{"id":"f4888594-22bc-4b75-822b-5a652628e7a3","arxiv_id":"1908.01852","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A new gPAV-based rotational velocity-correction scheme for incompressible flows with open boundaries is proven unconditionally stable and uses only constant-coefficient linear solves per step.","lead":"This paper presents a time-stepping scheme for incompressible fluid flows on domains with open boundaries that remains stable even at very large time steps, a regime where standard solvers often blow up. It matters to scientists and engineers who simulate blood flow, external aerodynamics, or other flows that exit a computational domain.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 2.1 bounds only the auxiliary scalar R, not the physical energy; equation (45) gives no lower bound on xi, so E[u] is not shown to be bounded.","rationale":"The construction in Section 2 is coherent, the algebra leading to the explicit xi formula is intricate, and the numerical experiments are extensive. The theorem as stated is not false, but it proves a decay statement about the auxiliary scalar R, not about the velocity energy. Because the abstract and introduction advertise unconditional energy stability for the flow itself, the missing link between R and E[u] is the load-bearing weakness. The reader's conditional verdict is appropriate: the method is plausible and likely functional, but the stability guarantee should be stated as stability of the augmented auxiliary energy unless a relationship between R and the physical energy is supplied. A simple diagnostic test tracking both quantities would settle whether the gap is only formal or also visible in practice.","tokens_in":30034,"tokens_out":19586,"duration_ms":197687,"concrete_test":"Re-implement Algorithm 1 for the homogeneous case (f=0, p0=0, fb=0, w=0) with a nontrivial initial vortex and a moderately large time step, e.g. Delta t = 0.1. At each time step record (R^{n+3/2})^2, xi, and the physical energy E[u^{n+1}] = int_Omega (1/2)|u^{n+1}|^2 dOmega + nu D0 int_{partial Omega_o} (1/2)|u^{n+1}|^2 dA + C0. If there is any time window in which (R^{n+3/2})^2 is non-increasing while E[u^{n+1}] increases by more than a few percent, or if min_n xi approaches zero while E[u^{n+1}] is large, then the missing lower bound on xi is active and Theorem 2.1 does not control physical stability.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim of unconditional energy stability rests on Theorem 2.1, whose inequality (26) controls only the auxiliary scalar R: (R^{n+3/2})^2 - (R^{n+1/2})^2 = -xi Delta t [nu ||grad bar u^{n+1}||^2 + nu D0 int_{partial Omega_o} (1/2)|bar u^{n+1}|^2 |n dot bar u^{n+1}|] <= 0. R is not the physical energy E[u]; it is a computed auxiliary variable. In the homogeneous case (f=0, p0=0, fb=0, w=0), equation (45) reduces to xi = (R^{n+1/2})^2 / (E[bar u^{n+3/2}] + Delta t (A0+B0)). The paper proves only that xi > 0, which is not enough: if E[bar u^{n+3/2}] grows large, xi can become arbitrarily small, and the right-hand side of (26) vanishes even when the velocity is large. The proof never establishes a lower bound xi >= c > 0, nor an inequality of the form E[u^{n+1}] <= C (R^{n+1/2})^2. This gap matters because the actual velocity is u^{n+1} = u1 + g(xi) u2, while the dissipation in (26) is evaluated on bar u^{n+1} = u1 + u2; these two fields can differ substantially when xi is small. Thus the theorem proves stability of the modified auxiliary energy R^2, not of the physical velocity energy, and the paper's headline claim is stronger than what is established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a gPAV-based, second-order rotational velocity-correction scheme for incompressible Navier-Stokes equations with energy-stable open/outflow boundaries. An auxiliary scalar R is introduced as the square root of a modified energy, the governing system is reformulated using g(R^2/E), and the resulting algorithm requires solving two pressure and two velocity subproblems, each with a constant precomputable coefficient matrix, plus an explicit formula for the scaling factor xi. Theorem 2.1 states a discrete inequality showing that the auxiliary quantities R^{n+3/2} decrease in the homogeneous case, and the numerical sections demonstrate spatial and temporal convergence, long-time simulations in backflow regimes, and stability at large time step sizes for bifurcation-channel, cylinder-wake, and impinging-jet flows.","tokens_in":30329,"tokens_out":13333,"duration_ms":137545,"significance":"The algorithmic construction is attractive: it avoids nonlinear solvers, uses only constant precomputable matrices, gives an explicit positive formula for the auxiliary variable, and the experiments indicate robust practical behavior in challenging open-boundary flows, including strong backflow. The paper also honestly reports accuracy deterioration at large time steps, which is a useful caution. However, the central stability claim needs re-scoping: Theorem 2.1 controls the auxiliary scalar R rather than the physical energy of the computed velocity field, so the headline claim of unconditional energy stability is currently stronger than what is proved.","major_comments":[{"comment":"The stability inequality (26) controls only the difference of squares of the extrapolated auxiliary quantities R^{n+3/2} and R^{n+1/2}; it does not control the physical energy E[u^{n+1}]. In the homogeneous case the explicit formula (45) gives xi = (R^{n+1/2})^2 / (E[bar u^{n+3/2}] + (A0+B0)Delta t), which guarantees only xi > 0, not a lower bound on xi. Since the actual velocity is u^{n+1} = u1 + g(xi)u2 (Eq. (42)) while the dissipation in (26) is evaluated on bar u^{n+1} = u1 + u2 (Eq. (43)), E[u^{n+1}] can in principle grow even while R remains bounded and the right-hand side of (26) tends to zero. The authors should either prove a bound such as E[u^{n+1}] <= C (R^{n+1/2})^2, or explicitly re-state the stability result as a property of the auxiliary/modified energy and adjust the abstract and title accordingly.","section":"§2.3, Theorem 2.1, Eqs. (26), (42), (43), (45)"},{"comment":"Theorem 2.1 is stated in the limit delta -> 0, while all numerical experiments use a finite smoothing parameter: delta = 0.05 in §3.2 and delta = 0.01 in §3.3 and §3.4. As written, the proof does not cover the finite-delta scheme actually tested. This gap is readily repairable: from (6) and (46), B0 = 1/2 int_{partial Omega_o} |bar u|^2 (n·bar u) tanh((n·bar u)/(U0 delta)) dA, which is nonnegative for every delta > 0. Thus the monotonicity conclusion (26) can be proved for finite delta without taking delta to 0. The authors should make this repair and state Theorem 2.1 for the finite-delta regime used in the computations.","section":"§2.3, Theorem 2.1, and §3 numerical experiments"}],"minor_comments":[{"comment":"The notation with doubled norm bars, || ||nabla bar u|| ||^2, appears to be a LaTeX artifact; please use a single norm notation throughout.","section":"Equations (26), (46)"},{"comment":"The text refers to the 'no-clip condition' on the walls; this should be the 'no-slip condition'.","section":"§3.2, bifurcation channel"},{"comment":"The claim that xi > 0 unconditionally should be qualified: in the homogeneous case S1 = 0 and the numerator of (45) is (R^{n+1/2})^2, so xi = 0 if R^{n+1/2} = 0. If this exceptional case is possible, the authors should either rule it out or state the positivity claim for R^{n+1} only, which follows from (47).","section":"§2.4, positivity of xi"},{"comment":"There are minor typographical errors: 'Collge of Infomation Science' in the author affiliation and 'International Jurnal for Numerical Methods in Fluids' in reference [53].","section":"Affiliation and references"}],"recommendation":"major_revision","confidential_remarks":"The paper fits the scope of a computational physics journal and the comparison with [37] is fair. The main issue for the editor is that the headline 'unconditionally energy-stable' should be scoped to the auxiliary energy unless a physical-energy bound is added or the claims are revised. The finite-delta gap in Theorem 2.1 is fixable and should be addressed during revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear Colleague,\n\nWhat you should know: this paper presents a genuinely practical scheme, and the numerical work is convincing, but the headline stability claim overreaches. The theorem proves decay of an auxiliary scalar R, not of the physical energy E[u].\n\nThe new thing here is the combination: gPAV + convective-like energy-stable open boundaries + rotational velocity-correction, giving per-step decoupled Helmholtz/Poisson solves with constant precomputable matrices and an explicitly computed, positivity-guaranteed auxiliary variable. That is a real advance in cost per step for open-boundary simulations. The experiments are thorough: bifurcation channel, cylinder, impinging jet, with resolution studies and force data matching published values. The scheme's stability at large dt in backflow-dominated regimes is convincingly demonstrated.\n\nThe soft spot is the proof. Theorem 2.1 gives (R^{n+3/2})^2 - (R^{n+1/2})^2 <= 0, where R is the auxiliary variable. The physical velocity is u^{n+1}=u1+g(xi)u2, while the dissipation in (26) is on bar u^{n+1}=u1+u2. Equation (45) defines xi via a quotient; only xi > 0 is shown, not a lower bound. If xi is small, bar u can have small dissipation while u is large. The authors themselves note (Section 2.2) that R^2 approximates E rather than equaling it. So the theorem is better stated as stability of the auxiliary energy, and the title/abstract overclaim unconditional energy stability in the usual physical sense.\n\nSecond, the \"first unconditionally stable rotational velocity-correction scheme\" claim ignores their own reference [18], Dong & Shen 2010, which proves unconditional stability for a rotational velocity-correction scheme (without open boundaries). The claim needs a qualifier.\n\nMinor issues: Theorem 2.1 holds in the delta -> 0 limit, tests use delta = 0.01-0.05; no code or data is provided, so reproducibility is limited. These are minor.\n\nThe core method is sound and the experiments back it. The flaws are in framing, not in the scheme. This deserves a serious referee. I'd send it to review and ask for (1) a precise statement of what energy is being bounded, and (2) a corrected novelty claim. The scheme itself is worth publishing.\n\nRecommendation: accept conditionally after revisions.","headline":"Useful gPAV scheme for open-boundary flows, but the stability theorem bounds an auxiliary scalar, not the physical energy.","tokens_in":30892,"tokens_out":5313,"would_cite":true,"duration_ms":52026,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65M12","65M60","65M70","76D05","76M10","76M22"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims an unconditionally energy-stable time-stepping scheme for incompressible flows on domains with outflow or open boundaries, where backflow can otherwise make simulations blow up.","keywords":["energy stability","unconditional stability","auxiliary variable","generalized positive auxiliary variable","open boundary condition","outflow boundary","incompressible Navier-Stokes","backflow instability"],"falsifier":"Run the paper's forced-free, zero-Dirichlet test (f=0, p0=0, fb=0, w=0) with a fixed large \\$\\Delta$ t and record E[\\bar $u^{{n+1}}$] = \\int_\\$\\Omega$ \\tfrac12 |\\bar $u^{{n+1}}$|^2 d\\$\\Omega$ + \\nu D_0 \\int_{\\partial \\Omega_o} \\tfrac12 |\\bar $u^{{n+1}}$|^2 dA alongside R; if E grows over a block of steps while ($R^{{n+3/2}}$)^2 - ($R^{{n+1/2}}$)^2 stays non-positive, the physical-energy interpretation of Theorem 2.1 is false.","tokens_in":29774,"feed_emoji":"🌊","tokens_out":6727,"duration_ms":64004,"temperature":0.7,"pith_summary":"The paper proposes a time-stepping scheme for the incompressible Navier–Stokes equations on domains with outflow/open boundaries, the setting where backflow instability can wreck standard simulations. It claims the scheme is unconditionally energy-stable: in the forced-free, zero-boundary case, a discrete auxiliary energy decreases at every step no matter how large the time-step size. The scheme avoids nonlinear iterations by solving two decoupled pressure problems and two decoupled velocity problems per step, each with a constant precomputable coefficient matrix, plus an explicit scalar formula for the auxiliary variable. The method is demonstrated on bifurcation channels, cylinder wakes, and impinging jets, where it stays stable at large time steps, while the authors also caution that accuracy deteriorates when the time step becomes very large.","feed_headline":"Open-boundary flow solver stays stable for any time-step size","feed_subtitle":"Vortices can cross the open boundary without numerical blowup at any time-step size.","key_machinery":"The central object is a scalar auxiliary variable R(t) = \\sqrt{E(t)} attached to a biased modified energy E(t) = \\int_\\$\\Omega$ \\tfrac12 |u|^2 d\\$\\Omega$ + \\nu D_0 \\int_{\\partial \\Omega_o} \\tfrac12 |u|^2 dA + C_0, together with the gPAV gating function g(\\chi) = \\chi for \\chi \\le 1 and g(\\chi) = 1 for \\chi > 1. The discrete scheme evaluates the ratio $R^{2}$/E one half-step later as \\xi = ($R^{{n+3/2}}$)^2 / E[\\bar $u^{{n+3/2}}$], solves the two pressure and two velocity subproblems whose coefficient matrices are constant and precomputable, and then reconstructs $p^{{n+1}}$ = $p_1^{{n+1}}$ + g(\\xi) $p_2^{{n+1}}$ and $u^{{n+1}}$ = $u_1^{{n+1}}$ + g(\\xi) $u_2^{{n+1}}$. Positivity of the computed R and \\xi follows from the explicit formula, so no Newton-type nonlinear solver is needed.","core_discovery":"The central claim is that the generalized Positive Auxiliary Variable (gPAV) reformulation of the incompressible Navier–Stokes equations with an energy-stable open boundary condition admits a rotational velocity-correction-type discretization that is unconditionally energy-stable. Specifically, Theorem 2.1 states that, in the absence of external forces and source terms and with homogeneous Dirichlet data, the discrete relation ($R^{{n+3/2}}$)^2 - ($R^{{n+1/2}}$)^2 = -\\xi \\$\\Delta$ t [ \\nu \\int_\\$\\Omega$ \\|\\nabla \\bar $u^{{n+1}}$\\|^2 d\\$\\Omega$ + \\nu D_0 \\int_{\\partial \\Omega_o} \\tfrac12 |\\bar $u^{{n+1}}$|^2 |n\\cdot \\bar $u^{{n+1}}$| dA ] \\le 0 holds for any time-step size. The key move is replacing the ratio $R^{2}$/E in the nonlinear and boundary terms by g($R^{2}$/E) = \\min($R^{2}$/E, 1), which lets the stability estimate survive the decoupling of pressure and velocity. As a result, each time step requires only two pressure solves and two velocity solves with constant, precomputable matrices, and the auxiliary variable is recovered from an explicit formula that guarantees positivity.","pith_inferences":["Inference: The proof bounds the auxiliary variable R, not the physical kinetic-energy integral; a direct monitoring of E across the paper's own forced-free test would reveal whether the scheme is energy-stable in the stricter physical sense—this check is not in the paper.","Inference: The gating function g(\\chi) = \\min(\\chi, 1) is a simple saturation mechanism that other auxiliary-variable schemes could adopt to preserve positivity without nonlinear solves; testing this on gradient-flow or SAV-type problems would be a natural extension.","Inference: The numerical evidence that accuracy degrades as \\Delta t grows suggests the practical value of the scheme lies in removing blowup rather than in allowing arbitrarily large steps; pairing it with adaptive time stepping that monitors the physical energy would be a testable production rule."],"forward_implications":["Outflow/open-boundary simulations at moderate or high Reynolds numbers can be run with large time steps without the backflow blowup that afflicts conventional semi-implicit velocity-correction schemes.","The per-step cost is roughly twice that of a typical semi-implicit scheme, not the full nonlinear solve of most energy-stable schemes, because all algebraic systems have constant precomputable matrices.","The scheme is not tied to a particular spatial discretization; the paper implements it with high-order spectral elements but states that other spatial methods can be used.","The stability guarantee is proven in the forced-free, zero-Dirichlet case, while numerical experiments also demonstrate stable behavior in flows with external pressure heads and multiple openings.","Large-time-step results should be regarded only as reference solutions; the authors explicitly state that convergence tests remain necessary in production calculations."],"supporting_citations":[{"why":"Establishes the convective-like energy-stable open boundary condition with H(n,u) that the scheme discretizes and uses in Theorem 2.1.","marker":"[14]"},{"why":"Documents the backflow instability at outflow boundaries that motivates the energy-stable boundary treatment and provides an earlier outflow condition to beat.","marker":"[16]"},{"why":"Supplies the generalized energy-stable open-boundary framework and the comparison drag/lift values used to validate the cylinder simulations.","marker":"[20]"},{"why":"Introduces the auxiliary-energy-variable strategy this work extends; the paper's decoupled, positivity-guaranteed treatment is explicitly contrasted with [37]'s Newton-based coupled version.","marker":"[37]"},{"why":"Sets out the gPAV roadmap, including the half-step evaluation of R^2/E and the gating function, which the scheme adopts.","marker":"[54]"}],"fun_headline_variants":["gPAV scheme: unconditionally stable for open-boundary flows","No time-step limit: gPAV tames backflow instability","Energy-stable open boundaries for any time-step size","Two solves per field: stable scheme for open boundaries","Backflow blowup fixed: unconditionally stable discretization"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The stability claim collapses if the computed ratio \\xi can become small even while remaining positive, because Theorem 2.1 only bounds $R^{{n+3/2}}$ and never directly bounds the physical energy E[\\bar u]; the proof needs \\xi to keep working as a positive factor, and the numerical tests also show accuracy degrades as \\$\\Delta$ t grows.","fun_headline_variants_meta":{"raw":{"variants":["gPAV scheme: unconditionally stable for open-boundary flows","No time-step limit: gPAV tames backflow instability","Energy-stable open boundaries for any time-step size","Two solves per field: stable scheme for open boundaries","Backflow blowup fixed: unconditionally stable discretization"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000217,"raw_usage":{"total_tokens":1464,"prompt_tokens":1001,"completion_tokens":463,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":617,"completion_tokens_details":{"reasoning_tokens":380}},"tokens_in":617,"tokens_out":463,"duration_ms":4840,"temperature":1.0,"reasoning_tokens":380,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:47:09.351029+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the paper's forced-free, zero-Dirichlet test (f=0, p0=0, fb=0, w=0) with a fixed large \\$\\Delta$ t and record E[\\bar $u^{{n+1}}$] = \\int_\\$\\Omega$ \\tfrac12 |\\bar $u^{{n+1}}$|^2 d\\$\\Omega$ + \\nu D_0 \\int_{\\partial \\Omega_o} \\tfrac12 |\\bar $u^{{n+1}}$|^2 dA alongside R; if E grows over a block of steps while ($R^{{n+3/2}}$)^2 - ($R^{{n+1/2}}$)^2 stays non-positive, the physical-energy interpretation of Theorem 2.1 is false.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the convective-like energy-stable open boundary condition with H(n,u) that the scheme discretizes and uses in Theorem 2.1."},{"cited_title":"Dong, G.E","cited_arxiv_id":null,"evidence_quote":"Documents the backflow instability at outflow boundaries that motivates the energy-stable boundary treatment and provides an earlier outflow condition to beat."},{"cited_title":"Dong and J","cited_arxiv_id":null,"evidence_quote":"Supplies the generalized energy-stable open-boundary framework and the comparison drag/lift values used to validate the cylinder simulations."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the auxiliary-energy-variable strategy this work extends; the paper's decoupled, positivity-guaranteed treatment is explicitly contrasted with [37]'s Newton-based coupled version."},{"cited_title":"A Roadmap for Discretely Energy-Stable Schemes for Dissipative Systems Based on a Generalized Auxiliary Variable with Guaranteed Positivity","cited_arxiv_id":"1904.00141","evidence_quote":"Sets out the gPAV roadmap, including the half-step evaluation of R^2/E and the gating function, which the scheme adopts."}],"review_version":1}