{"id":"ead84be0-7339-4ced-8e61-ddb39cd2fd87","arxiv_id":"2411.17524","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Every stationary measure of the one-dimensional Porous Medium Model is a mixture of frozen configurations and Bernoulli product measures.","lead":"Scientists found all long-term probability distributions for a simple one-dimensional particle model where a particle can jump only if another particle is within two sites: every distribution is either a stuck configuration or a random independent mix of particles. The result is a rigorous step toward understanding when constrained particle systems reach equilibrium, which is relevant for hydrodynamic limits and for models of glasses.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The proof of Theorem 1 rests on Lemma 1(2), whose only justification is a figure and an informal 'mobile cluster carries extra particle' claim; a rigorous proof or exhaustive check is needed before the classification is fully established.","rationale":"The reader's weakest_assumption identifies the same lemma, so my agreement is partial: the reader scores soundness 7 and accepts, while I think the missing case analysis warrants a conditional verdict. I found no independent error in the entropy argument or the de Finetti step: the sign issue with Φ appears to be a repairable typo, and the exchangeability proof works for arbitrary finite intervals by summing over complements in a large centered interval, so interval exchangeability implies finite-set exchangeability. The boundedness of rates under Assumption 1 is worth a footnote, but the central gap is the connectivity lemma. A conditional acceptance with a request for a rigorous proof or exhaustive verification of Lemma 1(2) is appropriate.","tokens_in":10014,"tokens_out":38501,"duration_ms":403117,"concrete_test":"Implement the allowed-jump graph for intervals Λ_L={1,...,L}, L up to 10 or 12, with the PMM rate c_x(σ)>0 iff σ(x−1)+σ(x+2)>0 and empty boundary outside. For each particle number k, compute the connected components of the subgraph induced by G_{Λ_L}; check that all configurations with |σ|=k lie in one component. If any L,k gives multiple components, Lemma 1(2) is false. If all small cases pass, the residual risk is the general induction, so also attempt a formal inductive proof or a proof-assistant formalization of the 'transport an extra particle' procedure shown in Figure 1.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Lemma 1(2) states that any two configurations in a finite interval with the same particle number are connected by allowed jumps, provided the interval contains a mobile cluster. The proof is not a full proof: it points to Figure 1 and asserts that a mobile cluster can transport an extra particle anywhere, then repeats the procedure to mass particles at the right. No induction is written for arbitrary interval lengths, arbitrary numbers and locations of additional particles, or arbitrary backgrounds. This lemma is used at every subsequent step: Lemma 2's propagation of zero probability, the exchangeability argument in Section 4 comparing σ·ζ and σ′·ζ for ζ∈G_{B_n}, the contradiction for F′/F′′ via (12), and the reflection argument showing α_E′=0. If Lemma 1(2) fails for some configuration, the equality ν(σ)=ν(σ′) for connected cylinder events is unjustified and Theorem 1 does not follow from the written argument. The gap is plausible to fill, but it is genuinely load-bearing: it is the single combinatorial fact on which all later equalities of cylinder probabilities depend.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","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).","tokens_in":10164,"tokens_out":11343,"duration_ms":103772,"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":[{"comment":"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.","section":"Section 3, Lemma 1(2)"},{"comment":"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.","section":"Section 4, Eq. (11)"},{"comment":"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.","section":"Section 5, Eq. (39)"}],"minor_comments":[{"comment":"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).","section":"Section 5, definition of Phi"},{"comment":"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.","section":"Section 4, after Eq. (14)"},{"comment":"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.","section":"Throughout"},{"comment":"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.","section":"Section 2.1, Definition 1"}],"recommendation":"major_revision","confidential_remarks":"The paper is well within the scope of the journal and the main result is likely correct, but the proof as written is incomplete at a load-bearing point (Lemma 1(2)). The other issues are more local. I would be willing to see a revised version with a rigorous proof of the connectivity lemma and the Section 4/5 clarifications."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the classification of stationary measures for the 1D porous medium model is new and the main theorem is likely true. The proof is a real adaptation of the Holley–Stroock entropy argument to a model with frozen configurations, and I found no circularity or hidden fitting. The main weakness is exactly where the stress-test note points: Lemma 1(2), the finite-interval connectivity statement, is load-bearing and its proof is a picture plus an informal 'mobile cluster carries an extra particle' claim. I think the lemma is true, but as written it is not a complete proof.\n\nWhat is good: Theorem 1 gives a clean decomposition into a frozen part and a mixture of Bernoulli product measures; that settles a natural question for a non-ergodic kinetically constrained model. The invariant-set decomposition into F, F′, F′′, E′, E is useful. Lemma 3 is substantial: the entropy bounds with boundary terms are the right adaptation and appear to work. Section 4's exchangeability step is elegant once Lemma 3 is granted. The citations are appropriate; [BES21] is used as an application, not as a black-box input.\n\nWhere the soft spots are: Lemma 1(2) is used in Lemma 2, in the exchangeability step for σ·ζ, in the contradiction for F′/F′′ near (12), and in the reflection argument for E′. If it fails for some background configuration, the equality of cylinder probabilities is unjustified. The proof should be replaced with a formal induction: show any extra particle can be moved past a mobile cluster, then repeat. That is a fixable gap, but it is real. The entropy section also has dense notation and a few typos, though those are minor.\n\nBottom line: this deserves a serious referee, and the paper should be accepted if the connectivity lemma is rigorously established. If it is not, the paper has the right idea but the written proof does not close. I would cite it once the gap is fixed, and I would bring it to a reading group now for the discussion of how to complete Lemma 1(2).","headline":"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.","tokens_in":10711,"tokens_out":2293,"would_cite":true,"duration_ms":22530,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60K35","82C22"],"pacs":[],"model":"deepseek-v4-flash","headline":"Every stationary measure of the one-dimensional Porous Medium Model is a mixture of frozen configurations and Bernoulli product measures.","keywords":["stationary measures","Porous Medium Model","kinetically constrained models","exclusion process","product measures","frozen configurations","entropy production","invariant sets"],"falsifier":"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.","tokens_in":9725,"feed_emoji":"🧊","tokens_out":5699,"duration_ms":55601,"temperature":0.7,"pith_summary":"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.","feed_headline":"All stationary states are frozen or product","feed_subtitle":"Classification of equilibria for a degenerate-rate exclusion process, with implications for hydrodynamic limits.","key_machinery":"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.","core_discovery":"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 $E'$.","pith_inferences":["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."],"forward_implications":["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 ($\\partial_t\\rho = \\Delta(\\rho^2)$)."],"supporting_citations":[{"why":"Supplies the entropy-production argument that a stationary measure is reversible with respect to allowed exchanges.","marker":"[HS77]"},{"why":"Provides the reformulated entropy argument and the technical lemma used to control boundary terms.","marker":"[Lig05]"},{"why":"Offers another reformulation of the entropy argument adapted in the proof.","marker":"[NS93]"},{"why":"Introduced the Porous Medium Model and its degenerate rates.","marker":"[BT04]"},{"why":"Establishes the hydrodynamic limit to the porous medium equation that motivates the stationary-measure question.","marker":"[GLT09]"}],"fun_headline_variants":["Stationary states: frozen or product","Porous Medium Model: only frozen and product equilibria","Stationary measures: frozen plus product mixtures","All equilibria are frozen or Bernoulli products","Porous Medium Model equilibria: frozen or product"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"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.","fun_headline_variants_meta":{"raw":{"variants":["Stationary states: frozen or product","Porous Medium Model: only frozen and product equilibria","Stationary measures: frozen plus product mixtures","All equilibria are frozen or Bernoulli products","Porous Medium Model equilibria: frozen or product"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000352,"raw_usage":{"total_tokens":1837,"prompt_tokens":784,"completion_tokens":1053,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":400,"completion_tokens_details":{"reasoning_tokens":983}},"tokens_in":400,"tokens_out":1053,"duration_ms":8105,"temperature":1.0,"reasoning_tokens":983,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T12:00:31.830142+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"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.","supporting_citations":[],"review_version":1}