{"id":"ee6cb2ce-efb1-48b1-a7bb-db2190acdc84","arxiv_id":"1908.01892","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A unified nonconforming Crouzeix-Raviart finite element method for the Stokes-Darcy problem is shown to be well-posed and to converge at first order in the mesh size.","lead":"This paper proposes a finite element method that uses the same nonconforming Crouzeix-Raviart element in both the free-fluid (Stokes) and porous-medium (Darcy) regions of a coupled flow problem. It proves a first-order error bound and shows one numerical example. A generalist might read it to see how numerical analysts build cheaper unified discretizations for fluid-porous coupling.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Interpolation lemma for the enlarged H_h is unproved; r_h may not map into H_h on Γ_d, so the Fortin argument and Theorem 3.4 are unsupported.","rationale":"The strongest claim is the O(h) error estimate in Theorem 3.4, so I checked the two ingredients that must hold for it to follow: the discrete inf-sup condition (Theorem 3.2) and the interpolation estimate (Lemma 3.2). Both depend on r_h. The paper acknowledges that H_h is larger than the space in [30] (page 9, after (28)), and the proof of Lemma 3.2 is deferred to [30], which is precisely where the enlargement matters. As written, (49)-(50) define r_h only on edges of the open subdomains Ω_s and Ω_d, so boundary edges are unconstrained; the Fortin proof in Theorem 3.2 nevertheless sums over Ω_s^+ and invokes (49), an internal mismatch. The most concrete failure is on Γ_d: v∈W has zero normal trace, but no interpolation condition enforces this on r_h v. Hence r_h(W)⊄H_h unless additional conditions are silently added. I do not claim the method is wrong; a corrected definition of r_h on all edges, or a direct proof of Lemma 3.2 for the normal-only space, may repair the argument, and the numerical experiment in Section 4 is consistent with the predicted rate. But until such a proof exists, Theorem 3.4 is unsupported. This echoes the reader's weakest assumption (agree), and the appropriate disposition remains conditional: accept only after Lemma 3.2 and the Fortin boundary treatment are fixed. I therefore affirm the reader's CONDITIONAL verdict rather than rejecting the paper outright.","tokens_in":16448,"tokens_out":11729,"duration_ms":120680,"concrete_test":"Take a triangle T in Ω_d with an edge E on Γ_d and let v∈W be supported in T with v·n_d=0 on E but v not identically zero on E. Compute the affine interpolant r_h v on T from (49)-(50), applied only to the edges in E_h(Ω_d); since no condition is imposed on E, (r_h v)·n_E is generally nonzero on E, so ∫_E (r_h v)·n_E ≠ 0 and r_h v violates (29). This decides whether r_h maps W into H_h; if it does not, Lemma 3.2, Theorem 3.2, and Theorem 3.4 fail as stated. If the authors intended (50) to include E_h(∂Ω_d), the same test with an interior Darcy edge shows that (50) enforces full-vector continuity, stronger than (29), so the enlargement claim still needs a separate proof of Lemma 3.2.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The proof of Theorem 3.4 is routed through Lemma 3.2 (boundedness of the interpolation operator r_h in the discrete norm) and the Fortin argument in Theorem 3.2. Lemma 3.2 is not proved; the proof is deferred to [30], but the space H_h here is larger: by (28)-(29), full-vector jump vanishing is required only on E_h(Ω_s^+), while on E_h(Ω_d) ∪ E_h(∂Ω_d) only the normal component of the jump is constrained. The interpolation operator defined in (49)-(50) uses full-vector edge moments over E_h(Ω_s) and E_h(Ω_d), i.e. only edges of the open subdomains; no condition is imposed on boundary edges Γ_s, Γ_d, Γ_I. For v∈W, r_h v therefore need not satisfy the H_h boundary conditions in (28)-(29); in particular v·n_d=0 on Γ_d is not transferred to r_h v because (50) does not include E_h(∂Ω_d). The Fortin proof in Theorem 3.2 then sums jumps over E_h(Ω_s^+) ∪ E_h(Ω_d) and invokes (49) for Ω_s^+ although (49) was stated only for Ω_s; the missing Γ_d boundary term Σ_{E⊂Γ_d} ∫_E q_h n_E·(r_h v) is generally nonzero. Since Lemma 3.2 and the Fortin argument are the only bridge from the enlarged space to the inf-sup and approximation estimates, the O(h) bound in Theorem 3.4 is unsupported until the interpolation map is corrected and Lemma 3.2 is proved for the actual H_h.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a unified stabilized mixed finite element method for the coupled Stokes–Darcy problem. The Darcy equation is modified by adding a div-div term, and both subproblems are discretized with a variant of the nonconforming Crouzeix–Raviart element in a single velocity space, with full-vector jump vanishing only in the Stokes region and normal-component jump vanishing in the Darcy region and on its boundary, together with a jump stabilization term. The paper claims well-posedness of the discrete problem, a discrete inf-sup condition, and an optimal first-order a priori error estimate under H^2/H^1 regularity, and reports one numerical experiment in two dimensions that shows the expected convergence behavior.","tokens_in":16817,"tokens_out":9912,"duration_ms":104335,"significance":"If the proof were completed, the result would be genuinely useful: it would justify a cheap, unified nonconforming discretization in which the Darcy velocity space only approximates H(div), which is more natural than enforcing full H^1 continuity in the porous medium. The scheme itself and the stabilization are reasonable, and the numerical experiment is consistent with the claimed rate. However, the central convergence argument depends on an interpolation operator that is not shown to map into the actual discrete space and is deferred to a reference that treats a smaller space. As written, the discrete inf-sup condition and the convergence theorem are not established. The contribution is promising, but the missing proof is load-bearing and requires a substantive repair.","major_comments":[{"comment":"The interpolation operator r_h defined by (49)-(50) is not shown to map W into the actual space H_h. H_h in (28)-(29) requires full-vector jump moments to vanish on E_h(Ω_s^+) and normal-component jump moments to vanish on E_h(Ω_d) ∪ E_h(∂Ω_d), whereas (49)-(50) only fix edge moments on E_h(Ω_s) and E_h(Ω_d). In particular, the boundary-edge conditions on Γ_s, Γ_d, and Γ_I are not imposed by (49)-(50): for v∈W with v=0 on Γ_s and v·n_d=0 on Γ_d, there is no reason that r_h v satisfies the corresponding discrete conditions. Since Lemma 3.2 is proved only by saying \"similar to [30]\" and the space in [30] is smaller, the boundedness of r_h in the discrete norm is unproved for the enlarged H_h. This lemma underpins both the Fortin argument and the approximation estimate in Lemma 3.4, so the O(h) claim in Theorem 3.4 is unsupported as written.","section":"Section 3.1, Lemma 3.2 and equations (28)-(29), (49)-(50)"},{"comment":"The Fortin proof integrates by parts over all elements and then restricts the resulting jump sum to E_h(Ω_s^+) ∪ E_h(Ω_d), omitting the edges in E_h(∂Ω_d). Because r_h v is not known to lie in H_h, the boundary terms on Γ_d and Γ_I are not zero, so the identity b_h(r_h v, q_h) = b_h(v, q_h) does not follow from (49)-(50). For v∈[H^1_0(Ω)]^N, the term Σ_{E⊂∂Ω_d} ∫_E q_h n_E·(r_h v) is generally nonzero unless additional conditions are imposed on r_h. Thus the discrete inf-sup condition (52) is not established.","section":"Section 3.1, Theorem 3.2 (Fortin argument)"},{"comment":"The coercivity proof contains some nontrivial steps that are only sketched: the transition from (42) to (43) and then to (44) relies on a discrete Korn inequality for a space with mixed jump conditions, but the precise hypotheses under which [5] applies to H_h are not stated. In addition, the inequalities in (45)-(47) use the symbol \">\" where a lower bound with a constant is intended. These issues are repairable, but they add to the impression that the well-posedness proof is not fully self-contained.","section":"Section 3.1, Theorem 3.1 and equations (43)-(47)"}],"minor_comments":[{"comment":"The expressions ([v_h·n_E, p_s]_E)_E and ([v_h·n_E, p_d]_E)_E are typographically confusing; they should be written as ([v_h·n_E]_E, p_s)_E and ([v_h·n_E]_E, p_d)_E, respectively.","section":"Section 3.2, Lemma 3.5"},{"comment":"The final bound in (72) contains the typo |u2,d; it should read |u|_{2,d}. The word \"satisfies\" is also misspelled in the theorem statement.","section":"Section 3.2, Theorem 3.4"},{"comment":"Only one numerical test is reported, and the observed convergence orders are not extracted or tabulated. Reporting the empirical orders for ‖u−u_h‖_h and ‖p−p_h‖ would make the numerical confirmation of the O(h) claim more convincing.","section":"Section 4, numerical experiments"},{"comment":"The decomposition E_h = E_h(Ω_s^+) ∪ E_h(Ω_d) ∪ E_h(∂Ω_d) in (24) should be stated with a short explanation that E_h(Ω_s^+) includes Γ_s, while E_h(∂Ω_d) includes Γ_d and Γ_I; this would help the reader see exactly which jump terms are penalized.","section":"Section 3.1, notation"}],"recommendation":"major_revision","confidential_remarks":"The central issue is the unproved interpolation lemma for the enlarged space H_h. I would not recommend acceptance until Lemma 3.2 is proved for the actual discrete space, including boundary-edge conditions, and the Fortin argument is repaired accordingly. If the interpolation operator cannot be extended in this way, the convergence theorem should be restricted to a space for which the argument is valid. The reliance on the author's prior work for Theorem 2.1 and on [30] for several key estimates is acceptable only if the proofs are actually transferable; here they are not."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The idea is genuinely new and sensible: enlarge the Darcy velocity space to require only normal continuity, add a div-div stabilization, and use the same Crouzeix-Raviart element in both subdomains. That gives a cheaper method, and the numerical experiments are consistent with first-order convergence. I think the author is onto something worth pursuing.\n\nThe problem is the proof. Lemma 3.2 is not proved; it says “similar to [30]” but [30] used a smaller space. The interpolation operator defined in (49)–(50) uses full-vector moments on subdomain edges and ignores boundary edges. H_h, however, requires normal-only continuity on ∂Ω_d and on Ω_d interior edges, so r_h v need not satisfy the H_h boundary conditions. In particular, v·n_d = 0 on Γ_d is lost; the Fortin proof sums jumps only over E_h(Ω_s^+) ∪ E_h(Ω_d), missing the Γ_d boundary term, which is generally nonzero. This breaks both the inf-sup condition and the approximation estimate, so Theorem 3.4 is not established.\n\nThere is also a step in the coercivity proof where φ(v_h) is bounded by J and a_h, and then a_h is bounded from below by something that includes φ^2—this looks circular or at least badly incomplete. The manuscript also has typos in the definitions of the linear forms and in the consistency terms that make it harder to check the argument (e.g., the term with div u in ~L and the pressure jump on Γ_I).\n\nNone of this is disqualifying in intent: the method is plausible and the proof is a standard a priori analysis with identifiable missing pieces. But the central theorem is unsupported as written. A serious referee should ask for a correct proof of Lemma 3.2 for the actual H_h, or a modification of the space/interpolation so that the Fortin argument works, plus a clean coercivity proof. I would not cite this paper until those repairs are made.\n\nMy recommendation: send it to peer review anyway. The idea is worth a second look, and a good referee can help the author fix the analysis. But it should not be accepted in current form.","headline":"Nice idea—cheaper unified Stokes–Darcy element—but the interpolation lemma for the enlarged space is unproved and the Fortin argument doesn't go through, so the convergence theorem is unsupported.","tokens_in":17310,"tokens_out":2497,"would_cite":false,"duration_ms":29214,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["74S05","74S10","74S15","74S20","74S25"],"pacs":[],"model":"deepseek-v4-flash","headline":"A single nonconforming Crouzeix-Raviart element, with normal-only continuity in the Darcy region, yields an optimally convergent stabilized mixed method for coupled Stokes-Darcy flow.","keywords":["Stokes-Darcy coupling","nonconforming finite element","Crouzeix-Raviart element","stabilized mixed method","Beavers-Joseph-Saffman condition","a priori error analysis","isotropic mesh"],"falsifier":"Compute the interpolation operator $r_h$ defined by (49)-(50) on a single isotropically refined mesh in two dimensions, using a Darcy-region velocity that has a nonzero tangential average jump across an interior edge; if $\\|r_h v\\|_h$ grows faster than a constant times $\\|v\\|_{1,d}$ as the mesh is refined, Lemma 3.2 is false and Theorem 3.4 is unsupported. Alternatively, run the Section 4 numerical test with a Darcy velocity in $H(\\mathrm{div})$ but not $H^1$ to see whether the claimed $O(h)$ rate collapses.","tokens_in":16270,"feed_emoji":"🌊","tokens_out":5432,"duration_ms":55364,"temperature":0.7,"pith_summary":"This paper develops and analyzes a unified stabilized mixed finite element method for the coupled Stokes-Darcy problem in two and three dimensions. The key move is to modify the Darcy equations by adding a divergence term and then use one variant of the nonconforming Crouzeix-Raviart element in both the free-fluid and porous regions, with a stabilization term that penalizes jumps across element edges. The author proves that the discrete problem is well-posed and that, under smoothness assumptions on the exact solution, the combined velocity-pressure error is $O(h)$ in the natural discrete norm. This matters because existing unified methods either use conforming elements in both regions or restrict the Darcy space more strongly, while the scheme here is cheaper and relies on a space that matches the natural $H(\\mathrm{div})$ regularity of Darcy flow.","feed_headline":"One nonconforming element handles both Stokes and Darcy flow","feed_subtitle":"A stabilized Crouzeix-Raviart scheme with normal-only Darcy continuity is proved to converge at optimal first order.","key_machinery":"The two devices that carry the argument are the modified Darcy formulation and the enlarged Crouzeix-Raviart space $H_h$. The Darcy modification replaces the raw momentum equation by $(\\mu K^{-1}u,v)_{\\Omega_d} + (\\mathrm{div}\\,u,\\mathrm{div}\\,v)_{\\Omega_d}$ plus pressure terms, which makes the bilinear form coercive over the relevant $H(\\mathrm{div})$ space and allows the same element in both regions. The space $H_h$ defined in (28)-(29) requires average continuity of the entire velocity across edges in the Stokes region and of only the normal component across edges in the Darcy region and on the outer Darcy boundary; a stabilization term $J(u,v)$ with $h^{-1}$ weights penalizes the remaining jumps. The convergence proof combines the discrete inf-sup condition with interpolation estimates for the four face-integral consistency errors.","core_discovery":"The central claim is Theorem 3.4: if the exact velocity is $H^2$ in each subdomain and the pressures are $H^1$, then the finite element solution satisfies $\\|u-u_h\\|_h + \\|p-p_h\\| \\leq C h (|u|_{2,s}+|u|_{2,d}+|p|_{1,s}+|p|_{1,d})$, with $C$ independent of the mesh size. The method achieves this by modifying the Darcy problem with an added $\\mathrm{div}$-$\\mathrm{div}$ term and by using a Crouzeix-Raviart space that enforces zero mean jump of the full velocity on Stokes-side edges but only zero mean jump of the normal component on Darcy-side edges, which is the natural continuity for $H(\\mathrm{div})$ velocity fields. The discrete inf-sup condition is proved via a Fortin argument, and the error analysis handles the resulting nonconformity through explicit face-integral estimates.","pith_inferences":["If the missing interpolation bound for $H_h$ can be established directly rather than quoted from a smaller space, the same scheme would be fully self-contained and likely extend to anisotropic meshes, since jump-penalty terms similar to discontinuous Galerkin methods often tolerate anisotropy.","The added $\\mathrm{div}$-$\\mathrm{div}$ term in the Darcy modification is a natural stabilization that vanishes on exactly divergence-free Darcy velocities; this suggests the method may transfer to a Brinkman-type model where the same term acts as a consistent penalty.","A test with Darcy velocity in $H(\\mathrm{div})$ but not $H^1$ would reveal whether the $O(h)$ rate depends on the stronger smoothness assumption or whether the normal-only continuity already captures the true regularity of the solution.","The face-integral estimates in Lemma 3.5 are written for isotropic meshes; porting them to anisotropic meshes would require a trace inequality with explicit dependence on element aspect ratios, which could be the next step toward a unified anisotropic analysis."],"forward_implications":["The discrete problem (31) is well-posed for any regular, interface-conforming triangulation, yielding a unique velocity-pressure pair in $H_h\\times Q_h$.","For smooth solutions the method reaches the optimal first-order convergence rate in the discrete norm, so halving the mesh size halves the combined velocity-pressure error.","Because the same nonconforming element is used in both regions and the pressure is piecewise constant, the implementation is simpler than methods that pair different stable elements in the Stokes and Darcy domains.","The stabilization term that penalizes element-edge jumps is essential to the coercivity proof, so the scheme avoids the need for separate inf-sup stable pairs in each subdomain.","The numerical experiment on a two-dimensional model problem reproduces the predicted first-order convergence rates for both velocity and pressure."],"supporting_citations":[{"why":"Provides the unified stabilized framework that this paper enlarges, including the interpolation and approximation lemmas (Lemma 3.2 and Lemma 3.4) quoted for the smaller space.","marker":"[30]"},{"why":"Introduces the idea of modifying the Darcy equations by adding a divergence term so that a single mixed formulation can be used in both regions.","marker":"[2]"},{"why":"Supplies the continuous mixed formulation and well-posedness result for the Stokes-Darcy problem that the discrete analysis starts from.","marker":"[27]"},{"why":"Gives the classical mixed method theory, the Fortin argument, and the divergence-lift result used to prove the discrete inf-sup condition.","marker":"[16]"},{"why":"Provides the Crouzeix-Raviart element and the projection/interpolation estimates (inequalities (64)-(65)) used to bound the consistency error terms.","marker":"[7]"},{"why":"Supplies the discrete Korn inequality for piecewise $H^1$ fields used in the coercivity proof of the stabilized bilinear form.","marker":"[5]"}],"fun_headline_variants":["One element, two flows: Stokes and Darcy unified","Optimal convergence for Stokes-Darcy with a single FE scheme","Modified Crouzeix-Raviart element tames Stokes-Darcy","Unified mixed method proves first-order accuracy for Stokes-Darcy","Stokes-Darcy made simple: one nonconforming element suffices"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The convergence proof assumes that the interpolation operator onto the enlarged space $H_h$ is bounded in the discrete norm and that the quoted approximation estimates hold, but these properties were proven for a smaller space in a cited paper and are not verified for the normal-continuity-only Darcy space actually used.","fun_headline_variants_meta":{"raw":{"variants":["One element, two flows: Stokes and Darcy unified","Optimal convergence for Stokes-Darcy with a single FE scheme","Modified Crouzeix-Raviart element tames Stokes-Darcy","Unified mixed method proves first-order accuracy for Stokes-Darcy","Stokes-Darcy made simple: one nonconforming element suffices"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000625,"raw_usage":{"total_tokens":2866,"prompt_tokens":890,"completion_tokens":1976,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":506,"completion_tokens_details":{"reasoning_tokens":1884}},"tokens_in":506,"tokens_out":1976,"duration_ms":14286,"temperature":1.0,"reasoning_tokens":1884,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:01:31.695244+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the interpolation operator $r_h$ defined by (49)-(50) on a single isotropically refined mesh in two dimensions, using a Darcy-region velocity that has a nonzero tangential average jump across an interior edge; if $\\|r_h v\\|_h$ grows faster than a constant times $\\|v\\|_{1,d}$ as the mesh is refined, Lemma 3.2 is false and Theorem 3.4 is unsupported. Alternatively, run the Section 4 numerical test with a Darcy velocity in $H(\\mathrm{div})$ but not $H^1$ to see whether the claimed $O(h)$ rate collapses.","supporting_citations":[{"cited_title":"Rui and R","cited_arxiv_id":null,"evidence_quote":"Provides the unified stabilized framework that this paper enlarges, including the interpolation and approximation lemmas (Lemma 3.2 and Lemma 3.4) quoted for the smaller space."},{"cited_title":"Nicaise, B","cited_arxiv_id":null,"evidence_quote":"Supplies the continuous mixed formulation and well-posedness result for the Stokes-Darcy problem that the discrete analysis starts from."},{"cited_title":"Girault and P.-A","cited_arxiv_id":null,"evidence_quote":"Gives the classical mixed method theory, the Fortin argument, and the divergence-lift result used to prove the discrete inf-sup condition."},{"cited_title":"Crouzeix and P","cited_arxiv_id":null,"evidence_quote":"Provides the Crouzeix-Raviart element and the projection/interpolation estimates (inequalities (64)-(65)) used to bound the consistency error terms."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the discrete Korn inequality for piecewise $H^1$ fields used in the coercivity proof of the stabilized bilinear form."}],"review_version":1}