{"id":"fad8de00-8aaa-4506-9047-53801516da0f","arxiv_id":"2512.18033","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A local Fortin projection for Scott-Vogelius k>=4 elements is constructed on general 2D shape-regular triangulations, including singular vertices, with divergence, trace, and stability properties.","lead":"This paper constructs a local Fortin projection for Scott-Vogelius finite elements on arbitrary two-dimensional meshes, including meshes with singular vertices. The projection preserves the divergence constraint and boundary data, a tool needed for error analysis of Stokes and non-Newtonian flow simulations.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 4's boundary singular collection case is asserted without proof; the local correction operator Π²_D may be undefined on general meshes.","rationale":"Reading the manuscript in good faith, the main construction is clear and the algebraic steps in Lemmas 5–6 and Theorem 7 are sound; the divergence preservation and trace preservation follow from the stated properties of Π_1 and Π_2. I found no internal inconsistency. The only serious gap is in Lemma 4, which is the foundation for the local correction operators Π²_D. The appendix gives a detailed construction for interior singular vertex patches but explicitly skips the boundary singular collections M^j_h with 'the latter is similar.' Since the paper's advertised novelty is handling singular vertices in general meshes, and boundary singular vertices are included in that claim, the central theorem is not fully proven for those meshes. The missing step is likely fillable—the two-triangle fan at a single boundary singular vertex is handled by a single edge bubble, and a chain should be treatable by summing such bubbles—but it is not just a notational variant of the interior case because the patch M^j_h has overlapping two-triangle fans and the mean constraints couple across the chain. The 'slight generalization' of [15, Lem. 6] in Step 1 is also stated without proof, adding a second unverified ingredient. The reader's conditional verdict is therefore appropriate; no evidence suggests the claim is false, only that the proof is incomplete as written.","tokens_in":14073,"tokens_out":10683,"duration_ms":97953,"concrete_test":"Write out the missing construction for D=M^j_h: take a maximal chain of boundary singular vertices, label the triangles in M^j_h, and check whether an explicit ψ∈˚V^k_h supported in M^j_h can be defined by summing edge bubble functions on the interior edges (as in (A.5)–(A.7)) so that ∫_T(q−div ψ)=0 for every T⊂M^j_h and ∥∇ψ∥_{L^p(M^j_h)} ≤ C∥q∥_{L^p(M^j_h)}. If no such ψ exists, then div ˚V^k_h(M^j_h) ≠ Q^{k-1}_h(M^j_h), and the dof system (3.2) is singular for boundary singular collections.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central construction defines Π²_D via the square dof system (3.2), whose well-posedness requires Lemma 4: for every D∈C_h, div maps ˚V^k_h(D) onto Q^{k-1}_h(D) with the stated L^p bound. The proof in Appendix A treats D=Ω_h(z) for interior singular vertices in detail (the four-triangle construction, (A.5)–(A.7)), but for D=M^j_h — the patch of a maximal chain of boundary singular vertices — it only says 'the latter is similar' and gives no construction. This is not a cosmetic omission: boundary singular vertices have two-triangle fans whose union M^j_h overlaps across shared boundary edges, and the edge-bubble construction must be organized along the entire chain; the mean constraints on q∈Q^{k-1}_h(M^j_h) couple across the chain in a way that differs from the interior four-triangle case. Since the abstract claims general meshes 'including singular vertices' (which includes boundary singular vertices), Theorem 7's claim for these meshes currently rests on an unproved assertion. Moreover, Step 1's 'slight generalization' of [15, Lem. 6] that supplies w with div w=q on all vertices and the L^p bound (A.1) is also stated without proof; if either step fails, Π²_D may not exist and the correction operator Π_2 is not well-defined.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs a local Fortin projection for the Scott–Vogelius pair (V^k_h, Q^{k-1}_h), k≥4, on arbitrary shape-regular 2D triangulations, including singular vertices. The construction splits the domain into a collection C_h (non-singular triangles, patches around interior singular vertices, and patches of maximal chains of boundary singular vertices) and defines a correction operator Π2 as a sum of local operators Π2_D obtained from a square dual dof system (3.2). Composing with a trace-preserving Scott–Zhang-type interpolant Π1 gives Π = Π1 + Π2(id − Π1). The paper proves Π is a projection, preserves divergence against Q^{k-1}_h, is locally stable in W^{1,p}, has approximation properties, and preserves boundary data, with a slip-boundary variant. The main analytical ingredients are a local discrete Bogovskii lemma (Lemma 4) and a local stability estimate for Π2_D.","tokens_in":14451,"tokens_out":9968,"duration_ms":98159,"significance":"If the main lemmas were fully proved, the paper would fill a genuine gap: previous local Fortin operators for Scott–Vogelius elements required meshes without singular vertices, while the present construction covers general meshes and gives explicit dependence on the local near-singularity measure Θ. The dof-based construction of Π2_D is elegant and the dimension-counting argument for well-posedness is sound. The stability and approximation results are stated with explicit constants in terms of 1/Θ and local patches, which is exactly what is needed for applications to L∞ error analysis and non-Newtonian flows. The slip-boundary variant broadens the applicability. However, the paper currently relies on unproved assertions in Lemma 4 and Lemma 13; these are not merely cosmetic because they concern the existence of the local surjectivity map underlying every local correction operator.","major_comments":[{"comment":"The proof of Lemma 4 treats only D = Ω_h(z) for an interior singular vertex and states that the boundary chain case D = M^j_h is 'similar'. This is not a minor omission. For D = M^j_h, the patch is a union of overlapping two-triangle fans along a chain of boundary singular vertices; the mean constraints ∫_T q dx couple along the chain, and the construction of ψ in (A.5)–(A.7) does not transfer directly. Since Lemma 5 and the definition of Π2_D require div ˚V^k_h(D) = Q^{k-1}_h(D) for every D ∈ C_h (2.13), Theorem 7's generality over meshes with boundary singular vertices depends on this missing proof. Please provide the chain construction or a reference containing it.","section":"Appendix A, Lemma 4 (Step 2)"},{"comment":"The first step claims that 'a slight generalization of [15, Lem. 6]' yields w ∈ ˚V^k_h with div w agreeing with q at all vertices, supp w ⊂ P(D), and the L^p bound (A.1). This generalization is not stated or proved. The published lemma does not, as written, cover arbitrary q ∈ Q^{k-1,⊥}_h(D) with support in a boundary chain patch, nor does it give the L^p stability estimate with constant 1 + 1/Θ(D). Because the remainder of Lemma 4 and hence all subsequent results reduce to this existence statement, the argument is incomplete at a load-bearing point. Please state the precise generalization and prove it or give a precise citation with the statement.","section":"Appendix A, Step 1 (A.1)"},{"comment":"In the slip-boundary variant, Lemma 13 is the analogue of [15, Lem. 6] and is used as the key input to Lemma 10. The proof splits boundary singular vertices into cases #T_h(z) = 2, 3, 1, but the case #T_h(z) = 1 is dismissed as 'similar' and omitted. A boundary vertex with a single triangle can occur at a corner of a polygonal domain, and the reflection/extension constructions used in the other cases do not immediately apply. Since Lemma 10 and Theorem 11 explicitly cover all D ∈ ˜C_h, this gap leaves the slip-boundary version of the main theorem unsupported for such vertices. Please complete the case or state an assumption excluding it.","section":"Appendix B, Lemma 13"}],"minor_comments":[{"comment":"The construction of the added triangle T4 is not fully specified: it is not clear which line is connected and whether T4 lies inside or outside Ω. Please clarify why z becomes an interior singular vertex of the enlarged triangulation.","section":"Appendix B, Lemma 13, case #T_h(z)=3"},{"comment":"The 'scaling argument' for p ∈ [1,∞] is compressed. For p = ∞ the intermediate power h_D^{2(1/p-1)} equals h_D^{-2}, which may confuse readers; writing the two cases p < ∞ and p = ∞ separately would improve readability.","section":"Lemma 5, proof of (3.5)"},{"comment":"The constant C_{P(Q(T))} is defined via Θ(P(Q(T))), but P(Q(T)) is a union of patches rather than a single sub-triangulation. Please clarify how Θ is defined for such a union, although the same definition as in (2.4) for M = ∪_{T∈M_h} T can be applied after collecting the triangles in the union.","section":"Theorem 7 (iii)"},{"comment":"On page 1 the header reads 'A LOCAL FOR TIN PROJECTION'; correct to 'Fortin'. Also, the name 'Bogovski ˘ ı' in the text before Remark 3 has a stray combining accent.","section":"Title/header"},{"comment":"For a boundary vertex with L = 1 triangle, Θ(z) is defined as a maximum over an empty set. The paper does not state whether such vertices can occur or how they are handled in Lemma 13; please add a clarifying sentence.","section":"Section 2, Definition 1"}],"recommendation":"major_revision","confidential_remarks":"To the editor: The main gap is not an error in the central construction but an unsupported extension of the authors' earlier work [15]. The reliance on a 'slight generalization' of [15, Lem. 6] and on an unproved boundary-chain case in Lemma 4 makes the current version conditional. I would not recommend rejection: the dof-counting framework is sound and the missing pieces are plausibly repairable. However, the authors should be asked to supply complete proofs or explicit references; in particular, the boundary singular chain case in Lemma 4 is nontrivial because the mean constraints couple along the chain. The slip-boundary case #T_h(z)=1 should also be completed in Lemma 13. The paper is purely theoretical; a short discussion of how the Θ-dependence enters in applications would strengthen its case."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper gives a local Fortin projection for Scott–Vogelius elements on general 2D shape-regular meshes, including singular vertices, and it preserves inhomogeneous boundary data. That is a real extension over Parker–Suli and over Tscherpel's zero-trace construction. The overall framework is clean: start from a trace-preserving Scott–Zhang interpolant, then add a local correction operator built from a domain decomposition into regular triangles, interior singular-vertex patches, and boundary chains of singular vertices. The proof of Theorem 7 is mostly rigorous; the stability and approximation estimates follow from the local Bogovskii lemma and inverse inequalities, and the projection property is argued correctly.\n\nThe key weakness is Lemma 4, the local discrete Bogovskii result. For an interior singular vertex, the appendix gives a concrete edge-bubble construction. For a boundary chain M^j_h, the proof says \"the latter is similar\" and stops. That is not a cosmetic omission. Boundary chains have overlapping vertex patches, the fans of triangles are arranged along a boundary segment, and the mean constraints on q couple across the chain. The simple four-triangle bubble construction does not obviously generalize, and the operator Π²_D is defined through a square dof system that requires this lemma. If the boundary-chain surjectivity fails, the correction operator is not well-defined. The reader marked this conditional, and I think that is the right call. The \"slight generalization\" of Guzmán–Scott Lem. 6 in Step 1 is also asserted without proof, and Lemma 13 omits one boundary case. These are likely fillable, but as written the full-generality claim rests on an undischarged assertion.\n\nThat said, the paper is not hiding anything—the omission is explicit, and the techniques available (reflection across the boundary, as used in Lemma 13) probably work. The interior singular-vertex case and the no-singular-vertex case are proven, and the structure of the argument is sound. This deserves a serious referee, but the referee should be told to press the authors for a complete proof of Lemma 4 on boundary chains, or to state the theorem with the boundary-chain condition explicitly excluded. If the gap is filled, this is a solid contribution to the finite-element toolbox.","headline":"A genuinely useful extension of local Fortin projections to meshes with singular vertices, but the proof's load-bearing boundary-chain case is asserted with \"similar\", which is a real gap that needs filling before the full-generality claim is solid.","tokens_in":14885,"tokens_out":5422,"would_cite":true,"duration_ms":55983,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65N30","65N12","76D07"],"pacs":[],"model":"deepseek-v4-flash","headline":"For k ≥ 4, the Scott–Vogelius finite element pair admits a local Fortin projection on any shape-regular 2D triangulation, including those with singular vertices.","keywords":["Scott–Vogelius elements","Fortin projection","local projection","singular vertices","divergence-free","finite element analysis","Stokes problem","shape-regular triangulation"],"falsifier":"Assemble the linear system (3.2) on the smallest boundary-singular patch, a single triangle at a boundary vertex (case #T_h(z) = 1), for k = 4 and a pressure q with nonzero triangle mean. If the system matrix is singular, or if the least-squares constant for ∥∇v∥_{L^p(P(D))} ≤ C(1 + 1/Θ(D))∥q∥_{L^p(D)} degenerates faster than 1/Θ(D) as Θ → 0, then Lemma 4 fails and Theorem 7 collapses. A run of meshes refining toward the singular vertex would show this in practice.","tokens_in":13942,"feed_emoji":"📐","tokens_out":9029,"duration_ms":87479,"temperature":0.7,"pith_summary":"This paper proves that the divergence-free Scott–Vogelius finite element spaces, in two dimensions and for polynomial degree k ≥ 4, have a local Fortin projection on arbitrary shape-regular triangulations, including meshes with singular vertices. The projection is built by correcting a Scott–Zhang-style interpolant with a correction operator obtained from local solves on a carefully chosen decomposition of the domain into patches. If the claim is correct, error analyses for Stokes and Navier–Stokes that need a local, trace-preserving Fortin operator now apply to general meshes, not merely to meshes without singular vertices. The construction also adapts to slip (no-penetration) boundary conditions.","feed_headline":"Divergence-free velocity spaces get a local projection on any 2D mesh","feed_subtitle":"Works even at singular vertices, enabling improved Stokes error analysis under inhomogeneous boundary conditions.","key_machinery":"The load-bearing mechanism is the local discrete Bogovskii operator supplied by Lemma 4: for every patch D in the decomposition and every pressure q supported in D, there exists a zero-trace Scott–Vogelius velocity v with div v = q, support in the enlarged patch P(D), and bound ∥∇v∥_{L^p(P(D))} ≤ C(1 + 1/Θ(D))∥q∥_{L^p(D)}. This surjectivity makes the correction operator Π₂^D well-defined: it is the unique solution of a square degree-of-freedom system that enforces both (3.2a) divergence preservation against the local pressure space and (3.2b) orthogonality to the local divergence-free subspace. Summing these local corrections over the decomposition C_h produces Π₂, and composing with the qua","core_discovery":"The central claim is Theorem 7: for k ≥ 4, the operator Π = Π₁ + Π₂(id − Π₁) is a linear projection from W^{1,1}(Ω) onto V_h^k that (i) preserves divergence against every pressure test in Q_h^{k-1}, (ii) satisfies local L^p stability, (iii) has order-h^r approximation, and (iv) preserves the boundary trace of any input that already equals a discrete velocity on ∂Ω. The construction partitions Ω into a decomposition C_h made up of triangles with only non-singular vertices, patches around interior singular vertices, and chains of boundary singular vertices. On each patch D, a local correction operator Π₂^D is defined as the solution of a square system (3.2) that asks for divergence preservatio","pith_inferences":["The structural template — a local surjectivity lemma plus a square correction system — looks transferable to other exact-divergence pairs, so the paper may serve as a recipe for building local Fortin projections whenever such a local Bogovskii result exists.","The stability constant involves 1/Θ(D), which can become large when a vertex is nearly singular; a natural follow-up is to ask whether this dependence is sharp, and whether local refinement near such vertices can keep the constant under control.","Because the projection preserves traces only for inputs that match a discrete function on the boundary, combining it with Nitsche-type treatments of boundary conditions seems a direct next step for inhomogeneous or dynamic boundary data.","A concrete computational check would be to assemble the correction system (3.2) on the minimal singular boundary patch (a single triangle at a boundary vertex) for k = 4 and verify the stability bound of Lemma 5 for a family of triangles shrinking toward degeneracy."],"forward_implications":["Divergence-preserving Fortin operators now exist for Scott–Vogelius elements on arbitrary shape-regular two-dimensional triangulations, including meshes with singular vertices, for polynomial degree k ≥ 4.","The operator preserves inhomogeneous boundary data, so analyses of Stokes and Navier–Stokes with non-homogeneous Dirichlet conditions no longer need to avoid singular vertices or split the mesh artificially.","Local stability and approximation properties in L^p for all p ∈ [1,∞] follow, which are the exact ingredients needed for max-norm and quasi-norm error estimates for incompressible flow.","The slip-boundary variant (Theorem 11) provides an analogous local Fortin projection for no-penetration conditions, extending the reach to free-surface and friction-type boundary models.","On Clough-Tocher and Powell-Sabin split triangulations the degree restriction drops to k ≥ 2 and k ≥ 1 respectively, making the construction available at lower order on those macro-element meshes."],"fun_headline_variants":["Local Fortin projection works on any 2D triangulation, even with singular vertices","New local projection for Scott-Vogelius handles singular vertices on 2D meshes","Divergence-preserving projection exists for Scott-Vogelius on all 2D meshes","A local Fortin projection for Scott-Vogelius on arbitrary 2D meshes","Scott-Vogelius k≥4 gets local projection on meshes with singular vertices"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The construction rests on Lemma 4, which asserts that on every patch in the decomposition — including patches touching boundary singular vertices — the divergence operator from the zero-trace local Scott–Vogelius space maps onto the local pressure space with stability constant C(1 + 1/Θ(D)); if this local surjectivity fails at some boundary singular patch, the correction operator is not well-defined and the projection as constructed does not exist.","fun_headline_variants_meta":{"raw":{"variants":["Local Fortin projection works on any 2D triangulation, even with singular vertices","New local projection for Scott-Vogelius handles singular vertices on 2D meshes","Divergence-preserving projection exists for Scott-Vogelius on all 2D meshes","A local Fortin projection for Scott-Vogelius on arbitrary 2D meshes","Scott-Vogelius k≥4 gets local projection on meshes with singular vertices"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000795,"raw_usage":{"total_tokens":3269,"prompt_tokens":605,"completion_tokens":2664,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":349,"completion_tokens_details":{"reasoning_tokens":2552}},"tokens_in":349,"tokens_out":2664,"duration_ms":19786,"temperature":1.0,"reasoning_tokens":2552,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T15:06:02.150373+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Assemble the linear system (3.2) on the smallest boundary-singular patch, a single triangle at a boundary vertex (case #T_h(z) = 1), for k = 4 and a pressure q with nonzero triangle mean. If the system matrix is singular, or if the least-squares constant for ∥∇v∥_{L^p(P(D))} ≤ C(1 + 1/Θ(D))∥q∥_{L^p(D)} degenerates faster than 1/Θ(D) as Θ → 0, then Lemma 4 fails and Theorem 7 collapses. A run of meshes refining toward the singular vertex would show this in practice.","supporting_citations":[],"review_version":1}