{"id":"c0e6d54f-c54b-4bb1-9216-19d42da214a9","arxiv_id":"2511.09490","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"New sharp lower and upper bounds for exterior Steklov eigenvalues are proved, showing first eigenvalues can blow up on fixed-volume convex domains in n≥3 while a Weinstock-type isoperimetric bound holds in 2D.","lead":"This paper studies the exterior Steklov eigenvalue problem, in which harmonic functions live outside a bounded obstacle and satisfy a boundary condition proportional to their boundary values. It unifies several competing formulations and proves new sharp geometric bounds, including a surprising divergence of the first eigenvalue for volume-normalized convex domains in three or more dimensions.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Equation (1.3) as printed makes L the reciprocal of the standard logarithmic mean; taken literally, Theorem 1.11 fails even for a ball.","rationale":"Read in good faith, the intended result is an Escobar-type lower bound using the usual logarithmic mean, and the normal-coordinate reduction is plausible for convex C^{1,1} boundaries: the nearest-point projection gives a bi-Lipschitz map and the Jacobian formula holds a.e. However, the paper's own definition (1.3) contradicts both its k=2 statement and its sharp ball case. The proof's final algebra only works if L is the reciprocal of the displayed expression; as printed, the theorem is false, and Theorem 1.18's divergence argument in Example 6.7 contains the same self-inconsistency (the formula as written would give a vanishing lower bound as a→0, while the following asymptotic uses the reciprocal). This is a load-bearing defect in the central claim, more fundamental than the coordinate-map concern identified by the reader. The intended mathematics appears recoverable by inserting the missing reciprocal, so the appropriate disposition is conditional acceptance pending correction of Eq. (1.3) and the analogous displayed formula in Example 6.7.","tokens_in":58044,"tokens_out":33691,"duration_ms":329648,"concrete_test":"Take Ω=B_ρ⊂R^4. Compute the RHS of (1.4) using (1.3) verbatim: for equal curvatures κ=1/ρ, I=∫_0^∞ (1+t/ρ)^{-3} dt = ρ/2, so L_p = I/(n-2) = ρ/4 and β=(n-2)L_p = ρ/2. Compare with σ1(B_ρ^ext)=(n-2)/ρ=2/ρ. For ρ=10 this gives 0.2≥5, a contradiction. Replacing (1.3) by the reciprocal expression L=1/[(n-2)Σ ... log ...] gives β=2/ρ and equality, confirming the intended definition.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The printed definition (1.3), L = 1/(k-1) Σ_j α_j^{k-2}/∏_{i≠j}(α_j-α_i) log α_j, does not reduce to the two-variable logarithmic mean stated immediately below: for k=2 it gives (log α1 - log α2)/(α1 - α2), the reciprocal of (α1-α2)/(log α1-log α2). This is not cosmetic. In §6.2 the one-dimensional Hardy constant is K(s)=1/I(s), where I(s)=∫_0^∞ 1/∏(1+κ_j(s)t) dt = Σ_j κ_j^{n-3}/∏_{i≠j}(κ_j-κ_i) log κ_j = (n-2)L_p(s) with L_p the printed quantity. Hence K(s)=1/[(n-2)L_p(s)], whereas the proof concludes K(s)=(n-2)L(s). Equality for a ball forces the intended L to be the reciprocal of the printed one. Concrete failure of the printed statement: for Ω=B_10⊂R^4, σ1(Ω_ext)=(4-2)/10=0.2; curvatures are all 1/10, I=5, printed L_p=5/2, so the printed RHS of (1.4) is (n-2)L_p=5 and the inequality 0.2≥5 is false. Example 6.3 uses the intended standard L, while the displayed formula in Example 6.7 repeats the missing reciprocal before its asymptotic line reverts to the correct value. The coordinate-map premise in Eq. (6.2) is not the main issue; the logarithmic mean itself must be corrected in the statement of the central theorem.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the exterior Steklov eigenvalue problem on bounded Euclidean domains from several equivalent viewpoints: finite-energy spaces, conformal mappings in two dimensions, truncated domains, Helmholtz regularisation, and layer potentials. Its main positive results are an Escobar-type lower bound for the first exterior Steklov eigenvalue on convex C^{1,1} domains in dimensions n≥3, expressed through the principal curvatures of the boundary (Theorem 1.11), a Weinstock-type upper bound in two dimensions (Theorem 1.16), spectral asymptotics, and a construction showing that in higher dimensions the first exterior Steklov eigenvalue can blow up for convex domains of fixed volume (Theorem 1.18). The proof of Theorem 1.11 reduces the Rayleigh quotient to a one-dimensional Hardy-type constant along normal rays and evaluates it in terms of a logarithmic mean of the principal curvatures.","tokens_in":58381,"tokens_out":12121,"duration_ms":107695,"significance":"If the logarithmic-mean definition is corrected as discussed below, this is a substantial contribution. The unification of the five formulations is careful and fills a genuine gap in the literature; the two-dimensional Weinstock proof is nontrivial because it has to control the image of infinity; and Theorem 1.18 gives a striking contrast with the interior Steklov problem. The variational reduction in §6.2 is elegant and self-contained, and the numerical comparisons with Xiong’s bound are useful. The accompanying scripts and examples are a further strength. However, the central lower bound is currently invalid as stated because of the reciprocal error in the definition of the logarithmic mean, so the manuscript needs a major correction before it can be evaluated as a final contribution.","major_comments":[{"comment":"Equation (1.3) is inconsistent with the displayed two-variable formula and, as printed, makes Theorem 1.11 false. For k=2, (1.3) gives (log α1-log α2)/(α1-α2), the reciprocal of the claimed L. In §6.2 the Hardy constant is K=1/I, with I=∫_0∞ [∏(1+κ_j t)]^{-1}dt=Σ κ_j^{n-3} log κ_j / ∏_{i≠j}(κ_j-κ_i). With L_p as in (1.3), K=1/[(n-2)L_p], not (n-2)L_p. For B_10⊂R^4, σ1=0.2, all κ=0.1, printed L_p=2.5, so the RHS of (1.4) is 5, and 0.2≥5 fails. The intended L must be the reciprocal of the sum, L = [ (n-2) Σ κ_j^{n-3} log κ_j / ∏_{i≠j}(κ_j-κ_i) ]^{-1}; this gives the standard two-variable mean and L(κ,...,κ)=κ. Correct (1.3) and all dependent displays, including Example 6.7 and the proof of Theorem 1.11.","section":"§1.3.2, Eq. (1.3); §6.2"},{"comment":"Corollary 1.12 claims equality iff Ω is a ball, but the proof invokes [Mu87, Theorem 2] to conclude that constant geometric mean of the principal curvatures implies the boundary is a sphere. Since Ω is only assumed C^{1,1} and the curvatures are defined only almost everywhere, please verify that the cited theorem applies at this regularity, or state the additional smoothness needed. This is secondary to Theorem 1.11 but is load-bearing for the equality statement of Corollary 1.12.","section":"§6.2, Corollary 1.12"}],"minor_comments":[{"comment":"The line 'L(α_1,...,α_2)' in the proof should read L(α_1,...,α_{n-1}).","section":"§1.3.2, Corollary 1.12"},{"comment":"Reference [HelKaNi25] contains a typo: 'and and F. Nicoleau'.","section":"References"},{"comment":"The displayed formula for the k=2 spheroid should be re-derived after the correction of (1.3); as written it appears to use the printed reciprocal definition before the asymptotic line reverts to the intended logarithmic mean.","section":"§6.3.2, Example 6.7"}],"recommendation":"major_revision","confidential_remarks":"The reciprocal error in the definition of the logarithmic mean is the main obstacle; it is local and fixable, so I recommend major revision rather than rejection. The authors should also ensure that the corrected definition is propagated through Examples 6.3 and 6.7 and the proof of Theorem 1.11. The rest of the paper is sound and well organized."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"One thing to know before anything else: do not quote Theorem 1.11 as printed. Definition (1.3) gives the reciprocal of the logarithmic mean the authors actually use. For k=2 it returns (log α1 - log α2)/(α1 - α2), while the text immediately below says the standard (α1 - α2)/(log α1 - log α2). The proof in §6.2 is internally consistent with the intended standard mean: it arrives at K(s,Ω) = 1/[(n-2)L_printed] = (n-2)L_standard, so the theorem is true after replacing (1.3) by its reciprocal. As printed, Theorem 1.11 fails for a ball: for B_10 ⊂ R^4, σ1 = 0.2 but the printed bound gives 5. This is a load-bearing typo in the statement, not a flaw in the variational argument; the same reciprocal error appears in the displayed formula in Example 6.7 before the asymptotics. The authors need to fix (1.3) and the surrounding text before this can be cited. Now the merits. The paper does a genuinely useful job of unifying the scattered definitions of the exterior Steklov problem — finite-energy spaces, truncations, Helmholtz regularisation, conformal mapping, and layer potentials — and proving they give the same spectrum. That matters for a topic where different communities keep writing down different problems. The two-dimensional Weinstock inequality is new in this setting, and the equality case requires care with the image of infinity; the proof handles it. The main phenomenon, fixed-volume convex domains in n≥3 whose first exterior Steklov eigenvalue diverges, is real and follows from the corrected lower bound applied to prolate spheroids. The numerical comparisons with Xiong's bound are useful. The proofs I checked are detailed and structurally sound; the reduction to a one-dimensional Hardy-type constant is clean. Soft spots, in proportion. The reader's worry about the normal-coordinate premise at (6.2) is not a real flaw: Theorem 1.11 explicitly assumes convex C^{1,1}, which is exactly the setting where the coordinate map is bi-Lipschitz. The equality characterisation in Theorem 1.11 is left open, and the authors are candid about it; only the geometric-mean version in Corollary 1.12 has a full equality proof. Proposition 6.9 is delegated by \"verbatim adaptation\" to Brisson's work, so it is really a cited result rather than a proved one; acceptable, but it should be labelled as such. The Robin corollary depends on Bundrock's duality, which is fine. Bottom line: the mathematical core is good and should be in the literature, but the central theorem needs a correction before the paper is in final form. Send it to referees, and ask the authors to fix the logarithmic mean definition and the k=2 claim. I would bring the corrected version to reading group, and I would cite the corrected version.","headline":"The paper is worth engaging with, but the central lower bound as printed uses the reciprocal of the intended logarithmic mean and fails for a ball; fix (1.3) before citing.","tokens_in":830,"tokens_out":1217,"would_cite":true,"duration_ms":65440,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35P05","35P15","35P20","31A10","31B10","47A75"],"pacs":[],"model":"deepseek-v4-flash","headline":"In dimensions three and higher, the first exterior Steklov eigenvalue of a convex domain is bounded below by the logarithmic mean of the boundary's principal curvatures — which forces eigenvalues to diverge for fixed-volume thin domains.","keywords":["Steklov eigenvalue","exterior domain","Dirichlet-to-Neumann operator","convex domain","principal curvatures","logarithmic mean","Weinstock inequality","Weyl asymptotics"],"falsifier":"Compute (numerically or analytically) the first exterior Steklov eigenvalue of a smooth convex domain — for example, a rounded cube or a very thin prolate spheroid — in R³ and compare it with (n−2) inf L(κ₁,κ₂). Any value below the bound, or a fixed-volume family whose σ₁ stays bounded while the bound diverges, would refute Theorem 1.11/1.18. For the 2D Weinstock part, a simply connected Lipschitz domain with σ₂|∂Ω| > 2π would refute Theorem 1.16; for the ball-rigidity in Corollary 1.12, a convex body with constant geometric mean curvature that is not a sphere would be a counterexample.","tokens_in":57892,"feed_emoji":"📐","tokens_out":7219,"duration_ms":65624,"temperature":0.7,"pith_summary":"The paper studies the Steklov eigenvalue problem in the unbounded exterior of a bounded Euclidean domain: harmonic functions on the outside whose normal derivative on the boundary is σ times their trace. Because harmonic extensions to unbounded domains are not unique, a definition needs extra conditions at infinity; the authors supply several formulations — finite-energy spaces, conformal inversion in two dimensions, truncated domains, a regularizing Helmholtz equation, and boundary layer potentials — and prove that they are all equivalent. The central geometric result is an Escobar-type lower bound in dimensions n≥3: for a bounded convex C^{1,1} domain, the first exterior Steklov eigenvalue is at least (n−2) times the infimum over the boundary of the logarithmic mean of the principal curvatures. The bound is sharp for a ball and implies that smooth convex domains of fixed volume can have first exterior Steklov eigenvalues tending to infinity, a phenomenon that cannot occur in the interior problem or in the two-dimensional exterior problem, where the paper proves a Weinstock-type isoperimetric inequality instead. The paper thereby establishes the exterior Steklov spectrum as a genuinely different regime from its interior counterpart.","feed_headline":"Thin convex solids push exterior Steklov eigenvalues to infinity","feed_subtitle":"New sharp bound shows fixed-volume convex exteriors can have arbitrarily large first eigenvalue — opposite of the 2D and interior cases.","key_machinery":"The central mechanism is the normal-coordinate parametrization of the exterior domain, Ψ(s,t)=s−tν(s), whose Jacobian determinant ζ(s,t)=∏_{j=1}^{n−1}(1+κ_j(s)t) measures how boundary length elements inflate along inward normal rays. This factor turns the exterior Rayleigh quotient into a weighted one-dimensional minimization along each ray, and the paper computes the optimal constant for the associated Hardy-type problem (f′ζ)′=0 with decay at infinity; the value (n−2)L(κ₁(s),…,κ_{n−1}(s)) — an identity involving the logarithmic mean and a partial-fraction decomposition of 1/∏(1+κ_j t) — is what ultimately bounds the eigenvalue. In two dimensions the load-bearing tool is different: the inve","core_discovery":"The load-bearing discovery is Theorem 1.11. For a bounded convex domain Ω ⊂ R^n, n≥3, with ∂Ω ∈ C^{1,1}, the first exterior Steklov eigenvalue satisfies σ₁(Ω^ext) ≥ (n−2) inf_{s∈∂Ω} L(κ₁(s),…,κ_{n−1}(s)), where L is the logarithmic mean of the principal curvatures; equality holds for a ball, and for the geometric-mean version equality characterizes balls. The proof reduces the exterior Rayleigh quotient to one-dimensional Hardy-type minimizations along normal rays, where the Jacobian factor ∏(1+κ_j(s)t) converts the problem into a family of one-dimensional constants, each evaluated explicitly as (n−2)L(κ₁(s),…,κ_{n−1}(s)). A direct corollary, Theorem 1.18, asserts that for every n≥3 there is","pith_inferences":["The quantity β(∂Ω)=(n−2) inf_s L(κ₁,…,κ_{n−1}) behaves like a curvature-concentration functional: it vanishes for any boundary that contains a flat patch and diverges when all curvatures grow simultaneously. This makes it a natural design parameter in applications such as diffusion-mediated surface reactions, where a larger β would suppress low-frequency boundary modes.","The normal-ray/Hardy-constant method is likely independent of the specific elliptic operator: the same Jacobian ζ(s,t) controls exterior Robin, p-Laplacian, or magnetic Schrödinger problems, and one could test whether the logarithmic mean appears there too.","The paper leaves open whether equality in Theorem 1.11 (logarithmic-mean version) forces a sphere; a testable conjecture is that any sufficiently smooth convex domain whose logarithmic-mean curvature is constant must be a ball, which could be checked by first-order perturbation theory around the sphere.","In view of Remark 1.17, the conformal dictionary suggests a general principle in dimension two: every isoperimetric inequality for the weighted interior Steklov problem has an exterior analogue; Theorem 1.16 is the first instance, and one could extend it to higher-order eigenvalues via Hersch–Payne–Schiffer inequalities."],"forward_implications":["There exist smooth convex bodies of fixed volume, in every dimension n≥3, whose first exterior Steklov eigenvalue tends to infinity; the same holds with surface area fixed.","In two dimensions the disk maximizes the first nonzero exterior Steklov eigenvalue among simply connected Lipschitz domains of given perimeter (σ₂|∂Ω| ≤ 2π) and of given area (σ₂|Ω|^{1/2} ≤ √π).","The exterior Robin criterion (1.8) turns the curvature bound into a statement about phase transitions: convex domains of prescribed volume can have −σ₁ arbitrarily large, so the negative-Robin-eigenvalue regime can be reached for arbitrarily strong couplings (Corollary 1.19).","The counting function obeys a Weyl law of the same order as the interior problem, N(σ)=ω_{n−1}|∂Ω|/(2π)^{n−1} σ^{n−1}+O(σ^{n−2}), so the blow-up of the first eigenvalue coexists with standard universal asymptotics at high frequency.","The first exterior Steklov eigenvalue is simple, and each k-th eigenfunction has at most k nodal domains, so the low-frequency spectral structure mirrors the interior problem even where the shapes are not comparable."],"fun_headline_variants":["Exterior Steklov: convex domains can have unbounded first eigenvalue","Convex exteriors: first Steklov eigenvalue can blow up in n≥3","Sharp bound: convex exterior eigenvalue diverges in dimensions ≥3","Logarithmic mean of curvatures bounds first exterior Steklov eigenvalue","Exterior Steklov eigenvalue → ∞ for convex solids with fixed volume"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The proof of the lower bound relies on the fact that for a convex C^{1,1} boundary, the normal-ray map (s,t) ↦ s−tν(s) is a bi-Lipschitz bijection onto the exterior with Jacobian ∏(1+κ_j(s)t); if the boundary loses convexity or C^{1,1} regularity, this coordinate representation fails and the derivation of Theorem 1.11 collapses.","fun_headline_variants_meta":{"raw":{"variants":["Exterior Steklov: convex domains can have unbounded first eigenvalue","Convex exteriors: first Steklov eigenvalue can blow up in n≥3","Sharp bound: convex exterior eigenvalue diverges in dimensions ≥3","Logarithmic mean of curvatures bounds first exterior Steklov eigenvalue","Exterior Steklov eigenvalue → ∞ for convex solids with fixed volume"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000587,"raw_usage":{"total_tokens":2580,"prompt_tokens":719,"completion_tokens":1861,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":463,"completion_tokens_details":{"reasoning_tokens":1778}},"tokens_in":463,"tokens_out":1861,"duration_ms":11762,"temperature":1.0,"reasoning_tokens":1778,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T22:35:44.050982+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute (numerically or analytically) the first exterior Steklov eigenvalue of a smooth convex domain — for example, a rounded cube or a very thin prolate spheroid — in R³ and compare it with (n−2) inf L(κ₁,κ₂). Any value below the bound, or a fixed-volume family whose σ₁ stays bounded while the bound diverges, would refute Theorem 1.11/1.18. For the 2D Weinstock part, a simply connected Lipschitz domain with σ₂|∂Ω| > 2π would refute Theorem 1.16; for the ball-rigidity in Corollary 1.12, a convex body with constant geometric mean curvature that is not a sphere would be a counterexample.","supporting_citations":[],"review_version":1}