{"id":"2acbaa1b-2efb-450d-a3cd-43097fc439df","arxiv_id":"2611.17200","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For subcritical particle systems with weak feedback, the absorbing barrier grows as K_a sqrt(t) and its t^{-1/4} fluctuations converge to an explicit Gaussian process identical to the continuous 1D-MDLA fluctuation limit.","lead":"A particle system with diffusing particles and a growing absorbing barrier is shown to converge, in a weak feedback regime, to the supercooled Stefan problem with an explicit square-root free boundary. The paper derives the Gaussian fluctuations around that limit and proves that a related aggregation model has identical fluctuations.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek/deepseek-v4-flash","headline":"The FCLT and SPDE rely on the classical solution regularity of the limiting free boundary Λ, which is imported from [17, Section 4] and [3, Theorem 1.1] via Proposition A.1; the paper does not verify that the truncated-data limit in Proposition A.1 yields the required uniform derivative bound…","rationale":"The reader's weakest_assumption correctly identifies the reliance on the classical solution regularity of Λ and the import from [17] and [3]. My stress-test deepens this: the specific concern is that the uniform derivative bound may not survive the limit from truncated to non-integrable data, because the constant in [17] likely depends on the L^1 norm of the initial data, which diverges for g≡a. The paper's Proposition A.1 asserts the existence of a classical solution on (0,∞) but does not provide a proof that the derivative bound is uniform in the truncation parameter n. Lemma B.1 then uses this bound as a black box. This is a load-bearing gap because the entire tightness argument for the fluctuation process depends on it. The paper acknowledges the reliance on [17] and [3], but does not address the potential n-dependence of the constant. Therefore the verdict should be CONDITIONAL: accept only if the derivative bound can be justified for the non-integrable case, either by a new argument or by showing that the constant in [17] depends only on ||g||_∞. The concrete test is to check this dependence explicitly.","tokens_in":40951,"tokens_out":938,"duration_ms":301930,"concrete_test":"Independently derive the derivative bound for the free boundary Λ corresponding to g≡a (the similarity solution K_a sqrt(t) is explicit, so the bound is known). Check whether the argument in Proposition A.1 can be made rigorous by showing that the constant in [17, Section 4] depends only on the L^∞ norm of g (which is a<1) and not on the L^1 norm, or by providing an alternative direct estimate for the non-integrable case. If the bound cannot be established, the tightness of √N(Λ^N - Λ) on [h,T] (Lemma B.1) and the subsequent FCLT are unsupported.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central fluctuation result (Theorem 4.1) and the SPDE (Theorem 6.1) depend on the limiting free boundary Λ being a classical solution, continuously differentiable on (0,T] with |dot Λ_t| ≤ C t^{-1/2}. This regularity enters Lemma B.1's control of the increments of c^N_t = √N(Λ^N_t - Λ_t) and the tightness argument. The paper's Proposition A.1 attempts to establish this by truncating the initial data g to g_n = g 1_{[1/n,n]}, citing [3, Theorem 1.1] for existence of a classical solution for each n, and then passing to the limit using [17, Section 4] to obtain a uniform derivative bound. However, the passage from the truncated data to the full non-integrable data g≡a is not fully justified: the bound |dot Λ^n_t| ≤ C t^{-1/2} from [17] is stated for the truncated data, but the constant C may depend on n through the L^1 norm of g_n, which grows like log n (since g≡a has infinite mass). The paper does not show that the constant can be chosen independently of n, nor does it address how the limit Λ inherits the derivative bound from the sequence Λ^n. The argument in Lemma B.1 then assumes that Λ is continuously differentiable on [h/3,T] with a uniform bound, which is precisely what is at stake. If this regularity fails, the density estimate near zero (used in Lemma 4.3, Corollary 4.6, and Lemma 5.4) would not hold, and the entire fluctuation theory would collapse.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek/deepseek-v4-flash","summary":"The paper studies a particle system on the positive real line with a moving barrier, connected to the supercooled Stefan problem. It establishes the long-time limit of the barrier (ξ_t/√t → K_a√s) and a functional central limit theorem for fluctuations, showing convergence to a Gaussian process solving a Volterra equation. The results are extended to 1D-MDLA and to an SPDE for the empirical measure fluctuations. The proofs use a Poisson random measure representation, martingale estimates, tightness, and contraction arguments.","tokens_in":41301,"tokens_out":2518,"duration_ms":22541,"significance":"If correct, the results provide a rigorous derivation of the scaling limit and fluctuations for a class of particle systems with non-integrable initial data, connecting to the supercooled Stefan problem. The FCLT and SPDE are significant contributions to the understanding of fluctuations in such systems. The paper also extends previous work to the weak feedback regime and non-integrable data.","major_comments":[{"comment":"The existence of a classical solution to the supercooled Stefan problem with non-integrable initial data g≡a is crucial for the entire fluctuation theory. The proof in Proposition A.1 truncates the initial data to g_n = g 1_{[1/n,n]}, cites [3, Theorem 1.1] for existence for each n, and then uses [17, Section 4] to obtain a uniform derivative bound |dot Λ^n_t| ≤ C t^{-1/2}. However, the constant C in [17] may depend on the L^1 norm of g_n, which grows like log n for g≡a. The paper does not demonstrate that the constant can be chosen independently of n, nor does it justify how the limit Λ inherits the derivative bound from the sequence Λ^n. This is a load-bearing issue because Lemma B.1 and the tightness argument (Lemma 4.7) rely on Λ being continuously differentiable on [h/3,T] with |dot Λ_t| ≤ C t^{-1/2}. Without this regularity, the density estimate near zero (used in Lemma 4.3, Corollary 4.6, Lemma 5.4) would not hold, and the central fluctuation theorem would collapse. The authors should either provide a direct proof of the required regularity for g≡a, or cite a source that establishes it explicitly for non-integrable data, or modify the argument to avoid reliance on this regularity.","section":"Appendix A, Proposition A.1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof of Lemma 4.8 uses the claim that sup_{s≤t}(W_s+Λ_s) is a Lebesgue point of g almost surely. The argument invokes Girsanov's theorem and the fact that Λ is increasing and differentiable on (0,t]. However, the differentiability of Λ is precisely what is at stake (see previous comment). Moreover, the argument that the supremum is attained at a time > δ with probability tending to 1 as δ→0 relies on continuity of Λ and properties of Brownian motion, but the differentiability of Λ is used to apply Girsanov. If Λ is not differentiable, the absolute continuity argument may fail. This needs to be clarified or the proof must be restructured to avoid reliance on differentiability of Λ.","section":"Section 4, Lemma 4.8","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof of Lemma 5.4 uses the similarity solution form for ˜V Λ(t,x) when g≡a, specifically the equality ˜V Λ(t,x) = a(1-Φ(K_a))[Φ(K_a + x/√t) - Φ(K_a)]. This relies on Λ_t = K_a√t being the exact similarity solution. However, the paper's main result for the long-time limit (Theorem 1.1) shows that ξ_t/√t → K_a√s, but the convergence is only in probability and the fluctuations are of order t^{-1/4}. The proof of Lemma 5.4 seems to assume that Λ_t = K_a√t exactly for all t, which is not true for the finite-N system; it is only the limit. The argument uses an inequality that bounds ˜V Λ^N(t-,x) by ˜V Λ(t, x+2N^{-1/2} sup_{s<t}|c^N_s|), but this inequality is derived from the similarity solution form for Λ (the limit). However, Λ^N is not equal to Λ; it is a random process that converges to Λ. The bound may still hold asymptotically, but the justification is insufficient. The authors should provide a more rigorous argument that does not assume the exact similarity form for the finite-N system, or clarify how the asymptotic bound is obtained.","section":"Section 5, Lemma 5.4","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof of Lemma 4.3 relies on the convergence of Λ^N to Λ uniformly in probability, and then uses the Skorokhod representation to consider versions that converge almost surely. The subsequent bound on the quadratic variation term involves sup_{s>δ} sup_{x≤2ϵ} ˜V^{Λ-ϵ}(s,x), which is claimed to decrease to zero as ϵ→0 because Λ is a classical solution. However, this again assumes regularity of Λ (specifically, that the density near zero is small). The dependence on the classical solution regularity should be explicitly stated and justified. If the regularity is not fully established (see first comment), this step is not justified.","section":"Section 4, Lemma 4.3 and Corollary 4.4","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof of Lemma 6.8 uses an upper bound on the jump size ∆Λ^N_t by comparison to the busy period of an M/D/∞ queue with arrival rate ||g||_∞ and service time one. This comparison is valid because ˜V^{Λ^N}(s-,x) ≤ ||g||_∞. However, the proof also uses the bound E((L^N_T)^2) ≤ C̃ E((Λ^N_T)^2) + C̃ N E(Λ^N_T). The derivation of this bound is not fully detailed; it relies on the independence of the Poisson processes and the Borel-Tanner distribution. The authors should provide a more explicit justification or a reference for the moment bounds of the Borel-Tanner distribution.","section":"Section 6, Lemma 6.8","authors_acknowledge":null,"authors_acknowledge_p":null}],"minor_comments":[{"comment":"The paper is generally well-written, but there are several typos and minor issues. For example, in the abstract, 'initiated from a Poisson point process' should be 'initialized from a Poisson point process'. In Section 1, 'W i0' should be 'W_i^0' for consistency. In equation (1.1), the definition of ρ_{t-} uses a sum over i, but the index i is not defined in that equation; it should be clear that it is the same i as in the particle positions.","section":"General","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The statement of Theorem 1.1 includes the tail bound P(sup_{s≤T} |ξ_{st}/√t - K_a√s| > x t^{-1/4}) ≤ 2 exp(-C(a,T)x^2). However, the condition 'for x < t^{1/4}' is unusual; typically such bounds hold for all x. The authors should clarify why the restriction x < t^{1/4} is needed, or whether it can be removed.","section":"Section 1.6, Theorem 1.1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The statement of Lemma 2.1 is somewhat unclear. The condition 'Γ^N(f_{.-} + x 1_{.≥t})_t ≥ f_t + x for all 0≤x≤Δf_t at any time t at which f jumps upwards' is difficult to parse. The authors should rephrase for clarity.","section":"Section 2, Lemma 2.1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof of Lemma 3.3 uses Doob's inequality for the submartingale exp(λ(Γ^N(Λ̃)_t - Γ(Λ̃)_t)). However, the exponential of a martingale is not necessarily a submartingale unless the martingale has independent increments or the exponential is convex. The authors should justify that the process is indeed a submartingale, or use a different inequality (e.g., the exponential martingale inequality for Poisson processes).","section":"Section 3.1, Lemma 3.3","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The representation (4.1) defines Λ^N_t in terms of a Poisson random measure Q and independent Poisson processes P^i. The definition of the jump size ΔΛ^N_s involves the infimum over x such that (1/N) P^{NΛ^N_{s-}}(N ˜V^{Λ^N}(s-,[0,x])) + 1/N - x < 0. This is a complicated expression; the authors should provide a brief explanation of why this correctly captures the jump condition.","section":"Section 4.1, equation (4.1)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof of Lemma 5.2 uses an inductive argument that assumes the property holds up to the i-th jump time. The induction step relies on the inequality ∂_x ˜V^{Λ^N}(t-,0) ≥ ∂_x ˜V^{¯Λ^N}(t-,0) to conclude that if ¯Λ^N jumps, then Λ^N must also jump. However, the jump condition for Λ^N involves the infimum over x, and it is not immediate that a larger derivative implies a jump occurs. The authors should provide a more detailed justification.","section":"Section 5, Lemma 5.2","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof of Lemma B.1 is quite involved and uses several approximations. The notation is sometimes confusing; for example, the use of Z^N(s,x) and the bound (B.3) could be explained more clearly. The authors should consider adding a brief outline of the proof strategy before diving into the details.","section":"Appendix B, proof of Lemma B.1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The reference list includes several unpublished works (e.g., [5], [22], [25]). The authors should ensure that these are cited appropriately and that the results used from them are clearly indicated. In particular, [5] is cited as 'ArXiv:2507.16703, 2025' and [22] as 'ArXiv:2409.20421, 2026'. The paper should state whether these are under review or accepted, and whether the results are reliable.","section":"References","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"major_revision"},"author_rebuttal":null,"desk_editor":{"model":"deepseek/deepseek-v4-flash","letter":"Tom,\n\nThis is a serious piece of work. The main result—a functional CLT for the barrier in the subcritical (a<1) particle system, with explicit t^{-1/4} scaling and Volterra covariance—is new and nontrivial. The connection to 1D-MDLA fluctuations and the companion SPDE for the empirical measure are nice bonuses. The proof strategy is coherent: Poisson random measure representation, martingale estimates, tightness, contraction argument. The constant K_a is not fitted; it comes from the similarity solution (1.4). The paper is honest about what it relies on.\n\nThe soft spot is the regularity of the limiting free boundary Λ. The FCLT and SPDE assume Λ is continuously differentiable on (0,T] with |dot Λ_t| ≤ C t^{-1/2}. This is imported from [17] and [3] via Proposition A.1, which truncates the initial data and passes to the limit. The stress-test note raises a fair point: the uniform derivative bound from [17] is stated for truncated data, and the constant may depend on the L^1 norm of the truncated density, which grows like log n for g≡a. The paper does not fully address whether the constant can be chosen independently of n, nor how the limit Λ inherits the bound. That said, the authors are aware of the issue—they state in Lemma B.1 that Λ is assumed continuously differentiable on [h/3,T] and note that this is what is at stake. The central argument would collapse if this regularity failed, but I think it's more likely that the bound can be made uniform via a more careful analysis (the weak feedback condition ||g||_∞<1 gives a lot of control). This is a gap, not a fatal flaw.\n\nOther concerns are minor: some lemmas (4.3, 6.3, 6.9) are sketched rather than fully proved, and the paper cites [22] (an unpublished arXiv preprint) for part of the well-posedness statement. But Proposition A.1 provides an independent route, so this is not load-bearing.\n\nWho is this for? Probabilists working on interacting particle systems, Stefan problems, and McKean-Vlasov limits. It deserves a serious referee—the core result is important and the proof program is sound. I'd recommend conditional acceptance: ask the authors to clarify the regularity argument for Λ, either by providing a self-contained proof or by citing a source that handles the non-integrable case explicitly.\n\nBest,\n[Your name]","headline":"Solid FCLT for the a<1 particle system; the regularity assumption on Λ is a real gap, but the paper acknowledges it and the central result likely holds.","tokens_in":41955,"tokens_out":632,"duration_ms":9845},"pith_extraction":null,"created_at":"2026-11-13T09:19:19.183963+00:00","model_set":{"reader":"deepseek/deepseek-v4-flash"},"pipeline_version":"pith-pipeline@v1.4.1-daily-deepseek-or-medium","input_kind":null,"input_chars":null,"input_total_chars":null,"passes_missing":null,"published_standing":null,"published_confidence":null,"objection_audit":null,"run_record":null,"basis":{"kind":"not_recorded","source":null,"chars_given":null,"chars_total":null},"receipt":null,"falsifier":null,"supporting_citations":[],"lane_accuracy":{"status":"not yet measured","lane":{"model":"deepseek/deepseek-v4-flash","input_kind":null},"url":"https://pith.science/accuracy"},"review_version":1}