{"id":"ddd746e2-8e80-4885-8366-fc5691fac913","arxiv_id":"2412.13985","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":0.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A comprehensive review of RANS turbulence models for hypersonic flows concludes that standard models, and even advanced Reynolds stress closures, routinely mispredict cold-wall heat transfer and shock-induced separation.","lead":"This review maps how well current RANS turbulence models predict hypersonic boundary layers and shock interactions, finding systematic overprediction of wall heat flux and unreliable separation sizes. A generalist should read it because hypersonic vehicle design still depends on RANS, so knowing which corrections and datasets are trustworthy matters for thermal protection and flight safety.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central claim overstates consistency: Table IX shows BL and SA underpredict peak heat flux, contradicting 'consistently overestimate' in Section VII A.","rationale":"Read in good faith: the paper is a comprehensive review of RANS turbulence modeling for hypersonic flows, and its main contribution is a synthesis and tabulated assessment of model performance. The central claim is intended to summarize model capabilities. For that claim to hold, the evidence must show systematic overprediction of heat flux across the model landscape. The paper's own Table IX does not show this: the BL and SA entries are underpredictions. This is an internal inconsistency, not a disagreement with external consensus; it does not require independent data, only a careful reading of Table IX and Section V C. The reader's weakest assumption (commensurability of error percentages across codes, grids, and experiments) is also valid and is acknowledged by the paper itself in Section VI B, but my concern is more direct: even taking the table at face value, the conclusion overreaches. The review remains useful as a qualitative guide, and the suggestion that heat-flux closures need work is well supported by the large overpredictions of most two-equation models. However, the conclusion as written is inaccurate and should be qualified. A conditional acceptance requiring revision of Section VII A and a clarifying caveat for Table IX is appropriate. I agree partially with the reader: the table's methodology needs clarification, but the more serious issue is that the stated conclusion contradicts the table's own entries.","tokens_in":53873,"tokens_out":4998,"duration_ms":44537,"concrete_test":"Construct a signed-error table from Table IX's Qwp column (positive = overprediction, negative = underprediction). Since BL (−10%) and SA (−25%) are negative while all two-equation models are positive, the 'consistently overestimate' statement is false as stated. To settle definitively, recompute the sign distribution and verify against Section V C; if the negative entries persist, the conclusion must be revised to 'two-equation models generally overpredict, while zero- and one-equation models underpredict.' Additionally, run a code-to-code comparison of a canonical SBLI (e.g., Schulein Mach 5, 14-degree generator) on identical grids using the same model versions to assess whether the Table IX spread exceeds the code variability acknowledged in Section VI B.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper's central conclusion (Section VII A) states that 'current turbulence models, while adequately predicting wall pressure distributions and separation zones, consistently overestimate heat flux at the wall surface.' This is not supported by the paper's own compilation. Table IX reports typical prediction errors for fully separated SBLIs: the Baldwin–Lomax model underpredicts peak wall heat transfer by roughly 10% and the Spalart–Allmaras model underpredicts it by roughly 25%, while the two-equation models (k-ε, k-ω, SST, k-ζ, etc.) overpredict Qwp by 50–250%. Section V C also explicitly states that the SA model 'does not result in the same type of over-predictions in wall heat flux in SBLIs as two-equation models' and instead underpredicts wall shear stress. Thus the universal 'consistently overestimate' assertion is contradicted by the data the review itself presents. The defensible claim is that two-equation eddy-viscosity models generally overpredict peak heat flux, while zero- and one-equation models tend to underpredict it. The same overgeneralization affects the 'adequately predicting separation zones' part: SA fails to predict separation (Section V C) and the collaborative flared-cone exercise (Ref. 153) showed large variation in separation onset. The conclusion therefore needs qualification; as written, it misrepresents the evidence.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This review paper surveys RANS turbulence modeling for hypersonic wall-bounded flows. It covers the Favre-averaged governing equations, eddy-viscosity closures of different levels (zero-, one-, two-equation, NLEVM/EARSM), Reynolds stress transport models, compressibility corrections, and modeling challenges specific to cold-wall boundary layers, shock/boundary-layer interactions, surface roughness and blowing, and thermochemical nonequilibrium. It also compiles experimental and DNS validation datasets, assesses the performance of fifteen turbulence models against SBLI cases (Tables VIII and IX), discusses compressibility corrections (Table X), and reviews data-driven and machine-learning approaches for closure improvement. The paper's main conclusion in Section VII A is that current turbulence models adequately predict wall pressure and separation zones but consistently overestimate wall heat flux, motivating a call for improved turbulent heat-flux closures.","tokens_in":54114,"tokens_out":3438,"duration_ms":32008,"significance":"The review is timely and comprehensive, providing a well-organized compilation of hypersonic validation datasets (Tables I–VII), a systematic classification of model families and corrections, and an up-to-date discussion of data-driven closure modeling. Its qualitative claim that standard two-equation eddy-viscosity models tend to overpredict peak heat flux in SBLIs is supported by many independent studies. The paper also honestly notes the scarcity of high-enthalpy data and code-to-code variability. However, the central conclusion as worded overgeneralizes the evidence, and the quantitative error table lacks a stated aggregation protocol; these issues affect the paper's main takeaway but are correctable within the scope of a review article.","major_comments":[{"comment":"The claim that 'current turbulence models ... consistently overestimate heat flux at the wall surface' is not supported by the paper's own compilation. Table IX reports that the Baldwin–Lomax model underpredicts peak wall heat transfer by roughly 10% and the Spalart–Allmaras model underpredicts it by roughly 25%, while Section V C explicitly states that SA 'does not result in the same type of over-predictions in wall heat flux in SBLIs as two-equation models.' The evidence supports the more defensible conclusion that two-equation eddy-viscosity models generally overpredict peak heat flux, whereas zero- and one-equation models tend to underpredict it, with some RSTM variants (e.g., SSG/LRR-ω) reporting good agreement. The sentence should be qualified accordingly. The companion claim that separation zones are 'adequately predicted' is also too strong given SA's failure to predict separation (Section V C) and the collaborative flared-cone exercise (Ref. 153) showing large variation in separation onset across codes.","section":"Section VII A; Table IX"},{"comment":"The quantitative error percentages in Table IX (and the intended performance ordering) lack a stated aggregation protocol. It is unclear how the 'typical' errors were computed across different experimental campaigns, CFD codes, grids, and model variants, and the paper itself concedes in Section VI B that 'simulation results may vary as a result of different implementations of the turbulence model within a code or between codes' and that multiple machine-zero converged solutions can arise from different initial guesses. As a result, the specific error magnitudes (e.g., SST 200% overprediction of Qwp) are not a controlled comparison, and the model rankings based on these numbers may not be robust. The authors should either state how the errors were aggregated (e.g., median or range across cases) or present them as qualitative trends rather than precise percentages.","section":"Table IX; Section VI B"}],"minor_comments":[{"comment":"The text refers to 'Bardin-Lomax' model; this should be 'Baldwin-Lomax'.","section":"Section VI C"},{"comment":"The name 'Sakar' appears twice; it should be 'Sarkar' (correctly spelled elsewhere in the paper).","section":"Section IV A"},{"comment":"The name 'Catrix' should be 'Catris' to match the cited reference (Catris and Aupoix).","section":"Section IV G"},{"comment":"The text around Eq. (9) contains garbled symbols in the provided manuscript; the authors should ensure the final typeset version displays the Reynolds stress decomposition cleanly.","section":"Section III"},{"comment":"The phrase 'blue underlined terms' appears in Section II but the manuscript does not use color/underline in the displayed equations; rephrase to 'unclosed terms' or use typographic emphasis consistently.","section":"Abstract and Section I"}],"recommendation":"major_revision","confidential_remarks":"This is a substantial review with a useful compilation of datasets and model assessments. The main revision needed is to qualify the headline conclusion and to add transparency to the aggregation of the error percentages in Table IX; once these are addressed, the paper should be suitable for publication. The authors' own SUQ-SST model is discussed fairly, including its overprediction of heat transfer, so self-citation is not a concern here."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a useful and mostly careful review, but the central conclusion is not supported by the paper's own table. The body is more nuanced than the summary. Section V C correctly notes that two-equation models tend to overpredict peak heat flux in SBLIs, while the SA model does not show the same behavior and instead underpredicts wall shear stress. The summary in Section VII A says current models 'consistently overestimate heat flux at the wall surface.' Table IX contradicts that: BL underpredicts peak Q_wp by ~10%, SA by ~25%, while most two-equation models overpredict by 50–250%. The same overgeneralization affects 'adequately predicting separation zones' — SA fails to predict separation in the Mach 8 ramp case, and the collaborative flared-cone exercise showed large variation in separation onset. The paper will be cited for that conclusion, so the authors should align it with their own evidence.\n\nThe real value is the compilation. Tables I–VII bring together the scattered experimental and DNS validation data for ZPG TBLs, 2D/axisymmetric/3D SBLIs, and rough walls; that saves a newcomer days of literature chasing. The coverage of model classes — zero- and one-equation, two-equation linear and nonlinear, RSTMs, compressibility corrections, and data-driven approaches — is broad, and the paper is honest about the weaknesses of existing models, including the authors' own SUQ-SST overpredicting heat transfer. The sections on roughness/blowing and thermochemical non-equilibrium are welcome additions beyond older reviews.\n\nSoft spots, in proportion: the quantitative error numbers in Table IX lack a stated aggregation protocol, and the paper itself concedes code-to-code variability and even multiple machine-zero converged solutions on the same grid. That means the precise percentages are not a controlled comparison. I would treat them as indicative, not definitive. Also, Eq. (9) in the manuscript appears garbled — presumably a rendering error, but it must be fixed. Both are fixable, not fatal.\n\nWho is this for? Anyone building or validating RANS models for hypersonic SBLIs, and engineers who need a map of the available validation datasets. It deserves a serious referee. I would send it to peer review and ask for moderate revision: rewrite Section VII A to match the body, and document how the typical errors in Table IX were computed. With those changes, it becomes a solid reference.","headline":"A genuinely useful review of hypersonic RANS turbulence modeling, but the headline conclusion overstates the evidence: Table IX itself shows BL and SA underpredict peak heat flux, so the blanket 'consistently overestimate' claim needs qualification.","tokens_in":54612,"tokens_out":2747,"would_cite":true,"duration_ms":27770,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["76F40","76F50","76K30"],"pacs":[],"model":"deepseek-v4-flash","headline":"This review finds that current RANS turbulence models consistently overestimate wall heat flux in hypersonic flows, even while wall pressure and separation zones are predicted adequately, and argues that improved turbulent heat-flux…","keywords":["hypersonic flows","turbulence modeling","RANS","shock/boundary layer interaction","wall heat flux","turbulent Prandtl number","eddy viscosity models","validation datasets"],"falsifier":"A controlled comparison of several RANS models on one cold-wall hypersonic SBLI case, using identical grids, boundary conditions, and solver settings across codes, would reveal whether heat-flux overprediction is a stable model feature or dominated by implementation differences; if the spread across implementations matches or exceeds the spread across models, the ranking-based conclusion is not settled.","tokens_in":53682,"feed_emoji":"🔥","tokens_out":5619,"duration_ms":50973,"temperature":0.7,"pith_summary":"This review surveys how Reynolds-averaged Navier-Stokes (RANS) turbulence models perform for hypersonic boundary layers and shock/boundary-layer interactions. Its central conclusion is that current models predict wall pressure distributions and separation zones reasonably well, but consistently overestimate wall heat flux at cold walls and in reattachment regions. The authors trace this to the way turbulent heat flux is closed: almost always a gradient-diffusion law with a constant turbulent Prandtl number ($\\mathrm{Pr}_t=0.9$), while DNS data show $\\mathrm{Pr}_t$ varies across the boundary layer and downstream of shocks. A sympathetic reader would come away with the thesis that the next gain in hypersonic aerothermal prediction will come less from refining Reynolds-stress closures and more from building credible turbulent heat-flux models and the validation data to support them. If true, this reframes where modeling effort should be spent in hypersonic vehicle design.","feed_headline":"RANS models overpredict heat flux on hypersonic walls, review finds","feed_subtitle":"Pressure and separation look right, but thermal loads are consistently too high — pointing to better heat-flux closures.","key_machinery":"The central object is the Favre-averaged RANS system for compressible flow, in which two unclosed terms carry the modeling load: the Reynolds-stress tensor $\\tau_{ij}$ (for momentum and separation) and the turbulent heat flux $q_j^T$ (for wall heating). The review's argument runs through the closure hierarchy for $\\tau_{ij}$—linear Boussinesq eddy viscosity, non-linear/explicit algebraic stress, and full Reynolds-stress transport—and contrasts it with the thin treatment of $q_j^T$, which in most models is $q_j^T=-\\mu_T C_p \\mathrm{Pr}_t^{-1}\\,\\partial\\tilde{T}/\\partial x_j$ with a single constant $\\mathrm{Pr}_t\\approx0.9$. The key mechanism identified is the near-universal reliance on this constant-$\\mathrm{Pr}_t$ gradient-diffusion law, together with length-scale errors in the $\\varepsilon$- or $\\omega$-equation, which the review shows is the reason heat-flux overprediction persists even when Reynolds-stress closures are upgraded.","core_discovery":"Across fifteen models—algebraic, one- and two-equation eddy-viscosity models, non-linear and explicit algebraic Reynolds-stress models, and Reynolds-stress transport models—the paper finds a common failure pattern for fully separated hypersonic shock/boundary-layer interactions: separation length and wall pressure are broadly captured (or under-predicted by standard two-equation models), while peak wall heat flux near reattachment is over-predicted by tens to hundreds of percent (Table IX). The review interprets this as evidence that the dominant source of error has shifted to the turbulent heat-flux closure, not the Reynolds-stress closure, and that the constant $\\mathrm{Pr}_t$ gradient-diffusion model is the weakest link. It also notes that Reynolds-stress transport models have not yet delivered consistent accuracy gains over eddy-viscosity models for these flows, that length-scale corrections and variable-$\\mathrm{Pr}_t$ models show the most promise, and that the scarcity of high-enthalpy and rough-wall data makes decisive validation impossible.","pith_inferences":["If the constant-$\\mathrm{Pr}_t$ assumption is the dominant cause of heat-flux overprediction, then data-driven or physics-based variable-$\\mathrm{Pr}_t$ closures may yield larger gains in wall-heating accuracy than any further refinement of Reynolds-stress anisotropy, which the review finds to be of limited payoff.","The review's model rankings are not a controlled comparison; the authors themselves note implementation and initial-guess variability, so a reader should treat the reported percentages as indicative rather than as an ordering of model quality.","A testable corollary is that the same turbulence model with a variable $\\mathrm{Pr}_t$ should reduce heat-flux error while leaving pressure and separation almost unchanged, which would isolate the heat-flux closure as the causal factor.","The overprediction pattern suggests that turbulent-heating margins in hypersonic design may be conservative, but the underprediction of separation by standard models could offset that margin at off-design conditions."],"forward_implications":["Hypersonic vehicle thermal design using standard RANS will keep overpredicting peak heat loads at compression corners and shock-impingement points until the heat-flux closure is replaced or corrected.","Separation zones and pressure loads, being less sensitive to the heat-flux closure, can be trusted more than wall heat transfer in current engineering prediction practice.","Variable turbulent Prandtl-number models and length-scale limiters should be tested more widely, since they attack the heat-flux error without necessarily disturbing the separation prediction.","Reynolds-stress transport models will not resolve the heat-flux problem; Table IX shows even advanced closures inherit the overprediction because the length-scale equation and heat-flux law stay the same.","New validation data—especially DNS and experiments at high enthalpy and on rough or ablating walls—are needed before any single model can be declared reliable for aerothermal design."],"supporting_citations":[{"why":"Prior comprehensive review of turbulence models for hypersonic flows; establishes the long-standing assessment problem and the 18-model baseline this review extends.","marker":"[13]"},{"why":"Provides the extended hypersonic SBLI experimental database used to evaluate model performance in earlier assessments.","marker":"[15]"},{"why":"Schülein's experiments give skin friction and heat-flux measurements for impinging SBLIs used repeatedly as validation cases.","marker":"[82]"},{"why":"Introduces the shock-unsteadiness correction that damps excessive TKE production at shocks, a key improvement strategy discussed throughout.","marker":"[85]"},{"why":"Proposes the length-scale correction that limits turbulent length scale in reattachment regions, one of the main fixes for heat-flux overprediction.","marker":"[110]"},{"why":"Develops a variable turbulent Prandtl number model for hypersonic SBLIs, the leading alternative to the constant-$\\mathrm{Pr}_t$ law.","marker":"[128]"},{"why":"A priori tests against DNS show the constant turbulent Schmidt/Prandtl number assumptions fail in hypersonic boundary layers.","marker":"[9]"},{"why":"Cold-wall hypersonic TBL DNS used in Figure 3 to demonstrate the systematic overprediction of wall heat flux and shear stress by SST and SA models.","marker":"[120]"}],"fun_headline_variants":["Hypersonic RANS heat flux overpredicted by up to hundreds of percent","Review: Heat-flux closure, not Reynolds stress, limits hypersonic RANS","RANS models: separation right, but heat flux off by 100s%","Hypersonic RANS: pressure accurate, heat transfer far too high","Turbulent heat flux closure is bottleneck for hypersonic RANS"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The review's error rankings assume that reported errors from different experiments, CFD codes, grids, and model variants are directly comparable; the authors themselves concede that different implementations and initial guesses can change results.","fun_headline_variants_meta":{"raw":{"variants":["Hypersonic RANS heat flux overpredicted by up to hundreds of percent","Review: Heat-flux closure, not Reynolds stress, limits hypersonic RANS","RANS models: separation right, but heat flux off by 100s%","Hypersonic RANS: pressure accurate, heat transfer far too high","Turbulent heat flux closure is bottleneck for hypersonic RANS"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000262,"raw_usage":{"total_tokens":1612,"prompt_tokens":979,"completion_tokens":633,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":595,"completion_tokens_details":{"reasoning_tokens":530}},"tokens_in":595,"tokens_out":633,"duration_ms":6337,"temperature":1.0,"reasoning_tokens":530,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T12:35:17.039129+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A controlled comparison of several RANS models on one cold-wall hypersonic SBLI case, using identical grids, boundary conditions, and solver settings across codes, would reveal whether heat-flux overprediction is a stable model feature or dominated by implementation differences; if the spread across implementations matches or exceeds the spread across models, the ranking-based conclusion is not settled.","supporting_citations":[],"review_version":1}