{"id":"a3309b24-d867-4970-b088-e4a5e7359a6d","arxiv_id":"1908.03395","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Reliable and efficient residual-type a posteriori error estimators are derived and analyzed for the mortar staggered DG method, with adaptive experiments on singular solutions.","lead":"Two new error estimators are derived for a mortar discontinuous Galerkin method used to solve elliptic problems on grids that do not match across subdomain boundaries. The paper proves these estimators guide adaptive meshing reliably and efficiently, without needing a saturation assumption common in earlier work.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 3.1 is not valid for the paper's discontinuous coefficients: it jumps ρ^{1/2}u_h rather than the ρ^{1/2}-weighted jump of u_h, so the nonconforming bound in Theorem 3.2 is unproven.","rationale":"The reader correctly identified Lemma 3.1 as load-bearing and noted that it is imported from [22,34] without proof for the mortar SDG space. My stress test sharpens this: the lemma is not merely unproven in the present setting, it is false as stated when ρ is discontinuous at the mortar interface, since a perfectly conforming u_h with a coefficient jump gives a positive right-hand side and zero left-hand side. This directly affects Theorem 3.2, the energy reliability result, which is central to the paper's claims. The issue is fixable by replacing the jump in Lemma 3.1 with a correctly weighted jump and by adding interface jump contributions with the appropriate coefficient (typically the maximum of the two adjacent ρ values) to the estimator in (3.8). Because this requires a technical modification of both the lemma and the estimator, the paper should not be accepted without that revision. The reader's CONDITIONAL verdict is therefore appropriate; my concern does not move the verdict, but it identifies a concrete mathematical gap rather than only a missing proof detail. The unstated elliptic regularity assumption in Theorem 3.1 remains a secondary presentation issue, but the discontinuous-coefficient jump problem is the more serious obstacle.","tokens_in":15956,"tokens_out":38422,"duration_ms":383957,"concrete_test":"Analytical check: on Ω=(0,2)×(0,1), split into Ω_1=(0,1)×(0,1) with ρ=ε and Ω_2=(1,2)×(0,1) with ρ=1, and take a piecewise-linear conforming u_h∈Vh that vanishes on ∂Ω and is continuous across x=1 (e.g., a hat function on the coarse mesh). Compute both sides of Lemma 3.1: the left side is 0 because u_h∈H01, while the right side equals (1−√ε)^2 h^{-1}‖u_h‖_{0,{1}×(0,1)}^2>0. This disproves the lemma as stated. If the intended meaning of ⟦ρ^{1/2}u_h⟧ is ρ^{1/2}⟦u_h⟧ with an edge value of ρ, repeat the test with the nonmortar side on the low-ρ subdomain and a function whose only defect is a nonmortar interface jump; compare the estimator's interface term with the true nonconforming distance to check whether the high-ρ contribution is missing.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Lemma 3.1 is the only control of the nonconforming part in Theorem 3.2, but as stated it cannot hold for the piecewise-constant ρ used in the paper. The lemma bounds min_{v∈H01}‖ρ^{1/2}∇(v−u_h)‖ by (Σ_{Fp∪T_{Γ,h}} h_e^{-1}‖⟦ρ^{1/2}u_h⟧‖^2)^{1/2}. With the standard jump definition in §2, on an interface Γ_{ij} one has ⟦ρ^{1/2}u_h⟧=ρ_i^{1/2}u_{h,i}−ρ_j^{1/2}u_{h,j}. If u_h is continuous across Γ and vanishes on ∂Ω, then u_h∈H01 and the left side is exactly zero, while the right side is positive whenever ρ_i≠ρ_j. Thus the stated lemma fails precisely for the interface problems the paper targets, including Example 5.3. A correct weighted nonconforming estimate would use the ρ^{1/2}-weighted jump of u_h, i.e. ‖ρ^{1/2}⟦u_h⟧‖^2, with contributions from both sides of the interface (or a maximum coefficient), not the jump of the weighted function ρ^{1/2}u_h. The estimator (3.8) only contains h^{-1}ρ‖⟦u_h⟧‖^2 on the nonmortar edges T_{Γ,h}, with ρ taken from that element; if the nonmortar side has the smaller ρ, this term can be far smaller than the contribution required from the larger-coefficient side. Consequently, Theorem 3.2 is not established for discontinuous coefficients as written.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript develops two residual-type a posteriori error estimators for the mortar staggered discontinuous Galerkin (SDG) discretization of the scalar second-order elliptic problem (2.1) with piecewise-constant coefficients and nonmatching subdomain meshes. The first estimator, η1 defined in (3.4), targets the L2 error and is derived by a duality argument; the second, η2 defined in (3.9), targets the ρ-weighted H1 seminorm and is derived with the help of an auxiliary H1_0 function that splits the energy error into conforming and nonconforming parts. The authors claim reliability of both estimators (Theorems 3.1 and 3.2), efficiency (Theorem 4.1), and the absence of saturation assumptions, and they present adaptive refinement experiments, including a transmission problem with a corner singularity.","tokens_in":16283,"tokens_out":20417,"duration_ms":179870,"significance":"If the proofs were correct, the paper would make a useful contribution: it would extend residual-based a posteriori control to mortar SDG on nonmatching grids without saturation assumptions and with explicit interface and mortar-flux terms in the estimators. The derivation is constructive and the estimators are not fitted to numerical data; the numerical experiments illustrate plausible adaptive behavior. However, the reliability proof for the energy estimator rests on a nonconforming estimate that is false for discontinuous coefficients, so the main claim is not established as stated. The paper has merit and the issue appears fixable, but a substantive revision is required.","major_comments":[{"comment":"Lemma 3.1 is not valid for the piecewise-constant coefficient ρ considered in §2, and Theorem 3.2 relies on it through (3.12). With the jump convention of §2, on an interface Γ_ij one has ⟦ρ^{1/2}u_h⟧ = ρ_i^{1/2}u_{h,i} − ρ_j^{1/2}u_{h,j}. Take any nonzero continuous function u_h ∈ H^1_0(Ω) that belongs to the SDG space; then min_{v∈H^1_0}‖ρ^{1/2}∇(v−u_h)‖ = 0, while the right-hand side of Lemma 3.1 is positive whenever ρ_i ≠ ρ_j and u_h does not vanish on Γ_ij. Thus the stated lemma is false. A correct nonconforming bound for the weighted energy norm must involve the jump of u_h weighted by a coefficient that accounts for both sides, for example a term of the form (ρ_i+ρ_j)h_e^{-1}‖⟦u_h⟧‖^2_{0,e}; it cannot use the jump of the weighted function ρ^{1/2}u_h. As written, the only interface contribution in η2, (3.8), is h_e^{-1}ρ‖⟦u_h⟧‖^2 with ρ taken from the element on the non-mortar side, and if that side has the smaller coefficient the estimator undercounts the contribution needed from the larger-coefficient side. Consequently Theorem 3.2 is not established for discontinuous coefficients, which is precisely the setting of Example 5.3.","section":"§3.2, Lemma 3.1 and Theorem 3.2"},{"comment":"The proof of Theorem 3.1 assumes the elliptic regularity estimate (3.6) for the dual problem, but the theorem as stated contains no such hypothesis. For the transmission problem (2.1) with interface corners, as in Example 5.3 where the solution has regularity 1+α < 2, the dual problem need not be H^2-regular. The L2 reliability bound is therefore conditional as written. The hypothesis should be stated explicitly and the theorem restricted accordingly, or an alternative proof that avoids full H^2 regularity of the dual problem should be supplied.","section":"§3.1, Theorem 3.1"}],"minor_comments":[{"comment":"In the edge terms of the local estimator, the text should clarify whether ρ is evaluated on the element τ or on a particular side of an interface; since ρ is discontinuous, an unambiguous convention is needed for the boundary and interface sums.","section":"§3.2, Eq. (3.8)"},{"comment":"The legends in Fig. 7 appear truncated, for example '|| 1/2 (u-uh)||', and the intended quantity is presumably ‖ρ^{1/2}∇(u−u_h)‖; the labels should be corrected so the reported convergence history is unambiguous.","section":"§5, Example 5.2"},{"comment":"There are several typographical errors, including 'Bot h' in the abstract and 'coefficients'; a careful proofreading pass is needed before publication.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The central issue is the incorrect nonconforming estimate in Lemma 3.1; this is a load-bearing technical gap in the proof of Theorem 3.2, not a stylistic matter. The mistake appears fixable within the scope of the manuscript by using a genuinely weighted jump of u_h and modifying η2 accordingly, so I would not recommend rejection outright. The numerical experiments are encouraging but do not compensate for the missing proof. I would ask the authors to correct the nonconforming estimate and restate Theorem 3.2, and to make the regularity hypothesis in Theorem 3.1 explicit."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, the good news: this is the first paper to give a posteriori error estimators for the mortar staggered DG method, and the L2 estimator (η1, Theorem 3.1) looks sound. The duality argument is standard, the residuals are the right ones, and the numerical experiments (including the singular example and the interface example) show the expected adaptive behavior. The efficiency bounds in Section 4 are also plausible and follow the usual bubble-function machinery.\n\nBut the energy-error estimator (η2, Theorem 3.2) has a load-bearing flaw. Lemma 3.1 claims that the H^1_0-conforming distance of u_h is bounded by the jumps of ρ^{1/2}u_h on F_p ∪ T_{Γ,h}. For the piecewise-constant coefficient ρ used throughout the paper, that is false. Take a u_h that is continuous across an interface where ρ jumps: the left side is zero (choose v = u_h), while the right side is positive because ⟦ρ^{1/2}u_h⟧ = ρ_i^{1/2}u_{h,i} − ρ_j^{1/2}u_{h,j} does not vanish even when ⟦u_h⟧ = 0. The correct quantity would be the weighted jump of u_h, ρ^{1/2}⟦u_h⟧, not the jump of the weighted function. The estimator (3.8) only contains h^{-1}ρ‖⟦u_h⟧‖^2 on the nonmortar edges, with ρ from the nonmortar side; if that side has the smaller coefficient, the term is weaker than what the lemma would need. So Theorem 3.2 is unproven for exactly the discontinuous-coefficient problems the paper advertises (Example 5.3). This is not a cosmetic issue; it breaks the reliability proof.\n\nAlso, Theorem 3.1 assumes elliptic regularity for the dual problem but only states it inside the proof. That's a minor fix—just add it to the hypotheses.\n\nFor constant ρ, the energy estimator is probably fine, and the L2 part stands on its own. But as written, the main result for variable ρ is not established. I'd recommend sending this to a serious referee, but the referee should be told to focus on Lemma 3.1. If the authors can replace it with a correct weighted-jump estimate and adjust η2 accordingly, the paper would be a solid contribution. If not, the energy estimator should be restricted to constant coefficients.\n\nThe citation pattern is fine: self-citations are for background and earlier SDG estimators, not for the disputed lemma. No serious circularity.","headline":"First a posteriori estimators for mortar SDG, but the energy-error reliability proof rests on a lemma that fails for the discontinuous coefficients the paper targets.","tokens_in":16820,"tokens_out":3818,"would_cite":false,"duration_ms":34589,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65N30","65N15","65N50"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that two residual-type error estimators for the mortar staggered discontinuous Galerkin method are both reliable and efficient for second-order elliptic problems, with no saturation assumptions.","keywords":["mortar method","staggered discontinuous Galerkin","a posteriori error estimates","residual-type estimators","nonmatching grids","adaptive mesh refinement","second-order elliptic equations","saturation assumption"],"falsifier":"Run the mortar staggered DG scheme on a sequence of nonmatching meshes for a smooth solution and compute the quantity $\\min_{v\\in H^1_0(\\Omega)}\\|\\rho^{1/2}\\nabla(v-u_h)\\|_0$ divided by $(\\sum_{e\\in F_p\\cup T_{\\Gamma,h}} h_e^{-1}\\|\\llbracket\\rho^{1/2}u_h\\rrbracket\\|^2_{0,e})^{1/2}$; if this ratio grows unboundedly as the mesh is refined, Lemma 3.1 fails and the energy reliability bound collapses.","tokens_in":15727,"feed_emoji":"📐","tokens_out":4487,"duration_ms":49240,"temperature":0.7,"pith_summary":"The paper aims to supply computable a posteriori error control for the mortar staggered discontinuous Galerkin method, which solves second-order elliptic equations on subdomains whose meshes do not match at interfaces. It constructs two residual-type estimators: one for the potential error $\\|u-u_h\\|_0$ and one for the weighted energy error $\\|\\rho^{1/2}\\nabla(u-u_h)\\|_0$. It proves each estimator is both reliable and efficient, with constants independent of mesh size, and it avoids the saturation assumptions that are common in the literature. If correct, this gives practitioners a rigorous basis for adaptive mesh refinement driven by these estimators.","feed_headline":"Mortar staggered DG gets proven error estimators","feed_subtitle":"Residual-based estimators bound both L2 and energy errors with no saturation assumptions.","key_machinery":"The two local estimators are assembled from element residuals, flux jumps, and mortar interface mismatch terms. For the potential $L^2$ error, the estimator $\\eta_1$ contains $h_\\tau^4\\|f+\\nabla\\cdot z_h\\|^2_{0,\\tau}$, $h_\\tau^2\\|\\rho^{-1}z_h-\\nabla u_h\\|^2_{0,\\tau}$, jump terms on the staggered edges, and mortar flux difference terms; its proof uses the duality argument with an auxiliary elliptic problem. For the energy error, the estimator $\\eta_2$ is built similarly with $\\rho$-weighted residuals, and the proof introduces an auxiliary function $s\\in H^1_0(\\Omega)$ defined by $(\\rho\\nabla s,\\nabla v)=(\\rho\\nabla u_h,\\nabla v)$, which decomposes the energy error into a conforming part and a nonconforming part. The nonconforming part is controlled by the jumps of $u_h$ on the subdivision edges and non-mortar interfaces, and the conforming part is handled by the Scott-Zhang local quasi-interpolation operator. Efficiency is proved with standard element and edge bubble functions.","core_discovery":"For the mortar staggered discontinuous Galerkin discretization of $-\\nabla\\cdot(\\rho\\nabla u)=f$ with $u=0$ on $\\partial\\Omega$, the paper constructs two residual-type error estimators and proves that they are both reliable and efficient. Theorem 3.1 shows that the $L^2$-norm error $\\|u-u_h\\|_0$ is bounded by the estimator $\\eta_1$ using a duality argument, and Theorem 3.2 shows that the weighted energy error $\\|\\rho^{1/2}\\nabla(u-u_h)\\|_0$ is bounded by the estimator $\\eta_2$ using an auxiliary function that splits the energy error into a conforming and a nonconforming part. Theorem 4.1 gives matching lower bounds for both estimators. The analysis requires no saturation assumptions and no mesh restrictions on the mortar and non-mortar sides of the interfaces.","pith_inferences":["The same conforming/nonconforming decomposition via an auxiliary function could transfer to other mortar or hybridizable discontinuous Galerkin methods on nonmatching meshes, as long as a subdomain-wise conforming interpolation operator is available.","The absence of saturation assumptions suggests the estimators could be used as ingredients in a full convergence proof for an adaptive algorithm with Dörfler marking, though the paper itself does not establish contraction or optimal decay rates.","If the hidden constants in the Scott-Zhang interpolation estimates and the imported distance lemma are made explicit, the reliability bounds could be turned into guaranteed upper bounds on the error; the present paper only establishes existence of constants independent of mesh size."],"forward_implications":["Adaptive mesh refinement can be driven directly by $\\eta_1$ or $\\eta_2$; the numerical experiments show that optimal convergence rates are recovered for solutions with limited regularity.","Both estimators give explicit upper and lower bounds, so they can be used to balance over- and under-refinement without relying on the usual saturation assumption.","The energy estimator explicitly monitors the nonconforming error through jumps on subdivision edges and the non-mortar interface, which is important on nonmatching grids.","The potential $L^2$ estimator includes mortar flux difference terms, extending reliable a posteriori control to the $L^2$ norm for nonmatching meshes.","Removing the saturation assumption means the reliability proofs do not require comparing the computed solution to a finer-grid reference solution."],"supporting_citations":[{"why":"Defines the mortar staggered discontinuous Galerkin formulation and its discrete spaces, which are the object of the error analysis.","marker":"[24]"},{"why":"Supplies Lemma 3.6, the imported bound on the H1-conforming distance used in the nonconforming part of the energy error estimate.","marker":"[34]"},{"why":"Supplies Theorem 2.2, a second source for the nonconforming-distance bound used in Lemma 3.1.","marker":"[22]"},{"why":"Provides the Scott-Zhang local quasi-interpolation operator and the approximation estimates used in the energy-reliability proof.","marker":"[30]"},{"why":"Provides the trace and interpolation approximation properties used to bound the duality terms in the potential error estimate.","marker":"[18]"}],"fun_headline_variants":["No-saturation error bounds for mortar staggered DG","Two residual estimators prove reliable and efficient for mortar DG","L2 and energy error estimators for mortar staggered DG","Duality-backed error estimators for mortar staggered DG"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The energy-error bound rests on an imported lemma, stated as Lemma 3.1, which asserts that the distance from the numerical solution to a genuinely continuous function is controlled by the size of its jumps across the interior subdivision edges and across the non-mortar interface; the paper does not prove this lemma for the mortar staggered DG spaces.","fun_headline_variants_meta":{"raw":{"variants":["No-saturation error bounds for mortar staggered DG","Two residual estimators prove reliable and efficient for mortar DG","L2 and energy error estimators for mortar staggered DG","Duality-backed error estimators for mortar staggered DG"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000193,"raw_usage":{"total_tokens":1294,"prompt_tokens":832,"completion_tokens":462,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":448,"completion_tokens_details":{"reasoning_tokens":401}},"tokens_in":448,"tokens_out":462,"duration_ms":5102,"temperature":1.0,"reasoning_tokens":401,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:17:20.997676+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the mortar staggered DG scheme on a sequence of nonmatching meshes for a smooth solution and compute the quantity $\\min_{v\\in H^1_0(\\Omega)}\\|\\rho^{1/2}\\nabla(v-u_h)\\|_0$ divided by $(\\sum_{e\\in F_p\\cup T_{\\Gamma,h}} h_e^{-1}\\|\\llbracket\\rho^{1/2}u_h\\rrbracket\\|^2_{0,e})^{1/2}$; if this ratio grows unboundedly as the mesh is refined, Lemma 3.1 fails and the energy reliability bound collapses.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the mortar staggered discontinuous Galerkin formulation and its discrete spaces, which are the object of the error analysis."},{"cited_title":"W ANG AND X","cited_arxiv_id":null,"evidence_quote":"Supplies Lemma 3.6, the imported bound on the H1-conforming distance used in the nonconforming part of the energy error estimate."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies Theorem 2.2, a second source for the nonconforming-distance bound used in Lemma 3.1."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the Scott-Zhang local quasi-interpolation operator and the approximation estimates used in the energy-reliability proof."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the trace and interpolation approximation properties used to bound the duality terms in the potential error estimate."}],"review_version":1}