{"id":"73b1d3df-6ab4-4bbf-9915-6c465eea1d90","arxiv_id":"1908.06124","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Existence, uniqueness, and a linear-in-K error estimate are established for a Robin-boundary approximation of an extended Cahn-Hilliard system with affine dynamic boundary conditions.","lead":"This paper proves existence and uniqueness for a Cahn-Hilliard system with a new affine-linear dynamic boundary condition, and shows that a Robin-boundary regularization converges to the limit model at a linear rate. The main value is a rigorous convergence and error estimate for a practical penalization approach, backed by finite-element simulations.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The O(K) convergence claim is proven only under assumptions that exclude the standard quartic surface potential in the κ=0 no-surface-diffusion case; the numerical-proxy claim is therefore narrower than the abstract suggests.","rationale":"I read Theorem 2.3 as the central claim: for affine H, the Robin solutions converge weakly to the unique Liu–Wu-type solution and, under extra regularity, satisfy the O(K) bound (2.14). The proof structure is standard and internally consistent; I found no algebraic contradiction in the main convergence argument. The soft spot is not an error but a scope restriction. The quantitative convergence result requires (2.6), and the κ=0 well-posedness of the Robin system requires (B1), i.e., a purely convex surface potential with no concave correction. The standard quartic double-well potential commonly used in Cahn–Hilliard models does not satisfy these conditions. Thus the paper's numerical-proxy statement, while true under its assumptions, does not cover a natural and physically relevant regime. The reader's weakest assumption already flagged the same structural restrictions, so my concern is a refinement rather than a new objection. Because the theorem is explicitly conditional, the existing CONDITIONAL verdict remains appropriate; no verdict change is needed. The proposed numerical test would clarify whether the restriction is essential or merely an artifact of the proof technique.","tokens_in":46237,"tokens_out":24464,"duration_ms":234719,"concrete_test":"Independently rerun the two finite-element schemes (7.2) and (7.4) for κ=0 with affine H(s)=αs+β and G(s)=1/4(s^2−1)^2, using the same discretization and K-sequence as Tables 1–5 (K=10^{-1},...,10^{-5}), and compute the experimental orders in (2.14). If the Robin scheme converges and the EOCs stay near 1, the excluded double-well potential is likely coverable by a relaxed theorem, so the restriction is a proof artifact; if the scheme fails to converge or the EOC degrades, the restriction is essential and the paper's numerical-proxy claim must be explicitly limited to convex G. A positive numerical result would not replace a proof, but it would decide whether the concern is merely a missing extension or a genuine limitation of the method.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 2.3's quantitative statement (2.14) is the load-bearing part of the paper: it is what licenses using (1.3) as a proxy for (1.2). For κ=0, the proof needs the Robin system to be well-posed, and Theorem 2.2 restricts this to (B1): G is purely convex (G2≡0) with G1'' of polynomial growth, and H affine. Moreover the error estimate in Section 6 is set up only after invoking (2.6), which for κ=0 requires |G1''|≤c1. The standard quartic double-well potential G(s)=1/4(s^2−1)^2 has the natural decomposition G1(s)=s^4/4, G2(s)=−s^2/2+1/4; this violates both G2≡0 and bounded G1''. So for the no-surface-diffusion case with the usual double-well surface energy—precisely the regime where the Robin regularisation was introduced to handle nonlinear H—the existence of the Robin solutions and the linear error bound are not established; only subsequential weak convergence is stated under the same restrictive (B1). The theorem is internally consistent, but the headline 'Robin approximation converges with O(K)' should be read as conditionally restricted, not as a statement for the original Liu–Wu model's standard potentials.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes a Cahn–Hilliard system with a dynamic boundary condition, extending the Liu–Wu model by allowing an affine transmission relation u=αv+β on the boundary. The main results are: (i) Theorem 2.1, global existence and uniqueness of weak solutions to the extended limit system (1.1)–(1.2), including the no-surface-diffusion case κ=0; (ii) Theorem 2.2, global existence and uniqueness for the Robin-penalized system (1.3) under assumptions (A1)–(A3), with additional restrictions (B1)–(B2) in the case κ=0; and (iii) Theorem 2.3, weak convergence as K→0 and an explicit O(K) error estimate for affine H, under the additional regularity hypothesis (2.6). The proofs use the gradient-flow structure of both systems and an implicit time discretization, following the framework of [19], with a Minty-trick argument for the κ=0 surface potential and elliptic regularity estimates for the boundary normal derivative. The paper also presents finite-element simulations that show linear experimental order of convergence for affine relations and for two nonlinear relations.","tokens_in":46460,"tokens_out":11855,"duration_ms":119077,"significance":"If the results are correct, they provide a rigorous justification for using the Robin problem (1.3) as a numerical proxy for the dynamic boundary condition in the affine case, with a quantified rate. The treatment of κ=0 is a genuine extension of previous work and the restriction to convex surface potentials with G2≡0 is stated transparently in the theorem hypotheses. The paper is built on standard compactness and elliptic-regularity arguments, and the main theorems are internally consistent; no parameter fitting is involved. The numerical experiments are consistent with the theoretical rate, although they only cover κ>0 and no code is supplied. The main caveat is that the quantitative O(K) result in the no-surface-diffusion case does not cover the standard quartic double-well potential; this should be flagged prominently so that the scope of the headline claim is not overstated.","major_comments":[],"minor_comments":[{"comment":"In the convergence statement of Theorem 2.3, the line “µK⇀µ in L2(0,T;H1(Γ))” should read L2(0,T;H1(Ω)), since µ is the bulk chemical potential; the surface limit is correctly stated for αµKΓ.","section":"Theorem 2.3, Section 2"},{"comment":"The abstract and conclusion state that the error estimate covers the no-surface-diffusion case, but for κ=0 the estimate is proved only under (B1) and (2.6), which require G2≡0 and |G1′′| bounded. These assumptions exclude the standard quartic double-well potential G(s)=1/4(s2−1)2, for which the natural decomposition has G2(s)=−s2/2+1/4 and G1′′(s)=3s2. Please state this restriction explicitly in the abstract and conclusion so that the scope of the O(K) claim is unambiguous.","section":"Abstract and Section 8"},{"comment":"The convergence assertions for the time-discrete approximations are delegated to [19, Lem. 8] with the statement that the proof is “almost analogous.” For κ=0 the compactness of vN in C([0,T];H1(Γ)′) and the absence of strong convergence in L2(ΣT) are new features that are not present in [19]; please provide a self-contained argument or at least spell out the compact embedding and the equicontinuity steps used in this case.","section":"Section 5, Step 5"},{"comment":"All numerical experiments for the convergence as K→0 are performed with κ>0 (κ=1 or κ=0.4), so the numerically observed O(K) rate is not evidence for the analytically harder κ=0 case. A sentence acknowledging this and explaining whether the κ=0 regime is numerically accessible with the stated assumptions would be helpful.","section":"Section 7"},{"comment":"The numerical implementation is described only through references to earlier work and MATLAB code; no code or data are provided. Please include a reproducibility statement or make the code available, since the convergence tables are otherwise difficult to verify independently.","section":"Section 7.1"},{"comment":"The uniform estimates in Step 2 are imported from [19] with the phrase “immediately infer.” Because the new affine transmission condition changes the boundary coupling and the κ=0 case changes the surface regularity, please specify exactly which estimates from [19] are being used and why the modifications do not affect them.","section":"Section 4, Step 2"}],"recommendation":"minor_revision","confidential_remarks":"The manuscript is a solid extension of published work and the main theorems are plausible and internally consistent. The principal concern is one of scope presentation: the κ=0 O(K) claim should be more carefully qualified. I recommend minor revision rather than rejection, provided the authors clarify the hypotheses and expand the κ=0 compactness argument as requested."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nRead Knopf-Lam on the Robin boundary approximation for the Liu-Wu Cahn-Hilliard system. Short version: this is a competent extension paper, and the central convergence claim is real but narrower than the abstract makes it sound. The genuinely new pieces are the affine-linear transmission condition u=αv+β and the proof that Robin regularization converges to the extended system, with an O(K) error bound in the affine case. The κ=0 no-surface-diffusion regime is also treated, and that is where the interesting restrictions live.\n\nWhat the paper does well: existence and uniqueness for both the limit and Robin systems are built on a natural gradient-flow structure, and the proofs are mostly clean. Theorem 2.3 is derived honestly: the error depends on ∂_nu∈L2 and the initial-data mismatch, and the K-rate is linear. The numerical experiments are detailed, with EOC tables for several parameter choices and two nonlinear H; they support the affine theorem and give suggestive evidence for the nonlinear case. No code or data is included, but for an analysis paper that is a minor omission, not a real defect.\n\nSoft spots, in order of importance. First, the κ=0 case is less general than the abstract implies. Theorem 2.3 for κ=0 requires B1: G2≡0 and G1'' of polynomial growth, plus the extra regularity condition (2.6) with |G1''| bounded for the error estimate. The standard quartic double-well potential cannot be decomposed that way with G2≡0 and G1 convex. So for the no-surface-diffusion case with the usual nonconvex surface energy, the paper proves only weak convergence under restrictive assumptions, and the O(K) proxy result is not established. The authors state these assumptions openly, so it is not a hidden flaw, but it should be flagged in the abstract and conclusion if the paper is revised.\n\nSecond, several load-bearing compactness and limit steps in Sections 4 and 5 are deferred to [19] and [8] with phrases like \"we omit the details.\" That is acceptable when the analogy is exact, and here it mostly is, but a referee should confirm those citations carry the weight.\n\nThird, the numerical section observes O(K) for nonlinear H, while the analysis only covers existence in that regime. The text is careful not to claim a theorem there, so this is just an honest observation.\n\nOverall, the theorems are internally consistent, the citations are to independent published results, and the paper is upfront about its restrictions. It deserves serious refereeing. If it were mine, I would ask the authors to state the κ=0 convexity restriction in the abstract and to expand the deferred arguments where they are load-bearing.","headline":"Solid extension paper: affine-linear transmission plus Robin convergence with O(K), but the κ=0 regime only covers convex surface potentials, not the standard double-well case.","tokens_in":47036,"tokens_out":2545,"would_cite":true,"duration_ms":27361,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35A01","35A02","35A35","35B40"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that a Robin boundary penalisation converges to a generalized Cahn–Hilliard dynamic boundary model with a first-order error estimate.","keywords":["Cahn–Hilliard system","dynamic boundary conditions","Robin boundary approximation","affine transmission condition","gradient flow","weak solutions","error estimates","finite element method"],"falsifier":"Compute the left side of (2.14) for a sequence of Robin solutions with affine $H$, a smooth domain, and initial data chosen so that the limit solution has $\\partial_\\nu u$ outside $L^2(\\Sigma_T)$, for instance by violating the growth condition (2.6) on $F_1'$. If the error $u^K-(\\alpha v^K+\\beta)$ in $L^2(\\Sigma_T)$ fails to be $O(K)$, the claimed estimate cannot hold with a constant independent of the solution; conversely, the same scheme run on a corner domain where elliptic regularity fails and the experimental order of convergence drops would directly test the regularity hypothesis.","tokens_in":45990,"feed_emoji":"🧮","tokens_out":8161,"duration_ms":78049,"temperature":0.7,"pith_summary":"The paper establishes that a Robin boundary condition can stand in for a nonlocal dynamic boundary condition in a Cahn–Hilliard system, with controlled error. For affine-linear transmission relations of the form $u|_{\\Sigma}=\\alpha v+\\beta$, the weak solutions of the Robin-regularised system converge as the penalty parameter $K\\to 0$ to the unique weak solution of the extended model (1.2), at a linear rate $O(K)$ whenever the boundary normal derivative of the limit solution is square-integrable. The paper also proves existence and uniqueness of weak solutions for both the extended model and the Robin regularisation, including the difficult case of no surface diffusion ($\\kappa=0$) under extra convexity restrictions. Finite-element experiments display the predicted first-order convergence and reveal boundary dynamics, including a competition between mass conservation and Cahn–Hilliard phase separation, that do not appear when the boundary and bulk variables are simply set equal.","feed_headline":"Robin penalty reproduces Cahn-Hilliard boundary dynamics at linear rate","feed_subtitle":"Rigorous bounds show the boundary coupling is recovered as the penalty parameter tends to zero.","key_machinery":"The load-bearing structure is the gradient-flow formulation of both systems, with energies $E^*$ and $E$, together with the associated implicit time discretisation in which each time step minimises a functional of the form $J_n(u)=\\frac{1}{2\\tau}\\|u-u_n\\|^2_{(V_0)'_*}+E(u)$ (or the analogue for $E$). The Robin penalty term $K^{-1}(H(v)-u)$ couples the bulk and surface variables and, as $K\\to 0$, enforces the affine transmission relation. The convergence proof combines uniform energy estimates, compactness from H\\\"older-in-time bounds, and, for $\\kappa=0$, Minty's monotonicity trick to identify the weak limit of $G'(v^K)$. The quantitative error estimate relies on a second-order regularity result giving $\\partial_\\nu u\\in L^2(\\Sigma_T)$, followed by a Gronwall argument applied to the difference of the two weak formulations.","core_discovery":"The central assertion is Theorem 2.3: with $H(s)=\\alpha s+\\beta$, if the initial data of the Robin system (1.3) are uniformly bounded in energy and converge in the natural weak sense to data of the limit system (1.2), then the Robin solutions converge, up to a subsequence, to the unique weak solution of the extended model. Under a supplementary elliptic-regularity condition on the limit solution, the error satisfies the estimate $$\\sup_{t\\in[0,T]}\\|(u^K-u,v^K-v)(t)\\|^2_{(H_0)'}+\\|u^K-u\\|^2_{$L^{4}$(0,T;$L^{2}$(\\$\\Omega$))}+\\|u^K-u\\|^2_{$L^{2}$(0,T;$H^{1}$(\\$\\Omega$))}+\\|v^K-v\\|^2_{$L^{4}$(0,T;$L^{2}$(\\Gamma))}+\\|v^K-v\\|^2_{$L^{2}$(0,T;$H^{1}$(\\Gamma))}+$K^{{-1}}$\\|u^K-(\\$\\alpha$ v^K+\\$\\beta$)\\|^2_{$L^{2}$(\\Sigma_T)}\\le C\\left(K\\|\\partial_\\nu u\\|^2_{$L^{2}$(\\Sigma_T)}+\\|(u^K_0-u_0,v^K_0-v_0)\\|^2_{(H_0)'}\\right).$$ This says that the affine transmission condition is recovered from the penalty term $K^{-1}(H(v)-u)$ at first order in $K$, with the error controlled by the boundary normal derivative of the limit solution and the mismatch in the initial data. Existence and uniqueness for both systems are established by exploiting their gradient-flow structure through an implicit time discretisation.","pith_inferences":["If the linear-rate convergence observed for nonlinear $H$ can be proved under assumptions that replace the affine structure, the Robin regularisation would become a general numerical route to nonlinear dynamic boundary conditions; a natural first target is monotone $H$ with controlled growth, where Minty's trick might replace the affine structure.","The error bound's dependence on $\\|\\partial_\\nu u\\|$ suggests that convergence will slow down when the limit solution develops steep boundary layers, making an adaptive choice of $K$ tied to the local normal-derivative size a testable extension.","The numerical competition between boundary mass conservation and phase separation suggests a predictive criterion based on whether the boundary mean lies between the stable minima of the shifted surface potential $G(\\alpha^{-1}(s-\\beta))$; the authors report explicit counterexamples, so a refined criterion involving profile shape or parameter ratios is still open.","The same gradient-flow discretisation used for existence can likely be turned into a structure-preserving scheme; proving convergence of the fully discrete scheme as the mesh size and time step go to zero simultaneously with $K$ would make the numerical proxy fully rigorous."],"forward_implications":["For affine-linear relations, the Robin system is a bona fide numerical surrogate: the transmission condition is recovered at rate $O(K)$ in $L^2(\\Sigma_T)$, and the other error norms obey the explicit bound (2.14).","With well-prepared initial data the error is controlled purely by $K\\|\\partial_\\nu u\\|^2_{L^2(\\Sigma_T)}$, so the rate is genuinely first order whenever the limit solution has a square-integrable normal boundary trace.","The extended model allows the surface and bulk order parameters to be opposite, as with $\\alpha=-1$, producing solutions such as $(u,v)=(\\pm 1,\\mp 1)$ that are impossible in the original model where $u=v$ on the boundary.","The numerical experiments show the same first-order behaviour for nonlinear relations such as $H(s)=\\sin s$ and $H(s)=3\\cos s+2$, suggesting that the affine-linear theory is the first instance of a wider principle.","In the absence of surface diffusion ($\\kappa=0$), existence, uniqueness, and convergence still hold, but the error estimate loses the surface $H^1$ and $L^4$-in-time bounds and requires a convex surface potential with $G_2\\equiv 0$."],"supporting_citations":[{"why":"Supplies the original Cahn–Hilliard model with dynamic boundary conditions that this paper extends to the affine relation $u|_{\\Sigma}=\\alpha v+\\beta$.","marker":"[25]"},{"why":"Provides the gradient-flow framework and implicit time-discretisation method used here to prove existence and uniqueness of weak solutions.","marker":"[19]"},{"why":"Establishes the corresponding Robin-approximation convergence and error estimate for the Allen–Cahn system, which the present argument adapts to the Cahn–Hilliard setting.","marker":"[8]"},{"why":"Gives the convexity-based Euler–Lagrange analysis used to derive the discrete weak formulation in the time discretisation.","marker":"[18]"},{"why":"Supplies the generalised Poincaré inequality used to bound the chemical potentials in the uniform estimates.","marker":"[1]"},{"why":"Provides the $H^2$-regularity argument strategy that yields $\\partial_\\nu u\\in L^2(\\Sigma_T)$ under condition (2.6).","marker":"[3]"},{"why":"Supplies a related transmission problem for Cahn–Hilliard systems with nonsmooth potentials, used as a comparison and motivation for the affine relation.","marker":"[9]"}],"fun_headline_variants":["Robin penalty recovers Cahn-Hilliard boundary coupling to first order","Affine boundary relation emerges from Robin limit in Cahn-Hilliard","Cahn-Hilliard Robin approximation: convergence and error estimate","Proving convergence of Robin regularization for dynamic Cahn-Hilliard","Linear-rate convergence for Robin-approximated Cahn-Hilliard"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The quantitative convergence result stands on structural assumptions not present in the original model: when there is no surface diffusion the surface potential must be purely convex with $G_2\\equiv 0$ and polynomially growing $G_1''$, and the $O(K)$ error bound requires the elliptic regularity condition (2.6) that puts the normal derivative of the limit solution in $L^2(\\Sigma_T)$; if either fails, the paper does not prove linear convergence.","fun_headline_variants_meta":{"raw":{"variants":["Robin penalty recovers Cahn-Hilliard boundary coupling to first order","Affine boundary relation emerges from Robin limit in Cahn-Hilliard","Cahn-Hilliard Robin approximation: convergence and error estimate","Proving convergence of Robin regularization for dynamic Cahn-Hilliard","Linear-rate convergence for Robin-approximated Cahn-Hilliard"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00031,"raw_usage":{"total_tokens":1833,"prompt_tokens":1077,"completion_tokens":756,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":693,"completion_tokens_details":{"reasoning_tokens":663}},"tokens_in":693,"tokens_out":756,"duration_ms":6474,"temperature":1.0,"reasoning_tokens":663,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:54:57.391990+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the left side of (2.14) for a sequence of Robin solutions with affine $H$, a smooth domain, and initial data chosen so that the limit solution has $\\partial_\\nu u$ outside $L^2(\\Sigma_T)$, for instance by violating the growth condition (2.6) on $F_1'$. If the error $u^K-(\\alpha v^K+\\beta)$ in $L^2(\\Sigma_T)$ fails to be $O(K)$, the claimed estimate cannot hold with a constant independent of the solution; conversely, the same scheme run on a corner domain where elliptic regularity fails and the experimental order of convergence drops would directly test the regularity hypothesis.","supporting_citations":[{"cited_title":"Liu and H","cited_arxiv_id":null,"evidence_quote":"Supplies the original Cahn–Hilliard model with dynamic boundary conditions that this paper extends to the affine relation $u|_{\\Sigma}=\\alpha v+\\beta$."},{"cited_title":"Garcke and P","cited_arxiv_id":null,"evidence_quote":"Provides the gradient-flow framework and implicit time-discretisation method used here to prove existence and uniqueness of weak solutions."},{"cited_title":"Colli, T","cited_arxiv_id":null,"evidence_quote":"Establishes the corresponding Robin-approximation convergence and error estimate for the Allen–Cahn system, which the present argument adapts to the Cahn–Hilliard setting."},{"cited_title":"Garcke, On Cahn–Hilliard systems with elasticity","cited_arxiv_id":null,"evidence_quote":"Gives the convexity-based Euler–Lagrange analysis used to derive the discrete weak formulation in the time discretisation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the generalised Poincaré inequality used to bound the chemical potentials in the uniform estimates."},{"cited_title":"Calatroni and P","cited_arxiv_id":null,"evidence_quote":"Provides the $H^2$-regularity argument strategy that yields $\\partial_\\nu u\\in L^2(\\Sigma_T)$ under condition (2.6)."},{"cited_title":"Computational universality of symmetry-protected topologically ordered cluster phases on 2D Archimedean lattices","cited_arxiv_id":"1907.13279","evidence_quote":"Supplies a related transmission problem for Cahn–Hilliard systems with nonsmooth potentials, used as a comparison and motivation for the affine relation."}],"review_version":1}