{"id":"7a8f0aa9-deb1-4004-8819-6f788031f28a","arxiv_id":"2608.04283","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A weighted upwind equilibrium distribution for vector kinetic lattice Boltzmann methods, using a smooth eigenvalue-based flux split, improves stability and accuracy for hyperbolic conservation laws.","lead":"This paper introduces a new version of a lattice-based numerical solver for hyperbolic conservation laws, the equations behind waves, shocks, and fluid and plasma flow. The new version smooths the way the solver splits left- and right-moving signals, which keeps it stable at sonic points and more accurate on benchmark tests.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The reported stability gain over discontinuous upwinding may be caused by the added c-diffusion rather than the sigmoid splitting; the paper omits c=0 control runs in all eigenvalue-sign-change benchmarks.","rationale":"The reader's weakest assumption is the heuristic extension of the c-bound (51) from the proven domain (homogeneous degree-one flux, constant Jacobian, k to infinity) to general nonlinear hyperbolic systems with finite k=2. This is a legitimate concern about parameter justification, and the paper explicitly discloses the limitation. However, I find a more load-bearing concern in the interpretation of the stability results: the weighted upwind set is always compared with c=0.05, while the discontinuous upwind set has no analogous diffusion term. Since the paper's own Theorem 4.2 and the finite-volume flux (49) identify c as a numerical diffusion coefficient, the stability improvement in eigenvalue-sign-change tests may be due to added diffusion rather than to the continuous sigmoid splitting. This directly affects the central mechanistic claim that the smooth flux split regularizes the sign-change instability. The c=0 control experiment is simple and would settle the attribution. The reader's c-bound concern, if it lands, would affect the theoretical guarantee but not necessarily the numerical performance; the c=0 control affects whether the headline contribution is genuinely the continuous splitting or merely the addition of a known diffusive term. The paper does provide credible algebraic verification of moment constraints, the finite-difference equivalence, and the finite-volume conservation form for omega=1, and the numerical benchmarks consistently show lower errors and improved stability for the full weighted upwind method. These elements support the accuracy and practical utility of the scheme as a whole, which is why I do not recommend moving away from CONDITIONAL. The missing artifact (no code or data) noted by the reader further supports a conditional acceptance. My concern reinforces the need for additional control experiments and would not by itself justify rejection, so the reader's CONDITIONAL verdict remains unchanged.","tokens_in":28990,"tokens_out":9277,"duration_ms":78806,"concrete_test":"Run the hydraulic flow and fixed-sonic-point Euler problems (Table 1, Table 2) with the weighted upwind equilibrium set (22)-(24) using c=0 and k=2, with all other settings identical (CFL=0.9, omega=1, same grids). Compare stability and L1 errors against the c=0.05 runs and the discontinuous upwind set. If c=0 reproduces the discontinuous upwind oscillations, the stability advantage is attributable to the c-diffusion term rather than the sigmoid splitting; if c=0 remains stable, the sigmoid alone is sufficient.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that the weighted upwind set (22)-(24), built from the smooth sigmoid split (25)-(27), is more stable than the discontinuous upwind set (12)-(14) when eigenvalues change sign. The paper's own analysis shows that c multiplies a Laplacian stencil (Theorem 4.2) and that the finite-volume flux (49) becomes a Rusanov/Lax-Friedrichs-type flux with diffusion coefficient c*xi. In every comparison where the discontinuous upwind set becomes unstable (hydraulic flow, compression flow, Leblanc, fixed sonic point, smooth Euler problems), the weighted upwind set is run with c=0.05 while the discontinuous set has no such term. No c=0 run of the weighted upwind set is reported for these problems. Consequently, the observed suppression of the eigenvalue-sign-change instability is confounded with the additional numerical diffusion. The smooth sigmoid may still contribute, but the evidence presented does not isolate its effect; the stability improvement could be mainly an artifact of adding a diffusion term, which would undercut the paper's emphasis on continuous flux splitting as the stabilizing mechanism.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a weighted upwind vector kinetic lattice Boltzmann (WU-VKLB) method for hyperbolic conservation laws. The equilibrium set (22)-(24) is built from the conserved state U and partial fluxes obtained from a smooth sigmoid flux splitting (25)-(27) based on the eigendecomposition of the flux Jacobian; it reduces to the centered and discontinuous upwind sets in limiting cases. The authors derive equivalent finite-difference and finite-volume formulations (Theorems 4.1-4.3), show that the c-term acts as a diffusive stencil and yields an LLF-type numerical flux near zero eigenvalues, and prove an upper bound on c under restrictive assumptions (Theorem 4.4). Numerical tests on shallow water, Euler, and ideal MHD benchmarks report improved stability and accuracy relative to the centered and discontinuous upwind VKLB sets, together with second-order convergence on smooth problems.","tokens_in":29102,"tokens_out":9195,"duration_ms":79674,"significance":"If the claims withstand scrutiny, the method is a useful and fairly general construction: it gives a conservative, eigendecomposition-based upwinding framework for VKLB schemes, with explicit analysis of the added dissipation and demonstrated second-order convergence on smooth problems. The manuscript's algebraic core is clean: the moment constraints (3) and (5) are satisfied by construction, the arithmetic in Theorems 4.2 and 4.3 checks out, and Theorem 4.4 is correct under its stated hypotheses. The numerical test suite is broad (shallow water, Euler, and MHD with divergence cleaning), and the Orszag-Tang runs with c=0 and c=0.05 are a good start toward separating the effect of the diffusion parameter. The main weakness is that the central stability claim, namely that the sigmoid splitting rather than the added c-diffusion suppresses eigenvalue-sign-change instabilities, is not yet cleanly demonstrated, and the parameter bound is used heuristically outside its proven domain.","major_comments":[{"comment":"The central claim that the continuous sigmoid splitting stabilizes the scheme when eigenvalues change sign is confounded by the added c-diffusion. In the hydraulic-flow, compression-flow, Leblanc, and fixed-sonic-point benchmarks, the weighted upwind set is run with c=0.05 while the discontinuous upwind set has no analogous c-term, and no c=0 control run of the weighted upwind set is reported for these problems. Theorem 4.2 and Eq. (50) show that c introduces a Laplacian stencil and an LLF-type numerical flux with diffusion coefficient c*xi, and c=0.05 is the maximum value allowed by (51) at nu=0.9, so this term is not negligible. The Orszag-Tang comparison with c=0 and c=0.05 (Fig. 16) is helpful but does not isolate the eigenvalue-sign-change behavior. Please add c=0 runs, and if feasible intermediate c values, for the sign-change benchmarks, or otherwise separate the stabilizing effect of the sigmoid splitting from that of the c-diffusion; without this, the paper's emphasized mechanism is not established.","section":"5.1, 5.2 (Figs. 8-9, 13-14)"},{"comment":"The bound 0 <= c <= (1/2)(1-nu) is proven only for fluxes homogeneous of degree one with constant Jacobian in the limit k -> infinity, yet it is used to set c=0.05 with k=2 in all benchmark systems, including shallow water and MHD where the homogeneity assumption fails, and for the 2D set (41) where no analogous bound is derived. The 'Taylor series type expansion argument' in Section 4.2 is only a heuristic, and the manuscript itself flags this at the end of Theorem 4.4. Because the stability and accuracy claims in Section 5 all depend on this parameter choice, the paper needs either a proof covering non-homogeneous fluxes, finite k, and the 2D set, or a systematic numerical verification that the chosen c remains within a stable and entropy-consistent regime for each system. As written, the theoretical underpinning of the main free parameter is incomplete.","section":"4.2, Theorem 4.4"},{"comment":"In the Brio-Wu test the comparison also mixes CFL and c: the discontinuous upwind set is limited to nu=0.45 while the weighted upwind set runs at nu=0.90 with c=0.05, so the reported stability gain at the larger CFL may again be due to c-diffusion rather than the sigmoid splitting. A matched comparison (weighted upwind at c=0 and nu=0.90, or discontinuous upwind with an equivalent added diffusion) would clarify whether the sigmoid contributes independently of the diffusion.","section":"5.3, Fig. 15"}],"minor_comments":[{"comment":"There are several typos and incorrect equation references: 'equations (2)' in Theorem 4.4 should be (27); 'in (4.4)' in Section 5 should be 'Theorem 4.4'; 'weighted upwind set (22)-(22)' in Section 5.3 should be (22)-(24); and 'discontinuos', 'T otal Energy', 'Fintie', 'the the', and 'and and' appear in the text.","section":"Throughout"},{"comment":"The algorithm in Figure 5 says the streaming speed is updated through 'CFL/subcharacteristic constraint,' but for the discontinuous and weighted upwind sets only the CFL constraint is analyzed in Section 3.2; please state explicitly which constraint is used for each equilibrium set.","section":"3.2, Figure 5"},{"comment":"Figure 8 (hydraulic flow) uses CFL=0.80 while Section 5 states that nu=0.90 unless otherwise noted; since the c-bound (51) is CFL-dependent, please state the CFL in each figure caption and confirm that c=0.05 is within the applicable bound for that nu.","section":"5.1, Figure 8"},{"comment":"The sentence 'This consistency argument can be applied verbatim to the weighted upwind equilibrium distribution function set (22)-(24) when c=0' is slightly misleading, because the partial fluxes (26) are not identical to the discontinuous ones at finite k; the consistency is standard for any consistent splitting, but the wording should be adjusted.","section":"4.1"},{"comment":"Equation (27) defines alpha using max_j |lambda_j|; when all eigenvalues are zero, this quotient is undefined. For completeness, specify the convention in that degenerate case (for example, alpha=1/2).","section":"3.1, Eq. (27)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is in scope for a numerical analysis journal and the construction is plausible. The main barrier is not the algebra but the experimental design separating the c-diffusion from the sigmoid splitting, together with the heuristic parameter bound. If the authors supply c=0 control runs for the sign-change benchmarks and either extend or more carefully qualify the c-bound, the paper would be a solid contribution. I would not reject on the current evidence."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague — quick read on arXiv:2608.04283. The genuinely new pieces are the weighted upwind equilibrium set (22)-(24) with the smooth sigmoid flux split (25)-(27), the finite-volume flux (49), and the c-bound (51). I checked the algebra: the moment constraints hold, the finite-difference form (46) follows, and the finite-volume form is consistent. The numerical work is extensive — shallow water, Euler, MHD — and the reported accuracy gains over the centered flux set are credible. The method is also not circular: c and k are chosen from stability/regularity reasoning, not fitted to the benchmarks.\n\nThe main soft spot is the central stability claim. In every 1D benchmark where the discontinuous upwind set blows up — hydraulic flow, compression flow, Leblanc, fixed sonic point, the smooth Euler problems — the weighted upwind set runs with c=0.05 and the discontinuous set has no c term at all. Theorem 4.2 shows c multiplies a Laplacian, and Theorem 4.3 shows the flux becomes an LLF-type flux with diffusion coefficient c*xi. So the added diffusion alone could explain the stability gain; the sigmoid splitting's contribution is not isolated. A c=0 control run for the weighted upwind set in those benchmarks would settle it. The one c=0 run shown, Orszag-Tang, is a 2D MHD problem where c=0 is actually stable and more accurate — which helps but doesn't directly address the 1D eigenvalue-sign-change cases.\n\nSecond soft spot: the c-bound (4.4) is proven only in the limit k to infinity, for homogeneous degree-one flux with constant Jacobian, then used heuristically for k=2 on nonlinear systems. The authors disclose this, but it means the choice c=0.05 rests on a Taylor-series-style guess outside the proven class. A sensitivity study in c and k for one or two nonlinear problems would tighten this considerably.\n\nThird, no code or data accompanies the preprint. For a method paper, that's a real limitation, though likely fixable on request.\n\nOverall, the paper is honest, well-written, and the core construction is sound. The confounding issue is real and should be addressed, but it's the kind of thing a good revision can fix. I'd send it to peer review. It deserves a serious referee, and with control runs plus a sensitivity study it could be a solid contribution to the VKLB/flux-splitting literature.","headline":"A genuinely new weighted upwind equilibrium set with a sigmoid flux split, but the stability gain over discontinuous upwinding is confounded with added c-diffusion; still referee-worthy.","tokens_in":29781,"tokens_out":3075,"would_cite":true,"duration_ms":26242,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35L65","65M08","65M12","76M28"],"pacs":[],"model":"deepseek-v4-flash","headline":"A sigmoid-smoothed eigenvalue flux split inside the vector kinetic lattice Boltzmann equilibrium set yields a scheme stable at sign-changing wave speeds and more accurate than both the upwind and centered parent sets.","keywords":["vector kinetic lattice Boltzmann","flux splitting","hyperbolic conservation laws","equilibrium distribution functions","upwind scheme","shallow water equations","Euler equations","ideal magnetohydrodynamics"],"falsifier":"Run the weighted upwind scheme on the shallow water hydraulic-flow problem or the Euler sonic-point problem with $c=0.05$, $k=2$, and $\\nu=0.9$: if spurious oscillations appear at the rarefaction or the pressure or density becomes negative, the claimed stability advantage over the discontinuous upwind set is not robust. More directly, compute the eigenvalues of $\\partial_U f_2^{eq}$ for the shallow water and MHD systems at representative states with $c=0.05$ and $k=2$; if any eigenvalue is negative, the monotonicity condition that justifies the $c$-bound is violated, and the heuristic extension on which the test settings rely is not sound.","tokens_in":28660,"feed_emoji":"🌊","tokens_out":12739,"duration_ms":103584,"temperature":0.7,"pith_summary":"This paper aims to give vector kinetic lattice Boltzmann (VKLB) methods a general and reliable upwinding mechanism for hyperbolic conservation laws. The central claim is that a weighted upwind equilibrium distribution set, built by splitting the physical flux continuously with a sigmoid weighted by the eigenvalues of the flux Jacobian, is stable where the discontinuous upwind equilibrium set develops oscillations (whenever an eigenvalue changes sign, as at a sonic point), and is more accurate than both the discontinuous upwind and centered flux equilibrium sets on shallow water, Euler, and ideal MHD tests. The scheme matters because it ties upwinding only to the eigenstructure of the system, so the same construction applies to any hyperbolic system with a well-defined eigendecomposition, and because the extra diffusive term proportional to the conserved vector is shown to be the source of the added stability. The paper also proves an equivalent finite volume form with telescoping internal fluxes and shows that the weighted set retains second-order accuracy on smooth problems.","feed_headline":"Sigmoid flux split stabilizes upwind lattice Boltzmann methods","feed_subtitle":"A weighted equilibrium beats the centered and discontinuous upwind versions on shallow water, Euler, and MHD tests.","key_machinery":"The load-bearing object is the weighted upwind equilibrium distribution set (22)-(24) together with the sigmoid flux split (25)-(27): a continuous decomposition of the physical flux obtained by expanding the flux in right eigenvectors of the flux Jacobian and weighting each characteristic coefficient by a sigmoid of the corresponding eigenvalue normalized by the largest wave speed. The sigmoid makes the split transition smoothly from centered to fully upwind as a characteristic speed moves away from zero, which is what regularizes the eigenvalue sign-change instability, while the parameter $c$, which controls how much of the conserved vector sits in the equilibria, acts as an additional diffusive stencil whose size is constrained by the monotonicity and entropy argument. The equivalent finite volume rewrite with interface flux $\\mathcal{F}_{i+1/2}=(\\mathcal{F}^{(+)}_i+\\mathcal{F}^{(-)}_{i+1})-c\\xi(U_{i+1}-U_i)$ shows that the scheme is conservative and that the additional term is a controllable local Lax-Friedrichs-type diffusion.","core_discovery":"On the paper's own terms, the discovery is a family of equilibrium distribution functions\n$$\n$f_1^{{eq}}$=cU+\\frac{1}{\\xi}\\mathcal{F}^{(+)}(U),\\qquad\n$f_2^{{eq}}$=(1-2c)U-\\frac{1}{\\xi}\\big(\\mathcal{F}^{(+)}(U)-\\mathcal{F}^{(-)}(U)\\big),\\qquad\n$f_3^{{eq}}$=cU-\\frac{1}{\\xi}\\mathcal{F}^{(-)}(U),\n$$\nin which the partial fluxes come from a continuous eigendecomposition-based splitting with coefficient $\\alpha(\\lambda/|\\lambda_{\\max}|;k)=1/(1+e^{-k\\lambda/\\max_j|\\lambda_j|})$. The parameter $c$ adds a diffusive stencil acting directly on the conserved variables, visible in the equivalent finite difference update (46), and must be nonnegative for that stencil to be well posed; the monotonicity criterion of Definition 1 gives the bound $0\\le c\\le \\tfrac{1}{2}(1-\\nu)$ under the stated restrictions of homogeneous degree-one fluxes, a constant Jacobian, and $k\\to\\infty$. The sigmoid splitting interpolates between the centered flux equilibrium set and the discontinuous upwind set, so the weighted set reduces to the centered set as an eigenvalue approaches zero and to the discontinuous upwind set as $k\\to\\infty$. Numerical tests on shallow water, Euler, and ideal MHD problems show that this interpolation suppresses the sign-change instabilities of the discontinuous upwind set, improves resolution relative to the centered set, and preserves second-order convergence on smooth problems.","pith_inferences":["The transition width $k$ in the sigmoid is fixed at $k=2$ in all tests; one could make $k$ depend on the local eigenvalue spread or the CFL number, letting the scheme become more upwind in smooth regions and more centered near sign changes.","Because the finite volume interface flux reduces to a local Lax-Friedrichs flux at eigenvalue sign changes, the weighted upwind construction could be analysed with the existing finite volume stability toolbox, not only with kinetic entropy arguments.","If the heuristic extension of the $c$-bound to non-homogeneous fluxes fails, a numerical check of entropy decay or of the monotonicity condition on the actual shallow water and MHD states would reveal how much of the stated stability margin is carried by the heuristic rather than the theorem.","The same sigmoid splitting could be combined with a positivity- and bounds-preserving limiter, which the authors identify as the natural next step, to make the stability gains useful for very strong MHD shocks where negative pressure is the practical failure mode."],"forward_implications":["Any hyperbolic system with a real eigendecomposition of its flux Jacobian can be given the same upwind weighted equilibrium construction, so the method is not tied to the particular equations tested.","On smooth problems the scheme keeps second-order accuracy with $\\omega=2$, while the discontinuous upwind set is unstable on the same smooth Euler tests, so the stability gain does not cost formal order.","In shock problems the weighted set runs at the same CFL number as the centered flux set, whereas the discontinuous upwind set required roughly half the CFL on the one-dimensional MHD shock tube, meaning the added dissipation does not degrade the time-step limit.","The scheme satisfies a discrete conservation law via telescoping internal fluxes in its finite volume form, which is the property needed for credible shock capturing.","The bound $0\\le c\\le \\tfrac{1}{2}(1-\\nu)$ gives a principled range for the extra diffusion when the assumptions of Theorem 4.4 hold, and the finite volume form identifies $c=\\nu/2$ with standard local Lax-Friedrichs diffusion."],"supporting_citations":[{"why":"Supplies the vector kinetic lattice Boltzmann framework and the centered and discontinuous upwind equilibrium sets that the weighted set generalizes.","marker":"[1]"},{"why":"Introduces the discrete vector kinetic (BGK) schemes for multidimensional conservation laws from which the VKLB discretization is built.","marker":"[3]"},{"why":"Gives the monotone non-decreasing criterion, Definition 1, used to derive the upper bound on the diffusion parameter c.","marker":"[9]"},{"why":"Provides the CFL constraint, flux splitting terminology, and finite volume conservation argument used in Sections 3.2 and 4.1.","marker":"[26]"},{"why":"Defines the classical flux-vector splitting whose discontinuous limit the weighted sigmoid splitting recovers for homogeneous degree-one fluxes.","marker":"[34]"},{"why":"Supplies the hyperbolic divergence cleaning that makes the ideal MHD system (58) a hyperbolic conservation law, one of the central test systems.","marker":"[13]"},{"why":"Provides the high-resolution reference solutions against which the MHD shock tube and two-dimensional vortex benchmark accuracy is measured.","marker":"[35]"},{"why":"Supplies the exact Riemann solution data used for the Euler convergence studies.","marker":"[39]"}],"fun_headline_variants":["Weighted upwind kinetic lattice Boltzmann tames hyperbolic shocks","Sigmoid flux split sharpens lattice Boltzmann shock resolution","Weighted upwind equilibrium boosts lattice Boltzmann stability","Interpolated upwind flux family stabilizes hyperbolic lattices"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the entropy and monotonicity bound $0\\le c\\le \\tfrac{1}{2}(1-\\nu)$, proved only for homogeneous degree-one fluxes with a constant Jacobian in the $k\\to\\infty$ limit, still holds approximately for the nonlinear shallow water, Euler, and ideal MHD systems at the finite value $k=2$ actually used in the numerical experiments.","fun_headline_variants_meta":{"raw":{"variants":["Weighted upwind kinetic lattice Boltzmann tames hyperbolic shocks","Sigmoid flux split sharpens lattice Boltzmann shock resolution","Weighted upwind equilibrium boosts lattice Boltzmann stability","Interpolated upwind flux family stabilizes hyperbolic lattices"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000397,"raw_usage":{"total_tokens":2115,"prompt_tokens":1018,"completion_tokens":1097,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":634,"completion_tokens_details":{"reasoning_tokens":1031}},"tokens_in":634,"tokens_out":1097,"duration_ms":7839,"temperature":1.0,"reasoning_tokens":1031,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T14:42:37.752947+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the weighted upwind scheme on the shallow water hydraulic-flow problem or the Euler sonic-point problem with $c=0.05$, $k=2$, and $\\nu=0.9$: if spurious oscillations appear at the rarefaction or the pressure or density becomes negative, the claimed stability advantage over the discontinuous upwind set is not robust. More directly, compute the eigenvalues of $\\partial_U f_2^{eq}$ for the shallow water and MHD systems at representative states with $c=0.05$ and $k=2$; if any eigenvalue is negative, the monotonicity condition that justifies the $c$-bound is violated, and the heuristic extension on which the test settings rely is not sound.","supporting_citations":[{"cited_title":"Discrete kinetic schemes for multidimensional systems of conservation laws","cited_arxiv_id":null,"evidence_quote":"Introduces the discrete vector kinetic (BGK) schemes for multidimensional conservation laws from which the VKLB discretization is built."},{"cited_title":"Finite volume methods for hyperbolic problems","cited_arxiv_id":null,"evidence_quote":"Provides the CFL constraint, flux splitting terminology, and finite volume conservation argument used in Sections 3.2 and 4.1."},{"cited_title":"Riemann Solvers and Numerical Methods for Fluid Dynamics: A Practical Introduction","cited_arxiv_id":null,"evidence_quote":"Supplies the exact Riemann solution data used for the Euler convergence studies."}],"review_version":1}