{"id":"74cce2ce-31f3-4e3d-8775-acd2e8ab1f65","arxiv_id":"1908.05801","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"The authors extend the factorization method to anisotropic periodic layers with two-sided near-field measurements and prove the existence of infinitely many transmission eigenvalues for the new quasi-periodic transmission eigenvalue problem.","lead":"This mathematics paper develops a shape-reconstruction algorithm, based on scattered waves, for periodic anisotropic layers, and proves that infinitely many special wave numbers, called transmission eigenvalues, exist for such layers. It is relevant to nondestructive testing of gratings and photonic crystals, where material fingerprints are needed.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Sign inconsistency in the central factorization: equations (7), (12), and (18) imply G=-ET, so WN=-H*TH rather than H*TH; the sign in the range-identity functional of Theorem 3.6 must be flipped.","rationale":"The reader's weakest assumption was the absorbing-material hypothesis in Assumption 3.1. My review identifies a different and more immediately load-bearing problem: the central factorization identity WN=H*TH appears to carry the wrong sign. This is not a dispute about scope or consensus; it is an internal algebraic inconsistency in the proof of Theorem 3.6. The computation is straightforward: B_A=B_I+∫Q∇u·∇v, (7) has a minus sign on the contrast term, and E is defined with the opposite sign, so G=−ET. Therefore WN=−H*TH unless the definition of T or the definition of E is changed. The range-identity step in Theorem 3.6 requires a positive operator (WN)_♯; with Lemma 3.5(b) the imaginary part of the correctly factorized WN has the opposite sign to what the formula |ReWN|−ImWN assumes. This is fixable by flipping a sign consistently, and the numerical examples may already use the correct convention in code, but as written the proof does not close. I do not find a comparable fatal gap in Theorem 4.3; the ball-transmission-eigenvalue identities check out once the boundary conditions are used. Because the sign error is concrete but likely reparable, the appropriate outcome remains a conditional acceptance pending correction, which matches the reader's verdict; hence UNCHANGED.","tokens_in":13050,"tokens_out":26654,"duration_ms":252866,"concrete_test":"Recompute the factorization from (7), (12), and (18) with f=H(a) for a single propagating mode in the configuration of Section 5 (e.g., k=5.85, absorbing Q). If the algebra or a numerical solve confirms B_I(u,v)=−∫(ReQ)^1/2 Tf·∇v (i.e., G=−ET), then evaluate the imaging functional (19) twice, once with (WN)_♯=|ReWN|−ImWN and once with |ReWN|+ImWN; only one will be positive semidefinite and reproduce the reconstructions in Figures 2–5. The theorem must be corrected to the sign that makes the operator positive and for which the range identity holds.","verdict_should_be":"UNCHANGED","load_bearing_attack":"With the sign conventions in (5), (7), (12), and (18), the factorization Lemma 3.4 has an algebraic sign error. Let B_A and B_I be the forms in (5) and (12). Since A=I+Q, B_A(u,v)=B_I(u,v)+∫_D Q∇u·∇v. For u=G(f), (7) gives B_A(u,v)=−∫_D Q(ReQ)^−1/2 f·∇v. Substituting Tf from (18), B_I(u,v)=−∫_D (ReQ)^1/2 Tf·∇v, so u is the negative of the solution E(Tf) defined by (15). Hence G=−ET, and because N=GH and H*=WE (up to the typo in (14), corrected by the H* formula), WN=−H*TH. Lemma 3.4 states WN=H*TH. The sign is not cosmetic: Lemma 3.5(b) asserts ImT≤0, which under the correct factorization makes Im(WN)≥0, whereas (WN)_♯=|ReWN|−ImWN in Theorem 3.6 is the formula appropriate to Im(WN)≤0. Unless T or ♯ is redefined with the opposite sign, the range identity used to prove (19) does not apply as written.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies two problems for a 2D anisotropic periodic layer in TM polarization: the inverse scattering problem of reconstructing the support D of the contrast from near-field data, and the interior transmission eigenvalue problem for non-absorbing anisotropic periodic layers. For the inverse problem, the authors propose a factorization method based on the two-sided near-field operator N and prove a characterization of D in terms of an eigensystem of a symmetrized operator (WN)_♯. They also provide numerical reconstructions for several layer geometries. For the transmission eigenvalue problem, they formulate a quasi-periodic interior transmission eigenvalue problem and prove the existence of infinitely many real transmission eigenvalues under the assumption Im Q = 0 and α ∉ Z.","tokens_in":13330,"tokens_out":36801,"duration_ms":319298,"significance":"If correct, the factorization result would provide a unique and computationally simple characterization of the shape of anisotropic periodic layers from near-field data, extending earlier half-space results to the full-space two-sided setting. The transmission eigenvalue existence result for quasi-periodic eigenfunctions is new and appears to be essentially sound. The paper is clearly written and the numerical experiments, despite the selection issue noted below, illustrate the method. However, the central factorization theorem in Section 3 contains a sign error that, as written, invalidates the range-identity argument, and key lemmas are stated without proof. The transmission eigenvalue part of the paper is the stronger contribution; the inverse part needs substantial correction before the claims can be accepted.","major_comments":[{"comment":"There is a sign error in the factorization. From (7), B_A(u,v) = -∫_D Q(ReQ)^{-1/2} f·∇v. Since A = I + Q, B_A(u,v) = B_I(u,v) + ∫_D Q∇u·∇v, and using (18) we obtain B_I(u,v) = -∫_D (ReQ)^{1/2} Tf·∇v. Therefore the unique solution of (7) satisfies Gf = -E Tf, not Gf = E Tf as claimed in the proof of Lemma 3.4. Consequently Lemma 3.4 should read WN = -H^* T H. This is load-bearing: with Im T ≤ 0 from Lemma 3.5(b), the identity WN = -H^* T H gives Im(WN) = -H^*(Im T)H ≥ 0, so the operator (WN)_♯ = |Re WN| - Im WN used in Theorem 3.6 need not be positive definite, and the range identity theorem [14] cannot be applied as written. The proof of Lemma 3.4 drops the minus sign when rewriting (7) as an identity for B_I; this must be corrected, and the definition of ♯ or T must be adjusted accordingly.","section":"§3.2, Lemma 3.4"},{"comment":"The Rayleigh coefficients of the quasi-periodic Green's function (16) are not those given in (17). For x2 > h, the coefficient of e^{iα_n x1 + iβ_n(x2-h)} in G is r+_n(z) = i/(4πβ_n) e^{-iα_n z1} e^{iβ_n(h-z2)}, and for x2 < -h it is r-_n(z) = i/(4πβ_n) e^{-iα_n z1} e^{iβ_n(z2+h)}. Equation (17) has the opposite sign in the z2-dependent phase, i.e., it gives the coefficients of the incoming (conjugate) Green's function. Since Lemma 3.3 and the test functional in Theorem 3.6 both use r_n(z), the characterization of D is not justified unless this formula is corrected and the proof of Lemma 3.3 is checked with the correct phase.","section":"§3.1, Eq. (17)"},{"comment":"These statements are load-bearing for the central inverse-scattering claim, yet they are not proved in the manuscript. Lemma 3.5 states that the proofs 'are omitted here, following [18]', and Theorem 3.6 states that 'the proof is similar to [18]'. The full-space two-sided problem has a different measurement operator, a different weighting operator W, and different boundary terms than the half-space problem in [18]; the sign issues identified above show that the adaptation is not automatic. Please provide complete proofs of Lemma 3.5 and Theorem 3.6, or a detailed step-by-step mapping to [18] that accounts for all sign conventions and the two-sided nature of the data.","section":"§3.2, Lemma 3.5 and Theorem 3.6"}],"minor_comments":[{"comment":"The matrix W in (14) does not match the formula for H^* in (11). To have H^* = W E, the first row of W should be ( ~w+_n, ~w+_n ), not ( ~w+_n, ~w-_n ). The determinant and invertibility argument are unaffected.","section":"§3.1, Eq. (14)"},{"comment":"In the proof of the z ∈ D direction, the displayed formula should read (Ef)_n = ( \\hatΦ^+_n, \\hatΦ^-_n )^T rather than ( \\hatΦ^+_n, \\hatΦ^+_n )^T.","section":"§3.1, Lemma 3.3"},{"comment":"Figures 2(d) and 3(d) are described as the best results out of 10 numerical experiments, and the worst cases are stated to show no reasonable reconstruction. As presented, this does not give a robust measure of the method's performance. Please report median or typical reconstructions and show at least one representative non-best case, or explain the selection criterion.","section":"§5, Numerical examples"},{"comment":"There are minor language and typographical issues, e.g., 'transmission eigenvalues in scattering have recently attracted' in the abstract and 'Lipchitz' in Section 6; these should be corrected.","section":"Abstract and §6"}],"recommendation":"major_revision","confidential_remarks":"The sign error in Lemma 3.4 appears to be a genuine algebraic slip rather than a harmless change of convention, and it affects the main inverse-scattering theorem. The r_n phase error in (17) compounds the problem. I would ask the authors to correct the signs, provide the missing proofs of Lemma 3.5 and Theorem 3.6, and re-run the numerical experiments with the corrected imaging functional. The transmission eigenvalue section seems largely sound and is a useful contribution; the revision should focus on the factorization part."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nHere's my read on arXiv:1908.05801. The paper does two things: adapts the factorization method to anisotropic periodic layers in full space with two-sided near-field data, and proves existence of infinitely many transmission eigenvalues for the quasi-periodic transmission eigenvalue problem. The first is a genuine extension of Nguyen's half-space work [18] and avoids the connectivity assumption in T.-P. Nguyen [19]. The second follows the Cakoni–Kirsch machinery but the formulation with quasi-periodic conditions is new. The numerical examples show the method works when it works, though the authors admit the reconstructions for the first two geometries are unstable and the displayed figures are best-of-10.\n\nNow the serious problem. The sign in the factorization is wrong. Let me walk through it. Equation (7) defines u via B(u,v;A) = -∫_D Q(ReQ)^-1/2 f·∇v. Since A=I+Q and Q is supported in D, B(u,v;A)=B(u,v;I)+∫_D Q∇u·∇v. Therefore B(u,v;I) = -∫_D Q((ReQ)^-1/2 f+∇u)·∇v. But T in (18) is defined so that (ReQ)^{1/2}Tf = Q((ReQ)^-1/2 f+∇u). Hence u solves B(u,v;I) = -∫_D (ReQ)^{1/2}Tf·∇v, while E in (15) is defined with the opposite sign: B(u,v;I)=+∫_D (ReQ)^{1/2}f·∇v. So G = -ET, not ET. Lemma 3.4's conclusion WN = H*TH should be WN = -H*TH. This is not cosmetic: Lemma 3.5(b) says ImT ≤ 0, so the correct factorization gives Im(WN) ≥ 0, and the range-identity functional in Theorem 3.6 should be |ReWN| + ImWN, not |ReWN| - ImWN. As written, the range identity from [14] does not apply. The error looks fixable by redefining T (or E) with a minus sign, but the proof of Theorem 3.6, as printed, doesn't hold.\n\nOther soft spots are minor by comparison. Lemma 3.5 is stated without proof (\"omitted here, following [18]\") and Theorem 3.6's proof is \"similar to [18]\"; that's acceptable for a subfield that knows the machinery, but it makes the paper less self-contained. The numerical section is honest about instability, but showing only the best of 10 runs is not a strong demonstration for the two unstable shapes.\n\nThe transmission eigenvalue part looks more solid: the operator L_k is self-adjoint, L_0 is coercive, L_k-L_0 compact, and the dimension-counting argument with shrinking balls gives infinitely many eigenvalues. I didn't spot an error there.\n\nWho is this for? Researchers working on factorization methods for periodic structures and on transmission eigenvalue problems for quasi-periodic settings. They should read it with a pencil. I'd send it to peer review—the sign error is real but likely fixable, and the contributions are worth having in the literature after correction. I would not cite it in its current form.","headline":"Worth a careful referee, but the central factorization has a sign error that propagates into the main characterization; fixable, but as written the proof of Theorem 3.6 does not go through.","tokens_in":13834,"tokens_out":6488,"would_cite":false,"duration_ms":53187,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35R30","78A46","65C20"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that a single series test over the eigensystem of the near-field operator decides membership in the support of an anisotropic periodic layer, and that the quasi-periodic transmission problem has infinitely many…","keywords":["inverse scattering","factorization method","transmission eigenvalues","anisotropic periodic structures","quasi-periodic layers","near-field data","shape reconstruction","TM polarization"],"falsifier":"Take a non-absorbing anisotropic periodic layer ($\\operatorname{Im} Q = 0$) and compute the near-field operator $N$; if the series in (19) converges at some point outside $D$ or diverges at some point inside $D$, the characterization fails. For the eigenvalue result, discretize $L_k$ for a layer with $A$ a constant positive-definite multiple of the identity: the theorem predicts infinitely many real eigenvalues, so observing only finitely many sign changes of $L_k$ would falsify Theorem 4.3.","tokens_in":12843,"feed_emoji":"🎯","tokens_out":14416,"duration_ms":122175,"temperature":0.7,"pith_summary":"This paper claims two things about anisotropic periodic layers in two dimensions. First, the support of the layer's contrast can be recovered from near-field scattering data by a pointwise series test: a point belongs to the support exactly when a certain sum over the eigensystem of the data operator converges. Second, the associated quasi-periodic interior transmission eigenvalue problem has infinitely many real eigenvalues. Both results matter for periodic structures that model photonic crystals and gratings, where knowing the layer's shape from scattering data is a nondestructive-testing task and transmission eigenvalues serve as material signatures. The paper supplies proofs for both claims and numerical examples for the shape-reconstruction method.","feed_headline":"One convergent series locates anisotropic scattering layers exactly","feed_subtitle":"Near-field data pin down the layer's shape; the proof also yields infinitely many transmission eigenvalues.","key_machinery":"The argument turns on the factorization $WN = H^* T H$, where $N$ is the near-field operator, $W$ is an explicit block weight matrix with a bounded inverse, $H$ is a compact injective map from sequence data into the layer, and $T$ is a contrast operator defined by $T f = (\\mathrm{Re}\\,Q)^{-1/2} Q((\\mathrm{Re}\\,Q)^{-1/2} f + \\nabla u)$. Under Assumption 3.1, $T$ is injective, has negative semidefinite imaginary part (strictly negative on nonzero vectors), and differs from a coercive operator by a compact perturbation; the range identity theorem then gives $\\operatorname{Range}(H^*) = \\operatorname{Range}((WN)_\\sharp^{1/2})$. The quasi-periodic Green function provides test vectors $r_n(z)$ whose Rayleigh coefficients are known explicitly, so membership of $z$ in $D$ is read off from the finiteness of the eigensystem sum. For the eigenvalue half, the key object is $L_k$, defined by a Riesz representation from a Hermitian sesquilinear form; its self-adjointness, coercivity at $k=0$, and compactness of $L_k - L_0$ allow the standard counting argument for transmission eigenvalues, and a family of disjoint balls inside $D$ produces the infinite-dimensional subspaces on which $L_k$ is non-positive.","core_discovery":"The central discovery is a complete characterization and a countable spectrum. Theorem 3.6 says that, under the absorbing-contrast Assumption 3.1, a point $z$ lies in the support $D$ exactly when the series $\\sum_j |\\langle r_n(z), \\psi_{n,j}\\rangle|_{\\ell^2(\\mathbb{Z})^2}^2 / \\lambda_j$ is finite, where $(\\lambda_j, \\psi_{n,j})$ is an orthonormal eigensystem of the self-adjoint positive operator $(WN)_\\sharp^{1/2}$ built from the measured near-field operator $N$. This is an if-and-only-if test, so it is both a uniqueness proof and a pixel-by-pixel imaging criterion. Theorem 4.3 states that, for a non-absorbing layer ($\\operatorname{Im} Q = 0$) with uniformly positive contrast and quasi-periodicity parameter $\\alpha$ not an integer, there exist infinitely many transmission eigenvalues. The proof uses an operator $L_k$ on $H^1_{0,\\alpha}(D)$ that is self-adjoint, coercive at $k=0$, and compactly different from $L_0$; positivity for small $k$ and non-positivity on an $M_\\epsilon$-dimensional subspace built from disjoint balls force the eigenvalue count to grow without bound. Together the two theorems give the first factorization-method justification and transmission-eigenvalue existence result for the anisotropic periodic-layer setting.","pith_inferences":["Because the proof of Lemma 3.5 is deferred to the half-space analogue, the self-containedness of the full-space argument rests on trusting that the half-space estimates carry over; a reader should verify the sign and coercivity estimates for the full-space Dirichlet-to-Neumann maps directly.","The paper reports stable reconstructions for ball- and cross-type layers but noise-sensitive ones for piecewise-linear and sinusoidal layers, and offers no justification; this suggests the imaging functional's stability depends on layer topology, which is a testable hypothesis.","The factorization structure $WN = H^* T H$ appears portable to full Maxwell biperiodic problems, where $W$ would encode polarization-dependent Dirichlet-to-Neumann data; the paper does not attempt this.","The disjoint-ball counting argument for transmission eigenvalues hints that the number of eigenvalues grows with the volume of $D$; a Weyl-type asymptotic is a natural next step."],"forward_implications":["The series test is a yes/no criterion for every point, so the layer's support can be imaged pixel by pixel from the measured near-field operator without knowing the contrast's amplitude.","Because the criterion is an equivalence, the near-field operator determines the support $D$ uniquely among layers satisfying Assumption 3.1.","The proof requires near-field data taken from both sides of the layer, and the numerical examples show that evanescent modes are needed for resolution.","The infinite sequence of real transmission eigenvalues gives countable material-dependent fingerprints that are in principle recoverable from scattering data.","The absorbing assumption behind the factorization and the transparency assumption behind the eigenvalue result delimit two complementary regimes: one for imaging, one for material probing."],"supporting_citations":[{"why":"Supplies the range identity theorem that converts the factorization into the range equality used for the pointwise test.","marker":"[14]"},{"why":"Provides the half-space anisotropic grating analysis whose proofs are adapted for the full-space factorization and final theorem.","marker":"[18]"},{"why":"Establishes well-posedness of the direct scattering problem and defines the exterior Dirichlet-to-Neumann operators used in the variational form.","marker":"[4]"},{"why":"Supplies the quasi-periodic Green function expansion and the trigonometric Galerkin method used to generate the numerical scattering data.","marker":"[17]"},{"why":"Gives the operator-theoretic framework (self-adjoint $L_k$, coercivity at zero, compact difference) for proving existence of transmission eigenvalues.","marker":"[8]"},{"why":"Provides the periodic-media transmission-eigenvalue theory that the authors follow for the quasi-periodic existence proof.","marker":"[7]"}],"fun_headline_variants":["A single series test pinpoints anisotropic layer shapes exactly","Infinite transmission eigenvalues proven for anisotropic periodic layers","Exact layer support from a convergent series in near-field data","First factorization-method proof for periodic anisotropic layers"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The inverse-scattering result collapses if the contrast is not absorbing: the proof needs $\\operatorname{Im} Q$ strictly negative definite so that $T$ is injective and the range identity theorem applies, and the paper explicitly notes this assumption excludes transmission eigenvalues.","fun_headline_variants_meta":{"raw":{"variants":["A single series test pinpoints anisotropic layer shapes exactly","Infinite transmission eigenvalues proven for anisotropic periodic layers","Exact layer support from a convergent series in near-field data","First factorization-method proof for periodic anisotropic layers"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001285,"raw_usage":{"total_tokens":5257,"prompt_tokens":957,"completion_tokens":4300,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":573,"completion_tokens_details":{"reasoning_tokens":4237}},"tokens_in":573,"tokens_out":4300,"duration_ms":28179,"temperature":1.0,"reasoning_tokens":4237,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:04:56.292364+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a non-absorbing anisotropic periodic layer ($\\operatorname{Im} Q = 0$) and compute the near-field operator $N$; if the series in (19) converges at some point outside $D$ or diverges at some point inside $D$, the characterization fails. For the eigenvalue result, discretize $L_k$ for a layer with $A$ a constant positive-definite multiple of the identity: the theorem predicts infinitely many real eigenvalues, so observing only finitely many sign changes of $L_k$ would falsify Theorem 4.3.","supporting_citations":[{"cited_title":"Kirsch and N","cited_arxiv_id":null,"evidence_quote":"Supplies the range identity theorem that converts the factorization into the range equality used for the pointwise test."},{"cited_title":"Nguyen , Shape identification of anisotropic diffraction gratings for TM- polarized electromagnetic waves, Appl","cited_arxiv_id":null,"evidence_quote":"Provides the half-space anisotropic grating analysis whose proofs are adapted for the full-space factorization and final theorem."},{"cited_title":"Bonnet-Bendhia and F","cited_arxiv_id":null,"evidence_quote":"Establishes well-posedness of the direct scattering problem and defines the exterior Dirichlet-to-Neumann operators used in the variational form."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the quasi-periodic Green function expansion and the trigonometric Galerkin method used to generate the numerical scattering data."},{"cited_title":"Cakoni and A","cited_arxiv_id":null,"evidence_quote":"Gives the operator-theoretic framework (self-adjoint $L_k$, coercivity at zero, compact difference) for proving existence of transmission eigenvalues."},{"cited_title":"Cakoni, H","cited_arxiv_id":null,"evidence_quote":"Provides the periodic-media transmission-eigenvalue theory that the authors follow for the quasi-periodic existence proof."}],"review_version":1}