{"id":"f8bcace6-4b41-4494-b4af-43209cd09237","arxiv_id":"2411.12368","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"A transition-path sampling algorithm for run-and-tumble particles is derived, validated, and used to show that tumbling reduces mean barrier-crossing time.","lead":"This paper adapts transition-path sampling, a rare-event simulation method, to run-and-tumble particles that alternate between passive diffusion and self-propelled runs. The method is validated against brute-force simulations and shows that tumbling can shorten the time to cross an energy barrier.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Acceptance rule Eq. (17) depends on a numerically estimated steady-state density ρ in the reactant basin; no error analysis is reported, and the benchmarks validate only the TPT distribution, leaving path-space observables untested.","rationale":"The paper's central methodological claim is that TPS can be extended to RTPs by defining a backward dynamics and deriving the Metropolis acceptance probability Eq. (17). The derivation of Eq. (17) is careful and the benchmarks against brute-force MD for the TPT distribution are strong support. The weakest point is the numerical steady-state density ρ(ω) in the reactant basin, which enters the acceptance rule directly. The paper acknowledges that no analytic form is known and estimates ρ from Langevin simulations, but gives no convergence or error analysis. This is load-bearing because an inaccurate ρ biases the entire path ensemble, and the reported benchmarks validate only the TPT distribution, not the full transition-path ensemble. The reader identified the same assumption; my concern does not change the verdict but reinforces the conditionality: the method is well-supported for the tested rates, but its general correctness depends on the accuracy of the numerical ρ input. The proposed concrete test would settle whether this concern is practically significant by comparing path-space observables from TPS and MD, both for a benchmarked and an unbenchmarked rate pair.","tokens_in":20515,"tokens_out":19483,"duration_ms":182890,"concrete_test":"Run the TPS algorithm for the benchmarked rate pair E (λ0→1=0.24/τ, λ1→0=0.70/τ) using two independent ρ estimates: (i) the original histogram, and (ii) a converged estimate from a 10× longer MD run with spatial bins of 0.1L and a reported autocorrelation time. Compare the TPS-sampled reactive density m(r) and current J(r) from both estimates against brute-force MD (10^6 reactive paths). If m(r) and J(r) deviate beyond statistical error, the numerical ρ is biasing the path ensemble even where the TPT distribution matched. Also perform the same comparison for an unbenchmarked rate pair (e.g., λ0→1=0.5/τ, λ1→0=0.5/τ).","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section III states that ρ(ω) is estimated as the relative occupation frequency of microstates in R from Langevin simulations, and Eq. (17) uses this estimate in the ratio ρ(ω0^new)/ρ(ω0^old). This is the only non-analytic input in the acceptance rule. Backward-shooting endpoints are typically at the boundary of R, where the stationary density may be small and poorly sampled; without reporting binning, equilibration, or statistical error, the estimate's accuracy is unknown. If ρ is biased, the sampled transition-path ensemble is biased for all observables, not just the TPT distribution. The benchmarks (Figs. 3 and 7) compare only P(t_TPT), a coarse projection of the ensemble. The reactive density m(r) and current J(r) in Fig. 5 are presented only for TPS, so a mismatch with brute-force MD cannot be ruled out. The claim that the method is valid for arbitrary rate pairs therefore rests on an unquantified numerical input.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript elaborates a transition-path-sampling (TPS) scheme for run-and-tumble particles (RTPs) that alternate between passive and active Brownian phases. The key theoretical step is the definition of a backward dynamics (Eqs. (10)-(12)) and the derivation of the corresponding Metropolis acceptance probability (Eq. (17)), which accounts for the lack of microscopic reversibility through a ratio involving the numerically estimated steady-state density ρ(ω0). The method is validated by comparing transition-path-time (TPT) distributions against brute-force Langevin simulations in a double-well potential (Figs. 3 and 7). The authors use the sampled transition-path ensemble to show that tumbling reduces the mean transition-path time below both the fully passive and fully active limits, with a minimum near λ0→1 ≈ 0.2/τ and λ1→0 ≈ 0.7/τ (Fig. 4).","tokens_in":20657,"tokens_out":9078,"duration_ms":90250,"significance":"The extension is significant because it brings a widely used rare-event sampling method to a paradigmatic active-matter model, complementing the earlier ABP adaptation of Ref. [32]. The derivation is self-contained and free of fitted parameters, and the validation against direct simulation is an appropriate external benchmark. The observation that tumble-induced randomization shortens transition paths is a concrete, testable prediction. The appendices (Fokker-Planck equations and stochastic path integral) strengthen the formal basis. The main caveat, discussed below, is that the acceptance rule relies on a numerical steady-state density whose accuracy is not quantified.","major_comments":[{"comment":"The acceptance probability uses ρ(ω0^new)/ρ(ω0^old), where ρ is estimated as the relative occupation frequency of microstates in the reactant basin. No binning, equilibration time, simulation length, or statistical uncertainty is reported for this estimate. Because backward-shooting endpoints typically lie on the boundary of R, where the stationary density is small, the estimate may be poorly converged there. A biased ρ biases the whole sampled transition-path ensemble, not only P(t_TPT). Please add a convergence analysis of ρ (e.g., dependence on bin size and sampling time) and report its statistical error, or show that the acceptance statistics are insensitive to this input.","section":"Sec. III, Eq. (17)"},{"comment":"The reactive density m(r) and current J(r) are presented only for TPS ensembles, with no comparison to direct Langevin simulation. Since Figs. 3 and 7 benchmark only the TPT distribution, which is a one-dimensional projection of the path ensemble, agreement in P(t_TPT) does not by itself establish that the sampled path-space distribution is unbiased. Please benchmark m(r) and J(r) against brute-force MD for at least the cases in Fig. 5, or provide an argument why the TPT match suffices.","section":"Sec. III, Fig. 5"},{"comment":"Fig. 4 reports ⟨t_TPT⟩ without error bars, and the text states that P(t_TPT) was checked 'across all combinations of rates λ0→1, λ1→0 ∈ [0,10/τ]' although only the points A–G are shown (Fig. 7). Please report statistical errors (e.g., block averages over the 10^8 reactive paths) and either show the full grid or qualify the statement to 'the tested set of rate combinations'.","section":"Sec. III, Fig. 4"}],"minor_comments":[{"comment":"There is a typo: 'findinds' should be 'findings', and 'RPTs' should be 'RTPs' in the sentence 'for RPTs not only the average TPT is decreased'.","section":"Sec. IV, Conclusions"},{"comment":"The optimal rates are given as λ0→1 = 0.2/τ in the text but as λ0→1 = 0.24/τ in the caption of Fig. 4; please reconcile these values.","section":"Sec. III, Fig. 4"},{"comment":"The index ℓ is not defined before its use in the product; please define it (presumably the length of the path segment).","section":"Eq. (18)"},{"comment":"The colorbar labels show '-0.0000' and '0.0000' for the minimum m(r); the negative zero is confusing and should be replaced by 0.","section":"Fig. 5"},{"comment":"The phrase 'RTP particles' is redundant; use 'RTPs'.","section":"Appendix B"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the scope of the journal. The analytic part of the derivation is sound, but the validation gap around the numerical steady-state density and the lack of error reporting justify a major revision in my view. I see no citation or novelty problems; the relationship to Ref. [32] is clearly stated."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read the Kiechl, Franosch, Caraglio paper on TPS for run-and-tumble particles. Bottom line: it does what it says, the algebra is right, and the tumble-speedup result is real within the parameters tested. It deserves peer review, but the authors should tighten two things before publication.\n\nWhat's actually new: this is the first TPS adaptation to RTPs with alternating passive and active phases. The acceptance probability derivation (Eq. 17) is self-contained, and the appendices give the Fokker-Planck equations and a stochastic path integral for the model. The passive limit recovers the standard TPS formula, and the fully active limit recovers the ABP result from Zanovello et al. Those checks are the right ones. The benchmarking is extensive: TPT distributions from TPS match brute-force MD across many rate pairs, including the extreme limits. That is solid evidence the algorithm works for those parameters.\n\nThe interesting physical finding is that tumbling reduces the mean transition-path time below both the passive and the fully active cases, with a minimum around 8.5τ for λ0→1=0.2/τ, λ1→0=0.7/τ. The explanation—tumbles let the particle detach from the barrier walls rather than surf along them—is plausible and consistent with the reactive density and current maps.\n\nNow the soft spots. The acceptance rule Eq. (17) uses the steady-state density ρ(ω) in the reactant basin, which is estimated numerically from Langevin simulations. There is no binning, equilibration, or error analysis for this estimate, and it is genuinely load-bearing: if it is biased, every observable sampled is biased. The benchmarks validate only P(t_TPT), which is a coarse projection; the reactive density m(r) and current J(r) are shown only for TPS, not compared with brute-force MD. I would want that comparison, or at least a clear statement that the path-space observables are predictions of the method under the numerical ρ. Also, Fig. 4, which supports the central speedup claim, has no error bars. With 10^8 paths the statistical error is probably tiny, but the authors should show it. And there is no code or data release, which makes re-implementation harder than it needs to be.\n\nNone of this undermines the derivation. The stress-test worry about ρ is valid but is an implementation limitation, not a flaw in the method's logic. This is the kind of paper that belongs in a serious journal after moderate revision.\n\nMy recommendation: send it to peer review, with requests for error bars, a path-space validation against MD, and a more prominent caveat about the numerical ρ.","headline":"A sound extension of TPS to run-and-tumble particles with a plausible speedup result, held back by missing error analysis and path-space validation.","tokens_in":21215,"tokens_out":3711,"would_cite":true,"duration_ms":34993,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Run-and-tumble particles can be studied with transition-path sampling, and alternating passive and active phases is found to minimize the mean barrier-crossing time.","keywords":["transition-path sampling","run-and-tumble particles","active matter","rare events","transition-path time","Metropolis acceptance probability","target search","microscopic reversibility"],"falsifier":"Run the TPS algorithm and a direct Langevin simulation for a rate pair outside the benchmarked set, for example $\\lambda_{0\\to1}=8/\\tau$ and $\\lambda_{1\\to0}=3/\\tau$, and compare the full transition-path-time distributions; a statistically significant mismatch would show that the acceptance rule or the numerical reactant density is not generally valid.","tokens_in":20279,"feed_emoji":"🦠","tokens_out":7730,"duration_ms":80682,"temperature":0.7,"pith_summary":"The paper generalizes transition-path sampling (TPS), a Monte Carlo method that samples the rare reactive trajectories connecting two basins, to run-and-tumble particles (RTPs) that switch between passive diffusion and self-propelled runs. Because RTPs lack microscopic reversibility, the standard backward dynamics used by TPS is not available; the authors instead prescribe a backward dynamics with a well-defined path probability and derive the corresponding Metropolis acceptance probability, Eq. (17). They benchmark the resulting sampler against direct Langevin simulation and obtain matching transition-path-time distributions across all tested rate pairs. The physics payoff is that alternating passive and active phases with tumbling shortens the mean transition-path time below both the fully passive and the fully active cases, with a minimum of about $8.5\\tau$ at rates $\\lambda_{0\\to1}=0.2/\\tau$ and $\\lambda_{1\\to0}=0.7/\\tau$. This matters because barrier-crossing events in active systems, such as bacteria searching for targets, are rare and otherwise costly to simulate directly.","feed_headline":"Alternating passive and active phases speeds barrier crossing","feed_subtitle":"A new path-sampling rule for run-and-tumble particles finds the fastest crossings at moderate tumbling rates.","key_machinery":"The load-bearing object is the backward propagator $\\bar{P}_{\\Delta t}(\\omega_i|\\omega_{i+1})$ of Eq. (13) together with the Metropolis factor of Eq. (17). The backward dynamics is chosen so that trajectories generated backward in time look like forward trajectories, and the passive-particle detailed balance relation (Eq. 16) lets the position propagators be rewritten in terms of Boltzmann weights and force--velocity exponentials. A telescoping identity (Eq. 18) collapses the phase-switching probabilities into endpoint factors, so the final acceptance ratio depends on the ratio of steady-state reactant densities, endpoint Boltzmann weights, and products over old and new path segments. In the passive limit the formula reduces to standard TPS, and in the fully active limit it reduces to the earlier active-Brownian-particle result, which is what allows shooting moves to sample the correct reactive-path ensemble.","core_discovery":"The central claim is that the irreversibility of RTP dynamics is not an obstacle to transition-path sampling. The authors define a backward dynamics (Eqs. 10--12) in which the self-propulsion velocity is odd under time reversal and the potential force is even, giving a single-step backward propagator with a well-defined path probability; joining forward and backward branches from an unperturbed shooting point, they derive the closed acceptance probability in Eq. (17). The only non-analytic input is the steady-state distribution $\\rho(\\omega_0)$ in the reactant basin, estimated numerically from Langevin simulations. With that input, the sampled transition-path-time distribution reproduces the brute-force result for rate pairs across $[0,10/\\tau]$, including the passive limit, the fully active limit, and the intermittent regime. The paper further claims that tumbling improves crossing by randomizing the propulsion direction and detaching trajectories from the potential walls, so a tuned passive-active mixture can outperform both pure passive diffusion and pure active swimming.","pith_inferences":["The numerical estimation of $\\rho(\\omega_0)$ is the practical bottleneck of the method; an on-the-fly or self-consistent estimate would make the sampler effectively black-box for arbitrary potentials, which the paper does not demonstrate.","The reported location of the TPT minimum ($\\lambda_{0\\to1}=0.2/\\tau$, $\\lambda_{1\\to0}=0.7/\\tau$) is specific to the double-well parameters used; scanning barrier height, Péclet number, and persistence would test whether the tumbling-speedup effect is universal or parameter-dependent.","Because TPS yields reactive trajectories without a reaction coordinate, the same algorithm could serve as a reference for forward-flux and machine-learning rare-event methods on active-particle problems, though no such comparison is made here.","The wall-surfing-versus-detachment mechanism suggests an experimental signature: in microfluidic barrier landscapes, tracking individual bacteria should show that cells with intermittent tumbling produce transition paths that are more concentrated near the minimum energy path than persistently swimming cells."],"forward_implications":["Rare barrier-crossing events of RTPs can be sampled efficiently without knowing a reaction coordinate beforehand, using the same trajectory-space Metropolis logic as equilibrium TPS.","Transition-path-time distributions for RTPs can be computed across the whole rate plane, so the effect of tumbling duration and frequency on rare transitions can be mapped systematically.","The finding that an intermittent passive-active mixture minimizes mean transition-path time identifies tumbling as a kinetic control parameter, not just a search-strategy parameter.","The authors argue the same backward-dynamics strategy applies to other irreversible active models, including chiral ABPs, anisotropic-diffusion particles, active Ornstein-Uhlenbeck particles, and gradient-orienting particles."],"supporting_citations":[{"why":"Supplies the original TPS framework and its path-space Metropolis acceptance condition, which the paper extends.","marker":"[8]"},{"why":"Provides the standard backward-dynamics construction for equilibrium TPS that the RTP generalization must replace.","marker":"[10]"},{"why":"Presents an earlier TPS–forward-flux hybrid for active particles and motivates the need for a backward-shooting alternative.","marker":"[31]"},{"why":"Demonstrates that any backward dynamics with a well-defined path probability can be used for active Brownian particles; the RTP acceptance rule follows that template.","marker":"[32]"},{"why":"Establishes the ABP target-search results and parameter choices used here for comparing transition-path-time behavior.","marker":"[43]"},{"why":"Justifies treating self-propulsion as odd and the potential force as even under time reversal when choosing the backward dynamics.","marker":"[66]"},{"why":"Documents the lack of a general analytic steady-state distribution for active particles, motivating the numerical estimate of $\\rho(\\omega_0)$.","marker":"[44]"}],"fun_headline_variants":["Moderate tumbling beats pure passive and active for barrier crossings","Path sampling for irreversible active particles","Tumbling rate tuned for fastest barrier crossing","Rare pathways in active matter sampled without reversibility"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The acceptance probability requires the steady-state distribution of microstates inside the reactant basin, which is not known analytically for active particles and is estimated here by counting occupation frequencies in long Langevin simulations; for rate pairs or potentials outside the benchmarked range, an inaccurate estimate would bias the sampled reactive trajectories.","fun_headline_variants_meta":{"raw":{"variants":["Moderate tumbling beats pure passive and active for barrier crossings","Path sampling for irreversible active particles","Tumbling rate tuned for fastest barrier crossing","Rare pathways in active matter sampled without reversibility"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00053,"raw_usage":{"total_tokens":2497,"prompt_tokens":833,"completion_tokens":1664,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":449,"completion_tokens_details":{"reasoning_tokens":1605}},"tokens_in":449,"tokens_out":1664,"duration_ms":12980,"temperature":1.0,"reasoning_tokens":1605,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T17:36:35.380810+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the TPS algorithm and a direct Langevin simulation for a rate pair outside the benchmarked set, for example $\\lambda_{0\\to1}=8/\\tau$ and $\\lambda_{1\\to0}=3/\\tau$, and compare the full transition-path-time distributions; a statistically significant mismatch would show that the acceptance rule or the numerical reactant density is not generally valid.","supporting_citations":[{"cited_title":"Transition path sampling and the calculation of rate constants,","cited_arxiv_id":null,"evidence_quote":"Supplies the original TPS framework and its path-space Metropolis acceptance condition, which the paper extends."},{"cited_title":"Transition path sampling,","cited_arxiv_id":null,"evidence_quote":"Provides the standard backward-dynamics construction for equilibrium TPS that the RTP generalization must replace."},{"cited_title":"Efficient Rare Event Sampling with Unsupervised Normalising Flows","cited_arxiv_id":"2401.01072","evidence_quote":"Presents an earlier TPS–forward-flux hybrid for active particles and motivates the need for a backward-shooting alternative."},{"cited_title":"Transition path sam- pling for non-equilibrium dynamics without predefined reaction coordinates,","cited_arxiv_id":null,"evidence_quote":"Demonstrates that any backward dynamics with a well-defined path probability can be used for active Brownian particles; the RTP acceptance rule follows that template."},{"cited_title":"Steady state of an active Brownian particle in a two-dimensional harmonic trap,","cited_arxiv_id":null,"evidence_quote":"Establishes the ABP target-search results and parameter choices used here for comparing transition-path-time behavior."},{"cited_title":"Transition path theory from biased simulations,","cited_arxiv_id":null,"evidence_quote":"Justifies treating self-propulsion as odd and the potential force as even under time reversal when choosing the backward dynamics."},{"cited_title":"Optimal navigation strategy of active Brownian par- ticles in target-search problems,","cited_arxiv_id":null,"evidence_quote":"Documents the lack of a general analytic steady-state distribution for active particles, motivating the numerical estimate of $\\rho(\\omega_0)$."}],"review_version":1}