{"id":"7388f511-e8cf-4d0c-8cf2-b577eb38999b","arxiv_id":"2411.14977","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A spectral element free-surface Navier-Stokes solver with p-multigrid accelerated Poisson solution demonstrates spectral convergence for nonlinear waves and O(n) scaling in 2D benchmark tests.","lead":"This paper presents a faster way to simulate nonlinear ocean waves by solving the full fluid equations with a moving water surface, using high-order mathematics and a solver that scales almost linearly with problem size. If the numerical tests hold up, engineers and oceanographers could predict steep waves over uneven seafloor with higher accuracy at lower cost.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The O(n) scalability claim rests on a frozen linearized preconditioner tested only in the small-amplitude regime; no experiment shows iteration counts and scaling persist for steep waves or long-time integration.","rationale":"The reader's weakest_assumption pinpoints the same load-bearing gap: the O(n) scalability result depends on a time-constant preconditioner derived from a small-amplitude linearization, and the paper provides no robustness study across wave steepness or time. My examination of the solver section confirms that Section 5.4 introduces the frozen preconditioner explicitly, and Section 6.3 tests it only at 30% of maximum steepness with a modest 17-wave window. This is the single most load-bearing concern because the abstract's headline claims include both 'spectral convergence for highly nonlinear waves' and 'O(n) computational scalability'; the spectral convergence claim is supported by the p-refinement study, and the Beji-Battjes agreement is a reasonable experimental validation, but the scalability claim is only demonstrated in a regime matching the preconditioner's design assumptions. The concern is addressable with additional numerical experiments, so the appropriate verdict remains CONDITIONAL rather than ACCEPT or REJECT. No internal inconsistency in the mixed-stage pressure-velocity coupling or the sigma-transformed weak forms was identified; those derivations are coherent and standard in structure. The absence of code and unreported filter parameters are reproducibility limitations, but they do not bear as directly on the central quantitative claims as the preconditioner robustness gap.","tokens_in":20421,"tokens_out":5973,"duration_ms":60260,"concrete_test":"Repeat the Section 6.3 scaling study for the same kh = 1 case but at 50%, 70%, and 90% of (H/L)_max, and for a 17-wave window as well as a longer window (e.g., 50 wave periods), recording PDC/GMRES iteration counts per RK stage and wall-clock time versus DoF. Also run one case at 90% steepness with the preconditioner rebuilt every stage to separate preconditioner-quality effects from solver implementation effects. If iteration counts and the time-versus-DoF slope remain essentially unchanged, the O(n) claim is supported; if they degrade, the Section 5.4 frozen-preconditioner assumption is the cause.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing concern is the O(n) scalability claim in Section 6.3, which rests entirely on the Section 5.4 device of freezing the p-multigrid preconditioner M to a linearized small-amplitude-wave operator so that multigrid operators are built only once. The actual mixed-stage Poisson operator changes at every RK stage through the sigma-transform, so bounded iteration counts are not automatic; they require M^{-1}A to remain a good preconditioner as the free surface evolves. The only scaling experiments use kh = 1 and (H/L) = 0.0301 (30% of maximum steepness), precisely the small-amplitude regime for which M was designed. No experiment varies wave steepness toward 90%, integrates over many periods, or otherwise perturbs the domain to test whether iteration counts and O(n) time scaling persist. If iteration counts grow with steepness or over time, the advertised O(n) scalability for nonlinear waves fails, and the claimed advantage of the solver strategy is not established. This is a testable efficiency-claim gap, not an inconsistency in the discretization itself.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a nodal spectral element method for the sigma-transformed free-surface incompressible Navier-Stokes equations, using explicit fourth-order Runge-Kutta time integration with a mixed-stage pressure-velocity coupling and a Poisson problem solved by a geometric p-multigrid preconditioner combined with PDC or GMRES. The model is verified against Rienecker-Fenton streamfunction wave solutions for a range of depths and steepnesses, validated against the Beji-Battjes submerged-bar experiment, and assessed for solver scalability. The central claims are spectral (exponential) convergence for highly nonlinear waves, excellent agreement with the bar experiment, and O(n) computational scalability of the p-multigrid-accelerated Poisson solver.","tokens_in":20665,"tokens_out":3250,"duration_ms":33544,"significance":"If fully substantiated, this is a useful contribution: it extends the single-domain pseudospectral model of Melander et al. to a geometrically flexible multi-domain spectral element setting, with a careful derivation of the mixed-stage Poisson problem and its weak form. The use of external analytical and experimental benchmarks is a strength, as is the explicit attention to the pressure Poisson problem as the main computational bottleneck. The p-multigrid transfer and smoothing framework follows established constructions and is presented in sufficient detail for implementation, apart from the missing parameter values noted below. However, the two headline claims are not fully supported by the evidence as presented: the high-steepness convergence plots do not clearly demonstrate exponential convergence, and the O(n) scalability claim rests on a single small-amplitude test case. The paper would be publishable after these load-bearing points are addressed and the missing numerical parameters are reported.","major_comments":[{"comment":"The abstract and conclusion claim spectral convergence in the velocity fields for highly nonlinear waves, but the 90% steepness cases in Figure 3(g)-(i) do not show an exponential error decay. The errors decrease from about 1e-2 to 1e-4 over P=2 to P=8, with the curves appearing to flatten at high P, which is more suggestive of algebraic or saturation behavior than exponential convergence. To support the claim, the authors should report the measured convergence rates (e.g., fitted slopes or exponential decay constants) for the 90% cases and should state the error tolerance, time-step size, and spectral filter parameters used in these runs, since any of these could limit the observed convergence.","section":"Section 6.1, Figure 3(g)-(i)"},{"comment":"The O(n) scalability claim is load-bearing and is not yet established for the nonlinear moving-domain problem. The preconditioner M is built once from a linearized small-amplitude-wave formulation, while the actual mixed-stage Poisson operator changes at every Runge-Kutta stage through the sigma-transform. The only scaling experiments, Table 1 and Figure 6, use kh=1 and (H/L)=0.0301, i.e., 30% of the maximum steepness, which is precisely the small-amplitude regime for which M was designed. No experiment varies the wave steepness toward the 90% cases used in Section 6.1, nor integrates over many periods, to show that iteration counts and O(n) time scaling persist when the domain deformation is large. Without such a robustness study, the advertised O(n) scalability for nonlinear waves remains a conjecture.","section":"Section 5.4 and Section 6.3"},{"comment":"Several numerical parameters that directly control the reported results are not specified, which prevents reproduction and weakens the quantitative claims. Section 4.4 defines the spectral filter S(i) with parameters alpha, beta, and Pc but never gives their values for the convergence or bar tests. Section 5.3 defines the additive Schwarz smoother with a user-determined overlap but does not report the overlap width or the number of smoothing iterations per multigrid level. These values affect both stability and the iteration counts in Table 1 and Figure 6, so they should be reported for every numerical experiment.","section":"Section 4.4 and Section 5.3"}],"minor_comments":[{"comment":"The wave period is stated as T=0.02 s, which is inconsistent with H=0.02 m and L=3.74 m in a wave tank; this is almost certainly a typo for a period near 2 s. Please correct it so the experimental setup is reproducible.","section":"Section 6.2"},{"comment":"Subfigures (c), (f), and (i) are labeled 'kh = 2' but the text in Section 6.1 states that the deep-water case is kh = 2π; the caption labels appear to be a typo and should be fixed.","section":"Figure 3 caption"},{"comment":"The statement that 'the time step size is chosen such that spatial errors dominate' would be more useful if the actual time step or CFL criterion were reported, especially because the convergence study is performed after one time step and temporal and spatial errors interact.","section":"Section 6.1"},{"comment":"Please correct minor typos and grammar issues, including 'Kroenecker' (should be 'Kronecker'), 'The slope of that bar' (should be 'The slope of the bar'), 'O(n) computational scalability simulations' in the abstract, and 'orthonomal' in Section 4.3.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper is technically credible and within the journal's scope, and the derivation of the mixed-stage Poisson formulation is a genuine strength. The main barriers to acceptance are the incomplete evidence for exponential convergence at high steepness and the narrow parameter range of the scalability study, together with the missing numerical parameters needed for reproduction. I would support acceptance after these points are addressed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short note on arXiv:2411.14977. This is a competent, incremental paper. The authors extend their single-domain pseudo-spectral INSE model [19] to multiple elements, using a nodal SEM discretization of the sigma-transformed free-surface Navier-Stokes equations and a geometric p-multigrid preconditioner for the mixed-stage Poisson problem. That combination is new, and it is a natural step toward wave-structure interaction. The equations are laid out carefully, the p-multigrid machinery is described concretely, and the verification uses genuinely external data: Rienecker-Fenton streamfunction solutions for convergence, and the Beji-Battjes bar experiment for physics. I also give credit for the honest efficiency comparison against the direct solver, which actually wins at strict tolerances for the small system they test.\n\nSoft spots, in increasing order of importance. Minor: the bar test lists T = 0.02 s, which must be 2.0 s; the filter parameters (alpha, beta, Pc), the Schwarz smoother overlap, and the time step sizes are never stated, which hurts reproducibility; no code or data are included. More substantive: 'spectral convergence... for highly nonlinear waves' overstates Figure 3g-i. At 90% of the breaking steepness, the error curves flatten around 1e-2 to 1e-4 and do not show exponential decay; that is high-order convergence at best. The biggest concern is the O(n) scalability claim. The p-multigrid preconditioner is frozen to a linearized small-amplitude operator (Section 5.4), and the only scaling experiments use kh = 1 and (H/L) = 0.0301, exactly the small-amplitude regime the preconditioner was built for. There is no test at larger steepness or over longer integration times, so we do not know whether iteration counts stay bounded as the free surface evolves and the mixed-stage operator moves. This is a testable efficiency-claim gap, not a flaw in the discretization. The authors should either run that experiment or explicitly restrict the O(n) claim to small-amplitude waves.\n\nThe paper is for people building high-order free-surface Navier-Stokes solvers or using p-multigrid on moving domains. It is useful and credible. I would send it to review with a request for missing parameter values, a reworded convergence claim, and a steepness robustness test for the preconditioner. Without that test, the O(n) claim should be qualified.","headline":"Useful incremental SEM-INSE solver, but the O(n) scaling claim is only tested at small amplitude and the steep-wave 'spectral convergence' wording oversells the results.","tokens_in":21207,"tokens_out":4663,"would_cite":true,"duration_ms":42873,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65M70","65M60","65N55","76B15","76D05","65F10"],"pacs":[],"model":"deepseek-v4-flash","headline":"A σ-transformed spectral element discretization of free-surface Navier-Stokes achieves spectral convergence for steep waves, reproduces the submerged-bar experiment, and solves the pressure Poisson problem at O(n) cost with a geometric…","keywords":["nonlinear water waves","free-surface flow","incompressible Navier-Stokes equations","nodal spectral element method","sigma-transform","geometric p-multigrid","pressure Poisson problem","time-domain simulation"],"falsifier":"Track the pressure-Poisson iteration count at every Runge-Kutta stage while simulating waves near 90% of the maximal breaking steepness for many wave periods, and repeat for the submerged-bar case. If the count climbs as the wave train evolves or steepens, the frozen linearized preconditioner of Section 5.4 is not doing the claimed work, and the $O(n)$ scaling measured at tolerance $10^{-6}$ in Section 6.3 would not persist.","tokens_in":20190,"feed_emoji":"🌊","tokens_out":13633,"duration_ms":116436,"temperature":0.7,"pith_summary":"The paper builds a numerical wave tank from the two-dimensional incompressible Navier-Stokes equations with a tracked free surface, and claims that a nodal spectral element discretization on a σ-transformed, time-frozen mesh achieves spectral (exponential) convergence for highly nonlinear waves. It further claims that the model reproduces the classical submerged-bar experiment on harmonic wave generation, and that the pressure Poisson equation—normally the computational bottleneck—can be solved at each Runge-Kutta stage by a geometric p-multigrid preconditioner with iteration counts that stay bounded as resolution grows, giving O(n) total cost. The broader aim is a solver that combines finite-element geometric flexibility with spectral accuracy and near-linear scaling as a stepping stone to three-dimensional wave-structure simulations.","feed_headline":"Spectral wave model matches bar experiment at near-linear cost","feed_subtitle":"A p-multigrid preconditioner keeps the Poisson solve flat as resolution grows, matching the classic bar experiment.","key_machinery":"The central mechanism is the σ-transform, $\\sigma=(z+h(x))/(\\eta(x,t)+h(x))$, mapping the time-dependent physical fluid domain to a fixed reference strip so that the mesh and operators do not need rebuilding at each time step. On the reference domain, a nodal spectral element method with tensor-product Legendre bases and Legendre-Gauss-Lobatto nodes provides the spatial discretization, with gradient recovery through global $L^2$ projections and a modal spectral filter to stabilize the nonlinear terms. The accuracy claim is carried by the mixed-stage pressure Poisson problem, whose operator blends metric coefficients from two consecutive Runge-Kutta stages; solving it at every stage enforces $\\nabla_\\sigma\\cdot u=0$ and defines the pressure-velocity coupling. The scalability claim is carried by the geometric p-multigrid V-cycle—coarsening by halving polynomial order, additive Schwarz smoothing, and modal transfer operators with restriction as the transpose of prolongation—used as a preconditioner for PDC and GMRES, with its operators assembled once from a linearized small-amplitude wave state rather than recomputed per stage.","core_discovery":"On the paper's own terms, the discovery is that a nodal spectral element discretization of the σ-transformed incompressible Navier-Stokes equations, coupled with a mixed-stage pressure-velocity projection, achieves spectral error convergence for highly nonlinear waves while preserving mass conservation. The mixed-stage Poisson problem for the dynamic pressure, $\\nabla_\\sigma^{(k)}\\cdot\\nabla_\\sigma^{(k-1)} p_D^{(k-1)} = \\cdots$, mixes metric coefficients from two consecutive Runge-Kutta stages and must be solved at every stage; the paper shows this bottleneck can be handled by a geometric p-multigrid V-cycle used as a preconditioner for PDC and GMRES, with the multigrid operators frozen from a linearized small-amplitude wave state at the start of the simulation. Numerical evidence is offered in three parts: velocity error after one time step falls exponentially with polynomial order $P=2,\\ldots,8$ at depths $kh=0.5,2,2\\pi$ and steepness up to 90% of the breaking limit; the free-surface signal matches the Beji-Battjes measurements at six gauge positions for harmonic generation over a submerged bar; and pressure-solve iteration counts stay flat between roughly $10^3$ and $10^5$ degrees of freedom, whether resolution is added through element count or polynomial order, which the authors read as $O(n)$ computational scalability.","pith_inferences":["The paper's own comparison table shows a direct solver beating the multigrid preconditioners at tight tolerances on the small test problem; a crossover study on larger and three-dimensional problems would show where the O(n) advantage takes over, a comparison the paper leaves implicit.","The exponential-convergence evidence is a single-step study; whether the spectral rate survives long-time integration, where the modal filter and temporal errors accumulate, is not quantified in the paper.","The scaling data come from a thin domain with only two vertical elements; three-dimensional grids or boundary-layer-refined meshes change the sparsity pattern, so flat iteration counts would need re-demonstration there.","The natural stress test of the frozen preconditioner is a long steep-wave run near 90% of the breaking limit: per-stage iteration counts would show whether the linearization degrades over time."],"forward_implications":["Spectral convergence means users can raise polynomial order instead of refining the mesh to reach engineering accuracy, and the p-multigrid preconditioner keeps the added cost near-linear.","The frozen preconditioner built from a linearized small-amplitude wave state removes the per-stage operator re-assembly that moving-domain formulations would otherwise require.","Close agreement with the submerged-bar data indicates the model captures shoaling and harmonic decomposition of nonlinear waves over uneven bathymetry.","The solver strategy is claimed to extend naturally to three space dimensions and to be highly suitable for massively parallel implementations, where the pressure Poisson problem dominates cost.","The p-multigrid approach applies equally when resolution grows through element count or through polynomial order, with iteration counts staying flat in both directions."],"supporting_citations":[{"why":"This work supplies the stabilised nodal spectral element formulation for fully nonlinear water waves that the present implementation is based on.","marker":"[1]"},{"why":"This work introduced the σ-transform for Navier-Stokes wave simulation, the coordinate mapping the whole scheme uses.","marker":"[13]"},{"why":"This work presents the high-order finite-difference free-surface model whose mixed-stage Poisson pressure-velocity coupling is extended here.","marker":"[17]"},{"why":"This work gives the single-domain pseudo-spectral free-surface Navier-Stokes model that the multi-domain spectral element scheme generalises.","marker":"[19]"},{"why":"This work demonstrates p-multigrid preconditioning for spectral element water-wave models, the strategy adapted to the pressure Poisson problem.","marker":"[34]"},{"why":"This work establishes the spectral element multigrid transfer operators, including restriction as the transpose of prolongation.","marker":"[38]"},{"why":"This work provides the preconditioned defect correction framework that the p-multigrid solver accelerates.","marker":"[40]"},{"why":"This work supplies the GMRES Krylov method used as the alternative iterative accelerator in the scaling tests.","marker":"[41]"},{"why":"This work provides the streamfunction theory giving exact periodic wave solutions used to measure spectral convergence.","marker":"[42]"},{"why":"This work reports the submerged-bar experiment whose measured data the model is compared against.","marker":"[44]"}],"fun_headline_variants":["p-Multigrid cracks wave solver's scaling wall","Spectral wave solver gains O(n) scaling via p-multigrid","p-Multigrid keeps wave pressure solves flat as grid grows","p-Multigrid accelerates nonlinear wave solver to O(n)"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The $O(n)$ scalability claim depends on the assumption that a preconditioner assembled once from a linearized small-amplitude wave state stays effective for the exact nonlinear, moving-domain pressure Poisson problem at every Runge-Kutta stage; the paper does not test whether iteration counts grow with wave steepness or with simulation time.","fun_headline_variants_meta":{"raw":{"variants":["p-Multigrid cracks wave solver's scaling wall","Spectral wave solver gains O(n) scaling via p-multigrid","p-Multigrid keeps wave pressure solves flat as grid grows","p-Multigrid accelerates nonlinear wave solver to O(n)"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00061,"raw_usage":{"total_tokens":2928,"prompt_tokens":1121,"completion_tokens":1807,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":737,"completion_tokens_details":{"reasoning_tokens":1733}},"tokens_in":737,"tokens_out":1807,"duration_ms":14156,"temperature":1.0,"reasoning_tokens":1733,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T14:39:54.393128+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Track the pressure-Poisson iteration count at every Runge-Kutta stage while simulating waves near 90% of the maximal breaking steepness for many wave periods, and repeat for the submerged-bar case. If the count climbs as the wave train evolves or steepens, the frozen linearized preconditioner of Section 5.4 is not doing the claimed work, and the $O(n)$ scaling measured at tolerance $10^{-6}$ in Section 6.3 would not persist.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"This work introduced the σ-transform for Navier-Stokes wave simulation, the coordinate mapping the whole scheme uses."},{"cited_title":"Engsig-Karup, A","cited_arxiv_id":null,"evidence_quote":"This work presents the high-order finite-difference free-surface model whose mixed-stage Poisson pressure-velocity coupling is extended here."},{"cited_title":"A High-Order Hybrid-Spectral Incompressible Navier-Stokes Model For Nonlinear Water Waves","cited_arxiv_id":"2406.00991","evidence_quote":"This work gives the single-domain pseudo-spectral free-surface Navier-Stokes model that the multi-domain spectral element scheme generalises."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"This work demonstrates p-multigrid preconditioning for spectral element water-wave models, the strategy adapted to the pressure Poisson problem."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"This work establishes the spectral element multigrid transfer operators, including restriction as the transpose of prolongation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"This work provides the preconditioned defect correction framework that the p-multigrid solver accelerates."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"This work provides the streamfunction theory giving exact periodic wave solutions used to measure spectral convergence."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"This work reports the submerged-bar experiment whose measured data the model is compared against."}],"review_version":1}