{"id":"8a8175e1-ed97-451c-b482-c8c7eee0177f","arxiv_id":"2508.21226","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Two finite difference schemes, one using entropy-correction artificial viscosity and one using knapsack limiting, satisfy discrete entropy inequalities; the knapsack variant also provably preserves positivity for the compressible Euler and Navier-Stokes equations.","lead":"This paper builds two new high-order finite difference methods for gas dynamics that enforce a discrete version of the second law of thermodynamics. They are designed to avoid tuning parameters and to keep density and pressure positive, which could make high-order simulations of shocks more robust.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"KL-FD positivity guarantee is asserted rather than proved: formula (5.1) is not derived, the proof is imported from prior DG work, and no Navier-Stokes discretization is even defined.","rationale":"The reader's conditional verdict is the right one. The entropy-stability core of ECAV-FD appears sound: Lemma 3.3's symmetrization argument is valid, and for LxF-type fluxes the nonnegativity of the coefficients needed for the knapsack reformulation follows from entropy stability of the low-order flux plus convexity. The real soft spot is the positivity-preservation guarantee for KL-FD, which the paper imports rather than proves. This is load-bearing because the abstract promises a proof of positivity, not just numerical evidence. Formula (5.1) is introduced without derivation and with ambiguous parentheses in the printed text; the cited positivity theory is from discontinuous Galerkin methods; and the numerical scheme switches to HLLC even though the θ-equivalence is only shown for LxF. The complete absence of any Navier-Stokes discretization makes the abstract's Navier-Stokes positivity claim especially hard to verify. None of this is an internal contradiction in the entropy-stability derivation, so a reject would be too harsh; but the manuscript needs a self-contained derivation of the limiting coefficients and a precise statement of the flux class and stencil conditions under which positivity holds. If those are supplied, the central claims would be substantially stronger.","tokens_in":16063,"tokens_out":15171,"duration_ms":156594,"concrete_test":"Independently derive formula (5.1) from (4.3): for a single forward-Euler stage with a fourth-order SBP operator, compute directly, for every node, the minimal θ needed for U(θ) ≥ αU^L (componentwise for density and pressure), and compare with the ℓc supplied by (5.1). Then repeat the comparison for the HLLC-based KL-FD update, without assuming the LxF reduction, and for a concrete viscous/Navier-Stokes flux. If any admissible state requires a larger limiting coefficient than (5.1) provides, or if the HLLC update cannot be expressed through the optimization (4.4), the positivity claim fails as stated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The abstract's headline claim that KL-FD 'provably preserves positivity for the compressible Euler and Navier-Stokes equations' rests entirely on Section 4. The relative positivity condition (4.3) and the limiting-coefficient formula (5.1) are stated without derivation. No proof is given that, for the SBP finite-difference stencils of Section 2 and the optimization (4.4), the symmetrized final coefficients θij + ℓcij imply (4.3). The text defers to [4] and [11], which are DG papers; the transfer to finite differences is not automatic because flux-differenced SBP stencils couple nodes differently than DG subcell limiting. Moreover, the equivalence f_KL(θ) = f^H + θ(f^L − f^H) is demonstrated only for the LxF flux (2.7), while the numerical section uses HLLC; for HLLC the reduction to the weighted diffusion coefficients in (3.5)/(4.4) is not established. Finally, the manuscript only defines the hyperbolic system (2.1) and never presents a Navier-Stokes or viscous discretization, so the Navier-Stokes positivity claim is unsupported on its face.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes two entropy-stable finite difference schemes on SBP operators: ECAV-FD, which adds minimal artificial diffusion determined by solving a nodal quadratic optimization, and KL-FD, which blends a low-order entropy-stable positivity-preserving flux with a high-order flux via knapsack limiting. Section 3 gives a self-contained derivation of the discrete nodal entropy inequality and proves that symmetrized diffusion coefficients θij = max(θ̂ij, θ̂ji) preserve entropy stability and conservation. Section 4 introduces the knapsack-limited scheme and claims positivity preservation through a relative positivity condition, with a limiting-coefficient formula (5.1) used in the numerical experiments. The numerical section reports spatial convergence for N=2,...,5, Shu-Osher, Leblanc shock tube, Woodward-Colella blast wave, and two-dimensional Kelvin-Helmholtz instability, including a relaxed entropy-inequality variant in the appendix.","tokens_in":16396,"tokens_out":10041,"duration_ms":101120,"significance":"If the positivity-preservation claim is fully established, the paper makes a valuable contribution: it transfers two recently developed DG stabilization strategies (entropy-correction artificial viscosity and knapsack limiting) to finite difference methods, gives an explicit and cheap optimization solution, and demonstrates high-order convergence and robustness in challenging benchmarks. The Section 3 entropy-stability derivation is clean and largely self-contained, and the symmetrization argument is simple and convincing. However, the headline positivity claim for KL-FD is asserted rather than proved in this manuscript, and the Navier-Stokes claim and the HLLC-based numerical scheme are not covered by the presented theory. These gaps substantially affect the significance of the paper as a proof-of-concept, though the numerical evidence suggests the methods work in practice.","major_comments":[{"comment":"The positivity-preservation guarantee for KL-FD is asserted, not derived. The text states that limiting coefficients ℓc can be derived so that θ ∈ [ℓc,1] implies the relative positivity condition (4.3), but no existence proof is given, and the computational formula (5.1) is introduced in Section 5 without derivation. The formula uses nodal totals r_i^H, r_i^L, while the actual scheme uses edgewise coefficients that are symmetrized and then re-optimized in (4.4); it is not shown that the nodal construction survives these steps for the SBP finite-difference stencils of Section 2. This is the core of the abstract's 'provably preserves positivity' claim and must be proved or explicitly reduced to a precise theorem.","section":"Section 4, Eqs. (4.3)–(4.4) and Eq. (5.1)"},{"comment":"The abstract claims positivity for the compressible Euler and Navier-Stokes equations, but the manuscript only defines the hyperbolic conservation law (2.1). No Navier-Stokes or viscous discretization, viscous flux, or corresponding boundary treatment is presented anywhere. Unless the claim is strictly inherited from the low-order scheme of [11], the transfer to the high-order KL-FD scheme is not established. The claim should be restricted to the compressible Euler equations or supported by a concrete viscous discretization and proof.","section":"Abstract and Section 4"},{"comment":"The key identity f_KL(θ) = (1−θ)f_H + θf_L is proved only for the Lax-Friedrichs flux, using Eq. (2.7). The paper then states that HLLC can be used and that 'the same strategies apply,' and all numerical positivity tests use KL-FD-HLLC. For HLLC, f_L − f_H is not proportional to u_i − u_j, so neither the convex blending used in (4.3) nor the nonnegativity of a_ij = ||n_ij||(v_j−v_i)^T(f_H−f_L) is automatic. The theory therefore does not currently cover the scheme that is actually tested.","section":"Section 4, HLLC paragraph"},{"comment":"The positivity proof and the one-iteration convergence of the quadratic knapsack algorithm are delegated to [4] and [11], both of which are discontinuous Galerkin/subcell-limiting papers. The transfer from DG subcell limiting to flux-differenced finite-difference SBP discretizations is not automatic and is not argued in the manuscript. Since the abstract's central claim is a provable positivity guarantee, the relevant theorem should be stated and proved here, or the claim should be marked as conditional on the cited DG theory.","section":"Section 4 and Appendix (8.1)–(8.2)"}],"minor_comments":[{"comment":"The limiting-coefficient formula (5.1) is not clamped to [0,1]. If the formula can produce values outside this range, the subsequent interpretation of θ + ℓc as a convex blending coefficient is unclear; please state the range or add clamping.","section":"Eq. (5.1)"},{"comment":"The entry '2.53¨ 105' should presumably be '2.53×10^{-5}'. Please correct the exponent formatting.","section":"Table 3"},{"comment":"The text refers to the 'Sod shocktube problem' and to Figure (1), but Figure (1) displays the Shu-Osher shock tube from Section 5.2. Please clarify whether the experiments in Section 8.1 are Shu-Osher or Sod, and adjust the figure references accordingly.","section":"Section 8.1"},{"comment":"The abstract describes both schemes as 'hyperparameter free,' but KL-FD with positivity uses the relative positivity parameter α, and the experiments use α = 0.5 and α = 0.1. Please either clarify that α is a user-specified safety factor rather than a fitted hyperparameter, or modify the hyperparameter-free claim.","section":"Abstract and Section 5.3"}],"recommendation":"major_revision","confidential_remarks":"The Section 3 entropy-stability construction is clean and could be publishable on its own. The main concern is that the paper's headline novelty—provable positivity for KL-FD, including Navier-Stokes—is not actually proved in this manuscript and depends on the authors' own preprint [4] and on DG subcell-limiting theory whose transfer to finite differences is not automatic. The HLLC discrepancy between theory and numerics is also significant. I do not recommend rejection because the gaps appear fillable, but they are substantive and require more than local edits."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The part that holds up is the entropy stability machinery. The derivation of the nodal entropy inequality (3.7), the explicit optimizer in Lemma 3.2, and the max-symmetrization that preserves both conservation and entropy stability (Lemma 3.3) are clean and self-contained. That is a genuine extension of ECAV to SBP finite differences, and it reads like a real result rather than an incremental tweak. The weak-boundary treatment is straightforward and useful.\n\nThe positivity story is the soft spot. The abstract promises provable positivity for Euler and Navier-Stokes, but the manuscript only sketches the knapsack machinery. Formula (5.1) for ℓ^c appears without derivation; the proof of the relative positivity condition is imported from the authors' DG papers [4, 11]. That transfer is not automatic: FD stencils couple nodes differently from DG subcell limiting, and the text itself notes the equivalence to the weighted diffusion form is only shown for LxF, while the numerics use HLLC. Also, the Navier-Stokes claim in the abstract has no discretization behind it in this paper. And the user-chosen α in the relative positivity constraint sits uneasily with the \"hyperparameter free\" label.\n\nNone of this kills the main contribution. The entropy-stable ECAV-FD part is provable and demonstrated. KL-FD is plausible and the experiments are encouraging, but the headline claim is ahead of the proof. The timing study mentioned in the introduction is also absent, though the paper does include convergence tables and several shock tests.\n\nWho should read it: anyone building entropy stable finite difference codes or comparing stabilization strategies. It deserves a serious referee; the revision should be asked to derive or properly reference the positivity lemma for the FD case, either prove formula (5.1) or state it as a lemma with a proof, define the Navier-Stokes discretization or soften the claim, and report timings. I would send it to review rather than desk reject.","headline":"Solid entropy-stability extension to finite differences; positivity claim needs a real proof before it carries weight.","tokens_in":16778,"tokens_out":2512,"would_cite":true,"duration_ms":22112,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["68Q25","68R10","68U05"],"pacs":[],"model":"deepseek-v4-flash","headline":"High-order finite difference schemes can be made entropy stable and positivity preserving by choosing diffusion coefficients from a tiny quadratic program.","keywords":["entropy stability","finite difference methods","artificial viscosity","knapsack limiting","positivity preservation","compressible Euler equations","Navier-Stokes equations","summation-by-parts operators"],"falsifier":"Run KL-FD with the stated limiting formula on a two-dimensional compressible Euler benchmark, such as the Kelvin-Helmholtz instability at order N=6 with 512 nodes per dimension and a timestep near the stability limit, and check whether any cell attains negative density or pressure; a single occurrence with the relative positivity parameter α > 0 would falsify the positivity-preservation claim as stated. Alternatively, verify the relative positivity inequality (4.3) cell-wise at every Runge-Kutta stage for a smooth periodic initial condition with N≥4; a violation would show the symmetrized limi","tokens_in":15983,"feed_emoji":"🌊","tokens_out":5885,"duration_ms":58219,"temperature":0.7,"pith_summary":"This paper transplants two stabilization techniques from discontinuous Galerkin methods to high-order finite difference discretizations of compressible flow: entropy-correction artificial viscosity and knapsack limiting. The resulting schemes satisfy a discrete entropy inequality, a discrete analog of the second law of thermodynamics, without any user-tuned hyperparameters. The knapsack-limited variant additionally provably keeps density and pressure positive for the compressible Euler and Navier-Stokes equations. If correct, this gives computational fluid dynamics a class of finite difference solvers that are simultaneously high-order accurate, entropy stable, positivity preserving, and computationally cheap.","feed_headline":"Per-node optimization yields entropy-stable, positivity-preserving finite difference schem","feed_subtitle":"Knapsack limiting extends the same rule to provably keep density and pressure positive for Euler and Navier-Stokes.","key_machinery":"The load-bearing object is the parametrized flux fij(θ) = f^H_ij + θ(ui - uj), where f^H is the high-order central flux and θ is a diffusion coefficient proportional to added dissipation. Setting θ = (λij/2)θ̃ recasts this flux as a convex combination (1-θ̃)f^H + θ̃ f^L, the knapsack-blended flux, where f^L is a low-order entropy-stable and positivity-preserving flux. A per-node quadratic program selects θ; the unconstrained problem has a closed-form solution, while the knapsack-constrained problem is solved by a one-pass algorithm. Symmetrization θij = max{θ̂ij, θ̂ji} ensures conservation and suffices for the entropy inequality.","core_discovery":"The central claim is that a high-order summation-by-parts finite difference discretization of a hyperbolic conservation law can be made entropy stable by adding diffusion coefficients θij to the central flux, with θij chosen as the minimal-norm solution of a per-node linear inequality a_i^T θ_i ≥ b_i. The paper proves in Lemma 3.2 that this optimization problem has a closed-form solution, so the scheme is hyperparameter free; symmetrization θij = max{θ̂ij, θ̂ji} preserves conservation while retaining entropy stability (Lemma 3.3). For the knapsack-limited scheme, the same flux is reinterpreted as a convex blend between the high-order central flux and a low-order positivity-preserving Lax-Fri","pith_inferences":["Because the machinery is written dimension-agnostically and the optimization is per node, the same per-node quadratic program could be applied directly to other summation-by-parts discretizations, such as discontinuous Galerkin spectral element methods, without modification; the paper gestures at this through its DG lineage but does not state it as a general recipe.","The entropy-stability proof only requires a_i ≥ 0 and symmetrization, so the argument likely extends to any convex entropy pair, not just the physical Euler entropy, suggesting applications to other hyperbolic systems such as magnetohydrodynamics provided a positivity-preserving low-order flux exists.","The relaxed entropy inequality in Section 8 introduces a dissipation-versus-oscillation trade-off that the paper documents numerically but does not rigorously bound; a natural next step would be to derive an a posteriori measure of entropy-dissipation error for the relaxed schemes.","The relative-positivity limiting formula (5.1) is imported from DG theory; a valuable test would be to verify it on genuinely multi-dimensional stencils and non-diagonal SBP norms, where the symmetrization step may interact differently with the limiting coefficients."],"forward_implications":["Entropy-stable finite difference solvers for the compressible Euler and Navier-Stokes equations can be constructed without tunable parameters, at computational cost comparable to standard stabilized schemes.","Positivity of density and pressure is guaranteed for high-order finite difference approximations, removing a common failure mode for under-resolved or shock-dominated flows.","The knapsack-flux viewpoint unifies artificial viscosity and flux limiting: the same diffusion coefficient enforces the entropy inequality and blends the low- and high-order fluxes.","High spatial order, up to the stencil accuracy N, is preserved for smooth solutions, so the stabilization does not degrade accuracy.","The construction extends to multi-dimensional summation-by-parts operators and weakly enforced Dirichlet boundary conditions."],"supporting_citations":[{"why":"Supplies the quadratic knapsack limiting algorithm and the positivity-preserving theory for discontinuous Galerkin methods that the finite difference knapsack scheme extends.","marker":"[4]"},{"why":"Supplies the positivity-preserving low-order scheme and the relative positivity condition (4.3) for the compressible Euler and Navier-Stokes equations.","marker":"[11]"},{"why":"Supplies the entropy correction artificial viscosity idea from high-order discontinuous Galerkin methods that the paper adapts to finite differences.","marker":"[1]"},{"why":"Supplies the split-form summation-by-parts flux differencing technique used to write the high-order finite difference discretization.","marker":"[6]"},{"why":"Supplies the entropy-stable flux framework and weakly enforced boundary treatment used in deriving the discrete entropy inequality.","marker":"[3]"},{"why":"Supplies the theory that subcell/knapsack limiting preserves high order accuracy in smooth regions, supporting the accuracy claim.","marker":"[20]"},{"why":"Supplies the HLLC flux used as the entropy-stable, positivity-preserving low-order flux in the KL-FD-HLLC scheme.","marker":"[9]"}],"fun_headline_variants":["Closed-form fix makes finite difference schemes entropy stable","Knapsack limiting brings positivity to finite difference methods","Entropy-stable finite differences with zero tuning parameters","Knapsack limiting proves positivity for Euler and Navier-Stokes","Per-node diffusion achieves entropy stability in finite difference schemes"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The positivity-preservation claim rests on formula (5.1) and the cited positivity theory, neither derived nor proven inside this paper, holding for arbitrary high-order finite difference stencils; if that limiting-coefficient formula fails for a given stencil, the positivity guarantee collapses even though the entropy inequality still holds.","fun_headline_variants_meta":{"raw":{"variants":["Closed-form fix makes finite difference schemes entropy stable","Knapsack limiting brings positivity to finite difference methods","Entropy-stable finite differences with zero tuning parameters","Knapsack limiting proves positivity for Euler and Navier-Stokes","Per-node diffusion achieves entropy stability in finite difference schemes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000207,"raw_usage":{"total_tokens":1248,"prompt_tokens":764,"completion_tokens":484,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":508,"completion_tokens_details":{"reasoning_tokens":416}},"tokens_in":508,"tokens_out":484,"duration_ms":4853,"temperature":1.0,"reasoning_tokens":416,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T14:28:02.131309+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run KL-FD with the stated limiting formula on a two-dimensional compressible Euler benchmark, such as the Kelvin-Helmholtz instability at order N=6 with 512 nodes per dimension and a timestep near the stability limit, and check whether any cell attains negative density or pressure; a single occurrence with the relative positivity parameter α > 0 would falsify the positivity-preservation claim as stated. Alternatively, verify the relative positivity inequality (4.3) cell-wise at every Runge-Kutta stage for a smooth periodic initial condition with N≥4; a violation would show the symmetrized limi","supporting_citations":[{"cited_title":"A positivity preserving strategy for entropy stable discontinuous Galerkin discretizations of the compressible Euler and Navier-Stokes equa- tions","cited_arxiv_id":null,"evidence_quote":"Supplies the positivity-preserving low-order scheme and the relative positivity condition (4.3) for the compressible Euler and Navier-Stokes equations."},{"cited_title":"Gassner, Andrew R","cited_arxiv_id":null,"evidence_quote":"Supplies the split-form summation-by-parts flux differencing technique used to write the high-order finite difference discretization."},{"cited_title":"Local subcell monolithic DG/FV convex property preserving scheme on un- structured grids and entropy consideration","cited_arxiv_id":null,"evidence_quote":"Supplies the theory that subcell/knapsack limiting preserves high order accuracy in smooth regions, supporting the accuracy claim."},{"cited_title":"Lax, and Bram van Leer","cited_arxiv_id":null,"evidence_quote":"Supplies the HLLC flux used as the entropy-stable, positivity-preserving low-order flux in the KL-FD-HLLC scheme."}],"review_version":1}