{"id":"8baf109f-6791-4964-ba6f-bbab6b6c5f44","arxiv_id":"1908.06604","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":3.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"An overview of the authors' own previously published HDG-NEFEM, HDG-Voigt and FCFV methods, illustrated with new tests in electrostatics, elasticity and Stokes flow.","lead":"This paper reviews three recent numerical methods that make discontinuous Galerkin simulations more practical: exact-geometry degree adaptivity, a symmetric-stress formulation for elasticity, and a lowest-order finite volume scheme for large flow problems. A generalist reader interested in how high-order finite element ideas are being made cheaper and more robust will find a concise map of this active area.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The NEFEM monotonic-decrease claim rests on exact-error curves whose reference solution is unspecified and on an error indicator that is never independently checked.","rationale":"The reader's CONDITIONAL verdict is appropriate. The strongest claim is empirical rather than theorem-based, so its support is the numerical experiment in Section 3. I looked for a mathematical flaw in Eq. (9) and did not find one: the ratio epsilon/E cancels the unknown constant C and the current degree, so the formula is internally consistent with Eq. (8). The real vulnerability is that the exact-error curves in Figure 3 are not defined and the error indicator is not separately validated. This is not an ad hominem concern; it is a standard reproducibility burden for a paper whose main new result is a numerical demonstration. The proposed check, reproducing the experiment with specified parameters and an independent reference, would settle whether the monotonic-decrease claim lands. Since the reader already assigned CONDITIONAL and my concern reinforces rather than changes that assessment, the verdict remains unchanged.","tokens_in":18877,"tokens_out":10886,"duration_ms":121850,"concrete_test":"Reproduce the Section 3 experiment with a fully specified setup: mesh size, initial polynomial degree, stabilization parameter tau, and the definition of the exact-error reference (e.g., a uniform degree-10 NEFEM solution on a twice-refined mesh or a manufactured solution). Independently recompute the exact and estimated error curves in Figure 3 and report the effectivity index of E_u^e at each iteration. If the NEFEM exact error is monotonically below the 5e-4 tolerance while the fixed-cubic exact error stagnates above it, the central claim stands; otherwise it fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim of Section 3 is that with HDG-NEFEM the exact error decreases monotonically to the prescribed tolerance, whereas degree adaptivity on a fixed cubic geometry converges to the solution of a different problem (Figure 3). The load-bearing support for this claim is the pair of 'exact error' curves, but the manuscript never defines how these errors are computed: there is no analytic solution for the filleted-square electrostatic problem, and the text does not state whether the reference is an overkill HDG-NEFEM solution, a different discretization, or a manufactured solution. Without that definition, the claimed monotonic decrease and the contrast with the cubic-geometry case cannot be independently checked. The local error indicator (7) and the degree-update formula (9) inherit this problem: (9) is derived from the a priori model (8) and presumes the estimator E_u^e is a faithful proxy for the true local error, but the only evidence offered is Figure 3 itself, and the cited analyses [82,113] are by the same group and concern standard polynomial elements, not NEFEM curved-boundary elements. Figure 3 (right) demonstrates how badly this assumption can fail when the geometry is wrong; the paper does not show the analogous effectivity-index check for the NEFEM run. The conclusion may be correct, but as written it is a single unreproducible numerical observation rather than an established property.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper revisits hybridizable discontinuous Galerkin (HDG) methods combined with exact geometry representation via NURBS-enhanced finite elements (NEFEM), degree adaptivity, a Voigt-based symmetric-tensor HDG formulation for linear elasticity, and the lowest-order face-centered finite volume (FCFV) method for large-scale Stokes flows. The central numerical claim is in Section 3: on a filleted-square electrostatic problem, degree adaptivity with HDG-NEFEM reduces the exact error monotonically to a prescribed tolerance, whereas degree adaptivity on a fixed cubic polynomial geometry converges instead to the solution of a different problem. The other sections summarize the authors' prior HDG-Voigt and FCFV developments and show representative benchmark results, including Cook's membrane and a 39-particle red-blood-cell Stokes flow.","tokens_in":19145,"tokens_out":6458,"duration_ms":65619,"significance":"If the monotonic-decrease and reliability claims hold, the HDG-NEFEM degree-adaptive strategy would be a practically valuable alternative to isoparametric or fixed-polynomial-geometry adaptivity for problems requiring tolerance-driven p-refinement on curved domains. The paper is clearly written and the formulations are standard. However, the central new evidence is a single numerical example whose reference ('exact') error is never defined, and the local error indicator's reliability for NEFEM elements is not independently verified. The manuscript also does not ship code or machine-checked proofs, so the numerical claims cannot be reproduced from the text alone. The contribution is best assessed as a well-structured survey plus one promising but under-validated numerical demonstration.","major_comments":[{"comment":"The exact-error curves in Figure 3 are not defined. The filleted-square problem has no known analytic solution, and the text does not state whether the reference solution is an overkill HDG-NEFEM computation, a different discretization, or a manufactured solution. Without this definition, the central claim that the HDG-NEFEM exact error decreases monotonically to the tolerance, and the contrast with the fixed-cubic-geometry case, cannot be independently checked. Please specify the reference solution, the discretization parameters used, and how the exact error is computed element-wise.","section":"Section 3, Figure 3"},{"comment":"The degree-update formula (9) presumes that the local indicator E_u^e defined in Eq. (7) is a faithful proxy for the true local error and that the a priori estimate (8), with the constant C estimated by Richardson extrapolation from two polynomial degrees, correctly predicts the error reduction when the degree is increased. No effectivity indices or per-iteration estimated-versus-exact error tables are given for the NEFEM run. This matters because Figure 3 (right) demonstrates that the same assumption can fail badly when the geometry is represented by fixed cubic polynomials. Please provide effectivity indices for both the NEFEM and cubic-geometry runs, and justify the validity of the exponent k+1+nsd/2 and the Richardson extrapolation over the degree range actually used.","section":"Section 3, Eqs. (8)-(9) and Figure 3"},{"comment":"The quantitative support for the HDG-Voigt and FCFV contributions is thin: the optimal-convergence claim for HDG-Voigt is illustrated only by the value of the displacement at one point of Cook's membrane, and the FCFV claim of first-order accuracy is not accompanied by any convergence table on a problem with a known solution. If these sections are intended as original numerical contributions rather than as summaries of prior published work, please add computed-versus-exact error tables and measured convergence rates. At minimum, state clearly that these are review sections and provide pointers to the original papers containing the full convergence studies.","section":"Sections 4-5, Figures 4, 6-7"}],"minor_comments":[{"comment":"The incompressibility constraint is written as ∫_{∂Ω_e\\Γ_D} υ̂·n dΓ = -∫_{∂Ω_e∩Γ_D} u_D·n dΓ = 0, but the final “= 0” is not true in general because the Dirichlet data u_D need not vanish on Γ_D. The equality should be just ∫_{∂Ω_e\\Γ_D} υ̂·n dΓ = -∫_{∂Ω_e∩Γ_D} u_D·n dΓ.","section":"Section 5, Eq. (26)"},{"comment":"The text says “the evaluation of the element-by-element solution in all 61 millions elements,” but the mesh consists of about 8.97 million tetrahedral elements; 61.5 million is the number of unknowns in the global system. Please correct this wording.","section":"Section 5, large-scale simulation paragraph"},{"comment":"In the degree-update formula, log(h_e) is negative when the nondimensional element size h_e is less than 1, so the sign convention in the ceiling expression should be explained explicitly, together with the nondimensionalization used to define h_e.","section":"Section 3, Eq. (9)"},{"comment":"The figures (Figures 1-7) are referenced in the captions but the images are not present in the manuscript text provided; please ensure the final version contains all figures with clearly readable axis labels and legends.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is largely a well-written summary of the authors' own prior work, with the only substantially new numerical claim being the Section 3 HDG-NEFEM degree-adaptivity demonstration. The self-citation pattern (Refs. [82,83,106,107,118]) is heavy, and the journal should consider whether this review-style paper fits its scope and whether the new evidence meets its standards. The main technical gap is the undefined reference solution for the 'exact error' and the unvalidated error indicator; these are fixable and should be required before acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this as a consolidated review of the authors' own HDG variants: HDG-NEFEM for exact geometry, HDG-Voigt for linear elasticity, and FCFV for large-scale Stokes flow. The genuinely new material is the degree-adaptivity comparison in Section 3 and the 39-RBC FCFV demonstration in Section 5. The formulations are standard and clearly explained; someone wanting an entry point to these methods will get a good map of the area and a sensible discussion of the HDG rationale.\n\nThe paper is honest about being an overview, and the cautionary negative result in Section 3—where a fixed cubic representation of the geometry lets the adaptive process converge to the solution of a different problem—is a genuinely useful warning. That part deserves credit.\n\nThe soft spot is exactly the one the stress-test flags. The central claim in Section 3 is that with HDG-NEFEM the exact error decreases monotonically to the prescribed tolerance. But the manuscript never defines how the 'exact error' is computed. There is no analytic solution for the filleted-square electrostatic problem, and the text does not say whether the reference is an overkill HDG-NEFEM solution, a different discretization, or a manufactured solution. Without that definition, the monotonic-decrease claim and the comparison with the cubic-geometry case cannot be independently checked. Relatedly, the local error indicator (7) and the degree-update formula (9) are used for NEFEM curved elements without an effectivity check; the cited analyses are by the same group and mostly concern standard polynomial elements. Figure 3 (right) shows how badly the indicator can fail when the geometry is wrong. The same degree of scrutiny is not shown for the NEFEM run. I find the qualitative conclusion plausible, but as written this is one unreproducible numerical observation, not an established property.\n\nThe rest of the paper is less problematic. The HDG-Voigt and FCFV sections are standard demonstrations, though they too rely on prior papers for the convergence analysis and do not include error tables with computed versus exact errors. For an overview that is acceptable, but a few tabulated numbers would help.\n\nThe citation pattern is self-heavy by design and not a flaw here; the prior work is published in reputable journals. Still, an independent reader has to go elsewhere for verification.\n\nBottom line: this is a useful review-style paper that deserves a serious referee, but the referee should ask for the exact-error definition in Section 3, a short effectivity study for the NEFEM error indicator, and ideally an artifact or a reproducible description of the experiments. I would not accept it as-is, but I would send it to review with those requests.","headline":"Useful consolidation of the authors' HDG variants, but the key new p-adaptivity claim rests on an undefined 'exact error' reference and needs a reproducibility pass before I'd trust the strong version.","tokens_in":19647,"tokens_out":3930,"would_cite":true,"duration_ms":37895,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65N30","65N50","74S05"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper argues that degree-adaptive error decreases monotonically to the tolerance only when the curved boundary is represented exactly (NURBS-enhanced HDG); a frozen polynomial geometry lets the loop stop on the wrong solution.","keywords":["hybridizable discontinuous Galerkin","degree adaptivity","NURBS-enhanced finite element method","exact geometry","HDG-Voigt formulation","superconvergent error indicator","face-centered finite volume","near-incompressible elasticity"],"falsifier":"Run the degree-adaptive loop of Section 3 on a smooth domain with a known analytic solution (for example, a harmonic function on a circular-arc domain representable both by NURBS and by polynomials) and compare, at every iteration, the estimated error from Eq. (7) with the true error against the manufactured solution, under both HDG-NEFEM and fixed-cubic subparametric geometry. If at any iteration the NEFEM path shows a true error far above the estimate, or fails to decrease monotonically to the prescribed tolerance, the central claim is refuted; if the fixed-cubic path again stops with a large true error while its estimate claims success, the claimed failure mode of polynomial-frozen adaptivity is reproduced.","tokens_in":177,"feed_emoji":"📉","tokens_out":11716,"duration_ms":130649,"temperature":0.7,"pith_summary":"This paper surveys three strands of hybridizable discontinuous Galerkin (HDG) methods for computational mechanics and makes a central argument about degree adaptivity: when a curved boundary is approximated once with fixed polynomials and only the solution degree is adapted, the error estimator can declare success while the true error stays far from the tolerance, because the iteration is converging to the solution of a different problem. The authors show that representing the geometry exactly with NURBS-enhanced finite elements (HDG-NEFEM) removes this failure, giving monotone decrease of the error to a user-specified tolerance on a coarse, fixed mesh. The same superconvergence machinery that builds the cheap error indicator also underlies an HDG-Voigt formulation for linear elasticity with pointwise symmetric stresses, optimal convergence at low polynomial degree, and no volumetric locking near the incompressible limit. The paper closes with the face-centered finite volume (FCFV) method, the lowest-order member of the HDG family, applied to a Stokes flow around 39 red-blood-cell-shaped particles in a mesh of nearly nine million tetrahedra.","feed_headline":"Error falls monotonically when geometry is exact","feed_subtitle":"Frozen polynomial boundaries make adaptive loops stop at false solutions; exact NURBS geometry restores convergence.","key_machinery":"The engine of the paper is the pairing of three existing constructions. First, the hybridizable discontinuous Galerkin (HDG) formulation rewrites a second-order PDE into local element problems whose only globally coupled unknowns live on mesh faces, which makes the degrees of freedom cheap and enables a local postprocessing step, Eqs. (5)–(6), that produces a superconvergent approximation $u^\\star$ of degree $k+1$ and accuracy order $k+2$. Second, the NURBS-enhanced finite element method (NEFEM) represents the domain boundary by the exact CAD NURBS curves and adapts the functional space near the boundary to that geometry, so changing the polynomial degree never requires re-approximating or regenerating the curved boundary. Third, the superconvergent $u^\\star$ feeds the elementwise error indicator $E_u^e$ of Eq. (7), and the a priori estimate of Eq. (8), $\\varepsilon_e \\le C h_e^{k_e+1+n_{\\mathrm{sd}}/2}$, combined with Richardson extrapolation of $C$, produces the degree-update rule $\\Delta k_e = \\left\\lceil \\log(\\varepsilon/E_u^e)/\\log(h_e) \\right\\rceil$ of Eq. (9). The HDG-Voigt formulation replaces the stress tensor by its independent entries in Voigt notation, enforcing symmetry of the mixed variable pointwise and yielding optimal convergence with nodal equal-order spaces; FCFV then appears as the constant-degree limit of the same HDG machinery, where all local unknowns admit closed-form expressions in terms of the face unknowns.","core_discovery":"The central claim is that in a degree-adaptive HDG process the geometry of the computational domain must be represented exactly for the iteration to converge to the intended solution. Using the electrostatic problem of a square inclusion with rounded corners, the paper shows that an isoparametric approach with a fixed cubic polynomial description of the rounded fillet stops after eight iterations because the estimated error reaches the tolerance, while the exact error remains far above it—the method is solving a problem with a slightly different, non-smooth geometry. With HDG-NEFEM, in which the NURBS description of the fillet is kept regardless of the polynomial degree of the functional approximation, the exact and estimated errors both decrease monotonically until the desired tolerance of $0.5\\times10^{-3}$ is achieved. The mechanism is the superconvergent postprocessed solution $u^\\star$, which yields a cheap elementwise error indicator; an a priori bound of the form $\\varepsilon_e \\le C h_e^{k_e+1+n_{\\mathrm{sd}}/2}$, with $C$ fixed by Richardson extrapolation, then dictates how much to raise the local polynomial degree. Supporting results establish an HDG-Voigt formulation for linear elasticity whose pointwise symmetric strain-rate variable converges optimally (order $k+1$) even for $k=1,2,3$ and drives error indicators based on either displacement or strain rate, as well as a lowest-order face-centered finite volume method that is LBB-stable, reconstruction-free, and insensitive to mesh distortion and cell stretching.","pith_inferences":["A practical diagnostic suggested by Figure 3 (right) but not stated in the paper: any existing p-adaptive code with a subparametric boundary could monitor the gap between its estimated and exact error on a representatively curved test case, and a persistent gap is a cheap signal that geometric error, not functional error, is dominating the iteration.","The clean separation NEFEM provides between geometric and functional approximation suggests a parameter-free way to size curved meshes: let the CAD model alone determine the boundary representation, so the geometry error is exactly zero rather than balanced against the functional error.","The HDG-Voigt strain-rate indicator could be carried into shape-optimization or fracture-preconditioning loops, where the quantity of interest is the stress concentration at a rounded notch—exactly the region the displacement-based indicator misses; this is an extension the paper motivates but does not perform.","If FCFV's closed-form local elimination is the reason for its cheap assembly, the same decoupling should make the global face system amenable to algebraic multigrid or domain-decomposition preconditioning at scales far beyond the reported serial experiment, a direct but untested consequence of the block structure in Eq. (26)."],"forward_implications":["Degree-adaptive HDG on a fixed coarse mesh can reach a user-prescribed tolerance in electrostatics without mesh regeneration, provided the curved boundary is the exact NURBS geometry; with a frozen cubic boundary it stops prematurely.","The superconvergent postprocessed displacement provides an inexpensive elementwise error indicator that drives p-adaptivity in linear elasticity for both displacement- and stress-accuracy goals, the strain-rate indicator localizing stress concentrations.","HDG-Voigt supplies a pointwise symmetric stress and strain-rate approximation that is optimally convergent for $k=1,2,3$ and free of volumetric locking for nearly incompressible materials, as demonstrated on Cook's membrane with $\\nu=0.499999975$.","The same HDG rationale, taken to constant degree, yields FCFV: a first-order, LBB-stable, reconstruction-free finite volume method insensitive to distorted and stretched cells, exercised here on a 61-million-unknown Stokes problem in a serial Matlab implementation.","Because the NEFEM boundary is degree-independent, the degree-adaptive strategy composes directly with h-refinement, yielding hp-adaptive processes that keep the exact CAD geometry throughout."],"supporting_citations":[{"why":"introduces the HDG-NEFEM degree-adaptive framework for Stokes flows that this paper extends to electrostatics and whose error-indicator machinery is reused.","marker":"[82]"},{"why":"defines the NURBS-enhanced finite element method (NEFEM) that supplies the exact-geometry boundary treatment used by HDG-NEFEM.","marker":"[114]"},{"why":"provides the local a priori error estimate $\\varepsilon_e \\le C h_e^{k+1+n_{\\mathrm{sd}}/2}$ that underlies the degree-update formula in Eq. (9).","marker":"[115]"},{"why":"represents the common practice of degree adaptation with fixed quadratic or cubic boundary polynomials whose failure is demonstrated in Figure 3.","marker":"[112]"},{"why":"extends that fixed-geometry degree-adaptive practice to the Navier-Stokes equations and is the immediate methodological contrast for the NEFEM result.","marker":"[113]"},{"why":"establishes the superconvergence of the postprocessed HDG primal solution used to construct the cheap error indicator of Eq. (7).","marker":"[35]"},{"why":"introduces the HDG-Voigt formulation for linear elasticity with strongly enforced pointwise symmetric stresses that Section 4 builds on.","marker":"[57]"},{"why":"introduces the face-centered finite volume (FCFV) method whose large-scale Stokes application is reported in Section 5.","marker":"[106]"}],"fun_headline_variants":["Exact NURBS geometry prevents false adaptive convergence","Isoparametric geometry stalls HDG adaptation—exact NURBS converges","HDG-NEFEM: exact geometry restores monotonic error reduction","Adaptive HDG fails with polynomial geometry—exact NURBS fixes it","Degree-adaptive HDG needs exact geometry for true convergence"],"cache_read_input_tokens":21760,"weakest_assumption_plain":"The whole degree-adaptive loop rests on the assumption that the error bound $\\varepsilon_e \\le C h_e^{k_e+1+n_{\\mathrm{sd}}/2}$, with the constant $C$ obtained by Richardson extrapolation from two computed polynomial degrees, correctly predicts how the error will fall when the degree is raised; the paper itself shows this prediction failing badly in Figure 3 (right) whenever the boundary is frozen as cubic polynomials, and it never independently verifies the reliability of the indicator for the NEFEM case.","fun_headline_variants_meta":{"raw":{"variants":["Exact NURBS geometry prevents false adaptive convergence","Isoparametric geometry stalls HDG adaptation—exact NURBS converges","HDG-NEFEM: exact geometry restores monotonic error reduction","Adaptive HDG fails with polynomial geometry—exact NURBS fixes it","Degree-adaptive HDG needs exact geometry for true convergence"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000403,"raw_usage":{"total_tokens":2132,"prompt_tokens":1006,"completion_tokens":1126,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":622,"completion_tokens_details":{"reasoning_tokens":1033}},"tokens_in":622,"tokens_out":1126,"duration_ms":10620,"temperature":1.0,"reasoning_tokens":1033,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:39:35.191806+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the degree-adaptive loop of Section 3 on a smooth domain with a known analytic solution (for example, a harmonic function on a circular-arc domain representable both by NURBS and by polynomials) and compare, at every iteration, the estimated error from Eq. (7) with the true error against the manufactured solution, under both HDG-NEFEM and fixed-cubic subparametric geometry. If at any iteration the NEFEM path shows a true error far above the estimate, or fails to decrease monotonically to the prescribed tolerance, the central claim is refuted; if the fixed-cubic path again stops with a large true error while its estimate claims success, the claimed failure mode of polynomial-frozen adaptivity is reproduced.","supporting_citations":[{"cited_title":"Sevilla and A","cited_arxiv_id":null,"evidence_quote":"introduces the HDG-NEFEM degree-adaptive framework for Stokes flows that this paper extends to electrostatics and whose error-indicator machinery is reused."},{"cited_title":"Sevilla, S","cited_arxiv_id":null,"evidence_quote":"defines the NURBS-enhanced finite element method (NEFEM) that supplies the exact-geometry boundary treatment used by HDG-NEFEM."},{"cited_title":"D´ ıez and A","cited_arxiv_id":null,"evidence_quote":"provides the local a priori error estimate $\\varepsilon_e \\le C h_e^{k+1+n_{\\mathrm{sd}}/2}$ that underlies the degree-update formula in Eq. (9)."},{"cited_title":"Giorgiani, S","cited_arxiv_id":null,"evidence_quote":"represents the common practice of degree adaptation with fixed quadratic or cubic boundary polynomials whose failure is demonstrated in Figure 3."},{"cited_title":"Giorgiani, S","cited_arxiv_id":null,"evidence_quote":"extends that fixed-geometry degree-adaptive practice to the Navier-Stokes equations and is the immediate methodological contrast for the NEFEM result."},{"cited_title":"Sevilla, M","cited_arxiv_id":null,"evidence_quote":"introduces the HDG-Voigt formulation for linear elasticity with strongly enforced pointwise symmetric stresses that Section 4 builds on."},{"cited_title":"Sevilla, M","cited_arxiv_id":null,"evidence_quote":"introduces the face-centered finite volume (FCFV) method whose large-scale Stokes application is reported in Section 5."}],"review_version":1}