{"id":"4932a3bc-7a99-41e2-9a97-71edc1a481c5","arxiv_id":"1908.02191","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Numerical eigenvalue analysis finds duct Poiseuille flow linearly stable up to Re=2500, with least-stable eigenmodes forming a wall boundary layer at high Re and a fitted critical perturbation amplitude scaling near Re^-0.6.","lead":"This paper computes the stability of three-dimensional Poiseuille flow in a finite-length square duct with a new finite-element eigenvalue method, finding that the flow is linearly stable for Reynolds numbers up to 2500. It also reports that the least-stable disturbance modes hug the duct wall at high Reynolds numbers, with large gradients that the authors argue make linear stability analysis unreliable.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Convergence analysis in Section 6 assumes smooth eigenmodes, so it cannot support the central claim of unbounded boundary-layer gradients; the finite-Re evidence for this singular structure is not resolution-tested.","rationale":"The reader identified the convergence analysis in Section 6 as the weakest assumption: it assumes smooth fields, while the computed high-Re eigenmodes are boundary-layer-like and the duct has sharp corners. My reading agrees and sharpens the point: the central new physical claim is not merely that the modes are boundary-layer-like, but that their gradient is unbounded as Re increases. That claim is used to explain pseudospectral sensitivity and to derive the critical-amplitude scaling in Section 4. The numerical support is a fixed-mesh sequence of peak positions in Figure 9, which does not establish an asymptotic trend. Moreover, Theorem 6.4 gives error bounds with constants depending on the unknown eigenmode; if the true eigenmode has unbounded gradient, the constants are not uniform and the error estimate cannot justify the computed mode shape or its gradient. Thus the load-bearing concern is that the paper's own numerical analysis does not cover the regime in which the new phenomenon is asserted. This does not undermine the more conventional result that duct Poiseuille flow is linearly stable up to Re = 2500, nor does it call into question the authors' good faith; it only means the singular-structure claim needs independent resolution study or a separate asymptotic argument. A conditional accept is appropriate, with the condition that the boundary-layer gradient claim be validated by mesh-refinement and, ideally, by a matched-asymptotic or other analysis that does not assume smoothness uniformly in Re.","tokens_in":16473,"tokens_out":3864,"duration_ms":44521,"concrete_test":"Reproduce the eigenvalue computation of Section 3 at Re = 2500 and Re = 5000 with the same SUPG/Q1 divergence-free scheme on two uniform meshes: Nx = 40/120 (refining in x, y, and z together) and Nx = 80/240. For the least-stable mode, report (i) the first eigenvalue lambda_1, (ii) sup_Omega |grad u_1| (or a mesh-convergent proxy), and (iii) the boundary-layer width, e.g. the x-coordinate of the peak of |w| at z = 2.9, as functions of h. If lambda_1 has converged but sup |grad u_1| changes by more than about 10% under refinement, or if the peak position does not move monotonically toward the wall with an h-independent trend, then the 'unbounded gradient' claim in the Abstract and Section 4 is not established. In addition, recompute the integrand in eq.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper's new assertion is that least-stable duct modes develop a wall boundary layer with unbounded gradient as Re grows, and that this structure drives pseudospectral sensitivity and early nonlinear effects (Abstract, Section 4). The numerical evidence is Figure 9, where at Re = 500, 1500, 2500 the peak of w moves from x ~ 0.1375 to x ~ 0.1 on a fixed mesh Nx = 40. This is an extrapolation, not a measured scaling of the gradient. More importantly, the convergence analysis in Section 6 starts by assuming a smooth domain boundary and smooth eigenmode, base flow, and pressure fields, then proves O(h^2) eigenvalue and O(h) eigenmode error bounds with constants C_{u,p} that depend on u and p (Theorem 6.4, eqs. 87-88). In exactly the regime where the paper claims ||grad u|| is unbounded, those constants are uncontrolled and the proof does not apply. The duct also has sharp corners, while the Q1 element and half-vortex outlet construction introduce additional non-smooth features that the analysis does not cover. No mesh-refinement study of the mode gradient is reported. The critical amplitude epsilon in eq. (25) is computed from the same unresolved eigenmode, so the claimed epsilon ~ Re^{-0.6} scaling is not independent of the questionable gradient data. Thus the central singular-structure claim rests on a numerical regime that the paper's own convergence theorem excludes. The linear-stability result for Re up to 2500 is less affected, because the eigenvalues are shown to converge and are consistent with prior studies; the fragile part is specifically the new boundary-layer/singularity interpretation.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the linear stability of three-dimensional Poiseuille flow through a finite-length square duct. The authors formulate the linearized eigenvalue problem using a weakly divergence-free finite element basis and propose a SUPG-based stabilized eigenvalue scheme. They compute the least-stable eigenvalues and eigenmodes for Reynolds numbers from 500 to 2500 and report that the flow is linearly stable in this range. The central new assertion is that at high Reynolds numbers the least-stable eigenmodes develop a boundary-layer structure near the duct walls, with gradients that grow without bound as Re increases. The paper argues that this singular structure, combined with the decreasing decay rate, makes the linearized Navier-Stokes operator pseudospectral and permits nonlinear effects to become significant while perturbation energy is still small. A critical perturbation amplitude scaling of approximately Re^-0.6 is inferred from the computed modes.","tokens_in":16781,"tokens_out":7138,"duration_ms":77663,"significance":"If the central claim were established, the paper would make a notable contribution by suggesting that standard linear stability analysis may be an inadequate model for high-Reynolds-number duct flow, and by connecting eigenmode structure to pseudospectral sensitivity and early nonlinearity. The numerical method itself, a divergence-free-basis FEM with a SUPG eigenvalue formulation, is potentially useful and is described in enough detail to be assessed. The eigenvalue computations for Re up to 2500 appear plausible and consistent with the existing understanding that duct Poiseuille flow is linearly stable. However, the paper's most important and novel conclusion, namely the unbounded gradient of the eigenmodes and its physical consequences, rests on numerical evidence that is not supported by the paper's own convergence analysis and lacks resolution checks.","major_comments":[{"comment":"The convergence proof assumes a smooth domain boundary and smooth base flow, eigenmode, and pressure fields, with constants C_{u,p} that depend on the solution. The central high-Re claim is precisely that the eigenmodes are boundary-layer-like with unbounded gradients, a regime in which the smoothness assumptions fail and the constants are uncontrolled. The duct also has sharp corners, which the analysis does not cover. No mesh-refinement study of the eigenmode gradients is reported; Figure 9 uses a fixed Nx=40 for all Reynolds numbers, so the observed shift of the peak from x about 0.1375 at Re=500 to x about 0.1 at Re=2500 cannot be distinguished from a resolution artifact, and it does not directly demonstrate an unbounded gradient.","section":"Section 6, first paragraph and Theorem 6.4, Eqs. (87)-(88)"},{"comment":"The critical amplitude epsilon and the Re^-0.6 scaling are computed from the same eigenmode whose gradient convergence is not verified, so the scaling is not independent evidence for the singular-structure scenario. The formula itself is ambiguous: the ratio |lambda_1 u_i^1 u_i^1 dOmega| / |u_k^1 grad_k u_i^1 u_i^1 dOmega| is written with dOmega inside the absolute values, making it unclear whether the quantity is a pointwise local ratio or an integral over an infinitesimal volume. If it is a local quantity, the global energy threshold in Eq. (24) should be derived from integrated quantities, and the two definitions may yield different scalings.","section":"Section 4, Eq. (25) and Figure 10"},{"comment":"The energy identity is derived from 'equation (1)' for the total velocity, but the forms A0 and C0 in Eq. (13) are the linearized operators acting on perturbation velocities. If u in Eq. (23) is the total velocity, pressure boundary terms at the inlet and outlet as well as the base-flow advection contribution are omitted. The derivation should be written explicitly for the perturbation velocity after subtracting the base flow; otherwise Eq. (24), which is the basis for the critical amplitude, is not justified.","section":"Section 4, Eqs. (23)-(24)"},{"comment":"All computations are performed for a single finite duct length L=3 and a single outlet condition. Figures 7 and 8 show that the least-stable modes are concentrated near the outlet, so the claimed boundary-layer structure and its extrapolation to very high Re may be controlled by the artificial outflow boundary rather than by intrinsic duct-flow dynamics. A study of the dependence on L and on the outlet boundary condition is needed before the result can be interpreted as a property of high-Reynolds-number duct Poiseuille flow.","section":"Section 3 and boundary conditions (2)-(4), (8)-(10)"}],"minor_comments":[{"comment":"The manuscript contains many typographical errors, including 'Navier-Stoker', 'pseudospectrua', 'condtions', 'soltuions', and 'satifised'. These should be corrected throughout.","section":"Abstract and throughout"},{"comment":"In the definition of M(u_h, w_h), the second term uses u^i instead of u_h^i; the notation should be consistent.","section":"Eq. (20)"},{"comment":"The remark states that m_supg(u,w) = -M(u,w) and that M is 'a bilinear form defined in equation', but the equation reference is missing.","section":"Section 6, Remark 6.0.2"},{"comment":"The convergence plot caption does not state the Reynolds number or the sequence of mesh sizes used. This information is needed to assess the claimed convergence.","section":"Figure 5"},{"comment":"The constant K0 is given to many significant figures without explanation of how it is determined from the total flux constraint; a brief derivation or reference would improve reproducibility.","section":"Eq. (6)"}],"recommendation":"major_revision","confidential_remarks":"The paper's central claim is interesting but currently overreaches its numerical evidence. The authors should be asked to either (a) provide a resolution study of the eigenmode gradients and a verification of the boundary-layer scaling, or (b) substantially soften the claims about unbounded gradients and failure of linearization. The manuscript would also benefit from a comparison with existing duct-flow stability computations (e.g., Tatsumi and Yoshimura) to calibrate the new numerical method."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this one for the eigenvalue solver and the up-to-Re=2500 stability result, not for the boundary-layer singularity story. The genuinely new pieces are the weakly-divergence-free basis applied to a 3D finite-length duct and the SUPG stabilization of the generalized eigenproblem. The eigenvalues converge in the mesh study and agree with the established picture that square-duct Poiseuille flow is linearly stable at these Reynolds numbers. The finding that the least-stable modes are wall-dominated and concentrated near the outlet is concrete and worth checking. To their credit, they admit the eigenmode initial condition is not likely to be physically realized, so the critical-amplitude numbers are explicitly illustrative.\n\nThe soft spot is exactly where the authors push their interpretation. The claim of unbounded mode gradients as Re grows is extrapolated from a fixed 40x40x120 mesh (Figure 9), with no mesh-refinement study of the gradient itself. Their own convergence theorem (Section 6) assumes a smooth domain boundary and smooth eigenmode, base flow, and pressure; the duct has sharp corners and the computed modes are boundary-layer-like, so the theorem's constants, which depend on the eigenmode, are not controlled in the regime where the claim lives. The critical-amplitude ~Re^-0.6 fit is computed from the same unresolved eigenmode, so it is not independent evidence. I also suspect the finite duct length and the do-nothing outlet condition are doing some of the work in localizing the modes near z=L; a longer duct or a different outlet treatment would tell you whether the structure is intrinsic or an artifact of the chosen boundary conditions.\n\nThe paper is a real numerical study and the linear-stability result is credible. But the burden of the new boundary-layer/pseudospectra story sits on an extrapolation that the paper's own convergence analysis does not cover. A serious referee should ask for a resolution study of the mode gradient and a separation of computed facts from the Re-to-infinity extrapolation. The method and stability diagram are publishable with those revisions. I would send it to review rather than desk-reject, but I would not cite the Re^-0.6 scaling or the unbounded-gradient conclusion until they are tested on finer meshes and a longer domain.","headline":"Solid numerical stability study whose boundary-layer singularity claim needs a resolution study before it can be taken seriously.","tokens_in":17302,"tokens_out":6029,"would_cite":false,"duration_ms":62435,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["76E05","65N30","65F15"],"pacs":["47.20.Ft","47.15.Rq"],"model":"deepseek-v4-flash","headline":"Three-dimensional duct Poiseuille flow is linearly stable up to Re=2500, but the least-stable modes form wall boundary layers with unbounded gradients at high Re, making linear stability analysis a poor model for physical disturbances.","keywords":["Poiseuille flow","duct flow stability","hydrodynamic stability","pseudospectra","SUPG finite element method","divergence-free basis","boundary-layer eigenmodes","Reynolds number transition"],"falsifier":"Run the same eigenvalue problem at Re=2500 with boundary-layer-refined meshes or a spectral method and track the normalized mode gradient and leading eigenvalue: if the peak gradient stops growing and converges to a finite value as the mesh is refined, or if the eigenvalue shifts significantly with the mesh, the unbounded-gradient and pseudospectral claims lose their support. Experimentally, one could seed an eigenmode-like disturbance and measure whether its energy initially grows at amplitudes where the linear prediction says it should decay.","tokens_in":16248,"feed_emoji":"🌀","tokens_out":8405,"duration_ms":83210,"temperature":0.7,"pith_summary":"The paper studies whether the textbook linear-stability picture of duct Poiseuille flow survives at high Reynolds numbers. It finds that the flow is linearly stable up to Re=2500, but the least-stable eigenmodes develop a thin wall boundary-layer structure whose gradient grows without bound as Re increases, even though the underlying base flow has no boundary layer. Because these modes hug the duct wall and are convected downstream, small geometry variations can shift eigenvalues substantially, and the nonlinear term u·∇u becomes active while the perturbation energy is still small. The paper concludes that at very high Re, linearization of the Navier-Stokes equations may not describe physical disturbances in a finite-length duct, and it gives a fitted critical amplitude scaling of roughly $Re^{{-0.6}}$ for eigenmode-like disturbances.","feed_headline":"Wall-layer modes make duct-flow linear stability fail at high Re","feed_subtitle":"Steep wall modes mean tiny disturbances go nonlinear, so decay rates alone do not predict transition.","key_machinery":"The central mechanism is a SUPG-stabilized finite-element eigenvalue formulation built on a weakly divergence-free basis. The basis is constructed from face vortices on cubic Q1 elements, with boundary 'half vortices' at the outflow, so the incompressibility constraint is built into the discrete velocity space and the pressure variable drops out of the weak form. A streamline-upwind Petrov-Galerkin (SUPG) term with parameter τ is added to stabilize the eigenvalue problem when the element Péclet number is large, reducing the linearized Navier-Stokes operator to a generalized matrix eigenvalue problem Hc=λMc. The paper proves O($h^{2}$) convergence for eigenvalues and O(h) for eigenmodes under smoothness assumptions. The argument then uses the energy identity d/dt(1/2⟨u,u⟩)=λ1ε²⟨u1,u1⟩-ε³⟨u1·∇u1,u1⟩ at t=0 to define a critical perturbation amplitude below which linearization stays a valid model.","core_discovery":"On its own terms, the paper's discovery is that the least-stable eigenmodes of the linearized Navier-Stokes operator for three-dimensional duct Poiseuille flow are not smooth global modes at high Re: they form a boundary layer near the duct wall, peak closer to the wall as Re rises, and have an unbounded normalized gradient in the inviscid limit, even though the Poiseuille base flow has no boundary-layer structure. The paper connects this singular mode shape to two consequences: the linear operator becomes pseudospectral, because small wall perturbations can greatly alter the steep modes and hence the eigenvalues, and the nonlinear mechanism acts at small disturbance amplitudes, so the linear decay rate alone does not bound the dynamics. The computed data show decay rates decreasing with Re and a critical amplitude for nonlinear energy production that falls roughly as $Re^{{-0.6}}$. The paper also notes that the leading modes do not have a simple normal-mode structure u=Ψ(x,y)$e^{{ikz}}$, so that common ansatz may be inadequate for finite-length duct stability.","pith_inferences":["If the wall-layer mechanism is generic, the same unbounded-gradient eigenmode behavior should appear in other finite-length shear flows, so the linear-stability breakdown would not be a quirk of square ducts; this could be tested by applying the same eigenvalue analysis to rectangular or circular cross-sections.","The ε∼Re^{-0.6} exponent is tied to an eigenmode initial condition and a finite duct length, so real disturbances triggered by noise or localized forcing may follow a different scaling; the paper's durable contribution is the mechanism, not a universal threshold.","The pseudospectral claim is qualitative; computing the resolvent norm of the discretized linearized operator at these eigenmodes would turn the 'tends to have pseudospectra' statement into a quantitative curve.","The outlet condition and finite length may themselves concentrate the modes near the outflow; testing longer ducts or periodic streamwise boundary conditions would separate finite-length effects from intrinsic wall-layer mechanics."],"forward_implications":["The normal-mode ansatz u=Ψ(x,y)e^{ikz} is not a safe starting point for finite-length duct stability, since the leading disturbances are wall-focused, outlet-concentrated, and genuinely three-dimensional.","Predictions of transition based only on the real part of the least-stable eigenvalue become unreliable as Re grows, because nonlinear energy production can overwhelm linear decay while perturbation energy is still small.","Small manufacturing roughness or geometric imperfections of the duct wall can substantially alter the least-stable eigenmodes and eigenvalues at high Re, giving a physical interpretation of pseudospectral sensitivity.","The stable region shrinks with Re for two compounding reasons: the decay rate of the leading mode falls and the nonlinear production rate rises, so any disturbance threshold below critical amplitude must also decrease.","The computed critical amplitude for eigenmode-like disturbances scales roughly as ε∼Re^{-0.6}, giving a quantitative, initial-condition-dependent boundary beyond which linear stability analysis stops being a good model."],"supporting_citations":[{"why":"Provides the high-Re pipe-flow spectral and pseudospectral reference that motivates the claim that duct modes become sensitive.","marker":"[10]"},{"why":"Supplies the experimental transition-threshold scaling against which the paper's computed critical amplitude is compared.","marker":"[14]"},{"why":"Gives the pseudospectra-without-eigenvalues argument that the paper connects to its boundary-layer mode structure.","marker":"[15]"},{"why":"Supplies the SUPG stabilization framework the eigenvalue solver is built on.","marker":"[20]"},{"why":"Supplies the weakly divergence-free vortex-basis construction used to eliminate pressure from the discrete problem.","marker":"[24]"},{"why":"Gives the stabilization parameter used in the SUPG eigenvalue formulation.","marker":"[27]"},{"why":"Supplies the eigenvalue and eigenspace convergence theory used to prove the scheme's error estimates.","marker":"[28]"},{"why":"Provides the finite-element eigenvalue approximation results used for the eigenmode error estimate.","marker":"[29]"}],"fun_headline_variants":["Singular wall modes break linear stability in ducts","Duct flow linearization fails at high Re","Wall-bound eigenmodes doom linear stability","Pseudospectral wall modes complicate duct flow","Linear decay not enough: duct flow goes nonlinear"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof of convergence assumes smooth duct walls and smooth eigenmodes, but the duct has sharp corners and the modes that carry the conclusion are thin boundary layers with steep gradients, so the reported singular structure could be a numerical artifact in exactly the regime where the argument depends on it.","fun_headline_variants_meta":{"raw":{"variants":["Singular wall modes break linear stability in ducts","Duct flow linearization fails at high Re","Wall-bound eigenmodes doom linear stability","Pseudospectral wall modes complicate duct flow","Linear decay not enough: duct flow goes nonlinear"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00025,"raw_usage":{"total_tokens":1594,"prompt_tokens":1027,"completion_tokens":567,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":643,"completion_tokens_details":{"reasoning_tokens":498}},"tokens_in":643,"tokens_out":567,"duration_ms":6034,"temperature":1.0,"reasoning_tokens":498,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:51:27.522976+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same eigenvalue problem at Re=2500 with boundary-layer-refined meshes or a spectral method and track the normalized mode gradient and leading eigenvalue: if the peak gradient stops growing and converges to a finite value as the mesh is refined, or if the eigenvalue shifts significantly with the mesh, the unbounded-gradient and pseudospectral claims lose their support. Experimentally, one could seed an eigenmode-like disturbance and measure whether its energy initially grows at amplitudes where the linear prediction says it should decay.","supporting_citations":[{"cited_title":"Meseguer, L","cited_arxiv_id":null,"evidence_quote":"Provides the high-Re pipe-flow spectral and pseudospectral reference that motivates the claim that duct modes become sensitive."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the experimental transition-threshold scaling against which the paper's computed critical amplitude is compared."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the pseudospectra-without-eigenvalues argument that the paper connects to its boundary-layer mode structure."},{"cited_title":"Hughes, Recent progress in the development and understanding of SUPG methods with special reference to the compressible Euler and Navier-Stokes equations , Internat","cited_arxiv_id":null,"evidence_quote":"Supplies the SUPG stabilization framework the eigenvalue solver is built on."},{"cited_title":"Fortin, Old and new finite elements for incompressible flows, Internat","cited_arxiv_id":null,"evidence_quote":"Supplies the weakly divergence-free vortex-basis construction used to eliminate pressure from the discrete problem."},{"cited_title":"Hughes, M","cited_arxiv_id":null,"evidence_quote":"Gives the stabilization parameter used in the SUPG eigenvalue formulation."},{"cited_title":"Babˇ uska and J","cited_arxiv_id":null,"evidence_quote":"Supplies the eigenvalue and eigenspace convergence theory used to prove the scheme's error estimates."},{"cited_title":"Boffi, Finite element approximation of eigenvalue problems , Acta Numerica (2010)","cited_arxiv_id":null,"evidence_quote":"Provides the finite-element eigenvalue approximation results used for the eigenmode error estimate."}],"review_version":1}