{"id":"768fddfe-d4d8-454c-b1a4-e144f77735c3","arxiv_id":"2411.16067","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A really pressure-robust virtual element method for the Brinkman problem is proven to have viscosity-independent velocity error, with residual a posteriori error bounds and adaptive mesh refinement.","lead":"This paper builds a virtual element method for the incompressible Brinkman equations whose velocity error does not depend on pressure or viscosity, and it adds an adaptive refinement loop driven by a proven error estimator. The method works on general polygonal meshes, which matters for porous-media flows with strongly varying permeability.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Efficiency theorem (Thm 5.2) bounds only η_r,E, not the full estimator; the data term h||f|| is never controlled, so the claimed local lower bound for η does not follow as stated.","rationale":"The reader's weakest assumption — that the interpolant u_I may not lie in Z_h — does not actually land: the VEM interpolant matches edge-normal degrees of freedom, ∇·v_I is piecewise constant on each element, and for divergence-free u one gets ∫_E ∇·u_I = ∮_{∂E} u_I·n = ∮_{∂E} u·n = ∫_E ∇·u = 0, hence u_I ∈ Z_h. The a priori proof in Theorem 4.1 does contain a repairable notational gap: a(u, R_hv_h − v_h) is used although R_hv_h is only in H(div), not H^1; however T1 can be bounded directly as ν(−Δu + κ^{-1}u, R_hv_h − v_h) using (3.9c), so the a priori estimate remains plausible after a correction. The genuinely load-bearing issue is in the a posteriori section: Theorem 5.2 proves a local upper bound for only η_r,E, not a lower bound for the full η_E, and no data-oscillation term is introduced to control η_f,E. This is a logical gap in a claim made in the abstract and used by the adaptive refinement algorithm, even though the numerical experiments are consistent with the estimator's practical behavior. The reader's CONDITIONAL verdict remains appropriate, but for the efficiency-gap reason rather than the interpolation concern.","tokens_in":24501,"tokens_out":16977,"duration_ms":158569,"concrete_test":"Run Example 6.4 on a locally refined sequence of meshes and record, on elements where the exact solution is smooth and well resolved, the ratio η_E / (ν9u−u_h9_E + ∥p−p_h∥_E). Because η_f,E = h_E∥f∥_E does not vanish while the local discretization error is near machine precision, the ratio will grow, demonstrating that the full estimator is not efficient in the sense claimed. Analytically, substitute (5.20) into (5.22) and attempt to close a bound for η²_E; the term Σ_E h²_E∥f∥²_E remains on the right-hand side with no oscillation term, so the claimed local lower bound for η cannot be derived from the theorem as stated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 5.2 proves (5.18)/(5.23): η²_r,E ≤ C_r Σ(ν²9u−u_h9²_E + ∥p−p_h∥²_E + η²_S,E + η²_f,E). This bounds only the residual sub-estimator η_r,E, while the adaptive algorithm and the abstract use the full estimator η_E = (η²_f,E + η²_S,E + η²_r,E)^{1/2}. The right-hand side contains η²_f,E = h²_E∥f∥²_E, and no later inequality bounds η²_f,E by the discretization error plus a data-oscillation term. Summing (5.23) and adding η²_f + η²_S to both sides does not close a bound of the form η²_E ≤ C(error² + osc²): the term h²∥f∥² is simply carried along, and for nonzero f it does not vanish even where the discrete solution is locally exact. Thus the abstract's claim that the estimator 'provides global upper and local lower bounds for the discretization error' is not established, and the stated efficiency theorem is weaker than the claim built on it. The numerical effectivity indices, which include all three components of η by construction, do not repair this logical gap.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a lowest-order virtual element method for the two-dimensional incompressible Brinkman problem. The right-hand side is discretized with a divergence-preserving Raviart-Thomas reconstruction operator, which makes the discrete velocity exactly divergence-free. The authors claim optimal a priori error estimates in which the velocity error in the energy norm is independent of both the continuous pressure and the viscosity, and they construct a residual-type a posteriori estimator with a global upper bound and a local lower bound. They also present an adaptive refinement algorithm and numerical experiments on polygonal meshes, including high-contrast permeability and singular geometries.","tokens_in":24771,"tokens_out":12008,"duration_ms":114428,"significance":"If the stated results are correct, the paper makes a useful contribution to pressure-robust virtual element methods for Brinkman flows on polygonal meshes: it extends divergence-preserving reconstruction ideas from Stokes-type problems to the Brinkman setting, gives explicit h-optimal rates with constants independent of ν and p in the velocity estimate, and provides an adaptive estimator with effectivity studies. The numerical section is reasonably broad, covering non-convex, Voronoi, and high-contrast permeability cases. The proof strategy is transparent and the reported experiments support the pressure-robustness claims. However, several load-bearing points in the a priori and a posteriori arguments are not justified as written.","major_comments":[{"comment":"The proof sets v_h = u_h - u_I and inserts this into the identity (4.2), which is asserted for arbitrary v_h in the discrete kernel Z_h. Lemma 4.1 supplies only approximation estimates and does not state or prove that the interpolant u_I is divergence-free, i.e. that u_I ∈ Z_h. The canonical VEM interpolant with matching edge-normal degrees of freedom would satisfy ∫_E div u_I = ∫_{∂E} u·n = 0, and since div v is piecewise constant for v ∈ V_h, this would imply u_I ∈ Z_h; but this property is not stated. Without it, identity (3.17) cannot be invoked at v_h = u_h - u_I, and the cancellation of the pressure term in T1 is not justified. Please add an explicit statement and proof that u_I is divergence-preserving before Eq. (4.2).","section":"§4, proof of Theorem 4.1, Eq. (4.2)"},{"comment":"The bound of ν(κ^{-1}u, R_h v_h - v_h) is displayed with the norm ∥κ^{1/2}u∥. The correct Cauchy-Schwarz pairing is (κ^{-1/2}u, κ^{-1/2}(R_h v_h - v_h)), giving a bound involving ∥κ^{-1/2}u∥ and ∥κ^{-1/2}∥∞, not ∥κ^{1/2}u∥. This error propagates into the explicit constant bC in (4.1a) and (4.6). The h-dependence and the independence of ν and p would survive the correction, but the displayed formula is not correct as written.","section":"§4, Eq. (4.3) and constant bC in (4.6)"},{"comment":"Theorem 5.2 bounds only the residual sub-estimator η_r,E, not the full local estimator η_E = (η_f,E² + η_S,E² + η_r,E²)^{1/2}. The right-hand side of (5.18)/(5.23) contains η_f,E² = h_E²∥f∥²_E and η_S,E², so adding these terms to both sides yields only an inequality in which the estimator is bounded by itself together with error terms. This does not establish a local lower bound for the full estimator in terms of the discretization error alone; in particular, h_E∥f∥_E need not vanish where the discrete solution is locally accurate. Since the adaptive marking (5.24) and the numerical effectivity indices use the full η, the abstract's claim that the estimator 'provides global upper and local lower bounds for the discretization error' is not established as stated. If the data term is intended as a data-oscillation term, this should be stated explicitly and the theorem and abstract rephrased accordingly.","section":"§5.2, Theorem 5.2 and Abstract"},{"comment":"The efficiency proof applies the bubble-function estimates of Lemma 5.2 to the vector function θ_E = κ^{-1}Π_{0,E}u_h. This is only valid if θ_E is a polynomial on E, which requires the permeability tensor κ to be piecewise constant on the mesh. The paper never states such an assumption on κ; for a general space-dependent symmetric positive definite tensor, θ_E is not a polynomial and Lemma 5.2 does not apply. Please add the missing coefficient regularity assumption or modify the proof accordingly.","section":"§5.2, Eqs. (5.19)-(5.21)"}],"minor_comments":[{"comment":"The phrase 'optimal priori error estimates' should be 'optimal a priori error estimates'; in the last paragraph of Section 1, 'The optimal a posteriori error estimate is derived in Section 4' should read 'a priori', since Section 4 concerns a priori estimates.","section":"Abstract and §1"},{"comment":"The inequality in (5.16b) has a dimension mismatch: the middle expression is a norm, not a squared norm, while the left and right bounds are written with ∥p1∥_E². It should presumably be ∥p1∥_E on both sides.","section":"§5.2, Lemma 5.2, Eq. (5.16b)"},{"comment":"The table caption says 'the errors for a series of Voronoi meshes', but the text states the test starts from the non-convex base mesh of Fig. 6.1(a); please make the mesh description consistent.","section":"§6.2, Example 6.4, Table 6.2"},{"comment":"The right-hand side of (5.18) references η_f,E and η_S,E before their definitions in (5.5a)-(5.5b) are recalled; a forward reference to (5.5) would improve readability.","section":"§5.2, Theorem 5.2 statement"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the scope of the journal, and I see no citation or novelty concern beyond the technical issues above. The self-citations [13,45] are used for background VEM facts and are not circular with respect to the main claims."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nQuick take: this is a competent pressure-robust VEM paper for the Brinkman problem, but the a priori proof has an interpolation gap and the a posteriori efficiency theorem does not prove what the abstract claims. Worth engaging, but it needs revision.\n\nWhat's actually new: the residual-type a posteriori estimator for this RT-reconstruction-based PR-VEM, the adaptive refinement loop, and a strong set of numerical experiments (high-contrast permeability, boundary layers, corner singularity, channel flow). The scheme itself is a close variant of the CW0 reconstruction for Brinkman in Wang et al. [40] and the RT-based reconstruction for Navier-Stokes in [42]; the authors acknowledge this. The a priori analysis aiming at nu- and p-independence is the right goal and largely plausible.\n\nSoft spots, in order of severity:\n\n1. The proof of Theorem 4.1 uses identity (4.2) for v_h in Z_h and then plugs in v_h = u_h - u_I. This requires u_I in Z_h. Lemma 4.1 only gives approximation estimates; it says nothing about divergence preservation. Without a divergence-preserving interpolant, the velocity bound (4.1a) doesn't follow. This is fixable—the literature [21,42] has such interpolants—but it needs to be stated and proven.\n\n2. In (4.3), the term ||kappa^{1/2}u|| should read ||kappa^{-1/2}u|| (or similar). As written, the kappa-power is backwards and would give a nonsensical dependence on the permeability.\n\n3. Theorem 5.2 proves a bound for eta_{r,E} only, not the full estimator eta_E = (eta_f^2 + eta_S^2 + eta_r^2)^{1/2}. The right-hand side contains eta_f,E, and the proof never controls that data term by error plus oscillation. So the abstract's claim that the estimator provides local lower bounds for the discretization error is not established by the stated theorem. The numerical effectivity indices are encouraging but don't repair the logical gap.\n\nOn the positive side: the reliability bound (Theorem 5.1) is argued carefully, the mesh refinement strategy is sensible, and the experiments are detailed and reproducible enough to be useful. The citations to [21] and [42] for the reconstruction are appropriate, and the self-citations are background, not load-bearing.\n\nThis paper deserves a serious referee. The right recommendation: send it out, and require the authors to close the Z_h interpolation question, correct the kappa-power, and either prove efficiency for the full estimator or restate the claim to cover only the residual part plus a data-oscillation term. With those changes it would be a solid contribution to adaptive VEM for Brinkman.","headline":"Solid PR-VEM paper for Brinkman with a fixable a priori gap and an overstated a posteriori efficiency claim; worth refereeing after revision.","tokens_in":25282,"tokens_out":3821,"would_cite":true,"duration_ms":33888,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65N15","65N30","76D07","35J47","35K55"],"pacs":[],"model":"deepseek-v4-flash","headline":"A virtual element method for the incompressible Brinkman problem is proved to keep the energy-norm velocity error independent of both pressure and viscosity, with a residual error estimator that drives adaptive refinement on polygonal…","keywords":["Brinkman equations","virtual element method","pressure-robust","locking-free","a priori error estimates","a posteriori error estimates","adaptive mesh refinement","divergence-preserving reconstruction"],"falsifier":"Solve a manufactured divergence-free problem with a large pressure component on a fixed non-convex polygonal mesh for $\\nu=10^{-4},10^{-8},10^{-12}$; if the energy-norm velocity error changes measurably, or if the computed $L^2$ norm of $\\nabla\\cdot u_I$ for the interpolation used in Theorem 4.1 is not zero to round-off, then the claimed independence from $\\nu$ and $p$ fails, and the proof's test-function step (4.2) is not valid.","tokens_in":24274,"feed_emoji":"🌊","tokens_out":13954,"duration_ms":113645,"temperature":0.7,"pith_summary":"The paper is trying to establish that the Brinkman flow equations can be solved on arbitrary polygonal meshes by a virtual element method whose velocity error does not blow up as the viscosity tends to zero and does not depend on the unknown pressure. Standard divergence-free virtual elements fail this test because the discrete right-hand side perturbs the divergence, so a consistency term proportional to $1/\\nu$ enters the velocity estimate and causes locking. The authors repair this by applying a divergence-preserving reconstruction operator to the load before computing the right-hand side, and they prove optimal $O(h)$ a priori error estimates whose velocity constant is independent of $\\nu$ and $p$. They also construct a residual-type a posteriori estimator, prove global upper and local lower bounds for the full discretization error, and use it in an adaptive refinement loop. If the analysis is right, the method is usable for porous-media flows spanning Darcy- and Stokes-dominated regions, where permeability contrasts are extreme.","feed_headline":"Velocity error free of viscosity and pressure in Brinkman solver","feed_subtitle":"Virtual element method stays accurate as viscosity vanishes and drives adaptive refinement.","key_machinery":"The load-bearing object is the divergence-preserving reconstruction operator $R_h$, defined elementwise on a regular subtriangulation of each polygon by the lowest-order H(div)-conforming space: its normal traces are matched to the virtual element's edge degrees of freedom, and it preserves the elementwise divergence of the virtual velocity. It is inserted only where the load meets the discrete test space, replacing $(f,v_h)$ by $(f,R_h v_h)$. On the discrete kernel $Z_h=\\{v_h:\\nabla\\cdot v_h=0\\}$, this yields $\\nabla\\cdot R_h v_h=0$, so the gradient part of the force is orthogonal to the discrete velocity space and cannot feed a $1/\\nu$ term into the velocity estimate. The stabilization terms in the discrete bilinear form are chosen to satisfy the usual consistency and stability estimates, and the discrete inf-sup condition supplies the pressure estimate; the main proof of Theorem 4.1 then isolates the interpolation error and the reconstruction error. This machinery is also what makes the a posteriori estimator treatable, because the residual equation inherits the same divergence-preserving structure.","core_discovery":"The central discovery is a really pressure-robust virtual element scheme whose velocity error in the energy norm satisfies $\\|u-u_h\\|_{\\text{energy}} \\le \\hat C h$ with $\\hat C$ independent of $h$, of the viscosity $\\nu$, and of the continuous pressure $p$, while the pressure error is $\\|p-p_h\\|\\le \\tilde C h$ with $\\tilde C$ independent of $h$. The mechanism is the reconstruction operator $R_h$: it is a piecewise divergence-preserving mapping into a lowest-order H(div)-conforming space on a subtriangulation, matched to the virtual degrees of freedom, and it is applied only to the load. Because $\\nabla\\cdot R_h v_h=0$ on the discrete divergence-free kernel $Z_h$, the gradient part of the load's decomposition no longer pollutes the velocity equation; this is exactly the orthogonality that the standard projection-based right-hand side destroys. The paper also proves reliability and efficiency of a residual estimator $\\eta$, in the form $\\nu\\|u-u_h\\|+\\|p-p_h\\|\\le C_\\eta\\eta$ and local lower bounds, and confirms the predicted rates on non-convex and centroidal-type polygonal meshes, including high-contrast permeability examples.","pith_inferences":["The same divergence-preserving right-hand side should transfer to other saddle-point flow formulations whose locking comes only from losing orthogonality to gradient forces, such as viscoelastic or magnetohydrodynamic models; a direct test is to apply the reconstruction to a Stokes problem with a large pressure gradient and check that no $1/\\nu$ term appears.","The reliability bound suggests a $\\nu$-independent adaptive stopping criterion can be set from the pressure error plus a weighted estimator; an experiment comparing effectivity indices for $\\nu=1$ and $\\nu=10^{-12}$ on the same mesh sequence would make this explicit.","The efficiency constant in Theorem 5.2 depends on $\\|\\kappa^{-1/2}\\|_\\infty$ through the stabilization bounds, so a checkerboard-permeability test with contrast above $10^6$ would probe whether the estimator remains reliable in the high-contrast regime the paper targets."],"forward_implications":["As $\\nu\\to0$, the velocity error in the energy norm stays bounded and first-order in $h$, so the method is locking-free across the Darcy-to-Stokes transition.","The discrete velocity is exactly divergence-free (since $\\nabla\\cdot V_h\\subset Q_h$), so mass conservation holds elementwise even on coarse polygonal meshes.","The residual estimator $\\eta$ gives a computable upper bound on $\\nu\\|u-u_h\\|+\\|p-p_h\\|$ and local lower bounds, allowing the adaptive loop Solve–Estimate–Mark–Refine to run on general polygonal meshes with guaranteed error control.","Optimal convergence rates $O(h)$ a priori and $O(\\mathrm{DOFs}^{-1/2})$ for adaptively refined meshes are observed for smooth solutions, boundary and interior layers, corner singularities, and flow past a cylinder."],"supporting_citations":[{"why":"Supplies the divergence-preserving reconstruction technique on polygons that the right-hand side discretization $(\\mathbf f, R_h v_h)$ is built on.","marker":"[21]"},{"why":"Establishes the really pressure-robust VEM framework for the Brinkman equations whose reconstruction properties (Lemma 3.1) are reused here.","marker":"[40]"},{"why":"Gives the implementation details of the reconstruction operator on a polygon, quoted as the source for conditions (3.8).","marker":"[42]"},{"why":"Provides the divergence-free H1-conforming VEM for Brinkman that is the baseline scheme whose $1/\\nu$ consistency error this paper removes.","marker":"[36]"},{"why":"The only previous a posteriori error analysis of a mixed VEM for a nonlinear Brinkman model, the comparison point for the estimator developed here.","marker":"[34]"},{"why":"Supplies the residual-type a posteriori error estimation strategy for the VEM Stokes problem that the proof in Section 5 follows.","marker":"[39]"},{"why":"Contributes the adaptive VEM framework and bubble-function techniques used for the local efficiency bounds.","marker":"[43]"},{"why":"Defines the basic virtual element spaces and mesh assumptions on which the discretization rests.","marker":"[15]"}],"fun_headline_variants":["Velocity error ignores viscosity and pressure in Brinkman VEM","Pressure-robust VEM drives adaptive mesh refinement","Brinkman solver: error bounds that don't care about pressure","VEM for Brinkman: truly pressure-robust, adaptively refined"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The velocity-error proof assumes that the virtual-element interpolant of the exact divergence-free solution is itself exactly divergence-free, so it lies in the discrete kernel; the paper states only approximation estimates for the interpolant and gives no separate proof of this preservation property.","fun_headline_variants_meta":{"raw":{"variants":["Velocity error ignores viscosity and pressure in Brinkman VEM","Pressure-robust VEM drives adaptive mesh refinement","Brinkman solver: error bounds that don't care about pressure","VEM for Brinkman: truly pressure-robust, adaptively refined"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000474,"raw_usage":{"total_tokens":2351,"prompt_tokens":940,"completion_tokens":1411,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":556,"completion_tokens_details":{"reasoning_tokens":1339}},"tokens_in":556,"tokens_out":1411,"duration_ms":10839,"temperature":1.0,"reasoning_tokens":1339,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:35:18.165804+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Solve a manufactured divergence-free problem with a large pressure component on a fixed non-convex polygonal mesh for $\\nu=10^{-4},10^{-8},10^{-12}$; if the energy-norm velocity error changes measurably, or if the computed $L^2$ norm of $\\nabla\\cdot u_I$ for the interpolation used in Theorem 4.1 is not zero to round-off, then the claimed independence from $\\nu$ and $p$ fails, and the proof's test-function step (4.2) is not valid.","supporting_citations":[{"cited_title":"Divergence-preserving reconstructions on polygons and a really pressure- robust virtual element method for the Stokes problem","cited_arxiv_id":null,"evidence_quote":"Supplies the divergence-preserving reconstruction technique on polygons that the right-hand side discretization $(\\mathbf f, R_h v_h)$ is built on."},{"cited_title":"Two robust virtual element methods for the Brinkman equations","cited_arxiv_id":null,"evidence_quote":"Establishes the really pressure-robust VEM framework for the Brinkman equations whose reconstruction properties (Lemma 3.1) are reused here."},{"cited_title":"A pressure-robust virtual element method for the Navier-Stokes problem on polygonal mesh","cited_arxiv_id":null,"evidence_quote":"Gives the implementation details of the reconstruction operator on a polygon, quoted as the source for conditions (3.8)."},{"cited_title":"An H1-conforming virtual element for Darcy and Brinkman equations","cited_arxiv_id":null,"evidence_quote":"Provides the divergence-free H1-conforming VEM for Brinkman that is the baseline scheme whose $1/\\nu$ consistency error this paper removes."},{"cited_title":"Sequeira","cited_arxiv_id":null,"evidence_quote":"The only previous a posteriori error analysis of a mixed VEM for a nonlinear Brinkman model, the comparison point for the estimator developed here."},{"cited_title":"A posteriori error estimates for the virtual element method for the Stokes problem","cited_arxiv_id":null,"evidence_quote":"Supplies the residual-type a posteriori error estimation strategy for the VEM Stokes problem that the proof in Section 5 follows."},{"cited_title":"An adaptive virtual element method for incompressible flow","cited_arxiv_id":null,"evidence_quote":"Contributes the adaptive VEM framework and bubble-function techniques used for the local efficiency bounds."},{"cited_title":"Basic principles of virtual element methods","cited_arxiv_id":null,"evidence_quote":"Defines the basic virtual element spaces and mesh assumptions on which the discretization rests."}],"review_version":1}