{"id":"7648a640-865f-4722-b260-a557069cdfb5","arxiv_id":"2506.03291","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A new Active Flux scheme for the 2D Euler equations combines an exactly solved locally linearized acoustic operator with a third-order advective evolution operator inside an additive splitting, using primitive variables and bound-preserving limiting.","lead":"The paper builds a new numerical method for the compressible Euler equations that treats sound waves with an exact acoustic evolution operator and advection with a third-order characteristic tracing step. It matters because it offers a one-stage Active Flux scheme with CFL up to 0.5, low numerical diffusion, and good behavior at low Mach number.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed maximal CFL 0.5 is not established for the additive-split scheme; stability of the sum of two stable evolution operators is not inherited and is not analyzed.","rationale":"The reader's verdict CONDITIONAL is appropriate. Among the unproven assumptions listed, the smoothness mismatch in Theorem 3.2 is a proof gap, but for a C0 piecewise-polynomial reconstruction that is pointwise O(h^3) accurate, the error introduced by evaluating the nested characteristic formula across cell boundaries is also O(h^3), so the third-order spatial accuracy is unlikely to be destroyed; the convergence tests support this. The stability of the additive-split scheme at CFL 0.5 is more directly load-bearing: the paper's only justification is an appeal to [CHK21] for a different, coupled linearized scheme, while the additive splitting can in principle have a larger amplification radius. The scheme is only tested at CFL 0.45, leaving the maximal CFL claim unverified. If a von Neumann analysis revealed instability below 0.5, the advertised CFL advantage over semi-discrete AF would be weakened, although the method would remain functional at lower CFL. I therefore keep the CONDITIONAL verdict and recommend adding a stability analysis or softening the CFL claim. The paper otherwise contains substantial independent support: detailed derivation of the acoustic evolution operator with pseudocode, a clean Taylor argument for the advection operator, and reproducible-looking numerical experiments across many test cases.","tokens_in":31789,"tokens_out":32325,"duration_ms":369894,"concrete_test":"Perform a von Neumann stability analysis of the linearized additive-split scheme: constant background (rho0, v0, p0), small perturbations, periodic Cartesian grid. Build the one-step amplification matrix for the three sets of degrees of freedom (cell averages, edge midpoints, nodes) using the exact acoustic evolution operator (Section 2) and the nested advection formula (71) linearized about the background, combined as S_ac + S_adv - I. Compute the spectral radius for CFL in (0, 0.5] with the same reconstruction and Simpson quadrature as the scheme. If the spectral radius exceeds 1 for any CFL < 0.5 (or at 0.5), the 'maximal CFL 0.5' claim is false and Section 3.3 must be revised; if it is <= 1 up to 0.5, the claim is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper states (Section 3.3) that the method has maximal CFL 0.5, citing [CHK21] for the stability of 'a combination of advection and acoustics' and the shortest degree-of-freedom separation. However, the actual point-value update uses additive operator splitting (Remark 3.1): q_new = S_ac(q_old) + S_adv(q_old) - q_old. Stability of the individual exact/approximate evolution operators does not imply stability of this sum; the one-step amplification matrix of the additive scheme is S_ac + S_adv - I, whose spectral radius can exceed 1 even when each factor is a contraction. No von Neumann analysis of this additive-split operator is given, and all numerical experiments are run at CFL 0.45, below the claimed limit. The cited [CHK21] analysis concerns a coupled (non-split) linearized evolution operator, not the present additive combination, which additionally omits the O(t^2) acoustic-advection cross terms. Thus the advertised 'large domain of stability' and the specific maximal CFL 0.5 are an extrapolation rather than a demonstrated property of the proposed scheme. This matters because the CFL advantage over semi-discrete Active Flux (CFL <= 0.2) is a central selling point.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a new one-stage Active Flux method for the two-dimensional Euler equations, based on an additive operator split into an acoustic subproblem (solved by the exact evolution operator for linear acoustics from a prior work, using locally linearized coefficients) and an advective subproblem (solved by a nested characteristic tracing formula). The method uses primitive variables for point values and for the reconstruction, a conservative update of cell averages, and a blended a priori/a posteriori bound-preserving limiter. The authors prove Theorems 3.1 and 3.2 giving third-order accuracy of the advective evolution operator for smooth data, note explicitly that the acoustic linearization and the splitting are only formally second-order in time, and present numerical tests (contact wave, moving vortex, spherical Sod, multi-dimensional Riemann problems, low-Mach vortex, Kelvin-Helmholtz instability) indicating third-order space convergence, low-Mach compliance, and low numerical diffusion. The paper also provides detailed pseudocode for an efficient implementation of the acoustic evolution operator.","tokens_in":32032,"tokens_out":8436,"duration_ms":90374,"significance":"If the claims hold, this is a valuable contribution: it offers a simple, efficient, fully discrete Active Flux scheme for multi-dimensional Euler equations that avoids the time-step restrictions of semi-discrete Active Flux methods, gives detailed implementation recipes for the exact acoustic evolution operator (Appendices B and C), and demonstrates competitive resolution of multi-dimensional Riemann problems and low-Mach flows, with apparently less diffusion than the semi-discrete method in the Kelvin-Helmholtz tests. The Taylor-level proofs in Theorems 3.1 and 3.2 are clean and self-contained, and the convergence tests against exact solutions support the accuracy claims. However, the advertised maximal CFL 0.5 is not demonstrated for the actual additive-split scheme, and the relationship between the formal second-order temporal error and the observed near-third-order convergence is not analyzed. These are load-bearing gaps that should be addressed before the claims can be accepted as stated.","major_comments":[{"comment":"The statement that the method has a maximal CFL number of 0.5 is not supported for the additive-split update (51). Stability of the acoustic and advective operators individually does not imply stability of the combined operator S_ac + S_adv − I; no von Neumann analysis of this one-step amplification operator is provided, and all numerical experiments in Section 4 are run at CFL 0.45, below the claimed limit. The citation to [CHK21] concerns the linear stability of a coupled (non-split) evolution operator, not the present additive combination, which also neglects the O(t^2) acoustic-advection cross terms. Because the large stability domain and the CFL advantage over the semi-discrete Active Flux method are central selling points (Sections 1 and 5), the authors should either prove stability for the additive-split operator or explicitly limit the claim to the observed CFL 0.45.","section":"Section 3.3 (CFL stability claim, Eq. (51), Remark 3.1)"},{"comment":"The observed near-third-order convergence is not reconciled with the formal O(Δt^2) error of the local acoustic linearization and of the additive splitting, as acknowledged in Section 3.4.1. All convergence tests refine space and time together at fixed CFL 0.45, so a genuine second-order temporal error would normally manifest as second-order global convergence unless its coefficient vanishes for the tested solutions (e.g., for the contact wave, p and v are constant and the acoustic linearization is exact). The paper does not provide a time-refinement study at fixed spatial resolution, nor an error decomposition that would justify the claim that 'the global error is not dominated by its second-order component' (Section 3.4.1). Please add such a study or a quantitative discussion of the leading error terms.","section":"Section 3.4.1 and Section 4.1"},{"comment":"Theorems 3.1 and 3.2 assume smooth initial data for the PDE, but in the numerical update they are applied to a piecewise biparabolic reconstruction that is only globally C^0 across cell edges (Section 3.1). The paper does not analyze the error introduced by this smoothness mismatch, including the situation where the characteristic foot crosses a cell boundary and the node update receives contributions from several cells. The third-order accuracy of the advective evolution operator for the actual numerical data is therefore not established by the theorems. The authors should state the local smoothness assumptions under which Theorem 3.2 is applied to the reconstruction and provide evidence that the observed convergence order is robust with respect to this mismatch.","section":"Theorems 3.1 and 3.2, applied in Section 3.3"}],"minor_comments":[{"comment":"Equation (54) contains a typographical error: the second argument of λ2 should read x − t(λ1(x)+λ2(x))/2, not 'x − t λ1(x)+λ2(x) 2' as printed.","section":"Section 3.4.2, Eq. (54)"},{"comment":"Equations (73)–(74) are hard to read because the fraction bars are missing; the terms '(f^x(q̄_{i+1,j}) − f^x(q̄_{ij}))/2' and 'λ_{HLL}/2 (q̄_{i+1,j} − 2 q^n_{i+1/2,j} + q̄_{ij})' should be typeset with explicit division bars.","section":"Section 3.5.1, Eqs. (73)–(74)"},{"comment":"The sentence 'a two-dimensional Simpson rule is inverted to find the value at cell center' would benefit from a few more details or a reference, since this step is not immediately obvious and is important for the reconstruction.","section":"Section 3.3, item 2"},{"comment":"The statement that convergence 'starts to deviate slightly towards finer grids' is left unexplained; given the formal second-order temporal error, a brief discussion of whether this deviation is consistent with the O(Δt^2) term would be useful.","section":"Section 4.1.2"},{"comment":"The pseudocode uses the variable name 'diffX' to differentiate with respect to the radius r (the polynomial variable), which is confusing; a comment or a different name (e.g., 'diffR') would improve readability.","section":"Appendix B, Listing 1"},{"comment":"The statement that the usage of operators that are exact for linearized problems 'guarantees von Neumann stability of the method' is too strong for the nonlinear additive-split scheme and should be qualified as in the major comment on CFL stability.","section":"Section 5"}],"recommendation":"major_revision","confidential_remarks":"The CFL 0.5 claim is the main issue I see for acceptance; the paper should either provide a rigorous von Neumann analysis of the additive-split operator or substantially soften the claim, since all tests are at CFL 0.45. The paper is otherwise well-executed and the implementation details are valuable. I would also suggest the editor ask the authors to clarify the relationship to [CHK21]—the cited stability analysis is for a different (non-split) scheme, and the current text appears to overstate its applicability."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this one if you care about Active Flux beyond 1-D. The paper builds a one-stage method for the 2-D Euler equations by additively splitting acoustics and advection, solving the linearized acoustic part with an exact evolution operator and the advective part with a new third-order nested-characteristic formula. The idea is sound and the execution is careful. Theorems 3.1 and 3.2 are simple Taylor arguments that correctly identify the missing t^2 term for the advection operator, and the implementation details for the exact acoustic evolution operator—precomputed spherical means, quadrant splitting, polynomial coefficients—are genuinely useful and appear in this form for the first time. The paper is also unusually honest: it states outright that the method is formally second-order in time and that the observed near-third-order convergence is experimental, not derived. The contact wave and moving vortex tests support the accuracy claims in the tested regime. The bound-preserving limiter is described in enough detail to reimplement.\n\nThe soft spots are real but localized. The biggest one is the CFL 0.5 claim. The paper says the stability reduces to [CHK21], but [CHK21] analyzed a coupled advection-acoustics operator, not the additive split used here (Remark 3.1 uses S_ac + S_adv - I). Stability of the sum is not inherited from stability of the parts, and no von Neumann analysis of the split operator is given. All experiments run at CFL 0.45. So the advertised large stability domain is an extrapolation, not a demonstrated property. This is a gap, not a demonstrated bug—the method may well be stable at 0.5—but it is the central selling point the paper uses to distinguish itself from semi-discrete Active Flux, so it deserves a direct stability analysis or at least a numerical test at the claimed limit.\n\nSecond, the order proofs assume smooth initial data, but the actual reconstruction is only C^0 across cell edges. The paper does not analyze the resulting error, nor does it prove that the O(t^2) acoustic linearization and the splitting error don't dominate. The experiments suggest they don't in the tested cases, and the paper says so. That's acceptable for a numerical methods paper of this type, but keep it in mind when reading the convergence tables.\n\nThird, the claimed advantages over [CHLM24] and [ABK25] are supported mostly by qualitative comparisons—the KH test shows earlier vortex formation, which is nice but not a quantitative claim. No code or data is shipped, though the pseudocode is detailed enough to reimplement.\n\nBottom line: this is a serious, honest paper for people who work on Active Flux and evolution-operator schemes. It deserves a serious referee. I would send it out, but I'd ask the author to nail down the stability analysis for the additive split and to include a direct comparison at matching CFL before publication.","headline":"A promising, honestly limited one-stage Active Flux method for multi-D Euler; the claimed CFL 0.5 is not actually supported by analysis.","tokens_in":32533,"tokens_out":3108,"would_cite":false,"duration_ms":35746,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65M08","65M70","76M12","35L45"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper constructs a one-stage Active Flux method for the multi-dimensional Euler equations by additively splitting an exactly solved, locally linearized acoustic operator from a third-order approximate advective evolution operator…","keywords":["Active Flux","Euler equations","evolution operator","operator splitting","linear acoustics","spherical means","bound preservation","low Mach number"],"falsifier":"A decisive test is a mesh-refinement study of the moving vortex, or a similar genuinely multi-dimensional smooth flow, at fixed CFL number on grids from roughly $50\\times50$ to $400\\times400$, computing both point-value and average errors: if the measured order of convergence moves toward 2 rather than 3 as the mesh is refined, or if halving $\\Delta t$ at fixed grid size isolates an $O(\\Delta t^2)$ acoustic-splitting term that dominates the error, then the effective-third-order claim would be disproved.","tokens_in":31538,"feed_emoji":"🌊","tokens_out":6283,"duration_ms":62802,"temperature":0.7,"pith_summary":"The paper aims to show that a practical, one-stage Active Flux method for the multi-dimensional Euler equations can be built by splitting the equations into an acoustic part and an advective part, solving the acoustic part exactly on a locally linearized problem and the advective part with a new third-order approximate evolution operator. If the construction works, it matters because evolution-operator-based Active Flux methods have so far struggled to reach high order for nonlinear systems in several space dimensions, and existing semi-discrete versions pay for generality with low CFL numbers and added numerical diffusion. The paper claims the resulting scheme is third-order accurate in space, formally second-order in time, stable up to CFL 0.5, low-Mach compliant, and less diffusive than the semi-discrete Active Flux method on subsonic tests. On smooth test problems the observed convergence is close to third order, and the method resolves multi-dimensional Riemann problems, low-Mach vortices, and Kelvin-Helmholtz instabilities.","feed_headline":"Exact acoustics plus third-order advection powers new Active Flux scheme","feed_subtitle":"One-stage scheme handles multi-dimensional Riemann problems, low-Mach vortices, and Kelvin-Helmholtz growth with less diffusion.","key_machinery":"The load-bearing object is the exact evolution operator for linear acoustics, given by formulas (5)-(6): spherical means of the initial data, together with $n$-weighted variants, are differentiated in the radius variable and evaluated at $r = ct$, producing the exact solution for piecewise polynomial data even when discontinuities are straight lines. The paper supplies the first efficient implementation details for this operator, including precomputed polynomials for wedge-shaped angular subdomains. The second ingredient is the additive operator splitting $S_2(q_0,t) + S_1(q_0,t) - q_0$, which lets each sub-operator act on the same initial data instead of sequentially, avoiding the need for conservative updates of the non-conservative sub-operators. The third ingredient is the third-order advective evolution operator from Theorem 3.2, a nested characteristic formula that generates the required second-order Taylor term without spatial derivatives. A biparabolic reconstruction in primitive variables provides globally continuous initial data, and a blended a priori/a posteriori bound-preserving limiter keeps density and pressure positive.","core_discovery":"The central discovery is that the acoustic and advective operators of the Euler equations can be treated asymmetrically inside one Active Flux step. The acoustic sub-operator is linearized by freezing density and sound speed at each point value and then evolved exactly, using the exact solution formula for linear acoustics in terms of spherical means of the initial data, so all multi-dimensional wave information is included. The advective sub-operator is evolved by a nested characteristic evaluation, $x \\mapsto x - v_0(x - v_0(x)t)t$, which Theorem 3.2 shows reproduces the correct $O(t^2)$ Taylor term for systems whose Jacobians are $v_x I, v_y I$, i.e. the advective part of Euler in primitive variables. The two updates are combined additively, and the cell average is then updated conservatively in conserved variables using the evolved point values. The paper reports third-order spatial convergence, stability to CFL 0.5, low-Mach compliance, and visibly less diffusion than the semi-discrete Active Flux method.","pith_inferences":["Because the acoustic sub-operator is exact for the locally linearized acoustics and the splitting is additive, the low-Mach compliance is plausibly inherited from the stationarity-preserving property of the exact acoustic Active Flux method; the same split may extend to other systems that admit an exact linear-acoustics solver, such as shallow water equations.","The nested characteristic update resembles a Runge-Kutta-style staged evaluation; substituting a higher-order acoustic treatment could in principle raise the method to full third order in time, at the cost of the extra computational effort the paper deliberately avoids.","A quantitative comparison against the semi-discrete Active Flux method at matched resolution and CFL on the Kelvin-Helmholtz and moving-vortex tests would isolate whether the reduced diffusion comes from the evolution operator, the splitting, or the primitive-variable reconstruction.","The bound-preserving limiter could be adapted to enforce additional physical constraints such as entropy or velocity bounds, since it already blends a priori and a posteriori strategies and is activated rarely."],"forward_implications":["The method is third-order accurate in space and shows near-third-order convergence on smooth tests such as the contact wave and the moving vortex, with point values of pressure and velocity reaching machine precision for the contact wave.","It remains stable at CFL numbers up to 0.5, the theoretical limit for the Cartesian Active Flux structure, whereas the semi-discrete Active Flux method used for comparison has lower CFL limits.","The scheme is low-Mach compliant: stationary vortices are preserved on an $\\epsilon$-independent time scale in the limit Mach number $\\to 0$, avoiding the artificial dissipation that plagues non-compliant finite volume methods.","In the Kelvin-Helmholtz tests the new method develops vortices earlier and more homogeneously than the semi-discrete Active Flux method, indicating substantially less numerical diffusion in the subsonic regime.","The average update is conservative because fluxes are evaluated in conserved variables using the primitive-variable point values, and the blended limiter guarantees positivity of density and pressure."],"supporting_citations":[{"why":"Supplies the exact solution formula for linear acoustics, with spherical means, that the locally linearized acoustic sub-operator uses.","marker":"[BK22]"},{"why":"Provides the Active Flux method for linear acoustics on Cartesian grids and the stationarity/low-Mach link that the new method reduces to in the linear limit.","marker":"[BHKR19]"},{"why":"Establishes the need to go beyond local linearization for third-order Active Flux and motivates the nested characteristic evolution used for advection.","marker":"[Bar21a]"},{"why":"Gives a second-order bicharacteristic Active Flux method that the present work compares against and improves on by using the exact acoustic operator.","marker":"[CHLM24]"},{"why":"Provides the semi-discrete Active Flux method for the Euler equations that serves as the main comparison for accuracy, diffusion, and CFL number.","marker":"[ABK25]"},{"why":"Analyzes linear stability of the Cartesian-grid Active Flux method and establishes the CFL 0.5 stability limit used here.","marker":"[CHK21]"},{"why":"Supplies the bound-preserving limiting strategy that blends a priori and a posteriori approaches, used for the average update.","marker":"[DBK25]"}],"fun_headline_variants":["Exact acoustics, third-order advection in one Active Flux step","Active Flux: exact acoustic waves plus third-order advection","Asymmetric splitting in Active Flux: exact acoustics, third-order advection","Active Flux for Euler: exact acoustic operator, third-order advection"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that applying a third-order advective update, designed and proved for smooth data, to the $C^0$ piecewise biparabolic reconstruction produces an error that stays below the formal second-order acoustic/splitting error; the paper observes near-third-order behavior experimentally but does not prove that the smoothness mismatch or the $O(t^2)$ acoustic term does not dominate on finer meshes.","fun_headline_variants_meta":{"raw":{"variants":["Exact acoustics, third-order advection in one Active Flux step","Active Flux: exact acoustic waves plus third-order advection","Asymmetric splitting in Active Flux: exact acoustics, third-order advection","Active Flux for Euler: exact acoustic operator, third-order advection"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000676,"raw_usage":{"total_tokens":3040,"prompt_tokens":872,"completion_tokens":2168,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":488,"completion_tokens_details":{"reasoning_tokens":2092}},"tokens_in":488,"tokens_out":2168,"duration_ms":16000,"temperature":1.0,"reasoning_tokens":2092,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T11:06:35.870955+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A decisive test is a mesh-refinement study of the moving vortex, or a similar genuinely multi-dimensional smooth flow, at fixed CFL number on grids from roughly $50\\times50$ to $400\\times400$, computing both point-value and average errors: if the measured order of convergence moves toward 2 rather than 3 as the mesh is refined, or if halving $\\Delta t$ at fixed grid size isolates an $O(\\Delta t^2)$ acoustic-splitting term that dominates the error, then the effective-third-order claim would be disproved.","supporting_citations":[],"review_version":1}