{"id":"99088beb-1c38-4e0d-a7f8-3a82d6551465","arxiv_id":"2509.04132","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"Simulations show ion velocity statistics in a model tokamak turbulence are approximately stationary, ergodic, and time-symmetric to about 1%, thanks to broad initial phase-space sampling.","lead":"The paper simulates ion motion in a model tokamak's turbulent electric field and checks whether particle statistics stay steady, random, and reversible even where the forces are uneven. The answer is yes, to about one percent, because particles start spread across wide conditions.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Super-ensemble construction may pre-impose Lagrangian stationarity; the shared-field test compares only D(t), not the two-time autocorrelation that anchors the central claim.","rationale":"The paper is a careful, honest numerical study with convergence checks and a genuine control experiment, but its headline claim about approximate Lagrangian stationarity in tokamak-relevant geometries overreaches the simulation design. The super-ensemble construction is the load-bearing premise: when each particle sees an independent realization of a time-stationary random field, the ensemble-averaged L(t0,t0+t) is heavily constrained to be approximately stationary even if the Eulerian drifts are inhomogeneous and compressible. The physical tokamak case is a single turbulent field shared by all particles; the only single-realization test in the paper compares D(t), not the two-time autocorrelation that actually defines the stationarity statement. This is not an internal inconsistency in the T3ST model, but it does mean the strongest claim—that the statistical assumptions underpinning reduced transport models hold in tokamak-like conditions—is not directly evidenced. The proposed single-realization autocorrelation test is a natural, computationally modest way to close that gap. If it passes, the paper's conclusions extend as stated; if it fails, the claims should be explicitly restricted to the synthetic super-ensemble. The reader identified essentially this same concern, so my assessment is in agreement: a conditional verdict, pending the shared-field stationarity test.","tokens_in":20721,"tokens_out":6947,"duration_ms":76363,"concrete_test":"Run the single-realization setup of Sec. III F (all Np particles sharing one turbulent field realization) and compute the radial Lagrangian velocity autocorrelation L(t0,t0+t) for t0 = 0, 5, 10, 20, 30 R0/vth, exactly as in Fig. 7 but using the particle average within that single realization. Repeat for 10-20 independent realizations to estimate the realization-to-realization spread. If the L(t0,t0+t) curves are mutually indistinguishable at the reported ~1% level, the super-ensemble stationarity is representative of a shared field and the central claim survives; if the spread across t0 or across realizations is comparable to or larger than the reported 5-20% sensitivity thresholds, the stationarity claim must be restricted to the super-ensemble.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central stationarity result is obtained with T3ST's super-ensemble: each particle evolves in its own independent realization of the random-phase Fourier field (Sec. II C, Eqs. 12-14), and the autocorrelation L(t0,t0+t) is averaged over that ensemble in Sec. III C. Because each field realization is time-stationary by construction and the independent-realization average erases realization-specific histories, the approximate independence of L(t0,t0+t) from t0 is largely a property of the ensemble, not of the inhomogeneous/compressible drift fields the paper claims to test. The only single-shared-field comparison, Sec. III F and Fig. 14, tests the running diffusion coefficient D(t), which temporally integrates and smooths the autocorrelation; it is far less sensitive to a time-origin dependence of L than L itself, as the paper's own Sec. III H demonstrates for DL vs Dd. Moreover, the single-trajectory test in Sec. III I, which also uses independent field realizations, shows that stationarity is broken when the initial phase-space sampling is narrow. This supports the super-ensemble attribution but does not establish that a single tokamak-like shared field would produce stationary L(t0,t0+t). The conclusion's extrapolation to 'tokamak-relevant geometries' therefore rests on an untested equivalence between the synthetic super-ensemble and a physical single realization.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper uses the T3ST test-particle code to construct a random-phase ITG-like turbulent electrostatic field ensemble and asks whether Lagrangian velocity statistics are stationary, ergodic, and time-symmetric despite the inhomogeneous and compressible character of the Eulerian gyrocenter drifts. The central numerical finding is that, in the standard T3ST setup (broad Maxwellian initial distribution plus an independent field realization per particle), the radial velocity autocorrelation L(t0,t0+t) is nearly independent of t0, the velocity variance drifts by about 1%, and forward/backward time simulations agree. The paper additionally reports a turbulent equipartition pinch V≈2D/R0, a ~5%–20% sensitivity of the long-time diffusion coefficient to the initial phase-space sampling, a close but imperfect agreement between MSD- and correlation-based diffusion estimates, and a single-trajectory control showing that stationarity is lost when the initial phase-space sampling is narrow. The conclusion is that the statistical assumptions underlying reduced transport models (Green-Kubo relations, DTM, Fick closures) hold to good approximation in tokamak-relevant geometries.","tokens_in":21114,"tokens_out":6184,"duration_ms":62173,"significance":"If the result holds as stated, it provides valuable numerical support for the use of stationary, ergodic statistical descriptions of turbulent transport in fusion plasmas, and it sharpens the known limitations of such descriptions. The paper has genuine strengths: it includes a control experiment (Section III I), convergence tests (Fig. 20), and an unusually candid self-assessment of the DL estimator (Section III H). The claims are appropriately hedged with 'approximately' and '≈1%'. The main significance is therefore conditional: the evidence is strong for the super-ensemble used in T3ST, but the extrapolation to a single physical turbulent realization is not yet tested at the level of the autocorrelation function, which is the quantity that anchors the paper's central claim.","major_comments":[{"comment":"The central stationarity result is obtained with the T3ST super-ensemble in which each of the Np particles evolves in its own random-phase field realization (Sec. II C, Eqs. 12–14). Because the field ensemble is stationary and homogeneous by construction, the near-independence of L(t0,t0+t) from t0 in Figs. 7b and 8b is largely an ensemble property. The only single-shared-field comparison, Fig. 14, tests the running D(t), not L(t0,t0+t); D(t) is a time integral of L and is therefore far less sensitive to a time-origin dependence, as the paper itself demonstrates in Sec. III H for DL vs Dd. To support the conclusion that Lagrangian stationarity holds in a tokamak-like single realization, the manuscript should compute L(t0,t0+t) for a broad initial distribution in a single shared field, or should explicitly restrict the claim to the super-ensemble.","section":"III C and III F"},{"comment":"The identification of the radial pinch as a turbulent equipartition pinch is presented as confirmed by the agreement with V≈2D/R0. However, starting from the linearization B(x)^-1 ≈ B(0)^-1(1 - X·∇lnB(0)) and the assumption that X is E×B-dominated, Eqs. (3)–(5) make both V(t) and 2D(t)/R0 time integrals of the same velocity autocorrelation ⟨Vr(t)Vr(τ)⟩. The agreement in Fig. 5 is therefore largely a consistency check of the linearization rather than an independent confirmation of the TEP mechanism. Please state this explicitly, or provide a test that does not use the same L(t,τ) on both sides (e.g., a direct measurement of the compressibility contribution).","section":"III B, Eq. (5), Fig. 5"},{"comment":"The single-trajectory experiment compares one fixed initial phase-space point across independent field realizations with the standard super-ensemble. It demonstrates that narrowing the initial phase-space sampling changes the outcome, but it does not isolate the physical mechanism: in the standard case the initial distribution is broad and each particle also sees an independent field realization, whereas in the single-trajectory case both are collapsed. The claim that 'the extensiveness of the initial kinetic distribution' causes the apparent stationarity still needs a control in which a broad initial distribution is evolved in a single shared field realization. Without that control, the attribution to phase-space sampling rather than to ensemble averaging over fields remains ambiguous.","section":"III I"}],"minor_comments":[{"comment":"The displayed formula for V(t|x) contains typographical errors in the braces and superscript: it should be V(t|x) = d/dt {⟨X(t|x)⟩}, with one closing brace and no stray '⟩}^2'. Please correct.","section":"Eq. (11)"},{"comment":"The text contains repeated typos: 'ask weather' should be 'ask whether', and 'a more consistent approach would be to initialize particles in either a known equilibrium state ... or a steady-state—like the one reached asymptotically under pure magnetic motion' contains a long dash that should be an em dash. These are minor but should be fixed.","section":"III C and III D"},{"comment":"The abbreviation TEP is introduced as 'Turbulence Equipartition Pinch' in Section III B but as 'Turbulent Equipartition' in the Conclusions. Please use one consistent name and define it at first use.","section":"III B and Conclusions"},{"comment":"The statement that D∞_L provides 'a smaller average value' than D∞_d is made without a quantitative bias estimate. Since this is directly relevant to the '≈1% stationarity' claim, please report the mean bias and its uncertainty, or explain why the bias is not statistically significant.","section":"III H, Figs. 19–20"}],"recommendation":"major_revision","confidential_remarks":"The stress-test concern from the reader largely lands. The central claim is demonstrated for the super-ensemble, and the only shared-field comparison (Fig. 14) is not sensitive to the quantity that matters. I would not reject the paper because the claim is hedged and the missing test is feasible within the manuscript's scope. The revision should add a single-realization autocorrelation analysis or explicitly restrict the conclusion, and it should relabel the TEP agreement as a consistency check. If the author can provide the shared-field L(t0,t0+t) test, the paper would be a solid contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this is a genuinely useful numerical verification paper, but the central claim needs a boundary drawn around it. What is new and good: the paper quantifies approximate Lagrangian stationarity to ~1% in a compressible, inhomogeneous gyrocenter model, runs a real control experiment (single trajectory, Sec. III I), includes convergence tests (Fig. 20), and even documents an uncomfortable result about the Green-Kubo estimator under-predicting diffusion (Sec. III H). The author hedges appropriately and puts real effort into checking numerical resolution. For the T3ST model as defined, the conclusions largely hold up.\n\nThe soft spot is the super-ensemble construction. Each particle evolves in its own independent random-phase field realization, which is time-stationary and Gaussian by construction. Averaging over independent realizations erases realization-specific histories, so the near-independence of L(t0,t0+t) from t0 is partly an ensemble artifact, not a discovery about inhomogeneous drift fields. The single-shared-field comparison (Fig. 14) only tests the running diffusion coefficient D(t), which temporally integrates and smooths the autocorrelation; it does not test two-time stationarity of L within one physical field. The single-trajectory test shows stationarity is broken when the initial sampling is narrow, which supports the super-ensemble attribution but also shows how much the result depends on how you build the ensemble. Relatedly, the TEP section is close to tautological: under the stated linearization, V(t) ≈ 2D(t)/R0 with both derived from the same measured autocorrelation, so Fig. 5 is a consistency check rather than independent confirmation.\n\nNone of this contradicts the presented data, and the paper is honest enough to flag several of its own limitations. The overreach is the abstract's phrase 'tokamak-relevant geometries': a single self-consistent gyrokinetic field with zonal flows, collisions, and non-Gaussian structures could behave differently. The fix is straightforward—explicitly state that the results characterize the T3ST synthetic-turbulence ensemble, and test L(t0,t0+t) stationarity within a single shared field realization.\n\nWho this is for: people working on test-particle transport models, Green-Kubo closures, and statistical ensembles for plasma turbulence. It deserves a serious referee; the numerics are careful and the control experiment is a good idea, but the referee should push on the ensemble-to-single-realization step. I would cite it for the model-specific results, not for the broad tokamak conclusion.","headline":"A careful, honest numerical study of approximate Lagrangian stationarity in a synthetic-turbulence test-particle model, but the abstract's tokamak extrapolation outruns the evidence because the super-ensemble construction builds in much of the stationarity that is then reported.","tokens_in":21607,"tokens_out":1687,"would_cite":true,"duration_ms":16408,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["52.25.Fi","52.35.Ra","52.65.-y"],"model":"deepseek-v4-flash","headline":"Tokamak turbulence keeps Lagrangian statistics nearly stationary despite inhomogeneous drifts","keywords":["Lagrangian statistics","turbulent transport","tokamak","gyrocenter test particles","ergodicity","stationarity","Green-Kubo relation","turbulence equipartition pinch"],"falsifier":"Run the same test-particle diagnostics in a global gyrokinetic simulation with a single shared, self-consistently saturated turbulent field (including zonal flows) and compare the two-time Lagrangian autocorrelation L(t0, t0+t) across different t0 values, along with the difference between MSD and Green-Kubo diffusion coefficients; if the deviations exceed the few-percent level in that setting, the paper's stationarity claim would not carry over to realistic turbulence.","tokens_in":1786,"feed_emoji":"🌪️","tokens_out":1695,"duration_ms":29059,"temperature":0.7,"pith_summary":"This paper asks whether the standard statistical assumptions used to model turbulent transport in tokamaks—stationarity, ergodicity, time-symmetry, and ensemble equivalence—still hold when the underlying drift fields are both compressible and spatially inhomogeneous. Using test-particle simulations with the T3ST code, the author finds that Lagrangian velocity statistics are approximately stationary, with deviations of about 1%, and that transport is nearly unchanged when the initial particle distribution is narrowed. The reason, the paper argues, is not that the drift fields are homogeneous, but that the broad initial phase-space distribution supports ergodic mixing. If true, this would justify the continued use of Green-Kubo-type relations, the Decorrelation Trajectory Method, and Fick-like closures in reduced tokamak transport models, even though the formal conditions for those tools are not exactly met.","feed_headline":"Tokamak turbulence passes a key statistical test","feed_subtitle":"Test-particle runs show Lagrangian stationarity holds to about 1%, backing standard reduced transport models.","key_machinery":"The central object is the Lagrangian velocity autocorrelation function L(t,t') for radial gyrocenter velocities, along with its two-time structure and its use in computing the diffusion coefficient via the Green-Kubo integral DL(t) = ∫₀ᵗ L(τ)dτ, compared against the mean-square-displacement definition Dd(t). The paper also uses the super-ensemble average over both initial phase-space coordinates and independent field realizations, and a linearized form of the E×B drift, B(x)⁻¹ ≈ B(0)⁻¹(1 - X·∇lnB(0)), to derive the turbulence equipartition pinch as V(t) ≈ (2D(t)/R₀) ∇lnB.","core_discovery":"The central claim is that, despite the Eulerian gyrocenter drifts being compressible and inhomogeneous, the Lagrangian statistics of test-particle transport in tokamak-like turbulence are approximately stationary, ergodic, and time-symmetric. The paper supports this with simulations of 5×10^5 particles in the Cyclone Base Case, showing that the Lagrangian velocity autocorrelation is nearly independent of the initial time, that forward and backward time evolutions produce essentially identical transport coefficients and particle distributions, and that single-realization and ensemble-average diffusion agree to about 5%. It also shows that the long-time diffusion coefficient changes by only 5-","pith_inferences":["The paper's mechanism suggests a testable extension for real tokamak turbulence: if a self-consistent, single-shared turbulent field (including zonal flows and non-Gaussian structures) is used instead of independent random-phase realizations, the measured Lagrangian stationarity may degrade because the ergodic mixing provided by initial conditions would have to compensate for a field that is no lo","The 1%-level stationarity found here is likely tied to the Gaussianity and homogeneity of the synthetic field ensemble; introducing drift-wave packets, streamers, or intermittent bursts could push the deviations beyond the few-percent range, providing a direct stress test of the paper's claim.","The TEP pinch expression V ∝ 2D/R₀ could be probed experimentally by comparing measured impurity or main-ion radial convection with independently inferred diffusion coefficients in discharges where turbulence is the dominant drive, isolating the geometric contribution.","The distinction between DL and Dd suggests a practical estimator for diagnosing non-stationarity in any test-particle or tracer simulation: the persistent gap between correlation-based and MSD-based diffusion coefficients measures the violation of Lagrangian stationarity directly."],"forward_implications":["Reduced transport models that assume Lagrangian stationarity, such as Green-Kubo relations and the Decorrelation Trajectory Method, can be applied in tokamak-relevant geometries with expected errors of roughly a percent.","The asymptotic diffusion coefficient is robust to changes in initial phase-space sampling, so test-particle codes can initialize particles from simpler distributions without materially altering predicted transport.","A finite radial pinch, here identified as a turbulence equipartition pinch proportional to the local diffusion coefficient, should be included in local transport closures even when temperature gradients and rotation are absent.","The time-integrated Green-Kubo estimator DL tends to underestimate the asymptotic diffusion relative to the MSD-based estimator, because the slight non-stationarity of the velocity amplitude is not captured by L(0,t); DL should be used with caution when stationarity is imperfect.","The apparent ergodicity of gyrocenter transport arises from broad initial phase-space sampling, so simulations with narrow or non-representative initial distributions may produce spurious transport statistics even if the field model is accurate."],"supporting_citations":[{"why":"Taylor's diffusion by continuous movements supplies the foundational relation between Lagrangian velocity correlations and the diffusion coefficient used throughout.","marker":"[2]"},{"why":"Monin, Yaglom, and Lumley's statistical fluid mechanics provides the symmetry properties of Lagrangian quantities for homogeneous, stationary, divergence-free Eulerian fields that the paper tests against.","marker":"[3]"},{"why":"Balescu's anomalous transport monograph supplies the statistical-ensemble framework for replacing a single turbulent field with an ensemble of random fields.","marker":"[4]"},{"why":"The T3ST code paper is the numerical platform whose test-particle simulations generate all the reported data.","marker":"[8]"},{"why":"Brizard and Hahm's foundations of nonlinear gyrokinetic theory provides the gyrocenter equations of motion that T3ST integrates.","marker":"[15]"},{"why":"The Cyclone Base Case parameters from Dimits et al. define the simulated DIII-D-like scenario.","marker":"[24]"},{"why":"The European turbulence code benchmarking effort supplies the gyrokinetic reference values for phase velocity, growth rates, and correlation times used to set the synthetic turbulence spectrum.","marker":"[25]"},{"why":"Isichenko, Gruzinov, and Diamond's invariant-measure theory provides the turbulence equipartition pinch concept that the paper's derived radial pinch is compared against.","marker":"[31]"},{"why":"Vlad et al.'s Decorrelation Trajectory Method is the reduced transport formalism whose validity the paper's stationarity results directly support.","marker":"[34]"}],"fun_headline_variants":["Tokamak turbulence passes ergodicity check","Test particles show tokamak turbulence is stationary","Lagrangian stats support tokamak transport models","Simulations confirm tokamak turbulence is ergodic","Tokamak turbulence: Lagrangian time-symmetry holds"],"cache_read_input_tokens":23168,"weakest_assumption_plain":"The load-bearing premise is that the statistics of the synthetic turbulence ensemble—independent random-phase Fourier realizations that are by construction Eulerian-stationary, homogeneous, and Gaussian—represent a real tokamak's single self-consistent turbulent field, which includes zonal flows, collisions, and non-Gaussian structures that the model omits.","fun_headline_variants_meta":{"raw":{"variants":["Tokamak turbulence passes ergodicity check","Test particles show tokamak turbulence is stationary","Lagrangian stats support tokamak transport models","Simulations confirm tokamak turbulence is ergodic","Tokamak turbulence: Lagrangian time-symmetry holds"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000537,"raw_usage":{"total_tokens":2332,"prompt_tokens":576,"completion_tokens":1756,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":320,"completion_tokens_details":{"reasoning_tokens":1692}},"tokens_in":320,"tokens_out":1756,"duration_ms":15471,"temperature":1.0,"reasoning_tokens":1692,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T10:22:33.834201+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same test-particle diagnostics in a global gyrokinetic simulation with a single shared, self-consistently saturated turbulent field (including zonal flows) and compare the two-time Lagrangian autocorrelation L(t0, t0+t) across different t0 values, along with the difference between MSD and Green-Kubo diffusion coefficients; if the deviations exceed the few-percent level in that setting, the paper's stationarity claim would not carry over to realistic turbulence.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Taylor's diffusion by continuous movements supplies the foundational relation between Lagrangian velocity correlations and the diffusion coefficient used throughout."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Monin, Yaglom, and Lumley's statistical fluid mechanics provides the symmetry properties of Lagrangian quantities for homogeneous, stationary, divergence-free Eulerian fields that the paper tests against."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Balescu's anomalous transport monograph supplies the statistical-ensemble framework for replacing a single turbulent field with an ensemble of random fields."},{"cited_title":"Palade and L.M","cited_arxiv_id":null,"evidence_quote":"The T3ST code paper is the numerical platform whose test-particle simulations generate all the reported data."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Brizard and Hahm's foundations of nonlinear gyrokinetic theory provides the gyrocenter equations of motion that T3ST integrates."},{"cited_title":"Collisional transport in plasma","cited_arxiv_id":null,"evidence_quote":"The Cyclone Base Case parameters from Dimits et al. define the simulated DIII-D-like scenario."},{"cited_title":"Dif-Pradalier, P","cited_arxiv_id":null,"evidence_quote":"The European turbulence code benchmarking effort supplies the gyrokinetic reference values for phase velocity, growth rates, and correlation times used to set the synthetic turbulence spectrum."},{"cited_title":"Jenko, W","cited_arxiv_id":null,"evidence_quote":"Isichenko, Gruzinov, and Diamond's invariant-measure theory provides the turbulence equipartition pinch concept that the paper's derived radial pinch is compared against."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Vlad et al.'s Decorrelation Trajectory Method is the reduced transport formalism whose validity the paper's stationarity results directly support."}],"review_version":1}