{"id":"a4f8814c-7b42-4041-bc5b-4fab8fde4dbf","arxiv_id":"2411.16367","paper_version":5,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"An RKDG method using flux-vector-splitting fluxes and an IS-L2 constrained-optimization TVB(D)-minmod limiter is constructed for hyperbolic conservation laws; its claimed advantages rest mainly on plotted comparisons.","lead":"This paper combines flux vector splitting with discontinuous Galerkin discretization for hyperbolic conservation laws, and adds an optimization-based limiter that mixes WENO smoothness and L2 error to control oscillations at high polynomial order. A generalist might read it as a candidate incremental improvement for robust high-order CFD, but the support is mostly visual, the limiter weights are tuned per problem, and no code is provided.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The limiter's balancing claim is not established: the IS-L2-TVB(D)-minmod limiter relies on per-problem tuned weights (Eq.","rationale":"The reader's weakest assumption is precisely the load-bearing weakness I find: the IS-L2-TVB(D)-minmod limiter's balancing property is not a property of the optimization formulation alone but of the chosen weights. The paper changes weights from example to example, so the central novelty claim is not predictive as stated. I additionally note that Tables 1–4 contain identical L2-error rows for different conservative variables, which is inconsistent with the exact solution of Example 8.1; this strengthens the need for independent numerical verification. I do not see a need to move the verdict: the paper could be salvaged by supplying reproducible code, corrected tables, a weight-selection strategy, or a demonstration that one fixed weight pair works across the test suite.","tokens_in":70456,"tokens_out":7066,"duration_ms":80288,"concrete_test":"Run Examples 8.4, 8.5, 8.8, 8.9, 8.10, 8.11, and 8.12 with one fixed weight pair, e.g. (omega_IS, omega_L2) = (0.75, 0.25), using the same TVB indicator parameter M = 1, and report L1/L2 errors plus an overshoot measure near discontinuities. If any test develops oscillations or loses accuracy relative to the paper's tuned weights, the balancing claim fails without a weight-selection rule; if all cases pass, the tuning concern is substantially weakened.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing claim is that the IS-L2-TVB(D)-minmod limiter “balances oscillation suppression and high precision” for arbitrary high-order DG polynomials (Sections 6.2, 9). The mechanism is the user-chosen convex combination in Eq. 6.28/6.29: minimize omega_IS*IS(u) + omega_L2*||u - u_old||^2 subject to TVB-modified endpoint values. The L2 term explicitly pulls the corrected polynomial back toward u_old, the oscillatory high-order solution the limiter is meant to cure, while the IS term pushes toward a heavily damped solution. Nothing in the formulation guarantees that the minimizer is simultaneously oscillation-free and high-order accurate; the balance is entirely determined by the two weights. The paper varies the weights across examples (Example 8.4: 1.0/0.0; Example 8.10: 0.75/0.25; Example 8.11: 0.8/0.2) and even reports that the pure L2 constraint can blow up on the Lax shock tube. No automatic selection rule, no sensitivity study, and no fixed-weight test across all problems is provided. Thus the numerical comparisons show that tuned runs work, not that the method itself balances the two goals. Separately, Tables 1–4 print identical L2 errors for rho, rho*u, and E, which is incompatible with the stated exact solution of Example 8.1 and further undermines the quantitative evidence.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a Runge-Kutta discontinuous Galerkin (RKDG) method in which the numerical flux is obtained from flux vector splitting (Jacobian-eigenvalue splitting or Mach-number splitting), giving what the authors call FVS-DG in one and two space dimensions. It also introduces a limiter that solves a constrained optimization problem: minimize a weighted combination of the WENO smoothness indicator IS and the L2 deviation from the original high-order solution, subject to the endpoint or boundary-average constraints produced by the TVB(D)-minmod limiter. A separate contribution is an interpolation-based characteristic transformation, with a proof of its equivalence to the moment-based transformation in one dimension when all components share the same basis. The numerical section contains accuracy tests for the Euler and shallow-water systems, scalar and system Riemann problems, and several two-dimensional test cases.","tokens_in":70739,"tokens_out":8369,"duration_ms":79768,"significance":"If the stated claims were fully supported, the paper would offer a simple FVS-type flux for DG and a limiter that can be applied to arbitrary high-order polynomial approximations while controlling oscillations. The derivation of the FVS-DG weak form is transparent, the equivalence proof in Section 7.2 is explicit, and Appendix C contains a useful derivation of a scalar Steger-Warming flux together with proofs of consistency, Lipschitz continuity, and monotonicity. However, the current quantitative support is not sufficient for the main claims: the balancing property of the limiter rests on per-problem tuned weights, and the accuracy tables contain an internal inconsistency. The significance is therefore conditional on a substantial revision of the evidence.","major_comments":[{"comment":"The central claim that the IS-L2-TVB(D)-minmod limiter balances oscillation suppression and high precision is not established by the evidence presented. The balance is controlled entirely by the user-specified weights omega_IS and omega_L2 in Eqs. (6.28) and (6.34), and these weights are changed from case to case: Example 8.4 uses (1.0, 0.0), Examples 8.7 and 8.10 use (0.75, 0.25), and Examples 8.11 and 8.13–8.15 use (0.8, 0.2). No automatic selection rule, no sensitivity study, and no fixed-weight test across all problems is provided. Moreover, the L2 term pulls the corrected polynomial back toward the original high-order solution, so for larger omega_L2 the limiter can in principle retain the oscillations it is meant to remove. Because the reported runs are tuned per problem, they demonstrate that suitable weights can be found, not that the proposed limiter itself balances the two objectives. Please provide a fixed-weight robustness study and quantitative measures such as overshoot magnitudes and smooth-region L1/L2 errors.","section":"§6.2.3, §8.2–8.3"},{"comment":"The accuracy tables contain an internal inconsistency that prevents verification of the claimed (K+1)-th order accuracy. For the exact solution of Example 8.1, rho*u = -0.7*rho and E = P/(gamma-1) + 0.5*rho*u^2 = 2.5 + 0.245*rho, so the errors of the three conservative variables cannot coincide. Nevertheless, Tables 1–4 report exactly the same L2 and L1 errors for rho, rho*u, and E; for instance Table 1 lists the L2-error 2.4737E-04 for all three variables at Mesh 10. This indicates either copy-paste errors or a defect in the error computation. Please provide corrected per-component errors and convergence orders for both the AUSM and Steger-Warming FVS-DG runs.","section":"Tables 1–4, Example 8.1"},{"comment":"The sentence that 'The FVS method inherently does not introduce any error' is not correct as stated. Flux splitting at the continuous PDE level is exact, but the discretized FVS-DG flux, e.g. hat-F_{i+1/2} = F^+(U^L_{i+1/2}) + F^-(U^R_{i+1/2}) in Eq. (4.9), is an approximate upwind numerical flux and introduces numerical dissipation, as the discussion of S-W versus L-F splitting in Appendix C itself indicates. The paper should remove or qualify this claim; otherwise the accuracy discussion is based on a false premise about the numerical flux.","section":"§1 and §4"}],"minor_comments":[{"comment":"There are numerous typos, including 'Numderical' (Example 8.1 and elsewhere), 'uniquely spaced' (Figure 1 caption and throughout), 'T VB' and 'Minmd' in Section 6, and 'setp4' in Section 7; these should be corrected in a revision.","section":"Throughout"},{"comment":"The weights are described as forming a convex combination, but the condition omega_IS + omega_L2 = 1 is never stated; please state the admissible range explicitly.","section":"§6.2.3"},{"comment":"Proposition 7.1 should explicitly assume that the interpolation points are distinct so that the matrices P^{(i)} in Eq. (7.14) are invertible; currently this assumption is implicit.","section":"§7.2"},{"comment":"The derivation in Section 4 eventually reduces to the standard DG weak form with a single numerical flux hat-F = hat-F^+ + hat-F^- (Eqs. (4.7)–(4.8)); the novelty discussion would be clearer if this reduction were stated explicitly rather than presenting the two sub-fluxes as a separate formulation.","section":"§4"},{"comment":"The claims about overshoot and smoothness in Examples 8.4–8.6 are based only on visual inspection of plots; adding quantitative overshoot values or error norms would make the comparisons reproducible.","section":"§8.2"}],"recommendation":"major_revision","confidential_remarks":"The two main concerns are load-bearing: the limiter's balancing claim depends on per-problem tuned weights with no fixed-weight or sensitivity study, and the accuracy tables contain an inconsistency that must be corrected before the optimal-order claim can be assessed. The paper would also be stronger if the authors placed the FVS-DG scheme in context by acknowledging that it is standard DG with a particular upwind FVS numerical flux, rather than a new weak formulation. With corrected tables and a robustness study, a revised version could be acceptable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this one if you care about high-order DG limiters. The main new thing is the IS-L2-TVB(D)-minmod limiter: take the classical TVB(D) endpoint corrections, and choose the modified polynomial as the minimizer of a weighted sum of the WENO smoothness indicator and an L2 pull-back toward the unfiltered DG solution, subject to those endpoint constraints. That's a reasonable idea, cleanly derived as a saddle-point linear system in 1D and 2D. The paper also proves the scalar Steger-Warming flux is consistent, Lipschitz, and monotone, which gives L2 stability; that's a genuine if small contribution, but it's relegated to an appendix.\n\nThe FVS-DG part is not new: it reduces to standard DG with the upwind split-flux numerical flux F^+(U_L)+F^-(U_R), and the claim to be the first to introduce FVS into DG is overstated.\n\nNow the soft spots, in order of severity.\n\nFirst, Tables 1-4 print the identical L2 and L1 errors for rho, rho*u, and E. For the stated exact solution, rho*u = -0.7*rho and E is an affine function of rho, so these errors cannot be equal. Either the tables are misreported or the accuracy test has a bug. That's a mechanical red flag that undermines the convergence claims.\n\nSecond, the limiter's balancing claim rests on the user-chosen weights omega_IS and omega_L2, which change from example to example (e.g., 1/0 in Example 8.4, 0.75/0.25 in 8.10, 0.8/0.2 in 8.11). There is no automatic selection rule, no sensitivity study, and no evidence that a fixed weight works across problems. As the stress-test note says, the L2 term pulls back toward the oscillatory high-order solution, so the balance is not guaranteed by the construction. The paper's own remark that pure L2 can blow up on the Lax tube shows the tuning is load-bearing. This is the weakest point: as submitted, the method demonstrates that tuned runs work, not that it balances the two objectives predictively.\n\nThird, the limiter experiments are shown only as plots. No error tables, no convergence orders for the limited solutions. For a paper whose central claim is accuracy preservation at P3-P5, that's a missing piece.\n\nWhat's good: the optimization formulation is careful and reproducible in principle, the equivalence between interpolation-based and moment characteristic transformations is spelled out, and the scalar SW flux proofs are correct as far as I checked. The paper engages seriously with the literature.\n\nBottom line: the paper deserves a serious referee, but it needs major revision: fix or explain the tables, add quantitative limiter results, and either provide a weight-selection strategy or show a fixed weight works across a broad test suite. I'd be skeptical of the load-bearing claim until then.\n\nBest, [Name]","headline":"A novel constrained-optimization limiter that deserves referee time, but the paper's quantitative evidence is unreliable and the weight-dependence undercuts its central claim.","tokens_in":71260,"tokens_out":3202,"would_cite":false,"duration_ms":29665,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65M60","65M99","35L65"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper brings flux vector splitting into the discontinuous Galerkin framework and adds a constrained-optimization TVB minmod limiter that keeps high-order accuracy while suppressing oscillations near shocks.","keywords":["flux vector splitting","RKDG","hyperbolic conservation law","TVB(D)-minmod limiter","Lagrange multiplier method","local characteristic decomposition","WENO smoothness indicator","Runge-Kutta discontinuous Galerkin"],"falsifier":"Fix the weights once, say $\\omega_{\\mathrm{IS}}=0.75$, $\\omega_{L^2}=0.25$, and run the limiter on the paper's own $P^3$ Burgers shock test, the $P^3$ Sod problem, and the $P^5$ Shu-Osher problem; if any of those runs shows visible overshoot, a loss of the $(K+1)$-order convergence rate on smooth parts, or a divergence, then the claimed balance depends on per-problem tuning rather than on the method itself.","tokens_in":70212,"feed_emoji":"🌊","tokens_out":6892,"duration_ms":66909,"temperature":0.7,"pith_summary":"The paper introduces flux vector splitting (FVS) into the discontinuous Galerkin (DG) framework for hyperbolic conservation laws, deriving a spatial semi-discrete scheme in which the numerical flux is built from positive and negative split sub-fluxes evaluated in their upwind directions. The central claim is that this FVS-DG scheme reaches the optimal $(K+1)$-th order of accuracy for $P^K$ polynomial approximations, tested on the Euler and shallow-water equations in one and two dimensions. To control oscillations near discontinuities, the paper replaces the underdetermined correction equations of the classical TVB(D)-minmod limiter by a quadratic optimization problem that minimizes a weighted combination of the WENO smoothness measure and an $L^2$-closeness term to the original high-order solution, subject to the minmod-modified boundary values. The resulting IS-$L^2$-TVB(D)-minmod limiter is claimed to apply to arbitrary high-order polynomial approximations and to balance oscillation suppression with accuracy. A reader would care because both ingredients are simple modifications of standard DG tools: a flux construction that avoids Riemann solvers, and a limiter that upgrades a classical low-order limiter to high order.","feed_headline":"Flux-split DG hits optimal order and a new limiter tames shocks","feed_subtitle":"Weighted smoothness-plus-L2 optimization extends TVB minmod to high-order polynomials without shock oscillations.","key_machinery":"The carrying object is the split flux identity $F = F^+ + F^-$, used both to build the DG numerical flux and to motivate the limiter's optimization. In one dimension the numerical flux at an interface is $\\hat{F}_{i+1/2} = \\hat{F}^+_{i+1/2} + \\hat{F}^-_{i+1/2}$, with $\\hat{F}^+$ taken from the left state and $\\hat{F}^-$ from the right state; in two dimensions the same splitting is applied to the normal flux. The limiter's machinery is a linear weighted-sum objective whose smoothness term is the WENO smoothness indicator (a scaled sum of squared derivatives), rewritten as a positive-semidefinite quadratic form in the modal coefficients, while the $L^2$ term is a diagonal quadratic form measuring distance to the original high-order solution. Lagrange multipliers enforce the minmod-modified boundary constraints, producing a small saddle-point linear system to solve per troubled cell. The interpolation-based characteristic transformation instead applies the left eigenvector matrix $L$ to sampled point values rather than to modal coefficient columns, and the paper proves this coincides with the moment transformation when all components share identical basis functions.","core_discovery":"On its own terms, the paper claims two new things. First, the FVS method—splitting the flux $F$ into $F^+$ and $F^-$ by Jacobian eigenvalue splitting or by Mach-number splitting, then applying upwind numerical sub-fluxes—has been incorporated for the first time into the DG weak formulation; the resulting FVS-DG scheme empirically delivers $(K+1)$-th order convergence for $P^K$ approximations on scalar and system equations in one and two dimensions. Second, the classical TVB(D)-minmod limiter, which gives only two correction conditions in 1D and four in 2D and therefore cannot determine all modal coefficients of high-order polynomials, is made well-posed by solving a constrained optimization problem: minimize $w_{\\mathrm{IS}}\\,\\mathrm{IS}(u^{\\mathrm{mod}}) + w_{L^2}\\|u^{\\mathrm{mod}} - u^{\\mathrm{old}}\\|_{L^2}$ subject to the cell average and the minmod-corrected boundary values. The paper further proposes an interpolation-based characteristic transformation that it proves equivalent to the moment characteristic transformation in 1D when all components share the same basis, and derives a Steger-Warming numerical flux for scalar equations from the Jacobian-FVS perspective.","pith_inferences":["A testable extension would replace the user-chosen weights with a local heuristic, for example setting $\\omega_{\\mathrm{IS}}$ high when a troubled-cell indicator flags a strong discontinuity and lowering it in mildly oscillatory regions; the paper gives no such rule.","The FVS-DG construction should transfer to other hyperbolic systems that admit homogeneous-flux or Mach-number splitting, such as magnetohydrodynamics, but the paper only demonstrates Euler and shallow-water equations; this is an extrapolation.","Because the equivalence proof for the interpolation-based characteristic transformation assumes all components share the same basis functions, the two transformations could differ for component-dependent bases; the paper leaves that case unexamined.","A cheaper production variant of the limiter might apply the smoothness-only objective near strong shocks and reserve the $L^2$ term for mild oscillations, since the paper itself notes that the smoothness constraint alone is preferable in extreme shock-tube cases."],"forward_implications":["If FVS-DG is correct, DG users can replace Lax-Friedrichs or Riemann-solver fluxes with Steger-Warming, van Leer, or AUSM split fluxes and still observe optimal $(K+1)$-order convergence, while inheriting the lower dissipation of FVS.","If the IS-$L^2$-TVB(D)-minmod limiter performs as claimed, the standard TVB(D)-minmod limiter can be extended to $P^K$ approximations for any $K$ by solving a small linear system per troubled cell, with the smoothness term preventing Gibbs oscillations and the $L^2$ term protecting accuracy.","The interpolation-based characteristic transformation gives an equivalent but simpler route to local characteristic decomposition in DG when all components share basis functions, and the paper's local-freezing proposal uses edge integral averages rather than cell averages.","The scalar Steger-Warming flux, proved consistent, Lipschitz, and monotone for linear advection and Burgers' equation, makes the DG scheme satisfy a cell entropy inequality and $L^2$ stability for those equations, with slightly lower dissipation than Lax-Friedrichs in the paper's long-time tests.","Numerical experiments on Sod, Lax, Shu-Osher, Blast, and two-dimensional Riemann problems indicate that $P^3$ and $P^5$ approximations with the new limiter capture discontinuities without visible overshoot on the tested meshes."],"supporting_citations":[{"why":"Supplies the L2-error constraint idea—keeping the reconstructed polynomial close to the original high-order scheme—that is combined with the smoothness factor in the bi-objective limiter.","marker":"[1]"},{"why":"Supplies the van Leer Mach-number-based flux splitting used as an FVS numerical flux in accuracy tests and system examples.","marker":"[3]"},{"why":"Supplies the Steger-Warming Jacobian eigenvalue splitting, defining $A^\\pm$ and $F^\\pm$, which is the basis of the FVS-DG flux and of the scalar Steger-Warming flux in Appendix C.","marker":"[5]"},{"why":"Establishes the TVB Runge-Kutta discontinuous Galerkin framework that the paper extends with FVS and the new limiter.","marker":"[14]"},{"why":"Supplies the WENO smoothness measurement function IS used as the objective in the IS-TVB and IS-L2-TVB limiter constructions.","marker":"[27]"},{"why":"Supplies Roe averaging, used for local freezing in the characteristic decomposition for systems.","marker":"[40]"},{"why":"Provides the hybrid limiter strategy that the paper contrasts with and extends; it recognized that the classical TVB(D)-minmod limiter only generates low-order polynomials.","marker":"[42]"},{"why":"Provides the OEDG limiter baseline compared in Burgers' tests, against which the paper claims no-overshoot behavior for its IS-TVB limiter.","marker":"[44]"}],"fun_headline_variants":["Flux vector splitting enters DG for first time","Optimized limiter tames shocks in high-order DG","FVS-DG: new limiter via constrained optimization","TVB-minmod goes high-order with weighted optimization","First flux-split DG scheme with shock-proof limiter"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The limiter's promised balance between accuracy and oscillation control rests on the user picking the two weights in the optimization objective well; the paper changes those weights from test to test and gives no automatic rule for choosing them.","fun_headline_variants_meta":{"raw":{"variants":["Flux vector splitting enters DG for first time","Optimized limiter tames shocks in high-order DG","FVS-DG: new limiter via constrained optimization","TVB-minmod goes high-order with weighted optimization","First flux-split DG scheme with shock-proof limiter"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000254,"raw_usage":{"total_tokens":1669,"prompt_tokens":1145,"completion_tokens":524,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":761,"completion_tokens_details":{"reasoning_tokens":446}},"tokens_in":761,"tokens_out":524,"duration_ms":5242,"temperature":1.0,"reasoning_tokens":446,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:12:33.897195+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Fix the weights once, say $\\omega_{\\mathrm{IS}}=0.75$, $\\omega_{L^2}=0.25$, and run the limiter on the paper's own $P^3$ Burgers shock test, the $P^3$ Sod problem, and the $P^5$ Shu-Osher problem; if any of those runs shows visible overshoot, a loss of the $(K+1)$-order convergence rate on smooth parts, or a divergence, then the claimed balance depends on per-problem tuning rather than on the method itself.","supporting_citations":[{"cited_title":"New two-dimensional slope limit ers for discontinuous Galerkin methods on arbitrary meshes [J]","cited_arxiv_id":null,"evidence_quote":"Supplies the L2-error constraint idea—keeping the reconstructed polynomial close to the original high-order scheme—that is combined with the smoothness factor in the bi-objective limiter."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the van Leer Mach-number-based flux splitting used as an FVS numerical flux in accuracy tests and system examples."},{"cited_title":"Flux vector splitting of the inviscid gasdyn amic equations with application to finite-difference methods [J]","cited_arxiv_id":null,"evidence_quote":"Supplies the Steger-Warming Jacobian eigenvalue splitting, defining $A^\\pm$ and $F^\\pm$, which is the basis of the FVS-DG flux and of the scalar Steger-Warming flux in Appendix C."},{"cited_title":"TVB Runge-Kutta local projection disco ntinuous Galerkin finite element method for conservation laws","cited_arxiv_id":null,"evidence_quote":"Establishes the TVB Runge-Kutta discontinuous Galerkin framework that the paper extends with FVS and the new limiter."},{"cited_title":"Efficient Implementation of Weighted ENO Sc hemes [J]","cited_arxiv_id":null,"evidence_quote":"Supplies the WENO smoothness measurement function IS used as the objective in the IS-TVB and IS-L2-TVB limiter constructions."},{"cited_title":"Approximate Riemann solvers, parameter vectors, an d difference schemes [J]","cited_arxiv_id":null,"evidence_quote":"Supplies Roe averaging, used for local freezing in the characteristic decomposition for systems."},{"cited_title":"An indicator-based hybrid limiter in discontinuous Ga lerkin methods for hyperbolic conservation laws [J]","cited_arxiv_id":null,"evidence_quote":"Provides the hybrid limiter strategy that the paper contrasts with and extends; it recognized that the classical TVB(D)-minmod limiter only generates low-order polynomials."},{"cited_title":"OEDG: Oscillation-eliminating discontinuous Ga lerkin method for hyperbolic conservation laws [J]","cited_arxiv_id":null,"evidence_quote":"Provides the OEDG limiter baseline compared in Burgers' tests, against which the paper claims no-overshoot behavior for its IS-TVB limiter."}],"review_version":1}