{"id":"c850e2ec-ea30-4a92-b239-74eb8de7f7e2","arxiv_id":"1908.03490","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"For the quadratic stochastic nonlinear wave and heat equations on the two-dimensional torus, standard Da Prato-Debussche solution theory fails at noise roughness alpha = 1/2 (wave) and alpha = 1 (heat), before the scaling-critical values 3/4 and 2.","lead":"This paper studies two equations that add random noise to vibrating strings and to spreading heat, and compares how rough the noise can be before standard solution methods fail. It finds the wave equation gives out at a lower noise roughness than the heat equation, despite scaling heuristics predicting the opposite.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the ill-posedness claim is explicitly scoped to Da Prato-Debussche-type expansions, and the divergence proof for Psi_N is sound within that scope.","rationale":"The paper's central claim is not that the SNLW equation has no solutions for alpha>=1/2 in any conceivable sense; it is that the standard Da Prato-Debussche decomposition and its higher-order variants break down. The authors state this explicitly in 'Theorem' 1.1 and the abstract, and footnote 4 concedes the cancellation loophole. I examined Proposition 1.6 closely: the divergence is produced by high-to-low frequency interactions with |k|~|n-k|>>|n|, the X_k(n,t) are independent, their variances have the lower bound (4.64), and Kolmogorov's three-series theorem with the zero-one law yields almost-sure non-convergence of each relevant Fourier coefficient at small fixed times. This is sufficient to prevent any subsequence from converging in C([0,T];D'). The positive well-posedness theorems are supported by standard contraction arguments and the displayed regularity estimates. Thus I find no load-bearing technical defect. The reader's weakest assumption identifies exactly the same scope caveat; I agree that the caveat weakens the rhetorical phrasing 'well-posedness breaks' but not the paper's actual, carefully qualified claim. No change to the verdict is needed.","tokens_in":42939,"tokens_out":16320,"duration_ms":189939,"concrete_test":"Recompute the second-moment lower bound (4.64) for alpha=1/2, n=0, and fixed small t from (4.59)-(4.63), and explicitly sum sum_{|k|<=N} |k|^{-2} ~ log N. Then apply Kolmogorov's three-series theorem to confirm P(lim_N hat(Psi_N)(0,t) exists)=0. This isolates the key divergence step of Proposition 1.6 and verifies the almost-sure non-convergence that underlies the negative half of the paper.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The only substantive caveat is the one the authors themselves record in footnote 4: divergence of the second-order stochastic term in C([0,T];D') does not by itself rule out a solution whose singularities cancel between Psi and v. However, the paper's 'Theorem' 1.1 and abstract define ill-posedness as the failure of the standard expansion method, so the claim is not overstated once that scope is read carefully. I checked the proof of Proposition 1.6 in detail. The divergence is driven by high-to-low interactions with |k|~|n-k|>>|n|; the random variables X_k(n,t) are independent, their second moments satisfy the lower bound (4.64), and the resulting logarithmic or power divergence of the variance sum triggers Kolmogorov's three-series theorem and the zero-one law. This gives almost-sure non-convergence of the Fourier coefficient at small fixed times, which rules out any convergent subsequence in C([0,T];D'). No internal inconsistency or missing step in this argument was found. The positive well-posedness side uses standard energy and Schauder estimates and matches the stated alpha-ranges. The remaining gap is therefore the acknowledged cancellation possibility, which is outside the paper's stated scope and does not undermine the central claim as formulated.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the two-dimensional stochastic nonlinear wave equation (SNLW) and stochastic nonlinear heat equation (SNLH) with a quadratic nonlinearity forced by an alpha-fractional derivative of space-time white noise. For SNLW it proves local well-posedness for 0=1/2, so that the standard expansion-based solution theory breaks down. For SNLH it proves local well-posedness for 0=1. The thresholds 1/2 and 1 are strictly below the scaling-critical values 3/4 and 2 predicted by probabilistic scaling. The divergence is established by identifying a high-to-low frequency interaction, estimating the variance of the relevant Fourier coefficients, and invoking Kolmogorov's three-series theorem and zero-one law.","tokens_in":109,"tokens_out":13117,"duration_ms":188981,"significance":"The paper makes a valuable and precise contribution to the singular SPDE literature. Its central negative result is cleanly scoped: it shows failure of the Da Prato-Debussche-type expansion machinery rather than asserting non-existence of solutions, and the authors explicitly record the cancellation caveat in footnote 4. The divergence thresholds are proved from first principles with detailed estimates, and the multilinear smoothing result in Proposition 1.4 is of independent interest. The comparison between the wave and heat equations, with the wave equation becoming worse at a lower value of alpha, is a substantive new phenomenon. The proofs use standard tools (Besov paraproducts, energy and Schauder estimates, Wiener chaos, and Kolmogorov criteria) and appear careful and complete; no fitted constants or ad-hoc assumptions are introduced.","major_comments":[],"minor_comments":[{"comment":"The phrases 'well-posedness theory breaks' and 'SNLW is ill-posed' could be over-read as a statement about existence or uniqueness of solutions; the precise content is failure of the Da Prato-Debussche-type expansion method. Since footnote 4 already limits the statement, I suggest adding a short qualifier in the abstract to make this conditional nature visible without reading the footnotes.","section":"Abstract and Theorem 1.1"},{"comment":"The lower bound for E[|Psi_N(n,t)|^2] is established only under the condition t >> |n|^{-2}. To cover arbitrary T>0 in the claim of divergence in C([0,T];D'), the proof should add a sentence explaining that one chooses n in Z^2 with |n|^{-2} << T and then applies the displayed estimate at t=T. As written, the reader may question the small-T case.","section":"§5.3, proof of Proposition 1.9(ii)"},{"comment":"The symbol Psi is used for the full paraproduct sum Psi = Psi^< + Psi^= + Psi^>, while the preceding text uses Psi^= specifically for the resonant product. This overloaded notation can be confusing; please introduce the full product with a distinct notation or an explicit definition such as Psi_full.","section":"§3.1, equation (3.16)"},{"comment":"Lemma 2.2 is stated for 1>|n|. The random variables X_k(n,t) are independent, their second moments satisfy the lower bound (4.64), and the variance sum diverges logarithmically at α=1/2 and as a power for α>1/2. Kolmogorov's three-series theorem plus the zero-one law gives almost-sure non-convergence of the Fourier coefficient, which rules out convergence in C([0,T];D'). The argument is sound. The heat-side divergence uses the same mechanism and is also correct.\n\nThe soft spot is exactly the one the authors flag in footnote 4: divergence of Ψ in C_t D'_x does not by itself exclude the possibility that singularities of Ψ and the remainder v cancel to produce a continuous distribution-valued solution. They define ill-posedness as failure of the standard Da Prato–Debussche-type expansion, so the headline claim is accurate within that stated scope. If one wants failure of well-posedness in the broadest sense, that remains open. This is a real limitation but not a flaw, because the paper is explicit about it.\n\nMinor points: the paper is a case study on T^2 with a quadratic nonlinearity, so its significance is within the singular SPDE subfield rather than as a universal principle. Some proofs say 'as in GKO' rather than giving every detail, but the omitted steps are parameter-free technical estimates, and the divergence thresholds themselves are proved here. The 'meta'-theorem in Section 1 is a slightly theatrical way to state a precise result, but it is harmless.\n\nWho gets value: researchers working on stochastic nonlinear wave equations, Da Prato–Debussche expansions, or the gap between scaling predictions and actual solution theory. The paper also provides a clean example of why higher-order expansions can fail before the probabilistically critical threshold. No code or data, which is standard for pure mathematics; the computations are detailed enough to be checked.\n\nThis deserves a serious referee. I would send it to review and expect a constructive report, with the main discussion centered on the scope of the ill-posedness claim rather than on the validity of the proofs.","headline":"A solid, carefully scoped case study showing that the second-order stochastic term diverges for SNLW at α≥1/2 and for SNLH at α≥1, with the caveat about cancellation explicitly acknowledged.","tokens_in":43716,"tokens_out":1738,"would_cite":true,"duration_ms":18469,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35L71","35K15","60H15"],"pacs":[],"model":"deepseek-v4-flash","headline":"For quadratic stochastic wave and heat equations on the two-dimensional torus, the second-order stochastic term diverges at α=1/2 (wave) and α=1 (heat), so the standard expansion-based solution theory breaks before the scaling-critical…","keywords":["stochastic nonlinear wave equation","stochastic nonlinear heat equation","quadratic nonlinearity","fractional space-time white noise","renormalization","ill-posedness","multilinear smoothing","singular stochastic PDEs"],"falsifier":"Fix a spatial frequency $n$ and time $t>0$. The paper's proof requires the variances $\\mathbb E[|\\widehat{\\Psi_N}(n,t)|^2]$ to diverge as $N\\to\\infty$ for $\\alpha\\ge\\tfrac12$ (wave) or $\\alpha\\ge1$ (heat), with rate $N^{-2+4\\alpha}$ (or $\\log N$ at the threshold). A direct computation finding a subsequence with uniformly bounded variance, or an alternative renormalization in which the singularities of $\\Psi$ and $v$ cancel to produce a continuous distribution-valued $u$, would falsify the claim that standard well-posedness breaks at those thresholds.","tokens_in":42674,"feed_emoji":"🌊","tokens_out":12951,"duration_ms":115468,"temperature":0.7,"pith_summary":"This paper compares two singular stochastic PDEs on the two-dimensional torus: the stochastic nonlinear wave equation (SNLW) and the stochastic nonlinear heat equation (SNLH), both with quadratic nonlinearity $u^2$ and additive noise $\\langle\\nabla\\rangle^\\alpha \\xi$, where $\\xi$ is space-time white noise. It establishes that the standard expansion-based solution theory breaks down at $\\alpha = \\tfrac12$ for SNLW and at $\\alpha = 1$ for SNLH: the second-order stochastic term, the Picard second iterate, ceases to exist as a continuous distribution-valued function of time. This matters because scaling analysis had predicted critical values $\\alpha = \\tfrac34$ (wave) and $\\alpha = 2$ (heat), so the theory fails strictly earlier than expected; it also provides the first example in which the stochastic nonlinear wave equation behaves less favorably than its heat counterpart. For $\\alpha$ below these thresholds, the paper proves local well-posedness, with a simplified argument in an intermediate wave range thanks to a new multilinear smoothing estimate.","feed_headline":"Standard solution theory breaks at α=1/2 (wave), α=1 (heat)","feed_subtitle":"Scaling predicted critical α=3/4 for waves and α=2 for heat; the second-order term already diverges at 1/2 and 1.","key_machinery":"The load-bearing object is the second-order stochastic term $\\Psi = I(:\\Phi^2:)$, where $\\Phi = I(\\langle\\nabla\\rangle^\\alpha \\xi)$ is the stochastic convolution (the first Picard iterate), $:\\Phi^2:$ is the renormalized square with the divergent variance subtracted, and $I$ is the Duhamel integral operator: $(\\partial_t^2 + (1-\\Delta))^{-1}$ for the wave equation and $(\\partial_t + (1-\\Delta))^{-1}$ for the heat equation. The divergence proof shows that for fixed spatial frequency $n$ and time $t$, the variance $\\mathbb E[|\\widehat{\\Psi_N}(n,t)|^2]$ of the truncated Fourier coefficient grows with $N$ (like $N^{-2+4\\alpha}$, or like $\\log N$ at the threshold), driven by high-to-low frequency interactions $|k|\\sim|n-k|\\gg|n|$; Kolmogorov's three-series theorem and zero-one law then force almost-sure divergence. On the well-posedness side, the paper exploits multilinear dispersion of the wave propagator to obtain extra regularity $s_\\alpha = 1-\\alpha$ for $\\alpha\\le \\tfrac14$ and $s_\\alpha = \\tfrac54 - 2\\alpha$ for $\\alpha>\\tfrac14$, which allows a simplified contraction argument for $\\tfrac13\\le\\alpha<\\tfrac{5}{12}$.","core_discovery":"On $\\mathbb T^2$, consider the stochastic nonlinear wave equation $\\partial_t^2 u + (1-\\Delta)u + u^2 = \\langle\\nabla\\rangle^\\alpha \\xi$ and the stochastic nonlinear heat equation $\\partial_t u + (1-\\Delta)u + u^2 = \\langle\\nabla\\rangle^\\alpha \\xi$. The paper proves that the truncated second-order stochastic terms $\\Psi_N$ form a divergent sequence in $C([0,T];\\mathcal D'(\\mathbb T^2))$ almost surely for $\\alpha\\ge \\tfrac12$ in the wave case and for $\\alpha\\ge 1$ in the heat case (Propositions 1.6 and 1.9(ii)). Below those thresholds the same objects converge, and local well-posedness of the renormalized equations holds; in the wave case the construction is simplified by an extra multilinear smoothing of order $\\tfrac14$ on $\\Psi$ (Proposition 1.4). The paper concludes that the standard expansion trick and its higher-order variants break at $\\alpha = \\tfrac12$ and $\\alpha = 1$, before the scaling-critical values $\\alpha = \\tfrac34$ and $\\alpha = 2$ predicted by probabilistic scaling.","pith_inferences":["A fair reading of the breakdown is method-specific: the paper proves that expansion-based solution theories cannot work, not that the stochastic PDE has no solution; footnote 4 explicitly leaves open cancellation between the singularities of $\\Psi$ and the remainder $v$.","The divergence mechanism—high-to-low energy transfer in the Picard second iterate—mirrors deterministic ill-posedness mechanisms; one could try to turn it into a norm-inflation construction for SNLW at $\\alpha\\ge\\tfrac12$, and to probe whether any non-expansion theory exists there.","The same gap between scaling-critical values and Picard-second-iterate divergence is expected for other low-degree dispersive equations with rough random data, such as quadratic nonlinear Schrödinger equations; the paper hints at this expectation, but testing it is a separate project."],"forward_implications":["For $\\alpha\\ge\\tfrac12$, no solution of the quadratic SNLW on $\\mathbb T^2$ can be constructed by the standard expansion trick or its higher-order variants; for $\\alpha\\ge1$ the same holds for SNLH.","The local well-posedness ranges $0<\\alpha<\\tfrac12$ (wave) and $0<\\alpha<1$ (heat) are sharp within this method, so the effective thresholds are $\\tfrac12$ and $1$, not the scaling-critical $\\tfrac34$ and $2$.","The wave equation behaves worse than the heat equation: $\\Psi$ diverges at the same $\\alpha=\\tfrac12$ as the renormalized square $:\\Phi^2:$, whereas in the heat case $\\Psi$ survives until $\\alpha=1$ even after $:\\Phi^2:$ fails at $\\alpha=\\tfrac12$.","For $\\tfrac13\\le\\alpha<\\tfrac{5}{12}$, the multilinear smoothing lets the wave equation be solved with a smaller enhanced data set, omitting the third-order object that would otherwise be needed.","On $\\mathbb T^d$, the divergence of $\\Psi$ holds for $\\alpha\\ge 1-\\tfrac d4$, so the wave breakdown precedes the scaling-critical value in dimensions $d=1,\\dots,5$."],"supporting_citations":[{"why":"Supplies the local well-posedness baseline for $0<\\alpha<1/2$ that the paper refines, along with an earlier variance-divergence result for $\\Psi_N$.","marker":"[22]"},{"why":"Shows how multilinear dispersion of the wave propagator yields extra smoothing for stochastic terms; the method is adapted in Proposition 1.4.","marker":"[30]"},{"why":"Gives the probabilistic scaling analysis that predicts the critical value $\\alpha=3/4$ which the paper shows is not attained.","marker":"[20]"},{"why":"Provides the scaling/regularity framework whose critical value $\\alpha=2$ for the heat equation the paper shows is too optimistic.","marker":"[33]"},{"why":"Introduces the expansion trick whose breakdown for $\\alpha\\ge1/2$ and $\\alpha\\ge1$ is the paper's main ill-posedness conclusion.","marker":"[17]"},{"why":"Supplies Kolmogorov's three-series theorem and zero-one law, which convert divergence of variances into almost-sure divergence.","marker":"[23]"},{"why":"Provides the stochastic regularity lemma used to pass from Fourier-coefficient moment bounds to Besov-space continuity and convergence.","marker":"[40]"},{"why":"Gives the parabolic example in which a renormalized square fails but its heat-smoothed version exists; the contrast highlights the wave equation's worse behavior.","marker":"[24]"}],"fun_headline_variants":["Wave ill-posed at α=1/2, heat at α=1, below scaling values","Stochastic wave/heat break at α=1/2 and α=1, not α=3/4,2","Rough noise: second-order term diverges at α=1/2 (wave), α=1 (heat)","Solution theory fails sooner: wave α=1/2, heat α=1","Wave worse than heat: α=1/2 vs α=1 breakdown"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument equates failure of the expansion method—specifically divergence of the second-order stochastic term $\\Psi$—with ill-posedness of the equation itself; if the singularities of $\\Psi$ and the residual term $v$ could cancel, a continuous distribution-valued solution might still exist, a case the paper explicitly sets aside.","fun_headline_variants_meta":{"raw":{"variants":["Wave ill-posed at α=1/2, heat at α=1, below scaling values","Stochastic wave/heat break at α=1/2 and α=1, not α=3/4,2","Rough noise: second-order term diverges at α=1/2 (wave), α=1 (heat)","Solution theory fails sooner: wave α=1/2, heat α=1","Wave worse than heat: α=1/2 vs α=1 breakdown"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000408,"raw_usage":{"total_tokens":2238,"prompt_tokens":1187,"completion_tokens":1051,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":803,"completion_tokens_details":{"reasoning_tokens":926}},"tokens_in":803,"tokens_out":1051,"duration_ms":9722,"temperature":1.0,"reasoning_tokens":926,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:11:21.892048+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Fix a spatial frequency $n$ and time $t>0$. The paper's proof requires the variances $\\mathbb E[|\\widehat{\\Psi_N}(n,t)|^2]$ to diverge as $N\\to\\infty$ for $\\alpha\\ge\\tfrac12$ (wave) or $\\alpha\\ge1$ (heat), with rate $N^{-2+4\\alpha}$ (or $\\log N$ at the threshold). A direct computation finding a subsequence with uniformly bounded variance, or an alternative renormalization in which the singularities of $\\Psi$ and $v$ cancel to produce a continuous distribution-valued $u$, would falsify the claim that standard well-posedness breaks at those thresholds.","supporting_citations":[{"cited_title":"Paracontrolled approach to the three-dimensional stochastic nonlinear wave equation with quadratic nonlinearity","cited_arxiv_id":"1811.07808","evidence_quote":"Shows how multilinear dispersion of the wave propagator yields extra smoothing for stochastic terms; the method is adapted in Proposition 1.4."},{"cited_title":"Da Prato, A","cited_arxiv_id":null,"evidence_quote":"Introduces the expansion trick whose breakdown for $\\alpha\\ge1/2$ and $\\alpha\\ge1$ is the paper's main ill-posedness conclusion."},{"cited_title":"Durrett, Probability–theory and examples, Fifth edition","cited_arxiv_id":null,"evidence_quote":"Supplies Kolmogorov's three-series theorem and zero-one law, which convert divergence of variances into almost-sure divergence."},{"cited_title":"Mourrat, H","cited_arxiv_id":null,"evidence_quote":"Provides the stochastic regularity lemma used to pass from Fourier-coefficient moment bounds to Besov-space continuity and convergence."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the parabolic example in which a renormalized square fails but its heat-smoothed version exists; the contrast highlights the wave equation's worse behavior."}],"review_version":1}