{"id":"9045cb93-404f-421c-bf54-295ae673c10e","arxiv_id":"2608.05251","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A non-equilibrium nucleation theory for vector reaction-diffusion systems shows front speed and front diffusivity play the roles of free-energy difference and surface tension, making invasion exponentially harder as interspecific competition strengthens.","lead":"The authors derive a nucleation theory for spatially extended bistable systems whose state is described by several coupled species abundances, not by a single scalar field, and which do not obey the energy-based rules of equilibrium. They show invasion probability is controlled by properties of propagating fronts, and predict that strongly competing species are exponentially harder to invade at equal competitive advantage.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central result inherits an unproven spectral-gap assumption for vector fronts; if the gap closes in any regime supporting Eq. 90, the effective-interface derivation of Eq. 34 and the quasipotential Eq. 45 fails.","rationale":"I re-read Secs. II B, IV A, and VI B. The central claim is not an empirical fit: it follows from a singular-perturbation reduction of Eq. (1) to the stochastic front equation (34), and then to the 1D droplet Langevin equation (40) whose stationary measure gives the quasipotential (45). The only step in that chain that is not derived from the structure of Eq. (1) is the assertion that vectorial bistable fronts are unique, linearly stable, and gapped. Everything downstream—the solvability condition (33), the projection formulas (35)-(36), the absence of resonant contamination of δu, and ultimately U_c ~ D^d/(g c^{d-1})—requires that L_b has an isolated zero eigenvalue and a spectral gap. The paper's own Sec. VIII admits that when fronts become oscillatory or chaotic the coarse-graining fails, confirming that the gap is a genuine condition rather than a formality.\n\nThe numerical evidence offered for this condition is thin: Fig. 9 shows one binodal front at a=b=2 and its first two eigenvalues. It verifies that Λ1=0 and Λ2<0 for that point, and that the result is robust to domain size/discretization. It does not cover the regimes where the paper makes its strongest quantitative claim (Eq. 90), particularly large a+b, where the front has two well-separated length scales, or the near-neutral limit where the front width diverges and the gap must close. A direct spectral sweep across the parameter range used in Fig. 20 would settle whether the singular-perturbation reduction is valid where the prediction is made.\n\nSecondary weaknesses (missing code DOI, fitted offset U in Figs. 16 and 19) do not touch the central derivation as directly and were already noted by the reader. I therefore agree with the reader's weakest-assumption identification and do not see a reason to change the CONDITIONAL verdict.","tokens_in":40803,"tokens_out":17953,"duration_ms":192487,"concrete_test":"Systematically compute the spectrum of L in Eq. (79) for LV fronts along and near the binodal over the parameter range used in Figs. 14 and 20 (e.g., (a+b)/2 = 1.05, 1.1, 1.25, 1.5, 1.75, 2, 3, 5, 10, 20, 50), varying domain length and grid spacing independently. Record the second eigenvalue Λ2 and the bottom of the essential spectrum. If Λ2 (or the essential spectrum) approaches zero in any regime supporting Eq. (90), the singular-perturbation projection used to derive Eq. (34) is not valid there; if the gap stays bounded away from zero across the whole range, the spectral condition underlying the central claim is verified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim that CNT survives out of equilibrium is derived by linearizing around a flat front and projecting Eq. (27) onto the kernel of L_b^† (Sec. IV A 2). This singular-perturbation step is valid only if the rest of the spectrum of L_b is gapped away from zero: otherwise the source terms Δf and Dκ excite continuum or additional zero modes, δu cannot be held to O(ε), the solvability condition Eq. (33) does not select a slow front coordinate, and Eq. (34) is not the correct reduced dynamics. The quasipotential Eq. (45) and critical-nucleus barrier Eq. (47), and hence the ecological prediction Eq. (90), stand or fall with this reduction.\n\nThe paper states in Sec. IV A 1 that uniqueness, linear stability, and the gap are 'taken for granted... as in the scalar bistable case,' and cites only numerical simulation of the LV model. The direct spectral evidence in Sec. VI B is a single binodal example (a=b=2, Fig. 9), with the statement that the gap is not a numerical artifact. This does not probe the parameter range used for the headline prediction, in particular the strong-competition regime where the front develops a two-scale structure (A on scale w, ξ on scale sqrt(D/(a+b))) and the near-neutral regime where the front width diverges. In the near-neutral limit the gap presumably closes as 1/w^2, so the reduction is not uniformly valid even on the binodal. Since the theory is advertised for a broad class of non-gradient vector systems, the missing proof or systematic verification of the gap is the single most load-bearing gap in the argument. The paper itself flags in Sec. VIII that fronts cease to be rigid, gapped objects in more complex dynamics, confirming that this is a condition, not a demonstrated fact.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a nucleation theory for stochastic reaction-diffusion systems with a non-conserved vector order parameter, without detailed balance. Near the binodal, the authors coarse-grain the field dynamics onto a stochastic interface equation for front motion, Eq. (34), and derive a quasipotential U(R) = (S_d/g) R^{d-1}(D - cR/d) whose structure mirrors classical nucleation theory: the front speed c plays the role of the free-energy difference and the diffusivity D that of the surface tension. Near the spinodal, they reduce the escape problem to one soft eigenmode and obtain a universal scaling U_c ~ ε^{(6-d)/2}. The theory is tested against shrinking-dimer and GMAM computations of the quasipotential in the two-species Lotka-Volterra competition model, and is used to obtain the ecological prediction ln p(R0) ~ -(a+b-2)/(a-b) for the invasion probability of a small inoculum in two dimensions.","tokens_in":41142,"tokens_out":7486,"duration_ms":83210,"significance":"If the results hold, this is an important extension of nucleation theory to non-equilibrium, multi-component systems, with a concrete and testable ecological prediction. The paper is also notable for what it does not hide: the central singular-perturbation derivation is presented in detail, the numerical methods are extensive and the code is promised to be available, and the recovery of classical CNT from the general formulas in Sec. IV A 4 is a genuine check of internal consistency. The spinodal calculation is parameter-free and agrees with numerical quasipotentials over a wide range, which is a substantive strength. The main limitations are the unproven spectral-gap assumption for vectorial fronts, the fitted additive constant in the binodal quasipotential, and the unproven claim that the spinodal action is exactly the leading-order quasipotential rather than an upper bound. These issues do not invalidate the central idea but they do mean that the paper's strongest claims are not yet established with full rigour.","major_comments":[{"comment":"The derivation of the reduced interface equation (34) and hence of the quasipotential (45) relies on the operator L_b having an isolated zero eigenvalue with a spectral gap. In Sec. IV A 1 the authors explicitly state that uniqueness, linear stability and the gap are 'taken for granted' and supported only by numerical simulation of the Lotka-Volterra model; the sole spectral evidence presented is the binodal example a=b=2 in Fig. 9. This does not cover the parameter regimes used for the headline prediction, in particular the near-neutral limit a,b→1, where the front width diverges as (a+b-2)^{-1/2} (Eq. B4) and the gap presumably closes as 1/w^2, and the strong-competition limit where the front develops a two-scale structure (A varying on scale √D, ξ on scale √(D/(a+b))). If the gap closes or additional soft modes appear, the projection onto the single zero mode is not the correct reduced dynamics, and Eqs. (34), (45) and (90) do not follow. The authors should either provide a proof or a systematic numerical verification of the gap and of the absence of other near-zero modes across the (a,b) range used, or restrict the claimed domain of validity.","section":"IV A 1 and IV A 2"},{"comment":"The binodal quasipotential comparison in Fig. 16 uses Eq. (72), U_c = K/ε^{d-1} + U, with K computed from the front profile independently of the GMAM data, but the constant U is fitted to the GMAM data (U = -0.60), and the demographic-noise offset U'(R0) in Eq. (88) is also fitted. The paper explicitly acknowledges in Sec. IV C 1 that the front-creation contribution has not been computed analytically. Thus the numerical validation establishes the predicted ε-dependence with an adjustable additive constant, not the absolute value of the nucleation barrier. For Eq. (90), the exponent is set by the parameter-free coefficient Λ, so the exponential suppression with (a+b-2)/(a-b) is on solid ground, but the absolute invasion probability retains a fitted offset per noise model and inoculum radius. The text should state precisely which predictions are parameter-free and which require the fitted U or U'(R0); ideally the additive constant should be derived or at least bounded.","section":"IV C 1 and VII B 1"},{"comment":"The derivation of the spinodal quasipotential (70) claims that restricting the action to trial paths u = u*_+ + δu_1 e_1 gives the leading-order exact value of the quasipotential, 'not an upper bound', because excursions along stiff directions yield sub-leading contributions. However, no estimate of these contributions is provided, and the non-equilibrium nature of the dynamics means that the minimizer of the Freidlin-Wentzell action could in principle mix modes and produce a lower action than the single-mode path. The numerical agreement in Fig. 16 is encouraging, but the analytical claim is stronger than what is proved. The authors should either supply the action estimate for the transverse modes or soften the statement to a variational upper bound that is numerically observed to be tight.","section":"IV B 3"}],"minor_comments":[{"comment":"There is a typo: 'equilbrium' should be 'equilibrium' in the sentence 'Since the front dynamics are the same for equilbrium and non-equilibrium field dynamics'.","section":"IV A 4"},{"comment":"The symbol T is used both for the noise amplitude (Section II) and for the rescaled time in Eq. (65) and the surrounding text; this overloaded notation is confusing and should be changed, for instance to τ.","section":"IV B 3"},{"comment":"The front width w is defined in words as the distance between the positions where v_fr = 0.9 and u_fr = 0.9; a concise formula or a clear reference point (e.g., the center x=0) would improve reproducibility.","section":"VI C"},{"comment":"The data availability statement contains a placeholder URL (https://zenodo.org/XXX); a working link should be provided before publication.","section":"IX"}],"recommendation":"major_revision","confidential_remarks":"The paper contains a promising and substantial framework, and the numerical evidence is extensive, but the main derivation rests on an explicitly unproven spectral-gap assumption for vectorial fronts, and the binodal validation uses a fitted constant. These are fixable within the manuscript's scope by adding a spectral verification across the parameter range and by a more precise statement of what is derived versus fitted. I would not reject the paper, but I would require these points to be addressed before publication. The authors should also be encouraged to make the claimed novelty relative to the concurrent scalar works (Refs. 31-32) more explicit, since both the solvability-condition derivation and the vector treatment are the distinctive contributions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The central result is real: for a broad class of non-conserved vector reaction-diffusion systems, the paper derives an effective interface motion whose form is identical to equilibrium, with front speed c and diffusivity D replacing bulk free-energy difference and surface tension. The solvability condition deriving the interface position is a genuine improvement over concurrent scalar treatments, and the vector structure matters: the depletion region in Lotka-Volterra fronts, invisible to scalar frequency descriptions, controls the speed and the nucleation barrier. The scaling ln p ~ -(a+b-2)/(a-b) is new, concrete, and testable in microbial systems.\n\nThe numerical validation is extensive and honest: GMAM on large grids, shrinking dimer, and analytical asymptotics in the neutral and strong-competition limits that match the numerics. The agreement between the front-based theory and the quasipotential near the binodal is good, and the spinodal prediction is also confirmed along the tested path.\n\nThe soft spots are real but not fatal. The spectral gap of the linearized front operator is load-bearing for the reduction, and it is 'taken for granted' with only one binodal numerical check. The stress-test concern is fair: near neutrality the front width diverges and the gap presumably closes, so the reduction is not uniformly valid across the whole binodal. The strong-competition regime where the headline ecological prediction lives has a finite gap, so the main prediction is less exposed, but the generality claim needs either a proof or a systematic numerical check of the gap across the parameter range. The constant U accounting for front creation is fitted rather than derived, as the paper admits; this affects absolute rates but not the leading scaling. The missing code artifact (placeholder DOI) should be fixed before publication. Minor point: the prefactor discussion is brief but appropriate.\n\nOverall the paper is a solid piece of theory with a concrete prediction. I'd send it to a serious referee: the derivation deserves close scrutiny, and the spectral-gap assumption should be pushed. It is not a desk reject.","headline":"A substantial extension of CNT to vector non-conserved fields via front speed and diffusivity; the spectral-gap assumption and fitted offsets are the main reservations, but the core result is credible and worth refereeing.","tokens_in":41724,"tokens_out":1884,"would_cite":true,"duration_ms":20755,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35K57","60H15","82C31","92D25"],"pacs":["05.40.-a","64.60.Q-","87.23.Cc"],"model":"deepseek-v4-flash","headline":"The paper derives a nucleation theory for bistable reaction-diffusion systems with vector, non-conserved order parameters that holds out of equilibrium, showing that classical nucleation formulas survive once free-energy differences and…","keywords":["nucleation","bistability","non-equilibrium","Lotka-Volterra","reaction-diffusion","fronts","quasipotential","classical nucleation theory"],"falsifier":"Numerically diagonalize the linearized front operator $\\mathcal{L}$ around a flat front in a vector reaction-diffusion system with a non-gradient reaction term; if the gap $\\Lambda_2 - \\Lambda_1$ between the translational mode and the next eigenmode closes as parameters approach the spinodal or strong competition, the front is not gapped and the solvability projection that yields $c_{\\mathrm{tot}} = c - D\\kappa + \\sqrt{2Tg}\\,\\xi$ breaks down.","tokens_in":40602,"feed_emoji":"🦠","tokens_out":5980,"duration_ms":54291,"temperature":0.7,"pith_summary":"The paper aims to answer a question no general theory has answered: how does a fluctuation-driven droplet of one stable state invade another in a bistable system whose dynamics violate detailed balance and whose order parameter is a vector of coupled abundances? It derives, from the stochastic reaction-diffusion dynamics, the quasipotential governing nucleation in two limits, near the binodal where invasion proceeds by fronts, and near the spinodal where the metastable state is about to lose stability. The central claim is that the mathematical structure of Classical Nucleation Theory survives: the front speed $c$ takes the role of the bulk free-energy difference $\\Delta\\phi$, the front's curvature response $D$ takes the role of the surface tension $\\gamma$, and the front mobility $g$ enters the activation barrier. The payoff is a concrete prediction for the two-species Lotka-Volterra model: at fixed competitive advantage, stronger interspecific competition makes invasion exponentially harder, because fronts develop a depletion region invisible to scalar descriptions.","feed_headline":"Front speed replaces energy to predict invasion odds","feed_subtitle":"Two dynamical numbers from an invasion front set the critical radius and the probability an inoculum takes over.","key_machinery":"The load-bearing object is the flat front $u_{\\mathrm{fr}}$ together with its linearization $\\mathcal{L} = D\\partial_x^2 + J(u_{\\mathrm{fr}})$ about that front. Translational invariance makes $\\mathcal{L}$ singular, with a zero mode $-\\partial_x u_{\\mathrm{fr}}$; the paper fixes the front's frame of reference by imposing the solvability condition $\\langle \\delta u_0, \\delta u \\rangle = 0$ with $\\delta u_0$ in the kernel of the adjoint $\\mathcal{L}^\\dagger$. Projecting the perturbed field equation onto $\\delta u_0$ eliminates the internal deformation $\\delta u$ and yields the effective interface equation, with $c$ coming from the integrated driving force, $D$ from curvature coupling, and $g$ from the noise overlap. This projection is the step that transfers the vectorial, non-equilibrium structure of the field into three scalar coefficients, and it is the step that reduces nucleation to a one-dimensional barrier crossing.","core_discovery":"The paper establishes a dictionary between equilibrium and non-equilibrium nucleation. For a general class of stochastic reaction-diffusion systems with a non-conserved vector order parameter, it shows that a flat front between the two stable states is coarse-grained into an effective interface whose motion is $c_{\\mathrm{tot}} = c - D\\kappa + \\sqrt{2Tg}\\,\\xi$, where $c$ is the flat-front speed, $D$ its response to curvature, and $g$ the noise mobility. Coarse-graining the droplet radius onto a one-dimensional Langevin equation restores detailed balance for the collective variable $R$, so the invasion probability obeys Arrhenius' law with quasipotential $U(R) = \\frac{S_d}{g} R^{d-1}\\left(D - \\frac{c}{d} R\\right)$, critical radius $R_c = (d-1)D/c$, and barrier $U_c = \\frac{(d-1)^{d-1} S_d D^d}{d\\, g\\, c^{d-1}}$. Close to the spinodal, the same large-deviation calculation yields a universal nucleus shape controlled by a single eigenvector of the Jacobian and a barrier scaling as $\\epsilon^{(d_c-d)/2}$ with $d_c = 6$ for cubic nonlinearities. Applied to the Lotka-Volterra competition model, the theory predicts that fronts carry a depletion region whose width and depth set the front speed and mobility, and that in two dimensions the invasion probability from a small inoculum scales as $\\ln p \\propto -(a+b-2)/(a-b)$.","pith_inferences":["Because the droplet-radius Langevin equation always satisfies one-dimensional detailed balance, the restored time-reversal symmetry is an artifact of coarse-graining the single collective variable; the MAP and relaxation path may still differ in the internal front profile $\\delta u$, which could be probed experimentally by tracking the abundance dip during a nucleation event.","The same front-based dictionary suggests a testable microscopic route to negative effective 'surface tension' $D$ in active or non-reciprocal systems: if the front's curvature response changes sign, stable nuclei rather than unstable critical droplets form, a hallmark recently discussed in conserved active systems and now predicted for non-conserved vector fields.","The prediction $\\ln p \\propto -(a+b-2)/(a-b)$ is directly testable in engineered killer-yeast communities by raising toxin production of both strains symmetrically, which the paper identifies as increasing $a+b$ at fixed $a-b$; measuring invasion probability as a function of $a+b$ would confirm or falsify the exponential suppression.","The spinodal universality, derived via an effective single-eigenvector reduction, should also appear in the statistics of extinction times in well-mixed metacommunities near the loss of stability, a regime where demographic noise cannot create individuals from zero abundance."],"forward_implications":["Invasion probability from an inoculum of radius $R_0$ is $p(R_0) \\propto \\exp[-(U_c - U(R_0))/T]$, so the critical radius and barrier are computable from three front measurements: speed, curvature response, and mobility.","Near the spinodal, the critical nucleus aligns with the soft eigenvector of the Jacobian and the nucleation barrier vanishes with the universal scaling $\\epsilon^{(d_c-d)/2}$, with $d_c = 6$ for cubic terms.","In the Lotka-Volterra model, the front develops a depletion region where $u+v$ falls below carrying capacity; at fixed competitive advantage $a-b$, stronger competition (larger $a+b$) slows fronts, enlarges the critical radius, and suppresses invasion exponentially.","Confinement reduces the critical nucleus to an elongated two-arc shape when domain size drops below twice the unconfined critical diameter, with a sharp limiting size $L_{\\mathrm{lim}}$ below which nucleation is homogeneous.","The theory reduces exactly to classical nucleation theory at equilibrium, recovering the energy barrier $E(R) = S_d R^{d-1}\\left(\\gamma - \\Delta\\phi\\, R/d\\right)$ from the front dictionary."],"supporting_citations":[{"why":"Pomeau's idea that the front speed plays the qualitative role of the free-energy difference out of equilibrium is the seed the paper makes quantitative.","marker":"[33]"},{"why":"Fife's method of linearizing around a flat front and treating the zero mode supplies the singular-perturbation and solvability technique.","marker":"[70]"},{"why":"Freidlin-Wentzell theory provides the large-deviation action and quasipotential that replace the energy landscape out of equilibrium.","marker":"[38]"},{"why":"Langer's field-theoretic nucleation theory is the equilibrium reference whose structure the paper extends.","marker":"[21]"},{"why":"Langer's decay-of-metastable-states theory supplies the escape-problem formulation and the shape-mode prefactor contributions.","marker":"[22]"},{"why":"Giometto, Nelson and Murray's killer-yeast experiments show the depletion region in fronts and provide the empirical anchor for the microbial predictions.","marker":"[14]"},{"why":"Muratov and Vanden-Eijnden's spinodal universality result fixes the upper critical dimension $d_c=6$ used in the spinodal scaling.","marker":"[50]"},{"why":"Cates and Nardini's CNT for active fluid phase separation is the closest prior non-equilibrium nucleation theory, contrasted as the conserved scalar case.","marker":"[23]"},{"why":"Bouchet and Reygner's generalized Eyring-Kramers formula underpins the claim that the MAP crosses the critical nucleus and sets the prefactor discussion.","marker":"[43]"},{"why":"Heymann and Vanden-Eijnden's geometric minimum action method is the numerical algorithm used to compute the quasipotential and validate the theory.","marker":"[93]"}],"fun_headline_variants":["Front speed replaces energy to set invasion odds","Strong competition makes invasion exponentially harder","Non-equilibrium nucleation explained by moving fronts","Invasion probability from front speed and curvature","Depletion in fronts blocks invasion: new law"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole binodal calculation assumes that a vector bistable system has a unique front that is linearly stable and gapped, meaning the translational zero mode is isolated from the rest of the relaxation spectrum; the paper states this is taken for granted, as in the scalar case, and supports it only by numerical simulation of one model.","fun_headline_variants_meta":{"raw":{"variants":["Front speed replaces energy to set invasion odds","Strong competition makes invasion exponentially harder","Non-equilibrium nucleation explained by moving fronts","Invasion probability from front speed and curvature","Depletion in fronts blocks invasion: new law"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000815,"raw_usage":{"total_tokens":3692,"prompt_tokens":1184,"completion_tokens":2508,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":800,"completion_tokens_details":{"reasoning_tokens":2443}},"tokens_in":800,"tokens_out":2508,"duration_ms":19457,"temperature":1.0,"reasoning_tokens":2443,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T17:05:22.930657+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Numerically diagonalize the linearized front operator $\\mathcal{L}$ around a flat front in a vector reaction-diffusion system with a non-gradient reaction term; if the gap $\\Lambda_2 - \\Lambda_1$ between the translational mode and the next eigenmode closes as parameters approach the spinodal or strong competition, the front is not gapped and the solvability projection that yields $c_{\\mathrm{tot}} = c - D\\kappa + \\sqrt{2Tg}\\,\\xi$ breaks down.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Fife's method of linearizing around a flat front and treating the zero mode supplies the singular-perturbation and solvability technique."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Muratov and Vanden-Eijnden's spinodal universality result fixes the upper critical dimension $d_c=6$ used in the spinodal scaling."},{"cited_title":"Tjhung, C","cited_arxiv_id":null,"evidence_quote":"Heymann and Vanden-Eijnden's geometric minimum action method is the numerical algorithm used to compute the quasipotential and validate the theory."}],"review_version":1}