{"id":"ad8d3bc3-c295-4909-8d84-64ff16ad87cd","arxiv_id":"2502.03847","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Optimal-order error bounds of order h^2 plus tau^q are proven for the fully discrete bulk-surface Cahn-Hilliard system with dynamic boundary conditions for BDF methods q=1,...,5, using an almost mass conservation argument.","lead":"This paper proves optimal-order error estimates for a finite element and linearly implicit backward difference discretization of the bulk-surface Cahn-Hilliard system with dynamic boundary conditions, for time-stepping orders one through five. It also extends the proof technique to the Cahn-Hilliard equation on evolving surfaces.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The stability proof tests the discrete error equation with a nonlinear residual that is not a finite element function; Eq. (34) is inadmissible and Theorem 5.1 is not established as written.","rationale":"The central claim of the paper, Theorem 5.1, rests on the stability estimate in Proposition 8.3. The proof of that proposition contains an inadmissible test-function step in Part C: the discrete Galerkin equation (20) is tested with the nonlinear residual (r^k_Ωh, r^k_Γh), which is not an element of the discrete finite element space V_h × S_h. For the double-well potential explicitly used in the numerical experiments, F'(u) = u^3 − u, the composition of F' with a piecewise-linear function is a cubic piecewise polynomial, not piecewise linear, so it is not in V_h. The equality displayed in (34) therefore does not follow from (20a). The same issue affects the norm appearing in the following estimate, since the discrete V-norm is defined only on V_h × S_h. This is an internal inconsistency in the proof, not a disagreement with the numerical-analysis literature or a missing external reference. The reader's identified concern about Lemma 6.1 is legitimate but does not capture the most immediate obstruction: even a fully proved Poincaré–Wirtinger inequality would not repair Part C, because the invalid test-function step occurs before Part D and feeds directly into the Grönwall argument in Part E. The paper provides no machine-checked proof or reproducible code that could independently certify Theorem 5.1, and the numerical experiments, while supportive of BDF2 convergence, do not test the higher-order cases or the special parameter regimes in Assumption (A) where the proof is most delicate. Therefore, as written, the central claim is not established, and the appropriate verdict is reject pending a corrected stability argument.","tokens_in":35180,"tokens_out":11695,"duration_ms":116178,"concrete_test":"In a 1D analogue, take Ω = (0,1), V_h the P1 space on a two-element mesh, F'(u) = u^3 − u, and choose discrete nodal vectors for ~u_h and ~u_* so that r = F'(~u_h) − F'(~u_*) is a cubic function not in V_h. Recompute the equality in (34) using (20a) tested with r; it will fail unless r is replaced by its L2 projection P_h r. Then bound the missing term m_h(r − P_h r, ∂τ e) by the defect/starting-value data; if this produces an uncontrolled O(h) or O(τ^{q−1}) term, the stability estimate (45) collapses. A successful resolution would re-derive Part C with P_h r (or a Clément interpolation) and show the extra consistency error is O(h^2 + τ^q) in the norms of Proposition 9.5.","verdict_should_be":"REJECT","load_bearing_attack":"Proposition 8.3 is the load-bearing stability estimate for Theorem 5.1. In Part C, the term (III) in (31) contains m_h((r^k_Ωh − η r^{k−1}_{Ωh}, r^k_Γh − η r^{k−1}_{Γh}), (∂τ_q e^k_u, ∂τ_q e^k_ψ)). To eliminate the discrete time derivative, the proof states in (34): 'Since L ≠ 0 we can test (20a) with (r^k_Ωh, r^k_Γh)'. But (20a) is a Galerkin equation valid only for test functions in V_{L,β}^h = V_h × S_h, and r^k_Ωh = F'_Ω(~u^k_h) − F'_Ω(~u^k_*) is the pointwise difference of the nonlinearity evaluated at piecewise-linear functions. Unless F'_Ω is affine, r^k_Ωh is not piecewise linear and does not lie in V_h; similarly for the surface. Hence the displayed equality in (34) is inadmissible, and the subsequent bound using the norm ‖(r^k_Ωh, r^k_Γh)‖_{V_{K,α}^h} is not a discrete norm as defined in (8). The continuous perturbation argument in Section 7 tests with r in the continuous space and does not transfer to the discrete Galerkin equation. To repair this one would need to insert an interpolation or projection P_h r and control an additional consistency term such as m_h(r − P_h r, ∂τ e); no such estimate appears. Therefore the stability estimate (45), and with it Theorem 5.1, does not follow from the written proof. The imported bulk–surface Poincaré–Wirtinger lemma (Lemma 6.1) is a secondary concern; even if it were fully proved, Part C remains invalid.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims optimal-order fully discrete error estimates for the bulk--surface Cahn--Hilliard system with dynamic boundary conditions, using linear bulk--surface finite elements in space and linearly implicit backward difference formulae of order 1 to 5 in time. The central result, Theorem 5.1, states that the error in the phase fields and chemical potentials is bounded by C(h^2 + tau^q) under a mild step-size restriction, for the parameter range K in [0, infinity) and L in (0, infinity), with extensions to K in (0, infinity), L = infinity and K = L = 0. The proof introduces a stability framework based on an 'almost mass conservation' of the error equations, combined with a bulk--surface Poincare--Wirtinger inequality, in place of the anti-symmetric structure used in prior work [13]. The manuscript also sketches an extension to the Cahn--Hilliard equation on evolving surfaces and presents numerical experiments.","tokens_in":35558,"tokens_out":7474,"duration_ms":72817,"significance":"If the proof were correct, this would be a significant contribution: it would supply the first fully discrete optimal-order error estimates for the general bulk--surface Cahn--Hilliard system with dynamic boundary conditions for BDF methods up to order 5. The almost-mass-conservation technique is a genuinely novel stability tool that plausibly transfers to other mass-conserving phase-field problems. The paper also includes instructive numerical experiments. However, the main stability proof contains a serious technical gap that prevents Theorem 5.1 from being established as written.","major_comments":[{"comment":"The displayed equality in (34) is inadmissible. The error equation (20a) is a Galerkin equation that holds only for test functions in V_{L,beta}^h = V_h × S_h. The function r^k_Omega h = F'_Omega(~u^k_h) − F'_Omega(~u^k_*) is the pointwise difference of the nonlinearity evaluated at piecewise-linear functions. For a nonlinear potential such as the double-well potential, this function is not piecewise linear and does not belong to V_h; similarly for r^k_Gamma h. Hence one cannot test (20a) with (r^k_Omega h, r^k_Gamma h) as stated. Consequently the equality in (34) is not justified, and the norm ||(r^k_Omega h, r^k_Gamma h)||_{V_{K,alpha}^h} is not a well-defined discrete norm for this argument. This step is the essential device used to bound term (III) and to proceed to (37) and the stability estimate (45). Without an additional argument, for example testing with an interpolation of r into V_h and bounding the resulting consistency error, Proposition 8.3 and Theorem 5.1 are not established.","section":"§8.4, Eq. (34)"},{"comment":"Lemma 6.1 is a load-bearing ingredient in Part D of the stability proof (Eqs. (38)–(40) and (43)), but its proof is delegated to a reference with the sentence 'The original proof in [45, Lemma A.1] suffices here up to the mean zero transformations.' The lemma is stated for the combined spaces V_{J,lambda} with the a_{J,lambda}-seminorm and a generalized mean M_{lambda,upsilon}; the needed discrete analogue, applied after geometric transfer, is asserted without a detailed derivation. The author should provide a complete proof of Lemma 6.1 and explicitly justify the h-independent discrete version used in (40) and (43).","section":"§6, Lemma 6.1"},{"comment":"The abstract states that the paper 'proves optimal-order fully discrete error estimates for the Cahn–Hilliard equation on evolving surfaces,' but Section 11 gives only a proof sketch: Parts A–D are described in several paragraphs with references to techniques in [53] and [33], and the consistency part says the defects are 'estimated similar to Section 9.' Theorem 11.1 is stated as a result, yet the accompanying argument is not a complete proof. Either the full details should be supplied, or the claim should be softened to an outline of a possible extension.","section":"§11, Theorem 11.1"}],"minor_comments":[{"comment":"The list of starting values reads '(u^0_h, psi^0_h), ..., (u^{q-1}_h, psi^0_h)'; the final component should presumably be psi^{q-1}_h.","section":"§4.2"},{"comment":"In the reformulation (42), the quantities s_j are used but not defined in the text; they should be specified.","section":"§8.4, Eq. (42)"},{"comment":"In the last sum of (36), the expression mixes \\partial_\\tau and \\partial_\\tau^q: it reads c\\tau \\sum_{i=1}^{q-1} \\|(\\partial_\\tau d^i_3, \\partial_\\tau^q d^i_4)\\|^2_{*,K,\\alpha}, which is likely a typo; both discrete derivatives should be of the same type.","section":"§8.4, Eq. (36)"},{"comment":"The phrase 'The second required initial data (u^1, psi^1)' is grammatically incorrect; it should be 'The second required initial data set' or 'The initial data (u^1, psi^1)'.","section":"§12.2"},{"comment":"Reference [2] has an incomplete author name: 'J., D.' should be replaced with the full name of the author.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The gap in Eq. (34) is serious and is the primary reason for the major-revision recommendation. If the author cannot provide a valid replacement for the inadmissible test function, the main theorem should be withdrawn or substantially weakened. The reference [13] by the same author is closely related; the novelty claim against [13] should be checked carefully in revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the almost-mass-conservation Poincaré technique is genuinely new and the numerical experiments are convincing. But the stability proof has a load-bearing gap. In Part C of Proposition 8.3, the error equation (20a) is tested with (r^k_Ωh, r^k_Γh), the pointwise difference of the nonlinear potentials. These functions are not piecewise linear and do not lie in the discrete test space. The displayed equality (34) is therefore inadmissible, and the bound that follows cannot be a discrete norm bound. The continuous perturbation argument in Section 7 tests with functions in H^1, which is fine, but transferring to the Galerkin setting requires an interpolation or projection step plus a consistency estimate for the remainder. No such estimate appears. So the stability estimate (45), and with it Theorem 5.1, does not follow from the written proof.\n\nWhat is new and good: the idea of exploiting almost mass conservation of the error to get a discrete Poincaré inequality is a real step forward, replacing the anti-symmetric structure argument from [13]. The parameter range K∈[0,∞), L∈(0,∞] is handled, and the evolving-surface result Theorem 11.1 with BDF up to order 5 is a useful extension. The numerical experiments confirm BDF2 convergence and replicate known dynamics.\n\nSoft spots: the Eq. (34) issue is the big one. The imported Poincaré–Wirtinger lemma is secondary; even if fully proved, Part C remains invalid. The proofs for the special cases K∈(0,∞), L=∞ and K=L=0 are only sketches, and the evolving-surface stability is presented as a 'similar' argument without full details.\n\nWho would benefit: numerical analysts working on Cahn–Hilliard with dynamic boundary conditions or on evolving surfaces. The technique may transfer to other mass-conserving problems, as the author suggests.\n\nRecommendation: this deserves a serious referee because the core idea is valuable and the paper is otherwise careful. But the current version does not establish the main theorem; it needs major revision, probably inserting a projection in Part C and proving the resulting consistency terms. I would not cite it as a proven result yet.","headline":"The paper has a good new idea but the main stability proof tests the Galerkin equation with a nonlinear function not in the finite element space, so Theorem 5.1 is not established as written.","tokens_in":36115,"tokens_out":2600,"would_cite":false,"duration_ms":24237,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65M60","65M15","35K55","65M12"],"pacs":[],"model":"deepseek-v4-flash","headline":"A fully discrete scheme for the bulk–surface Cahn–Hilliard system with dynamic boundary conditions is proven to converge with optimal-order error bounds for BDF orders 1 through 5.","keywords":["Cahn–Hilliard equation","dynamic boundary conditions","bulk–surface finite elements","linearly implicit backward difference formula","almost mass conservation","Poincaré-type inequality","fully discrete error estimates","evolving surfaces"],"falsifier":"Compute, for a sequence of quasi-uniform triangulations of a fixed curved domain with $\\alpha\\beta|\\Omega|+|\\Gamma|\\neq 0$, the smallest eigenvalue of the discrete generalized eigenvalue problem for the pair $(a_{K,\\alpha,h},\\|\\cdot\\|_{V_{K,\\alpha,h}})$; if this eigenvalue tends to $0$ as $h\\to 0$, the $h$-independent discrete Poincaré-type inequality used in Part D is false and Theorem 5.1 would not follow. A manufactured-solution run with BDF2 at fixed small $\\tau$ showing an $L^2$ error growing faster than $h^2$ as $h\\to 0$ would equally refute the spatial rate.","tokens_in":34975,"feed_emoji":"📐","tokens_out":11762,"duration_ms":101676,"temperature":0.7,"pith_summary":"The paper establishes that a fully discrete scheme—linear bulk–surface finite elements in space and linearly implicit backward difference time stepping of orders 1 through 5—converges at the optimal rate $h^2+\\tau^q$ for the bulk–surface Cahn–Hilliard system with dynamic boundary conditions, across essentially the full parameter range of transmission and reaction rates. The central claim is Theorem 5.1: the $L^2$ and $H^1$ errors of the phase fields and chemical potentials are bounded by $C(h^2+\\tau^q)$ under the mild step restriction $\\tau^q\\le C_0 h^2$ and sufficiently accurate starting values. This matters because no fully discrete error estimates existed for the general parameter range; the only prior fully discrete result was restricted to the GMS model $K=L=0$ and used an anti-symmetric structure that does not extend. The new tool is a discrete Poincaré-type inequality derived from the almost mass conservation of the error equations, and the same technique is shown to yield optimal-order estimates for the Cahn–Hilliard equation on evolving surfaces.","feed_headline":"Bulk–surface Cahn–Hilliard scheme hits optimal error order","feed_subtitle":"A new 'almost mass conservation' stability argument covers all dynamic-boundary parameter ranges, not just the classical GMS case.","key_machinery":"The carrying mechanism is the combination of an energy estimate for the error equations with a bulk–surface Poincaré–Wirtinger inequality (Lemma 6.1) fed by an almost mass conservation of the error. Because the system conserves the combined bulk–surface mass $\\beta\\int_\\Omega u + \\int_\\Gamma \\psi$, testing the error equation with the constant pair $(\\beta,1)$ shows that the combined mean of the error is controlled by the defect terms and the starting errors; Lemma 6.1 then bounds the full $V_{K,\\alpha}$-norm by the $a_{K,\\alpha}$-seminorm plus that combined mean, under the condition $\\alpha\\beta|\\Omega|+|\\Gamma|\\neq 0$. Higher-order BDF stability is transferred from the continuous energy estimate through Dahlquist's G-stability and the Nevanlinna–Odeh multiplier technique, which supply the discrete product-rule and positivity estimates for orders 1 to 5.","core_discovery":"On the paper's own terms, the discovery is an optimal-order convergence theorem for a fully discrete approximation of a diffuse-interface system with dynamic boundary conditions. For BDF order $q=1,\\ldots,5$, linear bulk–surface finite elements, and a sufficiently smooth exact solution, the scheme's error satisfies $\\|(u_h^n)^\\ell-u(t_n)\\|_{L^2(\\Omega)} + \\|(\\psi_h^n)^\\ell-\\psi(t_n)\\|_{L^2(\\Gamma)} + h(\\|(u_h^n)^\\ell-u(t_n)\\|_{H^1(\\Omega)} + \\|(\\psi_h^n)^\\ell-\\psi(t_n)\\|_{H^1(\\Gamma)}) \\le C(h^2+\\tau^q)$, with a similar bound for the chemical potentials, for all $h\\le h_0$ and $\\tau^q\\le C_0 h^2$, provided the starting values have error $O((\\tau^q+h^2)^2)$. The result covers the parameter range in assumption (A): $K\\in[0,\\infty)$ and $L\\in(0,\\infty)$ with $\\alpha\\beta|\\Omega|+|\\Gamma|\\neq 0$, the case $K\\in(0,\\infty)$, $L=\\infty$ with $\\alpha\\neq 0$, and the case $K=L=0$, thereby including transmission-rate and reaction-rate dependent dynamic boundary conditions as well as the classical GMS and LW models. The proof splits the error into Ritz projection error and scheme error, obtains stability by the almost-mass-conservation/Poincaré argument, and gets consistency from geometric approximation, Ritz-map, and BDF Peano-kernel estimates; only locally Lipschitz potentials are needed because the step restriction enables an $L^\\infty$ bootstrap.","pith_inferences":["Inference: the almost-mass-conservation/Poincaré template should apply to any BDF-discretized conserved gradient flow, such as the Allen–Cahn equation with homogeneous Neumann conditions, and should yield optimal-order $h^2+\\tau^q$ bounds there without needing anti-symmetric structure.","Inference: the excluded condition $\\alpha\\beta|\\Omega|+|\\Gamma|=0$ marks a genuine degeneracy in the combined mass; at that parameter combination a proof would need separate bulk and surface mass controls or a different norm.","Inference: since the stability proposition works under the weaker restriction $\\tau^q\\le C_0 h^\\kappa$ with $\\kappa>3/2$, the $h^2$ restriction in the theorem may be an artifact of the inverse-estimate bootstrap; a numerical test of the rates under the weaker restriction would settle whether it is removable."],"forward_implications":["For BDF orders $q=1,\\ldots,5$, the fully discrete bulk–surface finite element scheme converges with the optimal rates $O(h^2+\\tau^q)$ in $L^2$ and $H^1$ for phase fields and chemical potentials, for all parameters covered by assumption (A).","The transmission-rate and reaction-rate dependent dynamic boundary conditions are handled uniformly, excluding only the degenerate combination $\\alpha\\beta|\\Omega|+|\\Gamma|=0$ and the formally dashed limit lines in the parameter square.","The same stability strategy yields optimal-order fully discrete estimates for the Cahn–Hilliard equation on evolving surfaces (Theorem 11.1), with the same BDF orders and the same mild step restriction.","Only local Lipschitz continuity of $F_\\Omega$ and $F_\\Gamma$ and their first three derivatives is required, so the commonly used double-well potentials are admissible.","Starting values need only be accurate to order $O((\\tau^q+h^2)^2)$, and no shift by initial data is needed in the error analysis."],"supporting_citations":[{"why":"Supplies Lemma A.1, the bulk–surface Poincaré–Wirtinger inequality that the stability proof's Part D imports as Lemma 6.1.","marker":"[45]"},{"why":"Provides the predecessor fully discrete BDF analysis for the GMS model $K=L=0$, whose anti-symmetric stability argument this paper replaces.","marker":"[13]"},{"why":"G-stability equivalence used to obtain the BDF energy estimate for orders 1 to 5.","marker":"[25]"},{"why":"Multiplier technique with the $\\eta$ factors listed for $q=1,\\ldots,5$, used together with G-stability in the time-discrete energy estimates.","marker":"[26]"},{"why":"Abstract parabolic framework with dynamic boundary conditions, including Ritz map and spatial error estimates that the consistency analysis uses.","marker":"[23]"},{"why":"Isoparametric bulk–surface finite element construction and geometric approximation error estimates underlying the spatial discretization.","marker":"[21]"},{"why":"Defines the bulk–surface Cahn–Hilliard system with dynamic boundary conditions and supplies the well-posedness and regularity of the exact solution.","marker":"[8]"},{"why":"Previous semi-discrete error estimates for Cahn–Hilliard-type dynamic boundary conditions used in deriving the defect bounds.","marker":"[12]"},{"why":"Evolving surface finite element method for the Cahn–Hilliard equation that Section 11 builds on for the generalization.","marker":"[50]"}],"fun_headline_variants":["Optimal error order for dynamic boundary Cahn–Hilliard","Almost mass conservation yields optimal Cahn–Hilliard error estimates","Optimal-order error estimates for dynamic-boundary Cahn–Hilliard","Full discretization error bounds for Cahn–Hilliard with dynamic boundary","Optimal convergence for bulk–surface Cahn–Hilliard with dynamic BCs"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof assumes that the bulk–surface Poincaré–Wirtinger inequality (Lemma 6.1, imported from [45, Lemma A.1]) controls the full norm by the energy seminorm whenever $\\alpha\\beta|\\Omega|+|\\Gamma|\\neq 0$, and that this control carries over to the finite-element spaces with a constant independent of $h$; if that discrete control fails, the stability estimate and Theorem 5.1 collapse.","fun_headline_variants_meta":{"raw":{"variants":["Optimal error order for dynamic boundary Cahn–Hilliard","Almost mass conservation yields optimal Cahn–Hilliard error estimates","Optimal-order error estimates for dynamic-boundary Cahn–Hilliard","Full discretization error bounds for Cahn–Hilliard with dynamic boundary","Optimal convergence for bulk–surface Cahn–Hilliard with dynamic BCs"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000897,"raw_usage":{"total_tokens":3923,"prompt_tokens":1063,"completion_tokens":2860,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":679,"completion_tokens_details":{"reasoning_tokens":2761}},"tokens_in":679,"tokens_out":2860,"duration_ms":19424,"temperature":1.0,"reasoning_tokens":2761,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T00:30:55.377481+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute, for a sequence of quasi-uniform triangulations of a fixed curved domain with $\\alpha\\beta|\\Omega|+|\\Gamma|\\neq 0$, the smallest eigenvalue of the discrete generalized eigenvalue problem for the pair $(a_{K,\\alpha,h},\\|\\cdot\\|_{V_{K,\\alpha,h}})$; if this eigenvalue tends to $0$ as $h\\to 0$, the $h$-independent discrete Poincaré-type inequality used in Part D is false and Theorem 5.1 would not follow. A manufactured-solution run with BDF2 at fixed small $\\tau$ showing an $L^2$ error growing faster than $h^2$ as $h\\to 0$ would equally refute the spatial rate.","supporting_citations":[{"cited_title":"https://arxiv.org/abs/2407","cited_arxiv_id":null,"evidence_quote":"Provides the predecessor fully discrete BDF analysis for the GMS model $K=L=0$, whose anti-symmetric stability argument this paper replaces."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Multiplier technique with the $\\eta$ factors listed for $q=1,\\ldots,5$, used together with G-stability in the time-discrete energy estimates."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Abstract parabolic framework with dynamic boundary conditions, including Ritz map and spatial error estimates that the consistency analysis uses."},{"cited_title":"NoDEA Nonlinear Different ial Equations Appl","cited_arxiv_id":null,"evidence_quote":"Defines the bulk–surface Cahn–Hilliard system with dynamic boundary conditions and supplies the well-posedness and regularity of the exact solution."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Previous semi-discrete error estimates for Cahn–Hilliard-type dynamic boundary conditions used in deriving the defect bounds."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Evolving surface finite element method for the Cahn–Hilliard equation that Section 11 builds on for the generalization."}],"review_version":1}