{"id":"1bdebb47-0658-482f-9622-05a4d67f29ef","arxiv_id":"1908.08013","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"For sufficiently fine triangulations and small bulk parameter, the adaptive Morley finite element method for the von Kármán equations is proved to converge with optimal rates in the number of degrees of freedom.","lead":"This paper proves that an adaptive nonconforming Morley finite element method for the von Kármán plate equations achieves optimal convergence rates once the mesh is sufficiently fine and the Dörfler bulk parameter is small. The result is the first rate-optimal adaptive scheme for this semilinear fourth-order plate model and gives a theoretical basis for using adaptive Morley elements on nonconvex domains.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 4.3 is internally coherent, but the load-bearing risk is the unquantified and possibly impractical δ̄ required for unique discrete solvability, branch selection, and exact solves, as the paper itself flags in Remarks 4.1-4.2.","rationale":"The paper's central claim is a conditional existence theorem, and I read Section 5 as a good-faith proof of the four axioms. I spot-checked the key algebraic steps: the Fréchet-derivative identity (5.8) matches when the discrete equations are subtracted, the smallness condition (5.12) is satisfiable for δ sufficiently small, and the passage from (A4)_ε to (A4) closes without circularity, with δ_4 depending on θ through ε < (1−ρ_12)/Λ_12 but consistent for all θ ≤ θ̄. The novel piecewise H^1 a priori estimate (Theorem 3.1.c) is supported by a coherent duality argument; the proof leans on published results ([21], [23, Thm 6.19], [22, Lem 3.9b]) that I did not re-derive. The load-bearing weakness is not internal inconsistency but the gap between the theorem's hypotheses and any computable instance: δ_0, δ_1, δ_3, δ_4 are existential, the constants C_qo, Λ_3, Λ_12 entering (5.12) are unestimated, and δ̄ is polynomially forced toward zero as θ approaches θ_0, so the optimal-rate regime may lie below mesh sizes any implementation can reach; Remarks 4.1, 4.2, and 4.5 concede exactly this, and Remark 4.2 conjectures the smallness is essential. The exact-solve and branch-selection requirements are likewise acknowledged at Remark 4.1. Because I found no falsifying flaw, the conditional verdict stands; the concern should be recorded as a call for quantitative smallness estimates and a branch-aware implementation, not as grounds for rejection.","tokens_in":38010,"tokens_out":26609,"duration_ms":226674,"concrete_test":"Manufactured-solution benchmark on the L-shaped domain: choose a regular solution (u,v) of known closed form satisfying the clamped boundary conditions, set f := Δ²u − [u,v] and g := Δ²v + ½[u,u], and implement AMFEM exactly as in §4.2, initializing Newton's method on each T_ℓ from the Morley interpolation of the exact solution as the branch selector. Measure the smallest δ* for which Newton converges on every mesh to the branch with |||Ψ−Ψ_ℓ|||pw ≤ ε_0, and compare the empirical left-hand side of (4.4) with the optimal rate s = 1/2 (per degree of freedom) forced by the known exact solution at the nonconvex corner. If δ* ≥ 10⁻²·diam(Ω) and the rates track (4.4) over many iterations, the concern does not land; if branch selection fails or δ* ≤ 10⁻⁵·diam(Ω) is required, the hypotheses of Theorem 4.3 are not reachable in practice and Remark 4.2's conjecture is confirmed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"I verified the internal logic of the central argument and found no contradiction: the expansion (5.8) in the proof of Theorem 5.4 is correct algebra, the smallness selection (5.12) is satisfiable, and the passage from the perturbed quasiorthogonality (A4)_ε in Theorem 5.8 to (A4) in Corollary 5.9 closes without circularity. The load-bearing risk is therefore not a proof gap but the reachability of the theorem's hypotheses. Theorem 4.3 requires that for every triangulation T_ℓ in the adaptive loop, the discrete problem (2.6) has a unique solution within the ε_0-ball around Ψ and that AMFEM's solve step computes it exactly, selecting the intended branch (Remark 4.1). Theorem 3.1(a) guarantees existence only for mesh size ≤ δ_0, with δ_0 existential; moreover the proof of Theorem 5.8 shows the usable threshold must satisfy (5.12) with ε < (1−ρ_12)/Λ_12 (Corollary 5.9), so δ̄ ≤ C(1−ρ_12(θ̄))^{3/2} with constants C_qo, Λ_3, Λ_12 that are neither computed nor estimated. Since ρ_12 → 1 as θ → θ_0 := 1/(1+Λ_1²Λ_3), δ̄ shrinks polynomially as the marking parameter approaches the admissible bound; near a bifurcation point the paper itself (Remark 4.5) concedes δ may be forced very small, and Remark 4.2 conjectures the smallness is not a technical artefact. The theorem is true as an existence statement, but the advertised practical conclusion (method of choice for nonconvex domains) is not backed by any computable bound on δ̄, and if the Remark 4.2 conjecture is correct, optimal rates are attained only at mesh scales that uniform refinement cannot reach.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes an adaptive Morley finite element method for the von Kármán equations. It defines a discrete nonlinear problem, a residual-based error estimator, and an adaptive loop AMFEM with an initial uniform refinement phase. The main theorem (Theorem 4.3) asserts that, for any fixed regular solution and for sufficiently small mesh-size parameter δ and Dörfler parameter θ, the adaptive algorithm produces estimator sequences satisfying the optimal-rate equivalence (4.4). The proof verifies the axioms of adaptivity (A1)-(A4), introducing a new piecewise H1 a priori error estimate, a reduction argument for volume residuals, discrete reliability via a conforming companion, and quasiorthogonality via a perturbed variant (A4)_ε. The paper explicitly acknowledges that exact solves are assumed and that the smallness thresholds are existential.","tokens_in":38370,"tokens_out":13646,"duration_ms":125338,"significance":"If the central theorem holds, this is the first proof of rate-optimal adaptive convergence for the von Kármán equations, and it extends the axiomatic adaptivity framework to a nonconforming semilinear fourth-order problem. The paper contributes a genuinely new piecewise H1 a priori error estimate and carefully treats the trilinear nonlinearity. A notable strength is its transparency: Remarks 4.1, 4.2, and 4.5 explicitly identify the exact-solve idealization, the conjecture that the small-δ requirement may not be a technical artefact, and the possibly large pre-asymptotic range. These caveats limit the practical implications but do not, by themselves, undermine the mathematical claim as an existence theorem.","major_comments":[{"comment":"The proof of the piecewise H1 estimate for general ℓ∈N is not supplied: the final paragraph states 'further details are omitted' after invoking a higher-order companion operator from Remark 2.3, which is in turn deferred to reference [18]. Since Theorem 3.1(c) is stated for every ℓ∈N0 and Lemma 5.7 invokes it for arbitrary ℓ, this omission affects the axiomatic verification. Please either provide the complete argument or restrict the statement and all subsequent uses to the case ℓ=0, which is what the quasiorthogonality proof actually requires.","section":"Section 3, proof of Theorem 3.1(c)"},{"comment":"The proof selects δ with δ≤min{δ0,δ1} and satisfying (5.12), but it immediately uses discrete reliability (A3), which by Theorem 5.4 is guaranteed only for triangulations in T(δ3). The maximization should additionally impose δ≤δ3; the statement of Theorem 5.8 already promises δ≤δ3, so this is a small but real gap in the written proof rather than a substantive obstruction.","section":"Section 5.5, proof of Theorem 5.8"},{"comment":"The proof of Corollary 5.9 is condensed to a black-box invocation of [24, Thm. 4.1] for an unstated contraction property (A12) and [24, Thm. 3.1] to pass from (A4)_ε to (A4), with no statement of the hypotheses or of the dependence of ρ12 and Λ12 on θ. Because this is the final step connecting the perturbed quasiorthogonality to the optimal-rate theorem, the assumptions used from [24] should be restated precisely, or a self-contained argument should be provided for the present setting.","section":"Section 5.5, Corollary 5.9 and Theorem 4.3"}],"minor_comments":[{"comment":"The phrase 'method of choice for a nonconvex domain' is stronger than what the analysis supports: δ̄ and θ̄ are existential, exact solves are assumed, and Remark 4.2 conjectures that the smallness of δ is not a technical artefact. I recommend softening this to describe the result as an existence theorem with optimal rates under unquantified smallness and exact solves.","section":"Abstract and Section 1.3"},{"comment":"The final paragraph claims that the equivalence constants in (4.4) are independent of δ and θ, while Theorem 4.2 states that they depend on δ, θ, and T0. The argument that a closer inspection of [14,24] removes this dependence is plausible but not shown; please add the relevant details or explicitly state the weaker dependence that is actually proved.","section":"Section 5.6, proof of Theorem 4.3"},{"comment":"The notation T(δ) for triangulations with small mesh-size and T(N) for triangulations with a bounded number of triangles is overloaded and can be confusing, especially in Theorem 4.2 and Section 4.3. A brief notational table or a change of notation for the cardinality-bounded family would improve readability.","section":"Section 1.5 and Section 2.2"}],"recommendation":"major_revision","confidential_remarks":"The paper appears to be mathematically sound in its central architecture, and I found no circularity: the earlier results [21,24,35] predate this work and are cited appropriately. The main concerns are completeness of a stated proof (Theorem 3.1(c) for ℓ>0), a small parameter mismatch in the proof of Theorem 5.8, and the black-box use of [24] in Corollary 5.9. These are fixable within the scope of the manuscript. I would also encourage the editor to ensure that the practical claims in the abstract and introduction are reconciled with the paper's own caveats about unquantified thresholds and exact solves."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Zuzana, quick take on arXiv:1908.08013. This is a genuine contribution: Carstensen and Nataraj verify the four axioms of adaptivity (stability, reduction, discrete reliability, quasiorthogonality) for the nonconforming Morley FEM applied to the von Kármán equations, and thereby obtain optimal convergence rates. The main new technical piece is the piecewise H1 a priori estimate (Thm 3.1c) that lets the quasiorthogonality close despite the trilinear nonlinearity. The paper is well written and unusually honest about its own limitations.\n\nThe central theorem, Thm 4.3, says that for a regular solution and an initial triangulation there exist δ̄ and θ̄ such that AMFEM with any 0<δ≤δ̄, 0<θ≤θ̄ achieves the optimal-rate equivalence. I checked the logic of the proofs in Section 5, especially the perturbed quasiorthogonality (A4)_ε in Thm 5.8 and the passage to (A4) in Cor 5.9. The expansion (5.8) and the smallness selection (5.12) check out; there is no hidden circularity. The cited axioms framework from [14,24] and the a priori/posteriori estimates from [21,35] are used correctly.\n\nThe soft spots are exactly the ones the authors admit. The thresholds δ̄, θ̄ are existential; the proof gives no computable bound. Remark 4.1 says exact solve is assumed; Remark 4.2 conjectures the smallness requirement may be genuine, not a technical artefact. Near a bifurcation, δ may need to be very small (Rem 4.5). So the practical claim—\"method of choice\" for nonconvex domains—is not backed by any computable guarantee. But those are conditions for the theorem, not flaws in its internal logic. The result stands as an existence statement: for all sufficiently fine meshes and small enough Dörfler parameter, you get optimal rates.\n\nWho is this for? People doing adaptive FEM for nonlinear fourth-order problems, and anyone wanting to apply the axioms framework to a non-monotone semilinear problem. It deserves a serious referee. I would send it out, with the referee asked to verify the unquantified constants and to pressure the authors to at least estimate δ̄, even crudely, or to soften the \"method of choice\" language.","headline":"Solid, careful axioms-of-adaptivity paper for Morley FEM on the von Kármán equations; the optimal-rate result is real but sits behind unquantified smallness thresholds and an exact-solve assumption that the authors themselves flag.","tokens_in":38963,"tokens_out":1880,"would_cite":true,"duration_ms":18316,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65N30","65N12","65N50"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that the adaptive Morley finite element method, run after a short uniform-refinement phase, achieves optimal convergence rates for regular solutions of the von Kármán plate equations in the plane.","keywords":["von Kármán equations","adaptive finite element method","Morley finite element","nonconforming finite element method","optimal convergence rates","a posteriori error estimate","axioms of adaptivity","discrete reliability"],"falsifier":"The theorem would be falsified by exhibiting a regular solution $\\Psi$ and arbitrarily fine admissible triangulations $T$ on which the discrete problem (2.6) has two distinct solutions within the piecewise energy ball $|||\\Psi-\\cdot|||_{\\mathrm{pw}}\\le \\varepsilon_0$, since Theorem 3.1(a) asserts uniqueness for all $T\\in\\mathcal T(\\delta_0)$.","tokens_in":37739,"feed_emoji":"📐","tokens_out":10247,"duration_ms":86826,"temperature":0.7,"pith_summary":"The paper proves that the adaptive Morley finite element method, a nonconforming scheme built from quadratic polynomials, approximates a regular solution of the von Kármán plate equations at optimal adaptive convergence rates. The algorithm first refines uniformly until the mesh is fine enough that the nonlinear discrete problem has a unique solution near the exact solution, then runs the standard solve-estimate-mark-refine loop with bulk marking. The advertised result is Theorem 4.3: for every rate $s>0$, the estimator $\\eta_\\ell$ on the adaptively generated meshes decays as fast as the best possible estimator over all admissible refinements with the same number of triangles, up to constants independent of the smallness parameters. The paper thereby provides the first rate-optimal adaptive scheme for the von Kármán equations, and it removes the barrier that reduced elliptic regularity on nonconvex domains imposes on uniform meshes.","feed_headline":"First adaptive plate solver with proven optimal rates","feed_subtitle":"A nonconforming Morley scheme now provably matches near-best adaptive convergence for von Kármán plate equations.","key_machinery":"The mechanism that carries the argument is the abstract axioms-of-adaptivity framework, instantiated for the nonconforming Morley method. The central object is the Morley finite element space $M(T)$: piecewise quadratic functions continuous at triangle vertices and with continuous normal derivatives at edge midpoints, vanishing at boundary degrees of freedom. The key technical tools are a companion operator that maps Morley functions to conforming $H^2_0$ functions and controls the nonconformity gap; a discrete inf-sup condition at the regular solution, inherited from the continuous problem for sufficiently fine meshes; a new piecewise $H^1$ a priori error estimate for $\\Psi-\\Psi_M$; and a reduction property for the volume part of the residual estimator, which is not a higher-order term in this semilinear problem. The four axioms --- (A1) stability, (A2) reduction, (A3) discrete reliability, and (A4) quasiorthogonality --- are then verified for the explicit residual-based estimator $\\eta$, and the abstract theorem converts those axioms into the optimal-rate equivalence.","core_discovery":"The central claim, Theorem 4.3, states that a regular solution $\\Psi$ of the von Kármán equations can be approximated by the adaptive Morley FEM with optimal convergence rates. More precisely, given an initial triangulation $T_{\\mathrm{init}}$, there exist positive thresholds $\\bar\\delta$ and $\\bar\\theta$ such that, whenever the pre-asymptotic uniform refinement reaches a triangulation $T_0$ with maximal mesh-size $\\le\\bar\\delta$ and the bulk parameter $\\theta$ satisfies $0<\\theta\\le\\bar\\theta$, the output satisfies $$\\sup_{\\ell\\in\\mathbb N_0}(1+|T_\\ell|-|T_0|)^s\\,\\eta_\\ell \\approx \\sup_{n\\in\\mathbb N_0}(1+n)^s \\min_{T\\in\\mathcal T(T_0,n)}\\eta(T)$$ for every $s>0$, with equivalence constants that depend on $\\Psi$, $T_{\\mathrm{init}}$, $\\bar\\delta$, $\\bar\\theta$, and $s$ but not on the particular choice of $\\delta,\\theta$ below the thresholds. The paper establishes this by verifying the four axioms of adaptivity --- stability, reduction, discrete reliability, and quasiorthogonality --- for an explicit residual-based error estimator on the nonconforming Morley spaces.","pith_inferences":["If the theorem's threshold $\\bar\\delta$ is extremely small in practice, as the paper suspects may happen near bifurcation points, the pre-asymptotic uniform-refinement phase could dominate the computation; a practical route would be to use continuation or a good initial guess from a coarser level to enter the uniqueness regime earlier.","The piecewise $H^1$ a priori estimate and the volume-residual reduction may extend to other nonconforming or discontinuous Galerkin discretizations of fourth-order semilinear problems, since the arguments are formulated without relying on a Rayleigh-Ritz structure.","A natural testable consequence is that on convex domains, where the solution is smoother, the adaptive loop should never need to refine far from singularities; one could run AMFEM on a square and check that the marked set stays confined to regions where the estimator is large.","The optimal-rate equivalence for all $s>0$ suggests the algorithm is robust with respect to the marking parameter, but the dependence of the equivalence constants on $\\bar\\theta$ and $\\bar\\delta$ is not quantified; computing those constants for a model problem would tell how small the parameters must be in practice."],"forward_implications":["On nonconvex polygonal domains, where uniform refinement is limited by the elliptic regularity index $\\gamma<1$, the adaptive Morley FEM recovers near-best convergence rates in the estimator.","The equivalence (4.4) holds for every positive $s$, so the adaptive algorithm is not limited to a fixed algebraic rate; it adapts to the best possible approximation class.","The efficiency part of the a posteriori estimate (Theorem 4.1) converts the estimator-rate equivalence into rate optimality for the total error $|||\\Psi-\\Psi_M|||_{\\mathrm{pw}} + \\mathrm{osc}_0(f,T)$ with respect to nonlinear approximation classes.","Choices of $\\delta$ and $\\theta$ below the thresholds do not change the asymptotic rate, although smaller $\\delta$ lengthens the pre-asymptotic uniform-refinement phase.","The results assume exact solves of the discrete nonlinear problem; an inexact solver would need additional perturbation analysis to preserve the rates."],"supporting_citations":[{"why":"Supplies the abstract axioms-of-adaptivity theorem that converts (A1)-(A4) into the optimal-rate equivalence (4.4).","marker":"[14]"},{"why":"Provides the version of the axioms with combined marking used to derive quasiorthogonality (A4) from the perturbed form.","marker":"[24]"},{"why":"Establishes the discrete inf-sup condition, a priori energy estimates, and the explicit residual estimator for the von Kármán equations that this paper reuses.","marker":"[21]"},{"why":"Supplies the Morley nonconforming discretization and the energy-norm a priori error analysis for the von Kármán equations.","marker":"[35]"},{"why":"Gives the conforming companion construction and the discrete reliability technique for nonconforming FEMs used in the proof of (A3).","marker":"[23]"},{"why":"Provides the companion operator and Morley interpolation estimates, including the discrete Helmholtz decomposition used throughout the analysis.","marker":"[28]"},{"why":"Supplies Morley interpolation and the discrete Helmholtz decomposition with Morley functions, used for the reduction and reliability arguments.","marker":"[16]"},{"why":"Gives the complexity result for newest-vertex-bisection refinement that bounds the number of triangles added per marked triangle.","marker":"[4]"}],"fun_headline_variants":["Adaptive Morley FEM proven optimal for von Kármán equations","Optimal adaptive rates proven for von Kármán plate solver","Morley FEM attains optimal convergence in adaptive setting","First proof: nonconforming adaptive FEM optimal for plates"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof requires that a preliminary uniform-refinement phase reaches a mesh whose largest element is smaller than an unquantified threshold, and that on every such mesh the nonlinear discrete equations have a unique solution near the exact solution; the paper also assumes these discrete equations are solved exactly.","fun_headline_variants_meta":{"raw":{"variants":["Adaptive Morley FEM proven optimal for von Kármán equations","Optimal adaptive rates proven for von Kármán plate solver","Morley FEM attains optimal convergence in adaptive setting","First proof: nonconforming adaptive FEM optimal for plates"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00022,"raw_usage":{"total_tokens":1457,"prompt_tokens":968,"completion_tokens":489,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":584,"completion_tokens_details":{"reasoning_tokens":418}},"tokens_in":584,"tokens_out":489,"duration_ms":17137,"temperature":1.0,"reasoning_tokens":418,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:51:45.750478+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"The theorem would be falsified by exhibiting a regular solution $\\Psi$ and arbitrarily fine admissible triangulations $T$ on which the discrete problem (2.6) has two distinct solutions within the piecewise energy ball $|||\\Psi-\\cdot|||_{\\mathrm{pw}}\\le \\varepsilon_0$, since Theorem 3.1(a) asserts uniqueness for all $T\\in\\mathcal T(\\delta_0)$.","supporting_citations":[{"cited_title":"Carstensen, M","cited_arxiv_id":null,"evidence_quote":"Supplies the abstract axioms-of-adaptivity theorem that converts (A1)-(A4) into the optimal-rate equivalence (4.4)."},{"cited_title":"Carstensen and H","cited_arxiv_id":null,"evidence_quote":"Provides the version of the axioms with combined marking used to derive quasiorthogonality (A4) from the perturbed form."},{"cited_title":"Nonconforming Finite Element Discretisation for Semilinear Problems with Trilinear Nonlinearity","cited_arxiv_id":"1708.07627","evidence_quote":"Establishes the discrete inf-sup condition, a priori energy estimates, and the explicit residual estimator for the von Kármán equations that this paper reuses."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Morley nonconforming discretization and the energy-norm a priori error analysis for the von Kármán equations."},{"cited_title":"Carstensen and S","cited_arxiv_id":null,"evidence_quote":"Gives the conforming companion construction and the discrete reliability technique for nonconforming FEMs used in the proof of (A3)."},{"cited_title":"Gallistl, Morley finite element method for the eigenvalues of the bihar monic operator , IMA J","cited_arxiv_id":null,"evidence_quote":"Provides the companion operator and Morley interpolation estimates, including the discrete Helmholtz decomposition used throughout the analysis."},{"cited_title":"Carstensen, D","cited_arxiv_id":null,"evidence_quote":"Supplies Morley interpolation and the discrete Helmholtz decomposition with Morley functions, used for the reduction and reliability arguments."},{"cited_title":"Binev, W","cited_arxiv_id":null,"evidence_quote":"Gives the complexity result for newest-vertex-bisection refinement that bounds the number of triangles added per marked triangle."}],"review_version":1}