REVIEW 2 major objections 6 minor 1 cited by
Zero-noise selection and Large Deviations in $L^\infty_t L^p_x$ for the stochastic transport equation beyond DiPerna-Lions
T0 review · 2 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash
Pith's one-line read Rough Kraichnan noise forces the stochastic transport equation to select the DiPerna–Lions renormalized solution as the noise intensity goes to zero, with an ε² large-deviation rate.
desk verdict A genuine advance on zero-noise selection and large deviations for transport noise; the main theorems survive the stress-test concern about ε-uniform Sobolev bounds. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing object is the Sobolev regularization produced by the Kraichnan covariance: with α∈(0,1/2), the Itô correction term ε²Δρ gives the bound ε²E∫||ρ^ε||²_{$H^{{1−α−δ}}$} ≲ ||ρ_0||²_{L²} (estimate (A.10)). This estimate is what upgrades weak compactness to strong convergence in E in Section 2.2, and it is what makes the stability Proposition 3.5 work under Cameron–Martin perturbations of the drift, which is the heart of the large-deviation argument. The second mechanism is a new four-step reduction for uniform LDPs on non-separable metric spaces: tightness in a larger Polish space, measurability of distance-to-compact events, improvement of almost-sure convergence in the auxiliary space to convergence in E along sub-subsequences, and convergence of expectations of truncated distances to compact sets.
What would settle it
Replace the Kraichnan noise by a spatially smooth divergence-free noise (for example a finite sum of smooth modes). The coercive estimate (A.10) is known to fail, and the paper's proof of Theorem 1.7 explicitly uses that estimate; checking whether strong L∞_t L² convergence and the LDP still hold in that case would settle whether the rough-noise assumption is necessary. A direct numerical check on a 2D torus with α=0.25 would also confirm whether ε²E||ρ^ε||²_{$H^{{1−α−δ}}$} stays uniformly bounded as ε↓0.
Extended reading notes
Core claim
The central claim is Theorem 1.2: for every sequence of initial data ρ^ε_0 → ρ_0 in the closed unit ball B of L²_x∩L^p_x, the unique solution of the stochastic transport equation (1.2) converges in probability, as E-valued random variables, to the unique renormalized solution of (1.1) given by the DiPerna–Lions theory. If div b = 0, the convergence holds in the stronger, non-separable space E = L∞_t(L²_x∩L^p_x)∩C_t ilde H^−_x. Under the same divergence-free assumption, Theorem 1.7 gives a uniform Large Deviations Principle on E with speed ε² and rate function I_{ρ_0}(ρ) = inf{½||g||²_{L²_t $H^{0}$} : g∈L²_t $H^{0}$, ρ = $ρ^{{ρ_0,g}}$}, where $ρ^{{ρ_0,g}}$ is the renormalized solution of the transport equation with drift b+g and $H^{0}$ is the Cameron–Martin space of the noise, the space of drifts the noise can produce with finite energy. The rate function has compact level sets on compacts of B, so the LDP is uniform in the initial datum.
Load-bearing premise
The whole argument rests on the rough, spatially homogeneous Kraichnan covariance producing a coercive Sobolev regularization of size ε²; for spatially smooth noise this ingredient fails, so the strong convergence in E and the large-deviation result would not follow.
Editorial extensions
If this is right
- For every p∈(1,∞) and every compact set of initial data, the zero-noise limit of (1.2) is unique and equals the DiPerna–Lions renormalized solution; no other weak solution of the non-unique deterministic equation is selected by Kraichnan noise.
- Deviations of size δ from the selected limit have probability at most about exp(−ε^{-2}(m−o(1))) for all m up to any finite bound, and at least about exp(−ε^{-2}(I(ρ)+o(1))) near any trajectory ρ with finite rate.
- The dissipation measure of the L² norm, which records where and how much the noisy solutions lose L² mass, satisfies a pointwise LDP with rate function ∞ on every nonzero measure: nonzero dissipation is exponentially rare.
- The four-point method for non-separable spaces applies to any family of stochastic processes with tight laws in a Polish superspace, so the same scheme can prove sharp LDPs in the natural trajectory space of other SPDEs.
Reading between the lines
- If the coercive bound (A.10) is sharp, crossing the threshold α=1/2 or q=d/(2(1−α)) should destroy selection or change the speed: a spatially smoother or less integrable noise should allow non-renormalized limits to survive.
- The variational form of the rate function suggests a control interpretation of the selected solution as the small-noise limit of an optimal-control problem, which could yield quantitative estimates on the probability of selecting any particular weak solution.
- A natural testable extension is the joint limit ε,ν→0 with both noise and viscosity present; comparing the Kraichnan-selected limit with the vanishing-viscosity DiPerna–Lions limit would show whether the two regularizations commute.
- The same method should transfer to other transport-type SPDEs with coercive rough noise, such as generalized SQG or Boussinesq equations, giving uniform LDPs in their natural non-separable trajectory spaces.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper studies the stochastic transport equation (1.2) with drift b in L∞_t W^{1,q}_x and Kraichnan-type rough transport noise of intensity ε, in parameter regimes where the deterministic transport equation admits non-unique weak solutions. The main results are: (i) strong existence and pathwise uniqueness for the stochastic equation (Proposition 1.1); (ii) convergence in probability, as ε↓0, of the solutions to the unique DiPerna-Lions renormalized solution of the deterministic equation, in a Polish space E and, under div b=0, in the non-separable space E = L∞_t(L²_x∩L^p_x) (Theorem 1.2); (iii) a uniform Laplace/Large Deviations Principle on E with speed ε² and rate function I_{ρ0} given by the cost of the control g (Theorem 1.4); (iv) a uniform LDP on the non-separable space E under div b=0 (Theorem 1.7), obtained through a new extension of the weak-convergence approach; and (v) a pointwise LDP for the dissipation measure (Corollary 1.6). The proofs combine Fourier-based energy estimates adapted from [GGM24], DiPerna-Lions stability, a localization argument for strong convergence, and new measurability lemmas for E-valued random variables.
Significance. If the results are correct, they provide the first identification of a zero-noise selection principle for transport equations in a regime where deterministic weak solutions are non-unique, together with the corresponding large-deviation speed. The proof of the LDP on a non-separable space via the measurability and stability lemmas in Section 3 is an original methodological contribution that is likely to be reusable. The paper is careful about non-separability issues (Borel σ-field, Doob measurability, non-measurable events) and supplies detailed proofs, including a self-contained adaptation of the dissipation-measure construction from [DGP]. The main caveat is the uniformity in ε of the coercive Sobolev estimate (A.10), which is not established as written; my reading suggests the gap is local and fixable, but it affects the proof statements in Sections 2 and 3.
major comments (2)
- [Appendix A.2, Eq. (A.10); Section 2.1, Eq. (2.7)] The estimate ε² E‖ρ‖²_{L²_t H^{1-α-δ}_x} ≲ ‖ρ0‖²_{L²_x} is stated with an implicit constant that is used as uniform in ε in (2.7). The derivation does not establish this uniformity: the drift contribution C∫E‖ρ‖²_{H^{1-α-δ-λ}_x} ds is absorbed into the coercive term −ε²K∫E‖ρ‖²_{H^{1-α-δ}_x} ds by interpolation and Young's inequality, and the remaining constant is of order (ε²K)^{-(1-θ)/θ} with θ=(1-α-δ-λ)/(1-α-δ) when interpolating with L², which blows up as ε↓0; even the intermediate display in Appendix A.2 contains a constant C_{...,ε,R,T} and the Gronwall factor would be e^{C_ε T}. Consequently (2.7) as written is not justified. Since (2.7) is cited in Lemma 2.1 and 'bounds analogous to (2.7)' are invoked in Proposition 3.5, the manuscript must either prove the ε-uniform bound or explicitly remove the H^{1-α-δ} term from the uniform statements and verify that tightness and the strong-convergence arguments use only the C^γ_t H^{-σ} and L∞ bounds. My reading of Section 2.2 and Appendix B is that the fixed-ε version of (A.10) suffices for Lemma B.1 and that the uniform H^{1-α-δ} bound is not actually needed for the strong-convergence argument, so the gap appears fixable; but as written it affects the proof of the main theorems.
- [Appendix A.1, final display] In the pathwise-uniqueness proof, after absorption and Gronwall the displayed estimate reads E‖ρ_t‖²_{H^{-δ}_x} + K∫E‖ρ_s‖²_{H^{1-α-δ}_x} ds ≤ e^{CT}‖ρ0‖²_{H^{-δ}_x}; however, the coercive term carries the factor ε² and the Gronwall constant depends on ε through C_{...,ε,R}. This is not merely cosmetic: it is the step that would be needed to read (A.10) as ε-uniform, and the present display hides the ε-dependence. Please correct the formula and state explicitly which constants are uniform in ε and in κ.
minor comments (6)
- [Title and Abstract] The displayed title and abstract contain garbled spacing in 'L ∞ t Lp x FOR THE'; please fix the typography.
- [Introduction, Section 1] The sentence 'The proof of Proposition 1.1, given in the Appendix, is based on a technical adaptation...' is duplicated verbatim; one occurrence should be removed.
- [Corollary 1.6 proof] In the contradiction argument, 'ε²_n → ∞' should read 'ε_n^{-2} → ∞' (or equivalently 'ε_n² → 0'); as written the inequality is reversed.
- [Section 4.2, upper bound] In the approximate-minimizer display, the inequality '≤ E[...] - δ' should be '≤ E[...] + δ' for a near-minimizer; the subsequent liminf argument uses the correct sign, so this appears to be a typo.
- [Lemma A.4] The parameter θ is introduced twice in inconsistent ways ('0<λ<θ≪1' and later 'θ=α+δ+λ−1'); please define it once at the outset and use the same notation throughout.
- [Section 2.2 and Corollary 1.6] There are small typos: 'embdedding' should be 'embedding' in Lemma 2.2, and 'pointwise LPD' should be 'pointwise LDP' in the proof of Corollary 1.6.
Circularity Check
No significant circularity: zero-noise limit and LDP rate function are derived, not assumed.
full rationale
The derivation chain is self-contained, and the central claims are not equivalent to their inputs by construction. The coercive regularization (A.10), epsilon^2 E||rho||^2_{L^2_t H^{1-alpha-delta}} less than or similar to ||rho_0||^2_{L^2}, is the engine of the paper, and it is proven in Appendix A (Lemmas A.2-A.5 plus the Gronwall absorption argument) as a technical adaptation of the external reference [GGM24] (Galeati, Grotto, and Maurelli), whose authors do not overlap with the present paper; the estimate is not imported by citation alone. The zero-noise selection (Theorem 1.2) is a genuine limit: tightness of {rho^epsilon} in E follows from (2.7)-(2.9), the stochastic-integral term is shown to vanish in the identification step via the bound epsilon^2 E[...] -> 0, and the limit is identified with the unique renormalized solution using the external uniqueness theorem [DL89, Theorem II.3]. The strong-convergence upgrade on E uses the constancy of the L^2 norm of renormalized solutions when div b = 0 (an external DiPerna-Lions fact) together with the dissipation measure whose existence is proven in Lemma B.1; the citation [DGP] (Drivas, Galeati, and Pappalettera, sharing one author) is accompanied by an in-paper adaptation of the proof, so it is not load-bearing. The rate function I_{rho0}(rho) = inf{1/2 ||g||^2_{L^2_t H^0}: rho = rho^{rho0,g}} is the standard weak-convergence reduced cost, and the paper proves both sides of the LDP: the upper bound via the variational representation plus the stability Proposition 3.5, the lower bound by the explicit control g_* and the E-distance stability in Proposition 3.5. Nothing is calibrated to force the rate; the LDP lower bound requires showing that the stochastic solution with drift g_* has law close to the deterministic control solution, which is a deduced stability result rather than an input. The uniformity-in-epsilon concern about (A.10) is a mathematical correctness question about an analytic estimate, not a case of a prediction reducing to its own inputs, and it does not generate a circularity step.
Assumptions & free parameters
assumptions (7)
- standard math DiPerna-Lions theory of renormalized solutions and their stability [DL89, Theorems II.3 and II.7].
- standard math Budhiraja-Dupuis-Maroulas weak convergence approach and variational representation for infinite-dimensional Brownian noise [BDM08, BD19].
- standard math Girsanov theorem, Itô formula, and Burkholder-Davis-Gundy inequality for cylindrical Wiener processes.
- domain assumption Kraichnan noise covariance Q(z) with Fourier multiplier (1+|ξ|²)^{-(d/2+α)}, α∈(0,1/2), normalized by Q(0)=2I_d.
- domain assumption Drift regularity b∈L^∞_t W^{1,q}_x with d/(2(1-α)) < q ≤ 2 and div b∈L¹_t L^∞_x ∩ L^∞_t H^ϑ_x for some ϑ>0.
- domain assumption Initial data ρ0 ∈ L^2_x ∩ L^p_x with p∈(1,∞).
- domain assumption Existence of the dissipation measure for b=0 from [DGP] (cited as to appear).
Cite this review
Pith. "Pith review of Zero-noise selection and Large Deviations in $L^\infty_t L^p_x$ for the stochastic transport equation beyond DiPerna-Lions." pith.science (2026). https://pith.science/paper/XWQXWPMI
@misc{pith2026250606947,
author = {Pith},
title = {Pith review of: Zero-noise selection and Large Deviations in $L^\infty_t L^p_x$ for the stochastic transport equation beyond DiPerna-Lions},
year = {2026},
howpublished = {\url{https://pith.science/paper/XWQXWPMI}},
note = {Machine review of arXiv:2506.06947}
}
abstract
We consider $L^\infty_t L^p_x$ solutions of the stochastic transport equation with drift in $L^\infty_t W^{1,q}_x$. We show strong existence and pathwise uniqueness of solutions in a regime of parameters $p,q$ for which non-unique weak solutions of the deterministic transport equation exist. When the intensity of the noise goes to zero, we prove that the solutions of the stochastic transport equation converge to the unique renormalized solution of the transport equation in the sense of DiPerna-Lions. Furthermore, we show that the convergence is governed by a Large Deviations Principle in the space $L^\infty_t L^p_x$. Since the space $L^\infty_t L^p_x$ is not separable, the weak convergence approach to Large Deviations by Budhiraja, Dupuis, and Maroulas is not directly applicable.
Forward citations
Cited by 1 Pith paper
-
Failure of the Weak Sard property without Anomalous Dissipation
There exist autonomous C^α divergence-free planar fields that fail the weak Sard property but induce no anomalous dissipation for advection-diffusion.
Reference graph
Works this paper leans on
-
[9]
[CM24] Michele Coghi and Mario Maurelli. Existence and uniqueness by Kraichnan noise for 2D Euler equations with unbounded vorticity.arXiv 2308.03216,
-
[10]
[CP19] Sandra Cerrai and Nicholas Paskal. Large deviations for fast transport stochastic RDEs with applica- tions to the exit problem.The Annals of Applied Probability, 29(4):1993–2032,
work page 1993
-
[12]
[DPFR V16] Giuseppe Da Prato, Franco Flandoli, Michael R¨ ockner, and A. Yu. Veretennikov. Strong uniqueness for SDEs in Hilbert spaces with nonregular drift.The Annals of Probability, 44(3):1985–2023,
work page 1985
-
[20]
On the zero-noise limit for SDE’s singular at the initial time.arXiv preprint arXiv:2503.22905,
[Pit25] Jules Pitcho. On the zero-noise limit for SDE’s singular at the initial time.arXiv preprint arXiv:2503.22905,
-
[1981]
[Vis18a] Misha Vishik. Instability and non-uniqueness in the Cauchy problem for the Euler equations of an ideal incompressible fluid. Part I.arXiv preprint arXiv:1805.09426,
-
[1989]
Zero noise limit for multidimensional SDEs driven by a pointy gradient
[DM19] Fran¸ cois Delarue and Mario Maurelli. Zero noise limit for multidimensional SDEs driven by a pointy gradient.arXiv preprint arXiv:1909.08702,
work page Pith review arXiv 1909
-
[1995]
[Fla11] Franco Flandoli.Random perturbation of PDEs and fluid dynamic models, volume 2015 ofLecture Notes in Mathematics. Springer, Heidelberg,
work page 2015
-
[1996]
Weak well-posedness by transport noise for a class of 2D fluid dynamics equations.arXiv 2305.08761,
[GL23] Lucio Galeati and Dejun Luo. Weak well-posedness by transport noise for a class of 2D fluid dynamics equations.arXiv 2305.08761,
Show all 23 references
-
[1998]
Strong uniqueness by Kraichnan transport noise for the 2D Boussinesq equations with zero viscosity.arXiv:2504.19153,
ZERO-NOISE AND LDP FOR STOCHASTIC TRANSPORT 37 [JL25a] Shuaijie Jiao and Dejun Luo. Strong uniqueness by Kraichnan transport noise for the 2D Boussinesq equations with zero viscosity.arXiv:2504.19153,
-
[2002]
Anomalous regularization in Kraichnan’s passive scalar model.arXiv 2407.16668,
[GGM24] Lucio Galeati, Francesco Grotto, and Mario Maurelli. Anomalous regularization in Kraichnan’s passive scalar model.arXiv 2407.16668,
-
[2004]
Small random perturbations of Peano phenomena.Stochastics, 6(3- 4):279–292, 1981/82
[BB82] Roberto Bafico and Paolo Baldi. Small random perturbations of Peano phenomena.Stochastics, 6(3- 4):279–292, 1981/82. [BCC22] Paolo Bonicatto, Gennaro Ciampa, and Gianluca Crippa. On the advection-diffusion equation with rough coefficients: weak solutions and vanishing v...
1981
-
[2007]
On some recent results concerning non-uniqueness for the transport equation
[Mod20] Stefano Modena. On some recent results concerning non-uniqueness for the transport equation. In Hyperbolic problems: theory, numerics, applications, volume 10 ofAIMS Ser. Appl. Math., pages 562–568. Am. Inst. Math. Sci. (AIMS), Springfield, MO, [2020]©2020. [MP12] Carl...
2020
-
[2009]
Global smooth solutions by transport noise of 3D Navier-Stokes equations with small hyperviscosity.arXiv preprint arXiv:2406.09267,
[Agr24] Antonio Agresti. Global smooth solutions by transport noise of 3D Navier-Stokes equations with small hyperviscosity.arXiv preprint arXiv:2406.09267,
-
[2010]
[Ver81] Alexander Ju Veretennikov
Reprint of 1983 edition [MR0730762], Also published in 1983 by Birkh¨ auser Verlag [MR0781540]. [Ver81] Alexander Ju Veretennikov. On strong solutions and explicit formulas forsolutions of stochastic integral equations.Mathematics of the USSR-Sbornik, 39(3):387,
1983
-
[2011]
[Saint-Flour Probability Summer School]
Lectures from the 40th Probability Summer School held in Saint-Flour, 2010, ´Ecole d’ ´Et´ e de Probabilit´ es de Saint-Flour. [Saint-Flour Probability Summer School]. [FW12] Mark I. Freidlin and Alexander D. Wentzell.Random perturbations of dynamical systems, volume 260 of Gr...
2010
-
[2012]
[Gaw02] Krzysztof Gawedzki
Translated from the 1979 Russian original by Joseph Sz¨ ucs. [Gaw02] Krzysztof Gawedzki. Soluble models of turbulent advection.arXiv preprint nlin/0207058,
1979 arXiv
-
[2013]
A convex integration scheme for the continuity equation past the Sobolev embedding threshold.arXiv:2504.03578,
[CCK25] Maria Colombo, Roberto Colombo, and Anuj Kumar. A convex integration scheme for the continuity equation past the Sobolev embedding threshold.arXiv:2504.03578,
-
[2014]
Stochastic geometric wave equations with values in compact Riemannian homogeneous spaces.The Annals of Probability, 41(3B):1938–1977,
[BO13] Zdzis law Brze´ zniak and Martin Ondrej´ at. Stochastic geometric wave equations with values in compact Riemannian homogeneous spaces.The Annals of Probability, 41(3B):1938–1977,
1938
-
[2017]
Regularization by rough Kraichnan noise for the generalised SQG equations.arXiv 2405.12181,
[BGM24a] Marco Bagnara, Lucio Galeati, and Mario Maurelli. Regularization by rough Kraichnan noise for the generalised SQG equations.arXiv 2405.12181,
-
[2018]
Instability and non-uniqueness in the Cauchy problem for the Euler equations of an ideal incompressible fluid
[Vis18b] Misha Vishik. Instability and non-uniqueness in the Cauchy problem for the Euler equations of an ideal incompressible fluid. Part II.arXiv preprint arXiv:1805.09440,
-
[2019]
[AF09] Stefano Attanasio and Franco Flandoli
Publisher Copyright:© 2019, Springer Nature Switzerland AG. [AF09] Stefano Attanasio and Franco Flandoli. Zero-noise solutions of linear transport equations without uniqueness: an example.Comptes Rendus Mathematique, 347(13):753–756,
2019
-
[2021]
Sharp Nonuniqueness in the Transport Equation with Sobolev Velocity Field.arXiv preprint arXiv:2405.01670,
[BCK24] Elia Bru´ e, Maria Colombo, and Anuj Kumar. Sharp Nonuniqueness in the Transport Equation with Sobolev Velocity Field.arXiv preprint arXiv:2405.01670,
-
[2024]
Anomalous Regularization in Kazantsev- Kraichnan Model.arXiv:2411.09482,
[BGM24b] Marco Bagnara, Francesco Grotto, and Mario Maurelli. Anomalous Regularization in Kazantsev- Kraichnan Model.arXiv:2411.09482,
Reviewed August 7, 2026 · model on record in the stance chip above.
Discussion (0). Sign in to comment.