{"id":"752d8e69-86ab-4035-b045-70c27c41aeb2","arxiv_id":"2608.09615","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A compact, parameter-free centered viscous flux for the Spectral Difference method reduces the diffusive stencil from five to three elements and restores optimal convergence at even polynomial orders.","lead":"This paper presents a compact way to compute viscous, diffusive fluxes inside the Spectral Difference method, a high-order numerical technique used for fluid simulation. The new scheme needs fewer neighboring cells, fixes a known accuracy loss at even polynomial orders, and stays stable in turbulent test cases where the standard scheme crashes.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Multidimensional extension in Section 2.3 is the load-bearing gap: direction-by-direction one-sided corrections are untested for formal accuracy, and no 2D/3D convergence study supports the claim of restored order for all polynomial degrees.","rationale":"Read in good faith: the paper's core 1D derivation is algebraically plausible, the stencil reduction from five to three elements is explicitly exhibited, and the numerical evidence in 1D (Tables 1-2, Figures 4-8) is internally consistent with the claimed improvement at even solution-point counts. The eigenanalysis and TGV stability results provide independent support for the dissipation mechanism. I did not find a circular argument, parameter fitting, or an internal inconsistency in the 1D algebra. The weakest point is precisely the multidimensional extension: it is the only place where the abstract's 'all polynomial degrees' and 'attractive alternative' claims are extrapolated beyond the tested regime. The reader's weakest_assumption identifies the same gap. A single 2D/3D manufactured-solution convergence test would settle whether the direction-by-direction implementation preserves order. If it passes, the CONDITIONAL verdict should become ACCEPT; if it fails, the paper should be CONDITIONAL or REJECT depending on how central the multidimensional accuracy claim is. Since the current evidence is insufficient either way, the reader's CONDITIONAL verdict is the right recommendation.","tokens_in":49391,"tokens_out":10670,"duration_ms":104563,"concrete_test":"Implement the Section 2.3 compact scheme for the 2D diffusion equation on tensor-product quadrilateral or hexahedral meshes, including at least one non-affine, curved/deformed mesh, with a manufactured solution exercising non-separable dependence in x and y (e.g., u = sin(pi*x)*sin(pi*y) or a polynomial containing an xy term). Run p = 1 through 5 on four nested refinements (e.g., 16^2, 32^2, 64^2, 128^2 elements) and compute L2 convergence rates for the compact scheme against standard BR1. If the compact slope is one order below nominal for any polynomial degree, or if the rate degrades on the deformed mesh, the direction-by-direction extension fails to preserve the 1D accuracy claim. Passing this test would remove the main obstacle to ACCEPT.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim that the compact scheme restores the expected convergence order for all polynomial degrees and is an attractive multidimensional alternative depends on Section 2.3. The 1D construction (Section 2.2, Eqs. 21-26) is internally coherent and is supported by the 1D convergence study, the Dirac-delta test, and the PME tests. However, the multidimensional extension is asserted by analogy: in Eqs. (41)-(44), each component of the auxiliary gradient is computed by differentiating a one-sided corrected flux in one direction and the standard continuous flux in the transverse direction, then interpolating only at the relevant interface. No proof is given that this directional splitting preserves the 1D accuracy on tensor-product hexahedra, especially for cross-derivative terms and non-affine metric terms. No 2D or 3D convergence test appears anywhere in the paper. The 3D TGV runs demonstrate stability, not formal accuracy, so they cannot validate the accuracy claim. If the extension loses one order of accuracy, the abstract's 'all polynomial degrees' statement is only established in 1D, and the practical LES motivation weakens. This is the load-bearing gap: the headline accuracy property is claimed for multiple dimensions but is only evidenced in one dimension.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a compact, fully-centered viscous-flux formulation for the Spectral Difference (SD) method, inspired by Huynh's Flux Reconstruction approach. In one dimension it modifies the reconstruction of the auxiliary gradient so that the stencil shrinks from five to three elements while preserving the parameter-free, centered character of BR1. The authors give the 1D construction, a direction-by-direction extension to tensor-product hexahedral elements, a temporal eigenanalysis with and without interior penalty terms, and a series of tests: 1D Gaussian diffusion convergence, under-resolved Dirac-delta diffusion, the porous medium equation, and three-dimensional Taylor-Green vortex ILES. The 1D derivation and eigenanalysis are algebraically coherent, and the compact scheme's 1D convergence behavior is demonstrated for p=1 through p=5. The multidimensional accuracy claim, however, is asserted by analogy and is not supported by any 2D or 3D convergence test.","tokens_in":49575,"tokens_out":6010,"duration_ms":53745,"significance":"If the claims hold, this is a practically useful, parameter-free alternative to BR1 for the SD discretization of second-order fluxes, with particular relevance to implicit LES on hexahedral meshes. The manuscript has clear strengths: the compact scheme is parameter-free, the eigenanalysis is derived from the scheme's own operators, the 1D convergence rates are measured against exact solutions, and the TGV stability tables provide falsifiable comparisons. The current significance is limited by an internal contradiction in the parity of the suboptimal BR1 orders (the paper's own data show odd p, not even p, are suboptimal) and by the absence of any multidimensional accuracy validation, which leaves the headline 'all polynomial degrees' claim supported only in one dimension.","major_comments":[{"comment":"The text in §4.1 states that the standard scheme 'provides a much more degraded convergence rate' at even orders, listing p=2 and p=4 as even; however, Table 2 shows the opposite: the standard scheme's convergence rates are 0.998, 2.996, and 4.990 for p=1, p=3, and p=5 (all odd), while for p=2 and p=4 the rates are 3.001 and 5.050, which match the nominal orders. The same misstatement appears in the Abstract ('including even orders for which the standard BR1 scheme underperforms'). This is a load-bearing error in the central accuracy claim and must be corrected, and the abstract's 'even orders' must be revised to reflect the paper's own data.","section":"§4.1, Table 2 and surrounding text"},{"comment":"The multidimensional extension is presented by applying the one-sided compact corrections direction by direction (Eqs. (41)-(44)), but no 2D or 3D convergence study is reported anywhere in the manuscript. The Taylor-Green vortex tests demonstrate stability, not formal accuracy, and the text itself notes that in multiple dimensions the global stencil spans five elements. Consequently, the abstract's statement that the compact scheme 'restores the expected convergence order for all polynomial degrees' is only verified in one dimension. The authors should either supply multidimensional convergence evidence (e.g., a manufactured 2D or 3D diffusion problem with cross-derivative terms and non-affine metrics) or explicitly restrict the accuracy claim to the 1D case.","section":"§2.3 and Abstract"},{"comment":"The convergence-rate claims are based on a single smooth test problem (well-resolved Gaussian diffusion) with no error estimates or multiple test cases. While this is not by itself a fatal flaw, it is the only evidence for the central accuracy claim; adding a second 1D problem with a known exact solution, or fitting the rates with linear-regression standard errors, would strengthen the conclusion that the compact scheme 'consistently achieves the expected order of accuracy for all polynomial degrees.'","section":"§4.1, Table 2"}],"minor_comments":[{"comment":"The sentence beginning 'For even orders of approximation, p=1 p=2 p=3 p=4 p=5 compact 2.599...' is a fragment and mislabels the parity of the tabulated polynomial degrees; p=1,3,5 are odd and p=2,4 are even.","section":"§4.1"},{"comment":"The abstract's 'reducing the stencil from five to three elements' is a 1D statement; in §2.3, step 3, the text correctly notes that the global multidimensional stencil spans five elements. The abstract should qualify the stencil reduction as applying to the one-dimensional case to avoid overstatement.","section":"Abstract and §2.3"},{"comment":"The eigenanalysis section states that both schemes satisfy Re(e\\omega)=0 and then presents only dissipation curves; it would be clearer to state explicitly that the dispersion is identically zero for these centered schemes, so the imaginary part of the numerical frequency fully characterizes the scheme.","section":"§3"},{"comment":"The Fourier coefficient is computed using only the cosine transform; this is justified for the symmetric initial data used here, but the text should mention this restriction rather than presenting it as a general Fourier-space analysis.","section":"§4.2, Eq. (73)"},{"comment":"Reference [7] has a formatting inconsistency: the title and subtitle are run together, and the doi is duplicated; this should be normalized.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The parity error in §4.1 is a simple but consequential mistake: the paper's own table contradicts the abstract's 'even orders' claim. The more serious issue is the missing multidimensional accuracy validation, which the authors should address either by adding 2D/3D convergence tests or by revising the abstract's scope. The 1D contribution appears sound and worth publishing after revision; the manuscript is within the journal's scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a look. The paper does something real: it builds a compact, parameter-free, fully centered viscous flux for the Spectral Difference method, modeled transparently on Huynh's one-sided continuous-flux idea, and shows in 1D that it restores nominal convergence for all tested polynomial degrees where standard BR1 falls short by about one order at even approximation orders. The stencil drops from five elements to three. The derivation is clean and the eigenanalysis is internally coherent; the match between predicted dissipation peaks and the Fourier behavior in the Dirac-delta test is a nice piece of corroborating evidence. No parameters are fitted to make the results work. The PME and TGV tests are honest and the stability differences are credible.\n\nThe soft spot is exactly the one the stress-test flags: Section 2.3 extends the scheme to multiple dimensions direction by direction on tensor-product hexahedra, but there is no proof and no 2D or 3D convergence study that the split one-sided corrections preserve formal accuracy. The TGV runs show stability, not accuracy. As written, the abstract's 'all polynomial degrees' claim is only actually established in 1D. That is a real gap but a repairable one, so my verdict is conditional rather than dismissive. A couple of smaller things: the convergence table has no error estimates per run, and no code or data artifacts are provided, which makes independent verification slower. Neither issue undercuts the 1D result.\n\nWho gets value: anyone working on SD or FR viscous discretizations, especially people interested in ILES robustness and compact stencils for implicit solvers. It deserves serious refereeing. I would send it out and ask the authors to either prove the multidimensional accuracy bound or add a 2D/3D manufactured-solution convergence study before publication. If the multidimensional part holds up, this becomes a genuinely useful alternative to BR1 in the SD community.","headline":"A solid 1D compact viscous-flux construction for SD that fixes BR1's even-order convergence and adds high-wavenumber damping; the multidimensional extension is asserted, not demonstrated.","tokens_in":50120,"tokens_out":904,"would_cite":true,"duration_ms":11543,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65M70","65M60","76M22"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proposes a compact, fully centered reformulation of second-order fluxes for the Spectral Difference method that shrinks the viscous stencil from five to three elements and restores nominal convergence order at even approximation…","keywords":["Spectral Difference method","second-order fluxes","Bassi-Rebay flux","compact stencil","discontinuous spectral element methods","diffusion equation","implicit large-eddy simulation","Taylor-Green vortex"],"falsifier":"Run a two-dimensional manufactured-solution convergence study for the compact scheme on hexahedral elements, measuring L2 error for polynomial degrees p=1, 2, 3, and 4; the central claim would be refuted if any even approximation order (p+1 even, i.e., p odd) shows a convergence rate one order below nominal.","tokens_in":49154,"feed_emoji":"","tokens_out":8872,"duration_ms":72764,"temperature":0.7,"pith_summary":"The paper proposes a modification of how the Spectral Difference method handles second-order (viscous) fluxes, replacing the standard BR1 auxiliary-gradient construction with a compact, fully centered one. The goal is to keep BR1's simplicity and parameter-free character while curing its two known defects: sub-optimal convergence at even approximation orders and a wide five-element stencil. The new formulation, inspired by Huynh's Flux Reconstruction idea, builds one-sided continuous fluxes at each interface and averages them only there, which shrinks the viscous stencil to three elements in one dimension. Numerical tests on linear diffusion, an under-resolved Dirac delta, the porous medium equation, and the Taylor-Green vortex show the compact scheme recovering the expected order for all tested polynomial degrees, dampening spurious oscillations, and staying stable in turbulent cases where standard BR1 fails.","feed_headline":"Compact viscous flux restores even-order accuracy in SD methods","feed_subtitle":"A three-element stencil restores even-order accuracy and stabilizes under-resolved turbulent flow simulations.","key_machinery":"The load-bearing object is the pair of one-sided continuous flux reconstructions $H^{C,L}_e$ and $H^{C,R}_e$ (equivalently, the corrected gradient states $v^{S,L}_e$ and $v^{S,R}_e$) built from the correction vectors $\\hat{c}_L$ and $\\hat{c}_R$ at each element's interfaces. Unlike the standard BR1 construction, which first forms a globally continuous flux and then differentiates it, the compact construction applies each interface correction only on the side that shares that interface, so the auxiliary gradient at an interface depends only on the two adjacent elements. The interface gradient is then the average of these one-sided contributions. This is what cuts the viscous stencil from five to three elements and what the eigenanalysis shows to add dissipation in the medium- to high-wavenumber range.","core_discovery":"The central claim is that the compact, fully-centered formulation is an effective drop-in alternative to BR1 within the Spectral Difference method, achieving nominal convergence order for every polynomial degree tested and proving more robust in under-resolved and nonlinear regimes. The mechanism is the reconstruction of the auxiliary gradient from element-local one-sided continuous fluxes, with interface values obtained by averaging contributions from the two neighbors only. This preserves the centered, parameter-free character of BR1 while formally reducing the one-dimensional stencil from five to three elements and improving damping of high-wavenumber numerical modes. The paper supports this claim with a temporal eigenanalysis of the dissipation curves and with convergence studies, under-resolved diffusion tests, and implicit large-eddy simulations of the Taylor-Green vortex at Reynolds numbers 1600 and 5000.","pith_inferences":["The paper leaves implicit that the multidimensional accuracy gain is the main untested part of the scheme; a 2D or 3D manufactured-solution convergence study on hexahedral meshes would directly test the tensor-product extension, and none is reported.","The eigenanalysis shows that the compact flux adds dissipation mainly at medium and high wavenumbers, so one could quantify this as an effective implicit subgrid-scale viscosity by measuring spectral energy transfer in the Taylor-Green vortex, a diagnostic the paper does not compute.","The same one-sided, interface-averaged reconstruction idea could be transferred to other discontinuous spectral element formulations that inherit BR1's wide stencil, although the paper demonstrates it only within the Spectral Difference framework.","At higher resolution, the interior penalty term can degrade the compact scheme (as observed for the porous medium equation on 200 elements), suggesting that the compact flux may already supply enough dissipation and that penalty should be tuned lower when the two are combined."],"forward_implications":["For the linear diffusion equation, the compact scheme reaches convergence rates of 2.60, 3.01, 4.19, 5.04, and 6.37 for polynomial degrees p=1 through p=5, whereas standard BR1 drops to about 1.00, 3.00, 3.00, 5.05, and 4.99, so the even-order deficit at p=1, 3, and 5 is removed.","In under-resolved tests (Dirac delta diffusion and the porous medium equation), the compact scheme produces smaller L2 errors and markedly fewer spurious high-wavenumber oscillations than the standard flux, with the Fourier-space oscillation peaks matching the eigenanalysis predictions.","In implicit large-eddy simulations of the Taylor-Green vortex at Re=1600 and Re=5000, the compact scheme stays stable in configurations where the standard flux diverges, such as the 16^3 p7 case at Re=1600 and the 12^3 p6 case at Re=5000, and in stable runs it yields lower spurious enstrophy.","The reduced five-to-three-element stencil lowers the memory and data-exchange footprint of the viscous operator, which is a direct benefit for implicit time integration, adjoint-based optimization, and large-scale parallel implementations."],"supporting_citations":[{"why":"Supplies the one-sided continuous-flux idea from Flux Reconstruction that the compact SD formulation adapts to the auxiliary gradient.","marker":"[43]"},{"why":"Defines the parameter-free centered BR1 flux whose five-element stencil and even-order accuracy deficit the paper targets.","marker":"[9]"},{"why":"Introduces the staggered-grid Spectral Difference discretization into which the new compact flux is inserted.","marker":"[47]"},{"why":"Extends the SD scheme to multidomain compressible flows, establishing the flux-point/solution-point framework used throughout the paper.","marker":"[48]"},{"why":"Provides the standard SD treatment of interface fluxes for the auxiliary gradient that the compact scheme modifies.","marker":"[83]"},{"why":"Documents the sub-optimal even-order convergence of BR1-type diffusion discretizations in the DG literature.","marker":"[11]"},{"why":"Cockburn and Shu's LDG analysis is cited as further evidence of the even-order sub-optimal behavior that the compact scheme corrects.","marker":"[25]"},{"why":"Gives the interior penalty coefficient and the BR1-as-SIP relation used in the eigenanalysis and penalty-augmented tests.","marker":"[63]"},{"why":"Supplies the Taylor-Green vortex setup and flow diagnostics used in the implicit LES stability comparisons.","marker":"[14]"},{"why":"Provides a same-code Taylor-Green vortex simulation that was unstable without regularization, serving as the baseline for comparing the compact and standard fluxes.","marker":"[18]"}],"fun_headline_variants":["Three-element stencil fixes even-order SD accuracy","Compact SD viscous flux restores convergence order","Spectral Difference gets compact stability fix","Turbulent flows tamed by compact SD diffusion","Even-order accuracy restored with compact flux"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The multidimensional accuracy claim rests on assuming that applying the compact corrections direction by direction on tensor-product hexahedral elements preserves enough accuracy, and the paper does not include a 2D or 3D convergence test to verify this.","fun_headline_variants_meta":{"raw":{"variants":["Three-element stencil fixes even-order SD accuracy","Compact SD viscous flux restores convergence order","Spectral Difference gets compact stability fix","Turbulent flows tamed by compact SD diffusion","Even-order accuracy restored with compact flux"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000444,"raw_usage":{"total_tokens":2269,"prompt_tokens":987,"completion_tokens":1282,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":603,"completion_tokens_details":{"reasoning_tokens":1215}},"tokens_in":603,"tokens_out":1282,"duration_ms":9387,"temperature":1.0,"reasoning_tokens":1215,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T14:03:28.142026+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a two-dimensional manufactured-solution convergence study for the compact scheme on hexahedral elements, measuring L2 error for polynomial degrees p=1, 2, 3, and 4; the central claim would be refuted if any even approximation order (p+1 even, i.e., p odd) shows a convergence rate one order below nominal.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the one-sided continuous-flux idea from Flux Reconstruction that the compact SD formulation adapts to the auxiliary gradient."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the staggered-grid Spectral Difference discretization into which the new compact flux is inserted."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Extends the SD scheme to multidomain compressible flows, establishing the flux-point/solution-point framework used throughout the paper."},{"cited_title":"Wang, and Y","cited_arxiv_id":null,"evidence_quote":"Provides the standard SD treatment of interface fluxes for the auxiliary gradient that the compact scheme modifies."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the interior penalty coefficient and the BR1-as-SIP relation used in the eigenanalysis and penalty-augmented tests."}],"review_version":1}