{"id":"5d242320-5b8e-4ee5-b174-07344730e883","arxiv_id":"2506.08542","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A thin-domain asymptotic analysis of compressible two-phase Navier-Stokes with Navier interfacial slip yields a closed two-velocity one-pressure averaged model with drag and wall-friction source terms.","lead":"The authors derive averaged two-velocity, one-pressure models for compressible two-phase stratified flows by applying a thin-domain asymptotic expansion to the Navier-Stokes equations with a slip-and-friction interface condition. The result gives explicit source terms and error estimates that were previously obtained by heuristic closure arguments.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3.2's error estimates rely on unstated uniform-in-ε bounds for the strong solutions; without them the Landau O-terms are not justified and the closed model is only formal.","rationale":"The paper's objective is to derive two-velocity, one-pressure averaged models from a thin-domain two-phase Navier-Stokes system via asymptotic expansion. The central claim, Theorem 3.2, is conditional: given a strong solution and the scaling (3.3), the averaged variables satisfy the closed system up to stated error orders. The derivation is internally coherent: the averaging steps, interface-condition manipulations, and scaling constraints lead to equations (3.4)-(3.8), and the same structure extends naturally to the Navier-Stokes-Fourier case in Section 4.\n\nThe most load-bearing weakness I see is not the explicit stratification ansatz, which is a clearly stated modeling restriction, but the implicit assumption that the family of strong solutions admits uniform-in-ε bounds. Every Landau estimate in the proof has a constant controlled by solution derivatives such as ∂_zz v, ∂_t v, ∇p, and ∇v. If those derivatives are not bounded independently of ε, the error terms ε^τ and ε^{2−τ} may be multiplied by constants that blow up, and the closed system is not an approximation in any quantitative sense. The theorem's hypothesis 'strong solution' does not supply such bounds, and the paper does not derive them from the scaled equations. This is fixable by adding explicit regularity and uniform-bound hypotheses, but as written it is a missing load-bearing condition.\n\nThe reader's rationale does mention the need for 'regularity and uniform-bound hypotheses needed for the Landau error estimates,' while the reader's named weakest assumption is the persistent-stratification ansatz. I partially agree with the reader: the stratification issue is a limitation of scope rather than an internal flaw, whereas the uniform-bounds issue directly affects the validity of the stated error estimates. A concrete check is to redo Proposition 3.7 with explicit constants and see which norms need to be controlled; if they are not controlled by the scaled equations, the theorem needs an extra hypothesis.\n\nOverall the derivation is promising and the final models are plausible. The verdict should remain conditional: the statement should be tightened by adding the missing uniform-bounds hypotheses, but the argument does not contain a demonstrated fatal error.","tokens_in":21398,"tokens_out":20429,"duration_ms":239479,"concrete_test":"Re-derive Proposition 3.7 tracking every O-constant in terms of explicit Sobolev norms of the strong solution, e.g. ||v||_{W^{2,∞}}, ||∂_t v||_{L∞}, and ||∇p||_{L∞}. Then check whether the rescaled equations provide ε-independent bounds for these norms under the scaling (3.3). If any norm can grow faster than ε^{−(2−τ)}, the claimed error order fails; if the norms are bounded, add that as an explicit hypothesis to Theorems 3.2 and 4.1 and recompute the error terms.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim asserts that averaged quantities satisfy the closed system up to explicit error orders, e.g. ε^τ + ε^{2−τ} in Theorem 3.2. The proof obtains these orders via Landau estimates such as Lemma 3.8, v_h(z) = ⟨v_h⟩ + O((κε+ε²)/µ), and ∂_z p = O(µ), which are derived from equations like µ ∂_zz v_h = O(ε²). The constants in these O-terms are controlled by sup norms of ∂_t v, ∇v, ∇²v, ∇p, and similar. The theorem only assumes the microscopic system admits a strong solution for each ε, with no hypothesis that these norms are bounded independently of ε. Since the solution itself depends on ε through the rescaled coefficients, these norms can in principle grow, e.g. like ε^{−N}, and the claimed errors ε^τ and ε^{2−τ} would then not be small. The same issue affects the energy estimates in Section 4, where (4.28) and (4.30) additionally require uniform control of ∂_z ∇_h θ. This is a missing hypothesis rather than a demonstrated contradiction, but it is load-bearing: Theorems 3.2 and 4.1 as stated are not rigorous unless uniform-in-ε bounds are imposed. The stratification ansatz flagged by the reader is real but explicit; the uniform-bounds gap is implicit and affects the mathematical content of every theorem.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript derives two-dimensional averaged models for compressible two-phase stratified flows by performing a thin-domain asymptotic reduction of a three-dimensional two-layer Navier-Stokes system. The barotropic case yields the two-velocity, one-pressure system (3.4)-(3.8), claimed to hold up to explicit error orders under the scaling (3.3); the Navier-Stokes-Fourier case yields the two-velocity, one-pressure, two-temperature system (4.13)-(4.19) under the additional scaling (4.11) and structural assumption (4.12). The derivation proceeds by rescaling the microscopic equations, simplifying the vertical momentum equation and interface conditions, averaging in the vertical variable, and closing the resulting boundary terms through Landau-type estimates such as (3.16)-(3.17).","tokens_in":21781,"tokens_out":6142,"duration_ms":79123,"significance":"If the claims are taken as a formal asymptotic derivation, the paper gives a transparent route to two-velocity, one-pressure averaged models with explicit source terms, complementing earlier one-dimensional homogenization results that produce one-velocity models. The averaging is carried out in a multidimensional horizontal setting, the averaged mass equations are exact, and the error analysis is presented in enough detail to be checked. The main value is therefore in the derivation itself and in the explicit scaling conditions under which the interaction terms appear. The significance is conditional, however, because the theorem-level statements currently rely on unstated uniform-in-epsilon bounds on the microscopic solutions.","major_comments":[{"comment":"The error estimates (3.16)-(3.17) are Landau O-terms whose constants depend on sup norms of time and space derivatives of the velocity and pressure, such as ∂_t v, ∇v, ∇²v and ∇p. The theorem only assumes that a strong solution exists for each epsilon and imposes no uniform-in-epsilon bound on these norms. Since the solution itself depends on epsilon through the rescaled coefficients, these norms can in principle grow like epsilon^{-N}, in which case the claimed errors ε^τ + ε^{2-τ} and ε^τ do not follow. The same gap affects the energy estimates in Section 4. This is repairable by adding explicit hypotheses such as uniform-in-epsilon bounds on the relevant solution norms, but without such hypotheses Theorems 3.2 and 4.1 are not justified as stated.","section":"§3.2.2, Lemma 3.8 and Theorem 3.2"},{"comment":"The estimate ∂_z ∇_h θ_k = O(ε²/β) in assumption (4.12) is not derived from the energy equation; it is an additional structural condition on the microscopic solution, and it is load-bearing for the energy closure. Lemma 4.6 and the displayed error order ε^{2-γ} in (4.15) collapse without it. The paper should either prove (4.12) from the equation and uniform bounds, or state explicitly in Theorem 4.1 that the result is conditional on this solution-dependent assumption. As written, the distinction between assumptions on the constant coefficients in (3.3) and (4.11) and this assumption on the solution itself is not made prominent enough.","section":"§4.3.1, eq. (4.12) and Lemma 4.6"},{"comment":"The graph parametrization of the interface, α₁(t,x) ∈ (0,1), is assumed for all times, and every vertical integration in Theorems 3.2 and 4.1 depends on it. The microscopic model contains no gravity or surface tension, so the equations provide no mechanism preventing shear instabilities, overturning, or interface folding. If the interface ceases to be a graph, the averaging procedure and the resulting asymptotic model break down. The authors acknowledge this in Remark 2.1, but the final theorems should list this condition as an explicit hypothesis, and the abstract or introduction should describe the result as conditional on persistence of stratification.","section":"§2.1, eq. (2.2) and Remark 2.1"}],"minor_comments":[{"comment":"The running title on page 1 reads 'A VERAGED MODELS ...', and the abstract contains the typo 'two-ph ase'; both should be corrected.","section":"Abstract and running title"},{"comment":"The averaged energy equation uses ⟨E_{k,h}⟩, but E_{k,h} is only defined later in §4.3.1; the definition should be moved before the theorem statement.","section":"§4.2, Theorem 4.1"},{"comment":"The remark says the asymptotic limit corresponds to 'moderate Mach numbers and high Reynolds numbers', but no Mach number appears in the dimensionless equations; this statement should be made precise or removed.","section":"Remark 3.5"},{"comment":"The constraints on τ, ξ, ζ obtained from the remainder R are listed in prose; displaying them as a set of inequalities before the final scaling (3.3) would improve readability.","section":"§3.2.3"}],"recommendation":"major_revision","confidential_remarks":"The paper is a substantial formal-asymptotics contribution, and the main gap—missing uniform-in-epsilon bounds—is fixable by adding explicit hypotheses. The stratification assumption is acknowledged by the authors, so I would not treat it as grounds for rejection, but the theorem statements should be tightened so that the displayed error orders are controlled by stated norms. The paper fits a math.AP journal if the authors clearly frame the results as conditional asymptotic statements."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new thing here is the use of Navier-type interfacial slip instead of velocity continuity. With the chosen scalings of viscosity, wall friction, and interfacial friction, vertical averaging closes at the level of a two-velocity, one-pressure system with explicit error orders and source terms. The 1980s averaging papers relied on heuristic interface closures, and the homogenization route produced one-velocity models, so this gives a real alternative route to the two-velocity equations used in engineering.\n\nWhat the paper does well is the algebra. The averaged mass equations are exact, the momentum balance is tracked with Landau remainders, and the scalings in (3.3) are not guessed but derived from the requirement that the remainders vanish and the friction terms stay finite. The recovery of the one-velocity model as the interfacial friction limit is a nice check. The Navier-Stokes-Fourier extension follows the same pattern and gives a two-temperature averaged model, which is a useful addition.\n\nThe main soft spot is the one the stress-test note identifies. Theorem 3.2 assumes only that the microscopic system admits a strong solution for each epsilon, but the error estimates in Lemma 3.8 and elsewhere are controlled by sup norms of derivatives that could, in principle, grow like negative powers of epsilon. Without uniform-in-epsilon bounds, the claims of order epsilon^tau + epsilon^(2-tau) are not justified. This is a missing hypothesis rather than a demonstrated contradiction, but it is load-bearing: both Theorem 3.2 and Theorem 4.1 are formal as written. The same issue appears in Section 4, where the assumption on ∂_z ∇_h θ_k is an extra regularity hypothesis that also needs a uniform bound.\n\nThe stratification ansatz is real but explicit. The interface is assumed to stay a graph for all time, and the model has no gravity or surface tension to enforce that. The authors acknowledge this in Remark 2.1, so I read it as a stated limitation rather than a hidden flaw, though it does restrict the physical regime to flows that do not overturn.\n\nCitations look appropriate, and the self-citations are background rather than load-bearing. The paper is coherent, honest, and useful for people working on two-phase averaged models or thin-film asymptotics. It deserves a serious referee. I would send it out and ask the authors either to impose uniform-in-epsilon bounds or to label the derivation formal throughout.","headline":"A careful formal asymptotic derivation of two-velocity one-pressure models, worth sending to referees, but the theorems as stated need uniform-in-epsilon bounds that are never assumed.","tokens_in":22231,"tokens_out":2704,"would_cite":true,"duration_ms":35334,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["76T10","35Q30","35C20","76A20"],"pacs":[],"model":"deepseek-v4-flash","headline":"Vertical averaging of a thin two-layer Navier-Stokes system, with slip at the interface, produces closed two-velocity one-pressure two-phase flow models, with explicit error orders.","keywords":["compressible two-phase flow","stratified flow","thin domain asymptotic","vertical averaging","Navier-Stokes","two-velocity one-pressure model","Navier interface condition","Navier-Stokes-Fourier"],"falsifier":"Direct numerical simulation of the 3D two-phase Navier-Stokes system (2.4)-(2.13) with the scaling (3.3) and an initially flat interface should show the vertical averages satisfying (3.4)-(3.8) with residuals shrinking at the stated orders; if instead the residuals fail to vanish or the interface leaves the graph regime ($\\alpha_1$ leaves $(0,1)$) while $\\varepsilon$ is small, the central claim is refuted.","tokens_in":21252,"feed_emoji":"🌊","tokens_out":11608,"duration_ms":122236,"temperature":0.7,"pith_summary":"The paper derives averaged equations for compressible two-phase stratified flows from a microscopic three-dimensional description, rather than postulating them from closure heuristics. It considers two immiscible compressible fluids stacked in a thin domain of thickness ratio $\\varepsilon = D/L \\ll 1$, each governed by Navier-Stokes equations. The essential choice is to allow tangential slip between the fluids through a Navier friction condition at the interface and to scale viscosity and friction coefficients as powers of $\\varepsilon$. After vertical averaging, the paper obtains closed two-dimensional systems with two velocities and one pressure in the barotropic case, and two velocities, one pressure, and two temperatures in the full Navier-Stokes-Fourier case, with theorems stating the approximation errors. This matters because two-velocity models of this type were previously obtained by ad hoc interfacial closures, while here they emerge as an asymptotic limit of the three-dimensional equations.","feed_headline":"Thin-slab averaging yields closed two-velocity two-phase models","feed_subtitle":"Interfacial slip plus tuned friction makes two-velocity, one-pressure models the limit of 3D Navier-Stokes.","key_machinery":"The load-bearing machinery is the vertical average over each phase layer: the arithmetic average (3.1) for densities and pressures, and the density-weighted average (3.2) for velocities and energies. Closure comes from three estimates: the rescaled vertical momentum equation makes the pressure nearly independent of the vertical coordinate, so interface pressures can be replaced by any convex combination $p_i$ of the averaged pressures; the rescaled Navier interface condition converts the vertical flux of horizontal momentum into the drag source term $\\hat\\kappa_i(\\langle v_{1,h}\\rangle-\\langle v_{2,h}\\rangle)$; and Taylor expansions show the horizontal velocity of each phase is nearly constant across its layer, so products of velocities close as products of their averages. The scaling conditions (3.3) are what decide which terms survive as $\\varepsilon\\to 0$; varying the interface friction exponent is what interpolates between the two-velocity and one-velocity averaged models.","core_discovery":"On the paper's own terms, the central discovery is that the two-velocity, one-pressure system (1.1)-(1.4) is the leading-order vertical average of a thin two-layer compressible Navier-Stokes model, provided tangential slip is allowed at the interface and the coefficients are scaled as $\\mu_k=\\hat\\mu_k\\varepsilon^\\tau$ with $0<\\tau<2$, $\\kappa_i=\\hat\\kappa_i\\varepsilon$, and $\\kappa_k=\\hat\\kappa_k\\varepsilon^\\xi$ with $\\xi\\ge 1$ (Theorem 3.2). Under these scalings, the averaged densities, horizontal velocities, and pressures satisfy (3.4)-(3.8) up to errors of order $\\varepsilon^\\tau+\\varepsilon^{2-\\tau}+(1-\\delta_\\xi)\\varepsilon^{\\xi-1}$, where $\\delta_\\xi$ is 1 when $\\xi=1$ and 0 otherwise, and the equations of state and equality of pressures hold up to order $\\varepsilon^\\tau$. If the tangential interface condition is instead the no-slip Dirichlet condition, the two averaged velocities coincide up to order $\\varepsilon^{2-\\tau}$ and the one-velocity hyperbolic model of Theorem 3.13 is recovered. In the non-barotropic case, the same averaging applied to the Navier-Stokes-Fourier system, with thermal conductivities $\\beta_k=\\hat\\beta_k\\varepsilon^\\gamma$ ($0\\le\\gamma<2$) and contact conductance $h_c/\\varepsilon$, yields the closed two-velocity, one-pressure, two-temperature system (4.13)-(4.19) with the stated error orders (Theorem 4.1).","pith_inferences":["If gravity and surface tension were added to the microscopic model, the persistent-stratification assumption could become a consequence of the dynamics rather than a standing ansatz; the averaging machinery would likely survive unchanged, but this extension is not in the paper.","The error orders mix the viscosity exponent $\\tau$ with the friction exponents $\\xi$ and $\\gamma$, which suggests a numerical or experimental program that tunes these coefficients could identify which asymptotic regime real pipe flows occupy from measured slip lengths.","The same layer-by-layer averaging route could be applied to three or more stacked fluids or to stratified flows with phase change, since the interface conditions are local; those cases are natural but untested extensions.","Because the no-slip limit collapses the model to one velocity, the paper implies that the existence of a two-velocity regime is tied to measurable interfacial slip, so experiments or simulations that suppress slip should see the one-velocity model instead."],"forward_implications":["The two-velocity, one-pressure model is not an ad hoc closure: it is the $\\varepsilon\\to 0$ limit of the 3D Navier-Stokes system with Navier interfacial slip and the scaling (3.3), with explicit error bounds.","Changing the scaling of the interface friction changes the target model: friction of order $\\varepsilon$ gives two velocities, while slower-decaying friction (exponent $\\zeta<1$) forces $\\langle v_{1,h}\\rangle=\\langle v_{2,h}\\rangle$ and recovers the one-velocity hyperbolic model of Theorem 3.13.","The interfacial drag and wall friction source terms in the averaged momentum equations come directly from the microscopic boundary conditions, so no separate phenomenological closure is needed for them.","In the Navier-Stokes-Fourier case the averaged model has two temperatures with a heat-exchange term $\\hat h_c(\\theta_1-\\theta_2)$, and thermal conduction enters only when $\\gamma=0$; otherwise heat diffusion is a higher-order effect.","Because the derivation is carried out in a multidimensional horizontal domain, it covers averaged models in two dimensions, a setting that the existing one-dimensional homogenization derivations do not reach."],"supporting_citations":[{"why":"The classical averaging framework for two-phase flows that this paper revisits; its closure problem motivates the asymptotic derivation.","marker":"[20]"},{"why":"Earlier two-pressure averaging model that supplies the two-velocity model class and the heuristic interface closure this paper replaces.","marker":"[17]"},{"why":"The 1988 follow-up cited in the abstract; the other baseline averaging derivation for two-phase flow.","marker":"[18]"},{"why":"Rigorous homogenization derivation of a multi-fluid system that produces only one velocity, the limitation this paper overcomes by allowing slip.","marker":"[7]"},{"why":"Stability analysis showing standard two-fluid models have complex eigenvalues; this motivates studying two-velocity model classes.","marker":"[16]"},{"why":"Well-posedness properties of two-velocity two-pressure models, the class of averaged systems obtained here.","marker":"[8]"}],"fun_headline_variants":["Thin-slab averaging yields two-velocity one-pressure models","Slip at interface yields two-velocity two-phase models","Thin-domain asymptotics derive two-velocity one-pressure systems","Averaging thin films of two-phase flows yields two-fluid models","Two-velocity one-pressure models from thin-layer limit"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that the two fluids remain perfectly stratified forever: the interface must stay a graph $z = D\\alpha_1(t,x)$ with $0<\\alpha_1<1$, and the microscopic model contains no gravity or surface tension that would enforce this; if shear makes the interface overturn, the vertical averaging and the whole asymptotic model break down.","fun_headline_variants_meta":{"raw":{"variants":["Thin-slab averaging yields two-velocity one-pressure models","Slip at interface yields two-velocity two-phase models","Thin-domain asymptotics derive two-velocity one-pressure systems","Averaging thin films of two-phase flows yields two-fluid models","Two-velocity one-pressure models from thin-layer limit"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00068,"raw_usage":{"total_tokens":3129,"prompt_tokens":1022,"completion_tokens":2107,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":638,"completion_tokens_details":{"reasoning_tokens":2023}},"tokens_in":638,"tokens_out":2107,"duration_ms":21341,"temperature":1.0,"reasoning_tokens":2023,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T05:08:16.123547+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Direct numerical simulation of the 3D two-phase Navier-Stokes system (2.4)-(2.13) with the scaling (3.3) and an initially flat interface should show the vertical averages satisfying (3.4)-(3.8) with residuals shrinking at the stated orders; if instead the residuals fail to vanish or the interface leaves the graph regime ($\\alpha_1$ leaves $(0,1)$) while $\\varepsilon$ is small, the central claim is refuted.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The classical averaging framework for two-phase flows that this paper revisits; its closure problem motivates the asymptotic derivation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Earlier two-pressure averaging model that supplies the two-velocity model class and the heuristic interface closure this paper replaces."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The 1988 follow-up cited in the abstract; the other baseline averaging derivation for two-phase flow."},{"cited_title":"Bresch and X.-D","cited_arxiv_id":null,"evidence_quote":"Rigorous homogenization derivation of a multi-fluid system that produces only one velocity, the limitation this paper overcomes by allowing slip."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Stability analysis showing standard two-fluid models have complex eigenvalues; this motivates studying two-velocity model classes."},{"cited_title":"Coquel, J.-M","cited_arxiv_id":null,"evidence_quote":"Well-posedness properties of two-velocity two-pressure models, the class of averaged systems obtained here."}],"review_version":1}