{"id":"923253c6-b980-4f86-ad7e-714aa2295ef1","arxiv_id":"2504.15405","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"By promoting the internal CheY-P concentration to a slow macroscopic field, the authors derive memory-aware reaction-diffusion equations for chemotaxis that predict nonlocal smoothing of the bacterial density over lengths up to 170 micrometers in E. coli.","lead":"A kinetic-theory derivation replaces the standard Keller-Segel equations for bacterial chemotaxis with equations that account for E. coli's internal memory, predicting nonlocal responses over roughly 170 micrometers. This matters because microfluidic chips, soils, and marine nutrient patches often have chemical patterns on exactly that scale, where the old local equations become inaccurate.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Transport coefficients µ12 and µ22 diverge at finite memory times, and the paper does not show that the E. coli regime avoids these poles; the central two-field closure is therefore not established across the claimed 'appreciable memory' range.","rationale":"The reader’s weakest_assumption already identified the single-slow-variable reduction and the divergences of µ12 and µ22 as the main caveat. My stress-test agrees and sharpens it: the divergence is not merely a numerical curiosity but a direct failure of the slaving condition that the Chapman–Enskog method requires. Because the authors themselves state that the divergence indicates a lack of full separation of kinetic modes, the burden is on them to map the region of parameter space where Eqs. (91)–(93) are valid and to show that the E. coli values (λ=1.62, τ=4.2, α1=0.33) and the simulation values (λ=1, τ=1, α1=0) lie safely inside it. The present text reports the divergence but does not locate it for the parameters used in the quantitative claims, so the central claim is not established for the full 'appreciable memory' range it advertises. This does not require rejection: the derivation is careful, the step-profile simulations provide independent support in a specific regime, and the divergence may well lie far from the E. coli point. The appropriate disposition is conditional acceptance, exactly as the reader recommended, with the condition being a concrete domain-of-validity analysis and, where possible, a simulation check near a pole. I therefore leave the verdict unchanged.","tokens_in":36505,"tokens_out":10454,"duration_ms":103511,"concrete_test":"Using the numerical n=10 (and n=20) solution of Eqs. (80), compute the smallest positive τ at which the linear system for the coefficients R0, R1, R2 (and the full six-coefficient set) becomes singular, for the parameter triples (λ=1.62, α1=0.33), (λ=1.0, α1=0), and (λ=3.0, α1=0). If any pole falls within τ∈[0.5,10] for the E. coli parameters, run agent-based simulations at a τ adjacent to that pole and compare the step-profile density and tumble-rate predictions of Eqs. (91)–(93) with the simulation results; agreement would indicate the singularity is a removable Hermite-truncation artifact, while disagreement would demonstrate that the two-field closure breaks down inside the claimed 'appreciable memory' regime.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that Eqs. (91)–(93) replace Keller–Segel whenever the chemotactic memory time is appreciable. That claim requires the two-field Chapman–Enskog closure to be well defined in that regime. The paper shows otherwise: in Section IV.E, Figure 3, µ12 and µ22 diverge at specific values of τ, and the authors state that this divergence 'cannot be physical' and attribute it to 'a lack of full separation of the kinetic modes, a necessary condition for the Chapman–Enskog method.' This is an explicit admission that the necessary condition for the derivation is not met everywhere in the claimed regime. The divergence arises from resonant coupling between the retained mode (u1) and higher Hermite modes through the non-diagonal matrix c in Eqs. (80); it is not a harmless artifact of a particular parameter choice, since the text says it occurs for all α1. The paper never reports the locations of these poles for the E. coli parameters used in Section VI (λ=1.62, α1=0.33, τ=4.2) or for the simulation parameters in Figures 5–6 (λ=1.0, α1=0, τ=1.0). Without this information, the quantitative results—including the headline smoothing length L0=170 µm and the transport coefficients of Section VI—are not proven to lie in the convergent domain. The agent-based validation does help, but it is performed at λ=1, τ=1, which is not the E. coli operating point, and it does not probe the parameter region near a pole. Thus the paper’s principal claim is conditional on a domain-of-validity statement that is currently missing.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper derives macroscopic transport equations for run-and-tumble bacteria whose tumbling rate is controlled by an internal variable, the CheY-P concentration, with a finite memory time. Starting from a kinetic equation for the distribution f(r,n,X,t), the authors apply the Chapman-Enskog method in two regimes: for short memory they recover the Keller-Segel equation with explicit expressions for the diffusion coefficient and chemotactic mobility, and for long memory they treat the bacterial density rho and the CheY-P density rho_X as the two slow fields, obtaining the reaction-diffusion system in Eqs. (91)-(93). For the linear model with E. coli parameters the long-memory equations predict a nonlocal linear response with a Lorentzian smoothing kernel and a smoothing length L0=170 microns, and a nonmonotonic chemotactic current versus traveling-wave speed that is qualitatively consistent with the experiment in Ref. [72]. The paper also reports agent-based simulations for a step-like ligand profile that agree well with the density and tumbling-rate predictions of the two-field equations.","tokens_in":36899,"tokens_out":7147,"duration_ms":69377,"significance":"If the derivation is valid in the claimed regime, this is a substantial contribution: it provides a practical macroscopic description of chemotaxis with memory, reduces to Keller-Segel in the short-memory limit, gives transport coefficients entirely in terms of microscopics, and produces falsifiable predictions such as the 170 micron smoothing length and the traveling-wave current maximum. The systematic Chapman-Enskog procedure, the Onsager-type symmetry D12=D21, and the direct comparison with agent-based simulations are genuine strengths. The principal quantitative claims, however, rest on the domain of validity of the two-field closure, which the manuscript does not fully delimit because the transport coefficients mu12 and mu22 diverge at finite memory times.","major_comments":[{"comment":"The text states that mu12 and mu22 diverge for specific values of tau that depend on lambda and for all alpha1, and it attributes this to 'a lack of full separation of the kinetic modes, a necessary condition for the Chapman-Enskog method.' Because Eqs. (91)-(93) are claimed to replace Keller-Segel whenever the chemotactic memory time is appreciable, the closure must be well defined in exactly that regime. The manuscript never reports the pole locations in the (lambda, alpha1, tau) space. Please provide these pole locations, mark the E. coli operating point (lambda=1.62, alpha1=0.33, tau=4.2) and the simulation point (lambda=1.0, alpha1=0, tau=1.0), and show explicitly that the Section VI transport coefficients and L0 are evaluated away from these singularities.","section":"Section IV.E, Fig. 3"},{"comment":"The agent-based validation is performed only at lambda=1.0, alpha1=0, tau=1.0, which is not the E. coli point and does not approach the region near the mu12/mu22 poles. The good agreement at this parameter set supports the two-field closure in that neighborhood but does not validate the E. coli parameters used for Figure 7 and Section VI. Please add agent-based validation or a sensitivity analysis at the E. coli parameters and at a point near the nearest pole, or explain why the linear-response predictions are insensitive to the pole structure.","section":"Section V.C, Figs. 5-6"},{"comment":"The numerical values for E. coli, including L0 and the transport coefficients, come from the numerical solution of Eqs. (80) with a Hermite truncation at n=10, but no convergence check with respect to the truncation order is shown. Given the near-resonant structure that produces the mu12 and mu22 divergences, the truncation error could be significant even where the exact system is finite. Please report the dependence of D11, D12, D22, mu11, mu12, mu21, mu22, and L0 on the Hermite truncation order at the E. coli parameters, and state the accuracy of the reported values.","section":"Section VI and Eqs. (80)"}],"minor_comments":[{"comment":"The phrase 'Lorentz-type equation' should be replaced by a more standard descriptor such as 'linear Boltzmann-type collision operator' to avoid confusion with the Lorentz model of electrons.","section":"Eq. (9)"},{"comment":"There is a typo in 'Chey-P' in the introduction; it should be 'CheY-P' consistently.","section":"Section II.A"},{"comment":"The phrase 'the election of the signs' should read 'the choice of signs'.","section":"Section IV.E"},{"comment":"The caption does not state the normalization of the plotted current or the fact that the Keller-Segel curve is divided by 5; please add explicit axis labels and normalization details.","section":"Figure 7"},{"comment":"The expression for <X> in the step-profile case appears to mix orders in l1 within the denominator; please check the linearization and sign conventions, since this expression feeds into the tumbling-rate prediction Eq. (106a).","section":"Eq. (107)"},{"comment":"A notation table for the many transport coefficients (D11, D12, D21, D22, mu11, mu12, mu21, mu22, g1...g8) and their sign conventions would substantially improve readability, given the frequent sign choices in Eqs. (83)-(93).","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a strong derivation paper with a clear central claim and credible agent-based support in part of the parameter space. The main risk is the unquantified domain of validity of the two-field closure, specifically the mu12/mu22 poles and whether the E. coli parameters lie safely away from them. I believe the requested pole-location analysis, truncation-order checks, and additional validation at the E. coli point are feasible and would make the paper acceptable. I have no concerns about citation practice beyond normal reliance on the authors' prior work."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, the good news: this is a serious, systematic derivation. The authors take a kinetic equation for run-and-tumble bacteria with an internal CheY-P variable and apply the Chapman-Enskog method keeping the CheY-P density as a quasi-conserved field. The resulting closure, Eqs. (91) to (93), is new as far as I can tell; earlier work either used uncontrolled approximations or recovered the Keller-Segel limit. The equations reduce to Keller-Segel for small memory, give transport coefficients from microscopic parameters, and produce a nonlocal response with a smoothing length of about 170 micrometers for E. coli. The step-profile predictions match agent-based simulations, and the model captures the experimentally observed maximum in the chemotactic current versus traveling-wave speed. That is a real contribution. The soft spot is the one the stress-test flags. The transport coefficients mu12 and mu22 diverge at finite memory times. The paper itself says this cannot be physical and attributes it to a lack of full separation of kinetic modes, which is a necessary condition for the Chapman-Enskog method. That means the closure is not valid across the entire 'appreciable memory' regime claimed. The authors report finite values for the E. coli parameters, which suggests the operating point is not at a pole, but they never state where the poles are in parameter space. Without that, the quantitative headline numbers, including L0 = 170 micrometers, are not proven to be in the convergent domain. The agent-based simulations at lambda=1, tau=1 help, but they are not at the E. coli operating point and do not probe the divergent region. A second limitation is the single-variable linear model for CheY-P dynamics. The authors acknowledge that real fluctuations are not small and that higher kinetic modes can couple. This narrows the biological realism but is an honest limitation, not a hidden one. Who is this for? People working on bacterial chemotaxis, microfluidics, and coarse-grained descriptions of active matter with internal states. It deserves a serious referee. The referee should ask for a domain-of-validity analysis: where the poles are in (lambda, alpha1, tau), and a simulation test closer to the biological parameters. With that, I would support publication.","headline":"A genuinely new and mostly careful memory-aware closure for bacterial chemotaxis, but the divergent transport coefficients leave the closure's domain of validity under-specified.","tokens_in":37420,"tokens_out":3978,"would_cite":true,"duration_ms":34188,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["82C40","92C17","35K57"],"pacs":[],"model":"deepseek-v4-flash","headline":"Slow methylation makes bacterial chemotaxis nonlocal over 170 micrometres, requiring a new two-field macroscopic model.","keywords":["bacterial chemotaxis","memory effects","Keller-Segel equations","Chapman-Enskog","CheY-P","nonlocal response","run-and-tumble","reaction-diffusion"],"falsifier":"Measure the steady bacterial density profile of E. coli near a sharp chemoattractant step in a microfluidic channel and compare the width of the rounded accumulation with the predicted $L_0\\approx170$ $\\mu$m; a discontinuous Keller-Segel-like profile, or a width that does not grow with the methylation time $\\tau$, would falsify the nonlocal response. A second decisive check is the chemotactic current versus traveling-wave speed: the model predicts a maximum as a function of wave speed, whereas Keller-Segel predicts monotonic growth.","tokens_in":36282,"feed_emoji":"🦠","tokens_out":5449,"duration_ms":49894,"temperature":0.7,"pith_summary":"This paper claims that the standard Keller-Segel equations for bacterial chemotaxis break down when the chemotactic memory time is appreciable, and it derives the replacement macroscopic equations from a kinetic model. Treating the internal CheY-P protein concentration as a second slowly evolving field alongside bacterial density, the authors obtain coupled reaction-diffusion equations whose transport coefficients are computed from microscopic parameters. Applied to E. coli, the equations predict a nonlocal response to chemical signals with a smoothing length of roughly 170 micrometres, and a chemotactic current that peaks at a finite traveling-wave speed, matching experiment. The central point is that memory introduces an intrinsic length scale that local gradient-flux laws cannot capture.","feed_headline":"Memory rewrites the equations of bacterial chemotaxis","feed_subtitle":"E. coli's slow methylation adds a second field and a 170-micrometre smoothing length to Keller-Segel theory","key_machinery":"The engine is the kinetic equation (7) for $f(r,\\hat{n},X,t)$, where $X$ is a normalized internal variable (the CheY-P protein concentration in the linear model) evolving by a Fokker-Planck operator with a single long memory time $\\tau$, coupled to run-and-tumble reorientations. The Chapman-Enskog method is applied with $\\rho$ and $\\rho_X$ as the slow fields, treating $\\rho_X$ as a quasi-conserved field in analogy with granular temperature; the eigenfunctions of the Fokker-Planck operator provide the basis that converts the kinetic equation into algebraic equations for the transport coefficients $D_{ij}$ and $\\mu_{ij}$. The central output is a two-field reaction-diffusion system whose linear response contains a Lorentzian kernel, with smoothing length $L_0 = V\\tau\\sqrt{1+(1-\\alpha_1)(1+\\lambda^2)\\hat{\\tau}e^{\\lambda^2/2}}$, which sets the nonlocal scale absent from Keller-Segel.","core_discovery":"The central claim is that Eqs. (91) and (92) with the constitutive relations Eq. (93), derived by a two-field Chapman-Enskog expansion of the kinetic equation (7), are the macroscopic equations that replace the Keller-Segel equations when the chemotactic memory time is appreciable. The density $\\rho$ and the CheY-P density $\\rho_X$ evolve through coupled reaction-diffusion equations; the fluxes contain cross-diffusion terms, chemotactic mobilities, and the source term in the $\\rho_X$ equation includes an explicit ligand time-derivative coupling. For E. coli parameters, the static density response is a Lorentzian in wavevector with smoothing length $L_0 = 1.7\\times 10^2$ $\\mu$m, so a step in ligand concentration produces a rounded density profile over about 170 micrometres instead of the Keller-Segel discontinuity. The model also yields a maximum in the chemotactic current as a function of traveling-wave speed, a feature absent in Keller-Segel and observed in experiments. In the short-memory limit the equations reduce to the Keller-Segel equations with explicit expressions for the transport coefficients.","pith_inferences":["If the two-field equations are correct, a sharp chemoattractant step should produce a density profile whose edge width grows with the methylation time $\\tau$; this could be tested directly in microfluidic experiments by comparing the fitted width to $L_0$.","The divergences of $\\mu_{12}$ and $\\mu_{22}$ at specific memory times suggest that the two-field closure is incomplete for some parameter ranges; including higher CheY-P modes $u_2,u_3,\\dots$ might cure the divergence and extend the description to larger sensitivities $\\lambda$.","A transiently enhanced effective mobility after strong-gradient exposure is a prediction that could be probed with pulse-like ligand histories, connecting the memory mechanism to navigation in time-varying chemical fields.","The same two-field Chapman-Enskog structure could apply to other biological systems with one slow internal adaptation variable, such as eukaryotic chemotaxis or other run-and-tumble organisms, whenever that variable is not strictly conserved."],"forward_implications":["For E. coli parameters, the predicted smoothing length of about 170 micrometres makes memory relevant in microfluidic devices and porous media, and density profiles near chemical steps are rounded rather than discontinuous.","The chemotactic current as a function of traveling-wave speed is nonmonotonic, with a maximum at wave speeds of order 10 micrometres per second, a feature that Keller-Segel lacks and that experiments on E. coli observe.","In the absence of a ligand signal, the two-field system predicts scale-dependent diffusion: at wavelengths shorter than about $1.5\\times10^3$ micrometres the diffusion coefficient is reduced, slowing relaxation of density fluctuations.","The transport coefficients satisfy symmetries, including $D_{12}=D_{21}$ and, in the linear model, $\\mu_{21}=\\mu_{11}D_{22}/D_{12}$, giving testable relations among measurable transport coefficients.","When the memory time is short, the Chapman-Enskog expansion recovers the Keller-Segel equations and gives explicit microscopic expressions for the diffusion coefficient and chemotactic mobility, including the memoryless limit $\\mu = bV^2\\lambda/d$."],"supporting_citations":[{"why":"Supplies the run-and-tumble description of E. coli motion and the angular kernel parameter $\\alpha_1\\approx0.33$.","marker":"[1]"},{"why":"Supplies the two-state motor model that gives the exponential tumbling-rate dependence $\\nu=\\nu_0 e^{\\lambda X}$.","marker":"[4]"},{"why":"Defines the Keller-Segel model that this paper claims to replace in the long-memory regime.","marker":"[6]"},{"why":"Supplies the fitted E. coli parameters $V=27$ $\\mu$m/s, $\\nu_0=0.22$ s$^{-1}$, $\\tau=19$ s, and $\\lambda=1.62$ used for all numerical values.","marker":"[22]"},{"why":"Provides the earlier kinetic treatment of the same linear model whose stationary mobility and response are extended here to hydrodynamic equations.","marker":"[24]"},{"why":"Supplies the Fokker-Planck spectral properties and eigenfunction basis used to diagonalize the internal-variable dynamics.","marker":"[55]"},{"why":"Supplies the Chapman-Enskog expansion method that is the derivation's formal backbone.","marker":"[59]"},{"why":"Provides the experimental traveling-wave current maximum that the derived model reproduces and Keller-Segel does not.","marker":"[72]"}],"fun_headline_variants":["Bacterial chemotaxis gets a memory upgrade","Memory shifts bacterial chemotaxis beyond Keller-Segel","E. coli chemotaxis: memory matters at 170 microns","Two-field equations beat Keller-Segel for chemotaxis","Memory-driven smoothing in bacterial chemotaxis"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The derivation rests on assuming that the chemotactic pathway can be reduced to a single internal variable $X$ with one long memory time $\\tau$, with all other internal modes fast enough to be slaved; if CheY-P dynamics require additional slow variables or the memory spectrum is not cleanly separated, the two-field equations and the 170-micrometre smoothing length would need revision.","fun_headline_variants_meta":{"raw":{"variants":["Bacterial chemotaxis gets a memory upgrade","Memory shifts bacterial chemotaxis beyond Keller-Segel","E. coli chemotaxis: memory matters at 170 microns","Two-field equations beat Keller-Segel for chemotaxis","Memory-driven smoothing in bacterial chemotaxis"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000225,"raw_usage":{"total_tokens":1551,"prompt_tokens":1119,"completion_tokens":432,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":735,"completion_tokens_details":{"reasoning_tokens":353}},"tokens_in":735,"tokens_out":432,"duration_ms":3654,"temperature":1.0,"reasoning_tokens":353,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T11:27:18.174343+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the steady bacterial density profile of E. coli near a sharp chemoattractant step in a microfluidic channel and compare the width of the rounded accumulation with the predicted $L_0\\approx170$ $\\mu$m; a discontinuous Keller-Segel-like profile, or a width that does not grow with the methylation time $\\tau$, would falsify the nonlocal response. A second decisive check is the chemotactic current versus traveling-wave speed: the model predicts a maximum as a function of wave speed, whereas Keller-Segel predicts monotonic growth.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the run-and-tumble description of E. coli motion and the angular kernel parameter $\\alpha_1\\approx0.33$."},{"cited_title":"Korobkova, T","cited_arxiv_id":null,"evidence_quote":"Defines the Keller-Segel model that this paper claims to replace in the long-memory regime."},{"cited_title":"Ito and T","cited_arxiv_id":null,"evidence_quote":"Supplies the fitted E. coli parameters $V=27$ $\\mu$m/s, $\\nu_0=0.22$ s$^{-1}$, $\\tau=19$ s, and $\\lambda=1.62$ used for all numerical values."},{"cited_title":"Zhang, R","cited_arxiv_id":null,"evidence_quote":"Provides the earlier kinetic treatment of the same linear model whose stationary mobility and response are extended here to hydrodynamic equations."},{"cited_title":"Saragosti, P","cited_arxiv_id":null,"evidence_quote":"Supplies the Fokker-Planck spectral properties and eigenfunction basis used to diagonalize the internal-variable dynamics."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Chapman-Enskog expansion method that is the derivation's formal backbone."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the experimental traveling-wave current maximum that the derived model reproduces and Keller-Segel does not."}],"review_version":1}