{"id":"0869334b-1141-498b-af69-809635b44586","arxiv_id":"2608.08117","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A family of weighted compact least-squares schemes with boundary-value-difference penalties delivers high-order accuracy and low-dissipation shock capturing for compressible Navier-Stokes equations on curvilinear grids.","lead":"This paper develops three new high-resolution numerical schemes, from third to fifth order, for simulating compressible gas flows that contain both smooth turbulence and sharp shocks. The schemes aim to give engineers a more accurate, less dissipative way to compute flows around aircraft and inside engines.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Formal-order guarantee hinges on unproved theta = O(h^k) scaling at smooth extrema; Eq. (21) may give theta = O(h^{k-1}) there, leaving the 'even near smooth extrema' claim supported only by numerics.","rationale":"After a careful read, I agree with the Reader's weakest_assumption: the most load-bearing uncertainty is the BVD-penalty scaling near smooth extrema. The paper's Theorem 1 proof is a sufficient-condition argument, and Eq. (21) does not obviously deliver the required theta = O(h^k) when the denominator sum |u_i - u_m| is O(h^2). The numerical accuracy tests in Sec. 4.1 are genuinely reassuring and are the right kind of evidence: they use exactly the profiles (sin^m, m = 1, 2, 3) that create first- and second-order extrema and report full designed orders. That is why I do not escalate the verdict. What is missing is a direct measurement of theta scaling or a refined analytical bound; as written, the claim 'formal order is preserved even near smooth extrema' is supported by experiment rather than by the theorem the paper offers. Other potential concerns, such as the frozen-beta/theta GMRES linearization, the tuned constants, and the absence of deposited code, are secondary and do not change the conditional assessment. My proposed check would settle whether the concern is real or merely a proof gap: if theta actually scales as h^k at the extremum, the paper's mechanism is exactly as stated; if it scales as h^{k-1}, the numerical order preservation needs an explanation beyond Theorem 1, and the central claim would be conditional on tuning that is not analyzed.","tokens_in":32387,"tokens_out":14985,"duration_ms":159120,"concrete_test":"In the Sec. 4.1 problem, instrument Eq. (21) and record theta at the cell containing a smooth extremum (and at a generic cell) for N = 40, 80, 160, 320, 640. Fit log theta versus log h. If the slope is at least k, Theorem 1's hypothesis is met. If the slope is k-1, rerun the accuracy test with c_b in (21) increased by a factor of 20; a drop in observed order below k would confirm that the penalty is the cause and the formal-order claim requires qualification. This directly tests the paper's 'even near smooth extrema' claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing point is Section 2, Eq. (21) together with Theorem 1. The proof requires theta_n = O(Delta x^k) at every smooth interface, and the text justifies this by writing |p_i - p_n| = O(Delta x^{k+1}) and u_i - u_m = O(Delta x). Near a smooth extremum the second estimate fails: if the extremum lies in cell i, cell averages on the two sides differ by O(Delta x^2), not O(Delta x). At the same time, the smoothness indicator (12) for the extremal cell can be O(Delta x^7) while a neighbour is O(Delta x^5), so the weight beta_n in (14) deviates from 1/2 by O(1), making |1/2 - beta_n| = O(1). Then (21) gives theta_n = O(Delta x^{k-1}), violating the hypothesis of Theorem 1. The paper's assertion that the BVD penalty 'goes to zero ... even near smooth extrema' is therefore not established by the given argument. The accuracy tests in Sec. 4.1 do show designed order for sin, sin^2 and sin^3, so the concern is a gap in the proof rather than a demonstrated inconsistency; but the central claim of simultaneous high order and shock capture relies on this scaling, and the numerical tests are the only evidence at the extremum.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a family of third- to fifth-order weighted compact least-squares (WCLS) finite-volume schemes for compressible Navier-Stokes equations on curvilinear grids. The reconstruction weights the accuracy terms so that the compact reconstruction is performed only along smooth lines, adds a boundary-value-difference (BVD) penalty that is active near discontinuities, and embeds the nonlinear reconstruction into an implicit time-marching framework using GMRES. The accuracy theory is stated in Theorem 1, which requires the penalty coefficient theta_n to be O(Delta x^k) in smooth regions. The paper reports designed-order accuracy on smooth problems and improved resolution and lower dissipation than WENO3-JS, WENO5-Z, and WBAP-CLS schemes on a wide range of inviscid and viscous benchmark problems in one, two, and three dimensions.","tokens_in":32809,"tokens_out":7965,"duration_ms":76523,"significance":"If the accuracy claim holds, the WCLS framework offers a practical compromise between compact high-order resolution and robust shock capturing, extending the earlier third-order WCLS work to higher orders with an efficient implicit implementation. The method is novel in its combination of accuracy-preserving weighting, BVD-based penalties, and coupling of reconstruction with implicit iterations. The numerical validation is unusually broad, covering non-uniform curvilinear grids, viscous flows, and 3D turbulent benchmarks, and the reported efficiency comparisons are useful. The principal limitation is that the formal proof of the order-preservation property rests on a scaling assumption that is not rigorously established.","major_comments":[{"comment":"The theorem's hypothesis that theta_n = O(Delta x^k) is not established at smooth extrema. The argument following Eq. (21) balances the boundary-value difference |p_i(x_alpha)-p_n(x_alpha)| = O(Delta x^{k+1}) against the cell-average difference |u_i-u_m| = O(Delta x), but at a smooth extremum the cell-average difference can be O(Delta x^2) or smaller, and the smoothness indicators of Eq. (12) in the extremal cell and its neighbor can differ by O(1) amounts, making |1/2 - beta_n| = O(1). Under those conditions Eq. (21) gives theta_n = O(Delta x^{k-1}), which is outside the theorem's hypothesis. The accuracy tests in Sec. 4.1 bypass the shock detector and therefore do not close the gap; the central claim that the BVD penalty 'goes to zero ... even near smooth extrema' needs either a rigorous proof or a modification of Eq. (21) that guarantees the required scaling in the active-weighting regime.","section":"Section 2, Eq. (21) and Theorem 1"},{"comment":"The reconstruction system (11) is nonlinear because beta and theta in Eqs. (14) and (21) depend on the unknown reconstruction coefficients. The manuscript freezes these quantities at the beginning of each GMRES iteration (Algorithm I, lines 6-18) and updates them in subsequent inner iterations, but it provides no convergence analysis for this fixed-point iteration: no contractivity estimate, no proof that the inner iteration reaches a solution of the nonlinear system, and no examination of the dependence of the final solution on the iteration path. Since the numerical results in Section 4 rely on the convergence of this coupled process, the paper should provide either a convergence proof or a systematic residual-convergence study to rule out iteration-path-dependent accuracy.","section":"Section 3, Algorithm I and Eq. (11)"}],"minor_comments":[{"comment":"The subfigure labels in the Figure 5 caption are inconsistent: the caption lists (a), (b), (c) and then (c), (d), (f) for the second row; please renumber the panels.","section":"Figure 5 caption"},{"comment":"The sentence 'The reference result is taken from Li [43]' cites reference [43], which is Zhang and Shu; the citation appears to be misattributed.","section":"Section 4.13"},{"comment":"The piecewise display of Eq. (21) is formatted as a three-branch case but contains only the formula for theta_n and the two sensor definitions S_{i-1} and S_{i+1}; the typesetting should be clarified so that the reader can identify which branches correspond to which variables.","section":"Section 2, Eq. (21)"},{"comment":"The accuracy results are presented as error plots in Fig. 5; reporting the observed convergence rates in a table would make the claimed third-, fourth-, and fifth-order rates easier to verify.","section":"Section 4.1"},{"comment":"The genetic algorithm used for the spectral optimization is not referenced; please cite the implementation or a standard reference.","section":"Section 3"},{"comment":"The grid disturbance type 'ALT-RAND' from Ref. [35] is not defined; a one-sentence definition would help readability.","section":"Section 4.7"}],"recommendation":"major_revision","confidential_remarks":"The manuscript fits the journal's scope and the numerical study is extensive. The main risk is the unproved theta_n scaling in Section 2; the accuracy tests in Fig. 5 support the conclusion empirically but do not replace the missing bound. The revision should also address the convergence of the fixed-point reconstruction iteration. The duplicate label in Fig. 5 and the misattributed reference [43] should be corrected."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Zhi, quick take on 2608.08117. The paper does real work: it takes the third-order WCLS scheme from [16], extends it to third, fourth, and fifth order for the compressible Navier-Stokes equations on curvilinear grids, replaces the non-compact third-party smoothness indicator with a von-Neumann-neighbor-based one, adds BVD-based penalty matrices, and embeds the whole reconstruction in implicit time integration with GMRES. That last piece is genuinely useful: it avoids block-tridiagonal solves and makes the scheme competitive with WENO on cost. The numerical campaign is broad and mostly convincing: accuracy tests with sin^m for m=1,2,3 show designed order, shock/vortex/airfoil/TGV cases consistently show lower dissipation than WENO3-JS, WENO5-Z, and WBAP-CLS, and the Mach 2000 jet case shows robustness that the WBAP-limited CLS scheme lacks.\n\nThe soft spot is the one the stress test flags. Theorem 1 requires theta_n = O(h^k) to preserve k-th order. The paper argues this from the BVD term: |p_i-p_n|=O(h^{k+1}) over |u_i-u_m|=O(h). Near a smooth extremum the denominator can be O(h^2), and the smoothness indicator can push beta away from 1/2, so the scaling can degrade to O(h^{k-1}). The authors claim the penalty 'goes to zero even near smooth extrema,' but they don't prove it; the sin^2 and sin^3 accuracy tests are evidence but not a proof. That is a real gap, but it is not a demonstrated failure. The accuracy tests in Sec. 4.1 show the scheme does hit designed order on problems with extrema, so the practical claim likely holds. The paper would be stronger if it either proved the extremum case or explicitly stated that the formal guarantee is conditional on that scaling.\n\nMinor issues: no code or data deposited, only 'available upon request,' which limits reproducibility; several constants are tuned (c_h, c_s, sigma_0, spectral weights), which is normal for this class of scheme but worth stating more plainly.\n\nBottom line: this is a careful, well-tested extension of a promising framework. The central claim—simultaneous high order and sharp shock capture—is supported by the numerics, but the proof has a genuine hole at extrema. It deserves a serious referee; I'd send it to review and ask the authors to either close the gap or soften Theorem 1.","headline":"Solid incremental extension of the WCLS framework with broad numerical validation, but the formal order guarantee at smooth extrema rests on an unproved scaling assumption.","tokens_in":33257,"tokens_out":2224,"would_cite":true,"duration_ms":21358,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65M08","65M12","76N15"],"pacs":[],"model":"deepseek-v4-flash","headline":"A weighted compact least-squares reconstruction can preserve designed high-order accuracy in smooth regions and capture shocks and contacts sharply on curvilinear grids.","keywords":["weighted compact least-squares","finite volume method","compressible Navier-Stokes equations","shock capturing","high-order accuracy","curvilinear grids","implicit time integration","boundary value difference"],"falsifier":"Run the WCLS5 scheme on smooth advection with $\\sin^3(\\pi x)$ initial data on a sequence of refined grids, record the penalty coefficient in cells adjacent to the extrema, and check both whether it decays like $\\Delta x^5$ and whether the $L^\\infty$ error slope reaches 5; either failure would contradict the order-preservation claim.","tokens_in":32197,"feed_emoji":"🌪","tokens_out":16306,"duration_ms":147455,"temperature":0.7,"pith_summary":"This paper develops a family of third- to fifth-order finite-volume schemes for the compressible Navier-Stokes equations and claims that they resolve a long-standing trade-off: high-order accuracy in smooth regions no longer has to be sacrificed for non-oscillatory shock capture. The key move is to enforce the compact least-squares accuracy conditions only along reconstruction lines that are locally smooth, using face weights $\\beta_n$ that fade to zero across discontinuities, and to activate a boundary-value-difference penalty $\\theta_n$ only at the ends of those smooth lines. In smooth regions $\\theta_n$ is $O(\\Delta x^k)$, so the designed order is retained; near discontinuities it becomes $O(1)$ and damps Gibbs-type oscillations. Numerical experiments on one-, two-, and three-dimensional inviscid and viscous flows show sharper contacts, lower dissipation, and better small-scale structure preservation than the WENO3-JS, WENO5-Z, and WBAP-CLS reference schemes. If the claims hold, the method offers a practical route to high-resolution compressible turbulence simulations with strong shocks on curvilinear grids.","feed_headline":"3rd-to-5th-order solver keeps shocks sharp and smooth flow clean","feed_subtitle":"WCLS schemes preserve formal accuracy while out-resolving WENO references in shock and turbulence tests.","key_machinery":"The machinery is the weighted compact least-squares reconstruction: a per-control-volume polynomial whose coefficients are determined by minimizing a loss that couples derivatives across neighboring faces. Three ingredients carry the argument: the face weight $\\beta_n$, formed from a modified smoothness indicator that uses only von Neumann neighbors, which separates reconstruction lines into smooth segments; the penalty coefficient $\\theta_n$, whose boundary-value-difference scaling makes the penalty $O(\\Delta x^k)$ in smooth regions and $O(1)$ at discontinuities; and the coupling of the resulting nonlinear reconstruction with implicit time integration through GMRES, so that $\\beta$ and $\\theta$ are frozen during each Krylov iteration and updated between iterations. The linear weights $W_j$ are tuned by a genetic algorithm to maximize the resolved-wavenumber region, and the penalty parameters are calibrated separately for the third-order and the higher-order versions.","core_discovery":"The central claim is that the Gibbs phenomenon observed in compact least-squares schemes is not an unavoidable property of compact reconstruction but a consequence of enforcing accuracy constraints across discontinuities. The paper replaces the single global reconstruction line with smooth segments: each interface contribution to the loss is multiplied by $\\beta_n \\in [0,1]$, computed from a compact smoothness indicator that uses only von Neumann neighbors plus an inexpensive linear reconstruction. When $\\beta_n$ drops, the segment endpoint receives a penalty term whose coefficient $\\theta_n$ is proportional to the boundary-value difference $|p_i(x_\\alpha)-p_n(x_\\alpha)|$ divided by the jump in cell averages. Because this quantity is $O(\\Delta x^{k+1})/O(\\Delta x)=O(\\Delta x^k)$ in smooth regions and $O(1)$ near discontinuities, the penalty vanishes asymptotically where the solution is resolved, preserving formal order, and turns on where the solution is under-resolved, suppressing oscillations. The reconstruction system, now nonlinear through $\\beta_n$ and $\\theta_n$, is solved inside each implicit time step by one GMRES iteration, avoiding the block-tridiagonal solves and the separate limited-polynomial set of earlier compact least-squares schemes.","pith_inferences":["Beyond the paper, the compact smoothness indicator built only from von Neumann neighbors suggests that the same weighting could transfer to unstructured grids, but the paper does not demonstrate that extension.","Beyond the paper, the spectral-optimization procedure suggests that the free linear weights could be tuned for a given grid or flow to shift the resolved-wavenumber range, a use the paper does not explore.","Beyond the paper, the one-GMRES-step-per-iteration coupling suggests a natural synchronization strategy for parallel compact schemes: exchange reconstructed derivatives at halo cells during the Krylov iterations instead of solving a global block-tridiagonal system; the paper states this benefit but does not benchmark it in parallel.","Beyond the paper, the BVD-based penalty mechanism could be combined with a positivity-preserving or realizability fix to extend the method to multiphase or very extreme compressible flows."],"forward_implications":["If the central claim is right, the third-order WCLS scheme resolves contacts and small-scale vortices at a resolution comparable to the fifth-order WENO-Z scheme, so lower-order reconstructions become usable where fifth-order WENO was previously needed.","Formal third-, fourth-, and fifth-order convergence is maintained even when the weighting is active everywhere, including flows with first- and second-order smooth extrema, where WENO3-JS and WENO5-Z degrade.","The single-set-of-polynomials formulation removes the need for a separate limiter pass, and the GMRES coupling removes block-tridiagonal solves, so the per-iteration cost stays comparable to WENO inside an implicit solver.","The method retains low numerical dissipation in long-time compressible vortex simulations, preserving kinetic energy and dissipation peaks better than the WENO and WBAP-CLS references.","The penalty formulation remains stable in extreme regimes, including stationary shocks up to Mach 1000 and high-Mach jet flows, where the WBAP-CLS5 reference fails."],"supporting_citations":[{"why":"Provides the prior third-order WCLS formulation and the smooth-line reconstruction idea that the present schemes extend.","marker":"[16]"},{"why":"Supplies the compact least-squares variational reconstruction and the k-exact property behind Theorem 1.","marker":"[13]"},{"why":"Contributes the standard WENO smoothness indicator that Eq. (12) modifies to make the iteration converge near discontinuities.","marker":"[20]"},{"why":"Serves as the WENO-Z reference baseline for contact resolution, spectral behavior, and accuracy comparisons.","marker":"[17]"},{"why":"Establishes the compact high-order finite-volume formulation and the spectral analysis used for the WCLS reconstruction.","marker":"[11]"},{"why":"Introduces the boundary-value-difference concept that motivates the scaling of the penalty coefficient in Eq. (21).","marker":"[29]"},{"why":"Supplies the ESDIRK/SDIRK implicit Runge-Kutta time integrators used in the numerical experiments.","marker":"[24]"}],"fun_headline_variants":["Accuracy-preserving weights eliminate Gibbs in compact WENO","Single polynomial set boosts compact shock-capturing efficiency","Implicit WCLS damps high-freq noise, preserves formal order","Curvilinear WENO compact scheme tames Gibbs with beta weights","Sharp shocks, smooth fields: new implicit compact WCLS"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The accuracy guarantee rests on the assumption that the stabilizing penalty stays at most of order grid-spacing-to-the-k near smooth extrema, where the cell-averaged denominator used to define it can shrink to nearly zero; the paper provides numerical evidence but not a proof for that case.","fun_headline_variants_meta":{"raw":{"variants":["Accuracy-preserving weights eliminate Gibbs in compact WENO","Single polynomial set boosts compact shock-capturing efficiency","Implicit WCLS damps high-freq noise, preserves formal order","Curvilinear WENO compact scheme tames Gibbs with beta weights","Sharp shocks, smooth fields: new implicit compact WCLS"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000604,"raw_usage":{"total_tokens":2862,"prompt_tokens":1031,"completion_tokens":1831,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":647,"completion_tokens_details":{"reasoning_tokens":1743}},"tokens_in":647,"tokens_out":1831,"duration_ms":13139,"temperature":1.0,"reasoning_tokens":1743,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T00:23:47.964256+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the WCLS5 scheme on smooth advection with $\\sin^3(\\pi x)$ initial data on a sequence of refined grids, record the penalty coefficient in cells adjacent to the extrema, and check both whether it decays like $\\Delta x^5$ and whether the $L^\\infty$ error slope reaches 5; either failure would contradict the order-preservation claim.","supporting_citations":[{"cited_title":"A Third-Order Weighted Essentially Non-Oscillatory Compact Least-Squares Scheme for Hyperbolic Conservation Laws on Non-Uniform Grids","cited_arxiv_id":"2508.02033","evidence_quote":"Provides the prior third-order WCLS formulation and the smooth-line reconstruction idea that the present schemes extend."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the compact least-squares variational reconstruction and the k-exact property behind Theorem 1."},{"cited_title":"Jiang, C.-W","cited_arxiv_id":null,"evidence_quote":"Contributes the standard WENO smoothness indicator that Eq. (12) modifies to make the iteration converge near discontinuities."},{"cited_title":"Castro, B","cited_arxiv_id":null,"evidence_quote":"Serves as the WENO-Z reference baseline for contact resolution, spectral behavior, and accuracy comparisons."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the compact high-order finite-volume formulation and the spectral analysis used for the WCLS reconstruction."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the boundary-value-difference concept that motivates the scaling of the penalty coefficient in Eq. (21)."},{"cited_title":"Kennedy, M.H","cited_arxiv_id":null,"evidence_quote":"Supplies the ESDIRK/SDIRK implicit Runge-Kutta time integrators used in the numerical experiments."}],"review_version":1}