{"id":"24bbb644-f55b-4c67-846d-ae20885ef1ff","arxiv_id":"2505.20988","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"There exist compactly supported, smooth-before-blow-up solutions of the forced 2D Boussinesq equation that blow up in finite time with a uniformly C^{1,alpha} cap L^2 force for every alpha < sqrt(4/3)-1.","lead":"The paper constructs solutions of the forced 2D Boussinesq equations that stay smooth until a chosen time and then develop a singularity, while the external force remains as regular as C^{1,alpha} with alpha up to about 0.1547 up to the blow-up time. If correct, it shows that singularities can form with a force that is not itself singular, extending a recent program for Euler and porous media equations.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 10's uniform closeness estimate in the second half of each layer interval is the load-bearing step; the proof of its point 3 and the admissible parameter set for all α < sqrt(4/3)-1 are deferred to Section 6 and need independent verification.","rationale":"I read the paper in good faith: the mechanism is original, the toy pendulum model is explicit, and Lemma 2 and the formal layer decomposition are self-contained. The concern is not the existence of the toy model but whether the actual layer dynamics remain close enough to it through the part of each layer's life where the toy pendulum approaches the unstable point. The reader's weakest_assumption already flags the closeness of Ξ^(n) and k_n as load-bearing, and I agree with that diagnosis; I sharpen it by pointing to Proposition 10's third phase and to the exact numerical compatibility between the uniform error bound there and the entry condition for the next layer in Choice 13. The paper defers the final parameter optimization to Section 6, and without an explicit admissible tuple one cannot be certain that the inequalities close simultaneously for every α < sqrt(4/3)-1. This does not make the result implausible; it makes it conditional on a concrete quantitative verification. I therefore recommend keeping the reader's CONDITIONAL verdict rather than accepting or rejecting on the present text.","tokens_in":87770,"tokens_out":7318,"duration_ms":86470,"concrete_test":"Independently verify the deferred parameter optimization: extract all inequalities from Propositions 9–10 and Sections 4–5, and for representative values (e.g., α = 0.1 and α = 0.9·α*) find an explicit tuple (C,γ,kmax,δ,Λ,Y,ε,ζ,β,β',β'') satisfying them. For each n, check the third-phase requirement C^{-ββ' min{β'',1-β''}δγ(1/(1-γ))^{n-1}} ≤ (1/2) C^{-kmax(1/(1-γ))^n} and that the interval defining t*_n in inequality (108) is nonempty. If no such tuple exists, or the admissible γ is bounded away from 1 as α → α*, the claimed range fails.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central bridge is Proposition 10: the actual layer separations Ξ^(n)(t) must track the toy pendula Ξ_0^(n)(t) through each interval [t_n,1], because the force bounds in Sections 4–5 are computed for the toy configurations and the entry conditions for layer n+1 are taken from Choice 13/Remark 20. In the first half of [t_n,1] the error is controlled relative to sin(a_{n-1}(1)Ξ_0^(n)(t)); in the second half this relative control degenerates near π, and Proposition 10 switches to a uniform bound on [t*_n,1]. The definition of t*_n via inequality (108), 1/(π - a_{n-1}(1)Ξ^(n)(t)) ≤ K_n, and the assertion of point 3 belong to the most delicate part of the argument, and the final parameter optimization in Section 6 is deferred. The check that has to land is quantitative: with M_n = Y C^{δ(1/(1-γ))^n} (Choice 14), one must have K_n = C^{ββ'β''δγ(1/(1-γ))^{n-1}} and the resulting uniform error 14 C^{-ββ' min{β'',1-β''}δγ(1/(1-γ))^{n-1}} small compared with the distance from a_{n-1}(1)Ξ_0^(n)(1) to π, which is C^{-kmax(1/(1-γ))^n}. No contradiction is visible, but the interval of admissible (C,γ,kmax,δ,Λ,Y) has not been exhibited, and the claimed range of α below sqrt(4/3)-1 depends on it.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs compactly supported classical solutions of the forced inviscid 2D Boussinesq system on [0,1)×R^2 that develop a finite-time singularity at t=1, while the external forces retain Hölder regularity up to the singular time: the vorticity force belongs to C^0_t C^α_x and the density force to C^0_t C^{1,α}_x for every α<√(4/3)-1, with both forces compactly supported. The construction uses an infinite sequence of vorticity/density layers of decreasing spatial scale, introduced at times t_n accumulating at 1. The layer centers are designed to follow approximately the dynamics of a chain of degenerate inverted half-pendula, and the density is turned on and off in each layer so that the vorticity accumulates a hysteresis-like growth. The forces are defined as residuals of the layer equations, so their regularity is derived rather than assumed. The proof is organized as follows: Section 2 defines the stream function, density, and parameter choices; Section 3 proves that the actual layer separations Ξ^(n)(t) and the shape parameters k_n(t) stay close to the toy model; Sections 4 and 5 bound the residual forces; Section 6 performs the final parameter optimization and closes the proof of Theorem 1.","tokens_in":88163,"tokens_out":6419,"duration_ms":71811,"significance":"If the proof is completed, this would be a notable result: a pen-and-paper construction of a forced Boussinesq blow-up in the whole plane with forces that remain Hölder regular at the singular time, have compact support, and achieve the non-small Hölder range α<√(4/3)-1. The mechanism, based on an infinite chain of degenerate pendula and flickering density, is original and differs from the self-similar constructions of Elgindi–Pasqualotto and from the earlier forced constructions for 3D Euler and IPM. A strength of the paper is that the solution and the forces are explicitly constructed, with the forces obtained as residuals and no computer-assisted verification. The extension to axisymmetric 3D Euler announced in Remark 6, if realized, would increase the impact of the method. However, the current text leaves two load-bearing pieces incomplete: the proof of the uniform closeness estimate in Proposition 10 and the explicit parameter feasibility in Section 6. For this reason my assessment is conditional: the strategy is plausible and no internal contradiction is visible, but the main theorem is not yet fully established in the manuscript under review.","major_comments":[{"comment":"Proposition 10 is the load-bearing bridge between the actual layer separations Ξ^(n)(t) and the toy model Ξ_0^(n)(t). The proof of point 3 defines t*_n using inequality (108), 1/(π−a_{n-1}(1)Ξ^(n)(t)) ≤ K_n, and then asserts a uniform error 14C^{−ββ′min{β″,1−β″}δγ(1/(1−γ))^{n−1}} on [t*_n,1]. To close the argument one must show, with K_n = C^{ββ′β″δγ(1/(1−γ))^{n−1}} as proposed in the text, that the interval [t*_n,1] is nonempty and that this uniform error is small compared with the distance from a_{n-1}(1)Ξ_0^(n)(1) to π, which is C^{−k_max(1/(1−γ))^n} by Choices 12–14 and Remark 20. The manuscript does not verify these inequalities; it defers the admissible parameter set to Section 6. This is not a cosmetic omission, because the force bounds in Sections 4–5 are computed for the toy configurations and the entry conditions for layer n+1 are taken from Choice 13/Remark 20 on the basis of this closeness. I do not see a contradiction in the claimed estimates, but the parameter regime that makes them simultaneously true has not been exhibited.","section":"§3.2, Proposition 10 and Eq. (108)"},{"comment":"The theorem's quantitative range α∈(0,√(4/3)−1) depends on the existence of a nonempty admissible parameter set for C, γ, k_max, δ, Λ, Y, ε, ζ. Section 6 is presented as the place where the remaining parameters are optimized and the proof is closed, but the text provided stops short of displaying an explicit tuple of parameters or a chain of inequalities proving that all constraints coming from Proposition 9, Proposition 10, and the force bounds of Sections 4–5 are simultaneously satisfiable for every α below the stated threshold. In particular, the derivation of the critical exponent √(4/3)−1 from the parameter constraints is not shown. Since this exponent is the paper's headline improvement over previous forced constructions, the missing verification is load-bearing for Theorem 1.","section":"§6, final parameter optimization"},{"comment":"Remark 1 asserts that the local well-posedness argument of Chae–Kim–Nam extends to the forced case, but no proof is provided. If the theorem is meant to place the singularity 'in the well-posedness regime', the extension should be either proved or verified as a separate lemma; otherwise the claim in Remark 1 and the appeal to the blow-up criterion of [6] are not supported. The explicit construction itself may suffice to justify the existence of classical solutions on [0,1−ε], but the relationship between the constructed solutions and the asserted well-posedness theory should be made precise.","section":"§1.3, Remark 1"}],"minor_comments":[{"comment":"The line 'we choose kn(1) = kn(1) = 0' appears to contain a typo; it should presumably read 'kn(1)=0'.","section":"Choice 11, p. 55"},{"comment":"The abstract writes the Hölder exponent as C^{1,√(4/3)−1−ε}, while Theorem 1 states α∈(0,√(4/3)−1). It would be helpful to state explicitly that the ε in the abstract corresponds to the gap between α and the critical value.","section":"Abstract and Theorem 1"},{"comment":"The informal notation '∼' is used in several places to indicate order-of-magnitude equivalence without a precise definition. A short explanation of the convention would improve readability, especially for readers trying to follow the heuristic derivation of the toy model.","section":"§2.6, around Eq. (49)"}],"recommendation":"major_revision","confidential_remarks":"The stress-test concern about Proposition 10 and the deferred parameter optimization is, in my reading, valid and is the main reason I am not recommending acceptance. The construction is very intricate and the paper appears to be a serious attempt at a significant result; if the authors can complete Section 6 with an explicit admissible parameter set and a fully detailed proof of Proposition 10, the paper would likely be suitable for publication. I would also recommend that the authors clarify the status of Remark 1, since the paper currently both asserts the forced well-posedness extension and relies on it without proof."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper constructs compactly supported forced 2D Boussinesq solutions in R^2 that blow up at t=1, with f_rho in C^{1,alpha} and f_omega in C^alpha uniformly up to the blow-up time, for every alpha < sqrt(4/3)-1. The forces are residuals of the constructed fields, not fitted, and the compact supports are built in. That is a genuine new benchmark for singularities in the well-posedness regime with non-small Holder force regularity. The layer mechanism is clearly explained and honestly positioned as a sibling of [15], [17], [18]; the new twist is flickering density and degenerate pendula producing hysteresis in vorticity.\n\nThe proof is deliberately explicit: stream function, change of variables, toy models for layer centers and shapes, then long estimates for the forces. The paper flags several of its own gaps, which helps the reader: Remark 1 asserts without proof that local well-posedness extends to the forced case; Section 6 is a final parameter optimization that is not fully detailed; and Proposition 10 is the bridge between real dynamics and the toy models, with point 3 the delicate part. The stress-test note matches the text: no contradiction is visible in the parameter constraints, but the admissible set (C, gamma, k_max, delta, Lambda, Y) for all alpha below sqrt(4/3)-1 has not been exhibited. This is not a formality—the range of alpha comes out of that optimization.\n\nI did not verify the hundreds of inequalities, but the structure is honest: the force bounds are derived from the construction, and there is no visible circular step. The main concern is quantitative: uniform closeness in the second half of each layer interval, and whether the required K_n and error bounds fit with the distance to pi. That is exactly what a referee should check, and it is checkable.\n\nThis paper is for researchers working on singularity formation in Boussinesq and forced Euler-type systems. It deserves a serious referee; the likely outcome is heavy revision if the parameter feasibility needs more detail. I would send it out rather than desk-reject, and I would read the referee report on Section 6 and Proposition 10 carefully.","headline":"First whole-plane forced 2D Boussinesq blow-up with uniformly C^{1,alpha} force for all alpha below sqrt(4/3)-1; the construction is explicit and honest, but the load-bearing parameter feasibility in Section 6 needs referee scrutiny.","tokens_in":88738,"tokens_out":1889,"would_cite":true,"duration_ms":23689,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q31","35B44","76B03"],"pacs":[],"model":"deepseek-v4-flash","headline":"For every α below √(4/3)−1, smooth classical solutions of the forced 2D Boussinesq system blow up at t = 1 — the density-gradient integral diverges — while both forces keep uniform Hölder regularity up to the singular time.","keywords":["Boussinesq equations","finite-time blow-up","external force","Hölder regularity","degenerate pendulum","layer cascade","hysteresis","blow-up criterion"],"falsifier":"Evaluate the parameter constraints of Proposition 10 and Section 6 numerically for a fine grid of $\\alpha$ approaching $\\sqrt{4/3}-1$: if the admissible region (say, the maximum allowed $k_{\\max}$ as a function of $\\alpha$) closes strictly below $\\alpha^*$, the stated range of Hölder exponents collapses. Independently, simulate the two-layer toy model and compare the real displacement $\\Xi^{(n)}(t)$ with the explicit pendulum profile $\\sin(a_{n-1}(1)\\Xi^{(n)}_0(t)) = 1/\\cosh(\\hat t_{\\max}-\\hat t)$: the error should shrink as $n$ grows, and any deviation that grows with $n$ would contradict Proposition 10.","tokens_in":87566,"feed_emoji":"🌊","tokens_out":12383,"duration_ms":109782,"temperature":0.7,"pith_summary":"This paper tries to establish that the forced, inviscid, incompressible 2D Boussinesq equations admit classical solutions that develop a genuine singularity in finite time even though the external forces driving them stay perfectly regular up to (and at) the singular instant. Concretely, for every Hölder exponent $\\alpha \\in (0, \\sqrt{4/3}-1)$, there are smooth, compactly supported solutions on $[0,1) \\times \\mathbb{R}^2$ for which the integral of $\\|\\nabla \\rho\\|_{L^\\infty}$ diverges as $t \\to 1$ — a known blow-up criterion for this system — while the vorticity force remains in $C^\\alpha$ and the density force in $C^{1,\\alpha}$, uniformly in time with compact spatial support. The engine of the blow-up is a cascade of fluid layers of shrinking size, where each new density layer is switched on and then off and leaves behind a permanent increase in vorticity; the layer centers move according to an ideal model of 'degenerate' inverted half-pendulums, and the proof keeps the true dynamics close to that model. If the construction is correct, it is a finite-time singularity with a non-self-similar mechanism, produced by forcing that is more regular than the solution it destroys, and the same mechanism transfers to axisymmetric 3D Euler.","feed_headline":"Finite-time blow-up for 2D Boussinesq with regular forcing","feed_subtitle":"Its density gradient integrates to infinity at t = 1 while both forces keep their Hölder regularity throughout.","key_machinery":"The load-bearing mechanism is a hierarchy of fluid layers, each prescribed by a stream function $\\widetilde{\\psi}^{(n)}_n(t,x) = B_n(t)\\,\\varphi(\\lambda_n x_1)\\varphi(\\lambda_n x_2)\\sin(x_1)\\sin(x_2)$ written in coordinates that move and deform with the layer, so that the nonlinear transport is absorbed by a change of variables and the leading dynamics is read off from ODEs. The central identity is the degenerate-pendulum profile $\\sin F(t) = 1/\\cosh(t_{\\max}-t)$: after rescaling time, the relative displacement of consecutive layer centers solves $\\dot F = \\sin F$ with the distinguished energy $E=1$, so the sine of the displacement is a single hyperbolic pulse that peaks exactly at the midpoint of the layer's life. This pulse synchronizes the growth of the vorticity amplitude — the growth of a layer is the integral of the density pulse, which is why the vorticity increase survives after the density is switched off (hysteresis) — and it also fixes the shape parameter $k_n(t)$ through $\\int \\cos F\\,dt = \\ln(\\sin F(t)/\\sin F(0))$. The exponents $\\alpha^* = \\sqrt{4/3}-1$ come out of the balance among the superexponential layer scalings $a_n = C^{(1-k_n)(1/(1-\\gamma))^n}$, $b_n = C^{(1+k_n)(1/(1-\\gamma))^n}$, the superexponentially small cutoffs $\\lambda_n = C^{-\\Lambda(1/(1-\\gamma))^n}$ and the superexponentially short time steps $1-t_n \\sim C^{-\\delta(1/(1-\\gamma))^n}$, chosen so that every error term stays subordinate to the leading term up to $t=1$.","core_discovery":"The paper's central claim is Theorem 1: for every $\\alpha \\in (0, \\alpha^*)$ with $\\alpha^* = \\sqrt{4/3}-1$, there exist classical solutions $(u,\\rho)$ of the forced Boussinesq system on $[0,1)\\times\\mathbb{R}^2$ such that $u$ and $\\rho$ are smooth and compactly supported on every slab $[0,1-\\varepsilon]$, the vorticity force satisfies $f_\\omega \\in C^0_t C^{\\alpha}_{x,c}$ and the density force $f_\\rho \\in C^0_t C^{1,\\alpha}_{x,c}$ uniformly up to $t=1$, and $\\int_0^T \\|\\nabla\\rho(t,\\cdot)\\|_{L^\\infty}\\,dt \\to \\infty$ as $T\\to 1$. Since the divergence of that integral is the classical blow-up criterion for Boussinesq solutions, the singularity is genuine, and it occurs while both forces preserve their Hölder regularity at the blow-up time itself. The singular profile is odd-odd in the spatial variables, the forces are compactly supported, and the vorticity force is odd in $x_2$ so that the corresponding velocity force lies in $C^{1,\\alpha} \\cap L^2$.","pith_inferences":["The threshold $\\sqrt{4/3}-1$ is derived from the parameter balance rather than from the equation's scaling; it would be informative to test numerically whether the admissible parameter region (for instance, the maximum allowed $k_{\\max}$ as a function of $\\alpha$) actually closes up exactly at $\\alpha^*$, or whether a refined optimization could push the exponent higher.","The hysteresis principle suggests a transferable design rule: any forcing protocol that delivers the same localized pendulum pulse to each successive layer with well-separated scales should produce the same finite-time blow-up, so the specific density profile (the choice of the activation function $h^{(n)}$) is likely not essential.","Because the forces are compactly supported and the solution is classical away from the singular point, the construction offers a concrete target for numerical verification: the predicted layer displacement profile $\\sin(a_{n-1}(1)\\Xi^{(n)}_0(t)) = 1/\\cosh(\\hat t_{\\max} - \\hat t)$ is explicit enough to check in a two-layer simulation before testing the full cascade."],"forward_implications":["The integral of $\\|\\nabla \\rho\\|_{L^\\infty}$ — the established blow-up criterion for the Boussinesq system — is shown to diverge for classical solutions whose external forces stay uniformly Hölder ($C^\\alpha$ for the vorticity force, $C^{1,\\alpha}$ for the density force) up to the singular time, so the singularity is driven by the solution itself, not by the regularity of the forcing.","The blow-up happens in the well-posedness regime of the Boussinesq equations, so it is not an artifact of ill-posed initial data; the solution is smooth on every time slab $[0,1-\\varepsilon]$ and only loses regularity as $t\\to 1$.","At the blow-up time the solution loses regularity by an amount $r_{\\mathrm{loss}}$ that can be made arbitrarily small by choosing parameters, but with an upper bound that tends to $\\alpha^*/2$ as $\\alpha\\to 0$ and to $0$ as $\\alpha\\to\\alpha^*$.","The blow-up rate is not well defined: there are time sequences accumulating at $t=1$ along which $\\|\\omega\\|_{L^\\infty}$ and $\\|\\partial\\rho/\\partial x_2\\|_{L^\\infty}$ grow like different powers of $1/(1-t)$, including instants where $\\|\\partial\\rho/\\partial x_2\\|_{L^\\infty} = 0$.","The same layer construction, with the same degenerate pendulums, yields a finite-time blow-up for the axisymmetric 3D Euler equations, as the paper announces will appear in a forthcoming companion."],"supporting_citations":[{"why":"Supplies the blow-up criterion (divergence of the integral of the density-gradient L-infinity norm) that defines the singularity, and the local-existence class used in Remark 1.","marker":"[6]"},{"why":"The sibling force-based blow-up construction for 3D Euler with a uniform C^{1,1/2−ε} ∩ L² force, whose layer method this paper extends to Boussinesq.","marker":"[15]"},{"why":"Introduces the infinite chain of ODEs governing a family of vorticity bumps, the cascade idea behind the layer dynamics.","marker":"[16]"},{"why":"The sibling construction for the 2D porous media equation with smooth source, whose layer machinery is adapted here.","marker":"[18]"},{"why":"The closest existing finite-time blow-up for unforced 2D Boussinesq; the paper's main comparison point and the result it contrasts with on regularity and self-similarity.","marker":"[25, 26]"},{"why":"Establishes the correspondence between 2D Boussinesq and axisymmetric 3D Euler that supports the announced transfer of the construction.","marker":"[13]"}],"fun_headline_variants":["Blow-up for forced 2D Boussinesq despite Hölder-regular forcing","Infinite pendulum chain drives finite-time Boussinesq singularity","Regular forcing, singular outcome: Boussinesq blow-up at t=1","Hölder force persists as Boussinesq solution blows up in finite time","Wobbling pendula accumulate into 2D Boussinesq finite-time blow-up"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that a non-empty parameter regime exists in which the real layer-center dynamics stays close to the degenerate-pendulum toy model on every interval where the error estimates are made, while all the intertwined force-regularity inequalities hold at once — a feasibility check deferred to the final parameter optimization in Section 6 rather than exhibited in full.","fun_headline_variants_meta":{"raw":{"variants":["Blow-up for forced 2D Boussinesq despite Hölder-regular forcing","Infinite pendulum chain drives finite-time Boussinesq singularity","Regular forcing, singular outcome: Boussinesq blow-up at t=1","Hölder force persists as Boussinesq solution blows up in finite time","Wobbling pendula accumulate into 2D Boussinesq finite-time blow-up"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000172,"raw_usage":{"total_tokens":1273,"prompt_tokens":938,"completion_tokens":335,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":554,"completion_tokens_details":{"reasoning_tokens":243}},"tokens_in":554,"tokens_out":335,"duration_ms":3767,"temperature":1.0,"reasoning_tokens":243,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T13:42:52.082994+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the parameter constraints of Proposition 10 and Section 6 numerically for a fine grid of $\\alpha$ approaching $\\sqrt{4/3}-1$: if the admissible region (say, the maximum allowed $k_{\\max}$ as a function of $\\alpha$) closes strictly below $\\alpha^*$, the stated range of Hölder exponents collapses. Independently, simulate the two-layer toy model and compare the real displacement $\\Xi^{(n)}(t)$ with the explicit pendulum profile $\\sin(a_{n-1}(1)\\Xi^{(n)}_0(t)) = 1/\\cosh(\\hat t_{\\max}-\\hat t)$: the error should shrink as $n$ grows, and any deviation that grows with $n$ would contradict Proposition 10.","supporting_citations":[{"cited_title":"Chae, S-K","cited_arxiv_id":null,"evidence_quote":"Supplies the blow-up criterion (divergence of the integral of the density-gradient L-infinity norm) that defines the singularity, and the local-existence class used in Remark 1."},{"cited_title":"C´ ordoba, L","cited_arxiv_id":null,"evidence_quote":"Introduces the infinite chain of ODEs governing a family of vorticity bumps, the cascade idea behind the layer dynamics."},{"cited_title":"Constantin","cited_arxiv_id":null,"evidence_quote":"Establishes the correspondence between 2D Boussinesq and axisymmetric 3D Euler that supports the announced transfer of the construction."}],"review_version":1}