{"id":"48ec1ef4-84f2-4962-b81c-5e34819f78c9","arxiv_id":"2507.16647","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"high","formal_verification":"none","parameter_count":3,"one_line_summary":"A quasi-Monte Carlo plus localized multiscale method for the random Helmholtz equation is proposed, but the proof of its central wavenumber-explicit error estimate has serious gaps for d=2.","lead":"This paper combines a multiscale finite element method with quasi-Monte Carlo sampling to solve the Helmholtz wave equation in random media, reporting fast convergence in both space and randomness. The main advertised result, a wavenumber-explicit error bound with O(H^4) accuracy, rests on a stability proof that breaks down in two dimensions and on an unproven adjoint regularity step.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 3.2's boundary estimate (3.5) has the wrong sign relative to the definition (2.5), so the O(H^4) L²-superconvergence and the H²/V-error bounds are not proven as written.","rationale":"The reader's flagged assumptions (the ε0 parameter issue in Prop. 2.1 and the adjoint identity in Lemma 3.2) are both real problems. The ε0 issue is dimension-specific and potentially repairable by treating d=1,2 separately, so I do not regard it as the single most load-bearing gap. The adjoint identity is in the same Lemma 3.2 block, where I find an even more fundamental sign error: (3.5) contradicts (2.5) before the duality step is reached. This invalidates the boundary estimate at the start of Lemma 3.2, so the H² and O(H^4) L² superconvergence claims are unsupported as written. The qMC and truncation portions are standard conditional on wavenumber-explicit stability, but the spatial-error proof is internally inconsistent. Because the reader's REJECT verdict already reflects a high correctness risk, these findings do not change the verdict; they sharpen the basis for it.","tokens_in":23113,"tokens_out":20747,"duration_ms":211365,"concrete_test":"Take an arbitrary non-zero e_H and compute ℑ a(e_H,e_H) directly from (2.5). If it equals -κ ∫_{∂D} √n |e_H|² ds, then (3.5) is contradicted; if the authors intend a different sign for the Robin term, the definition (2.5), the boundary condition (2.2), and (3.5) must be made mutually consistent before Lemma 3.2's estimates can be used.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing, elementary defect is in the first estimate of Lemma 3.2. Equation (3.5) asserts κ√n1 ||e_H||²_{L²(∂D)} ≤ ℑ a(e_H,e_H). Using (2.5) with v=u=e_H gives ℑ a(e_H,e_H) = -κ ∫_{∂D} √n |e_H|² ds, which is negative for positive n. A negative number cannot be ≥ the positive left-hand side, so (3.5) is false. The subsequent chain—(3.6), the Nitsche duality estimate, and the claimed O(H^4) L² and H²/V-error bounds—all use this boundary control. Since Lemma 3.2 is the only in-paper derivation of the multiscale spatial error (Lemma 3.3 is outsourced), the advertised superconvergence and the H² term in Theorem 6.1 are not proven as written. This is not a missing constant or a parameter choice; it is an algebraic sign inconsistency with the paper's own variational formulation.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper proposes a quasi-Monte Carlo (qMC) multiscale method for the Helmholtz equation in a bounded domain with a random refractive index parameterized by an infinite series. The method combines dimension truncation, a localized orthogonal decomposition (LOD)-type multiscale discretization with a boundary corrector for the complex Robin condition, and qMC sampling. The main result, Theorem 6.1, asserts a combined error bound of order H^2 (with an L^2 superconvergence O(H^4)) in the physical space plus beta^ell localization error plus N^{-alpha} qMC error plus a dimension truncation term s^{1-2/p}. Numerical experiments in 1D and 2D are presented for the deterministic and random cases.","tokens_in":23400,"tokens_out":9028,"duration_ms":85785,"significance":"The proposed algorithm is a natural and potentially useful combination of known techniques; if the advertised rates were proven, the method would offer pollution-free, high-order UQ for random Helmholtz problems. The paper explicitly provides wavenumber-dependent constants and supports the theory with several numerical examples, including a heterogeneous test showing the boundary corrector's benefit. However, the central spatial-error analysis is not valid as written, so the main theorem is currently unsupported.","major_comments":[{"comment":"The proof chooses epsilon_0 = (d-2)/(2 C_0 mu kappa^2), which vanishes for d=2 and is negative for d=1, even though the paper states d=1,2,3 throughout. Cauchy's inequality requires epsilon_0 > 0, so the stability estimates (2.9)-(2.10) are not established for the dimensions used in the numerical experiments (1D and 2D).","section":"Section 2.3, Proposition 2.1"},{"comment":"From the definition (2.5), Im a(e_H,e_H) = -kappa int_{partial D} sqrt(n) |e_H|^2 ds, which is negative for n>0. Hence the claimed inequality kappa sqrt(n_1) ||e_H||^2_{L^2(partial D)} <= Im a(e_H,e_H) is false. In addition, the equality Im a(e_H,e_H) = Im f(e_H) is not justified: Galerkin orthogonality only gives a(e_H,v_H)=0 for v_H in Psi_H, and the sesquilinear form is not symmetric, so a(e_H,e_H) is not equal to f(e_H). Consequently the boundary estimate, the subsequent bound (3.6), and the resulting O(H^4) L^2-error and H^2/V-error claims in Lemma 3.2 are not proven.","section":"Section 3.2, Lemma 3.2, Eq. (3.5)"},{"comment":"The adjoint problem stated with grad w . nu = i kappa sqrt(n) w does not satisfy the duality identity a(v,w) = (v,rho). Integrating by parts for a(v,w) with the stated boundary condition on w yields an extra boundary contribution, so the identity used to introduce the adjoint solution is incorrect. The Nitsche argument that produces the L^2-error estimate therefore lacks a valid starting point.","section":"Lemma 3.2, Nitsche duality, Eq. (3.7)"}],"minor_comments":[{"comment":"The global functional a_hat in (3.3) contains +kappa int_{partial D} sqrt(n) (phi_R^2 - phi_I^2) ds, while the localized functional a_hat_ell in (3.8) contains -kappa int_{partial D cap D_ell} sqrt(n) (phi_R^2 - phi_I^2) ds; the sign change is not explained.","section":"Section 3.2.1 vs 3.2.2"},{"comment":"The notation int_D nabla u_s nabla v dx omits the conjugate on v; for consistency with (2.5) it should be nabla u_s . nabla bar(v).","section":"Section 5, Eq. (5.3)"},{"comment":"The localization error estimate is stated without proof; given the nonstandard Robin boundary condition and complex-valued basis, a short justification or a precise reference to the version in [41,45] would be useful.","section":"Section 3.2.2, Lemma 3.3"},{"comment":"The sentence 'Here the optimal convergence rates are O(H^4) for 1D and O(H^2), respectively' is ambiguous about which curve corresponds to which rate; the caption of Fig. 4 should be more explicit.","section":"Section 7.1, Example 7.4"},{"comment":"There are minor reference formatting issues (duplicated URLs in [1] and [3], incomplete entries); these do not affect the content.","section":"References"}],"recommendation":"reject","confidential_remarks":"The paper's new technical content is concentrated in the wavenumber-explicit stability analysis and the multiscale spatial error analysis. Both contain load-bearing defects (the dimension-dependent Cauchy parameter and the sign/duality errors in Lemma 3.2). The qMC and truncation components are largely imported from [19] and related work, so the manuscript cannot be accepted without a substantially revised analysis. I would not encourage resubmission in its present form unless the authors can provide a corrected proof of the spatial estimates."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the paper combines localized orthogonal decomposition and quasi-Monte Carlo for the Helmholtz problem with random refractive index and Robin boundary conditions. That combination is new and the numerical experiments are clean. But the main theorem, Theorem 6.1, is not proven as written. The first estimate in Lemma 3.2 has a sign error: with the sesquilinear form (2.5), ℑ a(e_H,e_H) = -κ ∫_{∂D} √n |e_H|² ds, which is negative, so the claimed lower bound κ√n1 ||e_H||²_{L²(∂D)} ≤ ℑ a(e_H,e_H) is false. Everything downstream—the L² superconvergence and the V-error bound—rests on that inequality. The adjoint boundary condition in the Nitsche duality is also stated incorrectly: for a complex inner product the adjoint Robin condition should be ∇w·ν = -iκ√n w, not +iκ√n w. And in Proposition 2.1, the Cauchy parameter ε0=(d-2)/(2C0 μ κ²) is zero in 2D and negative in 1D, which breaks the stability argument for d=1,2 even though the results claim d=1,2,3.\n\nThat said, the paper has real merits. The idea of building the randomness into the multiscale basis and using a boundary corrector for the random Robin condition is a legitimate and useful extension of the LOD program. The qMC analysis follows the template of Ganesh–Kuo–Sloan but is applied to a different spatial discretization, and the regularity estimates in Section 4 are carefully derived. The numerical experiments support the advertised convergence rates, so the method itself may well work.\n\nThe problems are not cosmetic. They hit the central claim. I would not accept the paper in its current form. But I would send it to a referee, because the combination is novel, the numerics are strong, and the errors in the proof look fixable—if the authors can rework Lemma 3.2 and the stability argument. The worst outcome would be to desk-reject a paper whose core idea is sound just because the proof needs serious repair.\n\nRecommendation: engage with it. A rigorous referee should be able to determine whether the sign error can be repaired without changing the method. If it can, the paper could be a solid contribution.","headline":"Promising LOD+qMC combination for random Helmholtz, but the central error estimate has a sign error that invalidates Theorem 6.1 as written.","tokens_in":23908,"tokens_out":3657,"would_cite":false,"duration_ms":37017,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["65N30","65N12","65N15","65C05"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that a boundary-corrected multiscale coarse-space discretization combined with quasi-Monte Carlo sampling solves the random Helmholtz equation with a total error bound $C(H^2 + \\beta^\\ell + N^{-\\alpha} +…","keywords":["uncertainty quantification","Helmholtz equation","random refractive index","quasi-Monte Carlo method","multiscale method","localized orthogonal decomposition","superconvergence","Robin boundary condition"],"falsifier":"Set $D=[0,1]^2$, $n=1$, $\\kappa=1$, choose complex polynomials $v,w$, and check numerically whether the stated adjoint Robin condition $\\nabla w \\cdot \\nu = i\\kappa\\sqrt{n}\\, w$ makes $a(v,w)$ equal to $(v,\\rho)$; if the boundary terms do not cancel, the duality identity behind the claimed $O(H^4)$ $L^2$ superconvergence fails in the exact setting where the paper predicts it.","tokens_in":22895,"feed_emoji":"🌊","tokens_out":12419,"duration_ms":114064,"temperature":0.7,"pith_summary":"The paper proposes a numerical method for the Helmholtz equation in a bounded domain whose refractive index is random, modeled as an infinite series of stochastic variables. The authors claim that their boundary-corrected multiscale discretization, when combined with quasi-Monte Carlo (qMC) sampling, achieves $O(H^4)$ accuracy in the $L^2$ norm, $O(H^2)$ in a wavenumber-weighted norm, and almost first-order convergence in the random variables, while removing the 'pollution' effect that normally ties mesh size to wavenumber. The central result is a wavenumber-explicit total error bound (Theorem 6.1) that combines spatial discretization, localization, dimension truncation, and qMC quadrature errors. If accurate, the method offers a practical route to uncertainty quantification for wave propagation in random media with far fewer sampling points than plain Monte Carlo.","feed_headline":"Random Helmholtz solver hits O(H^4) accuracy","feed_subtitle":"Boundary-corrected multiscale basis plus quasi-Monte Carlo sampling removes pollution and needs far fewer samples.","key_machinery":"The mechanism is a localized multiscale basis built by solving patch-wise optimal problems: for each coarse node $x_j$, the complex-valued basis $\\phi_j$ minimizes the modified energy $\\hat{a}(\\omega; \\phi, \\phi) = \\int_D |\\nabla \\phi|^2 - \\kappa^2 n |\\phi|^2 + \\kappa \\int_{\\partial D} \\sqrt{n}(\\Re \\phi \\Re v - \\Im \\phi \\Im v)\\, ds$ subject to a quasi-interpolation constraint, which builds the Robin boundary condition into the basis. Around this basis the paper layers a weighted quasi-interpolation operator, a duality argument for the $L^2$ superconvergence, and a dimension-truncation plus randomly shifted lattice rule estimate that converts parametric regularity into an almost first-order qMC rate.","core_discovery":"The paper's central claim is that the fully discrete solution obtained by truncating the random refractive index to $s$ terms, discretizing with a localized boundary-corrected multiscale space on a coarse mesh of size $H$, and sampling with a randomly shifted lattice qMC rule of $N$ points satisfies the total error bound $\\sqrt{\\mathbb{E}[|I(G(u))-Q_{s,N}(G(u^s_{H,\\ell}))|^2]} \\le C(H^2 + \\beta^\\ell + N^{-\\alpha} + (1+\\kappa)s^{1-2/p})$ with $\\alpha = \\min(1/p - 1/2, 1-\\delta)$. From this bound the authors derive $O(H^4)$ $L^2$ superconvergence, $O(H^2)$ convergence in the wavenumber-weighted $V$-norm, near-first-order convergence in the stochastic variables, and the elimination of the pollution effect, all with constants independent of $\\kappa$, $s$, $N$ and $H$.","pith_inferences":["Our inference: because $\\epsilon_0 = (d-2)/(2 C_0 \\mu \\kappa^2)$ is nonpositive in $d=1,2$, the wavenumber-explicit stability proof as written covers only $d=3$; the $d=1,2$ superconvergence claims should be read as numerically demonstrated but not yet proven by this argument.","Our inference: if the duality identity behind the $O(H^4)$ $L^2$ estimate fails for complex-valued test functions, that particular estimate would need an alternative proof, while the $V$-norm and qMC rates might still stand on the remaining arguments.","Our inference: the same boundary-corrected multiscale construction could be applied to impedance or scattering problems with stochastic boundary data, and the predicted superconvergence rate would be a testable extension beyond random volume coefficients.","Our inference: the total error bound implies a computational tradeoff between the oversampling size $\\ell$ (which shrinks $\\beta^\\ell$ but enlarges the constant) and the qMC sample size $N$; choosing $\\ell$ adaptively per wavenumber could lower the overall cost."],"forward_implications":["Under the claimed bound, the $L^2$ error decays as $H^4$ even on coarse meshes, so accurate wave statistics can be computed with far fewer degrees of freedom than standard finite elements.","The pollution effect is eliminated: the mesh-size condition is only $H\\kappa \\lesssim 1$ rather than resolving every wavelength.","qMC sampling gives almost first-order convergence in the stochastic dimension, so the number of required realizations grows far more slowly than with plain Monte Carlo.","The dimension truncation error decays as $s^{1-2/p}$ for $p\\in(0,1)$, making the method practical for problems with slowly decaying random-series expansions.","The boundary corrector is essential for the random Robin condition; without it the relative error in the heterogeneous test grows by roughly a factor of fifty."],"supporting_citations":[{"why":"Supplies the localized orthogonal decomposition framework that eliminates the pollution effect and provides the localization estimate used in Lemma 3.3.","marker":"[41]"},{"why":"Provides the super-localization strategy and the heterogeneous two-dimensional benchmark whose reported error the boundary-corrected method matches in Example 7.3.","marker":"[16]"},{"why":"Supplies the quasi-Monte Carlo dimension-truncation and lattice-rule error framework for wave propagation in heterogeneous random media, including the weight construction used in Lemma 5.2.","marker":"[19]"},{"why":"Provides the dimension-truncation estimate for series expansions, equation (5.4), that yields the $s^{1-2/p}$ truncation error.","marker":"[32]"},{"why":"Gives the component-by-component construction and root-mean-square error bound for randomly shifted lattice rules used for the qMC convergence rate.","marker":"[10]"},{"why":"Supplies the two vector identities used repeatedly in the wavenumber-explicit stability and regularity proofs.","marker":"[7]"},{"why":"Provides the complex-valued operator-adaptive multiscale basis construction for Helmholtz problems with nontrivial boundary conditions.","marker":"[24]"},{"why":"Establishes the series-parameterized random refractive index model that the paper truncates and discretizes.","marker":"[12]"}],"fun_headline_variants":["qMC multiscale method hits O(H^4) for random Helmholtz","Superconvergent solver removes pollution in random media","Random Helmholtz: boundary-corrected multiscale beats pollution","Near-first-order random convergence with O(H^2) V-error","Pollution-free Helmholtz solver for random refractive index"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Two premises carry the argument: the wavenumber-explicit stability proof needs $\\epsilon_0 = (d-2)/(2 C_0 \\mu \\kappa^2)$ to be positive, which fails in dimensions $d=1$ and $d=2$, and the duality identity $a(v,w) = (v,\\rho)$ used for the $O(H^4)$ $L^2$ estimate has boundary terms that do not cancel for complex-valued $v$ under the stated Robin adjoint condition.","fun_headline_variants_meta":{"raw":{"variants":["qMC multiscale method hits O(H^4) for random Helmholtz","Superconvergent solver removes pollution in random media","Random Helmholtz: boundary-corrected multiscale beats pollution","Near-first-order random convergence with O(H^2) V-error","Pollution-free Helmholtz solver for random refractive index"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000202,"raw_usage":{"total_tokens":1379,"prompt_tokens":941,"completion_tokens":438,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":557,"completion_tokens_details":{"reasoning_tokens":350}},"tokens_in":557,"tokens_out":438,"duration_ms":5026,"temperature":1.0,"reasoning_tokens":350,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T15:06:35.315580+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Set $D=[0,1]^2$, $n=1$, $\\kappa=1$, choose complex polynomials $v,w$, and check numerically whether the stated adjoint Robin condition $\\nabla w \\cdot \\nu = i\\kappa\\sqrt{n}\\, w$ makes $a(v,w)$ equal to $(v,\\rho)$; if the boundary terms do not cancel, the duality identity behind the claimed $O(H^4)$ $L^2$ superconvergence fails in the exact setting where the paper predicts it.","supporting_citations":[{"cited_title":"Peterseim , Eliminating the pollution effect in Helmholtz problems by local subscale correction , Mathematics of Computation, 86 (2017), pp","cited_arxiv_id":null,"evidence_quote":"Supplies the localized orthogonal decomposition framework that eliminates the pollution effect and provides the localization estimate used in Lemma 3.3."},{"cited_title":"Freese, M","cited_arxiv_id":null,"evidence_quote":"Provides the super-localization strategy and the heterogeneous two-dimensional benchmark whose reported error the boundary-corrected method matches in Example 7.3."},{"cited_title":"Ganesh, F","cited_arxiv_id":null,"evidence_quote":"Supplies the quasi-Monte Carlo dimension-truncation and lattice-rule error framework for wave propagation in heterogeneous random media, including the weight construction used in Lemma 5.2."},{"cited_title":"Cummings and X","cited_arxiv_id":null,"evidence_quote":"Supplies the two vector identities used repeatedly in the wavenumber-explicit stability and regularity proofs."},{"cited_title":"Hauck and D","cited_arxiv_id":null,"evidence_quote":"Provides the complex-valued operator-adaptive multiscale basis construction for Helmholtz problems with nontrivial boundary conditions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the series-parameterized random refractive index model that the paper truncates and discretizes."}],"review_version":1}