{"id":"e8f677cf-e377-48eb-b8e0-fa7ba191a45a","arxiv_id":"2501.11442","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Bubble areas in phase-separated active Brownian particles obey a Langevin equation with constant negative drift and noise proportional to the perimeter, reproducing area and lifetime statistics.","lead":"Using million-particle simulations, the authors find that vapor bubbles inside dense active fluids grow and shrink randomly, and that their area changes follow a simple random-walk equation. This could help explain a puzzling bubble phase in active matter and improve continuum theories of phase separation.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Markov closure of the bubble-area dynamics is untested: single-lag moment fits do not establish that Eq. (2) reproduces the full transition probabilities.","rationale":"The reader's weakest assumption identified the Markov closure and the single-time-lag fit; I agree that this is the most load-bearing point. I sharpen it by noting that the MFPT check provides no constraint on the noise term, so the only nontrivial tests of the multiplicative noise are the stationary distribution (which is tautological given the fit) and the lifetime distribution (which is partly supported but uses an extra parameter in the analytic approximation). The proposed test directly targets the Markov property by comparing full transition probabilities at multiple lags and checking lag-independence of the fitted coefficients. This would either resolve the concern or reveal that Eq. (2) is only an effective one-lag description. Since the concern is real and testable but not yet demonstrated to invalidate the model, the verdict remains CONDITIONAL, i.e., unchanged relative to the reader's assessment.","tokens_in":8280,"tokens_out":6527,"duration_ms":81491,"concrete_test":"Simulate or numerically solve Eq. (2) with the fitted ν = 19.3 σ²/τ and Γ = 348 σ³/τ to compute the conditional distribution P(A_t | A_0) for lags t ∈ {0.01τ, 0.05τ, 0.1τ, 0.5τ} and initial areas A_0 ∈ {50, 200, 500} σ². Measure the same conditional distributions in the ABP simulation, restricted to interior bubbles that do not merge, with the same binning. If the Kullback-Leibler divergence or the first two conditional moments deviate beyond statistical error for any lag, the closed Markov assumption fails. Additionally, recompute ν and Γ from increments with Δt = 0.01τ and 0.02τ; if the inferred coefficients change systematically with Δt, the Δt = 0.1τ values are not the continuum coefficients of a diffusion process.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that A_t is a closed Markov diffusion governed by Eq. (2), with drift -ν and noise √(2Γ A^{1/4}). The evidence is (i) single-time-lag moment fits at Δt = 0.1τ (Fig. 2b), (ii) the stationary distribution Eq. (4), (iii) the mean first-passage time T(A) = A/ν (Fig. 3b), and (iv) lifetime distributions from numerical integration of Eq. (2) (Fig. 3c,d). The load-bearing weakness is that this evidence does not actually test the Markov closure or the functional form of the noise. Point (iii) is independent of D(A): for any diffusion coefficient D(A) that vanishes at A=0, the MFPT to zero with constant drift -ν is T(A)=A/ν, so it validates only the constant drift, not the A^{1/4} noise. Point (ii) is a consistency check, not a prediction, because ν and Γ were fitted to the same increments that produce the stationary distribution, so agreement is guaranteed by construction. Point (iv) is a genuine out-of-sample test, but the full Langevin simulation is compared only for lifetime statistics at a single initial area (Fig. 3c) and for the population lifetime (Fig. 3d), whose derivation requires the additional tuned parameter A_eff = A0/2 for the inverse-Gaussian approximation. More fundamentally, the paper never checks whether the inferred coefficients are independent of the sampling lag or whether the two-parameter diffusion reproduces the full conditional distribution P(A_t | A_0). If the true dynamics is non-Markovian, e.g., because bubble shape or local liquid density relaxes on a timescale comparable to 0.1τ, the single-lag fits could coincidentally match the stationary and first-passage statistics while failing to predict the propagator. The Markov closure in Eq. (2) is thus the single assumption on which all derived statistics depend, and it is the least tested.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript studies vapor bubbles inside the liquid domains of phase-separated active Brownian particles using large-scale particle-based simulations. It reports that 97% of bubbles nucleate, grow, and dissolve within the bulk of the liquid domain, and proposes that the bubble area A_t obeys an Itô Langevin equation dA_t = -ν dt + sqrt(2Γ A^{1/4}) dW_t (Eq. 2), with fitted constants ν = 19.3 σ²/τ and Γ = 348 σ³/τ. From this equation the authors derive a stationary area distribution, a linear mean first-passage time T(A) = A/ν, and lifetime distributions, and they compare these against the simulations. They also introduce a lattice gas model for bubble shape and show that the mean asphericity decreases with area. The paper contrasts its findings with Active Model B+ and highlights limitations of continuum theories.","tokens_in":8538,"tokens_out":4951,"duration_ms":57002,"significance":"The paper addresses a genuinely open question in motility-induced phase separation: the stochastic dynamics of vapor bubbles inside dense domains. The proposed effective Langevin equation is simple, physically interpretable (noise proportional to the bubble perimeter), and it makes explicit predictions that are partly confirmed by a large data set of 30,012 simulated bubbles. The connection to a broader class of sub-demographic multiplicative noise processes is appealing and could be of interest beyond this specific system. However, the validation presented is mostly in-sample consistency rather than independent prediction: the drift and diffusion coefficients are fitted from the same simulation increments used to test the stationary distribution, the MFPT is insensitive to the functional form of the diffusion coefficient, and the lifetime distribution requires an additional tuned parameter A_eff. The Markov and Gaussian-white-noise assumptions underlying Eq. (2) are not directly tested. These issues currently limit the strength of the claim that Eq. (2) 'fully characterizes' the bubble statistics.","major_comments":[{"comment":"The stationary distribution p(A) ∝ A^{-1/2} exp(-2ν√A/Γ) is derived from the same fitted values of ν and Γ that were obtained from the simulation increments in Fig. 2(b). The agreement in Fig. 3(a) is therefore a consistency check, not an independent validation of the Langevin form. The same applies to the mean first-passage time T(A) = A/ν shown in Fig. 3(b): for any diffusion coefficient D(A) with D(0)=0, the equation -ν T' + D(A) T'' = -1 admits the solution T(A)=A/ν, so this result tests only the constant drift and not the perimeter-proportional noise. The authors should either explicitly label these as consistency tests or demonstrate prediction by, for example, fitting on one part of the data and predicting another part.","section":"Eq. (4) and Fig. 3(a)"},{"comment":"The Markov property and the Gaussian white-noise assumption are not tested. The drift and diffusion are estimated from moment fits at a single lag Δt = 0.1τ, and the paper does not check whether the inferred coefficients are independent of the sampling lag, whether the transition density P(A_t | A_0) satisfies the Chapman-Kolmogorov equation, or whether the conditional increment distribution is approximately Gaussian for small Δt. Without such checks, Eq. (2) is a Kramers-Moyal truncation whose validity as a closed stochastic evolution equation for A_t is unsupported. The authors should add at least one direct test of the Markov property, for example by comparing two-lag predictions with direct simulation or by showing that the inferred drift and diffusion do not change when Δt is varied.","section":"Eq. (2) and Fig. 2(b)"},{"comment":"The lifetime distribution in Eq. (5) uses a second fitted parameter A_eff = A0/2, and the population lifetime w(T) in Eq. (6) uses an additional cutoff Amin = 10σ², both chosen for best fit. The inverse-Gaussian curves in Fig. 3(c,d) are therefore not parameter-free predictions of Eq. (2). Moreover, the 'Langevin dynamics' curves in those panels are numerical solutions of the very equation that was fitted, so they cannot independently validate the model against the particle simulations. The authors should clarify the status of A_eff and Amin, ideally deriving them from the model or treating the lifetime comparison as a demonstration that the model can be tuned to match the data.","section":"Eq. (5) and Fig. 3(c,d)"},{"comment":"The small-area regime is explicitly identified as problematic: the first two drift data points deviate from the constant value, and the stationary distribution deviates for the smallest bubble sizes. Since bubble lifetimes are controlled by the dynamics near A=0, these deviations are not merely cosmetic. The paper should quantify the range of A over which Eq. (2) is validated and discuss whether the cutoff Amin (used for normalization and in Eq. (6)) biases the lifetime statistics. If the model is intended only for intermediate and large areas, the claim that it 'fully characterizes' lifetime statistics should be weakened accordingly.","section":"Fig. 2(b) and Fig. 3(a)"}],"minor_comments":[{"comment":"The supplementary material is referenced with a placeholder title and link ('link-provided-after-publication'); since many technical details and additional figures are deferred to [19], the manuscript is not self-contained for review. The authors should provide the supplemental material as part of the submission.","section":"Reference [19]"},{"comment":"It would be helpful to show a Gaussian fit (or other comparison) overlaid on the empirical distribution of ΔA for a few representative values of A, to support the Gaussian white-noise assumption used in Eq. (2).","section":"Fig. 2(a)"},{"comment":"The text contains a typo: 'motility-induces phase separation' should read 'motility-induced phase separation'. Also, the Péclet number is written as 'P´eclet' with inconsistent accents.","section":"Conclusion"},{"comment":"The lattice gas parameter ε = 1.1 is fitted to the asphericity data, but no error bars or sensitivity analysis are shown. A brief statement of how ε was determined and how sensitive the curves are to this parameter would strengthen the morphological comparison.","section":"Fig. 4(b)"},{"comment":"The general family dA_t = -νA^β dt + sqrt(2ΓA^α) dW_t is introduced at the end without any further analysis. This is fine as outlook, but the notation 'sub-demographic noise' is used without definition; a short explanation would help readers unfamiliar with demographic noise.","section":"Eq. (8)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is likely to be of interest to the soft-matter/active-matter community, and the simulation effort is substantial. My main concern is that the central claim of 'fully characterizing' the bubble statistics is not yet supported by independent validation: the stationary distribution and MFPT are derived from the same fitted parameters, and the lifetime distribution uses an extra tuned parameter. The authors should be encouraged to add explicit tests of Markovianity and to show that the inferred coefficients are lag-independent, or at least to clearly frame the comparison as consistency checks. The reliance on a placeholder supplemental reference is also a practical problem for the review process. If the authors address the validation issues, I would be happy to reconsider; as it stands, the paper overstates the predictive power of its effective Langevin description."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: the paper gives a clean, minimal description of vapor-bubble area dynamics in phase-separated ABPs: dA = -nu dt + sqrt(2Gamma A^{1/4}) dW, with nu and Gamma fitted from simulations. That is genuinely new and useful. Prior work saw bubbles and studied them in AMB+, but nobody had pulled out this reduced stochastic description with perimeter-proportional noise. The large simulation (30k bubbles) and the generalized Gibbs distribution are concrete. The lattice gas model is a bonus and captures the trend in asphericity. Credit where due: the paper is clearly written, the statistics are handled carefully for the bulk of the population, and the model does reproduce the stationary distribution and MFPT across the relevant range.\n\nNow the soft spots, in proportion. The stress-test note is right: the Markov closure is the load-bearing assumption and it is the least tested. The coefficients are fitted to single-lag moments at Delta t = 0.1 tau, and the stationary distribution is a consistency check, not a prediction, because it follows from the same fitted nu and Gamma. The MFPT T(A) = A/nu is independent of the noise amplitude for any D(A), so Fig. 3b validates only the constant drift, not the A^{1/4} noise. The lifetime comparison is a genuine out-of-sample test of the Langevin equation, but it needs the extra fitted A_eff = A0/2, which weakens it. The paper never checks whether the inferred nu and Gamma depend on the sampling lag, nor whether the Langevin equation reproduces the full conditional distribution P(A_t | A_0). That is a real gap, not a manufactured one.\n\nThat said, I don't think the paper's central conclusion is wrong. The deviations at small area are acknowledged. The comparison to AMB+ is suggestive but not direct; the claim that AMB+ has a peak in the area distribution is asserted, not shown. That's a minor issue. The unavailable supplement is an inconvenience, but the main text has enough for a referee to evaluate.\n\nBottom line: this is a useful empirical paper for people working on MIPS and active phase separation. The Langevin characterization is likely to be cited as a benchmark. It deserves peer review; the referee should ask for lag-dependence checks and a direct test of the propagator, and ideally make the supplement available. I'd bring it to a reading group and cite it.","headline":"A useful empirical characterization of MIPS bubbles as a one-variable Langevin process, but the central claim is under-tested: the fits use the same data and the Markov property is never checked.","tokens_in":9218,"tokens_out":2107,"would_cite":true,"duration_ms":22804,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["82C31","60H10","82C26"],"pacs":[],"model":"deepseek-v4-flash","headline":"Vapor bubbles in phase-separated active liquids mostly dissolve internally, and their areas obey a Langevin equation with constant negative drift and perimeter-proportional noise, yielding a generalized Gibbs distribution and lifetimes…","keywords":["active Brownian particles","motility-induced phase separation","vapor bubbles","Langevin dynamics","multiplicative noise","lattice gas model","bubble lifetime","sub-demographic noise"],"falsifier":"One concrete check is to measure the first and second moments of area increments at several lag times (e.g., 0.02, 0.05, 0.1, and 0.2 $\\tau$) and test whether the inferred $\\nu$ and $\\Gamma$ stay fixed and the transition probabilities obey the Chapman–Kolmogorov equation; if the coefficients change with lag time or the increment distribution is non-Gaussian, the Markovian Itô description fails. A second check is to follow the bubbles that merge with the domain boundary and test whether their area statistics obey the same $\\nu$ and $\\Gamma$; if not, the restriction to internally dissolving bubbles biases the result.","tokens_in":7978,"feed_emoji":"🫧","tokens_out":6902,"duration_ms":61605,"temperature":0.7,"pith_summary":"The paper studies vapor bubbles that appear inside liquid domains of phase-separated active Brownian particles. It claims that most such bubbles nucleate, grow, and dissolve entirely inside the liquid, and that their area fluctuations are not incidental but obey a precise stochastic law: a Langevin equation with constant negative drift and noise proportional to the bubble perimeter. If true, the entire area statistics—stationary distribution, mean first-passage time, and lifetime distribution—are determined by two fitted parameters. The paper also offers a lattice gas model that reproduces the observed decrease of bubble asphericity with area, and argues that current continuum theories, such as Active Model B+, miss this stochastic structure.","feed_headline":"Bubble sizes and lifetimes obey a single two-parameter law","feed_subtitle":"Simulations show bubble area noise scales with perimeter, so lifetimes grow linearly with area.","key_machinery":"The load-bearing object is the effective Langevin equation for bubble area, Eq. (2), written in Itô form. Its multiplicative noise amplitude, $\\sqrt{2\\Gamma A^{1/4}}$, is proportional to the square root of the bubble perimeter, so the diffusion coefficient $\\Gamma\\sqrt{A}$ scales with perimeter. From this equation, via the Fokker–Planck equation, the stationary distribution, first-passage time, and lifetime distribution follow. The lattice gas model—square blocks added or removed at boundary sites with rates set by the fitted drift and diffusion and a Boltzmann factor $e^{\\epsilon m_i}$ favoring more-coordinated sites—supplies the morphological predictions, reproducing the decline of asphericity with area.","core_discovery":"The central discovery is that the bubble area $A_t$ is a Markov process described by the Itô Langevin equation $dA_t = -\\nu\\,dt + \\sqrt{2\\Gamma A_t^{1/4}}\\,dW_t$, with $\\nu = 19.3\\,\\sigma^2/\\tau$ and $\\Gamma = 348\\,\\sigma^3/\\tau$ fitted from particle simulations. This equation is equivalent to a Fokker–Planck equation whose stationary solution is the generalized Gibbs distribution $p(A) \\propto A^{-1/2} \\exp(-2\\nu\\sqrt{A}/\\Gamma)$, which matches the simulated area distribution down to the resolution limit. It also yields the mean first-passage time $T(A) = A/\\nu$ and an inverse Gaussian lifetime distribution that fit the simulated lifetimes. A companion lattice gas model with perimeter-proportional addition and removal rates reproduces the measured decrease of asphericity with area, supporting the interpretation that boundary fluctuations drive the process. The paper contrasts these results with Active Model B+, which predicts a peaked area distribution not seen in the particle simulations.","pith_inferences":["Beyond the paper: if the same perimeter-proportional noise holds in three dimensions, the analogous area (volume) exponent should change from $1/4$ to $1/3$, since the diffusion coefficient would scale with surface area, i.e., $V^{2/3}$.","Beyond the paper: the measured discrepancy with Active Model B+ suggests continuum theories may need nonlocal terms; a testable extension is to compute the area distribution predicted by Active Model B+ in the same parameter regime and compare it quantitatively with Fig. 3(a).","Beyond the paper: the Markov property could be tested by checking higher-order moments of area increments, for example the ratio of third to second moments, which would reveal non-Gaussian noise and break the Itô Langevin description."],"forward_implications":["Once $\\nu$ and $\\Gamma$ are fitted, the stationary area distribution and the mean lifetime $T(A) = A/\\nu$ are predictions with no further free parameters.","The inverse Gaussian form of the lifetime distribution means that measuring lifetimes in experiments or simulations is enough to extract the drift and noise strength.","The effective line tension $\\lambda = \\nu/(\\sqrt{\\pi}\\,\\Gamma)$ turns the noise-dominated bubble statistic into a thermodynamic-like free energy, suggesting a consistent interfacial description even away from equilibrium.","The lattice gas model implies that local boundary events, not global rearrangements, set the shape statistics, connecting area noise to the same perimeter mechanism.","The general equation $dA_t = -\\nu A^{\\beta}\\,dt + \\sqrt{2\\Gamma A^{\\alpha}}\\,dW_t$ with $\\alpha = 1/4$, $\\beta = 0$ defines a class of sub-demographic noise processes whose statistical properties are interesting in their own right."],"supporting_citations":[{"why":"Established athermal phase separation in self-propelled particles, motivating the nonequilibrium setting.","marker":"[5]"},{"why":"Supplies the ABP simulation model and phase-separating dynamics that the present simulations extend.","marker":"[6]"},{"why":"Documented the existence of mesoscopic vapor bubbles in liquid domains, the phenomenon under study.","marker":"[11]"},{"why":"Provides curvature-dependent tension and tangential flows at the interfaces, relevant to bubble morphology.","marker":"[12]"},{"why":"Introduces Active Model B+, the continuum theory whose bubble-area predictions are compared and found wanting.","marker":"[13]"},{"why":"Gives the first-passage-time density for a Brownian particle to reach the origin, used for the inverse Gaussian lifetime distribution.","marker":"[21]"},{"why":"Companion first-passage calculation used in the same derivation of the lifetime distribution.","marker":"[22]"},{"why":"Identifies the lifetime distribution as the inverse Gaussian distribution.","marker":"[23]"}],"fun_headline_variants":["Bubble area noise scales with perimeter, not area","Active bubble lifetimes scale linearly with area","Langevin law describes bubble dynamics in active matter","Simulations show bubble area follows generalized Gibbs","Bubble dynamics in active matter follow a simple law"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The analysis assumes that the bubble area alone is a closed Markov process: the drift and noise depend only on the instantaneous area, with Gaussian white noise, and all other variables are either negligible or slaved to area; it also assumes that discarding the roughly 3% of bubbles that reach the boundary does not bias the fitted drift and diffusion.","fun_headline_variants_meta":{"raw":{"variants":["Bubble area noise scales with perimeter, not area","Active bubble lifetimes scale linearly with area","Langevin law describes bubble dynamics in active matter","Simulations show bubble area follows generalized Gibbs","Bubble dynamics in active matter follow a simple law"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000183,"raw_usage":{"total_tokens":1271,"prompt_tokens":861,"completion_tokens":410,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":477,"completion_tokens_details":{"reasoning_tokens":339}},"tokens_in":477,"tokens_out":410,"duration_ms":4413,"temperature":1.0,"reasoning_tokens":339,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T18:15:52.866047+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"One concrete check is to measure the first and second moments of area increments at several lag times (e.g., 0.02, 0.05, 0.1, and 0.2 $\\tau$) and test whether the inferred $\\nu$ and $\\Gamma$ stay fixed and the transition probabilities obey the Chapman–Kolmogorov equation; if the coefficients change with lag time or the increment distribution is non-Gaussian, the Markovian Itô description fails. A second check is to follow the bubbles that merge with the domain boundary and test whether their area statistics obey the same $\\nu$ and $\\Gamma$; if not, the restriction to internally dissolving bubbles biases the result.","supporting_citations":[{"cited_title":"Athermal phase separation of self-propelled particles with no alignment,","cited_arxiv_id":null,"evidence_quote":"Established athermal phase separation in self-propelled particles, motivating the nonequilibrium setting."},{"cited_title":"Structure and dynamics of a phase-separating active colloidal fluid,","cited_arxiv_id":null,"evidence_quote":"Supplies the ABP simulation model and phase-separating dynamics that the present simulations extend."},{"cited_title":"Phase behaviour of ac- tive brownian particles: the role of dimensionality,","cited_arxiv_id":null,"evidence_quote":"Documented the existence of mesoscopic vapor bubbles in liquid domains, the phenomenon under study."},{"cited_title":"Curvature-dependent tension and tangential flows at the interface of motility-induced phases,","cited_arxiv_id":null,"evidence_quote":"Provides curvature-dependent tension and tangential flows at the interfaces, relevant to bubble morphology."},{"cited_title":"Cluster phases and bubbly phase separation in active fluids: Reversal of the ostwald process,","cited_arxiv_id":null,"evidence_quote":"Introduces Active Model B+, the continuum theory whose bubble-area predictions are compared and found wanting."},{"cited_title":"Zur Theorie der Fall- und Steigversuche an Teilchen mit Brownscher Bewegung,","cited_arxiv_id":null,"evidence_quote":"Gives the first-passage-time density for a Brownian particle to reach the origin, used for the inverse Gaussian lifetime distribution."},{"cited_title":"Notiz ¨ uber die Berechnung der Brownschen Molekularbewegung bei der Ehrenhaft- Millikanschen Versuchsanordnung,","cited_arxiv_id":null,"evidence_quote":"Companion first-passage calculation used in the same derivation of the lifetime distribution."},{"cited_title":"The inverse gaussian distribution,","cited_arxiv_id":null,"evidence_quote":"Identifies the lifetime distribution as the inverse Gaussian distribution."}],"review_version":1}