{"id":"f9657cf5-4bdc-45b2-b14d-c99300b02252","arxiv_id":"2411.16436","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":5,"one_line_summary":"A filtered-threshold segmentation method extracts two-structure conditional averages in a simulated Rayleigh-Taylor layer, claimed as first reference data for two-structure RANS models.","lead":"This paper proposes a method to split the turbulent mixing zone of a Rayleigh-Taylor instability into upward-moving light and downward-moving heavy fluid structures by smoothing and thresholding the vertical velocity field. It reports conditional statistics from a high-resolution simulation and claims these are the first reference data for calibrating two-structure turbulence models.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central empirical claims are absent: Part 6 is an empty 'In progress' placeholder, so the abstract's 'first known reference data' and 'up to about 40%' statements are unsupported in this version.","rationale":"The reader's verdict of REJECT is justified, though for a slightly different primary reason than the stated weakest_assumption. The reader's weakest_assumption identifies the ad hoc b+ and b- segmentation as the main risk; that is a real substantive concern, but it can only be evaluated once the data are actually present. In this version the data are not present: Part 6 is a placeholder, and no measured conditional averages or directed-energy ratios are included. The theoretical framework in Parts 3-5 is elaborate and appears internally consistent; the algebraic decomposition in Eqs. (3.18)-(3.22) is standard conditional-averaging algebra, and the connection to the buoyancy-drag equation in Section 3.3 is plausible. I give credit for the careful specification of the filtering equation (5.1) and the optimization protocol in Section 5.4. However, none of that constitutes the empirical 'reference data' promised in the abstract. The manuscript itself signals this: Section 1.3 says 'Part 6 regroups the results', Section 5.3 says 'Ensuing two-structure correlations will be examined in part 6', and Section 6 is titled 'In progress'. The data availability statement is a placeholder. Because a central claim cannot be verified from the submitted manuscript, REJECT is the appropriate verdict for this version. If a complete revision appears, the decisive check would be reproduction of the approximate 40% directed-energy ratio and the conditional averages from the archived simulation; until then the concern is not about a wrong number but about a missing one.","tokens_in":55154,"tokens_out":4061,"duration_ms":39768,"concrete_test":"Inspect the full submitted source of arXiv:2411.16436v1 for any measurement content in Part 6. If it is empty, the abstract's claims are unsupported. If a revised version with Part 6 is provided, recompute the directed-to-turbulent energy ratio from the 1024^3 LES using the Eq. (5.8) coefficients and the definitions in Section 3.7, and compare with the claimed approximate 40% value; also verify that the conditional averages can be reproduced from the archived b+ and b- fields within the reported statistical uncertainty.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing problem is not a subtle flaw in the segmentation but a missing measurement. The abstract states that conditional averages were measured and 'provide the first known reference data' for two-structure RANS models, and that directed-to-turbulent energy reaches 'up to about 40%'. Yet Section 6, which is supposed to contain 'Two-structure field correlations from a simulated turbulent RT flow,' is literally an 'In progress' placeholder (p. 36). No conditional averages, no energy ratios, no profiles, no error bars, and no data-bearing figures are reported anywhere in the manuscript. Section 5.4 reports only optimized filter coefficients (C_omega=1.7, C_ell=0.15, C_wedge=30) and the resulting bimodality and interfacial-area diagnostics; it does not contain the physics results claimed in the abstract. The data availability statement is also a template ('openly available in [repository name] at http://doi.org/[doi]'). Thus the central claim cannot be checked. If the missing section were supplied, the secondary concern flagged by the reader (that b+ and b- are an ad hoc thresholding recipe per Section 2.3, optimized against their own bimodality coefficient) would become the decisive test, but on the submitted text the empirical content is simply absent.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript develops a two-structure-field conditional-averaging framework for Rayleigh-Taylor turbulent mixing layers, defines complementary presence fields b+ and b- through Lagrangian space-time filtering of the vertical velocity, and derives exact algebraic decompositions and energy identities for the resulting conditional statistics. Part 5 presents an optimization of the filtering coefficients against a bimodality criterion, and the abstract claims that the resulting conditional averages from a low-Atwood LES provide the first reference data for calibrating two-structure RANS models, with directed-to-turbulent energy ratios up to about 40%. However, the section that is supposed to report these simulation results, Section 6, is an empty placeholder titled \"In progress,\" and no conditional-average profiles, energy ratios, or error bars appear anywhere in the manuscript. The empirical claims in the abstract and introduction are therefore not supported by the submitted text.","tokens_in":55431,"tokens_out":3400,"duration_ms":32454,"significance":"If the missing measurements were supplied and the segmentation were shown to identify persistent physical structures, the framework could be a useful contribution: the algebraic derivation of the directed-flux decomposition (3.18)-(3.20) and the energy budget identities in Appendix C are internally consistent and provide a clean theoretical basis for interpreting two-structure RANS models. The paper also gives a broad and useful review of prior evidence for bimodality and directed energy in RT flows. Nevertheless, as submitted, the central deliverable is absent. The claim to provide the 'first known reference data' cannot be assessed, and the segmentation method is optimized against its own bimodality coefficient with no external validation, so even a completed Section 6 would need a convincing demonstration that the b+ and b- fields correspond to genuine structures rather than an artifact of the thresholding recipe.","major_comments":[{"comment":"Section 6, titled \"In progress: Two-structure field correlations from a simulated turbulent RT flow,\" contains no conditional-average profiles, no energy ratios, no profiles of the quantities announced in the introduction, and no error bars or data-bearing figures. The abstract nevertheless states that the measured conditional averages provide the first known reference data for two-structure RANS models and that directed-to-turbulent energy reaches up to about 40%. These are the paper's central empirical claims, and they are unsupported by the submitted text. This is a load-bearing omission, not a presentation issue.","section":"Section 6; Abstract"},{"comment":"Section 5.4 reports only the optimized filter coefficients (C_omega=1.7, C_ell=0.15, C_wedge=30) and the resulting mean bimodality and interfacial-area diagnostics. None of the structure-conditioned averages announced in Section 1.2 and Section 3.7 is given. Equations (5.8) cannot by itself substantiate the abstract's quantitative claims about directed energy, and cross-references such as \"as will be observed from the simulations in part 6\" (Section 3.6) and \"see part 6 and figure ??a\" (Section 3.7) point to material that is not present.","section":"Section 5.4, Eq. (5.8)"},{"comment":"The definition of b+ and b- is explicitly described as \"a somewhat ad hoc recipe\" (Section 2.3), and the coefficients in (5.8) are optimized to maximize a bimodality coefficient (Section 5.4) without comparison to an independent structure-identification method or to the visual eduction that motivates the approach. Because all conditional averages and directed-energy quantities inherit this segmentation, the missing external validation is a genuine correctness risk for the central claim. This concern would become decisive once the missing Section 6 is supplied, and should be addressed with, for example, a comparison to alternative structure-identification methods, a test of persistence in time, or a demonstration that the resulting conditional averages are insensitive to the chosen optimization criterion.","section":"Section 2.3 and Section 5.4"},{"comment":"The data availability statement reads \"The data that support the findings of this study are openly available in [repository name] at http://doi.org/[doi], reference number [reference number].\" This is a template placeholder and does not provide any way to verify the claimed simulation results. If the manuscript is resubmitted with the empirical section, a complete data citation will be required.","section":"Data availability statement"}],"minor_comments":[{"comment":"There is a typo in the abstract: \"first know reference data\" should be \"first known reference data.\"","section":"Abstract"},{"comment":"The text contains the typo \"two-structure field.s\" in the paragraph following Figure 2; it should read \"two-structure fields.\"","section":"Section 2.3"},{"comment":"Several figure captions contain placeholder remarks such as \"To be completed,\" \"Old. Update.,\" and \"Old. Replace.\" These must be replaced with proper descriptions before any resubmission.","section":"Figure 8 and Figure 9 captions"},{"comment":"Unresolved cross-references such as \"see part 6 and figure ??a\" and \"figure??\" appear in the text; these need to be resolved once the empirical section is written.","section":"Sections 3.7, 4.4, and 5.4"},{"comment":"The phrase \"Adapted from previous previous investigations\" contains a duplicated word and should read \"Adapted from previous investigations.\"","section":"Section 2.1"}],"recommendation":"reject","confidential_remarks":"The manuscript is an incomplete draft: the section that should contain the reported measurements is literally titled \"In progress,\" and several figure captions contain editing notes. The central claims in the abstract are not testable from the submitted text. This is not a case of a subtle technical flaw that could be remedied by a local revision; the empirical content of the paper is absent. If the authors complete Section 6 with the actual conditional statistics, perform an external validation of the segmentation, and replace the placeholder data-availability statement, the resulting paper could be a valuable contribution, but the present submission does not meet the standard for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this up front: the advertised central result is missing. The abstract says the paper provides 'the first known reference data' for two-structure RANS models and that directed energy reaches 'up to about 40%' of turbulent energy. But Section 6, which is supposed to contain those conditional averages, is literally an 'In progress' placeholder (p. 36). No profiles, no energy ratios, no error bars, no data-bearing figures. The data availability statement is a template with bracketed placeholders. So the paper as submitted is a theory/methods manuscript with a hole where the empirical payoff should be.\n\nWhat the paper does well is the scaffolding. The conditional-averaging algebra in Sections 3 and Appendix C is internally consistent; the decompositions (3.18)–(3.20) and the energy identities are derived cleanly from definitions, and the directed-energy concept is usefully formalized. The Lagrangian space-time filtering prescription in Section 5 is a genuinely new idea, with a sensible optimization strategy (bimodality coefficient plus interfacial area) and a sensitivity analysis around the chosen coefficients. Section 4's surrogate two-fluid estimates, based on published RT/RM/KH simulations, give a quantitative motivation for why directed effects matter in RT. The expected structure-volume-fraction profiles in Section 3.8 are a nice theoretical benchmark. All of that is real and mostly careful work.\n\nThe soft spot, beyond the missing section, is that the segmentation itself is admittedly ad hoc (Section 2.3 calls it 'a somewhat ad hoc recipe'), and the filter coefficients are optimized to maximize bimodality. That means the conditional averages, when they appear, will inherit that arbitrariness. Without an independent structure-identification method or some ground truth, the 'reference data' might be calibrating an artifact. But that is a secondary concern; the primary problem is simply that the data are not in this version.\n\nMy recommendation: do not send this to peer review as is. It is not a finished paper. The right move is to return it to the authors with a request to complete Section 6 with the actual conditional averages, error bars, and a usable data/code statement. If they do that, the theoretical framework and the novel segmentation method make it worth serious refereeing. Until then, the abstract overclaims and the manuscript underdelivers.","headline":"The paper's advertised measurements do not exist in the manuscript: Section 6 is an 'In progress' placeholder, so the abstract's claims of first reference data and 40% directed energy are unsupported by the submitted text.","tokens_in":56015,"tokens_out":2189,"would_cite":false,"duration_ms":23540,"reading_group":"no","serious_thinker":"yes","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["76F45","76F25","76T","76F55"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that the turbulent mixing layer produced by the Rayleigh-Taylor instability is organized into two persistent large-scale populations—light fluid moving up and heavy fluid moving down—and that thresholding a space-time…","keywords":["Rayleigh-Taylor instability","turbulent mixing layer","two-structure RANS models","directed energy","bimodality coefficient","Lagrangian space-time filtering","conditional averaging","turbulence modelling"],"falsifier":"Run the same two-structure segmentation on a simulation at higher Atwood number or with time-dependent acceleration and check whether the optimized coefficients and the measured directed-energy ratio stay in the same range; alternatively, compare the b^+ and b^- boundaries against an independent structure-identification method, such as Lagrangian coherent structures or bubble-tracking, in the same DNS and see whether the conditional statistics survive an independent structure definition.","tokens_in":54970,"feed_emoji":"🌊","tokens_out":7712,"duration_ms":68546,"temperature":0.7,"pith_summary":"The paper argues that fluctuations inside a Rayleigh-Taylor mixing layer are not just turbulent noise but are organized into two persistent, bimodally distributed large-scale populations: light fluid moving upward and heavy fluid moving downward. It develops a concrete segmentation recipe that thresholds a vertical velocity field smoothed by a Lagrangian convection-diffusion-relaxation filter, with filtering coefficients tuned to maximize a bimodality coefficient. Applied to a $1024^{3}$ simulation at Atwood number 0.01, the recipe yields structure-conditioned averages of concentration, momentum, and turbulent kinetic energy, which the paper presents as the first reference data for calibrating two-structure RANS models such as Youngs' and 2SFK. The central quantitative finding is that the directed energy—kinetic energy associated with the relative drift of the two structures—reaches about 40 percent of the total turbulent energy, well above earlier estimates and markedly higher than in shear or Richtmyer-Meshkov layers. This matters because the usual single-fluid gradient-diffusion closures cannot represent this directed contribution, so the new conditional averages give modellers a direct target to improve transport predictions in buoyancy-driven flows.","feed_headline":"Filtered velocity splits Rayleigh-Taylor mixing into two structures","feed_subtitle":"Conditional averages give two-structure RANS models their first calibration data; directed energy reaches 40 percent.","key_machinery":"The load-bearing object is the pair of presence fields b^+ and b^- produced by thresholding a filtered vertical velocity. The filter is a convection-diffusion-relaxation equation with a Lagrangian derivative, a local relaxation rate based on dissipation and longitudinal turbulent kinetic energy, and a global diffusion length based on transverse turbulent kinetic energy; the coefficients C_omega = 1.7, C_ell = 0.15, and C_wedge = 30 are optimized to maximize a bimodality coefficient over the mixing layer. A generalized Otsu threshold then defines the two structure fields. All of the paper's conditional averages, the decomposition of the single-fluid turbulent flux into per-structure fluxes plus a directed flux, and the measured directed-energy ratio inherit their definition from this segmentation, so the filtering-and-threshold procedure is what carries the argument.","core_discovery":"The central claim is that a turbulent Rayleigh-Taylor mixing layer at low Atwood number can be decomposed, without modelling assumptions, into two complementary presence fields, here denoted b^+ and b^-, defined by thresholding a space-time filtered vertical velocity. The resulting conditionally averaged equations separate the total turbulent flux into two per-structure fluxes plus an inter-structure term, the directed flux, and the simulation shows that the structure fields extend across the entire mixing layer. The probability density function of the filtered separator field is genuinely bimodal across the layer—the bimodality coefficient drops from about 1.6 for the unfiltered velocity to about 0.55 after optimized filtering—and the directed-to-turbulent energy ratio is measured at up to about 40 percent. On this basis the paper asserts that these structure-conditioned averages provide the first known reference data for validating and calibrating two-structure RANS turbulence models.","pith_inferences":["If the segmentation is physically sound, the same filtering-and-threshold recipe could be applied to experimental image pairs such as PLIF/PIV, where particle tracking cannot easily follow persistent volumes; the conditional averages would then have an experimental counterpart to these simulation data.","Because the method's parameters were tuned on a constant-acceleration, low-Atwood case, the 40 percent directed-energy figure may shift at finite Atwood number or under variable acceleration; measuring the ratio across those regimes would test whether two-structure modelling is the right general description.","The paper's decomposition implies, as a direct corollary it does not itself report, that the per-structure Reynolds stresses should be much closer to isotropic than the total Reynolds stress in a Rayleigh-Taylor layer, since the directed term is purely axial; that is a testable prediction from existing simulation data."],"forward_implications":["Two-structure RANS models can now be calibrated against directly measured conditional averages of concentration, momentum, and turbulent energies, instead of being tuned only through global growth-rate data.","Single-fluid models that close the total turbulent flux by gradient diffusion will systematically underestimate Rayleigh-Taylor transport, because the directed contribution to the flux—about 40 percent of the turbulent energy in this simulation—is non-diffusive.","The measured exchange terms between upward- and downward-moving structures give concrete targets for drag and entrainment closures in models such as Youngs' and 2SFK.","The bimodality and persistence criteria used to optimize the filter provide a quantitative way to judge whether a proposed structure segmentation captures genuine large-scale organization rather than small-scale noise."],"supporting_citations":[{"why":"The Youngs-type two-structure RANS model whose closure terms the new conditional averages are meant to calibrate; the abstract names it as the primary target.","marker":"Kokkinakis et al. (2015)"},{"why":"The 2SFK two-structure two-fluid model, the other named calibration target for the measured conditional averages.","marker":"Llor & Bailly (2003)"},{"why":"Supplies the two-structure conditional-averaging equations and the directed-energy/directed-flux decomposition on which the analysis in parts 3 and 4 rests.","marker":"Llor (2005)"},{"why":"Alpha-Group RT simulations provide the visual eduction (density and velocity maps) that motivates filtering vertical velocity and the qualitative persistence arguments.","marker":"Dimonte et al. (2004)"},{"why":"The threshold-selection method generalized in equation (5.7) and used to produce the two-structure fields from the filtered velocity.","marker":"Otsu (1979)"},{"why":"Gives the convection-diffusion-relaxation filtering rationale (fading memory along material paths) that underlies the persistence-preserving filter.","marker":"Lumley (1992)"},{"why":"High-resolution RT simulation used as the reference for the surrogate two-fluid estimate of the directed-energy ratio and other bulk quantities in table 1.","marker":"Soulard et al. (2016)"},{"why":"The buoyancy-drag Lagrangian whose kinetic term L(L')^2 is identified with directed energy, connecting the two-structure analysis to the growth-rate model.","marker":"Ramshaw (1998)"}],"fun_headline_variants":["Bimodal filter reveals two structures in Rayleigh-Taylor mixing","Two-structure decomposition of RTI turbulence via filtered velocity","First calibration data for two-structure RANS from conditional averages","Directed energy hits 40% in bimodal Rayleigh-Taylor structures","Filtered velocity splits RTI mixing into two persistent structures"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire dataset stands on the definition of the two structure fields: thresholding a filtered vertical velocity with hand-tuned parameters that the paper itself calls a somewhat ad hoc recipe, so if those fields do not correspond to real persistent physical structures, the conditional averages and the 40 percent directed-energy share are artifacts of the segmentation.","fun_headline_variants_meta":{"raw":{"variants":["Bimodal filter reveals two structures in Rayleigh-Taylor mixing","Two-structure decomposition of RTI turbulence via filtered velocity","First calibration data for two-structure RANS from conditional averages","Directed energy hits 40% in bimodal Rayleigh-Taylor structures","Filtered velocity splits RTI mixing into two persistent structures"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000954,"raw_usage":{"total_tokens":4132,"prompt_tokens":1073,"completion_tokens":3059,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":689,"completion_tokens_details":{"reasoning_tokens":2974}},"tokens_in":689,"tokens_out":3059,"duration_ms":21730,"temperature":1.0,"reasoning_tokens":2974,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:06:36.058981+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same two-structure segmentation on a simulation at higher Atwood number or with time-dependent acceleration and check whether the optimized coefficients and the measured directed-energy ratio stay in the same range; alternatively, compare the b^+ and b^- boundaries against an independent structure-identification method, such as Lagrangian coherent structures or bubble-tracking, in the same DNS and see whether the conditional statistics survive an independent structure definition.","supporting_citations":[],"review_version":1}