{"id":"f9862403-f5fb-4cee-8e6e-a25c7e1137fc","arxiv_id":"2501.06612","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Invariant measures of parabolic SPDEs with odd polynomial nonlinearities in the Da Prato-Debussche regime are non-Gaussian, proved by stationary generator identities.","lead":"A new algebraic proof shows that invariant measures of a family of singular stochastic PDEs, including the fractional Phi^4 models below dimension 14/5, are non-Gaussian. The method derives stationary identities from the SPDE generator instead of constructing the measures explicitly.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The theorem is conditional on Assumption 3.21, whose verification for the advertised Phi^4_delta measures is deferred; without it, Lemma 3.23 and hence Theorem 3.24 do not go through.","rationale":"We read the paper in good faith and focused on the central claim. The algebraic method is elegant and the proof of Theorem 3.24 is internally sound conditional on Assumptions 3.9, 3.19, and 3.21. The most fragile link is the moment assumption used to interchange limits in Lemma 3.23. This is exactly the assumption the reader identified. We found no hidden contradiction in the chaos computation or the covariance argument. The abstract's claim to cover Phi^4_delta measures is not fully supported because Assumption 3.21 is not verified or precisely cited for that family. This warrants a conditional verdict, which is what the reader recommended. Since our concern matches the reader's weakest assumption and does not suggest a different verdict, we leave the reader's CONDITIONAL verdict unchanged. If the proposed check succeeds, the paper can be upgraded; if not, the advertised scope should be narrowed.","tokens_in":22815,"tokens_out":24103,"duration_ms":229528,"concrete_test":"Verify Assumption 3.21 for the Phi^4_delta model with delta < 14/5: derive the uniform a priori bound ||v_1||_{C^gamma} <= (2+||Psi||_B)^m for the remainder equation (3.14), as in [MW20b, CMW23] (or [EW24] for the fractional setting), then apply Lemma 3.13 to deduce E_{u0~mu} 8u08_q^q < infinity for all q. If this bound is already available, cite the exact theorem and check that it covers the endpoint delta < 14/5; if not, the abstract and Theorem 3.24 should be restated as conditional on Assumption 3.21, or the missing estimates should be supplied.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing step is Lemma 3.23, which converts the pointwise generator formula of Lemma 3.20 into an identity under the invariant measure. Lemma 3.20 supplies the uniform-in-t bound |t^{-1}E[F(X_t)-F(X_0)]| <= (2+8u08_q)^q for t in (0,1/2). To pass t down to 0 inside E_{u0~mu} by dominated convergence, one needs E_{u0~mu} 8u08_q^q < infinity; this is precisely Assumption 3.21. The paper does not prove this assumption for the Phi^4_delta SPDEs advertised in the abstract (delta < 14/5). Remark 3.22 explicitly says verification is 'a separate task' and points only to external a priori estimates. Consequently, Theorem 3.24 is a conditional result: if Assumption 3.21 fails for one of the advertised models, the proof gives no non-Gaussianity conclusion for it, even though the statement may still be true. The rest of the argument appears internally consistent: conditional on Assumptions 3.9, 3.19 and 3.21, the Wiener-chaos computation leading to the contradiction integral (C phi)^p phi dx = 0 is sound. Thus the gap is not a mathematical error in the conditional theorem but an unfulfilled bridge between the theorem and the abstract's unconditional coverage claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper develops an elementary method, based on the generator equation at stationarity, to prove non-Gaussianity of invariant measures for parabolic SPDEs with polynomial nonlinearities in the Da Prato–Debussche (DPD) regime. The method is first presented in a classical one-dimensional setting (Section 2), where the generator identity is derived under mild assumptions and the Gaussianity contradiction is obtained via a Wiener–Itô chaos argument. The main result, Theorem 3.24, states that if Assumptions 3.9, 3.19, and 3.21 hold, then any invariant measure of the DPD-type SPDE (3.1) with odd-degree polynomial P of degree at least 3 is non-Gaussian. The paper also reviews the DPD solution theory, introduces a state space C for the Markov process, and discusses in Appendix A the relation between the DPD regime and singularity of the invariant measure with respect to the Gaussian free field.","tokens_in":23053,"tokens_out":9268,"duration_ms":75154,"significance":"Conditional on the assumptions, the theorem is clean and the method is genuinely different from existing non-Gaussianity proofs: it avoids skeleton inequalities and instead uses an integrated Dyson–Schwinger identity. The paper is also transparent about the role of the moment assumption, explicitly marking Assumption 3.21 as unverified for the Φ^4_δ models advertised in the abstract. If the missing moment estimate is supplied, the result would cover singular regimes where the invariant measure is mutually singular with respect to the Gaussian free field, which would be a valuable addition to the literature. The careful presentation of the DPD generator and the new state space C are useful contributions in their own right.","major_comments":[{"comment":"The proof of Theorem 3.24 depends critically on Assumption 3.21, since Lemma 3.23 passes the t→0 limit inside E_{u0∼µ} by dominated convergence using the bound from Lemma 3.20; however, Assumption 3.21 is not established for the Φ^4_δ measures with δ<14/5 that the abstract claims to cover. Remark 3.22 explicitly defers this verification to separate a priori estimates. The abstract's unconditional coverage claim is therefore not supported by the present proof; the theorem should be stated conditionally, or the moment estimate should be supplied.","section":"Section 3.3, Assumption 3.21 and Remark 3.22"},{"comment":"The global well-posedness and the bound (3.19) are assumed for the DPD equations, but the paper does not demonstrate that these hold for the Φ^4_δ models in the advertised range δ<14/5. The cited literature gives related a priori estimates, but the implication is not spelled out, so the applicability of Theorem 3.24 to those models is not fully established. This is a second assumption that stands between the theorem and the abstract's coverage claim.","section":"Section 3.1, Assumption 3.19"}],"minor_comments":[{"comment":"The statement that the state space has 'empty intersection with smooth functions' is inaccurate, since Lemma 3.13 implies that all sufficiently regular functions belong to C; please clarify the intended meaning.","section":"Introduction, page 3"},{"comment":"The existence of a nonzero eigenfunction of the covariance operator C and the approximation argument should be elaborated, since C is only established as a bounded operator L^q→L^{2p} and not as a compact operator on L^2.","section":"Section 3.3, proof of Theorem 3.24"},{"comment":"The mapping of parameters to [HKN24, Sec. 3] is quite compressed; a short explanation of the notation m, σ, k, and n_i would improve readability.","section":"Appendix A"},{"comment":"The measure µ in Lemma 2.5 is not introduced before its use; consider defining µ explicitly in the lemma statement.","section":"Section 2, Lemma 2.5"}],"recommendation":"major_revision","confidential_remarks":"The paper's main theorem is conditional on an unverified moment assumption that is load-bearing for the advertised applications. The authors should either prove the moment estimate for the Φ^4_δ models or carefully report the conditional status in the abstract and introduction. The mathematical core appears sound, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this one if you work on singular SPDEs or constructive QFT. The new thing is the method: instead of analysing the measure directly, they use the stationary generator equation to derive an infinite-dimensional Dyson–Schwinger identity and then push a Wiener-chaos computation to a contradiction. That is genuinely elementary and appears new. The classical d=1 section is a clean warm-up. In the DPD section they also do useful groundwork: a state space C of initial conditions for which the Markov process and generator make sense, stable under adding regular functions, and carrying full measure under the stationary linear solution. Nice.\n\nThe paper is honest about what is proved. Theorem 3.24 is a conditional theorem: if the invariant measure has finite moments of the DPD norm (Assumption 3.21), then it is non-Gaussian. The generator identity in Lemma 3.23 needs that moment bound to push t down to 0 inside the expectation, and that is exactly Assumption 3.21. The paper does not verify it for the Phi^4_delta measures advertised in the abstract; Remark 3.22 says verification is a separate task and points to a priori estimates elsewhere. So the abstract's claim to cover delta < 14/5 is too strong for what is proven here. This is a real gap, but it is a bridge, not a crack in the conditional theorem. The rest of the argument—the top-chaos computation and the eigenfunction contradiction—checks out as far as I can see.\n\nThe appendix relating the DPD regime to singularity with respect to the Gaussian free field is a good addition: it shows the non-Gaussianity result is not vacuous, since the invariant measure can be singular with respect to the linear solution's law. Citation practice is fair; they acknowledge prior non-Gaussianity proofs and do not oversell novelty. The Sine-Gordon discussion is speculative but appropriately flagged.\n\nWho is this for? Researchers working on stochastic quantisation, invariant measures of singular SPDEs, or constructive QFT. The method is simple enough to be taught and may transfer to other models. It deserves a serious referee. I would send it out, with the main request being: either prove or precisely import the missing moment estimates for Phi^4_delta, or soften the abstract and present Theorem 3.24 as the conditional statement it is. As written, it is a solid conditional result with an overreaching title claim.","headline":"Elegant algebraic proof of non-Gaussianity for DPD-regime SPDEs, but the advertised Phi^4_delta coverage rests on an unverified moment assumption.","tokens_in":23591,"tokens_out":1814,"would_cite":true,"duration_ms":18517,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60H15","60H40","81T08"],"pacs":[],"model":"deepseek-v4-flash","headline":"Every invariant measure of these singular SPDEs with odd polynomial nonlinearity of degree at least 3 is non-Gaussian.","keywords":["non-Gaussianity","invariant measures","stochastic quantization","Da Prato-Debussche regime","Wick powers","generator equation","Phi^4 measures","singular SPDEs"],"falsifier":"Because the argument reduces Gaussianity to the identity $\\int_{\\mathbb{T}^d}(C\\varphi)(x)^p\\varphi(x)\\,dx=0$ for all smooth $\\varphi$, where $C$ is the covariance operator of the stationary field and $p\\geq 3$ is odd, one can settle the claim by checking that identity for any candidate Gaussian invariant measure: for a translation-invariant $C$ on the torus, taking $\\varphi$ to be a positive-eigenvalue eigenfunction makes the integral strictly positive, so the identity fails and no such Gaussian measure can exist.","tokens_in":22575,"feed_emoji":"📐","tokens_out":8165,"duration_ms":74700,"temperature":0.7,"pith_summary":"This paper proposes an elementary, mostly algebraic method to prove that invariant (equilibrium) measures of parabolic stochastic PDEs with polynomial nonlinearities are not Gaussian, in the Da Prato–Debussche regime where solutions exist but are rough distributions. The method uses the generator equation of the SPDE at stationarity, an integrated Dyson–Schwinger identity, to derive moment recursions that no Gaussian law can satisfy when the polynomial has odd degree at least 3. The advertised application covers the $\\Phi^4_\\delta$ measures for $\\delta < \\frac{14}{5}$, including cases where the invariant measure is singular with respect to the Gaussian free field, so the result gives an elementary proof of interaction for stochastic quantization equations in a regime beyond absolute continuity.","feed_headline":"Singular SPDE equilibria proven non-Gaussian","feed_subtitle":"An algebraic generator trick proves non-Gaussianity for Phi^4_delta with delta < 14/5, even when singular.","key_machinery":"The load-bearing object is the generator equation at stationarity, $\\mathbb{E}_{u_0\\sim\\mu}[\\mathcal{L}F(u_0)] = 0$, where $\\mathcal{L}$ is the limit in Lemma 3.20: $$\\mathcal{L}F(u_0) = \\sum_i \\partial_i F(X_0)(\\langle u_0,(\\$\\Delta$-1)\\varphi_i\\rangle + \\langle :P(u_0):,\\varphi_i\\rangle) + \\frac12 \\sum_{i,j}\\$partial^{2}$_{ij}F(X_0)\\langle \\mathrm{Cov}_\\xi * \\varphi_i,\\varphi_j\\rangle.$$ This is an infinite-dimensional Euler–Lagrange equation, equivalently an integrated second-order integration-by-parts (Dyson–Schwinger) identity. Applied to $F(x)=x^k$, it yields the recursion $\\mathbb{E}[X^{k-1}(Y-Z)] = \\frac{k-1}{2}\\mathbb{E}[X^{k-2}]$ with $X=\\langle u_0,\\varphi\\rangle$, $Y=\\langle :P(u_0):,\\varphi\\rangle$, $Z=\\langle u_0,(\\Delta-1)\\varphi\\rangle$. Hermite polynomials $Q_k$ in the first Wiener chaos then force $\\mathbb{E}[Q_k(\\hat X)Y]=0$ for $k\\geq 2$, and at the top chaos $k=p$ the leading term $\\theta u_0^p$ of $:P(u_0):$ produces the integral $\\theta\\int (C\\varphi)^p\\varphi$, which cannot vanish for odd $p$.","core_discovery":"The central result is Theorem 3.24: under Assumptions 3.9, 3.19, and 3.21, if $P$ is a polynomial of odd degree $p \\geq 3$ and $u$ is the Markov process defined through the Da Prato–Debussche decomposition of $(\\partial_t + 1 - \\Delta)u = P(u) + \\xi$, then the invariant measure $\\mu$ is not Gaussian. The proof assumes $\\mu$ is Gaussian, writes the stationary field as $u_0 = g + \\Phi$ with $\\Phi$ a Gaussian free field and $g$ a regular remainder, and derives from the stationary generator equation the identity $\\int_{\\mathbb{T}^d} (C\\varphi)(x)^p \\varphi(x)\\,dx = 0$ for every smooth test function $\\varphi$, where $C$ is the covariance operator of $\\mu$. Because $p$ is odd and $C$ has positive eigenfunctions, the integral is strictly positive for a suitable $\\varphi$, a contradiction. The same argument, with a modified top-chaos computation, supplies non-Gaussianity for non-local polynomial models such as the $\\Phi^3_2$ measure.","pith_inferences":["We infer that if Assumption 3.21 is later established for $\\Phi^4_\\delta$ (for instance through coming-down-from-infinity estimates), Theorem 3.24 becomes unconditional for those measures; the paper explicitly leaves this verification open.","The top-chaos test is sensitive to the parity of $p$: for even-degree nonlinearities the same obstruction can vanish, suggesting that non-Gaussianity in those cases would need a different mechanism.","A similar generator-based obstruction could be sought for other singular Langevin dynamics whose nonlinearity has a single dominant Wick chaos, such as Yang–Mills–Higgs or tensor field theories, provided a state space and generator can be constructed."],"forward_implications":["For every $\\Phi^4_\\delta$ equilibrium measure with $\\delta < \\frac{14}{5}$, non-Gaussianity holds even when the measure is singular with respect to the Gaussian free field.","The stationary generator equation becomes a usable tool for SPDEs whose state space is a set of rough distributions, giving a second application of the Langevin dynamic to Euclidean quantum field theory.","The moment recursion in Lemma 3.23 is available as a quantitative stationarity identity for any invariant measure in this class, independent of the non-Gaussianity conclusion.","The method extends beyond local monomials to polynomial non-localities such as the normalised $\\Phi^3_2$ model, and the authors indicate that the sine-Gordon model is a plausible but nontrivial next case."],"supporting_citations":[{"why":"Supplies the Da Prato–Debussche decomposition u=v+Psi and local well-posedness that the paper extends to a Markov process on a rough state space.","marker":"[DPD03]"},{"why":"Identifies the generator equation as an integrated second-order Dyson–Schwinger/integration-by-parts identity, the algebraic core of the proof.","marker":"[BHST87]"},{"why":"Establishes that the Da Prato–Debussche regime is larger than the absolute-continuity regime and provides singularity results that frame the Phi^4_delta application.","marker":"[HKN24]"},{"why":"Provides the wavelet and Kolmogorov-type estimates used to prove convergence of Wick powers and that the state space C supports the generator.","marker":"[Hai14b]"},{"why":"Gives the coming-down-from-infinity estimates for Phi^4_2 that would verify Assumption 3.21 in that case, as noted in Remark 3.22.","marker":"[MW17b]"}],"fun_headline_variants":["Odd polynomial SPDEs lack Gaussian invariant measures","Algebraic generator argument rules out Gaussian equilibria","Singular SPDE measures proven non-Gaussian","Phi^4_delta with delta<14/5 is non-Gaussian","Non-Gaussian invariant measures from an algebraic identity"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof assumes the invariant measure has finite moments of the state-space norm, $\\mathbb{E}_{u_0\\sim\\mu}\\lVert u_0\\rVert_{\\mathcal{C}}^q < \\infty$ for every finite $q$ (Assumption 3.21), because only then can the $t\\to 0$ limit be interchanged with the expectation over $\\mu$ in Lemma 3.23; the paper does not verify this for the $\\Phi^4_\\delta$ measures and cites it as a separate task.","fun_headline_variants_meta":{"raw":{"variants":["Odd polynomial SPDEs lack Gaussian invariant measures","Algebraic generator argument rules out Gaussian equilibria","Singular SPDE measures proven non-Gaussian","Phi^4_delta with delta<14/5 is non-Gaussian","Non-Gaussian invariant measures from an algebraic identity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000858,"raw_usage":{"total_tokens":3690,"prompt_tokens":875,"completion_tokens":2815,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":491,"completion_tokens_details":{"reasoning_tokens":2738}},"tokens_in":491,"tokens_out":2815,"duration_ms":22233,"temperature":1.0,"reasoning_tokens":2738,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:55:55.346026+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Because the argument reduces Gaussianity to the identity $\\int_{\\mathbb{T}^d}(C\\varphi)(x)^p\\varphi(x)\\,dx=0$ for all smooth $\\varphi$, where $C$ is the covariance operator of the stationary field and $p\\geq 3$ is odd, one can settle the claim by checking that identity for any candidate Gaussian invariant measure: for a translation-invariant $C$ on the torus, taking $\\varphi$ to be a positive-eigenvalue eigenfunction makes the integral strictly positive, so the identity fails and no such Gaussian measure can exist.","supporting_citations":[],"review_version":1}