{"id":"7690b255-fd89-4c8a-9882-33cec40a92eb","arxiv_id":"2507.03575","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Solutions of the renormalized singular-degenerate stochastic porous media equation satisfy L1-based Besov regularity of order α and modelledness of order >1, uniformly in the regularization, for noise of regularity α−2 with α∈(2/3,1).","lead":"The paper proves a priori regularity bounds for a stochastic porous media equation that is both degenerate (diffusion vanishes with the solution) and singular (noise too rough for classical products). This removes a key obstacle to a solution theory for this class by combining kinetic methods with regularity structures and a new energy estimate.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The paper's a priori bounds are conditional on Assumption 1.2(ii); its verification for the two examples is partially deferred to the PhD thesis [dLF24], and the renormalized-model estimates are the least secure load-bearing point.","rationale":"I read the proof in good faith and traced the main chain: the kinetic formulation, the small/large velocity decomposition (Propositions 2.1 and 2.3), the reconstruction and integration lemmas of Section 3, the four large-velocity propositions, the renormalized energy inequality (Proposition 4.4), and the real interpolation argument of Section 5. The conditional argument is coherent: the exponent balance in (1.12) ensures the interval (2−α, 2α/(1+(M−1)α)) is nonempty and that the absorption in Lemma 2.8 closes; the counterterm cancellations in (1.16) and the renormalized energy identity are algebraically consistent; and the uses of the model bounds (1.8)-(1.9) appear in the right places with the right powers. I did not find an internal inconsistency or a concrete gap in the proof of the conditional theorems. The genuinely soft point is the verification of Assumption 1.2 for the examples, which the paper partially defers to the PhD thesis [dLF24]. This is exactly the load-bearing premise flagged by the reader: if the deferred estimates fail or are not uniform in the mollification scale, the theorems lose their concrete instances in the singular regime. The reader's verdict of ACCEPT with moderate confidence is consistent with this assessment; my stress-test does not change that verdict, but it does isolate the one check that would most increase confidence in the paper's headline claim.","tokens_in":68189,"tokens_out":52672,"duration_ms":528790,"concrete_test":"Independently recompute, for Example 1.11 (space white noise on T2), the Gaussian moment bounds for the pairings <∂_a^m Π_x[;a], φ^λ_x> and <∂_a^m Π_x[;a], φ^λ_x> with m=0,1, tracking the cutoff ε, the test scale λ, and the powers of a, and verify that (1.8)-(1.9) hold with constants independent of ε. Then verify the time-integrability of (1.10)-(1.11) by direct integration of the displayed bounds in Example 1.11. If every bound holds uniformly in ε, Assumption 1.2 is confirmed for the example and the central claim lands.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorems 1.4 and 1.5 are explicit conditional statements: if the enhanced model satisfying (1.8)-(1.9) and the counterterm closeness (1.10)-(1.11) exists, then the a priori bounds hold. The paper's claim to treat singular stochastic porous media equations therefore turns on Assumption 1.2 being satisfied for concrete noises. The two examples (space white noise on T2, and white-in-time coloured-in-space noise) provide explicit formulas for the counterterms and for the model, but several key estimates, in particular the verification of (1.8) for the renormalized symbols and the derivative-in-a bounds with m=1, are referred to the PhD thesis [dLF24] rather than proved in the paper. Since the cancellations in (1.16) and the renormalized energy inequality (Proposition 4.4) rely precisely on these bounds with the stated a-dependence and ε-uniformity, any error or non-uniformity in the deferred estimates would mean the theorem has no verified instance in the singular regime. Remark 1.8 likewise admits that the full list of necessary conditions (I)-(XVI) is long, and Assumption 1.1 is a simplification. This is the load-bearing premise of the paper's central claim, and it is the least secure point in the chain.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a priori Hölder-type regularity estimates in L1 for solutions of the regularized renormalized stochastic porous medium equation (1.1) with multiplicative noise of space-time Hölder regularity α−2 for α∈(2/3,1). Under Assumptions 1.1–1.3, Theorem 1.4 states that the solution u satisfies ∫_{D_y}|u(x+y)−u(x)| ≲ ||y||_s^α, and Theorem 1.5 states a higher-order modelledness estimate with respect to the solution of the linear stochastic heat equation. The proof uses the kinetic formulation, a splitting into small and large velocities, new reconstruction/integration lemmas for kernels composed with rough solutions, a renormalized energy inequality (Proposition 4.4), and a real interpolation argument for basepoint-dependent seminorms. The theorems are conditional on an enhanced model satisfying Assumption 1.2, which is verified for two examples (space white noise on T^2 and white-in-time coloured-in-space noise).","tokens_in":68445,"tokens_out":12201,"duration_ms":133552,"significance":"If correct, these are the first a priori estimates establishing positive space-time regularity and modelledness for SPDEs that are both degenerate and singular. The proof strategy is novel, combining kinetic formulations with the theory of regularity structures, and the paper is largely self-contained from Section 2 onward. The abstract results in Section 3, especially Lemmas 3.2–3.5, are stated with full proofs and carefully track the dependence on the diffusion coefficient. The main estimates are derived from explicit assumptions rather than fitted to the conclusion, so there is no circularity. The main caveat is that the verification of Assumption 1.2 for the two examples relies on the PhD thesis [dLF24] for several central bounds.","major_comments":[{"comment":"The verification of Assumption 1.2(ii) for the two examples is incomplete within the paper: the bound (1.8) for the renormalized symbols and the derivative-in-ā estimates with m=1 are deferred to the PhD thesis [dLF24] (e.g., 'we refer the reader to [dLF24, Section 3.3.1] for further details'). These bounds are load-bearing because the cancellations in (1.16) and the renormalized energy inequality in Proposition 4.4 rely on the exact ā-dependence and ε-uniformity of the enhanced model. Without them, the theorem has no fully verified instance in the singular regime. Please include the missing estimates or state precisely which results of [dLF24] are used and how they imply (1.8)–(1.9) with the required uniformity.","section":"§1.2, Examples 1.11 and 1.12"},{"comment":"The estimate of ∥ν∥_{L1} in the proof of Theorem 1.4 uses a uniform-in-time bound on ∥u(t)∥_{L1(T^d)}, citing the proof of Proposition 4.4. However, Proposition 4.4 as stated only gives an estimate for the final-time energy ∫ G(u(T)) dz. Please clarify that the energy identity in the proof of Proposition 4.4 is applied on an arbitrary interval [0,t] to obtain the sup_t bound needed here.","section":"§5, proof of Theorem 1.4"}],"minor_comments":[{"comment":"The notation for the two counterterms C_{\\bar a}(s) and C^{\\bar a}(s) is easy to confuse in equations such as (1.16); consider renaming them or adding a small table that fixes their definitions and homogeneities.","section":"§4.3 / Proposition 4.4"},{"comment":"The abuse of notation concerning the constant [ ]_{2α−2} in (4.5) is acknowledged but the precise combination [ ]_{2α−2} + [ ]_α[ ]_{α−2} could be written out in the display to help the reader track the provenance of constants.","section":"§4, equation (4.5)"},{"comment":"For the white-in-time coloured-in-space example, the verification of (1.9) for time increments and for derivatives in ā is only sketched; a few more details, or a precise pointer to the corresponding pages in [dLF24], would improve readability.","section":"§1.2, Example 1.12"},{"comment":"Since inequalities (I)–(XVI) are used throughout Section 4, it would be helpful to collect them in an appendix or table rather than leaving the reader to reconstruct them from the text, especially because Remark 1.8 states that Assumption 1.1 is a simplification of these conditions.","section":"Remark 1.8"}],"recommendation":"major_revision","confidential_remarks":"The main risk is the reliability of the estimates deferred to the PhD thesis [dLF24] for Examples 1.11 and 1.12. Since the thesis is from the same research group and is recent, I recommend that the editor arrange an independent check of the specific bounds used to verify Assumption 1.2, particularly (1.8) for the renormalized symbols and the m=1 derivative bounds. The remainder of the proof is detailed and appears sound; the requested additions to the paper should not change the architecture of the argument."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Markus,\n\nThis is a genuinely new result: the first a priori bounds of positive regularity/modelledness for an SPDE that is both singular and degenerate. The authors combine the kinetic formulation with regularity structures, introduce a renormalized energy inequality, and extend real interpolation to basepoint-dependent seminorms. That is real progress, and the paper is honest about what it does and does not prove.\n\nWhat is well done: the theorems are precise, the assumptions are explicit, and the proof is self-contained except for one point. The reconstruction and integration lemmas in Section 3 are careful and trace the dependence on the diffusion coefficient, which matters in the degenerate setting. The examples give concrete counterterm formulas and verify the integral conditions (1.10)-(1.11) directly. The paper also states its limitations clearly: the non-optimal restriction on M, the extra N ≥ M+1 assumption, and the simplification that collapses the long list of conditions into Assumption 1.1.\n\nNow the soft spot. The stress-test concern about Assumption 1.2(ii) is legitimate: the enhanced model bounds and counterterm closeness are load-bearing, and part of the verification for the two examples is deferred to a PhD thesis [dLF24]. But I think the concern is moderate, not fatal. The examples provide explicit formulas for the counterterms and outline the probabilistic estimates; what is left to the thesis are mostly moment bounds and Kolmogorov continuity arguments. That is standard practice in this literature, and the dependency is clearly stated. This is not a circular argument or a fitting procedure. A referee should ask to see the missing details, but the conditional structure is mathematically sound.\n\nOther weaknesses are minor. The proof is long and technical, but the structure is readable. The interpolation step in Section 5 is genuinely new and appears valid. I did not find a load-bearing flaw in the core estimate.\n\nThis paper is for researchers in singular SPDEs, especially those working on degenerate equations via kinetic methods. It deserves a serious referee. I would send it out and would expect the referee to spend time on the deferred model verification.\n\nBest","headline":"Serious and original a priori bounds for the first genuinely singular-and-degenerate regime; the proof is conditional on partly deferred model verification, but the core argument is sound and deserves a careful referee.","tokens_in":68961,"tokens_out":2107,"would_cite":true,"duration_ms":28901,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60H17","35K65","60L30","35B45"],"pacs":[],"model":"deepseek-v4-flash","headline":"Stochastic porous media equations with rough multiplicative noise satisfy explicit a priori L¹-Hölder-α space-time bounds, uniform in the smoothing, and a stronger modelledness estimate to order β.","keywords":["stochastic porous medium equation","degenerate singular SPDE","regularity structures","kinetic formulation","a priori bounds","modelledness","renormalized energy inequality","L1 Besov regularity"],"falsifier":"Evaluate the integrals in (1.10)–(1.11) for space white noise on T² using the explicit counterterms of Example 1.11 with σ(v)=|v|^M (violating N≥M+1) at the boundary case of (1.12): the paper's balancing inequality (II) fails at that point, so the integrals' divergence would confirm the restriction on σ is load-bearing, while their finiteness would show the admissible class is wider than stated.","tokens_in":67986,"feed_emoji":"🌊","tokens_out":17984,"duration_ms":188150,"temperature":0.7,"pith_summary":"The paper proves the first positive regularity estimates for a stochastic PDE that is simultaneously singular (noise too rough for classical products) and degenerate (the diffusion coefficient a(u)=M|u|^{M−1} vanishes where u=0): the stochastic porous media equation with multiplicative noise of space-time regularity α−2, α∈(2/3,1). The main result, Theorem 1.4, is an a priori bound: the solution of the regularized, renormalized equation satisfies ∫_{D_y}|u(x+y)−u(x)| ≲ ||y||_s^α, with the constant independent of the qualitative smoothness of the coefficients, noise, and initial data. Theorem 1.5 strengthens this to higher-order modelledness over the solution of the corresponding linear stochastic heat equation, to order β ∈ (2−α, 2α/(1+(M−1)α)) > 1. A sympathetic reader should care because uniform a priori bounds of this kind are the ingredient needed to pass from regularized problems to a genuine solution theory in the singular regime, and because the proof introduces a new renormalized energy inequality plus an L¹-based reconstruction machinery that should apply to other degenerate singular SPDEs.","feed_headline":"First regularity bounds for singular degenerate porous-media SPDEs","feed_subtitle":"Degenerate diffusion and rough noise: solutions retain Hölder-α regularity, uniformly in the smoothing.","key_machinery":"The argument runs on four connected mechanisms. The kinetic formulation: the indicator function χ(t,x,v)=1_{(−∞,u(t,x))}(v)−1_{(−∞,0)}(v) converts the degenerate equation into a linear transport equation (1.14), and the splitting u=u_<+u_> at a(u)≍δ separates the trivially controlled small-velocity part from the effectively non-degenerate large-velocity part. The regularity-structures model Π — the noise ξ, its heat-flow integral, the two renormalized singular products ξ·(integral) and |∇(integral)|², and the first-order symbol X — encodes the renormalized singular products σ(u)ξ and a(u)|∇u|²; the counterterms C̄_ā and C̄̄_ā enter through the closeness conditions (1.10)–(1.11). The renormalized energy inequality of Proposition 4.4 tests (1.1) with |u|^ε and bounds ∫ g′(u)a(u)|ν|², the good-sign quantity controlling the renormalized kinetic measure, using the fact that the counterterm for u(σ(u)ξ−σ′σC_a) and for aσ²|∇Π|² coincide up to sign and an integrable time blow-up. Finally, the reconstruction and integration lemmas of Section 3 (Lemmas 3.2–3.5) are L¹-based versions of reconstruction that track the blow-up in the diffusion coefficient a, and Lemma 5.1 extends real interpolation to basepoint-dependent seminorms, which is what turns the two-parameter δ-balancing into the final exponent β.","core_discovery":"The central claim is that in the first singular regime α∈(2/3,1), under the restriction M < 1+(3α−2)/(α(2−α)) and the enhanced-noise and counterterm-closeness assumptions of Assumption 1.2, the classical solution u of the regularized renormalized equation (1.1) satisfies the modelledness estimate of Theorem 1.5, hence the L¹ space-time regularity estimate ∫_{D_y}|u(x+y)−u(x)| dx ≲ ||y||_s^α of Theorem 1.4, with the optimal exponent α and with constants uniform in the qualitative smoothness of a, ξ, u0. The proof splits the solution into small and large velocities through the kinetic formulation, shows through the cancellation in (1.16) that the two singular products — the forcing σ(u)ξ and the kinetic measure a(u)|∇u|² — require the same counterterm up to an integrable blow-up at t=0, and controls the renormalized kinetic measure by a new energy inequality (Proposition 4.4) obtained by testing the equation with |u|^ε. Real interpolation, extended to basepoint-dependent modelledness seminorms (Lemma 5.1), combines the small-velocity decay $δ^{{1/(M−1)}}$ with the large-velocity blow-up $δ^{{−α−ε}}$, yielding the order 2α/(1+(M−1)(α+ε)), and a buckling argument proves modelledness of σ(u)ξ and hence, after integration, of u itself; the two requirements are compatible exactly when (1.12) holds.","pith_inferences":["The a priori bound appears to be the missing ingredient for a convergence theory: one expects that the same estimates, combined with compactness in L¹-Besov spaces, would produce kinetic solutions of the unrenormalized degenerate singular equation at α>2/3 — the paper stops at the bound, so the limit passage is not yet written down.","The counterterm cancellation between the forcing and the kinetic measure is plausibly a general design principle: any degenerate singular SPDE whose noise intensity satisfies σ²(v)≍a(v) near the degeneracy should admit an analogous renormalized energy inequality, suggesting the method transfers to degenerate p-Laplacian or thin-film type flows.","The integrable blow-up in conditions (1.10)–(1.11) is an artifact of starting the auxiliary heat flow at time zero; choosing a stationary model would make both differences vanish, so a stationary formulation could yield the same theorems with cleaner constants.","A numerical check is available: on T² with space white noise and a(u)=M|u|^{M−1}, the uniform L¹-Besov seminorm should stay bounded below the threshold (1.12) and blow up above it, which would directly test whether the restriction on M is sharp."],"forward_implications":["The a priori bounds pass to kinetic solutions constructed by approximation: by Remark 1.6 the estimate transmits immediately when u is obtained as a limit of classical solutions, giving the same L¹-Hölder-α control for the weaker notion of solution.","The space-time regularity exponent α is optimal for noise of regularity α−2, so the bound cannot be improved merely by refining the argument at fixed noise regularity.","The modelledness order β exceeds 2−α, hence exceeds one, meaning the solution is captured to first order by the linear stochastic heat flow plus a correction ν·y — a much finer description than a plain Hölder bound.","The two canonical noises — space white noise on the torus T² and white-in-time, spatially coloured noise — fall under Assumption 1.2, so the theorems give explicit, uniform a priori estimates for those cases.","The real-interpolation lemma for basepoint-dependent seminorms is a reusable device for turning regularity-structures estimates into L¹-Besov regularity in degenerate settings."],"supporting_citations":[{"why":"establishes the kinetic-formulation route to higher-order regularity for the deterministic porous medium equation; the present proof of Theorem 1.5 explicitly follows its strategy.","marker":"[Ges21]"},{"why":"obtains optimal time-space regularity for stochastic porous media equations with trace-class noise via Itô calculus and provides the L²-based mild-formulation estimates that the present L¹/rough-forcing argument extends.","marker":"[BGW22]"},{"why":"introduces the small/large velocity splitting and velocity averaging for kinetic formulations of quasi-linear PDEs that underpin Section 2.","marker":"[TT07]"},{"why":"proves optimal space-time regularity for the deterministic porous medium equation by combining the kinetic formulation with real interpolation, the combination extended here to modelledness seminorms.","marker":"[GST20]"},{"why":"supplies the regularity-structures formalism (models, modelled distributions, reconstruction) that the paper adapts to the present L¹ setting.","marker":"[Hai14]"},{"why":"introduced the kinetic formulation for scalar conservation laws, the starting point of the whole approach.","marker":"[LPT94]"},{"why":"a PhD thesis providing the model estimates and counterterm verification for the two worked examples (space white noise on T² and white-in-time coloured-in-space noise).","marker":"[dLF24]"},{"why":"the comparison point for the kernel-convolution strategy: convolution with Ψ(a(u(·)),·) versus Ψ(a(u(x)),·) is contrasted, and the singular products that the present work must renormalize are identified.","marker":"[GH19]"}],"fun_headline_variants":["Optimal a priori bounds for singular porous media","Degenerate porous-media SPDEs: optimal regularity","Rough noise tamed: optimal bounds for porous media","First a priori bounds for singular porous-media SPDEs","Renormalized energy gives optimal bounds for singular SPDEs"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the renormalization corrections, which carry a time dependence from starting the auxiliary heat flow at time zero, differ from their steady counterparts by an integrable-in-time amount (Assumption 1.2(ii), equations (1.10)–(1.11)) — if those differences diverge with the wrong rate or sign, the cancellation that keeps the singular products finite, and with it the entire energy estimate, collapses.","fun_headline_variants_meta":{"raw":{"variants":["Optimal a priori bounds for singular porous media","Degenerate porous-media SPDEs: optimal regularity","Rough noise tamed: optimal bounds for porous media","First a priori bounds for singular porous-media SPDEs","Renormalized energy gives optimal bounds for singular SPDEs"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001252,"raw_usage":{"total_tokens":5151,"prompt_tokens":983,"completion_tokens":4168,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":599,"completion_tokens_details":{"reasoning_tokens":4089}},"tokens_in":599,"tokens_out":4168,"duration_ms":31351,"temperature":1.0,"reasoning_tokens":4089,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T20:07:01.501977+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the integrals in (1.10)–(1.11) for space white noise on T² using the explicit counterterms of Example 1.11 with σ(v)=|v|^M (violating N≥M+1) at the boundary case of (1.12): the paper's balancing inequality (II) fails at that point, so the integrals' divergence would confirm the restriction on σ is load-bearing, while their finiteness would show the admissible class is wider than stated.","supporting_citations":[],"review_version":1}