescue it, but the paper does not show how. Since the nonlinear estimates depend on that linear bound, this is load-bearing.\n\nSecond, the bootstrap margins are razor-thin. The exponent m2 is 1.00003; the algebra is not shown. A small error would break time integrability. Several estimates are omitted or described as 'easier'. This requires careful independent checking.\n\nThe paper deserves a serious referee. The problem is important, the strategy is credible, the gaps are technical rather than conceptual. I would send it to peer review. Confidence in the proof as written is low, but the idea merits engagement.","headline":"A plausible solution to a notable open problem in anisotropic Boussinesq stability, but the linear dispersive estimate has a real gap around the Hilbert transform and the bootstrap margins are razor-thin.","tokens_in":828,"tokens_out":1404,"would_cite":false,"duration_ms":94906,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q35","35B35","35B40","76D03","76D50"],"pacs":[],"model":"deepseek-v4-flash","headline":"The hydrostatic balance of the 2D Boussinesq equations is globally stable under horizontal-only dissipation, with explicit decay rates.","keywords":["Boussinesq equations","hydrostatic equilibrium","horizontal dissipation","stability","large-time behavior","dispersive estimates","internal gravity waves","anisotropic decay"],"falsifier":"Evaluate numerically, for $j=0$, the oscillatory integral $I(t)=\\int e^{ix\\cdot\\xi}e^{i\\xi_1 t/|\\xi|}\\psi_2(\\xi_1)\\tilde\\varphi(\\xi)\\,d\\xi$ in the sector $|\\xi_1|\\le 1/3$, $|\\xi_2|\\approx 1$. The proof needs $|I(t)|\\lesssim (1+t)^{-1}$ uniformly in $x$; if the true decay is only $(1+t)^{-1/2}$ or worse, the kernel bound (3.7) is too optimistic and the bootstrap closing in Sections 4 and 5 collapses.","tokens_in":27414,"feed_emoji":"🌊","tokens_out":12638,"duration_ms":91544,"temperature":0.7,"pith_summary":"This paper is trying to prove that the hydrostatic equilibrium $(U,\\Theta)=(0,x_2)$ of the two-dimensional Boussinesq equations remains stable even when the only dissipation is horizontal. The claim is that for initial perturbations small in $H^k(\\mathbb{R}^2)\\cap W^{3,1}(\\mathbb{R}^2)$ with $k\\ge 14$, there is a unique global classical solution whose $H^k$ norm stays bounded, whose horizontal derivatives are square-integrable in time, and whose velocity and temperature decay at explicit componentwise rates. The reason this matters is that horizontal dissipation alone cannot control the nonlinearity by standard energy estimates; the paper argues that the coupling between velocity and temperature creates internal gravity waves whose dispersive decay supplies the missing stabilization. If the claim is right, it identifies wave propagation as a stabilizing mechanism that can compensate for incomplete dissipation in stratified fluids.","feed_headline":"Internal waves stabilize the Boussinesq hydrostatic balance","feed_subtitle":"Small perturbations of a stratified fluid decay to rest even with zero vertical dissipation.","key_machinery":"The load-bearing object is the coupled oscillatory-dissipative semigroup $S_\\sigma(t)=e^{t\\partial_1^2}e^{\\sigma t R_1}Q_\\sigma$, $\\sigma\\in\\{+, -\\}$, where $R_1$ is the Fourier multiplier $i|\\xi_1|/|\\xi|$ and $Q_\\sigma$ projects onto the eigenspace of the skew-adjoint coupling operator. The phase $\\varphi(\\xi)=\\xi_1/|\\xi|$ is anisotropic: its Hessian degenerates along $\\xi_2=0$ and the phase itself is not smooth at $\\xi_1=0$, so the proof decomposes frequency space and uses two mechanisms at once, stationary phase for the non-degenerate part and horizontal heat-kernel decay on Littlewood-Paley blocks for the degenerate part. The resulting kernel bound $2^{2j}(1+t)^{-1}+2^{2j}e^{-c2^{2j}t}(1+t)^{-1/2}$ is interpolated to the $L^4$ estimate, and this estimate is what makes the Duhamel nonlinear terms time-integrable in the bootstrap.","core_discovery":"The paper's central discovery is that the perturbation system around the hydrostatic balance has a hidden oscillatory structure that standard energy estimates miss. After writing the perturbation $(u_1,u_2,\\theta)$ as a vector and applying an extended Helmholtz projection, the linearized operator has eigenvalues $\\pm i|\\xi_1|/|\\xi|$ and $0$; the zero mode is killed by divergence freeness, so the linear flow is a sum of two semigroups $e^{t\\partial_1^2}e^{\\pm t R_1}$ acting on the corresponding eigenprojections, where $R_1$ is the Fourier multiplier with symbol $i|\\xi_1|/|\\xi|$. The paper establishes an $L^4$ dispersive estimate of order $(1+t)^{-1/2}$ for these semigroups on $W^{1+\\gamma,4/3}$ data, splitting frequency space: in the region where the phase $\\xi_1/|\\xi|$ is non-degenerate, stationary phase gives the decay, while in the degenerate region the one-dimensional horizontal heat kernel $e^{-\\xi_1^2 t}$ provides it. Combining this linear decay with a bootstrap of $H^k$ energy estimates yields the global existence, uniqueness, and componentwise decay rates of Theorem 1.1, including the faster decay of the vertical velocity.","pith_inferences":["One extension the authors do not pursue: the same semigroup mechanism should likely work for fractional horizontal dissipation $(-\\partial_1^2)^\\alpha$ with $\\alpha$ below 1, provided the degenerate-frequency kernel estimate is re-derived with the fractional heat kernel; the present paper treats the endpoint case $\\alpha=1$ in the horizontal direction.","The requirement $k\\ge 14$ likely reflects the interpolation-heavy bootstrap rather than a threshold set by the physics; sharper product estimates might lower the needed regularity, though the smallness condition would remain.","A direct numerical evaluation of the oscillatory integral at the degenerate sector would provide a clean independent test of the mechanism, since the linear decay rate is what carries the whole nonlinear argument."],"forward_implications":["If Theorem 1.1 is correct, the hydrostatic balance is globally nonlinearly stable on the whole space $\\mathbb{R}^2$ with only horizontal dissipation, with no periodicity or Poincaré inequality needed: the dispersive decay replaces that mechanism.","The vertical velocity $u_2$ decays faster than the horizontal velocity and the temperature componentwise, for example $\\|u_2(t)\\|_{L^2}\\le C_0\\varepsilon(1+t)^{-1/4}$ versus $\\|(u_1|\\theta)(t)\\|_{L^2}\\le C_0\\varepsilon(1+t)^{-1/8-\\eta}$, and horizontal derivatives decay faster still.","The proof yields uniform time-integrability of the horizontal dissipation: $\\int_0^t (\\|\\partial_1 u(\\tau)\\|_{H^k}^2 + \\|\\partial_1\\theta(\\tau)\\|_{H^k}^2)\\,d\\tau$ is bounded independently of $t$.","The same wave-dispersion mechanism separates the Boussinesq system from the anisotropically dissipative Navier–Stokes equations, for which horizontal-only dissipation has no known robust stabilizing mechanism."],"supporting_citations":[{"why":"Supplies the stationary-phase bound for oscillatory integrals used to obtain the $(1+t)^{-1/2}$ decay in the non-degenerate frequency region.","marker":"[32]"},{"why":"Gives the exponential decay $e^{-c2^{2j}t}$ of the heat semigroup on Littlewood-Paley blocks, used in the degenerate frequency region.","marker":"[2]"},{"why":"Provides the convolution-type time-integral estimates and the heat semigroup $L^p$-$L^q$ decay used throughout the Duhamel terms.","marker":"[28]"},{"why":"Provides the fractional Leibniz and commutator inequalities used to control nonlinear products in $H^k$.","marker":"[26]"},{"why":"Supplies Riesz–Thorin interpolation, converting the $L^2$ and $L^\\infty$ bounds into the $L^4$ estimate.","marker":"[15]"}],"fun_headline_variants":["Zero vertical dissipation: waves save Boussinesq stability","Internal waves compensate for missing vertical damping","Waves plus horizontal viscosity stabilize stratified flow","Boussinesq hydrostatic balance restored by wave decay","No vertical viscosity? Waves still stabilize the fluid"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof stands on the linear estimate that the oscillatory-dissipative semigroup decays like $(1+t)^{-1/2}$ in $L^4$; if the stationary-phase argument does not survive in the degenerate frequency region $|\\xi_1|\\le |\\xi_2|$, where the phase $|\\xi_1|/|\\xi|$ is not smooth at $\\xi_1=0$, the bootstrap exponents fail and the decay rates would not close.","fun_headline_variants_meta":{"raw":{"variants":["Zero vertical dissipation: waves save Boussinesq stability","Internal waves compensate for missing vertical damping","Waves plus horizontal viscosity stabilize stratified flow","Boussinesq hydrostatic balance restored by wave decay","No vertical viscosity? Waves still stabilize the fluid"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001356,"raw_usage":{"total_tokens":5551,"prompt_tokens":1037,"completion_tokens":4514,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":653,"completion_tokens_details":{"reasoning_tokens":4442}},"tokens_in":653,"tokens_out":4514,"duration_ms":27386,"temperature":1.0,"reasoning_tokens":4442,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T15:30:10.378607+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate numerically, for $j=0$, the oscillatory integral $I(t)=\\int e^{ix\\cdot\\xi}e^{i\\xi_1 t/|\\xi|}\\psi_2(\\xi_1)\\tilde\\varphi(\\xi)\\,d\\xi$ in the sector $|\\xi_1|\\le 1/3$, $|\\xi_2|\\approx 1$. The proof needs $|I(t)|\\lesssim (1+t)^{-1}$ uniformly in $x$; if the true decay is only $(1+t)^{-1/2}$ or worse, the kernel bound (3.7) is too optimistic and the bootstrap closing in Sections 4 and 5 collapses.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the stationary-phase bound for oscillatory integrals used to obtain the $(1+t)^{-1/2}$ decay in the non-degenerate frequency region."},{"cited_title":"Bahouri, J.-Y","cited_arxiv_id":null,"evidence_quote":"Gives the exponential decay $e^{-c2^{2j}t}$ of the heat semigroup on Littlewood-Paley blocks, used in the degenerate frequency region."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the convolution-type time-integral estimates and the heat semigroup $L^p$-$L^q$ decay used throughout the Duhamel terms."},{"cited_title":"Li, On Kato–Ponce and fractional Leibniz,Rev","cited_arxiv_id":null,"evidence_quote":"Provides the fractional Leibniz and commutator inequalities used to control nonlinear products in $H^k$."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies Riesz–Thorin interpolation, converting the $L^2$ and $L^\\infty$ bounds into the $L^4$ estimate."}],"review_version":2}