{"id":"cdbd885c-c550-4724-8e6a-6084dc679104","arxiv_id":"1908.06455","paper_version":3,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Steklov eigenvalues of curvilinear polygons are asymptotically equal to explicit quasi-eigenvalues built from side lengths and angles, with errors tending to zero.","lead":"This paper derives sharp asymptotic formulas for the Steklov eigenvalues of curvilinear polygons, showing they are determined by side lengths and corner angles up to errors that vanish in the high-frequency limit. It links the problem to quantum graphs and reveals that whether an angle is an even or odd part of pi/2 determines how eigenfunctions concentrate on the boundary.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Exceptional-case enumeration (Theorem 5.31) is explicitly sketched; the omitted total-loss computation in Proposition 5.32 is the load-bearing gap for Theorem 1.4.","rationale":"Theorem 1.4 is a statement about the full ordered sequences {λ_m} and {σ_m}. The quasimode construction alone only gives σ_m near some λ_{i_m}; the equality of indices i_m = m is supplied by the enumeration arguments of Section 5. For non-exceptional polygons this is done carefully (Theorems 5.20 and Corollary 5.18). For exceptional polygons the analogous statement is Theorem 5.31, and its proof is explicitly left as a sketch, with the base case Proposition 5.32 reduced to an unshown 'total loss' calculation. This is the point at which the exceptional-angle arithmetic (parities O(α)) enters the index matching. A mistake here would not be caught by the eigenvalue approximation estimates, because those only control |σ_m − λ_{i_m}|, not i_m = m. The numerical section confirms formulas for equilateral and right-isosceles triangles, which are exceptional but highly symmetric; in those cases large multiplicities make the ordered sequence less sensitive to small enumeration shifts. A generic exceptional polygon with incommensurable side lengths is the natural test. I therefore recommend conditional acceptance: the stated results are likely true and the overall structure is convincing, but the central claim is not fully supported until the omitted total-loss computation is supplied.","tokens_in":82810,"tokens_out":24843,"duration_ms":251398,"concrete_test":"Complete the base-case verification of Proposition 5.32. For each parity pair (odd/odd, odd/even, even/even) of the two exceptional angles of a straight two-sided zigzag, compute the quasi-eigenvalue counting function from Definition 5.28 by evaluating the lifted equation (5.62) at σ=0 and at large σ, and compare it with the counting function of the mixed Steklov problem after symmetrisation (as in the NN/DD cases of Proposition 5.10) or after the isospectral transplantation of Lemma 5.12 for ND/DN. If any parity case disagrees by a non-integer or by an integer different from the claimed total loss, Theorem 5.31 falls and Theorem 1.4 for exceptional polygons is not proved.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The load-bearing gap is in the enumeration theorem for exceptional zigzags. Theorem 1.4 for any polygon with an exceptional angle requires Theorem 5.35, and Theorem 5.35 is derived from Theorem 5.31. But Section 5.7 explicitly does not prove Theorem 5.31: after Definition 5.30 the text says the theorem is 'proved similarly to Theorem 2.39' and 'we outline the main steps ... leave the details to the reader.' The base case Proposition 5.32 (two equal straight sides with an exceptional angle) is dismissed by saying the result follows by 'explicitly computing the total loss of quasi-eigenvalues' using [LPPS17], but no computation is shown. The subsequent gluing propositions 5.33 and 5.34 are labelled 'straightforward adaptation.' The natural enumeration in Definition 5.30 assigns half-integer counting functions to odd-parity exceptional components and to endpoint components; a parity or sign error in the omitted total-loss computation would produce a wrong constant or a wrong asymptotically varying shift between the quasi-eigenvalues and the true Steklov eigenvalues. Since Theorem 5.35 transfers exactly this enumeration to exceptional polygons, the central claim Theorem 1.4 is not fully established for exceptional curvilinear polygons. The numerical examples in Section 9 test mostly symmetric cases such as the right isosceles triangle, where large multiplicities make index shifts less sensitive, not a generic exceptional polygon with mixed parities and incommensurable side lengths.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the asymptotic distribution of Steklov eigenvalues and eigenfunctions on curvilinear polygons, i.e., domains with smooth sides and corners of angle strictly between 0 and π. The main result (Theorem 1.4) asserts that the Steklov eigenvalues λ_m can be approximated, up to an error O(m^{-ε}), by a sequence of quasi-eigenvalues σ_m defined purely in terms of the side lengths and angles via vertex and side transfer matrices. The paper also proves a companion statement (Theorem 1.7) on the trigonometric form of boundary traces of eigenfunctions, and describes qualitatively different behaviour for polygons all of whose angles are 'special' (equidistribution) versus all 'exceptional' (concentration on a subset of sides). The proofs combine scattering Peters solutions in sectors, a quasimode construction, a delicate enumeration argument for quasi-eigenvalues, layer potential estimates for curved boundaries, and numerical illustrations.","tokens_in":83144,"tokens_out":3124,"duration_ms":34865,"significance":"If the results are fully established, this is a major contribution to the spectral geometry of non-smooth domains. The paper gives, for the first time, asymptotics with error tending to zero for Steklov eigenvalues of polygons, and reveals a surprising dependence on arithmetic properties of the angles. The quasi-eigenvalues are defined entirely from the geometry, with no reference to the actual Steklov spectrum, so the argument is not circular. The paper also provides a clean quantum-graph interpretation and, in Section 9, numerical benchmarks that confirm the formulas in concrete examples, including the equilateral triangle, the right isosceles triangle, and regular polygons. The proofs are detailed and mostly self-contained; the exceptional-angle enumeration is the one part that needs further support.","major_comments":[{"comment":"The proof of the exceptional-case enumeration theorem is incomplete in a way that affects the central claim. After Definition 5.30, the text states that Theorem 5.31 is proved 'similarly to Theorem 2.39' and says 'we outline the main steps ... leave the details to the reader.' Proposition 5.32, which is the base case for two equal straight sides with an exceptional angle, is dismissed with the sentence that the result follows 'by explicitly computing the total loss of quasi-eigenvalues' using [LPPS17]; however, no computation is shown. Since Theorem 5.35 (exceptional polygons) and hence Theorem 1.4 for any polygon with an exceptional angle rely on this enumeration, a parity or sign error in the omitted total-loss computation would change the constant or the index shift between σ_m and λ_m. Please provide the full computation for Proposition 5.32, or at least an explicit derivation of the total-loss formula and its comparison with Definition 5.30.","section":"5.7, Theorem 5.31 and Proposition 5.32"},{"comment":"The gluing propositions for exceptional components are introduced with 'a straightforward adaptation of the proof of Proposition 5.13', but no proof or even a precise statement of the required adaptation is given. These propositions are load-bearing for Theorem 5.31: they propagate the natural enumeration from the basic two-sided case to arbitrary exceptional zigzags, and the half-integer counting functions in Definition 5.30 make the correctness of this propagation non-obvious. Please either give the detailed proofs of Propositions 5.33 and 5.34 or state them as lemmas with the relevant analogue of Lemma 5.13 and its proof.","section":"5.7, Propositions 5.33 and 5.34"}],"minor_comments":[{"comment":"The proof of Proposition 2.41 is omitted with the note that it is 'almost identical' to that of Proposition 2.27. Since this proposition is used later in the proof of Theorem 4.30 for zigzag domains, a brief indication of the differences (e.g., the treatment of the endpoint conditions) would help the reader verify the result without reconstructing the whole argument.","section":"2.7, Proposition 2.41"},{"comment":"The numerical tests in Section 9 mostly concern symmetric cases such as the right isosceles triangle or regular polygons, where high multiplicities make index shifts less sensitive. Adding a numerical example with mixed-parity exceptional angles and incommensurable side lengths, where the half-integer shifts in Definition 5.30 are individually observable, would provide additional confidence in the exceptional-case enumeration.","section":"9, numerical examples"},{"comment":"The notation in Theorem 2.31 uses both ε0 (from Theorem 1.4) and a parameter written as '~ε'; the relation between the two is stated but could be made clearer by explicitly defining the admissible range for the exponent in the error term.","section":"2.6, Theorem 2.31"}],"recommendation":"major_revision","confidential_remarks":"The paper is substantial and the main ideas are convincing, but the exceptional-case enumeration (Section 5.7) contains a genuine gap: the base case Proposition 5.32 and the gluing propositions 5.33/5.34 are asserted without proof. Since Theorem 1.4 covers exceptional polygons, this gap must be closed before the paper can be accepted. If the authors provide the missing computation and details, I would support publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper delivers what it promises: for curvilinear polygons, the Steklov eigenvalues are approximated by explicit quasi-eigenvalues with error going to zero, not just o(m). The transfer-matrix formalism, the scattering Peters solutions, and the quantum graph reinterpretation are new and give a real computational handle. The angle-arithmetic distinction between exceptional and special angles, with its striking effect on eigenfunction localization, is clearly worked out and supported by numerics. This is a major step beyond the previous first-order Weyl law, and the main construction appears sound. The paper is honest about its debts to LPPS17, and the reliance is not circular.\n\nNow the soft spot, and it is real. Section 5.7, which handles exceptional polygons, does not actually prove its central enumeration theorem. Theorem 5.31 is dismissed as \"proved similarly to Theorem 2.39,\" with only an outline and the details left to the reader. More concretely, the base case Proposition 5.32 concludes by saying the result follows from \"explicitly computing the total loss of quasi-eigenvalues\" using LPPS17, but no computation is shown. The subsequent gluing propositions 5.33 and 5.34 are called \"straightforward adaptations.\" This matters because Theorem 5.35 transfers this enumeration to all exceptional polygons, and Theorem 1.4 depends on it. A wrong half-integer shift or a parity error in the omitted total-loss computation would change the index m, invalidating lambda_m = sigma_m + O(m^{-epsilon}) for exactly the polygons the paper claims to cover. The numerics in Section 9 mostly test symmetric cases like the right isosceles triangle, where large multiplicities mask such shifts; a generic exceptional polygon with mixed parities and incommensurable side lengths is not tested.\n\nI want to be clear: this is a gap in the written proof, not necessarily a flaw in the result. The non-exceptional case is treated in full detail and looks solid. For exceptional polygons, the theorem should currently be read as conditional on the omitted computation being correct. A referee should ask the authors to supply the full proof of Theorem 5.31, especially the computation in Proposition 5.32 and the details of the gluing. If that can be done, the paper is a significant contribution; if not, the exceptional case remains an interesting conjecture supported by partial evidence.\n\nWho is this for? Spectral geometers, people working on Steklov problems, and anyone interested in quantum graph models of boundary value problems. It deserves a serious referee, but the referee should push for the missing proof rather than accept the sketch on faith.","headline":"A genuinely new sharp Steklov asymptotic for curvilinear polygons, but the exceptional-angle enumeration is sketched rather than proved, and that is exactly where a wrong index shift would break the main theorem.","tokens_in":83669,"tokens_out":1855,"would_cite":true,"duration_ms":22892,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35P20","34B45"],"pacs":[],"model":"deepseek-v4-flash","headline":"The Steklov eigenvalues of a curvilinear polygon are determined to within O(m^{-ε}) by an explicitly computable sequence of quasi-eigenvalues built from the side lengths and angles.","keywords":["Steklov eigenvalues","curvilinear polygons","quasi-eigenvalues","transfer matrices","scattering Peters solutions","quantum graphs","exceptional angles","spectral asymptotics"],"falsifier":"Compute numerically the first several hundred Steklov eigenvalues of a curvilinear triangle with angles (π/3, π/4, 5π/12) and one curved side, then compare them with the roots of the trigonometric polynomial F_P(α,ℓ,σ); if the gap between a true eigenvalue and the nearest quasi-eigenvalue does not shrink like a negative power of m, Theorem 1.4 is false.","tokens_in":82634,"feed_emoji":"📐","tokens_out":4718,"duration_ms":45984,"temperature":0.7,"pith_summary":"This paper proves that the Steklov eigenvalues of a curvilinear polygon—a planar domain with smooth sides and corners—are approximated, up to an error that tends to zero like a negative power of the index m, by a sequence of quasi-eigenvalues computed only from the side lengths and the interior angles. The result replaces the crude o(m) Weyl law with a sharp asymptotic description that retains detailed geometric information. A surprising corollary is that the arithmetic nature of the angles controls the eigenfunctions: if every angle is π divided by an odd integer, the boundary traces are equidistributed, while if every angle is π divided by an even integer, they concentrate on a single side. The paper also shows that the quasi-eigenvalues are the square roots of eigenvalues of a quantum graph Laplacian on the boundary, and gives explicit trigonometric polynomials whose roots are these quasi-eigenvalues.","feed_headline":"Polygon corners fix Steklov eigenvalues down to O(m^{-ε})","feed_subtitle":"Angles' arithmetic decides whether eigenfunctions spread or concentrate, with a quantum-graph formula for the eigenvalues.","key_machinery":"The central object is the transfer matrix product T(α,ℓ,σ) formed by alternating the vertex matrix A(α) = [[csc(π²/2α), -i cot(π²/2α)],[i cot(π²/2α), csc(π²/2α)]] and the side matrix B(ℓ,σ) = diag($e^{{iℓσ}}$, $e^{{-iℓσ}}$) around the boundary. Quasi-eigenvalues are the σ for which T has eigenvalue 1 (non-exceptional case), or for which the exceptional-component condition U X·X⊥ = 0 holds. The proof that these quasi-eigenvalues actually enumerate the Steklov spectrum uses a lift of the transfer matrices to the universal cover of the punctured plane, where their argument is strictly monotone in σ, combined with Dirichlet-Neumann bracketing on auxiliary zigzag domains. The quasimodes themselves are assembled from scattering Peters solutions in sectors, whose existence and decay r = π/α come from the authors' earlier sloshing analysis.","core_discovery":"For any curvilinear polygon P with side lengths ℓ and angles α, the Steklov eigenvalues λ_m satisfy λ_m = σ_m + O($m^{{-ε}}$) for some ε>0 depending only on the angles, where σ_m is an explicitly defined sequence of quasi-eigenvalues. The quasi-eigenvalues are defined through products of 2×2 vertex transfer matrices A(α) and side transfer matrices B(ℓ,σ), equivalently as roots of trigonometric polynomials, and also as the square roots of eigenvalues of a quantum graph Laplacian on the boundary with angle-dependent matching conditions. The same construction gives asymptotic control of eigenfunctions: on each side, the boundary trace of u_m is, up to O($m^{{-ε}}$) in $L^{2}$, a trigonometric function of frequency σ_m. The arithmetic dichotomy is sharp: if all angles are special (π/(2k+1)), traces equidistribute; if all angles are exceptional (π/(2k)), each trace concentrates on one side, with a splitting of exceptional components governing which side.","pith_inferences":["The quantum-graph formulation suggests an inverse spectral program: since the quasi-eigenvalues are determined by side lengths and angles, a sufficiently long Steklov spectrum should in principle recover the polygon's geometry, as the authors indicate they plan to pursue.","The exceptional-angle concentration phenomenon could be tested experimentally in sloshing tanks with wedge-shaped corners: the free-surface eigenfunctions would localize on alternating walls depending on angle parity.","The transfer-matrix enumeration machinery may extend to higher-order Weyl corrections for polygons with rational angle ratios, where the lifts on the universal cover become periodic and a full asymptotic expansion might exist.","The same construction could be adapted to mixed Dirichlet-Neumann Steklov problems on keyhole domains and other non-simply connected configurations, following the paper's remark on conformal maps."],"forward_implications":["Two curvilinear polygons with the same angles and the same side lengths have Steklov spectra that differ only by O(m^{-ε}), so spectra encode the geometry at this resolution.","The Weyl law N(λ) = |∂P|/π λ + O(1) and the Riesz mean R(λ) = |∂P|/(2π) λ² + O(λ^{1-ε}) hold for all curvilinear polygons with angles less than π.","If all angles are special, the quasi-eigenvalues form periodic arithmetic progressions with double multiplicity, matching the smooth-domain pattern; even special angles can be removed without changing quasi-eigenvalues.","If all angles are exceptional, eigenfunctions concentrate on one side, whereas all-special angles force equidistribution on the boundary.","Quasi-eigenvalues can be computed explicitly as roots of trigonometric polynomials, giving a practical numerical recipe for the Steklov spectrum of cornered domains."],"supporting_citations":[{"why":"Supplies the sloshing-problem spectral asymptotics and the sector Peters solution remainder estimates that the quasimode construction relies on.","marker":"[LPPS17]"},{"why":"Original construction of the sloping beach solutions used to form the scattering Peters solutions in sectors.","marker":"[Pet50]"},{"why":"Provides the baseline Weyl law λ_m = πm/|∂P| + o(m) for piecewise C^1 boundaries that the paper improves to power-law error.","marker":"[Agr06]"},{"why":"Provides the quantum graph spectral theory, including the Weyl law and matching conditions, used to enumerate quasi-eigenvalues and prove their trigonometric-polynomial characterization.","marker":"[BeKu13]"},{"why":"Layer potential analysis adapted by the authors to control how curvature of sides perturbs the Steklov spectrum, needed for fully curvilinear polygons.","marker":"[Cos83]"},{"why":"Supplies comparison examples, such as the square spectrum obtained by separation of variables, used to check the exceptional-angle quasi-eigenvalue formulas.","marker":"[GiPo17]"}],"fun_headline_variants":["Angle arithmetic splits Steklov eigenfunctions: spread or concentrate","Quantum graph formula for Steklov spectra of curved polygons","Precise asymptotics for Steklov eigenvalues on sloshing polygons","Curved polygon corners: Steklov eigenvalues nailed to O(m^{-ε})","Steklov spectra of curved polygons: from sloshing to quantum graphs"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The construction assumes the sector scattering solutions from the earlier sloshing analysis exist with decay rate r = π/α; if those solutions decay more slowly or do not exist, the polygon quasimodes would not be nearly harmonic and the O($m^{{-ε}}$) error would fail.","fun_headline_variants_meta":{"raw":{"variants":["Angle arithmetic splits Steklov eigenfunctions: spread or concentrate","Quantum graph formula for Steklov spectra of curved polygons","Precise asymptotics for Steklov eigenvalues on sloshing polygons","Curved polygon corners: Steklov eigenvalues nailed to O(m^{-ε})","Steklov spectra of curved polygons: from sloshing to quantum graphs"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001682,"raw_usage":{"total_tokens":6640,"prompt_tokens":891,"completion_tokens":5749,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":507,"completion_tokens_details":{"reasoning_tokens":5654}},"tokens_in":507,"tokens_out":5749,"duration_ms":37165,"temperature":1.0,"reasoning_tokens":5654,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:44:48.996846+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute numerically the first several hundred Steklov eigenvalues of a curvilinear triangle with angles (π/3, π/4, 5π/12) and one curved side, then compare them with the roots of the trigonometric polynomial F_P(α,ℓ,σ); if the gap between a true eigenvalue and the nearest quasi-eigenvalue does not shrink like a negative power of m, Theorem 1.4 is false.","supporting_citations":[],"review_version":1}