Pith. sign in

REVIEW 3 major objections 4 minor 7 references

Stationary measures for the Porous Medium Model

T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Every stationary measure of the one-dimensional Porous Medium Model is a mixture of frozen configurations and Bernoulli product measures.

desk verdict A genuinely new classification of stationary measures for the porous medium model, with a real proof gap in the load-bearing connectivity lemma that a referee should require to be filled. read the letter →

arxiv 2411.17524 v1 pith:4E6VA4F6 submitted 2024-11-26 math.PR

Oriane Blondel This is my paper · ORCID
classification math.PR MSC 60K3582C22
keywords stationarymeasuresPorousMediumModelkineticallyconstrainedmodelsexclusionprocessproductfrozenconfigurationsentropyproductioninvariantsets
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper classifies the stationary probability measures of the one-dimensional Porous Medium Model, a kinetically constrained exclusion process in which two neighboring sites can exchange occupancy only when a particle sits one site away from the pair. The main theorem states that every stationary measure decomposes into a frozen part, supported on configurations whose particles are all isolated at distance at least three, plus a mixture of Bernoulli product measures over particle density. The result matters because degenerate-rate particle systems have infinitely many formal invariant sets, and knowing which measures actually survive stationarity is a prerequisite for hydrodynamic-limit and equilibrium questions. The proof works by showing that stationarity forces invariance under all allowed exchanges and then uses exchangeability to identify the surviving measures.

What carries the argument

The central combinatorial object is the mobile cluster: a pair of particles at distance one or two from each other, which can move through arbitrary backgrounds and can also transport one additional particle. Lemma 1(2) asserts that inside any finite interval containing a mobile cluster, two configurations with the same number of particles are connected by allowed jumps; this connectivity statement is what converts equality of connected probabilities into equality of cylinder probabilities. The analytic machinery is the entropy-production identity built on $\Phi(u,v) = \log(u/v)(v-u)$, whose convexity, homogeneity, and subadditivity force boundary terms to vanish and imply that the stationary measure is reversible with respect to allowed jumps.

What would settle it

An exhaustive computer search over finite intervals with empty boundary conditions, looking for two configurations with equal particle number and at least one mobile cluster but no allowed-jump path between them, would settle Lemma 1(2) and with it the classification.

Watch

Extended reading notes

Core claim

Under Assumption 1, Theorem 1 proves that any stationary probability measure $\nu$ for the generator (2) has the form $\nu = \alpha_F \nu_F + \alpha_E \nu_E$, where $\nu_F$ is supported on frozen configurations and, when $\alpha_E > 0$, $\nu_E = \int \mu_\rho \, d\lambda(\rho)$ for some probability measure $\lambda$ on $(0,1)$. In other words, the extremal stationary measures are exactly the point masses on frozen configurations and the translation-invariant product Bernoulli measures. An intermediate lemma, Lemma 3, is the key step: if two finite-window configurations are connected by allowed jumps, a stationary measure gives them the same probability; this turns stationarity into local exchangeability and rules out stationary measures concentrated on the exceptional invariant sets $F'$, $F''$, and $E'$.

Load-bearing premise

The whole classification rests on the unproved-by-case-analysis claim that a pair of nearby particles can be maneuvered anywhere inside a finite interval and can carry an extra particle, so that any two configurations with the same particle count are connected.

Editorial extensions

If this is right

  • The only extremal stationary measures are point masses on frozen configurations and Bernoulli product measures, so every stationary state is a statistical mixture of these.
  • No stationary probability measure can assign positive mass to the invariant sets with finitely many particles, finitely many holes, or finitely many active particles; such finite abnormalities escape to infinity.
  • Under any stationary measure, two finite-window configurations linked by a chain of allowed jumps have equal probability, so stationary weights are locally exchangeable.
  • The classification applies to all constraint families satisfying Assumption 1, not only the exact porous-medium rate $\eta(x-1)+\eta(x+2)$.
  • The result constrains the hydrodynamic behavior: the only invariant bulk measures are density mixtures, consistent with the porous medium equation ($\partial_t\rho = \Delta(\rho^2)$).

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural stress-test is the mobility lemma itself: the paper argues Lemma 1(2) from a figure and the phrase “as one can easily check,” so an exhaustive computer search over finite intervals could either confirm or break the classification.
  • The entropy-plus-exchangeability route would likely classify stationary measures for other one-dimensional kinetically constrained models, provided an analogous mobile-cluster connectivity lemma holds for their constraints.
  • Because the proof identifies stationary measures with mixtures of product measures, it suggests that in the hydrodynamic scaling the only accessible invariant densities are constant densities, reinforcing the connection to the porous medium equation.
  • The decomposition theorem leaves open the question of which mixtures $\lambda$ arise from natural initial states, so a next step would be to characterize the basin of attraction of each stationary measure.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper studies stationary probability measures for the one-dimensional Porous Medium Model and its variants under Assumption 1, a class of kinetically constrained exclusion processes with exchanges allowed at an edge only if a particle sits at a neighboring site. The main theorem, Theorem 1, states that every stationary measure decomposes as a frozen part concentrated on configurations without active particles and a part that is a mixture of Bernoulli product measures, so that the only extremal stationary measures are frozen point masses and product measures. The proof combines two ingredients: a Holley-Stroock-type entropy argument (Section 5) showing that stationarity forces invariance under all allowed exchanges on cylinder sets (Lemma 3), and combinatorial connectivity properties of configurations containing a mobile cluster (Lemma 1), which are then used to prove exchangeability and to rule out stationary measures concentrated on the invariant sets F', F'', and E' (Section 4).

Significance. If the proof is completed, this is a substantial contribution to the theory of kinetically constrained exclusion processes: it gives a complete classification of stationary measures for a degenerate-rate model, identifies the exotic invariant sets of zero product-measure probability, and provides a template for applying entropy arguments beyond reversible settings. The paper is clearly written and the overall strategy is convincing. The main proof ingredients — the entropy lemma and the connectivity lemma — are naturally separated, and the paper does not rely on fitted parameters or unstated external results. However, the written proof has a load-bearing gap in the combinatorial connectivity lemma, and a few technical definitions in the exchangeability step need to be made precise.

major comments (3)
  1. [Section 3, Lemma 1(2)] The proof of Lemma 1(2) is not a complete proof. It refers to Figure 1 and asserts that a mobile cluster can carry an additional particle anywhere, but no induction or case analysis is supplied for arbitrary interval lengths, arbitrary numbers and positions of additional particles, or arbitrary backgrounds. The lemma is used at every subsequent step: in Lemma 1(3), in the propagation of zero probability in Lemma 2, in the exchangeability argument of Section 4 comparing sigma·zeta and sigma'·zeta, and in the reflection argument proving alpha_E' = 0. If Lemma 1(2) fails for some configuration, the equality of cylinder probabilities under a stationary measure is not justified and Theorem 1 does not follow from the written argument. The gap is likely fillable, but it is genuinely load-bearing and needs a rigorous proof.
  2. [Section 4, Eq. (11)] The exchangeability step for nu_E uses the decomposition (11) over zeta in G_{B_n} \ G_{B_{n-1}}, with B_n = Lambda_{n0+n} \ Lambda. Since Lambda = Lambda_{n0}, the set B_n is a union of two disjoint intervals, not an interval, whereas G_Lambda was defined only for intervals of Z. The sets G_{B_n} and their differences must be defined, and the partition property used in (11) needs justification. This is a technical gap in the proof that nu(sigma) = nu(sigma') for all sigma, sigma' with equal particle number.
  3. [Section 5, Eq. (39)] In the case nu(F'_k) > 0, the modified entropy \tilde H_n is claimed to be well-defined by Lemma 2, but Lemma 2(1) only guarantees positivity on F'_k itself, not on all configurations sigma with |sigma| <= k appearing in the sum. The proof needs an explicit convention for terms with nu(sigma) = 0 and a justification that the derivative identity (16) and the boundary estimates (27)-(34) remain valid when some cylinder probabilities vanish. The sentence 'the rest of the proof can be carried as above' is not sufficient for this case.
minor comments (4)
  1. [Section 5, definition of Phi] The definition Phi(u,v) = log(u/v)(v-u) has the wrong sign: with this definition Phi is non-positive, not non-negative as stated. The subsequent inequalities require Phi(u,v) = log(u/v)(u-v).
  2. [Section 4, after Eq. (14)] The text writes 'For N >= (3k+1)N', which should presumably be 'N >= (3k+1)n'. Also, the sets Lambda'_N,2 and the phrase 'with empty boundary condition' in the definitions of B_N and B'_N are not defined precisely.
  3. [Throughout] There are several typographical errors, including 'M edium' and 'Po rous' in the abstract, and 'subadditive 1' with a stray footnote marker. These should be corrected.
  4. [Section 2.1, Definition 1] The notation G_Lambda is defined only for intervals, but later, especially in Section 4, it is used for subsets that are unions of intervals. Either extend the definition or adjust the notation.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the stationary-measure classification is derived from the generator via external entropy arguments, with no fitted inputs or load-bearing self-citation.

full rationale

The paper's central result, Theorem 1, is proved from the generator's definition and Assumption 1, not from any fitted parameter or imported conclusion. The entropy arguments are attributed to [HS77] and [Lig05], which are external methods; Lemma 3 is an adaptation of those arguments, and the proof of the lemma is carried out in Section 5 with explicit bounds. The only self-citation, [BES21], is cited as an application of such stationary-measure knowledge, not as an input to the proof. The combinatorial Lemma 1(2) is a genuinely stated fact about allowed jumps and mobile clusters; it is not defined in terms of stationary measures or the theorem's conclusion. The proof of Lemma 1(2) is informal and may warrant a rigorous case analysis, but an unproven or under-justified lemma is a correctness or rigor concern, not circularity. No equation is shown to be equal to its own input by construction, no parameter is fitted and then called a prediction, and no uniqueness claim is imported from the authors' prior work. Therefore the derivation chain is self-contained against the external benchmarks used, and the circularity score is 0.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The paper introduces no free parameters or new physical entities. The proof uses the model-defining Assumption 1, de Finetti's theorem, and entropy inequalities from the literature. The only fragile, paper-specific ingredient is the picture-based finite-box connectivity lemma.

assumptions (4)
  • domain assumption Assumption 1: rates are translation invariant, local, c0(eta) = c0(eta^{0,1}), and positive exactly when eta(-1) + eta(2) > 0.
    Defines the model class and guarantees reversibility with respect to every Bernoulli product measure; the theorem is stated under exactly this assumption.
  • ad hoc to paper Finite-box connectivity: any two configurations in G_Lambda with equal particle number are connected by allowed jumps inside Lambda (Lemma 1(2)).
    Justified by Figure 1 and a short argument rather than a full case analysis; all propagation of cylinder-probability equalities relies on this lemma.
  • standard math de Finetti's theorem: exchangeable probability measures on {0,1}^Z are mixtures of Bernoulli product measures.
    Used in Section 4 to convert finite-box exchangeability into the mixture representation (10).
  • standard math Holley-Stroock relative entropy identity: for a stationary measure, the time derivative of finite-box relative entropy vanishes, producing equation (16).
    Borrowed from [HS77] and [Lig05]; it is the engine of Lemma 3.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Stationary measures for the Porous Medium Model." pith.science (2026). https://pith.science/paper/4E6VA4F6

@misc{pith2026241117524,
  author       = {Pith},
  title        = {Pith review of: Stationary measures for the Porous Medium Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4E6VA4F6}},
  note         = {Machine review of arXiv:2411.17524}
}
abstract

We study the stationary measures for variants of the Porous Medium Model in dimension 1. These are exclusion processes that belong to the class of kinetically constrained models, in which an exchange can occur between $x$ and $x+1$ only if there is a particle either at $x-1$ or $x+2$. We show that any stationary probability measure can be decomposed into a frozen part and a mixture of product measures (although there exist invariant sets which have zero probability under these measures).

Figures

Figures reproduced from arXiv: 2411.17524 by the authors.

Figure 1
Figure 1. On the first line, a strategy for moving a mobile clus [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. On the first line are depicted Λ,Λ ′ appearing in the proof that αE′ = 0, with n = 2, k = 1. The configuration on top is in BN , and the bottom part of the picture gives its image under the application Φ defined in (14). For n ≥ 1, write Bn = Λn0+n \ Λ. For σ ∈ ΩΛ, write ν(σ) = X n≥2 X ζ∈GBn \GBn−1 ν(σ · ζ), (11) where σ · ζ denotes the configuration equal to σ in Λ and to ζ in Bn. This equality holds because, thank… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

7 extracted references · 7 canonical work pages

  1. [1]

    Stefan problem for a non-ergodic facilitated exclusion process

    Oriane Blondel , Cl \'e ment Erignoux , and Marielle Simon . Stefan problem for a non-ergodic facilitated exclusion process . Probability and Mathematical Physics , 2:127--178, 2021

    work page 2021

  2. [2]

    Exclusion processes with degenerate rates: convergence to equilibrium and tagged particle

    Lorenzo Bertini and Cristina Toninelli. Exclusion processes with degenerate rates: convergence to equilibrium and tagged particle. J. Statist. Phys. , 117(3-4):549--580, 2004

    work page 2004

  3. [3]

    T. Funaki. Free boundary problem from stochastic lattice gas model. Ann. Inst. H. Poincar\' e Probab. Statist. , 35(5):573--603, 1999

    work page 1999

  4. [4]

    Gon c alves, C

    P. Gon c alves, C. Landim, and C. Toninelli. Hydrodynamic limit for a particle system with degenerate rates. Ann. Inst. Henri Poincar\'e Probab. Stat. , 45(4):887--909, 2009

    work page 2009

  5. [5]

    R. A. Holley and D. W. Stroock. In one and two dimensions, every stationary measure for a stochastic I sing model is a G ibbs state. Comm. Math. Phys. , 55(1):37--45, 1977

    work page 1977

  6. [6]

    Thomas M. Liggett. Interacting particle systems. With a new postface. Class. Math. New York, NY: Springer, reprint of the 1985 edition edition, 2005

    work page 1985

  7. [7]

    The biased annihilating branching process

    Claudia Neuhauser and Aidan Sudbury. The biased annihilating branching process. Advances in Applied Probability , 25(1):24--38, 1993

    work page 1993

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.