{"id":"ccb3204e-89f0-4a58-a5b8-c7e3067fa4ab","arxiv_id":"2502.04185","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A nonperturbative lattice calculation finds that perturbative bubble nucleation rates in a tree-level barrier scalar theory agree only qualitatively, with |log Γ| off by 20% at one loop and 100% at tree level.","lead":"Lattice simulations of bubble nucleation in a scalar field theory with a tree-level barrier produce a rate that is about 20% lower in |log Γ| than the one-loop perturbative prediction. The result suggests that perturbative calculations used to forecast gravitational wave signals from cosmological phase transitions may carry larger uncertainties than previously assumed.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The real-time evolution step in Eq. (4) is the least-secure link: if the Gaussian-flux assumption or the γ=1/L Langevin dynamics biases the rate, the 20–100% discrepancy is a simulation artifact rather than a failure of perturbation theory.","rationale":"The paper's central claim is that a nonperturbative lattice computation of Γ disagrees with one-loop perturbation theory by 20% in |log Γ| while agreeing on the latent heat to <1%. The logic rests on the identification of Γ with the factorized expression in Eq. (4). The weakest link is the real-time evolution step: the flux is assumed Gaussian, the tunneling fraction is measured with a specific Langevin damping γ=1/L, and the order parameter is a quadratic combination φ^2−2Aφ rather than the linear field. The reader's weakest_assumption identified this same dynamical-model risk. Because the paper is a proceedings that defers full extrapolation details to the companion paper [19], and because the available reweighted points and extrapolation fits are sparse, the dynamical-system bias is the most load-bearing concern. The rest of the paper (multicanonical probability, latent-heat agreement, order-parameter choice) is well supported and the code/data are archived. The suggested check would settle whether the discrepancy is physical or a simulation artifact. Verdict remains CONDITIONAL since the concern does not invalidate the presentation but requires empirical confirmation of the real-time systematics.","tokens_in":7620,"tokens_out":1473,"duration_ms":13778,"concrete_test":"Recompute the rate with two independent variations: (1) repeat the real-time evolution at γ=0 (pure Hamiltonian, no momentum refresh) for the smallest lattices to see whether ⟨d⟩ and the final log Γ shift beyond the quoted 0.05–0.07 errors; (2) repeat the flux measurement with a direct numerical estimator of |Δφ/Δt| on the separatrix configurations, instead of the analytic Gaussian result √(8/π V(θ_c + A²)), and check whether the difference changes |log Γ| by more than the 20% discrepancy claimed against one-loop perturbation theory.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The rate in Eq. (4) is a factorization: P_c × (1/2)⟨flux⟩ × ⟨d⟩. The flux is computed analytically assuming Gaussian momentum and the projector |Δθ_op/Δt|; the tunneling fraction ⟨d⟩ is measured from Langevin trajectories with damping γ=1/L. The dynamical model must give the correct physical rate at leading order, and the paper reports only an empirical statement that a fourth-order symplectic algorithm was needed so that Monte Carlo and real-time stages agree on the separatrix. However, the finite timestep Δt, the momentum-refresh amplitude γ, and the projection |Δθ_op/Δt| (with θ_op = φ^2 − 2Aφ rather than the physical field φ) are all coupled. If the Gaussian-flux approximation or the γ→0 extrapolation is biased, then the 20–100% discrepancy with one-loop/tree-level perturbation theory is a simulation artifact, not evidence that perturbative nucleation calculations fail. The continuum and infinite-volume extrapolations are deferred to the companion paper; in this proceedings the points at different lattice spacings/volumes are limited, so systematics from the real-time step are the least-secure link in the central claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This proceedings paper reports a nonperturbative lattice computation of the bubble nucleation rate in a three-dimensional scalar field theory with a tree-level barrier, which is a toy model relevant for strong cosmological phase transitions. Using multicanonical simulations to sample near-critical configurations and real-time Langevin evolution to determine the tunneling fraction, the authors obtain a continuum-infinite-volume-extrapolated lattice rate that is substantially lower than the tree-level and one-loop perturbative predictions: the disagreement in |log Γ| is quoted as 100% at tree-level and 20% at one-loop, while the latent heat agrees to better than 1%. The paper interprets this as evidence that perturbative nucleation calculations, although expected to work well for tree-level-barrier scenarios, may only be in qualitative agreement with a fully nonperturbative treatment. Technical details, including the O(a^2) improvement and the detailed extrapolations, are deferred to a companion paper [19], and the data and code are made available.","tokens_in":7861,"tokens_out":12918,"duration_ms":125283,"significance":"If the result is correct, it is significant for the electroweak phase transition and gravitational-wave phenomenology, because it challenges the common assumption that perturbative bubble-nucleation rates are quantitatively reliable in strong, tree-level-barrier models. The calculation is a genuine nonperturbative test: the lattice simulation is independent of the perturbative curves, so the comparison is not circular. The paper also ships reproducible data (Zenodo DOI) and open source code (scalnuc), which strengthens the reliability of the reported numbers. The main caveat is that the rate measurement relies on a factorized dynamic prescription whose validation is only partially documented in this proceedings; this is the central issue assessed in the major comments.","major_comments":[{"comment":"The rate in Eq. (4) factorizes as P_c × (1/2)⟨flux⟩ × ⟨d⟩, where the flux is computed analytically assuming Gaussian momenta and ⟨d⟩ is measured from Langevin trajectories with damping γ = 1/L. The manuscript reports only an empirical statement that a high-order symplectic algorithm was needed so that the Monte Carlo and real-time stages agree on the separatrix; it gives no convergence tests in Δt, γ, or trajectory length, and no external validation of the factorized form. Since any bias in the flux or in ⟨d⟩ would propagate directly into Γ, the 20% and 100% discrepancies quoted in Sec. 4 are not fully supported by the evidence presented in this paper. The authors should either provide such tests or explicitly state that the discrepancy is conditional on this dynamical prescription and point to the validation in the companion paper [19].","section":"Sec. 2.1, Eqs. (4)-(12)"},{"comment":"The final lattice value used for the comparison in Fig. 4 is not clearly a simultaneous continuum and infinite-volume limit. The left panel of Fig. 3 shows a continuum extrapolation at fixed volume L λ_3 = 42, while the right panel shows an infinite-volume extrapolation at fixed lattice spacing a λ_3 = 1.5. Consequently, the quoted log(Γ/λ_3^4) ≈ -74.09(5) and the reweighted curves may carry undisclosed discretization or finite-volume systematics. The paper should specify which extrapolated value is used for the discrepancy percentages in Sec. 4 and what systematic error is assigned, or state that the combined continuum and infinite-volume extrapolation is deferred to [19].","section":"Sec. 3, Fig. 3"},{"comment":"The probability density P_c in Eq. (3) depends on the arbitrary order parameter θ_op = φ^2 − 2Aφ and on the window ε, and the flux formula in Eq. (4) also depends on this choice of θ_op. The text asserts that the exact choice of ε is compensated by ⟨d⟩, but no test of invariance of the final product P_c × (1/2)⟨flux⟩ × ⟨d⟩ under changes of θ_op is presented. The single comparison using the linear order parameter φ_lin in Fig. 3 is consistent within errors, but it is not described as a systematic check. A dependence of the final rate on the arbitrary projection would invalidate the method, so the authors should report such a check or refer to a concrete verification in [19].","section":"Eqs. (3) and (4)"}],"minor_comments":[{"comment":"The phrase 'relativistic and quantum theories theories' contains a duplicated word; please correct it.","section":"Sec. 1, Introduction"},{"comment":"The discussion of O(a^2) improvement mentions a parameter κ_lat and κ_MS, but Eq. (2) does not contain a κ parameter. Please rephrase this to refer to the lattice parameters that actually appear in Eq. (2), such as σ_lat, m^2_lat, and λ_lat.","section":"Sec. 1.1, near Eq. (2)"},{"comment":"The abstract states only that the agreement is 'qualitative'; consider adding the quantitative figures from Sec. 4 (20% at one-loop, 100% at tree-level) or a pointer to Sec. 4 so that the magnitude of the discrepancy is clear to the reader.","section":"Abstract"},{"comment":"Please state explicitly that Γ in Fig. 4 is the rate per unit volume, and define the constant A and the normalization convention for θ_op (e.g., volume-averaged) so that the flux formula can be checked dimensionally.","section":"Eq. (4)"},{"comment":"Please state how the initial momenta for the real-time evolution are sampled (for example, drawn from a Gaussian distribution) or provide a reference to the companion paper for this detail.","section":"Sec. 2.1"},{"comment":"The captions say the figures are reproduced from Ref. [19]; consider adding a sentence in the text noting that the lattice data shown here are identical to those in [19] and that the full fit details are given there.","section":"Figs. 3 and 4"}],"recommendation":"major_revision","confidential_remarks":"This is a proceedings contribution that condenses a fuller analysis in the companion paper [19]. The major comments are intended to make the proceedings self-contained enough for the central claim to be assessed; they are addressable by adding explicit statements and references to validations that, if present in [19], would resolve them. The data and code availability are clear strengths. The scope of the manuscript fits the journal, but the editor may wish to confirm that [19] has been reviewed or is in press, since several of the load-bearing details are deferred there."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The short version: this proceedings is the public preview of a companion paper (Ref. [19]) that reports the first nonperturbative lattice test of nucleation rates in a scalar theory with a tree-level barrier. The headline result is a genuine discrepancy: the lattice rate is lower than the one-loop perturbative rate by about 20% in |log Γ| (100% at tree level), with lattice statistical errors much smaller than the perturbative renormalization-scale bands. The latent heat agrees to better than 1%, so the thermodynamics is fine—it's specifically the rate that disagrees.\n\nWhat's actually new and worth taking seriously: the modified order parameter θ_op = φ^2 − 2Aφ is a genuinely useful practical trick for suppressing bulk phase fluctuations when identifying critical bubbles. The comparison is not circular—the lattice is an independent nonperturbative measurement and the perturbative curves are external calculations. The paper is openly a proceedings contribution, so it defers technical details to [19]; it also plainly states the limitations: one benchmark point, one toy model, and the possibility that a two-loop calculation could resolve the gap.\n\nThe soft spot, and the one I'd push on, is the real-time evolution step in Eqs. (4)–(12). The rate is a product of a multicanonical probability, an analytic Gaussian flux, and a tunneling fraction extracted from Langevin trajectories with damping γ=1/L. The stress-test worry—that a bias in the flux or the γ→0 dynamics would make the discrepancy a simulation artifact—is not a fatal objection on the face of the paper, because the real-time step is used only to classify which near-critical configurations actually tunnel, and the authors report an internal consistency check (Monte Carlo and real-time agree on the separatrix location). But that check is not a full validation: there is no direct test of the Gaussian momentum assumption, no γ-dependence scan, and no comparison of the measured flux to the analytic formula. Those checks are presumably in the companion paper, and that is where the claim stands or falls. Same goes for the continuum/infinite-volume extrapolations, which rest on a few points and exclude the coarsest lattice spacing from the cubic fit; the details are deferred.\n\nBottom line: this is for anyone who uses perturbative nucleation rates for gravitational-wave forecasts or for phase-transition phenomenology. It is a clear, honest, and useful benchmark, but it is a pointer to Ref. [19] rather than a self-contained validation. I would send it to a serious referee—the companion paper carries the weight, and this proceedings deserves scrutiny as the public face of that result. Read it, cite the modified order parameter, and go check the real-time systematics in the companion paper.","headline":"A short proceedings that points to a real and worrying result—the first lattice test of nucleation in a tree-level barrier model finds the lattice rate 20% below one-loop perturbation theory—but the systematics that would settle whether this is physics or an artifact live in the companion paper.","tokens_in":8385,"tokens_out":5100,"would_cite":true,"duration_ms":48006,"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":"A fully nonperturbative lattice computation of the bubble nucleation rate in a strong first-order phase transition finds that the one-loop perturbative result is too high by about 20% in $\\lvert\\log\\Gamma\\rvert$, and the tree-level result…","keywords":["bubble nucleation","first-order phase transition","lattice simulation","perturbation theory","nucleation rate","multicanonical method","real-time Langevin evolution","gravitational waves"],"falsifier":"Run a two-loop perturbative calculation at the same benchmark point: if the predicted rate still lies above the lattice value by about 20% in $\\lvert\\log\\Gamma\\rvert$, the loop expansion is not the explanation. Alternatively, repeat the real-time evolution with a different damping coefficient or integrator timestep; if the tunneling fraction shifts enough to change $\\log\\Gamma$ by the observed gap, the discrepancy is an artifact of the dynamical model.","tokens_in":7413,"feed_emoji":"🫧","tokens_out":12956,"duration_ms":106347,"temperature":0.7,"pith_summary":"The paper asks whether perturbative calculations of the bubble nucleation rate survive a fully nonperturbative test. The authors simulate the three-dimensional high-temperature effective theory of a single real scalar field with a tree-level barrier — the scenario where perturbation theory should be at its best — and compare the lattice nucleation rate with tree-level, local-potential-approximation, and one-loop results. The lattice rate agrees only qualitatively: the disagreement in $\\lvert\\log\\Gamma\\rvert$ is 20% at one loop and 100% at tree level, and the nonperturbative rate lies below the one-loop prediction throughout the reweightable temperature range. Since the latent heat in the same theory matches perturbation theory to better than 1%, the failure is specific to nucleation-rate calculations, which matter for gravitational-wave signals from cosmological phase transitions.","feed_headline":"Lattice simulation finds bubble nucleation rates 20-100% off","feed_subtitle":"A 20% error in the bubble rate shifts when and how strongly a cosmic phase transition rings gravitational waves.","key_machinery":"The rate computation rests on a factorisation of the nucleation rate into three pieces: the probability density of being near the critical bubble (the separatrix), an analytic Gaussian flux through the separatrix, and the fraction of near-critical configurations that actually tunnel when evolved in real time. The multicanonical weight function and the order parameter $\\theta_{\\mathrm{op}}=\\phi^2-2A\\phi$ isolate the suppressed critical-bubble peak while suppressing bulk phase fluctuations; a fourth-order symplectic integrator with momentum refresh then evolves the selected configurations forwards and backwards in a Langevin bath with damping $\\gamma=1/L$, so the Monte Carlo and real-time stages agree on where the separatrix lies.","core_discovery":"The central claim is that, for a strong first-order phase transition with a tree-level barrier in a single-scalar theory, the perturbative nucleation rate is not quantitatively reliable. Using multicanonical sampling to build critical-bubble configurations and real-time Langevin evolution to decide which of them tunnel, the authors extract the dimensionless rate $\\log(\\Gamma/\\lambda_3^4)$ on the lattice, continuum-extrapolate it, and compare it with analytic results. At the benchmark point, the lattice value is lower than the one-loop result by about 20% in $\\lvert\\log\\Gamma\\rvert$ and lower than the tree-level result by about 100%, with statistical errors much smaller than the perturbative renormalisation-scale bands. The paper does not claim to have identified the source of the mismatch; it concludes that higher-order perturbative calculations and further lattice studies are needed, mentioning possible extra saddle points and a potential breakdown of the saddle-point approximation as open possibilities.","pith_inferences":["If the lattice rate is the physical one, gravitational-wave spectra computed from one-loop rates would shift: a lower rate delays percolation, which changes both the peak frequency and the amplitude of the signal predicted for LISA-era detectors.","The size of the gap is consistent with the paper's suggested alternatives — extra saddle points beyond the critical bubble, or a breakdown of the saddle-point expansion — and would mean the standard bounce-action framework is missing a leading-order effect, not just loop corrections.","A sharper test would be to split the measured rate into the flux and tunneling-fraction factors and compute each nonperturbatively, since only the tunneling fraction depends on the real-time simulation and can be checked against direct Langevin nucleation studies at higher rates."],"forward_implications":["If the lattice result is correct, perturbative one-loop rates overestimate bubble nucleation for strong phase transitions by about 20% in $\\lvert\\log\\Gamma\\rvert$, so bubbles would form later and the transition would supercool more than one-loop estimates suggest.","Tree-level rate estimates are off by roughly a factor of about $e$ in the rate itself, making them inadequate for quantitative gravitational-wave phenomenology in models with a tree-level barrier.","The 20% one-loop gap gives a concrete target for two-loop calculations: they must move the predicted rate downward toward the lattice value to restore confidence in the perturbative expansion.","Because the latent heat agrees to better than 1%, thermodynamic quantities and nucleation rates can fail independently, so nonperturbative checks of the rate itself are needed even when bulk thermodynamics looks perturbative."],"supporting_citations":[{"why":"supplies the multicanonical separatrix method and the factorised approach to nonperturbative bubble nucleation that this work extends.","marker":"[14]"},{"why":"demonstrates the factorisation of the rate into critical-bubble probability, flux, and tunneling fraction in a nonperturbative computation.","marker":"[15]"},{"why":"is the preceding nonperturbative update of first-order electroweak phase transitions that motivates revisiting the method for tree-level-barrier models.","marker":"[16]"},{"why":"fixes the three-dimensional effective couplings and thermodynamic properties of the phase transition, supplying the benchmark point and the latent-heat comparison.","marker":"[17]"},{"why":"contains the full continuum and infinite-volume extrapolations and the figures the present paper summarises.","marker":"[19]"},{"why":"introduces the multicanonical ensemble algorithm used to sample the exponentially suppressed barrier region.","marker":"[20]"},{"why":"justifies using classical real-time evolution for hot scalar fields at leading order, grounding the Langevin step.","marker":"[22]"}],"fun_headline_variants":["Lattice test: nucleation rates 20-100% off","Nucleation rate mismatch: lattice vs theory up to 100%","Strong transitions: perturbative nucleation fails lattice check","Lattice simulation finds nucleation rate error 20-100%","Nucleation rates: perturbative vs lattice, 20-100% gap"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the real-time evolution step returns the true physical nucleation rate: if the way the simulation labels trajectories or computes the flux is wrong, the lattice rate is wrong and the discrepancy with perturbation theory is an artifact rather than a real failure of perturbation theory.","fun_headline_variants_meta":{"raw":{"variants":["Lattice test: nucleation rates 20-100% off","Nucleation rate mismatch: lattice vs theory up to 100%","Strong transitions: perturbative nucleation fails lattice check","Lattice simulation finds nucleation rate error 20-100%","Nucleation rates: perturbative vs lattice, 20-100% gap"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000632,"raw_usage":{"total_tokens":2857,"prompt_tokens":822,"completion_tokens":2035,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":438,"completion_tokens_details":{"reasoning_tokens":1946}},"tokens_in":438,"tokens_out":2035,"duration_ms":13983,"temperature":1.0,"reasoning_tokens":1946,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T23:12:43.133410+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a two-loop perturbative calculation at the same benchmark point: if the predicted rate still lies above the lattice value by about 20% in $\\lvert\\log\\Gamma\\rvert$, the loop expansion is not the explanation. Alternatively, repeat the real-time evolution with a different damping coefficient or integrator timestep; if the tunneling fraction shifts enough to change $\\log\\Gamma$ by the observed gap, the discrepancy is an artifact of the dynamical model.","supporting_citations":[{"cited_title":"Multicanonical Ensemble: A New Approach to Simulate First-order Phase Transitions","cited_arxiv_id":"hep-lat/9202004","evidence_quote":"introduces the multicanonical ensemble algorithm used to sample the exponentially suppressed barrier region."}],"review_version":1}